National Supercomputing Forum (NSCF-2017) was held on November 28th - December 1st, 2017 in Russia, Pereslavl, the Program Systems Institute of RAS.
The main difficulty in the development of parallel programs for the cluster is the need to make global decisions on the distribution of data and calculations taking into account the properties of the whole program, and then perform painstaking work on program modification and debugging. A large amount of code, multi-modularity, multifunctionality makes it difficult to make decisions on the coordinated distribution of data and calculations. To solve this problem the method of incremental or partial parallelization can be used. The idea of this method is that not the whole program is subjected to parallelization, but its parts (areas of parallelization) – additional copies of the required data are created in them, the distribution of the data and the corresponding calculations are performed. These areas can be built based on the times obtained by profiling a sequential program.
Presented by V.A. Bakhtin.
Performance results for benchmarks from NAS NPB 3.3 is updated. These benchmarks has been parallelized using Fortran-DVMH language and were executed on graphics accelerators NVIDIA, on 6-cores processor Intel and on 60-cores Intel Xeon Phi.
Parallel computational technologies (PCT) 2017 was held on April 3rd - 7th, 2017 in Russia, Kazan:Kazan, Kazan Federal University.
Automated parallelization of a simulation method of elastic wave propagation in media with complex 3D geometry surface on high-performance heterogeneous clusters
The paper considers application of DVM and SAPFOR systems in order to automate mapping of 3D elastic waves simulation method on high-performance heterogeneous clusters. A distinctive feature of the proposed method is the use of a curved threedimensional grid, which is consistent with the geometry of free surface. Usage of curved grids considerably complicates both manual and automated parallelization. Technique to map curved grid on a structured grid has been presented to solve this problem. The sequential program based on the finite difference method on a structured grid, has been parallelized using Fortran-DVMH programming language. Application of SAPFOR analysis tools simplified this parallelization process. The paper describes features of automated parallelization. Authors estimate efficiency and acceleration of the parallel program and compare performance of the DVMH-based program with a program obtained after manual parallelization using MPI programming technology.
Presented by A.S. Kolganov.
Parallel Large-Scale Graph Processing was held on March 2nd, 2017 in Russia, Moscow, Lomonosov Moscow State University.
DVM-system was designed to create parallel programs of scientific-technical calculations in C-DVMH and Fortran-DVMH languages. These languages use the same model of parallel programming (DVMH-model) and are the extensions of standard C and Fortran languages with parallelism specifications, implemented as compiler directives. DVMH-model allows to create efficient parallel programs for heterogeneous computational clusters, which nodes use as computing devices not only universal multi-core processors but also can use attached accelerators (GPUs or Intel Xeon Phi coprocessors). This report discusses new means to work with irregular grids, which were implemented in the C-DVMH compiler recently. Using the developed extension can significantly ease irregular grid applications parallelization on cluster.
Presented by A.S. Kolganov.
DVM system was installed at Siberian Supercomputer Center on NKS-30T supercomputer of Institute of Computational Mathematics and Mathematical Geophysics of Siberian Branch of Russian Academy of Sciences (ICMMG SB RAS). Instruction how to use DVM-sysytem on NKS-30T is available here. The instructions for all computer systems where DVM system is installed are here.