Speex dev - Feb 2010 - Manual scratch allocation : memory usage doubt

Hi,

I am currently encoding 32 Khz/Qual-10/UWB encode mode, with  
MANUAL_ALLOC enabled ( similar to the c5x/c6x configuration).

In file sb_celp.c, I noticed the scratch memory grow during 
recursive (UWB -> WB) calls to 'sb_encode'.

1. 'stack' was not tracked (with 'tmp_stack' as done at other
palaces)
   after/before - auto-correlation/Levinson-Durbin scope (~line 597 -
621).

2. Similarly in the sub-frame encode loop (last major loop) after/before
- 
   LSP-interpolation/weighted-signal-frm-synthesized-speech calculations
   (~line 762 - 950).

In <1>, the scope of the variable over after the braces, so I think that
memory allocated on the scratch may not be needed further. Is it right?

In <2>, I think the excitation, and other of one sub-frame will not be 
needed for the others. Is it ok?

So, is it ok to do reuse the above scratch stack?
This saves about 1.2 Kbyte of total scratch space required when running
in UWB/Qual-10 mode.

Regards
Sameer

Sameer,

I hope that you were able to resolve your problem, since you asked your 
question two months ago.  The MANUAL_ALLOC switch was only tested in 
narrowband mode for the c5x/c6x, so it would not be surprising if something 
were missing in wideband mode.

Regards,

Jim Crichton

----- Original Message ----- 
From: "Sameer Manchali" <sameerm at sanyo.co.in>
To: <speex-dev at xiph.org>
Sent: Saturday, February 20, 2010 4:31 AM
Subject: [Speex-dev] Manual scratch allocation : memory usage doubt

> Hi,
>
> I am currently encoding 32 Khz/Qual-10/UWB encode mode, with
> MANUAL_ALLOC enabled ( similar to the c5x/c6x configuration).
>
> In file sb_celp.c, I noticed the scratch memory grow during
> recursive (UWB -> WB) calls to 'sb_encode'.
>
> 1. 'stack' was not tracked (with 'tmp_stack' as done at
other palaces)
>   after/before - auto-correlation/Levinson-Durbin scope (~line 597 -
> 621).
>
> 2. Similarly in the sub-frame encode loop (last major loop) after/before
> -
>   LSP-interpolation/weighted-signal-frm-synthesized-speech calculations
>   (~line 762 - 950).
>
> In <1>, the scope of the variable over after the braces, so I think
that
> memory allocated on the scratch may not be needed further. Is it right?
>
> In <2>, I think the excitation, and other of one sub-frame will not
be
> needed for the others. Is it ok?
>
> So, is it ok to do reuse the above scratch stack?
> This saves about 1.2 Kbyte of total scratch space required when running
> in UWB/Qual-10 mode.
>
> Regards
> Sameer
>
>
> _______________________________________________
> Speex-dev mailing list
> Speex-dev at xiph.org
> http://lists.xiph.org/mailman/listinfo/speex-dev