Re: Problem with optimizations

["Albrecht Käfer" <albrecht_kaefer@yahoo.de>]

Olivier Scalbert schrieb::
> Hello,
>
> I have write a little program that try to solve Rubik's cube, just to
> practice a little Ada.
> The program is really stupid (brute force) but I have a strange problem.
> When I compile it with:
> gnatmake -O2 rubukmain, it works:
>
> Depth:  1  Positions Counter:                    12
> Depth:  2  Positions Counter:                   132
> Depth:  3  Positions Counter:                  1440
> Depth:  4  Positions Counter:                 15720
> Depth:  5  Positions Counter:                171600
> Depth:  6  Positions Counter:               1873200
> Depth:  7  Positions Counter:              20448000
> Depth:  8
> Solved !
>            1         12          8         10          2          6
>       3          3  Positions Counter:              18104705
>
> When I compile it with:
> gnatmake -O3 rubukmain, it does not work:
>
> Depth:  1  Positions Counter:                     1
> Then it stops.
>
> When, inside the Recursive_Find_Solution, I replace Is_Solved by
> Is_Solved1, then it works in all cases.
>
> I have try to understand what is happen with gdb and by having a look
> into the generated code, but without success.
>
> I have the same behaviour with:
> - Ubuntu 8.10 32 bits (GNATMAKE 4.3.2)
> - Debian testing AMD 64bits (GNATMAKE 4.3.3)

Works fine on Windows (GNAT GPL 2009).
However, I can't help but notice that you are doing premature
optimization. Did you *test* if it is faster?


Albrecht