jhuber6 added inline comments.
================ Comment at: clang/include/clang/Driver/Options.td:1056 +def libomptarget_nvptx_math_bc_path_EQ : Joined<["--"], "libomptarget-nvptx-math-bc-path=">, + Group<i_Group>, HelpText<"Path to libomptarget-nvptx math bitcode library">; def dD : Flag<["-"], "dD">, Group<d_Group>, Flags<[CC1Option]>, ---------------- JonChesterfield wrote: > jdoerfert wrote: > > FWIW, I don't believe this amdgpu/nvpxt separation is actually improving > > things, on the contrary. > libomptarget_gpu? Definitely simplifies clang which is a win. I think it's an > artefact of the previous device runtime layout. The device library has different triples so we don't get errors when linking, not sure if it's possible to compile to generic IR that doesn't care about the triple. And I thought the same but wasn't too concerned with addressing it in the off chance someone wants to compile for amdgpu and nvptx64 at the same time and also specify the libraries. Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D121468/new/ https://reviews.llvm.org/D121468 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
