banach-space wrote: > Where do you see the need for an RFC?
Either an RFC or a PSA would be welcome and help me understand the context. > The maintainers are all in agreement that this de-duplication is > wanted/appreciated, so there is no need for it as far as I know. If there was a public discussion where this was agreed on and I missed it, then apologies. That said, my understanding is that the role of maintainers in LLVM is primarily to help the community reach consensus, rather than for agreement among maintainers to substitute for broader community discussion: * https://llvm.org/docs/DeveloperPolicy.html#maintainers > > I am under the impression that some patches in this stack are actually > > independent? > > This definitely seems to be the case. > > > I'm not 100% sure that a generic CodeGenUtils.cpp file is the best > > approach. Ultimately, folks working on AMD GPU will not care about, e.g., > > Neon builtins, and vice versa. I'd be tempted to keep target-specific > > utilities separate and use CodeGenUtils.cpp only for hooks shared across > > targets. > > Haven't most of the stack split these up? I originally did the 'generic' one > when I did the first patch for these, but Henrich has continued to split > these up in what apperas to be a reasonable fashion. Which of my points are you referring to? * Re independent patches, this PR and https://github.com/llvm/llvm-project/pull/227260 look unrelated to me. * Re moving target-specific code to a generic file, https://github.com/llvm/llvm-project/pull/227260 moves code from `ARM.cpp` (target file) to `TargetUtils.cpp` (generic file) In case this wasn't clear from my initial message, I have no objection to the overall direction and appreciate the clean-up. I'm mostly asking for some broader context and trying to understand how the individual patches fit together. https://github.com/llvm/llvm-project/pull/227256 _______________________________________________ cfe-commits mailing list [email protected] https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
