Charles Randall wrote:
>What type of performance improvements did you see with this (under what
>workload)?
>
I don't have measured numbers yet, but I have projected numbers based
on the profile data that prompted the patch:
add_any_filter was about 0.15% of the non-syscall CPU time, and we were
averaging 4.3 string-compares per call (which makes sense, with a typical
filter list length of 8). Assuming 1 string-compare per call for the hash
table version, I'm estimating that the patch will reduce the httpd's overall
usr-mode CPU consumption by 0.12%.
--Brian
