And John Gustafson recently stated "The only people who said Amdahl's Law wasn't true were the PR people at Sandia, and a couple of managers who should have known better. Never the people who did the work." Check the video discussion, with Amdahl, Gustafson, and a couple of other folks.
If you see a white horse, it does not imply that all horses are white. If see an algorithm that scales, it does not imply that all algorithms will scale.
Some applications will map to 1000 cores nicely, but not all of them. Thinking Machines, Inmos, MasPar, BBN, and dozens of other companies have found this out the hard way. Intel and AMD also seem intent on finding this out the hard way. Of course, the smart thing to do is to ignore five decades of well documented history, and pretend that this is the first time anyone has tried parallel computing; it's much more fun that way.
Gotta put in a plug for Make magazine, which is a fun read, and full of good projects that anyone can do.
I teach an undergrad course in computer organization (basically beginner architecture), and I've gotten lots of ideas from Lady Ada and Evil Mad Scientist. We use AVR microcontrollers, and the cheap-o USB programmers from Lady Ada, to do a bunch of fun and easy projects.
My kids are 8 and 5, and are playing around a lot with LEDs and magnets. I probably won't let them solder until they're teen-agers (lead in solder sucks, but solder without lead also sucks), but they are getting to breadboard some stuff.
And of course, mentos and coke is always a good idea.
> Parallel processors on a single die (chip) is very different from Thinking Machines & beowulf clusters.
IMO, single or multiple die configuration is not a big distinction; the impact is a constant factor on communications latency (and throughput). One of the other posters mentioned the Transputer; while not on a single chip, they went to a lot of effort to minimize inter-chip communications times. It's easy to get caught up in implementation details; sort of like arguments over what programming language is best. It's the underlying algorithms and complexity that matter.
Anyway, this is all nit-picking. Whether or not Tilera survives depends a lot on what they're going to try to do with it. If the application is graphics (or, as you mention, video processing), the problem allows a lot of parallelism; multi-core makes a lot of sense here (nVidia and ATI have been doing this for years). Cisco-style router switches are good applications too. I heard Agarwal give a presentation at a conference earlier this year, where he was making his pitch; I thought there was way too much hand-waving about the application. The investors will find out the details soon enough, and in the mean time, I've got another stock to short!
Get investors to buy into the hair-brained scheme.
??? (Mention that programming is a problem to be solved shortly.)
Skip town with the cash (Profit!).
Hmmm. I think I'm missing something about a beowulf cluster, or maybe underpants.
It's scary how little history people know. Programming for multi-processor machines was part of the ACM recommended university curriculum back in 1968. Dozens of companies were going to revolutionize the world with parallel (anyone remember the Atari ATW? http://www.atarimuseum.com/computers/16bits/transp uter.html). If parallel worked, it would be really great; I'd like a big rock candy mountain and free energy, while we're at it. Amdahl's law http://en.wikipedia.org/wiki/Gene_Amdahl is from 1967 (this is the 40th anniversary, people!). Madness, sheer madness.
My wife is an architect; she went eco-crazy when we did our house. Straw bale, earth-based plaster, geo-thermal.... There's a lot that can be done with renewable (non-toxic) building materials to make a house energy efficient. We're in upstate NY; we get a big temperature swing, but can stay warm/cool without huge utility bills. I was chopping wood this morning (wood heat warms you twice...).
I hope things go well with Woz and his house. I'm sure he'll do something pretty cool (Woz! If you need an eco-crazy architect who can deal with geeks, send me email:-).
I'm running VMware Fusion (beta) on a low-end MacBook. The final release of Fusion just came out, and I'll probably upgrade tonight. Debian Etch and Windoze XP both run without trouble; no issues getting the networking to happen, and I can drag-and-drop files from the Mac to a VM desktop and back. I had 1gig of RAM; it was a bit slow (especially XP); this morning I upgraded to 2gig, and now it's very fast and smooth.
In general, I'm very happy with the setup. OSX has lots of Unix goodness, and the productivity stuff is nice (Pages and Keynote are slick, and I'm really glad they put out Numbers). I'm a huge fan of GarageBand; beats the daylights out of Cakewalk. My research group machines are all Debian, so having that as a VM lets me keep connected to what the students are up to; having XP supported is handy too. The Macbook is my main machine now; I use it with an external monitor and keyboard at home/office, and only use the laptop screen when I'm on the road.
I've heard good things about Parallels as well; one of my friends is running that, and has no complaints. The only thing I'd caution you on is to get a bunch of RAM; with only one gig, my laptop was swapping constantly. For the Macbook, it was really easy to drop in RAM and a big disk (good deals on both from Newegg).
I'm a CS prof (I teach both grad and undergrad, with my real job being research).
So... be brutally honest with yourself. Do you *really* understand the material, and just couldn't get it together for the exams? Or do you just think you understand the material? The number of people who are clueless to the point of being unaware of their cluelessness is staggering. Grades are an imperfect measure of what someone knows, but that doesn't mean that they're wrong.
If you know your stuff, then grades don't matter. If you don't know your stuff, high grades won't help you. If you've got a year left, and are confident that you actually are on top of things, then knock your last year out with straight As and by being the top student in every class. Recommendations from your professors will carry more weight than a GPA. And I'll agree with the parent post; a grad degree will get your foot in the door in many places, and gives you a clean GPA slate and the opportunity to gather a bunch of useful skills.
Trust in the Peter Principle. Your skills will determine how far you go.
So if there are only 2 possible bond configurations, and 1000 bonds, there are 2^1000 different configurations to test? Are they really trying to brute force it? This sounds like an NP-hard approach, and with massive parallelism, they might cut the run time down from 56 billion years to only 940 million.
And with folding, there have to be some structural constraints -- not every bond configuration may be reachable. I would think that the right way to go is to simulate the folding activity. Plenty of physics computations, but not that easy to send out in a massive distributed parallel run. You mentioned that the algorithm "walks" towards a solution -- how this walking happens is something I'd really like to know about.
Ok, I made the run time numbers up -- but is what they're doing really putting the "cure for cancer" on a human time scale, or is it a publicity stunt to get funding? I'll note that I'd love for them to find a cure, having lost several friends and a few family members myself. I'm just a bit skeptical of them making much progress with a brute force approach. Working on an important problem is good; working on an important problem in a way that can produce a solution is better.
[Side note -- I do optimization of large computational problems for a living, mostly in integrated circuit design. Folding seems like a really interesting problem, and I've been looking around to find a "problem formulation" that makes sense, without much luck.]
So a basic question (that I didn't really get from the F@H page). What exactly are the clients computing, and what gets sent back to the mother ship? There's lots of info on how I can install it and feel good about myself, but that's not what I want to know.
For things like crypto, clients are looking for the magic number in different ranges. Are the F@H clients looking at different protien chains, or the same single chain with different folding possibilities? Or something in between?
So I stuffed the intro section of a paper I recently published, and got 33%, Inauthentic. Other chunks (from published, peer reviewed, and heavily cited conference and journal papers) also seem to get low marks. The best paper at ISPD06 on robust extraction of spatial correlation gets 32.3 Inauthentic for the intro section. They've tuned their software to a specific set of papers, but it doesn't seem to be representative of the full spectrum.
Are they expecting full papers (4-8 pages of text)? Am I not feeding it enough?
I guess now I know why I like watching Blade Runner.
After browsing through, I spotted some names that were familiar. Yet another fine literary work with a New Mexico Tech grad co-author. After Dan McKay's award earlier this year, it's pretty clear the school is on a roll.
Put Dan's work into this format... now that would be cool.
Give them a break... the model is a PR stunt, but the whole nanotube/buckyball thing started at Rice. The feds have started pouring money into nanotech research; if Rice wants to get their fair share of the loot, they need to make sure no one forgets where the nanotube came from. Seems like a lot of/.ers don't know, which is kind of scary.
Most schools use their sports programs to get positive PR. Rice is doing their PR off of some very solid and useful research that happened on campus. Got a problem with that?
Should clarify the grim thing--it's going to get grim for IBM/Intel/AMD/... It'll be hard to sell a CPU with 4 cores if people never fully utilize the first two. And when these guys can't sell CPUs, lots of things will get interesting. No more push on Moore's law, for one.
Games is really the only consumer app where multi-core works -- and even here, it's tricky. You don't want a general purpose CPU--you want something tuned for 3D graphics. IBM has some good news; they make all the video game console chips at the moment, and the CELL is very cool. If we want to continue Moore's law, we need more games. Help us, John Carmack. You're our only hope.
Increased performance in CPUs has normally come from faster clock rates and more complex circuitry. As we all know, Intel (and the others) have bailed out on faster clocks. If you add more complex circuitry, the logic delay increases--to keep the clock rate up, you have to burn power.
What does this mean? The old-fashioned ways of getting more performance are dead--if you try it, the chip will burn up. It's easier to build two 1X MIP cores than one 2X MIP core. Like it or not, dual cores are the only solution; with transistor scaling, we'll have to go to 4, 8, and 16 cores in the next few years. IBM went dual-core with the PowerPC in 2001. Intel, AMD, and Sun are just following suit.
Not bummed out yet? Massive parallelism works well for people doing scientific computing, but for the average joe, it's useless. I don't care how fast a processor is--I usually have one task that will crush it--but rarely do I have two time-critical things to worry about at the same time. In the article referenced, they had to work hard to find things that would test the dual-core features. Parallel computing and multiple cores sounds great. History buffs will know about Thinking Machines, Meiko, Kendell Square, MasPar, NCUBE, Sequent, Transputer, Parsytec, Cray, and so on.... Not a happy ending.
So.... we can't get more single processor performance without bursting into flames. And parallel machines are only useful to a small market. IMO, it's gonna get grim. (And before anyone says new paradigm of computing to take advantage of the parallel resource, put down the crack pipe and think about it--we've been waiting for that paradigm for about 40 years. Remember occam? I thought not.)
The medical research community seems to be behind the curve on this stuff, but ACM http://portal.acm.org/ and IEEE http://ieee.org/ have been fairly progressive. Individual subscriptions to the ACM portal are reasonable, and the site-wide subscription for universities isn't that bad either. I'm involved with ACM SIGDA http://www.sigda.org/, and if you join (free), you get on-line access to all the SIGDA sponsored conference and journal articles. SIGDA will in fact mail you a DVD every year with the past 10 years worth of conference and journal publications -- something for nothing!
While I can't speak for ACM, the policy seems to be "you're smart enough to figure out a way to get it free, so we might as well make it cheap." For SIGDA, the board position is that if you're interested in the material, you should be able to get it with minimum hassle. The ACM portal is in fact working for ACM; they're getting lots of subscriptions, and have cut down on the admistrative grief.
Now if only the entertainment industry would get it....
ACM and IEEE journals can have page charges, but they're usually optional. If you've got a million bucks of government funding to do some research, it doesn't seem out of line for you to help subsidize the publication of research results. For conferences, much of the registration fee goes for rental of the meeting space, food, and so on; more than most people would expect. Conferences require authors to register so that the other attendees don't show up to an empty room. Conferences usually do a little better than break even, which helps cover the freebie DVD, and things like that.
We will be doing geothermal. It looks like the best option will be 5-ft deep trenches for the ground loop; we don't have a large body of water nearby, and drilling down is apparently more expensive than the trench. We're in upstate NY, and I've been getting killed by heating in our current house. This is most of the reason we're going with straw bale--great insulation for the price.
On a more serious note, I'll probably stick up a web cam during construction; probably interesting to some of the geek crowd, and then I can repurpose it to monitor my christmas lights over the net. We could even use the web cam to direct a nail gun during construction, and then repurpose it to deal with any wolves that might come by!
Thanks -- I was wondering about the wireless leakage. Any idea on what happens w/o the wire lathe? We're planning on using earth/lime, and applying it directly to the straw.
(And on an unrelated note--you wouldn't have been at a LEED workshop in PA a year or so ago? My wife thinks your linked site looks familiar).
There's a good article on the '02 Bell Labs scandal at http://www.aip.org/tip/INPHFA/vol-8/iss-6/p12.html. One of the researchers working there made up experimental data for at least 16 published (and peer reviewed) papers. He got caught because he didn't make up enough stuff, and used the same fake data to support two completely different experiments. The Schon incident is at the extreme end of the spectrum, but it's not an isolated event. IMO, around 30% of the journal papers in my field are intentionally misleading (or worse) with their experiments.
There's a lot of "publish or perish" pressure on academics, as well as the need to generate funding. My policy--if the experimental data is all favorable, it's faked. At least we can trust the news media and politicians to tell us the truth.
Slashdot Mirror
And John Gustafson recently stated "The only people who said Amdahl's Law wasn't true were the PR people at Sandia, and a couple of managers who should have known better. Never the people who did the work." Check the video discussion, with Amdahl, Gustafson, and a couple of other folks.
If you see a white horse, it does not imply that all horses are white. If see an algorithm that scales, it does not imply that all algorithms will scale.
Some applications will map to 1000 cores nicely, but not all of them. Thinking Machines, Inmos, MasPar, BBN, and dozens of other companies have found this out the hard way. Intel and AMD also seem intent on finding this out the hard way. Of course, the smart thing to do is to ignore five decades of well documented history, and pretend that this is the first time anyone has tried parallel computing; it's much more fun that way.
Gotta put in a plug for Make magazine, which is a fun read, and full of good projects that anyone can do.
I teach an undergrad course in computer organization (basically beginner architecture), and I've gotten lots of ideas from Lady Ada and Evil Mad Scientist. We use AVR microcontrollers, and the cheap-o USB programmers from Lady Ada, to do a bunch of fun and easy projects.
My kids are 8 and 5, and are playing around a lot with LEDs and magnets. I probably won't let them solder until they're teen-agers (lead in solder sucks, but solder without lead also sucks), but they are getting to breadboard some stuff.
And of course, mentos and coke is always a good idea.
IMO, single or multiple die configuration is not a big distinction; the impact is a constant factor on communications latency (and throughput). One of the other posters mentioned the Transputer; while not on a single chip, they went to a lot of effort to minimize inter-chip communications times. It's easy to get caught up in implementation details; sort of like arguments over what programming language is best. It's the underlying algorithms and complexity that matter.
Getting meaningful parallelism is difficult, even if you could get zero communications delay. It's amusing to me that Anant Agarwal is pushing this idea, when he had a CACM (Vol 34, No 3, March 1991, pp. 57-61). paper that mentions "for any algorithm in which the overhead due to parallelism increases with p, the scaled speedup is necessarily less than linear, and that when the parallelism overhead is linear or worse with p, there is actually a hard upper bound on the achievable scaled speedup." http://portal.acm.org/citation.cfm?id=102868.10287 1&coll=portal&dl=ACM&idx=102868&part=periodical&Wa ntType=periodical&title=Communications%20of%20the% 20ACM&CFID=32421333&CFTOKEN=50501291. These observations were based on some work by Richard Karp http://portal.acm.org/citation.cfm?id=894803&dl=AC M&coll=portal&CFID=32421333&CFTOKEN=50501291, who I have a lot of confidence in. What this means is that for most algorithms, Amdahl is right. If your communications time is non-zero, you're going to hit the wall. Better interconnect pushes the wall back, but not forever.
Anyway, this is all nit-picking. Whether or not Tilera survives depends a lot on what they're going to try to do with it. If the application is graphics (or, as you mention, video processing), the problem allows a lot of parallelism; multi-core makes a lot of sense here (nVidia and ATI have been doing this for years). Cisco-style router switches are good applications too. I heard Agarwal give a presentation at a conference earlier this year, where he was making his pitch; I thought there was way too much hand-waving about the application. The investors will find out the details soon enough, and in the mean time, I've got another stock to short!
It's good to see that MIT has perfected the technology.
- Build a machine with lots of processors.
- Get investors to buy into the hair-brained scheme.
- ??? (Mention that programming is a problem to be solved shortly.)
- Skip town with the cash (Profit!).
Hmmm. I think I'm missing something about a beowulf cluster, or maybe underpants.It's scary how little history people know. Programming for multi-processor machines was part of the ACM recommended university curriculum back in 1968. Dozens of companies were going to revolutionize the world with parallel (anyone remember the Atari ATW? http://www.atarimuseum.com/computers/16bits/trans
My wife is an architect; she went eco-crazy when we did our house. Straw bale, earth-based plaster, geo-thermal.... There's a lot that can be done with renewable (non-toxic) building materials to make a house energy efficient. We're in upstate NY; we get a big temperature swing, but can stay warm/cool without huge utility bills. I was chopping wood this morning (wood heat warms you twice...).
:-).
I hope things go well with Woz and his house. I'm sure he'll do something pretty cool (Woz! If you need an eco-crazy architect who can deal with geeks, send me email
I'm running VMware Fusion (beta) on a low-end MacBook. The final release of Fusion just came out, and I'll probably upgrade tonight. Debian Etch and Windoze XP both run without trouble; no issues getting the networking to happen, and I can drag-and-drop files from the Mac to a VM desktop and back. I had 1gig of RAM; it was a bit slow (especially XP); this morning I upgraded to 2gig, and now it's very fast and smooth.
In general, I'm very happy with the setup. OSX has lots of Unix goodness, and the productivity stuff is nice (Pages and Keynote are slick, and I'm really glad they put out Numbers). I'm a huge fan of GarageBand; beats the daylights out of Cakewalk. My research group machines are all Debian, so having that as a VM lets me keep connected to what the students are up to; having XP supported is handy too. The Macbook is my main machine now; I use it with an external monitor and keyboard at home/office, and only use the laptop screen when I'm on the road.
I've heard good things about Parallels as well; one of my friends is running that, and has no complaints. The only thing I'd caution you on is to get a bunch of RAM; with only one gig, my laptop was swapping constantly. For the Macbook, it was really easy to drop in RAM and a big disk (good deals on both from Newegg).
So... be brutally honest with yourself. Do you *really* understand the material, and just couldn't get it together for the exams? Or do you just think you understand the material? The number of people who are clueless to the point of being unaware of their cluelessness is staggering. Grades are an imperfect measure of what someone knows, but that doesn't mean that they're wrong.
If you know your stuff, then grades don't matter. If you don't know your stuff, high grades won't help you. If you've got a year left, and are confident that you actually are on top of things, then knock your last year out with straight As and by being the top student in every class. Recommendations from your professors will carry more weight than a GPA. And I'll agree with the parent post; a grad degree will get your foot in the door in many places, and gives you a clean GPA slate and the opportunity to gather a bunch of useful skills.
Trust in the Peter Principle. Your skills will determine how far you go.
So if there are only 2 possible bond configurations, and 1000 bonds, there are 2^1000 different configurations to test? Are they really trying to brute force it? This sounds like an NP-hard approach, and with massive parallelism, they might cut the run time down from 56 billion years to only 940 million.
And with folding, there have to be some structural constraints -- not every bond configuration may be reachable. I would think that the right way to go is to simulate the folding activity. Plenty of physics computations, but not that easy to send out in a massive distributed parallel run. You mentioned that the algorithm "walks" towards a solution -- how this walking happens is something I'd really like to know about.
Ok, I made the run time numbers up -- but is what they're doing really putting the "cure for cancer" on a human time scale, or is it a publicity stunt to get funding? I'll note that I'd love for them to find a cure, having lost several friends and a few family members myself. I'm just a bit skeptical of them making much progress with a brute force approach. Working on an important problem is good; working on an important problem in a way that can produce a solution is better.
[Side note -- I do optimization of large computational problems for a living, mostly in integrated circuit design. Folding seems like a really interesting problem, and I've been looking around to find a "problem formulation" that makes sense, without much luck.]
So a basic question (that I didn't really get from the F@H page). What exactly are the clients computing, and what gets sent back to the mother ship? There's lots of info on how I can install it and feel good about myself, but that's not what I want to know.
For things like crypto, clients are looking for the magic number in different ranges. Are the F@H clients looking at different protien chains, or the same single chain with different folding possibilities? Or something in between?
So I stuffed the intro section of a paper I recently published, and got 33%, Inauthentic. Other chunks (from published, peer reviewed, and heavily cited conference and journal papers) also seem to get low marks. The best paper at ISPD06 on robust extraction of spatial correlation gets 32.3 Inauthentic for the intro section. They've tuned their software to a specific set of papers, but it doesn't seem to be representative of the full spectrum. Are they expecting full papers (4-8 pages of text)? Am I not feeding it enough? I guess now I know why I like watching Blade Runner.
THX http://www.thx.com/ was founded in 1983 by George Lucas. I'd assume that it's a reference to his film THX 1138 http://www.thx1138movie.com/ http://www.imdb.com/title/tt0066434/.
THX 1138 is a documentary on the early 21st century.
After browsing through, I spotted some names that were familiar. Yet another fine literary work with a New Mexico Tech grad co-author. After Dan McKay's award earlier this year, it's pretty clear the school is on a roll.
Put Dan's work into this format... now that would be cool.
Most schools use their sports programs to get positive PR. Rice is doing their PR off of some very solid and useful research that happened on campus. Got a problem with that?
Should clarify the grim thing--it's going to get grim for IBM/Intel/AMD/... It'll be hard to sell a CPU with 4 cores if people never fully utilize the first two. And when these guys can't sell CPUs, lots of things will get interesting. No more push on Moore's law, for one.
Games is really the only consumer app where multi-core works -- and even here, it's tricky. You don't want a general purpose CPU--you want something tuned for 3D graphics. IBM has some good news; they make all the video game console chips at the moment, and the CELL is very cool. If we want to continue Moore's law, we need more games. Help us, John Carmack. You're our only hope.
I'll probably get flamed for this....
Increased performance in CPUs has normally come from faster clock rates and more complex circuitry. As we all know, Intel (and the others) have bailed out on faster clocks. If you add more complex circuitry, the logic delay increases--to keep the clock rate up, you have to burn power.
What does this mean? The old-fashioned ways of getting more performance are dead--if you try it, the chip will burn up. It's easier to build two 1X MIP cores than one 2X MIP core. Like it or not, dual cores are the only solution; with transistor scaling, we'll have to go to 4, 8, and 16 cores in the next few years. IBM went dual-core with the PowerPC in 2001. Intel, AMD, and Sun are just following suit.
Not bummed out yet? Massive parallelism works well for people doing scientific computing, but for the average joe, it's useless. I don't care how fast a processor is--I usually have one task that will crush it--but rarely do I have two time-critical things to worry about at the same time. In the article referenced, they had to work hard to find things that would test the dual-core features. Parallel computing and multiple cores sounds great. History buffs will know about Thinking Machines, Meiko, Kendell Square, MasPar, NCUBE, Sequent, Transputer, Parsytec, Cray, and so on.... Not a happy ending.
So.... we can't get more single processor performance without bursting into flames. And parallel machines are only useful to a small market. IMO, it's gonna get grim. (And before anyone says new paradigm of computing to take advantage of the parallel resource, put down the crack pipe and think about it--we've been waiting for that paradigm for about 40 years. Remember occam? I thought not.)
While I can't speak for ACM, the policy seems to be "you're smart enough to figure out a way to get it free, so we might as well make it cheap." For SIGDA, the board position is that if you're interested in the material, you should be able to get it with minimum hassle. The ACM portal is in fact working for ACM; they're getting lots of subscriptions, and have cut down on the admistrative grief.
Now if only the entertainment industry would get it....
ACM and IEEE journals can have page charges, but they're usually optional. If you've got a million bucks of government funding to do some research, it doesn't seem out of line for you to help subsidize the publication of research results. For conferences, much of the registration fee goes for rental of the meeting space, food, and so on; more than most people would expect. Conferences require authors to register so that the other attendees don't show up to an empty room. Conferences usually do a little better than break even, which helps cover the freebie DVD, and things like that.
We will be doing geothermal. It looks like the best option will be 5-ft deep trenches for the ground loop; we don't have a large body of water nearby, and drilling down is apparently more expensive than the trench. We're in upstate NY, and I've been getting killed by heating in our current house. This is most of the reason we're going with straw bale--great insulation for the price.
On a more serious note, I'll probably stick up a web cam during construction; probably interesting to some of the geek crowd, and then I can repurpose it to monitor my christmas lights over the net. We could even use the web cam to direct a nail gun during construction, and then repurpose it to deal with any wolves that might come by!
Thanks -- I was wondering about the wireless leakage. Any idea on what happens w/o the wire lathe? We're planning on using earth/lime, and applying it directly to the straw.
(And on an unrelated note--you wouldn't have been at a LEED workshop in PA a year or so ago? My wife thinks your linked site looks familiar).
There's a good article on the '02 Bell Labs scandal at http://www.aip.org/tip/INPHFA/vol-8/iss-6/p12.html . One of the researchers working there made up experimental data for at least 16 published (and peer reviewed) papers. He got caught because he didn't make up enough stuff, and used the same fake data to support two completely different experiments. The Schon incident is at the extreme end of the spectrum, but it's not an isolated event. IMO, around 30% of the journal papers in my field are intentionally misleading (or worse) with their experiments.
There's a lot of "publish or perish" pressure on academics, as well as the need to generate funding. My policy--if the experimental data is all favorable, it's faked. At least we can trust the news media and politicians to tell us the truth.