// This is "kernel 4" in a 4-kernel pipeline. It renders the commands
// in the per-tile command list to an image.

// Right now, this kernel stores the image in a buffer, but a better
// plan is to use a texture. This is because of limited support.

#version 450
#extension GL_GOOGLE_include_directive : enable

layout(local_size_x = 16, local_size_y = 16) in;

// Same concern that this should be readonly as in kernel 3.
layout(set = 0, binding = 0) buffer PtclBuf {
    uint[] ptcl;
};

layout(set = 0, binding = 1) buffer ImageBuf {
    uint[] image;
};

#include "ptcl.h"

#include "setup.h"

void main() {
    uint tile_ix = gl_WorkGroupID.y * WIDTH_IN_TILES + gl_WorkGroupID.x;
    CmdRef cmd_ref = CmdRef(tile_ix * PTCL_INITIAL_ALLOC);

    uvec2 xy_uint = gl_GlobalInvocationID.xy;
    vec2 xy = vec2(xy_uint);
    vec2 uv = xy * vec2(1.0 / IMAGE_WIDTH, 1.0 / IMAGE_HEIGHT);
    vec3 rgb = uv.xyy;

    while (true) {
        uint tag = Cmd_tag(cmd_ref);
        if (tag == Cmd_End) {
            break;
        }
        switch (tag) {
        case Cmd_Circle:
            CmdCircle circle = Cmd_Circle_read(cmd_ref);
            float r = length(xy + vec2(0.5, 0.5) - circle.center.xy);
            float alpha = clamp(0.5 + circle.radius - r, 0.0, 1.0);
            vec4 fg_rgba = unpackUnorm4x8(circle.rgba_color);
            // TODO: sRGB
            rgb = mix(rgb, fg_rgba.rgb, alpha * fg_rgba.a);
            break;
        case Cmd_Jump:
            cmd_ref = CmdRef(Cmd_Jump_read(cmd_ref).new_ref);
            continue;
        }
        cmd_ref.offset += Cmd_size;
    }

    // TODO: sRGB
    uvec4 s = uvec4(round(vec4(rgb, 1.0) * 255.0));
    uint rgba_packed = s.r | (s.g << 8) | (s.b << 16) | (s.a << 24);
    image[xy_uint.y * IMAGE_WIDTH + xy_uint.x] = rgba_packed;
}