While multiple cores are very useful, does anyone know if there are plans to scale up SIMD processing capability? Some operations don't require multiple instruction streams, instead they apply the same operation to a set of data. The chip logic to implement SIMD is much simpler than MIMD to it would make sense to add bigger SIMD engines to the processor cores, how about a 64 x 64 SIMD processor, this would allow matrix operations to be processed in chunks of 4,096.
Even if an application is single-threaded the Kernel isn't. If you are running several applications concurrently the total thread count on the machine is probably close to 100 because of the O.S. and services. While it's true that some of those threads may be tightly coupled and not able to be scheduled together, some won't and these will benefit from parallel processing. In the longer term when the O.S. has 16 or more processors it will make more sense to explcitly write data parallel code.
"Still... real-time data connectivity while in the air will greatly increase my productivity while flying."
Really? You mean you don't have anything equally important that you could work on that doesn't require real-time connectivity? I find that very hard to believe. Perhaps you're just one of those people that constantly checks email and IM instead of doing real work (the stuff you could do offline). How many technology jobs *require* constant real-time connections? I'm sure they exist, but I can't think of many.
That was my point! The analogy would only be correct if the turbo was stolen. Chaning a code to illegally upgrade the O.S. is stealing, it is not a customization. Suppose I download a 30 day eval version of Photoshop and patch the date checking code with a jmp, would Mr. Anon consider it a customization?
*stealing* a turbo charger and *stealing* a car are both *stealing*.
And a word to the wise, be careful with the name-calling, one day someone might shove that keyboard you're hiding behind down your throat. Now run back under your Mommie's skirt before I spank you.
Can someone explain the difference between hacking XP Home to turn it into XP Pro and downloading a warez copy? Is one slightly less wrong than the other?
I've been doing this for years. I leave "Guns and Ammo" and ballistics tables on my desk, when my boss asks me to work late I tell him, no, because that would make me late for my anger management couselling, which would then cut into my time at the gun range.
I disagree. Working out 2 weeks notice is a courtesy, not a requirement. Why should this guy stick around to be abused by his boss in front of colleagues? If I was him I would clear out my desk late in the evening and call HR early the next morning telling them I had decided to leave early because of my boss's hostility.
When someone is behaving irrationally the best thing to do is get out. His boss is already talking crap about the guy, so what's the difference? If everyone regards him as a nut then why would they listen to his stories?
I didn't assume your IQ was low, I merely pointed out that your sample code wasn't illustrative because it didn't show how control was passed to the stack code. I'm genuinely sorry if I hurt your pride. You then went off on a rant telling me that JMP ESP is an illegal instruction and that it would never appear.
Word to the wise, when someone tries to teach you something, do a little research before you tell them they're wrong. Now that the penny has dropped and you understand what I've been trying to tell you, you now have the gaul to tell me that your inability to comprehend is my fault. In any case, I'm glad you now understand how a buffer overflow exploit works. Now I suggest you go and write some non malicous code to demonstrate your knowledge. You see there's a difference between thinking you know something and actually knowing it.
Did it occur to you that the sequence JMP ESP could be there by pure accident? It's just two bytes, perhaps it could be there at an irregular offset of a compiler generated assembly sequence, or perhaps it could be there as part of an address in memory. Get it now?
"I know full well that I what I need to do is to put the machine code for the 'malicious' code in the actual buffer and to keep overflowing with nop instructions for the proper number of bytes so the final 4 bytes I copy into that overflowed buffer on any intel 32bit processor will align with the return address from that function and if it is set correctly by taking into account the proper stack trace, can be known beforehand and thus when the function returns, it resumes execution of code at the start of the buffer that you overflowed."
That's incorrect. Overwriting the returns address will not resume execution in your buffer because you don't know beforehand the value of the stack pointer at the point at which the function is called, so you do not know the address in memory of your buffer. The solution is to find code in memory that contains a JMP ESP instruction, fill in the address of this instruction and the function will return to JMP ESP which then passes control to your buffer code.
Try searching for a JMP ESP in the kernel, last time I tried I couldn't find on. That's because Microsoft have patched all occurences. I haven't tried this for a while though.
All you've done is overflow the stack, you could have actually achieved the same effect with one line of code. In any case, you've missed the point. The hard part is not overlflowing the stack, it is changing the instruction pointer to the address of the malicious code passed in the buffer. The prefered way to do this is do find a jmp csp in the kernel and overwrite the return address with its location. It will then pass control to the code at the stack pointer- the code in your buffer.
No, hyperthreading only replicates the CPU Architectural State, which is the registers but not the processing pipeline. With hyper-threading the context switch between threads is speeded up, but no parallel processing occurs. In a dual-core chip instructions are executed in parallel by the cores.
For some applications hyper-threading slows down processing, and that is why Dell ships HT machines with HT turned off.
(Proof: Suppose you have a compression algorithm that always shortens a number, and the corresponding decryption function. bla bla bla bla..Mr. Math gives us a long-winded proof and saves us from his oh so complicated wizardry- Math Overload; mathematicians should be able to fill in the blanks fairly easily
If the compression algorithm always shortens the number, applying it recursively with yield a 0 digit number. It's self evident that no decompression algorithm can decompress a zero length input to anything but a zero length output. QED motherfucker.
The point of using the tab character is it does not represent a fixed indent. If you like 2 char indents, set your tabs to 2; if you want 8, set it to 8.
Using hard coded spaces consumes more bytes and requires reformatting to change the indent. Use of Tabs is a no-brainer, but judging from the comments here and elsewhere people still don't understand the issue.
You clearly know nothing about file systems. Anyone can slap a bunch of functionality on an existing file system, but the result will be dog slow and no one will use it. If the file system is a dog, the O.S. will be a dog. What Microsoft is trying to do essentially is get a free lunch- add function to the file system while keeping it at least as fast as NTFS.
If (ptr==NULL) is poor style because it could lead to a = being deleted and overlooked, much safer to write if (NULL==ptr). if (!ptr) is equivalent in standard C++, however, both will compile to the same object even without optimization on most compilers.
I have installed a home network in our home (2700 square foot, 7 rooms). Here's the bottom line- wireless is unreliable and slow, you'll need plenty of repeaters to make it semi reliable. After trying different base stations, repeaters and cards I gave up and pulled cable. It was a hassle but I now have an extremely reliable gigabit LAN. Audio is streamed from a rackmount system in the garage to Audiotrons in the various rooms and video is streamed to a PC with an xcard in the living room. The video streaming was a semi-custom solution, I can play ripped DVD's in full quality mpreg2 over the network. Watching a DVD involves a menu click using the xcard remote control, the DVD starts playing almost instantaneously.
Wireless is great for certain applications, but if you own your home I would go with a gigabit copper LAN. It's cheap and it works even when your neighbor is microwaving a burrito.
Slashdot Mirror
While multiple cores are very useful, does anyone know if there are plans to scale up SIMD processing capability? Some operations don't require multiple instruction streams, instead they apply the same operation to a set of data. The chip logic to implement SIMD is much simpler than MIMD to it would make sense to add bigger SIMD engines to the processor cores, how about a 64 x 64 SIMD processor, this would allow matrix operations to be processed in chunks of 4,096.
Even if an application is single-threaded the Kernel isn't. If you are running several applications concurrently the total thread count on the machine is probably close to 100 because of the O.S. and services. While it's true that some of those threads may be tightly coupled and not able to be scheduled together, some won't and these will benefit from parallel processing. In the longer term when the O.S. has 16 or more processors it will make more sense to explcitly write data parallel code.
Really? You mean you don't have anything equally important that you could work on that doesn't require real-time connectivity? I find that very hard to believe. Perhaps you're just one of those people that constantly checks email and IM instead of doing real work (the stuff you could do offline). How many technology jobs *require* constant real-time connections? I'm sure they exist, but I can't think of many.
Why has it taken so long?
That was my point! The analogy would only be correct if the turbo was stolen. Chaning a code to illegally upgrade the O.S. is stealing, it is not a customization. Suppose I download a 30 day eval version of Photoshop and patch the date checking code with a jmp, would Mr. Anon consider it a customization?
*stealing* a turbo charger and *stealing* a car are both *stealing*.
And a word to the wise, be careful with the name-calling, one day someone might shove that keyboard you're hiding behind down your throat. Now run back under your Mommie's skirt before I spank you.
Can someone explain the difference between hacking XP Home to turn it into XP Pro and downloading a warez copy? Is one slightly less wrong than the other?
Alternatively swing by Best Buy and steal a copy.
I've been doing this for years. I leave "Guns and Ammo" and ballistics tables on my desk, when my boss asks me to work late I tell him, no, because that would make me late for my anger management couselling, which would then cut into my time at the gun range.
I disagree. Working out 2 weeks notice is a courtesy, not a requirement. Why should this guy stick around to be abused by his boss in front of colleagues? If I was him I would clear out my desk late in the evening and call HR early the next morning telling them I had decided to leave early because of my boss's hostility.
When someone is behaving irrationally the best thing to do is get out. His boss is already talking crap about the guy, so what's the difference? If everyone regards him as a nut then why would they listen to his stories?
I didn't assume your IQ was low, I merely pointed out that your sample code wasn't illustrative because it didn't show how control was passed to the stack code. I'm genuinely sorry if I hurt your pride. You then went off on a rant telling me that JMP ESP is an illegal instruction and that it would never appear.
Word to the wise, when someone tries to teach you something, do a little research before you tell them they're wrong. Now that the penny has dropped and you understand what I've been trying to tell you, you now have the gaul to tell me that your inability to comprehend is my fault. In any case, I'm glad you now understand how a buffer overflow exploit works. Now I suggest you go and write some non malicous code to demonstrate your knowledge. You see there's a difference between thinking you know something and actually knowing it.
Best regards,
Beatlebum
this is getting really tiresome.
Did it occur to you that the sequence JMP ESP could be there by pure accident? It's just two bytes, perhaps it could be there at an irregular offset of a compiler generated assembly sequence, or perhaps it could be there as part of an address in memory. Get it now?
You're diggging yourself deeper and deeper. JMP ESP is a perfectly valid instruction. Try a search a see the bugtraq hits:
h l=en&lr=&safe=off&rls=GGLD,GGLD:2004-41,GGLD:en&sa =N&tab=wg
http://groups-beta.google.com/groups?q=JMP%20ESP&
Actually you don't get it. You said
"I know full well that I what I need to do is to put the machine code for the 'malicious' code in the actual buffer and to keep overflowing with nop instructions for the proper number of bytes so the final 4 bytes I copy into that overflowed buffer on any intel 32bit processor will align with the return address from that function and if it is set correctly by taking into account the proper stack trace, can be known beforehand and thus when the function returns, it resumes execution of code at the start of the buffer that you overflowed."
That's incorrect. Overwriting the returns address will not resume execution in your buffer because you don't know beforehand the value of the stack pointer at the point at which the function is called, so you do not know the address in memory of your buffer. The solution is to find code in memory that contains a JMP ESP instruction, fill in the address of this instruction and the function will return to JMP ESP which then passes control to your buffer code.
Try searching for a JMP ESP in the kernel, last time I tried I couldn't find on. That's because Microsoft have patched all occurences. I haven't tried this for a while though.
All you've done is overflow the stack, you could have actually achieved the same effect with one line of code. In any case, you've missed the point. The hard part is not overlflowing the stack, it is changing the instruction pointer to the address of the malicious code passed in the buffer. The prefered way to do this is do find a jmp csp in the kernel and overwrite the return address with its location. It will then pass control to the code at the stack pointer- the code in your buffer.
Actually you don't seem to get it either. The cache is shared on a dual code. This is a good thing (tm).
No, hyperthreading only replicates the CPU Architectural State, which is the registers but not the processing pipeline. With hyper-threading the context switch between threads is speeded up, but no parallel processing occurs. In a dual-core chip instructions are executed in parallel by the cores.
For some applications hyper-threading slows down processing, and that is why Dell ships HT machines with HT turned off.
Right click on toobar, and uncheck "PDFMaker 7.0".
What's the big deal?
(Proof: Suppose you have a compression algorithm that always shortens a number, and the corresponding decryption function. bla bla bla bla ..Mr. Math gives us a long-winded proof and saves us from his oh so complicated wizardry- Math Overload; mathematicians should be able to fill in the blanks fairly easily
If the compression algorithm always shortens the number, applying it recursively with yield a 0 digit number. It's self evident that no decompression algorithm can decompress a zero length input to anything but a zero length output. QED motherfucker.
The point of using the tab character is it does not represent a fixed indent. If you like 2 char indents, set your tabs to 2; if you want 8, set it to 8.
Using hard coded spaces consumes more bytes and requires reformatting to change the indent. Use of Tabs is a no-brainer, but judging from the comments here and elsewhere people still don't understand the issue.
You clearly know nothing about file systems. Anyone can slap a bunch of functionality on an existing file system, but the result will be dog slow and no one will use it. If the file system is a dog, the O.S. will be a dog. What Microsoft is trying to do essentially is get a free lunch- add function to the file system while keeping it at least as fast as NTFS.
If (ptr==NULL) is poor style because it could lead to a = being deleted and overlooked, much safer to write if (NULL==ptr). if (!ptr) is equivalent in standard C++, however, both will compile to the same object even without optimization on most compilers.
I bet there are code comments inside CreateWindow saying feature X was added to assist Word/Excel/Powerpoint.
>> For example, if a truck suddenly pulls out in front of you, you will suddenly focus on it; your passenger will tend to notice this and stop talking
You're obviously not married.
I have installed a home network in our home (2700 square foot, 7 rooms). Here's the bottom line- wireless is unreliable and slow, you'll need plenty of repeaters to make it semi reliable. After trying different base stations, repeaters and cards I gave up and pulled cable. It was a hassle but I now have an extremely reliable gigabit LAN. Audio is streamed from a rackmount system in the garage to Audiotrons in the various rooms and video is streamed to a PC with an xcard in the living room. The video streaming was a semi-custom solution, I can play ripped DVD's in full quality mpreg2 over the network. Watching a DVD involves a menu click using the xcard remote control, the DVD starts playing almost instantaneously.
Wireless is great for certain applications, but if you own your home I would go with a gigabit copper LAN. It's cheap and it works even when your neighbor is microwaving a burrito.
IBM tech support is fantastic, so much so that I just purchased a 3 year extended warranty on my Thinkpad. I guess that's $100 down the drain.