AMDGPU: Fix trying to constrain physical registers in spill handling

[arsenm]

It's nonsensical to call constrainRegClass on a physical register,
and we should not see virtual registers here.

[arsenm]

• CodeGen: Stop checking for physregs in constrainRegClass #161795
• AMDGPU: Fix constrain register logic for physregs #161794
• AMDGPU: Fix trying to constrain physical registers in spill handling #161793 👈 (View in Graphite)
• AMDGPU: Stop trying to constrain register class of post-RA-pseudos #161792
• AMDGPU: Remove dead code trying to constrain a physical register #161790
• main

This stack of pull requests is managed by Graphite. Learn more about stacking.

[llvmbot]

@llvm/pr-subscribers-backend-amdgpu

Author: Matt Arsenault (arsenm)

Changes

It's nonsensical to call constrainRegClass on a physical register,
and we should not see virtual registers here.

---

Full diff: https://github.com/llvm/llvm-project/pull/161793.diff

2 Files Affected:

• (modified) llvm/lib/Target/AMDGPU/SIFrameLowering.cpp (+2-1)
• (modified) llvm/lib/Target/AMDGPU/SIRegisterInfo.cpp (+3-7)

diff --git a/llvm/lib/Target/AMDGPU/SIFrameLowering.cpp b/llvm/lib/Target/AMDGPU/SIFrameLowering.cpp
index e4b3528b432bb..0189e7b90ca94 100644
--- a/llvm/lib/Target/AMDGPU/SIFrameLowering.cpp
+++ b/llvm/lib/Target/AMDGPU/SIFrameLowering.cpp
@@ -306,7 +306,8 @@ class PrologEpilogSGPRSpillBuilder {
 
       buildEpilogRestore(ST, TRI, *FuncInfo, LiveUnits, MF, MBB, MI, DL,
                          TmpVGPR, FI, FrameReg, DwordOff);
-      MRI.constrainRegClass(SubReg, &AMDGPU::SReg_32_XM0RegClass);
+      assert(SubReg.isPhysical());
+
       BuildMI(MBB, MI, DL, TII->get(AMDGPU::V_READFIRSTLANE_B32), SubReg)
           .addReg(TmpVGPR, RegState::Kill);
       DwordOff += 4;
diff --git a/llvm/lib/Target/AMDGPU/SIRegisterInfo.cpp b/llvm/lib/Target/AMDGPU/SIRegisterInfo.cpp
index 205237fefe785..3c2dd4252c583 100644
--- a/llvm/lib/Target/AMDGPU/SIRegisterInfo.cpp
+++ b/llvm/lib/Target/AMDGPU/SIRegisterInfo.cpp
@@ -2222,8 +2222,6 @@ bool SIRegisterInfo::spillEmergencySGPR(MachineBasicBlock::iterator MI,
     // Don't need to write VGPR out.
   }
 
-  MachineRegisterInfo &MRI = MI->getMF()->getRegInfo();
-
   // Restore clobbered registers in the specified restore block.
   MI = RestoreMBB.end();
   SB.setMI(&RestoreMBB, MI);
@@ -2238,7 +2236,8 @@ bool SIRegisterInfo::spillEmergencySGPR(MachineBasicBlock::iterator MI,
           SB.NumSubRegs == 1
               ? SB.SuperReg
               : Register(getSubReg(SB.SuperReg, SB.SplitParts[i]));
-      MRI.constrainRegClass(SubReg, &AMDGPU::SReg_32_XM0RegClass);
+
+      assert(SubReg.isPhysical());
       bool LastSubReg = (i + 1 == e);
       auto MIB = BuildMI(*SB.MBB, MI, SB.DL, SB.TII.get(AMDGPU::V_READLANE_B32),
                          SubReg)
@@ -3059,8 +3058,7 @@ bool SIRegisterInfo::eliminateFrameIndex(MachineBasicBlock::iterator MI,
         if (IsSALU && LiveSCC) {
           Register NewDest;
           if (IsCopy) {
-            MF->getRegInfo().constrainRegClass(ResultReg,
-                                               &AMDGPU::SReg_32_XM0RegClass);
+            assert(ResultReg.isPhysical());
             NewDest = ResultReg;
           } else {
             NewDest = RS->scavengeRegisterBackwards(AMDGPU::SReg_32_XM0RegClass,
@@ -3190,8 +3188,6 @@ bool SIRegisterInfo::eliminateFrameIndex(MachineBasicBlock::iterator MI,
 
             Register NewDest;
             if (IsCopy) {
-              MF->getRegInfo().constrainRegClass(ResultReg,
-                                                 &AMDGPU::SReg_32_XM0RegClass);
               NewDest = ResultReg;
             } else {
               NewDest = RS->scavengeRegisterBackwards(