nikic wrote: > It seems better to associate the target-cpu and target-features with the asm > itself so that both the ModuleSymbolTable and the AsmPrinter know the correct > features.
That would make sense to me, though that probably requires not treating the module level asm as monolithic (i.e. allow different parts to have different target features). I think the idea behind this PR is that, from an architectural perspective, it shouldn't be necessary to parse inline assembly to fetch the symbol table of an object in the first place. If the goal is just to solve the original problem, then explicitly storing the features would probably be both simpler and more thorough. https://github.com/llvm/llvm-project/pull/174995 _______________________________________________ cfe-commits mailing list [email protected] https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
