asb added a comment. Herald added a subscriber: luke. In D141666#4050692 <https://reviews.llvm.org/D141666#4050692>, @eklepilkina wrote:
> We are interested in proper support for `Zicsr` and `Zifencei`. Could someone > look at patch or explain why these extensions were included by default? As Jessica says, the reason they're not split out is that when the backend was first implemented, they were part of the I specification. I'm not at all a fan of how this happened, but we are where we are, and as noted GCC has made this change. The key aspect this patch misses right now is the versioning issue - I is still marked as 2.0. So really, this needs someone to propose a what the change will look like in full. A few misc thoughts (and all of these would need some more discussion with others active in the RISC-V backend). e.g. do we need/want -misa-version, or is there a proposal to just move to I 2.1 by default? It might also be interesting to explore options that temporarly sidestep the versioning discussion and improve compatibility with GCC -march strings. e.g. an initial patch that adds zicsr and zifencei but enables it by default (or otherwise handles zicsr and zifencei as a no-op). I _think_ I'd be happy with something along those lines, but I can't speak for others involved in the RISC-V backend. Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D141666/new/ https://reviews.llvm.org/D141666 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
