From a8eb65a909ee7858bbc596447e26d3a7c8c3adb7 Mon Sep 17 00:00:00 2001
From: gretay-amd <gretay.amd@gmail.com>
Date: Thu, 30 Apr 2026 11:24:49 +0100
Subject: [PATCH] [AMDGPU][NFC] Use LaneMaskConstants for waterfall loops in
 AMDGPURegBankLegalizeHelper (#190792)

Use `LaneMaskConstants` for generating waterfall loops in
`AMDGPURegBankLegalizeHelper`.
No Functionality Change.
---
 .../AMDGPU/AMDGPURegBankLegalizeHelper.cpp    | 27 +++++++------------
 1 file changed, 9 insertions(+), 18 deletions(-)

diff --git a/llvm/lib/Target/AMDGPU/AMDGPURegBankLegalizeHelper.cpp b/llvm/lib/Target/AMDGPU/AMDGPURegBankLegalizeHelper.cpp
index 0782818b1c9d..090ffed1e1e5 100644
--- a/llvm/lib/Target/AMDGPU/AMDGPURegBankLegalizeHelper.cpp
+++ b/llvm/lib/Target/AMDGPU/AMDGPURegBankLegalizeHelper.cpp
@@ -14,6 +14,7 @@
 #include "AMDGPURegBankLegalizeHelper.h"
 #include "AMDGPUGlobalISelUtils.h"
 #include "AMDGPUInstrInfo.h"
+#include "AMDGPULaneMaskUtils.h"
 #include "AMDGPURegBankLegalizeRules.h"
 #include "AMDGPURegisterBankInfo.h"
 #include "GCNSubtarget.h"
@@ -95,20 +96,7 @@ bool RegBankLegalizeHelper::executeInWaterfallLoop(MachineIRBuilder &B,
 
   const SIRegisterInfo *TRI = ST.getRegisterInfo();
   const TargetRegisterClass *WaveRC = TRI->getWaveMaskRegClass();
-  unsigned MovExecOpc, MovExecTermOpc, XorTermOpc, AndSaveExecOpc, ExecReg;
-  if (IsWave32) {
-    MovExecOpc = AMDGPU::S_MOV_B32;
-    MovExecTermOpc = AMDGPU::S_MOV_B32_term;
-    XorTermOpc = AMDGPU::S_XOR_B32_term;
-    AndSaveExecOpc = AMDGPU::S_AND_SAVEEXEC_B32;
-    ExecReg = AMDGPU::EXEC_LO;
-  } else {
-    MovExecOpc = AMDGPU::S_MOV_B64;
-    MovExecTermOpc = AMDGPU::S_MOV_B64_term;
-    XorTermOpc = AMDGPU::S_XOR_B64_term;
-    AndSaveExecOpc = AMDGPU::S_AND_SAVEEXEC_B64;
-    ExecReg = AMDGPU::EXEC;
-  }
+  const AMDGPU::LaneMaskConstants &LMC = AMDGPU::LaneMaskConstants::get(ST);
 
 #ifndef NDEBUG
   const int OrigRangeSize = std::distance(BeginIt, EndIt);
@@ -270,7 +258,7 @@ bool RegBankLegalizeHelper::executeInWaterfallLoop(MachineIRBuilder &B,
   B.buildIntrinsic(Intrinsic::amdgcn_ballot, CondRegLM).addReg(CondReg);
 
   // Update EXEC, save the original EXEC value to SavedExec.
-  B.buildInstr(AndSaveExecOpc)
+  B.buildInstr(LMC.AndSaveExecOpc)
       .addDef(SavedExec)
       .addReg(CondRegLM, RegState::Kill);
   MRI.setSimpleHint(SavedExec, CondRegLM);
@@ -278,7 +266,10 @@ bool RegBankLegalizeHelper::executeInWaterfallLoop(MachineIRBuilder &B,
   B.setInsertPt(*BodyBB, BodyBB->end());
 
   // Update EXEC, switch all done bits to 0 and all todo bits to 1.
-  B.buildInstr(XorTermOpc).addDef(ExecReg).addReg(ExecReg).addReg(SavedExec);
+  B.buildInstr(LMC.XorTermOpc)
+      .addDef(LMC.ExecReg)
+      .addReg(LMC.ExecReg)
+      .addReg(SavedExec);
 
   // XXX - s_xor_b64 sets scc to 1 if the result is nonzero, so can we use
   // s_cbranch_scc0?
@@ -288,11 +279,11 @@ bool RegBankLegalizeHelper::executeInWaterfallLoop(MachineIRBuilder &B,
 
   // Save the EXEC mask before the loop.
   B.setInsertPt(MBB, MBB.end());
-  B.buildInstr(MovExecOpc).addDef(SaveExecReg).addReg(ExecReg);
+  B.buildInstr(LMC.MovOpc).addDef(SaveExecReg).addReg(LMC.ExecReg);
 
   // Restore the EXEC mask after the loop.
   B.setInsertPt(*RestoreExecBB, RestoreExecBB->begin());
-  B.buildInstr(MovExecTermOpc).addDef(ExecReg).addReg(SaveExecReg);
+  B.buildInstr(LMC.MovTermOpc).addDef(LMC.ExecReg).addReg(SaveExecReg);
 
   // Set the insert point after the original instruction, so any new
   // instructions will be in the remainder.