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

Re: [f-cpu] vhdl2c 0.1



hi !

Kim Enkovaara wrote:
> 
> >  - VHDL simulations are not easily parallelisable.
> > A globaly adressable, distributed memory multiprocessor might give better
> > results thanks to faster synchronisation, better internal bandwidth and
> > smaller OS overhead.
> 
> There has been many startups that have tried to create parallel
> simulators, but none of them have succeded in it well enough to create
> any real demand. And I'm quite sure that the big companies have also tried
> to make parallel simulators. I heard from one manufacturer that the memory
> latency and bandwidth needs negate the effects of parallelization.. It's
> more wise to run one simulator on each individual processor and just run
> many simulations in one machine, and not try to make one simulation
> faster.

Cedric also thinks that if we can translate VHDL to C, we can "speed up
the simulation" (but i showed a counter example) and make it run
on a non-PC platform. He mentionned Alpha, PPC and SUN examples
but i would like to have an account on such a multiprocessor machine
AND have the optimising, parallelizing compiler :-)

I would like him to speak about his own ideas on the list,
instead of me telling what he said... And in the end i remark that
his suggestions were already though about by a lot of companies
and the specialists (like Kim) know the results...
Frankly, i'm still happy with Vanilla and Simili.
When we will need speed, i'll dust off ncsim's documentations
and that's already a good start for running some BIOS or things
like that. Then, if we want to run mostly real-frequency sims,
we'll ask sponsors and friends for FPGAs and/or emulators...

Otherwise, like Cedric said, we can run the simulation on a 1GHz Athlon
all week long at his school ;-) I guess i could also ask some CPU
and quota from my university's admin on our largest computer.
i think it's a 4-CPU SUN but then we'd need another ncsim licence :-(

But before we can think about that, we have to finish the code.
I say "we" because i'm not alone ! Though sometimes, i think that
if Michael was more "active" i would feel less lonely ;-)
it's not against Michael but directed to all those who do not write code :-(
Damned, i think we're 80 on this list...

> --Kim
> ps. I would love to have a good parallel simulator :)

now you understand why i was so excited when i was hired at Meta Systems :-)
I hope i could still contact people there in the future.


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