HPC Hopscotch
Data locality is the key to efficient code.
Wednesday, June 17th, 2009
In my last column, I discussed a possible future for HPC programming. Briefly, I assumed that as cores per socket continue to grow, HPC programming models would diverge. I assume a small thread-based model would be used on multi-core systems (single memory domain) and a large model based on the current message passing (MPI) methodology used in clusters. It seems that I was not alone in thinking about this issue.
Joe Landman over at Scalability.org wrote about The coming bi(tri?)furcation in HPC, part 1 and John Leidel at InsideHPC.com even enshrined my name in his The Eadline Split: The Growing Chasm of Multicore Programming. I believe it is a discussion worth continuing. In the past, I have talked about the goodness that is MPI and at the same time I find it a bit low level for most programmers. Some may even call it…
Please log in to view this content.
Read More
- The State of Oracle/Sun Grid Engine
- A Cluster In Your Pocket
- Moving HPC Closer to The Desktop
- Cluster In The Clouds
- Summer Language Seminar: The R Project
|
