Will you optimze your Code allready to get more out of future cpus. By going for FMA and AVX2. Or do newer compiler optimazisation options allready do that? I thing both features could be really helpfull FMA and the new upcoming 256bit support for integer vetors.
Well they'd have to include executables for people who don't have the latest top of the line intel cpus.
Afaik FMA is allready in use by AMD. FMA3 and FMA4 And nearly double the performance for Integervectors from one generation to the next is something that is maybe usefull.