On Thu, Nov 26, 2015 at 8:50 AM, Hendrik Leppkes <h.lepp...@gmail.com> wrote: > On Thu, Nov 26, 2015 at 2:10 PM, Ganesh Ajjanagadde <gajja...@mit.edu> wrote: >> On Thu, Nov 26, 2015 at 2:49 AM, Hendrik Leppkes <h.lepp...@gmail.com> wrote: >>> On Thu, Nov 26, 2015 at 6:04 AM, Ganesh Ajjanagadde >>> <gajjanaga...@gmail.com> wrote: >>>> This is needed in order to obtain what is available for hardcoded >>>> table generation. >>>> >>>> A minimal avutil/host_libm.h is also created. >>>> >>> >>> I'm really not a fan of this change. configure is slow and complex >>> enough as it is, and host tools don't need to be micro-optimized, >>> since they are never run by a user. >> >> 2 comments: >> 1. I deliberately kept the host_math_func list very short because I >> recognized this concern regarding configure's slowness. >> 2. They are run by a lot of users, since tablegen is done at runtime >> by default. "Never run by a user" is thus incorrect. It is not a >> "micro-optimization", pow is often 4x slower than relevant function. >> > > These things are for host code, that is not run by users.
I see what you mean now, alternative solution by Ronald is far superior. Patch dropped, thanks all. > > - Hendrik > _______________________________________________ > ffmpeg-devel mailing list > ffmpeg-devel@ffmpeg.org > http://ffmpeg.org/mailman/listinfo/ffmpeg-devel _______________________________________________ ffmpeg-devel mailing list ffmpeg-devel@ffmpeg.org http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
