BILLERICA, Mass., 27 Jan. 2016. Abaco Systems in Billerica, Mass., is introducing an upgrade to the company's AXIS MPI high performance message passing software middleware to improve embedded computing performance and support GPUdirect.
AXIS MPI formerly was available only as part of Abaco's Advanced Multiprocessor Integrated Software (AXIS) software development environment, but now is available standalone. Abaco will support AXIS and AXIS MPI in the company's HPEC platforms and across the company's single-board computers.
AXIS and AXIS MPI are designed to optimize the development of sophisticated high performance, multiprocessor applications such as mission processors, radar, sonar, image processing, signals intelligence and electronic warfare.
AXIS MPI adheres to industry standard open architecture APIs, meaning that it benefits from broad compatibility and interoperability; is designed to use processor resources efficiently; is simple to use; provides flexibility of transports, including InfiniBand, RoCE, TCP, UDP, Posix Shared memory, KNEM shared memory, GPUDirect, and GPU IPC; supports heterogeneous architectures including multiple processor architectures (Power PC, Intel, ARM) and operating systems (Windows, Linux, VxWorks); can be optimized for minimal latency or minimal CPU usage (best SWaP performance), providing flexibility for a wide range of embedded applications; and benefits from the extensive customer support available through Abaco Systems' HPEC Center of Excellence.
Related: Software engineering tool for multicore HPEC embedded computing introduced by GE
AXIS MPI has a beneficial effect on power consumption and heat dissipation in certain applications, and to deliver substantially superior throughput compared with alternative message passing interfaces. Its support for direct data transfers between GPU memory over RDMA fabrics via GPUdirect technology provides a significant performance advantage.
For more information contact Abaco Systems online at www.abaco.com.
