#version 450

#define BARRIER() \
    memoryBarrierShared(); \
    barrier();

#extension GL_NV_mesh_shader : enable

layout(local_size_x = 32) in;

// test use of shared memory in task shaders:
layout(binding=0) writeonly uniform image2D uni_image;
uniform block0 {
    uint uni_value;
};
shared vec4 mem[10];

// test use of task memory in task shaders:
taskNV out Task {
    vec2 dummy;
    vec2 submesh[3];
} mytask;

void main()
{
    uint iid = gl_LocalInvocationID.x;
    uint gid = gl_WorkGroupID.x;

    // 1. shared memory load and stores
    for (uint i = 0; i < 10; ++i) {
        mem[i] = vec4(i + uni_value);
    }
    imageStore(uni_image, ivec2(iid), mem[gid]);
    imageStore(uni_image, ivec2(iid), mem[gid+1]);

    BARRIER();

    // 2. task memory stores

    mytask.dummy      = vec2(30.0, 31.0);
    mytask.submesh[0] = vec2(32.0, 33.0);
    mytask.submesh[1] = vec2(34.0, 35.0);
    mytask.submesh[2] = mytask.submesh[gid%2];

    BARRIER();

    // 3. set task count
    gl_TaskCountNV = 3;
}