[libcxx-commits] [libcxx] [libc++] Optimize std::min_element (PR #100616)

[Nikolas Klauser via libcxx-commits]

https://github.com/philnik777 commented:

Overall, I'd really like to compare this to an implementation which goes over N elements in parallel, saving the offset to a vector register and reducing only at the end. I think that would avoid a bunch of conditionals and we wouldn't have to search the range twice. Actually, is the current implementation standards-conforming? We're invoking projections and comparators a lot more than previously I think.

https://github.com/llvm/llvm-project/pull/100616