Aug
16
Wed
Using the Roofline Model and Intel Advisor
Aug 16 @ 1:00 pm – 2:00 pm

The IDEAS Productivity project, in partnership with the DOE Computing Facilities of the ALCF, OLCF, and NERSC and the DOE Exascale Computing Project (ECP), is resuming the webinar series on Best Practices for HPC Software Developers, which we began last year.

As part of this series, we will offer one-hour webinars on topics in scientific software development and high-performance computing, approximately once a month. Participation is free and open to the public, but registration will be required for each event.

The next webinar in the series was “Using the Roofline Model and Intel Advisor” with speaker Sam Williams of Lawrence Berkeley National Laboratory.  This webinar took place on August 16, 2017 and you can find a copy of the presentation and the video from the webinar in the Presentation Materials section below.

This webinar began by introducing the Roofline Model and its “Cache-Aware” variant.   Then, they moved to general guidelines and historical approaches to Roofline-based program analysis.  Next, they provided a short discussion of how changes in data locality and arithmetic intensity of two canonical benchmarks visually manifest in the context of these two Roofline formulations.  The speakers then conducted demonstrations of using Intel Advisor and the Roofline model within Intel Advisor.  The first demo was primarily instructive on how to compile, benchmark, and use Advisor. The second demo focused on using variants of a simple benchmark to highlight changes in the Roofline model as well as providing correlation to Advisor’s other capabilities.
Sep
20
Wed
Scalable Node Programming with OpenACC
Sep 20 @ 1:00 pm – 2:15 pm

The DOE Exascale Computing Project (ECP) is pleased to sponsor a webinar presented by Michael Wolfe of NVIDIA titled “Scalable Node Programming with OpenACC” that took place on Wednesday, September 20, at 1:00 pm ET.

Supercomputer nodes are becoming more parallel with each generation.  HPC applications must now utilize parallelism on-node as well as across many nodes for performance.  OpenACC is designed to expose and exploit parallelism, whether that be for a multicore, a manycore, or a GPU-accelerated node.

This webinar discussed:

  • what problems OpenACC addresses;
  • the important elements of OpenACC;
  • how and why OpenACC is being adopted and used in production science and engineering applications, including Gaussian, Fluent, VASP, and several weather and climate codes;
  • current status of OpenACC tools and of the specification;
  • building and using applications with OpenACC for different targets;
  • the OpenACC community and user events;
  • new features requested by users coming with OpenACC 2.6 and beyond.
Nov
8
Wed
TAU Performance System Webinar
Nov 8 @ 1:00 pm – 2:15 pm

The DOE Exascale Computing Project (ECP) is pleased to have sponsored a webinar presented by Sameer Shende of the University of Oregon on the TAU Performance System.  The webinar took place on Wednesday, November 8, at 1:00 pm ET.  You will find the presentation materials including the slides and video below.

Abstract:  The complex nature of HPC platforms and the application development environment, combining multiple languages, programming paradigms, hardware, and compilers, make effective performance engineering a challenging task. To meet the needs of computational scientists in performance engineering their codes, we present a webinar on the TAU Performance System. 

TAU is a powerful profiling and tracing toolkit that covers multiple aspects of performance instrumentation, measurement, and analysis. After describing and  demonstrating how performance data is collected using TAU’s automated instrumentation, the workshop will present ways to analyze the performance data collected and to drill down to find performance bottlenecks. Topics will cover generating performance profiles and traces with memory and system load utilization metrics, I/O, communication, and hardware performance counter data using PAPI, compiler-based instrumentation, callsite-profiling, generating OTF2 traces natively, using TAU for evaluating CUDA and Python based applications, and the advances in TAU that are being designed as part of the ECP Proteas project. The workshop will cover instrumentation of Hybrid MPI and OpenMP codes,  and the use of TAU for profiling Kokkos based applications using the Kokkos Profiling Interface.

Mar
26
Tue
Performance Portability with Kokkos Bootcamp March 2019 @ Oakland City Center Conference Center
Mar 26 @ 8:30 am – Mar 29 @ 12:30 pm

Performance Portability with Kokkos Bootcamp – Registration for this event is now closed.  

March 26-29, 2019

We are pleased to announce that we are hosting the next Performance Portability with Kokkos Bootcamp March 26-29, 2019 at the Oakland City Center Conference Center in Oakland, CA. This workshop is intended to teach new Kokkos users how to get started and to help existing Kokkos users to further improve their codes. The training will cover the minimum required topics to get your application started on using Kokkos, and Kokkos experts will be on hand to help the more advanced users.

What is Kokkos?
Kokkos is a programming model and library for writing performance portable code in C++. It includes abstractions for on-node parallel execution and data layout. These abstractions are mapped at compile time to fit a device’s architecture for best performance. It uses standard C++ in the same spirit as libraries such at Thrust and Thread Building Blocks.

Who should attend?
Anyone who has a C++ application, or would like to create C++ Kokkos kernels that hook onto an application, and would like to have a single source code run well on multiple platforms. We also encourage developers to bring applications that already use Kokkos since Kokkos experts will be available to help with more advanced use cases. Although we strongly suggest teams of two (or more) per application, please do not hesitate to apply if you are a single developer who wants attend this event.

What happens at the event?
We will have Kokkos experts to help you with your application. This event is a tutorial and a playground to experiment with integrating Kokkos with your application and to help optimize existing Kokkos applications.

What happens after the event?
Attendance to this event will help us create a relationship with your team that we hope to continue as you return home to continue your work. We plan to host regular office hours to tend to your teams questions in the initial stages and to help your team continue to make significant progress.

How should I prepare?
After signing up, we will contact you to discuss your application. If you are new to Kokkos, we can help you prepare a kernel for the event. If you have an existing Kokkos application, we would like to understand your needs before the event. We hope that doing this prep work will maximize your time learning from Kokkos experts.

How do I apply?
Registration for this event is now closed.   

If you have any questions, please contact one of the following organizers: