They have those. One brand is called LabJack. The downside is that you're paying extra for what amounts to an arduino that you can't (as?) easily re-program and/or embed into a gadget.
I would not be so quick to accuse people of writing poor code when you know very little about the problem they're working on. And remember, ~most code runs faster on CPUs. If you read some of Vasily Volkov's papers (he's the guy who wrote the early versions of CUBLAS), it is very clear that you might as well not bother with the GPU if you're mostly doing blas level 1-2 stuff, since the arithmetic intensity isn't high enough. For our application we had some specific operations we could combine and tricks we could use to reduce bandwidth, but they turned out to not be enough. This was a Ph.D. project that took several years; many parallelization strategies were tried on both CPU and GPU, and I got lots of feedback from colleagues working on gpu compiler tools.
Well, I suppose either my application (face recognition for hundreds of users) is under the threshold for your definition of HPC, or it's a notable exception. Our algorithm consists primarily of repeated BLAS level 1 and 2 operations on chunks of data that fit in CPU cache, but not GPU cache. Essentially, it's low arithmetic intensity operations performed repeatedly (hundreds of times) on gallery image sets that take up a couple megs at a time (and there are a couple hundred of those that can be computed in parallel). Under these conditions, we find that a dual socket quad-core Xeon machine is roughly comparable to a high-end Fermi. There is locality in our memory access pattern that the CPU has enough cache to exploit, but the GPU does not. I'm not a chip designer, so I can't ~really compare the opportunity cost (in $/flop) of adding more cache vs. widening a memory bus of a GPU, but I suspect the cache is cheaper, especially given that it need not be shared (again, for our application).
I'm curious about the memory model too. I'm pretty certain that bit about "cluster on a chip" is just marketing hyperbole, and it's actually still a shared memory system running one instance of the linux kernel. They're not going to make you run 50 linux kernel instances and communicate between them using network sockets.
Depends on how much cache is on the chip, and how big the problem being solved is. GPU's have a lot of FP units, but they have such a tiny amount of cache that they basically have to transfer ~everything they operate on over the memory bus. On a CPU, your dataset can be several MB and still fit on-chip, but of course you have fewer FP units. The algorithm I designed for my Ph.D. operate on the same few megabytes of data many times, and it ended up being about equally fast on both architectures, so I'm hoping KC will bridge the cache size chasm that exists between CPU's and GPU's.
Not that it has any bearing on whether children deserve to be beaten, but she ~does have ataxic cerebral palsy. It's in the second article you link to.
What's your take on elder abuse? Should we take off our belts and beat the crap out of our grandparents if they buy unauthorized copies of prescription drugs on the black market? Is a judge who beats his elderly parents for such behavior fit to decide cases involving elder abuse? Does the level of senility of a little old man really have any bearing on whether his children should ever beat him with a belt?
If your post is genuine, then this is a massive CPS WIN! They managed to teach you, even as an adult, that you can't get away with beating children. You are a excellent case study in how CPS helps to prevent abuse. You are a less violent person ~despite the abuse visited on you as a child, and ~thanks to the nanny state.
Her mother was also abused, and later divorced her husband, and has apologized to her daughter, who has forgiven her. This man would likely beat the crap out of his daughter again given the opportunity, while simultaneously claiming to be a paragon of morality. Don't blame victims for trying to expose their own abuse.
Letting customers prioritize the relative importance of their own data transfers sounds like a feature to me. This might help shift the bandwidth of the network from content users don't care much about to content they care more about.
Of course, I'm sure there are lots of ways to game the system; perhaps a devious app could wait in the background until the user clicks the "Turbo" button, and ~then download updated advertising content. And while it doesn't sound like this violates net neutrality, it might help create even more on an incentive to do so in the future. Advertisers are very interested in which content people are willing to pay extra for, both in aggregate and on a per-user basis. It's a well known fact that users pay more attention to ads inserted into content that they paid for. Magazines give away many free issues (while maintaining a high sticker price) to try to increase readership without hurting the perceived value of their publication. I'm sure web editions will be eager to advertise "Pre-boosted for free for a limited time only!"
I agree that it made the point, but I still feel it hurts his cause. Right now the biggest PR problem for atheists is that they come off as impolite jerks far too often, largely thanks to the style (not the substance) of public figures such as Dawkins. To quote the great Lebowski, "You're not wrong, Walter, you're just an ***hole."
Having watched the video and read the letters, here is my summary of how it got ugly:
In refuting the notion that the existence of religious scientists itself is evidence that science and religion are compatible, Hoyne uses the existence of Catholic pedophiles as an example of humans being able to hold incompatible beliefs in their heads. This was a strawman (Hoyne didn't use this argument), and a rather low blow; there are any number of non-humiliating examples that could have been used to make the same point.
Haught, was quite understandably offended by this, and didn't want the video published. If, say, the National Academy of Science had been caught systematically raping children and covering it up, I suspect Coyne would be just as offended if the scandal were needlessly mentioned by his opponent in a public, and supposedly good-natured academic debate.
Why does this article pre-suppose that multi-national corporations have any loyalty whatsoever to any country? Perhaps Dell screwed up by ceding it's manufacturing capability to it's main contractor (Asus) and then losing control, but that has nothing to do with America vs. China. Patriotism and Capitalism are inherently at odds.
I am a CUDA C++ programmer. My biggest complaint about programming tools for the GPU is that there are no dense linear algebra libraries that work at the SM level. For my application I had to re-implement a big chunk of BLAS and part of LAPACK from scratch so that each SM runs a different problem instance. On the CPU you can just use openmp + single threaded BLAS to achieve the same granularity of parallelism. Thrust API does not address this granularity of parallelism. I'm eager to see if the AMP API does.
There are people who memorize key commands for accessing menu items so thoroughly that they use the keyboard almost exclusively even in gui programs. Also, there ~are programs with a build-in application specific shell; cad programs have a long tradition of this. Gvim is also a good example, illustrating both a built-in shell, as well as menus that make a wide variety of commands easily discoverable for noobs.
Wow, the fragility of an encrypted file system plus the instability of a GPU, implemented in the kernel. Do not even read TFA without doing a full backup of your system.
Graphics processors' architecture resembles that of a CPU with hundreds (thousands?) of parallel cores.
Cpu is more like 6 cores, with 4 (or soon 8) single precision width vector units. For cuda, it's more like 16 cores, with 32 (or 64 depending on how you're counting) width single precision vector units. Nvidia marketing uses a funny definition of "core"; an cuda SMP is roughly analogous to a cpu core running at a much slower clock, but with a much wider vector unit.
I think a lot of people are downplaying the historic value of this device as the first stereo camera that the vast majority of consumers (and especially kids) will lay their hands on.
It is cheap, It fits in your pocket, It records images to standard media, it doubles as a glasses-free viewer for the images it takes.
The resolution isn't great, there is a better (but almost twice as expensive) digital stereo camera on the market for stereo enthusiasts, and film stereo photography is as old as the sun, but the kids don't care about any of that. For the most part they will view the 3DS as the birth of home stereo photography.
Just about any application involving computer vision will gobble up as many cores as you throw at it. Think face recognition, product recognition, augmented reality, computational photography, and so on. The list of conceived but not yet commercialized vision-related applications is rather long.
Slashdot Mirror
They have those. One brand is called LabJack. The downside is that you're paying extra for what amounts to an arduino that you can't (as?) easily re-program and/or embed into a gadget.
You win!
I would not be so quick to accuse people of writing poor code when you know very little about the problem they're working on. And remember, ~most code runs faster on CPUs. If you read some of Vasily Volkov's papers (he's the guy who wrote the early versions of CUBLAS), it is very clear that you might as well not bother with the GPU if you're mostly doing blas level 1-2 stuff, since the arithmetic intensity isn't high enough. For our application we had some specific operations we could combine and tricks we could use to reduce bandwidth, but they turned out to not be enough. This was a Ph.D. project that took several years; many parallelization strategies were tried on both CPU and GPU, and I got lots of feedback from colleagues working on gpu compiler tools.
Well, I suppose either my application (face recognition for hundreds of users) is under the threshold for your definition of HPC, or it's a notable exception. Our algorithm consists primarily of repeated BLAS level 1 and 2 operations on chunks of data that fit in CPU cache, but not GPU cache. Essentially, it's low arithmetic intensity operations performed repeatedly (hundreds of times) on gallery image sets that take up a couple megs at a time (and there are a couple hundred of those that can be computed in parallel). Under these conditions, we find that a dual socket quad-core Xeon machine is roughly comparable to a high-end Fermi. There is locality in our memory access pattern that the CPU has enough cache to exploit, but the GPU does not. I'm not a chip designer, so I can't ~really compare the opportunity cost (in $/flop) of adding more cache vs. widening a memory bus of a GPU, but I suspect the cache is cheaper, especially given that it need not be shared (again, for our application).
I'm curious about the memory model too. I'm pretty certain that bit about "cluster on a chip" is just marketing hyperbole, and it's actually still a shared memory system running one instance of the linux kernel. They're not going to make you run 50 linux kernel instances and communicate between them using network sockets.
Depends on how much cache is on the chip, and how big the problem being solved is. GPU's have a lot of FP units, but they have such a tiny amount of cache that they basically have to transfer ~everything they operate on over the memory bus. On a CPU, your dataset can be several MB and still fit on-chip, but of course you have fewer FP units. The algorithm I designed for my Ph.D. operate on the same few megabytes of data many times, and it ended up being about equally fast on both architectures, so I'm hoping KC will bridge the cache size chasm that exists between CPU's and GPU's.
I still don't quite get it. Could you please elaborate?
You're advocating for people beat their children more, and yet ~I'm the one "with little empathy for others"?
I'm not an expert on psychology or child development, but the experts sure seem to disagree with you:
http://www.apa.org/about/governance/council/policy/corporal-punishment.aspx
Not that it has any bearing on whether children deserve to be beaten, but she ~does have ataxic cerebral palsy. It's in the second article you link to.
What's your take on elder abuse? Should we take off our belts and beat the crap out of our grandparents if they buy unauthorized copies of prescription drugs on the black market? Is a judge who beats his elderly parents for such behavior fit to decide cases involving elder abuse? Does the level of senility of a little old man really have any bearing on whether his children should ever beat him with a belt?
http://abstrusegoose.com/114
"Beats his own children for violating RIAA sponsored copyright legislation"
Your response to watching video of a disabled girl getting beaten is "Whip her again"?
http://abstrusegoose.com/114
"Condones child abuse"
It doesn't matter if it's a boy or a girl.
If your post is genuine, then this is a massive CPS WIN! They managed to teach you, even as an adult, that you can't get away with beating children. You are a excellent case study in how CPS helps to prevent abuse. You are a less violent person ~despite the abuse visited on you as a child, and ~thanks to the nanny state.
Her mother was also abused, and later divorced her husband, and has apologized to her daughter, who has forgiven her. This man would likely beat the crap out of his daughter again given the opportunity, while simultaneously claiming to be a paragon of morality. Don't blame victims for trying to expose their own abuse.
Letting customers prioritize the relative importance of their own data transfers sounds like a feature to me. This might help shift the bandwidth of the network from content users don't care much about to content they care more about.
Of course, I'm sure there are lots of ways to game the system; perhaps a devious app could wait in the background until the user clicks the "Turbo" button, and ~then download updated advertising content. And while it doesn't sound like this violates net neutrality, it might help create even more on an incentive to do so in the future. Advertisers are very interested in which content people are willing to pay extra for, both in aggregate and on a per-user basis. It's a well known fact that users pay more attention to ads inserted into content that they paid for. Magazines give away many free issues (while maintaining a high sticker price) to try to increase readership without hurting the perceived value of their publication. I'm sure web editions will be eager to advertise "Pre-boosted for free for a limited time only!"
I agree that it made the point, but I still feel it hurts his cause. Right now the biggest PR problem for atheists is that they come off as impolite jerks far too often, largely thanks to the style (not the substance) of public figures such as Dawkins. To quote the great Lebowski, "You're not wrong, Walter, you're just an ***hole."
Having watched the video and read the letters, here is my summary of how it got ugly:
In refuting the notion that the existence of religious scientists itself is evidence that science and religion are compatible, Hoyne uses the existence of Catholic pedophiles as an example of humans being able to hold incompatible beliefs in their heads. This was a strawman (Hoyne didn't use this argument), and a rather low blow; there are any number of non-humiliating examples that could have been used to make the same point.
Haught, was quite understandably offended by this, and didn't want the video published. If, say, the National Academy of Science had been caught systematically raping children and covering it up, I suspect Coyne would be just as offended if the scandal were needlessly mentioned by his opponent in a public, and supposedly good-natured academic debate.
Why does this article pre-suppose that multi-national corporations have any loyalty whatsoever to any country? Perhaps Dell screwed up by ceding it's manufacturing capability to it's main contractor (Asus) and then losing control, but that has nothing to do with America vs. China. Patriotism and Capitalism are inherently at odds.
I am a CUDA C++ programmer. My biggest complaint about programming tools for the GPU is that there are no dense linear algebra libraries that work at the SM level. For my application I had to re-implement a big chunk of BLAS and part of LAPACK from scratch so that each SM runs a different problem instance. On the CPU you can just use openmp + single threaded BLAS to achieve the same granularity of parallelism. Thrust API does not address this granularity of parallelism. I'm eager to see if the AMP API does.
One thing I miss from my old 80's mobile... the ability to have it turn off at a certain hour and then turn on again automatically in the morning.
Amen. I cannot count the number of times my phone has woken me up in the middle of the night just to tell me it is running out of batteries.
There are people who memorize key commands for accessing menu items so thoroughly that they use the keyboard almost exclusively even in gui programs. Also, there ~are programs with a build-in application specific shell; cad programs have a long tradition of this. Gvim is also a good example, illustrating both a built-in shell, as well as menus that make a wide variety of commands easily discoverable for noobs.
Wow, the fragility of an encrypted file system plus the instability of a GPU, implemented in the kernel. Do not even read TFA without doing a full backup of your system.
Graphics processors' architecture resembles that of a CPU with hundreds (thousands?) of parallel cores.
Cpu is more like 6 cores, with 4 (or soon 8) single precision width vector units. For cuda, it's more like 16 cores, with 32 (or 64 depending on how you're counting) width single precision vector units. Nvidia marketing uses a funny definition of "core"; an cuda SMP is roughly analogous to a cpu core running at a much slower clock, but with a much wider vector unit.
They are targeting 1 Watt mobile applications to start with. For reference, a high-end gpu these days is ballpark 500W.
I think a lot of people are downplaying the historic value of this device as the first stereo camera that the vast majority of consumers (and especially kids) will lay their hands on.
It is cheap,
It fits in your pocket,
It records images to standard media,
it doubles as a glasses-free viewer for the images it takes.
The resolution isn't great, there is a better (but almost twice as expensive) digital stereo camera on the market for stereo enthusiasts, and film stereo photography is as old as the sun, but the kids don't care about any of that. For the most part they will view the 3DS as the birth of home stereo photography.
Just about any application involving computer vision will gobble up as many cores as you throw at it. Think face recognition, product recognition, augmented reality, computational photography, and so on. The list of conceived but not yet commercialized vision-related applications is rather long.