[MPlayer-cvslog] CVS: main/libmpcodecs vf_noise.c,1.17,1.18

Rich Felker dalias at aerifal.cx
Tue Apr 4 19:17:49 CEST 2006


On Tue, Apr 04, 2006 at 02:56:28PM +0300, Ivan Kalvachev wrote:
> 2006/4/4, Alex Beregszaszi <alex at fsn.hu>:
> > Hi,
> >
> > > > > -static int nonTempRandShift[MAX_RES]= {-1};
> > > > > +static int nonTempRandShift_init;
> > > >
> > > > For case of safety set this to zero. I know in theory everything is
> > > > perfect...
> > >
> > > No. All uninitialized data is zero. This is guaranteed by C and true
> > > on every single implementation. If you don't know C that's your
> > > problem. Explicit initialization just adds 4 bytes to the data segment
> > > for no reason.
> >
> > HAHAHAHA
> >
> > Rich, you are getting more crazy with every day. Do you think that 4
> > bytes hurt? MPlayer is such a bloatware that even your size reduction
> > patches cant help in that matter. Better design libavmunge instead
> > trolling. </trolling>
> 
> Maybe I am blind but I think that Rich actually added 4 bytes (`int
> foo_init` ) , not removed them.

Added 4 bytes to bss, not data. This does not increase the binary
size.

> afaik static arrays are filled at compiletime and are stored as they
> are in the binnary. (not sure if any compiler optimizes them out).

Only initialized data is stored in the binary.

> I would prefer dynamic allocation per filter. (no static arrays at all)

Nonsense. Why make multiple copies of constant data?

Rich




More information about the MPlayer-cvslog mailing list