Yeah seems that alternative compilers like pathscale don't work so well with libhugetlbfs and code segments. I'm currently trying to get intel and portland group compilers a test and we'll see what happens with them.
Thanks, for your help is there any chance you guys could get a hold of pathscale, intel, or pg compilers to do testing? If you have the time, we have the machine you could use. I'll have to get permission to see if its within our scope for the project to allow you access to do development but I can push it and see what happens. Thanks, - David Brown On Fri, Jul 10, 2009 at 10:38 AM, David Brown<dmlb2...@gmail.com> wrote: >> I have two quesitons. First, does pathscale include its own linker and if >> not, does >> it pass linker flags onto ld he same way that gcc does? Second, is it >> possible to test >> nwchem using the gnu tool chain? > > pathscale calls the system ld "same way" is difficult to say, all the > -L flags go first on the link line then all the -l flags for the > various libraries then the finish linker specific flags. However, not > sure if there's anything special the -m flag is also passed to ld > using pathscale. > > Yes it is possible to build nwchem using gnu tool chain but the > performance tanks. This is a very old fortran code and gfortran just > doesn't optimize well enough for nwchem. However, this would be a good > test to see if the pathscale compiler is causing issues, I'll try this > as well. > > Thanks, > - David Brown > ------------------------------------------------------------------------------ _______________________________________________ Libhugetlbfs-devel mailing list Libhugetlbfs-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/libhugetlbfs-devel
