Part of the  

Chip Design Magazine

  Network

About  |  Contact

Posts Tagged ‘SystemVerilog’

Portable Stimulus

Thursday, March 23rd, 2017

Gabe Moretti, Senior Editor

Portable Stimulus (PS) is not a new sex toy, and is not an Executable Specification either.  So what is it?  It is a method, or rather it will be once the work is finished to define inputs independently from the verification tool used.

As the complexity of a system increases, the cost of its functional verification increases at a more rapid pace.   Verification engineers must consider not only wanted scenarios but also erroneous one.  Increased complexity increases the number of unwanted scenarios.  To perform all the required tests, engineers use different tools, including logic simulation, accelerators and emulators, and FPGA prototyping tools as well.  To transport a test from one tool to another is a very time consuming job, which is also prone to errors.  The reason is simple.  Not only each different class of tools uses different syntax, in some cases it also uses different semantics.

The Accellera System Initiative, known commonly as simply Accellera is working on a solution.  It formed a Working Group to develop a way to define tests in a way that is independent of the tool used to perform the verification.  The group, made up of engineers and not of markting professionals, chose as its name what they are supposed to deliver, a Portable Stimulus since the verification tests are made up of stimuli to the device under test (DUT) and the stimuli will be portable among verification tools.

Adnan Hamid, CEO of Breker, gave me a demo at DVCon US this year.  Their product is trying to solve the same problem, but the standard being developed will only be similar, that is based on the same concept.  Both will be a descriptive language, Breker based on SystemC and PS based on SystemVerilog, but the approach the same.  The verification team develops a directed network where each node represents a test.  The Accellera work must, of course, be vendor independent, so their work is more complex.  The figure below may give you an idea of the complexity.

Once the working group is finished, and they expect to be finished no later than the end of 2017, each EDA vendor could then develop a generator that will translate the test described in PS language into the appropriate string of commands and stimuli required to actually perform the test with the tool in question.

The approach, of course, is such that the product of the Accellera work can then be easily submitted to the IEEE for standardization, since it will obey the IEEE requirements for standardization.

My question is: What about Formal Verification?  I believe that it would be possible to derive assertions from the PS language.  If this can be done it would be a wonderful result for the industry.  An IP vendor, for example, will then be able to provide only one definition of the test used to verify the IP, and the customer will be able to readily use it no matter which tool is appropriate at the time of acceptance and integration of the IP.

DVCon U.S. 2016 Is Around the Corner

Thursday, February 18th, 2016

Gabe Moretti, Senior Editor

Within the EDA industry, the Design & Verification Conference and Exhibition (DVCon) has created one of the most successful communities of the 21st century.  Started as a conference dealing with two design languages, Verilog and VHDL, DVCon has grown to cover all aspects of design and verification.  Beginning as a conference based in Silicon Valley, the conference is now held on three continents: America, Asia and Europe.  Both DVCon Europe and DVCon India have shown significant growth, and plans are well on their way to offer a DVCon in China as well.  As Yatin Trivedi, General Chair of this year’s DVCon U.S., says, “DVCon continues to be the premier conference for design and verification engineers of all experience levels. Compared to larger and more general conferences, DVCon affords attendees a concentrated menu of technical sessions – tutorials, papers, poster sessions and panels – focused on design and verification hot topics. In addition to participation in high quality technical sessions, DVCon attendees have the opportunity to take part in the many informal, but often intense, technical discussions that pop up around the conference venue among more than 800 design and verification engineers and engineering managers. This networking opportunity among peers is possibly the greatest benefit to DVCon attendees.”

Professionals attend DVCon to learn and to share, not just to show off their research achievements as a community.  The conference is focused on providing its attendees with the opportunity to learn by offering two days of tutorials as well as frequent networking opportunities.  The technical program offers engineers examples of how today’s problems have been solved under demanding development schedules and budgets.  Ambar Sarkar, Program Chair, offers this advice on the DVCon U.S. 2016 web site: “Find what your peers are working on and interact with the thought leaders in our industry. Learn where the trends are and become a thought leader yourself.”

Grown from the need to verify digital designs, verification technology now faces the need to verify heterogeneous systems that include analog, software, MEMS, and communication hardware and protocols.  Adapting to these new requirements is a task that the industry has not yet solved.

At the same time, methods and tools for mixed-signal or system-level design still need maturing.  The concept of system-level design is being revolutionized as architectures like those required for IoT applications demand heterogeneous systems.

Attendees to DVCon U.S. will find ample opportunity to consider, debate, and compare both requirements and solutions that impact near term projects.

Tutorials and Papers

As part of its mission to provide a learning venue for designers and verification engineers, DVCon U.S. offers two full days of tutorials.  The presentations of the 12 tutorial sessions are divided between Monday and Thursday, separate from the rest of the technical program so they do not conflict and force attendees to make difficult attendance choices.

Accellera has a unique approach to putting together its technical program.  I am slightly paraphrasing this year’s Program Chair, Ambar Sarkar, by stating that DVCon U.S. lets the industry set the agenda, not the conference asking for papers on selected topics.  He told me that the basic question is: “Can a practicing engineer get new ideas and try to use them in his or her upcoming project?” For this reason, the call for papers asks only for abstracts and those that do not meet the request are eliminated.  After a further selection, the authors of the chosen abstracts are asked to submit a full paper.  Those papers are then grouped according to their common subject areas into sessions.  The sessions that emerge automatically reflect the latest trends in the industry.

The paper presentations during Tuesday and Wednesday take the majority of the conference’s time and form the technical backbone of the event.

Of the 127 papers submitted, 36 were chosen to be presented in full.  There will be 13 sessions covering the following areas: UVM, Design and Modeling, Low Power, SystemVerilog, Fault Analysis, Emulation, Mixed-Signal, Resource Management, and Formal Techniques.  Each session offers from 3 to 4 individual papers.

Posters

Poster presentations are selected in the same manner as papers.  A poster presentation is less formal but has the advantage of giving the author the opportunity to interact with a small audience and thus the learning process can be bilateral.  There have been occasions in the past when an abstract submitted as a poster is switched to an oral presentation with the consent of the author.  Such operation is possible because the submitting and selecting process is similar and thus the poster has already been judged as presenting an approach that will be useful to the attendees.

Keynote

This year’s keynote will be delivered by Wally Rhines, the 2015 recipient of the Phil Kaufman Award.  Wally is well known in the EDA industry for both his insight and his track record as the Chairman and CEO of Mentor Graphics.  The title of his address is Design Verification Challenges: Past, Present, and Future.  Dr. Rhines will review the history of each major phase of verification evolution and then concentrate on the challenges of newly emerging problems. While functional verification still dominates the effort, new requirements for security and safety are becoming more important and will ultimately involve challenges that could be more difficult than those we have faced in the past.

Panels: One Good and One Suspect

There are two panels on the conference schedule.  One panel: “Emulation + Static Verification Will Replace Simulation”, scheduled for Wednesday March 2nd at 1:30 in the afternoon looks at the near future verification methods.  Both emulation and static verification use has been increasing significantly.  May be the verification paradigm of the future is to invest in high-end targeted static verification tools to get the design to a very high quality level, followed by very high-speed emulation or FPGA-prototyping for system-level functional verification. Where does that leave RTL simulation? Between a rock and a hard place! Gate-level simulation is already marginalized to doing basic sanity checks. May be RTL simulation will follow. Or will it?

The other panel scheduled for 8:30 in the morning of the same day concerns me a lot.  The title is “Redefining ESL” and the description of the panel is taken from a blog that Brian Bailey, the panel moderator, published on September 24 of 2015.  You can read the blog here: http://semiengineering.com/what-esl-is-really-about/.

In the blog Brian holds the point of view that ESL is not a design flow, it is a verification flow, and it will not take off until the industry recognizes that. Only now are we beginning to define what ESL verification means, but is it too little, too late?  There are a few problems with the panels committee accepting this panel.  To begin with ESL is an outdated concept.  Today’s systems include much more than digital design.  Modern SoCs, even small ones like those fund in IoT applications, include analog, firmware, and MEMS blocks.  All of these are outside of the ESL definition and fall within the System Level Design (SLD) market.

The statement made by Brian that ESL would not be made viable by the introduction of viable High Level Synthesis (HLS) tools is simply false.  ESL verification became a valuable tool only when designers began to use HLS products to automatically derive RTL models from ESL descriptions in SystemVerilog or C/C++ even if HLS covered mostly algorithms expressed in something else besides Verilog, VHDL, or SystemC.