Jan Wassenberg

> Is it uncommon for these instructions to be used? 32 bit [u]intfloat conversions are indeed common. Unfortunately, I doubt many developers are aware that signed is cheaper than unsigned...

@rrwinterton Nice, lots of goodies in AVX3, thanks for sharing :) Good idea, I'd be happy to help with testing AVX2 and 3 benefits using JPEG XL.

FYI AVX-512 and NEON have fabs() instructions.

+1 for these being useful, the math library being built for Highway requires these.

@Maratyszcza Interesting. Do I understand correctly that an engine would have to check whether the shift count matches the magic value, and then emit the corresponding code? It seems that...

@Maratyszcza @dtig @tlively I understand, thanks for explaining. Documentation + internal optimization in engines sounds good!

I guess the bit-select semantics were chosen because ARM does it that way? Sure would be nice if this mapped to a single blendv instruction on x86. Have we considered...

@tlively Right. I agree validating inside bitselect would be expensive. Is undefined behavior generally unacceptable? If so, we could define a Mask type guaranteed by the compiler to have that...

@abrown Yes, that's what I was thinking. Valid could mean all_true((mask == 0) | mask). @tlively I totally understand that this would be a large change with relatively low ROI....

Thanks @abrown :) I suppose a type could also be useful for swizzle() inputs that are known to be >= 0x80, which would allow using PSHUFB directly. Probably small potatoes,...