Most of the debugging I do is on IRIX using their cvd environment, ProDev Workshop. It's not great, but it has the following nice features:
There is a "Recompile" menu option that groks Makefiles. So if you make a quick change to a file, you can be up and running reasonably quickly. I know Visual Studio does this, too, but I haven't really delved into ddd, gdb, or Kdevelop to see how they work.
It can give you call trees. This is kinda nice, although it doesn't do the greatest job.
It can do profiling. For hardcore stuff you use perfex (like oprofile, except better). This can help you see your hotspots and so on.
Conditional breakpoints, watch windows of arbitrary length, etc. However, its watch windows aren't nearly as nice as those in Visual Studio for nested objects and arrays. The hideous behavior of nagging you for every damn step you do while an array is out of scope renders the array tool nearly useless.
What is missing are things like an array visualizer (I think): I had one of these for Digital Fortran on the Alpha. You could plot data from, say a 2D array, as a plane with the values in the array as the 3rd dimension. Nice.
Also, something like the KDE frontend for valgrind would be nice.
Smarter memory visualization would be nice. For example, on a simulation that we use, we have one machine that does the crunching, and another that acts as a communication frontend to some hardware. The cruncher and the communicator talk to each other over reflective memory. It would be nice to be able to describe the protocol they use to the debugger, and then see the message data in realtime. Or if you could simply see color coding for memory age or different colors for new vs. delete (or malloc vs. free), that would be okay.
Basically, I think the integration of a performance profiler, a memory visualizer, and a debugger with nice features would be very helpful. On the timing critical things that we do, adding printfs is not an option, because the delays they can induce absolutely break things. Sure, you can buffer them, and we often do that (even conditional breakpoints are often too heavyweight), but something that just gives you an idea of how things are progressing on the machine would be very welcome. Hell, for some things we output differential signals and then use a logic analyzer with advanced triggers to do bughunting.
I dunno. Sometimes I feel like I work in a unique environment, where something only really ``works'' when (a) it's functionally correct and (b) performance is good. For a lot of stuff, (a) is sufficient, and even that sometimes slides. I really like tweaking stuff to see if you can squeeze out those extra few microseconds.
A few thoughts from someone who has used and likes debuggers, which apparently is reasonably rare.
Postol's problems are kinda moot now, anyway. His problems were with a version of an interceptor from TRW that was ultimately down-selected (er, not picked). The missile defense program went with Raytheon's interceptor, the EKV (exo-atmospheric kill vehicle) which had a string of 9 intercepts in a row broken in mid-December, when the kill vehicle failed to detach from the booster (probably the booster, and therefore Boeing's, fault).
Basically, Postol is whining that the technology to discriminate decoys from real warheads will never be sufficient, to which basically every goddamn person in the universe who works on the stuff has replied, wait and see, weaknesses are being addressed.
Sorry, but face it, if you ARE on my System *I* am the one who tells you what you can do and can't do. I AM the cop on that system and if you don't behave I make sure you can't do much damage.
And this is why IT people get slapped down.
For the work I do, my program paid $$$ for a machine, and if I ever found some sysadmin weenie patching things I didn't want patched, or mucking about with configurations, by God he'd feel the Hand of Death closing in fast and sure. Yeah, I might be doing something stupid, and obviously the IT weenies are free to offer suggestions, but being stupid with your own property is a perogative of the buyer.
Now, on multiuser systems, there are obviously more complexities, and very likely the BOFH is both Lord and Master. But they should always remember that they are providing a service to the people paying them: as Customer, these people can be stupid with their own money.
Re:Let's hope this means the end of veal
on
Lab-Grown Steak
·
· Score: 1
I'm vegan, but even if you aren't, you'll be hard pressed to find people that support the truly unethical treatment of cows for the production of veal.
Here's one. Sure, I'd prefer it was humane to the cow, but what I'd prefer, and what I know isn't happening, is that it be:
Antiseptic
Safe for the employees
Cows not pumped full of antibiotics and crazy shit like that
I'd willingly pay more for cow if I could be assured that a better job was done in the raising, feeding, and slaughtering of the animal, but I could give two shits about the animal's feelings, really: thank god for the pointy bit of the food pyramid where I reside.
If they could get this slab-grown meat working correctly, then I think all of your concerns and all my concerns could be addressed. You wouldn't have the horrific killing floors (yeah, I read Fast Food Nation), and unless the animal rights crazies start up with ''Equal Protection for Cow Parts'' or something, everybody'd be happy.
I have no inherent reason to believe the latest results any less or any more than the results that came out of the Clinton Administration, "proving" that condom use reduced the incidence of STDs, or anything else of a political nature, for that matter.
I'll bite. How is showing that condoms reduce the incidence of STDs political? Either:
They do
They don't
No conclusion can be drawn at this time
Now, the actions you take given this information are almost certainly political, but the information itself is not.
Either way, I'd volunteer to do my part, in the interests of science of course.
No. Their L2 is probably optimized for density: that's the usual way. L1 optimized for latency, L2 for density.
Sure, it's not *small*, but even though it's a lot of transistors, it's not as big as you'd make it out. Anyway, it hardly matters because it's not like you were going down to S-mart and buy 2 shrink-wrapped ones. They'll just come in your hugely expensive SGI.
I should know, I bought one just recently and will be getting another soon. IRIX is the champ for very low latency, processor-intensive code.
Yes, the United States will spend approximately $320 billion on defense next year. Note, however, that about $200 billion or so is personnel and operations and stuff: there's about $63 billion for procurement and another $41 billion or so in Research and Development. From this source, the GDP of the U.S. is projected as $10.481 trillion dollars next year, which gives us defense spending as a percentage of GDP at 3%.
Whoopdy-fucking-do.
Look at Mandatory spending (projections, 2002 (table S-3)):
Social Security: $443 billion
Medicare and Medicaid: $362 billion
Means-tested entitlements: $119 billion
Other (I *love* this): $121 billion
For a total of $1.046 trillion dollars.
Remember, we have a volunteer military, and no one will sign up unless they get paid decently and have reasonable benefits, so those costs will always be higher than nations with compulsary service.
Note, too, that the United States makes every damn weapon everyone everywhere uses, and we like it that way. This of course means that the United States bears the brunt of these RD costs, but it also means we can limit who gets what equipment elsewhere, and with what degree of smarts.
There is that, but what you see day in and day out is more like this:
The company I work for has come kind of contract/union for guys who are supposed to move stuff: computers, gear, etc. which has resulted in this ridiculous situation where you have to either:
a) wait around for the union guy to come around and move the stuff you need to get on with *your* work b) move it anyway, on the sly
I'm talking stuff like PCs and power supplies and test gear. I got a dirty look from some guy for carrying a monitor from one end of a lab to a different desk outside (he was delivering power supplies).
In fact, it's sometimes so ridiculous it's fun: I actually had another employee run interference (go over and talk to a union guy), while we carried a laser printer around behind his back!
I've heard "Would you like it if we took *your* job?" See if you can, is my response. Just try to come up with the skills that I have, while your skills consist of picking shit up and then losing it somewhere.
And the union techs that we have aren't much better. Lots of the guys have 18+ years, so there's no way they're getting laid off, but on the other hand they've been doing the same thing for those 18 years, maybe, and they are *bored*. I understand this, but it leads to them sometimes doing maybe 4-10 hours of work a *week*. Now, I really like some of these guys (although I'm a little disappointed at this work ethic, even though I understand it), but this friendship serves a dual purpose, in that I can lean on it a little to get my stuff done on time.
On another front, we had a tech in a lab who'd been doing very good work, and everybody liked it. However, it turned out he didn't have sufficient seniority for this (apparently cherry) position, so he got swapped out for some older guy. Now, the replacement guy does okay work too, but everyone's a little cranky and unhappy that Mike (the 1st guy) got yanked away like that. When union's can't be flexible about shit like that, it just raises opinions against them.
For me, I'd never join a union. Never. They have their place, but they gell into this rigid, ugly structures of power and entitlement far too quickly. I dunno how the guys at the bottom of the seniority ladder hang on: I guess they must dream of the day when *they* can abuse their positions.
My problem with one of these is that Itanium2s are so damn hot, I wonder about the density of computational power. Granted, an individual Itanium2 smokes the hell out of an R14k, but since our 600MHz R14ks use on the order of 15W and I think an Itanium2 is up around 150W, we can get many times the computrons (R14ks aren't fast, but they aren't *that* slow) per Watt. Cooling is kind of limiting factor, because these machines go in labs, not prepared computer rooms.
Right now we have a couple of 8-way Onyx2s and we're in the process of building an 8-way Origin 300. For the kind of work we do (realtime simulations), where latency is king, we prefer to put each part of a task on its own CPU, so having 8 processors is nice, whereas having an 8-way Itanium2 would be prohibitively costly to cool, although it'd be nice for when we just crank: see 483/499 SpecInt/FP base for 600MHz R14k vs. 810/1350 SpecInt/FP base for 1GHz Itanium2.
I *would* like to move to Linux from IRIX, I think. I really like the IRIX realtime support (all the REACT stuff), but I am tired of poor tool support and limited lack of updates, etc. I think $500k worth of machines (in *that* lab) would warrant a better resolution of some issues we've had.
Finally, I very much anticipate the day that these Linux scalability improvements filter down into something like a 4-way Clawhammer system. That system could very likely do a lot of our work (at what, maybe $20k for a system?) that we now drop $50k on for SGIs.
Ha, you think that Ian Malcolm was pessimistic? Read some Vinge: in his universe, the amount of complexity and technological sophistication a civilization can handle is dependent upon their distance from the galactic core (gravitation-dependent? I don't think he ever clearly indicates). It's an interesting mechanism that lets him make faster-than-light travel possible, but only at the galactic fringe.
Anyhoo, overwhelming billions of civilization deeper into the galaxy rise to a high level of technical sophistication, but then the underlying infrastructure of their civilizations eventually become so rotted with teetering layers upon layers, each with obscure bugs in them, that at some point these bugs align and civilization collapses into barbarism. The hero in his second book in this universe, "A Deepness in the Sky" lives tens (maybe hundreds?) of thousands of years by mostly staying relativistic, and sees these collapses firsthand a number of times.
Sometimes, I think Vinge is right: do we have the capacity to handle the kinds of emergent complexity that we will face as we become ever-more reliant on technology? Because even if we build tools to handle this complexity (say, VHDL compilers and simulators), what do we do if they have bugs in them (we have found several in our tools: it was damn frustrating). What happens when our lives become entirely dependent upon the functioning of tools that no one understands, because it's been too long since there were problems with them: see fer instance some of the crazy problems with various air traffic control systems. What if wholesale replacement isn't an option?
Dunno, interesting questions. Sometimes I lean one way, sometimes the other.
Well, loosely. But note that they are at least using SMP (2-way Xeon) nodes: that probably helps if their algorithms are smart. It would be very hard to try and built a link that provides you with the same benefit as tons of processors on the same board. Note that: Big Iron has multiple chips that are physically close and are engineered to communicate with each other quickly.
You cannot dismiss this with a simply "blah, just fast NICs". No. They. Aren't. When you have an 8-way SGI, you have 2, 4-cpu nodes with their own local memories (CC-NUMA: cache coherent non-uniform memory architecture), where latency to local memory is maybe 100ns, and to go 1 hop to the next node is another 100ns or whatever.
I know there are starting to be NUMA Linux machines, but Beowulfs aren't these machines, so any hop from one machine to the next will have enormously more latency than something that can simply borrow some data from the next processor over.
So yes, maybe bandwidth will catch up, but I don't see how latency really can. You can try with hypercube arrangements of links and so forth, but you can wire together Big Iron like this too (see the SGI super-dense server story from Slashdot a day or two ago).
Or maybe you were just talking about really old supercomputer technology, about which I know jack, so I'll just shut up now.
Not quite. See the Quadrics webpage. This machine is using the Qsnet product. You stick a card in a PCI slot on each machine (it's not obvious if you *have* to have a switch in between or not for only a 2-machine network), and use the provided library to send data around.
Therefore, it does not appear to be transparent.
Also, from using a similar interface that is ~$4k per, I'd guess that these are Officially Not Cheap, although maybe if you buy in the quantities that these guys do, you can make out okay.
Really, for most user-level applications, it's hard to beat Fast Ethernet. I paid $26 each for my cards, and $90 for the hub.
Er, you know that your employer *has* to let you vote , right?
So do this: vote an absentee ballot, then use the magic combination of the words "long commute to assigned voting spot", "gotta vote", and "buh bye" to fuck right off to the bar of your choice after you mail your ballot (many days before, since you often have to have the ballot in the recorder's hands by election day).
(3) Africa is really screwed, because even the increased volume of melting ice isn't making up for other losses in river volume. Soon Africa will become a wasteland like Ireland, where I found it impossible to get ice.
A couple of people from Apple are working on adding this to the FSF distro: apparently they have it in the version they use, and it was up to 6x faster.
Dunno about a prospective release date. See the gcc mailing list for details.
Re:Here is a copy of the changes...
on
More MS EULA Fun
·
· Score: 1
You acknowledge and agree that Microsoft may automatically check the version of the OS Product and/or its components that you are utilizing and may provide upgrades or fixes to the OS Product that will be automatically downloaded to your computer
Your honor,
Of course it's automated -- I never expected a Microsoft employee to individually select and present to my attention the files I might want to upgrade my Windows 2000 installation. In my mind, and in the minds of millions of citizen-consumers of the wordld, however, 'automatically' does not indicate 'without my permission', it merely indicates that there is some piece of software at the Microsoft end making selections without the oversight of a human being.
And consider this: I work in defense, and we have some old, old, old hardware, because what happens is that, once a product is done and all the testing and simulation gear is ready, nobody wants to change anything. And since missiles and stuff like that are out in the field for ages, and the tools basically run themselves over time, people forget how to get things working again if stuff breaks -- or there's simply no *way* to get things working again if stuff breaks.
We have transputers and ancient VAXen and analog computers and crazy shit like that from the 70s and 80s that were used in the simulation of a certain Navy Program. The one guy who understood it all retired, so now there's about six guys in a lab somewhere who have to recreate an entirely new lab from just whatever specifications remain. Sure, people knew this guy was going to retire eventually, but since a lot of these efforts have to be funded by the military, who get yanked around by Congress every couple of weeks, nothing serious was ever done. Heck, I worked on emulating a DR-11W interface (from a PDP-11) for awhile on that program, and we even have a few computers that use MECL logic (!).
Anyway, if there's ever some kind of software fault discovered in this program, *no* *one* can test it to see what might be happening.
http://www.nytimes.com/slideshow/2002/06/06/maga zi ne/09median.slideshow_1.html
That one.
Now look at the amount in savings these people have on their $54,400 a year.
1st Family: he's 28, she's 29. Savings: $22.5k. I guess that's not *too* bad, but they bought a $16k boat even though they've only saved 1/2 a year of their combined incomes...
2nd Family: he's 35, she's 33. Savings: $1000 in a savings account (i.e., NOTHING). They bought a huge TV, but their kids are like 15, 12, and 8 or something. Sorry, no college for you.
3rd Family: he's 52, she's 45. Look at the hours that poor guy works: upto a 100 a week! For $54k a year! Only $20k in savings, but over $400k in debt.
4th Family: he's 39, she's 28 (good job, buddy). Savings: over $100k (plus college fund), and they own their house and are working on a second one. Best so far.
5th Family: he's 59, she's 53. No mention of savings, but they sold a diner they had, and he mentions that their 401k did pretty well.
6th Family: he's 43, she's 43. Savings: ~$40k + $7k for a college fund. He's a programmer (only technical person so far).
7th Family: She's 35. No $$ figure, but she puts $600 a *month* into a savings account.
So families 4 and 7 seem to be doing pretty well off toward retirement, and probably family 5 as well. Family 3 works hard, but farming is even tougher than I thought.
WHAT ABOUT THE REST OF YOU! WHERE ARE YOUR SAVINGS?!
That's what I can't believe about this country. No one ever saves any goddamn money. They either expect a miracle to happen, or that someone else will cover for them. Like me. And you know what: *fuck* that.
Family number 6: you're technical, you should *get* this. You make $54.4k in Fargo, N.D., which is probably a *lot* of money for that area. Yet you have less than one year of income's worth of savings, and apparently are still paying off your house. WHY?! I've only been working for two years and soon I'll have that much, and I've had to pay off large amounts of student and other debt. Why why why why why?
I can understand that some people just can't save money: that eating and having a place to sleep just eat up their entire income. But then there's people like Family 2: the price of their TV is twice their TOTAL savings. And when they reach retirement age, what do they expect will happen? Keep working until they die? Get Social Security from a generation probably smaller than theirs? What?
Take some responsiblity for you financial future, you goddamn losers. I am not your savior, and the rest of the American public shouldn't have to be, either.
You can't just go grocery shopping for parts, then go home and pile them all on the floor and say, 'yup, that there's a cruise missile'.
As someone who has worked on these things, it isn't easy. Primitive rockets are reasonably easy (see Hamas, fer instance), but any kind of guidance is relatively hard. Also, cruise missiles must either be:
a) airdropped (Tomahawk) b) launched with a rocket booster (Tomahawk) c) be able to fly itself up from the ground on a runway (unlikely).
The development cost of even a rude cruise missile is pretty significant, as is the requisite experience. Far cheaper to just buy or steal missiles from bit players like North Korea or whomever. Even the smallest nations could presumably build a missile. Getting one with a reasonable degree of accuracy, however...
Slashdot Mirror
Ah well, I missed all the discussion.
Most of the debugging I do is on IRIX using their cvd environment, ProDev Workshop. It's not great, but it has the following nice features:
What is missing are things like an array visualizer (I think): I had one of these for Digital Fortran on the Alpha. You could plot data from, say a 2D array, as a plane with the values in the array as the 3rd dimension. Nice.
Also, something like the KDE frontend for valgrind would be nice.
Smarter memory visualization would be nice. For example, on a simulation that we use, we have one machine that does the crunching, and another that acts as a communication frontend to some hardware. The cruncher and the communicator talk to each other over reflective memory. It would be nice to be able to describe the protocol they use to the debugger, and then see the message data in realtime. Or if you could simply see color coding for memory age or different colors for new vs. delete (or malloc vs. free), that would be okay.
Basically, I think the integration of a performance profiler, a memory visualizer, and a debugger with nice features would be very helpful. On the timing critical things that we do, adding printfs is not an option, because the delays they can induce absolutely break things. Sure, you can buffer them, and we often do that (even conditional breakpoints are often too heavyweight), but something that just gives you an idea of how things are progressing on the machine would be very welcome. Hell, for some things we output differential signals and then use a logic analyzer with advanced triggers to do bughunting.
I dunno. Sometimes I feel like I work in a unique environment, where something only really ``works'' when (a) it's functionally correct and (b) performance is good. For a lot of stuff, (a) is sufficient, and even that sometimes slides. I really like tweaking stuff to see if you can squeeze out those extra few microseconds.
A few thoughts from someone who has used and likes debuggers, which apparently is reasonably rare.
Postol's problems are kinda moot now, anyway. His problems were with a version of an interceptor from TRW that was ultimately down-selected (er, not picked). The missile defense program went with Raytheon's interceptor, the EKV (exo-atmospheric kill vehicle) which had a string of 9 intercepts in a row broken in mid-December, when the kill vehicle failed to detach from the booster (probably the booster, and therefore Boeing's, fault).
Basically, Postol is whining that the technology to discriminate decoys from real warheads will never be sufficient, to which basically every goddamn person in the universe who works on the stuff has replied, wait and see, weaknesses are being addressed.
And this is why IT people get slapped down.
For the work I do, my program paid $$$ for a machine, and if I ever found some sysadmin weenie patching things I didn't want patched, or mucking about with configurations, by God he'd feel the Hand of Death closing in fast and sure. Yeah, I might be doing something stupid, and obviously the IT weenies are free to offer suggestions, but being stupid with your own property is a perogative of the buyer.
Now, on multiuser systems, there are obviously more complexities, and very likely the BOFH is both Lord and Master. But they should always remember that they are providing a service to the people paying them: as Customer, these people can be stupid with their own money.
Here's one. Sure, I'd prefer it was humane to the cow, but what I'd prefer, and what I know isn't happening, is that it be:
I'd willingly pay more for cow if I could be assured that a better job was done in the raising, feeding, and slaughtering of the animal, but I could give two shits about the animal's feelings, really: thank god for the pointy bit of the food pyramid where I reside.
If they could get this slab-grown meat working correctly, then I think all of your concerns and all my concerns could be addressed. You wouldn't have the horrific killing floors (yeah, I read Fast Food Nation), and unless the animal rights crazies start up with ''Equal Protection for Cow Parts'' or something, everybody'd be happy.
Good luck, scientist-types.
I'll bite. How is showing that condoms reduce the incidence of STDs political? Either:
- They do
- They don't
- No conclusion can be drawn at this time
Now, the actions you take given this information are almost certainly political, but the information itself is not.Either way, I'd volunteer to do my part, in the interests of science of course.
No. Their L2 is probably optimized for density: that's the usual way. L1 optimized for latency, L2 for density.
Sure, it's not *small*, but even though it's a lot of transistors, it's not as big as you'd make it out. Anyway, it hardly matters because it's not like you were going down to S-mart and buy 2 shrink-wrapped ones. They'll just come in your hugely expensive SGI.
I should know, I bought one just recently and will be getting another soon. IRIX is the champ for very low latency, processor-intensive code.
I had a friend who used to support Tandems. There was a fire in a bank where they were using them. The question was:
"Uh, our Tandem's on fire. Now what?"
I think the answer was to leave it alone...
Er, what's your point?
Note: billion == 1E12
Yes, the United States will spend approximately $320 billion on defense next year. Note, however, that about $200 billion or so is personnel and operations and stuff: there's about $63 billion for procurement and another $41 billion or so in Research and Development. From this source, the GDP of the U.S. is projected as $10.481 trillion dollars next year, which gives us defense spending as a percentage of GDP at 3%.
Whoopdy-fucking-do.
Look at Mandatory spending (projections, 2002 (table S-3)):- Social Security: $443 billion
- Medicare and Medicaid: $362 billion
- Means-tested entitlements: $119 billion
- Other (I *love* this): $121 billion
For a total of $1.046 trillion dollars.Remember, we have a volunteer military, and no one will sign up unless they get paid decently and have reasonable benefits, so those costs will always be higher than nations with compulsary service.
Note, too, that the United States makes every damn weapon everyone everywhere uses, and we like it that way. This of course means that the United States bears the brunt of these RD costs, but it also means we can limit who gets what equipment elsewhere, and with what degree of smarts.
There is that, but what you see day in and day out is more like this:
The company I work for has come kind of contract/union for guys who are supposed to move stuff: computers, gear, etc. which has resulted in this ridiculous situation where you have to either:
a) wait around for the union guy to come around and move the stuff you need to get on with *your* work
b) move it anyway, on the sly
I'm talking stuff like PCs and power supplies and test gear. I got a dirty look from some guy for carrying a monitor from one end of a lab to a different desk outside (he was delivering power supplies).
In fact, it's sometimes so ridiculous it's fun: I actually had another employee run interference (go over and talk to a union guy), while we carried a laser printer around behind his back!
I've heard "Would you like it if we took *your* job?" See if you can, is my response. Just try to come up with the skills that I have, while your skills consist of picking shit up and then losing it somewhere.
And the union techs that we have aren't much better. Lots of the guys have 18+ years, so there's no way they're getting laid off, but on the other hand they've been doing the same thing for those 18 years, maybe, and they are *bored*. I understand this, but it leads to them sometimes doing maybe 4-10 hours of work a *week*. Now, I really like some of these guys (although I'm a little disappointed at this work ethic, even though I understand it), but this friendship serves a dual purpose, in that I can lean on it a little to get my stuff done on time.
On another front, we had a tech in a lab who'd been doing very good work, and everybody liked it. However, it turned out he didn't have sufficient seniority for this (apparently cherry) position, so he got swapped out for some older guy. Now, the replacement guy does okay work too, but everyone's a little cranky and unhappy that Mike (the 1st guy) got yanked away like that. When union's can't be flexible about shit like that, it just raises opinions against them.
For me, I'd never join a union. Never. They have their place, but they gell into this rigid, ugly structures of power and entitlement far too quickly. I dunno how the guys at the bottom of the seniority ladder hang on: I guess they must dream of the day when *they* can abuse their positions.
My problem with one of these is that Itanium2s are so damn hot, I wonder about the density of computational power. Granted, an individual Itanium2 smokes the hell out of an R14k, but since our 600MHz R14ks use on the order of 15W and I think an Itanium2 is up around 150W, we can get many times the computrons (R14ks aren't fast, but they aren't *that* slow) per Watt. Cooling is kind of limiting factor, because these machines go in labs, not prepared computer rooms.
Right now we have a couple of 8-way Onyx2s and we're in the process of building an 8-way Origin 300. For the kind of work we do (realtime simulations), where latency is king, we prefer to put each part of a task on its own CPU, so having 8 processors is nice, whereas having an 8-way Itanium2 would be prohibitively costly to cool, although it'd be nice for when we just crank: see 483/499 SpecInt/FP base for 600MHz R14k vs. 810/1350 SpecInt/FP base for 1GHz Itanium2.
I *would* like to move to Linux from IRIX, I think. I really like the IRIX realtime support (all the REACT stuff), but I am tired of poor tool support and limited lack of updates, etc. I think $500k worth of machines (in *that* lab) would warrant a better resolution of some issues we've had.
Finally, I very much anticipate the day that these Linux scalability improvements filter down into something like a 4-way Clawhammer system. That system could very likely do a lot of our work (at what, maybe $20k for a system?) that we now drop $50k on for SGIs.
Ha, you think that Ian Malcolm was pessimistic? Read some Vinge: in his universe, the amount of complexity and technological sophistication a civilization can handle is dependent upon their distance from the galactic core (gravitation-dependent? I don't think he ever clearly indicates). It's an interesting mechanism that lets him make faster-than-light travel possible, but only at the galactic fringe.
Anyhoo, overwhelming billions of civilization deeper into the galaxy rise to a high level of technical sophistication, but then the underlying infrastructure of their civilizations eventually become so rotted with teetering layers upon layers, each with obscure bugs in them, that at some point these bugs align and civilization collapses into barbarism. The hero in his second book in this universe, "A Deepness in the Sky" lives tens (maybe hundreds?) of thousands of years by mostly staying relativistic, and sees these collapses firsthand a number of times.
Sometimes, I think Vinge is right: do we have the capacity to handle the kinds of emergent complexity that we will face as we become ever-more reliant on technology? Because even if we build tools to handle this complexity (say, VHDL compilers and simulators), what do we do if they have bugs in them (we have found several in our tools: it was damn frustrating). What happens when our lives become entirely dependent upon the functioning of tools that no one understands, because it's been too long since there were problems with them: see fer instance some of the crazy problems with various air traffic control systems. What if wholesale replacement isn't an option?
Dunno, interesting questions. Sometimes I lean one way, sometimes the other.
Here's an interesting comment by JMS, the creator of Babylon 5, with respect to gross vs. net for that program:
Subject: Re: ATTN JMS: Gross or net?
Well, loosely. But note that they are at least using SMP (2-way Xeon) nodes: that probably helps if their algorithms are smart. It would be very hard to try and built a link that provides you with the same benefit as tons of processors on the same board. Note that: Big Iron has multiple chips that are physically close and are engineered to communicate with each other quickly.
You cannot dismiss this with a simply "blah, just fast NICs". No. They. Aren't. When you have an 8-way SGI, you have 2, 4-cpu nodes with their own local memories (CC-NUMA: cache coherent non-uniform memory architecture), where latency to local memory is maybe 100ns, and to go 1 hop to the next node is another 100ns or whatever.
I know there are starting to be NUMA Linux machines, but Beowulfs aren't these machines, so any hop from one machine to the next will have enormously more latency than something that can simply borrow some data from the next processor over.
So yes, maybe bandwidth will catch up, but I don't see how latency really can. You can try with hypercube arrangements of links and so forth, but you can wire together Big Iron like this too (see the SGI super-dense server story from Slashdot a day or two ago).
Or maybe you were just talking about really old supercomputer technology, about which I know jack, so I'll just shut up now.
Not quite. See the Quadrics webpage. This machine is using the Qsnet product. You stick a card in a PCI slot on each machine (it's not obvious if you *have* to have a switch in between or not for only a 2-machine network), and use the provided library to send data around.
Therefore, it does not appear to be transparent.
Also, from using a similar interface that is ~$4k per, I'd guess that these are Officially Not Cheap, although maybe if you buy in the quantities that these guys do, you can make out okay.
Really, for most user-level applications, it's hard to beat Fast Ethernet. I paid $26 each for my cards, and $90 for the hub.
Very very carefully.
I've seen it done, and making circuit cards survive 30,000 g's is not trivial. See also: ERGM, Excalibur, MRM.
Er, you know that your employer *has* to let you vote , right?
So do this: vote an absentee ballot, then use the magic combination of the words "long commute to assigned voting spot", "gotta vote", and "buh bye" to fuck right off to the bar of your choice after you mail your ballot (many days before, since you often have to have the ballot in the recorder's hands by election day).
(3) Africa is really screwed, because even the increased volume of melting ice isn't making up for other losses in river volume. Soon Africa will become a wasteland like Ireland, where I found it impossible to get ice.
Do we write mi/h? No. It's 50Mph or 100kph, or whathaveyou.
So in the real world, it's ?bps, with whatever prefix. 115.2kbps, 10Mbps, 10Gbps. Deal.
A couple of people from Apple are working on adding this to the FSF distro: apparently they have it in the version they use, and it was up to 6x faster.
Dunno about a prospective release date. See the gcc mailing list for details.
Oh no, your toast virus has a bug that causes it to be totally ineffectual. I will fix it.
// Note: fixed!
Toast Virus 0.99.1
if(Toast_Present == 1)
{
Turn_On_Coils();
While(Toast.OnFire() == 0)
{
if(Lever.Manual_Eject() == 1)Lever.Jam();
if(Power.Unplug() == 1)Power.Source = reserve_battery;
}
Eject_Flaming_Toast_At_User();
}
You acknowledge and agree that Microsoft may automatically check the version of the OS Product and/or its components that you are utilizing and may provide upgrades or fixes to the OS Product that will be automatically downloaded to your computer
Your honor, Of course it's automated -- I never expected a Microsoft employee to individually select and present to my attention the files I might want to upgrade my Windows 2000 installation. In my mind, and in the minds of millions of citizen-consumers of the wordld, however, 'automatically' does not indicate 'without my permission', it merely indicates that there is some piece of software at the Microsoft end making selections without the oversight of a human being.
That's my argument and I'm sticking it to it.
And consider this: I work in defense, and we have some old, old, old hardware, because what happens is that, once a product is done and all the testing and simulation gear is ready, nobody wants to change anything. And since missiles and stuff like that are out in the field for ages, and the tools basically run themselves over time, people forget how to get things working again if stuff breaks -- or there's simply no *way* to get things working again if stuff breaks.
We have transputers and ancient VAXen and analog computers and crazy shit like that from the 70s and 80s that were used in the simulation of a certain Navy Program. The one guy who understood it all retired, so now there's about six guys in a lab somewhere who have to recreate an entirely new lab from just whatever specifications remain. Sure, people knew this guy was going to retire eventually, but since a lot of these efforts have to be funded by the military, who get yanked around by Congress every couple of weeks, nothing serious was ever done. Heck, I worked on emulating a DR-11W interface (from a PDP-11) for awhile on that program, and we even have a few computers that use MECL logic (!).
Anyway, if there's ever some kind of software fault discovered in this program, *no* *one* can test it to see what might be happening.
Start with the one about median incomes:
a zi ne/09median.slideshow_1.html
http://www.nytimes.com/slideshow/2002/06/06/mag
That one.
Now look at the amount in savings these people have on their $54,400 a year.
1st Family: he's 28, she's 29. Savings: $22.5k.
I guess that's not *too* bad, but they bought a $16k boat even though they've only saved 1/2 a year of
their combined incomes...
2nd Family: he's 35, she's 33. Savings: $1000 in a savings account (i.e., NOTHING). They bought a huge TV, but their kids are like 15, 12, and 8 or something. Sorry, no college for you.
3rd Family: he's 52, she's 45. Look at the hours that poor guy works: upto a 100 a week! For $54k a year! Only $20k in savings, but over $400k in debt.
4th Family: he's 39, she's 28 (good job, buddy). Savings: over $100k (plus college fund), and they own their house and are working on a second one. Best so far.
5th Family: he's 59, she's 53. No mention of savings, but they sold a diner they had, and he mentions that their 401k did pretty well.
6th Family: he's 43, she's 43. Savings: ~$40k + $7k for a college fund. He's a programmer (only technical person so far).
7th Family: She's 35. No $$ figure, but she puts $600 a *month* into a savings account.
So families 4 and 7 seem to be doing pretty well off toward retirement, and probably family 5 as well. Family 3 works hard, but farming is even tougher than I thought.
WHAT ABOUT THE REST OF YOU! WHERE ARE YOUR SAVINGS?!
That's what I can't believe about this country. No one ever saves any goddamn money. They either expect a miracle to happen, or that someone else will cover for them. Like me. And you know what: *fuck* that.
Family number 6: you're technical, you should *get* this. You make $54.4k in Fargo, N.D., which is probably a *lot* of money for that area. Yet you have less than one year of income's worth of savings, and apparently are still paying off your house. WHY?! I've only been working for two years and soon I'll have that much, and I've had to pay off large amounts of student and other debt. Why why why why why?
I can understand that some people just can't save money: that eating and having a place to sleep just eat up their entire income. But then there's people like Family 2: the price of their TV is twice their TOTAL savings. And when they reach retirement age, what do they expect will happen? Keep working until they die? Get Social Security from a generation probably smaller than theirs? What?
Take some responsiblity for you financial future, you goddamn losers. I am not your savior, and the rest of the American public shouldn't have to be, either.
No.
You can't just go grocery shopping for parts, then go home and pile them all on the floor and say, 'yup, that there's a cruise missile'.
As someone who has worked on these things, it isn't easy. Primitive rockets are reasonably easy (see Hamas, fer instance), but any kind of guidance is relatively hard. Also, cruise missiles must either be:
a) airdropped (Tomahawk)
b) launched with a rocket booster (Tomahawk)
c) be able to fly itself up from the ground on a runway (unlikely).
The development cost of even a rude cruise missile is pretty significant, as is the requisite experience. Far cheaper to just buy or steal missiles from bit players like North Korea or whomever. Even the smallest nations could presumably build a missile. Getting one with a reasonable degree of accuracy, however...