Is it time to stop research in Computer Architecture ?
|
From: Bill Todd on 21 Oct 2009 07:20 nmm1(a)cam.ac.uk wrote: > In article <kKudnelkCfoGQEPXnZ2dnUVZ_t2dnZ2d(a)metrocastcablevision.com>, > Bill Todd <billtodd(a)metrocast.net> wrote: >>>> Do you really think that Merced in a then-current process "_would_ >>>> have been, by far, the fastest cpu on the planet" - especially for >>>> general (rather than heavily FP-specific) code? ... >>> Oh, yes, indeed, it would have been - if they had delivered in 1997 >>> what they were promising in 1995-6. >> That's irrelevant: the question was how an actual (shipped-in-mid-2001) >> Merced would have performed if delivered in 1997 (possibly 1998) using a >> then-current process, not how some vaporware being talked about earlier >> might have performed. > > You may have gathered that I agree with you from the rest of the > paragraph :-) Sometimes it's difficult to be sure. Ironically, both you and Robert have somewhat similarly subtle ways of conveying sarcasm. I'm actually kind of interested in whether I might be wrong here, since I don't often find myself on the opposite side of a technical issue from Terje (not that I viewed his assertion as necessarily any more than an off-the-cuff remark that he may not have thought much about). > > But the insoluble problems were software and not hardware - they > could, at least if they had handled the project excellently, have > delivered what they claimed in terms of hardware. Even that observation is not really relevant to the question here - because Terje's assertion (at least as I read it) refers to what they actually delivered (had it been delivered earlier), not to what they might have hoped to deliver. - bill
From: eternal september on 21 Oct 2009 07:23 Hello all, "Andy "Krazy" Glew" <ag-news(a)patten-glew.net> wrote in message news:4ADEA866.5090000(a)patten-glew.net... > I look forward to slowly, incrementally, increasing the scope of the > dataflow in OOO machines. > * Probably the next step is to make the window bigger, by multilevel > techniques. What is your favorite multilevel technique. I don't think I ever heard your opinion on HSW (Hierarchical Scheduling Windows http://portal.acm.org/citation.cfm?id=774861.774865)... Thanks, Ned
From: ChrisQ on 21 Oct 2009 07:24 kenney(a)cix.compulink.co.uk wrote: >> so I naturally wonder, >> what's happened in the meantime ?. > > Rant mode on > > Software bloat. The programs I use except for games have not visibly > increased in speed since my first PC. They have developed a lot more > bells and whistles but not got faster. DOS could run programs and a GUI > (GEM) in 512kb of memory. Windows 3.1 would run with a mb though it > needed 4mb to get maximum performance, I understand that Windows 7 has a > minimum requirement of 2gb. Just about all the increases in hardware > speed have been used to run more elaborate software at the same speed. > > Rant mode off. > > Ken Young Agreed, but the graphics is orders of magnitude better and that's partly where all the spare cpu power has gone. But little has changed in fundamental architectural terms for a very long time. When I take the lid off any new machine and see a device whose function I can't estimate within a minute or two, then I shall become curious about the progress of computing again. I caught the first 5 or 10 minutes of some physics character on tv a few days ago. The sort of pop science that the beeb do so well, with the dramatic music, graphics and odd angle shots of the presenter trying to look wise and intelligent. His thesis was that we would soon have intelligence in everything, with billions of microprocessors. My first reaction was who will write all the code for this ?, as there's already a shortage of embedded programmers who really know what they are doing. Probably India and China, but I digress. In a roundabout sort of way, he was right and it's already happening as more and more applications for embedded devices appear. If you track the development of computing, the story is one of an initial ivory towers world inhabited primarily by mathematicians, to a commodity used for amusement by everyone, but commodity means standardisation and dumbing down far enough to make it cheap. Now we are seeing the next stage, with some of the the functionality of the desktop becoming obsolete in favour of distributed computing. Separate appliances with lower throughput and less power hungry. In the old days, there were a wide variety of system designs, whereas now there are few. Why ? - well partly because the old machines were built from small scale devices. Bit slice, ttl, etc, with the wiring between them and perhaps microcode defining the overall architecture. The state of technology produced a level playing field that anyone with a little capital and a good idea could exploit. Now, the state of technology is such that only the largest corporations can afford to develop new devices. It's completely frozen out all the small vendors in terms of visibility. It's well known that the mainstream never really develops anything really new. It exists merely to maintain the status quo, perhaps with a few crumbs of improvement from time to time. It's the bits at the edges where all the interesting stuff happens, but there are effectively no edges left now. Mainstream computing has dug itself into a very deep hole, which is why i'm so pessimistic about future developments... Regards, Chris
From: Bernd Paysan on 21 Oct 2009 08:17 ChrisQ wrote: > In the old days, there were a wide variety of system designs, whereas > now there are few. Why ? - well partly because the old machines were > built from small scale devices. Bit slice, ttl, etc, with the wiring > between them and perhaps microcode defining the overall architecture. > The state of technology produced a level playing field that anyone with > a little capital and a good idea could exploit. Now, the state of > technology is such that only the largest corporations can afford to > develop new devices. Sorry, that's not true. Especially in the "small embedded intelligent device" area we are talking about. The scale of integration changed: You will produce a SoC for these applications. I.e. the parts are build from incredible small devices: GDS polygons, to be precise (people use more convenient building blocks, though). Most people who make SoCs embed some standard core like an ARM (e.g. Cortex M0) or an 8051 (shudder - takes the same area as a Cortex M0, but is horrible!), but that's because they chose so, not because it's not feasible to develop your own architecture. Architectures that usually don't surface to the user - e.g. when I embed a b16 in a device of ours, it's not user-programmable. It's not visible what kind of microprocessor there is or if there is any at all. -- Bernd Paysan "If you want it done right, you have to do it yourself" http://www.jwdt.com/~paysan/
From: nmm1 on 21 Oct 2009 08:27
In article <DHADm.42619$XI.25627(a)newsfe24.ams2>, ChrisQ <meru(a)devnull.com> wrote: > >But anyway, getting back on topic, how about von neumann is dead ?, and >what could replace it ?. While Von Neumann is dead, Von Neumann computing isn't, and current designs are almost all multiple Von Neumann threads with coherent shared memory. Except for GPUs and some specialist systems, where the coherence is patchy. Note that I am not saying that is a good approach, merely that it is what we have today. Regards, Nick Maclaren. |