Re: Fixed-point arithmetic

[David Bateman]

Le 18 avr. 2011 à 14:39, Chipmuenk a écrit :

Hi all,

I am using the "fixed" package for some time now to demonstrate the effects
of fixpoint coefficients and arithmetics in digital filters to my students.
It seems to me that this hidden gem of the octave packages is slowly falling
into oblivion as it is no longer actively maintained and hence no longer
appears at octave-forge.

Yeah I've been a bit slack. I haven't released any packages in octave-forge since octave-forge went to individual package releases and I'm not sure sure of the process. I've been confining my efforts with octave to the core functionality. Add to that that I don't need this package actually and my dev machine upped and died, I'm not sure I'm likely to do anything in the near future.

This is quite a pity, as this package is quite powerful in my opinion. As
far as I know, there is no other open source software supporting fixpoint
arithmetics. The only obvious alternative is Matlab's Fixpoint Toolbox but
of course Mathworks is aware of this fact: I cannot affort to provide
licenses for all of my students.

Not quite true. GMP supports fixed point and the matlab code is based on GMP (see http://gmplib.org/manual/). Laurent and I chose not to use GMP with the fixed point code for Octave and base it on 32 or 64 bit integers. In this way the code is much faster than the matlab code, though it can't handle large integers. The reality in communication systems is that more than 16 bits are rarely used, so frankly speed is the key here so I still think this choice was the right one.

Admittedly, this package is quite complex and maybe the purpose of this
package is not immediately clear. Its syntax is also incompatible to the
Matlab fixpoint toolbox (as it was developed at the same time) which makes
this package even more seem like a dead end. But everyone who has tried to
develop a digital fixpoint filter for a microcontroller or an FPGA knows how
important it is to take quantizing / truncation effects into account.

Its not compatible as it was written before matlab released their fixed point package, if it was in parallel I might have tried to copy their interface ;-). Its faster than the matlab fixed point code, so I hardly think its a dead end.

I am not a programmer but I could help rejuvenating the toolbox by providing
ideas, applications, documentations and examples. I also have access to
Matlab with the Fixpoint Toolbox to run comparisons.

Some Ideas I alright had were

- Adapt the code for NDArrays so more Octave code could be used with the package.

- Add different overflow and rounding options. The ones that are used in the code were the ones that were used in the CMOS process I was developing for.

Frankly I don't think its a good idea to make this package matlab compatible, unless we want to use GMP as well.

Please don't let this package rot away in some repository.

I'd rather it didn't and will get around to fixing it some time, as I have a certain desire to see my code used, but I can't guarantee to be quick and would be happy to see someone else get the fixed point package working with 3.4.0

David