Sudoku solver

[BartC]

On 29/03/2015 22:19, Mark Lawrence wrote:
> On 29/03/2015 21:59, BartC wrote:
>> On 29/03/2015 00:12, Chris Angelico wrote:
>>> On Sun, Mar 29, 2015 at 10:50 AM, BartC <bc at freeuk.com> wrote:
>>>> Using the OP's algorithm, and testing with the 'hard' puzzle posted
>>>> by Ian
>>>> Kelly, I got these approximate results:
>>>>
>>>> Python 3.1:     1700 seconds     (normal Python interpreter)
>>>> PyPy:           93 seconds
>>>> C unoptimised:  17 seconds       (gcc -O0 32-bit)
>>>> C optimised:    3.3 seconds      (gcc -O3 32-bit)
>>>> (X:             170 seconds)
>>>
>>> Nice stats. Any chance you can add CPython 3.4 or 3.5 to that? That's
>>> a pretty old CPython you're using.
>>
>> I've tried 3.4.3 and it's nearer 1900 seconds!
>>
>> Which wasn't too surprising as you don't expect new releases to be
>> faster, they tend to be slower.

> I simply do not believe those figures, that's roughly 12% slower.  If
> that happened in the real world you'd be able to hear the screams of
> anguish around the world.

You're right, it's wasn't 12% slower. It was 16%!

I didn't have time to run this very long benchmark so ran a different 
algorithm using 3.1, averaging 14.1 seconds for 3 runs. And ran the same 
code with 3.4.3, average 16.4 seconds.

Maybe most people don't run intensively computational benchmarks like these.

-- 
Bartc