Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

Yes, I'm aware of the math, but there are other aspects like 'does this form of the code help auto-vectorisation?', 'does it work across both gcc and clang?', 'what about optimisation levels', 'does my hand-rolled version beat the compiler's optimisation of the naive code?'. Etc. It's a lot of fun.


Oh. I'm skipping all of the fun by using SIMD intrinsics, I guess.

You could go to https://highload.fun/ to waste many evenings on this sort of thing.




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

Search: