The situation in America is a little bit different. In Europe you typically want to go to the center of a city. In Paris once you are in Gare du Nord, you are quickly anywhere you want provided you are not already where you want to go. And it is pretty much like that for every major city. Also, despite having a good HSR trian system, all trip that do not go toward or away from paris are a nightmare (try a Lyon-Bordeaux or Lyon-Strasbourg for instance).
In the US, you typically do not care about being downtown. I have been living in Columbus, OH (ok not the biggest city) for 4 years and I only went downtown 5 times (once to visit downtown, once to go to a museum, once to drop somebody at a justice court, once to go to social security administration and once for july 4th celebration). There is no point going downtown in the US for most city I believe. And since the cities are so spread out, you'll have to rent a car to go anywhere. So you'll end up needing to drive to and from the train station. So you don't save the overhead of going to/from the airport. That makes the train much less interesting. It would pretty much be a slower plane. And it will not be cheap. Actually if you aim at cheap, you have a bus system that connects major cities. It is much slower than a high speed train. But it is also much cheaper.
So a train would end up being a compromise between the slow/cheap bus and the fast/expensive plane. It is not clear there is a real market.
remember the debian ssh key fiasco. How many people still use some of these not-so-random keys? I agree that posting your authorized_keys should not be a problem. But the least information on your setup out there is the better.
faulty hardware is a common problem in clusters. We still have a 5 years old cluster at work that is falling in pieces due to faulty processor, faulty memory, faulty power supplies and faulty hard drives. Some errors are just weird. You never saw a machine having errors in memtest86 but stopping to have them once you swapped two dimms? I see that from time to time and I don't spend much time dealing directly with hardware.
so do you suggest guildwars incorporate a gcc testsuite run in parallel with the game?
Re:I don't believe 1% of computers give wrong answ
on
Whose Bug Is This Anyway?
·
· Score: 4, Informative
I actually believe it. I am sure they might have think of floating point precision problem. But most likely they only used integers. That's what prime 95 and memtest are doing. Integer and memory operations uncover most common hardware failure. I encountered many computers with faulty hardware when stressed. And I am sure guildwars was stressful.
I've been living in the US for 4 years now. And I still get mixed up from time to time. This convention makes little sense to me. Note that in English "July 4th, 2012" is a valid date, so is "the 4th of july, 2012". In French, there is no equivalent to the first form, only the second one.
I heard a psychologist on NPR last week saying that most massive shooting are long time premeditated actions. Almost no shooter just goes crazy take a gun and shoot everybody. They all spend weeks at it.
Well, and that's the difference between scale up and scale out in parallel computing. Throughput is typically given by many simple processing units. Latency is typically given by highly specialized processing units.
If it is throughput you care about, simple is the way to go.
This is not an help website. This is a technews site. As in, it is a newspaper. Some random guy does not know how to use a tool and it somewhat got posted here. Clearly the guy could learn. But somewhat it got posted here.
There is nothing interesting in what he has to say. He even explain how to measure the time a page took to generate in php. Seriously, some one here that might care does not know how to do that?
The guy is clueless and still somehow got posted on slashdot.
Even better than "pay the $3.99 and upload the app". Pay the $3.99, get the source. Then talk to upstream to merge the changes upstream. Then redistribute the app from upstream. You do not want to have to upkeep a separate code tree for android. Merging it to upstream might just give you free updates of the engine.
as said before it is more $1000 per household, still cheap. The main question is how to operate and upkeep that network, how much will that cost be? Also Who will control the network. Somewhat having 100% of the network infrastructure controlled by google does not sounds like a good thing.
There is an other reason to keep a phone on my desk. That is the only device that can call a phone number anywhere in the world and that I will not have to pay for. My institute will pay for it. There is no way, I'll call people to discuss work for hours out of state or out of country and pay the communication with my cell phone.
If I get a company skype account (or whatever) that can call and be called by regular phone number, sure, i'll use it. Otherwise, I need a phone on my desk.
Indeed, high gigaflops is easy, useful high gigaflops is hard. You can easily build a processor that only support float-addition and nothing else with a 1024 bit SIMD register clocked at 4 Ghz. And voila, you get 128Gflop/s per core. Problem is: it is useless.
The question is not how many adds or muls you can do per second in an ideal application for your architecture. The question is how many adds or muls (or whatever you need to measure) you can do per second on a real application.
For instance, the top-500 uses linpack, that measures how fast one can multiply dense matrices. That problem is only of interest to a small amount of people.
Exactly, I am doing parallel programming and HPC for a leaving. I do not believe in compiler automagic^Wautomatic parallelisation. Typing is only a small parts of the problem. Most algorithms need to be written VERY differently to be able to exploit the parallelism within. Some times, you actually need to have variable concurrency and race condition within your algorithm but you have ways to fix problems whenever they appear.
In brief, I do not think that compilers will be able to do anything more than quite basic stuff. What ever is proposed here do not appear to me to be significantly more powerful than Cilk (now Intel Cilk+).
I am frequently checking the state of tablet technology to do paper annotation. I write and annotate documents a lot. And any interface which is not paper and pencil like is typically useless to me. So all the tablet tend to be terrible on their own. I had a look at those stylus for ipad, that's better but still not enough. It is too imprecise which prevent proper annotation and drawings
Though, Itried a galaxy note (the phone one) with the spen, and that was a very convenient device to annotate a document. Except it is phone size so it is too small for real life use. The tablet version should be perfect. If you want to annotate stuff, you should check it out and see if it works for you.
In the humble bundle for android, a significant number of games were also released for the windows, os X and linux. I bought two of teh 4 bundles and more than half the games have been released on PC.
I just tried on my htc desire with cianogene mod. It is much slower than the "original" version. I need to click a link to read the summary, which completely defeat the purpose of having a homepage. If I want to click the stories I want to read, i'd use an RSS feed reader not slashdot home page.
Slashdot Mirror
The situation in America is a little bit different. In Europe you typically want to go to the center of a city. In Paris once you are in Gare du Nord, you are quickly anywhere you want provided you are not already where you want to go. And it is pretty much like that for every major city. Also, despite having a good HSR trian system, all trip that do not go toward or away from paris are a nightmare (try a Lyon-Bordeaux or Lyon-Strasbourg for instance).
In the US, you typically do not care about being downtown. I have been living in Columbus, OH (ok not the biggest city) for 4 years and I only went downtown 5 times (once to visit downtown, once to go to a museum, once to drop somebody at a justice court, once to go to social security administration and once for july 4th celebration). There is no point going downtown in the US for most city I believe. And since the cities are so spread out, you'll have to rent a car to go anywhere. So you'll end up needing to drive to and from the train station. So you don't save the overhead of going to/from the airport. That makes the train much less interesting. It would pretty much be a slower plane. And it will not be cheap. Actually if you aim at cheap, you have a bus system that connects major cities. It is much slower than a high speed train. But it is also much cheaper.
So a train would end up being a compromise between the slow/cheap bus and the fast/expensive plane. It is not clear there is a real market.
Actually, I do not think the Chinese built it. I believe they contracted european train builders.
remember the debian ssh key fiasco. How many people still use some of these not-so-random keys? I agree that posting your authorized_keys should not be a problem. But the least information on your setup out there is the better.
"I have had personal 2 week projects drag on for 6 months no problem. So if anyone has some great rules of thumb I would love to know."
You know what they say: times 2, round up.
2 weeks x 2 = 1 month. round up : 1 year.
faulty hardware is a common problem in clusters. We still have a 5 years old cluster at work that is falling in pieces due to faulty processor, faulty memory, faulty power supplies and faulty hard drives. Some errors are just weird. You never saw a machine having errors in memtest86 but stopping to have them once you swapped two dimms? I see that from time to time and I don't spend much time dealing directly with hardware.
so do you suggest guildwars incorporate a gcc testsuite run in parallel with the game?
I actually believe it. I am sure they might have think of floating point precision problem. But most likely they only used integers. That's what prime 95 and memtest are doing. Integer and memory operations uncover most common hardware failure. I encountered many computers with faulty hardware when stressed. And I am sure guildwars was stressful.
What are you talking about? They are going at least 50 miles an hour!
I've been living in the US for 4 years now. And I still get mixed up from time to time. This convention makes little sense to me. Note that in English "July 4th, 2012" is a valid date, so is "the 4th of july, 2012". In French, there is no equivalent to the first form, only the second one.
I heard a psychologist on NPR last week saying that most massive shooting are long time premeditated actions. Almost no shooter just goes crazy take a gun and shoot everybody. They all spend weeks at it.
That's the last straw Facebook! I've had it. I'm deleting my account!
Oh wait! I don't have one...
Well, and that's the difference between scale up and scale out in parallel computing. Throughput is typically given by many simple processing units. Latency is typically given by highly specialized processing units.
If it is throughput you care about, simple is the way to go.
This is not an help website. This is a technews site. As in, it is a newspaper. Some random guy does not know how to use a tool and it somewhat got posted here. Clearly the guy could learn. But somewhat it got posted here.
There is nothing interesting in what he has to say. He even explain how to measure the time a page took to generate in php. Seriously, some one here that might care does not know how to do that?
The guy is clueless and still somehow got posted on slashdot.
Even better than "pay the $3.99 and upload the app". Pay the $3.99, get the source. Then talk to upstream to merge the changes upstream. Then redistribute the app from upstream. You do not want to have to upkeep a separate code tree for android. Merging it to upstream might just give you free updates of the engine.
as said before it is more $1000 per household, still cheap. The main question is how to operate and upkeep that network, how much will that cost be? Also Who will control the network. Somewhat having 100% of the network infrastructure controlled by google does not sounds like a good thing.
There is an other reason to keep a phone on my desk. That is the only device that can call a phone number anywhere in the world and that I will not have to pay for. My institute will pay for it. There is no way, I'll call people to discuss work for hours out of state or out of country and pay the communication with my cell phone.
If I get a company skype account (or whatever) that can call and be called by regular phone number, sure, i'll use it. Otherwise, I need a phone on my desk.
better, sue for damages and set the price to "one millenium falcon"
most browser for android have an option "request the desktop version of the site"
That makes sense, Android users are doing really work, the iOS users just look at how shiny their toy is!
(Why am I modded -1 troll? :) )
Indeed, high gigaflops is easy, useful high gigaflops is hard. You can easily build a processor that only support float-addition and nothing else with a 1024 bit SIMD register clocked at 4 Ghz. And voila, you get 128Gflop/s per core. Problem is: it is useless.
The question is not how many adds or muls you can do per second in an ideal application for your architecture. The question is how many adds or muls (or whatever you need to measure) you can do per second on a real application.
For instance, the top-500 uses linpack, that measures how fast one can multiply dense matrices. That problem is only of interest to a small amount of people.
Exactly, I am doing parallel programming and HPC for a leaving. I do not believe in compiler automagic^Wautomatic parallelisation. Typing is only a small parts of the problem. Most algorithms need to be written VERY differently to be able to exploit the parallelism within. Some times, you actually need to have variable concurrency and race condition within your algorithm but you have ways to fix problems whenever they appear.
In brief, I do not think that compilers will be able to do anything more than quite basic stuff. What ever is proposed here do not appear to me to be significantly more powerful than Cilk (now Intel Cilk+).
I am frequently checking the state of tablet technology to do paper annotation. I write and annotate documents a lot. And any interface which is not paper and pencil like is typically useless to me. So all the tablet tend to be terrible on their own. I had a look at those stylus for ipad, that's better but still not enough. It is too imprecise which prevent proper annotation and drawings
Though, Itried a galaxy note (the phone one) with the spen, and that was a very convenient device to annotate a document. Except it is phone size so it is too small for real life use. The tablet version should be perfect. If you want to annotate stuff, you should check it out and see if it works for you.
crap, I want a drink now...
In the humble bundle for android, a significant number of games were also released for the windows, os X and linux. I bought two of teh 4 bundles and more than half the games have been released on PC.
I just tried on my htc desire with cianogene mod. It is much slower than the "original" version. I need to click a link to read the summary, which completely defeat the purpose of having a homepage. If I want to click the stories I want to read, i'd use an RSS feed reader not slashdot home page.
I won't use it.