Using 16 CPU cores the speed up for the variogram computation algorithm (gamv) is 19 times faster than the traditional implementation.
One of the most popular algorithms for conditional simulation receives a boost of 15.5x on 16 CPU cores and a 3.9x on a 4 CPU cores.
Over 3M points and using 96 realizations, the sequential gaussian simulation algorithm scales up to 14x.
We start with the GSLIB software package, originally presented by Deutsch and Journel (1992), that has been used in the geostatistical community for more than 30 years!
Minimal code modifications were added increasing as much as possible the elapsed time of execution of the studied routines. Also, we studied the run-time behavior of the applications using a profiler tool.
If multi-core processing is available, the user can activate special instructions in the code to speed up the execution using all resources of the CPU.
According to the obtained results, the scalability obtained is close to linear, and in some cases superlinear (due to the code optimizations added to the application).
A packaged version with the current algorithms: gamv, sisim and sgsim
Current status: Testing the release candidate.
Current status: In development.
Stay tuned for more updates.