On Tuesday, May 20, 1997 3:51 PM, Eric Scheirer wrote:
> >Or is the design of new functions and algorithms (in C)
> >out of the scope that SAOL covers?
>
> Yes, for the keyword "in C" -- it makes it impossible to deliver
> platform-independent content for MPEG-4 if you need to rely on
> an implementation having "special" non-standard built-in opcodes.
> But I'm not sure what your take is on things that *must* be
> implemented in C; the goal of SAOL is to make this largely
> unnecessary.
>
I absolutely agree. If SAOL can provide the complete set of functions,
that I need (also in future) for a specific algorithm, I will implement
it in SAOL, no problem. But, if not, I need to program in a descrete way
(like a=b*c). Do I have a problem with the performance, then? If you
look at
C-compilers for DSPs, you loose a factor of about 4 in performance
compared to assembler.
Another example: To implement a standard filter on a 64Bit floating
point processor you might be happy with a standard biquad. But, on a
16Bit DSP, you maybe better use another structure with error feedback,
where the feedback has some influence for the algorithm. How can I
handle such problems?
I am looking forward for the test-implementation!
regards, Matthias
-- __________________________________________________________ Studer Professional Audio AG, http://www.studer.ch/studer Matthias Rosenthal, Ph.D. <matthias.rosenthal@studer.ch> CH-8105 Regensdorf Tel +41-1-870-72-39 FAX +41-1-840-47-37
This archive was generated by hypermail 2b29 : Wed May 10 2000 - 12:11:08 EDT