[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Questions

From: Leslie P. Polzer
Subject: Re: Questions
Date: Tue, 13 Mar 2007 14:01:06 +0100 (CET)
User-agent: SquirrelMail/1.4.4

> Certainly I'm willing to entertain the idea of changing this.
> Optimisers tend to work best (andmost intuitively) when they have am
> idea of the world that matches reality.  For that reason I wanted
> est_success_rate to model the actual success rate of those predicates
> (on average).  Some of the tests have very low values, for example
> -inum is 1.0e-6.
> (While I think about it, the value for -quit is very wrong; fixed.)
I see.  But I believe we could stay with integers in normal operation
(since we need only number_of_predicates numbers to reorder the tests)
and use floats only for performance measurement.

  Would that match your plan?

Or perhaps you could tell me more about the architecture of the
optimizer you had in mind.  Should it be some adaptive system?


reply via email to

[Prev in Thread] Current Thread [Next in Thread]