What is the philosophy behind Shakti project, do you think it will be limited in scope because it’s an academic project?
As far as we are concerned at IIT Madras, all that we do are open source and we have made a conscious decision to release Shakti under BSD license which allows anyone to make anything including making profit from it and we will be happy. We are not looking at any revenue back from this. What we believe is that this will give our country the necessary technology.
We are doing it because we are an academic institute and we are in the juncture of a completely transforming digital world and at this point, we need to enable our country which is going to have a significant amount of electronic interference and manipulation; we need to enable our country with our designs which will make us feel comfortable. Comfort zone now comes with security.
Security is another point where you should forget completely about making money. There was a recent news article that an electronics-enabled vehicle could be hacked wherein when you press the break, it will accelerate and when you press the accelerator, it will break. With all these penetrations happening, we need to have a secure system. This whole project is trying to give open source environment where people can build boxes about which they know everything and there are absolutely no blind spots.
Are you looking at this from a security point of view alone like in a defense application or do you see Indian industry picking it up for civilian use? What applications is Shakti being put to?
We are looking at this as a processor for India, both strategic and for civilian and consumer electronics. We did a survey and found around 493 different Indian companies who can potentially use Shakti. We have met around 20 of them and we will try and meet rest of them and see how they can benefit.
The companies we have met are in the consumer electronics application and security applications space and we also have certain strategic partners who have now come to us to build an ecosystem. There are a couple of startups, one from our own lab, who want to start building an ecosystem based on Shakti. We are helping them at the IIT level. We are also looking at a RedHat model, where startups can offer customization of the Shakti platform for a fee.
We have a 180-nanometer fab in Chandigarh that gave us a lot of confidence. We developed a processor and it worked in the first instance; the first tape-out was successful. We are also working on an open source CAD tool that is essential for security reasons or while designing for high performance.
What’s your experience collaborating with other entities in this field and how much has Shakti gained from such collaborations?
We are playing a significant role in a worldwide open source initiative called the “Open Road Project”. It was initiated by DARPA. We are now taking our design to make an open source tool flow for our SEL Fab, so any startup in India who wants to make a chip need not invest anything. They need not get scared about money or investment. I can sit in a garage and start working on it and then make a chip, like the open source software got made. We are democratizing the hardware and the tools. This will be the biggest success model for us.
What is your motivation to start this project, how many man years did it take to achieve this level of accomplishment?
We conceived this project around 2012-13 inspired by Dr. Abdul Kalam and our stated goal was to give India a commercially viable processor ecosystem by 2020. It’s not just to make a processor, the goal was to make an ecosystem. If you don’t have an ecosystem, then it will be another project where we will deliver and probably show proof of concept and it will live and die in some academic paper.
We started off with Open Power but then it was not as open as we thought. Then RISC V started, we established a lot of contact with University of California, Berkeley. Rocket was the first chip made by Berkeley based on RISC V and then MIT came up with a chip.
In terms of architecture we are similar but what makes us different is that we need fault tolerance, security and intelligence. For fault tolerance we have tied up with Thallus. Thallus also came out with open source safety critical standards. We are now building up the world class safety critical standards which will be out in a year and a half. Thallus is funding that initiative under their open source funding. So, we’ll put out that work in open.
Security is not just putting crypto accelerators crypto is only 5%; it’s a very important thing, of course, as the entire route of trust depends upon crypto but in overall system building crypto is only playing a minor role although an important role. There are a lot of things that we need to do to handle in modern attacks like the SQL injection attacks and stuff like that. So, we are trying to build certain things in the spine of the microarchitecture. We have got a concept of tag destruction set architectures and we are giving lot more micro architectural aid to security. That will be our unique selling point. From an intelligence and security angle we don’t want to merely make it open, we want to do the best and make it open.
Coming to intelligence, there are two parts to intelligence, one is intelligence at the edge and the intelligence at the server. We are working on both. There is a center sponsored by Kris Gopalakrishnan, our alumnus, Center for Computational Brain Research where we are working on neuromorphic brain inspired computing and we have made a lot of progress on that. We have given a proposal to MEITY to fund us the next stage. The entire project has been funded by MEITY and I should thank them for their trust in us. We also thank Intel who came out and did a free 22 nanometer tape-out for us in their semiconductor laboratory in Chandigarh and Synopsis and Hindustan Computers Ltd. have helped us too
What’s next in the Shakti paradigm?
The next story for us is about supercomputing. We now have a very good grip over multicores. We have a framework where we can make N number of cores theoretically. We want to look at this as 32 to 64 core single-chip and we are also working with standards such as the Interconnect. This is going to be a single SOC yielding something like 2.5 gigahertz per core easily. We call it ParaShakti which means Parallel Shakti.
As of today, we have the understanding and the ability to roll out an FPGA level prototype of ParaShakti, it just costs a million dollars.
Finally, do you think your project can really be the precursor to an Indian processor ecosystem?
Of course, that has been our mission. I call it as India’s project. Some scientists at MEITY say it’s ‘your proposal’ and I say, ‘no this is our proposal’. I want to push that very clearly. From an academic point of view, we must do something constructive for the country. We know that there is a problem and we know God has given us the brains to solve that problem; we can’t keep quiet and that is the prime motivation for us. Beyond that as a country we need to solve this issue of security, indigenous development, democratizing hardware, democratizing software infrastructure and thus in the national security perspective, nothing in the country should stop because of lack of technology.