[PATCH] D45387: [CUDA] Revert defining __CUDA_ARCH__ for amdgcn targets

2018-04-09 Thread Yaxun Liu via Phabricator via cfe-commits 
This revision was automatically updated to reflect the committed changes.
Closed by commit rL329584: [CUDA] Revert defining __CUDA_ARCH__ for amdgcn 
targets (authored by yaxunl, committed by ).
Herald added a subscriber: llvm-commits.

Changed prior to commit:
  https://reviews.llvm.org/D45387?vs=141403=141664#toc

Repository:
  rL LLVM

https://reviews.llvm.org/D45387

Files:
  cfe/trunk/lib/Basic/Targets.cpp
  cfe/trunk/lib/Basic/Targets.h
  cfe/trunk/lib/Basic/Targets/AMDGPU.cpp
  cfe/trunk/lib/Basic/Targets/AMDGPU.h
  cfe/trunk/lib/Basic/Targets/NVPTX.cpp

Index: cfe/trunk/lib/Basic/Targets.h
===
--- cfe/trunk/lib/Basic/Targets.h
+++ cfe/trunk/lib/Basic/Targets.h
@@ -16,7 +16,6 @@
 #ifndef LLVM_CLANG_LIB_BASIC_TARGETS_H
 #define LLVM_CLANG_LIB_BASIC_TARGETS_H
 
-#include "clang/Basic/Cuda.h"
 #include "clang/Basic/LangOptions.h"
 #include "clang/Basic/MacroBuilder.h"
 #include "clang/Basic/TargetInfo.h"
@@ -47,9 +46,6 @@
 LLVM_LIBRARY_VISIBILITY
 void addCygMingDefines(const clang::LangOptions ,
clang::MacroBuilder );
-
-LLVM_LIBRARY_VISIBILITY
-void defineCudaArchMacro(CudaArch GPU, clang::MacroBuilder );
 } // namespace targets
 } // namespace clang
 #endif // LLVM_CLANG_LIB_BASIC_TARGETS_H
Index: cfe/trunk/lib/Basic/Targets/NVPTX.cpp
===
--- cfe/trunk/lib/Basic/Targets/NVPTX.cpp
+++ cfe/trunk/lib/Basic/Targets/NVPTX.cpp
@@ -153,8 +153,61 @@
MacroBuilder ) const {
   Builder.defineMacro("__PTX__");
   Builder.defineMacro("__NVPTX__");
-  if (Opts.CUDAIsDevice)
-defineCudaArchMacro(GPU, Builder);
+  if (Opts.CUDAIsDevice) {
+// Set __CUDA_ARCH__ for the GPU specified.
+std::string CUDAArchCode = [this] {
+  switch (GPU) {
+  case CudaArch::GFX600:
+  case CudaArch::GFX601:
+  case CudaArch::GFX700:
+  case CudaArch::GFX701:
+  case CudaArch::GFX702:
+  case CudaArch::GFX703:
+  case CudaArch::GFX704:
+  case CudaArch::GFX801:
+  case CudaArch::GFX802:
+  case CudaArch::GFX803:
+  case CudaArch::GFX810:
+  case CudaArch::GFX900:
+  case CudaArch::GFX902:
+  case CudaArch::LAST:
+break;
+  case CudaArch::UNKNOWN:
+assert(false && "No GPU arch when compiling CUDA device code.");
+return "";
+  case CudaArch::SM_20:
+return "200";
+  case CudaArch::SM_21:
+return "210";
+  case CudaArch::SM_30:
+return "300";
+  case CudaArch::SM_32:
+return "320";
+  case CudaArch::SM_35:
+return "350";
+  case CudaArch::SM_37:
+return "370";
+  case CudaArch::SM_50:
+return "500";
+  case CudaArch::SM_52:
+return "520";
+  case CudaArch::SM_53:
+return "530";
+  case CudaArch::SM_60:
+return "600";
+  case CudaArch::SM_61:
+return "610";
+  case CudaArch::SM_62:
+return "620";
+  case CudaArch::SM_70:
+return "700";
+  case CudaArch::SM_72:
+return "720";
+  }
+  llvm_unreachable("unhandled CudaArch");
+}();
+Builder.defineMacro("__CUDA_ARCH__", CUDAArchCode);
+  }
 }
 
 ArrayRef NVPTXTargetInfo::getTargetBuiltins() const {
Index: cfe/trunk/lib/Basic/Targets/AMDGPU.cpp
===
--- cfe/trunk/lib/Basic/Targets/AMDGPU.cpp
+++ cfe/trunk/lib/Basic/Targets/AMDGPU.cpp
@@ -12,7 +12,6 @@
 //===--===//
 
 #include "AMDGPU.h"
-#include "Targets.h"
 #include "clang/Basic/Builtins.h"
 #include "clang/Basic/LangOptions.h"
 #include "clang/Basic/MacroBuilder.h"
@@ -264,7 +263,6 @@
   resetDataLayout(isAMDGCN(getTriple()) ? DataLayoutStringAMDGCN
 : DataLayoutStringR600);
   assert(DataLayout->getAllocaAddrSpace() == Private);
-  GCN_Subarch = CudaArch::GFX803; // Default to fiji
 
   setAddressSpaceMap(Triple.getOS() == llvm::Triple::Mesa3D ||
  !isAMDGCN(Triple));
@@ -309,9 +307,6 @@
   if (GPU.Kind != GK_NONE)
 Builder.defineMacro(Twine("__") + Twine(GPU.CanonicalName) + Twine("__"));
 
-  if (Opts.CUDAIsDevice)
-defineCudaArchMacro(GCN_Subarch, Builder);
-
   // TODO: __HAS_FMAF__, __HAS_LDEXPF__, __HAS_FP64__ are deprecated and will be
   // removed in the near future.
   if (GPU.HasFMAF)
Index: cfe/trunk/lib/Basic/Targets/AMDGPU.h
===
--- cfe/trunk/lib/Basic/Targets/AMDGPU.h
+++ cfe/trunk/lib/Basic/Targets/AMDGPU.h
@@ -14,7 +14,6 @@
 #ifndef LLVM_CLANG_LIB_BASIC_TARGETS_AMDGPU_H
 #define LLVM_CLANG_LIB_BASIC_TARGETS_AMDGPU_H
 
-#include "clang/Basic/Cuda.h"
 #include "clang/Basic/TargetInfo.h"
 #include "clang/Basic/TargetOptions.h"
 #include "llvm/ADT/StringSet.h"
@@ -175,7 +174,6 @@
   static bool

[PATCH] D45387: [CUDA] Revert defining __CUDA_ARCH__ for amdgcn targets

2018-04-06 Thread Yaxun Liu via Phabricator via cfe-commits 
yaxunl added a comment.

In https://reviews.llvm.org/D45387#1060194, @tra wrote:

> You may want to mention in the commit log that this is a partial unroll of 
> r329232 / https://reviews.llvm.org/D45277.


Will do. Thanks.


https://reviews.llvm.org/D45387



___
cfe-commits mailing list
cfe-commits@lists.llvm.org
http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits

[PATCH] D45387: [CUDA] Revert defining __CUDA_ARCH__ for amdgcn targets

2018-04-06 Thread Artem Belevich via Phabricator via cfe-commits 
tra accepted this revision.
tra added a comment.
This revision is now accepted and ready to land.

You may want to mention in the commit log that this is a partial unroll of 
r329232 / https://reviews.llvm.org/D45277.


https://reviews.llvm.org/D45387



___
cfe-commits mailing list
cfe-commits@lists.llvm.org
http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits

[PATCH] D45387: [CUDA] Revert defining __CUDA_ARCH__ for amdgcn targets

2018-04-06 Thread Yaxun Liu via Phabricator via cfe-commits 
yaxunl created this revision.
yaxunl added a reviewer: tra.
Herald added subscribers: nhaehnle, jholewinski.

amdgcn targets only support HIP, which does not define `__CUDA_ARCH__`.

`__HIP_DEVICE_COMPILE__` depends on HIP language mode, therefore it needs to be 
added
in a different patch.

refer to discussions in https://reviews.llvm.org/D45277


https://reviews.llvm.org/D45387

Files:
  lib/Basic/Targets.cpp
  lib/Basic/Targets.h
  lib/Basic/Targets/AMDGPU.cpp
  lib/Basic/Targets/AMDGPU.h
  lib/Basic/Targets/NVPTX.cpp

Index: lib/Basic/Targets/NVPTX.cpp
===
--- lib/Basic/Targets/NVPTX.cpp
+++ lib/Basic/Targets/NVPTX.cpp
@@ -153,8 +153,61 @@
MacroBuilder ) const {
   Builder.defineMacro("__PTX__");
   Builder.defineMacro("__NVPTX__");
-  if (Opts.CUDAIsDevice)
-defineCudaArchMacro(GPU, Builder);
+  if (Opts.CUDAIsDevice) {
+// Set __CUDA_ARCH__ for the GPU specified.
+std::string CUDAArchCode = [this] {
+  switch (GPU) {
+  case CudaArch::GFX600:
+  case CudaArch::GFX601:
+  case CudaArch::GFX700:
+  case CudaArch::GFX701:
+  case CudaArch::GFX702:
+  case CudaArch::GFX703:
+  case CudaArch::GFX704:
+  case CudaArch::GFX801:
+  case CudaArch::GFX802:
+  case CudaArch::GFX803:
+  case CudaArch::GFX810:
+  case CudaArch::GFX900:
+  case CudaArch::GFX902:
+  case CudaArch::LAST:
+break;
+  case CudaArch::UNKNOWN:
+assert(false && "No GPU arch when compiling CUDA device code.");
+return "";
+  case CudaArch::SM_20:
+return "200";
+  case CudaArch::SM_21:
+return "210";
+  case CudaArch::SM_30:
+return "300";
+  case CudaArch::SM_32:
+return "320";
+  case CudaArch::SM_35:
+return "350";
+  case CudaArch::SM_37:
+return "370";
+  case CudaArch::SM_50:
+return "500";
+  case CudaArch::SM_52:
+return "520";
+  case CudaArch::SM_53:
+return "530";
+  case CudaArch::SM_60:
+return "600";
+  case CudaArch::SM_61:
+return "610";
+  case CudaArch::SM_62:
+return "620";
+  case CudaArch::SM_70:
+return "700";
+  case CudaArch::SM_72:
+return "720";
+  }
+  llvm_unreachable("unhandled CudaArch");
+}();
+Builder.defineMacro("__CUDA_ARCH__", CUDAArchCode);
+  }
 }
 
 ArrayRef NVPTXTargetInfo::getTargetBuiltins() const {
Index: lib/Basic/Targets/AMDGPU.h
===
--- lib/Basic/Targets/AMDGPU.h
+++ lib/Basic/Targets/AMDGPU.h
@@ -14,7 +14,6 @@
 #ifndef LLVM_CLANG_LIB_BASIC_TARGETS_AMDGPU_H
 #define LLVM_CLANG_LIB_BASIC_TARGETS_AMDGPU_H
 
-#include "clang/Basic/Cuda.h"
 #include "clang/Basic/TargetInfo.h"
 #include "clang/Basic/TargetOptions.h"
 #include "llvm/ADT/StringSet.h"
@@ -175,7 +174,6 @@
   static bool isAMDGCN(const llvm::Triple ) {
 return TT.getArch() == llvm::Triple::amdgcn;
   }
-  CudaArch GCN_Subarch;
 
 public:
   AMDGPUTargetInfo(const llvm::Triple , const TargetOptions );
@@ -332,7 +330,6 @@
 else
   GPU = parseR600Name(Name);
 
-GCN_Subarch = StringToCudaArch(Name);
 return GK_NONE != GPU.Kind;
   }
 
Index: lib/Basic/Targets/AMDGPU.cpp
===
--- lib/Basic/Targets/AMDGPU.cpp
+++ lib/Basic/Targets/AMDGPU.cpp
@@ -12,7 +12,6 @@
 //===--===//
 
 #include "AMDGPU.h"
-#include "Targets.h"
 #include "clang/Basic/Builtins.h"
 #include "clang/Basic/LangOptions.h"
 #include "clang/Basic/MacroBuilder.h"
@@ -264,7 +263,6 @@
   resetDataLayout(isAMDGCN(getTriple()) ? DataLayoutStringAMDGCN
 : DataLayoutStringR600);
   assert(DataLayout->getAllocaAddrSpace() == Private);
-  GCN_Subarch = CudaArch::GFX803; // Default to fiji
 
   setAddressSpaceMap(Triple.getOS() == llvm::Triple::Mesa3D ||
  !isAMDGCN(Triple));
@@ -309,9 +307,6 @@
   if (GPU.Kind != GK_NONE)
 Builder.defineMacro(Twine("__") + Twine(GPU.CanonicalName) + Twine("__"));
 
-  if (Opts.CUDAIsDevice)
-defineCudaArchMacro(GCN_Subarch, Builder);
-
   // TODO: __HAS_FMAF__, __HAS_LDEXPF__, __HAS_FP64__ are deprecated and will be
   // removed in the near future.
   if (GPU.HasFMAF)
Index: lib/Basic/Targets.h
===
--- lib/Basic/Targets.h
+++ lib/Basic/Targets.h
@@ -16,7 +16,6 @@
 #ifndef LLVM_CLANG_LIB_BASIC_TARGETS_H
 #define LLVM_CLANG_LIB_BASIC_TARGETS_H
 
-#include "clang/Basic/Cuda.h"
 #include "clang/Basic/LangOptions.h"
 #include "clang/Basic/MacroBuilder.h"
 #include "clang/Basic/TargetInfo.h"
@@ -47,9 +46,6 @@
 LLVM_LIBRARY_VISIBILITY
 void addCygMingDefines(const clang::LangOptions ,
clang::MacroBuilder );
-