( ESNUG 426 Item 5 ) -------------------------------------------- [03/31/04]
Subject: Ten More Letters On Mentor/Exemplar/Leonardo/Precision
> Scenario 1: Wally's Revenge
>
> Synplicity was founded by ex-Mentor synthesis gurus Ken McElvain and Andy
> Dauman. For Wally Rhines, the CEO of Mentor, it's a wee bit embarrassing
> to have his ex-employees best him in a market that Mentor used to own.
> If it turns out that Synplicity sales really did drop 31% while Mentor
> sales remained steady in FPGA synthesis, Wally gets to laugh as he watches
> Bernie Aronson, the CEO of Synplicity, do his best imitation of the
> Help-I've-Fallen-And-I-Can't-Get-Up lady. Sweet, sweet revenge....
>
> - from http://www.deepchip.com/gadfly/gad031204.html
From: Bin Liu <binliu=user domain=atmel spot balm>
To me, Leonardo Spectrum is a very good VHDL/Verilog Synthesis tool. In
the past, I tired different synthesis tools such Synplicify, DC shell
Synopsys, EDS, FPGA Express... etc. In fact, in comparison with other
synthesis tools, I found that Leonardo Specturm has more features than
other tools and it is easy to use.
However, Leonardo Specturm still contains bugs such as created incorrect
netlist for design code that is synthesized for targeting our AT40K FPGA
FreeRams. This problem has been reported to Mentor Graphics. However,
Mentor Graphics does not seem to be able to fixed it now. Also, some
standard VHDL syntax seem to be ingored by the compiler of the Leonardo.
So, in some cases, simulation and synthesis could be mismatch.
Nevertheless, I still like this tool because it is better than other
Synthesis tool that I have tried before.
For Precision, I saw the demo here before. However, I have not obtained
the license to try it yet. Our marketing group is working with Mentor
to see if we could obtain the evaluation license to try it here.
- Bin Liu
Atmel Corporation San Jose, CA
---- ---- ---- ---- ---- ---- ----
From: [ An Anon Engineer ]
Please keep me anonymous.
The project I am working on is my first end to end design. I have used
Synplify with Actel devices previously. For this program, I used
Precision RTL. I found it fairly easy to use, and it worked through my
design flow in the tool set nicely. I did have an issue with trying to
get parts of the design constrained properly for timing, which Mentor
helped up enormously. There are a few items I would like to see
enhanced. I had wanted to turn on register re-timing on just a specific
block, but the setup was not advantageous, so I changed my code instead.
I was synthesizing for Xilinx Virtex parts. I found the entire
experience good with the tool. There was a learning curve, but it was
fairly short. Once I got all of my SDC files and project in place, I
was able to crank out multiple design cycles in a single day (during I&T
of the design in the lab). Overall, I would rate the tool a B+ to A-.
I liked the ease of flow since I was using the Mentor Tool Suite for
everything (HDL Designer, ModelSIM, and Precision RTL). I am also
pleased with the results I had that I was able to flow into the Xilinx
PAR tools.
- [ An Anon Engineer ]
---- ---- ---- ---- ---- ---- ----
From: Debabrata Banerjee <dbanerjee=user domain=lucent spot balm>
John,
I have used Leonardo Spectrum for synthesis, FPGA Adv for block diagram
view of the VHDL implementation and also ModelSim Simulator. So far my
experience has been good. I have also ran ModelSim directly from
XILINX ISE.
Although I am happy about the performance I think Leonardo messages for
debugging can be improved. I have used Synplify tools and?I feel
that their debug messages are better than Leonardo. Their tech support
had been excellent specially Guy Wallin who had helped me so many times.
I like Mentor tools because I do both Synthesis and Simulation.
- Deb Banerjee
Lucent
---- ---- ---- ---- ---- ---- ----
From: [ An Anon Engineer ]
Hello John,
First of all, please keep my comment anonymous.
I didn't have any problem while using Exemplar for synthesis of up to
1 million gates.
My impression is that many strange problems appear when the part (FPGA)
is not chosen correctly for the specific application or if the code was
not written good enough.
My general approach is as follows: If the part matches the application
and the code is of high quality, the problem doesn't occur and has not
be solved. Designer has not to cause a problem. Exemplar has not to
solve a problem.
- [ An Anon Engineer ]
---- ---- ---- ---- ---- ---- ----
From: [ An Anon Engineer ]
John,
BTW, please keep anything I say anonymous.
I'm just finishing up on a Xilinx XC2VP30 design that is very full. I have
been using Leonardo Spectrum for a number of years, and am very comfortable
using it, so it is my synthesizer of choice. With it, I am reaching a LUT
utilization of 84%. Experience has shown that much more than this and I
won't be able to close timing.
So I decided to give Precision Synthesis a try. I couldn't get it to
successfully synthesize my design. Many attempts to work through the
problems with the local FAE were also unsuccessful. Eventually they're
bound to get Precision working. But as yet, it doesn't appear to be a
viable replacement for Leonardo.
Next I decided to give Synplicity a try. I was able to successfully
synthesize my design, but the LUT utilization came in at 100%. Well, I'm
new to this tool, so I decided to seek the help of the local FAE. He took
my design, worked on it for a while, and also couldn't do any better.
"I'm going to talk to the folks at the factory" he said. And that was the
last I heard from Synplicity.
So my design is going out of the door using Leonardo as a synthesizer. But
if Mentor has Leonardo on the dead end path that it appears, I don't know
what to use for the next project. I suspect it may be Synplicity since
they seem to be the up and coming player that everyone else is using.
- [ An Anon Engineer ]
---- ---- ---- ---- ---- ---- ----
From: Albert Lindmeier <albert.lindmeier=user domain=elca.de>
Hi, John,
I'd like to confirm that we are using using Precision-RTL successfully too
for months.
- Albert Lindmeier
El Camino GmbH Germany
---- ---- ---- ---- ---- ---- ----
From: Clive Rolston <rolston=user domain=radstone.co.uk>
Hi, John
Overall Precision works well, but I have the following problems with it.
1. Spurious crashes - "program error has occurred and program will
now close". Happens intermittently.
2. Stop button does not work. Try and stop it whilst compiling or
synthesising and it says "Tool cannot be closed while program is
running". Only way to stop it is using TASK Manager.
3. Adding Xilinx UCFs. Can't have any strange characters in UCF like
| IOSTANDARD = LVCMOS25
Precision doesn't like the |.
- Clive Rolston
Radstone Technology UK
---- ---- ---- ---- ---- ---- ----
From: [ An Anon Engineer ]
Hello John,
We use Leonardo Spectrum for FPGA design and IP/ASIC block checking.
Pros:
- For FPGA it is very easy to use. Reasonable results.
- The Insight graphical viewer is very useful to look at structure of
3rd-party IP and is reasonable for tracing signals through the
design heirarchy.
- For ASIC block checking, we run RTL through Spectrum to verify
synthesizability of our own IP. Its parser is good at catching most
structural & syntactical design gotchas.
- Works consistantly for VHDL & Verilog.
Cons:
- The Verilog 2001 support is not complete, with "instances of arrays"
causing errors. This has been flagged with customer support for a long
time, but has still not been resolved.
- No native support for Synopsys .LIB files/scripts, etc.
- The Mentor library convertor for Synopsys .LIB files is pretty poor,
and appears to be un-supported these days. This is something that we
have needed on a number of occasions, and have had problems with
incomplete library translation and errors. We have raised it with
customer support a number of times but it still hasn't been resolved.
Overall:
Lenoardo Spectrum is a reasonable FPGA synthesis tool with a good mix of
features, but does feel as though its support/development is being ramped
down, probably in favour of newer tools such as Precision Synthesis.
I hope this helps.
I would like to remain anonymous.
- [ An Anon Engineer ]
---- ---- ---- ---- ---- ---- ----
From: Kathryn Rickard <kathryn_rickard=user domain=xyratex spot balm>
Hi, John,
I've not been overly impressed with Precision so far. We have used
Leonardo Spectrum for many years for our FPGA designs with satisfactory
results. We have a couple of seats through Saros but also, as a mainly
Altera site we made good use of the OEM Leonardo, allowing each engineer
to have a seat. When this was withdrawn it was a big disappointment
especially as the stated reason was because Leonardo was being
superceded by Precision. We were given every indication that it was not
a good idea to stick with Leonardo because all development effort was
being targeted at Precision and Leonardo would be obsolete at some
point. Because of the incompatibility in scripting it needed to be an
all or nothing shift to Precision in order to allow efficient use of
licenses. We have bought an additional Precision seat and have a
temporary Precision license on our L3 Leonardo seat - for which I have
been intended to take advantage of the 'no-cost' 'upgrade' (quotes round
the former because it costs extra on the maintenance and quotes round
the latter because as described later its not necessarily an
improvement). Through recent discussion with Saros over maintenance
renewal I've been told that Leonardo is not intended to be obsoleted,
and that many customers are in fact staying with Leonardo.
Because of the legacy design situation, the expiration dates on the OEM
licences, and general tendency to stick with what you know we haven't
ported all designs to Precision yet. However, the designs that we have
tested through the new tool have disappointing results - in many cases
the design is larger or slower, the tool crashes, the language support
is not the same, and various inferences that would be expected are not
made - see comments from one of my collegues...
I've found Precision to be un-reliable, crashes with my X decode-filtering
block occasionally. It also fails to correctly recognize RAMs in the same
was as Leo currently does.
Neither Leo 2003a33 or Precision 2003c78 seam able to build the large
shift registers/delays using the Altera RAMs present in my decode-filtering
block. They do however, report finding the shift registers?
I've received opinions from our Altera FAEs that Leonardo/Precision are
falling behind in synthesis quality of results: they suggest Synplicity
is much better and that Altera's own synthesis under Quartus comes in
after Synplicity but ahead of Leonardo or Precision.
With all of this I'm generally unhappy with the whole synthesis thing.
We need to standardize on one tool and would like it to be Leonardo or
Precision because we already have investment in tools, legacy designs
and knowledge. However, it looks like this may not be the right thing
to do.
- Kathryn Rickard
Xyratex Havant, UK
---- ---- ---- ---- ---- ---- ----
From: Stephane Debieux <stephane.debieux=user domain=actis-computer spot balm>
Hi, John,
About two years ago we bought schematics and PCB tools from Mentor and as
part of the deal we get their FPGA package tool suite called FPGA Advantage
which includes ModelSim PE, Leonardo Spectrum Level 2 and HDL Designer.
I had to design a 60x-VME Bridge to interface the PowerQUICC II 60x bus
with the VMEbus. I designed in VHDL using ModelSim and Leonardo Spectrum
in standalone mode. Though HDL Designer provides a single frontend
interface with some kind of graphics capability for design from where you
can launch simulation and synthesis, I preferred to run the tools in
standalone mode using the most basic text editor. I wanted to be tool
independant without learning how to use HDL Designer and having the
complete control over my design to decide exactly how my VHDL source
code must be.
Leonardo Spectrum is relatively straightforward to use, you can either
select a basic flow 'push-button like' interface for simple designs or
an advanced flow for specifying detailed constraints.
The selected FPGA was an Altera Cyclone and Leonardo worked very well with
my design without requesting many iterations. I used the EDIF out of
Leonardo as input to Altera's P&R tool Quartus II.
Once I have run Leonardo with successful results, I can use a script in
case of source code change and get easily a new EDIF with a single click.
My opinion is that a big effort on Library Modeling still has to be made
to get synthesis results closer to what the P&R tool gives. I don't even
mention the still long existing gap between synthesis and P&R.
- Stephane Debieux
Actis Computer Switzerland
|
|
