Date: Sun, 19 Nov 2000 07:32:04 +1100 (EST) From: Bruce Evans <bde@zeta.org.au> To: Mike Smith <msmith@FreeBSD.ORG> Cc: Daniel Eischen <eischen@vigrid.com>, Mark Murray <mark@grondar.za>, arch@FreeBSD.ORG Subject: Re: "Monotonic" counter/register call - commit candidate. Message-ID: <Pine.BSF.4.21.0011190721390.604-100000@besplex.bde.org> In-Reply-To: <200011181356.eAIDuHF04876@mass.osd.bsdi.com>
next in thread | previous in thread | raw e-mail | index | archive | help
On Sat, 18 Nov 2000, Mike Smith wrote: > > > I considered renaming the function to marks(9), as I kinda liked > > > that name, but on reflection, it seems way too arrogant, so > > > Mike Smith's "get_jiffiecounter" won. > > > > I really hate "jiffie" and would prefer using just get_cyclecount > > or even better get_counter. > > "get_counter" is hopelessly vague. get_cyclecount would be OK too. I > don't care. Pick a name and stop bloody arguing about it. How about rdcdtsc() (read cpu-dependent timestamp counter)? :-). > > It would also be nice to know what > > resolution the counter was, perhaps get_counter_res(). Do we > > want an ID associated with the counter if there is more than > > one available on the hardware? > > Stop trying to make it into a timecounter. If you want a timecounter, > use a timecounter. We have perfectly good timecounters already. This is > not and never will be a facility any good for computing time. The value > may wrap unexpectedly, go backwards, proceed forwards erratically, etc. I hesitate to mention that we already have an imperfectly good function for access to certain machine-dependent counters: cputime(). It is only implemented on i386's and only used for profiling. It is almost as slow as a timecounter (not all that slow). Bruce To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-arch" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?Pine.BSF.4.21.0011190721390.604-100000>