[CUDA] Added -f[no-]cuda-short-ptr option

The option enables use of 32-bit pointers for accessing
const/local/shared memory. The feature is disabled by default.

Differential Revision: https://reviews.llvm.org/D46148

llvm-svn: 331938

clang/include/clang/Basic/TargetOptions.h

@@ -63,6 +63,10 @@ public:
 
   /// If given, enables support for __int128_t and __uint128_t types.
   bool ForceEnableInt128 = false;
+
+  /// \brief If enabled, use 32-bit pointers for accessing const/local/shared
+  /// address space.
+  bool NVPTXUseShortPointers = false;
 };
 
 }  // end namespace clang

clang/include/clang/Driver/Options.td

@@ -581,6 +581,9 @@ def fno_cuda_approx_transcendentals : Flag<["-"], "fno-cuda-approx-transcendenta
 def fcuda_rdc : Flag<["-"], "fcuda-rdc">, Flags<[CC1Option]>,
   HelpText<"Generate relocatable device code, also known as separate compilation mode.">;
 def fno_cuda_rdc : Flag<["-"], "fno-cuda-rdc">;
+def fcuda_short_ptr : Flag<["-"], "fcuda-short-ptr">, Flags<[CC1Option]>,
+  HelpText<"Use 32-bit pointers for accessing const/local/shared address spaces.">;
+def fno_cuda_short_ptr : Flag<["-"], "fno-cuda-short-ptr">;
 def dA : Flag<["-"], "dA">, Group<d_Group>;
 def dD : Flag<["-"], "dD">, Group<d_Group>, Flags<[CC1Option]>,
   HelpText<"Print macro definitions in -E mode in addition to normal output">;

clang/lib/Basic/Targets/NVPTX.cpp

@@ -68,6 +68,9 @@ NVPTXTargetInfo::NVPTXTargetInfo(const llvm::Triple &Triple,
 
   if (TargetPointerWidth == 32)
     resetDataLayout("e-p:32:32-i64:64-i128:128-v16:16-v32:32-n16:32:64");
+  else if (Opts.NVPTXUseShortPointers)
+    resetDataLayout(
+        "e-p3:32:32-p4:32:32-p5:32:32-i64:64-i128:128-v16:16-v32:32-n16:32:64");
   else
     resetDataLayout("e-i64:64-i128:128-v16:16-v32:32-n16:32:64");
 

clang/lib/Driver/ToolChains/Clang.cpp

@@ -4714,6 +4714,9 @@ void Clang::ConstructJob(Compilation &C, const JobAction &JA,
 
     if (Args.hasFlag(options::OPT_fcuda_rdc, options::OPT_fno_cuda_rdc, false))
       CmdArgs.push_back("-fcuda-rdc");
+    if (Args.hasFlag(options::OPT_fcuda_short_ptr,
+                     options::OPT_fno_cuda_short_ptr, false))
+      CmdArgs.push_back("-fcuda-short-ptr");
   }
 
   // OpenMP offloading device jobs take the argument -fopenmp-host-ir-file-path

clang/lib/Driver/ToolChains/Cuda.cpp

@@ -635,8 +635,10 @@ void CudaToolChain::addClangTargetOptions(
     // CUDA-9.0 uses new instructions that are only available in PTX6.0+
     PtxFeature = "+ptx60";
   }
-  CC1Args.push_back("-target-feature");
-  CC1Args.push_back(PtxFeature);
+  CC1Args.append({"-target-feature", PtxFeature});
+  if (DriverArgs.hasFlag(options::OPT_fcuda_short_ptr,
+                         options::OPT_fno_cuda_short_ptr, false))
+    CC1Args.append({"-mllvm", "--nvptx-short-ptr"});
 
   if (DeviceOffloadingKind == Action::OFK_OpenMP) {
     SmallVector<StringRef, 8> LibraryPaths;

clang/lib/Frontend/CompilerInvocation.cpp

@@ -2922,6 +2922,8 @@ static void ParseTargetArgs(TargetOptions &Opts, ArgList &Args,
     Opts.Triple = llvm::sys::getDefaultTargetTriple();
   Opts.OpenCLExtensionsAsWritten = Args.getAllArgValues(OPT_cl_ext_EQ);
   Opts.ForceEnableInt128 = Args.hasArg(OPT_fforce_enable_int128);
+  Opts.NVPTXUseShortPointers = Args.hasFlag(
+      options::OPT_fcuda_short_ptr, options::OPT_fno_cuda_short_ptr, false);
 }
 
 bool CompilerInvocation::CreateFromArgs(CompilerInvocation &Res,