[Olsr-users] Broadcast Packets & Windows Routin

[Aaron Kaplan]

>
> It is implementation-dependent what happens to your TSC when the
> kernel puts the CPU into idle.  With many x86 chipsets, the TSC will
> stop increasing in C2.

ah ok. Well, I always used that to measure some basic blocks of  
execution.

>
> What is more, on modern Linux kernels gettimeofday will transparently
> use the TSC or the ACPI timer (whichever happens to work on your
> particular machine) to give you sub-tick precision.  On a 1.2 GHz
> Pentium-M and a recent Linux 2.6 kernel, I have measured gettimeofday
> as having a granularity of roughly 3 us.

That is new for me. Neat! :)


--
L. Aaron Kaplan
nic.at   //   csirt.at // cert.at
(spam-protected)
+43 1 505 6416 / 46