Scientific computing and efficient memory usage

[torquil]

Hi!

I wonder if it is possible to get some book recommendations on the subject of efficient data structures and memory management in scientific computing? I do a bit of parallel programming in C/C++ with MPI, and I'm now considering trying my hand at some GPU programming as well.

When working with large arrays there is the possibility of introducing memory bottlenecks, so I would like to learn a bit more about efficient data structures and memory management in the context of scientific computing.

Best regards
Torquil

 

[rcgldr]

The main issue is the fact that ram memory throughput is greater when it's sequentially accessed or nearly so. How the cpu handles cache is another factor to consider.

There's a brief mention of this in a wiki article on merge sorting, not much detail, but it does contain links to other articles, such as locality of reference.

http://en.wikipedia.org/wiki/Merge_sort#Optimizing_merge_sort