Re: Rep:Re: [f-cpu] new langage ?

["Christophe" <christophe.avoinne@laposte.net>]

----- Original Message -----
From: cedric <cedric.bail@free.fr>
To: <f-cpu@seul.org>
Sent: Saturday, April 20, 2002 5:05 AM
Subject: Re: Rep:Re: [f-cpu] new langage ?


> > New language of a higher level is really needed to really have a code
> > portability without the problem of performance.
> >
> > BUT it's really hard to devellop a all new language ( or look at "D" ).
> >
> > I personnaly like the way systemC work. It's only a "big" librairy for
> > C++. We just have to design a librairy that work well with SIMD.
> >
> > In c++, one librairy do such thing : STL. If we port STL (in asm) for
> > the fpu we could accelerate all soon written programme and future one.
> Well, it's a good idea, but the problem is that the a big part of the STL
> class are template.

What is the problem with "template" ? it is the best compromise you can have in
fact so long as we cannot have a good SIMD compiler. Creating an ASM optimizer
with template will be cooler than a simple preprocessor in fact (yeah I must
consider creating such a language !).

> That mean that they are compiled with the application,
> and that will be difficult to optimise them.

What do you mean ??? template classes allows us to have a seperate
representation of data and algorithms : sometimes we could prefer to let the
compiler to optimize the code depending of the data the latter handles. It is
its main goal.

> Of course, class like string must be optimised, and all the libstdc++
> can be optimized. But like the libc and perhaps all the basic lib of each
> language we want to see on our chip.


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