Hi,
On Mon, Feb 21, 2011 at 9:06 PM, Ben Roberts <ben.roberts.geek.nz> wrote:
> I noticed that someone added this option to the C++ flags recently. Unfortunately, it seems to cause issues, at least when building with the Intel compilers on the Mac. Reduce and gleap, at least, don't take kindly to it.
This was from me. It was one of two things I found necessary for
compilation with the latest intel compilers (v12). I sent a note about
it earlier (here are the Cliff's notes):
1) The '-shared-intel' flag is needed during fortran (maybe also C,but
definitely fortran) link steps for dynamic linking of libraries
(
http://archive.ambermd.org/201012/0085.html).
2) The '-std=c++0x' flag is needed during compilation of c++ files
that include <iomanip> when the header files belong to gnu > 4.4
(
http://software.intel.com/en-us/forums/showthread.php?t=78677&p=1&wapkw=%28Multi+Threading+.NET*+Environment%29).
Seeing as how #2 is a very specific case there could either be a flag
to turn it on if necessary (or turn it off if you're on a Mac or
something).
I'm curious as to what issues come up. What exactly fails, and what
version of icc/icpc (and what version headers are being accessed)? I'm
wondering if this is compiler or architecture specific. On my fedora14
box with intel 12.0 and gnu headers 4.5.1 the -std=c++0x was necessary
for compilation.
> Given how close we are to the impending release of AT, is it worth taking that option out for the moment?
I'm more for figuring out how to make it so that AT will compile with
the v12 compilers, but if people don't think it's worth it I'm not
going to lose any sleep over it.
-Dan
_______________________________________________
AMBER-Developers mailing list
AMBER-Developers.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber-developers
Received on Mon Feb 21 2011 - 20:30:05 PST