[llvm-branch-commits] [llvm] [AMDGPU][NPM] Port SIMemoryLegalizer to NPM (PR #130060)

[Akshat Oke via llvm-branch-commits]

https://github.com/optimisan created https://github.com/llvm/llvm-project/pull/130060

None

>From 54641a84426002597f90f780e57575c4d31c6f58 Mon Sep 17 00:00:00 2001
From: Akshat Oke <Akshat.Oke at amd.com>
Date: Wed, 5 Mar 2025 11:06:40 +0000
Subject: [PATCH] [AMDGPU][NPM] Port SIMemoryLegalizer to NPM

---
 llvm/lib/Target/AMDGPU/AMDGPU.h               |  8 +++-
 llvm/lib/Target/AMDGPU/AMDGPUPassRegistry.def |  2 +-
 .../lib/Target/AMDGPU/AMDGPUTargetMachine.cpp |  5 ++-
 llvm/lib/Target/AMDGPU/SIMemoryLegalizer.cpp  | 43 ++++++++++++++-----
 4 files changed, 44 insertions(+), 14 deletions(-)

diff --git a/llvm/lib/Target/AMDGPU/AMDGPU.h b/llvm/lib/Target/AMDGPU/AMDGPU.h
index f208a8bb9964b..23b9aa0cf0523 100644
--- a/llvm/lib/Target/AMDGPU/AMDGPU.h
+++ b/llvm/lib/Target/AMDGPU/AMDGPU.h
@@ -363,6 +363,12 @@ class GCNCreateVOPDPass : public PassInfoMixin<GCNCreateVOPDPass> {
   PreservedAnalyses run(MachineFunction &MF, MachineFunctionAnalysisManager &AM);
 };
 
+class SIMemoryLegalizerPass : public PassInfoMixin<SIMemoryLegalizerPass> {
+public:
+  PreservedAnalyses run(MachineFunction &MF,
+                        MachineFunctionAnalysisManager &MFAM);
+};
+
 FunctionPass *createAMDGPUAnnotateUniformValuesLegacy();
 
 ModulePass *createAMDGPUPrintfRuntimeBinding();
@@ -427,7 +433,7 @@ class SIAnnotateControlFlowPass
 void initializeSIAnnotateControlFlowLegacyPass(PassRegistry &);
 extern char &SIAnnotateControlFlowLegacyPassID;
 
-void initializeSIMemoryLegalizerPass(PassRegistry&);
+void initializeSIMemoryLegalizerLegacyPass(PassRegistry &);
 extern char &SIMemoryLegalizerID;
 
 void initializeSIModeRegisterLegacyPass(PassRegistry &);
diff --git a/llvm/lib/Target/AMDGPU/AMDGPUPassRegistry.def b/llvm/lib/Target/AMDGPU/AMDGPUPassRegistry.def
index 0e3dcb4267ede..de959f8a2aa62 100644
--- a/llvm/lib/Target/AMDGPU/AMDGPUPassRegistry.def
+++ b/llvm/lib/Target/AMDGPU/AMDGPUPassRegistry.def
@@ -113,6 +113,7 @@ MACHINE_FUNCTION_PASS("si-load-store-opt", SILoadStoreOptimizerPass())
 MACHINE_FUNCTION_PASS("si-lower-control-flow", SILowerControlFlowPass())
 MACHINE_FUNCTION_PASS("si-lower-sgpr-spills", SILowerSGPRSpillsPass())
 MACHINE_FUNCTION_PASS("si-lower-wwm-copies", SILowerWWMCopiesPass())
+MACHINE_FUNCTION_PASS("si-memory-legalizer", SIMemoryLegalizerPass())
 MACHINE_FUNCTION_PASS("si-mode-register", SIModeRegisterPass())
 MACHINE_FUNCTION_PASS("si-opt-vgpr-liverange", SIOptimizeVGPRLiveRangePass())
 MACHINE_FUNCTION_PASS("si-optimize-exec-masking", SIOptimizeExecMaskingPass())
@@ -132,7 +133,6 @@ DUMMY_MACHINE_FUNCTION_PASS("amdgpu-set-wave-priority", AMDGPUSetWavePriorityPas
 DUMMY_MACHINE_FUNCTION_PASS("si-insert-hard-clauses", SIInsertHardClausesPass())
 DUMMY_MACHINE_FUNCTION_PASS("si-insert-waitcnts", SIInsertWaitcntsPass())
 DUMMY_MACHINE_FUNCTION_PASS("si-late-branch-lowering", SILateBranchLoweringPass())
-DUMMY_MACHINE_FUNCTION_PASS("si-memory-legalizer", SIMemoryLegalizerPass())
 DUMMY_MACHINE_FUNCTION_PASS("si-pre-emit-peephole", SIPreEmitPeepholePass())
 // TODO: Move amdgpu-preload-kern-arg-prolog to MACHINE_FUNCTION_PASS since it
 // already exists.
diff --git a/llvm/lib/Target/AMDGPU/AMDGPUTargetMachine.cpp b/llvm/lib/Target/AMDGPU/AMDGPUTargetMachine.cpp
index 73ae9135eb319..dbe212ad0a216 100644
--- a/llvm/lib/Target/AMDGPU/AMDGPUTargetMachine.cpp
+++ b/llvm/lib/Target/AMDGPU/AMDGPUTargetMachine.cpp
@@ -541,7 +541,7 @@ extern "C" LLVM_EXTERNAL_VISIBILITY void LLVMInitializeAMDGPUTarget() {
   initializeSILowerControlFlowLegacyPass(*PR);
   initializeSIPreEmitPeepholePass(*PR);
   initializeSILateBranchLoweringPass(*PR);
-  initializeSIMemoryLegalizerPass(*PR);
+  initializeSIMemoryLegalizerLegacyPass(*PR);
   initializeSIOptimizeExecMaskingLegacyPass(*PR);
   initializeSIPreAllocateWWMRegsLegacyPass(*PR);
   initializeSIFormMemoryClausesLegacyPass(*PR);
@@ -2151,7 +2151,8 @@ void AMDGPUCodeGenPassBuilder::addPreEmitPass(AddMachinePass &addPass) const {
   if (isPassEnabled(EnableVOPD, CodeGenOptLevel::Less)) {
     addPass(GCNCreateVOPDPass());
   }
-  // TODO: addPass(SIMemoryLegalizerPass());
+
+  addPass(SIMemoryLegalizerPass());
   // TODO: addPass(SIInsertWaitcntsPass());
 
   // TODO: addPass(SIModeRegisterPass());
diff --git a/llvm/lib/Target/AMDGPU/SIMemoryLegalizer.cpp b/llvm/lib/Target/AMDGPU/SIMemoryLegalizer.cpp
index 34953f9c08db7..1375ba201ec58 100644
--- a/llvm/lib/Target/AMDGPU/SIMemoryLegalizer.cpp
+++ b/llvm/lib/Target/AMDGPU/SIMemoryLegalizer.cpp
@@ -21,8 +21,10 @@
 #include "llvm/ADT/StringExtras.h"
 #include "llvm/CodeGen/MachineBasicBlock.h"
 #include "llvm/CodeGen/MachineFunctionPass.h"
+#include "llvm/CodeGen/MachinePassManager.h"
 #include "llvm/IR/DiagnosticInfo.h"
 #include "llvm/IR/MemoryModelRelaxationAnnotations.h"
+#include "llvm/IR/PassManager.h"
 #include "llvm/Support/AtomicOrdering.h"
 #include "llvm/TargetParser/TargetParser.h"
 
@@ -625,9 +627,9 @@ class SIGfx12CacheControl : public SIGfx11CacheControl {
   }
 };
 
-class SIMemoryLegalizer final : public MachineFunctionPass {
+class SIMemoryLegalizer final {
 private:
-
+  const MachineModuleInfo &MMI;
   /// Cache Control.
   std::unique_ptr<SICacheControl> CC = nullptr;
 
@@ -661,10 +663,16 @@ class SIMemoryLegalizer final : public MachineFunctionPass {
   bool expandAtomicCmpxchgOrRmw(const SIMemOpInfo &MOI,
                                 MachineBasicBlock::iterator &MI);
 
+public:
+  SIMemoryLegalizer(const MachineModuleInfo &MMI) : MMI(MMI) {};
+  bool run(MachineFunction &MF);
+};
+
+class SIMemoryLegalizerLegacy final : public MachineFunctionPass {
 public:
   static char ID;
 
-  SIMemoryLegalizer() : MachineFunctionPass(ID) {}
+  SIMemoryLegalizerLegacy() : MachineFunctionPass(ID) {}
 
   void getAnalysisUsage(AnalysisUsage &AU) const override {
     AU.setPreservesCFG();
@@ -2767,11 +2775,26 @@ bool SIMemoryLegalizer::expandAtomicCmpxchgOrRmw(const SIMemOpInfo &MOI,
   return Changed;
 }
 
-bool SIMemoryLegalizer::runOnMachineFunction(MachineFunction &MF) {
-  bool Changed = false;
-
+bool SIMemoryLegalizerLegacy::runOnMachineFunction(MachineFunction &MF) {
   const MachineModuleInfo &MMI =
       getAnalysis<MachineModuleInfoWrapperPass>().getMMI();
+  return SIMemoryLegalizer(MMI).run(MF);
+}
+
+PreservedAnalyses
+SIMemoryLegalizerPass::run(MachineFunction &MF,
+                           MachineFunctionAnalysisManager &MFAM) {
+  auto *MMI = MFAM.getResult<ModuleAnalysisManagerFunctionProxy>(MF)
+                  .getCachedResult<MachineModuleAnalysis>(
+                      *MF.getFunction().getParent());
+  assert(MMI && "MachineModuleAnalysis must be available");
+  if (!SIMemoryLegalizer(MMI->getMMI()).run(MF))
+    return PreservedAnalyses::all();
+  return getMachineFunctionPassPreservedAnalyses().preserveSet<CFGAnalyses>();
+}
+
+bool SIMemoryLegalizer::run(MachineFunction &MF) {
+  bool Changed = false;
 
   SIMemOpAccess MOA(MMI.getObjFileInfo<AMDGPUMachineModuleInfo>());
   CC = SICacheControl::create(MF.getSubtarget<GCNSubtarget>());
@@ -2812,11 +2835,11 @@ bool SIMemoryLegalizer::runOnMachineFunction(MachineFunction &MF) {
   return Changed;
 }
 
-INITIALIZE_PASS(SIMemoryLegalizer, DEBUG_TYPE, PASS_NAME, false, false)
+INITIALIZE_PASS(SIMemoryLegalizerLegacy, DEBUG_TYPE, PASS_NAME, false, false)
 
-char SIMemoryLegalizer::ID = 0;
-char &llvm::SIMemoryLegalizerID = SIMemoryLegalizer::ID;
+char SIMemoryLegalizerLegacy::ID = 0;
+char &llvm::SIMemoryLegalizerID = SIMemoryLegalizerLegacy::ID;
 
 FunctionPass *llvm::createSIMemoryLegalizerPass() {
-  return new SIMemoryLegalizer();
+  return new SIMemoryLegalizerLegacy();
 }