Automerge: [AMDGPU][NewPM] Port SILowerControlFlow pass into NPM. (#123045)

Author: cdevadas

llvm/lib/Target/AMDGPU/AMDGPU.h

@@ -207,8 +207,8 @@ extern char &SILoadStoreOptimizerLegacyID;
 void initializeSIWholeQuadModePass(PassRegistry &);
 extern char &SIWholeQuadModeID;
 
-void initializeSILowerControlFlowPass(PassRegistry &);
-extern char &SILowerControlFlowID;
+void initializeSILowerControlFlowLegacyPass(PassRegistry &);
+extern char &SILowerControlFlowLegacyID;
 
 void initializeSIPreEmitPeepholePass(PassRegistry &);
 extern char &SIPreEmitPeepholeID;

llvm/lib/Target/AMDGPU/AMDGPUPassRegistry.def

@@ -102,6 +102,7 @@ MACHINE_FUNCTION_PASS("si-i1-copies", SILowerI1CopiesPass())
 MACHINE_FUNCTION_PASS("si-fold-operands", SIFoldOperandsPass());
 MACHINE_FUNCTION_PASS("gcn-dpp-combine", GCNDPPCombinePass())
 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-opt-vgpr-liverange", SIOptimizeVGPRLiveRangePass())
 MACHINE_FUNCTION_PASS("si-peephole-sdwa", SIPeepholeSDWAPass())

llvm/lib/Target/AMDGPU/AMDGPUTargetMachine.cpp

@@ -38,6 +38,7 @@
 #include "SIFixSGPRCopies.h"
 #include "SIFoldOperands.h"
 #include "SILoadStoreOptimizer.h"
+#include "SILowerControlFlow.h"
 #include "SILowerSGPRSpills.h"
 #include "SIMachineFunctionInfo.h"
 #include "SIMachineScheduler.h"
@@ -523,7 +524,7 @@ extern "C" LLVM_EXTERNAL_VISIBILITY void LLVMInitializeAMDGPUTarget() {
   initializeSIInsertWaitcntsPass(*PR);
   initializeSIModeRegisterPass(*PR);
   initializeSIWholeQuadModePass(*PR);
-  initializeSILowerControlFlowPass(*PR);
+  initializeSILowerControlFlowLegacyPass(*PR);
   initializeSIPreEmitPeepholePass(*PR);
   initializeSILateBranchLoweringPass(*PR);
   initializeSIMemoryLegalizerPass(*PR);
@@ -1459,7 +1460,7 @@ void GCNPassConfig::addFastRegAlloc() {
   // This must be run immediately after phi elimination and before
   // TwoAddressInstructions, otherwise the processing of the tied operand of
   // SI_ELSE will introduce a copy of the tied operand source after the else.
-  insertPass(&PHIEliminationID, &SILowerControlFlowID);
+  insertPass(&PHIEliminationID, &SILowerControlFlowLegacyID);
 
   insertPass(&TwoAddressInstructionPassID, &SIWholeQuadModeID);
 
@@ -1480,7 +1481,7 @@ void GCNPassConfig::addOptimizedRegAlloc() {
   // This must be run immediately after phi elimination and before
   // TwoAddressInstructions, otherwise the processing of the tied operand of
   // SI_ELSE will introduce a copy of the tied operand source after the else.
-  insertPass(&PHIEliminationID, &SILowerControlFlowID);
+  insertPass(&PHIEliminationID, &SILowerControlFlowLegacyID);
 
   if (EnableRewritePartialRegUses)
     insertPass(&RenameIndependentSubregsID, &GCNRewritePartialRegUsesID);

llvm/lib/Target/AMDGPU/SILowerControlFlow.cpp

@@ -48,6 +48,7 @@
 /// %exec = S_OR_B64 %exec, %sgpr0     // Re-enable saved exec mask bits
 //===----------------------------------------------------------------------===//
 
+#include "SILowerControlFlow.h"
 #include "AMDGPU.h"
 #include "GCNSubtarget.h"
 #include "MCTargetDesc/AMDGPUMCTargetDesc.h"
@@ -68,7 +69,7 @@ RemoveRedundantEndcf("amdgpu-remove-redundant-endcf",
 
 namespace {
 
-class SILowerControlFlow : public MachineFunctionPass {
+class SILowerControlFlow {
 private:
   const SIRegisterInfo *TRI = nullptr;
   const SIInstrInfo *TII = nullptr;
@@ -135,10 +136,18 @@ class SILowerControlFlow : public MachineFunctionPass {
   // Remove redundant SI_END_CF instructions.
   void optimizeEndCf();
 
+public:
+  SILowerControlFlow(LiveIntervals *LIS, LiveVariables *LV,
+                     MachineDominatorTree *MDT)
+      : LIS(LIS), LV(LV), MDT(MDT) {}
+  bool run(MachineFunction &MF);
+};
+
+class SILowerControlFlowLegacy : public MachineFunctionPass {
 public:
   static char ID;
 
-  SILowerControlFlow() : MachineFunctionPass(ID) {}
+  SILowerControlFlowLegacy() : MachineFunctionPass(ID) {}
 
   bool runOnMachineFunction(MachineFunction &MF) override;
 
@@ -159,10 +168,10 @@ class SILowerControlFlow : public MachineFunctionPass {
 
 } // end anonymous namespace
 
-char SILowerControlFlow::ID = 0;
+char SILowerControlFlowLegacy::ID = 0;
 
-INITIALIZE_PASS(SILowerControlFlow, DEBUG_TYPE,
-               "SI lower control flow", false, false)
+INITIALIZE_PASS(SILowerControlFlowLegacy, DEBUG_TYPE, "SI lower control flow",
+                false, false)
 
 static void setImpSCCDefDead(MachineInstr &MI, bool IsDead) {
   MachineOperand &ImpDefSCC = MI.getOperand(3);
@@ -171,7 +180,7 @@ static void setImpSCCDefDead(MachineInstr &MI, bool IsDead) {
   ImpDefSCC.setIsDead(IsDead);
 }
 
-char &llvm::SILowerControlFlowID = SILowerControlFlow::ID;
+char &llvm::SILowerControlFlowLegacyID = SILowerControlFlowLegacy::ID;
 
 bool SILowerControlFlow::hasKill(const MachineBasicBlock *Begin,
                                  const MachineBasicBlock *End) {
@@ -753,21 +762,13 @@ bool SILowerControlFlow::removeMBBifRedundant(MachineBasicBlock &MBB) {
   return true;
 }
 
-bool SILowerControlFlow::runOnMachineFunction(MachineFunction &MF) {
+bool SILowerControlFlow::run(MachineFunction &MF) {
   const GCNSubtarget &ST = MF.getSubtarget<GCNSubtarget>();
   TII = ST.getInstrInfo();
   TRI = &TII->getRegisterInfo();
   EnableOptimizeEndCf = RemoveRedundantEndcf &&
                         MF.getTarget().getOptLevel() > CodeGenOptLevel::None;
 
-  // This doesn't actually need LiveIntervals, but we can preserve them.
-  auto *LISWrapper = getAnalysisIfAvailable<LiveIntervalsWrapperPass>();
-  LIS = LISWrapper ? &LISWrapper->getLIS() : nullptr;
-  // This doesn't actually need LiveVariables, but we can preserve them.
-  auto *LVWrapper = getAnalysisIfAvailable<LiveVariablesWrapperPass>();
-  LV = LVWrapper ? &LVWrapper->getLV() : nullptr;
-  auto *MDTWrapper = getAnalysisIfAvailable<MachineDominatorTreeWrapperPass>();
-  MDT = MDTWrapper ? &MDTWrapper->getDomTree() : nullptr;
   MRI = &MF.getRegInfo();
   BoolRC = TRI->getBoolRC();
 
@@ -864,3 +865,35 @@ bool SILowerControlFlow::runOnMachineFunction(MachineFunction &MF) {
 
   return Changed;
 }
+
+bool SILowerControlFlowLegacy::runOnMachineFunction(MachineFunction &MF) {
+  // This doesn't actually need LiveIntervals, but we can preserve them.
+  auto *LISWrapper = getAnalysisIfAvailable<LiveIntervalsWrapperPass>();
+  LiveIntervals *LIS = LISWrapper ? &LISWrapper->getLIS() : nullptr;
+  // This doesn't actually need LiveVariables, but we can preserve them.
+  auto *LVWrapper = getAnalysisIfAvailable<LiveVariablesWrapperPass>();
+  LiveVariables *LV = LVWrapper ? &LVWrapper->getLV() : nullptr;
+  auto *MDTWrapper = getAnalysisIfAvailable<MachineDominatorTreeWrapperPass>();
+  MachineDominatorTree *MDT = MDTWrapper ? &MDTWrapper->getDomTree() : nullptr;
+  return SILowerControlFlow(LIS, LV, MDT).run(MF);
+}
+
+PreservedAnalyses
+SILowerControlFlowPass::run(MachineFunction &MF,
+                            MachineFunctionAnalysisManager &MFAM) {
+  LiveIntervals *LIS = MFAM.getCachedResult<LiveIntervalsAnalysis>(MF);
+  LiveVariables *LV = MFAM.getCachedResult<LiveVariablesAnalysis>(MF);
+  MachineDominatorTree *MDT =
+      MFAM.getCachedResult<MachineDominatorTreeAnalysis>(MF);
+
+  bool Changed = SILowerControlFlow(LIS, LV, MDT).run(MF);
+  if (!Changed)
+    return PreservedAnalyses::all();
+
+  auto PA = getMachineFunctionPassPreservedAnalyses();
+  PA.preserve<MachineDominatorTreeAnalysis>();
+  PA.preserve<SlotIndexesAnalysis>();
+  PA.preserve<LiveIntervalsAnalysis>();
+  PA.preserve<LiveVariablesAnalysis>();
+  return PA;
+}

llvm/lib/Target/AMDGPU/SILowerControlFlow.h

@@ -0,0 +1,22 @@
+//===- SILowerControlFlow.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_SILOWERCONTROLFLOW_H
+#define LLVM_LIB_TARGET_AMDGPU_SILOWERCONTROLFLOW_H
+
+#include "llvm/CodeGen/MachinePassManager.h"
+
+namespace llvm {
+class SILowerControlFlowPass : public PassInfoMixin<SILowerControlFlowPass> {
+public:
+  PreservedAnalyses run(MachineFunction &MF,
+                        MachineFunctionAnalysisManager &MFAM);
+};
+} // namespace llvm
+
+#endif // LLVM_LIB_TARGET_AMDGPU_SILOWERCONTROLFLOW_H

llvm/test/CodeGen/AMDGPU/collapse-endcf.mir

@@ -1,5 +1,6 @@
 # NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py
 # RUN: llc -mtriple=amdgcn -verify-machineinstrs -run-pass=si-lower-control-flow -amdgpu-remove-redundant-endcf %s -o - | FileCheck -check-prefix=GCN %s
+# RUN: llc -mtriple=amdgcn -passes=si-lower-control-flow -amdgpu-remove-redundant-endcf %s -o - | FileCheck -check-prefix=GCN %s
 
 # Make sure dbg_value doesn't change codeegn when collapsing end_cf
 ---

llvm/test/CodeGen/AMDGPU/lower-control-flow-live-intervals.mir

@@ -1,5 +1,6 @@
 # NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py UTC_ARGS: --version 3
 # RUN: llc -run-pass=liveintervals -run-pass=si-lower-control-flow -mtriple=amdgcn--amdpal -mcpu=gfx1030 -verify-machineinstrs -o - %s | FileCheck %s
+# RUN: llc -passes='require<live-intervals>,si-lower-control-flow' -mtriple=amdgcn--amdpal -mcpu=gfx1030 -o - %s | FileCheck %s
 
 # Check that verifier passes for the following.
 

llvm/test/CodeGen/AMDGPU/lower-control-flow-other-terminators.mir

@@ -1,5 +1,6 @@
 # NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py
 # RUN: llc -mtriple=amdgcn -mcpu=fiji -verify-machineinstrs -run-pass=si-lower-control-flow -o - %s | FileCheck %s
+# RUN: llc -mtriple=amdgcn -mcpu=fiji -passes=si-lower-control-flow -o - %s | FileCheck %s
 
 # Test si-lower-control-flow insertion points when other terminator
 # instructions are present besides the control flow pseudo and a