[AMDGPU][NewPM] Port AMDGPUOpenCLEnqueuedBlockLowering to NPM

llvm/lib/Target/AMDGPU/AMDGPU.h

@@ -440,9 +440,9 @@ void initializeAMDGPUExternalAAWrapperPass(PassRegistry&);
 
 void initializeAMDGPUArgumentUsageInfoPass(PassRegistry &);
 
-ModulePass *createAMDGPUOpenCLEnqueuedBlockLoweringPass();
-void initializeAMDGPUOpenCLEnqueuedBlockLoweringPass(PassRegistry &);
-extern char &AMDGPUOpenCLEnqueuedBlockLoweringID;
+ModulePass *createAMDGPUOpenCLEnqueuedBlockLoweringLegacyPass();
+void initializeAMDGPUOpenCLEnqueuedBlockLoweringLegacyPass(PassRegistry &);
+extern char &AMDGPUOpenCLEnqueuedBlockLoweringLegacyID;
 
 void initializeGCNNSAReassignPass(PassRegistry &);
 extern char &GCNNSAReassignID;

llvm/lib/Target/AMDGPU/AMDGPUOpenCLEnqueuedBlockLowering.cpp

@@ -31,6 +31,7 @@
 //
 //===----------------------------------------------------------------------===//
 
+#include "AMDGPUOpenCLEnqueuedBlockLowering.h"
 #include "AMDGPU.h"
 #include "llvm/ADT/DenseSet.h"
 #include "llvm/ADT/SmallString.h"
@@ -48,31 +49,49 @@ using namespace llvm;
 namespace {
 
 /// Lower enqueued blocks.
-class AMDGPUOpenCLEnqueuedBlockLowering : public ModulePass {
+class AMDGPUOpenCLEnqueuedBlockLowering {
+public:
+  bool run(Module &M);
+};
+
+class AMDGPUOpenCLEnqueuedBlockLoweringLegacy : public ModulePass {
 public:
   static char ID;
 
-  explicit AMDGPUOpenCLEnqueuedBlockLowering() : ModulePass(ID) {}
+  explicit AMDGPUOpenCLEnqueuedBlockLoweringLegacy() : ModulePass(ID) {}
 
 private:
   bool runOnModule(Module &M) override;
 };
 
 } // end anonymous namespace
 
-char AMDGPUOpenCLEnqueuedBlockLowering::ID = 0;
+char AMDGPUOpenCLEnqueuedBlockLoweringLegacy::ID = 0;
 
-char &llvm::AMDGPUOpenCLEnqueuedBlockLoweringID =
-    AMDGPUOpenCLEnqueuedBlockLowering::ID;
+char &llvm::AMDGPUOpenCLEnqueuedBlockLoweringLegacyID =
+    AMDGPUOpenCLEnqueuedBlockLoweringLegacy::ID;
 
-INITIALIZE_PASS(AMDGPUOpenCLEnqueuedBlockLowering, DEBUG_TYPE,
+INITIALIZE_PASS(AMDGPUOpenCLEnqueuedBlockLoweringLegacy, DEBUG_TYPE,
                 "Lower OpenCL enqueued blocks", false, false)
 
-ModulePass* llvm::createAMDGPUOpenCLEnqueuedBlockLoweringPass() {
-  return new AMDGPUOpenCLEnqueuedBlockLowering();
+ModulePass *llvm::createAMDGPUOpenCLEnqueuedBlockLoweringLegacyPass() {
+  return new AMDGPUOpenCLEnqueuedBlockLoweringLegacy();
+}
+
+bool AMDGPUOpenCLEnqueuedBlockLoweringLegacy::runOnModule(Module &M) {
+  AMDGPUOpenCLEnqueuedBlockLowering Impl;
+  return Impl.run(M);
+}
+
+PreservedAnalyses
+AMDGPUOpenCLEnqueuedBlockLoweringPass::run(Module &M, ModuleAnalysisManager &) {
+  AMDGPUOpenCLEnqueuedBlockLowering Impl;
+  if (Impl.run(M))
+    return PreservedAnalyses::none();
+  return PreservedAnalyses::all();
 }
 
-bool AMDGPUOpenCLEnqueuedBlockLowering::runOnModule(Module &M) {
+bool AMDGPUOpenCLEnqueuedBlockLowering::run(Module &M) {
   DenseSet<Function *> Callers;
   auto &C = M.getContext();
   bool Changed = false;

llvm/lib/Target/AMDGPU/AMDGPUOpenCLEnqueuedBlockLowering.h

@@ -0,0 +1,23 @@
+//===- AMDGPUOpenCLEnqueuedBlockLowering.h -----------------------*- C++-*-===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef LLVM_LIB_TARGET_AMDGPU_OPENCLENQUEUEDBLOCKLOWERING_H
+#define LLVM_LIB_TARGET_AMDGPU_OPENCLENQUEUEDBLOCKLOWERING_H
+
+#include "llvm/IR/PassManager.h"
+
+namespace llvm {
+class AMDGPUOpenCLEnqueuedBlockLoweringPass
+    : public PassInfoMixin<AMDGPUOpenCLEnqueuedBlockLoweringPass> {
+public:
+  AMDGPUOpenCLEnqueuedBlockLoweringPass() = default;
+  PreservedAnalyses run(Module &M, ModuleAnalysisManager &MAM);
+};
+} // namespace llvm
+
+#endif // LLVM_LIB_TARGET_AMDGPU_OPENCLENQUEUEDBLOCKLOWERING_H

llvm/lib/Target/AMDGPU/AMDGPUPassRegistry.def

@@ -21,6 +21,7 @@ MODULE_PASS("amdgpu-lower-buffer-fat-pointers",
             AMDGPULowerBufferFatPointersPass(*this))
 MODULE_PASS("amdgpu-lower-ctor-dtor", AMDGPUCtorDtorLoweringPass())
 MODULE_PASS("amdgpu-sw-lower-lds", AMDGPUSwLowerLDSPass(*this))
+MODULE_PASS("amdgpu-lower-enqueued-block", AMDGPUOpenCLEnqueuedBlockLoweringPass())
 MODULE_PASS("amdgpu-lower-module-lds", AMDGPULowerModuleLDSPass(*this))
 MODULE_PASS("amdgpu-perf-hint",
             AMDGPUPerfHintAnalysisPass(

llvm/lib/Target/AMDGPU/AMDGPUTargetMachine.cpp

@@ -22,6 +22,7 @@
 #include "AMDGPUIGroupLP.h"
 #include "AMDGPUISelDAGToDAG.h"
 #include "AMDGPUMacroFusion.h"
+#include "AMDGPUOpenCLEnqueuedBlockLowering.h"
 #include "AMDGPUPerfHintAnalysis.h"
 #include "AMDGPUSplitModule.h"
 #include "AMDGPUTargetObjectFile.h"
@@ -499,7 +500,7 @@ extern "C" LLVM_EXTERNAL_VISIBILITY void LLVMInitializeAMDGPUTarget() {
   initializeAMDGPULowerKernelArgumentsPass(*PR);
   initializeAMDGPUPromoteKernelArgumentsPass(*PR);
   initializeAMDGPULowerKernelAttributesPass(*PR);
-  initializeAMDGPUOpenCLEnqueuedBlockLoweringPass(*PR);
+  initializeAMDGPUOpenCLEnqueuedBlockLoweringLegacyPass(*PR);
   initializeAMDGPUPostLegalizerCombinerPass(*PR);
   initializeAMDGPUPreLegalizerCombinerPass(*PR);
   initializeAMDGPURegBankCombinerPass(*PR);
@@ -1172,7 +1173,7 @@ void AMDGPUPassConfig::addIRPasses() {
     addPass(createR600OpenCLImageTypeLoweringPass());
 
   // Replace OpenCL enqueued block function pointers with global variables.
-  addPass(createAMDGPUOpenCLEnqueuedBlockLoweringPass());
+  addPass(createAMDGPUOpenCLEnqueuedBlockLoweringLegacyPass());
 
   // Lower LDS accesses to global memory pass if address sanitizer is enabled.
   if (EnableSwLowerLDS)
@@ -1939,7 +1940,7 @@ void AMDGPUCodeGenPassBuilder::addIRPasses(AddIRPass &addPass) const {
   addPass(AMDGPUAlwaysInlinePass());
   addPass(AlwaysInlinerPass());
 
-  // TODO: Missing OpenCLEnqueuedBlockLowering
+  addPass(AMDGPUOpenCLEnqueuedBlockLoweringPass());
 
   // Runs before PromoteAlloca so the latter can account for function uses
   if (EnableLowerModuleLDS)

llvm/test/CodeGen/AMDGPU/enqueue-kernel.ll

@@ -1,5 +1,6 @@
 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --function-signature --check-attributes --check-globals --include-generated-funcs
 ; RUN: opt -data-layout=A5 -amdgpu-lower-enqueued-block -S < %s | FileCheck %s
+; RUN: opt -data-layout=A5 -mtriple=amdgcn -passes=amdgpu-lower-enqueued-block -S < %s | FileCheck %s
 
 %struct.ndrange_t = type { i32 }
 %opencl.queue_t = type opaque