[Date Prev][Date Next] [Chronological] [Thread] [Top]

RE: An optimization for another day (and developer)



> -----Original Message-----
> From: Pierangelo Masarati [mailto:masarati@aero.polimi.it]

> > Well, I guess my last commit counts as a radical change to the API.
> > Anything that lets you avoid a malloc is a Good Thing. In this
> case, I've
> > changed the filter/syntax/indexer structures to embed the struct berval
> > in their parent structure, instead of using a struct berval pointer.
> > This cuts malloc calls almost in half, with a corresponding drop in
> > CPU time. I will be looking thru the code for other places to make
> > similar changes. (The Attribute->a_vals is a good candidate for gains,
> > but I left it out this time 'round because that was too much to change
> > at once.)

> I see very little commits on back-tcl, back-sql and so; I guess at some
> point all the API changes will need to be done at once :(.  I'm not
> usually compiling those backends, so I can have a look at them but
> I'm not sure I can catch everything.  Is anybody out there with more
> specific knowledge of the missing backends that wants to take care of
> them?
>
> Ando.

Hah, good question. I think back-tcl has been pretty much unsupported since
late 1999. With regard to these changes, I've been grep'ing the whole source
tree to see what needs adapting. Besides back-ldbm and back-bdb, very little
of the backend code was affected.

  -- Howard Chu
  Chief Architect, Symas Corp.       Director, Highland Sun
  http://www.symas.com               http://highlandsun.com/hyc
  Symas: Premier OpenSource Development and Support