added cpu optimization detection and avx optimization

[manuvi]

this is a proposal for avx optimization on many float loop computation inside DSP.cpp

[manuvi]

I guess, if this could be interesting, to refactor the makefiles to consider a simd extension, to enable id by pragma inside code and build an infrastructure for the various platform

[michaelwillis]

This change breaks the build for all processor architectures except for x86_64. It also obfuscates the code in the DSP.cpp files for fairly minimal performance improvement - the vast majority of the computation being done in these plugins happens in the freeverb3 code.

It would be better to allow the compiler to automatically apply such performance improvements, as described here: https://gcc.gnu.org/projects/tree-ssa/vectorization.html

[manuvi]

as far as I know, optimization has to be specified for every arch (like ffmpeg does) but I understand your pov (these opt are really rare to catch by compiler, you have to write a specific loop structures with specific loop increment...btv I don't see these structures).
It would be a good thing to open a separate branch and start to optimize all the possible flows, including all the makefiles, for every arch.
With this only optimization I got a speedup from 4% for one room-rev of DSP load (on Ardour) to 1%. Zen2 12core.
I know that I break lot of stuff, but my target was to open a branch and try to get the best.
However, thank you.