Spanish only in Spain, and Spain is more of a Second World country
and the US, which is (soon) also more of a 2nd world country! </irony> On a more serious note: doing interpolation from current data over about the USian population, Spanish will actually be the largest native language of the US somewhat soonish, let's just see if that or the downturn of the american economy comes first...
When even the creators of the machine are surprised by certain behaviors, THEN we will be getting close.
Uhm, quite a lot of real-world software already has that reaction, even among its creators. However, it tends to be more a sign of the creators not understanding what they have built rather than the software to actually be intelligent...
There are Ethernet-based PS/2 extension cables. Probably not really that cheap, but it would mean you would have a cable long enough too reach your TV set, and use any standard PS/2 based wireless desktop set (keyboard+mouse). I suppose that it relies on your server having PS/2 inputs though.
These extensions are commonly used in e.g. schools where the actual computers are hidden in a cupboard (i.e. to prevent theft) with only monitors, keyboards and mice on the desks.
"And that is how a scientific research's value is measured really. The more your research is cited in other publications, the greater the value of your research is."
Well, one quite good way to get cited is to have a result published, which is wrong, but for some reason at least is wrong in an interesting way. So, no, more cited works are typically not better, although a certain correlation exists.
Uh, there is still a lot of industrial (and embedded) equipment still being shipped which is still only accessible over serial connection. It's not even closely as obsolete as you suggest...
You're theoretically correct. It shouldn't get more than m times faster if m processors are used. Yay!
However, there is a limitation - the theories generally assume that processing is made at the same speed if it is done by a single processor, or divided by several processors. Due to the hardware architecture in modern processors - especially how cache is constructed - this assumption is not the reality, it's just an approximation.
To take a concrete example: One task takes 8 seconds to run on a single node, and uses 4 MB of memory during that time. Assuming the processor has 1 MB of cache, this will clearly be slow due to many memory accesses going to the main memory, not to the on-chip cache which is far faster.
If the work is split - assume on 8 machines, theoretically it shouldn't be possible to do it faster than 1s. However, depending on the problem, all processors may not need to access all data. Assuming each of the 8 processors only need to access 1 MB of data (which would just barely fit into the cache), the execution will be a bit faster because all memory accesses can go directly to the cache. Perhaps each processor will finish in just 0,8 ms (instead of the theoretical 1 ms), giving a speedup of 10, surpassing the theoretical maximum of 8.
This is very problem dependant though. Sometimes, all the processors will need to access all the memory even if work can be split. Further, the typical case is that the speedup is lower than the theoretical maximum, due to inefficiences. Data has to be distributed before processing can start, sometimes there are dependencies between different parts of the algorithm forcing different processors to wait on each other etcetra. Finally, for each algorithm and problem size, an "speedup with infinite processors" can be defined which marks an upper bound of the possible speedup, regardless of how many processors are used.
N/M seconds would be amount of time taken by a theoretically 'perfect' setup; it cannot be achieved in practice (I am fairly certain, can someone correct me if I'm wrong?).
You're wrong. It can be achieved (in particular cases). It can even sometimes be surpassed - but then it's because of some superlinear effect - like the work being exactly the right size to not fit into cache or RAM on one processor, but being small enough to fit when distributed, so when distribution is used the cache is effectively used or thrashing avoided. Theoretically, the limit shouldn't be surpassable (and very hard to actually reach).
As the parent was talking about ETH, I don't really think they are that relevant.
Erasmus
AFAIK this is only available to students at one European university going to another (I used to be an Erasmus student myself). This would still be a good opportunity - but might mean learning two languages in addition to English, not one. I may be a bit ignorant to the language knowledge of American people, but it might just be hoping for too much.:-) Still, I agree learning language is a good idea and starting off with a semester of language studies would probably be a good idea at least if the language is completely new (with some luck, it could perhaps be possible to find a summer language course?)
It all really depends a bit of which area you're interested in. KTH, Chalmers, LU are the big ones and the most well-recognized names (not MIT-level, but at least KTH and Chalmers tend to be included in the top 200 in different global rankings, LU is harder to judge as they are a general university and not a specialized technical university). The others often have good reputation in specialized areas only - and which university is good in which area is different. I'm not quite sure about LiU, unfortunately, so you'll have to do your research on your own...
Uppsala is big and well-known, but they don't really have anything worth counting in the engineering area (most others are covered). Lund is a good addition though.
Add Sweden to the list as well, all university education in Sweden is completely free of cost. The main issue with Germany and Switzerland might be that many BS programs start off in German. In Sweden many start in Swedish. Language just might be a small barrier here...
Yeah, I missed the part that it's supposedly a feature in my comment. Still it's quite obvious that the behaviour of MS' formatter is heavily dependant on how HP's printing driver works. Of course MS doesn't know how all printer makers' driver works, so that makes the behavior of the formatter unspecified. If MS calls it a "bug" or a "feature" I don't really care - to me it's an obvious design error, nothing else.
I'm pretty sure MS doesn't know what its Word formatter does, and I even have proof for it:
If I switch printers between the Adobe PDF and the HP printers at the office, the layout of my documents I edit in MS Word 2003 changes slightly (line lengths, row breaks, distance between rows, etc). This has been a major issue when I've had to submit papers etc. and switched to the PDF printer from the HP printers (I like to read drafts on paper as it is easier to correct), just to see the paper that I had just crafted to barely fit under the 8 page limit now is 11 pages long (with 4 of them left half blank due to formatting issues).:-(
Hmm, the letter you wrote didn't really cover the royalty back-licensing. I suppose they trust you well, I would have added a clause granting them the license right ahead anyway, just to make sure.
Well, they can complain all they want but I wasn't really keen on clearing the computer from spyware and other crap twice a year. If they wanted nothing to change (including the browser), I wouldn't have continued giving them free support...:) And except that no one really bothered so I'd say it was okay. Using Linux instead would be a bigger step though.:)
Probably works great for Linux, but with Windows you want the clean install.;-)
You're right about the psychological stuff. My dad even complained the bookmarks weren't there anymore, although I had imported them. Turned out that FF sorted them, while IE didn't. So probably, I'll wait for the hard drive failure, and blaim it on Windows to get the permission to reinstall with whatever Linux dist seems most suitable. Xubuntu is probably a nice one [checking Wikipedia entry...]
I'll do it eventually, I'm just too lazy to do a complete OS reinstall. As Ubuntu seems to be coming along perhaps it'll be a Linux computer when I finally get around.
Memory would be quite possible to upgrade, I even have a bunch of compatible memories in my own server computer (which really doesn't need 320 MB when it's even slower), although it's not really that prioritized right now. And despite the old hard drive, it runs FF quite well.:)
Slashdot Mirror
Spanish only in Spain, and Spain is more of a Second World country
and the US, which is (soon) also more of a 2nd world country! </irony> On a more serious note: doing interpolation from current data over about the USian population, Spanish will actually be the largest native language of the US somewhat soonish, let's just see if that or the downturn of the american economy comes first...
When even the creators of the machine are surprised by certain behaviors, THEN we will be getting close.
Uhm, quite a lot of real-world software already has that reaction, even among its creators. However, it tends to be more a sign of the creators not understanding what they have built rather than the software to actually be intelligent...
There are Ethernet-based PS/2 extension cables. Probably not really that cheap, but it would mean you would have a cable long enough too reach your TV set, and use any standard PS/2 based wireless desktop set (keyboard+mouse). I suppose that it relies on your server having PS/2 inputs though.
These extensions are commonly used in e.g. schools where the actual computers are hidden in a cupboard (i.e. to prevent theft) with only monitors, keyboards and mice on the desks.
"And that is how a scientific research's value is measured really. The more your research is cited in other publications, the greater the value of your research is."
Well, one quite good way to get cited is to have a result published, which is wrong, but for some reason at least is wrong in an interesting way. So, no, more cited works are typically not better, although a certain correlation exists.
"Zoom out"
"Now consider that it takes at least ten times as much energy to make a car than the car will use during its entire lifetime."
Uhm, wrong way around. It uses about ten times as much energy during its life-time than by manufacturing...
What exactly is Microsoft good at? =/
Buying up other companies making good products?
'(the frankly much better)'
Please show me a well-working 50 m USB cable. For many applications, USB is superior, but not all.
Uh, there is still a lot of industrial (and embedded) equipment still being shipped which is still only accessible over serial connection. It's not even closely as obsolete as you suggest...
Ok here's a try at an explanation:
You're theoretically correct. It shouldn't get more than m times faster if m processors are used. Yay!
However, there is a limitation - the theories generally assume that processing is made at the same speed if it is done by a single processor, or divided by several processors. Due to the hardware architecture in modern processors - especially how cache is constructed - this assumption is not the reality, it's just an approximation.
To take a concrete example: One task takes 8 seconds to run on a single node, and uses 4 MB of memory during that time. Assuming the processor has 1 MB of cache, this will clearly be slow due to many memory accesses going to the main memory, not to the on-chip cache which is far faster.
If the work is split - assume on 8 machines, theoretically it shouldn't be possible to do it faster than 1s. However, depending on the problem, all processors may not need to access all data. Assuming each of the 8 processors only need to access 1 MB of data (which would just barely fit into the cache), the execution will be a bit faster because all memory accesses can go directly to the cache. Perhaps each processor will finish in just 0,8 ms (instead of the theoretical 1 ms), giving a speedup of 10, surpassing the theoretical maximum of 8.
This is very problem dependant though. Sometimes, all the processors will need to access all the memory even if work can be split. Further, the typical case is that the speedup is lower than the theoretical maximum, due to inefficiences. Data has to be distributed before processing can start, sometimes there are dependencies between different parts of the algorithm forcing different processors to wait on each other etcetra. Finally, for each algorithm and problem size, an "speedup with infinite processors" can be defined which marks an upper bound of the possible speedup, regardless of how many processors are used.
If this is a topic that interests you, more information can be found at the course webpage for the parallel algorithms course that I took a couple of years ago.
N/M seconds would be amount of time taken by a theoretically 'perfect' setup; it cannot be achieved in practice (I am fairly certain, can someone correct me if I'm wrong?).
You're wrong. It can be achieved (in particular cases). It can even sometimes be surpassed - but then it's because of some superlinear effect - like the work being exactly the right size to not fit into cache or RAM on one processor, but being small enough to fit when distributed, so when distribution is used the cache is effectively used or thrashing avoided. Theoretically, the limit shouldn't be surpassable (and very hard to actually reach).
CH: Or French. Or Italian.
:-) Still, I agree learning language is a good idea and starting off with a semester of language studies would probably be a good idea at least if the language is completely new (with some luck, it could perhaps be possible to find a summer language course?)
As the parent was talking about ETH, I don't really think they are that relevant.
Erasmus
AFAIK this is only available to students at one European university going to another (I used to be an Erasmus student myself). This would still be a good opportunity - but might mean learning two languages in addition to English, not one. I may be a bit ignorant to the language knowledge of American people, but it might just be hoping for too much.
It all really depends a bit of which area you're interested in. KTH, Chalmers, LU are the big ones and the most well-recognized names (not MIT-level, but at least KTH and Chalmers tend to be included in the top 200 in different global rankings, LU is harder to judge as they are a general university and not a specialized technical university). The others often have good reputation in specialized areas only - and which university is good in which area is different. I'm not quite sure about LiU, unfortunately, so you'll have to do your research on your own...
Uppsala is big and well-known, but they don't really have anything worth counting in the engineering area (most others are covered). Lund is a good addition though.
Add Sweden to the list as well, all university education in Sweden is completely free of cost. The main issue with Germany and Switzerland might be that many BS programs start off in German. In Sweden many start in Swedish. Language just might be a small barrier here...
... which until today remains to be shown!
Yeah, I've also made sure to use that silly workaround. Fortunately, my work computer is equipped with Adobe...
They should rename it to WYSIWYGALAYDSP, methinks (What you see is what you get as long as you don't switch printers)... :-(
Yeah, I missed the part that it's supposedly a feature in my comment. Still it's quite obvious that the behaviour of MS' formatter is heavily dependant on how HP's printing driver works. Of course MS doesn't know how all printer makers' driver works, so that makes the behavior of the formatter unspecified. If MS calls it a "bug" or a "feature" I don't really care - to me it's an obvious design error, nothing else.
I'm pretty sure MS doesn't know what its Word formatter does, and I even have proof for it:
:-(
If I switch printers between the Adobe PDF and the HP printers at the office, the layout of my documents I edit in MS Word 2003 changes slightly (line lengths, row breaks, distance between rows, etc). This has been a major issue when I've had to submit papers etc. and switched to the PDF printer from the HP printers (I like to read drafts on paper as it is easier to correct), just to see the paper that I had just crafted to barely fit under the 8 page limit now is 11 pages long (with 4 of them left half blank due to formatting issues).
Hmm, the letter you wrote didn't really cover the royalty back-licensing. I suppose they trust you well, I would have added a clause granting them the license right ahead anyway, just to make sure.
Well, they can complain all they want but I wasn't really keen on clearing the computer from spyware and other crap twice a year. If they wanted nothing to change (including the browser), I wouldn't have continued giving them free support... :) And except that no one really bothered so I'd say it was okay. Using Linux instead would be a bigger step though. :)
Probably works great for Linux, but with Windows you want the clean install. ;-)
You're right about the psychological stuff. My dad even complained the bookmarks weren't there anymore, although I had imported them. Turned out that FF sorted them, while IE didn't. So probably, I'll wait for the hard drive failure, and blaim it on Windows to get the permission to reinstall with whatever Linux dist seems most suitable. Xubuntu is probably a nice one [checking Wikipedia entry...]
I'll do it eventually, I'm just too lazy to do a complete OS reinstall. As Ubuntu seems to be coming along perhaps it'll be a Linux computer when I finally get around.
:)
Memory would be quite possible to upgrade, I even have a bunch of compatible memories in my own server computer (which really doesn't need 320 MB when it's even slower), although it's not really that prioritized right now. And despite the old hard drive, it runs FF quite well.
"Well, if the likelihood of being wrong is higher than the likelihood of being right, then it's clear the front seats are safer."
No, it's not clear. They'd be wrong even if the risk was only identical, which is a special case not covered by you.