Re: [Gumstix-users] Using the Gumstix as a Audio encoder

[Keith O. <kso...@gm...>]

I think I still have a 80287 kicking around somewhere if you want to 
give it a try. <G>

Laker Net man wrote:

>So, I guess it begs the question.  Is there an FPU
>co-processor board of some kind the gumstix could talk
>to as a daughterboard? Ala, the old 8087 co-processor
>days. Ah, there's some memories... plugging in a
>second processor to do FP in *hardware* oooo, aaaahhh
>:D
>
>Laker
>
>--- Alexandre Pereira Nunes <al...@om...>
>wrote:
>
>  
>
>>Marc Olivier Chouinard escreveu:
>>
>>    
>>
>>>MAD is a mp3 decoder, and also explain why it
>>>      
>>>
>>included in the gumstix
>>    
>>
>>>buildroot.
>>>
>>>I have gaveup.  I do not think it can perform to
>>>      
>>>
>>do 128kbits encoding
>>    
>>
>>>sadly
>>>
>>>Thanks alot for your inputs in this.  Im sure I'll
>>>      
>>>
>>find another
>>    
>>
>>>project for the gumstix
>>>
>>>Marc O.
>>>
>>>Keith Olson wrote:
>>>      
>>>
>>With floating-point encoders, it surely can't,
>>because emulation is
>>quite expensive. With fixed-point, it probably still
>>can't. MP3 encoding
>>is a rather cpu-intensive task when compared to
>>decoding, and arm is a
>>RISC engine, which means it has mostly
>>timing-predictable simple
>>instructions, which means complex operations have to
>>be implemented in
>>software (even though the xscale family has some
>>support for vector
>>integer math through coprocessor 0, which could
>>improve things a bit,
>>but AFAIK nobody appears to use that on linux),
>>which finally means it
>>would probably require a higher cpu clock (you can
>>overclock your
>>gumstix quite a bit using cpufreq, I've done it, but
>>I can't say it's safe).
>>
>>I'm not convinced yet that it is impossible to reach
>>your goals, but I
>>don't know if it's worth the try.
>>
>>Alexandre
>>
>>    
>>


-- 
Keith Olson
K-Soft Consulting