Critics of Algorithm D about libdivide HOT 6 CLOSED

LOL, you even reported a bug in clang. Bravo. D:)

from libdivide.

Man what an asshole.

from libdivide.

Man what an asshole.

It is not my article, I am russian.

from libdivide.

Yes, understood, thank you for filing! I was referring to the blog author.

from libdivide.

Gonna publish my new algorithm tomorrow or perhaps later tonight!

from libdivide.

BTW, I read you newest article in your blog... If you use g++ instead of clang++ the divq algorithm will a little faster. Even on my nonrefreshed Haswell divq is the fastest, so apparently saying Ice Lake's Xeon is needed is wrong (unless there is some kind of checks needed for divq 128 bit /64). Also you can see that algorithms except for divq are all better on clang++ and even better (yep) when you do -Os in clang++, optimisation for size, in that case the algorithm in libdivide is better than branched version even. Of course -march=native also slightly tunes it.

from libdivide.