boomanaiden154 wrote: > Is the main difference that we use an explicit and dedicated cache now? Also, > how did you decide on the 2GB limit? For comparison, my build dir with mlir + > spirv tools in release_with_deb_info with asserts is 39GB.
It is an explicit cache now, but I don't think that's a major difference. The cache hits on workflows that run rarely like this one are pretty low (at least anecdotally, I don't have hard numbers) given we're using way over the storage limit (10GB) for all caches. Github takes a while to evict them though, so sometimes they can stick around. This is a release build, so the object file size is way smaller. I usually see <1GB for just building LLVM. I don't think MLIR adds that much more. This is intended to mostly be a NFC change. The main reason to do this is mostly to clean up some technical debt, namely `llvm-project-tests.yml`. It handles way more cases than what it is being used for in tree, so removing it will reduce quite a bit of complexity in my opinion. https://github.com/llvm/llvm-project/pull/153871 _______________________________________________ llvm-branch-commits mailing list llvm-branch-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-branch-commits
