( ESNUG 435 Item 8 ) -------------------------------------------- [12/08/04]
Subject: ( ESNUG 431 #5 ) Pessimism And PrimeTime-SI Path-Based Analysis
> We've been looking to reduce the pessimism in PrimeTime-SI. ... Our AE
> sent us a beta version of a SolvNet app note on their new path-based
> analysis.
>
> - Lydia Lee
> eSilicon Corp. Sunnyvale, CA
From: Chris Papademetrious <chrispy=user domain=synopsys spot calm>
Hi John,
In ESNUG 431 #5, Lydia Lee from eSilicon mentioned a beta version of an
applications note and Tcl procedure which she used to drive PrimeTime's
new path-based analysis feature. I am the PrimeTime corporate engineer
who wrote the appnote and developed the procedure. Please let your
audience know that this appnote and procedure have now been published
into SolvNet as article number 012134, entitled:
"Accurate sign-off analysis with PrimeTime's path-based analysis"
The article begins with a thorough introduction to the concept of slew
propagation, which is the fundamental mechanism underlying static timing
analysis. (I have received feedback from one user saying that the
article would be worth a read for this section alone!) I also cover the
difference between worst_slew and worst_arrival slew propagation modes,
which many users have been curious about. With these basic concepts
covered, I then explain the concept of path-based analysis and how it
can be applied to static timing analysis.
In addition, I have recently updated the article to cover some new
topics, as well as document some new features added to the procedure
based on feedback from Lydia Lee and other users. The most useful of
the improvements is a -verbose switch which allows the user to see
violating endpoints as soon as they are found by the path-based analysis
process.
Personally, I am very excited about path-based analysis. I think it is
one of the most significant STA technologies to come along in quite
awhile. Remember when simultaneous min/max optimization came out in
Design Compiler? Or when you first realized how much work case analysis
saved over manually disabling timing arcs? This is right up there.
Multiple customers have told me that path-based analysis has already
become a can't-live-without-it feature in their flows. I urge you to
spend a few minutes reading the article, and encourage you to submit any
feedback you have.
Okay, I confess that there are a few known bugs in the initial 2004.06
release of path-based analysis. All of these have been fixed for the
2004.06-SP1 release, which will be available this month. (The article
contains more details.)
By the way, Lydia also mentioned some significant improvements in
transition times in her 2004.06 PrimeTime-SI results, compared to
previous releases. I thought your readers might appreciate a little more
technical detail on this topic.
PrimeTime-SI performs delay calculation in "stages," which consist of a
cell arc along with the net being driven. For each stage, the waveform
responses at the load pins of the stage are measured at their trip points
to determine their transition times. These transition time measurements
are then propagated to the inputs of the subsequent stages.
This works well until strong aggressors begin to heavily distort the
response at the load pins with crosstalk-induced bumps. Because of the
distortions, measurements of this nonlinear waveform taken at the linear
transition time trip points resulted in pessimistic (but safe)
transition time values. In 2004.06, we introduced some new technology
which allows such waveforms to be more accurately propagated. As a
result, the downstream stages are stimulated more accurately than is
possible with a simple linear transition measurement of a nonlinear
distorted waveform, resulting in the more accurate slew values that
Lydia saw.
- Chris Papademetrious
Synopsys, Inc. Brodheadsville, PA
|
|
