Enhancing the in Situ Visualization of Performance Data in Parallel CFD Applications

Research output: Contribution to journalResearch articleContributedpeer-review

Abstract

This paper continues the work initiated by the authors on the feasibility of using ParaView as visualization software for the analysis of parallel CFD codes’ performance. Current performance tools are unable to show their data on top of complex simulation geometries (e.g. an aircraft engine). In our previous paper, a plugin for the open-source performance tool Score-P has been introduced, which intercepts an arbitrary number of manually selected code regions (mostly functions) and send their respective measurements – amount of executions and cumulative time spent – to ParaView (through its in situ library, Catalyst), as if they are any other flow-related variable. This paper adds to such plugin the capacity to also show communication data (messages sent between MPI ranks) on top of the CFD mesh. Testing is done again with Rolls-Royce’s in-house CFD code, Hydra. The plugin’s original feature (regions’ measurements) is here revisited, in a bigger test-case, which is also used to illustrate the new feature (communication data). The benefits and overhead of the tool are discussed.

Details

Original languageEnglish
Pages (from-to)16-31
Number of pages16
JournalSupercomputing frontiers and innovations : an international journal
Volume7
Issue number4
Publication statusPublished - 2020
Peer-reviewedYes

Keywords

Research priority areas of TU Dresden

Keywords

  • computational fluid dynamics, in situ processing, parallel computing, performance analysis

Library keywords