R&D partners of the GONT project

Interview with one of the key members of the GONT team, the head of the development team – Yury Fonin. Yury talks about the PPDL project, its development and plans for the future.

We are glad to welcome you to the pages of our blog, friends! Today we share interview with one of the key members of the GONT team, the head of the development team – Yury Fonin. Yury talks about the PPDL project, its development and plans for the future.

About the development.
How long has the project been developed, its history, milestones?

We started the PPDL project in 2002, when we did an internship as a post-graduate student at Infineon in Munich. We are Alexander Kravtsov, today the head of the systems department on the NIIME crystal and the Mikron plant and Yury Fonin. We then participated in the project ADS (architecture development systems) – the development of methods and technologies to accelerate the process of designing SoC-systems. On this occasion on Infineon came representatives of various universities and firms specializing in CAD to demonstrate their own developments, including high-level languages ​​describing the architecture. Among others, we remember the company LISA tech (now it has become part of Synopsys) with its language LISA. It was a complete and fairly high-quality product, but they asked for so much money for the license that our management rushed to look for alternatives. Including among startups and university projects.

Alexander and I decided to try to create a similar system. Firstly, it was interesting that such a writing of a similar language, that even one license costs fantastic money. Infineon – it’s not even Micron if they can not afford something, then this “something” really is worth a lot. Secondly, there was a desire to get a permanent position ahead of schedule with a good salary, a German version of the green card with the prospect of a residence permit, and other advantages available to people with a serious status in Germany. And we decided that PPDL is our chance to get all of the above, saving the company hundreds of thousands of euros.

We worked at night in parallel with the main work. And of course, by the time we had something to show, our management had found the means to buy LISA and nobody wanted to get involved in the project to develop another language.
In 2005, I (and a little later Alexander) returned to Russia. Shortly before that, I came to Russia on a vacation, where good friends told me about “friends of friends” who, thanks to expensive oil, earned millions and are looking to invest them. And since with the investment in Russia was not so simple, then from desperation they buy lots of flats. It was a “second chance !!!” to rise up by “due to our own purposefulness and outstanding business qualities.” Still it’s good to be young, naive and mobile. Investing in us for some reason no one then took, and the “friends of friends” were slightly inaccessible. And indeed, it seemed that the words “start-up”, “business angel” and “venture investment” mislead many people.

But all the same the movement was ahead. We managed to agree with the SEC “Module” that I will use PPDL to develop a simulator and assembler for them. And soon there was the first “combat” model of the processor, implemented on PPDL. I must say that we are still cooperating with STC “Module”. There works a wonderful team of developers who have something to learn and with whom it is simply pleasant to communicate.

In 2007, our application won the competition “START” and on this occasion we organized OOO “REDA-Soft” (Russian EDA).
To date, we have with Alexander more than ten years of experience in the design of systems on a chip. And Alexander specializes in “hardware”, and I – on the system software (operating systems, drivers, application optimization). Alexander managed as a project manager to go through the whole chain from the idea and requirements to the launch in production. This is a unique experience, I’m not sure that large companies have such specialists. And it’s not about Alexander’s talents, but about the fact that large stable companies have a clear division of labor: some model, others project, others produce. At us all are cooked in one boiler, therefore there is an opportunity to receive unique experience and to see a picture as a whole – as the microcircuit is designed, as it is tested, how it is made.

Are there any patents, patent applications?

No. We maybe found what to patent, but frankly frightens us the amount of body movements that need to be done to obtain a patent. And we do not yet have the means to pay for the services of the relevant specialists.

The main advantages of the package.

PPDL (processor and peripheral description language) is a language for describing the architectures of processor cores and special computing modules. The principle of PPDL can be formulated as “one description – several implementations”. From one holistic architecture description, the PPDL compiler can generate a Verilog description of the processor, from which it is possible to perform physical synthesis of the chip, as well as related software, such as:

1. Assembler
2. The Simulator
3. The interface of working with the debugger

In addition, in the near future we plan to add a test generator and a back-end compiler to the PPDL.

What does this ultimately give? Ultimately, this makes it possible to reduce the cost of designing the core of the microprocessor 3-5 times. In fact, what was previously created by a team of ten experts, now under the power of two or three developers, and in less time. If we take into account the reduction in communication costs (all the same, it is much easier for the two of us to agree on something than by a team of 10 people), in the end, we have an acceleration and a reduction in resources at times.

A separate topic is the adaptation of a ready-made kernel to application requirements. That is, the addition of several instructions to ensure that these or other algorithms work at the required speed. Such work using PPDL can generally be done by one specialist in a few days, and sometimes hours. Thus, “investing” once in the creation of one core, you can further adapt it to certain tasks with literally “cheap” labor.

About clients.
Who are the customers today?

So far we have only one permanent partner of SEC “Module”, for which we have developed models of NeroMatrix cores. We are negotiating with other design centers, but I’m not ready to talk about this in more detail.

Which customers would be interesting for you? What tasks are you ready to solve?

We are interested in all those who develop or plan to develop microcontroller cores. Especially it would be interesting to work with startups, creating something original in the field of microelectronics. I think such cooperation could help a lot in the development of the PPDL language, introduce into the language the designs that will allow the rapid development of next-generation PPDL solutions.

About the strategy.

While we are considering two main options: the sale of licenses and custom development of kernels on PPDL. In the second case, as part of custom development, we are also ready to transfer a license for PPDL, so that customers can adapt their architecture as quickly as possible to the requirements of certain applications. For example, we are ready to do for the customer all the routine work to create the kernel, leaving him to work on the implementation of key instructions, thanks to which its core will be fundamentally better than its counterparts. Since the extension of the instruction set to PPDL is extremely simple, the implementation of such “advanced” instructions can really take several days.

In addition, we are optimistic about “in the clouds”. I think that cloud services in the field of designing microelectronics will be very in demand and, perhaps, in general will be able to change the picture of the world electronic industry. Now, in order to start fully projecting chips, you need to invest from several tens of thousands to several million dollars. Imagine instead that there will be a certain “subscription fee” for which developers will have access to modern CAD systems, as well as to colossal computational resources for modeling. Moreover, access to this “happiness” will be from anywhere in the world where there is Internet.

Before thinking how to get out of business, it is necessary to enter thoroughly there. That is, my task is to develop the product and sell it as much as possible under different sauce. At the same time, I do not want to spend my energy trying to find potential buyers of the company. We will actively promote the product – buyers themselves will come running. On the other hand, the option of selling the company does not exclude. When there will be proposals – we will think.

About competitors.

Today our main competitor is Synopsys. Some time ago they bought CoWare, which, in turn, was the owner of LISA. As far as I know, several years ago Coware was a partner of Cadence, and Synopsys developed its high-level solutions.

There are, of course, solutions from other companies, but they are very far from the Russian market. In Russia there is PROSOFT, which does something similar, but so far I do not have accurate information about what their language and their product can do. Judging by rumors, they are more focused on creating a language for describing simulators for board-level development.

Who can be your partner? In which technological chains can you integrate, supplement them?

As already mentioned, the result of the work of the PPDL compiler is Verilog-description of the kernel + necessary system software. That is, development in PPDL is the first step in the technology chain from idea to crystal. Thus, our product is a superstructure over existing technological chains.

It is often asked how PPDL can be trusted, because there are probably errors in the compiler. And in case of development by “hands” and in the case of generation from PPDL, we have the result in the form of a Verilog-code, which needs to be tested. We do not replace synthesis – we simply make the development process faster and cheaper.

About the financing.

Currently we are looking for an investor, we are also working on a variant of participation in Skolkovo. In addition, we are trying together with other organizations to participate in various state. programs in the field of microelectronics. But we do not participate in leading roles, but as co-executors. So far it seems to me that it is impossible to “solo” in a more or less serious state tender for a small start-up company. And in partnership with a large firm with a solid reputation – quite feasible task.

Biographical reference

Fonin Yury (phonin@reda-soft.ru)
In 2001 he graduated from the Faculty of Automation and Computer Engineering of the Moscow Power Engineering Institute.
From December 1999 to May 2001 he worked as a programmer in the company “SPIRIT”.
From May 2001 to September 2005, he was interned as a graduate student in the research department of Infineon Technologies AG (Munich, Germany), where he studied the problems of designing complex systems on a chip and methods of automatic generation of system software. Since January 2006 he has collaborated with SEC “Module”, was developing software simulators for Nero Matrix cores.
From April 2006 to April 2008 he worked as a leading programmer of JSC “NIIME and Dr. Micron”.
In March 2007, together with Alexander Kravtsov, he established the firm “REDA-Soft”.
From December 2010 to January 2012 he worked in the company NVIDIA in the position of “engineer in architecture.”
From the beginning of 2012 he cooperates with MIPT, participates in research projects in the field of supercomputer computing.

Kravtsov Alexander (kravtsov_a_s@mail.ru)
In 2002 he graduated from the Faculty of Automation and Computer Engineering of the Moscow Power Engineering Institute.
From February 1999 to December 2000 he worked as a programmer at the MEI computer center.
From August 2001 to January 2002, he was interned as an engineer-trainee in the research department of Infineon Technologies AG (Munich, Germany), where he worked on the end-to-end method of modeling digital systems.
From June 2002 to December 2005 he worked in the research department of Infineon Technologies AG (Munich, Germany), where he worked on the development of a processor for a new generation of cell phones.
Since February 2006 to the present time he has been working as the head of the system development department on the chip of JSC “NIIME and Mikron Plant”, where he acquired a unique experience in project management in the field of system design on a chip, from the formulation of requirements to the launch in batch production.
Since 2007, has been cooperating with the STC “Module”, participated in the testing of multi-core architectures.
In March 2007, together with Yuri Fonin, established the firm “REDA-Soft”.

Materials are taken from the interview “LISA, PPDL, REDA-Soft”.
http://chiptalk.ru/lisa-ppdl-reda-soft/

More information about the R&D project partners on our website http://gont.io/page3199086.html

Thank you for attention! Good luck!
GONT

Leave a Reply