diff options
Diffstat (limited to 'test/CodeGen/SystemZ/frame-14.ll')
| -rw-r--r-- | test/CodeGen/SystemZ/frame-14.ll | 172 |
1 files changed, 88 insertions, 84 deletions
diff --git a/test/CodeGen/SystemZ/frame-14.ll b/test/CodeGen/SystemZ/frame-14.ll index d8ff0a54a761..3b48179c40b6 100644 --- a/test/CodeGen/SystemZ/frame-14.ll +++ b/test/CodeGen/SystemZ/frame-14.ll @@ -1,9 +1,13 @@ ; Test the handling of base + displacement addresses for large frames, ; in cases where both 12-bit and 20-bit displacements are allowed. +; The tests here assume z10 register pressure, without the high words +; being available. +; +; RUN: llc < %s -mtriple=s390x-linux-gnu -mcpu=z10 | \ +; RUN: FileCheck -check-prefix=CHECK-NOFP %s +; RUN: llc < %s -mtriple=s390x-linux-gnu -mcpu=z10 -disable-fp-elim | \ +; RUN: FileCheck -check-prefix=CHECK-FP %s ; -; RUN: llc < %s -mtriple=s390x-linux-gnu | FileCheck -check-prefix=CHECK-NOFP %s -; RUN: llc < %s -mtriple=s390x-linux-gnu -disable-fp-elim | FileCheck -check-prefix=CHECK-FP %s - ; This file tests what happens when a displacement is converted from ; being relative to the start of a frame object to being relative to ; the frame itself. In some cases the test is only possible if two @@ -16,21 +20,21 @@ ; First check the highest offset that is in range of the 12-bit form. ; -; The last in-range doubleword offset is 4088. Since the frame has an -; emergency spill slot at 160(%r15), the amount that we need to allocate -; in order to put another object at offset 4088 is 4088 - 168 = 3920 bytes. +; The last in-range doubleword offset is 4088. Since the frame has two +; emergency spill slots at 160(%r15), the amount that we need to allocate +; in order to put another object at offset 4088 is 4088 - 176 = 3912 bytes. define void @f1() { -; CHECK-NOFP: f1: +; CHECK-NOFP-LABEL: f1: ; CHECK-NOFP: mvi 4095(%r15), 42 ; CHECK-NOFP: br %r14 ; -; CHECK-FP: f1: +; CHECK-FP-LABEL: f1: ; CHECK-FP: mvi 4095(%r11), 42 ; CHECK-FP: br %r14 - %region1 = alloca [3920 x i8], align 8 - %region2 = alloca [3920 x i8], align 8 - %ptr1 = getelementptr inbounds [3920 x i8]* %region1, i64 0, i64 7 - %ptr2 = getelementptr inbounds [3920 x i8]* %region2, i64 0, i64 7 + %region1 = alloca [3912 x i8], align 8 + %region2 = alloca [3912 x i8], align 8 + %ptr1 = getelementptr inbounds [3912 x i8]* %region1, i64 0, i64 7 + %ptr2 = getelementptr inbounds [3912 x i8]* %region2, i64 0, i64 7 store volatile i8 42, i8 *%ptr1 store volatile i8 42, i8 *%ptr2 ret void @@ -38,17 +42,17 @@ define void @f1() { ; Test the first offset that is out-of-range of the 12-bit form. define void @f2() { -; CHECK-NOFP: f2: +; CHECK-NOFP-LABEL: f2: ; CHECK-NOFP: mviy 4096(%r15), 42 ; CHECK-NOFP: br %r14 ; -; CHECK-FP: f2: +; CHECK-FP-LABEL: f2: ; CHECK-FP: mviy 4096(%r11), 42 ; CHECK-FP: br %r14 - %region1 = alloca [3920 x i8], align 8 - %region2 = alloca [3920 x i8], align 8 - %ptr1 = getelementptr inbounds [3920 x i8]* %region1, i64 0, i64 8 - %ptr2 = getelementptr inbounds [3920 x i8]* %region2, i64 0, i64 8 + %region1 = alloca [3912 x i8], align 8 + %region2 = alloca [3912 x i8], align 8 + %ptr1 = getelementptr inbounds [3912 x i8]* %region1, i64 0, i64 8 + %ptr2 = getelementptr inbounds [3912 x i8]* %region2, i64 0, i64 8 store volatile i8 42, i8 *%ptr1 store volatile i8 42, i8 *%ptr2 ret void @@ -57,19 +61,19 @@ define void @f2() { ; Test the last offset that is in range of the 20-bit form. ; ; The last in-range doubleword offset is 524280, so by the same reasoning -; as above, we need to allocate objects of 524280 - 168 = 524122 bytes. +; as above, we need to allocate objects of 524280 - 176 = 524104 bytes. define void @f3() { -; CHECK-NOFP: f3: +; CHECK-NOFP-LABEL: f3: ; CHECK-NOFP: mviy 524287(%r15), 42 ; CHECK-NOFP: br %r14 ; -; CHECK-FP: f3: +; CHECK-FP-LABEL: f3: ; CHECK-FP: mviy 524287(%r11), 42 ; CHECK-FP: br %r14 - %region1 = alloca [524112 x i8], align 8 - %region2 = alloca [524112 x i8], align 8 - %ptr1 = getelementptr inbounds [524112 x i8]* %region1, i64 0, i64 7 - %ptr2 = getelementptr inbounds [524112 x i8]* %region2, i64 0, i64 7 + %region1 = alloca [524104 x i8], align 8 + %region2 = alloca [524104 x i8], align 8 + %ptr1 = getelementptr inbounds [524104 x i8]* %region1, i64 0, i64 7 + %ptr2 = getelementptr inbounds [524104 x i8]* %region2, i64 0, i64 7 store volatile i8 42, i8 *%ptr1 store volatile i8 42, i8 *%ptr2 ret void @@ -79,21 +83,21 @@ define void @f3() { ; and the offset is also out of LAY's range, so expect a constant load ; followed by an addition. define void @f4() { -; CHECK-NOFP: f4: +; CHECK-NOFP-LABEL: f4: ; CHECK-NOFP: llilh %r1, 8 ; CHECK-NOFP: agr %r1, %r15 ; CHECK-NOFP: mvi 0(%r1), 42 ; CHECK-NOFP: br %r14 ; -; CHECK-FP: f4: +; CHECK-FP-LABEL: f4: ; CHECK-FP: llilh %r1, 8 ; CHECK-FP: agr %r1, %r11 ; CHECK-FP: mvi 0(%r1), 42 ; CHECK-FP: br %r14 - %region1 = alloca [524112 x i8], align 8 - %region2 = alloca [524112 x i8], align 8 - %ptr1 = getelementptr inbounds [524112 x i8]* %region1, i64 0, i64 8 - %ptr2 = getelementptr inbounds [524112 x i8]* %region2, i64 0, i64 8 + %region1 = alloca [524104 x i8], align 8 + %region2 = alloca [524104 x i8], align 8 + %ptr1 = getelementptr inbounds [524104 x i8]* %region1, i64 0, i64 8 + %ptr2 = getelementptr inbounds [524104 x i8]* %region2, i64 0, i64 8 store volatile i8 42, i8 *%ptr1 store volatile i8 42, i8 *%ptr2 ret void @@ -102,21 +106,21 @@ define void @f4() { ; Add 4095 to the previous offset, to test the other end of the MVI range. ; The instruction will actually be STCY before frame lowering. define void @f5() { -; CHECK-NOFP: f5: +; CHECK-NOFP-LABEL: f5: ; CHECK-NOFP: llilh %r1, 8 ; CHECK-NOFP: agr %r1, %r15 ; CHECK-NOFP: mvi 4095(%r1), 42 ; CHECK-NOFP: br %r14 ; -; CHECK-FP: f5: +; CHECK-FP-LABEL: f5: ; CHECK-FP: llilh %r1, 8 ; CHECK-FP: agr %r1, %r11 ; CHECK-FP: mvi 4095(%r1), 42 ; CHECK-FP: br %r14 - %region1 = alloca [524112 x i8], align 8 - %region2 = alloca [524112 x i8], align 8 - %ptr1 = getelementptr inbounds [524112 x i8]* %region1, i64 0, i64 4103 - %ptr2 = getelementptr inbounds [524112 x i8]* %region2, i64 0, i64 4103 + %region1 = alloca [524104 x i8], align 8 + %region2 = alloca [524104 x i8], align 8 + %ptr1 = getelementptr inbounds [524104 x i8]* %region1, i64 0, i64 4103 + %ptr2 = getelementptr inbounds [524104 x i8]* %region2, i64 0, i64 4103 store volatile i8 42, i8 *%ptr1 store volatile i8 42, i8 *%ptr2 ret void @@ -124,21 +128,21 @@ define void @f5() { ; Test the next offset after that, which uses MVIY instead of MVI. define void @f6() { -; CHECK-NOFP: f6: +; CHECK-NOFP-LABEL: f6: ; CHECK-NOFP: llilh %r1, 8 ; CHECK-NOFP: agr %r1, %r15 ; CHECK-NOFP: mviy 4096(%r1), 42 ; CHECK-NOFP: br %r14 ; -; CHECK-FP: f6: +; CHECK-FP-LABEL: f6: ; CHECK-FP: llilh %r1, 8 ; CHECK-FP: agr %r1, %r11 ; CHECK-FP: mviy 4096(%r1), 42 ; CHECK-FP: br %r14 - %region1 = alloca [524112 x i8], align 8 - %region2 = alloca [524112 x i8], align 8 - %ptr1 = getelementptr inbounds [524112 x i8]* %region1, i64 0, i64 4104 - %ptr2 = getelementptr inbounds [524112 x i8]* %region2, i64 0, i64 4104 + %region1 = alloca [524104 x i8], align 8 + %region2 = alloca [524104 x i8], align 8 + %ptr1 = getelementptr inbounds [524104 x i8]* %region1, i64 0, i64 4104 + %ptr2 = getelementptr inbounds [524104 x i8]* %region2, i64 0, i64 4104 store volatile i8 42, i8 *%ptr1 store volatile i8 42, i8 *%ptr2 ret void @@ -149,21 +153,21 @@ define void @f6() { ; anchors 0x10000 bytes apart, so that the high part can be loaded using ; LLILH while still using MVI in more cases than 0x40000 anchors would. define void @f7() { -; CHECK-NOFP: f7: +; CHECK-NOFP-LABEL: f7: ; CHECK-NOFP: llilh %r1, 23 ; CHECK-NOFP: agr %r1, %r15 ; CHECK-NOFP: mviy 65535(%r1), 42 ; CHECK-NOFP: br %r14 ; -; CHECK-FP: f7: +; CHECK-FP-LABEL: f7: ; CHECK-FP: llilh %r1, 23 ; CHECK-FP: agr %r1, %r11 ; CHECK-FP: mviy 65535(%r1), 42 ; CHECK-FP: br %r14 - %region1 = alloca [1048408 x i8], align 8 - %region2 = alloca [1048408 x i8], align 8 - %ptr1 = getelementptr inbounds [1048408 x i8]* %region1, i64 0, i64 524287 - %ptr2 = getelementptr inbounds [1048408 x i8]* %region2, i64 0, i64 524287 + %region1 = alloca [1048400 x i8], align 8 + %region2 = alloca [1048400 x i8], align 8 + %ptr1 = getelementptr inbounds [1048400 x i8]* %region1, i64 0, i64 524287 + %ptr2 = getelementptr inbounds [1048400 x i8]* %region2, i64 0, i64 524287 store volatile i8 42, i8 *%ptr1 store volatile i8 42, i8 *%ptr2 ret void @@ -172,21 +176,21 @@ define void @f7() { ; Keep the object-relative offset the same but bump the size of the ; objects by one doubleword. define void @f8() { -; CHECK-NOFP: f8: +; CHECK-NOFP-LABEL: f8: ; CHECK-NOFP: llilh %r1, 24 ; CHECK-NOFP: agr %r1, %r15 ; CHECK-NOFP: mvi 7(%r1), 42 ; CHECK-NOFP: br %r14 ; -; CHECK-FP: f8: +; CHECK-FP-LABEL: f8: ; CHECK-FP: llilh %r1, 24 ; CHECK-FP: agr %r1, %r11 ; CHECK-FP: mvi 7(%r1), 42 ; CHECK-FP: br %r14 - %region1 = alloca [1048416 x i8], align 8 - %region2 = alloca [1048416 x i8], align 8 - %ptr1 = getelementptr inbounds [1048416 x i8]* %region1, i64 0, i64 524287 - %ptr2 = getelementptr inbounds [1048416 x i8]* %region2, i64 0, i64 524287 + %region1 = alloca [1048408 x i8], align 8 + %region2 = alloca [1048408 x i8], align 8 + %ptr1 = getelementptr inbounds [1048408 x i8]* %region1, i64 0, i64 524287 + %ptr2 = getelementptr inbounds [1048408 x i8]* %region2, i64 0, i64 524287 store volatile i8 42, i8 *%ptr1 store volatile i8 42, i8 *%ptr2 ret void @@ -200,56 +204,56 @@ define void @f8() { ; The LA then gets lowered into the LLILH/LA form. The exact sequence ; isn't that important though. define void @f9() { -; CHECK-NOFP: f9: +; CHECK-NOFP-LABEL: f9: ; CHECK-NOFP: llilh [[R1:%r[1-5]]], 16 ; CHECK-NOFP: la [[R2:%r[1-5]]], 8([[R1]],%r15) ; CHECK-NOFP: agfi [[R2]], 524288 ; CHECK-NOFP: mvi 0([[R2]]), 42 ; CHECK-NOFP: br %r14 ; -; CHECK-FP: f9: +; CHECK-FP-LABEL: f9: ; CHECK-FP: llilh [[R1:%r[1-5]]], 16 ; CHECK-FP: la [[R2:%r[1-5]]], 8([[R1]],%r11) ; CHECK-FP: agfi [[R2]], 524288 ; CHECK-FP: mvi 0([[R2]]), 42 ; CHECK-FP: br %r14 - %region1 = alloca [1048416 x i8], align 8 - %region2 = alloca [1048416 x i8], align 8 - %ptr1 = getelementptr inbounds [1048416 x i8]* %region1, i64 0, i64 524288 - %ptr2 = getelementptr inbounds [1048416 x i8]* %region2, i64 0, i64 524288 + %region1 = alloca [1048408 x i8], align 8 + %region2 = alloca [1048408 x i8], align 8 + %ptr1 = getelementptr inbounds [1048408 x i8]* %region1, i64 0, i64 524288 + %ptr2 = getelementptr inbounds [1048408 x i8]* %region2, i64 0, i64 524288 store volatile i8 42, i8 *%ptr1 store volatile i8 42, i8 *%ptr2 ret void } -; Repeat f4 in a case that needs the emergency spill slot (because all +; Repeat f4 in a case that needs the emergency spill slots (because all ; call-clobbered registers are live and no call-saved ones have been ; allocated). define void @f10(i32 *%vptr) { -; CHECK-NOFP: f10: -; CHECK-NOFP: stg [[REGISTER:%r[1-9][0-4]?]], 160(%r15) +; CHECK-NOFP-LABEL: f10: +; CHECK-NOFP: stg [[REGISTER:%r[1-9][0-4]?]], [[OFFSET:160|168]](%r15) ; CHECK-NOFP: llilh [[REGISTER]], 8 ; CHECK-NOFP: agr [[REGISTER]], %r15 ; CHECK-NOFP: mvi 0([[REGISTER]]), 42 -; CHECK-NOFP: lg [[REGISTER]], 160(%r15) +; CHECK-NOFP: lg [[REGISTER]], [[OFFSET]](%r15) ; CHECK-NOFP: br %r14 ; -; CHECK-FP: f10: -; CHECK-FP: stg [[REGISTER:%r[1-9][0-4]?]], 160(%r11) +; CHECK-FP-LABEL: f10: +; CHECK-FP: stg [[REGISTER:%r[1-9][0-4]?]], [[OFFSET:160|168]](%r11) ; CHECK-FP: llilh [[REGISTER]], 8 ; CHECK-FP: agr [[REGISTER]], %r11 ; CHECK-FP: mvi 0([[REGISTER]]), 42 -; CHECK-FP: lg [[REGISTER]], 160(%r11) +; CHECK-FP: lg [[REGISTER]], [[OFFSET]](%r11) ; CHECK-FP: br %r14 %i0 = load volatile i32 *%vptr %i1 = load volatile i32 *%vptr %i3 = load volatile i32 *%vptr %i4 = load volatile i32 *%vptr %i5 = load volatile i32 *%vptr - %region1 = alloca [524112 x i8], align 8 - %region2 = alloca [524112 x i8], align 8 - %ptr1 = getelementptr inbounds [524112 x i8]* %region1, i64 0, i64 8 - %ptr2 = getelementptr inbounds [524112 x i8]* %region2, i64 0, i64 8 + %region1 = alloca [524104 x i8], align 8 + %region2 = alloca [524104 x i8], align 8 + %ptr1 = getelementptr inbounds [524104 x i8]* %region1, i64 0, i64 8 + %ptr2 = getelementptr inbounds [524104 x i8]* %region2, i64 0, i64 8 store volatile i8 42, i8 *%ptr1 store volatile i8 42, i8 *%ptr2 store volatile i32 %i0, i32 *%vptr @@ -260,28 +264,28 @@ define void @f10(i32 *%vptr) { ret void } -; And again with maximum register pressure. The only spill slot that the -; NOFP case needs is the emergency one, so the offsets are the same as for f4. +; And again with maximum register pressure. The only spill slots that the +; NOFP case needs are the emergency ones, so the offsets are the same as for f4. ; However, the FP case uses %r11 as the frame pointer and must therefore ; spill a second register. This leads to an extra displacement of 8. define void @f11(i32 *%vptr) { -; CHECK-NOFP: f11: +; CHECK-NOFP-LABEL: f11: ; CHECK-NOFP: stmg %r6, %r15, -; CHECK-NOFP: stg [[REGISTER:%r[1-9][0-4]?]], 160(%r15) +; CHECK-NOFP: stg [[REGISTER:%r[1-9][0-4]?]], [[OFFSET:160|168]](%r15) ; CHECK-NOFP: llilh [[REGISTER]], 8 ; CHECK-NOFP: agr [[REGISTER]], %r15 ; CHECK-NOFP: mvi 0([[REGISTER]]), 42 -; CHECK-NOFP: lg [[REGISTER]], 160(%r15) +; CHECK-NOFP: lg [[REGISTER]], [[OFFSET]](%r15) ; CHECK-NOFP: lmg %r6, %r15, ; CHECK-NOFP: br %r14 ; -; CHECK-FP: f11: +; CHECK-FP-LABEL: f11: ; CHECK-FP: stmg %r6, %r15, -; CHECK-FP: stg [[REGISTER:%r[1-9][0-4]?]], 160(%r11) +; CHECK-FP: stg [[REGISTER:%r[1-9][0-4]?]], [[OFFSET:160|168]](%r11) ; CHECK-FP: llilh [[REGISTER]], 8 ; CHECK-FP: agr [[REGISTER]], %r11 ; CHECK-FP: mvi 8([[REGISTER]]), 42 -; CHECK-FP: lg [[REGISTER]], 160(%r11) +; CHECK-FP: lg [[REGISTER]], [[OFFSET]](%r11) ; CHECK-FP: lmg %r6, %r15, ; CHECK-FP: br %r14 %i0 = load volatile i32 *%vptr @@ -298,10 +302,10 @@ define void @f11(i32 *%vptr) { %i12 = load volatile i32 *%vptr %i13 = load volatile i32 *%vptr %i14 = load volatile i32 *%vptr - %region1 = alloca [524112 x i8], align 8 - %region2 = alloca [524112 x i8], align 8 - %ptr1 = getelementptr inbounds [524112 x i8]* %region1, i64 0, i64 8 - %ptr2 = getelementptr inbounds [524112 x i8]* %region2, i64 0, i64 8 + %region1 = alloca [524104 x i8], align 8 + %region2 = alloca [524104 x i8], align 8 + %ptr1 = getelementptr inbounds [524104 x i8]* %region1, i64 0, i64 8 + %ptr2 = getelementptr inbounds [524104 x i8]* %region2, i64 0, i64 8 store volatile i8 42, i8 *%ptr1 store volatile i8 42, i8 *%ptr2 store volatile i32 %i0, i32 *%vptr |