[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

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@331938 91177308-0d34-0410-b5e6-96231b3b80d8

diff --git a/include/clang/Basic/TargetOptions.h b/include/clang/Basic/TargetOptions.h
index cbe0953e74376046e57b4ceff078548afbe42116..31a67428791547600e20fd8b8e9cee929c4cc614 100644 (file)
--- a/include/clang/Basic/TargetOptions.h
+++ b/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

diff --git a/include/clang/Driver/Options.td b/include/clang/Driver/Options.td
index 52f2e34621dd5e3a2788b97190f1e5322bacb995..95ae9625d3a89b09c785c3a62982b7d975f78d27 100644 (file)
--- a/include/clang/Driver/Options.td
+++ b/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">;

diff --git a/lib/Basic/Targets/NVPTX.cpp b/lib/Basic/Targets/NVPTX.cpp
index 04293b3ce10caf34f02a8f41a38e4109232e30d3..fd4ee1606061194decfa5aad895e50f088b4696b 100644 (file)
--- a/lib/Basic/Targets/NVPTX.cpp
+++ b/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");
 

diff --git a/lib/Driver/ToolChains/Clang.cpp b/lib/Driver/ToolChains/Clang.cpp
index 4165afcfa9ca14575d7d4acad2dd14dfc603bc1e..cc26b5e04ec20cfee7b6b1c18c84fe5f44a87a01 100644 (file)
--- a/lib/Driver/ToolChains/Clang.cpp
+++ b/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

diff --git a/lib/Driver/ToolChains/Cuda.cpp b/lib/Driver/ToolChains/Cuda.cpp
index 3978cb27429bd6241a7873030f77b7b441d0e8d6..5916ad033c22acf14c4e9b028dccf5b4cc6c5625 100644 (file)
--- a/lib/Driver/ToolChains/Cuda.cpp
+++ b/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;

diff --git a/lib/Frontend/CompilerInvocation.cpp b/lib/Frontend/CompilerInvocation.cpp
index 71c946442d7cf2ba1989dc047b73063940b662c0..05e5196e32d26fc40c58314972c01fdfc62fceb9 100644 (file)
--- a/lib/Frontend/CompilerInvocation.cpp
+++ b/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,