dblaikie added a comment. In D143877#4156769 <https://reviews.llvm.org/D143877#4156769>, @dblaikie wrote:
>> Although we may not agree such ideas, we should offer an option for the >> users to give them the right to choose. > > I think @iains was bemoaning the large number of module flags recently - > perhaps he'd have some thoughts on this, but mine are that we should be more > prescriptive - figure out what a good solution is for most users and > implement that, rather than giving people lots of knobs to tune. > > For this issue - honestly I think it might be a better/safer default to not > import external module definitions at all/ever. Leave it to LTO to do that > kind of cross-module optimization, as we have in the past/non-modular builds. > > At least until/unless we implement something more like what the bug describes > - doing IR level optimizations first and importing the optimized IR, rather > than reoptimizing these external definitions in every user. Huh, this ended up on quite the wrong review... fascinating. Sorry about that. Disregard. I'll follow up with on-topic review feedback here shortly. Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D143877/new/ https://reviews.llvm.org/D143877 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
