Re: DD correctly simulated by HH --- never stops running without aborting its simulation

Liste des GroupesRevenir à theory 
Sujet : Re: DD correctly simulated by HH --- never stops running without aborting its simulation
De : richard (at) *nospam* damon-family.org (Richard Damon)
Groupes : comp.theory sci.logic
Date : 07. Jun 2024, 17:14:18
Autres entêtes
Organisation : i2pn2 (i2pn.org)
Message-ID : <v3v84a$39ri5$9@i2pn2.org>
References : 1 2 3
User-Agent : Mozilla Thunderbird
On 6/7/24 9:21 AM, olcott wrote:
On 6/7/2024 3:00 AM, Fred. Zwarts wrote:
Op 06.jun.2024 om 20:35 schreef olcott:
<MIT Professor Sipser agreed to ONLY these verbatim words 10/13/2022>
   If simulating halt decider H correctly simulates its input D
   until H correctly determines that its simulated D would never
   stop running unless aborted then
>
   H can abort its simulation of D and correctly report that D
   specifies a non-halting sequence of configurations.
</MIT Professor Sipser agreed to ONLY these verbatim words10/13/2022>
>
*Try to show how this DD correctly simulated by any HH ever*
*stops running without having its simulation aborted by HH*
>
_DD()
[00001e12] 55         push ebp
[00001e13] 8bec       mov  ebp,esp
[00001e15] 51         push ecx
[00001e16] 8b4508     mov  eax,[ebp+08]
[00001e19] 50         push eax      ; push DD
[00001e1a] 8b4d08     mov  ecx,[ebp+08]
[00001e1d] 51         push ecx      ; push DD
[00001e1e] e85ff5ffff call 00001382 ; call HH
>
>
Olcott has proven himself that HH reports false negatives, with the following example:
>
        typedef int (*ptr)();  // ptr is pointer to int function in C
>
        int H(ptr p, ptr i);
>
        int main()
        {
          return H(main, 0);
        }
>
He proved that main halts and that H reports non-halting.
>
 main() does meet the Sipser approved criteria and you cannot
possibly show otherwise.
     If simulating halt decider HH correctly simulates its input main
    until HH correctly determines that its simulated main would never
    stop running unless aborted then
Thus showing that H considers itself to be non-halting here, as main is just a proxy function to give H its parameters.
Also, why is the code for main so long below? it seems to be calling more than just H.
I suspect you aren't showing the code for the program above.

 _main()
[00001e42] 55         push ebp
[00001e43] 8bec       mov ebp,esp
[00001e45] 6a00       push +00
[00001e47] 68421e0000 push 00001e42
[00001e4c] e831f5ffff call 00001382
[00001e51] 83c408     add esp,+08
[00001e54] 50         push eax
[00001e55] 6843070000 push 00000743
[00001e5a] e843e9ffff call 000007a2
[00001e5f] 83c408     add esp,+08
[00001e62] eb79       jmp 00001edd
[00001e64] 6a05       push +05
[00001e66] 68321d0000 push 00001d32
[00001e6b] e8f2f6ffff call 00001562
[00001e70] 83c408     add esp,+08
[00001e73] 50         push eax
[00001e74] 6853070000 push 00000753
[00001e79] e824e9ffff call 000007a2
[00001e7e] 83c408     add esp,+08
[00001e81] 68621d0000 push 00001d62
[00001e86] e8d7f9ffff call 00001862
[00001e8b] 83c404     add esp,+04
[00001e8e] 50         push eax
[00001e8f] 6863070000 push 00000763
[00001e94] e809e9ffff call 000007a2
[00001e99] 83c408     add esp,+08
[00001e9c] 6a05       push +05
[00001e9e] 68f21c0000 push 00001cf2
[00001ea3] e8baf6ffff call 00001562
[00001ea8] 83c408     add esp,+08
[00001eab] 50         push eax
[00001eac] 6873070000 push 00000773
[00001eb1] e8ece8ffff call 000007a2
[00001eb6] 83c408     add esp,+08
[00001eb9] 68e21d0000 push 00001de2
[00001ebe] 68e21d0000 push 00001de2
[00001ec3] e89af6ffff call 00001562
[00001ec8] 83c408     add esp,+08
[00001ecb] 50         push eax
[00001ecc] 6883070000 push 00000783
[00001ed1] e8cce8ffff call 000007a2
[00001ed6] 83c408     add esp,+08
[00001ed9] 33c0       xor eax,eax
[00001edb] eb02       jmp 00001edf
[00001edd] 33c0       xor eax,eax
[00001edf] 5d         pop ebp
[00001ee0] c3         ret
Size in bytes:(0159) [00001ee0]
   machine   stack     stack     machine    assembly
  address   address   data      code       language
  ========  ========  ========  =========  =============
[00001e42][00103375][00000000] 55         push ebp
[00001e43][00103375][00000000] 8bec       mov ebp,esp
[00001e45][00103371][00000000] 6a00       push +00
[00001e47][0010336d][00001e42] 68421e0000 push 00001e42 ; push main
[00001e4c][00103369][00001e51] e831f5ffff call 00001382 ; call HH
New slave_stack at:103419
 Begin Local Halt Decider Simulation   Execution Trace Stored at:113421
[00001e42][0011340d][00113411] 55         push ebp
[00001e43][0011340d][00113411] 8bec       mov ebp,esp
[00001e45][00113409][00000000] 6a00       push +00
[00001e47][00113405][00001e42] 68421e0000 push 00001e42 ; push main
[00001e4c][00113401][00001e51] e831f5ffff call 00001382 ; call HH
New slave_stack at:14de41
And what does the above have to do with a correct instruction by instruction simulation of the input to HH? we should see the instructions that follow the eecution path of the call HH instruction.
I think you just don't understand how computers work.

[00001e42][0015de35][0015de39] 55         push ebp
[00001e43][0015de35][0015de39] 8bec       mov ebp,esp
[00001e45][0015de31][00000000] 6a00       push +00
[00001e47][0015de2d][00001e42] 68421e0000 push 00001e42 ; push main
[00001e4c][0015de29][00001e51] e831f5ffff call 00001382 ; call HH
Local Halt Decider: Infinite Recursion Detected Simulation Stopped
 [00001e51][00103375][00000000] 83c408     add esp,+08
[00001e54][00103371][00000000] 50         push eax
[00001e55][0010336d][00000743] 6843070000 push 00000743
[00001e5a][0010336d][00000743] e843e9ffff call 000007a2
Input_Halts = 0
[00001e5f][00103375][00000000] 83c408     add esp,+08
[00001e62][00103375][00000000] eb79       jmp 00001edd
[00001edd][00103375][00000000] 33c0       xor eax,eax
[00001edf][00103379][00000018] 5d         pop ebp
[00001ee0][0010337d][00000000] c3         ret
Number of Instructions Executed(11357) == 170 Pages
 
So, it should be no surprise that he finds other false negatives, like DD here.
>
Already earlier he claimed that his decider does not decide not about reality (the direct execution), but about its own decision process (the simulation). Another reason to expect false negatives. This makes his decider very uninteresting. Who wants to know whether a decider decides that its own process is correct? This is trivially correct. Just like the sentence "This sentence is true.": it might be true, but it is not informative.
 

Date Sujet#  Auteur
10 Nov 24 o 

Haut de la page

Les messages affichés proviennent d'usenet.

NewsPortal