This thread has been opened to share my experience of installing QBox on the SGI Altix at the PSC. The information here should be pertinent to all SGI-Altix 4700 machines.
The build environment that I used is:
icc/10.1.008
ifort/10.1.008
imkl/10.0.3.020 (for BLAS,LAPACK, and SCALAPACK)
fftw/2.1.5
xerxes/2.8.0 (had to build from source, download and follow the instructions, nothing tricky there)
I created a new mk file called altix_icc.mk, it is attached to this post.
The build procedure is a little tricky, because the Intel 10.1 compilers appear to break on StructureFactor.C on compilation with optimization.
1. Alter altix_icc.mk to correspond to your environment
2. Compile code with 'make TARGET=altix_icc'
3. rm StructureFactor.o
4. Compile StructureFactor with no optimization (uncomment line in altix_icc.mk)
The four example jobs run to completion when using this procedure.
More later...
QBox on Pople (the PSC SGI Altix 4700)
Forum rules
You must be a registered user to post in this forum. Registered users may also post new topics if they consider that their subject do not correspond to any topic already present on the forum.
You must be a registered user to post in this forum. Registered users may also post new topics if they consider that their subject do not correspond to any topic already present on the forum.
-
- Posts: 2
- Joined: Mon Aug 04, 2008 6:10 pm
QBox on Pople (the PSC SGI Altix 4700)
- Attachments
-
- altix_icc.mk
- (2.05 KiB) Downloaded 934 times
-
- Posts: 2
- Joined: Mon Aug 04, 2008 6:10 pm
Re: QBox on Pople (the PSC SGI Altix 4700)
A comment on running... one may need to set the following environment variables:
MPI_COMM_MAX 1024
MPI_GROUP_MAX 1024
You will know when this is necessary, as it will explicitly tell you. The need for this will depend on the size of molecular system and the number of cores used to run the computation.
MPI_COMM_MAX 1024
MPI_GROUP_MAX 1024
You will know when this is necessary, as it will explicitly tell you. The need for this will depend on the size of molecular system and the number of cores used to run the computation.