( DAC 04 Item 26 ) --------------------------------------------- [ 02/09/05 ]
Subject: Silicon Canvas Laker
DEATH TO VIRTUOSO! (Part I) -- Born in the shadows of the fabs of Taiwan,
Silicon Canvas Laker is the Virtuoso wannabe that seems to be getting the
most current interest from customers. I wonder how long it will be before
somebody buys Silicon Canvas.
Silicon Canvas Laker 3 is a layout editor very similar to Virtuoso.
Their Mcell is written in Tcl almost as powerful as Pcells are ready
now. Some cool Laker features:
- Query - polygon - info about area, perimeter, R and C based on a
table from process setup.
- Rule driven - info on DRC errors generated as you go.
- Easy reshape resistors, capacitors, guarding metal or active only.
- Now they've got a schematic editor from Cohesion.
- Interfaces with - AMS Designer for constraints forward and back
annotations.
- Direct interface with Milkyway - now with Magma - not with Cadence
as this is viewed as the real competitor.
- Very useful stick diagram with finger generation but no size, can
help for fast architecture exploration.
- Can copy symmetrical pattern from a group of devices to another;
equivalent to Cadence "clone" function.
- Can correct DRC - simple fixes.
- Associates imported layout with netlist and schematic - finds and
fixes errors.
- Matches creation with all orientation.
- Push wire - push and shove as CCT and VXL options.
- Biggest feature is that it can open big databases in minutes - very
useful for full chip revision or one layer touchup after P&R.
- Usability is the number one Laker priority - this is the place
Cadence lacks the most.
If you already have an infrastructure with Cadence Analog Artist and
Virtuoso XL it will be difficult for Laker to displace it. But Laker
can replace cheap seats. At a similar performance, Laker it is half
the price and Silicon Canvas is very aggressive to gain market share.
If they get some alliance with Sagantec, Magma, BindKey, etc., we may
see a new big player in the back end flow. Looking forward to see the
new competition.
Because of Silicon Canvas, Cadence already started to listen closer to
customers, work on quality and usability of their releases.
- Dan Clein, author of "CMOS IC Layout"
Comparing Laker and Virtuoso, even though IC 5.0 improved the
stream-in/stream-out performance and the speed of display, Laker
still has better performance in speed. Moreover, Virtuoso can't
provide enough editing layers and purpose numbers even with IC 5.0.
It is hard for me to prepare technology files for advanced processes.
I haven't had the chance to use the Pcell and Virtuoso XL, but the
demo of Virtuoso XL looks not bad.
- Brad Pu of TSMC
Like any tool it certainly will take some time to get used to the exact
look and feel, but Silicon Canvas did a good job in matching many of the
Cadence bindkeys/feel, so most Cadence users can do something pretty
quickly in Laker.
We are a Cadence-based group but did some work in Mentor IC Graph (under
duress I might add). Mentor was much harder for a cadence user to get
used to the look and feel.
The Laker strengths I think are the amount of features available in the
tool. To get a similar set of features you would need Cadence VXL+CCAR.
In addition you can edit Avanti mwlibs so your Cadence experts can
manipulate the data without needing to know Astro/Apollo.
Laker Cadence
------- ---------
device create/placement driven by the schematic VXL
schematic connectivity/flightlines VXL
guard ring create VXL MPP
guard ring editing (beyond chop) not sure about VXL
align to rows/columns VXL
align based on shapes (vertices/centerlines) not sure how well
VXL does this
point_to_point router CCAR (perhaps VXL)
router CCAR
bus create option off of basic path create VXL MPP + some
Skill code
DRC driven path create VXL
short finder not sure in Cadence
editing of Avanti mwlib no
interface to things like DeBussey no
large data stream/view another tool
(don't know name)
I think there is some user interface work that could be done in Laker,
but in general they appear to be very receptive to additions. We have
not asked for any since we are just evaluating. One example ... the
path command also allows for bus creation. You just change the number
of wires, etc. - sounds good but really creating a bus is separate
mindset for the layout people. You probably want a menu that remembers
your widths/spacing/via choices, etc. Bundling everything on the "path"
command is nice but makes that one menu somewhat complex. I would write
a new menu dedicated to bus create. Not that big a deal but the point
is they have a lot of functionality and I think it could be presented
in some simpler manner.
- Scott Lowrie of LSI Logic
I just tried CCAR, it's wonderful.
- [ An Anon Engineer ]
CCAR is OK
- [ An Anon Engineer ]
Laker L2 vs. Virtuoso 4.4.6: Laker has better performance in
stream-in/stream-out process, and the MCell of Laker is easier to
use than PCell of Virtuoso.
Laker provides functions that Virtuoso can't deliver:
- Real time rule-driven capability to produce DRC-free layout
- Efficient creation of bus and text array
- "Point to Point Route" with high-speed routing capability
- Practical "Align" function in implement layout,
- "Hierarchical Net Tracer" to debug and identify hierarchical
net at a glance
In the ability of measuring objects, Laker offers these measurement
functions to measure the object area, perimeter, distance, gate W/L
and the coordinate of pad window.
- Tommy Huang of UMC
Our group has been using both Virtuoso for a long time and we like the
flexibility of customizing the environment with SKILL. However, as we
start working on large GDSII files from customers, we found Virtuoso
became less efficient in dealing with their data.
Laker was recommended by an engineer who had experience with it.
We contacted Silicon Canvas and brought in Laker for evaluation. The
evaluation was done with the same GDSII file and with the same server
and configuration (Sun Ultra Blade dual 800MHz CPU & 4GB RAM machine.)
The only variable was the tool used.
Memory required for the stream-in process:
GDSII File Size Memory Usage
--------------- ------------
Laker ver 3.0 4.5 GB 0.433 GB
Virtuoso MSR3_IC50 4.5 GB 2.361 GB
Top-Cell load up time:
Laker ver 3.0 09:54
Virtuoso MSR3_IC50 13:14
All level top-Cell viewing Result
----------
Laker ver 3.0 successful
Virtuoso MSR3_IC50 failed. tool crashed.
After stream-in database size:
Laker ver 3.0 0.9216 GB
Virtuoso MSR3_IC50 2.8660 GB
Laker allows us to view the top cell and to support our customers.
After the evaluation, we adopted the Laker and use it as an alternative
to Virtuoso in our daily operation.
The downside in using Laker is switching cost. Similar to Cadence
Virtuoso, Laker requires a technology file to set up the appropriate
environment. For each process technology node, there are different
flavors (e.g. 1P6M, 1P7M, 1P8M for 0.13um.) Each flavor requires a
technology file for the setup. It takes time to develop the completed
sets of tech files to support different technologies. In addition,
Laker backbone language is Tcl/TK instead of SKILL. In order for us
to tailor Laker to the same level of customization as Virtuoso, our
engineers need to pick up Tck/TK and develop the equivalent SKILL
scripts in Tcl/TK.
- Winson Young of Chartered Semicunductor
I can only really give you an opinion on the circuit level stuff; I
think Laker has higher level capabilities, but since I live at the
electron level, that's mostly what I was looking at.
Basically I've been a Cadence user for over 13 years. I have become
more intimate with SKILL code than I care to think about, and have had
lots of experience coding up graphical Pcells, and have used some of
the newer ROD Pcells. We have dabbled a bit into Virtuoso XL, but it
always seems to have it's share of problems, coupled with the
limitations of Pcells.
Laker Pros:
The Laker stuff that I've seen and played with, albeit for only a day
or so, impressed the heck out of me. Their Mcell is so much better
than a Pcell, without the necessity of having a PDK team create it.
Much more flexibility than exists within a Pcell. I believe that making
customizations to it are not as complex as dealing with Pcells either.
However I haven't really gotten to that point, so it's just my first
high level impression.
Speed -- it seems that Laker is a little faster than Virtuoso, but
without having a complete apples-to-apples network structure I can't
totally say it's a Cadence issue in this case.
SKILL support -- Laker apparently has the ability to read in SKILL
code and preserve your investment of custom functions. I was able to
navigate around their database using SKILL-like commands, so support
teams wouldn't have a huge learning curve with a new language.
Interface -- as a Cadence user, Laker was very easy to navigate, many
of the same menu features, they have made it feel similar enough to
Cadence that you can jump right in and start working.
Parameterized Cells -- these are supported in Laker also, so you can
build custom cells.
Laker Cons:
Versioning -- As far as I know, there is no versioning system.
Mcells -- they re not perfect, and do have some limitations. (Still
better than Pcells, and actually exist vs. Cadence's future equivalent)
Open Access -- Silicon Canvas does not support Open Access, but they
do have integration into Milkyway.
VPA -- Cadence has volume purchase agreements, and it might be hard to
compete with price.
Overall, I am very intrigued by Laker. They also have a schematic entry
tool that I'd like to investigate. I only was able to play with Laker
for a short time, so there may be hidden issues I'm not aware of. I am
definitely going to bring the tool in for a closer look, and see if it
will help us out in out custom macro design work.
- Bill Goodall of Qualcomm
Laker integrated seamlessly into our flow and the Calibre interface
makes it seem as if Calibre is part of the tool. The Tcl programming
language is a nice tie to real world languages, but if a user has custom
SKILL scripts they will need to convert them to Tcl unless the function
already exists in Laker. I found some of the functions I had customized
in Virtuoso were already in place under other pull-downs.
Although both Virtuoso and Laker look & feel similar, I would recommend
even a seasoned custom designer take the time to explore all of the
functions in Laker. Don't compare the tools on a one-to-one command
basis, but look at the functions of the tools as a whole and advantages
of Laker will appear.
Laker support has reminded me of my days at IBM when we had EDA tools
group on our site. I've already had several suggestions show up in the
next release of Laker, something unheard of with other EDA vendors.
Laker technology files can be created with support by Silicon Canvas
AE's, but lately Laker tech files are showing up in fab house process
kits. It seems the tech files are running slightly behind the leading
edge, so users may want to see what is available at their fab.
We enjoyed a substantial cost savings. One of the hardest parts of
switching to Laker was to convince management it was truly a
replacement for Cadence Virtuoso.
- Rodney Hengst of eSilicon Corp.
I have used Laker, but in a non-mainstream way. Laker's main customer
base is physical designers, whereas I'm interested in Laker's ability to
link with a design debug tool like Novas Debussy/Verdi.
I have compared Laker mainly against Knights Maskview & Credence Layout
tool, which is what our customers are using. Laker's user interface is
not really tuned to Credence (which, in a nutshell, is a CAD link with
emission microscopes and laser voltage probers for silicon debug).
Laker's ability to link with Debussy after a very short processing phase
(10 to 15 minutes) did much better than Mentor DESIGNrev's link with
Debussy, which required a re-run of LVS, which can take hours and lots
more expertise than our users have.
Also, the fact that Laker reads database formats from all the big EDA
vendors: Mentor, Cadence, Magma, Synopsys... by reading LEF/DEF,
Milkyway, Open Access is a big plus. Our users can throw just about
anything at us. Credence equipment is everywhere.
- Cathy Kardach of Credence
Index
Next->Item
|
|
