have always assumed that the simulations of gravity and galaxy formation was done with relativistic mathematics. Instead they have used approximations using newtonian theories? WTF? No wonder they came out wrong!
Keep in mind, not only are relativistic simulations hard (or maybe just expensive as I think there are some good relativistic hydrocodes these days), but Newtonian Mechanics isn't that bad. This isn't like using epicycles and the like to calculate orbits. Newtonian mechanics works very well in some situations, and we have a pretty good handle on what those situations are. So astronomers weren't just being lazy by doing Newtonian simulations.
And this paper hasn't exactly been "proven" correct yet. There are lots of astronomers who would love to be rid of dark matter, but it fits lots of evidence, so it may not be so easy.
One minor correction to that. At least based on my experience, the people who don't want to change because they "know the ins and outs of MSIE" actually know no such thing. They are just stuck on something and too lazy/stuburn to learn new things. Web browsers aren't like a lot of other software. Most people do a very very small number of simple things with it and could probably do those same things with almost any browser. The forward and back buttons work the same on pretty much everything. Same goes for clicking on links. And that covers a lot of people (with some bookmarks thrown in maybe). They just do not like even having the chance of learning something new.
That being said, I do know Windows users who have completely switched to firefox over IE (my parents have done it on my advice for example). So not everyone is as bad as I make it sound above. (And to be totally fair, I still use pine for my email, so I'm not always at the front of the curve for picking up new programs either).
Why is it that the people who seem to complain about this are the ones who also complain about there being nothing good to watch on TV? If there isn't anything good, then why do you care if they put in a flag that prevents you from doing something with someone else's content that they paid to create and distribute?
I happen to like a lot of things on TV and think this is horrible, unjust, and dangerous. And, this type of flag will be used to prevent me from using something I paid for (my TiVo for example) the way I want to. If people are so afraid of someone watching or taping their content, maybe they just shouldn't ever broadcast or distribute it at all. That should keep it nice and safe without trampling on everyone else.
Why would customers "fear" this? Is it just a case of extreme addition to this gadget? It's very simple. From what I understand, they signed a usage agreement/contract, and if Tivo violates it, they just sue. If I watched TV and had one of these recorder thingies, I'd do just that. I don't understand why there'd be "fear" over Tivo changing the contract mid-stream.
I wish it worked that way, but the usage agreement with TiVo is almost certainly the type that says TiVo can change the terms whenever they want or something similar. Is it right? Of course not. I am happy about this as a TiVo user? Hell no. But do I think they broke the contract they made with their users? Probably not since it is almost definitely very one sided.
which is the size of your cluster?. because i'm just a poor student, my cluster only had 3 nodes.
I'm a poor grad student myself, but luckily, my department has a pretty nice cluster. We just got 13 new dual opterons bringing our cluster up to 29 dual processor nodes. There are a bunch of people using it, but I've used more hours than anyone:) (83367.00 cpu hours total so far, and my thesis isn't anywhere near done).
i used MPICH2 for my thesis (simulation of fluids by lattice boltzmann method, c++, ch_p4) if you are interested i can send it to you:-) .
Jumpshot is nice, you must compile MPICH with mpe enabled (the right flags are really tricky).
I'm definitely going to check out jumpshot, but since I pretty much have to use lam-mpi (I've had strange trouble with mpich on our beowulf), I fear it may not work for me. I'm actually doing my thesis on hydro sims of galaxy clusters using a F77 code. Now for shameless promotion, the code is Zeus-MP.
I'm really concerned about this windows 2003 cluster thing, they are going to make an incompatible version, they are NOT going to use SSH, or rsh since process managament is completly different, and a lot of things will broke. sure we differ in our views about minimum specs of mpi, but we agree about the broken windows-mpi thing.
I could not agree more. We've been lucky so far in the parallel/high performance computing world that Microsoft (or anyone else for that matter) hasn't tried to come and take over any standards we rely on. I really hope this isn't the beginning of the end of that.
python IS required for mpiexec and mpd, NFS is required, unless you copy your executables around in every node manually, and not use any Parallel-IO capability.
Java is needed for Jumpshot-4, a graphical utility that shows profiling data.
in your debunking points you don't say a word about ssh, why?.
I really meant it as just providing a little correction since I use mpi (lam specifically) every day, not a debunking.
I didn't mention ssh because I forgot to. But it isn't a requirement. You can use rsh/telnet. You would be nuts to do so, but you still can. Same with the nfs thing. You can do without it. It doesn't make sense, but you can.
And you may need java for a specific profiling tool, but you do not need it for mpi. I am of course interested in Jumpshot-4 now that I've heard about it. Of course probably just like all the other cool tools it won't work with my fortran 77 code.
Services for Unix provide python?, rsh/ssh?, nfs client and server?, fortran compilers?, automake autoconf?. java? because the last time i installed mpi, these were the requirements.
Actually, I don't think any of those are required for mpi (at least not for lam or mpich). Not sure about python since I don't use it, but it is always installed. Fortran is a compile time option (which for some strange reason is turned off by default in the Gentoo package). nfs isn't strictly needed. You just need some way for all processes to share a filesystem or files. And automake and company are only needed to compile mpi. java is no needed (not even sure if it is supported).
All that said, I can't imagine why anyone would want to do parallel computing on Windows (except maybe to prove that you code is totally cross platform even to crippled platforms).
What big-name projects is Perl being used for today?
Slashcode and fink come to mind. But big projects isn't what Perl is about. It is all the little scripts that people use to keep things running where Perl does its best.
Even Pugs, a Perl 6 compiler, is written in Haskell rather than Perl.
Well, since the Perl 5 interpreter is written in C, this isn't such a change. Perl just isn't meant for systems programming.
First off, an iPod battery lasts a hell of a longer then a walkman and it recharges relatively quickly. When I frequently used a walkman / diskman batteries definitely influenced how I used the device. Batteries only lasted so long, and I had to buy them from the store. That inconvenience prevented me from using the device on countless occasions.
If that is true then there is something seriously wrong with my ipod mini. I used to use a walkman (back in highschool) and I could go days without new batteries. My ipod mini just barely lasts me the day if I use it a lot and if I just use it during my walk in to work, I can get maybe two days out of it.
I would love to use hdf5 for the software I use and maintain (ZEUS-MP
but it only supports Fortran 90 (or C). So people like me stuck with Fortran 77 still have to use hdf4, which is nowhere near as nice as hdf5.
While I agree that etc-update is pretty bad, there is a really nice alternative that is in portage: dispatch-conf. It is a pretty nice way to deal with differences in config files, and can use rcs on you config files, which is a really great idea.
So far, there are only two companies that will even claim to have made a profit from open source. They are IBM, who may have reason to fudge the numbers, and Red Hat, who claims to have scraped some skin from its teeth. All the others are either losing money or folding.
Is that really a fair thing to say? Apple has used a ton of open source programs for OSX (even though the final product also contains lots of non-free stuff). And they have made money.
Google uses Linux which is free to make money. Tivo use Linux (although I don't know if they actually make money. Linksys sells (and I assume does pretty well) products like the WRT54g which run Linux.
I don't want to go crazy with examples, but the point is that lots of companies make money off of free software and some of them probably even give things back, they just don't always make money the way you expect a software company would.
Telling people that you won't serve them because you don't like their choice of browser makes non-technical people think that the open-source community is just a bunch of exclusionist assholes.
And, you know what? If they just looked at you, they would be right.
Okay, I understand that the original suggestion is not practical (even though I had some issues when I switched to XHTML1.0 that made me consider banning IE from my page), but the purpose is not to be exclusionist. It has nothing to do with liking someone's choice of browser.
All else being equal, I couldn't care less what browser you use. But, unfortunately Microsoft is the one not happy with that browser agnostic situation. They are the ones who step all over web standards and make things hard for everyone. So if someone could get away with banning IE from their site, it wouldn't be an exclusionist asshole thing to do. It would be just not letting a bully come in to make a mess and push everyone around.
So you and people "in the know" form an opinion once and never change it regardless of new information? Doesn't really sound like most of the people who are into technology I knew. Also sounds a bit scary. I guess I don't need to read any reviews of IE 7 or Windows vista since I remember using Windows 3.11 (is that the right number?) and didn't like it.
There is very little University money in researh that doesn't make money.
That isn't true in all fields (I hope). Fields like astronomy tend not to do too much directly profitable research, but somehow they still find money for us. Which by the way, we greatly appriciate.
Saying there isn't a God when there is is in fact rejecting God...
Actually, if we want to get really careful with our language, I will slightly correct what I said originally. I can (and should) reject things that are not real, e.g., the tooth fairy.
It is nothing personal against the tooth fairy. I just reject the notion that such a thing exists.
>What the editors and most people on slashdot fail to realize, likely because they cannot get over their own rejection of God, is that real Christians are not "self righteous" in fact the two are mutually exclusive.
Very impressive use of the "no true Scotsman" fallicy. That should invalidate your post too much for me to bother responding, but sometimes I can't help myself.
Non-belivers (or even liberal believers) can have problems with things that conservative Christians (or memebers of other religions) do because they disagree with it in principle. It doesn't have to involve some imagined guilt or chip on our shoulder for 'regecting god'. In fact, some of people don't even think there is a god to reject. And it is pretty hard to feel bad about rejecting something you don't think is there in the first place.
>So, the stable universal API for drivers exists, and it IS source code.
Based on talk I've heard (probably on Slashdot, so I realize it is pretty dumb to take that as representing reality) I sort of assumed the problem was the API could actually change. If that isn't the case, then things are completely the fault of these third party binary driver authors.
Not to mention the fact that binary drivers are a pretty bad idea anyway.
Which version? I think the comment about standards is related to the fact that there is no guaranteed stability in Linux's device driver interface, and this causes problems for binary drivers.
Actually, the post I was originally replying to mentioned distro specific (Red Hat and SuSe) kernels.
As for the lack of stability for the driver interface, I agree with you, that is potentially problematic, and one can argue for or against its usefulness.
My emachine intel extreme graphics chipset is NOT supported by UBUNTU . I go to the intel site and they only have drivers for a certain kernel for Redhat or Suse and no other distro. Get this. I read a suse messageboard and they can't even get it to work. Bahhhh.
We need to make it easier for hardware support which is another reason for people pulling their hair out with linux.
Device drivers are a kernel issue. And there is already a nice standard there: a vanilla kernel from kernal.org
So the problem isn't a lack of standards. It is a lack of people following the standards. And less choice in distros will not change that.
I agree that I want more bandwidth for different reasons that the "average" broadband user wants it. I would like it so when I log onto my work machine at home X-Windows forwarding and things like that work a little better (although I bet latency hurts me there too). I would like my remote backups to go a bit faster. I would like emerge --sync to go a bit faster too.
But an average user (say my mother) could definately use more bandwidth when she emails a picture to someone, or receives a picture in an email. A 5 megapixel digital camera makes a pretty big file. And most people with those 5 megapixel cameras take a whole lot of pictures.
Intel designed its compilers, which translate software programs into machine-readable language, to degrade a program's performance if operated on a computer powered by an AMD microprocessor.
I like AMD (nearly every machine I currently use has an AMD processor), and I tend to believe most of those statements, but the one about compilers just doesn't ring true based on my experience.
I use the Intel Fortran compiler for my simulation code on AMD machines, and I get better performance with it (although it takes forever to actually compile) than I get with any other compiler (at least 20-30% faster than with the Portland Group Compiler, which is supposed to be the AMD prefered compiler). It doesn't seem likely that this is actually a hobbled compiler that should produce even faster code.
There are definitely a LOT of things that require a real word processor. Some of them even need the auto-indexing and other fancy feature of Word.
I don't think I've ever seen anything that requires the features of a word processor (other than reading other people's word processor written documents).
A text editor and latex (plus bibtex) can do everything (and more) that a wordprocessor can do.
Slashdot Mirror
And this paper hasn't exactly been "proven" correct yet. There are lots of astronomers who would love to be rid of dark matter, but it fits lots of evidence, so it may not be so easy.
That being said, I do know Windows users who have completely switched to firefox over IE (my parents have done it on my advice for example). So not everyone is as bad as I make it sound above. (And to be totally fair, I still use pine for my email, so I'm not always at the front of the curve for picking up new programs either).
I could not agree more. We've been lucky so far in the parallel/high performance computing world that Microsoft (or anyone else for that matter) hasn't tried to come and take over any standards we rely on. I really hope this isn't the beginning of the end of that.
I didn't mention ssh because I forgot to. But it isn't a requirement. You can use rsh/telnet. You would be nuts to do so, but you still can. Same with the nfs thing. You can do without it. It doesn't make sense, but you can.
And you may need java for a specific profiling tool, but you do not need it for mpi. I am of course interested in Jumpshot-4 now that I've heard about it. Of course probably just like all the other cool tools it won't work with my fortran 77 code.
All that said, I can't imagine why anyone would want to do parallel computing on Windows (except maybe to prove that you code is totally cross platform even to crippled platforms).
Well, since the Perl 5 interpreter is written in C, this isn't such a change. Perl just isn't meant for systems programming.
I would love to use hdf5 for the software I use and maintain (ZEUS-MP but it only supports Fortran 90 (or C). So people like me stuck with Fortran 77 still have to use hdf4, which is nowhere near as nice as hdf5.
While I agree that etc-update is pretty bad, there is a really nice alternative that is in portage: dispatch-conf. It is a pretty nice way to deal with differences in config files, and can use rcs on you config files, which is a really great idea.
Google uses Linux which is free to make money. Tivo use Linux (although I don't know if they actually make money. Linksys sells (and I assume does pretty well) products like the WRT54g which run Linux.
I don't want to go crazy with examples, but the point is that lots of companies make money off of free software and some of them probably even give things back, they just don't always make money the way you expect a software company would.
All else being equal, I couldn't care less what browser you use. But, unfortunately Microsoft is the one not happy with that browser agnostic situation. They are the ones who step all over web standards and make things hard for everyone. So if someone could get away with banning IE from their site, it wouldn't be an exclusionist asshole thing to do. It would be just not letting a bully come in to make a mess and push everyone around.
Actually, if we want to get really careful with our language, I will slightly correct what I said originally. I can (and should) reject things that are not real, e.g., the tooth fairy.
It is nothing personal against the tooth fairy. I just reject the notion that such a thing exists.
Very impressive use of the "no true Scotsman" fallicy. That should invalidate your post too much for me to bother responding, but sometimes I can't help myself.
Non-belivers (or even liberal believers) can have problems with things that conservative Christians (or memebers of other religions) do because they disagree with it in principle. It doesn't have to involve some imagined guilt or chip on our shoulder for 'regecting god'. In fact, some of people don't even think there is a god to reject. And it is pretty hard to feel bad about rejecting something you don't think is there in the first place.
>So, the stable universal API for drivers exists, and it IS source code.
Based on talk I've heard (probably on Slashdot, so I realize it is pretty dumb to take that as representing reality) I sort of assumed the problem was the API could actually change. If that isn't the case, then things are completely the fault of these third party binary driver authors.
Not to mention the fact that binary drivers are a pretty bad idea anyway.
Actually, the post I was originally replying to mentioned distro specific (Red Hat and SuSe) kernels.
As for the lack of stability for the driver interface, I agree with you, that is potentially problematic, and one can argue for or against its usefulness.
Device drivers are a kernel issue. And there is already a nice standard there: a vanilla kernel from kernal.org
So the problem isn't a lack of standards. It is a lack of people following the standards. And less choice in distros will not change that.
I agree that I want more bandwidth for different reasons that the "average" broadband user wants it. I would like it so when I log onto my work machine at home X-Windows forwarding and things like that work a little better (although I bet latency hurts me there too). I would like my remote backups to go a bit faster. I would like emerge --sync to go a bit faster too.
But an average user (say my mother) could definately use more bandwidth when she emails a picture to someone, or receives a picture in an email. A 5 megapixel digital camera makes a pretty big file. And most people with those 5 megapixel cameras take a whole lot of pictures.
I like AMD (nearly every machine I currently use has an AMD processor), and I tend to believe most of those statements, but the one about compilers just doesn't ring true based on my experience.
I use the Intel Fortran compiler for my simulation code on AMD machines, and I get better performance with it (although it takes forever to actually compile) than I get with any other compiler (at least 20-30% faster than with the Portland Group Compiler, which is supposed to be the AMD prefered compiler). It doesn't seem likely that this is actually a hobbled compiler that should produce even faster code.
I don't think I've ever seen anything that requires the features of a word processor (other than reading other people's word processor written documents).
A text editor and latex (plus bibtex) can do everything (and more) that a wordprocessor can do.