( ESNUG 457 Item 3 ) -------------------------------------------- [10/05/06]
From: Jason Ware <jware=user domain=cadence spot mom>
Subject: Jason Ware's introduction of the new Cadence ETS PrimeTime-killer
Hey John,
This is a follow up to our conversation about our new stand-alone static
timing analysis tool, the Encounter Timing System (ETS). When we talked
on roll-out day I said I could not give you much detail since I had not
yet used the tool. I have only been at Cadence since February and have
been focused on our RTL Compiler and Conformal Constraint Designer. I
recently just completed training on ETS so I can now give you that more
"hands on" introduction you requested.
As you know ETS is our answer to Synopsys PrimeTime for signoff static
timing analysis. This is important for two reasons. First, those
customers wishing to have an all-Cadence flow (yes there are a growing
number of those type customers :), no longer have to bring in PrimeTime
for timing and signal integrity sign-off. Second, our ETS is based on
CTE, CeltIC, VotageStorm and SignalStorm which are the timing, SI, delay
calc and IR-drop engines for SOC Encounter. That means there are no
surprises when our customer completes his design and then times it for
signoff in ETS.
After training on the tool I said would like to give you my impressions.
A static timing tool must perform on two fronts. First, it must be
accurate and comprehensive. I haven't run enough designs to comment on
this, however, our partners have. ETS is now approved for signoff by
vendors like TSMC, UMC, Fujitsu and several others and from what I see
it looks good.
Secondly the tool must be easy to use in terms of analysis and debug.
Here is where I think ETS is a step ahead of PrimeTime. The menu system
within the GUI of ETS is well thought out and easy to navigate. I did
run into some "fit and finish" issues which R&D tells me are largely due
to the way the TK package behaves on different platforms and OS versions.
For example the "clock face" icon that shows ETS is working does not show
up when the cursor is within the window, only when it is to the side. If
you zoom in on the path histogram you should get scroll bars to move
around on the zoomed image.
In the Path Category window, if we scroll the window we loose the most
significant digits when the column is too narrow. For example I had
a TNS that was 197.xxxx, I mistakenly read it as 97.xxxx. We should
loose the decimal places rather than the MSD if the column is too narrow.
(Note it's displayed correctly in reports and the Summary Window.)
I would also like to see a default name for a category if you forget to
assign a name, rather than just a blank group name.
Yesterday I filed the Product Change Requests on these bugs and a few other
minor window issues that I am sure will be fixed in an upcoming release.
My favorite parts of the ETS analysis GUI are the Path Category window and
the Path SDC window. Path categories allow you to customize the histogram
to see how much of the slack is begin affected by paths of your choosing.
This is updated dynamically and can be color coded. For Path SDCs, after
selecting a critical path and viewing it's timing graph, you can click on
the Path SDC tab and ETS will show you the SDC statements (set_input_delay,
create_clock, etc.) that affect the path along with the line numbers from
the SDC file. This makes debugging much easier.
You can also launch a simulation run from the Timing Path Analyzer which
will build a SPICE deck and run a SPICE simulation or you can run it
natively with the onboard simulation engine. This gives you the ultimate
accuracy for analyzing your critical paths.
The Timing Path Analyzer gives you a color coded view of the path that
shows cell as well as net delay so you get a visual clue to any abnormally
long delays along a path. These are all updated when you do SI analysis
as well. My only suggestion to Cadence R&D is to add the clock waveforms
for the selected path. This would show at a glance if there is an error
in the definition of the clocks (incorrect create_clock statement, for
example). They like my idea on this and have already started working on
implementing it.
John, please ask your readers to give ETS a try. Trying the tool is easy
because we have a Programmable Timing Interface mode that will run existing
customer scripts. ETS also has a CeltIC interface as well, so current
CeltIC users would be able to use it without issue.
We are looking for user feedback on ETS and hope to transition as many
people as possible.
- Jason Ware
Cadence Design Systems Plano, TX
Editor's Note: Before all the EDA vendors give me a boatload of grief
on having a Cadence employee write about ETS, keep in mind that I've
know Jason since he authored the "Successful Synthesis Checklist" for
Synopsys Design Compiler back in 1999. I asked him for this write-up
because he's a known no-BS kind of guy and I felt it would be better
vs. an All-BS Cadence Corporate Marketing press release for such a big
announcement. It's not everyday that PrimeTime is challenged. Next,
I'd like to see some detailed hands-on user reviews of ETS. - John
Index
Next->Item
|
|
