I have looked at the assembly output of gcc, and have been horrified by how bad it is. Seriously, it's cringe-inducing. MSVC is better but still not great (though MSVC.NET 2005 seems noticeably improved).
Well, back in the day when we wrote large programs in assembly, we didn't optimize every line of code either. It's no different than C or other language, you only optimize the parts you need to. But back then even casually-written assembly was smaller and faster than the C and Pascal compilers generated, and if you buffed the code a bit you could really make it scream. But this was done only in certain places (video routines, memory management, etc). Nowadays we don't worry so much about size, so we code the non-critical bits in a higher-level language, and we don't worry so much about speed so we code the quick bits in C or C++.
I've seen this attitude a lot in recent years that assembly is something separate from the rest of computer science, and I'm curious where it's coming from. Assembly is just another language family, and it can be learned and programmed in like any other language, and the same principles of programming apply to it like any other language, you use design patterns similarly, etc. It's a bit verbose, and reads vertically instead of horizontally, but you get used to these quirks pretty quickly.
Spoken like someone who has never written assembly outside of class.
I did a fair chunk of it in my younger days doing embedded and DOS development, where it was common for 10%-20% of the code to be in assembly. I haven't had much need for it the last decade or so, but had to write a few hundred lines last year to solve a bottleneck that couldn't easily be beaten algorithmically. The assembly outperformed the C++ handily (~10x faster), and this was on a modern processor with all the bells and whistles that are supposed to make things difficult (64-bit Xeon).
Assembly is like any other language, it isn't really that hard once you're familiar with it. This wasn't necessarily true back in the early 80x86 days where you were seriously starved for registers, but once you get 10+ GP registers available then assembly gets pretty easy to write -- I always preferred the 680x0 over the 80x86 for this reason. Anyway, the hard part is the design and verification of the algorithm, the rest is just coding and debugging.
Your post didn't say what your clarification claims it said. You stated (smugly) that the swedish laws were superior because it was only illegal to leak classified information and not unclassified information. Sweden's laws in this situation are the same as the U.S. It is true that in the U.S. you can get fired for leaking non-classified information, but this is because you pissed off your boss and because there are few legal protections from getting fired, you're not getting fired because you broke the law. You may lose your job, but you won't go to jail.
In the U.S. if you leak information to expose criminal action then you are generally protected from losing your job. However, if the information you need to leak is itself classified, then there are special channels set up to handle this without breaking classification. These safeguards may or may not work well, but it is the system that *congress* put in place, so they are ultimately responsible.
Again, what Gonzales said should not be controversial -- he is simply stating the current situation in U.S. code: leaking classified information is a crime, and knowingly passing that information on is also a crime, whether you pass this on to everybody (journalist) or to just a few (spy). Most countries (including Sweden) have laws similar to this.
The laws in question here are similar. Leaking non-classified information is not generally a crime. In this case, Gonzales (the attorney general of the U.S.) stated that leaking and publishing *classified* information is a crime.
He didn't make this up one day -- this is the current state of U.S. law. This generally has only been enforced against the leakers and not against the press that published it, but the law states that knowingly leaking or publishing classified information is a crime. The press has gotten used to having a free pass on this, and is getting upset now that Gonzales is talking about closing that gate, but the gate was put there by Congress.
XBox Live Arcade is available whether you have an XBL subscription or not. The XBL subscription allows you to purchase additional games over the Live service, but the free "Silver" subscription level allows you to buy arcade games, download trailers and demos and patches, and buy additional game content (mmmm, useless horse barding). You just can't do multiplayer -- that takes the $$$ "Gold" subscription.
You seem to be under the delusion that the basic X360 costs $420 (it's $300, even the non-crippled version is $400). Or that x360 games cost $100 a pop. Granted, the original poster's math was off a bit as well, unless he was assuming some of those games were bought used, which may be a reasonable assumption since it's been out awhile and there is a decent used game selection at your local gamestop.
What is completely overlooked is that in the long run, such a system may have unintended/unforseen costs that make mockery of the current benefits. In addition, the hardening of the regulatory systems is amazingly costly for stifling innovation and other "inventions we can sell" features that helped create the boom economy of the previous system.
What we don't need is engineering sites spewing this kind of crap. Lots of engineers have lost their jobs due to the implementation of these systems, and the more criticism we can muster from the engineering press, the more iPods and Ferraris we can keep ourselves in.
S-Ox is disastrous. Please stop spewing this crap now.
I've had the DirecTV HD Tivo for nearly a year now after upgrading from an old series 1 unit. It can definitely record two HD signals at once, and it has a "ratio" button (upper left of the remote) for flipping through the various aspect ratios. It also has multiple options for viewing the guide, including a more normal format for the guide (although we stuck with the old TiVo guide format since we were already used to it).
The reason it doesn't mean "piss" to you is because you didn't grow up here. It's pretty exclusively a kid's word -- you'll never hear someone over the age of 6 or 7 use it unless they're talking to a child under the age or 6 or 7.
But as adults we still know what it means; it's just too bad Nintendo didn't:-(
I think you are mistaken about who made it to 64 bits first. I had a 64-bit DEC Alpha with Win NT 3.51 back in the mid-90's, and it ran just fine thank you. More recent descendants of Win NT have supported the Intel Itanium systems. And now XP runs just fine on the 64-bit Xeons and Durons.
Windows did make the jump to 64-bit x64 later than linux, but they jumped pretty quick once the EM64T was announced; they don't call the two companies WinTel for nothing.
I don't know about that. From the article, it seems like you could make a decent wheel by putting the revolution controller core into a wheel.
I can imagine a wheel controller without the mounting base on it, with paddle shifters on the backside. You just hold it out in front of you, turn it, push out to accelerate, pull back to decelerate. No expensive sensors needed, it's all in the revolution controller module. You need something to connect the paddle shifters to the a/b buttons, but that's no biggie. I seem to remember the article also mentioned that the controller supported force feedback as well.
Fencing is fun to watch if you know what to watch for, but you're right that the action is very fast.
In some ways it's similar to american football that way, brief periods of very complex action punctuated by long periods where everything is reset.
What fencing needs is (a) visual highlighting of the blades, like televised hockey does with the puck or football does with the LOS and (b) none of this reset-after-a-point BS -- just keep going -- wasn't that the whole point of the electric foils, to eliminate the slow judging process?
Besides, blade highlighting would make them look like lightsabers. Foil fencing = boring, lightsaber dueling = ratings. Encourage the players to paint their mask screen like darth maul.
I know this was supposed to be one of the "microsoft questions" at one time, but this answer is just silly.
The reason round manhole covers don't fall down round manholes because they are larger than the hole, not because of their shape. Square manhole covers on square manholes work fine as long as the cover is large enough (side-of-cover longer than sqrt(2)*side-of-hole, more-or-less, since the thickness of the cover plays a part as well). Similarly for any other shape.
Round covers simply solve the problem with a low material consumption. I suspect an equilateral triangle works well too, but is simply too inefficient from a usability perspective.
I got Doom III for the xbox. Spent a few hours playing it, then spent the next few days playing the xbox port of Doom I and II that came with it. Never have made it back to III.
But then, I still spend an inordinate amount of time in nethack and ctetris.
They're switching to Intel because they've been abandoned three times before. First with the 680x0, then with the Motorola PPC, and finally with IBM. Now they simply want a nice safe partner. Intel may not be exciting, but they'll be there for the long haul.
VPC is slower under the G5 than it was under the G4 -- the G4 can flip its byte ordering to match the x86.
I've got a 1ghz G4 ibook, and benchmarking some of my own code under VPC show it running at about 750 mhz equivalent. This is fast enough, btw, that apps compiled with the MSVC compiler can run faster under VPC than they do compiled natively with the GNU compiler, and before you flame about compile options, neither the MS nor GNU compilers were using any aggressive optimization options -- the point is that the CPU emulation (at least on the G4) is good enough that it falls into the noise category.
The IBM-sourced G5 in the new macs doesn't support the byte-order flag, and working around this causes a huge performance hit. What is still unknown is if the XBox360 version of the PPC supports the byte-order flag -- if they're serious about the emulation then it may well.
What *really* sucks about VPC are 1) the graphics emulation -- it currently emulates an old dumb S3 chipset, and 2) it takes a little bit before he's got a good working set loaded into his translated code cache, which means it can be sluggish for awhile. Hopefully VPC is getting much better graphics emulation capability for the next version.
And the lisp machines had something similar as well. the Symbolics Ivory processor used 40-bit words, 32-bits of address/data, and 8 bits of type info.
One of the neat side-affects of this is that there was only one opcode for, say, "add". It looked at the data types and figured out wether to do a fp or integer add, and could trap on anything else -- the standard trap handler would detect mismatched types and perform the appropriate coercions. Similarly, the array-reference opcode knew (a) whether it was a char array or an int array or something else, and the size, and handle the offset scaling and range checking appropriately.
There was also a magic "forwarding" type value, that said in effect "the thing that used to be here is now over there", where "there" was an address in the data field of that value. The hardware would automatically chase the pointer to the new location to get the real data. This meant that you could safely move objects around in memory without having to patch all of the pointers. There was a low-priority OS thread that walked memory finding pointers to redirected memory and patching them up, and would reclaim the memory occupied by the old object once no pointers to it were left around.
This was used by the paging system to rearrange objects around memory so that objects in the working set were compacted into pages in the working set. This allowed the symbolics machines to have virtual:physical memory ratios > 100:1 with acceptable performance.
All this complicated hardware wasn't the snappiest thing in the world, but it was pretty cool.
It indexed the CPP files on my system just fine. It seems to be content based, not extension based (makes sense given that the web google does the same thing).
It also seems to have brought back the finder bug where finder would crash if you tried to drag more than one item (This had been fixed in the most recent OS update)
Slashdot Mirror
I have looked at the assembly output of gcc, and have been horrified by how bad it is. Seriously, it's cringe-inducing. MSVC is better but still not great (though MSVC.NET 2005 seems noticeably improved).
Well, back in the day when we wrote large programs in assembly, we didn't optimize every line of code either. It's no different than C or other language, you only optimize the parts you need to. But back then even casually-written assembly was smaller and faster than the C and Pascal compilers generated, and if you buffed the code a bit you could really make it scream. But this was done only in certain places (video routines, memory management, etc). Nowadays we don't worry so much about size, so we code the non-critical bits in a higher-level language, and we don't worry so much about speed so we code the quick bits in C or C++.
I've seen this attitude a lot in recent years that assembly is something separate from the rest of computer science, and I'm curious where it's coming from. Assembly is just another language family, and it can be learned and programmed in like any other language, and the same principles of programming apply to it like any other language, you use design patterns similarly, etc. It's a bit verbose, and reads vertically instead of horizontally, but you get used to these quirks pretty quickly.
Spoken like someone who has never written assembly outside of class.
I did a fair chunk of it in my younger days doing embedded and DOS development, where it was common for 10%-20% of the code to be in assembly. I haven't had much need for it the last decade or so, but had to write a few hundred lines last year to solve a bottleneck that couldn't easily be beaten algorithmically. The assembly outperformed the C++ handily (~10x faster), and this was on a modern processor with all the bells and whistles that are supposed to make things difficult (64-bit Xeon).
Assembly is like any other language, it isn't really that hard once you're familiar with it. This wasn't necessarily true back in the early 80x86 days where you were seriously starved for registers, but once you get 10+ GP registers available then assembly gets pretty easy to write -- I always preferred the 680x0 over the 80x86 for this reason. Anyway, the hard part is the design and verification of the algorithm, the rest is just coding and debugging.
Your post didn't say what your clarification claims it said. You stated (smugly) that the swedish laws were superior because it was only illegal to leak classified information and not unclassified information. Sweden's laws in this situation are the same as the U.S. It is true that in the U.S. you can get fired for leaking non-classified information, but this is because you pissed off your boss and because there are few legal protections from getting fired, you're not getting fired because you broke the law. You may lose your job, but you won't go to jail.
In the U.S. if you leak information to expose criminal action then you are generally protected from losing your job. However, if the information you need to leak is itself classified, then there are special channels set up to handle this without breaking classification. These safeguards may or may not work well, but it is the system that *congress* put in place, so they are ultimately responsible.
Again, what Gonzales said should not be controversial -- he is simply stating the current situation in U.S. code: leaking classified information is a crime, and knowingly passing that information on is also a crime, whether you pass this on to everybody (journalist) or to just a few (spy). Most countries (including Sweden) have laws similar to this.
The laws in question here are similar. Leaking non-classified information is not generally a crime. In this case, Gonzales (the attorney general of the U.S.) stated that leaking and publishing *classified* information is a crime.
He didn't make this up one day -- this is the current state of U.S. law. This generally has only been enforced against the leakers and not against the press that published it, but the law states that knowingly leaking or publishing classified information is a crime. The press has gotten used to having a free pass on this, and is getting upset now that Gonzales is talking about closing that gate, but the gate was put there by Congress.
XBox Live Arcade is available whether you have an XBL subscription or not. The XBL subscription allows you to purchase additional games over the Live service, but the free "Silver" subscription level allows you to buy arcade games, download trailers and demos and patches, and buy additional game content (mmmm, useless horse barding). You just can't do multiplayer -- that takes the $$$ "Gold" subscription.
You seem to be under the delusion that the basic X360 costs $420 (it's $300, even the non-crippled version is $400). Or that x360 games cost $100 a pop. Granted, the original poster's math was off a bit as well, unless he was assuming some of those games were bought used, which may be a reasonable assumption since it's been out awhile and there is a decent used game selection at your local gamestop.
I would be very surprised, but not as surprised as the English.
What is completely overlooked is that in the long run, such a system may have unintended/unforseen costs that make mockery of the current benefits. In addition, the hardening of the regulatory systems is amazingly costly for stifling innovation and other "inventions we can sell" features that helped create the boom economy of the previous system.
What we don't need is engineering sites spewing this kind of crap. Lots of engineers have lost their jobs due to the implementation of these systems, and the more criticism we can muster from the engineering press, the more iPods and Ferraris we can keep ourselves in.
S-Ox is disastrous. Please stop spewing this crap now.
I've had the DirecTV HD Tivo for nearly a year now after upgrading from an old series 1 unit. It can definitely record two HD signals at once, and it has a "ratio" button (upper left of the remote) for flipping through the various aspect ratios. It also has multiple options for viewing the guide, including a more normal format for the guide (although we stuck with the old TiVo guide format since we were already used to it).
The reason it doesn't mean "piss" to you is because you didn't grow up here. It's pretty exclusively a kid's word -- you'll never hear someone over the age of 6 or 7 use it unless they're talking to a child under the age or 6 or 7.
:-(
But as adults we still know what it means; it's just too bad Nintendo didn't
I think you are mistaken about who made it to 64 bits first. I had a 64-bit DEC Alpha with Win NT 3.51 back in the mid-90's, and it ran just fine thank you. More recent descendants of Win NT have supported the Intel Itanium systems. And now XP runs just fine on the 64-bit Xeons and Durons.
Windows did make the jump to 64-bit x64 later than linux, but they jumped pretty quick once the EM64T was announced; they don't call the two companies WinTel for nothing.
I don't know about that. From the article, it seems like you could make a decent wheel by putting the revolution controller core into a wheel.
I can imagine a wheel controller without the mounting base on it, with paddle shifters on the backside. You just hold it out in front of you, turn it, push out to accelerate, pull back to decelerate. No expensive sensors needed, it's all in the revolution controller module. You need something to connect the paddle shifters to the a/b buttons, but that's no biggie. I seem to remember the article also mentioned that the controller supported force feedback as well.
Fencing is fun to watch if you know what to watch for, but you're right that the action is very fast.
In some ways it's similar to american football that way, brief periods of very complex action punctuated by long periods where everything is reset.
What fencing needs is (a) visual highlighting of the blades, like televised hockey does with the puck or football does with the LOS and (b) none of this reset-after-a-point BS -- just keep going -- wasn't that the whole point of the electric foils, to eliminate the slow judging process?
Besides, blade highlighting would make them look like lightsabers. Foil fencing = boring, lightsaber dueling = ratings. Encourage the players to paint their mask screen like darth maul.
I know this was supposed to be one of the "microsoft questions" at one time, but this answer is just silly.
The reason round manhole covers don't fall down round manholes because they are larger than the hole, not because of their shape. Square manhole covers on square manholes work fine as long as the cover is large enough (side-of-cover longer than sqrt(2)*side-of-hole, more-or-less, since the thickness of the cover plays a part as well). Similarly for any other shape.
Round covers simply solve the problem with a low material consumption. I suspect an equilateral triangle works well too, but is simply too inefficient from a usability perspective.
You mean like this?
I got Doom III for the xbox. Spent a few hours playing it, then spent the next few days playing the xbox port of Doom I and II that came with it. Never have made it back to III.
But then, I still spend an inordinate amount of time in nethack and ctetris.
That's the going rate here in Houston. More like $4 or $5 if you catch a midday show during the weekend.
They're switching to Intel because they've been abandoned three times before. First with the 680x0, then with the Motorola PPC, and finally with IBM. Now they simply want a nice safe partner. Intel may not be exciting, but they'll be there for the long haul.
VPC is slower under the G5 than it was under the G4 -- the G4 can flip its byte ordering to match the x86.
I've got a 1ghz G4 ibook, and benchmarking some of my own code under VPC show it running at about 750 mhz equivalent. This is fast enough, btw, that apps compiled with the MSVC compiler can run faster under VPC than they do compiled natively with the GNU compiler, and before you flame about compile options, neither the MS nor GNU compilers were using any aggressive optimization options -- the point is that the CPU emulation (at least on the G4) is good enough that it falls into the noise category.
The IBM-sourced G5 in the new macs doesn't support the byte-order flag, and working around this causes a huge performance hit. What is still unknown is if the XBox360 version of the PPC supports the byte-order flag -- if they're serious about the emulation then it may well.
What *really* sucks about VPC are 1) the graphics emulation -- it currently emulates an old dumb S3 chipset, and 2) it takes a little bit before he's got a good working set loaded into his translated code cache, which means it can be sluggish for awhile. Hopefully VPC is getting much better graphics emulation capability for the next version.
The lisp machines had (have, for those of us that still own one) this property.
And the lisp machines had something similar as well. the Symbolics Ivory processor used 40-bit words, 32-bits of address/data, and 8 bits of type info.
One of the neat side-affects of this is that there was only one opcode for, say, "add". It looked at the data types and figured out wether to do a fp or integer add, and could trap on anything else -- the standard trap handler would detect mismatched types and perform the appropriate coercions. Similarly, the array-reference opcode knew (a) whether it was a char array or an int array or something else, and the size, and handle the offset scaling and range checking appropriately.
There was also a magic "forwarding" type value, that said in effect "the thing that used to be here is now over there", where "there" was an address in the data field of that value. The hardware would automatically chase the pointer to the new location to get the real data. This meant that you could safely move objects around in memory without having to patch all of the pointers. There was a low-priority OS thread that walked memory finding pointers to redirected memory and patching them up, and would reclaim the memory occupied by the old object once no pointers to it were left around.
This was used by the paging system to rearrange objects around memory so that objects in the working set were compacted into pages in the working set. This allowed the symbolics machines to have virtual:physical memory ratios > 100:1 with acceptable performance.
All this complicated hardware wasn't the snappiest thing in the world, but it was pretty cool.
It indexed the CPP files on my system just fine. It seems to be content based, not extension based (makes sense given that the web google does the same thing).
Not necessarily -- MOV's are extremely common in code generated for register machines -- data movement is simply a very common activity.
It also seems to have brought back the finder bug where finder would crash if you tried to drag more than one item (This had been fixed in the most recent OS update)