[llvm] r364470 - [AMDGPU] Fix Livereg computation during epilogue insertion

[Matt Arsenault via llvm-commits]

Author: arsenm
Date: Wed Jun 26 13:35:18 2019
New Revision: 364470

URL: http://llvm.org/viewvc/llvm-project?rev=364470&view=rev
Log:
[AMDGPU] Fix Livereg computation during epilogue insertion

The LivePhysRegs calculated in order to find a scratch register in the
epilogue code wrongly uses 'LiveIns'. Instead, it should use the
'Liveout' sets.  For the liveness, also considering the operands of
the terminator (return) instruction which is the insertion point for
the scratch-exec-copy instruction.

Patch by Christudasan Devadasan

Modified:
    llvm/trunk/lib/Target/AMDGPU/SIFrameLowering.cpp
    llvm/trunk/test/CodeGen/AMDGPU/callee-special-input-sgprs.ll

Modified: llvm/trunk/lib/Target/AMDGPU/SIFrameLowering.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/AMDGPU/SIFrameLowering.cpp?rev=364470&r1=364469&r2=364470&view=diff
==============================================================================
--- llvm/trunk/lib/Target/AMDGPU/SIFrameLowering.cpp (original)
+++ llvm/trunk/lib/Target/AMDGPU/SIFrameLowering.cpp Wed Jun 26 13:35:18 2019
@@ -678,7 +678,8 @@ void SIFrameLowering::emitEpilogue(Machi
     if (ScratchExecCopy == AMDGPU::NoRegister) {
       // See emitPrologue
       LivePhysRegs LiveRegs(*ST.getRegisterInfo());
-      LiveRegs.addLiveIns(MBB);
+      LiveRegs.addLiveOuts(MBB);
+      LiveRegs.stepBackward(*MBBI);
 
       ScratchExecCopy
         = findScratchNonCalleeSaveRegister(MF, LiveRegs,

Modified: llvm/trunk/test/CodeGen/AMDGPU/callee-special-input-sgprs.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/AMDGPU/callee-special-input-sgprs.ll?rev=364470&r1=364469&r2=364470&view=diff
==============================================================================
--- llvm/trunk/test/CodeGen/AMDGPU/callee-special-input-sgprs.ll (original)
+++ llvm/trunk/test/CodeGen/AMDGPU/callee-special-input-sgprs.ll Wed Jun 26 13:35:18 2019
@@ -489,6 +489,7 @@ define amdgpu_kernel void @kern_indirect
 ; GCN-NOT: s[8:9]
 ; GCN-NOT: s[10:11]
 ; GCN-NOT: s[12:13]
+; GCN: s_or_saveexec_b64 s[6:7], -1
 define hidden void @func_indirect_use_every_sgpr_input() #1 {
   call void @use_every_sgpr_input()
   ret void