Should we really feel so good about this list? Should we really feel so good that such a significant portion of American computational resources is for warfare and the design weapons of mass destruction?
Look at the other machines at the top of the list. Where do other countries place their computational resources?
You are correct thout that the second CPU will (obviously) be warmer than the first one
Only if the water circulates so slowly that it warms up appreciably during the time it is in contact with the first CPU. For sufficiently fast circulation, the water will be essentially isothermal throughout the system and cooling efficiency will be essentially the same at each CPU.
While I do not know the quantitative engineering of the system, "sufficiently" most certainly is acheivable where "essentially" is small compared to other asymetries in the system.
There is a significant issue missing from this discussion.
There are two sorts of science journals: 1) for-profit journals from major publishing houses 2) non-profit journals from scientific societies (American Metorological Society, American Physical Society, The Royal Society). The second are most often the most highly regarded within a discipline -- even if others are more "prestigious" in some intangible way. The society officers and journal editors are selected by their colleagues and represent some of the most respected scientists in a field. Thus, scientists collectively charge and collect fees and own the copyrights.
The non-profit journals presage the best model for open source software: Free Software Foundation. Much open source has a lazy attitude toward copyright, with no real control or assurance of protection. This is the model advocated by many here for scientific publication. This is truly bad. Since this is slashdot, I assume you all know why XEmacs code was not incorporated into FSF Emacs (both are GNU, by the way) -- Richard Stallman needs all code copyright signed over to the FSF because only that organization can protect the freedom of code.
Likewise, when scientists sign copyright over to a scientific society, they are doing so in order that the society protect the openness of their work. Individual scientists cannot take on the task of preserving copyright. The societies are providing a service to open science. US Federal scientists do not sign over copyright as the government takes on this role.
As to costs, authors pay publication charges so that granting agencies subsidize the scientific societies. The more money you have for research, the more you support the societies. It is an acknowledged cost of doing research and an assumed budget item in proposals.
The system is pretty darned good and works probably better than any institution in existance. I don't expect it to change any time soon.
Basically, Fortran is still used because it's well adapted for the job it's doing.
This is why fortran will always exist and will always be the primary language used to code big numerical problems like weather models. There isn't anything any language can do, which is important to scientific programing, that fortran does not do as well or better. For, if there were, that feature would be implemented into fortran. It is a very dynamic language that takes advantage of the latest technologies.
It is because of this very feature of fortran that it appears marginalized. It is not, it is just highly specialized. Previously, the specialty it occupied was all there was in computing. Now, computing has become much more diverse than solving numerical models and other languages have arisen. Thus, the appearance that fortran has faded is an illusion. It is just that other tasks, using a vast array of new languages, have appeared.
"it seems that once someone learns how to program in Matlab, they never go back. It's just so damn quick to prototype ideas with."
Sure, Matlab is great for toy models and analysis, but these are not scientific programming tasks for which fortran exists.
Mostly, matlab converts never really learned modern fortran well. fortran90 allows all the syntax that makes Matlab quick to script plus more. I've found that any matlab script can be written in fewer, clearer lines of fortran. Better to invest the time learning a portable efficient programming language than a proprietary scripting language.
Yes. LINPACK says very little about many real-world simulations, and the top-500 list says very little about who has the best computers for certain tasks.
Climate modelling, one of the principal super computing applications, is a good example. Climate models do not parallelize over more than a few hundred nodes. Thus, machines with a few NEC SX-6 nodes are far better than those with many cheap nodes.
The weapons of mass destruction research that dominates US super computing is much better suited to massively parallel computing than more useful science. Unfortunately, that application has dominated US super computer design to the detriment of areas like climate modelling where European centers with Japanese super computers are leading.
Another thing to consider is that these machines don't run a single process that uses all nodes. Most have hundreds of processes running, each using a small subset of the available nodes. The gloming together of many nodes to run a single benchmark is pretty much an artificial excersize. The only real advantage of a large cluster is 1) bragging rights and 2) administrative.
-Eric
The statment "Mac OS X replacing Linux at Tokyo University" is misleading on so many levels, I won't even begin. The discussion has further degenerated around inaccurate speculation. Let's touch briefly on reality:
The new system, which includes the Macs and is provided by NEC, is described here.
These are "located in the Information Education Buildings, the Information Technology Center, and the main library" (ref) to be used "for general computer literacy education and programing education".
16% of the terminals will be Windows-based for lectures using Windows or CAD applications. What they replace are a mix of Windows NT and X window terminals (they write "X-terminal", but presumably these are actually linux boxes, not really X terminals).
What this decidedly has nothing to do with is the systems used by scientists, scholars, administrators, and clerical staff at the University of Tokyo. Nor what might be used in other teaching settings. So, the idea that Univ of Tokyo once used Linux monolithiocally is as wrong as to think Linux was replaced by OS X monolithically.
-Eric
Slashdot Mirror
Should we really feel so good about this list? Should we really feel so good that such a significant portion of American computational resources is for warfare and the design weapons of mass destruction?
Look at the other machines at the top of the list. Where do other countries place their computational resources?
Eric Salathe
You are correct thout that the second CPU will (obviously) be warmer than the first one
Only if the water circulates so slowly that it warms up appreciably during the time it is in contact with the first CPU. For sufficiently fast circulation, the water will be essentially isothermal throughout the system and cooling efficiency will be essentially the same at each CPU.
While I do not know the quantitative engineering of the system, "sufficiently" most certainly is acheivable where "essentially" is small compared to other asymetries in the system.
Eric Salathe
There is a significant issue missing from this discussion.
There are two sorts of science journals: 1) for-profit journals from major publishing houses 2) non-profit journals from scientific societies (American Metorological Society, American Physical Society, The Royal Society). The second are most often the most highly regarded within a discipline -- even if others are more "prestigious" in some intangible way. The society officers and journal editors are selected by their colleagues and represent some of the most respected scientists in a field. Thus, scientists collectively charge and collect fees and own the copyrights.
The non-profit journals presage the best model for open source software: Free Software Foundation. Much open source has a lazy attitude toward copyright, with no real control or assurance of protection. This is the model advocated by many here for scientific publication. This is truly bad. Since this is slashdot, I assume you all know why XEmacs code was not incorporated into FSF Emacs (both are GNU, by the way) -- Richard Stallman needs all code copyright signed over to the FSF because only that organization can protect the freedom of code.
Likewise, when scientists sign copyright over to a scientific society, they are doing so in order that the society protect the openness of their work. Individual scientists cannot take on the task of preserving copyright. The societies are providing a service to open science. US Federal scientists do not sign over copyright as the government takes on this role. As to costs, authors pay publication charges so that granting agencies subsidize the scientific societies. The more money you have for research, the more you support the societies. It is an acknowledged cost of doing research and an assumed budget item in proposals.
The system is pretty darned good and works probably better than any institution in existance. I don't expect it to change any time soon.
Eric Salathe
This is why fortran will always exist and will always be the primary language used to code big numerical problems like weather models. There isn't anything any language can do, which is important to scientific programing, that fortran does not do as well or better. For, if there were, that feature would be implemented into fortran. It is a very dynamic language that takes advantage of the latest technologies.
It is because of this very feature of fortran that it appears marginalized. It is not, it is just highly specialized. Previously, the specialty it occupied was all there was in computing. Now, computing has become much more diverse than solving numerical models and other languages have arisen. Thus, the appearance that fortran has faded is an illusion. It is just that other tasks, using a vast array of new languages, have appeared.
"it seems that once someone learns how to program in Matlab, they never go back. It's just so damn quick to prototype ideas with." Sure, Matlab is great for toy models and analysis, but these are not scientific programming tasks for which fortran exists. Mostly, matlab converts never really learned modern fortran well. fortran90 allows all the syntax that makes Matlab quick to script plus more. I've found that any matlab script can be written in fewer, clearer lines of fortran. Better to invest the time learning a portable efficient programming language than a proprietary scripting language.
Yes. LINPACK says very little about many real-world simulations, and the top-500 list says very little about who has the best computers for certain tasks. Climate modelling, one of the principal super computing applications, is a good example. Climate models do not parallelize over more than a few hundred nodes. Thus, machines with a few NEC SX-6 nodes are far better than those with many cheap nodes. The weapons of mass destruction research that dominates US super computing is much better suited to massively parallel computing than more useful science. Unfortunately, that application has dominated US super computer design to the detriment of areas like climate modelling where European centers with Japanese super computers are leading. Another thing to consider is that these machines don't run a single process that uses all nodes. Most have hundreds of processes running, each using a small subset of the available nodes. The gloming together of many nodes to run a single benchmark is pretty much an artificial excersize. The only real advantage of a large cluster is 1) bragging rights and 2) administrative. -Eric
The statment "Mac OS X replacing Linux at Tokyo University" is misleading on so many levels, I won't even begin. The discussion has further degenerated around inaccurate speculation. Let's touch briefly on reality:
The University-wide system is described here.
The new system, which includes the Macs and is provided by NEC, is described here.
These are "located in the Information Education Buildings, the Information Technology Center, and the main library" (ref) to be used "for general computer literacy education and programing education".
16% of the terminals will be Windows-based for lectures using Windows or CAD applications. What they replace are a mix of Windows NT and X window terminals (they write "X-terminal", but presumably these are actually linux boxes, not really X terminals).
What this decidedly has nothing to do with is the systems used by scientists, scholars, administrators, and clerical staff at the University of Tokyo. Nor what might be used in other teaching settings. So, the idea that Univ of Tokyo once used Linux monolithiocally is as wrong as to think Linux was replaced by OS X monolithically. -Eric