reiher@amethyst.Jpl.Nasa.Gov (Peter Reiher) (11/14/88)
In article <3485@hubcap.UUCP> khb@Sun.COM (Keith Bierman - Sun Tactical Engineering) writes: >In article <3465@hubcap.UUCP> wilson@uicbert.eecs.uic.edu writes: >> >> 2. Are there any explicitly optimistic programming systems, with >> language constructs that specify/allow optimistic execution? > >TimeWarp the first operating system with virtual time. Contact Brian >Beckman at JPL, Brian is still here, but his major focus is no longer Time Warp. I can probably answer any questions you have about the JPL version of Time Warp. Other research implementations of Time Warp exist at Rand and Rockwell, and perhaps elsewhere. A commercial version is in the works. >the idea was formulated by a UCLA professor (still there, I think), David Jefferson, who is indeed still at UCLA. He consults on the JPL Time Warp project. He's on the net, but I doubt if he spends much time looking at news. >the first hypercube implementation was done by Brian and some folks reporting >to him (mostly him, I think). . . .At one time there was a Mac simulator, a Sun >network implementation and a JPL Mark II hypercube implementation. I >don't know the projects current status. Currently, our version of Time Warp runs on a MarkIII hypercube, on a Butterfly, and on Suns (both 3's and 4's). Experimental versions have been ported to the Intel hypercube, to Ametek machines, and to a transputer-based system. The software is in pretty good shape, now, and the thrust of the research for the upcoming year is to apply dynamic load management to Time Warp. >> >> 3. How promising are optimistic techniques, really? > >Brian's results are extremely encouraging (as of a couple of years >ago). They built discrete simulations which worked. Got the speedup, >not infinite time loops. What more can we ask of a research project ? Good speedup? We've gotten that, too. We have several papers in the distributed session of next year's multiconference on simulation which present what we consider to be very good results on a couple of applications. It looks to me (an admittedly biased observer) that optimistic techniques are very promising indeed. >> 5. Do any optimistic systems let you play with a general heap, >> or do they impose major restrictions (like all data being >> allocated on a stack, or requiring that the compiler be >> able to determine statically when something will become >> garbage)? >TimeWarp is (in principle) object based, and built upon message >passing. So, IP, the programmer shouldn't know about the stack(s) etc. >Current reality is probably more primative, but it's research. We have added dynamic memory allocation to Time Warp, working more or less on the malloc() model. >>My own interest in this stems partly from the fact that I have devised a >>couple of checkpointing and recovery mechanisms that efficiently checkpoint >>heaps. I'd like to hear more about this work. Could the original poster send me more details in email? Peter Reiher reiher@amethyst.jpl.nasa.gov (DO NOT send to reiher@amethyst.uucp) . . . cit-vax!elroy!jato!jade!reiher