Technical News from PGI

 

In This Issue | NOV 2015

OpenACC for Multicore CPUs

Performance Portability from GPUs to CPUs with OpenACC

High Performance and Productivity with Unified Memory and OpenACC

Fortran Array Attributes: Pointer and Allocatable, Contiguous and Target

Customize CUDA Fortran Profiling with NVTX

Upcoming Events

HPC China November 9–12, Wuxi, Jiangsu, China

SC15 November 16–19, Austin, Texas

Computing Insights UK 2015 December 8–9, Coventry, UK

HiPC December 16–19, Bengaluru (Bangalore), India

Resources

Download Newest Release

PGInsider Back Issues

OpenACC Website

OpenACC Toolkit

Free PGI for OS X

CUDA Fortran

PGI User Forums

Recent News

PGI Accelerator Compilers Add OpenACC Support for x86 Multicore CPUs

Introducing the NVIDIA OpenACC Toolkit

Next Issue

Combined CPU and GPU Profiling with the new PGPROF

Using CUDA Libraries with OpenACC

Experiences Accelerating a Computation Electromagnets Code with OpenACC

PGI Compilers & Tools
Suite 100
20400 NW Amberwood Drive
Beaverton, OR 97006

Michael Wolfe

OpenACC for Multicore CPUs

Michael Wolfe's
Programming Guide

PGI 15.10 includes support for OpenACC on multicore x86 CPUs. Ccompiling OpenACC programs for parallel execution across all cores of a multicore CPU or multi-socket CPU server offers programmers several benefits including the ability to incrementally parallelize applications for multicore CPUs and GPUs, use of a single uniform parallel programming model across CPUs and GPUs in Fortran, C and C++, and the ability to write scalable OpenACC source code that will compile and run in parallel on NVIDIA GPUs, Radeon GPUs or multicore CPUs. In this article, Michael Wolfe looks at the features and potential implications of PGI's support for OpenACC on multicore. | Continue to the article…

Performance Portability from GPUs to CPUs with OpenACC

Over at the Parallel FORALL blog, Mark Harris takes a deeper look at OpenACC on multicore x86 including an example walk-through and performance comparisons. | Continue to the article…

High Performance and Productivity with Unified Memory and OpenACC

PGI's support for Unified Memory with OpenACC can significantly improve programmer productivity and simplify accelerating applications with GPUs. In this article, we look at using an MPI parallel Lattice Boltzmann Solver to describe a methodology for leveraging Unified Memory to improve productivity when accelerating applications with OpenACC. | Continue to the article…

Fortran Array Attributes: Pointer and Allocatable, Contiguous and Target

This article takes an in-depth look at differences and approaches to optimizing performance of modern Fortran dynamically allocated arrays. It covers such concepts as aliasing, contiguity, stride-1 access and more. | Continue to the article…

Customize CUDA Fortran Profiling with NVTX

Over at the NVIDIA Parallel Forall blog, Massimiliano Fatica takes a look at using the NVIDIA Tools Extension (NVTX) library to annotate profiles of Fortran codes—both CUDA Fortran and OpenACC. | Continue to the article…

Free OpenACC Course