diff options
| author | Roman Divacky <rdivacky@FreeBSD.org> | 2010-03-03 17:27:15 +0000 |
|---|---|---|
| committer | Roman Divacky <rdivacky@FreeBSD.org> | 2010-03-03 17:27:15 +0000 |
| commit | 67a71b3184ce20a901e874d0ee25e01397dd87ef (patch) | |
| tree | 836a05cff50ca46176117b86029f061fa4db54f0 /test/CodeGen/X86 | |
| parent | 6fe5c7aa327e188b7176daa5595bbf075a6b94df (diff) | |
Notes
Diffstat (limited to 'test/CodeGen/X86')
36 files changed, 1338 insertions, 324 deletions
diff --git a/test/CodeGen/X86/2005-01-17-CycleInDAG.ll b/test/CodeGen/X86/2005-01-17-CycleInDAG.ll index 32fafc61e8de..fe6674da041f 100644 --- a/test/CodeGen/X86/2005-01-17-CycleInDAG.ll +++ b/test/CodeGen/X86/2005-01-17-CycleInDAG.ll @@ -7,7 +7,7 @@ @GLOBAL = external global i32 ; <i32*> [#uses=1] -define i32 @test(i32* %P1, i32* %P2, i32* %P3) { +define i32 @test(i32* %P1, i32* %P2, i32* %P3) nounwind { %L = load i32* @GLOBAL ; <i32> [#uses=1] store i32 12, i32* %P2 %Y = load i32* %P3 ; <i32> [#uses=1] diff --git a/test/CodeGen/X86/2006-10-07-ScalarSSEMiscompile.ll b/test/CodeGen/X86/2006-10-07-ScalarSSEMiscompile.ll index bf9fa5782b06..d09d06147696 100644 --- a/test/CodeGen/X86/2006-10-07-ScalarSSEMiscompile.ll +++ b/test/CodeGen/X86/2006-10-07-ScalarSSEMiscompile.ll @@ -5,7 +5,7 @@ target datalayout = "e-p:32:32" target triple = "i686-apple-darwin8.7.2" -define <4 x float> @test(<4 x float> %A, <4 x float>* %B) { +define <4 x float> @test(<4 x float> %A, <4 x float>* %B) nounwind { %BV = load <4 x float>* %B ; <<4 x float>> [#uses=1] %tmp28 = tail call <4 x float> @llvm.x86.sse.sub.ss( <4 x float> %A, <4 x float> %BV ) ; <<4 x float>> [#uses=1] ret <4 x float> %tmp28 diff --git a/test/CodeGen/X86/2008-02-22-ReMatBug.ll b/test/CodeGen/X86/2008-02-22-ReMatBug.ll index 8d6bb0df1f6d..a91ac27f98dd 100644 --- a/test/CodeGen/X86/2008-02-22-ReMatBug.ll +++ b/test/CodeGen/X86/2008-02-22-ReMatBug.ll @@ -1,4 +1,4 @@ -; RUN: llc < %s -march=x86 -stats |& grep {Number of re-materialization} | grep 3 +; RUN: llc < %s -march=x86 -stats |& grep {Number of re-materialization} | grep 2 ; rdar://5761454 %struct.quad_struct = type { i32, i32, %struct.quad_struct*, %struct.quad_struct*, %struct.quad_struct*, %struct.quad_struct*, %struct.quad_struct* } diff --git a/test/CodeGen/X86/2009-02-12-DebugInfoVLA.ll b/test/CodeGen/X86/2009-02-12-DebugInfoVLA.ll index 72c7ee93a9d2..0dca14d064eb 100644 --- a/test/CodeGen/X86/2009-02-12-DebugInfoVLA.ll +++ b/test/CodeGen/X86/2009-02-12-DebugInfoVLA.ll @@ -3,74 +3,83 @@ ; PR3538 target datalayout = "e-p:32:32:32-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:32:64-f32:32:32-f64:32:64-v64:64:64-v128:128:128-a0:0:64-f80:128:128" target triple = "i386-apple-darwin9" - %llvm.dbg.anchor.type = type { i32, i32 } - %llvm.dbg.basictype.type = type { i32, { }*, i8*, { }*, i32, i64, i64, i64, i32, i32 } - %llvm.dbg.block.type = type { i32, { }* } - %llvm.dbg.compile_unit.type = type { i32, { }*, i32, i8*, i8*, i8*, i1, i1, i8* } - %llvm.dbg.composite.type = type { i32, { }*, i8*, { }*, i32, i64, i64, i64, i32, { }*, { }* } - %llvm.dbg.subprogram.type = type { i32, { }*, { }*, i8*, i8*, i8*, { }*, i32, { }*, i1, i1 } - %llvm.dbg.subrange.type = type { i32, i64, i64 } - %llvm.dbg.variable.type = type { i32, { }*, i8*, { }*, i32, { }* } -@llvm.dbg.compile_units = linkonce constant %llvm.dbg.anchor.type { i32 458752, i32 17 }, section "llvm.metadata" ; <%llvm.dbg.anchor.type*> [#uses=1] -@.str = internal constant [4 x i8] c"t.c\00", section "llvm.metadata" ; <[4 x i8]*> [#uses=1] -@.str1 = internal constant [2 x i8] c".\00", section "llvm.metadata" ; <[2 x i8]*> [#uses=1] -@.str2 = internal constant [6 x i8] c"clang\00", section "llvm.metadata" ; <[6 x i8]*> [#uses=1] -@llvm.dbg.compile_unit = internal constant %llvm.dbg.compile_unit.type { i32 458769, { }* bitcast (%llvm.dbg.anchor.type* @llvm.dbg.compile_units to { }*), i32 1, i8* getelementptr ([4 x i8]* @.str, i32 0, i32 0), i8* getelementptr ([2 x i8]* @.str1, i32 0, i32 0), i8* getelementptr ([6 x i8]* @.str2, i32 0, i32 0), i1 false, i1 false, i8* null }, section "llvm.metadata" ; <%llvm.dbg.compile_unit.type*> [#uses=1] -@.str3 = internal constant [4 x i8] c"int\00", section "llvm.metadata" ; <[4 x i8]*> [#uses=1] -@llvm.dbg.basictype = internal constant %llvm.dbg.basictype.type { i32 458788, { }* bitcast (%llvm.dbg.compile_unit.type* @llvm.dbg.compile_unit to { }*), i8* getelementptr ([4 x i8]* @.str3, i32 0, i32 0), { }* bitcast (%llvm.dbg.compile_unit.type* @llvm.dbg.compile_unit to { }*), i32 0, i64 32, i64 32, i64 0, i32 0, i32 5 }, section "llvm.metadata" ; <%llvm.dbg.basictype.type*> [#uses=1] -@llvm.dbg.subprograms = linkonce constant %llvm.dbg.anchor.type { i32 458752, i32 46 }, section "llvm.metadata" ; <%llvm.dbg.anchor.type*> [#uses=1] -@.str4 = internal constant [5 x i8] c"test\00", section "llvm.metadata" ; <[5 x i8]*> [#uses=1] -@llvm.dbg.subprogram = internal constant %llvm.dbg.subprogram.type { i32 458798, { }* bitcast (%llvm.dbg.anchor.type* @llvm.dbg.subprograms to { }*), { }* bitcast (%llvm.dbg.compile_unit.type* @llvm.dbg.compile_unit to { }*), i8* getelementptr ([5 x i8]* @.str4, i32 0, i32 0), i8* getelementptr ([5 x i8]* @.str4, i32 0, i32 0), i8* null, { }* bitcast (%llvm.dbg.compile_unit.type* @llvm.dbg.compile_unit to { }*), i32 3, { }* bitcast (%llvm.dbg.basictype.type* @llvm.dbg.basictype to { }*), i1 false, i1 true }, section "llvm.metadata" ; <%llvm.dbg.subprogram.type*> [#uses=1] -@.str5 = internal constant [2 x i8] c"X\00", section "llvm.metadata" ; <[2 x i8]*> [#uses=1] -@llvm.dbg.variable = internal constant %llvm.dbg.variable.type { i32 459009, { }* bitcast (%llvm.dbg.subprogram.type* @llvm.dbg.subprogram to { }*), i8* getelementptr ([2 x i8]* @.str5, i32 0, i32 0), { }* bitcast (%llvm.dbg.compile_unit.type* @llvm.dbg.compile_unit to { }*), i32 3, { }* bitcast (%llvm.dbg.basictype.type* @llvm.dbg.basictype to { }*) }, section "llvm.metadata" ; <%llvm.dbg.variable.type*> [#uses=1] -@llvm.dbg.block = internal constant %llvm.dbg.block.type { i32 458763, { }* bitcast (%llvm.dbg.subprogram.type* @llvm.dbg.subprogram to { }*) }, section "llvm.metadata" ; <%llvm.dbg.block.type*> [#uses=1] -@llvm.dbg.subrange = internal constant %llvm.dbg.subrange.type { i32 458785, i64 0, i64 0 }, section "llvm.metadata" ; <%llvm.dbg.subrange.type*> [#uses=1] -@llvm.dbg.array = internal constant [1 x { }*] [{ }* bitcast (%llvm.dbg.subrange.type* @llvm.dbg.subrange to { }*)], section "llvm.metadata" ; <[1 x { }*]*> [#uses=1] -@llvm.dbg.composite = internal constant %llvm.dbg.composite.type { i32 458753, { }* bitcast (%llvm.dbg.compile_unit.type* @llvm.dbg.compile_unit to { }*), i8* null, { }* null, i32 0, i64 0, i64 32, i64 0, i32 0, { }* bitcast (%llvm.dbg.basictype.type* @llvm.dbg.basictype to { }*), { }* bitcast ([1 x { }*]* @llvm.dbg.array to { }*) }, section "llvm.metadata" ; <%llvm.dbg.composite.type*> [#uses=1] -@.str6 = internal constant [2 x i8] c"Y\00", section "llvm.metadata" ; <[2 x i8]*> [#uses=1] -@llvm.dbg.variable7 = internal constant %llvm.dbg.variable.type { i32 459008, { }* bitcast (%llvm.dbg.block.type* @llvm.dbg.block to { }*), i8* getelementptr ([2 x i8]* @.str6, i32 0, i32 0), { }* bitcast (%llvm.dbg.compile_unit.type* @llvm.dbg.compile_unit to { }*), i32 4, { }* bitcast (%llvm.dbg.composite.type* @llvm.dbg.composite to { }*) }, section "llvm.metadata" ; <%llvm.dbg.variable.type*> [#uses=1] - -define i32 @test(i32 %X) nounwind { +define signext i8 @foo(i8* %s1) nounwind ssp { entry: - %retval = alloca i32 ; <i32*> [#uses=1] - %X.addr = alloca i32 ; <i32*> [#uses=3] - %saved_stack = alloca i8* ; <i8**> [#uses=2] - call void @llvm.dbg.func.start({ }* bitcast (%llvm.dbg.subprogram.type* @llvm.dbg.subprogram to { }*)) - store i32 %X, i32* %X.addr - %0 = bitcast i32* %X.addr to { }* ; <{ }*> [#uses=1] - call void @llvm.dbg.declare({ }* %0, { }* bitcast (%llvm.dbg.variable.type* @llvm.dbg.variable to { }*)) - call void @llvm.dbg.region.start({ }* bitcast (%llvm.dbg.block.type* @llvm.dbg.block to { }*)) - call void @llvm.dbg.stoppoint(i32 4, i32 3, { }* bitcast (%llvm.dbg.compile_unit.type* @llvm.dbg.compile_unit to { }*)) - %1 = call i8* @llvm.stacksave() ; <i8*> [#uses=1] - store i8* %1, i8** %saved_stack - %tmp = load i32* %X.addr ; <i32> [#uses=1] - %2 = mul i32 4, %tmp ; <i32> [#uses=1] - %vla = alloca i8, i32 %2 ; <i8*> [#uses=1] - %tmp1 = bitcast i8* %vla to i32* ; <i32*> [#uses=1] - %3 = bitcast i32* %tmp1 to { }* ; <{ }*> [#uses=1] - call void @llvm.dbg.declare({ }* %3, { }* bitcast (%llvm.dbg.variable.type* @llvm.dbg.variable7 to { }*)) - call void @llvm.dbg.stoppoint(i32 5, i32 1, { }* bitcast (%llvm.dbg.compile_unit.type* @llvm.dbg.compile_unit to { }*)) - call void @llvm.dbg.region.end({ }* bitcast (%llvm.dbg.block.type* @llvm.dbg.block to { }*)) - br label %cleanup + %s1_addr = alloca i8* ; <i8**> [#uses=2] + %retval = alloca i32 ; <i32*> [#uses=2] + %saved_stack.1 = alloca i8* ; <i8**> [#uses=2] + %0 = alloca i32 ; <i32*> [#uses=2] + %str.0 = alloca [0 x i8]* ; <[0 x i8]**> [#uses=3] + %1 = alloca i64 ; <i64*> [#uses=2] + %2 = alloca i64 ; <i64*> [#uses=1] + %3 = alloca i64 ; <i64*> [#uses=6] + %"alloca point" = bitcast i32 0 to i32 ; <i32> [#uses=0] + call void @llvm.dbg.declare(metadata !{i8** %s1_addr}, metadata !0), !dbg !7 + store i8* %s1, i8** %s1_addr + call void @llvm.dbg.declare(metadata !{[0 x i8]** %str.0}, metadata !8), !dbg !7 + %4 = call i8* @llvm.stacksave(), !dbg !7 ; <i8*> [#uses=1] + store i8* %4, i8** %saved_stack.1, align 8, !dbg !7 + %5 = load i8** %s1_addr, align 8, !dbg !13 ; <i8*> [#uses=1] + %6 = call i64 @strlen(i8* %5) nounwind readonly, !dbg !13 ; <i64> [#uses=1] + %7 = add i64 %6, 1, !dbg !13 ; <i64> [#uses=1] + store i64 %7, i64* %3, align 8, !dbg !13 + %8 = load i64* %3, align 8, !dbg !13 ; <i64> [#uses=1] + %9 = sub nsw i64 %8, 1, !dbg !13 ; <i64> [#uses=0] + %10 = load i64* %3, align 8, !dbg !13 ; <i64> [#uses=1] + %11 = mul i64 %10, 8, !dbg !13 ; <i64> [#uses=0] + %12 = load i64* %3, align 8, !dbg !13 ; <i64> [#uses=1] + store i64 %12, i64* %2, align 8, !dbg !13 + %13 = load i64* %3, align 8, !dbg !13 ; <i64> [#uses=1] + %14 = mul i64 %13, 8, !dbg !13 ; <i64> [#uses=0] + %15 = load i64* %3, align 8, !dbg !13 ; <i64> [#uses=1] + store i64 %15, i64* %1, align 8, !dbg !13 + %16 = load i64* %1, align 8, !dbg !13 ; <i64> [#uses=1] + %17 = trunc i64 %16 to i32, !dbg !13 ; <i32> [#uses=1] + %18 = alloca i8, i32 %17, !dbg !13 ; <i8*> [#uses=1] + %19 = bitcast i8* %18 to [0 x i8]*, !dbg !13 ; <[0 x i8]*> [#uses=1] + store [0 x i8]* %19, [0 x i8]** %str.0, align 8, !dbg !13 + %20 = load [0 x i8]** %str.0, align 8, !dbg !15 ; <[0 x i8]*> [#uses=1] + %21 = getelementptr inbounds [0 x i8]* %20, i64 0, i64 0, !dbg !15 ; <i8*> [#uses=1] + store i8 0, i8* %21, align 1, !dbg !15 + %22 = load [0 x i8]** %str.0, align 8, !dbg !16 ; <[0 x i8]*> [#uses=1] + %23 = getelementptr inbounds [0 x i8]* %22, i64 0, i64 0, !dbg !16 ; <i8*> [#uses=1] + %24 = load i8* %23, align 1, !dbg !16 ; <i8> [#uses=1] + %25 = sext i8 %24 to i32, !dbg !16 ; <i32> [#uses=1] + store i32 %25, i32* %0, align 4, !dbg !16 + %26 = load i8** %saved_stack.1, align 8, !dbg !16 ; <i8*> [#uses=1] + call void @llvm.stackrestore(i8* %26), !dbg !16 + %27 = load i32* %0, align 4, !dbg !16 ; <i32> [#uses=1] + store i32 %27, i32* %retval, align 4, !dbg !16 + br label %return, !dbg !16 -cleanup: ; preds = %entry - %tmp2 = load i8** %saved_stack ; <i8*> [#uses=1] - call void @llvm.stackrestore(i8* %tmp2) - call void @llvm.dbg.region.end({ }* bitcast (%llvm.dbg.subprogram.type* @llvm.dbg.subprogram to { }*)) - %4 = load i32* %retval ; <i32> [#uses=1] - ret i32 %4 +return: ; preds = %entry + %retval1 = load i32* %retval, !dbg !16 ; <i32> [#uses=1] + %retval12 = trunc i32 %retval1 to i8, !dbg !16 ; <i8> [#uses=1] + ret i8 %retval12, !dbg !16 } -declare void @llvm.dbg.func.start({ }*) nounwind - -declare void @llvm.dbg.declare({ }*, { }*) nounwind - -declare void @llvm.dbg.region.start({ }*) nounwind - -declare void @llvm.dbg.stoppoint(i32, i32, { }*) nounwind +declare void @llvm.dbg.declare(metadata, metadata) nounwind readnone declare i8* @llvm.stacksave() nounwind +declare i64 @strlen(i8*) nounwind readonly + declare void @llvm.stackrestore(i8*) nounwind -declare void @llvm.dbg.region.end({ }*) nounwind +!0 = metadata !{i32 459009, metadata !1, metadata !"s1", metadata !2, i32 2, metadata !6} ; [ DW_TAG_arg_variable ] +!1 = metadata !{i32 458798, i32 0, metadata !2, metadata !"foo", metadata !"foo", metadata !"foo", metadata !2, i32 2, metadata !3, i1 false, i1 true, i32 0, i32 0, null, i1 false} ; [ DW_TAG_subprogram ] +!2 = metadata !{i32 458769, i32 0, i32 1, metadata !"vla.c", metadata !"/tmp/", metadata !"4.2.1 (Based on Apple Inc. build 5658) (LLVM build)", i1 true, i1 false, metadata !"", i32 0} ; [ DW_TAG_compile_unit ] +!3 = metadata !{i32 458773, metadata !2, metadata !"", metadata !2, i32 0, i64 0, i64 0, i64 0, i32 0, null, metadata !4, i32 0, null} ; [ DW_TAG_subroutine_type ] +!4 = metadata !{metadata !5, metadata !6} +!5 = metadata !{i32 458788, metadata !2, metadata !"char", metadata !2, i32 0, i64 8, i64 8, i64 0, i32 0, i32 6} ; [ DW_TAG_base_type ] +!6 = metadata !{i32 458767, metadata !2, metadata !"", metadata !2, i32 0, i64 64, i64 64, i64 0, i32 0, metadata !5} ; [ DW_TAG_pointer_type ] +!7 = metadata !{i32 2, i32 0, metadata !1, null} +!8 = metadata !{i32 459008, metadata !1, metadata !"str.0", metadata !2, i32 3, metadata !9} ; [ DW_TAG_auto_variable ] +!9 = metadata !{i32 458767, metadata !2, metadata !"", metadata !2, i32 0, i64 64, i64 64, i64 0, i32 64, metadata !10} ; [ DW_TAG_pointer_type ] +!10 = metadata !{i32 458753, metadata !2, metadata !"", metadata !2, i32 0, i64 8, i64 8, i64 0, i32 0, metadata !5, metadata !11, i32 0, null} ; [ DW_TAG_array_type ] +!11 = metadata !{metadata !12} +!12 = metadata !{i32 458785, i64 0, i64 0} ; [ DW_TAG_subrange_type ] +!13 = metadata !{i32 3, i32 0, metadata !14, null} +!14 = metadata !{i32 458763, metadata !1, i32 0, i32 0} ; [ DW_TAG_lexical_block ] +!15 = metadata !{i32 4, i32 0, metadata !14, null} +!16 = metadata !{i32 5, i32 0, metadata !14, null} diff --git a/test/CodeGen/X86/2009-07-16-LoadFoldingBug.ll b/test/CodeGen/X86/2009-07-16-LoadFoldingBug.ll deleted file mode 100644 index e21c8923df4a..000000000000 --- a/test/CodeGen/X86/2009-07-16-LoadFoldingBug.ll +++ /dev/null @@ -1,102 +0,0 @@ -; RUN: llc < %s -mtriple=x86_64-apple-darwin10 | FileCheck %s - -; CHECK: _foo: -; CHECK: pavgw LCPI1_4(%rip) - -; rdar://7057804 - -define void @foo(i16* %out8x8, i16* %in8x8, i32 %lastrow) optsize ssp { -entry: - %0 = call <8 x i16> @llvm.x86.sse2.pmulh.w(<8 x i16> <i16 6518, i16 6518, i16 6518, i16 6518, i16 6518, i16 6518, i16 6518, i16 6518>, <8 x i16> undef) nounwind readnone ; <<8 x i16>> [#uses=2] - %1 = call <8 x i16> @llvm.x86.sse2.pcmpeq.w(<8 x i16> %0, <8 x i16> <i16 16384, i16 16384, i16 16384, i16 16384, i16 16384, i16 16384, i16 16384, i16 16384>) nounwind readnone ; <<8 x i16>> [#uses=1] - %2 = call <8 x i16> @llvm.x86.sse2.psrli.w(<8 x i16> zeroinitializer, i32 14) nounwind readnone ; <<8 x i16>> [#uses=1] - %3 = call <8 x i16> @llvm.x86.sse2.pavg.w(<8 x i16> %2, <8 x i16> zeroinitializer) nounwind readnone ; <<8 x i16>> [#uses=1] - %tmp.i.i10 = add <8 x i16> %0, %3 ; <<8 x i16>> [#uses=1] - %4 = call <8 x i16> @llvm.x86.sse2.psubs.w(<8 x i16> zeroinitializer, <8 x i16> %1) nounwind readnone ; <<8 x i16>> [#uses=1] - %5 = call <8 x i16> @llvm.x86.sse2.padds.w(<8 x i16> %tmp.i.i10, <8 x i16> %4) nounwind readnone ; <<8 x i16>> [#uses=3] - %6 = call <8 x i16> @llvm.x86.sse2.psubs.w(<8 x i16> %5, <8 x i16> undef) nounwind readnone ; <<8 x i16>> [#uses=1] - %7 = call <8 x i16> @llvm.x86.sse2.pmulh.w(<8 x i16> <i16 6518, i16 6518, i16 6518, i16 6518, i16 6518, i16 6518, i16 6518, i16 6518>, <8 x i16> undef) nounwind readnone ; <<8 x i16>> [#uses=2] - %8 = call <8 x i16> @llvm.x86.sse2.pcmpeq.w(<8 x i16> %7, <8 x i16> <i16 16384, i16 16384, i16 16384, i16 16384, i16 16384, i16 16384, i16 16384, i16 16384>) nounwind readnone ; <<8 x i16>> [#uses=1] - %9 = call <8 x i16> @llvm.x86.sse2.psrli.w(<8 x i16> zeroinitializer, i32 14) nounwind readnone ; <<8 x i16>> [#uses=1] - %10 = call <8 x i16> @llvm.x86.sse2.pavg.w(<8 x i16> %9, <8 x i16> zeroinitializer) nounwind readnone ; <<8 x i16>> [#uses=1] - %tmp.i.i8 = add <8 x i16> %7, %10 ; <<8 x i16>> [#uses=1] - %11 = call <8 x i16> @llvm.x86.sse2.psubs.w(<8 x i16> undef, <8 x i16> %8) nounwind readnone ; <<8 x i16>> [#uses=1] - %12 = call <8 x i16> @llvm.x86.sse2.padds.w(<8 x i16> %tmp.i.i8, <8 x i16> %11) nounwind readnone ; <<8 x i16>> [#uses=1] - %13 = call <8 x i16> @llvm.x86.sse2.padds.w(<8 x i16> undef, <8 x i16> undef) nounwind readnone ; <<8 x i16>> [#uses=1] - %14 = call <8 x i16> @llvm.x86.sse2.psubs.w(<8 x i16> %5, <8 x i16> undef) nounwind readnone ; <<8 x i16>> [#uses=1] - %15 = call <8 x i16> @llvm.x86.sse2.padds.w(<8 x i16> %5, <8 x i16> undef) nounwind readnone ; <<8 x i16>> [#uses=1] - %16 = call <8 x i16> @llvm.x86.sse2.padds.w(<8 x i16> %6, <8 x i16> undef) nounwind readnone ; <<8 x i16>> [#uses=1] - %17 = call <8 x i16> @llvm.x86.sse2.padds.w(<8 x i16> %12, <8 x i16> undef) nounwind readnone ; <<8 x i16>> [#uses=1] - %18 = call <8 x i16> @llvm.x86.sse2.psubs.w(<8 x i16> %13, <8 x i16> %15) nounwind readnone ; <<8 x i16>> [#uses=1] - %19 = call <8 x i16> @llvm.x86.sse2.psubs.w(<8 x i16> undef, <8 x i16> %14) nounwind readnone ; <<8 x i16>> [#uses=2] - %20 = call <8 x i16> @llvm.x86.sse2.psubs.w(<8 x i16> undef, <8 x i16> undef) nounwind readnone ; <<8 x i16>> [#uses=4] - %21 = call <8 x i16> @llvm.x86.sse2.psubs.w(<8 x i16> undef, <8 x i16> %17) nounwind readnone ; <<8 x i16>> [#uses=1] - %22 = bitcast <8 x i16> %21 to <2 x i64> ; <<2 x i64>> [#uses=1] - %23 = call <8 x i16> @llvm.x86.sse2.pmulh.w(<8 x i16> <i16 23170, i16 23170, i16 23170, i16 23170, i16 23170, i16 23170, i16 23170, i16 23170>, <8 x i16> undef) nounwind readnone ; <<8 x i16>> [#uses=2] - %24 = call <8 x i16> @llvm.x86.sse2.pcmpeq.w(<8 x i16> %23, <8 x i16> <i16 16384, i16 16384, i16 16384, i16 16384, i16 16384, i16 16384, i16 16384, i16 16384>) nounwind readnone ; <<8 x i16>> [#uses=1] - %25 = call <8 x i16> @llvm.x86.sse2.psrli.w(<8 x i16> zeroinitializer, i32 14) nounwind readnone ; <<8 x i16>> [#uses=1] - %26 = call <8 x i16> @llvm.x86.sse2.pavg.w(<8 x i16> %25, <8 x i16> zeroinitializer) nounwind readnone ; <<8 x i16>> [#uses=1] - %tmp.i.i6 = add <8 x i16> %23, %26 ; <<8 x i16>> [#uses=1] - %27 = call <8 x i16> @llvm.x86.sse2.psubs.w(<8 x i16> undef, <8 x i16> %24) nounwind readnone ; <<8 x i16>> [#uses=1] - %28 = call <8 x i16> @llvm.x86.sse2.padds.w(<8 x i16> %tmp.i.i6, <8 x i16> %27) nounwind readnone ; <<8 x i16>> [#uses=1] - %29 = call <8 x i16> @llvm.x86.sse2.pmulh.w(<8 x i16> <i16 -23170, i16 -23170, i16 -23170, i16 -23170, i16 -23170, i16 -23170, i16 -23170, i16 -23170>, <8 x i16> undef) nounwind readnone ; <<8 x i16>> [#uses=2] - %30 = call <8 x i16> @llvm.x86.sse2.pcmpeq.w(<8 x i16> %29, <8 x i16> <i16 16384, i16 16384, i16 16384, i16 16384, i16 16384, i16 16384, i16 16384, i16 16384>) nounwind readnone ; <<8 x i16>> [#uses=1] - %31 = call <8 x i16> @llvm.x86.sse2.psrli.w(<8 x i16> zeroinitializer, i32 14) nounwind readnone ; <<8 x i16>> [#uses=1] - %32 = call <8 x i16> @llvm.x86.sse2.pavg.w(<8 x i16> %31, <8 x i16> zeroinitializer) nounwind readnone ; <<8 x i16>> [#uses=1] - %tmp.i.i4 = add <8 x i16> %29, %32 ; <<8 x i16>> [#uses=1] - %33 = call <8 x i16> @llvm.x86.sse2.psubs.w(<8 x i16> undef, <8 x i16> %30) nounwind readnone ; <<8 x i16>> [#uses=1] - %34 = call <8 x i16> @llvm.x86.sse2.padds.w(<8 x i16> %tmp.i.i4, <8 x i16> %33) nounwind readnone ; <<8 x i16>> [#uses=1] - %35 = call <8 x i16> @llvm.x86.sse2.pmulh.w(<8 x i16> <i16 23170, i16 23170, i16 23170, i16 23170, i16 23170, i16 23170, i16 23170, i16 23170>, <8 x i16> %20) nounwind readnone ; <<8 x i16>> [#uses=2] - %tmp.i2.i1 = mul <8 x i16> %20, <i16 23170, i16 23170, i16 23170, i16 23170, i16 23170, i16 23170, i16 23170, i16 23170> ; <<8 x i16>> [#uses=1] - %36 = call <8 x i16> @llvm.x86.sse2.pcmpeq.w(<8 x i16> %35, <8 x i16> <i16 16384, i16 16384, i16 16384, i16 16384, i16 16384, i16 16384, i16 16384, i16 16384>) nounwind readnone ; <<8 x i16>> [#uses=1] - %37 = call <8 x i16> @llvm.x86.sse2.psrli.w(<8 x i16> %tmp.i2.i1, i32 14) nounwind readnone ; <<8 x i16>> [#uses=1] - %38 = call <8 x i16> @llvm.x86.sse2.pavg.w(<8 x i16> %37, <8 x i16> zeroinitializer) nounwind readnone ; <<8 x i16>> [#uses=1] - %tmp.i.i2 = add <8 x i16> %35, %38 ; <<8 x i16>> [#uses=1] - %39 = call <8 x i16> @llvm.x86.sse2.psubs.w(<8 x i16> %19, <8 x i16> %36) nounwind readnone ; <<8 x i16>> [#uses=1] - %40 = call <8 x i16> @llvm.x86.sse2.padds.w(<8 x i16> %tmp.i.i2, <8 x i16> %39) nounwind readnone ; <<8 x i16>> [#uses=1] - %41 = call <8 x i16> @llvm.x86.sse2.pmulh.w(<8 x i16> <i16 -23170, i16 -23170, i16 -23170, i16 -23170, i16 -23170, i16 -23170, i16 -23170, i16 -23170>, <8 x i16> %20) nounwind readnone ; <<8 x i16>> [#uses=2] - %tmp.i2.i = mul <8 x i16> %20, <i16 -23170, i16 -23170, i16 -23170, i16 -23170, i16 -23170, i16 -23170, i16 -23170, i16 -23170> ; <<8 x i16>> [#uses=1] - %42 = call <8 x i16> @llvm.x86.sse2.pcmpeq.w(<8 x i16> %41, <8 x i16> <i16 16384, i16 16384, i16 16384, i16 16384, i16 16384, i16 16384, i16 16384, i16 16384>) nounwind readnone ; <<8 x i16>> [#uses=1] - %43 = call <8 x i16> @llvm.x86.sse2.psrli.w(<8 x i16> %tmp.i2.i, i32 14) nounwind readnone ; <<8 x i16>> [#uses=1] - %44 = call <8 x i16> @llvm.x86.sse2.pavg.w(<8 x i16> %43, <8 x i16> zeroinitializer) nounwind readnone ; <<8 x i16>> [#uses=1] - %tmp.i.i = add <8 x i16> %41, %44 ; <<8 x i16>> [#uses=1] - %45 = call <8 x i16> @llvm.x86.sse2.psubs.w(<8 x i16> %19, <8 x i16> %42) nounwind readnone ; <<8 x i16>> [#uses=1] - %46 = call <8 x i16> @llvm.x86.sse2.padds.w(<8 x i16> %tmp.i.i, <8 x i16> %45) nounwind readnone ; <<8 x i16>> [#uses=1] - %47 = call <8 x i16> @llvm.x86.sse2.padds.w(<8 x i16> %18, <8 x i16> %16) nounwind readnone ; <<8 x i16>> [#uses=1] - %48 = bitcast <8 x i16> %47 to <2 x i64> ; <<2 x i64>> [#uses=1] - %49 = bitcast <8 x i16> %28 to <2 x i64> ; <<2 x i64>> [#uses=1] - %50 = getelementptr i16* %out8x8, i64 8 ; <i16*> [#uses=1] - %51 = bitcast i16* %50 to <2 x i64>* ; <<2 x i64>*> [#uses=1] - store <2 x i64> %49, <2 x i64>* %51, align 16 - %52 = bitcast <8 x i16> %40 to <2 x i64> ; <<2 x i64>> [#uses=1] - %53 = getelementptr i16* %out8x8, i64 16 ; <i16*> [#uses=1] - %54 = bitcast i16* %53 to <2 x i64>* ; <<2 x i64>*> [#uses=1] - store <2 x i64> %52, <2 x i64>* %54, align 16 - %55 = getelementptr i16* %out8x8, i64 24 ; <i16*> [#uses=1] - %56 = bitcast i16* %55 to <2 x i64>* ; <<2 x i64>*> [#uses=1] - store <2 x i64> %48, <2 x i64>* %56, align 16 - %57 = bitcast <8 x i16> %46 to <2 x i64> ; <<2 x i64>> [#uses=1] - %58 = getelementptr i16* %out8x8, i64 40 ; <i16*> [#uses=1] - %59 = bitcast i16* %58 to <2 x i64>* ; <<2 x i64>*> [#uses=1] - store <2 x i64> %57, <2 x i64>* %59, align 16 - %60 = bitcast <8 x i16> %34 to <2 x i64> ; <<2 x i64>> [#uses=1] - %61 = getelementptr i16* %out8x8, i64 48 ; <i16*> [#uses=1] - %62 = bitcast i16* %61 to <2 x i64>* ; <<2 x i64>*> [#uses=1] - store <2 x i64> %60, <2 x i64>* %62, align 16 - %63 = getelementptr i16* %out8x8, i64 56 ; <i16*> [#uses=1] - %64 = bitcast i16* %63 to <2 x i64>* ; <<2 x i64>*> [#uses=1] - store <2 x i64> %22, <2 x i64>* %64, align 16 - ret void -} - -declare <8 x i16> @llvm.x86.sse2.psubs.w(<8 x i16>, <8 x i16>) nounwind readnone - -declare <8 x i16> @llvm.x86.sse2.pmulh.w(<8 x i16>, <8 x i16>) nounwind readnone - -declare <8 x i16> @llvm.x86.sse2.pcmpeq.w(<8 x i16>, <8 x i16>) nounwind readnone - -declare <8 x i16> @llvm.x86.sse2.pavg.w(<8 x i16>, <8 x i16>) nounwind readnone - -declare <8 x i16> @llvm.x86.sse2.padds.w(<8 x i16>, <8 x i16>) nounwind readnone - -declare <8 x i16> @llvm.x86.sse2.psrli.w(<8 x i16>, i32) nounwind readnone diff --git a/test/CodeGen/X86/2009-09-07-CoalescerBug.ll b/test/CodeGen/X86/2009-09-07-CoalescerBug.ll index a5b4a79401ea..41b4bc087279 100644 --- a/test/CodeGen/X86/2009-09-07-CoalescerBug.ll +++ b/test/CodeGen/X86/2009-09-07-CoalescerBug.ll @@ -8,8 +8,7 @@ define i64 @hammer_time(i64 %modulep, i64 %physfree) nounwind ssp noredzone noimplicitfloat { ; CHECK: hammer_time: ; CHECK: movq $Xrsvd, %rax -; CHECK: movq $Xrsvd, %rsi -; CHECK: movq $Xrsvd, %rdi +; CHECK: movq $Xrsvd, %rcx entry: br i1 undef, label %if.then, label %if.end diff --git a/test/CodeGen/X86/2010-02-11-NonTemporal.ll b/test/CodeGen/X86/2010-02-11-NonTemporal.ll new file mode 100644 index 000000000000..5789a0b9847f --- /dev/null +++ b/test/CodeGen/X86/2010-02-11-NonTemporal.ll @@ -0,0 +1,22 @@ +; RUN: llc < %s -march=x86-64 | FileCheck %s +; CHECK: movnt +target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128" +target triple = "x86_64-unknown-linux-gnu" + +!0 = metadata !{ i32 1 } + +define void @sub_(i32* noalias %n) { +"file movnt.f90, line 2, bb1": + %n1 = alloca i32*, align 8 + %i = alloca i32, align 4 + %"$LCS_0" = alloca i64, align 8 + %"$LCS_S2" = alloca <2 x double>, align 16 + %r9 = load <2 x double>* %"$LCS_S2", align 8 + %r10 = load i64* %"$LCS_0", align 8 + %r11 = inttoptr i64 %r10 to <2 x double>* + store <2 x double> %r9, <2 x double>* %r11, align 16, !nontemporal !0 + br label %"file movnt.f90, line 18, bb5" + +"file movnt.f90, line 18, bb5": + ret void +} diff --git a/test/CodeGen/X86/2010-02-19-TailCallRetAddrBug.ll b/test/CodeGen/X86/2010-02-19-TailCallRetAddrBug.ll new file mode 100644 index 000000000000..eb21dc234a0d --- /dev/null +++ b/test/CodeGen/X86/2010-02-19-TailCallRetAddrBug.ll @@ -0,0 +1,55 @@ +; RUN: llc -mtriple=i386-apple-darwin -tailcallopt < %s | FileCheck %s +; Check that lowered argumens do not overwrite the return address before it is moved. +; Bug 6225 +; +; If a call is a fastcc tail call and tail call optimization is enabled, the +; caller frame is replaced by the callee frame. This can require that arguments are +; placed on the former return address stack slot. Special care needs to be taken +; taken that the return address is moved / or stored in a register before +; lowering of arguments potentially overwrites the value. +; +; Move return address (76(%esp)) to a temporary register (%ebp) +; CHECK: movl 76(%esp), %ebp +; Overwrite return addresss +; CHECK: movl %ecx, 76(%esp) +; Move return address from temporary register (%ebp) to new stack location (60(%esp)) +; CHECK: movl %ebp, 60(%esp) + +%tupl_p = type [9 x i32]* + +declare fastcc void @l297(i32 %r10, i32 %r9, i32 %r8, i32 %r7, i32 %r6, i32 %r5, i32 %r3, i32 %r2) noreturn nounwind +declare fastcc void @l298(i32 %r10, i32 %r9, i32 %r4) noreturn nounwind + +define fastcc void @l186(%tupl_p %r1) noreturn nounwind { +entry: + %ptr1 = getelementptr %tupl_p %r1, i32 0, i32 0 + %r2 = load i32* %ptr1 + %ptr3 = getelementptr %tupl_p %r1, i32 0, i32 1 + %r3 = load i32* %ptr3 + %ptr5 = getelementptr %tupl_p %r1, i32 0, i32 2 + %r4 = load i32* %ptr5 + %ptr7 = getelementptr %tupl_p %r1, i32 0, i32 3 + %r5 = load i32* %ptr7 + %ptr9 = getelementptr %tupl_p %r1, i32 0, i32 4 + %r6 = load i32* %ptr9 + %ptr11 = getelementptr %tupl_p %r1, i32 0, i32 5 + %r7 = load i32* %ptr11 + %ptr13 = getelementptr %tupl_p %r1, i32 0, i32 6 + %r8 = load i32* %ptr13 + %ptr15 = getelementptr %tupl_p %r1, i32 0, i32 7 + %r9 = load i32* %ptr15 + %ptr17 = getelementptr %tupl_p %r1, i32 0, i32 8 + %r10 = load i32* %ptr17 + %cond = icmp eq i32 %r10, 3 + br i1 %cond, label %true, label %false + +true: + tail call fastcc void @l297(i32 %r10, i32 %r9, i32 %r8, i32 %r7, i32 %r6, i32 %r5, i32 %r3, i32 %r2) noreturn nounwind + ret void + +false: + tail call fastcc void @l298(i32 %r10, i32 %r9, i32 %r4) noreturn nounwind + ret void +} + + diff --git a/test/CodeGen/X86/2010-02-23-DAGCombineBug.ll b/test/CodeGen/X86/2010-02-23-DAGCombineBug.ll new file mode 100644 index 000000000000..6a58e9e55182 --- /dev/null +++ b/test/CodeGen/X86/2010-02-23-DAGCombineBug.ll @@ -0,0 +1,18 @@ +; RUN: llc < %s -march=x86 | FileCheck %s + +define i32* @t() nounwind optsize ssp { +entry: +; CHECK: t: +; CHECK: testl %eax, %eax +; CHECK: js + %cmp = icmp slt i32 undef, 0 ; <i1> [#uses=1] + %outsearch.0 = select i1 %cmp, i1 false, i1 true ; <i1> [#uses=1] + br i1 %outsearch.0, label %if.then27, label %if.else29 + +if.then27: ; preds = %entry + ret i32* undef + +if.else29: ; preds = %entry + unreachable +} + diff --git a/test/CodeGen/X86/2010-02-23-DIV8rDefinesAX.ll b/test/CodeGen/X86/2010-02-23-DIV8rDefinesAX.ll new file mode 100644 index 000000000000..8543c80117b7 --- /dev/null +++ b/test/CodeGen/X86/2010-02-23-DIV8rDefinesAX.ll @@ -0,0 +1,20 @@ +; RUN: llc < %s +; PR6374 +; +; This test produces a DIV8r instruction and uses %AX instead of %AH and %AL. +; The DIV8r must have the right imp-defs for that to work. +target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64" +target triple = "x86_64-apple-darwin10.0.0" + +%struct._i386_state = type { %union.anon } +%union.anon = type { [0 x i8] } + +define void @i386_aam(%struct._i386_state* nocapture %cpustate) nounwind ssp { +entry: + %call = tail call fastcc signext i8 @FETCH() ; <i8> [#uses=1] + %rem = urem i8 0, %call ; <i8> [#uses=1] + store i8 %rem, i8* undef + ret void +} + +declare fastcc signext i8 @FETCH() nounwind readnone ssp diff --git a/test/CodeGen/X86/2010-02-23-RematImplicitSubreg.ll b/test/CodeGen/X86/2010-02-23-RematImplicitSubreg.ll new file mode 100644 index 000000000000..4a26ba088e5c --- /dev/null +++ b/test/CodeGen/X86/2010-02-23-RematImplicitSubreg.ll @@ -0,0 +1,49 @@ +; RUN: llc < %s +; PR6372 +; +; This test produces a move instruction with an implicitly defined super-register: +; +; %DL<def> = MOV8rr %reg1038<kill>, %RDX<imp-def> +; +; When %DL is rematerialized, we must remember to update live intervals for +; sub-registers %DX and %EDX. + +target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64" +target triple = "x86_64-apple-darwin10.0.0" + +define noalias i8* @foo() nounwind ssp { +entry: + br i1 undef, label %for.end, label %for.body + +for.body: ; preds = %if.end40, %entry + %tmp6 = load i8* undef, align 2 ; <i8> [#uses=3] + %conv11 = sext i8 %tmp6 to i64 ; <i64> [#uses=1] + %cmp15 = icmp slt i64 %conv11, undef ; <i1> [#uses=1] + br i1 %cmp15, label %if.end, label %if.then + +if.then: ; preds = %for.body + %conv18 = sext i8 %tmp6 to i32 ; <i32> [#uses=1] + %call = tail call i32 (...)* @invalid(i32 0, i32 0, i32 %conv18) nounwind ; <i32> [#uses=0] + br label %if.end + +if.end: ; preds = %if.then, %for.body + %index.0 = phi i8 [ 0, %if.then ], [ %tmp6, %for.body ] ; <i8> [#uses=1] + store i8 %index.0, i8* undef + %tmp24 = load i8* undef ; <i8> [#uses=2] + br i1 undef, label %if.end40, label %if.then36 + +if.then36: ; preds = %if.end + %conv38 = sext i8 %tmp24 to i32 ; <i32> [#uses=1] + %call39 = tail call i32 (...)* @invalid(i32 0, i32 0, i32 %conv38) nounwind ; <i32> [#uses=0] + br label %if.end40 + +if.end40: ; preds = %if.then36, %if.end + %index.1 = phi i8 [ 0, %if.then36 ], [ %tmp24, %if.end ] ; <i8> [#uses=1] + store i8 %index.1, i8* undef + br i1 false, label %for.body, label %for.end + +for.end: ; preds = %if.end40, %entry + ret i8* undef +} + +declare i32 @invalid(...) diff --git a/test/CodeGen/X86/2010-02-23-SingleDefPhiJoin.ll b/test/CodeGen/X86/2010-02-23-SingleDefPhiJoin.ll new file mode 100644 index 000000000000..aeed40146172 --- /dev/null +++ b/test/CodeGen/X86/2010-02-23-SingleDefPhiJoin.ll @@ -0,0 +1,146 @@ +; RUN: llc < %s +; PR6363 +; +; This test case creates a phi join register with a single definition. The other +; predecessor blocks are implicit-def. +; +; If LiveIntervalAnalysis fails to recognize this as a phi join, the coalescer +; will detect an infinity valno loop. +; +target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128" +target triple = "x86_64-unknown-linux-gnu" + +define i32 @decode(i8* nocapture %input, i32 %offset, i8* nocapture %output) nounwind { +entry: + br i1 undef, label %meshBB86, label %meshBB102 + +bb: ; preds = %meshBB106, %meshBB102 + br i1 false, label %bb9, label %meshBB90 + +bb.nph: ; preds = %meshBB90 + br label %meshBB114 + +bb.nph.fragment: ; preds = %meshBB114 + br label %meshBB118 + +bb1.fragment: ; preds = %meshBB118 + br i1 false, label %bb2, label %bb3 + +bb2: ; preds = %bb1.fragment + br label %meshBB74 + +bb2.fragment15: ; preds = %meshBB74 + br label %meshBB98 + +bb3: ; preds = %bb1.fragment + br i1 undef, label %meshBB, label %meshBB102 + +bb4: ; preds = %meshBB + br label %meshBB118 + +bb4.fragment: ; preds = %meshBB118 + br label %meshBB82 + +bb5: ; preds = %meshBB102, %meshBB82 + br i1 false, label %bb6, label %bb7 + +bb6: ; preds = %bb5 + br label %bb7 + +bb7: ; preds = %meshBB98, %bb6, %bb5 + br label %meshBB114 + +bb7.fragment: ; preds = %meshBB114 + br i1 undef, label %meshBB74, label %bb9 + +bb9: ; preds = %bb7.fragment, %bb + br label %bb1.i23 + +bb1.i23: ; preds = %meshBB110, %bb9 + br i1 undef, label %meshBB106, label %meshBB110 + +skip_to_newline.exit26: ; preds = %meshBB106 + br label %meshBB86 + +skip_to_newline.exit26.fragment: ; preds = %meshBB86 + br i1 false, label %meshBB90, label %meshBB106 + +bb11.fragment: ; preds = %meshBB90, %meshBB86 + br label %meshBB122 + +bb1.i: ; preds = %meshBB122, %meshBB + %ooffset.2.lcssa.phi.SV.phi203 = phi i32 [ 0, %meshBB122 ], [ %ooffset.2.lcssa.phi.SV.phi233, %meshBB ] ; <i32> [#uses=1] + br label %meshBB98 + +bb1.i.fragment: ; preds = %meshBB98 + br i1 undef, label %meshBB78, label %meshBB + +skip_to_newline.exit: ; preds = %meshBB78 + br i1 undef, label %bb12, label %meshBB110 + +bb12: ; preds = %skip_to_newline.exit + br label %meshBB94 + +bb12.fragment: ; preds = %meshBB94 + br i1 false, label %bb13, label %meshBB78 + +bb13: ; preds = %bb12.fragment + br label %meshBB82 + +bb13.fragment: ; preds = %meshBB82 + br i1 undef, label %meshBB94, label %meshBB122 + +bb14: ; preds = %meshBB94 + ret i32 %ooffset.2.lcssa.phi.SV.phi250 + +bb15: ; preds = %meshBB122, %meshBB110, %meshBB78 + unreachable + +meshBB: ; preds = %bb1.i.fragment, %bb3 + %ooffset.2.lcssa.phi.SV.phi233 = phi i32 [ undef, %bb3 ], [ %ooffset.2.lcssa.phi.SV.phi209, %bb1.i.fragment ] ; <i32> [#uses=1] + br i1 undef, label %bb1.i, label %bb4 + +meshBB74: ; preds = %bb7.fragment, %bb2 + br i1 false, label %meshBB118, label %bb2.fragment15 + +meshBB78: ; preds = %bb12.fragment, %bb1.i.fragment + %ooffset.2.lcssa.phi.SV.phi239 = phi i32 [ %ooffset.2.lcssa.phi.SV.phi209, %bb1.i.fragment ], [ %ooffset.2.lcssa.phi.SV.phi250, %bb12.fragment ] ; <i32> [#uses=1] + br i1 false, label %bb15, label %skip_to_newline.exit + +meshBB82: ; preds = %bb13, %bb4.fragment + br i1 false, label %bb5, label %bb13.fragment + +meshBB86: ; preds = %skip_to_newline.exit26, %entry + br i1 undef, label %skip_to_newline.exit26.fragment, label %bb11.fragment + +meshBB90: ; preds = %skip_to_newline.exit26.fragment, %bb + br i1 false, label %bb11.fragment, label %bb.nph + +meshBB94: ; preds = %bb13.fragment, %bb12 + %ooffset.2.lcssa.phi.SV.phi250 = phi i32 [ 0, %bb13.fragment ], [ %ooffset.2.lcssa.phi.SV.phi239, %bb12 ] ; <i32> [#uses=2] + br i1 false, label %bb12.fragment, label %bb14 + +meshBB98: ; preds = %bb1.i, %bb2.fragment15 + %ooffset.2.lcssa.phi.SV.phi209 = phi i32 [ undef, %bb2.fragment15 ], [ %ooffset.2.lcssa.phi.SV.phi203, %bb1.i ] ; <i32> [#uses=2] + br i1 undef, label %bb1.i.fragment, label %bb7 + +meshBB102: ; preds = %bb3, %entry + br i1 undef, label %bb5, label %bb + +meshBB106: ; preds = %skip_to_newline.exit26.fragment, %bb1.i23 + br i1 undef, label %bb, label %skip_to_newline.exit26 + +meshBB110: ; preds = %skip_to_newline.exit, %bb1.i23 + br i1 false, label %bb15, label %bb1.i23 + +meshBB114: ; preds = %bb7, %bb.nph + %meshStackVariable115.phi = phi i32 [ 19, %bb7 ], [ 8, %bb.nph ] ; <i32> [#uses=0] + br i1 undef, label %bb.nph.fragment, label %bb7.fragment + +meshBB118: ; preds = %meshBB74, %bb4, %bb.nph.fragment + %meshCmp121 = icmp eq i32 undef, 10 ; <i1> [#uses=1] + br i1 %meshCmp121, label %bb4.fragment, label %bb1.fragment + +meshBB122: ; preds = %bb13.fragment, %bb11.fragment + br i1 false, label %bb1.i, label %bb15 +} diff --git a/test/CodeGen/X86/and-or-fold.ll b/test/CodeGen/X86/and-or-fold.ll index 7733b8a5baaa..836b5f1551c7 100644 --- a/test/CodeGen/X86/and-or-fold.ll +++ b/test/CodeGen/X86/and-or-fold.ll @@ -1,14 +1,26 @@ -; RUN: llc < %s -march=x86 | grep and | count 1 +; RUN: llc < %s -mtriple=i686-apple-darwin | FileCheck -check-prefix=DARWIN %s +; RUN: opt < %s -O2 | llc -mtriple=x86_64-apple-darwin | FileCheck -check-prefix=DARWIN-OPT %s ; The dag combiner should fold together (x&127)|(y&16711680) -> (x|y)&c1 ; in this case. -define i32 @test6(i32 %x, i16 %y) { - %tmp1 = zext i16 %y to i32 ; <i32> [#uses=1] - %tmp2 = and i32 %tmp1, 127 ; <i32> [#uses=1] - %tmp4 = shl i32 %x, 16 ; <i32> [#uses=1] - %tmp5 = and i32 %tmp4, 16711680 ; <i32> [#uses=1] - %tmp6 = or i32 %tmp2, %tmp5 ; <i32> [#uses=1] - ret i32 %tmp6 +define i32 @test1(i32 %x, i16 %y) { + %tmp1 = zext i16 %y to i32 + %tmp2 = and i32 %tmp1, 127 + %tmp4 = shl i32 %x, 16 + %tmp5 = and i32 %tmp4, 16711680 + %tmp6 = or i32 %tmp2, %tmp5 + ret i32 %tmp6 +; DARWIN: andl $16711807, %eax } +; <rdar://problem/7529774> The optimizer shouldn't fold this into (and (or, C), D) +; if (C & D) == 0 +define i64 @test2(i64 %x) nounwind readnone ssp { +entry: + %tmp1 = and i64 %x, 123127 + %tmp2 = or i64 %tmp1, 3 + ret i64 %tmp2 +; DARWIN-OPT: andq $123124 +; DARWIN-OPT-NEXT: leaq 3 +} diff --git a/test/CodeGen/X86/code_placement_eh.ll b/test/CodeGen/X86/code_placement_eh.ll new file mode 100644 index 000000000000..172d5910d031 --- /dev/null +++ b/test/CodeGen/X86/code_placement_eh.ll @@ -0,0 +1,45 @@ +; RUN: llc < %s + +; CodePlacementOpt shouldn't try to modify this loop because +; it involves EH edges. + +target datalayout = "e-p:32:32:32-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:32:64-f32:32:32-f64:32:64-v64:64:64-v128:128:128-a0:0:64-f80:128:128-n8:16:32" +target triple = "i386-apple-darwin10.0" + +define void @foo() { +invcont5: + br label %bb15 + +.noexc3: ; preds = %bb15 + br i1 undef, label %bb18.i5.i, label %bb15 + +.noexc6.i.i: ; preds = %bb18.i5.i + %tmp2021 = invoke float @cosf(float 0.000000e+00) readonly + to label %bb18.i5.i unwind label %lpad.i.i ; <float> [#uses=0] + +bb18.i5.i: ; preds = %.noexc6.i.i, %bb51.i + %tmp2019 = invoke float @sinf(float 0.000000e+00) readonly + to label %.noexc6.i.i unwind label %lpad.i.i ; <float> [#uses=0] + +lpad.i.i: ; preds = %bb18.i5.i, %.noexc6.i.i + %eh_ptr.i.i = call i8* @llvm.eh.exception() ; <i8*> [#uses=1] + unreachable + +lpad59.i: ; preds = %bb15 + %eh_ptr60.i = call i8* @llvm.eh.exception() ; <i8*> [#uses=1] + unreachable + +bb15: ; preds = %.noexc3, %invcont5 + invoke fastcc void @_ZN28btHashedOverlappingPairCacheC2Ev() + to label %.noexc3 unwind label %lpad59.i +} + +declare i8* @llvm.eh.exception() nounwind readonly + +declare i32 @llvm.eh.selector(i8*, i8*, ...) nounwind + +declare float @sinf(float) readonly + +declare float @cosf(float) readonly + +declare fastcc void @_ZN28btHashedOverlappingPairCacheC2Ev() align 2 diff --git a/test/CodeGen/X86/critical-edge-split.ll b/test/CodeGen/X86/critical-edge-split.ll index 4fe554de75a0..f29cbf323e37 100644 --- a/test/CodeGen/X86/critical-edge-split.ll +++ b/test/CodeGen/X86/critical-edge-split.ll @@ -1,4 +1,4 @@ -; RUN: llc < %s -mtriple=i386-apple-darwin -tailcallopt=false -stats -info-output-file - | grep asm-printer | grep 31 +; RUN: llc < %s -mtriple=i386-apple-darwin -stats -info-output-file - | grep asm-printer | grep 29 %CC = type { %Register } %II = type { %"struct.XX::II::$_74" } diff --git a/test/CodeGen/X86/ins_subreg_coalesce-3.ll b/test/CodeGen/X86/ins_subreg_coalesce-3.ll index 627edc51c18d..8c1c40976605 100644 --- a/test/CodeGen/X86/ins_subreg_coalesce-3.ll +++ b/test/CodeGen/X86/ins_subreg_coalesce-3.ll @@ -1,4 +1,4 @@ -; RUN: llc < %s -march=x86-64 | grep mov | count 5 +; RUN: llc < %s -march=x86-64 | grep mov | count 3 %struct.COMPOSITE = type { i8, i16, i16 } %struct.FILE = type { i8*, i32, i32, i16, i16, %struct.__sbuf, i32, i8*, i32 (i8*)*, i32 (i8*, i8*, i32)*, i64 (i8*, i64, i32)*, i32 (i8*, i8*, i32)*, %struct.__sbuf, %struct.__sFILEX*, i32, [3 x i8], [1 x i8], %struct.__sbuf, i32, i64 } diff --git a/test/CodeGen/X86/licm-symbol.ll b/test/CodeGen/X86/licm-symbol.ll new file mode 100644 index 000000000000..d61bbfccbc96 --- /dev/null +++ b/test/CodeGen/X86/licm-symbol.ll @@ -0,0 +1,39 @@ +; RUN: llc < %s | FileCheck %s + +; MachineLICM should be able to hoist the sF reference out of the loop. + +; CHECK: pushl %esi +; CHECK: subl $8, %esp +; CHECK: movl $176, %esi +; CHECK: addl L___sF$non_lazy_ptr, %esi +; CHECK: .align 4, 0x90 + +target datalayout = "e-p:32:32:32-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:32:64-f32:32:32-f64:32:64-v64:64:64-v128:128:128-a0:0:64-f80:128:128-n8:16:32" +target triple = "i386-apple-darwin8" + +%struct.FILE = type { i8*, i32, i32, i16, i16, %struct.__sbuf, i32, i8*, i32 (i8*)*, i32 (i8*, i8*, i32)*, i64 (i8*, i64, i32)*, i32 (i8*, i8*, i32)*, %struct.__sbuf, %struct.__sFILEX*, i32, [3 x i8], [1 x i8], %struct.__sbuf, i32, i64 } +%struct.__sFILEX = type opaque +%struct.__sbuf = type { i8*, i32 } +%struct.gcov_ctr_summary = type { i32, i32, i64, i64, i64 } +%struct.gcov_summary = type { i32, [1 x %struct.gcov_ctr_summary] } + +@__sF = external global [0 x %struct.FILE] ; <[0 x %struct.FILE]*> [#uses=1] + +declare i32 @fprintf(%struct.FILE* nocapture) nounwind + +define void @gcov_exit() nounwind { +entry: + br label %bb151 + +bb151: ; preds = %bb59, %bb56, %bb14 + br i1 undef, label %bb56, label %bb59 + +bb56: ; preds = %bb151 + %t0 = call i32 (%struct.FILE*)* @fprintf(%struct.FILE* getelementptr inbounds ([0 x %struct.FILE]* @__sF, i32 0, i32 2)) nounwind + br label %bb151 + +bb59: ; preds = %bb151 + %t1 = call i32 (%struct.FILE*)* @fprintf(%struct.FILE* getelementptr inbounds ([0 x %struct.FILE]* @__sF, i32 0, i32 2)) nounwind + br label %bb151 +} + diff --git a/test/CodeGen/X86/lsr-overflow.ll b/test/CodeGen/X86/lsr-overflow.ll new file mode 100644 index 000000000000..0b0214c6d9f8 --- /dev/null +++ b/test/CodeGen/X86/lsr-overflow.ll @@ -0,0 +1,26 @@ +; RUN: llc < %s -march=x86-64 | FileCheck %s + +; The comparison uses the pre-inc value, which could lead LSR to +; try to compute -INT64_MIN. + +; CHECK: movabsq $-9223372036854775808, %rax +; CHECK: cmpq %rax, %rbx +; CHECK: sete %al + +declare i64 @bar() + +define i1 @foo() nounwind { +entry: + br label %for.cond.i + +for.cond.i: + %indvar = phi i64 [ 0, %entry ], [ %indvar.next, %for.cond.i ] + %t = call i64 @bar() + %indvar.next = add i64 %indvar, 1 + %s = icmp ne i64 %indvar.next, %t + br i1 %s, label %for.cond.i, label %__ABContainsLabel.exit + +__ABContainsLabel.exit: + %cmp = icmp eq i64 %indvar, 9223372036854775807 + ret i1 %cmp +} diff --git a/test/CodeGen/X86/lsr-reuse-trunc.ll b/test/CodeGen/X86/lsr-reuse-trunc.ll index d1d714491faa..a663a220e62d 100644 --- a/test/CodeGen/X86/lsr-reuse-trunc.ll +++ b/test/CodeGen/X86/lsr-reuse-trunc.ll @@ -1,10 +1,19 @@ -; RUN: llc < %s -march=x86-64 | FileCheck %s +; RUN: llc < %s -mtriple=x86_64-apple-darwin | FileCheck %s ; Full strength reduction wouldn't reduce register pressure, so LSR should ; stick with indexing here. -; CHECK: movaps (%rsi,%rax,4), %xmm3 -; CHECK: movaps %xmm3, (%rdi,%rax,4) +; Also checks andps and andnps shares the same constantpool. Previously llvm +; will codegen two andps, one using 0x80000000, the other 0x7fffffff. +; rdar://7323335 + +; CHECK: movaps LCPI1_0 +; CHECK: movaps LCPI1_1 +; CHECK-NOT: movaps LCPI1_2 +; CHECK: movaps (%rsi,%rax,4), %xmm2 +; CHECK: andps +; CHECK: andnps +; CHECK: movaps %xmm2, (%rdi,%rax,4) ; CHECK: addq $4, %rax ; CHECK: cmpl %eax, (%rdx) ; CHECK-NEXT: jg diff --git a/test/CodeGen/X86/lsr-reuse.ll b/test/CodeGen/X86/lsr-reuse.ll index 7f2b8cc8f832..2f6fb3fa8be1 100644 --- a/test/CodeGen/X86/lsr-reuse.ll +++ b/test/CodeGen/X86/lsr-reuse.ll @@ -1,4 +1,4 @@ -; RUN: llc < %s -march=x86-64 -O3 | FileCheck %s +; RUN: llc < %s -march=x86-64 -O3 -asm-verbose=false | FileCheck %s target datalayout = "e-p:64:64:64" target triple = "x86_64-unknown-unknown" @@ -384,3 +384,59 @@ loop: return: ret void } + +; LSR should use only one indvar for the inner loop. +; rdar://7657764 + +; CHECK: asd: +; CHECK: BB10_5: +; CHECK-NEXT: addl (%r{{[^,]*}},%rdi,4), %e +; CHECK-NEXT: incq %rdi +; CHECK-NEXT: cmpq %rdi, %r{{[^,]*}} +; CHECK-NEXT: jg + +%struct.anon = type { i32, [4200 x i32] } + +@bars = common global [123123 x %struct.anon] zeroinitializer, align 32 ; <[123123 x %struct.anon]*> [#uses=2] + +define i32 @asd(i32 %n) nounwind readonly { +entry: + %0 = icmp sgt i32 %n, 0 ; <i1> [#uses=1] + br i1 %0, label %bb.nph14, label %bb5 + +bb.nph14: ; preds = %entry + %tmp18 = zext i32 %n to i64 ; <i64> [#uses=1] + br label %bb + +bb: ; preds = %bb3, %bb.nph14 + %indvar16 = phi i64 [ 0, %bb.nph14 ], [ %indvar.next17, %bb3 ] ; <i64> [#uses=3] + %s.113 = phi i32 [ 0, %bb.nph14 ], [ %s.0.lcssa, %bb3 ] ; <i32> [#uses=2] + %scevgep2526 = getelementptr [123123 x %struct.anon]* @bars, i64 0, i64 %indvar16, i32 0 ; <i32*> [#uses=1] + %1 = load i32* %scevgep2526, align 4 ; <i32> [#uses=2] + %2 = icmp sgt i32 %1, 0 ; <i1> [#uses=1] + br i1 %2, label %bb.nph, label %bb3 + +bb.nph: ; preds = %bb + %tmp23 = sext i32 %1 to i64 ; <i64> [#uses=1] + br label %bb1 + +bb1: ; preds = %bb.nph, %bb1 + %indvar = phi i64 [ 0, %bb.nph ], [ %tmp19, %bb1 ] ; <i64> [#uses=2] + %s.07 = phi i32 [ %s.113, %bb.nph ], [ %4, %bb1 ] ; <i32> [#uses=1] + %c.08 = getelementptr [123123 x %struct.anon]* @bars, i64 0, i64 %indvar16, i32 1, i64 %indvar ; <i32*> [#uses=1] + %3 = load i32* %c.08, align 4 ; <i32> [#uses=1] + %4 = add nsw i32 %3, %s.07 ; <i32> [#uses=2] + %tmp19 = add i64 %indvar, 1 ; <i64> [#uses=2] + %5 = icmp sgt i64 %tmp23, %tmp19 ; <i1> [#uses=1] + br i1 %5, label %bb1, label %bb3 + +bb3: ; preds = %bb1, %bb + %s.0.lcssa = phi i32 [ %s.113, %bb ], [ %4, %bb1 ] ; <i32> [#uses=2] + %indvar.next17 = add i64 %indvar16, 1 ; <i64> [#uses=2] + %exitcond = icmp eq i64 %indvar.next17, %tmp18 ; <i1> [#uses=1] + br i1 %exitcond, label %bb5, label %bb + +bb5: ; preds = %bb3, %entry + %s.1.lcssa = phi i32 [ 0, %entry ], [ %s.0.lcssa, %bb3 ] ; <i32> [#uses=1] + ret i32 %s.1.lcssa +} diff --git a/test/CodeGen/X86/lsr-wrap.ll b/test/CodeGen/X86/lsr-wrap.ll new file mode 100644 index 000000000000..ec8db501ef34 --- /dev/null +++ b/test/CodeGen/X86/lsr-wrap.ll @@ -0,0 +1,37 @@ +; RUN: llc -march=x86-64 < %s | FileCheck %s + +; LSR would like to use a single IV for both of these, however it's +; not safe due to wraparound. + +; CHECK: addb $-4, %r +; CHECK: decw % + +@g_19 = common global i32 0 ; <i32*> [#uses=2] + +declare i32 @func_8(i8 zeroext) nounwind + +declare i32 @func_3(i8 signext) nounwind + +define void @func_1() nounwind { +entry: + br label %bb + +bb: ; preds = %bb, %entry + %indvar = phi i16 [ 0, %entry ], [ %indvar.next, %bb ] ; <i16> [#uses=2] + %tmp = sub i16 0, %indvar ; <i16> [#uses=1] + %tmp27 = trunc i16 %tmp to i8 ; <i8> [#uses=1] + %tmp1 = load i32* @g_19, align 4 ; <i32> [#uses=2] + %tmp2 = add i32 %tmp1, 1 ; <i32> [#uses=1] + store i32 %tmp2, i32* @g_19, align 4 + %tmp3 = trunc i32 %tmp1 to i8 ; <i8> [#uses=1] + %tmp4 = tail call i32 @func_8(i8 zeroext %tmp3) nounwind ; <i32> [#uses=0] + %tmp5 = shl i8 %tmp27, 2 ; <i8> [#uses=1] + %tmp6 = add i8 %tmp5, -112 ; <i8> [#uses=1] + %tmp7 = tail call i32 @func_3(i8 signext %tmp6) nounwind ; <i32> [#uses=0] + %indvar.next = add i16 %indvar, 1 ; <i16> [#uses=2] + %exitcond = icmp eq i16 %indvar.next, -28 ; <i1> [#uses=1] + br i1 %exitcond, label %return, label %bb + +return: ; preds = %bb + ret void +} diff --git a/test/CodeGen/X86/omit-label.ll b/test/CodeGen/X86/omit-label.ll deleted file mode 100644 index 0ec03ebace89..000000000000 --- a/test/CodeGen/X86/omit-label.ll +++ /dev/null @@ -1,57 +0,0 @@ -; RUN: llc < %s -asm-verbose=false -mtriple=x86_64-linux-gnu | FileCheck %s -; PR4126 -; PR4732 - -; Don't omit these labels' definitions. - -; CHECK: bux: -; CHECK: LBB1_1: - -define void @bux(i32 %p_53) nounwind optsize { -entry: - %0 = icmp eq i32 %p_53, 0 ; <i1> [#uses=1] - %1 = icmp sgt i32 %p_53, 0 ; <i1> [#uses=1] - %or.cond = and i1 %0, %1 ; <i1> [#uses=1] - br i1 %or.cond, label %bb.i, label %bb3 - -bb.i: ; preds = %entry - %2 = add i32 %p_53, 1 ; <i32> [#uses=1] - %3 = icmp slt i32 %2, 0 ; <i1> [#uses=0] - br label %bb3 - -bb3: ; preds = %bb.i, %entry - %4 = tail call i32 (...)* @baz(i32 0) nounwind ; <i32> [#uses=0] - ret void -} - -declare i32 @baz(...) - -; Don't omit this label in the assembly output. -; CHECK: int321: -; CHECK: LBB2_1 -; CHECK: LBB2_1 -; CHECK: LBB2_1: - -define void @int321(i8 signext %p_103, i32 %uint8p_104) nounwind readnone { -entry: - %tobool = icmp eq i8 %p_103, 0 ; <i1> [#uses=1] - %cmp.i = icmp sgt i8 %p_103, 0 ; <i1> [#uses=1] - %or.cond = and i1 %tobool, %cmp.i ; <i1> [#uses=1] - br i1 %or.cond, label %land.end.i, label %for.cond.preheader - -land.end.i: ; preds = %entry - %conv3.i = sext i8 %p_103 to i32 ; <i32> [#uses=1] - %div.i = sdiv i32 1, %conv3.i ; <i32> [#uses=1] - %tobool.i = icmp eq i32 %div.i, -2147483647 ; <i1> [#uses=0] - br label %for.cond.preheader - -for.cond.preheader: ; preds = %land.end.i, %entry - %cmp = icmp sgt i8 %p_103, 1 ; <i1> [#uses=1] - br i1 %cmp, label %for.end.split, label %for.cond - -for.cond: ; preds = %for.cond.preheader, %for.cond - br label %for.cond - -for.end.split: ; preds = %for.cond.preheader - ret void -} diff --git a/test/CodeGen/X86/pr3495-2.ll b/test/CodeGen/X86/pr3495-2.ll index 71aa5a04881e..98c064a07db9 100644 --- a/test/CodeGen/X86/pr3495-2.ll +++ b/test/CodeGen/X86/pr3495-2.ll @@ -1,4 +1,8 @@ -; RUN: llc < %s -march=x86 -relocation-model=pic -disable-fp-elim -stats |& grep {Number of reloads omited} +; RUN: llc < %s -march=x86 -relocation-model=pic -disable-fp-elim -stats |& grep {Number of loads added} | grep 1 +; PR3495 +; +; This test may not be testing what it was supposed to test. +; It used to have two spills and four reloads, but not it only has one spill and one reload. target datalayout = "e-p:32:32:32" target triple = "i386-apple-darwin9.6" diff --git a/test/CodeGen/X86/ptrtoint-constexpr.ll b/test/CodeGen/X86/ptrtoint-constexpr.ll index dd9790568abe..d1cb34bec8a0 100644 --- a/test/CodeGen/X86/ptrtoint-constexpr.ll +++ b/test/CodeGen/X86/ptrtoint-constexpr.ll @@ -9,6 +9,6 @@ ; CHECK: .globl x ; CHECK: x: -; CHECK: .quad 3 +; CHECK: .quad ((0+1)&4294967295)*3 @x = global i64 mul (i64 3, i64 ptrtoint (i2* getelementptr (i2* null, i64 1) to i64)) diff --git a/test/CodeGen/X86/sink-hoist.ll b/test/CodeGen/X86/sink-hoist.ll index 01d73736d6c2..e1d0fe76657d 100644 --- a/test/CodeGen/X86/sink-hoist.ll +++ b/test/CodeGen/X86/sink-hoist.ll @@ -63,7 +63,6 @@ entry: ; CHECK: vv: ; CHECK: LCPI4_0(%rip), %xmm0 ; CHECK: LCPI4_1(%rip), %xmm1 -; CHECK: LCPI4_2(%rip), %xmm2 ; CHECK: align ; CHECK-NOT: LCPI ; CHECK: ret diff --git a/test/CodeGen/X86/sse-minmax.ll b/test/CodeGen/X86/sse-minmax.ll index 17ffb5e464aa..19fbed015b2f 100644 --- a/test/CodeGen/X86/sse-minmax.ll +++ b/test/CodeGen/X86/sse-minmax.ll @@ -1,17 +1,26 @@ ; RUN: llc < %s -march=x86-64 -asm-verbose=false | FileCheck %s +; RUN: llc < %s -march=x86-64 -asm-verbose=false -enable-unsafe-fp-math | FileCheck -check-prefix=UNSAFE %s +; RUN: llc < %s -march=x86-64 -asm-verbose=false -enable-finite-only-fp-math | FileCheck -check-prefix=FINITE %s ; Some of these patterns can be matched as SSE min or max. Some of ; then can be matched provided that the operands are swapped. ; Some of them can't be matched at all and require a comparison ; and a conditional branch. -; The naming convention is {,x_}{o,u}{gt,lt,ge,le}{,_inverse} +; The naming convention is {,x_,y_}{o,u}{gt,lt,ge,le}{,_inverse} ; x_ : use 0.0 instead of %y +; y_ : use -0.0 instead of %y ; _inverse : swap the arms of the select. ; CHECK: ogt: ; CHECK-NEXT: maxsd %xmm1, %xmm0 ; CHECK-NEXT: ret +; UNSAFE: ogt: +; UNSAFE-NEXT: maxsd %xmm1, %xmm0 +; UNSAFE-NEXT: ret +; FINITE: ogt: +; FINITE-NEXT: maxsd %xmm1, %xmm0 +; FINITE-NEXT: ret define double @ogt(double %x, double %y) nounwind { %c = fcmp ogt double %x, %y %d = select i1 %c, double %x, double %y @@ -21,6 +30,12 @@ define double @ogt(double %x, double %y) nounwind { ; CHECK: olt: ; CHECK-NEXT: minsd %xmm1, %xmm0 ; CHECK-NEXT: ret +; UNSAFE: olt: +; UNSAFE-NEXT: minsd %xmm1, %xmm0 +; UNSAFE-NEXT: ret +; FINITE: olt: +; FINITE-NEXT: minsd %xmm1, %xmm0 +; FINITE-NEXT: ret define double @olt(double %x, double %y) nounwind { %c = fcmp olt double %x, %y %d = select i1 %c, double %x, double %y @@ -31,6 +46,14 @@ define double @olt(double %x, double %y) nounwind { ; CHECK-NEXT: minsd %xmm0, %xmm1 ; CHECK-NEXT: movapd %xmm1, %xmm0 ; CHECK-NEXT: ret +; UNSAFE: ogt_inverse: +; UNSAFE-NEXT: minsd %xmm0, %xmm1 +; UNSAFE-NEXT: movapd %xmm1, %xmm0 +; UNSAFE-NEXT: ret +; FINITE: ogt_inverse: +; FINITE-NEXT: minsd %xmm0, %xmm1 +; FINITE-NEXT: movapd %xmm1, %xmm0 +; FINITE-NEXT: ret define double @ogt_inverse(double %x, double %y) nounwind { %c = fcmp ogt double %x, %y %d = select i1 %c, double %y, double %x @@ -41,6 +64,14 @@ define double @ogt_inverse(double %x, double %y) nounwind { ; CHECK-NEXT: maxsd %xmm0, %xmm1 ; CHECK-NEXT: movapd %xmm1, %xmm0 ; CHECK-NEXT: ret +; UNSAFE: olt_inverse: +; UNSAFE-NEXT: maxsd %xmm0, %xmm1 +; UNSAFE-NEXT: movapd %xmm1, %xmm0 +; UNSAFE-NEXT: ret +; FINITE: olt_inverse: +; FINITE-NEXT: maxsd %xmm0, %xmm1 +; FINITE-NEXT: movapd %xmm1, %xmm0 +; FINITE-NEXT: ret define double @olt_inverse(double %x, double %y) nounwind { %c = fcmp olt double %x, %y %d = select i1 %c, double %y, double %x @@ -49,6 +80,12 @@ define double @olt_inverse(double %x, double %y) nounwind { ; CHECK: oge: ; CHECK-NEXT: ucomisd %xmm1, %xmm0 +; UNSAFE: oge: +; UNSAFE-NEXT: maxsd %xmm1, %xmm0 +; UNSAFE-NEXT: ret +; FINITE: oge: +; FINITE-NEXT: maxsd %xmm1, %xmm0 +; FINITE-NEXT: ret define double @oge(double %x, double %y) nounwind { %c = fcmp oge double %x, %y %d = select i1 %c, double %x, double %y @@ -57,6 +94,10 @@ define double @oge(double %x, double %y) nounwind { ; CHECK: ole: ; CHECK-NEXT: ucomisd %xmm0, %xmm1 +; UNSAFE: ole: +; UNSAFE-NEXT: minsd %xmm1, %xmm0 +; FINITE: ole: +; FINITE-NEXT: minsd %xmm1, %xmm0 define double @ole(double %x, double %y) nounwind { %c = fcmp ole double %x, %y %d = select i1 %c, double %x, double %y @@ -65,6 +106,14 @@ define double @ole(double %x, double %y) nounwind { ; CHECK: oge_inverse: ; CHECK-NEXT: ucomisd %xmm1, %xmm0 +; UNSAFE: oge_inverse: +; UNSAFE-NEXT: minsd %xmm0, %xmm1 +; UNSAFE-NEXT: movapd %xmm1, %xmm0 +; UNSAFE-NEXT: ret +; FINITE: oge_inverse: +; FINITE-NEXT: minsd %xmm0, %xmm1 +; FINITE-NEXT: movapd %xmm1, %xmm0 +; FINITE-NEXT: ret define double @oge_inverse(double %x, double %y) nounwind { %c = fcmp oge double %x, %y %d = select i1 %c, double %y, double %x @@ -73,6 +122,14 @@ define double @oge_inverse(double %x, double %y) nounwind { ; CHECK: ole_inverse: ; CHECK-NEXT: ucomisd %xmm0, %xmm1 +; UNSAFE: ole_inverse: +; UNSAFE-NEXT: maxsd %xmm0, %xmm1 +; UNSAFE-NEXT: movapd %xmm1, %xmm0 +; UNSAFE-NEXT: ret +; FINITE: ole_inverse: +; FINITE-NEXT: maxsd %xmm0, %xmm1 +; FINITE-NEXT: movapd %xmm1, %xmm0 +; FINITE-NEXT: ret define double @ole_inverse(double %x, double %y) nounwind { %c = fcmp ole double %x, %y %d = select i1 %c, double %y, double %x @@ -83,6 +140,14 @@ define double @ole_inverse(double %x, double %y) nounwind { ; CHECK-NEXT: pxor %xmm1, %xmm1 ; CHECK-NEXT: maxsd %xmm1, %xmm0 ; CHECK-NEXT: ret +; UNSAFE: x_ogt: +; UNSAFE-NEXT: pxor %xmm1, %xmm1 +; UNSAFE-NEXT: maxsd %xmm1, %xmm0 +; UNSAFE-NEXT: ret +; FINITE: x_ogt: +; FINITE-NEXT: pxor %xmm1, %xmm1 +; FINITE-NEXT: maxsd %xmm1, %xmm0 +; FINITE-NEXT: ret define double @x_ogt(double %x) nounwind { %c = fcmp ogt double %x, 0.000000e+00 %d = select i1 %c, double %x, double 0.000000e+00 @@ -93,6 +158,14 @@ define double @x_ogt(double %x) nounwind { ; CHECK-NEXT: pxor %xmm1, %xmm1 ; CHECK-NEXT: minsd %xmm1, %xmm0 ; CHECK-NEXT: ret +; UNSAFE: x_olt: +; UNSAFE-NEXT: pxor %xmm1, %xmm1 +; UNSAFE-NEXT: minsd %xmm1, %xmm0 +; UNSAFE-NEXT: ret +; FINITE: x_olt: +; FINITE-NEXT: pxor %xmm1, %xmm1 +; FINITE-NEXT: minsd %xmm1, %xmm0 +; FINITE-NEXT: ret define double @x_olt(double %x) nounwind { %c = fcmp olt double %x, 0.000000e+00 %d = select i1 %c, double %x, double 0.000000e+00 @@ -104,6 +177,16 @@ define double @x_olt(double %x) nounwind { ; CHECK-NEXT: minsd %xmm0, %xmm1 ; CHECK-NEXT: movapd %xmm1, %xmm0 ; CHECK-NEXT: ret +; UNSAFE: x_ogt_inverse: +; UNSAFE-NEXT: pxor %xmm1, %xmm1 +; UNSAFE-NEXT: minsd %xmm0, %xmm1 +; UNSAFE-NEXT: movapd %xmm1, %xmm0 +; UNSAFE-NEXT: ret +; FINITE: x_ogt_inverse: +; FINITE-NEXT: pxor %xmm1, %xmm1 +; FINITE-NEXT: minsd %xmm0, %xmm1 +; FINITE-NEXT: movapd %xmm1, %xmm0 +; FINITE-NEXT: ret define double @x_ogt_inverse(double %x) nounwind { %c = fcmp ogt double %x, 0.000000e+00 %d = select i1 %c, double 0.000000e+00, double %x @@ -115,6 +198,16 @@ define double @x_ogt_inverse(double %x) nounwind { ; CHECK-NEXT: maxsd %xmm0, %xmm1 ; CHECK-NEXT: movapd %xmm1, %xmm0 ; CHECK-NEXT: ret +; UNSAFE: x_olt_inverse: +; UNSAFE-NEXT: pxor %xmm1, %xmm1 +; UNSAFE-NEXT: maxsd %xmm0, %xmm1 +; UNSAFE-NEXT: movapd %xmm1, %xmm0 +; UNSAFE-NEXT: ret +; FINITE: x_olt_inverse: +; FINITE-NEXT: pxor %xmm1, %xmm1 +; FINITE-NEXT: maxsd %xmm0, %xmm1 +; FINITE-NEXT: movapd %xmm1, %xmm0 +; FINITE-NEXT: ret define double @x_olt_inverse(double %x) nounwind { %c = fcmp olt double %x, 0.000000e+00 %d = select i1 %c, double 0.000000e+00, double %x @@ -122,9 +215,15 @@ define double @x_olt_inverse(double %x) nounwind { } ; CHECK: x_oge: -; CHECK-NEXT: pxor %xmm1, %xmm1 -; CHECK-NEXT: maxsd %xmm1, %xmm0 -; CHECK-NEXT: ret +; CHECK: ucomisd %xmm1, %xmm0 +; UNSAFE: x_oge: +; UNSAFE-NEXT: pxor %xmm1, %xmm1 +; UNSAFE-NEXT: maxsd %xmm1, %xmm0 +; UNSAFE-NEXT: ret +; FINITE: x_oge: +; FINITE-NEXT: pxor %xmm1, %xmm1 +; FINITE-NEXT: maxsd %xmm1, %xmm0 +; FINITE-NEXT: ret define double @x_oge(double %x) nounwind { %c = fcmp oge double %x, 0.000000e+00 %d = select i1 %c, double %x, double 0.000000e+00 @@ -132,9 +231,15 @@ define double @x_oge(double %x) nounwind { } ; CHECK: x_ole: -; CHECK-NEXT: pxor %xmm1, %xmm1 -; CHECK-NEXT: minsd %xmm1, %xmm0 -; CHECK-NEXT: ret +; CHECK: ucomisd %xmm0, %xmm1 +; UNSAFE: x_ole: +; UNSAFE-NEXT: pxor %xmm1, %xmm1 +; UNSAFE-NEXT: minsd %xmm1, %xmm0 +; UNSAFE-NEXT: ret +; FINITE: x_ole: +; FINITE-NEXT: pxor %xmm1, %xmm1 +; FINITE-NEXT: minsd %xmm1, %xmm0 +; FINITE-NEXT: ret define double @x_ole(double %x) nounwind { %c = fcmp ole double %x, 0.000000e+00 %d = select i1 %c, double %x, double 0.000000e+00 @@ -142,10 +247,17 @@ define double @x_ole(double %x) nounwind { } ; CHECK: x_oge_inverse: -; CHECK-NEXT: pxor %xmm1, %xmm1 -; CHECK-NEXT: minsd %xmm0, %xmm1 -; CHECK-NEXT: movapd %xmm1, %xmm0 -; CHECK-NEXT: ret +; CHECK: ucomisd %xmm1, %xmm0 +; UNSAFE: x_oge_inverse: +; UNSAFE-NEXT: pxor %xmm1, %xmm1 +; UNSAFE-NEXT: minsd %xmm0, %xmm1 +; UNSAFE-NEXT: movapd %xmm1, %xmm0 +; UNSAFE-NEXT: ret +; FINITE: x_oge_inverse: +; FINITE-NEXT: pxor %xmm1, %xmm1 +; FINITE-NEXT: minsd %xmm0, %xmm1 +; FINITE-NEXT: movapd %xmm1, %xmm0 +; FINITE-NEXT: ret define double @x_oge_inverse(double %x) nounwind { %c = fcmp oge double %x, 0.000000e+00 %d = select i1 %c, double 0.000000e+00, double %x @@ -153,10 +265,17 @@ define double @x_oge_inverse(double %x) nounwind { } ; CHECK: x_ole_inverse: -; CHECK-NEXT: pxor %xmm1, %xmm1 -; CHECK-NEXT: maxsd %xmm0, %xmm1 -; CHECK-NEXT: movapd %xmm1, %xmm0 -; CHECK-NEXT: ret +; CHECK: ucomisd %xmm0, %xmm1 +; UNSAFE: x_ole_inverse: +; UNSAFE-NEXT: pxor %xmm1, %xmm1 +; UNSAFE-NEXT: maxsd %xmm0, %xmm1 +; UNSAFE-NEXT: movapd %xmm1, %xmm0 +; UNSAFE-NEXT: ret +; FINITE: x_ole_inverse: +; FINITE-NEXT: pxor %xmm1, %xmm1 +; FINITE-NEXT: maxsd %xmm0, %xmm1 +; FINITE-NEXT: movapd %xmm1, %xmm0 +; FINITE-NEXT: ret define double @x_ole_inverse(double %x) nounwind { %c = fcmp ole double %x, 0.000000e+00 %d = select i1 %c, double 0.000000e+00, double %x @@ -164,7 +283,13 @@ define double @x_ole_inverse(double %x) nounwind { } ; CHECK: ugt: -; CHECK-NEXT: ucomisd %xmm0, %xmm1 +; CHECK: ucomisd %xmm0, %xmm1 +; UNSAFE: ugt: +; UNSAFE-NEXT: maxsd %xmm1, %xmm0 +; UNSAFE-NEXT: ret +; FINITE: ugt: +; FINITE-NEXT: maxsd %xmm1, %xmm0 +; FINITE-NEXT: ret define double @ugt(double %x, double %y) nounwind { %c = fcmp ugt double %x, %y %d = select i1 %c, double %x, double %y @@ -172,7 +297,13 @@ define double @ugt(double %x, double %y) nounwind { } ; CHECK: ult: -; CHECK-NEXT: ucomisd %xmm1, %xmm0 +; CHECK: ucomisd %xmm1, %xmm0 +; UNSAFE: ult: +; UNSAFE-NEXT: minsd %xmm1, %xmm0 +; UNSAFE-NEXT: ret +; FINITE: ult: +; FINITE-NEXT: minsd %xmm1, %xmm0 +; FINITE-NEXT: ret define double @ult(double %x, double %y) nounwind { %c = fcmp ult double %x, %y %d = select i1 %c, double %x, double %y @@ -180,7 +311,15 @@ define double @ult(double %x, double %y) nounwind { } ; CHECK: ugt_inverse: -; CHECK-NEXT: ucomisd %xmm0, %xmm1 +; CHECK: ucomisd %xmm0, %xmm1 +; UNSAFE: ugt_inverse: +; UNSAFE-NEXT: minsd %xmm0, %xmm1 +; UNSAFE-NEXT: movapd %xmm1, %xmm0 +; UNSAFE-NEXT: ret +; FINITE: ugt_inverse: +; FINITE-NEXT: minsd %xmm0, %xmm1 +; FINITE-NEXT: movapd %xmm1, %xmm0 +; FINITE-NEXT: ret define double @ugt_inverse(double %x, double %y) nounwind { %c = fcmp ugt double %x, %y %d = select i1 %c, double %y, double %x @@ -188,7 +327,15 @@ define double @ugt_inverse(double %x, double %y) nounwind { } ; CHECK: ult_inverse: -; CHECK-NEXT: ucomisd %xmm1, %xmm0 +; CHECK: ucomisd %xmm1, %xmm0 +; UNSAFE: ult_inverse: +; UNSAFE-NEXT: maxsd %xmm0, %xmm1 +; UNSAFE-NEXT: movapd %xmm1, %xmm0 +; UNSAFE-NEXT: ret +; FINITE: ult_inverse: +; FINITE-NEXT: maxsd %xmm0, %xmm1 +; FINITE-NEXT: movapd %xmm1, %xmm0 +; FINITE-NEXT: ret define double @ult_inverse(double %x, double %y) nounwind { %c = fcmp ult double %x, %y %d = select i1 %c, double %y, double %x @@ -196,9 +343,15 @@ define double @ult_inverse(double %x, double %y) nounwind { } ; CHECK: uge: -; CHECK-NEXT: maxsd %xmm0, %xmm1 -; CHECK-NEXT: movapd %xmm1, %xmm0 +; CHECK-NEXT: maxsd %xmm0, %xmm1 +; CHECK-NEXT: movapd %xmm1, %xmm0 ; CHECK-NEXT: ret +; UNSAFE: uge: +; UNSAFE-NEXT: maxsd %xmm1, %xmm0 +; UNSAFE-NEXT: ret +; FINITE: uge: +; FINITE-NEXT: maxsd %xmm1, %xmm0 +; FINITE-NEXT: ret define double @uge(double %x, double %y) nounwind { %c = fcmp uge double %x, %y %d = select i1 %c, double %x, double %y @@ -209,6 +362,12 @@ define double @uge(double %x, double %y) nounwind { ; CHECK-NEXT: minsd %xmm0, %xmm1 ; CHECK-NEXT: movapd %xmm1, %xmm0 ; CHECK-NEXT: ret +; UNSAFE: ule: +; UNSAFE-NEXT: minsd %xmm1, %xmm0 +; UNSAFE-NEXT: ret +; FINITE: ule: +; FINITE-NEXT: minsd %xmm1, %xmm0 +; FINITE-NEXT: ret define double @ule(double %x, double %y) nounwind { %c = fcmp ule double %x, %y %d = select i1 %c, double %x, double %y @@ -218,6 +377,14 @@ define double @ule(double %x, double %y) nounwind { ; CHECK: uge_inverse: ; CHECK-NEXT: minsd %xmm1, %xmm0 ; CHECK-NEXT: ret +; UNSAFE: uge_inverse: +; UNSAFE-NEXT: minsd %xmm0, %xmm1 +; UNSAFE-NEXT: movapd %xmm1, %xmm0 +; UNSAFE-NEXT: ret +; FINITE: uge_inverse: +; FINITE-NEXT: minsd %xmm0, %xmm1 +; FINITE-NEXT: movapd %xmm1, %xmm0 +; FINITE-NEXT: ret define double @uge_inverse(double %x, double %y) nounwind { %c = fcmp uge double %x, %y %d = select i1 %c, double %y, double %x @@ -227,6 +394,14 @@ define double @uge_inverse(double %x, double %y) nounwind { ; CHECK: ule_inverse: ; CHECK-NEXT: maxsd %xmm1, %xmm0 ; CHECK-NEXT: ret +; UNSAFE: ule_inverse: +; UNSAFE-NEXT: maxsd %xmm0, %xmm1 +; UNSAFE-NEXT: movapd %xmm1, %xmm0 +; UNSAFE-NEXT: ret +; FINITE: ule_inverse: +; FINITE-NEXT: maxsd %xmm0, %xmm1 +; FINITE-NEXT: movapd %xmm1, %xmm0 +; FINITE-NEXT: ret define double @ule_inverse(double %x, double %y) nounwind { %c = fcmp ule double %x, %y %d = select i1 %c, double %y, double %x @@ -234,10 +409,15 @@ define double @ule_inverse(double %x, double %y) nounwind { } ; CHECK: x_ugt: -; CHECK-NEXT: pxor %xmm1, %xmm1 -; CHECK-NEXT: maxsd %xmm0, %xmm1 -; CHECK-NEXT: movapd %xmm1, %xmm0 -; CHECK-NEXT: ret +; CHECK: ucomisd %xmm0, %xmm1 +; UNSAFE: x_ugt: +; UNSAFE-NEXT: pxor %xmm1, %xmm1 +; UNSAFE-NEXT: maxsd %xmm1, %xmm0 +; UNSAFE-NEXT: ret +; FINITE: x_ugt: +; FINITE-NEXT: pxor %xmm1, %xmm1 +; FINITE-NEXT: maxsd %xmm1, %xmm0 +; FINITE-NEXT: ret define double @x_ugt(double %x) nounwind { %c = fcmp ugt double %x, 0.000000e+00 %d = select i1 %c, double %x, double 0.000000e+00 @@ -245,10 +425,15 @@ define double @x_ugt(double %x) nounwind { } ; CHECK: x_ult: -; CHECK-NEXT: pxor %xmm1, %xmm1 -; CHECK-NEXT: minsd %xmm0, %xmm1 -; CHECK-NEXT: movapd %xmm1, %xmm0 -; CHECK-NEXT: ret +; CHECK: ucomisd %xmm1, %xmm0 +; UNSAFE: x_ult: +; UNSAFE-NEXT: pxor %xmm1, %xmm1 +; UNSAFE-NEXT: minsd %xmm1, %xmm0 +; UNSAFE-NEXT: ret +; FINITE: x_ult: +; FINITE-NEXT: pxor %xmm1, %xmm1 +; FINITE-NEXT: minsd %xmm1, %xmm0 +; FINITE-NEXT: ret define double @x_ult(double %x) nounwind { %c = fcmp ult double %x, 0.000000e+00 %d = select i1 %c, double %x, double 0.000000e+00 @@ -256,9 +441,17 @@ define double @x_ult(double %x) nounwind { } ; CHECK: x_ugt_inverse: -; CHECK-NEXT: pxor %xmm1, %xmm1 -; CHECK-NEXT: minsd %xmm1, %xmm0 -; CHECK-NEXT: ret +; CHECK: ucomisd %xmm0, %xmm1 +; UNSAFE: x_ugt_inverse: +; UNSAFE-NEXT: pxor %xmm1, %xmm1 +; UNSAFE-NEXT: minsd %xmm0, %xmm1 +; UNSAFE-NEXT: movapd %xmm1, %xmm0 +; UNSAFE-NEXT: ret +; FINITE: x_ugt_inverse: +; FINITE-NEXT: pxor %xmm1, %xmm1 +; FINITE-NEXT: minsd %xmm0, %xmm1 +; FINITE-NEXT: movapd %xmm1, %xmm0 +; FINITE-NEXT: ret define double @x_ugt_inverse(double %x) nounwind { %c = fcmp ugt double %x, 0.000000e+00 %d = select i1 %c, double 0.000000e+00, double %x @@ -266,9 +459,17 @@ define double @x_ugt_inverse(double %x) nounwind { } ; CHECK: x_ult_inverse: -; CHECK-NEXT: pxor %xmm1, %xmm1 -; CHECK-NEXT: maxsd %xmm1, %xmm0 -; CHECK-NEXT: ret +; CHECK: ucomisd %xmm1, %xmm0 +; UNSAFE: x_ult_inverse: +; UNSAFE-NEXT: pxor %xmm1, %xmm1 +; UNSAFE-NEXT: maxsd %xmm0, %xmm1 +; UNSAFE-NEXT: movapd %xmm1, %xmm0 +; UNSAFE-NEXT: ret +; FINITE: x_ult_inverse: +; FINITE-NEXT: pxor %xmm1, %xmm1 +; FINITE-NEXT: maxsd %xmm0, %xmm1 +; FINITE-NEXT: movapd %xmm1, %xmm0 +; FINITE-NEXT: ret define double @x_ult_inverse(double %x) nounwind { %c = fcmp ult double %x, 0.000000e+00 %d = select i1 %c, double 0.000000e+00, double %x @@ -280,6 +481,14 @@ define double @x_ult_inverse(double %x) nounwind { ; CHECK-NEXT: maxsd %xmm0, %xmm1 ; CHECK-NEXT: movapd %xmm1, %xmm0 ; CHECK-NEXT: ret +; UNSAFE: x_uge: +; UNSAFE-NEXT: pxor %xmm1, %xmm1 +; UNSAFE-NEXT: maxsd %xmm1, %xmm0 +; UNSAFE-NEXT: ret +; FINITE: x_uge: +; FINITE-NEXT: pxor %xmm1, %xmm1 +; FINITE-NEXT: maxsd %xmm1, %xmm0 +; FINITE-NEXT: ret define double @x_uge(double %x) nounwind { %c = fcmp uge double %x, 0.000000e+00 %d = select i1 %c, double %x, double 0.000000e+00 @@ -291,6 +500,14 @@ define double @x_uge(double %x) nounwind { ; CHECK-NEXT: minsd %xmm0, %xmm1 ; CHECK-NEXT: movapd %xmm1, %xmm0 ; CHECK-NEXT: ret +; UNSAFE: x_ule: +; UNSAFE-NEXT: pxor %xmm1, %xmm1 +; UNSAFE-NEXT: minsd %xmm1, %xmm0 +; UNSAFE-NEXT: ret +; FINITE: x_ule: +; FINITE-NEXT: pxor %xmm1, %xmm1 +; FINITE-NEXT: minsd %xmm1, %xmm0 +; FINITE-NEXT: ret define double @x_ule(double %x) nounwind { %c = fcmp ule double %x, 0.000000e+00 %d = select i1 %c, double %x, double 0.000000e+00 @@ -301,6 +518,16 @@ define double @x_ule(double %x) nounwind { ; CHECK-NEXT: pxor %xmm1, %xmm1 ; CHECK-NEXT: minsd %xmm1, %xmm0 ; CHECK-NEXT: ret +; UNSAFE: x_uge_inverse: +; UNSAFE-NEXT: pxor %xmm1, %xmm1 +; UNSAFE-NEXT: minsd %xmm0, %xmm1 +; UNSAFE-NEXT: movapd %xmm1, %xmm0 +; UNSAFE-NEXT: ret +; FINITE: x_uge_inverse: +; FINITE-NEXT: pxor %xmm1, %xmm1 +; FINITE-NEXT: minsd %xmm0, %xmm1 +; FINITE-NEXT: movapd %xmm1, %xmm0 +; FINITE-NEXT: ret define double @x_uge_inverse(double %x) nounwind { %c = fcmp uge double %x, 0.000000e+00 %d = select i1 %c, double 0.000000e+00, double %x @@ -311,16 +538,301 @@ define double @x_uge_inverse(double %x) nounwind { ; CHECK-NEXT: pxor %xmm1, %xmm1 ; CHECK-NEXT: maxsd %xmm1, %xmm0 ; CHECK-NEXT: ret +; UNSAFE: x_ule_inverse: +; UNSAFE-NEXT: pxor %xmm1, %xmm1 +; UNSAFE-NEXT: maxsd %xmm0, %xmm1 +; UNSAFE-NEXT: movapd %xmm1, %xmm0 +; UNSAFE-NEXT: ret +; FINITE: x_ule_inverse: +; FINITE-NEXT: pxor %xmm1, %xmm1 +; FINITE-NEXT: maxsd %xmm0, %xmm1 +; FINITE-NEXT: movapd %xmm1, %xmm0 +; FINITE-NEXT: ret define double @x_ule_inverse(double %x) nounwind { %c = fcmp ule double %x, 0.000000e+00 %d = select i1 %c, double 0.000000e+00, double %x ret double %d } +; CHECK: y_ogt: +; CHECK-NEXT: maxsd {{[^,]*}}, %xmm0 +; CHECK-NEXT: ret +; UNSAFE: y_ogt: +; UNSAFE-NEXT: maxsd {{[^,]*}}, %xmm0 +; UNSAFE-NEXT: ret +; FINITE: y_ogt: +; FINITE-NEXT: maxsd {{[^,]*}}, %xmm0 +; FINITE-NEXT: ret +define double @y_ogt(double %x) nounwind { + %c = fcmp ogt double %x, -0.000000e+00 + %d = select i1 %c, double %x, double -0.000000e+00 + ret double %d +} + +; CHECK: y_olt: +; CHECK-NEXT: minsd {{[^,]*}}, %xmm0 +; CHECK-NEXT: ret +; UNSAFE: y_olt: +; UNSAFE-NEXT: minsd {{[^,]*}}, %xmm0 +; UNSAFE-NEXT: ret +; FINITE: y_olt: +; FINITE-NEXT: minsd {{[^,]*}}, %xmm0 +; FINITE-NEXT: ret +define double @y_olt(double %x) nounwind { + %c = fcmp olt double %x, -0.000000e+00 + %d = select i1 %c, double %x, double -0.000000e+00 + ret double %d +} + +; CHECK: y_ogt_inverse: +; CHECK-NEXT: movsd {{[^,]*}}, %xmm1 +; CHECK-NEXT: minsd %xmm0, %xmm1 +; CHECK-NEXT: movapd %xmm1, %xmm0 +; CHECK-NEXT: ret +; UNSAFE: y_ogt_inverse: +; UNSAFE-NEXT: movsd {{[^,]*}}, %xmm1 +; UNSAFE-NEXT: minsd %xmm0, %xmm1 +; UNSAFE-NEXT: movapd %xmm1, %xmm0 +; UNSAFE-NEXT: ret +; FINITE: y_ogt_inverse: +; FINITE-NEXT: movsd {{[^,]*}}, %xmm1 +; FINITE-NEXT: minsd %xmm0, %xmm1 +; FINITE-NEXT: movapd %xmm1, %xmm0 +; FINITE-NEXT: ret +define double @y_ogt_inverse(double %x) nounwind { + %c = fcmp ogt double %x, -0.000000e+00 + %d = select i1 %c, double -0.000000e+00, double %x + ret double %d +} + +; CHECK: y_olt_inverse: +; CHECK-NEXT: movsd {{[^,]*}}, %xmm1 +; CHECK-NEXT: maxsd %xmm0, %xmm1 +; CHECK-NEXT: movapd %xmm1, %xmm0 +; CHECK-NEXT: ret +; UNSAFE: y_olt_inverse: +; UNSAFE-NEXT: movsd {{[^,]*}}, %xmm1 +; UNSAFE-NEXT: maxsd %xmm0, %xmm1 +; UNSAFE-NEXT: movapd %xmm1, %xmm0 +; UNSAFE-NEXT: ret +; FINITE: y_olt_inverse: +; FINITE-NEXT: movsd {{[^,]*}}, %xmm1 +; FINITE-NEXT: maxsd %xmm0, %xmm1 +; FINITE-NEXT: movapd %xmm1, %xmm0 +; FINITE-NEXT: ret +define double @y_olt_inverse(double %x) nounwind { + %c = fcmp olt double %x, -0.000000e+00 + %d = select i1 %c, double -0.000000e+00, double %x + ret double %d +} + +; CHECK: y_oge: +; CHECK: ucomisd %xmm1, %xmm0 +; UNSAFE: y_oge: +; UNSAFE-NEXT: maxsd {{[^,]*}}, %xmm0 +; UNSAFE-NEXT: ret +; FINITE: y_oge: +; FINITE-NEXT: maxsd {{[^,]*}}, %xmm0 +; FINITE-NEXT: ret +define double @y_oge(double %x) nounwind { + %c = fcmp oge double %x, -0.000000e+00 + %d = select i1 %c, double %x, double -0.000000e+00 + ret double %d +} + +; CHECK: y_ole: +; CHECK: ucomisd %xmm0, %xmm1 +; UNSAFE: y_ole: +; UNSAFE-NEXT: minsd {{[^,]*}}, %xmm0 +; UNSAFE-NEXT: ret +; FINITE: y_ole: +; FINITE-NEXT: minsd {{[^,]*}}, %xmm0 +; FINITE-NEXT: ret +define double @y_ole(double %x) nounwind { + %c = fcmp ole double %x, -0.000000e+00 + %d = select i1 %c, double %x, double -0.000000e+00 + ret double %d +} + +; CHECK: y_oge_inverse: +; CHECK: ucomisd %xmm1, %xmm0 +; UNSAFE: y_oge_inverse: +; UNSAFE-NEXT: movsd {{[^,]*}}, %xmm1 +; UNSAFE-NEXT: minsd %xmm0, %xmm1 +; UNSAFE-NEXT: movapd %xmm1, %xmm0 +; UNSAFE-NEXT: ret +; FINITE: y_oge_inverse: +; FINITE-NEXT: movsd {{[^,]*}}, %xmm1 +; FINITE-NEXT: minsd %xmm0, %xmm1 +; FINITE-NEXT: movapd %xmm1, %xmm0 +; FINITE-NEXT: ret +define double @y_oge_inverse(double %x) nounwind { + %c = fcmp oge double %x, -0.000000e+00 + %d = select i1 %c, double -0.000000e+00, double %x + ret double %d +} + +; CHECK: y_ole_inverse: +; CHECK: ucomisd %xmm0, %xmm1 +; UNSAFE: y_ole_inverse: +; UNSAFE-NEXT: movsd {{[^,]*}}, %xmm1 +; UNSAFE-NEXT: maxsd %xmm0, %xmm1 +; UNSAFE-NEXT: movapd %xmm1, %xmm0 +; UNSAFE-NEXT: ret +; FINITE: y_ole_inverse: +; FINITE-NEXT: movsd {{[^,]*}}, %xmm1 +; FINITE-NEXT: maxsd %xmm0, %xmm1 +; FINITE-NEXT: movapd %xmm1, %xmm0 +; FINITE-NEXT: ret +define double @y_ole_inverse(double %x) nounwind { + %c = fcmp ole double %x, -0.000000e+00 + %d = select i1 %c, double -0.000000e+00, double %x + ret double %d +} + +; CHECK: y_ugt: +; CHECK: ucomisd %xmm0, %xmm1 +; UNSAFE: y_ugt: +; UNSAFE-NEXT: maxsd {{[^,]*}}, %xmm0 +; UNSAFE-NEXT: ret +; FINITE: y_ugt: +; FINITE-NEXT: maxsd {{[^,]*}}, %xmm0 +; FINITE-NEXT: ret +define double @y_ugt(double %x) nounwind { + %c = fcmp ugt double %x, -0.000000e+00 + %d = select i1 %c, double %x, double -0.000000e+00 + ret double %d +} + +; CHECK: y_ult: +; CHECK: ucomisd %xmm1, %xmm0 +; UNSAFE: y_ult: +; UNSAFE-NEXT: minsd {{[^,]*}}, %xmm0 +; UNSAFE-NEXT: ret +; FINITE: y_ult: +; FINITE-NEXT: minsd {{[^,]*}}, %xmm0 +; FINITE-NEXT: ret +define double @y_ult(double %x) nounwind { + %c = fcmp ult double %x, -0.000000e+00 + %d = select i1 %c, double %x, double -0.000000e+00 + ret double %d +} + +; CHECK: y_ugt_inverse: +; CHECK: ucomisd %xmm0, %xmm1 +; UNSAFE: y_ugt_inverse: +; UNSAFE-NEXT: movsd {{[^,]*}}, %xmm1 +; UNSAFE-NEXT: minsd %xmm0, %xmm1 +; UNSAFE-NEXT: movapd %xmm1, %xmm0 +; UNSAFE-NEXT: ret +; FINITE: y_ugt_inverse: +; FINITE-NEXT: movsd {{[^,]*}}, %xmm1 +; FINITE-NEXT: minsd %xmm0, %xmm1 +; FINITE-NEXT: movapd %xmm1, %xmm0 +; FINITE-NEXT: ret +define double @y_ugt_inverse(double %x) nounwind { + %c = fcmp ugt double %x, -0.000000e+00 + %d = select i1 %c, double -0.000000e+00, double %x + ret double %d +} + +; CHECK: y_ult_inverse: +; CHECK: ucomisd %xmm1, %xmm0 +; UNSAFE: y_ult_inverse: +; UNSAFE-NEXT: movsd {{[^,]*}}, %xmm1 +; UNSAFE-NEXT: maxsd %xmm0, %xmm1 +; UNSAFE-NEXT: movapd %xmm1, %xmm0 +; UNSAFE-NEXT: ret +; FINITE: y_ult_inverse: +; FINITE-NEXT: movsd {{[^,]*}}, %xmm1 +; FINITE-NEXT: maxsd %xmm0, %xmm1 +; FINITE-NEXT: movapd %xmm1, %xmm0 +; FINITE-NEXT: ret +define double @y_ult_inverse(double %x) nounwind { + %c = fcmp ult double %x, -0.000000e+00 + %d = select i1 %c, double -0.000000e+00, double %x + ret double %d +} + +; CHECK: y_uge: +; CHECK-NEXT: movsd {{[^,]*}}, %xmm1 +; CHECK-NEXT: maxsd %xmm0, %xmm1 +; CHECK-NEXT: movapd %xmm1, %xmm0 +; CHECK-NEXT: ret +; UNSAFE: y_uge: +; UNSAFE-NEXT: maxsd {{[^,]*}}, %xmm0 +; UNSAFE-NEXT: ret +; FINITE: y_uge: +; FINITE-NEXT: maxsd {{[^,]*}}, %xmm0 +; FINITE-NEXT: ret +define double @y_uge(double %x) nounwind { + %c = fcmp uge double %x, -0.000000e+00 + %d = select i1 %c, double %x, double -0.000000e+00 + ret double %d +} + +; CHECK: y_ule: +; CHECK-NEXT: movsd {{[^,]*}}, %xmm1 +; CHECK-NEXT: minsd %xmm0, %xmm1 +; CHECK-NEXT: movapd %xmm1, %xmm0 +; CHECK-NEXT: ret +; UNSAFE: y_ule: +; UNSAFE-NEXT: minsd {{[^,]*}}, %xmm0 +; UNSAFE-NEXT: ret +; FINITE: y_ule: +; FINITE-NEXT: minsd {{[^,]*}}, %xmm0 +; FINITE-NEXT: ret +define double @y_ule(double %x) nounwind { + %c = fcmp ule double %x, -0.000000e+00 + %d = select i1 %c, double %x, double -0.000000e+00 + ret double %d +} + +; CHECK: y_uge_inverse: +; CHECK-NEXT: minsd {{[^,]*}}, %xmm0 +; CHECK-NEXT: ret +; UNSAFE: y_uge_inverse: +; UNSAFE-NEXT: movsd {{[^,]*}}, %xmm1 +; UNSAFE-NEXT: minsd %xmm0, %xmm1 +; UNSAFE-NEXT: movapd %xmm1, %xmm0 +; UNSAFE-NEXT: ret +; FINITE: y_uge_inverse: +; FINITE-NEXT: movsd {{[^,]*}}, %xmm1 +; FINITE-NEXT: minsd %xmm0, %xmm1 +; FINITE-NEXT: movapd %xmm1, %xmm0 +; FINITE-NEXT: ret +define double @y_uge_inverse(double %x) nounwind { + %c = fcmp uge double %x, -0.000000e+00 + %d = select i1 %c, double -0.000000e+00, double %x + ret double %d +} + +; CHECK: y_ule_inverse: +; CHECK-NEXT: maxsd {{[^,]*}}, %xmm0 +; CHECK-NEXT: ret +; UNSAFE: y_ule_inverse: +; UNSAFE-NEXT: movsd {{[^,]*}}, %xmm1 +; UNSAFE-NEXT: maxsd %xmm0, %xmm1 +; UNSAFE-NEXT: movapd %xmm1, %xmm0 +; UNSAFE-NEXT: ret +; FINITE: y_ule_inverse: +; FINITE-NEXT: movsd {{[^,]*}}, %xmm1 +; FINITE-NEXT: maxsd %xmm0, %xmm1 +; FINITE-NEXT: movapd %xmm1, %xmm0 +; FINITE-NEXT: ret +define double @y_ule_inverse(double %x) nounwind { + %c = fcmp ule double %x, -0.000000e+00 + %d = select i1 %c, double -0.000000e+00, double %x + ret double %d +} ; Test a few more misc. cases. ; CHECK: clampTo3k_a: ; CHECK: minsd +; UNSAFE: clampTo3k_a: +; UNSAFE: minsd +; FINITE: clampTo3k_a: +; FINITE: minsd define double @clampTo3k_a(double %x) nounwind readnone { entry: %0 = fcmp ogt double %x, 3.000000e+03 ; <i1> [#uses=1] @@ -330,6 +842,10 @@ entry: ; CHECK: clampTo3k_b: ; CHECK: minsd +; UNSAFE: clampTo3k_b: +; UNSAFE: minsd +; FINITE: clampTo3k_b: +; FINITE: minsd define double @clampTo3k_b(double %x) nounwind readnone { entry: %0 = fcmp uge double %x, 3.000000e+03 ; <i1> [#uses=1] @@ -339,6 +855,10 @@ entry: ; CHECK: clampTo3k_c: ; CHECK: maxsd +; UNSAFE: clampTo3k_c: +; UNSAFE: maxsd +; FINITE: clampTo3k_c: +; FINITE: maxsd define double @clampTo3k_c(double %x) nounwind readnone { entry: %0 = fcmp olt double %x, 3.000000e+03 ; <i1> [#uses=1] @@ -348,6 +868,10 @@ entry: ; CHECK: clampTo3k_d: ; CHECK: maxsd +; UNSAFE: clampTo3k_d: +; UNSAFE: maxsd +; FINITE: clampTo3k_d: +; FINITE: maxsd define double @clampTo3k_d(double %x) nounwind readnone { entry: %0 = fcmp ule double %x, 3.000000e+03 ; <i1> [#uses=1] @@ -357,6 +881,10 @@ entry: ; CHECK: clampTo3k_e: ; CHECK: maxsd +; UNSAFE: clampTo3k_e: +; UNSAFE: maxsd +; FINITE: clampTo3k_e: +; FINITE: maxsd define double @clampTo3k_e(double %x) nounwind readnone { entry: %0 = fcmp olt double %x, 3.000000e+03 ; <i1> [#uses=1] @@ -366,6 +894,10 @@ entry: ; CHECK: clampTo3k_f: ; CHECK: maxsd +; UNSAFE: clampTo3k_f: +; UNSAFE: maxsd +; FINITE: clampTo3k_f: +; FINITE: maxsd define double @clampTo3k_f(double %x) nounwind readnone { entry: %0 = fcmp ule double %x, 3.000000e+03 ; <i1> [#uses=1] @@ -375,6 +907,10 @@ entry: ; CHECK: clampTo3k_g: ; CHECK: minsd +; UNSAFE: clampTo3k_g: +; UNSAFE: minsd +; FINITE: clampTo3k_g: +; FINITE: minsd define double @clampTo3k_g(double %x) nounwind readnone { entry: %0 = fcmp ogt double %x, 3.000000e+03 ; <i1> [#uses=1] @@ -384,6 +920,10 @@ entry: ; CHECK: clampTo3k_h: ; CHECK: minsd +; UNSAFE: clampTo3k_h: +; UNSAFE: minsd +; FINITE: clampTo3k_h: +; FINITE: minsd define double @clampTo3k_h(double %x) nounwind readnone { entry: %0 = fcmp uge double %x, 3.000000e+03 ; <i1> [#uses=1] diff --git a/test/CodeGen/X86/sse3.ll b/test/CodeGen/X86/sse3.ll index b2af7c947d98..921161e4a122 100644 --- a/test/CodeGen/X86/sse3.ll +++ b/test/CodeGen/X86/sse3.ll @@ -144,10 +144,9 @@ define void @t9(<4 x float>* %r, <2 x i32>* %A) nounwind { store <4 x float> %tmp13, <4 x float>* %r ret void ; X64: t9: -; X64: movsd (%rsi), %xmm0 -; X64: movaps (%rdi), %xmm1 -; X64: movlhps %xmm0, %xmm1 -; X64: movaps %xmm1, (%rdi) +; X64: movaps (%rdi), %xmm0 +; X64: movhps (%rsi), %xmm0 +; X64: movaps %xmm0, (%rdi) ; X64: ret } diff --git a/test/CodeGen/X86/stack-align.ll b/test/CodeGen/X86/stack-align.ll index cb65e9b50fe2..e971ef70dbd4 100644 --- a/test/CodeGen/X86/stack-align.ll +++ b/test/CodeGen/X86/stack-align.ll @@ -1,4 +1,4 @@ -; RUN: llc < %s -relocation-model=static -mcpu=yonah | grep {andpd.*4(%esp), %xmm} +; RUN: llc < %s -relocation-model=static -realign-stack=1 -mcpu=yonah | FileCheck %s ; The double argument is at 4(esp) which is 16-byte aligned, allowing us to ; fold the load into the andpd. @@ -12,6 +12,7 @@ entry: %tmp = getelementptr { double, double }* %z, i32 0, i32 0 ; <double*> [#uses=1] %tmp1 = load double* %tmp, align 8 ; <double> [#uses=1] %tmp2 = tail call double @fabs( double %tmp1 ) ; <double> [#uses=1] + ; CHECK: andpd{{.*}}4(%esp), %xmm %tmp3 = load double* @G, align 16 ; <double> [#uses=1] %tmp4 = tail call double @fabs( double %tmp3 ) ; <double> [#uses=1] %tmp6 = fadd double %tmp4, %tmp2 ; <double> [#uses=1] @@ -19,4 +20,20 @@ entry: ret void } +define void @test2() alignstack(16) { +entry: + ; CHECK: andl{{.*}}$-16, %esp + ret void +} + +; Use a call to force a spill. +define <2 x double> @test3(<2 x double> %x, <2 x double> %y) alignstack(32) { +entry: + ; CHECK: andl{{.*}}$-32, %esp + call void @test2() + %A = mul <2 x double> %x, %y + ret <2 x double> %A +} + declare double @fabs(double) + diff --git a/test/CodeGen/X86/store_op_load_fold.ll b/test/CodeGen/X86/store_op_load_fold.ll index 66d0e47c6d48..6e47eb397d1d 100644 --- a/test/CodeGen/X86/store_op_load_fold.ll +++ b/test/CodeGen/X86/store_op_load_fold.ll @@ -4,7 +4,7 @@ @X = internal global i16 0 ; <i16*> [#uses=2] -define void @foo() { +define void @foo() nounwind { %tmp.0 = load i16* @X ; <i16> [#uses=1] %tmp.3 = add i16 %tmp.0, 329 ; <i16> [#uses=1] store i16 %tmp.3, i16* @X diff --git a/test/CodeGen/X86/store_op_load_fold2.ll b/test/CodeGen/X86/store_op_load_fold2.ll index 0ccfe470db5f..46e59e95e53f 100644 --- a/test/CodeGen/X86/store_op_load_fold2.ll +++ b/test/CodeGen/X86/store_op_load_fold2.ll @@ -1,21 +1,12 @@ -; RUN: llc < %s -march=x86 -x86-asm-syntax=intel | \ -; RUN: grep {and DWORD PTR} | count 2 +; RUN: llc < %s -march=x86 -x86-asm-syntax=intel | FileCheck %s target datalayout = "e-p:32:32" %struct.Macroblock = type { i32, i32, i32, i32, i32, [8 x i32], %struct.Macroblock*, %struct.Macroblock*, i32, [2 x [4 x [4 x [2 x i32]]]], [16 x i8], [16 x i8], i32, i64, [4 x i32], [4 x i32], i64, i32, i32, i32, i32, i32, i32, i32, i32, i32, i32, i32, i32, i32, i32, i32, i16, double, i32, i32, i32, i32, i32, i32, i32, i32, i32 } -define internal fastcc i32 @dct_chroma(i32 %uv, i32 %cr_cbp) { -entry: - br i1 true, label %cond_true2732.preheader, label %cond_true129 -cond_true129: ; preds = %entry - ret i32 0 +define internal fastcc i32 @dct_chroma(i32 %uv, i32 %cr_cbp) nounwind { cond_true2732.preheader: ; preds = %entry %tmp2666 = getelementptr %struct.Macroblock* null, i32 0, i32 13 ; <i64*> [#uses=2] %tmp2674 = trunc i32 0 to i8 ; <i8> [#uses=1] - br i1 true, label %cond_true2732.preheader.split.us, label %cond_true2732.preheader.split -cond_true2732.preheader.split.us: ; preds = %cond_true2732.preheader - br i1 true, label %cond_true2732.outer.us.us, label %cond_true2732.outer.us -cond_true2732.outer.us.us: ; preds = %cond_true2732.preheader.split.us %tmp2667.us.us = load i64* %tmp2666 ; <i64> [#uses=1] %tmp2670.us.us = load i64* null ; <i64> [#uses=1] %shift.upgrd.1 = zext i8 %tmp2674 to i64 ; <i64> [#uses=1] @@ -24,11 +15,10 @@ cond_true2732.outer.us.us: ; preds = %cond_true2732.preheader.split %tmp2676.us.us = and i64 %tmp2667.us.us, %tmp2675not.us.us ; <i64> [#uses=1] store i64 %tmp2676.us.us, i64* %tmp2666 ret i32 0 -cond_true2732.outer.us: ; preds = %cond_true2732.preheader.split.us - ret i32 0 -cond_true2732.preheader.split: ; preds = %cond_true2732.preheader - ret i32 0 -cond_next2752: ; No predecessors! - ret i32 0 + +; CHECK: and {{E..}}, DWORD PTR [360] +; CHECK: and DWORD PTR [356], {{E..}} +; CHECK: mov DWORD PTR [360], {{E..}} + } diff --git a/test/CodeGen/X86/trunc-to-bool.ll b/test/CodeGen/X86/trunc-to-bool.ll index bfab1aef9014..60620841064f 100644 --- a/test/CodeGen/X86/trunc-to-bool.ll +++ b/test/CodeGen/X86/trunc-to-bool.ll @@ -3,13 +3,14 @@ ; value and as the operand of a branch. ; RUN: llc < %s -march=x86 | FileCheck %s -define i1 @test1(i32 %X) zeroext { +define i1 @test1(i32 %X) zeroext nounwind { %Y = trunc i32 %X to i1 ret i1 %Y } +; CHECK: test1: ; CHECK: andl $1, %eax -define i1 @test2(i32 %val, i32 %mask) { +define i1 @test2(i32 %val, i32 %mask) nounwind { entry: %shifted = ashr i32 %val, %mask %anded = and i32 %shifted, 1 @@ -20,9 +21,10 @@ ret_true: ret_false: ret i1 false } -; CHECK: testb $1, %al +; CHECK: test2: +; CHECK: btl %eax -define i32 @test3(i8* %ptr) { +define i32 @test3(i8* %ptr) nounwind { %val = load i8* %ptr %tmp = trunc i8 %val to i1 br i1 %tmp, label %cond_true, label %cond_false @@ -31,9 +33,10 @@ cond_true: cond_false: ret i32 42 } -; CHECK: testb $1, %al +; CHECK: test3: +; CHECK: testb $1, (%eax) -define i32 @test4(i8* %ptr) { +define i32 @test4(i8* %ptr) nounwind { %tmp = ptrtoint i8* %ptr to i1 br i1 %tmp, label %cond_true, label %cond_false cond_true: @@ -41,9 +44,10 @@ cond_true: cond_false: ret i32 42 } -; CHECK: testb $1, %al +; CHECK: test4: +; CHECK: testb $1, 4(%esp) -define i32 @test6(double %d) { +define i32 @test5(double %d) nounwind { %tmp = fptosi double %d to i1 br i1 %tmp, label %cond_true, label %cond_false cond_true: @@ -51,4 +55,5 @@ cond_true: cond_false: ret i32 42 } +; CHECK: test5: ; CHECK: testb $1 diff --git a/test/CodeGen/X86/vec_cast.ll b/test/CodeGen/X86/vec_cast.ll index 1f899b3c20aa..6f18d13cc9d3 100644 --- a/test/CodeGen/X86/vec_cast.ll +++ b/test/CodeGen/X86/vec_cast.ll @@ -31,11 +31,10 @@ define <1 x i32> @f(<1 x i16> %a) nounwind { ret <1 x i32> %c } -; TODO: Legalize doesn't yet handle this. -;define <8 x i16> @g(<8 x i32> %a) nounwind { -; %c = trunc <8 x i32> %a to <8 x i16> -; ret <8 x i16> %c -;} +define <8 x i16> @g(<8 x i32> %a) nounwind { + %c = trunc <8 x i32> %a to <8 x i16> + ret <8 x i16> %c +} define <3 x i16> @h(<3 x i32> %a) nounwind { %c = trunc <3 x i32> %a to <3 x i16> @@ -46,3 +45,12 @@ define <1 x i16> @i(<1 x i32> %a) nounwind { %c = trunc <1 x i32> %a to <1 x i16> ret <1 x i16> %c } + +; PR6438 +define void @__OpenCL_math_kernel4_kernel() nounwind { + %tmp12.i = and <4 x i32> zeroinitializer, <i32 2139095040, i32 2139095040, i32 2139095040, i32 2139095040> ; <<4 x i32>> [#uses=1] + %cmp13.i = icmp eq <4 x i32> %tmp12.i, <i32 2139095040, i32 2139095040, i32 2139095040, i32 2139095040> ; <<4 x i1>> [#uses=2] + %cmp.ext14.i = sext <4 x i1> %cmp13.i to <4 x i32> ; <<4 x i32>> [#uses=0] + %tmp2110.i = and <4 x i1> %cmp13.i, zeroinitializer ; <<4 x i1>> [#uses=0] + ret void +} diff --git a/test/CodeGen/X86/vec_insert.ll b/test/CodeGen/X86/vec_insert.ll index a7274a900044..4e5d445ff623 100644 --- a/test/CodeGen/X86/vec_insert.ll +++ b/test/CodeGen/X86/vec_insert.ll @@ -1,7 +1,7 @@ ; RUN: llc < %s -march=x86 -mattr=+sse2,-sse41 | grep movss | count 1 ; RUN: llc < %s -march=x86 -mattr=+sse2,-sse41 | not grep pinsrw -define void @test(<4 x float>* %F, i32 %I) { +define void @test(<4 x float>* %F, i32 %I) nounwind { %tmp = load <4 x float>* %F ; <<4 x float>> [#uses=1] %f = sitofp i32 %I to float ; <float> [#uses=1] %tmp1 = insertelement <4 x float> %tmp, float %f, i32 0 ; <<4 x float>> [#uses=2] @@ -10,7 +10,7 @@ define void @test(<4 x float>* %F, i32 %I) { ret void } -define void @test2(<4 x float>* %F, i32 %I, float %g) { +define void @test2(<4 x float>* %F, i32 %I, float %g) nounwind { %tmp = load <4 x float>* %F ; <<4 x float>> [#uses=1] %f = sitofp i32 %I to float ; <float> [#uses=1] %tmp1 = insertelement <4 x float> %tmp, float %f, i32 2 ; <<4 x float>> [#uses=1] diff --git a/test/CodeGen/X86/vec_shuffle-36.ll b/test/CodeGen/X86/vec_shuffle-36.ll index 8a93a7eeee3b..1ea37c881e71 100644 --- a/test/CodeGen/X86/vec_shuffle-36.ll +++ b/test/CodeGen/X86/vec_shuffle-36.ll @@ -1,9 +1,16 @@ -; RUN: llc < %s -march=x86 -mattr=sse41 -o %t -; RUN: grep pshufb %t | count 1 - +; RUN: llc < %s -march=x86-64 -mattr=sse41 | FileCheck %s define <8 x i16> @shuf6(<8 x i16> %T0, <8 x i16> %T1) nounwind readnone { +; CHECK: pshufb +; CHECK-NOT: pshufb +; CHECK: ret entry: - %tmp9 = shufflevector <8 x i16> %T0, <8 x i16> %T1, <8 x i32> < i32 3, i32 2, i32 0, i32 2, i32 1, i32 5, i32 6 , i32 undef > - ret <8 x i16> %tmp9 + %tmp9 = shufflevector <8 x i16> %T0, <8 x i16> %T1, <8 x i32> < i32 3, i32 2, i32 0, i32 2, i32 1, i32 5, i32 6 , i32 undef > + ret <8 x i16> %tmp9 } + +define <8 x i16> @shuf7(<8 x i16> %t0) { +; CHECK: pshufd + %tmp10 = shufflevector <8 x i16> %t0, <8 x i16> undef, <8 x i32> < i32 undef, i32 2, i32 2, i32 2, i32 2, i32 2, i32 undef, i32 undef > + ret <8 x i16> %tmp10 +}
\ No newline at end of file diff --git a/test/CodeGen/X86/vec_ss_load_fold.ll b/test/CodeGen/X86/vec_ss_load_fold.ll index b1613fb3a374..c8b2927b71f3 100644 --- a/test/CodeGen/X86/vec_ss_load_fold.ll +++ b/test/CodeGen/X86/vec_ss_load_fold.ll @@ -1,6 +1,4 @@ -; RUN: llc < %s -march=x86 -mattr=+sse,+sse2 -o %t -; RUN: grep minss %t | grep CPI | count 2 -; RUN: grep CPI %t | not grep movss +; RUN: llc < %s -march=x86 -mattr=+sse,+sse2,+sse41 | FileCheck %s target datalayout = "e-p:32:32" target triple = "i686-apple-darwin8.7.2" @@ -17,6 +15,10 @@ define i16 @test1(float %f) nounwind { %tmp.upgrd.1 = tail call i32 @llvm.x86.sse.cvttss2si( <4 x float> %tmp59 ) ; <i32> [#uses=1] %tmp69 = trunc i32 %tmp.upgrd.1 to i16 ; <i16> [#uses=1] ret i16 %tmp69 +; CHECK: test1: +; CHECK: subss LCPI1_ +; CHECK: mulss LCPI1_ +; CHECK: minss LCPI1_ } define i16 @test2(float %f) nounwind { @@ -28,6 +30,10 @@ define i16 @test2(float %f) nounwind { %tmp = tail call i32 @llvm.x86.sse.cvttss2si( <4 x float> %tmp59 ) ; <i32> [#uses=1] %tmp69 = trunc i32 %tmp to i16 ; <i16> [#uses=1] ret i16 %tmp69 +; CHECK: test2: +; CHECK: addss LCPI2_ +; CHECK: mulss LCPI2_ +; CHECK: minss LCPI2_ } declare <4 x float> @llvm.x86.sse.sub.ss(<4 x float>, <4 x float>) @@ -39,3 +45,28 @@ declare <4 x float> @llvm.x86.sse.min.ss(<4 x float>, <4 x float>) declare <4 x float> @llvm.x86.sse.max.ss(<4 x float>, <4 x float>) declare i32 @llvm.x86.sse.cvttss2si(<4 x float>) + + +declare <4 x float> @llvm.x86.sse41.round.ss(<4 x float>, <4 x float>, i32) +declare <4 x float> @f() + +define <4 x float> @test3(<4 x float> %A, float *%b, i32 %C) nounwind { + %a = load float *%b + %B = insertelement <4 x float> undef, float %a, i32 0 + %X = call <4 x float> @llvm.x86.sse41.round.ss(<4 x float> %A, <4 x float> %B, i32 4) + ret <4 x float> %X +; CHECK: test3: +; CHECK: roundss $4, (%eax), %xmm0 +} + +define <4 x float> @test4(<4 x float> %A, float *%b, i32 %C) nounwind { + %a = load float *%b + %B = insertelement <4 x float> undef, float %a, i32 0 + %q = call <4 x float> @f() + %X = call <4 x float> @llvm.x86.sse41.round.ss(<4 x float> %q, <4 x float> %B, i32 4) + ret <4 x float> %X +; CHECK: test4: +; CHECK: movss (%eax), %xmm +; CHECK: call +; CHECK: roundss $4, %xmm{{.*}}, %xmm0 +} diff --git a/test/CodeGen/X86/xor-icmp.ll b/test/CodeGen/X86/xor-icmp.ll index a6bdb13ec6b4..2d75c5d76201 100644 --- a/test/CodeGen/X86/xor-icmp.ll +++ b/test/CodeGen/X86/xor-icmp.ll @@ -1,5 +1,6 @@ ; RUN: llc < %s -march=x86 | FileCheck %s -check-prefix=X32 ; RUN: llc < %s -march=x86-64 | FileCheck %s -check-prefix=X64 +; rdar://7367229 define i32 @t(i32 %a, i32 %b) nounwind ssp { entry: @@ -34,3 +35,33 @@ bb1: ; preds = %entry declare i32 @foo(...) declare i32 @bar(...) + +define i32 @t2(i32 %x, i32 %y) nounwind ssp { +; X32: t2: +; X32: cmpl +; X32: sete +; X32: cmpl +; X32: sete +; X32-NOT: xor +; X32: je + +; X64: t2: +; X64: testl +; X64: sete +; X64: testl +; X64: sete +; X64-NOT: xor +; X64: je +entry: + %0 = icmp eq i32 %x, 0 ; <i1> [#uses=1] + %1 = icmp eq i32 %y, 0 ; <i1> [#uses=1] + %2 = xor i1 %1, %0 ; <i1> [#uses=1] + br i1 %2, label %bb, label %return + +bb: ; preds = %entry + %3 = tail call i32 (...)* @foo() nounwind ; <i32> [#uses=0] + ret i32 undef + +return: ; preds = %entry + ret i32 undef +} |