upvote
And that's why musl's allocator's performance isn't a huge concern for ripgrep
reply
That's to be properly measured. Assuming that the allocator will not cause issues there is to be proven.

The application I recently improved by changing the allocator had a similar profile (a C++ include scanner) and thread parallel I/O functions had terrible performance originally with mallocng. Adding threads almost had negative value because of the contention.

reply