Is it time to stop research in Computer Architecture ?
|
From: Robert Myers on 22 Oct 2009 14:37 On Oct 22, 5:38 am, Bill Todd <billt...(a)metrocast.net> wrote: > So I see no reason whatsoever to suspect that in 1994-5 Intel would have > considered the power requirements of a symbiotic full-fledged x86 core > to be a problem for Itanic - though I'd certainly be receptive to > credible information that could challenge that view. It's a little tedious to be challenged to defend a proposition you never made. You've put words in my mouth, proved to the satisfaction that the words you put into my mouth were wrong and then offered to let me rebut that claim. No thanks. Whatever Intel may have *thought* in 1994-1995 in making plans for Itanium, by the time they were shipping product, they had to have known they had a big problem. Thus, I interpreted "yup, we got x86 compatibility" as pure marketspeak denial of the reality of the situation. I may have been wrong in my interpretation, but x86 performance was wimpy and it seemed to me that power was a big problem. Intel was briefing to the issue of more and more complex cores and power constraints well before Itanium shipped. 1994-1995 was when Patterson was publishing about the problems industry was making for itself by making ever more complicated cores, and it would be pure fantasy to imagine that he didn't give that advice to Intel one way or another. Whatever Intel made of his advice in 1994-1995, by the time Intel shipped product, they had to know they had a big problem, which they may have thought they were going to fix, but it was a problem. As to Netburst, it had power constraints from the git-go as far as I've always understood it. I've been told that NetBurst went through some significant slimming down that cost it performance wise, and Bob Colwell basically admitted that it wasn't the design he had wanted to be, but hoped to fix that over time. That, of course, never happened. Robert.
From: Robert Myers on 22 Oct 2009 15:03 On Oct 22, 4:58 am, Andrew Reilly <andrew-newsp...(a)areilly.bpc- users.org> wrote: > On Wed, 21 Oct 2009 17:31:08 -0700, Robert Myers wrote: > I'm sorry. I think that I completely missed your point (or vice-versa).. > I don't believe that winning Spec performance numbers has anything to do > with success on the desktop. If intel had won race to 64 bits *and* had > a good backwards-compatibility story, and Microsoft had gone along for > the ride, (ie, both had wanted to), they'd have done it. I don't think > that the Itanium plan makes for a compelling laptop processor, though. > I don't know based on what other information you could claim that Itanium was the fastest processor on the planet. Intel/HP/SGI managed to keep Itanium at the top of the heap for SpecFP. Of course it has little to do with whether it makes a good desktop chip or not, but it was important to getting it into Top500 installations. If managers don't know how irrelevant Linpack benchmarked top 500 machines are to their needs, why would they understand the irrelevancy of SpecFP? Intel was most interested in selling Itanium as an enterprise chip. They just wanted the desktop market to support volume. > They can't have tried very hard, because it was pretty clear even then > that portability was important for client machines, and I don't recall > *ever* hearing a low-power story associated with ia64. Intel is a player in the portable market only because they had the resources to have a separate Banias effort. Whatever you may have understood about the importance of notebooks, Intel was slow to catch on. > > Didn't SGI open-source their own in-house itanium compiler, (open64 or > something like that)? Intel have their own compiler of course, and > that's both Linux and gcc compatible. Not sure if llvm does itanium. I > don't think that it was significant at the time, though. > Intel claims gcc compatibility for its compiler, but I don't know that anyone has ever built a linux distribution using anything other than gcc. I had to compile a newer version of gcc for Itanium from source. I had the intel compiler, but I never used anything but gcc to compile system code. So yeah, there were lots of compilers out there for Itanium on Linux and people were using them, including me, just not to build Linux. > > I still don't understand the argument. Sure, itanium wasn't going to > make anything any easier for open source, but there doesn't seem to be > any particular win for Microsoft, either. It has never seemed to me that > winning benchmarks was ever the argument for using open source, anyway. HPC is important to the Linux ecosystem. Microsoft wants that market if only to push Linux out. > Sure, but they're not the *only* folk who are about software, and I don't > recall them being any kind of force behind itanium compiler tech. That > was coming from intel, HP, SGI and a few others. That's a fair point, at least from what we know publicly. OTOH, I find it hard to believe that Microsoft was compiling software for Itanium using anyone's compiler but its own. Robert.
From: Robert Myers on 22 Oct 2009 15:23 On Oct 22, 10:42 am, an...(a)mips.complang.tuwien.ac.at (Anton Ertl) wrote: > Robert Myers <rbmyers...(a)gmail.com> writes: > >I've fiddled a little with the off-the-shelf Itanium compilers, but I > >always assumed that none of those compilers were even remotely good > >enough that you could expect just to run old software through them and > >get anything like hoped-for performance. John Dallman has had a bit > >to say on the subject here. > > >When I talked about rewriting code, I meant just that, not merely > >recompiling it. I wasn't all that interested in the standard task: > >how do you feed bad code to an Itanium compiler and get acceptable > >performance, because I was pretty sure that the answer was: you > >don't. > > Bad code? > It's a highly technical term, meaning code that is unnecessarily branchy and hard for a compiler to analyze for aliasing. > For most software, performance is not that much of an issue, and the > developers have left much more performance on the table than what > switching between IA-64 and other architectures, or between different > compilers for IA-64, or coding things in an IA-64-friendly manner > would have bought. > > To get an idea of how Itanium II performs compared to other CPUs on > code that's not tuned for it (at least not more than for any other > architecture), take a look at the first graphics (slide 4) on > > http://www.complang.tuwien.ac.at/anton/euroforth/ef09/papers/ertl-sli... > > This is performance per cycle, and the benchmarks are prety CPU-bound. > The compilers used here are various gcc versions (the fastest code > produced by the various gcc versions available on the test machines is > shown here). The only Gforth version that does not treat IA-64 as a > generic architecture is 0.7.0, and the only thing that's > IA-64-specific there is that it knows how to flush the I-cache. > > The performance per cycle of the Itanium II is not particularly good, > but also not particularly bad. > I'm not sure what you think this proves. It's been observed here more than once that, had Intel been able to clock Itanium as fast as it hoped to, it would have had a world-beater on its hands, at least for some kinds of tasks. Intel couldn't clock it fast because of all the architectural features, and, if you can't take advantage of those features, it's a dog. > >Could the open source community, essentially founded on > >x86, turn on a dime and compete with Microsoft running away with > >Itanium? Maybe with IBM's muscle behind Linux, open source would have > >stood a chance, but I'm not so sure. After all, IBM would always have > >preferred an Itanium-free world. Had I been at Microsoft, I might > >have seen a Wintanium future as really attractive. > > Microssoft obviously did not see it that way, because they eventually > decided against IA-64 and for AMD64. I don't know why they decided > that way, but I see two flaws in your scenario: > > To run away with IA-64, Windows software would have had to run on > IA-64 at all. Most of it is not controlled by Microsoft, and even the > software controlled by Microsoft does not appear to be that portable > (looking at the reported dearth of applications (including > applications from Microsoft) for Windows NT on Alpha). In contrast, > free software tends to be much more portable, so the situation on > IA-64 would have been: Windows with mostly emulated applications > against Linux with native applications. > > And would Microsoft have produced a better compiler for IA-64 than > SGI? > I'm assuming that those who used SGI's compiler used it for their own applications, not for building the system, as I commented elsewhere. Windows Server does run on Itanium. http://www.microsoft.com/servers/64bit/itanium/overview.mspx and it would appear that Microsoft is very happy to have those customers. After all, they are not running Linux or IBM proprietary on Power. Robert.
From: Bill Todd on 22 Oct 2009 18:41 Robert Myers wrote: > On Oct 22, 5:38 am, Bill Todd <billt...(a)metrocast.net> wrote: > >> So I see no reason whatsoever to suspect that in 1994-5 Intel would have >> considered the power requirements of a symbiotic full-fledged x86 core >> to be a problem for Itanic - though I'd certainly be receptive to >> credible information that could challenge that view. > > It's a little tedious to be challenged to defend a proposition you > never made. You've put words in my mouth, proved to the satisfaction > that the words you put into my mouth were wrong and then offered to > let me rebut that claim. No thanks. Sorry: I did you the courtesy of assuming that the words which you uttered were supposed to have some relevance to the subject you entered into discussing, and thus took them in that context - which developed as follows: You: As crazy as that sounds, it's the only way I can make sense of Intel's idea that Itanium would replace x86 as a desktop chip. [Clearly a statement about Intel's 1994-5 ideas, since that's when that idea was being bandied about. That's what I responded to, and have been responding to ever since.] Me: Did you forget that the original plan (implemented in Merced and I'm pretty sure McKinley as well) was to include x86 hardware on the chip to run existing code natively? [Still clearly talking about 1994-5.] You: I never took that capability seriously. Was I supposed to? I always thought it was a marketing gimmick. [Responding to my post regarding 1994-5.] Me: Why not? It ran x86 code natively in an integrated manner on a native Itanic OS. As with most things Merced the original cut wasn't impressive in terms of speed, but the relative sizes of the x86 and Itanic processors (especially given the amount of the chip area dedicated to cache) made it clear that full-fledged x86 cores could be included later if necessary as soon as the next process generations appeared. On-the-fly translation wasn't exactly common when the original plans were made (though the facilities for running x86 code on Alphas were probably well under way). Later on, after Intel had inherited a lot of those Alpha developers and the planned desktop takeover had become indefinitely postponed, it made more sense to do the work in software. [Still discussing why it was reasonable in 1994-5 (see the reference to 'the original plans') for Intel to believe that replacing x86 on the desktop was feasible - the original point under discussion (see above).] You: The die area may have been available, but I don't think the watts were. It's hard to remember with any accuracy what I knew when, but it's pretty easy to tell at least some of what Intel knew. By the second half of the nineties, Intel knew and briefed that power was going to be a big problem. [Still apparently referring to what Intel might reasonably have thought in 1994-5, since that was the explicit context of the post to which you were responding. I took the minor liberty of interpreting your phrase "By the second half of the nineties" as including the 1994-5 time-frame under discussion, since otherwise that part would have had no relevance to what we were talking about. Whatever you may have meant, you clearly were talking about a time *long* before "the time they were shipping product" (your phrasing in your most recent post).] Me: Not necessarily. Intel didn't have working silicon until some time in 1998 and were holding out hope for power reductions before shipping product well beyond that date (and further hope that McKinley would achieve whatever power targets Merced failed to). The decision to include the x86 core occurred far earlier (and Intel x86 cores at that time were still relatively stingy in their power requirements). [Yanking the discussion explicitly back to when the original plans were made.] You: I'm sorry. I should have been more explicit. Intel never admitted that power was a problem for Itanium, but I have a Gelsinger (Otellini?) briefing somewhere that extrapolates emitted power per area to that of a space shuttle heat shield tile for x86, c. 1997-1998. It was clear that they were headed to multicore, especially since Patterson was a consultant and he'd been saying similar for several years by then. [Given your current comments I can now see that this may not have been meant to address the point that had begun this sub-discussion, but - as I said at the start above - I did you the courtesy of assuming that it somehow was meant to and responded accordingly, which seems to have annoyed you.] Now, if you were just babbling away on your own without attempting to address the posts to which you were replying, never mind. Otherwise, reboot and interpret the discussion in the context of the original point: whether "Intel's idea that Itanium would replace x86 as a desktop chip" (an idea floated only in 1994-5 and conspicuously absent soon thereafter) was as 'crazy' as you thought it was. - bill
From: Robert Myers on 22 Oct 2009 20:03
On Oct 22, 6:41 pm, Bill Todd <billt...(a)metrocast.net> wrote: > Robert Myers wrote: > > On Oct 22, 5:38 am, Bill Todd <billt...(a)metrocast.net> wrote: > > >> So I see no reason whatsoever to suspect that in 1994-5 Intel would have > >> considered the power requirements of a symbiotic full-fledged x86 core > >> to be a problem for Itanic - though I'd certainly be receptive to > >> credible information that could challenge that view. > > > It's a little tedious to be challenged to defend a proposition you > > never made. You've put words in my mouth, proved to the satisfaction > > that the words you put into my mouth were wrong and then offered to > > let me rebut that claim. No thanks. > > Sorry: I did you the courtesy of assuming that the words which you > uttered were supposed to have some relevance to the subject you entered > into discussing, and thus took them in that context - which developed as > follows: > > You: As crazy as that sounds, it's the only way I can make sense of > Intel's idea that Itanium would replace x86 as a desktop chip. > > [Clearly a statement about Intel's 1994-5 ideas, since that's when that > idea was being bandied about. That's what I responded to, and have been > responding to ever since.] > This is yet another bit of blind men feeling different parts of the elephant. Itanium has a long and fascinating history, and each of us tends to emphasize different parts depending on what happened to grab our attention. You can ask different questions. Where did the delusion originate (your question)? Why did Intel hang onto that delusion so long in spite of evidence that it was a delusion (my question)? In groping for an answer to my question, I have wandered all over the history of Itanium, going so far as to speculate that Intel should have known better in 1994-95. My speculation could be way off target, but it is just that: speculation. I don't find legalistic discussions to be very helpful. History is messy. Technology is messy. If there are lessons to be learned, I'd like to learn them. I'm not all that interested in being right. Robert. |