( ESNUG 426 Item 2 ) -------------------------------------------- [03/31/04]
Subject: Ten Letters On Synplicity Synplify/Synplify Pro
> Scenario 2: Bernie's Vindication
>
> It's embarrassing to have Dataquest publically state you've slipped 31%
> while your biggest rival (Mentor) held steady. If you can publically
> catch Mentor lying or exaggerating here, it means that Bernie gets to
> laugh while Wally does his best explaination of a "wardrobe malfunction"
> to the entire industry. Sweet, sweet vindication...
>
> - from http://www.deepchip.com/gadfly/gad031204.html
From: Steve Weir <weirsp=user domain=atdial.net>
John,
I have been solidly in the Synplfiy camp since 1996 and have never turned
back. I looked at Exemplar a couple of times when Synplify had growing
pains around version 5. There is just no contest. I don't know of
anyone using Mentor. I know they are pushing Precision really hard, but
Synplify / Synplify Pro just seem unassailable. Ken is an absolute genius.
- Steve Weir
---- ---- ---- ---- ---- ---- ----
From: [ An Anon Engineer ]
John,
Better keep me & co. anon for the paranoid management.
What methodology did Gary Smith use to collect this data? Did he simply
asked the vendors?
We use Synplicity (not Pro we're too cheap) but now we have access to
Leonardo through Alcatels' corporate licensing. Is it possible Mentor
charges big corporations very little for Leo as part of a package deal
then reports higher numbers? (Of course, this could spoil the numbers
for other products).
Wouldn't number of seats be a better metric since Synplicity has essentially
3 products and each win is a win where Mentor has, what, 100 products? And
each (big corp.) win is 100 wins?
We could derate Mentor appropriately.
Besides, even with Mentor's advantage we have 1 Leo license (that never gets
used here in CA) vs. 4 Synplicity licenses that get used constantly.
- [ An Anon Engineer ]
---- ---- ---- ---- ---- ---- ----
From: Hans Roeven <hans.roeven=user domain=philips spot balm>
John,
I have been using both tools in the past year and a half, on two projects,
both involving prototyping of new processor designs on FPGA. Target
devices: Xilinx VirtexII and Altera Stratix.
Our "standard" tool has been Leonardo Spectrum for years. I use it only
with tcl scripts because in my opinion the graphical user interace is not
very good (to put it mildly ;-).
In the first project I had looked at Synplify Pro for the first time as
an alternative, but I had to revert to Leonardo because Synplify Pro gave
incorrect logic results (!). However I had major issues with Leonardo
because memory inference didn't work properly. Used "manual instantiation"
of Xilinx memory blocks in the end.
In my current project I started out with Leonardo again but now I *had* to
move to Synplify Pro because Leonardo kept bailing out on a certain portion
of the VHDL code, producing incomprehensible error messages. Could not
solve it. With Synplify Pro I had no problems compiling the same code,
although I did have problems with memory inference. The Altera RAM blocks
are not used optimally. Had to instantiate memory blocks generated with
Quartus II. This is quite easy to do but makes the design technology
dependent. And boy what a *great* GUI. Exemplary :-)
- Hans Roeven
Philips Sunnyvale, CA
---- ---- ---- ---- ---- ---- ----
From: Jay Dowling <jaydowling=user domain=aol spot balm>
John,
I used to use Leonardo (came free with the Altera tool, but then stopped),
as a sanity check. It had way too many tweeks to adjust (not that any of
them produced better results). Synplify blew it away in run time speed,
size and speed of results. Plus there were language support issues with
Leonardo so none of those runs ever made it into a product.
- Jay Dowling
---- ---- ---- ---- ---- ---- ----
From: Cliff Cummings <cliffc=user domain=sunburst-design spot balm>
Hi, John -
Great topic. My $0.02 below, but a similar topic that would be worth
exploring is, "What percentage of FPGA designs are being done with VHDL?"
I corresponded with Gary Smith in May 2003 and Gary reported:
"We are seeing more and more FPGA Designers using Verilog now, but
still the last time we checked (in 2000) 82% of all FPGA Designers
(that were using RTL simulation) were using VHDL."
In the absence of better data, I had to concede the point. My gut feel
(very scientific!) is that it is a lot closer to 50%-50% these days. I do
FPGA Verilog Synthesis training for so many large companies that I have a
hard time believing that Verilog FPGA design only accounts for 18%-30% of
the FPGA HDL pie.
Synplicity -vs- Leonardo-Exemplar
In the past two years, I have probably done ~10 FPGA Verilog Synthesis
classes and no Leonardo-Exemplar FPGA Synthesis classes. Going back past
two years, I taught a few Leonardo-Exemplar FPGA synthesis classes but the
ratio was still probably 20% Leonardo-Exemplar and 80% Synplicity.
I actually like both tools, but nobody has asked for the Leonardo-Exemplar
version of the training in at least two years. It could be that Mentor is
satisfying all of the RTL Synthesis training needs of the Leonardo-Exemplar
customers or it could be that the tool is just so easy to use that
engineers do not need RTL Synthesis training using Leonardo-Exemplar.
When I talk to FPGA designers, the tool set that is mentioned most often is
Synplicity. Once in a while an engineer mentions Leonardo-Exemplar.
Another data point is that Mentor has moved more towards their Precision
FPGA synthesis tools, which may be part of the Mentor numbers(?)
- Cliff Cummings
Sunburst Design, Inc. Beaverton, OR
---- ---- ---- ---- ---- ---- ----
From: [ An Anon Engineer ]
Hi John,
You'd best keep me anonymous.
On three occasions over the past 4 years we've done an evaluation of the
various synthesis tool options available to us (we use a mix of Altera Apex,
Apex2 and Stratix FPGAs).
About 4 years ago we did evaluations of Synopsys FPGA Compiler 2, Mentor
Leonardo Spectrum and Synplicity Synplify. We rejected FPGA compiler out of
hand due to poor language support and regular crashes. Leonardo Spectrum
and Synplify produced very similar results, so we went with Leonardo as at
the time Altera gave it away free with their Quartus software.
About 2 years ago we did evaluations of Leonardo Spectrum and Synplify.
This time round we found that Leonardo Spectrum produced netlist which were
20-50% bigger than those produced by Synplify, so we made the switch to
Synplify.
We've just completed an evaluation of Quartus native synthesis, Precision
and Synplify. The Quartus native synthesis has improved dramatically over
the last year (in that it actually manages to parse our code) but produces
netlists which are on average 50% bigger than with Synplify. Precision now
correctly compiles our code (early versions crashed pretty reliably) but
still produces netlists which are 20-50% bigger than those produced by
Synplify. We therefore have no plans to move away from Synplify in the near
future.
As an aside we also receive great support from Synplify, even though we
aren't based in the US. With Mentor our support questions just seemed to
disappear into a black hole (via our reseller) from which they never
reappeared.
- [ An Anon Engineer ]
---- ---- ---- ---- ---- ---- ----
From: Gerhard Cadek <cadek=user domain=oregano.at>
Hi John,
just to response to your questions on user's experience on Mentor's FPGA
synthesis tools. We compared in the past years Leonardo (Level 3) vs.
Synplify Pro (and later on Precision as well) several times.
We have run several test designs - most of them quite large and timing
driven. All our test designs showed the same results:
- Synplify had faster synthesis run times (factor 2 to 10!)
- Synplify results showed better timing (approx. 10-15% on average)
- Synplify also showed smaller area (while acheiving better timing)
- Leonardo reduced the gap in the past years, but there is still a
large gap
- Precision offered an improved user and scripting interface compared
to Leonardo, but the results were somewhat worse than those from
Leonardo
We also know some few designs with compareable results and some rare
designs with better results from Leonardo compared to Synplify. On the
average Synplify beats Leonardo and Presicion.
- Gerhard Cadek
Oregano Systems Vienna, Austria
---- ---- ---- ---- ---- ---- ----
From: Eric Robin <e.robin=user domain=giat-industries.fr>
Hi, John,
We've been using Synplify since 3 years now, because of its ease of use,
his uncomparable runtime and performances. Despite the very attractive
offers from Mentor Graphics to sell Leonardo, we remained faithfull to
Synplicity and have been standardized our design flow with Synplicity on
all our electronic development sites recently. The excellent relationship
entertained with Synplicity France and the support quality remain to us
we've done the best choice.
- Eric Robin
Giat Industries France
---- ---- ---- ---- ---- ---- ----
From: Nicolas Maury <tns=user domain=smpfrance spot balm>
Hi, John,
Here are the reasons about our migration from Mentor to Synplicity.
The incomparable performances, the dramatic speed of synthesis process
and the ease of use that has been proven with Synplify Pro use have
driven us to choose to migrate from Mentor to Synplicity. We consider
Synplify Pro as the most mature synthesis tool of the market with new
advanced technology functionnalities to increase productivity and QoR.
No matter the structured size you work for, you receive an excellent
technical support with customised advices and I really appreciate this
as well. All of this have guided me to choose Synplify Pro which is
the reference today to follow in the synthesis market.
Nicolas MAURY - ASIC / FPGA Designer - SMP for Syt=E8mes
Midi-Pyr=E9n=E9es -=20
For 30 years , SMP has implemented each new technology in order to meet
market requests in the field of telecommunications, data transmission =
and data processing (telecommand & telemetry) by supplying advanced =
equipments.=20
- Nicolas Maury
SMP Toulouse, France
---- ---- ---- ---- ---- ---- ----
From: [ An Anon Engineer ]
John
Please keep me anon. We use Synplicity for synthesis of Xilinx Virtex2
and Virtex2 Pro devices. In 2002, we received Leonardo for free with
ModelSim. We tried Leonardo and found it took longer to synthesize, and
produced a larger, slower result. We really like ModelSim; don't like
Leonardo; have not tried Precision yet, and like Synplify.
Regarding the 2002 FPGA synthesis market, my opinion is the following:
1. FPGA synthesis tools are cheap and easy to use
2. Low cost, ease of use means it is easy to switch between tools
3. Big companies buy perpetual licenses, pay yearly maintenance
4. Big companies buy all the synthesis tools because of reason 1
and 2
5. Start ups typically buy one tool under a time based lease
6. I don't have to tell you, 2002 was bad year for start ups
Based on the above, I suspect a lot of start ups bought one year
Synplify TBLs in 2001. In 2002, those TBLs were not renewed. Since the
big companies just continued paying maintenance on perpetual seats, and
they have both Mentor and Synplicity tools, perhaps the Dataquest
numbers are correct.
- [ An Anon EDA Engineer ]
|
|
