[tremor] Ogg/Vorbis report, FFT optimizations

[timmy brolin]

>> Let's stop on this subject?  It's not too useful to discuss here...
> I agree.

Me too.

> The target of the optimizations is a low memory DSP. The FFT based
> MDCT will probably speed up the execution on ARM, but a DSP is more
> beneficial in terms of exploting the regular structure of an FFT
> kernel.

Yes, it would be very interesting to see the results of this code on a ARM7. I'am not so sure the DSP is that much better at FFT than the ARM.. Sure, the dual pipelines and MACS help, but the ARM has an edge at multiplications when compared to a 16bit DSP.

> Also better power consumption for low power embedded systems.

ARM is among the best when it comes to MIPS per Watt.
A quick check in the datasheets tells me that the TI C5510 and ARM7 have about the same power consumpt)on.ID
In order for the C5510 to consume less power than a ARM7, it must do the same job with fewer cycles.
An educated guess based on the numbers in your paper is that a ARM7 will require about the same number of cycles (the same level of power consumption) as the C5510 if a ARM7 version of your FFT based MDCT is implemented.
It would be very interesting to see some hard numbers on your FFT based MDCT for ARM7. I think it could be a winner.

/Timmy

--- >8 ----
List archives:  http://www.xiph.org/archives/
Ogg project homepage: http://www.xiph.org/ogg/
To unsubscribe from this list, send a message to 'tremor-request at xiph.org'
containing only the word 'unsubscribe' in the body.  No subject is needed.
Unsubscribe messages sent to the list will be ignored/filtered.

<p><p>--- >8 ----
List archives:  http://www.xiph.org/archives/
Ogg project homepage: http://www.xiph.org/ogg/
To unsubscribe from this list, send a message to 'tremor-request at xiph.org'
containing only the word 'unsubscribe' in the body.  No subject is needed.
Unsubscribe messages sent to the list will be ignored/filtered.