Re: [PATCH] run_method performance improvement

[Randy Terbush <ra...@zyzzyva.com>]

Dean,

As long as you are breaking things tomorrow, you might also 
consider adding Doug's patch that adds SFIO and DCE support. I 
would like to play with that as well. 

Marc?

Have you come up with a benchmark setup that you are happy with? It 
would be nice to start testing some of these "performance" changes 
to see what is being accomplished.

Re: [PATCH] run_method performance improvement

[Dean Gaudet <dg...@arctic.org>]

Will do.  I've got another patch too for consideration, just cleaning it
up.

The run_method change I have profiled, and I've also profiled the upcoming
change (to make_dirstr and its callers).  Both are wins.  These I profiled
using httpd -X and gprof, running a 10k hit sample through the server from
www.arctic.org.  But I did that work about 4 months ago. 

The big http_main patch is only a theoretical improvement at this point. 
It's a hard one to profile because one of the wins in it is the removal of
scoreboard scanning in the children, and I haven't been happy with any
profiles I've ever got of the server + children.

Dean

On Sun, 29 Jun 1997, Randy Terbush wrote:

> Dean,
> 
> As long as you are breaking things tomorrow, you might also 
> consider adding Doug's patch that adds SFIO and DCE support. I 
> would like to play with that as well. 
> 
> Marc?
> 
> Have you come up with a benchmark setup that you are happy with? It 
> would be nice to start testing some of these "performance" changes 
> to see what is being accomplished.
> 
> 
> 
> 
>