Latest Loongson chip is another step in China's long road to semiconductor freedom China is slowly achieving its long-term goal of semiconductor self-sufficiency – homegrown chips for its computing devices, in other words. The Loongson processor family, developed in the Middle Kingdom, frees China from the hegemony of foreign chip makers, and gives the country more control of its information-technology …
Well, CentOS is a fork of RHEL (well sort of... I know the CentOS/RHEL relationship is incestuous enough it probably shouldn't quite be called a true fork)... and Loongnix is a new fork of that hot mess.
Given its country of origin, shouldn't that be a spork (as in, Sino-Fork)?
Where did that come from ?
Loognson's declarations are obviously written with a government and party representative. Every declaration must contain some form of jab at The West, even if it is totally groundless.
And as for the "independant analysis" of the architecture, I'm sure it was made by a party member as well.
It would seem that China's version of Pravda is working up a sweat.
The US is restricting technology exports to China, so I'd say "CPU architecture as a means of control" is a completely accurate statement. Nothing groundless about it.
@Androgynous Cupboard
OTOH by restricting sales of Intel processors, the US probably did China a massive favour. If you were starting a proc design now, knowing what can be achieved, you wouldn't start with x86. It was only the popularity of the PC™ that killed off many better designs such as the 68000 series. Thank god that ARM and MIPS stuck with the RISC strategy and found a business model that Intel couldn't kill off.
> That is a purely political decision
It's China, so everything is a political and party decision.
I don't quite follow one of the propaganda talking points put forth in the article: since these chips are - apparently - based on MIPS and RISC-V designs, how exactly is China breaking its dependency on US technology pilfering imports?
Both MIPS and RISC-V are American designs. Hennessy (MIPS) and Patterson (SPARC and RISC-V).
And now, let's go back to our objective analysis of them Anglo-American Imperialist Pigs.
That would be a valid point if we were talking about China's decision, but we weren't. That was the US's decision.
There is no issue with being based on other people's architectures. Everything is built on what came before, that's how technology works. The point is that there's nothing that needs to be licensed, so nothing that can be restricted by the whim of some foreign government.
Of course that's of no value to anyone outside China (except possibly in places like Iran that are subject to similar restrictions), but from their own point of view it's still worth crowing about. It's sending a message to the US in particular, that this is one less lever it has over China.
> That was the US's decision.
What was the US' decision? That China is using MIPS and RISC-V designs in their so-called homegrown chips? That's bullshit. They are using MIPS and RISC-V designs because (a) they aren't capable of designing a high-performance chip on their own [1] and (b) they got the designs for free from Github.
It's sending a message to the US in particular, that this is one less lever it has over China.
I don't think you realize how stupid and delusional that statement really is. Do you really think we give a shit about China's 15-years old MIPS chip?
The fact is, neither China nor Iran can get their hands on the chips they really want, in sufficient quantities to use them for what they really want to use them for. They might be able to sneak in a few Xeons through intel ops. But they do not have the ability - or capability - to reverse-engineer the designs, or the fab processes, and they are constantly playing 15-year catch-up.
That's the whole point. If you don't like the fact that we (US) don't feel like sharing this technology with some of the worst political regimes in the world, too bad. We do not owe it to the world to share our competitive advantage with our enemies.
In fact, give me one example of a military power - any military power - that has voluntarily shared military technology with the enemy. Which is precisely what you, and your fellow ideology commentards are pushing for and complaining about.
The part that I'm still unclear about is: why are you even trying? You cannot possibly be that stupid. That being the case, the only other possible explanation is that someone is paying you to say it.
-----
[1] The Chinese Communist Party hasn't come up with an ideologically correct chip design at its latest Central Committee Plenary Meeting. They are waiting for Winnie The Xi's input.
All RISC processors follow a straightforward three address instruction format. The differences between different manufacturers' implementation comes down to details like the number of registers, implementation of special registers (processor and cache control etc.) and 'special sauce' capabilities such as the skip on condition register bits you get with ARM. (I've used most RISC types in 'live' projects over the last two or three decades so I'm used to their idiosyncrasies -- and that includes the AMD 29K series.) What the Chinese have done is not ideological, its common sense. They will have taken an existing Harvard architecture (e/g/ MIPS), adapting and extending it as necessary, but once the RISC-V initiative got traction they would have converged on that. There's no magic, no politics, nothing -- this is exactly what I'd do if I were in a similar situation (and if i'd do it's certain that just about anyone else in the business would do so as well).
Its worth remembering that instruction sets like Intel's are emulated. The internal architecture of an Intel processor and its microcode are closely guarded trade secrets which is why most people aren't aware of this internal layer. But when you think about it from a logic design perspective this is pretty much the only way to design a CISC processor -- it makes absolutely no sense to try to implement it using pure logic (try it!).
The real action isn't in the processor anyway. Its in the cache / memory interface design, dynamic register assignment and renaming and multiprocessor interconnection.
As for the political angle, I think this attitude is riding for a fall which is why I've been very critical of it in the Reg. We're not good at long term investment in the US -- look at what's going on in DC to see how we can't (to use an English saying) "organize a piss-up in a brewery" there. We're living in the past, our politicians play on nostalgia (we've been doing it since Reagan, maybe before). Unless we change our attitude we're going to find that people like the Chinese will wipe the floor with us leaving us vulnerable to populist politicians who's only option is to throw our military weight about.
Well analysed and said: Personally I fully agree.
And what we see here in Europe is that most programmers in the West are lazy b*stards (I am sorry). Cut and paste policy without thinking. Glueing languages to another without questioning why, and if it is efficient at all... from a code execution point of view. What happens if code efficiency comes in? What happens in China if they open up a can of good Chinese programmers to really *optimize* things?. A good hardware engineer designs something for £2. A better one does it for £1. But what about the code? What hardware does it need? I see this killing the earlier mentioned engineering principle time and time again.
We have more than enough very well performing hardware. There is a real overspill, often with very bad worked out and supported tools, however. And peoplw who do not believe this, they should watch this video to convince themselves: The Rougol interview / presentation with Sophie Wilson:
https://www.youtube.com/watch?v=6lOnpQgn-9s
My conclusion: In the West we need good programmers, not new hardware. In the East they have all parameters set to overrule us in many ways. Very scary....
It was the US's decision to refuse exports of high performance Intel chips to China. Seriously, when someone uses a demonstrative pronoun, you should take a moment to read upthread and work out what they're talking about.
To the rest of your rant - sure, keep telling yourself that. Whatever.
> It was the US's decision to refuse exports of high performance Intel chips to China.
Yes, it was, and our export controls aren't limited to high performance Intel chips. UltraSPARC's and later are on the list too. Good for us. We're not as stupid as you and China might want us to be.
> To the rest of your rant - sure, keep telling yourself that. Whatever.
Awesome retort, bro.
"The US is restricting technology exports to China, so I'd say "CPU architecture as a means of control" is a completely accurate statement."
And how did you get from accurate clause 1 to proving opinion clause 2? Because they're really not all that related. They're not related for a few reasons. First, if the U.S.'s architectures are evil, then what's so wrong with ARM, designed in the UK (and a lot of other places)? Chinese manufacturers have been designing and using ARM chips for years without the U.S. controlling them.
But that's assuming there's something wrong with X86, which there isn't. The U.S. has been denying various technology exports, but chips are not among them. China can buy all the X86 chips they want. The export restrictions have not at all slowed the Chinese implementation of X86, most notably the efforts of Zhaoxin. Other companies can also do that. The U.S.'s export controls have had an effect on China's chips, but it's been in getting chips manufactured, not in their design. That issue is just as limiting for a Chinese-built architecture as any other.
The existing ISAs are basically open. While there is a technical license requirement for X86 and ARM, it can be violated without much difficulty; nothing blocked the use of MIPS in this architecture, after all. It's as if China tried to develop their own programming language because using C was too American; they can do it if they want, and it promotes competition so there's a mild benefit there, but their reasoning doesn't make sense.
if the U.S.'s architectures are evil
are you really that clueless ? Are you still living in a Bond-type villains era ? How could, even from a purely theoretical perspective, any "CPU architecture" be evil as a such ?
1) it's not about US architectures but about US politics
2) it's not about evil, it's about practical: if the US politics use some means for extortion – or political pressure – then the practical approach is to get rid of that means and replace it with some other means.
It's context: evil was the shorthand for being risky to use. It isn't risky to use for the reasons I explained in my post. Your idea that the U.S. might use it for extortion doesn't make any sense; if they deny exports of chips, then all China needs to do is start using their own manufactured chips (or those manufactured in countries other than the U.S.). Which ISA is used is of no importance in it. Encouraging local chip production because the U.S. could make chips more expensive is at least a logical plan which they have done. Invent a new ISA because somehow the U.S. will restrict them from understanding or implementing X86 is poor reasoning.
'Every declaration must contain some form of jab at The West, even if it is totally groundless.
And as for the "independant analysis" of the architecture, I'm sure it was made by a party member as well.'
Hmmm. I wonder whether they copied that approach from The West. (Well, parts of it).
Of course CPU architectures are a means of control. That much has been obvious at least since the Intel/ARM wars of the 80s.
The question is how this architecture will be any different. Seeing that it's clearly not even fully documented, let alone open source, my guess is that it won't. It's better for China, and only China - not for any other country, and not necessarily even for any given Chinese company or customer. For everyone else, it's just another option to be evaluated on its merits, whatever they are.
This post has been deleted by its author
Definitely the CCP. All day the Chinese govt tracks what I read and watch, monitors what I buy online and adjusts prices to make me spend more. They also show political messages to my fellow voters to influence the choice of govt.
If only chairman Pooh wasn't controlling all my internet things would be a lot saner here.
Err, can they? I suspect with this line of new chips, Chairman Xi will be baked into their silicon.
And Personally I'm quite happy to have permanently abstained from reading politically incorrect rants while avoiding being tracked, modelled and manipulated by the meta-twat.
Plastic trash is a more troublesome addiction but i'm working on it.
Given that when I do a Google translation of the web page in question I can't find any reference to the word "dictator" I suspect it's either the result of a dodgy machine translation or else El Reg are making an attempt at humour.
I found their reference to the "sexual development" of chips as being even more curious, although I suppose I should be open to the possibility that this is also just a dodgy machine translation.
A new journey from the "realm of necessity" of sexual development to the "realm of freedom" of independent development.
Given that they headline this chip as being a "Heavy Release", I would say that some things seem to not translate well if taken literally.
Supposedly the software would run in an emulator which translated the binary representation dynamically. I don't know if it actually works well enough to run something like Windows reliably however.
Apple did some sort of binary translation during one of their processor switches, and I believe that the same was done with Itanium as well.
These days its questionable as to how useful the idea really is given the rapid decline in the importance of the Windows PC.
> LoongArch
Can we all agree on a western pronunciation?
Do we go with "Long" or "Loon" +arch?
Colour me unconcerned. There was a lot of sabre rattling about the previous versions, which were generations behind and went nowhere as well. As long as the US can scare fabs with sanctions if they work with China, their special Arch-itectures won't even place. Once China shows they can build a cutting-edge fab (instead of just claiming Taiwan's accomplishments for its own) then things get interesting.
"Once China shows they can build a cutting-edge fab (instead of just claiming Taiwan's accomplishments for its own) then things get interesting."
Don't hold your breathe on the cutting edge fab from China - they are at least 10 years away from current lithography and I would be surprised if they get close to ASML equipment in that timeframe.
Of course they could always buy ASML equipment but second/third hand is much cheaper and you avoid the TSMC/Intel/Samsung queue.
China hasn't committed to a multi-billion dollar fab yet - they've announced them but only built test lines while pulling back from initial investment promises and likely struggle with quality from test lines versus what they can get from international sources. It's not easy to copy bleeding edge or even 5 year old processes...
Don't hold your breathe on the cutting edge fab from China - they are at least 10 years away from current lithography and I would be surprised if they get close to ASML equipment in that timeframe.
Just knowing that something like EUV-lithography is actually possible to achieve and being delivered today is a big step on the road to developing a knockoff that matches or at least closely approaches that capability. There is definitely "secret sauce" in ASML's boxes somewhere since the People's Number Five Rubber Novelty Products Factory down by the docks in Shenzen isn't spinning out 7nm wafers today but it's not White Man's Magic, it's optics and physics. Espionage may also play a part, of course.
"There is definitely "secret sauce" in ASML's boxes somewhere since the People's Number Five Rubber Novelty Products Factory down by the docks in Shenzen isn't spinning out 7nm wafers today but it's not White Man's Magic, it's optics and physics. Espionage may also play a part, of course."
And if the magic is "lots and lots of money that puts off existing competitors and newcomers alike"?
If things go wrong, no one wants to explain to the CCP why their $100bn 5+ year investment is a failure.
And they certainly don't get the opportunity to repeat the mistake.
Having said that, People's Number Five Rubber Novelty Products Factory may find making 16nm with third hand tools is actually quite profitable and squeeze a lot more competitors out of business with "good enough" ARM/memory/flash for mass market devices.
Or...
They know getting other platform's that perform better and are better optimised for easily available OS's is not hard and they don't have to use Chinese made CPU's/OS's. Yet.
It would be interesting to compare performance between these devices and RISC-V chips where development effort isn't primarily driven by major hardware developers that get significant revenue from CPU sales (i.e. ARM/Apple/Intel/IBM/AMD/nVidia). I suspect RISC-V is faster with less investment. Or that investment in Loongson is greatly overstated.
i.e. this s theatre and anyone in China wanting to do real work ignores them or can get exemptions for any requirement to use these systems.
This is a distraction from the Taiwan Fabrication Labs / Invasion threat.
Taiwan has the biggest collection of high end chip fab labs anywhere in the far east, or possible, the world. This is a market that china wants to dominate, so badly, you could almost taste it.
Like this one:- https://en.wikipedia.org/wiki/TSMC
This is where Apple CPU chips are made.
China does not want a slice of the action. It wants it all by' acquiring' the labs by any means, fair or foul.
If China invaded Taiwan, I suspect TSMC would be sitting idle within weeks without Korean and Japanese chemical supplies. And even if they used low quality chemicals, the current fabs would be dead within months without replacement US/European parts.
Here's an interesting (if rather log) discussion of the semiconductor supply chain - China make up around 1% of the overall supply chain and it is primarily in testing/packaging. In all key areas of production, China have no expertise in key manufacturing areas:
https://cset.georgetown.edu/wp-content/uploads/The-Semiconductor-Supply-Chain-Issue-Brief.pdf
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
