On the other hand, writing documentation for things is one of the best ways I've found to really flesh out my understanding of something, so I bet these students will come out of the project with a significantly improved understanding of the language.
Agreed. Some years ago, me and four classmates wrote a 100-page "compendium" (less ambitious than a textbook, more like supporting material) for our C++ course. This was the summer after we took it; we sold it to subsequent classes for $15 per copy, with the proceedings going to our student society. It improved my understanding of the subject a lot, even though I had aced the exam. The only thing I would have done differently today is write it in English, not our native language, so that we could share it with more people.
What are you going to do with 64 GB of RAM on a single node?
If you're actually using more than 2-4 GB per core, your
program is fully limited by RAM speed,
and you should REALLY be using MPI and more nodes
with less ram each.
You have to keep in mind that "inflation" measures published by the government are kept artificially low by excluding food, energy and other things that have been getting more expensive lately. The technical term is "core inflation", but it's just an excuse for the politicians to goalseek inflation into what they want it to be. Now, gasoline would still have a positive price increase with a realistic inflation measure, but less than your numbers.
Ehm, I have to say that ECC is over-hyped by server hardware vendors, especially for CFD applications. The failure rate for modern RAM is 1 bit error per 1 GB per 1 month of simulation. To be honest, a typical CFD code will have to handle much worse errors than that due to random programming bugs (if you think your 35,000 lines-of-code program is bug free, well... it's not) etc., such that if it crashes or becomes unphysical from 12 bit errors in a month, you're screwed anyway.
On the other hand, if you're running a COBOL program that calculates how much you're employees taxes are going to be, then you really want ECC.
Okay, I'll answer this. From my use of a Core i7, you can tell that I'll be coding a serial app with some OpenMP in the slow parts to utilize all 4 cores. Now that's much easier than writing the same code for a GPU. If I were serious in developing the "blazing fast" stuff (which I'm not, my focus is on implementing new multi-physics models) I could spend the same amount of effort and target MPI instead of GPUs, and then go run it on the 22,000 core cluster in the basement next door.
The good thing about manufacturers speeding up SSE/AVX/etc. is that the linear algebra libraries (specifically the ATLAS implementation of BLAS and LAPACK) usually release code that makes use of the new hawtness in about six months after release. Do you know how much software relies on BLAS and LAPACK for speed?
If you're doing numerics, what the fuck (if you'll pardon my French) are you doing buying Apple? I'm working on two-phase Navier-Stokes solvers myself, and I just bought a new rig consisting of 3 boxes each with a Intel Core i7 @ 3.7 GHz, 12 GB RAM, an SSD drive and a big-ass cooling system. In total that cost less than the Mac Pro with a single Core i7 @ 3.3 GHz listed in that article.You're paying 3x more than you should, and you get what extra? A shiny case? Puh-lease.
Agreed. My previous camera (Canon SX20) had to be repaired twice, and then replaced (with an SX40), since it came from one of the first production lines which used leadless solder. The solder was bad, so the voltage converter would stop delivering the correct voltage after about 10 000 shots.
It's not that a particle has a theoretical probability of being somewhere with some probable momentum, no it will be at a very real place at a very real time with a very actual momentum. It's just that practically it's so complicated to predict it, that the best way we have come up till now are quantum mechanics.
Nature tends to generate more data than can be quickly processed
FTFY. I would even amend that to "be quickly processed, or even stored". Think about a fluid dynamics simulation for instance. Say you want to store a velocity vector at each grid point on a 400x400x1000 grid at 500 time increments. How much harddisk space will you need? A quick calculation comes out at 1.8 terabytes of data! And that's just for the velocity field, you can easily triple that number in a realistic CFD case.
This. Top Gear is the show who
recommended the Skoda Yeti because
Sienna Miller's head fits in the
glovebox, routinely tests cars
against fighter planes, and
compares station wagons on how
many pounds of cheese they can carry.
I don't know what Tesla expected.
It's simple. Java was teh hawtness roughly ten years ago. It took four years for university CS departments to catch on, then four years for the first Java-loving graduates to start "entrepreneuring", and then two years for this to significantly influence Java usage. It's a basic convolution-type feedback.
Good designers make choices. Bad designers can't make up their minds and pass the buck down to users.
Look, it's not as if OSX has five toggles in the system menu, it has 29 separate tabs each with many options. It even has a tab called "Energy Saver", where this option would fit naturally. Stop trying to frame this omission as a conscious "design" choice when it's clearly not.
But if you include them all, you end up with a horrible mess of an options UI, that most people don't understand or get lost in.
Ah, but Apple, who are generally regarded as being a few orders of magnitude better at UI design than a random torrent-client devteam, should not have a problem here. A simple tickbox to enable "advanced settings" under the "Power management" tab in "System settings" (or whatever it is called) solves that problem. See e.g. VLC for a good demonstration - if you don't toggle from Simple to All in the settings window, there is no option for "which packetizer preprocesses the stream before muxing" and a million other such things. How many people do you think prefer to control the packetizer when looking at a movie? A grand total of ten out of six billion?
When it's only sensible for 70% of the use cases, there should be a software switch. How would it be a worse design decision to have the default behaviour as it is today, but a way to toggle it in the system settings? The original point in this thread was that users should be given maximum flexibility, especially when it's a matter of twenty lines of code in the OS. Apple fails to implement this, which is why many of us geeks use something else (cf. smartphones).
To clarify:
Point 1: I've never owned a Macbook, but I have friends who've been annoyed by this inability to close the lid while running.
Point 2: My current computer, a 4-core Dell Precision something-or-other, has compiled lots of code in my bacpack. I use my water bottle to ensure that the fan intakes on the bottom are free, and I open my backpack zipper on the side where the fan outlets are. Works like a charm (I logged the CPU temperature with a small script the first few times I tried it).
TL;DR:
I don't want my computer "protecting" me from doing something that could maybe be stupid. I want it to shut down if overheating, that's a much better solution that all laptops implement.
So you're saying if I buy a Mac, I have to give up on letting my code compile/run while my laptop is in my backpack on the way home? And that if I want to plug it into my stereo system and listen to music, I can't close it and put it in the shelf, but that it must take up space on the table?
Erm... why? No really... Why! The vast majority of accommodations we are talking about are software features. These are adjustments and settings which can be customised.
You're talking about a company which sells a computer OS which is mainly used on laptops, but where there is no toggle for whether closing the screen puts the laptop into sleep mode. It's a trivial feature found on all other OS's, but Apple seemingly can't be bothered to include it, and the people who buy the $999 Facebook machines adapt their usage patterns to their computers, not vice versa.
If the terrorists were really serious about levelling a small city, I think a massive fuel-air bomb would be much easier to execute than an actual nuke, even a small tactical one. Start out with a big tank of liquefied natural gas, use a small explosive to rupture the tank and let the LNG go through a rapid phase transition and subsequent turbulent mixing with air, wait a little before detonating the second small explosive to ignite the vapor cloud. Hang this from a blimp or whatever for added effect.
Also, unlike nukes, fuel-air explosives have a history of being used by terrorists.
Yes, and it's extremely annoying if you want to do anything to your own car. It's bad on the same level as proprietary connectors for phones and all that, but unfortunately the amount of people improving their own cars is too low to cause any consumer feedback to manufacturers.
And I don't mean adding stupid spoilers and boost chips and sillyness, I mean stuff like adding an extra pair of high beams that can be operated with the same button as the regular high beams. That will take some serious hacking on a modern car. If car manufacturers were good at making things, this wouldn't be a huge problem, but modern cars do so many things wrong that it's infuriating. Like putting lambertian leds in places where they should have put batwing ones, forcing me to put a diffuser in front of it so that my daughter is able to sleep in her car seat. Or making it a fifteen-minute job to remove the battery for charging it during the winter, when it should take two minutes. Or putting the light that activates when you open the trunk in the far left corner of the trunk, so that it doesn't light up anything if you actually have something in the trunk. I could go on about this for a while...
You're neglecting the energy which is dissipated by having the wheels turn the engine over (while coasting in gear). Coasting in neutral will get you much further, just test it yourself. Of course, the best solution is an engine kill switch, very popular with the hypermiling crowd.
Slashdot Mirror
On the other hand, writing documentation for things is one of the best ways I've found to really flesh out my understanding of something, so I bet these students will come out of the project with a significantly improved understanding of the language.
Agreed. Some years ago, me and four classmates wrote a 100-page "compendium" (less ambitious than a textbook, more like supporting material) for our C++ course. This was the summer after we took it; we sold it to subsequent classes for $15 per copy, with the proceedings going to our student society. It improved my understanding of the subject a lot, even though I had aced the exam. The only thing I would have done differently today is write it in English, not our native language, so that we could share it with more people.
What are you going to do with 64 GB of RAM on a single node? If you're actually using more than 2-4 GB per core, your program is fully limited by RAM speed, and you should REALLY be using MPI and more nodes with less ram each.
You have to keep in mind that "inflation" measures published by the government are kept artificially low by excluding food, energy and other things that have been getting more expensive lately. The technical term is "core inflation", but it's just an excuse for the politicians to goalseek inflation into what they want it to be. Now, gasoline would still have a positive price increase with a realistic inflation measure, but less than your numbers.
Ehm, I have to say that ECC is over-hyped by server hardware vendors, especially for CFD applications. The failure rate for modern RAM is 1 bit error per 1 GB per 1 month of simulation. To be honest, a typical CFD code will have to handle much worse errors than that due to random programming bugs (if you think your 35,000 lines-of-code program is bug free, well... it's not) etc., such that if it crashes or becomes unphysical from 12 bit errors in a month, you're screwed anyway.
On the other hand, if you're running a COBOL program that calculates how much you're employees taxes are going to be, then you really want ECC.
Okay, I'll answer this. From my use of a Core i7, you can tell that I'll be coding a serial app with some OpenMP in the slow parts to utilize all 4 cores. Now that's much easier than writing the same code for a GPU. If I were serious in developing the "blazing fast" stuff (which I'm not, my focus is on implementing new multi-physics models) I could spend the same amount of effort and target MPI instead of GPUs, and then go run it on the 22,000 core cluster in the basement next door.
The good thing about manufacturers speeding up SSE/AVX/etc. is that the linear algebra libraries (specifically the ATLAS implementation of BLAS and LAPACK) usually release code that makes use of the new hawtness in about six months after release. Do you know how much software relies on BLAS and LAPACK for speed?
If you're doing numerics, what the fuck (if you'll pardon my French) are you doing buying Apple? I'm working on two-phase Navier-Stokes solvers myself, and I just bought a new rig consisting of 3 boxes each with a Intel Core i7 @ 3.7 GHz, 12 GB RAM, an SSD drive and a big-ass cooling system. In total that cost less than the Mac Pro with a single Core i7 @ 3.3 GHz listed in that article.You're paying 3x more than you should, and you get what extra? A shiny case? Puh-lease.
Agreed. My previous camera (Canon SX20) had to be repaired twice, and then replaced (with an SX40), since it came from one of the first production lines which used leadless solder. The solder was bad, so the voltage converter would stop delivering the correct voltage after about 10 000 shots.
It's not that a particle has a theoretical probability of being somewhere with some probable momentum, no it will be at a very real place at a very real time with a very actual momentum. It's just that practically it's so complicated to predict it, that the best way we have come up till now are quantum mechanics .
Nope, you're wrong. Here are the experimental evidence which falsify your hypothesis. Bonus: Zombie Feynman.
Nature tends to generate more data than can be quickly processed
FTFY. I would even amend that to "be quickly processed, or even stored". Think about a fluid dynamics simulation for instance. Say you want to store a velocity vector at each grid point on a 400x400x1000 grid at 500 time increments. How much harddisk space will you need? A quick calculation comes out at 1.8 terabytes of data! And that's just for the velocity field, you can easily triple that number in a realistic CFD case.
This. Top Gear is the show who recommended the Skoda Yeti because Sienna Miller's head fits in the glovebox, routinely tests cars against fighter planes, and compares station wagons on how many pounds of cheese they can carry. I don't know what Tesla expected.
It's simple. Java was teh hawtness roughly ten years ago. It took four years for university CS departments to catch on, then four years for the first Java-loving graduates to start "entrepreneuring", and then two years for this to significantly influence Java usage. It's a basic convolution-type feedback.
Good designers make choices. Bad designers can't make up their minds and pass the buck down to users.
Look, it's not as if OSX has five toggles in the system menu, it has 29 separate tabs each with many options. It even has a tab called "Energy Saver", where this option would fit naturally. Stop trying to frame this omission as a conscious "design" choice when it's clearly not.
But if you include them all, you end up with a horrible mess of an options UI, that most people don't understand or get lost in.
Ah, but Apple, who are generally regarded as being a few orders of magnitude better at UI design than a random torrent-client devteam, should not have a problem here. A simple tickbox to enable "advanced settings" under the "Power management" tab in "System settings" (or whatever it is called) solves that problem. See e.g. VLC for a good demonstration - if you don't toggle from Simple to All in the settings window, there is no option for "which packetizer preprocesses the stream before muxing" and a million other such things. How many people do you think prefer to control the packetizer when looking at a movie? A grand total of ten out of six billion?
When it's only sensible for 70% of the use cases, there should be a software switch. How would it be a worse design decision to have the default behaviour as it is today, but a way to toggle it in the system settings? The original point in this thread was that users should be given maximum flexibility, especially when it's a matter of twenty lines of code in the OS. Apple fails to implement this, which is why many of us geeks use something else (cf. smartphones).
To clarify:
Point 1: I've never owned a Macbook, but I have friends who've been annoyed by this inability to close the lid while running.
Point 2: My current computer, a 4-core Dell Precision something-or-other, has compiled lots of code in my bacpack. I use my water bottle to ensure that the fan intakes on the bottom are free, and I open my backpack zipper on the side where the fan outlets are. Works like a charm (I logged the CPU temperature with a small script the first few times I tried it).
TL;DR:
I don't want my computer "protecting" me from doing something that could maybe be stupid. I want it to shut down if overheating, that's a much better solution that all laptops implement.
So you're saying if I buy a Mac, I have to give up on letting my code compile/run while my laptop is in my backpack on the way home? And that if I want to plug it into my stereo system and listen to music, I can't close it and put it in the shelf, but that it must take up space on the table?
Erm ... why? No really ... Why! The vast majority of accommodations we are talking about are software features. These are adjustments and settings which can be customised.
You're talking about a company which sells a computer OS which is mainly used on laptops, but where there is no toggle for whether closing the screen puts the laptop into sleep mode. It's a trivial feature found on all other OS's, but Apple seemingly can't be bothered to include it, and the people who buy the $999 Facebook machines adapt their usage patterns to their computers, not vice versa.
This needs to be modded up, it's the first sensible comment (by an AC nonetheless!) on this topic.
You're off by a factor of 1000. So it would provide power for roughly 60 people.
If the terrorists were really serious about levelling a small city, I think a massive fuel-air bomb would be much easier to execute than an actual nuke, even a small tactical one. Start out with a big tank of liquefied natural gas, use a small explosive to rupture the tank and let the LNG go through a rapid phase transition and subsequent turbulent mixing with air, wait a little before detonating the second small explosive to ignite the vapor cloud. Hang this from a blimp or whatever for added effect.
Also, unlike nukes, fuel-air explosives have a history of being used by terrorists.
Yes, and it's extremely annoying if you want to do anything to your own car. It's bad on the same level as proprietary connectors for phones and all that, but unfortunately the amount of people improving their own cars is too low to cause any consumer feedback to manufacturers.
And I don't mean adding stupid spoilers and boost chips and sillyness, I mean stuff like adding an extra pair of high beams that can be operated with the same button as the regular high beams. That will take some serious hacking on a modern car. If car manufacturers were good at making things, this wouldn't be a huge problem, but modern cars do so many things wrong that it's infuriating. Like putting lambertian leds in places where they should have put batwing ones, forcing me to put a diffuser in front of it so that my daughter is able to sleep in her car seat. Or making it a fifteen-minute job to remove the battery for charging it during the winter, when it should take two minutes. Or putting the light that activates when you open the trunk in the far left corner of the trunk, so that it doesn't light up anything if you actually have something in the trunk. I could go on about this for a while...
Interesting. I guess I'll have to check my car myself, but this sounds plausible.
Parent is right.
Also, it should be said that chemical -> electric -> kinetic is fairly common in large systems; see e.g. diesel-electric transmission.
You're neglecting the energy which is dissipated by having the wheels turn the engine over (while coasting in gear). Coasting in neutral will get you much further, just test it yourself. Of course, the best solution is an engine kill switch, very popular with the hypermiling crowd.