undefined | Better HN

0 pointssifar1y ago0 comments

One can use this techniques to optimize the multiplier inside the FP FMA unit. However this cannot be used to multiply two floating point numbers as FP arithmetic is not associative.

0 comments

saagarjha1y ago

I'm pretty sure a lot of people doing matrix multiplications are ok with non-associative floating point arithmetic?

j / k navigate · click thread line to collapse

0 comments

saagarjha1y ago

I'm pretty sure a lot of people doing matrix multiplications are ok with non-associative floating point arithmetic?

j / k navigate · click thread line to collapse