Slashdot Mirror

• Cut back gradually. Eliminate a cup or glassful a day rather than going "cold turkey."
• Keep a log to see how much caffeine you consume. Remember to count medications and supplements. Experiment with your intake to see how you feel both physically and psychologically
• Limit your intake to 200-300 mg of caffeine per day.
• Substitute herbal tea, hot or cider or decaf coffee for caffeinated drinks.
• Be active or be still - run, walk, bike ride, swim, do yoga or meditate.
• Eat regular meals
• Stop smoking - caffeine and cigarettes often go together.
• Ask others in your house or office to decrease their caffeine with you. There is strength in numbers.
• Remember that coffee does NOT help you to sober up after drinking alcohol.

cow milk is 87% water. Top link on google for "milk composition".

I thought even the trolls were able to use google these days...

There have been systems like this using DSPs for years; I've worked with some of them in the mid-late-80s. The Bell Labs DSP supercomputer had a tree-structured network of DSPs, and I think a more general-purpose processor at the root, back when 25 MFLOPS was blazingly fast for a single floating-point DSP, and the box had 127 of them, producing up to about 2 GFLOPS on applications like FFTs. It was one of the fastest ray-tracers around.

There have been systems like this using DSPs for years; I've worked with some of them in the mid-late-80s. The Bell Labs DSP supercomputer had a tree-structured network of DSPs, and I think a more general-purpose processor at the root, back when 25 MFLOPS was blazingly fast for a single floating-point DSP, and the box had 127 of them, producing up to about 2 GFLOPS on applications like FFTs. It was one of the fastest ray-tracers around.

http://brighton.ncsa.uiuc.edu/~prajlich/wall/8x5pi cs.html/

Also, extra credit if you can run this as a Beowulf cluster, though of course the really interesting Playstation cluster system uses the Emotion Engine graphics chip to do number crunching rather than the relatively tame CPU so it's not a straight Beowulf system. This new Playstation version is 1/3 the size and 1/2 the price of the ones NCSA used, so it's an even more attractive idea...

Someone has already mentioned apache, but the Internet itself was developed mainly in an open mindset.

What you don't see is the thousands of small utilities used inside companies that come from open-source. These utilities are not distributed to the public, so they are not affected by the GPL, but this cannot be done with closed source.

Also I can't think of anything that Microsoft made that was innovative, that they didn't steal and use their monopoly power to kill the original, that was usually better.

I'm running gnome and have lots of utilities that I don't have on MS. Some of these are available from third parties, but the quality is not as good. One main example is the multiple desktop. I use six different desktops to bounce around different projects that I work on in one day. This has helped me tremendously. Grant you, that this is old, but I first saw this with fvwm and that was opensource. Maybe it was copied from something else but that was not were I've seen it. I've found many utilities more easy to use in the opensource arena than the closed source.

Also where do you think IE came from? the same place as Mozilla, which is derived from Netscape which was derived from Mosaic which is another innovative opensource product. If all you look at is Word, Powerpoint, Excel and Photoshop, I can see you having this view, but there is a lot more out there that comes from opensource, but since it doesn't have a logo on it, you just don't see it.

Open your eyes.

Another thing to try--LLVM can use g++ as a front-end to compile C++, and can often do further optimizations to speed things up even more. LLVM also has a C back-end, so you can benchmark other C compilers against each other too!

Of course, the real fun will begin when it has completed Java and C# front-ends. :)

About 6 months ago, I decided to disprove the claim some people were making about spam increasing exponentially. So I started on a project of plotting my personal spam over the past few years. I was rather disturbed to discover the exponential fit was better than the quadratic fit. Since then, it's tapered off, but you might still check out the plot. Also, I started plotting spam and viruses system-wide. Lots more plots are available (though only for a few months history, rather than years).

About 6 months ago, I decided to disprove the claim some people were making about spam increasing exponentially. So I started on a project of plotting my personal spam over the past few years. I was rather disturbed to discover the exponential fit was better than the quadratic fit. Since then, it's tapered off, but you might still check out the plot. Also, I started plotting spam and viruses system-wide. Lots more plots are available (though only for a few months history, rather than years).

First, find the first 1, 3, 7, or 9 after the first ten digits after the decimal. Take the preceding 9 digits, and run it through a Prime Number Checker. (The algorithm is in the source).

Really, the hardest part is determining the farthest decimal points of e. Here's the formula: limn->infinity (1 + 1/n)n.

It's lazy, impatient, and full of hubris! BTW, I get a finder's fee.

Tom Siebel gave my school (UIUC) THIRTY million for a new building

... convince the engine that it should keep running.

Reminds me of Bomb #20

The marketing jargon goes a bit over the top, but it isn't impossible to translate code for one ISA to an intermediate form, optimize it, and then generate code for another ISA. I don't know that it's revolutionary either. Note that LLVM takes a similar approach, and has a very simple intermediate form. I hear someone on their team is working on a PPC front-end, and as for language front-ends, Java and C# is in the works.

Getting back to Transitives, in July 2001, they claimed to already be doing x86->MIPS translation, which bodes well for x86->PPC. However, doing things efficiently the other way around is tougher. And of course you need to support or translate a ton of the native OS API calls etc. It'll be interesting to see for Windows on Linux (for example) if they require a copy of Windows to run the binaries.

However the theoritical future has nothing to do with now. At this point, precomiled code is more efficient (in some cases quite a lot) and Java does not provide access to accelerated features. There is a REASON that Doom 3, UT 2004, etc are written in a language that precomiles to native code. Both seek to be cross platform (and UT 2004 is to an amazing extent) however Java is NOT the right tool for them at this time.
Thanks ALOT!
This is what I try to say here all the time, problem is that /. is overrun by java nuts.

I would also like to add:

1. JVM != Java. See gcj. Or mips2java. Of course, because the JVM architecture is optimized for Java code, C/C++ code doesn't get as efficiently compiled into JVM code. This is no advantage of Java, it is a poor design of the JVM.
2. Check out LLVM. These people are essentially trying to do (they are not yet there but it is one of their essential goals and no marketing blurb) what Java promised for years. Front-end language independent.

Last bad job I voluntarily left, back in 1991, I gave 2 weeks notice, then counted down the seconds. I still get nightmares about the place, sometimes.

Points well taken, though I seem to recall that Newton's predictions for Mercury's orbit didn't really fit until Einstein's G.R. accounted for the difference, and the solution did not involve matter, "dark" or "missing".

IBM Thinkpad A30p

'Nuff said

have you seen the facial detail and animation?

Actually, yes I have. It is almost as good as the Doom 3 facial animations. But I would certainly say the facial geometry of the HL2 models is significantly worse than the models for Doom 3. Not only that, but the unified lighthing system for Doom 3 accentuates the geometry making the faces incredibly lifelike.

Doom 3 face screenshot
Half-life 2 face screenshot

Doom 3 certainly blows HL2 out of the water, and HL2 hasn't even been released yet. And Carmack is already working on Doom 3 engine updates, some of which will significantly help in texturing skin to make people look as close to a photograph as you can get with this technology. Non-unified lighting will significantly hurt HL2 in the "engine war" if you can call it that. Pasted textures on low-geometry faces is not going to cut it, especially when directly comparing it to Doom 3.

have you seen the facial detail and animation?

Actually, yes I have. It is almost as good as the Doom 3 facial animations. But I would certainly say the facial geometry of the HL2 models is significantly worse than the models for Doom 3. Not only that, but the unified lighthing system for Doom 3 accentuates the geometry making the faces incredibly lifelike.

Doom 3 face screenshot
Half-life 2 face screenshot

Doom 3 certainly blows HL2 out of the water, and HL2 hasn't even been released yet. And Carmack is already working on Doom 3 engine updates, some of which will significantly help in texturing skin to make people look as close to a photograph as you can get with this technology. Non-unified lighting will significantly hurt HL2 in the "engine war" if you can call it that. Pasted textures on low-geometry faces is not going to cut it, especially when directly comparing it to Doom 3.

The earth is round....
http://www.talkorigins.org/faqs/flatearth.html

We'll never fly....
http://www.manwillneverfly.com/

We never went to the moon....
http://www.cen.uiuc.edu/~akapadia/moon.html

Gee.... This sounds awefully familiar.... Oh wait here...
http://slashdot.org/article.pl?sid=04/08/26/131921 5&tid=126&tid=1

LLVM Looks like cool stuff, I'd love a single tool that can show me a bunch of stuff about all my apps - call graphs, code paths, ultra esoteric remote super optimizers.

Imagine a system that would allow you to run the code analysis and optimization on a bunch of machines that might be partially idle (a la distcc, but for running apps not compiling them) - I'd love to have more computing power available to me to tune my code (while it is running) than I have to run it. Swwweeeeeet.

Of course by now I'm sure every floor has a Comp Sci major who has figured out how to break through the protections on the media service

As a well-heeled undergrad in CS at NIU, I would say that this is unlikely.

Most of our CS students are incompetent at any real hacking like the reverse-engineering problem you describe (including me). This is in part due to our outdated-since-1985 curriculum which strongly emphasizes coding ASM and COBOL on our IBM S/390 mainframe. Of course, almost none of our undergrad curriculum has anything to do with security, networking, copy-protection, or reverse-engineering. People in our Java course (CSCI480) don't know what an IP address is or why DNS is important, for god's sake! And don't even ask a junior or senior what even a "buffer overflow" is, despite having had 2 courses in C.

There's probably a couple dozen of our CS students who are actually thinking about getting around the protection on this service, and of those, 1 or 2 are actually trying to do something about it. The rest of them are probably doing homework or drinking and playing games, because most NIU CS'ers, in my experience, go into CS thinking that "playing games" == "writing games" (which obviously is not true). And most of them, in truth, want to play games, not write them - and that isn't the hacker spirit by any means.

Give the Ruckus Network to the genius kids who form the pride of Illinois -- then I'd agree with you. :)

Simple. It means that *this* dictionary has adopted it as a word, even though it wasn't. It does happen on occasion, like with words like "ain't". Here's a page that confirms the "non-wordness" (ha, ha) of the word.

Double extortion, in fact. The RIAA gets money, the uni is free from lawsuits.

It may be possible to ask for a refund of certain portions of your student fees. Illinois does that for the fees funding various registered student organizations or services (e.g. student legal dept).

Not only that, but they invented HAL!! Too bad it was at a university in Japan rather than at UIUC.

Though, It does appear that they're confusing evil robots with evil computers, at least as far as naming goes. But wait... This is Japan we're talking about? This might make some sense.

For anyone who happens to be reading this, my original rant was about people who go crazy trying to obfuscate their email addresses and have them fill out web-forms to contact them. I said I wasn't afraid to just put my email address (menscher@uiuc.edu) online. (You may also reach me at: menscher@fnal.gov, dmenscher@yahoo.com, and, if you want to bypass SpamAssassin/ClamAV, at menscher@mail.physics.uiuc.edu. )

Or, if you really like webforms, go to http://www.itg.uiuc.edu/people/ and click the "send email" link by my name.

SpamAssassin and ClamAV are free and work well. Speaking of which, there's a release candidate for SA-3.0.0 available now. Go test it!

For anyone who happens to be reading this, my original rant was about people who go crazy trying to obfuscate their email addresses and have them fill out web-forms to contact them. I said I wasn't afraid to just put my email address (menscher@uiuc.edu) online. (You may also reach me at: menscher@fnal.gov, dmenscher@yahoo.com, and, if you want to bypass SpamAssassin/ClamAV, at menscher@mail.physics.uiuc.edu. )

Or, if you really like webforms, go to http://www.itg.uiuc.edu/people/ and click the "send email" link by my name.

SpamAssassin and ClamAV are free and work well. Speaking of which, there's a release candidate for SA-3.0.0 available now. Go test it!

For anyone who happens to be reading this, my original rant was about people who go crazy trying to obfuscate their email addresses and have them fill out web-forms to contact them. I said I wasn't afraid to just put my email address (menscher@uiuc.edu) online. (You may also reach me at: menscher@fnal.gov, dmenscher@yahoo.com, and, if you want to bypass SpamAssassin/ClamAV, at menscher@mail.physics.uiuc.edu. )

Or, if you really like webforms, go to http://www.itg.uiuc.edu/people/ and click the "send email" link by my name.

SpamAssassin and ClamAV are free and work well. Speaking of which, there's a release candidate for SA-3.0.0 available now. Go test it!

Hello. I would like to contact you. Is menscher@uiuc.edu the best address to do so?

Thank you, good sir. I will be contacting you shortly at menscher@uiuc.edu.
menscher@uiuc.edu

Hello. I would like to contact you. Is menscher@uiuc.edu the best address to do so?

Thank you, good sir. I will be contacting you shortly at menscher@uiuc.edu.
menscher@uiuc.edu

Hello. I would like to contact you. Is menscher@uiuc.edu the best address to do so?

Thank you, good sir. I will be contacting you shortly at menscher@uiuc.edu.
menscher@uiuc.edu

IBM p-series 690

I feel like a huuuuuge nerd understanding that joke.

For those who don't, here's a quick view on red/blue shift

Possibly. There has been discussion in recent years that they were warm-blooded in some ways. More info here.

This was tried around 15 years ago by Virtuality. But back then, the helmets were absolutely massive, with a mini CRT monitors for each eye. At the companies peak, there were Virtuality booths close to every Underground station in central London. You could pay around 7 pounds for 20 minutes play. Although the games were simple, they were fun. One game was a first-person shooter, where you tried to shoot flying pterodactyls while trying to avoid being snatched or shooting other players.

Atari and the other console makers also jumped onto the VR bandwagon, even though the headsets were much lighter (later versions of the Virtuality helmet.

Obviously, you could do the same thing today, with consumer VR hardware, but the problem is cost. Consumers are more aware of the cost of playing in an arcade vs. playing at home. If the average game plays for one unit of currency for three minutes, and one person wants to play for three hours, thats 120 units of currency. For three months play, that amount of money would allow you to buy buy a PC + VR headset + broadband. Plus with headsets being as small and light as they are, they would very easily be stolen/broken. And that's not taking into account having to pay for parking, expensive drinks/snacks, worry about your belongings being stolen, your car being broken into, being mugged on the way home, or spend time finding a parking space.

I'm not arguing against distance learning per se, only against the type of people who so often seem to think it's a good idea, and the type of schools that seem to cater to them.

And here's some more links for security and privacy.

Second: you're almost certainly going to have to put it out to bid. For example, at UIUC, the bid limit is $28,100. Anything over that *must* go to bid unless you can provide a really good reason why you have to "sole source" it.

Now, you need to start thinking about stuff. First off, forget the number of nodes. You need to start by thinking about how they'll be used. Like, how much communication will there be? A few large packets, or many small ones? Myrinet is nice and fast, but will increase your costs by 50% over gigE. Similarly, you need to figure out how much ram to put in each node. How many processors, keeping in mind they'll be competing for the memory bandwidth. 32 bit vs 64 bit. The list goes on and on....

Of course, you didn't give any details, which means you probably don't have a clue. So maybe start by purchasing a couple of test systems and benchmarking your code on them, to see where your bottlenecks are.

Good luck. Sounds like you'll need it.

And perhaps another lesser known display is NCSA's 20-tile wall display that measures 12' x 9'. This is driven by Linux boxes and dates back at least 3 years.

According to this site Infiniband is about 2-4x slower than local CPU to CPU copy on an Itanium 2 (for values that fit into CPU cache, for some values larger than the cache it is actually faster to go across the Infiniband link!) and about 10x faster than messaging over GigE.

I had the opportunity to speak with him (among a group of probably 20 other people) after Mr. Gates spoke at the University of Illinois last semester. He seemed like a very pleasant man and I was very impressed. Kind of expecting an arrogant and rude man (especially after the security crew basically told us that he told them to get us the hell out of there), he was very friendly and was willing to answer some questions and take some photos. He probably spoke with us for 5-10 minutes before his security people ushered him away.

One of the guys in our Mac user group at Illinois was smart enough to bring a camera. (I, unfortunately, forgot mine that night. Oh well...) A picture can be found here. I was standing about 2 feet to the left of the field of view, so I'm not in it. Also, there are about 20 uniformed and plainclothed police in the general area.

<a href="http://www.cen.uiuc.edu/~akapadia/moon.html" >"fake" pic idiocy</a>

I'm a student at http://www.lis.uiuc.edu/, which is, in part, an Information Science school. Our infrastructure is almost entirely Microsoft. If I tried to reconfigure the bios on lab computers to boot from cd I'd be looking for a new computer lab. I luckily have an office across the street with my own machine running SUSE. I logon to the Windows lab machine, connect to a remote window on my Linux machine, and work in a browser window. With an ssh connection I can upload working documents to my ftp space and grab them from there to print in the lab.

Granted, this isn't the best solution and it won't suit those who can't run their own machine on a fast network, but if you have a Linux machine in a dorm room, it may work depending on the campus firewall configuration.

It's alway fun to see people react to Linux for the first time. I think they expect to see a Unix prompt or some arcane GUI.

Once it was clear that light moves at a finite speed, an English geologist, John Michell realized that one could imagine an object with a gravitational escape velocity greater than c. Such an object would appear black. Of course, the term "black hole" didn't appear until much later.

Google is your friend - http://archive.ncsa.uiuc.edu/Cyberia/NumRel/BlackH oleHistory.html

Paris recommends...

Does this mean that the applications running on the "old" clusters, presumably using some flavor of MPI to communicate between nodes, will have to be ported somehow to become multithreaded applications ?

NCSA still has plenty of "old" style clusters around. Two of the more aging clusters, Platinum and Titan are being retired, to make room for newer systems like Cobalt. Indeed, the official notice was made just recently--they're going down tommorrow. However, as the retirement notice points out, we still have Tungsten, Copper, and Mercury (Terragrid). Indeed, Tungsten is number 5 on the Top 500, so it should provide more than enough cycles for any message-passing jobs people require.

So, anyone has any insights as to why/how this matters for the programmers ?

What it means is that programming big jobs is easier. You no longer need to learn MPI, or figure out how to structure your job so that individual nodes are relatively loosely-coupled. Also, jobs that have more tightly-coupled parallelism are now possible. The older clusters used high-speed interconnects like Myrinet or Infiniband (NCSA doesn't own any Infiniband AFAIK, but we're looking at it for the next cluster supercomputer). Although they provided really good latency and bandwidth, they aren't as high-performing as shared memory. Also, Myrinet's ability to scale to huge numbers of nodes isn't all that great--Tugsten may have 1280 compute nodes, but a job that uses all 1280 nodes isn't practical. Indeed, untill recently the Myrinet didn't work at all, even after partitioning the cluster into smaller subclusters.

This new shared-memory machine will be more powerful, more convienient, and easier to maintain than the cluster-style supercomputers. Hopefully it will allow better scheduling algorithms than on the clusters too--an appaling number of cycles get thrown away because cluster scheduling is non-preemptive.

I'd also like to point out some errors in the Computerworld article. NCSA is *currently* storing 940 TB in near-line storage (Legato DiskXtender running on an obscenely big tape library), and growing at 2TB a week. The DiskXtender is licenced for up to 2 petabytes--we're coming close to half of that now. The article therefore vastly understates our storage capacity. On the other hand, I'd like to know where we're hiding all those teraflops of compute--35 TFLOPS after getting 6 TFLOPS from Cobalt sounds more than just a little high. That number smells of the most optimistic peak performance values of all currently connected compute nodes. I.e. - how many single-precision operations could the nodes do if they didn't have to communicate, everything was in L1 cache, we managed to schedule something on all of them, and they were all actually functioning. Realistically, I'd guess that we can clear maybe a quarter of that figure, given machines being down, jobs being non-ideal, etc. etc. etc.

As a disclaimer, I do work at NCSA, but in Security Research, not High-Performance Computing.

Does this mean that the applications running on the "old" clusters, presumably using some flavor of MPI to communicate between nodes, will have to be ported somehow to become multithreaded applications ?

NCSA still has plenty of "old" style clusters around. Two of the more aging clusters, Platinum and Titan are being retired, to make room for newer systems like Cobalt. Indeed, the official notice was made just recently--they're going down tommorrow. However, as the retirement notice points out, we still have Tungsten, Copper, and Mercury (Terragrid). Indeed, Tungsten is number 5 on the Top 500, so it should provide more than enough cycles for any message-passing jobs people require.

So, anyone has any insights as to why/how this matters for the programmers ?

What it means is that programming big jobs is easier. You no longer need to learn MPI, or figure out how to structure your job so that individual nodes are relatively loosely-coupled. Also, jobs that have more tightly-coupled parallelism are now possible. The older clusters used high-speed interconnects like Myrinet or Infiniband (NCSA doesn't own any Infiniband AFAIK, but we're looking at it for the next cluster supercomputer). Although they provided really good latency and bandwidth, they aren't as high-performing as shared memory. Also, Myrinet's ability to scale to huge numbers of nodes isn't all that great--Tugsten may have 1280 compute nodes, but a job that uses all 1280 nodes isn't practical. Indeed, untill recently the Myrinet didn't work at all, even after partitioning the cluster into smaller subclusters.

This new shared-memory machine will be more powerful, more convienient, and easier to maintain than the cluster-style supercomputers. Hopefully it will allow better scheduling algorithms than on the clusters too--an appaling number of cycles get thrown away because cluster scheduling is non-preemptive.

I'd also like to point out some errors in the Computerworld article. NCSA is *currently* storing 940 TB in near-line storage (Legato DiskXtender running on an obscenely big tape library), and growing at 2TB a week. The DiskXtender is licenced for up to 2 petabytes--we're coming close to half of that now. The article therefore vastly understates our storage capacity. On the other hand, I'd like to know where we're hiding all those teraflops of compute--35 TFLOPS after getting 6 TFLOPS from Cobalt sounds more than just a little high. That number smells of the most optimistic peak performance values of all currently connected compute nodes. I.e. - how many single-precision operations could the nodes do if they didn't have to communicate, everything was in L1 cache, we managed to schedule something on all of them, and they were all actually functioning. Realistically, I'd guess that we can clear maybe a quarter of that figure, given machines being down, jobs being non-ideal, etc. etc. etc.

As a disclaimer, I do work at NCSA, but in Security Research, not High-Performance Computing.

Does this mean that the applications running on the "old" clusters, presumably using some flavor of MPI to communicate between nodes, will have to be ported somehow to become multithreaded applications ?

NCSA still has plenty of "old" style clusters around. Two of the more aging clusters, Platinum and Titan are being retired, to make room for newer systems like Cobalt. Indeed, the official notice was made just recently--they're going down tommorrow. However, as the retirement notice points out, we still have Tungsten, Copper, and Mercury (Terragrid). Indeed, Tungsten is number 5 on the Top 500, so it should provide more than enough cycles for any message-passing jobs people require.

So, anyone has any insights as to why/how this matters for the programmers ?

What it means is that programming big jobs is easier. You no longer need to learn MPI, or figure out how to structure your job so that individual nodes are relatively loosely-coupled. Also, jobs that have more tightly-coupled parallelism are now possible. The older clusters used high-speed interconnects like Myrinet or Infiniband (NCSA doesn't own any Infiniband AFAIK, but we're looking at it for the next cluster supercomputer). Although they provided really good latency and bandwidth, they aren't as high-performing as shared memory. Also, Myrinet's ability to scale to huge numbers of nodes isn't all that great--Tugsten may have 1280 compute nodes, but a job that uses all 1280 nodes isn't practical. Indeed, untill recently the Myrinet didn't work at all, even after partitioning the cluster into smaller subclusters.

This new shared-memory machine will be more powerful, more convienient, and easier to maintain than the cluster-style supercomputers. Hopefully it will allow better scheduling algorithms than on the clusters too--an appaling number of cycles get thrown away because cluster scheduling is non-preemptive.

I'd also like to point out some errors in the Computerworld article. NCSA is *currently* storing 940 TB in near-line storage (Legato DiskXtender running on an obscenely big tape library), and growing at 2TB a week. The DiskXtender is licenced for up to 2 petabytes--we're coming close to half of that now. The article therefore vastly understates our storage capacity. On the other hand, I'd like to know where we're hiding all those teraflops of compute--35 TFLOPS after getting 6 TFLOPS from Cobalt sounds more than just a little high. That number smells of the most optimistic peak performance values of all currently connected compute nodes. I.e. - how many single-precision operations could the nodes do if they didn't have to communicate, everything was in L1 cache, we managed to schedule something on all of them, and they were all actually functioning. Realistically, I'd guess that we can clear maybe a quarter of that figure, given machines being down, jobs being non-ideal, etc. etc. etc.

As a disclaimer, I do work at NCSA, but in Security Research, not High-Performance Computing.