Traditionally parallel computers have provided a development environment based on the UNIX operating system. However in the past couple of years, the quality of the programming environment on personal computers has eclipsed that of UNIX workstations. The tiny size of the high performance parallel computing market precludes most commercial software companies from supporting their software on high end systems. However, by building parallel systems using commodity hardware and system software we will be able to use existing applications.
Parallel computing research has produced a wealth of ideas and of prototype software for carrying out data parallel computations on parallel computers and distributed computers.
Current projects using the PCHASM labratory include:
One project is to adapt parallel program coupling methods to microcomputer networks. This research should permit the coupling of commercial sequential applications with parallel applications via open interfaces such as Microsoft's OLE (Object Linking and Execution) and OMG's CORBA (Common Object Request Broker Architecture). This will leverage PC software for much of the functionality (e.g., user interfaces), yet employ parallel computation where higher performance is required.
As described above, we intend to create data parallel libraries that can be used by applications programmers. By using OLE and CORBA, we hope to create plug and play replacements for existing sequential libraries. If successful, this approach will permit application users (rather than programmers) to parallelize their applications by substituting data parallel libraries.