Intel don't have vpintersect. Its micro-coded dog slow but its such an obscure operation and of little use if any.
I'd rather AMD re-invest the hardware into speeding up useful operations, shave 1 cycle off divide or square root or something.
14 karma · joined May 16, 2023
I guess I'm good then.
The bottleneck are the 3 conditionals: - positive or negative - positive or negative exponent, x > 10.0 - correction for 1.xxxxx * 2^Y => fract(log10(2^Y)) 1.xxxxxxxx > 10.0
Where's a proper comparison between the fastest linux OS "CachyOS" and Windows 11 LTSC with all the crap removed, particularly the biggest offenders: Defender, Update, Edge, "fullscreen optimizations" and all the weird little services from print spooler to NPSV_xxx "now playing service" ?
lzcnt ecx,ecx
mov eax, 00100 0011 0010 0000 0001b shl ecx,2 shrx eax,eax,ecx and eax,15
Struggling proc \n lzcnt ecx,ecx \n test cl,cl \n setz al \n lea edx,[ecx-2] ;[0,1,2] \n cmp rdx,2 \n cmovle eax,ecx \n ret \n Struggling endp