bruce@kai.com (Bruce Leasure) (02/20/91)
> Subject: Info requested on Parallel Computing Forum (PCF) > Could anybody please send me information on the Parallel Computing Forum. My name is Bruce Leasure. I was the executive director of the Parallel Computing Forum during the time of its existance. > What are its primary goals? PCF had one primary goal: to achieve the standardization of both the syntax and the semantics of basic forms of parallelism. There was a strong effort to only standardize, not to invent. There was also a strong effort to concentrate on the basics, not anything advanced or special purpose. PCF decided to concentrate on the simplest problem first: parallelism for fortran like languages on a machine that presents a shared memory model to the programmer. The members of PCF believed that early standardization of the basics would enable the field to expand faster because of increased use. > Which companies support PCF? Almost every vendor of computer hardware from workstations to supercomputers in the US was involved either through active participation, or through the mailing list. There were many entries in the mailing list from Europe and Japan, though every company was not involved. The major suppliers of distributed memory machines were present on the mailing list, and a few participated in the meetings. > What is the difference between PCF-Fortran and Fortran 90? PCF-Fortran is concerned with parallel processing, while Fortran-90 has vector/array extensions. In some sense, Fortran-90 provides a mechanism to represent operation level parallelism (through array valued expressions) while PCF Fortran provides mechanisms for expressing parallelism between blocks of code. PCF made a strong effort to not attempt to form a standard for those concepts that Fortran-90 was going to standardize. > Which of them is better suited to vectorization PCF Fortran does not address vectorization at all. > resp. parallelization? Fortran-90 does not address parallelism between statements at all, not does it provide any parallel constructs other than operator level parallelism (array expressions). > Will PCF-Fortran probably become an ANSI or ISO standard? An ANSI committee has been formed (X3H5) and has been meeting for about a year. This committee is charged with defining a model for parallelism for high level procedural languages, and for providing a mapping of that model onto Fortran-77, Fortran-90 and C. Preliminary forms of all of these documents exist. If someone is interested in starting an ISO committee, I am sure X3H5 would support it. > Are there activities concerning parallel C or towards combining > C and Fortran in PCF? yes. see previous answer. ---- I hope this information helps. I have also enclosed a copy of the X3H5 document that explains how to join X3H5 ------ How to Join X3H5 ------------------------- Title: How to Join Author: Bruce Leasure, Walt Rudd Document Number: X3H5/90-0005-B Date Assigned: 09/15 Project Number: -- To join X3H5, Parallel Processing Constructs for High Level Programming Languages, simply send a letter, on company letterhead, to the Vice Chair stating that you wish to join, that you have an interest in parallel processing, and if you will be an active member or an observer. Include the following additional information for the membership list: Princible Member Name Mailing Address Voice Telephone Number Fax Telephone Number (if avaiable) Electronic Mail Connection (if available -- INTERNET) Alternate Member Name Mailing Address Voice Telephone Number Fax Telephone Number (if avaiable) Electronic Mail Connection (if available -- INTERNET) Both the princible and the alternate will be added to the membership list, and the names will be forwarded to ANSI X3. ANSI X3 will invoice the princible member for the $250 per year membership fee. The rules on voting at X3H5 are spelled out by X3. Voting is by active members only, with only one vote per company. To vote at an X3H5 meeting, the company (princible member or alternate) must have attended two of the last three meetings (counting the current meeting as one of the three meetings). If you don't come, you are sent a letter of warning, and if you still don't come, your membership status is automatically downgraded to observer. The Vice Chair may be contacted at: Walter G. Rudd Department of Computer Science 303 Dearborn Hall Oregon State University Corvallis, Oregon 97331-3202 rudd@cs.orst.edu 503-737-5553 fax: 503-737-3014