[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [f-cpu] 'make it work' or 'make it fast' ?


This evening i can't escape the mediatic bombing in the
living room. my family is almost hypnotized by tonight's events.
Fortunately, i can still speak with people who can think, not
only feel and react like trained animals.

Ben Franchuk wrote:
> Michael Riepe wrote:
> > On Sun, Oct 07, 2001 at 02:13:55AM +0100, Yann Guidon wrote:
> > [...]
> > > It would be cool if Alliance accepted VHDL'93 sources, or even
> > > FOR .. GENERATE loops !!!
> > Or library clauses... *sigh*
> >
> > I guess I'm only dreaming, but can't we feed Alliance a netlist?
> What about using a macro  pre-processer for all that stuff.
it would become extremely heavy :-/
VHDL and RTL in general is already not straight-forward to learn.
I have remarked that the work is easier when we stick to simple,
fully-standard VHDL source. The minimum requirement is to pass
the comilation with Vanilla, because it's the easiest to setup
for newcomers (before Simili'2 is released).

Maybe we could use m4 for generating the proper invocation for
latches and flip-flop, but it would not be useful : VHDL
already provides the necessary means with operator
overloading and component instanciation (you can redefine
the component without touching the interface).

> What about writing your own register-transfer-langage
> pre-processor for F-cpu. This way the output can be what ever
> format you want. It is not that I don't like high-level
> synthesis, it the fact that the two most popular languages are a
> bitch.
I agree with some of your points.
However it is not our goal to define a new RTL langage, and we
don't have time nor experience, i think. It is already great
to have some of the basic building blocks such as the minimum
VHDL tools and some execution units.

> Ben.

nicO wrote:
> Yann Guidon a écrit :
> <..>
> > > > It would be cool if Alliance accepted VHDL'93 sources, or even
> > > > FOR .. GENERATE loops !!!
> > > Alliance doesn't support "process", too...
> > that, too.
> > And they declare FlipFlops with ... guarded block :-(
> An complete unrecommended manner from synopsys point of view...
I don't know what they had in mind when they did that.
One "solution" if we want to use Alliance is to use external
components but it would force us to adopt a very low-level approach
(heavy and less readable).

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