[CMake] New FindMPI module has some backward-compatibility problems

Doug Gregor doug.gregor at gmail.com
Sun Mar 30 15:01:02 EDT 2008


On Fri, Mar 28, 2008 at 3:50 PM, Bill Hoffman <bill.hoffman at kitware.com> wrote:
>
> Doug Gregor wrote:
>  > I've been notified of some backward-compatibility issues in the
>  > FindMPI module, and have some other tweaks on hand to make it work
>  > better for more installations. I plan to finalize these changes this
>  > weekend. However, am I too late to fix this for the CMake 2.6.0
>  > release?
>  >
>
>  No, I am aware of the changes and am waiting for them. :)

I'm running into some issues with committing the updated FindMPI to
CVS [*], so here's the updated FindMPI module. I was able to test it a
bit on my end, but it could always use more testing. Changes in this
version:

  * MPI_COMPILER is the MPI compiler, rather than MPICXX. There's no
reason for this module to be C++-centric (that said, we'll still look
for the C++ drivers first)
  * Always respect MPI_LIBRARY and MPI_INCLUDE_PATH when they are set,
and don't reconfigure anything in this case (very important for
backward compatibility)
  * Don't put MPI_LIBRARIES into the cache; just set it as a normal
variable based on MPI_LIBRARY and MPI_EXTRA_LIBRARY.
  * Reconfiguring based on MPI_COMPILER clears out old compile flags/link flags
  * Support multiple include paths and link paths
  * Add equivalents to VTK_MPI_NUMPROC_FLAG, VTK_MPI_PREFLAGS,
VTK_MPI_POSTFLAGS, and VTK_MPI_MAX_NUMPROCS (but use MPIEXEC_* instead
of VTK_MPI_*, of course)
  * Added module documentation

Comments, bug reports fixes, etc, all welcome.

  - Doug

[*] I've dropped Zack a note about the CVS problem already.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: FindMPI.cmake
Type: application/octet-stream
Size: 11748 bytes
Desc: not available
URL: <http://www.cmake.org/pipermail/cmake/attachments/20080330/c41e43a2/attachment-0003.obj>


More information about the CMake mailing list