( SNUG 03 Item 16 ) ---------------------------------------------- [05/14/03]
Subject: Floorplan Compiler, Avanti Jupiter-XT, Cadence First Encounter
CADENCE'S TRUMP CARD: It's well known that Cadence fell flat on its face
when they tried to run their $260 million Ambit PKS pony in the physical
synthesis horse race and the thing couldn't work. (See SNUG 01 #7.) But
Cadence redeemed itself once again with its checkbook when snapped up
Silicon Perspectives for $132 million. This gave them ownership of the
very popular First Encounter ASIC floorplaning tool.
On the flip side, Synopsys has been struggling in the floorplanning niche.
First their home brew Chip Architect tool failed miserably. Then their
Hidden Dragon / Floorplan Compiler tool came out too late (and too green) to
market to be able to challenge a well entrenched Cadence First Encounter
user base. Then Synopsys picked up the Milkyway-based Jupiter-XT tool from
the Avanti merger -- but Jupiter-XT doesn't have FE's capacity. (Currently
users 2 to 1 prefer Jupiter-XT over Synopsys Floorplan Compiler.) So now
Synopsys is stuck trying to cobble together their Floorplan Compiler and
Jupiter-XT into a new tool that'll even be greener and even later to market.
And no matter how you cut it, Synopsys is royally screwed in floorplanning.
These failed Synopsys attempts in floorplanning reminds me of a joke by
comedian George Burns about what kind of sex life a 100 year old man has.
"It's like trying to shoot pool with a rope."
Dataquest FY 2001 ASIC Floorplanning Market (in $ Millions)
Cadence ######## $15.4 (52%)
Synopsys/Avanti #### $8.9 (30%)
Monterey # $2.1 (7%)
others ## $3.3 (11%)
"The layout team uses Jupiter, we use First Encounter -- not a good
marriage yet, we are each keeping our lovers on the side."
- Bob Lawrence of Agere Systems
"I really liked First Encounter. I think Jupiter-XT looks like a more
useful product than Floorplan Compiler."
- [ An Anon Engineer ]
"About Floorplan Compiler vs Jupiter-XT, I think Jupiter-XT is better.
While working I've seen lots of problem with FPC's feedthrough,
budgetting. The feedthrough generated from it creates lots of
problem while doing clock tree synthesis and I feel PrimeTime
budgetting is better than FPC's. Even with movebound creation I've
seen cases where there exists lots of top level cells, FPC has
generated abutted movebounds. And sometimes, it does not leave any
space for channel. And the option of rectilinear movebound creates
too much overlapped movebound. And it hardly places the block in
a fashion what we call a good floorplan. I've even seen Floorplan
Compiler crash very often. And also for Verilog assign statements
FPC directly shorts them."
- [ An Anon Engineer ]
"No hard comparison was done but we still like First Encounter better.
We looked at Jupiter-XT at its infancy 2 years ago and was not very
impressed. Monterey's Aristo IC Wizard was useless at that time, too.
Magma is luring its customer to depend on the new BlastPlan, which is
not right in the first place even though we are using more and more
BlastFusion. We want the freedom of choice on a floorplanner."
- John Zhang of Broadcom
"Keep Floorplan Compiler, and add Jupiter features. I would say that
Synopsys and Cadence is about the same.
- Bengt-Erik Embretsen of Zarlink Semiconductor
"I know little about Floorplan Compiler, but for me Jupiter-XT is good.
First Encounter is the best."
- Tie Li of Applause Technolgy
"Floorplan Compiler was developed with current large design requirements
in mind. Jupiter-XT wasn't. So Floorplan Compiler should be retained,
with a relatively long phase out/merging for Jupiter. I haven't used
FPC yet, but do know that Cadence First Encounter has a host of issues
that arise in large chips, too."
- [ An Anon Engineer ]
"If you are using PhysOpt, I think you should stick with Floorplan
Compiler and have the front-end synthesis team handle the floorplan.
Without physical synthesis, Jupiter-XT is the way to go, since it is
mostly compatible with Astro. We like the 'cookie cutter' feature,
where you do the power grid flat, at the top level, and then push it
to the lower level macros."
- Roberto Landrau of Mitre
"Can't speak about the whole Astro product, but Jupiter/PhysOpt works
well. I do like Jupiter better than Planet. Haven't used the latest
Cadence acquisitions, but I'm skeptical that they could have gotten
everything working well together in such a short time.
I'm a Jupiter-XT fan. I don't think Floorplan Compiler will survive,
except for the name. Most of the Jupiter functionality will win out."
- [ An Anon Engineer ]
"I think Floorplan Compiler is at least a year behind, for instance,
Cadence's First Encounter. Product stability, feature completeness and
ease of use are still major issues. Why is it still so difficult to
integrate a PhysOpt flow with FPC?? And they still have this funky
'cluster-mode' which to me seems more for the benefit of Synopsys FPC
program writers, rather than the user base. Although, I have heard
they're going to hide it from the GUI in some future release.
Jupiter may or may not be more mature. I never knew too many people
using it. I would say there is a long way to go for Synopsys, on the
design planning front. I think by far, Cadence has the best solution
in this space."
- Neel Das of Corrent Corp.
"I haven't used Jupiter-XT or Floorplan Compiler. I have some exposure
to First Encounter, but couldn't compare it with anything other than
Jupiter (to which it is most certainly superior). I haven't used
Magma, or Monterey."
- [ An Anon Engineer ]
"We're currently looking at Floorplan Compiler and it seems as though
Synopsys is planning to keep Floorplan Compiler. Floorplan Compiler
seems to have a simpler methodology than Jupiter-XT as far as the flow
to generate a chip floorplan is concerned. Plus, the availability of
a "PrimeTime-lite" timing engine, a relatively painless interface with
PhysOpt, and a promised seamless interface with the other Synopsys
implementation tools make it tempting.
We are also planning to look at First Encounter, as they seem to have
the lion's share of the floorplanning and prototyping market.
- [ An Anon Engineer ]
"We evaluated Avanti Jupiter-XT, Synopsys Floorplan Compiler, Cadence
First Encounter. Cadence came out ahead in terms of speed, easy of
use, and quality of result. Hands down winner."
- [ An Anon Engineer ]
"Cadence First Encounter means: No documentation, learn by guessing
commands, see what is (or most often) what is not happening with your
design. Be surprised by netlist (logic) changes by simple DEF
operations. Above all: DEF/LEF. An absolute must for Cadence First
Encounter: hack GIGs of ASCII lines with your sed/perl/vi editor."
- [ An Anon Engineer ]
"I think the First Encounter tool is still at the front of the pack.
From a flow point of view and a high level capability, Floorplan
Compiler has the lead over Jupiter-XT. Its design partitioning is
much more in tune with the DC front end, and the low level stuff done
in Jupiter can be done in Astro. The Magma BlastPlan product was
really close in the market but it did not cover the full flow window
that was needed. Their new BlastCreate product has a better feature
mix and should perform well - but we won't know about that until they
roll it out at the end of the month."
- Pallab Chatterjee of SiliconMap
"Synopsys should have only one tool. Last time I have looked at
Jupiter-XT, it was not ready. And Synopsys Floorplan Compiler could
not compare with Cadence First Encounter.
I am using First Encounter, now SoC Encounter. It is a good tool to
evaluate feasibility (with some margin of error), but is still having
problems with all the new features that Cadence is trying to put in it.
For example, the power supplies connecting is mostly done in layout
mode because of bugs!!!
I dream about a tool where the power supplies will follow you when you
move sub-blocks around (memories, cores, etc). I have seen a demo of
Monterey Aristo IC Wizard doing this but have not used it myself. That
would be great!"
- Philippe Duquennois of Philips
|
|
