11
February 2017 OPENACC MONTHLY HIGHLIGHTS

OpenACC Highlights - February

  • Upload
    nvidia

  • View
    425

  • Download
    1

Embed Size (px)

Citation preview

Page 1: OpenACC Highlights - February

February 2017

OPENACC MONTHLY HIGHLIGHTS

Page 2: OpenACC Highlights - February

OpenACC is a directives-based programming approach to parallel computing designed for performance and portability on CPUs and GPUs.

Page 3: OpenACC Highlights - February

PERFORMANCE PORTABLE PROGRAMMING MODEL FOR HPC

Powerful

LSDALTONSimulation of molecular energies

CPU GPU

1.0x

11.7x

Big PerformanceCCSD(T) Module, Alanine-3

Titan System: AMD CPU vs Tesla K20X

Spe

edup

vs

CP

U

Simple Portable

main() {

<serial code>

#pragma acc kernels

{

<parallel code>

}

}

Add Simple Compiler Hint ARM

PEZYPOWERSunwayx86 CPU

x86 Xeon PhiNVIDIA GPU

Quicker Development

Lines of Code Modified

<100 Lines

# of Weeks Required

1 Week

Page 4: OpenACC Highlights - February

Learn what’s new with OpenACC this month

Page 5: OpenACC Highlights - February

NEW PGI 2017 Parallel Fortran, C, C++, OpenACC Compilers and Tools

OpenACC 2.5 for Tesla P100 GPUs and multicore CPUs; CUDA 8.0 support

OpenPOWER CPU performance enhancements of 5-10%

Comprehensive C++14 support

GNU 6.2 interoperability

Additional tuned numerical intrinsic functions for Haswell and Broadwell

DOWNLOAD NOW

Page 6: OpenACC Highlights - February

3 STEPS TO MORE SCIENCE TUTORIALSimple tutorial designed to help scientists to start with GPUs using libraries and OpenACC directives.

3 Tutorial Videos:

1. Step 1: Analyze: Use a profiler to identify hot spots

2. Step 2: Parallelize: Expose Parallelism

3. Step 3: Optimize: Manage data movements

Free PGI Community EditionWATCH NOW

Page 7: OpenACC Highlights - February

PAPERSAcceleration of Element-by-Element Kernel in Unstructured Implicit Low-order Finite-element Earthquake Simulation using OpenACC on Pascal GPUs

“OpenACC porting approach enables high performance with relatively low development costs. When comparing eight K computer nodes and eight NVIDIA Pascal P100 GPUs, we achieve 23.1 times speedup for the element-by-element kernel, which leads to 16.7 times speedup for the 3 x 3 block Jacobi preconditional conjugate gradient finite-element solver.”

DOWNLOAD NOW

Page 8: OpenACC Highlights - February

Oak Ridge National Lab on OpenACC being a user-driven standard: Longtime User Requests Added to Next Version of OpenACC

Five Hackathons in 2017 to Train a New Generation on GPU Accelerated Computing

ARTICLES AND BLOG POSTS

READ NOW

READ NOW

Page 9: OpenACC Highlights - February

CALL FOR PAPERS

EVENT DUE DATE LINK

GPU Hackathon at Brookhaven National Lab March 15th, 2017 https://www.bnl.gov/gpuhackathon/.

2nd International Workshop on Performance Portable Programming Models for Accelerators

April 3, 2017 http://www.csm.ornl.gov/workshops/p3ma2017/

Page 10: OpenACC Highlights - February

UPCOMING HACKATHONS

LOCATION DATE LINK

Forschungzentrum Juelich, Juelich, Germany March 6th - 10th https://indico-jsc.fz-juelich.de/event/32/

Brookhaven National Laboratory, Upton, NY June 5th – 9th https://www.bnl.gov/gpuhackathon/.

For complete list of Hackathons visit: https://www.olcf.ornl.gov/training-event/2017-gpu-hackathons/

Page 11: OpenACC Highlights - February

FOR MORE INFORMATIONVisit OpenACC

EXPLORE NOW