Prev: [PATCHi v3] kconfig qconf: port to QT4
Next: posix-cpu-timers: Introduction of wall clocks
From: Stanislaw Gruszka on 30 Jul 2010 07:30
Hi Tomasz

On Fri, 30 Jul 2010 11:57:43 +0200
Tomasz Buchert <tomasz.buchert(a)inria.fr> wrote:
> During our work we encountered a problem of obtaining
> a reliable and fine-grained measurements of CPU time/wall time
> of a process/thread. The existing methods (taskstats, procfs,
> POSIX CPU clocks) have either unfriendly interface (taskstats)

Well, "unfriendly" is not technical word ... Taskstat was designed
as extensible method to provide statistics to user space, including
some special needs like yours. What your patches basically do, is get
the value of wall_time(p) = uptime - p->start_time . I do not see
any reason why this calculation can not be done in userspace, getting
uptime and p->start_time from kernel by existing methods (or adding
new one for p->start_time if precise method does not really exist).
Is enough to read p->start_time only once at the beginning, this value
does not change.

Anyway, I think we need better rationale, before we can consider your wall
clock patches inclusion into the kernel.

BTW: If you will repost please cc Thomas Gleixner as he is kernel/*timers*
maintainer.

Cheers
Stanislaw
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo(a)vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
From: Stanislaw Gruszka on 30 Jul 2010 08:00
On Fri, 30 Jul 2010 13:23:43 +0200
Stanislaw Gruszka <sgruszka(a)redhat.com> wrote:

> Anyway, I think we need better rationale, before we can consider your wall
> clock patches inclusion into the kernel.

To clarify. I'm not against your patches. I just would like to know
why patches are needed (or maybe even how they are used) to see if there
is no other/better way to solve your problem. We do not want add new
code to kernel (and what more important - maintain it), if problem can
be solved differently.

Stanislaw
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo(a)vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
 | 
Pages: 1
Prev: [PATCHi v3] kconfig qconf: port to QT4
Next: posix-cpu-timers: Introduction of wall clocks