* Posts by BlokeInTejas

34 posts • joined 3 Mar 2016

China's 7nm chip surprise reveals more than Beijing might like

BlokeInTejas

Re: Ours

"Ours"???

You mean in the UK?

But the UK has been controlled by EU (and local) greenery for decades. Fabs are not only very expensive, they're not particularly green. And they require lots of customers.

The EU is so far behind in 'high tech' they're spending billions of euros trying to attain "digital sovereignty". Going to be a very long slow expensive stern chase. Soon Europe (not American or Taiwanese fabs in Europe, but European-owned fabs) will be able to build stuff that the rest of the world was doing in 2005 or so. There are European semiconductor companies; NXP, ST and Infineon exist and build stuff. When they use their own fabs - NXP uses TSMC for their advanced stuff - it's not very state of the art (see the recent announcement about ST and Global Foundries building a fab near Grenoble).

Nowt to do with greed. Much more to do with the dead hand of the EU. The EU is much more concerned with diversity, keeping people employed regardless of the merits of the company, keeping the dammed furriners out (especially Americans, because they're just too effective), reducing inequality - all nonsense that really gets in the way of high tech (or indeed any business).

STMicroelectronics and GlobalFoundries to build wafer fab in France

BlokeInTejas

True.

It’s a bribe.

Apple's guy in charge of stopping insider trading guilty of … insider trading

BlokeInTejas

Ain't no cure for Stupidity

The simplest analysis is not that he's a vile opportunist, grabber of money while trampling the poor underfoot, nor any of that...

It's simply that he's an incompetent idiot.

Or that he had a giant payment to be made to someone with a trusty squad of knee-breakers to hand.

RISC-V International emits more open CPU specs

BlokeInTejas

Re: and the inter-web melts in China, Russia, Iran, Cuba, North Korea, ...

Architecture is just paperwork.

To replace your tank’s fire control system you need actual silicon. A bunch of pdfs won’t do much for you.

Vlad’s paradise doesn’t make chips.

New York to get first right-to-repair law for electronics

BlokeInTejas

The right to repair is a legal thing.

They'll sell you parts and info at a fair price....

Doesn't mean the thing is practically repairable...

You really think you can replace that M1-Ultra? and have it work? For how long? And how much is that M1-Ultra? Purchased in thousands, it's probably sorta OK - but thousands will cost quite a substantial sum, but one by one to an individual?

California Right-to-Repair bill quietly killed in committee

BlokeInTejas

..so are you. And Bill Gates. And a putative 'Right to Repair' orhganisation.

What's the problem?

But having a right to repair doesn't mean it's feasible to fix a specific thing. When a cellphone is just 4 chips (logic/memory, radio, flash, I/O) you'll need to remove the failing chip, buy a replacement, install it, test it.

If the bill demanded that things be repairable, that'd be different. That would stop the vendors using modern tech, and sticking with old, The products would probably not even be sellable..

Experts: AI inventors' designs should be protected in law

BlokeInTejas

Re: Duh...

Nope; you can't "patent a program".

You can patent a "method for doing X", or "a means for doing X".

The method has to be novel and non-obvious and useful, as does the means.

You can encode the method in a program.

All current AI systems work basically the same way, and so the method (give data to this program, which has been constructed by a statistical inference method) is known. You might be able to patent different statistical methods for training.

But once you've got the machine, you give it some data, it churns a lot, and out comes whatever. You haven't contributed anything, except the data, and the program was constructed using known methods.

Very hard to see how you can patent its output. Elucidate, please.

BlokeInTejas

Duh...

A patent can be granted for something which is non-obvious and of value.

If the something can be repeatedly invented by running a given program over and over again, then the combination of that program and its input data are mechanical, and its output must be deemed obvious.

Can you patent a program? Nope.

Can you patent data? Nope.

Nothing left to patent, then, is there?

RISC-V needs more than an open architecture to compete

BlokeInTejas

At last, some RISC V Sanity

I think the article is a useful breath of fresh air.

RISC V has exactly one unfair competitive advantage - the hype surrounding it. Technically, it's yet another '80's RISC, perhaps with a bit more forethought given to instruction encodings, and without a delay slot. Ho-hum. Nothing wrong, but no unfair technical advantage whatsoever.

Recall that when it started, it was said to have been brought into existence to simplify the lives of computer architecture researchers. Prior to RISC V, such researchers had to cast about for an architecture which they could modify to explore and then show the fruit of their research. There really weren't many available - MIPS was an obvious choice because it was 'open', at least for a while.

But the problem for the researchers was that before they could get their brainchild to do useful work, they had to create an OS for it, and a compiler, and a library. And these had to be as good - more or less - as the existing ones for existing architectures. This meant that interesting research could spend only a small fraction of the research effort on the neat new stuff - they also had to do all this adaptation/redevelopment work on the software infrastructure.

The big thing that RISC V promised was that - if everybody based their research on RISC-V - then there were no architecture licensing issues - it was free and open as an architecture; and that the infrastructure work of compilers and OS and library etc would be done once, and everybody could share. Thus - higher quality research. And for that use, being "just a MIPS without a delay slot" was an advantage - adding neat new architecture to a vanilla old machine was just the playground one would like to show how much better your architecture improvements made things.

But now, folk are reckoning to get rich selling IP, or maybe SoCs, or maybe both based on designs with RISC-V cores inside them.

These products have no inherent advantage over (say) equivalent ARM-based products. Anything RISC-V in silicon will have exactly the same supply chain problems as an equivalent ARM-based SoC. Or a MIPS-based SoC. It's fab capacity that's in short supply. Not .pdfs

And there's nothing in the RISC V architecture which makes it better in silicon than an equivalent ARM. There's no magic.

So, yes; RISC V is a fine thing for academia. It's unencumbered. You can modify it if you insist. There's an OS port or two. llvm works. You can experiment using it in systems which choose to do things differently - where the fact that it's a RISC V is largely irrelevant, but having a free and unencumbered processor is highly convenient.

But for products? For profitable business?

As the article says, entirely unclear how RISC V leads to business success for RISCV IP and silicon vendors. Doesn't mean it won't or can't happen; just that its completely unclear.

An international incident or just some finger trouble at the console?

BlokeInTejas

An old colleague at a European computer company said that the language spoken at work was "European English".

It had, he said, many words in common with actual English.

And some of them even meant the same.

.. A useful insight.

UK watchdogs ask how they can better regulate algorithms

BlokeInTejas

Algorithms aint nuttin new...

Algorithms....

Well, they've been in use.. forever.

Been employed? The employer had an algorithm to sort out the likely candidates from the less likely, and one to decide when to stop sorting.

Ever had a pay raise? Algorithm.

Ever got fired? Algorithm.

Ever got stopped for speeding? Algorithm.

Ever been convicted of a crime? Algorithm.

Ever paid tax? How much you were supposed to pay was set by an algorithm.

And all those and more were algorithms implemented by people. And apparently make society systematically racist. Or inegalitarian. Or rich. Or something that a whiner doesn't like.

I don't think the idiots muttering about algorithms have the vaguest idea of what an algorithm is.

But in the end, suppose that we have two companies (or equivalent large scale entities). One tunes its algorithms to find and hire mostly black candidates. The other tunes its algorithms to find and hire the best they can find, using the measures of competence and all the correlated data they've found.

Who likely gets the best additions to its workforce?

And if you don't like the wording 'to find and hire mostly black candidates', change 'black' for whatever. If you're not using the best search possible, you'll end up at a disadvantage.

So algorithms act in a competitive marketplace. So they'll get better.

Which leads to all sorts of issues, all of which are the same as if it the algorithms 'running' on humans instead of giant piles of silicon.

Study: How Amazon uses Echo smart speaker conversations to target ads

BlokeInTejas

Are people seriously complaining that when you purchase something from a vendor, the vendor takes note and tries to sell you related items, or the same thing later on?

Or that if you talk to a vendor about some product, then you're surprised when they notice and try to see you the things you were discussing?

Seems a ridiculous complaint to me.

Goldman Sachs reportedly set to head up $60bn Arm IPO

BlokeInTejas

Re: Yikes

Opens up a host of fun opportunities, then, eh?

As long as you avoid designing yer own chips for 7nm and below.... That's expensive.

But hot damn, 28nm let quite a lot of neat stuff get built.

Out of beta and ready for data: 64-bit Raspberry Pi OS is here

BlokeInTejas

Re: A silly idea from a silly person...

The correct answer to your question is- spend the 100 bucks and see for yourself

Now, you can't actually do that right now because of limited supply, but it's still the right answer because only you know how many zillion tabs you want and what the speed of an elk falling through molasses should be.

Open source, closed wallets, big profits – nobody wins the OSS rock, paper, scissors game

BlokeInTejas

TANSTAAFL

There ain't no such thing as a free lunch.

If you as an individual or a corporation choose to build your business on software that someone else wrote and offered with no warranty, no way to get fixes within a guaranteed time, and which you personally didn't test to within an inch of its life to the same standards as internally-developed stuff - then you're an idiot.

If you want stuff you didn't pay for to be reviewed for errors, but you're not prepared to do it or pay for it to be done - then you're an idiot.

What may fix the whole and complete nonsense of "yeah, it's software, but it comes with no warranty of correctness whatsoever" is somebody who embodies software in their product getting sued to within an inch of their lives - or beyond - for supplying stuff that didn't do the job it was clearly supposed to do.

I'd guess that this will eventually come about in the automotive space, because cars can kill people.

If there were a real threat of serious financial damage being done to you when you ship product embodying flawed software and said software fails, then perhaps things would change.

Until then, it really and truly doesn't matter. Stuff that matters has financial implications.

Please, no Moore: 'Law' that defined how chips have been made for decades has run itself into a cul-de-sac

BlokeInTejas

About time too

Moore's Law has been the ugly luxury that has lead to the death of computer architecture. What was the point producing something which was 10x better than an x86 when simply waiting three to four years gave you the same performance without having to rebuild toolchains/recompile/redesign/rewrite?

But now we live in interesting times. With a bit of luck, two things will happen:

- mainstream computing will realize that even a 1-time massive performance/power gain through improved architecture is a big win

- folk who think that the way you should classify things seen by humans is by using tera-operations per second to build unreliable "AI" systems will realize that's nuts, and find a much better way of doing it.

This may have an effect on how software's written, too

UK artists seek 'luvvie levy' on new gadgets to make up for all the media that consumers access online

BlokeInTejas

It's basically a demand that the gummint pay them for failing at their hobby (it must be hobby, because if it were work they'd be getting paid).

What we need is lions. Fierce, hungry lions. That'd fix this particular problem, and might work well in other areas, too.

BlokeInTejas

Citation needed.

Not only were half of an AI text adventure generator's sessions NSFW but some involved depictions of sex with children

BlokeInTejas

Ummm....

Dear El Reg, hy do you continue to refer to this garbage as "AI"?

It's just adaptive pattern matching.

Intel to finally scatter remaining ashes of Itanium to the wind in 2021: Final call for doomed server CPU line

BlokeInTejas

If Intel had been sure that Itanium was the architecture of the future - and it could have been: done right the silicon should have been simpler to design, simpler to verify and simpler to test, so products would come to market sooner and more cheaply, and the silicon would indeed suck less power and be cheaper - then they'd have had the courage of their convictions and launched it into the important market first- PCs and laptops. This would mean convincing some major PC makers up front, and making sure that the new machines could run x86 binaries. Then they could have tweaked one more thing, and persuaded Microsoft to distribute PC binaries in an architecture-independent form, with the final compilation to the actual computer being done at app installation time.

On the assumption that the architecture really did have commercially-useful benefits in implementation, this would have let AMD continue to make x86s, but lag further and further behind, and set up a fun universe where the upstart ARM could find a place in PCs (purely on merit) and oh, yeah, the server market was now a continuous part - as far as software was concerned - of the extremely high volume PC market. My, they might even have got into cellphones.

But, there y'go...

Another data-leaking Spectre CPU flaw among Intel's dirty dozen of security bug alerts today

BlokeInTejas

Re: You make your bed ....

Intel did not invent speculative execution. Almost certainly, IBM did.

Check your history. Google 'Tomasulu'.

Twerps.

Plans for half of Europeans to get 100Mbps by 2020 ain't gonna happen – report

BlokeInTejas

Rural little village in Normandy; too far from anywhere to get anything even approximating bad ADSL. Even the pigeons avoid the place...

So the solution is to buy a 4G/LTE box, which offers entirely acceptable up and down speeds (I'm not there at the moment, so I can't measure).

But the downside is - a budget of only 20GB per month. That goes very quickly, even though we don't stream music or movies. One software update to one phone once and you eat 10% of that. Ouch.

Apple store besieged by protesters in Paris 'die-in' over tax avoidance

BlokeInTejas

Idiots

If these bien-non-pensants have a beef with taxes paid by corporations in France, they should go and harass their own government, which is the entity that sets the tax paying rules in place.

Oh, they're French. Yep. No idea how the real world works (lookout the current SNCF strikes...)

Intel outside: Apple 'prepping' non-Chipzilla Macs by 2020 (stop us if you're having deja vu)

BlokeInTejas

No need for virtual machines....

Moving from x86 to ARM within a controlled environment like MacOS/OS X can be much simpler than deciding to have to run emulators/jit compilers and the like. Instead, long before the processor architecture swap is put in place, you change the form of the 'binaries' that the computer will accept and install.

And you change that to something equivalent in form and function to the llvm compiler's internal representation. And when you install that binary, you perform the last pass of compilation to whatever processor architecture your computer has installed. And everything runs at native speeds.

This can even be back ported to the appropriate last few OS releases; it will leave those who insist on staying the past somewhat high and dry, though.

London mayor: Self-driving cars? Not without jacked-up taxes, you don't!

BlokeInTejas

Suppose for a moment that we should tax the use of roads; then the solution is simple. The best indicator of actual road usage (including damage done etc) is the rate at which tyres are replaced on a vehicle. So slap an added purchase tax on tyres.

Tesla to charge for road trip 'leccy, promises it will cost less than petrol

BlokeInTejas

Taxing times indeed

Seems t'me that the simplest replacement for a petrol tax is to tax tires. They wear out, you have to replace them. And they wear out more when you go further. Or faster. Perfect.

Microsoft tries, fails to crush 'gender bias' lawsuit brought by its own women engineers

BlokeInTejas

I reckon, from observation, that company assessments are cruelly flawed. They have no reliable metrics. They have no reliable method of knowing the metrics were properly applied. They have no metrics to decide whether a pay rise of X should be given to A rather than B. They clearly have no metrics for measuring managers against baseline competence, nor one external candidate from another.

Personally, if I have a good view of an individual's work, I can generally rank them as "yeah, more or less ok", "bloody magic" or "bloody useless". Since there aren't generally many magic or useless folk, this ranking doesn't do much for fine gradations between employees. And, of course, I don't have good knowledge of most people's performance, so much of any ranking has to be the fourth, important, category "dunno".

Not at all clear that managers have, in general, even this degree of insight into people's performance.

oh well.

Zilog reveals very, very distant heir to the Z80 empire

BlokeInTejas

Z8000's, I'd wager, not Z80's.

Zilogs answer to the challenge of proper 16 bit microprocessors.

https://en.wikipedia.org/wiki/Zilog_Z8000

I well remember the days at ICL when all the 16 bit microprocessor vendors came trooping through Kidsgrove trying to sell us the magic things. Even TI, with the 99000, explaining that it were bloody good because it was software binary compatible with the 990 (??) minicomputer...

Wannabe Prime Minister Andrea Leadsom thinks all websites should be rated – just like movies

BlokeInTejas

Re: Stupid thinking similar to Donald Drump

"barmy", not "balmy"

Unless you're referencing the Mighty Steve

Surface Book nightmare: Microsoft won't fix 'Sleep of Death' bug

BlokeInTejas

Re: Sleep and Hibernate have always been iffy

Exactly what I was going to say. My MacBook Pro just stops doing things when I close its lid, and resumes when I open it, even days later. It's how it should work

Obama puts down his encrypted phone long enough to tell us: Knock it off with the encryption

BlokeInTejas

"If there is probable cause to think that you have abducted a child, or that you are engaging in a terrorist plot, or you are guilty of some serious crime, law enforcement can appear at your doorstep and say 'I have a warrant' and go into your bedroom to rifle through your underwear and see if there's any evidence of wrongdoing."

I thought that just as the FBI (or whoever) could get a warrant to look through your underwear, you could be required to unlock your phone so they could look therein.

That's NOT the same as saying they need access to everything just in case..

BlokeInTejas

Conspiracy Theory

Ah, yes, the simplistic idea that there's a small cadre of Those Who Control on whom we can blame everything.

It used to be evil secret societies from the deep past, or the Jews, or ... But now it's "the one percent".

last time I looked, none of the rich (except the vile Pelosi, perhaps) actually passed laws. So you're actually claiming, to the extent that your claims have any merit, that the elected representatives in all Western societies are corrupt.

Maybe that's a root cause to go worry about, rather than pointing at some group and saying "they're bribing".

Bruce Schneier: We're sleepwalking towards digital disaster and are too dumb to stop

BlokeInTejas

Sigh

"But software isn't like that. Instead you code fast and hard and then fix things when problems crop up. "

Nope; that's not basic; that's just what people do today.

Wait for the first billion dollar/billon euro successful lawsuit which guts somebody quite large (have to be large to pay that) for having defective software. Car manufacturers come to mind.

Habits will change when there's money at risk, not before.

SUBSCRIBE TO OUR WEEKLY TECH NEWSLETTER

Biting the hand that feeds IT © 1998–2022