Registration E3 - Osquarsbacke 14 - KTH Main Campus

Welcome to PDC and the Summer School

• Erwin Laure (PDC-HPC)

High-Performance Computer Architecture

• ERIK HAGERSTEN

High-Performance Computer Architecture

• Erik Hagersten

High-Performance Computer Architecture

• Erik Hagersten

Concepts and Algorithms for Scientific Computing

• Björn Engqvist

Concepts and Algorithms for Scientific Computing

• Björn ENGQUIST

Interconnection Networks

• Michael Schliephake (KTH)

Introduction to PDC's Environment

• Radovan Bast (PDC - KTH)

Shared memory programming, OpenMP

• Mats BRORSSON

Shared memory programming, OpenMP

• Mats BRORSSON

Shared memory programming, OpenMP

• Mats BRORSSON

Lab: Programming Exercises on OpenMP

• Stefano MARKIDIS

Lab: Programming Exercises on OpenMP

• Stefano MARKIDIS

Shared memory programming, OpenMP

• Mats BRORSSON

Shared memory programming, OpenMP

• Mats BRORSSON

Lab: OpenMP Advanced Project

• Stefano MARKIDIS

Lab: OpenMP Advanced Project

• Stefano MARKIDIS

GPU Architectures for Non-Graphics People

• David BLACK-SCHAFFER

GPUs: The Hype, The Reality, and The Future

• David BLACK-SCHAFFER

Introduction to CUDA

• Michael Schliephake (KTH)

Introduction to CUDA

• Michael Schliephake (KTH)

Lab: CUDA

• Michael Schliephake (KTH)

Lab: CUDA

• Michael Schliephake (KTH)

Introduction to CUDA

• Michael Schliephake (KTH)

MPI: Introduction, Basic Concepts, Point-to-Point Communication

• Erwin Laure (PDC-HPC)

Lab: CUDA

Lab: MPI Part 1/CUDA

MPI - Point-to-Point Communication

• Erwin Laure (PDC-HPC)

MPI - Collective Communication

• Erwin Laure (PDC-HPC)

MPI - Intermediate MPI

• Erwin Laure (PDC-HPC)

MPI: Intermediate MPI

• Erwin Laure (PDC-HPC)

Lab: MPI Part 2

Performance Engineering

• Pekka Manninen

Performance Engineering

• Pekka Manninen

Parallel Performance Engineering

• Pekka Manninen

Lab: Parallel Performance Engineering

• Pekka Manninen

Lab: Performance Engineering

• Pekka Manninen

Wrap up: Performance Engineering

• Pekka Manninen

MPI: Advanced Concepts

• Erwin Laure (PDC-HPC)

MPI: Advanced Concepts

• Erwin Laure (PDC-HPC)

Gyrokinetic modelling for fusion plasmas

• Pär Strand

Thermonuclear fusion energy is one of the most attractive future energy sources because of the widespread and abundant distribution and low cost of its fuel supplies, and because of its inherent safety and environmental features. A positive energy balance in a magnetic fusion device requires heating a Deuterium and Tritium mixture to around 100 million degrees and maintaining the hot plasma in magnetic confinement for sufficiently long time. Understanding and controlling the processes and instabilities inherent in a fusion energy grade plasma is therefore key to achieving a sustained nuclear fusion. Understanding the interaction between the macroscopic instabilities and the microscopic plasma perturbations is always a challenging issue due to the intrinsic multi-scale and nonlinear nature of the problem. It is, on the other hand, an almost unavoidable step towards realistic simulations of high temperature fusion plasmas. Microscopic instabilities have gained much attention over the last few decades mainly because they tend to define size of a viable fusion reactor. Hence a lot can be gained from a detailed understanding of the underlying physics - unfortunately this is a very hard problem and has only recently become tractable due to the rapid development of computing systems and more importantly theoretical developments. In this presentation: a) A brief general introduction to fusion b) A discussion on the theoretical model(s9 and technical implementations c) Overview of results from the Chalmers gropu using the GENE code

Lab: MPI Part 3

Future Programming Languages

• Stefano Markidis (PDC - KTH)

Future Programming Languages

• Stefano Markidis (PDC - KTH)

Abstractions for processing large data sets

• Jonas Yngvesson (Google Inc.)

At Google there is often a need to process very large data sets across many machines. Building efficient parallel processing programs is not trivial and to avoid the overhead of each engineer reinventing the wheel, Google has created several programming models to abstract away the complexities of parallelism and have the programmer concentrate on the core of his processing problem instead. I will talk mainly about the MapReduce model, the conceptual model the programmer get to work with and how it is implemented under the hood. I will also mention a few other models in use at Google, like Pregel, Dremel and Flume.

Open Lab and Project Work