( ESNUG 456 Item 12 ) ------------------------------------------- [07/17/06]
From: Jasper Lee <lgb=user domain=pgc.com.tw>
Subject: A user benchmarks Custom Wire Load Models vs. "Topographical" DC
Hi, John,
I recently spent some time evaluating topographical DC. What got me curious
about it is that it could eliminate custom wire load models (CWLM) from my
flow. One of our biggest headaches is that we are forced to use CWLM's to
close within a reasonable number of synthesis/placement iterations. Without
CWLM's, closing is painfully long and iterative process. To my surprise
topographical was able to closely match our complicated CWLM based flow.
Apparently DC topographical does some kind of layout during synthesis to
determine the actual wire lengths and then uses this info for optimization.
Our 130 nm test design was 125 MHz with 7 clocks and 28 derived clocks. It
had 164 IO's with 87 macros. The releases used: DC Ultra 2005.09 SP3 (topo
is only available on Ultra) & 2005.09 PhysOpt.
Timing, and area results were measure post-PhysOpt using our CWLM's and
topographical. Some changes were made to the script to include the physical
library and enable topographical, other then that the scripts were
identical. Both used top-down synthesis flow.
DC with CWLM's DC with Topographical
Timing Correlation 10.08% 6.02%
Area Correlation 2.8% 1.4%
Runtime (RTL to PG) 18.16 hr 17.7 hr
Memory (DC+PhysOpt) 4.08 GB 4.14 GB
Topo DC had better timing and area correlation compared to our manually
created CWLM's; to our surprise it also had better area post-PhysOpt.
With CWLM's No WLM's
(No topographical) (Topographical)
Timing (post-PhysOpt) -0.02 nsec -0.02 nsec
Area (post-PhysOpt) 15,037,748 gates 15,003,651 gates
I suspect that the better timing and area was due to the fact that the
quality of the netlist that goes into PhysOpt is fundamentally better.
There was one path group with a 16% timing correlation; these paths go
through macros, so without knowledge of these macros topo DC's prediction
was off. My Synopsys AC said that the next release of DC will have the
ability to read a floorplan input. This should help and I am curious to
see how this improves on correlation.
- Jasper Lee
Progate Group Corporation Taipei, Taiwan
Editor's Note: When I heard of this "Topographical" DC, three questions
immediately came to mind:
1) "Is this 'new' Topo DC just a Synopsys marketing hocus pocus
repackaging of PhysOpt-into-DC? Is Topo DC a dumbed down
version of PhysOpt or something actually new?"
2) "If I mix up revs of Topo DC and PhysOpt, does my area and timing
correlation fall apart?"
3) "How well or poorly do Topo DC generated netlists correlate with
Astro, Magma, and Cadence P&R flows? If Topo DC says I've closed
timing with X area, will these backends later say I didn't?"
I don't mean to throw stones at Topo; I'm just trying to gauge its value.
I'd really like to hear the Synopsys answers to these questions. - John
Index
Next->Item
|
|
