[AARCH64] Teach AArch64FrameLowering::getFrameIndexReferencePreferSP really prefer SP.

[skatkov]

Summary

Make more efforts to use sp even if fp is present in method.

This is used by STATEPOINT instruction.

[skatkov]

I wonder whether there are some idea who else from aarch64 world can be added to review this change?
Any suggestions are very appreciated.

[skatkov]

llvm/lib/Target/AArch64/AArch64FrameLowering.cpp

3539

Probably I need to check hasStackRealignment as well...

[reames]

Spent some time educating myself on AArch64 frame lowering. I'm mostly relying on the diagram at the top of AArch64FrameLowering.cpp. Review guidance is about specializing the case where the variable sized regions are empty.

llvm/lib/Target/AArch64/AArch64FrameLowering.cpp

3539

You do.

I'd also strongly prefer you specialize this for the case where there is no SVE stack. (e.g !getSVEStackSize()) Support for SVE can be added at a later time if needed, and would need it's own test coverage. I have no clue how to even suggest writing such a test at this time.

Once you specialize, the code here gets a lot simpler.

3540–3547

You should probably use an early return with the inverted condition above. This would reduce nesting in the new code. Once you simplify based on no SVE, it may be more of a wash though...

[skatkov]

please take a look.

[reames]

LGTM

llvm/lib/Target/AArch64/AArch64FrameLowering.cpp

StackOffset AArch64FrameLowering::getFrameIndexReferencePreferSP(
     bool IgnoreSPUpdates) const {
   const MachineFrameInfo &MFI = MF.getFrameInfo();
   if (IgnoreSPUpdates) {
     LLVM_DEBUG(dbgs() << "Offset from the SP for " << FI << " is "
                       << MFI.getObjectOffset(FI) << "\n");
     FrameReg = AArch64::SP;
     return StackOffset::getFixed(MFI.getObjectOffset(FI));
   }
 
+  // Go to common code if we cannot provide sp + offset.
+  if (MFI.hasVarSizedObjects() ||
+      MF.getInfo<AArch64FunctionInfo>()->getStackSizeSVE() ||
+      MF.getSubtarget().getRegisterInfo()->hasStackRealignment(MF))
     return getFrameIndexReference(MF, FI, FrameReg);
+
+  FrameReg = AArch64::SP;
+  return getStackOffset(MF, MFI.getObjectOffset(FI));
 }
 
 /// The parent frame offset (aka dispFrame) is only used on X86_64 to retrieve
 /// the parent's frame pointer
 unsigned AArch64FrameLowering::getWinEHParentFrameOffset(
     const MachineFunction &MF) const {
   return 0;
 }

llvm/test/CodeGen/AArch64/statepoint-call-lowering-sp.ll

+; RUN: llc -mtriple aarch64-none-linux-gnu -verify-machineinstrs -stop-after=prologepilog < %s | FileCheck %s
+
+; Check that STATEPOINT instruction prefer to use sp in presense of fp.
+target datalayout = "e-i64:64-f80:128-n8:16:32:64-S128"
+target triple = "aarch64-unknown-linux-gnu"
+
+declare void @consume(i32 addrspace(1)* %obj)
+
+define i1 @test(i32 addrspace(1)* %a) "frame-pointer"="all" gc "statepoint-example" {
+entry:
+  %safepoint_token = tail call token (i64, i32, i1 ()*, i32, i32, ...) @llvm.experimental.gc.statepoint.p0f_i1f(i64 0, i32 0, i1 ()* @return_i1, i32 0, i32 0, i32 0, i32 0) ["gc-live" (i32 addrspace(1)* %a)]
+; CHECK: STATEPOINT 0, 0, 0, @return_i1, 2, 0, 2, 0, 2, 0, 2, 1, 1, 8, $sp, 24, 2, 0, 2, 1, 0, 0
+  %call1 = call i32 addrspace(1)* @llvm.experimental.gc.relocate.p1i32(token %safepoint_token,  i32 0, i32 0)
+  %call2 = call zeroext i1 @llvm.experimental.gc.result.i1(token %safepoint_token)
+  call void @consume(i32 addrspace(1)* %call1)
+  ret i1 %call2
+}
+
+
+declare i1 @return_i1()
+declare token @llvm.experimental.gc.statepoint.p0f_i1f(i64, i32, i1 ()*, i32, i32, ...)
+declare i32 addrspace(1)* @llvm.experimental.gc.relocate.p1i32(token, i32, i32)
+declare i1 @llvm.experimental.gc.result.i1(token)