|
[Sponsors] |
CFD News and Announcements - Message Display |
Post Response | Return to Index | Read Prev Msg | Read Next Msg |
Posted By: | Numeca International |
Date: | Wed, 23 Nov 2016, 4:06 p.m. |
Numeca’s OpenACC accelerated CFD application achieves more accurate results in half the time, enabling much faster design cycles for its customers.
Numeca cooperated with NVIDIA to meet growing customer demand for GPUs. The challenge was to accelerate our FINE/Turbo application on GPUs, without having to rewrite the core code, providing faster time-to-solution for our customers.
NVIDIA's OpenACC is a directives-based parallel programming model designed for simplifying programming of massively parallel processors. The model provides performance portability across a wide range of platforms, including host-GPU, multi-core, and many-core processors.
The OpenACC and PGI Accelerator compiler enabled our developers to restructure and accelerate computationally intensive routines, while maintaining portability and eliminating the need to rewrite code.
Result: 10x or higher speed-ups on key routines, typical cases showed a full-application global speedup of up to 2.0x on the Oak Ridge Titan supercomputer. Customers can achieve more accurate results with the accelerated code in half the time, enabling much faster design cycles.
Read the full review here: NVIDIA - Porting Industrial Computational Fluid Dynamics Applications
Post Response | Return to Index | Read Prev Msg | Read Next Msg |