OpenCL info and testing program


ma_open_cl_info – OpenCL info and testing program


ma_open_cl_info [-h|--help] [-n] [-g] [-p <N>]


Allows the user to test the avaiability of OpenCL computation with GPUs and the CPU. The program is intended to give a rough estimate on the speed. Note that CPU and GPU timings are not comparable due to differences in implementation (see below for the details).


This program multiplies two random float/single-matrices A and B. A and B are nxn-matrices where n is the “problem size”. In order to caluclate the result, the OpenCL device needs to provide three times the memory necessary to hold one matrix (A, B and result = 3 Matrices)

To be able to compute a comparison-result on CPU we not simply calculate A ×\times B, but instead (A ×\times B)64. This way we can utilize exponentiation by squaring on the CPU. In order to run longer on the GPU (for more reliable measures), we do not use exponentiation by squaring on the OpenCL device.


-h Displays help screen.
-n Disables result verification (no CPU precalculation).
-g Requires OpenCL to be run on GPUs only. (Does not imply -n)
-p Configure problem size to be N, default 2048


Typical invocation on a laptop with Intel GPU and NVidia GPU available:

$ ./ma_open_cl_info -p 1024 OpenCL info and testing program 1.0.0, Copyright (c) 2016

Info: Changed problem size to 1024
System information
Platform 0: 1 devices.
  Device 0:
    Vendor:        Intel
    Name:          Intel(R) HD Graphics Haswell Ultrabook GT3 Mobile
    Compute Units: 40
    Global Memory: 2048 MiB
    Local Memory:  64 KiB
Platform 1: 1 devices.
  Device 0:
    Vendor:        NVIDIA
    Name:          NVD7
    Compute Units: 2
    Global Memory: 1048576 MiB
    Local Memory:  48 KiB

Initializing Tests (asz=4 MiB) ... talloc=0.0 trnd=0.0 tcalc=14.0 tS=14.0
Platform 0
  Device 0
    Comparing results... finished
    tinit=0.0, tmem=0.0, tcalc=32.0, tmem2=0.0, tcmp=0.0, tS/OCL=32.0, tS=32.0
Platform 1
  Device 0
Failed to run OpenCL statement, line 530, error code -11 / CL_BUILD_PROGRAM_FAILURE.
Failed to compile program: invalid source

Note that despite the fact that the NVIDIA GPU is reported, it fails to execute the OpenCL (the error message does not seem to be overly helpful). This is likely due to some missing parts of the proprietary NVIDIA driver which might be necessary to utilize that GPU. In the example, the tS times are relevant. Here, they were 14.0 sec for the initialization and 32.0 sec for the actual computation.


Compile the tool with ant or build the package with ant package. Dependencies for building OpenCL applications are often hard to satisfy. Check bdep in build.xml for a list of Debian packages which seemed to be sufficient for compilation at the time of this writing.

Bugs Website 5 (1.0.0) – no Flash, no JavaScript, no Webfont, no Copy Protection, no Mobile First. No bullshit. No GUI needed. Works with any browser.

Created: 2020/02/17 12:43:43 | Revised: 2020/02/17 23:58:50 | Tags: opencl, compute, c | Version: 1.0.0 | SRC (Pandoc MD) | GPL

Copyright (c) 2016, 2020 For further info send an e-mail to

This program is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version.
This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details.
You should have received a copy of the GNU General Public License along with this program. If not, see <>.