XB.N54@Forsythe.Stanford.EDU (SCA) (04/03/89)
Supercomputer Code Advancement (SCA) announces the availability of a flexible service to provide optimization of Fortran code currently running on Cray X-MP supercomputer. Due to the high cost of Cray cpu time, the financial burden incurred and the market effectiveness lost from the execution of suboptimal codes cannot be overestimated. User's code is analyzed on or off site to discover current inefficiencies. Consequently, areas which are found to execute at less than maximum performance are systematically restructured to capitalize on the Cray's unique architecture. Common optimization techniques such as vectorization, gather/scatter, and asynchronous I/O are used, as are more esoteric techniques such as interleaving and proprietary methods that take advantage of anomalies in the Cray's design. SCA accepts code for analysis and restructuring under the premise that if the code's speed is not at least doubled, there will be no charge for services rendered, however, enhanced code is always supplied to the user. All speed improvements are based on those actually realized on the user's Cray system. -- John Lau Phone: (415) 759-7714 Executive Vice President of Sales Telex: 402833 Supercomputer Code Advancement Internet: XB.N54@Forsythe.Stanford.Edu 2510 22nd Avenue Suite 407 Bitnet: XB.N54@Stanford San Francisco, CA 94116 EasyLink: 62773175