[Gluster-devel] Macros and small files (was Re: Gluster and GCC 5.1)

[Peter Portante]

Hi Jeff,

(Disclaimer, switched to gmail, responding via gmail web interface,
hopefully this formats well.

On Mon, Jul 6, 2015 at 8:34 AM, Jeff Darcy <jdarcy at redhat.com> wrote:

> > And in the past, if not now, are contributing factors to small file
> > performance issues.
>
> I'm not quite seeing the connection here.  Which macros are you
> thinking of, and how does the fact that they're macros instead of
> functions make them bad for small-file performance?  AFAIK the
> problem with some of the macros in storage/posix is the redundant
> xattr calls etc. that some of them make.  Is the idea here that
> with inline code (not inline functions) each instance could be
> more closely tuned to a particular need and avoid the redundant
> calls?
>

>From my past analysis of a small file workload strace,
https://gist.github.com/portante/5029518, I saw that the macros hid
gratuitous behaviors.  It seems that we need to be careful that macros are
used knowing what they are doing.  Digging through the strace (annotated
from code that I am sure has since changed) it was quite amazing to see
what happens for one operation.

-peter
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.gluster.org/pipermail/gluster-devel/attachments/20150706/0dae68d8/attachment.html>