The big customer is everyone who's buying PowerQuicc's and putting them in embedded spaces. PowerQuicc's with RapidIO connections, PowerQuicc's four-on-a-board, lots and lots of PowerPC chips going in lots and lots of embedded spaces.
I was recently at the Global Signal Processing Expo and it was amazing how many people were doing tasks involving heavy signal processing -- where you would expect DSPs and FPGAs -- on PowerPC chips. The interesting thing was that raw number-crunching power wasn't always the most important thing -- many times it is bandwith (what kind of interconnect you have to your processor makes a huge difference when you are trying to process gigabytes of information a second). Sometimes it is programmability that is the reason (use of familiar tools is a big plus). Sometimes you just want to use the same chip to do your signal processing as your network I/O.
Read the Lord of the Rings Trilogy (by JRR Tolkien) and the Narnia books (by CS Lewis) every year. Otherwise you'll grow older. Keeping the magic of your youth alive in you is essential for having an interesting, flavorful life.
I don't think you understand translation, or maybe you don't understand EPIC.
EPIC (or VLIW, which is pretty much the same thing) atchitectures define instructions to be executed (or, at least, that can be executed) in parallel in the encoding of the instructions. Most superscalar machines evolved from single-issue architectures (PPC, Alpha, x86) only have sequential instructions.
That being said, there are almost always instructions that can be executed in parallel. The only difference between EPIC/VLIW and sequential dynamically-parallel machines is whether the hardware or the assembler/compiler/optimizer determines whether or not instructions can be executed in parallel.
Now, software that interprets (for instance) java can utilize parallelism, because it can do things like lookup the place to jump for the next bytecode while it's executing the code to handle the current bytecode.
Software that translates (for instance) java to native code can utilize parallism, both in the translation phase, and in the native code output. Of course, scheduling code is a complex thing, but it doesn't mean it's impossible. And given that you have the scheduler for a good compiler for your architecture (which Intel does) you may have many of the problems already solved.
Translating x86, or PPC, or any other architecture is very similar... you have interpretation (usually somewhat slow), and varying levels of translate-to-native (which ends up with varying quality of output). Some architectures are easier to interpret than others (Java is nice, because it has no flags), but in the end the problem is pretty much the same between architectures.
Itanium's flaws are not in it's EPIC/VLIW nature. VLIW chips can be very small and efficient. Most high-performance embedded chips are migrating towards VLIW (see: TI DSPs, StarCore DSP, TransMeta) because you can get parallelism without having to add hardware complexity for out-of-order execution, hazard detection, etc. At the very least, you can get a VLIW that performs the same as a single-issue computer.
The problem with the Itanium is that it has lots of features. It's very complex. It has lots of predicates, lots of cache prefetch instructions, etc. that make it very cycle-efficient for either a hand programmer or a compiler that knows everything.
Unfortunately, the complexity makes design rather challanging, and many of the features aren't being used by production compilers, because they're hard to implement. Normal instructions are easy.
Sigh. It would be really great to see a good, simple VLIW with a good compiler on the market. In the end, however, it would only be a little more cycle-efficient and a little less costly than a superscalar chip with similar features. And the superscalar has the code base and the existing, proven compiler. Guess which one wins?
I think there's more momentum behind RapidIO for multidevice interconnect on high-bandwith processing (like wireless base station). Especially with PowerPCs coming out with RapidIO on borad.
I think that the major future switch fabric is PCIExpress. It is what's going to go into the consumer and semi-consumer markets (at least, that's my guess). It's going to be fast, serial, switched, and most importantly, has the 800lb gorilla of Intel behind it.
Anyway, network latency is determined mainly by processor speed, distance, layout, and bandwith of the network. On my 100bt network at home, going through a couple of 100bt switches, I usually get You can also generalize this to any sort of interconnect fabric, like RapidIO or HyperTransport or PCIExpress. There is always latency to the memory. Cache is rather low-latency, typically a few cycles, and is compensated for by the pipeline on your processor. L2 cache is higher latency -- maybe 5-10 clocks. Memory is, say, 30 clock cycles[0]. Disk is, give or take, 12 million clock cycles[1]. 6ms is a lot of time to send info over a network, even one where latency might be a few ms.
This is all one of those things where "it depends on the application". If you are on a fairly local, fairly high-bandwith, fairly low-latency network, it makes sense. If you're on a 28.8 dialup, obviously swapping to your disk will be faster.
But in many corporate environments, you have lots of machines on lots of desks that are attached to a fairly high-bandwith, low latency network. If engineers rarely and sporadically need large amounts of memory, this can be a more effective method than swapping to disk.
[0] Assuming 133Mhz/266DDR memory CAS-2 and 2Ghz CPU
There's lots of research about network shared memory for use in various things.
It's very interesting that using memory over the network is very much the same problem as cache coherency amongst processors. If you have multiple processors, you don't want to have to go out to the slow memory when the data you want is in your neighbors cache... so perhaps you grab it from the neighbor's cache.
Similarly, if you have many computers on a network, and you are out of RAM, and your nighbor has extra RAM, you don't want to page out to your slow disk when you can use your neighbor's memory.
NUMA machines are somewhere in between these two scenarios.
There are lots of problems: networks aren't very reliable, there's lots of network balancing issues, etc. But it's certainly interesting research, and can be useful for the right application, I guess.
Disk is slow, though... memory access time is measured in ns, disk access time is in ms... that's a 1,000,000x difference. So paging to someone else's RAM over the network can be more efficient.
I don't have any good papers handy, but I'm sure you can google for some.
Limited speed of signals acros the chip: If the clock frequency gets much larger a signal would require several buffer stages to reach the other side.
This is the case now. The p4 has several stages simply for driving data. I think it's for rather loaded busses, but wire delay is still very problematic.
I'm more hopeful that we might get away from the whole stupid clock idea and go asynchronos.
Clocks aren't so bad, they make a lot of things very simple. Asynchronous is getting easier, and there are lots of people working on it, but the end result isn't fantastically better -- you get average performance per-stage instead of worst-case performance per-stage. For most modern processors, that's not much of a difference; the stages are typically pretty well balanced. Stages that would be a very burdensome critical path are just split up.
Of course, there are always simpler operations that can get done a bit faster -- but as wire delay gets worse and transistors switch faster, routing information is becoming much more critical than computational delay. Calculation is pretty cheap; forwarding is expensive.
Branch prediction is easy. You can get very good performance (95% or so for many benchmarks) with a very small table and one-bit counter. More complex solutions involve even better results. Pipeline problems were taken into account for the papers linked to... that's why optimal floating point pipeline stages are even shorter than integer ones... there are fewer dependancies and hazards.
And, of course, the P4 can still do useful work on the other pseudo-processor whilst the first one is refilling on a mispredict or cache miss.
One of the problems with "leaky" parts is that the threshold voltages are kept very low. This makes the transistors switch much faster, but makes them leak current quite a bit.
You can fairly easily raise the threshold voltage (for a process). It makes the chip slower, but leaks less current (and therefore usually uses less power). This is one of the key elements of "Low Power" processes like CL013LP.
Interestingly, Using leaky transistors that switch faster has been a trick used for a very long time. One of the reasons the Cray computers took so much cooling was that they didn't use MOSFETs, their whole process was based on PNP and NPN junction transistors. For those who don't know much about transistors, FETs (or Field Effect Transistors) make a little capacitor that when you charge it up (or don't charge it up, depending), it lets current flow through on the other side. It takes a while to charge up the capacitor (time constant proportional to Resistance times Capacitance, remember!), but once it's charged there isn't any current (except the leakage current) that flows through.
At least, that's what I recall from my classes. I didn't do so well in the device physics and components classes.
..if Intel and AMD hadn't got locked into that stupid GHz battle and instead concentrated on optimizing their CPU design (rather than just ramping up the speed silly amounts)
If you don't think their designs are "optimized", What do you think the designs are?
AMD and Intel have gone to great lengths to give their processors deep pipeines. That's a GOOD thing. And they have both gone to great lengths to reduce the penalties associated with deep pipelines (like change-of-flow penalty).
This, in a nutshell, is what computer microarchitecture is. Recentstudies have shown that the optimal pipeline depth is very, very long, roughly twice the depth of the P4. So making the pipeline deeper seems to be the right solution.
Taking the totals from your statistics (I'm assuming they're correct):
(Per 100k)
Total Robberies:
Canada: 18+78 == 96
US: 63+102 == 165
Rate Increase: 1.7
Total Murders:
Canada: 0.5+0.23+1.3 ~= 2.0
US: 4.4+3.3+2.3 == 10.0
Rate Increase: 5x
It seems to me that Americans are much more criminal than Canada. Because Americans have laxer laws for guns, perhaps there are more guns involved. However, there is no causal evidence for this. That's the problem with information like this -- you can show what has happened, but not why it happened. If handguns were outlawed, perhaps crime would increase, as it is less likely that Average Joe would be armed.
I think it boils down to the supply/demand curve. Illegal file sharing is not good on face, but it will break the inflated CD cost model we have now, and probably increase music listnership and (I think) make better quality music in the end.
CD prices are extaordinarily high given the cost of production. Look at how many CDs AOL pushes out, or how much cheapbytes sells CDs for. If I could get CDs for $1 or $2, which is still 100% more than the cost of production for a big lot of CDs (I don't care about the artwork, just give me the CD in an envelope), I would buy a lot more CDs.
Another example... look at Emusic.com.
It's a great service. Fast, reliable. They give me a quality product, I give them what I feel is a reasonable amount of money.
And when record companies work somewhat like cheapbytes, producing bulk CDs, selling for $2, and giving back the artists $1, they won't have money to hype Boy Band of the Month, and will instead focus on lifting up good music that people want to listen to, music that sells itself.
Most of your requested features are planned for Segway v2. Not only will these features be added, but Segway will be safe to drive on normal streets. It will truely be a revolutionary thing. Entire countries will be redesigned around Segway v2. Families around the world will come to our office and bow before us, thanking us for the amazing accomplishment we have made.
The initial analysis is that Segway v2 will cost about $320,000, and will have a range of 70km (about 45 miles), or possibly even more. It will be able to run for up to 90 minutes, and when the power is exhausted a simple 18-hour process can be used to power the machine up again.
I think you will agree that $320,000 is a small price to pay for technology that will revolutionize the world.
There are actually some interesting papers out about optimal pipeline depth. At first they appear to have different conclusions, as they cover different architectures, but the conclusion is really sort of the same: optimum pipeline depth is about six fanout-of-four inverters per stage of work for integer paths and four for floating-point paths. Plus two (each) for overhead. That leads to crazy-long pipelines, I think the rough calculation for redoing the P4 pipeline came out to 50 stages or something.
If you do a google search on optimal pipeline depth you'll find some good results.
One of the great things about CDs is that they have become so cheap. Many of the blank CDs are used to store music from local bands or individuals. It's so easy to record a reasonably good mix of yourself or your small band, and copies come out of your computer for pennies. Almost every band that does local live music sells CDs.
And guess what? They're not copy protected.
So go out to the bars that have open mic night or are showing some sort of a local band. And buy their CDs. It's better than the boy-band-of-the-day or whatever crap is playing on your local Clear Channel station.
1. the schedule may specify different royalties for different language versions;
3
2. the schedule may specify reasonable volume discounts based upon the actual
volume of licenses of any Windows Operating System Product or any group of
such products; and
3. the schedule may include market development allowances, programs, or other
discounts in connection with Windows Operating System Products, provided
that:
a. such discounts are offered and available uniformly to all Covered OEMs,
except that Microsoft may establish one uniform discount schedule for the
ten largest Covered OEMs and a second uniform discount schedule for the
eleventh through twentieth largest Covered OEMs, where the size of the
OEM is measured by volume of licenses;
b. such discounts are based on objective, verifiable criteria that shall be
applied and enforced on a uniform basis for all Covered OEMs; and
c. such discounts or their award shall not be based on or impose any criterion
or requirement that is otherwise inconsistent with any portion of this Final
Judgment.
This is the best part. There's a standard schedule for costs for everyone. You can't brown-nose Microsoft and get special deals, and Microsoft can't punish you by raising your costs for licenses.
You buy licenses at the same rate as everyone else (who buys in the same quantities for the same languages as you do).
This means that if Dell is the largest buyer of MS Licenses, it doesn't matter what they do, they get the cheapest rate. Likewise, any company that buys 2,000 Windows licenses gets the same rate, whether they are MS Fanboys or Sun Microsystems.
Of course, a big "in theory" is added to all this.
I see people here on Slashdot a lot who dislike
the x86 processors because they do translation from the x86 ISA into internal opcodes.
Note that your new IBM chip is doing exactly that.
Intel and AMD have repeatedly shown that they can do whatever they like to implement top-notch internal architectures, and lopping on a translation unit only adds 10-20% die area and typically a very small performance hit over a traditional sequential RISC architecture. And they're free to change the internal architecture between revisions. And both Intel and AMD sell enough chips that they can spend a lot of money on designs and make them very good and still turn a profit.
If Russians are breaking into computers in the US, the crime is (perhaps, IANAL) committed in the US, and so the US has jurisdiction. If the FBI breaks into Russian computers (without consent from anyone) then, by the same token, the crime would be in Russia. So it would be up to the Russian government to prosecute the FBI (or investigators in the FBI), right?
Sometimes things aren't so "the-USA-is-really-bad" as Slashdot says they are (and sometimes they are, and sometimes they're probably worse).
Of course, if we use Law and Order as our legal source (and, though IANAL, I've watched a lot of L&O), then Jack McCoy would say that we have a responsibility to prosecute criminals when their own countries won't, and that as long as an element of the crime was taking place within jurisdiction of the court, the court should have prosecutorial powers. But in the episode where Jack and Carmichael were outside of the Supreme Court and the decision comes out, after attempting to prosecute a foreign diplomat for murder, they (frustratingly) don't tell us what the decision is. D'oh!
I don't want a napster I pay for. That means that the record labels would make money from other people's bandwith.
However, I would like a place where I could download very high quality, RAW.wav or Ogg Vorbis or MP3 files for, say, $0.50-$1.00 each. Maybe $5.00 for a whole album. From a fast server. That are not in some sort of DRM vault.
This way, I own the music. I can do whatever I like with it: burn it to a CD, put it in my portable player, whatever I want to do within my fair use rights. And I also don't have to (effectively) pay additional money by trying to hunt someone down with the file I want at the quality I want, with a good connection that won't stop halfway through the download.
Merely having the record industry collect money for "allowing" other people to share music peer-to-peer is not sufficient.
I thought the SuperDrive was the drive that could read and write 1.4M floppy disks.
I was really glad that it could... finally I could read and write DOS 1.4M floppy disks, Mac 1.4M floppy disks, 800K Mac floppies, and 720K Dos Floppies.
I can't wait to pull out my old 512Ke with the external SuperDrive and start burning DVDs!
Oh, you mean this is just another example of Marketing BuzzWord TermReuse? Crap.
It's a lot better to keep an independant board with their own political agenda who has no accountability to the people, instead of replacing it with a review board whose political agenda matches those of the officials elected by the people?
Does anyone see that maybe this isn't such a horrible idea?
I'm all for impartiality. I just don't think it exists.
Despite what slashdotters will tell you, they aren't going to put oil tycoons on review boards for hospital ethics. But they also might not put people in those review boards who have political agendas that few people want -- like (for instance) the governmentalization of health care.
Doesn't it make some sense to put people in boards who are going to look for solutions inside of the bounds of what the American voting population will tolerate?
Slashdot Mirror
The big customer is everyone who's buying PowerQuicc's and putting them in embedded spaces. PowerQuicc's with RapidIO connections, PowerQuicc's four-on-a-board, lots and lots of PowerPC chips going in lots and lots of embedded spaces.
I was recently at the Global Signal Processing Expo and it was amazing how many people were doing tasks involving heavy signal processing -- where you would expect DSPs and FPGAs -- on PowerPC chips. The interesting thing was that raw number-crunching power wasn't always the most important thing -- many times it is bandwith (what kind of interconnect you have to your processor makes a huge difference when you are trying to process gigabytes of information a second). Sometimes it is programmability that is the reason (use of familiar tools is a big plus). Sometimes you just want to use the same chip to do your signal processing as your network I/O.
Companies like Sky Computers are selling more PowerPCs than companies like Apple Computers.
Read the Lord of the Rings Trilogy (by JRR Tolkien) and the Narnia books (by CS Lewis) every year. Otherwise you'll grow older. Keeping the magic of your youth alive in you is essential for having an interesting, flavorful life.
EPIC (or VLIW, which is pretty much the same thing) atchitectures define instructions to be executed (or, at least, that can be executed) in parallel in the encoding of the instructions. Most superscalar machines evolved from single-issue architectures (PPC, Alpha, x86) only have sequential instructions.
That being said, there are almost always instructions that can be executed in parallel. The only difference between EPIC/VLIW and sequential dynamically-parallel machines is whether the hardware or the assembler/compiler/optimizer determines whether or not instructions can be executed in parallel.
Now, software that interprets (for instance) java can utilize parallelism, because it can do things like lookup the place to jump for the next bytecode while it's executing the code to handle the current bytecode.
Software that translates (for instance) java to native code can utilize parallism, both in the translation phase, and in the native code output. Of course, scheduling code is a complex thing, but it doesn't mean it's impossible. And given that you have the scheduler for a good compiler for your architecture (which Intel does) you may have many of the problems already solved.
Translating x86, or PPC, or any other architecture is very similar... you have interpretation (usually somewhat slow), and varying levels of translate-to-native (which ends up with varying quality of output). Some architectures are easier to interpret than others (Java is nice, because it has no flags), but in the end the problem is pretty much the same between architectures.
Itanium's flaws are not in it's EPIC/VLIW nature. VLIW chips can be very small and efficient. Most high-performance embedded chips are migrating towards VLIW (see: TI DSPs, StarCore DSP, TransMeta) because you can get parallelism without having to add hardware complexity for out-of-order execution, hazard detection, etc. At the very least, you can get a VLIW that performs the same as a single-issue computer.
The problem with the Itanium is that it has lots of features. It's very complex. It has lots of predicates, lots of cache prefetch instructions, etc. that make it very cycle-efficient for either a hand programmer or a compiler that knows everything.
Unfortunately, the complexity makes design rather challanging, and many of the features aren't being used by production compilers, because they're hard to implement. Normal instructions are easy.
Sigh. It would be really great to see a good, simple VLIW with a good compiler on the market. In the end, however, it would only be a little more cycle-efficient and a little less costly than a superscalar chip with similar features. And the superscalar has the code base and the existing, proven compiler. Guess which one wins?
Your architecture prof for the day,
I think that the major future switch fabric is PCIExpress. It is what's going to go into the consumer and semi-consumer markets (at least, that's my guess). It's going to be fast, serial, switched, and most importantly, has the 800lb gorilla of Intel behind it.
Anyway, network latency is determined mainly by processor speed, distance, layout, and bandwith of the network. On my 100bt network at home, going through a couple of 100bt switches, I usually get You can also generalize this to any sort of interconnect fabric, like RapidIO or HyperTransport or PCIExpress. There is always latency to the memory. Cache is rather low-latency, typically a few cycles, and is compensated for by the pipeline on your processor. L2 cache is higher latency -- maybe 5-10 clocks. Memory is, say, 30 clock cycles[0]. Disk is, give or take, 12 million clock cycles[1]. 6ms is a lot of time to send info over a network, even one where latency might be a few ms.
This is all one of those things where "it depends on the application". If you are on a fairly local, fairly high-bandwith, fairly low-latency network, it makes sense. If you're on a 28.8 dialup, obviously swapping to your disk will be faster.
But in many corporate environments, you have lots of machines on lots of desks that are attached to a fairly high-bandwith, low latency network. If engineers rarely and sporadically need large amounts of memory, this can be a more effective method than swapping to disk.
[0] Assuming 133Mhz/266DDR memory CAS-2 and 2Ghz CPU
[1] Assuming 6ms access time and 2Ghz CPU
It's very interesting that using memory over the network is very much the same problem as cache coherency amongst processors. If you have multiple processors, you don't want to have to go out to the slow memory when the data you want is in your neighbors cache... so perhaps you grab it from the neighbor's cache.
Similarly, if you have many computers on a network, and you are out of RAM, and your nighbor has extra RAM, you don't want to page out to your slow disk when you can use your neighbor's memory.
NUMA machines are somewhere in between these two scenarios.
There are lots of problems: networks aren't very reliable, there's lots of network balancing issues, etc. But it's certainly interesting research, and can be useful for the right application, I guess.
Disk is slow, though... memory access time is measured in ns, disk access time is in ms... that's a 1,000,000x difference. So paging to someone else's RAM over the network can be more efficient.
I don't have any good papers handy, but I'm sure you can google for some.
I dunno, maybe read a book instead of sending useless emails DELETED!!!
Of course, there are always simpler operations that can get done a bit faster -- but as wire delay gets worse and transistors switch faster, routing information is becoming much more critical than computational delay. Calculation is pretty cheap; forwarding is expensive.
And, of course, the P4 can still do useful work on the other pseudo-processor whilst the first one is refilling on a mispredict or cache miss.
You can fairly easily raise the threshold voltage (for a process). It makes the chip slower, but leaks less current (and therefore usually uses less power). This is one of the key elements of "Low Power" processes like CL013LP.
For more information, the Britney Spears' Guide to Semiconductor Physics is sure to help.
Interestingly, Using leaky transistors that switch faster has been a trick used for a very long time. One of the reasons the Cray computers took so much cooling was that they didn't use MOSFETs, their whole process was based on PNP and NPN junction transistors. For those who don't know much about transistors, FETs (or Field Effect Transistors) make a little capacitor that when you charge it up (or don't charge it up, depending), it lets current flow through on the other side. It takes a while to charge up the capacitor (time constant proportional to Resistance times Capacitance, remember!), but once it's charged there isn't any current (except the leakage current) that flows through.
At least, that's what I recall from my classes. I didn't do so well in the device physics and components classes.
AMD and Intel have gone to great lengths to give their processors deep pipeines. That's a GOOD thing. And they have both gone to great lengths to reduce the penalties associated with deep pipelines (like change-of-flow penalty).
This, in a nutshell, is what computer microarchitecture is. Recent studies have shown that the optimal pipeline depth is very, very long, roughly twice the depth of the P4. So making the pipeline deeper seems to be the right solution.
Total Robberies:
Total Murders:
It seems to me that Americans are much more criminal than Canada. Because Americans have laxer laws for guns, perhaps there are more guns involved. However, there is no causal evidence for this. That's the problem with information like this -- you can show what has happened, but not why it happened. If handguns were outlawed, perhaps crime would increase, as it is less likely that Average Joe would be armed.
No, you license the software that runs on it. The hardware is your property, and you (should be able to) do with it as you please.
CD prices are extaordinarily high given the cost of production. Look at how many CDs AOL pushes out, or how much cheapbytes sells CDs for. If I could get CDs for $1 or $2, which is still 100% more than the cost of production for a big lot of CDs (I don't care about the artwork, just give me the CD in an envelope), I would buy a lot more CDs.
Another example... look at Emusic.com. It's a great service. Fast, reliable. They give me a quality product, I give them what I feel is a reasonable amount of money.
And when record companies work somewhat like cheapbytes, producing bulk CDs, selling for $2, and giving back the artists $1, they won't have money to hype Boy Band of the Month, and will instead focus on lifting up good music that people want to listen to, music that sells itself.
Why do you care about your heirs (or the heirs of mankind) 5 billion years from now? Even if they exist, they don't care about you...
If all there is to your life is "getting off the planet" for some distant descendant, then God help you; your life is meaningless.
Most of your requested features are planned for Segway v2. Not only will these features be added, but Segway will be safe to drive on normal streets. It will truely be a revolutionary thing. Entire countries will be redesigned around Segway v2. Families around the world will come to our office and bow before us, thanking us for the amazing accomplishment we have made.
The initial analysis is that Segway v2 will cost about $320,000, and will have a range of 70km (about 45 miles), or possibly even more. It will be able to run for up to 90 minutes, and when the power is exhausted a simple 18-hour process can be used to power the machine up again.
I think you will agree that $320,000 is a small price to pay for technology that will revolutionize the world.
If you do a google search on optimal pipeline depth you'll find some good results.
And guess what? They're not copy protected.
So go out to the bars that have open mic night or are showing some sort of a local band. And buy their CDs. It's better than the boy-band-of-the-day or whatever crap is playing on your local Clear Channel station.
This is the best part. There's a standard schedule for costs for everyone. You can't brown-nose Microsoft and get special deals, and Microsoft can't punish you by raising your costs for licenses.
You buy licenses at the same rate as everyone else (who buys in the same quantities for the same languages as you do).
This means that if Dell is the largest buyer of MS Licenses, it doesn't matter what they do, they get the cheapest rate. Likewise, any company that buys 2,000 Windows licenses gets the same rate, whether they are MS Fanboys or Sun Microsystems.
Of course, a big "in theory" is added to all this.
Note that your new IBM chip is doing exactly that.
Intel and AMD have repeatedly shown that they can do whatever they like to implement top-notch internal architectures, and lopping on a translation unit only adds 10-20% die area and typically a very small performance hit over a traditional sequential RISC architecture. And they're free to change the internal architecture between revisions. And both Intel and AMD sell enough chips that they can spend a lot of money on designs and make them very good and still turn a profit.
Sometimes things aren't so "the-USA-is-really-bad" as Slashdot says they are (and sometimes they are, and sometimes they're probably worse).
Of course, if we use Law and Order as our legal source (and, though IANAL, I've watched a lot of L&O), then Jack McCoy would say that we have a responsibility to prosecute criminals when their own countries won't, and that as long as an element of the crime was taking place within jurisdiction of the court, the court should have prosecutorial powers. But in the episode where Jack and Carmichael were outside of the Supreme Court and the decision comes out, after attempting to prosecute a foreign diplomat for murder, they (frustratingly) don't tell us what the decision is. D'oh!
However, I would like a place where I could download very high quality, RAW .wav or Ogg Vorbis or MP3 files for, say, $0.50-$1.00 each. Maybe $5.00 for a whole album. From a fast server. That are not in some sort of DRM vault.
This way, I own the music. I can do whatever I like with it: burn it to a CD, put it in my portable player, whatever I want to do within my fair use rights. And I also don't have to (effectively) pay additional money by trying to hunt someone down with the file I want at the quality I want, with a good connection that won't stop halfway through the download.
Merely having the record industry collect money for "allowing" other people to share music peer-to-peer is not sufficient.
I was really glad that it could... finally I could read and write DOS 1.4M floppy disks, Mac 1.4M floppy disks, 800K Mac floppies, and 720K Dos Floppies.
I can't wait to pull out my old 512Ke with the external SuperDrive and start burning DVDs!
Oh, you mean this is just another example of Marketing BuzzWord TermReuse? Crap.
It's a lot better to keep an independant board with their own political agenda who has no accountability to the people, instead of replacing it with a review board whose political agenda matches those of the officials elected by the people?
Does anyone see that maybe this isn't such a horrible idea?
I'm all for impartiality. I just don't think it exists.
Despite what slashdotters will tell you, they aren't going to put oil tycoons on review boards for hospital ethics. But they also might not put people in those review boards who have political agendas that few people want -- like (for instance) the governmentalization of health care.
Doesn't it make some sense to put people in boards who are going to look for solutions inside of the bounds of what the American voting population will tolerate?