jhuber6 wrote: > It seems like this patch is not enough for a case, when we have multiple > offload-arch (one of which is SPIR-V) and -no-lto is passed - in this case > -no-lto will leak to SPIR-V compilation breaking it. > > I have 2 drafts for the fix: > > 1. Insert the !Triple.isSPIRV() in 2 more places to force ignore -no-lto; > > 2. Do something different actually by introducing an overload to -no-lto > option, which would also accept a string argument to specify for which > architecture apply it (the guards will be still needed, but the solution > would look a bit more pleasent). > > > While option 1 is easy and have a small blast, I personally lean towards > option 2. WDYT? > > @jhuber6 @AlexVlx @mgcarrasco @sarnex
Hm, truthfully this should be handled by something like `-Xarch_device -flto` and -foffload-lto should've never existed. https://github.com/llvm/llvm-project/pull/202986 _______________________________________________ cfe-commits mailing list [email protected] https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
