13.2 QW-OMP Simulator

QW-OMP version uses OpenMP programming standard (http://www.openmp.org/) to accelerate the FDTD loop. The default number of created threads in the Visual C++ implementation of the OpenMP standard is equal to the number of virtual processors, including hyperthreading CPUs. The user can control this number by setting “OMP_NUM_THREADS” system environment variable or in Preferences dialogue. The OMP version speedup depends on number of processors and computer architecture. Moreover, it is also model dependent. For large 3D models we have observed 1.8 - 2 speedup on PC computers with 2 physical processors (not cores, not HT). The user can run ker1omp.exe without any changes in the model prepared for the sequential QW-Simulator.

In consequence multicore processors for memory intensive applications like QuickWave now behaves similar to earlier multiprocessor computers and speed of calculations for FDTD processing increased more or less with the factor of 3. Now QWED recommends use QW‑OMP version of QuickWave with Intel processors (e.g. I7) or newer architecture and the FDTD performance is up to 300 mln FDTD cells calculated per second for such computer systems.