quentron.com

[sarkeizen]

*Yawn* Seemingly you're getting off on dragging your heels, or cowering or whatever you're doing  If you're lacking attention or something at home I promise to give you as much as I was giving you before but for the sake of actually giving your ego it's much needed trim.  Please answer the question:


So in other words a CPU without any timers, privileged mode or access to external circuitry except memory and the ability to load the program for analysis and the ability to tell you if the program will halt.  Can't solve the halting problem right?  If not, then list the self-induced EXTERNAL events (which is pathetically poorly defined btw) which are possible in this configuration.

[lumen]

*Yawn* Seemingly you're getting off on dragging your heels, or cowering or whatever you're doing  If you're lacking attention or something at home I promise to give you as much as I was giving you before but for the sake of actually giving your ego it's much needed trim.  Please answer the question:


So in other words a CPU without any timers, privileged mode or access to external circuitry except memory and the ability to load the program for analysis and the ability to tell you if the program will halt.  Can't solve the halting problem right?  If not, then list the self-induced EXTERNAL events (which is pathetically poorly defined btw) which are possible in this configuration.

Now, you admit that your theory has limited boundaries and you need me to fall into the area that works for you?

Ok, Ill say I can do it on any CPU as long as I have enough memory for my control program.

[DreamThinkBuild]

This halting theory sounds like bad programming practice to begin with. If there is no program there is no halting theory. If your going to have this problem with a program that does this why even bother writing it in the first place?

If your working with unknown data sets take a different approach. Like genetic algorithm/neural networks to approximate a best match formula from your test data then work backwards from there. See if it matches with reality.

[sarkeizen]

Now, you admit that your theory has limited boundaries and you need me to fall into the area that works for you?

Dude, you are the one who said that the halting problem was only an issue on old machines - not me - seriously get yourself checked out.  Similarly YOU said it was obsolete - generally in English that implies  that it at one time served a useful purpose anyway the point of all this discussion over hardware was simply positioning you for the kill.

Ok, Ill say I can do it on any CPU as long as I have enough memory for my control program.

Excellent.  Then it can work on a machine with no privileged mode right?  So your control program is essentially an interpreter right?   I look forward to kicking your teeth in (metaphorically).

This halting theory sounds like bad programming practice to begin with. If there is no program there is no halting theory. If your going to have this problem with a program that does this why even bother writing it in the first place?

Agreed, if a program is impossible to write - you shouldn't write it   This conversation between lumen and myself is that he doesn't yet realize that the program can not be written (and that is part of a larger argument in which lumen believes that the only way you can say if something will not work is if you understand it's operation to some level of detail.)

The purpose of computer science, or part of it anyway is to figure out boundary conditions like these.

If your working with unknown data sets take a different approach. Like genetic algorithm/neural networks to approximate a best match formula from your test data then work backwards from there. See if it matches with reality.

You almost have it.  If for some reason you really needed to determine if a program would terminate or not.  You need to change, not your algorithm (all algorithms which satisfy our original requirements for determinism will fail) but your requirements.  For example, you could say "I'd like to know if a program of 2 instructions or less will terminate".   Clearly such a program, on modern machines can be written.

[lumen]

I was having the same thought as DTB, in that such a program could allow code to run but never hang the CPU and still run up to any point where the problem would exist in the code.

This could be a basic start for random or trial by error learning on AI machines.

Hmm.. In that case I think it can be done! I have an idea to do such a problem.

(but don't tell sarkeizen or he'll flip out)