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

I think part of the challenge is to not exceed the word size, 16 bits in this case.


that's useful with explicit simd instructions which let you process 32 (pairs of corresponding) 16-bit pixels with one avx-512 instruction, or on a 16-bit machine like the 80286, but unnecessary if you're processing a single pair of 16-bit pixels with 32-bit instructions. on the other hand, you might have 32-bit vectors or something, for example for the standard axis-aligned bounding box test




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

Search: