[CUDA] Add explicit mapping from sm_XX to compute_YY.

Summary: This is used by D16082 when it invokes fatbinary.

Reviewers: tra

Subscribers: cfe-commits, jhen, echristo

Differential Revision: http://reviews.llvm.org/D16097

llvm-svn: 257530

3 files changed, 30 insertions, 16 deletions

diff --git a/clang/include/clang/Driver/Action.h b/clang/include/clang/Driver/Action.h
index 88d0def4448..c5b0f475509 100644
--- a/clang/include/clang/Driver/Action.h
+++ b/clang/include/clang/Driver/Action.h
@@ -146,6 +146,10 @@ public:
   CudaDeviceAction(Action *Input, const char *ArchName, bool AtTopLevel);
 
   const char *getGpuArchName() const { return GpuArchName; }
+
+  /// Gets the compute_XX that corresponds to getGpuArchName().
+  const char *getComputeArchName() const;
+
   bool isAtTopLevel() const { return AtTopLevel; }
 
   static bool IsValidGpuArchName(llvm::StringRef ArchName);
diff --git a/clang/lib/Driver/Action.cpp b/clang/lib/Driver/Action.cpp
index 0117f8ab0be..e9490e96db8 100644
--- a/clang/lib/Driver/Action.cpp
+++ b/clang/lib/Driver/Action.cpp
@@ -8,6 +8,7 @@
 //===----------------------------------------------------------------------===//
 
 #include "clang/Driver/Action.h"
+#include "llvm/ADT/StringSwitch.h"
 #include "llvm/Support/ErrorHandling.h"
 #include "llvm/Support/Regex.h"
 #include <cassert>
@@ -50,6 +51,24 @@ void BindArchAction::anchor() {}
 BindArchAction::BindArchAction(Action *Input, const char *_ArchName)
     : Action(BindArchClass, Input), ArchName(_ArchName) {}
 
+// Converts CUDA GPU architecture, e.g. "sm_21", to its corresponding virtual
+// compute arch, e.g. "compute_20".  Returns null if the input arch is null or
+// doesn't match an existing arch.
+static const char* GpuArchToComputeName(const char *ArchName) {
+  if (!ArchName)
+    return nullptr;
+  return llvm::StringSwitch<const char *>(ArchName)
+      .Cases("sm_20", "sm_21", "compute_20")
+      .Case("sm_30", "compute_30")
+      .Case("sm_32", "compute_32")
+      .Case("sm_35", "compute_35")
+      .Case("sm_37", "compute_37")
+      .Case("sm_50", "compute_50")
+      .Case("sm_52", "compute_52")
+      .Case("sm_53", "compute_53")
+      .Default(nullptr);
+}
+
 void CudaDeviceAction::anchor() {}
 
 CudaDeviceAction::CudaDeviceAction(Action *Input, const char *ArchName,
@@ -59,9 +78,12 @@ CudaDeviceAction::CudaDeviceAction(Action *Input, const char *ArchName,
   assert(IsValidGpuArchName(GpuArchName));
 }
 
+const char *CudaDeviceAction::getComputeArchName() const {
+  return GpuArchToComputeName(GpuArchName);
+}
+
 bool CudaDeviceAction::IsValidGpuArchName(llvm::StringRef ArchName) {
-  static llvm::Regex RE("^sm_[0-9]+$");
-  return RE.match(ArchName);
+  return GpuArchToComputeName(ArchName.data()) != nullptr;
 }
 
 void CudaHostAction::anchor() {}
diff --git a/clang/test/Driver/cuda-bad-arch.cu b/clang/test/Driver/cuda-bad-arch.cu
index 67bac30060c..f92bdcebd0d 100644
--- a/clang/test/Driver/cuda-bad-arch.cu
+++ b/clang/test/Driver/cuda-bad-arch.cu
@@ -5,28 +5,16 @@
 
 // RUN: %clang -### -target x86_64-linux-gnu --cuda-gpu-arch=compute_20 -c %s 2>&1 \
 // RUN: | FileCheck -check-prefix BAD %s
-// RUN: %clang -### -target x86_64-linux-gnu --cuda-gpu-arch=foo_20 -c %s 2>&1 \
-// RUN: | FileCheck -check-prefix BAD %s
 // RUN: %clang -### -target x86_64-linux-gnu --cuda-gpu-arch=sm20 -c %s 2>&1 \
 // RUN: | FileCheck -check-prefix BAD %s
-// RUN: %clang -### -target x86_64-linux-gnu --cuda-gpu-arch=sm_abc -c %s 2>&1 \
-// RUN: | FileCheck -check-prefix BAD %s
-// RUN: %clang -### -target x86_64-linux-gnu --cuda-gpu-arch=sm_20a -c %s 2>&1 \
-// RUN: | FileCheck -check-prefix BAD %s
-// RUN: %clang -### -target x86_64-linux-gnu --cuda-gpu-arch=sm_a20 -c %s 2>&1 \
-// RUN: | FileCheck -check-prefix BAD %s
-// RUN: %clang -### -target x86_64-linux-gnu --cuda-gpu-arch=ssm_20 -c %s 2>&1 \
-// RUN: | FileCheck -check-prefix BAD %s
-// RUN: %clang -### -target x86_64-linux-gnu --cuda-gpu-arch=sm_ -c %s 2>&1 \
+// RUN: %clang -### -target x86_64-linux-gnu --cuda-gpu-arch=sm_19 -c %s 2>&1 \
 // RUN: | FileCheck -check-prefix BAD %s
 
 // BAD: error: Unsupported CUDA gpu architecture
 
-// RUN: %clang -### -target x86_64-linux-gnu --cuda-gpu-arch=sm_2 -c %s 2>&1 \
-// RUN: | FileCheck -check-prefix OK %s
 // RUN: %clang -### -target x86_64-linux-gnu --cuda-gpu-arch=sm_20 -c %s 2>&1 \
 // RUN: | FileCheck -check-prefix OK %s
-// RUN: %clang -### -target x86_64-linux-gnu --cuda-gpu-arch=sm_999 -c %s 2>&1 \
+// RUN: %clang -### -target x86_64-linux-gnu --cuda-gpu-arch=sm_52 -c %s 2>&1 \
 // RUN: | FileCheck -check-prefix OK %s
 // RUN: %clang -### -target x86_64-linux-gnu -c %s 2>&1 \
 // RUN: | FileCheck -check-prefix OK %s