/* SPDX-License-Identifier: GPL-2.0-or-later
 * Copyright 2021 Blender Foundation.
 */

/** \file
 * \ingroup eevee
 *
 * Depth of field post process effect.
 *
 * There are 2 methods to achieve this effect.
 * - The first uses projection matrix offsetting and sample accumulation to give
 * reference quality depth of field. But this needs many samples to hide the
 * under-sampling.
 * - The second one is a post-processing based one. It follows the
 * implementation described in the presentation "Life of a Bokeh - Siggraph
 * 2018" from Guillaume Abadie. There are some difference with our actual
 * implementation that prioritize quality.
 */

#include "DRW_render.h"

#include "BKE_camera.h"
#include "DNA_camera_types.h"

#include "GPU_texture.h"
#include "GPU_uniform_buffer.h"

#include "eevee_camera.hh"
#include "eevee_instance.hh"
#include "eevee_sampling.hh"
#include "eevee_shader.hh"
#include "eevee_shader_shared.hh"

#include "eevee_depth_of_field.hh"

namespace blender::eevee {

/* -------------------------------------------------------------------- */
/** \name Depth of field
 * \{ */

void DepthOfField::init(void)
{
  const Instance &inst = inst_;
  const SceneEEVEE &sce_eevee = inst.scene->eevee;
  do_hq_slight_focus_ = (sce_eevee.flag & SCE_EEVEE_DOF_HQ_SLIGHT_FOCUS) != 0;
  do_jitter_ = (sce_eevee.flag & SCE_EEVEE_DOF_JITTER) != 0;
  user_overblur_ = sce_eevee.bokeh_overblur / 100.0f;
  fx_max_coc_ = sce_eevee.bokeh_max_size;
  data_.scatter_color_threshold = sce_eevee.bokeh_threshold;
  data_.scatter_neighbor_max_color = sce_eevee.bokeh_neighbor_max;
  data_.denoise_factor = sce_eevee.bokeh_denoise_fac;
  /* Default to no depth of field. */
  fx_radius_ = 0.0f;
  jitter_radius_ = 0.0f;
}

void DepthOfField::sync(const float4x4 &winmat, int2 input_extent)
{
  const Object *camera_object_eval = inst_.camera_eval_object;
  const ::Camera *cam = (camera_object_eval) ?
                            reinterpret_cast<const ::Camera *>(camera_object_eval->data) :
                            nullptr;

  if (cam == nullptr || (cam->dof.flag & CAM_DOF_ENABLED) == 0) {
    fx_radius_ = 0.0f;
    jitter_radius_ = 0.0f;
    return;
  }

  extent_ = input_extent;

  data_.bokeh_blades = cam->dof.aperture_blades;
  data_.bokeh_rotation = cam->dof.aperture_rotation;
  data_.bokeh_anisotropic_scale.x = clamp_f(1.0f / cam->dof.aperture_ratio, 0.00001f, 1.0f);
  data_.bokeh_anisotropic_scale.y = clamp_f(cam->dof.aperture_ratio, 0.00001f, 1.0f);
  data_.bokeh_anisotropic_scale_inv = 1.0f / data_.bokeh_anisotropic_scale;

  focus_distance_ = BKE_camera_object_dof_distance(camera_object_eval);
  float fstop = max_ff(cam->dof.aperture_fstop, 1e-5f);
  float aperture = 1.0f / (2.0f * fstop);
  if (cam->type == CAM_PERSP) {
    aperture *= cam->lens * 1e-3f;
  }

  if (cam->type == CAM_ORTHO) {
    /* FIXME: Why is this needed? Some kind of implicit unit conversion? */
    aperture *= 0.04f;
    /* Really strange behavior from Cycles but replicating. */
    focus_distance_ += cam->clip_start;
  }

  if (cam->type == CAM_PANO) {
    /* FIXME: Eyeballed. */
    aperture *= 0.185f;
  }

  if (cam->dof.aperture_ratio < 1.0) {
    /* If ratio is scaling the bokeh outwards, we scale the aperture so that
     * the gather kernel size will encompass the maximum axis. */
    aperture /= max_ff(cam->dof.aperture_ratio, 1e-5f);
  }

  /* Balance blur radius between fx dof and jitter dof. */
  if (do_jitter_ && (inst_.sampling.dof_ring_count_get() > 0) && (cam->type != CAM_PANO)) {
    /* Compute a minimal overblur radius to fill the gaps between the samples.
     * This is just the simplified form of dividing the area of the bokeh by
     * the number of samples. */
    float minimal_overblur = 1.0f / sqrtf(inst_.sampling.dof_sample_count_get());

    fx_radius_ = (minimal_overblur + user_overblur_) * aperture;
    /* Avoid dilating the shape. Over-blur only soften. */
    jitter_radius_ = max_ff(0.0f, aperture - fx_radius_);
  }
  else {
    jitter_radius_ = 0.0f;
    fx_radius_ = aperture;
  }

  if (fx_max_coc_ > 0.0f && fx_radius_ > 0.0f) {
    data_.camera_type = inst_.camera.data_get().type;
    /* OPTI(fclem) Could be optimized. */
    float jitter[3] = {fx_radius_, 0.0f, -focus_distance_};
    float center[3] = {0.0f, 0.0f, -focus_distance_};
    mul_project_m4_v3(winmat.ptr(), jitter);
    mul_project_m4_v3(winmat.ptr(), center);
    /* Simplify CoC calculation to a simple MADD. */
    if (data_.camera_type != CAMERA_ORTHO) {
      data_.coc_bias = -(center[0] - jitter[0]) * 0.5f * extent_[0];
      data_.coc_mul = focus_distance_ * data_.coc_bias;
    }
    else {
      data_.coc_mul = (center[0] - jitter[0]) * 0.5f * extent_[0];
      data_.coc_bias = focus_distance_ * data_.coc_mul;
    }

    float min_fg_coc = coc_radius_from_camera_depth(data_, -cam->clip_start);
    float max_bg_coc = coc_radius_from_camera_depth(data_, -cam->clip_end);
    if (data_.camera_type != CAMERA_ORTHO) {
      /* Background is at infinity so maximum CoC is the limit of the function at -inf. */
      /* NOTE: we only do this for perspective camera since orthographic coc limit is inf. */
      max_bg_coc = data_.coc_bias;
    }
    /* Clamp with user defined max. */
    data_.coc_abs_max = min_ff(max_ff(fabsf(min_fg_coc), fabsf(max_bg_coc)), fx_max_coc_);

    bokeh_lut_pass_sync();
    setup_pass_sync();
    tiles_prepare_pass_sync();
    reduce_pass_sync();
    convolve_pass_sync();
    resolve_pass_sync();

    data_.push_update();
  }
}

void DepthOfField::jitter_apply(float4x4 &winmat, float4x4 &viewmat)
{
  if (jitter_radius_ == 0.0f) {
    return;
  }
  float radius, theta;
  inst_.sampling.dof_disk_sample_get(&radius, &theta);

  if (data_.bokeh_blades >= 3.0f) {
    theta = circle_to_polygon_angle(data_.bokeh_blades, theta);
    radius *= circle_to_polygon_radius(data_.bokeh_blades, theta);
  }
  radius *= jitter_radius_;
  theta += data_.bokeh_rotation;

  /* Sample in View Space. */
  float2 sample = float2(radius * cosf(theta), radius * sinf(theta));
  sample *= data_.bokeh_anisotropic_scale;
  /* Convert to NDC Space. */
  float3 jitter = float3(UNPACK2(sample), -focus_distance_);
  float3 center = float3(0.0f, 0.0f, -focus_distance_);
  mul_project_m4_v3(winmat.ptr(), jitter);
  mul_project_m4_v3(winmat.ptr(), center);

  const bool is_ortho = (winmat[2][3] != -1.0f);
  if (is_ortho) {
    sample *= focus_distance_;
  }
  /* Translate origin. */
  sub_v2_v2(viewmat[3], sample);
  /* Skew winmat Z axis. */
  add_v2_v2(winmat[2], center - jitter);
}

/** Will swap input and output texture if rendering happens. The actual output of this function
 * is in intput_tx. */
void DepthOfField::render(GPUTexture *depth_tx, GPUTexture **input_tx, GPUTexture **output_tx)
{
  if (fx_radius_ == 0.0f || fx_max_coc_ < 0.5f) {
    return;
  }

  input_color_tx_ = *input_tx;
  input_depth_tx_ = depth_tx;

  DRW_stats_group_start("Depth of Field");

  bokeh_lut_pass_render();

  setup_pass_render();
  tiles_prepare_pass_render();
  reduce_pass_render();
  convolve_pass_render();
  resolve_pass_render(*output_tx);

  DRW_stats_group_end();

  /* Swap buffers so that next effect has the right input. */
  *input_tx = *output_tx;
  *output_tx = input_color_tx_;
}

/**
 * Creates bokeh texture.
 **/
void DepthOfField::bokeh_lut_pass_sync(void)
{
  const bool has_anisotropy = data_.bokeh_anisotropic_scale != float2(1.0f);
  if (has_anisotropy && (data_.bokeh_blades == 0.0)) {
    bokeh_gather_lut_tx_ = nullptr;
    bokeh_scatter_lut_tx_ = nullptr;
    bokeh_resolve_lut_tx_ = nullptr;
    bokeh_lut_ps_ = nullptr;
    return;
  }

  DrawEngineType *owner = (DrawEngineType *)view_name_.c_str();
  int res[2] = {DOF_BOKEH_LUT_SIZE, DOF_BOKEH_LUT_SIZE};

  DRW_PASS_CREATE(bokeh_lut_ps_, DRW_STATE_WRITE_COLOR);
  GPUShader *sh = inst_.shaders.static_shader_get(DOF_BOKEH_LUT);
  DRWShadingGroup *grp = DRW_shgroup_create(sh, bokeh_lut_ps_);
  DRW_shgroup_uniform_block(grp, "dof_buf", data_);
  DRW_shgroup_call_procedural_triangles(grp, nullptr, 1);

  bokeh_gather_lut_tx_ = DRW_texture_pool_query_2d(UNPACK2(res), GPU_RG16F, owner);
  bokeh_scatter_lut_tx_ = DRW_texture_pool_query_2d(UNPACK2(res), GPU_R16F, owner);
  bokeh_resolve_lut_tx_ = DRW_texture_pool_query_2d(UNPACK2(res), GPU_R16F, owner);

  bokeh_lut_fb_.ensure(GPU_ATTACHMENT_NONE,
                       GPU_ATTACHMENT_TEXTURE(bokeh_gather_lut_tx_),
                       GPU_ATTACHMENT_TEXTURE(bokeh_scatter_lut_tx_),
                       GPU_ATTACHMENT_TEXTURE(bokeh_resolve_lut_tx_));
}

void DepthOfField::bokeh_lut_pass_render(void)
{
  if (bokeh_lut_ps_) {
    GPU_framebuffer_bind(bokeh_lut_fb_);
    DRW_draw_pass(bokeh_lut_ps_);
  }
}

/**
 * Outputs halfResColorBuffer and halfResCocBuffer.
 **/
void DepthOfField::setup_pass_sync(void)
{
  DrawEngineType *owner = (DrawEngineType *)view_name_.c_str();
  uint res[2] = {divide_ceil_u(extent_[0], 2), divide_ceil_u(extent_[1], 2)};

  DRW_PASS_CREATE(setup_ps_, DRW_STATE_WRITE_COLOR);
  GPUShader *sh = inst_.shaders.static_shader_get(DOF_SETUP);
  DRWShadingGroup *grp = DRW_shgroup_create(sh, setup_ps_);

  eGPUSamplerState no_filter = GPU_SAMPLER_DEFAULT;
  DRW_shgroup_uniform_texture_ref_ex(grp, "color_tx", &input_color_tx_, no_filter);
  DRW_shgroup_uniform_texture_ref_ex(grp, "depth_tx", &input_depth_tx_, no_filter);
  DRW_shgroup_uniform_block(grp, "dof_buf", data_);
  DRW_shgroup_call_procedural_triangles(grp, nullptr, 1);

  setup_color_tx_ = DRW_texture_pool_query_2d(UNPACK2(res), GPU_RGBA16F, owner);
  setup_coc_tx_ = DRW_texture_pool_query_2d(UNPACK2(res), GPU_RG16F, owner);

  setup_fb_.ensure(GPU_ATTACHMENT_NONE,
                   GPU_ATTACHMENT_TEXTURE(setup_color_tx_),
                   GPU_ATTACHMENT_TEXTURE(setup_coc_tx_));
}

void DepthOfField::setup_pass_render(void)
{
  GPU_framebuffer_bind(setup_fb_);
  DRW_draw_pass(setup_ps_);
}

/**
 * Outputs min & max COC in each 8x8 half res pixel tiles (so 1/16th of full resolution).
 * Then dilates the min & max CoCs to cover maximum COC values.
 **/
void DepthOfField::tiles_prepare_pass_sync(void)
{
  DrawEngineType *owner = (DrawEngineType *)view_name_.c_str();
  uint res[2] = {divide_ceil_u(extent_[0], DOF_TILE_DIVISOR),
                 divide_ceil_u(extent_[1], DOF_TILE_DIVISOR)};
  /* WARNING: If you change this, make sure dof_tile_* GLSL constants can be properly encoded. */
  eGPUTextureFormat fg_tile_format = GPU_RGBA16F;
  eGPUTextureFormat bg_tile_format = GPU_R11F_G11F_B10F;

  eGPUSamplerState no_filter = GPU_SAMPLER_DEFAULT;
  {
    DRW_PASS_CREATE(tiles_flatten_ps_, DRW_STATE_WRITE_COLOR);

    GPUShader *sh = inst_.shaders.static_shader_get(DOF_TILES_FLATTEN);
    DRWShadingGroup *grp = DRW_shgroup_create(sh, tiles_flatten_ps_);
    DRW_shgroup_uniform_texture_ref_ex(grp, "coc_tx", &setup_coc_tx_, no_filter);
    DRW_shgroup_call_procedural_triangles(grp, nullptr, 1);

    tiles_fg_tx_ = DRW_texture_pool_query_2d(UNPACK2(res), fg_tile_format, owner);
    tiles_bg_tx_ = DRW_texture_pool_query_2d(UNPACK2(res), bg_tile_format, owner);

    tiles_flatten_fb_.ensure(GPU_ATTACHMENT_NONE,
                             GPU_ATTACHMENT_TEXTURE(tiles_fg_tx_),
                             GPU_ATTACHMENT_TEXTURE(tiles_bg_tx_));
  }
  {
    DRW_PASS_CREATE(tiles_dilate_minmax_ps_, DRW_STATE_WRITE_COLOR);
    DRW_PASS_CREATE(tiles_dilate_minabs_ps_, DRW_STATE_WRITE_COLOR);

    for (int pass = 0; pass < 2; pass++) {
      DRWPass *drw_pass = (pass == 0) ? tiles_dilate_minmax_ps_ : tiles_dilate_minabs_ps_;
      GPUShader *sh = inst_.shaders.static_shader_get((pass == 0) ? DOF_TILES_DILATE_MINMAX :
                                                                    DOF_TILES_DILATE_MINABS);
      DRWShadingGroup *grp = DRW_shgroup_create(sh, drw_pass);
      DRW_shgroup_uniform_texture_ref_ex(grp, "tiles_fg_tx", &tiles_fg_tx_, no_filter);
      DRW_shgroup_uniform_texture_ref_ex(grp, "tiles_bg_tx", &tiles_bg_tx_, no_filter);
      DRW_shgroup_uniform_bool(grp, "dilate_slight_focus", &tiles_dilate_slight_focus_, 1);
      DRW_shgroup_uniform_int(grp, "ring_count", &tiles_dilate_ring_count_, 1);
      DRW_shgroup_uniform_int(
          grp, "ring_width_multiplier", &tiles_dilate_ring_width_multiplier_, 1);
      DRW_shgroup_call_procedural_triangles(grp, nullptr, 1);
    }

    tiles_dilated_fg_tx_ = DRW_texture_pool_query_2d(UNPACK2(res), fg_tile_format, owner);
    tiles_dilated_bg_tx_ = DRW_texture_pool_query_2d(UNPACK2(res), bg_tile_format, owner);

    tiles_dilate_fb_.ensure(GPU_ATTACHMENT_NONE,
                            GPU_ATTACHMENT_TEXTURE(tiles_dilated_fg_tx_),
                            GPU_ATTACHMENT_TEXTURE(tiles_dilated_bg_tx_));
  }
}

void DepthOfField::tiles_prepare_pass_render(void)
{
  GPU_framebuffer_bind(tiles_flatten_fb_);
  DRW_draw_pass(tiles_flatten_ps_);

  /* Run dilation twice. One for minmax and one for minabs. */
  for (int pass = 0; pass < 2; pass++) {
    /* Error introduced by gather center jittering. */
    const float error_multiplier = 1.0f + 1.0f / (DOF_GATHER_RING_COUNT + 0.5f);
    int dilation_end_radius = ceilf((fx_max_coc_ * error_multiplier) / DOF_TILE_DIVISOR);

    /* This algorithm produce the exact dilation radius by dividing it in multiple passes. */
    int dilation_radius = 0;
    while (dilation_radius < dilation_end_radius) {
      /* Dilate slight focus only on first iteration. */
      tiles_dilate_slight_focus_ = (dilation_radius == 0) ? 1 : 0;

      int remainder = dilation_end_radius - dilation_radius;
      /* Do not step over any unvisited tile. */
      int max_multiplier = dilation_radius + 1;

      int ring_count = min_ii(DOF_DILATE_RING_COUNT, ceilf(remainder / (float)max_multiplier));
      int multiplier = min_ii(max_multiplier, floor(remainder / (float)ring_count));

      dilation_radius += ring_count * multiplier;

      tiles_dilate_ring_count_ = ring_count;
      tiles_dilate_ring_width_multiplier_ = multiplier;

      GPU_framebuffer_bind(tiles_dilate_fb_);
      DRW_draw_pass((pass == 0) ? tiles_dilate_minmax_ps_ : tiles_dilate_minabs_ps_);

      Framebuffer::swap(tiles_dilate_fb_, tiles_flatten_fb_);
      SWAP(GPUTexture *, tiles_dilated_bg_tx_, tiles_bg_tx_);
      SWAP(GPUTexture *, tiles_dilated_fg_tx_, tiles_fg_tx_);
    }
  }
  /* Swap again so that final textures are tiles_dilated_*_tx_. */
  Framebuffer::swap(tiles_dilate_fb_, tiles_flatten_fb_);
  SWAP(GPUTexture *, tiles_dilated_bg_tx_, tiles_bg_tx_);
  SWAP(GPUTexture *, tiles_dilated_fg_tx_, tiles_fg_tx_);
}

/**
 * Create mipmapped color & COC textures for gather passes.
 **/
void DepthOfField::reduce_pass_sync(void)
{
  DrawEngineType *owner = (DrawEngineType *)view_name_.c_str();
  eGPUSamplerState no_filter = GPU_SAMPLER_DEFAULT;
  /* Divide by 2 because dof_fx_max_coc is in fullres CoC radius and the reduce
   * texture begins at half resolution. */
  float max_space_between_sample = fx_max_coc_ * 0.5f / DOF_GATHER_RING_COUNT;

  int mip_count = max_ii(1, log2_ceil_u(max_space_between_sample));

  reduce_steps_ = mip_count - 1;
  /* This ensure the mipmaps are aligned for the needed 4 mip levels.
   * Starts at 2 because already at half resolution. */
  int multiple = 2 << (mip_count - 1);
  int2 res = (int2(divide_ceil_u(extent_[0], multiple), divide_ceil_u(extent_[1], multiple)) *
              multiple) /
             2;

  int2 quater_res = int2(divide_ceil_u(extent_[0], 4), divide_ceil_u(extent_[1], 4));

  /* TODO(fclem): Make this dependent of the quality of the gather pass. */
  data_.scatter_coc_threshold = 4.0f;

  /* Color needs to be signed format here. See note in shader for explanation. */
  /* Do not use texture pool because of needs mipmaps. */
  reduced_color_tx_.ensure_2d(GPU_RGBA16F, res, nullptr, mip_count);
  reduced_coc_tx_.ensure_2d(GPU_R16F, res, nullptr, mip_count);

  {
    DRW_PASS_CREATE(reduce_downsample_ps_, DRW_STATE_WRITE_COLOR);
    GPUShader *sh = inst_.shaders.static_shader_get(DOF_REDUCE_DOWNSAMPLE);
    DRWShadingGroup *grp = DRW_shgroup_create(sh, reduce_downsample_ps_);
    DRW_shgroup_uniform_texture_ref_ex(grp, "color_tx", &setup_color_tx_, no_filter);
    DRW_shgroup_uniform_texture_ref_ex(grp, "coc_tx", &setup_coc_tx_, no_filter);
    DRW_shgroup_call_procedural_triangles(grp, nullptr, 1);

    reduce_downsample_tx_ = DRW_texture_pool_query_2d(UNPACK2(quater_res), GPU_RGBA16F, owner);

    reduce_downsample_fb_.ensure(GPU_ATTACHMENT_NONE,
                                 GPU_ATTACHMENT_TEXTURE(reduce_downsample_tx_));
  }
  {
    DRW_PASS_CREATE(reduce_copy_ps_, DRW_STATE_WRITE_COLOR);
    GPUShader *sh = inst_.shaders.static_shader_get(DOF_REDUCE_COPY);
    DRWShadingGroup *grp = DRW_shgroup_create(sh, reduce_copy_ps_);
    DRW_shgroup_uniform_texture_ref_ex(grp, "color_tx", &setup_color_tx_, no_filter);
    DRW_shgroup_uniform_texture_ref_ex(grp, "coc_tx", &setup_coc_tx_, no_filter);
    DRW_shgroup_uniform_texture_ex(grp, "downsampled_tx", reduce_downsample_tx_, no_filter);
    DRW_shgroup_uniform_block(grp, "dof_buf", data_);
    DRW_shgroup_call_procedural_triangles(grp, nullptr, 1);

    scatter_src_tx_ = DRW_texture_pool_query_2d(UNPACK2(res), GPU_R11F_G11F_B10F, owner);
  }
  {
    DRW_PASS_CREATE(reduce_recursive_ps_, DRW_STATE_WRITE_COLOR);
    GPUShader *sh = inst_.shaders.static_shader_get(DOF_REDUCE_RECURSIVE);
    DRWShadingGroup *grp = DRW_shgroup_create(sh, reduce_recursive_ps_);
    DRW_shgroup_uniform_texture_ex(grp, "color_tx", reduced_color_tx_, no_filter);
    DRW_shgroup_uniform_texture_ex(grp, "coc_tx", reduced_coc_tx_, no_filter);
    DRW_shgroup_call_procedural_triangles(grp, nullptr, 1);
  }

  reduce_fb_.ensure(GPU_ATTACHMENT_NONE,
                    GPU_ATTACHMENT_TEXTURE(reduced_color_tx_),
                    GPU_ATTACHMENT_TEXTURE(reduced_coc_tx_));

  reduce_copy_fb_.ensure(GPU_ATTACHMENT_NONE,
                         GPU_ATTACHMENT_TEXTURE(reduced_color_tx_),
                         GPU_ATTACHMENT_TEXTURE(reduced_coc_tx_),
                         GPU_ATTACHMENT_TEXTURE(scatter_src_tx_));
}

void DepthOfField::reduce_recusive(void *thunk, int UNUSED(level))
{
  DepthOfField *dof = reinterpret_cast<DepthOfField *>(thunk);
  DRW_draw_pass(dof->reduce_recursive_ps_);
}

void DepthOfField::reduce_pass_render(void)
{
  GPU_framebuffer_bind(reduce_downsample_fb_);
  DRW_draw_pass(reduce_downsample_ps_);

  /* First step is just a copy. */
  GPU_framebuffer_bind(reduce_copy_fb_);
  DRW_draw_pass(reduce_copy_ps_);

  GPU_framebuffer_recursive_downsample(reduce_fb_, reduce_steps_, &reduce_recusive, this);
}

/**
 * Do the gather & scatter convolution. For each pixels we gather multiple pixels in its
 * neighborhood depending on the min & max CoC tiles. We apply a median filter on the output.
 * We also scatter a sprite for very bright pixels for high quality bokeh.
 **/
void DepthOfField::convolve_pass_sync(void)
{
  eGPUSamplerState no_filter = GPU_SAMPLER_DEFAULT;
  eGPUSamplerState with_filter = (GPU_SAMPLER_MIPMAP | GPU_SAMPLER_FILTER);
  DrawEngineType *owner = (DrawEngineType *)view_name_.c_str();
  uint res[2] = {divide_ceil_u(extent_[0], 2), divide_ceil_u(extent_[1], 2)};
  int input_size[2];
  GPU_texture_get_mipmap_size(reduced_color_tx_, 0, input_size);
  for (int i = 0; i < 2; i++) {
    data_.gather_uv_fac[i] = 1.0f / (float)input_size[i];
    data_.texel_size[i] = 1.0f / res[i];
  }

  /* Reuse textures from the setup pass. */
  /* NOTE: We could use the texture pool do that for us but it does not track
   * usage and it might backfire (it does in practice). */
  /* Since it is only used for scatter, and foreground is processed before background, we can
   * reuse the occlusion_tx for both field. */
  occlusion_tx_ = setup_coc_tx_;

  {
    DRW_PASS_CREATE(gather_holefill_ps_, DRW_STATE_WRITE_COLOR);
    GPUShader *sh = inst_.shaders.static_shader_get(DOF_GATHER_HOLEFILL);
    DRWShadingGroup *grp = DRW_shgroup_create(sh, gather_holefill_ps_);
    DRW_shgroup_uniform_texture_ex(grp, "color_bilinear_tx", reduced_color_tx_, with_filter);
    DRW_shgroup_uniform_texture_ex(grp, "color_tx", reduced_color_tx_, no_filter);
    DRW_shgroup_uniform_texture_ex(grp, "coc_tx", reduced_coc_tx_, no_filter);
    DRW_shgroup_uniform_texture(grp, "tiles_bg_tx", tiles_dilated_bg_tx_);
    DRW_shgroup_uniform_texture(grp, "tiles_fg_tx", tiles_dilated_fg_tx_);
    DRW_shgroup_uniform_block(grp, "dof_buf", data_);
    DRW_shgroup_uniform_block(grp, "sampling_buf", inst_.sampling.ubo_get());
    DRW_shgroup_call_procedural_triangles(grp, nullptr, 1);

    /* Reuse textures from the setup pass. */
    /* NOTE: We could use the texture pool do that for us but it does not track
     * usage and it might backfire (it does in practice). */
    color_holefill_tx_ = setup_color_tx_;
    weight_holefill_tx_ = DRW_texture_pool_query_2d(UNPACK2(res), GPU_R16F, owner);

    gather_holefill_fb_.ensure(GPU_ATTACHMENT_NONE,
                               GPU_ATTACHMENT_TEXTURE(color_holefill_tx_),
                               GPU_ATTACHMENT_TEXTURE(weight_holefill_tx_));
  }
  {
    DRW_PASS_CREATE(gather_fg_ps_, DRW_STATE_WRITE_COLOR);
    GPUShader *sh = inst_.shaders.static_shader_get(
        bokeh_gather_lut_tx_ ? DOF_GATHER_FOREGROUND_LUT : DOF_GATHER_FOREGROUND);
    DRWShadingGroup *grp = DRW_shgroup_create(sh, gather_fg_ps_);
    DRW_shgroup_uniform_texture_ex(grp, "color_bilinear_tx", reduced_color_tx_, with_filter);
    DRW_shgroup_uniform_texture_ex(grp, "color_tx", reduced_color_tx_, no_filter);
    DRW_shgroup_uniform_texture_ex(grp, "coc_tx", reduced_coc_tx_, no_filter);
    DRW_shgroup_uniform_texture(grp, "tiles_bg_tx", tiles_dilated_bg_tx_);
    DRW_shgroup_uniform_texture(grp, "tiles_fg_tx", tiles_dilated_fg_tx_);
    if (bokeh_gather_lut_tx_) {
      DRW_shgroup_uniform_texture_ref(grp, "bokeh_lut_tx", &bokeh_gather_lut_tx_);
    }
    DRW_shgroup_uniform_block(grp, "dof_buf", data_);
    DRW_shgroup_uniform_block(grp, "sampling_buf", inst_.sampling.ubo_get());
    DRW_shgroup_call_procedural_triangles(grp, nullptr, 1);

    color_fg_tx_ = DRW_texture_pool_query_2d(UNPACK2(res), GPU_RGBA16F, owner);
    weight_fg_tx_ = DRW_texture_pool_query_2d(UNPACK2(res), GPU_R16F, owner);
  }
  {
    DRW_PASS_CREATE(gather_bg_ps_, DRW_STATE_WRITE_COLOR);
    GPUShader *sh = inst_.shaders.static_shader_get(
        bokeh_gather_lut_tx_ ? DOF_GATHER_BACKGROUND_LUT : DOF_GATHER_BACKGROUND);
    DRWShadingGroup *grp = DRW_shgroup_create(sh, gather_bg_ps_);
    DRW_shgroup_uniform_texture_ex(grp, "color_bilinear_tx", reduced_color_tx_, with_filter);
    DRW_shgroup_uniform_texture_ex(grp, "color_tx", reduced_color_tx_, no_filter);
    DRW_shgroup_uniform_texture_ex(grp, "coc_tx", reduced_coc_tx_, no_filter);
    DRW_shgroup_uniform_texture(grp, "tiles_bg_tx", tiles_dilated_bg_tx_);
    DRW_shgroup_uniform_texture(grp, "tiles_fg_tx", tiles_dilated_fg_tx_);
    if (bokeh_gather_lut_tx_) {
      DRW_shgroup_uniform_texture_ref(grp, "bokeh_lut_tx", &bokeh_gather_lut_tx_);
    }
    DRW_shgroup_uniform_block(grp, "dof_buf", data_);
    DRW_shgroup_uniform_block(grp, "sampling_buf", inst_.sampling.ubo_get());
    DRW_shgroup_call_procedural_triangles(grp, nullptr, 1);

    color_bg_tx_ = DRW_texture_pool_query_2d(UNPACK2(res), GPU_RGBA16F, owner);
    weight_bg_tx_ = DRW_texture_pool_query_2d(UNPACK2(res), GPU_R16F, owner);
  }

  /* NOTE: First target is holefill texture so we can use the median filter on it and save some
   * texture memory. Both field use the same framebuffer. */
  gather_fb_.ensure(GPU_ATTACHMENT_NONE,
                    GPU_ATTACHMENT_TEXTURE(color_holefill_tx_),
                    GPU_ATTACHMENT_TEXTURE(weight_holefill_tx_),
                    GPU_ATTACHMENT_TEXTURE(occlusion_tx_));
  {
    /**
     * Filter an input buffer using a median filter to reduce noise.
     * NOTE: We use the holefill texture as our input to reduce memory usage.
     * Thus, the holefill pass cannot be filtered.
     **/
    DRW_PASS_CREATE(filter_ps_, DRW_STATE_WRITE_COLOR);
    GPUShader *sh = inst_.shaders.static_shader_get(DOF_FILTER);
    DRWShadingGroup *grp = DRW_shgroup_create(sh, filter_ps_);
    DRW_shgroup_uniform_texture_ex(grp, "color_tx", color_holefill_tx_, no_filter);
    DRW_shgroup_uniform_texture_ex(grp, "weight_tx", weight_holefill_tx_, no_filter);
    DRW_shgroup_call_procedural_triangles(grp, nullptr, 1);

    filter_fg_fb_.ensure(GPU_ATTACHMENT_NONE,
                         GPU_ATTACHMENT_TEXTURE(color_fg_tx_),
                         GPU_ATTACHMENT_TEXTURE(weight_fg_tx_));

    filter_bg_fb_.ensure(GPU_ATTACHMENT_NONE,
                         GPU_ATTACHMENT_TEXTURE(color_bg_tx_),
                         GPU_ATTACHMENT_TEXTURE(weight_bg_tx_));
  }

  /**
   * Do the Scatter convolution. A sprite is emitted for every 4 pixels but is only expanded
   * if the pixels are bright enough to be scattered.
   **/
  data_.scatter_sprite_per_row = input_size[0] / 2;
  int sprite_count = data_.scatter_sprite_per_row * (input_size[1] / 2);
  {
    DRW_PASS_CREATE(scatter_fg_ps_, DRW_STATE_WRITE_COLOR | DRW_STATE_BLEND_ADD_FULL);
    GPUShader *sh = inst_.shaders.static_shader_get(
        bokeh_gather_lut_tx_ ? DOF_SCATTER_FOREGROUND_LUT : DOF_SCATTER_FOREGROUND);
    DRWShadingGroup *grp = DRW_shgroup_create(sh, scatter_fg_ps_);
    DRW_shgroup_uniform_texture_ex(grp, "color_tx", scatter_src_tx_, no_filter);
    DRW_shgroup_uniform_texture_ex(grp, "coc_tx", reduced_coc_tx_, no_filter);
    DRW_shgroup_uniform_texture(grp, "occlusion_tx", occlusion_tx_);
    if (bokeh_scatter_lut_tx_) {
      DRW_shgroup_uniform_texture(grp, "bokehLut", bokeh_scatter_lut_tx_);
    }
    DRW_shgroup_uniform_block(grp, "dof_buf", data_);
    DRW_shgroup_call_procedural_triangles(grp, nullptr, sprite_count);

    scatter_fg_fb_.ensure(GPU_ATTACHMENT_NONE, GPU_ATTACHMENT_TEXTURE(color_fg_tx_));
  }
  {
    DRW_PASS_CREATE(scatter_bg_ps_, DRW_STATE_WRITE_COLOR | DRW_STATE_BLEND_ADD_FULL);
    GPUShader *sh = inst_.shaders.static_shader_get(
        bokeh_gather_lut_tx_ ? DOF_SCATTER_BACKGROUND_LUT : DOF_SCATTER_BACKGROUND);
    DRWShadingGroup *grp = DRW_shgroup_create(sh, scatter_bg_ps_);
    DRW_shgroup_uniform_texture_ex(grp, "color_tx", scatter_src_tx_, no_filter);
    DRW_shgroup_uniform_texture_ex(grp, "coc_tx", reduced_coc_tx_, no_filter);
    DRW_shgroup_uniform_texture(grp, "occlusion_tx", occlusion_tx_);
    if (bokeh_scatter_lut_tx_) {
      DRW_shgroup_uniform_texture(grp, "bokehLut", bokeh_scatter_lut_tx_);
    }
    DRW_shgroup_uniform_block(grp, "dof_buf", data_);
    DRW_shgroup_call_procedural_triangles(grp, nullptr, sprite_count);

    scatter_bg_fb_.ensure(GPU_ATTACHMENT_NONE, GPU_ATTACHMENT_TEXTURE(color_bg_tx_));
  }
}

void DepthOfField::convolve_pass_render(void)
{
  DRW_stats_group_start("Foreground convolution");
  GPU_framebuffer_bind(gather_fb_);
  DRW_draw_pass(gather_fg_ps_);

  GPU_framebuffer_bind(filter_fg_fb_);
  DRW_draw_pass(filter_ps_);

  GPU_framebuffer_bind(scatter_fg_fb_);
  DRW_draw_pass(scatter_fg_ps_);
  DRW_stats_group_end();

  DRW_stats_group_start("Background convolution");
  GPU_framebuffer_bind(gather_fb_);
  DRW_draw_pass(gather_bg_ps_);

  GPU_framebuffer_bind(filter_bg_fb_);
  DRW_draw_pass(filter_ps_);

  GPU_framebuffer_bind(scatter_bg_fb_);
  DRW_draw_pass(scatter_bg_ps_);
  DRW_stats_group_end();

  DRW_stats_group_start("Background convolution");
  /* Hole-fill convolution. */
  GPU_framebuffer_bind(gather_holefill_fb_);
  DRW_draw_pass(gather_holefill_ps_);
  /* NOTE: We do not filter the hole-fill pass as we use it as out filter input
   * buffer. Also effect is likely to not be noticeable. */
  DRW_stats_group_end();
}

/**
 * Recombine the result of the foreground and background processing.
 * Also perform a slight out of focus gather to improve geometric continuity.
 **/
void DepthOfField::resolve_pass_sync(void)
{
  eGPUSamplerState no_filter = GPU_SAMPLER_DEFAULT;
  eGPUSamplerState with_filter = GPU_SAMPLER_FILTER;
  eShaderType sh_type = (bokeh_resolve_lut_tx_) ?
                            (do_hq_slight_focus_ ? DOF_RESOLVE_LUT_HQ : DOF_RESOLVE_LUT) :
                            (do_hq_slight_focus_ ? DOF_RESOLVE_HQ : DOF_RESOLVE);

  DRW_PASS_CREATE(resolve_ps_, DRW_STATE_WRITE_COLOR);
  GPUShader *sh = inst_.shaders.static_shader_get(sh_type);
  DRWShadingGroup *grp = DRW_shgroup_create(sh, resolve_ps_);
  DRW_shgroup_uniform_texture_ref_ex(grp, "depth_tx", &input_depth_tx_, no_filter);
  DRW_shgroup_uniform_texture_ref_ex(grp, "color_tx", &input_color_tx_, no_filter);
  DRW_shgroup_uniform_texture_ex(grp, "color_bg_tx", color_bg_tx_, with_filter);
  DRW_shgroup_uniform_texture_ex(grp, "color_fg_tx", color_fg_tx_, with_filter);
  DRW_shgroup_uniform_texture_ex(grp, "color_holefill_tx", color_holefill_tx_, with_filter);
  DRW_shgroup_uniform_texture(grp, "tiles_bg_tx", tiles_dilated_bg_tx_);
  DRW_shgroup_uniform_texture(grp, "tiles_fg_tx", tiles_dilated_fg_tx_);
  DRW_shgroup_uniform_texture_ex(grp, "weight_bg_tx", weight_bg_tx_, with_filter);
  DRW_shgroup_uniform_texture_ex(grp, "weight_fg_tx", weight_fg_tx_, with_filter);
  DRW_shgroup_uniform_texture_ex(grp, "weight_holefill_tx", weight_holefill_tx_, with_filter);
  DRW_shgroup_uniform_block(grp, "dof_buf", data_);
  DRW_shgroup_uniform_block(grp, "sampling_buf", inst_.sampling.ubo_get());
  if (bokeh_resolve_lut_tx_) {
    DRW_shgroup_uniform_texture_ref(grp, "bokeh_lut_tx", &bokeh_resolve_lut_tx_);
  }
  DRW_shgroup_call_procedural_triangles(grp, nullptr, 1);
}

void DepthOfField::resolve_pass_render(GPUTexture *output_tx)
{
  resolve_fb_.ensure(GPU_ATTACHMENT_NONE, GPU_ATTACHMENT_TEXTURE(output_tx));

  GPU_framebuffer_bind(resolve_fb_);
  DRW_draw_pass(resolve_ps_);
}

/** \} */

}  // namespace blender::eevee