The Parallel and Distributed System Performance TRC Laboratory and Multi-Processor Computing Hitachi Laboratory developed the "OpenMP compiler and automatic parallelization software" that simplifies parallel processing programming, and issued a press release at the Tsukuba Research Center (TRC) on November 8,1999.The press release is outlined here; for a detailed explanation. visit: http://www.rwcp.or.jp/press-release/press-99-09shiryo.html
|
-Parallel processing programming simplified by developing the parallel programming language OpenMP compiler and automatic parallelization tool that transforms existing programs into OpenMP programs - |
Real World Computing Partnership (RWCP) is conducting research on the development of a parallel processing technology that achieves high-speed information processing beyond the performance limit of a single CPU. In comparison with the programming of a single CPU computer, programming for parallel processing is very difficult. However, TRC and the Multi-Processor Computing Hitachi Laboratory of the RWCP have developed Omni OpenMP and WPP as software that simplifies parallel processing.
Omni OpenMP is a compiler of the OpenMP parallel programming language, which is compatible with Linux and can be used on various SMP computers with comparative ease. Although the OpenMP compiler currently available as a commercial product was only made in the U.S., in order to solve the problems of conventional use of a parallel computer, we analyzed a Fortran program written for a usual computer and developed WPP (Whole Program Parallelizer) software that automatically transforms it into a parallel program (OpenMP). By analyzing the whole program intellectually we successfully enabled efficient programming.
These software tools make programming for a parallel computer, which has been conventionally difficult, as simple as programming for regular computers, and greatly reduces the burden on the user of the parallel computer. This should help expand the parallel computer market.
These software tools are demonstrated at the research exhibition International Supercomputing Conference (SC99)* held in Portland, USA. in this November.
*See page 53 for details of SC99.