Dear vasp master,
I have some query regarding parallel compilation of VASP 5.2 on our Intell Xeon cluster
I am using the following compiler and library to compile.
OS : fc12.x86_64
Intel ifort 11.0.83
Openmpi 1.4.3
MKL Library 10.1.2.024 and FFTW library from Intel.
The system was Intel Xeon X5570 @ 2.93GHz with cache size : 8192 KB
Connectivity in between nodes through infiniband
My query is what will be the optimal DCACHE_SIZE and DMPI_BLOCK values for the compilation. And also for other system architecture how to decide the optimum values of the above two variables.
Thanks in advance .......
Optimum values for DCACHE_SIZE and DMPI_BLOCK value
Moderators: Global Moderator, Moderator
- shaldar
- Newbie
- Posts: 7
- Joined: Tue Oct 20, 2009 5:08 am
- License Nr.: 20-0458
- Location: Kiel, Germany
Optimum values for DCACHE_SIZE and DMPI_BLOCK value
Last edited by shaldar on Tue Oct 12, 2010 12:20 pm, edited 1 time in total.
Dr. Soumyajyoti Haldar
Postdoctoral Researcher
Christian-Albrechts-Universität zu Kiel
Institut für Theoretische Physik und Astrophysik
Leibnizstraße 15, D-24098 Kiel, Germany
Postdoctoral Researcher
Christian-Albrechts-Universität zu Kiel
Institut für Theoretische Physik und Astrophysik
Leibnizstraße 15, D-24098 Kiel, Germany
-
- Administrator
- Posts: 2921
- Joined: Tue Aug 03, 2004 8:18 am
- License Nr.: 458
Optimum values for DCACHE_SIZE and DMPI_BLOCK value
please make benchmark tests to determine the optimum values
for your system and applications. The values set in the makefiles
are suggesttions which work well usually
for your system and applications. The values set in the makefiles
are suggesttions which work well usually
Last edited by admin on Fri Oct 29, 2010 10:21 am, edited 1 time in total.
-
- Newbie
- Posts: 18
- Joined: Thu Nov 25, 2010 5:41 am
Optimum values for DCACHE_SIZE and DMPI_BLOCK value
[quote="shaldar"]Dear vasp master,
I have some query regarding parallel compilation of VASP 5.2 on our Intell Xeon cluster
I am using the following compiler and library to compile.
OS : fc12.x86_64
Intel ifort 11.0.83
Openmpi 1.4.3
MKL Library 10.1.2.024 and FFTW library from Intel.
The system was Intel Xeon X5570 @ 2.93GHz with cache size : 8192 KB
Connectivity in between nodes through infiniband
My query is what will be the optimal DCACHE_SIZE and DMPI_BLOCK values for the compilation. And also for other system architecture how to decide the optimum values of the above two variables.
Thanks in advance ....... [/quote]
Hi, I am a new user of VASP, and it seems that we have the similar hard ware configuration, however, now the VASP does not run very well on my cluster, Once I run the VASP on more than two nodes, it crashed. So could you share your makefile with me? Thanks a lot....
I have some query regarding parallel compilation of VASP 5.2 on our Intell Xeon cluster
I am using the following compiler and library to compile.
OS : fc12.x86_64
Intel ifort 11.0.83
Openmpi 1.4.3
MKL Library 10.1.2.024 and FFTW library from Intel.
The system was Intel Xeon X5570 @ 2.93GHz with cache size : 8192 KB
Connectivity in between nodes through infiniband
My query is what will be the optimal DCACHE_SIZE and DMPI_BLOCK values for the compilation. And also for other system architecture how to decide the optimum values of the above two variables.
Thanks in advance ....... [/quote]
Hi, I am a new user of VASP, and it seems that we have the similar hard ware configuration, however, now the VASP does not run very well on my cluster, Once I run the VASP on more than two nodes, it crashed. So could you share your makefile with me? Thanks a lot....
Last edited by Gu Chenjie on Tue Feb 08, 2011 5:15 am, edited 1 time in total.
-
- Administrator
- Posts: 2921
- Joined: Tue Aug 03, 2004 8:18 am
- License Nr.: 458
Optimum values for DCACHE_SIZE and DMPI_BLOCK value
this rather indicates an error in the installation/setup of your local MPI, the combination Intel Compiler /Openmpi is a very stable one, if you use the standard makefiles for the Intel compiler
Last edited by admin on Tue Feb 08, 2011 1:45 pm, edited 1 time in total.