Hi Vitaliy,
Thanks very much for these performance measures. That'll be something I
can point people to when they ask about Speex on ARM.
Cheers,
Jean-Marc
Vit Kanevsky a ?crit :>
>
>>Dear Speex developers
>
>>I am going to port Speex on LPC2368
>>I tested Speex encoding and the mesurments shows ~40ms cpu time for one
>>frame
>
>>Do you know who ported speex to NXP or other 32bit platform?
>
>>Best Regards
>>Zohar fox
>
> I have an own optimisation for Speex for different versions of ARM
> architecture.
>
>
>
> Here is Measurements of performance for STM32 72 MHz.
>
>
>
> 4000 kbit/sec
>
> Encoder - 59%
>
> Decoder - 3%
>
>
>
> 6000 kbit/sec
>
> Encoder - 43%
>
> Decoder ? 3.8%
>
>
>
> *8000 kbit/sec*
>
> *Encoder - 47% *
>
> *Decoder - 4%*
>
> **
>
>
>
> 11200 kbit/sec
>
> Encoder - 65%
>
> Decoder ? 4.1%
>
>
>
> 15000 kbit/sec
>
> Encoder - 53%
>
> Decoder ? 4.2%
>
>
>
> Except 11200 mode it can work on 48 MHz STM32 USB Access Line. (11200
> probably can work in non full duplex application).
>
>
>
> Version for ARM v5E (STR9) is faster for about of 25% for comparable
> processor?s frequencies, by using of DSP commands.
>
>
>
> Version for ARM v4 (STR7, LPC21XX/23XX/28XX) can work even on 60 MHz.
> (but not all speex modes. Just 6,8,15).
>
>
>
> Here is Measurements of performance for LPC2138 - 60 MHz.
>
>
>
> 4000 kbit/sec (quality = 1)
>
> Encoder - 17.9 msec/frame 59%
>
> Decoder - 0.97 msec/frame 3%
>
>
>
> 6000 kbit/sec (quality = 2)
>
> Encoder - 13.5 msec/frame 43%
>
> Decoder ? 1.11 msec/frame 3.8%
>
>
>
> 8000 kbit/sec (quality = 3)
>
> Encoder - 15.07 msec/frame 47%
>
> Decoder - 1.14 msec/frame 4%
>
>
>
> 11200 kbit/sec (quality = 5)
>
> Encoder - 21 msec/frame 65%
>
> Decoder ? 1.19 msec/frame 4.1%
>
>
>
> 15000 kbit/sec (quality = 7)
>
> Encoder - 16.71 msec/frame 53%
>
> Decoder ? 1.24 msec/frame 4.2%
>
>
>
> on LPC2368 - 72 MHz NXP All modes work ( in case of using DMA for
> collect 160 input samples)
>
>
>
> Best Regards
> Vitaliy Kanevsky
>
>
> ------------------------------------------------------------------------
>
> _______________________________________________
> Speex-dev mailing list
> Speex-dev at xiph.org
> http://lists.xiph.org/mailman/listinfo/speex-dev