isuckatcs wrote: > Consider downstream users that might use this reporting system and have their > own trackers. (We don't at Sonar, but pretend), then they would need to > remove one more unjust assert.
This is also a fair point, though as far as I know, LLVM doesn't keep API compatibility either between releases, so downstream users should expect that their code will need adjustments. > Now we could save the cost for building a call stack, a comparison, a > conditional jump based on the result of the previous comparison and also the > deconstruction of the stack. Actually a release build compiled with `-O2` or `-O3` should have this function inlined already, so the stack frame is not created. As for the others, I'm not sure if that will be a significant or even a measurable performance improvement. When it comes to speeding up execution, the rule of thumb is to always measure it though. [D131707](https://reviews.llvm.org/D131707) is/was a patch, where improving runtime speed came up as a discussion too. It contains some points about performance bottlenecks in the analyzer, tips on how to get started with profiling, etc. If you're interested in such stuff, it might worth taking a look at. https://github.com/llvm/llvm-project/pull/75076 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
