Multi-threading

[GoogleCodeExporter]

I've noticed though quite a bit of playing around with both the compiled 
program and the source code that it is tends to be a single thread to crunch 
all the data. Though the performance is responsive when using only several 
items between four to seven regions, when doing a check between 200 and 500 
items between four regions becomes time consuming just looking at the reports, 
doing the route search takes even longer, averaging around 3 to 5 minutes 
(gathered over several tests) on an AMD Pheonom II 1055T overclocked at 3.6 GHz 
(from 2.8 GHz).

Taking full advantage of all cores and / or cpu's (if running on a multi socket 
system or virtual machine) would theoretically speed results up significantly 
over the current means.

Along with that, sometime in the future as well looking into possibly CUDA or 
Stream to do the calculations on a much higher threaded level, given of course 
the GPU or GPGPU is not being utilize heavily by other processes.

Original issue reported on code.google.com by gac64...@gmail.com on 1 Aug 2011 at 9:49

[GoogleCodeExporter]

As a side note, using 56,749 market exports using issue 13's solution to dump 
data into the market share folder. Taking this with data less then 22 hours old 
from the following regions:

The Forge
Metropolis
Domain
Sinq Laison
Fountain

This was compiled utilizing 6 clients between two locally networked computer 
and two other clients with data being fully synchronized via dropbox. Compiling 
a search using the following settings took 61 minutes, 43 seconds:

Isk: 1,000,000,000
Cargo Size: 981,000 m3

Memory usage at the end was allocated at 1,289,748 KB cross referencing between 
Task Manager and Resource Monitor.

OS: Windows 7 Ultimate x64
CPU: AMD Phenom II 1055T overclocked at 3.6 GHz
Memory: 8 GB DDR3-1600 (4 x 2 GB)

Original comment by gac64...@gmail.com on 1 Sep 2011 at 6:50