diff options
Diffstat (limited to 'servers/rendering/renderer_rd/shaders')
40 files changed, 1009 insertions, 646 deletions
diff --git a/servers/rendering/renderer_rd/shaders/SCsub b/servers/rendering/renderer_rd/shaders/SCsub index d352743908..5405985741 100644 --- a/servers/rendering/renderer_rd/shaders/SCsub +++ b/servers/rendering/renderer_rd/shaders/SCsub @@ -18,3 +18,5 @@ if "RD_GLSL" in env["BUILDERS"]: SConscript("effects/SCsub") SConscript("environment/SCsub") +SConscript("forward_clustered/SCsub") +SConscript("forward_mobile/SCsub") diff --git a/servers/rendering/renderer_rd/shaders/canvas.glsl b/servers/rendering/renderer_rd/shaders/canvas.glsl index f8e9020f9f..eb5f68849e 100644 --- a/servers/rendering/renderer_rd/shaders/canvas.glsl +++ b/servers/rendering/renderer_rd/shaders/canvas.glsl @@ -191,48 +191,6 @@ void main() { uv += 1e-5; } -#ifdef USE_ATTRIBUTES -#if 0 - if (bool(draw_data.flags & FLAGS_USE_SKELETON) && bone_weights != vec4(0.0)) { //must be a valid bone - //skeleton transform - ivec4 bone_indicesi = ivec4(bone_indices); - - uvec2 tex_ofs = bone_indicesi.x * 2; - - mat2x4 m; - m = mat2x4( - texelFetch(skeleton_buffer, tex_ofs + 0), - texelFetch(skeleton_buffer, tex_ofs + 1)) * - bone_weights.x; - - tex_ofs = bone_indicesi.y * 2; - - m += mat2x4( - texelFetch(skeleton_buffer, tex_ofs + 0), - texelFetch(skeleton_buffer, tex_ofs + 1)) * - bone_weights.y; - - tex_ofs = bone_indicesi.z * 2; - - m += mat2x4( - texelFetch(skeleton_buffer, tex_ofs + 0), - texelFetch(skeleton_buffer, tex_ofs + 1)) * - bone_weights.z; - - tex_ofs = bone_indicesi.w * 2; - - m += mat2x4( - texelFetch(skeleton_buffer, tex_ofs + 0), - texelFetch(skeleton_buffer, tex_ofs + 1)) * - bone_weights.w; - - mat4 bone_matrix = skeleton_data.skeleton_transform * transpose(mat4(m[0], m[1], vec4(0.0, 0.0, 1.0, 0.0), vec4(0.0, 0.0, 0.0, 1.0))) * skeleton_data.skeleton_transform_inverse; - - //outvec = bone_matrix * outvec; - } -#endif -#endif - vertex = (canvas_data.canvas_transform * vec4(vertex, 0.0, 1.0)).xy; vertex_interp = vertex; @@ -313,6 +271,14 @@ vec4 light_compute( vec2 uv, vec4 color, bool is_directional) { vec4 light = vec4(0.0); + vec3 light_direction = vec3(0.0); + + if (is_directional) { + light_direction = normalize(mix(vec3(light_position.xy, 0.0), vec3(0, 0, 1), light_position.z)); + light_position = vec3(0.0); + } else { + light_direction = normalize(light_position - light_vertex); + } #CODE : LIGHT @@ -509,7 +475,13 @@ void main() { float a = clamp(d * px_size + 0.5, 0.0, 1.0); color.a = a * color.a; } - + } else if (bool(draw_data.flags & FLAGS_USE_LCD)) { + vec4 lcd_sample = texture(sampler2D(color_texture, texture_sampler), uv); + if (lcd_sample.a == 1.0) { + color.rgb = lcd_sample.rgb * color.a; + } else { + color = vec4(0.0, 0.0, 0.0, 0.0); + } } else { #else { @@ -584,14 +556,14 @@ void main() { normal = normalize((canvas_data.canvas_normal_transform * vec4(normal, 0.0)).xyz); } - vec3 base_color = color.rgb; + vec4 base_color = color; if (bool(draw_data.flags & FLAGS_USING_LIGHT_MASK)) { color = vec4(0.0); //invisible by default due to using light mask } #ifdef MODE_LIGHT_ONLY color = vec4(0.0); -#else +#elif !defined(MODE_UNSHADED) color *= canvas_data.canvas_modulation; #endif @@ -608,12 +580,14 @@ void main() { #ifdef LIGHT_CODE_USED vec4 shadow_modulate = vec4(1.0); - light_color = light_compute(light_vertex, vec3(direction, light_array.data[light_base].height), normal, light_color, light_color.a, specular_shininess, shadow_modulate, screen_uv, uv, color, true); + light_color = light_compute(light_vertex, vec3(direction, light_array.data[light_base].height), normal, light_color, light_color.a, specular_shininess, shadow_modulate, screen_uv, uv, base_color, true); #else if (normal_used) { vec3 light_vec = normalize(mix(vec3(direction, 0.0), vec3(0, 0, 1), light_array.data[light_base].height)); - light_color.rgb = light_normal_compute(light_vec, normal, base_color, light_color.rgb, specular_shininess, specular_shininess_used); + light_color.rgb = light_normal_compute(light_vec, normal, base_color.rgb, light_color.rgb, specular_shininess, specular_shininess_used); + } else { + light_color.rgb *= base_color.rgb; } #endif @@ -639,20 +613,7 @@ void main() { if (i >= light_count) { break; } - uint light_base; - if (i < 8) { - if (i < 4) { - light_base = draw_data.lights[0]; - } else { - light_base = draw_data.lights[1]; - } - } else { - if (i < 12) { - light_base = draw_data.lights[2]; - } else { - light_base = draw_data.lights[3]; - } - } + uint light_base = draw_data.lights[i >> 2]; light_base >>= (i & 3) * 8; light_base &= 0xFF; @@ -667,7 +628,7 @@ void main() { vec3 light_position = vec3(light_array.data[light_base].position, light_array.data[light_base].height); light_color.rgb *= light_base_color.rgb; - light_color = light_compute(light_vertex, light_position, normal, light_color, light_base_color.a, specular_shininess, shadow_modulate, screen_uv, uv, color, false); + light_color = light_compute(light_vertex, light_position, normal, light_color, light_base_color.a, specular_shininess, shadow_modulate, screen_uv, uv, base_color, false); #else light_color.rgb *= light_base_color.rgb * light_base_color.a; @@ -676,9 +637,10 @@ void main() { vec3 light_pos = vec3(light_array.data[light_base].position, light_array.data[light_base].height); vec3 pos = light_vertex; vec3 light_vec = normalize(light_pos - pos); - float cNdotL = max(0.0, dot(normal, light_vec)); - light_color.rgb = light_normal_compute(light_vec, normal, base_color, light_color.rgb, specular_shininess, specular_shininess_used); + light_color.rgb = light_normal_compute(light_vec, normal, base_color.rgb, light_color.rgb, specular_shininess, specular_shininess_used); + } else { + light_color.rgb *= base_color.rgb; } #endif if (any(lessThan(tex_uv, vec2(0.0, 0.0))) || any(greaterThanEqual(tex_uv, vec2(1.0, 1.0)))) { diff --git a/servers/rendering/renderer_rd/shaders/canvas_uniforms_inc.glsl b/servers/rendering/renderer_rd/shaders/canvas_uniforms_inc.glsl index 12f57b0178..1b627a3e81 100644 --- a/servers/rendering/renderer_rd/shaders/canvas_uniforms_inc.glsl +++ b/servers/rendering/renderer_rd/shaders/canvas_uniforms_inc.glsl @@ -25,6 +25,7 @@ #define FLAGS_DEFAULT_SPECULAR_MAP_USED (1 << 27) #define FLAGS_USE_MSDF (1 << 28) +#define FLAGS_USE_LCD (1 << 29) #define SAMPLER_NEAREST_CLAMP 0 #define SAMPLER_LINEAR_CLAMP 1 @@ -138,10 +139,10 @@ layout(set = 0, binding = 7) uniform texture2D sdf_texture; layout(set = 0, binding = 8) uniform sampler material_samplers[12]; -layout(set = 0, binding = 9, std430) restrict readonly buffer GlobalVariableData { +layout(set = 0, binding = 9, std430) restrict readonly buffer GlobalShaderUniformData { vec4 data[]; } -global_variables; +global_shader_uniforms; /* SET1: Is reserved for the material */ diff --git a/servers/rendering/renderer_rd/shaders/cluster_render.glsl b/servers/rendering/renderer_rd/shaders/cluster_render.glsl index 2fe230f0bf..8c26a67926 100644 --- a/servers/rendering/renderer_rd/shaders/cluster_render.glsl +++ b/servers/rendering/renderer_rd/shaders/cluster_render.glsl @@ -64,7 +64,7 @@ void main() { #version 450 #VERSION_DEFINES - +#ifndef MOLTENVK_USED // Metal will corrupt GPU state otherwise #if defined(has_GL_KHR_shader_subgroup_ballot) && defined(has_GL_KHR_shader_subgroup_arithmetic) && defined(has_GL_KHR_shader_subgroup_vote) #extension GL_KHR_shader_subgroup_ballot : enable @@ -73,6 +73,7 @@ void main() { #define USE_SUBGROUPS #endif +#endif layout(location = 0) in float depth_interp; layout(location = 1) in flat uint element_index; @@ -141,7 +142,11 @@ void main() { } } #else - if (!gl_HelperInvocation) { +// MoltenVK/Metal fails to compile shaders using gl_HelperInvocation for some GPUs +#ifndef MOLTENVK_USED + if (!gl_HelperInvocation) +#endif + { atomicOr(cluster_render.data[usage_write_offset], usage_write_bit); } #endif @@ -161,7 +166,11 @@ void main() { } } #else - if (!gl_HelperInvocation) { +// MoltenVK/Metal fails to compile shaders using gl_HelperInvocation for some GPUs +#ifndef MOLTENVK_USED + if (!gl_HelperInvocation) +#endif + { atomicOr(cluster_render.data[z_write_offset], z_write_bit); } #endif diff --git a/servers/rendering/renderer_rd/shaders/effects/SCsub b/servers/rendering/renderer_rd/shaders/effects/SCsub index 741da8fe69..f06a2d86e2 100644 --- a/servers/rendering/renderer_rd/shaders/effects/SCsub +++ b/servers/rendering/renderer_rd/shaders/effects/SCsub @@ -4,7 +4,7 @@ Import("env") if "RD_GLSL" in env["BUILDERS"]: # find all include files - gl_include_files = [str(f) for f in Glob("*_inc.glsl")] + gl_include_files = [str(f) for f in Glob("*_inc.glsl")] + [str(f) for f in Glob("../*_inc.glsl")] # find all shader code(all glsl files excluding our include files) glsl_files = [str(f) for f in Glob("*.glsl") if str(f) not in gl_include_files] diff --git a/servers/rendering/renderer_rd/shaders/effects/blur_raster.glsl b/servers/rendering/renderer_rd/shaders/effects/blur_raster.glsl index 96f5c3e9f2..31aabbe9d2 100644 --- a/servers/rendering/renderer_rd/shaders/effects/blur_raster.glsl +++ b/servers/rendering/renderer_rd/shaders/effects/blur_raster.glsl @@ -53,30 +53,31 @@ void main() { #ifdef MODE_GAUSSIAN_BLUR - // Simpler blur uses SIGMA2 for the gaussian kernel for a stronger effect - - // note, for blur blur.luminance_multiplier is irrelavant, we would be multiplying and then dividing by this amount. - - if (bool(blur.flags & FLAG_HORIZONTAL)) { - vec2 pix_size = blur.pixel_size; - pix_size *= 0.5; //reading from larger buffer, so use more samples - vec4 color = texture(source_color, uv_interp + vec2(0.0, 0.0) * pix_size) * 0.214607; - color += texture(source_color, uv_interp + vec2(1.0, 0.0) * pix_size) * 0.189879; - color += texture(source_color, uv_interp + vec2(2.0, 0.0) * pix_size) * 0.131514; - color += texture(source_color, uv_interp + vec2(3.0, 0.0) * pix_size) * 0.071303; - color += texture(source_color, uv_interp + vec2(-1.0, 0.0) * pix_size) * 0.189879; - color += texture(source_color, uv_interp + vec2(-2.0, 0.0) * pix_size) * 0.131514; - color += texture(source_color, uv_interp + vec2(-3.0, 0.0) * pix_size) * 0.071303; - frag_color = color; - } else { - vec2 pix_size = blur.pixel_size; - vec4 color = texture(source_color, uv_interp + vec2(0.0, 0.0) * pix_size) * 0.38774; - color += texture(source_color, uv_interp + vec2(0.0, 1.0) * pix_size) * 0.24477; - color += texture(source_color, uv_interp + vec2(0.0, 2.0) * pix_size) * 0.06136; - color += texture(source_color, uv_interp + vec2(0.0, -1.0) * pix_size) * 0.24477; - color += texture(source_color, uv_interp + vec2(0.0, -2.0) * pix_size) * 0.06136; - frag_color = color; - } + // For Gaussian Blur we use 13 taps in a single pass instead of 12 taps over 2 passes. + // This minimizes the number of times we change framebuffers which is very important for mobile. + // Source: http://www.iryoku.com/next-generation-post-processing-in-call-of-duty-advanced-warfare + vec4 A = texture(source_color, uv_interp + blur.pixel_size * vec2(-1.0, -1.0)); + vec4 B = texture(source_color, uv_interp + blur.pixel_size * vec2(0.0, -1.0)); + vec4 C = texture(source_color, uv_interp + blur.pixel_size * vec2(1.0, -1.0)); + vec4 D = texture(source_color, uv_interp + blur.pixel_size * vec2(-0.5, -0.5)); + vec4 E = texture(source_color, uv_interp + blur.pixel_size * vec2(0.5, -0.5)); + vec4 F = texture(source_color, uv_interp + blur.pixel_size * vec2(-1.0, 0.0)); + vec4 G = texture(source_color, uv_interp); + vec4 H = texture(source_color, uv_interp + blur.pixel_size * vec2(1.0, 0.0)); + vec4 I = texture(source_color, uv_interp + blur.pixel_size * vec2(-0.5, 0.5)); + vec4 J = texture(source_color, uv_interp + blur.pixel_size * vec2(0.5, 0.5)); + vec4 K = texture(source_color, uv_interp + blur.pixel_size * vec2(-1.0, 1.0)); + vec4 L = texture(source_color, uv_interp + blur.pixel_size * vec2(0.0, 1.0)); + vec4 M = texture(source_color, uv_interp + blur.pixel_size * vec2(1.0, 1.0)); + + float base_weight = 0.5 / 4.0; + float lesser_weight = 0.125 / 4.0; + + frag_color = (D + E + I + J) * base_weight; + frag_color += (A + B + G + F) * lesser_weight; + frag_color += (B + C + H + G) * lesser_weight; + frag_color += (F + G + L + K) * lesser_weight; + frag_color += (G + H + M + L) * lesser_weight; #endif #ifdef MODE_GAUSSIAN_GLOW @@ -129,7 +130,7 @@ void main() { #ifdef GLOW_USE_AUTO_EXPOSURE - frag_color /= texelFetch(source_auto_exposure, ivec2(0, 0), 0).r / blur.glow_auto_exposure_grey; + frag_color /= texelFetch(source_auto_exposure, ivec2(0, 0), 0).r / blur.glow_auto_exposure_scale; #endif frag_color *= blur.glow_exposure; diff --git a/servers/rendering/renderer_rd/shaders/effects/blur_raster_inc.glsl b/servers/rendering/renderer_rd/shaders/effects/blur_raster_inc.glsl index 730504571a..06ca198f37 100644 --- a/servers/rendering/renderer_rd/shaders/effects/blur_raster_inc.glsl +++ b/servers/rendering/renderer_rd/shaders/effects/blur_raster_inc.glsl @@ -16,7 +16,7 @@ layout(push_constant, std430) uniform Blur { float glow_exposure; // 04 - 36 float glow_white; // 04 - 40 float glow_luminance_cap; // 04 - 44 - float glow_auto_exposure_grey; // 04 - 48 + float glow_auto_exposure_scale; // 04 - 48 float luminance_multiplier; // 04 - 52 float res1; // 04 - 56 diff --git a/servers/rendering/renderer_rd/shaders/effects/bokeh_dof.glsl b/servers/rendering/renderer_rd/shaders/effects/bokeh_dof.glsl index 0438671dd2..fe770ac065 100644 --- a/servers/rendering/renderer_rd/shaders/effects/bokeh_dof.glsl +++ b/servers/rendering/renderer_rd/shaders/effects/bokeh_dof.glsl @@ -30,7 +30,7 @@ layout(set = 1, binding = 0) uniform sampler2D source_bokeh; #ifdef MODE_GEN_BLUR_SIZE float get_depth_at_pos(vec2 uv) { - float depth = textureLod(source_depth, uv, 0.0).x; + float depth = textureLod(source_depth, uv, 0.0).x * 2.0 - 1.0; if (params.orthogonal) { depth = ((depth + (params.z_far + params.z_near) / (params.z_far - params.z_near)) * (params.z_far - params.z_near)) / 2.0; } else { @@ -41,11 +41,25 @@ float get_depth_at_pos(vec2 uv) { float get_blur_size(float depth) { if (params.blur_near_active && depth < params.blur_near_begin) { - return -(1.0 - smoothstep(params.blur_near_end, params.blur_near_begin, depth)) * params.blur_size - DEPTH_GAP; //near blur is negative + if (params.use_physical_near) { + // Physically-based. + float d = abs(params.blur_near_begin - depth); + return -(d / (params.blur_near_begin - d)) * params.blur_size_near - DEPTH_GAP; // Near blur is negative. + } else { + // Non-physically-based. + return -(1.0 - smoothstep(params.blur_near_end, params.blur_near_begin, depth)) * params.blur_size - DEPTH_GAP; // Near blur is negative. + } } if (params.blur_far_active && depth > params.blur_far_begin) { - return smoothstep(params.blur_far_begin, params.blur_far_end, depth) * params.blur_size + DEPTH_GAP; + if (params.use_physical_far) { + // Physically-based. + float d = abs(params.blur_far_begin - depth); + return (d / (params.blur_far_begin + d)) * params.blur_size_far + DEPTH_GAP; + } else { + // Non-physically-based. + return smoothstep(params.blur_far_begin, params.blur_far_end, depth) * params.blur_size + DEPTH_GAP; + } } return 0.0; @@ -172,6 +186,7 @@ void main() { uv += pixel_size * 0.5; //half pixel to read centers vec4 color = texture(color_texture, uv); + float initial_blur = color.a; float accum = 1.0; float radius = params.blur_scale; @@ -179,8 +194,8 @@ void main() { vec2 suv = uv + vec2(cos(ang), sin(ang)) * pixel_size * radius; vec4 sample_color = texture(color_texture, suv); float sample_size = abs(sample_color.a); - if (sample_color.a > color.a) { - sample_size = clamp(sample_size, 0.0, abs(color.a) * 2.0); + if (sample_color.a > initial_blur) { + sample_size = clamp(sample_size, 0.0, abs(initial_blur) * 2.0); } float m = smoothstep(radius - 0.5, radius + 0.5, sample_size); diff --git a/servers/rendering/renderer_rd/shaders/effects/bokeh_dof_inc.glsl b/servers/rendering/renderer_rd/shaders/effects/bokeh_dof_inc.glsl index b90a527554..4a2b0edc18 100644 --- a/servers/rendering/renderer_rd/shaders/effects/bokeh_dof_inc.glsl +++ b/servers/rendering/renderer_rd/shaders/effects/bokeh_dof_inc.glsl @@ -20,6 +20,11 @@ layout(push_constant, std430) uniform Params { bool use_jitter; float jitter_seed; + bool use_physical_near; + bool use_physical_far; + + float blur_size_near; + float blur_size_far; uint pad[2]; } params; diff --git a/servers/rendering/renderer_rd/shaders/effects/bokeh_dof_raster.glsl b/servers/rendering/renderer_rd/shaders/effects/bokeh_dof_raster.glsl index a3b3938ee9..1b487835d2 100644 --- a/servers/rendering/renderer_rd/shaders/effects/bokeh_dof_raster.glsl +++ b/servers/rendering/renderer_rd/shaders/effects/bokeh_dof_raster.glsl @@ -52,7 +52,7 @@ layout(set = 2, binding = 0) uniform sampler2D original_weight; #ifdef MODE_GEN_BLUR_SIZE float get_depth_at_pos(vec2 uv) { - float depth = textureLod(source_depth, uv, 0.0).x; + float depth = textureLod(source_depth, uv, 0.0).x * 2.0 - 1.0; if (params.orthogonal) { depth = ((depth + (params.z_far + params.z_near) / (params.z_far - params.z_near)) * (params.z_far - params.z_near)) / 2.0; } else { @@ -63,11 +63,25 @@ float get_depth_at_pos(vec2 uv) { float get_blur_size(float depth) { if (params.blur_near_active && depth < params.blur_near_begin) { - return -(1.0 - smoothstep(params.blur_near_end, params.blur_near_begin, depth)) * params.blur_size - DEPTH_GAP; //near blur is negative + if (params.use_physical_near) { + // Physically-based. + float d = abs(params.blur_near_begin - depth); + return -(d / (params.blur_near_begin - d)) * params.blur_size_near - DEPTH_GAP; // Near blur is negative. + } else { + // Non-physically-based. + return -(1.0 - smoothstep(params.blur_near_end, params.blur_near_begin, depth)) * params.blur_size - DEPTH_GAP; // Near blur is negative. + } } if (params.blur_far_active && depth > params.blur_far_begin) { - return smoothstep(params.blur_far_begin, params.blur_far_end, depth) * params.blur_size + DEPTH_GAP; + if (params.use_physical_far) { + // Physically-based. + float d = abs(params.blur_far_begin - depth); + return (d / (params.blur_far_begin + d)) * params.blur_size_far + DEPTH_GAP; + } else { + // Non-physically-based. + return smoothstep(params.blur_far_begin, params.blur_far_end, depth) * params.blur_size + DEPTH_GAP; + } } return 0.0; @@ -207,12 +221,9 @@ void main() { vec4 sample_color = texture(source_color, uv_adj); sample_color.a = texture(source_weight, uv_adj).r; - float limit; - - if (sample_color.a < color.a) { - limit = abs(sample_color.a); - } else { - limit = abs(color.a); + float limit = abs(sample_color.a); + if (sample_color.a > color.a) { + limit = clamp(limit, 0.0, abs(color.a) * 2.0); } limit -= DEPTH_GAP; diff --git a/servers/rendering/renderer_rd/shaders/effects/copy.glsl b/servers/rendering/renderer_rd/shaders/effects/copy.glsl index 3a4ef86ef0..5cc2ed7622 100644 --- a/servers/rendering/renderer_rd/shaders/effects/copy.glsl +++ b/servers/rendering/renderer_rd/shaders/effects/copy.glsl @@ -31,7 +31,7 @@ layout(push_constant, std430) uniform Params { float glow_exposure; float glow_white; float glow_luminance_cap; - float glow_auto_exposure_grey; + float glow_auto_exposure_scale; // DOF. float camera_z_far; float camera_z_near; @@ -185,7 +185,7 @@ void main() { if (bool(params.flags & FLAG_GLOW_FIRST_PASS)) { #ifdef GLOW_USE_AUTO_EXPOSURE - color /= texelFetch(source_auto_exposure, ivec2(0, 0), 0).r / params.glow_auto_exposure_grey; + color /= texelFetch(source_auto_exposure, ivec2(0, 0), 0).r / params.glow_auto_exposure_scale; #endif color *= params.glow_exposure; @@ -194,10 +194,10 @@ void main() { color = min(color * feedback, vec4(params.glow_luminance_cap)); } -#endif +#endif // MODE_GLOW imageStore(dest_buffer, pos + params.target, color); -#endif +#endif // MODE_GAUSSIAN_BLUR #ifdef MODE_SIMPLE_COPY @@ -227,7 +227,7 @@ void main() { imageStore(dest_buffer, pos + params.target, color); -#endif +#endif // MODE_SIMPLE_COPY #ifdef MODE_SIMPLE_COPY_DEPTH @@ -239,7 +239,7 @@ void main() { imageStore(dest_buffer, pos + params.target, vec4(color.r)); -#endif +#endif // MODE_SIMPLE_COPY_DEPTH #ifdef MODE_LINEARIZE_DEPTH_COPY @@ -253,7 +253,7 @@ void main() { } imageStore(dest_buffer, pos + params.target, color); -#endif +#endif // MODE_LINEARIZE_DEPTH_COPY #if defined(MODE_CUBEMAP_TO_PANORAMA) || defined(MODE_CUBEMAP_ARRAY_TO_PANORAMA) @@ -276,7 +276,7 @@ void main() { vec4 color = textureLod(source_color, vec4(normal, params.camera_z_far), 0.0); //the biggest the lod the least the acne #endif imageStore(dest_buffer, pos + params.target, color); -#endif +#endif // defined(MODE_CUBEMAP_TO_PANORAMA) || defined(MODE_CUBEMAP_ARRAY_TO_PANORAMA) #ifdef MODE_SET_COLOR imageStore(dest_buffer, pos + params.target, params.set_color); diff --git a/servers/rendering/renderer_rd/shaders/effects/copy_to_fb.glsl b/servers/rendering/renderer_rd/shaders/effects/copy_to_fb.glsl index 1c17eabb56..46bb99794d 100644 --- a/servers/rendering/renderer_rd/shaders/effects/copy_to_fb.glsl +++ b/servers/rendering/renderer_rd/shaders/effects/copy_to_fb.glsl @@ -13,6 +13,14 @@ #endif // has_VK_KHR_multiview #endif //MULTIVIEW +#define FLAG_FLIP_Y (1 << 0) +#define FLAG_USE_SECTION (1 << 1) +#define FLAG_FORCE_LUMINANCE (1 << 2) +#define FLAG_ALPHA_TO_ZERO (1 << 3) +#define FLAG_SRGB (1 << 4) +#define FLAG_ALPHA_TO_ONE (1 << 5) +#define FLAG_LINEAR (1 << 6) + #ifdef MULTIVIEW layout(location = 0) out vec3 uv_interp; #else @@ -22,11 +30,10 @@ layout(location = 0) out vec2 uv_interp; layout(push_constant, std430) uniform Params { vec4 section; vec2 pixel_size; - bool flip_y; - bool use_section; + float luminance_multiplier; + uint flags; - bool force_luminance; - uint pad[3]; + vec4 color; } params; @@ -37,13 +44,13 @@ void main() { uv_interp.z = ViewIndex; #endif vec2 vpos = uv_interp.xy; - if (params.use_section) { + if (bool(params.flags & FLAG_USE_SECTION)) { vpos = params.section.xy + vpos * params.section.zw; } gl_Position = vec4(vpos * 2.0 - 1.0, 0.0, 1.0); - if (params.flip_y) { + if (bool(params.flags & FLAG_FLIP_Y)) { uv_interp.y = 1.0 - uv_interp.y; } } @@ -63,16 +70,21 @@ void main() { #endif // has_VK_KHR_multiview #endif //MULTIVIEW +#define FLAG_FLIP_Y (1 << 0) +#define FLAG_USE_SECTION (1 << 1) +#define FLAG_FORCE_LUMINANCE (1 << 2) +#define FLAG_ALPHA_TO_ZERO (1 << 3) +#define FLAG_SRGB (1 << 4) +#define FLAG_ALPHA_TO_ONE (1 << 5) +#define FLAG_LINEAR (1 << 6) + layout(push_constant, std430) uniform Params { vec4 section; vec2 pixel_size; - bool flip_y; - bool use_section; + float luminance_multiplier; + uint flags; - bool force_luminance; - bool alpha_to_zero; - bool srgb; - uint pad; + vec4 color; } params; @@ -104,7 +116,15 @@ vec3 linear_to_srgb(vec3 color) { return mix((vec3(1.0f) + a) * pow(color.rgb, vec3(1.0f / 2.4f)) - a, 12.92f * color.rgb, lessThan(color.rgb, vec3(0.0031308f))); } +vec3 srgb_to_linear(vec3 color) { + return mix(pow((color.rgb + vec3(0.055)) * (1.0 / (1.0 + 0.055)), vec3(2.4)), color.rgb * (1.0 / 12.92), lessThan(color.rgb, vec3(0.04045))); +} + void main() { +#ifdef MODE_SET_COLOR + frag_color = params.color; +#else + #ifdef MULTIVIEW vec3 uv = uv_interp; #else @@ -155,15 +175,22 @@ void main() { #endif /* MODE_TWO_SOURCES */ #endif /* MULTIVIEW */ - if (params.force_luminance) { + if (bool(params.flags & FLAG_FORCE_LUMINANCE)) { color.rgb = vec3(max(max(color.r, color.g), color.b)); } - if (params.alpha_to_zero) { + if (bool(params.flags & FLAG_ALPHA_TO_ZERO)) { color.rgb *= color.a; } - if (params.srgb) { + if (bool(params.flags & FLAG_SRGB)) { color.rgb = linear_to_srgb(color.rgb); } + if (bool(params.flags & FLAG_ALPHA_TO_ONE)) { + color.a = 1.0; + } + if (bool(params.flags & FLAG_LINEAR)) { + color.rgb = srgb_to_linear(color.rgb); + } - frag_color = color; + frag_color = color / params.luminance_multiplier; +#endif // MODE_SET_COLOR } diff --git a/servers/rendering/renderer_rd/shaders/fsr_upscale.glsl b/servers/rendering/renderer_rd/shaders/effects/fsr_upscale.glsl index c8eb78a2f0..c8eb78a2f0 100644 --- a/servers/rendering/renderer_rd/shaders/fsr_upscale.glsl +++ b/servers/rendering/renderer_rd/shaders/effects/fsr_upscale.glsl diff --git a/servers/rendering/renderer_rd/shaders/effects/screen_space_reflection.glsl b/servers/rendering/renderer_rd/shaders/effects/screen_space_reflection.glsl index d85ab3af2e..9d8294a7da 100644 --- a/servers/rendering/renderer_rd/shaders/effects/screen_space_reflection.glsl +++ b/servers/rendering/renderer_rd/shaders/effects/screen_space_reflection.glsl @@ -30,12 +30,7 @@ layout(push_constant, std430) uniform Params { bool orthogonal; float filter_mipmap_levels; bool use_half_res; - uint metallic_mask; - uint view_index; - uint pad1; - uint pad2; - uint pad3; } params; @@ -126,7 +121,7 @@ void main() { // clip z and w advance to line advance vec2 line_advance = normalize(line_dir); // down to pixel - float step_size = length(line_advance) / length(line_dir); + float step_size = 1.0 / length(line_dir); float z_advance = z_dir * step_size; // adapt z advance to line advance float w_advance = w_dir * step_size; // adapt w advance to line advance @@ -144,6 +139,14 @@ void main() { float depth; vec2 prev_pos = pos; + if (ivec2(pos + line_advance - 0.5) == ssC) { + // It is possible for rounding to cause our first pixel to check to be the pixel we're reflecting. + // Make sure we skip it + pos += line_advance; + z += z_advance; + w += w_advance; + } + bool found = false; float steps_taken = 0.0; @@ -154,8 +157,8 @@ void main() { w += w_advance; // convert to linear depth - - depth = imageLoad(source_depth, ivec2(pos - 0.5)).r; + ivec2 test_pos = ivec2(pos - 0.5); + depth = imageLoad(source_depth, test_pos).r; if (sc_multiview) { depth = depth * 2.0 - 1.0; depth = 2.0 * params.camera_z_near * params.camera_z_far / (params.camera_z_far + params.camera_z_near - depth * (params.camera_z_far - params.camera_z_near)); @@ -166,13 +169,21 @@ void main() { z_to = z / w; if (depth > z_to) { - // if depth was surpassed - if (depth <= max(z_to, z_from) + params.depth_tolerance && -depth < params.camera_z_far) { - // check the depth tolerance and far clip - // check that normal is valid - found = true; + // Test if our ray is hitting the "right" side of the surface, if not we're likely self reflecting and should skip. + vec4 test_normal_roughness = imageLoad(source_normal_roughness, test_pos); + vec3 test_normal = test_normal_roughness.xyz * 2.0 - 1.0; + test_normal = normalize(test_normal); + test_normal.y = -test_normal.y; //because this code reads flipped + + if (dot(ray_dir, test_normal) < 0.001) { + // if depth was surpassed + if (depth <= max(z_to, z_from) + params.depth_tolerance && -depth < params.camera_z_far * 0.95) { + // check the depth tolerance and far clip + // check that normal is valid + found = true; + } + break; } - break; } steps_taken += 1.0; @@ -182,17 +193,18 @@ void main() { if (found) { float margin_blend = 1.0; - vec2 margin = vec2((params.screen_size.x + params.screen_size.y) * 0.5 * 0.05); // make a uniform margin - if (any(bvec4(lessThan(pos, -margin), greaterThan(pos, params.screen_size + margin)))) { - // clip outside screen + margin + vec2 margin = vec2((params.screen_size.x + params.screen_size.y) * 0.05); // make a uniform margin + if (any(bvec4(lessThan(pos, vec2(0.0, 0.0)), greaterThan(pos, params.screen_size)))) { + // clip at the screen edges imageStore(ssr_image, ssC, vec4(0.0)); return; } { - //blend fading out towards external margin - vec2 margin_grad = mix(pos - params.screen_size, -pos, lessThan(pos, vec2(0.0))); - margin_blend = 1.0 - smoothstep(0.0, margin.x, max(margin_grad.x, margin_grad.y)); + //blend fading out towards inner margin + // 0.5 = midpoint of reflection + vec2 margin_grad = mix(params.screen_size - pos, pos, lessThan(pos, params.screen_size * 0.5)); + margin_blend = smoothstep(0.0, margin.x * margin.y, margin_grad.x * margin_grad.y); //margin_blend = 1.0; } @@ -230,18 +242,20 @@ void main() { } } - // Isn't this going to be overwritten after our endif? - final_color = imageLoad(source_diffuse, ivec2((final_pos - 0.5) * pixel_size)); - imageStore(blur_radius_image, ssC, vec4(blur_radius / 255.0)); //stored in r8 #endif // MODE_ROUGH final_color = vec4(imageLoad(source_diffuse, ivec2(final_pos - 0.5)).rgb, fade * margin_blend); - //change blend by metallic - vec4 metallic_mask = unpackUnorm4x8(params.metallic_mask); - final_color.a *= dot(metallic_mask, texelFetch(source_metallic, ssC << 1, 0)); + // Schlick term. + float metallic = texelFetch(source_metallic, ssC << 1, 0).w; + float f0 = mix(0.04, 1.0, metallic); // Assume a "specular" amount of 0.5 + normal.y = -normal.y; + float m = clamp(1.0 - dot(normalize(normal), -view_dir), 0.0, 1.0); + float m2 = m * m; + m = m2 * m2 * m; // pow(m,5) + final_color.a *= f0 + (1.0 - f0) * m; // Fresnel Schlick term. imageStore(ssr_image, ssC, final_color); diff --git a/servers/rendering/renderer_rd/shaders/subsurface_scattering.glsl b/servers/rendering/renderer_rd/shaders/effects/subsurface_scattering.glsl index fb35d3cde6..fb35d3cde6 100644 --- a/servers/rendering/renderer_rd/shaders/subsurface_scattering.glsl +++ b/servers/rendering/renderer_rd/shaders/effects/subsurface_scattering.glsl diff --git a/servers/rendering/renderer_rd/shaders/taa_resolve.glsl b/servers/rendering/renderer_rd/shaders/effects/taa_resolve.glsl index b0a0839836..02566d8e35 100644 --- a/servers/rendering/renderer_rd/shaders/taa_resolve.glsl +++ b/servers/rendering/renderer_rd/shaders/effects/taa_resolve.glsl @@ -32,7 +32,9 @@ // Based on Spartan Engine's TAA implementation (without TAA upscale). // <https://github.com/PanosK92/SpartanEngine/blob/a8338d0609b85dc32f3732a5c27fb4463816a3b9/Data/shaders/temporal_antialiasing.hlsl> +#ifndef MOLTENVK_USED #define USE_SUBGROUPS +#endif // MOLTENVK_USED #define GROUP_SIZE 8 #define FLT_MIN 0.00000001 diff --git a/servers/rendering/renderer_rd/shaders/effects/tonemap.glsl b/servers/rendering/renderer_rd/shaders/effects/tonemap.glsl index 62a7b0e7d7..52aee8b648 100644 --- a/servers/rendering/renderer_rd/shaders/effects/tonemap.glsl +++ b/servers/rendering/renderer_rd/shaders/effects/tonemap.glsl @@ -75,7 +75,7 @@ layout(push_constant, std430) uniform Params { float exposure; float white; - float auto_exposure_grey; + float auto_exposure_scale; float luminance_multiplier; vec2 pixel_size; @@ -360,15 +360,15 @@ vec3 do_fxaa(vec3 color, float exposure, vec2 uv_interp) { const float FXAA_SPAN_MAX = 8.0; #ifdef MULTIVIEW - vec3 rgbNW = textureLod(source_color, vec3(uv_interp + vec2(-1.0, -1.0) * params.pixel_size, ViewIndex), 0.0).xyz * exposure * params.luminance_multiplier; - vec3 rgbNE = textureLod(source_color, vec3(uv_interp + vec2(1.0, -1.0) * params.pixel_size, ViewIndex), 0.0).xyz * exposure * params.luminance_multiplier; - vec3 rgbSW = textureLod(source_color, vec3(uv_interp + vec2(-1.0, 1.0) * params.pixel_size, ViewIndex), 0.0).xyz * exposure * params.luminance_multiplier; - vec3 rgbSE = textureLod(source_color, vec3(uv_interp + vec2(1.0, 1.0) * params.pixel_size, ViewIndex), 0.0).xyz * exposure * params.luminance_multiplier; + vec3 rgbNW = textureLod(source_color, vec3(uv_interp + vec2(-0.5, -0.5) * params.pixel_size, ViewIndex), 0.0).xyz * exposure * params.luminance_multiplier; + vec3 rgbNE = textureLod(source_color, vec3(uv_interp + vec2(0.5, -0.5) * params.pixel_size, ViewIndex), 0.0).xyz * exposure * params.luminance_multiplier; + vec3 rgbSW = textureLod(source_color, vec3(uv_interp + vec2(-0.5, 0.5) * params.pixel_size, ViewIndex), 0.0).xyz * exposure * params.luminance_multiplier; + vec3 rgbSE = textureLod(source_color, vec3(uv_interp + vec2(0.5, 0.5) * params.pixel_size, ViewIndex), 0.0).xyz * exposure * params.luminance_multiplier; #else - vec3 rgbNW = textureLod(source_color, uv_interp + vec2(-1.0, -1.0) * params.pixel_size, 0.0).xyz * exposure * params.luminance_multiplier; - vec3 rgbNE = textureLod(source_color, uv_interp + vec2(1.0, -1.0) * params.pixel_size, 0.0).xyz * exposure * params.luminance_multiplier; - vec3 rgbSW = textureLod(source_color, uv_interp + vec2(-1.0, 1.0) * params.pixel_size, 0.0).xyz * exposure * params.luminance_multiplier; - vec3 rgbSE = textureLod(source_color, uv_interp + vec2(1.0, 1.0) * params.pixel_size, 0.0).xyz * exposure * params.luminance_multiplier; + vec3 rgbNW = textureLod(source_color, uv_interp + vec2(-0.5, -0.5) * params.pixel_size, 0.0).xyz * exposure * params.luminance_multiplier; + vec3 rgbNE = textureLod(source_color, uv_interp + vec2(0.5, -0.5) * params.pixel_size, 0.0).xyz * exposure * params.luminance_multiplier; + vec3 rgbSW = textureLod(source_color, uv_interp + vec2(-0.5, 0.5) * params.pixel_size, 0.0).xyz * exposure * params.luminance_multiplier; + vec3 rgbSE = textureLod(source_color, uv_interp + vec2(0.5, 0.5) * params.pixel_size, 0.0).xyz * exposure * params.luminance_multiplier; #endif vec3 rgbM = color; vec3 luma = vec3(0.299, 0.587, 0.114); @@ -440,7 +440,7 @@ void main() { #ifndef SUBPASS if (params.use_auto_exposure) { - exposure *= 1.0 / (texelFetch(source_auto_exposure, ivec2(0, 0), 0).r * params.luminance_multiplier / params.auto_exposure_grey); + exposure *= 1.0 / (texelFetch(source_auto_exposure, ivec2(0, 0), 0).r * params.luminance_multiplier / params.auto_exposure_scale); } #endif @@ -462,12 +462,6 @@ void main() { } #endif - if (params.use_debanding) { - // For best results, debanding should be done before tonemapping. - // Otherwise, we're adding noise to an already-quantized image. - color.rgb += screen_space_dither(gl_FragCoord.xy); - } - color.rgb = apply_tonemapping(color.rgb, params.white); color.rgb = linear_to_srgb(color.rgb); // regular linear -> SRGB conversion @@ -498,5 +492,11 @@ void main() { color.rgb = apply_color_correction(color.rgb); } + if (params.use_debanding) { + // Debanding should be done at the end of tonemapping, but before writing to the LDR buffer. + // Otherwise, we're adding noise to an already-quantized image. + color.rgb += screen_space_dither(gl_FragCoord.xy); + } + frag_color = color; } diff --git a/servers/rendering/renderer_rd/shaders/effects/vrs.glsl b/servers/rendering/renderer_rd/shaders/effects/vrs.glsl index 5ef83c0b44..b450bb9fe9 100644 --- a/servers/rendering/renderer_rd/shaders/effects/vrs.glsl +++ b/servers/rendering/renderer_rd/shaders/effects/vrs.glsl @@ -63,10 +63,18 @@ void main() { #ifdef MULTIVIEW vec4 color = textureLod(source_color, uv, 0.0); + frag_color = uint(color.r * 255.0); #else /* MULTIVIEW */ vec4 color = textureLod(source_color, uv, 0.0); -#endif /* MULTIVIEW */ - // See if we can change the sampler to one that returns int... - frag_color = uint(color.r * 256.0); + // for user supplied VRS map we do a color mapping + color.r *= 3.0; + frag_color = int(color.r) << 2; + + color.g *= 3.0; + frag_color += int(color.g); + + // note 1x4, 4x1, 1x8, 8x1, 2x8 and 8x2 are not supported + // 4x8, 8x4 and 8x8 are only available on some GPUs +#endif /* MULTIVIEW */ } diff --git a/servers/rendering/renderer_rd/shaders/environment/SCsub b/servers/rendering/renderer_rd/shaders/environment/SCsub index 741da8fe69..f06a2d86e2 100644 --- a/servers/rendering/renderer_rd/shaders/environment/SCsub +++ b/servers/rendering/renderer_rd/shaders/environment/SCsub @@ -4,7 +4,7 @@ Import("env") if "RD_GLSL" in env["BUILDERS"]: # find all include files - gl_include_files = [str(f) for f in Glob("*_inc.glsl")] + gl_include_files = [str(f) for f in Glob("*_inc.glsl")] + [str(f) for f in Glob("../*_inc.glsl")] # find all shader code(all glsl files excluding our include files) glsl_files = [str(f) for f in Glob("*.glsl") if str(f) not in gl_include_files] diff --git a/servers/rendering/renderer_rd/shaders/environment/gi.glsl b/servers/rendering/renderer_rd/shaders/environment/gi.glsl index 6ea8cb1377..ab927df678 100644 --- a/servers/rendering/renderer_rd/shaders/environment/gi.glsl +++ b/servers/rendering/renderer_rd/shaders/environment/gi.glsl @@ -32,6 +32,8 @@ struct ProbeCascadeData { float to_probe; ivec3 probe_world_offset; float to_cell; // 1/bounds * grid_size + vec3 pad; + float exposure_normalization; }; layout(rgba16f, set = 0, binding = 9) uniform restrict writeonly image2D ambient_buffer; @@ -83,6 +85,9 @@ struct VoxelGIData { float normal_bias; // 4 - 88 bool blend_ambient; // 4 - 92 uint mipmaps; // 4 - 96 + + vec3 pad; // 12 - 108 + float exposure_normalization; // 4 - 112 }; layout(set = 0, binding = 16, std140) uniform VoxelGIs { @@ -241,7 +246,7 @@ void sdfvoxel_gi_process(uint cascade, vec3 cascade_pos, vec3 cam_pos, vec3 cam_ pos_uvw.x += float(offset.z) * sdfgi.lightprobe_uv_offset.z; diffuse = textureLod(sampler2DArray(lightprobe_texture, linear_sampler), pos_uvw, 0.0).rgb; - diffuse_accum += vec4(diffuse * weight, weight); + diffuse_accum += vec4(diffuse * weight * sdfgi.cascades[cascade].exposure_normalization, weight); { vec3 specular = vec3(0.0); @@ -255,7 +260,7 @@ void sdfvoxel_gi_process(uint cascade, vec3 cascade_pos, vec3 cam_pos, vec3 cam_ specular = mix(specular, textureLod(sampler2DArray(lightprobe_texture, linear_sampler), pos_uvw, 0.0).rgb, (roughness - 0.2) * 1.25); } - specular_accum += specular * weight; + specular_accum += specular * weight * sdfgi.cascades[cascade].exposure_normalization; } } @@ -574,7 +579,7 @@ void voxel_gi_compute(uint index, vec3 position, vec3 normal, vec3 ref_vec, mat3 } } - light.rgb *= voxel_gi_instances.data[index].dynamic_range; + light.rgb *= voxel_gi_instances.data[index].dynamic_range * voxel_gi_instances.data[index].exposure_normalization; if (!voxel_gi_instances.data[index].blend_ambient) { light.a = 1.0; } @@ -583,7 +588,7 @@ void voxel_gi_compute(uint index, vec3 position, vec3 normal, vec3 ref_vec, mat3 //radiance vec4 irr_light = voxel_cone_trace(voxel_gi_textures[index], cell_size, position, ref_vec, tan(roughness * 0.5 * M_PI * 0.99), max_distance, voxel_gi_instances.data[index].bias); - irr_light.rgb *= voxel_gi_instances.data[index].dynamic_range; + irr_light.rgb *= voxel_gi_instances.data[index].dynamic_range * voxel_gi_instances.data[index].exposure_normalization; if (!voxel_gi_instances.data[index].blend_ambient) { irr_light.a = 1.0; } diff --git a/servers/rendering/renderer_rd/shaders/environment/sdfgi_debug.glsl b/servers/rendering/renderer_rd/shaders/environment/sdfgi_debug.glsl index af5f7d0a58..177dab16c7 100644 --- a/servers/rendering/renderer_rd/shaders/environment/sdfgi_debug.glsl +++ b/servers/rendering/renderer_rd/shaders/environment/sdfgi_debug.glsl @@ -21,6 +21,7 @@ struct CascadeData { float to_cell; // 1/bounds * grid_size ivec3 probe_world_offset; uint pad; + vec4 pad2; }; layout(set = 0, binding = 9, std140) uniform Cascades { @@ -37,16 +38,14 @@ layout(push_constant, std430) uniform Params { uint max_cascades; ivec2 screen_size; - bool use_occlusion; float y_mult; - int probe_axis_size; float z_near; - float reserved1; - float reserved2; - mat4 cam_transform; - mat4 inv_projection; + mat3x4 inv_projection; + // We pack these more tightly than mat3 and vec3, which will require some reconstruction trickery. + float cam_basis[3][3]; + float cam_origin[3]; } params; @@ -82,13 +81,21 @@ void main() { vec3 ray_pos; vec3 ray_dir; { - ray_pos = params.cam_transform[3].xyz; + ray_pos = vec3(params.cam_origin[0], params.cam_origin[1], params.cam_origin[2]); 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); + ray_dir = (vec4(ray_dir, 1.0) * mat4(params.inv_projection)).xyz; + + mat3 cam_basis; + { + vec3 c0 = vec3(params.cam_basis[0][0], params.cam_basis[0][1], params.cam_basis[0][2]); + vec3 c1 = vec3(params.cam_basis[1][0], params.cam_basis[1][1], params.cam_basis[1][2]); + vec3 c2 = vec3(params.cam_basis[2][0], params.cam_basis[2][1], params.cam_basis[2][2]); + cam_basis = mat3(c0, c1, c2); + } + ray_dir = normalize(cam_basis * ray_dir); } ray_pos.y *= params.y_mult; diff --git a/servers/rendering/renderer_rd/shaders/environment/sdfgi_debug_probes.glsl b/servers/rendering/renderer_rd/shaders/environment/sdfgi_debug_probes.glsl index 75b1ad2130..a0ef169f03 100644 --- a/servers/rendering/renderer_rd/shaders/environment/sdfgi_debug_probes.glsl +++ b/servers/rendering/renderer_rd/shaders/environment/sdfgi_debug_probes.glsl @@ -73,6 +73,7 @@ struct CascadeData { float to_cell; // 1/bounds * grid_size ivec3 probe_world_offset; uint pad; + vec4 pad2; }; layout(set = 0, binding = 1, std140) uniform Cascades { diff --git a/servers/rendering/renderer_rd/shaders/environment/sdfgi_direct_light.glsl b/servers/rendering/renderer_rd/shaders/environment/sdfgi_direct_light.glsl index b95fad650e..9f7449b8aa 100644 --- a/servers/rendering/renderer_rd/shaders/environment/sdfgi_direct_light.glsl +++ b/servers/rendering/renderer_rd/shaders/environment/sdfgi_direct_light.glsl @@ -45,6 +45,7 @@ struct CascadeData { float to_cell; // 1/bounds * grid_size ivec3 probe_world_offset; uint pad; + vec4 pad2; }; layout(set = 0, binding = 8, std140) uniform Cascades { diff --git a/servers/rendering/renderer_rd/shaders/environment/sdfgi_integrate.glsl b/servers/rendering/renderer_rd/shaders/environment/sdfgi_integrate.glsl index 9c03297f5c..4bdb0dcc72 100644 --- a/servers/rendering/renderer_rd/shaders/environment/sdfgi_integrate.glsl +++ b/servers/rendering/renderer_rd/shaders/environment/sdfgi_integrate.glsl @@ -20,6 +20,7 @@ struct CascadeData { float to_cell; // 1/bounds * grid_size ivec3 probe_world_offset; uint pad; + vec4 pad2; }; layout(set = 0, binding = 7, std140) uniform Cascades { diff --git a/servers/rendering/renderer_rd/shaders/environment/sky.glsl b/servers/rendering/renderer_rd/shaders/environment/sky.glsl index 5b4594da99..d523461600 100644 --- a/servers/rendering/renderer_rd/shaders/environment/sky.glsl +++ b/servers/rendering/renderer_rd/shaders/environment/sky.glsl @@ -15,10 +15,10 @@ layout(location = 0) out vec2 uv_interp; layout(push_constant, std430) uniform Params { mat3 orientation; vec4 projections[MAX_VIEWS]; - vec4 position_multiplier; + vec3 position; float time; + vec3 pad; float luminance_multiplier; - float pad[2]; } params; @@ -55,10 +55,10 @@ layout(location = 0) in vec2 uv_interp; layout(push_constant, std430) uniform Params { mat3 orientation; vec4 projections[MAX_VIEWS]; - vec4 position_multiplier; + vec3 position; float time; + vec3 pad; float luminance_multiplier; - float pad[2]; } params; @@ -77,26 +77,29 @@ params; layout(set = 0, binding = 0) uniform sampler material_samplers[12]; -layout(set = 0, binding = 1, std430) restrict readonly buffer GlobalVariableData { +layout(set = 0, binding = 1, std430) restrict readonly buffer GlobalShaderUniformData { vec4 data[]; } -global_variables; +global_shader_uniforms; layout(set = 0, binding = 2, std140) uniform SceneData { - bool volumetric_fog_enabled; - float volumetric_fog_inv_length; - float volumetric_fog_detail_spread; - - float fog_aerial_perspective; - - vec3 fog_light_color; - float fog_sun_scatter; - - bool fog_enabled; - float fog_density; - - float z_far; - uint directional_light_count; + bool volumetric_fog_enabled; // 4 - 4 + float volumetric_fog_inv_length; // 4 - 8 + float volumetric_fog_detail_spread; // 4 - 12 + float volumetric_fog_sky_affect; // 4 - 16 + + bool fog_enabled; // 4 - 20 + float fog_sky_affect; // 4 - 24 + float fog_density; // 4 - 28 + float fog_sun_scatter; // 4 - 32 + + vec3 fog_light_color; // 12 - 44 + float fog_aerial_perspective; // 4 - 48 + + float z_far; // 4 - 52 + uint directional_light_count; // 4 - 56 + uint pad1; // 4 - 60 + uint pad2; // 4 - 64 } scene_data; @@ -150,6 +153,15 @@ layout(set = 3, binding = 0) uniform texture3D volumetric_fog_texture; layout(location = 0) out vec4 frag_color; +#ifdef USE_DEBANDING +// https://www.iryoku.com/next-generation-post-processing-in-call-of-duty-advanced-warfare +vec3 interleaved_gradient_noise(vec2 pos) { + const vec3 magic = vec3(0.06711056f, 0.00583715f, 52.9829189f); + float res = fract(magic.z * fract(dot(pos, magic.xy))) * 2.0 - 1.0; + return vec3(res, -res, res) / 255.0; +} +#endif + vec4 volumetric_fog_process(vec2 screen_uv) { vec3 fog_pos = vec3(screen_uv, 1.0); @@ -169,9 +181,7 @@ vec4 fog_process(vec3 view, vec3 sky_color) { } } - float fog_amount = clamp(1.0 - exp(-scene_data.z_far * scene_data.fog_density), 0.0, 1.0); - - return vec4(fog_color, fog_amount); + return vec4(fog_color, 1.0); } void main() { @@ -200,17 +210,17 @@ void main() { #ifdef USE_CUBEMAP_PASS #ifdef USES_HALF_RES_COLOR - half_res_color = texture(samplerCube(half_res, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), 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]), 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 - half_res_color = textureLod(sampler2D(half_res, material_samplers[SAMPLER_LINEAR_CLAMP]), uv, 0.0) * params.luminance_multiplier; + half_res_color = textureLod(sampler2D(half_res, material_samplers[SAMPLER_LINEAR_CLAMP]), uv, 0.0) / params.luminance_multiplier; #endif #ifdef USES_QUARTER_RES_COLOR - quarter_res_color = textureLod(sampler2D(quarter_res, material_samplers[SAMPLER_LINEAR_CLAMP]), uv, 0.0) * params.luminance_multiplier; + quarter_res_color = textureLod(sampler2D(quarter_res, material_samplers[SAMPLER_LINEAR_CLAMP]), uv, 0.0) / params.luminance_multiplier; #endif #endif @@ -220,7 +230,7 @@ void main() { } - frag_color.rgb = color * params.position_multiplier.w; + frag_color.rgb = color; frag_color.a = alpha; #if !defined(DISABLE_FOG) && !defined(USE_CUBEMAP_PASS) @@ -228,12 +238,12 @@ void main() { // Draw "fixed" fog before volumetric fog to ensure volumetric fog can appear in front of the sky. if (scene_data.fog_enabled) { vec4 fog = fog_process(cube_normal, frag_color.rgb); - frag_color.rgb = mix(frag_color.rgb, fog.rgb, fog.a); + frag_color.rgb = mix(frag_color.rgb, fog.rgb, fog.a * scene_data.fog_sky_affect); } if (scene_data.volumetric_fog_enabled) { vec4 fog = volumetric_fog_process(uv); - frag_color.rgb = mix(frag_color.rgb, fog.rgb, fog.a); + frag_color.rgb = mix(frag_color.rgb, fog.rgb, fog.a * scene_data.volumetric_fog_sky_affect); } if (custom_fog.a > 0.0) { @@ -242,12 +252,17 @@ void main() { #endif // DISABLE_FOG - // Blending is disabled for Sky, so alpha doesn't blend - // alpha is used for subsurface scattering so make sure it doesn't get applied to Sky + // Blending is disabled for Sky, so alpha doesn't blend. + // Alpha is used for subsurface scattering so make sure it doesn't get applied to Sky. if (!AT_CUBEMAP_PASS && !AT_HALF_RES_PASS && !AT_QUARTER_RES_PASS) { frag_color.a = 0.0; } - // For mobile renderer we're dividing by 2.0 as we're using a UNORM buffer - frag_color.rgb = frag_color.rgb / params.luminance_multiplier; + // For mobile renderer we're multiplying by 0.5 as we're using a UNORM buffer. + // For both mobile and clustered, we also bake in the exposure value for the environment and camera. + frag_color.rgb = frag_color.rgb * params.luminance_multiplier; + +#ifdef USE_DEBANDING + frag_color.rgb += interleaved_gradient_noise(gl_FragCoord.xy); +#endif } diff --git a/servers/rendering/renderer_rd/shaders/environment/volumetric_fog.glsl b/servers/rendering/renderer_rd/shaders/environment/volumetric_fog.glsl index fb3c725b1f..4658afd02d 100644 --- a/servers/rendering/renderer_rd/shaders/environment/volumetric_fog.glsl +++ b/servers/rendering/renderer_rd/shaders/environment/volumetric_fog.glsl @@ -28,10 +28,10 @@ layout(local_size_x = 4, local_size_y = 4, local_size_z = 4) in; layout(set = 0, binding = 1) uniform sampler material_samplers[12]; -layout(set = 0, binding = 2, std430) restrict readonly buffer GlobalVariableData { +layout(set = 0, binding = 2, std430) restrict readonly buffer GlobalShaderUniformData { vec4 data[]; } -global_variables; +global_shader_uniforms; layout(push_constant, std430) uniform Params { vec3 position; diff --git a/servers/rendering/renderer_rd/shaders/environment/volumetric_fog_process.glsl b/servers/rendering/renderer_rd/shaders/environment/volumetric_fog_process.glsl index e74cfad65c..28507e6c12 100644 --- a/servers/rendering/renderer_rd/shaders/environment/volumetric_fog_process.glsl +++ b/servers/rendering/renderer_rd/shaders/environment/volumetric_fog_process.glsl @@ -84,6 +84,9 @@ struct VoxelGIData { float normal_bias; // 4 - 88 bool blend_ambient; // 4 - 92 uint mipmaps; // 4 - 96 + + vec3 pad; // 12 - 108 + float exposure_normalization; // 4 - 112 }; layout(set = 0, binding = 11, std140) uniform VoxelGIs { @@ -105,6 +108,8 @@ struct SDFVoxelGICascadeData { float to_probe; ivec3 probe_world_offset; float to_cell; // 1/bounds * grid_size + vec3 pad; + float exposure_normalization; }; layout(set = 1, binding = 0, std140) uniform SDFGI { @@ -270,6 +275,9 @@ const vec3 halton_map[TEMPORAL_FRAMES] = vec3[]( vec3(0.9375, 0.25925926, 0.12), vec3(0.03125, 0.59259259, 0.32)); +// Higher values will make light in volumetric fog fade out sooner when it's occluded by shadow. +const float INV_FOG_FADE = 10.0; + void main() { vec3 fog_cell_size = 1.0 / vec3(params.fog_volume_size); @@ -373,48 +381,50 @@ void main() { float cell_depth_size = abs(view_pos.z - get_depth_at_pos(fog_cell_size.z, pos.z + 1)); //compute directional lights - if (total_density > 0.001) { + if (total_density > 0.00005) { for (uint i = 0; i < params.directional_light_count; i++) { - vec3 shadow_attenuation = vec3(1.0); - - if (directional_lights.data[i].shadow_enabled) { - float depth_z = -view_pos.z; - - vec4 pssm_coord; - vec3 light_dir = directional_lights.data[i].direction; - vec4 v = vec4(view_pos, 1.0); - float z_range; - - if (depth_z < directional_lights.data[i].shadow_split_offsets.x) { - pssm_coord = (directional_lights.data[i].shadow_matrix1 * v); - pssm_coord /= pssm_coord.w; - z_range = directional_lights.data[i].shadow_z_range.x; - - } else if (depth_z < directional_lights.data[i].shadow_split_offsets.y) { - pssm_coord = (directional_lights.data[i].shadow_matrix2 * v); - pssm_coord /= pssm_coord.w; - z_range = directional_lights.data[i].shadow_z_range.y; - - } else if (depth_z < directional_lights.data[i].shadow_split_offsets.z) { - pssm_coord = (directional_lights.data[i].shadow_matrix3 * v); - pssm_coord /= pssm_coord.w; - z_range = directional_lights.data[i].shadow_z_range.z; - - } else { - pssm_coord = (directional_lights.data[i].shadow_matrix4 * v); - pssm_coord /= pssm_coord.w; - z_range = directional_lights.data[i].shadow_z_range.w; - } + if (directional_lights.data[i].volumetric_fog_energy > 0.001) { + vec3 shadow_attenuation = vec3(1.0); + + if (directional_lights.data[i].shadow_opacity > 0.001) { + float depth_z = -view_pos.z; + + vec4 pssm_coord; + vec3 light_dir = directional_lights.data[i].direction; + vec4 v = vec4(view_pos, 1.0); + float z_range; + + if (depth_z < directional_lights.data[i].shadow_split_offsets.x) { + pssm_coord = (directional_lights.data[i].shadow_matrix1 * v); + pssm_coord /= pssm_coord.w; + z_range = directional_lights.data[i].shadow_z_range.x; + + } else if (depth_z < directional_lights.data[i].shadow_split_offsets.y) { + pssm_coord = (directional_lights.data[i].shadow_matrix2 * v); + pssm_coord /= pssm_coord.w; + z_range = directional_lights.data[i].shadow_z_range.y; + + } else if (depth_z < directional_lights.data[i].shadow_split_offsets.z) { + pssm_coord = (directional_lights.data[i].shadow_matrix3 * v); + pssm_coord /= pssm_coord.w; + z_range = directional_lights.data[i].shadow_z_range.z; + + } else { + pssm_coord = (directional_lights.data[i].shadow_matrix4 * v); + pssm_coord /= pssm_coord.w; + z_range = directional_lights.data[i].shadow_z_range.w; + } - float depth = texture(sampler2D(directional_shadow_atlas, linear_sampler), pssm_coord.xy).r; - float shadow = exp(min(0.0, (depth - pssm_coord.z)) * z_range * directional_lights.data[i].shadow_volumetric_fog_fade); + float depth = texture(sampler2D(directional_shadow_atlas, linear_sampler), pssm_coord.xy).r; + float shadow = exp(min(0.0, (depth - pssm_coord.z)) * z_range * INV_FOG_FADE); - 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 = 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(vec3(0.0), vec3(1.0), shadow); - } + shadow_attenuation = mix(vec3(1.0 - directional_lights.data[i].shadow_opacity), 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); + 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) * directional_lights.data[i].volumetric_fog_energy; + } } // Compute light from sky @@ -481,12 +491,12 @@ void main() { float d = distance(omni_lights.data[light_index].position, view_pos); float shadow_attenuation = 1.0; - if (d * omni_lights.data[light_index].inv_radius < 1.0) { + if (omni_lights.data[light_index].volumetric_fog_energy > 0.001 && d * omni_lights.data[light_index].inv_radius < 1.0) { float attenuation = get_omni_attenuation(d, omni_lights.data[light_index].inv_radius, omni_lights.data[light_index].attenuation); vec3 light = omni_lights.data[light_index].color; - if (omni_lights.data[light_index].shadow_enabled) { + if (omni_lights.data[light_index].shadow_opacity > 0.001) { //has shadow vec4 uv_rect = omni_lights.data[light_index].atlas_rect; vec2 flip_offset = omni_lights.data[light_index].direction.xy; @@ -509,9 +519,9 @@ void main() { float depth = texture(sampler2D(shadow_atlas, linear_sampler), pos.xy).r; - shadow_attenuation = exp(min(0.0, (depth - pos.z)) / omni_lights.data[light_index].inv_radius * omni_lights.data[light_index].shadow_volumetric_fog_fade); + shadow_attenuation = mix(1.0 - omni_lights.data[light_index].shadow_opacity, 1.0, exp(min(0.0, (depth - pos.z)) / omni_lights.data[light_index].inv_radius * INV_FOG_FADE)); } - total_light += light * attenuation * shadow_attenuation * henyey_greenstein(dot(normalize(light_pos - view_pos), normalize(view_pos)), params.phase_g); + total_light += light * attenuation * shadow_attenuation * henyey_greenstein(dot(normalize(light_pos - view_pos), normalize(view_pos)), params.phase_g) * omni_lights.data[light_index].volumetric_fog_energy; } } } @@ -562,7 +572,7 @@ void main() { float d = length(light_rel_vec); float shadow_attenuation = 1.0; - if (d * spot_lights.data[light_index].inv_radius < 1.0) { + if (spot_lights.data[light_index].volumetric_fog_energy > 0.001 && d * spot_lights.data[light_index].inv_radius < 1.0) { float attenuation = get_omni_attenuation(d, spot_lights.data[light_index].inv_radius, spot_lights.data[light_index].attenuation); vec3 spot_dir = spot_lights.data[light_index].direction; @@ -572,7 +582,7 @@ void main() { vec3 light = spot_lights.data[light_index].color; - if (spot_lights.data[light_index].shadow_enabled) { + if (spot_lights.data[light_index].shadow_opacity > 0.001) { //has shadow vec4 uv_rect = spot_lights.data[light_index].atlas_rect; vec2 flip_offset = spot_lights.data[light_index].direction.xy; @@ -595,9 +605,9 @@ void main() { float depth = texture(sampler2D(shadow_atlas, linear_sampler), pos.xy).r; - shadow_attenuation = exp(min(0.0, (depth - pos.z)) / spot_lights.data[light_index].inv_radius * spot_lights.data[light_index].shadow_volumetric_fog_fade); + shadow_attenuation = mix(1.0 - spot_lights.data[light_index].shadow_opacity, 1.0, exp(min(0.0, (depth - pos.z)) / spot_lights.data[light_index].inv_radius * INV_FOG_FADE)); } - total_light += light * attenuation * shadow_attenuation * henyey_greenstein(dot(normalize(light_rel_vec), normalize(view_pos)), params.phase_g); + total_light += light * attenuation * shadow_attenuation * henyey_greenstein(dot(normalize(light_rel_vec), normalize(view_pos)), params.phase_g) * spot_lights.data[light_index].volumetric_fog_energy; } } } @@ -619,7 +629,7 @@ void main() { light += a * slight; } - light.rgb *= voxel_gi_instances.data[i].dynamic_range * params.gi_inject; + light.rgb *= voxel_gi_instances.data[i].dynamic_range * params.gi_inject * voxel_gi_instances.data[i].exposure_normalization; total_light += light.rgb; } @@ -686,7 +696,7 @@ void main() { vec3 ambient = texelFetch(sampler2DArray(sdfgi_ambient_texture, linear_sampler), uvw, 0).rgb; - ambient_accum.rgb += ambient * weight; + ambient_accum.rgb += ambient * weight * sdfgi.cascades[i].exposure_normalization; ambient_accum.a += weight; } diff --git a/servers/rendering/renderer_rd/shaders/forward_clustered/SCsub b/servers/rendering/renderer_rd/shaders/forward_clustered/SCsub new file mode 100644 index 0000000000..f06a2d86e2 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/forward_clustered/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")] + [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/scene_forward_clustered.glsl b/servers/rendering/renderer_rd/shaders/forward_clustered/scene_forward_clustered.glsl index 5947fc5351..896f51ca01 100644 --- a/servers/rendering/renderer_rd/shaders/scene_forward_clustered.glsl +++ b/servers/rendering/renderer_rd/shaders/forward_clustered/scene_forward_clustered.glsl @@ -15,11 +15,11 @@ layout(location = 0) in vec3 vertex_attrib; //only for pure render depth when normal is not used #ifdef NORMAL_USED -layout(location = 1) in vec3 normal_attrib; +layout(location = 1) in vec2 normal_attrib; #endif #if defined(TANGENT_USED) || defined(NORMAL_MAP_USED) || defined(LIGHT_ANISOTROPY_USED) -layout(location = 2) in vec4 tangent_attrib; +layout(location = 2) in vec2 tangent_attrib; #endif #if defined(COLOR_USED) @@ -58,6 +58,13 @@ layout(location = 10) in uvec4 bone_attrib; layout(location = 11) in vec4 weight_attrib; #endif +vec3 oct_to_vec3(vec2 e) { + vec3 v = vec3(e.xy, 1.0 - abs(e.x) - abs(e.y)); + float t = max(-v.z, 0.0); + v.xy += t * -sign(v.xy); + return normalize(v); +} + /* Varyings */ layout(location = 0) out vec3 vertex_interp; @@ -90,9 +97,7 @@ layout(location = 8) out vec4 prev_screen_position; #ifdef MATERIAL_UNIFORMS_USED layout(set = MATERIAL_UNIFORM_SET, binding = 0, std140) uniform MaterialUniforms{ - #MATERIAL_UNIFORMS - } material; #endif @@ -122,12 +127,52 @@ invariant gl_Position; #GLOBALS -void vertex_shader(in uint instance_index, in bool is_multimesh, in SceneData scene_data, in mat4 model_matrix, out vec4 screen_pos) { +#ifdef USE_DOUBLE_PRECISION +// Helper functions for emulating double precision when adding floats. +vec3 quick_two_sum(vec3 a, vec3 b, out vec3 out_p) { + vec3 s = a + b; + out_p = b - (s - a); + return s; +} + +vec3 two_sum(vec3 a, vec3 b, out vec3 out_p) { + vec3 s = a + b; + vec3 v = s - a; + out_p = (a - (s - v)) + (b - v); + return s; +} + +vec3 double_add_vec3(vec3 base_a, vec3 prec_a, vec3 base_b, vec3 prec_b, out vec3 out_precision) { + vec3 s, t, se, te; + s = two_sum(base_a, base_b, se); + t = two_sum(prec_a, prec_b, te); + se += t; + s = quick_two_sum(s, se, se); + se += te; + s = quick_two_sum(s, se, out_precision); + return s; +} +#endif + +void vertex_shader(in uint instance_index, in bool is_multimesh, in uint multimesh_offset, 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 + mat4 inv_view_matrix = scene_data.inv_view_matrix; + +#ifdef USE_DOUBLE_PRECISION + vec3 model_precision = vec3(model_matrix[0][3], model_matrix[1][3], model_matrix[2][3]); + model_matrix[0][3] = 0.0; + model_matrix[1][3] = 0.0; + model_matrix[2][3] = 0.0; + vec3 view_precision = vec3(inv_view_matrix[0][3], inv_view_matrix[1][3], inv_view_matrix[2][3]); + inv_view_matrix[0][3] = 0.0; + inv_view_matrix[1][3] = 0.0; + inv_view_matrix[2][3] = 0.0; +#endif + mat3 model_normal_matrix; if (bool(instances.data[instance_index].flags & INSTANCE_FLAGS_NON_UNIFORM_SCALE)) { model_normal_matrix = transpose(inverse(mat3(model_matrix))); @@ -135,11 +180,12 @@ void vertex_shader(in uint instance_index, in bool is_multimesh, in SceneData sc model_normal_matrix = mat3(model_matrix); } + mat4 matrix; + mat4 read_model_matrix = model_matrix; + if (is_multimesh) { //multimesh, instances are for it - mat4 matrix; - #ifdef USE_PARTICLE_TRAILS uint trail_size = (instances.data[instance_index].flags >> INSTANCE_FLAGS_PARTICLE_TRAIL_SHIFT) & INSTANCE_FLAGS_PARTICLE_TRAIL_MASK; uint stride = 3 + 1 + 1; //particles always uses this format @@ -201,7 +247,7 @@ void vertex_shader(in uint instance_index, in bool is_multimesh, in SceneData sc } } - uint offset = stride * gl_InstanceIndex; + uint offset = stride * (gl_InstanceIndex + multimesh_offset); if (bool(instances.data[instance_index].flags & INSTANCE_FLAGS_MULTIMESH_FORMAT_2D)) { matrix = mat4(transforms.data[offset + 0], transforms.data[offset + 1], vec4(0.0, 0.0, 1.0, 0.0), vec4(0.0, 0.0, 0.0, 1.0)); @@ -225,18 +271,26 @@ void vertex_shader(in uint instance_index, in bool is_multimesh, in SceneData sc #endif //transpose matrix = transpose(matrix); - model_matrix = model_matrix * matrix; +#if !defined(USE_DOUBLE_PRECISION) || defined(SKIP_TRANSFORM_USED) || defined(VERTEX_WORLD_COORDS_USED) || defined(MODEL_MATRIX_USED) + // Normally we can bake the multimesh transform into the model matrix, but when using double precision + // we avoid baking it in so we can emulate high precision. + read_model_matrix = model_matrix * matrix; +#if !defined(USE_DOUBLE_PRECISION) || defined(SKIP_TRANSFORM_USED) || defined(VERTEX_WORLD_COORDS_USED) + model_matrix = read_model_matrix; +#endif // !defined(USE_DOUBLE_PRECISION) || defined(SKIP_TRANSFORM_USED) || defined(VERTEX_WORLD_COORDS_USED) +#endif // !defined(USE_DOUBLE_PRECISION) || defined(SKIP_TRANSFORM_USED) || defined(VERTEX_WORLD_COORDS_USED) || defined(MODEL_MATRIX_USED) model_normal_matrix = model_normal_matrix * mat3(matrix); } vec3 vertex = vertex_attrib; #ifdef NORMAL_USED - vec3 normal = normal_attrib * 2.0 - 1.0; + vec3 normal = oct_to_vec3(normal_attrib * 2.0 - 1.0); #endif #if defined(TANGENT_USED) || defined(NORMAL_MAP_USED) || defined(LIGHT_ANISOTROPY_USED) - vec3 tangent = tangent_attrib.xyz * 2.0 - 1.0; - float binormalf = tangent_attrib.a * 2.0 - 1.0; + vec2 signed_tangent_attrib = tangent_attrib * 2.0 - 1.0; + vec3 tangent = oct_to_vec3(vec2(signed_tangent_attrib.x, abs(signed_tangent_attrib.y) * 2.0 - 1.0)); + float binormalf = sign(signed_tangent_attrib.y); vec3 binormal = normalize(cross(normal, tangent) * binormalf); #endif @@ -289,7 +343,22 @@ void vertex_shader(in uint instance_index, in bool is_multimesh, in SceneData sc // using local coordinates (default) #if !defined(SKIP_TRANSFORM_USED) && !defined(VERTEX_WORLD_COORDS_USED) +#ifdef USE_DOUBLE_PRECISION + // We separate the basis from the origin because the basis is fine with single point precision. + // Then we combine the translations from the model matrix and the view matrix using emulated doubles. + // We add the result to the vertex and ignore the final lost precision. + vec3 model_origin = model_matrix[3].xyz; + if (is_multimesh) { + vertex = mat3(matrix) * vertex; + model_origin = double_add_vec3(model_origin, model_precision, matrix[3].xyz, vec3(0.0), model_precision); + } + vertex = mat3(model_matrix) * vertex; + vec3 temp_precision; // Will be ignored. + vertex += double_add_vec3(model_origin, model_precision, scene_data.inv_view_matrix[3].xyz, view_precision, temp_precision); + vertex = mat3(scene_data.view_matrix) * vertex; +#else vertex = (modelview * vec4(vertex, 1.0)).xyz; +#endif #ifdef NORMAL_USED normal = modelview_normal * normal; #endif @@ -318,10 +387,6 @@ void vertex_shader(in uint instance_index, in bool is_multimesh, in SceneData sc vertex_interp = vertex; -#ifdef MOTION_VECTORS - screen_pos = projection_matrix * vec4(vertex_interp, 1.0); -#endif - #ifdef NORMAL_USED normal_interp = normal; #endif @@ -359,6 +424,10 @@ void vertex_shader(in uint instance_index, in bool is_multimesh, in SceneData sc gl_Position = projection_matrix * vec4(vertex_interp, 1.0); #endif +#ifdef MOTION_VECTORS + screen_pos = gl_Position; +#endif + #ifdef MODE_RENDER_DEPTH if (scene_data.pancake_shadows) { if (gl_Position.z <= 0.00001) { @@ -389,13 +458,13 @@ void main() { 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); + vertex_shader(instance_index, is_multimesh, draw_call.multimesh_motion_vectors_previous_offset, 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); + vertex_shader(instance_index, is_multimesh, draw_call.multimesh_motion_vectors_current_offset, 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); + vertex_shader(instance_index, is_multimesh, draw_call.multimesh_motion_vectors_current_offset, scene_data_block.data, model_matrix, screen_position); #endif } @@ -482,7 +551,6 @@ layout(location = 10) in flat uint instance_index_interp; //defines to keep compatibility with vertex -#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] @@ -545,7 +613,7 @@ layout(location = 0) out vec4 frag_color; layout(location = 2) out vec2 motion_vector; #endif -#include "scene_forward_aa_inc.glsl" +#include "../scene_forward_aa_inc.glsl" #if !defined(MODE_RENDER_DEPTH) && !defined(MODE_UNSHADED) @@ -554,20 +622,20 @@ layout(location = 2) out vec2 motion_vector; #define SPECULAR_SCHLICK_GGX #endif -#include "scene_forward_lights_inc.glsl" +#include "../scene_forward_lights_inc.glsl" -#include "scene_forward_gi_inc.glsl" +#include "../scene_forward_gi_inc.glsl" #endif //!defined(MODE_RENDER_DEPTH) && !defined(MODE_UNSHADED) #ifndef MODE_RENDER_DEPTH vec4 volumetric_fog_process(vec2 screen_uv, float z) { - vec3 fog_pos = vec3(screen_uv, z * scene_data_block.data.volumetric_fog_inv_length); + vec3 fog_pos = vec3(screen_uv, z * implementation_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_block.data.volumetric_fog_detail_spread); + fog_pos.z = pow(fog_pos.z, implementation_data.volumetric_fog_detail_spread); } return texture(sampler3D(volumetric_fog_texture, material_samplers[SAMPLER_LINEAR_CLAMP]), fog_pos); @@ -621,7 +689,7 @@ vec4 fog_process(vec3 vertex) { void cluster_get_item_range(uint p_offset, out uint item_min, out uint item_max, out uint item_from, out uint item_to) { uint item_min_max = cluster_buffer.data[p_offset]; - item_min = item_min_max & 0xFFFF; + item_min = item_min_max & 0xFFFFu; item_max = item_min_max >> 16; item_from = item_min >> 5; @@ -729,6 +797,17 @@ void fragment_shader(in SceneData scene_data) { vec2 alpha_texture_coordinate = vec2(0.0, 0.0); #endif // ALPHA_ANTIALIASING_EDGE_USED + mat4 inv_view_matrix = scene_data.inv_view_matrix; + mat4 read_model_matrix = instances.data[instance_index].transform; +#ifdef USE_DOUBLE_PRECISION + read_model_matrix[0][3] = 0.0; + read_model_matrix[1][3] = 0.0; + read_model_matrix[2][3] = 0.0; + inv_view_matrix[0][3] = 0.0; + inv_view_matrix[1][3] = 0.0; + inv_view_matrix[2][3] = 0.0; +#endif + { #CODE : FRAGMENT } @@ -747,7 +826,8 @@ void fragment_shader(in SceneData scene_data) { // alpha hash can be used in unison with alpha antialiasing #ifdef ALPHA_HASH_USED - if (alpha < compute_alpha_hash_threshold(vertex, alpha_hash_scale)) { + vec3 object_pos = (inverse(read_model_matrix) * inv_view_matrix * vec4(vertex, 1.0)).xyz; + if (alpha < compute_alpha_hash_threshold(object_pos, alpha_hash_scale)) { discard; } #endif // ALPHA_HASH_USED @@ -813,7 +893,7 @@ void fragment_shader(in SceneData scene_data) { fog = fog_process(vertex); } - if (scene_data.volumetric_fog_enabled) { + if (implementation_data.volumetric_fog_enabled) { vec4 volumetric_fog = volumetric_fog_process(screen_uv, -vertex.z); if (scene_data.fog_enabled) { //must use the full blending equation here to blend fogs @@ -841,8 +921,8 @@ void fragment_shader(in SceneData scene_data) { #ifndef MODE_RENDER_DEPTH - uvec2 cluster_pos = uvec2(gl_FragCoord.xy) >> scene_data.cluster_shift; - uint cluster_offset = (scene_data.cluster_width * cluster_pos.y + cluster_pos.x) * (scene_data.max_cluster_element_count_div_32 + 32); + uvec2 cluster_pos = uvec2(gl_FragCoord.xy) >> implementation_data.cluster_shift; + uint cluster_offset = (implementation_data.cluster_width * cluster_pos.y + cluster_pos.x) * (implementation_data.max_cluster_element_count_div_32 + 32); uint cluster_z = uint(clamp((-vertex.z / scene_data.z_far) * 32.0, 0.0, 31.0)); @@ -852,14 +932,14 @@ void fragment_shader(in SceneData scene_data) { { // process decals - uint cluster_decal_offset = cluster_offset + scene_data.cluster_type_size * 2; + uint cluster_decal_offset = cluster_offset + implementation_data.cluster_type_size * 2; uint item_min; uint item_max; uint item_from; uint item_to; - cluster_get_item_range(cluster_decal_offset + scene_data.max_cluster_element_count_div_32 + cluster_z, item_min, item_max, item_from, item_to); + cluster_get_item_range(cluster_decal_offset + implementation_data.max_cluster_element_count_div_32 + cluster_z, item_min, item_max, item_from, item_to); #ifdef USE_SUBGROUPS item_from = subgroupBroadcastFirst(subgroupMin(item_from)); @@ -877,9 +957,9 @@ void fragment_shader(in SceneData scene_data) { while (merged_mask != 0) { uint bit = findMSB(merged_mask); - merged_mask &= ~(1 << bit); + merged_mask &= ~(1u << bit); #ifdef USE_SUBGROUPS - if (((1 << bit) & mask) == 0) { //do not process if not originally here + if (((1u << bit) & mask) == 0) { //do not process if not originally here continue; } #endif @@ -947,9 +1027,9 @@ void fragment_shader(in SceneData scene_data) { if (decals.data[decal_index].emission_rect != vec4(0.0)) { //emission is additive, so its independent from albedo if (sc_decal_use_mipmaps) { - emission += textureGrad(sampler2D(decal_atlas_srgb, decal_sampler), uv_local.xz * decals.data[decal_index].emission_rect.zw + decals.data[decal_index].emission_rect.xy, ddx * decals.data[decal_index].emission_rect.zw, ddy * decals.data[decal_index].emission_rect.zw).xyz * decals.data[decal_index].emission_energy * fade; + emission += textureGrad(sampler2D(decal_atlas_srgb, decal_sampler), uv_local.xz * decals.data[decal_index].emission_rect.zw + decals.data[decal_index].emission_rect.xy, ddx * decals.data[decal_index].emission_rect.zw, ddy * decals.data[decal_index].emission_rect.zw).xyz * decals.data[decal_index].modulate.rgb * decals.data[decal_index].emission_energy * fade; } else { - emission += textureLod(sampler2D(decal_atlas_srgb, decal_sampler), uv_local.xz * decals.data[decal_index].emission_rect.zw + decals.data[decal_index].emission_rect.xy, 0.0).xyz * decals.data[decal_index].emission_energy * fade; + emission += textureLod(sampler2D(decal_atlas_srgb, decal_sampler), uv_local.xz * decals.data[decal_index].emission_rect.zw + decals.data[decal_index].emission_rect.xy, 0.0).xyz * decals.data[decal_index].modulate.rgb * decals.data[decal_index].emission_energy * fade; } } } @@ -978,6 +1058,11 @@ void fragment_shader(in SceneData scene_data) { vec3 diffuse_light = vec3(0.0, 0.0, 0.0); vec3 ambient_light = vec3(0.0, 0.0, 0.0); +#ifndef MODE_UNSHADED + // Used in regular draw pass and when drawing SDFs for SDFGI and materials for VoxelGI. + emission *= scene_data.emissive_exposure_normalization; +#endif + #if !defined(MODE_RENDER_DEPTH) && !defined(MODE_UNSHADED) if (scene_data.use_reflection_cubemap) { @@ -988,8 +1073,10 @@ void fragment_shader(in SceneData scene_data) { 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); + ref_vec = mix(ref_vec, bent_normal, roughness * roughness); #else vec3 ref_vec = reflect(-view, normal); + ref_vec = mix(ref_vec, normal, roughness * roughness); #endif float horizon = min(1.0 + dot(ref_vec, normal), 1.0); @@ -1005,6 +1092,7 @@ void fragment_shader(in SceneData scene_data) { specular_light = textureLod(samplerCube(radiance_cubemap, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), ref_vec, roughness * MAX_ROUGHNESS_LOD).rgb; #endif //USE_RADIANCE_CUBEMAP_ARRAY + specular_light *= scene_data.IBL_exposure_normalization; specular_light *= horizon * horizon; specular_light *= scene_data.ambient_light_color_energy.a; } @@ -1025,7 +1113,7 @@ void fragment_shader(in SceneData scene_data) { #else vec3 cubemap_ambient = textureLod(samplerCube(radiance_cubemap, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), ambient_dir, MAX_ROUGHNESS_LOD).rgb; #endif //USE_RADIANCE_CUBEMAP_ARRAY - + cubemap_ambient *= scene_data.IBL_exposure_normalization; ambient_light = mix(ambient_light, cubemap_ambient * scene_data.ambient_light_color_energy.a, scene_data.ambient_color_sky_mix); } } @@ -1046,6 +1134,7 @@ void fragment_shader(in SceneData scene_data) { ambient_light *= attenuation; specular_light *= attenuation; + ref_vec = mix(ref_vec, n, clearcoat_roughness * clearcoat_roughness); 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; @@ -1083,15 +1172,16 @@ void fragment_shader(in SceneData scene_data) { const float c4 = 0.886227; const float c5 = 0.247708; ambient_light += (c1 * lightmap_captures.data[index].sh[8].rgb * (wnormal.x * wnormal.x - wnormal.y * wnormal.y) + - c3 * lightmap_captures.data[index].sh[6].rgb * wnormal.z * wnormal.z + - c4 * lightmap_captures.data[index].sh[0].rgb - - c5 * lightmap_captures.data[index].sh[6].rgb + - 2.0 * c1 * lightmap_captures.data[index].sh[4].rgb * wnormal.x * wnormal.y + - 2.0 * c1 * lightmap_captures.data[index].sh[7].rgb * wnormal.x * wnormal.z + - 2.0 * c1 * lightmap_captures.data[index].sh[5].rgb * wnormal.y * wnormal.z + - 2.0 * c2 * lightmap_captures.data[index].sh[3].rgb * wnormal.x + - 2.0 * c2 * lightmap_captures.data[index].sh[1].rgb * wnormal.y + - 2.0 * c2 * lightmap_captures.data[index].sh[2].rgb * wnormal.z); + c3 * lightmap_captures.data[index].sh[6].rgb * wnormal.z * wnormal.z + + c4 * lightmap_captures.data[index].sh[0].rgb - + c5 * lightmap_captures.data[index].sh[6].rgb + + 2.0 * c1 * lightmap_captures.data[index].sh[4].rgb * wnormal.x * wnormal.y + + 2.0 * c1 * lightmap_captures.data[index].sh[7].rgb * wnormal.x * wnormal.z + + 2.0 * c1 * lightmap_captures.data[index].sh[5].rgb * wnormal.y * wnormal.z + + 2.0 * c2 * lightmap_captures.data[index].sh[3].rgb * wnormal.x + + 2.0 * c2 * lightmap_captures.data[index].sh[1].rgb * wnormal.y + + 2.0 * c2 * lightmap_captures.data[index].sh[2].rgb * wnormal.z) * + scene_data.emissive_exposure_normalization; } else if (bool(instances.data[instance_index].flags & INSTANCE_FLAGS_USE_LIGHTMAP)) { // has actual lightmap bool uses_sh = bool(instances.data[instance_index].flags & INSTANCE_FLAGS_USE_SH_LIGHTMAP); @@ -1109,20 +1199,22 @@ void fragment_shader(in SceneData scene_data) { uint idx = instances.data[instance_index].gi_offset >> 20; vec3 n = normalize(lightmaps.data[idx].normal_xform * normal); + float en = lightmaps.data[idx].exposure_normalization; - ambient_light += lm_light_l0 * 0.282095f; - ambient_light += lm_light_l1n1 * 0.32573 * n.y; - ambient_light += lm_light_l1_0 * 0.32573 * n.z; - ambient_light += lm_light_l1p1 * 0.32573 * n.x; + ambient_light += lm_light_l0 * 0.282095f * en; + ambient_light += lm_light_l1n1 * 0.32573 * n.y * en; + ambient_light += lm_light_l1_0 * 0.32573 * n.z * en; + ambient_light += lm_light_l1p1 * 0.32573 * n.x * en; if (metallic > 0.01) { // since the more direct bounced light is lost, we can kind of fake it with this trick vec3 r = reflect(normalize(-vertex), normal); - specular_light += lm_light_l1n1 * 0.32573 * r.y; - specular_light += lm_light_l1_0 * 0.32573 * r.z; - specular_light += lm_light_l1p1 * 0.32573 * r.x; + specular_light += lm_light_l1n1 * 0.32573 * r.y * en; + specular_light += lm_light_l1_0 * 0.32573 * r.z * en; + specular_light += lm_light_l1p1 * 0.32573 * r.x * en; } } else { - ambient_light += textureLod(sampler2DArray(lightmap_textures[ofs], material_samplers[SAMPLER_LINEAR_CLAMP]), uvw, 0.0).rgb; + uint idx = instances.data[instance_index].gi_offset >> 20; + ambient_light += textureLod(sampler2DArray(lightmap_textures[ofs], material_samplers[SAMPLER_LINEAR_CLAMP]), uvw, 0.0).rgb * lightmaps.data[idx].exposure_normalization; } } #else @@ -1203,6 +1295,7 @@ void fragment_shader(in SceneData scene_data) { uint index1 = instances.data[instance_index].gi_offset & 0xFFFF; vec3 ref_vec = normalize(reflect(-view, normal)); + ref_vec = mix(ref_vec, normal, roughness * roughness); //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)); @@ -1235,7 +1328,7 @@ void fragment_shader(in SceneData scene_data) { vec2 coord; - if (scene_data.gi_upscale_for_msaa) { + if (implementation_data.gi_upscale_for_msaa) { vec2 base_coord = screen_uv; vec2 closest_coord = base_coord; #ifdef USE_MULTIVIEW @@ -1277,10 +1370,10 @@ void fragment_shader(in SceneData scene_data) { } #endif // !USE_LIGHTMAP - if (bool(scene_data.ss_effects_flags & SCREEN_SPACE_EFFECTS_FLAGS_USE_SSAO)) { + if (bool(implementation_data.ss_effects_flags & SCREEN_SPACE_EFFECTS_FLAGS_USE_SSAO)) { float ssao = texture(sampler2D(ao_buffer, material_samplers[SAMPLER_LINEAR_CLAMP]), screen_uv).r; ao = min(ao, ssao); - ao_light_affect = mix(ao_light_affect, max(ao_light_affect, scene_data.ssao_light_affect), scene_data.ssao_ao_affect); + ao_light_affect = mix(ao_light_affect, max(ao_light_affect, implementation_data.ssao_light_affect), implementation_data.ssao_ao_affect); } { // process reflections @@ -1288,20 +1381,32 @@ void fragment_shader(in SceneData scene_data) { vec4 reflection_accum = vec4(0.0, 0.0, 0.0, 0.0); vec4 ambient_accum = vec4(0.0, 0.0, 0.0, 0.0); - uint cluster_reflection_offset = cluster_offset + scene_data.cluster_type_size * 3; + uint cluster_reflection_offset = cluster_offset + implementation_data.cluster_type_size * 3; uint item_min; uint item_max; uint item_from; uint item_to; - cluster_get_item_range(cluster_reflection_offset + scene_data.max_cluster_element_count_div_32 + cluster_z, item_min, item_max, item_from, item_to); + cluster_get_item_range(cluster_reflection_offset + implementation_data.max_cluster_element_count_div_32 + cluster_z, item_min, item_max, item_from, item_to); #ifdef USE_SUBGROUPS item_from = subgroupBroadcastFirst(subgroupMin(item_from)); item_to = subgroupBroadcastFirst(subgroupMax(item_to)); #endif +#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 + vec3 ref_vec = normalize(reflect(-view, bent_normal)); + ref_vec = mix(ref_vec, bent_normal, roughness * roughness); + for (uint i = item_from; i < item_to; i++) { uint mask = cluster_buffer.data[cluster_reflection_offset + i]; mask &= cluster_get_range_clip_mask(i, item_min, item_max); @@ -1313,9 +1418,9 @@ void fragment_shader(in SceneData scene_data) { while (merged_mask != 0) { uint bit = findMSB(merged_mask); - merged_mask &= ~(1 << bit); + merged_mask &= ~(1u << bit); #ifdef USE_SUBGROUPS - if (((1 << bit) & mask) == 0) { //do not process if not originally here + if (((1u << bit) & mask) == 0) { //do not process if not originally here continue; } #endif @@ -1324,16 +1429,8 @@ void fragment_shader(in SceneData scene_data) { if (!bool(reflections.data[reflection_index].mask & instances.data[instance_index].layer_mask)) { continue; //not masked } -#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); + + reflection_process(reflection_index, vertex, ref_vec, normal, roughness, ambient_light, specular_light, ambient_accum, reflection_accum); } } @@ -1349,18 +1446,24 @@ void fragment_shader(in SceneData scene_data) { } //finalize ambient light here - ambient_light *= albedo.rgb; - ambient_light *= ao; + { +#if defined(AMBIENT_LIGHT_DISABLED) + ambient_light = vec3(0.0, 0.0, 0.0); +#else + ambient_light *= albedo.rgb; + ambient_light *= ao; + + if (bool(implementation_data.ss_effects_flags & SCREEN_SPACE_EFFECTS_FLAGS_USE_SSIL)) { + vec4 ssil = textureLod(sampler2D(ssil_buffer, material_samplers[SAMPLER_LINEAR_CLAMP]), screen_uv, 0.0); + ambient_light *= 1.0 - ssil.a; + ambient_light += ssil.rgb * albedo.rgb; + } +#endif // AMBIENT_LIGHT_DISABLED + } // convert ao to direct light ao ao = mix(1.0, ao, ao_light_affect); - if (bool(scene_data.ss_effects_flags & SCREEN_SPACE_EFFECTS_FLAGS_USE_SSIL)) { - vec4 ssil = textureLod(sampler2D(ssil_buffer, material_samplers[SAMPLER_LINEAR_CLAMP]), screen_uv, 0.0); - ambient_light *= 1.0 - ssil.a; - ambient_light += ssil.rgb * albedo.rgb; - } - //this saves some VGPRs vec3 f0 = F0(metallic, specular, albedo); @@ -1381,7 +1484,7 @@ void fragment_shader(in SceneData scene_data) { float a004 = min(r.x * r.x, exp2(-9.28 * ndotv)) * r.x + r.y; vec2 env = vec2(-1.04, 1.04) * a004 + r.zw; - specular_light *= env.x * f0 + env.y; + specular_light *= env.x * f0 + env.y * clamp(50.0 * f0.g, 0.0, 1.0); #endif } @@ -1417,7 +1520,7 @@ void fragment_shader(in SceneData scene_data) { float shadow = 1.0; - if (directional_lights.data[i].shadow_enabled) { + if (directional_lights.data[i].shadow_opacity > 0.001) { float depth_z = -vertex.z; 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)))); @@ -1626,7 +1729,7 @@ void fragment_shader(in SceneData scene_data) { #ifdef LIGHT_TRANSMITTANCE_USED float transmittance_z = transmittance_depth; - if (directional_lights.data[i].shadow_enabled) { + if (directional_lights.data[i].shadow_opacity > 0.001) { float depth_z = -vertex.z; if (depth_z < directional_lights.data[i].shadow_split_offsets.x) { @@ -1677,10 +1780,12 @@ void fragment_shader(in SceneData scene_data) { float shadow = 1.0; #ifndef SHADOWS_DISABLED if (i < 4) { - shadow = float(shadow0 >> (i * 8) & 0xFF) / 255.0; + shadow = float(shadow0 >> (i * 8u) & 0xFFu) / 255.0; } else { - shadow = float(shadow1 >> ((i - 4) * 8) & 0xFF) / 255.0; + shadow = float(shadow1 >> ((i - 4u) * 8u) & 0xFFu) / 255.0; } + + shadow = shadow * directional_lights.data[i].shadow_opacity + 1.0 - directional_lights.data[i].shadow_opacity; #endif blur_shadow(shadow); @@ -1721,7 +1826,7 @@ void fragment_shader(in SceneData scene_data) { uint item_from; uint item_to; - cluster_get_item_range(cluster_omni_offset + scene_data.max_cluster_element_count_div_32 + cluster_z, item_min, item_max, item_from, item_to); + cluster_get_item_range(cluster_omni_offset + implementation_data.max_cluster_element_count_div_32 + cluster_z, item_min, item_max, item_from, item_to); #ifdef USE_SUBGROUPS item_from = subgroupBroadcastFirst(subgroupMin(item_from)); @@ -1739,9 +1844,9 @@ void fragment_shader(in SceneData scene_data) { while (merged_mask != 0) { uint bit = findMSB(merged_mask); - merged_mask &= ~(1 << bit); + merged_mask &= ~(1u << bit); #ifdef USE_SUBGROUPS - if (((1 << bit) & mask) == 0) { //do not process if not originally here + if (((1u << bit) & mask) == 0) { //do not process if not originally here continue; } #endif @@ -1785,14 +1890,14 @@ void fragment_shader(in SceneData scene_data) { { //spot lights - uint cluster_spot_offset = cluster_offset + scene_data.cluster_type_size; + uint cluster_spot_offset = cluster_offset + implementation_data.cluster_type_size; uint item_min; uint item_max; uint item_from; uint item_to; - cluster_get_item_range(cluster_spot_offset + scene_data.max_cluster_element_count_div_32 + cluster_z, item_min, item_max, item_from, item_to); + cluster_get_item_range(cluster_spot_offset + implementation_data.max_cluster_element_count_div_32 + cluster_z, item_min, item_max, item_from, item_to); #ifdef USE_SUBGROUPS item_from = subgroupBroadcastFirst(subgroupMin(item_from)); @@ -1810,9 +1915,9 @@ void fragment_shader(in SceneData scene_data) { while (merged_mask != 0) { uint bit = findMSB(merged_mask); - merged_mask &= ~(1 << bit); + merged_mask &= ~(1u << bit); #ifdef USE_SUBGROUPS - if (((1 << bit) & mask) == 0) { //do not process if not originally here + if (((1u << bit) & mask) == 0) { //do not process if not originally here continue; } #endif @@ -1882,8 +1987,8 @@ void fragment_shader(in SceneData scene_data) { #ifdef MODE_RENDER_SDF { - vec3 local_pos = (scene_data.sdf_to_bounds * vec4(vertex, 1.0)).xyz; - ivec3 grid_pos = scene_data.sdf_offset + ivec3(local_pos * vec3(scene_data.sdf_size)); + vec3 local_pos = (implementation_data.sdf_to_bounds * vec4(vertex, 1.0)).xyz; + ivec3 grid_pos = implementation_data.sdf_offset + ivec3(local_pos * vec3(implementation_data.sdf_size)); uint albedo16 = 0x1; //solid flag albedo16 |= clamp(uint(albedo.r * 31.0), 0, 31) << 11; @@ -1965,7 +2070,7 @@ void fragment_shader(in SceneData scene_data) { float sGreen = floor((cGreen / pow(2.0f, exps - B - N)) + 0.5f); float sBlue = floor((cBlue / pow(2.0f, exps - B - N)) + 0.5f); //store as 8985 to have 2 extra neighbour bits - uint light_rgbe = ((uint(sRed) & 0x1FF) >> 1) | ((uint(sGreen) & 0x1FF) << 8) | (((uint(sBlue) & 0x1FF) >> 1) << 17) | ((uint(exps) & 0x1F) << 25); + uint light_rgbe = ((uint(sRed) & 0x1FFu) >> 1) | ((uint(sGreen) & 0x1FFu) << 8) | (((uint(sBlue) & 0x1FFu) >> 1) << 17) | ((uint(exps) & 0x1Fu) << 25); imageStore(emission_grid, grid_pos, uvec4(light_rgbe)); imageStore(emission_aniso_grid, grid_pos, uvec4(light_aniso)); @@ -1999,8 +2104,8 @@ void fragment_shader(in SceneData scene_data) { if (bool(instances.data[instance_index].flags & INSTANCE_FLAGS_USE_VOXEL_GI)) { // process voxel_gi_instances uint index1 = instances.data[instance_index].gi_offset & 0xFFFF; uint index2 = instances.data[instance_index].gi_offset >> 16; - voxel_gi_buffer.x = index1 & 0xFF; - voxel_gi_buffer.y = index2 & 0xFF; + voxel_gi_buffer.x = index1 & 0xFFu; + voxel_gi_buffer.y = index2 & 0xFFu; } else { voxel_gi_buffer.x = 0xFF; voxel_gi_buffer.y = 0xFF; diff --git a/servers/rendering/renderer_rd/shaders/scene_forward_clustered_inc.glsl b/servers/rendering/renderer_rd/shaders/forward_clustered/scene_forward_clustered_inc.glsl index 0c23de96c3..3a45ab0059 100644 --- a/servers/rendering/renderer_rd/shaders/scene_forward_clustered_inc.glsl +++ b/servers/rendering/renderer_rd/shaders/forward_clustered/scene_forward_clustered_inc.glsl @@ -4,21 +4,23 @@ #define MAX_VOXEL_GI_INSTANCES 8 #define MAX_VIEWS 2 +#ifndef MOLTENVK_USED #if defined(has_GL_KHR_shader_subgroup_ballot) && defined(has_GL_KHR_shader_subgroup_arithmetic) #extension GL_KHR_shader_subgroup_ballot : enable #extension GL_KHR_shader_subgroup_arithmetic : enable #define USE_SUBGROUPS - #endif +#endif // MOLTENVK_USED #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" +#include "../cluster_data_inc.glsl" +#include "../decal_data_inc.glsl" +#include "../scene_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) || defined(LIGHT_ANISOTROPY_USED) #ifndef NORMAL_USED @@ -29,8 +31,8 @@ layout(push_constant, std430) uniform DrawCall { uint instance_index; uint uv_offset; - uint pad0; - uint pad1; + uint multimesh_motion_vectors_current_offset; + uint multimesh_motion_vectors_previous_offset; } draw_call; @@ -38,7 +40,7 @@ draw_call; /* Set 0: Base Pass (never changes) */ -#include "light_data_inc.glsl" +#include "../light_data_inc.glsl" #define SAMPLER_NEAREST_CLAMP 0 #define SAMPLER_LINEAR_CLAMP 1 @@ -61,13 +63,14 @@ layout(set = 0, binding = 3) uniform sampler decal_sampler; layout(set = 0, binding = 4) uniform sampler light_projector_sampler; -#define INSTANCE_FLAGS_NON_UNIFORM_SCALE (1 << 5) -#define INSTANCE_FLAGS_USE_GI_BUFFERS (1 << 6) -#define INSTANCE_FLAGS_USE_SDFGI (1 << 7) -#define INSTANCE_FLAGS_USE_LIGHTMAP_CAPTURE (1 << 8) -#define INSTANCE_FLAGS_USE_LIGHTMAP (1 << 9) -#define INSTANCE_FLAGS_USE_SH_LIGHTMAP (1 << 10) -#define INSTANCE_FLAGS_USE_VOXEL_GI (1 << 11) +#define INSTANCE_FLAGS_NON_UNIFORM_SCALE (1 << 4) +#define INSTANCE_FLAGS_USE_GI_BUFFERS (1 << 5) +#define INSTANCE_FLAGS_USE_SDFGI (1 << 6) +#define INSTANCE_FLAGS_USE_LIGHTMAP_CAPTURE (1 << 7) +#define INSTANCE_FLAGS_USE_LIGHTMAP (1 << 8) +#define INSTANCE_FLAGS_USE_SH_LIGHTMAP (1 << 9) +#define INSTANCE_FLAGS_USE_VOXEL_GI (1 << 10) +#define INSTANCE_FLAGS_PARTICLES (1 << 11) #define INSTANCE_FLAGS_MULTIMESH (1 << 12) #define INSTANCE_FLAGS_MULTIMESH_FORMAT_2D (1 << 13) #define INSTANCE_FLAGS_MULTIMESH_HAS_COLOR (1 << 14) @@ -105,6 +108,8 @@ directional_lights; struct Lightmap { mat3 normal_xform; + vec3 pad; + float exposure_normalization; }; layout(set = 0, binding = 9, std140) restrict readonly buffer Lightmaps { @@ -129,16 +134,18 @@ layout(set = 0, binding = 13, std430) restrict readonly buffer Decals { } decals; -layout(set = 0, binding = 14, std430) restrict readonly buffer GlobalVariableData { +layout(set = 0, binding = 14, std430) restrict readonly buffer GlobalShaderUniformData { vec4 data[]; } -global_variables; +global_shader_uniforms; struct SDFVoxelGICascadeData { vec3 position; float to_probe; ivec3 probe_world_offset; float to_cell; // 1/bounds * grid_size + vec3 pad; + float exposure_normalization; }; layout(set = 0, binding = 15, std140) uniform SDFGI { @@ -171,62 +178,27 @@ sdfgi; /* Set 1: Render Pass (changes per render pass) */ -struct SceneData { - mat4 projection_matrix; - mat4 inv_projection_matrix; - mat4 inv_view_matrix; - mat4 view_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; +layout(set = 1, binding = 0, std140) uniform SceneDataBlock { + SceneData data; + SceneData prev_data; +} +scene_data_block; +struct ImplementationData { uint cluster_shift; uint cluster_width; uint cluster_type_size; uint max_cluster_element_count_div_32; - // Use vec4s because std140 doesn't play nice with vec2s, z and w are wasted. - vec4 directional_penumbra_shadow_kernel[32]; - vec4 directional_soft_shadow_kernel[32]; - vec4 penumbra_shadow_kernel[32]; - vec4 soft_shadow_kernel[32]; - - vec4 ambient_light_color_energy; - - float ambient_color_sky_mix; - bool use_ambient_light; - bool use_ambient_cubemap; - bool use_reflection_cubemap; - - mat3 radiance_inverse_xform; - - vec2 shadow_atlas_pixel_size; - vec2 directional_shadow_pixel_size; - - uint directional_light_count; - float dual_paraboloid_side; - float z_far; - float z_near; - uint ss_effects_flags; float ssao_light_affect; float ssao_ao_affect; - bool roughness_limiter_enabled; - - float roughness_limiter_amount; - float roughness_limiter_limit; - float opaque_prepass_threshold; - uint roughness_limiter_pad; + uint pad1; mat4 sdf_to_bounds; ivec3 sdf_offset; - bool material_uv2_mode; + uint pad2; ivec3 sdf_size; bool gi_upscale_for_msaa; @@ -235,30 +207,14 @@ struct SceneData { float volumetric_fog_inv_length; float volumetric_fog_detail_spread; uint volumetric_fog_pad; - - bool fog_enabled; - float fog_density; - float fog_height; - float fog_height_density; - - vec3 fog_light_color; - float fog_sun_scatter; - - float fog_aerial_perspective; - - float time; - 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; +layout(set = 1, binding = 1, std140) uniform ImplementationDataBlock { + ImplementationData data; } -scene_data_block; +implementation_data_block; + +#define implementation_data implementation_data_block.data struct InstanceData { mat4 transform; @@ -270,42 +226,42 @@ struct InstanceData { vec4 lightmap_uv_scale; }; -layout(set = 1, binding = 1, std430) buffer restrict readonly InstanceDataBuffer { +layout(set = 1, binding = 2, std430) buffer restrict readonly InstanceDataBuffer { InstanceData data[]; } instances; #ifdef USE_RADIANCE_CUBEMAP_ARRAY -layout(set = 1, binding = 2) uniform textureCubeArray radiance_cubemap; +layout(set = 1, binding = 3) uniform textureCubeArray radiance_cubemap; #else -layout(set = 1, binding = 2) uniform textureCube radiance_cubemap; +layout(set = 1, binding = 3) uniform textureCube radiance_cubemap; #endif -layout(set = 1, binding = 3) uniform textureCubeArray reflection_atlas; +layout(set = 1, binding = 4) uniform textureCubeArray reflection_atlas; -layout(set = 1, binding = 4) uniform texture2D shadow_atlas; +layout(set = 1, binding = 5) uniform texture2D shadow_atlas; -layout(set = 1, binding = 5) uniform texture2D directional_shadow_atlas; +layout(set = 1, binding = 6) uniform texture2D directional_shadow_atlas; -layout(set = 1, binding = 6) uniform texture2DArray lightmap_textures[MAX_LIGHTMAP_TEXTURES]; +layout(set = 1, binding = 7) uniform texture2DArray lightmap_textures[MAX_LIGHTMAP_TEXTURES]; -layout(set = 1, binding = 7) uniform texture3D voxel_gi_textures[MAX_VOXEL_GI_INSTANCES]; +layout(set = 1, binding = 8) uniform texture3D voxel_gi_textures[MAX_VOXEL_GI_INSTANCES]; -layout(set = 1, binding = 8, std430) buffer restrict readonly ClusterBuffer { +layout(set = 1, binding = 9, std430) buffer restrict readonly ClusterBuffer { uint data[]; } cluster_buffer; #ifdef MODE_RENDER_SDF -layout(r16ui, set = 1, binding = 9) uniform restrict writeonly uimage3D albedo_volume_grid; -layout(r32ui, set = 1, binding = 10) uniform restrict writeonly uimage3D emission_grid; -layout(r32ui, set = 1, binding = 11) uniform restrict writeonly uimage3D emission_aniso_grid; -layout(r32ui, set = 1, binding = 12) uniform restrict uimage3D geom_facing_grid; +layout(r16ui, set = 1, binding = 10) uniform restrict writeonly uimage3D albedo_volume_grid; +layout(r32ui, set = 1, binding = 11) uniform restrict writeonly uimage3D emission_grid; +layout(r32ui, set = 1, binding = 12) uniform restrict writeonly uimage3D emission_aniso_grid; +layout(r32ui, set = 1, binding = 13) uniform restrict uimage3D geom_facing_grid; //still need to be present for shaders that use it, so remap them to something #define depth_buffer shadow_atlas @@ -314,21 +270,21 @@ layout(r32ui, set = 1, binding = 12) uniform restrict uimage3D geom_facing_grid; #else -layout(set = 1, binding = 9) uniform texture2D depth_buffer; -layout(set = 1, binding = 10) uniform texture2D color_buffer; +layout(set = 1, binding = 10) uniform texture2D depth_buffer; +layout(set = 1, binding = 11) 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; +layout(set = 1, binding = 12) uniform texture2DArray normal_roughness_buffer; +layout(set = 1, binding = 14) uniform texture2DArray ambient_buffer; +layout(set = 1, binding = 15) uniform texture2DArray reflection_buffer; #else // USE_MULTIVIEW -layout(set = 1, binding = 11) uniform texture2D normal_roughness_buffer; -layout(set = 1, binding = 13) uniform texture2D ambient_buffer; -layout(set = 1, binding = 14) uniform texture2D reflection_buffer; +layout(set = 1, binding = 12) uniform texture2D normal_roughness_buffer; +layout(set = 1, binding = 14) uniform texture2D ambient_buffer; +layout(set = 1, binding = 15) 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; +layout(set = 1, binding = 13) uniform texture2D ao_buffer; +layout(set = 1, binding = 16) uniform texture2DArray sdfgi_lightprobe_texture; +layout(set = 1, binding = 17) uniform texture3D sdfgi_occlusion_cascades; struct VoxelGIData { mat4 xform; // 64 - 64 @@ -340,16 +296,19 @@ struct VoxelGIData { float normal_bias; // 4 - 88 bool blend_ambient; // 4 - 92 uint mipmaps; // 4 - 96 + + vec3 pad; // 12 - 108 + float exposure_normalization; // 4 - 112 }; -layout(set = 1, binding = 17, std140) uniform VoxelGIs { +layout(set = 1, binding = 18, std140) uniform VoxelGIs { VoxelGIData data[MAX_VOXEL_GI_INSTANCES]; } voxel_gi_instances; -layout(set = 1, binding = 18) uniform texture3D volumetric_fog_texture; +layout(set = 1, binding = 19) uniform texture3D volumetric_fog_texture; -layout(set = 1, binding = 19) uniform texture2D ssil_buffer; +layout(set = 1, binding = 20) uniform texture2D ssil_buffer; #endif diff --git a/servers/rendering/renderer_rd/shaders/forward_mobile/SCsub b/servers/rendering/renderer_rd/shaders/forward_mobile/SCsub new file mode 100644 index 0000000000..f06a2d86e2 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/forward_mobile/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")] + [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/scene_forward_mobile.glsl b/servers/rendering/renderer_rd/shaders/forward_mobile/scene_forward_mobile.glsl index 26d0de46c2..d50749306e 100644 --- a/servers/rendering/renderer_rd/shaders/scene_forward_mobile.glsl +++ b/servers/rendering/renderer_rd/shaders/forward_mobile/scene_forward_mobile.glsl @@ -16,11 +16,11 @@ layout(location = 0) in vec3 vertex_attrib; //only for pure render depth when normal is not used #ifdef NORMAL_USED -layout(location = 1) in vec3 normal_attrib; +layout(location = 1) in vec2 normal_attrib; #endif #if defined(TANGENT_USED) || defined(NORMAL_MAP_USED) || defined(LIGHT_ANISOTROPY_USED) -layout(location = 2) in vec4 tangent_attrib; +layout(location = 2) in vec2 tangent_attrib; #endif #if defined(COLOR_USED) @@ -59,6 +59,13 @@ layout(location = 10) in uvec4 bone_attrib; layout(location = 11) in vec4 weight_attrib; #endif +vec3 oct_to_vec3(vec2 e) { + vec3 v = vec3(e.xy, 1.0 - abs(e.x) - abs(e.y)); + float t = max(-v.z, 0.0); + v.xy += t * -sign(v.xy); + return normalize(v); +} + /* Varyings */ layout(location = 0) highp out vec3 vertex_interp; @@ -94,7 +101,7 @@ layout(set = MATERIAL_UNIFORM_SET, binding = 0, std140) uniform MaterialUniforms #ifdef MODE_DUAL_PARABOLOID -layout(location = 8) out highp float dp_clip; +layout(location = 9) out highp float dp_clip; #endif @@ -116,6 +123,33 @@ invariant gl_Position; #define scene_data scene_data_block.data +#ifdef USE_DOUBLE_PRECISION +// Helper functions for emulating double precision when adding floats. +vec3 quick_two_sum(vec3 a, vec3 b, out vec3 out_p) { + vec3 s = a + b; + out_p = b - (s - a); + return s; +} + +vec3 two_sum(vec3 a, vec3 b, out vec3 out_p) { + vec3 s = a + b; + vec3 v = s - a; + out_p = (a - (s - v)) + (b - v); + return s; +} + +vec3 double_add_vec3(vec3 base_a, vec3 prec_a, vec3 base_b, vec3 prec_b, out vec3 out_precision) { + vec3 s, t, se, te; + s = two_sum(base_a, base_b, se); + t = two_sum(prec_a, prec_b, te); + se += t; + s = quick_two_sum(s, se, se); + se += te; + s = quick_two_sum(s, se, out_precision); + return s; +} +#endif + void main() { vec4 instance_custom = vec4(0.0); #if defined(COLOR_USED) @@ -125,6 +159,17 @@ void main() { bool is_multimesh = bool(draw_call.flags & INSTANCE_FLAGS_MULTIMESH); mat4 model_matrix = draw_call.transform; + mat4 inv_view_matrix = scene_data.inv_view_matrix; +#ifdef USE_DOUBLE_PRECISION + vec3 model_precision = vec3(model_matrix[0][3], model_matrix[1][3], model_matrix[2][3]); + model_matrix[0][3] = 0.0; + model_matrix[1][3] = 0.0; + model_matrix[2][3] = 0.0; + vec3 view_precision = vec3(inv_view_matrix[0][3], inv_view_matrix[1][3], inv_view_matrix[2][3]); + inv_view_matrix[0][3] = 0.0; + inv_view_matrix[1][3] = 0.0; + inv_view_matrix[2][3] = 0.0; +#endif mat3 model_normal_matrix; if (bool(draw_call.flags & INSTANCE_FLAGS_NON_UNIFORM_SCALE)) { @@ -133,11 +178,12 @@ void main() { model_normal_matrix = mat3(model_matrix); } + mat4 matrix; + mat4 read_model_matrix = model_matrix; + if (is_multimesh) { //multimesh, instances are for it - mat4 matrix; - #ifdef USE_PARTICLE_TRAILS uint trail_size = (draw_call.flags >> INSTANCE_FLAGS_PARTICLE_TRAIL_SHIFT) & INSTANCE_FLAGS_PARTICLE_TRAIL_MASK; uint stride = 3 + 1 + 1; //particles always uses this format @@ -223,18 +269,27 @@ void main() { #endif //transpose matrix = transpose(matrix); - model_matrix = model_matrix * matrix; + +#if !defined(USE_DOUBLE_PRECISION) || defined(SKIP_TRANSFORM_USED) || defined(VERTEX_WORLD_COORDS_USED) || defined(MODEL_MATRIX_USED) + // Normally we can bake the multimesh transform into the model matrix, but when using double precision + // we avoid baking it in so we can emulate high precision. + read_model_matrix = model_matrix * matrix; +#if !defined(USE_DOUBLE_PRECISION) || defined(SKIP_TRANSFORM_USED) || defined(VERTEX_WORLD_COORDS_USED) + model_matrix = read_model_matrix; +#endif // !defined(USE_DOUBLE_PRECISION) || defined(SKIP_TRANSFORM_USED) || defined(VERTEX_WORLD_COORDS_USED) +#endif // !defined(USE_DOUBLE_PRECISION) || defined(SKIP_TRANSFORM_USED) || defined(VERTEX_WORLD_COORDS_USED) || defined(MODEL_MATRIX_USED) model_normal_matrix = model_normal_matrix * mat3(matrix); } vec3 vertex = vertex_attrib; #ifdef NORMAL_USED - vec3 normal = normal_attrib * 2.0 - 1.0; + vec3 normal = oct_to_vec3(normal_attrib * 2.0 - 1.0); #endif #if defined(TANGENT_USED) || defined(NORMAL_MAP_USED) || defined(LIGHT_ANISOTROPY_USED) - vec3 tangent = tangent_attrib.xyz * 2.0 - 1.0; - float binormalf = tangent_attrib.a * 2.0 - 1.0; + vec2 signed_tangent_attrib = tangent_attrib * 2.0 - 1.0; + vec3 tangent = oct_to_vec3(vec2(signed_tangent_attrib.x, abs(signed_tangent_attrib.y) * 2.0 - 1.0)); + float binormalf = sign(signed_tangent_attrib.y); vec3 binormal = normalize(cross(normal, tangent) * binormalf); #endif @@ -289,7 +344,22 @@ void main() { // using local coordinates (default) #if !defined(SKIP_TRANSFORM_USED) && !defined(VERTEX_WORLD_COORDS_USED) +#ifdef USE_DOUBLE_PRECISION + // We separate the basis from the origin because the basis is fine with single point precision. + // Then we combine the translations from the model matrix and the view matrix using emulated doubles. + // We add the result to the vertex and ignore the final lost precision. + vec3 model_origin = model_matrix[3].xyz; + if (is_multimesh) { + vertex = mat3(matrix) * vertex; + model_origin = double_add_vec3(model_origin, model_precision, matrix[3].xyz, vec3(0.0), model_precision); + } + vertex = mat3(model_matrix) * vertex; + vec3 temp_precision; + vertex += double_add_vec3(model_origin, model_precision, scene_data.inv_view_matrix[3].xyz, view_precision, temp_precision); + vertex = mat3(scene_data.view_matrix) * vertex; +#else vertex = (modelview * vec4(vertex, 1.0)).xyz; +#endif #ifdef NORMAL_USED normal = modelview_normal * normal; #endif @@ -442,7 +512,7 @@ layout(location = 6) mediump in vec3 binormal_interp; #ifdef MODE_DUAL_PARABOLOID -layout(location = 8) highp in float dp_clip; +layout(location = 9) highp in float dp_clip; #endif @@ -460,11 +530,12 @@ layout(location = 8) highp in float dp_clip; //defines to keep compatibility with vertex -#define model_matrix draw_call.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 #if defined(ENABLE_SSS) && defined(ENABLE_TRANSMITTANCE) @@ -509,7 +580,7 @@ layout(location = 0) out mediump vec4 frag_color; #endif // RENDER DEPTH -#include "scene_forward_aa_inc.glsl" +#include "../scene_forward_aa_inc.glsl" #if !defined(MODE_RENDER_DEPTH) && !defined(MODE_UNSHADED) @@ -518,7 +589,7 @@ layout(location = 0) out mediump vec4 frag_color; #define SPECULAR_SCHLICK_GGX #endif -#include "scene_forward_lights_inc.glsl" +#include "../scene_forward_lights_inc.glsl" #endif //!defined(MODE_RENDER_DEPTH) && !defined(MODE_UNSHADED) @@ -675,6 +746,17 @@ void main() { vec2 alpha_texture_coordinate = vec2(0.0, 0.0); #endif // ALPHA_ANTIALIASING_EDGE_USED + mat4 inv_view_matrix = scene_data.inv_view_matrix; + mat4 read_model_matrix = draw_call.transform; +#ifdef USE_DOUBLE_PRECISION + read_model_matrix[0][3] = 0.0; + read_model_matrix[1][3] = 0.0; + read_model_matrix[2][3] = 0.0; + inv_view_matrix[0][3] = 0.0; + inv_view_matrix[1][3] = 0.0; + inv_view_matrix[2][3] = 0.0; +#endif + { #CODE : FRAGMENT } @@ -697,7 +779,8 @@ void main() { // alpha hash can be used in unison with alpha antialiasing #ifdef ALPHA_HASH_USED - if (alpha < compute_alpha_hash_threshold(vertex, alpha_hash_scale)) { + vec3 object_pos = (inverse(read_model_matrix) * inv_view_matrix * vec4(vertex, 1.0)).xyz; + if (alpha < compute_alpha_hash_threshold(object_pos, alpha_hash_scale)) { discard; } #endif // ALPHA_HASH_USED @@ -879,6 +962,11 @@ void main() { vec3 diffuse_light = vec3(0.0, 0.0, 0.0); vec3 ambient_light = vec3(0.0, 0.0, 0.0); +#ifndef MODE_UNSHADED + // Used in regular draw pass and when drawing SDFs for SDFGI and materials for VoxelGI. + emission *= scene_data.emissive_exposure_normalization; +#endif + #if !defined(MODE_RENDER_DEPTH) && !defined(MODE_UNSHADED) if (scene_data.use_reflection_cubemap) { @@ -889,8 +977,10 @@ void main() { 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); + ref_vec = mix(ref_vec, bent_normal, roughness * roughness); #else vec3 ref_vec = reflect(-view, normal); + ref_vec = mix(ref_vec, normal, roughness * roughness); #endif float horizon = min(1.0 + dot(ref_vec, normal), 1.0); ref_vec = scene_data.radiance_inverse_xform * ref_vec; @@ -905,6 +995,8 @@ void main() { specular_light = textureLod(samplerCube(radiance_cubemap, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), ref_vec, roughness * MAX_ROUGHNESS_LOD).rgb; #endif //USE_RADIANCE_CUBEMAP_ARRAY + specular_light *= sc_luminance_multiplier; + specular_light *= scene_data.IBL_exposure_normalization; specular_light *= horizon * horizon; specular_light *= scene_data.ambient_light_color_energy.a; } @@ -925,7 +1017,8 @@ void main() { #else vec3 cubemap_ambient = textureLod(samplerCube(radiance_cubemap, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), ambient_dir, MAX_ROUGHNESS_LOD).rgb; #endif //USE_RADIANCE_CUBEMAP_ARRAY - + cubemap_ambient *= sc_luminance_multiplier; + cubemap_ambient *= scene_data.IBL_exposure_normalization; ambient_light = mix(ambient_light, cubemap_ambient * scene_data.ambient_light_color_energy.a, scene_data.ambient_color_sky_mix); } } @@ -940,6 +1033,7 @@ void main() { 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); + ref_vec = mix(ref_vec, n, clearcoat_roughness * clearcoat_roughness); // 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; @@ -982,15 +1076,16 @@ void main() { const float c4 = 0.886227; const float c5 = 0.247708; ambient_light += (c1 * lightmap_captures.data[index].sh[8].rgb * (wnormal.x * wnormal.x - wnormal.y * wnormal.y) + - c3 * lightmap_captures.data[index].sh[6].rgb * wnormal.z * wnormal.z + - c4 * lightmap_captures.data[index].sh[0].rgb - - c5 * lightmap_captures.data[index].sh[6].rgb + - 2.0 * c1 * lightmap_captures.data[index].sh[4].rgb * wnormal.x * wnormal.y + - 2.0 * c1 * lightmap_captures.data[index].sh[7].rgb * wnormal.x * wnormal.z + - 2.0 * c1 * lightmap_captures.data[index].sh[5].rgb * wnormal.y * wnormal.z + - 2.0 * c2 * lightmap_captures.data[index].sh[3].rgb * wnormal.x + - 2.0 * c2 * lightmap_captures.data[index].sh[1].rgb * wnormal.y + - 2.0 * c2 * lightmap_captures.data[index].sh[2].rgb * wnormal.z); + c3 * lightmap_captures.data[index].sh[6].rgb * wnormal.z * wnormal.z + + c4 * lightmap_captures.data[index].sh[0].rgb - + c5 * lightmap_captures.data[index].sh[6].rgb + + 2.0 * c1 * lightmap_captures.data[index].sh[4].rgb * wnormal.x * wnormal.y + + 2.0 * c1 * lightmap_captures.data[index].sh[7].rgb * wnormal.x * wnormal.z + + 2.0 * c1 * lightmap_captures.data[index].sh[5].rgb * wnormal.y * wnormal.z + + 2.0 * c2 * lightmap_captures.data[index].sh[3].rgb * wnormal.x + + 2.0 * c2 * lightmap_captures.data[index].sh[1].rgb * wnormal.y + + 2.0 * c2 * lightmap_captures.data[index].sh[2].rgb * wnormal.z) * + scene_data.emissive_exposure_normalization; } else if (bool(draw_call.flags & INSTANCE_FLAGS_USE_LIGHTMAP)) { // has actual lightmap bool uses_sh = bool(draw_call.flags & INSTANCE_FLAGS_USE_SH_LIGHTMAP); @@ -999,6 +1094,8 @@ void main() { uvw.xy = uv2 * draw_call.lightmap_uv_scale.zw + draw_call.lightmap_uv_scale.xy; uvw.z = float((draw_call.gi_offset >> 16) & 0xFFFF); + uint idx = draw_call.gi_offset >> 20; + if (uses_sh) { uvw.z *= 4.0; //SH textures use 4 times more data vec3 lm_light_l0 = textureLod(sampler2DArray(lightmap_textures[ofs], material_samplers[SAMPLER_LINEAR_CLAMP]), uvw + vec3(0.0, 0.0, 0.0), 0.0).rgb; @@ -1006,22 +1103,22 @@ void main() { vec3 lm_light_l1_0 = textureLod(sampler2DArray(lightmap_textures[ofs], material_samplers[SAMPLER_LINEAR_CLAMP]), uvw + vec3(0.0, 0.0, 2.0), 0.0).rgb; vec3 lm_light_l1p1 = textureLod(sampler2DArray(lightmap_textures[ofs], material_samplers[SAMPLER_LINEAR_CLAMP]), uvw + vec3(0.0, 0.0, 3.0), 0.0).rgb; - uint idx = draw_call.gi_offset >> 20; vec3 n = normalize(lightmaps.data[idx].normal_xform * normal); + float exposure_normalization = lightmaps.data[idx].exposure_normalization; ambient_light += lm_light_l0 * 0.282095f; - ambient_light += lm_light_l1n1 * 0.32573 * n.y; - ambient_light += lm_light_l1_0 * 0.32573 * n.z; - ambient_light += lm_light_l1p1 * 0.32573 * n.x; + ambient_light += lm_light_l1n1 * 0.32573 * n.y * exposure_normalization; + ambient_light += lm_light_l1_0 * 0.32573 * n.z * exposure_normalization; + ambient_light += lm_light_l1p1 * 0.32573 * n.x * exposure_normalization; if (metallic > 0.01) { // since the more direct bounced light is lost, we can kind of fake it with this trick vec3 r = reflect(normalize(-vertex), normal); - specular_light += lm_light_l1n1 * 0.32573 * r.y; - specular_light += lm_light_l1_0 * 0.32573 * r.z; - specular_light += lm_light_l1p1 * 0.32573 * r.x; + specular_light += lm_light_l1n1 * 0.32573 * r.y * exposure_normalization; + specular_light += lm_light_l1_0 * 0.32573 * r.z * exposure_normalization; + specular_light += lm_light_l1p1 * 0.32573 * r.x * exposure_normalization; } } else { - ambient_light += textureLod(sampler2DArray(lightmap_textures[ofs], material_samplers[SAMPLER_LINEAR_CLAMP]), uvw, 0.0).rgb; + ambient_light += textureLod(sampler2DArray(lightmap_textures[ofs], material_samplers[SAMPLER_LINEAR_CLAMP]), uvw, 0.0).rgb * lightmaps.data[idx].exposure_normalization; } } @@ -1036,6 +1133,19 @@ void main() { vec4 ambient_accum = vec4(0.0, 0.0, 0.0, 0.0); uint reflection_indices = draw_call.reflection_probes.x; + +#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 + vec3 ref_vec = normalize(reflect(-view, bent_normal)); + ref_vec = mix(ref_vec, bent_normal, roughness * roughness); + for (uint i = 0; i < 8; i++) { uint reflection_index = reflection_indices & 0xFF; if (i == 4) { @@ -1047,16 +1157,8 @@ void main() { if (reflection_index == 0xFF) { break; } -#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); + + reflection_process(reflection_index, vertex, ref_vec, bent_normal, roughness, ambient_light, specular_light, ambient_accum, reflection_accum); } if (reflection_accum.a > 0.0) { @@ -1071,8 +1173,14 @@ void main() { } //Reflection probes // finalize ambient light here - ambient_light *= albedo.rgb; - ambient_light *= ao; + { +#if defined(AMBIENT_LIGHT_DISABLED) + ambient_light = vec3(0.0, 0.0, 0.0); +#else + ambient_light *= albedo.rgb; + ambient_light *= ao; +#endif // AMBIENT_LIGHT_DISABLED + } // convert ao to direct light ao ao = mix(1.0, ao, ao_light_affect); @@ -1134,7 +1242,7 @@ void main() { #ifdef USE_SOFT_SHADOWS //version with soft shadows, more expensive - if (directional_lights.data[i].shadow_enabled) { + if (directional_lights.data[i].shadow_opacity > 0.001) { float depth_z = -vertex.z; vec4 pssm_coord; @@ -1286,7 +1394,7 @@ void main() { #else // Soft shadow disabled version - if (directional_lights.data[i].shadow_enabled) { + if (directional_lights.data[i].shadow_opacity > 0.001) { float depth_z = -vertex.z; vec4 pssm_coord; diff --git a/servers/rendering/renderer_rd/shaders/scene_forward_mobile_inc.glsl b/servers/rendering/renderer_rd/shaders/forward_mobile/scene_forward_mobile_inc.glsl index 7413d8730a..631ff0575b 100644 --- a/servers/rendering/renderer_rd/shaders/scene_forward_mobile_inc.glsl +++ b/servers/rendering/renderer_rd/shaders/forward_mobile/scene_forward_mobile_inc.glsl @@ -5,7 +5,8 @@ #extension GL_EXT_multiview : enable #endif -#include "decal_data_inc.glsl" +#include "../decal_data_inc.glsl" +#include "../scene_data_inc.glsl" #if !defined(MODE_RENDER_DEPTH) || defined(MODE_RENDER_MATERIAL) || defined(TANGENT_USED) || defined(NORMAL_MAP_USED) || defined(LIGHT_ANISOTROPY_USED) #ifndef NORMAL_USED @@ -32,7 +33,7 @@ draw_call; /* Set 0: Base Pass (never changes) */ -#include "light_data_inc.glsl" +#include "../light_data_inc.glsl" #define SAMPLER_NEAREST_CLAMP 0 #define SAMPLER_LINEAR_CLAMP 1 @@ -54,13 +55,14 @@ layout(set = 0, binding = 2) uniform sampler shadow_sampler; layout(set = 0, binding = 3) uniform sampler decal_sampler; layout(set = 0, binding = 4) uniform sampler light_projector_sampler; -#define INSTANCE_FLAGS_NON_UNIFORM_SCALE (1 << 5) -#define INSTANCE_FLAGS_USE_GI_BUFFERS (1 << 6) -#define INSTANCE_FLAGS_USE_SDFGI (1 << 7) -#define INSTANCE_FLAGS_USE_LIGHTMAP_CAPTURE (1 << 8) -#define INSTANCE_FLAGS_USE_LIGHTMAP (1 << 9) -#define INSTANCE_FLAGS_USE_SH_LIGHTMAP (1 << 10) -#define INSTANCE_FLAGS_USE_VOXEL_GI (1 << 11) +#define INSTANCE_FLAGS_NON_UNIFORM_SCALE (1 << 4) +#define INSTANCE_FLAGS_USE_GI_BUFFERS (1 << 5) +#define INSTANCE_FLAGS_USE_SDFGI (1 << 6) +#define INSTANCE_FLAGS_USE_LIGHTMAP_CAPTURE (1 << 7) +#define INSTANCE_FLAGS_USE_LIGHTMAP (1 << 8) +#define INSTANCE_FLAGS_USE_SH_LIGHTMAP (1 << 9) +#define INSTANCE_FLAGS_USE_VOXEL_GI (1 << 10) +#define INSTANCE_FLAGS_PARTICLES (1 << 11) #define INSTANCE_FLAGS_MULTIMESH (1 << 12) #define INSTANCE_FLAGS_MULTIMESH_FORMAT_2D (1 << 13) #define INSTANCE_FLAGS_MULTIMESH_HAS_COLOR (1 << 14) @@ -94,6 +96,8 @@ directional_lights; struct Lightmap { mediump mat3 normal_xform; + vec3 pad; + float exposure_normalization; }; layout(set = 0, binding = 9, std140) restrict readonly buffer Lightmaps { @@ -118,82 +122,16 @@ layout(set = 0, binding = 13, std430) restrict readonly buffer Decals { } decals; -layout(set = 0, binding = 14, std430) restrict readonly buffer GlobalVariableData { +layout(set = 0, binding = 14, std430) restrict readonly buffer GlobalShaderUniformData { highp vec4 data[]; } -global_variables; +global_shader_uniforms; /* Set 1: Render Pass (changes per render pass) */ -struct SceneData { - highp mat4 projection_matrix; - highp mat4 inv_projection_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; - - // Use vec4s because std140 doesn't play nice with vec2s, z and w are wasted. - highp vec4 directional_penumbra_shadow_kernel[32]; - highp vec4 directional_soft_shadow_kernel[32]; - highp vec4 penumbra_shadow_kernel[32]; - highp vec4 soft_shadow_kernel[32]; - - mediump vec4 ambient_light_color_energy; - - mediump float ambient_color_sky_mix; - bool use_ambient_light; - bool use_ambient_cubemap; - bool use_reflection_cubemap; - - mediump mat3 radiance_inverse_xform; - - highp vec2 shadow_atlas_pixel_size; - highp vec2 directional_shadow_pixel_size; - - uint directional_light_count; - mediump float dual_paraboloid_side; - highp float z_far; - highp float z_near; - - bool ssao_enabled; - mediump float ssao_light_affect; - mediump float ssao_ao_affect; - bool roughness_limiter_enabled; - - mediump float roughness_limiter_amount; - mediump float roughness_limiter_limit; - mediump float opaque_prepass_threshold; - uint roughness_limiter_pad; - - bool fog_enabled; - highp float fog_density; - highp float fog_height; - highp float fog_height_density; - - mediump vec3 fog_light_color; - mediump float fog_sun_scatter; - - mediump float fog_aerial_perspective; - bool material_uv2_mode; - - highp float time; - mediump float reflection_multiplier; // one normally, zero when rendering reflections - - bool pancake_shadows; - uint pad1; - uint pad2; - uint pad3; -}; - layout(set = 1, binding = 0, std140) uniform SceneDataBlock { SceneData data; + SceneData prev_data; } scene_data_block; diff --git a/servers/rendering/renderer_rd/shaders/light_data_inc.glsl b/servers/rendering/renderer_rd/shaders/light_data_inc.glsl index 61c8488a05..7488a3f2c7 100644 --- a/servers/rendering/renderer_rd/shaders/light_data_inc.glsl +++ b/servers/rendering/renderer_rd/shaders/light_data_inc.glsl @@ -15,7 +15,7 @@ struct LightData { //this structure needs to be as packed as possible mediump float cone_attenuation; mediump float cone_angle; mediump float specular_amount; - bool shadow_enabled; + mediump float shadow_opacity; highp vec4 atlas_rect; // rect in the shadow atlas highp mat4 shadow_matrix; @@ -25,7 +25,7 @@ struct LightData { //this structure needs to be as packed as possible highp float soft_shadow_size; // for spot, it's the size in uv coordinates of the light, for omni it's the span angle highp float soft_shadow_scale; // scales the shadow kernel for blurrier shadows uint mask; - mediump float shadow_volumetric_fog_fade; + mediump float volumetric_fog_energy; uint bake_mode; highp vec4 projector_rect; //projector rect in srgb decal atlas }; @@ -44,7 +44,7 @@ struct ReflectionData { bool exterior; bool box_project; uint ambient_mode; - uint pad; + float exposure_normalization; //0-8 is intensity,8-9 is ambient, mode highp mat4 local_matrix; // up to here for spot and omni, rest is for directional // notes: for ambientblend, use distance to edge to blend between already existing global environment @@ -52,7 +52,7 @@ struct ReflectionData { struct DirectionalLightData { mediump vec3 direction; - mediump float energy; + highp float energy; // needs to be highp to avoid NaNs being created with high energy values (i.e. when using physical light units and over-exposing the image) mediump vec3 color; mediump float size; mediump float specular; @@ -60,12 +60,12 @@ struct DirectionalLightData { highp float softshadow_angle; highp float soft_shadow_scale; bool blend_splits; - bool shadow_enabled; + mediump float shadow_opacity; highp float fade_from; highp float fade_to; uvec2 pad; uint bake_mode; - mediump float shadow_volumetric_fog_fade; + mediump float volumetric_fog_energy; highp vec4 shadow_bias; highp vec4 shadow_normal_bias; highp vec4 shadow_transmittance_bias; diff --git a/servers/rendering/renderer_rd/shaders/particles.glsl b/servers/rendering/renderer_rd/shaders/particles.glsl index acb62b812e..9f6aa7adc0 100644 --- a/servers/rendering/renderer_rd/shaders/particles.glsl +++ b/servers/rendering/renderer_rd/shaders/particles.glsl @@ -25,10 +25,10 @@ layout(local_size_x = 64, local_size_y = 1, local_size_z = 1) in; layout(set = 0, binding = 1) uniform sampler material_samplers[12]; -layout(set = 0, binding = 2, std430) restrict readonly buffer GlobalVariableData { +layout(set = 0, binding = 2, std430) restrict readonly buffer GlobalShaderUniformData { vec4 data[]; } -global_variables; +global_shader_uniforms; /* Set 1: FRAME AND PARTICLE DATA */ @@ -458,11 +458,11 @@ void main() { } break; case ATTRACTOR_TYPE_VECTOR_FIELD: { - vec3 uvw_pos = (local_pos / FRAME.attractors[i].extents) * 2.0 - 1.0; + vec3 uvw_pos = (local_pos / FRAME.attractors[i].extents + 1.0) * 0.5; if (any(lessThan(uvw_pos, vec3(0.0))) || any(greaterThan(uvw_pos, vec3(1.0)))) { continue; } - vec3 s = texture(sampler3D(sdf_vec_textures[FRAME.attractors[i].texture_index], material_samplers[SAMPLER_LINEAR_CLAMP]), uvw_pos).xyz; + vec3 s = texture(sampler3D(sdf_vec_textures[FRAME.attractors[i].texture_index], material_samplers[SAMPLER_LINEAR_CLAMP]), uvw_pos).xyz * -2.0 + 1.0; dir = mat3(FRAME.attractors[i].transform) * safe_normalize(s); //revert direction amount = length(s); @@ -475,7 +475,7 @@ void main() { float particle_size = FRAME.particle_size; -#ifdef USE_COLLISON_SCALE +#ifdef USE_COLLISION_SCALE particle_size *= dot(vec3(length(PARTICLE.xform[0].xyz), length(PARTICLE.xform[1].xyz), length(PARTICLE.xform[2].xyz)), vec3(0.33333333333)); diff --git a/servers/rendering/renderer_rd/shaders/scene_data_inc.glsl b/servers/rendering/renderer_rd/shaders/scene_data_inc.glsl new file mode 100644 index 0000000000..048257e9ef --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/scene_data_inc.glsl @@ -0,0 +1,69 @@ +// Scene data stores all our 3D rendering globals for a frame such as our matrices +// where this information is independent of the different RD implementations. +// This enables us to use this UBO in our main scene render shaders but also in +// effects that need access to this data. + +struct SceneData { + highp mat4 projection_matrix; + highp mat4 inv_projection_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; + + // Use vec4s because std140 doesn't play nice with vec2s, z and w are wasted. + highp vec4 directional_penumbra_shadow_kernel[32]; + highp vec4 directional_soft_shadow_kernel[32]; + highp vec4 penumbra_shadow_kernel[32]; + highp vec4 soft_shadow_kernel[32]; + + mediump mat3 radiance_inverse_xform; + + mediump vec4 ambient_light_color_energy; + + mediump float ambient_color_sky_mix; + bool use_ambient_light; + bool use_ambient_cubemap; + bool use_reflection_cubemap; + + highp vec2 shadow_atlas_pixel_size; + highp vec2 directional_shadow_pixel_size; + + uint directional_light_count; + mediump float dual_paraboloid_side; + highp float z_far; + highp float z_near; + + bool roughness_limiter_enabled; + mediump float roughness_limiter_amount; + mediump float roughness_limiter_limit; + mediump float opaque_prepass_threshold; + + bool fog_enabled; + highp float fog_density; + highp float fog_height; + highp float fog_height_density; + + mediump vec3 fog_light_color; + mediump float fog_sun_scatter; + + mediump float fog_aerial_perspective; + highp float time; + mediump float reflection_multiplier; // one normally, zero when rendering reflections + bool material_uv2_mode; + + vec2 taa_jitter; + float emissive_exposure_normalization; + float IBL_exposure_normalization; + + bool pancake_shadows; + uint pad1; + uint pad2; + uint pad3; +}; diff --git a/servers/rendering/renderer_rd/shaders/scene_forward_aa_inc.glsl b/servers/rendering/renderer_rd/shaders/scene_forward_aa_inc.glsl index 97c913d489..71510ee06a 100644 --- a/servers/rendering/renderer_rd/shaders/scene_forward_aa_inc.glsl +++ b/servers/rendering/renderer_rd/shaders/scene_forward_aa_inc.glsl @@ -11,7 +11,8 @@ float hash_3d(vec3 p) { float compute_alpha_hash_threshold(vec3 pos, float hash_scale) { vec3 dx = dFdx(pos); - vec3 dy = dFdx(pos); + vec3 dy = dFdy(pos); + float delta_max_sqr = max(length(dx), length(dy)); float pix_scale = 1.0 / (hash_scale * delta_max_sqr); @@ -32,9 +33,9 @@ float compute_alpha_hash_threshold(vec3 pos, float hash_scale) { 1.0 - ((1.0 - a_interp) * (1.0 - a_interp) / (2.0 * min_lerp * (1.0 - min_lerp)))); float alpha_hash_threshold = - (lerp_factor < (1.0 - min_lerp)) ? ((lerp_factor < min_lerp) ? cases.x : cases.y) : cases.z; + (a_interp < (1.0 - min_lerp)) ? ((a_interp < min_lerp) ? cases.x : cases.y) : cases.z; - return clamp(alpha_hash_threshold, 0.0, 1.0); + return clamp(alpha_hash_threshold, 0.00001, 1.0); } #endif // ALPHA_HASH_USED diff --git a/servers/rendering/renderer_rd/shaders/scene_forward_gi_inc.glsl b/servers/rendering/renderer_rd/shaders/scene_forward_gi_inc.glsl index c88bd0a14b..ae5e1b7251 100644 --- a/servers/rendering/renderer_rd/shaders/scene_forward_gi_inc.glsl +++ b/servers/rendering/renderer_rd/shaders/scene_forward_gi_inc.glsl @@ -94,7 +94,7 @@ void voxel_gi_compute(uint index, vec3 position, vec3 normal, vec3 ref_vec, mat3 light += cone_weights[i] * cone_light.rgb; } - light *= voxel_gi_instances.data[index].dynamic_range; + light *= voxel_gi_instances.data[index].dynamic_range * voxel_gi_instances.data[index].exposure_normalization; out_diff += vec4(light * blend, blend); //irradiance @@ -102,7 +102,7 @@ void voxel_gi_compute(uint index, vec3 position, vec3 normal, vec3 ref_vec, mat3 if (voxel_gi_instances.data[index].blend_ambient) { irr_light.rgb = mix(environment, irr_light.rgb, min(1.0, irr_light.a / 0.95)); } - irr_light.rgb *= voxel_gi_instances.data[index].dynamic_range; + irr_light.rgb *= voxel_gi_instances.data[index].dynamic_range * voxel_gi_instances.data[index].exposure_normalization; //irr_light=vec3(0.0); out_spec += vec4(irr_light.rgb * blend, blend); @@ -189,7 +189,7 @@ void sdfgi_process(uint cascade, vec3 cascade_pos, vec3 cam_pos, vec3 cam_normal pos_uvw.x += float(offset.z) * sdfgi.lightprobe_uv_offset.z; diffuse = textureLod(sampler2DArray(sdfgi_lightprobe_texture, material_samplers[SAMPLER_LINEAR_CLAMP]), pos_uvw, 0.0).rgb; - diffuse_accum += vec4(diffuse * weight, weight); + diffuse_accum += vec4(diffuse * weight * sdfgi.cascades[cascade].exposure_normalization, weight); if (use_specular) { vec3 specular = vec3(0.0); @@ -203,7 +203,7 @@ void sdfgi_process(uint cascade, vec3 cascade_pos, vec3 cam_pos, vec3 cam_normal specular = mix(specular, textureLod(sampler2DArray(sdfgi_lightprobe_texture, material_samplers[SAMPLER_LINEAR_CLAMP]), pos_uvw, 0.0).rgb, (roughness - 0.5) * 2.0); } - specular_accum += specular * weight; + specular_accum += specular * weight * sdfgi.cascades[cascade].exposure_normalization; } } 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 c92b29b14a..2fba1351f7 100644 --- a/servers/rendering/renderer_rd/shaders/scene_forward_lights_inc.glsl +++ b/servers/rendering/renderer_rd/shaders/scene_forward_lights_inc.glsl @@ -97,11 +97,12 @@ void light_compute(vec3 N, vec3 L, vec3 V, float A, vec3 light_color, float atte float diffuse_brdf_NL; // BRDF times N.L for calculating diffuse radiance #if defined(DIFFUSE_LAMBERT_WRAP) - // energy conserving lambert wrap shader - diffuse_brdf_NL = max(0.0, (NdotL + roughness) / ((1.0 + roughness) * (1.0 + roughness))); + // Energy conserving lambert wrap shader. + // https://web.archive.org/web/20210228210901/http://blog.stevemcauley.com/2011/12/03/energy-conserving-wrapped-diffuse/ + diffuse_brdf_NL = max(0.0, (NdotL + roughness) / ((1.0 + roughness) * (1.0 + roughness))) * (1.0 / M_PI); #elif defined(DIFFUSE_TOON) - diffuse_brdf_NL = smoothstep(-roughness, max(roughness, 0.01), NdotL); + diffuse_brdf_NL = smoothstep(-roughness, max(roughness, 0.01), NdotL) * (1.0 / M_PI); #elif defined(DIFFUSE_BURLEY) @@ -199,7 +200,10 @@ void light_compute(vec3 N, vec3 L, vec3 V, float A, vec3 light_color, float atte #endif // LIGHT_ANISOTROPY_USED // F float cLdotH5 = SchlickFresnel(cLdotH); - vec3 F = mix(vec3(cLdotH5), vec3(1.0), f0); + // Calculate Fresnel using specular occlusion term from Filament: + // https://google.github.io/filament/Filament.html#lighting/occlusion/specularocclusion + float f90 = clamp(dot(f0, vec3(50.0 * 0.33)), 0.0, 1.0); + vec3 F = f0 + (f90 - f0) * cLdotH5; vec3 specular_brdf_NL = cNdotL * D * F * G; @@ -392,7 +396,7 @@ float get_omni_attenuation(float distance, float inv_range, float decay) { float light_process_omni_shadow(uint idx, vec3 vertex, vec3 normal) { #ifndef SHADOWS_DISABLED - if (omni_lights.data[idx].shadow_enabled) { + if (omni_lights.data[idx].shadow_opacity > 0.001) { // there is a shadowmap vec2 texel_size = scene_data_block.data.shadow_atlas_pixel_size; vec4 base_uv_rect = omni_lights.data[idx].atlas_rect; @@ -495,6 +499,7 @@ float light_process_omni_shadow(uint idx, vec3 vertex, vec3 normal) { } shadow /= float(sc_penumbra_shadow_samples); + shadow = mix(1.0, shadow, omni_lights.data[idx].shadow_opacity); } else { //no blockers found, so no shadow @@ -513,7 +518,7 @@ float light_process_omni_shadow(uint idx, vec3 vertex, vec3 normal) { vec2 pos = shadow_sample.xy / shadow_sample.z; float depth = shadow_len - omni_lights.data[idx].shadow_bias; depth *= omni_lights.data[idx].inv_radius; - shadow = sample_omni_pcf_shadow(shadow_atlas, omni_lights.data[idx].soft_shadow_scale / shadow_sample.z, pos, uv_rect, flip_offset, depth); + shadow = mix(1.0, sample_omni_pcf_shadow(shadow_atlas, omni_lights.data[idx].soft_shadow_scale / shadow_sample.z, pos, uv_rect, flip_offset, depth), omni_lights.data[idx].shadow_opacity); } return shadow; @@ -671,7 +676,7 @@ void light_process_omni(uint idx, vec3 vertex, vec3 eye_vec, vec3 normal, vec3 v float light_process_spot_shadow(uint idx, vec3 vertex, vec3 normal) { #ifndef SHADOWS_DISABLED - if (spot_lights.data[idx].shadow_enabled) { + if (spot_lights.data[idx].shadow_opacity > 0.001) { vec3 light_rel_vec = spot_lights.data[idx].position - vertex; float light_length = length(light_rel_vec); vec3 spot_dir = spot_lights.data[idx].direction; @@ -732,6 +737,7 @@ float light_process_spot_shadow(uint idx, vec3 vertex, vec3 normal) { } shadow /= float(sc_penumbra_shadow_samples); + shadow = mix(1.0, shadow, spot_lights.data[idx].shadow_opacity); } else { //no blockers found, so no shadow @@ -740,7 +746,7 @@ 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_block.data.shadow_atlas_pixel_size, shadow_uv); + shadow = mix(1.0, sample_pcf_shadow(shadow_atlas, spot_lights.data[idx].soft_shadow_scale * scene_data_block.data.shadow_atlas_pixel_size, shadow_uv), spot_lights.data[idx].shadow_opacity); } return shadow; @@ -869,7 +875,7 @@ 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 view, 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 vertex, vec3 ref_vec, 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; @@ -877,8 +883,6 @@ void reflection_process(uint ref_index, vec3 view, vec3 vertex, vec3 normal, flo return; } - 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)); //make blend more rounded @@ -906,7 +910,7 @@ void reflection_process(uint ref_index, vec3 view, vec3 vertex, vec3 normal, flo vec4 reflection; reflection.rgb = textureLod(samplerCubeArray(reflection_atlas, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), vec4(local_ref_vec, reflections.data[ref_index].index), roughness * MAX_ROUGHNESS_LOD).rgb * sc_luminance_multiplier; - + reflection.rgb *= reflections.data[ref_index].exposure_normalization; if (reflections.data[ref_index].exterior) { reflection.rgb = mix(specular_light, reflection.rgb, blend); } @@ -929,6 +933,7 @@ void reflection_process(uint ref_index, vec3 view, vec3 vertex, vec3 normal, flo vec4 ambient_out; ambient_out.rgb = textureLod(samplerCubeArray(reflection_atlas, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), vec4(local_amb_vec, reflections.data[ref_index].index), MAX_ROUGHNESS_LOD).rgb; + ambient_out.rgb *= reflections.data[ref_index].exposure_normalization; ambient_out.a = blend; if (reflections.data[ref_index].exterior) { ambient_out.rgb = mix(ambient_light, ambient_out.rgb, blend); diff --git a/servers/rendering/renderer_rd/shaders/skeleton.glsl b/servers/rendering/renderer_rd/shaders/skeleton.glsl index a893a66c94..f5b233cca0 100644 --- a/servers/rendering/renderer_rd/shaders/skeleton.glsl +++ b/servers/rendering/renderer_rd/shaders/skeleton.glsl @@ -54,14 +54,54 @@ layout(push_constant, std430) uniform Params { } params; -vec4 decode_abgr_2_10_10_10(uint base) { - uvec4 abgr_2_10_10_10 = (uvec4(base) >> uvec4(0, 10, 20, 30)) & uvec4(0x3FF, 0x3FF, 0x3FF, 0x3); - return vec4(abgr_2_10_10_10) / vec4(1023.0, 1023.0, 1023.0, 3.0) * 2.0 - 1.0; +vec2 uint_to_vec2(uint base) { + uvec2 decode = (uvec2(base) >> uvec2(0, 16)) & uvec2(0xFFFF, 0xFFFF); + return vec2(decode) / vec2(65535.0, 65535.0) * 2.0 - 1.0; } -uint encode_abgr_2_10_10_10(vec4 base) { - uvec4 abgr_2_10_10_10 = uvec4(clamp(ivec4((base * 0.5 + 0.5) * vec4(1023.0, 1023.0, 1023.0, 3.0)), ivec4(0), ivec4(0x3FF, 0x3FF, 0x3FF, 0x3))) << uvec4(0, 10, 20, 30); - return abgr_2_10_10_10.x | abgr_2_10_10_10.y | abgr_2_10_10_10.z | abgr_2_10_10_10.w; +vec3 oct_to_vec3(vec2 oct) { + vec3 v = vec3(oct.xy, 1.0 - abs(oct.x) - abs(oct.y)); + float t = max(-v.z, 0.0); + v.xy += t * -sign(v.xy); + return normalize(v); +} + +vec3 decode_uint_oct_to_norm(uint base) { + return oct_to_vec3(uint_to_vec2(base)); +} + +vec4 decode_uint_oct_to_tang(uint base) { + vec2 oct_sign_encoded = uint_to_vec2(base); + // Binormal sign encoded in y component + vec2 oct = vec2(oct_sign_encoded.x, abs(oct_sign_encoded.y) * 2.0 - 1.0); + return vec4(oct_to_vec3(oct), sign(oct_sign_encoded.y)); +} + +vec2 signNotZero(vec2 v) { + return mix(vec2(-1.0), vec2(1.0), greaterThanEqual(v.xy, vec2(0.0))); +} + +uint vec2_to_uint(vec2 base) { + uvec2 enc = uvec2(clamp(ivec2(base * vec2(65535, 65535)), ivec2(0), ivec2(0xFFFF, 0xFFFF))) << uvec2(0, 16); + return enc.x | enc.y; +} + +vec2 vec3_to_oct(vec3 e) { + e /= abs(e.x) + abs(e.y) + abs(e.z); + vec2 oct = e.z >= 0.0f ? e.xy : (vec2(1.0f) - abs(e.yx)) * signNotZero(e.xy); + return oct * 0.5f + 0.5f; +} + +uint encode_norm_to_uint_oct(vec3 base) { + return vec2_to_uint(vec3_to_oct(base)); +} + +uint encode_tang_to_uint_oct(vec4 base) { + vec2 oct = vec3_to_oct(base.xyz); + // Encode binormal sign in y component + oct.y = oct.y * 0.5f + 0.5f; + oct.y = base.w >= 0.0f ? oct.y : 1 - oct.y; + return vec2_to_uint(oct); } void main() { @@ -103,8 +143,8 @@ void main() { uint skin_offset = params.skin_stride * index; uvec2 bones = uvec2(src_bone_weights.data[skin_offset + 0], src_bone_weights.data[skin_offset + 1]); - uvec2 bones_01 = uvec2(bones.x & 0xFFFF, bones.x >> 16) * 3; //pre-add xform offset - uvec2 bones_23 = uvec2(bones.y & 0xFFFF, bones.y >> 16) * 3; + uvec2 bones_01 = uvec2(bones.x & 0xFFFF, bones.x >> 16) * 2; //pre-add xform offset + uvec2 bones_23 = uvec2(bones.y & 0xFFFF, bones.y >> 16) * 2; skin_offset += params.skin_weight_offset; @@ -121,6 +161,13 @@ void main() { //reverse order because its transposed vertex = (vec4(vertex, 0.0, 1.0) * m).xy; } + + uint dst_offset = index * params.vertex_stride; + + uvec2 uvertex = floatBitsToUint(vertex); + dst_vertices.data[dst_offset + 0] = uvertex.x; + dst_vertices.data[dst_offset + 1] = uvertex.y; + #else vec3 vertex; vec3 normal; @@ -131,12 +178,12 @@ void main() { src_offset += 3; if (params.has_normal) { - normal = decode_abgr_2_10_10_10(src_vertices.data[src_offset]).rgb; + normal = decode_uint_oct_to_norm(src_vertices.data[src_offset]); src_offset++; } if (params.has_tangent) { - tangent = decode_abgr_2_10_10_10(src_vertices.data[src_offset]); + tangent = decode_uint_oct_to_tang(src_vertices.data[src_offset]); } if (params.has_blend_shape) { @@ -155,12 +202,12 @@ void main() { base_offset += 3; if (params.has_normal) { - blend_normal += decode_abgr_2_10_10_10(src_blend_shapes.data[base_offset]).rgb * w; + blend_normal += decode_uint_oct_to_norm(src_blend_shapes.data[base_offset]) * w; base_offset++; } if (params.has_tangent) { - blend_tangent += decode_abgr_2_10_10_10(src_blend_shapes.data[base_offset]).rgb * w; + blend_tangent += decode_uint_oct_to_tang(src_blend_shapes.data[base_offset]).rgb * w; } blend_total += w; @@ -234,12 +281,12 @@ void main() { dst_offset += 3; if (params.has_normal) { - dst_vertices.data[dst_offset] = encode_abgr_2_10_10_10(vec4(normal, 0.0)); + dst_vertices.data[dst_offset] = encode_norm_to_uint_oct(normal); dst_offset++; } if (params.has_tangent) { - dst_vertices.data[dst_offset] = encode_abgr_2_10_10_10(tangent); + dst_vertices.data[dst_offset] = encode_tang_to_uint_oct(tangent); } #endif |