That's a little bit overstated. For Linux, the changes are a couple thousand lines of code. That's a number I took from their research papers.
But what do you have to change ? First of all, the system has to be made aware that it's not the "top top". Its physical memory is no longer contiguous (you ask Xen for memory pages, and it gives them to you in arbitrary places), it also has to be aware of absolute time that's no longer tick++. Second, you need drivers for the abstract network card and disk. Those are generally easy to write, because you just delegate the real work to Xen. VMWare is already doing something similar with its vmxnet driver for Windoze.
I'd really expect these kind of changes to the OS to be incorporated in the main linux tree, as they mature.
What do you gain from all this ? Well, SPEED. I mean, SPEED. Take a look at their research papers (wrong suggestion for the "I won't RTFA" crowd, but still...). Their slowdowns/throughput losses (they run Postgres and Apache on a couple of virtual nodes, as opposed to a single, consolidated machine), are negligible (less than 10%). On some configurations they even got performance improvements! At the same time, VMWare and UML do considerably worse.
In general, it's very easy to "virtualize" stuff that's running mostly in user space. As soon as you have considerable OS+I/O overhead, your performance drops significantly. The para-virtualization approach (employed by Xen), pretty much gets you the best of both worlds.
Dell has tried at some point to work with AMD, but AMD just didn't deliver (b/c of fabrication problems). In the end they didn't get a better deal, and pissed Intel at the same time.
Having a better product from an architectural standpoint doesn't mean much if you can't mass-produce it. With their new plants they're doing better, but the memory of their past failures is still fresh.
IANAP (P=psychiatrist, psychologist), but here's my theory: games are very often a way for one to consume his/her domination instinct. Clearly, it's a surrogate, there's more satisfaction from dominating real people in real life, but it's still better than nothing. One way of consuming your domination instinct is just being very evil.
To conclude, I really don't think you're evil. Mr. pointy-hairy doesn't need to play computer games, he is the Satan of the office.
It was actually much less than that (probably comparable to a Z80). 486 would have been PFM (pure magik) for those times.
FYI: even today, you can't send more than a 486 in orbit, mostly because of feature sizes. The smaller the feature size, the easier for cosmic radiation to screw things up.
They also release a lot of volcanic ash in the atmosphere, which has the opposite effect. The largest observed eruptions, Agung in 1963, and Pinatubo in 1991, have in fact cooled down the earth by 1 or 2 degrees.
Cars do kill the environment, despite what dubious, paid-by oil companies, "scientists" say.
Check
this out. It's written by the same people who claimed that Linux was not written by Linux (and that it's Minix instead).
The problem with HSQL is that it lives completely in memory; consequently it just can't possibly meet the Durability requirement from ACID (IOW, once you committed a transaction it's not the case that it's on permanent storage). Disk is one of the biggest bottlenecks with databases, so I would expect HSQL to actually be significantly faster with update transactions.
Most importantly, it can address up 2^64 bytes of memory. And yes, that generally implies 64-bit integer GPRS. BTW, vector operations on x86 (MMX) also operate with 64 bit registers, but it can only access 4G (32G if you use the extended bits "hack").
With the second official Sharp ROM, it became impossible to synchronize Linux with Zaurus... (I'm not talking about home-grown scripts, I'm talking about the full thing). It became in fact easier to synchronize WinCE devices than Zaurus.
My impression was that they thought the open source community would take care of everything for them as far as software is concerned, and they wouldn't have to lift a finger...
In the end the only player left in the market was theKompany.
People like having a number that allows for quick comparisons between two products. Unfortunately, frequency is only one part of the equation.
performance = Frequency X Instructions per Cycle
Granted, if you keep the IPC the same, increasing frequency will help linearly. The problems with keeping IPC are: 1. complexity, which negatively impacts clock speed. 2. memory, which increases at a much slower pace (so if you miss out of the L2 a lot, increasing CPU frequency won't do much)
So, Intel said - people can't comprehend IPC, let's just ramp up the frequency. Their processors work ok for applications which are not memory intensive and don't have many branch mispredictions. AMD said - we can get better performance by improving the IPC (e.g. have a considerably better design for the memory system than Intel's). AMD processors would thus crush Intel processors working at the same frequency, but people don't really see that, because there's no aggregate performance number associated with the processor.
I personally like AMD's approach better, because higher frequency means higher power.
Yes, there are things you can't do in SQL. Most mentioned example is transitive closure.
As opposed to regular programming languages, SQL is not Turing-complete, meaning that there are computations that can be expressed in a Turing-machine language, but not in SQL
Turing completeness does come for a price though - it's undecidable whether the execution of the program will even finish (instead of infinitely looping). SQL OTOH is guaranteed to complete in poly time of the input dataset. Different tasks - different hammers
We know quite well that Mao murdered 30 million "counter-revolutionaries". It's true that modern-day China is no longer the one of the "Cultural Revolution", but still, it's a similar regime (not necessarily from a economics standpoint, but from a totalitarian/human-rights one; Tienanmen anybody ?).
OTOH what Google is doing is very legitimate. They're doing whatever they can to prevent themselves from being firewalled-out while offering a decent service to Chinese users. It should be, anyway, the responsibility of the US government to at the very least condemn censorship in China if not impose penalties on companies that help China censor.
On a totally different note: do you think press is free (as in uncensored) in the States ? Read in parallel dailynews.yahoo.com and bbcnews.com for a month.
Sure, in theory getting rid of a dictator like Saddam is a good idea. In practice, the moment the coallition troops get out of Iraq the country will quickly descend into civil war, and it might as well disintegrate.
The idea that you can just impose democracy in a country and everybody will be ecstatic about it is simply idiotic. The citizens of Iraq, btw, consider themselves to be under occupation, and not liberated (see the response of the Iraqi Olympic Team Leader for instance).
is pretty obvious. Easy tampering with election results, in a way that's very inconspicuous and less susceptible to suspicion (people believe it's much safer than paper). It doesn't even have to be similar to the communist-ran elections in the late 40s/early 50s - when the communist parties were made victors without even passing the electoral threshold - these days you can influence the outcome of an election by mingling with one-digit percentages (which are generally within the error margins of polls).
The reason Solaris does so well on many processors is because Sun's strategy regarding "big iron" was "ultraslows, but many". It's very true that their target workloads (server-based) can exploit multiprocessing easily. Having good support for multiprocessors was simply crucial for the platform.
In the x86 world things are quite different. Having been a desktop-oriented architecture for a long time, the main x86 chips (Opteron/Pentium IV) are pretty much the best these days at executing single-threaded stuff (see spec.org if you don't believe me). Multiprocessing was more of an "after-thought" than an initial requirement. Consequently, you can easily get 4-way SMPs for x86s, but not more than that (Sun AFAIK scales considerably better).
This reflects on x86 OSes as well. There's not that much need to do well on more than 8 execution contexts (4way SMP x2 - hyperthreading), and consequently having an operating system that scales better won't have that much of an impact on x86. Sure, in the "big iron" category things will be different, but not for the dominant architecture
The two main power "consumers" are the processor and LCD (they generally account for ~90% of a laptop's power consumption). Strangely enough, the hard drive, altough mechanical and spinning very fast, is not nearly as bad.
Most "features" of a laptop don't really consume extra power if not utilized. For instance the DVD drive only consumes power if it's actually spinning (and mencoder can can take care of that).
To reduce the power consumption of the cpu simply put it to the lowest frequency (speedstep). 600MHz is generally enough to play a movie (DVD or.avi). As far as the LCD screen is concerned - simply reduce its brightness.
I'd personally recommend the Centrino processor line - good perfomance at reasonable power levels (as opposed to Pentium 4 Mobile).
Slashdot Mirror
But what do you have to change ? First of all, the system has to be made aware that it's not the "top top". Its physical memory is no longer contiguous (you ask Xen for memory pages, and it gives them to you in arbitrary places), it also has to be aware of absolute time that's no longer tick++. Second, you need drivers for the abstract network card and disk. Those are generally easy to write, because you just delegate the real work to Xen. VMWare is already doing something similar with its vmxnet driver for Windoze.
I'd really expect these kind of changes to the OS to be incorporated in the main linux tree, as they mature.
What do you gain from all this ? Well, SPEED. I mean, SPEED. Take a look at their research papers (wrong suggestion for the "I won't RTFA" crowd, but still ...). Their slowdowns/throughput losses (they run Postgres and Apache on a couple of virtual nodes, as opposed to a single, consolidated machine), are negligible (less than 10%). On some configurations they even got performance improvements! At the same time, VMWare and UML do considerably worse.
In general, it's very easy to "virtualize" stuff that's running mostly in user space. As soon as you have considerable OS+I/O overhead, your performance drops significantly. The para-virtualization approach (employed by Xen), pretty much gets you the best of both worlds.
You only have to store it for the duration of your office (4-8-whatever years). After that, it becomes Someone Else's Problem.
This proposal will have the opposite effect
Yeah, Europe and Canada are nice ... :)
Does "thou shalt not steal" ring any bell :) ?
Having a better product from an architectural standpoint doesn't mean much if you can't mass-produce it. With their new plants they're doing better, but the memory of their past failures is still fresh.
To conclude, I really don't think you're evil. Mr. pointy-hairy doesn't need to play computer games, he is the Satan of the office.
FYI: even today, you can't send more than a 486 in orbit, mostly because of feature sizes. The smaller the feature size, the easier for cosmic radiation to screw things up.
Cars do kill the environment, despite what dubious, paid-by oil companies, "scientists" say. Check this out. It's written by the same people who claimed that Linux was not written by Linux (and that it's Minix instead).
So, no, the comparison isn't fair at all.
Most importantly, it can address up 2^64 bytes of memory. And yes, that generally implies 64-bit integer GPRS. BTW, vector operations on x86 (MMX) also operate with 64 bit registers, but it can only access 4G (32G if you use the extended bits "hack").
My impression was that they thought the open source community would take care of everything for them as far as software is concerned, and they wouldn't have to lift a finger ...
In the end the only player left in the market was theKompany.
What would you do if you needed heart surgery ?
performance = Frequency X Instructions per Cycle
Granted, if you keep the IPC the same, increasing frequency will help linearly. The problems with keeping IPC are: 1. complexity, which negatively impacts clock speed. 2. memory, which increases at a much slower pace (so if you miss out of the L2 a lot, increasing CPU frequency won't do much)
So, Intel said - people can't comprehend IPC, let's just ramp up the frequency. Their processors work ok for applications which are not memory intensive and don't have many branch mispredictions. AMD said - we can get better performance by improving the IPC (e.g. have a considerably better design for the memory system than Intel's). AMD processors would thus crush Intel processors working at the same frequency, but people don't really see that, because there's no aggregate performance number associated with the processor.
I personally like AMD's approach better, because higher frequency means higher power.
As opposed to regular programming languages, SQL is not Turing-complete, meaning that there are computations that can be expressed in a Turing-machine language, but not in SQL
Turing completeness does come for a price though - it's undecidable whether the execution of the program will even finish (instead of infinitely looping). SQL OTOH is guaranteed to complete in poly time of the input dataset. Different tasks - different hammers
Need I say more :) ?
... with HOWTOs, trubleshooting and changelogs, and, um ... hardware upgrades!
Not to mention living costs in London.
People are far less idealistic than you think
Need I say more?
OTOH what Google is doing is very legitimate. They're doing whatever they can to prevent themselves from being firewalled-out while offering a decent service to Chinese users. It should be, anyway, the responsibility of the US government to at the very least condemn censorship in China if not impose penalties on companies that help China censor.
On a totally different note: do you think press is free (as in uncensored) in the States ? Read in parallel dailynews.yahoo.com and bbcnews.com for a month.
The idea that you can just impose democracy in a country and everybody will be ecstatic about it is simply idiotic. The citizens of Iraq, btw, consider themselves to be under occupation, and not liberated (see the response of the Iraqi Olympic Team Leader for instance).
is pretty obvious. Easy tampering with election results, in a way that's very inconspicuous and less susceptible to suspicion (people believe it's much safer than paper). It doesn't even have to be similar to the communist-ran elections in the late 40s/early 50s - when the communist parties were made victors without even passing the electoral threshold - these days you can influence the outcome of an election by mingling with one-digit percentages (which are generally within the error margins of polls).
It's not a P4 Mobile, it's a P3 spin-off. That's why a Centrino processor beats a P4 if running at the same frequency.
In the x86 world things are quite different. Having been a desktop-oriented architecture for a long time, the main x86 chips (Opteron/Pentium IV) are pretty much the best these days at executing single-threaded stuff (see spec.org if you don't believe me). Multiprocessing was more of an "after-thought" than an initial requirement. Consequently, you can easily get 4-way SMPs for x86s, but not more than that (Sun AFAIK scales considerably better).
This reflects on x86 OSes as well. There's not that much need to do well on more than 8 execution contexts (4way SMP x2 - hyperthreading), and consequently having an operating system that scales better won't have that much of an impact on x86. Sure, in the "big iron" category things will be different, but not for the dominant architecture
Most "features" of a laptop don't really consume extra power if not utilized. For instance the DVD drive only consumes power if it's actually spinning (and mencoder can can take care of that).
To reduce the power consumption of the cpu simply put it to the lowest frequency (speedstep). 600MHz is generally enough to play a movie (DVD or .avi). As far as the LCD screen is concerned - simply reduce its brightness.
I'd personally recommend the Centrino processor line - good perfomance at reasonable power levels (as opposed to Pentium 4 Mobile).