scholarly journals Distribution of computations in hybrid computing systems when translating NORMA language programs

Author(s):  
А.Н. Андрианов ◽  
Т.П. Баранова ◽  
А.Б. Бугеря ◽  
К.Н. Ефимкин

Рассмотрены методы распределения вычислительной нагрузки при трансляции программ с непроцедурного (декларативного) языка НОРМА в исполняемые программы для различных параллельных архитектур. Приведены краткие характеристики языка НОРМА и основные возможности компилятора программ на языке НОРМА. Описаны способы автоматического распределения вычислительной нагрузки при генерации исполняемых программ следующих типов: OpenMP, NVIDIA CUDA, MPIOpenMP и MPIOpenMPNVIDIA CUDA. Рассмотрена задача динамической балансировки вычислительной нагрузки, возникающая в случае неоднородной вычислительной среды MPIOpenMPNVIDIA CUDA, и предложен метод ее решения. Приведены результаты практического применения компилятора программ на языке НОРМА для решения двух различных задач и оценена скорость выполнения получаемых при этом исполняемых программ для различных параллельных архитектур. The methods of computational load distribution when translating programs from the nonprocedural (declarative) NORMA language into executable programs for various parallel architectures are discussed. Some brief characteristics of the NORMA language and the main features of the compiler for programs in NORMA language are given. The methods of automatic distribution of computational load when generating executable programs of the following types are described: OpenMP, NVIDIA CUDA, MPIOpenMP, and MPIOpenMPNVIDIA CUDA. The problem of dynamic computational load balancing arising in the case of the heterogeneous computing environment MPIOpenMPNVIDIA CUDA is considered and a method of solving it is proposed. The results of practical application of the compiler for the programs in NORMA language for solving two different mathematical problems are given and the performance of the resulting executable programs is estimated for various parallel architectures.

Author(s):  
Nikolay Kondratyuk ◽  
Vsevolod Nikolskiy ◽  
Daniil Pavlov ◽  
Vladimir Stegailov

Classical molecular dynamics (MD) calculations represent a significant part of the utilization time of high-performance computing systems. As usual, the efficiency of such calculations is based on an interplay of software and hardware that are nowadays moving to hybrid GPU-based technologies. Several well-developed open-source MD codes focused on GPUs differ both in their data management capabilities and in performance. In this work, we analyze the performance of LAMMPS, GROMACS and OpenMM MD packages with different GPU backends on Nvidia Volta and AMD Vega20 GPUs. We consider the efficiency of solving two identical MD models (generic for material science and biomolecular studies) using different software and hardware combinations. We describe our experience in porting the CUDA backend of LAMMPS to ROCm HIP that shows considerable benefits for AMD GPUs comparatively to the OpenCL backend.


Sign in / Sign up

Export Citation Format

Share Document