PIRA: Performance Instrumentation Refinement Automation
In this paper we present PIRA – an infrastructure for automatic instrumentation refinement for performance analysis. It automates the process to generate an initial performance overview measurement and gradually refines it, based on the recorded runtime information. This relieves a performance analyst from the time consuming and largely manual, yet mechanical, task of selecting which functions to capture in subsequent measurements. PIRA implements an existing aggregation strategy that heuristically determines which functions to include or exclude for initial overview measurements. Moreover, it implements a newly developed heuristic to incorporate profile information and expand instrumentation in hot-spot regions only. The approach is evaluated on different benchmarks, including the SU2 multi-physics solver package. PIRA is able to generate instrumentation configurations that contain the application’s hot-spot, but generate significantly less overhead when compared to the Score-P reference measurement.
Tue 6 NovDisplayed time zone: Guadalajara, Mexico City, Monterrey change
08:00 - 10:00 | AI SEPSAI-SEPS at Cabot Chair(s): Ali Jannesari Iowa State University, Yukinori Sato Toyohashi University of Technology | ||
08:00 50mTalk | Deep Learning at ScaleKeynote AI-SEPS Prabhat NERSC, Berkeley Lab | ||
08:50 25mTalk | PIRA: Performance Instrumentation Refinement Automation AI-SEPS Jan-Patrick Lehr Graduate School of Computational Engineering, TU Darmstadt, Alexander Hück Institute for Scientific Computing, TU Darmstadt, Christian Bischof Scientific Computing, TU Darmstadt | ||
09:15 15mTalk | PyGA: A Python to FPGA compiler prototype AI-SEPS | ||
09:30 30mTalk | Panel discussion AI-SEPS P: Yukinori Sato Toyohashi University of Technology, P: Ali Jannesari Iowa State University, P: Shigeru Chiba The University of Tokyo |