Merging part of r259297:

We need to correctly initialize the AMDGPUPromoteAlloca pass, because
later commits will add tests that try to pass the -amdgpu-promote-alloca
flag to opt.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/branches/release_38@271591 91177308-0d34-0410-b5e6-96231b3b80d8

diff --git a/lib/Target/AMDGPU/AMDGPU.h b/lib/Target/AMDGPU/AMDGPU.h
index a6f49e8c326b97ed785301ab13e7988c007412a8..0cff46295885aa745a0a1ad8a332352673555806 100644 (file)
--- a/lib/Target/AMDGPU/AMDGPU.h
+++ b/lib/Target/AMDGPU/AMDGPU.h
@@ -70,7 +70,10 @@ void initializeSILoadStoreOptimizerPass(PassRegistry &);
 extern char &SILoadStoreOptimizerID;
 
 // Passes common to R600 and SI
-FunctionPass *createAMDGPUPromoteAlloca(const AMDGPUSubtarget &ST);
+FunctionPass *createAMDGPUPromoteAlloca(const TargetMachine *TM = nullptr);
+void initializeAMDGPUPromoteAllocaPass(PassRegistry&);
+extern char &AMDGPUPromoteAllocaID;
+
 Pass *createAMDGPUStructurizeCFGPass();
 FunctionPass *createAMDGPUISelDag(TargetMachine &tm);
 ModulePass *createAMDGPUAlwaysInlinePass();

diff --git a/lib/Target/AMDGPU/AMDGPUPromoteAlloca.cpp b/lib/Target/AMDGPU/AMDGPUPromoteAlloca.cpp
index ce2a4529d10d60140feef4f037590706f85602a0..852934231a48a3bc138725920ff7389e0013b99d 100644 (file)
--- a/lib/Target/AMDGPU/AMDGPUPromoteAlloca.cpp
+++ b/lib/Target/AMDGPU/AMDGPUPromoteAlloca.cpp
@@ -27,16 +27,21 @@ using namespace llvm;
 namespace {
 
 class AMDGPUPromoteAlloca : public FunctionPass,
-                       public InstVisitor<AMDGPUPromoteAlloca> {
-
-  static char ID;
+                            public InstVisitor<AMDGPUPromoteAlloca> {
+private:
+  const TargetMachine *TM;
   Module *Mod;
-  const AMDGPUSubtarget &ST;
   int LocalMemAvailable;
 
 public:
-  AMDGPUPromoteAlloca(const AMDGPUSubtarget &st) : FunctionPass(ID), ST(st),
-                                                   LocalMemAvailable(0) { }
+  static char ID;
+
+  AMDGPUPromoteAlloca(const TargetMachine *TM_ = nullptr) :
+    FunctionPass(ID),
+    TM (TM_),
+    Mod(nullptr),
+    LocalMemAvailable(0) { }
+
   bool doInitialization(Module &M) override;
   bool runOnFunction(Function &F) override;
   const char *getPassName() const override { return "AMDGPU Promote Alloca"; }
@@ -47,12 +52,24 @@ public:
 
 char AMDGPUPromoteAlloca::ID = 0;
 
+INITIALIZE_TM_PASS(AMDGPUPromoteAlloca, DEBUG_TYPE,
+                   "AMDGPU promote alloca to vector or LDS", false, false)
+
+char &llvm::AMDGPUPromoteAllocaID = AMDGPUPromoteAlloca::ID;
+
 bool AMDGPUPromoteAlloca::doInitialization(Module &M) {
+  if (!TM)
+    return false;
+
   Mod = &M;
   return false;
 }
 
 bool AMDGPUPromoteAlloca::runOnFunction(Function &F) {
+  if (!TM)
+    return false;
+
+  const AMDGPUSubtarget &ST = TM->getSubtarget<AMDGPUSubtarget>(F);
 
   FunctionType *FTy = F.getFunctionType();
 
@@ -428,6 +445,6 @@ void AMDGPUPromoteAlloca::visitAlloca(AllocaInst &I) {
   }
 }
 
-FunctionPass *llvm::createAMDGPUPromoteAlloca(const AMDGPUSubtarget &ST) {
-  return new AMDGPUPromoteAlloca(ST);
+FunctionPass *llvm::createAMDGPUPromoteAlloca(const TargetMachine *TM) {
+  return new AMDGPUPromoteAlloca(TM);
 }

diff --git a/lib/Target/AMDGPU/AMDGPUTargetMachine.cpp b/lib/Target/AMDGPU/AMDGPUTargetMachine.cpp
index 5a91be442771be81d89e66985d721db7414fabd9..7b4fbea64efb5433882d8964df56190743c93f07 100644 (file)
--- a/lib/Target/AMDGPU/AMDGPUTargetMachine.cpp
+++ b/lib/Target/AMDGPU/AMDGPUTargetMachine.cpp
@@ -52,6 +52,7 @@ extern "C" void LLVMInitializeAMDGPUTarget() {
   initializeSILoadStoreOptimizerPass(*PR);
   initializeAMDGPUAnnotateKernelFeaturesPass(*PR);
   initializeAMDGPUAnnotateUniformValuesPass(*PR);
+  initializeAMDGPUPromoteAllocaPass(*PR);
   initializeSIAnnotateControlFlowPass(*PR);
 }
 
@@ -228,7 +229,7 @@ void AMDGPUPassConfig::addIRPasses() {
 void AMDGPUPassConfig::addCodeGenPrepare() {
   const AMDGPUSubtarget &ST = *getAMDGPUTargetMachine().getSubtargetImpl();
   if (ST.isPromoteAllocaEnabled()) {
-    addPass(createAMDGPUPromoteAlloca(ST));
+    addPass(createAMDGPUPromoteAlloca(TM));
     addPass(createSROAPass());
   }
   TargetPassConfig::addCodeGenPrepare();