Quote:
Originally Posted by Romario
How you optimised these LAME builds ? With ICL15 or newest MSCS 2017 ?
While ICL work always much better on Intel processors.
|
ICL, whatever version was up-to-date when I last built LAME.
Note that ICL has an option to add an "optimized" code path (for all compatible processors, including AMD) and a separate option to "optimize" specifically for Intel processors (potentially breaking the binary for Non-Intel processors).
I always use the former
But, as said before, those
compiler optimizations have a lesser effect! All the "hot" code, where ~99% of the CPU time is spent, is hand-optimized assembler code that will
not be effected by compiler optimizations at all.