You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I've been enjoying using hermetic_cc_toolchain for the last year, but was exploring some changes to LLVM, so set up my Bazel project to use it directly as an alternative, and discovered along the way that my iteration cycle on a large test was much faster.
e.g. if I change one file, which forces a re-link:
Straight clang (that I built from source) takes ~65 seconds
hermetic_cc_toolchain takes ~110 seconds.
Thus clang is running ~40% faster. That being said, I'm using clang 18, and, while I'm not certain, zig seems to be using clang 17.
So I'm happy to investigate further (e.g. try older clang), or try to create a reproducible example, but was first curious if this was known or expected.
Thanks!
The text was updated successfully, but these errors were encountered:
Ah, this was it. It took my smaller test (which is easier to see the difference in build times since there is less variance) from ~30 seconds -> ~6 seconds. Thanks!
Perhaps it would be worth having it export a warning to Bazel in these cases so it is easier to notice.
Hi,
I've been enjoying using hermetic_cc_toolchain for the last year, but was exploring some changes to LLVM, so set up my Bazel project to use it directly as an alternative, and discovered along the way that my iteration cycle on a large test was much faster.
e.g. if I change one file, which forces a re-link:
Thus clang is running ~40% faster. That being said, I'm using clang 18, and, while I'm not certain, zig seems to be using clang 17.
So I'm happy to investigate further (e.g. try older clang), or try to create a reproducible example, but was first curious if this was known or expected.
Thanks!
The text was updated successfully, but these errors were encountered: