[Tremor] Re: Reducing code size

Ripinder Singh ripinder.singh at gmail.com
Wed Sep 21 21:25:49 PDT 2005


Try removing the floor0.c functions from internal memory, placing all
those functions explicitly in to external mem. Floor0 has not been
used to encode from a very long time as I read from this group. So it
should be totally not required for any song decoding.

Also what kind of file-system have you implemented, i.e is the file
read thru a JTAG support by ur toolset or you have implemented one
yourself in the external mem. If a JTAG support to read files from the
Host PC is provided you may even try one by one placing the functions
such as ov_open and its call chain ov_close etc which are used during
initialization and deinitialization.

Also information functions such as ov_bitrate, ov_streams, etc and
test functions such as ov_test_open can be intelligently avoided in
internal memory.

Other functions for init, deinit are :-
vorbis_dsp_destroy
vorbis_dsp_create
vorbis_dsp_restart
...


And all those that do not get called from the ov_read's call chain.

File info.c is also a good caandidate for external SDRAM although some
funcs may be called in the first call to ov_read when the codebook and
other stuff are decoded I think.

I think using these and some toolset options you may be able to fit
the intensive parts into Internal memory.

Ciao,
Rip



On 9/19/05, s22088840 at tuks.co.za <s22088840 at tuks.co.za> wrote:
> Hello everyone
>
> I'm implementing a very simple audio player on DSP as a University project.
> I have quite a fast DSP evaluation board to work with but I am unable to
> achieve real time decoding.  The main problem is that I only have 64KB of
> internal RAM to work with and I thus have to move some of the code to the
> slow external SDRAM.
>
> I basically would like to know where I could save some space, is there code
> that I could remove?  I only need a very simple decoder.  I am decoding
> files as Ogg streams from a memory buffer and am not planning on
> implementing
> seeking.  Also, backwards compatability is unnecessary, as I will encode
> the audio files with the latest reference encoder.
>
> Thank you!
>
> Daniel vN
> _______________________________________________
> Tremor mailing list
> Tremor at xiph.org
> http://lists.xiph.org/mailman/listinfo/tremor
>


--
Regards,
Ripinder Singh


More information about the Tremor mailing list