Re: GEQO: ERX - Mailing list pgsql-hackers

From Tobias Zahn
Subject Re: GEQO: ERX
Date
Msg-id 49FC6A92.7000302@arcor.de
Whole thread Raw
In response to GEQO: ERX  (Tobias Zahn <tobias-zahn@arcor.de>)
Responses Re: GEQO: ERX  (Tom Lane <tgl@sss.pgh.pa.us>)
List pgsql-hackers
Hello again,
I didn't not get any response to my initial message below. Now I am
wondering if nobody is into the optimizer or if my question was just too
stupid. Could you please give me some clues? Your help would really be
appreciated.

Regards,
Tobias

> Hello,
> I was digging through the optimizer code and have a question regarding
> the edge recombination crossover (ERX) of the GEQO.  It might be
> completely stupid and therefore I apologize for this in advance.
> 
> As far as I understand it, the idea of the ERX is the minimization of
> edge failures. When reading in geqo_main.c and geqo_erx.c, it seams that
> every iterative round (generation) it is checked by gimme_tour() if
> there where any edge failures. When I understand the algorithm right,
> there should be no edge failures.
> Therefore I think about NOT checking for edge failures anymore, to save
> some time. (If it just to make sure, it could be done only once in the
> end.) Might that work or do I have some errors in my thoughts?
> 
> Thanks in advance,
> Tobias
> 



pgsql-hackers by date:

Previous
From: Nikhil Sontakke
Date:
Subject: Re: Throw some low-level C scutwork at me
Next
From: Andrew Dunstan
Date:
Subject: Re: windows shared memory error