Re: Avoiding microtime(9) global state

Martin Husemann <[email protected]> Thu, 1 Jul 2004 16:04:48 +0200
Newsgroups gmane.os.netbsd.devel.smp
Message-ID <[email protected]>
On Wed, Jun 30, 2004 at 06:09:05PM -0700, Jason Thorpe wrote:
> What is the point of having microtime() or any microtime()-like call if 
> the call's value is meaningless?  That's what you're proposing.

No, what I'm proposing is a GetTickCount()^Wnanotime() function
that is very similar to the current cc_microtime implementation,
but that

 - has nanosecond granularity
 - does not promise strong montonicity
 - is not affected by adjtime

It's value would roughly be "time since boot"; the implementation could
probably avoid any locking. I suppose most callers of microtime() could
be easily converted (the scheduler, call wheel, timeout calculations
for ISDN connections :-}).

Martin