jhuber6 added a comment. In D150998#4403444 <https://reviews.llvm.org/D150998#4403444>, @yaxunl wrote:
> what happens to > > clang -xhip a.c --offload-arch=gfx90a:xnack+ --offload-arch=gfx90a:xnack- > --offload-new-driver -fgpu-rdc > clang -xhip b.c --offload-arch=gfx90a:xnack+ --offload-arch=gfx90a:xnack- > --offload-new-driver -fgpu-rdc > clang --offload-link a.o b.o > > Basically gfx90a:xnack+ and gfx90a:xnack- need to be treated as distinct GPU > arch's and the fat binary should contain different code objects for them. In this logic they would both map to `gfx90a` but be given distinct images and then be linked together into a single one with `gfx90a` metadata. This is the part that I'm saying should probably be a follow-up fix to handle this correctly in the linker wrapper. Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D150998/new/ https://reviews.llvm.org/D150998 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
