PostgreSQL at least uses write ahead logging, and the manual claims that it's a standard technique. See here.
I guess you're correct that the anticipatory scheduler can reduce write performance when you're waiting on fsync(), but the logic seems to be that the improved read performance will more than compensate for this (some other poster said 15 % improvement for "typical" database loads).
This is from an ab initio program. One of the better ones actually, VASP. Did anybody say that functions should have a limited number of arguments? Huh?
But otherwise, IMHO VASP source is rather nice for a fortran program.
Having used a quite large number of statistics program, I would like to recommend R in addition to the ones in the list you linked to. R originally was a clone of S-plus, but I wouldn't be surprised if it has surpassed S-plus today. One of the nice features of R is that it's crossplatform (most Linux distros include it) and free.
The bad thing, at least for people used to excel, is that R has no fancy GUI.
The DRI drivers only support 3d acceleration with <= 9200 cards. If you want 3d acceleration with the newer cards you have to use the closed source driver.
Personally, I have a radeon 9200 and I use the DRI drivers. Works just fine for me.
Maybe you're supposed to use the old russian trick of making a bonfire under the engine?;-)
Re:A great success story of Linux on the desktop..
on
Rome Moving to Linux
·
· Score: 2, Funny
They switch the name of openwriter to "Cervantes", the famous Spanish writer, author of Don Quijote. They did the same for all of the applications and...
I don't know about you, but I'd find it quite confusing if every application was named "Cervantes".;-)
I don't think it will work. Reliability, bandwidth and latency tend to be the biggest problems with purpose built clusters. I imagine a setup like the one which you describe is going to be much worse in these aspects. Not to mention that most simulation software is written for homogeneous clusters, i.e. all the computers in the clusters should have an identical software environment, and the slowest node limits the speed of the entire calculation.
A similar but slightly less pie-in-the-sky thing would be to use the lab computers for batch processing during the night. Lab computers tend to be centrally administered and perhaps identical computers could be grouped together so that CPU cycles aren't wasted to the extent that they would be in a homogeneous environment.
Most western democracies have a so called parliamentary system, where the actual leader of the country is the prime minister and not the president/king/queen. The big exeptions are IIRC USA, France and Russia where the president has lots of power.
The crucial difference is that in a parliamentary system parliament can fire the PM and his cabinet any time it wishes. This helps keep the policies of the PM and his cabinet in line with the best interests of the country, whereas in a presidential system the president can do pretty much anything until the next election approaches.
I say this, because RMS warned several years ago, I think, that putting your code under XFree licence (so it must have been mandatory) was putting it at odds, and was very dangerous, in case the project behaves badly. That's just what happened.
No, what RMS is afraid of is a proprietary fork, which the BSD style licenses allow. You still own the copyright to your own contributions, regardless whether they are GPL or XFree86 licenced. Without your permission, a project cannot change the license of something you contributed (unless you gave them the copyright). The difference is that with the GPL a proprietary fork is not allowed.
One could use a voting method like Condorcet where strategic voting doesn't work. Of course, back in real life something like that will never happen because it would reduce the power of the two major parties.:(
Yeah, it's quite a lame idea. I mean, if you're going to learn to program, the time required to install your programming environment of choice is minuscule compared to the time you have to spend actually learning how to program.
Heck, I first installed linux back in 1995 or so because I wanted a real programming enviroment.
I got started with C++ (ok I had done some hello world kind of stuff in gwbasic, qbasic, pascal and such), and armed with my current level of hindsight I have to say I disagree.
My reason is that C++ is a very complicated language. There are a zillion completely non-obvious and subtle ways in which C++ manages to bite your ass.
Now for the question about which language is the best to get started in, I guess it depends on what you value as important to teach first. If you think it's important that people get some understanding in how computers actually work, know how to implement e.g. linked lists then C might be a good choice. OTOH, if you feel it's more important that they learn more high-level concepts like OO then a language like Python would probably be better.
I mean, look at the most popular shows on tv - home improvement, auctions, win a million dollars, 'lifestyle' programs... I think I'll move to Europe.
Sorry, tv is as bad over here.. The current fad seems to be "reality tv". I mean, how totally uninterresting can it be to watch a bunch of blokes and a few fat bitches hang around in an appartment, doing nothing particular? And still people watch it. In droves. It's so sad, so sad..
Luckily a few of us have seen the light and instead of wasting our time watching tv we contribute to the interresting discussions at slashdot.org instead.
Yup, I think Intel is between a rock and a hard place.
- drop ia64 prices to compete with amd64 => forget the wet dreams about huge profit margins on the high end stuff.
- introduce something amd64-like (yamhill?) => ia64 market fails, no huge profit margins on high end stuff. Also big PR loss. Lots of money wasted on ia64 project.
- keep going as is. amd64 commodizes high end market => no huge profit margins.
Regardless of what they do, they lose! Not a nice place to be in. IMHO the least bad option for intel (and perhaps users too) would be the first one, but it appears that they are going for the second one?
Slashdot Mirror
Umm, why? In the brave new world of Unicode, there are lots of symbols to choose from, albeit unpronouncable.
A bit like Prince, you know?
PostgreSQL at least uses write ahead logging, and the manual claims that it's a standard technique. See here.
I guess you're correct that the anticipatory scheduler can reduce write performance when you're waiting on fsync(), but the logic seems to be that the improved read performance will more than compensate for this (some other poster said 15 % improvement for "typical" database loads).
Try this:
L ATT_INI,EXCTAB, &T , &, E2, &
SUBROUTINE ELMIN( &
P,WDES,NONLR_S,NONL_S,W,W_F,W_G,WUP,WDW,LATT_CUR,
T_INFO,DYN,INFO,IO,MIX,KPOINTS,SYMM,GRID,GRID_SOF
GRIDC,GRIDB,GRIDUS,C_TO_US,B_TO_C,SOFT_TO_C,DIP,E
CHTOT,CHTOTL,DENCOR,CVTOT,CSTRF, &
CDIJ,CQIJ,CRHODE,N_MIX_PAW,RHOLM,RHOLM_LAST, &
CHDEN,SV,DOS,DOSI,CHF,CHAM,DESUM,XCSIF, &
NSTEP,NELMLS,LMDIM,NIOND,IRDMAX,NBLK,NEDOS, &
TOTEN,TOTENL,EFERMI,LDIMP,LMDIMP,LTRUNC)
This is from an ab initio program. One of the better ones actually, VASP. Did anybody say that functions should have a limited number of arguments? Huh?
But otherwise, IMHO VASP source is rather nice for a fortran program.
Having used a quite large number of statistics program, I would like to recommend R in addition to the ones in the list you linked to. R originally was a clone of S-plus, but I wouldn't be surprised if it has surpassed S-plus today. One of the nice features of R is that it's crossplatform (most Linux distros include it) and free.
The bad thing, at least for people used to excel, is that R has no fancy GUI.
Hear Hear!
Guess how I found about the terrorist attacks on 2001-09-11?
"-Why the #$%& does slashdot load so slowly today?"
"-Oh..."
You know, that's the real selling point of those wifi equipped laptops: You can stream porn from your server to the toilet even if your wife is home.
;-)
Right?
I wrote a short article about this at the other site a while ago. Especially, take a look at this article, which goes into considerable detail.
PostgreSQL has support for writing stored procs in python. I haven't used it, so I can't comment on the quality of the implementation.
The DRI drivers only support 3d acceleration with <= 9200 cards. If you want 3d acceleration with the newer cards you have to use the closed source driver.
Personally, I have a radeon 9200 and I use the DRI drivers. Works just fine for me.
Maybe you're supposed to use the old russian trick of making a bonfire under the engine? ;-)
They switch the name of openwriter to "Cervantes", the famous Spanish writer, author of Don Quijote. They did the same for all of the applications and
I don't know about you, but I'd find it quite confusing if every application was named "Cervantes".
I don't think it will work. Reliability, bandwidth and latency tend to be the biggest problems with purpose built clusters. I imagine a setup like the one which you describe is going to be much worse in these aspects. Not to mention that most simulation software is written for homogeneous clusters, i.e. all the computers in the clusters should have an identical software environment, and the slowest node limits the speed of the entire calculation.
A similar but slightly less pie-in-the-sky thing would be to use the lab computers for batch processing during the night. Lab computers tend to be centrally administered and perhaps identical computers could be grouped together so that CPU cycles aren't wasted to the extent that they would be in a homogeneous environment.
Why do I, as a non-native english speaker, have to correct people who supposedly speak english as their native language?
Most western democracies have a so called parliamentary system, where the actual leader of the country is the prime minister and not the president/king/queen. The big exeptions are IIRC USA, France and Russia where the president has lots of power.
The crucial difference is that in a parliamentary system parliament can fire the PM and his cabinet any time it wishes. This helps keep the policies of the PM and his cabinet in line with the best interests of the country, whereas in a presidential system the president can do pretty much anything until the next election approaches.
It lacked a centralized CVS
It still lacks a centralized revision control system, since one of the prime virtues of BitKeeper is that it is decentralized.
You might hit a bump and spill your drink!
I say this, because RMS warned several years ago, I think, that putting your code under XFree licence (so it must have been mandatory) was putting it at odds, and was very dangerous, in case the project behaves badly.
That's just what happened.
No, what RMS is afraid of is a proprietary fork, which the BSD style licenses allow. You still own the copyright to your own contributions, regardless whether they are GPL or XFree86 licenced. Without your permission, a project cannot change the license of something you contributed (unless you gave them the copyright). The difference is that with the GPL a proprietary fork is not allowed.
One could use a voting method like Condorcet where strategic voting doesn't work. Of course, back in real life something like that will never happen because it would reduce the power of the two major parties. :(
It's just troll feed..
I'd like to have tab completion for apt-get before that happens!
Hmm, maybe one could cook together something with zsh?
Yeah, it's quite a lame idea. I mean, if you're going to learn to program, the time required to install your programming environment of choice is minuscule compared to the time you have to spend actually learning how to program.
Heck, I first installed linux back in 1995 or so because I wanted a real programming enviroment.
I got started with C++ (ok I had done some hello world kind of stuff in gwbasic, qbasic, pascal and such), and armed with my current level of hindsight I have to say I disagree.
My reason is that C++ is a very complicated language. There are a zillion completely non-obvious and subtle ways in which C++ manages to bite your ass.
Now for the question about which language is the best to get started in, I guess it depends on what you value as important to teach first. If you think it's important that people get some understanding in how computers actually work, know how to implement e.g. linked lists then C might be a good choice. OTOH, if you feel it's more important that they learn more high-level concepts like OO then a language like Python would probably be better.
I mean, look at the most popular shows on tv - home improvement, auctions, win a million dollars, 'lifestyle' programs... I think I'll move to Europe.
Sorry, tv is as bad over here.. The current fad seems to be "reality tv". I mean, how totally uninterresting can it be to watch a bunch of blokes and a few fat bitches hang around in an appartment, doing nothing particular? And still people watch it. In droves. It's so sad, so sad..
Luckily a few of us have seen the light and instead of wasting our time watching tv we contribute to the interresting discussions at slashdot.org instead.
Uh oh, forget that last paragraph..
Awwwwwww....
Yum! I'm sure they taste good!
Yup, I think Intel is between a rock and a hard place.
- drop ia64 prices to compete with amd64 => forget the wet dreams about huge profit margins on the high end stuff.
- introduce something amd64-like (yamhill?) => ia64 market fails, no huge profit margins on high end stuff. Also big PR loss. Lots of money wasted on ia64 project.
- keep going as is. amd64 commodizes high end market => no huge profit margins.
Regardless of what they do, they lose! Not a nice place to be in. IMHO the least bad option for intel (and perhaps users too) would be the first one, but it appears that they are going for the second one?