|
Scientific Visualization
| Issue Year: | 2014 |
| Quarter: | 4 |
| Volume: | 6 |
| Number: | 4 |
| Pages: | 1-10 |
|
| Article Name: |
APPLICATION OF NVIDIA OPTIX TO CARRY OUT NUMERICAL EXPERIMENTS |
| Authors: |
V.A. Debelov (Russian Federation) |
| |
The paper is recommended by program committee of 24th International Conference on Computer Graphics and Vision GraphiCon’2014. |
| Address: |
V.A. Debelov
debelov@oapmg.sscc.ru
Institute of Computational Mathematics and Mathematical
Geophysics SB RAS, Novosibirsk, Russia |
| Abstract: |
A lot of research programs are implemented in order to carry
out numerical experiments while debugging of an algorithm or obtaining
different data of an investigated computer model. As a rule the data are
visualized by means of scientific visualization tools. In these paper the
author considers reasons to apply Nvidia OptiX as a tool to carry out numerical
experiments using graphical processing units (GPU) GeForce.
Although OptiX is built on CUDA and runs on a graphical unit,
a researcher does not have to know programming elements such as an architecture
of GPU, CUDA, OpenGL, DirectX. By the way, a knowledge of programming for a cluster
– for example, MPI (Message Passing Interface) – is also not needed.
OptiX is considered as an effective tool to control a run of
a loop
i = 0 .. N
j = 0 .. M
k = 0 .. K
f(i, j, k, <additional parameters >),
where f(i, j, k, < additional parameters >)
is an algoritm being debugged or tested computer model.
From the author’s point of view GPU considered can be imagined
as a desktop cluster including multiple processors working in parallel on
shared memory. A researcher can use a PC in parallel at his discretion while
his program runs on GPU.
To justify a suggested approach the author shows with
several tests a speedup rate of an OptiX program in comparison with a
sequential program. |
| Language: |
Russian |
|
|
|
