Re: [PATCH v2 3/3] X86: Add a thread cpu time implementation to vDSO

[David Ahern]

On 1/2/15 10:17 AM, Shaohua Li wrote:
> On Fri, Jan 02, 2015 at 10:09:10AM -0700, David Ahern wrote:
> > On 1/2/15 10:02 AM, Shaohua Li wrote:
> > > On Fri, Jan 02, 2015 at 08:31:33AM -0700, David Ahern wrote:
> > > > On 1/1/15 7:59 PM, Shaohua Li wrote:
> > > > > I'm wondering how we could use the perf to implament a clock_gettime.
> > > > > reading the perf fd or using ioctl is slow so reading the mmap
> > > > > ringbuffer is the only option. But as far as I know the ringbuffer has
> > > > > data only when an event is generated. Between two events, there is
> > > > > nothing we can read from the ringbuffer. Then how can application get
> > > > > time info in the interval?
> > > >
> > > > Are you wanting to read perf_clock from userspace?
> > >
> > > Yep, in some sort of form. Basically I want to read the time a task
> > > runs. Peter suggests we can read the activation time of a perf event.
> > > But I don't want to use any system call, as it's slow and likes
> > > clock_gettime.
> >
> > Since we cannot get the capability committed upstream a number of
> > folks are using this method:
> >
> >      https://github.com/dsahern/linux/blob/perf-full-monty/README.ahern
> >
> > ie., a KLM exports perf_clock and apps can use:
> >
> > #define CLOCK_PERF 14
> > if (clock_gettime(CLOCK_PERF, &ts) != 0) {
> > }
> >
> > No vdso acceleration, but works with an unmodified kernel.
>
> no, that's not what I want. as I said, we don't want any syscall (unless
> it's vdso based).

A number of clocks are vdso based (e.g., CLOCK_MONOTONIC). Those are not 
good enough?

David
--
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/