Re: [f-cpu] Re: Floating-Point?

[Yann Guidon <whygee@f-cpu.org>]

hi,

nicO wrote:
> Michael Riepe a écrit :
> > If you have enough room.  Do you have an idea how big the FP unit will be?
> It's big but not as a 256 Ko SRAM cache. So it will be the choice of the
> user. Duplicate big unit could save a lot of computing power (because
> you make a single run).

After all, the files are under GPL so they can be modified if needed.

> For your 16 bit fp processing, could not be better to use log number ?
> For audio processing it could be enough (no DC stuff).

i'd rather use 32-b ints or 32-b FP instead of 16-b FP.
unless you really don't care about the huge rounding noise that you will
create. 16-b and 32-b LNS could be used too, but i know that some people
are not very happy with it.

> For the following text, it could be nice to update quickly the manual.
> If whygee is ok.

the "little" problem is that Olivier Jean has not submitted his work.

i'm ok for modifying the manual, of course, but it is a very heavy
and long work and if Olivier comes with a "forked" revision,
the branch merge will be another nightmare.

- i hope that olivier will send some news about what he has done,
  is doing and will do.
- we can prepare "patches" that will be applied to the manual in either
  new or old version. these would be "manual patches" in case a new
  manual revision comes.
  

> nicO
WHYGEE
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
*************************************************************
To unsubscribe, send an e-mail to majordomo@seul.org with
unsubscribe f-cpu       in the body. http://f-cpu.seul.org/