Path: utzoo!attcan!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!swrinde!zaphod.mps.ohio-state.edu!sol.ctr.columbia.edu!emory!hubcap!pratt From: pratt@cs.stanford.edu Newsgroups: comp.parallel Subject: Re: parallel programming model Message-ID: <12316@hubcap.clemson.edu> Date: 17 Dec 90 17:53:38 GMT Sender: fpst@hubcap.clemson.edu Lines: 35 Approved: parallel@hubcap.clemson.edu In-Reply-To: Your message of Mon, 17 Dec 90 10:37:04 -0500. <9012171537.AA02225@hubcap.clemson.edu> From: caliban!ig@uunet.UU.NET (Iain Bason) I don't see the point of this argument. The original poster was... ... In fact, Von Neumann machines and parallel processors are the same at some level (e.g., they are both digital computers). I don't see the point of this argument in this context. The effect of levels on complexity is conjunctive: if all levels save one are distributed and that one is shared memory, and the shared memory turns out to be a bottleneck, then it kills your performance no matter how fast the other levels are. At the level at which Von Neumann machines and parallel computers are the same, both are distributed. The discussion was whether a particular level was distributed or centralized (i.e. von Neumann). Assuming this is the only level that hasn't yet committed to distributed, that's the only level that pertains to the discussion. Protocol implementations that mistake protocol levels for an implementation guideline instead of merely a way of structuring the semantics take a big performance hit due to this conjunctiveness of per-level computational complexity: every level slows you down a little bit more. If *any* level of Internet were implemented by looking things up over at NIC per packet it would slow you down a *lot* more. However it would not impact the main purpose of networking, which is not performance but rather simply to spare people the inconvenience, cost, and unreliability of moving data around on magnetic tapes and other portable media. Great performance is great, but not as great as convenience, economy, reliability, or sex. Vaughan Pratt