TotalView Debugging on CORAL Early Access Systems (and more)
May 3, 2017
|Date / Time||This event has already occurred. Page is archival.|
Building 453 Room 1001 (Armadillo Room)
|Title||TotalView debugging CORAL, NVIDIA GPU and OpenMP 4, TotalView/Python integration, TotalView Split DWARF support, and the TotalView Roadmap|
This presentation by the Rogue Wave software development team will demonstrate some of the high impact improvements made as part of projects conducted in collaboration with LLNL. These projects focus on improving CORAL NVIDIA CUDA GPU and OpenMP 4 debugging with TotalView, and TotalView/Python debugging integration.
Learn some tips and techniques for debugging NVIDIA CUDA GPU and IBM XL OpenMP 4 codes running in MPI jobs on the CORAL EA systems. We’ll present the TotalView CUDA GPU debugging model, basic GPU debugging (start up, breakpoint planting, CUDA context selection, CUDA logical/hardware focus navigation), and how to enable CUDA-MEMCHECK and TotalView memory debugging. Come hear about the TotalView/Python debugging integration, which allows debugging mixed Python interpreted and compiled-language (C/C++/Fortran) programs in the same TotalView debugging session.
In addition, learn about recent work to support GNU DebugFission "Split DWARF" in TotalView, along with a brief Roadmap update.
Finally, learn more about the support activities of TotalView specialist Peter Thompson, the LLNL Dedicated Technical Account Manager (DTAM) at firstname.lastname@example.org.
|Presenters||John DelSignore, Bill Burns, Marty Bakal and Peter Thompson from Rogue Wave Software|
|What to Bring||This is a presentation-only workshop - no hands-on. Attendees do not need to bring anything.|
|Level/Prerequisites||Introductory level. No prerequisites, though an existing knowledge of using TotalView would be beneficial.|
No registration is required
|Questions?||Please call or send e-mail to Dong Ahn (925-422-1939 / email@example.com).|