2019-10-24 Performance Portability Programming with Kokkos (HKOK1W19)

The many-core revolution can be characterized by increasing thread counts, decreasing memory per thread, and diversity of continually evolving many-core architectures.High performance computing (HPC) applications and libraries must exploit increasingly finer levels of parallelism within their codes to sustain scalability on these devices. A major obstacle to performance portability is the diverse and conflicting set of constraints on memory access patterns across devices. Contemporary portable programming models address many-core parallelism (e.g., OpenMP, OpenACC, OpenCL) but fail to address memory access patterns.

The Kokkos C++ library enables applications and domain libraries to achieve performance portability on diverse many-core architectures by unifying abstractions for both fine-grain data parallelism and memory access patterns. In this tutorial we describe Kokkos’ abstractions, summarize its application programmer interface (API), present performance results for unit-test kernels and mini-applications, and outline an incremental strategy for migrating legacy C++ codes to Kokkos. The Kokkos library is under active research and development to incorporate capabilities from new generations of many-core architectures, and to address a growing list of applications and domain libraries.

This is a tutorial is open and intended for people interested in computational science and programming models. The tutorial is not limited to C++ experts, but also gives a broad introduction into performance portability. PhD students are especially invited.

Hands-on sessions will be done on Amazon Cloud GPU resources.

The workshop is organised by the Leibniz Supercomputing Centre in collaboration with Sandia National Lab and the Technical University of Munich.

Please bring your own laptop with ssh support for the hands-on session!