jhuber6 wrote: > > but we now assume that if the user specified --offload-arch= on the link > > job, they definitely want that architecture to be used if it exists. > > That would be my assumption, too. Do we currently just ignore > `--offload-arch=` for the linking phase? > > With the patch, what's expected to happen when link-time `--offload-arch=` > specifies an architecture that the libraries were not built for and therefore > have no corresponding blobs to extract. I assume it would be diagnosed. If > that's the case, we may want to add a test case for that.
Right now we infer the ones to use from the contents of the `.o` files. This is more of a hint I guess? It's not a fatal error but I suppose it could be warning. https://github.com/llvm/llvm-project/pull/147823 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits