07-18-2015, 08:16 AM
(07-17-2015, 07:11 PM)Nobbs66 Wrote: [ -> ]AVX is slower in HW. AVX2 is about the same as SSE4.1 in HW and faster in SW.
well, that actually would be dependent on the Implementation. both of them are actually quite similar with AVX just offering instruction sets for YMM registers. therefore making computation of precision floating point numbers easily. also, AVX supports three operands so there really isn't any theoretical reason to say that SSE4.1 is faster.