Problem with structures

Accelerated Computing CUDA CUDA Programming and Performance
1

Hi guys,

I’m wondering how to handle arrays of structures correctly in opencl. I have an Triangle structure which contains the three vertices of a triangle. These variables are also structures of Point3D. So my structures look like this:

typedef struct Point3D {

	float x, y, z;

} Point3D;

typedef struct Triangle {

	Point3D	v0, v1, v2;   	// vertices 

} Triangle;

If I use it in the kernel like this:

__kernel testKernel(__global Triangle* triangles) {

   Triangle t = triangles[0];

}

the compilation fails and points to a line where no error appears. But if I just get one of the vertices and assign it like

__kernel testKernel(__global Triangle* triangles) {

   Point3D p = triangles[0].v0;

}

the compilations doesn’t throw an error. What am I doing wrong?

Thanks in advance

Daniel

2

Doesn’t anyone here work with structures??? Strange, how do you organize your data?

3

It looks like you forgot to specify the void return type in your kernel definition. It should be:

__kernel void testKernel(__global Triangle* triangles) {

   Triangle t = triangles[0];

}

Recent versions of the compiler should give an error message about this.