----------------------------------------
> Date: Thu, 31 Jul 2008 14:47:02 -0700
> From: [EMAIL PROTECTED]
> To: [email protected]
> Subject: Re: Any company ever publish software that mimics their CPUs 
> *exactly*?
> 
> On Thu, Jul 31, 2008 at 02:50:48PM -0500, Gabriel Sechan wrote:
> 
>>> Every ARM emulator accepts 32 bit instructions.  32 bits is "only" 4 
>>> billion.
>>> I just need to compare the state of RAM, registers and CPU before and after 
>>> a
>>> 32 bit instruction for a measly 4 billion of them!!!  Maybe it would finish
>>> overnight and so I could wake up with proof in hand!
>>> 
>>> Of course I'd have to do this for various initial states of CPU, RAM and
>>> registers since hardware isn't stateless like HTML.
>>
>>Sure.  Now lets say we have 10 registers of 32 bits.  Since we don't
>>know how they may effect state we have to try all combinations, that's
>>a mere 2^320 states per instruction, for a mere 2^352 total tests.  Let
>>me know when they finish, and we can start testing ram configurations
>>:)
> 
> Why not make it even worse (and still realistic).  The CPU has a
> pipeline, so the instructions don't execute independently.  With modern
> non-strongly ordered memory, you have to check possibly large sequences
> of instructions to make sure they do the same things with memory
> interactions.
> 
> I think, generally once a "solution" requires more compute steps than
> the size/age of the known universe it is safe to consider it
> "impractical".
> 
Bah, we just need computers that can go back in time and start calculating form 
the beginning of the known universe.  I think I read a sci-fi book about that 
once although they only could go back to their creation date, and hence got 
smarter over time. Yeah, I'm basically posting this due to the sci-fi reference.

Gabe 

--
[email protected]
http://www.kernel-panic.org/cgi-bin/mailman/listinfo/kplug-list

Reply via email to