CAMP first GPU solver: a solution to accelerate chemistry in atmospheric models
Visualitza/Obre
Estadístiques de LA Referencia / Recolecta
Inclou dades d'ús des de 2022
Cita com:
hdl:2117/383420
Tipus de documentText en actes de congrés
Data publicació2022-05
EditorBarcelona Supercomputing Center
Condicions d'accésAccés obert
Llevat que s'hi indiqui el contrari, els
continguts d'aquesta obra estan subjectes a la llicència de Creative Commons
:
Reconeixement-NoComercial-SenseObraDerivada 4.0 Internacional
Abstract
Atmospheric models are a representation of dynamical, physical, chemical, dynamical, and radiative processes in the atmosphere [1]. The load of these models is often spread across multiple processes in HPC environments. Most of this load comes from the resolution of chemical processes, which can take up to 90% of the total time execution [2]. Recent studies reported a relevant speedup by translating a chemical module to GPUs [3] [4]. This study is based in some previous works of the authors. These works are tested in the Chemistry Accross Multiple Phases (CAMP) module [5] simulating the conditions of an atmospheric model experiment. In our first approach we present an strategy to efficiently integrate GPU routines without needing to translate the entire chemical module to GPU code [6]. In our second and last work, we integrated a GPU version of the linear solver used in CAMP and evaluated multiple kernel configurations, achieving up to 34x speedup from the base CPU linear solver in a singlethread execution, in addition to a 2.7x for an equivalent MPI implementation with the maximum number of physical cores available on a node (40) [7]. The main objective of this work is to develop a GPU version of the entire CAMP solving algorithm. Our second objective is to evaluate the performance of our work, comparing the results with other state of the art GPU chemical modules.
CitacióGuzman-Ruiz, C.; Acosta Cobos, M.C.; Jorba, O. CAMP first GPU solver: a solution to accelerate chemistry in atmospheric models. A: . Barcelona Supercomputing Center, 2022, p. 49-50.
Fitxers | Descripció | Mida | Format | Visualitza |
---|---|---|---|---|
9BSCDS_18_CAMP First GPU.pdf | 784,6Kb | Visualitza/Obre |