[time-nuts] Time stamping with a PICPET

[Hal Murray]

albertson.chris at gmail.com said:
> I'm pretty certain the outer that the PPS interrupt handler reads in on the
> CPU chip.   The counter counts nanoseconds but the lower digits are always
> zero.  I don't think a PCI bus is involved 

On most CPU architectures, the low level hardware has a register that counts 
on the CPU clock.

On Intel systems, it's read by the rdtsc instruction which reads a 48 bit 
counter: TSC, time stamp counter.
  http://en.wikipedia.org/wiki/Time_Stamp_Counter

On DEC/Alpha it was rpcc, 32 bits.  (read processor cycle counter)  MIPS had 
a similar counter, but it wasn't available from user space.  I'm not familiar 
with other architectures.

There was a tangle when Intel started changing the CPU clock frequency on the 
fly to save power.  I don't know how they fixed that.


POSIX/Linux/*BSD has 2 calls to read the (software) clock.  gettimeofday 
returns seconds and microseconds in a struct timeval.  clock_gettime returns 
seconds and nanoseconds in a struct timespec.  (nanoseconds fit in a 32 bit 
word)


-- 
These are my opinions.  I hate spam.