[prev in list] [next in list] [prev in thread] [next in thread] 

List:       gmp-discuss
Subject:    Tuning
From:       user42 () zip ! com ! au (Kevin Ryde)
Date:       2004-06-20 0:29:01
Message-ID: 87y8mji3ni.fsf () zip ! com ! au
[Download RAW message or body]

Josh Liu <zliu2@student.gsu.edu> writes:
>
> I want to know how the precision parameter comes into play as well.

Enough repetitions of a call are made to ensure 1 in prec accuracy,
having regard to the timebase unit.  For instance if it's microseconds
then the default precision is 1000 and enough repetitions are made to
take 1000 microseconds, ie. 1 millisecond.  See speed_measure in
common.c.

> The other possibility is to do curve fitting of the time function

It's not necessarily very smooth.


Torbjorn Granlund <tg at swox.com> writes:
>
> minimizing a "badness" metric.

Yep.  Comments in tuneup.c.

[prev in list] [next in list] [prev in thread] [next in thread]