No one is going to throw away everything they have to switch over any time soon.
The interesting part is that one can change architectures when building a Beowulf cluster - most of the codes are compiled from scratch anyway and performance per dollar is all that matters.
Nevertheless, for the past N years most of these were x86 based - this year the winner is, of course, Opteron.
This would *NEVER* happen in the field. Firstly, in any modern OS the process would have been pre-empted long before any problem could occur (causing other instructions to run and hence stopping the overheating). Secondly, no real-world program would ever do this sort of thing as there would always be a comparison in the loop within the timeframe.
I am not so sure. The TFA said millions of instructions and the chips are capable of billions.
So with HZ=100 there is room enough for 28e6 instructions to be executed - plenty enough to trigger the problem.
Also, the TFA mentioned specifically FPU, so it is possible that an integer loop counter gets processed fast enough and in parallel so that FPU is always loaded.
In particular, suppose you have an array of 10 million floats and you want to find the mean - would this trigger the bug ?
I do agree that it is very nice that they checked and announced the problem.
Ok, I don't know for sure about 2.6.x, but I am certain that Linux 2.4.x used COW for fork.
What the discussion is about is a method to do zero-copy I/O - i.e. share buffers between userland and kernel.
Apparently, one could try to make a hack where the buffer is marked COW, so that when userland starts to write into it before the I/O finishes it is copied.
The alternative (which I find quite sensible) is to simply communicate to the application when the kernel is done with the buffer.
The attitude is prevalent, and ignores completely the service aspect of providing a positive customer experience.
I want to point out that for many the reason they are working on a project is *NOT* to "provide positive customer experience", but to celebrate skill and experience of writing complicated software.
In this case, fellow developers are welcomed most, people who want to learn about the software are welcomed second and "users" who want a canned answer
as if they are trying to cheat on a multiple-choice exam are listened to last - and get really annoying if they insist it is their due.
Remember, that when you deal with most corporations you don't even get a chance to talk to the developers. So be polite by recognizing what people value and if you think that you can do a better job - please go ahead. The source is availaible.
she couldn't open some of the attachements. some were.wmv and.ppt files and some were.exe
I heard someplace that Microsoft Office Mac won't open files created with the Windows version and vice versa.
I suggest you tell her to try OpenOffice - it will open most Microsoft formats (I did not have any fail on me yet) and one can then use OpenOffice to save in either Windows and Mac flavor.
As for *.exe - if these are just archives there some utilities for extracting them, however, AFAIK, it is not a good idea to be sending exe files around - what if they get infected ?
The computer students I met there learnt *everything* voluntarily outside of class. They generally seemed competent.
The economics students were toss. Seriously. I cannot begin to describe it. I met several, and tried discussing with them. This is probably partly due to a much lower general awareness of politics and economics (not that I blame them, 70 years of disagreement being fatal *will* affect a society)
I spoke better English than the Language students. I wrote better English than the English teachers.
I did not spend much time with Math/Physics students though.
Ahh, that explains it. The economics department was not the place where one learns business (at least between 1990-1995), but rather where one learned the science of economics, in whatever form it survived under the communists. So not only was it non-practical to begin with, but also, as you point out, got seriously screwed up by the government.
The people who wanted to learn business enrolled in outside classes - there was some sort of 2-year course which was also a good opportunity to go study in Brandeis for a few years.
As far as language, there was a standing joke that there are three branches of English language: American English, British English and Moscow English.
At some point the government took part in how English is taught as well (especially with the need to translate communist ideology), on top of which being isolated from native speakers did not help much.
On the other hand, I had an excellent English teacher through my years of math classes and I seriously doubt she had any problems with pronunciation, but, as you can guess from her teaching math students, people like that did not get easily into any departments government kept a close watch on.
Lastly, while I can rant for hours about things that could be improved (for example, they did not show us any algebraic geometry - unless one took a special course) and that I believe Independent University of Moscow to be the best anywhere, whatever courses (math/physics) they taught in Moscow state were generally taught well (with few exceptions) and the really bad part was administration.
2. If you ever see Russian State Universities at the top of anything, be very, very cautious. I studied at MGU (Moscow State University) for a little while, and it was frankly appaling.
Hi, Hyfe, you might be very, very cautious about making overbroad comments.
Now I could possibly believe that computer science of Moscow State was not that good (in 1993 they were doing assembler on Abat3 (Z80 clone) when 386 were widespread), but, at the very least, the Math and Physics departments are excellent in comparison to what passes for "education" in most places in USA.
The only university in USA that I personally know can compare is University of Chicago. The rest can do the job if you know which courses to take (you'll need to start with 300-400 level instead of Calc 101) but tough luck if you want to follow standard progress - or if you want to have fellow students of your own age.
wow, and I mean just fucking WOW at the processing power alone.
This thing makes echelon look like a toy.
Could someone explain to me what is so amazing about a system that pushes through 10 Gbit per second ?
Yes, it is a lot - but not impossible or even very expensive. Split this up into 30 1 Gbit lines, stick a $2000 dual-core Athlons on the end of each and you get a $60K cluster that has a budget of 118 64-bit instructions per byte that comes in. Plenty enough to examine packet headers. If you need more cpu cycles (say for speech recognition) just add nodes.
Now, if they were able to brute-force decrypt 1000 IPSec connections in real time - this would be something to worry about.
And you know what? I'll take enlightened self-interest over kindness any day. At least, then I know where I stand. But if someone is being kind to you for no apparent reason, you really have to wonder about their actual agenda.
I see you have chosen your nick "ScrewMaster" for a reason..
Is this not the same company that sued Handspring over the shape of the buttons on their Treo keyboards?
Yes. Hopefully this is a good thing.
You see, there is no better way to be certain that something is possible but to do it yourself.
These guys tried to be a patent troll, got mugged by a fouler troll and want to clean up the countryside. They are not going to easily accept the usual of "well, the system mostly works with some exceptions" - not only it happened to them, but they have experience in doing it to others.
Even if they stuffed a PII 400 mhz and had a 12" screen, it would be very usefull.
This is an excellent point.
When I was doing undergrad in Moscow I had two friends whose specialization was hydrodynamics.
Obviously they needed to write and run some code, but computer time was hard to come by. So they put their savings together and bought an IBM XT clone for $5. It was that cheap because at that time 386 were already low end. That XT machine was still very useful - and all theirs.
In a similar fashion, what Negroponte is going after is not performance but capability - a device that has a screen, a keyboard, some processing unit and a wireless card makes a whole lot of difference versus the absence of such device.
microsoft has also started offering its own proprietary air for free, in an attempt to muscle out the Earth's atmosphere from its traditional strength position in the marketplace.
Beware - besides regular dust this air contains RFID chips and they will measure how much of their air you breathed by scanning your lungs..
Not a particle physicist, but here is my take on it:
As you know mass enters equations through kinetic energy - mv^2/2
If you have several particles then their kinetic terms add up - and you get a diagonal matrix for mass.
A question comes up though is why are we sure that particle 1 is really particle 1 and not 99% particle 1 and 1% particle 2 ?
Mixtures can happen and do happen very often. Consider a photon and an electron - two particles we can usually separate pretty easily. However, if a photon travels inside transparent but weakly conductive medium then it will interact with electrons there.
In fact, if one writes out equations it would turn out that the solutions that are preserved by time evolution are a combination of photons and electrons - a plasmon. So if we fire a photon into the medium with time it will turn into electron movement which will after some time emit a photon again and so on.
One of the effects of this is that speed of light in the medium becomes smaller than in vacuum.
Back to neutrinos - we have a diagonal mass matrix, but we are not sure that particles we formed it from are really preserved by time evolution and they don't turn into each other - i.e. oscillate.
What we do know is that if all the masses are the same then there is no way to tell one particle from another - and thus there is no oscillation.
In particular, if neutrino masses are 0 they are, obviously, identical and thus no oscillation can occur.
Another interesting point is that what we think of "mass" in matter is mostly (>95%) just bound energy - it comes about from kinetic energy of quarks comprising protons and neutrons via E=mc^2
Jesus, watch the friggin movies!! You forgot the midichlorians -- ASS!
I did - and found it very enlightening that neither the democratic republic nor the Jedi counsel were concerned in the slightest about ethical issues of using an army of brainwashed human clones. Enlightening - both from the point of view of the story and the society it was made for.
I remember getting scolded by some fanboys for suggesting that lightsabers should cast no shadows (apparently they must cast shadows since shadows were present in Episode 4...) because the cutting edge - whether plasma or whatever else - would need to be hairline-thin regardless of temperature in order to slice through things without causing unmanageable explosions of melted and vaporized target material.
Well, assume for a moment that it works. To melt metal one needs a lot of energy - so it likely comes from a nuclear source.
1kg of steel has specific heat of 448 joules per degree Kelvin.
Energy from fusion of hydrogen atoms is at most 8 Mev, the energy stored in Hafnium atom is 3 Mev - let's assume that the agent used has weight of Hafnium but produces 1 Mev per atom.
Thus 1kg of energy agent stores 9e10 Joules - plenty enough to heat 20e3 tonnes of steel to 10000 degrees - cool !
So, as long as I am having fun, here is a "complete" light saber design - just so that no one tries to patent something that obvious:
Handle - let's separate in two compartments - one contains energy agent and the other initiator that bombards that agent with nuclear particles.
In response to bombardment energy agent produces new particles in much greater proportion - this is a sticky point as single pass stimulated emission amplification is likely not that efficient - but then we have power to spare ! In fact this might be a feature as the handle will last very long time - the amplification medium will deplete slowly and from one end.
the particles are passed through moderator which limits their mean path in air to desired length.
put peltier element around the energy agent and moderator and feed the energy into the initiator.
initiator could be made as short pulse laser striking metal foil - these have been tested as tabletop devices already and should be capable to produce 3Mev gamma rays.
move the initiator around as energy agent is used up.
The particle fountain would be very narrow - but it will heat up the air and that would produce the glow. Oh - and plasmas are opaque to light so there will be a shadow.
For games I'd rather see them decrease latency than increase bandwidth. Not sure to what extent that would be possible though.
You really want both as the formula is fixed_latency+data_size/bandwidth.
The last part is not insignificant - for a chunk of 8KB the pipe with 1.5Mbps contributes 44 milliseconds while a pipe doing 6Mpbs contributes only 11 milliseconds.
Slashdot Mirror
The bug is recent, a little time passed and it got caught and fixed.
The trick is that SNS produces a lot more of them and in a beam. You can't focus neutrons as efficiently as you can light or electrons.
The interesting part is that one can change architectures when building a Beowulf cluster - most of the codes are compiled from scratch anyway and performance per dollar is all that matters.
Nevertheless, for the past N years most of these were x86 based - this year the winner is, of course, Opteron.
I am not so sure. The TFA said millions of instructions and the chips are capable of billions. So with HZ=100 there is room enough for 28e6 instructions to be executed - plenty enough to trigger the problem.
Also, the TFA mentioned specifically FPU, so it is possible that an integer loop counter gets processed fast enough and in parallel so that FPU is always loaded.
In particular, suppose you have an array of 10 million floats and you want to find the mean - would this trigger the bug ?
I do agree that it is very nice that they checked and announced the problem.
What the discussion is about is a method to do zero-copy I/O - i.e. share buffers between userland and kernel.
Apparently, one could try to make a hack where the buffer is marked COW, so that when userland starts to write into it before the I/O finishes it is copied.
The alternative (which I find quite sensible) is to simply communicate to the application when the kernel is done with the buffer.
I want to point out that for many the reason they are working on a project is *NOT* to "provide positive customer experience", but to celebrate skill and experience of writing complicated software.
In this case, fellow developers are welcomed most, people who want to learn about the software are welcomed second and "users" who want a canned answer as if they are trying to cheat on a multiple-choice exam are listened to last - and get really annoying if they insist it is their due.
Remember, that when you deal with most corporations you don't even get a chance to talk to the developers. So be polite by recognizing what people value and if you think that you can do a better job - please go ahead. The source is availaible.
I heard someplace that Microsoft Office Mac won't open files created with the Windows version and vice versa.
I suggest you tell her to try OpenOffice - it will open most Microsoft formats (I did not have any fail on me yet) and one can then use OpenOffice to save in either Windows and Mac flavor.
As for *.exe - if these are just archives there some utilities for extracting them, however, AFAIK, it is not a good idea to be sending exe files around - what if they get infected ?
The computer students I met there learnt *everything* voluntarily outside of class. They generally seemed competent.
The economics students were toss. Seriously. I cannot begin to describe it. I met several, and tried discussing with them. This is probably partly due to a much lower general awareness of politics and economics (not that I blame them, 70 years of disagreement being fatal *will* affect a society)
I spoke better English than the Language students. I wrote better English than the English teachers.
I did not spend much time with Math/Physics students though.
Ahh, that explains it. The economics department was not the place where one learns business (at least between 1990-1995), but rather where one learned the science of economics, in whatever form it survived under the communists. So not only was it non-practical to begin with, but also, as you point out, got seriously screwed up by the government.
The people who wanted to learn business enrolled in outside classes - there was some sort of 2-year course which was also a good opportunity to go study in Brandeis for a few years.
As far as language, there was a standing joke that there are three branches of English language: American English, British English and Moscow English.
At some point the government took part in how English is taught as well (especially with the need to translate communist ideology), on top of which being isolated from native speakers did not help much.
On the other hand, I had an excellent English teacher through my years of math classes and I seriously doubt she had any problems with pronunciation, but, as you can guess from her teaching math students, people like that did not get easily into any departments government kept a close watch on.
Lastly, while I can rant for hours about things that could be improved (for example, they did not show us any algebraic geometry - unless one took a special course) and that I believe Independent University of Moscow to be the best anywhere, whatever courses (math/physics) they taught in Moscow state were generally taught well (with few exceptions) and the really bad part was administration.
Hi, Hyfe, you might be very, very cautious about making overbroad comments.
Now I could possibly believe that computer science of Moscow State was not that good (in 1993 they were doing assembler on Abat3 (Z80 clone) when 386 were widespread), but, at the very least, the Math and Physics departments are excellent in comparison to what passes for "education" in most places in USA.
The only university in USA that I personally know can compare is University of Chicago. The rest can do the job if you know which courses to take (you'll need to start with 300-400 level instead of Calc 101) but tough luck if you want to follow standard progress - or if you want to have fellow students of your own age.
Hi nacnud75 :)
Would you know who won the World Series in 2006 ? Just curious ! Thank you !
Could someone explain to me what is so amazing about a system that pushes through 10 Gbit per second ?
Yes, it is a lot - but not impossible or even very expensive. Split this up into 30 1 Gbit lines, stick a $2000 dual-core Athlons on the end of each and you get a $60K cluster that has a budget of 118 64-bit instructions per byte that comes in. Plenty enough to examine packet headers. If you need more cpu cycles (say for speech recognition) just add nodes.
Now, if they were able to brute-force decrypt 1000 IPSec connections in real time - this would be something to worry about.
I see you have chosen your nick "ScrewMaster" for a reason..
Any ethnic conflict can be solved with application of sufficiently large quantities of high explosive.
Yes. Hopefully this is a good thing.
You see, there is no better way to be certain that something is possible but to do it yourself.
These guys tried to be a patent troll, got mugged by a fouler troll and want to clean up the countryside. They are not going to easily accept the usual of "well, the system mostly works with some exceptions" - not only it happened to them, but they have experience in doing it to others.
You mean an exoskeleton ?
This is an excellent point.
When I was doing undergrad in Moscow I had two friends whose specialization was hydrodynamics.
Obviously they needed to write and run some code, but computer time was hard to come by. So they put their savings together and bought an IBM XT clone for $5. It was that cheap because at that time 386 were already low end. That XT machine was still very useful - and all theirs.
In a similar fashion, what Negroponte is going after is not performance but capability - a device that has a screen, a keyboard, some processing unit and a wireless card makes a whole lot of difference versus the absence of such device.
Beware - besides regular dust this air contains RFID chips and they will measure how much of their air you breathed by scanning your lungs..
Wonderful article, thank you very much !
As you know mass enters equations through kinetic energy - mv^2/2
If you have several particles then their kinetic terms add up - and you get a diagonal matrix for mass.
A question comes up though is why are we sure that particle 1 is really particle 1 and not 99% particle 1 and 1% particle 2 ?
Mixtures can happen and do happen very often. Consider a photon and an electron - two particles we can usually separate pretty easily. However, if a photon travels inside transparent but weakly conductive medium then it will interact with electrons there.
In fact, if one writes out equations it would turn out that the solutions that are preserved by time evolution are a combination of photons and electrons - a plasmon. So if we fire a photon into the medium with time it will turn into electron movement which will after some time emit a photon again and so on. One of the effects of this is that speed of light in the medium becomes smaller than in vacuum.
Back to neutrinos - we have a diagonal mass matrix, but we are not sure that particles we formed it from are really preserved by time evolution and they don't turn into each other - i.e. oscillate.
What we do know is that if all the masses are the same then there is no way to tell one particle from another - and thus there is no oscillation.
In particular, if neutrino masses are 0 they are, obviously, identical and thus no oscillation can occur.
Hope this made some sense.
Weird, ha ?
I did - and found it very enlightening that neither the democratic republic nor the Jedi counsel were concerned in the slightest about ethical issues of using an army of brainwashed human clones. Enlightening - both from the point of view of the story and the society it was made for.
Or should I say, m'lady?
Fellow hacker would suffice.. Also RTF link.
Well, assume for a moment that it works. To melt metal one needs a lot of energy - so it likely comes from a nuclear source.
1kg of steel has specific heat of 448 joules per degree Kelvin.
Energy from fusion of hydrogen atoms is at most 8 Mev, the energy stored in Hafnium atom is 3 Mev - let's assume that the agent used has weight of Hafnium but produces 1 Mev per atom.
Thus 1kg of energy agent stores 9e10 Joules - plenty enough to heat 20e3 tonnes of steel to 10000 degrees - cool !
So, as long as I am having fun, here is a "complete" light saber design - just so that no one tries to patent something that obvious:
- Handle - let's separate in two compartments - one contains energy agent and the other initiator that bombards that agent with nuclear particles.
- In response to bombardment energy agent produces new particles in much greater proportion - this is a sticky point as single pass stimulated emission amplification is likely not that efficient - but then we have power to spare ! In fact this might be a feature as the handle will last very long time - the amplification medium will deplete slowly and from one end.
- the particles are passed through moderator which limits their mean path in air to desired length.
- put peltier element around the energy agent and moderator and feed the energy into the initiator.
- initiator could be made as short pulse laser striking metal foil - these have been tested as tabletop devices already and should be capable to produce 3Mev gamma rays.
- move the initiator around as energy agent is used up.
The particle fountain would be very narrow - but it will heat up the air and that would produce the glow. Oh - and plasmas are opaque to light so there will be a shadow.You really want both as the formula is fixed_latency+data_size/bandwidth.
The last part is not insignificant - for a chunk of 8KB the pipe with 1.5Mbps contributes 44 milliseconds while a pipe doing 6Mpbs contributes only 11 milliseconds.
Math ? - No, but behold the Power of Marketing!