How to write velocity clamp in openCL for pyro sim

[Ekaterina Sky]

Hi, I have a openCL question. How can I write a velocity clamp in openCL for pyro sim? In VEX, I write like this:

float speed = length(v@vel);

vector dir = normalize(v@vel);

@vel = dir * clamp(speed, 0, 20); 

However, I have no idea how to write it in openCL for a minimal solver mode. I suceeded to use ChatGPT to write a minor change in exisitng code, but not for a full writing. 

I am expecting to use a Gas openCL node, and not only write a code but also set options, bindings, Generated code tabs. Thus, it would appriciate if someone provide an example hip file. 

Thanks for your assistance!

[Librarian]

Just find White dog on Twitter (X) he posted Some Files regarding that and everything its in openCL @Ekaterina Sky

[Ekaterina Sky]

here is his answer and the code is working;

kernel void kernelName(
    int stride_x,
    int stride_y,
    int stride_z,
    int stride_offset,
    global float * vel_1, global float * vel_2, global float * vel_3
)
{
    int gidx = get_global_id(0);
    int gidy = get_global_id(1);
    int gidz = get_global_id(2);
    int xres = get_global_size(0);
    int yres = get_global_size(1);
    int zres = get_global_size(2);
    int idx = stride_offset + gidx * stride_x + gidy * stride_y + gidz * stride_z;
    float3 vel = (float3)(vel_1[idx], vel_2[idx], vel_3[idx]);
    float speed = length(vel);
    float3 dir = normalize(vel);
    vel = dir * clamp(speed, 0.0f, 20.0f);
    vel_1[idx] = vel.x;
    vel_2[idx] = vel.y;
    vel_3[idx] = vel.z;
}

 

OpenCL is more difficult than VEX, but may figure out some formats and patterns when comparing other openCL codes, like an inside of gasTurbulence node.