APPLICATIONS

PathScale, developer of innovative software and hardware solutions to accelerate the performance and efficiency of Linux clusters, announces the PathScale OptiPath MPI Acceleration Tools. OptiPath incorporates the best MPI programming expertise available, and gives scientists, researchers and MPI experts direct visibility into complex MPl program and cluster behavior. It automates performance analysis and provides specific expert-level tuning advice that can be used to improve time to results and cluster efficiency. This announcement is being made at the SC2004 SuperComputing Conference in Pittsburgh. While MPI is the most widely used standard for developing parallel applications, writing efficient MPI code is extremely difficult. PathScale OptiPath, the first software of its kind, addresses a key weakness in MPI -- the need to manually analyze and improve application and cluster-dependent performance. By automating the performance analysis process, OptiPath lets scientists focus on science rather than on tuning MPI code. "We've been using the PathScale EKOPath compiler suite for the past year and the performance improvement to our code has had a dramatic impact on efficiency," said Martin Cuma, Scientific Applications Programmer for the University of Utah. "There is a clear, long-standing need for tools that make MPI as efficient as possible, and we're looking forward to being a beta tester for OptiPath, and to deriving the same level of benefits we've gotten from EKOPath." Writing Efficient MPI Applications PathScale OptiPath MPI Acceleration Tools distill years of MPI learning and expert knowledge to lead developers to understand the root cause of their performance problems. It automates MPI program testing and guides the software developer to the most important performance problems that impact efficiency. It then directs the developer to proposed solutions and allows them to see potential results before investing in the changes. "MPI is a very specialized, complex programming environment, and with use of Linux clusters continuing to grow, the need for efficient MPI code is compounded," said Scott Metcalf, CEO of PathScale. "By automating analysis and resolution of the most common problems, organizations will free up their scientists from the drudgery of MPI coding and enable them to get back to doing what they do best." Availability "There is significant, pent-up demand for tools to make MPI more efficient and easier to use," said Winston Prather, VP/GM of HPC at Hewlett Packard. "OptiPath has the potential to significantly ease the use of clusters in parallel programming, and we're very excited about making this tool available to our customers." More than 25 Linux cluster technology providers in North America, Europe and Asia have joined the PathScale FastPath reseller program since it began in April, 2004. PathScale will utilize this same world-class distribution channel to sell OptiPath when it becomes generally available during the spring of 2005. OptiPath supports both 32-bit and 64-bit x86-compatible AMD and Intel architectures, including the AMD64 and EM64T processors. OptiPath supports the major Linux distributions used in HPCRed Hat, including Fedora Core, SuSE Enterprise and SuSE Professional. PathScale OptiPath supports MPICH and other popular MPI implementations and works with PathScale's InfiniPath interconnect, Gigabit Ethernet, and Infiniband host adapters and switches.