diff options
Diffstat (limited to 'lib/builtins')
347 files changed, 8636 insertions, 9774 deletions
diff --git a/lib/builtins/aarch64/chkstk.S b/lib/builtins/aarch64/chkstk.S index 89ec90b08a13..01f90366f030 100644 --- a/lib/builtins/aarch64/chkstk.S +++ b/lib/builtins/aarch64/chkstk.S @@ -1,5 +1,6 @@ -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception #include "../assembly.h" diff --git a/lib/builtins/absvdi2.c b/lib/builtins/absvdi2.c index 682c2355d2ac..b9566cd874fe 100644 --- a/lib/builtins/absvdi2.c +++ b/lib/builtins/absvdi2.c @@ -1,29 +1,25 @@ -/*===-- absvdi2.c - Implement __absvdi2 -----------------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - *===----------------------------------------------------------------------=== - * - * This file implements __absvdi2 for the compiler_rt library. - * - *===----------------------------------------------------------------------=== - */ +//===-- absvdi2.c - Implement __absvdi2 -----------------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements __absvdi2 for the compiler_rt library. +// +//===----------------------------------------------------------------------===// #include "int_lib.h" -/* Returns: absolute value */ +// Returns: absolute value -/* Effects: aborts if abs(x) < 0 */ +// Effects: aborts if abs(x) < 0 -COMPILER_RT_ABI di_int -__absvdi2(di_int a) -{ - const int N = (int)(sizeof(di_int) * CHAR_BIT); - if (a == ((di_int)1 << (N-1))) - compilerrt_abort(); - const di_int t = a >> (N - 1); - return (a ^ t) - t; +COMPILER_RT_ABI di_int __absvdi2(di_int a) { + const int N = (int)(sizeof(di_int) * CHAR_BIT); + if (a == ((di_int)1 << (N - 1))) + compilerrt_abort(); + const di_int t = a >> (N - 1); + return (a ^ t) - t; } diff --git a/lib/builtins/absvsi2.c b/lib/builtins/absvsi2.c index 4812af815982..44ada169e7e6 100644 --- a/lib/builtins/absvsi2.c +++ b/lib/builtins/absvsi2.c @@ -1,29 +1,25 @@ -/* ===-- absvsi2.c - Implement __absvsi2 -----------------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - * - * This file implements __absvsi2 for the compiler_rt library. - * - * ===----------------------------------------------------------------------=== - */ +//===-- absvsi2.c - Implement __absvsi2 -----------------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements __absvsi2 for the compiler_rt library. +// +//===----------------------------------------------------------------------===// #include "int_lib.h" -/* Returns: absolute value */ +// Returns: absolute value -/* Effects: aborts if abs(x) < 0 */ +// Effects: aborts if abs(x) < 0 -COMPILER_RT_ABI si_int -__absvsi2(si_int a) -{ - const int N = (int)(sizeof(si_int) * CHAR_BIT); - if (a == (1 << (N-1))) - compilerrt_abort(); - const si_int t = a >> (N - 1); - return (a ^ t) - t; +COMPILER_RT_ABI si_int __absvsi2(si_int a) { + const int N = (int)(sizeof(si_int) * CHAR_BIT); + if (a == (1 << (N - 1))) + compilerrt_abort(); + const si_int t = a >> (N - 1); + return (a ^ t) - t; } diff --git a/lib/builtins/absvti2.c b/lib/builtins/absvti2.c index 7927770c9ab3..491d99d7ce0f 100644 --- a/lib/builtins/absvti2.c +++ b/lib/builtins/absvti2.c @@ -1,34 +1,29 @@ -/* ===-- absvti2.c - Implement __absvdi2 -----------------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - * - * This file implements __absvti2 for the compiler_rt library. - * - * ===----------------------------------------------------------------------=== - */ +//===-- absvti2.c - Implement __absvdi2 -----------------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements __absvti2 for the compiler_rt library. +// +//===----------------------------------------------------------------------===// #include "int_lib.h" #ifdef CRT_HAS_128BIT -/* Returns: absolute value */ +// Returns: absolute value -/* Effects: aborts if abs(x) < 0 */ +// Effects: aborts if abs(x) < 0 -COMPILER_RT_ABI ti_int -__absvti2(ti_int a) -{ - const int N = (int)(sizeof(ti_int) * CHAR_BIT); - if (a == ((ti_int)1 << (N-1))) - compilerrt_abort(); - const ti_int s = a >> (N - 1); - return (a ^ s) - s; +COMPILER_RT_ABI ti_int __absvti2(ti_int a) { + const int N = (int)(sizeof(ti_int) * CHAR_BIT); + if (a == ((ti_int)1 << (N - 1))) + compilerrt_abort(); + const ti_int s = a >> (N - 1); + return (a ^ s) - s; } -#endif /* CRT_HAS_128BIT */ - +#endif // CRT_HAS_128BIT diff --git a/lib/builtins/adddf3.c b/lib/builtins/adddf3.c index 9a3901312e51..f2727fafcabe 100644 --- a/lib/builtins/adddf3.c +++ b/lib/builtins/adddf3.c @@ -1,9 +1,8 @@ //===-- lib/adddf3.c - Double-precision addition ------------------*- C -*-===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// // @@ -15,16 +14,12 @@ #define DOUBLE_PRECISION #include "fp_add_impl.inc" -COMPILER_RT_ABI double __adddf3(double a, double b){ - return __addXf3__(a, b); -} +COMPILER_RT_ABI double __adddf3(double a, double b) { return __addXf3__(a, b); } #if defined(__ARM_EABI__) #if defined(COMPILER_RT_ARMHF_TARGET) -AEABI_RTABI double __aeabi_dadd(double a, double b) { - return __adddf3(a, b); -} +AEABI_RTABI double __aeabi_dadd(double a, double b) { return __adddf3(a, b); } #else -AEABI_RTABI double __aeabi_dadd(double a, double b) COMPILER_RT_ALIAS(__adddf3); +COMPILER_RT_ALIAS(__adddf3, __aeabi_dadd) #endif #endif diff --git a/lib/builtins/addsf3.c b/lib/builtins/addsf3.c index c5c1a41c3611..8fe8622aadd9 100644 --- a/lib/builtins/addsf3.c +++ b/lib/builtins/addsf3.c @@ -1,9 +1,8 @@ //===-- lib/addsf3.c - Single-precision addition ------------------*- C -*-===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// // @@ -15,16 +14,12 @@ #define SINGLE_PRECISION #include "fp_add_impl.inc" -COMPILER_RT_ABI float __addsf3(float a, float b) { - return __addXf3__(a, b); -} +COMPILER_RT_ABI float __addsf3(float a, float b) { return __addXf3__(a, b); } #if defined(__ARM_EABI__) #if defined(COMPILER_RT_ARMHF_TARGET) -AEABI_RTABI float __aeabi_fadd(float a, float b) { - return __addsf3(a, b); -} +AEABI_RTABI float __aeabi_fadd(float a, float b) { return __addsf3(a, b); } #else -AEABI_RTABI float __aeabi_fadd(float a, float b) COMPILER_RT_ALIAS(__addsf3); +COMPILER_RT_ALIAS(__addsf3, __aeabi_fadd) #endif #endif diff --git a/lib/builtins/addtf3.c b/lib/builtins/addtf3.c index e4bbe0227ae6..570472a14554 100644 --- a/lib/builtins/addtf3.c +++ b/lib/builtins/addtf3.c @@ -1,9 +1,8 @@ //===-- lib/addtf3.c - Quad-precision addition --------------------*- C -*-===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// // @@ -18,8 +17,8 @@ #if defined(CRT_HAS_128BIT) && defined(CRT_LDBL_128BIT) #include "fp_add_impl.inc" -COMPILER_RT_ABI long double __addtf3(long double a, long double b){ - return __addXf3__(a, b); +COMPILER_RT_ABI long double __addtf3(long double a, long double b) { + return __addXf3__(a, b); } #endif diff --git a/lib/builtins/addvdi3.c b/lib/builtins/addvdi3.c index 0da38945679d..28661fda8498 100644 --- a/lib/builtins/addvdi3.c +++ b/lib/builtins/addvdi3.c @@ -1,36 +1,29 @@ -/* ===-- addvdi3.c - Implement __addvdi3 -----------------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - * - * This file implements __addvdi3 for the compiler_rt library. - * - * ===----------------------------------------------------------------------=== - */ +//===-- addvdi3.c - Implement __addvdi3 -----------------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements __addvdi3 for the compiler_rt library. +// +//===----------------------------------------------------------------------===// #include "int_lib.h" -/* Returns: a + b */ +// Returns: a + b -/* Effects: aborts if a + b overflows */ +// Effects: aborts if a + b overflows -COMPILER_RT_ABI di_int -__addvdi3(di_int a, di_int b) -{ - di_int s = (du_int) a + (du_int) b; - if (b >= 0) - { - if (s < a) - compilerrt_abort(); - } - else - { - if (s >= a) - compilerrt_abort(); - } - return s; +COMPILER_RT_ABI di_int __addvdi3(di_int a, di_int b) { + di_int s = (du_int)a + (du_int)b; + if (b >= 0) { + if (s < a) + compilerrt_abort(); + } else { + if (s >= a) + compilerrt_abort(); + } + return s; } diff --git a/lib/builtins/addvsi3.c b/lib/builtins/addvsi3.c index 94ca726f42b9..40400237527b 100644 --- a/lib/builtins/addvsi3.c +++ b/lib/builtins/addvsi3.c @@ -1,36 +1,29 @@ -/* ===-- addvsi3.c - Implement __addvsi3 -----------------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - * - * This file implements __addvsi3 for the compiler_rt library. - * - * ===----------------------------------------------------------------------=== - */ +//===-- addvsi3.c - Implement __addvsi3 -----------------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements __addvsi3 for the compiler_rt library. +// +//===----------------------------------------------------------------------===// #include "int_lib.h" -/* Returns: a + b */ +// Returns: a + b -/* Effects: aborts if a + b overflows */ +// Effects: aborts if a + b overflows -COMPILER_RT_ABI si_int -__addvsi3(si_int a, si_int b) -{ - si_int s = (su_int) a + (su_int) b; - if (b >= 0) - { - if (s < a) - compilerrt_abort(); - } - else - { - if (s >= a) - compilerrt_abort(); - } - return s; +COMPILER_RT_ABI si_int __addvsi3(si_int a, si_int b) { + si_int s = (su_int)a + (su_int)b; + if (b >= 0) { + if (s < a) + compilerrt_abort(); + } else { + if (s >= a) + compilerrt_abort(); + } + return s; } diff --git a/lib/builtins/addvti3.c b/lib/builtins/addvti3.c index c224de60aab0..aa709875d243 100644 --- a/lib/builtins/addvti3.c +++ b/lib/builtins/addvti3.c @@ -1,40 +1,33 @@ -/* ===-- addvti3.c - Implement __addvti3 -----------------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - * - * This file implements __addvti3 for the compiler_rt library. - * - * ===----------------------------------------------------------------------=== - */ +//===-- addvti3.c - Implement __addvti3 -----------------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements __addvti3 for the compiler_rt library. +// +//===----------------------------------------------------------------------===// #include "int_lib.h" #ifdef CRT_HAS_128BIT -/* Returns: a + b */ +// Returns: a + b -/* Effects: aborts if a + b overflows */ +// Effects: aborts if a + b overflows -COMPILER_RT_ABI ti_int -__addvti3(ti_int a, ti_int b) -{ - ti_int s = (tu_int) a + (tu_int) b; - if (b >= 0) - { - if (s < a) - compilerrt_abort(); - } - else - { - if (s >= a) - compilerrt_abort(); - } - return s; +COMPILER_RT_ABI ti_int __addvti3(ti_int a, ti_int b) { + ti_int s = (tu_int)a + (tu_int)b; + if (b >= 0) { + if (s < a) + compilerrt_abort(); + } else { + if (s >= a) + compilerrt_abort(); + } + return s; } -#endif /* CRT_HAS_128BIT */ +#endif // CRT_HAS_128BIT diff --git a/lib/builtins/apple_versioning.c b/lib/builtins/apple_versioning.c index 3797a1ab02da..f87b42820c15 100644 --- a/lib/builtins/apple_versioning.c +++ b/lib/builtins/apple_versioning.c @@ -1,47 +1,42 @@ -/* ===-- apple_versioning.c - Adds versioning symbols for ld ---------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - */ - +//===-- apple_versioning.c - Adds versioning symbols for ld ---------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// #if __APPLE__ - #include <Availability.h> - - #if __IPHONE_OS_VERSION_MIN_REQUIRED - #define NOT_HERE_BEFORE_10_6(sym) - #define NOT_HERE_IN_10_8_AND_EARLIER(sym) \ - extern const char sym##_tmp61 __asm("$ld$hide$os6.1$_" #sym ); \ - __attribute__((visibility("default"))) const char sym##_tmp61 = 0; \ - extern const char sym##_tmp60 __asm("$ld$hide$os6.0$_" #sym ); \ - __attribute__((visibility("default"))) const char sym##_tmp60 = 0; \ - extern const char sym##_tmp51 __asm("$ld$hide$os5.1$_" #sym ); \ - __attribute__((visibility("default"))) const char sym##_tmp51 = 0; \ - extern const char sym##_tmp50 __asm("$ld$hide$os5.0$_" #sym ); \ - __attribute__((visibility("default"))) const char sym##_tmp50 = 0; - #else - #define NOT_HERE_BEFORE_10_6(sym) \ - extern const char sym##_tmp4 __asm("$ld$hide$os10.4$_" #sym ); \ - __attribute__((visibility("default"))) const char sym##_tmp4 = 0; \ - extern const char sym##_tmp5 __asm("$ld$hide$os10.5$_" #sym ); \ - __attribute__((visibility("default"))) const char sym##_tmp5 = 0; - #define NOT_HERE_IN_10_8_AND_EARLIER(sym) \ - extern const char sym##_tmp8 __asm("$ld$hide$os10.8$_" #sym ); \ - __attribute__((visibility("default"))) const char sym##_tmp8 = 0; \ - extern const char sym##_tmp7 __asm("$ld$hide$os10.7$_" #sym ); \ - __attribute__((visibility("default"))) const char sym##_tmp7 = 0; \ - extern const char sym##_tmp6 __asm("$ld$hide$os10.6$_" #sym ); \ - __attribute__((visibility("default"))) const char sym##_tmp6 = 0; - #endif +#include <Availability.h> +#if __IPHONE_OS_VERSION_MIN_REQUIRED +#define NOT_HERE_BEFORE_10_6(sym) +#define NOT_HERE_IN_10_8_AND_EARLIER(sym) \ + extern const char sym##_tmp61 __asm("$ld$hide$os6.1$_" #sym); \ + __attribute__((visibility("default"))) const char sym##_tmp61 = 0; \ + extern const char sym##_tmp60 __asm("$ld$hide$os6.0$_" #sym); \ + __attribute__((visibility("default"))) const char sym##_tmp60 = 0; \ + extern const char sym##_tmp51 __asm("$ld$hide$os5.1$_" #sym); \ + __attribute__((visibility("default"))) const char sym##_tmp51 = 0; \ + extern const char sym##_tmp50 __asm("$ld$hide$os5.0$_" #sym); \ + __attribute__((visibility("default"))) const char sym##_tmp50 = 0; +#else +#define NOT_HERE_BEFORE_10_6(sym) \ + extern const char sym##_tmp4 __asm("$ld$hide$os10.4$_" #sym); \ + __attribute__((visibility("default"))) const char sym##_tmp4 = 0; \ + extern const char sym##_tmp5 __asm("$ld$hide$os10.5$_" #sym); \ + __attribute__((visibility("default"))) const char sym##_tmp5 = 0; +#define NOT_HERE_IN_10_8_AND_EARLIER(sym) \ + extern const char sym##_tmp8 __asm("$ld$hide$os10.8$_" #sym); \ + __attribute__((visibility("default"))) const char sym##_tmp8 = 0; \ + extern const char sym##_tmp7 __asm("$ld$hide$os10.7$_" #sym); \ + __attribute__((visibility("default"))) const char sym##_tmp7 = 0; \ + extern const char sym##_tmp6 __asm("$ld$hide$os10.6$_" #sym); \ + __attribute__((visibility("default"))) const char sym##_tmp6 = 0; +#endif -/* Symbols in libSystem.dylib in 10.6 and later, - * but are in libgcc_s.dylib in earlier versions - */ +// Symbols in libSystem.dylib in 10.6 and later, +// but are in libgcc_s.dylib in earlier versions NOT_HERE_BEFORE_10_6(__absvdi2) NOT_HERE_BEFORE_10_6(__absvsi2) @@ -143,14 +138,13 @@ NOT_HERE_BEFORE_10_6(__udivti3) NOT_HERE_BEFORE_10_6(__umoddi3) NOT_HERE_BEFORE_10_6(__umodti3) - #if __ppc__ NOT_HERE_BEFORE_10_6(__gcc_qadd) NOT_HERE_BEFORE_10_6(__gcc_qdiv) NOT_HERE_BEFORE_10_6(__gcc_qmul) NOT_HERE_BEFORE_10_6(__gcc_qsub) NOT_HERE_BEFORE_10_6(__trampoline_setup) -#endif /* __ppc__ */ +#endif // __ppc__ NOT_HERE_IN_10_8_AND_EARLIER(__atomic_compare_exchange) NOT_HERE_IN_10_8_AND_EARLIER(__atomic_compare_exchange_1) @@ -201,24 +195,23 @@ NOT_HERE_IN_10_8_AND_EARLIER(__atomic_store_2) NOT_HERE_IN_10_8_AND_EARLIER(__atomic_store_4) NOT_HERE_IN_10_8_AND_EARLIER(__atomic_store_8) - #if __arm__ && __DYNAMIC__ - #define NOT_HERE_UNTIL_AFTER_4_3(sym) \ - extern const char sym##_tmp1 __asm("$ld$hide$os3.0$_" #sym ); \ - __attribute__((visibility("default"))) const char sym##_tmp1 = 0; \ - extern const char sym##_tmp2 __asm("$ld$hide$os3.1$_" #sym ); \ - __attribute__((visibility("default"))) const char sym##_tmp2 = 0; \ - extern const char sym##_tmp3 __asm("$ld$hide$os3.2$_" #sym ); \ - __attribute__((visibility("default"))) const char sym##_tmp3 = 0; \ - extern const char sym##_tmp4 __asm("$ld$hide$os4.0$_" #sym ); \ - __attribute__((visibility("default"))) const char sym##_tmp4 = 0; \ - extern const char sym##_tmp5 __asm("$ld$hide$os4.1$_" #sym ); \ - __attribute__((visibility("default"))) const char sym##_tmp5 = 0; \ - extern const char sym##_tmp6 __asm("$ld$hide$os4.2$_" #sym ); \ - __attribute__((visibility("default"))) const char sym##_tmp6 = 0; \ - extern const char sym##_tmp7 __asm("$ld$hide$os4.3$_" #sym ); \ - __attribute__((visibility("default"))) const char sym##_tmp7 = 0; - +#define NOT_HERE_UNTIL_AFTER_4_3(sym) \ + extern const char sym##_tmp1 __asm("$ld$hide$os3.0$_" #sym); \ + __attribute__((visibility("default"))) const char sym##_tmp1 = 0; \ + extern const char sym##_tmp2 __asm("$ld$hide$os3.1$_" #sym); \ + __attribute__((visibility("default"))) const char sym##_tmp2 = 0; \ + extern const char sym##_tmp3 __asm("$ld$hide$os3.2$_" #sym); \ + __attribute__((visibility("default"))) const char sym##_tmp3 = 0; \ + extern const char sym##_tmp4 __asm("$ld$hide$os4.0$_" #sym); \ + __attribute__((visibility("default"))) const char sym##_tmp4 = 0; \ + extern const char sym##_tmp5 __asm("$ld$hide$os4.1$_" #sym); \ + __attribute__((visibility("default"))) const char sym##_tmp5 = 0; \ + extern const char sym##_tmp6 __asm("$ld$hide$os4.2$_" #sym); \ + __attribute__((visibility("default"))) const char sym##_tmp6 = 0; \ + extern const char sym##_tmp7 __asm("$ld$hide$os4.3$_" #sym); \ + __attribute__((visibility("default"))) const char sym##_tmp7 = 0; + NOT_HERE_UNTIL_AFTER_4_3(__absvdi2) NOT_HERE_UNTIL_AFTER_4_3(__absvsi2) NOT_HERE_UNTIL_AFTER_4_3(__adddf3) @@ -339,12 +332,8 @@ NOT_HERE_UNTIL_AFTER_4_3(__divmodsi4) NOT_HERE_UNTIL_AFTER_4_3(__udivmodsi4) #endif // __arm__ && __DYNAMIC__ - - - - -#else /* !__APPLE__ */ +#else // !__APPLE__ extern int avoid_empty_file; -#endif /* !__APPLE__*/ +#endif // !__APPLE__ diff --git a/lib/builtins/arm/adddf3vfp.S b/lib/builtins/arm/adddf3vfp.S index 8e476cad1624..1a271db0847c 100644 --- a/lib/builtins/arm/adddf3vfp.S +++ b/lib/builtins/arm/adddf3vfp.S @@ -1,20 +1,18 @@ //===-- adddf3vfp.S - Implement adddf3vfp ---------------------------------===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// #include "../assembly.h" -// // double __adddf3vfp(double a, double b) { return a + b; } // // Adds two double precision floating point numbers using the Darwin // calling convention where double arguments are passsed in GPR pairs -// + .syntax unified .p2align 2 DEFINE_COMPILERRT_FUNCTION(__adddf3vfp) @@ -23,7 +21,7 @@ DEFINE_COMPILERRT_FUNCTION(__adddf3vfp) #else vmov d6, r0, r1 // move first param from r0/r1 pair into d6 vmov d7, r2, r3 // move second param from r2/r3 pair into d7 - vadd.f64 d6, d6, d7 + vadd.f64 d6, d6, d7 vmov r0, r1, d6 // move result back to r0/r1 pair #endif bx lr diff --git a/lib/builtins/arm/addsf3.S b/lib/builtins/arm/addsf3.S index 74723cbeff74..aa4d40473edb 100644 --- a/lib/builtins/arm/addsf3.S +++ b/lib/builtins/arm/addsf3.S @@ -1,17 +1,16 @@ -/*===-- addsf3.S - Adds two single precision floating pointer numbers-----===// - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - *===----------------------------------------------------------------------===// - * - * This file implements the __addsf3 (single precision floating pointer number - * addition with the IEEE-754 default rounding (to nearest, ties to even) - * function for the ARM Thumb1 ISA. - * - *===----------------------------------------------------------------------===*/ +//===-- addsf3.S - Adds two single precision floating pointer numbers-----===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements the __addsf3 (single precision floating pointer number +// addition with the IEEE-754 default rounding (to nearest, ties to even) +// function for the ARM Thumb1 ISA. +// +//===----------------------------------------------------------------------===// #include "../assembly.h" #define significandBits 23 @@ -29,9 +28,9 @@ DEFINE_COMPILERRT_THUMB_FUNCTION(__addsf3) // Get the absolute value of a and b. lsls r2, r0, #1 lsls r3, r1, #1 - lsrs r2, r2, #1 /* aAbs */ + lsrs r2, r2, #1 // aAbs beq LOCAL_LABEL(a_zero_nan_inf) - lsrs r3, r3, #1 /* bAbs */ + lsrs r3, r3, #1 // bAbs beq LOCAL_LABEL(zero_nan_inf) // Detect if a or b is infinity or Nan. @@ -55,9 +54,9 @@ LOCAL_LABEL(no_swap): // Get the significands and shift them to give us round, guard and sticky. lsls r4, r0, #(typeWidth - significandBits) - lsrs r4, r4, #(typeWidth - significandBits - 3) /* aSignificand << 3 */ + lsrs r4, r4, #(typeWidth - significandBits - 3) // aSignificand << 3 lsls r5, r1, #(typeWidth - significandBits) - lsrs r5, r5, #(typeWidth - significandBits - 3) /* bSignificand << 3 */ + lsrs r5, r5, #(typeWidth - significandBits - 3) // bSignificand << 3 // Get the implicitBit. movs r6, #1 @@ -199,7 +198,7 @@ LOCAL_LABEL(do_substraction): beq 1f movs r7, #1 1: - lsrs r4, r6 /* aSignificand >> shift */ + lsrs r4, r6 // aSignificand >> shift orrs r4, r7 b LOCAL_LABEL(form_result) diff --git a/lib/builtins/arm/addsf3vfp.S b/lib/builtins/arm/addsf3vfp.S index 8871efdcc5d1..c9d1fd1509a8 100644 --- a/lib/builtins/arm/addsf3vfp.S +++ b/lib/builtins/arm/addsf3vfp.S @@ -1,9 +1,8 @@ //===-- addsf3vfp.S - Implement addsf3vfp ---------------------------------===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// diff --git a/lib/builtins/arm/aeabi_cdcmp.S b/lib/builtins/arm/aeabi_cdcmp.S index adc2d55d90f5..bd039a0329ea 100644 --- a/lib/builtins/arm/aeabi_cdcmp.S +++ b/lib/builtins/arm/aeabi_cdcmp.S @@ -1,9 +1,8 @@ //===-- aeabi_cdcmp.S - EABI cdcmp* implementation ------------------------===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// diff --git a/lib/builtins/arm/aeabi_cdcmpeq_check_nan.c b/lib/builtins/arm/aeabi_cdcmpeq_check_nan.c index 7578433a1df7..7bae8743fcf2 100644 --- a/lib/builtins/arm/aeabi_cdcmpeq_check_nan.c +++ b/lib/builtins/arm/aeabi_cdcmpeq_check_nan.c @@ -1,16 +1,15 @@ //===-- lib/arm/aeabi_cdcmpeq_helper.c - Helper for cdcmpeq ---------------===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// -#include <stdint.h> #include "../int_lib.h" +#include <stdint.h> -AEABI_RTABI __attribute__((visibility("hidden"))) -int __aeabi_cdcmpeq_check_nan(double a, double b) { - return __builtin_isnan(a) || __builtin_isnan(b); +AEABI_RTABI __attribute__((visibility("hidden"))) int +__aeabi_cdcmpeq_check_nan(double a, double b) { + return __builtin_isnan(a) || __builtin_isnan(b); } diff --git a/lib/builtins/arm/aeabi_cfcmp.S b/lib/builtins/arm/aeabi_cfcmp.S index 4b1de997687f..a26cb2a3ce16 100644 --- a/lib/builtins/arm/aeabi_cfcmp.S +++ b/lib/builtins/arm/aeabi_cfcmp.S @@ -1,9 +1,8 @@ //===-- aeabi_cfcmp.S - EABI cfcmp* implementation ------------------------===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// diff --git a/lib/builtins/arm/aeabi_cfcmpeq_check_nan.c b/lib/builtins/arm/aeabi_cfcmpeq_check_nan.c index 43dde9a49597..25407337d075 100644 --- a/lib/builtins/arm/aeabi_cfcmpeq_check_nan.c +++ b/lib/builtins/arm/aeabi_cfcmpeq_check_nan.c @@ -1,16 +1,15 @@ //===-- lib/arm/aeabi_cfcmpeq_helper.c - Helper for cdcmpeq ---------------===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// -#include <stdint.h> #include "../int_lib.h" +#include <stdint.h> -AEABI_RTABI __attribute__((visibility("hidden"))) -int __aeabi_cfcmpeq_check_nan(float a, float b) { - return __builtin_isnan(a) || __builtin_isnan(b); +AEABI_RTABI __attribute__((visibility("hidden"))) int +__aeabi_cfcmpeq_check_nan(float a, float b) { + return __builtin_isnan(a) || __builtin_isnan(b); } diff --git a/lib/builtins/arm/aeabi_dcmp.S b/lib/builtins/arm/aeabi_dcmp.S index 9fa78b461248..5f720670ddd7 100644 --- a/lib/builtins/arm/aeabi_dcmp.S +++ b/lib/builtins/arm/aeabi_dcmp.S @@ -1,9 +1,8 @@ //===-- aeabi_dcmp.S - EABI dcmp* implementation ---------------------------===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// diff --git a/lib/builtins/arm/aeabi_div0.c b/lib/builtins/arm/aeabi_div0.c index dc3031326e37..7e8862321d82 100644 --- a/lib/builtins/arm/aeabi_div0.c +++ b/lib/builtins/arm/aeabi_div0.c @@ -1,34 +1,30 @@ -/* ===-- aeabi_div0.c - ARM Runtime ABI support routines for compiler-rt ---=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - * - * This file implements the division by zero helper routines as specified by the - * Run-time ABI for the ARM Architecture. - * - * ===----------------------------------------------------------------------=== - */ +//===-- aeabi_div0.c - ARM Runtime ABI support routines for compiler-rt ---===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements the division by zero helper routines as specified by the +// Run-time ABI for the ARM Architecture. +// +//===----------------------------------------------------------------------===// -/* - * RTABI 4.3.2 - Division by zero - * - * The *div0 functions: - * - Return the value passed to them as a parameter - * - Or, return a fixed value defined by the execution environment (such as 0) - * - Or, raise a signal (often SIGFPE) or throw an exception, and do not return - * - * An application may provide its own implementations of the *div0 functions to - * for a particular behaviour from the *div and *divmod functions called out of - * line. - */ +// RTABI 4.3.2 - Division by zero +// +// The *div0 functions: +// - Return the value passed to them as a parameter +// - Or, return a fixed value defined by the execution environment (such as 0) +// - Or, raise a signal (often SIGFPE) or throw an exception, and do not return +// +// An application may provide its own implementations of the *div0 functions to +// for a particular behaviour from the *div and *divmod functions called out of +// line. #include "../int_lib.h" -/* provide an unused declaration to pacify pendantic compilation */ +// provide an unused declaration to pacify pendantic compilation extern unsigned char declaration; #if defined(__ARM_EABI__) @@ -37,9 +33,8 @@ __aeabi_idiv0(int return_value) { return return_value; } -AEABI_RTABI long long __attribute__((weak)) __attribute__((visibility("hidden"))) -__aeabi_ldiv0(long long return_value) { +AEABI_RTABI long long __attribute__((weak)) +__attribute__((visibility("hidden"))) __aeabi_ldiv0(long long return_value) { return return_value; } #endif - diff --git a/lib/builtins/arm/aeabi_drsub.c b/lib/builtins/arm/aeabi_drsub.c index 1254886086fb..e4e8dc051468 100644 --- a/lib/builtins/arm/aeabi_drsub.c +++ b/lib/builtins/arm/aeabi_drsub.c @@ -1,19 +1,14 @@ //===-- lib/arm/aeabi_drsub.c - Double-precision subtraction --------------===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// #define DOUBLE_PRECISION #include "../fp_lib.h" -AEABI_RTABI fp_t -__aeabi_dsub(fp_t, fp_t); +AEABI_RTABI fp_t __aeabi_dsub(fp_t, fp_t); -AEABI_RTABI fp_t -__aeabi_drsub(fp_t a, fp_t b) { - return __aeabi_dsub(b, a); -} +AEABI_RTABI fp_t __aeabi_drsub(fp_t a, fp_t b) { return __aeabi_dsub(b, a); } diff --git a/lib/builtins/arm/aeabi_fcmp.S b/lib/builtins/arm/aeabi_fcmp.S index ea5b96c21d57..cd311b417096 100644 --- a/lib/builtins/arm/aeabi_fcmp.S +++ b/lib/builtins/arm/aeabi_fcmp.S @@ -1,9 +1,8 @@ //===-- aeabi_fcmp.S - EABI fcmp* implementation ---------------------------===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// diff --git a/lib/builtins/arm/aeabi_frsub.c b/lib/builtins/arm/aeabi_frsub.c index 34f2303745bc..9a363248f14f 100644 --- a/lib/builtins/arm/aeabi_frsub.c +++ b/lib/builtins/arm/aeabi_frsub.c @@ -1,19 +1,14 @@ //===-- lib/arm/aeabi_frsub.c - Single-precision subtraction --------------===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// #define SINGLE_PRECISION #include "../fp_lib.h" -AEABI_RTABI fp_t -__aeabi_fsub(fp_t, fp_t); +AEABI_RTABI fp_t __aeabi_fsub(fp_t, fp_t); -AEABI_RTABI fp_t -__aeabi_frsub(fp_t a, fp_t b) { - return __aeabi_fsub(b, a); -} +AEABI_RTABI fp_t __aeabi_frsub(fp_t a, fp_t b) { return __aeabi_fsub(b, a); } diff --git a/lib/builtins/arm/aeabi_idivmod.S b/lib/builtins/arm/aeabi_idivmod.S index 9c9c80ab5a7b..bb80e4b96fc1 100644 --- a/lib/builtins/arm/aeabi_idivmod.S +++ b/lib/builtins/arm/aeabi_idivmod.S @@ -1,9 +1,8 @@ //===-- aeabi_idivmod.S - EABI idivmod implementation ---------------------===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// diff --git a/lib/builtins/arm/aeabi_ldivmod.S b/lib/builtins/arm/aeabi_ldivmod.S index 038ae5d723a3..d0d06be6f329 100644 --- a/lib/builtins/arm/aeabi_ldivmod.S +++ b/lib/builtins/arm/aeabi_ldivmod.S @@ -1,9 +1,8 @@ //===-- aeabi_ldivmod.S - EABI ldivmod implementation ---------------------===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// diff --git a/lib/builtins/arm/aeabi_memcmp.S b/lib/builtins/arm/aeabi_memcmp.S index e86d6113760e..41637289c1bc 100644 --- a/lib/builtins/arm/aeabi_memcmp.S +++ b/lib/builtins/arm/aeabi_memcmp.S @@ -1,9 +1,8 @@ //===-- aeabi_memcmp.S - EABI memcmp implementation -----------------------===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// diff --git a/lib/builtins/arm/aeabi_memcpy.S b/lib/builtins/arm/aeabi_memcpy.S index e83c5fd4dbb3..93e1b05d54ce 100644 --- a/lib/builtins/arm/aeabi_memcpy.S +++ b/lib/builtins/arm/aeabi_memcpy.S @@ -1,9 +1,8 @@ //===-- aeabi_memcpy.S - EABI memcpy implementation -----------------------===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// diff --git a/lib/builtins/arm/aeabi_memmove.S b/lib/builtins/arm/aeabi_memmove.S index ee28300e46f2..c2f0fa4cd8ff 100644 --- a/lib/builtins/arm/aeabi_memmove.S +++ b/lib/builtins/arm/aeabi_memmove.S @@ -1,9 +1,8 @@ //===-- aeabi_memmove.S - EABI memmove implementation --------------------===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===---------------------------------------------------------------------===// diff --git a/lib/builtins/arm/aeabi_memset.S b/lib/builtins/arm/aeabi_memset.S index 0a678d7627e7..2aa8ec0c4679 100644 --- a/lib/builtins/arm/aeabi_memset.S +++ b/lib/builtins/arm/aeabi_memset.S @@ -1,9 +1,8 @@ //===-- aeabi_memset.S - EABI memset implementation -----------------------===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// diff --git a/lib/builtins/arm/aeabi_uidivmod.S b/lib/builtins/arm/aeabi_uidivmod.S index 88a4a6d8bc12..df030769fd40 100644 --- a/lib/builtins/arm/aeabi_uidivmod.S +++ b/lib/builtins/arm/aeabi_uidivmod.S @@ -1,9 +1,8 @@ //===-- aeabi_uidivmod.S - EABI uidivmod implementation -------------------===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// diff --git a/lib/builtins/arm/aeabi_uldivmod.S b/lib/builtins/arm/aeabi_uldivmod.S index be343b6bc826..4fc97704d3c5 100644 --- a/lib/builtins/arm/aeabi_uldivmod.S +++ b/lib/builtins/arm/aeabi_uldivmod.S @@ -1,9 +1,8 @@ //===-- aeabi_uldivmod.S - EABI uldivmod implementation -------------------===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// diff --git a/lib/builtins/arm/bswapdi2.S b/lib/builtins/arm/bswapdi2.S index e9db8bac7994..271df8bd6ad0 100644 --- a/lib/builtins/arm/bswapdi2.S +++ b/lib/builtins/arm/bswapdi2.S @@ -1,9 +1,8 @@ //===------- bswapdi2 - Implement bswapdi2 --------------------------------===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// diff --git a/lib/builtins/arm/bswapsi2.S b/lib/builtins/arm/bswapsi2.S index 1f6eed5c1bbf..07cc3d89190e 100644 --- a/lib/builtins/arm/bswapsi2.S +++ b/lib/builtins/arm/bswapsi2.S @@ -1,9 +1,8 @@ //===------- bswapsi2 - Implement bswapsi2 --------------------------------===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// diff --git a/lib/builtins/arm/chkstk.S b/lib/builtins/arm/chkstk.S index e3002105897e..c5c9ebe0a531 100644 --- a/lib/builtins/arm/chkstk.S +++ b/lib/builtins/arm/chkstk.S @@ -1,5 +1,6 @@ -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception #include "../assembly.h" diff --git a/lib/builtins/arm/clzdi2.S b/lib/builtins/arm/clzdi2.S index fc03b385cdfa..685668b11b8e 100644 --- a/lib/builtins/arm/clzdi2.S +++ b/lib/builtins/arm/clzdi2.S @@ -1,16 +1,15 @@ -/* ===-- clzdi2.c - Implement __clzdi2 -------------------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - * - * This file implements count leading zeros for 64bit arguments. - * - * ===----------------------------------------------------------------------=== - */ +//===-- clzdi2.c - Implement __clzdi2 -------------------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements count leading zeros for 64bit arguments. +// +//===----------------------------------------------------------------------===// + #include "../assembly.h" .syntax unified @@ -35,14 +34,12 @@ DEFINE_COMPILERRT_FUNCTION(__clzdi2) #endif JMP(lr) #else - /* Assumption: n != 0 */ + // Assumption: n != 0 - /* - * r0: n - * r1: upper half of n, overwritten after check - * r1: count of leading zeros in n + 1 - * r2: scratch register for shifted r0 - */ + // r0: n + // r1: upper half of n, overwritten after check + // r1: count of leading zeros in n + 1 + // r2: scratch register for shifted r0 #ifdef __ARMEB__ cmp r0, 0 moveq r0, r1 @@ -53,14 +50,12 @@ DEFINE_COMPILERRT_FUNCTION(__clzdi2) movne r1, 1 moveq r1, 33 - /* - * Basic block: - * if ((r0 >> SHIFT) == 0) - * r1 += SHIFT; - * else - * r0 >>= SHIFT; - * for descending powers of two as SHIFT. - */ + // Basic block: + // if ((r0 >> SHIFT) == 0) + // r1 += SHIFT; + // else + // r0 >>= SHIFT; + // for descending powers of two as SHIFT. #define BLOCK(shift) \ lsrs r2, r0, shift; \ movne r0, r2; \ @@ -71,18 +66,16 @@ DEFINE_COMPILERRT_FUNCTION(__clzdi2) BLOCK(4) BLOCK(2) - /* - * The basic block invariants at this point are (r0 >> 2) == 0 and - * r0 != 0. This means 1 <= r0 <= 3 and 0 <= (r0 >> 1) <= 1. - * - * r0 | (r0 >> 1) == 0 | (r0 >> 1) == 1 | -(r0 >> 1) | 1 - (r0 >> 1) - * ---+----------------+----------------+------------+-------------- - * 1 | 1 | 0 | 0 | 1 - * 2 | 0 | 1 | -1 | 0 - * 3 | 0 | 1 | -1 | 0 - * - * The r1's initial value of 1 compensates for the 1 here. - */ + // The basic block invariants at this point are (r0 >> 2) == 0 and + // r0 != 0. This means 1 <= r0 <= 3 and 0 <= (r0 >> 1) <= 1. + // + // r0 | (r0 >> 1) == 0 | (r0 >> 1) == 1 | -(r0 >> 1) | 1 - (r0 >> 1) + // ---+----------------+----------------+------------+-------------- + // 1 | 1 | 0 | 0 | 1 + // 2 | 0 | 1 | -1 | 0 + // 3 | 0 | 1 | -1 | 0 + // + // The r1's initial value of 1 compensates for the 1 here. sub r0, r1, r0, lsr #1 JMP(lr) diff --git a/lib/builtins/arm/clzsi2.S b/lib/builtins/arm/clzsi2.S index f2ce59c90119..5d86fe486edd 100644 --- a/lib/builtins/arm/clzsi2.S +++ b/lib/builtins/arm/clzsi2.S @@ -1,16 +1,15 @@ -/* ===-- clzsi2.c - Implement __clzsi2 -------------------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - * - * This file implements count leading zeros for 32bit arguments. - * - * ===----------------------------------------------------------------------=== - */ +//===-- clzsi2.c - Implement __clzsi2 -------------------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements count leading zeros for 32bit arguments. +// +//===----------------------------------------------------------------------===// + #include "../assembly.h" .syntax unified @@ -23,23 +22,19 @@ DEFINE_COMPILERRT_FUNCTION(__clzsi2) clz r0, r0 JMP(lr) #else - /* Assumption: n != 0 */ + // Assumption: n != 0 - /* - * r0: n - * r1: count of leading zeros in n + 1 - * r2: scratch register for shifted r0 - */ + // r0: n + // r1: count of leading zeros in n + 1 + // r2: scratch register for shifted r0 mov r1, 1 - /* - * Basic block: - * if ((r0 >> SHIFT) == 0) - * r1 += SHIFT; - * else - * r0 >>= SHIFT; - * for descending powers of two as SHIFT. - */ + // Basic block: + // if ((r0 >> SHIFT) == 0) + // r1 += SHIFT; + // else + // r0 >>= SHIFT; + // for descending powers of two as SHIFT. #define BLOCK(shift) \ lsrs r2, r0, shift; \ @@ -51,18 +46,16 @@ DEFINE_COMPILERRT_FUNCTION(__clzsi2) BLOCK(4) BLOCK(2) - /* - * The basic block invariants at this point are (r0 >> 2) == 0 and - * r0 != 0. This means 1 <= r0 <= 3 and 0 <= (r0 >> 1) <= 1. - * - * r0 | (r0 >> 1) == 0 | (r0 >> 1) == 1 | -(r0 >> 1) | 1 - (r0 >> 1) - * ---+----------------+----------------+------------+-------------- - * 1 | 1 | 0 | 0 | 1 - * 2 | 0 | 1 | -1 | 0 - * 3 | 0 | 1 | -1 | 0 - * - * The r1's initial value of 1 compensates for the 1 here. - */ + // The basic block invariants at this point are (r0 >> 2) == 0 and + // r0 != 0. This means 1 <= r0 <= 3 and 0 <= (r0 >> 1) <= 1. + // + // r0 | (r0 >> 1) == 0 | (r0 >> 1) == 1 | -(r0 >> 1) | 1 - (r0 >> 1) + // ---+----------------+----------------+------------+-------------- + // 1 | 1 | 0 | 0 | 1 + // 2 | 0 | 1 | -1 | 0 + // 3 | 0 | 1 | -1 | 0 + // + // The r1's initial value of 1 compensates for the 1 here. sub r0, r1, r0, lsr #1 JMP(lr) diff --git a/lib/builtins/arm/comparesf2.S b/lib/builtins/arm/comparesf2.S index c6c4cc067f07..24b85d2fee15 100644 --- a/lib/builtins/arm/comparesf2.S +++ b/lib/builtins/arm/comparesf2.S @@ -1,9 +1,8 @@ //===-- comparesf2.S - Implement single-precision soft-float comparisons --===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// // @@ -38,14 +37,12 @@ //===----------------------------------------------------------------------===// #include "../assembly.h" + .syntax unified .text DEFINE_CODE_STATE -@ int __eqsf2(float a, float b) - - .p2align 2 -DEFINE_COMPILERRT_FUNCTION(__eqsf2) + .macro COMPARESF2_FUNCTION_BODY handle_nan:req #if defined(COMPILER_RT_ARMHF_TARGET) vmov r0, s0 vmov r1, s1 @@ -113,9 +110,9 @@ DEFINE_COMPILERRT_FUNCTION(__eqsf2) // b < 0 ? 1 : -1. Same if a and b have the opposite sign (ignoring Nan). movs r0, #1 lsrs r1, #31 - bne LOCAL_LABEL(CHECK_NAN) + bne LOCAL_LABEL(CHECK_NAN\@) negs r0, r0 - b LOCAL_LABEL(CHECK_NAN) + b LOCAL_LABEL(CHECK_NAN\@) 1: #else it lo @@ -131,7 +128,7 @@ DEFINE_COMPILERRT_FUNCTION(__eqsf2) // Here both have the same sign and absA > absB. movs r0, #1 lsrs r1, #31 - beq LOCAL_LABEL(CHECK_NAN) + beq LOCAL_LABEL(CHECK_NAN\@) negs r0, r0 1: #else @@ -152,7 +149,7 @@ DEFINE_COMPILERRT_FUNCTION(__eqsf2) // Finally, we need to deal with NaNs. If either argument is NaN, replace // the value in r0 with 1. #if defined(USE_THUMB_1) -LOCAL_LABEL(CHECK_NAN): +LOCAL_LABEL(CHECK_NAN\@): movs r6, #0xff lsls r6, #24 cmp r2, r6 @@ -160,92 +157,60 @@ LOCAL_LABEL(CHECK_NAN): cmp r3, r6 1: bls 2f - movs r0, #1 + \handle_nan 2: pop {r6, pc} #else cmp r2, #0xff000000 ite ls cmpls r3, #0xff000000 - movhi r0, #1 + \handle_nan JMP(lr) #endif + .endm + +@ int __eqsf2(float a, float b) + + .p2align 2 +DEFINE_COMPILERRT_FUNCTION(__eqsf2) + + .macro __eqsf2_handle_nan +#if defined(USE_THUMB_1) + movs r0, #1 +#else + movhi r0, #1 +#endif + .endm + +COMPARESF2_FUNCTION_BODY __eqsf2_handle_nan + END_COMPILERRT_FUNCTION(__eqsf2) DEFINE_COMPILERRT_FUNCTION_ALIAS(__lesf2, __eqsf2) DEFINE_COMPILERRT_FUNCTION_ALIAS(__ltsf2, __eqsf2) DEFINE_COMPILERRT_FUNCTION_ALIAS(__nesf2, __eqsf2) +#if defined(__ELF__) +// Alias for libgcc compatibility +DEFINE_COMPILERRT_FUNCTION_ALIAS(__cmpsf2, __lesf2) +#endif + @ int __gtsf2(float a, float b) .p2align 2 DEFINE_COMPILERRT_FUNCTION(__gtsf2) - // Identical to the preceding except in that we return -1 for NaN values. - // Given that the two paths share so much code, one might be tempted to - // unify them; however, the extra code needed to do so makes the code size - // to performance tradeoff very hard to justify for such small functions. -#if defined(COMPILER_RT_ARMHF_TARGET) - vmov r0, s0 - vmov r1, s1 -#endif + + .macro __gtsf2_handle_nan #if defined(USE_THUMB_1) - push {r6, lr} - lsls r2, r0, #1 - lsls r3, r1, #1 - lsrs r6, r3, #1 - orrs r6, r2 - beq 1f - movs r6, r0 - eors r6, r1 -1: - bmi 2f - subs r0, r2, r3 -2: - bhs 3f - movs r0, #1 - lsrs r1, #31 - bne LOCAL_LABEL(CHECK_NAN_2) - negs r0, r0 - b LOCAL_LABEL(CHECK_NAN_2) -3: - bls 4f - movs r0, #1 - lsrs r1, #31 - beq LOCAL_LABEL(CHECK_NAN_2) - negs r0, r0 -4: -LOCAL_LABEL(CHECK_NAN_2): - movs r6, #0xff - lsls r6, #24 - cmp r2, r6 - bhi 5f - cmp r3, r6 -5: - bls 6f movs r0, #1 negs r0, r0 -6: - pop {r6, pc} #else - mov r2, r0, lsl #1 - mov r3, r1, lsl #1 - orrs r12, r2, r3, lsr #1 - it ne - eorsne r12, r0, r1 - it pl - subspl r0, r2, r3 - it lo - mvnlo r0, r1, asr #31 - it hi - movhi r0, r1, asr #31 - it ne - orrne r0, r0, #1 - cmp r2, #0xff000000 - ite ls - cmpls r3, #0xff000000 movhi r0, #-1 - JMP(lr) #endif + .endm + +COMPARESF2_FUNCTION_BODY __gtsf2_handle_nan + END_COMPILERRT_FUNCTION(__gtsf2) DEFINE_COMPILERRT_FUNCTION_ALIAS(__gesf2, __gtsf2) @@ -283,11 +248,11 @@ DEFINE_COMPILERRT_FUNCTION(__unordsf2) END_COMPILERRT_FUNCTION(__unordsf2) #if defined(COMPILER_RT_ARMHF_TARGET) -DEFINE_COMPILERRT_FUNCTION(__aeabi_fcmpum) +DEFINE_COMPILERRT_FUNCTION(__aeabi_fcmpun) vmov s0, r0 vmov s1, r1 b SYMBOL_NAME(__unordsf2) -END_COMPILERRT_FUNCTION(__aeabi_fcmpum) +END_COMPILERRT_FUNCTION(__aeabi_fcmpun) #else DEFINE_AEABI_FUNCTION_ALIAS(__aeabi_fcmpun, __unordsf2) #endif diff --git a/lib/builtins/arm/divdf3vfp.S b/lib/builtins/arm/divdf3vfp.S index 776ba4f24b47..ad50b57a651d 100644 --- a/lib/builtins/arm/divdf3vfp.S +++ b/lib/builtins/arm/divdf3vfp.S @@ -1,9 +1,8 @@ //===-- divdf3vfp.S - Implement divdf3vfp ---------------------------------===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// diff --git a/lib/builtins/arm/divmodsi4.S b/lib/builtins/arm/divmodsi4.S index 8a027b741efe..f94438dfd593 100644 --- a/lib/builtins/arm/divmodsi4.S +++ b/lib/builtins/arm/divmodsi4.S @@ -1,17 +1,16 @@ -/*===-- divmodsi4.S - 32-bit signed integer divide and modulus ------------===// - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - *===----------------------------------------------------------------------===// - * - * This file implements the __divmodsi4 (32-bit signed integer divide and - * modulus) function for the ARM architecture. A naive digit-by-digit - * computation is employed for simplicity. - * - *===----------------------------------------------------------------------===*/ +//===-- divmodsi4.S - 32-bit signed integer divide and modulus ------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements the __divmodsi4 (32-bit signed integer divide and +// modulus) function for the ARM architecture. A naive digit-by-digit +// computation is employed for simplicity. +// +//===----------------------------------------------------------------------===// #include "../assembly.h" diff --git a/lib/builtins/arm/divsf3vfp.S b/lib/builtins/arm/divsf3vfp.S index 130318f0c37b..958a6724bbe3 100644 --- a/lib/builtins/arm/divsf3vfp.S +++ b/lib/builtins/arm/divsf3vfp.S @@ -1,9 +1,8 @@ //===-- divsf3vfp.S - Implement divsf3vfp ---------------------------------===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// diff --git a/lib/builtins/arm/divsi3.S b/lib/builtins/arm/divsi3.S index 19757af177eb..761bf49d3ed0 100644 --- a/lib/builtins/arm/divsi3.S +++ b/lib/builtins/arm/divsi3.S @@ -1,16 +1,15 @@ -/*===-- divsi3.S - 32-bit signed integer divide ---------------------------===// - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - *===----------------------------------------------------------------------===// - * - * This file implements the __divsi3 (32-bit signed integer divide) function - * for the ARM architecture as a wrapper around the unsigned routine. - * - *===----------------------------------------------------------------------===*/ +//===-- divsi3.S - 32-bit signed integer divide ---------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements the __divsi3 (32-bit signed integer divide) function +// for the ARM architecture as a wrapper around the unsigned routine. +// +//===----------------------------------------------------------------------===// #include "../assembly.h" diff --git a/lib/builtins/arm/eqdf2vfp.S b/lib/builtins/arm/eqdf2vfp.S index d50706570916..2a0a64b97e7d 100644 --- a/lib/builtins/arm/eqdf2vfp.S +++ b/lib/builtins/arm/eqdf2vfp.S @@ -1,21 +1,19 @@ //===-- eqdf2vfp.S - Implement eqdf2vfp -----------------------------------===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// #include "../assembly.h" -// // extern int __eqdf2vfp(double a, double b); // // Returns one iff a == b and neither is NaN. -// Uses Darwin calling convention where double precision arguments are passsed +// Uses Darwin calling convention where double precision arguments are passsed // like in GPR pairs. -// + .syntax unified .p2align 2 DEFINE_COMPILERRT_FUNCTION(__eqdf2vfp) @@ -24,7 +22,7 @@ DEFINE_COMPILERRT_FUNCTION(__eqdf2vfp) #else vmov d6, r0, r1 // load r0/r1 pair in double register vmov d7, r2, r3 // load r2/r3 pair in double register - vcmp.f64 d6, d7 + vcmp.f64 d6, d7 #endif vmrs apsr_nzcv, fpscr ITE(eq) diff --git a/lib/builtins/arm/eqsf2vfp.S b/lib/builtins/arm/eqsf2vfp.S index fd72b2fdbdee..5fefe7b71762 100644 --- a/lib/builtins/arm/eqsf2vfp.S +++ b/lib/builtins/arm/eqsf2vfp.S @@ -1,9 +1,8 @@ //===-- eqsf2vfp.S - Implement eqsf2vfp -----------------------------------===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// @@ -13,7 +12,7 @@ // extern int __eqsf2vfp(float a, float b); // // Returns one iff a == b and neither is NaN. -// Uses Darwin calling convention where single precision arguments are passsed +// Uses Darwin calling convention where single precision arguments are passsed // like 32-bit ints // .syntax unified diff --git a/lib/builtins/arm/extendsfdf2vfp.S b/lib/builtins/arm/extendsfdf2vfp.S index 1079f977bae6..37c8be8dcd9c 100644 --- a/lib/builtins/arm/extendsfdf2vfp.S +++ b/lib/builtins/arm/extendsfdf2vfp.S @@ -1,9 +1,8 @@ //===-- extendsfdf2vfp.S - Implement extendsfdf2vfp -----------------------===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// @@ -13,7 +12,7 @@ // extern double __extendsfdf2vfp(float a); // // Converts single precision float to double precision result. -// Uses Darwin calling convention where a single precision parameter is +// Uses Darwin calling convention where a single precision parameter is // passed in a GPR and a double precision result is returned in R0/R1 pair. // .syntax unified diff --git a/lib/builtins/arm/fixdfsivfp.S b/lib/builtins/arm/fixdfsivfp.S index 5d7b0f856549..af1d4f4fa5f5 100644 --- a/lib/builtins/arm/fixdfsivfp.S +++ b/lib/builtins/arm/fixdfsivfp.S @@ -1,9 +1,8 @@ //===-- fixdfsivfp.S - Implement fixdfsivfp -----------------------===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// @@ -13,7 +12,7 @@ // extern int __fixdfsivfp(double a); // // Converts double precision float to a 32-bit int rounding towards zero. -// Uses Darwin calling convention where a double precision parameter is +// Uses Darwin calling convention where a double precision parameter is // passed in GPR register pair. // .syntax unified diff --git a/lib/builtins/arm/fixsfsivfp.S b/lib/builtins/arm/fixsfsivfp.S index 805a277afa34..30b2f3cf9364 100644 --- a/lib/builtins/arm/fixsfsivfp.S +++ b/lib/builtins/arm/fixsfsivfp.S @@ -1,9 +1,8 @@ //===-- fixsfsivfp.S - Implement fixsfsivfp -----------------------===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// @@ -13,7 +12,7 @@ // extern int __fixsfsivfp(float a); // // Converts single precision float to a 32-bit int rounding towards zero. -// Uses Darwin calling convention where a single precision parameter is +// Uses Darwin calling convention where a single precision parameter is // passed in a GPR.. // .syntax unified diff --git a/lib/builtins/arm/fixunsdfsivfp.S b/lib/builtins/arm/fixunsdfsivfp.S index 4f1b2c8cefdc..44e6dbd4989e 100644 --- a/lib/builtins/arm/fixunsdfsivfp.S +++ b/lib/builtins/arm/fixunsdfsivfp.S @@ -1,9 +1,8 @@ //===-- fixunsdfsivfp.S - Implement fixunsdfsivfp -------------------------===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// @@ -12,9 +11,9 @@ // // extern unsigned int __fixunsdfsivfp(double a); // -// Converts double precision float to a 32-bit unsigned int rounding towards +// Converts double precision float to a 32-bit unsigned int rounding towards // zero. All negative values become zero. -// Uses Darwin calling convention where a double precision parameter is +// Uses Darwin calling convention where a double precision parameter is // passed in GPR register pair. // .syntax unified diff --git a/lib/builtins/arm/fixunssfsivfp.S b/lib/builtins/arm/fixunssfsivfp.S index e5d778236879..5d6ee7cce27f 100644 --- a/lib/builtins/arm/fixunssfsivfp.S +++ b/lib/builtins/arm/fixunssfsivfp.S @@ -1,9 +1,8 @@ //===-- fixunssfsivfp.S - Implement fixunssfsivfp -------------------------===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// @@ -12,9 +11,9 @@ // // extern unsigned int __fixunssfsivfp(float a); // -// Converts single precision float to a 32-bit unsigned int rounding towards +// Converts single precision float to a 32-bit unsigned int rounding towards // zero. All negative values become zero. -// Uses Darwin calling convention where a single precision parameter is +// Uses Darwin calling convention where a single precision parameter is // passed in a GPR.. // .syntax unified diff --git a/lib/builtins/arm/floatsidfvfp.S b/lib/builtins/arm/floatsidfvfp.S index 3297ad44d8cd..ae8d2465889c 100644 --- a/lib/builtins/arm/floatsidfvfp.S +++ b/lib/builtins/arm/floatsidfvfp.S @@ -1,9 +1,8 @@ //===-- floatsidfvfp.S - Implement floatsidfvfp ---------------------------===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// @@ -13,7 +12,7 @@ // extern double __floatsidfvfp(int a); // // Converts a 32-bit int to a double precision float. -// Uses Darwin calling convention where a double precision result is +// Uses Darwin calling convention where a double precision result is // return in GPR register pair. // .syntax unified diff --git a/lib/builtins/arm/floatsisfvfp.S b/lib/builtins/arm/floatsisfvfp.S index 65408b54b8d4..a36bc5ee5ee0 100644 --- a/lib/builtins/arm/floatsisfvfp.S +++ b/lib/builtins/arm/floatsisfvfp.S @@ -1,9 +1,8 @@ //===-- floatsisfvfp.S - Implement floatsisfvfp ---------------------------===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// @@ -13,7 +12,7 @@ // extern float __floatsisfvfp(int a); // // Converts single precision float to a 32-bit int rounding towards zero. -// Uses Darwin calling convention where a single precision result is +// Uses Darwin calling convention where a single precision result is // return in a GPR.. // .syntax unified diff --git a/lib/builtins/arm/floatunssidfvfp.S b/lib/builtins/arm/floatunssidfvfp.S index d7a7024a25b8..0932dab2bdb9 100644 --- a/lib/builtins/arm/floatunssidfvfp.S +++ b/lib/builtins/arm/floatunssidfvfp.S @@ -1,9 +1,8 @@ //===-- floatunssidfvfp.S - Implement floatunssidfvfp ---------------------===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// @@ -13,7 +12,7 @@ // extern double __floatunssidfvfp(unsigned int a); // // Converts a 32-bit int to a double precision float. -// Uses Darwin calling convention where a double precision result is +// Uses Darwin calling convention where a double precision result is // return in GPR register pair. // .syntax unified diff --git a/lib/builtins/arm/floatunssisfvfp.S b/lib/builtins/arm/floatunssisfvfp.S index 1ca856519a92..9578546f40d6 100644 --- a/lib/builtins/arm/floatunssisfvfp.S +++ b/lib/builtins/arm/floatunssisfvfp.S @@ -1,9 +1,8 @@ //===-- floatunssisfvfp.S - Implement floatunssisfvfp ---------------------===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// @@ -13,7 +12,7 @@ // extern float __floatunssisfvfp(unsigned int a); // // Converts single precision float to a 32-bit int rounding towards zero. -// Uses Darwin calling convention where a single precision result is +// Uses Darwin calling convention where a single precision result is // return in a GPR.. // .syntax unified diff --git a/lib/builtins/arm/gedf2vfp.S b/lib/builtins/arm/gedf2vfp.S index 364fc5b24cd1..2af9d909967b 100644 --- a/lib/builtins/arm/gedf2vfp.S +++ b/lib/builtins/arm/gedf2vfp.S @@ -1,9 +1,8 @@ //===-- gedf2vfp.S - Implement gedf2vfp -----------------------------------===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// @@ -13,7 +12,7 @@ // extern int __gedf2vfp(double a, double b); // // Returns one iff a >= b and neither is NaN. -// Uses Darwin calling convention where double precision arguments are passsed +// Uses Darwin calling convention where double precision arguments are passsed // like in GPR pairs. // .syntax unified diff --git a/lib/builtins/arm/gesf2vfp.S b/lib/builtins/arm/gesf2vfp.S index 346c3473ae4c..cedd1e13eb22 100644 --- a/lib/builtins/arm/gesf2vfp.S +++ b/lib/builtins/arm/gesf2vfp.S @@ -1,9 +1,8 @@ //===-- gesf2vfp.S - Implement gesf2vfp -----------------------------------===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// @@ -13,7 +12,7 @@ // extern int __gesf2vfp(float a, float b); // // Returns one iff a >= b and neither is NaN. -// Uses Darwin calling convention where single precision arguments are passsed +// Uses Darwin calling convention where single precision arguments are passsed // like 32-bit ints // .syntax unified diff --git a/lib/builtins/arm/gtdf2vfp.S b/lib/builtins/arm/gtdf2vfp.S index 3389c3ad9737..782ad8cac013 100644 --- a/lib/builtins/arm/gtdf2vfp.S +++ b/lib/builtins/arm/gtdf2vfp.S @@ -1,9 +1,8 @@ //===-- gtdf2vfp.S - Implement gtdf2vfp -----------------------------------===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// @@ -13,7 +12,7 @@ // extern double __gtdf2vfp(double a, double b); // // Returns one iff a > b and neither is NaN. -// Uses Darwin calling convention where double precision arguments are passsed +// Uses Darwin calling convention where double precision arguments are passsed // like in GPR pairs. // .syntax unified diff --git a/lib/builtins/arm/gtsf2vfp.S b/lib/builtins/arm/gtsf2vfp.S index afdba8b018e2..1cc2bd14f810 100644 --- a/lib/builtins/arm/gtsf2vfp.S +++ b/lib/builtins/arm/gtsf2vfp.S @@ -1,9 +1,8 @@ //===-- gtsf2vfp.S - Implement gtsf2vfp -----------------------------------===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// @@ -13,7 +12,7 @@ // extern int __gtsf2vfp(float a, float b); // // Returns one iff a > b and neither is NaN. -// Uses Darwin calling convention where single precision arguments are passsed +// Uses Darwin calling convention where single precision arguments are passsed // like 32-bit ints // .syntax unified diff --git a/lib/builtins/arm/ledf2vfp.S b/lib/builtins/arm/ledf2vfp.S index 4bbe4c86837c..0097e4b6c129 100644 --- a/lib/builtins/arm/ledf2vfp.S +++ b/lib/builtins/arm/ledf2vfp.S @@ -1,9 +1,8 @@ //===-- ledf2vfp.S - Implement ledf2vfp -----------------------------------===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// @@ -13,7 +12,7 @@ // extern double __ledf2vfp(double a, double b); // // Returns one iff a <= b and neither is NaN. -// Uses Darwin calling convention where double precision arguments are passsed +// Uses Darwin calling convention where double precision arguments are passsed // like in GPR pairs. // .syntax unified diff --git a/lib/builtins/arm/lesf2vfp.S b/lib/builtins/arm/lesf2vfp.S index 51232bd8cedc..2052d3869145 100644 --- a/lib/builtins/arm/lesf2vfp.S +++ b/lib/builtins/arm/lesf2vfp.S @@ -1,9 +1,8 @@ //===-- lesf2vfp.S - Implement lesf2vfp -----------------------------------===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// @@ -13,7 +12,7 @@ // extern int __lesf2vfp(float a, float b); // // Returns one iff a <= b and neither is NaN. -// Uses Darwin calling convention where single precision arguments are passsed +// Uses Darwin calling convention where single precision arguments are passsed // like 32-bit ints // .syntax unified diff --git a/lib/builtins/arm/ltdf2vfp.S b/lib/builtins/arm/ltdf2vfp.S index 8e2928c813d2..a126aa9e0536 100644 --- a/lib/builtins/arm/ltdf2vfp.S +++ b/lib/builtins/arm/ltdf2vfp.S @@ -1,9 +1,8 @@ //===-- ltdf2vfp.S - Implement ltdf2vfp -----------------------------------===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// @@ -13,7 +12,7 @@ // extern double __ltdf2vfp(double a, double b); // // Returns one iff a < b and neither is NaN. -// Uses Darwin calling convention where double precision arguments are passsed +// Uses Darwin calling convention where double precision arguments are passsed // like in GPR pairs. // .syntax unified diff --git a/lib/builtins/arm/ltsf2vfp.S b/lib/builtins/arm/ltsf2vfp.S index 59c00c6bab67..ba10d71d2095 100644 --- a/lib/builtins/arm/ltsf2vfp.S +++ b/lib/builtins/arm/ltsf2vfp.S @@ -1,9 +1,8 @@ //===-- ltsf2vfp.S - Implement ltsf2vfp -----------------------------------===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// @@ -13,7 +12,7 @@ // extern int __ltsf2vfp(float a, float b); // // Returns one iff a < b and neither is NaN. -// Uses Darwin calling convention where single precision arguments are passsed +// Uses Darwin calling convention where single precision arguments are passsed // like 32-bit ints // .syntax unified diff --git a/lib/builtins/arm/modsi3.S b/lib/builtins/arm/modsi3.S index be263834d7f1..5312f5b41c4d 100644 --- a/lib/builtins/arm/modsi3.S +++ b/lib/builtins/arm/modsi3.S @@ -1,16 +1,15 @@ -/*===-- modsi3.S - 32-bit signed integer modulus --------------------------===// - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - *===----------------------------------------------------------------------===// - * - * This file implements the __modsi3 (32-bit signed integer modulus) function - * for the ARM architecture as a wrapper around the unsigned routine. - * - *===----------------------------------------------------------------------===*/ +//===-- modsi3.S - 32-bit signed integer modulus --------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements the __modsi3 (32-bit signed integer modulus) function +// for the ARM architecture as a wrapper around the unsigned routine. +// +//===----------------------------------------------------------------------===// #include "../assembly.h" diff --git a/lib/builtins/arm/muldf3vfp.S b/lib/builtins/arm/muldf3vfp.S index aa7b23495034..9adc937bcb3f 100644 --- a/lib/builtins/arm/muldf3vfp.S +++ b/lib/builtins/arm/muldf3vfp.S @@ -1,9 +1,8 @@ //===-- muldf3vfp.S - Implement muldf3vfp ---------------------------------===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// diff --git a/lib/builtins/arm/mulsf3vfp.S b/lib/builtins/arm/mulsf3vfp.S index a1da789dcade..a94131bebb92 100644 --- a/lib/builtins/arm/mulsf3vfp.S +++ b/lib/builtins/arm/mulsf3vfp.S @@ -1,9 +1,8 @@ //===-- mulsf3vfp.S - Implement mulsf3vfp ---------------------------------===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// diff --git a/lib/builtins/arm/nedf2vfp.S b/lib/builtins/arm/nedf2vfp.S index aef72eb00974..32d35c41d466 100644 --- a/lib/builtins/arm/nedf2vfp.S +++ b/lib/builtins/arm/nedf2vfp.S @@ -1,21 +1,19 @@ //===-- nedf2vfp.S - Implement nedf2vfp -----------------------------------===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// #include "../assembly.h" -// // extern double __nedf2vfp(double a, double b); // // Returns zero if a and b are unequal and neither is NaN. -// Uses Darwin calling convention where double precision arguments are passsed +// Uses Darwin calling convention where double precision arguments are passsed // like in GPR pairs. -// + .syntax unified .p2align 2 DEFINE_COMPILERRT_FUNCTION(__nedf2vfp) @@ -24,7 +22,7 @@ DEFINE_COMPILERRT_FUNCTION(__nedf2vfp) #else vmov d6, r0, r1 // load r0/r1 pair in double register vmov d7, r2, r3 // load r2/r3 pair in double register - vcmp.f64 d6, d7 + vcmp.f64 d6, d7 #endif vmrs apsr_nzcv, fpscr ITE(ne) diff --git a/lib/builtins/arm/negdf2vfp.S b/lib/builtins/arm/negdf2vfp.S index 81f0ab8eec1d..b7cf91877e38 100644 --- a/lib/builtins/arm/negdf2vfp.S +++ b/lib/builtins/arm/negdf2vfp.S @@ -1,9 +1,8 @@ //===-- negdf2vfp.S - Implement negdf2vfp ---------------------------------===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// @@ -12,7 +11,7 @@ // // extern double __negdf2vfp(double a, double b); // -// Returns the negation a double precision floating point numbers using the +// Returns the negation a double precision floating point numbers using the // Darwin calling convention where double arguments are passsed in GPR pairs. // .syntax unified diff --git a/lib/builtins/arm/negsf2vfp.S b/lib/builtins/arm/negsf2vfp.S index 46ab4a9cf164..b6d3c618927f 100644 --- a/lib/builtins/arm/negsf2vfp.S +++ b/lib/builtins/arm/negsf2vfp.S @@ -1,9 +1,8 @@ //===-- negsf2vfp.S - Implement negsf2vfp ---------------------------------===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// @@ -12,7 +11,7 @@ // // extern float __negsf2vfp(float a); // -// Returns the negation of a single precision floating point numbers using the +// Returns the negation of a single precision floating point numbers using the // Darwin calling convention where single arguments are passsed like 32-bit ints // .syntax unified diff --git a/lib/builtins/arm/nesf2vfp.S b/lib/builtins/arm/nesf2vfp.S index 50d60f493005..34c8bb4896c2 100644 --- a/lib/builtins/arm/nesf2vfp.S +++ b/lib/builtins/arm/nesf2vfp.S @@ -1,9 +1,8 @@ //===-- nesf2vfp.S - Implement nesf2vfp -----------------------------------===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// @@ -13,7 +12,7 @@ // extern int __nesf2vfp(float a, float b); // // Returns one iff a != b and neither is NaN. -// Uses Darwin calling convention where single precision arguments are passsed +// Uses Darwin calling convention where single precision arguments are passsed // like 32-bit ints // .syntax unified diff --git a/lib/builtins/arm/restore_vfp_d8_d15_regs.S b/lib/builtins/arm/restore_vfp_d8_d15_regs.S index 0692cf3e1b77..fd6d59bb3c38 100644 --- a/lib/builtins/arm/restore_vfp_d8_d15_regs.S +++ b/lib/builtins/arm/restore_vfp_d8_d15_regs.S @@ -1,9 +1,8 @@ //===-- save_restore_regs.S - Implement save/restore* ---------------------===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// diff --git a/lib/builtins/arm/save_vfp_d8_d15_regs.S b/lib/builtins/arm/save_vfp_d8_d15_regs.S index 544dd5467a4d..5eb3a2fbaae7 100644 --- a/lib/builtins/arm/save_vfp_d8_d15_regs.S +++ b/lib/builtins/arm/save_vfp_d8_d15_regs.S @@ -1,9 +1,8 @@ //===-- save_restore_regs.S - Implement save/restore* ---------------------===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// diff --git a/lib/builtins/arm/softfloat-alias.list b/lib/builtins/arm/softfloat-alias.list index cc6a4b3cdd2e..ab6ed21e54b0 100644 --- a/lib/builtins/arm/softfloat-alias.list +++ b/lib/builtins/arm/softfloat-alias.list @@ -1,5 +1,5 @@ # -# These are soft float functions which can be +# These are soft float functions which can be # aliased to the *vfp functions on arm processors # that support floating point instructions. # diff --git a/lib/builtins/arm/subdf3vfp.S b/lib/builtins/arm/subdf3vfp.S index 2b6f2bdbfdd5..f4eaf9af1afe 100644 --- a/lib/builtins/arm/subdf3vfp.S +++ b/lib/builtins/arm/subdf3vfp.S @@ -1,9 +1,8 @@ //===-- subdf3vfp.S - Implement subdf3vfp ---------------------------------===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// @@ -12,7 +11,7 @@ // // extern double __subdf3vfp(double a, double b); // -// Returns difference between two double precision floating point numbers using +// Returns difference between two double precision floating point numbers using // the Darwin calling convention where double arguments are passsed in GPR pairs // .syntax unified @@ -23,7 +22,7 @@ DEFINE_COMPILERRT_FUNCTION(__subdf3vfp) #else vmov d6, r0, r1 // move first param from r0/r1 pair into d6 vmov d7, r2, r3 // move second param from r2/r3 pair into d7 - vsub.f64 d6, d6, d7 + vsub.f64 d6, d6, d7 vmov r0, r1, d6 // move result back to r0/r1 pair #endif bx lr diff --git a/lib/builtins/arm/subsf3vfp.S b/lib/builtins/arm/subsf3vfp.S index 3e83ea26507d..80e69f2e8e88 100644 --- a/lib/builtins/arm/subsf3vfp.S +++ b/lib/builtins/arm/subsf3vfp.S @@ -1,9 +1,8 @@ //===-- subsf3vfp.S - Implement subsf3vfp ---------------------------------===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// diff --git a/lib/builtins/arm/switch16.S b/lib/builtins/arm/switch16.S index df9e38e176ce..a4b568da59da 100644 --- a/lib/builtins/arm/switch16.S +++ b/lib/builtins/arm/switch16.S @@ -1,9 +1,8 @@ //===-- switch.S - Implement switch* --------------------------------------===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// diff --git a/lib/builtins/arm/switch32.S b/lib/builtins/arm/switch32.S index d97b5361436d..f2a5af5275cf 100644 --- a/lib/builtins/arm/switch32.S +++ b/lib/builtins/arm/switch32.S @@ -1,9 +1,8 @@ //===-- switch.S - Implement switch* --------------------------------------===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// diff --git a/lib/builtins/arm/switch8.S b/lib/builtins/arm/switch8.S index 4d9e0eaff845..0db875c32914 100644 --- a/lib/builtins/arm/switch8.S +++ b/lib/builtins/arm/switch8.S @@ -1,9 +1,8 @@ //===-- switch.S - Implement switch* --------------------------------------===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// diff --git a/lib/builtins/arm/switchu8.S b/lib/builtins/arm/switchu8.S index 4ffe35f0549b..551abebc6083 100644 --- a/lib/builtins/arm/switchu8.S +++ b/lib/builtins/arm/switchu8.S @@ -1,9 +1,8 @@ //===-- switch.S - Implement switch* --------------------------------------===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// diff --git a/lib/builtins/arm/sync-ops.h b/lib/builtins/arm/sync-ops.h index ee02c30c6eaa..c9623249e5d2 100644 --- a/lib/builtins/arm/sync-ops.h +++ b/lib/builtins/arm/sync-ops.h @@ -1,64 +1,61 @@ -/*===-- sync-ops.h - --===// - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - *===----------------------------------------------------------------------===// - * - * This file implements outline macros for the __sync_fetch_and_* - * operations. Different instantiations will generate appropriate assembly for - * ARM and Thumb-2 versions of the functions. - * - *===----------------------------------------------------------------------===*/ +//===-- sync-ops.h - --===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements outline macros for the __sync_fetch_and_* +// operations. Different instantiations will generate appropriate assembly for +// ARM and Thumb-2 versions of the functions. +// +//===----------------------------------------------------------------------===// #include "../assembly.h" -#define SYNC_OP_4(op) \ - .p2align 2 ; \ - .thumb ; \ - .syntax unified ; \ - DEFINE_COMPILERRT_THUMB_FUNCTION(__sync_fetch_and_ ## op) \ - dmb ; \ - mov r12, r0 ; \ - LOCAL_LABEL(tryatomic_ ## op): \ - ldrex r0, [r12] ; \ - op(r2, r0, r1) ; \ - strex r3, r2, [r12] ; \ - cmp r3, #0 ; \ - bne LOCAL_LABEL(tryatomic_ ## op) ; \ - dmb ; \ - bx lr +#define SYNC_OP_4(op) \ + .p2align 2; \ + .thumb; \ + .syntax unified; \ + DEFINE_COMPILERRT_THUMB_FUNCTION(__sync_fetch_and_##op) \ + dmb; \ + mov r12, r0; \ + LOCAL_LABEL(tryatomic_##op) : ldrex r0, [r12]; \ + op(r2, r0, r1); \ + strex r3, r2, [r12]; \ + cmp r3, #0; \ + bne LOCAL_LABEL(tryatomic_##op); \ + dmb; \ + bx lr -#define SYNC_OP_8(op) \ - .p2align 2 ; \ - .thumb ; \ - .syntax unified ; \ - DEFINE_COMPILERRT_THUMB_FUNCTION(__sync_fetch_and_ ## op) \ - push {r4, r5, r6, lr} ; \ - dmb ; \ - mov r12, r0 ; \ - LOCAL_LABEL(tryatomic_ ## op): \ - ldrexd r0, r1, [r12] ; \ - op(r4, r5, r0, r1, r2, r3) ; \ - strexd r6, r4, r5, [r12] ; \ - cmp r6, #0 ; \ - bne LOCAL_LABEL(tryatomic_ ## op) ; \ - dmb ; \ - pop {r4, r5, r6, pc} +#define SYNC_OP_8(op) \ + .p2align 2; \ + .thumb; \ + .syntax unified; \ + DEFINE_COMPILERRT_THUMB_FUNCTION(__sync_fetch_and_##op) \ + push {r4, r5, r6, lr}; \ + dmb; \ + mov r12, r0; \ + LOCAL_LABEL(tryatomic_##op) : ldrexd r0, r1, [r12]; \ + op(r4, r5, r0, r1, r2, r3); \ + strexd r6, r4, r5, [r12]; \ + cmp r6, #0; \ + bne LOCAL_LABEL(tryatomic_##op); \ + dmb; \ + pop { r4, r5, r6, pc } -#define MINMAX_4(rD, rN, rM, cmp_kind) \ - cmp rN, rM ; \ - mov rD, rM ; \ - it cmp_kind ; \ - mov##cmp_kind rD, rN +#define MINMAX_4(rD, rN, rM, cmp_kind) \ + cmp rN, rM; \ + mov rD, rM; \ + it cmp_kind; \ + mov##cmp_kind rD, rN -#define MINMAX_8(rD_LO, rD_HI, rN_LO, rN_HI, rM_LO, rM_HI, cmp_kind) \ - cmp rN_LO, rM_LO ; \ - sbcs rN_HI, rM_HI ; \ - mov rD_LO, rM_LO ; \ - mov rD_HI, rM_HI ; \ - itt cmp_kind ; \ - mov##cmp_kind rD_LO, rN_LO ; \ - mov##cmp_kind rD_HI, rN_HI +#define MINMAX_8(rD_LO, rD_HI, rN_LO, rN_HI, rM_LO, rM_HI, cmp_kind) \ + cmp rN_LO, rM_LO; \ + sbcs rN_HI, rM_HI; \ + mov rD_LO, rM_LO; \ + mov rD_HI, rM_HI; \ + itt cmp_kind; \ + mov##cmp_kind rD_LO, rN_LO; \ + mov##cmp_kind rD_HI, rN_HI diff --git a/lib/builtins/arm/sync_fetch_and_add_4.S b/lib/builtins/arm/sync_fetch_and_add_4.S index 7877d6c46c11..0d55975b7a3a 100644 --- a/lib/builtins/arm/sync_fetch_and_add_4.S +++ b/lib/builtins/arm/sync_fetch_and_add_4.S @@ -1,20 +1,19 @@ -/*===-- sync_fetch_and_add_4.S - ------------------------------------------===// - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - *===----------------------------------------------------------------------===// - * - * This file implements the __sync_fetch_and_add_4 function for the ARM - * architecture. - * - *===----------------------------------------------------------------------===*/ +//===-- sync_fetch_and_add_4.S - ------------------------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements the __sync_fetch_and_add_4 function for the ARM +// architecture. +// +//===----------------------------------------------------------------------===// #include "sync-ops.h" -/* "adds" is 2 bytes shorter than "add". */ +// "adds" is 2 bytes shorter than "add". #define add_4(rD, rN, rM) add rD, rN, rM SYNC_OP_4(add_4) diff --git a/lib/builtins/arm/sync_fetch_and_add_8.S b/lib/builtins/arm/sync_fetch_and_add_8.S index 1df07a342a1b..18bdd875b8b7 100644 --- a/lib/builtins/arm/sync_fetch_and_add_8.S +++ b/lib/builtins/arm/sync_fetch_and_add_8.S @@ -1,16 +1,15 @@ -/*===-- sync_fetch_and_add_8.S - ------------------------------------------===// - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - *===----------------------------------------------------------------------===// - * - * This file implements the __sync_fetch_and_add_8 function for the ARM - * architecture. - * - *===----------------------------------------------------------------------===*/ +//===-- sync_fetch_and_add_8.S - ------------------------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements the __sync_fetch_and_add_8 function for the ARM +// architecture. +// +//===----------------------------------------------------------------------===// #include "sync-ops.h" diff --git a/lib/builtins/arm/sync_fetch_and_and_4.S b/lib/builtins/arm/sync_fetch_and_and_4.S index 720ff02279cd..3a76acca671e 100644 --- a/lib/builtins/arm/sync_fetch_and_and_4.S +++ b/lib/builtins/arm/sync_fetch_and_and_4.S @@ -1,16 +1,15 @@ -/*===-- sync_fetch_and_and_4.S - ------------------------------------------===// - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - *===----------------------------------------------------------------------===// - * - * This file implements the __sync_fetch_and_and_4 function for the ARM - * architecture. - * - *===----------------------------------------------------------------------===*/ +//===-- sync_fetch_and_and_4.S - ------------------------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements the __sync_fetch_and_and_4 function for the ARM +// architecture. +// +//===----------------------------------------------------------------------===// #include "sync-ops.h" diff --git a/lib/builtins/arm/sync_fetch_and_and_8.S b/lib/builtins/arm/sync_fetch_and_and_8.S index 4f7b5ca7ab29..3716eff809d5 100644 --- a/lib/builtins/arm/sync_fetch_and_and_8.S +++ b/lib/builtins/arm/sync_fetch_and_and_8.S @@ -1,16 +1,15 @@ -/*===-- sync_fetch_and_and_8.S - ------------------------------------------===// - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - *===----------------------------------------------------------------------===// - * - * This file implements the __sync_fetch_and_and_8 function for the ARM - * architecture. - * - *===----------------------------------------------------------------------===*/ +//===-- sync_fetch_and_and_8.S - ------------------------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements the __sync_fetch_and_and_8 function for the ARM +// architecture. +// +//===----------------------------------------------------------------------===// #include "sync-ops.h" diff --git a/lib/builtins/arm/sync_fetch_and_max_4.S b/lib/builtins/arm/sync_fetch_and_max_4.S index 43da9c7d4067..b9cee45219de 100644 --- a/lib/builtins/arm/sync_fetch_and_max_4.S +++ b/lib/builtins/arm/sync_fetch_and_max_4.S @@ -1,16 +1,15 @@ -/*===-- sync_fetch_and_max_4.S - ------------------------------------------===// - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - *===----------------------------------------------------------------------===// - * - * This file implements the __sync_fetch_and_max_4 function for the ARM - * architecture. - * - *===----------------------------------------------------------------------===*/ +//===-- sync_fetch_and_max_4.S - ------------------------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements the __sync_fetch_and_max_4 function for the ARM +// architecture. +// +//===----------------------------------------------------------------------===// #include "sync-ops.h" diff --git a/lib/builtins/arm/sync_fetch_and_max_8.S b/lib/builtins/arm/sync_fetch_and_max_8.S index 898fc6202ac8..06115ab55246 100644 --- a/lib/builtins/arm/sync_fetch_and_max_8.S +++ b/lib/builtins/arm/sync_fetch_and_max_8.S @@ -1,16 +1,15 @@ -/*===-- sync_fetch_and_max_8.S - ------------------------------------------===// - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - *===----------------------------------------------------------------------===// - * - * This file implements the __sync_fetch_and_max_8 function for the ARM - * architecture. - * - *===----------------------------------------------------------------------===*/ +//===-- sync_fetch_and_max_8.S - ------------------------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements the __sync_fetch_and_max_8 function for the ARM +// architecture. +// +//===----------------------------------------------------------------------===// #include "sync-ops.h" diff --git a/lib/builtins/arm/sync_fetch_and_min_4.S b/lib/builtins/arm/sync_fetch_and_min_4.S index bba31a03aace..60d435a0bd3f 100644 --- a/lib/builtins/arm/sync_fetch_and_min_4.S +++ b/lib/builtins/arm/sync_fetch_and_min_4.S @@ -1,16 +1,15 @@ -/*===-- sync_fetch_and_min_4.S - ------------------------------------------===// - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - *===----------------------------------------------------------------------===// - * - * This file implements the __sync_fetch_and_min_4 function for the ARM - * architecture. - * - *===----------------------------------------------------------------------===*/ +//===-- sync_fetch_and_min_4.S - ------------------------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements the __sync_fetch_and_min_4 function for the ARM +// architecture. +// +//===----------------------------------------------------------------------===// #include "sync-ops.h" diff --git a/lib/builtins/arm/sync_fetch_and_min_8.S b/lib/builtins/arm/sync_fetch_and_min_8.S index e7ccf9fb60ef..4f3e299d95cc 100644 --- a/lib/builtins/arm/sync_fetch_and_min_8.S +++ b/lib/builtins/arm/sync_fetch_and_min_8.S @@ -1,16 +1,15 @@ -/*===-- sync_fetch_and_min_8.S - ------------------------------------------===// - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - *===----------------------------------------------------------------------===// - * - * This file implements the __sync_fetch_and_min_8 function for the ARM - * architecture. - * - *===----------------------------------------------------------------------===*/ +//===-- sync_fetch_and_min_8.S - ------------------------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements the __sync_fetch_and_min_8 function for the ARM +// architecture. +// +//===----------------------------------------------------------------------===// #include "sync-ops.h" diff --git a/lib/builtins/arm/sync_fetch_and_nand_4.S b/lib/builtins/arm/sync_fetch_and_nand_4.S index c13dd394588f..5a04be0f6455 100644 --- a/lib/builtins/arm/sync_fetch_and_nand_4.S +++ b/lib/builtins/arm/sync_fetch_and_nand_4.S @@ -1,16 +1,15 @@ -/*===-- sync_fetch_and_nand_4.S - -----------------------------------------===// - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - *===----------------------------------------------------------------------===// - * - * This file implements the __sync_fetch_and_nand_4 function for the ARM - * architecture. - * - *===----------------------------------------------------------------------===*/ +//===-- sync_fetch_and_nand_4.S - -----------------------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements the __sync_fetch_and_nand_4 function for the ARM +// architecture. +// +//===----------------------------------------------------------------------===// #include "sync-ops.h" diff --git a/lib/builtins/arm/sync_fetch_and_nand_8.S b/lib/builtins/arm/sync_fetch_and_nand_8.S index e8107ab3a33c..425c94474af7 100644 --- a/lib/builtins/arm/sync_fetch_and_nand_8.S +++ b/lib/builtins/arm/sync_fetch_and_nand_8.S @@ -1,16 +1,15 @@ -/*===-- sync_fetch_and_nand_8.S - ------------------------------------------===// - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - *===----------------------------------------------------------------------===// - * - * This file implements the __sync_fetch_and_nand_8 function for the ARM - * architecture. - * - *===----------------------------------------------------------------------===*/ +//===-- sync_fetch_and_nand_8.S - ------------------------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements the __sync_fetch_and_nand_8 function for the ARM +// architecture. +// +//===----------------------------------------------------------------------===// #include "sync-ops.h" diff --git a/lib/builtins/arm/sync_fetch_and_or_4.S b/lib/builtins/arm/sync_fetch_and_or_4.S index 6726571a944f..f44751b9a9f2 100644 --- a/lib/builtins/arm/sync_fetch_and_or_4.S +++ b/lib/builtins/arm/sync_fetch_and_or_4.S @@ -1,16 +1,15 @@ -/*===-- sync_fetch_and_or_4.S - -------------------------------------------===// - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - *===----------------------------------------------------------------------===// - * - * This file implements the __sync_fetch_and_or_4 function for the ARM - * architecture. - * - *===----------------------------------------------------------------------===*/ +//===-- sync_fetch_and_or_4.S - -------------------------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements the __sync_fetch_and_or_4 function for the ARM +// architecture. +// +//===----------------------------------------------------------------------===// #include "sync-ops.h" diff --git a/lib/builtins/arm/sync_fetch_and_or_8.S b/lib/builtins/arm/sync_fetch_and_or_8.S index f7f162c7c3b3..4f18dcf84df9 100644 --- a/lib/builtins/arm/sync_fetch_and_or_8.S +++ b/lib/builtins/arm/sync_fetch_and_or_8.S @@ -1,16 +1,15 @@ -/*===-- sync_fetch_and_or_8.S - -------------------------------------------===// - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - *===----------------------------------------------------------------------===// - * - * This file implements the __sync_fetch_and_or_8 function for the ARM - * architecture. - * - *===----------------------------------------------------------------------===*/ +//===-- sync_fetch_and_or_8.S - -------------------------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements the __sync_fetch_and_or_8 function for the ARM +// architecture. +// +//===----------------------------------------------------------------------===// #include "sync-ops.h" diff --git a/lib/builtins/arm/sync_fetch_and_sub_4.S b/lib/builtins/arm/sync_fetch_and_sub_4.S index b9326b14cdd5..999d48c28aec 100644 --- a/lib/builtins/arm/sync_fetch_and_sub_4.S +++ b/lib/builtins/arm/sync_fetch_and_sub_4.S @@ -1,20 +1,19 @@ -/*===-- sync_fetch_and_sub_4.S - ------------------------------------------===// - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - *===----------------------------------------------------------------------===// - * - * This file implements the __sync_fetch_and_sub_4 function for the ARM - * architecture. - * - *===----------------------------------------------------------------------===*/ +//===-- sync_fetch_and_sub_4.S - ------------------------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements the __sync_fetch_and_sub_4 function for the ARM +// architecture. +// +//===----------------------------------------------------------------------===// #include "sync-ops.h" -/* "subs" is 2 bytes shorter than "sub". */ +// "subs" is 2 bytes shorter than "sub". #define sub_4(rD, rN, rM) sub rD, rN, rM SYNC_OP_4(sub_4) diff --git a/lib/builtins/arm/sync_fetch_and_sub_8.S b/lib/builtins/arm/sync_fetch_and_sub_8.S index 6ce743e5ee9f..25a4a1076555 100644 --- a/lib/builtins/arm/sync_fetch_and_sub_8.S +++ b/lib/builtins/arm/sync_fetch_and_sub_8.S @@ -1,16 +1,15 @@ -/*===-- sync_fetch_and_sub_8.S - ------------------------------------------===// - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - *===----------------------------------------------------------------------===// - * - * This file implements the __sync_fetch_and_sub_8 function for the ARM - * architecture. - * - *===----------------------------------------------------------------------===*/ +//===-- sync_fetch_and_sub_8.S - ------------------------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements the __sync_fetch_and_sub_8 function for the ARM +// architecture. +// +//===----------------------------------------------------------------------===// #include "sync-ops.h" diff --git a/lib/builtins/arm/sync_fetch_and_umax_4.S b/lib/builtins/arm/sync_fetch_and_umax_4.S index b8d19ff35057..a7b233b15305 100644 --- a/lib/builtins/arm/sync_fetch_and_umax_4.S +++ b/lib/builtins/arm/sync_fetch_and_umax_4.S @@ -1,16 +1,15 @@ -/*===-- sync_fetch_and_umax_4.S - ------------------------------------------===// - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - *===----------------------------------------------------------------------===// - * - * This file implements the __sync_fetch_and_umax_4 function for the ARM - * architecture. - * - *===----------------------------------------------------------------------===*/ +//===-- sync_fetch_and_umax_4.S - ------------------------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements the __sync_fetch_and_umax_4 function for the ARM +// architecture. +// +//===----------------------------------------------------------------------===// #include "sync-ops.h" diff --git a/lib/builtins/arm/sync_fetch_and_umax_8.S b/lib/builtins/arm/sync_fetch_and_umax_8.S index 34442fd77454..aa5213ff1def 100644 --- a/lib/builtins/arm/sync_fetch_and_umax_8.S +++ b/lib/builtins/arm/sync_fetch_and_umax_8.S @@ -1,16 +1,15 @@ -/*===-- sync_fetch_and_umax_8.S - ------------------------------------------===// - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - *===----------------------------------------------------------------------===// - * - * This file implements the __sync_fetch_and_umax_8 function for the ARM - * architecture. - * - *===----------------------------------------------------------------------===*/ +//===-- sync_fetch_and_umax_8.S - ------------------------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements the __sync_fetch_and_umax_8 function for the ARM +// architecture. +// +//===----------------------------------------------------------------------===// #include "sync-ops.h" diff --git a/lib/builtins/arm/sync_fetch_and_umin_4.S b/lib/builtins/arm/sync_fetch_and_umin_4.S index 0998e3e10f58..c7a9c89df60a 100644 --- a/lib/builtins/arm/sync_fetch_and_umin_4.S +++ b/lib/builtins/arm/sync_fetch_and_umin_4.S @@ -1,16 +1,15 @@ -/*===-- sync_fetch_and_umin_4.S - ------------------------------------------===// - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - *===----------------------------------------------------------------------===// - * - * This file implements the __sync_fetch_and_umin_4 function for the ARM - * architecture. - * - *===----------------------------------------------------------------------===*/ +//===-- sync_fetch_and_umin_4.S - ------------------------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements the __sync_fetch_and_umin_4 function for the ARM +// architecture. +// +//===----------------------------------------------------------------------===// #include "sync-ops.h" diff --git a/lib/builtins/arm/sync_fetch_and_umin_8.S b/lib/builtins/arm/sync_fetch_and_umin_8.S index 558f91390512..8b40541ab47d 100644 --- a/lib/builtins/arm/sync_fetch_and_umin_8.S +++ b/lib/builtins/arm/sync_fetch_and_umin_8.S @@ -1,16 +1,15 @@ -/*===-- sync_fetch_and_umin_8.S - ------------------------------------------===// - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - *===----------------------------------------------------------------------===// - * - * This file implements the __sync_fetch_and_umin_8 function for the ARM - * architecture. - * - *===----------------------------------------------------------------------===*/ +//===-- sync_fetch_and_umin_8.S - ------------------------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements the __sync_fetch_and_umin_8 function for the ARM +// architecture. +// +//===----------------------------------------------------------------------===// #include "sync-ops.h" diff --git a/lib/builtins/arm/sync_fetch_and_xor_4.S b/lib/builtins/arm/sync_fetch_and_xor_4.S index 824f49146880..f509191d0175 100644 --- a/lib/builtins/arm/sync_fetch_and_xor_4.S +++ b/lib/builtins/arm/sync_fetch_and_xor_4.S @@ -1,16 +1,15 @@ -/*===-- sync_fetch_and_xor_4.S - ------------------------------------------===// - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - *===----------------------------------------------------------------------===// - * - * This file implements the __sync_fetch_and_xor_4 function for the ARM - * architecture. - * - *===----------------------------------------------------------------------===*/ +//===-- sync_fetch_and_xor_4.S - ------------------------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements the __sync_fetch_and_xor_4 function for the ARM +// architecture. +// +//===----------------------------------------------------------------------===// #include "sync-ops.h" diff --git a/lib/builtins/arm/sync_fetch_and_xor_8.S b/lib/builtins/arm/sync_fetch_and_xor_8.S index 073fb9c20f25..7436eb1d4cae 100644 --- a/lib/builtins/arm/sync_fetch_and_xor_8.S +++ b/lib/builtins/arm/sync_fetch_and_xor_8.S @@ -1,16 +1,15 @@ -/*===-- sync_fetch_and_xor_8.S - ------------------------------------------===// - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - *===----------------------------------------------------------------------===// - * - * This file implements the __sync_fetch_and_xor_8 function for the ARM - * architecture. - * - *===----------------------------------------------------------------------===*/ +//===-- sync_fetch_and_xor_8.S - ------------------------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements the __sync_fetch_and_xor_8 function for the ARM +// architecture. +// +//===----------------------------------------------------------------------===// #include "sync-ops.h" diff --git a/lib/builtins/arm/sync_synchronize.S b/lib/builtins/arm/sync_synchronize.S index 61d1db910f0d..dd06e7191f9b 100644 --- a/lib/builtins/arm/sync_synchronize.S +++ b/lib/builtins/arm/sync_synchronize.S @@ -1,20 +1,17 @@ //===-- sync_synchronize - Implement memory barrier * ----------------------===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// #include "../assembly.h" -// // When compiling a use of the gcc built-in __sync_synchronize() in thumb1 mode -// the compiler may emit a call to __sync_synchronize. -// On Darwin the implementation jumps to an OS supplied function named +// the compiler may emit a call to __sync_synchronize. +// On Darwin the implementation jumps to an OS supplied function named // OSMemoryBarrier -// .text .syntax unified @@ -31,7 +28,7 @@ END_COMPILERRT_FUNCTION(__sync_synchronize) // tell linker it can break up file at label boundaries .subsections_via_symbols - + #endif NO_EXEC_STACK_DIRECTIVE diff --git a/lib/builtins/arm/truncdfsf2vfp.S b/lib/builtins/arm/truncdfsf2vfp.S index 682e54d3d294..a3c0a73466e9 100644 --- a/lib/builtins/arm/truncdfsf2vfp.S +++ b/lib/builtins/arm/truncdfsf2vfp.S @@ -1,9 +1,8 @@ //===-- truncdfsf2vfp.S - Implement truncdfsf2vfp -------------------------===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// @@ -13,7 +12,7 @@ // extern float __truncdfsf2vfp(double a); // // Converts double precision float to signle precision result. -// Uses Darwin calling convention where a double precision parameter is +// Uses Darwin calling convention where a double precision parameter is // passed in a R0/R1 pair and a signle precision result is returned in R0. // .syntax unified diff --git a/lib/builtins/arm/udivmodsi4.S b/lib/builtins/arm/udivmodsi4.S index ee3950c9b0eb..0f40575fe99e 100644 --- a/lib/builtins/arm/udivmodsi4.S +++ b/lib/builtins/arm/udivmodsi4.S @@ -1,16 +1,15 @@ -/*===-- udivmodsi4.S - 32-bit unsigned integer divide and modulus ---------===// - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - *===----------------------------------------------------------------------===// - * - * This file implements the __udivmodsi4 (32-bit unsigned integer divide and - * modulus) function for the ARM 32-bit architecture. - * - *===----------------------------------------------------------------------===*/ +//===-- udivmodsi4.S - 32-bit unsigned integer divide and modulus ---------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements the __udivmodsi4 (32-bit unsigned integer divide and +// modulus) function for the ARM 32-bit architecture. +// +//===----------------------------------------------------------------------===// #include "../assembly.h" @@ -39,26 +38,25 @@ DEFINE_COMPILERRT_FUNCTION(__udivmodsi4) beq LOCAL_LABEL(divby1) cmp r0, r1 bcc LOCAL_LABEL(quotient0) - /* - * Implement division using binary long division algorithm. - * - * r0 is the numerator, r1 the denominator. - * - * The code before JMP computes the correct shift I, so that - * r0 and (r1 << I) have the highest bit set in the same position. - * At the time of JMP, ip := .Ldiv0block - 12 * I. - * This depends on the fixed instruction size of block. - * For ARM mode, this is 12 Bytes, for THUMB mode 14 Bytes. - * - * block(shift) implements the test-and-update-quotient core. - * It assumes (r0 << shift) can be computed without overflow and - * that (r0 << shift) < 2 * r1. The quotient is stored in r3. - */ + + // Implement division using binary long division algorithm. + // + // r0 is the numerator, r1 the denominator. + // + // The code before JMP computes the correct shift I, so that + // r0 and (r1 << I) have the highest bit set in the same position. + // At the time of JMP, ip := .Ldiv0block - 12 * I. + // This depends on the fixed instruction size of block. + // For ARM mode, this is 12 Bytes, for THUMB mode 14 Bytes. + // + // block(shift) implements the test-and-update-quotient core. + // It assumes (r0 << shift) can be computed without overflow and + // that (r0 << shift) < 2 * r1. The quotient is stored in r3. # ifdef __ARM_FEATURE_CLZ clz ip, r0 clz r3, r1 - /* r0 >= r1 implies clz(r0) <= clz(r1), so ip <= r3. */ + // r0 >= r1 implies clz(r0) <= clz(r1), so ip <= r3. sub r3, r3, ip # if defined(USE_THUMB_2) adr ip, LOCAL_LABEL(div0block) + 1 @@ -99,11 +97,11 @@ DEFINE_COMPILERRT_FUNCTION(__udivmodsi4) movhs r4, r3 subhs ip, ip, #(2 * 12) - /* Last block, no need to update r3 or r4. */ + // Last block, no need to update r3 or r4. cmp r1, r4, lsr #1 subls ip, ip, #(1 * 12) - ldr r4, [sp], #8 /* restore r4, we are done with it. */ + ldr r4, [sp], #8 // restore r4, we are done with it. mov r3, #0 JMP(ip) @@ -164,7 +162,7 @@ LOCAL_LABEL(divby1): mov r3, #0 str r3, [r2] JMP(lr) -#endif /* __ARM_ARCH_EXT_IDIV__ */ +#endif // __ARM_ARCH_EXT_IDIV__ LOCAL_LABEL(divby0): mov r0, #0 diff --git a/lib/builtins/arm/udivsi3.S b/lib/builtins/arm/udivsi3.S index 6dea27d404ff..9b1b035b33d6 100644 --- a/lib/builtins/arm/udivsi3.S +++ b/lib/builtins/arm/udivsi3.S @@ -1,16 +1,15 @@ -/*===-- udivsi3.S - 32-bit unsigned integer divide ------------------------===// - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - *===----------------------------------------------------------------------===// - * - * This file implements the __udivsi3 (32-bit unsigned integer divide) - * function for the ARM 32-bit architecture. - * - *===----------------------------------------------------------------------===*/ +//===-- udivsi3.S - 32-bit unsigned integer divide ------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements the __udivsi3 (32-bit unsigned integer divide) +// function for the ARM 32-bit architecture. +// +//===----------------------------------------------------------------------===// #include "../assembly.h" @@ -40,7 +39,7 @@ LOCAL_LABEL(divby0): JMP(lr) # endif -#else /* ! __ARM_ARCH_EXT_IDIV__ */ +#else // ! __ARM_ARCH_EXT_IDIV__ cmp r1, #1 bcc LOCAL_LABEL(divby0) #if defined(USE_THUMB_1) @@ -63,26 +62,24 @@ LOCAL_LABEL(num_ge_denom): JMPc(lr, cc) #endif - /* - * Implement division using binary long division algorithm. - * - * r0 is the numerator, r1 the denominator. - * - * The code before JMP computes the correct shift I, so that - * r0 and (r1 << I) have the highest bit set in the same position. - * At the time of JMP, ip := .Ldiv0block - 12 * I. - * This depends on the fixed instruction size of block. - * For ARM mode, this is 12 Bytes, for THUMB mode 14 Bytes. - * - * block(shift) implements the test-and-update-quotient core. - * It assumes (r0 << shift) can be computed without overflow and - * that (r0 << shift) < 2 * r1. The quotient is stored in r3. - */ + // Implement division using binary long division algorithm. + // + // r0 is the numerator, r1 the denominator. + // + // The code before JMP computes the correct shift I, so that + // r0 and (r1 << I) have the highest bit set in the same position. + // At the time of JMP, ip := .Ldiv0block - 12 * I. + // This depends on the fixed instruction size of block. + // For ARM mode, this is 12 Bytes, for THUMB mode 14 Bytes. + // + // block(shift) implements the test-and-update-quotient core. + // It assumes (r0 << shift) can be computed without overflow and + // that (r0 << shift) < 2 * r1. The quotient is stored in r3. # if defined(__ARM_FEATURE_CLZ) clz ip, r0 clz r3, r1 - /* r0 >= r1 implies clz(r0) <= clz(r1), so ip <= r3. */ + // r0 >= r1 implies clz(r0) <= clz(r1), so ip <= r3. sub r3, r3, ip # if defined(USE_THUMB_2) adr ip, LOCAL_LABEL(div0block) + 1 @@ -94,7 +91,7 @@ LOCAL_LABEL(num_ge_denom): sub ip, ip, r3, lsl #3 mov r3, #0 bx ip -# else /* No CLZ Feature */ +# else // No CLZ Feature # if defined(USE_THUMB_2) # error THUMB mode requires CLZ or UDIV # endif @@ -160,7 +157,7 @@ LOCAL_LABEL(skip_2): subhs ip, ip, #(2 * BLOCK_SIZE) # endif - /* Last block, no need to update r2 or r3. */ + // Last block, no need to update r2 or r3. # if defined(USE_THUMB_1) lsrs r3, r2, #1 cmp r3, r1 @@ -180,12 +177,12 @@ LOCAL_LABEL(skip_1): JMP(ip) # endif -# endif /* __ARM_FEATURE_CLZ */ +# endif // __ARM_FEATURE_CLZ #define IMM # - /* due to the range limit of branch in Thumb1, we have to place the - block closer */ + // due to the range limit of branch in Thumb1, we have to place the + // block closer LOCAL_LABEL(divby0): movs r0, #0 # if defined(__ARM_EABI__) @@ -204,13 +201,13 @@ LOCAL_LABEL(divby0): blo LOCAL_LABEL(block_skip_##shift); \ subs r0, r0, r2; \ LOCAL_LABEL(block_skip_##shift) :; \ - adcs r3, r3 /* same as ((r3 << 1) | Carry). Carry is set if r0 >= r2. */ + adcs r3, r3 // same as ((r3 << 1) | Carry). Carry is set if r0 >= r2. - /* TODO: if current location counter is not not word aligned, we don't - need the .p2align and nop */ - /* Label div0block must be word-aligned. First align block 31 */ + // TODO: if current location counter is not not word aligned, we don't + // need the .p2align and nop + // Label div0block must be word-aligned. First align block 31 .p2align 2 - nop /* Padding to align div0block as 31 blocks = 310 bytes */ + nop // Padding to align div0block as 31 blocks = 310 bytes #else #define block(shift) \ @@ -256,7 +253,7 @@ LOCAL_LABEL(div0block): mov r0, r3 JMP(lr) -#endif /* __ARM_ARCH_EXT_IDIV__ */ +#endif // __ARM_ARCH_EXT_IDIV__ END_COMPILERRT_FUNCTION(__udivsi3) diff --git a/lib/builtins/arm/umodsi3.S b/lib/builtins/arm/umodsi3.S index 069fad34cb9c..5ab78de17853 100644 --- a/lib/builtins/arm/umodsi3.S +++ b/lib/builtins/arm/umodsi3.S @@ -1,16 +1,15 @@ -/*===-- umodsi3.S - 32-bit unsigned integer modulus -----------------------===// - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - *===----------------------------------------------------------------------===// - * - * This file implements the __umodsi3 (32-bit unsigned integer modulus) - * function for the ARM 32-bit architecture. - * - *===----------------------------------------------------------------------===*/ +//===-- umodsi3.S - 32-bit unsigned integer modulus -----------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements the __umodsi3 (32-bit unsigned integer modulus) +// function for the ARM 32-bit architecture. +// +//===----------------------------------------------------------------------===// #include "../assembly.h" @@ -38,26 +37,25 @@ DEFINE_COMPILERRT_FUNCTION(__umodsi3) cmp r0, r1 IT(cc) JMPc(lr, cc) - /* - * Implement division using binary long division algorithm. - * - * r0 is the numerator, r1 the denominator. - * - * The code before JMP computes the correct shift I, so that - * r0 and (r1 << I) have the highest bit set in the same position. - * At the time of JMP, ip := .Ldiv0block - 8 * I. - * This depends on the fixed instruction size of block. - * For ARM mode, this is 8 Bytes, for THUMB mode 10 Bytes. - * - * block(shift) implements the test-and-update-quotient core. - * It assumes (r0 << shift) can be computed without overflow and - * that (r0 << shift) < 2 * r1. The quotient is stored in r3. - */ + + // Implement division using binary long division algorithm. + // + // r0 is the numerator, r1 the denominator. + // + // The code before JMP computes the correct shift I, so that + // r0 and (r1 << I) have the highest bit set in the same position. + // At the time of JMP, ip := .Ldiv0block - 8 * I. + // This depends on the fixed instruction size of block. + // For ARM mode, this is 8 Bytes, for THUMB mode 10 Bytes. + // + // block(shift) implements the test-and-update-quotient core. + // It assumes (r0 << shift) can be computed without overflow and + // that (r0 << shift) < 2 * r1. The quotient is stored in r3. # ifdef __ARM_FEATURE_CLZ clz ip, r0 clz r3, r1 - /* r0 >= r1 implies clz(r0) <= clz(r1), so ip <= r3. */ + // r0 >= r1 implies clz(r0) <= clz(r1), so ip <= r3. sub r3, r3, ip # if defined(USE_THUMB_2) adr ip, LOCAL_LABEL(div0block) + 1 @@ -94,7 +92,7 @@ DEFINE_COMPILERRT_FUNCTION(__umodsi3) movhs r2, r3 subhs ip, ip, #(2 * 8) - /* Last block, no need to update r2 or r3. */ + // Last block, no need to update r2 or r3. cmp r1, r2, lsr #1 subls ip, ip, #(1 * 8) @@ -142,7 +140,7 @@ DEFINE_COMPILERRT_FUNCTION(__umodsi3) LOCAL_LABEL(div0block): block(0) JMP(lr) -#endif /* __ARM_ARCH_EXT_IDIV__ */ +#endif // __ARM_ARCH_EXT_IDIV__ LOCAL_LABEL(divby0): mov r0, #0 diff --git a/lib/builtins/arm/unorddf2vfp.S b/lib/builtins/arm/unorddf2vfp.S index 6625fa8a3119..ea36a1cb5594 100644 --- a/lib/builtins/arm/unorddf2vfp.S +++ b/lib/builtins/arm/unorddf2vfp.S @@ -1,9 +1,8 @@ //===-- unorddf2vfp.S - Implement unorddf2vfp ------------------------------===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// @@ -13,7 +12,7 @@ // extern int __unorddf2vfp(double a, double b); // // Returns one iff a or b is NaN -// Uses Darwin calling convention where double precision arguments are passsed +// Uses Darwin calling convention where double precision arguments are passsed // like in GPR pairs. // .syntax unified diff --git a/lib/builtins/arm/unordsf2vfp.S b/lib/builtins/arm/unordsf2vfp.S index 0b5da2ba3e17..7311297329dc 100644 --- a/lib/builtins/arm/unordsf2vfp.S +++ b/lib/builtins/arm/unordsf2vfp.S @@ -1,9 +1,8 @@ //===-- unordsf2vfp.S - Implement unordsf2vfp -----------------------------===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// @@ -13,7 +12,7 @@ // extern int __unordsf2vfp(float a, float b); // // Returns one iff a or b is NaN -// Uses Darwin calling convention where single precision arguments are passsed +// Uses Darwin calling convention where single precision arguments are passsed // like 32-bit ints // .syntax unified diff --git a/lib/builtins/ashldi3.c b/lib/builtins/ashldi3.c index a5c1836006b9..7c81057a2284 100644 --- a/lib/builtins/ashldi3.c +++ b/lib/builtins/ashldi3.c @@ -1,45 +1,38 @@ -/* ====-- ashldi3.c - Implement __ashldi3 -----------------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - * - * This file implements __ashldi3 for the compiler_rt library. - * - * ===----------------------------------------------------------------------=== - */ +// ====-- ashldi3.c - Implement __ashldi3 ---------------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements __ashldi3 for the compiler_rt library. +// +//===----------------------------------------------------------------------===// #include "int_lib.h" -/* Returns: a << b */ +// Returns: a << b -/* Precondition: 0 <= b < bits_in_dword */ +// Precondition: 0 <= b < bits_in_dword -COMPILER_RT_ABI di_int -__ashldi3(di_int a, si_int b) -{ - const int bits_in_word = (int)(sizeof(si_int) * CHAR_BIT); - dwords input; - dwords result; - input.all = a; - if (b & bits_in_word) /* bits_in_word <= b < bits_in_dword */ - { - result.s.low = 0; - result.s.high = input.s.low << (b - bits_in_word); - } - else /* 0 <= b < bits_in_word */ - { - if (b == 0) - return a; - result.s.low = input.s.low << b; - result.s.high = (input.s.high << b) | (input.s.low >> (bits_in_word - b)); - } - return result.all; +COMPILER_RT_ABI di_int __ashldi3(di_int a, si_int b) { + const int bits_in_word = (int)(sizeof(si_int) * CHAR_BIT); + dwords input; + dwords result; + input.all = a; + if (b & bits_in_word) /* bits_in_word <= b < bits_in_dword */ { + result.s.low = 0; + result.s.high = input.s.low << (b - bits_in_word); + } else /* 0 <= b < bits_in_word */ { + if (b == 0) + return a; + result.s.low = input.s.low << b; + result.s.high = (input.s.high << b) | (input.s.low >> (bits_in_word - b)); + } + return result.all; } #if defined(__ARM_EABI__) -AEABI_RTABI di_int __aeabi_llsl(di_int a, si_int b) COMPILER_RT_ALIAS(__ashldi3); +COMPILER_RT_ALIAS(__ashldi3, __aeabi_llsl) #endif diff --git a/lib/builtins/ashlti3.c b/lib/builtins/ashlti3.c index 638ae845ff04..2d7bd4a89380 100644 --- a/lib/builtins/ashlti3.c +++ b/lib/builtins/ashlti3.c @@ -1,45 +1,38 @@ -/* ===-- ashlti3.c - Implement __ashlti3 -----------------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - * - * This file implements __ashlti3 for the compiler_rt library. - * - * ===----------------------------------------------------------------------=== - */ +//===-- ashlti3.c - Implement __ashlti3 -----------------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements __ashlti3 for the compiler_rt library. +// +//===----------------------------------------------------------------------===// #include "int_lib.h" #ifdef CRT_HAS_128BIT -/* Returns: a << b */ +// Returns: a << b -/* Precondition: 0 <= b < bits_in_tword */ +// Precondition: 0 <= b < bits_in_tword -COMPILER_RT_ABI ti_int -__ashlti3(ti_int a, si_int b) -{ - const int bits_in_dword = (int)(sizeof(di_int) * CHAR_BIT); - twords input; - twords result; - input.all = a; - if (b & bits_in_dword) /* bits_in_dword <= b < bits_in_tword */ - { - result.s.low = 0; - result.s.high = input.s.low << (b - bits_in_dword); - } - else /* 0 <= b < bits_in_dword */ - { - if (b == 0) - return a; - result.s.low = input.s.low << b; - result.s.high = (input.s.high << b) | (input.s.low >> (bits_in_dword - b)); - } - return result.all; +COMPILER_RT_ABI ti_int __ashlti3(ti_int a, si_int b) { + const int bits_in_dword = (int)(sizeof(di_int) * CHAR_BIT); + twords input; + twords result; + input.all = a; + if (b & bits_in_dword) /* bits_in_dword <= b < bits_in_tword */ { + result.s.low = 0; + result.s.high = input.s.low << (b - bits_in_dword); + } else /* 0 <= b < bits_in_dword */ { + if (b == 0) + return a; + result.s.low = input.s.low << b; + result.s.high = (input.s.high << b) | (input.s.low >> (bits_in_dword - b)); + } + return result.all; } -#endif /* CRT_HAS_128BIT */ +#endif // CRT_HAS_128BIT diff --git a/lib/builtins/ashrdi3.c b/lib/builtins/ashrdi3.c index 84619965eca0..b9939132205c 100644 --- a/lib/builtins/ashrdi3.c +++ b/lib/builtins/ashrdi3.c @@ -1,46 +1,39 @@ -/*===-- ashrdi3.c - Implement __ashrdi3 -----------------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - * - * This file implements __ashrdi3 for the compiler_rt library. - * - * ===----------------------------------------------------------------------=== - */ +//===-- ashrdi3.c - Implement __ashrdi3 -----------------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements __ashrdi3 for the compiler_rt library. +// +//===----------------------------------------------------------------------===// #include "int_lib.h" -/* Returns: arithmetic a >> b */ +// Returns: arithmetic a >> b -/* Precondition: 0 <= b < bits_in_dword */ +// Precondition: 0 <= b < bits_in_dword -COMPILER_RT_ABI di_int -__ashrdi3(di_int a, si_int b) -{ - const int bits_in_word = (int)(sizeof(si_int) * CHAR_BIT); - dwords input; - dwords result; - input.all = a; - if (b & bits_in_word) /* bits_in_word <= b < bits_in_dword */ - { - /* result.s.high = input.s.high < 0 ? -1 : 0 */ - result.s.high = input.s.high >> (bits_in_word - 1); - result.s.low = input.s.high >> (b - bits_in_word); - } - else /* 0 <= b < bits_in_word */ - { - if (b == 0) - return a; - result.s.high = input.s.high >> b; - result.s.low = (input.s.high << (bits_in_word - b)) | (input.s.low >> b); - } - return result.all; +COMPILER_RT_ABI di_int __ashrdi3(di_int a, si_int b) { + const int bits_in_word = (int)(sizeof(si_int) * CHAR_BIT); + dwords input; + dwords result; + input.all = a; + if (b & bits_in_word) /* bits_in_word <= b < bits_in_dword */ { + // result.s.high = input.s.high < 0 ? -1 : 0 + result.s.high = input.s.high >> (bits_in_word - 1); + result.s.low = input.s.high >> (b - bits_in_word); + } else /* 0 <= b < bits_in_word */ { + if (b == 0) + return a; + result.s.high = input.s.high >> b; + result.s.low = (input.s.high << (bits_in_word - b)) | (input.s.low >> b); + } + return result.all; } #if defined(__ARM_EABI__) -AEABI_RTABI di_int __aeabi_lasr(di_int a, si_int b) COMPILER_RT_ALIAS(__ashrdi3); +COMPILER_RT_ALIAS(__ashrdi3, __aeabi_lasr) #endif diff --git a/lib/builtins/ashrti3.c b/lib/builtins/ashrti3.c index f78205d961e4..f573b6d6ccba 100644 --- a/lib/builtins/ashrti3.c +++ b/lib/builtins/ashrti3.c @@ -1,46 +1,39 @@ -/* ===-- ashrti3.c - Implement __ashrti3 -----------------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - * - * This file implements __ashrti3 for the compiler_rt library. - * - * ===----------------------------------------------------------------------=== - */ +//===-- ashrti3.c - Implement __ashrti3 -----------------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements __ashrti3 for the compiler_rt library. +// +//===----------------------------------------------------------------------===// #include "int_lib.h" #ifdef CRT_HAS_128BIT -/* Returns: arithmetic a >> b */ +// Returns: arithmetic a >> b -/* Precondition: 0 <= b < bits_in_tword */ +// Precondition: 0 <= b < bits_in_tword -COMPILER_RT_ABI ti_int -__ashrti3(ti_int a, si_int b) -{ - const int bits_in_dword = (int)(sizeof(di_int) * CHAR_BIT); - twords input; - twords result; - input.all = a; - if (b & bits_in_dword) /* bits_in_dword <= b < bits_in_tword */ - { - /* result.s.high = input.s.high < 0 ? -1 : 0 */ - result.s.high = input.s.high >> (bits_in_dword - 1); - result.s.low = input.s.high >> (b - bits_in_dword); - } - else /* 0 <= b < bits_in_dword */ - { - if (b == 0) - return a; - result.s.high = input.s.high >> b; - result.s.low = (input.s.high << (bits_in_dword - b)) | (input.s.low >> b); - } - return result.all; +COMPILER_RT_ABI ti_int __ashrti3(ti_int a, si_int b) { + const int bits_in_dword = (int)(sizeof(di_int) * CHAR_BIT); + twords input; + twords result; + input.all = a; + if (b & bits_in_dword) /* bits_in_dword <= b < bits_in_tword */ { + // result.s.high = input.s.high < 0 ? -1 : 0 + result.s.high = input.s.high >> (bits_in_dword - 1); + result.s.low = input.s.high >> (b - bits_in_dword); + } else /* 0 <= b < bits_in_dword */ { + if (b == 0) + return a; + result.s.high = input.s.high >> b; + result.s.low = (input.s.high << (bits_in_dword - b)) | (input.s.low >> b); + } + return result.all; } -#endif /* CRT_HAS_128BIT */ +#endif // CRT_HAS_128BIT diff --git a/lib/builtins/assembly.h b/lib/builtins/assembly.h index 3f5e59b25442..f437cb87f60a 100644 --- a/lib/builtins/assembly.h +++ b/lib/builtins/assembly.h @@ -1,17 +1,15 @@ -/* ===-- assembly.h - compiler-rt assembler support macros -----------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - * - * This file defines macros for use in compiler-rt assembler source. - * This file is not part of the interface of this library. - * - * ===----------------------------------------------------------------------=== - */ +//===-- assembly.h - compiler-rt assembler support macros -----------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file defines macros for use in compiler-rt assembler source. +// This file is not part of the interface of this library. +// +//===----------------------------------------------------------------------===// #ifndef COMPILERRT_ASSEMBLY_H #define COMPILERRT_ASSEMBLY_H @@ -69,11 +67,9 @@ #if defined(__arm__) -/* - * Determine actual [ARM][THUMB[1][2]] ISA using compiler predefined macros: - * - for '-mthumb -march=armv6' compiler defines '__thumb__' - * - for '-mthumb -march=armv7' compiler defines '__thumb__' and '__thumb2__' - */ +// Determine actual [ARM][THUMB[1][2]] ISA using compiler predefined macros: +// - for '-mthumb -march=armv6' compiler defines '__thumb__' +// - for '-mthumb -march=armv7' compiler defines '__thumb__' and '__thumb2__' #if defined(__thumb2__) || defined(__thumb__) #define DEFINE_CODE_STATE .thumb SEPARATOR #define DECLARE_FUNC_ENCODING .thumb_func SEPARATOR @@ -201,4 +197,4 @@ #define END_COMPILERRT_FUNCTION(name) #endif -#endif /* COMPILERRT_ASSEMBLY_H */ +#endif // COMPILERRT_ASSEMBLY_H diff --git a/lib/builtins/atomic.c b/lib/builtins/atomic.c index ee35e342eda5..0f82803a6416 100644 --- a/lib/builtins/atomic.c +++ b/lib/builtins/atomic.c @@ -1,29 +1,27 @@ -/*===-- atomic.c - Implement support functions for atomic operations.------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - *===----------------------------------------------------------------------=== - * - * atomic.c defines a set of functions for performing atomic accesses on - * arbitrary-sized memory locations. This design uses locks that should - * be fast in the uncontended case, for two reasons: - * - * 1) This code must work with C programs that do not link to anything - * (including pthreads) and so it should not depend on any pthread - * functions. - * 2) Atomic operations, rather than explicit mutexes, are most commonly used - * on code where contended operations are rate. - * - * To avoid needing a per-object lock, this code allocates an array of - * locks and hashes the object pointers to find the one that it should use. - * For operations that must be atomic on two locations, the lower lock is - * always acquired first, to avoid deadlock. - * - *===----------------------------------------------------------------------=== - */ +//===-- atomic.c - Implement support functions for atomic operations.------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// atomic.c defines a set of functions for performing atomic accesses on +// arbitrary-sized memory locations. This design uses locks that should +// be fast in the uncontended case, for two reasons: +// +// 1) This code must work with C programs that do not link to anything +// (including pthreads) and so it should not depend on any pthread +// functions. +// 2) Atomic operations, rather than explicit mutexes, are most commonly used +// on code where contended operations are rate. +// +// To avoid needing a per-object lock, this code allocates an array of +// locks and hashes the object pointers to find the one that it should use. +// For operations that must be atomic on two locations, the lower lock is +// always acquired first, to avoid deadlock. +// +//===----------------------------------------------------------------------===// #include <stdint.h> #include <string.h> @@ -35,13 +33,14 @@ #pragma redefine_extname __atomic_load_c SYMBOL_NAME(__atomic_load) #pragma redefine_extname __atomic_store_c SYMBOL_NAME(__atomic_store) #pragma redefine_extname __atomic_exchange_c SYMBOL_NAME(__atomic_exchange) -#pragma redefine_extname __atomic_compare_exchange_c SYMBOL_NAME(__atomic_compare_exchange) +#pragma redefine_extname __atomic_compare_exchange_c SYMBOL_NAME( \ + __atomic_compare_exchange) /// Number of locks. This allocates one page on 32-bit platforms, two on /// 64-bit. This can be specified externally if a different trade between /// memory usage and contention probability is required for a given platform. #ifndef SPINLOCK_COUNT -#define SPINLOCK_COUNT (1<<10) +#define SPINLOCK_COUNT (1 << 10) #endif static const long SPINLOCK_MASK = SPINLOCK_COUNT - 1; @@ -52,38 +51,35 @@ static const long SPINLOCK_MASK = SPINLOCK_COUNT - 1; //////////////////////////////////////////////////////////////////////////////// #ifdef __FreeBSD__ #include <errno.h> -#include <sys/types.h> #include <machine/atomic.h> +#include <sys/types.h> #include <sys/umtx.h> typedef struct _usem Lock; __inline static void unlock(Lock *l) { - __c11_atomic_store((_Atomic(uint32_t)*)&l->_count, 1, __ATOMIC_RELEASE); + __c11_atomic_store((_Atomic(uint32_t) *)&l->_count, 1, __ATOMIC_RELEASE); __c11_atomic_thread_fence(__ATOMIC_SEQ_CST); if (l->_has_waiters) - _umtx_op(l, UMTX_OP_SEM_WAKE, 1, 0, 0); + _umtx_op(l, UMTX_OP_SEM_WAKE, 1, 0, 0); } __inline static void lock(Lock *l) { uint32_t old = 1; - while (!__c11_atomic_compare_exchange_weak((_Atomic(uint32_t)*)&l->_count, &old, - 0, __ATOMIC_ACQUIRE, __ATOMIC_RELAXED)) { + while (!__c11_atomic_compare_exchange_weak((_Atomic(uint32_t) *)&l->_count, + &old, 0, __ATOMIC_ACQUIRE, + __ATOMIC_RELAXED)) { _umtx_op(l, UMTX_OP_SEM_WAIT, 0, 0, 0); old = 1; } } /// locks for atomic operations -static Lock locks[SPINLOCK_COUNT] = { [0 ... SPINLOCK_COUNT-1] = {0,1,0} }; +static Lock locks[SPINLOCK_COUNT] = {[0 ... SPINLOCK_COUNT - 1] = {0, 1, 0}}; #elif defined(__APPLE__) #include <libkern/OSAtomic.h> typedef OSSpinLock Lock; -__inline static void unlock(Lock *l) { - OSSpinLockUnlock(l); -} +__inline static void unlock(Lock *l) { OSSpinLockUnlock(l); } /// Locks a lock. In the current implementation, this is potentially /// unbounded in the contended case. -__inline static void lock(Lock *l) { - OSSpinLockLock(l); -} +__inline static void lock(Lock *l) { OSSpinLockLock(l); } static Lock locks[SPINLOCK_COUNT]; // initialized to OS_SPINLOCK_INIT which is 0 #else @@ -97,20 +93,19 @@ __inline static void unlock(Lock *l) { __inline static void lock(Lock *l) { uintptr_t old = 0; while (!__c11_atomic_compare_exchange_weak(l, &old, 1, __ATOMIC_ACQUIRE, - __ATOMIC_RELAXED)) + __ATOMIC_RELAXED)) old = 0; } /// locks for atomic operations static Lock locks[SPINLOCK_COUNT]; #endif - -/// Returns a lock to use for a given pointer. +/// Returns a lock to use for a given pointer. static __inline Lock *lock_for_pointer(void *ptr) { intptr_t hash = (intptr_t)ptr; // Disregard the lowest 4 bits. We want all values that may be part of the // same memory operation to hash to the same value and therefore use the same - // lock. + // lock. hash >>= 4; // Use the next bits as the basis for the hash intptr_t low = hash & SPINLOCK_MASK; @@ -133,36 +128,44 @@ static __inline Lock *lock_for_pointer(void *ptr) { /// Macro that calls the compiler-generated lock-free versions of functions /// when they exist. -#define LOCK_FREE_CASES() \ - do {\ - switch (size) {\ - case 2:\ - if (IS_LOCK_FREE_2) {\ - LOCK_FREE_ACTION(uint16_t);\ - }\ - case 4:\ - if (IS_LOCK_FREE_4) {\ - LOCK_FREE_ACTION(uint32_t);\ - }\ - case 8:\ - if (IS_LOCK_FREE_8) {\ - LOCK_FREE_ACTION(uint64_t);\ - }\ - case 16:\ - if (IS_LOCK_FREE_16) {\ - /* FIXME: __uint128_t isn't available on 32 bit platforms. - LOCK_FREE_ACTION(__uint128_t);*/\ - }\ - }\ +#define LOCK_FREE_CASES() \ + do { \ + switch (size) { \ + case 1: \ + if (IS_LOCK_FREE_1) { \ + LOCK_FREE_ACTION(uint8_t); \ + } \ + break; \ + case 2: \ + if (IS_LOCK_FREE_2) { \ + LOCK_FREE_ACTION(uint16_t); \ + } \ + break; \ + case 4: \ + if (IS_LOCK_FREE_4) { \ + LOCK_FREE_ACTION(uint32_t); \ + } \ + break; \ + case 8: \ + if (IS_LOCK_FREE_8) { \ + LOCK_FREE_ACTION(uint64_t); \ + } \ + break; \ + case 16: \ + if (IS_LOCK_FREE_16) { \ + /* FIXME: __uint128_t isn't available on 32 bit platforms. \ + LOCK_FREE_ACTION(__uint128_t);*/ \ + } \ + break; \ + } \ } while (0) - /// An atomic load operation. This is atomic with respect to the source /// pointer only. void __atomic_load_c(int size, void *src, void *dest, int model) { -#define LOCK_FREE_ACTION(type) \ - *((type*)dest) = __c11_atomic_load((_Atomic(type)*)src, model);\ - return; +#define LOCK_FREE_ACTION(type) \ + *((type *)dest) = __c11_atomic_load((_Atomic(type) *)src, model); \ + return; LOCK_FREE_CASES(); #undef LOCK_FREE_ACTION Lock *l = lock_for_pointer(src); @@ -174,9 +177,9 @@ void __atomic_load_c(int size, void *src, void *dest, int model) { /// An atomic store operation. This is atomic with respect to the destination /// pointer only. void __atomic_store_c(int size, void *dest, void *src, int model) { -#define LOCK_FREE_ACTION(type) \ - __c11_atomic_store((_Atomic(type)*)dest, *(type*)dest, model);\ - return; +#define LOCK_FREE_ACTION(type) \ + __c11_atomic_store((_Atomic(type) *)dest, *(type *)src, model); \ + return; LOCK_FREE_CASES(); #undef LOCK_FREE_ACTION Lock *l = lock_for_pointer(dest); @@ -189,12 +192,13 @@ void __atomic_store_c(int size, void *dest, void *src, int model) { /// to the value at *expected, then this copies value at *desired to *ptr. If /// they are not, then this stores the current value from *ptr in *expected. /// -/// This function returns 1 if the exchange takes place or 0 if it fails. +/// This function returns 1 if the exchange takes place or 0 if it fails. int __atomic_compare_exchange_c(int size, void *ptr, void *expected, - void *desired, int success, int failure) { -#define LOCK_FREE_ACTION(type) \ - return __c11_atomic_compare_exchange_strong((_Atomic(type)*)ptr, (type*)expected,\ - *(type*)desired, success, failure) + void *desired, int success, int failure) { +#define LOCK_FREE_ACTION(type) \ + return __c11_atomic_compare_exchange_strong( \ + (_Atomic(type) *)ptr, (type *)expected, *(type *)desired, success, \ + failure) LOCK_FREE_CASES(); #undef LOCK_FREE_ACTION Lock *l = lock_for_pointer(ptr); @@ -212,10 +216,10 @@ int __atomic_compare_exchange_c(int size, void *ptr, void *expected, /// Performs an atomic exchange operation between two pointers. This is atomic /// with respect to the target address. void __atomic_exchange_c(int size, void *ptr, void *val, void *old, int model) { -#define LOCK_FREE_ACTION(type) \ - *(type*)old = __c11_atomic_exchange((_Atomic(type)*)ptr, *(type*)val,\ - model);\ - return; +#define LOCK_FREE_ACTION(type) \ + *(type *)old = \ + __c11_atomic_exchange((_Atomic(type) *)ptr, *(type *)val, model); \ + return; LOCK_FREE_CASES(); #undef LOCK_FREE_ACTION Lock *l = lock_for_pointer(ptr); @@ -230,96 +234,96 @@ void __atomic_exchange_c(int size, void *ptr, void *val, void *old, int model) { // specialised versions of the above functions. //////////////////////////////////////////////////////////////////////////////// #ifdef __SIZEOF_INT128__ -#define OPTIMISED_CASES\ - OPTIMISED_CASE(1, IS_LOCK_FREE_1, uint8_t)\ - OPTIMISED_CASE(2, IS_LOCK_FREE_2, uint16_t)\ - OPTIMISED_CASE(4, IS_LOCK_FREE_4, uint32_t)\ - OPTIMISED_CASE(8, IS_LOCK_FREE_8, uint64_t)\ +#define OPTIMISED_CASES \ + OPTIMISED_CASE(1, IS_LOCK_FREE_1, uint8_t) \ + OPTIMISED_CASE(2, IS_LOCK_FREE_2, uint16_t) \ + OPTIMISED_CASE(4, IS_LOCK_FREE_4, uint32_t) \ + OPTIMISED_CASE(8, IS_LOCK_FREE_8, uint64_t) \ OPTIMISED_CASE(16, IS_LOCK_FREE_16, __uint128_t) #else -#define OPTIMISED_CASES\ - OPTIMISED_CASE(1, IS_LOCK_FREE_1, uint8_t)\ - OPTIMISED_CASE(2, IS_LOCK_FREE_2, uint16_t)\ - OPTIMISED_CASE(4, IS_LOCK_FREE_4, uint32_t)\ +#define OPTIMISED_CASES \ + OPTIMISED_CASE(1, IS_LOCK_FREE_1, uint8_t) \ + OPTIMISED_CASE(2, IS_LOCK_FREE_2, uint16_t) \ + OPTIMISED_CASE(4, IS_LOCK_FREE_4, uint32_t) \ OPTIMISED_CASE(8, IS_LOCK_FREE_8, uint64_t) #endif -#define OPTIMISED_CASE(n, lockfree, type)\ -type __atomic_load_##n(type *src, int model) {\ - if (lockfree)\ - return __c11_atomic_load((_Atomic(type)*)src, model);\ - Lock *l = lock_for_pointer(src);\ - lock(l);\ - type val = *src;\ - unlock(l);\ - return val;\ -} +#define OPTIMISED_CASE(n, lockfree, type) \ + type __atomic_load_##n(type *src, int model) { \ + if (lockfree) \ + return __c11_atomic_load((_Atomic(type) *)src, model); \ + Lock *l = lock_for_pointer(src); \ + lock(l); \ + type val = *src; \ + unlock(l); \ + return val; \ + } OPTIMISED_CASES #undef OPTIMISED_CASE -#define OPTIMISED_CASE(n, lockfree, type)\ -void __atomic_store_##n(type *dest, type val, int model) {\ - if (lockfree) {\ - __c11_atomic_store((_Atomic(type)*)dest, val, model);\ - return;\ - }\ - Lock *l = lock_for_pointer(dest);\ - lock(l);\ - *dest = val;\ - unlock(l);\ - return;\ -} +#define OPTIMISED_CASE(n, lockfree, type) \ + void __atomic_store_##n(type *dest, type val, int model) { \ + if (lockfree) { \ + __c11_atomic_store((_Atomic(type) *)dest, val, model); \ + return; \ + } \ + Lock *l = lock_for_pointer(dest); \ + lock(l); \ + *dest = val; \ + unlock(l); \ + return; \ + } OPTIMISED_CASES #undef OPTIMISED_CASE -#define OPTIMISED_CASE(n, lockfree, type)\ -type __atomic_exchange_##n(type *dest, type val, int model) {\ - if (lockfree)\ - return __c11_atomic_exchange((_Atomic(type)*)dest, val, model);\ - Lock *l = lock_for_pointer(dest);\ - lock(l);\ - type tmp = *dest;\ - *dest = val;\ - unlock(l);\ - return tmp;\ -} +#define OPTIMISED_CASE(n, lockfree, type) \ + type __atomic_exchange_##n(type *dest, type val, int model) { \ + if (lockfree) \ + return __c11_atomic_exchange((_Atomic(type) *)dest, val, model); \ + Lock *l = lock_for_pointer(dest); \ + lock(l); \ + type tmp = *dest; \ + *dest = val; \ + unlock(l); \ + return tmp; \ + } OPTIMISED_CASES #undef OPTIMISED_CASE -#define OPTIMISED_CASE(n, lockfree, type)\ -int __atomic_compare_exchange_##n(type *ptr, type *expected, type desired,\ - int success, int failure) {\ - if (lockfree)\ - return __c11_atomic_compare_exchange_strong((_Atomic(type)*)ptr, expected, desired,\ - success, failure);\ - Lock *l = lock_for_pointer(ptr);\ - lock(l);\ - if (*ptr == *expected) {\ - *ptr = desired;\ - unlock(l);\ - return 1;\ - }\ - *expected = *ptr;\ - unlock(l);\ - return 0;\ -} +#define OPTIMISED_CASE(n, lockfree, type) \ + int __atomic_compare_exchange_##n(type *ptr, type *expected, type desired, \ + int success, int failure) { \ + if (lockfree) \ + return __c11_atomic_compare_exchange_strong( \ + (_Atomic(type) *)ptr, expected, desired, success, failure); \ + Lock *l = lock_for_pointer(ptr); \ + lock(l); \ + if (*ptr == *expected) { \ + *ptr = desired; \ + unlock(l); \ + return 1; \ + } \ + *expected = *ptr; \ + unlock(l); \ + return 0; \ + } OPTIMISED_CASES #undef OPTIMISED_CASE //////////////////////////////////////////////////////////////////////////////// // Atomic read-modify-write operations for integers of various sizes. //////////////////////////////////////////////////////////////////////////////// -#define ATOMIC_RMW(n, lockfree, type, opname, op) \ -type __atomic_fetch_##opname##_##n(type *ptr, type val, int model) {\ - if (lockfree) \ - return __c11_atomic_fetch_##opname((_Atomic(type)*)ptr, val, model);\ - Lock *l = lock_for_pointer(ptr);\ - lock(l);\ - type tmp = *ptr;\ - *ptr = tmp op val;\ - unlock(l);\ - return tmp;\ -} +#define ATOMIC_RMW(n, lockfree, type, opname, op) \ + type __atomic_fetch_##opname##_##n(type *ptr, type val, int model) { \ + if (lockfree) \ + return __c11_atomic_fetch_##opname((_Atomic(type) *)ptr, val, model); \ + Lock *l = lock_for_pointer(ptr); \ + lock(l); \ + type tmp = *ptr; \ + *ptr = tmp op val; \ + unlock(l); \ + return tmp; \ + } #define OPTIMISED_CASE(n, lockfree, type) ATOMIC_RMW(n, lockfree, type, add, +) OPTIMISED_CASES diff --git a/lib/builtins/atomic_flag_clear.c b/lib/builtins/atomic_flag_clear.c index da912af64312..983e5d7f076b 100644 --- a/lib/builtins/atomic_flag_clear.c +++ b/lib/builtins/atomic_flag_clear.c @@ -1,16 +1,14 @@ -/*===-- atomic_flag_clear.c -------------------------------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - *===------------------------------------------------------------------------=== - * - * This file implements atomic_flag_clear from C11's stdatomic.h. - * - *===------------------------------------------------------------------------=== - */ +//===-- atomic_flag_clear.c -----------------------------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements atomic_flag_clear from C11's stdatomic.h. +// +//===----------------------------------------------------------------------===// #ifndef __has_include #define __has_include(inc) 0 diff --git a/lib/builtins/atomic_flag_clear_explicit.c b/lib/builtins/atomic_flag_clear_explicit.c index 1059b787f169..e61c0647684e 100644 --- a/lib/builtins/atomic_flag_clear_explicit.c +++ b/lib/builtins/atomic_flag_clear_explicit.c @@ -1,16 +1,14 @@ -/*===-- atomic_flag_clear_explicit.c ----------------------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - *===------------------------------------------------------------------------=== - * - * This file implements atomic_flag_clear_explicit from C11's stdatomic.h. - * - *===------------------------------------------------------------------------=== - */ +//===-- atomic_flag_clear_explicit.c --------------------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements atomic_flag_clear_explicit from C11's stdatomic.h. +// +//===----------------------------------------------------------------------===// #ifndef __has_include #define __has_include(inc) 0 diff --git a/lib/builtins/atomic_flag_test_and_set.c b/lib/builtins/atomic_flag_test_and_set.c index e8811d39ef25..ee22b08b5669 100644 --- a/lib/builtins/atomic_flag_test_and_set.c +++ b/lib/builtins/atomic_flag_test_and_set.c @@ -1,16 +1,14 @@ -/*===-- atomic_flag_test_and_set.c ------------------------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - *===------------------------------------------------------------------------=== - * - * This file implements atomic_flag_test_and_set from C11's stdatomic.h. - * - *===------------------------------------------------------------------------=== - */ +//===-- atomic_flag_test_and_set.c ----------------------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements atomic_flag_test_and_set from C11's stdatomic.h. +// +//===----------------------------------------------------------------------===// #ifndef __has_include #define __has_include(inc) 0 diff --git a/lib/builtins/atomic_flag_test_and_set_explicit.c b/lib/builtins/atomic_flag_test_and_set_explicit.c index 5c8c2df90543..8c9d03994267 100644 --- a/lib/builtins/atomic_flag_test_and_set_explicit.c +++ b/lib/builtins/atomic_flag_test_and_set_explicit.c @@ -1,16 +1,14 @@ -/*===-- atomic_flag_test_and_set_explicit.c ---------------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - *===------------------------------------------------------------------------=== - * - * This file implements atomic_flag_test_and_set_explicit from C11's stdatomic.h - * - *===------------------------------------------------------------------------=== - */ +//===-- atomic_flag_test_and_set_explicit.c -------------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements atomic_flag_test_and_set_explicit from C11's stdatomic.h +// +//===----------------------------------------------------------------------===// #ifndef __has_include #define __has_include(inc) 0 diff --git a/lib/builtins/atomic_signal_fence.c b/lib/builtins/atomic_signal_fence.c index 9ccc2ae60ad8..f4f5169d3008 100644 --- a/lib/builtins/atomic_signal_fence.c +++ b/lib/builtins/atomic_signal_fence.c @@ -1,16 +1,14 @@ -/*===-- atomic_signal_fence.c -----------------------------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - *===------------------------------------------------------------------------=== - * - * This file implements atomic_signal_fence from C11's stdatomic.h. - * - *===------------------------------------------------------------------------=== - */ +//===-- atomic_signal_fence.c ---------------------------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements atomic_signal_fence from C11's stdatomic.h. +// +//===----------------------------------------------------------------------===// #ifndef __has_include #define __has_include(inc) 0 diff --git a/lib/builtins/atomic_thread_fence.c b/lib/builtins/atomic_thread_fence.c index d22560151bc8..5659ecb0b1f9 100644 --- a/lib/builtins/atomic_thread_fence.c +++ b/lib/builtins/atomic_thread_fence.c @@ -1,16 +1,14 @@ -/*===-- atomic_thread_fence.c -----------------------------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - *===------------------------------------------------------------------------=== - * - * This file implements atomic_thread_fence from C11's stdatomic.h. - * - *===------------------------------------------------------------------------=== - */ +//===-- atomic_thread_fence.c ---------------------------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements atomic_thread_fence from C11's stdatomic.h. +// +//===----------------------------------------------------------------------===// #ifndef __has_include #define __has_include(inc) 0 diff --git a/lib/builtins/bswapdi2.c b/lib/builtins/bswapdi2.c index eb220007bb10..cd049f58e328 100644 --- a/lib/builtins/bswapdi2.c +++ b/lib/builtins/bswapdi2.c @@ -1,16 +1,14 @@ -/* ===-- bswapdi2.c - Implement __bswapdi2 ---------------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - * - * This file implements __bswapdi2 for the compiler_rt library. - * - * ===----------------------------------------------------------------------=== - */ +//===-- bswapdi2.c - Implement __bswapdi2 ---------------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements __bswapdi2 for the compiler_rt library. +// +//===----------------------------------------------------------------------===// #include "int_lib.h" diff --git a/lib/builtins/bswapsi2.c b/lib/builtins/bswapsi2.c index 5d941e69f7c4..ec566d61a2b0 100644 --- a/lib/builtins/bswapsi2.c +++ b/lib/builtins/bswapsi2.c @@ -1,23 +1,20 @@ -/* ===-- bswapsi2.c - Implement __bswapsi2 ---------------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - * - * This file implements __bswapsi2 for the compiler_rt library. - * - * ===----------------------------------------------------------------------=== - */ +//===-- bswapsi2.c - Implement __bswapsi2 ---------------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements __bswapsi2 for the compiler_rt library. +// +//===----------------------------------------------------------------------===// #include "int_lib.h" COMPILER_RT_ABI uint32_t __bswapsi2(uint32_t u) { - return ( - (((u)&0xff000000) >> 24) | - (((u)&0x00ff0000) >> 8) | - (((u)&0x0000ff00) << 8) | - (((u)&0x000000ff) << 24)); + return ((((u)&0xff000000) >> 24) | + (((u)&0x00ff0000) >> 8) | + (((u)&0x0000ff00) << 8) | + (((u)&0x000000ff) << 24)); } diff --git a/lib/builtins/clear_cache.c b/lib/builtins/clear_cache.c index 9dcab344ad1b..76dc1968cc7e 100644 --- a/lib/builtins/clear_cache.c +++ b/lib/builtins/clear_cache.c @@ -1,179 +1,164 @@ -/* ===-- clear_cache.c - Implement __clear_cache ---------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - */ +//===-- clear_cache.c - Implement __clear_cache ---------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// #include "int_lib.h" #include <assert.h> #include <stddef.h> #if __APPLE__ - #include <libkern/OSCacheControl.h> +#include <libkern/OSCacheControl.h> #endif #if defined(_WIN32) -/* Forward declare Win32 APIs since the GCC mode driver does not handle the - newer SDKs as well as needed. */ +// Forward declare Win32 APIs since the GCC mode driver does not handle the +// newer SDKs as well as needed. uint32_t FlushInstructionCache(uintptr_t hProcess, void *lpBaseAddress, uintptr_t dwSize); uintptr_t GetCurrentProcess(void); #endif #if defined(__FreeBSD__) && defined(__arm__) - #include <sys/types.h> - #include <machine/sysarch.h> +#include <machine/sysarch.h> +#include <sys/types.h> #endif #if defined(__NetBSD__) && defined(__arm__) - #include <machine/sysarch.h> +#include <machine/sysarch.h> #endif #if defined(__OpenBSD__) && defined(__mips__) - #include <sys/types.h> - #include <machine/sysarch.h> +#include <machine/sysarch.h> +#include <sys/types.h> #endif #if defined(__linux__) && defined(__mips__) - #include <sys/cachectl.h> - #include <sys/syscall.h> - #include <unistd.h> - #if defined(__ANDROID__) && defined(__LP64__) - /* - * clear_mips_cache - Invalidates instruction cache for Mips. - */ - static void clear_mips_cache(const void* Addr, size_t Size) { - __asm__ volatile ( - ".set push\n" - ".set noreorder\n" - ".set noat\n" - "beq %[Size], $zero, 20f\n" /* If size == 0, branch around. */ - "nop\n" - "daddu %[Size], %[Addr], %[Size]\n" /* Calculate end address + 1 */ - "rdhwr $v0, $1\n" /* Get step size for SYNCI. - $1 is $HW_SYNCI_Step */ - "beq $v0, $zero, 20f\n" /* If no caches require - synchronization, branch - around. */ - "nop\n" - "10:\n" - "synci 0(%[Addr])\n" /* Synchronize all caches around - address. */ - "daddu %[Addr], %[Addr], $v0\n" /* Add step size. */ - "sltu $at, %[Addr], %[Size]\n" /* Compare current with end - address. */ - "bne $at, $zero, 10b\n" /* Branch if more to do. */ - "nop\n" - "sync\n" /* Clear memory hazards. */ - "20:\n" - "bal 30f\n" - "nop\n" - "30:\n" - "daddiu $ra, $ra, 12\n" /* $ra has a value of $pc here. - Add offset of 12 to point to the - instruction after the last nop. - */ - "jr.hb $ra\n" /* Return, clearing instruction - hazards. */ - "nop\n" - ".set pop\n" - : [Addr] "+r"(Addr), [Size] "+r"(Size) - :: "at", "ra", "v0", "memory" - ); - } - #endif +#include <sys/cachectl.h> +#include <sys/syscall.h> +#include <unistd.h> +#if defined(__ANDROID__) && defined(__LP64__) +// clear_mips_cache - Invalidates instruction cache for Mips. +static void clear_mips_cache(const void *Addr, size_t Size) { + __asm__ volatile( + ".set push\n" + ".set noreorder\n" + ".set noat\n" + "beq %[Size], $zero, 20f\n" // If size == 0, branch around. + "nop\n" + "daddu %[Size], %[Addr], %[Size]\n" // Calculate end address + 1 + "rdhwr $v0, $1\n" // Get step size for SYNCI. + // $1 is $HW_SYNCI_Step + "beq $v0, $zero, 20f\n" // If no caches require + // synchronization, branch + // around. + "nop\n" + "10:\n" + "synci 0(%[Addr])\n" // Synchronize all caches around + // address. + "daddu %[Addr], %[Addr], $v0\n" // Add step size. + "sltu $at, %[Addr], %[Size]\n" // Compare current with end + // address. + "bne $at, $zero, 10b\n" // Branch if more to do. + "nop\n" + "sync\n" // Clear memory hazards. + "20:\n" + "bal 30f\n" + "nop\n" + "30:\n" + "daddiu $ra, $ra, 12\n" // $ra has a value of $pc here. + // Add offset of 12 to point to the + // instruction after the last nop. + // + "jr.hb $ra\n" // Return, clearing instruction + // hazards. + "nop\n" + ".set pop\n" + : [ Addr ] "+r"(Addr), [ Size ] "+r"(Size)::"at", "ra", "v0", "memory"); +} +#endif #endif -/* - * The compiler generates calls to __clear_cache() when creating - * trampoline functions on the stack for use with nested functions. - * It is expected to invalidate the instruction cache for the - * specified range. - */ +// The compiler generates calls to __clear_cache() when creating +// trampoline functions on the stack for use with nested functions. +// It is expected to invalidate the instruction cache for the +// specified range. void __clear_cache(void *start, void *end) { #if __i386__ || __x86_64__ || defined(_M_IX86) || defined(_M_X64) -/* - * Intel processors have a unified instruction and data cache - * so there is nothing to do - */ +// Intel processors have a unified instruction and data cache +// so there is nothing to do #elif defined(_WIN32) && (defined(__arm__) || defined(__aarch64__)) - FlushInstructionCache(GetCurrentProcess(), start, end - start); + FlushInstructionCache(GetCurrentProcess(), start, end - start); #elif defined(__arm__) && !defined(__APPLE__) - #if defined(__FreeBSD__) || defined(__NetBSD__) - struct arm_sync_icache_args arg; - - arg.addr = (uintptr_t)start; - arg.len = (uintptr_t)end - (uintptr_t)start; - - sysarch(ARM_SYNC_ICACHE, &arg); - #elif defined(__linux__) - /* - * We used to include asm/unistd.h for the __ARM_NR_cacheflush define, but - * it also brought many other unused defines, as well as a dependency on - * kernel headers to be installed. - * - * This value is stable at least since Linux 3.13 and should remain so for - * compatibility reasons, warranting it's re-definition here. - */ - #define __ARM_NR_cacheflush 0x0f0002 - register int start_reg __asm("r0") = (int) (intptr_t) start; - const register int end_reg __asm("r1") = (int) (intptr_t) end; - const register int flags __asm("r2") = 0; - const register int syscall_nr __asm("r7") = __ARM_NR_cacheflush; - __asm __volatile("svc 0x0" - : "=r"(start_reg) - : "r"(syscall_nr), "r"(start_reg), "r"(end_reg), - "r"(flags)); - assert(start_reg == 0 && "Cache flush syscall failed."); - #else - compilerrt_abort(); - #endif +#if defined(__FreeBSD__) || defined(__NetBSD__) + struct arm_sync_icache_args arg; + + arg.addr = (uintptr_t)start; + arg.len = (uintptr_t)end - (uintptr_t)start; + + sysarch(ARM_SYNC_ICACHE, &arg); +#elif defined(__linux__) +// We used to include asm/unistd.h for the __ARM_NR_cacheflush define, but +// it also brought many other unused defines, as well as a dependency on +// kernel headers to be installed. +// +// This value is stable at least since Linux 3.13 and should remain so for +// compatibility reasons, warranting it's re-definition here. +#define __ARM_NR_cacheflush 0x0f0002 + register int start_reg __asm("r0") = (int)(intptr_t)start; + const register int end_reg __asm("r1") = (int)(intptr_t)end; + const register int flags __asm("r2") = 0; + const register int syscall_nr __asm("r7") = __ARM_NR_cacheflush; + __asm __volatile("svc 0x0" + : "=r"(start_reg) + : "r"(syscall_nr), "r"(start_reg), "r"(end_reg), "r"(flags)); + assert(start_reg == 0 && "Cache flush syscall failed."); +#else + compilerrt_abort(); +#endif #elif defined(__linux__) && defined(__mips__) - const uintptr_t start_int = (uintptr_t) start; - const uintptr_t end_int = (uintptr_t) end; - #if defined(__ANDROID__) && defined(__LP64__) - // Call synci implementation for short address range. - const uintptr_t address_range_limit = 256; - if ((end_int - start_int) <= address_range_limit) { - clear_mips_cache(start, (end_int - start_int)); - } else { - syscall(__NR_cacheflush, start, (end_int - start_int), BCACHE); - } - #else - syscall(__NR_cacheflush, start, (end_int - start_int), BCACHE); - #endif + const uintptr_t start_int = (uintptr_t)start; + const uintptr_t end_int = (uintptr_t)end; +#if defined(__ANDROID__) && defined(__LP64__) + // Call synci implementation for short address range. + const uintptr_t address_range_limit = 256; + if ((end_int - start_int) <= address_range_limit) { + clear_mips_cache(start, (end_int - start_int)); + } else { + syscall(__NR_cacheflush, start, (end_int - start_int), BCACHE); + } +#else + syscall(__NR_cacheflush, start, (end_int - start_int), BCACHE); +#endif #elif defined(__mips__) && defined(__OpenBSD__) cacheflush(start, (uintptr_t)end - (uintptr_t)start, BCACHE); #elif defined(__aarch64__) && !defined(__APPLE__) - uint64_t xstart = (uint64_t)(uintptr_t) start; - uint64_t xend = (uint64_t)(uintptr_t) end; + uint64_t xstart = (uint64_t)(uintptr_t)start; + uint64_t xend = (uint64_t)(uintptr_t)end; uint64_t addr; // Get Cache Type Info uint64_t ctr_el0; __asm __volatile("mrs %0, ctr_el0" : "=r"(ctr_el0)); - /* - * dc & ic instructions must use 64bit registers so we don't use - * uintptr_t in case this runs in an IPL32 environment. - */ + // dc & ic instructions must use 64bit registers so we don't use + // uintptr_t in case this runs in an IPL32 environment. const size_t dcache_line_size = 4 << ((ctr_el0 >> 16) & 15); for (addr = xstart & ~(dcache_line_size - 1); addr < xend; addr += dcache_line_size) - __asm __volatile("dc cvau, %0" :: "r"(addr)); + __asm __volatile("dc cvau, %0" ::"r"(addr)); __asm __volatile("dsb ish"); const size_t icache_line_size = 4 << ((ctr_el0 >> 0) & 15); for (addr = xstart & ~(icache_line_size - 1); addr < xend; addr += icache_line_size) - __asm __volatile("ic ivau, %0" :: "r"(addr)); + __asm __volatile("ic ivau, %0" ::"r"(addr)); __asm __volatile("isb sy"); -#elif defined (__powerpc64__) +#elif defined(__powerpc64__) const size_t line_size = 32; const size_t len = (uintptr_t)end - (uintptr_t)start; @@ -189,11 +174,11 @@ void __clear_cache(void *start, void *end) { __asm__ volatile("icbi 0, %0" : : "r"(line)); __asm__ volatile("isync"); #else - #if __APPLE__ - /* On Darwin, sys_icache_invalidate() provides this functionality */ - sys_icache_invalidate(start, end-start); - #else - compilerrt_abort(); - #endif +#if __APPLE__ + // On Darwin, sys_icache_invalidate() provides this functionality + sys_icache_invalidate(start, end - start); +#else + compilerrt_abort(); +#endif #endif } diff --git a/lib/builtins/clzdi2.c b/lib/builtins/clzdi2.c index 1819e6be436b..a0bacb2ae39e 100644 --- a/lib/builtins/clzdi2.c +++ b/lib/builtins/clzdi2.c @@ -1,40 +1,35 @@ -/* ===-- clzdi2.c - Implement __clzdi2 -------------------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - * - * This file implements __clzdi2 for the compiler_rt library. - * - * ===----------------------------------------------------------------------=== - */ +//===-- clzdi2.c - Implement __clzdi2 -------------------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements __clzdi2 for the compiler_rt library. +// +//===----------------------------------------------------------------------===// #include "int_lib.h" -/* Returns: the number of leading 0-bits */ +// Returns: the number of leading 0-bits #if !defined(__clang__) && \ - ((defined(__sparc__) && defined(__arch64__)) || \ - defined(__mips64) || \ + ((defined(__sparc__) && defined(__arch64__)) || defined(__mips64) || \ (defined(__riscv) && __SIZEOF_POINTER__ >= 8)) -/* On 64-bit architectures with neither a native clz instruction nor a native - * ctz instruction, gcc resolves __builtin_clz to __clzdi2 rather than - * __clzsi2, leading to infinite recursion. */ +// On 64-bit architectures with neither a native clz instruction nor a native +// ctz instruction, gcc resolves __builtin_clz to __clzdi2 rather than +// __clzsi2, leading to infinite recursion. #define __builtin_clz(a) __clzsi2(a) extern si_int __clzsi2(si_int); #endif -/* Precondition: a != 0 */ +// Precondition: a != 0 -COMPILER_RT_ABI si_int -__clzdi2(di_int a) -{ - dwords x; - x.all = a; - const si_int f = -(x.s.high == 0); - return __builtin_clz((x.s.high & ~f) | (x.s.low & f)) + - (f & ((si_int)(sizeof(si_int) * CHAR_BIT))); +COMPILER_RT_ABI si_int __clzdi2(di_int a) { + dwords x; + x.all = a; + const si_int f = -(x.s.high == 0); + return __builtin_clz((x.s.high & ~f) | (x.s.low & f)) + + (f & ((si_int)(sizeof(si_int) * CHAR_BIT))); } diff --git a/lib/builtins/clzsi2.c b/lib/builtins/clzsi2.c index 25b8ed2c4c24..3f9f27f41331 100644 --- a/lib/builtins/clzsi2.c +++ b/lib/builtins/clzsi2.c @@ -1,53 +1,48 @@ -/* ===-- clzsi2.c - Implement __clzsi2 -------------------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - * - * This file implements __clzsi2 for the compiler_rt library. - * - * ===----------------------------------------------------------------------=== - */ +//===-- clzsi2.c - Implement __clzsi2 -------------------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements __clzsi2 for the compiler_rt library. +// +//===----------------------------------------------------------------------===// #include "int_lib.h" -/* Returns: the number of leading 0-bits */ +// Returns: the number of leading 0-bits -/* Precondition: a != 0 */ +// Precondition: a != 0 -COMPILER_RT_ABI si_int -__clzsi2(si_int a) -{ - su_int x = (su_int)a; - si_int t = ((x & 0xFFFF0000) == 0) << 4; /* if (x is small) t = 16 else 0 */ - x >>= 16 - t; /* x = [0 - 0xFFFF] */ - su_int r = t; /* r = [0, 16] */ - /* return r + clz(x) */ - t = ((x & 0xFF00) == 0) << 3; - x >>= 8 - t; /* x = [0 - 0xFF] */ - r += t; /* r = [0, 8, 16, 24] */ - /* return r + clz(x) */ - t = ((x & 0xF0) == 0) << 2; - x >>= 4 - t; /* x = [0 - 0xF] */ - r += t; /* r = [0, 4, 8, 12, 16, 20, 24, 28] */ - /* return r + clz(x) */ - t = ((x & 0xC) == 0) << 1; - x >>= 2 - t; /* x = [0 - 3] */ - r += t; /* r = [0 - 30] and is even */ - /* return r + clz(x) */ -/* switch (x) - * { - * case 0: - * return r + 2; - * case 1: - * return r + 1; - * case 2: - * case 3: - * return r; - * } - */ - return r + ((2 - x) & -((x & 2) == 0)); +COMPILER_RT_ABI si_int __clzsi2(si_int a) { + su_int x = (su_int)a; + si_int t = ((x & 0xFFFF0000) == 0) << 4; // if (x is small) t = 16 else 0 + x >>= 16 - t; // x = [0 - 0xFFFF] + su_int r = t; // r = [0, 16] + // return r + clz(x) + t = ((x & 0xFF00) == 0) << 3; + x >>= 8 - t; // x = [0 - 0xFF] + r += t; // r = [0, 8, 16, 24] + // return r + clz(x) + t = ((x & 0xF0) == 0) << 2; + x >>= 4 - t; // x = [0 - 0xF] + r += t; // r = [0, 4, 8, 12, 16, 20, 24, 28] + // return r + clz(x) + t = ((x & 0xC) == 0) << 1; + x >>= 2 - t; // x = [0 - 3] + r += t; // r = [0 - 30] and is even + // return r + clz(x) + // switch (x) + // { + // case 0: + // return r + 2; + // case 1: + // return r + 1; + // case 2: + // case 3: + // return r; + // } + return r + ((2 - x) & -((x & 2) == 0)); } diff --git a/lib/builtins/clzti2.c b/lib/builtins/clzti2.c index 15a7b3c90004..0c787104caa2 100644 --- a/lib/builtins/clzti2.c +++ b/lib/builtins/clzti2.c @@ -1,33 +1,29 @@ -/* ===-- clzti2.c - Implement __clzti2 -------------------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - * - * This file implements __clzti2 for the compiler_rt library. - * - * ===----------------------------------------------------------------------=== - */ +//===-- clzti2.c - Implement __clzti2 -------------------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements __clzti2 for the compiler_rt library. +// +//===----------------------------------------------------------------------===// #include "int_lib.h" #ifdef CRT_HAS_128BIT -/* Returns: the number of leading 0-bits */ +// Returns: the number of leading 0-bits -/* Precondition: a != 0 */ +// Precondition: a != 0 -COMPILER_RT_ABI si_int -__clzti2(ti_int a) -{ - twords x; - x.all = a; - const di_int f = -(x.s.high == 0); - return __builtin_clzll((x.s.high & ~f) | (x.s.low & f)) + - ((si_int)f & ((si_int)(sizeof(di_int) * CHAR_BIT))); +COMPILER_RT_ABI si_int __clzti2(ti_int a) { + twords x; + x.all = a; + const di_int f = -(x.s.high == 0); + return __builtin_clzll((x.s.high & ~f) | (x.s.low & f)) + + ((si_int)f & ((si_int)(sizeof(di_int) * CHAR_BIT))); } -#endif /* CRT_HAS_128BIT */ +#endif // CRT_HAS_128BIT diff --git a/lib/builtins/cmpdi2.c b/lib/builtins/cmpdi2.c index 52634d9c3362..951db85b5062 100644 --- a/lib/builtins/cmpdi2.c +++ b/lib/builtins/cmpdi2.c @@ -1,51 +1,42 @@ -/* ===-- cmpdi2.c - Implement __cmpdi2 -------------------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - * - * This file implements __cmpdi2 for the compiler_rt library. - * - * ===----------------------------------------------------------------------=== - */ +//===-- cmpdi2.c - Implement __cmpdi2 -------------------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements __cmpdi2 for the compiler_rt library. +// +//===----------------------------------------------------------------------===// #include "int_lib.h" -/* Returns: if (a < b) returns 0 -* if (a == b) returns 1 -* if (a > b) returns 2 -*/ +// Returns: if (a < b) returns 0 +// if (a == b) returns 1 +// if (a > b) returns 2 -COMPILER_RT_ABI si_int -__cmpdi2(di_int a, di_int b) -{ - dwords x; - x.all = a; - dwords y; - y.all = b; - if (x.s.high < y.s.high) - return 0; - if (x.s.high > y.s.high) - return 2; - if (x.s.low < y.s.low) - return 0; - if (x.s.low > y.s.low) - return 2; - return 1; +COMPILER_RT_ABI si_int __cmpdi2(di_int a, di_int b) { + dwords x; + x.all = a; + dwords y; + y.all = b; + if (x.s.high < y.s.high) + return 0; + if (x.s.high > y.s.high) + return 2; + if (x.s.low < y.s.low) + return 0; + if (x.s.low > y.s.low) + return 2; + return 1; } #ifdef __ARM_EABI__ -/* Returns: if (a < b) returns -1 -* if (a == b) returns 0 -* if (a > b) returns 1 -*/ -COMPILER_RT_ABI si_int -__aeabi_lcmp(di_int a, di_int b) -{ - return __cmpdi2(a, b) - 1; +// Returns: if (a < b) returns -1 +// if (a == b) returns 0 +// if (a > b) returns 1 +COMPILER_RT_ABI si_int __aeabi_lcmp(di_int a, di_int b) { + return __cmpdi2(a, b) - 1; } #endif - diff --git a/lib/builtins/cmpti2.c b/lib/builtins/cmpti2.c index 2c8b56e29a06..7f0ee1b5159d 100644 --- a/lib/builtins/cmpti2.c +++ b/lib/builtins/cmpti2.c @@ -1,42 +1,37 @@ -/* ===-- cmpti2.c - Implement __cmpti2 -------------------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - * - * This file implements __cmpti2 for the compiler_rt library. - * - * ===----------------------------------------------------------------------=== - */ +//===-- cmpti2.c - Implement __cmpti2 -------------------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements __cmpti2 for the compiler_rt library. +// +//===----------------------------------------------------------------------===// #include "int_lib.h" #ifdef CRT_HAS_128BIT -/* Returns: if (a < b) returns 0 - * if (a == b) returns 1 - * if (a > b) returns 2 - */ +// Returns: if (a < b) returns 0 +// if (a == b) returns 1 +// if (a > b) returns 2 -COMPILER_RT_ABI si_int -__cmpti2(ti_int a, ti_int b) -{ - twords x; - x.all = a; - twords y; - y.all = b; - if (x.s.high < y.s.high) - return 0; - if (x.s.high > y.s.high) - return 2; - if (x.s.low < y.s.low) - return 0; - if (x.s.low > y.s.low) - return 2; - return 1; +COMPILER_RT_ABI si_int __cmpti2(ti_int a, ti_int b) { + twords x; + x.all = a; + twords y; + y.all = b; + if (x.s.high < y.s.high) + return 0; + if (x.s.high > y.s.high) + return 2; + if (x.s.low < y.s.low) + return 0; + if (x.s.low > y.s.low) + return 2; + return 1; } -#endif /* CRT_HAS_128BIT */ +#endif // CRT_HAS_128BIT diff --git a/lib/builtins/comparedf2.c b/lib/builtins/comparedf2.c index 44e5d2b288a6..58290d87de65 100644 --- a/lib/builtins/comparedf2.c +++ b/lib/builtins/comparedf2.c @@ -1,9 +1,8 @@ //===-- lib/comparedf2.c - Double-precision comparisons -----------*- C -*-===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// // @@ -40,79 +39,93 @@ #define DOUBLE_PRECISION #include "fp_lib.h" -enum LE_RESULT { - LE_LESS = -1, - LE_EQUAL = 0, - LE_GREATER = 1, - LE_UNORDERED = 1 -}; +enum LE_RESULT { LE_LESS = -1, LE_EQUAL = 0, LE_GREATER = 1, LE_UNORDERED = 1 }; + +COMPILER_RT_ABI enum LE_RESULT __ledf2(fp_t a, fp_t b) { + + const srep_t aInt = toRep(a); + const srep_t bInt = toRep(b); + const rep_t aAbs = aInt & absMask; + const rep_t bAbs = bInt & absMask; + + // If either a or b is NaN, they are unordered. + if (aAbs > infRep || bAbs > infRep) + return LE_UNORDERED; + + // If a and b are both zeros, they are equal. + if ((aAbs | bAbs) == 0) + return LE_EQUAL; -COMPILER_RT_ABI enum LE_RESULT -__ledf2(fp_t a, fp_t b) { - - const srep_t aInt = toRep(a); - const srep_t bInt = toRep(b); - const rep_t aAbs = aInt & absMask; - const rep_t bAbs = bInt & absMask; - - // If either a or b is NaN, they are unordered. - if (aAbs > infRep || bAbs > infRep) return LE_UNORDERED; - - // If a and b are both zeros, they are equal. - if ((aAbs | bAbs) == 0) return LE_EQUAL; - - // If at least one of a and b is positive, we get the same result comparing - // a and b as signed integers as we would with a floating-point compare. - if ((aInt & bInt) >= 0) { - if (aInt < bInt) return LE_LESS; - else if (aInt == bInt) return LE_EQUAL; - else return LE_GREATER; - } - - // Otherwise, both are negative, so we need to flip the sense of the - // comparison to get the correct result. (This assumes a twos- or ones- - // complement integer representation; if integers are represented in a - // sign-magnitude representation, then this flip is incorrect). - else { - if (aInt > bInt) return LE_LESS; - else if (aInt == bInt) return LE_EQUAL; - else return LE_GREATER; - } + // If at least one of a and b is positive, we get the same result comparing + // a and b as signed integers as we would with a floating-point compare. + if ((aInt & bInt) >= 0) { + if (aInt < bInt) + return LE_LESS; + else if (aInt == bInt) + return LE_EQUAL; + else + return LE_GREATER; + } + + // Otherwise, both are negative, so we need to flip the sense of the + // comparison to get the correct result. (This assumes a twos- or ones- + // complement integer representation; if integers are represented in a + // sign-magnitude representation, then this flip is incorrect). + else { + if (aInt > bInt) + return LE_LESS; + else if (aInt == bInt) + return LE_EQUAL; + else + return LE_GREATER; + } } #if defined(__ELF__) // Alias for libgcc compatibility -FNALIAS(__cmpdf2, __ledf2); +COMPILER_RT_ALIAS(__ledf2, __cmpdf2) #endif +COMPILER_RT_ALIAS(__ledf2, __eqdf2) +COMPILER_RT_ALIAS(__ledf2, __ltdf2) +COMPILER_RT_ALIAS(__ledf2, __nedf2) enum GE_RESULT { - GE_LESS = -1, - GE_EQUAL = 0, - GE_GREATER = 1, - GE_UNORDERED = -1 // Note: different from LE_UNORDERED + GE_LESS = -1, + GE_EQUAL = 0, + GE_GREATER = 1, + GE_UNORDERED = -1 // Note: different from LE_UNORDERED }; -COMPILER_RT_ABI enum GE_RESULT -__gedf2(fp_t a, fp_t b) { - - const srep_t aInt = toRep(a); - const srep_t bInt = toRep(b); - const rep_t aAbs = aInt & absMask; - const rep_t bAbs = bInt & absMask; - - if (aAbs > infRep || bAbs > infRep) return GE_UNORDERED; - if ((aAbs | bAbs) == 0) return GE_EQUAL; - if ((aInt & bInt) >= 0) { - if (aInt < bInt) return GE_LESS; - else if (aInt == bInt) return GE_EQUAL; - else return GE_GREATER; - } else { - if (aInt > bInt) return GE_LESS; - else if (aInt == bInt) return GE_EQUAL; - else return GE_GREATER; - } +COMPILER_RT_ABI enum GE_RESULT __gedf2(fp_t a, fp_t b) { + + const srep_t aInt = toRep(a); + const srep_t bInt = toRep(b); + const rep_t aAbs = aInt & absMask; + const rep_t bAbs = bInt & absMask; + + if (aAbs > infRep || bAbs > infRep) + return GE_UNORDERED; + if ((aAbs | bAbs) == 0) + return GE_EQUAL; + if ((aInt & bInt) >= 0) { + if (aInt < bInt) + return GE_LESS; + else if (aInt == bInt) + return GE_EQUAL; + else + return GE_GREATER; + } else { + if (aInt > bInt) + return GE_LESS; + else if (aInt == bInt) + return GE_EQUAL; + else + return GE_GREATER; + } } +COMPILER_RT_ALIAS(__gedf2, __gtdf2) + COMPILER_RT_ABI int __unorddf2(fp_t a, fp_t b) { const rep_t aAbs = toRep(a) & absMask; @@ -120,34 +133,19 @@ __unorddf2(fp_t a, fp_t b) { return aAbs > infRep || bAbs > infRep; } -// The following are alternative names for the preceding routines. - -COMPILER_RT_ABI enum LE_RESULT -__eqdf2(fp_t a, fp_t b) { - return __ledf2(a, b); -} - -COMPILER_RT_ABI enum LE_RESULT -__ltdf2(fp_t a, fp_t b) { - return __ledf2(a, b); -} - -COMPILER_RT_ABI enum LE_RESULT -__nedf2(fp_t a, fp_t b) { - return __ledf2(a, b); -} - -COMPILER_RT_ABI enum GE_RESULT -__gtdf2(fp_t a, fp_t b) { - return __gedf2(a, b); -} - #if defined(__ARM_EABI__) #if defined(COMPILER_RT_ARMHF_TARGET) -AEABI_RTABI int __aeabi_dcmpun(fp_t a, fp_t b) { - return __unorddf2(a, b); -} +AEABI_RTABI int __aeabi_dcmpun(fp_t a, fp_t b) { return __unorddf2(a, b); } #else -AEABI_RTABI int __aeabi_dcmpun(fp_t a, fp_t b) COMPILER_RT_ALIAS(__unorddf2); +COMPILER_RT_ALIAS(__unorddf2, __aeabi_dcmpun) #endif #endif + +#if defined(_WIN32) && !defined(__MINGW32__) +// The alias mechanism doesn't work on Windows except for MinGW, so emit +// wrapper functions. +int __eqdf2(fp_t a, fp_t b) { return __ledf2(a, b); } +int __ltdf2(fp_t a, fp_t b) { return __ledf2(a, b); } +int __nedf2(fp_t a, fp_t b) { return __ledf2(a, b); } +int __gtdf2(fp_t a, fp_t b) { return __gedf2(a, b); } +#endif diff --git a/lib/builtins/comparesf2.c b/lib/builtins/comparesf2.c index 43cd6a6a7003..1cb99e468c18 100644 --- a/lib/builtins/comparesf2.c +++ b/lib/builtins/comparesf2.c @@ -1,9 +1,8 @@ //===-- lib/comparesf2.c - Single-precision comparisons -----------*- C -*-===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// // @@ -40,79 +39,93 @@ #define SINGLE_PRECISION #include "fp_lib.h" -enum LE_RESULT { - LE_LESS = -1, - LE_EQUAL = 0, - LE_GREATER = 1, - LE_UNORDERED = 1 -}; +enum LE_RESULT { LE_LESS = -1, LE_EQUAL = 0, LE_GREATER = 1, LE_UNORDERED = 1 }; + +COMPILER_RT_ABI enum LE_RESULT __lesf2(fp_t a, fp_t b) { + + const srep_t aInt = toRep(a); + const srep_t bInt = toRep(b); + const rep_t aAbs = aInt & absMask; + const rep_t bAbs = bInt & absMask; + + // If either a or b is NaN, they are unordered. + if (aAbs > infRep || bAbs > infRep) + return LE_UNORDERED; + + // If a and b are both zeros, they are equal. + if ((aAbs | bAbs) == 0) + return LE_EQUAL; -COMPILER_RT_ABI enum LE_RESULT -__lesf2(fp_t a, fp_t b) { - - const srep_t aInt = toRep(a); - const srep_t bInt = toRep(b); - const rep_t aAbs = aInt & absMask; - const rep_t bAbs = bInt & absMask; - - // If either a or b is NaN, they are unordered. - if (aAbs > infRep || bAbs > infRep) return LE_UNORDERED; - - // If a and b are both zeros, they are equal. - if ((aAbs | bAbs) == 0) return LE_EQUAL; - - // If at least one of a and b is positive, we get the same result comparing - // a and b as signed integers as we would with a fp_ting-point compare. - if ((aInt & bInt) >= 0) { - if (aInt < bInt) return LE_LESS; - else if (aInt == bInt) return LE_EQUAL; - else return LE_GREATER; - } - - // Otherwise, both are negative, so we need to flip the sense of the - // comparison to get the correct result. (This assumes a twos- or ones- - // complement integer representation; if integers are represented in a - // sign-magnitude representation, then this flip is incorrect). - else { - if (aInt > bInt) return LE_LESS; - else if (aInt == bInt) return LE_EQUAL; - else return LE_GREATER; - } + // If at least one of a and b is positive, we get the same result comparing + // a and b as signed integers as we would with a fp_ting-point compare. + if ((aInt & bInt) >= 0) { + if (aInt < bInt) + return LE_LESS; + else if (aInt == bInt) + return LE_EQUAL; + else + return LE_GREATER; + } + + // Otherwise, both are negative, so we need to flip the sense of the + // comparison to get the correct result. (This assumes a twos- or ones- + // complement integer representation; if integers are represented in a + // sign-magnitude representation, then this flip is incorrect). + else { + if (aInt > bInt) + return LE_LESS; + else if (aInt == bInt) + return LE_EQUAL; + else + return LE_GREATER; + } } #if defined(__ELF__) // Alias for libgcc compatibility -FNALIAS(__cmpsf2, __lesf2); +COMPILER_RT_ALIAS(__lesf2, __cmpsf2) #endif +COMPILER_RT_ALIAS(__lesf2, __eqsf2) +COMPILER_RT_ALIAS(__lesf2, __ltsf2) +COMPILER_RT_ALIAS(__lesf2, __nesf2) enum GE_RESULT { - GE_LESS = -1, - GE_EQUAL = 0, - GE_GREATER = 1, - GE_UNORDERED = -1 // Note: different from LE_UNORDERED + GE_LESS = -1, + GE_EQUAL = 0, + GE_GREATER = 1, + GE_UNORDERED = -1 // Note: different from LE_UNORDERED }; -COMPILER_RT_ABI enum GE_RESULT -__gesf2(fp_t a, fp_t b) { - - const srep_t aInt = toRep(a); - const srep_t bInt = toRep(b); - const rep_t aAbs = aInt & absMask; - const rep_t bAbs = bInt & absMask; - - if (aAbs > infRep || bAbs > infRep) return GE_UNORDERED; - if ((aAbs | bAbs) == 0) return GE_EQUAL; - if ((aInt & bInt) >= 0) { - if (aInt < bInt) return GE_LESS; - else if (aInt == bInt) return GE_EQUAL; - else return GE_GREATER; - } else { - if (aInt > bInt) return GE_LESS; - else if (aInt == bInt) return GE_EQUAL; - else return GE_GREATER; - } +COMPILER_RT_ABI enum GE_RESULT __gesf2(fp_t a, fp_t b) { + + const srep_t aInt = toRep(a); + const srep_t bInt = toRep(b); + const rep_t aAbs = aInt & absMask; + const rep_t bAbs = bInt & absMask; + + if (aAbs > infRep || bAbs > infRep) + return GE_UNORDERED; + if ((aAbs | bAbs) == 0) + return GE_EQUAL; + if ((aInt & bInt) >= 0) { + if (aInt < bInt) + return GE_LESS; + else if (aInt == bInt) + return GE_EQUAL; + else + return GE_GREATER; + } else { + if (aInt > bInt) + return GE_LESS; + else if (aInt == bInt) + return GE_EQUAL; + else + return GE_GREATER; + } } +COMPILER_RT_ALIAS(__gesf2, __gtsf2) + COMPILER_RT_ABI int __unordsf2(fp_t a, fp_t b) { const rep_t aAbs = toRep(a) & absMask; @@ -120,34 +133,19 @@ __unordsf2(fp_t a, fp_t b) { return aAbs > infRep || bAbs > infRep; } -// The following are alternative names for the preceding routines. - -COMPILER_RT_ABI enum LE_RESULT -__eqsf2(fp_t a, fp_t b) { - return __lesf2(a, b); -} - -COMPILER_RT_ABI enum LE_RESULT -__ltsf2(fp_t a, fp_t b) { - return __lesf2(a, b); -} - -COMPILER_RT_ABI enum LE_RESULT -__nesf2(fp_t a, fp_t b) { - return __lesf2(a, b); -} - -COMPILER_RT_ABI enum GE_RESULT -__gtsf2(fp_t a, fp_t b) { - return __gesf2(a, b); -} - #if defined(__ARM_EABI__) #if defined(COMPILER_RT_ARMHF_TARGET) -AEABI_RTABI int __aeabi_fcmpun(fp_t a, fp_t b) { - return __unordsf2(a, b); -} +AEABI_RTABI int __aeabi_fcmpun(fp_t a, fp_t b) { return __unordsf2(a, b); } #else -AEABI_RTABI int __aeabi_fcmpun(fp_t a, fp_t b) COMPILER_RT_ALIAS(__unordsf2); +COMPILER_RT_ALIAS(__unordsf2, __aeabi_fcmpun) #endif #endif + +#if defined(_WIN32) && !defined(__MINGW32__) +// The alias mechanism doesn't work on Windows except for MinGW, so emit +// wrapper functions. +int __eqsf2(fp_t a, fp_t b) { return __lesf2(a, b); } +int __ltsf2(fp_t a, fp_t b) { return __lesf2(a, b); } +int __nesf2(fp_t a, fp_t b) { return __lesf2(a, b); } +int __gtsf2(fp_t a, fp_t b) { return __gesf2(a, b); } +#endif diff --git a/lib/builtins/comparetf2.c b/lib/builtins/comparetf2.c index c0ad8ed0aecd..2eb34cf37fbc 100644 --- a/lib/builtins/comparetf2.c +++ b/lib/builtins/comparetf2.c @@ -1,9 +1,8 @@ //===-- lib/comparetf2.c - Quad-precision comparisons -------------*- C -*-===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// // @@ -41,98 +40,95 @@ #include "fp_lib.h" #if defined(CRT_HAS_128BIT) && defined(CRT_LDBL_128BIT) -enum LE_RESULT { - LE_LESS = -1, - LE_EQUAL = 0, - LE_GREATER = 1, - LE_UNORDERED = 1 -}; +enum LE_RESULT { LE_LESS = -1, LE_EQUAL = 0, LE_GREATER = 1, LE_UNORDERED = 1 }; COMPILER_RT_ABI enum LE_RESULT __letf2(fp_t a, fp_t b) { - const srep_t aInt = toRep(a); - const srep_t bInt = toRep(b); - const rep_t aAbs = aInt & absMask; - const rep_t bAbs = bInt & absMask; - - // If either a or b is NaN, they are unordered. - if (aAbs > infRep || bAbs > infRep) return LE_UNORDERED; - - // If a and b are both zeros, they are equal. - if ((aAbs | bAbs) == 0) return LE_EQUAL; - - // If at least one of a and b is positive, we get the same result comparing - // a and b as signed integers as we would with a floating-point compare. - if ((aInt & bInt) >= 0) { - if (aInt < bInt) return LE_LESS; - else if (aInt == bInt) return LE_EQUAL; - else return LE_GREATER; - } - else { - // Otherwise, both are negative, so we need to flip the sense of the - // comparison to get the correct result. (This assumes a twos- or ones- - // complement integer representation; if integers are represented in a - // sign-magnitude representation, then this flip is incorrect). - if (aInt > bInt) return LE_LESS; - else if (aInt == bInt) return LE_EQUAL; - else return LE_GREATER; - } + const srep_t aInt = toRep(a); + const srep_t bInt = toRep(b); + const rep_t aAbs = aInt & absMask; + const rep_t bAbs = bInt & absMask; + + // If either a or b is NaN, they are unordered. + if (aAbs > infRep || bAbs > infRep) + return LE_UNORDERED; + + // If a and b are both zeros, they are equal. + if ((aAbs | bAbs) == 0) + return LE_EQUAL; + + // If at least one of a and b is positive, we get the same result comparing + // a and b as signed integers as we would with a floating-point compare. + if ((aInt & bInt) >= 0) { + if (aInt < bInt) + return LE_LESS; + else if (aInt == bInt) + return LE_EQUAL; + else + return LE_GREATER; + } else { + // Otherwise, both are negative, so we need to flip the sense of the + // comparison to get the correct result. (This assumes a twos- or ones- + // complement integer representation; if integers are represented in a + // sign-magnitude representation, then this flip is incorrect). + if (aInt > bInt) + return LE_LESS; + else if (aInt == bInt) + return LE_EQUAL; + else + return LE_GREATER; + } } #if defined(__ELF__) // Alias for libgcc compatibility -FNALIAS(__cmptf2, __letf2); +COMPILER_RT_ALIAS(__letf2, __cmptf2) #endif +COMPILER_RT_ALIAS(__letf2, __eqtf2) +COMPILER_RT_ALIAS(__letf2, __lttf2) +COMPILER_RT_ALIAS(__letf2, __netf2) enum GE_RESULT { - GE_LESS = -1, - GE_EQUAL = 0, - GE_GREATER = 1, - GE_UNORDERED = -1 // Note: different from LE_UNORDERED + GE_LESS = -1, + GE_EQUAL = 0, + GE_GREATER = 1, + GE_UNORDERED = -1 // Note: different from LE_UNORDERED }; COMPILER_RT_ABI enum GE_RESULT __getf2(fp_t a, fp_t b) { - const srep_t aInt = toRep(a); - const srep_t bInt = toRep(b); - const rep_t aAbs = aInt & absMask; - const rep_t bAbs = bInt & absMask; - - if (aAbs > infRep || bAbs > infRep) return GE_UNORDERED; - if ((aAbs | bAbs) == 0) return GE_EQUAL; - if ((aInt & bInt) >= 0) { - if (aInt < bInt) return GE_LESS; - else if (aInt == bInt) return GE_EQUAL; - else return GE_GREATER; - } else { - if (aInt > bInt) return GE_LESS; - else if (aInt == bInt) return GE_EQUAL; - else return GE_GREATER; - } -} - -COMPILER_RT_ABI int __unordtf2(fp_t a, fp_t b) { - const rep_t aAbs = toRep(a) & absMask; - const rep_t bAbs = toRep(b) & absMask; - return aAbs > infRep || bAbs > infRep; -} - -// The following are alternative names for the preceding routines. - -COMPILER_RT_ABI enum LE_RESULT __eqtf2(fp_t a, fp_t b) { - return __letf2(a, b); + const srep_t aInt = toRep(a); + const srep_t bInt = toRep(b); + const rep_t aAbs = aInt & absMask; + const rep_t bAbs = bInt & absMask; + + if (aAbs > infRep || bAbs > infRep) + return GE_UNORDERED; + if ((aAbs | bAbs) == 0) + return GE_EQUAL; + if ((aInt & bInt) >= 0) { + if (aInt < bInt) + return GE_LESS; + else if (aInt == bInt) + return GE_EQUAL; + else + return GE_GREATER; + } else { + if (aInt > bInt) + return GE_LESS; + else if (aInt == bInt) + return GE_EQUAL; + else + return GE_GREATER; + } } -COMPILER_RT_ABI enum LE_RESULT __lttf2(fp_t a, fp_t b) { - return __letf2(a, b); -} - -COMPILER_RT_ABI enum LE_RESULT __netf2(fp_t a, fp_t b) { - return __letf2(a, b); -} +COMPILER_RT_ALIAS(__getf2, __gttf2) -COMPILER_RT_ABI enum GE_RESULT __gttf2(fp_t a, fp_t b) { - return __getf2(a, b); +COMPILER_RT_ABI int __unordtf2(fp_t a, fp_t b) { + const rep_t aAbs = toRep(a) & absMask; + const rep_t bAbs = toRep(b) & absMask; + return aAbs > infRep || bAbs > infRep; } #endif diff --git a/lib/builtins/cpu_model.c b/lib/builtins/cpu_model.c index fb2b899fc70a..f953aed959e5 100644 --- a/lib/builtins/cpu_model.c +++ b/lib/builtins/cpu_model.c @@ -1,9 +1,8 @@ //===-- cpu_model.c - Support for __cpu_model builtin ------------*- C -*-===// // -// The LLVM Compiler Infrastructure -// -// This file is distributed under the University of Illinois Open Source -// License. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// // @@ -13,8 +12,8 @@ // //===----------------------------------------------------------------------===// -#if (defined(__i386__) || defined(_M_IX86) || \ - defined(__x86_64__) || defined(_M_X64)) && \ +#if (defined(__i386__) || defined(_M_IX86) || defined(__x86_64__) || \ + defined(_M_X64)) && \ (defined(__GNUC__) || defined(__clang__) || defined(_MSC_VER)) #include <assert.h> @@ -32,8 +31,8 @@ #endif enum VendorSignatures { - SIG_INTEL = 0x756e6547 /* Genu */, - SIG_AMD = 0x68747541 /* Auth */ + SIG_INTEL = 0x756e6547, // Genu + SIG_AMD = 0x68747541, // Auth }; enum ProcessorVendors { @@ -81,6 +80,8 @@ enum ProcessorSubtypes { INTEL_COREI7_CANNONLAKE, INTEL_COREI7_ICELAKE_CLIENT, INTEL_COREI7_ICELAKE_SERVER, + AMDFAM17H_ZNVER2, + INTEL_COREI7_CASCADELAKE, CPU_SUBTYPE_MAX }; @@ -266,10 +267,11 @@ static void detectX86FamilyModel(unsigned EAX, unsigned *Family, } } -static void -getIntelProcessorTypeAndSubtype(unsigned Family, unsigned Model, - unsigned Brand_id, unsigned Features, - unsigned *Type, unsigned *Subtype) { +static void getIntelProcessorTypeAndSubtype(unsigned Family, unsigned Model, + unsigned Brand_id, + unsigned Features, + unsigned Features2, unsigned *Type, + unsigned *Subtype) { if (Brand_id != 0) return; switch (Family) { @@ -295,7 +297,7 @@ getIntelProcessorTypeAndSubtype(unsigned Family, unsigned Model, case 0x1e: // Intel(R) Core(TM) i7 CPU 870 @ 2.93GHz. // As found in a Summer 2010 model iMac. case 0x1f: - case 0x2e: // Nehalem EX + case 0x2e: // Nehalem EX *Type = INTEL_COREI7; // "nehalem" *Subtype = INTEL_COREI7_NEHALEM; break; @@ -313,7 +315,7 @@ getIntelProcessorTypeAndSubtype(unsigned Family, unsigned Model, *Subtype = INTEL_COREI7_SANDYBRIDGE; break; case 0x3a: - case 0x3e: // Ivy Bridge EP + case 0x3e: // Ivy Bridge EP *Type = INTEL_COREI7; // "ivybridge" *Subtype = INTEL_COREI7_IVYBRIDGE; break; @@ -337,10 +339,10 @@ getIntelProcessorTypeAndSubtype(unsigned Family, unsigned Model, break; // Skylake: - case 0x4e: // Skylake mobile - case 0x5e: // Skylake desktop - case 0x8e: // Kaby Lake mobile - case 0x9e: // Kaby Lake desktop + case 0x4e: // Skylake mobile + case 0x5e: // Skylake desktop + case 0x8e: // Kaby Lake mobile + case 0x9e: // Kaby Lake desktop *Type = INTEL_COREI7; // "skylake" *Subtype = INTEL_COREI7_SKYLAKE; break; @@ -348,7 +350,10 @@ getIntelProcessorTypeAndSubtype(unsigned Family, unsigned Model, // Skylake Xeon: case 0x55: *Type = INTEL_COREI7; - *Subtype = INTEL_COREI7_SKYLAKE_AVX512; // "skylake-avx512" + if (Features2 & (1 << (FEATURE_AVX512VNNI - 32))) + *Subtype = INTEL_COREI7_CASCADELAKE; // "cascadelake" + else + *Subtype = INTEL_COREI7_SKYLAKE_AVX512; // "skylake-avx512" break; // Cannonlake: @@ -357,6 +362,20 @@ getIntelProcessorTypeAndSubtype(unsigned Family, unsigned Model, *Subtype = INTEL_COREI7_CANNONLAKE; // "cannonlake" break; + // Icelake: + case 0x7d: + case 0x7e: + *Type = INTEL_COREI7; + *Subtype = INTEL_COREI7_ICELAKE_CLIENT; // "icelake-client" + break; + + // Icelake Xeon: + case 0x6a: + case 0x6c: + *Type = INTEL_COREI7; + *Subtype = INTEL_COREI7_ICELAKE_SERVER; // "icelake-server" + break; + case 0x1c: // Most 45 nm Intel Atom processors case 0x26: // 45 nm Atom Lincroft case 0x27: // 32 nm Atom Medfield @@ -382,6 +401,9 @@ getIntelProcessorTypeAndSubtype(unsigned Family, unsigned Model, case 0x7a: *Type = INTEL_GOLDMONT_PLUS; break; + case 0x86: + *Type = INTEL_TREMONT; + break; case 0x57: *Type = INTEL_KNL; // knl @@ -393,7 +415,7 @@ getIntelProcessorTypeAndSubtype(unsigned Family, unsigned Model, default: // Unknown family 6 CPU. break; - break; + break; } default: break; // Unknown. @@ -401,8 +423,8 @@ getIntelProcessorTypeAndSubtype(unsigned Family, unsigned Model, } static void getAMDProcessorTypeAndSubtype(unsigned Family, unsigned Model, - unsigned Features, unsigned *Type, - unsigned *Subtype) { + unsigned Features, unsigned Features2, + unsigned *Type, unsigned *Subtype) { // FIXME: this poorly matches the generated SubtargetFeatureKV table. There // appears to be no way to generate the wide variety of AMD-specific targets // from the information returned from CPUID. @@ -448,7 +470,14 @@ static void getAMDProcessorTypeAndSubtype(unsigned Family, unsigned Model, break; // "btver2" case 23: *Type = AMDFAM17H; - *Subtype = AMDFAM17H_ZNVER1; + if (Model >= 0x30 && Model <= 0x3f) { + *Subtype = AMDFAM17H_ZNVER2; + break; // "znver2"; 30h-3fh: Zen2 + } + if (Model <= 0x0f) { + *Subtype = AMDFAM17H_ZNVER1; + break; // "znver1"; 00h-0Fh: Zen1 + } break; default: break; // "generic" @@ -462,12 +491,12 @@ static void getAvailableFeatures(unsigned ECX, unsigned EDX, unsigned MaxLeaf, unsigned Features2 = 0; unsigned EAX, EBX; -#define setFeature(F) \ - do { \ - if (F < 32) \ - Features |= 1U << (F & 0x1f); \ - else if (F < 64) \ - Features2 |= 1U << ((F - 32) & 0x1f); \ +#define setFeature(F) \ + do { \ + if (F < 32) \ + Features |= 1U << (F & 0x1f); \ + else if (F < 64) \ + Features2 |= 1U << ((F - 32) & 0x1f); \ } while (0) if ((EDX >> 15) & 1) @@ -580,24 +609,33 @@ static void getAvailableFeatures(unsigned ECX, unsigned EDX, unsigned MaxLeaf, #define CONSTRUCTOR_ATTRIBUTE #endif +#ifndef _WIN32 +__attribute__((visibility("hidden"))) +#endif int __cpu_indicator_init(void) CONSTRUCTOR_ATTRIBUTE; +#ifndef _WIN32 +__attribute__((visibility("hidden"))) +#endif struct __processor_model { unsigned int __cpu_vendor; unsigned int __cpu_type; unsigned int __cpu_subtype; unsigned int __cpu_features[1]; } __cpu_model = {0, 0, 0, {0}}; + +#ifndef _WIN32 +__attribute__((visibility("hidden"))) +#endif unsigned int __cpu_features2; -/* A constructor function that is sets __cpu_model and __cpu_features2 with - the right values. This needs to run only once. This constructor is - given the highest priority and it should run before constructors without - the priority set. However, it still runs after ifunc initializers and - needs to be called explicitly there. */ +// A constructor function that is sets __cpu_model and __cpu_features2 with +// the right values. This needs to run only once. This constructor is +// given the highest priority and it should run before constructors without +// the priority set. However, it still runs after ifunc initializers and +// needs to be called explicitly there. -int CONSTRUCTOR_ATTRIBUTE -__cpu_indicator_init(void) { +int CONSTRUCTOR_ATTRIBUTE __cpu_indicator_init(void) { unsigned EAX, EBX, ECX, EDX; unsigned MaxLeaf = 5; unsigned Vendor; @@ -605,14 +643,14 @@ __cpu_indicator_init(void) { unsigned Features = 0; unsigned Features2 = 0; - /* This function needs to run just once. */ + // This function needs to run just once. if (__cpu_model.__cpu_vendor) return 0; if (!isCpuIdSupported()) return -1; - /* Assume cpuid insn present. Run in level 0 to get vendor id. */ + // Assume cpuid insn present. Run in level 0 to get vendor id. if (getX86CpuIDAndInfo(0, &MaxLeaf, &Vendor, &ECX, &EDX) || MaxLeaf < 1) { __cpu_model.__cpu_vendor = VENDOR_OTHER; return -1; @@ -621,20 +659,20 @@ __cpu_indicator_init(void) { detectX86FamilyModel(EAX, &Family, &Model); Brand_id = EBX & 0xff; - /* Find available features. */ + // Find available features. getAvailableFeatures(ECX, EDX, MaxLeaf, &Features, &Features2); __cpu_model.__cpu_features[0] = Features; __cpu_features2 = Features2; if (Vendor == SIG_INTEL) { - /* Get CPU type. */ + // Get CPU type. getIntelProcessorTypeAndSubtype(Family, Model, Brand_id, Features, - &(__cpu_model.__cpu_type), + Features2, &(__cpu_model.__cpu_type), &(__cpu_model.__cpu_subtype)); __cpu_model.__cpu_vendor = VENDOR_INTEL; } else if (Vendor == SIG_AMD) { - /* Get CPU type. */ - getAMDProcessorTypeAndSubtype(Family, Model, Features, + // Get CPU type. + getAMDProcessorTypeAndSubtype(Family, Model, Features, Features2, &(__cpu_model.__cpu_type), &(__cpu_model.__cpu_subtype)); __cpu_model.__cpu_vendor = VENDOR_AMD; diff --git a/lib/builtins/ctzdi2.c b/lib/builtins/ctzdi2.c index ef6d7fea1365..9384aa6055a1 100644 --- a/lib/builtins/ctzdi2.c +++ b/lib/builtins/ctzdi2.c @@ -1,40 +1,35 @@ -/* ===-- ctzdi2.c - Implement __ctzdi2 -------------------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - * - * This file implements __ctzdi2 for the compiler_rt library. - * - * ===----------------------------------------------------------------------=== - */ +//===-- ctzdi2.c - Implement __ctzdi2 -------------------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements __ctzdi2 for the compiler_rt library. +// +//===----------------------------------------------------------------------===// #include "int_lib.h" -/* Returns: the number of trailing 0-bits */ +// Returns: the number of trailing 0-bits #if !defined(__clang__) && \ - ((defined(__sparc__) && defined(__arch64__)) || \ - defined(__mips64) || \ + ((defined(__sparc__) && defined(__arch64__)) || defined(__mips64) || \ (defined(__riscv) && __SIZEOF_POINTER__ >= 8)) -/* On 64-bit architectures with neither a native clz instruction nor a native - * ctz instruction, gcc resolves __builtin_ctz to __ctzdi2 rather than - * __ctzsi2, leading to infinite recursion. */ +// On 64-bit architectures with neither a native clz instruction nor a native +// ctz instruction, gcc resolves __builtin_ctz to __ctzdi2 rather than +// __ctzsi2, leading to infinite recursion. #define __builtin_ctz(a) __ctzsi2(a) extern si_int __ctzsi2(si_int); #endif -/* Precondition: a != 0 */ +// Precondition: a != 0 -COMPILER_RT_ABI si_int -__ctzdi2(di_int a) -{ - dwords x; - x.all = a; - const si_int f = -(x.s.low == 0); - return __builtin_ctz((x.s.high & f) | (x.s.low & ~f)) + - (f & ((si_int)(sizeof(si_int) * CHAR_BIT))); +COMPILER_RT_ABI si_int __ctzdi2(di_int a) { + dwords x; + x.all = a; + const si_int f = -(x.s.low == 0); + return __builtin_ctz((x.s.high & f) | (x.s.low & ~f)) + + (f & ((si_int)(sizeof(si_int) * CHAR_BIT))); } diff --git a/lib/builtins/ctzsi2.c b/lib/builtins/ctzsi2.c index c69486ea445f..09c6863b74e3 100644 --- a/lib/builtins/ctzsi2.c +++ b/lib/builtins/ctzsi2.c @@ -1,57 +1,53 @@ -/* ===-- ctzsi2.c - Implement __ctzsi2 -------------------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - * - * This file implements __ctzsi2 for the compiler_rt library. - * - * ===----------------------------------------------------------------------=== - */ +//===-- ctzsi2.c - Implement __ctzsi2 -------------------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements __ctzsi2 for the compiler_rt library. +// +//===----------------------------------------------------------------------===// #include "int_lib.h" -/* Returns: the number of trailing 0-bits */ +// Returns: the number of trailing 0-bits -/* Precondition: a != 0 */ +// Precondition: a != 0 -COMPILER_RT_ABI si_int -__ctzsi2(si_int a) -{ - su_int x = (su_int)a; - si_int t = ((x & 0x0000FFFF) == 0) << 4; /* if (x has no small bits) t = 16 else 0 */ - x >>= t; /* x = [0 - 0xFFFF] + higher garbage bits */ - su_int r = t; /* r = [0, 16] */ - /* return r + ctz(x) */ - t = ((x & 0x00FF) == 0) << 3; - x >>= t; /* x = [0 - 0xFF] + higher garbage bits */ - r += t; /* r = [0, 8, 16, 24] */ - /* return r + ctz(x) */ - t = ((x & 0x0F) == 0) << 2; - x >>= t; /* x = [0 - 0xF] + higher garbage bits */ - r += t; /* r = [0, 4, 8, 12, 16, 20, 24, 28] */ - /* return r + ctz(x) */ - t = ((x & 0x3) == 0) << 1; - x >>= t; - x &= 3; /* x = [0 - 3] */ - r += t; /* r = [0 - 30] and is even */ - /* return r + ctz(x) */ +COMPILER_RT_ABI si_int __ctzsi2(si_int a) { + su_int x = (su_int)a; + si_int t = ((x & 0x0000FFFF) == 0) + << 4; // if (x has no small bits) t = 16 else 0 + x >>= t; // x = [0 - 0xFFFF] + higher garbage bits + su_int r = t; // r = [0, 16] + // return r + ctz(x) + t = ((x & 0x00FF) == 0) << 3; + x >>= t; // x = [0 - 0xFF] + higher garbage bits + r += t; // r = [0, 8, 16, 24] + // return r + ctz(x) + t = ((x & 0x0F) == 0) << 2; + x >>= t; // x = [0 - 0xF] + higher garbage bits + r += t; // r = [0, 4, 8, 12, 16, 20, 24, 28] + // return r + ctz(x) + t = ((x & 0x3) == 0) << 1; + x >>= t; + x &= 3; // x = [0 - 3] + r += t; // r = [0 - 30] and is even + // return r + ctz(x) -/* The branch-less return statement below is equivalent - * to the following switch statement: - * switch (x) - * { - * case 0: - * return r + 2; - * case 2: - * return r + 1; - * case 1: - * case 3: - * return r; - * } - */ - return r + ((2 - (x >> 1)) & -((x & 1) == 0)); + // The branch-less return statement below is equivalent + // to the following switch statement: + // switch (x) + // { + // case 0: + // return r + 2; + // case 2: + // return r + 1; + // case 1: + // case 3: + // return r; + // } + return r + ((2 - (x >> 1)) & -((x & 1) == 0)); } diff --git a/lib/builtins/ctzti2.c b/lib/builtins/ctzti2.c index 45de682700cc..2a1312c8437d 100644 --- a/lib/builtins/ctzti2.c +++ b/lib/builtins/ctzti2.c @@ -1,33 +1,29 @@ -/* ===-- ctzti2.c - Implement __ctzti2 -------------------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - * - * This file implements __ctzti2 for the compiler_rt library. - * - * ===----------------------------------------------------------------------=== - */ +//===-- ctzti2.c - Implement __ctzti2 -------------------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements __ctzti2 for the compiler_rt library. +// +//===----------------------------------------------------------------------===// #include "int_lib.h" #ifdef CRT_HAS_128BIT -/* Returns: the number of trailing 0-bits */ +// Returns: the number of trailing 0-bits -/* Precondition: a != 0 */ +// Precondition: a != 0 -COMPILER_RT_ABI si_int -__ctzti2(ti_int a) -{ - twords x; - x.all = a; - const di_int f = -(x.s.low == 0); - return __builtin_ctzll((x.s.high & f) | (x.s.low & ~f)) + - ((si_int)f & ((si_int)(sizeof(di_int) * CHAR_BIT))); +COMPILER_RT_ABI si_int __ctzti2(ti_int a) { + twords x; + x.all = a; + const di_int f = -(x.s.low == 0); + return __builtin_ctzll((x.s.high & f) | (x.s.low & ~f)) + + ((si_int)f & ((si_int)(sizeof(di_int) * CHAR_BIT))); } -#endif /* CRT_HAS_128BIT */ +#endif // CRT_HAS_128BIT diff --git a/lib/builtins/divdc3.c b/lib/builtins/divdc3.c index 392d6ecacde1..c2cf62874603 100644 --- a/lib/builtins/divdc3.c +++ b/lib/builtins/divdc3.c @@ -1,62 +1,53 @@ -/* ===-- divdc3.c - Implement __divdc3 -------------------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - * - * This file implements __divdc3 for the compiler_rt library. - * - * ===----------------------------------------------------------------------=== - */ +//===-- divdc3.c - Implement __divdc3 -------------------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements __divdc3 for the compiler_rt library. +// +//===----------------------------------------------------------------------===// #define DOUBLE_PRECISION #include "fp_lib.h" #include "int_lib.h" #include "int_math.h" -/* Returns: the quotient of (a + ib) / (c + id) */ +// Returns: the quotient of (a + ib) / (c + id) -COMPILER_RT_ABI Dcomplex -__divdc3(double __a, double __b, double __c, double __d) -{ - int __ilogbw = 0; - double __logbw = __compiler_rt_logb(crt_fmax(crt_fabs(__c), crt_fabs(__d))); - if (crt_isfinite(__logbw)) - { - __ilogbw = (int)__logbw; - __c = crt_scalbn(__c, -__ilogbw); - __d = crt_scalbn(__d, -__ilogbw); +COMPILER_RT_ABI Dcomplex __divdc3(double __a, double __b, double __c, + double __d) { + int __ilogbw = 0; + double __logbw = __compiler_rt_logb(crt_fmax(crt_fabs(__c), crt_fabs(__d))); + if (crt_isfinite(__logbw)) { + __ilogbw = (int)__logbw; + __c = crt_scalbn(__c, -__ilogbw); + __d = crt_scalbn(__d, -__ilogbw); + } + double __denom = __c * __c + __d * __d; + Dcomplex z; + COMPLEX_REAL(z) = crt_scalbn((__a * __c + __b * __d) / __denom, -__ilogbw); + COMPLEX_IMAGINARY(z) = + crt_scalbn((__b * __c - __a * __d) / __denom, -__ilogbw); + if (crt_isnan(COMPLEX_REAL(z)) && crt_isnan(COMPLEX_IMAGINARY(z))) { + if ((__denom == 0.0) && (!crt_isnan(__a) || !crt_isnan(__b))) { + COMPLEX_REAL(z) = crt_copysign(CRT_INFINITY, __c) * __a; + COMPLEX_IMAGINARY(z) = crt_copysign(CRT_INFINITY, __c) * __b; + } else if ((crt_isinf(__a) || crt_isinf(__b)) && crt_isfinite(__c) && + crt_isfinite(__d)) { + __a = crt_copysign(crt_isinf(__a) ? 1.0 : 0.0, __a); + __b = crt_copysign(crt_isinf(__b) ? 1.0 : 0.0, __b); + COMPLEX_REAL(z) = CRT_INFINITY * (__a * __c + __b * __d); + COMPLEX_IMAGINARY(z) = CRT_INFINITY * (__b * __c - __a * __d); + } else if (crt_isinf(__logbw) && __logbw > 0.0 && crt_isfinite(__a) && + crt_isfinite(__b)) { + __c = crt_copysign(crt_isinf(__c) ? 1.0 : 0.0, __c); + __d = crt_copysign(crt_isinf(__d) ? 1.0 : 0.0, __d); + COMPLEX_REAL(z) = 0.0 * (__a * __c + __b * __d); + COMPLEX_IMAGINARY(z) = 0.0 * (__b * __c - __a * __d); } - double __denom = __c * __c + __d * __d; - Dcomplex z; - COMPLEX_REAL(z) = crt_scalbn((__a * __c + __b * __d) / __denom, -__ilogbw); - COMPLEX_IMAGINARY(z) = crt_scalbn((__b * __c - __a * __d) / __denom, -__ilogbw); - if (crt_isnan(COMPLEX_REAL(z)) && crt_isnan(COMPLEX_IMAGINARY(z))) - { - if ((__denom == 0.0) && (!crt_isnan(__a) || !crt_isnan(__b))) - { - COMPLEX_REAL(z) = crt_copysign(CRT_INFINITY, __c) * __a; - COMPLEX_IMAGINARY(z) = crt_copysign(CRT_INFINITY, __c) * __b; - } - else if ((crt_isinf(__a) || crt_isinf(__b)) && - crt_isfinite(__c) && crt_isfinite(__d)) - { - __a = crt_copysign(crt_isinf(__a) ? 1.0 : 0.0, __a); - __b = crt_copysign(crt_isinf(__b) ? 1.0 : 0.0, __b); - COMPLEX_REAL(z) = CRT_INFINITY * (__a * __c + __b * __d); - COMPLEX_IMAGINARY(z) = CRT_INFINITY * (__b * __c - __a * __d); - } - else if (crt_isinf(__logbw) && __logbw > 0.0 && - crt_isfinite(__a) && crt_isfinite(__b)) - { - __c = crt_copysign(crt_isinf(__c) ? 1.0 : 0.0, __c); - __d = crt_copysign(crt_isinf(__d) ? 1.0 : 0.0, __d); - COMPLEX_REAL(z) = 0.0 * (__a * __c + __b * __d); - COMPLEX_IMAGINARY(z) = 0.0 * (__b * __c - __a * __d); - } - } - return z; + } + return z; } diff --git a/lib/builtins/divdf3.c b/lib/builtins/divdf3.c index 411c82ebb87a..1dea3b534f5a 100644 --- a/lib/builtins/divdf3.c +++ b/lib/builtins/divdf3.c @@ -1,9 +1,8 @@ //===-- lib/divdf3.c - Double-precision division ------------------*- C -*-===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// // @@ -19,175 +18,193 @@ #define DOUBLE_PRECISION #include "fp_lib.h" -COMPILER_RT_ABI fp_t -__divdf3(fp_t a, fp_t b) { - - const unsigned int aExponent = toRep(a) >> significandBits & maxExponent; - const unsigned int bExponent = toRep(b) >> significandBits & maxExponent; - const rep_t quotientSign = (toRep(a) ^ toRep(b)) & signBit; - - rep_t aSignificand = toRep(a) & significandMask; - rep_t bSignificand = toRep(b) & significandMask; - int scale = 0; - - // Detect if a or b is zero, denormal, infinity, or NaN. - if (aExponent-1U >= maxExponent-1U || bExponent-1U >= maxExponent-1U) { - - const rep_t aAbs = toRep(a) & absMask; - const rep_t bAbs = toRep(b) & absMask; - - // NaN / anything = qNaN - if (aAbs > infRep) return fromRep(toRep(a) | quietBit); - // anything / NaN = qNaN - if (bAbs > infRep) return fromRep(toRep(b) | quietBit); - - if (aAbs == infRep) { - // infinity / infinity = NaN - if (bAbs == infRep) return fromRep(qnanRep); - // infinity / anything else = +/- infinity - else return fromRep(aAbs | quotientSign); - } - - // anything else / infinity = +/- 0 - if (bAbs == infRep) return fromRep(quotientSign); - - if (!aAbs) { - // zero / zero = NaN - if (!bAbs) return fromRep(qnanRep); - // zero / anything else = +/- zero - else return fromRep(quotientSign); - } - // anything else / zero = +/- infinity - if (!bAbs) return fromRep(infRep | quotientSign); - - // one or both of a or b is denormal, the other (if applicable) is a - // normal number. Renormalize one or both of a and b, and set scale to - // include the necessary exponent adjustment. - if (aAbs < implicitBit) scale += normalize(&aSignificand); - if (bAbs < implicitBit) scale -= normalize(&bSignificand); - } - - // Or in the implicit significand bit. (If we fell through from the - // denormal path it was already set by normalize( ), but setting it twice - // won't hurt anything.) - aSignificand |= implicitBit; - bSignificand |= implicitBit; - int quotientExponent = aExponent - bExponent + scale; - - // Align the significand of b as a Q31 fixed-point number in the range - // [1, 2.0) and get a Q32 approximate reciprocal using a small minimax - // polynomial approximation: reciprocal = 3/4 + 1/sqrt(2) - b/2. This - // is accurate to about 3.5 binary digits. - const uint32_t q31b = bSignificand >> 21; - uint32_t recip32 = UINT32_C(0x7504f333) - q31b; - - // Now refine the reciprocal estimate using a Newton-Raphson iteration: - // - // x1 = x0 * (2 - x0 * b) - // - // This doubles the number of correct binary digits in the approximation - // with each iteration, so after three iterations, we have about 28 binary - // digits of accuracy. - uint32_t correction32; - correction32 = -((uint64_t)recip32 * q31b >> 32); - recip32 = (uint64_t)recip32 * correction32 >> 31; - correction32 = -((uint64_t)recip32 * q31b >> 32); - recip32 = (uint64_t)recip32 * correction32 >> 31; - correction32 = -((uint64_t)recip32 * q31b >> 32); - recip32 = (uint64_t)recip32 * correction32 >> 31; - - // recip32 might have overflowed to exactly zero in the preceding - // computation if the high word of b is exactly 1.0. This would sabotage - // the full-width final stage of the computation that follows, so we adjust - // recip32 downward by one bit. - recip32--; - - // We need to perform one more iteration to get us to 56 binary digits; - // The last iteration needs to happen with extra precision. - const uint32_t q63blo = bSignificand << 11; - uint64_t correction, reciprocal; - correction = -((uint64_t)recip32*q31b + ((uint64_t)recip32*q63blo >> 32)); - uint32_t cHi = correction >> 32; - uint32_t cLo = correction; - reciprocal = (uint64_t)recip32*cHi + ((uint64_t)recip32*cLo >> 32); - - // We already adjusted the 32-bit estimate, now we need to adjust the final - // 64-bit reciprocal estimate downward to ensure that it is strictly smaller - // than the infinitely precise exact reciprocal. Because the computation - // of the Newton-Raphson step is truncating at every step, this adjustment - // is small; most of the work is already done. - reciprocal -= 2; - - // The numerical reciprocal is accurate to within 2^-56, lies in the - // interval [0.5, 1.0), and is strictly smaller than the true reciprocal - // of b. Multiplying a by this reciprocal thus gives a numerical q = a/b - // in Q53 with the following properties: - // - // 1. q < a/b - // 2. q is in the interval [0.5, 2.0) - // 3. the error in q is bounded away from 2^-53 (actually, we have a - // couple of bits to spare, but this is all we need). - - // We need a 64 x 64 multiply high to compute q, which isn't a basic - // operation in C, so we need to be a little bit fussy. - rep_t quotient, quotientLo; - wideMultiply(aSignificand << 2, reciprocal, "ient, "ientLo); - - // Two cases: quotient is in [0.5, 1.0) or quotient is in [1.0, 2.0). - // In either case, we are going to compute a residual of the form - // - // r = a - q*b - // - // We know from the construction of q that r satisfies: - // - // 0 <= r < ulp(q)*b - // - // if r is greater than 1/2 ulp(q)*b, then q rounds up. Otherwise, we - // already have the correct result. The exact halfway case cannot occur. - // We also take this time to right shift quotient if it falls in the [1,2) - // range and adjust the exponent accordingly. - rep_t residual; - if (quotient < (implicitBit << 1)) { - residual = (aSignificand << 53) - quotient * bSignificand; - quotientExponent--; - } else { - quotient >>= 1; - residual = (aSignificand << 52) - quotient * bSignificand; +COMPILER_RT_ABI fp_t __divdf3(fp_t a, fp_t b) { + + const unsigned int aExponent = toRep(a) >> significandBits & maxExponent; + const unsigned int bExponent = toRep(b) >> significandBits & maxExponent; + const rep_t quotientSign = (toRep(a) ^ toRep(b)) & signBit; + + rep_t aSignificand = toRep(a) & significandMask; + rep_t bSignificand = toRep(b) & significandMask; + int scale = 0; + + // Detect if a or b is zero, denormal, infinity, or NaN. + if (aExponent - 1U >= maxExponent - 1U || + bExponent - 1U >= maxExponent - 1U) { + + const rep_t aAbs = toRep(a) & absMask; + const rep_t bAbs = toRep(b) & absMask; + + // NaN / anything = qNaN + if (aAbs > infRep) + return fromRep(toRep(a) | quietBit); + // anything / NaN = qNaN + if (bAbs > infRep) + return fromRep(toRep(b) | quietBit); + + if (aAbs == infRep) { + // infinity / infinity = NaN + if (bAbs == infRep) + return fromRep(qnanRep); + // infinity / anything else = +/- infinity + else + return fromRep(aAbs | quotientSign); } - const int writtenExponent = quotientExponent + exponentBias; + // anything else / infinity = +/- 0 + if (bAbs == infRep) + return fromRep(quotientSign); - if (writtenExponent >= maxExponent) { - // If we have overflowed the exponent, return infinity. - return fromRep(infRep | quotientSign); - } - - else if (writtenExponent < 1) { - // Flush denormals to zero. In the future, it would be nice to add - // code to round them correctly. + if (!aAbs) { + // zero / zero = NaN + if (!bAbs) + return fromRep(qnanRep); + // zero / anything else = +/- zero + else return fromRep(quotientSign); } - - else { - const bool round = (residual << 1) > bSignificand; - // Clear the implicit bit - rep_t absResult = quotient & significandMask; - // Insert the exponent - absResult |= (rep_t)writtenExponent << significandBits; - // Round - absResult += round; - // Insert the sign and return - const double result = fromRep(absResult | quotientSign); - return result; + // anything else / zero = +/- infinity + if (!bAbs) + return fromRep(infRep | quotientSign); + + // One or both of a or b is denormal. The other (if applicable) is a + // normal number. Renormalize one or both of a and b, and set scale to + // include the necessary exponent adjustment. + if (aAbs < implicitBit) + scale += normalize(&aSignificand); + if (bAbs < implicitBit) + scale -= normalize(&bSignificand); + } + + // Set the implicit significand bit. If we fell through from the + // denormal path it was already set by normalize( ), but setting it twice + // won't hurt anything. + aSignificand |= implicitBit; + bSignificand |= implicitBit; + int quotientExponent = aExponent - bExponent + scale; + + // Align the significand of b as a Q31 fixed-point number in the range + // [1, 2.0) and get a Q32 approximate reciprocal using a small minimax + // polynomial approximation: reciprocal = 3/4 + 1/sqrt(2) - b/2. This + // is accurate to about 3.5 binary digits. + const uint32_t q31b = bSignificand >> 21; + uint32_t recip32 = UINT32_C(0x7504f333) - q31b; + // 0x7504F333 / 2^32 + 1 = 3/4 + 1/sqrt(2) + + // Now refine the reciprocal estimate using a Newton-Raphson iteration: + // + // x1 = x0 * (2 - x0 * b) + // + // This doubles the number of correct binary digits in the approximation + // with each iteration. + uint32_t correction32; + correction32 = -((uint64_t)recip32 * q31b >> 32); + recip32 = (uint64_t)recip32 * correction32 >> 31; + correction32 = -((uint64_t)recip32 * q31b >> 32); + recip32 = (uint64_t)recip32 * correction32 >> 31; + correction32 = -((uint64_t)recip32 * q31b >> 32); + recip32 = (uint64_t)recip32 * correction32 >> 31; + + // The reciprocal may have overflowed to zero if the upper half of b is + // exactly 1.0. This would sabatoge the full-width final stage of the + // computation that follows, so we adjust the reciprocal down by one bit. + recip32--; + + // We need to perform one more iteration to get us to 56 binary digits. + // The last iteration needs to happen with extra precision. + const uint32_t q63blo = bSignificand << 11; + uint64_t correction, reciprocal; + correction = -((uint64_t)recip32 * q31b + ((uint64_t)recip32 * q63blo >> 32)); + uint32_t cHi = correction >> 32; + uint32_t cLo = correction; + reciprocal = (uint64_t)recip32 * cHi + ((uint64_t)recip32 * cLo >> 32); + + // Adjust the final 64-bit reciprocal estimate downward to ensure that it is + // strictly smaller than the infinitely precise exact reciprocal. Because + // the computation of the Newton-Raphson step is truncating at every step, + // this adjustment is small; most of the work is already done. + reciprocal -= 2; + + // The numerical reciprocal is accurate to within 2^-56, lies in the + // interval [0.5, 1.0), and is strictly smaller than the true reciprocal + // of b. Multiplying a by this reciprocal thus gives a numerical q = a/b + // in Q53 with the following properties: + // + // 1. q < a/b + // 2. q is in the interval [0.5, 2.0) + // 3. The error in q is bounded away from 2^-53 (actually, we have a + // couple of bits to spare, but this is all we need). + + // We need a 64 x 64 multiply high to compute q, which isn't a basic + // operation in C, so we need to be a little bit fussy. + rep_t quotient, quotientLo; + wideMultiply(aSignificand << 2, reciprocal, "ient, "ientLo); + + // Two cases: quotient is in [0.5, 1.0) or quotient is in [1.0, 2.0). + // In either case, we are going to compute a residual of the form + // + // r = a - q*b + // + // We know from the construction of q that r satisfies: + // + // 0 <= r < ulp(q)*b + // + // If r is greater than 1/2 ulp(q)*b, then q rounds up. Otherwise, we + // already have the correct result. The exact halfway case cannot occur. + // We also take this time to right shift quotient if it falls in the [1,2) + // range and adjust the exponent accordingly. + rep_t residual; + if (quotient < (implicitBit << 1)) { + residual = (aSignificand << 53) - quotient * bSignificand; + quotientExponent--; + } else { + quotient >>= 1; + residual = (aSignificand << 52) - quotient * bSignificand; + } + + const int writtenExponent = quotientExponent + exponentBias; + + if (writtenExponent >= maxExponent) { + // If we have overflowed the exponent, return infinity. + return fromRep(infRep | quotientSign); + } + + else if (writtenExponent < 1) { + if (writtenExponent == 0) { + // Check whether the rounded result is normal. + const bool round = (residual << 1) > bSignificand; + // Clear the implicit bit. + rep_t absResult = quotient & significandMask; + // Round. + absResult += round; + if (absResult & ~significandMask) { + // The rounded result is normal; return it. + return fromRep(absResult | quotientSign); + } } + // Flush denormals to zero. In the future, it would be nice to add + // code to round them correctly. + return fromRep(quotientSign); + } + + else { + const bool round = (residual << 1) > bSignificand; + // Clear the implicit bit. + rep_t absResult = quotient & significandMask; + // Insert the exponent. + absResult |= (rep_t)writtenExponent << significandBits; + // Round. + absResult += round; + // Insert the sign and return. + const double result = fromRep(absResult | quotientSign); + return result; + } } #if defined(__ARM_EABI__) #if defined(COMPILER_RT_ARMHF_TARGET) -AEABI_RTABI fp_t __aeabi_ddiv(fp_t a, fp_t b) { - return __divdf3(a, b); -} +AEABI_RTABI fp_t __aeabi_ddiv(fp_t a, fp_t b) { return __divdf3(a, b); } #else -AEABI_RTABI fp_t __aeabi_ddiv(fp_t a, fp_t b) COMPILER_RT_ALIAS(__divdf3); +COMPILER_RT_ALIAS(__divdf3, __aeabi_ddiv) #endif #endif diff --git a/lib/builtins/divdi3.c b/lib/builtins/divdi3.c index b8eebcb20465..ee08d6557783 100644 --- a/lib/builtins/divdi3.c +++ b/lib/builtins/divdi3.c @@ -1,29 +1,25 @@ -/* ===-- divdi3.c - Implement __divdi3 -------------------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - * - * This file implements __divdi3 for the compiler_rt library. - * - * ===----------------------------------------------------------------------=== - */ +//===-- divdi3.c - Implement __divdi3 -------------------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements __divdi3 for the compiler_rt library. +// +//===----------------------------------------------------------------------===// #include "int_lib.h" -/* Returns: a / b */ +// Returns: a / b -COMPILER_RT_ABI di_int -__divdi3(di_int a, di_int b) -{ - const int bits_in_dword_m1 = (int)(sizeof(di_int) * CHAR_BIT) - 1; - di_int s_a = a >> bits_in_dword_m1; /* s_a = a < 0 ? -1 : 0 */ - di_int s_b = b >> bits_in_dword_m1; /* s_b = b < 0 ? -1 : 0 */ - a = (a ^ s_a) - s_a; /* negate if s_a == -1 */ - b = (b ^ s_b) - s_b; /* negate if s_b == -1 */ - s_a ^= s_b; /*sign of quotient */ - return (__udivmoddi4(a, b, (du_int*)0) ^ s_a) - s_a; /* negate if s_a == -1 */ +COMPILER_RT_ABI di_int __divdi3(di_int a, di_int b) { + const int bits_in_dword_m1 = (int)(sizeof(di_int) * CHAR_BIT) - 1; + di_int s_a = a >> bits_in_dword_m1; // s_a = a < 0 ? -1 : 0 + di_int s_b = b >> bits_in_dword_m1; // s_b = b < 0 ? -1 : 0 + a = (a ^ s_a) - s_a; // negate if s_a == -1 + b = (b ^ s_b) - s_b; // negate if s_b == -1 + s_a ^= s_b; // sign of quotient + return (__udivmoddi4(a, b, (du_int *)0) ^ s_a) - s_a; // negate if s_a == -1 } diff --git a/lib/builtins/divmoddi4.c b/lib/builtins/divmoddi4.c index 0d4df67a63e0..7f333510c003 100644 --- a/lib/builtins/divmoddi4.c +++ b/lib/builtins/divmoddi4.c @@ -1,25 +1,21 @@ -/*===-- divmoddi4.c - Implement __divmoddi4 --------------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - * - * This file implements __divmoddi4 for the compiler_rt library. - * - * ===----------------------------------------------------------------------=== - */ +//===-- divmoddi4.c - Implement __divmoddi4 -------------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements __divmoddi4 for the compiler_rt library. +// +//===----------------------------------------------------------------------===// #include "int_lib.h" -/* Returns: a / b, *rem = a % b */ +// Returns: a / b, *rem = a % b -COMPILER_RT_ABI di_int -__divmoddi4(di_int a, di_int b, di_int* rem) -{ - di_int d = __divdi3(a,b); - *rem = a - (d*b); +COMPILER_RT_ABI di_int __divmoddi4(di_int a, di_int b, di_int *rem) { + di_int d = __divdi3(a, b); + *rem = a - (d * b); return d; } diff --git a/lib/builtins/divmodsi4.c b/lib/builtins/divmodsi4.c index dabe2874397c..402eed22fe7a 100644 --- a/lib/builtins/divmodsi4.c +++ b/lib/builtins/divmodsi4.c @@ -1,27 +1,22 @@ -/*===-- divmodsi4.c - Implement __divmodsi4 --------------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - * - * This file implements __divmodsi4 for the compiler_rt library. - * - * ===----------------------------------------------------------------------=== - */ +//===-- divmodsi4.c - Implement __divmodsi4 +//--------------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements __divmodsi4 for the compiler_rt library. +// +//===----------------------------------------------------------------------===// #include "int_lib.h" -/* Returns: a / b, *rem = a % b */ +// Returns: a / b, *rem = a % b -COMPILER_RT_ABI si_int -__divmodsi4(si_int a, si_int b, si_int* rem) -{ - si_int d = __divsi3(a,b); - *rem = a - (d*b); - return d; +COMPILER_RT_ABI si_int __divmodsi4(si_int a, si_int b, si_int *rem) { + si_int d = __divsi3(a, b); + *rem = a - (d * b); + return d; } - - diff --git a/lib/builtins/divsc3.c b/lib/builtins/divsc3.c index 0d18a256c3de..1a63634dde21 100644 --- a/lib/builtins/divsc3.c +++ b/lib/builtins/divsc3.c @@ -1,63 +1,53 @@ -/*===-- divsc3.c - Implement __divsc3 -------------------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - * - * This file implements __divsc3 for the compiler_rt library. - * - *===----------------------------------------------------------------------=== - */ +//===-- divsc3.c - Implement __divsc3 -------------------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements __divsc3 for the compiler_rt library. +// +//===----------------------------------------------------------------------===// #define SINGLE_PRECISION #include "fp_lib.h" #include "int_lib.h" #include "int_math.h" -/* Returns: the quotient of (a + ib) / (c + id) */ +// Returns: the quotient of (a + ib) / (c + id) -COMPILER_RT_ABI Fcomplex -__divsc3(float __a, float __b, float __c, float __d) -{ - int __ilogbw = 0; - float __logbw = - __compiler_rt_logbf(crt_fmaxf(crt_fabsf(__c), crt_fabsf(__d))); - if (crt_isfinite(__logbw)) - { - __ilogbw = (int)__logbw; - __c = crt_scalbnf(__c, -__ilogbw); - __d = crt_scalbnf(__d, -__ilogbw); +COMPILER_RT_ABI Fcomplex __divsc3(float __a, float __b, float __c, float __d) { + int __ilogbw = 0; + float __logbw = + __compiler_rt_logbf(crt_fmaxf(crt_fabsf(__c), crt_fabsf(__d))); + if (crt_isfinite(__logbw)) { + __ilogbw = (int)__logbw; + __c = crt_scalbnf(__c, -__ilogbw); + __d = crt_scalbnf(__d, -__ilogbw); + } + float __denom = __c * __c + __d * __d; + Fcomplex z; + COMPLEX_REAL(z) = crt_scalbnf((__a * __c + __b * __d) / __denom, -__ilogbw); + COMPLEX_IMAGINARY(z) = + crt_scalbnf((__b * __c - __a * __d) / __denom, -__ilogbw); + if (crt_isnan(COMPLEX_REAL(z)) && crt_isnan(COMPLEX_IMAGINARY(z))) { + if ((__denom == 0) && (!crt_isnan(__a) || !crt_isnan(__b))) { + COMPLEX_REAL(z) = crt_copysignf(CRT_INFINITY, __c) * __a; + COMPLEX_IMAGINARY(z) = crt_copysignf(CRT_INFINITY, __c) * __b; + } else if ((crt_isinf(__a) || crt_isinf(__b)) && crt_isfinite(__c) && + crt_isfinite(__d)) { + __a = crt_copysignf(crt_isinf(__a) ? 1 : 0, __a); + __b = crt_copysignf(crt_isinf(__b) ? 1 : 0, __b); + COMPLEX_REAL(z) = CRT_INFINITY * (__a * __c + __b * __d); + COMPLEX_IMAGINARY(z) = CRT_INFINITY * (__b * __c - __a * __d); + } else if (crt_isinf(__logbw) && __logbw > 0 && crt_isfinite(__a) && + crt_isfinite(__b)) { + __c = crt_copysignf(crt_isinf(__c) ? 1 : 0, __c); + __d = crt_copysignf(crt_isinf(__d) ? 1 : 0, __d); + COMPLEX_REAL(z) = 0 * (__a * __c + __b * __d); + COMPLEX_IMAGINARY(z) = 0 * (__b * __c - __a * __d); } - float __denom = __c * __c + __d * __d; - Fcomplex z; - COMPLEX_REAL(z) = crt_scalbnf((__a * __c + __b * __d) / __denom, -__ilogbw); - COMPLEX_IMAGINARY(z) = crt_scalbnf((__b * __c - __a * __d) / __denom, -__ilogbw); - if (crt_isnan(COMPLEX_REAL(z)) && crt_isnan(COMPLEX_IMAGINARY(z))) - { - if ((__denom == 0) && (!crt_isnan(__a) || !crt_isnan(__b))) - { - COMPLEX_REAL(z) = crt_copysignf(CRT_INFINITY, __c) * __a; - COMPLEX_IMAGINARY(z) = crt_copysignf(CRT_INFINITY, __c) * __b; - } - else if ((crt_isinf(__a) || crt_isinf(__b)) && - crt_isfinite(__c) && crt_isfinite(__d)) - { - __a = crt_copysignf(crt_isinf(__a) ? 1 : 0, __a); - __b = crt_copysignf(crt_isinf(__b) ? 1 : 0, __b); - COMPLEX_REAL(z) = CRT_INFINITY * (__a * __c + __b * __d); - COMPLEX_IMAGINARY(z) = CRT_INFINITY * (__b * __c - __a * __d); - } - else if (crt_isinf(__logbw) && __logbw > 0 && - crt_isfinite(__a) && crt_isfinite(__b)) - { - __c = crt_copysignf(crt_isinf(__c) ? 1 : 0, __c); - __d = crt_copysignf(crt_isinf(__d) ? 1 : 0, __d); - COMPLEX_REAL(z) = 0 * (__a * __c + __b * __d); - COMPLEX_IMAGINARY(z) = 0 * (__b * __c - __a * __d); - } - } - return z; + } + return z; } diff --git a/lib/builtins/divsf3.c b/lib/builtins/divsf3.c index a74917fd1de5..593f93b45ac2 100644 --- a/lib/builtins/divsf3.c +++ b/lib/builtins/divsf3.c @@ -1,9 +1,8 @@ //===-- lib/divsf3.c - Single-precision division ------------------*- C -*-===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// // @@ -19,159 +18,177 @@ #define SINGLE_PRECISION #include "fp_lib.h" -COMPILER_RT_ABI fp_t -__divsf3(fp_t a, fp_t b) { - - const unsigned int aExponent = toRep(a) >> significandBits & maxExponent; - const unsigned int bExponent = toRep(b) >> significandBits & maxExponent; - const rep_t quotientSign = (toRep(a) ^ toRep(b)) & signBit; - - rep_t aSignificand = toRep(a) & significandMask; - rep_t bSignificand = toRep(b) & significandMask; - int scale = 0; - - // Detect if a or b is zero, denormal, infinity, or NaN. - if (aExponent-1U >= maxExponent-1U || bExponent-1U >= maxExponent-1U) { - - const rep_t aAbs = toRep(a) & absMask; - const rep_t bAbs = toRep(b) & absMask; - - // NaN / anything = qNaN - if (aAbs > infRep) return fromRep(toRep(a) | quietBit); - // anything / NaN = qNaN - if (bAbs > infRep) return fromRep(toRep(b) | quietBit); - - if (aAbs == infRep) { - // infinity / infinity = NaN - if (bAbs == infRep) return fromRep(qnanRep); - // infinity / anything else = +/- infinity - else return fromRep(aAbs | quotientSign); - } - - // anything else / infinity = +/- 0 - if (bAbs == infRep) return fromRep(quotientSign); - - if (!aAbs) { - // zero / zero = NaN - if (!bAbs) return fromRep(qnanRep); - // zero / anything else = +/- zero - else return fromRep(quotientSign); - } - // anything else / zero = +/- infinity - if (!bAbs) return fromRep(infRep | quotientSign); - - // one or both of a or b is denormal, the other (if applicable) is a - // normal number. Renormalize one or both of a and b, and set scale to - // include the necessary exponent adjustment. - if (aAbs < implicitBit) scale += normalize(&aSignificand); - if (bAbs < implicitBit) scale -= normalize(&bSignificand); - } - - // Or in the implicit significand bit. (If we fell through from the - // denormal path it was already set by normalize( ), but setting it twice - // won't hurt anything.) - aSignificand |= implicitBit; - bSignificand |= implicitBit; - int quotientExponent = aExponent - bExponent + scale; - - // Align the significand of b as a Q31 fixed-point number in the range - // [1, 2.0) and get a Q32 approximate reciprocal using a small minimax - // polynomial approximation: reciprocal = 3/4 + 1/sqrt(2) - b/2. This - // is accurate to about 3.5 binary digits. - uint32_t q31b = bSignificand << 8; - uint32_t reciprocal = UINT32_C(0x7504f333) - q31b; - - // Now refine the reciprocal estimate using a Newton-Raphson iteration: - // - // x1 = x0 * (2 - x0 * b) - // - // This doubles the number of correct binary digits in the approximation - // with each iteration, so after three iterations, we have about 28 binary - // digits of accuracy. - uint32_t correction; - correction = -((uint64_t)reciprocal * q31b >> 32); - reciprocal = (uint64_t)reciprocal * correction >> 31; - correction = -((uint64_t)reciprocal * q31b >> 32); - reciprocal = (uint64_t)reciprocal * correction >> 31; - correction = -((uint64_t)reciprocal * q31b >> 32); - reciprocal = (uint64_t)reciprocal * correction >> 31; - - // Exhaustive testing shows that the error in reciprocal after three steps - // is in the interval [-0x1.f58108p-31, 0x1.d0e48cp-29], in line with our - // expectations. We bump the reciprocal by a tiny value to force the error - // to be strictly positive (in the range [0x1.4fdfp-37,0x1.287246p-29], to - // be specific). This also causes 1/1 to give a sensible approximation - // instead of zero (due to overflow). - reciprocal -= 2; - - // The numerical reciprocal is accurate to within 2^-28, lies in the - // interval [0x1.000000eep-1, 0x1.fffffffcp-1], and is strictly smaller - // than the true reciprocal of b. Multiplying a by this reciprocal thus - // gives a numerical q = a/b in Q24 with the following properties: - // - // 1. q < a/b - // 2. q is in the interval [0x1.000000eep-1, 0x1.fffffffcp0) - // 3. the error in q is at most 2^-24 + 2^-27 -- the 2^24 term comes - // from the fact that we truncate the product, and the 2^27 term - // is the error in the reciprocal of b scaled by the maximum - // possible value of a. As a consequence of this error bound, - // either q or nextafter(q) is the correctly rounded - rep_t quotient = (uint64_t)reciprocal*(aSignificand << 1) >> 32; - - // Two cases: quotient is in [0.5, 1.0) or quotient is in [1.0, 2.0). - // In either case, we are going to compute a residual of the form - // - // r = a - q*b - // - // We know from the construction of q that r satisfies: - // - // 0 <= r < ulp(q)*b - // - // if r is greater than 1/2 ulp(q)*b, then q rounds up. Otherwise, we - // already have the correct result. The exact halfway case cannot occur. - // We also take this time to right shift quotient if it falls in the [1,2) - // range and adjust the exponent accordingly. - rep_t residual; - if (quotient < (implicitBit << 1)) { - residual = (aSignificand << 24) - quotient * bSignificand; - quotientExponent--; - } else { - quotient >>= 1; - residual = (aSignificand << 23) - quotient * bSignificand; +COMPILER_RT_ABI fp_t __divsf3(fp_t a, fp_t b) { + + const unsigned int aExponent = toRep(a) >> significandBits & maxExponent; + const unsigned int bExponent = toRep(b) >> significandBits & maxExponent; + const rep_t quotientSign = (toRep(a) ^ toRep(b)) & signBit; + + rep_t aSignificand = toRep(a) & significandMask; + rep_t bSignificand = toRep(b) & significandMask; + int scale = 0; + + // Detect if a or b is zero, denormal, infinity, or NaN. + if (aExponent - 1U >= maxExponent - 1U || + bExponent - 1U >= maxExponent - 1U) { + + const rep_t aAbs = toRep(a) & absMask; + const rep_t bAbs = toRep(b) & absMask; + + // NaN / anything = qNaN + if (aAbs > infRep) + return fromRep(toRep(a) | quietBit); + // anything / NaN = qNaN + if (bAbs > infRep) + return fromRep(toRep(b) | quietBit); + + if (aAbs == infRep) { + // infinity / infinity = NaN + if (bAbs == infRep) + return fromRep(qnanRep); + // infinity / anything else = +/- infinity + else + return fromRep(aAbs | quotientSign); } - const int writtenExponent = quotientExponent + exponentBias; + // anything else / infinity = +/- 0 + if (bAbs == infRep) + return fromRep(quotientSign); - if (writtenExponent >= maxExponent) { - // If we have overflowed the exponent, return infinity. - return fromRep(infRep | quotientSign); - } - - else if (writtenExponent < 1) { - // Flush denormals to zero. In the future, it would be nice to add - // code to round them correctly. + if (!aAbs) { + // zero / zero = NaN + if (!bAbs) + return fromRep(qnanRep); + // zero / anything else = +/- zero + else return fromRep(quotientSign); } - - else { - const bool round = (residual << 1) > bSignificand; - // Clear the implicit bit - rep_t absResult = quotient & significandMask; - // Insert the exponent - absResult |= (rep_t)writtenExponent << significandBits; - // Round - absResult += round; - // Insert the sign and return + // anything else / zero = +/- infinity + if (!bAbs) + return fromRep(infRep | quotientSign); + + // One or both of a or b is denormal. The other (if applicable) is a + // normal number. Renormalize one or both of a and b, and set scale to + // include the necessary exponent adjustment. + if (aAbs < implicitBit) + scale += normalize(&aSignificand); + if (bAbs < implicitBit) + scale -= normalize(&bSignificand); + } + + // Set the implicit significand bit. If we fell through from the + // denormal path it was already set by normalize( ), but setting it twice + // won't hurt anything. + aSignificand |= implicitBit; + bSignificand |= implicitBit; + int quotientExponent = aExponent - bExponent + scale; + // 0x7504F333 / 2^32 + 1 = 3/4 + 1/sqrt(2) + + // Align the significand of b as a Q31 fixed-point number in the range + // [1, 2.0) and get a Q32 approximate reciprocal using a small minimax + // polynomial approximation: reciprocal = 3/4 + 1/sqrt(2) - b/2. This + // is accurate to about 3.5 binary digits. + uint32_t q31b = bSignificand << 8; + uint32_t reciprocal = UINT32_C(0x7504f333) - q31b; + + // Now refine the reciprocal estimate using a Newton-Raphson iteration: + // + // x1 = x0 * (2 - x0 * b) + // + // This doubles the number of correct binary digits in the approximation + // with each iteration. + uint32_t correction; + correction = -((uint64_t)reciprocal * q31b >> 32); + reciprocal = (uint64_t)reciprocal * correction >> 31; + correction = -((uint64_t)reciprocal * q31b >> 32); + reciprocal = (uint64_t)reciprocal * correction >> 31; + correction = -((uint64_t)reciprocal * q31b >> 32); + reciprocal = (uint64_t)reciprocal * correction >> 31; + + // Adust the final 32-bit reciprocal estimate downward to ensure that it is + // strictly smaller than the infinitely precise exact reciprocal. Because + // the computation of the Newton-Raphson step is truncating at every step, + // this adjustment is small; most of the work is already done. + reciprocal -= 2; + + // The numerical reciprocal is accurate to within 2^-28, lies in the + // interval [0x1.000000eep-1, 0x1.fffffffcp-1], and is strictly smaller + // than the true reciprocal of b. Multiplying a by this reciprocal thus + // gives a numerical q = a/b in Q24 with the following properties: + // + // 1. q < a/b + // 2. q is in the interval [0x1.000000eep-1, 0x1.fffffffcp0) + // 3. The error in q is at most 2^-24 + 2^-27 -- the 2^24 term comes + // from the fact that we truncate the product, and the 2^27 term + // is the error in the reciprocal of b scaled by the maximum + // possible value of a. As a consequence of this error bound, + // either q or nextafter(q) is the correctly rounded. + rep_t quotient = (uint64_t)reciprocal * (aSignificand << 1) >> 32; + + // Two cases: quotient is in [0.5, 1.0) or quotient is in [1.0, 2.0). + // In either case, we are going to compute a residual of the form + // + // r = a - q*b + // + // We know from the construction of q that r satisfies: + // + // 0 <= r < ulp(q)*b + // + // If r is greater than 1/2 ulp(q)*b, then q rounds up. Otherwise, we + // already have the correct result. The exact halfway case cannot occur. + // We also take this time to right shift quotient if it falls in the [1,2) + // range and adjust the exponent accordingly. + rep_t residual; + if (quotient < (implicitBit << 1)) { + residual = (aSignificand << 24) - quotient * bSignificand; + quotientExponent--; + } else { + quotient >>= 1; + residual = (aSignificand << 23) - quotient * bSignificand; + } + + const int writtenExponent = quotientExponent + exponentBias; + + if (writtenExponent >= maxExponent) { + // If we have overflowed the exponent, return infinity. + return fromRep(infRep | quotientSign); + } + + else if (writtenExponent < 1) { + if (writtenExponent == 0) { + // Check whether the rounded result is normal. + const bool round = (residual << 1) > bSignificand; + // Clear the implicit bit. + rep_t absResult = quotient & significandMask; + // Round. + absResult += round; + if (absResult & ~significandMask) { + // The rounded result is normal; return it. return fromRep(absResult | quotientSign); + } } + // Flush denormals to zero. In the future, it would be nice to add + // code to round them correctly. + return fromRep(quotientSign); + } + + else { + const bool round = (residual << 1) > bSignificand; + // Clear the implicit bit. + rep_t absResult = quotient & significandMask; + // Insert the exponent. + absResult |= (rep_t)writtenExponent << significandBits; + // Round. + absResult += round; + // Insert the sign and return. + return fromRep(absResult | quotientSign); + } } #if defined(__ARM_EABI__) #if defined(COMPILER_RT_ARMHF_TARGET) -AEABI_RTABI fp_t __aeabi_fdiv(fp_t a, fp_t b) { - return __divsf3(a, b); -} +AEABI_RTABI fp_t __aeabi_fdiv(fp_t a, fp_t b) { return __divsf3(a, b); } #else -AEABI_RTABI fp_t __aeabi_fdiv(fp_t a, fp_t b) COMPILER_RT_ALIAS(__divsf3); +COMPILER_RT_ALIAS(__divsf3, __aeabi_fdiv) #endif #endif diff --git a/lib/builtins/divsi3.c b/lib/builtins/divsi3.c index 75aea008ddc1..b97e11119f0e 100644 --- a/lib/builtins/divsi3.c +++ b/lib/builtins/divsi3.c @@ -1,39 +1,35 @@ -/* ===-- divsi3.c - Implement __divsi3 -------------------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - * - * This file implements __divsi3 for the compiler_rt library. - * - * ===----------------------------------------------------------------------=== - */ +//===-- divsi3.c - Implement __divsi3 -------------------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements __divsi3 for the compiler_rt library. +// +//===----------------------------------------------------------------------===// #include "int_lib.h" -/* Returns: a / b */ +// Returns: a / b -COMPILER_RT_ABI si_int -__divsi3(si_int a, si_int b) -{ - const int bits_in_word_m1 = (int)(sizeof(si_int) * CHAR_BIT) - 1; - si_int s_a = a >> bits_in_word_m1; /* s_a = a < 0 ? -1 : 0 */ - si_int s_b = b >> bits_in_word_m1; /* s_b = b < 0 ? -1 : 0 */ - a = (a ^ s_a) - s_a; /* negate if s_a == -1 */ - b = (b ^ s_b) - s_b; /* negate if s_b == -1 */ - s_a ^= s_b; /* sign of quotient */ - /* - * On CPUs without unsigned hardware division support, - * this calls __udivsi3 (notice the cast to su_int). - * On CPUs with unsigned hardware division support, - * this uses the unsigned division instruction. - */ - return ((su_int)a/(su_int)b ^ s_a) - s_a; /* negate if s_a == -1 */ +COMPILER_RT_ABI si_int __divsi3(si_int a, si_int b) { + const int bits_in_word_m1 = (int)(sizeof(si_int) * CHAR_BIT) - 1; + si_int s_a = a >> bits_in_word_m1; // s_a = a < 0 ? -1 : 0 + si_int s_b = b >> bits_in_word_m1; // s_b = b < 0 ? -1 : 0 + a = (a ^ s_a) - s_a; // negate if s_a == -1 + b = (b ^ s_b) - s_b; // negate if s_b == -1 + s_a ^= s_b; // sign of quotient + // + // On CPUs without unsigned hardware division support, + // this calls __udivsi3 (notice the cast to su_int). + // On CPUs with unsigned hardware division support, + // this uses the unsigned division instruction. + // + return ((su_int)a / (su_int)b ^ s_a) - s_a; // negate if s_a == -1 } #if defined(__ARM_EABI__) -AEABI_RTABI si_int __aeabi_idiv(si_int a, si_int b) COMPILER_RT_ALIAS(__divsi3); +COMPILER_RT_ALIAS(__divsi3, __aeabi_idiv) #endif diff --git a/lib/builtins/divtc3.c b/lib/builtins/divtc3.c index e5ea00d841e3..37c71400e370 100644 --- a/lib/builtins/divtc3.c +++ b/lib/builtins/divtc3.c @@ -1,63 +1,54 @@ -/*===-- divtc3.c - Implement __divtc3 -------------------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - * - * This file implements __divtc3 for the compiler_rt library. - * - *===----------------------------------------------------------------------=== - */ +//===-- divtc3.c - Implement __divtc3 -------------------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements __divtc3 for the compiler_rt library. +// +//===----------------------------------------------------------------------===// #define QUAD_PRECISION #include "fp_lib.h" #include "int_lib.h" #include "int_math.h" -/* Returns: the quotient of (a + ib) / (c + id) */ +// Returns: the quotient of (a + ib) / (c + id) -COMPILER_RT_ABI Lcomplex -__divtc3(long double __a, long double __b, long double __c, long double __d) -{ - int __ilogbw = 0; - long double __logbw = - __compiler_rt_logbl(crt_fmaxl(crt_fabsl(__c), crt_fabsl(__d))); - if (crt_isfinite(__logbw)) - { - __ilogbw = (int)__logbw; - __c = crt_scalbnl(__c, -__ilogbw); - __d = crt_scalbnl(__d, -__ilogbw); +COMPILER_RT_ABI Lcomplex __divtc3(long double __a, long double __b, + long double __c, long double __d) { + int __ilogbw = 0; + long double __logbw = + __compiler_rt_logbl(crt_fmaxl(crt_fabsl(__c), crt_fabsl(__d))); + if (crt_isfinite(__logbw)) { + __ilogbw = (int)__logbw; + __c = crt_scalbnl(__c, -__ilogbw); + __d = crt_scalbnl(__d, -__ilogbw); + } + long double __denom = __c * __c + __d * __d; + Lcomplex z; + COMPLEX_REAL(z) = crt_scalbnl((__a * __c + __b * __d) / __denom, -__ilogbw); + COMPLEX_IMAGINARY(z) = + crt_scalbnl((__b * __c - __a * __d) / __denom, -__ilogbw); + if (crt_isnan(COMPLEX_REAL(z)) && crt_isnan(COMPLEX_IMAGINARY(z))) { + if ((__denom == 0.0) && (!crt_isnan(__a) || !crt_isnan(__b))) { + COMPLEX_REAL(z) = crt_copysignl(CRT_INFINITY, __c) * __a; + COMPLEX_IMAGINARY(z) = crt_copysignl(CRT_INFINITY, __c) * __b; + } else if ((crt_isinf(__a) || crt_isinf(__b)) && crt_isfinite(__c) && + crt_isfinite(__d)) { + __a = crt_copysignl(crt_isinf(__a) ? 1.0 : 0.0, __a); + __b = crt_copysignl(crt_isinf(__b) ? 1.0 : 0.0, __b); + COMPLEX_REAL(z) = CRT_INFINITY * (__a * __c + __b * __d); + COMPLEX_IMAGINARY(z) = CRT_INFINITY * (__b * __c - __a * __d); + } else if (crt_isinf(__logbw) && __logbw > 0.0 && crt_isfinite(__a) && + crt_isfinite(__b)) { + __c = crt_copysignl(crt_isinf(__c) ? 1.0 : 0.0, __c); + __d = crt_copysignl(crt_isinf(__d) ? 1.0 : 0.0, __d); + COMPLEX_REAL(z) = 0.0 * (__a * __c + __b * __d); + COMPLEX_IMAGINARY(z) = 0.0 * (__b * __c - __a * __d); } - long double __denom = __c * __c + __d * __d; - Lcomplex z; - COMPLEX_REAL(z) = crt_scalbnl((__a * __c + __b * __d) / __denom, -__ilogbw); - COMPLEX_IMAGINARY(z) = crt_scalbnl((__b * __c - __a * __d) / __denom, -__ilogbw); - if (crt_isnan(COMPLEX_REAL(z)) && crt_isnan(COMPLEX_IMAGINARY(z))) - { - if ((__denom == 0.0) && (!crt_isnan(__a) || !crt_isnan(__b))) - { - COMPLEX_REAL(z) = crt_copysignl(CRT_INFINITY, __c) * __a; - COMPLEX_IMAGINARY(z) = crt_copysignl(CRT_INFINITY, __c) * __b; - } - else if ((crt_isinf(__a) || crt_isinf(__b)) && - crt_isfinite(__c) && crt_isfinite(__d)) - { - __a = crt_copysignl(crt_isinf(__a) ? 1.0 : 0.0, __a); - __b = crt_copysignl(crt_isinf(__b) ? 1.0 : 0.0, __b); - COMPLEX_REAL(z) = CRT_INFINITY * (__a * __c + __b * __d); - COMPLEX_IMAGINARY(z) = CRT_INFINITY * (__b * __c - __a * __d); - } - else if (crt_isinf(__logbw) && __logbw > 0.0 && - crt_isfinite(__a) && crt_isfinite(__b)) - { - __c = crt_copysignl(crt_isinf(__c) ? 1.0 : 0.0, __c); - __d = crt_copysignl(crt_isinf(__d) ? 1.0 : 0.0, __d); - COMPLEX_REAL(z) = 0.0 * (__a * __c + __b * __d); - COMPLEX_IMAGINARY(z) = 0.0 * (__b * __c - __a * __d); - } - } - return z; + } + return z; } diff --git a/lib/builtins/divtf3.c b/lib/builtins/divtf3.c index e81dab826bdd..6e61d2e31b75 100644 --- a/lib/builtins/divtf3.c +++ b/lib/builtins/divtf3.c @@ -1,9 +1,8 @@ //===-- lib/divtf3.c - Quad-precision division --------------------*- C -*-===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// // @@ -22,182 +21,201 @@ #if defined(CRT_HAS_128BIT) && defined(CRT_LDBL_128BIT) COMPILER_RT_ABI fp_t __divtf3(fp_t a, fp_t b) { - const unsigned int aExponent = toRep(a) >> significandBits & maxExponent; - const unsigned int bExponent = toRep(b) >> significandBits & maxExponent; - const rep_t quotientSign = (toRep(a) ^ toRep(b)) & signBit; - - rep_t aSignificand = toRep(a) & significandMask; - rep_t bSignificand = toRep(b) & significandMask; - int scale = 0; - - // Detect if a or b is zero, denormal, infinity, or NaN. - if (aExponent-1U >= maxExponent-1U || bExponent-1U >= maxExponent-1U) { - - const rep_t aAbs = toRep(a) & absMask; - const rep_t bAbs = toRep(b) & absMask; - - // NaN / anything = qNaN - if (aAbs > infRep) return fromRep(toRep(a) | quietBit); - // anything / NaN = qNaN - if (bAbs > infRep) return fromRep(toRep(b) | quietBit); - - if (aAbs == infRep) { - // infinity / infinity = NaN - if (bAbs == infRep) return fromRep(qnanRep); - // infinity / anything else = +/- infinity - else return fromRep(aAbs | quotientSign); - } - - // anything else / infinity = +/- 0 - if (bAbs == infRep) return fromRep(quotientSign); - - if (!aAbs) { - // zero / zero = NaN - if (!bAbs) return fromRep(qnanRep); - // zero / anything else = +/- zero - else return fromRep(quotientSign); - } - // anything else / zero = +/- infinity - if (!bAbs) return fromRep(infRep | quotientSign); - - // one or both of a or b is denormal, the other (if applicable) is a - // normal number. Renormalize one or both of a and b, and set scale to - // include the necessary exponent adjustment. - if (aAbs < implicitBit) scale += normalize(&aSignificand); - if (bAbs < implicitBit) scale -= normalize(&bSignificand); - } - - // Or in the implicit significand bit. (If we fell through from the - // denormal path it was already set by normalize( ), but setting it twice - // won't hurt anything.) - aSignificand |= implicitBit; - bSignificand |= implicitBit; - int quotientExponent = aExponent - bExponent + scale; - - // Align the significand of b as a Q63 fixed-point number in the range - // [1, 2.0) and get a Q64 approximate reciprocal using a small minimax - // polynomial approximation: reciprocal = 3/4 + 1/sqrt(2) - b/2. This - // is accurate to about 3.5 binary digits. - const uint64_t q63b = bSignificand >> 49; - uint64_t recip64 = UINT64_C(0x7504f333F9DE6484) - q63b; - // 0x7504f333F9DE6484 / 2^64 + 1 = 3/4 + 1/sqrt(2) - - // Now refine the reciprocal estimate using a Newton-Raphson iteration: - // - // x1 = x0 * (2 - x0 * b) - // - // This doubles the number of correct binary digits in the approximation - // with each iteration. - uint64_t correction64; - correction64 = -((rep_t)recip64 * q63b >> 64); - recip64 = (rep_t)recip64 * correction64 >> 63; - correction64 = -((rep_t)recip64 * q63b >> 64); - recip64 = (rep_t)recip64 * correction64 >> 63; - correction64 = -((rep_t)recip64 * q63b >> 64); - recip64 = (rep_t)recip64 * correction64 >> 63; - correction64 = -((rep_t)recip64 * q63b >> 64); - recip64 = (rep_t)recip64 * correction64 >> 63; - correction64 = -((rep_t)recip64 * q63b >> 64); - recip64 = (rep_t)recip64 * correction64 >> 63; - - // recip64 might have overflowed to exactly zero in the preceeding - // computation if the high word of b is exactly 1.0. This would sabotage - // the full-width final stage of the computation that follows, so we adjust - // recip64 downward by one bit. - recip64--; - - // We need to perform one more iteration to get us to 112 binary digits; - // The last iteration needs to happen with extra precision. - const uint64_t q127blo = bSignificand << 15; - rep_t correction, reciprocal; - - // NOTE: This operation is equivalent to __multi3, which is not implemented - // in some architechure - rep_t r64q63, r64q127, r64cH, r64cL, dummy; - wideMultiply((rep_t)recip64, (rep_t)q63b, &dummy, &r64q63); - wideMultiply((rep_t)recip64, (rep_t)q127blo, &dummy, &r64q127); - - correction = -(r64q63 + (r64q127 >> 64)); - - uint64_t cHi = correction >> 64; - uint64_t cLo = correction; - - wideMultiply((rep_t)recip64, (rep_t)cHi, &dummy, &r64cH); - wideMultiply((rep_t)recip64, (rep_t)cLo, &dummy, &r64cL); - - reciprocal = r64cH + (r64cL >> 64); - - // We already adjusted the 64-bit estimate, now we need to adjust the final - // 128-bit reciprocal estimate downward to ensure that it is strictly smaller - // than the infinitely precise exact reciprocal. Because the computation - // of the Newton-Raphson step is truncating at every step, this adjustment - // is small; most of the work is already done. - reciprocal -= 2; - - // The numerical reciprocal is accurate to within 2^-112, lies in the - // interval [0.5, 1.0), and is strictly smaller than the true reciprocal - // of b. Multiplying a by this reciprocal thus gives a numerical q = a/b - // in Q127 with the following properties: - // - // 1. q < a/b - // 2. q is in the interval [0.5, 2.0) - // 3. the error in q is bounded away from 2^-113 (actually, we have a - // couple of bits to spare, but this is all we need). - - // We need a 128 x 128 multiply high to compute q, which isn't a basic - // operation in C, so we need to be a little bit fussy. - rep_t quotient, quotientLo; - wideMultiply(aSignificand << 2, reciprocal, "ient, "ientLo); - - // Two cases: quotient is in [0.5, 1.0) or quotient is in [1.0, 2.0). - // In either case, we are going to compute a residual of the form - // - // r = a - q*b - // - // We know from the construction of q that r satisfies: - // - // 0 <= r < ulp(q)*b - // - // if r is greater than 1/2 ulp(q)*b, then q rounds up. Otherwise, we - // already have the correct result. The exact halfway case cannot occur. - // We also take this time to right shift quotient if it falls in the [1,2) - // range and adjust the exponent accordingly. - rep_t residual; - rep_t qb; - - if (quotient < (implicitBit << 1)) { - wideMultiply(quotient, bSignificand, &dummy, &qb); - residual = (aSignificand << 113) - qb; - quotientExponent--; - } else { - quotient >>= 1; - wideMultiply(quotient, bSignificand, &dummy, &qb); - residual = (aSignificand << 112) - qb; + const unsigned int aExponent = toRep(a) >> significandBits & maxExponent; + const unsigned int bExponent = toRep(b) >> significandBits & maxExponent; + const rep_t quotientSign = (toRep(a) ^ toRep(b)) & signBit; + + rep_t aSignificand = toRep(a) & significandMask; + rep_t bSignificand = toRep(b) & significandMask; + int scale = 0; + + // Detect if a or b is zero, denormal, infinity, or NaN. + if (aExponent - 1U >= maxExponent - 1U || + bExponent - 1U >= maxExponent - 1U) { + + const rep_t aAbs = toRep(a) & absMask; + const rep_t bAbs = toRep(b) & absMask; + + // NaN / anything = qNaN + if (aAbs > infRep) + return fromRep(toRep(a) | quietBit); + // anything / NaN = qNaN + if (bAbs > infRep) + return fromRep(toRep(b) | quietBit); + + if (aAbs == infRep) { + // infinity / infinity = NaN + if (bAbs == infRep) + return fromRep(qnanRep); + // infinity / anything else = +/- infinity + else + return fromRep(aAbs | quotientSign); } - const int writtenExponent = quotientExponent + exponentBias; + // anything else / infinity = +/- 0 + if (bAbs == infRep) + return fromRep(quotientSign); - if (writtenExponent >= maxExponent) { - // If we have overflowed the exponent, return infinity. - return fromRep(infRep | quotientSign); - } - else if (writtenExponent < 1) { - // Flush denormals to zero. In the future, it would be nice to add - // code to round them correctly. + if (!aAbs) { + // zero / zero = NaN + if (!bAbs) + return fromRep(qnanRep); + // zero / anything else = +/- zero + else return fromRep(quotientSign); } - else { - const bool round = (residual << 1) >= bSignificand; - // Clear the implicit bit - rep_t absResult = quotient & significandMask; - // Insert the exponent - absResult |= (rep_t)writtenExponent << significandBits; - // Round - absResult += round; - // Insert the sign and return - const long double result = fromRep(absResult | quotientSign); - return result; + // anything else / zero = +/- infinity + if (!bAbs) + return fromRep(infRep | quotientSign); + + // One or both of a or b is denormal. The other (if applicable) is a + // normal number. Renormalize one or both of a and b, and set scale to + // include the necessary exponent adjustment. + if (aAbs < implicitBit) + scale += normalize(&aSignificand); + if (bAbs < implicitBit) + scale -= normalize(&bSignificand); + } + + // Set the implicit significand bit. If we fell through from the + // denormal path it was already set by normalize( ), but setting it twice + // won't hurt anything. + aSignificand |= implicitBit; + bSignificand |= implicitBit; + int quotientExponent = aExponent - bExponent + scale; + + // Align the significand of b as a Q63 fixed-point number in the range + // [1, 2.0) and get a Q64 approximate reciprocal using a small minimax + // polynomial approximation: reciprocal = 3/4 + 1/sqrt(2) - b/2. This + // is accurate to about 3.5 binary digits. + const uint64_t q63b = bSignificand >> 49; + uint64_t recip64 = UINT64_C(0x7504f333F9DE6484) - q63b; + // 0x7504f333F9DE6484 / 2^64 + 1 = 3/4 + 1/sqrt(2) + + // Now refine the reciprocal estimate using a Newton-Raphson iteration: + // + // x1 = x0 * (2 - x0 * b) + // + // This doubles the number of correct binary digits in the approximation + // with each iteration. + uint64_t correction64; + correction64 = -((rep_t)recip64 * q63b >> 64); + recip64 = (rep_t)recip64 * correction64 >> 63; + correction64 = -((rep_t)recip64 * q63b >> 64); + recip64 = (rep_t)recip64 * correction64 >> 63; + correction64 = -((rep_t)recip64 * q63b >> 64); + recip64 = (rep_t)recip64 * correction64 >> 63; + correction64 = -((rep_t)recip64 * q63b >> 64); + recip64 = (rep_t)recip64 * correction64 >> 63; + correction64 = -((rep_t)recip64 * q63b >> 64); + recip64 = (rep_t)recip64 * correction64 >> 63; + + // The reciprocal may have overflowed to zero if the upper half of b is + // exactly 1.0. This would sabatoge the full-width final stage of the + // computation that follows, so we adjust the reciprocal down by one bit. + recip64--; + + // We need to perform one more iteration to get us to 112 binary digits; + // The last iteration needs to happen with extra precision. + const uint64_t q127blo = bSignificand << 15; + rep_t correction, reciprocal; + + // NOTE: This operation is equivalent to __multi3, which is not implemented + // in some architechure + rep_t r64q63, r64q127, r64cH, r64cL, dummy; + wideMultiply((rep_t)recip64, (rep_t)q63b, &dummy, &r64q63); + wideMultiply((rep_t)recip64, (rep_t)q127blo, &dummy, &r64q127); + + correction = -(r64q63 + (r64q127 >> 64)); + + uint64_t cHi = correction >> 64; + uint64_t cLo = correction; + + wideMultiply((rep_t)recip64, (rep_t)cHi, &dummy, &r64cH); + wideMultiply((rep_t)recip64, (rep_t)cLo, &dummy, &r64cL); + + reciprocal = r64cH + (r64cL >> 64); + + // Adjust the final 128-bit reciprocal estimate downward to ensure that it + // is strictly smaller than the infinitely precise exact reciprocal. Because + // the computation of the Newton-Raphson step is truncating at every step, + // this adjustment is small; most of the work is already done. + reciprocal -= 2; + + // The numerical reciprocal is accurate to within 2^-112, lies in the + // interval [0.5, 1.0), and is strictly smaller than the true reciprocal + // of b. Multiplying a by this reciprocal thus gives a numerical q = a/b + // in Q127 with the following properties: + // + // 1. q < a/b + // 2. q is in the interval [0.5, 2.0) + // 3. The error in q is bounded away from 2^-113 (actually, we have a + // couple of bits to spare, but this is all we need). + + // We need a 128 x 128 multiply high to compute q, which isn't a basic + // operation in C, so we need to be a little bit fussy. + rep_t quotient, quotientLo; + wideMultiply(aSignificand << 2, reciprocal, "ient, "ientLo); + + // Two cases: quotient is in [0.5, 1.0) or quotient is in [1.0, 2.0). + // In either case, we are going to compute a residual of the form + // + // r = a - q*b + // + // We know from the construction of q that r satisfies: + // + // 0 <= r < ulp(q)*b + // + // If r is greater than 1/2 ulp(q)*b, then q rounds up. Otherwise, we + // already have the correct result. The exact halfway case cannot occur. + // We also take this time to right shift quotient if it falls in the [1,2) + // range and adjust the exponent accordingly. + rep_t residual; + rep_t qb; + + if (quotient < (implicitBit << 1)) { + wideMultiply(quotient, bSignificand, &dummy, &qb); + residual = (aSignificand << 113) - qb; + quotientExponent--; + } else { + quotient >>= 1; + wideMultiply(quotient, bSignificand, &dummy, &qb); + residual = (aSignificand << 112) - qb; + } + + const int writtenExponent = quotientExponent + exponentBias; + + if (writtenExponent >= maxExponent) { + // If we have overflowed the exponent, return infinity. + return fromRep(infRep | quotientSign); + } else if (writtenExponent < 1) { + if (writtenExponent == 0) { + // Check whether the rounded result is normal. + const bool round = (residual << 1) > bSignificand; + // Clear the implicit bit. + rep_t absResult = quotient & significandMask; + // Round. + absResult += round; + if (absResult & ~significandMask) { + // The rounded result is normal; return it. + return fromRep(absResult | quotientSign); + } } + // Flush denormals to zero. In the future, it would be nice to add + // code to round them correctly. + return fromRep(quotientSign); + } else { + const bool round = (residual << 1) >= bSignificand; + // Clear the implicit bit. + rep_t absResult = quotient & significandMask; + // Insert the exponent. + absResult |= (rep_t)writtenExponent << significandBits; + // Round. + absResult += round; + // Insert the sign and return. + const long double result = fromRep(absResult | quotientSign); + return result; + } } #endif diff --git a/lib/builtins/divti3.c b/lib/builtins/divti3.c index c73eae28fe08..6d007fe34654 100644 --- a/lib/builtins/divti3.c +++ b/lib/builtins/divti3.c @@ -1,33 +1,29 @@ -/* ===-- divti3.c - Implement __divti3 -------------------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - * - * This file implements __divti3 for the compiler_rt library. - * - * ===----------------------------------------------------------------------=== - */ +//===-- divti3.c - Implement __divti3 -------------------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements __divti3 for the compiler_rt library. +// +//===----------------------------------------------------------------------===// #include "int_lib.h" #ifdef CRT_HAS_128BIT -/* Returns: a / b */ +// Returns: a / b -COMPILER_RT_ABI ti_int -__divti3(ti_int a, ti_int b) -{ - const int bits_in_tword_m1 = (int)(sizeof(ti_int) * CHAR_BIT) - 1; - ti_int s_a = a >> bits_in_tword_m1; /* s_a = a < 0 ? -1 : 0 */ - ti_int s_b = b >> bits_in_tword_m1; /* s_b = b < 0 ? -1 : 0 */ - a = (a ^ s_a) - s_a; /* negate if s_a == -1 */ - b = (b ^ s_b) - s_b; /* negate if s_b == -1 */ - s_a ^= s_b; /* sign of quotient */ - return (__udivmodti4(a, b, (tu_int*)0) ^ s_a) - s_a; /* negate if s_a == -1 */ +COMPILER_RT_ABI ti_int __divti3(ti_int a, ti_int b) { + const int bits_in_tword_m1 = (int)(sizeof(ti_int) * CHAR_BIT) - 1; + ti_int s_a = a >> bits_in_tword_m1; // s_a = a < 0 ? -1 : 0 + ti_int s_b = b >> bits_in_tword_m1; // s_b = b < 0 ? -1 : 0 + a = (a ^ s_a) - s_a; // negate if s_a == -1 + b = (b ^ s_b) - s_b; // negate if s_b == -1 + s_a ^= s_b; // sign of quotient + return (__udivmodti4(a, b, (tu_int *)0) ^ s_a) - s_a; // negate if s_a == -1 } -#endif /* CRT_HAS_128BIT */ +#endif // CRT_HAS_128BIT diff --git a/lib/builtins/divxc3.c b/lib/builtins/divxc3.c index 6f49280e5f61..97ffd2eac211 100644 --- a/lib/builtins/divxc3.c +++ b/lib/builtins/divxc3.c @@ -1,63 +1,55 @@ -/* ===-- divxc3.c - Implement __divxc3 -------------------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - * - * This file implements __divxc3 for the compiler_rt library. - * - */ +//===-- divxc3.c - Implement __divxc3 -------------------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements __divxc3 for the compiler_rt library. +// +//===----------------------------------------------------------------------===// #if !_ARCH_PPC #include "int_lib.h" #include "int_math.h" -/* Returns: the quotient of (a + ib) / (c + id) */ +// Returns: the quotient of (a + ib) / (c + id) -COMPILER_RT_ABI Lcomplex -__divxc3(long double __a, long double __b, long double __c, long double __d) -{ - int __ilogbw = 0; - long double __logbw = crt_logbl(crt_fmaxl(crt_fabsl(__c), crt_fabsl(__d))); - if (crt_isfinite(__logbw)) - { - __ilogbw = (int)__logbw; - __c = crt_scalbnl(__c, -__ilogbw); - __d = crt_scalbnl(__d, -__ilogbw); +COMPILER_RT_ABI Lcomplex __divxc3(long double __a, long double __b, + long double __c, long double __d) { + int __ilogbw = 0; + long double __logbw = crt_logbl(crt_fmaxl(crt_fabsl(__c), crt_fabsl(__d))); + if (crt_isfinite(__logbw)) { + __ilogbw = (int)__logbw; + __c = crt_scalbnl(__c, -__ilogbw); + __d = crt_scalbnl(__d, -__ilogbw); + } + long double __denom = __c * __c + __d * __d; + Lcomplex z; + COMPLEX_REAL(z) = crt_scalbnl((__a * __c + __b * __d) / __denom, -__ilogbw); + COMPLEX_IMAGINARY(z) = + crt_scalbnl((__b * __c - __a * __d) / __denom, -__ilogbw); + if (crt_isnan(COMPLEX_REAL(z)) && crt_isnan(COMPLEX_IMAGINARY(z))) { + if ((__denom == 0) && (!crt_isnan(__a) || !crt_isnan(__b))) { + COMPLEX_REAL(z) = crt_copysignl(CRT_INFINITY, __c) * __a; + COMPLEX_IMAGINARY(z) = crt_copysignl(CRT_INFINITY, __c) * __b; + } else if ((crt_isinf(__a) || crt_isinf(__b)) && crt_isfinite(__c) && + crt_isfinite(__d)) { + __a = crt_copysignl(crt_isinf(__a) ? 1 : 0, __a); + __b = crt_copysignl(crt_isinf(__b) ? 1 : 0, __b); + COMPLEX_REAL(z) = CRT_INFINITY * (__a * __c + __b * __d); + COMPLEX_IMAGINARY(z) = CRT_INFINITY * (__b * __c - __a * __d); + } else if (crt_isinf(__logbw) && __logbw > 0 && crt_isfinite(__a) && + crt_isfinite(__b)) { + __c = crt_copysignl(crt_isinf(__c) ? 1 : 0, __c); + __d = crt_copysignl(crt_isinf(__d) ? 1 : 0, __d); + COMPLEX_REAL(z) = 0 * (__a * __c + __b * __d); + COMPLEX_IMAGINARY(z) = 0 * (__b * __c - __a * __d); } - long double __denom = __c * __c + __d * __d; - Lcomplex z; - COMPLEX_REAL(z) = crt_scalbnl((__a * __c + __b * __d) / __denom, -__ilogbw); - COMPLEX_IMAGINARY(z) = crt_scalbnl((__b * __c - __a * __d) / __denom, -__ilogbw); - if (crt_isnan(COMPLEX_REAL(z)) && crt_isnan(COMPLEX_IMAGINARY(z))) - { - if ((__denom == 0) && (!crt_isnan(__a) || !crt_isnan(__b))) - { - COMPLEX_REAL(z) = crt_copysignl(CRT_INFINITY, __c) * __a; - COMPLEX_IMAGINARY(z) = crt_copysignl(CRT_INFINITY, __c) * __b; - } - else if ((crt_isinf(__a) || crt_isinf(__b)) && - crt_isfinite(__c) && crt_isfinite(__d)) - { - __a = crt_copysignl(crt_isinf(__a) ? 1 : 0, __a); - __b = crt_copysignl(crt_isinf(__b) ? 1 : 0, __b); - COMPLEX_REAL(z) = CRT_INFINITY * (__a * __c + __b * __d); - COMPLEX_IMAGINARY(z) = CRT_INFINITY * (__b * __c - __a * __d); - } - else if (crt_isinf(__logbw) && __logbw > 0 && - crt_isfinite(__a) && crt_isfinite(__b)) - { - __c = crt_copysignl(crt_isinf(__c) ? 1 : 0, __c); - __d = crt_copysignl(crt_isinf(__d) ? 1 : 0, __d); - COMPLEX_REAL(z) = 0 * (__a * __c + __b * __d); - COMPLEX_IMAGINARY(z) = 0 * (__b * __c - __a * __d); - } - } - return z; + } + return z; } #endif diff --git a/lib/builtins/emutls.c b/lib/builtins/emutls.c index ef95a1c260c1..da58feb7b906 100644 --- a/lib/builtins/emutls.c +++ b/lib/builtins/emutls.c @@ -1,37 +1,35 @@ -/* ===---------- emutls.c - Implements __emutls_get_address ---------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - */ +//===---------- emutls.c - Implements __emutls_get_address ---------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// + #include <stdint.h> #include <stdlib.h> #include <string.h> #include "int_lib.h" -#include "int_util.h" #ifdef __BIONIC__ -/* There are 4 pthread key cleanup rounds on Bionic. Delay emutls deallocation - to round 2. We need to delay deallocation because: - - Android versions older than M lack __cxa_thread_atexit_impl, so apps - use a pthread key destructor to call C++ destructors. - - Apps might use __thread/thread_local variables in pthread destructors. - We can't wait until the final two rounds, because jemalloc needs two rounds - after the final malloc/free call to free its thread-specific data (see - https://reviews.llvm.org/D46978#1107507). */ +// There are 4 pthread key cleanup rounds on Bionic. Delay emutls deallocation +// to round 2. We need to delay deallocation because: +// - Android versions older than M lack __cxa_thread_atexit_impl, so apps +// use a pthread key destructor to call C++ destructors. +// - Apps might use __thread/thread_local variables in pthread destructors. +// We can't wait until the final two rounds, because jemalloc needs two rounds +// after the final malloc/free call to free its thread-specific data (see +// https://reviews.llvm.org/D46978#1107507). #define EMUTLS_SKIP_DESTRUCTOR_ROUNDS 1 #else #define EMUTLS_SKIP_DESTRUCTOR_ROUNDS 0 #endif typedef struct emutls_address_array { - uintptr_t skip_destructor_rounds; - uintptr_t size; /* number of elements in the 'data' array */ - void* data[]; + uintptr_t skip_destructor_rounds; + uintptr_t size; // number of elements in the 'data' array + void *data[]; } emutls_address_array; static void emutls_shutdown(emutls_address_array *array); @@ -47,359 +45,339 @@ static bool emutls_key_created = false; typedef unsigned int gcc_word __attribute__((mode(word))); typedef unsigned int gcc_pointer __attribute__((mode(pointer))); -/* Default is not to use posix_memalign, so systems like Android - * can use thread local data without heavier POSIX memory allocators. - */ +// Default is not to use posix_memalign, so systems like Android +// can use thread local data without heavier POSIX memory allocators. #ifndef EMUTLS_USE_POSIX_MEMALIGN #define EMUTLS_USE_POSIX_MEMALIGN 0 #endif static __inline void *emutls_memalign_alloc(size_t align, size_t size) { - void *base; + void *base; #if EMUTLS_USE_POSIX_MEMALIGN - if (posix_memalign(&base, align, size) != 0) - abort(); + if (posix_memalign(&base, align, size) != 0) + abort(); #else - #define EXTRA_ALIGN_PTR_BYTES (align - 1 + sizeof(void*)) - char* object; - if ((object = (char*)malloc(EXTRA_ALIGN_PTR_BYTES + size)) == NULL) - abort(); - base = (void*)(((uintptr_t)(object + EXTRA_ALIGN_PTR_BYTES)) - & ~(uintptr_t)(align - 1)); - - ((void**)base)[-1] = object; +#define EXTRA_ALIGN_PTR_BYTES (align - 1 + sizeof(void *)) + char *object; + if ((object = (char *)malloc(EXTRA_ALIGN_PTR_BYTES + size)) == NULL) + abort(); + base = (void *)(((uintptr_t)(object + EXTRA_ALIGN_PTR_BYTES)) & + ~(uintptr_t)(align - 1)); + + ((void **)base)[-1] = object; #endif - return base; + return base; } static __inline void emutls_memalign_free(void *base) { #if EMUTLS_USE_POSIX_MEMALIGN - free(base); + free(base); #else - /* The mallocated address is in ((void**)base)[-1] */ - free(((void**)base)[-1]); + // The mallocated address is in ((void**)base)[-1] + free(((void **)base)[-1]); #endif } static __inline void emutls_setspecific(emutls_address_array *value) { - pthread_setspecific(emutls_pthread_key, (void*) value); + pthread_setspecific(emutls_pthread_key, (void *)value); } -static __inline emutls_address_array* emutls_getspecific() { - return (emutls_address_array*) pthread_getspecific(emutls_pthread_key); +static __inline emutls_address_array *emutls_getspecific() { + return (emutls_address_array *)pthread_getspecific(emutls_pthread_key); } -static void emutls_key_destructor(void* ptr) { - emutls_address_array *array = (emutls_address_array*)ptr; - if (array->skip_destructor_rounds > 0) { - /* emutls is deallocated using a pthread key destructor. These - * destructors are called in several rounds to accommodate destructor - * functions that (re)initialize key values with pthread_setspecific. - * Delay the emutls deallocation to accommodate other end-of-thread - * cleanup tasks like calling thread_local destructors (e.g. the - * __cxa_thread_atexit fallback in libc++abi). - */ - array->skip_destructor_rounds--; - emutls_setspecific(array); - } else { - emutls_shutdown(array); - free(ptr); - } +static void emutls_key_destructor(void *ptr) { + emutls_address_array *array = (emutls_address_array *)ptr; + if (array->skip_destructor_rounds > 0) { + // emutls is deallocated using a pthread key destructor. These + // destructors are called in several rounds to accommodate destructor + // functions that (re)initialize key values with pthread_setspecific. + // Delay the emutls deallocation to accommodate other end-of-thread + // cleanup tasks like calling thread_local destructors (e.g. the + // __cxa_thread_atexit fallback in libc++abi). + array->skip_destructor_rounds--; + emutls_setspecific(array); + } else { + emutls_shutdown(array); + free(ptr); + } } static __inline void emutls_init(void) { - if (pthread_key_create(&emutls_pthread_key, emutls_key_destructor) != 0) - abort(); - emutls_key_created = true; + if (pthread_key_create(&emutls_pthread_key, emutls_key_destructor) != 0) + abort(); + emutls_key_created = true; } static __inline void emutls_init_once(void) { - static pthread_once_t once = PTHREAD_ONCE_INIT; - pthread_once(&once, emutls_init); + static pthread_once_t once = PTHREAD_ONCE_INIT; + pthread_once(&once, emutls_init); } -static __inline void emutls_lock() { - pthread_mutex_lock(&emutls_mutex); -} +static __inline void emutls_lock() { pthread_mutex_lock(&emutls_mutex); } -static __inline void emutls_unlock() { - pthread_mutex_unlock(&emutls_mutex); -} +static __inline void emutls_unlock() { pthread_mutex_unlock(&emutls_mutex); } -#else /* _WIN32 */ +#else // _WIN32 -#include <windows.h> +#include <assert.h> #include <malloc.h> #include <stdio.h> -#include <assert.h> +#include <windows.h> static LPCRITICAL_SECTION emutls_mutex; static DWORD emutls_tls_index = TLS_OUT_OF_INDEXES; typedef uintptr_t gcc_word; -typedef void * gcc_pointer; +typedef void *gcc_pointer; static void win_error(DWORD last_err, const char *hint) { - char *buffer = NULL; - if (FormatMessageA(FORMAT_MESSAGE_ALLOCATE_BUFFER | - FORMAT_MESSAGE_FROM_SYSTEM | - FORMAT_MESSAGE_MAX_WIDTH_MASK, - NULL, last_err, 0, (LPSTR)&buffer, 1, NULL)) { - fprintf(stderr, "Windows error: %s\n", buffer); - } else { - fprintf(stderr, "Unkown Windows error: %s\n", hint); - } - LocalFree(buffer); + char *buffer = NULL; + if (FormatMessageA(FORMAT_MESSAGE_ALLOCATE_BUFFER | + FORMAT_MESSAGE_FROM_SYSTEM | + FORMAT_MESSAGE_MAX_WIDTH_MASK, + NULL, last_err, 0, (LPSTR)&buffer, 1, NULL)) { + fprintf(stderr, "Windows error: %s\n", buffer); + } else { + fprintf(stderr, "Unkown Windows error: %s\n", hint); + } + LocalFree(buffer); } static __inline void win_abort(DWORD last_err, const char *hint) { - win_error(last_err, hint); - abort(); + win_error(last_err, hint); + abort(); } static __inline void *emutls_memalign_alloc(size_t align, size_t size) { - void *base = _aligned_malloc(size, align); - if (!base) - win_abort(GetLastError(), "_aligned_malloc"); - return base; + void *base = _aligned_malloc(size, align); + if (!base) + win_abort(GetLastError(), "_aligned_malloc"); + return base; } -static __inline void emutls_memalign_free(void *base) { - _aligned_free(base); -} +static __inline void emutls_memalign_free(void *base) { _aligned_free(base); } static void emutls_exit(void) { - if (emutls_mutex) { - DeleteCriticalSection(emutls_mutex); - _aligned_free(emutls_mutex); - emutls_mutex = NULL; - } - if (emutls_tls_index != TLS_OUT_OF_INDEXES) { - emutls_shutdown((emutls_address_array*)TlsGetValue(emutls_tls_index)); - TlsFree(emutls_tls_index); - emutls_tls_index = TLS_OUT_OF_INDEXES; - } + if (emutls_mutex) { + DeleteCriticalSection(emutls_mutex); + _aligned_free(emutls_mutex); + emutls_mutex = NULL; + } + if (emutls_tls_index != TLS_OUT_OF_INDEXES) { + emutls_shutdown((emutls_address_array *)TlsGetValue(emutls_tls_index)); + TlsFree(emutls_tls_index); + emutls_tls_index = TLS_OUT_OF_INDEXES; + } } -#pragma warning (push) -#pragma warning (disable : 4100) +#pragma warning(push) +#pragma warning(disable : 4100) static BOOL CALLBACK emutls_init(PINIT_ONCE p0, PVOID p1, PVOID *p2) { - emutls_mutex = (LPCRITICAL_SECTION)_aligned_malloc(sizeof(CRITICAL_SECTION), 16); - if (!emutls_mutex) { - win_error(GetLastError(), "_aligned_malloc"); - return FALSE; - } - InitializeCriticalSection(emutls_mutex); - - emutls_tls_index = TlsAlloc(); - if (emutls_tls_index == TLS_OUT_OF_INDEXES) { - emutls_exit(); - win_error(GetLastError(), "TlsAlloc"); - return FALSE; - } - atexit(&emutls_exit); - return TRUE; + emutls_mutex = + (LPCRITICAL_SECTION)_aligned_malloc(sizeof(CRITICAL_SECTION), 16); + if (!emutls_mutex) { + win_error(GetLastError(), "_aligned_malloc"); + return FALSE; + } + InitializeCriticalSection(emutls_mutex); + + emutls_tls_index = TlsAlloc(); + if (emutls_tls_index == TLS_OUT_OF_INDEXES) { + emutls_exit(); + win_error(GetLastError(), "TlsAlloc"); + return FALSE; + } + atexit(&emutls_exit); + return TRUE; } static __inline void emutls_init_once(void) { - static INIT_ONCE once; - InitOnceExecuteOnce(&once, emutls_init, NULL, NULL); + static INIT_ONCE once; + InitOnceExecuteOnce(&once, emutls_init, NULL, NULL); } -static __inline void emutls_lock() { - EnterCriticalSection(emutls_mutex); -} +static __inline void emutls_lock() { EnterCriticalSection(emutls_mutex); } -static __inline void emutls_unlock() { - LeaveCriticalSection(emutls_mutex); -} +static __inline void emutls_unlock() { LeaveCriticalSection(emutls_mutex); } static __inline void emutls_setspecific(emutls_address_array *value) { - if (TlsSetValue(emutls_tls_index, (LPVOID) value) == 0) - win_abort(GetLastError(), "TlsSetValue"); + if (TlsSetValue(emutls_tls_index, (LPVOID)value) == 0) + win_abort(GetLastError(), "TlsSetValue"); } -static __inline emutls_address_array* emutls_getspecific() { - LPVOID value = TlsGetValue(emutls_tls_index); - if (value == NULL) { - const DWORD err = GetLastError(); - if (err != ERROR_SUCCESS) - win_abort(err, "TlsGetValue"); - } - return (emutls_address_array*) value; +static __inline emutls_address_array *emutls_getspecific() { + LPVOID value = TlsGetValue(emutls_tls_index); + if (value == NULL) { + const DWORD err = GetLastError(); + if (err != ERROR_SUCCESS) + win_abort(err, "TlsGetValue"); + } + return (emutls_address_array *)value; } -/* Provide atomic load/store functions for emutls_get_index if built with MSVC. - */ +// Provide atomic load/store functions for emutls_get_index if built with MSVC. #if !defined(__ATOMIC_RELEASE) #include <intrin.h> enum { __ATOMIC_ACQUIRE = 2, __ATOMIC_RELEASE = 3 }; static __inline uintptr_t __atomic_load_n(void *ptr, unsigned type) { - assert(type == __ATOMIC_ACQUIRE); - // These return the previous value - but since we do an OR with 0, - // it's equivalent to a plain load. + assert(type == __ATOMIC_ACQUIRE); + // These return the previous value - but since we do an OR with 0, + // it's equivalent to a plain load. #ifdef _WIN64 - return InterlockedOr64(ptr, 0); + return InterlockedOr64(ptr, 0); #else - return InterlockedOr(ptr, 0); + return InterlockedOr(ptr, 0); #endif } static __inline void __atomic_store_n(void *ptr, uintptr_t val, unsigned type) { - assert(type == __ATOMIC_RELEASE); - InterlockedExchangePointer((void *volatile *)ptr, (void *)val); + assert(type == __ATOMIC_RELEASE); + InterlockedExchangePointer((void *volatile *)ptr, (void *)val); } -#endif /* __ATOMIC_RELEASE */ +#endif // __ATOMIC_RELEASE -#pragma warning (pop) +#pragma warning(pop) -#endif /* _WIN32 */ +#endif // _WIN32 -static size_t emutls_num_object = 0; /* number of allocated TLS objects */ +static size_t emutls_num_object = 0; // number of allocated TLS objects -/* Free the allocated TLS data - */ +// Free the allocated TLS data static void emutls_shutdown(emutls_address_array *array) { - if (array) { - uintptr_t i; - for (i = 0; i < array->size; ++i) { - if (array->data[i]) - emutls_memalign_free(array->data[i]); - } + if (array) { + uintptr_t i; + for (i = 0; i < array->size; ++i) { + if (array->data[i]) + emutls_memalign_free(array->data[i]); } + } } -/* For every TLS variable xyz, - * there is one __emutls_control variable named __emutls_v.xyz. - * If xyz has non-zero initial value, __emutls_v.xyz's "value" - * will point to __emutls_t.xyz, which has the initial value. - */ +// For every TLS variable xyz, +// there is one __emutls_control variable named __emutls_v.xyz. +// If xyz has non-zero initial value, __emutls_v.xyz's "value" +// will point to __emutls_t.xyz, which has the initial value. typedef struct __emutls_control { - /* Must use gcc_word here, instead of size_t, to match GCC. When - gcc_word is larger than size_t, the upper extra bits are all - zeros. We can use variables of size_t to operate on size and - align. */ - gcc_word size; /* size of the object in bytes */ - gcc_word align; /* alignment of the object in bytes */ - union { - uintptr_t index; /* data[index-1] is the object address */ - void* address; /* object address, when in single thread env */ - } object; - void* value; /* null or non-zero initial value for the object */ + // Must use gcc_word here, instead of size_t, to match GCC. When + // gcc_word is larger than size_t, the upper extra bits are all + // zeros. We can use variables of size_t to operate on size and + // align. + gcc_word size; // size of the object in bytes + gcc_word align; // alignment of the object in bytes + union { + uintptr_t index; // data[index-1] is the object address + void *address; // object address, when in single thread env + } object; + void *value; // null or non-zero initial value for the object } __emutls_control; -/* Emulated TLS objects are always allocated at run-time. */ +// Emulated TLS objects are always allocated at run-time. static __inline void *emutls_allocate_object(__emutls_control *control) { - /* Use standard C types, check with gcc's emutls.o. */ - COMPILE_TIME_ASSERT(sizeof(uintptr_t) == sizeof(gcc_pointer)); - COMPILE_TIME_ASSERT(sizeof(uintptr_t) == sizeof(void*)); - - size_t size = control->size; - size_t align = control->align; - void* base; - if (align < sizeof(void*)) - align = sizeof(void*); - /* Make sure that align is power of 2. */ - if ((align & (align - 1)) != 0) - abort(); - - base = emutls_memalign_alloc(align, size); - if (control->value) - memcpy(base, control->value, size); - else - memset(base, 0, size); - return base; -} + // Use standard C types, check with gcc's emutls.o. + COMPILE_TIME_ASSERT(sizeof(uintptr_t) == sizeof(gcc_pointer)); + COMPILE_TIME_ASSERT(sizeof(uintptr_t) == sizeof(void *)); + + size_t size = control->size; + size_t align = control->align; + void *base; + if (align < sizeof(void *)) + align = sizeof(void *); + // Make sure that align is power of 2. + if ((align & (align - 1)) != 0) + abort(); + base = emutls_memalign_alloc(align, size); + if (control->value) + memcpy(base, control->value, size); + else + memset(base, 0, size); + return base; +} -/* Returns control->object.index; set index if not allocated yet. */ +// Returns control->object.index; set index if not allocated yet. static __inline uintptr_t emutls_get_index(__emutls_control *control) { - uintptr_t index = __atomic_load_n(&control->object.index, __ATOMIC_ACQUIRE); + uintptr_t index = __atomic_load_n(&control->object.index, __ATOMIC_ACQUIRE); + if (!index) { + emutls_init_once(); + emutls_lock(); + index = control->object.index; if (!index) { - emutls_init_once(); - emutls_lock(); - index = control->object.index; - if (!index) { - index = ++emutls_num_object; - __atomic_store_n(&control->object.index, index, __ATOMIC_RELEASE); - } - emutls_unlock(); + index = ++emutls_num_object; + __atomic_store_n(&control->object.index, index, __ATOMIC_RELEASE); } - return index; + emutls_unlock(); + } + return index; } -/* Updates newly allocated thread local emutls_address_array. */ +// Updates newly allocated thread local emutls_address_array. static __inline void emutls_check_array_set_size(emutls_address_array *array, uintptr_t size) { - if (array == NULL) - abort(); - array->size = size; - emutls_setspecific(array); + if (array == NULL) + abort(); + array->size = size; + emutls_setspecific(array); } -/* Returns the new 'data' array size, number of elements, - * which must be no smaller than the given index. - */ +// Returns the new 'data' array size, number of elements, +// which must be no smaller than the given index. static __inline uintptr_t emutls_new_data_array_size(uintptr_t index) { - /* Need to allocate emutls_address_array with extra slots - * to store the header. - * Round up the emutls_address_array size to multiple of 16. - */ - uintptr_t header_words = sizeof(emutls_address_array) / sizeof(void *); - return ((index + header_words + 15) & ~((uintptr_t)15)) - header_words; + // Need to allocate emutls_address_array with extra slots + // to store the header. + // Round up the emutls_address_array size to multiple of 16. + uintptr_t header_words = sizeof(emutls_address_array) / sizeof(void *); + return ((index + header_words + 15) & ~((uintptr_t)15)) - header_words; } -/* Returns the size in bytes required for an emutls_address_array with - * N number of elements for data field. - */ +// Returns the size in bytes required for an emutls_address_array with +// N number of elements for data field. static __inline uintptr_t emutls_asize(uintptr_t N) { - return N * sizeof(void *) + sizeof(emutls_address_array); + return N * sizeof(void *) + sizeof(emutls_address_array); } -/* Returns the thread local emutls_address_array. - * Extends its size if necessary to hold address at index. - */ +// Returns the thread local emutls_address_array. +// Extends its size if necessary to hold address at index. static __inline emutls_address_array * emutls_get_address_array(uintptr_t index) { - emutls_address_array* array = emutls_getspecific(); - if (array == NULL) { - uintptr_t new_size = emutls_new_data_array_size(index); - array = (emutls_address_array*) malloc(emutls_asize(new_size)); - if (array) { - memset(array->data, 0, new_size * sizeof(void*)); - array->skip_destructor_rounds = EMUTLS_SKIP_DESTRUCTOR_ROUNDS; - } - emutls_check_array_set_size(array, new_size); - } else if (index > array->size) { - uintptr_t orig_size = array->size; - uintptr_t new_size = emutls_new_data_array_size(index); - array = (emutls_address_array*) realloc(array, emutls_asize(new_size)); - if (array) - memset(array->data + orig_size, 0, - (new_size - orig_size) * sizeof(void*)); - emutls_check_array_set_size(array, new_size); + emutls_address_array *array = emutls_getspecific(); + if (array == NULL) { + uintptr_t new_size = emutls_new_data_array_size(index); + array = (emutls_address_array *)malloc(emutls_asize(new_size)); + if (array) { + memset(array->data, 0, new_size * sizeof(void *)); + array->skip_destructor_rounds = EMUTLS_SKIP_DESTRUCTOR_ROUNDS; } - return array; + emutls_check_array_set_size(array, new_size); + } else if (index > array->size) { + uintptr_t orig_size = array->size; + uintptr_t new_size = emutls_new_data_array_size(index); + array = (emutls_address_array *)realloc(array, emutls_asize(new_size)); + if (array) + memset(array->data + orig_size, 0, + (new_size - orig_size) * sizeof(void *)); + emutls_check_array_set_size(array, new_size); + } + return array; } -void* __emutls_get_address(__emutls_control* control) { - uintptr_t index = emutls_get_index(control); - emutls_address_array* array = emutls_get_address_array(index--); - if (array->data[index] == NULL) - array->data[index] = emutls_allocate_object(control); - return array->data[index]; +void *__emutls_get_address(__emutls_control *control) { + uintptr_t index = emutls_get_index(control); + emutls_address_array *array = emutls_get_address_array(index--); + if (array->data[index] == NULL) + array->data[index] = emutls_allocate_object(control); + return array->data[index]; } #ifdef __BIONIC__ -/* Called by Bionic on dlclose to delete the emutls pthread key. */ -__attribute__((visibility("hidden"))) -void __emutls_unregister_key(void) { - if (emutls_key_created) { - pthread_key_delete(emutls_pthread_key); - emutls_key_created = false; - } +// Called by Bionic on dlclose to delete the emutls pthread key. +__attribute__((visibility("hidden"))) void __emutls_unregister_key(void) { + if (emutls_key_created) { + pthread_key_delete(emutls_pthread_key); + emutls_key_created = false; + } } #endif diff --git a/lib/builtins/enable_execute_stack.c b/lib/builtins/enable_execute_stack.c index 327d460b4253..e18de4eaebf2 100644 --- a/lib/builtins/enable_execute_stack.c +++ b/lib/builtins/enable_execute_stack.c @@ -1,12 +1,10 @@ -/* ===-- enable_execute_stack.c - Implement __enable_execute_stack ---------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - */ +//===-- enable_execute_stack.c - Implement __enable_execute_stack ---------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// #include "int_lib.h" @@ -14,10 +12,9 @@ #include <sys/mman.h> #endif -/* #include "config.h" - * FIXME: CMake - include when cmake system is ready. - * Remove #define HAVE_SYSCONF 1 line. - */ +// #include "config.h" +// FIXME: CMake - include when cmake system is ready. +// Remove #define HAVE_SYSCONF 1 line. #define HAVE_SYSCONF 1 #ifdef _WIN32 @@ -26,47 +23,45 @@ #else #ifndef __APPLE__ #include <unistd.h> -#endif /* __APPLE__ */ -#endif /* _WIN32 */ +#endif // __APPLE__ +#endif // _WIN32 #if __LP64__ - #define TRAMPOLINE_SIZE 48 +#define TRAMPOLINE_SIZE 48 #else - #define TRAMPOLINE_SIZE 40 +#define TRAMPOLINE_SIZE 40 #endif -/* - * The compiler generates calls to __enable_execute_stack() when creating - * trampoline functions on the stack for use with nested functions. - * It is expected to mark the page(s) containing the address - * and the next 48 bytes as executable. Since the stack is normally rw- - * that means changing the protection on those page(s) to rwx. - */ +// The compiler generates calls to __enable_execute_stack() when creating +// trampoline functions on the stack for use with nested functions. +// It is expected to mark the page(s) containing the address +// and the next 48 bytes as executable. Since the stack is normally rw- +// that means changing the protection on those page(s) to rwx. -COMPILER_RT_ABI void -__enable_execute_stack(void* addr) -{ +COMPILER_RT_ABI void __enable_execute_stack(void *addr) { #if _WIN32 - MEMORY_BASIC_INFORMATION mbi; - if (!VirtualQuery (addr, &mbi, sizeof(mbi))) - return; /* We should probably assert here because there is no return value */ - VirtualProtect (mbi.BaseAddress, mbi.RegionSize, PAGE_EXECUTE_READWRITE, &mbi.Protect); + MEMORY_BASIC_INFORMATION mbi; + if (!VirtualQuery(addr, &mbi, sizeof(mbi))) + return; // We should probably assert here because there is no return value + VirtualProtect(mbi.BaseAddress, mbi.RegionSize, PAGE_EXECUTE_READWRITE, + &mbi.Protect); #else #if __APPLE__ - /* On Darwin, pagesize is always 4096 bytes */ - const uintptr_t pageSize = 4096; + // On Darwin, pagesize is always 4096 bytes + const uintptr_t pageSize = 4096; #elif !defined(HAVE_SYSCONF) #error "HAVE_SYSCONF not defined! See enable_execute_stack.c" #else - const uintptr_t pageSize = sysconf(_SC_PAGESIZE); -#endif /* __APPLE__ */ + const uintptr_t pageSize = sysconf(_SC_PAGESIZE); +#endif // __APPLE__ - const uintptr_t pageAlignMask = ~(pageSize-1); - uintptr_t p = (uintptr_t)addr; - unsigned char* startPage = (unsigned char*)(p & pageAlignMask); - unsigned char* endPage = (unsigned char*)((p+TRAMPOLINE_SIZE+pageSize) & pageAlignMask); - size_t length = endPage - startPage; - (void) mprotect((void *)startPage, length, PROT_READ | PROT_WRITE | PROT_EXEC); + const uintptr_t pageAlignMask = ~(pageSize - 1); + uintptr_t p = (uintptr_t)addr; + unsigned char *startPage = (unsigned char *)(p & pageAlignMask); + unsigned char *endPage = + (unsigned char *)((p + TRAMPOLINE_SIZE + pageSize) & pageAlignMask); + size_t length = endPage - startPage; + (void)mprotect((void *)startPage, length, PROT_READ | PROT_WRITE | PROT_EXEC); #endif } diff --git a/lib/builtins/eprintf.c b/lib/builtins/eprintf.c index 89f34b154577..89fb0e315b2e 100644 --- a/lib/builtins/eprintf.c +++ b/lib/builtins/eprintf.c @@ -1,35 +1,27 @@ -/* ===---------- eprintf.c - Implements __eprintf --------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - */ - - +//===---------- eprintf.c - Implements __eprintf --------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// #include "int_lib.h" #include <stdio.h> - -/* - * __eprintf() was used in an old version of <assert.h>. - * It can eventually go away, but it is needed when linking - * .o files built with the old <assert.h>. - * - * It should never be exported from a dylib, so it is marked - * visibility hidden. - */ +// __eprintf() was used in an old version of <assert.h>. +// It can eventually go away, but it is needed when linking +// .o files built with the old <assert.h>. +// +// It should never be exported from a dylib, so it is marked +// visibility hidden. #ifndef _WIN32 __attribute__((visibility("hidden"))) #endif COMPILER_RT_ABI void -__eprintf(const char* format, const char* assertion_expression, - const char* line, const char* file) -{ - fprintf(stderr, format, assertion_expression, line, file); - fflush(stderr); - compilerrt_abort(); +__eprintf(const char *format, const char *assertion_expression, + const char *line, const char *file) { + fprintf(stderr, format, assertion_expression, line, file); + fflush(stderr); + compilerrt_abort(); } diff --git a/lib/builtins/extenddftf2.c b/lib/builtins/extenddftf2.c index 86dab8f03a8f..849a39da1915 100644 --- a/lib/builtins/extenddftf2.c +++ b/lib/builtins/extenddftf2.c @@ -1,12 +1,10 @@ //===-- lib/extenddftf2.c - double -> quad conversion -------------*- C -*-===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// -// #define QUAD_PRECISION #include "fp_lib.h" @@ -17,7 +15,7 @@ #include "fp_extend_impl.inc" COMPILER_RT_ABI long double __extenddftf2(double a) { - return __extendXfYf2__(a); + return __extendXfYf2__(a); } #endif diff --git a/lib/builtins/extendhfsf2.c b/lib/builtins/extendhfsf2.c index d9c0db84b0ce..7c1a76eb5851 100644 --- a/lib/builtins/extendhfsf2.c +++ b/lib/builtins/extendhfsf2.c @@ -1,12 +1,10 @@ //===-- lib/extendhfsf2.c - half -> single conversion -------------*- C -*-===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// -// #define SRC_HALF #define DST_SINGLE @@ -15,19 +13,15 @@ // Use a forwarding definition and noinline to implement a poor man's alias, // as there isn't a good cross-platform way of defining one. COMPILER_RT_ABI NOINLINE float __extendhfsf2(uint16_t a) { - return __extendXfYf2__(a); + return __extendXfYf2__(a); } -COMPILER_RT_ABI float __gnu_h2f_ieee(uint16_t a) { - return __extendhfsf2(a); -} +COMPILER_RT_ABI float __gnu_h2f_ieee(uint16_t a) { return __extendhfsf2(a); } #if defined(__ARM_EABI__) #if defined(COMPILER_RT_ARMHF_TARGET) -AEABI_RTABI float __aeabi_h2f(uint16_t a) { - return __extendhfsf2(a); -} +AEABI_RTABI float __aeabi_h2f(uint16_t a) { return __extendhfsf2(a); } #else -AEABI_RTABI float __aeabi_h2f(uint16_t a) COMPILER_RT_ALIAS(__extendhfsf2); +COMPILER_RT_ALIAS(__extendhfsf2, __aeabi_h2f) #endif #endif diff --git a/lib/builtins/extendsfdf2.c b/lib/builtins/extendsfdf2.c index 3d84529a6c53..8132d57e6bf6 100644 --- a/lib/builtins/extendsfdf2.c +++ b/lib/builtins/extendsfdf2.c @@ -1,27 +1,21 @@ //===-- lib/extendsfdf2.c - single -> double conversion -----------*- C -*-===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// -// #define SRC_SINGLE #define DST_DOUBLE #include "fp_extend_impl.inc" -COMPILER_RT_ABI double __extendsfdf2(float a) { - return __extendXfYf2__(a); -} +COMPILER_RT_ABI double __extendsfdf2(float a) { return __extendXfYf2__(a); } #if defined(__ARM_EABI__) #if defined(COMPILER_RT_ARMHF_TARGET) -AEABI_RTABI double __aeabi_f2d(float a) { - return __extendsfdf2(a); -} +AEABI_RTABI double __aeabi_f2d(float a) { return __extendsfdf2(a); } #else -AEABI_RTABI double __aeabi_f2d(float a) COMPILER_RT_ALIAS(__extendsfdf2); +COMPILER_RT_ALIAS(__extendsfdf2, __aeabi_f2d) #endif #endif diff --git a/lib/builtins/extendsftf2.c b/lib/builtins/extendsftf2.c index 2eeeba284845..c6368406dde1 100644 --- a/lib/builtins/extendsftf2.c +++ b/lib/builtins/extendsftf2.c @@ -1,12 +1,10 @@ //===-- lib/extendsftf2.c - single -> quad conversion -------------*- C -*-===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// -// #define QUAD_PRECISION #include "fp_lib.h" @@ -17,7 +15,7 @@ #include "fp_extend_impl.inc" COMPILER_RT_ABI long double __extendsftf2(float a) { - return __extendXfYf2__(a); + return __extendXfYf2__(a); } #endif diff --git a/lib/builtins/ffsdi2.c b/lib/builtins/ffsdi2.c index a5ac9900ff10..9c1a24260956 100644 --- a/lib/builtins/ffsdi2.c +++ b/lib/builtins/ffsdi2.c @@ -1,33 +1,27 @@ -/* ===-- ffsdi2.c - Implement __ffsdi2 -------------------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - * - * This file implements __ffsdi2 for the compiler_rt library. - * - * ===----------------------------------------------------------------------=== - */ +//===-- ffsdi2.c - Implement __ffsdi2 -------------------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements __ffsdi2 for the compiler_rt library. +// +//===----------------------------------------------------------------------===// #include "int_lib.h" -/* Returns: the index of the least significant 1-bit in a, or - * the value zero if a is zero. The least significant bit is index one. - */ +// Returns: the index of the least significant 1-bit in a, or +// the value zero if a is zero. The least significant bit is index one. -COMPILER_RT_ABI si_int -__ffsdi2(di_int a) -{ - dwords x; - x.all = a; - if (x.s.low == 0) - { - if (x.s.high == 0) - return 0; - return __builtin_ctz(x.s.high) + (1 + sizeof(si_int) * CHAR_BIT); - } - return __builtin_ctz(x.s.low) + 1; +COMPILER_RT_ABI si_int __ffsdi2(di_int a) { + dwords x; + x.all = a; + if (x.s.low == 0) { + if (x.s.high == 0) + return 0; + return __builtin_ctz(x.s.high) + (1 + sizeof(si_int) * CHAR_BIT); + } + return __builtin_ctz(x.s.low) + 1; } diff --git a/lib/builtins/ffssi2.c b/lib/builtins/ffssi2.c index e5180eff5e08..cba1f72fdc61 100644 --- a/lib/builtins/ffssi2.c +++ b/lib/builtins/ffssi2.c @@ -1,29 +1,23 @@ -/* ===-- ffssi2.c - Implement __ffssi2 -------------------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - * - * This file implements __ffssi2 for the compiler_rt library. - * - * ===----------------------------------------------------------------------=== - */ +//===-- ffssi2.c - Implement __ffssi2 -------------------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements __ffssi2 for the compiler_rt library. +// +//===----------------------------------------------------------------------===// #include "int_lib.h" -/* Returns: the index of the least significant 1-bit in a, or - * the value zero if a is zero. The least significant bit is index one. - */ +// Returns: the index of the least significant 1-bit in a, or +// the value zero if a is zero. The least significant bit is index one. -COMPILER_RT_ABI si_int -__ffssi2(si_int a) -{ - if (a == 0) - { - return 0; - } - return __builtin_ctz(a) + 1; +COMPILER_RT_ABI si_int __ffssi2(si_int a) { + if (a == 0) { + return 0; + } + return __builtin_ctz(a) + 1; } diff --git a/lib/builtins/ffsti2.c b/lib/builtins/ffsti2.c index dcdb3bd7f807..a2d7ce08ada1 100644 --- a/lib/builtins/ffsti2.c +++ b/lib/builtins/ffsti2.c @@ -1,37 +1,31 @@ -/* ===-- ffsti2.c - Implement __ffsti2 -------------------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - * - * This file implements __ffsti2 for the compiler_rt library. - * - * ===----------------------------------------------------------------------=== - */ +//===-- ffsti2.c - Implement __ffsti2 -------------------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements __ffsti2 for the compiler_rt library. +// +//===----------------------------------------------------------------------===// #include "int_lib.h" #ifdef CRT_HAS_128BIT -/* Returns: the index of the least significant 1-bit in a, or - * the value zero if a is zero. The least significant bit is index one. - */ +// Returns: the index of the least significant 1-bit in a, or +// the value zero if a is zero. The least significant bit is index one. -COMPILER_RT_ABI si_int -__ffsti2(ti_int a) -{ - twords x; - x.all = a; - if (x.s.low == 0) - { - if (x.s.high == 0) - return 0; - return __builtin_ctzll(x.s.high) + (1 + sizeof(di_int) * CHAR_BIT); - } - return __builtin_ctzll(x.s.low) + 1; +COMPILER_RT_ABI si_int __ffsti2(ti_int a) { + twords x; + x.all = a; + if (x.s.low == 0) { + if (x.s.high == 0) + return 0; + return __builtin_ctzll(x.s.high) + (1 + sizeof(di_int) * CHAR_BIT); + } + return __builtin_ctzll(x.s.low) + 1; } -#endif /* CRT_HAS_128BIT */ +#endif // CRT_HAS_128BIT diff --git a/lib/builtins/fixdfdi.c b/lib/builtins/fixdfdi.c index 54e312d3c8f7..2ed5261c51b7 100644 --- a/lib/builtins/fixdfdi.c +++ b/lib/builtins/fixdfdi.c @@ -1,55 +1,44 @@ -/* ===-- fixdfdi.c - Implement __fixdfdi -----------------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - */ +//===-- fixdfdi.c - Implement __fixdfdi -----------------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// #define DOUBLE_PRECISION #include "fp_lib.h" #ifndef __SOFT_FP__ -/* Support for systems that have hardware floating-point; can set the invalid - * flag as a side-effect of computation. - */ +// Support for systems that have hardware floating-point; can set the invalid +// flag as a side-effect of computation. COMPILER_RT_ABI du_int __fixunsdfdi(double a); -COMPILER_RT_ABI di_int -__fixdfdi(double a) -{ - if (a < 0.0) { - return -__fixunsdfdi(-a); - } - return __fixunsdfdi(a); +COMPILER_RT_ABI di_int __fixdfdi(double a) { + if (a < 0.0) { + return -__fixunsdfdi(-a); + } + return __fixunsdfdi(a); } #else -/* Support for systems that don't have hardware floating-point; there are no - * flags to set, and we don't want to code-gen to an unknown soft-float - * implementation. - */ +// Support for systems that don't have hardware floating-point; there are no +// flags to set, and we don't want to code-gen to an unknown soft-float +// implementation. typedef di_int fixint_t; typedef du_int fixuint_t; #include "fp_fixint_impl.inc" -COMPILER_RT_ABI di_int -__fixdfdi(fp_t a) { - return __fixint(a); -} +COMPILER_RT_ABI di_int __fixdfdi(fp_t a) { return __fixint(a); } #endif #if defined(__ARM_EABI__) #if defined(COMPILER_RT_ARMHF_TARGET) -AEABI_RTABI di_int __aeabi_d2lz(fp_t a) { - return __fixdfdi(a); -} +AEABI_RTABI di_int __aeabi_d2lz(fp_t a) { return __fixdfdi(a); } #else -AEABI_RTABI di_int __aeabi_d2lz(fp_t a) COMPILER_RT_ALIAS(__fixdfdi); +COMPILER_RT_ALIAS(__fixdfdi, __aeabi_d2lz) #endif #endif diff --git a/lib/builtins/fixdfsi.c b/lib/builtins/fixdfsi.c index 5b9588175717..f54649993c7f 100644 --- a/lib/builtins/fixdfsi.c +++ b/lib/builtins/fixdfsi.c @@ -1,12 +1,10 @@ -/* ===-- fixdfsi.c - Implement __fixdfsi -----------------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - */ +//===-- fixdfsi.c - Implement __fixdfsi -----------------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// #define DOUBLE_PRECISION #include "fp_lib.h" @@ -14,17 +12,12 @@ typedef si_int fixint_t; typedef su_int fixuint_t; #include "fp_fixint_impl.inc" -COMPILER_RT_ABI si_int -__fixdfsi(fp_t a) { - return __fixint(a); -} +COMPILER_RT_ABI si_int __fixdfsi(fp_t a) { return __fixint(a); } #if defined(__ARM_EABI__) #if defined(COMPILER_RT_ARMHF_TARGET) -AEABI_RTABI si_int __aeabi_d2iz(fp_t a) { - return __fixdfsi(a); -} +AEABI_RTABI si_int __aeabi_d2iz(fp_t a) { return __fixdfsi(a); } #else -AEABI_RTABI si_int __aeabi_d2iz(fp_t a) COMPILER_RT_ALIAS(__fixdfsi); +COMPILER_RT_ALIAS(__fixdfsi, __aeabi_d2iz) #endif #endif diff --git a/lib/builtins/fixdfti.c b/lib/builtins/fixdfti.c index aaf225e74f86..90ca8959dae1 100644 --- a/lib/builtins/fixdfti.c +++ b/lib/builtins/fixdfti.c @@ -1,12 +1,10 @@ -/* ===-- fixdfti.c - Implement __fixdfti -----------------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - */ +//===-- fixdfti.c - Implement __fixdfti -----------------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// #include "int_lib.h" @@ -18,9 +16,6 @@ typedef ti_int fixint_t; typedef tu_int fixuint_t; #include "fp_fixint_impl.inc" -COMPILER_RT_ABI ti_int -__fixdfti(fp_t a) { - return __fixint(a); -} +COMPILER_RT_ABI ti_int __fixdfti(fp_t a) { return __fixint(a); } -#endif /* CRT_HAS_128BIT */ +#endif // CRT_HAS_128BIT diff --git a/lib/builtins/fixsfdi.c b/lib/builtins/fixsfdi.c index 32e87c60889f..615e93d4f8d9 100644 --- a/lib/builtins/fixsfdi.c +++ b/lib/builtins/fixsfdi.c @@ -1,55 +1,44 @@ -/* ===-- fixsfdi.c - Implement __fixsfdi -----------------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - */ +//===-- fixsfdi.c - Implement __fixsfdi -----------------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// #define SINGLE_PRECISION #include "fp_lib.h" #ifndef __SOFT_FP__ -/* Support for systems that have hardware floating-point; can set the invalid - * flag as a side-effect of computation. - */ +// Support for systems that have hardware floating-point; can set the invalid +// flag as a side-effect of computation. COMPILER_RT_ABI du_int __fixunssfdi(float a); -COMPILER_RT_ABI di_int -__fixsfdi(float a) -{ - if (a < 0.0f) { - return -__fixunssfdi(-a); - } - return __fixunssfdi(a); +COMPILER_RT_ABI di_int __fixsfdi(float a) { + if (a < 0.0f) { + return -__fixunssfdi(-a); + } + return __fixunssfdi(a); } #else -/* Support for systems that don't have hardware floating-point; there are no - * flags to set, and we don't want to code-gen to an unknown soft-float - * implementation. - */ +// Support for systems that don't have hardware floating-point; there are no +// flags to set, and we don't want to code-gen to an unknown soft-float +// implementation. typedef di_int fixint_t; typedef du_int fixuint_t; #include "fp_fixint_impl.inc" -COMPILER_RT_ABI di_int -__fixsfdi(fp_t a) { - return __fixint(a); -} +COMPILER_RT_ABI di_int __fixsfdi(fp_t a) { return __fixint(a); } #endif #if defined(__ARM_EABI__) #if defined(COMPILER_RT_ARMHF_TARGET) -AEABI_RTABI di_int __aeabi_f2lz(fp_t a) { - return __fixsfdi(a); -} +AEABI_RTABI di_int __aeabi_f2lz(fp_t a) { return __fixsfdi(a); } #else -AEABI_RTABI di_int __aeabi_f2lz(fp_t a) COMPILER_RT_ALIAS(__fixsfdi); +COMPILER_RT_ALIAS(__fixsfdi, __aeabi_f2lz) #endif #endif diff --git a/lib/builtins/fixsfsi.c b/lib/builtins/fixsfsi.c index e94e5f3dcd68..d83d7e722271 100644 --- a/lib/builtins/fixsfsi.c +++ b/lib/builtins/fixsfsi.c @@ -1,12 +1,10 @@ -/* ===-- fixsfsi.c - Implement __fixsfsi -----------------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - */ +//===-- fixsfsi.c - Implement __fixsfsi -----------------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// #define SINGLE_PRECISION #include "fp_lib.h" @@ -14,17 +12,12 @@ typedef si_int fixint_t; typedef su_int fixuint_t; #include "fp_fixint_impl.inc" -COMPILER_RT_ABI si_int -__fixsfsi(fp_t a) { - return __fixint(a); -} +COMPILER_RT_ABI si_int __fixsfsi(fp_t a) { return __fixint(a); } #if defined(__ARM_EABI__) #if defined(COMPILER_RT_ARMHF_TARGET) -AEABI_RTABI si_int __aeabi_f2iz(fp_t a) { - return __fixsfsi(a); -} +AEABI_RTABI si_int __aeabi_f2iz(fp_t a) { return __fixsfsi(a); } #else -AEABI_RTABI si_int __aeabi_f2iz(fp_t a) COMPILER_RT_ALIAS(__fixsfsi); +COMPILER_RT_ALIAS(__fixsfsi, __aeabi_f2iz) #endif #endif diff --git a/lib/builtins/fixsfti.c b/lib/builtins/fixsfti.c index 3a159b3e18e4..3c01b75e2b23 100644 --- a/lib/builtins/fixsfti.c +++ b/lib/builtins/fixsfti.c @@ -1,12 +1,10 @@ -/* ===-- fixsfti.c - Implement __fixsfti -----------------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - */ +//===-- fixsfti.c - Implement __fixsfti -----------------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// #include "int_lib.h" @@ -18,9 +16,6 @@ typedef ti_int fixint_t; typedef tu_int fixuint_t; #include "fp_fixint_impl.inc" -COMPILER_RT_ABI ti_int -__fixsfti(fp_t a) { - return __fixint(a); -} +COMPILER_RT_ABI ti_int __fixsfti(fp_t a) { return __fixint(a); } -#endif /* CRT_HAS_128BIT */ +#endif // CRT_HAS_128BIT diff --git a/lib/builtins/fixtfdi.c b/lib/builtins/fixtfdi.c index bc9dea1f4f81..fe570e6b3755 100644 --- a/lib/builtins/fixtfdi.c +++ b/lib/builtins/fixtfdi.c @@ -1,12 +1,10 @@ -/* ===-- fixtfdi.c - Implement __fixtfdi -----------------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - */ +//===-- fixtfdi.c - Implement __fixtfdi -----------------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// #define QUAD_PRECISION #include "fp_lib.h" @@ -16,8 +14,5 @@ typedef di_int fixint_t; typedef du_int fixuint_t; #include "fp_fixint_impl.inc" -COMPILER_RT_ABI di_int -__fixtfdi(fp_t a) { - return __fixint(a); -} +COMPILER_RT_ABI di_int __fixtfdi(fp_t a) { return __fixint(a); } #endif diff --git a/lib/builtins/fixtfsi.c b/lib/builtins/fixtfsi.c index feb3de885090..a32bd964caa3 100644 --- a/lib/builtins/fixtfsi.c +++ b/lib/builtins/fixtfsi.c @@ -1,12 +1,10 @@ -/* ===-- fixtfsi.c - Implement __fixtfsi -----------------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - */ +//===-- fixtfsi.c - Implement __fixtfsi -----------------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// #define QUAD_PRECISION #include "fp_lib.h" @@ -16,8 +14,5 @@ typedef si_int fixint_t; typedef su_int fixuint_t; #include "fp_fixint_impl.inc" -COMPILER_RT_ABI si_int -__fixtfsi(fp_t a) { - return __fixint(a); -} +COMPILER_RT_ABI si_int __fixtfsi(fp_t a) { return __fixint(a); } #endif diff --git a/lib/builtins/fixtfti.c b/lib/builtins/fixtfti.c index ee4ada85cb4a..19f84ce38907 100644 --- a/lib/builtins/fixtfti.c +++ b/lib/builtins/fixtfti.c @@ -1,12 +1,10 @@ -/* ===-- fixtfti.c - Implement __fixtfti -----------------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - */ +//===-- fixtfti.c - Implement __fixtfti -----------------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// #define QUAD_PRECISION #include "fp_lib.h" @@ -16,8 +14,5 @@ typedef ti_int fixint_t; typedef tu_int fixuint_t; #include "fp_fixint_impl.inc" -COMPILER_RT_ABI ti_int -__fixtfti(fp_t a) { - return __fixint(a); -} +COMPILER_RT_ABI ti_int __fixtfti(fp_t a) { return __fixint(a); } #endif diff --git a/lib/builtins/fixunsdfdi.c b/lib/builtins/fixunsdfdi.c index bfe4dbb25656..d2ba73825cfe 100644 --- a/lib/builtins/fixunsdfdi.c +++ b/lib/builtins/fixunsdfdi.c @@ -1,52 +1,42 @@ -/* ===-- fixunsdfdi.c - Implement __fixunsdfdi -----------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - */ +//===-- fixunsdfdi.c - Implement __fixunsdfdi -----------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// #define DOUBLE_PRECISION #include "fp_lib.h" #ifndef __SOFT_FP__ -/* Support for systems that have hardware floating-point; can set the invalid - * flag as a side-effect of computation. - */ - -COMPILER_RT_ABI du_int -__fixunsdfdi(double a) -{ - if (a <= 0.0) return 0; - su_int high = a / 4294967296.f; /* a / 0x1p32f; */ - su_int low = a - (double)high * 4294967296.f; /* high * 0x1p32f; */ - return ((du_int)high << 32) | low; +// Support for systems that have hardware floating-point; can set the invalid +// flag as a side-effect of computation. + +COMPILER_RT_ABI du_int __fixunsdfdi(double a) { + if (a <= 0.0) + return 0; + su_int high = a / 4294967296.f; // a / 0x1p32f; + su_int low = a - (double)high * 4294967296.f; // high * 0x1p32f; + return ((du_int)high << 32) | low; } #else -/* Support for systems that don't have hardware floating-point; there are no - * flags to set, and we don't want to code-gen to an unknown soft-float - * implementation. - */ +// Support for systems that don't have hardware floating-point; there are no +// flags to set, and we don't want to code-gen to an unknown soft-float +// implementation. typedef du_int fixuint_t; #include "fp_fixuint_impl.inc" -COMPILER_RT_ABI du_int -__fixunsdfdi(fp_t a) { - return __fixuint(a); -} +COMPILER_RT_ABI du_int __fixunsdfdi(fp_t a) { return __fixuint(a); } #endif #if defined(__ARM_EABI__) #if defined(COMPILER_RT_ARMHF_TARGET) -AEABI_RTABI du_int __aeabi_d2ulz(fp_t a) { - return __fixunsdfdi(a); -} +AEABI_RTABI du_int __aeabi_d2ulz(fp_t a) { return __fixunsdfdi(a); } #else -AEABI_RTABI du_int __aeabi_d2ulz(fp_t a) COMPILER_RT_ALIAS(__fixunsdfdi); +COMPILER_RT_ALIAS(__fixunsdfdi, __aeabi_d2ulz) #endif #endif diff --git a/lib/builtins/fixunsdfsi.c b/lib/builtins/fixunsdfsi.c index 3c5355beae1a..3db2adec09eb 100644 --- a/lib/builtins/fixunsdfsi.c +++ b/lib/builtins/fixunsdfsi.c @@ -1,29 +1,22 @@ -/* ===-- fixunsdfsi.c - Implement __fixunsdfsi -----------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - */ +//===-- fixunsdfsi.c - Implement __fixunsdfsi -----------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// #define DOUBLE_PRECISION #include "fp_lib.h" typedef su_int fixuint_t; #include "fp_fixuint_impl.inc" -COMPILER_RT_ABI su_int -__fixunsdfsi(fp_t a) { - return __fixuint(a); -} +COMPILER_RT_ABI su_int __fixunsdfsi(fp_t a) { return __fixuint(a); } #if defined(__ARM_EABI__) #if defined(COMPILER_RT_ARMHF_TARGET) -AEABI_RTABI su_int __aeabi_d2uiz(fp_t a) { - return __fixunsdfsi(a); -} +AEABI_RTABI su_int __aeabi_d2uiz(fp_t a) { return __fixunsdfsi(a); } #else -AEABI_RTABI su_int __aeabi_d2uiz(fp_t a) COMPILER_RT_ALIAS(__fixunsdfsi); +COMPILER_RT_ALIAS(__fixunsdfsi, __aeabi_d2uiz) #endif #endif diff --git a/lib/builtins/fixunsdfti.c b/lib/builtins/fixunsdfti.c index f8046a02632b..be497d0e16af 100644 --- a/lib/builtins/fixunsdfti.c +++ b/lib/builtins/fixunsdfti.c @@ -1,12 +1,10 @@ -/* ===-- fixunsdfti.c - Implement __fixunsdfti -----------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - */ +//===-- fixunsdfti.c - Implement __fixunsdfti -----------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// #include "int_lib.h" @@ -16,8 +14,5 @@ typedef tu_int fixuint_t; #include "fp_fixuint_impl.inc" -COMPILER_RT_ABI tu_int -__fixunsdfti(fp_t a) { - return __fixuint(a); -} -#endif /* CRT_HAS_128BIT */ +COMPILER_RT_ABI tu_int __fixunsdfti(fp_t a) { return __fixuint(a); } +#endif // CRT_HAS_128BIT diff --git a/lib/builtins/fixunssfdi.c b/lib/builtins/fixunssfdi.c index 080a25bb1e99..2b90dafad8c2 100644 --- a/lib/builtins/fixunssfdi.c +++ b/lib/builtins/fixunssfdi.c @@ -1,53 +1,43 @@ -/* ===-- fixunssfdi.c - Implement __fixunssfdi -----------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - */ +//===-- fixunssfdi.c - Implement __fixunssfdi -----------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// #define SINGLE_PRECISION #include "fp_lib.h" #ifndef __SOFT_FP__ -/* Support for systems that have hardware floating-point; can set the invalid - * flag as a side-effect of computation. - */ - -COMPILER_RT_ABI du_int -__fixunssfdi(float a) -{ - if (a <= 0.0f) return 0; - double da = a; - su_int high = da / 4294967296.f; /* da / 0x1p32f; */ - su_int low = da - (double)high * 4294967296.f; /* high * 0x1p32f; */ - return ((du_int)high << 32) | low; +// Support for systems that have hardware floating-point; can set the invalid +// flag as a side-effect of computation. + +COMPILER_RT_ABI du_int __fixunssfdi(float a) { + if (a <= 0.0f) + return 0; + double da = a; + su_int high = da / 4294967296.f; // da / 0x1p32f; + su_int low = da - (double)high * 4294967296.f; // high * 0x1p32f; + return ((du_int)high << 32) | low; } #else -/* Support for systems that don't have hardware floating-point; there are no - * flags to set, and we don't want to code-gen to an unknown soft-float - * implementation. - */ +// Support for systems that don't have hardware floating-point; there are no +// flags to set, and we don't want to code-gen to an unknown soft-float +// implementation. typedef du_int fixuint_t; #include "fp_fixuint_impl.inc" -COMPILER_RT_ABI du_int -__fixunssfdi(fp_t a) { - return __fixuint(a); -} +COMPILER_RT_ABI du_int __fixunssfdi(fp_t a) { return __fixuint(a); } #endif #if defined(__ARM_EABI__) #if defined(COMPILER_RT_ARMHF_TARGET) -AEABI_RTABI du_int __aeabi_f2ulz(fp_t a) { - return __fixunssfdi(a); -} +AEABI_RTABI du_int __aeabi_f2ulz(fp_t a) { return __fixunssfdi(a); } #else -AEABI_RTABI du_int __aeabi_f2ulz(fp_t a) COMPILER_RT_ALIAS(__fixunssfdi); +COMPILER_RT_ALIAS(__fixunssfdi, __aeabi_f2ulz) #endif #endif diff --git a/lib/builtins/fixunssfsi.c b/lib/builtins/fixunssfsi.c index eca2916a5c88..738c1bb95631 100644 --- a/lib/builtins/fixunssfsi.c +++ b/lib/builtins/fixunssfsi.c @@ -1,33 +1,26 @@ -/* ===-- fixunssfsi.c - Implement __fixunssfsi -----------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - * - * This file implements __fixunssfsi for the compiler_rt library. - * - * ===----------------------------------------------------------------------=== - */ +//===-- fixunssfsi.c - Implement __fixunssfsi -----------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements __fixunssfsi for the compiler_rt library. +// +//===----------------------------------------------------------------------===// #define SINGLE_PRECISION #include "fp_lib.h" typedef su_int fixuint_t; #include "fp_fixuint_impl.inc" -COMPILER_RT_ABI su_int -__fixunssfsi(fp_t a) { - return __fixuint(a); -} +COMPILER_RT_ABI su_int __fixunssfsi(fp_t a) { return __fixuint(a); } #if defined(__ARM_EABI__) #if defined(COMPILER_RT_ARMHF_TARGET) -AEABI_RTABI su_int __aeabi_f2uiz(fp_t a) { - return __fixunssfsi(a); -} +AEABI_RTABI su_int __aeabi_f2uiz(fp_t a) { return __fixunssfsi(a); } #else -AEABI_RTABI su_int __aeabi_f2uiz(fp_t a) COMPILER_RT_ALIAS(__fixunssfsi); +COMPILER_RT_ALIAS(__fixunssfsi, __aeabi_f2uiz) #endif #endif diff --git a/lib/builtins/fixunssfti.c b/lib/builtins/fixunssfti.c index 862d7bd6c7af..5525d77f26f6 100644 --- a/lib/builtins/fixunssfti.c +++ b/lib/builtins/fixunssfti.c @@ -1,16 +1,14 @@ -/* ===-- fixunssfti.c - Implement __fixunssfti -----------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - * - * This file implements __fixunssfti for the compiler_rt library. - * - * ===----------------------------------------------------------------------=== - */ +//===-- fixunssfti.c - Implement __fixunssfti -----------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements __fixunssfti for the compiler_rt library. +// +//===----------------------------------------------------------------------===// #define SINGLE_PRECISION #include "fp_lib.h" @@ -19,8 +17,5 @@ typedef tu_int fixuint_t; #include "fp_fixuint_impl.inc" -COMPILER_RT_ABI tu_int -__fixunssfti(fp_t a) { - return __fixuint(a); -} +COMPILER_RT_ABI tu_int __fixunssfti(fp_t a) { return __fixuint(a); } #endif diff --git a/lib/builtins/fixunstfdi.c b/lib/builtins/fixunstfdi.c index b2995f65834a..a0805e63db82 100644 --- a/lib/builtins/fixunstfdi.c +++ b/lib/builtins/fixunstfdi.c @@ -1,12 +1,10 @@ -/* ===-- fixunstfdi.c - Implement __fixunstfdi -----------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - */ +//===-- fixunstfdi.c - Implement __fixunstfdi -----------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// #define QUAD_PRECISION #include "fp_lib.h" @@ -15,8 +13,5 @@ typedef du_int fixuint_t; #include "fp_fixuint_impl.inc" -COMPILER_RT_ABI du_int -__fixunstfdi(fp_t a) { - return __fixuint(a); -} +COMPILER_RT_ABI du_int __fixunstfdi(fp_t a) { return __fixuint(a); } #endif diff --git a/lib/builtins/fixunstfsi.c b/lib/builtins/fixunstfsi.c index b5d3f6a7d38d..3a1320ed3e0a 100644 --- a/lib/builtins/fixunstfsi.c +++ b/lib/builtins/fixunstfsi.c @@ -1,12 +1,10 @@ -/* ===-- fixunstfsi.c - Implement __fixunstfsi -----------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - */ +//===-- fixunstfsi.c - Implement __fixunstfsi -----------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// #define QUAD_PRECISION #include "fp_lib.h" @@ -15,8 +13,5 @@ typedef su_int fixuint_t; #include "fp_fixuint_impl.inc" -COMPILER_RT_ABI su_int -__fixunstfsi(fp_t a) { - return __fixuint(a); -} +COMPILER_RT_ABI su_int __fixunstfsi(fp_t a) { return __fixuint(a); } #endif diff --git a/lib/builtins/fixunstfti.c b/lib/builtins/fixunstfti.c index 22ff9dfc0339..23cd1ab615a7 100644 --- a/lib/builtins/fixunstfti.c +++ b/lib/builtins/fixunstfti.c @@ -1,12 +1,10 @@ -/* ===-- fixunstfsi.c - Implement __fixunstfsi -----------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - */ +//===-- fixunstfsi.c - Implement __fixunstfsi -----------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// #define QUAD_PRECISION #include "fp_lib.h" @@ -15,8 +13,5 @@ typedef tu_int fixuint_t; #include "fp_fixuint_impl.inc" -COMPILER_RT_ABI tu_int -__fixunstfti(fp_t a) { - return __fixuint(a); -} +COMPILER_RT_ABI tu_int __fixunstfti(fp_t a) { return __fixuint(a); } #endif diff --git a/lib/builtins/fixunsxfdi.c b/lib/builtins/fixunsxfdi.c index 075304e78dc9..75c4f093794f 100644 --- a/lib/builtins/fixunsxfdi.c +++ b/lib/builtins/fixunsxfdi.c @@ -1,46 +1,39 @@ -/* ===-- fixunsxfdi.c - Implement __fixunsxfdi -----------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - * - * This file implements __fixunsxfdi for the compiler_rt library. - * - * ===----------------------------------------------------------------------=== - */ +//===-- fixunsxfdi.c - Implement __fixunsxfdi -----------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements __fixunsxfdi for the compiler_rt library. +// +//===----------------------------------------------------------------------===// #if !_ARCH_PPC #include "int_lib.h" -/* Returns: convert a to a unsigned long long, rounding toward zero. - * Negative values all become zero. - */ +// Returns: convert a to a unsigned long long, rounding toward zero. +// Negative values all become zero. -/* Assumption: long double is an intel 80 bit floating point type padded with 6 bytes - * du_int is a 64 bit integral type - * value in long double is representable in du_int or is negative - * (no range checking performed) - */ +// Assumption: long double is an intel 80 bit floating point type padded with 6 +// bytes du_int is a 64 bit integral type value in long double is representable +// in du_int or is negative (no range checking performed) -/* gggg gggg gggg gggg gggg gggg gggg gggg | gggg gggg gggg gggg seee eeee eeee eeee | - * 1mmm mmmm mmmm mmmm mmmm mmmm mmmm mmmm | mmmm mmmm mmmm mmmm mmmm mmmm mmmm mmmm - */ +// gggg gggg gggg gggg gggg gggg gggg gggg | gggg gggg gggg gggg seee eeee eeee +// eeee | 1mmm mmmm mmmm mmmm mmmm mmmm mmmm mmmm | mmmm mmmm mmmm mmmm mmmm +// mmmm mmmm mmmm -COMPILER_RT_ABI du_int -__fixunsxfdi(long double a) -{ - long_double_bits fb; - fb.f = a; - int e = (fb.u.high.s.low & 0x00007FFF) - 16383; - if (e < 0 || (fb.u.high.s.low & 0x00008000)) - return 0; - if ((unsigned)e > sizeof(du_int) * CHAR_BIT) - return ~(du_int)0; - return fb.u.low.all >> (63 - e); +COMPILER_RT_ABI du_int __fixunsxfdi(long double a) { + long_double_bits fb; + fb.f = a; + int e = (fb.u.high.s.low & 0x00007FFF) - 16383; + if (e < 0 || (fb.u.high.s.low & 0x00008000)) + return 0; + if ((unsigned)e > sizeof(du_int) * CHAR_BIT) + return ~(du_int)0; + return fb.u.low.all >> (63 - e); } #endif diff --git a/lib/builtins/fixunsxfsi.c b/lib/builtins/fixunsxfsi.c index c3c70f743de8..1432d8ba92d2 100644 --- a/lib/builtins/fixunsxfsi.c +++ b/lib/builtins/fixunsxfsi.c @@ -1,45 +1,39 @@ -/* ===-- fixunsxfsi.c - Implement __fixunsxfsi -----------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - * - * This file implements __fixunsxfsi for the compiler_rt library. - * - * ===----------------------------------------------------------------------=== - */ +//===-- fixunsxfsi.c - Implement __fixunsxfsi -----------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements __fixunsxfsi for the compiler_rt library. +// +//===----------------------------------------------------------------------===// #if !_ARCH_PPC #include "int_lib.h" -/* Returns: convert a to a unsigned int, rounding toward zero. - * Negative values all become zero. - */ +// Returns: convert a to a unsigned int, rounding toward zero. +// Negative values all become zero. -/* Assumption: long double is an intel 80 bit floating point type padded with 6 bytes - * su_int is a 32 bit integral type - * value in long double is representable in su_int or is negative - */ +// Assumption: long double is an intel 80 bit floating point type padded with 6 +// bytes su_int is a 32 bit integral type value in long double is representable +// in su_int or is negative -/* gggg gggg gggg gggg gggg gggg gggg gggg | gggg gggg gggg gggg seee eeee eeee eeee | - * 1mmm mmmm mmmm mmmm mmmm mmmm mmmm mmmm | mmmm mmmm mmmm mmmm mmmm mmmm mmmm mmmm - */ +// gggg gggg gggg gggg gggg gggg gggg gggg | gggg gggg gggg gggg seee eeee eeee +// eeee | 1mmm mmmm mmmm mmmm mmmm mmmm mmmm mmmm | mmmm mmmm mmmm mmmm mmmm +// mmmm mmmm mmmm -COMPILER_RT_ABI su_int -__fixunsxfsi(long double a) -{ - long_double_bits fb; - fb.f = a; - int e = (fb.u.high.s.low & 0x00007FFF) - 16383; - if (e < 0 || (fb.u.high.s.low & 0x00008000)) - return 0; - if ((unsigned)e > sizeof(su_int) * CHAR_BIT) - return ~(su_int)0; - return fb.u.low.s.high >> (31 - e); +COMPILER_RT_ABI su_int __fixunsxfsi(long double a) { + long_double_bits fb; + fb.f = a; + int e = (fb.u.high.s.low & 0x00007FFF) - 16383; + if (e < 0 || (fb.u.high.s.low & 0x00008000)) + return 0; + if ((unsigned)e > sizeof(su_int) * CHAR_BIT) + return ~(su_int)0; + return fb.u.low.s.high >> (31 - e); } -#endif /* !_ARCH_PPC */ +#endif // !_ARCH_PPC diff --git a/lib/builtins/fixunsxfti.c b/lib/builtins/fixunsxfti.c index fb39d00ff5b2..508554e4f8f6 100644 --- a/lib/builtins/fixunsxfti.c +++ b/lib/builtins/fixunsxfti.c @@ -1,50 +1,44 @@ -/* ===-- fixunsxfti.c - Implement __fixunsxfti -----------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - * - * This file implements __fixunsxfti for the compiler_rt library. - * - * ===----------------------------------------------------------------------=== - */ +//===-- fixunsxfti.c - Implement __fixunsxfti -----------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements __fixunsxfti for the compiler_rt library. +// +//===----------------------------------------------------------------------===// #include "int_lib.h" #ifdef CRT_HAS_128BIT -/* Returns: convert a to a unsigned long long, rounding toward zero. - * Negative values all become zero. - */ +// Returns: convert a to a unsigned long long, rounding toward zero. +// Negative values all become zero. -/* Assumption: long double is an intel 80 bit floating point type padded with 6 bytes - * tu_int is a 128 bit integral type - * value in long double is representable in tu_int or is negative - */ +// Assumption: long double is an intel 80 bit floating point type padded with 6 +// bytes tu_int is a 128 bit integral type value in long double is representable +// in tu_int or is negative -/* gggg gggg gggg gggg gggg gggg gggg gggg | gggg gggg gggg gggg seee eeee eeee eeee | - * 1mmm mmmm mmmm mmmm mmmm mmmm mmmm mmmm | mmmm mmmm mmmm mmmm mmmm mmmm mmmm mmmm - */ +// gggg gggg gggg gggg gggg gggg gggg gggg | gggg gggg gggg gggg seee eeee eeee +// eeee | 1mmm mmmm mmmm mmmm mmmm mmmm mmmm mmmm | mmmm mmmm mmmm mmmm mmmm +// mmmm mmmm mmmm -COMPILER_RT_ABI tu_int -__fixunsxfti(long double a) -{ - long_double_bits fb; - fb.f = a; - int e = (fb.u.high.s.low & 0x00007FFF) - 16383; - if (e < 0 || (fb.u.high.s.low & 0x00008000)) - return 0; - if ((unsigned)e > sizeof(tu_int) * CHAR_BIT) - return ~(tu_int)0; - tu_int r = fb.u.low.all; - if (e > 63) - r <<= (e - 63); - else - r >>= (63 - e); - return r; +COMPILER_RT_ABI tu_int __fixunsxfti(long double a) { + long_double_bits fb; + fb.f = a; + int e = (fb.u.high.s.low & 0x00007FFF) - 16383; + if (e < 0 || (fb.u.high.s.low & 0x00008000)) + return 0; + if ((unsigned)e > sizeof(tu_int) * CHAR_BIT) + return ~(tu_int)0; + tu_int r = fb.u.low.all; + if (e > 63) + r <<= (e - 63); + else + r >>= (63 - e); + return r; } -#endif /* CRT_HAS_128BIT */ +#endif // CRT_HAS_128BIT diff --git a/lib/builtins/fixxfdi.c b/lib/builtins/fixxfdi.c index 011787f9e4b4..4783c0101740 100644 --- a/lib/builtins/fixxfdi.c +++ b/lib/builtins/fixxfdi.c @@ -1,48 +1,43 @@ -/* ===-- fixxfdi.c - Implement __fixxfdi -----------------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - * - * This file implements __fixxfdi for the compiler_rt library. - * - * ===----------------------------------------------------------------------=== - */ +//===-- fixxfdi.c - Implement __fixxfdi -----------------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements __fixxfdi for the compiler_rt library. +// +//===----------------------------------------------------------------------===// #if !_ARCH_PPC #include "int_lib.h" -/* Returns: convert a to a signed long long, rounding toward zero. */ +// Returns: convert a to a signed long long, rounding toward zero. -/* Assumption: long double is an intel 80 bit floating point type padded with 6 bytes - * di_int is a 64 bit integral type - * value in long double is representable in di_int (no range checking performed) - */ +// Assumption: long double is an intel 80 bit floating point type padded with 6 +// bytes di_int is a 64 bit integral type value in long double is representable +// in di_int (no range checking performed) -/* gggg gggg gggg gggg gggg gggg gggg gggg | gggg gggg gggg gggg seee eeee eeee eeee | - * 1mmm mmmm mmmm mmmm mmmm mmmm mmmm mmmm | mmmm mmmm mmmm mmmm mmmm mmmm mmmm mmmm - */ +// gggg gggg gggg gggg gggg gggg gggg gggg | gggg gggg gggg gggg seee eeee eeee +// eeee | 1mmm mmmm mmmm mmmm mmmm mmmm mmmm mmmm | mmmm mmmm mmmm mmmm mmmm +// mmmm mmmm mmmm -COMPILER_RT_ABI di_int -__fixxfdi(long double a) -{ - const di_int di_max = (di_int)((~(du_int)0) / 2); - const di_int di_min = -di_max - 1; - long_double_bits fb; - fb.f = a; - int e = (fb.u.high.s.low & 0x00007FFF) - 16383; - if (e < 0) - return 0; - if ((unsigned)e >= sizeof(di_int) * CHAR_BIT) - return a > 0 ? di_max : di_min; - di_int s = -(si_int)((fb.u.high.s.low & 0x00008000) >> 15); - di_int r = fb.u.low.all; - r = (du_int)r >> (63 - e); - return (r ^ s) - s; +COMPILER_RT_ABI di_int __fixxfdi(long double a) { + const di_int di_max = (di_int)((~(du_int)0) / 2); + const di_int di_min = -di_max - 1; + long_double_bits fb; + fb.f = a; + int e = (fb.u.high.s.low & 0x00007FFF) - 16383; + if (e < 0) + return 0; + if ((unsigned)e >= sizeof(di_int) * CHAR_BIT) + return a > 0 ? di_max : di_min; + di_int s = -(si_int)((fb.u.high.s.low & 0x00008000) >> 15); + di_int r = fb.u.low.all; + r = (du_int)r >> (63 - e); + return (r ^ s) - s; } -#endif /* !_ARCH_PPC */ +#endif // !_ARCH_PPC diff --git a/lib/builtins/fixxfti.c b/lib/builtins/fixxfti.c index 968a4f0d5eea..90e03116e7ca 100644 --- a/lib/builtins/fixxfti.c +++ b/lib/builtins/fixxfti.c @@ -1,51 +1,46 @@ -/* ===-- fixxfti.c - Implement __fixxfti -----------------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - * - * This file implements __fixxfti for the compiler_rt library. - * - * ===----------------------------------------------------------------------=== - */ +//===-- fixxfti.c - Implement __fixxfti -----------------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements __fixxfti for the compiler_rt library. +// +//===----------------------------------------------------------------------===// #include "int_lib.h" #ifdef CRT_HAS_128BIT -/* Returns: convert a to a signed long long, rounding toward zero. */ +// Returns: convert a to a signed long long, rounding toward zero. -/* Assumption: long double is an intel 80 bit floating point type padded with 6 bytes - * ti_int is a 128 bit integral type - * value in long double is representable in ti_int - */ +// Assumption: long double is an intel 80 bit floating point type padded with 6 +// bytes ti_int is a 128 bit integral type value in long double is representable +// in ti_int -/* gggg gggg gggg gggg gggg gggg gggg gggg | gggg gggg gggg gggg seee eeee eeee eeee | - * 1mmm mmmm mmmm mmmm mmmm mmmm mmmm mmmm | mmmm mmmm mmmm mmmm mmmm mmmm mmmm mmmm - */ +// gggg gggg gggg gggg gggg gggg gggg gggg | gggg gggg gggg gggg seee eeee eeee +// eeee | 1mmm mmmm mmmm mmmm mmmm mmmm mmmm mmmm | mmmm mmmm mmmm mmmm mmmm +// mmmm mmmm mmmm -COMPILER_RT_ABI ti_int -__fixxfti(long double a) -{ - const ti_int ti_max = (ti_int)((~(tu_int)0) / 2); - const ti_int ti_min = -ti_max - 1; - long_double_bits fb; - fb.f = a; - int e = (fb.u.high.s.low & 0x00007FFF) - 16383; - if (e < 0) - return 0; - ti_int s = -(si_int)((fb.u.high.s.low & 0x00008000) >> 15); - ti_int r = fb.u.low.all; - if ((unsigned)e >= sizeof(ti_int) * CHAR_BIT) - return a > 0 ? ti_max : ti_min; - if (e > 63) - r <<= (e - 63); - else - r >>= (63 - e); - return (r ^ s) - s; +COMPILER_RT_ABI ti_int __fixxfti(long double a) { + const ti_int ti_max = (ti_int)((~(tu_int)0) / 2); + const ti_int ti_min = -ti_max - 1; + long_double_bits fb; + fb.f = a; + int e = (fb.u.high.s.low & 0x00007FFF) - 16383; + if (e < 0) + return 0; + ti_int s = -(si_int)((fb.u.high.s.low & 0x00008000) >> 15); + ti_int r = fb.u.low.all; + if ((unsigned)e >= sizeof(ti_int) * CHAR_BIT) + return a > 0 ? ti_max : ti_min; + if (e > 63) + r <<= (e - 63); + else + r >>= (63 - e); + return (r ^ s) - s; } -#endif /* CRT_HAS_128BIT */ +#endif // CRT_HAS_128BIT diff --git a/lib/builtins/floatdidf.c b/lib/builtins/floatdidf.c index 36b856e078d4..8f887314b9e1 100644 --- a/lib/builtins/floatdidf.c +++ b/lib/builtins/floatdidf.c @@ -1,115 +1,103 @@ -/*===-- floatdidf.c - Implement __floatdidf -------------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - *===----------------------------------------------------------------------=== - * - * This file implements __floatdidf for the compiler_rt library. - * - *===----------------------------------------------------------------------=== - */ +//===-- floatdidf.c - Implement __floatdidf -------------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements __floatdidf for the compiler_rt library. +// +//===----------------------------------------------------------------------===// #include "int_lib.h" -/* Returns: convert a to a double, rounding toward even. */ +// Returns: convert a to a double, rounding toward even. -/* Assumption: double is a IEEE 64 bit floating point type - * di_int is a 64 bit integral type - */ +// Assumption: double is a IEEE 64 bit floating point type +// di_int is a 64 bit integral type -/* seee eeee eeee mmmm mmmm mmmm mmmm mmmm | mmmm mmmm mmmm mmmm mmmm mmmm mmmm mmmm */ +// seee eeee eeee mmmm mmmm mmmm mmmm mmmm | mmmm mmmm mmmm mmmm mmmm mmmm mmmm +// mmmm #ifndef __SOFT_FP__ -/* Support for systems that have hardware floating-point; we'll set the inexact flag - * as a side-effect of this computation. - */ +// Support for systems that have hardware floating-point; we'll set the inexact +// flag as a side-effect of this computation. -COMPILER_RT_ABI double -__floatdidf(di_int a) -{ - static const double twop52 = 4503599627370496.0; // 0x1.0p52 - static const double twop32 = 4294967296.0; // 0x1.0p32 +COMPILER_RT_ABI double __floatdidf(di_int a) { + static const double twop52 = 4503599627370496.0; // 0x1.0p52 + static const double twop32 = 4294967296.0; // 0x1.0p32 - union { int64_t x; double d; } low = { .d = twop52 }; + union { + int64_t x; + double d; + } low = {.d = twop52}; - const double high = (int32_t)(a >> 32) * twop32; - low.x |= a & INT64_C(0x00000000ffffffff); + const double high = (int32_t)(a >> 32) * twop32; + low.x |= a & INT64_C(0x00000000ffffffff); - const double result = (high - twop52) + low.d; - return result; + const double result = (high - twop52) + low.d; + return result; } #else -/* Support for systems that don't have hardware floating-point; there are no flags to - * set, and we don't want to code-gen to an unknown soft-float implementation. - */ +// Support for systems that don't have hardware floating-point; there are no +// flags to set, and we don't want to code-gen to an unknown soft-float +// implementation. -COMPILER_RT_ABI double -__floatdidf(di_int a) -{ - if (a == 0) - return 0.0; - const unsigned N = sizeof(di_int) * CHAR_BIT; - const di_int s = a >> (N-1); - a = (a ^ s) - s; - int sd = N - __builtin_clzll(a); /* number of significant digits */ - int e = sd - 1; /* exponent */ - if (sd > DBL_MANT_DIG) - { - /* start: 0000000000000000000001xxxxxxxxxxxxxxxxxxxxxxPQxxxxxxxxxxxxxxxxxx - * finish: 000000000000000000000000000000000000001xxxxxxxxxxxxxxxxxxxxxxPQR - * 12345678901234567890123456 - * 1 = msb 1 bit - * P = bit DBL_MANT_DIG-1 bits to the right of 1 - * Q = bit DBL_MANT_DIG bits to the right of 1 - * R = "or" of all bits to the right of Q - */ - switch (sd) - { - case DBL_MANT_DIG + 1: - a <<= 1; - break; - case DBL_MANT_DIG + 2: - break; - default: - a = ((du_int)a >> (sd - (DBL_MANT_DIG+2))) | - ((a & ((du_int)(-1) >> ((N + DBL_MANT_DIG+2) - sd))) != 0); - }; - /* finish: */ - a |= (a & 4) != 0; /* Or P into R */ - ++a; /* round - this step may add a significant bit */ - a >>= 2; /* dump Q and R */ - /* a is now rounded to DBL_MANT_DIG or DBL_MANT_DIG+1 bits */ - if (a & ((du_int)1 << DBL_MANT_DIG)) - { - a >>= 1; - ++e; - } - /* a is now rounded to DBL_MANT_DIG bits */ +COMPILER_RT_ABI double __floatdidf(di_int a) { + if (a == 0) + return 0.0; + const unsigned N = sizeof(di_int) * CHAR_BIT; + const di_int s = a >> (N - 1); + a = (a ^ s) - s; + int sd = N - __builtin_clzll(a); // number of significant digits + int e = sd - 1; // exponent + if (sd > DBL_MANT_DIG) { + // start: 0000000000000000000001xxxxxxxxxxxxxxxxxxxxxxPQxxxxxxxxxxxxxxxxxx + // finish: 000000000000000000000000000000000000001xxxxxxxxxxxxxxxxxxxxxxPQR + // 12345678901234567890123456 + // 1 = msb 1 bit + // P = bit DBL_MANT_DIG-1 bits to the right of 1 + // Q = bit DBL_MANT_DIG bits to the right of 1 + // R = "or" of all bits to the right of Q + switch (sd) { + case DBL_MANT_DIG + 1: + a <<= 1; + break; + case DBL_MANT_DIG + 2: + break; + default: + a = ((du_int)a >> (sd - (DBL_MANT_DIG + 2))) | + ((a & ((du_int)(-1) >> ((N + DBL_MANT_DIG + 2) - sd))) != 0); + }; + // finish: + a |= (a & 4) != 0; // Or P into R + ++a; // round - this step may add a significant bit + a >>= 2; // dump Q and R + // a is now rounded to DBL_MANT_DIG or DBL_MANT_DIG+1 bits + if (a & ((du_int)1 << DBL_MANT_DIG)) { + a >>= 1; + ++e; } - else - { - a <<= (DBL_MANT_DIG - sd); - /* a is now rounded to DBL_MANT_DIG bits */ - } - double_bits fb; - fb.u.s.high = ((su_int)s & 0x80000000) | /* sign */ - ((e + 1023) << 20) | /* exponent */ - ((su_int)(a >> 32) & 0x000FFFFF); /* mantissa-high */ - fb.u.s.low = (su_int)a; /* mantissa-low */ - return fb.f; + // a is now rounded to DBL_MANT_DIG bits + } else { + a <<= (DBL_MANT_DIG - sd); + // a is now rounded to DBL_MANT_DIG bits + } + double_bits fb; + fb.u.s.high = ((su_int)s & 0x80000000) | // sign + ((e + 1023) << 20) | // exponent + ((su_int)(a >> 32) & 0x000FFFFF); // mantissa-high + fb.u.s.low = (su_int)a; // mantissa-low + return fb.f; } #endif #if defined(__ARM_EABI__) #if defined(COMPILER_RT_ARMHF_TARGET) -AEABI_RTABI double __aeabi_l2d(di_int a) { - return __floatdidf(a); -} +AEABI_RTABI double __aeabi_l2d(di_int a) { return __floatdidf(a); } #else -AEABI_RTABI double __aeabi_l2d(di_int a) COMPILER_RT_ALIAS(__floatdidf); +COMPILER_RT_ALIAS(__floatdidf, __aeabi_l2d) #endif #endif diff --git a/lib/builtins/floatdisf.c b/lib/builtins/floatdisf.c index a2f09eb2ed2c..cd9e0a3b78a5 100644 --- a/lib/builtins/floatdisf.c +++ b/lib/builtins/floatdisf.c @@ -1,88 +1,75 @@ -/*===-- floatdisf.c - Implement __floatdisf -------------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - *===----------------------------------------------------------------------=== - * - * This file implements __floatdisf for the compiler_rt library. - * - *===----------------------------------------------------------------------=== - */ +//===-- floatdisf.c - Implement __floatdisf -------------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements __floatdisf for the compiler_rt library. +// +//===----------------------------------------------------------------------===// -/* Returns: convert a to a float, rounding toward even.*/ +// Returns: convert a to a float, rounding toward even. -/* Assumption: float is a IEEE 32 bit floating point type - * di_int is a 64 bit integral type - */ +// Assumption: float is a IEEE 32 bit floating point type +// di_int is a 64 bit integral type -/* seee eeee emmm mmmm mmmm mmmm mmmm mmmm */ +// seee eeee emmm mmmm mmmm mmmm mmmm mmmm #include "int_lib.h" -COMPILER_RT_ABI float -__floatdisf(di_int a) -{ - if (a == 0) - return 0.0F; - const unsigned N = sizeof(di_int) * CHAR_BIT; - const di_int s = a >> (N-1); - a = (a ^ s) - s; - int sd = N - __builtin_clzll(a); /* number of significant digits */ - int e = sd - 1; /* exponent */ - if (sd > FLT_MANT_DIG) - { - /* start: 0000000000000000000001xxxxxxxxxxxxxxxxxxxxxxPQxxxxxxxxxxxxxxxxxx - * finish: 000000000000000000000000000000000000001xxxxxxxxxxxxxxxxxxxxxxPQR - * 12345678901234567890123456 - * 1 = msb 1 bit - * P = bit FLT_MANT_DIG-1 bits to the right of 1 - * Q = bit FLT_MANT_DIG bits to the right of 1 - * R = "or" of all bits to the right of Q - */ - switch (sd) - { - case FLT_MANT_DIG + 1: - a <<= 1; - break; - case FLT_MANT_DIG + 2: - break; - default: - a = ((du_int)a >> (sd - (FLT_MANT_DIG+2))) | - ((a & ((du_int)(-1) >> ((N + FLT_MANT_DIG+2) - sd))) != 0); - }; - /* finish: */ - a |= (a & 4) != 0; /* Or P into R */ - ++a; /* round - this step may add a significant bit */ - a >>= 2; /* dump Q and R */ - /* a is now rounded to FLT_MANT_DIG or FLT_MANT_DIG+1 bits */ - if (a & ((du_int)1 << FLT_MANT_DIG)) - { - a >>= 1; - ++e; - } - /* a is now rounded to FLT_MANT_DIG bits */ +COMPILER_RT_ABI float __floatdisf(di_int a) { + if (a == 0) + return 0.0F; + const unsigned N = sizeof(di_int) * CHAR_BIT; + const di_int s = a >> (N - 1); + a = (a ^ s) - s; + int sd = N - __builtin_clzll(a); // number of significant digits + int e = sd - 1; // exponent + if (sd > FLT_MANT_DIG) { + // start: 0000000000000000000001xxxxxxxxxxxxxxxxxxxxxxPQxxxxxxxxxxxxxxxxxx + // finish: 000000000000000000000000000000000000001xxxxxxxxxxxxxxxxxxxxxxPQR + // 12345678901234567890123456 + // 1 = msb 1 bit + // P = bit FLT_MANT_DIG-1 bits to the right of 1 + // Q = bit FLT_MANT_DIG bits to the right of 1 + // R = "or" of all bits to the right of Q + switch (sd) { + case FLT_MANT_DIG + 1: + a <<= 1; + break; + case FLT_MANT_DIG + 2: + break; + default: + a = ((du_int)a >> (sd - (FLT_MANT_DIG + 2))) | + ((a & ((du_int)(-1) >> ((N + FLT_MANT_DIG + 2) - sd))) != 0); + }; + // finish: + a |= (a & 4) != 0; // Or P into R + ++a; // round - this step may add a significant bit + a >>= 2; // dump Q and R + // a is now rounded to FLT_MANT_DIG or FLT_MANT_DIG+1 bits + if (a & ((du_int)1 << FLT_MANT_DIG)) { + a >>= 1; + ++e; } - else - { - a <<= (FLT_MANT_DIG - sd); - /* a is now rounded to FLT_MANT_DIG bits */ - } - float_bits fb; - fb.u = ((su_int)s & 0x80000000) | /* sign */ - ((e + 127) << 23) | /* exponent */ - ((su_int)a & 0x007FFFFF); /* mantissa */ - return fb.f; + // a is now rounded to FLT_MANT_DIG bits + } else { + a <<= (FLT_MANT_DIG - sd); + // a is now rounded to FLT_MANT_DIG bits + } + float_bits fb; + fb.u = ((su_int)s & 0x80000000) | // sign + ((e + 127) << 23) | // exponent + ((su_int)a & 0x007FFFFF); // mantissa + return fb.f; } #if defined(__ARM_EABI__) #if defined(COMPILER_RT_ARMHF_TARGET) -AEABI_RTABI float __aeabi_l2f(di_int a) { - return __floatdisf(a); -} +AEABI_RTABI float __aeabi_l2f(di_int a) { return __floatdisf(a); } #else -AEABI_RTABI float __aeabi_l2f(di_int a) COMPILER_RT_ALIAS(__floatdisf); +COMPILER_RT_ALIAS(__floatdisf, __aeabi_l2f) #endif #endif diff --git a/lib/builtins/floatditf.c b/lib/builtins/floatditf.c index cd51dd8aade4..9b07b65825b8 100644 --- a/lib/builtins/floatditf.c +++ b/lib/builtins/floatditf.c @@ -1,9 +1,8 @@ //===-- lib/floatditf.c - integer -> quad-precision conversion ----*- C -*-===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// // @@ -19,32 +18,32 @@ #if defined(CRT_HAS_128BIT) && defined(CRT_LDBL_128BIT) COMPILER_RT_ABI fp_t __floatditf(di_int a) { - const int aWidth = sizeof a * CHAR_BIT; + const int aWidth = sizeof a * CHAR_BIT; - // Handle zero as a special case to protect clz - if (a == 0) - return fromRep(0); + // Handle zero as a special case to protect clz + if (a == 0) + return fromRep(0); - // All other cases begin by extracting the sign and absolute value of a - rep_t sign = 0; - du_int aAbs = (du_int)a; - if (a < 0) { - sign = signBit; - aAbs = ~(du_int)a + 1U; - } + // All other cases begin by extracting the sign and absolute value of a + rep_t sign = 0; + du_int aAbs = (du_int)a; + if (a < 0) { + sign = signBit; + aAbs = ~(du_int)a + 1U; + } - // Exponent of (fp_t)a is the width of abs(a). - const int exponent = (aWidth - 1) - __builtin_clzll(aAbs); - rep_t result; + // Exponent of (fp_t)a is the width of abs(a). + const int exponent = (aWidth - 1) - __builtin_clzll(aAbs); + rep_t result; - // Shift a into the significand field, rounding if it is a right-shift - const int shift = significandBits - exponent; - result = (rep_t)aAbs << shift ^ implicitBit; + // Shift a into the significand field, rounding if it is a right-shift + const int shift = significandBits - exponent; + result = (rep_t)aAbs << shift ^ implicitBit; - // Insert the exponent - result += (rep_t)(exponent + exponentBias) << significandBits; - // Insert the sign bit and return - return fromRep(result | sign); + // Insert the exponent + result += (rep_t)(exponent + exponentBias) << significandBits; + // Insert the sign bit and return + return fromRep(result | sign); } #endif diff --git a/lib/builtins/floatdixf.c b/lib/builtins/floatdixf.c index d39e81d7ca75..ad5deb2d4bf5 100644 --- a/lib/builtins/floatdixf.c +++ b/lib/builtins/floatdixf.c @@ -1,46 +1,41 @@ -/* ===-- floatdixf.c - Implement __floatdixf -------------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - * - * This file implements __floatdixf for the compiler_rt library. - * - * ===----------------------------------------------------------------------=== - */ +//===-- floatdixf.c - Implement __floatdixf -------------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements __floatdixf for the compiler_rt library. +// +//===----------------------------------------------------------------------===// #if !_ARCH_PPC #include "int_lib.h" -/* Returns: convert a to a long double, rounding toward even. */ +// Returns: convert a to a long double, rounding toward even. -/* Assumption: long double is a IEEE 80 bit floating point type padded to 128 bits - * di_int is a 64 bit integral type - */ +// Assumption: long double is a IEEE 80 bit floating point type padded to 128 +// bits di_int is a 64 bit integral type -/* gggg gggg gggg gggg gggg gggg gggg gggg | gggg gggg gggg gggg seee eeee eeee eeee | - * 1mmm mmmm mmmm mmmm mmmm mmmm mmmm mmmm | mmmm mmmm mmmm mmmm mmmm mmmm mmmm mmmm - */ +// gggg gggg gggg gggg gggg gggg gggg gggg | gggg gggg gggg gggg seee eeee eeee +// eeee | 1mmm mmmm mmmm mmmm mmmm mmmm mmmm mmmm | mmmm mmmm mmmm mmmm mmmm +// mmmm mmmm mmmm -COMPILER_RT_ABI long double -__floatdixf(di_int a) -{ - if (a == 0) - return 0.0; - const unsigned N = sizeof(di_int) * CHAR_BIT; - const di_int s = a >> (N-1); - a = (a ^ s) - s; - int clz = __builtin_clzll(a); - int e = (N - 1) - clz ; /* exponent */ - long_double_bits fb; - fb.u.high.s.low = ((su_int)s & 0x00008000) | /* sign */ - (e + 16383); /* exponent */ - fb.u.low.all = a << clz; /* mantissa */ - return fb.f; +COMPILER_RT_ABI long double __floatdixf(di_int a) { + if (a == 0) + return 0.0; + const unsigned N = sizeof(di_int) * CHAR_BIT; + const di_int s = a >> (N - 1); + a = (a ^ s) - s; + int clz = __builtin_clzll(a); + int e = (N - 1) - clz; // exponent + long_double_bits fb; + fb.u.high.s.low = ((su_int)s & 0x00008000) | // sign + (e + 16383); // exponent + fb.u.low.all = a << clz; // mantissa + return fb.f; } -#endif /* !_ARCH_PPC */ +#endif // !_ARCH_PPC diff --git a/lib/builtins/floatsidf.c b/lib/builtins/floatsidf.c index fe051123ce7c..2c66167d794d 100644 --- a/lib/builtins/floatsidf.c +++ b/lib/builtins/floatsidf.c @@ -1,9 +1,8 @@ //===-- lib/floatsidf.c - integer -> double-precision conversion --*- C -*-===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// // @@ -18,44 +17,41 @@ #include "int_lib.h" -COMPILER_RT_ABI fp_t -__floatsidf(int a) { - - const int aWidth = sizeof a * CHAR_BIT; - - // Handle zero as a special case to protect clz - if (a == 0) - return fromRep(0); - - // All other cases begin by extracting the sign and absolute value of a - rep_t sign = 0; - if (a < 0) { - sign = signBit; - a = -a; - } - - // Exponent of (fp_t)a is the width of abs(a). - const int exponent = (aWidth - 1) - __builtin_clz(a); - rep_t result; - - // Shift a into the significand field and clear the implicit bit. Extra - // cast to unsigned int is necessary to get the correct behavior for - // the input INT_MIN. - const int shift = significandBits - exponent; - result = (rep_t)(unsigned int)a << shift ^ implicitBit; - - // Insert the exponent - result += (rep_t)(exponent + exponentBias) << significandBits; - // Insert the sign bit and return - return fromRep(result | sign); +COMPILER_RT_ABI fp_t __floatsidf(int a) { + + const int aWidth = sizeof a * CHAR_BIT; + + // Handle zero as a special case to protect clz + if (a == 0) + return fromRep(0); + + // All other cases begin by extracting the sign and absolute value of a + rep_t sign = 0; + if (a < 0) { + sign = signBit; + a = -a; + } + + // Exponent of (fp_t)a is the width of abs(a). + const int exponent = (aWidth - 1) - __builtin_clz(a); + rep_t result; + + // Shift a into the significand field and clear the implicit bit. Extra + // cast to unsigned int is necessary to get the correct behavior for + // the input INT_MIN. + const int shift = significandBits - exponent; + result = (rep_t)(unsigned int)a << shift ^ implicitBit; + + // Insert the exponent + result += (rep_t)(exponent + exponentBias) << significandBits; + // Insert the sign bit and return + return fromRep(result | sign); } #if defined(__ARM_EABI__) #if defined(COMPILER_RT_ARMHF_TARGET) -AEABI_RTABI fp_t __aeabi_i2d(int a) { - return __floatsidf(a); -} +AEABI_RTABI fp_t __aeabi_i2d(int a) { return __floatsidf(a); } #else -AEABI_RTABI fp_t __aeabi_i2d(int a) COMPILER_RT_ALIAS(__floatsidf); +COMPILER_RT_ALIAS(__floatsidf, __aeabi_i2d) #endif #endif diff --git a/lib/builtins/floatsisf.c b/lib/builtins/floatsisf.c index bf087ee3c295..fe060407755b 100644 --- a/lib/builtins/floatsisf.c +++ b/lib/builtins/floatsisf.c @@ -1,9 +1,8 @@ //===-- lib/floatsisf.c - integer -> single-precision conversion --*- C -*-===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// // @@ -18,50 +17,49 @@ #include "int_lib.h" -COMPILER_RT_ABI fp_t -__floatsisf(int a) { - - const int aWidth = sizeof a * CHAR_BIT; - - // Handle zero as a special case to protect clz - if (a == 0) - return fromRep(0); - - // All other cases begin by extracting the sign and absolute value of a - rep_t sign = 0; - if (a < 0) { - sign = signBit; - a = -a; - } - - // Exponent of (fp_t)a is the width of abs(a). - const int exponent = (aWidth - 1) - __builtin_clz(a); - rep_t result; - - // Shift a into the significand field, rounding if it is a right-shift - if (exponent <= significandBits) { - const int shift = significandBits - exponent; - result = (rep_t)a << shift ^ implicitBit; - } else { - const int shift = exponent - significandBits; - result = (rep_t)a >> shift ^ implicitBit; - rep_t round = (rep_t)a << (typeWidth - shift); - if (round > signBit) result++; - if (round == signBit) result += result & 1; - } - - // Insert the exponent - result += (rep_t)(exponent + exponentBias) << significandBits; - // Insert the sign bit and return - return fromRep(result | sign); +COMPILER_RT_ABI fp_t __floatsisf(int a) { + + const int aWidth = sizeof a * CHAR_BIT; + + // Handle zero as a special case to protect clz + if (a == 0) + return fromRep(0); + + // All other cases begin by extracting the sign and absolute value of a + rep_t sign = 0; + if (a < 0) { + sign = signBit; + a = -a; + } + + // Exponent of (fp_t)a is the width of abs(a). + const int exponent = (aWidth - 1) - __builtin_clz(a); + rep_t result; + + // Shift a into the significand field, rounding if it is a right-shift + if (exponent <= significandBits) { + const int shift = significandBits - exponent; + result = (rep_t)a << shift ^ implicitBit; + } else { + const int shift = exponent - significandBits; + result = (rep_t)a >> shift ^ implicitBit; + rep_t round = (rep_t)a << (typeWidth - shift); + if (round > signBit) + result++; + if (round == signBit) + result += result & 1; + } + + // Insert the exponent + result += (rep_t)(exponent + exponentBias) << significandBits; + // Insert the sign bit and return + return fromRep(result | sign); } #if defined(__ARM_EABI__) #if defined(COMPILER_RT_ARMHF_TARGET) -AEABI_RTABI fp_t __aeabi_i2f(int a) { - return __floatsisf(a); -} +AEABI_RTABI fp_t __aeabi_i2f(int a) { return __floatsisf(a); } #else -AEABI_RTABI fp_t __aeabi_i2f(int a) COMPILER_RT_ALIAS(__floatsisf); +COMPILER_RT_ALIAS(__floatsisf, __aeabi_i2f) #endif #endif diff --git a/lib/builtins/floatsitf.c b/lib/builtins/floatsitf.c index f0abca363b5e..f56063f368d9 100644 --- a/lib/builtins/floatsitf.c +++ b/lib/builtins/floatsitf.c @@ -1,9 +1,8 @@ //===-- lib/floatsitf.c - integer -> quad-precision conversion ----*- C -*-===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// // @@ -19,32 +18,32 @@ #if defined(CRT_HAS_128BIT) && defined(CRT_LDBL_128BIT) COMPILER_RT_ABI fp_t __floatsitf(int a) { - const int aWidth = sizeof a * CHAR_BIT; + const int aWidth = sizeof a * CHAR_BIT; - // Handle zero as a special case to protect clz - if (a == 0) - return fromRep(0); + // Handle zero as a special case to protect clz + if (a == 0) + return fromRep(0); - // All other cases begin by extracting the sign and absolute value of a - rep_t sign = 0; - unsigned aAbs = (unsigned)a; - if (a < 0) { - sign = signBit; - aAbs = ~(unsigned)a + 1U; - } + // All other cases begin by extracting the sign and absolute value of a + rep_t sign = 0; + unsigned aAbs = (unsigned)a; + if (a < 0) { + sign = signBit; + aAbs = ~(unsigned)a + 1U; + } - // Exponent of (fp_t)a is the width of abs(a). - const int exponent = (aWidth - 1) - __builtin_clz(aAbs); - rep_t result; + // Exponent of (fp_t)a is the width of abs(a). + const int exponent = (aWidth - 1) - __builtin_clz(aAbs); + rep_t result; - // Shift a into the significand field and clear the implicit bit. - const int shift = significandBits - exponent; - result = (rep_t)aAbs << shift ^ implicitBit; + // Shift a into the significand field and clear the implicit bit. + const int shift = significandBits - exponent; + result = (rep_t)aAbs << shift ^ implicitBit; - // Insert the exponent - result += (rep_t)(exponent + exponentBias) << significandBits; - // Insert the sign bit and return - return fromRep(result | sign); + // Insert the exponent + result += (rep_t)(exponent + exponentBias) << significandBits; + // Insert the sign bit and return + return fromRep(result | sign); } #endif diff --git a/lib/builtins/floattidf.c b/lib/builtins/floattidf.c index 2702a3c8a2db..0a1c04bec82e 100644 --- a/lib/builtins/floattidf.c +++ b/lib/builtins/floattidf.c @@ -1,83 +1,73 @@ -/* ===-- floattidf.c - Implement __floattidf -------------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - * - * This file implements __floattidf for the compiler_rt library. - * - * ===----------------------------------------------------------------------=== - */ +//===-- floattidf.c - Implement __floattidf -------------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements __floattidf for the compiler_rt library. +// +//===----------------------------------------------------------------------===// #include "int_lib.h" #ifdef CRT_HAS_128BIT -/* Returns: convert a to a double, rounding toward even.*/ +// Returns: convert a to a double, rounding toward even. -/* Assumption: double is a IEEE 64 bit floating point type - * ti_int is a 128 bit integral type - */ +// Assumption: double is a IEEE 64 bit floating point type +// ti_int is a 128 bit integral type -/* seee eeee eeee mmmm mmmm mmmm mmmm mmmm | mmmm mmmm mmmm mmmm mmmm mmmm mmmm mmmm */ +// seee eeee eeee mmmm mmmm mmmm mmmm mmmm | mmmm mmmm mmmm mmmm mmmm mmmm mmmm +// mmmm -COMPILER_RT_ABI double -__floattidf(ti_int a) -{ - if (a == 0) - return 0.0; - const unsigned N = sizeof(ti_int) * CHAR_BIT; - const ti_int s = a >> (N-1); - a = (a ^ s) - s; - int sd = N - __clzti2(a); /* number of significant digits */ - int e = sd - 1; /* exponent */ - if (sd > DBL_MANT_DIG) - { - /* start: 0000000000000000000001xxxxxxxxxxxxxxxxxxxxxxPQxxxxxxxxxxxxxxxxxx - * finish: 000000000000000000000000000000000000001xxxxxxxxxxxxxxxxxxxxxxPQR - * 12345678901234567890123456 - * 1 = msb 1 bit - * P = bit DBL_MANT_DIG-1 bits to the right of 1 - * Q = bit DBL_MANT_DIG bits to the right of 1 - * R = "or" of all bits to the right of Q - */ - switch (sd) - { - case DBL_MANT_DIG + 1: - a <<= 1; - break; - case DBL_MANT_DIG + 2: - break; - default: - a = ((tu_int)a >> (sd - (DBL_MANT_DIG+2))) | - ((a & ((tu_int)(-1) >> ((N + DBL_MANT_DIG+2) - sd))) != 0); - }; - /* finish: */ - a |= (a & 4) != 0; /* Or P into R */ - ++a; /* round - this step may add a significant bit */ - a >>= 2; /* dump Q and R */ - /* a is now rounded to DBL_MANT_DIG or DBL_MANT_DIG+1 bits */ - if (a & ((tu_int)1 << DBL_MANT_DIG)) - { - a >>= 1; - ++e; - } - /* a is now rounded to DBL_MANT_DIG bits */ +COMPILER_RT_ABI double __floattidf(ti_int a) { + if (a == 0) + return 0.0; + const unsigned N = sizeof(ti_int) * CHAR_BIT; + const ti_int s = a >> (N - 1); + a = (a ^ s) - s; + int sd = N - __clzti2(a); // number of significant digits + int e = sd - 1; // exponent + if (sd > DBL_MANT_DIG) { + // start: 0000000000000000000001xxxxxxxxxxxxxxxxxxxxxxPQxxxxxxxxxxxxxxxxxx + // finish: 000000000000000000000000000000000000001xxxxxxxxxxxxxxxxxxxxxxPQR + // 12345678901234567890123456 + // 1 = msb 1 bit + // P = bit DBL_MANT_DIG-1 bits to the right of 1 + // Q = bit DBL_MANT_DIG bits to the right of 1 + // R = "or" of all bits to the right of Q + switch (sd) { + case DBL_MANT_DIG + 1: + a <<= 1; + break; + case DBL_MANT_DIG + 2: + break; + default: + a = ((tu_int)a >> (sd - (DBL_MANT_DIG + 2))) | + ((a & ((tu_int)(-1) >> ((N + DBL_MANT_DIG + 2) - sd))) != 0); + }; + // finish: + a |= (a & 4) != 0; // Or P into R + ++a; // round - this step may add a significant bit + a >>= 2; // dump Q and R + // a is now rounded to DBL_MANT_DIG or DBL_MANT_DIG+1 bits + if (a & ((tu_int)1 << DBL_MANT_DIG)) { + a >>= 1; + ++e; } - else - { - a <<= (DBL_MANT_DIG - sd); - /* a is now rounded to DBL_MANT_DIG bits */ - } - double_bits fb; - fb.u.s.high = ((su_int)s & 0x80000000) | /* sign */ - ((e + 1023) << 20) | /* exponent */ - ((su_int)(a >> 32) & 0x000FFFFF); /* mantissa-high */ - fb.u.s.low = (su_int)a; /* mantissa-low */ - return fb.f; + // a is now rounded to DBL_MANT_DIG bits + } else { + a <<= (DBL_MANT_DIG - sd); + // a is now rounded to DBL_MANT_DIG bits + } + double_bits fb; + fb.u.s.high = ((su_int)s & 0x80000000) | // sign + ((e + 1023) << 20) | // exponent + ((su_int)(a >> 32) & 0x000FFFFF); // mantissa-high + fb.u.s.low = (su_int)a; // mantissa-low + return fb.f; } -#endif /* CRT_HAS_128BIT */ +#endif // CRT_HAS_128BIT diff --git a/lib/builtins/floattisf.c b/lib/builtins/floattisf.c index f1b585f2c326..a8fcdbe14c07 100644 --- a/lib/builtins/floattisf.c +++ b/lib/builtins/floattisf.c @@ -1,82 +1,71 @@ -/* ===-- floattisf.c - Implement __floattisf -------------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - * - * This file implements __floattisf for the compiler_rt library. - * - * ===----------------------------------------------------------------------=== - */ +//===-- floattisf.c - Implement __floattisf -------------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements __floattisf for the compiler_rt library. +// +//===----------------------------------------------------------------------===// #include "int_lib.h" #ifdef CRT_HAS_128BIT -/* Returns: convert a to a float, rounding toward even. */ +// Returns: convert a to a float, rounding toward even. -/* Assumption: float is a IEEE 32 bit floating point type - * ti_int is a 128 bit integral type - */ +// Assumption: float is a IEEE 32 bit floating point type +// ti_int is a 128 bit integral type -/* seee eeee emmm mmmm mmmm mmmm mmmm mmmm */ +// seee eeee emmm mmmm mmmm mmmm mmmm mmmm -COMPILER_RT_ABI float -__floattisf(ti_int a) -{ - if (a == 0) - return 0.0F; - const unsigned N = sizeof(ti_int) * CHAR_BIT; - const ti_int s = a >> (N-1); - a = (a ^ s) - s; - int sd = N - __clzti2(a); /* number of significant digits */ - int e = sd - 1; /* exponent */ - if (sd > FLT_MANT_DIG) - { - /* start: 0000000000000000000001xxxxxxxxxxxxxxxxxxxxxxPQxxxxxxxxxxxxxxxxxx - * finish: 000000000000000000000000000000000000001xxxxxxxxxxxxxxxxxxxxxxPQR - * 12345678901234567890123456 - * 1 = msb 1 bit - * P = bit FLT_MANT_DIG-1 bits to the right of 1 - * Q = bit FLT_MANT_DIG bits to the right of 1 - * R = "or" of all bits to the right of Q - */ - switch (sd) - { - case FLT_MANT_DIG + 1: - a <<= 1; - break; - case FLT_MANT_DIG + 2: - break; - default: - a = ((tu_int)a >> (sd - (FLT_MANT_DIG+2))) | - ((a & ((tu_int)(-1) >> ((N + FLT_MANT_DIG+2) - sd))) != 0); - }; - /* finish: */ - a |= (a & 4) != 0; /* Or P into R */ - ++a; /* round - this step may add a significant bit */ - a >>= 2; /* dump Q and R */ - /* a is now rounded to FLT_MANT_DIG or FLT_MANT_DIG+1 bits */ - if (a & ((tu_int)1 << FLT_MANT_DIG)) - { - a >>= 1; - ++e; - } - /* a is now rounded to FLT_MANT_DIG bits */ +COMPILER_RT_ABI float __floattisf(ti_int a) { + if (a == 0) + return 0.0F; + const unsigned N = sizeof(ti_int) * CHAR_BIT; + const ti_int s = a >> (N - 1); + a = (a ^ s) - s; + int sd = N - __clzti2(a); // number of significant digits + int e = sd - 1; // exponent + if (sd > FLT_MANT_DIG) { + // start: 0000000000000000000001xxxxxxxxxxxxxxxxxxxxxxPQxxxxxxxxxxxxxxxxxx + // finish: 000000000000000000000000000000000000001xxxxxxxxxxxxxxxxxxxxxxPQR + // 12345678901234567890123456 + // 1 = msb 1 bit + // P = bit FLT_MANT_DIG-1 bits to the right of 1 + // Q = bit FLT_MANT_DIG bits to the right of 1 + // R = "or" of all bits to the right of Q + switch (sd) { + case FLT_MANT_DIG + 1: + a <<= 1; + break; + case FLT_MANT_DIG + 2: + break; + default: + a = ((tu_int)a >> (sd - (FLT_MANT_DIG + 2))) | + ((a & ((tu_int)(-1) >> ((N + FLT_MANT_DIG + 2) - sd))) != 0); + }; + // finish: + a |= (a & 4) != 0; // Or P into R + ++a; // round - this step may add a significant bit + a >>= 2; // dump Q and R + // a is now rounded to FLT_MANT_DIG or FLT_MANT_DIG+1 bits + if (a & ((tu_int)1 << FLT_MANT_DIG)) { + a >>= 1; + ++e; } - else - { - a <<= (FLT_MANT_DIG - sd); - /* a is now rounded to FLT_MANT_DIG bits */ - } - float_bits fb; - fb.u = ((su_int)s & 0x80000000) | /* sign */ - ((e + 127) << 23) | /* exponent */ - ((su_int)a & 0x007FFFFF); /* mantissa */ - return fb.f; + // a is now rounded to FLT_MANT_DIG bits + } else { + a <<= (FLT_MANT_DIG - sd); + // a is now rounded to FLT_MANT_DIG bits + } + float_bits fb; + fb.u = ((su_int)s & 0x80000000) | // sign + ((e + 127) << 23) | // exponent + ((su_int)a & 0x007FFFFF); // mantissa + return fb.f; } -#endif /* CRT_HAS_128BIT */ +#endif // CRT_HAS_128BIT diff --git a/lib/builtins/floattitf.c b/lib/builtins/floattitf.c index 994fded3947b..196cbdae14e0 100644 --- a/lib/builtins/floattitf.c +++ b/lib/builtins/floattitf.c @@ -1,9 +1,8 @@ //===-- lib/floattitf.c - int128 -> quad-precision conversion -----*- C -*-===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// // @@ -17,66 +16,63 @@ #include "fp_lib.h" #include "int_lib.h" -/* Returns: convert a ti_int to a fp_t, rounding toward even. */ +// Returns: convert a ti_int to a fp_t, rounding toward even. -/* Assumption: fp_t is a IEEE 128 bit floating point type - * ti_int is a 128 bit integral type - */ +// Assumption: fp_t is a IEEE 128 bit floating point type +// ti_int is a 128 bit integral type -/* seee eeee eeee eeee mmmm mmmm mmmm mmmm | mmmm mmmm mmmm mmmm mmmm mmmm mmmm mmmm | - * mmmm mmmm mmmm mmmm mmmm mmmm mmmm mmmm | mmmm mmmm mmmm mmmm mmmm mmmm mmmm mmmm - */ +// seee eeee eeee eeee mmmm mmmm mmmm mmmm | mmmm mmmm mmmm mmmm mmmm mmmm mmmm +// mmmm | mmmm mmmm mmmm mmmm mmmm mmmm mmmm mmmm | mmmm mmmm mmmm mmmm mmmm +// mmmm mmmm mmmm #if defined(CRT_HAS_128BIT) && defined(CRT_LDBL_128BIT) -COMPILER_RT_ABI fp_t -__floattitf(ti_int a) { - if (a == 0) - return 0.0; - const unsigned N = sizeof(ti_int) * CHAR_BIT; - const ti_int s = a >> (N-1); - a = (a ^ s) - s; - int sd = N - __clzti2(a); /* number of significant digits */ - int e = sd - 1; /* exponent */ - if (sd > LDBL_MANT_DIG) { - /* start: 0000000000000000000001xxxxxxxxxxxxxxxxxxxxxxPQxxxxxxxxxxxxxxxxxx - * finish: 000000000000000000000000000000000000001xxxxxxxxxxxxxxxxxxxxxxPQR - * 12345678901234567890123456 - * 1 = msb 1 bit - * P = bit LDBL_MANT_DIG-1 bits to the right of 1 - * Q = bit LDBL_MANT_DIG bits to the right of 1 - * R = "or" of all bits to the right of Q - */ - switch (sd) { - case LDBL_MANT_DIG + 1: - a <<= 1; - break; - case LDBL_MANT_DIG + 2: - break; - default: - a = ((tu_int)a >> (sd - (LDBL_MANT_DIG+2))) | - ((a & ((tu_int)(-1) >> ((N + LDBL_MANT_DIG+2) - sd))) != 0); - }; - /* finish: */ - a |= (a & 4) != 0; /* Or P into R */ - ++a; /* round - this step may add a significant bit */ - a >>= 2; /* dump Q and R */ - /* a is now rounded to LDBL_MANT_DIG or LDBL_MANT_DIG+1 bits */ - if (a & ((tu_int)1 << LDBL_MANT_DIG)) { - a >>= 1; - ++e; - } - /* a is now rounded to LDBL_MANT_DIG bits */ - } else { - a <<= (LDBL_MANT_DIG - sd); - /* a is now rounded to LDBL_MANT_DIG bits */ +COMPILER_RT_ABI fp_t __floattitf(ti_int a) { + if (a == 0) + return 0.0; + const unsigned N = sizeof(ti_int) * CHAR_BIT; + const ti_int s = a >> (N - 1); + a = (a ^ s) - s; + int sd = N - __clzti2(a); // number of significant digits + int e = sd - 1; // exponent + if (sd > LDBL_MANT_DIG) { + // start: 0000000000000000000001xxxxxxxxxxxxxxxxxxxxxxPQxxxxxxxxxxxxxxxxxx + // finish: 000000000000000000000000000000000000001xxxxxxxxxxxxxxxxxxxxxxPQR + // 12345678901234567890123456 + // 1 = msb 1 bit + // P = bit LDBL_MANT_DIG-1 bits to the right of 1 + // Q = bit LDBL_MANT_DIG bits to the right of 1 + // R = "or" of all bits to the right of Q + switch (sd) { + case LDBL_MANT_DIG + 1: + a <<= 1; + break; + case LDBL_MANT_DIG + 2: + break; + default: + a = ((tu_int)a >> (sd - (LDBL_MANT_DIG + 2))) | + ((a & ((tu_int)(-1) >> ((N + LDBL_MANT_DIG + 2) - sd))) != 0); + }; + // finish: + a |= (a & 4) != 0; // Or P into R + ++a; // round - this step may add a significant bit + a >>= 2; // dump Q and R + // a is now rounded to LDBL_MANT_DIG or LDBL_MANT_DIG+1 bits + if (a & ((tu_int)1 << LDBL_MANT_DIG)) { + a >>= 1; + ++e; } + // a is now rounded to LDBL_MANT_DIG bits + } else { + a <<= (LDBL_MANT_DIG - sd); + // a is now rounded to LDBL_MANT_DIG bits + } - long_double_bits fb; - fb.u.high.all = (s & 0x8000000000000000LL) /* sign */ - | (du_int)(e + 16383) << 48 /* exponent */ - | ((a >> 64) & 0x0000ffffffffffffLL); /* significand */ - fb.u.low.all = (du_int)(a); - return fb.f; + long_double_bits fb; + fb.u.high.all = (s & 0x8000000000000000LL) // sign + | (du_int)(e + 16383) << 48 // exponent + | ((a >> 64) & 0x0000ffffffffffffLL); // significand + fb.u.low.all = (du_int)(a); + return fb.f; } #endif diff --git a/lib/builtins/floattixf.c b/lib/builtins/floattixf.c index 1203b3a96e75..23796f1bb56f 100644 --- a/lib/builtins/floattixf.c +++ b/lib/builtins/floattixf.c @@ -1,84 +1,73 @@ -/* ===-- floattixf.c - Implement __floattixf -------------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - * - * This file implements __floattixf for the compiler_rt library. - * - * ===----------------------------------------------------------------------=== - */ +//===-- floattixf.c - Implement __floattixf -------------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements __floattixf for the compiler_rt library. +// +//===----------------------------------------------------------------------===// #include "int_lib.h" #ifdef CRT_HAS_128BIT -/* Returns: convert a to a long double, rounding toward even. */ +// Returns: convert a to a long double, rounding toward even. -/* Assumption: long double is a IEEE 80 bit floating point type padded to 128 bits - * ti_int is a 128 bit integral type - */ +// Assumption: long double is a IEEE 80 bit floating point type padded to 128 +// bits ti_int is a 128 bit integral type -/* gggg gggg gggg gggg gggg gggg gggg gggg | gggg gggg gggg gggg seee eeee eeee eeee | - * 1mmm mmmm mmmm mmmm mmmm mmmm mmmm mmmm | mmmm mmmm mmmm mmmm mmmm mmmm mmmm mmmm - */ +// gggg gggg gggg gggg gggg gggg gggg gggg | gggg gggg gggg gggg seee eeee eeee +// eeee | 1mmm mmmm mmmm mmmm mmmm mmmm mmmm mmmm | mmmm mmmm mmmm mmmm mmmm +// mmmm mmmm mmmm -COMPILER_RT_ABI long double -__floattixf(ti_int a) -{ - if (a == 0) - return 0.0; - const unsigned N = sizeof(ti_int) * CHAR_BIT; - const ti_int s = a >> (N-1); - a = (a ^ s) - s; - int sd = N - __clzti2(a); /* number of significant digits */ - int e = sd - 1; /* exponent */ - if (sd > LDBL_MANT_DIG) - { - /* start: 0000000000000000000001xxxxxxxxxxxxxxxxxxxxxxPQxxxxxxxxxxxxxxxxxx - * finish: 000000000000000000000000000000000000001xxxxxxxxxxxxxxxxxxxxxxPQR - * 12345678901234567890123456 - * 1 = msb 1 bit - * P = bit LDBL_MANT_DIG-1 bits to the right of 1 - * Q = bit LDBL_MANT_DIG bits to the right of 1 - * R = "or" of all bits to the right of Q - */ - switch (sd) - { - case LDBL_MANT_DIG + 1: - a <<= 1; - break; - case LDBL_MANT_DIG + 2: - break; - default: - a = ((tu_int)a >> (sd - (LDBL_MANT_DIG+2))) | - ((a & ((tu_int)(-1) >> ((N + LDBL_MANT_DIG+2) - sd))) != 0); - }; - /* finish: */ - a |= (a & 4) != 0; /* Or P into R */ - ++a; /* round - this step may add a significant bit */ - a >>= 2; /* dump Q and R */ - /* a is now rounded to LDBL_MANT_DIG or LDBL_MANT_DIG+1 bits */ - if (a & ((tu_int)1 << LDBL_MANT_DIG)) - { - a >>= 1; - ++e; - } - /* a is now rounded to LDBL_MANT_DIG bits */ +COMPILER_RT_ABI long double __floattixf(ti_int a) { + if (a == 0) + return 0.0; + const unsigned N = sizeof(ti_int) * CHAR_BIT; + const ti_int s = a >> (N - 1); + a = (a ^ s) - s; + int sd = N - __clzti2(a); // number of significant digits + int e = sd - 1; // exponent + if (sd > LDBL_MANT_DIG) { + // start: 0000000000000000000001xxxxxxxxxxxxxxxxxxxxxxPQxxxxxxxxxxxxxxxxxx + // finish: 000000000000000000000000000000000000001xxxxxxxxxxxxxxxxxxxxxxPQR + // 12345678901234567890123456 + // 1 = msb 1 bit + // P = bit LDBL_MANT_DIG-1 bits to the right of 1 + // Q = bit LDBL_MANT_DIG bits to the right of 1 + // R = "or" of all bits to the right of Q + switch (sd) { + case LDBL_MANT_DIG + 1: + a <<= 1; + break; + case LDBL_MANT_DIG + 2: + break; + default: + a = ((tu_int)a >> (sd - (LDBL_MANT_DIG + 2))) | + ((a & ((tu_int)(-1) >> ((N + LDBL_MANT_DIG + 2) - sd))) != 0); + }; + // finish: + a |= (a & 4) != 0; // Or P into R + ++a; // round - this step may add a significant bit + a >>= 2; // dump Q and R + // a is now rounded to LDBL_MANT_DIG or LDBL_MANT_DIG+1 bits + if (a & ((tu_int)1 << LDBL_MANT_DIG)) { + a >>= 1; + ++e; } - else - { - a <<= (LDBL_MANT_DIG - sd); - /* a is now rounded to LDBL_MANT_DIG bits */ - } - long_double_bits fb; - fb.u.high.s.low = ((su_int)s & 0x8000) | /* sign */ - (e + 16383); /* exponent */ - fb.u.low.all = (du_int)a; /* mantissa */ - return fb.f; + // a is now rounded to LDBL_MANT_DIG bits + } else { + a <<= (LDBL_MANT_DIG - sd); + // a is now rounded to LDBL_MANT_DIG bits + } + long_double_bits fb; + fb.u.high.s.low = ((su_int)s & 0x8000) | // sign + (e + 16383); // exponent + fb.u.low.all = (du_int)a; // mantissa + return fb.f; } -#endif /* CRT_HAS_128BIT */ +#endif // CRT_HAS_128BIT diff --git a/lib/builtins/floatundidf.c b/lib/builtins/floatundidf.c index 8bc2a096324f..e7c6aae5ce38 100644 --- a/lib/builtins/floatundidf.c +++ b/lib/builtins/floatundidf.c @@ -1,114 +1,106 @@ -/* ===-- floatundidf.c - Implement __floatundidf ---------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - * - * This file implements __floatundidf for the compiler_rt library. - * - * ===----------------------------------------------------------------------=== - */ +//===-- floatundidf.c - Implement __floatundidf ---------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements __floatundidf for the compiler_rt library. +// +//===----------------------------------------------------------------------===// -/* Returns: convert a to a double, rounding toward even. */ +// Returns: convert a to a double, rounding toward even. -/* Assumption: double is a IEEE 64 bit floating point type - * du_int is a 64 bit integral type - */ +// Assumption: double is a IEEE 64 bit floating point type +// du_int is a 64 bit integral type -/* seee eeee eeee mmmm mmmm mmmm mmmm mmmm | mmmm mmmm mmmm mmmm mmmm mmmm mmmm mmmm */ +// seee eeee eeee mmmm mmmm mmmm mmmm mmmm | mmmm mmmm mmmm mmmm mmmm mmmm mmmm +// mmmm #include "int_lib.h" #ifndef __SOFT_FP__ -/* Support for systems that have hardware floating-point; we'll set the inexact flag - * as a side-effect of this computation. - */ +// Support for systems that have hardware floating-point; we'll set the inexact +// flag as a side-effect of this computation. -COMPILER_RT_ABI double -__floatundidf(du_int a) -{ - static const double twop52 = 4503599627370496.0; // 0x1.0p52 - static const double twop84 = 19342813113834066795298816.0; // 0x1.0p84 - static const double twop84_plus_twop52 = 19342813118337666422669312.0; // 0x1.00000001p84 +COMPILER_RT_ABI double __floatundidf(du_int a) { + static const double twop52 = 4503599627370496.0; // 0x1.0p52 + static const double twop84 = 19342813113834066795298816.0; // 0x1.0p84 + static const double twop84_plus_twop52 = + 19342813118337666422669312.0; // 0x1.00000001p84 - union { uint64_t x; double d; } high = { .d = twop84 }; - union { uint64_t x; double d; } low = { .d = twop52 }; + union { + uint64_t x; + double d; + } high = {.d = twop84}; + union { + uint64_t x; + double d; + } low = {.d = twop52}; - high.x |= a >> 32; - low.x |= a & UINT64_C(0x00000000ffffffff); + high.x |= a >> 32; + low.x |= a & UINT64_C(0x00000000ffffffff); - const double result = (high.d - twop84_plus_twop52) + low.d; - return result; + const double result = (high.d - twop84_plus_twop52) + low.d; + return result; } #else -/* Support for systems that don't have hardware floating-point; there are no flags to - * set, and we don't want to code-gen to an unknown soft-float implementation. - */ +// Support for systems that don't have hardware floating-point; there are no +// flags to set, and we don't want to code-gen to an unknown soft-float +// implementation. -COMPILER_RT_ABI double -__floatundidf(du_int a) -{ - if (a == 0) - return 0.0; - const unsigned N = sizeof(du_int) * CHAR_BIT; - int sd = N - __builtin_clzll(a); /* number of significant digits */ - int e = sd - 1; /* exponent */ - if (sd > DBL_MANT_DIG) - { - /* start: 0000000000000000000001xxxxxxxxxxxxxxxxxxxxxxPQxxxxxxxxxxxxxxxxxx - * finish: 000000000000000000000000000000000000001xxxxxxxxxxxxxxxxxxxxxxPQR - * 12345678901234567890123456 - * 1 = msb 1 bit - * P = bit DBL_MANT_DIG-1 bits to the right of 1 - * Q = bit DBL_MANT_DIG bits to the right of 1 - * R = "or" of all bits to the right of Q - */ - switch (sd) - { - case DBL_MANT_DIG + 1: - a <<= 1; - break; - case DBL_MANT_DIG + 2: - break; - default: - a = (a >> (sd - (DBL_MANT_DIG+2))) | - ((a & ((du_int)(-1) >> ((N + DBL_MANT_DIG+2) - sd))) != 0); - }; - /* finish: */ - a |= (a & 4) != 0; /* Or P into R */ - ++a; /* round - this step may add a significant bit */ - a >>= 2; /* dump Q and R */ - /* a is now rounded to DBL_MANT_DIG or DBL_MANT_DIG+1 bits */ - if (a & ((du_int)1 << DBL_MANT_DIG)) - { - a >>= 1; - ++e; - } - /* a is now rounded to DBL_MANT_DIG bits */ +COMPILER_RT_ABI double __floatundidf(du_int a) { + if (a == 0) + return 0.0; + const unsigned N = sizeof(du_int) * CHAR_BIT; + int sd = N - __builtin_clzll(a); // number of significant digits + int e = sd - 1; // exponent + if (sd > DBL_MANT_DIG) { + // start: 0000000000000000000001xxxxxxxxxxxxxxxxxxxxxxPQxxxxxxxxxxxxxxxxxx + // finish: 000000000000000000000000000000000000001xxxxxxxxxxxxxxxxxxxxxxPQR + // 12345678901234567890123456 + // 1 = msb 1 bit + // P = bit DBL_MANT_DIG-1 bits to the right of 1 + // Q = bit DBL_MANT_DIG bits to the right of 1 + // R = "or" of all bits to the right of Q + switch (sd) { + case DBL_MANT_DIG + 1: + a <<= 1; + break; + case DBL_MANT_DIG + 2: + break; + default: + a = (a >> (sd - (DBL_MANT_DIG + 2))) | + ((a & ((du_int)(-1) >> ((N + DBL_MANT_DIG + 2) - sd))) != 0); + }; + // finish: + a |= (a & 4) != 0; // Or P into R + ++a; // round - this step may add a significant bit + a >>= 2; // dump Q and R + // a is now rounded to DBL_MANT_DIG or DBL_MANT_DIG+1 bits + if (a & ((du_int)1 << DBL_MANT_DIG)) { + a >>= 1; + ++e; } - else - { - a <<= (DBL_MANT_DIG - sd); - /* a is now rounded to DBL_MANT_DIG bits */ - } - double_bits fb; - fb.u.s.high = ((e + 1023) << 20) | /* exponent */ - ((su_int)(a >> 32) & 0x000FFFFF); /* mantissa-high */ - fb.u.s.low = (su_int)a; /* mantissa-low */ - return fb.f; + // a is now rounded to DBL_MANT_DIG bits + } else { + a <<= (DBL_MANT_DIG - sd); + // a is now rounded to DBL_MANT_DIG bits + } + double_bits fb; + fb.u.s.high = ((e + 1023) << 20) | // exponent + ((su_int)(a >> 32) & 0x000FFFFF); // mantissa-high + fb.u.s.low = (su_int)a; // mantissa-low + return fb.f; } #endif #if defined(__ARM_EABI__) #if defined(COMPILER_RT_ARMHF_TARGET) -AEABI_RTABI double __aeabi_ul2d(du_int a) { - return __floatundidf(a); -} +AEABI_RTABI double __aeabi_ul2d(du_int a) { return __floatundidf(a); } #else -AEABI_RTABI double __aeabi_ul2d(du_int a) COMPILER_RT_ALIAS(__floatundidf); +COMPILER_RT_ALIAS(__floatundidf, __aeabi_ul2d) #endif #endif diff --git a/lib/builtins/floatundisf.c b/lib/builtins/floatundisf.c index 844786ea7777..87841b761ded 100644 --- a/lib/builtins/floatundisf.c +++ b/lib/builtins/floatundisf.c @@ -1,85 +1,72 @@ -/*===-- floatundisf.c - Implement __floatundisf ---------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - * - * This file implements __floatundisf for the compiler_rt library. - * - *===----------------------------------------------------------------------=== - */ +//===-- floatundisf.c - Implement __floatundisf ---------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements __floatundisf for the compiler_rt library. +// +//===----------------------------------------------------------------------===// -/* Returns: convert a to a float, rounding toward even. */ +// Returns: convert a to a float, rounding toward even. -/* Assumption: float is a IEEE 32 bit floating point type - * du_int is a 64 bit integral type - */ +// Assumption: float is a IEEE 32 bit floating point type +// du_int is a 64 bit integral type -/* seee eeee emmm mmmm mmmm mmmm mmmm mmmm */ +// seee eeee emmm mmmm mmmm mmmm mmmm mmmm #include "int_lib.h" -COMPILER_RT_ABI float -__floatundisf(du_int a) -{ - if (a == 0) - return 0.0F; - const unsigned N = sizeof(du_int) * CHAR_BIT; - int sd = N - __builtin_clzll(a); /* number of significant digits */ - int e = sd - 1; /* 8 exponent */ - if (sd > FLT_MANT_DIG) - { - /* start: 0000000000000000000001xxxxxxxxxxxxxxxxxxxxxxPQxxxxxxxxxxxxxxxxxx - * finish: 000000000000000000000000000000000000001xxxxxxxxxxxxxxxxxxxxxxPQR - * 12345678901234567890123456 - * 1 = msb 1 bit - * P = bit FLT_MANT_DIG-1 bits to the right of 1 - * Q = bit FLT_MANT_DIG bits to the right of 1 - * R = "or" of all bits to the right of Q - */ - switch (sd) - { - case FLT_MANT_DIG + 1: - a <<= 1; - break; - case FLT_MANT_DIG + 2: - break; - default: - a = (a >> (sd - (FLT_MANT_DIG+2))) | - ((a & ((du_int)(-1) >> ((N + FLT_MANT_DIG+2) - sd))) != 0); - }; - /* finish: */ - a |= (a & 4) != 0; /* Or P into R */ - ++a; /* round - this step may add a significant bit */ - a >>= 2; /* dump Q and R */ - /* a is now rounded to FLT_MANT_DIG or FLT_MANT_DIG+1 bits */ - if (a & ((du_int)1 << FLT_MANT_DIG)) - { - a >>= 1; - ++e; - } - /* a is now rounded to FLT_MANT_DIG bits */ +COMPILER_RT_ABI float __floatundisf(du_int a) { + if (a == 0) + return 0.0F; + const unsigned N = sizeof(du_int) * CHAR_BIT; + int sd = N - __builtin_clzll(a); // number of significant digits + int e = sd - 1; // 8 exponent + if (sd > FLT_MANT_DIG) { + // start: 0000000000000000000001xxxxxxxxxxxxxxxxxxxxxxPQxxxxxxxxxxxxxxxxxx + // finish: 000000000000000000000000000000000000001xxxxxxxxxxxxxxxxxxxxxxPQR + // 12345678901234567890123456 + // 1 = msb 1 bit + // P = bit FLT_MANT_DIG-1 bits to the right of 1 + // Q = bit FLT_MANT_DIG bits to the right of 1 + // R = "or" of all bits to the right of Q + switch (sd) { + case FLT_MANT_DIG + 1: + a <<= 1; + break; + case FLT_MANT_DIG + 2: + break; + default: + a = (a >> (sd - (FLT_MANT_DIG + 2))) | + ((a & ((du_int)(-1) >> ((N + FLT_MANT_DIG + 2) - sd))) != 0); + }; + // finish: + a |= (a & 4) != 0; // Or P into R + ++a; // round - this step may add a significant bit + a >>= 2; // dump Q and R + // a is now rounded to FLT_MANT_DIG or FLT_MANT_DIG+1 bits + if (a & ((du_int)1 << FLT_MANT_DIG)) { + a >>= 1; + ++e; } - else - { - a <<= (FLT_MANT_DIG - sd); - /* a is now rounded to FLT_MANT_DIG bits */ - } - float_bits fb; - fb.u = ((e + 127) << 23) | /* exponent */ - ((su_int)a & 0x007FFFFF); /* mantissa */ - return fb.f; + // a is now rounded to FLT_MANT_DIG bits + } else { + a <<= (FLT_MANT_DIG - sd); + // a is now rounded to FLT_MANT_DIG bits + } + float_bits fb; + fb.u = ((e + 127) << 23) | // exponent + ((su_int)a & 0x007FFFFF); // mantissa + return fb.f; } #if defined(__ARM_EABI__) #if defined(COMPILER_RT_ARMHF_TARGET) -AEABI_RTABI float __aeabi_ul2f(du_int a) { - return __floatundisf(a); -} +AEABI_RTABI float __aeabi_ul2f(du_int a) { return __floatundisf(a); } #else -AEABI_RTABI float __aeabi_ul2f(du_int a) COMPILER_RT_ALIAS(__floatundisf); +COMPILER_RT_ALIAS(__floatundisf, __aeabi_ul2f) #endif #endif diff --git a/lib/builtins/floatunditf.c b/lib/builtins/floatunditf.c index 8098e95e82bc..8d310851e179 100644 --- a/lib/builtins/floatunditf.c +++ b/lib/builtins/floatunditf.c @@ -1,9 +1,8 @@ //===-- lib/floatunditf.c - uint -> quad-precision conversion -----*- C -*-===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// // @@ -19,22 +18,23 @@ #if defined(CRT_HAS_128BIT) && defined(CRT_LDBL_128BIT) COMPILER_RT_ABI fp_t __floatunditf(du_int a) { - const int aWidth = sizeof a * CHAR_BIT; + const int aWidth = sizeof a * CHAR_BIT; - // Handle zero as a special case to protect clz - if (a == 0) return fromRep(0); + // Handle zero as a special case to protect clz + if (a == 0) + return fromRep(0); - // Exponent of (fp_t)a is the width of abs(a). - const int exponent = (aWidth - 1) - __builtin_clzll(a); - rep_t result; + // Exponent of (fp_t)a is the width of abs(a). + const int exponent = (aWidth - 1) - __builtin_clzll(a); + rep_t result; - // Shift a into the significand field and clear the implicit bit. - const int shift = significandBits - exponent; - result = (rep_t)a << shift ^ implicitBit; + // Shift a into the significand field and clear the implicit bit. + const int shift = significandBits - exponent; + result = (rep_t)a << shift ^ implicitBit; - // Insert the exponent - result += (rep_t)(exponent + exponentBias) << significandBits; - return fromRep(result); + // Insert the exponent + result += (rep_t)(exponent + exponentBias) << significandBits; + return fromRep(result); } #endif diff --git a/lib/builtins/floatundixf.c b/lib/builtins/floatundixf.c index ca5e06d64dcd..85264adac197 100644 --- a/lib/builtins/floatundixf.c +++ b/lib/builtins/floatundixf.c @@ -1,42 +1,37 @@ -/* ===-- floatundixf.c - Implement __floatundixf ---------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - * - * This file implements __floatundixf for the compiler_rt library. - * - * ===----------------------------------------------------------------------=== - */ +//===-- floatundixf.c - Implement __floatundixf ---------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements __floatundixf for the compiler_rt library. +// +//===----------------------------------------------------------------------===// #if !_ARCH_PPC #include "int_lib.h" -/* Returns: convert a to a long double, rounding toward even. */ +// Returns: convert a to a long double, rounding toward even. -/* Assumption: long double is a IEEE 80 bit floating point type padded to 128 bits - * du_int is a 64 bit integral type - */ +// Assumption: long double is a IEEE 80 bit floating point type padded to 128 +// bits du_int is a 64 bit integral type -/* gggg gggg gggg gggg gggg gggg gggg gggg | gggg gggg gggg gggg seee eeee eeee eeee | - * 1mmm mmmm mmmm mmmm mmmm mmmm mmmm mmmm | mmmm mmmm mmmm mmmm mmmm mmmm mmmm mmmm - */ -COMPILER_RT_ABI long double -__floatundixf(du_int a) -{ - if (a == 0) - return 0.0; - const unsigned N = sizeof(du_int) * CHAR_BIT; - int clz = __builtin_clzll(a); - int e = (N - 1) - clz ; /* exponent */ - long_double_bits fb; - fb.u.high.s.low = (e + 16383); /* exponent */ - fb.u.low.all = a << clz; /* mantissa */ - return fb.f; +// gggg gggg gggg gggg gggg gggg gggg gggg | gggg gggg gggg gggg seee eeee eeee +// eeee | 1mmm mmmm mmmm mmmm mmmm mmmm mmmm mmmm | mmmm mmmm mmmm mmmm mmmm +// mmmm mmmm mmmm +COMPILER_RT_ABI long double __floatundixf(du_int a) { + if (a == 0) + return 0.0; + const unsigned N = sizeof(du_int) * CHAR_BIT; + int clz = __builtin_clzll(a); + int e = (N - 1) - clz; // exponent + long_double_bits fb; + fb.u.high.s.low = (e + 16383); // exponent + fb.u.low.all = a << clz; // mantissa + return fb.f; } -#endif /* _ARCH_PPC */ +#endif // _ARCH_PPC diff --git a/lib/builtins/floatunsidf.c b/lib/builtins/floatunsidf.c index 75cf6b9177df..2c01c3041434 100644 --- a/lib/builtins/floatunsidf.c +++ b/lib/builtins/floatunsidf.c @@ -1,9 +1,8 @@ //===-- lib/floatunsidf.c - uint -> double-precision conversion ---*- C -*-===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// // @@ -18,33 +17,31 @@ #include "int_lib.h" -COMPILER_RT_ABI fp_t -__floatunsidf(unsigned int a) { - - const int aWidth = sizeof a * CHAR_BIT; - - // Handle zero as a special case to protect clz - if (a == 0) return fromRep(0); - - // Exponent of (fp_t)a is the width of abs(a). - const int exponent = (aWidth - 1) - __builtin_clz(a); - rep_t result; - - // Shift a into the significand field and clear the implicit bit. - const int shift = significandBits - exponent; - result = (rep_t)a << shift ^ implicitBit; - - // Insert the exponent - result += (rep_t)(exponent + exponentBias) << significandBits; - return fromRep(result); +COMPILER_RT_ABI fp_t __floatunsidf(unsigned int a) { + + const int aWidth = sizeof a * CHAR_BIT; + + // Handle zero as a special case to protect clz + if (a == 0) + return fromRep(0); + + // Exponent of (fp_t)a is the width of abs(a). + const int exponent = (aWidth - 1) - __builtin_clz(a); + rep_t result; + + // Shift a into the significand field and clear the implicit bit. + const int shift = significandBits - exponent; + result = (rep_t)a << shift ^ implicitBit; + + // Insert the exponent + result += (rep_t)(exponent + exponentBias) << significandBits; + return fromRep(result); } #if defined(__ARM_EABI__) #if defined(COMPILER_RT_ARMHF_TARGET) -AEABI_RTABI fp_t __aeabi_ui2d(unsigned int a) { - return __floatunsidf(a); -} +AEABI_RTABI fp_t __aeabi_ui2d(unsigned int a) { return __floatunsidf(a); } #else -AEABI_RTABI fp_t __aeabi_ui2d(unsigned int a) COMPILER_RT_ALIAS(__floatunsidf); +COMPILER_RT_ALIAS(__floatunsidf, __aeabi_ui2d) #endif #endif diff --git a/lib/builtins/floatunsisf.c b/lib/builtins/floatunsisf.c index 29525ccedbbe..33a1b5ae2a63 100644 --- a/lib/builtins/floatunsisf.c +++ b/lib/builtins/floatunsisf.c @@ -1,9 +1,8 @@ //===-- lib/floatunsisf.c - uint -> single-precision conversion ---*- C -*-===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// // @@ -18,41 +17,41 @@ #include "int_lib.h" -COMPILER_RT_ABI fp_t -__floatunsisf(unsigned int a) { - - const int aWidth = sizeof a * CHAR_BIT; - - // Handle zero as a special case to protect clz - if (a == 0) return fromRep(0); - - // Exponent of (fp_t)a is the width of abs(a). - const int exponent = (aWidth - 1) - __builtin_clz(a); - rep_t result; - - // Shift a into the significand field, rounding if it is a right-shift - if (exponent <= significandBits) { - const int shift = significandBits - exponent; - result = (rep_t)a << shift ^ implicitBit; - } else { - const int shift = exponent - significandBits; - result = (rep_t)a >> shift ^ implicitBit; - rep_t round = (rep_t)a << (typeWidth - shift); - if (round > signBit) result++; - if (round == signBit) result += result & 1; - } - - // Insert the exponent - result += (rep_t)(exponent + exponentBias) << significandBits; - return fromRep(result); +COMPILER_RT_ABI fp_t __floatunsisf(unsigned int a) { + + const int aWidth = sizeof a * CHAR_BIT; + + // Handle zero as a special case to protect clz + if (a == 0) + return fromRep(0); + + // Exponent of (fp_t)a is the width of abs(a). + const int exponent = (aWidth - 1) - __builtin_clz(a); + rep_t result; + + // Shift a into the significand field, rounding if it is a right-shift + if (exponent <= significandBits) { + const int shift = significandBits - exponent; + result = (rep_t)a << shift ^ implicitBit; + } else { + const int shift = exponent - significandBits; + result = (rep_t)a >> shift ^ implicitBit; + rep_t round = (rep_t)a << (typeWidth - shift); + if (round > signBit) + result++; + if (round == signBit) + result += result & 1; + } + + // Insert the exponent + result += (rep_t)(exponent + exponentBias) << significandBits; + return fromRep(result); } #if defined(__ARM_EABI__) #if defined(COMPILER_RT_ARMHF_TARGET) -AEABI_RTABI fp_t __aeabi_ui2f(unsigned int a) { - return __floatunsisf(a); -} +AEABI_RTABI fp_t __aeabi_ui2f(unsigned int a) { return __floatunsisf(a); } #else -AEABI_RTABI fp_t __aeabi_ui2f(unsigned int a) COMPILER_RT_ALIAS(__floatunsisf); +COMPILER_RT_ALIAS(__floatunsisf, __aeabi_ui2f) #endif #endif diff --git a/lib/builtins/floatunsitf.c b/lib/builtins/floatunsitf.c index 1cd1842e709a..a4bf0f65fe1c 100644 --- a/lib/builtins/floatunsitf.c +++ b/lib/builtins/floatunsitf.c @@ -1,9 +1,8 @@ //===-- lib/floatunsitf.c - uint -> quad-precision conversion -----*- C -*-===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// // @@ -19,22 +18,23 @@ #if defined(CRT_HAS_128BIT) && defined(CRT_LDBL_128BIT) COMPILER_RT_ABI fp_t __floatunsitf(unsigned int a) { - const int aWidth = sizeof a * CHAR_BIT; + const int aWidth = sizeof a * CHAR_BIT; - // Handle zero as a special case to protect clz - if (a == 0) return fromRep(0); + // Handle zero as a special case to protect clz + if (a == 0) + return fromRep(0); - // Exponent of (fp_t)a is the width of abs(a). - const int exponent = (aWidth - 1) - __builtin_clz(a); - rep_t result; + // Exponent of (fp_t)a is the width of abs(a). + const int exponent = (aWidth - 1) - __builtin_clz(a); + rep_t result; - // Shift a into the significand field and clear the implicit bit. - const int shift = significandBits - exponent; - result = (rep_t)a << shift ^ implicitBit; + // Shift a into the significand field and clear the implicit bit. + const int shift = significandBits - exponent; + result = (rep_t)a << shift ^ implicitBit; - // Insert the exponent - result += (rep_t)(exponent + exponentBias) << significandBits; - return fromRep(result); + // Insert the exponent + result += (rep_t)(exponent + exponentBias) << significandBits; + return fromRep(result); } #endif diff --git a/lib/builtins/floatuntidf.c b/lib/builtins/floatuntidf.c index 960265d80772..e69e65c1ace4 100644 --- a/lib/builtins/floatuntidf.c +++ b/lib/builtins/floatuntidf.c @@ -1,80 +1,70 @@ -/* ===-- floatuntidf.c - Implement __floatuntidf ---------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - * - * This file implements __floatuntidf for the compiler_rt library. - * - * ===----------------------------------------------------------------------=== - */ +//===-- floatuntidf.c - Implement __floatuntidf ---------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements __floatuntidf for the compiler_rt library. +// +//===----------------------------------------------------------------------===// #include "int_lib.h" #ifdef CRT_HAS_128BIT -/* Returns: convert a to a double, rounding toward even. */ +// Returns: convert a to a double, rounding toward even. -/* Assumption: double is a IEEE 64 bit floating point type - * tu_int is a 128 bit integral type - */ +// Assumption: double is a IEEE 64 bit floating point type +// tu_int is a 128 bit integral type -/* seee eeee eeee mmmm mmmm mmmm mmmm mmmm | mmmm mmmm mmmm mmmm mmmm mmmm mmmm mmmm */ +// seee eeee eeee mmmm mmmm mmmm mmmm mmmm | mmmm mmmm mmmm mmmm mmmm mmmm mmmm +// mmmm -COMPILER_RT_ABI double -__floatuntidf(tu_int a) -{ - if (a == 0) - return 0.0; - const unsigned N = sizeof(tu_int) * CHAR_BIT; - int sd = N - __clzti2(a); /* number of significant digits */ - int e = sd - 1; /* exponent */ - if (sd > DBL_MANT_DIG) - { - /* start: 0000000000000000000001xxxxxxxxxxxxxxxxxxxxxxPQxxxxxxxxxxxxxxxxxx - * finish: 000000000000000000000000000000000000001xxxxxxxxxxxxxxxxxxxxxxPQR - * 12345678901234567890123456 - * 1 = msb 1 bit - * P = bit DBL_MANT_DIG-1 bits to the right of 1 - * Q = bit DBL_MANT_DIG bits to the right of 1 - * R = "or" of all bits to the right of Q - */ - switch (sd) - { - case DBL_MANT_DIG + 1: - a <<= 1; - break; - case DBL_MANT_DIG + 2: - break; - default: - a = (a >> (sd - (DBL_MANT_DIG+2))) | - ((a & ((tu_int)(-1) >> ((N + DBL_MANT_DIG+2) - sd))) != 0); - }; - /* finish: */ - a |= (a & 4) != 0; /* Or P into R */ - ++a; /* round - this step may add a significant bit */ - a >>= 2; /* dump Q and R */ - /* a is now rounded to DBL_MANT_DIG or DBL_MANT_DIG+1 bits */ - if (a & ((tu_int)1 << DBL_MANT_DIG)) - { - a >>= 1; - ++e; - } - /* a is now rounded to DBL_MANT_DIG bits */ +COMPILER_RT_ABI double __floatuntidf(tu_int a) { + if (a == 0) + return 0.0; + const unsigned N = sizeof(tu_int) * CHAR_BIT; + int sd = N - __clzti2(a); // number of significant digits + int e = sd - 1; // exponent + if (sd > DBL_MANT_DIG) { + // start: 0000000000000000000001xxxxxxxxxxxxxxxxxxxxxxPQxxxxxxxxxxxxxxxxxx + // finish: 000000000000000000000000000000000000001xxxxxxxxxxxxxxxxxxxxxxPQR + // 12345678901234567890123456 + // 1 = msb 1 bit + // P = bit DBL_MANT_DIG-1 bits to the right of 1 + // Q = bit DBL_MANT_DIG bits to the right of 1 + // R = "or" of all bits to the right of Q + switch (sd) { + case DBL_MANT_DIG + 1: + a <<= 1; + break; + case DBL_MANT_DIG + 2: + break; + default: + a = (a >> (sd - (DBL_MANT_DIG + 2))) | + ((a & ((tu_int)(-1) >> ((N + DBL_MANT_DIG + 2) - sd))) != 0); + }; + // finish: + a |= (a & 4) != 0; // Or P into R + ++a; // round - this step may add a significant bit + a >>= 2; // dump Q and R + // a is now rounded to DBL_MANT_DIG or DBL_MANT_DIG+1 bits + if (a & ((tu_int)1 << DBL_MANT_DIG)) { + a >>= 1; + ++e; } - else - { - a <<= (DBL_MANT_DIG - sd); - /* a is now rounded to DBL_MANT_DIG bits */ - } - double_bits fb; - fb.u.s.high = ((e + 1023) << 20) | /* exponent */ - ((su_int)(a >> 32) & 0x000FFFFF); /* mantissa-high */ - fb.u.s.low = (su_int)a; /* mantissa-low */ - return fb.f; + // a is now rounded to DBL_MANT_DIG bits + } else { + a <<= (DBL_MANT_DIG - sd); + // a is now rounded to DBL_MANT_DIG bits + } + double_bits fb; + fb.u.s.high = ((e + 1023) << 20) | // exponent + ((su_int)(a >> 32) & 0x000FFFFF); // mantissa-high + fb.u.s.low = (su_int)a; // mantissa-low + return fb.f; } -#endif /* CRT_HAS_128BIT */ +#endif // CRT_HAS_128BIT diff --git a/lib/builtins/floatuntisf.c b/lib/builtins/floatuntisf.c index c0dd0275ddba..9dec0ab5c58f 100644 --- a/lib/builtins/floatuntisf.c +++ b/lib/builtins/floatuntisf.c @@ -1,79 +1,68 @@ -/* ===-- floatuntisf.c - Implement __floatuntisf ---------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - * - * This file implements __floatuntisf for the compiler_rt library. - * - * ===----------------------------------------------------------------------=== - */ +//===-- floatuntisf.c - Implement __floatuntisf ---------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements __floatuntisf for the compiler_rt library. +// +//===----------------------------------------------------------------------===// #include "int_lib.h" #ifdef CRT_HAS_128BIT -/* Returns: convert a to a float, rounding toward even. */ +// Returns: convert a to a float, rounding toward even. -/* Assumption: float is a IEEE 32 bit floating point type - * tu_int is a 128 bit integral type - */ +// Assumption: float is a IEEE 32 bit floating point type +// tu_int is a 128 bit integral type -/* seee eeee emmm mmmm mmmm mmmm mmmm mmmm */ +// seee eeee emmm mmmm mmmm mmmm mmmm mmmm -COMPILER_RT_ABI float -__floatuntisf(tu_int a) -{ - if (a == 0) - return 0.0F; - const unsigned N = sizeof(tu_int) * CHAR_BIT; - int sd = N - __clzti2(a); /* number of significant digits */ - int e = sd - 1; /* exponent */ - if (sd > FLT_MANT_DIG) - { - /* start: 0000000000000000000001xxxxxxxxxxxxxxxxxxxxxxPQxxxxxxxxxxxxxxxxxx - * finish: 000000000000000000000000000000000000001xxxxxxxxxxxxxxxxxxxxxxPQR - * 12345678901234567890123456 - * 1 = msb 1 bit - * P = bit FLT_MANT_DIG-1 bits to the right of 1 - * Q = bit FLT_MANT_DIG bits to the right of 1 - * R = "or" of all bits to the right of Q - */ - switch (sd) - { - case FLT_MANT_DIG + 1: - a <<= 1; - break; - case FLT_MANT_DIG + 2: - break; - default: - a = (a >> (sd - (FLT_MANT_DIG+2))) | - ((a & ((tu_int)(-1) >> ((N + FLT_MANT_DIG+2) - sd))) != 0); - }; - /* finish: */ - a |= (a & 4) != 0; /* Or P into R */ - ++a; /* round - this step may add a significant bit */ - a >>= 2; /* dump Q and R */ - /* a is now rounded to FLT_MANT_DIG or FLT_MANT_DIG+1 bits */ - if (a & ((tu_int)1 << FLT_MANT_DIG)) - { - a >>= 1; - ++e; - } - /* a is now rounded to FLT_MANT_DIG bits */ +COMPILER_RT_ABI float __floatuntisf(tu_int a) { + if (a == 0) + return 0.0F; + const unsigned N = sizeof(tu_int) * CHAR_BIT; + int sd = N - __clzti2(a); // number of significant digits + int e = sd - 1; // exponent + if (sd > FLT_MANT_DIG) { + // start: 0000000000000000000001xxxxxxxxxxxxxxxxxxxxxxPQxxxxxxxxxxxxxxxxxx + // finish: 000000000000000000000000000000000000001xxxxxxxxxxxxxxxxxxxxxxPQR + // 12345678901234567890123456 + // 1 = msb 1 bit + // P = bit FLT_MANT_DIG-1 bits to the right of 1 + // Q = bit FLT_MANT_DIG bits to the right of 1 + // R = "or" of all bits to the right of Q + switch (sd) { + case FLT_MANT_DIG + 1: + a <<= 1; + break; + case FLT_MANT_DIG + 2: + break; + default: + a = (a >> (sd - (FLT_MANT_DIG + 2))) | + ((a & ((tu_int)(-1) >> ((N + FLT_MANT_DIG + 2) - sd))) != 0); + }; + // finish: + a |= (a & 4) != 0; // Or P into R + ++a; // round - this step may add a significant bit + a >>= 2; // dump Q and R + // a is now rounded to FLT_MANT_DIG or FLT_MANT_DIG+1 bits + if (a & ((tu_int)1 << FLT_MANT_DIG)) { + a >>= 1; + ++e; } - else - { - a <<= (FLT_MANT_DIG - sd); - /* a is now rounded to FLT_MANT_DIG bits */ - } - float_bits fb; - fb.u = ((e + 127) << 23) | /* exponent */ - ((su_int)a & 0x007FFFFF); /* mantissa */ - return fb.f; + // a is now rounded to FLT_MANT_DIG bits + } else { + a <<= (FLT_MANT_DIG - sd); + // a is now rounded to FLT_MANT_DIG bits + } + float_bits fb; + fb.u = ((e + 127) << 23) | // exponent + ((su_int)a & 0x007FFFFF); // mantissa + return fb.f; } -#endif /* CRT_HAS_128BIT */ +#endif // CRT_HAS_128BIT diff --git a/lib/builtins/floatuntitf.c b/lib/builtins/floatuntitf.c index e2518c93f234..d308d3118d03 100644 --- a/lib/builtins/floatuntitf.c +++ b/lib/builtins/floatuntitf.c @@ -1,9 +1,8 @@ //===-- lib/floatuntitf.c - uint128 -> quad-precision conversion --*- C -*-===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// // @@ -17,63 +16,60 @@ #include "fp_lib.h" #include "int_lib.h" -/* Returns: convert a tu_int to a fp_t, rounding toward even. */ +// Returns: convert a tu_int to a fp_t, rounding toward even. -/* Assumption: fp_t is a IEEE 128 bit floating point type - * tu_int is a 128 bit integral type - */ +// Assumption: fp_t is a IEEE 128 bit floating point type +// tu_int is a 128 bit integral type -/* seee eeee eeee eeee mmmm mmmm mmmm mmmm | mmmm mmmm mmmm mmmm mmmm mmmm mmmm mmmm | - * mmmm mmmm mmmm mmmm mmmm mmmm mmmm mmmm | mmmm mmmm mmmm mmmm mmmm mmmm mmmm mmmm - */ +// seee eeee eeee eeee mmmm mmmm mmmm mmmm | mmmm mmmm mmmm mmmm mmmm mmmm mmmm +// mmmm | mmmm mmmm mmmm mmmm mmmm mmmm mmmm mmmm | mmmm mmmm mmmm mmmm mmmm +// mmmm mmmm mmmm #if defined(CRT_HAS_128BIT) && defined(CRT_LDBL_128BIT) -COMPILER_RT_ABI fp_t -__floatuntitf(tu_int a) { - if (a == 0) - return 0.0; - const unsigned N = sizeof(tu_int) * CHAR_BIT; - int sd = N - __clzti2(a); /* number of significant digits */ - int e = sd - 1; /* exponent */ - if (sd > LDBL_MANT_DIG) { - /* start: 0000000000000000000001xxxxxxxxxxxxxxxxxxxxxxPQxxxxxxxxxxxxxxxxxx - * finish: 000000000000000000000000000000000000001xxxxxxxxxxxxxxxxxxxxxxPQR - * 12345678901234567890123456 - * 1 = msb 1 bit - * P = bit LDBL_MANT_DIG-1 bits to the right of 1 - * Q = bit LDBL_MANT_DIG bits to the right of 1 - * R = "or" of all bits to the right of Q - */ - switch (sd) { - case LDBL_MANT_DIG + 1: - a <<= 1; - break; - case LDBL_MANT_DIG + 2: - break; - default: - a = (a >> (sd - (LDBL_MANT_DIG+2))) | - ((a & ((tu_int)(-1) >> ((N + LDBL_MANT_DIG+2) - sd))) != 0); - }; - /* finish: */ - a |= (a & 4) != 0; /* Or P into R */ - ++a; /* round - this step may add a significant bit */ - a >>= 2; /* dump Q and R */ - /* a is now rounded to LDBL_MANT_DIG or LDBL_MANT_DIG+1 bits */ - if (a & ((tu_int)1 << LDBL_MANT_DIG)) { - a >>= 1; - ++e; - } - /* a is now rounded to LDBL_MANT_DIG bits */ - } else { - a <<= (LDBL_MANT_DIG - sd); - /* a is now rounded to LDBL_MANT_DIG bits */ +COMPILER_RT_ABI fp_t __floatuntitf(tu_int a) { + if (a == 0) + return 0.0; + const unsigned N = sizeof(tu_int) * CHAR_BIT; + int sd = N - __clzti2(a); // number of significant digits + int e = sd - 1; // exponent + if (sd > LDBL_MANT_DIG) { + // start: 0000000000000000000001xxxxxxxxxxxxxxxxxxxxxxPQxxxxxxxxxxxxxxxxxx + // finish: 000000000000000000000000000000000000001xxxxxxxxxxxxxxxxxxxxxxPQR + // 12345678901234567890123456 + // 1 = msb 1 bit + // P = bit LDBL_MANT_DIG-1 bits to the right of 1 + // Q = bit LDBL_MANT_DIG bits to the right of 1 + // R = "or" of all bits to the right of Q + switch (sd) { + case LDBL_MANT_DIG + 1: + a <<= 1; + break; + case LDBL_MANT_DIG + 2: + break; + default: + a = (a >> (sd - (LDBL_MANT_DIG + 2))) | + ((a & ((tu_int)(-1) >> ((N + LDBL_MANT_DIG + 2) - sd))) != 0); + }; + // finish: + a |= (a & 4) != 0; // Or P into R + ++a; // round - this step may add a significant bit + a >>= 2; // dump Q and R + // a is now rounded to LDBL_MANT_DIG or LDBL_MANT_DIG+1 bits + if (a & ((tu_int)1 << LDBL_MANT_DIG)) { + a >>= 1; + ++e; } + // a is now rounded to LDBL_MANT_DIG bits + } else { + a <<= (LDBL_MANT_DIG - sd); + // a is now rounded to LDBL_MANT_DIG bits + } - long_double_bits fb; - fb.u.high.all = (du_int)(e + 16383) << 48 /* exponent */ - | ((a >> 64) & 0x0000ffffffffffffLL); /* significand */ - fb.u.low.all = (du_int)(a); - return fb.f; + long_double_bits fb; + fb.u.high.all = (du_int)(e + 16383) << 48 // exponent + | ((a >> 64) & 0x0000ffffffffffffLL); // significand + fb.u.low.all = (du_int)(a); + return fb.f; } #endif diff --git a/lib/builtins/floatuntixf.c b/lib/builtins/floatuntixf.c index ea81cb1bcdae..efd8a27a0875 100644 --- a/lib/builtins/floatuntixf.c +++ b/lib/builtins/floatuntixf.c @@ -1,81 +1,70 @@ -/* ===-- floatuntixf.c - Implement __floatuntixf ---------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - * - * This file implements __floatuntixf for the compiler_rt library. - * - * ===----------------------------------------------------------------------=== - */ +//===-- floatuntixf.c - Implement __floatuntixf ---------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements __floatuntixf for the compiler_rt library. +// +//===----------------------------------------------------------------------===// #include "int_lib.h" #ifdef CRT_HAS_128BIT -/* Returns: convert a to a long double, rounding toward even. */ +// Returns: convert a to a long double, rounding toward even. -/* Assumption: long double is a IEEE 80 bit floating point type padded to 128 bits - * tu_int is a 128 bit integral type - */ +// Assumption: long double is a IEEE 80 bit floating point type padded to 128 +// bits tu_int is a 128 bit integral type -/* gggg gggg gggg gggg gggg gggg gggg gggg | gggg gggg gggg gggg seee eeee eeee eeee | - * 1mmm mmmm mmmm mmmm mmmm mmmm mmmm mmmm | mmmm mmmm mmmm mmmm mmmm mmmm mmmm mmmm - */ +// gggg gggg gggg gggg gggg gggg gggg gggg | gggg gggg gggg gggg seee eeee eeee +// eeee | 1mmm mmmm mmmm mmmm mmmm mmmm mmmm mmmm | mmmm mmmm mmmm mmmm mmmm +// mmmm mmmm mmmm -COMPILER_RT_ABI long double -__floatuntixf(tu_int a) -{ - if (a == 0) - return 0.0; - const unsigned N = sizeof(tu_int) * CHAR_BIT; - int sd = N - __clzti2(a); /* number of significant digits */ - int e = sd - 1; /* exponent */ - if (sd > LDBL_MANT_DIG) - { - /* start: 0000000000000000000001xxxxxxxxxxxxxxxxxxxxxxPQxxxxxxxxxxxxxxxxxx - * finish: 000000000000000000000000000000000000001xxxxxxxxxxxxxxxxxxxxxxPQR - * 12345678901234567890123456 - * 1 = msb 1 bit - * P = bit LDBL_MANT_DIG-1 bits to the right of 1 - * Q = bit LDBL_MANT_DIG bits to the right of 1 - * R = "or" of all bits to the right of Q - */ - switch (sd) - { - case LDBL_MANT_DIG + 1: - a <<= 1; - break; - case LDBL_MANT_DIG + 2: - break; - default: - a = (a >> (sd - (LDBL_MANT_DIG+2))) | - ((a & ((tu_int)(-1) >> ((N + LDBL_MANT_DIG+2) - sd))) != 0); - }; - /* finish: */ - a |= (a & 4) != 0; /* Or P into R */ - ++a; /* round - this step may add a significant bit */ - a >>= 2; /* dump Q and R */ - /* a is now rounded to LDBL_MANT_DIG or LDBL_MANT_DIG+1 bits */ - if (a & ((tu_int)1 << LDBL_MANT_DIG)) - { - a >>= 1; - ++e; - } - /* a is now rounded to LDBL_MANT_DIG bits */ +COMPILER_RT_ABI long double __floatuntixf(tu_int a) { + if (a == 0) + return 0.0; + const unsigned N = sizeof(tu_int) * CHAR_BIT; + int sd = N - __clzti2(a); // number of significant digits + int e = sd - 1; // exponent + if (sd > LDBL_MANT_DIG) { + // start: 0000000000000000000001xxxxxxxxxxxxxxxxxxxxxxPQxxxxxxxxxxxxxxxxxx + // finish: 000000000000000000000000000000000000001xxxxxxxxxxxxxxxxxxxxxxPQR + // 12345678901234567890123456 + // 1 = msb 1 bit + // P = bit LDBL_MANT_DIG-1 bits to the right of 1 + // Q = bit LDBL_MANT_DIG bits to the right of 1 + // R = "or" of all bits to the right of Q + switch (sd) { + case LDBL_MANT_DIG + 1: + a <<= 1; + break; + case LDBL_MANT_DIG + 2: + break; + default: + a = (a >> (sd - (LDBL_MANT_DIG + 2))) | + ((a & ((tu_int)(-1) >> ((N + LDBL_MANT_DIG + 2) - sd))) != 0); + }; + // finish: + a |= (a & 4) != 0; // Or P into R + ++a; // round - this step may add a significant bit + a >>= 2; // dump Q and R + // a is now rounded to LDBL_MANT_DIG or LDBL_MANT_DIG+1 bits + if (a & ((tu_int)1 << LDBL_MANT_DIG)) { + a >>= 1; + ++e; } - else - { - a <<= (LDBL_MANT_DIG - sd); - /* a is now rounded to LDBL_MANT_DIG bits */ - } - long_double_bits fb; - fb.u.high.s.low = (e + 16383); /* exponent */ - fb.u.low.all = (du_int)a; /* mantissa */ - return fb.f; + // a is now rounded to LDBL_MANT_DIG bits + } else { + a <<= (LDBL_MANT_DIG - sd); + // a is now rounded to LDBL_MANT_DIG bits + } + long_double_bits fb; + fb.u.high.s.low = (e + 16383); // exponent + fb.u.low.all = (du_int)a; // mantissa + return fb.f; } #endif diff --git a/lib/builtins/fp_add_impl.inc b/lib/builtins/fp_add_impl.inc index b47be1b648e6..da8639341703 100644 --- a/lib/builtins/fp_add_impl.inc +++ b/lib/builtins/fp_add_impl.inc @@ -1,9 +1,8 @@ //===----- lib/fp_add_impl.inc - floaing point addition -----------*- C -*-===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// // @@ -15,130 +14,144 @@ #include "fp_lib.h" static __inline fp_t __addXf3__(fp_t a, fp_t b) { - rep_t aRep = toRep(a); - rep_t bRep = toRep(b); - const rep_t aAbs = aRep & absMask; - const rep_t bAbs = bRep & absMask; - - // Detect if a or b is zero, infinity, or NaN. - if (aAbs - REP_C(1) >= infRep - REP_C(1) || - bAbs - REP_C(1) >= infRep - REP_C(1)) { - // NaN + anything = qNaN - if (aAbs > infRep) return fromRep(toRep(a) | quietBit); - // anything + NaN = qNaN - if (bAbs > infRep) return fromRep(toRep(b) | quietBit); - - if (aAbs == infRep) { - // +/-infinity + -/+infinity = qNaN - if ((toRep(a) ^ toRep(b)) == signBit) return fromRep(qnanRep); - // +/-infinity + anything remaining = +/- infinity - else return a; - } - - // anything remaining + +/-infinity = +/-infinity - if (bAbs == infRep) return b; - - // zero + anything = anything - if (!aAbs) { - // but we need to get the sign right for zero + zero - if (!bAbs) return fromRep(toRep(a) & toRep(b)); - else return b; - } - - // anything + zero = anything - if (!bAbs) return a; + rep_t aRep = toRep(a); + rep_t bRep = toRep(b); + const rep_t aAbs = aRep & absMask; + const rep_t bAbs = bRep & absMask; + + // Detect if a or b is zero, infinity, or NaN. + if (aAbs - REP_C(1) >= infRep - REP_C(1) || + bAbs - REP_C(1) >= infRep - REP_C(1)) { + // NaN + anything = qNaN + if (aAbs > infRep) + return fromRep(toRep(a) | quietBit); + // anything + NaN = qNaN + if (bAbs > infRep) + return fromRep(toRep(b) | quietBit); + + if (aAbs == infRep) { + // +/-infinity + -/+infinity = qNaN + if ((toRep(a) ^ toRep(b)) == signBit) + return fromRep(qnanRep); + // +/-infinity + anything remaining = +/- infinity + else + return a; } - // Swap a and b if necessary so that a has the larger absolute value. - if (bAbs > aAbs) { - const rep_t temp = aRep; - aRep = bRep; - bRep = temp; + // anything remaining + +/-infinity = +/-infinity + if (bAbs == infRep) + return b; + + // zero + anything = anything + if (!aAbs) { + // We need to get the sign right for zero + zero. + if (!bAbs) + return fromRep(toRep(a) & toRep(b)); + else + return b; } - // Extract the exponent and significand from the (possibly swapped) a and b. - int aExponent = aRep >> significandBits & maxExponent; - int bExponent = bRep >> significandBits & maxExponent; - rep_t aSignificand = aRep & significandMask; - rep_t bSignificand = bRep & significandMask; - - // Normalize any denormals, and adjust the exponent accordingly. - if (aExponent == 0) aExponent = normalize(&aSignificand); - if (bExponent == 0) bExponent = normalize(&bSignificand); - - // The sign of the result is the sign of the larger operand, a. If they - // have opposite signs, we are performing a subtraction; otherwise addition. - const rep_t resultSign = aRep & signBit; - const bool subtraction = (aRep ^ bRep) & signBit; - - // Shift the significands to give us round, guard and sticky, and or in the - // implicit significand bit. (If we fell through from the denormal path it - // was already set by normalize( ), but setting it twice won't hurt - // anything.) - aSignificand = (aSignificand | implicitBit) << 3; - bSignificand = (bSignificand | implicitBit) << 3; - - // Shift the significand of b by the difference in exponents, with a sticky - // bottom bit to get rounding correct. - const unsigned int align = aExponent - bExponent; - if (align) { - if (align < typeWidth) { - const bool sticky = bSignificand << (typeWidth - align); - bSignificand = bSignificand >> align | sticky; - } else { - bSignificand = 1; // sticky; b is known to be non-zero. - } - } - if (subtraction) { - aSignificand -= bSignificand; - // If a == -b, return +zero. - if (aSignificand == 0) return fromRep(0); - - // If partial cancellation occured, we need to left-shift the result - // and adjust the exponent: - if (aSignificand < implicitBit << 3) { - const int shift = rep_clz(aSignificand) - rep_clz(implicitBit << 3); - aSignificand <<= shift; - aExponent -= shift; - } + // anything + zero = anything + if (!bAbs) + return a; + } + + // Swap a and b if necessary so that a has the larger absolute value. + if (bAbs > aAbs) { + const rep_t temp = aRep; + aRep = bRep; + bRep = temp; + } + + // Extract the exponent and significand from the (possibly swapped) a and b. + int aExponent = aRep >> significandBits & maxExponent; + int bExponent = bRep >> significandBits & maxExponent; + rep_t aSignificand = aRep & significandMask; + rep_t bSignificand = bRep & significandMask; + + // Normalize any denormals, and adjust the exponent accordingly. + if (aExponent == 0) + aExponent = normalize(&aSignificand); + if (bExponent == 0) + bExponent = normalize(&bSignificand); + + // The sign of the result is the sign of the larger operand, a. If they + // have opposite signs, we are performing a subtraction. Otherwise, we + // perform addition. + const rep_t resultSign = aRep & signBit; + const bool subtraction = (aRep ^ bRep) & signBit; + + // Shift the significands to give us round, guard and sticky, and set the + // implicit significand bit. If we fell through from the denormal path it + // was already set by normalize( ), but setting it twice won't hurt + // anything. + aSignificand = (aSignificand | implicitBit) << 3; + bSignificand = (bSignificand | implicitBit) << 3; + + // Shift the significand of b by the difference in exponents, with a sticky + // bottom bit to get rounding correct. + const unsigned int align = aExponent - bExponent; + if (align) { + if (align < typeWidth) { + const bool sticky = bSignificand << (typeWidth - align); + bSignificand = bSignificand >> align | sticky; + } else { + bSignificand = 1; // Set the sticky bit. b is known to be non-zero. } - else /* addition */ { - aSignificand += bSignificand; - - // If the addition carried up, we need to right-shift the result and - // adjust the exponent: - if (aSignificand & implicitBit << 4) { - const bool sticky = aSignificand & 1; - aSignificand = aSignificand >> 1 | sticky; - aExponent += 1; - } + } + if (subtraction) { + aSignificand -= bSignificand; + // If a == -b, return +zero. + if (aSignificand == 0) + return fromRep(0); + + // If partial cancellation occured, we need to left-shift the result + // and adjust the exponent. + if (aSignificand < implicitBit << 3) { + const int shift = rep_clz(aSignificand) - rep_clz(implicitBit << 3); + aSignificand <<= shift; + aExponent -= shift; } - - // If we have overflowed the type, return +/- infinity: - if (aExponent >= maxExponent) return fromRep(infRep | resultSign); - - if (aExponent <= 0) { - // Result is denormal before rounding; the exponent is zero and we - // need to shift the significand. - const int shift = 1 - aExponent; - const bool sticky = aSignificand << (typeWidth - shift); - aSignificand = aSignificand >> shift | sticky; - aExponent = 0; + } else /* addition */ { + aSignificand += bSignificand; + + // If the addition carried up, we need to right-shift the result and + // adjust the exponent. + if (aSignificand & implicitBit << 4) { + const bool sticky = aSignificand & 1; + aSignificand = aSignificand >> 1 | sticky; + aExponent += 1; } - - // Low three bits are round, guard, and sticky. - const int roundGuardSticky = aSignificand & 0x7; - - // Shift the significand into place, and mask off the implicit bit. - rep_t result = aSignificand >> 3 & significandMask; - - // Insert the exponent and sign. - result |= (rep_t)aExponent << significandBits; - result |= resultSign; - - // Final rounding. The result may overflow to infinity, but that is the - // correct result in that case. - if (roundGuardSticky > 0x4) result++; - if (roundGuardSticky == 0x4) result += result & 1; - return fromRep(result); + } + + // If we have overflowed the type, return +/- infinity. + if (aExponent >= maxExponent) + return fromRep(infRep | resultSign); + + if (aExponent <= 0) { + // The result is denormal before rounding. The exponent is zero and we + // need to shift the significand. + const int shift = 1 - aExponent; + const bool sticky = aSignificand << (typeWidth - shift); + aSignificand = aSignificand >> shift | sticky; + aExponent = 0; + } + + // Low three bits are round, guard, and sticky. + const int roundGuardSticky = aSignificand & 0x7; + + // Shift the significand into place, and mask off the implicit bit. + rep_t result = aSignificand >> 3 & significandMask; + + // Insert the exponent and sign. + result |= (rep_t)aExponent << significandBits; + result |= resultSign; + + // Perform the final rounding. The result may overflow to infinity, but + // that is the correct result in that case. + if (roundGuardSticky > 0x4) + result++; + if (roundGuardSticky == 0x4) + result += result & 1; + return fromRep(result); } diff --git a/lib/builtins/fp_extend.h b/lib/builtins/fp_extend.h index 6d95a0680709..d2083c426722 100644 --- a/lib/builtins/fp_extend.h +++ b/lib/builtins/fp_extend.h @@ -1,9 +1,9 @@ -//===-lib/fp_extend.h - low precision -> high precision conversion -*- C -*-===// +//===-lib/fp_extend.h - low precision -> high precision conversion -*- C +//-*-===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// // @@ -30,12 +30,12 @@ typedef uint64_t src_rep_t; static const int srcSigBits = 52; static __inline int src_rep_t_clz(src_rep_t a) { #if defined __LP64__ - return __builtin_clzl(a); + return __builtin_clzl(a); #else - if (a & REP_C(0xffffffff00000000)) - return __builtin_clz(a >> 32); - else - return 32 + __builtin_clz(a & REP_C(0xffffffff)); + if (a & REP_C(0xffffffff00000000)) + return __builtin_clz(a >> 32); + else + return 32 + __builtin_clz(a & REP_C(0xffffffff)); #endif } @@ -48,7 +48,7 @@ static const int srcSigBits = 10; #else #error Source should be half, single, or double precision! -#endif //end source precision +#endif // end source precision #if defined DST_SINGLE typedef float dst_t; @@ -70,20 +70,26 @@ static const int dstSigBits = 112; #else #error Destination should be single, double, or quad precision! -#endif //end destination precision +#endif // end destination precision // End of specialization parameters. Two helper routines for conversion to and // from the representation of floating-point data as integer values follow. static __inline src_rep_t srcToRep(src_t x) { - const union { src_t f; src_rep_t i; } rep = {.f = x}; - return rep.i; + const union { + src_t f; + src_rep_t i; + } rep = {.f = x}; + return rep.i; } static __inline dst_t dstFromRep(dst_rep_t x) { - const union { dst_t f; dst_rep_t i; } rep = {.i = x}; - return rep.f; + const union { + dst_t f; + dst_rep_t i; + } rep = {.i = x}; + return rep.f; } // End helper routines. Conversion implementation follows. -#endif //FP_EXTEND_HEADER +#endif // FP_EXTEND_HEADER diff --git a/lib/builtins/fp_extend_impl.inc b/lib/builtins/fp_extend_impl.inc index b785cc7687ad..d1c9c02a00c5 100644 --- a/lib/builtins/fp_extend_impl.inc +++ b/lib/builtins/fp_extend_impl.inc @@ -1,9 +1,8 @@ //=-lib/fp_extend_impl.inc - low precision -> high precision conversion -*-- -// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// // @@ -28,81 +27,81 @@ // // Finally, the following assumptions are made: // -// 1. floating-point types and integer types have the same endianness on the -// target platform +// 1. Floating-point types and integer types have the same endianness on the +// target platform. // -// 2. quiet NaNs, if supported, are indicated by the leading bit of the -// significand field being set +// 2. Quiet NaNs, if supported, are indicated by the leading bit of the +// significand field being set. // //===----------------------------------------------------------------------===// #include "fp_extend.h" static __inline dst_t __extendXfYf2__(src_t a) { - // Various constants whose values follow from the type parameters. - // Any reasonable optimizer will fold and propagate all of these. - const int srcBits = sizeof(src_t)*CHAR_BIT; - const int srcExpBits = srcBits - srcSigBits - 1; - const int srcInfExp = (1 << srcExpBits) - 1; - const int srcExpBias = srcInfExp >> 1; + // Various constants whose values follow from the type parameters. + // Any reasonable optimizer will fold and propagate all of these. + const int srcBits = sizeof(src_t) * CHAR_BIT; + const int srcExpBits = srcBits - srcSigBits - 1; + const int srcInfExp = (1 << srcExpBits) - 1; + const int srcExpBias = srcInfExp >> 1; - const src_rep_t srcMinNormal = SRC_REP_C(1) << srcSigBits; - const src_rep_t srcInfinity = (src_rep_t)srcInfExp << srcSigBits; - const src_rep_t srcSignMask = SRC_REP_C(1) << (srcSigBits + srcExpBits); - const src_rep_t srcAbsMask = srcSignMask - 1; - const src_rep_t srcQNaN = SRC_REP_C(1) << (srcSigBits - 1); - const src_rep_t srcNaNCode = srcQNaN - 1; + const src_rep_t srcMinNormal = SRC_REP_C(1) << srcSigBits; + const src_rep_t srcInfinity = (src_rep_t)srcInfExp << srcSigBits; + const src_rep_t srcSignMask = SRC_REP_C(1) << (srcSigBits + srcExpBits); + const src_rep_t srcAbsMask = srcSignMask - 1; + const src_rep_t srcQNaN = SRC_REP_C(1) << (srcSigBits - 1); + const src_rep_t srcNaNCode = srcQNaN - 1; - const int dstBits = sizeof(dst_t)*CHAR_BIT; - const int dstExpBits = dstBits - dstSigBits - 1; - const int dstInfExp = (1 << dstExpBits) - 1; - const int dstExpBias = dstInfExp >> 1; + const int dstBits = sizeof(dst_t) * CHAR_BIT; + const int dstExpBits = dstBits - dstSigBits - 1; + const int dstInfExp = (1 << dstExpBits) - 1; + const int dstExpBias = dstInfExp >> 1; - const dst_rep_t dstMinNormal = DST_REP_C(1) << dstSigBits; + const dst_rep_t dstMinNormal = DST_REP_C(1) << dstSigBits; - // Break a into a sign and representation of the absolute value - const src_rep_t aRep = srcToRep(a); - const src_rep_t aAbs = aRep & srcAbsMask; - const src_rep_t sign = aRep & srcSignMask; - dst_rep_t absResult; + // Break a into a sign and representation of the absolute value. + const src_rep_t aRep = srcToRep(a); + const src_rep_t aAbs = aRep & srcAbsMask; + const src_rep_t sign = aRep & srcSignMask; + dst_rep_t absResult; - // If sizeof(src_rep_t) < sizeof(int), the subtraction result is promoted - // to (signed) int. To avoid that, explicitly cast to src_rep_t. - if ((src_rep_t)(aAbs - srcMinNormal) < srcInfinity - srcMinNormal) { - // a is a normal number. - // Extend to the destination type by shifting the significand and - // exponent into the proper position and rebiasing the exponent. - absResult = (dst_rep_t)aAbs << (dstSigBits - srcSigBits); - absResult += (dst_rep_t)(dstExpBias - srcExpBias) << dstSigBits; - } + // If sizeof(src_rep_t) < sizeof(int), the subtraction result is promoted + // to (signed) int. To avoid that, explicitly cast to src_rep_t. + if ((src_rep_t)(aAbs - srcMinNormal) < srcInfinity - srcMinNormal) { + // a is a normal number. + // Extend to the destination type by shifting the significand and + // exponent into the proper position and rebiasing the exponent. + absResult = (dst_rep_t)aAbs << (dstSigBits - srcSigBits); + absResult += (dst_rep_t)(dstExpBias - srcExpBias) << dstSigBits; + } - else if (aAbs >= srcInfinity) { - // a is NaN or infinity. - // Conjure the result by beginning with infinity, then setting the qNaN - // bit (if needed) and right-aligning the rest of the trailing NaN - // payload field. - absResult = (dst_rep_t)dstInfExp << dstSigBits; - absResult |= (dst_rep_t)(aAbs & srcQNaN) << (dstSigBits - srcSigBits); - absResult |= (dst_rep_t)(aAbs & srcNaNCode) << (dstSigBits - srcSigBits); - } + else if (aAbs >= srcInfinity) { + // a is NaN or infinity. + // Conjure the result by beginning with infinity, then setting the qNaN + // bit (if needed) and right-aligning the rest of the trailing NaN + // payload field. + absResult = (dst_rep_t)dstInfExp << dstSigBits; + absResult |= (dst_rep_t)(aAbs & srcQNaN) << (dstSigBits - srcSigBits); + absResult |= (dst_rep_t)(aAbs & srcNaNCode) << (dstSigBits - srcSigBits); + } - else if (aAbs) { - // a is denormal. - // renormalize the significand and clear the leading bit, then insert - // the correct adjusted exponent in the destination type. - const int scale = src_rep_t_clz(aAbs) - src_rep_t_clz(srcMinNormal); - absResult = (dst_rep_t)aAbs << (dstSigBits - srcSigBits + scale); - absResult ^= dstMinNormal; - const int resultExponent = dstExpBias - srcExpBias - scale + 1; - absResult |= (dst_rep_t)resultExponent << dstSigBits; - } + else if (aAbs) { + // a is denormal. + // renormalize the significand and clear the leading bit, then insert + // the correct adjusted exponent in the destination type. + const int scale = src_rep_t_clz(aAbs) - src_rep_t_clz(srcMinNormal); + absResult = (dst_rep_t)aAbs << (dstSigBits - srcSigBits + scale); + absResult ^= dstMinNormal; + const int resultExponent = dstExpBias - srcExpBias - scale + 1; + absResult |= (dst_rep_t)resultExponent << dstSigBits; + } - else { - // a is zero. - absResult = 0; - } + else { + // a is zero. + absResult = 0; + } - // Apply the signbit to (dst_t)abs(a). - const dst_rep_t result = absResult | (dst_rep_t)sign << (dstBits - srcBits); - return dstFromRep(result); + // Apply the signbit to the absolute value. + const dst_rep_t result = absResult | (dst_rep_t)sign << (dstBits - srcBits); + return dstFromRep(result); } diff --git a/lib/builtins/fp_fixint_impl.inc b/lib/builtins/fp_fixint_impl.inc index da70d4d39301..2196d712f05f 100644 --- a/lib/builtins/fp_fixint_impl.inc +++ b/lib/builtins/fp_fixint_impl.inc @@ -1,9 +1,8 @@ //===-- lib/fixdfsi.c - Double-precision -> integer conversion ----*- C -*-===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// // @@ -15,27 +14,27 @@ #include "fp_lib.h" static __inline fixint_t __fixint(fp_t a) { - const fixint_t fixint_max = (fixint_t)((~(fixuint_t)0) / 2); - const fixint_t fixint_min = -fixint_max - 1; - // Break a into sign, exponent, significand - const rep_t aRep = toRep(a); - const rep_t aAbs = aRep & absMask; - const fixint_t sign = aRep & signBit ? -1 : 1; - const int exponent = (aAbs >> significandBits) - exponentBias; - const rep_t significand = (aAbs & significandMask) | implicitBit; + const fixint_t fixint_max = (fixint_t)((~(fixuint_t)0) / 2); + const fixint_t fixint_min = -fixint_max - 1; + // Break a into sign, exponent, significand parts. + const rep_t aRep = toRep(a); + const rep_t aAbs = aRep & absMask; + const fixint_t sign = aRep & signBit ? -1 : 1; + const int exponent = (aAbs >> significandBits) - exponentBias; + const rep_t significand = (aAbs & significandMask) | implicitBit; - // If exponent is negative, the result is zero. - if (exponent < 0) - return 0; + // If exponent is negative, the result is zero. + if (exponent < 0) + return 0; - // If the value is too large for the integer type, saturate. - if ((unsigned)exponent >= sizeof(fixint_t) * CHAR_BIT) - return sign == 1 ? fixint_max : fixint_min; + // If the value is too large for the integer type, saturate. + if ((unsigned)exponent >= sizeof(fixint_t) * CHAR_BIT) + return sign == 1 ? fixint_max : fixint_min; - // If 0 <= exponent < significandBits, right shift to get the result. - // Otherwise, shift left. - if (exponent < significandBits) - return sign * (significand >> (significandBits - exponent)); - else - return sign * ((fixint_t)significand << (exponent - significandBits)); + // If 0 <= exponent < significandBits, right shift to get the result. + // Otherwise, shift left. + if (exponent < significandBits) + return sign * (significand >> (significandBits - exponent)); + else + return sign * ((fixint_t)significand << (exponent - significandBits)); } diff --git a/lib/builtins/fp_fixuint_impl.inc b/lib/builtins/fp_fixuint_impl.inc index d68ccf27a79c..cb2bf54ffaf5 100644 --- a/lib/builtins/fp_fixuint_impl.inc +++ b/lib/builtins/fp_fixuint_impl.inc @@ -1,9 +1,8 @@ //===-- lib/fixdfsi.c - Double-precision -> integer conversion ----*- C -*-===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// // @@ -15,25 +14,25 @@ #include "fp_lib.h" static __inline fixuint_t __fixuint(fp_t a) { - // Break a into sign, exponent, significand - const rep_t aRep = toRep(a); - const rep_t aAbs = aRep & absMask; - const int sign = aRep & signBit ? -1 : 1; - const int exponent = (aAbs >> significandBits) - exponentBias; - const rep_t significand = (aAbs & significandMask) | implicitBit; + // Break a into sign, exponent, significand parts. + const rep_t aRep = toRep(a); + const rep_t aAbs = aRep & absMask; + const int sign = aRep & signBit ? -1 : 1; + const int exponent = (aAbs >> significandBits) - exponentBias; + const rep_t significand = (aAbs & significandMask) | implicitBit; - // If either the value or the exponent is negative, the result is zero. - if (sign == -1 || exponent < 0) - return 0; + // If either the value or the exponent is negative, the result is zero. + if (sign == -1 || exponent < 0) + return 0; - // If the value is too large for the integer type, saturate. - if ((unsigned)exponent >= sizeof(fixuint_t) * CHAR_BIT) - return ~(fixuint_t)0; + // If the value is too large for the integer type, saturate. + if ((unsigned)exponent >= sizeof(fixuint_t) * CHAR_BIT) + return ~(fixuint_t)0; - // If 0 <= exponent < significandBits, right shift to get the result. - // Otherwise, shift left. - if (exponent < significandBits) - return significand >> (significandBits - exponent); - else - return (fixuint_t)significand << (exponent - significandBits); + // If 0 <= exponent < significandBits, right shift to get the result. + // Otherwise, shift left. + if (exponent < significandBits) + return significand >> (significandBits - exponent); + else + return (fixuint_t)significand << (exponent - significandBits); } diff --git a/lib/builtins/fp_lib.h b/lib/builtins/fp_lib.h index a0e19ab6a8f6..d1a988ea4713 100644 --- a/lib/builtins/fp_lib.h +++ b/lib/builtins/fp_lib.h @@ -1,9 +1,8 @@ //===-- lib/fp_lib.h - Floating-point utilities -------------------*- C -*-===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// // @@ -21,22 +20,22 @@ #ifndef FP_LIB_HEADER #define FP_LIB_HEADER -#include <stdint.h> -#include <stdbool.h> -#include <limits.h> #include "int_lib.h" #include "int_math.h" +#include <limits.h> +#include <stdbool.h> +#include <stdint.h> // x86_64 FreeBSD prior v9.3 define fixed-width types incorrectly in // 32-bit mode. #if defined(__FreeBSD__) && defined(__i386__) -# include <sys/param.h> -# if __FreeBSD_version < 903000 // v9.3 -# define uint64_t unsigned long long -# define int64_t long long -# undef UINT64_C -# define UINT64_C(c) (c ## ULL) -# endif +#include <sys/param.h> +#if __FreeBSD_version < 903000 // v9.3 +#define uint64_t unsigned long long +#define int64_t long long +#undef UINT64_C +#define UINT64_C(c) (c##ULL) +#endif #endif #if defined SINGLE_PRECISION @@ -47,15 +46,13 @@ typedef float fp_t; #define REP_C UINT32_C #define significandBits 23 -static __inline int rep_clz(rep_t a) { - return __builtin_clz(a); -} +static __inline int rep_clz(rep_t a) { return __builtin_clz(a); } // 32x32 --> 64 bit multiply static __inline void wideMultiply(rep_t a, rep_t b, rep_t *hi, rep_t *lo) { - const uint64_t product = (uint64_t)a*b; - *hi = product >> 32; - *lo = product; + const uint64_t product = (uint64_t)a * b; + *hi = product >> 32; + *lo = product; } COMPILER_RT_ABI fp_t __addsf3(fp_t a, fp_t b); @@ -69,12 +66,12 @@ typedef double fp_t; static __inline int rep_clz(rep_t a) { #if defined __LP64__ - return __builtin_clzl(a); + return __builtin_clzl(a); #else - if (a & REP_C(0xffffffff00000000)) - return __builtin_clz(a >> 32); - else - return 32 + __builtin_clz(a & REP_C(0xffffffff)); + if (a & REP_C(0xffffffff00000000)) + return __builtin_clz(a >> 32); + else + return 32 + __builtin_clz(a & REP_C(0xffffffff)); #endif } @@ -85,17 +82,17 @@ static __inline int rep_clz(rep_t a) { // many 64-bit platforms have this operation, but they tend to have hardware // floating-point, so we don't bother with a special case for them here. static __inline void wideMultiply(rep_t a, rep_t b, rep_t *hi, rep_t *lo) { - // Each of the component 32x32 -> 64 products - const uint64_t plolo = loWord(a) * loWord(b); - const uint64_t plohi = loWord(a) * hiWord(b); - const uint64_t philo = hiWord(a) * loWord(b); - const uint64_t phihi = hiWord(a) * hiWord(b); - // Sum terms that contribute to lo in a way that allows us to get the carry - const uint64_t r0 = loWord(plolo); - const uint64_t r1 = hiWord(plolo) + loWord(plohi) + loWord(philo); - *lo = r0 + (r1 << 32); - // Sum terms contributing to hi with the carry from lo - *hi = hiWord(plohi) + hiWord(philo) + hiWord(r1) + phihi; + // Each of the component 32x32 -> 64 products + const uint64_t plolo = loWord(a) * loWord(b); + const uint64_t plohi = loWord(a) * hiWord(b); + const uint64_t philo = hiWord(a) * loWord(b); + const uint64_t phihi = hiWord(a) * hiWord(b); + // Sum terms that contribute to lo in a way that allows us to get the carry + const uint64_t r0 = loWord(plolo); + const uint64_t r1 = hiWord(plolo) + loWord(plohi) + loWord(philo); + *lo = r0 + (r1 << 32); + // Sum terms contributing to hi with the carry from lo + *hi = hiWord(plohi) + hiWord(philo) + hiWord(r1) + phihi; } #undef loWord #undef hiWord @@ -103,7 +100,7 @@ static __inline void wideMultiply(rep_t a, rep_t b, rep_t *hi, rep_t *lo) { COMPILER_RT_ABI fp_t __adddf3(fp_t a, fp_t b); #elif defined QUAD_PRECISION -#if __LDBL_MANT_DIG__ == 113 +#if __LDBL_MANT_DIG__ == 113 && defined(__SIZEOF_INT128__) #define CRT_LDBL_128BIT typedef __uint128_t rep_t; typedef __int128_t srep_t; @@ -114,32 +111,34 @@ typedef long double fp_t; #define significandBits 112 static __inline int rep_clz(rep_t a) { - const union - { - __uint128_t ll; + const union { + __uint128_t ll; #if _YUGA_BIG_ENDIAN - struct { uint64_t high, low; } s; + struct { + uint64_t high, low; + } s; #else - struct { uint64_t low, high; } s; + struct { + uint64_t low, high; + } s; #endif - } uu = { .ll = a }; + } uu = {.ll = a}; - uint64_t word; - uint64_t add; + uint64_t word; + uint64_t add; - if (uu.s.high){ - word = uu.s.high; - add = 0; - } - else{ - word = uu.s.low; - add = 64; - } - return __builtin_clzll(word) + add; + if (uu.s.high) { + word = uu.s.high; + add = 0; + } else { + word = uu.s.low; + add = 64; + } + return __builtin_clzll(word) + add; } -#define Word_LoMask UINT64_C(0x00000000ffffffff) -#define Word_HiMask UINT64_C(0xffffffff00000000) +#define Word_LoMask UINT64_C(0x00000000ffffffff) +#define Word_HiMask UINT64_C(0xffffffff00000000) #define Word_FullMask UINT64_C(0xffffffffffffffff) #define Word_1(a) (uint64_t)((a >> 96) & Word_LoMask) #define Word_2(a) (uint64_t)((a >> 64) & Word_LoMask) @@ -151,55 +150,41 @@ static __inline int rep_clz(rep_t a) { // floating-point, so we don't bother with a special case for them here. static __inline void wideMultiply(rep_t a, rep_t b, rep_t *hi, rep_t *lo) { - const uint64_t product11 = Word_1(a) * Word_1(b); - const uint64_t product12 = Word_1(a) * Word_2(b); - const uint64_t product13 = Word_1(a) * Word_3(b); - const uint64_t product14 = Word_1(a) * Word_4(b); - const uint64_t product21 = Word_2(a) * Word_1(b); - const uint64_t product22 = Word_2(a) * Word_2(b); - const uint64_t product23 = Word_2(a) * Word_3(b); - const uint64_t product24 = Word_2(a) * Word_4(b); - const uint64_t product31 = Word_3(a) * Word_1(b); - const uint64_t product32 = Word_3(a) * Word_2(b); - const uint64_t product33 = Word_3(a) * Word_3(b); - const uint64_t product34 = Word_3(a) * Word_4(b); - const uint64_t product41 = Word_4(a) * Word_1(b); - const uint64_t product42 = Word_4(a) * Word_2(b); - const uint64_t product43 = Word_4(a) * Word_3(b); - const uint64_t product44 = Word_4(a) * Word_4(b); - - const __uint128_t sum0 = (__uint128_t)product44; - const __uint128_t sum1 = (__uint128_t)product34 + - (__uint128_t)product43; - const __uint128_t sum2 = (__uint128_t)product24 + - (__uint128_t)product33 + - (__uint128_t)product42; - const __uint128_t sum3 = (__uint128_t)product14 + - (__uint128_t)product23 + - (__uint128_t)product32 + - (__uint128_t)product41; - const __uint128_t sum4 = (__uint128_t)product13 + - (__uint128_t)product22 + - (__uint128_t)product31; - const __uint128_t sum5 = (__uint128_t)product12 + - (__uint128_t)product21; - const __uint128_t sum6 = (__uint128_t)product11; - - const __uint128_t r0 = (sum0 & Word_FullMask) + - ((sum1 & Word_LoMask) << 32); - const __uint128_t r1 = (sum0 >> 64) + - ((sum1 >> 32) & Word_FullMask) + - (sum2 & Word_FullMask) + - ((sum3 << 32) & Word_HiMask); - - *lo = r0 + (r1 << 64); - *hi = (r1 >> 64) + - (sum1 >> 96) + - (sum2 >> 64) + - (sum3 >> 32) + - sum4 + - (sum5 << 32) + - (sum6 << 64); + const uint64_t product11 = Word_1(a) * Word_1(b); + const uint64_t product12 = Word_1(a) * Word_2(b); + const uint64_t product13 = Word_1(a) * Word_3(b); + const uint64_t product14 = Word_1(a) * Word_4(b); + const uint64_t product21 = Word_2(a) * Word_1(b); + const uint64_t product22 = Word_2(a) * Word_2(b); + const uint64_t product23 = Word_2(a) * Word_3(b); + const uint64_t product24 = Word_2(a) * Word_4(b); + const uint64_t product31 = Word_3(a) * Word_1(b); + const uint64_t product32 = Word_3(a) * Word_2(b); + const uint64_t product33 = Word_3(a) * Word_3(b); + const uint64_t product34 = Word_3(a) * Word_4(b); + const uint64_t product41 = Word_4(a) * Word_1(b); + const uint64_t product42 = Word_4(a) * Word_2(b); + const uint64_t product43 = Word_4(a) * Word_3(b); + const uint64_t product44 = Word_4(a) * Word_4(b); + + const __uint128_t sum0 = (__uint128_t)product44; + const __uint128_t sum1 = (__uint128_t)product34 + (__uint128_t)product43; + const __uint128_t sum2 = + (__uint128_t)product24 + (__uint128_t)product33 + (__uint128_t)product42; + const __uint128_t sum3 = (__uint128_t)product14 + (__uint128_t)product23 + + (__uint128_t)product32 + (__uint128_t)product41; + const __uint128_t sum4 = + (__uint128_t)product13 + (__uint128_t)product22 + (__uint128_t)product31; + const __uint128_t sum5 = (__uint128_t)product12 + (__uint128_t)product21; + const __uint128_t sum6 = (__uint128_t)product11; + + const __uint128_t r0 = (sum0 & Word_FullMask) + ((sum1 & Word_LoMask) << 32); + const __uint128_t r1 = (sum0 >> 64) + ((sum1 >> 32) & Word_FullMask) + + (sum2 & Word_FullMask) + ((sum3 << 32) & Word_HiMask); + + *lo = r0 + (r1 << 64); + *hi = (r1 >> 64) + (sum1 >> 96) + (sum2 >> 64) + (sum3 >> 32) + sum4 + + (sum5 << 32) + (sum6 << 64); } #undef Word_1 #undef Word_2 @@ -208,63 +193,70 @@ static __inline void wideMultiply(rep_t a, rep_t b, rep_t *hi, rep_t *lo) { #undef Word_HiMask #undef Word_LoMask #undef Word_FullMask -#endif // __LDBL_MANT_DIG__ == 113 +#endif // __LDBL_MANT_DIG__ == 113 && __SIZEOF_INT128__ #else #error SINGLE_PRECISION, DOUBLE_PRECISION or QUAD_PRECISION must be defined. #endif -#if defined(SINGLE_PRECISION) || defined(DOUBLE_PRECISION) || defined(CRT_LDBL_128BIT) -#define typeWidth (sizeof(rep_t)*CHAR_BIT) -#define exponentBits (typeWidth - significandBits - 1) -#define maxExponent ((1 << exponentBits) - 1) -#define exponentBias (maxExponent >> 1) +#if defined(SINGLE_PRECISION) || defined(DOUBLE_PRECISION) || \ + defined(CRT_LDBL_128BIT) +#define typeWidth (sizeof(rep_t) * CHAR_BIT) +#define exponentBits (typeWidth - significandBits - 1) +#define maxExponent ((1 << exponentBits) - 1) +#define exponentBias (maxExponent >> 1) -#define implicitBit (REP_C(1) << significandBits) +#define implicitBit (REP_C(1) << significandBits) #define significandMask (implicitBit - 1U) -#define signBit (REP_C(1) << (significandBits + exponentBits)) -#define absMask (signBit - 1U) -#define exponentMask (absMask ^ significandMask) -#define oneRep ((rep_t)exponentBias << significandBits) -#define infRep exponentMask -#define quietBit (implicitBit >> 1) -#define qnanRep (exponentMask | quietBit) +#define signBit (REP_C(1) << (significandBits + exponentBits)) +#define absMask (signBit - 1U) +#define exponentMask (absMask ^ significandMask) +#define oneRep ((rep_t)exponentBias << significandBits) +#define infRep exponentMask +#define quietBit (implicitBit >> 1) +#define qnanRep (exponentMask | quietBit) static __inline rep_t toRep(fp_t x) { - const union { fp_t f; rep_t i; } rep = {.f = x}; - return rep.i; + const union { + fp_t f; + rep_t i; + } rep = {.f = x}; + return rep.i; } static __inline fp_t fromRep(rep_t x) { - const union { fp_t f; rep_t i; } rep = {.i = x}; - return rep.f; + const union { + fp_t f; + rep_t i; + } rep = {.i = x}; + return rep.f; } static __inline int normalize(rep_t *significand) { - const int shift = rep_clz(*significand) - rep_clz(implicitBit); - *significand <<= shift; - return 1 - shift; + const int shift = rep_clz(*significand) - rep_clz(implicitBit); + *significand <<= shift; + return 1 - shift; } static __inline void wideLeftShift(rep_t *hi, rep_t *lo, int count) { - *hi = *hi << count | *lo >> (typeWidth - count); - *lo = *lo << count; + *hi = *hi << count | *lo >> (typeWidth - count); + *lo = *lo << count; } -static __inline void wideRightShiftWithSticky(rep_t *hi, rep_t *lo, unsigned int count) { - if (count < typeWidth) { - const bool sticky = *lo << (typeWidth - count); - *lo = *hi << (typeWidth - count) | *lo >> count | sticky; - *hi = *hi >> count; - } - else if (count < 2*typeWidth) { - const bool sticky = *hi << (2*typeWidth - count) | *lo; - *lo = *hi >> (count - typeWidth) | sticky; - *hi = 0; - } else { - const bool sticky = *hi | *lo; - *lo = sticky; - *hi = 0; - } +static __inline void wideRightShiftWithSticky(rep_t *hi, rep_t *lo, + unsigned int count) { + if (count < typeWidth) { + const bool sticky = *lo << (typeWidth - count); + *lo = *hi << (typeWidth - count) | *lo >> count | sticky; + *hi = *hi >> count; + } else if (count < 2 * typeWidth) { + const bool sticky = *hi << (2 * typeWidth - count) | *lo; + *lo = *hi >> (count - typeWidth) | sticky; + *hi = 0; + } else { + const bool sticky = *hi | *lo; + *lo = sticky; + *hi = 0; + } } // Implements logb methods (logb, logbf, logbl) for IEEE-754. This avoids @@ -280,9 +272,9 @@ static __inline fp_t __compiler_rt_logbX(fp_t x) { // 2) 0.0 returns -inf if (exp == maxExponent) { if (((rep & signBit) == 0) || (x != x)) { - return x; // NaN or +inf: return x + return x; // NaN or +inf: return x } else { - return -x; // -inf: return -x + return -x; // -inf: return -x } } else if (x == 0.0) { // 0.0: return -inf @@ -291,13 +283,13 @@ static __inline fp_t __compiler_rt_logbX(fp_t x) { if (exp != 0) { // Normal number - return exp - exponentBias; // Unbias exponent + return exp - exponentBias; // Unbias exponent } else { // Subnormal number; normalize and repeat rep &= absMask; const int shift = 1 - normalize(&rep); exp = (rep & exponentMask) >> significandBits; - return exp - exponentBias - shift; // Unbias exponent + return exp - exponentBias - shift; // Unbias exponent } } #endif @@ -311,17 +303,17 @@ static __inline fp_t __compiler_rt_logb(fp_t x) { return __compiler_rt_logbX(x); } #elif defined(QUAD_PRECISION) - #if defined(CRT_LDBL_128BIT) +#if defined(CRT_LDBL_128BIT) static __inline fp_t __compiler_rt_logbl(fp_t x) { return __compiler_rt_logbX(x); } - #else +#else // The generic implementation only works for ieee754 floating point. For other // floating point types, continue to rely on the libm implementation for now. static __inline long double __compiler_rt_logbl(long double x) { return crt_logbl(x); } - #endif +#endif #endif #endif // FP_LIB_HEADER diff --git a/lib/builtins/fp_mul_impl.inc b/lib/builtins/fp_mul_impl.inc index b34aa1b8f544..a93f2d78ad6b 100644 --- a/lib/builtins/fp_mul_impl.inc +++ b/lib/builtins/fp_mul_impl.inc @@ -1,9 +1,8 @@ //===---- lib/fp_mul_impl.inc - floating point multiplication -----*- C -*-===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// // @@ -15,102 +14,115 @@ #include "fp_lib.h" static __inline fp_t __mulXf3__(fp_t a, fp_t b) { - const unsigned int aExponent = toRep(a) >> significandBits & maxExponent; - const unsigned int bExponent = toRep(b) >> significandBits & maxExponent; - const rep_t productSign = (toRep(a) ^ toRep(b)) & signBit; - - rep_t aSignificand = toRep(a) & significandMask; - rep_t bSignificand = toRep(b) & significandMask; - int scale = 0; - - // Detect if a or b is zero, denormal, infinity, or NaN. - if (aExponent-1U >= maxExponent-1U || bExponent-1U >= maxExponent-1U) { - - const rep_t aAbs = toRep(a) & absMask; - const rep_t bAbs = toRep(b) & absMask; - - // NaN * anything = qNaN - if (aAbs > infRep) return fromRep(toRep(a) | quietBit); - // anything * NaN = qNaN - if (bAbs > infRep) return fromRep(toRep(b) | quietBit); - - if (aAbs == infRep) { - // infinity * non-zero = +/- infinity - if (bAbs) return fromRep(aAbs | productSign); - // infinity * zero = NaN - else return fromRep(qnanRep); - } - - if (bAbs == infRep) { - //? non-zero * infinity = +/- infinity - if (aAbs) return fromRep(bAbs | productSign); - // zero * infinity = NaN - else return fromRep(qnanRep); - } - - // zero * anything = +/- zero - if (!aAbs) return fromRep(productSign); - // anything * zero = +/- zero - if (!bAbs) return fromRep(productSign); - - // one or both of a or b is denormal, the other (if applicable) is a - // normal number. Renormalize one or both of a and b, and set scale to - // include the necessary exponent adjustment. - if (aAbs < implicitBit) scale += normalize(&aSignificand); - if (bAbs < implicitBit) scale += normalize(&bSignificand); + const unsigned int aExponent = toRep(a) >> significandBits & maxExponent; + const unsigned int bExponent = toRep(b) >> significandBits & maxExponent; + const rep_t productSign = (toRep(a) ^ toRep(b)) & signBit; + + rep_t aSignificand = toRep(a) & significandMask; + rep_t bSignificand = toRep(b) & significandMask; + int scale = 0; + + // Detect if a or b is zero, denormal, infinity, or NaN. + if (aExponent - 1U >= maxExponent - 1U || + bExponent - 1U >= maxExponent - 1U) { + + const rep_t aAbs = toRep(a) & absMask; + const rep_t bAbs = toRep(b) & absMask; + + // NaN * anything = qNaN + if (aAbs > infRep) + return fromRep(toRep(a) | quietBit); + // anything * NaN = qNaN + if (bAbs > infRep) + return fromRep(toRep(b) | quietBit); + + if (aAbs == infRep) { + // infinity * non-zero = +/- infinity + if (bAbs) + return fromRep(aAbs | productSign); + // infinity * zero = NaN + else + return fromRep(qnanRep); } - // Or in the implicit significand bit. (If we fell through from the - // denormal path it was already set by normalize( ), but setting it twice - // won't hurt anything.) - aSignificand |= implicitBit; - bSignificand |= implicitBit; - - // Get the significand of a*b. Before multiplying the significands, shift - // one of them left to left-align it in the field. Thus, the product will - // have (exponentBits + 2) integral digits, all but two of which must be - // zero. Normalizing this result is just a conditional left-shift by one - // and bumping the exponent accordingly. - rep_t productHi, productLo; - wideMultiply(aSignificand, bSignificand << exponentBits, - &productHi, &productLo); - - int productExponent = aExponent + bExponent - exponentBias + scale; - - // Normalize the significand, adjust exponent if needed. - if (productHi & implicitBit) productExponent++; - else wideLeftShift(&productHi, &productLo, 1); - - // If we have overflowed the type, return +/- infinity. - if (productExponent >= maxExponent) return fromRep(infRep | productSign); - - if (productExponent <= 0) { - // Result is denormal before rounding - // - // If the result is so small that it just underflows to zero, return - // a zero of the appropriate sign. Mathematically there is no need to - // handle this case separately, but we make it a special case to - // simplify the shift logic. - const unsigned int shift = REP_C(1) - (unsigned int)productExponent; - if (shift >= typeWidth) return fromRep(productSign); - - // Otherwise, shift the significand of the result so that the round - // bit is the high bit of productLo. - wideRightShiftWithSticky(&productHi, &productLo, shift); - } - else { - // Result is normal before rounding; insert the exponent. - productHi &= significandMask; - productHi |= (rep_t)productExponent << significandBits; + if (bAbs == infRep) { + // non-zero * infinity = +/- infinity + if (aAbs) + return fromRep(bAbs | productSign); + // zero * infinity = NaN + else + return fromRep(qnanRep); } - // Insert the sign of the result: - productHi |= productSign; - - // Final rounding. The final result may overflow to infinity, or underflow - // to zero, but those are the correct results in those cases. We use the - // default IEEE-754 round-to-nearest, ties-to-even rounding mode. - if (productLo > signBit) productHi++; - if (productLo == signBit) productHi += productHi & 1; - return fromRep(productHi); + // zero * anything = +/- zero + if (!aAbs) + return fromRep(productSign); + // anything * zero = +/- zero + if (!bAbs) + return fromRep(productSign); + + // One or both of a or b is denormal. The other (if applicable) is a + // normal number. Renormalize one or both of a and b, and set scale to + // include the necessary exponent adjustment. + if (aAbs < implicitBit) + scale += normalize(&aSignificand); + if (bAbs < implicitBit) + scale += normalize(&bSignificand); + } + + // Set the implicit significand bit. If we fell through from the + // denormal path it was already set by normalize( ), but setting it twice + // won't hurt anything. + aSignificand |= implicitBit; + bSignificand |= implicitBit; + + // Perform a basic multiplication on the significands. One of them must be + // shifted beforehand to be aligned with the exponent. + rep_t productHi, productLo; + wideMultiply(aSignificand, bSignificand << exponentBits, &productHi, + &productLo); + + int productExponent = aExponent + bExponent - exponentBias + scale; + + // Normalize the significand and adjust the exponent if needed. + if (productHi & implicitBit) + productExponent++; + else + wideLeftShift(&productHi, &productLo, 1); + + // If we have overflowed the type, return +/- infinity. + if (productExponent >= maxExponent) + return fromRep(infRep | productSign); + + if (productExponent <= 0) { + // The result is denormal before rounding. + // + // If the result is so small that it just underflows to zero, return + // zero with the appropriate sign. Mathematically, there is no need to + // handle this case separately, but we make it a special case to + // simplify the shift logic. + const unsigned int shift = REP_C(1) - (unsigned int)productExponent; + if (shift >= typeWidth) + return fromRep(productSign); + + // Otherwise, shift the significand of the result so that the round + // bit is the high bit of productLo. + wideRightShiftWithSticky(&productHi, &productLo, shift); + } else { + // The result is normal before rounding. Insert the exponent. + productHi &= significandMask; + productHi |= (rep_t)productExponent << significandBits; + } + + // Insert the sign of the result. + productHi |= productSign; + + // Perform the final rounding. The final result may overflow to infinity, + // or underflow to zero, but those are the correct results in those cases. + // We use the default IEEE-754 round-to-nearest, ties-to-even rounding mode. + if (productLo > signBit) + productHi++; + if (productLo == signBit) + productHi += productHi & 1; + return fromRep(productHi); } diff --git a/lib/builtins/fp_trunc.h b/lib/builtins/fp_trunc.h index d5e79bb5b863..aca4c9b6e677 100644 --- a/lib/builtins/fp_trunc.h +++ b/lib/builtins/fp_trunc.h @@ -1,9 +1,8 @@ //=== lib/fp_trunc.h - high precision -> low precision conversion *- C -*-===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// // @@ -36,7 +35,7 @@ static const int srcSigBits = 112; #else #error Source should be double precision or quad precision! -#endif //end source precision +#endif // end source precision #if defined DST_DOUBLE typedef double dst_t; @@ -58,19 +57,25 @@ static const int dstSigBits = 10; #else #error Destination should be single precision or double precision! -#endif //end destination precision +#endif // end destination precision // End of specialization parameters. Two helper routines for conversion to and // from the representation of floating-point data as integer values follow. static __inline src_rep_t srcToRep(src_t x) { - const union { src_t f; src_rep_t i; } rep = {.f = x}; - return rep.i; + const union { + src_t f; + src_rep_t i; + } rep = {.f = x}; + return rep.i; } static __inline dst_t dstFromRep(dst_rep_t x) { - const union { dst_t f; dst_rep_t i; } rep = {.i = x}; - return rep.f; + const union { + dst_t f; + dst_rep_t i; + } rep = {.i = x}; + return rep.f; } #endif // FP_TRUNC_HEADER diff --git a/lib/builtins/fp_trunc_impl.inc b/lib/builtins/fp_trunc_impl.inc index d88ae060913f..133c8bbe5c2f 100644 --- a/lib/builtins/fp_trunc_impl.inc +++ b/lib/builtins/fp_trunc_impl.inc @@ -1,9 +1,8 @@ //= lib/fp_trunc_impl.inc - high precision -> low precision conversion *-*-===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// // @@ -29,107 +28,105 @@ // // Finally, the following assumptions are made: // -// 1. floating-point types and integer types have the same endianness on the -// target platform +// 1. Floating-point types and integer types have the same endianness on the +// target platform. // -// 2. quiet NaNs, if supported, are indicated by the leading bit of the -// significand field being set +// 2. Quiet NaNs, if supported, are indicated by the leading bit of the +// significand field being set. // //===----------------------------------------------------------------------===// #include "fp_trunc.h" static __inline dst_t __truncXfYf2__(src_t a) { - // Various constants whose values follow from the type parameters. - // Any reasonable optimizer will fold and propagate all of these. - const int srcBits = sizeof(src_t)*CHAR_BIT; - const int srcExpBits = srcBits - srcSigBits - 1; - const int srcInfExp = (1 << srcExpBits) - 1; - const int srcExpBias = srcInfExp >> 1; + // Various constants whose values follow from the type parameters. + // Any reasonable optimizer will fold and propagate all of these. + const int srcBits = sizeof(src_t) * CHAR_BIT; + const int srcExpBits = srcBits - srcSigBits - 1; + const int srcInfExp = (1 << srcExpBits) - 1; + const int srcExpBias = srcInfExp >> 1; - const src_rep_t srcMinNormal = SRC_REP_C(1) << srcSigBits; - const src_rep_t srcSignificandMask = srcMinNormal - 1; - const src_rep_t srcInfinity = (src_rep_t)srcInfExp << srcSigBits; - const src_rep_t srcSignMask = SRC_REP_C(1) << (srcSigBits + srcExpBits); - const src_rep_t srcAbsMask = srcSignMask - 1; - const src_rep_t roundMask = (SRC_REP_C(1) << (srcSigBits - dstSigBits)) - 1; - const src_rep_t halfway = SRC_REP_C(1) << (srcSigBits - dstSigBits - 1); - const src_rep_t srcQNaN = SRC_REP_C(1) << (srcSigBits - 1); - const src_rep_t srcNaNCode = srcQNaN - 1; + const src_rep_t srcMinNormal = SRC_REP_C(1) << srcSigBits; + const src_rep_t srcSignificandMask = srcMinNormal - 1; + const src_rep_t srcInfinity = (src_rep_t)srcInfExp << srcSigBits; + const src_rep_t srcSignMask = SRC_REP_C(1) << (srcSigBits + srcExpBits); + const src_rep_t srcAbsMask = srcSignMask - 1; + const src_rep_t roundMask = (SRC_REP_C(1) << (srcSigBits - dstSigBits)) - 1; + const src_rep_t halfway = SRC_REP_C(1) << (srcSigBits - dstSigBits - 1); + const src_rep_t srcQNaN = SRC_REP_C(1) << (srcSigBits - 1); + const src_rep_t srcNaNCode = srcQNaN - 1; - const int dstBits = sizeof(dst_t)*CHAR_BIT; - const int dstExpBits = dstBits - dstSigBits - 1; - const int dstInfExp = (1 << dstExpBits) - 1; - const int dstExpBias = dstInfExp >> 1; + const int dstBits = sizeof(dst_t) * CHAR_BIT; + const int dstExpBits = dstBits - dstSigBits - 1; + const int dstInfExp = (1 << dstExpBits) - 1; + const int dstExpBias = dstInfExp >> 1; - const int underflowExponent = srcExpBias + 1 - dstExpBias; - const int overflowExponent = srcExpBias + dstInfExp - dstExpBias; - const src_rep_t underflow = (src_rep_t)underflowExponent << srcSigBits; - const src_rep_t overflow = (src_rep_t)overflowExponent << srcSigBits; + const int underflowExponent = srcExpBias + 1 - dstExpBias; + const int overflowExponent = srcExpBias + dstInfExp - dstExpBias; + const src_rep_t underflow = (src_rep_t)underflowExponent << srcSigBits; + const src_rep_t overflow = (src_rep_t)overflowExponent << srcSigBits; - const dst_rep_t dstQNaN = DST_REP_C(1) << (dstSigBits - 1); - const dst_rep_t dstNaNCode = dstQNaN - 1; + const dst_rep_t dstQNaN = DST_REP_C(1) << (dstSigBits - 1); + const dst_rep_t dstNaNCode = dstQNaN - 1; - // Break a into a sign and representation of the absolute value - const src_rep_t aRep = srcToRep(a); - const src_rep_t aAbs = aRep & srcAbsMask; - const src_rep_t sign = aRep & srcSignMask; - dst_rep_t absResult; + // Break a into a sign and representation of the absolute value. + const src_rep_t aRep = srcToRep(a); + const src_rep_t aAbs = aRep & srcAbsMask; + const src_rep_t sign = aRep & srcSignMask; + dst_rep_t absResult; - if (aAbs - underflow < aAbs - overflow) { - // The exponent of a is within the range of normal numbers in the - // destination format. We can convert by simply right-shifting with - // rounding and adjusting the exponent. - absResult = aAbs >> (srcSigBits - dstSigBits); - absResult -= (dst_rep_t)(srcExpBias - dstExpBias) << dstSigBits; + if (aAbs - underflow < aAbs - overflow) { + // The exponent of a is within the range of normal numbers in the + // destination format. We can convert by simply right-shifting with + // rounding and adjusting the exponent. + absResult = aAbs >> (srcSigBits - dstSigBits); + absResult -= (dst_rep_t)(srcExpBias - dstExpBias) << dstSigBits; - const src_rep_t roundBits = aAbs & roundMask; - // Round to nearest - if (roundBits > halfway) - absResult++; - // Ties to even - else if (roundBits == halfway) - absResult += absResult & 1; - } - else if (aAbs > srcInfinity) { - // a is NaN. - // Conjure the result by beginning with infinity, setting the qNaN - // bit and inserting the (truncated) trailing NaN field. - absResult = (dst_rep_t)dstInfExp << dstSigBits; - absResult |= dstQNaN; - absResult |= ((aAbs & srcNaNCode) >> (srcSigBits - dstSigBits)) & dstNaNCode; - } - else if (aAbs >= overflow) { - // a overflows to infinity. - absResult = (dst_rep_t)dstInfExp << dstSigBits; - } - else { - // a underflows on conversion to the destination type or is an exact - // zero. The result may be a denormal or zero. Extract the exponent - // to get the shift amount for the denormalization. - const int aExp = aAbs >> srcSigBits; - const int shift = srcExpBias - dstExpBias - aExp + 1; + const src_rep_t roundBits = aAbs & roundMask; + // Round to nearest. + if (roundBits > halfway) + absResult++; + // Tie to even. + else if (roundBits == halfway) + absResult += absResult & 1; + } else if (aAbs > srcInfinity) { + // a is NaN. + // Conjure the result by beginning with infinity, setting the qNaN + // bit and inserting the (truncated) trailing NaN field. + absResult = (dst_rep_t)dstInfExp << dstSigBits; + absResult |= dstQNaN; + absResult |= + ((aAbs & srcNaNCode) >> (srcSigBits - dstSigBits)) & dstNaNCode; + } else if (aAbs >= overflow) { + // a overflows to infinity. + absResult = (dst_rep_t)dstInfExp << dstSigBits; + } else { + // a underflows on conversion to the destination type or is an exact + // zero. The result may be a denormal or zero. Extract the exponent + // to get the shift amount for the denormalization. + const int aExp = aAbs >> srcSigBits; + const int shift = srcExpBias - dstExpBias - aExp + 1; - const src_rep_t significand = (aRep & srcSignificandMask) | srcMinNormal; + const src_rep_t significand = (aRep & srcSignificandMask) | srcMinNormal; - // Right shift by the denormalization amount with sticky. - if (shift > srcSigBits) { - absResult = 0; - } else { - const bool sticky = significand << (srcBits - shift); - src_rep_t denormalizedSignificand = significand >> shift | sticky; - absResult = denormalizedSignificand >> (srcSigBits - dstSigBits); - const src_rep_t roundBits = denormalizedSignificand & roundMask; - // Round to nearest - if (roundBits > halfway) - absResult++; - // Ties to even - else if (roundBits == halfway) - absResult += absResult & 1; - } + // Right shift by the denormalization amount with sticky. + if (shift > srcSigBits) { + absResult = 0; + } else { + const bool sticky = significand << (srcBits - shift); + src_rep_t denormalizedSignificand = significand >> shift | sticky; + absResult = denormalizedSignificand >> (srcSigBits - dstSigBits); + const src_rep_t roundBits = denormalizedSignificand & roundMask; + // Round to nearest + if (roundBits > halfway) + absResult++; + // Ties to even + else if (roundBits == halfway) + absResult += absResult & 1; } + } - // Apply the signbit to (dst_t)abs(a). - const dst_rep_t result = absResult | sign >> (srcBits - dstBits); - return dstFromRep(result); + // Apply the signbit to the absolute value. + const dst_rep_t result = absResult | sign >> (srcBits - dstBits); + return dstFromRep(result); } diff --git a/lib/builtins/gcc_personality_v0.c b/lib/builtins/gcc_personality_v0.c index 68581ef16434..d12ee03c49fe 100644 --- a/lib/builtins/gcc_personality_v0.c +++ b/lib/builtins/gcc_personality_v0.c @@ -1,145 +1,135 @@ -/* ===-- gcc_personality_v0.c - Implement __gcc_personality_v0 -------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - * - */ +//===-- gcc_personality_v0.c - Implement __gcc_personality_v0 -------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// #include "int_lib.h" #include <unwind.h> -#if defined(__arm__) && !defined(__ARM_DWARF_EH__) && !defined(__USING_SJLJ_EXCEPTIONS__) -/* - * When building with older compilers (e.g. clang <3.9), it is possible that we - * have a version of unwind.h which does not provide the EHABI declarations - * which are quired for the C personality to conform to the specification. In - * order to provide forward compatibility for such compilers, we re-declare the - * necessary interfaces in the helper to permit a standalone compilation of the - * builtins (which contains the C unwinding personality for historical reasons). - */ +#if defined(__arm__) && !defined(__ARM_DWARF_EH__) && \ + !defined(__USING_SJLJ_EXCEPTIONS__) +// When building with older compilers (e.g. clang <3.9), it is possible that we +// have a version of unwind.h which does not provide the EHABI declarations +// which are quired for the C personality to conform to the specification. In +// order to provide forward compatibility for such compilers, we re-declare the +// necessary interfaces in the helper to permit a standalone compilation of the +// builtins (which contains the C unwinding personality for historical reasons). #include "unwind-ehabi-helpers.h" #endif -/* - * Pointer encodings documented at: - * http://refspecs.freestandards.org/LSB_1.3.0/gLSB/gLSB/ehframehdr.html - */ - -#define DW_EH_PE_omit 0xff /* no data follows */ - -#define DW_EH_PE_absptr 0x00 -#define DW_EH_PE_uleb128 0x01 -#define DW_EH_PE_udata2 0x02 -#define DW_EH_PE_udata4 0x03 -#define DW_EH_PE_udata8 0x04 -#define DW_EH_PE_sleb128 0x09 -#define DW_EH_PE_sdata2 0x0A -#define DW_EH_PE_sdata4 0x0B -#define DW_EH_PE_sdata8 0x0C - -#define DW_EH_PE_pcrel 0x10 -#define DW_EH_PE_textrel 0x20 -#define DW_EH_PE_datarel 0x30 -#define DW_EH_PE_funcrel 0x40 -#define DW_EH_PE_aligned 0x50 -#define DW_EH_PE_indirect 0x80 /* gcc extension */ - - - -/* read a uleb128 encoded value and advance pointer */ -static uintptr_t readULEB128(const uint8_t** data) -{ - uintptr_t result = 0; - uintptr_t shift = 0; - unsigned char byte; - const uint8_t* p = *data; - do { - byte = *p++; - result |= (byte & 0x7f) << shift; - shift += 7; - } while (byte & 0x80); - *data = p; - return result; +// Pointer encodings documented at: +// http://refspecs.freestandards.org/LSB_1.3.0/gLSB/gLSB/ehframehdr.html + +#define DW_EH_PE_omit 0xff // no data follows + +#define DW_EH_PE_absptr 0x00 +#define DW_EH_PE_uleb128 0x01 +#define DW_EH_PE_udata2 0x02 +#define DW_EH_PE_udata4 0x03 +#define DW_EH_PE_udata8 0x04 +#define DW_EH_PE_sleb128 0x09 +#define DW_EH_PE_sdata2 0x0A +#define DW_EH_PE_sdata4 0x0B +#define DW_EH_PE_sdata8 0x0C + +#define DW_EH_PE_pcrel 0x10 +#define DW_EH_PE_textrel 0x20 +#define DW_EH_PE_datarel 0x30 +#define DW_EH_PE_funcrel 0x40 +#define DW_EH_PE_aligned 0x50 +#define DW_EH_PE_indirect 0x80 // gcc extension + +// read a uleb128 encoded value and advance pointer +static uintptr_t readULEB128(const uint8_t **data) { + uintptr_t result = 0; + uintptr_t shift = 0; + unsigned char byte; + const uint8_t *p = *data; + do { + byte = *p++; + result |= (byte & 0x7f) << shift; + shift += 7; + } while (byte & 0x80); + *data = p; + return result; } -/* read a pointer encoded value and advance pointer */ -static uintptr_t readEncodedPointer(const uint8_t** data, uint8_t encoding) -{ - const uint8_t* p = *data; - uintptr_t result = 0; - - if ( encoding == DW_EH_PE_omit ) - return 0; - - /* first get value */ - switch (encoding & 0x0F) { - case DW_EH_PE_absptr: - result = *((const uintptr_t*)p); - p += sizeof(uintptr_t); - break; - case DW_EH_PE_uleb128: - result = readULEB128(&p); - break; - case DW_EH_PE_udata2: - result = *((const uint16_t*)p); - p += sizeof(uint16_t); - break; - case DW_EH_PE_udata4: - result = *((const uint32_t*)p); - p += sizeof(uint32_t); - break; - case DW_EH_PE_udata8: - result = *((const uint64_t*)p); - p += sizeof(uint64_t); - break; - case DW_EH_PE_sdata2: - result = *((const int16_t*)p); - p += sizeof(int16_t); - break; - case DW_EH_PE_sdata4: - result = *((const int32_t*)p); - p += sizeof(int32_t); - break; - case DW_EH_PE_sdata8: - result = *((const int64_t*)p); - p += sizeof(int64_t); - break; - case DW_EH_PE_sleb128: - default: - /* not supported */ - compilerrt_abort(); - break; - } - - /* then add relative offset */ - switch ( encoding & 0x70 ) { - case DW_EH_PE_absptr: - /* do nothing */ - break; - case DW_EH_PE_pcrel: - result += (uintptr_t)(*data); - break; - case DW_EH_PE_textrel: - case DW_EH_PE_datarel: - case DW_EH_PE_funcrel: - case DW_EH_PE_aligned: - default: - /* not supported */ - compilerrt_abort(); - break; - } - - /* then apply indirection */ - if (encoding & DW_EH_PE_indirect) { - result = *((const uintptr_t*)result); - } - - *data = p; - return result; +// read a pointer encoded value and advance pointer +static uintptr_t readEncodedPointer(const uint8_t **data, uint8_t encoding) { + const uint8_t *p = *data; + uintptr_t result = 0; + + if (encoding == DW_EH_PE_omit) + return 0; + + // first get value + switch (encoding & 0x0F) { + case DW_EH_PE_absptr: + result = *((const uintptr_t *)p); + p += sizeof(uintptr_t); + break; + case DW_EH_PE_uleb128: + result = readULEB128(&p); + break; + case DW_EH_PE_udata2: + result = *((const uint16_t *)p); + p += sizeof(uint16_t); + break; + case DW_EH_PE_udata4: + result = *((const uint32_t *)p); + p += sizeof(uint32_t); + break; + case DW_EH_PE_udata8: + result = *((const uint64_t *)p); + p += sizeof(uint64_t); + break; + case DW_EH_PE_sdata2: + result = *((const int16_t *)p); + p += sizeof(int16_t); + break; + case DW_EH_PE_sdata4: + result = *((const int32_t *)p); + p += sizeof(int32_t); + break; + case DW_EH_PE_sdata8: + result = *((const int64_t *)p); + p += sizeof(int64_t); + break; + case DW_EH_PE_sleb128: + default: + // not supported + compilerrt_abort(); + break; + } + + // then add relative offset + switch (encoding & 0x70) { + case DW_EH_PE_absptr: + // do nothing + break; + case DW_EH_PE_pcrel: + result += (uintptr_t)(*data); + break; + case DW_EH_PE_textrel: + case DW_EH_PE_datarel: + case DW_EH_PE_funcrel: + case DW_EH_PE_aligned: + default: + // not supported + compilerrt_abort(); + break; + } + + // then apply indirection + if (encoding & DW_EH_PE_indirect) { + result = *((const uintptr_t *)result); + } + + *data = p; + return result; } #if defined(__arm__) && !defined(__USING_SJLJ_EXCEPTIONS__) && \ @@ -153,99 +143,92 @@ static inline _Unwind_Reason_Code continueUnwind(struct _Unwind_Exception *exceptionObject, struct _Unwind_Context *context) { #if USING_ARM_EHABI - /* - * On ARM EHABI the personality routine is responsible for actually - * unwinding a single stack frame before returning (ARM EHABI Sec. 6.1). - */ - if (__gnu_unwind_frame(exceptionObject, context) != _URC_OK) - return _URC_FAILURE; + // On ARM EHABI the personality routine is responsible for actually + // unwinding a single stack frame before returning (ARM EHABI Sec. 6.1). + if (__gnu_unwind_frame(exceptionObject, context) != _URC_OK) + return _URC_FAILURE; #endif - return _URC_CONTINUE_UNWIND; + return _URC_CONTINUE_UNWIND; } -/* - * The C compiler makes references to __gcc_personality_v0 in - * the dwarf unwind information for translation units that use - * __attribute__((cleanup(xx))) on local variables. - * This personality routine is called by the system unwinder - * on each frame as the stack is unwound during a C++ exception - * throw through a C function compiled with -fexceptions. - */ +// The C compiler makes references to __gcc_personality_v0 in +// the dwarf unwind information for translation units that use +// __attribute__((cleanup(xx))) on local variables. +// This personality routine is called by the system unwinder +// on each frame as the stack is unwound during a C++ exception +// throw through a C function compiled with -fexceptions. #if __USING_SJLJ_EXCEPTIONS__ -/* the setjump-longjump based exceptions personality routine has a - * different name */ -COMPILER_RT_ABI _Unwind_Reason_Code -__gcc_personality_sj0(int version, _Unwind_Action actions, - uint64_t exceptionClass, struct _Unwind_Exception* exceptionObject, - struct _Unwind_Context *context) +// the setjump-longjump based exceptions personality routine has a +// different name +COMPILER_RT_ABI _Unwind_Reason_Code __gcc_personality_sj0( + int version, _Unwind_Action actions, uint64_t exceptionClass, + struct _Unwind_Exception *exceptionObject, struct _Unwind_Context *context) #elif USING_ARM_EHABI -/* The ARM EHABI personality routine has a different signature. */ +// The ARM EHABI personality routine has a different signature. COMPILER_RT_ABI _Unwind_Reason_Code __gcc_personality_v0( - _Unwind_State state, struct _Unwind_Exception *exceptionObject, - struct _Unwind_Context *context) + _Unwind_State state, struct _Unwind_Exception *exceptionObject, + struct _Unwind_Context *context) #else -COMPILER_RT_ABI _Unwind_Reason_Code -__gcc_personality_v0(int version, _Unwind_Action actions, - uint64_t exceptionClass, struct _Unwind_Exception* exceptionObject, - struct _Unwind_Context *context) +COMPILER_RT_ABI _Unwind_Reason_Code __gcc_personality_v0( + int version, _Unwind_Action actions, uint64_t exceptionClass, + struct _Unwind_Exception *exceptionObject, struct _Unwind_Context *context) #endif { - /* Since C does not have catch clauses, there is nothing to do during */ - /* phase 1 (the search phase). */ + // Since C does not have catch clauses, there is nothing to do during + // phase 1 (the search phase). #if USING_ARM_EHABI - /* After resuming from a cleanup we should also continue on to the next - * frame straight away. */ - if ((state & _US_ACTION_MASK) != _US_UNWIND_FRAME_STARTING) + // After resuming from a cleanup we should also continue on to the next + // frame straight away. + if ((state & _US_ACTION_MASK) != _US_UNWIND_FRAME_STARTING) #else - if ( actions & _UA_SEARCH_PHASE ) + if (actions & _UA_SEARCH_PHASE) #endif - return continueUnwind(exceptionObject, context); - - /* There is nothing to do if there is no LSDA for this frame. */ - const uint8_t* lsda = (uint8_t*)_Unwind_GetLanguageSpecificData(context); - if ( lsda == (uint8_t*) 0 ) - return continueUnwind(exceptionObject, context); + return continueUnwind(exceptionObject, context); - uintptr_t pc = (uintptr_t)_Unwind_GetIP(context)-1; - uintptr_t funcStart = (uintptr_t)_Unwind_GetRegionStart(context); - uintptr_t pcOffset = pc - funcStart; + // There is nothing to do if there is no LSDA for this frame. + const uint8_t *lsda = (uint8_t *)_Unwind_GetLanguageSpecificData(context); + if (lsda == (uint8_t *)0) + return continueUnwind(exceptionObject, context); - /* Parse LSDA header. */ - uint8_t lpStartEncoding = *lsda++; - if (lpStartEncoding != DW_EH_PE_omit) { - readEncodedPointer(&lsda, lpStartEncoding); - } - uint8_t ttypeEncoding = *lsda++; - if (ttypeEncoding != DW_EH_PE_omit) { - readULEB128(&lsda); - } - /* Walk call-site table looking for range that includes current PC. */ - uint8_t callSiteEncoding = *lsda++; - uint32_t callSiteTableLength = readULEB128(&lsda); - const uint8_t* callSiteTableStart = lsda; - const uint8_t* callSiteTableEnd = callSiteTableStart + callSiteTableLength; - const uint8_t* p=callSiteTableStart; - while (p < callSiteTableEnd) { - uintptr_t start = readEncodedPointer(&p, callSiteEncoding); - uintptr_t length = readEncodedPointer(&p, callSiteEncoding); - uintptr_t landingPad = readEncodedPointer(&p, callSiteEncoding); - readULEB128(&p); /* action value not used for C code */ - if ( landingPad == 0 ) - continue; /* no landing pad for this entry */ - if ( (start <= pcOffset) && (pcOffset < (start+length)) ) { - /* Found landing pad for the PC. - * Set Instruction Pointer to so we re-enter function - * at landing pad. The landing pad is created by the compiler - * to take two parameters in registers. - */ - _Unwind_SetGR(context, __builtin_eh_return_data_regno(0), - (uintptr_t)exceptionObject); - _Unwind_SetGR(context, __builtin_eh_return_data_regno(1), 0); - _Unwind_SetIP(context, (funcStart + landingPad)); - return _URC_INSTALL_CONTEXT; - } + uintptr_t pc = (uintptr_t)_Unwind_GetIP(context) - 1; + uintptr_t funcStart = (uintptr_t)_Unwind_GetRegionStart(context); + uintptr_t pcOffset = pc - funcStart; + + // Parse LSDA header. + uint8_t lpStartEncoding = *lsda++; + if (lpStartEncoding != DW_EH_PE_omit) { + readEncodedPointer(&lsda, lpStartEncoding); + } + uint8_t ttypeEncoding = *lsda++; + if (ttypeEncoding != DW_EH_PE_omit) { + readULEB128(&lsda); + } + // Walk call-site table looking for range that includes current PC. + uint8_t callSiteEncoding = *lsda++; + uint32_t callSiteTableLength = readULEB128(&lsda); + const uint8_t *callSiteTableStart = lsda; + const uint8_t *callSiteTableEnd = callSiteTableStart + callSiteTableLength; + const uint8_t *p = callSiteTableStart; + while (p < callSiteTableEnd) { + uintptr_t start = readEncodedPointer(&p, callSiteEncoding); + uintptr_t length = readEncodedPointer(&p, callSiteEncoding); + uintptr_t landingPad = readEncodedPointer(&p, callSiteEncoding); + readULEB128(&p); // action value not used for C code + if (landingPad == 0) + continue; // no landing pad for this entry + if ((start <= pcOffset) && (pcOffset < (start + length))) { + // Found landing pad for the PC. + // Set Instruction Pointer to so we re-enter function + // at landing pad. The landing pad is created by the compiler + // to take two parameters in registers. + _Unwind_SetGR(context, __builtin_eh_return_data_regno(0), + (uintptr_t)exceptionObject); + _Unwind_SetGR(context, __builtin_eh_return_data_regno(1), 0); + _Unwind_SetIP(context, (funcStart + landingPad)); + return _URC_INSTALL_CONTEXT; } + } - /* No landing pad found, continue unwinding. */ - return continueUnwind(exceptionObject, context); + // No landing pad found, continue unwinding. + return continueUnwind(exceptionObject, context); } diff --git a/lib/builtins/hexagon/common_entry_exit_abi1.S b/lib/builtins/hexagon/common_entry_exit_abi1.S index d5479d2a52fa..23fed01c6eef 100644 --- a/lib/builtins/hexagon/common_entry_exit_abi1.S +++ b/lib/builtins/hexagon/common_entry_exit_abi1.S @@ -1,14 +1,13 @@ //===----------------------Hexagon builtin routine ------------------------===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// -/* Functions that implement common sequences in function prologues and epilogues - used to save code size */ +// Functions that implement common sequences in function prologues and epilogues +// used to save code size .macro FUNCTION_BEGIN name .text @@ -33,16 +32,16 @@ -/* Save r25:24 at fp+#-8 and r27:26 at fp+#-16. */ +// Save r25:24 at fp+#-8 and r27:26 at fp+#-16. -/* The compiler knows that the __save_* functions clobber LR. No other - registers should be used without informing the compiler. */ +// The compiler knows that the __save_* functions clobber LR. No other +// registers should be used without informing the compiler. -/* Since we can only issue one store per packet, we don't hurt performance by - simply jumping to the right point in this sequence of stores. */ +// Since we can only issue one store per packet, we don't hurt performance by +// simply jumping to the right point in this sequence of stores. FUNCTION_BEGIN __save_r24_through_r27 memd(fp+#-16) = r27:26 @@ -56,10 +55,10 @@ FUNCTION_END __save_r24_through_r25 -/* For each of the *_before_tailcall functions, jumpr lr is executed in parallel - with deallocframe. That way, the return gets the old value of lr, which is - where these functions need to return, and at the same time, lr gets the value - it needs going into the tail call. */ +// For each of the *_before_tailcall functions, jumpr lr is executed in parallel +// with deallocframe. That way, the return gets the old value of lr, which is +// where these functions need to return, and at the same time, lr gets the value +// it needs going into the tail call. FUNCTION_BEGIN __restore_r24_through_r27_and_deallocframe_before_tailcall r27:26 = memd(fp+#-16) @@ -74,8 +73,8 @@ FUNCTION_END __restore_r24_through_r25_and_deallocframe_before_tailcall -/* Here we use the extra load bandwidth to restore LR early, allowing the return - to occur in parallel with the deallocframe. */ +// Here we use the extra load bandwidth to restore LR early, allowing the return +// to occur in parallel with the deallocframe. FUNCTION_BEGIN __restore_r24_through_r27_and_deallocframe { @@ -92,7 +91,7 @@ FUNCTION_END __restore_r24_through_r27_and_deallocframe -/* Here the load bandwidth is maximized. */ +// Here the load bandwidth is maximized. FUNCTION_BEGIN __restore_r24_through_r25_and_deallocframe { diff --git a/lib/builtins/hexagon/common_entry_exit_abi2.S b/lib/builtins/hexagon/common_entry_exit_abi2.S index 6f470343db49..3b85aea2f600 100644 --- a/lib/builtins/hexagon/common_entry_exit_abi2.S +++ b/lib/builtins/hexagon/common_entry_exit_abi2.S @@ -1,14 +1,13 @@ //===----------------------Hexagon builtin routine ------------------------===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// -/* Functions that implement common sequences in function prologues and epilogues - used to save code size */ +// Functions that implement common sequences in function prologues and epilogues +// used to save code size .macro FUNCTION_BEGIN name .p2align 2 @@ -33,10 +32,10 @@ -/* Save r17:16 at fp+#-8, r19:18 at fp+#-16, r21:20 at fp+#-24, r23:22 at - fp+#-32, r25:24 at fp+#-40, and r27:26 at fp+#-48. - The compiler knows that the __save_* functions clobber LR. No other - registers should be used without informing the compiler. */ +// Save r17:16 at fp+#-8, r19:18 at fp+#-16, r21:20 at fp+#-24, r23:22 at +// fp+#-32, r25:24 at fp+#-40, and r27:26 at fp+#-48. +// The compiler knows that the __save_* functions clobber LR. No other +// registers should be used without informing the compiler. FUNCTION_BEGIN __save_r16_through_r27 { @@ -107,10 +106,10 @@ FUNCTION_BEGIN __save_r16_through_r17 } FUNCTION_END __save_r16_through_r17 -/* For each of the *_before_tailcall functions, jumpr lr is executed in parallel - with deallocframe. That way, the return gets the old value of lr, which is - where these functions need to return, and at the same time, lr gets the value - it needs going into the tail call. */ +// For each of the *_before_tailcall functions, jumpr lr is executed in parallel +// with deallocframe. That way, the return gets the old value of lr, which is +// where these functions need to return, and at the same time, lr gets the value +// it needs going into the tail call. FUNCTION_BEGIN __restore_r16_through_r27_and_deallocframe_before_tailcall diff --git a/lib/builtins/hexagon/common_entry_exit_legacy.S b/lib/builtins/hexagon/common_entry_exit_legacy.S index 3258f15a3267..8a6044573236 100644 --- a/lib/builtins/hexagon/common_entry_exit_legacy.S +++ b/lib/builtins/hexagon/common_entry_exit_legacy.S @@ -1,15 +1,14 @@ //===----------------------Hexagon builtin routine ------------------------===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// -/* Functions that implement common sequences in function prologues and epilogues - used to save code size */ +// Functions that implement common sequences in function prologues and epilogues +// used to save code size .macro FUNCTION_BEGIN name .text @@ -34,17 +33,17 @@ -/* Save r27:26 at fp+#-8, r25:24 at fp+#-16, r23:22 at fp+#-24, r21:20 at - fp+#-32, r19:18 at fp+#-40, and r17:16 at fp+#-48. */ +// Save r27:26 at fp+#-8, r25:24 at fp+#-16, r23:22 at fp+#-24, r21:20 at +// fp+#-32, r19:18 at fp+#-40, and r17:16 at fp+#-48. -/* The compiler knows that the __save_* functions clobber LR. No other - registers should be used without informing the compiler. */ +// The compiler knows that the __save_* functions clobber LR. No other +// registers should be used without informing the compiler. -/* Since we can only issue one store per packet, we don't hurt performance by - simply jumping to the right point in this sequence of stores. */ +// Since we can only issue one store per packet, we don't hurt performance by +// simply jumping to the right point in this sequence of stores. FUNCTION_BEGIN __save_r27_through_r16 memd(fp+#-48) = r17:16 @@ -65,10 +64,10 @@ FUNCTION_END __save_r27_through_r24 -/* For each of the *_before_sibcall functions, jumpr lr is executed in parallel - with deallocframe. That way, the return gets the old value of lr, which is - where these functions need to return, and at the same time, lr gets the value - it needs going into the sibcall. */ +// For each of the *_before_sibcall functions, jumpr lr is executed in parallel +// with deallocframe. That way, the return gets the old value of lr, which is +// where these functions need to return, and at the same time, lr gets the value +// it needs going into the sibcall. FUNCTION_BEGIN __restore_r27_through_r20_and_deallocframe_before_sibcall { @@ -108,8 +107,8 @@ FUNCTION_END __restore_r27_through_r26_and_deallocframe_before_sibcall -/* Here we use the extra load bandwidth to restore LR early, allowing the return - to occur in parallel with the deallocframe. */ +// Here we use the extra load bandwidth to restore LR early, allowing the return +// to occur in parallel with the deallocframe. FUNCTION_BEGIN __restore_r27_through_r16_and_deallocframe { @@ -136,7 +135,7 @@ FUNCTION_END __restore_r27_through_r24_and_deallocframe -/* Here the load bandwidth is maximized for all three functions. */ +// Here the load bandwidth is maximized for all three functions. FUNCTION_BEGIN __restore_r27_through_r18_and_deallocframe { diff --git a/lib/builtins/hexagon/dfaddsub.S b/lib/builtins/hexagon/dfaddsub.S index 4173f86a4f54..1b0d34550ff3 100644 --- a/lib/builtins/hexagon/dfaddsub.S +++ b/lib/builtins/hexagon/dfaddsub.S @@ -1,13 +1,12 @@ //===----------------------Hexagon builtin routine ------------------------===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// -/* Double Precision Multiply */ +// Double Precision Multiply #define A r1:0 #define AH r1 @@ -179,18 +178,17 @@ __hexagon_subdf3: .Ladd_ovf_unf: // Overflow or Denormal is possible // Good news: Underflow flag is not possible! - /* - * ATMP has 2's complement value - * - * EXPA has A's exponent, EXPB has EXPA-BIAS-60 - * - * Convert, extract exponent, add adjustment. - * If > 2046, overflow - * If <= 0, denormal - * - * Note that we've not done our zero check yet, so do that too - * - */ + + // ATMP has 2's complement value + // + // EXPA has A's exponent, EXPB has EXPA-BIAS-60 + // + // Convert, extract exponent, add adjustment. + // If > 2046, overflow + // If <= 0, denormal + // + // Note that we've not done our zero check yet, so do that too + { A = convert_d2df(ATMP) p0 = cmp.eq(ATMPH,#0) diff --git a/lib/builtins/hexagon/dfdiv.S b/lib/builtins/hexagon/dfdiv.S index 0c5dbe272c89..202965ec4789 100644 --- a/lib/builtins/hexagon/dfdiv.S +++ b/lib/builtins/hexagon/dfdiv.S @@ -1,13 +1,12 @@ //===----------------------Hexagon builtin routine ------------------------===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// -/* Double Precision Divide */ +// Double Precision Divide #define A r1:0 #define AH r1 @@ -237,10 +236,10 @@ __hexagon_divdf3: P_TMP = cmp.gt(EXPA,#0) if (P_TMP.new) jump:nt .Lpossible_unf // round up to normal possible... } - /* Underflow */ - /* We know what the infinite range exponent should be (EXPA) */ - /* Q is 2's complement, PROD is abs(Q) */ - /* Normalize Q, shift right, add a high bit, convert, change exponent */ + // Underflow + // We know what the infinite range exponent should be (EXPA) + // Q is 2's complement, PROD is abs(Q) + // Normalize Q, shift right, add a high bit, convert, change exponent #define FUDGE1 7 // how much to shift right #define FUDGE2 4 // how many guard/round to keep at lsbs @@ -287,8 +286,8 @@ __hexagon_divdf3: .Lpossible_unf: - /* If upper parts of Q were all F's, but abs(A) == 0x00100000_00000000, we rounded up to min_normal */ - /* The answer is correct, but we need to raise Underflow */ + // If upper parts of Q were all F's, but abs(A) == 0x00100000_00000000, we rounded up to min_normal + // The answer is correct, but we need to raise Underflow { B = extractu(A,#63,#0) TMPPAIR = combine(##0x00100000,#0) // min normal @@ -321,9 +320,9 @@ __hexagon_divdf3: } .Ldiv_ovf: - /* - * Raise Overflow, and choose the correct overflow value (saturated normal or infinity) - */ + + // Raise Overflow, and choose the correct overflow value (saturated normal or infinity) + { TMP = USR B = combine(##0x7fefffff,#-1) @@ -389,8 +388,8 @@ __hexagon_divdf3: if (!P_ZERO) jump .Ldiv_zero_result if (!P_INF) jump .Ldiv_inf_result } - /* Now we've narrowed it down to (de)normal / (de)normal */ - /* Set up A/EXPA B/EXPB and go back */ + // Now we've narrowed it down to (de)normal / (de)normal + // Set up A/EXPA B/EXPB and go back #undef P_ZERO #undef P_INF #define P_TMP2 p1 diff --git a/lib/builtins/hexagon/dffma.S b/lib/builtins/hexagon/dffma.S index 97b885a3bf27..c201d3d8be5e 100644 --- a/lib/builtins/hexagon/dffma.S +++ b/lib/builtins/hexagon/dffma.S @@ -1,16 +1,15 @@ //===----------------------Hexagon builtin routine ------------------------===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// #define Q6_ALIAS(TAG) .global __qdsp_##TAG ; .set __qdsp_##TAG, __hexagon_##TAG #define END(TAG) .size TAG,.-TAG -/* Double Precision Multiply */ +// Double Precision Multiply #define A r1:0 @@ -76,33 +75,29 @@ #define SR_ROUND_OFF 22 #endif - /* - * First, classify for normal values, and abort if abnormal - * - * Next, unpack mantissa into 0x1000_0000_0000_0000 + mant<<8 - * - * Since we know that the 2 MSBs of the H registers is zero, we should never carry - * the partial products that involve the H registers - * - * Try to buy X slots, at the expense of latency if needed - * - * We will have PP_HH with the upper bits of the product, PP_LL with the lower - * PP_HH can have a maximum of 0x03FF_FFFF_FFFF_FFFF or thereabouts - * PP_HH can have a minimum of 0x0100_0000_0000_0000 - * - * 0x0100_0000_0000_0000 has EXP of EXPA+EXPB-BIAS - * - * We need to align CTMP. - * If CTMP >> PP, convert PP to 64 bit with sticky, align CTMP, and follow normal add - * If CTMP << PP align CTMP and add 128 bits. Then compute sticky - * If CTMP ~= PP, align CTMP and add 128 bits. May have massive cancellation. - * - * Convert partial product and CTMP to 2's complement prior to addition - * - * After we add, we need to normalize into upper 64 bits, then compute sticky. - * - * - */ + // First, classify for normal values, and abort if abnormal + // + // Next, unpack mantissa into 0x1000_0000_0000_0000 + mant<<8 + // + // Since we know that the 2 MSBs of the H registers is zero, we should never carry + // the partial products that involve the H registers + // + // Try to buy X slots, at the expense of latency if needed + // + // We will have PP_HH with the upper bits of the product, PP_LL with the lower + // PP_HH can have a maximum of 0x03FF_FFFF_FFFF_FFFF or thereabouts + // PP_HH can have a minimum of 0x0100_0000_0000_0000 + // + // 0x0100_0000_0000_0000 has EXP of EXPA+EXPB-BIAS + // + // We need to align CTMP. + // If CTMP >> PP, convert PP to 64 bit with sticky, align CTMP, and follow normal add + // If CTMP << PP align CTMP and add 128 bits. Then compute sticky + // If CTMP ~= PP, align CTMP and add 128 bits. May have massive cancellation. + // + // Convert partial product and CTMP to 2's complement prior to addition + // + // After we add, we need to normalize into upper 64 bits, then compute sticky. .text .global __hexagon_fmadf4 @@ -182,14 +177,12 @@ fma: #define EXPCA r19:18 EXPC = extractu(CH,#EXPBITS,#HI_MANTBITS) } - /* PP_HH:PP_LL now has product */ - /* CTMP is negated */ - /* EXPA,B,C are extracted */ - /* - * We need to negate PP - * Since we will be adding with carry later, if we need to negate, - * just invert all bits now, which we can do conditionally and in parallel - */ + // PP_HH:PP_LL now has product + // CTMP is negated + // EXPA,B,C are extracted + // We need to negate PP + // Since we will be adding with carry later, if we need to negate, + // just invert all bits now, which we can do conditionally and in parallel #define PP_HH_TMP r15:14 #define PP_LL_TMP r7:6 { @@ -274,18 +267,16 @@ fma: PP_HH = add(CTMP,PP_HH,P_CARRY):carry TMP = #62 } - /* - * PP_HH:PP_LL now holds the sum - * We may need to normalize left, up to ??? bits. - * - * I think that if we have massive cancellation, the range we normalize by - * is still limited - */ + // PP_HH:PP_LL now holds the sum + // We may need to normalize left, up to ??? bits. + // + // I think that if we have massive cancellation, the range we normalize by + // is still limited { LEFTSHIFT = add(clb(PP_HH),#-2) if (!cmp.eq(LEFTSHIFT.new,TMP)) jump:t 1f // all sign bits? } - /* We had all sign bits, shift left by 62. */ + // We had all sign bits, shift left by 62. { CTMP = extractu(PP_LL,#62,#2) PP_LL = asl(PP_LL,#62) @@ -330,7 +321,7 @@ fma: if (!P_TMP) dealloc_return // not zero, return } .Ladd_yields_zero: - /* We had full cancellation. Return +/- zero (-0 when round-down) */ + // We had full cancellation. Return +/- zero (-0 when round-down) { TMP = USR A = #0 @@ -408,9 +399,9 @@ fma: EXPA = sub(#1+5,TMP) // Amount to right shift to denormalize p3 = cmp.gt(CTMPH,#-1) } - /* Underflow */ - /* We know that the infinte range exponent should be EXPA */ - /* CTMP is 2's complement, ATMP is abs(CTMP) */ + // Underflow + // We know that the infinte range exponent should be EXPA + // CTMP is 2's complement, ATMP is abs(CTMP) { EXPA = add(EXPA,EXPB) // how much to shift back right ATMP = asl(ATMP,EXPB) // shift left @@ -593,7 +584,7 @@ fma: p1 = dfclass(C,#0x08) if (p1.new) jump:nt .Lfma_inf_plus_inf } - /* A*B is +/- inf, C is finite. Return A */ + // A*B is +/- inf, C is finite. Return A { jumpr r31 } @@ -649,7 +640,7 @@ fma: if (!p0) A = C // If C is not zero, return C if (!p0) jumpr r31 } - /* B has correctly signed zero, C is also zero */ + // B has correctly signed zero, C is also zero .Lzero_plus_zero: { p0 = cmp.eq(B,C) // yes, scalar equals. +0++0 or -0+-0 @@ -674,8 +665,8 @@ fma: #define CTMP r11:10 .falign .Lfma_abnormal_c: - /* We know that AB is normal * normal */ - /* C is not normal: zero, subnormal, inf, or NaN. */ + // We know that AB is normal * normal + // C is not normal: zero, subnormal, inf, or NaN. { p0 = dfclass(C,#0x10) // is C NaN? if (p0.new) jump:nt .Lnan diff --git a/lib/builtins/hexagon/dfminmax.S b/lib/builtins/hexagon/dfminmax.S index 41122911f183..44f031ba11ba 100644 --- a/lib/builtins/hexagon/dfminmax.S +++ b/lib/builtins/hexagon/dfminmax.S @@ -1,9 +1,8 @@ //===----------------------Hexagon builtin routine ------------------------===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// @@ -15,17 +14,14 @@ #define Q6_ALIAS(TAG) .global __qdsp_##TAG ; .set __qdsp_##TAG, __hexagon_##TAG #define END(TAG) .size TAG,.-TAG -/* - * Min and Max return A if B is NaN, or B if A is NaN - * Otherwise, they return the smaller or bigger value - * - * If values are equal, we want to favor -0.0 for min and +0.0 for max. - */ +// Min and Max return A if B is NaN, or B if A is NaN +// Otherwise, they return the smaller or bigger value +// +// If values are equal, we want to favor -0.0 for min and +0.0 for max. + +// Compares always return false for NaN +// if (isnan(A)) A = B; if (A > B) A = B will only trigger at most one of those options. -/* - * Compares always return false for NaN - * if (isnan(A)) A = B; if (A > B) A = B will only trigger at most one of those options. - */ .text .global __hexagon_mindf3 .global __hexagon_maxdf3 @@ -51,7 +47,7 @@ fmin: p2 = dfcmp.eq(A,B) // if A == B if (!p2.new) jumpr:t r31 } - /* A == B, return A|B to select -0.0 over 0.0 */ + // A == B, return A|B to select -0.0 over 0.0 { A = or(ATMP,B) jumpr r31 @@ -71,7 +67,7 @@ fmax: p2 = dfcmp.eq(A,B) if (!p2.new) jumpr:t r31 } - /* A == B, return A&B to select 0.0 over -0.0 */ + // A == B, return A&B to select 0.0 over -0.0 { A = and(ATMP,B) jumpr r31 diff --git a/lib/builtins/hexagon/dfmul.S b/lib/builtins/hexagon/dfmul.S index fde6d77bdcff..e6f62c3515f4 100644 --- a/lib/builtins/hexagon/dfmul.S +++ b/lib/builtins/hexagon/dfmul.S @@ -1,13 +1,12 @@ //===----------------------Hexagon builtin routine ------------------------===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// -/* Double Precision Multiply */ +// Double Precision Multiply #define A r1:0 #define AH r1 #define AL r0 @@ -47,8 +46,8 @@ #define BIAS 1024 #define MANTISSA_TO_INT_BIAS 52 -/* Some constant to adjust normalization amount in error code */ -/* Amount to right shift the partial product to get to a denorm */ +// Some constant to adjust normalization amount in error code +// Amount to right shift the partial product to get to a denorm #define FUDGE 5 #define Q6_ALIAS(TAG) .global __qdsp_##TAG ; .set __qdsp_##TAG, __hexagon_##TAG @@ -80,10 +79,10 @@ __hexagon_muldf3: PP_ODD = mpyu(BTMPL,ATMPH) BTMP = insert(ONE,#2,#62) } - /* since we know that the MSB of the H registers is zero, we should never carry */ - /* H <= 2^31-1. L <= 2^32-1. Therefore, HL <= 2^63-2^32-2^31+1 */ - /* Adding 2 HLs, we get 2^64-3*2^32+2 maximum. */ - /* Therefore, we can add 3 2^32-1 values safely without carry. We only need one. */ + // since we know that the MSB of the H registers is zero, we should never carry + // H <= 2^31-1. L <= 2^32-1. Therefore, HL <= 2^63-2^32-2^31+1 + // Adding 2 HLs, we get 2^64-3*2^32+2 maximum. + // Therefore, we can add 3 2^32-1 values safely without carry. We only need one. { PP_LL = mpyu(ATMPL,BTMPL) PP_ODD += mpyu(ATMPL,BTMPH) @@ -99,10 +98,10 @@ __hexagon_muldf3: p1 = cmp.eq(PP_LL_L,#0) // 64 lsb's 0? p1 = cmp.eq(PP_ODD_L,#0) // 64 lsb's 0? } - /* - * PP_HH can have a maximum of 0x3FFF_FFFF_FFFF_FFFF or thereabouts - * PP_HH can have a minimum of 0x1000_0000_0000_0000 or so - */ + + // PP_HH can have a maximum of 0x3FFF_FFFF_FFFF_FFFF or thereabouts + // PP_HH can have a minimum of 0x1000_0000_0000_0000 or so + #undef PP_ODD #undef PP_ODD_H #undef PP_ODD_L @@ -137,15 +136,15 @@ __hexagon_muldf3: .falign .Lpossible_unf: - /* We end up with a positive exponent */ - /* But we may have rounded up to an exponent of 1. */ - /* If the exponent is 1, if we rounded up to it - * we need to also raise underflow - * Fortunately, this is pretty easy to detect, we must have +/- 0x0010_0000_0000_0000 - * And the PP should also have more than one bit set - */ - /* Note: ATMP should have abs(PP_HH) */ - /* Note: BTMPL should have 0x7FEFFFFF */ + // We end up with a positive exponent + // But we may have rounded up to an exponent of 1. + // If the exponent is 1, if we rounded up to it + // we need to also raise underflow + // Fortunately, this is pretty easy to detect, we must have +/- 0x0010_0000_0000_0000 + // And the PP should also have more than one bit set + // + // Note: ATMP should have abs(PP_HH) + // Note: BTMPL should have 0x7FEFFFFF { p0 = cmp.eq(AL,#0) p0 = bitsclr(AH,BTMPL) @@ -194,29 +193,25 @@ __hexagon_muldf3: BTMPH = sub(EXP0,BTMPH) TMP = #63 // max amount to shift } - /* Underflow */ - /* - * PP_HH has the partial product with sticky LSB. - * PP_HH can have a maximum of 0x3FFF_FFFF_FFFF_FFFF or thereabouts - * PP_HH can have a minimum of 0x1000_0000_0000_0000 or so - * The exponent of PP_HH is in EXP1, which is non-positive (0 or negative) - * That's the exponent that happens after the normalization - * - * EXP0 has the exponent that, when added to the normalized value, is out of range. - * - * Strategy: - * - * * Shift down bits, with sticky bit, such that the bits are aligned according - * to the LZ count and appropriate exponent, but not all the way to mantissa - * field, keep around the last few bits. - * * Put a 1 near the MSB - * * Check the LSBs for inexact; if inexact also set underflow - * * Convert [u]d2df -- will correctly round according to rounding mode - * * Replace exponent field with zero - * - * - */ - + // Underflow + // + // PP_HH has the partial product with sticky LSB. + // PP_HH can have a maximum of 0x3FFF_FFFF_FFFF_FFFF or thereabouts + // PP_HH can have a minimum of 0x1000_0000_0000_0000 or so + // The exponent of PP_HH is in EXP1, which is non-positive (0 or negative) + // That's the exponent that happens after the normalization + // + // EXP0 has the exponent that, when added to the normalized value, is out of range. + // + // Strategy: + // + // * Shift down bits, with sticky bit, such that the bits are aligned according + // to the LZ count and appropriate exponent, but not all the way to mantissa + // field, keep around the last few bits. + // * Put a 1 near the MSB + // * Check the LSBs for inexact; if inexact also set underflow + // * Convert [u]d2df -- will correctly round according to rounding mode + // * Replace exponent field with zero { BTMPL = #0 // offset for extract diff --git a/lib/builtins/hexagon/dfsqrt.S b/lib/builtins/hexagon/dfsqrt.S index 027d9e1fde43..f1435e868319 100644 --- a/lib/builtins/hexagon/dfsqrt.S +++ b/lib/builtins/hexagon/dfsqrt.S @@ -1,13 +1,12 @@ //===----------------------Hexagon builtin routine ------------------------===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// -/* Double Precision square root */ +// Double Precision square root #define EXP r28 @@ -169,9 +168,9 @@ __hexagon_sqrt: #define P_CARRY1 p2 #define P_CARRY2 p3 - /* Iteration 0 */ - /* Maybe we can save a cycle by starting with ERROR=asl(fracrad), then as we multiply */ - /* We can shift and subtract instead of shift and add? */ + // Iteration 0 + // Maybe we can save a cycle by starting with ERROR=asl(fracrad), then as we multiply + // We can shift and subtract instead of shift and add? { ERROR = asl(FRACRAD,#15) PROD = mpyu(ROOTHI,ROOTHI) @@ -194,7 +193,7 @@ __hexagon_sqrt: SHIFTAMT = add(SHIFTAMT,#16) ERROR = asl(FRACRAD,#31) // for next iter } - /* Iteration 1 */ + // Iteration 1 { PROD = mpyu(ROOTHI,ROOTHI) ERROR -= mpyu(ROOTHI,ROOTLO) // amount is 31, no shift needed @@ -214,7 +213,7 @@ __hexagon_sqrt: SHIFTAMT = add(SHIFTAMT,#16) ERROR = asl(FRACRAD,#47) // for next iter } - /* Iteration 2 */ + // Iteration 2 { PROD = mpyu(ROOTHI,ROOTHI) } @@ -245,7 +244,7 @@ __hexagon_sqrt: #undef RECIPEST #undef SHIFTAMT #define TWOROOT_LO r9:8 - /* Adjust Root */ + // Adjust Root { HL = mpyu(ROOTHI,ROOTLO) LL = mpyu(ROOTLO,ROOTLO) diff --git a/lib/builtins/hexagon/divdi3.S b/lib/builtins/hexagon/divdi3.S index 49ee8104f305..770601a470f7 100644 --- a/lib/builtins/hexagon/divdi3.S +++ b/lib/builtins/hexagon/divdi3.S @@ -1,9 +1,8 @@ //===----------------------Hexagon builtin routine ------------------------===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// diff --git a/lib/builtins/hexagon/divsi3.S b/lib/builtins/hexagon/divsi3.S index 8e159baa192f..5f406524e88a 100644 --- a/lib/builtins/hexagon/divsi3.S +++ b/lib/builtins/hexagon/divsi3.S @@ -1,9 +1,8 @@ //===----------------------Hexagon builtin routine ------------------------===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// diff --git a/lib/builtins/hexagon/fabs_opt.S b/lib/builtins/hexagon/fabs_opt.S index b09b00734d98..6bf9b84b3d20 100644 --- a/lib/builtins/hexagon/fabs_opt.S +++ b/lib/builtins/hexagon/fabs_opt.S @@ -1,9 +1,8 @@ //===----------------------Hexagon builtin routine ------------------------===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// diff --git a/lib/builtins/hexagon/fastmath2_dlib_asm.S b/lib/builtins/hexagon/fastmath2_dlib_asm.S index 9286df06c26d..574a04432f2c 100644 --- a/lib/builtins/hexagon/fastmath2_dlib_asm.S +++ b/lib/builtins/hexagon/fastmath2_dlib_asm.S @@ -1,9 +1,8 @@ //===----------------------Hexagon builtin routine ------------------------===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// /* ==================================================================== */ diff --git a/lib/builtins/hexagon/fastmath2_ldlib_asm.S b/lib/builtins/hexagon/fastmath2_ldlib_asm.S index 4192555351a6..cf623f94c849 100644 --- a/lib/builtins/hexagon/fastmath2_ldlib_asm.S +++ b/lib/builtins/hexagon/fastmath2_ldlib_asm.S @@ -1,9 +1,8 @@ //===----------------------Hexagon builtin routine ------------------------===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// /* ==================================================================== * diff --git a/lib/builtins/hexagon/fastmath_dlib_asm.S b/lib/builtins/hexagon/fastmath_dlib_asm.S index 215936b783ca..3e59526c1e23 100644 --- a/lib/builtins/hexagon/fastmath_dlib_asm.S +++ b/lib/builtins/hexagon/fastmath_dlib_asm.S @@ -1,9 +1,8 @@ //===----------------------Hexagon builtin routine ------------------------===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// /* ==================================================================== */ diff --git a/lib/builtins/hexagon/fma_opt.S b/lib/builtins/hexagon/fma_opt.S index 12378f0da04e..7f566adffd6a 100644 --- a/lib/builtins/hexagon/fma_opt.S +++ b/lib/builtins/hexagon/fma_opt.S @@ -1,9 +1,8 @@ //===----------------------Hexagon builtin routine ------------------------===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// diff --git a/lib/builtins/hexagon/fmax_opt.S b/lib/builtins/hexagon/fmax_opt.S index f3a218c9769b..81d711dff8d2 100644 --- a/lib/builtins/hexagon/fmax_opt.S +++ b/lib/builtins/hexagon/fmax_opt.S @@ -1,9 +1,8 @@ //===----------------------Hexagon builtin routine ------------------------===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// diff --git a/lib/builtins/hexagon/fmin_opt.S b/lib/builtins/hexagon/fmin_opt.S index ef9b0ff854a2..d043f1d7a698 100644 --- a/lib/builtins/hexagon/fmin_opt.S +++ b/lib/builtins/hexagon/fmin_opt.S @@ -1,9 +1,8 @@ //===----------------------Hexagon builtin routine ------------------------===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// diff --git a/lib/builtins/hexagon/memcpy_forward_vp4cp4n2.S b/lib/builtins/hexagon/memcpy_forward_vp4cp4n2.S index fbe09086cd33..10b81f653343 100644 --- a/lib/builtins/hexagon/memcpy_forward_vp4cp4n2.S +++ b/lib/builtins/hexagon/memcpy_forward_vp4cp4n2.S @@ -1,9 +1,8 @@ //===----------------------Hexagon builtin routine ------------------------===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// // diff --git a/lib/builtins/hexagon/memcpy_likely_aligned.S b/lib/builtins/hexagon/memcpy_likely_aligned.S index bbc85c22db08..492298f10326 100644 --- a/lib/builtins/hexagon/memcpy_likely_aligned.S +++ b/lib/builtins/hexagon/memcpy_likely_aligned.S @@ -1,9 +1,8 @@ //===------------------------- memcopy routines ---------------------------===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// diff --git a/lib/builtins/hexagon/moddi3.S b/lib/builtins/hexagon/moddi3.S index 12a0595fe465..d4246b61bdc3 100644 --- a/lib/builtins/hexagon/moddi3.S +++ b/lib/builtins/hexagon/moddi3.S @@ -1,9 +1,8 @@ //===----------------------Hexagon builtin routine ------------------------===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// diff --git a/lib/builtins/hexagon/modsi3.S b/lib/builtins/hexagon/modsi3.S index 5afda9e2978b..4015d5e06891 100644 --- a/lib/builtins/hexagon/modsi3.S +++ b/lib/builtins/hexagon/modsi3.S @@ -1,9 +1,8 @@ //===----------------------Hexagon builtin routine ------------------------===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// diff --git a/lib/builtins/hexagon/sfdiv_opt.S b/lib/builtins/hexagon/sfdiv_opt.S index 6bdd4808c2b8..7c9ae14b718f 100644 --- a/lib/builtins/hexagon/sfdiv_opt.S +++ b/lib/builtins/hexagon/sfdiv_opt.S @@ -1,9 +1,8 @@ //===----------------------Hexagon builtin routine ------------------------===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// diff --git a/lib/builtins/hexagon/sfsqrt_opt.S b/lib/builtins/hexagon/sfsqrt_opt.S index 7f619002774f..532df9a06d2d 100644 --- a/lib/builtins/hexagon/sfsqrt_opt.S +++ b/lib/builtins/hexagon/sfsqrt_opt.S @@ -1,9 +1,8 @@ //===----------------------Hexagon builtin routine ------------------------===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// diff --git a/lib/builtins/hexagon/udivdi3.S b/lib/builtins/hexagon/udivdi3.S index 1ca326b75208..23f931d4f587 100644 --- a/lib/builtins/hexagon/udivdi3.S +++ b/lib/builtins/hexagon/udivdi3.S @@ -1,9 +1,8 @@ //===----------------------Hexagon builtin routine ------------------------===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// diff --git a/lib/builtins/hexagon/udivmoddi4.S b/lib/builtins/hexagon/udivmoddi4.S index deb5aae0924d..6dbfc59bd650 100644 --- a/lib/builtins/hexagon/udivmoddi4.S +++ b/lib/builtins/hexagon/udivmoddi4.S @@ -1,9 +1,8 @@ //===----------------------Hexagon builtin routine ------------------------===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// diff --git a/lib/builtins/hexagon/udivmodsi4.S b/lib/builtins/hexagon/udivmodsi4.S index 25bbe7cd5925..9e231212d6c8 100644 --- a/lib/builtins/hexagon/udivmodsi4.S +++ b/lib/builtins/hexagon/udivmodsi4.S @@ -1,9 +1,8 @@ //===----------------------Hexagon builtin routine ------------------------===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// diff --git a/lib/builtins/hexagon/udivsi3.S b/lib/builtins/hexagon/udivsi3.S index 54f0aa409f93..d68599a8e04d 100644 --- a/lib/builtins/hexagon/udivsi3.S +++ b/lib/builtins/hexagon/udivsi3.S @@ -1,9 +1,8 @@ //===----------------------Hexagon builtin routine ------------------------===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// diff --git a/lib/builtins/hexagon/umoddi3.S b/lib/builtins/hexagon/umoddi3.S index f091521414af..646ca128dd96 100644 --- a/lib/builtins/hexagon/umoddi3.S +++ b/lib/builtins/hexagon/umoddi3.S @@ -1,9 +1,8 @@ //===----------------------Hexagon builtin routine ------------------------===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// diff --git a/lib/builtins/hexagon/umodsi3.S b/lib/builtins/hexagon/umodsi3.S index a8270c2030d5..a92394486209 100644 --- a/lib/builtins/hexagon/umodsi3.S +++ b/lib/builtins/hexagon/umodsi3.S @@ -1,9 +1,8 @@ //===----------------------Hexagon builtin routine ------------------------===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// diff --git a/lib/builtins/i386/ashldi3.S b/lib/builtins/i386/ashldi3.S index 6f05dcf74443..7ba912692f0d 100644 --- a/lib/builtins/i386/ashldi3.S +++ b/lib/builtins/i386/ashldi3.S @@ -1,5 +1,6 @@ -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception #include "../assembly.h" diff --git a/lib/builtins/i386/ashrdi3.S b/lib/builtins/i386/ashrdi3.S index 206369f360aa..3cca4782aea3 100644 --- a/lib/builtins/i386/ashrdi3.S +++ b/lib/builtins/i386/ashrdi3.S @@ -1,5 +1,6 @@ -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception #include "../assembly.h" @@ -22,10 +23,10 @@ DEFINE_COMPILERRT_FUNCTION(__ashrdi3) #endif psrlq %xmm2, %xmm0 // unsigned shift input by count - + testl %eax, %eax // check the sign-bit of the input jns 1f // early out for positive inputs - + // If the input is negative, we need to construct the shifted sign bit // to or into the result, as xmm does not have a signed right shift. pcmpeqb %xmm1, %xmm1 // -1ULL @@ -35,7 +36,7 @@ DEFINE_COMPILERRT_FUNCTION(__ashrdi3) psubq %xmm1, %xmm2 // 64 - count psllq %xmm2, %xmm1 // -1 << (64 - count) = leading sign bits por %xmm1, %xmm0 - + // Move the result back to the general purpose registers and return 1: movd %xmm0, %eax psrlq $32, %xmm0 @@ -51,14 +52,14 @@ DEFINE_COMPILERRT_FUNCTION(__ashrdi3) movl 12(%esp), %ecx // Load count movl 8(%esp), %edx // Load high movl 4(%esp), %eax // Load low - + testl $0x20, %ecx // If count >= 32 jnz 1f // goto 1 shrdl %cl, %edx, %eax // right shift low by count sarl %cl, %edx // right shift high by count ret - + 1: movl %edx, %eax // Move high to low sarl $31, %edx // clear high sarl %cl, %eax // shift low by count - 32 diff --git a/lib/builtins/i386/chkstk.S b/lib/builtins/i386/chkstk.S index b59974868f21..f0bea2187457 100644 --- a/lib/builtins/i386/chkstk.S +++ b/lib/builtins/i386/chkstk.S @@ -1,5 +1,6 @@ -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception #include "../assembly.h" diff --git a/lib/builtins/i386/chkstk2.S b/lib/builtins/i386/chkstk2.S index 7d65bb088928..5d6cbdfa5c99 100644 --- a/lib/builtins/i386/chkstk2.S +++ b/lib/builtins/i386/chkstk2.S @@ -1,5 +1,6 @@ -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception #include "../assembly.h" diff --git a/lib/builtins/i386/divdi3.S b/lib/builtins/i386/divdi3.S index 2fb4bdcad90d..09e1e42eb719 100644 --- a/lib/builtins/i386/divdi3.S +++ b/lib/builtins/i386/divdi3.S @@ -1,5 +1,6 @@ -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception #include "../assembly.h" @@ -22,9 +23,9 @@ .balign 4 DEFINE_COMPILERRT_FUNCTION(__divdi3) -/* This is currently implemented by wrapping the unsigned divide up in an absolute - value, then restoring the correct sign at the end of the computation. This could - certainly be improved upon. */ +// This is currently implemented by wrapping the unsigned divide up in an absolute +// value, then restoring the correct sign at the end of the computation. This could +// certainly be improved upon. pushl %esi movl 20(%esp), %edx // high word of b @@ -38,7 +39,7 @@ DEFINE_COMPILERRT_FUNCTION(__divdi3) movl %edx, 20(%esp) movl %eax, 16(%esp) // store abs(b) back to stack movl %ecx, %esi // set aside sign of b - + movl 12(%esp), %edx // high word of b movl 8(%esp), %eax // low word of b movl %edx, %ecx @@ -55,11 +56,11 @@ DEFINE_COMPILERRT_FUNCTION(__divdi3) movl 24(%esp), %ebx // Find the index i of the leading bit in b. bsrl %ebx, %ecx // If the high word of b is zero, jump to jz 9f // the code to handle that special case [9]. - - /* High word of b is known to be non-zero on this branch */ - + + // High word of b is known to be non-zero on this branch + movl 20(%esp), %eax // Construct bhi, containing bits [1+i:32+i] of b - + shrl %cl, %eax // Practically, this means that bhi is given by: shrl %eax // notl %ecx // bhi = (high word of b) << (31 - i) | @@ -68,10 +69,10 @@ DEFINE_COMPILERRT_FUNCTION(__divdi3) movl 16(%esp), %edx // Load the high and low words of a, and jump movl 12(%esp), %eax // to [1] if the high word is larger than bhi cmpl %ebx, %edx // to avoid overflowing the upcoming divide. - jae 1f - - /* High word of a is greater than or equal to (b >> (1 + i)) on this branch */ - + jae 1f + + // High word of a is greater than or equal to (b >> (1 + i)) on this branch + divl %ebx // eax <-- qs, edx <-- r such that ahi:alo = bs*qs + r pushl %edi @@ -90,7 +91,7 @@ DEFINE_COMPILERRT_FUNCTION(__divdi3) sbbl $0, %edi // decrement q if remainder is negative xorl %edx, %edx movl %edi, %eax - + addl %esi, %eax // Restore correct sign to result adcl %esi, %edx xorl %esi, %eax @@ -101,8 +102,8 @@ DEFINE_COMPILERRT_FUNCTION(__divdi3) retl // Return -1: /* High word of a is greater than or equal to (b >> (1 + i)) on this branch */ - +1: // High word of a is greater than or equal to (b >> (1 + i)) on this branch + subl %ebx, %edx // subtract bhi from ahi so that divide will not divl %ebx // overflow, and find q and r such that // @@ -128,7 +129,7 @@ DEFINE_COMPILERRT_FUNCTION(__divdi3) sbbl $0, %edi // decrement q if remainder is negative xorl %edx, %edx movl %edi, %eax - + addl %esi, %eax // Restore correct sign to result adcl %esi, %edx xorl %esi, %eax @@ -138,8 +139,8 @@ DEFINE_COMPILERRT_FUNCTION(__divdi3) popl %esi retl // Return - -9: /* High word of b is zero on this branch */ + +9: // High word of b is zero on this branch movl 16(%esp), %eax // Find qhi and rhi such that movl 20(%esp), %ecx // @@ -149,7 +150,7 @@ DEFINE_COMPILERRT_FUNCTION(__divdi3) movl 12(%esp), %eax // Find qlo such that divl %ecx // movl %ebx, %edx // rhi:alo = qlo*b + rlo with 0 ≤ rlo < b - + addl %esi, %eax // Restore correct sign to result adcl %esi, %edx xorl %esi, %eax diff --git a/lib/builtins/i386/floatdidf.S b/lib/builtins/i386/floatdidf.S index d75dfe62d6a7..ab7422c312dc 100644 --- a/lib/builtins/i386/floatdidf.S +++ b/lib/builtins/i386/floatdidf.S @@ -1,5 +1,6 @@ -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception #include "../assembly.h" diff --git a/lib/builtins/i386/floatdisf.S b/lib/builtins/i386/floatdisf.S index 0874eaaa9a98..d91f14e9f330 100644 --- a/lib/builtins/i386/floatdisf.S +++ b/lib/builtins/i386/floatdisf.S @@ -1,5 +1,6 @@ -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception #include "../assembly.h" diff --git a/lib/builtins/i386/floatdixf.S b/lib/builtins/i386/floatdixf.S index 1044ef55a1a8..df70f5f9e6e3 100644 --- a/lib/builtins/i386/floatdixf.S +++ b/lib/builtins/i386/floatdixf.S @@ -1,5 +1,6 @@ -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception #include "../assembly.h" diff --git a/lib/builtins/i386/floatundidf.S b/lib/builtins/i386/floatundidf.S index fe032348e829..8b1b666ce1ec 100644 --- a/lib/builtins/i386/floatundidf.S +++ b/lib/builtins/i386/floatundidf.S @@ -1,9 +1,8 @@ //===-- floatundidf.S - Implement __floatundidf for i386 ------------------===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// // diff --git a/lib/builtins/i386/floatundisf.S b/lib/builtins/i386/floatundisf.S index 16000b576026..44301719edda 100644 --- a/lib/builtins/i386/floatundisf.S +++ b/lib/builtins/i386/floatundisf.S @@ -1,5 +1,6 @@ -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception #include "../assembly.h" @@ -48,7 +49,7 @@ END_COMPILERRT_FUNCTION(__floatundisf) */ -/* branch-free, x87-free implementation - faster at the expense of code size */ +// branch-free, x87-free implementation - faster at the expense of code size #ifdef __i386__ @@ -78,7 +79,7 @@ DEFINE_COMPILERRT_FUNCTION(__floatundisf) movd 8(%esp), %xmm1 movd 4(%esp), %xmm0 punpckldq %xmm1, %xmm0 - + calll 0f 0: popl %ecx shrl %eax // high 31 bits of input as sint32 diff --git a/lib/builtins/i386/floatundixf.S b/lib/builtins/i386/floatundixf.S index c935670cb52f..30b4d9f4b96c 100644 --- a/lib/builtins/i386/floatundixf.S +++ b/lib/builtins/i386/floatundixf.S @@ -1,5 +1,6 @@ -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception #include "../assembly.h" diff --git a/lib/builtins/i386/lshrdi3.S b/lib/builtins/i386/lshrdi3.S index 53e95cf76527..896633e85d35 100644 --- a/lib/builtins/i386/lshrdi3.S +++ b/lib/builtins/i386/lshrdi3.S @@ -1,5 +1,6 @@ -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception #include "../assembly.h" @@ -41,14 +42,14 @@ DEFINE_COMPILERRT_FUNCTION(__lshrdi3) movl 12(%esp), %ecx // Load count movl 8(%esp), %edx // Load high movl 4(%esp), %eax // Load low - + testl $0x20, %ecx // If count >= 32 jnz 1f // goto 1 shrdl %cl, %edx, %eax // right shift low by count shrl %cl, %edx // right shift high by count ret - + 1: movl %edx, %eax // Move high to low xorl %edx, %edx // clear high shrl %cl, %eax // shift low by count - 32 diff --git a/lib/builtins/i386/moddi3.S b/lib/builtins/i386/moddi3.S index a5bf9ce8ea0f..4580f20ead15 100644 --- a/lib/builtins/i386/moddi3.S +++ b/lib/builtins/i386/moddi3.S @@ -1,5 +1,6 @@ -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception #include "../assembly.h" @@ -23,8 +24,8 @@ .balign 4 DEFINE_COMPILERRT_FUNCTION(__moddi3) -/* This is currently implemented by wrapping the unsigned modulus up in an absolute - value. This could certainly be improved upon. */ +// This is currently implemented by wrapping the unsigned modulus up in an absolute +// value. This could certainly be improved upon. pushl %esi movl 20(%esp), %edx // high word of b @@ -37,7 +38,7 @@ DEFINE_COMPILERRT_FUNCTION(__moddi3) sbbl %ecx, %edx // EDX:EAX = abs(b) movl %edx, 20(%esp) movl %eax, 16(%esp) // store abs(b) back to stack - + movl 12(%esp), %edx // high word of b movl 8(%esp), %eax // low word of b movl %edx, %ecx @@ -54,11 +55,11 @@ DEFINE_COMPILERRT_FUNCTION(__moddi3) movl 24(%esp), %ebx // Find the index i of the leading bit in b. bsrl %ebx, %ecx // If the high word of b is zero, jump to jz 9f // the code to handle that special case [9]. - - /* High word of b is known to be non-zero on this branch */ - + + // High word of b is known to be non-zero on this branch + movl 20(%esp), %eax // Construct bhi, containing bits [1+i:32+i] of b - + shrl %cl, %eax // Practically, this means that bhi is given by: shrl %eax // notl %ecx // bhi = (high word of b) << (31 - i) | @@ -67,10 +68,10 @@ DEFINE_COMPILERRT_FUNCTION(__moddi3) movl 16(%esp), %edx // Load the high and low words of a, and jump movl 12(%esp), %eax // to [2] if the high word is larger than bhi cmpl %ebx, %edx // to avoid overflowing the upcoming divide. - jae 2f - - /* High word of a is greater than or equal to (b >> (1 + i)) on this branch */ - + jae 2f + + // High word of a is greater than or equal to (b >> (1 + i)) on this branch + divl %ebx // eax <-- qs, edx <-- r such that ahi:alo = bs*qs + r pushl %edi @@ -86,13 +87,13 @@ DEFINE_COMPILERRT_FUNCTION(__moddi3) movl 28(%esp), %eax imull %edi, %eax // q*bhi subl %eax, %ecx // ECX:EBX = a - q*b - + jnc 1f // if positive, this is the result. addl 24(%esp), %ebx // otherwise adcl 28(%esp), %ecx // ECX:EBX = a - (q-1)*b = result 1: movl %ebx, %eax movl %ecx, %edx - + addl %esi, %eax // Restore correct sign to result adcl %esi, %edx xorl %esi, %eax @@ -102,8 +103,8 @@ DEFINE_COMPILERRT_FUNCTION(__moddi3) popl %esi retl // Return -2: /* High word of a is greater than or equal to (b >> (1 + i)) on this branch */ - +2: // High word of a is greater than or equal to (b >> (1 + i)) on this branch + subl %ebx, %edx // subtract bhi from ahi so that divide will not divl %ebx // overflow, and find q and r such that // @@ -132,7 +133,7 @@ DEFINE_COMPILERRT_FUNCTION(__moddi3) adcl 28(%esp), %ecx // ECX:EBX = a - (q-1)*b = result 3: movl %ebx, %eax movl %ecx, %edx - + addl %esi, %eax // Restore correct sign to result adcl %esi, %edx xorl %esi, %eax @@ -141,8 +142,8 @@ DEFINE_COMPILERRT_FUNCTION(__moddi3) popl %ebx popl %esi retl // Return - -9: /* High word of b is zero on this branch */ + +9: // High word of b is zero on this branch movl 16(%esp), %eax // Find qhi and rhi such that movl 20(%esp), %ecx // diff --git a/lib/builtins/i386/muldi3.S b/lib/builtins/i386/muldi3.S index 12394606421c..a898e2414670 100644 --- a/lib/builtins/i386/muldi3.S +++ b/lib/builtins/i386/muldi3.S @@ -1,5 +1,6 @@ -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception #include "../assembly.h" @@ -14,15 +15,15 @@ DEFINE_COMPILERRT_FUNCTION(__muldi3) movl 16(%esp), %eax // b.lo movl 12(%esp), %ecx // a.hi imull %eax, %ecx // b.lo * a.hi - + movl 8(%esp), %edx // a.lo movl 20(%esp), %ebx // b.hi imull %edx, %ebx // a.lo * b.hi - + mull %edx // EDX:EAX = a.lo * b.lo addl %ecx, %ebx // EBX = (a.lo*b.hi + a.hi*b.lo) addl %ebx, %edx - + popl %ebx retl END_COMPILERRT_FUNCTION(__muldi3) diff --git a/lib/builtins/i386/udivdi3.S b/lib/builtins/i386/udivdi3.S index 727613639b12..ca390245eabc 100644 --- a/lib/builtins/i386/udivdi3.S +++ b/lib/builtins/i386/udivdi3.S @@ -1,5 +1,6 @@ -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception #include "../assembly.h" @@ -26,11 +27,11 @@ DEFINE_COMPILERRT_FUNCTION(__udivdi3) movl 20(%esp), %ebx // Find the index i of the leading bit in b. bsrl %ebx, %ecx // If the high word of b is zero, jump to jz 9f // the code to handle that special case [9]. - - /* High word of b is known to be non-zero on this branch */ - + + // High word of b is known to be non-zero on this branch + movl 16(%esp), %eax // Construct bhi, containing bits [1+i:32+i] of b - + shrl %cl, %eax // Practically, this means that bhi is given by: shrl %eax // notl %ecx // bhi = (high word of b) << (31 - i) | @@ -39,10 +40,10 @@ DEFINE_COMPILERRT_FUNCTION(__udivdi3) movl 12(%esp), %edx // Load the high and low words of a, and jump movl 8(%esp), %eax // to [1] if the high word is larger than bhi cmpl %ebx, %edx // to avoid overflowing the upcoming divide. - jae 1f - - /* High word of a is greater than or equal to (b >> (1 + i)) on this branch */ - + jae 1f + + // High word of a is greater than or equal to (b >> (1 + i)) on this branch + divl %ebx // eax <-- qs, edx <-- r such that ahi:alo = bs*qs + r pushl %edi @@ -66,8 +67,8 @@ DEFINE_COMPILERRT_FUNCTION(__udivdi3) retl -1: /* High word of a is greater than or equal to (b >> (1 + i)) on this branch */ - +1: // High word of a is greater than or equal to (b >> (1 + i)) on this branch + subl %ebx, %edx // subtract bhi from ahi so that divide will not divl %ebx // overflow, and find q and r such that // @@ -97,8 +98,8 @@ DEFINE_COMPILERRT_FUNCTION(__udivdi3) popl %ebx retl - -9: /* High word of b is zero on this branch */ + +9: // High word of b is zero on this branch movl 12(%esp), %eax // Find qhi and rhi such that movl 16(%esp), %ecx // diff --git a/lib/builtins/i386/umoddi3.S b/lib/builtins/i386/umoddi3.S index 763e821946c0..2717e7e5da68 100644 --- a/lib/builtins/i386/umoddi3.S +++ b/lib/builtins/i386/umoddi3.S @@ -1,5 +1,6 @@ -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception #include "../assembly.h" @@ -27,11 +28,11 @@ DEFINE_COMPILERRT_FUNCTION(__umoddi3) movl 20(%esp), %ebx // Find the index i of the leading bit in b. bsrl %ebx, %ecx // If the high word of b is zero, jump to jz 9f // the code to handle that special case [9]. - - /* High word of b is known to be non-zero on this branch */ - + + // High word of b is known to be non-zero on this branch + movl 16(%esp), %eax // Construct bhi, containing bits [1+i:32+i] of b - + shrl %cl, %eax // Practically, this means that bhi is given by: shrl %eax // notl %ecx // bhi = (high word of b) << (31 - i) | @@ -40,10 +41,10 @@ DEFINE_COMPILERRT_FUNCTION(__umoddi3) movl 12(%esp), %edx // Load the high and low words of a, and jump movl 8(%esp), %eax // to [2] if the high word is larger than bhi cmpl %ebx, %edx // to avoid overflowing the upcoming divide. - jae 2f - - /* High word of a is greater than or equal to (b >> (1 + i)) on this branch */ - + jae 2f + + // High word of a is greater than or equal to (b >> (1 + i)) on this branch + divl %ebx // eax <-- qs, edx <-- r such that ahi:alo = bs*qs + r pushl %edi @@ -59,20 +60,20 @@ DEFINE_COMPILERRT_FUNCTION(__umoddi3) movl 24(%esp), %eax imull %edi, %eax // q*bhi subl %eax, %ecx // ECX:EBX = a - q*b - + jnc 1f // if positive, this is the result. addl 20(%esp), %ebx // otherwise adcl 24(%esp), %ecx // ECX:EBX = a - (q-1)*b = result 1: movl %ebx, %eax movl %ecx, %edx - + popl %edi popl %ebx retl -2: /* High word of a is greater than or equal to (b >> (1 + i)) on this branch */ - +2: // High word of a is greater than or equal to (b >> (1 + i)) on this branch + subl %ebx, %edx // subtract bhi from ahi so that divide will not divl %ebx // overflow, and find q and r such that // @@ -101,14 +102,14 @@ DEFINE_COMPILERRT_FUNCTION(__umoddi3) adcl 24(%esp), %ecx // ECX:EBX = a - (q-1)*b = result 3: movl %ebx, %eax movl %ecx, %edx - + popl %edi popl %ebx retl - -9: /* High word of b is zero on this branch */ + +9: // High word of b is zero on this branch movl 12(%esp), %eax // Find qhi and rhi such that movl 16(%esp), %ecx // @@ -120,7 +121,7 @@ DEFINE_COMPILERRT_FUNCTION(__umoddi3) movl %edx, %eax // rhi:alo = qlo*b + rlo with 0 ≤ rlo < b popl %ebx // xorl %edx, %edx // and return 0:rlo - retl // + retl // END_COMPILERRT_FUNCTION(__umoddi3) #endif // __i386__ diff --git a/lib/builtins/int_endianness.h b/lib/builtins/int_endianness.h index e2586c56bac8..def046c34a62 100644 --- a/lib/builtins/int_endianness.h +++ b/lib/builtins/int_endianness.h @@ -1,51 +1,49 @@ -/* ===-- int_endianness.h - configuration header for compiler-rt ------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - * - * This file is a configuration header for compiler-rt. - * This file is not part of the interface of this library. - * - * ===----------------------------------------------------------------------=== - */ +//===-- int_endianness.h - configuration header for compiler-rt -----------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file is a configuration header for compiler-rt. +// This file is not part of the interface of this library. +// +//===----------------------------------------------------------------------===// #ifndef INT_ENDIANNESS_H #define INT_ENDIANNESS_H -#if defined(__BYTE_ORDER__) && defined(__ORDER_BIG_ENDIAN__) && \ +#if defined(__BYTE_ORDER__) && defined(__ORDER_BIG_ENDIAN__) && \ defined(__ORDER_LITTLE_ENDIAN__) -/* Clang and GCC provide built-in endianness definitions. */ +// Clang and GCC provide built-in endianness definitions. #if __BYTE_ORDER__ == __ORDER_BIG_ENDIAN__ #define _YUGA_LITTLE_ENDIAN 0 -#define _YUGA_BIG_ENDIAN 1 +#define _YUGA_BIG_ENDIAN 1 #elif __BYTE_ORDER__ == __ORDER_LITTLE_ENDIAN__ #define _YUGA_LITTLE_ENDIAN 1 -#define _YUGA_BIG_ENDIAN 0 -#endif /* __BYTE_ORDER__ */ +#define _YUGA_BIG_ENDIAN 0 +#endif // __BYTE_ORDER__ -#else /* Compilers other than Clang or GCC. */ +#else // Compilers other than Clang or GCC. #if defined(__SVR4) && defined(__sun) #include <sys/byteorder.h> #if defined(_BIG_ENDIAN) #define _YUGA_LITTLE_ENDIAN 0 -#define _YUGA_BIG_ENDIAN 1 +#define _YUGA_BIG_ENDIAN 1 #elif defined(_LITTLE_ENDIAN) #define _YUGA_LITTLE_ENDIAN 1 -#define _YUGA_BIG_ENDIAN 0 -#else /* !_LITTLE_ENDIAN */ +#define _YUGA_BIG_ENDIAN 0 +#else // !_LITTLE_ENDIAN #error "unknown endianness" -#endif /* !_LITTLE_ENDIAN */ +#endif // !_LITTLE_ENDIAN -#endif /* Solaris and AuroraUX. */ +#endif // Solaris and AuroraUX. -/* .. */ +// .. #if defined(__FreeBSD__) || defined(__NetBSD__) || defined(__DragonFly__) || \ defined(__minix) @@ -53,64 +51,64 @@ #if _BYTE_ORDER == _BIG_ENDIAN #define _YUGA_LITTLE_ENDIAN 0 -#define _YUGA_BIG_ENDIAN 1 +#define _YUGA_BIG_ENDIAN 1 #elif _BYTE_ORDER == _LITTLE_ENDIAN #define _YUGA_LITTLE_ENDIAN 1 -#define _YUGA_BIG_ENDIAN 0 -#endif /* _BYTE_ORDER */ +#define _YUGA_BIG_ENDIAN 0 +#endif // _BYTE_ORDER -#endif /* *BSD */ +#endif // *BSD #if defined(__OpenBSD__) #include <machine/endian.h> #if _BYTE_ORDER == _BIG_ENDIAN #define _YUGA_LITTLE_ENDIAN 0 -#define _YUGA_BIG_ENDIAN 1 +#define _YUGA_BIG_ENDIAN 1 #elif _BYTE_ORDER == _LITTLE_ENDIAN #define _YUGA_LITTLE_ENDIAN 1 -#define _YUGA_BIG_ENDIAN 0 -#endif /* _BYTE_ORDER */ +#define _YUGA_BIG_ENDIAN 0 +#endif // _BYTE_ORDER -#endif /* OpenBSD */ +#endif // OpenBSD -/* .. */ +// .. -/* Mac OSX has __BIG_ENDIAN__ or __LITTLE_ENDIAN__ automatically set by the - * compiler (at least with GCC) */ -#if defined(__APPLE__) || defined(__ellcc__ ) +// Mac OSX has __BIG_ENDIAN__ or __LITTLE_ENDIAN__ automatically set by the +// compiler (at least with GCC) +#if defined(__APPLE__) || defined(__ellcc__) #ifdef __BIG_ENDIAN__ #if __BIG_ENDIAN__ #define _YUGA_LITTLE_ENDIAN 0 -#define _YUGA_BIG_ENDIAN 1 +#define _YUGA_BIG_ENDIAN 1 #endif -#endif /* __BIG_ENDIAN__ */ +#endif // __BIG_ENDIAN__ #ifdef __LITTLE_ENDIAN__ #if __LITTLE_ENDIAN__ #define _YUGA_LITTLE_ENDIAN 1 -#define _YUGA_BIG_ENDIAN 0 +#define _YUGA_BIG_ENDIAN 0 #endif -#endif /* __LITTLE_ENDIAN__ */ +#endif // __LITTLE_ENDIAN__ -#endif /* Mac OSX */ +#endif // Mac OSX -/* .. */ +// .. #if defined(_WIN32) #define _YUGA_LITTLE_ENDIAN 1 -#define _YUGA_BIG_ENDIAN 0 +#define _YUGA_BIG_ENDIAN 0 -#endif /* Windows */ +#endif // Windows -#endif /* Clang or GCC. */ +#endif // Clang or GCC. -/* . */ +// . #if !defined(_YUGA_LITTLE_ENDIAN) || !defined(_YUGA_BIG_ENDIAN) #error Unable to determine endian -#endif /* Check we found an endianness correctly. */ +#endif // Check we found an endianness correctly. -#endif /* INT_ENDIANNESS_H */ +#endif // INT_ENDIANNESS_H diff --git a/lib/builtins/int_lib.h b/lib/builtins/int_lib.h index fe8a3bdedc0e..3092f68c084a 100644 --- a/lib/builtins/int_lib.h +++ b/lib/builtins/int_lib.h @@ -1,44 +1,33 @@ -/* ===-- int_lib.h - configuration header for compiler-rt -----------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - * - * This file is a configuration header for compiler-rt. - * This file is not part of the interface of this library. - * - * ===----------------------------------------------------------------------=== - */ +//===-- int_lib.h - configuration header for compiler-rt -----------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file is a configuration header for compiler-rt. +// This file is not part of the interface of this library. +// +//===----------------------------------------------------------------------===// #ifndef INT_LIB_H #define INT_LIB_H -/* Assumption: Signed integral is 2's complement. */ -/* Assumption: Right shift of signed negative is arithmetic shift. */ -/* Assumption: Endianness is little or big (not mixed). */ +// Assumption: Signed integral is 2's complement. +// Assumption: Right shift of signed negative is arithmetic shift. +// Assumption: Endianness is little or big (not mixed). -#if defined(__ELF__) -#define FNALIAS(alias_name, original_name) \ - void alias_name() __attribute__((__alias__(#original_name))) -#define COMPILER_RT_ALIAS(aliasee) __attribute__((__alias__(#aliasee))) -#else -#define FNALIAS(alias, name) _Pragma("GCC error(\"alias unsupported on this file format\")") -#define COMPILER_RT_ALIAS(aliasee) _Pragma("GCC error(\"alias unsupported on this file format\")") -#endif - -/* ABI macro definitions */ +// ABI macro definitions #if __ARM_EABI__ -# ifdef COMPILER_RT_ARMHF_TARGET -# define COMPILER_RT_ABI -# else -# define COMPILER_RT_ABI __attribute__((__pcs__("aapcs"))) -# endif +#ifdef COMPILER_RT_ARMHF_TARGET +#define COMPILER_RT_ABI +#else +#define COMPILER_RT_ABI __attribute__((__pcs__("aapcs"))) +#endif #else -# define COMPILER_RT_ABI +#define COMPILER_RT_ABI #endif #define AEABI_RTABI __attribute__((__pcs__("aapcs"))) @@ -55,26 +44,44 @@ #define UNUSED __attribute__((unused)) #endif +#define STR(a) #a +#define XSTR(a) STR(a) +#define SYMBOL_NAME(name) XSTR(__USER_LABEL_PREFIX__) #name + +#if defined(__ELF__) || defined(__MINGW32__) || defined(__wasm__) +#define COMPILER_RT_ALIAS(name, aliasname) \ + COMPILER_RT_ABI __typeof(name) aliasname __attribute__((__alias__(#name))); +#elif defined(__APPLE__) +#define COMPILER_RT_ALIAS(name, aliasname) \ + __asm__(".globl " SYMBOL_NAME(aliasname)); \ + __asm__(SYMBOL_NAME(aliasname) " = " SYMBOL_NAME(name)); \ + COMPILER_RT_ABI __typeof(name) aliasname; +#elif defined(_WIN32) +#define COMPILER_RT_ALIAS(name, aliasname) +#else +#error Unsupported target +#endif + #if defined(__NetBSD__) && (defined(_KERNEL) || defined(_STANDALONE)) -/* - * Kernel and boot environment can't use normal headers, - * so use the equivalent system headers. - */ -# include <machine/limits.h> -# include <sys/stdint.h> -# include <sys/types.h> +// +// Kernel and boot environment can't use normal headers, +// so use the equivalent system headers. +// +#include <machine/limits.h> +#include <sys/stdint.h> +#include <sys/types.h> #else -/* Include the standard compiler builtin headers we use functionality from. */ -# include <limits.h> -# include <stdint.h> -# include <stdbool.h> -# include <float.h> +// Include the standard compiler builtin headers we use functionality from. +#include <float.h> +#include <limits.h> +#include <stdbool.h> +#include <stdint.h> #endif -/* Include the commonly used internal type definitions. */ +// Include the commonly used internal type definitions. #include "int_types.h" -/* Include internal utility function declarations. */ +// Include internal utility function declarations. #include "int_util.h" COMPILER_RT_ABI si_int __paritysi2(si_int a); @@ -84,14 +91,14 @@ COMPILER_RT_ABI di_int __divdi3(di_int a, di_int b); COMPILER_RT_ABI si_int __divsi3(si_int a, si_int b); COMPILER_RT_ABI su_int __udivsi3(su_int n, su_int d); -COMPILER_RT_ABI su_int __udivmodsi4(su_int a, su_int b, su_int* rem); -COMPILER_RT_ABI du_int __udivmoddi4(du_int a, du_int b, du_int* rem); +COMPILER_RT_ABI su_int __udivmodsi4(su_int a, su_int b, su_int *rem); +COMPILER_RT_ABI du_int __udivmoddi4(du_int a, du_int b, du_int *rem); #ifdef CRT_HAS_128BIT COMPILER_RT_ABI si_int __clzti2(ti_int a); -COMPILER_RT_ABI tu_int __udivmodti4(tu_int a, tu_int b, tu_int* rem); +COMPILER_RT_ABI tu_int __udivmodti4(tu_int a, tu_int b, tu_int *rem); #endif -/* Definitions for builtins unavailable on MSVC */ +// Definitions for builtins unavailable on MSVC #if defined(_MSC_VER) && !defined(__clang__) #include <intrin.h> @@ -129,6 +136,6 @@ uint32_t __inline __builtin_clzll(uint64_t value) { #endif #define __builtin_clzl __builtin_clzll -#endif /* defined(_MSC_VER) && !defined(__clang__) */ +#endif // defined(_MSC_VER) && !defined(__clang__) -#endif /* INT_LIB_H */ +#endif // INT_LIB_H diff --git a/lib/builtins/int_math.h b/lib/builtins/int_math.h index aa3d0721a8ab..58d8990f31b1 100644 --- a/lib/builtins/int_math.h +++ b/lib/builtins/int_math.h @@ -1,34 +1,31 @@ -/* ===-- int_math.h - internal math inlines ---------------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===-----------------------------------------------------------------------=== - * - * This file is not part of the interface of this library. - * - * This file defines substitutes for the libm functions used in some of the - * compiler-rt implementations, defined in such a way that there is not a direct - * dependency on libm or math.h. Instead, we use the compiler builtin versions - * where available. This reduces our dependencies on the system SDK by foisting - * the responsibility onto the compiler. - * - * ===-----------------------------------------------------------------------=== - */ +//===-- int_math.h - internal math inlines --------------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file is not part of the interface of this library. +// +// This file defines substitutes for the libm functions used in some of the +// compiler-rt implementations, defined in such a way that there is not a direct +// dependency on libm or math.h. Instead, we use the compiler builtin versions +// where available. This reduces our dependencies on the system SDK by foisting +// the responsibility onto the compiler. +// +//===----------------------------------------------------------------------===// #ifndef INT_MATH_H #define INT_MATH_H #ifndef __has_builtin -# define __has_builtin(x) 0 +#define __has_builtin(x) 0 #endif #if defined(_MSC_VER) && !defined(__clang__) #include <math.h> #include <stdlib.h> -#include <ymath.h> #endif #if defined(_MSC_VER) && !defined(__clang__) @@ -42,24 +39,23 @@ #define crt_isinf(x) !_finite((x)) #define crt_isnan(x) _isnan((x)) #else -/* Define crt_isfinite in terms of the builtin if available, otherwise provide - * an alternate version in terms of our other functions. This supports some - * versions of GCC which didn't have __builtin_isfinite. - */ +// Define crt_isfinite in terms of the builtin if available, otherwise provide +// an alternate version in terms of our other functions. This supports some +// versions of GCC which didn't have __builtin_isfinite. #if __has_builtin(__builtin_isfinite) -# define crt_isfinite(x) __builtin_isfinite((x)) +#define crt_isfinite(x) __builtin_isfinite((x)) #elif defined(__GNUC__) -# define crt_isfinite(x) \ - __extension__(({ \ - __typeof((x)) x_ = (x); \ - !crt_isinf(x_) && !crt_isnan(x_); \ - })) +#define crt_isfinite(x) \ + __extension__(({ \ + __typeof((x)) x_ = (x); \ + !crt_isinf(x_) && !crt_isnan(x_); \ + })) #else -# error "Do not know how to check for infinity" -#endif /* __has_builtin(__builtin_isfinite) */ +#error "Do not know how to check for infinity" +#endif // __has_builtin(__builtin_isfinite) #define crt_isinf(x) __builtin_isinf((x)) #define crt_isnan(x) __builtin_isnan((x)) -#endif /* _MSC_VER */ +#endif // _MSC_VER #if defined(_MSC_VER) && !defined(__clang__) #define crt_copysign(x, y) copysign((x), (y)) @@ -107,4 +103,4 @@ #define crt_scalbnl(x, y) __builtin_scalbnl((x), (y)) #endif -#endif /* INT_MATH_H */ +#endif // INT_MATH_H diff --git a/lib/builtins/int_types.h b/lib/builtins/int_types.h index 9f8da56cb77b..f89220d54350 100644 --- a/lib/builtins/int_types.h +++ b/lib/builtins/int_types.h @@ -1,63 +1,57 @@ -/* ===-- int_lib.h - configuration header for compiler-rt -----------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - * - * This file is not part of the interface of this library. - * - * This file defines various standard types, most importantly a number of unions - * used to access parts of larger types. - * - * ===----------------------------------------------------------------------=== - */ +//===-- int_lib.h - configuration header for compiler-rt -----------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file is not part of the interface of this library. +// +// This file defines various standard types, most importantly a number of unions +// used to access parts of larger types. +// +//===----------------------------------------------------------------------===// #ifndef INT_TYPES_H #define INT_TYPES_H #include "int_endianness.h" -/* si_int is defined in Linux sysroot's asm-generic/siginfo.h */ +// si_int is defined in Linux sysroot's asm-generic/siginfo.h #ifdef si_int #undef si_int #endif -typedef int si_int; +typedef int si_int; typedef unsigned su_int; -typedef long long di_int; +typedef long long di_int; typedef unsigned long long du_int; -typedef union -{ - di_int all; - struct - { +typedef union { + di_int all; + struct { #if _YUGA_LITTLE_ENDIAN - su_int low; - si_int high; + su_int low; + si_int high; #else - si_int high; - su_int low; -#endif /* _YUGA_LITTLE_ENDIAN */ - }s; + si_int high; + su_int low; +#endif // _YUGA_LITTLE_ENDIAN + } s; } dwords; -typedef union -{ - du_int all; - struct - { +typedef union { + du_int all; + struct { #if _YUGA_LITTLE_ENDIAN - su_int low; - su_int high; + su_int low; + su_int high; #else - su_int high; - su_int low; -#endif /* _YUGA_LITTLE_ENDIAN */ - }s; + su_int high; + su_int low; +#endif // _YUGA_LITTLE_ENDIAN + } s; } udwords; #if defined(__LP64__) || defined(__wasm__) || defined(__mips64) || \ @@ -65,103 +59,93 @@ typedef union #define CRT_HAS_128BIT #endif -/* MSVC doesn't have a working 128bit integer type. Users should really compile - * compiler-rt with clang, but if they happen to be doing a standalone build for - * asan or something else, disable the 128 bit parts so things sort of work. - */ +// MSVC doesn't have a working 128bit integer type. Users should really compile +// compiler-rt with clang, but if they happen to be doing a standalone build for +// asan or something else, disable the 128 bit parts so things sort of work. #if defined(_MSC_VER) && !defined(__clang__) #undef CRT_HAS_128BIT #endif #ifdef CRT_HAS_128BIT -typedef int ti_int __attribute__ ((mode (TI))); -typedef unsigned tu_int __attribute__ ((mode (TI))); - -typedef union -{ - ti_int all; - struct - { +typedef int ti_int __attribute__((mode(TI))); +typedef unsigned tu_int __attribute__((mode(TI))); + +typedef union { + ti_int all; + struct { #if _YUGA_LITTLE_ENDIAN - du_int low; - di_int high; + du_int low; + di_int high; #else - di_int high; - du_int low; -#endif /* _YUGA_LITTLE_ENDIAN */ - }s; + di_int high; + du_int low; +#endif // _YUGA_LITTLE_ENDIAN + } s; } twords; -typedef union -{ - tu_int all; - struct - { +typedef union { + tu_int all; + struct { #if _YUGA_LITTLE_ENDIAN - du_int low; - du_int high; + du_int low; + du_int high; #else - du_int high; - du_int low; -#endif /* _YUGA_LITTLE_ENDIAN */ - }s; + du_int high; + du_int low; +#endif // _YUGA_LITTLE_ENDIAN + } s; } utwords; static __inline ti_int make_ti(di_int h, di_int l) { - twords r; - r.s.high = h; - r.s.low = l; - return r.all; + twords r; + r.s.high = h; + r.s.low = l; + return r.all; } static __inline tu_int make_tu(du_int h, du_int l) { - utwords r; - r.s.high = h; - r.s.low = l; - return r.all; + utwords r; + r.s.high = h; + r.s.low = l; + return r.all; } -#endif /* CRT_HAS_128BIT */ +#endif // CRT_HAS_128BIT -typedef union -{ - su_int u; - float f; +typedef union { + su_int u; + float f; } float_bits; -typedef union -{ - udwords u; - double f; +typedef union { + udwords u; + double f; } double_bits; -typedef struct -{ +typedef struct { #if _YUGA_LITTLE_ENDIAN - udwords low; - udwords high; + udwords low; + udwords high; #else - udwords high; - udwords low; -#endif /* _YUGA_LITTLE_ENDIAN */ + udwords high; + udwords low; +#endif // _YUGA_LITTLE_ENDIAN } uqwords; -/* Check if the target supports 80 bit extended precision long doubles. - * Notably, on x86 Windows, MSVC only provides a 64-bit long double, but GCC - * still makes it 80 bits. Clang will match whatever compiler it is trying to - * be compatible with. - */ -#if ((defined(__i386__) || defined(__x86_64__)) && !defined(_MSC_VER)) || \ +// Check if the target supports 80 bit extended precision long doubles. +// Notably, on x86 Windows, MSVC only provides a 64-bit long double, but GCC +// still makes it 80 bits. Clang will match whatever compiler it is trying to +// be compatible with. +#if ((defined(__i386__) || defined(__x86_64__)) && !defined(_MSC_VER)) || \ defined(__m68k__) || defined(__ia64__) #define HAS_80_BIT_LONG_DOUBLE 1 #else #define HAS_80_BIT_LONG_DOUBLE 0 #endif -typedef union -{ - uqwords u; - long double f; +typedef union { + uqwords u; + long double f; } long_double_bits; #if __STDC_VERSION__ >= 199901L @@ -172,14 +156,19 @@ typedef long double _Complex Lcomplex; #define COMPLEX_REAL(x) __real__(x) #define COMPLEX_IMAGINARY(x) __imag__(x) #else -typedef struct { float real, imaginary; } Fcomplex; +typedef struct { + float real, imaginary; +} Fcomplex; -typedef struct { double real, imaginary; } Dcomplex; +typedef struct { + double real, imaginary; +} Dcomplex; -typedef struct { long double real, imaginary; } Lcomplex; +typedef struct { + long double real, imaginary; +} Lcomplex; #define COMPLEX_REAL(x) (x).real #define COMPLEX_IMAGINARY(x) (x).imaginary #endif -#endif /* INT_TYPES_H */ - +#endif // INT_TYPES_H diff --git a/lib/builtins/int_util.c b/lib/builtins/int_util.c index 752f2015580e..226a6e93440d 100644 --- a/lib/builtins/int_util.c +++ b/lib/builtins/int_util.c @@ -1,25 +1,21 @@ -/* ===-- int_util.c - Implement internal utilities --------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - */ +//===-- int_util.c - Implement internal utilities -------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// #include "int_lib.h" -#include "int_util.h" -/* NOTE: The definitions in this file are declared weak because we clients to be - * able to arbitrarily package individual functions into separate .a files. If - * we did not declare these weak, some link situations might end up seeing - * duplicate strong definitions of the same symbol. - * - * We can't use this solution for kernel use (which may not support weak), but - * currently expect that when built for kernel use all the functionality is - * packaged into a single library. - */ +// NOTE: The definitions in this file are declared weak because we clients to be +// able to arbitrarily package individual functions into separate .a files. If +// we did not declare these weak, some link situations might end up seeing +// duplicate strong definitions of the same symbol. +// +// We can't use this solution for kernel use (which may not support weak), but +// currently expect that when built for kernel use all the functionality is +// packaged into a single library. #ifdef KERNEL_USE @@ -33,7 +29,7 @@ void __compilerrt_abort_impl(const char *file, int line, const char *function) { #elif __APPLE__ -/* from libSystem.dylib */ +// from libSystem.dylib NORETURN extern void __assert_rtn(const char *func, const char *file, int line, const char *message); @@ -57,7 +53,7 @@ void __compilerrt_abort_impl(const char *file, int line, const char *function) { #else -/* Get the system definition of abort() */ +// Get the system definition of abort() #include <stdlib.h> #ifndef _WIN32 diff --git a/lib/builtins/int_util.h b/lib/builtins/int_util.h index c3c87381ad8a..5fbdfb57c1e7 100644 --- a/lib/builtins/int_util.h +++ b/lib/builtins/int_util.h @@ -1,25 +1,23 @@ -/* ===-- int_util.h - internal utility functions ----------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===-----------------------------------------------------------------------=== - * - * This file is not part of the interface of this library. - * - * This file defines non-inline utilities which are available for use in the - * library. The function definitions themselves are all contained in int_util.c - * which will always be compiled into any compiler-rt library. - * - * ===-----------------------------------------------------------------------=== - */ +//===-- int_util.h - internal utility functions ---------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file is not part of the interface of this library. +// +// This file defines non-inline utilities which are available for use in the +// library. The function definitions themselves are all contained in int_util.c +// which will always be compiled into any compiler-rt library. +// +//===----------------------------------------------------------------------===// #ifndef INT_UTIL_H #define INT_UTIL_H -/** \brief Trigger a program abort (or panic for kernel code). */ +/// \brief Trigger a program abort (or panic for kernel code). #define compilerrt_abort() __compilerrt_abort_impl(__FILE__, __LINE__, __func__) NORETURN void __compilerrt_abort_impl(const char *file, int line, @@ -30,4 +28,4 @@ NORETURN void __compilerrt_abort_impl(const char *file, int line, #define COMPILE_TIME_ASSERT2(expr, cnt) \ typedef char ct_assert_##cnt[(expr) ? 1 : -1] UNUSED -#endif /* INT_UTIL_H */ +#endif // INT_UTIL_H diff --git a/lib/builtins/lshrdi3.c b/lib/builtins/lshrdi3.c index 67b2a7668345..97e08e1e9ba0 100644 --- a/lib/builtins/lshrdi3.c +++ b/lib/builtins/lshrdi3.c @@ -1,45 +1,38 @@ -/* ===-- lshrdi3.c - Implement __lshrdi3 -----------------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - * - * This file implements __lshrdi3 for the compiler_rt library. - * - * ===----------------------------------------------------------------------=== - */ +//===-- lshrdi3.c - Implement __lshrdi3 -----------------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements __lshrdi3 for the compiler_rt library. +// +//===----------------------------------------------------------------------===// #include "int_lib.h" -/* Returns: logical a >> b */ +// Returns: logical a >> b -/* Precondition: 0 <= b < bits_in_dword */ +// Precondition: 0 <= b < bits_in_dword -COMPILER_RT_ABI di_int -__lshrdi3(di_int a, si_int b) -{ - const int bits_in_word = (int)(sizeof(si_int) * CHAR_BIT); - udwords input; - udwords result; - input.all = a; - if (b & bits_in_word) /* bits_in_word <= b < bits_in_dword */ - { - result.s.high = 0; - result.s.low = input.s.high >> (b - bits_in_word); - } - else /* 0 <= b < bits_in_word */ - { - if (b == 0) - return a; - result.s.high = input.s.high >> b; - result.s.low = (input.s.high << (bits_in_word - b)) | (input.s.low >> b); - } - return result.all; +COMPILER_RT_ABI di_int __lshrdi3(di_int a, si_int b) { + const int bits_in_word = (int)(sizeof(si_int) * CHAR_BIT); + udwords input; + udwords result; + input.all = a; + if (b & bits_in_word) /* bits_in_word <= b < bits_in_dword */ { + result.s.high = 0; + result.s.low = input.s.high >> (b - bits_in_word); + } else /* 0 <= b < bits_in_word */ { + if (b == 0) + return a; + result.s.high = input.s.high >> b; + result.s.low = (input.s.high << (bits_in_word - b)) | (input.s.low >> b); + } + return result.all; } #if defined(__ARM_EABI__) -AEABI_RTABI di_int __aeabi_llsr(di_int a, si_int b) COMPILER_RT_ALIAS(__lshrdi3); +COMPILER_RT_ALIAS(__lshrdi3, __aeabi_llsr) #endif diff --git a/lib/builtins/lshrti3.c b/lib/builtins/lshrti3.c index e4170ff84a5a..d00a22095993 100644 --- a/lib/builtins/lshrti3.c +++ b/lib/builtins/lshrti3.c @@ -1,45 +1,38 @@ -/* ===-- lshrti3.c - Implement __lshrti3 -----------------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - * - * This file implements __lshrti3 for the compiler_rt library. - * - * ===----------------------------------------------------------------------=== - */ +//===-- lshrti3.c - Implement __lshrti3 -----------------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements __lshrti3 for the compiler_rt library. +// +//===----------------------------------------------------------------------===// #include "int_lib.h" #ifdef CRT_HAS_128BIT -/* Returns: logical a >> b */ +// Returns: logical a >> b -/* Precondition: 0 <= b < bits_in_tword */ +// Precondition: 0 <= b < bits_in_tword -COMPILER_RT_ABI ti_int -__lshrti3(ti_int a, si_int b) -{ - const int bits_in_dword = (int)(sizeof(di_int) * CHAR_BIT); - utwords input; - utwords result; - input.all = a; - if (b & bits_in_dword) /* bits_in_dword <= b < bits_in_tword */ - { - result.s.high = 0; - result.s.low = input.s.high >> (b - bits_in_dword); - } - else /* 0 <= b < bits_in_dword */ - { - if (b == 0) - return a; - result.s.high = input.s.high >> b; - result.s.low = (input.s.high << (bits_in_dword - b)) | (input.s.low >> b); - } - return result.all; +COMPILER_RT_ABI ti_int __lshrti3(ti_int a, si_int b) { + const int bits_in_dword = (int)(sizeof(di_int) * CHAR_BIT); + utwords input; + utwords result; + input.all = a; + if (b & bits_in_dword) /* bits_in_dword <= b < bits_in_tword */ { + result.s.high = 0; + result.s.low = input.s.high >> (b - bits_in_dword); + } else /* 0 <= b < bits_in_dword */ { + if (b == 0) + return a; + result.s.high = input.s.high >> b; + result.s.low = (input.s.high << (bits_in_dword - b)) | (input.s.low >> b); + } + return result.all; } -#endif /* CRT_HAS_128BIT */ +#endif // CRT_HAS_128BIT diff --git a/lib/builtins/mingw_fixfloat.c b/lib/builtins/mingw_fixfloat.c index c462e0dbf654..945be9d4344a 100644 --- a/lib/builtins/mingw_fixfloat.c +++ b/lib/builtins/mingw_fixfloat.c @@ -1,12 +1,10 @@ -/* ===-- mingw_fixfloat.c - Wrap int/float conversions for arm/windows -----=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - */ +//===-- mingw_fixfloat.c - Wrap int/float conversions for arm/windows -----===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// #include "int_lib.h" diff --git a/lib/builtins/moddi3.c b/lib/builtins/moddi3.c index a04279e3875f..92b0996077c6 100644 --- a/lib/builtins/moddi3.c +++ b/lib/builtins/moddi3.c @@ -1,30 +1,26 @@ -/*===-- moddi3.c - Implement __moddi3 -------------------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - * - * This file implements __moddi3 for the compiler_rt library. - * - * ===----------------------------------------------------------------------=== - */ +//===-- moddi3.c - Implement __moddi3 -------------------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements __moddi3 for the compiler_rt library. +// +//===----------------------------------------------------------------------===// #include "int_lib.h" -/* Returns: a % b */ +// Returns: a % b -COMPILER_RT_ABI di_int -__moddi3(di_int a, di_int b) -{ - const int bits_in_dword_m1 = (int)(sizeof(di_int) * CHAR_BIT) - 1; - di_int s = b >> bits_in_dword_m1; /* s = b < 0 ? -1 : 0 */ - b = (b ^ s) - s; /* negate if s == -1 */ - s = a >> bits_in_dword_m1; /* s = a < 0 ? -1 : 0 */ - a = (a ^ s) - s; /* negate if s == -1 */ - du_int r; - __udivmoddi4(a, b, &r); - return ((di_int)r ^ s) - s; /* negate if s == -1 */ +COMPILER_RT_ABI di_int __moddi3(di_int a, di_int b) { + const int bits_in_dword_m1 = (int)(sizeof(di_int) * CHAR_BIT) - 1; + di_int s = b >> bits_in_dword_m1; // s = b < 0 ? -1 : 0 + b = (b ^ s) - s; // negate if s == -1 + s = a >> bits_in_dword_m1; // s = a < 0 ? -1 : 0 + a = (a ^ s) - s; // negate if s == -1 + du_int r; + __udivmoddi4(a, b, &r); + return ((di_int)r ^ s) - s; // negate if s == -1 } diff --git a/lib/builtins/modsi3.c b/lib/builtins/modsi3.c index 86c73ce13777..e443b8a59096 100644 --- a/lib/builtins/modsi3.c +++ b/lib/builtins/modsi3.c @@ -1,23 +1,19 @@ -/* ===-- modsi3.c - Implement __modsi3 -------------------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - * - * This file implements __modsi3 for the compiler_rt library. - * - * ===----------------------------------------------------------------------=== - */ +//===-- modsi3.c - Implement __modsi3 -------------------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements __modsi3 for the compiler_rt library. +// +//===----------------------------------------------------------------------===// #include "int_lib.h" -/* Returns: a % b */ +// Returns: a % b -COMPILER_RT_ABI si_int -__modsi3(si_int a, si_int b) -{ - return a - __divsi3(a, b) * b; +COMPILER_RT_ABI si_int __modsi3(si_int a, si_int b) { + return a - __divsi3(a, b) * b; } diff --git a/lib/builtins/modti3.c b/lib/builtins/modti3.c index d505c07ac162..d11fe220b769 100644 --- a/lib/builtins/modti3.c +++ b/lib/builtins/modti3.c @@ -1,34 +1,30 @@ -/* ===-- modti3.c - Implement __modti3 -------------------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - * - * This file implements __modti3 for the compiler_rt library. - * - * ===----------------------------------------------------------------------=== - */ +//===-- modti3.c - Implement __modti3 -------------------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements __modti3 for the compiler_rt library. +// +//===----------------------------------------------------------------------===// #include "int_lib.h" #ifdef CRT_HAS_128BIT -/*Returns: a % b */ +// Returns: a % b -COMPILER_RT_ABI ti_int -__modti3(ti_int a, ti_int b) -{ - const int bits_in_tword_m1 = (int)(sizeof(ti_int) * CHAR_BIT) - 1; - ti_int s = b >> bits_in_tword_m1; /* s = b < 0 ? -1 : 0 */ - b = (b ^ s) - s; /* negate if s == -1 */ - s = a >> bits_in_tword_m1; /* s = a < 0 ? -1 : 0 */ - a = (a ^ s) - s; /* negate if s == -1 */ - tu_int r; - __udivmodti4(a, b, &r); - return ((ti_int)r ^ s) - s; /* negate if s == -1 */ +COMPILER_RT_ABI ti_int __modti3(ti_int a, ti_int b) { + const int bits_in_tword_m1 = (int)(sizeof(ti_int) * CHAR_BIT) - 1; + ti_int s = b >> bits_in_tword_m1; // s = b < 0 ? -1 : 0 + b = (b ^ s) - s; // negate if s == -1 + s = a >> bits_in_tword_m1; // s = a < 0 ? -1 : 0 + a = (a ^ s) - s; // negate if s == -1 + tu_int r; + __udivmodti4(a, b, &r); + return ((ti_int)r ^ s) - s; // negate if s == -1 } -#endif /* CRT_HAS_128BIT */ +#endif // CRT_HAS_128BIT diff --git a/lib/builtins/muldc3.c b/lib/builtins/muldc3.c index 16d8e98390a3..0ea7041c3706 100644 --- a/lib/builtins/muldc3.c +++ b/lib/builtins/muldc3.c @@ -1,73 +1,65 @@ -/* ===-- muldc3.c - Implement __muldc3 -------------------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - * - * This file implements __muldc3 for the compiler_rt library. - * - * ===----------------------------------------------------------------------=== - */ +//===-- muldc3.c - Implement __muldc3 -------------------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements __muldc3 for the compiler_rt library. +// +//===----------------------------------------------------------------------===// #include "int_lib.h" #include "int_math.h" -/* Returns: the product of a + ib and c + id */ +// Returns: the product of a + ib and c + id -COMPILER_RT_ABI Dcomplex -__muldc3(double __a, double __b, double __c, double __d) -{ - double __ac = __a * __c; - double __bd = __b * __d; - double __ad = __a * __d; - double __bc = __b * __c; - Dcomplex z; - COMPLEX_REAL(z) = __ac - __bd; - COMPLEX_IMAGINARY(z) = __ad + __bc; - if (crt_isnan(COMPLEX_REAL(z)) && crt_isnan(COMPLEX_IMAGINARY(z))) - { - int __recalc = 0; - if (crt_isinf(__a) || crt_isinf(__b)) - { - __a = crt_copysign(crt_isinf(__a) ? 1 : 0, __a); - __b = crt_copysign(crt_isinf(__b) ? 1 : 0, __b); - if (crt_isnan(__c)) - __c = crt_copysign(0, __c); - if (crt_isnan(__d)) - __d = crt_copysign(0, __d); - __recalc = 1; - } - if (crt_isinf(__c) || crt_isinf(__d)) - { - __c = crt_copysign(crt_isinf(__c) ? 1 : 0, __c); - __d = crt_copysign(crt_isinf(__d) ? 1 : 0, __d); - if (crt_isnan(__a)) - __a = crt_copysign(0, __a); - if (crt_isnan(__b)) - __b = crt_copysign(0, __b); - __recalc = 1; - } - if (!__recalc && (crt_isinf(__ac) || crt_isinf(__bd) || - crt_isinf(__ad) || crt_isinf(__bc))) - { - if (crt_isnan(__a)) - __a = crt_copysign(0, __a); - if (crt_isnan(__b)) - __b = crt_copysign(0, __b); - if (crt_isnan(__c)) - __c = crt_copysign(0, __c); - if (crt_isnan(__d)) - __d = crt_copysign(0, __d); - __recalc = 1; - } - if (__recalc) - { - COMPLEX_REAL(z) = CRT_INFINITY * (__a * __c - __b * __d); - COMPLEX_IMAGINARY(z) = CRT_INFINITY * (__a * __d + __b * __c); - } +COMPILER_RT_ABI Dcomplex __muldc3(double __a, double __b, double __c, + double __d) { + double __ac = __a * __c; + double __bd = __b * __d; + double __ad = __a * __d; + double __bc = __b * __c; + Dcomplex z; + COMPLEX_REAL(z) = __ac - __bd; + COMPLEX_IMAGINARY(z) = __ad + __bc; + if (crt_isnan(COMPLEX_REAL(z)) && crt_isnan(COMPLEX_IMAGINARY(z))) { + int __recalc = 0; + if (crt_isinf(__a) || crt_isinf(__b)) { + __a = crt_copysign(crt_isinf(__a) ? 1 : 0, __a); + __b = crt_copysign(crt_isinf(__b) ? 1 : 0, __b); + if (crt_isnan(__c)) + __c = crt_copysign(0, __c); + if (crt_isnan(__d)) + __d = crt_copysign(0, __d); + __recalc = 1; } - return z; + if (crt_isinf(__c) || crt_isinf(__d)) { + __c = crt_copysign(crt_isinf(__c) ? 1 : 0, __c); + __d = crt_copysign(crt_isinf(__d) ? 1 : 0, __d); + if (crt_isnan(__a)) + __a = crt_copysign(0, __a); + if (crt_isnan(__b)) + __b = crt_copysign(0, __b); + __recalc = 1; + } + if (!__recalc && (crt_isinf(__ac) || crt_isinf(__bd) || crt_isinf(__ad) || + crt_isinf(__bc))) { + if (crt_isnan(__a)) + __a = crt_copysign(0, __a); + if (crt_isnan(__b)) + __b = crt_copysign(0, __b); + if (crt_isnan(__c)) + __c = crt_copysign(0, __c); + if (crt_isnan(__d)) + __d = crt_copysign(0, __d); + __recalc = 1; + } + if (__recalc) { + COMPLEX_REAL(z) = CRT_INFINITY * (__a * __c - __b * __d); + COMPLEX_IMAGINARY(z) = CRT_INFINITY * (__a * __d + __b * __c); + } + } + return z; } diff --git a/lib/builtins/muldf3.c b/lib/builtins/muldf3.c index 1bb103e38c13..f64b5228d7b7 100644 --- a/lib/builtins/muldf3.c +++ b/lib/builtins/muldf3.c @@ -1,9 +1,8 @@ //===-- lib/muldf3.c - Double-precision multiplication ------------*- C -*-===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// // @@ -15,16 +14,12 @@ #define DOUBLE_PRECISION #include "fp_mul_impl.inc" -COMPILER_RT_ABI fp_t __muldf3(fp_t a, fp_t b) { - return __mulXf3__(a, b); -} +COMPILER_RT_ABI fp_t __muldf3(fp_t a, fp_t b) { return __mulXf3__(a, b); } #if defined(__ARM_EABI__) #if defined(COMPILER_RT_ARMHF_TARGET) -AEABI_RTABI fp_t __aeabi_dmul(fp_t a, fp_t b) { - return __muldf3(a, b); -} +AEABI_RTABI fp_t __aeabi_dmul(fp_t a, fp_t b) { return __muldf3(a, b); } #else -AEABI_RTABI fp_t __aeabi_dmul(fp_t a, fp_t b) COMPILER_RT_ALIAS(__muldf3); +COMPILER_RT_ALIAS(__muldf3, __aeabi_dmul) #endif #endif diff --git a/lib/builtins/muldi3.c b/lib/builtins/muldi3.c index a187315e9165..013f669a1833 100644 --- a/lib/builtins/muldi3.c +++ b/lib/builtins/muldi3.c @@ -1,58 +1,51 @@ -/* ===-- muldi3.c - Implement __muldi3 -------------------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - * - * This file implements __muldi3 for the compiler_rt library. - * - * ===----------------------------------------------------------------------=== - */ +//===-- muldi3.c - Implement __muldi3 -------------------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements __muldi3 for the compiler_rt library. +// +//===----------------------------------------------------------------------===// #include "int_lib.h" -/* Returns: a * b */ +// Returns: a * b -static -di_int -__muldsi3(su_int a, su_int b) -{ - dwords r; - const int bits_in_word_2 = (int)(sizeof(si_int) * CHAR_BIT) / 2; - const su_int lower_mask = (su_int)~0 >> bits_in_word_2; - r.s.low = (a & lower_mask) * (b & lower_mask); - su_int t = r.s.low >> bits_in_word_2; - r.s.low &= lower_mask; - t += (a >> bits_in_word_2) * (b & lower_mask); - r.s.low += (t & lower_mask) << bits_in_word_2; - r.s.high = t >> bits_in_word_2; - t = r.s.low >> bits_in_word_2; - r.s.low &= lower_mask; - t += (b >> bits_in_word_2) * (a & lower_mask); - r.s.low += (t & lower_mask) << bits_in_word_2; - r.s.high += t >> bits_in_word_2; - r.s.high += (a >> bits_in_word_2) * (b >> bits_in_word_2); - return r.all; +static di_int __muldsi3(su_int a, su_int b) { + dwords r; + const int bits_in_word_2 = (int)(sizeof(si_int) * CHAR_BIT) / 2; + const su_int lower_mask = (su_int)~0 >> bits_in_word_2; + r.s.low = (a & lower_mask) * (b & lower_mask); + su_int t = r.s.low >> bits_in_word_2; + r.s.low &= lower_mask; + t += (a >> bits_in_word_2) * (b & lower_mask); + r.s.low += (t & lower_mask) << bits_in_word_2; + r.s.high = t >> bits_in_word_2; + t = r.s.low >> bits_in_word_2; + r.s.low &= lower_mask; + t += (b >> bits_in_word_2) * (a & lower_mask); + r.s.low += (t & lower_mask) << bits_in_word_2; + r.s.high += t >> bits_in_word_2; + r.s.high += (a >> bits_in_word_2) * (b >> bits_in_word_2); + return r.all; } -/* Returns: a * b */ +// Returns: a * b -COMPILER_RT_ABI di_int -__muldi3(di_int a, di_int b) -{ - dwords x; - x.all = a; - dwords y; - y.all = b; - dwords r; - r.all = __muldsi3(x.s.low, y.s.low); - r.s.high += x.s.high * y.s.low + x.s.low * y.s.high; - return r.all; +COMPILER_RT_ABI di_int __muldi3(di_int a, di_int b) { + dwords x; + x.all = a; + dwords y; + y.all = b; + dwords r; + r.all = __muldsi3(x.s.low, y.s.low); + r.s.high += x.s.high * y.s.low + x.s.low * y.s.high; + return r.all; } #if defined(__ARM_EABI__) -AEABI_RTABI di_int __aeabi_lmul(di_int a, di_int b) COMPILER_RT_ALIAS(__muldi3); +COMPILER_RT_ALIAS(__muldi3, __aeabi_lmul) #endif diff --git a/lib/builtins/mulodi4.c b/lib/builtins/mulodi4.c index d2fd7db2bcd3..23f5571ac468 100644 --- a/lib/builtins/mulodi4.c +++ b/lib/builtins/mulodi4.c @@ -1,58 +1,49 @@ -/*===-- mulodi4.c - Implement __mulodi4 -----------------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - * - * This file implements __mulodi4 for the compiler_rt library. - * - * ===----------------------------------------------------------------------=== - */ +//===-- mulodi4.c - Implement __mulodi4 -----------------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements __mulodi4 for the compiler_rt library. +// +//===----------------------------------------------------------------------===// #include "int_lib.h" -/* Returns: a * b */ +// Returns: a * b -/* Effects: sets *overflow to 1 if a * b overflows */ +// Effects: sets *overflow to 1 if a * b overflows -COMPILER_RT_ABI di_int -__mulodi4(di_int a, di_int b, int* overflow) -{ - const int N = (int)(sizeof(di_int) * CHAR_BIT); - const di_int MIN = (di_int)1 << (N-1); - const di_int MAX = ~MIN; - *overflow = 0; - di_int result = a * b; - if (a == MIN) - { - if (b != 0 && b != 1) - *overflow = 1; - return result; - } - if (b == MIN) - { - if (a != 0 && a != 1) - *overflow = 1; - return result; - } - di_int sa = a >> (N - 1); - di_int abs_a = (a ^ sa) - sa; - di_int sb = b >> (N - 1); - di_int abs_b = (b ^ sb) - sb; - if (abs_a < 2 || abs_b < 2) - return result; - if (sa == sb) - { - if (abs_a > MAX / abs_b) - *overflow = 1; - } - else - { - if (abs_a > MIN / -abs_b) - *overflow = 1; - } +COMPILER_RT_ABI di_int __mulodi4(di_int a, di_int b, int *overflow) { + const int N = (int)(sizeof(di_int) * CHAR_BIT); + const di_int MIN = (di_int)1 << (N - 1); + const di_int MAX = ~MIN; + *overflow = 0; + di_int result = a * b; + if (a == MIN) { + if (b != 0 && b != 1) + *overflow = 1; return result; + } + if (b == MIN) { + if (a != 0 && a != 1) + *overflow = 1; + return result; + } + di_int sa = a >> (N - 1); + di_int abs_a = (a ^ sa) - sa; + di_int sb = b >> (N - 1); + di_int abs_b = (b ^ sb) - sb; + if (abs_a < 2 || abs_b < 2) + return result; + if (sa == sb) { + if (abs_a > MAX / abs_b) + *overflow = 1; + } else { + if (abs_a > MIN / -abs_b) + *overflow = 1; + } + return result; } diff --git a/lib/builtins/mulosi4.c b/lib/builtins/mulosi4.c index 422528085c44..fea4311296f8 100644 --- a/lib/builtins/mulosi4.c +++ b/lib/builtins/mulosi4.c @@ -1,58 +1,49 @@ -/*===-- mulosi4.c - Implement __mulosi4 -----------------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - * - * This file implements __mulosi4 for the compiler_rt library. - * - * ===----------------------------------------------------------------------=== - */ +//===-- mulosi4.c - Implement __mulosi4 -----------------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements __mulosi4 for the compiler_rt library. +// +//===----------------------------------------------------------------------===// #include "int_lib.h" -/* Returns: a * b */ +// Returns: a * b -/* Effects: sets *overflow to 1 if a * b overflows */ +// Effects: sets *overflow to 1 if a * b overflows -COMPILER_RT_ABI si_int -__mulosi4(si_int a, si_int b, int* overflow) -{ - const int N = (int)(sizeof(si_int) * CHAR_BIT); - const si_int MIN = (si_int)1 << (N-1); - const si_int MAX = ~MIN; - *overflow = 0; - si_int result = a * b; - if (a == MIN) - { - if (b != 0 && b != 1) - *overflow = 1; - return result; - } - if (b == MIN) - { - if (a != 0 && a != 1) - *overflow = 1; - return result; - } - si_int sa = a >> (N - 1); - si_int abs_a = (a ^ sa) - sa; - si_int sb = b >> (N - 1); - si_int abs_b = (b ^ sb) - sb; - if (abs_a < 2 || abs_b < 2) - return result; - if (sa == sb) - { - if (abs_a > MAX / abs_b) - *overflow = 1; - } - else - { - if (abs_a > MIN / -abs_b) - *overflow = 1; - } +COMPILER_RT_ABI si_int __mulosi4(si_int a, si_int b, int *overflow) { + const int N = (int)(sizeof(si_int) * CHAR_BIT); + const si_int MIN = (si_int)1 << (N - 1); + const si_int MAX = ~MIN; + *overflow = 0; + si_int result = a * b; + if (a == MIN) { + if (b != 0 && b != 1) + *overflow = 1; return result; + } + if (b == MIN) { + if (a != 0 && a != 1) + *overflow = 1; + return result; + } + si_int sa = a >> (N - 1); + si_int abs_a = (a ^ sa) - sa; + si_int sb = b >> (N - 1); + si_int abs_b = (b ^ sb) - sb; + if (abs_a < 2 || abs_b < 2) + return result; + if (sa == sb) { + if (abs_a > MAX / abs_b) + *overflow = 1; + } else { + if (abs_a > MIN / -abs_b) + *overflow = 1; + } + return result; } diff --git a/lib/builtins/muloti4.c b/lib/builtins/muloti4.c index 16b218920321..9bdd5b649908 100644 --- a/lib/builtins/muloti4.c +++ b/lib/builtins/muloti4.c @@ -1,62 +1,53 @@ -/*===-- muloti4.c - Implement __muloti4 -----------------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - * - * This file implements __muloti4 for the compiler_rt library. - * - * ===----------------------------------------------------------------------=== - */ +//===-- muloti4.c - Implement __muloti4 -----------------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements __muloti4 for the compiler_rt library. +// +//===----------------------------------------------------------------------===// #include "int_lib.h" #ifdef CRT_HAS_128BIT -/* Returns: a * b */ +// Returns: a * b -/* Effects: sets *overflow to 1 if a * b overflows */ +// Effects: sets *overflow to 1 if a * b overflows -COMPILER_RT_ABI ti_int -__muloti4(ti_int a, ti_int b, int* overflow) -{ - const int N = (int)(sizeof(ti_int) * CHAR_BIT); - const ti_int MIN = (ti_int)1 << (N-1); - const ti_int MAX = ~MIN; - *overflow = 0; - ti_int result = a * b; - if (a == MIN) - { - if (b != 0 && b != 1) - *overflow = 1; - return result; - } - if (b == MIN) - { - if (a != 0 && a != 1) - *overflow = 1; - return result; - } - ti_int sa = a >> (N - 1); - ti_int abs_a = (a ^ sa) - sa; - ti_int sb = b >> (N - 1); - ti_int abs_b = (b ^ sb) - sb; - if (abs_a < 2 || abs_b < 2) - return result; - if (sa == sb) - { - if (abs_a > MAX / abs_b) - *overflow = 1; - } - else - { - if (abs_a > MIN / -abs_b) - *overflow = 1; - } +COMPILER_RT_ABI ti_int __muloti4(ti_int a, ti_int b, int *overflow) { + const int N = (int)(sizeof(ti_int) * CHAR_BIT); + const ti_int MIN = (ti_int)1 << (N - 1); + const ti_int MAX = ~MIN; + *overflow = 0; + ti_int result = a * b; + if (a == MIN) { + if (b != 0 && b != 1) + *overflow = 1; return result; + } + if (b == MIN) { + if (a != 0 && a != 1) + *overflow = 1; + return result; + } + ti_int sa = a >> (N - 1); + ti_int abs_a = (a ^ sa) - sa; + ti_int sb = b >> (N - 1); + ti_int abs_b = (b ^ sb) - sb; + if (abs_a < 2 || abs_b < 2) + return result; + if (sa == sb) { + if (abs_a > MAX / abs_b) + *overflow = 1; + } else { + if (abs_a > MIN / -abs_b) + *overflow = 1; + } + return result; } -#endif /* CRT_HAS_128BIT */ +#endif // CRT_HAS_128BIT diff --git a/lib/builtins/mulsc3.c b/lib/builtins/mulsc3.c index c89cfd247a15..60653174e440 100644 --- a/lib/builtins/mulsc3.c +++ b/lib/builtins/mulsc3.c @@ -1,73 +1,64 @@ -/* ===-- mulsc3.c - Implement __mulsc3 -------------------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - * - * This file implements __mulsc3 for the compiler_rt library. - * - * ===----------------------------------------------------------------------=== - */ +//===-- mulsc3.c - Implement __mulsc3 -------------------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements __mulsc3 for the compiler_rt library. +// +//===----------------------------------------------------------------------===// #include "int_lib.h" #include "int_math.h" -/* Returns: the product of a + ib and c + id */ +// Returns: the product of a + ib and c + id -COMPILER_RT_ABI Fcomplex -__mulsc3(float __a, float __b, float __c, float __d) -{ - float __ac = __a * __c; - float __bd = __b * __d; - float __ad = __a * __d; - float __bc = __b * __c; - Fcomplex z; - COMPLEX_REAL(z) = __ac - __bd; - COMPLEX_IMAGINARY(z) = __ad + __bc; - if (crt_isnan(COMPLEX_REAL(z)) && crt_isnan(COMPLEX_IMAGINARY(z))) - { - int __recalc = 0; - if (crt_isinf(__a) || crt_isinf(__b)) - { - __a = crt_copysignf(crt_isinf(__a) ? 1 : 0, __a); - __b = crt_copysignf(crt_isinf(__b) ? 1 : 0, __b); - if (crt_isnan(__c)) - __c = crt_copysignf(0, __c); - if (crt_isnan(__d)) - __d = crt_copysignf(0, __d); - __recalc = 1; - } - if (crt_isinf(__c) || crt_isinf(__d)) - { - __c = crt_copysignf(crt_isinf(__c) ? 1 : 0, __c); - __d = crt_copysignf(crt_isinf(__d) ? 1 : 0, __d); - if (crt_isnan(__a)) - __a = crt_copysignf(0, __a); - if (crt_isnan(__b)) - __b = crt_copysignf(0, __b); - __recalc = 1; - } - if (!__recalc && (crt_isinf(__ac) || crt_isinf(__bd) || - crt_isinf(__ad) || crt_isinf(__bc))) - { - if (crt_isnan(__a)) - __a = crt_copysignf(0, __a); - if (crt_isnan(__b)) - __b = crt_copysignf(0, __b); - if (crt_isnan(__c)) - __c = crt_copysignf(0, __c); - if (crt_isnan(__d)) - __d = crt_copysignf(0, __d); - __recalc = 1; - } - if (__recalc) - { - COMPLEX_REAL(z) = CRT_INFINITY * (__a * __c - __b * __d); - COMPLEX_IMAGINARY(z) = CRT_INFINITY * (__a * __d + __b * __c); - } +COMPILER_RT_ABI Fcomplex __mulsc3(float __a, float __b, float __c, float __d) { + float __ac = __a * __c; + float __bd = __b * __d; + float __ad = __a * __d; + float __bc = __b * __c; + Fcomplex z; + COMPLEX_REAL(z) = __ac - __bd; + COMPLEX_IMAGINARY(z) = __ad + __bc; + if (crt_isnan(COMPLEX_REAL(z)) && crt_isnan(COMPLEX_IMAGINARY(z))) { + int __recalc = 0; + if (crt_isinf(__a) || crt_isinf(__b)) { + __a = crt_copysignf(crt_isinf(__a) ? 1 : 0, __a); + __b = crt_copysignf(crt_isinf(__b) ? 1 : 0, __b); + if (crt_isnan(__c)) + __c = crt_copysignf(0, __c); + if (crt_isnan(__d)) + __d = crt_copysignf(0, __d); + __recalc = 1; } - return z; + if (crt_isinf(__c) || crt_isinf(__d)) { + __c = crt_copysignf(crt_isinf(__c) ? 1 : 0, __c); + __d = crt_copysignf(crt_isinf(__d) ? 1 : 0, __d); + if (crt_isnan(__a)) + __a = crt_copysignf(0, __a); + if (crt_isnan(__b)) + __b = crt_copysignf(0, __b); + __recalc = 1; + } + if (!__recalc && (crt_isinf(__ac) || crt_isinf(__bd) || crt_isinf(__ad) || + crt_isinf(__bc))) { + if (crt_isnan(__a)) + __a = crt_copysignf(0, __a); + if (crt_isnan(__b)) + __b = crt_copysignf(0, __b); + if (crt_isnan(__c)) + __c = crt_copysignf(0, __c); + if (crt_isnan(__d)) + __d = crt_copysignf(0, __d); + __recalc = 1; + } + if (__recalc) { + COMPLEX_REAL(z) = CRT_INFINITY * (__a * __c - __b * __d); + COMPLEX_IMAGINARY(z) = CRT_INFINITY * (__a * __d + __b * __c); + } + } + return z; } diff --git a/lib/builtins/mulsf3.c b/lib/builtins/mulsf3.c index 1e2cf3e717c9..b9cf39abc784 100644 --- a/lib/builtins/mulsf3.c +++ b/lib/builtins/mulsf3.c @@ -1,9 +1,8 @@ //===-- lib/mulsf3.c - Single-precision multiplication ------------*- C -*-===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// // @@ -15,16 +14,12 @@ #define SINGLE_PRECISION #include "fp_mul_impl.inc" -COMPILER_RT_ABI fp_t __mulsf3(fp_t a, fp_t b) { - return __mulXf3__(a, b); -} +COMPILER_RT_ABI fp_t __mulsf3(fp_t a, fp_t b) { return __mulXf3__(a, b); } #if defined(__ARM_EABI__) #if defined(COMPILER_RT_ARMHF_TARGET) -AEABI_RTABI fp_t __aeabi_fmul(fp_t a, fp_t b) { - return __mulsf3(a, b); -} +AEABI_RTABI fp_t __aeabi_fmul(fp_t a, fp_t b) { return __mulsf3(a, b); } #else -AEABI_RTABI fp_t __aeabi_fmul(fp_t a, fp_t b) COMPILER_RT_ALIAS(__mulsf3); +COMPILER_RT_ALIAS(__mulsf3, __aeabi_fmul) #endif #endif diff --git a/lib/builtins/multc3.c b/lib/builtins/multc3.c index 0518bc2569f1..bb7f6aabfe2c 100644 --- a/lib/builtins/multc3.c +++ b/lib/builtins/multc3.c @@ -1,68 +1,65 @@ -/* ===-- multc3.c - Implement __multc3 -------------------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - * - * This file implements __multc3 for the compiler_rt library. - * - * ===----------------------------------------------------------------------=== - */ +//===-- multc3.c - Implement __multc3 -------------------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements __multc3 for the compiler_rt library. +// +//===----------------------------------------------------------------------===// #include "int_lib.h" #include "int_math.h" -/* Returns: the product of a + ib and c + id */ +// Returns: the product of a + ib and c + id -COMPILER_RT_ABI long double _Complex -__multc3(long double a, long double b, long double c, long double d) -{ - long double ac = a * c; - long double bd = b * d; - long double ad = a * d; - long double bc = b * c; - long double _Complex z; - __real__ z = ac - bd; - __imag__ z = ad + bc; - if (crt_isnan(__real__ z) && crt_isnan(__imag__ z)) { - int recalc = 0; - if (crt_isinf(a) || crt_isinf(b)) { - a = crt_copysignl(crt_isinf(a) ? 1 : 0, a); - b = crt_copysignl(crt_isinf(b) ? 1 : 0, b); - if (crt_isnan(c)) - c = crt_copysignl(0, c); - if (crt_isnan(d)) - d = crt_copysignl(0, d); - recalc = 1; - } - if (crt_isinf(c) || crt_isinf(d)) { - c = crt_copysignl(crt_isinf(c) ? 1 : 0, c); - d = crt_copysignl(crt_isinf(d) ? 1 : 0, d); - if (crt_isnan(a)) - a = crt_copysignl(0, a); - if (crt_isnan(b)) - b = crt_copysignl(0, b); - recalc = 1; - } - if (!recalc && (crt_isinf(ac) || crt_isinf(bd) || - crt_isinf(ad) || crt_isinf(bc))) { - if (crt_isnan(a)) - a = crt_copysignl(0, a); - if (crt_isnan(b)) - b = crt_copysignl(0, b); - if (crt_isnan(c)) - c = crt_copysignl(0, c); - if (crt_isnan(d)) - d = crt_copysignl(0, d); - recalc = 1; - } - if (recalc) { - __real__ z = CRT_INFINITY * (a * c - b * d); - __imag__ z = CRT_INFINITY * (a * d + b * c); - } +COMPILER_RT_ABI long double _Complex __multc3(long double a, long double b, + long double c, long double d) { + long double ac = a * c; + long double bd = b * d; + long double ad = a * d; + long double bc = b * c; + long double _Complex z; + __real__ z = ac - bd; + __imag__ z = ad + bc; + if (crt_isnan(__real__ z) && crt_isnan(__imag__ z)) { + int recalc = 0; + if (crt_isinf(a) || crt_isinf(b)) { + a = crt_copysignl(crt_isinf(a) ? 1 : 0, a); + b = crt_copysignl(crt_isinf(b) ? 1 : 0, b); + if (crt_isnan(c)) + c = crt_copysignl(0, c); + if (crt_isnan(d)) + d = crt_copysignl(0, d); + recalc = 1; } - return z; + if (crt_isinf(c) || crt_isinf(d)) { + c = crt_copysignl(crt_isinf(c) ? 1 : 0, c); + d = crt_copysignl(crt_isinf(d) ? 1 : 0, d); + if (crt_isnan(a)) + a = crt_copysignl(0, a); + if (crt_isnan(b)) + b = crt_copysignl(0, b); + recalc = 1; + } + if (!recalc && + (crt_isinf(ac) || crt_isinf(bd) || crt_isinf(ad) || crt_isinf(bc))) { + if (crt_isnan(a)) + a = crt_copysignl(0, a); + if (crt_isnan(b)) + b = crt_copysignl(0, b); + if (crt_isnan(c)) + c = crt_copysignl(0, c); + if (crt_isnan(d)) + d = crt_copysignl(0, d); + recalc = 1; + } + if (recalc) { + __real__ z = CRT_INFINITY * (a * c - b * d); + __imag__ z = CRT_INFINITY * (a * d + b * c); + } + } + return z; } diff --git a/lib/builtins/multf3.c b/lib/builtins/multf3.c index 0b915923ea09..0626fb8c7fc9 100644 --- a/lib/builtins/multf3.c +++ b/lib/builtins/multf3.c @@ -1,9 +1,8 @@ //===-- lib/multf3.c - Quad-precision multiplication --------------*- C -*-===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// // @@ -18,8 +17,6 @@ #if defined(CRT_HAS_128BIT) && defined(CRT_LDBL_128BIT) #include "fp_mul_impl.inc" -COMPILER_RT_ABI fp_t __multf3(fp_t a, fp_t b) { - return __mulXf3__(a, b); -} +COMPILER_RT_ABI fp_t __multf3(fp_t a, fp_t b) { return __mulXf3__(a, b); } #endif diff --git a/lib/builtins/multi3.c b/lib/builtins/multi3.c index e0d52d430b66..d9d8b59cd3ef 100644 --- a/lib/builtins/multi3.c +++ b/lib/builtins/multi3.c @@ -1,58 +1,51 @@ -/* ===-- multi3.c - Implement __multi3 -------------------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - - * This file implements __multi3 for the compiler_rt library. - * - * ===----------------------------------------------------------------------=== - */ +//===-- multi3.c - Implement __multi3 -------------------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements __multi3 for the compiler_rt library. +// +//===----------------------------------------------------------------------===// #include "int_lib.h" #ifdef CRT_HAS_128BIT -/* Returns: a * b */ - -static -ti_int -__mulddi3(du_int a, du_int b) -{ - twords r; - const int bits_in_dword_2 = (int)(sizeof(di_int) * CHAR_BIT) / 2; - const du_int lower_mask = (du_int)~0 >> bits_in_dword_2; - r.s.low = (a & lower_mask) * (b & lower_mask); - du_int t = r.s.low >> bits_in_dword_2; - r.s.low &= lower_mask; - t += (a >> bits_in_dword_2) * (b & lower_mask); - r.s.low += (t & lower_mask) << bits_in_dword_2; - r.s.high = t >> bits_in_dword_2; - t = r.s.low >> bits_in_dword_2; - r.s.low &= lower_mask; - t += (b >> bits_in_dword_2) * (a & lower_mask); - r.s.low += (t & lower_mask) << bits_in_dword_2; - r.s.high += t >> bits_in_dword_2; - r.s.high += (a >> bits_in_dword_2) * (b >> bits_in_dword_2); - return r.all; +// Returns: a * b + +static ti_int __mulddi3(du_int a, du_int b) { + twords r; + const int bits_in_dword_2 = (int)(sizeof(di_int) * CHAR_BIT) / 2; + const du_int lower_mask = (du_int)~0 >> bits_in_dword_2; + r.s.low = (a & lower_mask) * (b & lower_mask); + du_int t = r.s.low >> bits_in_dword_2; + r.s.low &= lower_mask; + t += (a >> bits_in_dword_2) * (b & lower_mask); + r.s.low += (t & lower_mask) << bits_in_dword_2; + r.s.high = t >> bits_in_dword_2; + t = r.s.low >> bits_in_dword_2; + r.s.low &= lower_mask; + t += (b >> bits_in_dword_2) * (a & lower_mask); + r.s.low += (t & lower_mask) << bits_in_dword_2; + r.s.high += t >> bits_in_dword_2; + r.s.high += (a >> bits_in_dword_2) * (b >> bits_in_dword_2); + return r.all; } -/* Returns: a * b */ - -COMPILER_RT_ABI ti_int -__multi3(ti_int a, ti_int b) -{ - twords x; - x.all = a; - twords y; - y.all = b; - twords r; - r.all = __mulddi3(x.s.low, y.s.low); - r.s.high += x.s.high * y.s.low + x.s.low * y.s.high; - return r.all; +// Returns: a * b + +COMPILER_RT_ABI ti_int __multi3(ti_int a, ti_int b) { + twords x; + x.all = a; + twords y; + y.all = b; + twords r; + r.all = __mulddi3(x.s.low, y.s.low); + r.s.high += x.s.high * y.s.low + x.s.low * y.s.high; + return r.all; } -#endif /* CRT_HAS_128BIT */ +#endif // CRT_HAS_128BIT diff --git a/lib/builtins/mulvdi3.c b/lib/builtins/mulvdi3.c index e63249e0a04c..cecc97ccf22e 100644 --- a/lib/builtins/mulvdi3.c +++ b/lib/builtins/mulvdi3.c @@ -1,56 +1,47 @@ -/*===-- mulvdi3.c - Implement __mulvdi3 -----------------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - * - * This file implements __mulvdi3 for the compiler_rt library. - * - * ===----------------------------------------------------------------------=== - */ +//===-- mulvdi3.c - Implement __mulvdi3 -----------------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements __mulvdi3 for the compiler_rt library. +// +//===----------------------------------------------------------------------===// #include "int_lib.h" -/* Returns: a * b */ +// Returns: a * b -/* Effects: aborts if a * b overflows */ +// Effects: aborts if a * b overflows -COMPILER_RT_ABI di_int -__mulvdi3(di_int a, di_int b) -{ - const int N = (int)(sizeof(di_int) * CHAR_BIT); - const di_int MIN = (di_int)1 << (N-1); - const di_int MAX = ~MIN; - if (a == MIN) - { - if (b == 0 || b == 1) - return a * b; - compilerrt_abort(); - } - if (b == MIN) - { - if (a == 0 || a == 1) - return a * b; - compilerrt_abort(); - } - di_int sa = a >> (N - 1); - di_int abs_a = (a ^ sa) - sa; - di_int sb = b >> (N - 1); - di_int abs_b = (b ^ sb) - sb; - if (abs_a < 2 || abs_b < 2) - return a * b; - if (sa == sb) - { - if (abs_a > MAX / abs_b) - compilerrt_abort(); - } - else - { - if (abs_a > MIN / -abs_b) - compilerrt_abort(); - } +COMPILER_RT_ABI di_int __mulvdi3(di_int a, di_int b) { + const int N = (int)(sizeof(di_int) * CHAR_BIT); + const di_int MIN = (di_int)1 << (N - 1); + const di_int MAX = ~MIN; + if (a == MIN) { + if (b == 0 || b == 1) + return a * b; + compilerrt_abort(); + } + if (b == MIN) { + if (a == 0 || a == 1) + return a * b; + compilerrt_abort(); + } + di_int sa = a >> (N - 1); + di_int abs_a = (a ^ sa) - sa; + di_int sb = b >> (N - 1); + di_int abs_b = (b ^ sb) - sb; + if (abs_a < 2 || abs_b < 2) return a * b; + if (sa == sb) { + if (abs_a > MAX / abs_b) + compilerrt_abort(); + } else { + if (abs_a > MIN / -abs_b) + compilerrt_abort(); + } + return a * b; } diff --git a/lib/builtins/mulvsi3.c b/lib/builtins/mulvsi3.c index 74ea4f2da226..0d6b18ad01a4 100644 --- a/lib/builtins/mulvsi3.c +++ b/lib/builtins/mulvsi3.c @@ -1,56 +1,47 @@ -/* ===-- mulvsi3.c - Implement __mulvsi3 -----------------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - * - * This file implements __mulvsi3 for the compiler_rt library. - * - * ===----------------------------------------------------------------------=== - */ +//===-- mulvsi3.c - Implement __mulvsi3 -----------------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements __mulvsi3 for the compiler_rt library. +// +//===----------------------------------------------------------------------===// #include "int_lib.h" -/* Returns: a * b */ +// Returns: a * b -/* Effects: aborts if a * b overflows */ +// Effects: aborts if a * b overflows -COMPILER_RT_ABI si_int -__mulvsi3(si_int a, si_int b) -{ - const int N = (int)(sizeof(si_int) * CHAR_BIT); - const si_int MIN = (si_int)1 << (N-1); - const si_int MAX = ~MIN; - if (a == MIN) - { - if (b == 0 || b == 1) - return a * b; - compilerrt_abort(); - } - if (b == MIN) - { - if (a == 0 || a == 1) - return a * b; - compilerrt_abort(); - } - si_int sa = a >> (N - 1); - si_int abs_a = (a ^ sa) - sa; - si_int sb = b >> (N - 1); - si_int abs_b = (b ^ sb) - sb; - if (abs_a < 2 || abs_b < 2) - return a * b; - if (sa == sb) - { - if (abs_a > MAX / abs_b) - compilerrt_abort(); - } - else - { - if (abs_a > MIN / -abs_b) - compilerrt_abort(); - } +COMPILER_RT_ABI si_int __mulvsi3(si_int a, si_int b) { + const int N = (int)(sizeof(si_int) * CHAR_BIT); + const si_int MIN = (si_int)1 << (N - 1); + const si_int MAX = ~MIN; + if (a == MIN) { + if (b == 0 || b == 1) + return a * b; + compilerrt_abort(); + } + if (b == MIN) { + if (a == 0 || a == 1) + return a * b; + compilerrt_abort(); + } + si_int sa = a >> (N - 1); + si_int abs_a = (a ^ sa) - sa; + si_int sb = b >> (N - 1); + si_int abs_b = (b ^ sb) - sb; + if (abs_a < 2 || abs_b < 2) return a * b; + if (sa == sb) { + if (abs_a > MAX / abs_b) + compilerrt_abort(); + } else { + if (abs_a > MIN / -abs_b) + compilerrt_abort(); + } + return a * b; } diff --git a/lib/builtins/mulvti3.c b/lib/builtins/mulvti3.c index f4c7d1612ba9..03963a0ca694 100644 --- a/lib/builtins/mulvti3.c +++ b/lib/builtins/mulvti3.c @@ -1,60 +1,51 @@ -/* ===-- mulvti3.c - Implement __mulvti3 -----------------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - * - * This file implements __mulvti3 for the compiler_rt library. - * - * ===----------------------------------------------------------------------=== - */ +//===-- mulvti3.c - Implement __mulvti3 -----------------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements __mulvti3 for the compiler_rt library. +// +//===----------------------------------------------------------------------===// #include "int_lib.h" #ifdef CRT_HAS_128BIT -/* Returns: a * b */ +// Returns: a * b -/* Effects: aborts if a * b overflows */ +// Effects: aborts if a * b overflows -COMPILER_RT_ABI ti_int -__mulvti3(ti_int a, ti_int b) -{ - const int N = (int)(sizeof(ti_int) * CHAR_BIT); - const ti_int MIN = (ti_int)1 << (N-1); - const ti_int MAX = ~MIN; - if (a == MIN) - { - if (b == 0 || b == 1) - return a * b; - compilerrt_abort(); - } - if (b == MIN) - { - if (a == 0 || a == 1) - return a * b; - compilerrt_abort(); - } - ti_int sa = a >> (N - 1); - ti_int abs_a = (a ^ sa) - sa; - ti_int sb = b >> (N - 1); - ti_int abs_b = (b ^ sb) - sb; - if (abs_a < 2 || abs_b < 2) - return a * b; - if (sa == sb) - { - if (abs_a > MAX / abs_b) - compilerrt_abort(); - } - else - { - if (abs_a > MIN / -abs_b) - compilerrt_abort(); - } +COMPILER_RT_ABI ti_int __mulvti3(ti_int a, ti_int b) { + const int N = (int)(sizeof(ti_int) * CHAR_BIT); + const ti_int MIN = (ti_int)1 << (N - 1); + const ti_int MAX = ~MIN; + if (a == MIN) { + if (b == 0 || b == 1) + return a * b; + compilerrt_abort(); + } + if (b == MIN) { + if (a == 0 || a == 1) + return a * b; + compilerrt_abort(); + } + ti_int sa = a >> (N - 1); + ti_int abs_a = (a ^ sa) - sa; + ti_int sb = b >> (N - 1); + ti_int abs_b = (b ^ sb) - sb; + if (abs_a < 2 || abs_b < 2) return a * b; + if (sa == sb) { + if (abs_a > MAX / abs_b) + compilerrt_abort(); + } else { + if (abs_a > MIN / -abs_b) + compilerrt_abort(); + } + return a * b; } -#endif /* CRT_HAS_128BIT */ +#endif // CRT_HAS_128BIT diff --git a/lib/builtins/mulxc3.c b/lib/builtins/mulxc3.c index ba3221691821..2f7f14c28453 100644 --- a/lib/builtins/mulxc3.c +++ b/lib/builtins/mulxc3.c @@ -1,77 +1,69 @@ -/* ===-- mulxc3.c - Implement __mulxc3 -------------------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - * - * This file implements __mulxc3 for the compiler_rt library. - * - * ===----------------------------------------------------------------------=== - */ +//===-- mulxc3.c - Implement __mulxc3 -------------------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements __mulxc3 for the compiler_rt library. +// +//===----------------------------------------------------------------------===// #if !_ARCH_PPC #include "int_lib.h" #include "int_math.h" -/* Returns: the product of a + ib and c + id */ +// Returns: the product of a + ib and c + id -COMPILER_RT_ABI Lcomplex -__mulxc3(long double __a, long double __b, long double __c, long double __d) -{ - long double __ac = __a * __c; - long double __bd = __b * __d; - long double __ad = __a * __d; - long double __bc = __b * __c; - Lcomplex z; - COMPLEX_REAL(z) = __ac - __bd; - COMPLEX_IMAGINARY(z) = __ad + __bc; - if (crt_isnan(COMPLEX_REAL(z)) && crt_isnan(COMPLEX_IMAGINARY(z))) - { - int __recalc = 0; - if (crt_isinf(__a) || crt_isinf(__b)) - { - __a = crt_copysignl(crt_isinf(__a) ? 1 : 0, __a); - __b = crt_copysignl(crt_isinf(__b) ? 1 : 0, __b); - if (crt_isnan(__c)) - __c = crt_copysignl(0, __c); - if (crt_isnan(__d)) - __d = crt_copysignl(0, __d); - __recalc = 1; - } - if (crt_isinf(__c) || crt_isinf(__d)) - { - __c = crt_copysignl(crt_isinf(__c) ? 1 : 0, __c); - __d = crt_copysignl(crt_isinf(__d) ? 1 : 0, __d); - if (crt_isnan(__a)) - __a = crt_copysignl(0, __a); - if (crt_isnan(__b)) - __b = crt_copysignl(0, __b); - __recalc = 1; - } - if (!__recalc && (crt_isinf(__ac) || crt_isinf(__bd) || - crt_isinf(__ad) || crt_isinf(__bc))) - { - if (crt_isnan(__a)) - __a = crt_copysignl(0, __a); - if (crt_isnan(__b)) - __b = crt_copysignl(0, __b); - if (crt_isnan(__c)) - __c = crt_copysignl(0, __c); - if (crt_isnan(__d)) - __d = crt_copysignl(0, __d); - __recalc = 1; - } - if (__recalc) - { - COMPLEX_REAL(z) = CRT_INFINITY * (__a * __c - __b * __d); - COMPLEX_IMAGINARY(z) = CRT_INFINITY * (__a * __d + __b * __c); - } +COMPILER_RT_ABI Lcomplex __mulxc3(long double __a, long double __b, + long double __c, long double __d) { + long double __ac = __a * __c; + long double __bd = __b * __d; + long double __ad = __a * __d; + long double __bc = __b * __c; + Lcomplex z; + COMPLEX_REAL(z) = __ac - __bd; + COMPLEX_IMAGINARY(z) = __ad + __bc; + if (crt_isnan(COMPLEX_REAL(z)) && crt_isnan(COMPLEX_IMAGINARY(z))) { + int __recalc = 0; + if (crt_isinf(__a) || crt_isinf(__b)) { + __a = crt_copysignl(crt_isinf(__a) ? 1 : 0, __a); + __b = crt_copysignl(crt_isinf(__b) ? 1 : 0, __b); + if (crt_isnan(__c)) + __c = crt_copysignl(0, __c); + if (crt_isnan(__d)) + __d = crt_copysignl(0, __d); + __recalc = 1; } - return z; + if (crt_isinf(__c) || crt_isinf(__d)) { + __c = crt_copysignl(crt_isinf(__c) ? 1 : 0, __c); + __d = crt_copysignl(crt_isinf(__d) ? 1 : 0, __d); + if (crt_isnan(__a)) + __a = crt_copysignl(0, __a); + if (crt_isnan(__b)) + __b = crt_copysignl(0, __b); + __recalc = 1; + } + if (!__recalc && (crt_isinf(__ac) || crt_isinf(__bd) || crt_isinf(__ad) || + crt_isinf(__bc))) { + if (crt_isnan(__a)) + __a = crt_copysignl(0, __a); + if (crt_isnan(__b)) + __b = crt_copysignl(0, __b); + if (crt_isnan(__c)) + __c = crt_copysignl(0, __c); + if (crt_isnan(__d)) + __d = crt_copysignl(0, __d); + __recalc = 1; + } + if (__recalc) { + COMPLEX_REAL(z) = CRT_INFINITY * (__a * __c - __b * __d); + COMPLEX_IMAGINARY(z) = CRT_INFINITY * (__a * __d + __b * __c); + } + } + return z; } #endif diff --git a/lib/builtins/negdf2.c b/lib/builtins/negdf2.c index f0bfaad24743..f9ceaa374542 100644 --- a/lib/builtins/negdf2.c +++ b/lib/builtins/negdf2.c @@ -1,9 +1,8 @@ //===-- lib/negdf2.c - double-precision negation ------------------*- C -*-===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// // @@ -14,17 +13,12 @@ #define DOUBLE_PRECISION #include "fp_lib.h" -COMPILER_RT_ABI fp_t -__negdf2(fp_t a) { - return fromRep(toRep(a) ^ signBit); -} +COMPILER_RT_ABI fp_t __negdf2(fp_t a) { return fromRep(toRep(a) ^ signBit); } #if defined(__ARM_EABI__) #if defined(COMPILER_RT_ARMHF_TARGET) -AEABI_RTABI fp_t __aeabi_dneg(fp_t a) { - return __negdf2(a); -} +AEABI_RTABI fp_t __aeabi_dneg(fp_t a) { return __negdf2(a); } #else -AEABI_RTABI fp_t __aeabi_dneg(fp_t a) COMPILER_RT_ALIAS(__negdf2); +COMPILER_RT_ALIAS(__negdf2, __aeabi_dneg) #endif #endif diff --git a/lib/builtins/negdi2.c b/lib/builtins/negdi2.c index 3d49ba2899d2..5a525d4b0e55 100644 --- a/lib/builtins/negdi2.c +++ b/lib/builtins/negdi2.c @@ -1,26 +1,21 @@ -/* ===-- negdi2.c - Implement __negdi2 -------------------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - * - * This file implements __negdi2 for the compiler_rt library. - * - * ===----------------------------------------------------------------------=== - */ +//===-- negdi2.c - Implement __negdi2 -------------------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements __negdi2 for the compiler_rt library. +// +//===----------------------------------------------------------------------===// #include "int_lib.h" -/* Returns: -a */ +// Returns: -a -COMPILER_RT_ABI di_int -__negdi2(di_int a) -{ - /* Note: this routine is here for API compatibility; any sane compiler - * should expand it inline. - */ - return -a; +COMPILER_RT_ABI di_int __negdi2(di_int a) { + // Note: this routine is here for API compatibility; any sane compiler + // should expand it inline. + return -a; } diff --git a/lib/builtins/negsf2.c b/lib/builtins/negsf2.c index 05c97d4d5a11..d59dfe7cf91b 100644 --- a/lib/builtins/negsf2.c +++ b/lib/builtins/negsf2.c @@ -1,9 +1,8 @@ //===-- lib/negsf2.c - single-precision negation ------------------*- C -*-===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// // @@ -14,17 +13,12 @@ #define SINGLE_PRECISION #include "fp_lib.h" -COMPILER_RT_ABI fp_t -__negsf2(fp_t a) { - return fromRep(toRep(a) ^ signBit); -} +COMPILER_RT_ABI fp_t __negsf2(fp_t a) { return fromRep(toRep(a) ^ signBit); } #if defined(__ARM_EABI__) #if defined(COMPILER_RT_ARMHF_TARGET) -AEABI_RTABI fp_t __aeabi_fneg(fp_t a) { - return __negsf2(a); -} +AEABI_RTABI fp_t __aeabi_fneg(fp_t a) { return __negsf2(a); } #else -AEABI_RTABI fp_t __aeabi_fneg(fp_t a) COMPILER_RT_ALIAS(__negsf2); +COMPILER_RT_ALIAS(__negsf2, __aeabi_fneg) #endif #endif diff --git a/lib/builtins/negti2.c b/lib/builtins/negti2.c index 9b00b303f856..d52ba4e13a46 100644 --- a/lib/builtins/negti2.c +++ b/lib/builtins/negti2.c @@ -1,30 +1,25 @@ -/* ===-- negti2.c - Implement __negti2 -------------------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - * - * This file implements __negti2 for the compiler_rt library. - * - * ===----------------------------------------------------------------------=== - */ +//===-- negti2.c - Implement __negti2 -------------------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements __negti2 for the compiler_rt library. +// +//===----------------------------------------------------------------------===// #include "int_lib.h" #ifdef CRT_HAS_128BIT -/* Returns: -a */ +// Returns: -a -COMPILER_RT_ABI ti_int -__negti2(ti_int a) -{ - /* Note: this routine is here for API compatibility; any sane compiler - * should expand it inline. - */ - return -a; +COMPILER_RT_ABI ti_int __negti2(ti_int a) { + // Note: this routine is here for API compatibility; any sane compiler + // should expand it inline. + return -a; } -#endif /* CRT_HAS_128BIT */ +#endif // CRT_HAS_128BIT diff --git a/lib/builtins/negvdi2.c b/lib/builtins/negvdi2.c index e336ecf28f0d..5c52b3ec2aa6 100644 --- a/lib/builtins/negvdi2.c +++ b/lib/builtins/negvdi2.c @@ -1,28 +1,24 @@ -/* ===-- negvdi2.c - Implement __negvdi2 -----------------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - * - * This file implements __negvdi2 for the compiler_rt library. - * - * ===----------------------------------------------------------------------=== - */ +//===-- negvdi2.c - Implement __negvdi2 -----------------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements __negvdi2 for the compiler_rt library. +// +//===----------------------------------------------------------------------===// #include "int_lib.h" -/* Returns: -a */ +// Returns: -a -/* Effects: aborts if -a overflows */ +// Effects: aborts if -a overflows -COMPILER_RT_ABI di_int -__negvdi2(di_int a) -{ - const di_int MIN = (di_int)1 << ((int)(sizeof(di_int) * CHAR_BIT)-1); - if (a == MIN) - compilerrt_abort(); - return -a; +COMPILER_RT_ABI di_int __negvdi2(di_int a) { + const di_int MIN = (di_int)1 << ((int)(sizeof(di_int) * CHAR_BIT) - 1); + if (a == MIN) + compilerrt_abort(); + return -a; } diff --git a/lib/builtins/negvsi2.c b/lib/builtins/negvsi2.c index b9e93fef06c5..cccdee6dc5e5 100644 --- a/lib/builtins/negvsi2.c +++ b/lib/builtins/negvsi2.c @@ -1,28 +1,24 @@ -/* ===-- negvsi2.c - Implement __negvsi2 -----------------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - * - * This file implements __negvsi2 for the compiler_rt library. - * - * ===----------------------------------------------------------------------=== - */ +//===-- negvsi2.c - Implement __negvsi2 -----------------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements __negvsi2 for the compiler_rt library. +// +//===----------------------------------------------------------------------===// #include "int_lib.h" -/* Returns: -a */ +// Returns: -a -/* Effects: aborts if -a overflows */ +// Effects: aborts if -a overflows -COMPILER_RT_ABI si_int -__negvsi2(si_int a) -{ - const si_int MIN = (si_int)1 << ((int)(sizeof(si_int) * CHAR_BIT)-1); - if (a == MIN) - compilerrt_abort(); - return -a; +COMPILER_RT_ABI si_int __negvsi2(si_int a) { + const si_int MIN = (si_int)1 << ((int)(sizeof(si_int) * CHAR_BIT) - 1); + if (a == MIN) + compilerrt_abort(); + return -a; } diff --git a/lib/builtins/negvti2.c b/lib/builtins/negvti2.c index 85f9f7d19d9d..8f92e1046d0c 100644 --- a/lib/builtins/negvti2.c +++ b/lib/builtins/negvti2.c @@ -1,32 +1,28 @@ -/*===-- negvti2.c - Implement __negvti2 -----------------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - *===----------------------------------------------------------------------=== - * - *This file implements __negvti2 for the compiler_rt library. - * - *===----------------------------------------------------------------------=== - */ +//===-- negvti2.c - Implement __negvti2 -----------------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements __negvti2 for the compiler_rt library. +// +//===----------------------------------------------------------------------===// #include "int_lib.h" #ifdef CRT_HAS_128BIT -/* Returns: -a */ +// Returns: -a -/* Effects: aborts if -a overflows */ +// Effects: aborts if -a overflows -COMPILER_RT_ABI ti_int -__negvti2(ti_int a) -{ - const ti_int MIN = (ti_int)1 << ((int)(sizeof(ti_int) * CHAR_BIT)-1); - if (a == MIN) - compilerrt_abort(); - return -a; +COMPILER_RT_ABI ti_int __negvti2(ti_int a) { + const ti_int MIN = (ti_int)1 << ((int)(sizeof(ti_int) * CHAR_BIT) - 1); + if (a == MIN) + compilerrt_abort(); + return -a; } -#endif /* CRT_HAS_128BIT */ +#endif // CRT_HAS_128BIT diff --git a/lib/builtins/os_version_check.c b/lib/builtins/os_version_check.c index e0d40edc7e38..3794b979434c 100644 --- a/lib/builtins/os_version_check.c +++ b/lib/builtins/os_version_check.c @@ -1,17 +1,15 @@ -/* ===-- os_version_check.c - OS version checking -------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - * - * This file implements the function __isOSVersionAtLeast, used by - * Objective-C's @available - * - * ===----------------------------------------------------------------------=== - */ +//===-- os_version_check.c - OS version checking -------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements the function __isOSVersionAtLeast, used by +// Objective-C's @available +// +//===----------------------------------------------------------------------===// #ifdef __APPLE__ @@ -23,15 +21,15 @@ #include <stdlib.h> #include <string.h> -/* These three variables hold the host's OS version. */ +// These three variables hold the host's OS version. static int32_t GlobalMajor, GlobalMinor, GlobalSubminor; static dispatch_once_t DispatchOnceCounter; -/* We can't include <CoreFoundation/CoreFoundation.h> directly from here, so - * just forward declare everything that we need from it. */ +// We can't include <CoreFoundation/CoreFoundation.h> directly from here, so +// just forward declare everything that we need from it. typedef const void *CFDataRef, *CFAllocatorRef, *CFPropertyListRef, - *CFStringRef, *CFDictionaryRef, *CFTypeRef, *CFErrorRef; + *CFStringRef, *CFDictionaryRef, *CFTypeRef, *CFErrorRef; #if __LLP64__ typedef unsigned long long CFTypeID; @@ -48,9 +46,9 @@ typedef _Bool Boolean; typedef CFIndex CFPropertyListFormat; typedef uint32_t CFStringEncoding; -/* kCFStringEncodingASCII analog. */ +// kCFStringEncodingASCII analog. #define CF_STRING_ENCODING_ASCII 0x0600 -/* kCFStringEncodingUTF8 analog. */ +// kCFStringEncodingUTF8 analog. #define CF_STRING_ENCODING_UTF8 0x08000100 #define CF_PROPERTY_LIST_IMMUTABLE 0 @@ -74,10 +72,10 @@ typedef Boolean (*CFStringGetCStringFuncTy)(CFStringRef, char *, CFIndex, CFStringEncoding); typedef void (*CFReleaseFuncTy)(CFTypeRef); -/* Find and parse the SystemVersion.plist file. */ +// Find and parse the SystemVersion.plist file. static void parseSystemVersionPList(void *Unused) { (void)Unused; - /* Load CoreFoundation dynamically */ + // Load CoreFoundation dynamically const void *NullAllocator = dlsym(RTLD_DEFAULT, "kCFAllocatorNull"); if (!NullAllocator) return; @@ -88,18 +86,18 @@ static void parseSystemVersionPList(void *Unused) { if (!CFDataCreateWithBytesNoCopyFunc) return; CFPropertyListCreateWithDataFuncTy CFPropertyListCreateWithDataFunc = - (CFPropertyListCreateWithDataFuncTy)dlsym( - RTLD_DEFAULT, "CFPropertyListCreateWithData"); -/* CFPropertyListCreateWithData was introduced only in macOS 10.6+, so it - * will be NULL on earlier OS versions. */ + (CFPropertyListCreateWithDataFuncTy)dlsym(RTLD_DEFAULT, + "CFPropertyListCreateWithData"); +// CFPropertyListCreateWithData was introduced only in macOS 10.6+, so it +// will be NULL on earlier OS versions. #pragma clang diagnostic push #pragma clang diagnostic ignored "-Wdeprecated-declarations" CFPropertyListCreateFromXMLDataFuncTy CFPropertyListCreateFromXMLDataFunc = (CFPropertyListCreateFromXMLDataFuncTy)dlsym( RTLD_DEFAULT, "CFPropertyListCreateFromXMLData"); #pragma clang diagnostic pop - /* CFPropertyListCreateFromXMLDataFunc is deprecated in macOS 10.10, so it - * might be NULL in future OS versions. */ + // CFPropertyListCreateFromXMLDataFunc is deprecated in macOS 10.10, so it + // might be NULL in future OS versions. if (!CFPropertyListCreateWithDataFunc && !CFPropertyListCreateFromXMLDataFunc) return; CFStringCreateWithCStringNoCopyFuncTy CFStringCreateWithCStringNoCopyFunc = @@ -143,7 +141,7 @@ static void parseSystemVersionPList(void *Unused) { if (!PropertyList) return; - /* Dynamically allocated stuff. */ + // Dynamically allocated stuff. CFDictionaryRef PListRef = NULL; CFDataRef FileContentsRef = NULL; UInt8 *PListBuf = NULL; @@ -162,8 +160,8 @@ static void parseSystemVersionPList(void *Unused) { if (NumRead != (size_t)PListFileSize) goto Fail; - /* Get the file buffer into CF's format. We pass in a null allocator here * - * because we free PListBuf ourselves */ + // Get the file buffer into CF's format. We pass in a null allocator here * + // because we free PListBuf ourselves FileContentsRef = (*CFDataCreateWithBytesNoCopyFunc)( NULL, PListBuf, (CFIndex)NumRead, AllocatorNull); if (!FileContentsRef) @@ -204,7 +202,7 @@ Fail: } int32_t __isOSVersionAtLeast(int32_t Major, int32_t Minor, int32_t Subminor) { - /* Populate the global version variables, if they haven't already. */ + // Populate the global version variables, if they haven't already. dispatch_once_f(&DispatchOnceCounter, NULL, parseSystemVersionPList); if (Major < GlobalMajor) @@ -220,7 +218,7 @@ int32_t __isOSVersionAtLeast(int32_t Major, int32_t Minor, int32_t Subminor) { #else -/* Silence an empty translation unit warning. */ +// Silence an empty translation unit warning. typedef int unused; #endif diff --git a/lib/builtins/paritydi2.c b/lib/builtins/paritydi2.c index 8ea5ab4214e9..dd9d45e63ea4 100644 --- a/lib/builtins/paritydi2.c +++ b/lib/builtins/paritydi2.c @@ -1,25 +1,21 @@ -/* ===-- paritydi2.c - Implement __paritydi2 -------------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - * - * This file implements __paritydi2 for the compiler_rt library. - * - * ===----------------------------------------------------------------------=== - */ +//===-- paritydi2.c - Implement __paritydi2 -------------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements __paritydi2 for the compiler_rt library. +// +//===----------------------------------------------------------------------===// #include "int_lib.h" -/* Returns: 1 if number of bits is odd else returns 0 */ +// Returns: 1 if number of bits is odd else returns 0 -COMPILER_RT_ABI si_int -__paritydi2(di_int a) -{ - dwords x; - x.all = a; - return __paritysi2(x.s.high ^ x.s.low); +COMPILER_RT_ABI si_int __paritydi2(di_int a) { + dwords x; + x.all = a; + return __paritysi2(x.s.high ^ x.s.low); } diff --git a/lib/builtins/paritysi2.c b/lib/builtins/paritysi2.c index 599984663849..3efa961f2f85 100644 --- a/lib/builtins/paritysi2.c +++ b/lib/builtins/paritysi2.c @@ -1,27 +1,23 @@ -/* ===-- paritysi2.c - Implement __paritysi2 -------------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - * - * This file implements __paritysi2 for the compiler_rt library. - * - * ===----------------------------------------------------------------------=== - */ +//===-- paritysi2.c - Implement __paritysi2 -------------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements __paritysi2 for the compiler_rt library. +// +//===----------------------------------------------------------------------===// #include "int_lib.h" -/* Returns: 1 if number of bits is odd else returns 0 */ +// Returns: 1 if number of bits is odd else returns 0 -COMPILER_RT_ABI si_int -__paritysi2(si_int a) -{ - su_int x = (su_int)a; - x ^= x >> 16; - x ^= x >> 8; - x ^= x >> 4; - return (0x6996 >> (x & 0xF)) & 1; +COMPILER_RT_ABI si_int __paritysi2(si_int a) { + su_int x = (su_int)a; + x ^= x >> 16; + x ^= x >> 8; + x ^= x >> 4; + return (0x6996 >> (x & 0xF)) & 1; } diff --git a/lib/builtins/parityti2.c b/lib/builtins/parityti2.c index 5a4fe492486f..f3942ba8378c 100644 --- a/lib/builtins/parityti2.c +++ b/lib/builtins/parityti2.c @@ -1,29 +1,25 @@ -/* ===-- parityti2.c - Implement __parityti2 -------------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - * - * This file implements __parityti2 for the compiler_rt library. - * - * ===----------------------------------------------------------------------=== - */ +//===-- parityti2.c - Implement __parityti2 -------------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements __parityti2 for the compiler_rt library. +// +//===----------------------------------------------------------------------===// #include "int_lib.h" #ifdef CRT_HAS_128BIT -/* Returns: 1 if number of bits is odd else returns 0 */ +// Returns: 1 if number of bits is odd else returns 0 -COMPILER_RT_ABI si_int -__parityti2(ti_int a) -{ - twords x; - x.all = a; - return __paritydi2(x.s.high ^ x.s.low); +COMPILER_RT_ABI si_int __parityti2(ti_int a) { + twords x; + x.all = a; + return __paritydi2(x.s.high ^ x.s.low); } -#endif /* CRT_HAS_128BIT */ +#endif // CRT_HAS_128BIT diff --git a/lib/builtins/popcountdi2.c b/lib/builtins/popcountdi2.c index 5e8a62f075eb..9bbc39c6608a 100644 --- a/lib/builtins/popcountdi2.c +++ b/lib/builtins/popcountdi2.c @@ -1,36 +1,32 @@ -/* ===-- popcountdi2.c - Implement __popcountdi2 ----------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - * - * This file implements __popcountdi2 for the compiler_rt library. - * - * ===----------------------------------------------------------------------=== - */ +//===-- popcountdi2.c - Implement __popcountdi2 ---------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements __popcountdi2 for the compiler_rt library. +// +//===----------------------------------------------------------------------===// #include "int_lib.h" -/* Returns: count of 1 bits */ +// Returns: count of 1 bits -COMPILER_RT_ABI si_int -__popcountdi2(di_int a) -{ - du_int x2 = (du_int)a; - x2 = x2 - ((x2 >> 1) & 0x5555555555555555uLL); - /* Every 2 bits holds the sum of every pair of bits (32) */ - x2 = ((x2 >> 2) & 0x3333333333333333uLL) + (x2 & 0x3333333333333333uLL); - /* Every 4 bits holds the sum of every 4-set of bits (3 significant bits) (16) */ - x2 = (x2 + (x2 >> 4)) & 0x0F0F0F0F0F0F0F0FuLL; - /* Every 8 bits holds the sum of every 8-set of bits (4 significant bits) (8) */ - su_int x = (su_int)(x2 + (x2 >> 32)); - /* The lower 32 bits hold four 16 bit sums (5 significant bits). */ - /* Upper 32 bits are garbage */ - x = x + (x >> 16); - /* The lower 16 bits hold two 32 bit sums (6 significant bits). */ - /* Upper 16 bits are garbage */ - return (x + (x >> 8)) & 0x0000007F; /* (7 significant bits) */ +COMPILER_RT_ABI si_int __popcountdi2(di_int a) { + du_int x2 = (du_int)a; + x2 = x2 - ((x2 >> 1) & 0x5555555555555555uLL); + // Every 2 bits holds the sum of every pair of bits (32) + x2 = ((x2 >> 2) & 0x3333333333333333uLL) + (x2 & 0x3333333333333333uLL); + // Every 4 bits holds the sum of every 4-set of bits (3 significant bits) (16) + x2 = (x2 + (x2 >> 4)) & 0x0F0F0F0F0F0F0F0FuLL; + // Every 8 bits holds the sum of every 8-set of bits (4 significant bits) (8) + su_int x = (su_int)(x2 + (x2 >> 32)); + // The lower 32 bits hold four 16 bit sums (5 significant bits). + // Upper 32 bits are garbage + x = x + (x >> 16); + // The lower 16 bits hold two 32 bit sums (6 significant bits). + // Upper 16 bits are garbage + return (x + (x >> 8)) & 0x0000007F; // (7 significant bits) } diff --git a/lib/builtins/popcountsi2.c b/lib/builtins/popcountsi2.c index 44544ff49890..75e592a778d9 100644 --- a/lib/builtins/popcountsi2.c +++ b/lib/builtins/popcountsi2.c @@ -1,33 +1,29 @@ -/* ===-- popcountsi2.c - Implement __popcountsi2 ---------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - * - * This file implements __popcountsi2 for the compiler_rt library. - * - * ===----------------------------------------------------------------------=== - */ +//===-- popcountsi2.c - Implement __popcountsi2 ---------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements __popcountsi2 for the compiler_rt library. +// +//===----------------------------------------------------------------------===// #include "int_lib.h" -/* Returns: count of 1 bits */ +// Returns: count of 1 bits -COMPILER_RT_ABI si_int -__popcountsi2(si_int a) -{ - su_int x = (su_int)a; - x = x - ((x >> 1) & 0x55555555); - /* Every 2 bits holds the sum of every pair of bits */ - x = ((x >> 2) & 0x33333333) + (x & 0x33333333); - /* Every 4 bits holds the sum of every 4-set of bits (3 significant bits) */ - x = (x + (x >> 4)) & 0x0F0F0F0F; - /* Every 8 bits holds the sum of every 8-set of bits (4 significant bits) */ - x = (x + (x >> 16)); - /* The lower 16 bits hold two 8 bit sums (5 significant bits).*/ - /* Upper 16 bits are garbage */ - return (x + (x >> 8)) & 0x0000003F; /* (6 significant bits) */ +COMPILER_RT_ABI si_int __popcountsi2(si_int a) { + su_int x = (su_int)a; + x = x - ((x >> 1) & 0x55555555); + // Every 2 bits holds the sum of every pair of bits + x = ((x >> 2) & 0x33333333) + (x & 0x33333333); + // Every 4 bits holds the sum of every 4-set of bits (3 significant bits) + x = (x + (x >> 4)) & 0x0F0F0F0F; + // Every 8 bits holds the sum of every 8-set of bits (4 significant bits) + x = (x + (x >> 16)); + // The lower 16 bits hold two 8 bit sums (5 significant bits). + // Upper 16 bits are garbage + return (x + (x >> 8)) & 0x0000003F; // (6 significant bits) } diff --git a/lib/builtins/popcountti2.c b/lib/builtins/popcountti2.c index 7451bbb286b7..853fd722309e 100644 --- a/lib/builtins/popcountti2.c +++ b/lib/builtins/popcountti2.c @@ -1,44 +1,43 @@ -/* ===-- popcountti2.c - Implement __popcountti2 ----------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - * - * This file implements __popcountti2 for the compiler_rt library. - * - * ===----------------------------------------------------------------------=== - */ +//===-- popcountti2.c - Implement __popcountti2 +//----------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements __popcountti2 for the compiler_rt library. +// +//===----------------------------------------------------------------------===// #include "int_lib.h" #ifdef CRT_HAS_128BIT -/* Returns: count of 1 bits */ +// Returns: count of 1 bits -COMPILER_RT_ABI si_int -__popcountti2(ti_int a) -{ - tu_int x3 = (tu_int)a; - x3 = x3 - ((x3 >> 1) & (((tu_int)0x5555555555555555uLL << 64) | - 0x5555555555555555uLL)); - /* Every 2 bits holds the sum of every pair of bits (64) */ - x3 = ((x3 >> 2) & (((tu_int)0x3333333333333333uLL << 64) | 0x3333333333333333uLL)) - + (x3 & (((tu_int)0x3333333333333333uLL << 64) | 0x3333333333333333uLL)); - /* Every 4 bits holds the sum of every 4-set of bits (3 significant bits) (32) */ - x3 = (x3 + (x3 >> 4)) - & (((tu_int)0x0F0F0F0F0F0F0F0FuLL << 64) | 0x0F0F0F0F0F0F0F0FuLL); - /* Every 8 bits holds the sum of every 8-set of bits (4 significant bits) (16) */ - du_int x2 = (du_int)(x3 + (x3 >> 64)); - /* Every 8 bits holds the sum of every 8-set of bits (5 significant bits) (8) */ - su_int x = (su_int)(x2 + (x2 >> 32)); - /* Every 8 bits holds the sum of every 8-set of bits (6 significant bits) (4) */ - x = x + (x >> 16); - /* Every 8 bits holds the sum of every 8-set of bits (7 significant bits) (2) */ - /* Upper 16 bits are garbage */ - return (x + (x >> 8)) & 0xFF; /* (8 significant bits) */ +COMPILER_RT_ABI si_int __popcountti2(ti_int a) { + tu_int x3 = (tu_int)a; + x3 = x3 - ((x3 >> 1) & + (((tu_int)0x5555555555555555uLL << 64) | 0x5555555555555555uLL)); + // Every 2 bits holds the sum of every pair of bits (64) + x3 = ((x3 >> 2) & + (((tu_int)0x3333333333333333uLL << 64) | 0x3333333333333333uLL)) + + (x3 & (((tu_int)0x3333333333333333uLL << 64) | 0x3333333333333333uLL)); + // Every 4 bits holds the sum of every 4-set of bits (3 significant bits) (32) + x3 = (x3 + (x3 >> 4)) & + (((tu_int)0x0F0F0F0F0F0F0F0FuLL << 64) | 0x0F0F0F0F0F0F0F0FuLL); + // Every 8 bits holds the sum of every 8-set of bits (4 significant bits) (16) + du_int x2 = (du_int)(x3 + (x3 >> 64)); + // Every 8 bits holds the sum of every 8-set of bits (5 significant bits) (8) + su_int x = (su_int)(x2 + (x2 >> 32)); + // Every 8 bits holds the sum of every 8-set of bits (6 significant bits) (4) + x = x + (x >> 16); + // Every 8 bits holds the sum of every 8-set of bits (7 significant bits) (2) + // + // Upper 16 bits are garbage + return (x + (x >> 8)) & 0xFF; // (8 significant bits) } -#endif /* CRT_HAS_128BIT */ +#endif // CRT_HAS_128BIT diff --git a/lib/builtins/powidf2.c b/lib/builtins/powidf2.c index ac13b172b043..9697588484e7 100644 --- a/lib/builtins/powidf2.c +++ b/lib/builtins/powidf2.c @@ -1,34 +1,29 @@ -/* ===-- powidf2.cpp - Implement __powidf2 ---------------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - * - * This file implements __powidf2 for the compiler_rt library. - * - * ===----------------------------------------------------------------------=== - */ +//===-- powidf2.cpp - Implement __powidf2 ---------------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements __powidf2 for the compiler_rt library. +// +//===----------------------------------------------------------------------===// #include "int_lib.h" -/* Returns: a ^ b */ +// Returns: a ^ b -COMPILER_RT_ABI double -__powidf2(double a, si_int b) -{ - const int recip = b < 0; - double r = 1; - while (1) - { - if (b & 1) - r *= a; - b /= 2; - if (b == 0) - break; - a *= a; - } - return recip ? 1/r : r; +COMPILER_RT_ABI double __powidf2(double a, si_int b) { + const int recip = b < 0; + double r = 1; + while (1) { + if (b & 1) + r *= a; + b /= 2; + if (b == 0) + break; + a *= a; + } + return recip ? 1 / r : r; } diff --git a/lib/builtins/powisf2.c b/lib/builtins/powisf2.c index 0c400ec6dd6a..469402348825 100644 --- a/lib/builtins/powisf2.c +++ b/lib/builtins/powisf2.c @@ -1,34 +1,29 @@ -/*===-- powisf2.cpp - Implement __powisf2 ---------------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - * - * This file implements __powisf2 for the compiler_rt library. - * - * ===----------------------------------------------------------------------=== - */ +//===-- powisf2.cpp - Implement __powisf2 ---------------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements __powisf2 for the compiler_rt library. +// +//===----------------------------------------------------------------------===// #include "int_lib.h" -/* Returns: a ^ b */ +// Returns: a ^ b -COMPILER_RT_ABI float -__powisf2(float a, si_int b) -{ - const int recip = b < 0; - float r = 1; - while (1) - { - if (b & 1) - r *= a; - b /= 2; - if (b == 0) - break; - a *= a; - } - return recip ? 1/r : r; +COMPILER_RT_ABI float __powisf2(float a, si_int b) { + const int recip = b < 0; + float r = 1; + while (1) { + if (b & 1) + r *= a; + b /= 2; + if (b == 0) + break; + a *= a; + } + return recip ? 1 / r : r; } diff --git a/lib/builtins/powitf2.c b/lib/builtins/powitf2.c index 172f29f58f25..fcbdb4c2ee2a 100644 --- a/lib/builtins/powitf2.c +++ b/lib/builtins/powitf2.c @@ -1,38 +1,33 @@ -/* ===-- powitf2.cpp - Implement __powitf2 ---------------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - * - * This file implements __powitf2 for the compiler_rt library. - * - * ===----------------------------------------------------------------------=== - */ +//===-- powitf2.cpp - Implement __powitf2 ---------------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements __powitf2 for the compiler_rt library. +// +//===----------------------------------------------------------------------===// #include "int_lib.h" #if _ARCH_PPC -/* Returns: a ^ b */ +// Returns: a ^ b -COMPILER_RT_ABI long double -__powitf2(long double a, si_int b) -{ - const int recip = b < 0; - long double r = 1; - while (1) - { - if (b & 1) - r *= a; - b /= 2; - if (b == 0) - break; - a *= a; - } - return recip ? 1/r : r; +COMPILER_RT_ABI long double __powitf2(long double a, si_int b) { + const int recip = b < 0; + long double r = 1; + while (1) { + if (b & 1) + r *= a; + b /= 2; + if (b == 0) + break; + a *= a; + } + return recip ? 1 / r : r; } #endif diff --git a/lib/builtins/powixf2.c b/lib/builtins/powixf2.c index 0fd96e503e7b..b7b52095afa1 100644 --- a/lib/builtins/powixf2.c +++ b/lib/builtins/powixf2.c @@ -1,38 +1,33 @@ -/* ===-- powixf2.cpp - Implement __powixf2 ---------------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - * - * This file implements __powixf2 for the compiler_rt library. - * - * ===----------------------------------------------------------------------=== - */ +//===-- powixf2.cpp - Implement __powixf2 ---------------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements __powixf2 for the compiler_rt library. +// +//===----------------------------------------------------------------------===// #if !_ARCH_PPC #include "int_lib.h" -/* Returns: a ^ b */ +// Returns: a ^ b -COMPILER_RT_ABI long double -__powixf2(long double a, si_int b) -{ - const int recip = b < 0; - long double r = 1; - while (1) - { - if (b & 1) - r *= a; - b /= 2; - if (b == 0) - break; - a *= a; - } - return recip ? 1/r : r; +COMPILER_RT_ABI long double __powixf2(long double a, si_int b) { + const int recip = b < 0; + long double r = 1; + while (1) { + if (b & 1) + r *= a; + b /= 2; + if (b == 0) + break; + a *= a; + } + return recip ? 1 / r : r; } #endif diff --git a/lib/builtins/ppc/DD.h b/lib/builtins/ppc/DD.h index 3e5f9e58c138..8f31a962fc77 100644 --- a/lib/builtins/ppc/DD.h +++ b/lib/builtins/ppc/DD.h @@ -4,20 +4,20 @@ #include "../int_lib.h" typedef union { - long double ld; - struct { - double hi; - double lo; - }s; + long double ld; + struct { + double hi; + double lo; + } s; } DD; -typedef union { - double d; - uint64_t x; +typedef union { + double d; + uint64_t x; } doublebits; -#define LOWORDER(xy,xHi,xLo,yHi,yLo) \ - (((((xHi)*(yHi) - (xy)) + (xHi)*(yLo)) + (xLo)*(yHi)) + (xLo)*(yLo)) +#define LOWORDER(xy, xHi, xLo, yHi, yLo) \ + (((((xHi) * (yHi) - (xy)) + (xHi) * (yLo)) + (xLo) * (yHi)) + (xLo) * (yLo)) static __inline ALWAYS_INLINE double local_fabs(double x) { doublebits result = {.d = x}; @@ -42,4 +42,4 @@ long double __gcc_qsub(long double, long double); long double __gcc_qmul(long double, long double); long double __gcc_qdiv(long double, long double); -#endif /* COMPILERRT_DD_HEADER */ +#endif // COMPILERRT_DD_HEADER diff --git a/lib/builtins/ppc/divtc3.c b/lib/builtins/ppc/divtc3.c index ef532b841145..afaccf5a8fd6 100644 --- a/lib/builtins/ppc/divtc3.c +++ b/lib/builtins/ppc/divtc3.c @@ -1,9 +1,9 @@ -/* This file is distributed under the University of Illinois Open Source - * License. See LICENSE.TXT for details. - */ +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception -#include "DD.h" #include "../int_math.h" +#include "DD.h" // Use DOUBLE_PRECISION because the soft-fp method we use is logb (on the upper // half of the long doubles), even though this file defines complex division for // 128-bit floats. @@ -12,86 +12,85 @@ #if !defined(CRT_INFINITY) && defined(HUGE_VAL) #define CRT_INFINITY HUGE_VAL -#endif /* CRT_INFINITY */ +#endif // CRT_INFINITY + +#define makeFinite(x) \ + { \ + (x).s.hi = crt_copysign(crt_isinf((x).s.hi) ? 1.0 : 0.0, (x).s.hi); \ + (x).s.lo = 0.0; \ + } + +long double _Complex __divtc3(long double a, long double b, long double c, + long double d) { + DD cDD = {.ld = c}; + DD dDD = {.ld = d}; + + int ilogbw = 0; + const double logbw = + __compiler_rt_logb(crt_fmax(crt_fabs(cDD.s.hi), crt_fabs(dDD.s.hi))); + + if (crt_isfinite(logbw)) { + ilogbw = (int)logbw; + + cDD.s.hi = crt_scalbn(cDD.s.hi, -ilogbw); + cDD.s.lo = crt_scalbn(cDD.s.lo, -ilogbw); + dDD.s.hi = crt_scalbn(dDD.s.hi, -ilogbw); + dDD.s.lo = crt_scalbn(dDD.s.lo, -ilogbw); + } + + const long double denom = + __gcc_qadd(__gcc_qmul(cDD.ld, cDD.ld), __gcc_qmul(dDD.ld, dDD.ld)); + const long double realNumerator = + __gcc_qadd(__gcc_qmul(a, cDD.ld), __gcc_qmul(b, dDD.ld)); + const long double imagNumerator = + __gcc_qsub(__gcc_qmul(b, cDD.ld), __gcc_qmul(a, dDD.ld)); + + DD real = {.ld = __gcc_qdiv(realNumerator, denom)}; + DD imag = {.ld = __gcc_qdiv(imagNumerator, denom)}; + + real.s.hi = crt_scalbn(real.s.hi, -ilogbw); + real.s.lo = crt_scalbn(real.s.lo, -ilogbw); + imag.s.hi = crt_scalbn(imag.s.hi, -ilogbw); + imag.s.lo = crt_scalbn(imag.s.lo, -ilogbw); -#define makeFinite(x) { \ - (x).s.hi = crt_copysign(crt_isinf((x).s.hi) ? 1.0 : 0.0, (x).s.hi); \ - (x).s.lo = 0.0; \ + if (crt_isnan(real.s.hi) && crt_isnan(imag.s.hi)) { + DD aDD = {.ld = a}; + DD bDD = {.ld = b}; + DD rDD = {.ld = denom}; + + if ((rDD.s.hi == 0.0) && (!crt_isnan(aDD.s.hi) || !crt_isnan(bDD.s.hi))) { + real.s.hi = crt_copysign(CRT_INFINITY, cDD.s.hi) * aDD.s.hi; + real.s.lo = 0.0; + imag.s.hi = crt_copysign(CRT_INFINITY, cDD.s.hi) * bDD.s.hi; + imag.s.lo = 0.0; + } + + else if ((crt_isinf(aDD.s.hi) || crt_isinf(bDD.s.hi)) && + crt_isfinite(cDD.s.hi) && crt_isfinite(dDD.s.hi)) { + makeFinite(aDD); + makeFinite(bDD); + real.s.hi = CRT_INFINITY * (aDD.s.hi * cDD.s.hi + bDD.s.hi * dDD.s.hi); + real.s.lo = 0.0; + imag.s.hi = CRT_INFINITY * (bDD.s.hi * cDD.s.hi - aDD.s.hi * dDD.s.hi); + imag.s.lo = 0.0; + } + + else if ((crt_isinf(cDD.s.hi) || crt_isinf(dDD.s.hi)) && + crt_isfinite(aDD.s.hi) && crt_isfinite(bDD.s.hi)) { + makeFinite(cDD); + makeFinite(dDD); + real.s.hi = + crt_copysign(0.0, (aDD.s.hi * cDD.s.hi + bDD.s.hi * dDD.s.hi)); + real.s.lo = 0.0; + imag.s.hi = + crt_copysign(0.0, (bDD.s.hi * cDD.s.hi - aDD.s.hi * dDD.s.hi)); + imag.s.lo = 0.0; + } } -long double _Complex -__divtc3(long double a, long double b, long double c, long double d) -{ - DD cDD = { .ld = c }; - DD dDD = { .ld = d }; - - int ilogbw = 0; - const double logbw = __compiler_rt_logb( - crt_fmax(crt_fabs(cDD.s.hi), crt_fabs(dDD.s.hi))); - - if (crt_isfinite(logbw)) - { - ilogbw = (int)logbw; - - cDD.s.hi = crt_scalbn(cDD.s.hi, -ilogbw); - cDD.s.lo = crt_scalbn(cDD.s.lo, -ilogbw); - dDD.s.hi = crt_scalbn(dDD.s.hi, -ilogbw); - dDD.s.lo = crt_scalbn(dDD.s.lo, -ilogbw); - } - - const long double denom = __gcc_qadd(__gcc_qmul(cDD.ld, cDD.ld), __gcc_qmul(dDD.ld, dDD.ld)); - const long double realNumerator = __gcc_qadd(__gcc_qmul(a,cDD.ld), __gcc_qmul(b,dDD.ld)); - const long double imagNumerator = __gcc_qsub(__gcc_qmul(b,cDD.ld), __gcc_qmul(a,dDD.ld)); - - DD real = { .ld = __gcc_qdiv(realNumerator, denom) }; - DD imag = { .ld = __gcc_qdiv(imagNumerator, denom) }; - - real.s.hi = crt_scalbn(real.s.hi, -ilogbw); - real.s.lo = crt_scalbn(real.s.lo, -ilogbw); - imag.s.hi = crt_scalbn(imag.s.hi, -ilogbw); - imag.s.lo = crt_scalbn(imag.s.lo, -ilogbw); - - if (crt_isnan(real.s.hi) && crt_isnan(imag.s.hi)) - { - DD aDD = { .ld = a }; - DD bDD = { .ld = b }; - DD rDD = { .ld = denom }; - - if ((rDD.s.hi == 0.0) && (!crt_isnan(aDD.s.hi) || - !crt_isnan(bDD.s.hi))) - { - real.s.hi = crt_copysign(CRT_INFINITY,cDD.s.hi) * aDD.s.hi; - real.s.lo = 0.0; - imag.s.hi = crt_copysign(CRT_INFINITY,cDD.s.hi) * bDD.s.hi; - imag.s.lo = 0.0; - } - - else if ((crt_isinf(aDD.s.hi) || crt_isinf(bDD.s.hi)) && - crt_isfinite(cDD.s.hi) && crt_isfinite(dDD.s.hi)) - { - makeFinite(aDD); - makeFinite(bDD); - real.s.hi = CRT_INFINITY * (aDD.s.hi*cDD.s.hi + bDD.s.hi*dDD.s.hi); - real.s.lo = 0.0; - imag.s.hi = CRT_INFINITY * (bDD.s.hi*cDD.s.hi - aDD.s.hi*dDD.s.hi); - imag.s.lo = 0.0; - } - - else if ((crt_isinf(cDD.s.hi) || crt_isinf(dDD.s.hi)) && - crt_isfinite(aDD.s.hi) && crt_isfinite(bDD.s.hi)) - { - makeFinite(cDD); - makeFinite(dDD); - real.s.hi = crt_copysign(0.0,(aDD.s.hi*cDD.s.hi + bDD.s.hi*dDD.s.hi)); - real.s.lo = 0.0; - imag.s.hi = crt_copysign(0.0,(bDD.s.hi*cDD.s.hi - aDD.s.hi*dDD.s.hi)); - imag.s.lo = 0.0; - } - } - - long double _Complex z; - __real__ z = real.ld; - __imag__ z = imag.ld; - - return z; + long double _Complex z; + __real__ z = real.ld; + __imag__ z = imag.ld; + + return z; } diff --git a/lib/builtins/ppc/fixtfdi.c b/lib/builtins/ppc/fixtfdi.c index 2c7c0f8e279f..a97aaf095846 100644 --- a/lib/builtins/ppc/fixtfdi.c +++ b/lib/builtins/ppc/fixtfdi.c @@ -1,104 +1,98 @@ -/* This file is distributed under the University of Illinois Open Source - * License. See LICENSE.TXT for details. - */ +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception -/* int64_t __fixunstfdi(long double x); - * This file implements the PowerPC 128-bit double-double -> int64_t conversion - */ +// int64_t __fixunstfdi(long double x); +// This file implements the PowerPC 128-bit double-double -> int64_t conversion -#include "DD.h" #include "../int_math.h" +#include "DD.h" + +uint64_t __fixtfdi(long double input) { + const DD x = {.ld = input}; + const doublebits hibits = {.d = x.s.hi}; + + const uint32_t absHighWord = + (uint32_t)(hibits.x >> 32) & UINT32_C(0x7fffffff); + const uint32_t absHighWordMinusOne = absHighWord - UINT32_C(0x3ff00000); + + // If (1.0 - tiny) <= input < 0x1.0p63: + if (UINT32_C(0x03f00000) > absHighWordMinusOne) { + // Do an unsigned conversion of the absolute value, then restore the sign. + const int unbiasedHeadExponent = absHighWordMinusOne >> 20; + + int64_t result = hibits.x & INT64_C(0x000fffffffffffff); // mantissa(hi) + result |= INT64_C(0x0010000000000000); // matissa(hi) with implicit bit + result <<= 10; // mantissa(hi) with one zero preceding bit. + + const int64_t hiNegationMask = ((int64_t)(hibits.x)) >> 63; + + // If the tail is non-zero, we need to patch in the tail bits. + if (0.0 != x.s.lo) { + const doublebits lobits = {.d = x.s.lo}; + int64_t tailMantissa = lobits.x & INT64_C(0x000fffffffffffff); + tailMantissa |= INT64_C(0x0010000000000000); + + // At this point we have the mantissa of |tail| + // We need to negate it if head and tail have different signs. + const int64_t loNegationMask = ((int64_t)(lobits.x)) >> 63; + const int64_t negationMask = loNegationMask ^ hiNegationMask; + tailMantissa = (tailMantissa ^ negationMask) - negationMask; + + // Now we have the mantissa of tail as a signed 2s-complement integer + + const int biasedTailExponent = (int)(lobits.x >> 52) & 0x7ff; + + // Shift the tail mantissa into the right position, accounting for the + // bias of 10 that we shifted the head mantissa by. + tailMantissa >>= + (unbiasedHeadExponent - (biasedTailExponent - (1023 - 10))); + + result += tailMantissa; + } + + result >>= (62 - unbiasedHeadExponent); + + // Restore the sign of the result and return + result = (result ^ hiNegationMask) - hiNegationMask; + return result; + } + + // Edge cases handled here: + + // |x| < 1, result is zero. + if (1.0 > crt_fabs(x.s.hi)) + return INT64_C(0); + + // x very close to INT64_MIN, care must be taken to see which side we are on. + if (x.s.hi == -0x1.0p63) { + + int64_t result = INT64_MIN; + + if (0.0 < x.s.lo) { + // If the tail is positive, the correct result is something other than + // INT64_MIN. we'll need to figure out what it is. + + const doublebits lobits = {.d = x.s.lo}; + int64_t tailMantissa = lobits.x & INT64_C(0x000fffffffffffff); + tailMantissa |= INT64_C(0x0010000000000000); + + // Now we negate the tailMantissa + tailMantissa = (tailMantissa ^ INT64_C(-1)) + INT64_C(1); + + // And shift it by the appropriate amount + const int biasedTailExponent = (int)(lobits.x >> 52) & 0x7ff; + tailMantissa >>= 1075 - biasedTailExponent; + + result -= tailMantissa; + } + + return result; + } -uint64_t __fixtfdi(long double input) -{ - const DD x = { .ld = input }; - const doublebits hibits = { .d = x.s.hi }; - - const uint32_t absHighWord = (uint32_t)(hibits.x >> 32) & UINT32_C(0x7fffffff); - const uint32_t absHighWordMinusOne = absHighWord - UINT32_C(0x3ff00000); - - /* If (1.0 - tiny) <= input < 0x1.0p63: */ - if (UINT32_C(0x03f00000) > absHighWordMinusOne) - { - /* Do an unsigned conversion of the absolute value, then restore the sign. */ - const int unbiasedHeadExponent = absHighWordMinusOne >> 20; - - int64_t result = hibits.x & INT64_C(0x000fffffffffffff); /* mantissa(hi) */ - result |= INT64_C(0x0010000000000000); /* matissa(hi) with implicit bit */ - result <<= 10; /* mantissa(hi) with one zero preceding bit. */ - - const int64_t hiNegationMask = ((int64_t)(hibits.x)) >> 63; - - /* If the tail is non-zero, we need to patch in the tail bits. */ - if (0.0 != x.s.lo) - { - const doublebits lobits = { .d = x.s.lo }; - int64_t tailMantissa = lobits.x & INT64_C(0x000fffffffffffff); - tailMantissa |= INT64_C(0x0010000000000000); - - /* At this point we have the mantissa of |tail| */ - /* We need to negate it if head and tail have different signs. */ - const int64_t loNegationMask = ((int64_t)(lobits.x)) >> 63; - const int64_t negationMask = loNegationMask ^ hiNegationMask; - tailMantissa = (tailMantissa ^ negationMask) - negationMask; - - /* Now we have the mantissa of tail as a signed 2s-complement integer */ - - const int biasedTailExponent = (int)(lobits.x >> 52) & 0x7ff; - - /* Shift the tail mantissa into the right position, accounting for the - * bias of 10 that we shifted the head mantissa by. - */ - tailMantissa >>= (unbiasedHeadExponent - (biasedTailExponent - (1023 - 10))); - - result += tailMantissa; - } - - result >>= (62 - unbiasedHeadExponent); - - /* Restore the sign of the result and return */ - result = (result ^ hiNegationMask) - hiNegationMask; - return result; - - } - - /* Edge cases handled here: */ - - /* |x| < 1, result is zero. */ - if (1.0 > crt_fabs(x.s.hi)) - return INT64_C(0); - - /* x very close to INT64_MIN, care must be taken to see which side we are on. */ - if (x.s.hi == -0x1.0p63) { - - int64_t result = INT64_MIN; - - if (0.0 < x.s.lo) - { - /* If the tail is positive, the correct result is something other than INT64_MIN. - * we'll need to figure out what it is. - */ - - const doublebits lobits = { .d = x.s.lo }; - int64_t tailMantissa = lobits.x & INT64_C(0x000fffffffffffff); - tailMantissa |= INT64_C(0x0010000000000000); - - /* Now we negate the tailMantissa */ - tailMantissa = (tailMantissa ^ INT64_C(-1)) + INT64_C(1); - - /* And shift it by the appropriate amount */ - const int biasedTailExponent = (int)(lobits.x >> 52) & 0x7ff; - tailMantissa >>= 1075 - biasedTailExponent; - - result -= tailMantissa; - } - - return result; - } - - /* Signed overflows, infinities, and NaNs */ - if (x.s.hi > 0.0) - return INT64_MAX; - else - return INT64_MIN; + // Signed overflows, infinities, and NaNs + if (x.s.hi > 0.0) + return INT64_MAX; + else + return INT64_MIN; } diff --git a/lib/builtins/ppc/fixunstfdi.c b/lib/builtins/ppc/fixunstfdi.c index 5e6e2cedf6ac..8d53f372527a 100644 --- a/lib/builtins/ppc/fixunstfdi.c +++ b/lib/builtins/ppc/fixunstfdi.c @@ -1,59 +1,57 @@ -/* This file is distributed under the University of Illinois Open Source - * License. See LICENSE.TXT for details. - */ +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception -/* uint64_t __fixunstfdi(long double x); */ -/* This file implements the PowerPC 128-bit double-double -> uint64_t conversion */ +// uint64_t __fixunstfdi(long double x); +// This file implements the PowerPC 128-bit double-double -> uint64_t conversion #include "DD.h" -uint64_t __fixunstfdi(long double input) -{ - const DD x = { .ld = input }; - const doublebits hibits = { .d = x.s.hi }; - - const uint32_t highWordMinusOne = (uint32_t)(hibits.x >> 32) - UINT32_C(0x3ff00000); - - /* If (1.0 - tiny) <= input < 0x1.0p64: */ - if (UINT32_C(0x04000000) > highWordMinusOne) - { - const int unbiasedHeadExponent = highWordMinusOne >> 20; - - uint64_t result = hibits.x & UINT64_C(0x000fffffffffffff); /* mantissa(hi) */ - result |= UINT64_C(0x0010000000000000); /* matissa(hi) with implicit bit */ - result <<= 11; /* mantissa(hi) left aligned in the int64 field. */ - - /* If the tail is non-zero, we need to patch in the tail bits. */ - if (0.0 != x.s.lo) - { - const doublebits lobits = { .d = x.s.lo }; - int64_t tailMantissa = lobits.x & INT64_C(0x000fffffffffffff); - tailMantissa |= INT64_C(0x0010000000000000); - - /* At this point we have the mantissa of |tail| */ - - const int64_t negationMask = ((int64_t)(lobits.x)) >> 63; - tailMantissa = (tailMantissa ^ negationMask) - negationMask; - - /* Now we have the mantissa of tail as a signed 2s-complement integer */ - - const int biasedTailExponent = (int)(lobits.x >> 52) & 0x7ff; - - /* Shift the tail mantissa into the right position, accounting for the - * bias of 11 that we shifted the head mantissa by. - */ - tailMantissa >>= (unbiasedHeadExponent - (biasedTailExponent - (1023 - 11))); - - result += tailMantissa; - } - - result >>= (63 - unbiasedHeadExponent); - return result; - } - - /* Edge cases are handled here, with saturation. */ - if (1.0 > x.s.hi) - return UINT64_C(0); - else - return UINT64_MAX; +uint64_t __fixunstfdi(long double input) { + const DD x = {.ld = input}; + const doublebits hibits = {.d = x.s.hi}; + + const uint32_t highWordMinusOne = + (uint32_t)(hibits.x >> 32) - UINT32_C(0x3ff00000); + + // If (1.0 - tiny) <= input < 0x1.0p64: + if (UINT32_C(0x04000000) > highWordMinusOne) { + const int unbiasedHeadExponent = highWordMinusOne >> 20; + + uint64_t result = hibits.x & UINT64_C(0x000fffffffffffff); // mantissa(hi) + result |= UINT64_C(0x0010000000000000); // matissa(hi) with implicit bit + result <<= 11; // mantissa(hi) left aligned in the int64 field. + + // If the tail is non-zero, we need to patch in the tail bits. + if (0.0 != x.s.lo) { + const doublebits lobits = {.d = x.s.lo}; + int64_t tailMantissa = lobits.x & INT64_C(0x000fffffffffffff); + tailMantissa |= INT64_C(0x0010000000000000); + + // At this point we have the mantissa of |tail| + + const int64_t negationMask = ((int64_t)(lobits.x)) >> 63; + tailMantissa = (tailMantissa ^ negationMask) - negationMask; + + // Now we have the mantissa of tail as a signed 2s-complement integer + + const int biasedTailExponent = (int)(lobits.x >> 52) & 0x7ff; + + // Shift the tail mantissa into the right position, accounting for the + // bias of 11 that we shifted the head mantissa by. + tailMantissa >>= + (unbiasedHeadExponent - (biasedTailExponent - (1023 - 11))); + + result += tailMantissa; + } + + result >>= (63 - unbiasedHeadExponent); + return result; + } + + // Edge cases are handled here, with saturation. + if (1.0 > x.s.hi) + return UINT64_C(0); + else + return UINT64_MAX; } diff --git a/lib/builtins/ppc/fixunstfti.c b/lib/builtins/ppc/fixunstfti.c index fa21084cb576..1d19e01e3a91 100644 --- a/lib/builtins/ppc/fixunstfti.c +++ b/lib/builtins/ppc/fixunstfti.c @@ -1,9 +1,8 @@ //===-- lib/builtins/ppc/fixunstfti.c - Convert long double->int128 *-C -*-===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// // @@ -15,10 +14,10 @@ #include "../int_math.h" #define BIAS 1023 -/* Convert long double into an unsigned 128-bit integer. */ +// Convert long double into an unsigned 128-bit integer. __uint128_t __fixunstfti(long double input) { - /* If we are trying to convert a NaN, return the NaN bit pattern. */ + // If we are trying to convert a NaN, return the NaN bit pattern. if (crt_isnan(input)) { return ((__uint128_t)0x7FF8000000000000ll) << 64 | (__uint128_t)0x0000000000000000ll; @@ -26,59 +25,59 @@ __uint128_t __fixunstfti(long double input) { __uint128_t result, hiResult, loResult; int hiExponent, loExponent, shift; - /* The long double representation, with the high and low portions of - * the long double, and the corresponding bit patterns of each double. */ + // The long double representation, with the high and low portions of + // the long double, and the corresponding bit patterns of each double. union { long double ld; - double d[2]; /* [0] is the high double, [1] is the low double. */ - unsigned long long ull[2]; /* High and low doubles as 64-bit integers. */ + double d[2]; // [0] is the high double, [1] is the low double. + unsigned long long ull[2]; // High and low doubles as 64-bit integers. } ldUnion; - /* If the long double is less than 1.0 or negative, - * return 0.0. */ + // If the long double is less than 1.0 or negative, + // return 0.0. if (input < 1.0) return 0.0; - /* Retrieve the 64-bit patterns of high and low doubles. - * Compute the unbiased exponent of both high and low doubles by - * removing the signs, isolating the exponent, and subtracting - * the bias from it. */ + // Retrieve the 64-bit patterns of high and low doubles. + // Compute the unbiased exponent of both high and low doubles by + // removing the signs, isolating the exponent, and subtracting + // the bias from it. ldUnion.ld = input; hiExponent = ((ldUnion.ull[0] & 0x7FFFFFFFFFFFFFFFll) >> 52) - BIAS; loExponent = ((ldUnion.ull[1] & 0x7FFFFFFFFFFFFFFFll) >> 52) - BIAS; - /* Convert each double into int64; they will be added to the int128 result. - * CASE 1: High or low double fits in int64 - * - Convert the each double normally into int64. - * - * CASE 2: High or low double does not fit in int64 - * - Scale the double to fit within a 64-bit integer - * - Calculate the shift (amount to scale the double by in the int128) - * - Clear all the bits of the exponent (with 0x800FFFFFFFFFFFFF) - * - Add BIAS+53 (0x4350000000000000) to exponent to correct the value - * - Scale (move) the double to the correct place in the int128 - * (Move it by 2^53 places) - * - * Note: If the high double is assumed to be positive, an unsigned conversion - * from long double to 64-bit integer is needed. The low double can be either - * positive or negative, so a signed conversion is needed to retain the result - * of the low double and to ensure it does not simply get converted to 0. */ + // Convert each double into int64; they will be added to the int128 result. + // CASE 1: High or low double fits in int64 + // - Convert the each double normally into int64. + // + // CASE 2: High or low double does not fit in int64 + // - Scale the double to fit within a 64-bit integer + // - Calculate the shift (amount to scale the double by in the int128) + // - Clear all the bits of the exponent (with 0x800FFFFFFFFFFFFF) + // - Add BIAS+53 (0x4350000000000000) to exponent to correct the value + // - Scale (move) the double to the correct place in the int128 + // (Move it by 2^53 places) + // + // Note: If the high double is assumed to be positive, an unsigned conversion + // from long double to 64-bit integer is needed. The low double can be either + // positive or negative, so a signed conversion is needed to retain the result + // of the low double and to ensure it does not simply get converted to 0. - /* CASE 1 - High double fits in int64. */ + // CASE 1 - High double fits in int64. if (hiExponent < 63) { hiResult = (unsigned long long)ldUnion.d[0]; } else if (hiExponent < 128) { - /* CASE 2 - High double does not fit in int64, scale and convert it. */ + // CASE 2 - High double does not fit in int64, scale and convert it. shift = hiExponent - 54; ldUnion.ull[0] &= 0x800FFFFFFFFFFFFFll; ldUnion.ull[0] |= 0x4350000000000000ll; hiResult = (unsigned long long)ldUnion.d[0]; hiResult <<= shift; } else { - /* Detect cases for overflow. When the exponent of the high - * double is greater than 128 bits and when the long double - * input is positive, return the max 128-bit integer. - * For negative inputs with exponents > 128, return 1, like gcc. */ + // Detect cases for overflow. When the exponent of the high + // double is greater than 128 bits and when the long double + // input is positive, return the max 128-bit integer. + // For negative inputs with exponents > 128, return 1, like gcc. if (ldUnion.d[0] > 0) { return ((__uint128_t)0xFFFFFFFFFFFFFFFFll) << 64 | (__uint128_t)0xFFFFFFFFFFFFFFFFll; @@ -88,11 +87,11 @@ __uint128_t __fixunstfti(long double input) { } } - /* CASE 1 - Low double fits in int64. */ + // CASE 1 - Low double fits in int64. if (loExponent < 63) { loResult = (long long)ldUnion.d[1]; } else { - /* CASE 2 - Low double does not fit in int64, scale and convert it. */ + // CASE 2 - Low double does not fit in int64, scale and convert it. shift = loExponent - 54; ldUnion.ull[1] &= 0x800FFFFFFFFFFFFFll; ldUnion.ull[1] |= 0x4350000000000000ll; @@ -100,7 +99,7 @@ __uint128_t __fixunstfti(long double input) { loResult <<= shift; } - /* Add the high and low doublewords together to form a 128 bit integer. */ + // Add the high and low doublewords together to form a 128 bit integer. result = loResult + hiResult; return result; } diff --git a/lib/builtins/ppc/floatditf.c b/lib/builtins/ppc/floatditf.c index beabdd017422..4c365418f082 100644 --- a/lib/builtins/ppc/floatditf.c +++ b/lib/builtins/ppc/floatditf.c @@ -1,36 +1,33 @@ -/* This file is distributed under the University of Illinois Open Source - * License. See LICENSE.TXT for details. - */ +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception -/* long double __floatditf(long long x); */ -/* This file implements the PowerPC long long -> long double conversion */ +// long double __floatditf(long long x); +// This file implements the PowerPC long long -> long double conversion #include "DD.h" long double __floatditf(int64_t a) { - - static const double twop32 = 0x1.0p32; - static const double twop52 = 0x1.0p52; - - doublebits low = { .d = twop52 }; - low.x |= a & UINT64_C(0x00000000ffffffff); /* 0x1.0p52 + low 32 bits of a. */ - - const double high_addend = (double)((int32_t)(a >> 32))*twop32 - twop52; - - /* At this point, we have two double precision numbers - * high_addend and low.d, and we wish to return their sum - * as a canonicalized long double: - */ - - /* This implementation sets the inexact flag spuriously. - * This could be avoided, but at some substantial cost. - */ - - DD result; - - result.s.hi = high_addend + low.d; - result.s.lo = (high_addend - result.s.hi) + low.d; - - return result.ld; - + + static const double twop32 = 0x1.0p32; + static const double twop52 = 0x1.0p52; + + doublebits low = {.d = twop52}; + low.x |= a & UINT64_C(0x00000000ffffffff); // 0x1.0p52 + low 32 bits of a. + + const double high_addend = (double)((int32_t)(a >> 32)) * twop32 - twop52; + + // At this point, we have two double precision numbers + // high_addend and low.d, and we wish to return their sum + // as a canonicalized long double: + + // This implementation sets the inexact flag spuriously. + // This could be avoided, but at some substantial cost. + + DD result; + + result.s.hi = high_addend + low.d; + result.s.lo = (high_addend - result.s.hi) + low.d; + + return result.ld; } diff --git a/lib/builtins/ppc/floattitf.c b/lib/builtins/ppc/floattitf.c index b8e297b6b8d9..6deac6498128 100644 --- a/lib/builtins/ppc/floattitf.c +++ b/lib/builtins/ppc/floattitf.c @@ -1,9 +1,8 @@ //===-- lib/builtins/ppc/floattitf.c - Convert int128->long double -*-C -*-===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// // @@ -14,35 +13,34 @@ #include <stdint.h> -/* Conversions from signed and unsigned 64-bit int to long double. */ +// Conversions from signed and unsigned 64-bit int to long double. long double __floatditf(int64_t); long double __floatunditf(uint64_t); -/* Convert a signed 128-bit integer to long double. - * This uses the following property: Let hi and lo be 64-bits each, - * and let signed_val_k() and unsigned_val_k() be the value of the - * argument interpreted as a signed or unsigned k-bit integer. Then, - * - * signed_val_128(hi,lo) = signed_val_64(hi) * 2^64 + unsigned_val_64(lo) - * = (long double)hi * 2^64 + (long double)lo, - * - * where (long double)hi and (long double)lo are signed and - * unsigned 64-bit integer to long double conversions, respectively. - */ +// Convert a signed 128-bit integer to long double. +// This uses the following property: Let hi and lo be 64-bits each, +// and let signed_val_k() and unsigned_val_k() be the value of the +// argument interpreted as a signed or unsigned k-bit integer. Then, +// +// signed_val_128(hi,lo) = signed_val_64(hi) * 2^64 + unsigned_val_64(lo) +// = (long double)hi * 2^64 + (long double)lo, +// +// where (long double)hi and (long double)lo are signed and +// unsigned 64-bit integer to long double conversions, respectively. long double __floattitf(__int128_t arg) { - /* Split the int128 argument into 64-bit high and low int64 parts. */ + // Split the int128 argument into 64-bit high and low int64 parts. int64_t ArgHiPart = (int64_t)(arg >> 64); uint64_t ArgLoPart = (uint64_t)arg; - /* Convert each 64-bit part into long double. The high part - * must be a signed conversion and the low part an unsigned conversion - * to ensure the correct result. */ + // Convert each 64-bit part into long double. The high part + // must be a signed conversion and the low part an unsigned conversion + // to ensure the correct result. long double ConvertedHiPart = __floatditf(ArgHiPart); long double ConvertedLoPart = __floatunditf(ArgLoPart); - /* The low bit of ArgHiPart corresponds to the 2^64 bit in arg. - * Multiply the high part by 2^64 to undo the right shift by 64-bits - * done in the splitting. Then, add to the low part to obtain the - * final result. */ + // The low bit of ArgHiPart corresponds to the 2^64 bit in arg. + // Multiply the high part by 2^64 to undo the right shift by 64-bits + // done in the splitting. Then, add to the low part to obtain the + // final result. return ((ConvertedHiPart * 0x1.0p64) + ConvertedLoPart); } diff --git a/lib/builtins/ppc/floatunditf.c b/lib/builtins/ppc/floatunditf.c index b12e1e738fd0..fb4cd3f91d86 100644 --- a/lib/builtins/ppc/floatunditf.c +++ b/lib/builtins/ppc/floatunditf.c @@ -1,41 +1,39 @@ -/* This file is distributed under the University of Illinois Open Source - * License. See LICENSE.TXT for details. - */ +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception -/* long double __floatunditf(unsigned long long x); */ -/* This file implements the PowerPC unsigned long long -> long double conversion */ +// long double __floatunditf(unsigned long long x); +// This file implements the PowerPC unsigned long long -> long double conversion #include "DD.h" long double __floatunditf(uint64_t a) { - - /* Begins with an exact copy of the code from __floatundidf */ - - static const double twop52 = 0x1.0p52; - static const double twop84 = 0x1.0p84; - static const double twop84_plus_twop52 = 0x1.00000001p84; - - doublebits high = { .d = twop84 }; - doublebits low = { .d = twop52 }; - - high.x |= a >> 32; /* 0x1.0p84 + high 32 bits of a */ - low.x |= a & UINT64_C(0x00000000ffffffff); /* 0x1.0p52 + low 32 bits of a */ - - const double high_addend = high.d - twop84_plus_twop52; - - /* At this point, we have two double precision numbers - * high_addend and low.d, and we wish to return their sum - * as a canonicalized long double: - */ - - /* This implementation sets the inexact flag spuriously. */ - /* This could be avoided, but at some substantial cost. */ - - DD result; - - result.s.hi = high_addend + low.d; - result.s.lo = (high_addend - result.s.hi) + low.d; - - return result.ld; - + + // Begins with an exact copy of the code from __floatundidf + + static const double twop52 = 0x1.0p52; + static const double twop84 = 0x1.0p84; + static const double twop84_plus_twop52 = 0x1.00000001p84; + + doublebits high = {.d = twop84}; + doublebits low = {.d = twop52}; + + high.x |= a >> 32; // 0x1.0p84 + high 32 bits of a + low.x |= a & UINT64_C(0x00000000ffffffff); // 0x1.0p52 + low 32 bits of a + + const double high_addend = high.d - twop84_plus_twop52; + + // At this point, we have two double precision numbers + // high_addend and low.d, and we wish to return their sum + // as a canonicalized long double: + + // This implementation sets the inexact flag spuriously. + // This could be avoided, but at some substantial cost. + + DD result; + + result.s.hi = high_addend + low.d; + result.s.lo = (high_addend - result.s.hi) + low.d; + + return result.ld; } diff --git a/lib/builtins/ppc/gcc_qadd.c b/lib/builtins/ppc/gcc_qadd.c index 32e16e9d1d11..6e1e63cb530e 100644 --- a/lib/builtins/ppc/gcc_qadd.c +++ b/lib/builtins/ppc/gcc_qadd.c @@ -1,76 +1,74 @@ -/* This file is distributed under the University of Illinois Open Source - * License. See LICENSE.TXT for details. - */ +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception -/* long double __gcc_qadd(long double x, long double y); - * This file implements the PowerPC 128-bit double-double add operation. - * This implementation is shamelessly cribbed from Apple's DDRT, circa 1993(!) - */ +// long double __gcc_qadd(long double x, long double y); +// This file implements the PowerPC 128-bit double-double add operation. +// This implementation is shamelessly cribbed from Apple's DDRT, circa 1993(!) #include "DD.h" -long double __gcc_qadd(long double x, long double y) -{ - static const uint32_t infinityHi = UINT32_C(0x7ff00000); - - DD dst = { .ld = x }, src = { .ld = y }; - - register double A = dst.s.hi, a = dst.s.lo, - B = src.s.hi, b = src.s.lo; - - /* If both operands are zero: */ - if ((A == 0.0) && (B == 0.0)) { - dst.s.hi = A + B; - dst.s.lo = 0.0; - return dst.ld; - } - - /* If either operand is NaN or infinity: */ - const doublebits abits = { .d = A }; - const doublebits bbits = { .d = B }; - if ((((uint32_t)(abits.x >> 32) & infinityHi) == infinityHi) || - (((uint32_t)(bbits.x >> 32) & infinityHi) == infinityHi)) { - dst.s.hi = A + B; - dst.s.lo = 0.0; - return dst.ld; - } - - /* If the computation overflows: */ - /* This may be playing things a little bit fast and loose, but it will do for a start. */ - const double testForOverflow = A + (B + (a + b)); - const doublebits testbits = { .d = testForOverflow }; - if (((uint32_t)(testbits.x >> 32) & infinityHi) == infinityHi) { - dst.s.hi = testForOverflow; - dst.s.lo = 0.0; - return dst.ld; - } - - double H, h; - double T, t; - double W, w; - double Y; - - H = B + (A - (A + B)); - T = b + (a - (a + b)); - h = A + (B - (A + B)); - t = a + (b - (a + b)); - - if (local_fabs(A) <= local_fabs(B)) - w = (a + b) + h; - else - w = (a + b) + H; - - W = (A + B) + w; - Y = (A + B) - W; - Y += w; - - if (local_fabs(a) <= local_fabs(b)) - w = t + Y; - else - w = T + Y; - - dst.s.hi = Y = W + w; - dst.s.lo = (W - Y) + w; - - return dst.ld; +long double __gcc_qadd(long double x, long double y) { + static const uint32_t infinityHi = UINT32_C(0x7ff00000); + + DD dst = {.ld = x}, src = {.ld = y}; + + register double A = dst.s.hi, a = dst.s.lo, B = src.s.hi, b = src.s.lo; + + // If both operands are zero: + if ((A == 0.0) && (B == 0.0)) { + dst.s.hi = A + B; + dst.s.lo = 0.0; + return dst.ld; + } + + // If either operand is NaN or infinity: + const doublebits abits = {.d = A}; + const doublebits bbits = {.d = B}; + if ((((uint32_t)(abits.x >> 32) & infinityHi) == infinityHi) || + (((uint32_t)(bbits.x >> 32) & infinityHi) == infinityHi)) { + dst.s.hi = A + B; + dst.s.lo = 0.0; + return dst.ld; + } + + // If the computation overflows: + // This may be playing things a little bit fast and loose, but it will do for + // a start. + const double testForOverflow = A + (B + (a + b)); + const doublebits testbits = {.d = testForOverflow}; + if (((uint32_t)(testbits.x >> 32) & infinityHi) == infinityHi) { + dst.s.hi = testForOverflow; + dst.s.lo = 0.0; + return dst.ld; + } + + double H, h; + double T, t; + double W, w; + double Y; + + H = B + (A - (A + B)); + T = b + (a - (a + b)); + h = A + (B - (A + B)); + t = a + (b - (a + b)); + + if (local_fabs(A) <= local_fabs(B)) + w = (a + b) + h; + else + w = (a + b) + H; + + W = (A + B) + w; + Y = (A + B) - W; + Y += w; + + if (local_fabs(a) <= local_fabs(b)) + w = t + Y; + else + w = T + Y; + + dst.s.hi = Y = W + w; + dst.s.lo = (W - Y) + w; + + return dst.ld; } diff --git a/lib/builtins/ppc/gcc_qdiv.c b/lib/builtins/ppc/gcc_qdiv.c index 70aa00b64400..35a3cbc3d3f8 100644 --- a/lib/builtins/ppc/gcc_qdiv.c +++ b/lib/builtins/ppc/gcc_qdiv.c @@ -1,55 +1,52 @@ -/* This file is distributed under the University of Illinois Open Source - * License. See LICENSE.TXT for details. - */ +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception -/* long double __gcc_qdiv(long double x, long double y); - * This file implements the PowerPC 128-bit double-double division operation. - * This implementation is shamelessly cribbed from Apple's DDRT, circa 1993(!) - */ +// long double __gcc_qdiv(long double x, long double y); +// This file implements the PowerPC 128-bit double-double division operation. +// This implementation is shamelessly cribbed from Apple's DDRT, circa 1993(!) #include "DD.h" -long double __gcc_qdiv(long double a, long double b) -{ - static const uint32_t infinityHi = UINT32_C(0x7ff00000); - DD dst = { .ld = a }, src = { .ld = b }; - - register double x = dst.s.hi, x1 = dst.s.lo, - y = src.s.hi, y1 = src.s.lo; - - double yHi, yLo, qHi, qLo; - double yq, tmp, q; - - q = x / y; - - /* Detect special cases */ - if (q == 0.0) { - dst.s.hi = q; - dst.s.lo = 0.0; - return dst.ld; - } - - const doublebits qBits = { .d = q }; - if (((uint32_t)(qBits.x >> 32) & infinityHi) == infinityHi) { - dst.s.hi = q; - dst.s.lo = 0.0; - return dst.ld; - } - - yHi = high26bits(y); - qHi = high26bits(q); - - yq = y * q; - yLo = y - yHi; - qLo = q - qHi; - - tmp = LOWORDER(yq, yHi, yLo, qHi, qLo); - tmp = (x - yq) - tmp; - tmp = ((tmp + x1) - y1 * q) / y; - x = q + tmp; - - dst.s.lo = (q - x) + tmp; - dst.s.hi = x; - +long double __gcc_qdiv(long double a, long double b) { + static const uint32_t infinityHi = UINT32_C(0x7ff00000); + DD dst = {.ld = a}, src = {.ld = b}; + + register double x = dst.s.hi, x1 = dst.s.lo, y = src.s.hi, y1 = src.s.lo; + + double yHi, yLo, qHi, qLo; + double yq, tmp, q; + + q = x / y; + + // Detect special cases + if (q == 0.0) { + dst.s.hi = q; + dst.s.lo = 0.0; + return dst.ld; + } + + const doublebits qBits = {.d = q}; + if (((uint32_t)(qBits.x >> 32) & infinityHi) == infinityHi) { + dst.s.hi = q; + dst.s.lo = 0.0; return dst.ld; + } + + yHi = high26bits(y); + qHi = high26bits(q); + + yq = y * q; + yLo = y - yHi; + qLo = q - qHi; + + tmp = LOWORDER(yq, yHi, yLo, qHi, qLo); + tmp = (x - yq) - tmp; + tmp = ((tmp + x1) - y1 * q) / y; + x = q + tmp; + + dst.s.lo = (q - x) + tmp; + dst.s.hi = x; + + return dst.ld; } diff --git a/lib/builtins/ppc/gcc_qmul.c b/lib/builtins/ppc/gcc_qmul.c index fb4c5164ccb5..75f519aad6f0 100644 --- a/lib/builtins/ppc/gcc_qmul.c +++ b/lib/builtins/ppc/gcc_qmul.c @@ -1,53 +1,50 @@ -/* This file is distributed under the University of Illinois Open Source - * License. See LICENSE.TXT for details. - */ +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception -/* long double __gcc_qmul(long double x, long double y); - * This file implements the PowerPC 128-bit double-double multiply operation. - * This implementation is shamelessly cribbed from Apple's DDRT, circa 1993(!) - */ +// long double __gcc_qmul(long double x, long double y); +// This file implements the PowerPC 128-bit double-double multiply operation. +// This implementation is shamelessly cribbed from Apple's DDRT, circa 1993(!) #include "DD.h" -long double __gcc_qmul(long double x, long double y) -{ - static const uint32_t infinityHi = UINT32_C(0x7ff00000); - DD dst = { .ld = x }, src = { .ld = y }; - - register double A = dst.s.hi, a = dst.s.lo, - B = src.s.hi, b = src.s.lo; - - double aHi, aLo, bHi, bLo; - double ab, tmp, tau; - - ab = A * B; - - /* Detect special cases */ - if (ab == 0.0) { - dst.s.hi = ab; - dst.s.lo = 0.0; - return dst.ld; - } - - const doublebits abBits = { .d = ab }; - if (((uint32_t)(abBits.x >> 32) & infinityHi) == infinityHi) { - dst.s.hi = ab; - dst.s.lo = 0.0; - return dst.ld; - } - - /* Generic cases handled here. */ - aHi = high26bits(A); - bHi = high26bits(B); - aLo = A - aHi; - bLo = B - bHi; - - tmp = LOWORDER(ab, aHi, aLo, bHi, bLo); - tmp += (A * b + a * B); - tau = ab + tmp; - - dst.s.lo = (ab - tau) + tmp; - dst.s.hi = tau; - +long double __gcc_qmul(long double x, long double y) { + static const uint32_t infinityHi = UINT32_C(0x7ff00000); + DD dst = {.ld = x}, src = {.ld = y}; + + register double A = dst.s.hi, a = dst.s.lo, B = src.s.hi, b = src.s.lo; + + double aHi, aLo, bHi, bLo; + double ab, tmp, tau; + + ab = A * B; + + // Detect special cases + if (ab == 0.0) { + dst.s.hi = ab; + dst.s.lo = 0.0; + return dst.ld; + } + + const doublebits abBits = {.d = ab}; + if (((uint32_t)(abBits.x >> 32) & infinityHi) == infinityHi) { + dst.s.hi = ab; + dst.s.lo = 0.0; return dst.ld; + } + + // Generic cases handled here. + aHi = high26bits(A); + bHi = high26bits(B); + aLo = A - aHi; + bLo = B - bHi; + + tmp = LOWORDER(ab, aHi, aLo, bHi, bLo); + tmp += (A * b + a * B); + tau = ab + tmp; + + dst.s.lo = (ab - tau) + tmp; + dst.s.hi = tau; + + return dst.ld; } diff --git a/lib/builtins/ppc/gcc_qsub.c b/lib/builtins/ppc/gcc_qsub.c index c092e24dbda1..ac08120be0bd 100644 --- a/lib/builtins/ppc/gcc_qsub.c +++ b/lib/builtins/ppc/gcc_qsub.c @@ -1,76 +1,74 @@ -/* This file is distributed under the University of Illinois Open Source - * License. See LICENSE.TXT for details. - */ +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception -/* long double __gcc_qsub(long double x, long double y); - * This file implements the PowerPC 128-bit double-double add operation. - * This implementation is shamelessly cribbed from Apple's DDRT, circa 1993(!) - */ +// long double __gcc_qsub(long double x, long double y); +// This file implements the PowerPC 128-bit double-double add operation. +// This implementation is shamelessly cribbed from Apple's DDRT, circa 1993(!) #include "DD.h" -long double __gcc_qsub(long double x, long double y) -{ - static const uint32_t infinityHi = UINT32_C(0x7ff00000); - - DD dst = { .ld = x }, src = { .ld = y }; - - register double A = dst.s.hi, a = dst.s.lo, - B = -src.s.hi, b = -src.s.lo; - - /* If both operands are zero: */ - if ((A == 0.0) && (B == 0.0)) { - dst.s.hi = A + B; - dst.s.lo = 0.0; - return dst.ld; - } - - /* If either operand is NaN or infinity: */ - const doublebits abits = { .d = A }; - const doublebits bbits = { .d = B }; - if ((((uint32_t)(abits.x >> 32) & infinityHi) == infinityHi) || - (((uint32_t)(bbits.x >> 32) & infinityHi) == infinityHi)) { - dst.s.hi = A + B; - dst.s.lo = 0.0; - return dst.ld; - } - - /* If the computation overflows: */ - /* This may be playing things a little bit fast and loose, but it will do for a start. */ - const double testForOverflow = A + (B + (a + b)); - const doublebits testbits = { .d = testForOverflow }; - if (((uint32_t)(testbits.x >> 32) & infinityHi) == infinityHi) { - dst.s.hi = testForOverflow; - dst.s.lo = 0.0; - return dst.ld; - } - - double H, h; - double T, t; - double W, w; - double Y; - - H = B + (A - (A + B)); - T = b + (a - (a + b)); - h = A + (B - (A + B)); - t = a + (b - (a + b)); - - if (local_fabs(A) <= local_fabs(B)) - w = (a + b) + h; - else - w = (a + b) + H; - - W = (A + B) + w; - Y = (A + B) - W; - Y += w; - - if (local_fabs(a) <= local_fabs(b)) - w = t + Y; - else - w = T + Y; - - dst.s.hi = Y = W + w; - dst.s.lo = (W - Y) + w; - - return dst.ld; +long double __gcc_qsub(long double x, long double y) { + static const uint32_t infinityHi = UINT32_C(0x7ff00000); + + DD dst = {.ld = x}, src = {.ld = y}; + + register double A = dst.s.hi, a = dst.s.lo, B = -src.s.hi, b = -src.s.lo; + + // If both operands are zero: + if ((A == 0.0) && (B == 0.0)) { + dst.s.hi = A + B; + dst.s.lo = 0.0; + return dst.ld; + } + + // If either operand is NaN or infinity: + const doublebits abits = {.d = A}; + const doublebits bbits = {.d = B}; + if ((((uint32_t)(abits.x >> 32) & infinityHi) == infinityHi) || + (((uint32_t)(bbits.x >> 32) & infinityHi) == infinityHi)) { + dst.s.hi = A + B; + dst.s.lo = 0.0; + return dst.ld; + } + + // If the computation overflows: + // This may be playing things a little bit fast and loose, but it will do for + // a start. + const double testForOverflow = A + (B + (a + b)); + const doublebits testbits = {.d = testForOverflow}; + if (((uint32_t)(testbits.x >> 32) & infinityHi) == infinityHi) { + dst.s.hi = testForOverflow; + dst.s.lo = 0.0; + return dst.ld; + } + + double H, h; + double T, t; + double W, w; + double Y; + + H = B + (A - (A + B)); + T = b + (a - (a + b)); + h = A + (B - (A + B)); + t = a + (b - (a + b)); + + if (local_fabs(A) <= local_fabs(B)) + w = (a + b) + h; + else + w = (a + b) + H; + + W = (A + B) + w; + Y = (A + B) - W; + Y += w; + + if (local_fabs(a) <= local_fabs(b)) + w = t + Y; + else + w = T + Y; + + dst.s.hi = Y = W + w; + dst.s.lo = (W - Y) + w; + + return dst.ld; } diff --git a/lib/builtins/ppc/multc3.c b/lib/builtins/ppc/multc3.c index 9dd79c975dde..f1fd6816d6c8 100644 --- a/lib/builtins/ppc/multc3.c +++ b/lib/builtins/ppc/multc3.c @@ -1,90 +1,85 @@ -/* This file is distributed under the University of Illinois Open Source - * License. See LICENSE.TXT for details. - */ +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception -#include "DD.h" #include "../int_math.h" +#include "DD.h" -#define makeFinite(x) { \ - (x).s.hi = crt_copysign(crt_isinf((x).s.hi) ? 1.0 : 0.0, (x).s.hi); \ - (x).s.lo = 0.0; \ +#define makeFinite(x) \ + { \ + (x).s.hi = crt_copysign(crt_isinf((x).s.hi) ? 1.0 : 0.0, (x).s.hi); \ + (x).s.lo = 0.0; \ } -#define zeroNaN(x) { \ - if (crt_isnan((x).s.hi)) { \ - (x).s.hi = crt_copysign(0.0, (x).s.hi); \ - (x).s.lo = 0.0; \ - } \ +#define zeroNaN(x) \ + { \ + if (crt_isnan((x).s.hi)) { \ + (x).s.hi = crt_copysign(0.0, (x).s.hi); \ + (x).s.lo = 0.0; \ + } \ } -long double _Complex -__multc3(long double a, long double b, long double c, long double d) -{ - long double ac = __gcc_qmul(a,c); - long double bd = __gcc_qmul(b,d); - long double ad = __gcc_qmul(a,d); - long double bc = __gcc_qmul(b,c); - - DD real = { .ld = __gcc_qsub(ac,bd) }; - DD imag = { .ld = __gcc_qadd(ad,bc) }; - - if (crt_isnan(real.s.hi) && crt_isnan(imag.s.hi)) - { - int recalc = 0; - - DD aDD = { .ld = a }; - DD bDD = { .ld = b }; - DD cDD = { .ld = c }; - DD dDD = { .ld = d }; - - if (crt_isinf(aDD.s.hi) || crt_isinf(bDD.s.hi)) - { - makeFinite(aDD); - makeFinite(bDD); - zeroNaN(cDD); - zeroNaN(dDD); - recalc = 1; - } - - if (crt_isinf(cDD.s.hi) || crt_isinf(dDD.s.hi)) - { - makeFinite(cDD); - makeFinite(dDD); - zeroNaN(aDD); - zeroNaN(bDD); - recalc = 1; - } - - if (!recalc) - { - DD acDD = { .ld = ac }; - DD bdDD = { .ld = bd }; - DD adDD = { .ld = ad }; - DD bcDD = { .ld = bc }; - - if (crt_isinf(acDD.s.hi) || crt_isinf(bdDD.s.hi) || - crt_isinf(adDD.s.hi) || crt_isinf(bcDD.s.hi)) - { - zeroNaN(aDD); - zeroNaN(bDD); - zeroNaN(cDD); - zeroNaN(dDD); - recalc = 1; - } - } - - if (recalc) - { - real.s.hi = CRT_INFINITY * (aDD.s.hi*cDD.s.hi - bDD.s.hi*dDD.s.hi); - real.s.lo = 0.0; - imag.s.hi = CRT_INFINITY * (aDD.s.hi*dDD.s.hi + bDD.s.hi*cDD.s.hi); - imag.s.lo = 0.0; - } - } - - long double _Complex z; - __real__ z = real.ld; - __imag__ z = imag.ld; - - return z; +long double _Complex __multc3(long double a, long double b, long double c, + long double d) { + long double ac = __gcc_qmul(a, c); + long double bd = __gcc_qmul(b, d); + long double ad = __gcc_qmul(a, d); + long double bc = __gcc_qmul(b, c); + + DD real = {.ld = __gcc_qsub(ac, bd)}; + DD imag = {.ld = __gcc_qadd(ad, bc)}; + + if (crt_isnan(real.s.hi) && crt_isnan(imag.s.hi)) { + int recalc = 0; + + DD aDD = {.ld = a}; + DD bDD = {.ld = b}; + DD cDD = {.ld = c}; + DD dDD = {.ld = d}; + + if (crt_isinf(aDD.s.hi) || crt_isinf(bDD.s.hi)) { + makeFinite(aDD); + makeFinite(bDD); + zeroNaN(cDD); + zeroNaN(dDD); + recalc = 1; + } + + if (crt_isinf(cDD.s.hi) || crt_isinf(dDD.s.hi)) { + makeFinite(cDD); + makeFinite(dDD); + zeroNaN(aDD); + zeroNaN(bDD); + recalc = 1; + } + + if (!recalc) { + DD acDD = {.ld = ac}; + DD bdDD = {.ld = bd}; + DD adDD = {.ld = ad}; + DD bcDD = {.ld = bc}; + + if (crt_isinf(acDD.s.hi) || crt_isinf(bdDD.s.hi) || + crt_isinf(adDD.s.hi) || crt_isinf(bcDD.s.hi)) { + zeroNaN(aDD); + zeroNaN(bDD); + zeroNaN(cDD); + zeroNaN(dDD); + recalc = 1; + } + } + + if (recalc) { + real.s.hi = CRT_INFINITY * (aDD.s.hi * cDD.s.hi - bDD.s.hi * dDD.s.hi); + real.s.lo = 0.0; + imag.s.hi = CRT_INFINITY * (aDD.s.hi * dDD.s.hi + bDD.s.hi * cDD.s.hi); + imag.s.lo = 0.0; + } + } + + long double _Complex z; + __real__ z = real.ld; + __imag__ z = imag.ld; + + return z; } diff --git a/lib/builtins/ppc/restFP.S b/lib/builtins/ppc/restFP.S index 507e756e18b1..02317bd6a649 100644 --- a/lib/builtins/ppc/restFP.S +++ b/lib/builtins/ppc/restFP.S @@ -1,9 +1,8 @@ //===-- restFP.S - Implement restFP ---------------------------------------===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// diff --git a/lib/builtins/ppc/saveFP.S b/lib/builtins/ppc/saveFP.S index 20b06fff53d9..1ef5532c8a83 100644 --- a/lib/builtins/ppc/saveFP.S +++ b/lib/builtins/ppc/saveFP.S @@ -1,9 +1,8 @@ //===-- saveFP.S - Implement saveFP ---------------------------------------===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// diff --git a/lib/builtins/riscv/mulsi3.S b/lib/builtins/riscv/mulsi3.S index a58d237040b6..5464919b26b9 100644 --- a/lib/builtins/riscv/mulsi3.S +++ b/lib/builtins/riscv/mulsi3.S @@ -1,9 +1,8 @@ //===--- mulsi3.S - Integer multiplication routines routines ---===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// diff --git a/lib/builtins/subdf3.c b/lib/builtins/subdf3.c index a892fa603cf2..5346dbc970fa 100644 --- a/lib/builtins/subdf3.c +++ b/lib/builtins/subdf3.c @@ -1,9 +1,8 @@ //===-- lib/adddf3.c - Double-precision subtraction ---------------*- C -*-===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// // @@ -16,17 +15,14 @@ #include "fp_lib.h" // Subtraction; flip the sign bit of b and add. -COMPILER_RT_ABI fp_t -__subdf3(fp_t a, fp_t b) { - return __adddf3(a, fromRep(toRep(b) ^ signBit)); +COMPILER_RT_ABI fp_t __subdf3(fp_t a, fp_t b) { + return __adddf3(a, fromRep(toRep(b) ^ signBit)); } #if defined(__ARM_EABI__) #if defined(COMPILER_RT_ARMHF_TARGET) -AEABI_RTABI fp_t __aeabi_dsub(fp_t a, fp_t b) { - return __subdf3(a, b); -} +AEABI_RTABI fp_t __aeabi_dsub(fp_t a, fp_t b) { return __subdf3(a, b); } #else -AEABI_RTABI fp_t __aeabi_dsub(fp_t a, fp_t b) COMPILER_RT_ALIAS(__subdf3); +COMPILER_RT_ALIAS(__subdf3, __aeabi_dsub) #endif #endif diff --git a/lib/builtins/subsf3.c b/lib/builtins/subsf3.c index 4b2786177dcc..85bde029b5bc 100644 --- a/lib/builtins/subsf3.c +++ b/lib/builtins/subsf3.c @@ -1,9 +1,8 @@ //===-- lib/subsf3.c - Single-precision subtraction ---------------*- C -*-===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// // @@ -16,17 +15,14 @@ #include "fp_lib.h" // Subtraction; flip the sign bit of b and add. -COMPILER_RT_ABI fp_t -__subsf3(fp_t a, fp_t b) { - return __addsf3(a, fromRep(toRep(b) ^ signBit)); +COMPILER_RT_ABI fp_t __subsf3(fp_t a, fp_t b) { + return __addsf3(a, fromRep(toRep(b) ^ signBit)); } #if defined(__ARM_EABI__) #if defined(COMPILER_RT_ARMHF_TARGET) -AEABI_RTABI fp_t __aeabi_fsub(fp_t a, fp_t b) { - return __subsf3(a, b); -} +AEABI_RTABI fp_t __aeabi_fsub(fp_t a, fp_t b) { return __subsf3(a, b); } #else -AEABI_RTABI fp_t __aeabi_fsub(fp_t a, fp_t b) COMPILER_RT_ALIAS(__subsf3); +COMPILER_RT_ALIAS(__subsf3, __aeabi_fsub) #endif #endif diff --git a/lib/builtins/subtf3.c b/lib/builtins/subtf3.c index 609b816f41e3..c96814692d2c 100644 --- a/lib/builtins/subtf3.c +++ b/lib/builtins/subtf3.c @@ -1,9 +1,8 @@ //===-- lib/subtf3.c - Quad-precision subtraction -----------------*- C -*-===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// // @@ -19,9 +18,8 @@ COMPILER_RT_ABI fp_t __addtf3(fp_t a, fp_t b); // Subtraction; flip the sign bit of b and add. -COMPILER_RT_ABI fp_t -__subtf3(fp_t a, fp_t b) { - return __addtf3(a, fromRep(toRep(b) ^ signBit)); +COMPILER_RT_ABI fp_t __subtf3(fp_t a, fp_t b) { + return __addtf3(a, fromRep(toRep(b) ^ signBit)); } #endif diff --git a/lib/builtins/subvdi3.c b/lib/builtins/subvdi3.c index 71fc70ffa92d..d7d78f1bb34f 100644 --- a/lib/builtins/subvdi3.c +++ b/lib/builtins/subvdi3.c @@ -1,36 +1,29 @@ -/* ===-- subvdi3.c - Implement __subvdi3 -----------------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - * - * This file implements __subvdi3 for the compiler_rt library. - * - * ===----------------------------------------------------------------------=== - */ +//===-- subvdi3.c - Implement __subvdi3 -----------------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements __subvdi3 for the compiler_rt library. +// +//===----------------------------------------------------------------------===// #include "int_lib.h" -/* Returns: a - b */ +// Returns: a - b -/* Effects: aborts if a - b overflows */ +// Effects: aborts if a - b overflows -COMPILER_RT_ABI di_int -__subvdi3(di_int a, di_int b) -{ - di_int s = (du_int) a - (du_int) b; - if (b >= 0) - { - if (s > a) - compilerrt_abort(); - } - else - { - if (s <= a) - compilerrt_abort(); - } - return s; +COMPILER_RT_ABI di_int __subvdi3(di_int a, di_int b) { + di_int s = (du_int)a - (du_int)b; + if (b >= 0) { + if (s > a) + compilerrt_abort(); + } else { + if (s <= a) + compilerrt_abort(); + } + return s; } diff --git a/lib/builtins/subvsi3.c b/lib/builtins/subvsi3.c index e6c0fb688c97..c3cb6e8a217c 100644 --- a/lib/builtins/subvsi3.c +++ b/lib/builtins/subvsi3.c @@ -1,36 +1,29 @@ -/* ===-- subvsi3.c - Implement __subvsi3 -----------------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - * - * This file implements __subvsi3 for the compiler_rt library. - * - * ===----------------------------------------------------------------------=== - */ +//===-- subvsi3.c - Implement __subvsi3 -----------------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements __subvsi3 for the compiler_rt library. +// +//===----------------------------------------------------------------------===// #include "int_lib.h" -/* Returns: a - b */ +// Returns: a - b -/* Effects: aborts if a - b overflows */ +// Effects: aborts if a - b overflows -COMPILER_RT_ABI si_int -__subvsi3(si_int a, si_int b) -{ - si_int s = (su_int) a - (su_int) b; - if (b >= 0) - { - if (s > a) - compilerrt_abort(); - } - else - { - if (s <= a) - compilerrt_abort(); - } - return s; +COMPILER_RT_ABI si_int __subvsi3(si_int a, si_int b) { + si_int s = (su_int)a - (su_int)b; + if (b >= 0) { + if (s > a) + compilerrt_abort(); + } else { + if (s <= a) + compilerrt_abort(); + } + return s; } diff --git a/lib/builtins/subvti3.c b/lib/builtins/subvti3.c index a6804d2d7b95..91ac18834043 100644 --- a/lib/builtins/subvti3.c +++ b/lib/builtins/subvti3.c @@ -1,40 +1,33 @@ -/* ===-- subvti3.c - Implement __subvti3 -----------------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - * - * This file implements __subvti3 for the compiler_rt library. - * - * ===----------------------------------------------------------------------=== - */ +//===-- subvti3.c - Implement __subvti3 -----------------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements __subvti3 for the compiler_rt library. +// +//===----------------------------------------------------------------------===// #include "int_lib.h" #ifdef CRT_HAS_128BIT -/* Returns: a - b */ +// Returns: a - b -/* Effects: aborts if a - b overflows */ +// Effects: aborts if a - b overflows -COMPILER_RT_ABI ti_int -__subvti3(ti_int a, ti_int b) -{ - ti_int s = (tu_int) a - (tu_int) b; - if (b >= 0) - { - if (s > a) - compilerrt_abort(); - } - else - { - if (s <= a) - compilerrt_abort(); - } - return s; +COMPILER_RT_ABI ti_int __subvti3(ti_int a, ti_int b) { + ti_int s = (tu_int)a - (tu_int)b; + if (b >= 0) { + if (s > a) + compilerrt_abort(); + } else { + if (s <= a) + compilerrt_abort(); + } + return s; } -#endif /* CRT_HAS_128BIT */ +#endif // CRT_HAS_128BIT diff --git a/lib/builtins/trampoline_setup.c b/lib/builtins/trampoline_setup.c index 25b627ab7658..a62431723d78 100644 --- a/lib/builtins/trampoline_setup.c +++ b/lib/builtins/trampoline_setup.c @@ -1,48 +1,43 @@ -/* ===----- trampoline_setup.c - Implement __trampoline_setup -------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - */ +//===----- trampoline_setup.c - Implement __trampoline_setup -------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// #include "int_lib.h" -extern void __clear_cache(void* start, void* end); +extern void __clear_cache(void *start, void *end); -/* - * The ppc compiler generates calls to __trampoline_setup() when creating - * trampoline functions on the stack for use with nested functions. - * This function creates a custom 40-byte trampoline function on the stack - * which loads r11 with a pointer to the outer function's locals - * and then jumps to the target nested function. - */ +// The ppc compiler generates calls to __trampoline_setup() when creating +// trampoline functions on the stack for use with nested functions. +// This function creates a custom 40-byte trampoline function on the stack +// which loads r11 with a pointer to the outer function's locals +// and then jumps to the target nested function. #if __ppc__ && !defined(__powerpc64__) -COMPILER_RT_ABI void -__trampoline_setup(uint32_t* trampOnStack, int trampSizeAllocated, - const void* realFunc, void* localsPtr) -{ - /* should never happen, but if compiler did not allocate */ - /* enough space on stack for the trampoline, abort */ - if ( trampSizeAllocated < 40 ) - compilerrt_abort(); - - /* create trampoline */ - trampOnStack[0] = 0x7c0802a6; /* mflr r0 */ - trampOnStack[1] = 0x4800000d; /* bl Lbase */ - trampOnStack[2] = (uint32_t)realFunc; - trampOnStack[3] = (uint32_t)localsPtr; - trampOnStack[4] = 0x7d6802a6; /* Lbase: mflr r11 */ - trampOnStack[5] = 0x818b0000; /* lwz r12,0(r11) */ - trampOnStack[6] = 0x7c0803a6; /* mtlr r0 */ - trampOnStack[7] = 0x7d8903a6; /* mtctr r12 */ - trampOnStack[8] = 0x816b0004; /* lwz r11,4(r11) */ - trampOnStack[9] = 0x4e800420; /* bctr */ - - /* clear instruction cache */ - __clear_cache(trampOnStack, &trampOnStack[10]); +COMPILER_RT_ABI void __trampoline_setup(uint32_t *trampOnStack, + int trampSizeAllocated, + const void *realFunc, void *localsPtr) { + // should never happen, but if compiler did not allocate + // enough space on stack for the trampoline, abort + if (trampSizeAllocated < 40) + compilerrt_abort(); + + // create trampoline + trampOnStack[0] = 0x7c0802a6; // mflr r0 + trampOnStack[1] = 0x4800000d; // bl Lbase + trampOnStack[2] = (uint32_t)realFunc; + trampOnStack[3] = (uint32_t)localsPtr; + trampOnStack[4] = 0x7d6802a6; // Lbase: mflr r11 + trampOnStack[5] = 0x818b0000; // lwz r12,0(r11) + trampOnStack[6] = 0x7c0803a6; // mtlr r0 + trampOnStack[7] = 0x7d8903a6; // mtctr r12 + trampOnStack[8] = 0x816b0004; // lwz r11,4(r11) + trampOnStack[9] = 0x4e800420; // bctr + + // clear instruction cache + __clear_cache(trampOnStack, &trampOnStack[10]); } -#endif /* __ppc__ && !defined(__powerpc64__) */ +#endif // __ppc__ && !defined(__powerpc64__) diff --git a/lib/builtins/truncdfhf2.c b/lib/builtins/truncdfhf2.c index 8354a41b8b6f..90c418a4387f 100644 --- a/lib/builtins/truncdfhf2.c +++ b/lib/builtins/truncdfhf2.c @@ -1,9 +1,8 @@ //===-- lib/truncdfhf2.c - double -> half conversion --------------*- C -*-===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// @@ -11,16 +10,12 @@ #define DST_HALF #include "fp_trunc_impl.inc" -COMPILER_RT_ABI uint16_t __truncdfhf2(double a) { - return __truncXfYf2__(a); -} +COMPILER_RT_ABI uint16_t __truncdfhf2(double a) { return __truncXfYf2__(a); } #if defined(__ARM_EABI__) #if defined(COMPILER_RT_ARMHF_TARGET) -AEABI_RTABI uint16_t __aeabi_d2h(double a) { - return __truncdfhf2(a); -} +AEABI_RTABI uint16_t __aeabi_d2h(double a) { return __truncdfhf2(a); } #else -AEABI_RTABI uint16_t __aeabi_d2h(double a) COMPILER_RT_ALIAS(__truncdfhf2); +COMPILER_RT_ALIAS(__truncdfhf2, __aeabi_d2h) #endif #endif diff --git a/lib/builtins/truncdfsf2.c b/lib/builtins/truncdfsf2.c index 195d3e0656e7..44a1299e30ae 100644 --- a/lib/builtins/truncdfsf2.c +++ b/lib/builtins/truncdfsf2.c @@ -1,9 +1,8 @@ //===-- lib/truncdfsf2.c - double -> single conversion ------------*- C -*-===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// @@ -11,16 +10,12 @@ #define DST_SINGLE #include "fp_trunc_impl.inc" -COMPILER_RT_ABI float __truncdfsf2(double a) { - return __truncXfYf2__(a); -} +COMPILER_RT_ABI float __truncdfsf2(double a) { return __truncXfYf2__(a); } #if defined(__ARM_EABI__) #if defined(COMPILER_RT_ARMHF_TARGET) -AEABI_RTABI float __aeabi_d2f(double a) { - return __truncdfsf2(a); -} +AEABI_RTABI float __aeabi_d2f(double a) { return __truncdfsf2(a); } #else -AEABI_RTABI float __aeabi_d2f(double a) COMPILER_RT_ALIAS(__truncdfsf2); +COMPILER_RT_ALIAS(__truncdfsf2, __aeabi_d2f) #endif #endif diff --git a/lib/builtins/truncsfhf2.c b/lib/builtins/truncsfhf2.c index 9c84ab4f938a..1f17194c38e5 100644 --- a/lib/builtins/truncsfhf2.c +++ b/lib/builtins/truncsfhf2.c @@ -1,9 +1,8 @@ //===-- lib/truncsfhf2.c - single -> half conversion --------------*- C -*-===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// @@ -14,19 +13,15 @@ // Use a forwarding definition and noinline to implement a poor man's alias, // as there isn't a good cross-platform way of defining one. COMPILER_RT_ABI NOINLINE uint16_t __truncsfhf2(float a) { - return __truncXfYf2__(a); + return __truncXfYf2__(a); } -COMPILER_RT_ABI uint16_t __gnu_f2h_ieee(float a) { - return __truncsfhf2(a); -} +COMPILER_RT_ABI uint16_t __gnu_f2h_ieee(float a) { return __truncsfhf2(a); } #if defined(__ARM_EABI__) #if defined(COMPILER_RT_ARMHF_TARGET) -AEABI_RTABI uint16_t __aeabi_f2h(float a) { - return __truncsfhf2(a); -} +AEABI_RTABI uint16_t __aeabi_f2h(float a) { return __truncsfhf2(a); } #else -AEABI_RTABI uint16_t __aeabi_f2h(float a) COMPILER_RT_ALIAS(__truncsfhf2); +COMPILER_RT_ALIAS(__truncsfhf2, __aeabi_f2h) #endif #endif diff --git a/lib/builtins/trunctfdf2.c b/lib/builtins/trunctfdf2.c index 741a71b33c5a..6857ea54d8a5 100644 --- a/lib/builtins/trunctfdf2.c +++ b/lib/builtins/trunctfdf2.c @@ -1,9 +1,8 @@ //===-- lib/truncdfsf2.c - quad -> double conversion --------------*- C -*-===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// @@ -15,8 +14,6 @@ #define DST_DOUBLE #include "fp_trunc_impl.inc" -COMPILER_RT_ABI double __trunctfdf2(long double a) { - return __truncXfYf2__(a); -} +COMPILER_RT_ABI double __trunctfdf2(long double a) { return __truncXfYf2__(a); } #endif diff --git a/lib/builtins/trunctfsf2.c b/lib/builtins/trunctfsf2.c index de96c1decf63..0261b1e90f5d 100644 --- a/lib/builtins/trunctfsf2.c +++ b/lib/builtins/trunctfsf2.c @@ -1,9 +1,8 @@ //===-- lib/trunctfsf2.c - quad -> single conversion --------------*- C -*-===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// @@ -15,8 +14,6 @@ #define DST_SINGLE #include "fp_trunc_impl.inc" -COMPILER_RT_ABI float __trunctfsf2(long double a) { - return __truncXfYf2__(a); -} +COMPILER_RT_ABI float __trunctfsf2(long double a) { return __truncXfYf2__(a); } #endif diff --git a/lib/builtins/ucmpdi2.c b/lib/builtins/ucmpdi2.c index 40af23613b1f..0f2edde8358b 100644 --- a/lib/builtins/ucmpdi2.c +++ b/lib/builtins/ucmpdi2.c @@ -1,51 +1,42 @@ -/* ===-- ucmpdi2.c - Implement __ucmpdi2 -----------------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - * - * This file implements __ucmpdi2 for the compiler_rt library. - * - * ===----------------------------------------------------------------------=== - */ +//===-- ucmpdi2.c - Implement __ucmpdi2 -----------------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements __ucmpdi2 for the compiler_rt library. +// +//===----------------------------------------------------------------------===// #include "int_lib.h" -/* Returns: if (a < b) returns 0 - * if (a == b) returns 1 - * if (a > b) returns 2 - */ +// Returns: if (a < b) returns 0 +// if (a == b) returns 1 +// if (a > b) returns 2 -COMPILER_RT_ABI si_int -__ucmpdi2(du_int a, du_int b) -{ - udwords x; - x.all = a; - udwords y; - y.all = b; - if (x.s.high < y.s.high) - return 0; - if (x.s.high > y.s.high) - return 2; - if (x.s.low < y.s.low) - return 0; - if (x.s.low > y.s.low) - return 2; - return 1; +COMPILER_RT_ABI si_int __ucmpdi2(du_int a, du_int b) { + udwords x; + x.all = a; + udwords y; + y.all = b; + if (x.s.high < y.s.high) + return 0; + if (x.s.high > y.s.high) + return 2; + if (x.s.low < y.s.low) + return 0; + if (x.s.low > y.s.low) + return 2; + return 1; } #ifdef __ARM_EABI__ -/* Returns: if (a < b) returns -1 -* if (a == b) returns 0 -* if (a > b) returns 1 -*/ -COMPILER_RT_ABI si_int -__aeabi_ulcmp(di_int a, di_int b) -{ - return __ucmpdi2(a, b) - 1; +// Returns: if (a < b) returns -1 +// if (a == b) returns 0 +// if (a > b) returns 1 +COMPILER_RT_ABI si_int __aeabi_ulcmp(di_int a, di_int b) { + return __ucmpdi2(a, b) - 1; } #endif - diff --git a/lib/builtins/ucmpti2.c b/lib/builtins/ucmpti2.c index bda8083bb2a1..4eb6655b05af 100644 --- a/lib/builtins/ucmpti2.c +++ b/lib/builtins/ucmpti2.c @@ -1,42 +1,37 @@ -/* ===-- ucmpti2.c - Implement __ucmpti2 -----------------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - * - * This file implements __ucmpti2 for the compiler_rt library. - * - * ===----------------------------------------------------------------------=== - */ +//===-- ucmpti2.c - Implement __ucmpti2 -----------------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements __ucmpti2 for the compiler_rt library. +// +//===----------------------------------------------------------------------===// #include "int_lib.h" #ifdef CRT_HAS_128BIT -/* Returns: if (a < b) returns 0 - * if (a == b) returns 1 - * if (a > b) returns 2 - */ +// Returns: if (a < b) returns 0 +// if (a == b) returns 1 +// if (a > b) returns 2 -COMPILER_RT_ABI si_int -__ucmpti2(tu_int a, tu_int b) -{ - utwords x; - x.all = a; - utwords y; - y.all = b; - if (x.s.high < y.s.high) - return 0; - if (x.s.high > y.s.high) - return 2; - if (x.s.low < y.s.low) - return 0; - if (x.s.low > y.s.low) - return 2; - return 1; +COMPILER_RT_ABI si_int __ucmpti2(tu_int a, tu_int b) { + utwords x; + x.all = a; + utwords y; + y.all = b; + if (x.s.high < y.s.high) + return 0; + if (x.s.high > y.s.high) + return 2; + if (x.s.low < y.s.low) + return 0; + if (x.s.low > y.s.low) + return 2; + return 1; } -#endif /* CRT_HAS_128BIT */ +#endif // CRT_HAS_128BIT diff --git a/lib/builtins/udivdi3.c b/lib/builtins/udivdi3.c index dc68e154b109..a23139ec947f 100644 --- a/lib/builtins/udivdi3.c +++ b/lib/builtins/udivdi3.c @@ -1,23 +1,19 @@ -/* ===-- udivdi3.c - Implement __udivdi3 -----------------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - * - * This file implements __udivdi3 for the compiler_rt library. - * - * ===----------------------------------------------------------------------=== - */ +//===-- udivdi3.c - Implement __udivdi3 -----------------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements __udivdi3 for the compiler_rt library. +// +//===----------------------------------------------------------------------===// #include "int_lib.h" -/* Returns: a / b */ +// Returns: a / b -COMPILER_RT_ABI du_int -__udivdi3(du_int a, du_int b) -{ - return __udivmoddi4(a, b, 0); +COMPILER_RT_ABI du_int __udivdi3(du_int a, du_int b) { + return __udivmoddi4(a, b, 0); } diff --git a/lib/builtins/udivmoddi4.c b/lib/builtins/udivmoddi4.c index 0c8b4ff46474..2914cc0fb46d 100644 --- a/lib/builtins/udivmoddi4.c +++ b/lib/builtins/udivmoddi4.c @@ -1,231 +1,189 @@ -/* ===-- udivmoddi4.c - Implement __udivmoddi4 -----------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - * - * This file implements __udivmoddi4 for the compiler_rt library. - * - * ===----------------------------------------------------------------------=== - */ +//===-- udivmoddi4.c - Implement __udivmoddi4 -----------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements __udivmoddi4 for the compiler_rt library. +// +//===----------------------------------------------------------------------===// #include "int_lib.h" -/* Effects: if rem != 0, *rem = a % b - * Returns: a / b - */ +// Effects: if rem != 0, *rem = a % b +// Returns: a / b -/* Translated from Figure 3-40 of The PowerPC Compiler Writer's Guide */ +// Translated from Figure 3-40 of The PowerPC Compiler Writer's Guide -COMPILER_RT_ABI du_int -__udivmoddi4(du_int a, du_int b, du_int* rem) -{ - const unsigned n_uword_bits = sizeof(su_int) * CHAR_BIT; - const unsigned n_udword_bits = sizeof(du_int) * CHAR_BIT; - udwords n; - n.all = a; - udwords d; - d.all = b; - udwords q; - udwords r; - unsigned sr; - /* special cases, X is unknown, K != 0 */ - if (n.s.high == 0) - { - if (d.s.high == 0) - { - /* 0 X - * --- - * 0 X - */ - if (rem) - *rem = n.s.low % d.s.low; - return n.s.low / d.s.low; - } - /* 0 X - * --- - * K X - */ - if (rem) - *rem = n.s.low; - return 0; +COMPILER_RT_ABI du_int __udivmoddi4(du_int a, du_int b, du_int *rem) { + const unsigned n_uword_bits = sizeof(su_int) * CHAR_BIT; + const unsigned n_udword_bits = sizeof(du_int) * CHAR_BIT; + udwords n; + n.all = a; + udwords d; + d.all = b; + udwords q; + udwords r; + unsigned sr; + // special cases, X is unknown, K != 0 + if (n.s.high == 0) { + if (d.s.high == 0) { + // 0 X + // --- + // 0 X + if (rem) + *rem = n.s.low % d.s.low; + return n.s.low / d.s.low; + } + // 0 X + // --- + // K X + if (rem) + *rem = n.s.low; + return 0; + } + // n.s.high != 0 + if (d.s.low == 0) { + if (d.s.high == 0) { + // K X + // --- + // 0 0 + if (rem) + *rem = n.s.high % d.s.low; + return n.s.high / d.s.low; + } + // d.s.high != 0 + if (n.s.low == 0) { + // K 0 + // --- + // K 0 + if (rem) { + r.s.high = n.s.high % d.s.high; + r.s.low = 0; + *rem = r.all; + } + return n.s.high / d.s.high; + } + // K K + // --- + // K 0 + if ((d.s.high & (d.s.high - 1)) == 0) /* if d is a power of 2 */ { + if (rem) { + r.s.low = n.s.low; + r.s.high = n.s.high & (d.s.high - 1); + *rem = r.all; + } + return n.s.high >> __builtin_ctz(d.s.high); } - /* n.s.high != 0 */ - if (d.s.low == 0) - { - if (d.s.high == 0) - { - /* K X - * --- - * 0 0 - */ - if (rem) - *rem = n.s.high % d.s.low; - return n.s.high / d.s.low; - } - /* d.s.high != 0 */ - if (n.s.low == 0) - { - /* K 0 - * --- - * K 0 - */ - if (rem) - { - r.s.high = n.s.high % d.s.high; - r.s.low = 0; - *rem = r.all; - } - return n.s.high / d.s.high; - } - /* K K - * --- - * K 0 - */ - if ((d.s.high & (d.s.high - 1)) == 0) /* if d is a power of 2 */ - { - if (rem) - { - r.s.low = n.s.low; - r.s.high = n.s.high & (d.s.high - 1); - *rem = r.all; - } - return n.s.high >> __builtin_ctz(d.s.high); - } - /* K K - * --- - * K 0 - */ - sr = __builtin_clz(d.s.high) - __builtin_clz(n.s.high); - /* 0 <= sr <= n_uword_bits - 2 or sr large */ - if (sr > n_uword_bits - 2) - { - if (rem) - *rem = n.all; - return 0; - } - ++sr; - /* 1 <= sr <= n_uword_bits - 1 */ - /* q.all = n.all << (n_udword_bits - sr); */ + // K K + // --- + // K 0 + sr = __builtin_clz(d.s.high) - __builtin_clz(n.s.high); + // 0 <= sr <= n_uword_bits - 2 or sr large + if (sr > n_uword_bits - 2) { + if (rem) + *rem = n.all; + return 0; + } + ++sr; + // 1 <= sr <= n_uword_bits - 1 + // q.all = n.all << (n_udword_bits - sr); + q.s.low = 0; + q.s.high = n.s.low << (n_uword_bits - sr); + // r.all = n.all >> sr; + r.s.high = n.s.high >> sr; + r.s.low = (n.s.high << (n_uword_bits - sr)) | (n.s.low >> sr); + } else /* d.s.low != 0 */ { + if (d.s.high == 0) { + // K X + // --- + // 0 K + if ((d.s.low & (d.s.low - 1)) == 0) /* if d is a power of 2 */ { + if (rem) + *rem = n.s.low & (d.s.low - 1); + if (d.s.low == 1) + return n.all; + sr = __builtin_ctz(d.s.low); + q.s.high = n.s.high >> sr; + q.s.low = (n.s.high << (n_uword_bits - sr)) | (n.s.low >> sr); + return q.all; + } + // K X + // --- + // 0 K + sr = 1 + n_uword_bits + __builtin_clz(d.s.low) - __builtin_clz(n.s.high); + // 2 <= sr <= n_udword_bits - 1 + // q.all = n.all << (n_udword_bits - sr); + // r.all = n.all >> sr; + if (sr == n_uword_bits) { + q.s.low = 0; + q.s.high = n.s.low; + r.s.high = 0; + r.s.low = n.s.high; + } else if (sr < n_uword_bits) /* 2 <= sr <= n_uword_bits - 1 */ { q.s.low = 0; q.s.high = n.s.low << (n_uword_bits - sr); - /* r.all = n.all >> sr; */ r.s.high = n.s.high >> sr; r.s.low = (n.s.high << (n_uword_bits - sr)) | (n.s.low >> sr); + } else /* n_uword_bits + 1 <= sr <= n_udword_bits - 1 */ { + q.s.low = n.s.low << (n_udword_bits - sr); + q.s.high = (n.s.high << (n_udword_bits - sr)) | + (n.s.low >> (sr - n_uword_bits)); + r.s.high = 0; + r.s.low = n.s.high >> (sr - n_uword_bits); + } + } else { + // K X + // --- + // K K + sr = __builtin_clz(d.s.high) - __builtin_clz(n.s.high); + // 0 <= sr <= n_uword_bits - 1 or sr large + if (sr > n_uword_bits - 1) { + if (rem) + *rem = n.all; + return 0; + } + ++sr; + // 1 <= sr <= n_uword_bits + // q.all = n.all << (n_udword_bits - sr); + q.s.low = 0; + if (sr == n_uword_bits) { + q.s.high = n.s.low; + r.s.high = 0; + r.s.low = n.s.high; + } else { + q.s.high = n.s.low << (n_uword_bits - sr); + r.s.high = n.s.high >> sr; + r.s.low = (n.s.high << (n_uword_bits - sr)) | (n.s.low >> sr); + } } - else /* d.s.low != 0 */ - { - if (d.s.high == 0) - { - /* K X - * --- - * 0 K - */ - if ((d.s.low & (d.s.low - 1)) == 0) /* if d is a power of 2 */ - { - if (rem) - *rem = n.s.low & (d.s.low - 1); - if (d.s.low == 1) - return n.all; - sr = __builtin_ctz(d.s.low); - q.s.high = n.s.high >> sr; - q.s.low = (n.s.high << (n_uword_bits - sr)) | (n.s.low >> sr); - return q.all; - } - /* K X - * --- - * 0 K - */ - sr = 1 + n_uword_bits + __builtin_clz(d.s.low) - __builtin_clz(n.s.high); - /* 2 <= sr <= n_udword_bits - 1 - * q.all = n.all << (n_udword_bits - sr); - * r.all = n.all >> sr; - */ - if (sr == n_uword_bits) - { - q.s.low = 0; - q.s.high = n.s.low; - r.s.high = 0; - r.s.low = n.s.high; - } - else if (sr < n_uword_bits) // 2 <= sr <= n_uword_bits - 1 - { - q.s.low = 0; - q.s.high = n.s.low << (n_uword_bits - sr); - r.s.high = n.s.high >> sr; - r.s.low = (n.s.high << (n_uword_bits - sr)) | (n.s.low >> sr); - } - else // n_uword_bits + 1 <= sr <= n_udword_bits - 1 - { - q.s.low = n.s.low << (n_udword_bits - sr); - q.s.high = (n.s.high << (n_udword_bits - sr)) | - (n.s.low >> (sr - n_uword_bits)); - r.s.high = 0; - r.s.low = n.s.high >> (sr - n_uword_bits); - } - } - else - { - /* K X - * --- - * K K - */ - sr = __builtin_clz(d.s.high) - __builtin_clz(n.s.high); - /* 0 <= sr <= n_uword_bits - 1 or sr large */ - if (sr > n_uword_bits - 1) - { - if (rem) - *rem = n.all; - return 0; - } - ++sr; - /* 1 <= sr <= n_uword_bits */ - /* q.all = n.all << (n_udword_bits - sr); */ - q.s.low = 0; - if (sr == n_uword_bits) - { - q.s.high = n.s.low; - r.s.high = 0; - r.s.low = n.s.high; - } - else - { - q.s.high = n.s.low << (n_uword_bits - sr); - r.s.high = n.s.high >> sr; - r.s.low = (n.s.high << (n_uword_bits - sr)) | (n.s.low >> sr); - } - } - } - /* Not a special case - * q and r are initialized with: - * q.all = n.all << (n_udword_bits - sr); - * r.all = n.all >> sr; - * 1 <= sr <= n_udword_bits - 1 - */ - su_int carry = 0; - for (; sr > 0; --sr) - { - /* r:q = ((r:q) << 1) | carry */ - r.s.high = (r.s.high << 1) | (r.s.low >> (n_uword_bits - 1)); - r.s.low = (r.s.low << 1) | (q.s.high >> (n_uword_bits - 1)); - q.s.high = (q.s.high << 1) | (q.s.low >> (n_uword_bits - 1)); - q.s.low = (q.s.low << 1) | carry; - /* carry = 0; - * if (r.all >= d.all) - * { - * r.all -= d.all; - * carry = 1; - * } - */ - const di_int s = (di_int)(d.all - r.all - 1) >> (n_udword_bits - 1); - carry = s & 1; - r.all -= d.all & s; - } - q.all = (q.all << 1) | carry; - if (rem) - *rem = r.all; - return q.all; + } + // Not a special case + // q and r are initialized with: + // q.all = n.all << (n_udword_bits - sr); + // r.all = n.all >> sr; + // 1 <= sr <= n_udword_bits - 1 + su_int carry = 0; + for (; sr > 0; --sr) { + // r:q = ((r:q) << 1) | carry + r.s.high = (r.s.high << 1) | (r.s.low >> (n_uword_bits - 1)); + r.s.low = (r.s.low << 1) | (q.s.high >> (n_uword_bits - 1)); + q.s.high = (q.s.high << 1) | (q.s.low >> (n_uword_bits - 1)); + q.s.low = (q.s.low << 1) | carry; + // carry = 0; + // if (r.all >= d.all) + // { + // r.all -= d.all; + // carry = 1; + // } + const di_int s = (di_int)(d.all - r.all - 1) >> (n_udword_bits - 1); + carry = s & 1; + r.all -= d.all & s; + } + q.all = (q.all << 1) | carry; + if (rem) + *rem = r.all; + return q.all; } diff --git a/lib/builtins/udivmodsi4.c b/lib/builtins/udivmodsi4.c index 789c4b5061e4..753ad6dd96db 100644 --- a/lib/builtins/udivmodsi4.c +++ b/lib/builtins/udivmodsi4.c @@ -1,27 +1,21 @@ -/*===-- udivmodsi4.c - Implement __udivmodsi4 ------------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - * - * This file implements __udivmodsi4 for the compiler_rt library. - * - * ===----------------------------------------------------------------------=== - */ +//===-- udivmodsi4.c - Implement __udivmodsi4 -----------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements __udivmodsi4 for the compiler_rt library. +// +//===----------------------------------------------------------------------===// #include "int_lib.h" -/* Returns: a / b, *rem = a % b */ +// Returns: a / b, *rem = a % b -COMPILER_RT_ABI su_int -__udivmodsi4(su_int a, su_int b, su_int* rem) -{ - si_int d = __udivsi3(a,b); - *rem = a - (d*b); +COMPILER_RT_ABI su_int __udivmodsi4(su_int a, su_int b, su_int *rem) { + si_int d = __udivsi3(a, b); + *rem = a - (d * b); return d; } - - diff --git a/lib/builtins/udivmodti4.c b/lib/builtins/udivmodti4.c index 803168849c6c..dd14a8b579ca 100644 --- a/lib/builtins/udivmodti4.c +++ b/lib/builtins/udivmodti4.c @@ -1,238 +1,195 @@ -/* ===-- udivmodti4.c - Implement __udivmodti4 -----------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - * - * This file implements __udivmodti4 for the compiler_rt library. - * - * ===----------------------------------------------------------------------=== - */ +//===-- udivmodti4.c - Implement __udivmodti4 -----------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements __udivmodti4 for the compiler_rt library. +// +//===----------------------------------------------------------------------===// #include "int_lib.h" #ifdef CRT_HAS_128BIT -/* Effects: if rem != 0, *rem = a % b - * Returns: a / b - */ +// Effects: if rem != 0, *rem = a % b +// Returns: a / b -/* Translated from Figure 3-40 of The PowerPC Compiler Writer's Guide */ +// Translated from Figure 3-40 of The PowerPC Compiler Writer's Guide -COMPILER_RT_ABI tu_int -__udivmodti4(tu_int a, tu_int b, tu_int* rem) -{ - const unsigned n_udword_bits = sizeof(du_int) * CHAR_BIT; - const unsigned n_utword_bits = sizeof(tu_int) * CHAR_BIT; - utwords n; - n.all = a; - utwords d; - d.all = b; - utwords q; - utwords r; - unsigned sr; - /* special cases, X is unknown, K != 0 */ - if (n.s.high == 0) - { - if (d.s.high == 0) - { - /* 0 X - * --- - * 0 X - */ - if (rem) - *rem = n.s.low % d.s.low; - return n.s.low / d.s.low; - } - /* 0 X - * --- - * K X - */ - if (rem) - *rem = n.s.low; - return 0; +COMPILER_RT_ABI tu_int __udivmodti4(tu_int a, tu_int b, tu_int *rem) { + const unsigned n_udword_bits = sizeof(du_int) * CHAR_BIT; + const unsigned n_utword_bits = sizeof(tu_int) * CHAR_BIT; + utwords n; + n.all = a; + utwords d; + d.all = b; + utwords q; + utwords r; + unsigned sr; + // special cases, X is unknown, K != 0 + if (n.s.high == 0) { + if (d.s.high == 0) { + // 0 X + // --- + // 0 X + if (rem) + *rem = n.s.low % d.s.low; + return n.s.low / d.s.low; + } + // 0 X + // --- + // K X + if (rem) + *rem = n.s.low; + return 0; + } + // n.s.high != 0 + if (d.s.low == 0) { + if (d.s.high == 0) { + // K X + // --- + // 0 0 + if (rem) + *rem = n.s.high % d.s.low; + return n.s.high / d.s.low; + } + // d.s.high != 0 + if (n.s.low == 0) { + // K 0 + // --- + // K 0 + if (rem) { + r.s.high = n.s.high % d.s.high; + r.s.low = 0; + *rem = r.all; + } + return n.s.high / d.s.high; + } + // K K + // --- + // K 0 + if ((d.s.high & (d.s.high - 1)) == 0) /* if d is a power of 2 */ { + if (rem) { + r.s.low = n.s.low; + r.s.high = n.s.high & (d.s.high - 1); + *rem = r.all; + } + return n.s.high >> __builtin_ctzll(d.s.high); } - /* n.s.high != 0 */ - if (d.s.low == 0) - { - if (d.s.high == 0) - { - /* K X - * --- - * 0 0 - */ - if (rem) - *rem = n.s.high % d.s.low; - return n.s.high / d.s.low; - } - /* d.s.high != 0 */ - if (n.s.low == 0) - { - /* K 0 - * --- - * K 0 - */ - if (rem) - { - r.s.high = n.s.high % d.s.high; - r.s.low = 0; - *rem = r.all; - } - return n.s.high / d.s.high; - } - /* K K - * --- - * K 0 - */ - if ((d.s.high & (d.s.high - 1)) == 0) /* if d is a power of 2 */ - { - if (rem) - { - r.s.low = n.s.low; - r.s.high = n.s.high & (d.s.high - 1); - *rem = r.all; - } - return n.s.high >> __builtin_ctzll(d.s.high); - } - /* K K - * --- - * K 0 - */ - sr = __builtin_clzll(d.s.high) - __builtin_clzll(n.s.high); - /* 0 <= sr <= n_udword_bits - 2 or sr large */ - if (sr > n_udword_bits - 2) - { - if (rem) - *rem = n.all; - return 0; - } - ++sr; - /* 1 <= sr <= n_udword_bits - 1 */ - /* q.all = n.all << (n_utword_bits - sr); */ + // K K + // --- + // K 0 + sr = __builtin_clzll(d.s.high) - __builtin_clzll(n.s.high); + // 0 <= sr <= n_udword_bits - 2 or sr large + if (sr > n_udword_bits - 2) { + if (rem) + *rem = n.all; + return 0; + } + ++sr; + // 1 <= sr <= n_udword_bits - 1 + // q.all = n.all << (n_utword_bits - sr); + q.s.low = 0; + q.s.high = n.s.low << (n_udword_bits - sr); + // r.all = n.all >> sr; + r.s.high = n.s.high >> sr; + r.s.low = (n.s.high << (n_udword_bits - sr)) | (n.s.low >> sr); + } else /* d.s.low != 0 */ { + if (d.s.high == 0) { + // K X + // --- + // 0 K + if ((d.s.low & (d.s.low - 1)) == 0) /* if d is a power of 2 */ { + if (rem) + *rem = n.s.low & (d.s.low - 1); + if (d.s.low == 1) + return n.all; + sr = __builtin_ctzll(d.s.low); + q.s.high = n.s.high >> sr; + q.s.low = (n.s.high << (n_udword_bits - sr)) | (n.s.low >> sr); + return q.all; + } + // K X + // --- + // 0 K + sr = 1 + n_udword_bits + __builtin_clzll(d.s.low) - + __builtin_clzll(n.s.high); + // 2 <= sr <= n_utword_bits - 1 + // q.all = n.all << (n_utword_bits - sr); + // r.all = n.all >> sr; + if (sr == n_udword_bits) { + q.s.low = 0; + q.s.high = n.s.low; + r.s.high = 0; + r.s.low = n.s.high; + } else if (sr < n_udword_bits) /* 2 <= sr <= n_udword_bits - 1 */ { q.s.low = 0; q.s.high = n.s.low << (n_udword_bits - sr); - /* r.all = n.all >> sr; */ r.s.high = n.s.high >> sr; r.s.low = (n.s.high << (n_udword_bits - sr)) | (n.s.low >> sr); + } else /* n_udword_bits + 1 <= sr <= n_utword_bits - 1 */ { + q.s.low = n.s.low << (n_utword_bits - sr); + q.s.high = (n.s.high << (n_utword_bits - sr)) | + (n.s.low >> (sr - n_udword_bits)); + r.s.high = 0; + r.s.low = n.s.high >> (sr - n_udword_bits); + } + } else { + // K X + // --- + // K K + sr = __builtin_clzll(d.s.high) - __builtin_clzll(n.s.high); + // 0 <= sr <= n_udword_bits - 1 or sr large + if (sr > n_udword_bits - 1) { + if (rem) + *rem = n.all; + return 0; + } + ++sr; + // 1 <= sr <= n_udword_bits + // q.all = n.all << (n_utword_bits - sr); + // r.all = n.all >> sr; + q.s.low = 0; + if (sr == n_udword_bits) { + q.s.high = n.s.low; + r.s.high = 0; + r.s.low = n.s.high; + } else { + r.s.high = n.s.high >> sr; + r.s.low = (n.s.high << (n_udword_bits - sr)) | (n.s.low >> sr); + q.s.high = n.s.low << (n_udword_bits - sr); + } } - else /* d.s.low != 0 */ - { - if (d.s.high == 0) - { - /* K X - * --- - * 0 K - */ - if ((d.s.low & (d.s.low - 1)) == 0) /* if d is a power of 2 */ - { - if (rem) - *rem = n.s.low & (d.s.low - 1); - if (d.s.low == 1) - return n.all; - sr = __builtin_ctzll(d.s.low); - q.s.high = n.s.high >> sr; - q.s.low = (n.s.high << (n_udword_bits - sr)) | (n.s.low >> sr); - return q.all; - } - /* K X - * --- - * 0 K - */ - sr = 1 + n_udword_bits + __builtin_clzll(d.s.low) - - __builtin_clzll(n.s.high); - /* 2 <= sr <= n_utword_bits - 1 - * q.all = n.all << (n_utword_bits - sr); - * r.all = n.all >> sr; - */ - if (sr == n_udword_bits) - { - q.s.low = 0; - q.s.high = n.s.low; - r.s.high = 0; - r.s.low = n.s.high; - } - else if (sr < n_udword_bits) // 2 <= sr <= n_udword_bits - 1 - { - q.s.low = 0; - q.s.high = n.s.low << (n_udword_bits - sr); - r.s.high = n.s.high >> sr; - r.s.low = (n.s.high << (n_udword_bits - sr)) | (n.s.low >> sr); - } - else // n_udword_bits + 1 <= sr <= n_utword_bits - 1 - { - q.s.low = n.s.low << (n_utword_bits - sr); - q.s.high = (n.s.high << (n_utword_bits - sr)) | - (n.s.low >> (sr - n_udword_bits)); - r.s.high = 0; - r.s.low = n.s.high >> (sr - n_udword_bits); - } - } - else - { - /* K X - * --- - * K K - */ - sr = __builtin_clzll(d.s.high) - __builtin_clzll(n.s.high); - /*0 <= sr <= n_udword_bits - 1 or sr large */ - if (sr > n_udword_bits - 1) - { - if (rem) - *rem = n.all; - return 0; - } - ++sr; - /* 1 <= sr <= n_udword_bits - * q.all = n.all << (n_utword_bits - sr); - * r.all = n.all >> sr; - */ - q.s.low = 0; - if (sr == n_udword_bits) - { - q.s.high = n.s.low; - r.s.high = 0; - r.s.low = n.s.high; - } - else - { - r.s.high = n.s.high >> sr; - r.s.low = (n.s.high << (n_udword_bits - sr)) | (n.s.low >> sr); - q.s.high = n.s.low << (n_udword_bits - sr); - } - } - } - /* Not a special case - * q and r are initialized with: - * q.all = n.all << (n_utword_bits - sr); - * r.all = n.all >> sr; - * 1 <= sr <= n_utword_bits - 1 - */ - su_int carry = 0; - for (; sr > 0; --sr) - { - /* r:q = ((r:q) << 1) | carry */ - r.s.high = (r.s.high << 1) | (r.s.low >> (n_udword_bits - 1)); - r.s.low = (r.s.low << 1) | (q.s.high >> (n_udword_bits - 1)); - q.s.high = (q.s.high << 1) | (q.s.low >> (n_udword_bits - 1)); - q.s.low = (q.s.low << 1) | carry; - /* carry = 0; - * if (r.all >= d.all) - * { - * r.all -= d.all; - * carry = 1; - * } - */ - const ti_int s = (ti_int)(d.all - r.all - 1) >> (n_utword_bits - 1); - carry = s & 1; - r.all -= d.all & s; - } - q.all = (q.all << 1) | carry; - if (rem) - *rem = r.all; - return q.all; + } + // Not a special case + // q and r are initialized with: + // q.all = n.all << (n_utword_bits - sr); + // r.all = n.all >> sr; + // 1 <= sr <= n_utword_bits - 1 + su_int carry = 0; + for (; sr > 0; --sr) { + // r:q = ((r:q) << 1) | carry + r.s.high = (r.s.high << 1) | (r.s.low >> (n_udword_bits - 1)); + r.s.low = (r.s.low << 1) | (q.s.high >> (n_udword_bits - 1)); + q.s.high = (q.s.high << 1) | (q.s.low >> (n_udword_bits - 1)); + q.s.low = (q.s.low << 1) | carry; + // carry = 0; + // if (r.all >= d.all) + // { + // r.all -= d.all; + // carry = 1; + // } + const ti_int s = (ti_int)(d.all - r.all - 1) >> (n_utword_bits - 1); + carry = s & 1; + r.all -= d.all & s; + } + q.all = (q.all << 1) | carry; + if (rem) + *rem = r.all; + return q.all; } -#endif /* CRT_HAS_128BIT */ +#endif // CRT_HAS_128BIT diff --git a/lib/builtins/udivsi3.c b/lib/builtins/udivsi3.c index bb720f8c382b..18cc96c1b2e0 100644 --- a/lib/builtins/udivsi3.c +++ b/lib/builtins/udivsi3.c @@ -1,68 +1,62 @@ -/* ===-- udivsi3.c - Implement __udivsi3 -----------------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - * - * This file implements __udivsi3 for the compiler_rt library. - * - * ===----------------------------------------------------------------------=== - */ +//===-- udivsi3.c - Implement __udivsi3 -----------------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements __udivsi3 for the compiler_rt library. +// +//===----------------------------------------------------------------------===// #include "int_lib.h" -/* Returns: a / b */ +// Returns: a / b -/* Translated from Figure 3-40 of The PowerPC Compiler Writer's Guide */ +// Translated from Figure 3-40 of The PowerPC Compiler Writer's Guide -/* This function should not call __divsi3! */ -COMPILER_RT_ABI su_int -__udivsi3(su_int n, su_int d) -{ - const unsigned n_uword_bits = sizeof(su_int) * CHAR_BIT; - su_int q; - su_int r; - unsigned sr; - /* special cases */ - if (d == 0) - return 0; /* ?! */ - if (n == 0) - return 0; - sr = __builtin_clz(d) - __builtin_clz(n); - /* 0 <= sr <= n_uword_bits - 1 or sr large */ - if (sr > n_uword_bits - 1) /* d > r */ - return 0; - if (sr == n_uword_bits - 1) /* d == 1 */ - return n; - ++sr; - /* 1 <= sr <= n_uword_bits - 1 */ - /* Not a special case */ - q = n << (n_uword_bits - sr); - r = n >> sr; - su_int carry = 0; - for (; sr > 0; --sr) - { - /* r:q = ((r:q) << 1) | carry */ - r = (r << 1) | (q >> (n_uword_bits - 1)); - q = (q << 1) | carry; - /* carry = 0; - * if (r.all >= d.all) - * { - * r.all -= d.all; - * carry = 1; - * } - */ - const si_int s = (si_int)(d - r - 1) >> (n_uword_bits - 1); - carry = s & 1; - r -= d & s; - } +// This function should not call __divsi3! +COMPILER_RT_ABI su_int __udivsi3(su_int n, su_int d) { + const unsigned n_uword_bits = sizeof(su_int) * CHAR_BIT; + su_int q; + su_int r; + unsigned sr; + // special cases + if (d == 0) + return 0; // ?! + if (n == 0) + return 0; + sr = __builtin_clz(d) - __builtin_clz(n); + // 0 <= sr <= n_uword_bits - 1 or sr large + if (sr > n_uword_bits - 1) // d > r + return 0; + if (sr == n_uword_bits - 1) // d == 1 + return n; + ++sr; + // 1 <= sr <= n_uword_bits - 1 + // Not a special case + q = n << (n_uword_bits - sr); + r = n >> sr; + su_int carry = 0; + for (; sr > 0; --sr) { + // r:q = ((r:q) << 1) | carry + r = (r << 1) | (q >> (n_uword_bits - 1)); q = (q << 1) | carry; - return q; + // carry = 0; + // if (r.all >= d.all) + // { + // r.all -= d.all; + // carry = 1; + // } + const si_int s = (si_int)(d - r - 1) >> (n_uword_bits - 1); + carry = s & 1; + r -= d & s; + } + q = (q << 1) | carry; + return q; } #if defined(__ARM_EABI__) -AEABI_RTABI su_int __aeabi_uidiv(su_int n, su_int d) COMPILER_RT_ALIAS(__udivsi3); +COMPILER_RT_ALIAS(__udivsi3, __aeabi_uidiv) #endif diff --git a/lib/builtins/udivti3.c b/lib/builtins/udivti3.c index ec94673e25b0..4c82040b88cb 100644 --- a/lib/builtins/udivti3.c +++ b/lib/builtins/udivti3.c @@ -1,27 +1,23 @@ -/* ===-- udivti3.c - Implement __udivti3 -----------------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - * - * This file implements __udivti3 for the compiler_rt library. - * - * ===----------------------------------------------------------------------=== - */ +//===-- udivti3.c - Implement __udivti3 -----------------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements __udivti3 for the compiler_rt library. +// +//===----------------------------------------------------------------------===// #include "int_lib.h" #ifdef CRT_HAS_128BIT -/* Returns: a / b */ +// Returns: a / b -COMPILER_RT_ABI tu_int -__udivti3(tu_int a, tu_int b) -{ - return __udivmodti4(a, b, 0); +COMPILER_RT_ABI tu_int __udivti3(tu_int a, tu_int b) { + return __udivmodti4(a, b, 0); } -#endif /* CRT_HAS_128BIT */ +#endif // CRT_HAS_128BIT diff --git a/lib/builtins/umoddi3.c b/lib/builtins/umoddi3.c index d513f080a1e9..965cf8fc01bd 100644 --- a/lib/builtins/umoddi3.c +++ b/lib/builtins/umoddi3.c @@ -1,25 +1,21 @@ -/* ===-- umoddi3.c - Implement __umoddi3 -----------------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - * - * This file implements __umoddi3 for the compiler_rt library. - * - * ===----------------------------------------------------------------------=== - */ +//===-- umoddi3.c - Implement __umoddi3 -----------------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements __umoddi3 for the compiler_rt library. +// +//===----------------------------------------------------------------------===// #include "int_lib.h" -/* Returns: a % b */ +// Returns: a % b -COMPILER_RT_ABI du_int -__umoddi3(du_int a, du_int b) -{ - du_int r; - __udivmoddi4(a, b, &r); - return r; +COMPILER_RT_ABI du_int __umoddi3(du_int a, du_int b) { + du_int r; + __udivmoddi4(a, b, &r); + return r; } diff --git a/lib/builtins/umodsi3.c b/lib/builtins/umodsi3.c index d5fda4a6af1f..ce9abcd94ef7 100644 --- a/lib/builtins/umodsi3.c +++ b/lib/builtins/umodsi3.c @@ -1,23 +1,19 @@ -/* ===-- umodsi3.c - Implement __umodsi3 -----------------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - * - * This file implements __umodsi3 for the compiler_rt library. - * - * ===----------------------------------------------------------------------=== - */ +//===-- umodsi3.c - Implement __umodsi3 -----------------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements __umodsi3 for the compiler_rt library. +// +//===----------------------------------------------------------------------===// #include "int_lib.h" -/* Returns: a % b */ +// Returns: a % b -COMPILER_RT_ABI su_int -__umodsi3(su_int a, su_int b) -{ - return a - __udivsi3(a, b) * b; +COMPILER_RT_ABI su_int __umodsi3(su_int a, su_int b) { + return a - __udivsi3(a, b) * b; } diff --git a/lib/builtins/umodti3.c b/lib/builtins/umodti3.c index 6d1ca7a8cf6f..8cc5cb6b88f7 100644 --- a/lib/builtins/umodti3.c +++ b/lib/builtins/umodti3.c @@ -1,29 +1,25 @@ -/* ===-- umodti3.c - Implement __umodti3 -----------------------------------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===----------------------------------------------------------------------=== - * - * This file implements __umodti3 for the compiler_rt library. - * - * ===----------------------------------------------------------------------=== - */ +//===-- umodti3.c - Implement __umodti3 -----------------------------------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===----------------------------------------------------------------------===// +// +// This file implements __umodti3 for the compiler_rt library. +// +//===----------------------------------------------------------------------===// #include "int_lib.h" #ifdef CRT_HAS_128BIT -/* Returns: a % b */ +// Returns: a % b -COMPILER_RT_ABI tu_int -__umodti3(tu_int a, tu_int b) -{ - tu_int r; - __udivmodti4(a, b, &r); - return r; +COMPILER_RT_ABI tu_int __umodti3(tu_int a, tu_int b) { + tu_int r; + __udivmodti4(a, b, &r); + return r; } -#endif /* CRT_HAS_128BIT */ +#endif // CRT_HAS_128BIT diff --git a/lib/builtins/unwind-ehabi-helpers.h b/lib/builtins/unwind-ehabi-helpers.h index ccb0765975a9..1b48cdb75e1b 100644 --- a/lib/builtins/unwind-ehabi-helpers.h +++ b/lib/builtins/unwind-ehabi-helpers.h @@ -1,43 +1,40 @@ -/* ===-- arm-ehabi-helpers.h - Supplementary ARM EHABI declarations --------=== - * - * The LLVM Compiler Infrastructure - * - * This file is dual licensed under the MIT and the University of Illinois Open - * Source Licenses. See LICENSE.TXT for details. - * - * ===--------------------------------------------------------------------=== */ +//===-- arm-ehabi-helpers.h - Supplementary ARM EHABI declarations --------===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===--------------------------------------------------------------------===// #ifndef UNWIND_EHABI_HELPERS_H #define UNWIND_EHABI_HELPERS_H #include <stdint.h> -/* NOTE: see reasoning for this inclusion below */ +// NOTE: see reasoning for this inclusion below #include <unwind.h> #if !defined(__ARM_EABI_UNWINDER__) -/* - * NOTE: _URC_OK, _URC_FAILURE must be present as preprocessor tokens. This - * allows for a substitution of a constant which can be cast into the - * appropriate enumerated type. This header is expected to always be included - * AFTER unwind.h (which is why it is forcefully included above). This ensures - * that we do not overwrite the token for the enumeration. Subsequent uses of - * the token would be clean to rewrite with constant values. - * - * The typedef redeclaration should be safe. Due to the protection granted to - * us by the `__ARM_EABI_UNWINDER__` above, we are guaranteed that we are in a - * header not vended by gcc. The HP unwinder (being an itanium unwinder) does - * not support EHABI, and the GNU unwinder, derived from the HP unwinder, also - * does not support EHABI as of the introduction of this header. As such, we - * are fairly certain that we are in the LLVM case. Here, _Unwind_State is a - * typedef, and so we can get away with a redeclaration. - * - * Guarded redefinitions of the needed unwind state prevent the redefinition of - * those states. - */ - -#define _URC_OK 0 -#define _URC_FAILURE 9 +// NOTE: _URC_OK, _URC_FAILURE must be present as preprocessor tokens. This +// allows for a substitution of a constant which can be cast into the +// appropriate enumerated type. This header is expected to always be included +// AFTER unwind.h (which is why it is forcefully included above). This ensures +// that we do not overwrite the token for the enumeration. Subsequent uses of +// the token would be clean to rewrite with constant values. +// +// The typedef redeclaration should be safe. Due to the protection granted to +// us by the `__ARM_EABI_UNWINDER__` above, we are guaranteed that we are in a +// header not vended by gcc. The HP unwinder (being an itanium unwinder) does +// not support EHABI, and the GNU unwinder, derived from the HP unwinder, also +// does not support EHABI as of the introduction of this header. As such, we +// are fairly certain that we are in the LLVM case. Here, _Unwind_State is a +// typedef, and so we can get away with a redeclaration. +// +// Guarded redefinitions of the needed unwind state prevent the redefinition of +// those states. + +#define _URC_OK 0 +#define _URC_FAILURE 9 typedef uint32_t _Unwind_State; @@ -52,4 +49,3 @@ typedef uint32_t _Unwind_State; #endif #endif - diff --git a/lib/builtins/x86_64/chkstk.S b/lib/builtins/x86_64/chkstk.S index 4149ac63d9d0..ad7953a116ac 100644 --- a/lib/builtins/x86_64/chkstk.S +++ b/lib/builtins/x86_64/chkstk.S @@ -1,5 +1,6 @@ -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception #include "../assembly.h" diff --git a/lib/builtins/x86_64/chkstk2.S b/lib/builtins/x86_64/chkstk2.S index ac1eb920e0e8..33d10d5b63be 100644 --- a/lib/builtins/x86_64/chkstk2.S +++ b/lib/builtins/x86_64/chkstk2.S @@ -1,5 +1,6 @@ -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception #include "../assembly.h" diff --git a/lib/builtins/x86_64/floatdidf.c b/lib/builtins/x86_64/floatdidf.c index dead0ed42c65..f83f53a38e9e 100644 --- a/lib/builtins/x86_64/floatdidf.c +++ b/lib/builtins/x86_64/floatdidf.c @@ -1,16 +1,13 @@ -/* This file is distributed under the University of Illinois Open Source - * License. See LICENSE.TXT for details. - */ +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception -/* double __floatdidf(di_int a); */ +// double __floatdidf(di_int a); #if defined(__x86_64__) || defined(_M_X64) #include "../int_lib.h" -double __floatdidf(int64_t a) -{ - return (double)a; -} +double __floatdidf(int64_t a) { return (double)a; } -#endif /* __x86_64__ */ +#endif // __x86_64__ diff --git a/lib/builtins/x86_64/floatdisf.c b/lib/builtins/x86_64/floatdisf.c index 99d5621c6327..06c118cfa18d 100644 --- a/lib/builtins/x86_64/floatdisf.c +++ b/lib/builtins/x86_64/floatdisf.c @@ -1,14 +1,11 @@ -/* This file is distributed under the University of Illinois Open Source - * License. See LICENSE.TXT for details. - */ +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception #if defined(__x86_64__) || defined(_M_X64) #include "../int_lib.h" -float __floatdisf(int64_t a) -{ - return (float)a; -} +float __floatdisf(int64_t a) { return (float)a; } -#endif /* __x86_64__ */ +#endif // __x86_64__ diff --git a/lib/builtins/x86_64/floatdixf.c b/lib/builtins/x86_64/floatdixf.c index c01193a82b5e..cf8450ce6f42 100644 --- a/lib/builtins/x86_64/floatdixf.c +++ b/lib/builtins/x86_64/floatdixf.c @@ -1,16 +1,13 @@ -/* This file is distributed under the University of Illinois Open Source - * License. See LICENSE.TXT for details. - */ +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception -/* long double __floatdixf(di_int a); */ +// long double __floatdixf(di_int a); #ifdef __x86_64__ #include "../int_lib.h" -long double __floatdixf(int64_t a) -{ - return (long double)a; -} +long double __floatdixf(int64_t a) { return (long double)a; } -#endif /* __i386__ */ +#endif // __i386__ diff --git a/lib/builtins/x86_64/floatundidf.S b/lib/builtins/x86_64/floatundidf.S index 094a68dc3cd4..7f6ef3bbb6bf 100644 --- a/lib/builtins/x86_64/floatundidf.S +++ b/lib/builtins/x86_64/floatundidf.S @@ -1,9 +1,8 @@ //===-- floatundidf.S - Implement __floatundidf for x86_64 ----------------===// // -// The LLVM Compiler Infrastructure -// -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// // diff --git a/lib/builtins/x86_64/floatundisf.S b/lib/builtins/x86_64/floatundisf.S index 7c9f75e188eb..246bdff5a4c3 100644 --- a/lib/builtins/x86_64/floatundisf.S +++ b/lib/builtins/x86_64/floatundisf.S @@ -1,5 +1,6 @@ -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception #include "../assembly.h" @@ -23,7 +24,7 @@ DEFINE_COMPILERRT_FUNCTION(__floatundisf) js 1f cvtsi2ssq %rdi, %xmm0 ret - + 1: andq %rdi, %rsi shrq %rdi orq %rsi, %rdi diff --git a/lib/builtins/x86_64/floatundixf.S b/lib/builtins/x86_64/floatundixf.S index 28a096b71373..9e3bcedcb7e4 100644 --- a/lib/builtins/x86_64/floatundixf.S +++ b/lib/builtins/x86_64/floatundixf.S @@ -1,5 +1,6 @@ -// This file is dual licensed under the MIT and the University of Illinois Open -// Source Licenses. See LICENSE.TXT for details. +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception #include "../assembly.h" @@ -57,7 +58,7 @@ DEFINE_COMPILERRT_FUNCTION(__floatundixf) orq REL_ADDR(twop52), %rsi // 2^52 + lo (as a double) movq %rdi, -8(%rsp) movq %rsi, -16(%rsp) - fldl REL_ADDR(twop84_plus_twop52_neg) + fldl REL_ADDR(twop84_plus_twop52_neg) faddl -8(%rsp) // hi - 2^52 (as double extended, no rounding occurs) faddl -16(%rsp) // hi + lo (as double extended) ret |