Re: [f-cpu] Verification, Testing, And Random Numbers

[Michael Riepe <michael+fcpu@stud.uni-hannover.de>]

[heck, it bounced again :(]

Hi F-gang,

Pierre Tardy wrote:
> On Tue, 03 Feb 2004 17:18:30 +0100
> Michael Riepe <michael+fcpu@stud.uni-hannover.de> wrote:
>
>
> > But in general, there is no "golden" code. There is only another 
> > implementation of the same problem that may or may not be correct. You 
> > have no way to verify that, unless you use one of the methods outlined 
> > above. In that case, you can as well verify your circuit directly.
>
> If you have a difference between the result of the two models, you know that there is a problem, then you can study step by step the behavior of the suspected implementation, reacting to this particular stimulation.
> So this is not totally unusefull, it can be used to detacted cases that you have not thought about. 
That's the way I have been "debugging" my designs: processing units and
testbenches use different approaches to obtain the result, and if
there's a difference, at least one of them must be wrong. In that case,
I examined both of them more closely, and I usually found an error in
one of them. But it's not a 100% reliable way to verify a design.

Michael.



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