[MPlayer-dev-eng] mplayer's size

[Lauri Kasanen]

On Sat, 13 Feb 2016 16:41:28 +0100
Reimar Döffinger <Reimar.Doeffinger at gmx.de> wrote:

> On Sat, Feb 13, 2016 at 04:58:22PM +0200, Lauri Kasanen wrote:
> > My ticket about mplayer's excessive size increase was closed as
> > invalid. However, a 18mb binary is in no way acceptable, it results in
> > a startup time of several seconds with cold caches. Mplayer used to be
> > 13mb in 2013, and 8mb in 2011.
> 
> Uh, if you mind the size you should maybe link against system
> FFmpeg though, as that is most of the size.
> Admittedly that is not necessarily the most recommended configuration,
> but that would be most effective.

I have no system ffmpeg in this case, and I care about startup speed
mainly. Using a shared ffmpeg may or may not help that.

> > - the unnecessary encoding functions are present in mplayer
> 
> That is most likely an incorrect assumption.
> There are among other things video filters using encoding functions.
> It is possible that some are indeed not necessary, or not necessary
> with certain configurations, but in general they are in fact needed.

Mplayer will never need to encode H.264, ROQ, AAC, Sorenson, etc ;)

0x000000000003853 t encode_frame
0x000000000003a29 t aac_encode_frame
0x00000000000415a t ff_mpv_encode_picture
0x000000000004162 t g723_1_encode_frame
0x000000000004fb7 t encode_mb_hq.isra.6
0x000000000005405 t wavpack_encode_frame
0x0000000000054d4 t roq_encode_frame
0x000000000005f6b t encode_frame
0x000000000006d4e t svq1_encode_plane
0x000000000007bd9 t encode_thread

The non-compressed audio encoding options and some image filter options
are necessary like you say, but most of them are not. They're included
wholesale.

> > Most of rela.dyn seems to be from variables, likely from .data.
> 
> Large tables with pointers.
> Not compiling as PIE should avoid most of them I'd guess.

Thanks, I'll add --disable-relocatable to my configure args.

- Lauri