header-dsp Curtiss Wright
Contact Sales Contact Support Print Page

Continuum IPC

Real-time sensor based systems require status, control and bulk data movement to coexist within the same communication architecture. The Interprocessor Communication (Continuum IPC) Library provides all the capabilities needed to control applications running on multiple processors having data movement requirements. The software hides the hardware architectural details from the application designer. For example, message passing functions send messages using the same application interface and parameters, whether the destination task resides on the same processor, same board, or any other board within the fabric. The hardware abstraction provided by Continuum IPC ensures that applications programs written for current generation processors and fabrics will be easily portable to next generation processors and fabrics. Currently supported fabrics and interconnects include PCI, StarFabric, Serial RapidIO and PCI Express (PCIe).

The software is designed to dynamically determine routing paths at run time. This means that designers need not pre-configure routing or entity tables. The library determines the location and routing for source and destination as applications open endpoints for input (receiving messages) or output (sending messages). The library determines the location of source and destination endpoints at run time without using a central database. The absence of the central database eliminates single points of failure, since routing information is stored within each processor.

Features

  • High performance, low latency message passing library
  • POSIX style application interface
  • Message passing features priorities, time-outs, and copyless tranfers
  • Bulk data transfer from any processor’s memory to any other processor’s memory
  • Memoryless signaling from any processor to any other processor
  • Global buffer and global semaphore API
  • Segmented block transfer capability supports memory striding, matrix manipulation and submatrix selection
  • Support for multi-core processors
  • Supported under VxWorks on CHAMP multi-processor boards and SVME-183 and VPX-185 single board computers