This course covers performance engineering approaches on the compute node level. Even application developers who are fluent in OpenMP and MPI often lack a good grasp of how much performance could at best be achieved by their code. This is because parallelism takes us only half the way to good performance. Even worse, slow serial code tends to scale very well, hiding the fact that resources are wasted. This course conveys the required knowledge to develop a thorough understanding of the interactions between software and hardware. This process must start at the core, socket, and node level, where the code gets executed that does the actual computational work. We introduce the basic architectural features and bottlenecks of modern processors and compute nodes. Pipelining, SIMD, superscalarity, caches, memory interfaces, ccNUMA, etc., are covered. A cornerstone of node-level performance analysis is the Roofline model, which is introduced in due detail and applied to various examples from computational science. We also show how simple software tools can be used to acquire knowledge about the system, run code in a reproducible way, and validate hypotheses about resource consumption. Finally, once the architectural requirements of a code are understood and correlated with performance measurements, the potential benefit of code changes can often be predicted, replacing hope-for-the-best optimizations by a scientific process.
This course provides - via lectures, demos, and hands-on labs - scientific training in Computational Science, and in addition, the scientific exchange of the participants among themselves.
Online course Organizer: HLRS, University of Stuttgart, Germany
Jul 12, 2021
Jul 15, 2021
Back to list
08:45 Enter the Zoom meeting 09:00 Welcome – Intro 09:30 Computer architecture for software developers part 1 10:00 Coffee break 10:15 Computer architecture for software developers part 1 cont. 10:45 Hands-on: Warmup (login etc.) 11:00 Computer architecture for software developers part 2 11:45 Hands-on: Divide benchmark 12:30 Lunch 13:30 Tools: Topology and affinity, frequency 14:00 Hands-on: likwid-topology and likwid-pin 14:15 Roofline Model: Basics 15:15 Hands-on: Microarchitectural exploration 16:00- Open end Day 2
08:45 Enter the Zoom meeting 09:00 Tools: Performance counters 09:30 Hands-on: performance counters and memory bandwidth 10:00 Optimal use of parallel resources: SIMD 11:00 Coffee break 11:15 Hands-on: SIMD in MiniMD 12:15 Optimal use of parallel resources: ccNUMA 13:00 Lunch 14:00 Performance Engineering: Basic skills 14:45 Coffee break 15:00 Hands-on: Dense Matrix Vector Multiplication 16:00- Open end Day 3
08:45 Enter the Zoom meeting 09:00 Roofline case study: Jacobi smoother 10:00 Coffee break 10:15 Hands-on: Matrix free CG solver part 1 11:15 Coffee break 11:30 Roofline case study: SpMVM 12:30 Lunch 13:30 ECM Performance model 14:15 Coffee break 14:30 Hands-on: Matrix free CG solver part 2 (optimizatons) 16:00 Farewell & feedback
Day 4 (Tools Day)
08:45 Enter the Zoom meeting 09:00 Performance engineering from the application point of view 09:30 Monitoring with Score-P 09:45 Hands-on: MiniMD with Score-P 10:15 Coffee break 10:30 Profile exploration and measurement scoring 10:45 Trace exploration with Vampir 11:15 Coffee break 11:30 Hands-on: Load imbalance: SpMV 12:15 Wrap-up
Tools topology & affinity in multicore environments
Microbenchmarking for architectural exploration
Roofline model: basics
Tools: hardware performance counters
Roofline case studies
Optimal use of parallel resources
Extending Roofline: The ECM performance model
Optional: Pattern-based performance engineering
Participants should be able to work on the Linux command line. They must have basic knowledge in programming with Fortran or C, and basic OpenMP.
Learn more about course curricula and content levels.
The course language is English.
Dr. habil. Georg Hager and Dr.-Ing. Jan Eitzinger (formerly Treibig) (RRZE/HPC, Uni. Erlangen) Bert Wesarg (ZIH Uni. Dresden) for the tools-day
Before the course, the course material and an updated agenda will be available here. An older version of this course with most of the material (including the audio information) can also be viewed in the ONLINE Parallel Programming Workshop.
Registration is closed.
for registration is June 27, 2021 (extended deadline).
Late registrations after the deadline are still possible but not later than July 4, 2021 and maybe with reduced quality of the service.
Members of German universities and public research institutes: none. Members of universities and public research institutes within Europe or
see our How to find us page.
HLRS is part of the Gauss Centre for Supercomputing (GCS), which is one of the six PRACE Advanced Training Centres (PATCs) that started in Feb. 2012. The mandate for the PATCs is as follows: "The PRACE Advanced Training Centres will serve as European hubs of advanced, world-class training for researchers working in the computational sciences." (see D3.2.3) This course is a PATC course, see also the PRACE Training Portal and Events. For participants from public research institutions in PRACE countries, the course fee is sponsored through the PRACE PATC program.
HLRS is also member of the Baden-Württemberg initiative bwHPC-C5. This course is also provided within the framework of the bwHPC-C5 user Support.
Rolf Rabenseifner phone 0711 685 65530, rabenseifner(at)hlrs.de Lucienne Dettki phone 0711 685 63894, dettki(at)hlrs.de
https://www.hlrs.de/training/2021/NLP and PATC web page: https://events.prace-ri.eu/e/HLRS-2021-NLP
https://www.hlrs.de/training/ and https://www.hlrs.de/training/overview/
Back to list
High-Performance Computing Center Stuttgart
Nobelstraße 19, 70569 Stuttgart, Germany
+49 (0) 711 / 685-87 209
A member of the Gauss Centre for Supercomputing, HLRS is one of three German national centers for high-performance computing.
HLRS is a central unit of the University of Stuttgart.