Again: Most audio encoders already contain hand-optimized assembler code for the "hot" functions, i.e. the functions where like ~99% of the work is done.
If a particular audio encoder contains optimized assembler code for your CPU, then the encoder will
automatically enable/use the optimized assembler code on your system – that is regardless of what compiler or compiler settings were used to compile the plain C/C++ code. If, on the other hand, a particular audio encoder does
not contain optimized assembler code for your CPU, then there is
nothing I could do about that. You'd have to ask the developer of the particular encoder
All that
I can do is change the compiler settings that are used to compile the plain C/C++ code. But, as pointed out several times, this has a lesser effect on the overall performance, compared to the assembler stuff...
(That plus: I already create "generic", "sse2" and "avx" builds of the encoders and LameXP will invoke the appropriate one on each system)