2

In a x86/amd64 CPU could I find an instruction or a combination of instructions that leak inside the system? Could some other physics properties (not accounted by the design) be measured or interacted with?

dddvvv
  • 23
  • 1
    I don't understand this question. What do you mean by "leak inside the system"? And, anyway, this seems more like an (computer) engineering question to me. – ACuriousMind May 31 '15 at 18:35
  • I am partially inspired by this article https://docs.google.com/document/d/1ez4qVKucyZN_-if5UIvMpfjdqeaFbg8RnBJOA8pOA_0/edit Leak inside the system refers to the CPU as an embedded system and the outside world as the container system – dddvvv May 31 '15 at 18:38
  • 1
    What does untamed physics mean? If the circuit "wires" in the cpu are too close together, say less than 5nm I think, then you might get quantum tunneling, that's as near to untamed physics as I can imagine. –  May 31 '15 at 18:52
  • 1
    By untamed I mean an instruction or combination of instructions that produce some effect that was not planned in the design of the CPU. – dddvvv May 31 '15 at 18:55
  • 1
    No offence intended, but that does not seem remotely like a physics based question, imo. –  May 31 '15 at 18:58
  • 1
    Is not an answer but as an example RdRand http://spectrum.ieee.org/computing/hardware/behind-intels-new-randomnumber-generator It uses thermal noise – dddvvv May 31 '15 at 20:34
  • Ok, from that article I kinda see your point, but could I suggest you pick out the physics part, because there's a lot of pure math in that article, and then include it in a reworded question, but I would not, imo, use computer terminology in the title. Best of luck with it tho. –  May 31 '15 at 21:15

2 Answers2

1

As with any universal Turing machine, there is no general way to predict the behavior of most software. That includes the behavior of your "CPU + memory" system. One specific behavior that you cannot predict is if a given input will result in an output or if the machine will keep computing, or running, forever. Of course, you can predict that for simple software examples but not in general (see halting problem).

I am not sure what other kinds of specific behaviors you have in mind, but I'll be glad to try to answer it if you can make your question more specific.

1

yes, it is possible but unlikely. Manufacturers do the best to deliver hardwares matching their specifications , using as many errors corrections algorithms that needed for this purpose.

Some people look for expected quantum signals at the i/o devices level. For example, a physicist claims to use an ordinary camera to get quantum random numbers, while others try to count the occurences of some crc corrections. I am not convinced that is what is expected.