How can any country justify having CEO's that make $10,000,000+ a year, and having janitors who make $5.75 an hour?
Hi, welcome to America. I'll be giving you a brief synopsis of how a free market works.
In a free and open job market, you have people and you have jobs. Many people have the ability to do more than one kind of job. Likewise, most jobs could potentially be filed by one of a relatively large set of people.
Some jobs are able to be done by lots of people, like being a cook in a fast food restaraunt. Some people are willing to take that job because they do not have a skill set that enables them to work a different job, or maybe because they like it. Anyway, the worker and the business agree on compensation for the job that is agreeable to both the business and the worker. If there is someone willing to do the same job for less, then the worker should get paied less. If there is nobody that can do as good of a job for the same price, the worker can ask for more money and may get it.
For a job like a CEO or Baseball player, there is incredible demand to get the absolute best people for those positions. If you are the best baseball player, you can demand a lot of money, because there is demand to have the absolute best player on your baseball team: it would help you win, wins help you sell more tickets and advertisements, and people want their team to win. If your team is a perpetual loser, you won't sell as many tickets. (Unless you are the Cubs, because of Cubs Fans like Me).
Likewise, Good CEOs can demand a premium. Someone who can manage a business effectively has a skillset that is in high demand. If you can make your company that earns 1 Billion / year, and grow that profit by 20%, you are worth a lot of money. Stockholders are willing to pay these high salaries because there is the perception that a good CEO can lead a company to be more profitable.
Of course, this is all based on perception. If killing the R&D saves money now, but you lose out on the next major innovation, you've screwed yourself. But that's how the market works, and your competition (who did the right R&D) will beat you. As they should.
Now you may notice that America isn't exactly a free market. Even when Southwest airlines is making money, the Government gives billions of dollars to airlines who didn't make good business decisions. Unfortunately, the Government has, over the last several decades, gotten more involved with business. This leads to situations where bad businesses are rewarded, in many ways at the expense of better businesses. This makes those better businesses less competitive in the global marketplace. Which means that jobs will flow out of America into places where the work / pay ratio is better.
The good news is that you have the freedom to do things you think will be successful. Is there demand for baseball games for blue-collar folks? Start up your own baseball league, sell tickets for $5, sell beer and hot dogs for $1 each, and pay amateurs a few extra bucks to come out and play in front of some people. Will people come out to your park? Maybe. But maybe people won't really want to come to your baseball stadium and see fourth-rate baseball played. And that's the risk you take in a capitalistic society. If you want the best baseball players to come and play for you, you'll have to pay them what they demand. And if you have to pay them a lot, you'll have to earn lots of money. It's just the way things work.
Of course, there's another way to do things, the socialist way. The government makes sure everyone has a good lifestyle, regardless of what they do. The government also tries its best to make sure the best people don't get too far ahead of the worst people. The government has a hand in almost every part of the economy. Europe does this a lot, and the lowest standard of living in Europe tends to be better than the lowest standard of living here. And w
Only Crays and DSPs used to have pipelines that long
Show me one fixed-point DSP with anywhere close to a 32-stage pipeline.
It is rare to see fixed point DSP pipelines longer than 12 stages, and most of that is just memory latency. That's short compared to just about any modern microprocessor.
Not even floating point DSPs, which are irrelevant; nobody uses them. Well, maybe not nobody, but certainly the vast majority of the market is fixed-point with 16x16 multipliers. Pipelines for modern versions of these DSPs are usually in the 5-12 stage range.
Try turning down your MTU. This is a common problem when packets get encapsulated somewhere along the link.
Some sites disallow ICMP. This means that people can't use ICMP ("ping") to do malicious stuff, but it also means that the TCP Max MTU discovery doesn't work. Manually turning down your max MTU can solve this problem.
I see this problem a lot on LANs that connect to the internet through an PPPoE connection. The PPPoE has a Max MTU of 1492 or so, because of the PPP encapsulation. The LAN has a max MTU of 1500, and that causes some problems. Turning down the MTU to 1480 made everything run smoothly.
Then again, you could be having some other problem.
Re:But will it be competitive against WiMax?
on
Tetherless Wireless
·
· Score: 1
802.16 still has big problems with doppler shifting at automobile speeds and faster, and it also has multipath problems. It's fine if you are a stationary thing (like a house) with a nice antenna (and I think it will become a decent alternative to DSL and Cable Modem), but if you want to move around and use the network you might be disappointed.
Also, the next revision of EVDO, EVDO Rev A, is starting to be integrated into devices and infrastructure, and has multi-megabit download and faster upload speeds.
Another interesting problem: will people deploying WiMax buy up frequency, or will you have interference problems from 802.11 networks, cordless phones, and other devices using unlicensed frequencies? One of the big reasons your mobile phone works so well is that the carriers have chunks of spectrum that won't be used by anyone else.
I think it will be interesting as it plays out. Frequency is expensive, and the CDMA wireless networks are designed for handling lots of users, and do a pretty good job at it. They handle voice very well, and getting better at data all the time. 802.16 is slightly faster than CDMA 1x EV-DO RevA (20-30% iirc), when stationary, but has lots of real-world problems: QoS, Doppler, Multipath, and things like billing that the people wanting to deploy the infrastructure really care about.
A lot of it is just marketing hype, though. Intel wants a big piece of the wireless market and all their chipsets have been failures so far. They certainly don't have the advantages in the CDMA space that Qualcomm has. So they shout about how great 802.16 is and hope people get on the wagon so they can be more competitive.
Then again, it worked with Itanium: intel shouts about how great ia64 is and it will kill all the 64-bit RISCs, and look who gave up their RISCs to get on the Intel wagon: SGI (MIPS), HP/Compaq (PA-RISC, Alpha). Imagine where Alpha could be now. The only folks who didn't buy into the Intel plan were the PowerPC folks, and now Apple has jumped ship.
Anyway, Intel knows marketing hype works, and so they're trying to do it in the huge and relatively high-margin wireless industry. Time will tell how things will unfurl.
Substantially higher bandwith for both forward and reverse links. Both Verizon and Sprint still have a lot of infrastructure yet to deploy. DO and DO rev A infrastructure should be more and more common as
carriers add to and replace systems, just as DO and DO rev A will become more common in handsets.
But at least Verizon is somewhat on the ball with modern wireless data telecommunications. Why are the US GSM carriers so slow in deploying WCDMA? If you think EV-DO sucks, go look at EDGE.
I used to live out east. I was skeptical of Texas Barbecue. But I have come to realize:
Cows are better than pigs for eating, and Texas barbecue is better than East Coast Barbecue.
Low country barbecue is great, I like pulled pork just fine, but it's dogfood compared to some succulent Texas Brisket. Maybe with some Rudy's "sause" and some pickles and onions... or maybe head down to the Salt Lick...
But even pork loin cooked up Texas style is, IMNSHO, better than pulled pork.
You're making me hungry... early lunch today! Wohoo!
802.16 is designed precisely for this purpose -- point to point, largeish distances, and especially good for fixed stations with line-of-sight.
The jury is still out for whether 802.16 can be used for mobile phones, but it is exactly designed for wireless MANs.. Metropolitan Area Networks.
When 802.16 comes down in price substantially, and Intel and the other WiMax folks are working hard on it, 802.16 will (in my opinion) go from a useless buzzword to something that companies can actually use for network distribution.
When that happens, I have faith in the US Entrepreneurial Spirit, and I think new service providers will spring up, offering wireless internet access at lower costs than Cable or DSL. And the Cable and DSL operators will either drop
their prices to compete, or will lose customers.
There's a lot of infrastructure with cable and phone lines... lots of physical wires to maintain, route, replace equipment on... and it costs money. Wireless distribution can be much cheaper.
I predict 802.16 antennas will start to pop up on houses over the next 10 years, just like the DirecTV dishes over the past 10.
Take, for instance, MindSpring. A local ISP, that grew to become a major, nation-wide ISP, which is now struggling to become a player in the broadband market. I think we'll see similar life cycle with the wired broadband, and then with 802.16.
Or maybe a wireless provider will provide cheap 1xEV-DO or HSDPA. We can only hope.
If you are looking at getting a Masters, there are two types of studies, at most Universities: a classroom-only track and a research-also track.
If you are planning on doing research, go to a university with a good research program, preferably in a field you are interested in. If you want to do classroom only, go to a university that has a reputation for being a good university.
For computer engineering, in the US, that means a list that starts out with schools like University of Texas at Austin, University of Wisconsin, MIT, Georgia Tech, CalTech, maybe University of California at Berkely, University of Illinois, and several others. But even these schools vary: for Computer Architecture (what I do) UT and UW are very good, but for Digital Signal Processing (also what I do, I guess), Georgia Tech and MIT are very good.
I don't know too much about schools outside the US. I know that Cambridge does stuff with Xen and virtualization that is very interesting. I know that University of New South Wales in Australia does interesting stuff with the L4 microkernel. But I list these only because they have touched what I've done at work, I'm positive that there are other UK/Aus schools that do other interesting research.
Basically, I would suggest familiarizing yourself with what research you are interested in and go to a school that does research in that area. At least, if you want a research-based degree. If you just want to take a bunch of classes, pick a good school and go.
But a graduate degree from a mediocre school is probably pretty useless.
For those not so versed in the way things work, here's what this does:
getuid() and geteuid() get the "user ID" and "effective User ID". This scripts makes a library that makes these functions return 0 -- signifying that you are the superuser. It doesn't actually make you the superuser, but it implements functions that pretend you are.
Anyway, then it compiles these functions into a library and tells the loader to link that library first. So when you start up a shell, it will call getuid() to figure out your userid, and your new library will tell it that you are userid 0. The shell prints out the root prompt instead of the regular one.
But this in no way actually gives you root privs, if you go to edit/etc/passwd you will not be able to.
Pretty funny joke, though, kind of like telling everyone that they can ftp to 127.0.0.1 using their own regular userid and password and find all the porn they could ever want...
Well, maybe a bit of explaning here. Most modern java VMs use "dynamic binary translation" to translate the java to the host instruction set. They are able to do many of the same kind of optimizations a compiler does at the same time.
This is the same thing that Transmeta does with their code morphing, but the underneath architecture is not exposed. And instead of Java bytecodes (which is good at code size, but rather poor for untranslated performance due to the stack architecture) they are translating x86 instructions (which are somewhat better for native performance, but can still be improved with translation).
Now, AMD and Intel are also doing this translation -- but instead of doing it in software like Transmeta or a Java VM, they do it with a hardware block. And instead of allocating memory for the results, they just re-translate them later (or, in the P4, they might cache them in the instruction trace cache).
As it turns out, out-of-order hardware scheduling tends to beat software scheduling, many times fairly substantially. And, as it turns out, translation is fairly effective. So it is, in my opinion, likely that a decent translating java VM on a good ol' Athlon or P3 might have better performance than a Transmeta doing only one translation from Java to their internal VLIW. Though either one might win, and the performance difference would probably not be so big.
To top it all off, java bytecode performance usually isn't the bottleneck. Sometimes it is, but for most uses of java, most of the time is not spent in executing bytecodes, but in executing the libraries, which are usually compiled for the host machine.
In conclusion, if you have a decent VM, you are probably already executing java bytecodes natively right next to your native x86 Linux apps, and you even do your "code morphing" -- but that is done by the java VM, not embedded code that you can't see.
Thought everyone might like an explanation of what Xen does, in comparison to VMware.
VMware works with a host operating system to provide a complete x86 virtual environment for a guest operating system.
Xen is an operating system in its own right. It's a "virtual machine monitor" or "hypervisor". It can spawn multiple guest operating virtual machines.
x86 is not a very good architecture for virtualization. To have a virtualizeable architecture, anything a user-level program can do should behave the same way it would in supervisor mode, or it should trap so the virtual machine monitor can emulate it. x86 has instructions that don't quite follow this guideline -- for instance, you can see what protection ring you are currently in. In supervisor mode, you would get something like ring 0. In user mode, you get ring 3. So an operating system trying to see what ring it was in would get ring 3, but you are trying to fool it into thinking it is in ring 0.
Anyway, Xen modifies the guest architecture. It disallows these "sensitive" instructions and creates some virtual devices that are easier to emulate (like a simple software-programmed TLB). This allows the performance to be very very good, faster than VMware, but it requires you to fiddle with your operating system a bit. Which, of course, is easy to do with Linux.
IIRC, PowerPC was engineered to be backwards compatible with 68k. To preserve apple's software. The main dis-advantage of this is that you'd have to support the umpteen billion addressing modes.
No, you're completely wrong. PPC is a fairly traditional RISC, 32 registers plus some special ones, and a rich instruction set, but ALU operations operate on registers, not memory.
Apple did have a 68k emulator in their OS to be able to run 68k binaries.
On the other hand, the early coldfire processors, like the ones in the palm pilots, are68k chips. The newer ones are 68k-assembly-compatible, and have memory operands, a la CISC architectures.
But I can understand the confusion. CISC and RISC don't really mean anything anymore, they are merely encoding strategies that are translated to some similar (or very different) internal representation.
Perhaps we need to start developing patents where everyone is free to use the innovation, and the licensing terms require all patents using the disclosed innovation to also be licensed freely on the same terms.
This is very similar to the GPL, which uses copyright law to enforce derivitave works also be GPL'd.
Sounds good. FSF, are you listening? Get your lawyers to work on the legal licensing terms and let's get pantenting!
One considerable aspect to consider is game developers.
Are you going to want to invest the money for a Xbox game if it won't run on next year's xbox2? So game developers wait, and fewer new games come out for the existing platform, causing people to be more interested in the competition...
Re:Why do oil companies fund this research?
on
Ethanol From Waste Straw
·
· Score: 4, Informative
Take solar energy. Oil companies own somewhere around 90% of the patents on solar energy. Why do you think they do this? Simple, better to fund the research themselves so they own the patents. This prevents anyone else from actually inventing something new and possibly marketing it.
Let's see some links. Or did you just hear this from some guy?
Oil companies do lots of research into natural gas and ethanol and the like because they know that one day, many many years from now, the oil production will not be able to meet demand. The company that can provide the fuel via another method will be the one making the profit. It just makes sense.
Anyway, if you get a patent on something, it is made public knowledge, and it is available for public use by anyone after a few years. And, in the meantime, the knowledge is used to further the state of the art.
So what you are saying is: Oil companies fund alternative fuel resource research and that knowledge is made public, furthering the state of the art and making us more independant from oil. They own the rights to the inventions for a while, but they make the invention public knowledge and the invention is released to the public after a period of time.
Well, that sounds pretty reasonable! Maybe these companies aren't the evil entities the propaganda you listen to would leave you to believe? Maybe they are normal people, trying to make some money, and concerned about the future.
On the flip side, when was the last time someone who went "off the grid" contributed to the state of the art in energy production?
Oooh! Add it to the list
on
TiVo Will Die
·
· Score: 5, Funny
Another one to add to the "foo is dying" list:
Apple is dying
Linux is dying
Real is dying
*BSD is dying
Tivo is dying
America is dying
Europe is dying
Morality and Ethics are dying
People who color fabric are dyeing
Which will pull through? Which won't? Who's going to be next? Place your bets!
Seriously, though, I think that licensing to DirecTV et al will help out TiVo in a pretty substantial way.
Slashdot Mirror
Hi, welcome to America. I'll be giving you a brief synopsis of how a free market works.
In a free and open job market, you have people and you have jobs. Many people have the ability to do more than one kind of job. Likewise, most jobs could potentially be filed by one of a relatively large set of people.
Some jobs are able to be done by lots of people, like being a cook in a fast food restaraunt. Some people are willing to take that job because they do not have a skill set that enables them to work a different job, or maybe because they like it. Anyway, the worker and the business agree on compensation for the job that is agreeable to both the business and the worker. If there is someone willing to do the same job for less, then the worker should get paied less. If there is nobody that can do as good of a job for the same price, the worker can ask for more money and may get it.
For a job like a CEO or Baseball player, there is incredible demand to get the absolute best people for those positions. If you are the best baseball player, you can demand a lot of money, because there is demand to have the absolute best player on your baseball team: it would help you win, wins help you sell more tickets and advertisements, and people want their team to win. If your team is a perpetual loser, you won't sell as many tickets. (Unless you are the Cubs, because of Cubs Fans like Me).
Likewise, Good CEOs can demand a premium. Someone who can manage a business effectively has a skillset that is in high demand. If you can make your company that earns 1 Billion / year, and grow that profit by 20%, you are worth a lot of money. Stockholders are willing to pay these high salaries because there is the perception that a good CEO can lead a company to be more profitable.
Of course, this is all based on perception. If killing the R&D saves money now, but you lose out on the next major innovation, you've screwed yourself. But that's how the market works, and your competition (who did the right R&D) will beat you. As they should.
Now you may notice that America isn't exactly a free market. Even when Southwest airlines is making money, the Government gives billions of dollars to airlines who didn't make good business decisions. Unfortunately, the Government has, over the last several decades, gotten more involved with business. This leads to situations where bad businesses are rewarded, in many ways at the expense of better businesses. This makes those better businesses less competitive in the global marketplace. Which means that jobs will flow out of America into places where the work / pay ratio is better.
The good news is that you have the freedom to do things you think will be successful. Is there demand for baseball games for blue-collar folks? Start up your own baseball league, sell tickets for $5, sell beer and hot dogs for $1 each, and pay amateurs a few extra bucks to come out and play in front of some people. Will people come out to your park? Maybe. But maybe people won't really want to come to your baseball stadium and see fourth-rate baseball played. And that's the risk you take in a capitalistic society. If you want the best baseball players to come and play for you, you'll have to pay them what they demand. And if you have to pay them a lot, you'll have to earn lots of money. It's just the way things work.
Of course, there's another way to do things, the socialist way. The government makes sure everyone has a good lifestyle, regardless of what they do. The government also tries its best to make sure the best people don't get too far ahead of the worst people. The government has a hand in almost every part of the economy. Europe does this a lot, and the lowest standard of living in Europe tends to be better than the lowest standard of living here. And w
Get in line for what?
Horray! Intelligent humor on slashdot!
Show me one fixed-point DSP with anywhere close to a 32-stage pipeline.
It is rare to see fixed point DSP pipelines longer than 12 stages, and most of that is just memory latency. That's short compared to just about any modern microprocessor.
Not even floating point DSPs, which are irrelevant; nobody uses them. Well, maybe not nobody, but certainly the vast majority of the market is fixed-point with 16x16 multipliers. Pipelines for modern versions of these DSPs are usually in the 5-12 stage range.
Perhaps you are thinking of graphics pipelines?
Some sites disallow ICMP. This means that people can't use ICMP ("ping") to do malicious stuff, but it also means that the TCP Max MTU discovery doesn't work. Manually turning down your max MTU can solve this problem.
I see this problem a lot on LANs that connect to the internet through an PPPoE connection. The PPPoE has a Max MTU of 1492 or so, because of the PPP encapsulation. The LAN has a max MTU of 1500, and that causes some problems. Turning down the MTU to 1480 made everything run smoothly.
Then again, you could be having some other problem.
Also, the next revision of EVDO, EVDO Rev A, is starting to be integrated into devices and infrastructure, and has multi-megabit download and faster upload speeds.
Another interesting problem: will people deploying WiMax buy up frequency, or will you have interference problems from 802.11 networks, cordless phones, and other devices using unlicensed frequencies? One of the big reasons your mobile phone works so well is that the carriers have chunks of spectrum that won't be used by anyone else.
I think it will be interesting as it plays out. Frequency is expensive, and the CDMA wireless networks are designed for handling lots of users, and do a pretty good job at it. They handle voice very well, and getting better at data all the time. 802.16 is slightly faster than CDMA 1x EV-DO RevA (20-30% iirc), when stationary, but has lots of real-world problems: QoS, Doppler, Multipath, and things like billing that the people wanting to deploy the infrastructure really care about.
A lot of it is just marketing hype, though. Intel wants a big piece of the wireless market and all their chipsets have been failures so far. They certainly don't have the advantages in the CDMA space that Qualcomm has. So they shout about how great 802.16 is and hope people get on the wagon so they can be more competitive.
Then again, it worked with Itanium: intel shouts about how great ia64 is and it will kill all the 64-bit RISCs, and look who gave up their RISCs to get on the Intel wagon: SGI (MIPS), HP/Compaq (PA-RISC, Alpha). Imagine where Alpha could be now. The only folks who didn't buy into the Intel plan were the PowerPC folks, and now Apple has jumped ship.
Anyway, Intel knows marketing hype works, and so they're trying to do it in the huge and relatively high-margin wireless industry. Time will tell how things will unfurl.
Wake me up when they've solved the multipath and doppler problems at the high datarate.
Huh? What lower userid?
Substantially higher bandwith for both forward and reverse links. Both Verizon and Sprint still have a lot of infrastructure yet to deploy. DO and DO rev A infrastructure should be more and more common as carriers add to and replace systems, just as DO and DO rev A will become more common in handsets.
But at least Verizon is somewhat on the ball with modern wireless data telecommunications. Why are the US GSM carriers so slow in deploying WCDMA? If you think EV-DO sucks, go look at EDGE.
Low country barbecue is great, I like pulled pork just fine, but it's dogfood compared to some succulent Texas Brisket. Maybe with some Rudy's "sause" and some pickles and onions... or maybe head down to the Salt Lick...
But even pork loin cooked up Texas style is, IMNSHO, better than pulled pork.
You're making me hungry... early lunch today! Wohoo!
The jury is still out for whether 802.16 can be used for mobile phones, but it is exactly designed for wireless MANs.. Metropolitan Area Networks.
When 802.16 comes down in price substantially, and Intel and the other WiMax folks are working hard on it, 802.16 will (in my opinion) go from a useless buzzword to something that companies can actually use for network distribution.
When that happens, I have faith in the US Entrepreneurial Spirit, and I think new service providers will spring up, offering wireless internet access at lower costs than Cable or DSL. And the Cable and DSL operators will either drop their prices to compete, or will lose customers.
There's a lot of infrastructure with cable and phone lines... lots of physical wires to maintain, route, replace equipment on... and it costs money. Wireless distribution can be much cheaper. I predict 802.16 antennas will start to pop up on houses over the next 10 years, just like the DirecTV dishes over the past 10.
Take, for instance, MindSpring. A local ISP, that grew to become a major, nation-wide ISP, which is now struggling to become a player in the broadband market. I think we'll see similar life cycle with the wired broadband, and then with 802.16.
Or maybe a wireless provider will provide cheap 1xEV-DO or HSDPA. We can only hope.
When great books get turned into movies, the directors use the books as toilet paper. LOTR was no exception.
I despise Peter "Dwarves are Clowns" Jackson.
65nm wiring is really slow. What we're seeing from TSMC is still bouncing around a bit for 65nm low power, but wires are slower than we were hoping.
It looks as though people won't switch to 65nm because 65nm produces faster devices, instead people will go to 65nm for cost and capacity.
Yes, but you can middle-click in the middle of the view window, which will follow the pasted URL.
If you are planning on doing research, go to a university with a good research program, preferably in a field you are interested in. If you want to do classroom only, go to a university that has a reputation for being a good university.
For computer engineering, in the US, that means a list that starts out with schools like University of Texas at Austin, University of Wisconsin, MIT, Georgia Tech, CalTech, maybe University of California at Berkely, University of Illinois, and several others. But even these schools vary: for Computer Architecture (what I do) UT and UW are very good, but for Digital Signal Processing (also what I do, I guess), Georgia Tech and MIT are very good.
I don't know too much about schools outside the US. I know that Cambridge does stuff with Xen and virtualization that is very interesting. I know that University of New South Wales in Australia does interesting stuff with the L4 microkernel. But I list these only because they have touched what I've done at work, I'm positive that there are other UK/Aus schools that do other interesting research.
Basically, I would suggest familiarizing yourself with what research you are interested in and go to a school that does research in that area. At least, if you want a research-based degree. If you just want to take a bunch of classes, pick a good school and go.
But a graduate degree from a mediocre school is probably pretty useless.
getuid() and geteuid() get the "user ID" and "effective User ID". This scripts makes a library that makes these functions return 0 -- signifying that you are the superuser. It doesn't actually make you the superuser, but it implements functions that pretend you are.
Anyway, then it compiles these functions into a library and tells the loader to link that library first. So when you start up a shell, it will call getuid() to figure out your userid, and your new library will tell it that you are userid 0. The shell prints out the root prompt instead of the regular one.
But this in no way actually gives you root privs, if you go to edit /etc/passwd you will not be able to.
Pretty funny joke, though, kind of like telling everyone that they can ftp to 127.0.0.1 using their own regular userid and password and find all the porn they could ever want...
Well, maybe a bit of explaning here. Most modern java VMs use "dynamic binary translation" to translate the java to the host instruction set. They are able to do many of the same kind of optimizations a compiler does at the same time.
This is the same thing that Transmeta does with their code morphing, but the underneath architecture is not exposed. And instead of Java bytecodes (which is good at code size, but rather poor for untranslated performance due to the stack architecture) they are translating x86 instructions (which are somewhat better for native performance, but can still be improved with translation).
Now, AMD and Intel are also doing this translation -- but instead of doing it in software like Transmeta or a Java VM, they do it with a hardware block. And instead of allocating memory for the results, they just re-translate them later (or, in the P4, they might cache them in the instruction trace cache).
As it turns out, out-of-order hardware scheduling tends to beat software scheduling, many times fairly substantially. And, as it turns out, translation is fairly effective. So it is, in my opinion, likely that a decent translating java VM on a good ol' Athlon or P3 might have better performance than a Transmeta doing only one translation from Java to their internal VLIW. Though either one might win, and the performance difference would probably not be so big.
To top it all off, java bytecode performance usually isn't the bottleneck. Sometimes it is, but for most uses of java, most of the time is not spent in executing bytecodes, but in executing the libraries, which are usually compiled for the host machine.
In conclusion, if you have a decent VM, you are probably already executing java bytecodes natively right next to your native x86 Linux apps, and you even do your "code morphing" -- but that is done by the java VM, not embedded code that you can't see.
Oh, and dynamic binary translation is cool.
VMware works with a host operating system to provide a complete x86 virtual environment for a guest operating system.
Xen is an operating system in its own right. It's a "virtual machine monitor" or "hypervisor". It can spawn multiple guest operating virtual machines.
x86 is not a very good architecture for virtualization. To have a virtualizeable architecture, anything a user-level program can do should behave the same way it would in supervisor mode, or it should trap so the virtual machine monitor can emulate it. x86 has instructions that don't quite follow this guideline -- for instance, you can see what protection ring you are currently in. In supervisor mode, you would get something like ring 0. In user mode, you get ring 3. So an operating system trying to see what ring it was in would get ring 3, but you are trying to fool it into thinking it is in ring 0.
Anyway, Xen modifies the guest architecture. It disallows these "sensitive" instructions and creates some virtual devices that are easier to emulate (like a simple software-programmed TLB). This allows the performance to be very very good, faster than VMware, but it requires you to fiddle with your operating system a bit. Which, of course, is easy to do with Linux.
Apple did have a 68k emulator in their OS to be able to run 68k binaries.
On the other hand, the early coldfire processors, like the ones in the palm pilots, are68k chips. The newer ones are 68k-assembly-compatible, and have memory operands, a la CISC architectures.
But I can understand the confusion. CISC and RISC don't really mean anything anymore, they are merely encoding strategies that are translated to some similar (or very different) internal representation.
This is very similar to the GPL, which uses copyright law to enforce derivitave works also be GPL'd.
Sounds good. FSF, are you listening? Get your lawyers to work on the legal licensing terms and let's get pantenting!
What I want to know is... why did they waste so much opcode space with preshift operands? Do you really "ldr r3,[r0, r1 ROR r2]" all that often?
Woo! Truly, the geek's cooking magazine. Plus, no ads. It's great. I love it.
Are you going to want to invest the money for a Xbox game if it won't run on next year's xbox2? So game developers wait, and fewer new games come out for the existing platform, causing people to be more interested in the competition...
Oil companies do lots of research into natural gas and ethanol and the like because they know that one day, many many years from now, the oil production will not be able to meet demand. The company that can provide the fuel via another method will be the one making the profit. It just makes sense.
Anyway, if you get a patent on something, it is made public knowledge, and it is available for public use by anyone after a few years. And, in the meantime, the knowledge is used to further the state of the art.
So what you are saying is: Oil companies fund alternative fuel resource research and that knowledge is made public, furthering the state of the art and making us more independant from oil. They own the rights to the inventions for a while, but they make the invention public knowledge and the invention is released to the public after a period of time.
Well, that sounds pretty reasonable! Maybe these companies aren't the evil entities the propaganda you listen to would leave you to believe? Maybe they are normal people, trying to make some money, and concerned about the future.
On the flip side, when was the last time someone who went "off the grid" contributed to the state of the art in energy production?
Which will pull through? Which won't? Who's going to be next? Place your bets!
Seriously, though, I think that licensing to DirecTV et al will help out TiVo in a pretty substantial way.