Mixed precision and minimal data transfer algorithms: simple and efficient implementations

• Grant number: 24/09172-3
• Support Opportunities: Scholarships in Brazil - Master
• Start date: August 01, 2024
• Status: Discontinued
• Field of knowledge: Physical Sciences and Mathematics - Mathematics - Applied Mathematics
• Agreement: Equinor (former Statoil)
• Principal Investigator: Alberto Vazquez Saa
• Grantee: Mateus Orsoni Cabral
• Host Institution: Instituto de Matemática, Estatística e Computação Científica (IMECC). Universidade Estadual de Campinas (UNICAMP). Campinas , SP, Brazil
• Company: Universidade Estadual de Campinas (UNICAMP). Faculdade de Engenharia Mecânica (FEM)
• Associated research grant: 17/15736-3 - Engineering Research Centre in Reservoir and Production Management, AP.PCPE
• Associated scholarship(s): 25/14902-3 - Enhancing multi-GPU physical simulations with mixed-precision and minimal data transfer techniques, BE.EP.MS

Abstract

This Work Package (WP) is a Ph.D. research project to explore new mathematical andcomputational ideas for high-performance computational applications for reservoir simulation,aiming possible implementations in the Open Porous Media (OPM) initiative [1], aninternational project devoted to develop and disseminate open-source alternatives for realisticreservoir simulation.Reservoir simulation has become an essential tool for the oil and gas industry. Roughly, thesetools are necessary to predict fluid dynamics based on physical models of the reservoirs andfluids, which are commonly built from different collected field data and laboratorymeasurements. In a realistic field operation, typically a huge number of different strategies aretested to find optimal results. From a computational point of view, these simulations areenormously time and memory demanding, and the exploration of high-performancecomputational schemes for these applications is a timely fundamental research topic for boththe Industry and Academia.Despite the unequivocal importance of the commercial software packages available today forreservoir simulation, the nature of their licenses (proprietary) prevents the proper developmentof their capabilities in the Industry and Academia. Proprietary software simulators aretypically distributed without the underlying source code, which obviously prevents the usercommunity from proposing, for instance, new features or improvements on the alreadyavailable packages. The open-source option, on the other hand, not only allows theparticipation of the user community but indeed promotes user contributions, leading typicallya much faster and more reliable code development. The oil and gas industry and Academiawill certainly benefit from open-source projects like the OPM initiative, a potentiallydisruptive project that proposes the development and popularization of open-sourcealternatives for the most commonly used proprietary reservoir simulation software packagesavailable today. The present WP aims to explore, from a mathematical and computationalpoint of view, some mixed precision and minimal data transfer algorithms with possibleapplication to the reservoir simulation package of the OPM Initiative.The methodology of mathematical and computational research as the proposed in this WP isbased on extensive literature revisions and testing of new ideas with prototypical problemsaiming for possible applications to more realistic situations.