zeyi2 wrote:

> Well that's a bit weird.

Yeah, that does look a bit odd.

My guess is that this is more likely benchmark noise than something caused by 
this change.

A few reasons:

- The two benchmark runs were not done back-to-back under the exact same 
machine state.
- The upstream revision also differed between the two runs. I reused the data 
from https://github.com/llvm/llvm-project/pull/204045#issuecomment-4744770346.
- When `-lifetime-safety-lifetimebound-macro=` is set, this path uses the 
spelling from `LangOpts` and skips macro lookup entirely. So a small slowdown 
in this row does not seem like something that would be explained by the cache 
lookup change.


https://github.com/llvm/llvm-project/pull/205250
_______________________________________________
cfe-commits mailing list
[email protected]
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits

Reply via email to