( ESNUG 455 Item 3 ) -------------------------------------------- [06/29/06]
Subject: Users on the Apache RedHawk FAO option
NEW & LESS IFFY -- The bulk of the respondants to my survey sort of knew
about RedHawk's FAO option, but they seemed to have a wait-and-see view
of it. They like the *idea* of noise repair; they're just waiting for
other users to debug it first (I guess).
"Do you use RedHawk's FAO option for noise repair? How did it work?"
We used Apache's FAO (Fix And Optimize) option on two ASIC designs;
130 nm and 90 nm. FAO creates an ECO script for the Magma Blast Fusion
layout tool to implement changes for reducing Dynamic Voltage Drop
(DvD), using decap insertion, cells swapping, and power grid changes.
We checked the runtime and capacity for both Apache FAO and Sequence's
CoolTime fixing capabilities.
1. FAO was approximately 50 percent faster on large designs.
2. FAO achieved a 2% to 25% DvD reduction over CoolTime's
equivalent.
Here is how FAO works. Layout tools can often group high-drive cells
into clusters in low cell utilization areas. In this case, cell
swapping can help DvD and enable decap insertion in the cluster area.
In designs with more than 85% cell-utilization, the layout tool must
reserve area for decoupling capacitance insertion in order to reduce
DvD. This is where we use FAO's power grid enhancement options and
decap insertion. After cell placement optimization, decap insertion
and power grid enhancements, RedHawk has a design rule check that
verifies poly/metal width/spacing/overlapping. It also removes
ineffective decaps to ensure the die area is utilized in the most
advantageous way.
- Lief Erickson of VikingStarTech, Inc.
No, but review seems to indicate it is a valuable option
- [ An Anon Engineer ]
Used FAO option. It does its job well but is not tied closely to other
vendor P&R tools. Some of the fixes may cause other problems like
timing and congestion. Closer P&R and STA interfaces will be required
to make FAO really useful.
- [ An Anon Engineer ]
The decap insertion flow was good. The grid resizing was not tested
enough, but gave some interesting results on the trial.
- [ An Anon Engineer ]
I have not used these options yet. I am excited about FAO and
believe it will be a great asset once learned.
- [ An Anon Engineer ]
Not yet for both.
- [ An Anon Engineer ]
Not yet.
- [ An Anon Engineer ]
No. But should be a very good feature.
- [ An Anon Engineer ]
No
- [ An Anon Engineer ]
No, we don't use FAO for signal-noise repair. However, we are rolling
out usage of FAO for DVD repair. We've been able to work with Apache's
development team to refine the way FAO works to best suit our repair
needs. With these refinements, we see consistently good repair results
post-analysis.
- [ An Anon Engineer ]
No
- [ An Anon Engineer ]
I didn't use, or even don't know the PowerGate or FAO. We are just
running static and dynamic IR drop and electromigration analysis, no
backannotation on signal delay.
- [ An Anon Engineer ]
Index
Next->Item
|
|
