diff options
Diffstat (limited to 'servers/rendering/renderer_rd/shaders')
61 files changed, 1626 insertions, 483 deletions
diff --git a/servers/rendering/renderer_rd/shaders/SCsub b/servers/rendering/renderer_rd/shaders/SCsub index fc513d3fb9..d352743908 100644 --- a/servers/rendering/renderer_rd/shaders/SCsub +++ b/servers/rendering/renderer_rd/shaders/SCsub @@ -10,8 +10,11 @@ if "RD_GLSL" in env["BUILDERS"]: glsl_files = [str(f) for f in Glob("*.glsl") if str(f) not in gl_include_files] # make sure we recompile shaders if include files change - env.Depends([f + ".gen.h" for f in glsl_files], gl_include_files) + env.Depends([f + ".gen.h" for f in glsl_files], gl_include_files + ["#glsl_builders.py"]) # compile shaders for glsl_file in glsl_files: env.RD_GLSL(glsl_file) + +SConscript("effects/SCsub") +SConscript("environment/SCsub") diff --git a/servers/rendering/renderer_rd/shaders/blur_raster_inc.glsl b/servers/rendering/renderer_rd/shaders/blur_raster_inc.glsl deleted file mode 100644 index e7a2e18323..0000000000 --- a/servers/rendering/renderer_rd/shaders/blur_raster_inc.glsl +++ /dev/null @@ -1,21 +0,0 @@ -#define FLAG_HORIZONTAL (1 << 0) -#define FLAG_USE_ORTHOGONAL_PROJECTION (1 << 1) -#define FLAG_GLOW_FIRST_PASS (1 << 2) - -layout(push_constant, std430) uniform Blur { - vec2 pixel_size; - uint flags; - uint pad; - - // Glow. - float glow_strength; - float glow_bloom; - float glow_hdr_threshold; - float glow_hdr_scale; - - float glow_exposure; - float glow_white; - float glow_luminance_cap; - float glow_auto_exposure_grey; -} -blur; diff --git a/servers/rendering/renderer_rd/shaders/canvas.glsl b/servers/rendering/renderer_rd/shaders/canvas.glsl index 65a621b203..f8e9020f9f 100644 --- a/servers/rendering/renderer_rd/shaders/canvas.glsl +++ b/servers/rendering/renderer_rd/shaders/canvas.glsl @@ -82,7 +82,7 @@ void main() { #endif - mat4 world_matrix = mat4(vec4(draw_data.world_x, 0.0, 0.0), vec4(draw_data.world_y, 0.0, 0.0), vec4(0.0, 0.0, 1.0, 0.0), vec4(draw_data.world_ofs, 0.0, 1.0)); + mat4 model_matrix = mat4(vec4(draw_data.world_x, 0.0, 0.0), vec4(draw_data.world_y, 0.0, 0.0), vec4(0.0, 0.0, 1.0, 0.0), vec4(draw_data.world_ofs, 0.0, 1.0)); #define FLAGS_INSTANCING_MASK 0x7F #define FLAGS_INSTANCING_HAS_COLORS (1 << 7) @@ -156,7 +156,7 @@ void main() { } matrix = transpose(matrix); - world_matrix = world_matrix * matrix; + model_matrix = model_matrix * matrix; } } @@ -179,7 +179,7 @@ void main() { #endif #if !defined(SKIP_TRANSFORM_USED) - vertex = (world_matrix * vec4(vertex, 0.0, 1.0)).xy; + vertex = (model_matrix * vec4(vertex, 0.0, 1.0)).xy; #endif color_interp = color; @@ -461,10 +461,6 @@ float msdf_median(float r, float g, float b, float a) { return min(max(min(r, g), min(max(r, g), b)), a); } -vec2 msdf_map(vec2 value, vec2 in_min, vec2 in_max, vec2 out_min, vec2 out_max) { - return out_min + (out_max - out_min) * (value - in_min) / (in_max - in_min); -} - void main() { vec4 color = color_interp; vec2 uv = uv_interp; diff --git a/servers/rendering/renderer_rd/shaders/effects/SCsub b/servers/rendering/renderer_rd/shaders/effects/SCsub new file mode 100644 index 0000000000..741da8fe69 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/effects/SCsub @@ -0,0 +1,17 @@ +#!/usr/bin/env python + +Import("env") + +if "RD_GLSL" in env["BUILDERS"]: + # find all include files + gl_include_files = [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] + + # make sure we recompile shaders if include files change + env.Depends([f + ".gen.h" for f in glsl_files], gl_include_files + ["#glsl_builders.py"]) + + # compile shaders + for glsl_file in glsl_files: + env.RD_GLSL(glsl_file) diff --git a/servers/rendering/renderer_rd/shaders/blur_raster.glsl b/servers/rendering/renderer_rd/shaders/effects/blur_raster.glsl index f8b4e3f610..96f5c3e9f2 100644 --- a/servers/rendering/renderer_rd/shaders/blur_raster.glsl +++ b/servers/rendering/renderer_rd/shaders/effects/blur_raster.glsl @@ -53,7 +53,9 @@ void main() { #ifdef MODE_GAUSSIAN_BLUR - //Simpler blur uses SIGMA2 for the gaussian kernel for a stronger effect + // 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; @@ -94,6 +96,7 @@ void main() { 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) * 0.174938; GLOW_ADD(vec2(1.0, 0.0), 0.165569); GLOW_ADD(vec2(2.0, 0.0), 0.140367); @@ -101,7 +104,10 @@ void main() { GLOW_ADD(vec2(-1.0, 0.0), 0.165569); GLOW_ADD(vec2(-2.0, 0.0), 0.140367); GLOW_ADD(vec2(-3.0, 0.0), 0.106595); + + // only do this in the horizontal pass, if we also do this in the vertical pass we're doubling up. color *= blur.glow_strength; + frag_color = color; } else { vec2 pix_size = blur.pixel_size; @@ -110,13 +116,17 @@ void main() { GLOW_ADD(vec2(0.0, 2.0), 0.122581); GLOW_ADD(vec2(0.0, -1.0), 0.233062); GLOW_ADD(vec2(0.0, -2.0), 0.122581); - color *= blur.glow_strength; + frag_color = color; } #undef GLOW_ADD if (bool(blur.flags & FLAG_GLOW_FIRST_PASS)) { + // In the first pass bring back to correct color range else we're applying the wrong threshold + // in subsequent passes we can use it as is as we'd just be undoing it right after. + frag_color *= blur.luminance_multiplier; + #ifdef GLOW_USE_AUTO_EXPOSURE frag_color /= texelFetch(source_auto_exposure, ivec2(0, 0), 0).r / blur.glow_auto_exposure_grey; @@ -126,10 +136,10 @@ void main() { float luminance = max(frag_color.r, max(frag_color.g, frag_color.b)); float feedback = max(smoothstep(blur.glow_hdr_threshold, blur.glow_hdr_threshold + blur.glow_hdr_scale, luminance), blur.glow_bloom); - frag_color = min(frag_color * feedback, vec4(blur.glow_luminance_cap)); + frag_color = min(frag_color * feedback, vec4(blur.glow_luminance_cap)) / blur.luminance_multiplier; } -#endif +#endif // MODE_GAUSSIAN_GLOW #ifdef MODE_COPY vec4 color = textureLod(source_color, uv_interp, 0.0); diff --git a/servers/rendering/renderer_rd/shaders/effects/blur_raster_inc.glsl b/servers/rendering/renderer_rd/shaders/effects/blur_raster_inc.glsl new file mode 100644 index 0000000000..730504571a --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/effects/blur_raster_inc.glsl @@ -0,0 +1,26 @@ +#define FLAG_HORIZONTAL (1 << 0) +#define FLAG_USE_ORTHOGONAL_PROJECTION (1 << 1) +#define FLAG_GLOW_FIRST_PASS (1 << 2) + +layout(push_constant, std430) uniform Blur { + vec2 pixel_size; // 08 - 08 + uint flags; // 04 - 12 + uint pad; // 04 - 16 + + // Glow. + float glow_strength; // 04 - 20 + float glow_bloom; // 04 - 24 + float glow_hdr_threshold; // 04 - 28 + float glow_hdr_scale; // 04 - 32 + + float glow_exposure; // 04 - 36 + float glow_white; // 04 - 40 + float glow_luminance_cap; // 04 - 44 + float glow_auto_exposure_grey; // 04 - 48 + + float luminance_multiplier; // 04 - 52 + float res1; // 04 - 56 + float res2; // 04 - 60 + float res3; // 04 - 64 +} +blur; diff --git a/servers/rendering/renderer_rd/shaders/bokeh_dof.glsl b/servers/rendering/renderer_rd/shaders/effects/bokeh_dof.glsl index 0438671dd2..0438671dd2 100644 --- a/servers/rendering/renderer_rd/shaders/bokeh_dof.glsl +++ b/servers/rendering/renderer_rd/shaders/effects/bokeh_dof.glsl diff --git a/servers/rendering/renderer_rd/shaders/bokeh_dof_inc.glsl b/servers/rendering/renderer_rd/shaders/effects/bokeh_dof_inc.glsl index b90a527554..b90a527554 100644 --- a/servers/rendering/renderer_rd/shaders/bokeh_dof_inc.glsl +++ b/servers/rendering/renderer_rd/shaders/effects/bokeh_dof_inc.glsl diff --git a/servers/rendering/renderer_rd/shaders/bokeh_dof_raster.glsl b/servers/rendering/renderer_rd/shaders/effects/bokeh_dof_raster.glsl index a3b3938ee9..a3b3938ee9 100644 --- a/servers/rendering/renderer_rd/shaders/bokeh_dof_raster.glsl +++ b/servers/rendering/renderer_rd/shaders/effects/bokeh_dof_raster.glsl diff --git a/servers/rendering/renderer_rd/shaders/copy.glsl b/servers/rendering/renderer_rd/shaders/effects/copy.glsl index 4563ac7af9..3a4ef86ef0 100644 --- a/servers/rendering/renderer_rd/shaders/copy.glsl +++ b/servers/rendering/renderer_rd/shaders/effects/copy.glsl @@ -189,7 +189,7 @@ void main() { #endif color *= params.glow_exposure; - float luminance = dot(color.rgb, vec3(0.299, 0.587, 0.114)); + float luminance = max(color.r, max(color.g, color.b)); float feedback = max(smoothstep(params.glow_hdr_threshold, params.glow_hdr_threshold + params.glow_hdr_scale, luminance), params.glow_bloom); color = min(color * feedback, vec4(params.glow_luminance_cap)); diff --git a/servers/rendering/renderer_rd/shaders/copy_to_fb.glsl b/servers/rendering/renderer_rd/shaders/effects/copy_to_fb.glsl index 2f1f9c4765..1c17eabb56 100644 --- a/servers/rendering/renderer_rd/shaders/copy_to_fb.glsl +++ b/servers/rendering/renderer_rd/shaders/effects/copy_to_fb.glsl @@ -4,7 +4,20 @@ #VERSION_DEFINES +#ifdef MULTIVIEW +#ifdef has_VK_KHR_multiview +#extension GL_EXT_multiview : enable +#define ViewIndex gl_ViewIndex +#else // has_VK_KHR_multiview +#define ViewIndex 0 +#endif // has_VK_KHR_multiview +#endif //MULTIVIEW + +#ifdef MULTIVIEW +layout(location = 0) out vec3 uv_interp; +#else layout(location = 0) out vec2 uv_interp; +#endif layout(push_constant, std430) uniform Params { vec4 section; @@ -19,9 +32,11 @@ params; 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]; - - vec2 vpos = uv_interp; + uv_interp.xy = base_arr[gl_VertexIndex]; +#ifdef MULTIVIEW + uv_interp.z = ViewIndex; +#endif + vec2 vpos = uv_interp.xy; if (params.use_section) { vpos = params.section.xy + vpos * params.section.zw; } @@ -39,6 +54,15 @@ void main() { #VERSION_DEFINES +#ifdef MULTIVIEW +#ifdef has_VK_KHR_multiview +#extension GL_EXT_multiview : enable +#define ViewIndex gl_ViewIndex +#else // has_VK_KHR_multiview +#define ViewIndex 0 +#endif // has_VK_KHR_multiview +#endif //MULTIVIEW + layout(push_constant, std430) uniform Params { vec4 section; vec2 pixel_size; @@ -52,12 +76,25 @@ layout(push_constant, std430) uniform Params { } params; +#ifdef MULTIVIEW +layout(location = 0) in vec3 uv_interp; +#else layout(location = 0) in vec2 uv_interp; +#endif +#ifdef MULTIVIEW +layout(set = 0, binding = 0) uniform sampler2DArray source_color; +#ifdef MODE_TWO_SOURCES +layout(set = 1, binding = 0) uniform sampler2DArray source_depth; +layout(location = 1) out float depth; +#endif /* MODE_TWO_SOURCES */ +#else /* MULTIVIEW */ layout(set = 0, binding = 0) uniform sampler2D source_color; #ifdef MODE_TWO_SOURCES layout(set = 1, binding = 0) uniform sampler2D source_color2; -#endif +#endif /* MODE_TWO_SOURCES */ +#endif /* MULTIVIEW */ + layout(location = 0) out vec4 frag_color; vec3 linear_to_srgb(vec3 color) { @@ -68,9 +105,14 @@ vec3 linear_to_srgb(vec3 color) { } void main() { +#ifdef MULTIVIEW + vec3 uv = uv_interp; +#else vec2 uv = uv_interp; +#endif #ifdef MODE_PANORAMA_TO_DP + // Note, multiview and panorama should not be mixed at this time //obtain normal from dual paraboloid uv #define M_PI 3.14159265359 @@ -97,11 +139,22 @@ void main() { //uv.y = 1.0 - uv.y; uv = 1.0 - uv; } -#endif +#endif /* MODE_PANORAMA_TO_DP */ + +#ifdef MULTIVIEW + vec4 color = textureLod(source_color, uv, 0.0); +#ifdef MODE_TWO_SOURCES + // In multiview our 2nd input will be our depth map + depth = textureLod(source_depth, uv, 0.0).r; +#endif /* MODE_TWO_SOURCES */ + +#else /* MULTIVIEW */ vec4 color = textureLod(source_color, uv, 0.0); #ifdef MODE_TWO_SOURCES color += textureLod(source_color2, uv, 0.0); -#endif +#endif /* MODE_TWO_SOURCES */ +#endif /* MULTIVIEW */ + if (params.force_luminance) { color.rgb = vec3(max(max(color.r, color.g), color.b)); } @@ -111,5 +164,6 @@ void main() { if (params.srgb) { color.rgb = linear_to_srgb(color.rgb); } + frag_color = color; } diff --git a/servers/rendering/renderer_rd/shaders/cube_to_dp.glsl b/servers/rendering/renderer_rd/shaders/effects/cube_to_dp.glsl index e77d0de719..e77d0de719 100644 --- a/servers/rendering/renderer_rd/shaders/cube_to_dp.glsl +++ b/servers/rendering/renderer_rd/shaders/effects/cube_to_dp.glsl diff --git a/servers/rendering/renderer_rd/shaders/cubemap_downsampler.glsl b/servers/rendering/renderer_rd/shaders/effects/cubemap_downsampler.glsl index 63f0ce690e..63f0ce690e 100644 --- a/servers/rendering/renderer_rd/shaders/cubemap_downsampler.glsl +++ b/servers/rendering/renderer_rd/shaders/effects/cubemap_downsampler.glsl diff --git a/servers/rendering/renderer_rd/shaders/cubemap_downsampler_inc.glsl b/servers/rendering/renderer_rd/shaders/effects/cubemap_downsampler_inc.glsl index 641e0906f5..641e0906f5 100644 --- a/servers/rendering/renderer_rd/shaders/cubemap_downsampler_inc.glsl +++ b/servers/rendering/renderer_rd/shaders/effects/cubemap_downsampler_inc.glsl diff --git a/servers/rendering/renderer_rd/shaders/cubemap_downsampler_raster.glsl b/servers/rendering/renderer_rd/shaders/effects/cubemap_downsampler_raster.glsl index 0828ffd921..0828ffd921 100644 --- a/servers/rendering/renderer_rd/shaders/cubemap_downsampler_raster.glsl +++ b/servers/rendering/renderer_rd/shaders/effects/cubemap_downsampler_raster.glsl diff --git a/servers/rendering/renderer_rd/shaders/cubemap_filter.glsl b/servers/rendering/renderer_rd/shaders/effects/cubemap_filter.glsl index 2a774b0eb4..2a774b0eb4 100644 --- a/servers/rendering/renderer_rd/shaders/cubemap_filter.glsl +++ b/servers/rendering/renderer_rd/shaders/effects/cubemap_filter.glsl diff --git a/servers/rendering/renderer_rd/shaders/cubemap_filter_raster.glsl b/servers/rendering/renderer_rd/shaders/effects/cubemap_filter_raster.glsl index 0990dc7c2f..0990dc7c2f 100644 --- a/servers/rendering/renderer_rd/shaders/cubemap_filter_raster.glsl +++ b/servers/rendering/renderer_rd/shaders/effects/cubemap_filter_raster.glsl diff --git a/servers/rendering/renderer_rd/shaders/cubemap_roughness.glsl b/servers/rendering/renderer_rd/shaders/effects/cubemap_roughness.glsl index 1d46f59408..1d46f59408 100644 --- a/servers/rendering/renderer_rd/shaders/cubemap_roughness.glsl +++ b/servers/rendering/renderer_rd/shaders/effects/cubemap_roughness.glsl diff --git a/servers/rendering/renderer_rd/shaders/cubemap_roughness_inc.glsl b/servers/rendering/renderer_rd/shaders/effects/cubemap_roughness_inc.glsl index 1bee428a6f..1bee428a6f 100644 --- a/servers/rendering/renderer_rd/shaders/cubemap_roughness_inc.glsl +++ b/servers/rendering/renderer_rd/shaders/effects/cubemap_roughness_inc.glsl diff --git a/servers/rendering/renderer_rd/shaders/cubemap_roughness_raster.glsl b/servers/rendering/renderer_rd/shaders/effects/cubemap_roughness_raster.glsl index c29accd8a7..c29accd8a7 100644 --- a/servers/rendering/renderer_rd/shaders/cubemap_roughness_raster.glsl +++ b/servers/rendering/renderer_rd/shaders/effects/cubemap_roughness_raster.glsl diff --git a/servers/rendering/renderer_rd/shaders/resolve.glsl b/servers/rendering/renderer_rd/shaders/effects/resolve.glsl index 0e086331c0..0e086331c0 100644 --- a/servers/rendering/renderer_rd/shaders/resolve.glsl +++ b/servers/rendering/renderer_rd/shaders/effects/resolve.glsl diff --git a/servers/rendering/renderer_rd/shaders/screen_space_reflection.glsl b/servers/rendering/renderer_rd/shaders/effects/screen_space_reflection.glsl index a416891ff2..d85ab3af2e 100644 --- a/servers/rendering/renderer_rd/shaders/screen_space_reflection.glsl +++ b/servers/rendering/renderer_rd/shaders/effects/screen_space_reflection.glsl @@ -32,12 +32,17 @@ layout(push_constant, std430) uniform Params { bool use_half_res; uint metallic_mask; - mat4 projection; + uint view_index; + uint pad1; + uint pad2; + uint pad3; } params; +#include "screen_space_reflection_inc.glsl" + vec2 view_to_screen(vec3 view_pos, out float w) { - vec4 projected = params.projection * vec4(view_pos, 1.0); + vec4 projected = scene_data.projection[params.view_index] * vec4(view_pos, 1.0); projected.xyz /= projected.w; projected.xy = projected.xy * 0.5 + 0.5; w = projected.w; @@ -46,24 +51,16 @@ vec2 view_to_screen(vec3 view_pos, out float w) { #define M_PI 3.14159265359 -vec3 reconstructCSPosition(vec2 S, float z) { - if (params.orthogonal) { - return vec3((S.xy * params.proj_info.xy + params.proj_info.zw), z); - } else { - return vec3((S.xy * params.proj_info.xy + params.proj_info.zw) * z, z); - } -} - void main() { // Pixel being shaded ivec2 ssC = ivec2(gl_GlobalInvocationID.xy); - if (any(greaterThanEqual(ssC, params.screen_size))) { //too large, do nothing + if (any(greaterThanEqual(ssC.xy, params.screen_size))) { //too large, do nothing return; } vec2 pixel_size = 1.0 / vec2(params.screen_size); - vec2 uv = vec2(ssC) * pixel_size; + vec2 uv = vec2(ssC.xy) * pixel_size; uv += pixel_size * 0.5; @@ -77,7 +74,12 @@ void main() { normal = normalize(normal); normal.y = -normal.y; //because this code reads flipped - vec3 view_dir = normalize(vertex); + vec3 view_dir; + if (sc_multiview) { + view_dir = normalize(vertex + scene_data.eye_offset[params.view_index].xyz); + } else { + view_dir = normalize(vertex); + } vec3 ray_dir = normalize(reflect(view_dir, normal)); if (dot(ray_dir, normal) < 0.001) { @@ -154,6 +156,11 @@ void main() { // convert to linear depth depth = imageLoad(source_depth, ivec2(pos - 0.5)).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)); + depth = -depth; + } z_from = z_to; z_to = z / w; @@ -222,13 +229,16 @@ void main() { blur_radius = (a * (sqrt(a2 + fh2) - a)) / (4.0f * h); } } + + // 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 +#endif // MODE_ROUGH final_color = vec4(imageLoad(source_diffuse, ivec2(final_pos - 0.5)).rgb, fade * margin_blend); + //change blend by metallic vec4 metallic_mask = unpackUnorm4x8(params.metallic_mask); final_color.a *= dot(metallic_mask, texelFetch(source_metallic, ssC << 1, 0)); diff --git a/servers/rendering/renderer_rd/shaders/screen_space_reflection_filter.glsl b/servers/rendering/renderer_rd/shaders/effects/screen_space_reflection_filter.glsl index 20e1712496..a63d60e0b2 100644 --- a/servers/rendering/renderer_rd/shaders/screen_space_reflection_filter.glsl +++ b/servers/rendering/renderer_rd/shaders/effects/screen_space_reflection_filter.glsl @@ -22,7 +22,7 @@ layout(push_constant, std430) uniform Params { bool orthogonal; float edge_tolerance; int increment; - uint pad; + uint view_index; ivec2 screen_size; bool vertical; @@ -30,6 +30,8 @@ layout(push_constant, std430) uniform Params { } params; +#include "screen_space_reflection_inc.glsl" + #define GAUSS_TABLE_SIZE 15 const float gauss_table[GAUSS_TABLE_SIZE + 1] = float[]( @@ -64,14 +66,6 @@ float gauss_weight(float p_val) { #define M_PI 3.14159265359 -vec3 reconstructCSPosition(vec2 S, float z) { - if (params.orthogonal) { - return vec3((S.xy * params.proj_info.xy + params.proj_info.zw), z); - } else { - return vec3((S.xy * params.proj_info.xy + params.proj_info.zw) * z, z); - } -} - void do_filter(inout vec4 accum, inout float accum_radius, inout float divisor, ivec2 texcoord, ivec2 increment, vec3 p_pos, vec3 normal, float p_limit_radius) { for (int i = 1; i < params.steps; i++) { float d = float(i * params.increment); @@ -110,7 +104,7 @@ void main() { // Pixel being shaded ivec2 ssC = ivec2(gl_GlobalInvocationID.xy); - if (any(greaterThanEqual(ssC, params.screen_size))) { //too large, do nothing + if (any(greaterThanEqual(ssC.xy, params.screen_size))) { //too large, do nothing return; } @@ -130,13 +124,13 @@ void main() { ivec2 direction = ivec2(params.increment, 0); #endif float depth = imageLoad(source_depth, ssC).r; - vec3 pos = reconstructCSPosition(vec2(ssC) + 0.5, depth); + vec3 pos = reconstructCSPosition(vec2(ssC.xy) + 0.5, depth); vec3 normal = imageLoad(source_normal, ssC).xyz * 2.0 - 1.0; normal = normalize(normal); normal.y = -normal.y; - do_filter(accum, accum_radius, divisor, ssC, direction, pos, normal, radius); - do_filter(accum, accum_radius, divisor, ssC, -direction, pos, normal, radius); + do_filter(accum, accum_radius, divisor, ssC.xy, direction, pos, normal, radius); + do_filter(accum, accum_radius, divisor, ssC.xy, -direction, pos, normal, radius); if (divisor > 0.0) { accum /= divisor; diff --git a/servers/rendering/renderer_rd/shaders/effects/screen_space_reflection_inc.glsl b/servers/rendering/renderer_rd/shaders/effects/screen_space_reflection_inc.glsl new file mode 100644 index 0000000000..26405ab040 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/effects/screen_space_reflection_inc.glsl @@ -0,0 +1,28 @@ +layout(constant_id = 0) const bool sc_multiview = false; + +layout(set = 4, binding = 0, std140) uniform SceneData { + mat4x4 projection[2]; + mat4x4 inv_projection[2]; + vec4 eye_offset[2]; +} +scene_data; + +vec3 reconstructCSPosition(vec2 screen_pos, float z) { + if (sc_multiview) { + vec4 pos; + pos.xy = (2.0 * vec2(screen_pos) / vec2(params.screen_size)) - 1.0; + pos.z = z * 2.0 - 1.0; + pos.w = 1.0; + + pos = scene_data.inv_projection[params.view_index] * pos; + pos.xyz /= pos.w; + + return pos.xyz; + } else { + if (params.orthogonal) { + return vec3((screen_pos.xy * params.proj_info.xy + params.proj_info.zw), z); + } else { + return vec3((screen_pos.xy * params.proj_info.xy + params.proj_info.zw) * z, z); + } + } +} diff --git a/servers/rendering/renderer_rd/shaders/screen_space_reflection_scale.glsl b/servers/rendering/renderer_rd/shaders/effects/screen_space_reflection_scale.glsl index 3f537e273a..a7da0812df 100644 --- a/servers/rendering/renderer_rd/shaders/screen_space_reflection_scale.glsl +++ b/servers/rendering/renderer_rd/shaders/effects/screen_space_reflection_scale.glsl @@ -6,6 +6,11 @@ layout(local_size_x = 8, local_size_y = 8, local_size_z = 1) in; +/* Specialization Constants (Toggles) */ + +layout(constant_id = 0) const bool sc_multiview = false; + +/* inputs */ layout(set = 0, binding = 0) uniform sampler2D source_ssr; layout(set = 1, binding = 0) uniform sampler2D source_depth; layout(set = 1, binding = 1) uniform sampler2D source_normal; @@ -28,7 +33,7 @@ void main() { // Pixel being shaded ivec2 ssC = ivec2(gl_GlobalInvocationID.xy); - if (any(greaterThanEqual(ssC, params.screen_size))) { //too large, do nothing + if (any(greaterThanEqual(ssC.xy, params.screen_size))) { //too large, do nothing return; } //do not filter, SSR will generate arctifacts if this is done @@ -57,13 +62,19 @@ void main() { normal.xyz += nr.xyz * 2.0 - 1.0; normal.w += nr.w; - d = d * 2.0 - 1.0; - if (params.orthogonal) { - d = ((d + (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; + if (sc_multiview) { + // we're doing a full unproject so we need the value as is. + depth += d; } else { - d = 2.0 * params.camera_z_near * params.camera_z_far / (params.camera_z_far + params.camera_z_near - d * (params.camera_z_far - params.camera_z_near)); + // unproject our Z value so we can use it directly. + d = d * 2.0 - 1.0; + if (params.orthogonal) { + d = ((d + (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; + } else { + d = 2.0 * params.camera_z_near * params.camera_z_far / (params.camera_z_far + params.camera_z_near - d * (params.camera_z_far - params.camera_z_near)); + } + depth += -d; } - depth += -d; } color /= 4.0; @@ -71,17 +82,22 @@ void main() { normal.xyz = normalize(normal.xyz / 4.0) * 0.5 + 0.5; normal.w /= 4.0; } else { - color = texelFetch(source_ssr, ssC << 1, 0); - depth = texelFetch(source_depth, ssC << 1, 0).r; - normal = texelFetch(source_normal, ssC << 1, 0); - - depth = depth * 2.0 - 1.0; - 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; - } 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)); + ivec2 ofs = ssC << 1; + + color = texelFetch(source_ssr, ofs, 0); + depth = texelFetch(source_depth, ofs, 0).r; + normal = texelFetch(source_normal, ofs, 0); + + if (!sc_multiview) { + // unproject our Z value so we can use it directly. + depth = depth * 2.0 - 1.0; + 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; + } 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 = -depth; } - depth = -depth; } imageStore(dest_ssr, ssC, color); diff --git a/servers/rendering/renderer_rd/shaders/effects/specular_merge.glsl b/servers/rendering/renderer_rd/shaders/effects/specular_merge.glsl new file mode 100644 index 0000000000..c62144fdaf --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/effects/specular_merge.glsl @@ -0,0 +1,112 @@ +#[vertex] + +#version 450 + +#VERSION_DEFINES + +#if defined(USE_MULTIVIEW) && defined(has_VK_KHR_multiview) +#extension GL_EXT_multiview : enable +#endif + +#ifdef USE_MULTIVIEW +#ifdef has_VK_KHR_multiview +#define ViewIndex gl_ViewIndex +#else // has_VK_KHR_multiview +// !BAS! This needs to become an input once we implement our fallback! +#define ViewIndex 0 +#endif // has_VK_KHR_multiview +#else // USE_MULTIVIEW +// Set to zero, not supported in non stereo +#define ViewIndex 0 +#endif //USE_MULTIVIEW + +#ifdef USE_MULTIVIEW +layout(location = 0) out vec3 uv_interp; +#else // USE_MULTIVIEW +layout(location = 0) out vec2 uv_interp; +#endif //USE_MULTIVIEW + +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)); + +#ifdef USE_MULTIVIEW + uv_interp = vec3(base_arr[gl_VertexIndex], ViewIndex); + + gl_Position = vec4(uv_interp.xy * 2.0 - 1.0, 0.0, 1.0); +#else + uv_interp = base_arr[gl_VertexIndex]; + + gl_Position = vec4(uv_interp * 2.0 - 1.0, 0.0, 1.0); +#endif +} + +#[fragment] + +#version 450 + +#VERSION_DEFINES + +#if defined(USE_MULTIVIEW) && defined(has_VK_KHR_multiview) +#extension GL_EXT_multiview : enable +#endif + +#ifdef USE_MULTIVIEW +#ifdef has_VK_KHR_multiview +#define ViewIndex gl_ViewIndex +#else // has_VK_KHR_multiview +// !BAS! This needs to become an input once we implement our fallback! +#define ViewIndex 0 +#endif // has_VK_KHR_multiview +#else // USE_MULTIVIEW +// Set to zero, not supported in non stereo +#define ViewIndex 0 +#endif //USE_MULTIVIEW + +#ifdef USE_MULTIVIEW +layout(location = 0) in vec3 uv_interp; +#else // USE_MULTIVIEW +layout(location = 0) in vec2 uv_interp; +#endif //USE_MULTIVIEW + +#ifdef USE_MULTIVIEW +layout(set = 0, binding = 0) uniform sampler2DArray specular; +#else // USE_MULTIVIEW +layout(set = 0, binding = 0) uniform sampler2D specular; +#endif //USE_MULTIVIEW + +#ifdef MODE_SSR + +#ifdef USE_MULTIVIEW +layout(set = 1, binding = 0) uniform sampler2DArray ssr; +#else // USE_MULTIVIEW +layout(set = 1, binding = 0) uniform sampler2D ssr; +#endif //USE_MULTIVIEW + +#endif + +#ifdef MODE_MERGE + +#ifdef USE_MULTIVIEW +layout(set = 2, binding = 0) uniform sampler2DArray diffuse; +#else // USE_MULTIVIEW +layout(set = 2, binding = 0) uniform sampler2D diffuse; +#endif //USE_MULTIVIEW + +#endif + +layout(location = 0) out vec4 frag_color; + +void main() { + frag_color.rgb = texture(specular, uv_interp).rgb; + frag_color.a = 0.0; +#ifdef MODE_SSR + + vec4 ssr_color = texture(ssr, uv_interp); + frag_color.rgb = mix(frag_color.rgb, ssr_color.rgb, ssr_color.a); +#endif + +#ifdef MODE_MERGE + frag_color += texture(diffuse, uv_interp); +#endif + //added using additive blend +} diff --git a/servers/rendering/renderer_rd/shaders/ss_effects_downsample.glsl b/servers/rendering/renderer_rd/shaders/effects/ss_effects_downsample.glsl index 134aae5ce7..134aae5ce7 100644 --- a/servers/rendering/renderer_rd/shaders/ss_effects_downsample.glsl +++ b/servers/rendering/renderer_rd/shaders/effects/ss_effects_downsample.glsl diff --git a/servers/rendering/renderer_rd/shaders/ssao.glsl b/servers/rendering/renderer_rd/shaders/effects/ssao.glsl index 2a87e273bc..2a87e273bc 100644 --- a/servers/rendering/renderer_rd/shaders/ssao.glsl +++ b/servers/rendering/renderer_rd/shaders/effects/ssao.glsl diff --git a/servers/rendering/renderer_rd/shaders/ssao_blur.glsl b/servers/rendering/renderer_rd/shaders/effects/ssao_blur.glsl index f42734c46d..f42734c46d 100644 --- a/servers/rendering/renderer_rd/shaders/ssao_blur.glsl +++ b/servers/rendering/renderer_rd/shaders/effects/ssao_blur.glsl diff --git a/servers/rendering/renderer_rd/shaders/ssao_importance_map.glsl b/servers/rendering/renderer_rd/shaders/effects/ssao_importance_map.glsl index 04f98964e8..04f98964e8 100644 --- a/servers/rendering/renderer_rd/shaders/ssao_importance_map.glsl +++ b/servers/rendering/renderer_rd/shaders/effects/ssao_importance_map.glsl diff --git a/servers/rendering/renderer_rd/shaders/ssao_interleave.glsl b/servers/rendering/renderer_rd/shaders/effects/ssao_interleave.glsl index f6a9a92fac..f6a9a92fac 100644 --- a/servers/rendering/renderer_rd/shaders/ssao_interleave.glsl +++ b/servers/rendering/renderer_rd/shaders/effects/ssao_interleave.glsl diff --git a/servers/rendering/renderer_rd/shaders/ssil.glsl b/servers/rendering/renderer_rd/shaders/effects/ssil.glsl index 513791dfbf..513791dfbf 100644 --- a/servers/rendering/renderer_rd/shaders/ssil.glsl +++ b/servers/rendering/renderer_rd/shaders/effects/ssil.glsl diff --git a/servers/rendering/renderer_rd/shaders/ssil_blur.glsl b/servers/rendering/renderer_rd/shaders/effects/ssil_blur.glsl index ee21d46a74..47c56571f6 100644 --- a/servers/rendering/renderer_rd/shaders/ssil_blur.glsl +++ b/servers/rendering/renderer_rd/shaders/effects/ssil_blur.glsl @@ -1,3 +1,23 @@ +/////////////////////////////////////////////////////////////////////////////////////////////////////////////////////// +// Copyright (c) 2016, Intel Corporation +// 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) +// 2016-09-07: filip.strugar@intel.com: first commit +// 2020-12-05: clayjohn: convert to Vulkan and Godot +// 2021-05-27: clayjohn: convert SSAO to SSIL +/////////////////////////////////////////////////////////////////////////////////////////////////////////////////////// + #[compute] #version 450 diff --git a/servers/rendering/renderer_rd/shaders/ssil_importance_map.glsl b/servers/rendering/renderer_rd/shaders/effects/ssil_importance_map.glsl index 8818f8cada..6b6b02739d 100644 --- a/servers/rendering/renderer_rd/shaders/ssil_importance_map.glsl +++ b/servers/rendering/renderer_rd/shaders/effects/ssil_importance_map.glsl @@ -15,6 +15,7 @@ // File changes (yyyy-mm-dd) // 2016-09-07: filip.strugar@intel.com: first commit // 2020-12-05: clayjohn: convert to Vulkan and Godot +// 2021-05-27: clayjohn: convert SSAO to SSIL /////////////////////////////////////////////////////////////////////////////////////////////////////////////////////// #[compute] diff --git a/servers/rendering/renderer_rd/shaders/ssil_interleave.glsl b/servers/rendering/renderer_rd/shaders/effects/ssil_interleave.glsl index fa4309353d..9e86ac0cf0 100644 --- a/servers/rendering/renderer_rd/shaders/ssil_interleave.glsl +++ b/servers/rendering/renderer_rd/shaders/effects/ssil_interleave.glsl @@ -1,3 +1,23 @@ +/////////////////////////////////////////////////////////////////////////////////////////////////////////////////////// +// Copyright (c) 2016, Intel Corporation +// 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) +// 2016-09-07: filip.strugar@intel.com: first commit +// 2020-12-05: clayjohn: convert to Vulkan and Godot +// 2021-05-27: clayjohn: convert SSAO to SSIL +/////////////////////////////////////////////////////////////////////////////////////////////////////////////////////// + #[compute] #version 450 diff --git a/servers/rendering/renderer_rd/shaders/tonemap.glsl b/servers/rendering/renderer_rd/shaders/effects/tonemap.glsl index 19a9350137..62a7b0e7d7 100644 --- a/servers/rendering/renderer_rd/shaders/tonemap.glsl +++ b/servers/rendering/renderer_rd/shaders/effects/tonemap.glsl @@ -44,7 +44,11 @@ layout(set = 0, binding = 0) uniform sampler2D source_color; #endif layout(set = 1, binding = 0) uniform sampler2D source_auto_exposure; +#ifdef MULTIVIEW +layout(set = 2, binding = 0) uniform sampler2DArray source_glow; +#else layout(set = 2, binding = 0) uniform sampler2D source_glow; +#endif layout(set = 2, binding = 1) uniform sampler2D glow_map; #ifdef USE_1D_LUT @@ -118,6 +122,36 @@ float h1(float a) { return 1.0f + w3(a) / (w2(a) + w3(a)); } +#ifdef MULTIVIEW +vec4 texture2D_bicubic(sampler2DArray tex, vec2 uv, int p_lod) { + float lod = float(p_lod); + vec2 tex_size = vec2(params.glow_texture_size >> p_lod); + vec2 pixel_size = vec2(1.0f) / tex_size; + + uv = uv * tex_size + vec2(0.5f); + + vec2 iuv = floor(uv); + vec2 fuv = fract(uv); + + float g0x = g0(fuv.x); + float g1x = g1(fuv.x); + float h0x = h0(fuv.x); + float h1x = h1(fuv.x); + float h0y = h0(fuv.y); + float h1y = h1(fuv.y); + + vec3 p0 = vec3((vec2(iuv.x + h0x, iuv.y + h0y) - vec2(0.5f)) * pixel_size, ViewIndex); + vec3 p1 = vec3((vec2(iuv.x + h1x, iuv.y + h0y) - vec2(0.5f)) * pixel_size, ViewIndex); + vec3 p2 = vec3((vec2(iuv.x + h0x, iuv.y + h1y) - vec2(0.5f)) * pixel_size, ViewIndex); + vec3 p3 = vec3((vec2(iuv.x + h1x, iuv.y + h1y) - vec2(0.5f)) * pixel_size, ViewIndex); + + return (g0(fuv.y) * (g0x * textureLod(tex, p0, lod) + g1x * textureLod(tex, p1, lod))) + + (g1(fuv.y) * (g0x * textureLod(tex, p2, lod) + g1x * textureLod(tex, p3, lod))); +} + +#define GLOW_TEXTURE_SAMPLE(m_tex, m_uv, m_lod) texture2D_bicubic(m_tex, m_uv, m_lod) +#else // MULTIVIEW + vec4 texture2D_bicubic(sampler2D tex, vec2 uv, int p_lod) { float lod = float(p_lod); vec2 tex_size = vec2(params.glow_texture_size >> p_lod); @@ -145,12 +179,17 @@ vec4 texture2D_bicubic(sampler2D tex, vec2 uv, int p_lod) { } #define GLOW_TEXTURE_SAMPLE(m_tex, m_uv, m_lod) texture2D_bicubic(m_tex, m_uv, m_lod) +#endif // !MULTIVIEW -#else +#else // USE_GLOW_FILTER_BICUBIC +#ifdef MULTIVIEW +#define GLOW_TEXTURE_SAMPLE(m_tex, m_uv, m_lod) textureLod(m_tex, vec3(m_uv, ViewIndex), float(m_lod)) +#else // MULTIVIEW #define GLOW_TEXTURE_SAMPLE(m_tex, m_uv, m_lod) textureLod(m_tex, m_uv, float(m_lod)) +#endif // !MULTIVIEW -#endif +#endif // !USE_GLOW_FILTER_BICUBIC vec3 tonemap_filmic(vec3 color, float white) { // exposure bias: input scale (color *= bias, white *= bias) to make the brightness consistent with other tonemappers @@ -231,7 +270,11 @@ vec3 apply_tonemapping(vec3 color, float white) { // inputs are LINEAR, always o } } +#ifdef MULTIVIEW +vec3 gather_glow(sampler2DArray tex, vec2 uv) { // sample all selected glow levels, view is added to uv later +#else vec3 gather_glow(sampler2D tex, vec2 uv) { // sample all selected glow levels +#endif // defined(MULTIVIEW) vec3 glow = vec3(0.0f); if (params.glow_levels[0] > 0.0001) { @@ -383,12 +426,13 @@ vec3 screen_space_dither(vec2 frag_coord) { void main() { #ifdef SUBPASS // SUBPASS and MULTIVIEW can be combined but in that case we're already reading from the correct layer - vec3 color = subpassLoad(input_color).rgb * params.luminance_multiplier; + vec4 color = subpassLoad(input_color); #elif defined(MULTIVIEW) - vec3 color = textureLod(source_color, vec3(uv_interp, ViewIndex), 0.0f).rgb * params.luminance_multiplier; + vec4 color = textureLod(source_color, vec3(uv_interp, ViewIndex), 0.0f); #else - vec3 color = textureLod(source_color, uv_interp, 0.0f).rgb * params.luminance_multiplier; + vec4 color = textureLod(source_color, uv_interp, 0.0f); #endif + color.rgb *= params.luminance_multiplier; // Exposure @@ -400,10 +444,15 @@ void main() { } #endif - color *= exposure; + color.rgb *= exposure; // Early Tonemap & SRGB Conversion #ifndef SUBPASS + if (params.use_fxaa) { + // FXAA must be performed before glow to preserve the "bleed" effect of glow. + color.rgb = do_fxaa(color.rgb, exposure, uv_interp); + } + if (params.use_glow && params.glow_mode == GLOW_MODE_MIX) { vec3 glow = gather_glow(source_glow, uv_interp) * params.luminance_multiplier; if (params.glow_map_strength > 0.001) { @@ -411,21 +460,17 @@ void main() { } color.rgb = mix(color.rgb, glow, params.glow_intensity); } - - if (params.use_fxaa) { - color = do_fxaa(color, exposure, uv_interp); - } #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 += screen_space_dither(gl_FragCoord.xy); + color.rgb += screen_space_dither(gl_FragCoord.xy); } - color = apply_tonemapping(color, params.white); + color.rgb = apply_tonemapping(color.rgb, params.white); - color = linear_to_srgb(color); // regular linear -> SRGB conversion + color.rgb = linear_to_srgb(color.rgb); // regular linear -> SRGB conversion #ifndef SUBPASS // Glow @@ -439,19 +484,19 @@ void main() { glow = apply_tonemapping(glow, params.white); glow = linear_to_srgb(glow); - color = apply_glow(color, glow); + color.rgb = apply_glow(color.rgb, glow); } #endif // Additional effects if (params.use_bcs) { - color = apply_bcs(color, params.bcs); + color.rgb = apply_bcs(color.rgb, params.bcs); } if (params.use_color_correction) { - color = apply_color_correction(color); + color.rgb = apply_color_correction(color.rgb); } - frag_color = vec4(color, 1.0f); + frag_color = color; } diff --git a/servers/rendering/renderer_rd/shaders/effects/vrs.glsl b/servers/rendering/renderer_rd/shaders/effects/vrs.glsl new file mode 100644 index 0000000000..5ef83c0b44 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/effects/vrs.glsl @@ -0,0 +1,72 @@ +#[vertex] + +#version 450 + +#VERSION_DEFINES + +#ifdef MULTIVIEW +#ifdef has_VK_KHR_multiview +#extension GL_EXT_multiview : enable +#define ViewIndex gl_ViewIndex +#else // has_VK_KHR_multiview +#define ViewIndex 0 +#endif // has_VK_KHR_multiview +#endif //MULTIVIEW + +#ifdef MULTIVIEW +layout(location = 0) out vec3 uv_interp; +#else +layout(location = 0) out vec2 uv_interp; +#endif + +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.xy = base_arr[gl_VertexIndex]; +#ifdef MULTIVIEW + uv_interp.z = ViewIndex; +#endif + + gl_Position = vec4(uv_interp.xy * 2.0 - 1.0, 0.0, 1.0); +} + +#[fragment] + +#version 450 + +#VERSION_DEFINES + +#ifdef MULTIVIEW +#ifdef has_VK_KHR_multiview +#extension GL_EXT_multiview : enable +#define ViewIndex gl_ViewIndex +#else // has_VK_KHR_multiview +#define ViewIndex 0 +#endif // has_VK_KHR_multiview +#endif //MULTIVIEW + +#ifdef MULTIVIEW +layout(location = 0) in vec3 uv_interp; +layout(set = 0, binding = 0) uniform sampler2DArray source_color; +#else /* MULTIVIEW */ +layout(location = 0) in vec2 uv_interp; +layout(set = 0, binding = 0) uniform sampler2D source_color; +#endif /* MULTIVIEW */ + +layout(location = 0) out uint frag_color; + +void main() { +#ifdef MULTIVIEW + vec3 uv = uv_interp; +#else + vec2 uv = uv_interp; +#endif + +#ifdef MULTIVIEW + vec4 color = textureLod(source_color, uv, 0.0); +#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); +} diff --git a/servers/rendering/renderer_rd/shaders/environment/SCsub b/servers/rendering/renderer_rd/shaders/environment/SCsub new file mode 100644 index 0000000000..741da8fe69 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/environment/SCsub @@ -0,0 +1,17 @@ +#!/usr/bin/env python + +Import("env") + +if "RD_GLSL" in env["BUILDERS"]: + # find all include files + gl_include_files = [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] + + # make sure we recompile shaders if include files change + env.Depends([f + ".gen.h" for f in glsl_files], gl_include_files + ["#glsl_builders.py"]) + + # compile shaders + for glsl_file in glsl_files: + env.RD_GLSL(glsl_file) diff --git a/servers/rendering/renderer_rd/shaders/gi.glsl b/servers/rendering/renderer_rd/shaders/environment/gi.glsl index 0c7f08813b..6ea8cb1377 100644 --- a/servers/rendering/renderer_rd/shaders/gi.glsl +++ b/servers/rendering/renderer_rd/shaders/environment/gi.glsl @@ -8,6 +8,12 @@ layout(local_size_x = 8, local_size_y = 8, local_size_z = 1) in; #define M_PI 3.141592 +/* Specialization Constants (Toggles) */ + +layout(constant_id = 0) const bool sc_half_res = false; +layout(constant_id = 1) const bool sc_use_full_projection_matrix = false; +layout(constant_id = 2) const bool sc_use_vrs = false; + #define SDFGI_MAX_CASCADES 8 //set 0 for SDFGI and render buffers @@ -86,19 +92,31 @@ voxel_gi_instances; layout(set = 0, binding = 17) uniform texture3D voxel_gi_textures[MAX_VOXEL_GI_INSTANCES]; -layout(push_constant, std430) uniform Params { +layout(set = 0, binding = 18, std140) uniform SceneData { + mat4x4 inv_projection[2]; + mat4x4 cam_transform; + vec4 eye_offset[2]; + ivec2 screen_size; - float z_near; - float z_far; + float pad1; + float pad2; +} +scene_data; - vec4 proj_info; +layout(r8ui, set = 0, binding = 19) uniform restrict readonly uimage2D vrs_buffer; +layout(push_constant, std430) uniform Params { uint max_voxel_gi_instances; bool high_quality_vct; bool orthogonal; - uint pad; + uint view_index; + + vec4 proj_info; - mat3x4 cam_rotation; + float z_near; + float z_far; + float pad2; + float pad3; } params; @@ -130,23 +148,34 @@ vec4 blend_color(vec4 src, vec4 dst) { } vec3 reconstruct_position(ivec2 screen_pos) { - vec3 pos; - pos.z = texelFetch(sampler2D(depth_buffer, linear_sampler), screen_pos, 0).r; + if (sc_use_full_projection_matrix) { + vec4 pos; + pos.xy = (2.0 * vec2(screen_pos) / vec2(scene_data.screen_size)) - 1.0; + pos.z = texelFetch(sampler2D(depth_buffer, linear_sampler), screen_pos, 0).r * 2.0 - 1.0; + pos.w = 1.0; + + pos = scene_data.inv_projection[params.view_index] * pos; - pos.z = pos.z * 2.0 - 1.0; - if (params.orthogonal) { - pos.z = ((pos.z + (params.z_far + params.z_near) / (params.z_far - params.z_near)) * (params.z_far - params.z_near)) / 2.0; + return pos.xyz / pos.w; } else { - pos.z = 2.0 * params.z_near * params.z_far / (params.z_far + params.z_near - pos.z * (params.z_far - params.z_near)); - } - pos.z = -pos.z; + vec3 pos; + pos.z = texelFetch(sampler2D(depth_buffer, linear_sampler), screen_pos, 0).r; + + pos.z = pos.z * 2.0 - 1.0; + if (params.orthogonal) { + pos.z = ((pos.z + (params.z_far + params.z_near) / (params.z_far - params.z_near)) * (params.z_far - params.z_near)) / 2.0; + } else { + pos.z = 2.0 * params.z_near * params.z_far / (params.z_far + params.z_near - pos.z * (params.z_far - params.z_near)); + } + pos.z = -pos.z; - pos.xy = vec2(screen_pos) * params.proj_info.xy + params.proj_info.zw; - if (!params.orthogonal) { - pos.xy *= pos.z; - } + pos.xy = vec2(screen_pos) * params.proj_info.xy + params.proj_info.zw; + if (!params.orthogonal) { + pos.xy *= pos.z; + } - return pos; + return pos; + } } void sdfvoxel_gi_process(uint cascade, vec3 cascade_pos, vec3 cam_pos, vec3 cam_normal, vec3 cam_specular_normal, float roughness, out vec3 diffuse_light, out vec3 specular_light) { @@ -566,7 +595,6 @@ void voxel_gi_compute(uint index, vec3 position, vec3 normal, vec3 ref_vec, mat3 vec4 fetch_normal_and_roughness(ivec2 pos) { vec4 normal_roughness = texelFetch(sampler2D(normal_roughness_buffer, linear_sampler), pos, 0); - normal_roughness.xyz = normalize(normal_roughness.xyz * 2.0 - 1.0); return normal_roughness; } @@ -579,9 +607,10 @@ void process_gi(ivec2 pos, vec3 vertex, inout vec4 ambient_light, inout vec4 ref if (normal.length() > 0.5) { //valid normal, can do GI float roughness = normal_roughness.w; - vertex = mat3(params.cam_rotation) * vertex; - normal = normalize(mat3(params.cam_rotation) * normal); - vec3 reflection = normalize(reflect(normalize(vertex), normal)); + vec3 view = -normalize(mat3(scene_data.cam_transform) * (vertex - scene_data.eye_offset[gl_GlobalInvocationID.z].xyz)); + vertex = mat3(scene_data.cam_transform) * vertex; + normal = normalize(mat3(scene_data.cam_transform) * normal); + vec3 reflection = normalize(reflect(-view, normal)); #ifdef USE_SDFGI sdfgi_process(vertex, normal, reflection, roughness, ambient_light, reflection_light); @@ -626,10 +655,36 @@ void process_gi(ivec2 pos, vec3 vertex, inout vec4 ambient_light, inout vec4 ref void main() { ivec2 pos = ivec2(gl_GlobalInvocationID.xy); -#ifdef MODE_HALF_RES - pos <<= 1; -#endif - if (any(greaterThanEqual(pos, params.screen_size))) { //too large, do nothing + uint vrs_x, vrs_y; + if (sc_use_vrs) { + ivec2 vrs_pos; + + // Currently we use a 16x16 texel, possibly some day make this configurable. + if (sc_half_res) { + vrs_pos = pos >> 3; + } else { + vrs_pos = pos >> 4; + } + + uint vrs_texel = imageLoad(vrs_buffer, vrs_pos).r; + // note, valid values for vrs_x and vrs_y are 1, 2 and 4. + vrs_x = 1 << ((vrs_texel >> 2) & 3); + vrs_y = 1 << (vrs_texel & 3); + + if (mod(pos.x, vrs_x) != 0) { + return; + } + + if (mod(pos.y, vrs_y) != 0) { + return; + } + } + + if (sc_half_res) { + pos <<= 1; + } + + if (any(greaterThanEqual(pos, scene_data.screen_size))) { //too large, do nothing return; } @@ -641,10 +696,69 @@ void main() { process_gi(pos, vertex, ambient_light, reflection_light); -#ifdef MODE_HALF_RES - pos >>= 1; -#endif + if (sc_half_res) { + pos >>= 1; + } imageStore(ambient_buffer, pos, ambient_light); imageStore(reflection_buffer, pos, reflection_light); + + if (sc_use_vrs) { + if (vrs_x > 1) { + imageStore(ambient_buffer, pos + ivec2(1, 0), ambient_light); + imageStore(reflection_buffer, pos + ivec2(1, 0), reflection_light); + } + + if (vrs_x > 2) { + imageStore(ambient_buffer, pos + ivec2(2, 0), ambient_light); + imageStore(reflection_buffer, pos + ivec2(2, 0), reflection_light); + + imageStore(ambient_buffer, pos + ivec2(3, 0), ambient_light); + imageStore(reflection_buffer, pos + ivec2(3, 0), reflection_light); + } + + if (vrs_y > 1) { + imageStore(ambient_buffer, pos + ivec2(0, 1), ambient_light); + imageStore(reflection_buffer, pos + ivec2(0, 1), reflection_light); + } + + if (vrs_y > 1 && vrs_x > 1) { + imageStore(ambient_buffer, pos + ivec2(1, 1), ambient_light); + imageStore(reflection_buffer, pos + ivec2(1, 1), reflection_light); + } + + if (vrs_y > 1 && vrs_x > 2) { + imageStore(ambient_buffer, pos + ivec2(2, 1), ambient_light); + imageStore(reflection_buffer, pos + ivec2(2, 1), reflection_light); + + imageStore(ambient_buffer, pos + ivec2(3, 1), ambient_light); + imageStore(reflection_buffer, pos + ivec2(3, 1), reflection_light); + } + + if (vrs_y > 2) { + imageStore(ambient_buffer, pos + ivec2(0, 2), ambient_light); + imageStore(reflection_buffer, pos + ivec2(0, 2), reflection_light); + imageStore(ambient_buffer, pos + ivec2(0, 3), ambient_light); + imageStore(reflection_buffer, pos + ivec2(0, 3), reflection_light); + } + + if (vrs_y > 2 && vrs_x > 1) { + imageStore(ambient_buffer, pos + ivec2(1, 2), ambient_light); + imageStore(reflection_buffer, pos + ivec2(1, 2), reflection_light); + imageStore(ambient_buffer, pos + ivec2(1, 3), ambient_light); + imageStore(reflection_buffer, pos + ivec2(1, 3), reflection_light); + } + + if (vrs_y > 2 && vrs_x > 2) { + imageStore(ambient_buffer, pos + ivec2(2, 2), ambient_light); + imageStore(reflection_buffer, pos + ivec2(2, 2), reflection_light); + imageStore(ambient_buffer, pos + ivec2(2, 3), ambient_light); + imageStore(reflection_buffer, pos + ivec2(2, 3), reflection_light); + + imageStore(ambient_buffer, pos + ivec2(3, 2), ambient_light); + imageStore(reflection_buffer, pos + ivec2(3, 2), reflection_light); + imageStore(ambient_buffer, pos + ivec2(3, 3), ambient_light); + imageStore(reflection_buffer, pos + ivec2(3, 3), reflection_light); + } + } } diff --git a/servers/rendering/renderer_rd/shaders/sdfgi_debug.glsl b/servers/rendering/renderer_rd/shaders/environment/sdfgi_debug.glsl index 802a410825..af5f7d0a58 100644 --- a/servers/rendering/renderer_rd/shaders/sdfgi_debug.glsl +++ b/servers/rendering/renderer_rd/shaders/environment/sdfgi_debug.glsl @@ -40,10 +40,13 @@ layout(push_constant, std430) uniform Params { bool use_occlusion; float y_mult; - vec3 cam_extent; int probe_axis_size; + float z_near; + float reserved1; + float reserved2; mat4 cam_transform; + mat4 inv_projection; } params; @@ -81,8 +84,9 @@ void main() { { ray_pos = params.cam_transform[3].xyz; - ray_dir.xy = params.cam_extent.xy * ((vec2(screen_pos) / vec2(params.screen_size)) * 2.0 - 1.0); - ray_dir.z = params.cam_extent.z; + ray_dir.xy = ((vec2(screen_pos) / vec2(params.screen_size)) * 2.0 - 1.0); + ray_dir.z = params.z_near; + ray_dir = (params.inv_projection * vec4(ray_dir, 1.0)).xyz; ray_dir = normalize(mat3(params.cam_transform) * ray_dir); } diff --git a/servers/rendering/renderer_rd/shaders/sdfgi_debug_probes.glsl b/servers/rendering/renderer_rd/shaders/environment/sdfgi_debug_probes.glsl index e0be0bca12..75b1ad2130 100644 --- a/servers/rendering/renderer_rd/shaders/sdfgi_debug_probes.glsl +++ b/servers/rendering/renderer_rd/shaders/environment/sdfgi_debug_probes.glsl @@ -2,13 +2,28 @@ #version 450 +#if defined(USE_MULTIVIEW) && defined(has_VK_KHR_multiview) +#extension GL_EXT_multiview : enable +#endif + +#ifdef USE_MULTIVIEW +#ifdef has_VK_KHR_multiview +#define ViewIndex gl_ViewIndex +#else // has_VK_KHR_multiview +// !BAS! This needs to become an input once we implement our fallback! +#define ViewIndex 0 +#endif // has_VK_KHR_multiview +#else // USE_MULTIVIEW +// Set to zero, not supported in non stereo +#define ViewIndex 0 +#endif //USE_MULTIVIEW + #VERSION_DEFINES #define MAX_CASCADES 8 +#define MAX_VIEWS 2 layout(push_constant, std430) uniform Params { - mat4 projection; - uint band_power; uint sections_in_band; uint band_mask; @@ -68,6 +83,11 @@ cascades; layout(set = 0, binding = 4) uniform texture3D occlusion_texture; layout(set = 0, binding = 3) uniform sampler linear_sampler; +layout(set = 0, binding = 5, std140) uniform SceneData { + mat4 projection[MAX_VIEWS]; +} +scene_data; + void main() { #ifdef MODE_PROBES probe_index = gl_InstanceIndex; @@ -85,7 +105,7 @@ void main() { vertex += (cascades.data[params.cascade].offset + vec3(probe_cell) * probe_cell_size) / vec3(1.0, params.y_mult, 1.0); - gl_Position = params.projection * vec4(vertex, 1.0); + gl_Position = scene_data.projection[ViewIndex] * vec4(vertex, 1.0); #endif #ifdef MODE_VISIBILITY @@ -144,7 +164,7 @@ void main() { visibility = dot(texelFetch(sampler3D(occlusion_texture, linear_sampler), tex_pos, 0), layer_axis[occlusion_layer]); - gl_Position = params.projection * vec4(vertex, 1.0); + gl_Position = scene_data.projection[ViewIndex] * vec4(vertex, 1.0); #endif } @@ -153,16 +173,32 @@ void main() { #version 450 +#if defined(USE_MULTIVIEW) && defined(has_VK_KHR_multiview) +#extension GL_EXT_multiview : enable +#endif + +#ifdef USE_MULTIVIEW +#ifdef has_VK_KHR_multiview +#define ViewIndex gl_ViewIndex +#else // has_VK_KHR_multiview +// !BAS! This needs to become an input once we implement our fallback! +#define ViewIndex 0 +#endif // has_VK_KHR_multiview +#else // USE_MULTIVIEW +// Set to zero, not supported in non stereo +#define ViewIndex 0 +#endif //USE_MULTIVIEW + #VERSION_DEFINES +#define MAX_VIEWS 2 + layout(location = 0) out vec4 frag_color; layout(set = 0, binding = 2) uniform texture2DArray lightprobe_texture; layout(set = 0, binding = 3) uniform sampler linear_sampler; layout(push_constant, std430) uniform Params { - mat4 projection; - uint band_power; uint sections_in_band; uint band_mask; diff --git a/servers/rendering/renderer_rd/shaders/sdfgi_direct_light.glsl b/servers/rendering/renderer_rd/shaders/environment/sdfgi_direct_light.glsl index 5bda15236c..b95fad650e 100644 --- a/servers/rendering/renderer_rd/shaders/sdfgi_direct_light.glsl +++ b/servers/rendering/renderer_rd/shaders/environment/sdfgi_direct_light.glsl @@ -70,8 +70,6 @@ struct Light { float cos_spot_angle; float inv_spot_attenuation; float radius; - - vec4 shadow_color; }; layout(set = 0, binding = 9, std140) buffer restrict readonly Lights { diff --git a/servers/rendering/renderer_rd/shaders/sdfgi_integrate.glsl b/servers/rendering/renderer_rd/shaders/environment/sdfgi_integrate.glsl index 9c03297f5c..9c03297f5c 100644 --- a/servers/rendering/renderer_rd/shaders/sdfgi_integrate.glsl +++ b/servers/rendering/renderer_rd/shaders/environment/sdfgi_integrate.glsl diff --git a/servers/rendering/renderer_rd/shaders/sdfgi_preprocess.glsl b/servers/rendering/renderer_rd/shaders/environment/sdfgi_preprocess.glsl index bce98f4054..bce98f4054 100644 --- a/servers/rendering/renderer_rd/shaders/sdfgi_preprocess.glsl +++ b/servers/rendering/renderer_rd/shaders/environment/sdfgi_preprocess.glsl diff --git a/servers/rendering/renderer_rd/shaders/sky.glsl b/servers/rendering/renderer_rd/shaders/environment/sky.glsl index b258e89c66..5b4594da99 100644 --- a/servers/rendering/renderer_rd/shaders/sky.glsl +++ b/servers/rendering/renderer_rd/shaders/environment/sky.glsl @@ -180,12 +180,11 @@ void main() { cube_normal.x = (cube_normal.z * (-uv_interp.x - params.projections[ViewIndex].x)) / params.projections[ViewIndex].y; cube_normal.y = -(cube_normal.z * (-uv_interp.y - params.projections[ViewIndex].z)) / params.projections[ViewIndex].w; cube_normal = mat3(params.orientation) * cube_normal; - cube_normal.z = -cube_normal.z; cube_normal = normalize(cube_normal); vec2 uv = uv_interp * 0.5 + 0.5; - vec2 panorama_coords = vec2(atan(cube_normal.x, cube_normal.z), acos(cube_normal.y)); + vec2 panorama_coords = vec2(atan(cube_normal.x, -cube_normal.z), acos(cube_normal.y)); if (panorama_coords.x < 0.0) { panorama_coords.x += M_PI * 2.0; @@ -200,13 +199,11 @@ void main() { vec4 custom_fog = vec4(0.0); #ifdef USE_CUBEMAP_PASS - vec3 inverted_cube_normal = cube_normal; - inverted_cube_normal.z *= -1.0; #ifdef USES_HALF_RES_COLOR - half_res_color = texture(samplerCube(half_res, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), inverted_cube_normal) * params.luminance_multiplier; + half_res_color = texture(samplerCube(half_res, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), cube_normal) * params.luminance_multiplier; #endif #ifdef USES_QUARTER_RES_COLOR - quarter_res_color = texture(samplerCube(quarter_res, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), inverted_cube_normal) * params.luminance_multiplier; + quarter_res_color = texture(samplerCube(quarter_res, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), cube_normal) * params.luminance_multiplier; #endif #else #ifdef USES_HALF_RES_COLOR diff --git a/servers/rendering/renderer_rd/shaders/volumetric_fog.glsl b/servers/rendering/renderer_rd/shaders/environment/volumetric_fog.glsl index a2a4c91894..fb3c725b1f 100644 --- a/servers/rendering/renderer_rd/shaders/volumetric_fog.glsl +++ b/servers/rendering/renderer_rd/shaders/environment/volumetric_fog.glsl @@ -21,8 +21,8 @@ layout(local_size_x = 4, local_size_y = 4, local_size_z = 4) in; #define DENSITY_SCALE 1024.0 -#include "cluster_data_inc.glsl" -#include "light_data_inc.glsl" +#include "../cluster_data_inc.glsl" +#include "../light_data_inc.glsl" #define M_PI 3.14159265359 @@ -186,12 +186,31 @@ void main() { float sdf = -1.0; if (params.shape == 0) { - //Ellipsoid + // Ellipsoid // https://www.shadertoy.com/view/tdS3DG float k0 = length(local_pos.xyz / params.extents); float k1 = length(local_pos.xyz / (params.extents * params.extents)); sdf = k0 * (k0 - 1.0) / k1; } else if (params.shape == 1) { + // Cone + // https://iquilezles.org/www/articles/distfunctions/distfunctions.htm + + // Compute the cone angle automatically to fit within the volume's extents. + float inv_height = 1.0 / max(0.001, params.extents.y); + float radius = 1.0 / max(0.001, (min(params.extents.x, params.extents.z) * 0.5)); + float hypotenuse = sqrt(radius * radius + inv_height * inv_height); + float rsin = radius / hypotenuse; + float rcos = inv_height / hypotenuse; + vec2 c = vec2(rsin, rcos); + + float q = length(local_pos.xz); + sdf = max(dot(c, vec2(q, local_pos.y - params.extents.y)), -params.extents.y - local_pos.y); + } else if (params.shape == 2) { + // Cylinder + // https://iquilezles.org/www/articles/distfunctions/distfunctions.htm + vec2 d = abs(vec2(length(local_pos.xz), local_pos.y)) - vec2(min(params.extents.x, params.extents.z), params.extents.y); + sdf = min(max(d.x, d.y), 0.0) + length(max(d, 0.0)); + } else if (params.shape == 3) { // Box // https://iquilezles.org/www/articles/distfunctions/distfunctions.htm vec3 q = abs(local_pos.xyz) - params.extents; @@ -199,7 +218,7 @@ void main() { } float cull_mask = 1.0; //used to cull cells that do not contribute - if (params.shape <= 1) { + if (params.shape <= 3) { #ifndef SDF_USED cull_mask = 1.0 - smoothstep(-0.1, 0.0, sdf); #endif diff --git a/servers/rendering/renderer_rd/shaders/volumetric_fog_process.glsl b/servers/rendering/renderer_rd/shaders/environment/volumetric_fog_process.glsl index 7a0cea421e..e74cfad65c 100644 --- a/servers/rendering/renderer_rd/shaders/volumetric_fog_process.glsl +++ b/servers/rendering/renderer_rd/shaders/environment/volumetric_fog_process.glsl @@ -19,8 +19,8 @@ layout(local_size_x = 4, local_size_y = 4, local_size_z = 4) in; layout(local_size_x = 8, local_size_y = 8, local_size_z = 1) in; #endif -#include "cluster_data_inc.glsl" -#include "light_data_inc.glsl" +#include "../cluster_data_inc.glsl" +#include "../light_data_inc.glsl" #define M_PI 3.14159265359 @@ -53,7 +53,6 @@ layout(set = 0, binding = 7) uniform sampler linear_sampler; #ifdef MODE_DENSITY layout(rgba16f, set = 0, binding = 8) uniform restrict writeonly image3D density_map; -layout(rgba16f, set = 0, binding = 9) uniform restrict readonly image3D fog_map; //unused #endif #ifdef MODE_FOG @@ -382,7 +381,6 @@ void main() { float depth_z = -view_pos.z; vec4 pssm_coord; - vec3 shadow_color = directional_lights.data[i].shadow_color1.rgb; vec3 light_dir = directional_lights.data[i].direction; vec4 v = vec4(view_pos, 1.0); float z_range; @@ -413,7 +411,7 @@ void main() { shadow = mix(shadow, 1.0, smoothstep(directional_lights.data[i].fade_from, directional_lights.data[i].fade_to, view_pos.z)); //done with negative values for performance - shadow_attenuation = mix(shadow_color, vec3(1.0), shadow); + shadow_attenuation = mix(vec3(0.0), vec3(1.0), shadow); } total_light += shadow_attenuation * directional_lights.data[i].color * directional_lights.data[i].energy * henyey_greenstein(dot(normalize(view_pos), normalize(directional_lights.data[i].direction)), params.phase_g); diff --git a/servers/rendering/renderer_rd/shaders/voxel_gi.glsl b/servers/rendering/renderer_rd/shaders/environment/voxel_gi.glsl index 577c6d0cd0..577c6d0cd0 100644 --- a/servers/rendering/renderer_rd/shaders/voxel_gi.glsl +++ b/servers/rendering/renderer_rd/shaders/environment/voxel_gi.glsl diff --git a/servers/rendering/renderer_rd/shaders/voxel_gi_debug.glsl b/servers/rendering/renderer_rd/shaders/environment/voxel_gi_debug.glsl index fd7a2bf8ad..fd7a2bf8ad 100644 --- a/servers/rendering/renderer_rd/shaders/voxel_gi_debug.glsl +++ b/servers/rendering/renderer_rd/shaders/environment/voxel_gi_debug.glsl diff --git a/servers/rendering/renderer_rd/shaders/voxel_gi_sdf.glsl b/servers/rendering/renderer_rd/shaders/environment/voxel_gi_sdf.glsl index 47a611a543..47a611a543 100644 --- a/servers/rendering/renderer_rd/shaders/voxel_gi_sdf.glsl +++ b/servers/rendering/renderer_rd/shaders/environment/voxel_gi_sdf.glsl diff --git a/servers/rendering/renderer_rd/shaders/light_data_inc.glsl b/servers/rendering/renderer_rd/shaders/light_data_inc.glsl index 52787bb204..61c8488a05 100644 --- a/servers/rendering/renderer_rd/shaders/light_data_inc.glsl +++ b/servers/rendering/renderer_rd/shaders/light_data_inc.glsl @@ -76,10 +76,6 @@ struct DirectionalLightData { highp mat4 shadow_matrix2; highp mat4 shadow_matrix3; highp mat4 shadow_matrix4; - mediump vec4 shadow_color1; - mediump vec4 shadow_color2; - mediump vec4 shadow_color3; - mediump vec4 shadow_color4; highp vec2 uv_scale1; highp vec2 uv_scale2; highp vec2 uv_scale3; diff --git a/servers/rendering/renderer_rd/shaders/particles.glsl b/servers/rendering/renderer_rd/shaders/particles.glsl index 1b1051ecfa..acb62b812e 100644 --- a/servers/rendering/renderer_rd/shaders/particles.glsl +++ b/servers/rendering/renderer_rd/shaders/particles.glsl @@ -228,6 +228,14 @@ bool emit_subparticle(mat4 p_xform, vec3 p_velocity, vec4 p_color, vec4 p_custom return true; } +vec3 safe_normalize(vec3 direction) { + const float EPSILON = 0.001; + if (length(direction) < EPSILON) { + return vec3(0.0); + } + return normalize(direction); +} + #GLOBALS void main() { @@ -431,7 +439,7 @@ void main() { switch (FRAME.attractors[i].type) { case ATTRACTOR_TYPE_SPHERE: { - dir = normalize(rel_vec); + dir = safe_normalize(rel_vec); float d = length(local_pos) / FRAME.attractors[i].extents.x; if (d > 1.0) { continue; @@ -439,7 +447,7 @@ void main() { amount = max(0.0, 1.0 - d); } break; case ATTRACTOR_TYPE_BOX: { - dir = normalize(rel_vec); + dir = safe_normalize(rel_vec); vec3 abs_pos = abs(local_pos / FRAME.attractors[i].extents); float d = max(abs_pos.x, max(abs_pos.y, abs_pos.z)); @@ -455,13 +463,13 @@ void main() { continue; } vec3 s = texture(sampler3D(sdf_vec_textures[FRAME.attractors[i].texture_index], material_samplers[SAMPLER_LINEAR_CLAMP]), uvw_pos).xyz; - dir = mat3(FRAME.attractors[i].transform) * normalize(s); //revert direction + dir = mat3(FRAME.attractors[i].transform) * safe_normalize(s); //revert direction amount = length(s); } break; } amount = pow(amount, FRAME.attractors[i].attenuation); - dir = normalize(mix(dir, FRAME.attractors[i].transform[2].xyz, FRAME.attractors[i].directionality)); + dir = safe_normalize(mix(dir, FRAME.attractors[i].transform[2].xyz, FRAME.attractors[i].directionality)); attractor_force -= amount * dir * FRAME.attractors[i].strength; } diff --git a/servers/rendering/renderer_rd/shaders/particles_copy.glsl b/servers/rendering/renderer_rd/shaders/particles_copy.glsl index b991880cd9..afbd5a9caa 100644 --- a/servers/rendering/renderer_rd/shaders/particles_copy.glsl +++ b/servers/rendering/renderer_rd/shaders/particles_copy.glsl @@ -61,6 +61,8 @@ layout(push_constant, std430) uniform Params { uint lifetime_split; bool lifetime_reverse; bool copy_mode_2d; + + mat4 inv_emission_transform; } params; @@ -199,6 +201,12 @@ void main() { txform = txform * trail_bind_poses.data[part_ofs]; } + if (params.copy_mode_2d) { + // In global mode, bring 2D particles to local coordinates + // as they will be drawn with the node position as origin. + txform = params.inv_emission_transform * txform; + } + txform = transpose(txform); } else { txform = mat4(vec4(0.0), vec4(0.0), vec4(0.0), vec4(0.0)); //zero scale, becomes invisible diff --git a/servers/rendering/renderer_rd/shaders/scene_forward_clustered.glsl b/servers/rendering/renderer_rd/shaders/scene_forward_clustered.glsl index 5d65b00bee..5947fc5351 100644 --- a/servers/rendering/renderer_rd/shaders/scene_forward_clustered.glsl +++ b/servers/rendering/renderer_rd/shaders/scene_forward_clustered.glsl @@ -83,6 +83,11 @@ layout(location = 5) out vec3 tangent_interp; layout(location = 6) out vec3 binormal_interp; #endif +#ifdef MOTION_VECTORS +layout(location = 7) out vec4 screen_position; +layout(location = 8) out vec4 prev_screen_position; +#endif + #ifdef MATERIAL_UNIFORMS_USED layout(set = MATERIAL_UNIFORM_SET, binding = 0, std140) uniform MaterialUniforms{ @@ -91,40 +96,43 @@ layout(set = MATERIAL_UNIFORM_SET, binding = 0, std140) uniform MaterialUniforms } material; #endif +float global_time; + #ifdef MODE_DUAL_PARABOLOID -layout(location = 8) out float dp_clip; +layout(location = 9) out float dp_clip; #endif -layout(location = 9) out flat uint instance_index_interp; +layout(location = 10) out flat uint instance_index_interp; + +#ifdef USE_MULTIVIEW +#ifdef has_VK_KHR_multiview +#define ViewIndex gl_ViewIndex +#else // has_VK_KHR_multiview +// !BAS! This needs to become an input once we implement our fallback! +#define ViewIndex 0 +#endif // has_VK_KHR_multiview +#else // USE_MULTIVIEW +// Set to zero, not supported in non stereo +#define ViewIndex 0 +#endif //USE_MULTIVIEW invariant gl_Position; #GLOBALS -void main() { +void vertex_shader(in uint instance_index, in bool is_multimesh, in SceneData scene_data, in mat4 model_matrix, out vec4 screen_pos) { vec4 instance_custom = vec4(0.0); #if defined(COLOR_USED) color_interp = color_attrib; #endif - uint instance_index = draw_call.instance_index; - - bool is_multimesh = bool(instances.data[instance_index].flags & INSTANCE_FLAGS_MULTIMESH); - if (!is_multimesh) { - instance_index += gl_InstanceIndex; - } - - instance_index_interp = instance_index; - - mat4 world_matrix = instances.data[instance_index].transform; - - mat3 world_normal_matrix; + mat3 model_normal_matrix; if (bool(instances.data[instance_index].flags & INSTANCE_FLAGS_NON_UNIFORM_SCALE)) { - world_normal_matrix = transpose(inverse(mat3(world_matrix))); + model_normal_matrix = transpose(inverse(mat3(model_matrix))); } else { - world_normal_matrix = mat3(world_matrix); + model_normal_matrix = mat3(model_matrix); } if (is_multimesh) { @@ -217,8 +225,8 @@ void main() { #endif //transpose matrix = transpose(matrix); - world_matrix = world_matrix * matrix; - world_normal_matrix = world_normal_matrix * mat3(matrix); + model_matrix = model_matrix * matrix; + model_normal_matrix = model_normal_matrix * mat3(matrix); } vec3 vertex = vertex_attrib; @@ -244,29 +252,35 @@ void main() { vec4 position; #endif +#ifdef USE_MULTIVIEW + mat4 projection_matrix = scene_data.projection_matrix_view[ViewIndex]; + mat4 inv_projection_matrix = scene_data.inv_projection_matrix_view[ViewIndex]; +#else mat4 projection_matrix = scene_data.projection_matrix; + mat4 inv_projection_matrix = scene_data.inv_projection_matrix; +#endif //USE_MULTIVIEW //using world coordinates #if !defined(SKIP_TRANSFORM_USED) && defined(VERTEX_WORLD_COORDS_USED) - vertex = (world_matrix * vec4(vertex, 1.0)).xyz; + vertex = (model_matrix * vec4(vertex, 1.0)).xyz; #ifdef NORMAL_USED - normal = world_normal_matrix * normal; + normal = model_normal_matrix * normal; #endif #if defined(TANGENT_USED) || defined(NORMAL_MAP_USED) || defined(LIGHT_ANISOTROPY_USED) - tangent = world_normal_matrix * tangent; - binormal = world_normal_matrix * binormal; + tangent = model_normal_matrix * tangent; + binormal = model_normal_matrix * binormal; #endif #endif float roughness = 1.0; - mat4 modelview = scene_data.inv_camera_matrix * world_matrix; - mat3 modelview_normal = mat3(scene_data.inv_camera_matrix) * world_normal_matrix; + mat4 modelview = scene_data.view_matrix * model_matrix; + mat3 modelview_normal = mat3(scene_data.view_matrix) * model_normal_matrix; { #CODE : VERTEX @@ -291,18 +305,23 @@ void main() { //using world coordinates #if !defined(SKIP_TRANSFORM_USED) && defined(VERTEX_WORLD_COORDS_USED) - vertex = (scene_data.inv_camera_matrix * vec4(vertex, 1.0)).xyz; + vertex = (scene_data.view_matrix * vec4(vertex, 1.0)).xyz; #ifdef NORMAL_USED - normal = (scene_data.inv_camera_matrix * vec4(normal, 0.0)).xyz; + normal = (scene_data.view_matrix * vec4(normal, 0.0)).xyz; #endif #if defined(TANGENT_USED) || defined(NORMAL_MAP_USED) || defined(LIGHT_ANISOTROPY_USED) - binormal = (scene_data.inv_camera_matrix * vec4(binormal, 0.0)).xyz; - tangent = (scene_data.inv_camera_matrix * vec4(tangent, 0.0)).xyz; + binormal = (scene_data.view_matrix * vec4(binormal, 0.0)).xyz; + tangent = (scene_data.view_matrix * vec4(tangent, 0.0)).xyz; #endif #endif vertex_interp = vertex; + +#ifdef MOTION_VECTORS + screen_pos = projection_matrix * vec4(vertex_interp, 1.0); +#endif + #ifdef NORMAL_USED normal_interp = normal; #endif @@ -357,6 +376,29 @@ void main() { #endif } +void main() { + uint instance_index = draw_call.instance_index; + + bool is_multimesh = bool(instances.data[instance_index].flags & INSTANCE_FLAGS_MULTIMESH); + if (!is_multimesh) { + instance_index += gl_InstanceIndex; + } + + instance_index_interp = instance_index; + + mat4 model_matrix = instances.data[instance_index].transform; +#if defined(MOTION_VECTORS) + global_time = scene_data_block.prev_data.time; + vertex_shader(instance_index, is_multimesh, scene_data_block.prev_data, instances.data[instance_index].prev_transform, prev_screen_position); + global_time = scene_data_block.data.time; + vertex_shader(instance_index, is_multimesh, scene_data_block.data, model_matrix, screen_position); +#else + global_time = scene_data_block.data.time; + vec4 screen_position; + vertex_shader(instance_index, is_multimesh, scene_data_block.data, model_matrix, screen_position); +#endif +} + #[fragment] #version 450 @@ -413,18 +455,43 @@ layout(location = 5) in vec3 tangent_interp; layout(location = 6) in vec3 binormal_interp; #endif +#ifdef MOTION_VECTORS +layout(location = 7) in vec4 screen_position; +layout(location = 8) in vec4 prev_screen_position; +#endif + #ifdef MODE_DUAL_PARABOLOID -layout(location = 8) in float dp_clip; +layout(location = 9) in float dp_clip; #endif -layout(location = 9) in flat uint instance_index_interp; +layout(location = 10) in flat uint instance_index_interp; + +#ifdef USE_MULTIVIEW +#ifdef has_VK_KHR_multiview +#define ViewIndex gl_ViewIndex +#else // has_VK_KHR_multiview +// !BAS! This needs to become an input once we implement our fallback! +#define ViewIndex 0 +#endif // has_VK_KHR_multiview +#else // USE_MULTIVIEW +// Set to zero, not supported in non stereo +#define ViewIndex 0 +#endif //USE_MULTIVIEW //defines to keep compatibility with vertex -#define world_matrix instances.data[instance_index].transform +#define model_matrix instances.data[draw_call.instance_index].transform +#ifdef USE_MULTIVIEW +#define projection_matrix scene_data.projection_matrix_view[ViewIndex] +#define inv_projection_matrix scene_data.inv_projection_matrix_view[ViewIndex] +#else #define projection_matrix scene_data.projection_matrix +#define inv_projection_matrix scene_data.inv_projection_matrix +#endif + +#define global_time scene_data_block.data.time #if defined(ENABLE_SSS) && defined(ENABLE_TRANSMITTANCE) //both required for transmittance to be enabled @@ -463,23 +530,27 @@ layout(location = 1) out uvec2 voxel_gi_buffer; #endif //MODE_RENDER_NORMAL #else // RENDER DEPTH -#ifdef MODE_MULTIPLE_RENDER_TARGETS +#ifdef MODE_SEPARATE_SPECULAR layout(location = 0) out vec4 diffuse_buffer; //diffuse (rgb) and roughness layout(location = 1) out vec4 specular_buffer; //specular and SSS (subsurface scatter) #else layout(location = 0) out vec4 frag_color; -#endif // MODE_MULTIPLE_RENDER_TARGETS +#endif // MODE_SEPARATE_SPECULAR #endif // RENDER DEPTH +#ifdef MOTION_VECTORS +layout(location = 2) out vec2 motion_vector; +#endif + #include "scene_forward_aa_inc.glsl" #if !defined(MODE_RENDER_DEPTH) && !defined(MODE_UNSHADED) -/* Make a default specular mode SPECULAR_SCHLICK_GGX. */ -#if !defined(SPECULAR_DISABLED) && !defined(SPECULAR_SCHLICK_GGX) && !defined(SPECULAR_BLINN) && !defined(SPECULAR_PHONG) && !defined(SPECULAR_TOON) +// Default to SPECULAR_SCHLICK_GGX. +#if !defined(SPECULAR_DISABLED) && !defined(SPECULAR_SCHLICK_GGX) && !defined(SPECULAR_TOON) #define SPECULAR_SCHLICK_GGX #endif @@ -492,24 +563,24 @@ layout(location = 0) out vec4 frag_color; #ifndef MODE_RENDER_DEPTH vec4 volumetric_fog_process(vec2 screen_uv, float z) { - vec3 fog_pos = vec3(screen_uv, z * scene_data.volumetric_fog_inv_length); + vec3 fog_pos = vec3(screen_uv, z * scene_data_block.data.volumetric_fog_inv_length); if (fog_pos.z < 0.0) { return vec4(0.0); } else if (fog_pos.z < 1.0) { - fog_pos.z = pow(fog_pos.z, scene_data.volumetric_fog_detail_spread); + fog_pos.z = pow(fog_pos.z, scene_data_block.data.volumetric_fog_detail_spread); } return texture(sampler3D(volumetric_fog_texture, material_samplers[SAMPLER_LINEAR_CLAMP]), fog_pos); } vec4 fog_process(vec3 vertex) { - vec3 fog_color = scene_data.fog_light_color; + vec3 fog_color = scene_data_block.data.fog_light_color; - if (scene_data.fog_aerial_perspective > 0.0) { + if (scene_data_block.data.fog_aerial_perspective > 0.0) { vec3 sky_fog_color = vec3(0.0); - vec3 cube_view = scene_data.radiance_inverse_xform * vertex; + vec3 cube_view = scene_data_block.data.radiance_inverse_xform * vertex; // mip_level always reads from the second mipmap and higher so the fog is always slightly blurred - float mip_level = mix(1.0 / MAX_ROUGHNESS_LOD, 1.0, 1.0 - (abs(vertex.z) - scene_data.z_near) / (scene_data.z_far - scene_data.z_near)); + float mip_level = mix(1.0 / MAX_ROUGHNESS_LOD, 1.0, 1.0 - (abs(vertex.z) - scene_data_block.data.z_near) / (scene_data_block.data.z_far - scene_data_block.data.z_near)); #ifdef USE_RADIANCE_CUBEMAP_ARRAY float lod, blend; blend = modf(mip_level * MAX_ROUGHNESS_LOD, lod); @@ -518,29 +589,29 @@ vec4 fog_process(vec3 vertex) { #else sky_fog_color = textureLod(samplerCube(radiance_cubemap, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), cube_view, mip_level * MAX_ROUGHNESS_LOD).rgb; #endif //USE_RADIANCE_CUBEMAP_ARRAY - fog_color = mix(fog_color, sky_fog_color, scene_data.fog_aerial_perspective); + fog_color = mix(fog_color, sky_fog_color, scene_data_block.data.fog_aerial_perspective); } - if (scene_data.fog_sun_scatter > 0.001) { + if (scene_data_block.data.fog_sun_scatter > 0.001) { vec4 sun_scatter = vec4(0.0); float sun_total = 0.0; vec3 view = normalize(vertex); - for (uint i = 0; i < scene_data.directional_light_count; i++) { + for (uint i = 0; i < scene_data_block.data.directional_light_count; i++) { vec3 light_color = directional_lights.data[i].color * directional_lights.data[i].energy; float light_amount = pow(max(dot(view, directional_lights.data[i].direction), 0.0), 8.0); - fog_color += light_color * light_amount * scene_data.fog_sun_scatter; + fog_color += light_color * light_amount * scene_data_block.data.fog_sun_scatter; } } - float fog_amount = 1.0 - exp(min(0.0, -length(vertex) * scene_data.fog_density)); + float fog_amount = 1.0 - exp(min(0.0, -length(vertex) * scene_data_block.data.fog_density)); - if (abs(scene_data.fog_height_density) >= 0.0001) { - float y = (scene_data.camera_matrix * vec4(vertex, 1.0)).y; + if (abs(scene_data_block.data.fog_height_density) >= 0.0001) { + float y = (scene_data_block.data.inv_view_matrix * vec4(vertex, 1.0)).y; - float y_dist = y - scene_data.fog_height; + float y_dist = y - scene_data_block.data.fog_height; - float vfog_amount = 1.0 - exp(min(0.0, y_dist * scene_data.fog_height_density)); + float vfog_amount = 1.0 - exp(min(0.0, y_dist * scene_data_block.data.fog_height_density)); fog_amount = max(vfog_amount, fog_amount); } @@ -565,18 +636,16 @@ uint cluster_get_range_clip_mask(uint i, uint z_min, uint z_max) { #endif //!MODE_RENDER DEPTH -void main() { -#ifdef MODE_DUAL_PARABOLOID - - if (dp_clip > 0.0) - discard; -#endif - +void fragment_shader(in SceneData scene_data) { uint instance_index = instance_index_interp; //lay out everything, whatever is unused is optimized away anyway vec3 vertex = vertex_interp; +#ifdef USE_MULTIVIEW + vec3 view = -normalize(vertex_interp - scene_data.eye_offset[ViewIndex].xyz); +#else vec3 view = -normalize(vertex_interp); +#endif vec3 albedo = vec3(1.0); vec3 backlight = vec3(0.0); vec4 transmittance_color = vec4(0.0, 0.0, 0.0, 1.0); @@ -589,7 +658,7 @@ void main() { float rim = 0.0; float rim_tint = 0.0; float clearcoat = 0.0; - float clearcoat_gloss = 0.0; + float clearcoat_roughness = 0.0; float anisotropy = 0.0; vec2 anisotropy_flow = vec2(1.0, 0.0); vec4 fog = vec4(0.0); @@ -643,7 +712,7 @@ void main() { float normal_map_depth = 1.0; - vec2 screen_uv = gl_FragCoord.xy * scene_data.screen_pixel_size + scene_data.screen_pixel_size * 0.5; //account for center + vec2 screen_uv = gl_FragCoord.xy * scene_data.screen_pixel_size; float sss_strength = 0.0; @@ -912,7 +981,17 @@ void main() { #if !defined(MODE_RENDER_DEPTH) && !defined(MODE_UNSHADED) if (scene_data.use_reflection_cubemap) { +#ifdef LIGHT_ANISOTROPY_USED + // https://google.github.io/filament/Filament.html#lighting/imagebasedlights/anisotropy + vec3 anisotropic_direction = anisotropy >= 0.0 ? binormal : tangent; + vec3 anisotropic_tangent = cross(anisotropic_direction, view); + vec3 anisotropic_normal = cross(anisotropic_tangent, anisotropic_direction); + vec3 bent_normal = normalize(mix(normal, anisotropic_normal, abs(anisotropy) * clamp(5.0 * roughness, 0.0, 1.0))); + vec3 ref_vec = reflect(-view, bent_normal); +#else vec3 ref_vec = reflect(-view, normal); +#endif + float horizon = min(1.0 + dot(ref_vec, normal), 1.0); ref_vec = scene_data.radiance_inverse_xform * ref_vec; #ifdef USE_RADIANCE_CUBEMAP_ARRAY @@ -954,6 +1033,36 @@ void main() { #if defined(CUSTOM_IRRADIANCE_USED) ambient_light = mix(ambient_light, custom_irradiance.rgb, custom_irradiance.a); #endif + +#ifdef LIGHT_CLEARCOAT_USED + + if (scene_data.use_reflection_cubemap) { + vec3 n = normalize(normal_interp); // We want to use geometric normal, not normal_map + float NoV = max(dot(n, view), 0.0001); + vec3 ref_vec = reflect(-view, n); + // The clear coat layer assumes an IOR of 1.5 (4% reflectance) + float Fc = clearcoat * (0.04 + 0.96 * SchlickFresnel(NoV)); + float attenuation = 1.0 - Fc; + ambient_light *= attenuation; + specular_light *= attenuation; + + float horizon = min(1.0 + dot(ref_vec, normal), 1.0); + ref_vec = scene_data.radiance_inverse_xform * ref_vec; + float roughness_lod = mix(0.001, 0.1, clearcoat_roughness) * MAX_ROUGHNESS_LOD; +#ifdef USE_RADIANCE_CUBEMAP_ARRAY + + float lod, blend; + blend = modf(roughness_lod, lod); + vec3 clearcoat_light = texture(samplerCubeArray(radiance_cubemap, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), vec4(ref_vec, lod)).rgb; + clearcoat_light = mix(clearcoat_light, texture(samplerCubeArray(radiance_cubemap, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), vec4(ref_vec, lod + 1)).rgb, blend); + +#else + vec3 clearcoat_light = textureLod(samplerCube(radiance_cubemap, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), ref_vec, roughness_lod).rgb; + +#endif //USE_RADIANCE_CUBEMAP_ARRAY + specular_light += clearcoat_light * horizon * horizon * Fc * scene_data.ambient_light_color_energy.a; + } +#endif #endif //!defined(MODE_RENDER_DEPTH) && !defined(MODE_UNSHADED) //radiance @@ -967,7 +1076,7 @@ void main() { if (bool(instances.data[instance_index].flags & INSTANCE_FLAGS_USE_LIGHTMAP_CAPTURE)) { //has lightmap capture uint index = instances.data[instance_index].gi_offset; - vec3 wnormal = mat3(scene_data.camera_matrix) * normal; + vec3 wnormal = mat3(scene_data.inv_view_matrix) * normal; const float c1 = 0.429043; const float c2 = 0.511664; const float c3 = 0.743125; @@ -1021,9 +1130,9 @@ void main() { if (sc_use_forward_gi && bool(instances.data[instance_index].flags & INSTANCE_FLAGS_USE_SDFGI)) { //has lightmap capture //make vertex orientation the world one, but still align to camera - vec3 cam_pos = mat3(scene_data.camera_matrix) * vertex; - vec3 cam_normal = mat3(scene_data.camera_matrix) * normal; - vec3 cam_reflection = mat3(scene_data.camera_matrix) * reflect(-view, normal); + vec3 cam_pos = mat3(scene_data.inv_view_matrix) * vertex; + vec3 cam_normal = mat3(scene_data.inv_view_matrix) * normal; + vec3 cam_reflection = mat3(scene_data.inv_view_matrix) * reflect(-view, normal); //apply y-mult cam_pos.y *= sdfgi.y_mult; @@ -1093,7 +1202,7 @@ void main() { if (sc_use_forward_gi && bool(instances.data[instance_index].flags & INSTANCE_FLAGS_USE_VOXEL_GI)) { // process voxel_gi_instances uint index1 = instances.data[instance_index].gi_offset & 0xFFFF; - vec3 ref_vec = normalize(reflect(normalize(vertex), normal)); + vec3 ref_vec = normalize(reflect(-view, normal)); //find arbitrary tangent and bitangent, then build a matrix vec3 v0 = abs(normal.z) < 0.999 ? vec3(0.0, 0.0, 1.0) : vec3(0.0, 1.0, 0.0); vec3 tangent = normalize(cross(v0, normal)); @@ -1129,12 +1238,20 @@ void main() { if (scene_data.gi_upscale_for_msaa) { vec2 base_coord = screen_uv; vec2 closest_coord = base_coord; +#ifdef USE_MULTIVIEW + float closest_ang = dot(normal, textureLod(sampler2DArray(normal_roughness_buffer, material_samplers[SAMPLER_LINEAR_CLAMP]), vec3(base_coord, ViewIndex), 0.0).xyz * 2.0 - 1.0); +#else // USE_MULTIVIEW float closest_ang = dot(normal, textureLod(sampler2D(normal_roughness_buffer, material_samplers[SAMPLER_LINEAR_CLAMP]), base_coord, 0.0).xyz * 2.0 - 1.0); +#endif // USE_MULTIVIEW for (int i = 0; i < 4; i++) { const vec2 neighbours[4] = vec2[](vec2(-1, 0), vec2(1, 0), vec2(0, -1), vec2(0, 1)); vec2 neighbour_coord = base_coord + neighbours[i] * scene_data.screen_pixel_size; +#ifdef USE_MULTIVIEW + float neighbour_ang = dot(normal, textureLod(sampler2DArray(normal_roughness_buffer, material_samplers[SAMPLER_LINEAR_CLAMP]), vec3(neighbour_coord, ViewIndex), 0.0).xyz * 2.0 - 1.0); +#else // USE_MULTIVIEW float neighbour_ang = dot(normal, textureLod(sampler2D(normal_roughness_buffer, material_samplers[SAMPLER_LINEAR_CLAMP]), neighbour_coord, 0.0).xyz * 2.0 - 1.0); +#endif // USE_MULTIVIEW if (neighbour_ang > closest_ang) { closest_ang = neighbour_ang; closest_coord = neighbour_coord; @@ -1147,8 +1264,13 @@ void main() { coord = screen_uv; } +#ifdef USE_MULTIVIEW + vec4 buffer_ambient = textureLod(sampler2DArray(ambient_buffer, material_samplers[SAMPLER_LINEAR_CLAMP]), vec3(coord, ViewIndex), 0.0); + vec4 buffer_reflection = textureLod(sampler2DArray(reflection_buffer, material_samplers[SAMPLER_LINEAR_CLAMP]), vec3(coord, ViewIndex), 0.0); +#else // USE_MULTIVIEW vec4 buffer_ambient = textureLod(sampler2D(ambient_buffer, material_samplers[SAMPLER_LINEAR_CLAMP]), coord, 0.0); vec4 buffer_reflection = textureLod(sampler2D(reflection_buffer, material_samplers[SAMPLER_LINEAR_CLAMP]), coord, 0.0); +#endif // USE_MULTIVIEW ambient_light = mix(ambient_light, buffer_ambient.rgb, buffer_ambient.a); specular_light = mix(specular_light, buffer_reflection.rgb, buffer_reflection.a); @@ -1202,8 +1324,16 @@ void main() { if (!bool(reflections.data[reflection_index].mask & instances.data[instance_index].layer_mask)) { continue; //not masked } - - reflection_process(reflection_index, vertex, normal, roughness, ambient_light, specular_light, ambient_accum, reflection_accum); +#ifdef LIGHT_ANISOTROPY_USED + // https://google.github.io/filament/Filament.html#lighting/imagebasedlights/anisotropy + vec3 anisotropic_direction = anisotropy >= 0.0 ? binormal : tangent; + vec3 anisotropic_tangent = cross(anisotropic_direction, view); + vec3 anisotropic_normal = cross(anisotropic_tangent, anisotropic_direction); + vec3 bent_normal = normalize(mix(normal, anisotropic_normal, abs(anisotropy) * clamp(5.0 * roughness, 0.0, 1.0))); +#else + vec3 bent_normal = normal; +#endif + reflection_process(reflection_index, view, vertex, bent_normal, roughness, ambient_light, specular_light, ambient_accum, reflection_accum); } } @@ -1395,62 +1525,78 @@ void main() { } else { //no soft shadows vec4 pssm_coord; + float blur_factor; + if (depth_z < directional_lights.data[i].shadow_split_offsets.x) { vec4 v = vec4(vertex, 1.0); BIAS_FUNC(v, 0) pssm_coord = (directional_lights.data[i].shadow_matrix1 * v); + blur_factor = 1.0; } else if (depth_z < directional_lights.data[i].shadow_split_offsets.y) { vec4 v = vec4(vertex, 1.0); BIAS_FUNC(v, 1) pssm_coord = (directional_lights.data[i].shadow_matrix2 * v); + // Adjust shadow blur with reference to the first split to reduce discrepancy between shadow splits. + blur_factor = directional_lights.data[i].shadow_split_offsets.x / directional_lights.data[i].shadow_split_offsets.y; } else if (depth_z < directional_lights.data[i].shadow_split_offsets.z) { vec4 v = vec4(vertex, 1.0); BIAS_FUNC(v, 2) pssm_coord = (directional_lights.data[i].shadow_matrix3 * v); - + // Adjust shadow blur with reference to the first split to reduce discrepancy between shadow splits. + blur_factor = directional_lights.data[i].shadow_split_offsets.x / directional_lights.data[i].shadow_split_offsets.z; } else { vec4 v = vec4(vertex, 1.0); BIAS_FUNC(v, 3) pssm_coord = (directional_lights.data[i].shadow_matrix4 * v); + // Adjust shadow blur with reference to the first split to reduce discrepancy between shadow splits. + blur_factor = directional_lights.data[i].shadow_split_offsets.x / directional_lights.data[i].shadow_split_offsets.w; } pssm_coord /= pssm_coord.w; - shadow = sample_directional_pcf_shadow(directional_shadow_atlas, scene_data.directional_shadow_pixel_size * directional_lights.data[i].soft_shadow_scale, pssm_coord); + shadow = sample_directional_pcf_shadow(directional_shadow_atlas, scene_data.directional_shadow_pixel_size * directional_lights.data[i].soft_shadow_scale * blur_factor, pssm_coord); if (directional_lights.data[i].blend_splits) { float pssm_blend; + float blur_factor2; if (depth_z < directional_lights.data[i].shadow_split_offsets.x) { vec4 v = vec4(vertex, 1.0); BIAS_FUNC(v, 1) pssm_coord = (directional_lights.data[i].shadow_matrix2 * v); pssm_blend = smoothstep(0.0, directional_lights.data[i].shadow_split_offsets.x, depth_z); + // Adjust shadow blur with reference to the first split to reduce discrepancy between shadow splits. + blur_factor2 = directional_lights.data[i].shadow_split_offsets.x / directional_lights.data[i].shadow_split_offsets.y; } else if (depth_z < directional_lights.data[i].shadow_split_offsets.y) { vec4 v = vec4(vertex, 1.0); BIAS_FUNC(v, 2) pssm_coord = (directional_lights.data[i].shadow_matrix3 * v); pssm_blend = smoothstep(directional_lights.data[i].shadow_split_offsets.x, directional_lights.data[i].shadow_split_offsets.y, depth_z); + // Adjust shadow blur with reference to the first split to reduce discrepancy between shadow splits. + blur_factor2 = directional_lights.data[i].shadow_split_offsets.x / directional_lights.data[i].shadow_split_offsets.z; } else if (depth_z < directional_lights.data[i].shadow_split_offsets.z) { vec4 v = vec4(vertex, 1.0); BIAS_FUNC(v, 3) pssm_coord = (directional_lights.data[i].shadow_matrix4 * v); pssm_blend = smoothstep(directional_lights.data[i].shadow_split_offsets.y, directional_lights.data[i].shadow_split_offsets.z, depth_z); + // Adjust shadow blur with reference to the first split to reduce discrepancy between shadow splits. + blur_factor2 = directional_lights.data[i].shadow_split_offsets.x / directional_lights.data[i].shadow_split_offsets.w; } else { pssm_blend = 0.0; //if no blend, same coord will be used (divide by z will result in same value, and already cached) + blur_factor2 = 1.0; } pssm_coord /= pssm_coord.w; - float shadow2 = sample_directional_pcf_shadow(directional_shadow_atlas, scene_data.directional_shadow_pixel_size * directional_lights.data[i].soft_shadow_scale, pssm_coord); + float shadow2 = sample_directional_pcf_shadow(directional_shadow_atlas, scene_data.directional_shadow_pixel_size * directional_lights.data[i].soft_shadow_scale * blur_factor2, pssm_coord); shadow = mix(shadow, shadow2, pssm_blend); } } @@ -1555,10 +1701,11 @@ void main() { rim, rim_tint, #endif #ifdef LIGHT_CLEARCOAT_USED - clearcoat, clearcoat_gloss, + clearcoat, clearcoat_roughness, normalize(normal_interp), #endif #ifdef LIGHT_ANISOTROPY_USED - binormal, tangent, anisotropy, + binormal, + tangent, anisotropy, #endif diffuse_light, specular_light); @@ -1626,7 +1773,7 @@ void main() { rim_tint, #endif #ifdef LIGHT_CLEARCOAT_USED - clearcoat, clearcoat_gloss, + clearcoat, clearcoat_roughness, normalize(normal_interp), #endif #ifdef LIGHT_ANISOTROPY_USED tangent, binormal, anisotropy, @@ -1698,10 +1845,11 @@ void main() { rim_tint, #endif #ifdef LIGHT_CLEARCOAT_USED - clearcoat, clearcoat_gloss, + clearcoat, clearcoat_roughness, normalize(normal_interp), #endif #ifdef LIGHT_ANISOTROPY_USED - tangent, binormal, anisotropy, + tangent, + binormal, anisotropy, #endif diffuse_light, specular_light); } @@ -1753,7 +1901,7 @@ void main() { vec3(0, -1, 0), vec3(0, 0, -1)); - vec3 cam_normal = mat3(scene_data.camera_matrix) * normalize(normal_interp); + vec3 cam_normal = mat3(scene_data.inv_view_matrix) * normalize(normal_interp); float closest_dist = -1e20; @@ -1880,7 +2028,7 @@ void main() { //restore fog fog = vec4(unpackHalf2x16(fog_rg), unpackHalf2x16(fog_ba)); -#ifdef MODE_MULTIPLE_RENDER_TARGETS +#ifdef MODE_SEPARATE_SPECULAR #ifdef MODE_UNSHADED diffuse_buffer = vec4(albedo.rgb, 0.0); @@ -1898,19 +2046,38 @@ void main() { diffuse_buffer.rgb = mix(diffuse_buffer.rgb, fog.rgb, fog.a); specular_buffer.rgb = mix(specular_buffer.rgb, vec3(0.0), fog.a); -#else //MODE_MULTIPLE_RENDER_TARGETS +#else //MODE_SEPARATE_SPECULAR #ifdef MODE_UNSHADED frag_color = vec4(albedo, alpha); #else frag_color = vec4(emission + ambient_light + diffuse_light + specular_light, alpha); - //frag_color = vec4(1.0); +//frag_color = vec4(1.0); #endif //USE_NO_SHADING // Draw "fixed" fog before volumetric fog to ensure volumetric fog can appear in front of the sky. frag_color.rgb = mix(frag_color.rgb, fog.rgb, fog.a); -#endif //MODE_MULTIPLE_RENDER_TARGETS +#endif //MODE_SEPARATE_SPECULAR #endif //MODE_RENDER_DEPTH +#ifdef MOTION_VECTORS + vec2 position_clip = (screen_position.xy / screen_position.w) - scene_data.taa_jitter; + vec2 prev_position_clip = (prev_screen_position.xy / prev_screen_position.w) - scene_data_block.prev_data.taa_jitter; + + vec2 position_uv = position_clip * vec2(0.5, 0.5); + vec2 prev_position_uv = prev_position_clip * vec2(0.5, 0.5); + + motion_vector = position_uv - prev_position_uv; +#endif +} + +void main() { +#ifdef MODE_DUAL_PARABOLOID + + if (dp_clip > 0.0) + discard; +#endif + + fragment_shader(scene_data_block.data); } diff --git a/servers/rendering/renderer_rd/shaders/scene_forward_clustered_inc.glsl b/servers/rendering/renderer_rd/shaders/scene_forward_clustered_inc.glsl index 084e2a0673..0c23de96c3 100644 --- a/servers/rendering/renderer_rd/shaders/scene_forward_clustered_inc.glsl +++ b/servers/rendering/renderer_rd/shaders/scene_forward_clustered_inc.glsl @@ -2,6 +2,7 @@ #define ROUGHNESS_MAX_LOD 5 #define MAX_VOXEL_GI_INSTANCES 8 +#define MAX_VIEWS 2 #if defined(has_GL_KHR_shader_subgroup_ballot) && defined(has_GL_KHR_shader_subgroup_arithmetic) @@ -12,10 +13,14 @@ #endif +#if defined(USE_MULTIVIEW) && defined(has_VK_KHR_multiview) +#extension GL_EXT_multiview : enable +#endif + #include "cluster_data_inc.glsl" #include "decal_data_inc.glsl" -#if !defined(MODE_RENDER_DEPTH) || defined(MODE_RENDER_MATERIAL) || defined(MODE_RENDER_SDF) || defined(MODE_RENDER_NORMAL_ROUGHNESS) || defined(MODE_RENDER_VOXEL_GI) || defined(TANGENT_USED) || defined(NORMAL_MAP_USED) +#if !defined(MODE_RENDER_DEPTH) || defined(MODE_RENDER_MATERIAL) || defined(MODE_RENDER_SDF) || defined(MODE_RENDER_NORMAL_ROUGHNESS) || defined(MODE_RENDER_VOXEL_GI) || defined(TANGENT_USED) || defined(NORMAL_MAP_USED) || defined(LIGHT_ANISOTROPY_USED) #ifndef NORMAL_USED #define NORMAL_USED #endif @@ -166,12 +171,16 @@ sdfgi; /* Set 1: Render Pass (changes per render pass) */ -layout(set = 1, binding = 0, std140) uniform SceneData { +struct SceneData { mat4 projection_matrix; mat4 inv_projection_matrix; + mat4 inv_view_matrix; + mat4 view_matrix; - mat4 camera_matrix; - mat4 inv_camera_matrix; + // only used for multiview + mat4 projection_matrix_view[MAX_VIEWS]; + mat4 inv_projection_matrix_view[MAX_VIEWS]; + vec4 eye_offset[MAX_VIEWS]; vec2 viewport_size; vec2 screen_pixel_size; @@ -241,11 +250,19 @@ layout(set = 1, binding = 0, std140) uniform SceneData { float reflection_multiplier; // one normally, zero when rendering reflections bool pancake_shadows; + vec2 taa_jitter; + uvec2 pad2; +}; + +layout(set = 1, binding = 0, std140) uniform SceneDataBlock { + SceneData data; + SceneData prev_data; } -scene_data; +scene_data_block; struct InstanceData { mat4 transform; + mat4 prev_transform; uint flags; uint instance_uniforms_ofs; //base offset in global buffer for instance variables uint gi_offset; //GI information when using lightmapping (VCT or lightmap index) @@ -300,10 +317,16 @@ layout(r32ui, set = 1, binding = 12) uniform restrict uimage3D geom_facing_grid; layout(set = 1, binding = 9) uniform texture2D depth_buffer; layout(set = 1, binding = 10) uniform texture2D color_buffer; +#ifdef USE_MULTIVIEW +layout(set = 1, binding = 11) uniform texture2DArray normal_roughness_buffer; +layout(set = 1, binding = 13) uniform texture2DArray ambient_buffer; +layout(set = 1, binding = 14) uniform texture2DArray reflection_buffer; +#else // USE_MULTIVIEW layout(set = 1, binding = 11) uniform texture2D normal_roughness_buffer; -layout(set = 1, binding = 12) uniform texture2D ao_buffer; layout(set = 1, binding = 13) uniform texture2D ambient_buffer; layout(set = 1, binding = 14) uniform texture2D reflection_buffer; +#endif +layout(set = 1, binding = 12) uniform texture2D ao_buffer; layout(set = 1, binding = 15) uniform texture2DArray sdfgi_lightprobe_texture; layout(set = 1, binding = 16) uniform texture3D sdfgi_occlusion_cascades; diff --git a/servers/rendering/renderer_rd/shaders/scene_forward_lights_inc.glsl b/servers/rendering/renderer_rd/shaders/scene_forward_lights_inc.glsl index 16f77fb91a..c92b29b14a 100644 --- a/servers/rendering/renderer_rd/shaders/scene_forward_lights_inc.glsl +++ b/servers/rendering/renderer_rd/shaders/scene_forward_lights_inc.glsl @@ -1,55 +1,29 @@ // Functions related to lighting -// This returns the G_GGX function divided by 2 cos_theta_m, where in practice cos_theta_m is either N.L or N.V. -// We're dividing this factor off because the overall term we'll end up looks like -// (see, for example, the first unnumbered equation in B. Burley, "Physically Based Shading at Disney", SIGGRAPH 2012): -// -// F(L.V) D(N.H) G(N.L) G(N.V) / (4 N.L N.V) -// -// We're basically regouping this as -// -// F(L.V) D(N.H) [G(N.L)/(2 N.L)] [G(N.V) / (2 N.V)] -// -// and thus, this function implements the [G(N.m)/(2 N.m)] part with m = L or V. -// -// The contents of the D and G (G1) functions (GGX) are taken from -// E. Heitz, "Understanding the Masking-Shadowing Function in Microfacet-Based BRDFs", J. Comp. Graph. Tech. 3 (2) (2014). -// Eqns 71-72 and 85-86 (see also Eqns 43 and 80). - -float G_GGX_2cos(float cos_theta_m, float alpha) { - // Schlick's approximation - // C. Schlick, "An Inexpensive BRDF Model for Physically-based Rendering", Computer Graphics Forum. 13 (3): 233 (1994) - // Eq. (19), although see Heitz (2014) the about the problems with his derivation. - // It nevertheless approximates GGX well with k = alpha/2. - float k = 0.5 * alpha; - return 0.5 / (cos_theta_m * (1.0 - k) + k); - - // float cos2 = cos_theta_m * cos_theta_m; - // float sin2 = (1.0 - cos2); - // return 1.0 / (cos_theta_m + sqrt(cos2 + alpha * alpha * sin2)); -} - float D_GGX(float cos_theta_m, float alpha) { - float alpha2 = alpha * alpha; - float d = 1.0 + (alpha2 - 1.0) * cos_theta_m * cos_theta_m; - return alpha2 / (M_PI * d * d); + float a = cos_theta_m * alpha; + float k = alpha / (1.0 - cos_theta_m * cos_theta_m + a * a); + return k * k * (1.0 / M_PI); } -float G_GGX_anisotropic_2cos(float cos_theta_m, float alpha_x, float alpha_y, float cos_phi, float sin_phi) { - float cos2 = cos_theta_m * cos_theta_m; - float sin2 = (1.0 - cos2); - float s_x = alpha_x * cos_phi; - float s_y = alpha_y * sin_phi; - return 1.0 / max(cos_theta_m + sqrt(cos2 + (s_x * s_x + s_y * s_y) * sin2), 0.001); +// From Earl Hammon, Jr. "PBR Diffuse Lighting for GGX+Smith Microsurfaces" https://www.gdcvault.com/play/1024478/PBR-Diffuse-Lighting-for-GGX +float V_GGX(float NdotL, float NdotV, float alpha) { + return 0.5 / mix(2.0 * NdotL * NdotV, NdotL + NdotV, alpha); } float D_GGX_anisotropic(float cos_theta_m, float alpha_x, float alpha_y, float cos_phi, float sin_phi) { - float cos2 = cos_theta_m * cos_theta_m; - float sin2 = (1.0 - cos2); - float r_x = cos_phi / alpha_x; - float r_y = sin_phi / alpha_y; - float d = cos2 + sin2 * (r_x * r_x + r_y * r_y); - return 1.0 / max(M_PI * alpha_x * alpha_y * d * d, 0.001); + float alpha2 = alpha_x * alpha_y; + highp vec3 v = vec3(alpha_y * cos_phi, alpha_x * sin_phi, alpha2 * cos_theta_m); + highp float v2 = dot(v, v); + float w2 = alpha2 / v2; + float D = alpha2 * w2 * w2 * (1.0 / M_PI); + return D; +} + +float V_GGX_anisotropic(float alpha_x, float alpha_y, float TdotV, float TdotL, float BdotV, float BdotL, float NdotV, float NdotL) { + float Lambda_V = NdotL * length(vec3(alpha_x * TdotV, alpha_y * BdotV, NdotV)); + float Lambda_L = NdotV * length(vec3(alpha_x * TdotL, alpha_y * BdotL, NdotL)); + return 0.5 / (Lambda_V + Lambda_L); } float SchlickFresnel(float u) { @@ -58,14 +32,6 @@ float SchlickFresnel(float u) { return m2 * m2 * m; // pow(m,5) } -float GTR1(float NdotH, float a) { - if (a >= 1.0) - return 1.0 / M_PI; - float a2 = a * a; - float t = 1.0 + (a2 - 1.0) * NdotH * NdotH; - return (a2 - 1.0) / (M_PI * log(a2) * t); -} - vec3 F0(float metallic, float specular, vec3 albedo) { float dielectric = 0.16 * specular * specular; // use albedo * metallic as colored specular reflectance at 0 angle for metallic materials; @@ -87,7 +53,7 @@ void light_compute(vec3 N, vec3 L, vec3 V, float A, vec3 light_color, float atte float rim, float rim_tint, #endif #ifdef LIGHT_CLEARCOAT_USED - float clearcoat, float clearcoat_gloss, + float clearcoat, float clearcoat_roughness, vec3 vertex_normal, #endif #ifdef LIGHT_ANISOTROPY_USED vec3 B, vec3 T, float anisotropy, @@ -113,13 +79,13 @@ void light_compute(vec3 N, vec3 L, vec3 V, float A, vec3 light_color, float atte float NdotL = min(A + dot(N, L), 1.0); float cNdotL = max(NdotL, 0.0); // clamped NdotL float NdotV = dot(N, V); - float cNdotV = max(NdotV, 0.0); + float cNdotV = max(NdotV, 1e-4); -#if defined(DIFFUSE_BURLEY) || defined(SPECULAR_BLINN) || defined(SPECULAR_SCHLICK_GGX) || defined(LIGHT_CLEARCOAT_USED) +#if defined(DIFFUSE_BURLEY) || defined(SPECULAR_SCHLICK_GGX) || defined(LIGHT_CLEARCOAT_USED) vec3 H = normalize(V + L); #endif -#if defined(SPECULAR_BLINN) || defined(SPECULAR_SCHLICK_GGX) || defined(LIGHT_CLEARCOAT_USED) +#if defined(SPECULAR_SCHLICK_GGX) float cNdotH = clamp(A + dot(N, H), 0.0, 1.0); #endif @@ -203,26 +169,7 @@ void light_compute(vec3 N, vec3 L, vec3 V, float A, vec3 light_color, float atte // D -#if defined(SPECULAR_BLINN) - - //normalized blinn - float shininess = exp2(15.0 * (1.0 - roughness) + 1.0) * 0.25; - float blinn = pow(cNdotH, shininess); - blinn *= (shininess + 2.0) * (1.0 / (8.0 * M_PI)); - - specular_light += light_color * attenuation * specular_amount * blinn * f0 * orms_unpacked.w; - -#elif defined(SPECULAR_PHONG) - - vec3 R = normalize(-reflect(L, N)); - float cRdotV = clamp(A + dot(R, V), 0.0, 1.0); - float shininess = exp2(15.0 * (1.0 - roughness) + 1.0) * 0.25; - float phong = pow(cRdotV, shininess); - phong *= (shininess + 1.0) * (1.0 / (8.0 * M_PI)); - - specular_light += light_color * attenuation * specular_amount * phong * f0 * orms_unpacked.w; - -#elif defined(SPECULAR_TOON) +#if defined(SPECULAR_TOON) vec3 R = normalize(-reflect(L, N)); float RdotV = dot(R, V); @@ -236,24 +183,21 @@ void light_compute(vec3 N, vec3 L, vec3 V, float A, vec3 light_color, float atte #elif defined(SPECULAR_SCHLICK_GGX) // shlick+ggx as default - + float alpha_ggx = roughness * roughness; #if defined(LIGHT_ANISOTROPY_USED) - float alpha_ggx = roughness * roughness; float aspect = sqrt(1.0 - anisotropy * 0.9); float ax = alpha_ggx / aspect; float ay = alpha_ggx * aspect; float XdotH = dot(T, H); float YdotH = dot(B, H); float D = D_GGX_anisotropic(cNdotH, ax, ay, XdotH, YdotH); - float G = G_GGX_anisotropic_2cos(cNdotL, ax, ay, XdotH, YdotH) * G_GGX_anisotropic_2cos(cNdotV, ax, ay, XdotH, YdotH); - -#else - float alpha_ggx = roughness * roughness; + float G = V_GGX_anisotropic(ax, ay, dot(T, V), dot(T, L), dot(B, V), dot(B, L), cNdotV, cNdotL); +#else // LIGHT_ANISOTROPY_USED float D = D_GGX(cNdotH, alpha_ggx); - float G = G_GGX_2cos(cNdotL, alpha_ggx) * G_GGX_2cos(cNdotV, alpha_ggx); -#endif - // F + float G = V_GGX(cNdotL, cNdotV, alpha_ggx); +#endif // LIGHT_ANISOTROPY_USED + // F float cLdotH5 = SchlickFresnel(cLdotH); vec3 F = mix(vec3(cLdotH5), vec3(1.0), f0); @@ -263,18 +207,23 @@ void light_compute(vec3 N, vec3 L, vec3 V, float A, vec3 light_color, float atte #endif #if defined(LIGHT_CLEARCOAT_USED) + // Clearcoat ignores normal_map, use vertex normal instead + float ccNdotL = max(min(A + dot(vertex_normal, L), 1.0), 0.0); + float ccNdotH = clamp(A + dot(vertex_normal, H), 0.0, 1.0); + float ccNdotV = max(dot(vertex_normal, V), 1e-4); #if !defined(SPECULAR_SCHLICK_GGX) float cLdotH5 = SchlickFresnel(cLdotH); #endif - float Dr = GTR1(cNdotH, mix(.1, .001, clearcoat_gloss)); + float Dr = D_GGX(ccNdotH, mix(0.001, 0.1, clearcoat_roughness)); + float Gr = 0.25 / (cLdotH * cLdotH); float Fr = mix(.04, 1.0, cLdotH5); - float Gr = G_GGX_2cos(cNdotL, .25) * G_GGX_2cos(cNdotV, .25); - - float clearcoat_specular_brdf_NL = 0.25 * clearcoat * Gr * Fr * Dr * cNdotL; + float clearcoat_specular_brdf_NL = clearcoat * Gr * Fr * Dr * cNdotL; specular_light += clearcoat_specular_brdf_NL * light_color * attenuation * specular_amount; -#endif + // TODO: Clearcoat adds light to the scene right now (it is non-energy conserving), both diffuse and specular need to be scaled by (1.0 - FR) + // but to do so we need to rearrange this entire function +#endif // LIGHT_CLEARCOAT_USED } #ifdef USE_SHADOW_TO_OPACITY @@ -313,7 +262,7 @@ float sample_directional_pcf_shadow(texture2D shadow, vec2 shadow_pixel_size, ve float avg = 0.0; for (uint i = 0; i < sc_directional_soft_shadow_samples; i++) { - avg += textureProj(sampler2DShadow(shadow, shadow_sampler), vec4(pos + shadow_pixel_size * (disk_rotation * scene_data.directional_soft_shadow_kernel[i].xy), depth, 1.0)); + avg += textureProj(sampler2DShadow(shadow, shadow_sampler), vec4(pos + shadow_pixel_size * (disk_rotation * scene_data_block.data.directional_soft_shadow_kernel[i].xy), depth, 1.0)); } return avg * (1.0 / float(sc_directional_soft_shadow_samples)); @@ -339,7 +288,7 @@ float sample_pcf_shadow(texture2D shadow, vec2 shadow_pixel_size, vec3 coord) { float avg = 0.0; for (uint i = 0; i < sc_soft_shadow_samples; i++) { - avg += textureProj(sampler2DShadow(shadow, shadow_sampler), vec4(pos + shadow_pixel_size * (disk_rotation * scene_data.soft_shadow_kernel[i].xy), depth, 1.0)); + avg += textureProj(sampler2DShadow(shadow, shadow_sampler), vec4(pos + shadow_pixel_size * (disk_rotation * scene_data_block.data.soft_shadow_kernel[i].xy), depth, 1.0)); } return avg * (1.0 / float(sc_soft_shadow_samples)); @@ -362,10 +311,10 @@ float sample_omni_pcf_shadow(texture2D shadow, float blur_scale, vec2 coord, vec } float avg = 0.0; - vec2 offset_scale = blur_scale * 2.0 * scene_data.shadow_atlas_pixel_size / uv_rect.zw; + vec2 offset_scale = blur_scale * 2.0 * scene_data_block.data.shadow_atlas_pixel_size / uv_rect.zw; for (uint i = 0; i < sc_soft_shadow_samples; i++) { - vec2 offset = offset_scale * (disk_rotation * scene_data.soft_shadow_kernel[i].xy); + vec2 offset = offset_scale * (disk_rotation * scene_data_block.data.soft_shadow_kernel[i].xy); vec2 sample_coord = coord + offset; float sample_coord_length_sqaured = dot(sample_coord, sample_coord); @@ -402,7 +351,7 @@ float sample_directional_soft_shadow(texture2D shadow, vec3 pssm_coord, vec2 tex } for (uint i = 0; i < sc_directional_penumbra_shadow_samples; i++) { - vec2 suv = pssm_coord.xy + (disk_rotation * scene_data.directional_penumbra_shadow_kernel[i].xy) * tex_scale; + vec2 suv = pssm_coord.xy + (disk_rotation * scene_data_block.data.directional_penumbra_shadow_kernel[i].xy) * tex_scale; float d = textureLod(sampler2D(shadow, material_samplers[SAMPLER_LINEAR_CLAMP]), suv, 0.0).r; if (d < pssm_coord.z) { blocker_average += d; @@ -418,7 +367,7 @@ float sample_directional_soft_shadow(texture2D shadow, vec3 pssm_coord, vec2 tex float s = 0.0; for (uint i = 0; i < sc_directional_penumbra_shadow_samples; i++) { - vec2 suv = pssm_coord.xy + (disk_rotation * scene_data.directional_penumbra_shadow_kernel[i].xy) * tex_scale; + vec2 suv = pssm_coord.xy + (disk_rotation * scene_data_block.data.directional_penumbra_shadow_kernel[i].xy) * tex_scale; s += textureProj(sampler2DShadow(shadow, shadow_sampler), vec4(suv, pssm_coord.z, 1.0)); } @@ -445,7 +394,7 @@ float light_process_omni_shadow(uint idx, vec3 vertex, vec3 normal) { #ifndef SHADOWS_DISABLED if (omni_lights.data[idx].shadow_enabled) { // there is a shadowmap - vec2 texel_size = scene_data.shadow_atlas_pixel_size; + vec2 texel_size = scene_data_block.data.shadow_atlas_pixel_size; vec4 base_uv_rect = omni_lights.data[idx].atlas_rect; base_uv_rect.xy += texel_size; base_uv_rect.zw -= texel_size * 2.0; @@ -489,7 +438,7 @@ float light_process_omni_shadow(uint idx, vec3 vertex, vec3 normal) { bitangent *= omni_lights.data[idx].soft_shadow_size * omni_lights.data[idx].soft_shadow_scale; for (uint i = 0; i < sc_penumbra_shadow_samples; i++) { - vec2 disk = disk_rotation * scene_data.penumbra_shadow_kernel[i].xy; + vec2 disk = disk_rotation * scene_data_block.data.penumbra_shadow_kernel[i].xy; vec3 pos = local_vert + tangent * disk.x + bitangent * disk.y; @@ -525,7 +474,7 @@ float light_process_omni_shadow(uint idx, vec3 vertex, vec3 normal) { shadow = 0.0; for (uint i = 0; i < sc_penumbra_shadow_samples; i++) { - vec2 disk = disk_rotation * scene_data.penumbra_shadow_kernel[i].xy; + vec2 disk = disk_rotation * scene_data_block.data.penumbra_shadow_kernel[i].xy; vec3 pos = local_vert + tangent * disk.x + bitangent * disk.y; pos = normalize(pos); @@ -587,7 +536,7 @@ void light_process_omni(uint idx, vec3 vertex, vec3 eye_vec, vec3 normal, vec3 v float rim, float rim_tint, #endif #ifdef LIGHT_CLEARCOAT_USED - float clearcoat, float clearcoat_gloss, + float clearcoat, float clearcoat_roughness, vec3 vertex_normal, #endif #ifdef LIGHT_ANISOTROPY_USED vec3 binormal, vec3 tangent, float anisotropy, @@ -630,7 +579,7 @@ void light_process_omni(uint idx, vec3 vertex, vec3 eye_vec, vec3 normal, vec3 v splane.xy = splane.xy * 0.5 + 0.5; splane.z = shadow_len * omni_lights.data[idx].inv_radius; splane.xy = clamp_rect.xy + splane.xy * clamp_rect.zw; - // splane.xy = clamp(splane.xy,clamp_rect.xy + scene_data.shadow_atlas_pixel_size,clamp_rect.xy + clamp_rect.zw - scene_data.shadow_atlas_pixel_size ); + // splane.xy = clamp(splane.xy,clamp_rect.xy + scene_data_block.data.shadow_atlas_pixel_size,clamp_rect.xy + clamp_rect.zw - scene_data_block.data.shadow_atlas_pixel_size ); splane.w = 1.0; //needed? i think it should be 1 already float shadow_z = textureLod(sampler2D(shadow_atlas, material_samplers[SAMPLER_LINEAR_CLAMP]), splane.xy, 0.0).r; @@ -711,7 +660,7 @@ void light_process_omni(uint idx, vec3 vertex, vec3 eye_vec, vec3 normal, vec3 v rim * omni_attenuation, rim_tint, #endif #ifdef LIGHT_CLEARCOAT_USED - clearcoat, clearcoat_gloss, + clearcoat, clearcoat_roughness, vertex_normal, #endif #ifdef LIGHT_ANISOTROPY_USED binormal, tangent, anisotropy, @@ -760,7 +709,7 @@ float light_process_spot_shadow(uint idx, vec3 vertex, vec3 normal) { float uv_size = spot_lights.data[idx].soft_shadow_size * z_norm * spot_lights.data[idx].soft_shadow_scale; vec2 clamp_max = spot_lights.data[idx].atlas_rect.xy + spot_lights.data[idx].atlas_rect.zw; for (uint i = 0; i < sc_penumbra_shadow_samples; i++) { - vec2 suv = shadow_uv + (disk_rotation * scene_data.penumbra_shadow_kernel[i].xy) * uv_size; + vec2 suv = shadow_uv + (disk_rotation * scene_data_block.data.penumbra_shadow_kernel[i].xy) * uv_size; suv = clamp(suv, spot_lights.data[idx].atlas_rect.xy, clamp_max); float d = textureLod(sampler2D(shadow_atlas, material_samplers[SAMPLER_LINEAR_CLAMP]), suv, 0.0).r; if (d < splane.z) { @@ -777,7 +726,7 @@ float light_process_spot_shadow(uint idx, vec3 vertex, vec3 normal) { shadow = 0.0; for (uint i = 0; i < sc_penumbra_shadow_samples; i++) { - vec2 suv = shadow_uv + (disk_rotation * scene_data.penumbra_shadow_kernel[i].xy) * uv_size; + vec2 suv = shadow_uv + (disk_rotation * scene_data_block.data.penumbra_shadow_kernel[i].xy) * uv_size; suv = clamp(suv, spot_lights.data[idx].atlas_rect.xy, clamp_max); shadow += textureProj(sampler2DShadow(shadow_atlas, shadow_sampler), vec4(suv, splane.z, 1.0)); } @@ -791,7 +740,7 @@ float light_process_spot_shadow(uint idx, vec3 vertex, vec3 normal) { } else { //hard shadow vec3 shadow_uv = vec3(splane.xy * spot_lights.data[idx].atlas_rect.zw + spot_lights.data[idx].atlas_rect.xy, splane.z); - shadow = sample_pcf_shadow(shadow_atlas, spot_lights.data[idx].soft_shadow_scale * scene_data.shadow_atlas_pixel_size, shadow_uv); + shadow = sample_pcf_shadow(shadow_atlas, spot_lights.data[idx].soft_shadow_scale * scene_data_block.data.shadow_atlas_pixel_size, shadow_uv); } return shadow; @@ -827,7 +776,7 @@ void light_process_spot(uint idx, vec3 vertex, vec3 eye_vec, vec3 normal, vec3 v float rim, float rim_tint, #endif #ifdef LIGHT_CLEARCOAT_USED - float clearcoat, float clearcoat_gloss, + float clearcoat, float clearcoat_roughness, vec3 vertex_normal, #endif #ifdef LIGHT_ANISOTROPY_USED vec3 binormal, vec3 tangent, float anisotropy, @@ -912,7 +861,7 @@ void light_process_spot(uint idx, vec3 vertex, vec3 eye_vec, vec3 normal, vec3 v rim * spot_attenuation, rim_tint, #endif #ifdef LIGHT_CLEARCOAT_USED - clearcoat, clearcoat_gloss, + clearcoat, clearcoat_roughness, vertex_normal, #endif #ifdef LIGHT_ANISOTROPY_USED binormal, tangent, anisotropy, @@ -920,7 +869,7 @@ void light_process_spot(uint idx, vec3 vertex, vec3 eye_vec, vec3 normal, vec3 v diffuse_light, specular_light); } -void reflection_process(uint ref_index, vec3 vertex, vec3 normal, float roughness, vec3 ambient_light, vec3 specular_light, inout vec4 ambient_accum, inout vec4 reflection_accum) { +void reflection_process(uint ref_index, vec3 view, vec3 vertex, vec3 normal, float roughness, vec3 ambient_light, vec3 specular_light, inout vec4 ambient_accum, inout vec4 reflection_accum) { vec3 box_extents = reflections.data[ref_index].box_extents; vec3 local_pos = (reflections.data[ref_index].local_matrix * vec4(vertex, 1.0)).xyz; @@ -928,7 +877,7 @@ void reflection_process(uint ref_index, vec3 vertex, vec3 normal, float roughnes return; } - vec3 ref_vec = normalize(reflect(vertex, normal)); + vec3 ref_vec = normalize(reflect(-view, normal)); vec3 inner_pos = abs(local_pos / box_extents); float blend = max(inner_pos.x, max(inner_pos.y, inner_pos.z)); diff --git a/servers/rendering/renderer_rd/shaders/scene_forward_mobile.glsl b/servers/rendering/renderer_rd/shaders/scene_forward_mobile.glsl index 4d6a3b5864..26d0de46c2 100644 --- a/servers/rendering/renderer_rd/shaders/scene_forward_mobile.glsl +++ b/servers/rendering/renderer_rd/shaders/scene_forward_mobile.glsl @@ -114,6 +114,8 @@ invariant gl_Position; #GLOBALS +#define scene_data scene_data_block.data + void main() { vec4 instance_custom = vec4(0.0); #if defined(COLOR_USED) @@ -122,13 +124,13 @@ void main() { bool is_multimesh = bool(draw_call.flags & INSTANCE_FLAGS_MULTIMESH); - mat4 world_matrix = draw_call.transform; + mat4 model_matrix = draw_call.transform; - mat3 world_normal_matrix; + mat3 model_normal_matrix; if (bool(draw_call.flags & INSTANCE_FLAGS_NON_UNIFORM_SCALE)) { - world_normal_matrix = transpose(inverse(mat3(world_matrix))); + model_normal_matrix = transpose(inverse(mat3(model_matrix))); } else { - world_normal_matrix = mat3(world_matrix); + model_normal_matrix = mat3(model_matrix); } if (is_multimesh) { @@ -221,8 +223,8 @@ void main() { #endif //transpose matrix = transpose(matrix); - world_matrix = world_matrix * matrix; - world_normal_matrix = world_normal_matrix * mat3(matrix); + model_matrix = model_matrix * matrix; + model_normal_matrix = model_normal_matrix * mat3(matrix); } vec3 vertex = vertex_attrib; @@ -259,24 +261,24 @@ void main() { //using world coordinates #if !defined(SKIP_TRANSFORM_USED) && defined(VERTEX_WORLD_COORDS_USED) - vertex = (world_matrix * vec4(vertex, 1.0)).xyz; + vertex = (model_matrix * vec4(vertex, 1.0)).xyz; #ifdef NORMAL_USED - normal = world_normal_matrix * normal; + normal = model_normal_matrix * normal; #endif #if defined(TANGENT_USED) || defined(NORMAL_MAP_USED) || defined(LIGHT_ANISOTROPY_USED) - tangent = world_normal_matrix * tangent; - binormal = world_normal_matrix * binormal; + tangent = model_normal_matrix * tangent; + binormal = model_normal_matrix * binormal; #endif #endif float roughness = 1.0; - mat4 modelview = scene_data.inv_camera_matrix * world_matrix; - mat3 modelview_normal = mat3(scene_data.inv_camera_matrix) * world_normal_matrix; + mat4 modelview = scene_data.view_matrix * model_matrix; + mat3 modelview_normal = mat3(scene_data.view_matrix) * model_normal_matrix; { #CODE : VERTEX @@ -303,14 +305,14 @@ void main() { //using world coordinates #if !defined(SKIP_TRANSFORM_USED) && defined(VERTEX_WORLD_COORDS_USED) - vertex = (scene_data.inv_camera_matrix * vec4(vertex, 1.0)).xyz; + vertex = (scene_data.view_matrix * vec4(vertex, 1.0)).xyz; #ifdef NORMAL_USED - normal = (scene_data.inv_camera_matrix * vec4(normal, 0.0)).xyz; + normal = (scene_data.view_matrix * vec4(normal, 0.0)).xyz; #endif #if defined(TANGENT_USED) || defined(NORMAL_MAP_USED) || defined(LIGHT_ANISOTROPY_USED) - binormal = (scene_data.inv_camera_matrix * vec4(binormal, 0.0)).xyz; - tangent = (scene_data.inv_camera_matrix * vec4(tangent, 0.0)).xyz; + binormal = (scene_data.view_matrix * vec4(binormal, 0.0)).xyz; + tangent = (scene_data.view_matrix * vec4(tangent, 0.0)).xyz; #endif #endif @@ -458,7 +460,7 @@ layout(location = 8) highp in float dp_clip; //defines to keep compatibility with vertex -#define world_matrix draw_call.transform +#define model_matrix draw_call.transform #ifdef USE_MULTIVIEW #define projection_matrix scene_data.projection_matrix_view[ViewIndex] #else @@ -511,8 +513,8 @@ layout(location = 0) out mediump vec4 frag_color; #if !defined(MODE_RENDER_DEPTH) && !defined(MODE_UNSHADED) -/* Make a default specular mode SPECULAR_SCHLICK_GGX. */ -#if !defined(SPECULAR_DISABLED) && !defined(SPECULAR_SCHLICK_GGX) && !defined(SPECULAR_BLINN) && !defined(SPECULAR_PHONG) && !defined(SPECULAR_TOON) +// Default to SPECULAR_SCHLICK_GGX. +#if !defined(SPECULAR_DISABLED) && !defined(SPECULAR_SCHLICK_GGX) && !defined(SPECULAR_TOON) #define SPECULAR_SCHLICK_GGX #endif @@ -527,13 +529,13 @@ layout(location = 0) out mediump vec4 frag_color; */ vec4 fog_process(vec3 vertex) { - vec3 fog_color = scene_data.fog_light_color; + vec3 fog_color = scene_data_block.data.fog_light_color; - if (scene_data.fog_aerial_perspective > 0.0) { + if (scene_data_block.data.fog_aerial_perspective > 0.0) { vec3 sky_fog_color = vec3(0.0); - vec3 cube_view = scene_data.radiance_inverse_xform * vertex; + vec3 cube_view = scene_data_block.data.radiance_inverse_xform * vertex; // mip_level always reads from the second mipmap and higher so the fog is always slightly blurred - float mip_level = mix(1.0 / MAX_ROUGHNESS_LOD, 1.0, 1.0 - (abs(vertex.z) - scene_data.z_near) / (scene_data.z_far - scene_data.z_near)); + float mip_level = mix(1.0 / MAX_ROUGHNESS_LOD, 1.0, 1.0 - (abs(vertex.z) - scene_data_block.data.z_near) / (scene_data_block.data.z_far - scene_data_block.data.z_near)); #ifdef USE_RADIANCE_CUBEMAP_ARRAY float lod, blend; blend = modf(mip_level * MAX_ROUGHNESS_LOD, lod); @@ -542,29 +544,29 @@ vec4 fog_process(vec3 vertex) { #else sky_fog_color = textureLod(samplerCube(radiance_cubemap, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), cube_view, mip_level * MAX_ROUGHNESS_LOD).rgb; #endif //USE_RADIANCE_CUBEMAP_ARRAY - fog_color = mix(fog_color, sky_fog_color, scene_data.fog_aerial_perspective); + fog_color = mix(fog_color, sky_fog_color, scene_data_block.data.fog_aerial_perspective); } - if (scene_data.fog_sun_scatter > 0.001) { + if (scene_data_block.data.fog_sun_scatter > 0.001) { vec4 sun_scatter = vec4(0.0); float sun_total = 0.0; vec3 view = normalize(vertex); - for (uint i = 0; i < scene_data.directional_light_count; i++) { + for (uint i = 0; i < scene_data_block.data.directional_light_count; i++) { vec3 light_color = directional_lights.data[i].color * directional_lights.data[i].energy; float light_amount = pow(max(dot(view, directional_lights.data[i].direction), 0.0), 8.0); - fog_color += light_color * light_amount * scene_data.fog_sun_scatter; + fog_color += light_color * light_amount * scene_data_block.data.fog_sun_scatter; } } - float fog_amount = 1.0 - exp(min(0.0, -length(vertex) * scene_data.fog_density)); + float fog_amount = 1.0 - exp(min(0.0, -length(vertex) * scene_data_block.data.fog_density)); - if (abs(scene_data.fog_height_density) >= 0.0001) { - float y = (scene_data.camera_matrix * vec4(vertex, 1.0)).y; + if (abs(scene_data_block.data.fog_height_density) >= 0.0001) { + float y = (scene_data_block.data.inv_view_matrix * vec4(vertex, 1.0)).y; - float y_dist = y - scene_data.fog_height; + float y_dist = y - scene_data_block.data.fog_height; - float vfog_amount = 1.0 - exp(min(0.0, y_dist * scene_data.fog_height_density)); + float vfog_amount = 1.0 - exp(min(0.0, y_dist * scene_data_block.data.fog_height_density)); fog_amount = max(vfog_amount, fog_amount); } @@ -574,6 +576,8 @@ vec4 fog_process(vec3 vertex) { #endif //!MODE_RENDER DEPTH +#define scene_data scene_data_block.data + void main() { #ifdef MODE_DUAL_PARABOLOID @@ -583,7 +587,11 @@ void main() { //lay out everything, whatever is unused is optimized away anyway vec3 vertex = vertex_interp; +#ifdef USE_MULTIVIEW + vec3 view = -normalize(vertex_interp - scene_data.eye_offset[ViewIndex].xyz); +#else vec3 view = -normalize(vertex_interp); +#endif vec3 albedo = vec3(1.0); vec3 backlight = vec3(0.0); vec4 transmittance_color = vec4(0.0); @@ -596,7 +604,7 @@ void main() { float rim = 0.0; float rim_tint = 0.0; float clearcoat = 0.0; - float clearcoat_gloss = 0.0; + float clearcoat_roughness = 0.0; float anisotropy = 0.0; vec2 anisotropy_flow = vec2(1.0, 0.0); vec4 fog = vec4(0.0); @@ -650,7 +658,7 @@ void main() { float normal_map_depth = 1.0; - vec2 screen_uv = gl_FragCoord.xy * scene_data.screen_pixel_size + scene_data.screen_pixel_size * 0.5; //account for center + vec2 screen_uv = gl_FragCoord.xy * scene_data.screen_pixel_size; float sss_strength = 0.0; @@ -874,7 +882,16 @@ void main() { #if !defined(MODE_RENDER_DEPTH) && !defined(MODE_UNSHADED) if (scene_data.use_reflection_cubemap) { +#ifdef LIGHT_ANISOTROPY_USED + // https://google.github.io/filament/Filament.html#lighting/imagebasedlights/anisotropy + vec3 anisotropic_direction = anisotropy >= 0.0 ? binormal : tangent; + vec3 anisotropic_tangent = cross(anisotropic_direction, view); + vec3 anisotropic_normal = cross(anisotropic_tangent, anisotropic_direction); + vec3 bent_normal = normalize(mix(normal, anisotropic_normal, abs(anisotropy) * clamp(5.0 * roughness, 0.0, 1.0))); + vec3 ref_vec = reflect(-view, bent_normal); +#else vec3 ref_vec = reflect(-view, normal); +#endif float horizon = min(1.0 + dot(ref_vec, normal), 1.0); ref_vec = scene_data.radiance_inverse_xform * ref_vec; #ifdef USE_RADIANCE_CUBEMAP_ARRAY @@ -915,9 +932,37 @@ void main() { #endif // !USE_LIGHTMAP #if defined(CUSTOM_IRRADIANCE_USED) - ambient_light = mix(specular_light, custom_irradiance.rgb, custom_irradiance.a); + ambient_light = mix(ambient_light, custom_irradiance.rgb, custom_irradiance.a); #endif // CUSTOM_IRRADIANCE_USED +#ifdef LIGHT_CLEARCOAT_USED + + if (scene_data.use_reflection_cubemap) { + vec3 n = normalize(normal_interp); // We want to use geometric normal, not normal_map + float NoV = max(dot(n, view), 0.0001); + vec3 ref_vec = reflect(-view, n); + // The clear coat layer assumes an IOR of 1.5 (4% reflectance) + float Fc = clearcoat * (0.04 + 0.96 * SchlickFresnel(NoV)); + float attenuation = 1.0 - Fc; + ambient_light *= attenuation; + specular_light *= attenuation; + + float horizon = min(1.0 + dot(ref_vec, normal), 1.0); + ref_vec = scene_data.radiance_inverse_xform * ref_vec; + float roughness_lod = mix(0.001, 0.1, clearcoat_roughness) * MAX_ROUGHNESS_LOD; +#ifdef USE_RADIANCE_CUBEMAP_ARRAY + + float lod, blend; + blend = modf(roughness_lod, lod); + vec3 clearcoat_light = texture(samplerCubeArray(radiance_cubemap, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), vec4(ref_vec, lod)).rgb; + clearcoat_light = mix(clearcoat_light, texture(samplerCubeArray(radiance_cubemap, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), vec4(ref_vec, lod + 1)).rgb, blend); + +#else + vec3 clearcoat_light = textureLod(samplerCube(radiance_cubemap, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), ref_vec, roughness_lod).rgb; +#endif //USE_RADIANCE_CUBEMAP_ARRAY + specular_light += clearcoat_light * horizon * horizon * Fc * scene_data.ambient_light_color_energy.a; + } +#endif #endif //!defined(MODE_RENDER_DEPTH) && !defined(MODE_UNSHADED) //radiance @@ -930,7 +975,7 @@ void main() { if (bool(draw_call.flags & INSTANCE_FLAGS_USE_LIGHTMAP_CAPTURE)) { //has lightmap capture uint index = draw_call.gi_offset; - vec3 wnormal = mat3(scene_data.camera_matrix) * normal; + vec3 wnormal = mat3(scene_data.inv_view_matrix) * normal; const float c1 = 0.429043; const float c2 = 0.511664; const float c3 = 0.743125; @@ -1002,13 +1047,27 @@ void main() { if (reflection_index == 0xFF) { break; } - - reflection_process(reflection_index, vertex, normal, roughness, ambient_light, specular_light, ambient_accum, reflection_accum); +#ifdef LIGHT_ANISOTROPY_USED + // https://google.github.io/filament/Filament.html#lighting/imagebasedlights/anisotropy + vec3 anisotropic_direction = anisotropy >= 0.0 ? binormal : tangent; + vec3 anisotropic_tangent = cross(anisotropic_direction, view); + vec3 anisotropic_normal = cross(anisotropic_tangent, anisotropic_direction); + vec3 bent_normal = normalize(mix(normal, anisotropic_normal, abs(anisotropy) * clamp(5.0 * roughness, 0.0, 1.0))); +#else + vec3 bent_normal = normal; +#endif + reflection_process(reflection_index, view, vertex, bent_normal, roughness, ambient_light, specular_light, ambient_accum, reflection_accum); } if (reflection_accum.a > 0.0) { specular_light = reflection_accum.rgb / reflection_accum.a; } + +#if !defined(USE_LIGHTMAP) + if (ambient_accum.a > 0.0) { + ambient_light = ambient_accum.rgb / ambient_accum.a; + } +#endif } //Reflection probes // finalize ambient light here @@ -1079,7 +1138,6 @@ void main() { float depth_z = -vertex.z; vec4 pssm_coord; - vec3 shadow_color = vec3(0.0); vec3 light_dir = directional_lights.data[i].direction; #define BIAS_FUNC(m_var, m_idx) \ @@ -1105,9 +1163,6 @@ void main() { } else { shadow = sample_directional_pcf_shadow(directional_shadow_atlas, scene_data.directional_shadow_pixel_size * directional_lights.data[i].soft_shadow_scale, pssm_coord); } - - shadow_color = directional_lights.data[i].shadow_color1.rgb; - } else if (depth_z < directional_lights.data[i].shadow_split_offsets.y) { vec4 v = vec4(vertex, 1.0); @@ -1125,8 +1180,6 @@ void main() { } else { shadow = sample_directional_pcf_shadow(directional_shadow_atlas, scene_data.directional_shadow_pixel_size * directional_lights.data[i].soft_shadow_scale, pssm_coord); } - - shadow_color = directional_lights.data[i].shadow_color2.rgb; } else if (depth_z < directional_lights.data[i].shadow_split_offsets.z) { vec4 v = vec4(vertex, 1.0); @@ -1144,9 +1197,6 @@ void main() { } else { shadow = sample_directional_pcf_shadow(directional_shadow_atlas, scene_data.directional_shadow_pixel_size * directional_lights.data[i].soft_shadow_scale, pssm_coord); } - - shadow_color = directional_lights.data[i].shadow_color3.rgb; - } else { vec4 v = vec4(vertex, 1.0); @@ -1164,12 +1214,9 @@ void main() { } else { shadow = sample_directional_pcf_shadow(directional_shadow_atlas, scene_data.directional_shadow_pixel_size * directional_lights.data[i].soft_shadow_scale, pssm_coord); } - - shadow_color = directional_lights.data[i].shadow_color4.rgb; } if (directional_lights.data[i].blend_splits) { - vec3 shadow_color_blend = vec3(0.0); float pssm_blend; float shadow2; @@ -1190,7 +1237,6 @@ void main() { } pssm_blend = smoothstep(0.0, directional_lights.data[i].shadow_split_offsets.x, depth_z); - shadow_color_blend = directional_lights.data[i].shadow_color2.rgb; } else if (depth_z < directional_lights.data[i].shadow_split_offsets.y) { vec4 v = vec4(vertex, 1.0); BIAS_FUNC(v, 2) @@ -1208,8 +1254,6 @@ void main() { } pssm_blend = smoothstep(directional_lights.data[i].shadow_split_offsets.x, directional_lights.data[i].shadow_split_offsets.y, depth_z); - - shadow_color_blend = directional_lights.data[i].shadow_color3.rgb; } else if (depth_z < directional_lights.data[i].shadow_split_offsets.z) { vec4 v = vec4(vertex, 1.0); BIAS_FUNC(v, 3) @@ -1226,7 +1270,6 @@ void main() { } pssm_blend = smoothstep(directional_lights.data[i].shadow_split_offsets.y, directional_lights.data[i].shadow_split_offsets.z, depth_z); - shadow_color_blend = directional_lights.data[i].shadow_color4.rgb; } else { pssm_blend = 0.0; //if no blend, same coord will be used (divide by z will result in same value, and already cached) } @@ -1234,7 +1277,6 @@ void main() { pssm_blend = sqrt(pssm_blend); shadow = mix(shadow, shadow2, pssm_blend); - shadow_color = mix(shadow_color, shadow_color_blend, pssm_blend); } shadow = mix(shadow, 1.0, smoothstep(directional_lights.data[i].fade_from, directional_lights.data[i].fade_to, vertex.z)); //done with negative values for performance @@ -1248,6 +1290,7 @@ void main() { float depth_z = -vertex.z; vec4 pssm_coord; + float blur_factor; vec3 light_dir = directional_lights.data[i].direction; vec3 base_normal_bias = normalize(normal_interp) * (1.0 - max(0.0, dot(light_dir, -normalize(normal_interp)))); @@ -1263,56 +1306,71 @@ void main() { BIAS_FUNC(v, 0) pssm_coord = (directional_lights.data[i].shadow_matrix1 * v); + blur_factor = 1.0; } else if (depth_z < directional_lights.data[i].shadow_split_offsets.y) { vec4 v = vec4(vertex, 1.0); BIAS_FUNC(v, 1) pssm_coord = (directional_lights.data[i].shadow_matrix2 * v); + // Adjust shadow blur with reference to the first split to reduce discrepancy between shadow splits. + blur_factor = directional_lights.data[i].shadow_split_offsets.x / directional_lights.data[i].shadow_split_offsets.y; + ; } else if (depth_z < directional_lights.data[i].shadow_split_offsets.z) { vec4 v = vec4(vertex, 1.0); BIAS_FUNC(v, 2) pssm_coord = (directional_lights.data[i].shadow_matrix3 * v); - + // Adjust shadow blur with reference to the first split to reduce discrepancy between shadow splits. + blur_factor = directional_lights.data[i].shadow_split_offsets.x / directional_lights.data[i].shadow_split_offsets.z; } else { vec4 v = vec4(vertex, 1.0); BIAS_FUNC(v, 3) pssm_coord = (directional_lights.data[i].shadow_matrix4 * v); + // Adjust shadow blur with reference to the first split to reduce discrepancy between shadow splits. + blur_factor = directional_lights.data[i].shadow_split_offsets.x / directional_lights.data[i].shadow_split_offsets.w; } pssm_coord /= pssm_coord.w; - shadow = sample_directional_pcf_shadow(directional_shadow_atlas, scene_data.directional_shadow_pixel_size * directional_lights.data[i].soft_shadow_scale, pssm_coord); + shadow = sample_directional_pcf_shadow(directional_shadow_atlas, scene_data.directional_shadow_pixel_size * directional_lights.data[i].soft_shadow_scale * blur_factor, pssm_coord); if (directional_lights.data[i].blend_splits) { float pssm_blend; + float blur_factor2; if (depth_z < directional_lights.data[i].shadow_split_offsets.x) { vec4 v = vec4(vertex, 1.0); BIAS_FUNC(v, 1) pssm_coord = (directional_lights.data[i].shadow_matrix2 * v); pssm_blend = smoothstep(0.0, directional_lights.data[i].shadow_split_offsets.x, depth_z); + // Adjust shadow blur with reference to the first split to reduce discrepancy between shadow splits. + blur_factor2 = directional_lights.data[i].shadow_split_offsets.x / directional_lights.data[i].shadow_split_offsets.y; } else if (depth_z < directional_lights.data[i].shadow_split_offsets.y) { vec4 v = vec4(vertex, 1.0); BIAS_FUNC(v, 2) pssm_coord = (directional_lights.data[i].shadow_matrix3 * v); pssm_blend = smoothstep(directional_lights.data[i].shadow_split_offsets.x, directional_lights.data[i].shadow_split_offsets.y, depth_z); + // Adjust shadow blur with reference to the first split to reduce discrepancy between shadow splits. + blur_factor2 = directional_lights.data[i].shadow_split_offsets.x / directional_lights.data[i].shadow_split_offsets.z; } else if (depth_z < directional_lights.data[i].shadow_split_offsets.z) { vec4 v = vec4(vertex, 1.0); BIAS_FUNC(v, 3) pssm_coord = (directional_lights.data[i].shadow_matrix4 * v); pssm_blend = smoothstep(directional_lights.data[i].shadow_split_offsets.y, directional_lights.data[i].shadow_split_offsets.z, depth_z); + // Adjust shadow blur with reference to the first split to reduce discrepancy between shadow splits. + blur_factor2 = directional_lights.data[i].shadow_split_offsets.x / directional_lights.data[i].shadow_split_offsets.w; } else { pssm_blend = 0.0; //if no blend, same coord will be used (divide by z will result in same value, and already cached) + blur_factor2 = 1.0; } pssm_coord /= pssm_coord.w; - float shadow2 = sample_directional_pcf_shadow(directional_shadow_atlas, scene_data.directional_shadow_pixel_size * directional_lights.data[i].soft_shadow_scale, pssm_coord); + float shadow2 = sample_directional_pcf_shadow(directional_shadow_atlas, scene_data.directional_shadow_pixel_size * directional_lights.data[i].soft_shadow_scale * blur_factor2, pssm_coord); shadow = mix(shadow, shadow2, pssm_blend); } @@ -1368,7 +1426,7 @@ void main() { rim, rim_tint, #endif #ifdef LIGHT_CLEARCOAT_USED - clearcoat, clearcoat_gloss, + clearcoat, clearcoat_roughness, normalize(normal_interp), #endif #ifdef LIGHT_ANISOTROPY_USED binormal, tangent, anisotropy, @@ -1415,10 +1473,11 @@ void main() { rim_tint, #endif #ifdef LIGHT_CLEARCOAT_USED - clearcoat, clearcoat_gloss, + clearcoat, clearcoat_roughness, normalize(normal_interp), #endif #ifdef LIGHT_ANISOTROPY_USED - tangent, binormal, anisotropy, + tangent, + binormal, anisotropy, #endif diffuse_light, specular_light); } @@ -1459,10 +1518,11 @@ void main() { rim_tint, #endif #ifdef LIGHT_CLEARCOAT_USED - clearcoat, clearcoat_gloss, + clearcoat, clearcoat_roughness, normalize(normal_interp), #endif #ifdef LIGHT_ANISOTROPY_USED - tangent, binormal, anisotropy, + tangent, + binormal, anisotropy, #endif diffuse_light, specular_light); } diff --git a/servers/rendering/renderer_rd/shaders/scene_forward_mobile_inc.glsl b/servers/rendering/renderer_rd/shaders/scene_forward_mobile_inc.glsl index 541c0b0603..7413d8730a 100644 --- a/servers/rendering/renderer_rd/shaders/scene_forward_mobile_inc.glsl +++ b/servers/rendering/renderer_rd/shaders/scene_forward_mobile_inc.glsl @@ -7,7 +7,7 @@ #include "decal_data_inc.glsl" -#if !defined(MODE_RENDER_DEPTH) || defined(MODE_RENDER_MATERIAL) || defined(TANGENT_USED) || defined(NORMAL_MAP_USED) +#if !defined(MODE_RENDER_DEPTH) || defined(MODE_RENDER_MATERIAL) || defined(TANGENT_USED) || defined(NORMAL_MAP_USED) || defined(LIGHT_ANISOTROPY_USED) #ifndef NORMAL_USED #define NORMAL_USED #endif @@ -125,15 +125,16 @@ global_variables; /* Set 1: Render Pass (changes per render pass) */ -layout(set = 1, binding = 0, std140) uniform SceneData { +struct SceneData { highp mat4 projection_matrix; highp mat4 inv_projection_matrix; - highp mat4 camera_matrix; - highp mat4 inv_camera_matrix; + highp mat4 inv_view_matrix; + highp mat4 view_matrix; // only used for multiview highp mat4 projection_matrix_view[MAX_VIEWS]; highp mat4 inv_projection_matrix_view[MAX_VIEWS]; + highp vec4 eye_offset[MAX_VIEWS]; highp vec2 viewport_size; highp vec2 screen_pixel_size; @@ -189,8 +190,12 @@ layout(set = 1, binding = 0, std140) uniform SceneData { uint pad1; uint pad2; uint pad3; +}; + +layout(set = 1, binding = 0, std140) uniform SceneDataBlock { + SceneData data; } -scene_data; +scene_data_block; #ifdef USE_RADIANCE_CUBEMAP_ARRAY diff --git a/servers/rendering/renderer_rd/shaders/skeleton.glsl b/servers/rendering/renderer_rd/shaders/skeleton.glsl index 4ef6a26443..a893a66c94 100644 --- a/servers/rendering/renderer_rd/shaders/skeleton.glsl +++ b/servers/rendering/renderer_rd/shaders/skeleton.glsl @@ -160,7 +160,7 @@ void main() { } if (params.has_tangent) { - blend_tangent += decode_abgr_2_10_10_10(src_blend_shapes.data[base_offset]).rgb; + blend_tangent += decode_abgr_2_10_10_10(src_blend_shapes.data[base_offset]).rgb * w; } blend_total += w; @@ -174,8 +174,8 @@ void main() { } vertex += blend_vertex; - normal += normalize(normal + blend_normal); - tangent.rgb += normalize(tangent.rgb + blend_tangent); + normal = normalize(normal + blend_normal); + tangent.rgb = normalize(tangent.rgb + blend_tangent); } if (params.has_skeleton) { diff --git a/servers/rendering/renderer_rd/shaders/specular_merge.glsl b/servers/rendering/renderer_rd/shaders/specular_merge.glsl deleted file mode 100644 index 3579c35cce..0000000000 --- a/servers/rendering/renderer_rd/shaders/specular_merge.glsl +++ /dev/null @@ -1,53 +0,0 @@ -#[vertex] - -#version 450 - -#VERSION_DEFINES - -layout(location = 0) out vec2 uv_interp; - -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); -} - -#[fragment] - -#version 450 - -#VERSION_DEFINES - -layout(location = 0) in vec2 uv_interp; - -layout(set = 0, binding = 0) uniform sampler2D specular; - -#ifdef MODE_SSR - -layout(set = 1, binding = 0) uniform sampler2D ssr; - -#endif - -#ifdef MODE_MERGE - -layout(set = 2, binding = 0) uniform sampler2D diffuse; - -#endif - -layout(location = 0) out vec4 frag_color; - -void main() { - frag_color.rgb = texture(specular, uv_interp).rgb; - frag_color.a = 0.0; -#ifdef MODE_SSR - - vec4 ssr_color = texture(ssr, uv_interp); - frag_color.rgb = mix(frag_color.rgb, ssr_color.rgb, ssr_color.a); -#endif - -#ifdef MODE_MERGE - frag_color += texture(diffuse, uv_interp); -#endif - //added using additive blend -} diff --git a/servers/rendering/renderer_rd/shaders/taa_resolve.glsl b/servers/rendering/renderer_rd/shaders/taa_resolve.glsl new file mode 100644 index 0000000000..b0a0839836 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/taa_resolve.glsl @@ -0,0 +1,394 @@ +/////////////////////////////////////////////////////////////////////////////////// +// 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> + +#define USE_SUBGROUPS + +#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)); +} |