Dear all!

I am afraid I may re-raise an issue debated previously, but I would be 
interested if there maybe are some new thoughts.

I am one of the developers of NEST, the Neural Simulation Tool 
(www.nest-initiative.org). Our users work on everything from extremely large 
computers to laptops running Linux or OSX. In the past, supporting NEST on OSX 
was reasonably easy using MacPorts, but over the past year two developments 
have been working against us: 

- We have started to use OpenMP in NEST.
- Apple has made Clang the standard compiler, and Clang does not support OpenMP.

While NEST can be built without OpenMP, it loses performance, so we really 
would like to build NEST with OpenMP. We can do that with either llvm-gcc42 or 
with the one of the MacPorts GCCs, but this leads to a number of problems:

- NEST has a Python UI. Building NEST with a different compiler than Python 
usually leads to trouble.
- NEST can also be compiled with OpenMPI. For this to work well, NEST should be 
built with the same compiler as OpenMPI.
- NEST links against GSL. Here we have experience little trouble, but ideally 
we should use the same compiler for GSL and NEST, too.

Everything would be solved if I could download MacPorts, configure it to use, 
e.g., MacPorts gcc 4.7 to build everything, and then install all the ports I 
need for NEST---I could even write a Portfile for NEST :). But as things stand, 
Python and OpenMPI are compiled with Clang, and as far as I can see there 
aren't even variants that allow you to switch the compiler to GCC (the +gccNN 
variants of OpenMPI only select the Fortran compiler).

I much appreciate MacPorts philosophy of making itself maximally independent 
from tools shipped with OSX. Therefore, I have never quite understood why 
MacPorts insists on using the system compiler. I very well understand that it 
is virtually impossible to support a range of compilers, but why not use 
MacPorts own GCC as default then? As most of the packages provided through 
MacPorts originate from the Unix/Linux-world, I'd expect them to be 
GCC-friendly.

I'd much appreciate some information on why MacPorts has made the choices it 
has made and whether there are any plans for change.

Best regards,
Hans
 

--

Dr. Hans Ekkehard Plesser, Associate Professor
Head, Basic Science Section

Dept. of Mathematical Sciences and Technology
Norwegian University of Life Sciences
PO Box 5003, 1432 Aas, Norway

Phone +47 6496 5467
Fax   +47 6496 5401
Email [email protected]
Home  http://arken.umb.no/~plesser

_______________________________________________
macports-users mailing list
[email protected]
https://lists.macosforge.org/mailman/listinfo/macports-users

Reply via email to