[concurrency-interest] Matrix Multiply examples

[Mark Thornton]

I have created a simple example of matrix multiplication using the 
current fork join frame work.
The source is here: 
http://homepage.ntlworld.com/m.p.thornton/MatrixMultiply.zip (less than 
8KB).

Actually a set of 7 matrix multipliers with progressively better 
performance, starting with the obvious trivial version. It could no 
doubt be improved further. The simplest code achieves about 90 MFLops, 
the best serial code 1600 MFlops, and the concurrent code 6000 MFlops. 
All for 1200x1200 matrices on a quad core Q6600. This matrix size was 
chosen to ensure it didn't fit in the processor cache (2 x 4MB). The 
code does not require square matrices or round sizes, but has not been 
tested extensively.

Mark Thornton