[PATCH] D17591: AMDGPU/SI: Enable frame index scavenging during PrologEpilogueInserter

[Tom Stellard via llvm-commits]

tstellarAMD created this revision.
tstellarAMD added a reviewer: arsenm.
tstellarAMD added a subscriber: llvm-commits.
Herald added a subscriber: arsenm.

This allows us to use virtual registers when we need extra registers
for inserting spill instructions in SIRegisterInfo:eliminateFrameIndex().

Once all the frame indices have been eliminated, the
PrologEpilogueInserter does an extra pass over the program to replace
all virtual registers with physical ones.

This allows us to make more efficient use of our emergency spill slots,
so we only need to create one.

http://reviews.llvm.org/D17591

Files:
  lib/Target/AMDGPU/SIRegisterInfo.cpp
  lib/Target/AMDGPU/SIRegisterInfo.h

Index: lib/Target/AMDGPU/SIRegisterInfo.h
===================================================================
--- lib/Target/AMDGPU/SIRegisterInfo.h
+++ lib/Target/AMDGPU/SIRegisterInfo.h
@@ -49,6 +49,8 @@
 
   bool requiresRegisterScavenging(const MachineFunction &Fn) const override;
 
+  bool requiresFrameIndexScavenging(const MachineFunction &MF) const override;
+
   void eliminateFrameIndex(MachineBasicBlock::iterator MI, int SPAdj,
                            unsigned FIOperandNum,
                            RegScavenger *RS) const override;
@@ -162,7 +164,7 @@
   void buildScratchLoadStore(MachineBasicBlock::iterator MI,
                              unsigned LoadStoreOp, unsigned Value,
                              unsigned ScratchRsrcReg, unsigned ScratchOffset,
-                             int64_t Offset, RegScavenger *RS) const;
+                             int64_t Offset) const;
 };
 
 } // End namespace llvm
Index: lib/Target/AMDGPU/SIRegisterInfo.cpp
===================================================================
--- lib/Target/AMDGPU/SIRegisterInfo.cpp
+++ lib/Target/AMDGPU/SIRegisterInfo.cpp
@@ -182,6 +182,11 @@
   return Fn.getFrameInfo()->hasStackObjects();
 }
 
+bool
+SIRegisterInfo::requiresFrameIndexScavenging(const MachineFunction &MF) const {
+  return MF.getFrameInfo()->hasStackObjects();
+}
+
 static unsigned getNumSubRegsForSpillOp(unsigned Op) {
 
   switch (Op) {
@@ -222,11 +227,11 @@
                                            unsigned Value,
                                            unsigned ScratchRsrcReg,
                                            unsigned ScratchOffset,
-                                           int64_t Offset,
-                                           RegScavenger *RS) const {
+                                           int64_t Offset) const {
 
   MachineBasicBlock *MBB = MI->getParent();
-  const MachineFunction *MF = MI->getParent()->getParent();
+  MachineFunction *MF = MI->getParent()->getParent();
+  MachineRegisterInfo &MRI = MF->getRegInfo();
   const SIInstrInfo *TII =
       static_cast<const SIInstrInfo *>(MF->getSubtarget().getInstrInfo());
   LLVMContext &Ctx = MF->getFunction()->getContext();
@@ -241,7 +246,7 @@
   unsigned Size = NumSubRegs * 4;
 
   if (!isUInt<12>(Offset + Size)) {
-    SOffset = RS->scavengeRegister(&AMDGPU::SGPR_32RegClass, MI, 0);
+    SOffset = MRI.createVirtualRegister(&AMDGPU::SGPR_32RegClass);
     if (SOffset == AMDGPU::NoRegister) {
       RanOutOfSGPRs = true;
       SOffset = AMDGPU::SGPR0;
@@ -283,6 +288,7 @@
                                         int SPAdj, unsigned FIOperandNum,
                                         RegScavenger *RS) const {
   MachineFunction *MF = MI->getParent()->getParent();
+  MachineRegisterInfo &MRI = MF->getRegInfo();
   MachineBasicBlock *MBB = MI->getParent();
   SIMachineFunctionInfo *MFI = MF->getInfo<SIMachineFunctionInfo>();
   MachineFrameInfo *FrameInfo = MF->getFrameInfo();
@@ -375,7 +381,7 @@
             TII->getNamedOperand(*MI, AMDGPU::OpName::src)->getReg(),
             TII->getNamedOperand(*MI, AMDGPU::OpName::scratch_rsrc)->getReg(),
             TII->getNamedOperand(*MI, AMDGPU::OpName::scratch_offset)->getReg(),
-             FrameInfo->getObjectOffset(Index), RS);
+             FrameInfo->getObjectOffset(Index));
       MI->eraseFromParent();
       break;
     case AMDGPU::SI_SPILL_V32_RESTORE:
@@ -388,16 +394,16 @@
             TII->getNamedOperand(*MI, AMDGPU::OpName::dst)->getReg(),
             TII->getNamedOperand(*MI, AMDGPU::OpName::scratch_rsrc)->getReg(),
             TII->getNamedOperand(*MI, AMDGPU::OpName::scratch_offset)->getReg(),
-            FrameInfo->getObjectOffset(Index), RS);
+            FrameInfo->getObjectOffset(Index));
       MI->eraseFromParent();
       break;
     }
 
     default: {
       int64_t Offset = FrameInfo->getObjectOffset(Index);
       FIOp.ChangeToImmediate(Offset);
       if (!TII->isImmOperandLegal(MI, FIOperandNum, FIOp)) {
-        unsigned TmpReg = RS->scavengeRegister(&AMDGPU::VGPR_32RegClass, MI, SPAdj);
+        unsigned TmpReg = MRI.createVirtualRegister(&AMDGPU::VGPR_32RegClass);
         BuildMI(*MBB, MI, MI->getDebugLoc(),
                 TII->get(AMDGPU::V_MOV_B32_e32), TmpReg)
                 .addImm(Offset);


-------------- next part --------------
A non-text attachment was scrubbed...
Name: D17591.49002.patch
Type: text/x-patch
Size: 4307 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20160225/f788e5d1/attachment-0001.bin>