From: Aaron W. LaFramboise (aaronrabiddog51_at_[hidden])
Date: 2004-05-18 06:35:56
Johan Nilsson wrote:
>[sort-of-a-shameless-plug but on-topic]:
>Questions are welcome.
Thank you for the excellent and informative article.
This is definitely the sort of thing I had in mind, and you seem to have
explored it a lot further than I did.
However, encorporating this functionality into Boost date_time may be,
in some ways, more difficult than it was in time_provider, as Boost
probably can't do things like creating a background thread to keep the
time synchronized, and has some additional portability constraints. I'm
also concerned by your remarks--as well as other information--that
suggest that the performance counter may be unsuitable as a general time
Without having a synchronization monitor thread, it seems like it might
be a lot more difficult to attempt to synchronize the time, especially
with regard to intermittant "spikes" caused by certain hardware
activity. Perhaps it might be necessary to develop a much more
substancial clock framework such as your time_provider template for
Boost in order to accomidate the demands of getting accurate time on
operating systems such as Windows. While it may add size, complexity,
and weight to the library, it might be worth it; what good is a timer
library that can't accurately tell time?
I'm going to spend the next few days analyzing your code, in particular,
Aaron W. LaFramboise
Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk