I do agree that machines should spit better code than humans do, but we are just not there yet. When will we get there? Maybe tomorrow, maybe next month, maybe next year, maybe next century. Nobody knows. Suppose we do get there tomorrow, I believe it will not be Java + JIT that made it, for the following reasons:
Java is an imperative language, which are very hard to reason about because of the lack of referential transparency. Yes, I know that it is a known fact in computation theory that it is impossible to reason about arbitrary programs, but the fact is that we do not write arbitrary programs. Imperative programs with destructive updates are hard to reason about, and as a result we are almost always restricted to local optimizations. Local optimizations are great, but I am afraid that that alone will not be enough to break the barrier.
Vast amounts of information is lost once the source program is compiled into bytecode, therefore making effective optimization much harder. You can tell the intention of the programmer from the source code, but it is hard to do so for machine instructions (or even preprocessed source code, which anyone who had read the nVidia XFree86 source code must agree). The JIT compiler must create code that acts exactly the same as the bytecode, and that restriction severely reduced its options.
If that breakthrough were to come tomorrow, I believe it is most likely to come from the functional programming people. I believe that is the way to go for the future.
Besides, optimizing compilers are good, but not that good, someone has to write them, and when was the last time that you wrote a program that can solve complex creative problems better than you can?
I haven't done that myself, but some other people had. The FFTW people have created a DFT library written in ANSI C that outperforms nearly all other DFT libraries, many of which are provided by the hardware vendors who hire the best programmers and use assembly code extensively. Most of the code in the library are generated by a program written in OCaml.
And judging from the number of patents and papers in IEEE journals, I believe that creating a fast DFT library is indeed a complex and creative problem.
I'm not sure if I'm reading this right, but I guess in the past your patent term began when you filed the patent? So I guess here you get extra days for every day the patent office is unnecessarily slow in getting it processed. Maybe someone has a better analysis.
I think it was the other way around in the past, ie. the term began when your patent is granted. I remembered reading in Applied Cryptography about the patent stunts pulled out by the NSA and the PTO: the NSA always applies patents for each and every one of their cryptographic inventions, but the USPTO never grants them immediately (so that it need not be disclosed to the public). Then, when somebody outside came up with the same idea and applied for a patent (maybe a good 10 years later), the USPTO reveals the NSA patent, cite it as prior art, grants the NSA patent, and the 17 year clock starts ticking. Real smart, isn't it?
This is just another symptom of the collapse of the educational system in the United States, which the educational establishment is frantically trying to conceal through dumbing down courses and inflating grades. The SAT is a useless as an historical benchmark anymore because a) it is much easier than it ever used to be, and b) on top of this students are now allowed to use calculators on the dumb thing! Sheesh! The stupid test is MULTIPLE CHOICE, for pete's sake!!! Why would you need a CALCULATOR???
Not that it really mattered. For the sake of comparison let's take a look the college admission system in Taiwan (I am a Taiwanese) for example. Basically the whole admission process is centered on the JCEE (Joint College Entrance Examination), so high school graduates take this test on July each year, state their preferences, and then get admitted to different universities and majors by the combination of both.
The JCEE is tough, even by our standards. Sure, we also have multiple choices: you are given five options, one to five of them could be right, and no partial scores (or each wrong option incurs additional penalty so that the expected value is zero, can't remember which). Math includes theorem proving and calculations on elementary arithmetic, geometry, probability, and calculus. Physics includes mechanics, optics, electro-magnetics, and a little relativity and quantum physics. Chemistry includes basic chemical reaction principles, equilibrium, acids/bases reactions, electrochemistry, bonding, and chemical kinetics. All the calculations in phycics and chemistry (equations solving or whatsoever) requires 3-digit accuracy without the use of calculators.
Most students spend a whole year preparing for the exam. They need to be able to recite the order of reduction potentials and electronegavitity of elements, common logarithm values, and important constants in physics. They need to perform complicated symbolic and numeric calculations (you do know that equilibrium constants and atomic weights are not nice integers) fast and free of error. They need to know lots of things that you might not have heard of.
In comparison SAT is easy. I have taken SAT I (November '96) and SAT II (December '96) back in the high school days; nothing can possibly be easier than that. I made no preparations or whatsoever except for skimming through the sample tests provided upon registration. I still remembered the day I took SAT I, about how surprised I was to find the formula of rectangle area on the back of front cover, and how amazed I was to see other students (mostly from Taipei American School) busy typing on their big, Gameboy-like Casio fx-9800 calculators in math sections. Alas, I only used my fx-3600 once that day because I was too lazy to do two-digit multiplies. I even have the score report in my drawer: SAT I verbal 540, math 800, SAT II writing 550, math IIC 800, physics 800.
So what difference does it make? Nothing. Many people simply do not learn, although they have proved that they are capable of learning. They do not care whether they can learn something or not, being able to pass is good enough for them. Some people simply do not show up except for midterms and finals. Cutting classes has become so common that practically no teachers offer classes in early mornings these days. The situation is basically the same over here.
Whether the tests are simple or difficult really doesn't matter. The whole story is not about how much you have learned, it is about how much more do you still want to learn. You can force them to learn, but that will not be of much use in the long run. Without willingness to learn, they are doomed to fail one way or the other, and there is nothing we can do about it.
Slashdot Mirror
Suse supports UDMA/66 installation out of the box.
I do agree that machines should spit better code than humans do, but we are just not there yet. When will we get there? Maybe tomorrow, maybe next month, maybe next year, maybe next century. Nobody knows. Suppose we do get there tomorrow, I believe it will not be Java + JIT that made it, for the following reasons:
If that breakthrough were to come tomorrow, I believe it is most likely to come from the functional programming people. I believe that is the way to go for the future.
Besides, optimizing compilers are good, but not that good, someone has to write them, and when was the last time that you wrote a program that can solve complex creative problems better than you can?
I haven't done that myself, but some other people had. The FFTW people have created a DFT library written in ANSI C that outperforms nearly all other DFT libraries, many of which are provided by the hardware vendors who hire the best programmers and use assembly code extensively. Most of the code in the library are generated by a program written in OCaml.
And judging from the number of patents and papers in IEEE journals, I believe that creating a fast DFT library is indeed a complex and creative problem.
I'm not sure if I'm reading this right, but I guess in the past your patent term began when you filed the patent? So I guess here you get extra days for every day the patent office is unnecessarily slow in getting it processed. Maybe someone has a better analysis.
I think it was the other way around in the past, ie. the term began when your patent is granted. I remembered reading in Applied Cryptography about the patent stunts pulled out by the NSA and the PTO: the NSA always applies patents for each and every one of their cryptographic inventions, but the USPTO never grants them immediately (so that it need not be disclosed to the public). Then, when somebody outside came up with the same idea and applied for a patent (maybe a good 10 years later), the USPTO reveals the NSA patent, cite it as prior art, grants the NSA patent, and the 17 year clock starts ticking. Real smart, isn't it?
This is just another symptom of the collapse of the educational system in the United States, which the educational establishment is frantically trying to conceal through dumbing down courses and inflating grades. The SAT is a useless as an historical benchmark anymore because a) it is much easier than it ever used to be, and b) on top of this students are now allowed to use calculators on the dumb thing! Sheesh! The stupid test is MULTIPLE CHOICE, for pete's sake!!! Why would you need a CALCULATOR???
Not that it really mattered. For the sake of comparison let's take a look the college admission system in Taiwan (I am a Taiwanese) for example. Basically the whole admission process is centered on the JCEE (Joint College Entrance Examination), so high school graduates take this test on July each year, state their preferences, and then get admitted to different universities and majors by the combination of both.
The JCEE is tough, even by our standards. Sure, we also have multiple choices: you are given five options, one to five of them could be right, and no partial scores (or each wrong option incurs additional penalty so that the expected value is zero, can't remember which). Math includes theorem proving and calculations on elementary arithmetic, geometry, probability, and calculus. Physics includes mechanics, optics, electro-magnetics, and a little relativity and quantum physics. Chemistry includes basic chemical reaction principles, equilibrium, acids/bases reactions, electrochemistry, bonding, and chemical kinetics. All the calculations in phycics and chemistry (equations solving or whatsoever) requires 3-digit accuracy without the use of calculators.
Most students spend a whole year preparing for the exam. They need to be able to recite the order of reduction potentials and electronegavitity of elements, common logarithm values, and important constants in physics. They need to perform complicated symbolic and numeric calculations (you do know that equilibrium constants and atomic weights are not nice integers) fast and free of error. They need to know lots of things that you might not have heard of.
In comparison SAT is easy. I have taken SAT I (November '96) and SAT II (December '96) back in the high school days; nothing can possibly be easier than that. I made no preparations or whatsoever except for skimming through the sample tests provided upon registration. I still remembered the day I took SAT I, about how surprised I was to find the formula of rectangle area on the back of front cover, and how amazed I was to see other students (mostly from Taipei American School) busy typing on their big, Gameboy-like Casio fx-9800 calculators in math sections. Alas, I only used my fx-3600 once that day because I was too lazy to do two-digit multiplies. I even have the score report in my drawer: SAT I verbal 540, math 800, SAT II writing 550, math IIC 800, physics 800.
So what difference does it make? Nothing. Many people simply do not learn, although they have proved that they are capable of learning. They do not care whether they can learn something or not, being able to pass is good enough for them. Some people simply do not show up except for midterms and finals. Cutting classes has become so common that practically no teachers offer classes in early mornings these days. The situation is basically the same over here.
Whether the tests are simple or difficult really doesn't matter. The whole story is not about how much you have learned, it is about how much more do you still want to learn. You can force them to learn, but that will not be of much use in the long run. Without willingness to learn, they are doomed to fail one way or the other, and there is nothing we can do about it.