Friday, 9 September 2011

Reply to Joseph Wigley

I tried to respond to a link and a piece  from Joseph on his blog (  http://www.josephwquigley.com/internal-reflections/2011/09/mobile-cores-vs-desktop-cores-a-visual-comparison/   ) but it kept failing so I am posting it here - hope that you get it here Joseph:

"I am interested by your perspective, too. There are still a lot of issues. It seems apparent at present that mobile (ie phone) devices aren't keeping pace with the sort of leapfrog that you talk about. The  mobile embedded stuff is already looking ahead of what you predict, at least a bit. The deep embedded stuff however is leading the way.

The real issue for the vendors is programmability and changes in architecture. If the number of cores increases, then so does contention for shared memory. The logical step is then to distribute the memory closer to the processors. Fine. That means that a lot of work has to go into models of and tools for programming - and "true" parallel programming is somewhat more difficult than regular concurrent programming.

Shared memory is a natural evolutionary path from uniprocessors, but it has a finite lifetime. The problem is that the majors don't know how to address it."