I don't see any possible motive for Jobs to do this other than to see just how well the Reality Distortion Field(tm) works.
I don't think you understand how the billionaire mind works. They don't think of much of motivation when dealing with such trivial amounts of money. When you find a quarter in the vending machine change slot, how much effort do you exert to find the proper owner. What were your motivations in putting it in your pocket?
I'm sure Steve feels the same way. He had a bunch of worthless options, he traded them in for some that aren't worthless. So he had to fake a board meeting. What's the big deal? It's only 1/10,000th of the value of the company that he's pocketing. Divide the amount by the (my estimate) 3 million Apple share holders and it's only a bit more than 2 bucks a piece. They can afford that. Besides, he's entitled to that money because he doesn't take a salary and after all when they gave him the (now worthless) options a while back, they were worth something. I'm sure Steve feels that he was owed...
A friend of mine is in health care. She stated she would never want to go to a private, for profit, hospital. Why? Because every decision they make is based on profitability. She has seen both professionally, and a non-profit hospital put the patient first and the cost second. For profits focus on what they can bill.
My wife is in health care. She works for a non-profit hospital (actually a chain of non-profit hospitals mostly owned by a larger non-profit.) And you know what? Every decision they make is based upon profitability. Every department is judged by whether they turn a profit. If they don't, the non-profit hospital will close the department. They are closing emergency rooms left and right because poor people without insurance use the emergency rooms. Therefore there's no-profit on emergency rooms. Unless, of course, there's trauma center, because rich people get into auto accidents, too. The other departments, fearing for their jobs, try their hardest to avoid taking patients with Medi-Cal (California's version of Medicaid, state health care for the impoverished) because Medi-Cal reimburses pennies on the dollar.
But trying to avoid them is better than another "non-profit" chain that just got a several billion dollar state bond issue approved by the voters. That chain doesn't take Medi-Cal at all for non-urgent care. But it's got a nice friendly name, and a reputation for treating diseases that nobody ever gets, so I guess that makes it all right.
Your friend is a utopian if she thinks her hospital is any different. To some extent it's understandable. The hospital needs to break even to stay open. Staying open takes precidence over patient care.
The difference between that and a for profit hospital is that a for profit hospital needs to add 5% on top of that. And, of course, the CEO of a for-profit hospital makes $5M per year, whereas the impoverished CEO of a non-profit hospital might only make $1M. Tough jobs they have, too. From what I can see the job of a hospital CEO is to make sure the hospital loses enough money to be a nice take-over target, but not so much that it's not attractive to buyers. Once the takeover happens, a lot of money appears in the CEO's swiss bank account from out of nowhere and the CEO retires to the Carribean.
Frankly, health care in the US has been in a state of collapse for decades. The collapse will continue until the people that own congress realize THEY might not make it to the nearest open emergency room. And that emergency room might not take them without a credit check.
Given that Linux doesn't anything but the powerpc portion of the cell processor and is limited to 512M, it's not entirely surprising that its performance is unimpressive. Call me when you've got all 7 SPUs and the GPU humming along.
Uhm, aren't those approximations? You still need to do 2 or 3 iterations of newton-raphson to get full float precision. Do you know why these instructions return approximations? Is it maybe because they are a hardware implementation of
Some timings (in 20 second runs on exponentially distributed random floats) for reciprocal sqrt for 32 bit floats with differences only allowed in last bit including proper handling of sqrt(-0.0) (== -0.000)
UltraSPARC II 400 MHz. Carmack: 165.2 ns, FPU 163.7 ns
On intel machines, Carmack is often, but not always, faster then the FPU. In fact, for some it can be faster to use Carmack to compute the non-reciprocal square root. And, no. I don't quite know why G5 sucks so bad in this test. The same source code is used for all of them.
What is causing it, however, is another matter... some say there is proof that humans are causing it, others will say it's merely circumstantial...
And then there are the idiots who think that because there are two sides, then sides must be of equal validity, and therefore we know nothing. Of course, the truth is that the jury came in long ago. That jury was composed of people who study climate, and far more than 11 out of 12 of them think that a majority of the warming in the last century is due to human activity. The verdict was "guilty as charged" and the penalty is "fix it or die."
The fact is, your view that NRAO tried to get earmarks is not correct. Did we (actually our managing organization AUI) lobby congress for funds? Yes. But ask for earmarks? No.
Well, given that earmarks are the only way an organization funded by NSF is going to get funds outside of the normal NSF review process, to "lobby for funds" is the same as to "ask for earmarks." That lobbying was a way to bypass the way NSF normally allocates their funds, which is in theory based upon scientific merit. NSF is prohibited from giving NRAO the funding that it merits if that funding is less than the earmark. Did NRAO honestly think that their request was going to generate an increase in the total astronomy funding? Is anyone that naive?
Did anyone notice that Puerto Rico doesn't have senators? What a surprise that they are on the chopping block.
Do you honestly think that we like to take a heap of flak from the rest of the astronomical community and endanger our future funding?
I think the people involved in the decision thought that the dollars were worth the flak. Anyhow, it seems a little to late to worry about that. And yes, it may color the way proposals from investigators at NRAO are viewed during the normal proposal review process, especially in the near future. But since NRAO is swimming in it, do they really need to propose?
Although lowering tax rates now will increase taxes that have to be paid later, the lower tax rates increase the total economy, and before long more taxes are paid at the lower rate than would have been paid at the higher rate
Yes, and zero tax rates will lead to infinite revenue.
I am an astrophysicist that doesn't limit myself to radio observations and I have to disagree with the above sentiment. The root cause of this issue is not the new instruments that are coming online. The long term cause is flat budgets. The current administration doesn't value this kind of research.
The short term cause is the Director of the National Radio Astronomy Observatories (NRAO) convincing a senator to earmark a significant portion of the NSF astronomy funding for NRAO. Since most of that money is going to be spent in the home state of that senator, he earmarked 20% more than NRAO requested. Of course these earmarks don't change the total funding for astronomy. They just restrict how it can be spent. In other words, the director of NRAO and the senator from New Mexico joined forces to fsck over the rest of the astronomy community.
It is possible that the senior review chose Arecibo for the cuts because of the public outcry that might result could cause an increase in the total astronomy budget.
We are a non-profit group inside a non-profit University. Commercial advertising is against university policy and would threaten our non-profit status. Any non-profit that does use commercial advertising should think twice, as it is clearly a violation of the rules set out for income earned by non-profit entities. The IRS could come a knocking...
Provisions frequently get inserted at the last minute, for the express purpose of preventing anyone from reading it (or debating it) before voting.
The current congress has elevated this to a new level: Adding provisions after the last minute.
It used to be that after a bill was passed, there would be a reconcilliation committee which would reconcile differences between the version passed by the house and the versions passed by the senate. These committies used to represent the party makeup of the houses. Now they are entirely made up of members of tha majority party. They used to only modify existing provisions in the bills. Now they add new provisions that were not in the legislation that was voted upon.
If I have a choice between two candidates who seem equally skilled/motivated/whatever, but one of them came out of MIT and one of them came out of Michigan State, I'm probably going to go with the one from MIT.
Even though I am a graduate of two of the top ten schools in my field, I disagree with that. If two candidates seem equal in ability and seem equal in potential, you should consider how they got that way. The candidate at Michigan State probably had to learn to excel without being pushed to do so. On the other hand, the one from MIT has already shown the ability to bullshit his way through a relatively meaningless interview process. Only question is, can he bullshit his way through this job interview.
The advice I got when selecting graduate schools was that choice of school doesn't matter. I pass that advice along to my students. A student either makes their own education or they don't. Choice of venue is a second order effect.
A GPU is an enormous parallel processor. A cpu is a small (mostly) serial processor. You are an idiot.
Wow that's insightful. Have you written scientific code for a GPU before? I have.
A GPU is a special purpose parallel processor that is very good at doing the calculations required for 3D graphics. It's notoriously difficult to translate real scientific problems into a reasonably small number of vertex computations and texture lookups. They are great at performing a small sequence of identical operations on a large number of input data points. They are awful at performing an arbitrary and dynamically changing stream of operations on a wide variety of input data. Guess which one of these scheme represents the majority of scientific problems?
Since, AFAIHBT, ATI is funding the port, a generalized GPU client might not happen for a while.
Most of the claims in TFA hinge on beleiving that the GPU client is (as Stanford has claimed) 20 to 40 times faster than the CPU client. It would be nice, and certainly beneficial to ATI, if the FAH team would allow the same work units to be processed by both the GPU and CPU versions. As it is, there is no way to test their claims, and it seems they've gone out of their way to be sure there is no way to test their claims.
The SETI@home team answer is that they won't release versions that use specific optimizations for specific hardware because they're worried about the integrity of the results
Time to tune into the new century. SETI@home has been available under the GPL for several years now. Nothing prevents you from modifying it and using the modified version.
I keep asking for people to send me processor specific optimizations and so far only a Mac/PPC version has shown up. I'm ending up writing the SSE version myself which is taking a damn long time. I'd even like to see versions that use the GPU. The problem is not that we are unwilling, it's that I'm one guy with limited time. So far nobody is sending me code...
So put your money where your mouth is and send me some code. Join the boinc_opt mailing list. If no one is willing to put in the work then why the hell should I bother doing it myself.
But if the PS3 Linux kernel is modified to use only those Linux executables that have been signed by Sony (as is the case on PSP), and the PS3 bootloader is modified to use only those Linux kernels that have been signed by Sony, then how do I make my own working executables without being a big enough corporation to become a Sony licensee?
First thing is to drop an email to Linus saying that maybe GPL v3 isn't such a bad thing. Shortly after GPL v3 gets attached to the GNU toolset, Sony will have to cease providing updates to anything that has made the license switch. Or they will have to allow unsigned binaries to run. It's their option.
Unfortunately, Linus doesn't mind if you have to run a Sony signed kernel and would apparently like to allow Sony the option of requiring that you use Sony signed binaries. About the only thing you can do is drop him a line and explain, nicely, why you would prefer the ability to run unsigned binaries.
Write the number (x) in IEEE 32 bit single precision binary format.
Convert it to a 32 bit integer. (i=*(long *)&x;)
Divide the integer by two (shift right 1 bit).
subtract it from 0x5f3759df.
Convert back to single precision floating pont (f=*(float *)&i;).
The result is the reciprocal of the square root to about 3% precision.
If you need better accuracy perform Newton-Raphson iteration (f=f*(1.5-0.5*x*f*f))
Multiply by the original number to get the square root of the original number.
Of course this method can be used to avoid division any time. To find the reciprocal of a number (x), square it and find the approximate reciprocal of the square root of the square using the method above. Then use Newton-Raphson iteration (f=f*(2-x*f)) to get the precision you require.
Slashdot Mirror
I don't think you understand how the billionaire mind works. They don't think of much of motivation when dealing with such trivial amounts of money. When you find a quarter in the vending machine change slot, how much effort do you exert to find the proper owner. What were your motivations in putting it in your pocket?
I'm sure Steve feels the same way. He had a bunch of worthless options, he traded them in for some that aren't worthless. So he had to fake a board meeting. What's the big deal? It's only 1/10,000th of the value of the company that he's pocketing. Divide the amount by the (my estimate) 3 million Apple share holders and it's only a bit more than 2 bucks a piece. They can afford that. Besides, he's entitled to that money because he doesn't take a salary and after all when they gave him the (now worthless) options a while back, they were worth something. I'm sure Steve feels that he was owed...
Actually, rumour has it that Apple has a separate product, the iWC, that's designed specifically for the UK market.
My wife is in health care. She works for a non-profit hospital (actually a chain of non-profit hospitals mostly owned by a larger non-profit.) And you know what? Every decision they make is based upon profitability. Every department is judged by whether they turn a profit. If they don't, the non-profit hospital will close the department. They are closing emergency rooms left and right because poor people without insurance use the emergency rooms. Therefore there's no-profit on emergency rooms. Unless, of course, there's trauma center, because rich people get into auto accidents, too. The other departments, fearing for their jobs, try their hardest to avoid taking patients with Medi-Cal (California's version of Medicaid, state health care for the impoverished) because Medi-Cal reimburses pennies on the dollar.
But trying to avoid them is better than another "non-profit" chain that just got a several billion dollar state bond issue approved by the voters. That chain doesn't take Medi-Cal at all for non-urgent care. But it's got a nice friendly name, and a reputation for treating diseases that nobody ever gets, so I guess that makes it all right.
Your friend is a utopian if she thinks her hospital is any different. To some extent it's understandable. The hospital needs to break even to stay open. Staying open takes precidence over patient care.
The difference between that and a for profit hospital is that a for profit hospital needs to add 5% on top of that. And, of course, the CEO of a for-profit hospital makes $5M per year, whereas the impoverished CEO of a non-profit hospital might only make $1M. Tough jobs they have, too. From what I can see the job of a hospital CEO is to make sure the hospital loses enough money to be a nice take-over target, but not so much that it's not attractive to buyers. Once the takeover happens, a lot of money appears in the CEO's swiss bank account from out of nowhere and the CEO retires to the Carribean.
Frankly, health care in the US has been in a state of collapse for decades. The collapse will continue until the people that own congress realize THEY might not make it to the nearest open emergency room. And that emergency room might not take them without a credit check.
Given that Linux doesn't anything but the powerpc portion of the cell processor and is limited to 512M, it's not entirely surprising that its performance is unimpressive. Call me when you've got all 7 SPUs and the GPU humming along.
On intel machines, Carmack is often, but not always, faster then the FPU. In fact, for some it can be faster to use Carmack to compute the non-reciprocal square root. And, no. I don't quite know why G5 sucks so bad in this test. The same source code is used for all of them.
Why run for president when you can buy congress for a whole lot less.
And then there are the idiots who think that because there are two sides, then sides must be of equal validity, and therefore we know nothing. Of course, the truth is that the jury came in long ago. That jury was composed of people who study climate, and far more than 11 out of 12 of them think that a majority of the warming in the last century is due to human activity. The verdict was "guilty as charged" and the penalty is "fix it or die."
Thus far the defendant has chosen "die."
Well, given that earmarks are the only way an organization funded by NSF is going to get funds outside of the normal NSF review process, to "lobby for funds" is the same as to "ask for earmarks." That lobbying was a way to bypass the way NSF normally allocates their funds, which is in theory based upon scientific merit. NSF is prohibited from giving NRAO the funding that it merits if that funding is less than the earmark. Did NRAO honestly think that their request was going to generate an increase in the total astronomy funding? Is anyone that naive?
Did anyone notice that Puerto Rico doesn't have senators? What a surprise that they are on the chopping block.
I think the people involved in the decision thought that the dollars were worth the flak. Anyhow, it seems a little to late to worry about that. And yes, it may color the way proposals from investigators at NRAO are viewed during the normal proposal review process, especially in the near future. But since NRAO is swimming in it, do they really need to propose?
Yes, and zero tax rates will lead to infinite revenue.
You should stop smoking what you are selling...
You'd be amazed at what 100,000 letters can do to congress.
The short term cause is the Director of the National Radio Astronomy Observatories (NRAO) convincing a senator to earmark a significant portion of the NSF astronomy funding for NRAO. Since most of that money is going to be spent in the home state of that senator, he earmarked 20% more than NRAO requested. Of course these earmarks don't change the total funding for astronomy. They just restrict how it can be spent. In other words, the director of NRAO and the senator from New Mexico joined forces to fsck over the rest of the astronomy community.
It is possible that the senior review chose Arecibo for the cuts because of the public outcry that might result could cause an increase in the total astronomy budget.
We are a non-profit group inside a non-profit University. Commercial advertising is against university policy and would threaten our non-profit status. Any non-profit that does use commercial advertising should think twice, as it is clearly a violation of the rules set out for income earned by non-profit entities. The IRS could come a knocking...
Where did you learn the Queen's english? "Doughnut" is how the lower classes spell "doughnought".
The current congress has elevated this to a new level: Adding provisions after the last minute.
It used to be that after a bill was passed, there would be a reconcilliation committee which would reconcile differences between the version passed by the house and the versions passed by the senate. These committies used to represent the party makeup of the houses. Now they are entirely made up of members of tha majority party. They used to only modify existing provisions in the bills. Now they add new provisions that were not in the legislation that was voted upon.
Even though I am a graduate of two of the top ten schools in my field, I disagree with that. If two candidates seem equal in ability and seem equal in potential, you should consider how they got that way. The candidate at Michigan State probably had to learn to excel without being pushed to do so. On the other hand, the one from MIT has already shown the ability to bullshit his way through a relatively meaningless interview process. Only question is, can he bullshit his way through this job interview.
The advice I got when selecting graduate schools was that choice of school doesn't matter. I pass that advice along to my students. A student either makes their own education or they don't. Choice of venue is a second order effect.
A GPU is a special purpose parallel processor that is very good at doing the calculations required for 3D graphics. It's notoriously difficult to translate real scientific problems into a reasonably small number of vertex computations and texture lookups. They are great at performing a small sequence of identical operations on a large number of input data points. They are awful at performing an arbitrary and dynamically changing stream of operations on a wide variety of input data. Guess which one of these scheme represents the majority of scientific problems?
Since, AFAIHBT, ATI is funding the port, a generalized GPU client might not happen for a while.
Most of the claims in TFA hinge on beleiving that the GPU client is (as Stanford has claimed) 20 to 40 times faster than the CPU client. It would be nice, and certainly beneficial to ATI, if the FAH team would allow the same work units to be processed by both the GPU and CPU versions. As it is, there is no way to test their claims, and it seems they've gone out of their way to be sure there is no way to test their claims.
Call me skeptical.
I hope they come up with a new acronym. PNI is already used for Prescott New Instructions.
Isn't that 9 named storms and 5 hurricanes? A named storm doesn't stop being a named storm when it becomes a hurricane.
Time to tune into the new century. SETI@home has been available under the GPL for several years now. Nothing prevents you from modifying it and using the modified version.
I keep asking for people to send me processor specific optimizations and so far only a Mac/PPC version has shown up. I'm ending up writing the SSE version myself which is taking a damn long time. I'd even like to see versions that use the GPU. The problem is not that we are unwilling, it's that I'm one guy with limited time. So far nobody is sending me code...
So put your money where your mouth is and send me some code. Join the boinc_opt mailing list. If no one is willing to put in the work then why the hell should I bother doing it myself.
Unfortunately, Linus doesn't mind if you have to run a Sony signed kernel and would apparently like to allow Sony the option of requiring that you use Sony signed binaries. About the only thing you can do is drop him a line and explain, nicely, why you would prefer the ability to run unsigned binaries.
Of course this method can be used to avoid division any time. To find the reciprocal of a number (x), square it and find the approximate reciprocal of the square root of the square using the method above. Then use Newton-Raphson iteration (f=f*(2-x*f)) to get the precision you require.
(*): all answers are provided in base sqrt(2).