On Tue, Apr 10, 2001 at 01:12:14PM +0100, Alan Cox wrote:
> Measure the number of clocks executing a timer interrupt. rdtsc is fast. Now
> consider the fact that out of this you get KHz or better scheduling
> resolution required for games and midi. I'd say it looks good. I agree
And measure the number of cycles a gigahertz CPU can do between a 1ms timer.
And then check how often the typical application executes something like
gettimeofday.
> the accounting of user/system time needs care to avoid slowing down syscall
> paths
It's also all interrupts, not only syscalls, and also context switch if you
want to be accurate.
On modern PC hardware it might be possible to do user/system accounting using
performance MSRs. They have a bit in the performance counter that allows to
only account user or system. If you find a count that is near equivalent to
the cycles you have both: total = rdtsc, user = msr, system = rdtsc-msr.
At least PPro derived have event 0x16, number of instructions executed, which
might be good enough when multiplied with a factor if your instruction mix is not
too unusual.
Still even with that the more complex checking in add_timer doesn't look too good.
-Andi
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
This archive was generated by hypermail 2b29 : Sun Apr 15 2001 - 21:00:13 EST