Vampir 9.10

Vampir meets SC '13

65_vampir_news_sc-13.pdf 18
Nov
2013

Welcome to the autumn edition of Vampir News, designed to keep you informed about the latest developments in our performance analysis environment. This includes the visualization and analysis framework Vampir and the performance monitor Score-P.

Vampir

Vampir 8.2 improves the overall usability and stability of the tool chain and is recommended to all Vampir and VampirServer users. It incorporates the latest developments of the collaborative performance monitor Score-P. This release provides the following updates for Vampir and VampirServer:

  • Added D-Bus remote interfaces for partial trace file loading and counter selection.
  • Improved representation of remote memory access (RMA) operations e.g., originating from MPI one-sided operations or PGAS communication operations from GASPI or OpenSHMEM.
  • New ways of combining different performance metrics with arithmetic operations for construc- ting self-defined metrics.
  • Enhancements for the interactive usage of VampirServer in the context of batch systems.
  • Simplified the usage of Vampir-Proxy, which allows fast and easy remote connection setups.

For more details visit http://www.vampir.eu.

Score-P

Score-P is the primary code instrumentation and run-time measurement framework for Vampir 8, and also works natively with Scalasca, TAU, and Periscope. The latest version 1.3 provides a number of improvements and new features:

  • Support for the directive-based accelerator programming standards HMPP (Hybrid Multi- core Parallel Programming) as inspired by CAPS and OmpSs by BSC was added.
  • A new and generic event model for remote memory access (RMA) operations in Score-P and OTF2 to support the trace-based performance analysis of PGAS parallelization models. Actual support for GASPI and OpenSHMEM is underway.
  • A new and generic threading event model in OTF2 covering fork-join threading models such as OpenMP or the create-wait threading models such as POSIX Threads.
  • Improved support for instrumenting and monitoring of CUDA applications.
  • The new metric plugin interface opens up a broad range of new sources of measurement data for Score-P. Simple plugins allow to record many kinds of metrics over time, for example plattform specific hardware performance counters, internals from the OS kernel, or power measurements from external power meters.

For more information and downloads visit http://www.score-p.org.

Download news PDF