Summer School on Hardware-aware Scientific Computing

Date and Place

October 4-15, 2021

Interdisciplinary Center for Scientific Computing (IWR), Heidelberg University (UHD), Germany and Department of Computational and Data Sciences (CDS) at Indian Institute of Science Bangalore (IISc).

The school will be a hybrid event with live sessions at both places and streaming to offsite participants.

This summer school is part of our HASC grant in the Indo-german Partnership programme that we have together with the Department of Computational and Data Sciences (CDS) at Indian Institute of Science Bangalore (IISc).


Bring together Ph.D. candidates from Mathematics, Computer Science and Scientific Computing to study the interplay of efficient numerical algorithms with modern computer hardware.


First week October 4-8 is filled with lectures covering all aspects from hardware, programming models and algorithms to applications.

Second week October 11-15 features small projects lead by a pair of supervisors from India and Germany.


  • Aspects of modern processor architectures
  • Programming models and accelerator programming
  • Scalable methods for solving partial differential equations (PDEs)
  • Optimal control of PDEs
  • Large-scale Bayesian interference and data assimilation
  • Energy-aware numerical methods


  • Konduri Aditya (CDS, IISc)
  • Peter Bastian (IWR, UHD)
  • Praveen Chandrashekar (TIFR CAM)
  • Sashikumaar Ganesan (CDS, IISc)
  • Roland Herzog (IWR, UHD)
  • Vincent Heuveline (IWR, UHD)
  • Guido Kanschat (IWR, UHD)
  • Phani Motamarri (CDS, IISc)
  • Dirk Pleiter (KTH, Stockholm)
  • Robert Scheichl (IAM/IWR, UHD)
  • Deepak Subramani (CDS, IISc)
  • Jakob Zech (IWR, UHD)


The official registration form is closed now. Late registrations are possible, but with a restriction on in person participation. Please write an email to

Lecture Schedule October 4-8

Monday, October 4, 2021 Tuesday, October 5, 2021 Wednesday, October 6, 2021 Thursday, October 7, 2021 Friday, October 8, 2021
Time India Time Germany
11:30 - 13:00 8:00 - 9:30 Lecture 1 Peter Bastian Dirk Pleiter Konduri Aditya Dirk Pleiter Deepak Subramani
Introduction & Programming Models Performance engineering on modern processor architectures A scalable asynchronous computing method for solving PDEs at extreme scale Introduction to accelerated HPC architectures Bayesian Data Assimilation, Inverse Problems and Digital Twins
13:30 - 15:00 10:00 - 11:30 Lecture 2 Sashikumaar Ganesan Peter Bastian Roland Herzog Vincent Heuveline Jakob Zech
AI & ML in Scientific Computing: A Boon or a Bane? Multilevel Spectral Domain Decomposition Methods Mesh-independent algorithms for optimal control of partial differential equations Energy-aware numerical methods for CFD Large-Scale Bayesian Inference: Sparse-grid and Transport methods
16:00 - 17:30 12:30 - 14:00 Lecture 3 Deepak Subramani Sashikumaar Ganesan Robert Scheichl Praveen Chandrashekar Phani Motamarri
GPU accelerated Optimal Path Planning of Marine Robots GPU-Accelerated Algebraic Multigrid Implementations Generalised Multiscale FEs: Efficient Offline/Online HPC Implementation Talk Supplement Lax-Wendroff flux reconstruction method for convection-dominated problems Large-scale nonlinear eigenvalue problems on heterogenous parallel architectures
18:00 - 19:30 14:30 - 16:00 Lecture 4 Guido Kanschat 17:00 (german time) Joint Dinner Robert Scheichl
Efficient block smoothers in multigrid Large-Scale Bayesian Inference: Efficient Sampling-Based Approaches

Mini Projects October 11-15

In the second week the following projects will be organized:

  1. Vectorization and parallelization of a matrix multiplication kernel on AMD EpycTM 7713
  2. Performance analysis of the CFD code HiFlow3
  3. Performance of a tensor train approximation code
  4. Uncertainty quantification with the code MUQ


This summer school is funded by German Academic Exchange Service (DAAD) and University Grants Commission (UGC) as part of the Indo-German Partnership Programme.

Impressum  |  Haftungsausschluss  |  Datenschutzerkl√§rung  |  generated with Hugo v0.80.0