Advanced OpenACC Profiling - Lecture 1

Explore advanced OpenACC profiling techniques in this first lecture of the Advanced OpenACC Course from May 2016. Dive into three ways to accelerate applications, learn about OpenACC profiling tools, and discover PGPROF, the OpenACC CPU and GPU profiler. Examine the exploration and production workflow in seismic data processing, including data acquisition, imaging techniques, and model verification. Analyze the pseudo-code for Kirchhoff migration, assess baseline CPU profiles, and learn parallelization strategies. Understand how Unified Memory improves productivity and its integration with OpenACC. Explore optimization techniques, including multi-core, OMP, and pthread implementations, and compare deployment results. Gain access to homework assignments and learn how to set up Seismic Unix for practical application.

Intro
ADVANCED PROFILING OF OPENACC CODE
3 WAYS TO ACCELERATE APPLICATIONS
OPENACC PROFILING TOOLS
PGPROF: OPENACC CPU AND GPU PROFILER Available with the latest OpenACC Toolkit
EXPLORATION & PRODUCTION WORKFLOW
ACQUIRE SEISMIC DATA
REAL ROCK VS SEISMIC REFLECTION Imaging, signal processing, filtering, ray tracing
USE KNOWN MODEL Verify accuracy of imaging algorithms
PROCESS SEISMIC DATA Open Source Seismic Unix Package
PSEUDO-CODE FOR KIRCHHOFF MIGRATION
ASSESS BASELINE CPU PROFILE
PARALLELIZE
UNIFIED MEMORY IMPROVES PRODUCTIVITY
OPENACC AND UNIFIED MEMORY
OPTIMIZATION How about Multi-Core / OMP / pthread!
DEPLOY How do the results compare?
ACCESS TO HOMEWORK
SETUP SEISMIC UNIX