Success Stories

With 10x or higher speedups on key routines, typical cases showed a full application speedup of up to 2.0x on Titan supercomputer.

Using OpenACC, ADSCFD GPU-accelerated their density-based, explicit time-marching code for multistage steady and unsteady turbomachinery analysis.

Strong and weak scaling on up to 16,384 GPUs compared with 262,144 CPU cores achieved a 2.5x speedup over a highly tuned CPU-only version.

A quantum chemist without any formal education in computer science achieves 12x speed-up modifying less than 100 lines of code in one week.

Using OpenACC, the team ported the implicit solver on the GPGPU. OpenACC enabled the translation of mathematical formulations with minimal coding.

OpenACC made it efficient to develop for GPU-based hardware while retaining a single source for almost all the COSMO physics code.

An OpenACC GPU version of CloverLeaf with no code changes runs on a CPU with performance equal to an optimized OpenMP version.

Researchers can now run more scientifically interesting simulations by modeling larger nuclear reaction networks in 3D using OpenACC.