[AArch64] Fix Local Deallocation for Homogeneous Prolog/Epilog

[kyulee]

Summary

The stack adjustment for local deallocation was incorrectly ported.

[MaskRay]

llvm/test/CodeGen/AArch64/arm64-homogeneous-prolog-epilog-local.ll

1 ↗

(On Diff #361482)

arm64-* tests are historical. New tests shouldn't need the prefix since all tests in this directory are related to aarch64.

20 ↗

(On Diff #361482)

remove ssp "frame-pointer"="non-leaf" which are irrelevant.

[kyulee]

Update the change per feedbacks.

[MaskRay]

LG

llvm/test/CodeGen/AArch64/homogeneous-prolog-epilog-local.ll

1 ↗

(On Diff #361483)

hmm, i don't know whether it is necessary to explicitly test two configurations. having one seems sufficient.

5 ↗

(On Diff #361483)

It's useful to check sub as well because the intention is to check the sub+add pair.

[kyulee]

Update the change per feedbacks.

• Add sub/add pair check.
• Check against the one target.

[MaskRay]

llvm/test/CodeGen/AArch64/homogeneous-prolog-epilog-local.ll

8 ↗

(On Diff #361488)

Now I believe this test should just be added into arm64-homogeneous-prolog-epilog.ll

This reduces the number of RUN lines which can improve performance. Every llc process has some small initial setup overhead.

[kyulee]

Update the change

• Combine the test into the existing one.

llvm/lib/Target/AArch64/AArch64FrameLowering.cpp

if (homogeneousPrologEpilog(MF, &MBB)) {
     if (LastPopI != MBB.begin()) {
       auto HomogeneousEpilog = std::prev(LastPopI);
       if (HomogeneousEpilog->getOpcode() == AArch64::HOM_Epilog)
         LastPopI = HomogeneousEpilog;
     }
 
     // Adjust local stack
     emitFrameOffset(MBB, LastPopI, DL, AArch64::SP, AArch64::SP,
-                    StackOffset::getFixed(-AFI->getLocalStackSize()), TII,
+                    StackOffset::getFixed(AFI->getLocalStackSize()), TII,
                     MachineInstr::FrameDestroy, false, NeedsWinCFI);
 
     // SP has been already adjusted while restoring callee save regs.
     // We've bailed-out the case with adjusting SP for arguments.
     assert(AfterCSRPopSize == 0);
     return;
   }
   bool CombineSPBump = shouldCombineCSRLocalStackBumpInEpilogue(MBB, NumBytes);

llvm/test/CodeGen/AArch64/arm64-homogeneous-prolog-epilog.ll

define i32 @_Z3hooii(i32 %b, i32 %a) nounwind ssp minsize {
   %4 = add i32 %3, %1
   %5 = add i32 %4, %2
   %6 = tail call i32 @_Z3gooi(i32 %5)
   ret i32 %6
 }
 
 declare i32 @_Z3gooi(i32);
 
+; CHECK-LABEL: _foo:
+; CHECK:        sub sp, sp, #16
+; CHECK:        bl  _goo
+; CHECK:        add sp, sp, #16
+
+define i32 @foo(i32 %c) nounwind minsize {
+entry:
+  %buffer = alloca [1 x i32], align 4
+  %0 = bitcast [1 x i32]* %buffer to i8*
+  call void @llvm.lifetime.start.p0i8(i64 4, i8* nonnull %0)
+  %arraydecay = getelementptr inbounds [1 x i32], [1 x i32]* %buffer, i64 0, i64 0
+  %call = call i32 @goo(i32* nonnull %arraydecay)
+  %sub = sub nsw i32 %c, %call
+  call void @llvm.lifetime.end.p0i8(i64 4, i8* nonnull %0)
+
+  ret i32 %sub
+}
+
+declare void @llvm.lifetime.start.p0i8(i64 immarg, i8* nocapture)
+declare i32 @goo(i32*)
+declare void @llvm.lifetime.end.p0i8(i64 immarg, i8* nocapture)
 
 ; CHECK-LABEL: _OUTLINED_FUNCTION_PROLOG_x30x29x19x20x21x22:
 ; CHECK:      stp     x22, x21, [sp, #-32]!
 ; CHECK-NEXT: stp     x20, x19, [sp, #16]
 ; CHECK-NEXT: ret
 
 ; CHECK-LABEL: _OUTLINED_FUNCTION_EPILOG_x30x29x19x20x21x22:
 ; CHECK:      mov     x16, x30