[PATCH] D16079: [CUDA] Reject values for --cuda-gpu-arch that are not of the form /sm_\d+/.

[Justin Lebar via cfe-commits]

jlebar created this revision.
jlebar added a reviewer: tra.
jlebar added subscribers: echristo, jhen, cfe-commits.

http://reviews.llvm.org/D16079

Files:
  include/clang/Driver/Action.h
  lib/Driver/Action.cpp
  lib/Driver/Driver.cpp
  test/Driver/cuda-bad-arch.cu


Index: test/Driver/cuda-bad-arch.cu
===================================================================
--- /dev/null
+++ test/Driver/cuda-bad-arch.cu
@@ -0,0 +1,34 @@
+// Checks errors generated by passing a bad value for --cuda-gpu-arch.
+// REQUIRES: clang-driver
+// REQUIRES: x86-registered-target
+// REQUIRES: nvptx-registered-target
+
+// 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: | 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: | FileCheck -check-prefix OK %s
+// RUN: %clang -### -target x86_64-linux-gnu -c %s 2>&1 \
+// RUN: | FileCheck -check-prefix OK %s
+
+// OK-NOT: error: Unsupported CUDA gpu architecture
Index: lib/Driver/Driver.cpp
===================================================================
--- lib/Driver/Driver.cpp
+++ lib/Driver/Driver.cpp
@@ -1297,8 +1297,12 @@
     if (!A->getOption().matches(options::OPT_cuda_gpu_arch_EQ))
       continue;
     A->claim();
-    if (GpuArchNames.insert(A->getValue()).second)
-      GpuArchList.push_back(A->getValue());
+
+    const auto& Arch = A->getValue();
+    if (!CudaDeviceAction::IsValidGpuArchName(Arch))
+      C.getDriver().Diag(clang::diag::err_drv_cuda_bad_gpu_arch) << Arch;
+    else if (GpuArchNames.insert(Arch).second)
+      GpuArchList.push_back(Arch);
   }
 
   // Default to sm_20 which is the lowest common denominator for supported GPUs.
Index: lib/Driver/Action.cpp
===================================================================
--- lib/Driver/Action.cpp
+++ lib/Driver/Action.cpp
@@ -9,6 +9,7 @@
 
 #include "clang/Driver/Action.h"
 #include "llvm/Support/ErrorHandling.h"
+#include "llvm/Support/Regex.h"
 #include <cassert>
 using namespace clang::driver;
 using namespace llvm::opt;
@@ -54,7 +55,14 @@
 CudaDeviceAction::CudaDeviceAction(Action *Input, const char *ArchName,
                                    bool AtTopLevel)
     : Action(CudaDeviceClass, Input), GpuArchName(ArchName),
-      AtTopLevel(AtTopLevel) {}
+      AtTopLevel(AtTopLevel) {
+  assert(IsValidGpuArchName(GpuArchName));
+}
+
+bool CudaDeviceAction::IsValidGpuArchName(llvm::StringRef ArchName) {
+  static llvm::Regex RE("^sm_[0-9]+$");
+  return RE.match(ArchName);
+}
 
 void CudaHostAction::anchor() {}
 
Index: include/clang/Driver/Action.h
===================================================================
--- include/clang/Driver/Action.h
+++ include/clang/Driver/Action.h
@@ -15,6 +15,9 @@
 #include "llvm/ADT/SmallVector.h"
 
 namespace llvm {
+
+class StringRef;
+
 namespace opt {
   class Arg;
 }
@@ -133,7 +136,7 @@
 
 class CudaDeviceAction : public Action {
   virtual void anchor();
-  /// GPU architecture to bind -- e.g 'sm_35'.
+  /// GPU architecture to bind.  Always of the form /sm_\d+/.
   const char *GpuArchName;
   /// True when action results are not consumed by the host action (e.g when
   /// -fsyntax-only or --cuda-device-only options are used).
@@ -145,6 +148,8 @@
   const char *getGpuArchName() const { return GpuArchName; }
   bool isAtTopLevel() const { return AtTopLevel; }
 
+  static bool IsValidGpuArchName(llvm::StringRef ArchName);
+
   static bool classof(const Action *A) {
     return A->getKind() == CudaDeviceClass;
   }

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