How to have OpenCL and gcc in same project with different targets?

[devent]

Hi. I’m trying to learn OpenCL. I have a background in C/C++ and Java programming. I want to compile my project first just in gcc and test out the functions and then later I want to use the same code in OpenCL. It should be possible to have both C and OpenCL code in the same project since OpenCL uses C as the language. I can then specify with a precompiler flag that I want to use gcc or OpenCL kernel to run. This should be similar how I used to create both Linux and Windows executable from the same code base.

Something like

// Use gcc
#ifndef USE_OPENCI
#include <math.h>
#include <stdlib.h>
#include <CL/cl_platform.h>
#else
// Use OpenCI kernel
#endif

float4 do_vector_stuff(double* data) {
}

Is this actually possible? Is it even what I should be doing? Again, I just want to test my code with gcc and then at a later stage to run it in OpenCL.

[hterrolle]

hi,

the question is not very clear.

you want the same project code for two different OS. It wont work. only the OpenCL code “Kernel” may be portable, it depends ;)).

[devent]

Thank you for the reply. I actually already finished what I asked here. Using the clang compiler and the float2, float3, etc. extensions it was quite easy. I used a precompiler flag USE_OPENCL to have the (-almost- it’s actually the same code because the opencl_utils.h will act as a bridge) same code for both C and OpenCL. Now I can use the code base and test it with google-test using the clang compiler and then use OpenCL to run the kernel.

You can check out the code here:
https://github.com/devent/anl-opencl/tree/develop/anlopencl/src/main/cpp
And the tests:
https://github.com/devent/anl-opencl/tree/develop/anlopencl/src/test/cpp
This header will bridge the code between clang and OpenCL:
https://github.com/devent/anl-opencl/blob/develop/anlopencl/src/main/cpp/opencl_utils.h