Hi Martin,

Welcome back.

-j number should be controlled by the user who defined MAKEOPTS, right? We
are not explicitly define -j n in the patched code.


On Fri, Aug 13, 2010 at 1:57 PM, Martin Bligh <[email protected]> wrote:

> you have to be a bit careful just doing -j
> for large compiles (eg linux kernel) on small boxes, it tends to kill
> the whole box.
> I've found it safer to do '-j N' where N is twice the number of CPUs?
> (we have a utils
> function to get that).
>
> On Fri, Aug 13, 2010 at 4:49 PM, Eric Li(李咏竹) <[email protected]> wrote:
> > The idea here is to add -j option to most of our client tests at
> pre-build
> > stage. From our experiment, it reduced almost 50% of our build  time.
> > The original author of this CL is CCed.
> >
> > Index: client/common_lib/utils.py
> > ===================================================================
> > --- client/common_lib/utils.py (revision 4747)
> > +++ client/common_lib/utils.py (working copy)
> > @@ -1165,6 +1165,16 @@
> >      system('%s %s' % (configure, ' '.join(args)))
> >
> >
> > +def make(extra='', make='make', timeout=None, ignore_status=False):
> > +    """
> > +    Run make, adding MAKEOPTS to the list of options.
> > +
> > +    @param extra: extra command line arguments to pass to make.
> > +    """
> > +    cmd = '%s %s %s' % (make, os.environ.get("MAKEOPTS", ""), extra)
> > +    return system(cmd, timeout=timeout, ignore_status=ignore_status)
> > +
> > +
> >  def compare_versions(ver1, ver2):
> >      """Version number comparison between ver1 and ver2 strings.
> >
> > --
> > Eric Li
> > 李咏竹
> > Google Kirkland
> >
> >
> >
>



-- 
Eric Li
李咏竹
Google Kirkland
_______________________________________________
Autotest mailing list
[email protected]
http://test.kernel.org/cgi-bin/mailman/listinfo/autotest

Reply via email to