"Just as an aside, to give you an interesting benchmark—on roughly the same system [(compared to the original Burroughs B5000)], roughly optimized the same way, a benchmark from 1979 at Xerox PARC runs only 50 times faster today. Moore’s law has given us somewhere between 40,000 and 60,000 times improvement in that time. So there’s approximately a factor of 1,000 in efficiency that has been lost by bad CPU architectures.
The myth that it doesn’t matter what your processor architecture is—that Moore’s law will take care of you—is totally false."
Wow. Really?! I'd be interested in knowing the specifics of what Burroughs did right.
On the other hand, computer hardware has also gotten drastically cheaper. How much did one of those Burroughs B5000 systems cost? Tens of thousands of dollars? (Hundreds?) I expect that spending that much on modern hardware and clustering it together would come quite a bit closer to what Moore's law would suggest.
Not everything can just be made parallel at the drop of a hat, of course, and the architecture itself is probably partially to blame for the difference, but there are also other major differences.
The point is that if modern computers had been based on that architecture, which was way, way more efficient for execution of high-level languages, today we'd have perhaps the same speed in GHZ but we could do way more because the high-level languages wouldn't need to execute at a level so far removed from machine language.
Worrying about how much those machines costed at the time is like worrying about how much a GB of ram costed at the time.
I agree with your underlying point, I just think that it's a little deceptive to compare a modern, $750 computer's processor to a $100,000+ mainframe's without acknowledging that they were also designed with a very different budget.
I was wondering about how much the processor itself influences things a few days ago while reading an interview with Charles Moore in which he described Forth-native processors, but I've done very little with any assembly languages directly, so it's kind of beyond my ken. (Most of what I know about different architectures has to do with porting C programs.)
Not only that, but he does a pretty good job of civilly saying, "Please prove me wrong, I'm seriously interested in how this could work." (Good link, thanks.)
Since we're on a meta-discussion, I'd point out that this raised some flags for me:
“A large project in C would implement the Lisp run time?” Oh really? You mean each variable will have the type LispObject (or PyObject or whatever)? Never happens, unless the C code is written by a deeply disturbed Lisp weenie (gcc and especially BetaPlayer, I’m talking about you). The fact that some people write C code as if they were a Lisp back-end is their personal problem, nothing more, nothing less.
No, that's obviously not what people who say that mean. Usually, they're talking about what I've most recently seen sold as an "IoC frameworks" like Java's Spring.
Meta-comment: He is awfully confident at ridiculing a claim he doesn't understand. If it's intentional, it's straw man rhetoric. Either way, it raises a flag.
Great interview. The page formatting was off, but the interviewer asked some interesting questions which go beyond the usual contemporary Java/Web material.
The myth that it doesn’t matter what your processor architecture is—that Moore’s law will take care of you—is totally false."
Wow. Really?! I'd be interested in knowing the specifics of what Burroughs did right.