I think I sent this to the wrong place, I've cc'd petsc-maint on this reply which includes the patch. There are a couple of other things that have to be done that I couldn't include in the patch because I believe it requires modifying the files used to generate the configure script:
1) On windows, petsc expects libml.lib not libml.a and so the Makefile has to be changed to build libml.lib. 2) I believe you can pass AR to the configure script, but in any case, you need AR="$PETSC_DIR/bin/win32fe/win32fe lib -a" and RANLIB=/usr/bin/true in the Makefile 3) You have to add "-DICL" to CFLAGS and CXXFLAGS. 4) The patch I've attached, which changes sleep to Sleep and excludes unistd.h when ICL is defined. It also removes "--verbose" from the things that configure tries, because the result isn't parsed correctly (it makes configure think that "-link" needs to be added to the link line, which is erroneous). I believe those are all the changes I made. I hope this patch can be applied to the ml tarball so that windows users have a better shot at using ml+petsc. John On Wed, Jun 1, 2011 at 10:22 AM, John Fettig <john.fettig at gmail.com> wrote: > I have been working on compiling ML on windows using the tarball that > is downloaded through configure.py from petsc 3.1 p8. ?There are > several changes to the source code and to the configure/Makefile that > I found necessary to get it to compile with the intel compilers. > Would you be interested in merging these changes into your tarball? > If so, I will put together a patch. > > Cheers, > John > -------------- next part -------------- A non-text attachment was scrubbed... Name: ml.patch Type: application/octet-stream Size: 4025 bytes Desc: not available URL: <http://lists.mcs.anl.gov/pipermail/petsc-dev/attachments/20110602/aaac3cfc/attachment.obj>
