Messages in this thread | | | Date | Tue, 25 Aug 1998 14:00:54 -0400 | From | Roger Espel Llima <> | Subject | counters that wrap around |
| |
Mitchell Blank Jr wrote: > I think using 64-bit counters on i386 is unreasonable. These counters > are often incremented in extremely performance-critical areas. Using > a polled system where the kernel remembers when 32-bit counters overflowed > might be workable.
Is it that unreasonable to do something like (forgive the Intel asm syntax..)
add variable, 1 adc variable64, 0
instead of
inc variable
like we're (I assume) doing now?
Sure, it's a few cycles more, but it's not a horribly complicated test with conditional jumps... the performance guys should know if the timing difference would be lost in the noise, or if it'd be actually noticeable.
If it'd be lost in the noise, it's probably worth keeping these things, at least for uptime and network/vm/etc stats.
otherwise, your polled system sounds like a good way to do it without affecting the critical paths at all.
-- Roger Espel Llima, espel@llaic.u-clermont1.fr -o) http://www.eleves.ens.fr:8080/home/espel/index.html /\\ _\_v
- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu Please read the FAQ at http://www.altern.org/andrebalsa/doc/lkml-faq.html
| |