Re: [fluid-dev] A question of memory

[Dave Serls]

On Tue, 21 Nov 2006 19:45:57 +0000
Paul Millar <address@hidden> wrote:

> Hi all,
> 
> I've written (what I hope is) a minimal, functioning test code for running 
> fluidsynth, linking against libfluidsynth.  Its available from:
> 
> http://ppewww.physics.gla.ac.uk/~paul/fluidsynth/simple-fluidsynth-test.c
> 
> Running this test test through valgrind's memcheck tool is quite interesting:
> 
> Valgrind reports that the alsa driver is calling libasound with some 
> uninitialised bytes in a few places.
> 
> Also, after calling delete_fluid_* functions, there's still some 33 MB of 
> memory in use when the program terminates.
> 
> I'm loading a 30 MB SoundFont file, so I'm guessing that delete_fluid_synth() 
> doesn't free memory allocated by fluid_synth_sfload().  There's lots of other 
> wee bits of leaked memory here and there (strdup()s, etc...).

  When mlock=1 is enabled for realtime kernels, that memory is pinned 
(page-fixed)
  and very large soundfonts combined with multiple engines (via qsynth) can
  deplete real memory into the OOM-killer range.  
  I'm running a poor patch to qsynth to re-use soundfonts across synth engines.
  It's not correct, but it prevents disaster.
> 
> The output from valgrind is available from:
> http://ppewww.physics.gla.ac.uk/~paul/fluidsynth/valgrind.out
> 
> But, I hope the test-case above is pretty easy to check this.
> 
> Cheers,
> 
> Paul.


-- 
************************************************************************
*   Dave Serls                                 Littleton, CO, USA      *
*   dashs.denver.co.us                         dashs.com               *
************************************************************************