ARM India, armed to face the future from India
Date: Thursday , January 31, 2008
Last year in August, the entire ARM community, along with its partners had gathered in a university campus in Cambridge for the annual ARM Partner Meeting (APM). ARM, whose comprehensive product offering includes 16 and 32-bit RISC microprocessors, data engines, 3D processors, digital libraries, embedded memories, peripherals, software, and development tools et al was to share its technology roadmap with partners. A part of the initiative was a competition for the ‘Most Innovative Idea’. Over 120 ARM engineers from across the world had submitted their entries and guess what? The idea from India, centered on using an ARM processor to make an intelligent helmet was chosen as the winner.
The winner received a generous cash award. “But,” says Anil Gupta, Managing Director of ARM’s India Operations, “the recognition from the CEO, CTO, and the rest of the top management and the entire partner community was much more satisfying.”
The India development center
The award at the APM was not a one off occurrence. Such has been the performance of ARM India development center which is now trusted with ownership of many parts of the product development cycle.
ARM is organized into 5 business groups: processor, silicon library, media processing, system design, and support. All of these have a presence at the India development center, each at a different stage of evolution.
The library division at five years is the oldest. Its basic work is developing silicon libraries for SoCs. The processor division, started in early 2005 in ARM India, does some very critical work in the development of processor cores. Gupta notes that depending on the way the processor core interacts with the other subsystems like memory, the silicon library architecture can have a significant impact on the performance. In order to help realize optimum performance, power, or area, tweaking might be required either on the memory or the processor side, or both. The two divisions work closely to help realize this.
Incidentally, the media processing group is engaged in enabling graphics, driver coding, and embedded software work. Operating systems support and critical software for ARM’s MALI® family of graphics processors and security architecture get implemented by this team. The system design group is engaged in development and validation of ARM RealView® tools and models that help customers to implement their designs with ARM processors. And the support team deals with technical support on processors and libraries to customers at the local as well as global levels.
The India center has been quite a success, notes Gupta, based on the fact that ARM India continues to expand. “We have grown quickly from a 40+ team in 2005, to over 300 now. Today, more than 20 percent of ARM’s R&D workforce is located in India”. With time, teams at the India center are gaining full ownership and accountability for projects. For example, the earlier ARM model had Sunnyvale as the center of library processing. Today, that responsibility is shared by six centers located worldwide, India being one of the primary centers. “ARM came to India due to the scarcity of necessary skills and resources in the U.S. and U.K. The biggest contribution of the India center has been to enable and support the growth of ARM business with adequate R&D,” he says.
If that be the case, one may ask, for instance, how many library projects does ARM India do in a year? And what percentage of ARM’s global library processing work does it constitute?
Gupta insists that the number of projects can never give a correct picture since projects range from six weeks to nine months in duration. While something like tweaking an existing library meant for designing a chip for industrial use to enable it to work in different environmental conditions might get done in six weeks, projects pertaining to the latest process technology can take nine months or more since the technology is still evolving.
In fact, engineers at ARM India, after having soiled their hands with 65nm and 45nm works are now focusing on 32nm.
Focus on innovation matters
That this has been achieved in spite of the hue and cry about talent crunch and fresh graduates being low on employability ratings is remarkable. The management at ARM says that the stewardship of experienced people in the team plays a critical role in grooming youngsters with less experience. “Besides,” says Gupta, “youngsters are encouraged to ask all sorts of questions, even ones that appear out right silly.” Questioning “status quo” is the first step to innovation.
For example, while working on the ARM 11 processor, the team was busy rewriting the ARM 9 tests for the new processor. There was nothing odd about this. It was a necessary tradition for every new generation of processors to ensure that it came out without bugs. However, this time around, one of the engineers quipped, “This is stupid! We must be able to take these tests forward without modification since the basic architecture in all processors remains the same.”
This question snowballed into a full-fledged discussion and the engineers created the abstraction layer — one which separates the upper level from the basic lower levels. This has made it easy to keep the upper level unchanged from generation to generation while enabling the minimal tweaks necessary at the lower levels. More importantly, it has taken away the need for laborious rewriting of test codes, which was the reason for a significant amount of lost time in the development of earlier processor generations.
Remaining focused on innovation, ARM India continues to hire the best young talent from the universities and grooming it. Gupta states that ARM is in the midst of establishing alliances with the universities. Under this program, he envisages a few ARM professionals engaging in part-time teaching and a few professors working in ARM part-time. Technology is moving in the direction of multi-core and it is essential that the programming paradigm shifts from its uni-processor orientation to parallel programming. “With the right stewardship and industry initiatives, universities in India could turn into powerhouses for parallel programming. Significant innovation in parallel programming could come from India,” says Gupta.
Going ahead, as PC growth stunts and mobile computing thrives, other companies in the industry are busy creating products to capture the mobile market. It’s both a matter of concern as well as an opportunity for ARM. Concern, since competition is technically and financially strong and opportunity, since competition should only spur ARM to do the best it can in what it does – creating the best IP for its partners in the semiconductor industry that build industry leading chips with its IP. Can the engineers at ARM India help make a difference? With its engineers prepared to ask apparently silly questions and the management awarding innovations with adequate recognition at the global level, that doesn’t seem so difficult.