I didn't see people mentioning that there is a lot of optimization code for Pentium-type CPUs (how well it works, I can't attest to).
In my experience it's pretty touchy. Enabling -march=i686 on my machine (with some gcc 2.95.3 snapshot that I should probably upgrade) slows down some parts of my code while speeding up other parts.
Also, AFAIK, there isn't any real support for P-III, just PPro/P-II and I think K7 in the latest snapshots.
I suspect the main reason there is a lot of explicit code for the Pentium (besides it's popularity, obviously), is that it is the only CPU in common use today that isn't a RISC machine with a ton of registers. Optimizing for an Alpha is a lot like optimizing for a MIPS, which is a lot like optimizing for a SPARC, PowerPC, and so on (yes, there are often fairly significant differences, but nothing like the diff between any of those and x86).
Whatever the case the may be, the day gcc generates working 64-bit code I'll drink a few beers for the guys working on gcc. As it stands now, gcc can't generate a decent (maybe I should say working) 64-bit binary for both the SGI and SPARC platforms:( (I haven't tried it on an Alpha yet.)
I haven't tried 64-bit code with GCC on an SGI, but it works OK on a Sun [assuming you have 64-bit Solaris or Linux on it]. And *no* worries on the Alpha, I've been building code on those guys for a while and it works fine. It's not as fast as Compaq's compiler, but I guess that's not really a suprise.
Gcc is a nice tool; it's free, and it works well. Unfortunately, even with -O3 -funroll-loops, it can't optimize for beans. I had to study this in detail as a project for one of my grad courses, and I was appalled when I found out just how many potential optimizations it wouldn't catch.
So true. I've benchmarked crypto code [code that can take great advangage of pipelining and good register allocation] I've written with gcc and a few commerical compilers (all running on Linux on the same system), and in some cases I would see 3x-4x speed increase. And if you have gcc dump the asm, you'll see many silly things, even with full optimizations. This is totally from memory, but gives you an idea of what I'm talking about:
add esp,-4
[some instructions that just use registers, don't read or write to esp]
add esp,-8
I hand optimized the code (removing the second instruction and changing the first to add esp,-12), and it worked fine. This is of course a trivial example (yeah, I saved on cycle!), but in a large program things like this could mean tens of millions of cycles (think inner loops of long running programs).
If your algorithm is already in pipeline-friendly form, you'll generally be OK, but AFAIK, you're right about GCC not rearranging instructions to handle pipelining (but I haven't looked into this too carefully).
Of course I figure for heavy numerical work G4 will kick an x86's butt, just on the basis that a G4 has a reasonable number of registers. I'm amazed that Intel hasn't added a new extension like MMX or SSE that gives programs a few more GPRs; it would really be useful to a wide variety of programs.
Re:A sign of changing times
on
NSA Inside?
·
· Score: 2
I agree though that the NSA thinks a strong defense is a good offense. But think about this: how many of the really really important stuff do you think is on a web connected machine? How much is actually sent over the internet?
None. Zip. Zero. That's not the point. MAC is primarily used by military and intelligence places. MAC is not about preventing people from rooting your system (OpenBSD is probably equivalent to a B1 or B2 system as far as penetration resistance is concerned). MAC is about preventing user A, who knows secret S, from giving it to B, who is not authorized to know S. Computer security != crypto. Having a license-free, open-sourced, B2 rated OS would be a real win for the federal government as far as costs go.
Targets using 128-bit or higher encryption- how long would that take to crack on 10k clustered machines each with 64 Alpha chips?
A very, very long time. Here's an example: say you constructed a Dyson sphere around the sun, and used _all_ of the energy you collected to power a computer whose only purpose was to increment a 128-bit counter from 0 until 0xFFFF....FFFF. That would not finish before you die (this derives from thermodynamic arguments, not current or forecasted computing power). Actually running a cipher key schedule, decrypting the blocks, and figuring out if they're the plaintext will take much more time.
128 bit crypto is "good enough", because it will be much simpler to break your security in other ways. If I want your data bad enough, I'm not going to try to brute force even a 64 bit key; I'm going to come to your house and start breaking your fingers until your tell me what I want to know. Or I'll root your systems through conventional means and trojan your binaries. Or break into your office building and replace your custom hardware with stuff that is really weak. Or (if I'm the government), arrest you and give the options: a) tell us everything, or b) go to jail for a long time. All of these options are much more economical.
The analogy commonly used for this situation is putting a giant spike in the middle of your yard in the hopes that someone will run into it (this spike is called "crypto"). It doesn't matter if the spike is a mile long or 2 miles long, if someone can walk right around it and climb through your unlocked window.
P.S., I'm always fascinated how people hold a distaste for organized Christianity and a love for the Lord of the Rings and Siilmarillion simultaneously, given the unashamedly Catholic
imagery woven throughout those two books.
AFAIK, no one has ever killed hundreds of thousands of innocent people because they didn't like LotR. I have no particular problem with Christianity in general, it's assholes like these guys that I can't stand.
(I can't read the article right now, it's/.ed, so this may be totally wrong).
Hasn't this sort of thing already been known? Chaitan's Omega Number stuff has been known for some time (he's written a book about it, which I think is this one, but Fatbrain doesn't have descriptions so I could be wrong).
Goedel proved back in the 30's that there were many things (an infinite number?) which were true but for which proofs cannot be provided. OTOH Chaitin is a well known mathemetician (in some circles, anyway). Presumably he has something interesting to say, but I doubt it's as revolutionary as the post makes it sound.
They have 2000 now, which is *gasp* really good. Now, they might not be able to scale to 64 bits, but I don' think they'd be THAT dumb.
Maybe so. But I have yet to see the code. Where is W2K running on Alpha, MIPS, UltraSPARC, POWER, HP-PA, etc?
Linux has the advantage of being first, and having to simply recompile it's software, Microsoft the benefit of barrels of money.
I think Linux will win. Pretty much everyone who writes Unix software makes reasonably sure that the software will run on 64-bit platforms, big-endian machines, etc, because enough people run Unix on them that you'll find out fairly quickly if you code is broken on them.
Consider how few commercial apps are done for Linux on x86, even - the numbers just don't make it commercially viable (this is not my opinion, but what is obviously the opinion of a lot of software shops out there). W2K on (insert 64-bit RISC chip) is going to have far, far, far fewer users than Linux on x86. The only people who are going to even consider doings apps for is MS.
Also, I saw on the Register an article stating that any ports to IA-64 would be using 32-bit pointers and longs. Why? Because Windows programmers (this probably includes MS itself), have long assumed that they're running on a 32-bit little-endian CPU, and moving the code breaks. I can't find the article now, however, or I would give a link, sorry.
You may be thinking, "Only a real moron would assume a 32-bit CPU, etc". Story time:
Last month, Micrsoft gave a presentation on XP here. XP has support to "emulate" 98 or NT for applications that need it. Seems that some software (*cough* *Tomb Raider* *cough*), instead of accessing COM objects the normal, sane way (ie, using the documented access functions), simply looked into the object at a certain byte offset and expected the data to be there. I suspect this sort of thing happens a lot on Windows software.
Well, in any case, the emergence of 64-bit CPUs on the desktop should make things very interesting.
What are they using for a compiler, I wonder? The VLIW beast needs an optimizer with near-omniscience to get halfway decent performance.
A few years ago Intel bought KAI, a C++ compiler vendor (they do really good stuff, BTW). I supect the main reason for the purchase was so the people there would work on an IA-64 compiler for them.
Even with a good compiler, I think Merced will still probably suck. HP's implementation looks like it may actually be usaeble, however.
I actually had a rather lengthy argument with my computer sciences teacher about this -- it is impossible to generate a truely random number.
That depends entirely on how you define a "random" number. If you want to be a big philosopher and claim that nothing is "random", be my guest (BTW, Quatum Mechanics guarantees that there is real, true randomness in the world - presuming it's possible to sample that).
I (and most people working on crypto in academia [which I am not one of, but just pointing out there are plenty of informed people who think this]) adopt the more practial view that if there does not exist a polynomial time algorithm to decide if a given string is truly random or the output of the random number generator, it's "random". There are plenty of things that will do this just fine.
It is kind of like trying to prove something can't be done.
Huh? There are tons of things you can prove cannot be done. For example, I can prove that you cannot possibly find a integer n such that 2*n+1 is evenly divisible by two. Could you please explain what you meant by this?
In fact, there are things such that you can provide proofs for both of the following:
1) It is not possible to prove that X exists.
2) It is not possible to prove that X does not exist.
An example is an existence of a set S such that |N|<|S|<|R| (where N is the natural numbers and R is the real numbers).
Yeah, that confused me. And the fact that the Societas Eruditorium was poised to actually play a role in the plot, but didn't.
It's my understanding that Cryptonomicon is the first part of a series of books. Makes sense, when you consider the vast amount of material which was vaugely referenced but never used (who were the black and Indian guys with Rudy and Otto on the boat (he never even gives their names!)? What's up with that secret society? What was Root doing working for the NSA during the 50s? If he faked his death, why? If not, how is he still alive? Not to mention the data haven; plenty of interesting things could happen there).
You know, I didn't get ths part at all. Did he fake his own death? Was I reading this part too late at night and I missed something? Mod me off-topic, just answer me:)
I dunno. That secret society Root was in seemed to know a thing or two about healing (remember when Amy was about to lose her leg from an infection, and Root did something which she refused to discuss but healed her). And we know Rudy is a member of the same society.
And remember that Shaftoe witnessed the death. He was a fscking marine, he should probably be able to tell when someone is dead (or close to it).
My theory is that Root was not-quite-but-almost dead, the society has some technique or technology (nano-tech in the 40's?) that can heal people, and that's what brought him back. Guess we'll find out.
Well, IANAA (I Am Not an Astronomer), but basically, instead of detecting photons like a normal telescope, it detects radio waves. That makes it fun for studying things like stars, galaxies, quasars, and other things that emit lots of radio energy.
we were all running computers that needed two-phase 208v and air conditioning.
This is slashdot. Most of us probably are.;)
Seriously, I think you're right. Personally, I seem way to interested in Transmeta considering how much I'm looking forward to Sledgehammer, and actually don't like laptops at all (ie, I will almost certainly never buy a Transmeta chip).
The authors leave out the hordes of lesser known Unicies. I'm sure the graph would be completely unreadable if any of these were included.
I also didn't see A/UX (Apples UNIX - don't ask me, I've just saw once that Perl had been ported to it). Not to mention all the early variants: I heard that there was a JHUNIX, a variant created at, as you would expect, JHU. The main Unix box (an SGI) is still called jhunix, but that is all that's left.:(
SGI, on the other hand, has the right idea. By giving up IRIX and supporting Linux development
Maybe in the long term, IRIX is on it's way out, but I think SGI will be doing lots of good stuff with it for at least a few more years, until Linux is ready to handle the big SGI boxen that are currently running IRIX.
And for projects SGI is helping with, don't forget the STL: the GCC STL implementation is based heavily on SGI code (read the copyrights in the headers).
Do you know how sore your vocal chords would be after dictating all day?
Not to mention that fact that it would IRRITATE THE HELL out of everybody within hearing distance. I mean, COME ON! I would hate to work in a cube, surrounded by a couple dozen people, all talking to their computers (well, we talk to our computers here, but mostly to swear at them). The clicking of a keyboard is pretty easy to ignore, because it's not particularly interesting to listen to, just a bunch of clicks.
Also, if I had to talk to post all my slashdot comments, I'd probably be fired because people would finally realize how little time I spend working.:P
Promise has an ATA-RAID controller than can bind up to four IDE disks together for about $100. Use something like this, and you could cut your load time down by half or better.
Slashdot Mirror
I think he also had something to cycle the graveyard back if needed.
Tormod's Crypt?
I didn't see people mentioning that there is a lot of optimization code for Pentium-type CPUs (how well it works, I can't attest to).
In my experience it's pretty touchy. Enabling -march=i686 on my machine (with some gcc 2.95.3 snapshot that I should probably upgrade) slows down some parts of my code while speeding up other parts.
Also, AFAIK, there isn't any real support for P-III, just PPro/P-II and I think K7 in the latest snapshots.
I suspect the main reason there is a lot of explicit code for the Pentium (besides it's popularity, obviously), is that it is the only CPU in common use today that isn't a RISC machine with a ton of registers. Optimizing for an Alpha is a lot like optimizing for a MIPS, which is a lot like optimizing for a SPARC, PowerPC, and so on (yes, there are often fairly significant differences, but nothing like the diff between any of those and x86).
Whatever the case the may be, the day gcc generates working 64-bit code I'll drink a few beers for the guys working on gcc. As it stands now, gcc can't generate a decent (maybe I should say working) 64-bit binary for both the SGI and SPARC platforms :( (I haven't tried it on an Alpha yet.)
I haven't tried 64-bit code with GCC on an SGI, but it works OK on a Sun [assuming you have 64-bit Solaris or Linux on it]. And *no* worries on the Alpha, I've been building code on those guys for a while and it works fine. It's not as fast as Compaq's compiler, but I guess that's not really a suprise.
Gcc is a nice tool; it's free, and it works well. Unfortunately, even with -O3 -funroll-loops, it can't optimize for beans. I had to study this in detail as a project for one of my grad courses, and I was appalled when I found out just how many potential optimizations it wouldn't catch.
So true. I've benchmarked crypto code [code that can take great advangage of pipelining and good register allocation] I've written with gcc and a few commerical compilers (all running on Linux on the same system), and in some cases I would see 3x-4x speed increase. And if you have gcc dump the asm, you'll see many silly things, even with full optimizations. This is totally from memory, but gives you an idea of what I'm talking about:
add esp,-4
[some instructions that just use registers, don't read or write to esp]
add esp,-8
I hand optimized the code (removing the second instruction and changing the first to add esp,-12), and it worked fine. This is of course a trivial example (yeah, I saved on cycle!), but in a large program things like this could mean tens of millions of cycles (think inner loops of long running programs).
If your algorithm is already in pipeline-friendly form, you'll generally be OK, but AFAIK, you're right about GCC not rearranging instructions to handle pipelining (but I haven't looked into this too carefully).
Of course I figure for heavy numerical work G4 will kick an x86's butt, just on the basis that a G4 has a reasonable number of registers. I'm amazed that Intel hasn't added a new extension like MMX or SSE that gives programs a few more GPRs; it would really be useful to a wide variety of programs.
I agree though that the NSA thinks a strong defense is a good offense. But think about this: how many of the really really important stuff do you think is on a web connected machine? How much is actually sent over the internet?
None. Zip. Zero. That's not the point. MAC is primarily used by military and intelligence places. MAC is not about preventing people from rooting your system (OpenBSD is probably equivalent to a B1 or B2 system as far as penetration resistance is concerned). MAC is about preventing user A, who knows secret S, from giving it to B, who is not authorized to know S. Computer security != crypto. Having a license-free, open-sourced, B2 rated OS would be a real win for the federal government as far as costs go.
Targets using 128-bit or higher encryption- how long would that take to crack on 10k clustered machines each with 64 Alpha chips?
A very, very long time. Here's an example: say you constructed a Dyson sphere around the sun, and used _all_ of the energy you collected to power a computer whose only purpose was to increment a 128-bit counter from 0 until 0xFFFF....FFFF. That would not finish before you die (this derives from thermodynamic arguments, not current or forecasted computing power). Actually running a cipher key schedule, decrypting the blocks, and figuring out if they're the plaintext will take much more time.
128 bit crypto is "good enough", because it will be much simpler to break your security in other ways. If I want your data bad enough, I'm not going to try to brute force even a 64 bit key; I'm going to come to your house and start breaking your fingers until your tell me what I want to know. Or I'll root your systems through conventional means and trojan your binaries. Or break into your office building and replace your custom hardware with stuff that is really weak. Or (if I'm the government), arrest you and give the options: a) tell us everything, or b) go to jail for a long time. All of these options are much more economical.
The analogy commonly used for this situation is putting a giant spike in the middle of your yard in the hopes that someone will run into it (this spike is called "crypto"). It doesn't matter if the spike is a mile long or 2 miles long, if someone can walk right around it and climb through your unlocked window.
Tripwire is now a pay for play product, so I suggest using something like this which is open source/free and just as good
IIRC Tripwire is GPL now. But in any case I prefer AIDE myself.
P.S., I'm always fascinated how people hold a distaste for organized Christianity and a love for the Lord of the Rings and Siilmarillion simultaneously, given the unashamedly Catholic
imagery woven throughout those two books.
AFAIK, no one has ever killed hundreds of thousands of innocent people because they didn't like LotR. I have no particular problem with Christianity in general, it's assholes like these guys that I can't stand.
(I can't read the article right now, it's /.ed, so this may be totally wrong).
Hasn't this sort of thing already been known? Chaitan's Omega Number stuff has been known for some time (he's written a book about it, which I think is this one, but Fatbrain doesn't have descriptions so I could be wrong).
Goedel proved back in the 30's that there were many things (an infinite number?) which were true but for which proofs cannot be provided. OTOH Chaitin is a well known mathemetician (in some circles, anyway). Presumably he has something interesting to say, but I doubt it's as revolutionary as the post makes it sound.
They have 2000 now, which is *gasp* really good. Now, they might not be able to scale to 64 bits, but I don' think they'd be THAT dumb.
Maybe so. But I have yet to see the code. Where is W2K running on Alpha, MIPS, UltraSPARC, POWER, HP-PA, etc?
Linux has the advantage of being first, and having to simply recompile it's software, Microsoft the benefit of barrels of money.
I think Linux will win. Pretty much everyone who writes Unix software makes reasonably sure that the software will run on 64-bit platforms, big-endian machines, etc, because enough people run Unix on them that you'll find out fairly quickly if you code is broken on them.
Consider how few commercial apps are done for Linux on x86, even - the numbers just don't make it commercially viable (this is not my opinion, but what is obviously the opinion of a lot of software shops out there). W2K on (insert 64-bit RISC chip) is going to have far, far, far fewer users than Linux on x86. The only people who are going to even consider doings apps for is MS.
Also, I saw on the Register an article stating that any ports to IA-64 would be using 32-bit pointers and longs. Why? Because Windows programmers (this probably includes MS itself), have long assumed that they're running on a 32-bit little-endian CPU, and moving the code breaks. I can't find the article now, however, or I would give a link, sorry.
You may be thinking, "Only a real moron would assume a 32-bit CPU, etc". Story time:
Last month, Micrsoft gave a presentation on XP here. XP has support to "emulate" 98 or NT for applications that need it. Seems that some software (*cough* *Tomb Raider* *cough*), instead of accessing COM objects the normal, sane way (ie, using the documented access functions), simply looked into the object at a certain byte offset and expected the data to be there. I suspect this sort of thing happens a lot on Windows software.
Well, in any case, the emergence of 64-bit CPUs on the desktop should make things very interesting.
What are they using for a compiler, I wonder? The VLIW beast needs an optimizer with near-omniscience to get halfway decent performance.
A few years ago Intel bought KAI, a C++ compiler vendor (they do really good stuff, BTW). I supect the main reason for the purchase was so the people there would work on an IA-64 compiler for them.
Even with a good compiler, I think Merced will still probably suck. HP's implementation looks like it may actually be usaeble, however.
I actually had a rather lengthy argument with my computer sciences teacher about this -- it is impossible to generate a truely random number.
That depends entirely on how you define a "random" number. If you want to be a big philosopher and claim that nothing is "random", be my guest (BTW, Quatum Mechanics guarantees that there is real, true randomness in the world - presuming it's possible to sample that).
I (and most people working on crypto in academia [which I am not one of, but just pointing out there are plenty of informed people who think this]) adopt the more practial view that if there does not exist a polynomial time algorithm to decide if a given string is truly random or the output of the random number generator, it's "random". There are plenty of things that will do this just fine.
It is kind of like trying to prove something can't be done.
Huh? There are tons of things you can prove cannot be done. For example, I can prove that you cannot possibly find a integer n such that 2*n+1 is evenly divisible by two. Could you please explain what you meant by this?
In fact, there are things such that you can provide proofs for both of the following:
1) It is not possible to prove that X exists.
2) It is not possible to prove that X does not exist.
An example is an existence of a set S such that |N|<|S|<|R| (where N is the natural numbers and R is the real numbers).
Give me the Sci-Fi Channel, 24/7.
Cartoon Network!
Yeah, that confused me. And the fact that the Societas Eruditorium was poised to actually play a role in the plot, but didn't.
It's my understanding that Cryptonomicon is the first part of a series of books. Makes sense, when you consider the vast amount of material which was vaugely referenced but never used (who were the black and Indian guys with Rudy and Otto on the boat (he never even gives their names!)? What's up with that secret society? What was Root doing working for the NSA during the 50s? If he faked his death, why? If not, how is he still alive? Not to mention the data haven; plenty of interesting things could happen there).
I dunno. That secret society Root was in seemed to know a thing or two about healing (remember when Amy was about to lose her leg from an infection, and Root did something which she refused to discuss but healed her). And we know Rudy is a member of the same society.
And remember that Shaftoe witnessed the death. He was a fscking marine, he should probably be able to tell when someone is dead (or close to it).
My theory is that Root was not-quite-but-almost dead, the society has some technique or technology (nano-tech in the 40's?) that can heal people, and that's what brought him back. Guess we'll find out.
ppl tend to forget about temp files which can contain an unencrypted copy.
[...]
he only simple solution i've found is an app called evidence eliminator
I think W2K has encrypted filesystems built in, so as long as you remember to encrypt your swapfile and temp dirs, you're OK there.
OpenBSD (and soon Linux) has support for encrypted swap and of course most Unices have encrypted filesystems.
Sadly, consumer windows doesn't have any of that (not to mention anyone can come up and tamper with your binaries).
Not always - Unix system programming is always in style (unless you think Unix is already obsolete, of course).
Avoid those courses that centre on specific technologies. They'll be obsolete by the time you graduate.
I'd buy it just for Metroid. Hell, I bought N64 just for MarioKart and Zelda.
I'd do the same. I bough my N64 just for Zelda, but now that I've played Mario Kart I'm glad that's on there too.
Now all we really need is Shadowrun, god I miss that game.
What exactly is a 'radio telescope' anyway?
Well, IANAA (I Am Not an Astronomer), but basically, instead of detecting photons like a normal telescope, it detects radio waves. That makes it fun for studying things like stars, galaxies, quasars, and other things that emit lots of radio energy.
we were all running computers that needed two-phase 208v and air conditioning.
;)
This is slashdot. Most of us probably are.
Seriously, I think you're right. Personally, I seem way to interested in Transmeta considering how much I'm looking forward to Sledgehammer, and actually don't like laptops at all (ie, I will almost certainly never buy a Transmeta chip).
There are people who believe the earth is hollow.
Well, DUH! Where else would the Advanced Supersonic Alluminum Nazi Hell Creatures from Hollow Earth live?
But it could have been filled in seconds before you looked and hollowed out again as soon as you turn your back.
Yeah, shit like that happens to me a lot for some reason. HOLY GOD WHAT IS THAT THING... oh, it's gone now... nevermind.
No bash? Sheesh, I guess the judges knew I would really clean up if they let me in there. ;) [It's a joke, laugh]
The authors leave out the hordes of lesser known Unicies. I'm sure the graph would be completely unreadable if any of these were included.
:(
I also didn't see A/UX (Apples UNIX - don't ask me, I've just saw once that Perl had been ported to it). Not to mention all the early variants: I heard that there was a JHUNIX, a variant created at, as you would expect, JHU. The main Unix box (an SGI) is still called jhunix, but that is all that's left.
SGI, on the other hand, has the right idea. By giving up IRIX and supporting Linux development
Maybe in the long term, IRIX is on it's way out, but I think SGI will be doing lots of good stuff with it for at least a few more years, until Linux is ready to handle the big SGI boxen that are currently running IRIX.
And for projects SGI is helping with, don't forget the STL: the GCC STL implementation is based heavily on SGI code (read the copyrights in the headers).
Do you know how sore your vocal chords would be after dictating all day?
:P
Not to mention that fact that it would IRRITATE THE HELL out of everybody within hearing distance. I mean, COME ON! I would hate to work in a cube, surrounded by a couple dozen people, all talking to their computers (well, we talk to our computers here, but mostly to swear at them). The clicking of a keyboard is pretty easy to ignore, because it's not particularly interesting to listen to, just a bunch of clicks.
Also, if I had to talk to post all my slashdot comments, I'd probably be fired because people would finally realize how little time I spend working.
Promise has an ATA-RAID controller than can bind up to four IDE disks together for about $100. Use something like this, and you could cut your load time down by half or better.
Or get a KT7 mobo.