Posts by Dominic Sweetman

Poppadom

Itanic was doomed by the time it was announced. The idea of VLIW was to increase parallel execution, but out-of-order CPU implementations exploited the same opportunities much better. Itanium was conceived at a time when out-of-order was first being contemplated and fought, and found to be terrifyingly difficult; I got the impression that early OOO attempts like the MIPS 8000 burned out half a generation of brilliant engineers... But as often happens, people worked it out and suddenly it all seemed possible.

The conspiracy theorist in me suspected that the smart people at Intel were not really blind to the obvious flaws, but saw it as a great way to divert corporates who might otherwise turn to RISC architectures, at least until Intel could make something which worked (as it turned out, until AMD could make something that worked). But it might have been that the decision makers got caught up in their own hype.

At least the Register called it pretty early...

Poppadom

5W is not much power, so won't cook anything. 30-300GHz waves (see wikipedia) won't travel long distances through air, so are not likely to be used for regulated communication. They'll have no biological effect except for heating. Presumably the transmitter only turns up the power when it finds a friendly chargee, so walk through the beam and it turns off. Small-device chargers are never going to represent more than a small fraction of anyone's electricity use, so efficiency isn't critical.

Too clever by half? Perhaps. And since all companies always want to make incompatible chargers, your ceiling will be festooned with battling chargers...

As the man said, Ubuntu is available and has the explicit goal of being easy to get into at the cost of strict "free-ness" and less flexibility. So having Debian feeding Ubuntu (and in details, often vice-versa) works for all of us. Though perhaps the Debian install pages should note that wi-fi usually won't work on a laptop without fiddling, and the user will do well to invest in an ethernet cable to keep going until the fiddling is done.

At a few GHz, CPUs are memory-limited. Once you get on-chip caches you can make the CPU faster, until cache misses dominate the workload. A 2GHz wide-issue CPU with cunning insides can probably perform 3-4 instructions per nanosecond. In classical PC a DRAM access must go off chip, off module, through a connector, across the tracks and through a DRAM interface. That's going to take perhaps 80-100 nanoseconds: could be longer, and the more expandable the memory is the longer it will take. That represents about 250-400 instructions you didn't execute because you're waiting for memory. Big caches lower cache miss rates, but only to the low single-digits of percentage. Fast CPUs running one thread (most laptops have one impatient user waiting for one thing to happen) spend the great majority of their time waiting for memory.

Meanwhlie, Moore's law continues to work for memory density. A laptop with 16GB of memory sounds pretty usable, and should get you a big performance and battery use. A trade-off well worth making.

An opportunity...

I'm guessing that router software is complicated, but well understood. The trick would be to use Huawei hardware with the firmware rebuilt from source code by a team of open-source-world programmers, under Ross Anderson's supervision. If you could negotiate and build such a thing, you'd have everything the UK needs. If you could keep GCHQ out of it, it might also be an international hit.

Old MIPS hacker

Early RISC CPUs (late-80s MIPS and SPARC) really were a lot faster than contemporary x86. That was because the instruction set was redesigned to pipeline nicely. Once out-of-order CPUs were worked out and built successfully (though perhaps on the graves of burnt-out CPU designers...) the RISCs had only a small edge on x86, which Intel's better design/silicon easily overwhelmed.

The approach described is an attempt to create an instruction set which better uses an out-of-order core. This is an interesting idea. Probably only Microsoft, Apple and Google have a big enough codebase and captive audience to develop such a thing.

The 40,000 drivers came from the same Ubertweet as the 3.5M users.

The 3.5M users would be rather more than half of Londoners who have both a credit card and a smartphone. It is wholly incredible, if they mean any sensible definition of "user" (might be "has called Uber twice"). I suspect they're adding everyone who ever downloaded the app plus all the people in every cab, then multiplying by two.

40,000 drivers? I doubt it. Halve it, then assume most only ever worked a few hours.

Simple RISC CPUs (one pipeline, in-order) work quite well while the primary cache returns data in a very small number of cycles -- perhaps two. At GHz speeds, that's impossible. Out-of-order execution makes your brain hurt, but it keeps a CPU reasonably busy while waiting for the data.

CO2 levels are 60% higher than in the late 18thC, and are still rising fast. There's a very simple mechanism which causes it to raise global temperature. Climate is far more complicated than we can understand: perhaps there is some not-yet-known feedback mechanism which means we won't get much warmer. But given the least worst science available, most likely it will make it quite a lot warmer rather fast, which is likely to be difficult for our grandchildren. If we had another planet or two for us to move to if it turns out badly, that might be a reasonable chance to take.

But we haven't got another planet.

We know how to stop adding CO2. It wouldn't even cost that much. We won't do it. That's amazingly stupid, and we don't like admitting to stupidity, so we grasp at straws. Or scandinavian tree-rings, of course.

Battery charging at home

A correspondent does the maths implicit in your article... Yes, charging my phone's 1Ah battery in 9 seconds would require a current of 400A. But that's only at 3.7V, so the amount of power involved is about 1.5kW, half of an electric kettle and well within the maximum of a regular 13A/240V plug. Surprisingly high currents are provided in domestic environments -- your PC power supply feeds its CPU chip with over 90A of current. Nine seconds: hardly. Three minutes? there's no obvious power supply reason why not.

Why choose

A MIPS CPU is a lot simpler than an x86, which means less transistors in the CPU's heart. But almost all the transistors on a CPU chip are cache and interface anyway, so it won't ever make up for the huge economies of scale in the x86 clone market. But while you can't buy a chinese x86 yet (for legal and political reasons, you can get the chinese MIPS.

It will never run Vista, so you won't even be tempted. That's got to be a plus, too...

CPUs need low latency, not high bandwidth

Fast CPUs need their interconnect to provide the data they need, fast (they're probably stalled waiting for that data). While too-low bandwidth gets in the way, the critical factor is more often the time until the first datum you asked for arrives ("latency"). Prof Roger Needham used to say that "bandwidth can be made by man, but God makes latency".

Nobody ever quotes the minimum latency which is practically achievable with their interconnect, because as the interconnects get more amazing, it generally gets worse... So this bandwidth claim, like almost all other such, is irrelevant.

BTW, optical enthusiasts should probably try to convince us that four electric/optical transitions in their favourite path doesn't add delay.

Copyright on scrabble

If it was just a trademark, "Scrabulous" could just change its name (assuming it's plausible that it confuses -- it seems pretty distinct to me).

Any patent (there was a 1956 US patent relevant to the scrabble board) expired long ago.

So what's left is copyright. The inventor of scrabble died only in 1993 and -- in the UK at any rate -- I believe this is treated like an artistic work, so that's 70 years after the death of the author. But apparently courts are inclined to limit the copyright to fairly precise reproduction of artwork... this could run on.