Dear all, I've installed ALPS2.0 on a 12-core Linux Server Computer. When running programs using DMRG, one program usually uses more than one CPU, e.g., even 390% CPU. I'd like to know if there're some places we should pay attention to in the installation or the using process. I doubt whether the ALPS2.0 version has applied the Parallel Programming Algorithm. Does the usage of 390% CPU mean the speeding up of the computation? What I have to do if I'd like one DMRG program uses just one CPU? Yours sincerely, L. H. Pan
Hi,
2011/10/24 潘丽华 lhpan@yzu.edu.cn:
I've installed ALPS2.0 on a 12-core Linux Server Computer. When running programs using DMRG, one program usually uses more than one CPU, e.g., even 390% CPU. I'd like to know if there're some places we should pay attention to in the installation or the using process. I doubt whether the ALPS2.0 version has applied the Parallel Programming Algorithm. Does the usage of 390% CPU mean the speeding up of the computation? What I have to do if I'd like one DMRG program uses just one CPU? Yours sincerely,
This comes from the shared-memory parallelization of the LAPACK routine by OpenMP (Intel MKL, etc.) or by pthread (ATLAS).
Look at the usage guide of your LAPACK library which you linked against the ALPS library/applications.
Best regards,
comp-phys-alps-users@lists.phys.ethz.ch