🚀 go-pugleaf

Johnny <john_wr100@> schrieb in im Newsbeitrag:
...
> One thing that concerns me is that the evaluation kit is limited to
> 8kb program size.  ie 2048 Words.

It's actually less: If code size is limited to 8kB = 65536 Bit then you only
have 1365 instruction words (one instruction word = 48 bit).


> Should I assume this is going to
> be more than enough for my FIR filter implementation?
>
> I wonder, is there separate memory space to store my filter
> co-efficients... Because I will probably have nearly 2000 filter
> coeficients which wouldn't leave a lot of space for program.

I just checked a program I wrote for the 21160 a while back. It's a 8
channel three-band crossover,  designed to work at 96kHz with 24bit quality
(regardig SNR of the lowpass FIRs). It uses a two-stage downsampling
architecture, and the lowest band frequency is 41 Hz.

I program in assembler. The whole program is 1080 instructions big, so 1365
are plenty ;-). There was room enough in the internal memory for everything,
including filter coefficients (although the 21160 has a built-in SDRAM
interface, which makes memory even less of an issue). I strongly suggest you
learn the assembler. I quite like it, and you are a lot more efficient in
program speed / size. Also, the C / C++ compiler / debugger is notoriously
buggy. Check out
http://groups.yahoo.com/group/adsp/.

> Also I suppose the program must be stored in a low-speed flash memory
> anyway, and probably gets loaded by the boot loader or external
> micro..

That's the way we do it.


Regards,
Andor