Scientific Computing has published an article called ‘Breaking through the programming wall‘, written by Steve Conway, research VP for high performance computing (HPC) at IDC. The piece provides an excellent summary of the multicore challenge, so if anyone asks you what the big deal is about multicore, you can send them the link as a starting point.
The piece says that although most HPC applications are inherently parallel, they have been written to run on a single processor with direct access to memory. As a result, IDC research shows that 56% do not scale beyond eight threads and only 6% can exploit more than 128 threads. This is at a time when the biggest supercomputers have hundreds of thousands of cores and IDC expects their successors to have millions of cores in a few years.
There was one assertion in the article that caught my eye particularly. Conway says that “a fair number” of HPC applications will need to be rewritten over the next five to ten years to take account of highly parallel supercomputers. He adds: “There aren’t enough people on Planet Earth with the right kind of brainpower and experience to tackle this, but there are some. We’ll need to motivate universities to produce more.”
I’ve often seen industry insiders note that universities have been slow to educate their students on parallel programming, and that this has slowed down progress in the industry, but I was surprised to see this problem expressed so strongly. Do we really lack the brainpower to crack parallel programming? Is it a matter of education or does parallel programming require a particular type of inherent intelligence? Or is parallel programming just something that will always be restricted to the elite programmers, in the way that fabulous riches and public performance are (mostly) restricted to the best musicians?
