OpenCL

From UFRC
Revision as of 19:28, 11 March 2013 by Prescott (talk | contribs)
Jump to navigation Jump to search

Description

opencl website  

OpenCL™ (Open Computing Language) is a low-level API for heterogeneous computing that runs on GPUs. Using the OpenCL API, developers can launch compute kernels written using a limited subset of the C programming language on a GPU.

Required Modules

Serial

  • opencl

System Variables

  • HPC_{{#uppercase:opencl}}_DIR
  • HPC_{{#uppercase:opencl}}_INC
  • HPC_{{#uppercase:opencl}}_LIB

Additional Information

The following example shows how to build a sample code with OpenCL.

[prescott@r11a-s17 ~]$ mkdir opencl
[prescott@r11a-s17 ~]$ cd opencl
[prescott@r11a-s17 opencl]$ module load opencl
[prescott@r11a-s17 opencl]$ wget -q https://raw.github.com/smistad/OpenCL-Getting-Started/master/main.c
[prescott@r11a-s17 opencl]$ wget -q https://raw.github.com/smistad/OpenCL-Getting-Started/master/vector_add_kernel.cl
[prescott@r11a-s17 opencl]$ gcc -c -I$HPC_OPENCL_INC main.c -o main.o
[prescott@r11a-s17 opencl]$ gcc main.o -o myopenclprog -l OpenCL

PBS Script Examples

See the OpenCL_PBS page for opencl PBS script examples.

Usage Policy

OpenCL is a trademark of Apple Inc., used under license by Khronos.


Installation

The OpenCL implementation installed on the cluster is the one included with the NVIDIA CUDA Toolkit.