Patch by Anton Lokhmotov to add OpenCL work group size attributes.

llvm-svn: 159965

2 files changed, 56 insertions, 8 deletions

diff --git a/clang/lib/CodeGen/CodeGenFunction.cpp b/clang/lib/CodeGen/CodeGenFunction.cpp
index 68593946f6d..c40b42193f9 100644
--- a/clang/lib/CodeGen/CodeGenFunction.cpp
+++ b/clang/lib/CodeGen/CodeGenFunction.cpp
@@ -252,6 +252,51 @@ void CodeGenFunction::EmitMCountInstrumentation() {
   Builder.CreateCall(MCountFn);
 }
 
+void CodeGenFunction::EmitOpenCLKernelMetadata(const FunctionDecl *FD, 
+                                               llvm::Function *Fn)
+{
+  if (!FD->hasAttr<OpenCLKernelAttr>())
+    return;
+
+  llvm::LLVMContext &Context = getLLVMContext();
+
+  llvm::SmallVector <llvm::Value*, 5> kernelMDArgs;
+  kernelMDArgs.push_back(Fn);
+
+  if (FD->hasAttr<WorkGroupSizeHintAttr>()) {
+    llvm::SmallVector <llvm::Value*, 5> attrMDArgs;
+    attrMDArgs.push_back(llvm::MDString::get(Context, "work_group_size_hint"));
+    WorkGroupSizeHintAttr *attr = FD->getAttr<WorkGroupSizeHintAttr>();
+    llvm::Type *iTy = llvm::IntegerType::get(Context, 32);
+    attrMDArgs.push_back(llvm::ConstantInt::get(iTy,
+       llvm::APInt(32, (uint64_t)attr->getXDim())));
+    attrMDArgs.push_back(llvm::ConstantInt::get(iTy,
+       llvm::APInt(32, (uint64_t)attr->getYDim())));
+    attrMDArgs.push_back(llvm::ConstantInt::get(iTy,
+       llvm::APInt(32, (uint64_t)attr->getZDim())));
+    kernelMDArgs.push_back(llvm::MDNode::get(Context, attrMDArgs));
+  }
+
+  if (FD->hasAttr<ReqdWorkGroupSizeAttr>()) {
+    llvm::SmallVector <llvm::Value*, 5> attrMDArgs;
+    attrMDArgs.push_back(llvm::MDString::get(Context, "reqd_work_group_size"));
+    ReqdWorkGroupSizeAttr *attr = FD->getAttr<ReqdWorkGroupSizeAttr>();
+    llvm::Type *iTy = llvm::IntegerType::get(Context, 32);
+    attrMDArgs.push_back(llvm::ConstantInt::get(iTy,
+       llvm::APInt(32, (uint64_t)attr->getXDim())));
+    attrMDArgs.push_back(llvm::ConstantInt::get(iTy,
+       llvm::APInt(32, (uint64_t)attr->getYDim())));
+    attrMDArgs.push_back(llvm::ConstantInt::get(iTy,
+       llvm::APInt(32, (uint64_t)attr->getZDim())));
+    kernelMDArgs.push_back(llvm::MDNode::get(Context, attrMDArgs));
+  }
+
+  llvm::MDNode *kernelMDNode = llvm::MDNode::get(Context, kernelMDArgs);
+  llvm::NamedMDNode *OpenCLKernelMetadata =
+    CGM.getModule().getOrInsertNamedMetadata("opencl.kernels");
+  OpenCLKernelMetadata->addOperand(kernelMDNode);
+}
+
 void CodeGenFunction::StartFunction(GlobalDecl GD, QualType RetTy,
                                     llvm::Function *Fn,
                                     const CGFunctionInfo &FnInfo,
@@ -280,14 +325,7 @@ void CodeGenFunction::StartFunction(GlobalDecl GD, QualType RetTy,
   if (getContext().getLangOpts().OpenCL) {
     // Add metadata for a kernel function.
     if (const FunctionDecl *FD = dyn_cast_or_null<FunctionDecl>(D))
-      if (FD->hasAttr<OpenCLKernelAttr>()) {
-        llvm::LLVMContext &Context = getLLVMContext();
-        llvm::NamedMDNode *OpenCLMetadata = 
-          CGM.getModule().getOrInsertNamedMetadata("opencl.kernels");
-          
-        llvm::Value *Op = Fn;
-        OpenCLMetadata->addOperand(llvm::MDNode::get(Context, Op));
-      }
+      EmitOpenCLKernelMetadata(FD, Fn);
   }
 
   llvm::BasicBlock *EntryBB = createBasicBlock("entry", CurFn);
diff --git a/clang/lib/CodeGen/CodeGenFunction.h b/clang/lib/CodeGen/CodeGenFunction.h
index a3d1acf5917..b737a9ed7ab 100644
--- a/clang/lib/CodeGen/CodeGenFunction.h
+++ b/clang/lib/CodeGen/CodeGenFunction.h
@@ -1197,6 +1197,16 @@ private:
   llvm::BasicBlock *TerminateHandler;
   llvm::BasicBlock *TrapBB;
 
+  /// Add a kernel metadata node to the named metadata node 'opencl.kernels'.
+  /// In the kernel metadata node, reference the kernel function and metadata 
+  /// nodes for its optional attribute qualifiers (OpenCL 1.1 6.7.2):
+  /// - A node for the work_group_size_hint(X,Y,Z) qualifier contains string 
+  ///   "work_group_size_hint", and three 32-bit integers X, Y and Z.
+  /// - A node for the reqd_work_group_size(X,Y,Z) qualifier contains string 
+  ///   "reqd_work_group_size", and three 32-bit integers X, Y and Z.
+  void EmitOpenCLKernelMetadata(const FunctionDecl *FD, 
+                                llvm::Function *Fn);
+
 public:
   CodeGenFunction(CodeGenModule &cgm, bool suppressNewContext=false);
   ~CodeGenFunction();