// Copyright (c) 2023 Xiamen Yaji Software Co., Ltd.

CCEffect %{
  techniques:
  - name: opaque
    passes:
    - compute: cluster-main
      pass: cluster-build-cs
}%


CCProgram cluster-main %{
  #include <common/math/coordinates>

  precision highp float;
  #define LOCAL_SIZE_X 16u
  #define LOCAL_SIZE_Y 8u
  #define LOCAL_SIZE_Z 1u

  layout (local_size_x = LOCAL_SIZE_X, local_size_y = LOCAL_SIZE_Y, local_size_z = LOCAL_SIZE_Z) in;

  #pragma rate CCConst pass
  layout(std140) uniform CCConst {
    vec4 cc_nearFar;
    vec4 cc_viewPort;
    vec4 cc_workGroup;
    mat4 cc_matView;
    mat4 cc_matProjInv;
  };
  #pragma rate b_clustersBuffer pass
  layout(std430) buffer b_clustersBuffer { vec4 b_clusters[]; };

  vec4 screen2Eye(vec4 coord) {
    vec3 ndc = vec3(
      2.0 * (coord.x - cc_viewPort.x) / cc_viewPort.z - 1.0,
      2.0 * (coord.y - cc_viewPort.y) / cc_viewPort.w - 1.0,
      2.0 * coord.z - 1.0);

    CC_HANDLE_SAMPLE_NDC_FLIP_STATIC(ndc.y);

    vec4 eye = ((cc_matProjInv) * (vec4(ndc, 1.0)));
    eye      = eye / eye.w;
    return eye;
  }

  void main () {

    uint clusterIndex = gl_GlobalInvocationID.z * gl_WorkGroupSize.x * gl_WorkGroupSize.y +
      gl_GlobalInvocationID.y * gl_WorkGroupSize.x +
      gl_GlobalInvocationID.x;

    float clusterSizeX = ceil(cc_viewPort.z / cc_workGroup.x);
    float clusterSizeY = ceil(cc_viewPort.w / cc_workGroup.y);
    vec4  minScreen    = vec4(vec2(gl_GlobalInvocationID.xy) * vec2(clusterSizeX, clusterSizeY), 1.0, 1.0);
    vec4  maxScreen    = vec4(vec2(gl_GlobalInvocationID.xy + uvec2(1, 1)) * vec2(clusterSizeX, clusterSizeY), 1.0, 1.0);
    vec3  minEye       = screen2Eye(minScreen).xyz;
    vec3  maxEye       = screen2Eye(maxScreen).xyz;
    float clusterNear  = -cc_nearFar.x * pow(cc_nearFar.y / cc_nearFar.x, float(gl_GlobalInvocationID.z) / cc_workGroup.z);
    float clusterFar   = -cc_nearFar.x * pow(cc_nearFar.y / cc_nearFar.x, float(gl_GlobalInvocationID.z + 1u) / cc_workGroup.z);
    vec3  minNear      = minEye * clusterNear / minEye.z;
    vec3  minFar       = minEye * clusterFar / minEye.z;
    vec3  maxNear      = maxEye * clusterNear / maxEye.z;
    vec3  maxFar       = maxEye * clusterFar / maxEye.z;
    vec3  minBounds    = min(min(minNear, minFar), min(maxNear, maxFar));
    vec3  maxBounds    = max(max(minNear, minFar), max(maxNear, maxFar));

    b_clusters[2u * clusterIndex + 0u] = vec4(minBounds, 1.0);
    b_clusters[2u * clusterIndex + 1u] = vec4(maxBounds, 1.0);
  }
}%