Re: [PATCH] virtual sched_clock() for s390

[Martin Schwidefsky]

On Mon, 2007-07-23 at 09:15 +0000, Jan Glauber wrote:
> > > As with s390, 64-bit PowerPC also uses CONFIG_VIRT_CPU_ACCOUNTING. 
> > > That affects how tsk->utime and tsk->stime are accumulated (we call 
> > > account_user_time and account_system_time directly rather than calling 
> > > update_process_times) as well as the system hardirq/softirq time, idle 
> > > time, and stolen time.
> > 
> > tsk->utime and tsk->stime is only used for a single purpose: to 
> > determine the 'split' factor of how to split up the precise total time 
> > between user and system time.

At least for s390 and powerpc the utime and stime already contain a very
precise value how much time was spent in the user and system context.
For s390 the granularity is a microsecond. The other values nice, idle,
iowait, irq, softirq and steal are precise as well.

> > > When you say "precise task statistics for /proc", where are they 
> > > accumulated?  I don't see any changes to the way that tsk->utime and 
> > > ctime are computed.
> > 
> > we now use p->se.sum_exec_runtime that measures (in nanoseconds) the 
> > precise amount of time spent executing (sum of system and user time) - 
> > and ->stime and ->utime is used to determine the 'split'. [this allows 
> > us to gather ->stime and ->utime via low-resolution sampling, while 
> > keeping the 'total' precise. Accounting at every system entry point 
> > would be quite expensive on most platforms.]

With the exact accounting of utime and stime that would mean that
p->se.sum_exec_runtime is utime + stime, no?
Precise Accounting at every cpu context switch has some cost, but for
s390 it is not as bad as it sounds. We do 2 store-cpu-timer (STPT)
instructions, 2 64 bit adds and 2 64 bit subtracts. In terms of cycles
it is less than 30 cycles on each system entry on the latest machine.

> Using se.sum_exec_runtime to generate ->utime and ->stime breaks
> the process accounting we have on s390 (and probably on PowerPC too).
> With CONFIG_VIRT_CPU_ACCOUNTING we already have precise values in
> ->utime and ->stime. Can we make the calculation of the CFS-based time
> values conditional by CONFIG_VIRT_CPU_ACCOUNTING?

Imho, we just have to update utime and stime when the process accounting
values are requested and set se.sum_exec_runtime to the sum of utime and
stime for CONFIG_VIRT_CPU_ACCOUNTING=y.

-- 
blue skies,
  Martin.

"Reality continues to ruin my life." - Calvin.



-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/