( ESNUG 443 Item 3 ) -------------------------------------------- [04/15/05]
Subject: ( ESNUG 440 #4 ) A Verification Flow Using Both Jasper & Magellan
> When Jasper finds a false proof, you can get a nice trace. Their current
> GUI is a little weak. It's Tcl/java based. Not the prettiest looking
> but gets the job done. They are developing a new GUI though.
>
> You start at the end result and work backwards. You can click and get
> all signals that are part of the fanin for a signal, or you can source
> browse with a "why" button that shows why the value is that way.
>
> - Kevin Normoyle
> Azul Systems, Inc. Mountain View, CA
From: Erik Maier <erik.maier=user domain=tundra spot gone>
Hi John,
I found Jasper fairly difficult to get up and running meaningful cycles on
my own. The documentation is severely lacking so I was able to bring the
tool up but that was pretty much it. Jasper seems to work by sending a rep
out to get you going to insure that you are doing things correctly.
I found this help outstanding. They worked very hard and within a week
we were looking at waves that meant something.
I believe this Jasper tool is really useful at the early stages of design
with someone who has a design background driving it. We had a Mentor type
setup with one of our head designers and myself, so both of us were
familiar with the internals of the design. Our project was pulled before
I was able to complete all the desired proofs but we still found bugs/issues
with Jasper. I think the most useful aspect of the Jasper tool was the
process it forces you to take to constrain the search regions. We found
that by doing this Jasper forced us to re-evaluate design decisions we had
made and also see things that we hadn't thought of before. But by being
forced to be so intimate with the design to make those assumptions, I don't
believe some of our verification engineers would find it easy to do.
Another use we found that completely saved us was we used the Jasper tool
to debug a silicon problem we had. We were able to setup what was
happening on the outputs and Jasper pretty quickly showed us a combination
on some inputs we couldn't see before and lead us to the root cause.
Granted there was some luck involved because this combination happened
to show up almost immediately but it was a good debug tool that could be
setup quicker than experiments in the lab or on a tester.
I personally haven't used Magellan yet, Synopsys is currently all over me,
but we do use it. From what I have heard we are using Magellan with some
success. We see both tools having a place in our flow. Jasper being used
in the early development stages to help debug the design, and Magellan
assertions to help debug test errors later in the flow. Jasper can
systematically find bugs where Magellan is more of a side-band thing that
can flag errors if you hit them. So both play in the formal space but
approach it differently.
One other note, the head designer is one of our more stubborn guys and
hated the Jasper tool back when it was TempestQuest. He pretty much
discarded any results from the tool because we couldn't constrain things
properly. This last time through, with the latest Jasper tool, the same
guy came around and found Jasper useful. His thing is "show me waves I
can make sense of and I'll figure out the problem", and Jasper was able
to do that. Hopefully I haven't rambled too much. Let me know if you
want or need more information and I will gladly provide it.
- Erik Maier
Tundra Semiconductor South Portland, ME
Index
Next->Item
|
|
