Subject: Re: [boost] [chrono] steady_clock efficiency
From: Kenneth Porter (shiva.blacklist_at_[hidden])
Date: 2011-11-30 00:10:24
Kenneth Porter <shiva.blacklist_at_[hidden]> wrote in
> So about 650 nanoseconds using a 32-bit build on Vista 64 Ultimate.
I recoded now() to eliminate the math with no improvement. The Windows API
call totally dominates the time spent, so there's no significant savings to
deferring the conversion.
The lower numbers on XP may be due to it being single-core, and not needing
to deal with which core is reading the counters. My Vista system is dual-
core and I understand there's logic in the API to worry about that.
Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk