[CUDA] Added partial support for CUDA-9.1

Clang can use CUDA-9.1 now, though new APIs (are not implemented yet.

The major change is that headers in CUDA-9.1 went through substantial
changes that started in CUDA-9.0 which required substantial changes
in the cuda compatibility headers provided by clang.

There are two major issues:
* CUDA SDK no longer provides declarations for libdevice functions.
* A lot of device-side functions have become nvcc's builtins and
  CUDA headers no longer contain their implementations.

This patch changes the way CUDA headers are handled if we compile
with CUDA 9.x. Both 9.0 and 9.1 are affected.

* Clang provides its own declarations of libdevice functions.
* For CUDA-9.x clang now provides implementation of device-side
  'standard library' functions using libdevice.

This patch should not affect compilation with CUDA-8. There may be
some observable differences for CUDA-9.0, though they are not expected
to affect functionality.

Tested: CUDA test-suite tests for all supported combinations of:
        CUDA: 7.0,7.5,8.0,9.0,9.1
        GPU: sm_20, sm_35, sm_60, sm_70

Differential Revision: https://reviews.llvm.org/D42513

llvm-svn: 323713

clang/include/clang/Basic/Cuda.h

@@ -22,7 +22,8 @@ enum class CudaVersion {
   CUDA_75,
   CUDA_80,
   CUDA_90,
-  LATEST = CUDA_90,
+  CUDA_91,
+  LATEST = CUDA_91,
 };
 const char *CudaVersionToString(CudaVersion V);
 
@@ -44,6 +45,7 @@ enum class CudaArch {
   SM_61,
   SM_62,
   SM_70,
+  SM_72,
 };
 const char *CudaArchToString(CudaArch A);
 
@@ -64,6 +66,7 @@ enum class CudaVirtualArch {
   COMPUTE_61,
   COMPUTE_62,
   COMPUTE_70,
+  COMPUTE_72,
 };
 const char *CudaVirtualArchToString(CudaVirtualArch A);
 

clang/lib/Basic/Cuda.cpp

@@ -18,6 +18,8 @@ const char *CudaVersionToString(CudaVersion V) {
     return "8.0";
   case CudaVersion::CUDA_90:
     return "9.0";
+  case CudaVersion::CUDA_91:
+    return "9.1";
   }
   llvm_unreachable("invalid enum");
 }
@@ -52,6 +54,8 @@ const char *CudaArchToString(CudaArch A) {
     return "sm_62";
   case CudaArch::SM_70:
     return "sm_70";
+  case CudaArch::SM_72:
+    return "sm_72";
   }
   llvm_unreachable("invalid enum");
 }
@@ -71,6 +75,7 @@ CudaArch StringToCudaArch(llvm::StringRef S) {
       .Case("sm_61", CudaArch::SM_61)
       .Case("sm_62", CudaArch::SM_62)
       .Case("sm_70", CudaArch::SM_70)
+      .Case("sm_72", CudaArch::SM_72)
       .Default(CudaArch::UNKNOWN);
 }
 
@@ -102,6 +107,8 @@ const char *CudaVirtualArchToString(CudaVirtualArch A) {
     return "compute_62";
   case CudaVirtualArch::COMPUTE_70:
     return "compute_70";
+  case CudaVirtualArch::COMPUTE_72:
+    return "compute_72";
   }
   llvm_unreachable("invalid enum");
 }
@@ -120,6 +127,7 @@ CudaVirtualArch StringToCudaVirtualArch(llvm::StringRef S) {
       .Case("compute_61", CudaVirtualArch::COMPUTE_61)
       .Case("compute_62", CudaVirtualArch::COMPUTE_62)
       .Case("compute_70", CudaVirtualArch::COMPUTE_70)
+      .Case("compute_72", CudaVirtualArch::COMPUTE_72)
       .Default(CudaVirtualArch::UNKNOWN);
 }
 
@@ -152,6 +160,8 @@ CudaVirtualArch VirtualArchForCudaArch(CudaArch A) {
     return CudaVirtualArch::COMPUTE_62;
   case CudaArch::SM_70:
     return CudaVirtualArch::COMPUTE_70;
+  case CudaArch::SM_72:
+    return CudaVirtualArch::COMPUTE_72;
   }
   llvm_unreachable("invalid enum");
 }
@@ -176,6 +186,8 @@ CudaVersion MinVersionForCudaArch(CudaArch A) {
     return CudaVersion::CUDA_80;
   case CudaArch::SM_70:
     return CudaVersion::CUDA_90;
+  case CudaArch::SM_72:
+    return CudaVersion::CUDA_91;
   }
   llvm_unreachable("invalid enum");
 }

clang/lib/Basic/Targets/NVPTX.cpp

@@ -186,6 +186,8 @@ void NVPTXTargetInfo::getTargetDefines(const LangOptions &Opts,
         return "620";
       case CudaArch::SM_70:
         return "700";
+      case CudaArch::SM_72:
+        return "720";
       }
       llvm_unreachable("unhandled CudaArch");
     }();

clang/lib/Driver/ToolChains/Cuda.cpp

@@ -52,6 +52,8 @@ static CudaVersion ParseCudaVersionFile(llvm::StringRef V) {
     return CudaVersion::CUDA_80;
   if (Major == 9 && Minor == 0)
     return CudaVersion::CUDA_90;
+  if (Major == 9 && Minor == 1)
+    return CudaVersion::CUDA_91;
   return CudaVersion::UNKNOWN;
 }
 
@@ -119,14 +121,18 @@ CudaInstallationDetector::CudaInstallationDetector(
       Version = ParseCudaVersionFile((*VersionFile)->getBuffer());
     }
 
-    if (Version == CudaVersion::CUDA_90) {
-      // CUDA-9 uses single libdevice file for all GPU variants.
+    if (Version >= CudaVersion::CUDA_90) {
+      // CUDA-9+ uses single libdevice file for all GPU variants.
       std::string FilePath = LibDevicePath + "/libdevice.10.bc";
       if (FS.exists(FilePath)) {
-        for (const char *GpuArch :
+        for (const char *GpuArchName :
              {"sm_20", "sm_30", "sm_32", "sm_35", "sm_50", "sm_52", "sm_53",
-              "sm_60", "sm_61", "sm_62", "sm_70"})
-          LibDeviceMap[GpuArch] = FilePath;
+                   "sm_60", "sm_61", "sm_62", "sm_70", "sm_72"}) {
+          const CudaArch GpuArch = StringToCudaArch(GpuArchName);
+          if (Version >= MinVersionForCudaArch(GpuArch) &&
+              Version <= MaxVersionForCudaArch(GpuArch))
+            LibDeviceMap[GpuArchName] = FilePath;
+        }
       }
     } else {
       std::error_code EC;

clang/lib/Headers/CMakeLists.txt

@@ -34,7 +34,9 @@ set(files
   __clang_cuda_builtin_vars.h
   __clang_cuda_cmath.h
   __clang_cuda_complex_builtins.h
+  __clang_cuda_device_functions.h
   __clang_cuda_intrinsics.h
+  __clang_cuda_libdevice_declares.h
   __clang_cuda_math_forward_declares.h
   __clang_cuda_runtime_wrapper.h
   cetintrin.h