Reply by Noffke, Pat March 8, 20012001-03-08
I should think it would be enough. You might also consider doing the
wavelet transform in place, as much as you can, anyway. You should also
consider how you will store your 8-bit data. If you pack it 4 bytes to a
32-bit word, that will add overhead for processing time, and a little for
code as well. Unless you need to save the original data though, doing it in
place would be the way to go. Store your 8-bit data as one byte per 32-bit
word, and then the conversion to float and the processing of the data is
much easier. Then you just run the wavelet filter on one row/column at a
time, and stuff the results back where you got them.

Pat

> -----Original Message-----
> From: Jack Galanda [mailto:]
> Sent: Thursday, March 08, 2001 4:06 AM
> To:
> Subject: [adsp] ADSP-21160 Wavelet Algorithm > Hi
> I am newcomer to DSP programming.
> I would like to use the ADSP-21160 to implement the 2
> dimensional wavelet
> algorithm to compress 256*256 integer matrices (8 bits
> per pixel).
> I have to use 256*256*8 bits to store an input integer
> matrix and
> 256*256*32 bits to store the floating wavelet
> transformed matrix;
> in order to use the ADSP-21160 4Mbits of internal
> SRAM, I am left with
> 4Mbits - (256*256*8 + 256*256*32)bits = 1.5 Mbits for
> the program.
> Do you think that 1.5 Mbits are enough for a quite
> simple algorithm
> (12k of C code) like the wavelet one?
> Thanks
> M. > ______________________________________________________________________
>
> _____________________________________
> Note: If you do a simple "reply" with your email client, only
> the author of this message will receive your answer. You
> need to do a "reply all" if you want your answer to be
> distributed to the entire group.
>
> _____________________________________
> About this discussion group:
>
> To Join: Send an email to
>
> To Post: Send an email to
>
> To Leave: Send an email to
>
> Archives: http://www.egroups.com/group/adsp
>
> Other Groups: http://www.dsprelated.com > ">http://docs.yahoo.com/info/terms/

This message has been scanned for viruses with Trend Micro's Interscan
VirusWall.


Reply by Jack Galanda March 8, 20012001-03-08
Hi
I am newcomer to DSP programming.
I would like to use the ADSP-21160 to implement the 2
dimensional wavelet
algorithm to compress 256*256 integer matrices (8 bits
per pixel).
I have to use 256*256*8 bits to store an input integer
matrix and
256*256*32 bits to store the floating wavelet
transformed matrix;
in order to use the ADSP-21160 4Mbits of internal
SRAM, I am left with
4Mbits - (256*256*8 + 256*256*32)bits = 1.5 Mbits for
the program.
Do you think that 1.5 Mbits are enough for a quite
simple algorithm
(12k of C code) like the wavelet one?
Thanks
M. ______________________________________________________________________