Yes, but a user process is not allowed to set the scheduler to real-time priority The reason is that an infinite loop can usually hang the system. Only 2.6.12 (with rt-rlimits) allows a constrained real-time scheduling even to normal users.
Audio doesn't need microseconds, but 1-5 milliseconds is important for some applications. Unfortunately, Linux isn't yet there (dunno whether others are). Even with the latest RT preempt patches, I'm having problems with 1 ms latency. Also, prior to 2.6.12, you couldn't even get real-time priority (required even for 20ms latency) when running as a user (non-root).
better to work on creating something done right, or to object to it on moral grounds?
How about work create a lot of different standards done wrong, so the whole thing gets scrapped once people get frustrated with the stuff just not working.
But if the president's goal of universal, affordable high-speed Internet access by 2007 is to be achieved, policymakers in Washington must change course.
That's a very good idea... in theory. I suspect the implementation would be terrible. Why? The problem is with who decides whether an interface is public or private. If anyone is allowed to change that part of the code, it means that someone who wants to use the private interface can just change it to public, defeating the purpose of the thing. On the other hand, if you allow an author to create a private interface and prevent anyone from making it public, you create a mess worse than of the GFDL (see invariant sections). Not only do you add restrictions to how one can modify the code (possibly making it non-free), but you probably make it incompatible with the current GPl. Also, the possibility of adding public interfaces might even allow people to create a public interface to something you originally intended private (but never bothered to tag as private). I'm sure I'm missing other reasons why this wouldn't work.
You probably couldn't do it. Remember, space is big. You could almost fit another three earths in between us and a geostationary satellite.
Actually, I wasn't refering to geostationary satellites, but the ones on a low orbit (~200 km). That's much smaller and happens to be where most of the junk is. There's satellites going in all kinds of intersecting orbits (tilted, polar). Low orbit is also where most of the military satellites are AFAIK. You can't take very good pictures from geostationary orbit.
That's exactly what I was thinking. Actually, if you go a bit further, you can wonder how many satellites you need to destroy before the debris end up destroying even more satellites create a chain reaction. I'm sure I've seen people research that, but I don't know the result. With the number of satellites up there, I would expect just a few would be enough.
Sure. Except that kernel 2.0 came out early 1996 (or was it late 1995), while XP came out in 2001. It would be more like dumping support for kernel 2.2 (which was still in stable Debian until Sarge was released a few months ago).
If it really works, then it's not just useful "as long as defect rate is high". Think about it, current technology assumes that everything needs to be perfect, but if you can tolerate some defects, then you can be a lot more aggressive in the design. That means using a smaller features, lower voltage, higher clock rate,...
Wouldn't want to disappoint you, but you most likely ARE paying for advertising whenever you buy a product that is shown in an ad (even if you didn't actually see it).
But I agree about your main topic: TV is not an essential component of live (even though I do own one, which I watch ~1 hour a week these days).
If the L1 was as fast as registers, people wouldn't bother putting more registers on their chip. While the L1 can give you a throughput of 1/cycle, this doesn't count latency (~3 cycles on recent Intel chips I think), the fact that at least one x86 instruction has to be a register operant, and the fact that a PPC would *already* be using the L1 for other things.
About renaming, the PPC does it too (so it has even more registers), so you still have less registers. Also, the renaming is mainly there to allow the pipelines to work correctly. You still only have eight "logical" registers to put stuff in.
Last thing, AltiVec must be pretty hard to do efficiently on x86. First, it does twice the amount of computation as SSE does per cycle, but also because it does a MAC, which would have a 9-cycle latency if implemented in SSE.
I really doubt they can run PPC code on an x86 faster than it is on the PPC. Not only are new PPC close to x86 at native code, but the translation isn't easy at all. I could see a PPC doing a decent job at x86 emulation, but for the reverse there's a problem: registers. If you have a piece of PPC code that uses more registers than the x86 has (I expect this is true of any decent code), then you need to replace registers by memory (L1 at least) accesses. That will cost a lot.
What? Are you somehow implying that the US is not the greatest democracy in the world? You must be some kind of old Europe commie!
What has always amazed me is not that much the fact that group X dumps a pile of money on politician Y (that happens everywhere), but the fact that in the US you can do it without even hiding. "We don't have any corruption here, our politicians just obey free market rules."
(Note: I am Canadian and the funding of political parties are, at least in theory, strictly controlled here)
You do realize that in order to have something look the size of the moon (still rather small) at geo-synch orbit, it would have to be one tenth of the size of the moon! That's pretty huge. The only way you can make something visible on Earth is at low orbit.
Slashdot Mirror
Yes, but a user process is not allowed to set the scheduler to real-time priority The reason is that an infinite loop can usually hang the system. Only 2.6.12 (with rt-rlimits) allows a constrained real-time scheduling even to normal users.
It says 5 microseconds, while the clock period is around 500 ps. That's 10,000 cycles, not 10 million.
Audio doesn't need microseconds, but 1-5 milliseconds is important for some applications. Unfortunately, Linux isn't yet there (dunno whether others are). Even with the latest RT preempt patches, I'm having problems with 1 ms latency. Also, prior to 2.6.12, you couldn't even get real-time priority (required even for 20ms latency) when running as a user (non-root).
better to work on creating something done right, or to object to it on moral grounds?
How about work create a lot of different standards done wrong, so the whole thing gets scrapped once people get frustrated with the stuff just not working.
But if the president's goal of universal, affordable high-speed Internet access by 2007 is to be achieved, policymakers in Washington must change course.
Nah, just redefine "universal".
Guess what. It's survival of the fittest.
Wow! It's the best reason I've ever heard for killing all other species on earth.
That's a very good idea... in theory. I suspect the implementation would be terrible. Why? The problem is with who decides whether an interface is public or private. If anyone is allowed to change that part of the code, it means that someone who wants to use the private interface can just change it to public, defeating the purpose of the thing. On the other hand, if you allow an author to create a private interface and prevent anyone from making it public, you create a mess worse than of the GFDL (see invariant sections). Not only do you add restrictions to how one can modify the code (possibly making it non-free), but you probably make it incompatible with the current GPl. Also, the possibility of adding public interfaces might even allow people to create a public interface to something you originally intended private (but never bothered to tag as private). I'm sure I'm missing other reasons why this wouldn't work.
You probably couldn't do it. Remember, space is big. You could almost fit another three earths in between us and a geostationary satellite.
Actually, I wasn't refering to geostationary satellites, but the ones on a low orbit (~200 km). That's much smaller and happens to be where most of the junk is. There's satellites going in all kinds of intersecting orbits (tilted, polar). Low orbit is also where most of the military satellites are AFAIK. You can't take very good pictures from geostationary orbit.
That's exactly what I was thinking. Actually, if you go a bit further, you can wonder how many satellites you need to destroy before the debris end up destroying even more satellites create a chain reaction. I'm sure I've seen people research that, but I don't know the result. With the number of satellites up there, I would expect just a few would be enough.
Thanks. Can you make one for me while you're at it?
It's a minor problem, which ordinarily would have gone unnoticed.
You mean stuff like slightly damaged booster seals and small pieces of foam hitting the wing that happens all the time?
Nevertheless, I expect that Google will obey, because the consequences of getting caught not doing it can be dire.
Not quite. The contract is between MS and the former employee. Google isn't bound by anything. The employee might get in trouble though.
No, the US only bombs countries *after* they made sure that said country doesn't have WMDs.
Who here run pre-2.0 kernel on their Linux boxes?
Sure. Except that kernel 2.0 came out early 1996 (or was it late 1995), while XP came out in 2001. It would be more like dumping support for kernel 2.2 (which was still in stable Debian until Sarge was released a few months ago).
That's exactly what I was thinking. Maybe if they use slower transistors or decrease the density?
If it really works, then it's not just useful "as long as defect rate is high". Think about it, current technology assumes that everything needs to be perfect, but if you can tolerate some defects, then you can be a lot more aggressive in the design. That means using a smaller features, lower voltage, higher clock rate, ...
We save lots of money not paying for advertising.
Wouldn't want to disappoint you, but you most likely ARE paying for advertising whenever you buy a product that is shown in an ad (even if you didn't actually see it).
But I agree about your main topic: TV is not an essential component of live (even though I do own one, which I watch ~1 hour a week these days).
I don't see how the contraption can both be small and deliver at a high pressure while operating off of one battery.
Because you're already at that pressure, any device will produce O2 at that pressure. It would actually be *harder* to get it atmospheric pressure.
Also, now that I think about it, I think the US navy has some pure O2 underwater low depth breathing rigs like this.
I don't think anyone uses pure O2. When going past a certain dept, I think it's mainly a O2 + Helium mix, hence divers sounding like Donard Duck.
If the L1 was as fast as registers, people wouldn't bother putting more registers on their chip. While the L1 can give you a throughput of 1/cycle, this doesn't count latency (~3 cycles on recent Intel chips I think), the fact that at least one x86 instruction has to be a register operant, and the fact that a PPC would *already* be using the L1 for other things.
About renaming, the PPC does it too (so it has even more registers), so you still have less registers. Also, the renaming is mainly there to allow the pipelines to work correctly. You still only have eight "logical" registers to put stuff in.
Last thing, AltiVec must be pretty hard to do efficiently on x86. First, it does twice the amount of computation as SSE does per cycle, but also because it does a MAC, which would have a 9-cycle latency if implemented in SSE.
I really doubt they can run PPC code on an x86 faster than it is on the PPC. Not only are new PPC close to x86 at native code, but the translation isn't easy at all. I could see a PPC doing a decent job at x86 emulation, but for the reverse there's a problem: registers. If you have a piece of PPC code that uses more registers than the x86 has (I expect this is true of any decent code), then you need to replace registers by memory (L1 at least) accesses. That will cost a lot.
someone could mention that terrorists are applying for a lot of patents too?
What? Are you somehow implying that the US is not the greatest democracy in the world? You must be some kind of old Europe commie!
What has always amazed me is not that much the fact that group X dumps a pile of money on politician Y (that happens everywhere), but the fact that in the US you can do it without even hiding. "We don't have any corruption here, our politicians just obey free market rules."
(Note: I am Canadian and the funding of political parties are, at least in theory, strictly controlled here)
Yup. I can confirm seeing that for the Developers headlines yesterday.
One thing you seem to forget is that we'd also need much more robust displays. Current displays won't stand being dropped or kicked.
You do realize that in order to have something look the size of the moon (still rather small) at geo-synch orbit, it would have to be one tenth of the size of the moon! That's pretty huge. The only way you can make something visible on Earth is at low orbit.