Timing Facilities

Timing Facilities

[Steve Poole]

I have written a small assembly fragment to read the TB register that runs
under Linux (YDL/BLL). What is the definition of a "tick" or "jiffie" under
Linux. When some code is run under MacOS (400MHz G4) the results are
*faster* (the numbers returned are lower) than when I run on a 500MHz G4. I
appreciate any input.

Steve...
Steve Poole
IS2 / PeerTechnologies
Phone 713.806.0329
eFax   240.539.7413


** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

Re: Timing Facilities

[Joseph P. Garcia]

On Wed, 21 Feb 2001 23:24:57 -0600, Steve Poole <sp@is2-isnot.com> wrote:

>
> I have written a small assembly fragment to read the TB register that runs
> under Linux (YDL/BLL). What is the definition of a "tick" or "jiffie" under
> Linux. When some code is run under MacOS (400MHz G4) the results are
> *faster* (the numbers returned are lower) than when I run on a 500MHz G4. I
> appreciate any input.

I beleive there are 1*HZ jiffies in one second.  Used in the kernel's timer_list struct (to name one use), which basically has the kernel run a function at a specific time.  ex, setting the 'expires' value to (jiffies + HZ) will cause it to run one second from now.

--
Joseph P. Garcia
http://www.execpc.com/~jpgarcia

** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

Re: Timing Facilities

[Iain Sandoe]

Thu, Feb 22, 2001,Joseph P. Garcia wrote:
> On Wed, 21 Feb 2001 23:24:57 -0600, Steve Poole <sp@is2-isnot.com> wrote:
>> I have written a small assembly fragment to read the TB register that runs
>> under Linux (YDL/BLL). What is the definition of a "tick" or "jiffie" under
>> Linux. When some code is run under MacOS (400MHz G4) the results are
>> *faster* (the numbers returned are lower) than when I run on a 500MHz G4. I
>> appreciate any input.
>
> I beleive there are 1*HZ jiffies in one second.  Used in the kernel's
> timer_list struct (to name one use), which basically has the kernel run a
> function at a specific time.  ex, setting the 'expires' value to (jiffies +
> HZ) will cause it to run one second from now.

the VEA TB rate is not fixed to the processor clock - but is related to the
bus freq (IIRC).  You can find the TB rate from OF (but beware of a report
that says the values stored are not always correct).

I did a similar thing a while ago - and, because high precision was not
required, just used gettime() and a few seconds sleep... to calibrate.

ciao,
Iain.

** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/