What good is a new chip, no matter how fast it is, if you can't run anything on it?
There is this really neat group of operating systems called Unix/Linuxes. They have a major advantage in that you only need a small amount of assembler to get going on a new chip, then the rest can be ported over in C/C++. This has been the situation for decades - Unix (and now Linux) has been the initial OS for almost all new chips.
How fast will this chip be at general purpose stuff? Who cares if it can do 100GFLOPS on a couple operations.
Reasonable point, but FLOPs are a good general measure of the speed, as they are pretty complex operations. We all used to measure speed in MIPS (Million Instructions Per Second), but as chips got so diverse, one chip's instruction could not be easily compared with another's (particularly if RISC chips were involved, where the instructions could be very minimal). FLOPs are a better measure, as a divide is a divide and a multiply is a multiply no matter what chip architecture you use.
Otherwise it is all fake, serving this or that special groop
Science is ALL fake? You obviously have no understanding of science at all. Science is about what works. Its about what can be proven, and about what is useful long-term. Interpretations can be fake, or serving special groups, but in the long term, fake science dies as it is simply of no use.
Also, many of my friends who are working in science to actually help people by studying disease (and devising treatments) would probably be very angry at your thoughtless and ignorant comment.
Why in the world would one consider the *possibility* of unsafe code in the CLR a security hole?
Because of the way in which many coders work..Net makes the use of raw memory access and pointers easy - just use a keyword. It can make the use of potentially insecure legacy code embedded in.Net routine. Its a classic example of Microsoft designing for the convenience of the developer and not for security.
Enriching uranium to get this mass is very tricky, as the two isotopes of uranium have very little difference in mass.
assemble it in a way that they'll be pushed together again by the conv. explosion
That is hard. You need to assemble a critical mass with the right geometry and at the right speed or else you will end up with either nothing or a nasty hot radioactive mess but no large explosion.
On the other hand, they are in a position to tell us what meaning people have in the past inferred from their observations of the universe--which ideas are widespread, which are not, criticisms which have been levied against one or another, etc.. They study these things in seminary.
This is only for some priests and only for a very few religions.
The hugely more common situation is for the leaders of religions to be instructing humanity how they should relate to life and the cosmos, often with severe condemnation of those who disagree with their position.
"We have no way of telling that a massive release of gas on Mars would not eventually come back to haunt us here on Earth."
We do. We are constantly bombarded by massive releases of gas and particles from the Sun, and it has only very minor effects.
Also, there is the not-so-little matter of the gravity of Mars - any gas we release there is not going to drift off in bulk and effect us.
Perhaps someone could honestly explain why the above post, which I hoped was calmly correcting what I felt was a mistaken comment using facts and gentle humour, was rated 'Troll'?
I try to avoid being offensive on Slashdot. The advice would be useful so I could avoid doing this again.
Having a colony on Mars would not save the human race were the sun to be destroyed, for example.
So if a colony on Mars won't protect us from ALL disasters, only some, its not worth it?
Asteroid collisions and vulcanism are very common on our planet. Disasters happen every few hundreds of thousands of years. We can surely learn to deal with whatever happens to the Sun on a timescale of millions of years. It would be neat to survive until then.
More importantly, what about the uncertainties? Like nuclear war? Worldwide plague? Asteroid strike?
And, of course, mass vulcanism. That nearly destroyed all life about 250 million years ago, and can't be easily stopped (like an asterioid).
The fact is, Earth is a single point of failure for the human race, and we can't predict when it will fail or what will cause the failure. The only safe solution is redundancy. Terraforming Mars is the only remotely feasible option in the near future.
I give up. Please provide me with the scientific explanations of what man is and what the universe is.
Man is an ape - the result of 4 billion years of evolution.
The Universe is more difficult. At least science gives some idea of the properties of the universe - the particles that make it up, its size and composition.
At least religion, rooted in fact or not, makes some effort to explain what these things are and why they exist.
Making an effort is pointless if you have no idea what you are trying to explain. An example is the milky way - one religion has said that this is the milk of a goddess, and explained why this milk was split. The problem is that the milky was is not the milk of a goddess, its 100 billion stars. Look at how religion tried to suppress the truth of religion and how the Earth goes around the Sun.
As for why the Universe was created, why should a priest have any better answer that, say, a plumber or policeman? On what basis (other than irrational faith) can a priest claim authority and knowledge?
If you don't give a shit about security (closed system) and want to sqeeze every cycle from the cpu C (with a bit of inline ASM if need be) is stronger than Java.
A few years ago I would definitely have agreed, but now I have changed my mind. It seems that the Sun Java Hotspot system is basically an expert system for generating machine code for whatever processor you are on. It does a phenomenal amount of work profiling and fine-tuning the generated machine code. I honestly think that in a few years it will be able to do far better than any pre-compiled C with hand-coded assembler added.
Another plus for C is the defacto-standard thing, for whatever reason windows & *nix are written in C.
Heh, things are definitely better these days, but when I was coding C even 10 years ago, the amount of conditional code I had to include to deal with processors and compilers was considerable. (I even had problems with different versions of gcc).
The Java OS was a proof of concept, and probably not practical, but I really think the way forward is virtual machines almost everywhere.. eventually!
just because he is a scientist, he does not have to be competent to pass judgements about everything in the universe.
Well, actually, being a scientist pretty much does make you competent, as you have experience of judging evidence and discussing ideas rationally. For example, one of the things he most objects to is religious leaders saying that they have special qualifications to discuss matters such as cosmology, biology and the origin of the Universe etc. I totally support him in this. In a world where there is increasing religious influence and attempts to protect religious views from critical thinking, I applaud Dawkins as someone brave enough to speak out for rationality.
Shakespeare is not an idiot because he did not pass through the peer review process and got his works published without them.
Has Dawkins commented on Shakespeare?
I believe that art, religion and science are all equally valid ways that man uses to understand his place in the universe.
How can they possibly be equally valid? Without science we would have virtually no idea of what man is or what the universe is. Unlike religion, science progresses and expands its understanding.
Slander of other disciplines takes us nowhere.
I don't think Dawkins has slandered other disciplines, simply asked that they are measured by the same standard as scientific ideas.
You don't believe that all publishers out there have a lot of money to spend on the army of experts you suggest, do you?
If they don't, then they are not a quality academic publisher, and don't deserve that reputation.
narrowmindedly decline to publish because this is not your field of expertise
A journal chooses the field of expertise it is intending to publish. It is not narrowminded to decline to publish - it is intellectual honesty.
Sokal's attack was so successful because it highlighted the already published work which often included scientific nonsense, written by authors who were ignorant of the terms they were using. Sokal was not ignorant of the math and physics terms he included, so he was in an unassailable position to declare his paper nonsense. The damaging 'payload' of his paper was the false science he included. If he had simply written some arbitrary post-modern analysis, the editors could have simply responded that he did not understand the full meaning of his own work. They could not refute his understanding of the science.
posturing by the artists or that of the loud-mouthed scientists like Richard Dawson.
I would be interested to know how clear thinking about Darwinism by Dawkins - one of the most successful promoters of rationality - can be labelled 'posturing'. It it is, I wish there were a lot more!
Java is much quicker than a P-CODE interpreter but it still takes time to do stuff at run-time, native apps do this at compile time. Any "security" checking, no matter what the language, will eat up instructions.
Yes, but the way most modern Java systems work is to run the compilation as a background thread. This allows the program to get started immediately and then this thread performs tasks such as code analysis to allow security checking to be optimised out in many cases, then it profiles the application to produce very highly optimised machine code (even to the extent of re-ordering instructions to suit the pipeline structure of specific processors). The whole system has been very highly tuned to avoid noticeable delay for the user.
why not take a look at Sun's Java site and check out the white papers on Hotspot technology? You'll see its a pretty awesome process.
HoHoHeHeHaHa, you seriously think we should write operating systems and device drivers in Java?
It has already been done: JJos.
Java is just a language. There is no reason why it shouldn't be used almost everywhere that C is. It doesn't even have to be interpreted - there are plenty of compilers around if that is what you prefer.
So since such detailed knowledge is virtually impossible when your end users are all running different CPUs, are you actually saying that assembly langauge is slow?
Of course it can be. It is very easy to code ineffeciently in assembler! On some processors, with multiple pipelines and complex register sets it can be very hard to write efficient assembler.
Ahem. Do you mean to imply that Java is faster than assembler? That a good Java program can execute the same task faster than an assembly program? And I am not comparing code written by a good Java programmer and a bad assembly programmer. I guess that is what you are saying. I find that difficult to believe.
Why? It can be very hard to produce effective fast assembler on today's complex processors. The Sun Java Hotspot system includes features such as automatic instruction scheduling, re-organising machine code instructions for improved performance on specific processors. This requires no effort from the programmer.
I am not very familiar with Java. I have avoided it due to its reputation for inefficiency. But I have studied.NET a bit and found (IIRC) that it does not actually produce native code per se. That it is still essentially an interpreted language.
No..Net produces native code at runtime via a JIT (Just In Time) compiler. It is not interpreted.
If Java is fully compiled (to machine code) at runtime, then the only performance penalty (in theory) would be waiting for it to compile before you can run it which for large programs could be significant of course. If this is how it works it is kind of an interesting compromise. While the user may have to wait 10 minutes for the program to compile before he uses it, in some cases the increased machine code optimisations could be worth it if the program is run for long periods.
What happens with Hotspot is roughly like this (the exact process depends on switches than can be supplied to the VM): The program starts interpreted, then as it runs background processes determine the speed-critical sections and translate them to highly optimised machine code, including things such as loop unrolling and method inlining where appropriate. This way there is no wait for compilation when the application is started. This profiling and compiling is an on-going process, continously fine-tuning the performance of the application.
And it was just as true then. The truth is that it is much easier to write fast applications in C or assembly than in these "advanced" new languages that try to make programming into fingerpainting.
I think that time is long past. To write fast code in assembler you would need to have detailed knowledge of the register set and pipelining of the specific processor you are using. The same goes for C - you have to tell the compiler all kinds of information about the processor - the type, word length etc., and even then the compiler can only make a good guess at how to optimise the code. The advantage of the Java Hotspot (as used in Sun's VMs) approach is that the compiler can produce native code for the specific processor you are using at the time you run the application, and then the machine code can be optimised by profiling as the application runs. In some cases, this can provide performance faster than C. As this kind of run-time optimisation gets better that boost in performance over the traditional approach should become routine.
Yeah, whatever dude. Say what you like, but the bottom line is that all the Java apps I use are slow. This was true five years ago when people were prattling on about how Java wasn't slow, and it's still true today.
Substitute 'C++' for 'Java' and this is exactly what I was hearing 20 years ago, when all 'true coders' used assembler.
Java is much more sophisticated than the "P-Code" that was used in the 70's but the basic principle is the same.
No it isn't. The P-Code for Java is the way that Java apps are distributed, but on almost all VMs it is translated to pure and optimised native code. This is nothing like the 70s situation at all.
The vast majority of code that goes into Win/nix O/S's is C because in most cases it is a fast and flexible as assembly.
That is not the reason at all! The vast majority of this code is in C for historical reasons. Truly high performance Java has only been around for a few years, so it would be a bit late to write Windows and Linux in it!
Java is excellent for writing secure distributed apps fast, it trades speed to do this! Lets see a Java VM written in Java and see how fast it runs Java apps?
There is no connection between security and speed. Matters of security (range checking and byte code validation) can usually be done once for most code then optimised away at run time.
As for the Java VM in Java - IBM has already done this, and it works fine!
If you can solve your problem in an hour anyway, I dont think its worth the time to have a grid computer do it. You might as well just run it on your own system, however big.
But if you can solve your problem in an hour on a grid, what is the point of having your own system for the rest of the time?
Slashdot Mirror
What good is a new chip, no matter how fast it is, if you can't run anything on it?
There is this really neat group of operating systems called Unix/Linuxes. They have a major advantage in that you only need a small amount of assembler to get going on a new chip, then the rest can be ported over in C/C++. This has been the situation for decades - Unix (and now Linux) has been the initial OS for almost all new chips.
How fast will this chip be at general purpose stuff? Who cares if it can do 100GFLOPS on a couple operations.
Reasonable point, but FLOPs are a good general measure of the speed, as they are pretty complex operations. We all used to measure speed in MIPS (Million Instructions Per Second), but as chips got so diverse, one chip's instruction could not be easily compared with another's (particularly if RISC chips were involved, where the instructions could be very minimal). FLOPs are a better measure, as a divide is a divide and a multiply is a multiply no matter what chip architecture you use.
Otherwise it is all fake, serving this or that special groop
Science is ALL fake? You obviously have no understanding of science at all. Science is about what works. Its about what can be proven, and about what is useful long-term. Interpretations can be fake, or serving special groups, but in the long term, fake science dies as it is simply of no use.
Also, many of my friends who are working in science to actually help people by studying disease (and devising treatments) would probably be very angry at your thoughtless and ignorant comment.
Why in the world would one consider the *possibility* of unsafe code in the CLR a security hole?
.Net makes the use of raw memory access and pointers easy - just use a keyword. It can make the use of potentially insecure legacy code embedded in .Net routine. Its a classic example of Microsoft designing for the convenience of the developer and not for security.
Because of the way in which many coders work.
It is easy!
No it isn't.
You just need a critical mass of U238
Enriching uranium to get this mass is very tricky, as the two isotopes of uranium have very little difference in mass.
assemble it in a way that they'll be pushed together again by the conv. explosion
That is hard. You need to assemble a critical mass with the right geometry and at the right speed or else you will end up with either nothing or a nasty hot radioactive mess but no large explosion.
On the other hand, they are in a position to tell us what meaning people have in the past inferred from their observations of the universe--which ideas are widespread, which are not, criticisms which have been levied against one or another, etc.. They study these things in seminary.
This is only for some priests and only for a very few religions.
The hugely more common situation is for the leaders of religions to be instructing humanity how they should relate to life and the cosmos, often with severe condemnation of those who disagree with their position.
"We have no way of telling that a massive release of gas on Mars would not eventually come back to haunt us here on Earth."
We do. We are constantly bombarded by massive releases of gas and particles from the Sun, and it has only very minor effects.
Also, there is the not-so-little matter of the gravity of Mars - any gas we release there is not going to drift off in bulk and effect us.
Perhaps someone could honestly explain why the above post, which I hoped was calmly correcting what I felt was a mistaken comment using facts and gentle humour, was rated 'Troll'?
I try to avoid being offensive on Slashdot. The advice would be useful so I could avoid doing this again.
Having a colony on Mars would not save the human race were the sun to be destroyed, for example.
So if a colony on Mars won't protect us from ALL disasters, only some, its not worth it?
Asteroid collisions and vulcanism are very common on our planet. Disasters happen every few hundreds of thousands of years. We can surely learn to deal with whatever happens to the Sun on a timescale of millions of years. It would be neat to survive until then.
More importantly, what about the uncertainties? Like nuclear war? Worldwide plague? Asteroid strike?
And, of course, mass vulcanism. That nearly destroyed all life about 250 million years ago, and can't be easily stopped (like an asterioid).
The fact is, Earth is a single point of failure for the human race, and we can't predict when it will fail or what will cause the failure. The only safe solution is redundancy. Terraforming Mars is the only remotely feasible option in the near future.
Perfectly expressed.
We have no way of telling that a massive release of gas on Mars would not eventually come back to haunt us here on Earth.
We do. We are constantly bombarded by massive releases of gas and particles from the Sun, and it has only very minor effects.
Also, there is the not-so-little matter of the gravity of Mars - any gas we release there is not going to drift off in bulk and effect us.
I give up. Please provide me with the scientific explanations of what man is and what the universe is.
Man is an ape - the result of 4 billion years of evolution.
The Universe is more difficult. At least science gives some idea of the properties of the universe - the particles that make it up, its size and composition.
At least religion, rooted in fact or not, makes some effort to explain what these things are and why they exist.
Making an effort is pointless if you have no idea what you are trying to explain. An example is the milky way - one religion has said that this is the milk of a goddess, and explained why this milk was split. The problem is that the milky was is not the milk of a goddess, its 100 billion stars. Look at how religion tried to suppress the truth of religion and how the Earth goes around the Sun.
As for why the Universe was created, why should a priest have any better answer that, say, a plumber or policeman? On what basis (other than irrational faith) can a priest claim authority and knowledge?
If you don't give a shit about security (closed system) and want to sqeeze every cycle from the cpu C (with a bit of inline ASM if need be) is stronger than Java.
A few years ago I would definitely have agreed, but now I have changed my mind. It seems that the Sun Java Hotspot system is basically an expert system for generating machine code for whatever processor you are on. It does a phenomenal amount of work profiling and fine-tuning the generated machine code. I honestly think that in a few years it will be able to do far better than any pre-compiled C with hand-coded assembler added.
Another plus for C is the defacto-standard thing, for whatever reason windows & *nix are written in C.
Heh, things are definitely better these days, but when I was coding C even 10 years ago, the amount of conditional code I had to include to deal with processors and compilers was considerable. (I even had problems with different versions of gcc).
The Java OS was a proof of concept, and probably not practical, but I really think the way forward is virtual machines almost everywhere.. eventually!
just because he is a scientist, he does not have to be competent to pass judgements about everything in the universe.
Well, actually, being a scientist pretty much does make you competent, as you have experience of judging evidence and discussing ideas rationally. For example, one of the things he most objects to is religious leaders saying that they have special qualifications to discuss matters such as cosmology, biology and the origin of the Universe etc. I totally support him in this. In a world where there is increasing religious influence and attempts to protect religious views from critical thinking, I applaud Dawkins as someone brave enough to speak out for rationality.
Shakespeare is not an idiot because he did not pass through the peer review process and got his works published without them.
Has Dawkins commented on Shakespeare?
I believe that art, religion and science are all equally valid ways that man uses to understand his place in the universe.
How can they possibly be equally valid? Without science we would have virtually no idea of what man is or what the universe is. Unlike religion, science progresses and expands its understanding.
Slander of other disciplines takes us nowhere.
I don't think Dawkins has slandered other disciplines, simply asked that they are measured by the same standard as scientific ideas.
You don't believe that all publishers out there have a lot of money to spend on the army of experts you suggest, do you?
If they don't, then they are not a quality academic publisher, and don't deserve that reputation.
narrowmindedly decline to publish because this is not your field of expertise
A journal chooses the field of expertise it is intending to publish. It is not narrowminded to decline to publish - it is intellectual honesty.
Sokal's attack was so successful because it highlighted the already published work which often included scientific nonsense, written by authors who were ignorant of the terms they were using.
Sokal was not ignorant of the math and physics terms he included, so he was in an unassailable position to declare his paper nonsense. The damaging 'payload' of his paper was the false science he included. If he had simply written some arbitrary post-modern analysis, the editors could have simply responded that he did not understand the full meaning of his own work. They could not refute his understanding of the science.
posturing by the artists or that of the loud-mouthed scientists like Richard Dawson.
I would be interested to know how clear thinking about Darwinism by Dawkins - one of the most successful promoters of rationality - can be labelled 'posturing'. It it is, I wish there were a lot more!
Incompetence leads to trouble in any language. ....
- churn up 2GB swap in java (or something similar)
Yes but some languages limit the trouble. Your example of churning up swap in Java is wrong because you can set a limit for memory use with the VM.
Java is much quicker than a P-CODE interpreter but it still takes time to do stuff at run-time, native apps do this at compile time. Any "security" checking, no matter what the language, will eat up instructions.
Yes, but the way most modern Java systems work is to run the compilation as a background thread. This allows the program to get started immediately and then this thread performs tasks such as code analysis to allow security checking to be optimised out in many cases, then it profiles the application to produce very highly optimised machine code (even to the extent of re-ordering instructions to suit the pipeline structure of specific processors). The whole system has been very highly tuned to avoid noticeable delay for the user.
why not take a look at Sun's Java site and check out the white papers on Hotspot technology? You'll see its a pretty awesome process.
HoHoHeHeHaHa, you seriously think we should write operating systems and device drivers in Java?
It has already been done: JJos.
Java is just a language. There is no reason why it shouldn't be used almost everywhere that C is. It doesn't even have to be interpreted - there are plenty of compilers around if that is what you prefer.
Try running WindowsXP on a 286. Then you shall see the light.
How does this impossibility relate to the speed of Java?
So since such detailed knowledge is virtually impossible when your end users are all running different CPUs, are you actually saying that assembly langauge is slow?
.NET a bit and found (IIRC) that it does not actually produce native code per se. That it is still essentially an interpreted language.
.Net produces native code at runtime via a JIT (Just In Time) compiler. It is not interpreted.
Of course it can be. It is very easy to code ineffeciently in assembler! On some processors, with multiple pipelines and complex register sets it can be very hard to write efficient assembler.
Ahem. Do you mean to imply that Java is faster than assembler? That a good Java program can execute the same task faster than an assembly program? And I am not comparing code written by a good Java programmer and a bad assembly programmer. I guess that is what you are saying. I find that difficult to believe.
Why? It can be very hard to produce effective fast assembler on today's complex processors. The Sun Java Hotspot system includes features such as automatic instruction scheduling, re-organising machine code instructions for improved performance on specific processors. This requires no effort from the programmer.
I am not very familiar with Java. I have avoided it due to its reputation for inefficiency. But I have studied
No.
If Java is fully compiled (to machine code) at runtime, then the only performance penalty (in theory) would be waiting for it to compile before you can run it which for large programs could be significant of course. If this is how it works it is kind of an interesting compromise. While the user may have to wait 10 minutes for the program to compile before he uses it, in some cases the increased machine code optimisations could be worth it if the program is run for long periods.
What happens with Hotspot is roughly like this (the exact process depends on switches than can be supplied to the VM): The program starts interpreted, then as it runs background processes determine the speed-critical sections and translate them to highly optimised machine code, including things such as loop unrolling and method inlining where appropriate. This way there is no wait for compilation when the application is started. This profiling and compiling is an on-going process, continously fine-tuning the performance of the application.
And it was just as true then. The truth is that it is much easier to write fast applications in C or assembly than in these "advanced" new languages that try to make programming into fingerpainting.
I think that time is long past. To write fast code in assembler you would need to have detailed knowledge of the register set and pipelining of the specific processor you are using. The same goes for C - you have to tell the compiler all kinds of information about the processor - the type, word length etc., and even then the compiler can only make a good guess at how to optimise the code. The advantage of the Java Hotspot (as used in Sun's VMs) approach is that the compiler can produce native code for the specific processor you are using at the time you run the application, and then the machine code can be optimised by profiling as the application runs. In some cases, this can provide performance faster than C. As this kind of run-time optimisation gets better that boost in performance over the traditional approach should become routine.
Didn't know that it could do that instantaneously.
It doesn't, but why does this matter?
Yeah, whatever dude. Say what you like, but the bottom line is that all the Java apps I use are slow. This was true five years ago when people were prattling on about how Java wasn't slow, and it's still true today.
Substitute 'C++' for 'Java' and this is exactly what I was hearing 20 years ago, when all 'true coders' used assembler.
Java is much more sophisticated than the "P-Code" that was used in the 70's but the basic principle is the same.
No it isn't. The P-Code for Java is the way that Java apps are distributed, but on almost all VMs it is translated to pure and optimised native code. This is nothing like the 70s situation at all.
The vast majority of code that goes into Win/nix O/S's is C because in most cases it is a fast and flexible as assembly.
That is not the reason at all! The vast majority of this code is in C for historical reasons. Truly high performance Java has only been around for a few years, so it would be a bit late to write Windows and Linux in it!
Java is excellent for writing secure distributed apps fast, it trades speed to do this! Lets see a Java VM written in Java and see how fast it runs Java apps?
There is no connection between security and speed. Matters of security (range checking and byte code validation) can usually be done once for most code then optimised away at run time.
As for the Java VM in Java - IBM has already done this, and it works fine!
Wrong. Java is slow no matter who codes it.
Nonsense. Java is translated to native code that is profiled and optimised at run-time.
If you can solve your problem in an hour anyway, I dont think its worth the time to have a grid computer do it. You might as well just run it on your own system, however big.
But if you can solve your problem in an hour on a grid, what is the point of having your own system for the rest of the time?
[Sorry buddy, but I already have a patent on the process of threatening someone with lawyers in order to get money from something.]
And I'm sure you are aware that you are infringing my patent on threats in italics.