Multicore Tool Demo: Cray2014-09-12T09:33:02+00:00

sponsor-icon-cray

Conference:Multicore Challenge 2014 (click here to see full programme)
Session Type:Multicore Tools Demonstration
Session Title:Introducing Cray Reveal: A tool for accelerating directive development
Presenter/s:Tom Edwards
Designation:Application Analyst Cray Centre of Excellence for ARCHER
Abstract:Most applications running on large High Performance Computer systems, like the Cray XC30, use a distributed programming model like MPI or UPC, but don’t take advantage of shared memory programming possibilities within a multi-core node. To move forward and harness the power of the next generation systems, application developers face significant efforts as they prepare their programs to take advantage of the increasing number of cores and threads available to them. Cray Reveal tool extends Cray’s existing performance measurement, analysis and visualization technology by combining runtime performance statistics and program source code visualization with Cray compiler optimization feedback. Reveal provides the ability to easily navigate through source code to highlighted dependencies or bottlenecks during the optimization phase of program tuning. Using a program library provided by the Cray Compiling Environment (CCE) and performance data collected by the Cray Performance, Measurement, Analysis, and Porting Tools (CPMAT), the user can browse his or her source code to quickly retrieve compiler optimization information and determine which high-level loops could most benefit from OpenMP parallelism.
Reveal provides dependency analysis and variable scoping information for targeted loops and assists the user with the analysis and ultimately with the forming OpenMP directives.

  • Exploiting multi-core through shared memory OpenMP directives
  • Integrating profile data with application source code
  • Automated dependency analysis and scoping of OpenMP for loops
BiographyTom Edwards is an Application Analyst at the Cray Centre of Excellence in Edinburgh having worked for the past five years first on HECToR and now on ARCHER, the UK’s largest Academic Supercomputer. Tom Edwards’ expertise in High Performance Computing and in particular in porting applications on multicore systems has benefited a wide spectrum of scientific users and developers across the environmental, engineering and academic space. Before joining Cray Tom Edwards was at the Met Office working on weather models before completing his PhD in High Performance Computational Magneto-hydrodynamics at the University of Edinburgh.
Follow: