intriguing , especially their claims about compiler technology, how much of that will they develop proprietary vs open... could any of this be transferable to the epiphany architecture. would they collaborate (similar goal) or be hostile to competition.
I'd like to know more details about their software approach.
From a lot of thinking about this I have wondered how much of that '60%' its' really possible to eliminate; with this kind of design theres' still work done getting getting the right data in the right place. I wonder if a halfway house is possible with more cache control instructions and memory hints (e.g generating data then flagging it as read-only etc.).
C++11 helps a lot with lambdas but I still think this needs a new language to exploit in a natural way so there's a chicken/egg situation