DivergencyAnalysis patch for review

llvm-svn: 305494

5 files changed, 20 insertions, 2 deletions

diff --git a/llvm/lib/Analysis/DivergenceAnalysis.cpp b/llvm/lib/Analysis/DivergenceAnalysis.cpp
index 1b36569f7a0..2d39a0b0215 100644
--- a/llvm/lib/Analysis/DivergenceAnalysis.cpp
+++ b/llvm/lib/Analysis/DivergenceAnalysis.cpp
@@ -241,7 +241,7 @@ void DivergencePropagator::exploreDataDependency(Value *V) {
   // Follow def-use chains of V.
   for (User *U : V->users()) {
     Instruction *UserInst = cast<Instruction>(U);
-    if (DV.insert(UserInst).second)
+    if (!TTI.isAlwaysUniform(U) && DV.insert(UserInst).second)
       Worklist.push_back(UserInst);
   }
 }
diff --git a/llvm/lib/Analysis/TargetTransformInfo.cpp b/llvm/lib/Analysis/TargetTransformInfo.cpp
index 488cb332a0b..92328f6e5ef 100644
--- a/llvm/lib/Analysis/TargetTransformInfo.cpp
+++ b/llvm/lib/Analysis/TargetTransformInfo.cpp
@@ -103,6 +103,10 @@ bool TargetTransformInfo::isSourceOfDivergence(const Value *V) const {
   return TTIImpl->isSourceOfDivergence(V);
 }
 
+bool llvm::TargetTransformInfo::isAlwaysUniform(const Value *V) const {
+  return TTIImpl->isAlwaysUniform(V);
+}
+
 unsigned TargetTransformInfo::getFlatAddressSpace() const {
   return TTIImpl->getFlatAddressSpace();
 }
diff --git a/llvm/lib/Target/AMDGPU/AMDGPUAnnotateUniformValues.cpp b/llvm/lib/Target/AMDGPU/AMDGPUAnnotateUniformValues.cpp
index 3c788fa1dce..6f002860044 100644
--- a/llvm/lib/Target/AMDGPU/AMDGPUAnnotateUniformValues.cpp
+++ b/llvm/lib/Target/AMDGPU/AMDGPUAnnotateUniformValues.cpp
@@ -107,7 +107,7 @@ bool AMDGPUAnnotateUniformValues::isClobberedInFunction(LoadInst * Load) {
 
   DFS(Start, Checklist);
   for (auto &BB : Checklist) {
-    BasicBlock::iterator StartIt = (BB == Load->getParent()) ?
+    BasicBlock::iterator StartIt = (!L && (BB == Load->getParent())) ?
      BasicBlock::iterator(Load) : BB->end();
      if (MDR->getPointerDependencyFrom(MemoryLocation(Ptr),
        true, StartIt, BB, Load).isClobber())
diff --git a/llvm/lib/Target/AMDGPU/AMDGPUTargetTransformInfo.cpp b/llvm/lib/Target/AMDGPU/AMDGPUTargetTransformInfo.cpp
index 65dba7d6055..0d6689bd04c 100644
--- a/llvm/lib/Target/AMDGPU/AMDGPUTargetTransformInfo.cpp
+++ b/llvm/lib/Target/AMDGPU/AMDGPUTargetTransformInfo.cpp
@@ -489,6 +489,19 @@ bool AMDGPUTTIImpl::isSourceOfDivergence(const Value *V) const {
   return false;
 }
 
+bool AMDGPUTTIImpl::isAlwaysUniform(const Value *V) const {
+  if (const IntrinsicInst *Intrinsic = dyn_cast<IntrinsicInst>(V)) {
+    switch (Intrinsic->getIntrinsicID()) {
+    default:
+      return false;
+    case Intrinsic::amdgcn_readfirstlane:
+    case Intrinsic::amdgcn_readlane:
+      return true;
+    }
+  }
+  return false;
+}
+
 unsigned AMDGPUTTIImpl::getShuffleCost(TTI::ShuffleKind Kind, Type *Tp, int Index,
                                        Type *SubTp) {
   if (ST->hasVOP3PInsts()) {
diff --git a/llvm/lib/Target/AMDGPU/AMDGPUTargetTransformInfo.h b/llvm/lib/Target/AMDGPU/AMDGPUTargetTransformInfo.h
index 8466f118395..a60b1bb1b59 100644
--- a/llvm/lib/Target/AMDGPU/AMDGPUTargetTransformInfo.h
+++ b/llvm/lib/Target/AMDGPU/AMDGPUTargetTransformInfo.h
@@ -103,6 +103,7 @@ public:
 
   int getVectorInstrCost(unsigned Opcode, Type *ValTy, unsigned Index);
   bool isSourceOfDivergence(const Value *V) const;
+  bool isAlwaysUniform(const Value *V) const;
 
   unsigned getFlatAddressSpace() const {
     // Don't bother running InferAddressSpaces pass on graphics shaders which