[PATCH] D20168: [CodeGen] Handle structs directly in AMDGPUABIInfo

[Vedran Miletić via cfe-commits]

rivanvx created this revision.
rivanvx added reviewers: arsenm, tstellarAMD.
rivanvx added a subscriber: cfe-commits.
Herald added a subscriber: kzhuravl.

Structs are currently handled as pointer + byval, which makes AMDGPU LLVM backend generate incorrect code when structs are used. This patch changes struct argument to be handled directly and without flattening, which Clover (Mesa 3D Gallium OpenCL state tracker) will be able to handle. Flattening would expand the struct to individual elements and pass each as a separate argument, which Clover can not handle. Furthermore, such expansion does not fit the OpenCL programming model which requires to explicitely specify each argument index, size and memory location.

This patch is a modification of a patch provided by Matt Arsenault.

http://reviews.llvm.org/D20168

Files:
  lib/CodeGen/TargetInfo.cpp

Index: lib/CodeGen/TargetInfo.cpp
===================================================================
--- lib/CodeGen/TargetInfo.cpp
+++ lib/CodeGen/TargetInfo.cpp
@@ -6808,10 +6808,41 @@
 
 namespace {
 
+class AMDGPUABIInfo final : public DefaultABIInfo {
+public:
+  explicit AMDGPUABIInfo(CodeGen::CodeGenTypes &CGT) : DefaultABIInfo(CGT) {}
+
+private:
+  ABIArgInfo classifyArgumentType(QualType Ty) const;
+
+  void computeInfo(CGFunctionInfo &FI) const override;
+};
+
+void AMDGPUABIInfo::computeInfo(CGFunctionInfo &FI) const {
+  if (!getCXXABI().classifyReturnType(FI))
+    FI.getReturnInfo() = classifyReturnType(FI.getReturnType());
+
+  for (auto &Arg : FI.arguments())
+    Arg.info = classifyArgumentType(Arg.type);
+}
+
+/// \brief Classify argument of given type \p Ty.
+ABIArgInfo AMDGPUABIInfo::classifyArgumentType(QualType Ty) const {
+  llvm::StructType *StrTy = dyn_cast<llvm::StructType>(CGT.ConvertType(Ty));
+  if (!StrTy) {
+    return DefaultABIInfo::classifyArgumentType(Ty);
+  }
+
+  // If we set CanBeFlattened to true, CodeGen will expand the struct to its
+  // individual elements, which confuses the Clover OpenCL backend; therefore we
+  // have to set it to false here. Other args of getDirect() are just defaults.
+  return ABIArgInfo::getDirect(nullptr, 0, nullptr, false);
+}
+
 class AMDGPUTargetCodeGenInfo : public TargetCodeGenInfo {
 public:
   AMDGPUTargetCodeGenInfo(CodeGenTypes &CGT)
-    : TargetCodeGenInfo(new DefaultABIInfo(CGT)) {}
+    : TargetCodeGenInfo(new AMDGPUABIInfo(CGT)) {}
   void setTargetAttributes(const Decl *D, llvm::GlobalValue *GV,
                            CodeGen::CodeGenModule &M) const override;
 };


-------------- next part --------------
A non-text attachment was scrubbed...
Name: D20168.56901.patch
Type: text/x-patch
Size: 1683 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/cfe-commits/attachments/20160511/f1a1d18c/attachment-0001.bin>