Integer AVX
Aug. 21st, 2010 05:27 pm![[personal profile]](https://www.dreamwidth.org/img/silk/identity/user.png){kind=small}
izardIt is a well known fact that AVX only supports floating point (single and double precision and scalar) operations in SNB u-arch CPUs family.
However some don't realize that all YMM register permutation functions and logical functions (XOR, ANDNOT, OR) can be used over integers, stored in 256 bit SIMD registers. This is enough to implements some of crypto routines, especially in bit-sliced case, in AVX, practically doubling the throughput.
However some don't realize that all YMM register permutation functions and logical functions (XOR, ANDNOT, OR) can be used over integers, stored in 256 bit SIMD registers. This is enough to implements some of crypto routines, especially in bit-sliced case, in AVX, practically doubling the throughput.
