http://www.stoneridgetechnology.com/services/high-performance-computing/
Application developers looking for performance on modern CPUs must understand how to code for multi-core processors; not just one of them but many connected together either on the same motherboard or over a network. Typical performance chips today come with four to eight cores and that number will continue to increase in the future. Some of the tools used for scalable parallel programming are MPI, for communication between computing nodes and OpenMP, Intel TBB or pthreads for communication between cores running under the same OS either on the same chip or different chips. In addition a fully optimized application must consider the use of SSE, the Streaming SIMD Extension instructions available on all modern x86 processors that enable low level data parallelism. This three level model of parallelism (between compute nodes, inside the compute node and in the data representation) can be daunting. Stone Ridge Technology has 15 years experience developing challenging parallel codes for large technical applications using these and other tools and achieving superior results. If you need help navigating through the options for parallel implementations on multi-core processors please call us. Our code evaluation services are free of charge. Please also check out the HPC Toolshed where you will find more information related to the development of high performance parallel applications. INDUSTRY APPLICATIONS:
CASE STUDIES: |
地質探勘 >