[CELT-dev] fixed point code

Jean-Marc Valin jmvalin at jmvalin.ca
Thu Mar 3 14:12:17 PST 2011


They are in static_modes_fixed.c, but they should actually be used  
automatically.

     Jean-Marc

"J.B. (Jae) Hong" <jbhong at broadcom.com> a écrit :

> Thanks for your prompt response.
>
>>> ROM tables are already included for frame sizes 120,240,480,960 at 48kHz.
> Could you let me know where to find those tables?
>
> Kindly regards,
> //JB Hong
>
> -----Original Message-----
> From: Jean-Marc Valin [mailto:jmvalin at jmvalin.ca]
> Sent: Thursday, March 03, 2011 1:26 PM
> To: J.B. (Jae) Hong
> Cc: celt-dev at xiph.org
> Subject: Re: [CELT-dev] fixed point code
>
> Hi,
>
> "J.B. (Jae) Hong" <jbhong at broadcom.com> a écrit :
>> Could you provide us with fixed-point code for initialization of
>> this array? Or you may kindly provide us with ROM tables of the
>> array for each possible combination of frame size and sampling
>> freq...?
>
> ROM tables are already included for frame sizes 120,240,480,960 at 48
> kHz. For the others, you can use the dump_modes utility with the
> sampling rate(s) and frame size(s) you need.
>
>> [2] Usage on global stack memory
>>
>> In "arch.h", there is a definition on the following constant for
>> global processing memory (stack RAM).
>>
>> #define GLOBAL_STACK_SIZE 100000
>
> This "global stack" is only a fall-back when neither C99
> variable-length arrays, nor alloca() are available. See if you can use
> one of those instead. For that, all you need is define either
> VAR_ARRAYS or USE_ALLOCA.
>
>> -          Peak state memory (static RAM) size?
>>
>> -          Peak processing RAM size including local stack RAM
>> (100,000 bytes?)?
>
> I have not measured those, but I'm sure it's under 100k. Also, it's
> probably possible to make CELT more efficient than it currently is
> when it comes to stack size.
>
>> -          Minimum input bit-stream buffer size with maximum
>> bit-rate per frame considered at 48KHz sampling rate?
>
> If by that you mean the largest compressed packet that can exist, then
> the answer is 1275 bytes.
>
>     Jean-Marc
>
>
>
> _______________________________________________
> celt-dev mailing list
> celt-dev at xiph.org
> http://lists.xiph.org/mailman/listinfo/celt-dev
>






More information about the celt-dev mailing list