U3364521@ucsvc.ucs.unimelb.edu.au (Lou Cavallo) (03/29/90)
G'day, Please note that this is essentially a repost of an article I posted to comp. sys.amiga. I am reposting it to this newsgroup at the invitation of Valentin Pepelea (sp?). For those who may not have read the original posting it is in- cluded below (apologies for the bandwidth duplication). From comp.sys.amiga, article "Distributed processing on the Amiga. Is it possible?" > I'm not asking for help this time {I usually seem to :-)}. I'd like to ask a > (compound) question of the c.s.a readership. {Just for the kick of it! :-)}. > > Does/should the Amiga OS have the potential/capability in future for distri- > buted processing? > > By this I'm suggesting more than just networking. For example, is Amiga's OS > likely to be able to split up a job of work (say 3-D rendering of an object) > to be run concurrently on multiple Amigas? > > Do I ask too much? Seriously that is. > > {I debated whether I should post this to comp.sys.amiga.technical. I hope to > see some discussion of uses as well as the "fitness" of the existing AmigaOS > for this purpose. I would like comments from "application types" who may not > necessarily read that newsgroup.} > > {For Randell Jesup (sorry if I misspell your name). Is it in there? :-) ...} > > P.S:: Andy Tanenbaum has just announced a new book for Amoeba, a distributed > systems OS which sounds quite exciting. This was a motivation for my asking. > > Please note that I am not myself knowledgeable of the OS technical details. Some e-mail discussion (apologies, I won't repeat your names here) reminded me of Amiga X however I'm not aware of how/whether one could control other Amigas with X. Any details here somebody? I've since considered the question of distributed processing within an AmigaOS framework from a user perspective. I can't convince myself that enough home PC users would want it. I think they'd rather other features and would consider that this stuff would clutter up the OS (in code size and efficiency terms). How should that be handled? Perhaps with library code (distributed.library ?) so that it would be used only if asked for? {I'm guessing here folks, I don't know the Amiga internals from the back of my hand. :-)}. Any ideas? Any interest? :-) yours truly, Lou Cavallo.
jvmiller@ptolemy0.rdrc.rpi.edu (Jim Miller) (03/30/90)
This is slightly off the beaten path, but this discussion hits upon an idea I had a few months ago. I was using an AIS-5000 SIMD computer which had 384 processors (one processor for each column of an image). Since this was a single user machine and there were 10 of us in the class, it became apparent that it would be nice if there were tools for testing code meant for parallel machines on sequential machine. My idea (which does't look as though I will have the time to bring to life any time soon) was to create a series of handlers that would be virtual processors. There would be an instance of the handler for each processor you would like use. The operating system would handle all the time slicing of processes and transmission of data between processors. Since the AIS uses simple 1 bit processors, it shouldn't be too difficult to implement a handler that behaves in a similar fashion. I also thought that these "handlers" do not have to be all on the same Amiga. Soem of them could be set up as actually devices, or links to handlers on some other amiga. Since I am busy working on my thesis or keeping track of my sanity most the time, I will not have time to try anything like this for a while. I just wanted to see what people thought, if it was doable, or if I was just wasting my time. Jimmy Miller