There's a fairly detailed interview with bruce eckel ("famous" guy who has written c++ and java books and sits on the C++ standards committee) at artima on why he likes python (I linked to the last part of the article series, since that contains links to the previous ones).
ASCI Q and Pink where both in fact installed in the same time frame. Actually, both of these can be classified as clusters. I guess the point was that ASCI Q is a cluster of DEC, uh Compaq, sorry, HP Alpha machines (=expensive) while Pink is a x86 cluster running Linux.
ethernet has about 100 us (microseconds) latency, regardless if it's 100 or 1000 Mbit. Specialized cluster interconnects, like myrinet, scali and quadrics, have about 3-7 us. Of course, those interconnects cost a fortune too.
Imagine having the fastest processesor on earth, and then take that chip and use it to do the calculation of x1++ (thats x1 = x1 +1 for you non-C'ers)and looping it a few Trillion times.
Well, regardless of the chip performing the calculation, I'd prefer doing
x1 += a_few_trillion;
instead of looping.;-)
But yes, I get your point. "Supercomputer CPU's", like in the earth simulator, are optimised for SIMD operations. Given a vectorizing FORTRAN compiler and a BLAS library handwritten in assembler to take advantage of SIMD features on the chip, I guess you can get some decent performance from such chips.
Re:Supporting travelling users
on
Replacing SMTP?
·
· Score: 1
That's what SMTP Auth and similar stuff is for, i.e. all your mail goes through your own/companys/whatever mailserver (you authenticate to the mailserver and possibly the connection can be encrypted too) wihtout having to allow open relaying on the mailserver.
One of the benefits of the RMX proposal is that it would create and incentive to deploy SMTP Auth too.
Actually, from FORTRAN 90 onward, pointers (somewhat restricted compared to C, but still) as well as dynamic memory allocation are part of the language. Making FORTRAN as slow as C..:-)
Anyway, I don't think the difference between C and fortran is so big anymore. Compiler tech has improved, and in fact most fortran compilers have identical backends with the C/C++ compiler from the same manufacturer.
Exocet is obsolete. Modern ASM:s fly at mach 3 and do evasive manouvers. There's no chance that a phalanx (or goalkeeper, for that matter) will shoot something like that down, except by pure chance.
Of course, that's why phalanx is being replaced with RAM.
Contrast that with Red Hat for example, who are yanking support for their 'personal' operating systems 12 months from the time of their release. It's kind of sobering to think that Red Hat 8.0, 7.3, 7.2, 7.1 are end of lifed in six months from now and 9.0 a mere four months after that.
While this might save Red Hat money in the short term I have to wonder what impact it will have on customer confidence. Even assuming you bought it on the very day of release at best you get twelve months maximum of bug fixes, which isn't very much especially if you were planning on deploying it. If some horrible exploit is discovered ten months from now you're screwed. You might appeal to the community to produce an updated patch, but you still forfeit any QA testing or automated RHN update that you would have gotten before.
Well that's why RH has introduced their Enterprise server/workstation/advanced server/whatever line of products, with all kinds of support options. They have, IIRC, a 5 year support lifetime. Granted, they cost a lot too, but RH has to make a profit just like any other company. IMHO, their strategy is entirely reasonable:
(a) Use the "normal" RH distro, get the latest and greatest software for free, and help RH and the free software community improve the software (by filing bug reports, if nothing else).
(b) Buy the Enterprise/blahblah products and get a high quality product with a long lifetime and support.
(c) If you're just a parasite expecting to get everything for free without contributing anything, sod off.
#3 on the list is a machine made by linux networx for LLNL. It consists of 1100+ dual P4 nodes with quadrics interconnect. And linuxBIOS on all the nodes.
An interesting thing about this is that it doesn't consist of the ususal 1U rack cases, but they are special cases where the boards are mounted vertically (better airflow).
Re:Irrelevant? Absolutely. I kicked the habit!
on
Is 3G Irrelevant?
·
· Score: 1
...But you couldn't kick your slashdot habit, huh?:)
But yes, I understand you. It's really irritating when you're trying to concentrate on something and people keep calling you, babbling about some shit that you really couldn't care less about.
Well, if there only was a RFID tag taped on the guns, what's stopping you from just removing it?
Over here in the civilized world, you need a licence to own and operate a gun. Seems to work quite well. Yes, the government has a database of all guns and their owners. Boo!! hiss!! So what? The idea that a gun-equipped public could prevent the government from turning against its own people was perhaps a good idea 150-200 years ago when wars were won by the side who had the most guys with handguns. That time is long gone.
Most scientific applications store their data in some proprietary format, or with the help of something like netcdf or hdf. It's not like the article is suggesting that they should turn all supercomputers into relational databases.
No, the next version will be XXP. And as diversifying for different customers is all the rage we'll have different version, such as:
XXP home edition - without network transparency for those people who are too stupid to understand that X doesn't use tcp/ip sockets on localhost, and thus believe that taking away network transparency would improve performance.
XXP professional - X as we know it.
XXP enterprise advanced datacenter server edition - Very expensive version for those that like to pay through their noses for no reason. This is also targeted for servers, but as the server and client concept is somewhat backward for X, the only things this includes are xterm, xclock and xeyes.
...So you could grab Apache (is that GPL? assuming it is)...
Umm, no, Apache is licensed under the (drumroll...) Apache License! Basically it's the BSD license with an added clause saying that if you sell a derived product, you may not call it apache.
But other than that, or point is correct. You don't have to release your modifications to GPL code if you keep it for yourself. IIRC RMS supported himself by doing proprietary in-house modifications to emacs back in the eighties (or was it the early nineties?).
Well, they say the CPU is liquid-cooled, not necessarily water-cooled. Looking at the pictures there's no pump, so I guess it's some kind of heat pipe which transports heat from the cpu to the case by convection only. Better pictures at http://www.signum-data.de/pdf/ct_signum_engl.pdf
The hush technologies itx box was already mentioned. Another german thing which is very similar, but uses a P4, is the Signum Data FutureClient. Expensive as hell, though.
Re:Does linux support hypertrheading?
on
Linux SMP Round-Up
·
· Score: 1
As others have already said, Linux does support HT. Does anyone have experience from both HT and "real" SMP machines for desktops? I mean, people constantly rave about how nice SMP:s are for desktop work with low latency etc. How does a single HT processor compare to a real SMP box in the interactivity department?
Sadly, it looks like this project is mainly for bomb research. I read that the tungsten wires used to generate the field melt because of the high currents, so there's no way to run these bursts quickly enough for commercial power generation, because you must change the wire array after every burst.
Not to mention the NV42: Just put on your ear plugs, point your computers exhaust towards that pesky neighbours house, start doom IV and blow him away!
But with the Nvidia drivers you're not really "building from source" - you're building a wrapper around the drivers from source, the drivers are still a big.o binary file.
This matters to me. I would like to see Open Software on Open Hardware, and thus would rather not give my money to NVidia.
Unfortunately I think the ATI drivers are also closed src. I don't know if any of the current generation cards offer full support for their features with open source drivers.
Slashdot Mirror
There's a fairly detailed interview with bruce eckel ("famous" guy who has written c++ and java books and sits on the C++ standards committee) at artima on why he likes python (I linked to the last part of the article series, since that contains links to the previous ones).
ASCI Q and Pink where both in fact installed in the same time frame. Actually, both of these can be classified as clusters. I guess the point was that ASCI Q is a cluster of DEC, uh Compaq, sorry, HP Alpha machines (=expensive) while Pink is a x86 cluster running Linux.
ethernet has about 100 us (microseconds) latency, regardless if it's 100 or 1000 Mbit. Specialized cluster interconnects, like myrinet, scali and quadrics, have about 3-7 us. Of course, those interconnects cost a fortune too.
Imagine having the fastest processesor on earth, and then take that chip and use it to do the calculation of x1++ (thats x1 = x1 +1 for you non-C'ers)and looping it a few Trillion times.
Well, regardless of the chip performing the calculation, I'd prefer doing
x1 += a_few_trillion;
instead of looping.
But yes, I get your point. "Supercomputer CPU's", like in the earth simulator, are optimised for SIMD operations. Given a vectorizing FORTRAN compiler and a BLAS library handwritten in assembler to take advantage of SIMD features on the chip, I guess you can get some decent performance from such chips.
That's what SMTP Auth and similar stuff is for, i.e. all your mail goes through your own/companys/whatever mailserver (you authenticate to the mailserver and possibly the connection can be encrypted too) wihtout having to allow open relaying on the mailserver.
One of the benefits of the RMX proposal is that it would create and incentive to deploy SMTP Auth too.
Actually, from FORTRAN 90 onward, pointers (somewhat restricted compared to C, but still) as well as dynamic memory allocation are part of the language. Making FORTRAN as slow as C.. :-)
Anyway, I don't think the difference between C and fortran is so big anymore. Compiler tech has improved, and in fact most fortran compilers have identical backends with the C/C++ compiler from the same manufacturer.
Exocet is obsolete. Modern ASM:s fly at mach 3 and do evasive manouvers. There's no chance that a phalanx (or goalkeeper, for that matter) will shoot something like that down, except by pure chance.
Of course, that's why phalanx is being replaced with RAM.
FORTRAN is still kicking in the engineering market. A large fraction of numerical code is still being written in fortran.
Contrast that with Red Hat for example, who are yanking support for their 'personal' operating systems 12 months from the time of their release. It's kind of sobering to think that Red Hat 8.0, 7.3, 7.2, 7.1 are end of lifed in six months from now and 9.0 a mere four months after that.
While this might save Red Hat money in the short term I have to wonder what impact it will have on customer confidence. Even assuming you bought it on the very day of release at best you get twelve months maximum of bug fixes, which isn't very much especially if you were planning on deploying it. If some horrible exploit is discovered ten months from now you're screwed. You might appeal to the community to produce an updated patch, but you still forfeit any QA testing or automated RHN update that you would have gotten before.
Well that's why RH has introduced their Enterprise server/workstation/advanced server/whatever line of products, with all kinds of support options. They have, IIRC, a 5 year support lifetime. Granted, they cost a lot too, but RH has to make a profit just like any other company. IMHO, their strategy is entirely reasonable:
(a) Use the "normal" RH distro, get the latest and greatest software for free, and help RH and the free software community improve the software (by filing bug reports, if nothing else).
(b) Buy the Enterprise/blahblah products and get a high quality product with a long lifetime and support.
(c) If you're just a parasite expecting to get everything for free without contributing anything, sod off.
#3 on the list is a machine made by linux networx for LLNL. It consists of 1100+ dual P4 nodes with quadrics interconnect. And linuxBIOS on all the nodes.
An interesting thing about this is that it doesn't consist of the ususal 1U rack cases, but they are special cases where the boards are mounted vertically (better airflow).
...But you couldn't kick your slashdot habit, huh? :)
But yes, I understand you. It's really irritating when you're trying to concentrate on something and people keep calling you, babbling about some shit that you really couldn't care less about.
Well, if there only was a RFID tag taped on the guns, what's stopping you from just removing it?
Over here in the civilized world, you need a licence to own and operate a gun. Seems to work quite well. Yes, the government has a database of all guns and their owners. Boo!! hiss!! So what? The idea that a gun-equipped public could prevent the government from turning against its own people was perhaps a good idea 150-200 years ago when wars were won by the side who had the most guys with handguns. That time is long gone.
The MTA is dead, sadly. It was an interesting architecture, but Cray never did any work on it after Tera bought cray and changed name.
Most scientific applications store their data in some proprietary format, or with the help of something like netcdf or hdf. It's not like the article is suggesting that they should turn all supercomputers into relational databases.
Yes, but fuel cells aren't 100 % efficient either. IIRC, they are usually around 60-70 % efficient.
No, the next version will be XXP. And as diversifying for different customers is all the rage we'll have different version, such as:
XXP home edition - without network transparency for those people who are too stupid to understand that X doesn't use tcp/ip sockets on localhost, and thus believe that taking away network transparency would improve performance.
XXP professional - X as we know it.
XXP enterprise advanced datacenter server edition - Very expensive version for those that like to pay through their noses for no reason. This is also targeted for servers, but as the server and client concept is somewhat backward for X, the only things this includes are xterm, xclock and xeyes.
Uh, does the US military have to come up with acronyms for absolutely everything?
:)
When I was in the army, we cleaned our barracks 3 times per day (if we were there of course), and we had no damn acronym for it!
Umm, no, Apache is licensed under the (drumroll...) Apache License! Basically it's the BSD license with an added clause saying that if you sell a derived product, you may not call it apache.
But other than that, or point is correct. You don't have to release your modifications to GPL code if you keep it for yourself. IIRC RMS supported himself by doing proprietary in-house modifications to emacs back in the eighties (or was it the early nineties?).
Well, they say the CPU is liquid-cooled, not necessarily water-cooled. Looking at the pictures there's no pump, so I guess it's some kind of heat pipe which transports heat from the cpu to the case by convection only. Better pictures at http://www.signum-data.de/pdf/ct_signum_engl.pdf
The hush technologies itx box was already mentioned. Another german thing which is very similar, but uses a P4, is the Signum Data FutureClient. Expensive as hell, though.
As others have already said, Linux does support HT. Does anyone have experience from both HT and "real" SMP machines for desktops? I mean, people constantly rave about how nice SMP:s are for desktop work with low latency etc. How does a single HT processor compare to a real SMP box in the interactivity department?
Sadly, it looks like this project is mainly for bomb research. I read that the tungsten wires used to generate the field melt because of the high currents, so there's no way to run these bursts quickly enough for commercial power generation, because you must change the wire array after every burst.
But still, it's a cool device. yay!
Yes. You can get it here. :)
Not to mention the NV42: Just put on your ear plugs, point your computers exhaust towards that pesky neighbours house, start doom IV and blow him away!
But with the Nvidia drivers you're not really "building from source" - you're building a wrapper around the drivers from source, the drivers are still a big
This matters to me. I would like to see Open Software on Open Hardware, and thus would rather not give my money to NVidia.
Unfortunately I think the ATI drivers are also closed src. I don't know if any of the current generation cards offer full support for their features with open source drivers.