narem-james@yale.UUCP (02/10/89)
From: "James E. Narem Jr." <narem-james> Announcing the Linda Mailing List Recently we've gotten many inquiries regarding Linda newsgroups or mailing lists. We feel a newsgroup would be premature, but we've started a mailing list. If you'd like to be on it, fill in the form appended. Just What is Linda? Linda is a set of operators for explicitly parallel programming. The opera- tors are embedded into various conventional programming languages to produce parallel programming languages. Most of our work has been based on Linda embedded in C: C-Linda. We have an experimental Fortran-Linda system; an Ada-Linda project is of interest to a number of sites, and may start soon. Efforts we know of at other places involve PostScript-Linda, C++-Linda, Modula-2-Linda, Scheme-Linda, Common Lisp-Linda, and some object-oriented Linda-like systems. Linda is "low-level" in the sense that it does not attempt to hide parallelism from you: We don't claim to take a "dusty deck" and produce a highly parallel program from it. But Linda is very "high-level" in the sense that it hides the details of the hardware, unless you insist on looking. Parallel programs written in Linda are *portable*, across machines with very different archi- tectures. We have programs we regularly run, unmodified, both on Encore shared-memory machines and Intel hypercubes. Why Linda? People tell us they like Linda because it's powerful but simple. Its simpli- city means that it's easy to get an implementation of some kind up quickly, and many people have done this. Building an *efficient* implementation, how- ever, can be very tricky. For groups seriously interested in Linda on hard- ware or in language environments we don't support, a good approach is to start by porting our system rather than building from scratch. We're happy to ar- range for this kind of thing, and have done so with several groups. Linda isn't wedded to any single supporting language; it is capable of thriv- ing in a wide variety of environments, in climates ranging from C and Fortran through object-oriented systems and logic languages. This is one reason why there's been a surprisingly wide spectrum of interesting Linda-related work at various places. Most of the non-Yale Linda projects are still fairly early on, but we'll be happy to disseminate information about these to people who are interested. Linda has been used for a pretty wide variety of purposes, ranging from nume- rical codes, to simulation, to theoretical parallel algorithm design, to bio- logical applications, to on-line intelligent monitoring. The more, and more diverse, the range of applications, the more we learn; so we are always inte- rested in talking to people who want to try out the language. Well over a dozen papers about Linda have appeared in the literature, most recently in last summer's ASPLOS issue of IEEE Transactions on Computing, last summer's SIGARCH conference and SIGPLAN Parallel Programming symposium (PPEALS), Supercomputing '88 and November Byte. Except for the Supercomputing paper, which deals with the VAX network system, everything in print about Linda is out of date. If you want up-to-date reports or users' manuals, drop us a line. How Can I Get Linda? At present, beta test versions of C-Linda are available for the Sequent Bal- ance and Symmetry and the Encore Multimax. Also available is a compatible single-processor simulator for Sun workstations. For further information about these implementations, contact us or Dennis Philbin at Scientific Com- puting Associates in New Haven (203 777-7442). The Encore system is available directly through Encore as well. C-Linda for the Alliant FX/8, the Intel iPSC/2, and VAX/VMS multiprocessors and networks has been distributed for research purposes to cooperating groups. The Linda Users Mailing List To be added to the Linda users mailing list, please fill out the following form and send it to: linda-users-request@yale.cs.edu !Internet linda-users-request@yalecs.bitnet !BITNET [ucbvax!]decvax!yale!linda-users-request !UUCP Only the first two items on the form are really necessary, but we'd appreciate information on the rest as well. Omit any you don't want to answer. Please provide enough address in your Email address so that we can be sure of getting mail to you - if we can't use the address you give us, there isn't much we can do to tell you so! To send mail to the list, use the appropriate address form as above, but send to "linda-users": linda-users@yale.cs.edu !Internet linda-users@yalecs.bitnet !BITNET [ucbvax!]decvax!yale!linda-users !UUCP Gives us a couple of days (say, until February 15th) to prime the list before sending to it. We've had interest from non-network sites in a hardcopy newsletter; if you are interested in an initial mailing, please include your US Mail address. -----------------------------Cut Here and Return----------------------------- Your name: Your Email address: Your institutional affiliation: Your mailing address: Parallel processing systems/languages you have used: Parallel processing hardware at your site: Parallel processing systems/languages you have used: Parallel processing systems/languages at your site: Kinds of parallel programming you are interested in: