3 ms·
While you’re at it, allow for a few more ulp (units of least place) error for floating point ops.
by basementcat 2y ago
While you’re at it, allow for a few more ulp (units of least place) error for floating point ops.
- basementcat 2y agoKidding aside, my understanding is this sort of thing cannot be microcode patched. But I would be pleased to be proven wrong.
- Lockal 2y agoIt can be patched, see example in "constant-time hardware division" section in https://misc0110.net/files/cpu_woot23.pdf https://misc0110.net/files/cpu_woot23.pdf (code: https://github.com/pietroborrello/CustomProcessingUnit/blob/master/bios/ucode_patches/ctdiv.u https://github.com/pietroborrello/CustomProcessingUnit/blob/...). You probably won't observe any improvement unless you know an algo massively better than reciprocals+newton.