Fourth International Workshop on Visual Performance Analysis (VPA 17)
Denver, Colorado, USA
November 17, 2017
Held in conjunction with SC17: The International Conference on High Performance Computing, Networking, Storage and Analysis, in cooperation with TCHPC: The IEEE Computer Society Technical Consortium on High Performance Computing
Over the last decades an incredible amount of resources has been devoted to building ever more powerful supercomputers. However, exploiting the full capabilities of these machines is becoming exponentially more difficult with each new generation of hardware. To help understand and optimize the behavior of massively parallel simulations the performance analysis community has created a wide range of tools and APIs to collect performance data, such as flop counts, network traffic or cache behavior at the largest scale. However, this success has created a new challenge, as the resulting data is far too large and too complex to be analyzed in a straightforward manner. Therefore, new automatic analysis and visualization approaches must be developed to allow application developers to intuitively understand the multiple, interdependent effects that their algorithmic choices have on the final performance.
This workshop will bring together researchers from the fields of performance analysis and visualization to discuss new approaches of applying visualization and visual analytics techniques to large scale applications.
Workshop Topics
- Scalable displays of performance data
- Data models to enable scalable visualization
- Graph representation of unstructured performance data
- Presentation of high-dimensional data
- Visual correlations between multiple data source
- Human-Computer Interfaces for exploring performance data
- Multi-scale representations of performance data for visual exploration
Previous Workshops
Papers
Call for Papers
We solicit 8-page full papers as well as 4-page short papers that focus on techniques at the intersection of performance analysis and visualization, and either use visualization techniques to display large scale performance data or that develop new visualization or visual analytics methods that help create new insights.
Papers must be submitted in PDF format (readable by Adobe Acrobat Reader 5.0 and higher) and formatted for 8.5” x 11” (U.S. Letter). Submissions are limited to 8 pages in the ACM format, using the sample-sigconf template. The 8-page limit includes figures, tables, and references.
All papers must be submitted through Easychair at:
Dates
Important Dates
- Submission deadline (extended): August 21, 2017 (AoE)
- Notification of acceptance: September 18, 2017 (AoE)
- Camera-ready deadline: October 9, 2017 (AoE)
Program
Technical Program
All times refer to Friday, Nov. 17, 2017
8:30am - 8:35am | ---- Welcome and Introduction ---- |
8:35am - 9:35am |
Keynote Talk: Visual Performance Analysis for Extremely Heterogeneous Systems
|
9:35am - 10:00am |
Paper Talk: Chad Wood, Matthew Larsen, Alfredo Gimenez, Cyrus Harrison, Todd Gamblin and Allen Malony. Projecting Performance Data Over Simulation Geometry Using SOSflow and AlpineThe performance of HPC simulation codes is often tied to their simulated domains; e.g., properties of the input decks, boundaries of the underlying meshes, and parallel decomposition of the simulation space. A variety of research efforts have demonstrated the utility of projecting performance data onto the simulation geometry to enable analysis of these kinds of performance problems. However, current methods to do so are largely ad-hoc and limited in terms of extensibility and scalability. Furthermore, few methods enable this projection online, resulting in large storage and processing requirements for offline analysis. We present a general, extensible, and scalable solution for in-situ (online) visualization of performance data projected onto the underlying geometry of simulation codes. Our solution employs the scalable observation system SOSflow with the in-situ visualization framework ALPINE to automatically extract simulation geometry and stream aggregated performance metrics to respective locations within the geometry at runtime. Our system decouples the resources and mechanisms to collect, aggregate, project, and visualize the resulting data, thus mitigating overhead and enabling online analysis at large scales. Furthermore, our method requires minimal user input and modification of existing code, enabling general and widespread adoption. [PDF]
|
10:00am - 10:30am | ---- Coffee Break ---- |
10:30am - 11:10am |
Panel Discussion: Challenges and the Future of HPC Performance Visualization
Panelists:
|
11:10am - 11:35am |
Paper Talk: Nico Reissmann, Magnus Jahre and Ananya Muddukrishna. Towards Aggregated Grain GraphsGrain graphs simplify OpenMP performance analysis by visualizing performance problems from a fork-join perspective that is familiar to programmers. However, it is tedious to navigate and diagnose problems in large grain graphs with thousands of task and parallel for-loop chunk instances. We present an aggregation method that matches recurring patterns in grain graphs and groups related nodes together, reducing graphs of any size to one root group. The aggregated grain graph is then navigated by progressively uncovering groups and analyzing only those groups that have problems. This enhances productivity by enabling programmers to understand program structure and problems in large grain graphs with less effort than before. [PDF]
|
11:35am - 12:00pm |
Paper Talk: Matthias Diener, Sam White and Laxmikant Kale. Visualizing, measuring, and tuning Adaptive MPI parametersAdaptive MPI (AMPI) is an advanced MPI runtime environment that offers several features over traditional MPI runtimes, which can lead to a better utilization of the underlying hardware platform and therefore higher performance. These features are overdecomposition through virtualization, and load balancing via rank migration. Choosing which of these features to use, and finding the optimal parameters for them is a challenging task however, since different applications and systems may require different options. Furthermore, there is a lack of information about the impact of each option. In this paper, we present a new visualization of AMPI in its companion Projections tool, which depicts the operation of an MPI application and details the impact of the different AMPI features on its resource usage. We show how these visualizations can help to improve the efficiency and execution time of an MPI application. Applying optimizations indicated by the performance analysis to two MPI-based applications results in performance improvements of up 18% from overdecomposition and load balancing. [PDF]
|
---- Closing ---- |
Committees
Steering Committee
Peer-Timo Bremer, Lawrence Livermore National Laboratory
Bernd Mohr, Juelich Supercomputing Center
Valerio Pascucci, University of Utah
Martin Schulz, Lawrence Livermore National Laboratory
Workshop Chairs
Fabian Beck, University of Duisburg-Essen
Abhinav Bhatele, Lawrence Livermore National Laboratory
Judit Gimenez, Barcelona Supercomputing Center
Joshua A. Levine, University of Arizona
Program Committee
Harsh Bhatia, Lawrence Livermore National Laboratory
Holger Brunst, TU Dresden
Alexandru Calotoiu, Technical University Darmstadt
Todd Gamblin, Lawrence Livermore National Laboratory
Marc-Andre Hermanns, Juelich Supercomputing Center
Kevin Huck, University of Oregon
Katherine Isaacs, University of Arizona
Yarden Livnat, University of Utah
Naoya Maruyama, Lawrence Livermore National Laboratory
Bernd Mohr, Juelich Supercomputing Center
Ananya Muddukrishna, KTH Royal Institute of Technology
Matthias Mueller, RWTH Aachen University
Valerio Pascucci, University of Utah
Paul Rosen, University of South Florida
Carlos Scheidegger, University of Arizona
Chad Steed, Oak Ridge National Laboratory