On Sat, Oct 20, 2001 at 12:20:17AM -0400, Christopher D. Carothers wrote:
> Gary -- use the "rdtsc" instruction which stands for read
> timestamp counter. This instruction will return the number of cycles the
> processor has been up since it was last rebooted. It is 64 bits so the
> processor will turn to slag before it overflows.
>
> Below is the inline assembly language that will compile under gcc.
>
> __inline__ unsigned long long int rdtsc()
> {
> unsigned long long int x;
> __asm__ volatile (".byte 0x0f, 0x31" : "=A" (x));
> return x;
> }
>
> Note, that to get the seconds you need to divide by the speed of
> the processor. Inside the kernel you can get the speed directly by using
> the "cpu_khz" variable and multiple by 1000.0.
>
> Final word, on SMP boxes each processor will have it own counter
> register and because of clock drift (and slightly offset boot times) these
> clocks will vary -- however, my experience is that for high-end/high-speed
> systems this amount is not that noticeable -- but they will differ...
>
> Hope that helps,
> Chris
Fabulous. I'm using it to find out exactly how long a given RT task
takes by adding these bits to the schedulers.
But I also have to ask. There are other counters, there, that I've
read about. There are 67 things that they can count, and they can do
it at no cost to the system.
eg, cache hits&misses, cycles since counter was started, etc, etc.
I'm also interested in these as a more general thing. The number of
ticks is all well and good, but I'd also like to know other stuff.
Thank-you very much,
Gary (-;
PS It's getting added later today (=
-- [rtl] ---
To unsubscribe:
echo "unsubscribe rtl" | mail [EMAIL PROTECTED] OR
echo "unsubscribe rtl <Your_email>" | mail [EMAIL PROTECTED]
--
For more information on Real-Time Linux see:
http://www.rtlinux.org/
