ALCF Developer Sessions

Webinar
ALCF Developer Sessions

 

The monthly ALCF Developer Sessions series is aimed at increasing the dialogue between ALCF users and the developers of leadership-class systems and software. Conducted in an interactive format, attendees are encouraged to bring any questions related to ALCF computing resources. Stay tuned to this webpage for details on upcoming sessions.

2019 Developer Sessions

  • March 2019 
    Quick Start: Using Apache Spark for Large-Scale Data Processing 
    Xiao-Yong Jin, Argonne Leadership Computing Facility, Argonne National Laboratory
    Description: Apache Spark provides a high-level framework for parallel data processing, with built-in support for fault tolerance and data replication. It has become increasingly popular in cloud computing environments and commercial data centers due to its ease of use and integration with existing libraries in Java, Scala, Python, R, and SQL.

  • February 2019 
    Singularity on Theta: How to Build and Scale Contianers at the ALCF  
    Taylor Childers, Argonne Leadership Computing Facility, Argonne National Laboratory
    Description: The ALCF supports containerized software on its Theta and Cooley systems, and plans to support containers on future systems as well. Containers not only provide an easy way to run the same software across systems with different environments, but their use can also lead to improved performance on shared filesystems. This presentation will discuss the performance benefits of using containerized software and provide instruction on how to produce containers.

  • January 2019 
    SENSEI Cross-Platform View of in Situ Analytics 
    Silvio Rizzi and Sergei Shudler, Argonne Leadership Computing Facility, Argonne National Laboratory
    Description: This talk will cover the design and use of SENSEI, a platform for in situ visualization and analysis. In situ techniques eliminate the need to write large simulation state files or other data that prevents applications from scaling to large machines. On the one hand, the SENSEI platform allows developers of simulation applications to instrument their code with the simulation API and gain access to a wide variety of I/O, analysis, and visualization tools. On the other hand, tool developers can plug their tools into SENSEI through the analysis API, thereby allowing these tools to analyze a wide range of simulations.

2018 Developer Sessions

  • November 2018 
    Workflow managment with Balsam [slides] [video]
    Misha Salim, Argonne Leadership Computing Facility, Argonne National Laboratory
    Description: This session provides an overview of Balsam, a workflow manager that automatically packages application runs for parallel execution, provides strong tolerance to task faults, and tracks the workflow status and usage statistics in a user-managed database.

  • October 2018 
    Preparing an application for Hybrid Supercomputing using Cray's Tool Suite [slides] [video]
    John Levesque, Director Cray’s Supercomputing Center of Excellence
    Description: This session covers a computation fluid dynamics application called Leslie3D. Cray's memory analysis tool is used to identify areas where the memory bandwidth is limiting the performance.

  • September 2018 
    Run-to-Run Variability on Theta and Best Practices for Performance Benchmarking [slides] [video]
    Sudheer Chunduri - Performance Engineer, Argonne Leadership Computing Facility, Argonne National Laboratory
    Description: The increasing complexity of HPC systems has introduced new sources of variability, which can contribute to significant differenrces in run-to-run perfromance of applications. 

  • August 2018 
    Boosting Power Efficiency of HPC Applications with GEOPM [slides] [video]
    Jonathan Eastep, Principal Engineer in the Data Center Group at Intel
    Description: This session provides an introduction to the Global Extensible Open Power Manager (GEOPM), an open source runtime framework that is available on the ALCF's Theta system for optimization of application power efficiency and exploration of advanced power management strategies for future systems. 

  • July 2018 
    TensorFlow Optimizations on Modern Intel Architectures [slides] [video]
    Vamsi Sripathi - Senior Software Engineer, Intel and Vikram Saletore - Principal Engineer, Intel
    Description: The session provides an overview of the optimizations implemented by Intel to accelerate the performance of the TensorFlow framework for Intel Xeon and Xeon Phi architectures.

  • June 2018 
    Profiling Python Workloads with Intel VTune Amplifier [slides] [video]
    Paulius Velesko - Intel
    Description: This talk covers efficient profiling techniques that can help to dramatically improve the performance of code by identifying CPU and memory bottlenecks. 

  • May 2018 
    Leveraging Optimized FFT on Intel Xeon Platforms [slides] [video]
    Jeongnim Kim - Data Center Group, Intel
    Description: This session explores the use patterns arising from real applications, the factors that are critical for the performance, and how to use the basic FFT library to maximize the performance on Intel Xeon and Xeon Phi clusters.

  • April 2018
    A Field Guide to KNL Memory Modes and OpenMP Affinity Control [slides] [video]
    Jeff Hammond, System Architect at Intel 
    Description: This session covers different usage models for MCDRAM in Intel Xeon Phi processors (Knights Landing), as well as OpenMP affinity control for Intel Xeon Scalable processors (Skylake). 

  • Februrary 2018
    Analyzing Parallel Program Performance using HPCToolkit  [slides] [video]
    John Mellor-Crummey, Department of Computer Science, Rice University
    Description: This session is focused on HPCToolkit, an integrated suite of multiplatform tools for measuring and analyzing the performance of parallel programs on scalable parallel systems.

  • January 2018 
    Performance Profiling on KNL with Cray Perftools-lite  [slides] [video]
    Heidi Poxon, Cray
    Description: Cray performance tools provide a simple interface that allows users to easily obtain a rich set of information to help identify important performance bottlenecks