[gmx-users] Another Compiling Adventure

Hank deHaan hdh at jupiter.physics.uoguelph.ca
Fri Dec 14 02:04:29 CET 2001


Thanks for your reply,
you convinced me

However,
I've still got problems

1) FFTW won't install with the --enable-mpi option
   error message:
	checking for mpicc... no
	checking for hcc... no
	checking for MPI_Init... no
	checking for MPI_Init in -lmpi... no
	configure: error: couldn't find mpi library for --enable-mpi



2) G3 won't configure with the --enable-mpi option
   error message:
	checking for mpxlc... no
	checking for mpicc... no
	checking for mpcc... no
	checking for hcc... no
	checking whether the MPI cc command works... configure: error:
	Cannot compile and link MPI code with ccc

3) Installing G3 without --enable-mpi option, I still
   receive the following error all over the place:
	In the declaration of "__data", zero cannot be used as an element
	count specifier.  The specifier will be ignored, (leaving the
	member/parameter with an incomplete array type) in this context.
	(zeroelements)

   However, it does seem to compile, and even work.  Do I need to worrry
   about this?

The bulk of my problems seem to be in recognizing the MPI package while
trying to compile in parallel.  The package that is installed is the RMS
package (?) with command "prun".

I don't need PME (at least immediately), so i could remove the calls to
fftw so that gromacs can compile without it (how do i do this, btw).  But
i do need gromacs to install in parallel so i need it to talk to those
MPI libraries.

Thanks for all your help,

			Hendrick de Haan





On Thu, 13 Dec 2001, K.A.Feenstra wrote:

> Hank deHaan wrote:
> >
> > Hello All
> >
> > I'm attempting to compile gromacs (version 2.0) on a Beowulf built from
> > Alpha chips running Linux.  The compaq C/Fortran compilers are installed,
> > but when i go to build it, I get the following error:
>
> Sorry, but we don't support 2.0 any more.
> I hope my comments below will convince you to use 3.0 in stead.
>
> > Another question,
> > I'd like to install Gromacs version 3,
> > but i don't think the system admins will want to change the lam version
> > for me, is there any way to complile with a (slightly) different
> > version than the one required by Gromaces 3?
>
> Compiling agains a different lam version is no problem, it is just
> that the pre-compiled .rpm requires the exact lam version.
>
> > how about getting around having those fast-fourier libraries installed?
>
> If you leave the fft out, you can't use PME long range interactions.
> So if you don't need PME, you don't need fft.
>
>





More information about the gromacs.org_gmx-users mailing list