See how other developers are solving their computational problems with OpenACC GPU directives
and the application performance speed-ups they achieved.
Get started by registering for the free 30-day trial license.
Understanding the Origins of the Universe
Panos Labropoulos, University of Groningen and ASTRON
Speed-up: 5.6x in 5 days
My main interest is the calibration and imaging using data from radio arrays and in particular LOFAR (www.lofar.org
). Such an array consists of tens to hundreds of thousands on relatively simple dipole antenna elements that collect low frequency astronomical signals. With directives and GPUs, we were able to accelerate the code by nearly 6x in 5 days. This speed-up is significant because it allows us to process this enormous amount of data (using many nodes, each with a couple GPUs), in a reasonable amount of time. It will save us years of work.
Real Time Object Detection in Images Sequences
Global Manufacturer of Navigation Systems
Speed-up 5x in 7 days
My application does scene object detection in image sequences using advanced imaging algorithms. I was able to achieve 5x speed-up of my application. Using PGI directives is quite easy, especially compared to CPU threads creation or writing CUDA kernels manually. The most important thing is avoiding big restructuring of existing code, which is risky for production applications.
I've convinced my company to buy processing computers with two more GPUs due to my work.
Understanding Micromagnetics with GPUs
Prof. M Amin Kayali, University of Houston
Speed-up: > 20x in less than 2 days
In micromagnetics, the calculation of the long range magnetostatic interaction (dipole-dipole interaction) is computationally very expensive and hence it limits our ability to effectively study large systems in details. I have written micromagnetic codes (written in Fortran 90) to study the properties of two and three dimensional magnetic systems. The PGI directives approach enabled me to port my existing code with ease to perform my computations on the GPU which resulted in a significant speedup (more than 20 times) of the computation.
Read more stories
Simulating Solvents on Proteins
Bharat Medasani, University of Texas at San Antonio
Speed-up: 5x in 1 day
I am developing a code to include the effects of solvents on biomolecules. The code development is in Matlab & Fortran. see more details
We are porting the codes to GPU. Accelerator directives are the first step in the process of porting. By applying directives, I accelerated my code by 5x in a single day.
Phytoplankton Prediction Over Time
Prof. Kerry Black, University of Melbourne
Speed-up: > 65x in 2 days
In marine biology and oceanography, processes are based on particles -- tiny fish, water molecules, phytoplankton. Computers, however, have been too slow to deal with the micro-scale. see more details
Now using the PGI compiler and Tesla GPU, science can duplicate the particles of nature more closely. Our project is the world's first combined particle-based model that deals with the currents, primary production, zooplankton and fish larvae. The goal is to understand the complex reasons behind the boom and bust cycles of snapper fish in Victoria's marine food bowl of Port Phillip Bay. The research to help feed future populations is funded by the Australian Research Council.
Finding Oil in the Largest Reservoirs in the World
Global Oil Company
Speed-up: 3x in 7 days
Simulating the largest petroleum reservoirs in the world at high resolution requires computational models consisting of hundreds of millions of cells see more details
and this implies solving systems of billions of equations iteratively. Our work investigating GPUs to accelerate the solution of these equations reached speedups of an order of magnitude by using PGI Fortran.
High Performance Geostatistical Simulator of Oilfields
Prof. Arthur Yuldashev, Ufa State Aviation Technical University
Speed-up: 7x in less than 4 Weeks
We are working on high-performance geostatistical simulator that is intended for generation of stochastic geological models of oilfield reservoirs conditioned by borehole data. see more details
With the help of PGI Accelerator C compiler we have accelerated several hotspots of the multi-threaded (OpenMP) version of our simulator including covariance matrix construction, conditional stochastic modeling and FFT.
Consequently we achieved 25x speed-up on modern hybrid servers when running on 1 CPU core + 1 Tesla GPU comparing to execution time on1 CPU core. Moreover we got almost linear performance boost when increasing number of GPUs utilized by geostatistical simulator.
Submit Your GPU Directives Story
Tell your story and show your friends how you accelerated your science with directives.