Re: [boost] [log] filtering performance

From: Andrey Semashev (andrey.semashev_at_[hidden])
Date: 2008-09-01 11:25:25

Igor Nazarenko wrote:
> It seems that the library does quite a bit of work before filtering
> away a record, unless the logging is turned off entirely. There's at
> least two mutex locks, some memory allocation, etc. Callgrind claims
> ~2500 instructions when the record is thrown away by the global
> severity filter. Shortened callgrind output is in-line below; full
> output and the test program attached. Compiled by gcc 4.2.3 using
> boost.build variant=release threading=multi.

Although I didn't extensively profile the code, I believe that is quite
possible.