X86, ARM, GPUs!

Today's Programming Environment for the Complex, Many-choice Platforms in HPC


To be held as a full-day workshop at ISC 2019

June 20, 2019

 

A key component for a fast and balanced supercomputer is the programming environment. The scale and complexity of current and future high-end systems brings a new set of challenges for application developers. The technology changes in the supercomputing industry, coupled with the emergence of new processor and accelerator architectures, force computational scientists to face new critical system characteristics that will significantly impact the performance and scalability of applications. With the next generation of supercomputers, application developers need sophisticated compilers, tools, libraries, and adaptive runtime systems that can help maximize programmability with low porting and tuning efforts, while not losing sight of performance portability across a wide range of processors and architectures. Application developers need programming environments that can address, as well as hide, the issues of scale and complexity of high-end HPC systems. This workshop will focus on the programming environments that target the emerging processors and accelerators for the exascale class of supercomputers. It will have presentations from industry, as well as end-user experiences with the available programming environment. The keynote will discuss the programming environment needs for the US DOE CORAL systems and the new features in OpenMP 5.0 that are most expected to impact its use.

09:00-09:30  The Cray Programming Environment for the Exascale Class of Supercomputers  Luiz DeRose / Cray Inc.
09:30-10:00  One year on: User Experiences with Isambard, the World’s First Production Arm-based Supercomputer  Simon McIntosh-Smith / University of Bristol
10:00-11:00  Keynote: OpenMP and CORAL  Bronis R. de Supinski / LLNL
11:00-11:30  Break
11:30-12:00  The Radeon Open Compute Platform for HPC and Machine Learning  Michael Schulte / AMD
12:00-12:30  Cross Platform Performance Engineering for Python  Patrick Wohlschlegel / Arm
12:30-13:00  Intel’s Approach to Simplifying the Increasingly Complex Programming Environment for Extreme-Scale Computing  James Cownie / Intel
13:00-14:00  Lunch 
14:00-14:30  Task and PGAS: Programming Model for Manycore Extreme-Scale Computing  Mitsuhisa Sato / Riken
14:30-15:00  Using OpenMP 4.5: Application Experiences  Oscar Hernandez / ORNL
15:00-15:30  Kokkos and the Future of C++: Parallel Programming for Heterogeneous Architectures  Christian Trott / SNL
15:30-16:00  RAJA: Enabling Portability for Large-Scale C++ Scientific Applications  Jeff Hittinger / LLNL
16:00-16:30  Break
16:30-17:00  Scaling Forward on GPU-based Systems with Greater Effectiveness  C.J. Newburn / NVIDIA
17:00-17:30  Tips and Tricks for Improving Vectorization for Arm  John Linford / Arm
17:30-18:00  Profiling Applications for Exascale Systems Using Cray Tools  Heidi Poxon / Cray Inc.

Keynote Speaker

Dr. Bronis R. de Supinski, LLNL

OpenMP and CORAL


Abstract:

This talk will provide a detailed look at the design of Sierra, which is currently number 2 on the Top500, as well as early application and benchmark results on it. It will also discuss the new features in OpenMP 5.0 that are most expected to impact its use. Finally, it will explore how our experiences with Sierra are shaping El Capitan, which will be procured through the CORAL2 RFP to replace Sierra in 2023.