[OpenMP] Customize CUDA-based tool chain selection

Summary: This patch provides a generic way of selecting CUDA based tool chains as host-device pairs.

Reviewers: arpith-jacob, caomhin, carlo.bertolli, ABataev, Hahnfeld, jlebar, hfinkel, tstellar

Reviewed By: Hahnfeld

Subscribers: rengolin, cfe-commits

Differential Revision: https://reviews.llvm.org/D29658

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@307271 91177308-0d34-0410-b5e6-96231b3b80d8

diff --git a/lib/Driver/Driver.cpp b/lib/Driver/Driver.cpp
index faced0697ed94ff1d6c97f1d49e43e3aba61d67b..8031d78a21e943ef64d1cfffdbb932a5b276fc74 100644 (file)
--- a/lib/Driver/Driver.cpp
+++ b/lib/Driver/Driver.cpp
@@ -572,8 +572,22 @@ void Driver::CreateOffloadingDeviceToolChains(Compilation &C,
           if (TT.getArch() == llvm::Triple::UnknownArch)
             Diag(clang::diag::err_drv_invalid_omp_target) << Val;
           else {
-            const ToolChain &TC = getToolChain(C.getInputArgs(), TT);
-            C.addOffloadDeviceToolChain(&TC, Action::OFK_OpenMP);
+            const ToolChain *TC;
+            // CUDA toolchains have to be selected differently. They pair host
+            // and device in their implementation.
+            if (TT.isNVPTX()) {
+              const ToolChain *HostTC =
+                  C.getSingleOffloadToolChain<Action::OFK_Host>();
+              assert(HostTC && "Host toolchain should be always defined.");
+              auto &CudaTC =
+                  ToolChains[TT.str() + "/" + HostTC->getTriple().str()];
+              if (!CudaTC)
+                CudaTC = llvm::make_unique<toolchains::CudaToolChain>(
+                    *this, TT, *HostTC, C.getInputArgs());
+              TC = CudaTC.get();
+            } else
+              TC = &getToolChain(C.getInputArgs(), TT);
+            C.addOffloadDeviceToolChain(TC, Action::OFK_OpenMP);
           }
         }
       } else