OpenCL – adding `float4` to `float4` array does nothing

Question

I have an OpenGL VBO containing cl_float4 vertices and I'm trying to update the VBO via OpenCL (I'm rendering the VBO contents as GL_POINTS). I pass cl_mem object representing the VBO as a kernel argument (the buffer is set as CL_MEM_READ_WRITE).

Unfortunately I cannot update one vertex's float4 data at once.

Following snippet doesn't work (i.e. the rendered points doesn't move):

__kernel void update(__global float4* particle_positions)
{
    int gid = get_global_id(0);

    particle_positions[gid] += float4(0.1, 0.1, 0.1, 0.0);
}

Following snippet does work (i.e. the rendered points move):

__kernel void update(__global float4* particle_positions)
{ 
    int gid = get_global_id(0);

    particle_positions[gid].x += 0.1;
    particle_positions[gid].y += 0.1;
    particle_positions[gid].z += 0.1;
}

Answer 1

It is needed to be in paranthesis like

(float4)(1,1,1,1)

to have a float4 type. Or you can use other types too

(float4)((float2)(1,1),(float2)(1,1))

or mixed type

(float4)((float2)(1,1),1,1)

acts like an overloaded function.