Designing an HPC infrastructure with the greatest qualities and at a reasonable cost requires optimising HPC systems based on performance variables and bottlenecks. Only a thorough examination of existing HPC systems and the execution of their workloads can provide such information. The "Quinde I" is Ecuador's only and most powerful supercomputer, currently ranking third in South America. It was constructed using IBM Power 8 servers. We tested its performance using several High-Performance Computing (HPC) parameters and compared it to theoretical values and values derived from testing on similar models in this study. We generated and performed multiple benchmarks with particular optimization flags for Power 8 to acquire the most performance with the existing configuration of the vendor-installed hardware to test its performance. Benchmarking is a technique used in computer science to evaluate the performance of a system or one of its components. The benchmarks' inputs were adjusted to see how they affected system performance. Furthermore, we assemble and compare the performance of two dense matrix multiplication algorithms, SRUMMA and DGEMM.
Author(S) Details
Eduardo Patricio Estévez Ruiz
Grupo de Polímeros, Departamento de Física y Ciencias de la Tierra, Escuela Universitaria Politécnica, Universidade da Coruña, 15471 Ferrol, Spain.
Giovanny Eduardo Caluña Chicaiza
School of Mathematical Sciences and Information Technology, Yachay Tech University, Imbabura, 100119, Ecuador.
Fabian Rodolfo Jiménez Patiño
Department of Technological Operations, Alluriam Inc Co., Orlando, FL 32801, USA.
Joaquín Cayetano López Lago
Grupo de Polímeros, Departamento de Física y Ciencias de la Tierra, Escuela Universitaria Politécnica, Universidade da Coruña, 15471 Ferrol, Spain.
Saravana Prakash Thirumuruganandham
Centro de Investigación en Mecatrónica y Sistemas Interactivos—MIST del Instituto de Investigación, Desarrollo e Innovación, Universidad Tecnológica Indoamérica (UTI), Ambato 180103, Ecuador.
View Book:- https://stm.bookpi.org/RAMRCS-V5/article/view/4972
No comments:
Post a Comment