Register or Login
News Categories
Advertisements
Synopsys: The Queen of Virtual Prototyping
In the last couple of weeks Synopsys has purchased VaST Systems and signed a definitive agreement to purchase CoWare Inc. Two things come to mind. First of all I have been writing for almost a year that Synopsys had too much cash on hand and should do something useful with it. Second, with both acquisition, Synopsys has practically cornered the system level virtual prototyping market.
Using Up the Cash
I started saying the Synopsys should use the cash in more productive ways than just get interest payments when its cash position was around $800 million. Now it is around $1.17 billion, and so it is probably time to do something before the company gets regulated as a bank by the Federal Government. Here is how Jay Vleeschhouwer commented the acquisition:
"Synopsys has announced two small acquisitions (VaST Systems, CoWare) to build its position in electronic in system-level design and "virtual prototyping". The combined transaction values, we believe, represents only a small portion of the company's nearly $1.2 billion in cash. VaST will be immaterial to results while CoWare, a roughly $30 million company, may have some impact."
The acquisition is certainly good news for CoWare, a company that has always had great ambitions in the ESL market, but that, like many smaller EDA companies, found it both hard and expensive to compete with the likes of Synopsys and Mentor for the attention of very large customers. The problem, in this case, is the usual one that keeps the EDA industry from growing as it should.
Pricing in this industry has always been dominated by "bundling" strategies. Cadence, Mentor, and Synopsys have, and continue to maintain their priority accounts with licensing policies that significantly discount the individual prices of their tools in order to obtain a "corporate wide" purchase order from their large customers. This means that small companies cannot compete in those accounts and stay in business, since they can only offer one tool that may be a bit superior to the incumbent tool, but must be priced higher.
Software producers in other markets, who implement much simpler algorithms and data management solutions, have much better margins that EDA companies that are at the forefront of numerical analysis, compiler, and data handling techniques, simply because the latter find it necessary to discount their products. I think that after so many years, the only thing standing between the large EDA companies and better margins is FEAR. Fear to loose an account, fear that the large companies will grow their own internal CAD departments. Neither of these events would materialize. There might be some early and temporary shifting of accounts, but the improved bottom line would reward the EDA companies in the financial markets. Let's compete in technology not in price! The problems solved are so complex and require such organizational structure, that electronics companies would take years to develop equivalent solutions to the design and manufacturing problems they face.
May be Synopsys, with these two acquisitions, has found a way to finally raise the prices in the Virtual Prototyping market.
A Big Fish in a Small Market
To the best of my knowledge once the CoWare acquisition is completed, Synopsys will have only to acquire Imperas in order to be the only vendor in the field. It is true that there are at least a couple of companies, Tensilica and Carbon come to mind, that offer products that can be seen as similar; but Tensilica produces other things beside software models of IP cores, and Carbon has a different approach to the market. Do the acquisitions make sense for Synopsys? Absolutely. Synopsys has been a pioneer in the IP development, having started as far back as 1990 developing DesignWare. Now it has a complete inventory of cores, from simple to very complex, and having the internal capability to develop and market development environments to support the IP products makes a lot of sense.
What will happen to Imperas? How long can they survive to compete with Synopsys? And an even more important question: how can they achieve a financial level of maturity to be a viable prospect for IPO in Great Britain where they are based?
Imperas counts on its Open Virtual Platform (OVP) initiative and their open source strategy to be viable. I am not sure that Synopsys needs to join or even acknowledge OVP. Neither VaST nor CoWare were part of OVP.
What can come next? How about MIPS? Synopsys already purchased the analog IP development group, based in Portugal, from MIPS. That could use up a nice portion of that cash that is still laying in the Synopsys vaults!
Comments
Virtual Platform Vendors
There is one more large independent virtual platform company in the industry capable of delivering a comprehensive Virtual Platform and TLM solution. For some years now, ASTC has been quietly developing and delivering advanced Virtual Platform and TLM products, solutions and services, in automotive, wireless, and multimedia. It has for some time been bigger than VaST and as large in engineering and R&D capabilities as CoWare. ASTC has an innovative comprehensive technology solution framework, called OSCAR, for TLM model and Virtual Platform creation, test and run time operation, with SystemC/TLM as the standard exchange format, including integration with HW verification environments and flows. ASTC also has developed and deployed a broad library of TLM models of very fast instruction set simulators, on- and off-chip IP models, pre-integrated SoC and MCU platforms, ECU and handset platforms, and test objects and tools. ASTC has continuously avoided any form of publicity and promotion, instead focusing on R&D and on working with its key customers.
ASTC stands for Australian Semiconductor Technology Company, with offices in Australia, US, and Japan.
Re: FEAR in EDA/ESL
It might be interesting to compare IBM and Synopsys (or Cadence). IBM has survived through many paradigm shifts (since 1885) and moves into new business areas as necessary and abandons old ones. Likewise Intel abandoned DRAM. Synopsys seems to still be tied to RTL design/synthesis as it was at its inception (1986). I.e. the focus and business model of IBM has shifted considerably over time, Synopsys doesn't seem to have changed very much apart from getting bigger - not unlike the dinosaurs.
Personally I think the acquisition of CoWare is good news for other players in ESL, since it is somewhat unlikely that Synopsys can rationalize its disparate array of tools into something coherent that works for ESL->Silicon.
Safety In Numbers
While I think that the EDA industry is broken because it grew up using the enterprise software business model (yielding a Big Three, maybe Four, but lacking the number of customers as SAP or CA), I wonder if SNPS sees its size as enabling sales of CoWare, VaST, Virtio (Innovator) at customers reluctant to deal with a SME.
Where's the money?
You are right, Synopsys has cornered the market, but is there a sustainable business here? A virtual prototype is comprised of a fast ISS (free from QEMU or OVP) surrounded by a SystemC representation of the hardware (the simulator is free from OSCI). The hardware and software are debugged using GDB and Eclipse (all free) running on Linux (free). Do designers really need an EDA vendor to hold their hand while they assemble these free parts? What does Synopsys bring to the party that the designer really needs? Models? In a virtual prototype, the models need to be so abstract to deliver sufficient performance that they must be simple (read low-value, easy to roll your own) - the notable exception being the CPU, but that's already free.
And if it is SystemC IP that Synopsys is bringing to the party, won't it just get thrown in with the "bundle"? Will Synopsys really get more or bigger bundles because they can throw in a SystemC IP library? I'm dubious.
Is OVP truly open source?
While the OVP models are 'open source', using them requires their simulator which is no longer free (except for evaluation or education). And you need the Imperas simulator if you use the OVP API.
At least that's my takeaway from reading the OVP World site.
There is one more company
You seem to have forgotten about one major virtual platform product, admittedly not in the core EDA space: Virtutech Simics. Also, you should not discount the usefulness of Qemu to build virtual platforms for software development.
If your dealing with this
If your dealing with this questions you might do something to solve this problems. You see there are lots of competitions especially large corporations that were able to maintain their standards. If you want to get into the position you must strive really hard to work out on this.
The Virtual platform scenario ..
The semiconductor industry is yet to see an ESL tool which provides the real value to the customer and which allows to create the ESL infrastructure re-usable in long term. Synopsys, with acquisition of VaST, Coware, and with its internal ESL offerings (Innovator), will hopefully combine the best pieces from all these and will come up with some real valuable ESL offerings. We are eagerly waiting to see that. Earlier the strategy of Synopsys was to provide the SystemC / TLM models of its own IP blocks, and not the models of third party IP. That shows their commitment to the standards, ie.. their customer can source the third party IP models from other sources, and if these models are created by using standards (SystemC/TLM2.0), then these should work fine with Synopsys tools. We hope that after acquisition of Coware, their strategy remains same and does not dilute their commitment to standards. Coware sells lots of third party IP models that work with their tools (not sure if these can work with other ESL tools also).
However, that leaves only one serious ESL player in the industry which is not good. The announcement of OSCAR from ASTC has come at the perfectly right time. With their strong background in SoC modeling I am sure this company will come out with something very meaningful.
In my opinion, Imperas has to play a slightly different role rather than projecting itself as virtual platform company. Its core strength seems to be the next generation embedded software tools that works with virtual platforms. These are very much required to realize the true potential of using a virtual platform for embedded software development instead of using FPGA board. The traditional embedded development tols are designed to work with FPGA board, and also does not work well with the muti core SoCs. OVP is opensource, and can work well with the SystemC / TLM2.0 based models (atleast that’s what they claim). Imperas should ensure that its tools should work equally well with any virtual platform created by using any ESL tool (as long as virtual platform conforms to the standards: SystemC, TLM2.0, CCI..)
Virtual platforms were being created even long back by using the proprietary technologies. OSCI has played an important role in defining the standards for virtual platform creation. That allows the integration of IP models from different sources. If any player tries to promote a virtual platform methodology which moves the industry away from standards, that should not be welcomed in the industry. Any ESL tool should work equally well with the standards (TLM2.0, CCI) based models from any source, and not just with the models created by using their technologies. OSCI TLM2.0 has played an important role to ensure that IP models from different sources can be mixed in the virtual platform. The OSCI CCI working group will take the interoperability to the next level. If all the players show their commitment to these standards, we can be sure that time for ESL has arrived NOW.
GreenSocs is an standards based open source infrastructure which allows to create the tool independent models. It is playing an important role in the System level design domain. The greenbus technology was adopted by OSCI, and was the base for definition of OSCI TLM2.0. The OSCI CCI working group has also adopted the technology from GreenSocs. The only industry ready TLM Kit (OCP-IP TLM Kit) is based on the GreenSocs infrastructure. GreenSocs have defined a methodology for using the QEMU processor emulator in the SystemC / TLM2.0 based virtual platform. I think after filling some gaps, GreenSocs infrastructure has the potential to become Linux of the hardware world.
We at CircuitSutra, provides the SystemC modeling services to create virtual platforms of the SoC at various abstraction levels, and also provides the embedded software development services using the virtual platforms. We are highly committed to adoption of standards (SystemC, OSCI TLM2.0, OSCI CCI, OCP-IP TLM Kit, STARC TL Guidelines.). We have been able to succesfuly create the virtual platform of complete SoC for several customers by using the QEMU processor emulator and other GreenSocs infrastructure. Our methodology allows the customer to quick start their ESL activity.
A paper was presented at IPESC 09 conference, Grenoble about this methodology:
http://www.ocpip.org/uploads/dynamic_areas/Ct9Rr6XmkN84Y6MvTouu/947/IPES...
One of the use case of virtual platforms is for embedded software development. These can also be used for : Architectural exploration, performance optimization, RTL verification etc.. The platform has to be implemented at right abstraction level for it to be suitable for certain use case. One of the missing component is that there is not clear definition of abstraction levels accepted universally.
The most clear definition of the abstraction level we have seen is the one provided by OCP-IP. It defines TL4, TL3, TL2 & TL1 abstraction levels. The STARC TL Guidelines v2 from STARC, Japan discusses in detail about the various terminologies popular in the industry, their comparison, and their mapping to the use case. CircuitSutra has recently presented on this topic at ESCUG:
http://www.ocpip.org/uploads/dynamic_areas/Ct9Rr6XmkN84Y6MvTouu/947/ESCU...
We work with our customers to create their ESL infrastructure which is based on standards and can be re-used in long term. We are also in the process of partnering with other vendors in the ecosystem so as to provide the maximum value to our customers.
Regards,
Umesh Sisodia
CircuitSutra Technologies Pvt Ltd
www.circuitsutra.com