budd@mist.cs.orst.edu (Tim Budd) (05/24/89)
Regarding APL and Parallelism I've been an advocate of APL for parallel processors for a long time (see my TOPLAS paper in 84 and my book ``An APL Compiler'' published by Springer in 1988), so I was interested in Leigh's comments passed on by Eugene. Some comments on his remarks: (1) My students and I continue to pursue approaches to generating efficient code from APL for a variety of machines. Our latest technique involves a nice intermediate form based on the lambda calculus. We can translate APL (and FP too!) into this form; perform transformations and optimizations at this intermediate form level, then generate code from the intermediate form. The interesting thing is that the intermediate form does not involve any explicit control flow, and thus we can introduce parallelism during code generation in a variety of ways. This allows us to generate fine grain SIMD (connection-machine style) code by introducing parallelism one way, and coarse grain, MIMD (sequent-style) code by introducing parallelism in another way. There are also interesting time/space tradeoffs which we are treating as a *code-generation* problem and not something the programmer needs to deal with. If you are intersted in finding out more about this let me know and I can send you technical reports. (2) APL is, despite being an anathema to the academic communnity (for reasons which I have never quite understood) still alive and well. Following the publication of my book I was contacted by several commercial wall street types offering to throw large sums of money my direction if I would develop a commercial quality compiler for APL. Being a silly college professor, I resisted these because (1) I don't think I understand the problems well enough to make such a system yet and (2) there are a few dark corners of the language I've been able to ignore since I'm just ``doing research''; if I were serious about a commercial product I would have to really look into this, and (3) if I was interested in making money I would never have become a college professor in the first place. --tim budd, budd@cs.orst.edu
phmb@genrad.com (Peter Brooks) (05/25/89)
I am very interested in promoting APL. I feel that it is the most natural language to teach programming in for one thing. What are these 'dark corners' you refer to? I think I might enjoy looking into some dark corners - unless they are to do with screen formatting or some such non sexy issue. Please let me know. Peter Brooks