EarthJet.Net Performance

Preliminary tests reveal that EarthJet.Net for a 1000 particles has an increase in performance 4 times that of version one.

I ran version one with a 1000 particles on a test chute. On a regular AMD 64 3800+ the first particle to hit the deflector plate took 2:25 minutes.

latest version

Then I ran the same chute with the same particle size, tonnage, and a 1000 particles using the latest version, and it only took 37 seconds for the first particle to hit the deflector plate. It is quite surprising that such a performance boost is even possible. Yes, the new version was compiled using the latest Visual Studio compiler with fast floating point optimization setting, while version one was compiled with VS 6.0. But the biggest performance factor is the code optimization itself. The Discrete Element Method incorporates for particle localizations what is called, "cell memory", or at least something like that of a honeycombed memory packaging structure in which particles get located by the cell's address. The so-called cell memory grid structure provides a means for optimized particle calculations. EarthJet.Net's cell memory model has been totally reworked and simplified thus giving it the extra performance boost. I have not even attempted using OpenMP (API for Parallel Processing) yet. Will be interesting indeed.

By Frank Neubecker