Hacker News new | past | comments | ask | show | jobs | submit login

I experimented with the proposed parallel data type extensions to the C++ standard library. I got impressive performance gains for calculating APFS fletcher checksums without resorting to compiler intrinsics or inline assembly.

Gains were even more impressive when adding some simple loop unrolling: https://jtsylve.blog/post/2022/12/24/Blazingly-Fast-er-SIMD-...




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: