https://github.com/NagyDonat commented:
Thanks for the updates! I added some inline remarks, but overall I'm satisfied with the state of this PR and I hope that it can be merged soon. However, I became a bit uncertain about the plans for follow-up changes, because it seems that there are several locations where the difference between "inherently in `<memory space X>`" and "originally in UnknownSpace, but we deduced that it's in `<memory space X>`" is significant. (I'm not entirely certain about this, because e.g. the schematics and applications of "deducing a memory space" are not yet fixed. This might deserve some additional thinking.) Previously I strongly supported the idea that in the future we should store the `MemSpace` in a single location, as a state trait -- but if the inherent/acquired difference turns out to be relevant, then perhaps we should "stop after this PR" and stabilize this "hybrid representation" of memory spaces. https://github.com/llvm/llvm-project/pull/123003 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
