On 2009-09-29 19:46-0400 Bill Hoffman wrote:
Alan W. Irwin wrote:
On 2009-09-29 16:36-0400 Bill Hoffman wrote:
make -j N is only supported with the all target.
How difficult would it be to implement parallel build support for more than
just the "all" target? Note, if there is some limitation that makes it
impractical or inefficient to implement this current "all" target
capability
for every target, you could instead implement this capability just for the
targets where this capability is wanted (as designated, say, by a
PARALLEL_BUILD=ON target property).
It is very hard to implement on top of make. Let me restate what I mean, it
is not the all target that is special. You can do make -j N target. You
just can't do make -jN target1 target2 if target1 and target2 have a common
target that they depend on.
Hi Bill:
I agree that is not too bad a limitation (compared to what you first
stated). When using my new test framework, I have tended to use one custom
target which depended on a whole group of other custom test targets. That
scenario satisfies your criterion of running parallel builds with only one
target at a time, and indeed I didn't notice any parallel build issues for
that case. So if I am able to anticipate the usual combinations of tests
our developers will want to run, and create a custom target for each such
combination, we should be okay.
The /fast targets could be used to avoid the issue you are having
maybe...
make -j2 x01c/fast x02c
Then it will only trace the depends of one of them... I have not tested that
but it might work... :)
Just as a matter of interest that make command didn't work (because the
required plplotd dependency of x01c was missing) but
make -j2 x01c x02c/fast
worked without any attempts to double build plplotd. But I think that
worked only because x01c and x02c have absolutely common dependencies and
/fast won't work for our more usual case where the tests have combinations
of unique dependencies as well as some dependencies that are common with
other tests. But as I said above, I think we are okay, now that we know the
exact limitation on parallel builds.
Thanks for clarifying that limitation.
Alan
__________________________
Alan W. Irwin
Astronomical research affiliation with Department of Physics and Astronomy,
University of Victoria (astrowww.phys.uvic.ca).
Programming affiliations with the FreeEOS equation-of-state implementation
for stellar interiors (freeeos.sf.net); PLplot scientific plotting software
package (plplot.org); the libLASi project (unifont.org/lasi); the Loads of
Linux Links project (loll.sf.net); and the Linux Brochure Project
(lbproject.sf.net).
__________________________
Linux-powered Science
__________________________
_______________________________________________
Powered by www.kitware.com
Visit other Kitware open-source projects at
http://www.kitware.com/opensource/opensource.html
Please keep messages on-topic and check the CMake FAQ at:
http://www.cmake.org/Wiki/CMake_FAQ
Follow this link to subscribe/unsubscribe:
http://www.cmake.org/mailman/listinfo/cmake
