BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//CERN//INDICO//EN
BEGIN:VEVENT
SUMMARY:HPC Workshop "Elements of Performance Engineering"\, by Dr. Georg 
 Hager
DTSTART:20260317T080000Z
DTEND:20260317T160000Z
DTSTAMP:20260609T060000Z
UID:indico-event-1305@indico.hiskp.uni-bonn.de
CONTACT:training@hpc.uni-bonn.de
DESCRIPTION:Speakers: Georg Hager\n\nDescription\nThis course covers perfo
 rmance engineering approaches on the compute node level. It is a shortened
  version of an originally 3-day-course.\nEven application developers who a
 re fluent in OpenMP and MPI often lack a good grasp of how much performanc
 e could at best be achieved by their code. This is because parallelism tak
 es 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. Thi
 s course conveys the required knowledge to develop a thorough understandin
 g of the interactions between software and hardware. This process must sta
 rt at the core\, socket\, and node level\, where the code gets executed th
 at does the actual computational work. We introduce the basic architectura
 l features and bottlenecks of modern processors and compute nodes. Pipelin
 ing\, SIMD\, superscalarity\, caches\, memory interfaces\, ccNUMA\, etc.\,
  are covered. A cornerstone of node-level performance analysis is the Roof
 line model\, which is introduced in due detail and applied to various exam
 ples from computational science. We also show how simple software tools ca
 n be used to acquire knowledge about the system\, run code in a reproducib
 le way\, and validate hypotheses about resource consumption. Finally\, onc
 e the architectural requirements of a code are understood and correlated w
 ith performance measurements\, the potential benefit of code changes can o
 ften be predicted\, replacing hope-for-the-best optimizations by a scienti
 fic process.\nNote also our hackathon from Dr. Hager on the following day 
 (March 18):\n"Bring-Your-Own-Code Hackathon"\, with special focus on perfo
 rmance measurement and optimization using LIKWID https://indico.hiskp.uni
 -bonn.de/event/1306/\n \nWhen and where is the workshop held?\nThe worksh
 op is held in-person (no online participation possible) at the Hochschulre
 chenzentrum\, Wegelerstr. 6. \nIt takes place from 9:00 to 17:00\, on Thu
 rsday\, March 17\, 2026\nYou need to register to participate.\nDetailed sc
 hedule:\n\n9:00-10:30 Part 1\n10:30-11:00 Break\n11:00-12:30 Part 2\n12:30
 -13:30 Lunch break\n13:30-15:00 Part 3\n15:00-15:30 Break\n15:30-17:00 Par
 t 4\n\nWho can register\, and how?\nThe workshop is open to both Uni Bonn 
 members (employees and students) and interested guests from other organiza
 tions. However\, if there are too many registrations\, Uni Bonn members ha
 ve priority.\nWhen you register\, you will be put on a waiting list. You w
 ill be notified by March 10 whether you were accepted.\nPrerequisites and 
 equipment\nCaution: This is not a beginner workshop! Basic knowledge of th
 e following topics is recommended:\n\nC programming and compilation\nBasic
  use of an HPC cluster\, particularly the Linux console\, SSH and SLURM\n\
 n \nYou do not need to bring anything. You will be using the course room 
 PCs and the supercomputer Marvin.\nYou do not need to have a Marvin accoun
 t prior to the course\, you will get temporary access.\n\nhttps://indico.h
 iskp.uni-bonn.de/event/1305/
URL:https://indico.hiskp.uni-bonn.de/event/1305/
END:VEVENT
END:VCALENDAR