On Sun, Dec 27, 2015 at 5:38 AM, Ronald S. Bultje <[email protected]> wrote: > Hi, > > On Thu, Dec 24, 2015 at 1:32 PM, Ganesh Ajjanagadde <[email protected]> > wrote: >> >> In the standard library, these are functions. We should match it; there >> is no reason for these to be macros. >> >> While at it, add some trivial comments for readability and correct an >> incorrect (at standard double precision) M_LN2 constant used in the exp2 >> fallback. > > > For bisect purposes, the M_LN2 change should be a separate patch IMO. I > don't have objections to that one.
The ln2 change went in already: 5630ed5be64fef3fd70cb93a7623d46afa0c83e6 with some very minor stuff (comment additions), > > As for the rest, I'm not against it, but this stuff is extremely brittle so > please test it extremely carefully on various configs. Yes, I have dropped the rest on my end since: 1. I lack an array of configs (only GNU/Linux+clang/gcc, Mac OS X with difficulty). 2. I am not knowledgable on this aspect. 3. It should be done slowly IMHO to minimize scope of regressions. For a start, maybe focusing on the float variants (expf, etc) may be useful. > > Ronald _______________________________________________ ffmpeg-devel mailing list [email protected] http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
