It is amazing that this is so common. We had a similar situation in our lab here. Our espresso machine had tripped a circuit breaker in our lab. No biggie, but we have to call a union guy to flip it back for us. So he comes down and doesn't even know where the box is. I tell him and he goes off. A minute later, all of our computers go off. Apparently he wasn't sure what was what and just flipped every breaker in the box to make sure...
90/10 is not a law or empirical result, just a rule of thumb, and probably rarely true within a reasonable enough margin of error to plug it into another equation.
90/10 is a rule of thumb, but I think it is quite a bit better than you give it credit for.
Obviously some codes scale far beyond an order of magnitude or there wouldn't be supercomputers with hundreds or thousands of processors.
As a computational chemist, I am quite aware of these special cases that violate this rule of thumb. I do think that the original poster was too absolute in his/her assertion, but I still think it is essentially true. For example, tons of processors doesn't give you much more than an order of magnitude speed up for a molecular dynamics simulation. In other words, if you can simulate a certain sized system for 1 ns, doubling the number of processors may not give you 2 ns. However, more processors will help you much more for larger systems. It is all about how parallelizable the problem is, and 90/10 is a good rule of thumb.
How do you know the 90% is serializable and the 10% isn't? Answer: you don't, there is no relationship whatsoever.
You are right, maybe you can't parallelize the 10% of the code that is costing you 90% of your time. That makes s greater than 10 and p less than 90. The limit of that as n goes to infinity is less than 10; that is the assertion of the original poster, that a 10 times speed up is about the best you can hope for when parallelizing a code.
The parent post should be ignored. The information content, while real, is misapplied, and that "10" number is pulled out of his ass.
That is what I thought at first, too. But the orignal poster is right (in a way), a factor of 10 is about the best you can hope for when parallelizing code. Since Amdahl's (or some other guy's) law also says something like 90% of the time is spent in 10% of the code. That makes s=10 and p=90. The limit of his equation, (s+p)/(s+p/n), as n goes to infinity is 10. A number not pulled out of anyone's ass.
Maybe the original poster should be moderated down because I don't think the stuff here is really about parallelization (they talk about speed ups on uniproc systems too), but for the parallel case, he seems to be right.
If that's what you want, get the MPEG-LA to lighten up on MPEG-4 licensing.
Then you won't have a problem. It's Apple's goal, after all, to have the most open, standards-based platform. It's not quite their choice to hold Sorenson codecs from Linux.
Why don't they choose not to use closed, fascist licensed codecs. Then they wouldn't be in this position. (And neither would we.)
Yeah, I have been disappointed by printing in mozilla. I found a trick that seems to help with 0.9.8 though. Print to a file and then run the file through ps2ps and print that. Works everytime so far...
so we don't have to try to find a dusty old copy of DOS 5.0 or 3.2 to update the bios on a shiny new P4 motherboard.
But you don't. Just download FreeDOS and follow the instructions to make a boot disk. Download your motherboard's BIOS update program and new BIOS to the boot disk and you are good to go. We recently did exactly this to update a half a dozen ASUS based machines here.
You are confusing popular with good. How about Jeanette Winterson, William Gibson, Isaac Asimov, Arthur C. Clarke, J.R.R. Tolkien, Robert Heinlein, Ray Bradbury?
Personally I think jms's comment above could quite possibly be the most insightful thing I have ever read on slashdot.
A version control system must provide security. All too often management uses the SCM repository as kind of a shared directory (BAD, BAD, BAD) -- and people who have no need to see or modify the code, do... implicitly.
Since you seem to know a lot about this, or at least have thought about it more than I have, let me ask you a question about this point. Shortly after I finally convinced my boss to use CVS (the way he remembers it now we always used CVS...) he when a little nutty with it and started using in in just the way you described, as a "shared directory" of stuff people needed. As our sys. admin. I had a "bad feeling" about this, but he was nuts about it and it seems to work so i didn't argue.
So my question is why is this so bad? Should I really push for CVS to not be used this way? Why?
The one thing I hate about disk based backup solutions is they are not disaster proof. With a tape system I can take a set of tapes out of rotation and put them off-site, like in a safe-deposit box. I guess for home use that isn't so important for most people, but when your thesis work is on a computer, you want to know you won't loose it all, even if your house gets hit by a tornado or something.
Just look at how many of these worms have had little bugs in them, like not attacking when the were supposed to, or emailing the wrong drop and stuff. All we need is some cowboy thinking he is going to clean up the internet and messing up even more stuff.
Ever see that movie Office Space? One wrong decimal point could mean big trouble. It is bad enough these people have to run Microsoft's buggy code. But at least they chose to do that. They shouldn't be forced to run your buggy code too, even if you are trying to help.
What exactly does GNOME have to do with argonomics?
I don't know? What is argonomics?
The last time I checked, a desktop environment has no bearing whatsoever on soil or plnat sciences.
Yeah, most desktop environments don't work so well if you soil them, and there was soil before desktop environments, but what the hell is plnat sciences?
I recently upgraded to 4.1.0, and now I get "snow" on my screen whenever I scroll. It goes away when I stop scrolling. 4.0.3 was fine. Does anyone else have this problem?
admittedly, their motivation is probably more economical than moral
Why do these two things have to be diametrically opposed in everyone's minds? This is the market working. Sure, it would be great if the movie guys weren't such bastards. It is their irrational self-interest that is destroying them. The could make a ton of money if they played nice, but instead they choose to dick us over. Too bad the market isn't letting that happen.
To paraphrase Princess Leia, the tighter they squeeze, the more money slips through their fingers.
If you want to sweeten the deal (if you have a candidate you really want to woe) you can allow them to spend some amount of work time on Open Source projects.
You want to woo them, or maybe wow them, but definitely don't woe them.
Bugs and incompatibilities are taken care of by the maintainer of the software. If RedHat has a patch to fix a problem, they can submit it.
The great thing about free software is YOU can take care of these problems. If your answer was "Bugs and incompatibilities are taken care of by the maintainer of the software. If RedHat has a patch to fix a problem, they can beg us to fix it." one might almost think you are talking about Microsoft or something. Free software is about control, specifically that fact that YOU control the code, not some "maintainer."
I think djbdns and qmail are probably great for the compile it yourself crowd who can make any modifications locally and be satisfied, but for the people trying to add value, it sucks.
I like Linux a lot and use it at home and at work. (Mandrake at home and RedHat at work). In both cases I like Linux because it helps me get my work done, not because I like fucking around with it (which I do, but i can't let that get in the way of my work). So I rely on distributors to make my life easier. Anything that makes life harder for them makes life harder for me.
A lot of people are yelling that there aren't holes in qmail and/or djbdns. Okay, there probably aren't, but that cash reward is small consolation for RedHat and its customers in the unlikely case the shit does hit the fan.
But what about bugs? Incompatabilities? Features that RedHat customers want that these programs might not have natively? Can't do anything about that can they? I suppose they could ship the source and patches and build them during install to get around the "distributing modified binaries" clause, but what a pain.
Imagine if RedHat could only ship the Linus kernel binary, or you had to build the modified kernel during every installation (makes installation of large clusters quite a pain). The great thing about the license of the kernel is that RedHat can modify the kernel, give those changes back to the community at large as source, and to their customers as easy to use binaries.
they had to invent a new size prefix to describe how large a filesystem XFS could accomodate ("exo-byte" = 1024 Gb)
This is a slight exaggeration. A kilobyte is
1024 bytes, a megabyte is 1024 kilobytes, a gigabyte is 1024 megabytes, a terabyte is 1024 gigabytes. Oh wow, it is not even an exaggeration, it is just wrong, I guess. Anyway, if we keep going up the table of SI prefixes in my CRC, we could have petabytes, which would be 1024 terabytes, and then exabytes = 1024 petabytes, and so on. The is no need to invent things, it is all taken care of, from 10E-24 (yocto-) to 10E24 (yotta) (now thats a lotta stuff);)
It's a good thing Moore didn't have a pompous ass like you as an instructor, or he might have been too traumatized to make the observation that processing capacity doubles every 18 months.
If Moore's observation was correct (which most people seem to think has been shown by the history of the industry) there are ways to "prove" it, like trying different fitting functions and looking at their errors and/or correlation coefficients. In the case of Moore's law, one whould find that an exponential is the best fitting function.
For most of the graphs in the article, linear or simple polynomial (e.g., quadratic) would appear to give better/comparable fits to the presented data. It seems they chose exponential because it is more impressive to say "this is growing expoentially!" than to say "we fit the growth to a quadratic with coefficients blah-blah-blah."
Slashdot Mirror
It is amazing that this is so common. We had a similar situation in our lab here. Our espresso machine had tripped a circuit breaker in our lab. No biggie, but we have to call a union guy to flip it back for us. So he comes down and doesn't even know where the box is. I tell him and he goes off. A minute later, all of our computers go off. Apparently he wasn't sure what was what and just flipped every breaker in the box to make sure...
90/10 is a rule of thumb, but I think it is quite a bit better than you give it credit for.
As a computational chemist, I am quite aware of these special cases that violate this rule of thumb. I do think that the original poster was too absolute in his/her assertion, but I still think it is essentially true. For example, tons of processors doesn't give you much more than an order of magnitude speed up for a molecular dynamics simulation. In other words, if you can simulate a certain sized system for 1 ns, doubling the number of processors may not give you 2 ns. However, more processors will help you much more for larger systems. It is all about how parallelizable the problem is, and 90/10 is a good rule of thumb.
You are right, maybe you can't parallelize the 10% of the code that is costing you 90% of your time. That makes s greater than 10 and p less than 90. The limit of that as n goes to infinity is less than 10; that is the assertion of the original poster, that a 10 times speed up is about the best you can hope for when parallelizing a code.
That is what I thought at first, too. But the orignal poster is right (in a way), a factor of 10 is about the best you can hope for when parallelizing code. Since Amdahl's (or some other guy's) law also says something like 90% of the time is spent in 10% of the code. That makes s=10 and p=90. The limit of his equation, (s+p)/(s+p/n), as n goes to infinity is 10. A number not pulled out of anyone's ass.
Maybe the original poster should be moderated down because I don't think the stuff here is really about parallelization (they talk about speed ups on uniproc systems too), but for the parallel case, he seems to be right.
Why don't they choose not to use closed, fascist licensed codecs. Then they wouldn't be in this position. (And neither would we.)
Yeah, I have been disappointed by printing in mozilla. I found a trick that seems to help with 0.9.8 though. Print to a file and then run the file through ps2ps and print that. Works everytime so far...
They say
But the gun works by repeatedly smashing these same delicate magnets with steel balls! If the magnets are really so fragile, wouldn't they break?
so we don't have to try to find a dusty old copy of DOS 5.0 or 3.2 to update the bios on a shiny new P4 motherboard.
But you don't. Just download FreeDOS and follow the instructions to make a boot disk. Download your motherboard's BIOS update program and new BIOS to the boot disk and you are good to go. We recently did exactly this to update a half a dozen ASUS based machines here.
You are confusing popular with good. How about Jeanette Winterson, William Gibson, Isaac Asimov, Arthur C. Clarke, J.R.R. Tolkien, Robert Heinlein, Ray Bradbury?
Personally I think jms's comment above could quite possibly be the most insightful thing I have ever read on slashdot.
A version control system must provide security. All too often management uses the SCM repository as kind of a shared directory (BAD, BAD, BAD) -- and people who have no need to see or modify the code, do... implicitly.
Since you seem to know a lot about this, or at least have thought about it more than I have, let me ask you a question about this point. Shortly after I finally convinced my boss to use CVS (the way he remembers it now we always used CVS...) he when a little nutty with it and started using in in just the way you described, as a "shared directory" of stuff people needed. As our sys. admin. I had a "bad feeling" about this, but he was nuts about it and it seems to work so i didn't argue.
So my question is why is this so bad? Should I really push for CVS to not be used this way? Why?
Thanks in advance.
In this decade, we don't use casettes, and some of us rarely use cd's.
And what do you use to back up your machines?
I use my ipaq for music with mp3's.
That is great if all you care about are your mp3s.
This machine is for changing tapes in a computer backup system, not some kind of fancy jukebox.
The one thing I hate about disk based backup solutions is they are not disaster proof. With a tape system I can take a set of tapes out of rotation and put them off-site, like in a safe-deposit box. I guess for home use that isn't so important for most people, but when your thesis work is on a computer, you want to know you won't loose it all, even if your house gets hit by a tornado or something.
There is a flaw in your reasoning
The author can always change the license.
Just look at how many of these worms have had little bugs in them, like not attacking when the were supposed to, or emailing the wrong drop and stuff. All we need is some cowboy thinking he is going to clean up the internet and messing up even more stuff.
Ever see that movie Office Space? One wrong decimal point could mean big trouble. It is bad enough these people have to run Microsoft's buggy code. But at least they chose to do that. They shouldn't be forced to run your buggy code too, even if you are trying to help.
What exactly does GNOME have to do with argonomics?
I don't know? What is argonomics?
The last time I checked, a desktop environment has no bearing whatsoever on soil or plnat sciences.
Yeah, most desktop environments don't work so well if you soil them, and there was soil before desktop environments, but what the hell is plnat sciences?
I recently upgraded to 4.1.0, and now I get "snow" on my screen whenever I scroll. It goes away when I stop scrolling. 4.0.3 was fine. Does anyone else have this problem?
admittedly, their motivation is probably more economical than moral
Why do these two things have to be diametrically opposed in everyone's minds? This is the market working. Sure, it would be great if the movie guys weren't such bastards. It is their irrational self-interest that is destroying them. The could make a ton of money if they played nice, but instead they choose to dick us over. Too bad the market isn't letting that happen.
To paraphrase Princess Leia, the tighter they squeeze, the more money slips through their fingers.
Not on Linux it doesn't.
If you want to sweeten the deal (if you have a candidate you really want to woe) you can allow them to spend some amount of work time on Open Source projects.
You want to woo them, or maybe wow them, but definitely don't woe them.
Bugs and incompatibilities are taken care of by the maintainer of the software. If RedHat has a patch to fix a problem, they can submit it.
The great thing about free software is YOU can take care of these problems. If your answer was "Bugs and incompatibilities are taken care of by the maintainer of the software. If RedHat has a patch to fix a problem, they can beg us to fix it." one might almost think you are talking about Microsoft or something. Free software is about control, specifically that fact that YOU control the code, not some "maintainer."
I think djbdns and qmail are probably great for the compile it yourself crowd who can make any modifications locally and be satisfied, but for the people trying to add value, it sucks.
I like Linux a lot and use it at home and at work. (Mandrake at home and RedHat at work). In both cases I like Linux because it helps me get my work done, not because I like fucking around with it (which I do, but i can't let that get in the way of my work). So I rely on distributors to make my life easier. Anything that makes life harder for them makes life harder for me.
A lot of people are yelling that there aren't holes in qmail and/or djbdns. Okay, there probably aren't, but that cash reward is small consolation for RedHat and its customers in the unlikely case the shit does hit the fan.
But what about bugs? Incompatabilities? Features that RedHat customers want that these programs might not have natively? Can't do anything about that can they? I suppose they could ship the source and patches and build them during install to get around the "distributing modified binaries" clause, but what a pain.
Imagine if RedHat could only ship the Linus kernel binary, or you had to build the modified kernel during every installation (makes installation of large clusters quite a pain). The great thing about the license of the kernel is that RedHat can modify the kernel, give those changes back to the community at large as source, and to their customers as easy to use binaries.
I am not affilated with RedHat in anyway.
This isn't specifically for bytes, but the list of SI prefixes is here
The new prefixes for binary units (which nobody uses (the prefixes, not binary units)) are here
they had to invent a new size prefix to describe how large a filesystem XFS could accomodate ("exo-byte" = 1024 Gb)
This is a slight exaggeration. A kilobyte is 1024 bytes, a megabyte is 1024 kilobytes, a gigabyte is 1024 megabytes, a terabyte is 1024 gigabytes. Oh wow, it is not even an exaggeration, it is just wrong, I guess. Anyway, if we keep going up the table of SI prefixes in my CRC, we could have petabytes, which would be 1024 terabytes, and then exabytes = 1024 petabytes, and so on. The is no need to invent things, it is all taken care of, from 10E-24 (yocto-) to 10E24 (yotta) (now thats a lotta stuff) ;)
It's a good thing Moore didn't have a pompous ass like you as an instructor, or he might have been too traumatized to make the observation that processing capacity doubles every 18 months.
If Moore's observation was correct (which most people seem to think has been shown by the history of the industry) there are ways to "prove" it, like trying different fitting functions and looking at their errors and/or correlation coefficients. In the case of Moore's law, one whould find that an exponential is the best fitting function.
For most of the graphs in the article, linear or simple polynomial (e.g., quadratic) would appear to give better/comparable fits to the presented data. It seems they chose exponential because it is more impressive to say "this is growing expoentially!" than to say "we fit the growth to a quadratic with coefficients blah-blah-blah."
Lies, damn lies, and statistics.
What community college do you teach at, stupid arrogant cock-sucker?
Very eloquent. I teach at the University of Minnesota, Twin Cities.