BBN Technologies

High-throughput, real-time signal and information processing typically requires the use of multiple CPUs and array processors operating in parallel--and has typically involved high development and maintenance costs. The SPE software uses a Distributed Data Flow (DDF) architecture to allow creation of systems which are efficient, maintainable, and scalable.

Distributed Data Flow (DDF) Processing
DDF processing is data-driven and distributed across processors rather than pre-scheduled or centrally scheduled. This results in an efficient use of multiple, parallel processors, yielding high processing throughput without a cumbersome scheduling mechanism.

Reusable Modules
A large library of reusable signal processing modules, developed in C and C++, supports the SPE. A simple, high-level, scripting language that is integral to SPE defines and assigns processing modules to specific processors.

Easily Modified
As requirements change, the user can modify the system easily by adding hardware or rescripting, without any software recoding. Newly-devised algorithms can be incorporated as new modules in the library with minimal development effort and made available for reuse.

Scalable
SPE can run on one processor or a thousand processors to meet your requirements in a cost-effective manner using inexpensive Pentium or other available processors. The Intel Vector Processing Library is extremely efficient -- often eliminating the need for specialized vector processors required by systems built a year or two ago.

Re-hostable
A key feature of the SPE and its DDF core is the ease with which the software can be re-hosted on new COTS processors and operating systems.

Reduced Vendor Dependence
The adaptability of the SPE architecture makes reliance on vendor-specific, proprietary capabilities unnecessary. Users can move between standards-compliant COTS hardware and operating systems with a minimum of recoding.