[petsc-dev] Building PETSc with PGI Compilers

[Jed Brown]

On Thu, Dec 22, 2011 at 23:52, Dave Nystrom
<Dave.Nystrom at tachyonlogic.com>wrote:

> So, should I specify each of the compiler environment variables this way?
> That is,
>
> CC=/path/to/pgcc
> CXX=/path/to/pgCC
> FC=/path/to/pgfortran
>

Yes


>
> Should I also add --with-gnu-compilers=0 ?
>

Doesn't matter, you can drop this option entirely.


>  > >  I'm also interested in seeing what
>  > > difference PETSc in general would see in performance for PGI versus
> GNU.
>  >
>  > I have always found PGI to be a waste of time.
>
> OK.  What about Intel or any other vendor compilers?
>

IBM compilers are mandatory on Blue Gene because the GCC people don't have
resources to optimize for that platform. Clang (open source, part of LLVM)
is totally worthwhile for the better error messages, and it has fully
compatible command line options to GCC. Intel compilers are usually fairly
straightforward to try. I usually don't see a big difference in end-to-end
performance, but sometimes Intel is a clear winner in micro-benchmarks. The
problem with PGI is that they are usually a bad user experience (weird
environment, poor diagonstics, often takes significant effort to build
code) and don't offer much if anything in performance. In contrast, Intel
compilers usually don't require much effort to try.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/petsc-dev/attachments/20111223/259181bf/attachment.html>