Remove -ffast-math. (Better to focus on correctness by default, and only sacrifice it for performance in limited profiler-indicated areas.)

This was SVN commit r7246.
This commit is contained in:
Ykkrosh 2010-01-05 20:54:33 +00:00
parent c9a6af2134
commit c8fed35c52

View File

@ -190,10 +190,6 @@ function package_set_build_flags()
-- enable SSE intrinsics
"-msse",
-- speed up math functions by inlining. warning: this may result in
-- non-IEEE-conformant results, but haven't noticed any trouble so far.
"-ffast-math",
-- don't omit frame pointers (for now), because performance will be impacted
-- negatively by the way this breaks profilers more than it will be impacted
-- positively by the optimisation