SM: How much do people need to learn to be able to optimize programs on the Tilera chips? AA: That depends on the applications and the domain. The good news is you have gotten something working and running. Once you have done that, you can then try various optimizations. You don’t have to read 50
Here Anant discusses his final area of innovation in Tilera, which is the piece which really supports their go-to-market strategy and allows companies to become early adopters of multicore processors. SM: What is the final innovation? AA: The fifth and final innovation is in software. The third “P” is programmability. There, we have done some
SM: What came after Alewife and VMW? AA: I did VMW in 1994 – 1995, and in 1996 I came back to MIT. I started the Raw effort in 1996. Looking at processor design, we felt that in another 10 years we would have chips with billions of transistors and we wanted to discover how