In the last couple of years, if you had been to any CAD related conference or had been reading related news portals, it’d have been hard to miss the euphoria around this whole ESL thing. There are several ESL tools in the market now, including Bluespec, Catapult etc. The technology has advanced considerably to the point of being deployed on a growing number of designs.
While most of these ESL companies push their tools tirelessly, I am finding a lot of inertia from the designers’ front in adopting these tools and their associated methodologies.
Firstly, most of the design community is comfortable with Verilog and the foremost question on their minds is why adopt a new methodology at all. Verilog works fine for design and with the advent of SystemVerilog, hopefully most of its limitations should be addressed. When that happens for designs is anyone’s guess and a different topic altogether.
One of the biggest USPs for ESL tools is fewer lines of code at a higher level of abstraction of the design intent. Seemingly, this means reduced verification effort. However, as far as I know, most ESL tools focus only on design and don’t provide the verification infrastructure support that is very critical. Without a solid platform for verifying the “high-level design”, you end up relying on the generated Verilog or VHDL to accomplish verification. This by itself is not essentially bad because the chances of bugs in the high-level design is still less owing to the fewer lines of code.
But, how about coverage and other metrics that are conventionally used to qualify verification? Are these metrics going to be defined with respect to the original hand-written high-level design or the generated Verilog design?
Nevertheless, ESL is probably the future for designs. I reckon new design engineers will easily embrace it and will lay the foundation of its success rather than the experienced designers. Based on my interaction with several expert designers, it is going to be hard for them to adopt anything that is new unless they are the type that is looking for resume material.
Otherwise, they are going to keep asking the rhetoric “Why fix something that ain’t broken?”. Well, only time will tell if the current experienced designers will become ancient overnight and the new fresh designers will take over in a catastrophic change.
Here is a list of companies that work on VLSI or are primarily semiconductor companies that have a presence in Hyderabad
– Tundra Semiconductors (www.tundra.com)
Tundra established its design center in Hyderabad through the acquisition of Chip Engines Ltd., a former subsidiary of Alliance Semiconductors. Tundra works on system interconnect based bridges and other products revolving around PCI, PCI-X, PCI Express, Hyper Transport and Rapid IO. I think they also have an IP portfolio around these products. Some of the bridges include Hyper Transport to PCI and PCI-X bridges.
– AMD (www.amd.com)
Well, AMD, ATI, ATI-AMD whatever…… Its now AMD and whether the acquisition made sense or not – time will tell. The former ATI Hyderabad was established by acquiring CuTe Technologies. CuTe was predominantly working on compression/decompression of audio and video data for game developers for mobile phones and PDAs.
– Conexant (www.conexant.co.in/hyderabad)
Conexant has operations in Noida,Pune, Hyderabad and Bangalore but has made Hyderabad its base for expansion. They are totally about 775 strong and plan to expand to 2000 in the next couple of years.
– Mindspeed Technologies (www.mindspeed.com)
Mindspeed Technologies, Inc. is a leading supplier of semiconductor solutions for network infrastructure applications according to their website. The press release at the time of their opening the Hyderabad center, says the center will provide development and customer support for a broad range of Mindspeed communication products including the company’s flagship Comcerto([TM]) voice-over-Internet Protocol (VoIP) processors. They are starting with a small team of 15.
– Cypress Semiconductors (www.cypress.com)
The facility in Hyderabad is Cypress’s second in India. Cypress is one of the first after TI to enter India in the early 90s and have been in Bangalore since. I think their Asia Pacific head, Nagendra Cherukupalli, operates out of Hyderabad. I know a team there works on functional verification.
– Xilinx (www.xilinx.com)
Xilinx partnered with CMC before going on their own in Hyderabad. The CMC team produced more than 35 core IPs for Xilinx in less than three years. The center is about 60 strong.
– Renesas Technology (www.renesas.com)
Renesas Technology set up its first ODC at KPIT Cummins Infosystems to strengthen its design and development capabilities. KPIT Cummins previously executed design projects for Renesas in several areas, including SoC products for use in digital consumer electronics, analog circuit designs and embedded software. The Renesas ODC at KPIT Cummins will employ about 100 engineers.
– Mentor Graphics (www.mentor.com)
Mentor Graphics India was set up in 1997 and is now the second largest operation for Mentor outside the US. I have very little information on this and I don’t know if it is just a sales presence in Hyderabad.
– Synopsys (www.synopsys.com)
Synopsys has been in Hyderabad for a while now and I believe its mostly sales and AE oriented. I don’t think they do cutting edge EDA work.
– NANO-TECH Silicon India Pvt Ltd (NSTI)
NSTI plans to set up a mega semiconductor fab facility. I think it is associated with SemIndia
And there is SemIndia (www.semindia.in)
Companies under the SemIndia umbrella
– SemIndia Systems Private Limited – Focused on end user products (ADSL modems; GPONs, Wi-Max; Set-top boxes etc)
– SemIndia Fab Ltd
SemIndia Wafer Fab
SemIndia Assembly and Test Operations
– SemIndia Fab City Ltd
Focused on creating the entire Semiconductor Ecosystem in India
Here is a list of companies that work on VLSI or are primarily semiconductor companies that have a presence in Pune.
LSI Logic (www.lsi.com)
LSI’s presence in Pune came through the acquisition of Metta technologies, a multimedia SOC and software company. They do a lot of work on ASIC design and verification
They are a design and verification services company, with focus on the latter. They established the Pune center to set up an ODC for Qlogic around 2004. Since then, I hear that they have expanded.
I couldn’t get too much information but Conexant came to Pune through the acquisition of CG-CoreEL systems.
They work on IP cores and exercisers/analyzers for standard protocols like PCI Express, USB 2.0, IEEE 1394 etc.
Started its Pune center in 2004 to focus on VLSI design including Analog and Mixed-Signal designs.
Tensilica’s Pune design center was set up in 2006 though I know that they had few consultants working out of Pune much before that. According to their news release, the Pune team is working on a number of projects, including next-generation integrated processors, software and system design automation tools, and advanced multimedia architecture/applications for cell phones, portable media and digital TV. It also works on the company’s TIE (Tensilica Instruction Extension) compiler for automated hardware and software generation from high-level processor description.
I know they were involved in the initial phase in designing a complex SoC for one of the top two semiconductor companies in the world 🙂 I am sure they are expanding their VLSI presence in a large way
nVidia came to Pune through the acquisition of Pace Soft Silicon, which was a multimedia software application company. I remember hearing about Pace in my earlier life that they had some cool software products that work on TI and Intel powered mobile processors. nVidia now only has a software division.
A friend of mine, a verification expert, pointed out to me about a new verification tool that is apparently making some ripples in the design community. The tool, Certitude, launched by Certess uses a novel mutation based identification of testbench holes. Although he hasn’t played around with it yet, my friend was so impressed with the technology that he thought I should write a post on it.
Here’s how it works at a high level: The tool injects errors into a design and checks if the existing test bench and test suite detects the errors. If the errors go undetected, it exposes a hole in the verification. Certess has patents pending for the technology and their claim to fame so far has been roping in ST Microelectronics as a customer.
While the jury is still out, here is my take based on my knowledge of the tool. I believe that Certitude is an attempt to augment code and functional coverage tools. I don’t think the tool can replace these well established methods to qualify verification progress.
For example, I guess the tool will be able to inject errors into the design functionality only by accident because of its lack of knowledge about the design’s intent. As far as I can imagine, the tool will probably end up flipping some bits here and there, change some gates but nothing more than that. There is no doubt that these changes may result in considerable changes in the design that may manifest into a serious bug.
However, the tool can only identify broad areas of the design that have not been verified properly. I reckon these areas will be anyway identified by code and functional coverage holes.
Moreover, corner cases in the design are going to be hard to identify in a deterministic fashion. I am also not sure how the tool is going to provide data on the entire design. If it goes through each and every line of code in the design in order to inject errors, the result may be no different than code coverage result; only difference is that it will take an enormous amount of time to generate it.
It will be very interesting to see Certitude results when deployed on a design that is already mature – 100% code coverage and 100% functional coverage (assuming coverage points are defined thoroughly).
At first glance, Certitude looks like a promising tool to verify your verification environment but how methodical and exhaustive it can get, will determine its success. I believe that identifying holes here and there is not going to be compelling.
I am trying to put together a list of companies working on semiconductor space in and around Indian cities. I thought I’ll start with Chennai and also attempted to categorize them based on whether they offer products or services.
Their website claims to provide the industry’s first message stream processing hardware solution. I know of their founders and I believe its mostly the same team that ran the now-dead RealChip. They seem to be well funded by some of the industry’s best known VC fund houses.
Texas Instruments (www.ti.com)
TI acquired most of the engineers with DSP expertise from Nulife semiconductors in Chennai when Nulife was folding down. I know a lot of companies were trying to recruit these talented engineers and finally TI hired a whole bunch of them.
I think TI wanted to strategically locate itself close to the upcoming manufacturing facilities of Nokia and Motorola in Chennai. TI, Chennai is small and how commited they are to this facility remains to be seen.
Scolis Technologies (http://www.scolistech.com/index.html)
Scolis provides Contactless Smartcards and is founded by a couple of veterans from SCM Microsystems, Chennai. I believe one of these founders (Balu) came to SCM through the acquisition of Shuttle Technologies in Pondicherry and the other from WIPRO. I have heard very high opinions about the founders, especially Balu who is also the engineering manager. They just started a few months back, so its a real start-up.
Atheros Communications (www.atheros.com)
Atheros is into wireless communication products. It is a very good company with a strong leadership.
Cypress Semiconductors (www.cypress.com)
I have very little information about Cypress in Chennai but I know they are trying to set up an office, if not already.
Most of the VLSI team in L&T Infotech came through the acquisition of GDA Technologies, a semiconductor services and IP provider based out of San Jose, CA. Their IP portfolio comprises of standard interconnects like Hypertransport and other interconnect technologies. They are also reseller parters with Rambus.
Classic services company. You most likely won’t know what you will be working on until you start working.
There may be other companies working on VLSI in Chennai that I am not aware of. If you happen to know, please feel free to post on these pages.
I will gather information on Hyderabad, Noida in the coming days. I will update on Bangalore later as most companies are located there anyway.
During my conversation with a veteran technologist entrepreneur recently, I was talking about the lack of ground-breaking innovation happening in India, citing the lack of Apples and SONYs in India. He had a different interesting perspective on the topic.
His argument was that Indian companies are still technically and technologically competitive but lack the marketing prowess or the inclination. Perhaps we get overwhelmed by the US marketing giants.
The theory is that an Infosys can indeed make similar innovative products but they refrain from it assuming the market is still the US or Japan or Europe. After all, Steve Jobs is more of a marketer than a technologist.
Additionally our country with its population offers a huge market by itself that we do not need to depend elsewhere.
Generally, Indians have a tendency to play down their achievements as opposed to say, Americans. In fact, most of us were taught to be modest and humble right from childhood and we took it to our heart. I think it reflects itself in our marketing capabilities and the ability to think big. Of course, there are exceptions like Reliance and Tatas.
I hope we will see the next wave of Apples and Googles from India. We do have the manpower and a local market to thrive on. I think its the mindset that needs to change.
So, where is the Google going to come from, in India?
Continued from the previous post (Should Bangalore be Bangalore-d – Part I)
So, what are the alternatives to Bangalore?
Off the top of my head, there is a bunch of VLSI companies like Cadence, Freescale, Mentor and ST who have been in the NCR (Delhi, Noida, Gurgaon etc.) for a while now.
I was talking to some people from these companies and they proudly announced that they don’t face such levels of attrition as in Bangalore, in spite of higher salaries. Infrastructure development is rapid in the region too.
Hyderabad and Chennai are places where you find lot of engineers graduating every year in addition to a reasonable infrastructure.
Weather is a problem but Bangalore is no longer the pensioners’ paradise that it was, several years ago.
Chennai is traditionally conservative and job-hopping is not common. Pune offers a good alternative too.
Moreover, lot of VLSI engineers are originally from Hyderabad, Chennai and Maharashtra and hence wouldn’t mind being employed in these cities. You can also attract a lot of returning NRIs because most of them originated from Hyderabad and Chennai!
There are a few VLSI companies in all these cities but it is still not as big an industry as in Bangalore. I hope this will change because Bangalore in its current state just cannot accomodate more people.
If you are a software engineer, you will find tons of options anywhere in the country. Not so for a VLSI engineer currently.
As an employer, you are better off starting your operations elsewhere. TI for example, has set up a shop in Chennai through its acquisition of Nulife engineers. Mentor, Conexant, Cypress, ATI/AMD and a few others already have operations in Hyderabad. Now, VLSI-based companies have started in towns like Manipal and Hubli too though I personally think that’s carrying a bit too far!
People are slowly but surely realizing that Bangalore is not living up to its brand name and the hype.
So, if you are considering relocating to Bangalore, re-consider!