Benchmark test of accelerated multi-slice simulation by GPGPU.

Abstract A fast multi-slice image simulation by parallelized computation using a graphics processing unit (GPU) has been developed.　The image simulation contains multiple sets of computing steps, such as Fourier transform and pixel-to-pixel operation. The efficiency of GPU varies depending on the type of calculation. In the effective case of utilizing GPU, the calculation speed is conducted hundreds of times faster than a central processing unit (CPU). The benchmark test of parallelized multi-slice was performed, and the results of contents, such as TEM imaging, STEM imaging and CBD calculation are reported. Some features of the simulation software are also introduced.