CUDA: add -fcuda-is-device flag

This frontend-only flag is used by the IR generator to determine
whether to filter CUDA declarations for the host or for the device.

llvm-svn: 141301

2 files changed, 20 insertions, 0 deletions

diff --git a/clang/lib/CodeGen/CodeGenModule.cpp b/clang/lib/CodeGen/CodeGenModule.cpp
index f5e37897bcb..d8b9c9d853c 100644
--- a/clang/lib/CodeGen/CodeGenModule.cpp
+++ b/clang/lib/CodeGen/CodeGenModule.cpp
@@ -785,6 +785,23 @@ void CodeGenModule::EmitGlobal(GlobalDecl GD) {
   if (Global->hasAttr<AliasAttr>())
     return EmitAliasDefinition(GD);
 
+  // If this is CUDA, be selective about which declarations we emit.
+  if (Features.CUDA) {
+    if (CodeGenOpts.CUDAIsDevice) {
+      if (!Global->hasAttr<CUDADeviceAttr>() &&
+          !Global->hasAttr<CUDAGlobalAttr>() &&
+          !Global->hasAttr<CUDAConstantAttr>() &&
+          !Global->hasAttr<CUDASharedAttr>())
+        return;
+    } else {
+      if (!Global->hasAttr<CUDAHostAttr>() && (
+            Global->hasAttr<CUDADeviceAttr>() ||
+            Global->hasAttr<CUDAConstantAttr>() ||
+            Global->hasAttr<CUDASharedAttr>()))
+        return;
+    }
+  }
+
   // Ignore declarations, they will be emitted on their first use.
   if (const FunctionDecl *FD = dyn_cast<FunctionDecl>(Global)) {
     // Forward declarations are emitted lazily on first use.
diff --git a/clang/lib/Frontend/CompilerInvocation.cpp b/clang/lib/Frontend/CompilerInvocation.cpp
index 4fb3ee62f04..b35bc65c7e7 100644
--- a/clang/lib/Frontend/CompilerInvocation.cpp
+++ b/clang/lib/Frontend/CompilerInvocation.cpp
@@ -183,6 +183,8 @@ static void CodeGenOptsToArgs(const CodeGenOptions &Opts,
     Res.push_back("-mcode-model");
     Res.push_back(Opts.CodeModel);
   }
+  if (Opts.CUDAIsDevice)
+    Res.push_back("-fcuda-is-device");
   if (!Opts.CXAAtExit)
     Res.push_back("-fno-use-cxa-atexit");
   if (Opts.CXXCtorDtorAliases)
@@ -1028,6 +1030,7 @@ static void ParseCodeGenArgs(CodeGenOptions &Opts, ArgList &Args, InputKind IK,
   Opts.ObjCAutoRefCountExceptions = Args.hasArg(OPT_fobjc_arc_exceptions);
   Opts.ObjCRuntimeHasARC = Args.hasArg(OPT_fobjc_runtime_has_arc);
   Opts.ObjCRuntimeHasTerminate = Args.hasArg(OPT_fobjc_runtime_has_terminate);
+  Opts.CUDAIsDevice = Args.hasArg(OPT_fcuda_is_device);
   Opts.CXAAtExit = !Args.hasArg(OPT_fno_use_cxa_atexit);
   Opts.CXXCtorDtorAliases = Args.hasArg(OPT_mconstructor_aliases);
   Opts.CodeModel = Args.getLastArgValue(OPT_mcode_model);