summaryrefslogtreecommitdiff
path: root/servers/rendering/renderer_rd/shaders/effects
diff options
context:
space:
mode:
Diffstat (limited to 'servers/rendering/renderer_rd/shaders/effects')
-rw-r--r--servers/rendering/renderer_rd/shaders/effects/SCsub2
-rw-r--r--servers/rendering/renderer_rd/shaders/effects/blur_raster.glsl51
-rw-r--r--servers/rendering/renderer_rd/shaders/effects/blur_raster_inc.glsl2
-rw-r--r--servers/rendering/renderer_rd/shaders/effects/bokeh_dof.glsl25
-rw-r--r--servers/rendering/renderer_rd/shaders/effects/bokeh_dof_inc.glsl5
-rw-r--r--servers/rendering/renderer_rd/shaders/effects/bokeh_dof_raster.glsl29
-rw-r--r--servers/rendering/renderer_rd/shaders/effects/copy.glsl42
-rw-r--r--servers/rendering/renderer_rd/shaders/effects/copy_to_fb.glsl61
-rw-r--r--servers/rendering/renderer_rd/shaders/effects/cubemap_roughness_inc.glsl11
-rw-r--r--servers/rendering/renderer_rd/shaders/effects/fsr_upscale.glsl173
-rw-r--r--servers/rendering/renderer_rd/shaders/effects/luminance_reduce.glsl82
-rw-r--r--servers/rendering/renderer_rd/shaders/effects/luminance_reduce_raster.glsl74
-rw-r--r--servers/rendering/renderer_rd/shaders/effects/luminance_reduce_raster_inc.glsl11
-rw-r--r--servers/rendering/renderer_rd/shaders/effects/screen_space_reflection.glsl87
-rw-r--r--servers/rendering/renderer_rd/shaders/effects/subsurface_scattering.glsl189
-rw-r--r--servers/rendering/renderer_rd/shaders/effects/taa_resolve.glsl396
-rw-r--r--servers/rendering/renderer_rd/shaders/effects/tonemap.glsl32
-rw-r--r--servers/rendering/renderer_rd/shaders/effects/vrs.glsl14
18 files changed, 1142 insertions, 144 deletions
diff --git a/servers/rendering/renderer_rd/shaders/effects/SCsub b/servers/rendering/renderer_rd/shaders/effects/SCsub
index 741da8fe69..f06a2d86e2 100644
--- a/servers/rendering/renderer_rd/shaders/effects/SCsub
+++ b/servers/rendering/renderer_rd/shaders/effects/SCsub
@@ -4,7 +4,7 @@ Import("env")
if "RD_GLSL" in env["BUILDERS"]:
# find all include files
- gl_include_files = [str(f) for f in Glob("*_inc.glsl")]
+ gl_include_files = [str(f) for f in Glob("*_inc.glsl")] + [str(f) for f in Glob("../*_inc.glsl")]
# find all shader code(all glsl files excluding our include files)
glsl_files = [str(f) for f in Glob("*.glsl") if str(f) not in gl_include_files]
diff --git a/servers/rendering/renderer_rd/shaders/effects/blur_raster.glsl b/servers/rendering/renderer_rd/shaders/effects/blur_raster.glsl
index 96f5c3e9f2..31aabbe9d2 100644
--- a/servers/rendering/renderer_rd/shaders/effects/blur_raster.glsl
+++ b/servers/rendering/renderer_rd/shaders/effects/blur_raster.glsl
@@ -53,30 +53,31 @@ void main() {
#ifdef MODE_GAUSSIAN_BLUR
- // Simpler blur uses SIGMA2 for the gaussian kernel for a stronger effect
-
- // note, for blur blur.luminance_multiplier is irrelavant, we would be multiplying and then dividing by this amount.
-
- if (bool(blur.flags & FLAG_HORIZONTAL)) {
- vec2 pix_size = blur.pixel_size;
- pix_size *= 0.5; //reading from larger buffer, so use more samples
- vec4 color = texture(source_color, uv_interp + vec2(0.0, 0.0) * pix_size) * 0.214607;
- color += texture(source_color, uv_interp + vec2(1.0, 0.0) * pix_size) * 0.189879;
- color += texture(source_color, uv_interp + vec2(2.0, 0.0) * pix_size) * 0.131514;
- color += texture(source_color, uv_interp + vec2(3.0, 0.0) * pix_size) * 0.071303;
- color += texture(source_color, uv_interp + vec2(-1.0, 0.0) * pix_size) * 0.189879;
- color += texture(source_color, uv_interp + vec2(-2.0, 0.0) * pix_size) * 0.131514;
- color += texture(source_color, uv_interp + vec2(-3.0, 0.0) * pix_size) * 0.071303;
- frag_color = color;
- } else {
- vec2 pix_size = blur.pixel_size;
- vec4 color = texture(source_color, uv_interp + vec2(0.0, 0.0) * pix_size) * 0.38774;
- color += texture(source_color, uv_interp + vec2(0.0, 1.0) * pix_size) * 0.24477;
- color += texture(source_color, uv_interp + vec2(0.0, 2.0) * pix_size) * 0.06136;
- color += texture(source_color, uv_interp + vec2(0.0, -1.0) * pix_size) * 0.24477;
- color += texture(source_color, uv_interp + vec2(0.0, -2.0) * pix_size) * 0.06136;
- frag_color = color;
- }
+ // For Gaussian Blur we use 13 taps in a single pass instead of 12 taps over 2 passes.
+ // This minimizes the number of times we change framebuffers which is very important for mobile.
+ // Source: http://www.iryoku.com/next-generation-post-processing-in-call-of-duty-advanced-warfare
+ vec4 A = texture(source_color, uv_interp + blur.pixel_size * vec2(-1.0, -1.0));
+ vec4 B = texture(source_color, uv_interp + blur.pixel_size * vec2(0.0, -1.0));
+ vec4 C = texture(source_color, uv_interp + blur.pixel_size * vec2(1.0, -1.0));
+ vec4 D = texture(source_color, uv_interp + blur.pixel_size * vec2(-0.5, -0.5));
+ vec4 E = texture(source_color, uv_interp + blur.pixel_size * vec2(0.5, -0.5));
+ vec4 F = texture(source_color, uv_interp + blur.pixel_size * vec2(-1.0, 0.0));
+ vec4 G = texture(source_color, uv_interp);
+ vec4 H = texture(source_color, uv_interp + blur.pixel_size * vec2(1.0, 0.0));
+ vec4 I = texture(source_color, uv_interp + blur.pixel_size * vec2(-0.5, 0.5));
+ vec4 J = texture(source_color, uv_interp + blur.pixel_size * vec2(0.5, 0.5));
+ vec4 K = texture(source_color, uv_interp + blur.pixel_size * vec2(-1.0, 1.0));
+ vec4 L = texture(source_color, uv_interp + blur.pixel_size * vec2(0.0, 1.0));
+ vec4 M = texture(source_color, uv_interp + blur.pixel_size * vec2(1.0, 1.0));
+
+ float base_weight = 0.5 / 4.0;
+ float lesser_weight = 0.125 / 4.0;
+
+ frag_color = (D + E + I + J) * base_weight;
+ frag_color += (A + B + G + F) * lesser_weight;
+ frag_color += (B + C + H + G) * lesser_weight;
+ frag_color += (F + G + L + K) * lesser_weight;
+ frag_color += (G + H + M + L) * lesser_weight;
#endif
#ifdef MODE_GAUSSIAN_GLOW
@@ -129,7 +130,7 @@ void main() {
#ifdef GLOW_USE_AUTO_EXPOSURE
- frag_color /= texelFetch(source_auto_exposure, ivec2(0, 0), 0).r / blur.glow_auto_exposure_grey;
+ frag_color /= texelFetch(source_auto_exposure, ivec2(0, 0), 0).r / blur.glow_auto_exposure_scale;
#endif
frag_color *= blur.glow_exposure;
diff --git a/servers/rendering/renderer_rd/shaders/effects/blur_raster_inc.glsl b/servers/rendering/renderer_rd/shaders/effects/blur_raster_inc.glsl
index 730504571a..06ca198f37 100644
--- a/servers/rendering/renderer_rd/shaders/effects/blur_raster_inc.glsl
+++ b/servers/rendering/renderer_rd/shaders/effects/blur_raster_inc.glsl
@@ -16,7 +16,7 @@ layout(push_constant, std430) uniform Blur {
float glow_exposure; // 04 - 36
float glow_white; // 04 - 40
float glow_luminance_cap; // 04 - 44
- float glow_auto_exposure_grey; // 04 - 48
+ float glow_auto_exposure_scale; // 04 - 48
float luminance_multiplier; // 04 - 52
float res1; // 04 - 56
diff --git a/servers/rendering/renderer_rd/shaders/effects/bokeh_dof.glsl b/servers/rendering/renderer_rd/shaders/effects/bokeh_dof.glsl
index 0438671dd2..fe770ac065 100644
--- a/servers/rendering/renderer_rd/shaders/effects/bokeh_dof.glsl
+++ b/servers/rendering/renderer_rd/shaders/effects/bokeh_dof.glsl
@@ -30,7 +30,7 @@ layout(set = 1, binding = 0) uniform sampler2D source_bokeh;
#ifdef MODE_GEN_BLUR_SIZE
float get_depth_at_pos(vec2 uv) {
- float depth = textureLod(source_depth, uv, 0.0).x;
+ float depth = textureLod(source_depth, uv, 0.0).x * 2.0 - 1.0;
if (params.orthogonal) {
depth = ((depth + (params.z_far + params.z_near) / (params.z_far - params.z_near)) * (params.z_far - params.z_near)) / 2.0;
} else {
@@ -41,11 +41,25 @@ float get_depth_at_pos(vec2 uv) {
float get_blur_size(float depth) {
if (params.blur_near_active && depth < params.blur_near_begin) {
- return -(1.0 - smoothstep(params.blur_near_end, params.blur_near_begin, depth)) * params.blur_size - DEPTH_GAP; //near blur is negative
+ if (params.use_physical_near) {
+ // Physically-based.
+ float d = abs(params.blur_near_begin - depth);
+ return -(d / (params.blur_near_begin - d)) * params.blur_size_near - DEPTH_GAP; // Near blur is negative.
+ } else {
+ // Non-physically-based.
+ return -(1.0 - smoothstep(params.blur_near_end, params.blur_near_begin, depth)) * params.blur_size - DEPTH_GAP; // Near blur is negative.
+ }
}
if (params.blur_far_active && depth > params.blur_far_begin) {
- return smoothstep(params.blur_far_begin, params.blur_far_end, depth) * params.blur_size + DEPTH_GAP;
+ if (params.use_physical_far) {
+ // Physically-based.
+ float d = abs(params.blur_far_begin - depth);
+ return (d / (params.blur_far_begin + d)) * params.blur_size_far + DEPTH_GAP;
+ } else {
+ // Non-physically-based.
+ return smoothstep(params.blur_far_begin, params.blur_far_end, depth) * params.blur_size + DEPTH_GAP;
+ }
}
return 0.0;
@@ -172,6 +186,7 @@ void main() {
uv += pixel_size * 0.5; //half pixel to read centers
vec4 color = texture(color_texture, uv);
+ float initial_blur = color.a;
float accum = 1.0;
float radius = params.blur_scale;
@@ -179,8 +194,8 @@ void main() {
vec2 suv = uv + vec2(cos(ang), sin(ang)) * pixel_size * radius;
vec4 sample_color = texture(color_texture, suv);
float sample_size = abs(sample_color.a);
- if (sample_color.a > color.a) {
- sample_size = clamp(sample_size, 0.0, abs(color.a) * 2.0);
+ if (sample_color.a > initial_blur) {
+ sample_size = clamp(sample_size, 0.0, abs(initial_blur) * 2.0);
}
float m = smoothstep(radius - 0.5, radius + 0.5, sample_size);
diff --git a/servers/rendering/renderer_rd/shaders/effects/bokeh_dof_inc.glsl b/servers/rendering/renderer_rd/shaders/effects/bokeh_dof_inc.glsl
index b90a527554..4a2b0edc18 100644
--- a/servers/rendering/renderer_rd/shaders/effects/bokeh_dof_inc.glsl
+++ b/servers/rendering/renderer_rd/shaders/effects/bokeh_dof_inc.glsl
@@ -20,6 +20,11 @@ layout(push_constant, std430) uniform Params {
bool use_jitter;
float jitter_seed;
+ bool use_physical_near;
+ bool use_physical_far;
+
+ float blur_size_near;
+ float blur_size_far;
uint pad[2];
}
params;
diff --git a/servers/rendering/renderer_rd/shaders/effects/bokeh_dof_raster.glsl b/servers/rendering/renderer_rd/shaders/effects/bokeh_dof_raster.glsl
index a3b3938ee9..1b487835d2 100644
--- a/servers/rendering/renderer_rd/shaders/effects/bokeh_dof_raster.glsl
+++ b/servers/rendering/renderer_rd/shaders/effects/bokeh_dof_raster.glsl
@@ -52,7 +52,7 @@ layout(set = 2, binding = 0) uniform sampler2D original_weight;
#ifdef MODE_GEN_BLUR_SIZE
float get_depth_at_pos(vec2 uv) {
- float depth = textureLod(source_depth, uv, 0.0).x;
+ float depth = textureLod(source_depth, uv, 0.0).x * 2.0 - 1.0;
if (params.orthogonal) {
depth = ((depth + (params.z_far + params.z_near) / (params.z_far - params.z_near)) * (params.z_far - params.z_near)) / 2.0;
} else {
@@ -63,11 +63,25 @@ float get_depth_at_pos(vec2 uv) {
float get_blur_size(float depth) {
if (params.blur_near_active && depth < params.blur_near_begin) {
- return -(1.0 - smoothstep(params.blur_near_end, params.blur_near_begin, depth)) * params.blur_size - DEPTH_GAP; //near blur is negative
+ if (params.use_physical_near) {
+ // Physically-based.
+ float d = abs(params.blur_near_begin - depth);
+ return -(d / (params.blur_near_begin - d)) * params.blur_size_near - DEPTH_GAP; // Near blur is negative.
+ } else {
+ // Non-physically-based.
+ return -(1.0 - smoothstep(params.blur_near_end, params.blur_near_begin, depth)) * params.blur_size - DEPTH_GAP; // Near blur is negative.
+ }
}
if (params.blur_far_active && depth > params.blur_far_begin) {
- return smoothstep(params.blur_far_begin, params.blur_far_end, depth) * params.blur_size + DEPTH_GAP;
+ if (params.use_physical_far) {
+ // Physically-based.
+ float d = abs(params.blur_far_begin - depth);
+ return (d / (params.blur_far_begin + d)) * params.blur_size_far + DEPTH_GAP;
+ } else {
+ // Non-physically-based.
+ return smoothstep(params.blur_far_begin, params.blur_far_end, depth) * params.blur_size + DEPTH_GAP;
+ }
}
return 0.0;
@@ -207,12 +221,9 @@ void main() {
vec4 sample_color = texture(source_color, uv_adj);
sample_color.a = texture(source_weight, uv_adj).r;
- float limit;
-
- if (sample_color.a < color.a) {
- limit = abs(sample_color.a);
- } else {
- limit = abs(color.a);
+ float limit = abs(sample_color.a);
+ if (sample_color.a > color.a) {
+ limit = clamp(limit, 0.0, abs(color.a) * 2.0);
}
limit -= DEPTH_GAP;
diff --git a/servers/rendering/renderer_rd/shaders/effects/copy.glsl b/servers/rendering/renderer_rd/shaders/effects/copy.glsl
index 3a4ef86ef0..3a82861057 100644
--- a/servers/rendering/renderer_rd/shaders/effects/copy.glsl
+++ b/servers/rendering/renderer_rd/shaders/effects/copy.glsl
@@ -14,8 +14,7 @@ layout(local_size_x = 8, local_size_y = 8, local_size_z = 1) in;
#define FLAG_FLIP_Y (1 << 5)
#define FLAG_FORCE_LUMINANCE (1 << 6)
#define FLAG_COPY_ALL_SOURCE (1 << 7)
-#define FLAG_HIGH_QUALITY_GLOW (1 << 8)
-#define FLAG_ALPHA_TO_ONE (1 << 9)
+#define FLAG_ALPHA_TO_ONE (1 << 8)
layout(push_constant, std430) uniform Params {
ivec4 section;
@@ -31,7 +30,7 @@ layout(push_constant, std430) uniform Params {
float glow_exposure;
float glow_white;
float glow_luminance_cap;
- float glow_auto_exposure_grey;
+ float glow_auto_exposure_scale;
// DOF.
float camera_z_far;
float camera_z_near;
@@ -93,25 +92,14 @@ void main() {
#ifdef MODE_GAUSSIAN_BLUR
// First pass copy texture into 16x16 local memory for every 8x8 thread block
- vec2 quad_center_uv = clamp(vec2(gl_GlobalInvocationID.xy + gl_LocalInvocationID.xy - 3.5) / params.section.zw, vec2(0.5 / params.section.zw), vec2(1.0 - 1.5 / params.section.zw));
+ vec2 quad_center_uv = clamp(vec2(params.section.xy + gl_GlobalInvocationID.xy + gl_LocalInvocationID.xy - 3.5) / params.section.zw, vec2(0.5 / params.section.zw), vec2(1.0 - 1.5 / params.section.zw));
uint dest_index = gl_LocalInvocationID.x * 2 + gl_LocalInvocationID.y * 2 * 16;
-#ifdef MODE_GLOW
- if (bool(params.flags & FLAG_HIGH_QUALITY_GLOW)) {
- vec2 quad_offset_uv = clamp((vec2(gl_GlobalInvocationID.xy + gl_LocalInvocationID.xy - 3.0)) / params.section.zw, vec2(0.5 / params.section.zw), vec2(1.0 - 1.5 / params.section.zw));
-
- local_cache[dest_index] = (textureLod(source_color, quad_center_uv, 0) + textureLod(source_color, quad_offset_uv, 0)) * 0.5;
- local_cache[dest_index + 1] = (textureLod(source_color, quad_center_uv + vec2(1.0 / params.section.z, 0.0), 0) + textureLod(source_color, quad_offset_uv + vec2(1.0 / params.section.z, 0.0), 0)) * 0.5;
- local_cache[dest_index + 16] = (textureLod(source_color, quad_center_uv + vec2(0.0, 1.0 / params.section.w), 0) + textureLod(source_color, quad_offset_uv + vec2(0.0, 1.0 / params.section.w), 0)) * 0.5;
- local_cache[dest_index + 16 + 1] = (textureLod(source_color, quad_center_uv + vec2(1.0 / params.section.zw), 0) + textureLod(source_color, quad_offset_uv + vec2(1.0 / params.section.zw), 0)) * 0.5;
- } else
-#endif
- {
- local_cache[dest_index] = textureLod(source_color, quad_center_uv, 0);
- local_cache[dest_index + 1] = textureLod(source_color, quad_center_uv + vec2(1.0 / params.section.z, 0.0), 0);
- local_cache[dest_index + 16] = textureLod(source_color, quad_center_uv + vec2(0.0, 1.0 / params.section.w), 0);
- local_cache[dest_index + 16 + 1] = textureLod(source_color, quad_center_uv + vec2(1.0 / params.section.zw), 0);
- }
+ local_cache[dest_index] = textureLod(source_color, quad_center_uv, 0);
+ local_cache[dest_index + 1] = textureLod(source_color, quad_center_uv + vec2(1.0 / params.section.z, 0.0), 0);
+ local_cache[dest_index + 16] = textureLod(source_color, quad_center_uv + vec2(0.0, 1.0 / params.section.w), 0);
+ local_cache[dest_index + 16 + 1] = textureLod(source_color, quad_center_uv + vec2(1.0 / params.section.zw), 0);
+
#ifdef MODE_GLOW
if (bool(params.flags & FLAG_GLOW_FIRST_PASS)) {
// Tonemap initial samples to reduce weight of fireflies: https://graphicrants.blogspot.com/2013/12/tone-mapping.html
@@ -185,7 +173,7 @@ void main() {
if (bool(params.flags & FLAG_GLOW_FIRST_PASS)) {
#ifdef GLOW_USE_AUTO_EXPOSURE
- color /= texelFetch(source_auto_exposure, ivec2(0, 0), 0).r / params.glow_auto_exposure_grey;
+ color /= texelFetch(source_auto_exposure, ivec2(0, 0), 0).r / params.glow_auto_exposure_scale;
#endif
color *= params.glow_exposure;
@@ -194,10 +182,10 @@ void main() {
color = min(color * feedback, vec4(params.glow_luminance_cap));
}
-#endif
+#endif // MODE_GLOW
imageStore(dest_buffer, pos + params.target, color);
-#endif
+#endif // MODE_GAUSSIAN_BLUR
#ifdef MODE_SIMPLE_COPY
@@ -227,7 +215,7 @@ void main() {
imageStore(dest_buffer, pos + params.target, color);
-#endif
+#endif // MODE_SIMPLE_COPY
#ifdef MODE_SIMPLE_COPY_DEPTH
@@ -239,7 +227,7 @@ void main() {
imageStore(dest_buffer, pos + params.target, vec4(color.r));
-#endif
+#endif // MODE_SIMPLE_COPY_DEPTH
#ifdef MODE_LINEARIZE_DEPTH_COPY
@@ -253,7 +241,7 @@ void main() {
}
imageStore(dest_buffer, pos + params.target, color);
-#endif
+#endif // MODE_LINEARIZE_DEPTH_COPY
#if defined(MODE_CUBEMAP_TO_PANORAMA) || defined(MODE_CUBEMAP_ARRAY_TO_PANORAMA)
@@ -276,7 +264,7 @@ void main() {
vec4 color = textureLod(source_color, vec4(normal, params.camera_z_far), 0.0); //the biggest the lod the least the acne
#endif
imageStore(dest_buffer, pos + params.target, color);
-#endif
+#endif // defined(MODE_CUBEMAP_TO_PANORAMA) || defined(MODE_CUBEMAP_ARRAY_TO_PANORAMA)
#ifdef MODE_SET_COLOR
imageStore(dest_buffer, pos + params.target, params.set_color);
diff --git a/servers/rendering/renderer_rd/shaders/effects/copy_to_fb.glsl b/servers/rendering/renderer_rd/shaders/effects/copy_to_fb.glsl
index 1c17eabb56..6137224162 100644
--- a/servers/rendering/renderer_rd/shaders/effects/copy_to_fb.glsl
+++ b/servers/rendering/renderer_rd/shaders/effects/copy_to_fb.glsl
@@ -13,6 +13,14 @@
#endif // has_VK_KHR_multiview
#endif //MULTIVIEW
+#define FLAG_FLIP_Y (1 << 0)
+#define FLAG_USE_SECTION (1 << 1)
+#define FLAG_FORCE_LUMINANCE (1 << 2)
+#define FLAG_ALPHA_TO_ZERO (1 << 3)
+#define FLAG_SRGB (1 << 4)
+#define FLAG_ALPHA_TO_ONE (1 << 5)
+#define FLAG_LINEAR (1 << 6)
+
#ifdef MULTIVIEW
layout(location = 0) out vec3 uv_interp;
#else
@@ -22,11 +30,10 @@ layout(location = 0) out vec2 uv_interp;
layout(push_constant, std430) uniform Params {
vec4 section;
vec2 pixel_size;
- bool flip_y;
- bool use_section;
+ float luminance_multiplier;
+ uint flags;
- bool force_luminance;
- uint pad[3];
+ vec4 color;
}
params;
@@ -37,13 +44,13 @@ void main() {
uv_interp.z = ViewIndex;
#endif
vec2 vpos = uv_interp.xy;
- if (params.use_section) {
+ if (bool(params.flags & FLAG_USE_SECTION)) {
vpos = params.section.xy + vpos * params.section.zw;
}
gl_Position = vec4(vpos * 2.0 - 1.0, 0.0, 1.0);
- if (params.flip_y) {
+ if (bool(params.flags & FLAG_FLIP_Y)) {
uv_interp.y = 1.0 - uv_interp.y;
}
}
@@ -63,19 +70,25 @@ void main() {
#endif // has_VK_KHR_multiview
#endif //MULTIVIEW
+#define FLAG_FLIP_Y (1 << 0)
+#define FLAG_USE_SECTION (1 << 1)
+#define FLAG_FORCE_LUMINANCE (1 << 2)
+#define FLAG_ALPHA_TO_ZERO (1 << 3)
+#define FLAG_SRGB (1 << 4)
+#define FLAG_ALPHA_TO_ONE (1 << 5)
+#define FLAG_LINEAR (1 << 6)
+
layout(push_constant, std430) uniform Params {
vec4 section;
vec2 pixel_size;
- bool flip_y;
- bool use_section;
+ float luminance_multiplier;
+ uint flags;
- bool force_luminance;
- bool alpha_to_zero;
- bool srgb;
- uint pad;
+ vec4 color;
}
params;
+#ifndef MODE_SET_COLOR
#ifdef MULTIVIEW
layout(location = 0) in vec3 uv_interp;
#else
@@ -94,6 +107,7 @@ layout(set = 0, binding = 0) uniform sampler2D source_color;
layout(set = 1, binding = 0) uniform sampler2D source_color2;
#endif /* MODE_TWO_SOURCES */
#endif /* MULTIVIEW */
+#endif /* !SET_COLOR */
layout(location = 0) out vec4 frag_color;
@@ -104,7 +118,15 @@ vec3 linear_to_srgb(vec3 color) {
return mix((vec3(1.0f) + a) * pow(color.rgb, vec3(1.0f / 2.4f)) - a, 12.92f * color.rgb, lessThan(color.rgb, vec3(0.0031308f)));
}
+vec3 srgb_to_linear(vec3 color) {
+ return mix(pow((color.rgb + vec3(0.055)) * (1.0 / (1.0 + 0.055)), vec3(2.4)), color.rgb * (1.0 / 12.92), lessThan(color.rgb, vec3(0.04045)));
+}
+
void main() {
+#ifdef MODE_SET_COLOR
+ frag_color = params.color;
+#else
+
#ifdef MULTIVIEW
vec3 uv = uv_interp;
#else
@@ -155,15 +177,22 @@ void main() {
#endif /* MODE_TWO_SOURCES */
#endif /* MULTIVIEW */
- if (params.force_luminance) {
+ if (bool(params.flags & FLAG_FORCE_LUMINANCE)) {
color.rgb = vec3(max(max(color.r, color.g), color.b));
}
- if (params.alpha_to_zero) {
+ if (bool(params.flags & FLAG_ALPHA_TO_ZERO)) {
color.rgb *= color.a;
}
- if (params.srgb) {
+ if (bool(params.flags & FLAG_SRGB)) {
color.rgb = linear_to_srgb(color.rgb);
}
+ if (bool(params.flags & FLAG_ALPHA_TO_ONE)) {
+ color.a = 1.0;
+ }
+ if (bool(params.flags & FLAG_LINEAR)) {
+ color.rgb = srgb_to_linear(color.rgb);
+ }
- frag_color = color;
+ frag_color = color / params.luminance_multiplier;
+#endif // MODE_SET_COLOR
}
diff --git a/servers/rendering/renderer_rd/shaders/effects/cubemap_roughness_inc.glsl b/servers/rendering/renderer_rd/shaders/effects/cubemap_roughness_inc.glsl
index 1bee428a6f..c0597fe3f3 100644
--- a/servers/rendering/renderer_rd/shaders/effects/cubemap_roughness_inc.glsl
+++ b/servers/rendering/renderer_rd/shaders/effects/cubemap_roughness_inc.glsl
@@ -70,17 +70,6 @@ float DistributionGGX(float NdotH, float roughness4) {
return roughness4 / denom;
}
-// https://graphicrants.blogspot.com.au/2013/08/specular-brdf-reference.html
-float GGX(float NdotV, float a) {
- float k = a / 2.0;
- return NdotV / (NdotV * (1.0 - k) + k);
-}
-
-// https://graphicrants.blogspot.com.au/2013/08/specular-brdf-reference.html
-float G_Smith(float a, float nDotV, float nDotL) {
- return GGX(nDotL, a * a) * GGX(nDotV, a * a);
-}
-
float radicalInverse_VdC(uint bits) {
bits = (bits << 16u) | (bits >> 16u);
bits = ((bits & 0x55555555u) << 1u) | ((bits & 0xAAAAAAAAu) >> 1u);
diff --git a/servers/rendering/renderer_rd/shaders/effects/fsr_upscale.glsl b/servers/rendering/renderer_rd/shaders/effects/fsr_upscale.glsl
new file mode 100644
index 0000000000..221e97bece
--- /dev/null
+++ b/servers/rendering/renderer_rd/shaders/effects/fsr_upscale.glsl
@@ -0,0 +1,173 @@
+/**************************************************************************/
+/* fsr_upscale.glsl */
+/**************************************************************************/
+/* This file is part of: */
+/* GODOT ENGINE */
+/* https://godotengine.org */
+/**************************************************************************/
+/* Copyright (c) 2014-present Godot Engine contributors (see AUTHORS.md). */
+/* Copyright (c) 2007-2014 Juan Linietsky, Ariel Manzur. */
+/* */
+/* Permission is hereby granted, free of charge, to any person obtaining */
+/* a copy of this software and associated documentation files (the */
+/* "Software"), to deal in the Software without restriction, including */
+/* without limitation the rights to use, copy, modify, merge, publish, */
+/* distribute, sublicense, and/or sell copies of the Software, and to */
+/* permit persons to whom the Software is furnished to do so, subject to */
+/* the following conditions: */
+/* */
+/* The above copyright notice and this permission notice shall be */
+/* included in all copies or substantial portions of the Software. */
+/* */
+/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, */
+/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF */
+/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. */
+/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY */
+/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, */
+/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE */
+/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */
+/**************************************************************************/
+
+#[compute]
+
+#version 450
+
+#VERSION_DEFINES
+
+#define A_GPU
+#define A_GLSL
+
+#ifdef MODE_FSR_UPSCALE_NORMAL
+
+#define A_HALF
+
+#endif
+
+#include "thirdparty/amd-fsr/ffx_a.h"
+
+layout(local_size_x = 64, local_size_y = 1, local_size_z = 1) in;
+
+layout(rgba16f, set = 1, binding = 0) uniform restrict writeonly image2D fsr_image;
+layout(set = 0, binding = 0) uniform sampler2D source_image;
+
+#define FSR_UPSCALE_PASS_TYPE_EASU 0
+#define FSR_UPSCALE_PASS_TYPE_RCAS 1
+
+layout(push_constant, std430) uniform Params {
+ float resolution_width;
+ float resolution_height;
+ float upscaled_width;
+ float upscaled_height;
+ float sharpness;
+ int pass;
+}
+params;
+
+AU4 Const0, Const1, Const2, Const3;
+
+#ifdef MODE_FSR_UPSCALE_FALLBACK
+
+#define FSR_EASU_F
+AF4 FsrEasuRF(AF2 p) {
+ AF4 res = textureGather(source_image, p, 0);
+ return res;
+}
+AF4 FsrEasuGF(AF2 p) {
+ AF4 res = textureGather(source_image, p, 1);
+ return res;
+}
+AF4 FsrEasuBF(AF2 p) {
+ AF4 res = textureGather(source_image, p, 2);
+ return res;
+}
+
+#define FSR_RCAS_F
+AF4 FsrRcasLoadF(ASU2 p) {
+ return AF4(texelFetch(source_image, ASU2(p), 0));
+}
+void FsrRcasInputF(inout AF1 r, inout AF1 g, inout AF1 b) {}
+
+#else
+
+#define FSR_EASU_H
+AH4 FsrEasuRH(AF2 p) {
+ AH4 res = AH4(textureGather(source_image, p, 0));
+ return res;
+}
+AH4 FsrEasuGH(AF2 p) {
+ AH4 res = AH4(textureGather(source_image, p, 1));
+ return res;
+}
+AH4 FsrEasuBH(AF2 p) {
+ AH4 res = AH4(textureGather(source_image, p, 2));
+ return res;
+}
+
+#define FSR_RCAS_H
+AH4 FsrRcasLoadH(ASW2 p) {
+ return AH4(texelFetch(source_image, ASU2(p), 0));
+}
+void FsrRcasInputH(inout AH1 r, inout AH1 g, inout AH1 b) {}
+
+#endif
+
+#include "thirdparty/amd-fsr/ffx_fsr1.h"
+
+void fsr_easu_pass(AU2 pos) {
+#ifdef MODE_FSR_UPSCALE_NORMAL
+
+ AH3 Gamma2Color = AH3(0, 0, 0);
+ FsrEasuH(Gamma2Color, pos, Const0, Const1, Const2, Const3);
+ imageStore(fsr_image, ASU2(pos), AH4(Gamma2Color, 1));
+
+#else
+
+ AF3 Gamma2Color = AF3(0, 0, 0);
+ FsrEasuF(Gamma2Color, pos, Const0, Const1, Const2, Const3);
+ imageStore(fsr_image, ASU2(pos), AF4(Gamma2Color, 1));
+
+#endif
+}
+
+void fsr_rcas_pass(AU2 pos) {
+#ifdef MODE_FSR_UPSCALE_NORMAL
+
+ AH3 Gamma2Color = AH3(0, 0, 0);
+ FsrRcasH(Gamma2Color.r, Gamma2Color.g, Gamma2Color.b, pos, Const0);
+ imageStore(fsr_image, ASU2(pos), AH4(Gamma2Color, 1));
+
+#else
+
+ AF3 Gamma2Color = AF3(0, 0, 0);
+ FsrRcasF(Gamma2Color.r, Gamma2Color.g, Gamma2Color.b, pos, Const0);
+ imageStore(fsr_image, ASU2(pos), AF4(Gamma2Color, 1));
+
+#endif
+}
+
+void fsr_pass(AU2 pos) {
+ if (params.pass == FSR_UPSCALE_PASS_TYPE_EASU) {
+ fsr_easu_pass(pos);
+ } else if (params.pass == FSR_UPSCALE_PASS_TYPE_RCAS) {
+ fsr_rcas_pass(pos);
+ }
+}
+
+void main() {
+ // Clang does not like unused functions. If ffx_a.h is included in the binary, clang will throw a fit and not compile so we must configure FSR in this shader
+ if (params.pass == FSR_UPSCALE_PASS_TYPE_EASU) {
+ FsrEasuCon(Const0, Const1, Const2, Const3, params.resolution_width, params.resolution_height, params.resolution_width, params.resolution_height, params.upscaled_width, params.upscaled_height);
+ } else if (params.pass == FSR_UPSCALE_PASS_TYPE_RCAS) {
+ FsrRcasCon(Const0, params.sharpness);
+ }
+
+ AU2 gxy = ARmp8x8(gl_LocalInvocationID.x) + AU2(gl_WorkGroupID.x << 4u, gl_WorkGroupID.y << 4u);
+
+ fsr_pass(gxy);
+ gxy.x += 8u;
+ fsr_pass(gxy);
+ gxy.y += 8u;
+ fsr_pass(gxy);
+ gxy.x -= 8u;
+ fsr_pass(gxy);
+}
diff --git a/servers/rendering/renderer_rd/shaders/effects/luminance_reduce.glsl b/servers/rendering/renderer_rd/shaders/effects/luminance_reduce.glsl
new file mode 100644
index 0000000000..0ee4cf6e31
--- /dev/null
+++ b/servers/rendering/renderer_rd/shaders/effects/luminance_reduce.glsl
@@ -0,0 +1,82 @@
+#[compute]
+
+#version 450
+
+#VERSION_DEFINES
+
+#define BLOCK_SIZE 8
+
+layout(local_size_x = BLOCK_SIZE, local_size_y = BLOCK_SIZE, local_size_z = 1) in;
+
+shared float tmp_data[BLOCK_SIZE * BLOCK_SIZE];
+
+#ifdef READ_TEXTURE
+
+//use for main texture
+layout(set = 0, binding = 0) uniform sampler2D source_texture;
+
+#else
+
+//use for intermediate textures
+layout(r32f, set = 0, binding = 0) uniform restrict readonly image2D source_luminance;
+
+#endif
+
+layout(r32f, set = 1, binding = 0) uniform restrict writeonly image2D dest_luminance;
+
+#ifdef WRITE_LUMINANCE
+layout(set = 2, binding = 0) uniform sampler2D prev_luminance;
+#endif
+
+layout(push_constant, std430) uniform Params {
+ ivec2 source_size;
+ float max_luminance;
+ float min_luminance;
+ float exposure_adjust;
+ float pad[3];
+}
+params;
+
+void main() {
+ uint t = gl_LocalInvocationID.y * BLOCK_SIZE + gl_LocalInvocationID.x;
+ ivec2 pos = ivec2(gl_GlobalInvocationID.xy);
+
+ if (any(lessThan(pos, params.source_size))) {
+#ifdef READ_TEXTURE
+ vec3 v = texelFetch(source_texture, pos, 0).rgb;
+ tmp_data[t] = max(v.r, max(v.g, v.b));
+#else
+ tmp_data[t] = imageLoad(source_luminance, pos).r;
+#endif
+ } else {
+ tmp_data[t] = 0.0;
+ }
+
+ groupMemoryBarrier();
+ barrier();
+
+ uint size = (BLOCK_SIZE * BLOCK_SIZE) >> 1;
+
+ do {
+ if (t < size) {
+ tmp_data[t] += tmp_data[t + size];
+ }
+ groupMemoryBarrier();
+ barrier();
+
+ size >>= 1;
+ } while (size >= 1);
+
+ if (t == 0) {
+ //compute rect size
+ ivec2 rect_size = min(params.source_size - pos, ivec2(BLOCK_SIZE));
+ float avg = tmp_data[0] / float(rect_size.x * rect_size.y);
+ //float avg = tmp_data[0] / float(BLOCK_SIZE*BLOCK_SIZE);
+ pos /= ivec2(BLOCK_SIZE);
+#ifdef WRITE_LUMINANCE
+ float prev_lum = texelFetch(prev_luminance, ivec2(0, 0), 0).r; //1 pixel previous exposure
+ avg = clamp(prev_lum + (avg - prev_lum) * params.exposure_adjust, params.min_luminance, params.max_luminance);
+#endif
+ imageStore(dest_luminance, pos, vec4(avg));
+ }
+}
diff --git a/servers/rendering/renderer_rd/shaders/effects/luminance_reduce_raster.glsl b/servers/rendering/renderer_rd/shaders/effects/luminance_reduce_raster.glsl
new file mode 100644
index 0000000000..29ebd74a90
--- /dev/null
+++ b/servers/rendering/renderer_rd/shaders/effects/luminance_reduce_raster.glsl
@@ -0,0 +1,74 @@
+/* clang-format off */
+#[vertex]
+
+#version 450
+
+#VERSION_DEFINES
+
+#include "luminance_reduce_raster_inc.glsl"
+
+layout(location = 0) out vec2 uv_interp;
+/* clang-format on */
+
+void main() {
+ vec2 base_arr[4] = vec2[](vec2(0.0, 0.0), vec2(0.0, 1.0), vec2(1.0, 1.0), vec2(1.0, 0.0));
+ uv_interp = base_arr[gl_VertexIndex];
+
+ gl_Position = vec4(uv_interp * 2.0 - 1.0, 0.0, 1.0);
+}
+
+/* clang-format off */
+#[fragment]
+
+#version 450
+
+#VERSION_DEFINES
+
+#include "luminance_reduce_raster_inc.glsl"
+
+layout(location = 0) in vec2 uv_interp;
+/* clang-format on */
+
+layout(set = 0, binding = 0) uniform sampler2D source_exposure;
+
+#ifdef FINAL_PASS
+layout(set = 1, binding = 0) uniform sampler2D prev_luminance;
+#endif
+
+layout(location = 0) out highp float luminance;
+
+void main() {
+ ivec2 dest_pos = ivec2(uv_interp * settings.dest_size);
+ ivec2 src_pos = ivec2(uv_interp * settings.source_size);
+
+ ivec2 next_pos = (dest_pos + ivec2(1)) * settings.source_size / settings.dest_size;
+ next_pos = max(next_pos, src_pos + ivec2(1)); //so it at least reads one pixel
+
+ highp vec3 source_color = vec3(0.0);
+ for (int i = src_pos.x; i < next_pos.x; i++) {
+ for (int j = src_pos.y; j < next_pos.y; j++) {
+ source_color += texelFetch(source_exposure, ivec2(i, j), 0).rgb;
+ }
+ }
+
+ source_color /= float((next_pos.x - src_pos.x) * (next_pos.y - src_pos.y));
+
+#ifdef FIRST_PASS
+ luminance = max(source_color.r, max(source_color.g, source_color.b));
+
+ // This formula should be more "accurate" but gave an overexposed result when testing.
+ // Leaving it here so we can revisit it if we want.
+ // luminance = source_color.r * 0.21 + source_color.g * 0.71 + source_color.b * 0.07;
+#else
+ luminance = source_color.r;
+#endif
+
+#ifdef FINAL_PASS
+ // Obtain our target luminance
+ luminance = clamp(luminance, settings.min_luminance, settings.max_luminance);
+
+ // Now smooth to our transition
+ highp float prev_lum = texelFetch(prev_luminance, ivec2(0, 0), 0).r; //1 pixel previous luminance
+ luminance = prev_lum + (luminance - prev_lum) * clamp(settings.exposure_adjust, 0.0, 1.0);
+#endif
+}
diff --git a/servers/rendering/renderer_rd/shaders/effects/luminance_reduce_raster_inc.glsl b/servers/rendering/renderer_rd/shaders/effects/luminance_reduce_raster_inc.glsl
new file mode 100644
index 0000000000..b8860f6518
--- /dev/null
+++ b/servers/rendering/renderer_rd/shaders/effects/luminance_reduce_raster_inc.glsl
@@ -0,0 +1,11 @@
+
+layout(push_constant, std430) uniform PushConstant {
+ ivec2 source_size;
+ ivec2 dest_size;
+
+ float exposure_adjust;
+ float min_luminance;
+ float max_luminance;
+ uint pad1;
+}
+settings;
diff --git a/servers/rendering/renderer_rd/shaders/effects/screen_space_reflection.glsl b/servers/rendering/renderer_rd/shaders/effects/screen_space_reflection.glsl
index d85ab3af2e..631d1968b0 100644
--- a/servers/rendering/renderer_rd/shaders/effects/screen_space_reflection.glsl
+++ b/servers/rendering/renderer_rd/shaders/effects/screen_space_reflection.glsl
@@ -30,12 +30,7 @@ layout(push_constant, std430) uniform Params {
bool orthogonal;
float filter_mipmap_levels;
bool use_half_res;
- uint metallic_mask;
-
uint view_index;
- uint pad1;
- uint pad2;
- uint pad3;
}
params;
@@ -71,6 +66,19 @@ void main() {
vec4 normal_roughness = imageLoad(source_normal_roughness, ssC);
vec3 normal = normal_roughness.xyz * 2.0 - 1.0;
+ float roughness = normal_roughness.w;
+
+ // The roughness cutoff of 0.6 is chosen to match the roughness fadeout from GH-69828.
+ if (roughness > 0.6) {
+ // Do not compute SSR for rough materials to improve performance at the cost of
+ // subtle artifacting.
+#ifdef MODE_ROUGH
+ imageStore(blur_radius_image, ssC, vec4(0.0));
+#endif
+ imageStore(ssr_image, ssC, vec4(0.0));
+ return;
+ }
+
normal = normalize(normal);
normal.y = -normal.y; //because this code reads flipped
@@ -86,8 +94,6 @@ void main() {
imageStore(ssr_image, ssC, vec4(0.0));
return;
}
- //ray_dir = normalize(view_dir - normal * dot(normal,view_dir) * 2.0);
- //ray_dir = normalize(vec3(1.0, 1.0, -1.0));
////////////////
@@ -126,7 +132,7 @@ void main() {
// clip z and w advance to line advance
vec2 line_advance = normalize(line_dir); // down to pixel
- float step_size = length(line_advance) / length(line_dir);
+ float step_size = 1.0 / length(line_dir);
float z_advance = z_dir * step_size; // adapt z advance to line advance
float w_advance = w_dir * step_size; // adapt w advance to line advance
@@ -144,6 +150,14 @@ void main() {
float depth;
vec2 prev_pos = pos;
+ if (ivec2(pos + line_advance - 0.5) == ssC) {
+ // It is possible for rounding to cause our first pixel to check to be the pixel we're reflecting.
+ // Make sure we skip it
+ pos += line_advance;
+ z += z_advance;
+ w += w_advance;
+ }
+
bool found = false;
float steps_taken = 0.0;
@@ -154,8 +168,8 @@ void main() {
w += w_advance;
// convert to linear depth
-
- depth = imageLoad(source_depth, ivec2(pos - 0.5)).r;
+ ivec2 test_pos = ivec2(pos - 0.5);
+ depth = imageLoad(source_depth, test_pos).r;
if (sc_multiview) {
depth = depth * 2.0 - 1.0;
depth = 2.0 * params.camera_z_near * params.camera_z_far / (params.camera_z_far + params.camera_z_near - depth * (params.camera_z_far - params.camera_z_near));
@@ -166,13 +180,21 @@ void main() {
z_to = z / w;
if (depth > z_to) {
- // if depth was surpassed
- if (depth <= max(z_to, z_from) + params.depth_tolerance && -depth < params.camera_z_far) {
- // check the depth tolerance and far clip
- // check that normal is valid
- found = true;
+ // Test if our ray is hitting the "right" side of the surface, if not we're likely self reflecting and should skip.
+ vec4 test_normal_roughness = imageLoad(source_normal_roughness, test_pos);
+ vec3 test_normal = test_normal_roughness.xyz * 2.0 - 1.0;
+ test_normal = normalize(test_normal);
+ test_normal.y = -test_normal.y; //because this code reads flipped
+
+ if (dot(ray_dir, test_normal) < 0.001) {
+ // if depth was surpassed
+ if (depth <= max(z_to, z_from) + params.depth_tolerance && -depth < params.camera_z_far * 0.95) {
+ // check the depth tolerance and far clip
+ // check that normal is valid
+ found = true;
+ }
+ break;
}
- break;
}
steps_taken += 1.0;
@@ -182,17 +204,18 @@ void main() {
if (found) {
float margin_blend = 1.0;
- vec2 margin = vec2((params.screen_size.x + params.screen_size.y) * 0.5 * 0.05); // make a uniform margin
- if (any(bvec4(lessThan(pos, -margin), greaterThan(pos, params.screen_size + margin)))) {
- // clip outside screen + margin
+ vec2 margin = vec2((params.screen_size.x + params.screen_size.y) * 0.05); // make a uniform margin
+ if (any(bvec4(lessThan(pos, vec2(0.0, 0.0)), greaterThan(pos, params.screen_size)))) {
+ // clip at the screen edges
imageStore(ssr_image, ssC, vec4(0.0));
return;
}
{
- //blend fading out towards external margin
- vec2 margin_grad = mix(pos - params.screen_size, -pos, lessThan(pos, vec2(0.0)));
- margin_blend = 1.0 - smoothstep(0.0, margin.x, max(margin_grad.x, margin_grad.y));
+ //blend fading out towards inner margin
+ // 0.5 = midpoint of reflection
+ vec2 margin_grad = mix(params.screen_size - pos, pos, lessThan(pos, params.screen_size * 0.5));
+ margin_blend = smoothstep(0.0, margin.x * margin.y, margin_grad.x * margin_grad.y);
//margin_blend = 1.0;
}
@@ -200,6 +223,9 @@ void main() {
float grad = (steps_taken + 1.0) / float(params.num_steps);
float initial_fade = params.curve_fade_in == 0.0 ? 1.0 : pow(clamp(grad, 0.0, 1.0), params.curve_fade_in);
float fade = pow(clamp(1.0 - grad, 0.0, 1.0), params.distance_fade) * initial_fade;
+ // This is an ad-hoc term to fade out the SSR as roughness increases. Values used
+ // are meant to match the visual appearance of a ReflectionProbe.
+ float roughness_fade = smoothstep(0.4, 0.7, 1.0 - normal_roughness.w);
final_pos = pos;
vec4 final_color;
@@ -208,7 +234,6 @@ void main() {
// if roughness is enabled, do screen space cone tracing
float blur_radius = 0.0;
- float roughness = normal_roughness.w;
if (roughness > 0.001) {
float cone_angle = min(roughness, 0.999) * M_PI * 0.5;
@@ -230,18 +255,20 @@ void main() {
}
}
- // Isn't this going to be overwritten after our endif?
- final_color = imageLoad(source_diffuse, ivec2((final_pos - 0.5) * pixel_size));
-
imageStore(blur_radius_image, ssC, vec4(blur_radius / 255.0)); //stored in r8
#endif // MODE_ROUGH
- final_color = vec4(imageLoad(source_diffuse, ivec2(final_pos - 0.5)).rgb, fade * margin_blend);
+ final_color = vec4(imageLoad(source_diffuse, ivec2(final_pos - 0.5)).rgb, fade * margin_blend * roughness_fade);
- //change blend by metallic
- vec4 metallic_mask = unpackUnorm4x8(params.metallic_mask);
- final_color.a *= dot(metallic_mask, texelFetch(source_metallic, ssC << 1, 0));
+ // Schlick term.
+ float metallic = texelFetch(source_metallic, ssC << 1, 0).w;
+ float f0 = mix(0.04, 1.0, metallic); // Assume a "specular" amount of 0.5
+ normal.y = -normal.y;
+ float m = clamp(1.0 - dot(normalize(normal), -view_dir), 0.0, 1.0);
+ float m2 = m * m;
+ m = m2 * m2 * m; // pow(m,5)
+ final_color.a *= f0 + (1.0 - f0) * m; // Fresnel Schlick term.
imageStore(ssr_image, ssC, final_color);
diff --git a/servers/rendering/renderer_rd/shaders/effects/subsurface_scattering.glsl b/servers/rendering/renderer_rd/shaders/effects/subsurface_scattering.glsl
new file mode 100644
index 0000000000..fb35d3cde6
--- /dev/null
+++ b/servers/rendering/renderer_rd/shaders/effects/subsurface_scattering.glsl
@@ -0,0 +1,189 @@
+#[compute]
+
+#version 450
+
+#VERSION_DEFINES
+
+layout(local_size_x = 8, local_size_y = 8, local_size_z = 1) in;
+
+#ifdef USE_25_SAMPLES
+const int kernel_size = 13;
+
+const vec2 kernel[kernel_size] = vec2[](
+ vec2(0.530605, 0.0),
+ vec2(0.0211412, 0.0208333),
+ vec2(0.0402784, 0.0833333),
+ vec2(0.0493588, 0.1875),
+ vec2(0.0410172, 0.333333),
+ vec2(0.0263642, 0.520833),
+ vec2(0.017924, 0.75),
+ vec2(0.0128496, 1.02083),
+ vec2(0.0094389, 1.33333),
+ vec2(0.00700976, 1.6875),
+ vec2(0.00500364, 2.08333),
+ vec2(0.00333804, 2.52083),
+ vec2(0.000973794, 3.0));
+
+const vec4 skin_kernel[kernel_size] = vec4[](
+ vec4(0.530605, 0.613514, 0.739601, 0),
+ vec4(0.0211412, 0.0459286, 0.0378196, 0.0208333),
+ vec4(0.0402784, 0.0657244, 0.04631, 0.0833333),
+ vec4(0.0493588, 0.0367726, 0.0219485, 0.1875),
+ vec4(0.0410172, 0.0199899, 0.0118481, 0.333333),
+ vec4(0.0263642, 0.0119715, 0.00684598, 0.520833),
+ vec4(0.017924, 0.00711691, 0.00347194, 0.75),
+ vec4(0.0128496, 0.00356329, 0.00132016, 1.02083),
+ vec4(0.0094389, 0.00139119, 0.000416598, 1.33333),
+ vec4(0.00700976, 0.00049366, 0.000151938, 1.6875),
+ vec4(0.00500364, 0.00020094, 5.28848e-005, 2.08333),
+ vec4(0.00333804, 7.85443e-005, 1.2945e-005, 2.52083),
+ vec4(0.000973794, 1.11862e-005, 9.43437e-007, 3));
+
+#endif //USE_25_SAMPLES
+
+#ifdef USE_17_SAMPLES
+const int kernel_size = 9;
+const vec2 kernel[kernel_size] = vec2[](
+ vec2(0.536343, 0.0),
+ vec2(0.0324462, 0.03125),
+ vec2(0.0582416, 0.125),
+ vec2(0.0571056, 0.28125),
+ vec2(0.0347317, 0.5),
+ vec2(0.0216301, 0.78125),
+ vec2(0.0144609, 1.125),
+ vec2(0.0100386, 1.53125),
+ vec2(0.00317394, 2.0));
+
+const vec4 skin_kernel[kernel_size] = vec4[](
+ vec4(0.536343, 0.624624, 0.748867, 0),
+ vec4(0.0324462, 0.0656718, 0.0532821, 0.03125),
+ vec4(0.0582416, 0.0659959, 0.0411329, 0.125),
+ vec4(0.0571056, 0.0287432, 0.0172844, 0.28125),
+ vec4(0.0347317, 0.0151085, 0.00871983, 0.5),
+ vec4(0.0216301, 0.00794618, 0.00376991, 0.78125),
+ vec4(0.0144609, 0.00317269, 0.00106399, 1.125),
+ vec4(0.0100386, 0.000914679, 0.000275702, 1.53125),
+ vec4(0.00317394, 0.000134823, 3.77269e-005, 2));
+#endif //USE_17_SAMPLES
+
+#ifdef USE_11_SAMPLES
+const int kernel_size = 6;
+const vec2 kernel[kernel_size] = vec2[](
+ vec2(0.560479, 0.0),
+ vec2(0.0771802, 0.08),
+ vec2(0.0821904, 0.32),
+ vec2(0.03639, 0.72),
+ vec2(0.0192831, 1.28),
+ vec2(0.00471691, 2.0));
+
+const vec4 skin_kernel[kernel_size] = vec4[](
+
+ vec4(0.560479, 0.669086, 0.784728, 0),
+ vec4(0.0771802, 0.113491, 0.0793803, 0.08),
+ vec4(0.0821904, 0.0358608, 0.0209261, 0.32),
+ vec4(0.03639, 0.0130999, 0.00643685, 0.72),
+ vec4(0.0192831, 0.00282018, 0.00084214, 1.28),
+ vec4(0.00471691, 0.000184771, 5.07565e-005, 2));
+
+#endif //USE_11_SAMPLES
+
+layout(push_constant, std430) uniform Params {
+ ivec2 screen_size;
+ float camera_z_far;
+ float camera_z_near;
+
+ bool vertical;
+ bool orthogonal;
+ float unit_size;
+ float scale;
+
+ float depth_scale;
+ uint pad[3];
+}
+params;
+
+layout(set = 0, binding = 0) uniform sampler2D source_image;
+layout(rgba16f, set = 1, binding = 0) uniform restrict writeonly image2D dest_image;
+layout(set = 2, binding = 0) uniform sampler2D source_depth;
+
+void do_filter(inout vec3 color_accum, inout vec3 divisor, vec2 uv, vec2 step, bool p_skin) {
+ // Accumulate the other samples:
+ for (int i = 1; i < kernel_size; i++) {
+ // Fetch color and depth for current sample:
+ vec2 offset = uv + kernel[i].y * step;
+ vec4 color = texture(source_image, offset);
+
+ if (abs(color.a) < 0.001) {
+ break; //mix no more
+ }
+
+ vec3 w;
+ if (p_skin) {
+ //skin
+ w = skin_kernel[i].rgb;
+ } else {
+ w = vec3(kernel[i].x);
+ }
+
+ color_accum += color.rgb * w;
+ divisor += w;
+ }
+}
+
+void main() {
+ // Pixel being shaded
+ ivec2 ssC = ivec2(gl_GlobalInvocationID.xy);
+
+ if (any(greaterThanEqual(ssC, params.screen_size))) { //too large, do nothing
+ return;
+ }
+
+ vec2 uv = (vec2(ssC) + 0.5) / vec2(params.screen_size);
+
+ // Fetch color of current pixel:
+ vec4 base_color = texture(source_image, uv);
+ float strength = abs(base_color.a);
+
+ if (strength > 0.0) {
+ vec2 dir = params.vertical ? vec2(0.0, 1.0) : vec2(1.0, 0.0);
+
+ // Fetch linear depth of current pixel:
+ float depth = texture(source_depth, uv).r * 2.0 - 1.0;
+ float depth_scale;
+
+ if (params.orthogonal) {
+ depth = ((depth + (params.camera_z_far + params.camera_z_near) / (params.camera_z_far - params.camera_z_near)) * (params.camera_z_far - params.camera_z_near)) / 2.0;
+ depth_scale = params.unit_size; //remember depth is negative by default in OpenGL
+ } else {
+ depth = 2.0 * params.camera_z_near * params.camera_z_far / (params.camera_z_far + params.camera_z_near - depth * (params.camera_z_far - params.camera_z_near));
+ depth_scale = params.unit_size / depth; //remember depth is negative by default in OpenGL
+ }
+
+ float scale = mix(params.scale, depth_scale, params.depth_scale);
+
+ // Calculate the final step to fetch the surrounding pixels:
+ vec2 step = scale * dir;
+ step *= strength;
+ step /= 3.0;
+ // Accumulate the center sample:
+
+ vec3 divisor;
+ bool skin = bool(base_color.a < 0.0);
+
+ if (skin) {
+ //skin
+ divisor = skin_kernel[0].rgb;
+ } else {
+ divisor = vec3(kernel[0].x);
+ }
+
+ vec3 color = base_color.rgb * divisor;
+
+ do_filter(color, divisor, uv, step, skin);
+ do_filter(color, divisor, uv, -step, skin);
+
+ base_color.rgb = color / divisor;
+ }
+
+ imageStore(dest_image, ssC, base_color);
+}
diff --git a/servers/rendering/renderer_rd/shaders/effects/taa_resolve.glsl b/servers/rendering/renderer_rd/shaders/effects/taa_resolve.glsl
new file mode 100644
index 0000000000..02566d8e35
--- /dev/null
+++ b/servers/rendering/renderer_rd/shaders/effects/taa_resolve.glsl
@@ -0,0 +1,396 @@
+///////////////////////////////////////////////////////////////////////////////////
+// Copyright(c) 2016-2022 Panos Karabelas
+//
+// Permission is hereby granted, free of charge, to any person obtaining a copy
+// of this software and associated documentation files (the "Software"), to deal
+// in the Software without restriction, including without limitation the rights
+// to use, copy, modify, merge, publish, distribute, sublicense, and / or sell
+// copies of the Software, and to permit persons to whom the Software is furnished
+// to do so, subject to the following conditions :
+//
+// The above copyright notice and this permission notice shall be included in
+// all copies or substantial portions of the Software.
+//
+// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS
+// FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.IN NO EVENT SHALL THE AUTHORS OR
+// COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER
+// IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+// CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+///////////////////////////////////////////////////////////////////////////////////
+// File changes (yyyy-mm-dd)
+// 2022-05-06: Panos Karabelas: first commit
+// 2020-12-05: Joan Fons: convert to Vulkan and Godot
+///////////////////////////////////////////////////////////////////////////////////
+
+#[compute]
+
+#version 450
+
+#VERSION_DEFINES
+
+// Based on Spartan Engine's TAA implementation (without TAA upscale).
+// <https://github.com/PanosK92/SpartanEngine/blob/a8338d0609b85dc32f3732a5c27fb4463816a3b9/Data/shaders/temporal_antialiasing.hlsl>
+
+#ifndef MOLTENVK_USED
+#define USE_SUBGROUPS
+#endif // MOLTENVK_USED
+
+#define GROUP_SIZE 8
+#define FLT_MIN 0.00000001
+#define FLT_MAX 32767.0
+#define RPC_9 0.11111111111
+#define RPC_16 0.0625
+
+#ifdef USE_SUBGROUPS
+layout(local_size_x = GROUP_SIZE, local_size_y = GROUP_SIZE, local_size_z = 1) in;
+#endif
+
+layout(rgba16f, set = 0, binding = 0) uniform restrict readonly image2D color_buffer;
+layout(set = 0, binding = 1) uniform sampler2D depth_buffer;
+layout(rg16f, set = 0, binding = 2) uniform restrict readonly image2D velocity_buffer;
+layout(rg16f, set = 0, binding = 3) uniform restrict readonly image2D last_velocity_buffer;
+layout(set = 0, binding = 4) uniform sampler2D history_buffer;
+layout(rgba16f, set = 0, binding = 5) uniform restrict writeonly image2D output_buffer;
+
+layout(push_constant, std430) uniform Params {
+ vec2 resolution;
+ float disocclusion_threshold; // 0.1 / max(params.resolution.x, params.resolution.y
+ float disocclusion_scale;
+}
+params;
+
+const ivec2 kOffsets3x3[9] = {
+ ivec2(-1, -1),
+ ivec2(0, -1),
+ ivec2(1, -1),
+ ivec2(-1, 0),
+ ivec2(0, 0),
+ ivec2(1, 0),
+ ivec2(-1, 1),
+ ivec2(0, 1),
+ ivec2(1, 1),
+};
+
+/*------------------------------------------------------------------------------
+ THREAD GROUP SHARED MEMORY (LDS)
+------------------------------------------------------------------------------*/
+
+const int kBorderSize = 1;
+const int kGroupSize = GROUP_SIZE;
+const int kTileDimension = kGroupSize + kBorderSize * 2;
+const int kTileDimension2 = kTileDimension * kTileDimension;
+
+vec3 reinhard(vec3 hdr) {
+ return hdr / (hdr + 1.0);
+}
+vec3 reinhard_inverse(vec3 sdr) {
+ return sdr / (1.0 - sdr);
+}
+
+float get_depth(ivec2 thread_id) {
+ return texelFetch(depth_buffer, thread_id, 0).r;
+}
+
+#ifdef USE_SUBGROUPS
+shared vec3 tile_color[kTileDimension][kTileDimension];
+shared float tile_depth[kTileDimension][kTileDimension];
+
+vec3 load_color(uvec2 group_thread_id) {
+ group_thread_id += kBorderSize;
+ return tile_color[group_thread_id.x][group_thread_id.y];
+}
+
+void store_color(uvec2 group_thread_id, vec3 color) {
+ tile_color[group_thread_id.x][group_thread_id.y] = color;
+}
+
+float load_depth(uvec2 group_thread_id) {
+ group_thread_id += kBorderSize;
+ return tile_depth[group_thread_id.x][group_thread_id.y];
+}
+
+void store_depth(uvec2 group_thread_id, float depth) {
+ tile_depth[group_thread_id.x][group_thread_id.y] = depth;
+}
+
+void store_color_depth(uvec2 group_thread_id, ivec2 thread_id) {
+ // out of bounds clamp
+ thread_id = clamp(thread_id, ivec2(0, 0), ivec2(params.resolution) - ivec2(1, 1));
+
+ store_color(group_thread_id, imageLoad(color_buffer, thread_id).rgb);
+ store_depth(group_thread_id, get_depth(thread_id));
+}
+
+void populate_group_shared_memory(uvec2 group_id, uint group_index) {
+ // Populate group shared memory
+ ivec2 group_top_left = ivec2(group_id) * kGroupSize - kBorderSize;
+ if (group_index < (kTileDimension2 >> 2)) {
+ ivec2 group_thread_id_1 = ivec2(group_index % kTileDimension, group_index / kTileDimension);
+ ivec2 group_thread_id_2 = ivec2((group_index + (kTileDimension2 >> 2)) % kTileDimension, (group_index + (kTileDimension2 >> 2)) / kTileDimension);
+ ivec2 group_thread_id_3 = ivec2((group_index + (kTileDimension2 >> 1)) % kTileDimension, (group_index + (kTileDimension2 >> 1)) / kTileDimension);
+ ivec2 group_thread_id_4 = ivec2((group_index + kTileDimension2 * 3 / 4) % kTileDimension, (group_index + kTileDimension2 * 3 / 4) / kTileDimension);
+
+ store_color_depth(group_thread_id_1, group_top_left + group_thread_id_1);
+ store_color_depth(group_thread_id_2, group_top_left + group_thread_id_2);
+ store_color_depth(group_thread_id_3, group_top_left + group_thread_id_3);
+ store_color_depth(group_thread_id_4, group_top_left + group_thread_id_4);
+ }
+
+ // Wait for group threads to load store data.
+ groupMemoryBarrier();
+ barrier();
+}
+#else
+vec3 load_color(uvec2 screen_pos) {
+ return imageLoad(color_buffer, ivec2(screen_pos)).rgb;
+}
+
+float load_depth(uvec2 screen_pos) {
+ return get_depth(ivec2(screen_pos));
+}
+#endif
+
+/*------------------------------------------------------------------------------
+ VELOCITY
+------------------------------------------------------------------------------*/
+
+void depth_test_min(uvec2 pos, inout float min_depth, inout uvec2 min_pos) {
+ float depth = load_depth(pos);
+
+ if (depth < min_depth) {
+ min_depth = depth;
+ min_pos = pos;
+ }
+}
+
+// Returns velocity with closest depth (3x3 neighborhood)
+void get_closest_pixel_velocity_3x3(in uvec2 group_pos, uvec2 group_top_left, out vec2 velocity) {
+ float min_depth = 1.0;
+ uvec2 min_pos = group_pos;
+
+ depth_test_min(group_pos + kOffsets3x3[0], min_depth, min_pos);
+ depth_test_min(group_pos + kOffsets3x3[1], min_depth, min_pos);
+ depth_test_min(group_pos + kOffsets3x3[2], min_depth, min_pos);
+ depth_test_min(group_pos + kOffsets3x3[3], min_depth, min_pos);
+ depth_test_min(group_pos + kOffsets3x3[4], min_depth, min_pos);
+ depth_test_min(group_pos + kOffsets3x3[5], min_depth, min_pos);
+ depth_test_min(group_pos + kOffsets3x3[6], min_depth, min_pos);
+ depth_test_min(group_pos + kOffsets3x3[7], min_depth, min_pos);
+ depth_test_min(group_pos + kOffsets3x3[8], min_depth, min_pos);
+
+ // Velocity out
+ velocity = imageLoad(velocity_buffer, ivec2(group_top_left + min_pos)).xy;
+}
+
+/*------------------------------------------------------------------------------
+ HISTORY SAMPLING
+------------------------------------------------------------------------------*/
+
+vec3 sample_catmull_rom_9(sampler2D stex, vec2 uv, vec2 resolution) {
+ // Source: https://gist.github.com/TheRealMJP/c83b8c0f46b63f3a88a5986f4fa982b1
+ // License: https://gist.github.com/TheRealMJP/bc503b0b87b643d3505d41eab8b332ae
+
+ // We're going to sample a 4x4 grid of texels surrounding the target UV coordinate. We'll do this by rounding
+ // down the sample location to get the exact center of our "starting" texel. The starting texel will be at
+ // location [1, 1] in the grid, where [0, 0] is the top left corner.
+ vec2 sample_pos = uv * resolution;
+ vec2 texPos1 = floor(sample_pos - 0.5f) + 0.5f;
+
+ // Compute the fractional offset from our starting texel to our original sample location, which we'll
+ // feed into the Catmull-Rom spline function to get our filter weights.
+ vec2 f = sample_pos - texPos1;
+
+ // Compute the Catmull-Rom weights using the fractional offset that we calculated earlier.
+ // These equations are pre-expanded based on our knowledge of where the texels will be located,
+ // which lets us avoid having to evaluate a piece-wise function.
+ vec2 w0 = f * (-0.5f + f * (1.0f - 0.5f * f));
+ vec2 w1 = 1.0f + f * f * (-2.5f + 1.5f * f);
+ vec2 w2 = f * (0.5f + f * (2.0f - 1.5f * f));
+ vec2 w3 = f * f * (-0.5f + 0.5f * f);
+
+ // Work out weighting factors and sampling offsets that will let us use bilinear filtering to
+ // simultaneously evaluate the middle 2 samples from the 4x4 grid.
+ vec2 w12 = w1 + w2;
+ vec2 offset12 = w2 / (w1 + w2);
+
+ // Compute the final UV coordinates we'll use for sampling the texture
+ vec2 texPos0 = texPos1 - 1.0f;
+ vec2 texPos3 = texPos1 + 2.0f;
+ vec2 texPos12 = texPos1 + offset12;
+
+ texPos0 /= resolution;
+ texPos3 /= resolution;
+ texPos12 /= resolution;
+
+ vec3 result = vec3(0.0f, 0.0f, 0.0f);
+
+ result += textureLod(stex, vec2(texPos0.x, texPos0.y), 0.0).xyz * w0.x * w0.y;
+ result += textureLod(stex, vec2(texPos12.x, texPos0.y), 0.0).xyz * w12.x * w0.y;
+ result += textureLod(stex, vec2(texPos3.x, texPos0.y), 0.0).xyz * w3.x * w0.y;
+
+ result += textureLod(stex, vec2(texPos0.x, texPos12.y), 0.0).xyz * w0.x * w12.y;
+ result += textureLod(stex, vec2(texPos12.x, texPos12.y), 0.0).xyz * w12.x * w12.y;
+ result += textureLod(stex, vec2(texPos3.x, texPos12.y), 0.0).xyz * w3.x * w12.y;
+
+ result += textureLod(stex, vec2(texPos0.x, texPos3.y), 0.0).xyz * w0.x * w3.y;
+ result += textureLod(stex, vec2(texPos12.x, texPos3.y), 0.0).xyz * w12.x * w3.y;
+ result += textureLod(stex, vec2(texPos3.x, texPos3.y), 0.0).xyz * w3.x * w3.y;
+
+ return max(result, 0.0f);
+}
+
+/*------------------------------------------------------------------------------
+ HISTORY CLIPPING
+------------------------------------------------------------------------------*/
+
+// Based on "Temporal Reprojection Anti-Aliasing" - https://github.com/playdeadgames/temporal
+vec3 clip_aabb(vec3 aabb_min, vec3 aabb_max, vec3 p, vec3 q) {
+ vec3 r = q - p;
+ vec3 rmax = (aabb_max - p.xyz);
+ vec3 rmin = (aabb_min - p.xyz);
+
+ if (r.x > rmax.x + FLT_MIN)
+ r *= (rmax.x / r.x);
+ if (r.y > rmax.y + FLT_MIN)
+ r *= (rmax.y / r.y);
+ if (r.z > rmax.z + FLT_MIN)
+ r *= (rmax.z / r.z);
+
+ if (r.x < rmin.x - FLT_MIN)
+ r *= (rmin.x / r.x);
+ if (r.y < rmin.y - FLT_MIN)
+ r *= (rmin.y / r.y);
+ if (r.z < rmin.z - FLT_MIN)
+ r *= (rmin.z / r.z);
+
+ return p + r;
+}
+
+// Clip history to the neighbourhood of the current sample
+vec3 clip_history_3x3(uvec2 group_pos, vec3 color_history, vec2 velocity_closest) {
+ // Sample a 3x3 neighbourhood
+ vec3 s1 = load_color(group_pos + kOffsets3x3[0]);
+ vec3 s2 = load_color(group_pos + kOffsets3x3[1]);
+ vec3 s3 = load_color(group_pos + kOffsets3x3[2]);
+ vec3 s4 = load_color(group_pos + kOffsets3x3[3]);
+ vec3 s5 = load_color(group_pos + kOffsets3x3[4]);
+ vec3 s6 = load_color(group_pos + kOffsets3x3[5]);
+ vec3 s7 = load_color(group_pos + kOffsets3x3[6]);
+ vec3 s8 = load_color(group_pos + kOffsets3x3[7]);
+ vec3 s9 = load_color(group_pos + kOffsets3x3[8]);
+
+ // Compute min and max (with an adaptive box size, which greatly reduces ghosting)
+ vec3 color_avg = (s1 + s2 + s3 + s4 + s5 + s6 + s7 + s8 + s9) * RPC_9;
+ vec3 color_avg2 = ((s1 * s1) + (s2 * s2) + (s3 * s3) + (s4 * s4) + (s5 * s5) + (s6 * s6) + (s7 * s7) + (s8 * s8) + (s9 * s9)) * RPC_9;
+ float box_size = mix(0.0f, 2.5f, smoothstep(0.02f, 0.0f, length(velocity_closest)));
+ vec3 dev = sqrt(abs(color_avg2 - (color_avg * color_avg))) * box_size;
+ vec3 color_min = color_avg - dev;
+ vec3 color_max = color_avg + dev;
+
+ // Variance clipping
+ vec3 color = clip_aabb(color_min, color_max, clamp(color_avg, color_min, color_max), color_history);
+
+ // Clamp to prevent NaNs
+ color = clamp(color, FLT_MIN, FLT_MAX);
+
+ return color;
+}
+
+/*------------------------------------------------------------------------------
+ TAA
+------------------------------------------------------------------------------*/
+
+const vec3 lumCoeff = vec3(0.299f, 0.587f, 0.114f);
+
+float luminance(vec3 color) {
+ return max(dot(color, lumCoeff), 0.0001f);
+}
+
+float get_factor_disocclusion(vec2 uv_reprojected, vec2 velocity) {
+ vec2 velocity_previous = imageLoad(last_velocity_buffer, ivec2(uv_reprojected * params.resolution)).xy;
+ vec2 velocity_texels = velocity * params.resolution;
+ vec2 prev_velocity_texels = velocity_previous * params.resolution;
+ float disocclusion = length(prev_velocity_texels - velocity_texels) - params.disocclusion_threshold;
+ return clamp(disocclusion * params.disocclusion_scale, 0.0, 1.0);
+}
+
+vec3 temporal_antialiasing(uvec2 pos_group_top_left, uvec2 pos_group, uvec2 pos_screen, vec2 uv, sampler2D tex_history) {
+ // Get the velocity of the current pixel
+ vec2 velocity = imageLoad(velocity_buffer, ivec2(pos_screen)).xy;
+
+ // Get reprojected uv
+ vec2 uv_reprojected = uv - velocity;
+
+ // Get input color
+ vec3 color_input = load_color(pos_group);
+
+ // Get history color (catmull-rom reduces a lot of the blurring that you get under motion)
+ vec3 color_history = sample_catmull_rom_9(tex_history, uv_reprojected, params.resolution).rgb;
+
+ // Clip history to the neighbourhood of the current sample (fixes a lot of the ghosting).
+ vec2 velocity_closest = vec2(0.0); // This is best done by using the velocity with the closest depth.
+ get_closest_pixel_velocity_3x3(pos_group, pos_group_top_left, velocity_closest);
+ color_history = clip_history_3x3(pos_group, color_history, velocity_closest);
+
+ // Compute blend factor
+ float blend_factor = RPC_16; // We want to be able to accumulate as many jitter samples as we generated, that is, 16.
+ {
+ // If re-projected UV is out of screen, converge to current color immediatel
+ float factor_screen = any(lessThan(uv_reprojected, vec2(0.0))) || any(greaterThan(uv_reprojected, vec2(1.0))) ? 1.0 : 0.0;
+
+ // Increase blend factor when there is disocclusion (fixes a lot of the remaining ghosting).
+ float factor_disocclusion = get_factor_disocclusion(uv_reprojected, velocity);
+
+ // Add to the blend factor
+ blend_factor = clamp(blend_factor + factor_screen + factor_disocclusion, 0.0, 1.0);
+ }
+
+ // Resolve
+ vec3 color_resolved = vec3(0.0);
+ {
+ // Tonemap
+ color_history = reinhard(color_history);
+ color_input = reinhard(color_input);
+
+ // Reduce flickering
+ float lum_color = luminance(color_input);
+ float lum_history = luminance(color_history);
+ float diff = abs(lum_color - lum_history) / max(lum_color, max(lum_history, 1.001));
+ diff = 1.0 - diff;
+ diff = diff * diff;
+ blend_factor = mix(0.0, blend_factor, diff);
+
+ // Lerp/blend
+ color_resolved = mix(color_history, color_input, blend_factor);
+
+ // Inverse tonemap
+ color_resolved = reinhard_inverse(color_resolved);
+ }
+
+ return color_resolved;
+}
+
+void main() {
+#ifdef USE_SUBGROUPS
+ populate_group_shared_memory(gl_WorkGroupID.xy, gl_LocalInvocationIndex);
+#endif
+
+ // Out of bounds check
+ if (any(greaterThanEqual(vec2(gl_GlobalInvocationID.xy), params.resolution))) {
+ return;
+ }
+
+#ifdef USE_SUBGROUPS
+ const uvec2 pos_group = gl_LocalInvocationID.xy;
+ const uvec2 pos_group_top_left = gl_WorkGroupID.xy * kGroupSize - kBorderSize;
+#else
+ const uvec2 pos_group = gl_GlobalInvocationID.xy;
+ const uvec2 pos_group_top_left = uvec2(0, 0);
+#endif
+ const uvec2 pos_screen = gl_GlobalInvocationID.xy;
+ const vec2 uv = (gl_GlobalInvocationID.xy + 0.5f) / params.resolution;
+
+ vec3 result = temporal_antialiasing(pos_group_top_left, pos_group, pos_screen, uv, history_buffer);
+ imageStore(output_buffer, ivec2(gl_GlobalInvocationID.xy), vec4(result, 1.0));
+}
diff --git a/servers/rendering/renderer_rd/shaders/effects/tonemap.glsl b/servers/rendering/renderer_rd/shaders/effects/tonemap.glsl
index 62a7b0e7d7..52aee8b648 100644
--- a/servers/rendering/renderer_rd/shaders/effects/tonemap.glsl
+++ b/servers/rendering/renderer_rd/shaders/effects/tonemap.glsl
@@ -75,7 +75,7 @@ layout(push_constant, std430) uniform Params {
float exposure;
float white;
- float auto_exposure_grey;
+ float auto_exposure_scale;
float luminance_multiplier;
vec2 pixel_size;
@@ -360,15 +360,15 @@ vec3 do_fxaa(vec3 color, float exposure, vec2 uv_interp) {
const float FXAA_SPAN_MAX = 8.0;
#ifdef MULTIVIEW
- vec3 rgbNW = textureLod(source_color, vec3(uv_interp + vec2(-1.0, -1.0) * params.pixel_size, ViewIndex), 0.0).xyz * exposure * params.luminance_multiplier;
- vec3 rgbNE = textureLod(source_color, vec3(uv_interp + vec2(1.0, -1.0) * params.pixel_size, ViewIndex), 0.0).xyz * exposure * params.luminance_multiplier;
- vec3 rgbSW = textureLod(source_color, vec3(uv_interp + vec2(-1.0, 1.0) * params.pixel_size, ViewIndex), 0.0).xyz * exposure * params.luminance_multiplier;
- vec3 rgbSE = textureLod(source_color, vec3(uv_interp + vec2(1.0, 1.0) * params.pixel_size, ViewIndex), 0.0).xyz * exposure * params.luminance_multiplier;
+ vec3 rgbNW = textureLod(source_color, vec3(uv_interp + vec2(-0.5, -0.5) * params.pixel_size, ViewIndex), 0.0).xyz * exposure * params.luminance_multiplier;
+ vec3 rgbNE = textureLod(source_color, vec3(uv_interp + vec2(0.5, -0.5) * params.pixel_size, ViewIndex), 0.0).xyz * exposure * params.luminance_multiplier;
+ vec3 rgbSW = textureLod(source_color, vec3(uv_interp + vec2(-0.5, 0.5) * params.pixel_size, ViewIndex), 0.0).xyz * exposure * params.luminance_multiplier;
+ vec3 rgbSE = textureLod(source_color, vec3(uv_interp + vec2(0.5, 0.5) * params.pixel_size, ViewIndex), 0.0).xyz * exposure * params.luminance_multiplier;
#else
- vec3 rgbNW = textureLod(source_color, uv_interp + vec2(-1.0, -1.0) * params.pixel_size, 0.0).xyz * exposure * params.luminance_multiplier;
- vec3 rgbNE = textureLod(source_color, uv_interp + vec2(1.0, -1.0) * params.pixel_size, 0.0).xyz * exposure * params.luminance_multiplier;
- vec3 rgbSW = textureLod(source_color, uv_interp + vec2(-1.0, 1.0) * params.pixel_size, 0.0).xyz * exposure * params.luminance_multiplier;
- vec3 rgbSE = textureLod(source_color, uv_interp + vec2(1.0, 1.0) * params.pixel_size, 0.0).xyz * exposure * params.luminance_multiplier;
+ vec3 rgbNW = textureLod(source_color, uv_interp + vec2(-0.5, -0.5) * params.pixel_size, 0.0).xyz * exposure * params.luminance_multiplier;
+ vec3 rgbNE = textureLod(source_color, uv_interp + vec2(0.5, -0.5) * params.pixel_size, 0.0).xyz * exposure * params.luminance_multiplier;
+ vec3 rgbSW = textureLod(source_color, uv_interp + vec2(-0.5, 0.5) * params.pixel_size, 0.0).xyz * exposure * params.luminance_multiplier;
+ vec3 rgbSE = textureLod(source_color, uv_interp + vec2(0.5, 0.5) * params.pixel_size, 0.0).xyz * exposure * params.luminance_multiplier;
#endif
vec3 rgbM = color;
vec3 luma = vec3(0.299, 0.587, 0.114);
@@ -440,7 +440,7 @@ void main() {
#ifndef SUBPASS
if (params.use_auto_exposure) {
- exposure *= 1.0 / (texelFetch(source_auto_exposure, ivec2(0, 0), 0).r * params.luminance_multiplier / params.auto_exposure_grey);
+ exposure *= 1.0 / (texelFetch(source_auto_exposure, ivec2(0, 0), 0).r * params.luminance_multiplier / params.auto_exposure_scale);
}
#endif
@@ -462,12 +462,6 @@ void main() {
}
#endif
- if (params.use_debanding) {
- // For best results, debanding should be done before tonemapping.
- // Otherwise, we're adding noise to an already-quantized image.
- color.rgb += screen_space_dither(gl_FragCoord.xy);
- }
-
color.rgb = apply_tonemapping(color.rgb, params.white);
color.rgb = linear_to_srgb(color.rgb); // regular linear -> SRGB conversion
@@ -498,5 +492,11 @@ void main() {
color.rgb = apply_color_correction(color.rgb);
}
+ if (params.use_debanding) {
+ // Debanding should be done at the end of tonemapping, but before writing to the LDR buffer.
+ // Otherwise, we're adding noise to an already-quantized image.
+ color.rgb += screen_space_dither(gl_FragCoord.xy);
+ }
+
frag_color = color;
}
diff --git a/servers/rendering/renderer_rd/shaders/effects/vrs.glsl b/servers/rendering/renderer_rd/shaders/effects/vrs.glsl
index 5ef83c0b44..b450bb9fe9 100644
--- a/servers/rendering/renderer_rd/shaders/effects/vrs.glsl
+++ b/servers/rendering/renderer_rd/shaders/effects/vrs.glsl
@@ -63,10 +63,18 @@ void main() {
#ifdef MULTIVIEW
vec4 color = textureLod(source_color, uv, 0.0);
+ frag_color = uint(color.r * 255.0);
#else /* MULTIVIEW */
vec4 color = textureLod(source_color, uv, 0.0);
-#endif /* MULTIVIEW */
- // See if we can change the sampler to one that returns int...
- frag_color = uint(color.r * 256.0);
+ // for user supplied VRS map we do a color mapping
+ color.r *= 3.0;
+ frag_color = int(color.r) << 2;
+
+ color.g *= 3.0;
+ frag_color += int(color.g);
+
+ // note 1x4, 4x1, 1x8, 8x1, 2x8 and 8x2 are not supported
+ // 4x8, 8x4 and 8x8 are only available on some GPUs
+#endif /* MULTIVIEW */
}