Coalescing the fragmented information associated with each event of a high-energy physics detector and then processing it at the rates demanded by accelerator systems are among the largest challenges in the physics community. New detector systems continue to increase the number of measured parameters and simultaneously increase the resolution of detector components so that the volume of generated information is exploding. Data acquisition and computation systems now face staggering amounts of data to both move and process. A peripheral component interconnection (PCI)/ scalable coherent interface (SCI) bridge (PSB) that solves both of these data acquisition and computation issues will be developed. This bridge links the high performance PCI, the emerging standard for the internal bus of desktop computers, to the recently developed SCI IEEE standard 1596 1992. Scalable farms (groups) of inexpensive PCIequipped desktop computers will connect to large detector systems through the PSB and the highly configurable SCI network to satisfy data rates and to process events from thousands of distributed subsystems. The Phase I research will develop a specification for the PCI/SCI bridge following the form of the proposed 1596.1 SCI VME bridge standard. This work will ensure compliance with both the PCI Local Bus Specification Revision 2.0 and IEEE Standard 1596- 1992. A preliminary design minimizing cost through use of commercial components will be developed, PCVSCI transactions simulated, and bridge latency and throughput performance estimated. Anticipated Results /Potential Commercial Applications as described by the awardee:It is anticipated that the high performance and versatile architecture of the PCI/SCI bridge will make it an essential building block in the data acquisition and computation systems for high-energy physics experiments now in the planning stages for upgrades or for new construction in the United States and elsewhere. Because of its low latency and high bandwidth, this bridge will be a component of low cost supercomputing clusters in a broad scientific and business market. Such cluster computing accelerates execution of software programs except those that make significant use of vector processing.
