Opened 12 years ago
Closed 12 years ago
#35457 closed enhancement (fixed)
mpich2 variant for hdf5-18
| Reported by: | numaryu | Owned by: | eborisch (Eric A. Borisch) |
|---|---|---|---|
| Priority: | Normal | Milestone: | |
| Component: | ports | Version: | 2.1.2 |
| Keywords: | haspatch | Cc: | mamoll (Mark Moll) |
| Port: | hdf5-18 |
Description
I'd like to add +mpich2 variant for hdf5-18.
Attachments (3)
Change History (14)
Changed 12 years ago by numaryu
| Attachment: | Portfile.diff added |
|---|
comment:1 Changed 12 years ago by ryandesign (Ryan Carsten Schmidt)
| Owner: | changed from macports-tickets@… to eborisch@… |
|---|
comment:2 Changed 12 years ago by mamoll (Mark Moll)
comment:3 Changed 12 years ago by eborisch (Eric A. Borisch)
It is one of the ways to build parallel hdf5; just set the cc and cxx [fc] to the mpi compiler/wrapper scripts.
From http://www.hdfgroup.org/ftp/HDF5/current/src/unpacked/release_docs/INSTALL
A parallel version of HDF5 can be built by specifying `mpicc'
as the C compiler. (The `--enable-parallel' flag documented
below is optional in this case.) Using the `mpicc' compiler
will insure that the correct MPI and MPI-IO header files and
libraries are used.
$ CC=/usr/local/mpi/bin/mpicc ./configure
I'll try to look some night this week...
comment:4 Changed 12 years ago by eborisch (Eric A. Borisch)
I'm not sure if I should be the owner here. If you (mmoll) would like me to, I can take care of it. (Like I just mentioned, will likely be some evening this week.)
comment:5 Changed 12 years ago by mamoll (Mark Moll)
Sorry, I should have been more specific. I am referring to lines like this:
configure.compiler macports-gcc-4.5
comment:6 Changed 12 years ago by eborisch (Eric A. Borisch)
Hrmm. I guess that's a choice. It might be worth changing the gccNN variants to build everything with the selected gccNN (cc, cxx) compilers, rather than just using it for Fortran. I know some other ports work that way...
Discussion?
comment:7 Changed 12 years ago by numaryu
I found configure.compiler sets some FLAGS nicely as well. So, I set this to avoid the -arch x86_64 flag, otherwise compilation fails.
comment:8 Changed 12 years ago by mamoll (Mark Moll)
But this means that you *HAVE* to enable a Fortran variant for the mpich2 variant to compile. This is not the right solution. Your patch also enables unsupported combinations of extensions (C++ with parallel HDF5). Have you checked if "make check" still passes?
comment:9 Changed 12 years ago by eborisch (Eric A. Borisch)
Here's a (very modified) Portfile that I worked on this weekend. It supports lots of combinations (including "unsupported" versions - which the user is warned about)
It also takes the approach that a +gccXX variant means "compile C/CXX/Fortran variants (as enabled) with gccNN package compilers." (NOT just Fortran.)
Changed 12 years ago by eborisch (Eric A. Borisch)
Changed 12 years ago by eborisch (Eric A. Borisch)
| Attachment: | Portfile.2.diff added |
|---|
The diff for ease of discussion...
comment:10 Changed 12 years ago by mamoll (Mark Moll)
This looks good except for a missing conflict between the openmpi and mpich2 variant. You can commit this, if you like.
comment:11 Changed 12 years ago by eborisch (Eric A. Borisch)
| Resolution: | → fixed |
|---|---|
| Status: | new → closed |
mpich2 variant (and other changes) available in r97394.
Why does the patch change the compiler for the non-Fortran code?