Subject: Re: changing traffic counters to quad_t (long long)
To: Jonathan Stone <jonathan@DSG.Stanford.EDU>
From: Perry E. Metzger <firstname.lastname@example.org>
Date: 03/24/1998 13:42:44
Jonathan Stone writes:
> >It has been observed in PR#1515 and PR#3732 that our 32 bit traffic
> >counters overflow a bit quickly on a busy, stable server host.
> >Why would it be bad thing to change the various counters to quad_t?
> IIANM, last time this came up, people with old, slow machines (e.g.,
> 68020s, vaxes) were not enthralled with the extra cost of quad_t adds.
Given how much cost there is in just copying the packet around during
processing, the extra three instructions needed will not be noticed by
If someone really, really cares, we could always make these things an
MD type and put the definition in some md include file.