[llvm] r365846 - [AMDGPU] Skip calculating callee saved registers for entry function.

[Michael Liao via llvm-commits]

Author: hliao
Date: Thu Jul 11 16:53:30 2019
New Revision: 365846

URL: http://llvm.org/viewvc/llvm-project?rev=365846&view=rev
Log:
[AMDGPU] Skip calculating callee saved registers for entry function.

Reviewers: arsenm

Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, dstuttard, tpr, t-tye, hiraditya, llvm-commits

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D64596

Added:
    llvm/trunk/test/CodeGen/AMDGPU/frame-lowering-entry-all-sgpr-used.mir
Modified:
    llvm/trunk/lib/Target/AMDGPU/SIFrameLowering.cpp

Modified: llvm/trunk/lib/Target/AMDGPU/SIFrameLowering.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/AMDGPU/SIFrameLowering.cpp?rev=365846&r1=365845&r2=365846&view=diff
==============================================================================
--- llvm/trunk/lib/Target/AMDGPU/SIFrameLowering.cpp (original)
+++ llvm/trunk/lib/Target/AMDGPU/SIFrameLowering.cpp Thu Jul 11 16:53:30 2019
@@ -975,8 +975,10 @@ void SIFrameLowering::determineCalleeSav
                                            BitVector &SavedVGPRs,
                                            RegScavenger *RS) const {
   TargetFrameLowering::determineCalleeSaves(MF, SavedVGPRs, RS);
-
   SIMachineFunctionInfo *MFI = MF.getInfo<SIMachineFunctionInfo>();
+  if (MFI->isEntryFunction())
+    return;
+
   const MachineFrameInfo &FrameInfo = MF.getFrameInfo();
   const GCNSubtarget &ST = MF.getSubtarget<GCNSubtarget>();
   const SIRegisterInfo *TRI = ST.getRegisterInfo();
@@ -1049,6 +1051,8 @@ void SIFrameLowering::determineCalleeSav
                                                RegScavenger *RS) const {
   TargetFrameLowering::determineCalleeSaves(MF, SavedRegs, RS);
   const SIMachineFunctionInfo *MFI = MF.getInfo<SIMachineFunctionInfo>();
+  if (MFI->isEntryFunction())
+    return;
 
   const GCNSubtarget &ST = MF.getSubtarget<GCNSubtarget>();
   const SIRegisterInfo *TRI = ST.getRegisterInfo();

Added: llvm/trunk/test/CodeGen/AMDGPU/frame-lowering-entry-all-sgpr-used.mir
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/AMDGPU/frame-lowering-entry-all-sgpr-used.mir?rev=365846&view=auto
==============================================================================
--- llvm/trunk/test/CodeGen/AMDGPU/frame-lowering-entry-all-sgpr-used.mir (added)
+++ llvm/trunk/test/CodeGen/AMDGPU/frame-lowering-entry-all-sgpr-used.mir Thu Jul 11 16:53:30 2019
@@ -0,0 +1,54 @@
+# RUN: llc -mtriple=amdgcn-amd-amdhsa -verify-machineinstrs -run-pass=prologepilog %s -o - | FileCheck %s
+
+# CHECK-LABEL: all_sgpr_used
+# CHECK: V_CMP_LT_U32_e64
+--- |
+  define amdgpu_kernel void @all_sgpr_used() #0 {
+    ret void
+  }
+  attributes #0 = { "amdgpu-num-sgpr"="8" "frame-pointer"="all"}
+...
+---
+name:            all_sgpr_used
+tracksRegLiveness: true
+liveins:         
+  - { reg: '$vgpr0' }
+  - { reg: '$vgpr1' }
+  - { reg: '$vgpr2' }
+  - { reg: '$sgpr4_sgpr5' }
+  - { reg: '$sgpr6_sgpr7' }
+  - { reg: '$sgpr8' }
+  - { reg: '$sgpr9' }
+machineFunctionInfo: 
+  explicitKernArgSize: 84
+  maxKernArgAlign: 8
+  ldsSize:         20496
+  isEntryFunction: true
+  waveLimiter:     true
+  scratchRSrcReg:  '$sgpr96_sgpr97_sgpr98_sgpr99'
+  scratchWaveOffsetReg: '$sgpr101'
+  frameOffsetReg:  '$sgpr101'
+  stackPtrOffsetReg: '$sgpr32'
+  argumentInfo:    
+    privateSegmentBuffer: { reg: '$sgpr0_sgpr1_sgpr2_sgpr3' }
+    dispatchPtr:     { reg: '$sgpr4_sgpr5' }
+    kernargSegmentPtr: { reg: '$sgpr6_sgpr7' }
+    workGroupIDX:    { reg: '$sgpr8' }
+    workGroupIDY:    { reg: '$sgpr9' }
+    privateSegmentWaveByteOffset: { reg: '$sgpr10' }
+    workItemIDX:     { reg: '$vgpr0' }
+    workItemIDY:     { reg: '$vgpr1' }
+    workItemIDZ:     { reg: '$vgpr2' }
+body:             |
+  bb.0:
+    liveins: $sgpr8, $sgpr9, $vgpr0, $vgpr1, $vgpr2, $sgpr4_sgpr5, $sgpr6_sgpr7
+    $sgpr0 = S_MOV_B32 0
+    $sgpr1 = S_MOV_B32 0
+    $sgpr2 = S_MOV_B32 0
+    $sgpr3 = S_MOV_B32 0
+    $sgpr4 = S_MOV_B32 0
+    $sgpr5 = S_MOV_B32 0
+    $sgpr6 = S_MOV_B32 0
+    $sgpr7 = S_MOV_B32 0
+    $vcc = V_CMP_LT_U32_e64 $sgpr8, $vgpr1, implicit $exec
+...