Quoting James Almer (2015-07-15 19:26:08) > On 15/07/15 2:06 PM, Anton Khirnov wrote: > > Quoting James Almer (2015-07-15 18:26:28) > >> On 14/07/15 7:36 AM, Luca Barbato wrote: > >>> --- > >>> > >>> Properly rebased version > >>> > >>> libavcodec/hevc_ps.c | 258 > >>> ++++++++++++++++++++++++++------------------------- > >>> 1 file changed, 131 insertions(+), 127 deletions(-) > >>> > >>> diff --git a/libavcodec/hevc_ps.c b/libavcodec/hevc_ps.c > >>> index a5a2ace..2112265 100644 > >>> --- a/libavcodec/hevc_ps.c > >>> +++ b/libavcodec/hevc_ps.c > >>> @@ -1039,14 +1039,139 @@ static void hevc_pps_free(void *opaque, uint8_t > >>> *data) > >>> av_freep(&pps); > >>> } > >>> > >>> +static int setup_pps(AVCodecContext *avctx, GetBitContext *gb, > >> > >> If this is done only for readability's sake and not because this function > >> will be > >> used somewhere else, then making sure the compiler always inlines this > >> function may > >> be a good idea. > >> > > > > Why should that be important in this specific case? > > Function call overhead. ff_hevc_decode_nal_pps() is not an init function, it > may > or may not be called for every decoded frame.
If it gets called every frame, then the function call overhead is the least of our worries. Worrying about such microoptimizations is IMO counterproductive, unless there's a good reason to think they actually have a measurable effect. -- Anton Khirnov _______________________________________________ libav-devel mailing list [email protected] https://lists.libav.org/mailman/listinfo/libav-devel
