Re: multiple cpu's

[Benoit SIGOURE]

On Aug 20, 2007, at 5:15 PM, Jim Meyering wrote:

Benoit SIGOURE <[EMAIL PROTECTED]> wrote:

On Aug 12, 2007, at 11:09 PM, Ralf Wildenhues wrote:

Hello Jim,

* Jim Meyering wrote on Sun, Aug 12, 2007 at 12:46:20PM CEST:

On a related note, it'd sure be nice if "make -j4 check" would run
its
tests in parallel.  Especially on an SMP system, and/or when some
tests sleep for a few (interminable :-) seconds.

Of course, this might not be trivial, especially trying to keep
the resulting Makefile.in portable.  However, even if it means
I'd need to use a special option to generate a GNU-make-only
Makefile.in, I'd happily use the result for day to day work.

but on my list it's a bit on the far end.

IIRC, Akim Demaille posted a patch a while ago on an automake list for this, GNU make-specific. I don't recall how much work was left to do,

Yep:
http://lists.gnu.org/archive/html/automake/2007-02/msg00015.html
The whole idea is to have one input file per test (say a .chk file)
that produces a .log file.  We can then benefit from the parallelism
of make with implicit rules (exactly like for .c -> .o).

The file has slightly evolved since then but I can post an updated
version if someone is interested.

That'd be great.

FYI, in testing this, I've been using check.mk from vaucanson-trunk.
Is that what you mean?

http://www.tsunanet.net/~tsuna/check.mk


I hope this helps.

Cheers,

--
Benoit Sigoure aka Tsuna
EPITA Research and Development Laboratory

Not exactly, in fact Akim and I use this piece of Makefile in several other projects, some of which are unfortunately not open source. You can find a more up to date version at: It has a couple of bug fixes and enhancements compared to that in vaucanson HEAD (which ought to be updated).

PGP.sig
Description: This is a digitally signed message part