[FFmpeg-devel] [PATCH] Move cbrt tables to a separate cbrt_data(_fixed).c files.

[Reimar Döffinger]

On Sun, Mar 13, 2016 at 06:14:18PM +0100, Reimar Döffinger wrote:
> On Sun, Mar 13, 2016 at 01:12:57PM -0400, Ganesh Ajjanagadde wrote:
> > On Sun, Mar 13, 2016 at 12:49 PM, Reimar Döffinger
> > <Reimar.Doeffinger at gmx.de> wrote:
> > > On Sun, Mar 13, 2016 at 12:24:25PM -0400, Ganesh Ajjanagadde wrote:
> > >> On Sat, Mar 12, 2016 at 1:24 PM, Reimar Döffinger
> > >> <Reimar.Doeffinger at gmx.de> wrote:
> > >> >          for (i = 0; i < 1<<13; i++)
> > >> > -            cbrt_tab[i] = CBRT(cbrt_tab_dbl[i]);
> > >> > +            AAC_RENAME(ff_cbrt_tab)[i] = CBRT(cbrt_tab_dbl[i]);
> > >> >      }
> > >> >  }
> > >>
> > >> Note that cbrt_tab_dbl is really intended to be shared by both the
> > >> fixed/floating table inits. This was another thing my patch achieved:
> > >> only doing the more expensive double table init once across
> > >> float/fixed, and then doing the cheap conversion to uint32_t via
> > >> av_float2int or lrint(x*8192). Please change; it could go into a
> > >> separate patch if you prefer.
> > >
> > > IMHO you really need to argue why that would be desirable.
> > > The only case I can see this as useful is if someone
> > > runs at the same time fixed-point AAC decode and AAC encode,
> > > where it saves a bit of startup time.
> > > In all other cases you waste time and memory initializing
> > > a table that will never be used.
> > 
> > I don't understand the waste; the double init anyway needs to be done,
> > all tables are derived from it. This dates to an approach I did in
> > commit:  07a11ebcab9b31e9fc784029e5d24e6fbf486ff3.
> 
> The waste is in writing to the table that will never be used.
> As long as it is left as 0, it will not use memory, as soon
> as you touch it it uses up memory.
> But see my patch, if people think it's better it can be added.

And looking at that commit, I actually realize now that the
cbrt_tab_dbl will also permanently waste 64 kB of memory despite
never being used again.
Using malloc/free for it would potentially be better (though
in general I have some doubts about the wisdom of optimizing
for initialization time at the detriment of runtime performance
or memory usage).