Matrices - parallelise the build process

[GoogleCodeExporter]

Matrix building is currently a linear process on a single core.  It should be 
possible to parallelise it using MCE, with prediction of existing elements done 
using binary searches on a sorted array based on the current write-to-file 
prediction approach.

Might need to turn it off in the event we have asymmetrical spatial conditions 
where a does not contain b, but b does contain a.  In these cases the sorting 
approach will be invalid unless we check a's neighbours when processing b.  

Original issue reported on code.google.com by shawnlaffan on 15 Oct 2014 at 10:28

[GoogleCodeExporter]

Need to add an option to Spatial.pm sort the neighbour lists to support this 
approach.  Adding the sort to the neighbour search would potentially slow 
things down, as well as impeding any future approach which supports neighbour 
weights, as that would use a hash approach.  

Original comment by shawnlaffan on 16 Oct 2014 at 10:48

• Added labels: ****
• Removed labels: ****