[maemo-developers] Ogg vorbis/tremor dsp task questions

[From: Simon Pickering]

>> Don't get too excited, I'm writing code, I don't have anything working.
>> I should also add that I need to check about the copyright for some of
>> the ASM code I'm using before I can release anything very much, so this
>> is a theoretical discussion more than anything else.

> Unfortunately I can't help you either, but if you could put your code
> somewhere with instructions about how to compile and use it I might
> come up with some ideas.

I will do so, I just have to check that I'm able to publish the source 
I'm basing the DSP-side code on. Looks like it should be okay though so 
hopefully in the next day or two.

By way of a status update, I've got the ARM-side compiling fine, the 
DSP-side has a couple of errors in one of the header files. I've no 
idea why they are being produced, but I imagine someone will have seen 
the same kind of thing and will be able to suggest a fix once I put the 
source up.

I should note that the code hasn't been run and in all probability will 
not work as hoped without further work (so don't get too excited yet) 
including:

* Alterations to C54 asm to make it work on the C55, or go back to the 
original C code;
* Structural changes/additions of parts of the code that I 
removed/moved in my belief that they were extraneous/could be placed 
elsewhere (I'm not an ogg vorbis expert by any stretch of the 
imagination).
* DSP-side allocation from a fixed buffer - I'm not sure whether this 
will work as is as it uses global vars to hold the buffer data and 
pointers, I don't know how this will work with the dspgateway as I 
thought all task data should be stored in udata? This can be ficed by 
passing the struct to all the allocation routines, but that's lots of 
searching and adding parameters so I've left it alone for the time 
being to see if I can avoid this. I should also note that kulve's Speex 
dsp task uses a similar method and appears to work so hopefully this 
will do too.
* Same as above for my parameter buffer (for passing function return 
values and parameters back and forth between the DSP and ARM) - though 
this can be easily fixed by passing the task struct to these functions 
and there aren't many calls to these functions to be fixed.
* Establishing buffer sizes that both fit and work
* I've no idea whether it will work in real-time, or indeed whether it 
will even run faster than Tremor code on ARM.

> Also I think you might find useful to check OpenMAX DL:
>
> OpenMAX DL (Development Layer) APIs contains a comprehensive set of
> audio, video and imaging functions that can be implemented and
> optimized on new CPUs , hardware engines, and DSPs and then used for a
> wide range of accelerated codec functionality such as MPEG-4, H.264,
> MP3, AAC and JPEG.
>
> http://www.khronos.org/openmax/

Yes this looks interesting. I plan on looking at it once the vorbis dsp 
task is proven to work (or to not work for that matter). From what I've 
heard about OpenMAX it's the way forward.

> I'm witch Krischan, I also appreciate what you are doing, it's very 
> interesting.
>
> Cheers!

No probs, something interesting for me to learn too! :)


Simon