[RFC] use CLOCK_MONOTONIC_COARSE posix timer instead of CLOCK_MONOTONIC when calling the function of clock_gettime in xserver/os/util.c

Adam Jackson ajax at nwnk.net
Fri Aug 13 09:04:16 PDT 2010


On Fri, 2010-08-13 at 11:30 -0400, Adam Jackson wrote:
> On Fri, 2010-08-13 at 13:08 +0800, ykzhao wrote:
> > On Fri, 2010-08-13 at 11:32 +0800, Julien Cristau wrote:
> > > On Thu, Aug 12, 2010 at 09:47:23 +0800, ykzhao wrote:
> > > 
> > > > when playing flash h.264 cif video workload, the cpu C0 usage is reduced
> > > > from 43% to 42% while the video quality is not affected.
> > > > 
> > > Is 43% vs 42% a really significant win?  How close is it to measurement
> > > noise?
> > 
> > The purpose of using CLOCK_MONOTONIC_COARSE posix timer is to reduce
> > the overhead of read_hpet. 
> > 
> > Yes. The 1% ratio seems tiny. But the tiny ratio is related with that
> > the total overhead of read_hpet is not very high on the CPU usage(about
> > 3%) when playing the video workload. In such case it is not easy to get
> > one higher improvement.
> 
> On second thought, this seems like it's pretty awkward to use if we
> don't know the granularity.  And given that we're assuming milliseconds
> everywhere, you'd need CONFIG_HZ >= 1000 to get that.  Do we have a way
> of getting HZ from userspace?

Naturally, right after I say this I find clock_getres().

- ajax
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 198 bytes
Desc: This is a digitally signed message part
URL: <http://lists.x.org/archives/xorg-devel/attachments/20100813/d4c28821/attachment.pgp>


More information about the xorg-devel mailing list