source/blender/draw/engines/eevee_next/shaders/eevee_reflection_probe_lib.glsl

@@ -7,65 +7,53 @@ void light_world_eval(ClosureReflection reflection, vec3 P, vec3 V, inout vec3 o
    * eventually this should based on a user setting. */
   float lod_cube_max = 12.0;
 
+  /* Pow2f to distributed across lod more evenly */
+  float roughness = clamp(pow2f(reflection.roughness), 1e-4f, 0.9999f); /* Avoid artifacts */
+
+#if defined(GPU_COMPUTE_SHADER)
+  vec2 frag_coord = vec2(gl_GlobalInvocationID.xy) + 0.5;
+#else
+  vec2 frag_coord = gl_FragCoord.xy;
+#endif
+  float noise = utility_tx_fetch(utility_tx, frag_coord, UTIL_BLUE_NOISE_LAYER).g;
+  vec2 rand = fract(vec2(noise) + sampling_rng_2D_get(SAMPLING_RAYTRACE_U));
+
+  vec3 Xi = sample_cylinder(rand);
+
+  /* Microfacet normal */
   vec3 T, B;
   make_orthonormal_basis(reflection.N, T, B);
+  float pdf;
+  vec3 H = sample_ggx(Xi, roughness, V, reflection.N, T, B, pdf);
 
-  float weight = 0.0;
-  vec3 out_radiance = vec3(0.0);
-  /* Note: this was dynamic based on the mipmap level that was read from. */
-  /* lod 0 => 1
-   *     1 => 32
-   *     2 => 40
-   *     3 => 64
-   *  else => 128
-   * multiplied by the filter quality. */
-  const int max_sample_count = 32;
-  /* Note this is dynamically based on the mip map level.
-   pinfo->lodfactor = bias + 0.5f * log(square_f(target_size) / pinfo->samples_len) / log(2);
-   */
-  /* Pow4f for Disney Roughness distributed across lod more evenly */
-  float roughness = clamp(pow4f(reflection.roughness), 1e-4f, 0.9999f); /* Avoid artifacts */
+  vec3 L = -reflect(V, H);
+  float NL = dot(reflection.N, L);
 
-  const float lod_factor = 1.0 +
-                           0.5 * log(float(square_i(texture_size.x)) / float(max_sample_count)) /
-                               log(2);
-  /* We should find a math formular that would approx max_sample_count and bias. */
-  for (int i = 0; i < max_sample_count; i++) {
-    vec3 Xi = sample_cylinder(hammersley_2d(float(i), max_sample_count));
+  if (NL > 0.0) {
+    float NH = max(1e-8, dot(reflection.N, H)); /*cosTheta */
 
-    /* Microfacet normal */
-    float pdf;
-    vec3 H = sample_ggx(Xi, roughness, V, reflection.N, T, B, pdf);
+    /* Coarse Approximation of the mapping distortion
+     * Unit Sphere -> Cubemap Face */
+    const float dist = 4.0 * M_PI / 6.0;
 
-    vec3 L = -reflect(V, H);
-    float NL = dot(reflection.N, L);
+    /* http://http.developer.nvidia.com/GPUGems3/gpugems3_ch20.html : Equation 13 */
+    /* const float bias = 2.0;
+     const float lod_factor = bias + 0.5 * log(float(square_i(texture_size.x))) / log(2);
+     float lod = clamp(lod_factor - 0.5 * log2(pdf * dist), 0.0, lod_cube_max);
+ */
+    vec3 l_col = textureLod_cubemapArray(reflectionProbes, vec4(L, 0.0), 0.0).rgb;
 
-    if (NL > 0.0) {
-      float NH = max(1e-8, dot(reflection.N, H)); /*cosTheta */
+    /* Clamped brightness. */
+    // float luma = max(1e-8, max_v3(l_col));
 
-      /* Coarse Approximation of the mapping distortion
-       * Unit Sphere -> Cubemap Face */
-      const float dist = 4.0 * M_PI / 6.0;
-      /* http://http.developer.nvidia.com/GPUGems3/gpugems3_ch20.html : Equation 13 */
-      float lod = clamp(lod_factor - 0.5 * log2(pdf * dist), 0.0, lod_cube_max);
+    /* For artistic freedom this should be read from the scene/reflection probe.
+     * Note: Eevee-legacy read the firefly_factor from gi_glossy_clamp. */
+    // const float firefly_factor = 1e16;
+    // l_col *= 1.0 - max(0.0, luma - firefly_factor) / luma;
 
-      vec3 l_col = textureLod_cubemapArray(reflectionProbes, vec4(L, 0.0), lod).rgb;
-
-      /* Clamped brightness. */
-      float luma = max(1e-8, max_v3(l_col));
-
-      /* For artistic freedom this should be read from the scene/reflection probe.
-       * Note: Eevee-legacy read the firefly_factor from gi_glossy_clamp. */
-      const float firefly_factor = 1e16;
-      l_col *= 1.0 - max(0.0, luma - firefly_factor) / luma;
-
-      out_radiance += l_col * NL;
-      weight += NL;
-    }
+    /* TODO: for artistic freedom want to read this from the reflection probe. That can be added as
+     * part of the reflection probe patch. */
+    const float intensity_factor = 1.0;
+    out_specular += vec3(intensity_factor * l_col);
   }
-
-  /* TODO: for artistic freedom want to read this from the reflection probe. That can be added as
-   * part of the reflection probe patch. */
-  const float intensity_factor = 1.0;
-  out_specular += vec3(intensity_factor * out_radiance / weight);
 }

source/blender/draw/engines/eevee_next/shaders/infos/eevee_deferred_info.hh

@@ -36,6 +36,7 @@ GPU_SHADER_CREATE_INFO(eevee_deferred_light)
     .sampler(1, ImageType::FLOAT_2D_ARRAY, "gbuffer_color_tx")
     .additional_info("eevee_shared",
                      "eevee_utility_texture",
+                     "eevee_sampling_data",
                      "eevee_light_data",
                      "eevee_reflection_probe_data",
                      "eevee_lightprobe_data",