Attend an informative office hour session introducing OpenACC, a parallel computing standard for simplifying GPU programming. Learn about the differences between CUDA, OpenMP, and OpenACC, supported operating systems, and libraries. Explore topics such as multiple GPU usage, Matlab integration, kernel generation, and available resources for assistance. Gain insights into compiler options, memory types, and comparisons with other parallel computing technologies. Discover OpenACC's capabilities in handling multi-dimensional arrays, gang and vector parallelism, synchronization, and zero-copy operations. Address questions about OpenACC's effectiveness, compatibility, and support for various accelerators, helping you determine whether to use CUDA or OpenACC for your parallel computing needs.
Overview
Syllabus
Introduction
Justin Lutem
CUDA vs OpenACC
OpenMP vs OpenACC
Supported OS
PGI
Libraries
Multiple GPUs
Matlab with Parallel Toolbox
How many dimensions
Two applications run in parallel
How many kernels can I generate
Where can you find help
Stack Overflow
Parallel for All
GPU Technology Conference
Free Fortran Compilers
STL Vectors
Shared or Distributed Memory
Xeon Phi
Can we use PGI
Does the data region need to go outside of the Colonels region
OpenACC compiler for GCC
Thank you
Nvidia GPU Toolkit
Does OpenACC support a 4 dimensional array
Does OpenACC support multiple dimension array
Find Questions
Gangs Vectors
Inner Loop Vectors
OpenACC vs CUDA
Learning is OpenACC easier
Synchronization
Manuel Gang Work
Zero Copy
Does OpenACC support Zero Copy
Is FFT1 optimized for OpenACC
How effective is OpenACC
Async claws
Compatibility
Does OpenACC automatically use multiple GPUs
What accelerators are supported by OpenACC
CUDA or OpenACC
Questions
Taught by
NVIDIA Developer
