Sorry, I can't call it anything else, even if it's a PhD thesis. Sure, you can argue that copyright infringement is not theft, but then don't call people who abuse GPL software "thieves".
The lower education system was created solely for indoctrination purposes. Unfortunately, it still has that aspect.
History, for example, can be a very controversial subject. Yet they teach the US version of history to kids as 100% true facts. This is not happening just in the US - everybody brainwashes children this way.
The only way for local graduates to be hired will be to offer their services for lesser pay.
If this is really the case, why do we still have any IT jobs in the States? Don't tell me that there aren't any, quite a few of my friends who are very good coders have gotten jobs in the last year (by good coding I don't mean "oh, I know VB/C# and how to write a php script", btw). Sure, the companies that hired them made them jump through hoops when interviewing them, but still...
Being in the states has a major advantage. The money is still made here (after all, the gross national income per capita is still about 100 times larger in the States than in India), so outsourcing is going to be less than ideal in quite a few cases.
For example, "lions can spend as much time as 20 hours per day sleeping" -- wikipedia. At the same time, a gnu antilope (not GNU/Antilope) needs only about 6 hours of sleep per night. And yes, this is because of their diet.
Furthermore, the chain for a carnivore is simply longer by one (plants->herbivores->carnivore)
Memory is so slow these days, that the cost of an L2 miss for can be as high as 400 cycles.
Basically, a load that hits in the L1 can be as fast as 2-3 cycles of latency. If it misses all the way to the main memory, it's ~400cycles. Two orders of magnitude.
I was talking about gcc, and gcc alone. After all, this is a story about gcc 4.0. I'm very well aware that compilers like icc are more capable of exploiting differences between old processors and newer ones. It's just that gcc is not. Please show me one single benchmark -- from a respectable suite, not something concocted to prove a point, and not something that has inlined assembly -- where tuning for 686 in gcc gives more than 5% performance improvement! My very strong belief is that the whole Gentoo "ultra-optimization" is simply a myth, because gcc is incapable of exploiting the differences.
And if you bothered to do a little bit of googling, you'd have figured out that I'm not exactly foreign to computer architecture.
vectorization very rarely works
on
GCC 4.0.0 Released
·
· Score: 4, Insightful
The main problem is the C language. While vectorizing a loop is generally not that difficult, figuring out if it's the right thing to do is extremely tough. To do that, you have to "prove" that iterations of a loop are independent of each other. This, in turn, requires good pointer alias analysis, and gcc isn't doing it well enough yet. BTW... a language like Fortran, that doesn't have pointers at all, is much easier to vectorize; that's one of the reasons a lot of scientific codes are still in Fortran.
Without automatic vectorization, the performance benefit of compiling for 686 as opposed to 386 is simply minimal. A lot of people have done benchmarks on this, and found out that tuning for 686 with gcc only provides 1-2% improvements in the best case. Keep in mind that current X86 processors execute instructions out-of-order, so instruction scheduling for a specific pipeline is not going to do much (it's very important for in-order machines, though)
While Sun hardware is very stable and reliable, their processors just suck. They work well for some type of workloads (webserving, oltp), but for pretty much everything else AMD and Intel chips just kick their asses. Sure, you can scale more with Sun, but in general it's preferable to have a fast chip than multiple chips that are considerably slower. And it's not just clockspeed. Intel/AMD chips are doing out-of-order execution for 3 generations now (PPro, PII/PIII, PIV and K6, K7, K8), Sun -- well, they're still in-order.
Why do you think Sun is doing Opteron servers these days ?
My university, too, is mid-way switching from Sun to Linux. With Sun hardware you pay a premium for a slow product (at least CPU-wise, which, for the kind of stuff university people do, is the most important). Simply not worth it.
The main problem was that the first JVMs from Sun were outright dumb (it wasn't until 1.3 that they actually included a good JIT in it). First impressions are strong.
Current JVMs are reasonably fast. In fact, most performance gap between java and C comes from its safety features.
Safety, however, is very important for web apps. Writing regular sites in C is just asking for C-style memory attacks, like buffer overflows.
Finally, nobody really complains about python's or perl's speed.
To me, EU is like the Polish-Lithuanian Republic - a beautiful concept, eventually brought to its knees by undemocratic technicalities like the Liberum Veto
One big problem with the free market model is that it actually has failure modes. Sure, if the market had perfect competition, and everybody made rational decisions, we wouldn't have these problems. But these assumptions are 99% of the time wrong.
The biggest failure mode of the free market is a monopoly, or an economic singularity, if you will. That's why we have the Sherman anti-trust law - to break such black holes. Well, guess what, it's not working with Microsoft. They will continue, undeterred, to use their existing monopoly to get other monopolies.
IANAQP, but, here's the thing: With his double slit experiment, you can tell which way the photon went, because on receptor A you only get photons from slit B, and the other way around. Nevertheless, the photon still interferes
with its "clone" from the other slit, making interference patterns that have minima (no energy) at the wires, thus bypassing them (that way, the 3rd image is very close to the first one, the small error comes from the fact that the wires aren't zero-width).
If the experiment is right (and is seems to be, the "one photon at a time" validation also worked), it basically destroys the Copenhagen/Multiple Worlds interpretations of Schrodinger's equation. The transactional interpretation still holds, but it's even weirder than Copenhagen/MWI (it relies on waves travelling back in time).
Slashdot Mirror
But the whole point of trackerless P2P is. For legitimate P2P (e.g. downloading FC) you don't need all this.
Sorry, I can't call it anything else, even if it's a PhD thesis. Sure, you can argue that copyright infringement is not theft, but then don't call people who abuse GPL software "thieves".
Yeah, it's a "diet" thing
History, for example, can be a very controversial subject. Yet they teach the US version of history to kids as 100% true facts. This is not happening just in the US - everybody brainwashes children this way.
This is one of the best news I've heard in years.
If this is really the case, why do we still have any IT jobs in the States? Don't tell me that there aren't any, quite a few of my friends who are very good coders have gotten jobs in the last year (by good coding I don't mean "oh, I know VB/C# and how to write a php script", btw). Sure, the companies that hired them made them jump through hoops when interviewing them, but still ...
Being in the states has a major advantage. The money is still made here (after all, the gross national income per capita is still about 100 times larger in the States than in India), so outsourcing is going to be less than ideal in quite a few cases.
i.e., they don't rot, like DVDs do.
This is actually doable with todays' algorithms, with reasonably large keys. Sure, it won't be realtime, but still ...
Yes, I should be modded -1 sexist.
Furthermore, the chain for a carnivore is simply longer by one (plants->herbivores->carnivore)
If you do a poor one, you're gonna have to do it all over again next year (i.e. more money).
Memory is so slow these days, that the cost of an L2 miss for can be as high as 400 cycles. Basically, a load that hits in the L1 can be as fast as 2-3 cycles of latency. If it misses all the way to the main memory, it's ~400cycles. Two orders of magnitude.
Small is fast; large is slow.
Yeap, you're right. The main difference though is that P4 has a trace-cache, so decoding happens only on a miss (for most codes, that's very rarely).
I was talking about gcc, and gcc alone. After all, this is a story about gcc 4.0. I'm very well aware that compilers like icc are more capable of exploiting differences between old processors and newer ones. It's just that gcc is not. Please show me one single benchmark -- from a respectable suite, not something concocted to prove a point, and not something that has inlined assembly -- where tuning for 686 in gcc gives more than 5% performance improvement! My very strong belief is that the whole Gentoo "ultra-optimization" is simply a myth, because gcc is incapable of exploiting the differences.
And if you bothered to do a little bit of googling, you'd have figured out that I'm not exactly foreign to computer architecture.
Without automatic vectorization, the performance benefit of compiling for 686 as opposed to 386 is simply minimal. A lot of people have done benchmarks on this, and found out that tuning for 686 with gcc only provides 1-2% improvements in the best case. Keep in mind that current X86 processors execute instructions out-of-order, so instruction scheduling for a specific pipeline is not going to do much (it's very important for in-order machines, though)
Preventing OSS adoption through patent/indemnization FUD campaigns: priceless
Patents is really the only ace they have right now against OSS.
Why do you think Sun is doing Opteron servers these days ?
My university, too, is mid-way switching from Sun to Linux. With Sun hardware you pay a premium for a slow product (at least CPU-wise, which, for the kind of stuff university people do, is the most important). Simply not worth it.
And that means ... TCPA/Palladium ... bwhahaha
Current JVMs are reasonably fast. In fact, most performance gap between java and C comes from its safety features.
Safety, however, is very important for web apps. Writing regular sites in C is just asking for C-style memory attacks, like buffer overflows.
Finally, nobody really complains about python's or perl's speed.
To me, EU is like the Polish-Lithuanian Republic - a beautiful concept, eventually brought to its knees by undemocratic technicalities like the Liberum Veto
The biggest failure mode of the free market is a monopoly, or an economic singularity, if you will. That's why we have the Sherman anti-trust law - to break such black holes. Well, guess what, it's not working with Microsoft. They will continue, undeterred, to use their existing monopoly to get other monopolies.
The way things evolve, other people might have to use the parent post as prior art
(Sorry, I meant SCOXE)
While not directly saying it, what most employers appreciate is the ability to deliver. Best way to achieve this in college? An opensource project.
IANAQP, but, here's the thing: With his double slit experiment, you can tell which way the photon went, because on receptor A you only get photons from slit B, and the other way around. Nevertheless, the photon still interferes with its "clone" from the other slit, making interference patterns that have minima (no energy) at the wires, thus bypassing them (that way, the 3rd image is very close to the first one, the small error comes from the fact that the wires aren't zero-width).
If the experiment is right (and is seems to be, the "one photon at a time" validation also worked), it basically destroys the Copenhagen/Multiple Worlds interpretations of Schrodinger's equation. The transactional interpretation still holds, but it's even weirder than Copenhagen/MWI (it relies on waves travelling back in time).