( ESNUG 426 Item 4 ) -------------------------------------------- [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: Laurent Hausammann <lhausammann=user domain=yminds spot balm>
Hi John,
My experience with LeonardoSpectrum and Precision Synthesis is not big
but what I can say today is:
- I'm using both tools everytime in order to compare results. So, there
is no sensible difference in the result. Therefore the manner of how to
set the constraints and everything around, before starting a synthese,
is more intuitive in the Leonardo compare to the Precision. The
Precision constraint settings are very poors (Or I still miss something
in the procedure to do it).
- The Precision Synthesis has the Verilog 2001 support which is one of
the reason why I will use it instead of Leonardo. I heard that there is
no plan for Leonardo to support Verilog 2001.
- The Leonardo Spectrum is more stable.
Due to the fact that Verilog 2001 is a big advantage, I will push myself to
use the Precision Synthesis as the first tool to use.
My configuration today is a licencied Precision Synthesis tool + an
archival license Leonardo Spectrum.
- Laurent Hausammann
Yminds SA Switzerland
---- ---- ---- ---- ---- ---- ----
From: Aaron Byman <aaron.byman=user domain=elektrobit spot balm>
> So, if you're currently using Mentor/Exemplar/Precision/Leonardo, please
> send me an email. Did you intentially buy the Mentor tool?
Yes we intentionally bought the tool. We as a company have been using
Leonardo for many years with great success. I find it the best tool for
running FPGA synthesis in batch mode via tcl scripts. I don't need the fancy
GUIs or the "one-button" synthesis very often. I open the Leonardo GUI very
seldom (to view the RTL schematic) if there is a problem with how it has
translated my code. Other than that I run the tool in batch as part of my
incremental FPGA build script.
> What FPGAs do you synthesize in Exemplar/Precision/Leonardo?
Xilinx: Virtex, -II, -IIPro, Spartan, -II, CPLDs. Some Altera as well though
not as often as Xilinx.
> How does Mentor FPGA synthesis stack up against Synplicity?
I tried Synplify while we still had a license around. As far as results I
found it to be similar to Leonardo on the designs that I tried. The one
thing I didn't like at the time was that Synplicity didn't allow the setting
of top-level VHDL generics at synthesis compile time. Leonardo was the only
tool that allowed this. This feature is useful when running a bottom-up
synthesis flow. Many sub-blocks that contain generics can be "reconfigured"
at synthesis compile time. As luck would have it as soon as our company had
decided not to renew the Synplify license they added support for this very
feature... As a side note Mentor's new tool, Precision RTL, does _NOT_
include this functionality. Precision does not support bottom-up flow very
well. I have asked them about it a number of times but it does not seem to
be high on the priority list.
> Given the choice, would you buy Mentor/Exemplar/Precision/Leonardo
> again today? Why or why not?
I would renew my Leonardo license but I would not buy a Precision RTL
lisence. In addition the the above mentioned reason my attempts to run
Precision have been unsatisfactory. This is somewhat tolerable as the tool
is quite young still. I was expecting a little more as they had a decent
starting point with Leonardo.
That said, I am quite comfortable buying from Mentor as the support is
fantastic. Questions are answered the same day. Many times a response is
recieved within an hour. (and I mean a real response, not the automatic
reciept type) The support people seem to care about your problem and work
hard to find a solution.
- Aaron Byman
Elektrobit Ltd.
---- ---- ---- ---- ---- ---- ----
From: Siegmar Zander <szander=user domain=husware.de>
Hi, John,
I use Leonardo Spectrum Level 2 2003b.25. I use this tool in conjunction
with the HDL Designer and overall I have no or only small problems.
Normally I use the edif-output as input for Quartus. The quality of output
is good, but for not too big and not too complex designs the sysntesis is
not much better than the systesis of QuartusII V3.0 with SP2.
You can have problems when using buses with not symetrical use of all
lines. For example if you have an 16 bit bus with 8 lower bits read,
write, 3-state direction and 8 higher bits which only have the write
direction enabled or 3-state than leonard will split the bus and renames
the lines to single bits. If you after this will use the gate level
netlist in a testbench, the signal names will not match with the orginal
names. Mentor will fix this, and there are also a workaround, but I use
for this design now the Quartus syntesis.
- Siegmar Zander
---- ---- ---- ---- ---- ---- ----
From: Georg Dollinger <georg.dollinger=user domain=siemens spot balm>
Hello, John,
I am a Hardware Design Engineer at Siemens and we are using Precision-RTL
successfully for months. We used it mainly for XILINX FPGA's (Gigabit
Ethernet Linecards) VirtexII and Virtex Pro with Core Clock frequency
of 104 MHz!
- Georg Dollinger
Siemens
---- ---- ---- ---- ---- ---- ----
From: Hans Sahm <hsahm=user domain=slds1.de.lucent spot balm>
Hi John,
We at Lucent are using the Mentor FPGA synthesis tools Leonardo and
Precision-RTL for our projects. The tools are available as part of a
global Mentor-Lucent contract.
- Hans Sahm
Lucent Technologies Nuernberg, Germany
---- ---- ---- ---- ---- ---- ----
From: [ An Anon Engineer ]
Hi John,
Please keep me anonymous but here is what one of our engineers thinks of
Leonardo:
Leonardo Spectrum on Linux (FPGA synthesis only, no ASIC) v2003b65
Very good GUI, most commands are there where you expect them to be.
Exists in win32 and Linux platforms (I had worked with the Windows version
some years ago, and the Linux has the same feel)
Reasonably fast synthesis for large designs (filling up XC2V6000 FPGAs)
During optimization, window refreshing is intermittent, and it appears to
hang on large modules.
Mixed design support (Verilog/VHDL), one has to be careful to discern what
is actually happening. Leonardo can remove blocks without any warning (in
our case, the synthesis script had the VHDL wrapper after the Verilog module
it was "shelling" and the Verilog module was just removed, not even creating
a blackbox! This was solved by putting the wrapper before the module in the
script)
Leonardo can happily let problems go through the flow. We had a clock
divider using a state machine that seemed to be synthesized OK, but it
crashed place and route. Upon closer inspection (synthesis of this module
only), this module was not properly synthesised. No warnings again, but
schematics obviously not matching RTL. Solved by rewriting this clock
divider.
I cannot really compare with other tools for optimization (both duration
and area), but I think Leonardo could be pessimistic with area figures.
A design using 97% function generators utilization was still placed and
routed using Xilinx ISE, and a limited number of constraints. It could
be that other tools get better (more optimistic) figures but in fact are
too optimistic and the resulting EDIF cannot be placed and routed properly.
This is just a supposition.
The design we have is extremely large (filling several of the largest Xilinx
FPGAs) and an auto-splitting tool would be welcome. Leonardo is
future-ready, and even includes options to synthesize for devices not yet
available, like the king sized VirtexII Pro FPGA.
All in all, I find Leonardo is a great tool. But the log files need to be
a bit more detailed, to highlight what's happening especially in mixed
designs. Support has been excellent, the engineer even helping us to carry
on with Xilinx ISE, and the website being very useful.
- [ An Anon Engineer ]
---- ---- ---- ---- ---- ---- ----
From: John Lynch <jdlynch=user domain=cse.ogi.edu>
John,
We use Mentor Graphics tools in teaching digital design courses. Mentor
Graphics has been very generous in supporting our graduate education
program. We had been using Leonardo for FPGA synthesis, but recently
switched to Precision. Precision has an intuitive GUI which is easy for
students to learn; it infers Xilinx block RAMs and carry chains nicely,
and handles large (100K gate) designs easily. We haven't looked into
Synplicity's university program.
- John Lynch, Instructor
Oregon Health & Science University
---- ---- ---- ---- ---- ---- ----
From: [ An Anon Engineer ]
Hi John,
We originally purchased LeonardoSpectrum about 5 years ago before it was
aquired by Mentor Graphics. Even so, it was purchased as part of a
package deal with ModelSim and we have maintained maintenance on both
products ever since. Our original need was synthesis and simulation for
development of a gate array design (with Chip Express). The fact that
LeonardoSpectrum would also synthesize FPGAs was considered a bonus but
we really didn't know at the time if we would ever use it for FPGAs.
In the meantime we have done one more gate array design with Leonardo
based on the first design but about twice as big (~100K gates). I have
only used the tool sporadically since then. Another engineer here has
used it for quite a few designs in Xilinx Spartan (I,II,IIE,III) parts.
Currently I've been using Xilinx almost exclusively for FPGAs and the
XST synthesis tool has been doing the job. Most of my latest designs
have been in Virtex E. Once in a while, if I'm really pressed for real
estate I'll try running a design through Leonardo. One design I'm
currently working on is a very tight design in a Spartan IIE. Here we
had to use Leonardo over XST to get the design to fit, but the clock
speed is only 16Mhz. At the moment I'm much more comfortable
implementing detailed timing constraints in XST than in Leonardo. If
this was an 80 Mhz design with multi-cycle paths I don't think I would
be using Leonardo. One problem I've had in this regard is the
difficulty in specifying paths in Leonardo. I don't understand the whole
concept of the various "views" of a design, and why I as a user should
have to know anything about them. I would like to reference parts of
the design as close as possible to the way I entered them (as in XST).
We were told at one point that Mentor was going to change their approach
to specifying paths, but I don't know if they have. I confess I haven't
been very diligent in upgrading to the latest versions-always something
more important to do.
I guess I'd like to exercise my option to remain anonymous. Thanks.
- [ An Anon Engineer ]
---- ---- ---- ---- ---- ---- ----
From: Robert Morrison <robert.l.morrison=user domain=nasa.gov>
As a first time user of this software with no experience or training, I
was unable to utilize Leonardo Spectrum to synthesis my design. I did
not know enough then nor do I probably know enough now.
However as an input I would mention the difficulty I had in creating
internal memory modules through the Altera Megawizard. I chose a 240
pin FLEX10KE and was unable to create a simulation module for a dual
port RAM that could also be synthesized in Leonardo Spectrum.
I eventually resorted to compiling my program in FPGA Advantage then
synthesizing the resulting *.v files as a separate project in the Altera
Quartus II package. I had to rebuild the RAM module from within
Quartus II to get it to work.
I will certainly try Leonardo Spectrum in the next project, but my first
one lead me away from using it.
- Robert Morrison
NASA JFK Space Center, FL
---- ---- ---- ---- ---- ---- ----
From: [ An Anon Engineer ]
John,
Please keep these comments anonymous.
We purchased Precision as part of a package deal for Modelsim licenses.
We planned originally to get Synplicity, but we thought we'd give
Precision a try since our timing goals for the FPGA were very aggressive
and we liked the ability to perform some level of Physical Synthesis.
As we got further into the design, we found that the tool had a lot of
bugs and quirks, and was difficult to use, even as a Synthesis tool
only. Some of the bugs were so basic, we assumed the user base must be
very small for these problems to be there. Using the tool as an
integrated FPGA environment with Xilinx was not practical, since many of
the native options in the Xilinx Project Navigator could not be set from
the Precision environment, or the ability to do it was undocumented for
numerous options. The flow to use Mentor Physical Synthesis and do the
Place and Route in Xilinx stand-alone environment and iterate back and
forth was also very problematic. We were unsuccessful in making it work
iteratively between the 2 stand-alone tools or as an integrated
environment. In the end, we gave up and used it only as a Synthesis
tool. If we knew in the beginning that we weren't even going to use the
Physical Synthesis, we would have bought Synplicity instead.
We still like the concept of FPGA Physical Synthesis, but Mentor has a
ways to go to make it work. At one point we were considering going to
Synplicity, even though we had already purchased Precision. If we had
the time, and a better Synplicity sales guy, we would have liked to
evaluate Synplicity's version of Physical synthesis. If the Synplicity
sales guy wasn't so stingy with eval licenses, he may have sold us.
- [ An Anon Engineer ]
|
|
