diff options
Diffstat (limited to 'servers/rendering/renderer_rd/shaders')
76 files changed, 8992 insertions, 4011 deletions
diff --git a/servers/rendering/renderer_rd/shaders/SCsub b/servers/rendering/renderer_rd/shaders/SCsub index fc513d3fb9..acb843bfb6 100644 --- a/servers/rendering/renderer_rd/shaders/SCsub +++ b/servers/rendering/renderer_rd/shaders/SCsub @@ -15,3 +15,5 @@ if "RD_GLSL" in env["BUILDERS"]: # compile shaders for glsl_file in glsl_files: env.RD_GLSL(glsl_file) + +SConscript("effects/SCsub") diff --git a/servers/rendering/renderer_rd/shaders/blit.glsl b/servers/rendering/renderer_rd/shaders/blit.glsl new file mode 100644 index 0000000000..14f190a49f --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/blit.glsl @@ -0,0 +1,97 @@ +#[vertex] + +#version 450 + +#VERSION_DEFINES + +layout(push_constant, std140) uniform Pos { + vec4 src_rect; + vec4 dst_rect; + + vec2 eye_center; + float k1; + float k2; + + float upscale; + float aspect_ratio; + uint layer; + uint pad1; +} +data; + +layout(location = 0) out vec2 uv; + +void main() { + vec2 base_arr[4] = vec2[](vec2(0.0, 0.0), vec2(0.0, 1.0), vec2(1.0, 1.0), vec2(1.0, 0.0)); + uv = data.src_rect.xy + base_arr[gl_VertexIndex] * data.src_rect.zw; + vec2 vtx = data.dst_rect.xy + base_arr[gl_VertexIndex] * data.dst_rect.zw; + gl_Position = vec4(vtx * 2.0 - 1.0, 0.0, 1.0); +} + +#[fragment] + +#version 450 + +#VERSION_DEFINES + +layout(push_constant, std140) uniform Pos { + vec4 src_rect; + vec4 dst_rect; + + vec2 eye_center; + float k1; + float k2; + + float upscale; + float aspect_ratio; + uint layer; + uint pad1; +} +data; + +layout(location = 0) in vec2 uv; + +layout(location = 0) out vec4 color; + +#ifdef USE_LAYER +layout(binding = 0) uniform sampler2DArray src_rt; +#else +layout(binding = 0) uniform sampler2D src_rt; +#endif + +void main() { +#ifdef APPLY_LENS_DISTORTION + vec2 coords = uv * 2.0 - 1.0; + vec2 offset = coords - data.eye_center; + + // take aspect ratio into account + offset.y /= data.aspect_ratio; + + // distort + vec2 offset_sq = offset * offset; + float radius_sq = offset_sq.x + offset_sq.y; + float radius_s4 = radius_sq * radius_sq; + float distortion_scale = 1.0 + (data.k1 * radius_sq) + (data.k2 * radius_s4); + offset *= distortion_scale; + + // reapply aspect ratio + offset.y *= data.aspect_ratio; + + // add our eye center back in + coords = offset + data.eye_center; + coords /= data.upscale; + + // and check our color + if (coords.x < -1.0 || coords.y < -1.0 || coords.x > 1.0 || coords.y > 1.0) { + color = vec4(0.0, 0.0, 0.0, 1.0); + } else { + // layer is always used here + coords = (coords + vec2(1.0)) / vec2(2.0); + color = texture(src_rt, vec3(coords, data.layer)); + } +#elif defined(USE_LAYER) + color = texture(src_rt, vec3(uv, data.layer)); +#else + color = texture(src_rt, uv); +#endif +} diff --git a/servers/rendering/renderer_rd/shaders/canvas.glsl b/servers/rendering/renderer_rd/shaders/canvas.glsl index 3b39edc70e..f8e9020f9f 100644 --- a/servers/rendering/renderer_rd/shaders/canvas.glsl +++ b/servers/rendering/renderer_rd/shaders/canvas.glsl @@ -2,7 +2,7 @@ #version 450 -VERSION_DEFINES +#VERSION_DEFINES #ifdef USE_ATTRIBUTES layout(location = 0) in vec2 vertex_attrib; @@ -26,17 +26,15 @@ layout(location = 3) out vec2 pixel_size_interp; #endif -#ifdef USE_MATERIAL_UNIFORMS +#ifdef MATERIAL_UNIFORMS_USED layout(set = 1, binding = 0, std140) uniform MaterialUniforms{ - /* clang-format off */ -MATERIAL_UNIFORMS - /* clang-format on */ + +#MATERIAL_UNIFORMS + } material; #endif -/* clang-format off */ -VERTEX_SHADER_GLOBALS -/* clang-format on */ +#GLOBALS void main() { vec4 instance_custom = vec4(0.0); @@ -67,7 +65,7 @@ void main() { #elif defined(USE_ATTRIBUTES) vec2 vertex = vertex_attrib; - vec4 color = color_attrib; + vec4 color = color_attrib * draw_data.modulation; vec2 uv = uv_attrib; uvec4 bones = bone_attrib; @@ -84,43 +82,84 @@ void main() { #endif - mat4 world_matrix = mat4(vec4(draw_data.world_x, 0.0, 0.0), vec4(draw_data.world_y, 0.0, 0.0), vec4(0.0, 0.0, 1.0, 0.0), vec4(draw_data.world_ofs, 0.0, 1.0)); + mat4 model_matrix = mat4(vec4(draw_data.world_x, 0.0, 0.0), vec4(draw_data.world_y, 0.0, 0.0), vec4(0.0, 0.0, 1.0, 0.0), vec4(draw_data.world_ofs, 0.0, 1.0)); -#if 0 - if (draw_data.flags & FLAGS_INSTANCING_ENABLED) { - uint offset = draw_data.flags & FLAGS_INSTANCING_STRIDE_MASK; - offset *= gl_InstanceIndex; - mat4 instance_xform = mat4( - vec4(texelFetch(instancing_buffer, offset + 0), texelFetch(instancing_buffer, offset + 1), 0.0, texelFetch(instancing_buffer, offset + 3)), - vec4(texelFetch(instancing_buffer, offset + 4), texelFetch(instancing_buffer, offset + 5), 0.0, texelFetch(instancing_buffer, offset + 7)), - vec4(0.0, 0.0, 1.0, 0.0), - vec4(0.0, 0.0, 0.0, 1.0)); - offset += 8; - if (draw_data.flags & FLAGS_INSTANCING_HAS_COLORS) { - vec4 instance_color; - if (draw_data.flags & FLAGS_INSTANCING_COLOR_8_BIT) { - uint bits = floatBitsToUint(texelFetch(instancing_buffer, offset)); - instance_color = unpackUnorm4x8(bits); +#define FLAGS_INSTANCING_MASK 0x7F +#define FLAGS_INSTANCING_HAS_COLORS (1 << 7) +#define FLAGS_INSTANCING_HAS_CUSTOM_DATA (1 << 8) + + uint instancing = draw_data.flags & FLAGS_INSTANCING_MASK; + +#ifdef USE_ATTRIBUTES + if (instancing > 1) { + // trails + + uint stride = 2 + 1 + 1; //particles always uses this format + + uint trail_size = instancing; + + uint offset = trail_size * stride * gl_InstanceIndex; + + vec4 pcolor; + vec2 new_vertex; + { + uint boffset = offset + bone_attrib.x * stride; + new_vertex = (vec4(vertex, 0.0, 1.0) * mat4(transforms.data[boffset + 0], transforms.data[boffset + 1], vec4(0.0, 0.0, 1.0, 0.0), vec4(0.0, 0.0, 0.0, 1.0))).xy * weight_attrib.x; + pcolor = transforms.data[boffset + 2] * weight_attrib.x; + } + if (weight_attrib.y > 0.001) { + uint boffset = offset + bone_attrib.y * stride; + new_vertex += (vec4(vertex, 0.0, 1.0) * mat4(transforms.data[boffset + 0], transforms.data[boffset + 1], vec4(0.0, 0.0, 1.0, 0.0), vec4(0.0, 0.0, 0.0, 1.0))).xy * weight_attrib.y; + pcolor += transforms.data[boffset + 2] * weight_attrib.y; + } + if (weight_attrib.z > 0.001) { + uint boffset = offset + bone_attrib.z * stride; + new_vertex += (vec4(vertex, 0.0, 1.0) * mat4(transforms.data[boffset + 0], transforms.data[boffset + 1], vec4(0.0, 0.0, 1.0, 0.0), vec4(0.0, 0.0, 0.0, 1.0))).xy * weight_attrib.z; + pcolor += transforms.data[boffset + 2] * weight_attrib.z; + } + if (weight_attrib.w > 0.001) { + uint boffset = offset + bone_attrib.w * stride; + new_vertex += (vec4(vertex, 0.0, 1.0) * mat4(transforms.data[boffset + 0], transforms.data[boffset + 1], vec4(0.0, 0.0, 1.0, 0.0), vec4(0.0, 0.0, 0.0, 1.0))).xy * weight_attrib.w; + pcolor += transforms.data[boffset + 2] * weight_attrib.w; + } + + instance_custom = transforms.data[offset + 3]; + + vertex = new_vertex; + color *= pcolor; + } else +#endif // USE_ATTRIBUTES + { + if (instancing == 1) { + uint stride = 2; + { + if (bool(draw_data.flags & FLAGS_INSTANCING_HAS_COLORS)) { + stride += 1; + } + if (bool(draw_data.flags & FLAGS_INSTANCING_HAS_CUSTOM_DATA)) { + stride += 1; + } + } + + uint offset = stride * gl_InstanceIndex; + + mat4 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)); + offset += 2; + + if (bool(draw_data.flags & FLAGS_INSTANCING_HAS_COLORS)) { + color *= transforms.data[offset]; offset += 1; - } else { - instance_color = vec4(texelFetch(instancing_buffer, offset + 0), texelFetch(instancing_buffer, offset + 1), texelFetch(instancing_buffer, offset + 2), texelFetch(instancing_buffer, offset + 3)); - offset += 4; } - color *= instance_color; - } - if (draw_data.flags & FLAGS_INSTANCING_HAS_CUSTOM_DATA) { - if (draw_data.flags & FLAGS_INSTANCING_CUSTOM_DATA_8_BIT) { - uint bits = floatBitsToUint(texelFetch(instancing_buffer, offset)); - instance_custom = unpackUnorm4x8(bits); - } else { - instance_custom = vec4(texelFetch(instancing_buffer, offset + 0), texelFetch(instancing_buffer, offset + 1), texelFetch(instancing_buffer, offset + 2), texelFetch(instancing_buffer, offset + 3)); + if (bool(draw_data.flags & FLAGS_INSTANCING_HAS_CUSTOM_DATA)) { + instance_custom = transforms.data[offset]; } + + matrix = transpose(matrix); + model_matrix = model_matrix * matrix; } } -#endif - #if !defined(USE_ATTRIBUTES) && !defined(USE_PRIMITIVE) if (bool(draw_data.flags & FLAGS_USING_PARTICLES)) { //scale by texture size @@ -132,9 +171,7 @@ void main() { float point_size = 1.0; #endif { - /* clang-format off */ -VERTEX_SHADER_CODE - /* clang-format on */ +#CODE : VERTEX } #ifdef USE_NINEPATCH @@ -142,7 +179,7 @@ VERTEX_SHADER_CODE #endif #if !defined(SKIP_TRANSFORM_USED) - vertex = (world_matrix * vec4(vertex, 0.0, 1.0)).xy; + vertex = (model_matrix * vec4(vertex, 0.0, 1.0)).xy; #endif color_interp = color; @@ -212,7 +249,7 @@ VERTEX_SHADER_CODE #version 450 -VERSION_DEFINES +#VERSION_DEFINES #include "canvas_uniforms_inc.glsl" @@ -228,11 +265,11 @@ layout(location = 3) in vec2 pixel_size_interp; layout(location = 0) out vec4 frag_color; -#ifdef USE_MATERIAL_UNIFORMS +#ifdef MATERIAL_UNIFORMS_USED layout(set = 1, binding = 0, std140) uniform MaterialUniforms{ - /* clang-format off */ -MATERIAL_UNIFORMS - /* clang-format on */ + +#MATERIAL_UNIFORMS + } material; #endif @@ -243,7 +280,7 @@ vec2 screen_uv_to_sdf(vec2 p_uv) { float texture_sdf(vec2 p_sdf) { vec2 uv = p_sdf * canvas_data.sdf_to_tex.xy + canvas_data.sdf_to_tex.zw; float d = texture(sampler2D(sdf_texture, material_samplers[SAMPLER_LINEAR_CLAMP]), uv).r; - d = d * SDF_MAX_LENGTH - 1.0; + d *= SDF_MAX_LENGTH; return d * canvas_data.tex_to_sdf; } @@ -260,11 +297,9 @@ vec2 sdf_to_screen_uv(vec2 p_sdf) { return p_sdf * canvas_data.sdf_to_screen; } -/* clang-format off */ -FRAGMENT_SHADER_GLOBALS -/* clang-format on */ +#GLOBALS -#ifdef LIGHT_SHADER_CODE_USED +#ifdef LIGHT_CODE_USED vec4 light_compute( vec3 light_vertex, @@ -278,9 +313,9 @@ vec4 light_compute( vec2 uv, vec4 color, bool is_directional) { vec4 light = vec4(0.0); - /* clang-format off */ -LIGHT_SHADER_CODE - /* clang-format on */ + +#CODE : LIGHT + return light; } @@ -356,7 +391,7 @@ vec3 light_normal_compute(vec3 light_vec, vec3 normal, vec3 base_color, vec3 lig //float distance = length(shadow_pos); vec4 light_shadow_compute(uint light_base, vec4 light_color, vec4 shadow_uv -#ifdef LIGHT_SHADER_CODE_USED +#ifdef LIGHT_CODE_USED , vec3 shadow_modulate #endif @@ -395,7 +430,7 @@ vec4 light_shadow_compute(uint light_base, vec4 light_color, vec4 shadow_uv } vec4 shadow_color = unpackUnorm4x8(light_array.data[light_base].shadow_color); -#ifdef LIGHT_SHADER_CODE_USED +#ifdef LIGHT_CODE_USED shadow_color.rgb *= shadow_modulate; #endif @@ -422,6 +457,10 @@ void light_blend_compute(uint light_base, vec4 light_color, inout vec3 color) { #endif +float msdf_median(float r, float g, float b, float a) { + return min(max(min(r, g), min(max(r, g), b)), a); +} + void main() { vec4 color = color_interp; vec2 uv = uv_interp; @@ -449,7 +488,34 @@ void main() { #endif - color *= texture(sampler2D(color_texture, texture_sampler), uv); +#ifndef USE_PRIMITIVE + if (bool(draw_data.flags & FLAGS_USE_MSDF)) { + float px_range = draw_data.ninepatch_margins.x; + float outline_thickness = draw_data.ninepatch_margins.y; + //float reserved1 = draw_data.ninepatch_margins.z; + //float reserved2 = draw_data.ninepatch_margins.w; + + vec4 msdf_sample = texture(sampler2D(color_texture, texture_sampler), uv); + vec2 msdf_size = vec2(textureSize(sampler2D(color_texture, texture_sampler), 0)); + vec2 dest_size = vec2(1.0) / fwidth(uv); + float px_size = max(0.5 * dot((vec2(px_range) / msdf_size), dest_size), 1.0); + float d = msdf_median(msdf_sample.r, msdf_sample.g, msdf_sample.b, msdf_sample.a) - 0.5; + + if (outline_thickness > 0) { + float cr = clamp(outline_thickness, 0.0, px_range / 2) / px_range; + float a = clamp((d + cr) * px_size, 0.0, 1.0); + color.a = a * color.a; + } else { + float a = clamp(d * px_size + 0.5, 0.0, 1.0); + color.a = a * color.a; + } + + } else { +#else + { +#endif + color *= texture(sampler2D(color_texture, texture_sampler), uv); + } uint light_count = (draw_data.flags >> FLAGS_LIGHT_COUNT_SHIFT) & 0xF; //max 16 lights bool using_light = light_count > 0 || canvas_data.directional_light_count > 0; @@ -504,11 +570,7 @@ void main() { normal_used = true; #endif - /* clang-format off */ - -FRAGMENT_SHADER_CODE - - /* clang-format on */ +#CODE : FRAGMENT #if defined(NORMAL_MAP_USED) normal = mix(vec3(0.0, 0.0, 1.0), normal_map * vec3(2.0, -2.0, 1.0) - vec3(1.0, -1.0, 0.0), normal_map_depth); @@ -543,7 +605,7 @@ FRAGMENT_SHADER_CODE vec2 direction = light_array.data[light_base].position; vec4 light_color = light_array.data[light_base].color; -#ifdef LIGHT_SHADER_CODE_USED +#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); @@ -561,7 +623,7 @@ FRAGMENT_SHADER_CODE vec4 shadow_uv = vec4(shadow_pos.x, light_array.data[light_base].shadow_y_ofs, shadow_pos.y * light_array.data[light_base].shadow_zfar_inv, 1.0); light_color = light_shadow_compute(light_base, light_color, shadow_uv -#ifdef LIGHT_SHADER_CODE_USED +#ifdef LIGHT_CODE_USED , shadow_modulate.rgb #endif @@ -599,7 +661,7 @@ FRAGMENT_SHADER_CODE vec4 light_color = textureLod(sampler2D(atlas_texture, texture_sampler), tex_uv_atlas, 0.0); vec4 light_base_color = light_array.data[light_base].color; -#ifdef LIGHT_SHADER_CODE_USED +#ifdef LIGHT_CODE_USED vec4 shadow_modulate = vec4(1.0); vec3 light_position = vec3(light_array.data[light_base].position, light_array.data[light_base].height); @@ -657,7 +719,7 @@ FRAGMENT_SHADER_CODE vec4 shadow_uv = vec4(tex_ofs, light_array.data[light_base].shadow_y_ofs, distance, 1.0); light_color = light_shadow_compute(light_base, light_color, shadow_uv -#ifdef LIGHT_SHADER_CODE_USED +#ifdef LIGHT_CODE_USED , shadow_modulate.rgb #endif diff --git a/servers/rendering/renderer_rd/shaders/canvas_occlusion.glsl b/servers/rendering/renderer_rd/shaders/canvas_occlusion.glsl index 5c25235c58..930cf792cb 100644 --- a/servers/rendering/renderer_rd/shaders/canvas_occlusion.glsl +++ b/servers/rendering/renderer_rd/shaders/canvas_occlusion.glsl @@ -2,11 +2,11 @@ #version 450 -VERSION_DEFINES +#VERSION_DEFINES layout(location = 0) in highp vec3 vertex; -layout(push_constant, binding = 0, std430) uniform Constants { +layout(push_constant, std430) uniform Constants { mat4 projection; mat2x4 modelview; vec2 direction; @@ -32,9 +32,9 @@ void main() { #version 450 -VERSION_DEFINES +#VERSION_DEFINES -layout(push_constant, binding = 0, std430) uniform Constants { +layout(push_constant, std430) uniform Constants { mat4 projection; mat2x4 modelview; vec2 direction; diff --git a/servers/rendering/renderer_rd/shaders/canvas_sdf.glsl b/servers/rendering/renderer_rd/shaders/canvas_sdf.glsl index 302ad03b41..0fafc7d486 100644 --- a/servers/rendering/renderer_rd/shaders/canvas_sdf.glsl +++ b/servers/rendering/renderer_rd/shaders/canvas_sdf.glsl @@ -2,17 +2,17 @@ #version 450 -VERSION_DEFINES +#VERSION_DEFINES layout(local_size_x = 8, local_size_y = 8, local_size_z = 1) in; layout(r8, set = 0, binding = 1) uniform restrict readonly image2D src_pixels; -layout(r16, set = 0, binding = 2) uniform restrict writeonly image2D dst_sdf; +layout(r16_snorm, set = 0, binding = 2) uniform restrict writeonly image2D dst_sdf; layout(rg16i, set = 0, binding = 3) uniform restrict readonly iimage2D src_process; layout(rg16i, set = 0, binding = 4) uniform restrict writeonly iimage2D dst_process; -layout(push_constant, binding = 0, std430) uniform Params { +layout(push_constant, std430) uniform Params { ivec2 size; int stride; int shift; @@ -32,7 +32,7 @@ void main() { #ifdef MODE_LOAD bool solid = imageLoad(src_pixels, pos).r > 0.5; - imageStore(dst_process, pos, solid ? ivec4(pos, 0, 0) : ivec4(ivec2(32767), 0, 0)); + imageStore(dst_process, pos, solid ? ivec4(ivec2(-32767), 0, 0) : ivec4(ivec2(32767), 0, 0)); #endif #ifdef MODE_LOAD_SHRINK @@ -43,6 +43,8 @@ void main() { ivec2 rel = ivec2(32767); float d = 1e20; + int found = 0; + int solid_found = 0; for (int i = 0; i < s; i++) { for (int j = 0; j < s; j++) { ivec2 src_pos = base + ivec2(i, j); @@ -56,10 +58,17 @@ void main() { d = dist; rel = src_pos; } + solid_found++; } + found++; } } + if (solid_found == found) { + //mark solid only if all are solid + rel = ivec2(-32767); + } + imageStore(dst_process, pos, ivec4(rel, 0, 0)); #endif @@ -70,6 +79,12 @@ void main() { ivec2 rel = imageLoad(src_process, pos).xy; + bool solid = rel.x < 0; + + if (solid) { + rel = -rel - ivec2(1); + } + if (center != rel) { //only process if it does not point to itself const int ofs_table_size = 8; @@ -92,6 +107,15 @@ void main() { continue; } ivec2 src_rel = imageLoad(src_process, src_pos).xy; + bool src_solid = src_rel.x < 0; + if (src_solid) { + src_rel = -src_rel - ivec2(1); + } + + if (src_solid != solid) { + src_rel = ivec2(src_pos << params.shift); //point to itself if of different type + } + float src_dist = length(vec2(src_rel - center)); if (src_dist < dist) { dist = src_dist; @@ -100,18 +124,31 @@ void main() { } } + if (solid) { + rel = -rel - ivec2(1); + } + imageStore(dst_process, pos, ivec4(rel, 0, 0)); #endif #ifdef MODE_STORE ivec2 rel = imageLoad(src_process, pos).xy; + + bool solid = rel.x < 0; + + if (solid) { + rel = -rel - ivec2(1); + } + float d = length(vec2(rel - pos)); - if (d > 0.01) { - d += 1.0; //make it signed + + if (solid) { + d = -d; } + d /= SDF_MAX_LENGTH; - d = clamp(d, 0.0, 1.0); + d = clamp(d, -1.0, 1.0); imageStore(dst_sdf, pos, vec4(d)); #endif @@ -122,13 +159,20 @@ void main() { ivec2 center = base + ivec2(params.shift); ivec2 rel = imageLoad(src_process, pos).xy; + + bool solid = rel.x < 0; + + if (solid) { + rel = -rel - ivec2(1); + } + float d = length(vec2(rel - center)); - if (d > 0.01) { - d += 1.0; //make it signed + if (solid) { + d = -d; } d /= SDF_MAX_LENGTH; - d = clamp(d, 0.0, 1.0); + d = clamp(d, -1.0, 1.0); imageStore(dst_sdf, pos, vec4(d)); #endif diff --git a/servers/rendering/renderer_rd/shaders/canvas_uniforms_inc.glsl b/servers/rendering/renderer_rd/shaders/canvas_uniforms_inc.glsl index cf7678ea31..12f57b0178 100644 --- a/servers/rendering/renderer_rd/shaders/canvas_uniforms_inc.glsl +++ b/servers/rendering/renderer_rd/shaders/canvas_uniforms_inc.glsl @@ -5,12 +5,10 @@ #define SDF_MAX_LENGTH 16384.0 -#define FLAGS_INSTANCING_STRIDE_MASK 0xF -#define FLAGS_INSTANCING_ENABLED (1 << 4) -#define FLAGS_INSTANCING_HAS_COLORS (1 << 5) -#define FLAGS_INSTANCING_COLOR_8BIT (1 << 6) -#define FLAGS_INSTANCING_HAS_CUSTOM_DATA (1 << 7) -#define FLAGS_INSTANCING_CUSTOM_DATA_8_BIT (1 << 8) +//1 means enabled, 2+ means trails in use +#define FLAGS_INSTANCING_MASK 0x7F +#define FLAGS_INSTANCING_HAS_COLORS (1 << 7) +#define FLAGS_INSTANCING_HAS_CUSTOM_DATA (1 << 8) #define FLAGS_CLIP_RECT_UV (1 << 9) #define FLAGS_TRANSPOSE_RECT (1 << 10) @@ -26,6 +24,8 @@ #define FLAGS_DEFAULT_NORMAL_MAP_USED (1 << 26) #define FLAGS_DEFAULT_SPECULAR_MAP_USED (1 << 27) +#define FLAGS_USE_MSDF (1 << 28) + #define SAMPLER_NEAREST_CLAMP 0 #define SAMPLER_LINEAR_CLAMP 1 #define SAMPLER_NEAREST_WITH_MIPMAPS_CLAMP 2 @@ -41,7 +41,7 @@ // Push Constant -layout(push_constant, binding = 0, std430) uniform DrawData { +layout(push_constant, std430) uniform DrawData { vec2 world_x; vec2 world_y; vec2 world_ofs; diff --git a/servers/rendering/renderer_rd/shaders/cluster_data_inc.glsl b/servers/rendering/renderer_rd/shaders/cluster_data_inc.glsl index 3a4bf4da07..8e616ebe1f 100644 --- a/servers/rendering/renderer_rd/shaders/cluster_data_inc.glsl +++ b/servers/rendering/renderer_rd/shaders/cluster_data_inc.glsl @@ -1,105 +1,3 @@ - #define CLUSTER_COUNTER_SHIFT 20 #define CLUSTER_POINTER_MASK ((1 << CLUSTER_COUNTER_SHIFT) - 1) #define CLUSTER_COUNTER_MASK 0xfff - -struct LightData { //this structure needs to be as packed as possible - vec3 position; - float inv_radius; - - vec3 direction; - float size; - - vec3 color; - float attenuation; - - float cone_attenuation; - float cone_angle; - float specular_amount; - bool shadow_enabled; - - vec4 atlas_rect; // rect in the shadow atlas - mat4 shadow_matrix; - float shadow_bias; - float shadow_normal_bias; - float transmittance_bias; - float soft_shadow_size; // for spot, it's the size in uv coordinates of the light, for omni it's the span angle - float soft_shadow_scale; // scales the shadow kernel for blurrier shadows - uint mask; - float shadow_volumetric_fog_fade; - uint pad; - vec4 projector_rect; //projector rect in srgb decal atlas -}; - -#define REFLECTION_AMBIENT_DISABLED 0 -#define REFLECTION_AMBIENT_ENVIRONMENT 1 -#define REFLECTION_AMBIENT_COLOR 2 - -struct ReflectionData { - vec3 box_extents; - float index; - vec3 box_offset; - uint mask; - vec3 ambient; // ambient color - float intensity; - bool exterior; - bool box_project; - uint ambient_mode; - uint pad; - //0-8 is intensity,8-9 is ambient, mode - 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 -}; - -struct DirectionalLightData { - vec3 direction; - float energy; - vec3 color; - float size; - float specular; - uint mask; - float softshadow_angle; - float soft_shadow_scale; - bool blend_splits; - bool shadow_enabled; - float fade_from; - float fade_to; - uvec3 pad; - float shadow_volumetric_fog_fade; - vec4 shadow_bias; - vec4 shadow_normal_bias; - vec4 shadow_transmittance_bias; - vec4 shadow_z_range; - vec4 shadow_range_begin; - vec4 shadow_split_offsets; - mat4 shadow_matrix1; - mat4 shadow_matrix2; - mat4 shadow_matrix3; - mat4 shadow_matrix4; - vec4 shadow_color1; - vec4 shadow_color2; - vec4 shadow_color3; - vec4 shadow_color4; - vec2 uv_scale1; - vec2 uv_scale2; - vec2 uv_scale3; - vec2 uv_scale4; -}; - -struct DecalData { - mat4 xform; //to decal transform - vec3 inv_extents; - float albedo_mix; - vec4 albedo_rect; - vec4 normal_rect; - vec4 orm_rect; - vec4 emission_rect; - vec4 modulate; - float emission_energy; - uint mask; - float upper_fade; - float lower_fade; - mat3x4 normal_xform; - vec3 normal; - float normal_fade; -}; diff --git a/servers/rendering/renderer_rd/shaders/cluster_debug.glsl b/servers/rendering/renderer_rd/shaders/cluster_debug.glsl index 70a875192c..0034de8c91 100644 --- a/servers/rendering/renderer_rd/shaders/cluster_debug.glsl +++ b/servers/rendering/renderer_rd/shaders/cluster_debug.glsl @@ -2,7 +2,7 @@ #version 450 -VERSION_DEFINES +#VERSION_DEFINES layout(local_size_x = 8, local_size_y = 8, local_size_z = 1) in; @@ -40,7 +40,7 @@ const vec3 usage_gradient[33] = vec3[]( // 1 (none) + 32 vec3(0.83, 0.22, 0.27), vec3(0.83, 0.22, 0.32), vec3(1.00, 0.63, 0.70)); -layout(push_constant, binding = 0, std430) uniform Params { +layout(push_constant, std430) uniform Params { uvec2 screen_size; uvec2 cluster_screen_size; diff --git a/servers/rendering/renderer_rd/shaders/cluster_render.glsl b/servers/rendering/renderer_rd/shaders/cluster_render.glsl index 8723ea78e4..2fe230f0bf 100644 --- a/servers/rendering/renderer_rd/shaders/cluster_render.glsl +++ b/servers/rendering/renderer_rd/shaders/cluster_render.glsl @@ -2,14 +2,14 @@ #version 450 -VERSION_DEFINES +#VERSION_DEFINES layout(location = 0) in vec3 vertex_attrib; layout(location = 0) out float depth_interp; layout(location = 1) out flat uint element_index; -layout(push_constant, binding = 0, std430) uniform Params { +layout(push_constant, std430) uniform Params { uint base_index; uint pad0; uint pad1; @@ -63,9 +63,9 @@ void main() { #version 450 -VERSION_DEFINES +#VERSION_DEFINES -#if defined(GL_KHR_shader_subgroup_ballot) && defined(GL_KHR_shader_subgroup_arithmetic) && defined(GL_KHR_shader_subgroup_vote) +#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 #extension GL_KHR_shader_subgroup_arithmetic : enable @@ -117,7 +117,7 @@ void main() { uint cluster_thread_group_index; if (!gl_HelperInvocation) { - //http://advances.realtimerendering.com/s2017/2017_Sig_Improved_Culling_final.pdf + //https://advances.realtimerendering.com/s2017/2017_Sig_Improved_Culling_final.pdf uvec4 mask; diff --git a/servers/rendering/renderer_rd/shaders/cluster_store.glsl b/servers/rendering/renderer_rd/shaders/cluster_store.glsl index 5be0893c4f..64a145f3c6 100644 --- a/servers/rendering/renderer_rd/shaders/cluster_store.glsl +++ b/servers/rendering/renderer_rd/shaders/cluster_store.glsl @@ -2,11 +2,11 @@ #version 450 -VERSION_DEFINES +#VERSION_DEFINES layout(local_size_x = 8, local_size_y = 8, local_size_z = 1) in; -layout(push_constant, binding = 0, std430) uniform Params { +layout(push_constant, std430) uniform Params { uint cluster_render_data_size; // how much data for a single cluster takes uint max_render_element_count_div_32; //divided by 32 uvec2 cluster_screen_size; diff --git a/servers/rendering/renderer_rd/shaders/cube_to_dp.glsl b/servers/rendering/renderer_rd/shaders/cube_to_dp.glsl index c3ac0bee57..e77d0de719 100644 --- a/servers/rendering/renderer_rd/shaders/cube_to_dp.glsl +++ b/servers/rendering/renderer_rd/shaders/cube_to_dp.glsl @@ -2,13 +2,12 @@ #version 450 -VERSION_DEFINES +#VERSION_DEFINES -layout(push_constant, binding = 1, std430) uniform Params { +layout(push_constant, std430) uniform Params { float z_far; float z_near; - bool z_flip; - uint pad; + vec2 texel_size; vec4 screen_rect; } params; @@ -26,31 +25,32 @@ void main() { #version 450 -VERSION_DEFINES +#VERSION_DEFINES layout(location = 0) in vec2 uv_interp; layout(set = 0, binding = 0) uniform samplerCube source_cube; -layout(push_constant, binding = 1, std430) uniform Params { +layout(push_constant, std430) uniform Params { float z_far; float z_near; - bool z_flip; - uint pad; + vec2 texel_size; vec4 screen_rect; } params; void main() { vec2 uv = uv_interp; + vec2 texel_size = abs(params.texel_size); - vec3 normal = vec3(uv * 2.0 - 1.0, 0.0); + uv = clamp(uv * (1.0 + 2.0 * texel_size) - texel_size, vec2(0.0), vec2(1.0)); - normal.z = 0.5 - 0.5 * ((normal.x * normal.x) + (normal.y * normal.y)); + vec3 normal = vec3(uv * 2.0 - 1.0, 0.0); + normal.z = 0.5 * (1.0 - dot(normal.xy, normal.xy)); // z = 1/2 - 1/2 * (x^2 + y^2) normal = normalize(normal); normal.y = -normal.y; //needs to be flipped to match projection matrix - if (!params.z_flip) { + if (params.texel_size.x >= 0.0) { // Sign is used to encode Z flip normal.z = -normal.z; } diff --git a/servers/rendering/renderer_rd/shaders/cubemap_downsampler.glsl b/servers/rendering/renderer_rd/shaders/cubemap_downsampler.glsl index 7f269b7af3..63f0ce690e 100644 --- a/servers/rendering/renderer_rd/shaders/cubemap_downsampler.glsl +++ b/servers/rendering/renderer_rd/shaders/cubemap_downsampler.glsl @@ -22,7 +22,7 @@ #version 450 -VERSION_DEFINES +#VERSION_DEFINES #define BLOCK_SIZE 8 @@ -32,53 +32,7 @@ layout(set = 0, binding = 0) uniform samplerCube source_cubemap; layout(rgba16f, set = 1, binding = 0) uniform restrict writeonly imageCube dest_cubemap; -layout(push_constant, binding = 1, std430) uniform Params { - uint face_size; -} -params; - -#define M_PI 3.14159265359 - -void get_dir_0(out vec3 dir, in float u, in float v) { - dir[0] = 1.0; - dir[1] = v; - dir[2] = -u; -} - -void get_dir_1(out vec3 dir, in float u, in float v) { - dir[0] = -1.0; - dir[1] = v; - dir[2] = u; -} - -void get_dir_2(out vec3 dir, in float u, in float v) { - dir[0] = u; - dir[1] = 1.0; - dir[2] = -v; -} - -void get_dir_3(out vec3 dir, in float u, in float v) { - dir[0] = u; - dir[1] = -1.0; - dir[2] = v; -} - -void get_dir_4(out vec3 dir, in float u, in float v) { - dir[0] = u; - dir[1] = v; - dir[2] = 1.0; -} - -void get_dir_5(out vec3 dir, in float u, in float v) { - dir[0] = -u; - dir[1] = v; - dir[2] = -1.0; -} - -float calcWeight(float u, float v) { - float val = u * u + v * v + 1.0; - return val * sqrt(val); -} +#include "cubemap_downsampler_inc.glsl" void main() { uvec3 id = gl_GlobalInvocationID; diff --git a/servers/rendering/renderer_rd/shaders/cubemap_downsampler_inc.glsl b/servers/rendering/renderer_rd/shaders/cubemap_downsampler_inc.glsl new file mode 100644 index 0000000000..641e0906f5 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/cubemap_downsampler_inc.glsl @@ -0,0 +1,48 @@ +layout(push_constant, std430) uniform Params { + uint face_size; + uint face_id; // only used in raster shader +} +params; + +#define M_PI 3.14159265359 + +void get_dir_0(out vec3 dir, in float u, in float v) { + dir[0] = 1.0; + dir[1] = v; + dir[2] = -u; +} + +void get_dir_1(out vec3 dir, in float u, in float v) { + dir[0] = -1.0; + dir[1] = v; + dir[2] = u; +} + +void get_dir_2(out vec3 dir, in float u, in float v) { + dir[0] = u; + dir[1] = 1.0; + dir[2] = -v; +} + +void get_dir_3(out vec3 dir, in float u, in float v) { + dir[0] = u; + dir[1] = -1.0; + dir[2] = v; +} + +void get_dir_4(out vec3 dir, in float u, in float v) { + dir[0] = u; + dir[1] = v; + dir[2] = 1.0; +} + +void get_dir_5(out vec3 dir, in float u, in float v) { + dir[0] = -u; + dir[1] = v; + dir[2] = -1.0; +} + +float calcWeight(float u, float v) { + float val = u * u + v * v + 1.0; + return val * sqrt(val); +} diff --git a/servers/rendering/renderer_rd/shaders/cubemap_downsampler_raster.glsl b/servers/rendering/renderer_rd/shaders/cubemap_downsampler_raster.glsl new file mode 100644 index 0000000000..0828ffd921 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/cubemap_downsampler_raster.glsl @@ -0,0 +1,163 @@ +// Copyright 2016 Activision Publishing, Inc. +// +// Permission is hereby granted, free of charge, to any person obtaining +// a copy of this software and associated documentation files (the "Software"), +// to deal in the Software without restriction, including without limitation +// the rights to use, copy, modify, merge, publish, distribute, sublicense, +// and/or sell copies of the Software, and to permit persons to whom the Software +// is furnished to do so, subject to the following conditions: +// +// The above copyright notice and this permission notice shall be included in all +// copies or substantial portions of the Software. +// +// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR +// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, +// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE +// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER +// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, +// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE +// SOFTWARE. + +/* clang-format off */ +#[vertex] + +#version 450 + +#VERSION_DEFINES + +#include "cubemap_downsampler_inc.glsl" + +layout(location = 0) out vec2 uv_interp; +/* clang-format on */ + +void main() { + vec2 base_arr[4] = vec2[](vec2(0.0, 0.0), vec2(0.0, 1.0), vec2(1.0, 1.0), vec2(1.0, 0.0)); + uv_interp = base_arr[gl_VertexIndex] * float(params.face_size); + gl_Position = vec4(base_arr[gl_VertexIndex] * 2.0 - 1.0, 0.0, 1.0); +} + +/* clang-format off */ +#[fragment] + +#version 450 + +#VERSION_DEFINES + +#include "cubemap_downsampler_inc.glsl" + +layout(set = 0, binding = 0) uniform samplerCube source_cubemap; + +layout(location = 0) in vec2 uv_interp; +layout(location = 0) out vec4 frag_color; +/* clang-format on */ + +void main() { + // Converted from compute shader which uses absolute coordinates. + // Could possibly simplify this + float face_size = float(params.face_size); + + if (uv_interp.x < face_size && uv_interp.y < face_size) { + float inv_face_size = 1.0 / face_size; + + float u0 = (uv_interp.x * 2.0 + 1.0 - 0.75) * inv_face_size - 1.0; + float u1 = (uv_interp.x * 2.0 + 1.0 + 0.75) * inv_face_size - 1.0; + + float v0 = (uv_interp.y * 2.0 + 1.0 - 0.75) * -inv_face_size + 1.0; + float v1 = (uv_interp.y * 2.0 + 1.0 + 0.75) * -inv_face_size + 1.0; + + float weights[4]; + weights[0] = calcWeight(u0, v0); + weights[1] = calcWeight(u1, v0); + weights[2] = calcWeight(u0, v1); + weights[3] = calcWeight(u1, v1); + + const float wsum = 0.5 / (weights[0] + weights[1] + weights[2] + weights[3]); + for (int i = 0; i < 4; i++) { + weights[i] = weights[i] * wsum + .125; + } + + vec3 dir; + vec4 color; + switch (params.face_id) { + case 0: + get_dir_0(dir, u0, v0); + color = textureLod(source_cubemap, normalize(dir), 0.0) * weights[0]; + + get_dir_0(dir, u1, v0); + color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[1]; + + get_dir_0(dir, u0, v1); + color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[2]; + + get_dir_0(dir, u1, v1); + color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[3]; + break; + case 1: + get_dir_1(dir, u0, v0); + color = textureLod(source_cubemap, normalize(dir), 0.0) * weights[0]; + + get_dir_1(dir, u1, v0); + color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[1]; + + get_dir_1(dir, u0, v1); + color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[2]; + + get_dir_1(dir, u1, v1); + color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[3]; + break; + case 2: + get_dir_2(dir, u0, v0); + color = textureLod(source_cubemap, normalize(dir), 0.0) * weights[0]; + + get_dir_2(dir, u1, v0); + color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[1]; + + get_dir_2(dir, u0, v1); + color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[2]; + + get_dir_2(dir, u1, v1); + color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[3]; + break; + case 3: + get_dir_3(dir, u0, v0); + color = textureLod(source_cubemap, normalize(dir), 0.0) * weights[0]; + + get_dir_3(dir, u1, v0); + color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[1]; + + get_dir_3(dir, u0, v1); + color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[2]; + + get_dir_3(dir, u1, v1); + color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[3]; + break; + case 4: + get_dir_4(dir, u0, v0); + color = textureLod(source_cubemap, normalize(dir), 0.0) * weights[0]; + + get_dir_4(dir, u1, v0); + color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[1]; + + get_dir_4(dir, u0, v1); + color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[2]; + + get_dir_4(dir, u1, v1); + color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[3]; + break; + default: + get_dir_5(dir, u0, v0); + color = textureLod(source_cubemap, normalize(dir), 0.0) * weights[0]; + + get_dir_5(dir, u1, v0); + color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[1]; + + get_dir_5(dir, u0, v1); + color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[2]; + + get_dir_5(dir, u1, v1); + color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[3]; + break; + } + frag_color = color; + } +} diff --git a/servers/rendering/renderer_rd/shaders/cubemap_filter.glsl b/servers/rendering/renderer_rd/shaders/cubemap_filter.glsl index 987545fb76..2a774b0eb4 100644 --- a/servers/rendering/renderer_rd/shaders/cubemap_filter.glsl +++ b/servers/rendering/renderer_rd/shaders/cubemap_filter.glsl @@ -22,7 +22,7 @@ #version 450 -VERSION_DEFINES +#VERSION_DEFINES #define GROUP_SIZE 64 diff --git a/servers/rendering/renderer_rd/shaders/cubemap_filter_raster.glsl b/servers/rendering/renderer_rd/shaders/cubemap_filter_raster.glsl new file mode 100644 index 0000000000..0990dc7c2f --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/cubemap_filter_raster.glsl @@ -0,0 +1,256 @@ +// Copyright 2016 Activision Publishing, Inc. +// +// Permission is hereby granted, free of charge, to any person obtaining +// a copy of this software and associated documentation files (the "Software"), +// to deal in the Software without restriction, including without limitation +// the rights to use, copy, modify, merge, publish, distribute, sublicense, +// and/or sell copies of the Software, and to permit persons to whom the Software +// is furnished to do so, subject to the following conditions: +// +// The above copyright notice and this permission notice shall be included in all +// copies or substantial portions of the Software. +// +// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR +// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, +// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE +// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER +// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, +// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE +// SOFTWARE. + +/* clang-format off */ +#[vertex] + +#version 450 + +#VERSION_DEFINES + +layout(push_constant, std430) uniform Params { + int mip_level; + uint face_id; +} +params; + +layout(location = 0) out vec2 uv_interp; +/* clang-format on */ + +void main() { + vec2 base_arr[4] = vec2[](vec2(0.0, 0.0), vec2(0.0, 1.0), vec2(1.0, 1.0), vec2(1.0, 0.0)); + uv_interp = base_arr[gl_VertexIndex]; + gl_Position = vec4(base_arr[gl_VertexIndex] * 2.0 - 1.0, 0.0, 1.0); +} + +/* clang-format off */ +#[fragment] + +#version 450 + +#VERSION_DEFINES + +layout(push_constant, std430) uniform Params { + int mip_level; + uint face_id; +} +params; + +layout(set = 0, binding = 0) uniform samplerCube source_cubemap; + +layout(location = 0) in vec2 uv_interp; +layout(location = 0) out vec4 frag_color; + +/* clang-format on */ + +#ifdef USE_HIGH_QUALITY +#define NUM_TAPS 32 +#else +#define NUM_TAPS 8 +#endif + +#define BASE_RESOLUTION 128 + +#ifdef USE_HIGH_QUALITY +layout(set = 1, binding = 0, std430) buffer restrict readonly Data { + vec4[7][5][3][24] coeffs; +} +data; +#else +layout(set = 1, binding = 0, std430) buffer restrict readonly Data { + vec4[7][5][6] coeffs; +} +data; +#endif + +void get_dir(out vec3 dir, in vec2 uv, in uint face) { + switch (face) { + case 0: + dir = vec3(1.0, uv[1], -uv[0]); + break; + case 1: + dir = vec3(-1.0, uv[1], uv[0]); + break; + case 2: + dir = vec3(uv[0], 1.0, -uv[1]); + break; + case 3: + dir = vec3(uv[0], -1.0, uv[1]); + break; + case 4: + dir = vec3(uv[0], uv[1], 1.0); + break; + default: + dir = vec3(-uv[0], uv[1], -1.0); + break; + } +} + +void main() { + // determine dir / pos for the texel + vec3 dir, adir, frameZ; + { + vec2 uv; + uv.x = uv_interp.x; + uv.y = 1.0 - uv_interp.y; + uv = uv * 2.0 - 1.0; + + get_dir(dir, uv, params.face_id); + frameZ = normalize(dir); + + adir = abs(dir); + } + + // determine which texel this is + // NOTE (macOS/MoltenVK): Do not rename, "level" variable name conflicts with the Metal "level(float lod)" mipmap sampling function name. + int mip_level = 0; + + if (params.mip_level < 0) { + // return as is + frag_color.rgb = textureLod(source_cubemap, frameZ, 0.0).rgb; + frag_color.a = 1.0; + return; + } else if (params.mip_level > 6) { + // maximum level + mip_level = 6; + } else { + mip_level = params.mip_level; + } + + // GGX gather colors + vec4 color = vec4(0.0); + for (int axis = 0; axis < 3; axis++) { + const int otherAxis0 = 1 - (axis & 1) - (axis >> 1); + const int otherAxis1 = 2 - (axis >> 1); + + float frameweight = (max(adir[otherAxis0], adir[otherAxis1]) - .75) / .25; + if (frameweight > 0.0) { + // determine frame + vec3 UpVector; + switch (axis) { + case 0: + UpVector = vec3(1, 0, 0); + break; + case 1: + UpVector = vec3(0, 1, 0); + break; + default: + UpVector = vec3(0, 0, 1); + break; + } + + vec3 frameX = normalize(cross(UpVector, frameZ)); + vec3 frameY = cross(frameZ, frameX); + + // calculate parametrization for polynomial + float Nx = dir[otherAxis0]; + float Ny = dir[otherAxis1]; + float Nz = adir[axis]; + + float NmaxXY = max(abs(Ny), abs(Nx)); + Nx /= NmaxXY; + Ny /= NmaxXY; + + float theta; + if (Ny < Nx) { + if (Ny <= -0.999) + theta = Nx; + else + theta = Ny; + } else { + if (Ny >= 0.999) + theta = -Nx; + else + theta = -Ny; + } + + float phi; + if (Nz <= -0.999) + phi = -NmaxXY; + else if (Nz >= 0.999) + phi = NmaxXY; + else + phi = Nz; + + float theta2 = theta * theta; + float phi2 = phi * phi; + + // sample + for (int iSuperTap = 0; iSuperTap < NUM_TAPS / 4; iSuperTap++) { + const int index = (NUM_TAPS / 4) * axis + iSuperTap; + +#ifdef USE_HIGH_QUALITY + vec4 coeffsDir0[3]; + vec4 coeffsDir1[3]; + vec4 coeffsDir2[3]; + vec4 coeffsLevel[3]; + vec4 coeffsWeight[3]; + + for (int iCoeff = 0; iCoeff < 3; iCoeff++) { + coeffsDir0[iCoeff] = data.coeffs[mip_level][0][iCoeff][index]; + coeffsDir1[iCoeff] = data.coeffs[mip_level][1][iCoeff][index]; + coeffsDir2[iCoeff] = data.coeffs[mip_level][2][iCoeff][index]; + coeffsLevel[iCoeff] = data.coeffs[mip_level][3][iCoeff][index]; + coeffsWeight[iCoeff] = data.coeffs[mip_level][4][iCoeff][index]; + } + + for (int iSubTap = 0; iSubTap < 4; iSubTap++) { + // determine sample attributes (dir, weight, mip_level) + vec3 sample_dir = frameX * (coeffsDir0[0][iSubTap] + coeffsDir0[1][iSubTap] * theta2 + coeffsDir0[2][iSubTap] * phi2) + frameY * (coeffsDir1[0][iSubTap] + coeffsDir1[1][iSubTap] * theta2 + coeffsDir1[2][iSubTap] * phi2) + frameZ * (coeffsDir2[0][iSubTap] + coeffsDir2[1][iSubTap] * theta2 + coeffsDir2[2][iSubTap] * phi2); + + float sample_level = coeffsLevel[0][iSubTap] + coeffsLevel[1][iSubTap] * theta2 + coeffsLevel[2][iSubTap] * phi2; + + float sample_weight = coeffsWeight[0][iSubTap] + coeffsWeight[1][iSubTap] * theta2 + coeffsWeight[2][iSubTap] * phi2; +#else + vec4 coeffsDir0 = data.coeffs[mip_level][0][index]; + vec4 coeffsDir1 = data.coeffs[mip_level][1][index]; + vec4 coeffsDir2 = data.coeffs[mip_level][2][index]; + vec4 coeffsLevel = data.coeffs[mip_level][3][index]; + vec4 coeffsWeight = data.coeffs[mip_level][4][index]; + + for (int iSubTap = 0; iSubTap < 4; iSubTap++) { + // determine sample attributes (dir, weight, mip_level) + vec3 sample_dir = frameX * coeffsDir0[iSubTap] + frameY * coeffsDir1[iSubTap] + frameZ * coeffsDir2[iSubTap]; + + float sample_level = coeffsLevel[iSubTap]; + + float sample_weight = coeffsWeight[iSubTap]; +#endif + + sample_weight *= frameweight; + + // adjust for jacobian + sample_dir /= max(abs(sample_dir[0]), max(abs(sample_dir[1]), abs(sample_dir[2]))); + sample_level += 0.75 * log2(dot(sample_dir, sample_dir)); + // sample cubemap + color.xyz += textureLod(source_cubemap, normalize(sample_dir), sample_level).xyz * sample_weight; + color.w += sample_weight; + } + } + } + } + color /= color.w; + + // write color + color.xyz = max(vec3(0.0), color.xyz); + color.w = 1.0; + + frag_color = color; +} diff --git a/servers/rendering/renderer_rd/shaders/cubemap_roughness.glsl b/servers/rendering/renderer_rd/shaders/cubemap_roughness.glsl index 5cbb00baa4..1d46f59408 100644 --- a/servers/rendering/renderer_rd/shaders/cubemap_roughness.glsl +++ b/servers/rendering/renderer_rd/shaders/cubemap_roughness.glsl @@ -2,7 +2,7 @@ #version 450 -VERSION_DEFINES +#VERSION_DEFINES #define GROUP_SIZE 8 @@ -12,100 +12,7 @@ layout(set = 0, binding = 0) uniform samplerCube source_cube; layout(rgba16f, set = 1, binding = 0) uniform restrict writeonly imageCube dest_cubemap; -layout(push_constant, binding = 1, std430) uniform Params { - uint face_id; - uint sample_count; - float roughness; - bool use_direct_write; - float face_size; -} -params; - -#define M_PI 3.14159265359 - -vec3 texelCoordToVec(vec2 uv, uint faceID) { - mat3 faceUvVectors[6]; - - // -x - faceUvVectors[1][0] = vec3(0.0, 0.0, 1.0); // u -> +z - faceUvVectors[1][1] = vec3(0.0, -1.0, 0.0); // v -> -y - faceUvVectors[1][2] = vec3(-1.0, 0.0, 0.0); // -x face - - // +x - faceUvVectors[0][0] = vec3(0.0, 0.0, -1.0); // u -> -z - faceUvVectors[0][1] = vec3(0.0, -1.0, 0.0); // v -> -y - faceUvVectors[0][2] = vec3(1.0, 0.0, 0.0); // +x face - - // -y - faceUvVectors[3][0] = vec3(1.0, 0.0, 0.0); // u -> +x - faceUvVectors[3][1] = vec3(0.0, 0.0, -1.0); // v -> -z - faceUvVectors[3][2] = vec3(0.0, -1.0, 0.0); // -y face - - // +y - faceUvVectors[2][0] = vec3(1.0, 0.0, 0.0); // u -> +x - faceUvVectors[2][1] = vec3(0.0, 0.0, 1.0); // v -> +z - faceUvVectors[2][2] = vec3(0.0, 1.0, 0.0); // +y face - - // -z - faceUvVectors[5][0] = vec3(-1.0, 0.0, 0.0); // u -> -x - faceUvVectors[5][1] = vec3(0.0, -1.0, 0.0); // v -> -y - faceUvVectors[5][2] = vec3(0.0, 0.0, -1.0); // -z face - - // +z - faceUvVectors[4][0] = vec3(1.0, 0.0, 0.0); // u -> +x - faceUvVectors[4][1] = vec3(0.0, -1.0, 0.0); // v -> -y - faceUvVectors[4][2] = vec3(0.0, 0.0, 1.0); // +z face - - // out = u * s_faceUv[0] + v * s_faceUv[1] + s_faceUv[2]. - vec3 result = (faceUvVectors[faceID][0] * uv.x) + (faceUvVectors[faceID][1] * uv.y) + faceUvVectors[faceID][2]; - return normalize(result); -} - -vec3 ImportanceSampleGGX(vec2 Xi, float Roughness, vec3 N) { - float a = Roughness * Roughness; // DISNEY'S ROUGHNESS [see Burley'12 siggraph] - - // Compute distribution direction - float Phi = 2.0 * M_PI * Xi.x; - float CosTheta = sqrt((1.0 - Xi.y) / (1.0 + (a * a - 1.0) * Xi.y)); - float SinTheta = sqrt(1.0 - CosTheta * CosTheta); - - // Convert to spherical direction - vec3 H; - H.x = SinTheta * cos(Phi); - H.y = SinTheta * sin(Phi); - H.z = CosTheta; - - vec3 UpVector = abs(N.z) < 0.999 ? vec3(0.0, 0.0, 1.0) : vec3(1.0, 0.0, 0.0); - vec3 TangentX = normalize(cross(UpVector, N)); - vec3 TangentY = cross(N, TangentX); - - // Tangent to world space - return TangentX * H.x + TangentY * H.y + N * H.z; -} - -// http://graphicrants.blogspot.com.au/2013/08/specular-brdf-reference.html -float GGX(float NdotV, float a) { - float k = a / 2.0; - return NdotV / (NdotV * (1.0 - k) + k); -} - -// http://graphicrants.blogspot.com.au/2013/08/specular-brdf-reference.html -float G_Smith(float a, float nDotV, float nDotL) { - return GGX(nDotL, a * a) * GGX(nDotV, a * a); -} - -float radicalInverse_VdC(uint bits) { - bits = (bits << 16u) | (bits >> 16u); - bits = ((bits & 0x55555555u) << 1u) | ((bits & 0xAAAAAAAAu) >> 1u); - bits = ((bits & 0x33333333u) << 2u) | ((bits & 0xCCCCCCCCu) >> 2u); - bits = ((bits & 0x0F0F0F0Fu) << 4u) | ((bits & 0xF0F0F0F0u) >> 4u); - bits = ((bits & 0x00FF00FFu) << 8u) | ((bits & 0xFF00FF00u) >> 8u); - return float(bits) * 2.3283064365386963e-10; // / 0x100000000 -} - -vec2 Hammersley(uint i, uint N) { - return vec2(float(i) / float(N), radicalInverse_VdC(i)); -} +#include "cubemap_roughness_inc.glsl" void main() { uvec3 id = gl_GlobalInvocationID; @@ -114,24 +21,38 @@ void main() { vec2 uv = ((vec2(id.xy) * 2.0 + 1.0) / (params.face_size) - 1.0); vec3 N = texelCoordToVec(uv, id.z); - //vec4 color = color_interp; - if (params.use_direct_write) { imageStore(dest_cubemap, ivec3(id), vec4(texture(source_cube, N).rgb, 1.0)); } else { vec4 sum = vec4(0.0, 0.0, 0.0, 0.0); + float solid_angle_texel = 4.0 * M_PI / (6.0 * params.face_size * params.face_size); + float roughness2 = params.roughness * params.roughness; + float roughness4 = roughness2 * roughness2; + vec3 UpVector = abs(N.z) < 0.999 ? vec3(0.0, 0.0, 1.0) : vec3(1.0, 0.0, 0.0); + mat3 T; + T[0] = normalize(cross(UpVector, N)); + T[1] = cross(N, T[0]); + T[2] = N; + for (uint sampleNum = 0u; sampleNum < params.sample_count; sampleNum++) { vec2 xi = Hammersley(sampleNum, params.sample_count); - vec3 H = ImportanceSampleGGX(xi, params.roughness, N); - vec3 V = N; - vec3 L = (2.0 * dot(V, H) * H - V); + vec3 H = T * ImportanceSampleGGX(xi, roughness4); + float NdotH = dot(N, H); + vec3 L = (2.0 * NdotH * H - N); float ndotl = clamp(dot(N, L), 0.0, 1.0); if (ndotl > 0.0) { - sum.rgb += textureLod(source_cube, L, 0.0).rgb * ndotl; + float D = DistributionGGX(NdotH, roughness4); + float pdf = D * NdotH / (4.0 * NdotH) + 0.0001; + + float solid_angle_sample = 1.0 / (float(params.sample_count) * pdf + 0.0001); + + float mipLevel = params.roughness == 0.0 ? 0.0 : 0.5 * log2(solid_angle_sample / solid_angle_texel); + + sum.rgb += textureLod(source_cube, L, mipLevel).rgb * ndotl; sum.a += ndotl; } } diff --git a/servers/rendering/renderer_rd/shaders/cubemap_roughness_inc.glsl b/servers/rendering/renderer_rd/shaders/cubemap_roughness_inc.glsl new file mode 100644 index 0000000000..1bee428a6f --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/cubemap_roughness_inc.glsl @@ -0,0 +1,95 @@ +#define M_PI 3.14159265359 + +layout(push_constant, std430) uniform Params { + uint face_id; + uint sample_count; + float roughness; + bool use_direct_write; + float face_size; +} +params; + +vec3 texelCoordToVec(vec2 uv, uint faceID) { + mat3 faceUvVectors[6]; + + // -x + faceUvVectors[1][0] = vec3(0.0, 0.0, 1.0); // u -> +z + faceUvVectors[1][1] = vec3(0.0, -1.0, 0.0); // v -> -y + faceUvVectors[1][2] = vec3(-1.0, 0.0, 0.0); // -x face + + // +x + faceUvVectors[0][0] = vec3(0.0, 0.0, -1.0); // u -> -z + faceUvVectors[0][1] = vec3(0.0, -1.0, 0.0); // v -> -y + faceUvVectors[0][2] = vec3(1.0, 0.0, 0.0); // +x face + + // -y + faceUvVectors[3][0] = vec3(1.0, 0.0, 0.0); // u -> +x + faceUvVectors[3][1] = vec3(0.0, 0.0, -1.0); // v -> -z + faceUvVectors[3][2] = vec3(0.0, -1.0, 0.0); // -y face + + // +y + faceUvVectors[2][0] = vec3(1.0, 0.0, 0.0); // u -> +x + faceUvVectors[2][1] = vec3(0.0, 0.0, 1.0); // v -> +z + faceUvVectors[2][2] = vec3(0.0, 1.0, 0.0); // +y face + + // -z + faceUvVectors[5][0] = vec3(-1.0, 0.0, 0.0); // u -> -x + faceUvVectors[5][1] = vec3(0.0, -1.0, 0.0); // v -> -y + faceUvVectors[5][2] = vec3(0.0, 0.0, -1.0); // -z face + + // +z + faceUvVectors[4][0] = vec3(1.0, 0.0, 0.0); // u -> +x + faceUvVectors[4][1] = vec3(0.0, -1.0, 0.0); // v -> -y + faceUvVectors[4][2] = vec3(0.0, 0.0, 1.0); // +z face + + // out = u * s_faceUv[0] + v * s_faceUv[1] + s_faceUv[2]. + vec3 result = (faceUvVectors[faceID][0] * uv.x) + (faceUvVectors[faceID][1] * uv.y) + faceUvVectors[faceID][2]; + return normalize(result); +} + +vec3 ImportanceSampleGGX(vec2 xi, float roughness4) { + // Compute distribution direction + float Phi = 2.0 * M_PI * xi.x; + float CosTheta = sqrt((1.0 - xi.y) / (1.0 + (roughness4 - 1.0) * xi.y)); + float SinTheta = sqrt(1.0 - CosTheta * CosTheta); + + // Convert to spherical direction + vec3 H; + H.x = SinTheta * cos(Phi); + H.y = SinTheta * sin(Phi); + H.z = CosTheta; + + return H; +} + +float DistributionGGX(float NdotH, float roughness4) { + float NdotH2 = NdotH * NdotH; + float denom = (NdotH2 * (roughness4 - 1.0) + 1.0); + denom = M_PI * denom * denom; + + return roughness4 / denom; +} + +// https://graphicrants.blogspot.com.au/2013/08/specular-brdf-reference.html +float GGX(float NdotV, float a) { + float k = a / 2.0; + return NdotV / (NdotV * (1.0 - k) + k); +} + +// https://graphicrants.blogspot.com.au/2013/08/specular-brdf-reference.html +float G_Smith(float a, float nDotV, float nDotL) { + return GGX(nDotL, a * a) * GGX(nDotV, a * a); +} + +float radicalInverse_VdC(uint bits) { + bits = (bits << 16u) | (bits >> 16u); + bits = ((bits & 0x55555555u) << 1u) | ((bits & 0xAAAAAAAAu) >> 1u); + bits = ((bits & 0x33333333u) << 2u) | ((bits & 0xCCCCCCCCu) >> 2u); + bits = ((bits & 0x0F0F0F0Fu) << 4u) | ((bits & 0xF0F0F0F0u) >> 4u); + bits = ((bits & 0x00FF00FFu) << 8u) | ((bits & 0xFF00FF00u) >> 8u); + return float(bits) * 2.3283064365386963e-10; // / 0x100000000 +} + +vec2 Hammersley(uint i, uint N) { + return vec2(float(i) / float(N), radicalInverse_VdC(i)); +} diff --git a/servers/rendering/renderer_rd/shaders/cubemap_roughness_raster.glsl b/servers/rendering/renderer_rd/shaders/cubemap_roughness_raster.glsl new file mode 100644 index 0000000000..c29accd8a7 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/cubemap_roughness_raster.glsl @@ -0,0 +1,79 @@ +/* clang-format off */ +#[vertex] + +#version 450 + +#VERSION_DEFINES + +#include "cubemap_roughness_inc.glsl" + +layout(location = 0) out vec2 uv_interp; +/* clang-format on */ + +void main() { + vec2 base_arr[4] = vec2[](vec2(0.0, 0.0), vec2(0.0, 1.0), vec2(1.0, 1.0), vec2(1.0, 0.0)); + uv_interp = base_arr[gl_VertexIndex]; + gl_Position = vec4(uv_interp * 2.0 - 1.0, 0.0, 1.0); +} + +/* clang-format off */ +#[fragment] + +#version 450 + +#VERSION_DEFINES + +#include "cubemap_roughness_inc.glsl" + +layout(location = 0) in vec2 uv_interp; + +layout(set = 0, binding = 0) uniform samplerCube source_cube; + +layout(location = 0) out vec4 frag_color; +/* clang-format on */ + +void main() { + vec3 N = texelCoordToVec(uv_interp * 2.0 - 1.0, params.face_id); + + //vec4 color = color_interp; + + if (params.use_direct_write) { + frag_color = vec4(texture(source_cube, N).rgb, 1.0); + } else { + vec4 sum = vec4(0.0, 0.0, 0.0, 0.0); + + float solid_angle_texel = 4.0 * M_PI / (6.0 * params.face_size * params.face_size); + float roughness2 = params.roughness * params.roughness; + float roughness4 = roughness2 * roughness2; + vec3 UpVector = abs(N.z) < 0.999 ? vec3(0.0, 0.0, 1.0) : vec3(1.0, 0.0, 0.0); + mat3 T; + T[0] = normalize(cross(UpVector, N)); + T[1] = cross(N, T[0]); + T[2] = N; + + for (uint sampleNum = 0u; sampleNum < params.sample_count; sampleNum++) { + vec2 xi = Hammersley(sampleNum, params.sample_count); + + vec3 H = T * ImportanceSampleGGX(xi, roughness4); + float NdotH = dot(N, H); + vec3 L = (2.0 * NdotH * H - N); + + float ndotl = clamp(dot(N, L), 0.0, 1.0); + + if (ndotl > 0.0) { + float D = DistributionGGX(NdotH, roughness4); + float pdf = D * NdotH / (4.0 * NdotH) + 0.0001; + + float solid_angle_sample = 1.0 / (float(params.sample_count) * pdf + 0.0001); + + float mipLevel = params.roughness == 0.0 ? 0.0 : 0.5 * log2(solid_angle_sample / solid_angle_texel); + + sum.rgb += textureLod(source_cube, L, mipLevel).rgb * ndotl; + sum.a += ndotl; + } + } + sum /= sum.a; + + frag_color = vec4(sum.rgb, 1.0); + } +} diff --git a/servers/rendering/renderer_rd/shaders/decal_data_inc.glsl b/servers/rendering/renderer_rd/shaders/decal_data_inc.glsl new file mode 100644 index 0000000000..158096d3c7 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/decal_data_inc.glsl @@ -0,0 +1,18 @@ + +struct DecalData { + highp mat4 xform; //to decal transform + highp vec3 inv_extents; + mediump float albedo_mix; + highp vec4 albedo_rect; + highp vec4 normal_rect; + highp vec4 orm_rect; + highp vec4 emission_rect; + highp vec4 modulate; + mediump float emission_energy; + uint mask; + mediump float upper_fade; + mediump float lower_fade; + mediump mat3x4 normal_xform; + mediump vec3 normal; + mediump float normal_fade; +}; diff --git a/servers/rendering/renderer_rd/shaders/effects/SCsub b/servers/rendering/renderer_rd/shaders/effects/SCsub new file mode 100644 index 0000000000..fc513d3fb9 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/effects/SCsub @@ -0,0 +1,17 @@ +#!/usr/bin/env python + +Import("env") + +if "RD_GLSL" in env["BUILDERS"]: + # find all include files + gl_include_files = [str(f) for f in Glob("*_inc.glsl")] + + # find all shader code(all glsl files excluding our include files) + glsl_files = [str(f) for f in Glob("*.glsl") if str(f) not in gl_include_files] + + # make sure we recompile shaders if include files change + env.Depends([f + ".gen.h" for f in glsl_files], gl_include_files) + + # compile shaders + for glsl_file in glsl_files: + env.RD_GLSL(glsl_file) diff --git a/servers/rendering/renderer_rd/shaders/effects/blur_raster.glsl b/servers/rendering/renderer_rd/shaders/effects/blur_raster.glsl new file mode 100644 index 0000000000..96f5c3e9f2 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/effects/blur_raster.glsl @@ -0,0 +1,148 @@ +/* clang-format off */ +#[vertex] + +#version 450 + +#VERSION_DEFINES + +#include "blur_raster_inc.glsl" + +layout(location = 0) out vec2 uv_interp; +/* clang-format on */ + +void main() { + vec2 base_arr[4] = vec2[](vec2(0.0, 0.0), vec2(0.0, 1.0), vec2(1.0, 1.0), vec2(1.0, 0.0)); + uv_interp = base_arr[gl_VertexIndex]; + + gl_Position = vec4(uv_interp * 2.0 - 1.0, 0.0, 1.0); +} + +/* clang-format off */ +#[fragment] + +#version 450 + +#VERSION_DEFINES + +#include "blur_raster_inc.glsl" + +layout(location = 0) in vec2 uv_interp; +/* clang-format on */ + +layout(set = 0, binding = 0) uniform sampler2D source_color; + +#ifdef GLOW_USE_AUTO_EXPOSURE +layout(set = 1, binding = 0) uniform sampler2D source_auto_exposure; +#endif + +layout(location = 0) out vec4 frag_color; + +void main() { + // We do not apply our color scale for our mobile renderer here, we'll leave our colors at half brightness and apply scale in the tonemap raster. + +#ifdef MODE_MIPMAP + + vec2 pix_size = blur.pixel_size; + vec4 color = texture(source_color, uv_interp + vec2(-0.5, -0.5) * pix_size); + color += texture(source_color, uv_interp + vec2(0.5, -0.5) * pix_size); + color += texture(source_color, uv_interp + vec2(0.5, 0.5) * pix_size); + color += texture(source_color, uv_interp + vec2(-0.5, 0.5) * pix_size); + frag_color = color / 4.0; + +#endif + +#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; + } +#endif + +#ifdef MODE_GAUSSIAN_GLOW + + //Glow uses larger sigma 1 for a more rounded blur effect + +#define GLOW_ADD(m_ofs, m_mult) \ + { \ + vec2 ofs = uv_interp + m_ofs * pix_size; \ + vec4 c = texture(source_color, ofs) * m_mult; \ + if (any(lessThan(ofs, vec2(0.0))) || any(greaterThan(ofs, vec2(1.0)))) { \ + c *= 0.0; \ + } \ + color += c; \ + } + + if (bool(blur.flags & FLAG_HORIZONTAL)) { + vec2 pix_size = blur.pixel_size; + pix_size *= 0.5; //reading from larger buffer, so use more samples + + vec4 color = texture(source_color, uv_interp + vec2(0.0, 0.0) * pix_size) * 0.174938; + GLOW_ADD(vec2(1.0, 0.0), 0.165569); + GLOW_ADD(vec2(2.0, 0.0), 0.140367); + GLOW_ADD(vec2(3.0, 0.0), 0.106595); + GLOW_ADD(vec2(-1.0, 0.0), 0.165569); + GLOW_ADD(vec2(-2.0, 0.0), 0.140367); + GLOW_ADD(vec2(-3.0, 0.0), 0.106595); + + // only do this in the horizontal pass, if we also do this in the vertical pass we're doubling up. + color *= blur.glow_strength; + + frag_color = color; + } else { + vec2 pix_size = blur.pixel_size; + vec4 color = texture(source_color, uv_interp + vec2(0.0, 0.0) * pix_size) * 0.288713; + GLOW_ADD(vec2(0.0, 1.0), 0.233062); + GLOW_ADD(vec2(0.0, 2.0), 0.122581); + GLOW_ADD(vec2(0.0, -1.0), 0.233062); + GLOW_ADD(vec2(0.0, -2.0), 0.122581); + + frag_color = color; + } + +#undef GLOW_ADD + + if (bool(blur.flags & FLAG_GLOW_FIRST_PASS)) { + // In the first pass bring back to correct color range else we're applying the wrong threshold + // in subsequent passes we can use it as is as we'd just be undoing it right after. + frag_color *= blur.luminance_multiplier; + +#ifdef GLOW_USE_AUTO_EXPOSURE + + frag_color /= texelFetch(source_auto_exposure, ivec2(0, 0), 0).r / blur.glow_auto_exposure_grey; +#endif + frag_color *= blur.glow_exposure; + + float luminance = max(frag_color.r, max(frag_color.g, frag_color.b)); + float feedback = max(smoothstep(blur.glow_hdr_threshold, blur.glow_hdr_threshold + blur.glow_hdr_scale, luminance), blur.glow_bloom); + + frag_color = min(frag_color * feedback, vec4(blur.glow_luminance_cap)) / blur.luminance_multiplier; + } + +#endif // MODE_GAUSSIAN_GLOW + +#ifdef MODE_COPY + vec4 color = textureLod(source_color, uv_interp, 0.0); + frag_color = color; +#endif +} diff --git a/servers/rendering/renderer_rd/shaders/effects/blur_raster_inc.glsl b/servers/rendering/renderer_rd/shaders/effects/blur_raster_inc.glsl new file mode 100644 index 0000000000..730504571a --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/effects/blur_raster_inc.glsl @@ -0,0 +1,26 @@ +#define FLAG_HORIZONTAL (1 << 0) +#define FLAG_USE_ORTHOGONAL_PROJECTION (1 << 1) +#define FLAG_GLOW_FIRST_PASS (1 << 2) + +layout(push_constant, std430) uniform Blur { + vec2 pixel_size; // 08 - 08 + uint flags; // 04 - 12 + uint pad; // 04 - 16 + + // Glow. + float glow_strength; // 04 - 20 + float glow_bloom; // 04 - 24 + float glow_hdr_threshold; // 04 - 28 + float glow_hdr_scale; // 04 - 32 + + float glow_exposure; // 04 - 36 + float glow_white; // 04 - 40 + float glow_luminance_cap; // 04 - 44 + float glow_auto_exposure_grey; // 04 - 48 + + float luminance_multiplier; // 04 - 52 + float res1; // 04 - 56 + float res2; // 04 - 60 + float res3; // 04 - 64 +} +blur; diff --git a/servers/rendering/renderer_rd/shaders/bokeh_dof.glsl b/servers/rendering/renderer_rd/shaders/effects/bokeh_dof.glsl index 63f086a83d..0438671dd2 100644 --- a/servers/rendering/renderer_rd/shaders/bokeh_dof.glsl +++ b/servers/rendering/renderer_rd/shaders/effects/bokeh_dof.glsl @@ -2,7 +2,7 @@ #version 450 -VERSION_DEFINES +#VERSION_DEFINES #define BLOCK_SIZE 8 @@ -25,34 +25,7 @@ layout(set = 1, binding = 0) uniform sampler2D source_bokeh; // based on https://www.shadertoy.com/view/Xd3GDl -layout(push_constant, binding = 1, std430) uniform Params { - ivec2 size; - float z_far; - float z_near; - - bool orthogonal; - float blur_size; - float blur_scale; - int blur_steps; - - bool blur_near_active; - float blur_near_begin; - float blur_near_end; - bool blur_far_active; - - float blur_far_begin; - float blur_far_end; - bool second_pass; - bool half_size; - - bool use_jitter; - float jitter_seed; - uint pad[2]; -} -params; - -//used to work around downsampling filter -#define DEPTH_GAP 0.0 +#include "bokeh_dof_inc.glsl" #ifdef MODE_GEN_BLUR_SIZE @@ -80,15 +53,6 @@ float get_blur_size(float depth) { #endif -const float GOLDEN_ANGLE = 2.39996323; - -//note: uniform pdf rand [0;1[ -float hash12n(vec2 p) { - p = fract(p * vec2(5.3987, 5.4421)); - p += dot(p.yx, p.xy + vec2(21.5351, 14.3137)); - return fract(p.x * p.y * 95.4307); -} - #if defined(MODE_BOKEH_BOX) || defined(MODE_BOKEH_HEXAGONAL) vec4 weighted_filter_dir(vec2 dir, vec2 uv, vec2 pixel_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 new file mode 100644 index 0000000000..b90a527554 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/effects/bokeh_dof_inc.glsl @@ -0,0 +1,37 @@ +layout(push_constant, std430) uniform Params { + ivec2 size; + float z_far; + float z_near; + + bool orthogonal; + float blur_size; + float blur_scale; + int blur_steps; + + bool blur_near_active; + float blur_near_begin; + float blur_near_end; + bool blur_far_active; + + float blur_far_begin; + float blur_far_end; + bool second_pass; + bool half_size; + + bool use_jitter; + float jitter_seed; + uint pad[2]; +} +params; + +//used to work around downsampling filter +#define DEPTH_GAP 0.0 + +const float GOLDEN_ANGLE = 2.39996323; + +//note: uniform pdf rand [0;1[ +float hash12n(vec2 p) { + p = fract(p * vec2(5.3987, 5.4421)); + p += dot(p.yx, p.xy + vec2(21.5351, 14.3137)); + return fract(p.x * p.y * 95.4307); +} diff --git a/servers/rendering/renderer_rd/shaders/effects/bokeh_dof_raster.glsl b/servers/rendering/renderer_rd/shaders/effects/bokeh_dof_raster.glsl new file mode 100644 index 0000000000..a3b3938ee9 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/effects/bokeh_dof_raster.glsl @@ -0,0 +1,253 @@ +/* clang-format off */ +#[vertex] + +#version 450 + +#VERSION_DEFINES + +#include "bokeh_dof_inc.glsl" + +layout(location = 0) out vec2 uv_interp; +/* clang-format on */ + +void main() { + vec2 base_arr[4] = vec2[](vec2(0.0, 0.0), vec2(0.0, 1.0), vec2(1.0, 1.0), vec2(1.0, 0.0)); + uv_interp = base_arr[gl_VertexIndex]; + + gl_Position = vec4(uv_interp * 2.0 - 1.0, 0.0, 1.0); +} + +/* clang-format off */ +#[fragment] + +#version 450 + +#VERSION_DEFINES + +#include "bokeh_dof_inc.glsl" + +layout(location = 0) in vec2 uv_interp; +/* clang-format on */ + +#ifdef MODE_GEN_BLUR_SIZE +layout(location = 0) out float weight; + +layout(set = 0, binding = 0) uniform sampler2D source_depth; +#else +layout(location = 0) out vec4 frag_color; +#ifdef OUTPUT_WEIGHT +layout(location = 1) out float weight; +#endif + +layout(set = 0, binding = 0) uniform sampler2D source_color; +layout(set = 1, binding = 0) uniform sampler2D source_weight; +#ifdef MODE_COMPOSITE_BOKEH +layout(set = 2, binding = 0) uniform sampler2D original_weight; +#endif +#endif + +//DOF +// Bokeh single pass implementation based on https://tuxedolabs.blogspot.com/2018/05/bokeh-depth-of-field-in-single-pass.html + +#ifdef MODE_GEN_BLUR_SIZE + +float get_depth_at_pos(vec2 uv) { + float depth = textureLod(source_depth, uv, 0.0).x; + 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 { + depth = 2.0 * params.z_near * params.z_far / (params.z_far + params.z_near - depth * (params.z_far - params.z_near)); + } + return depth; +} + +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.blur_far_active && depth > params.blur_far_begin) { + return smoothstep(params.blur_far_begin, params.blur_far_end, depth) * params.blur_size + DEPTH_GAP; + } + + return 0.0; +} + +#endif + +#if defined(MODE_BOKEH_BOX) || defined(MODE_BOKEH_HEXAGONAL) + +vec4 weighted_filter_dir(vec2 dir, vec2 uv, vec2 pixel_size) { + dir *= pixel_size; + vec4 color = texture(source_color, uv); + color.a = texture(source_weight, uv).r; + + vec4 accum = color; + float total = 1.0; + + float blur_scale = params.blur_size / float(params.blur_steps); + + if (params.use_jitter) { + uv += dir * (hash12n(uv + params.jitter_seed) - 0.5); + } + + for (int i = -params.blur_steps; i <= params.blur_steps; i++) { + if (i == 0) { + continue; + } + float radius = float(i) * blur_scale; + vec2 suv = uv + dir * radius; + radius = abs(radius); + + vec4 sample_color = texture(source_color, suv); + sample_color.a = texture(source_weight, suv).r; + float limit; + + if (sample_color.a < color.a) { + limit = abs(sample_color.a); + } else { + limit = abs(color.a); + } + + limit -= DEPTH_GAP; + + float m = smoothstep(radius - 0.5, radius + 0.5, limit); + + accum += mix(color, sample_color, m); + + total += 1.0; + } + + return accum / total; +} + +#endif + +void main() { + vec2 pixel_size = 1.0 / vec2(params.size); + vec2 uv = uv_interp; + +#ifdef MODE_GEN_BLUR_SIZE + uv += pixel_size * 0.5; + float center_depth = get_depth_at_pos(uv); + weight = get_blur_size(center_depth); +#endif + +#ifdef MODE_BOKEH_BOX + //pixel_size*=0.5; //resolution is doubled + if (params.second_pass || !params.half_size) { + uv += pixel_size * 0.5; //half pixel to read centers + } else { + uv += pixel_size * 0.25; //half pixel to read centers from full res + } + + float alpha = texture(source_color, uv).a; // retain this + vec2 dir = (params.second_pass ? vec2(0.0, 1.0) : vec2(1.0, 0.0)); + + vec4 color = weighted_filter_dir(dir, uv, pixel_size); + + frag_color = color; + frag_color.a = alpha; // attempt to retain this in case we have a transparent background, ignored if half_size +#ifdef OUTPUT_WEIGHT + weight = color.a; +#endif + +#endif + +#ifdef MODE_BOKEH_HEXAGONAL + + //pixel_size*=0.5; //resolution is doubled + if (params.second_pass || !params.half_size) { + uv += pixel_size * 0.5; //half pixel to read centers + } else { + uv += pixel_size * 0.25; //half pixel to read centers from full res + } + + float alpha = texture(source_color, uv).a; // retain this + + vec2 dir = (params.second_pass ? normalize(vec2(1.0, 0.577350269189626)) : vec2(0.0, 1.0)); + + vec4 color = weighted_filter_dir(dir, uv, pixel_size); + + if (params.second_pass) { + dir = normalize(vec2(-1.0, 0.577350269189626)); + + vec4 color2 = weighted_filter_dir(dir, uv, pixel_size); + + color.rgb = min(color.rgb, color2.rgb); + color.a = (color.a + color2.a) * 0.5; + } + + frag_color = color; + frag_color.a = alpha; // attempt to retain this in case we have a transparent background, ignored if half_size +#ifdef OUTPUT_WEIGHT + weight = color.a; +#endif + +#endif + +#ifdef MODE_BOKEH_CIRCULAR + if (params.half_size) { + pixel_size *= 0.5; //resolution is doubled + } + + uv += pixel_size * 0.5; //half pixel to read centers + + vec4 color = texture(source_color, uv); + float alpha = color.a; // retain this + color.a = texture(source_weight, uv).r; + + vec4 color_accum = color; + float accum = 1.0; + + float radius = params.blur_scale; + for (float ang = 0.0; radius < params.blur_size; ang += GOLDEN_ANGLE) { + vec2 uv_adj = uv + vec2(cos(ang), sin(ang)) * pixel_size * radius; + + 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); + } + + limit -= DEPTH_GAP; + + float m = smoothstep(radius - 0.5, radius + 0.5, limit); + color_accum += mix(color_accum / accum, sample_color, m); + accum += 1.0; + + radius += params.blur_scale / radius; + } + + color_accum = color_accum / accum; + + frag_color.rgb = color_accum.rgb; + frag_color.a = alpha; // attempt to retain this in case we have a transparent background, ignored if half_size +#ifdef OUTPUT_WEIGHT + weight = color_accum.a; +#endif + +#endif + +#ifdef MODE_COMPOSITE_BOKEH + frag_color.rgb = texture(source_color, uv).rgb; + + float center_weigth = texture(source_weight, uv).r; + float sample_weight = texture(original_weight, uv).r; + + float mix_amount; + if (sample_weight < center_weigth) { + mix_amount = min(1.0, max(0.0, max(abs(center_weigth), abs(sample_weight)) - DEPTH_GAP)); + } else { + mix_amount = min(1.0, max(0.0, abs(center_weigth) - DEPTH_GAP)); + } + + // let alpha blending take care of mixing + frag_color.a = mix_amount; +#endif +} diff --git a/servers/rendering/renderer_rd/shaders/copy.glsl b/servers/rendering/renderer_rd/shaders/effects/copy.glsl index cdd35dfb3f..3a4ef86ef0 100644 --- a/servers/rendering/renderer_rd/shaders/copy.glsl +++ b/servers/rendering/renderer_rd/shaders/effects/copy.glsl @@ -2,7 +2,7 @@ #version 450 -VERSION_DEFINES +#VERSION_DEFINES layout(local_size_x = 8, local_size_y = 8, local_size_z = 1) in; @@ -17,7 +17,7 @@ layout(local_size_x = 8, local_size_y = 8, local_size_z = 1) in; #define FLAG_HIGH_QUALITY_GLOW (1 << 8) #define FLAG_ALPHA_TO_ONE (1 << 9) -layout(push_constant, binding = 1, std430) uniform Params { +layout(push_constant, std430) uniform Params { ivec4 section; ivec2 target; uint flags; @@ -61,7 +61,7 @@ layout(rgba8, set = 3, binding = 0) uniform restrict writeonly image2D dest_buff layout(rgba32f, set = 3, binding = 0) uniform restrict writeonly image2D dest_buffer; #endif -#ifdef MODE_GAUSSIAN_GLOW +#ifdef MODE_GAUSSIAN_BLUR shared vec4 local_cache[256]; shared vec4 temp_cache[128]; #endif @@ -70,7 +70,7 @@ void main() { // Pixel being shaded ivec2 pos = ivec2(gl_GlobalInvocationID.xy); -#ifndef MODE_GAUSSIAN_GLOW // Glow needs the extra threads +#ifndef MODE_GAUSSIAN_BLUR // Gaussian blur needs the extra threads if (any(greaterThanEqual(pos, params.section.zw))) { //too large, do nothing return; } @@ -84,41 +84,19 @@ void main() { color += texelFetch(source_color, base_pos + ivec2(1, 0), 0); color += texelFetch(source_color, base_pos + ivec2(1, 1), 0); color /= 4.0; + color = mix(color, vec4(100.0, 100.0, 100.0, 1.0), isinf(color)); + color = mix(color, vec4(100.0, 100.0, 100.0, 1.0), isnan(color)); imageStore(dest_buffer, pos + params.target, color); #endif #ifdef MODE_GAUSSIAN_BLUR - //Simpler blur uses SIGMA2 for the gaussian kernel for a stronger effect - - if (bool(params.flags & FLAG_HORIZONTAL)) { - ivec2 base_pos = (pos + params.section.xy) << 1; - vec4 color = texelFetch(source_color, base_pos + ivec2(0, 0), 0) * 0.214607; - color += texelFetch(source_color, base_pos + ivec2(1, 0), 0) * 0.189879; - color += texelFetch(source_color, base_pos + ivec2(2, 0), 0) * 0.131514; - color += texelFetch(source_color, base_pos + ivec2(3, 0), 0) * 0.071303; - color += texelFetch(source_color, base_pos + ivec2(-1, 0), 0) * 0.189879; - color += texelFetch(source_color, base_pos + ivec2(-2, 0), 0) * 0.131514; - color += texelFetch(source_color, base_pos + ivec2(-3, 0), 0) * 0.071303; - imageStore(dest_buffer, pos + params.target, color); - } else { - ivec2 base_pos = (pos + params.section.xy); - vec4 color = texelFetch(source_color, base_pos + ivec2(0, 0), 0) * 0.38774; - color += texelFetch(source_color, base_pos + ivec2(0, 1), 0) * 0.24477; - color += texelFetch(source_color, base_pos + ivec2(0, 2), 0) * 0.06136; - color += texelFetch(source_color, base_pos + ivec2(0, -1), 0) * 0.24477; - color += texelFetch(source_color, base_pos + ivec2(0, -2), 0) * 0.06136; - imageStore(dest_buffer, pos + params.target, color); - } -#endif - -#ifdef MODE_GAUSSIAN_GLOW - // First pass copy texture into 16x16 local memory for every 8x8 thread block vec2 quad_center_uv = clamp(vec2(gl_GlobalInvocationID.xy + gl_LocalInvocationID.xy - 3.5) / params.section.zw, vec2(0.5 / params.section.zw), vec2(1.0 - 1.5 / params.section.zw)); uint dest_index = gl_LocalInvocationID.x * 2 + gl_LocalInvocationID.y * 2 * 16; +#ifdef MODE_GLOW if (bool(params.flags & FLAG_HIGH_QUALITY_GLOW)) { vec2 quad_offset_uv = clamp((vec2(gl_GlobalInvocationID.xy + gl_LocalInvocationID.xy - 3.0)) / params.section.zw, vec2(0.5 / params.section.zw), vec2(1.0 - 1.5 / params.section.zw)); @@ -126,35 +104,49 @@ void main() { local_cache[dest_index + 1] = (textureLod(source_color, quad_center_uv + vec2(1.0 / params.section.z, 0.0), 0) + textureLod(source_color, quad_offset_uv + vec2(1.0 / params.section.z, 0.0), 0)) * 0.5; local_cache[dest_index + 16] = (textureLod(source_color, quad_center_uv + vec2(0.0, 1.0 / params.section.w), 0) + textureLod(source_color, quad_offset_uv + vec2(0.0, 1.0 / params.section.w), 0)) * 0.5; local_cache[dest_index + 16 + 1] = (textureLod(source_color, quad_center_uv + vec2(1.0 / params.section.zw), 0) + textureLod(source_color, quad_offset_uv + vec2(1.0 / params.section.zw), 0)) * 0.5; - } else { + } else +#endif + { local_cache[dest_index] = textureLod(source_color, quad_center_uv, 0); local_cache[dest_index + 1] = textureLod(source_color, quad_center_uv + vec2(1.0 / params.section.z, 0.0), 0); local_cache[dest_index + 16] = textureLod(source_color, quad_center_uv + vec2(0.0, 1.0 / params.section.w), 0); local_cache[dest_index + 16 + 1] = textureLod(source_color, quad_center_uv + vec2(1.0 / params.section.zw), 0); } - +#ifdef MODE_GLOW + if (bool(params.flags & FLAG_GLOW_FIRST_PASS)) { + // Tonemap initial samples to reduce weight of fireflies: https://graphicrants.blogspot.com/2013/12/tone-mapping.html + local_cache[dest_index] /= 1.0 + dot(local_cache[dest_index].rgb, vec3(0.299, 0.587, 0.114)); + local_cache[dest_index + 1] /= 1.0 + dot(local_cache[dest_index + 1].rgb, vec3(0.299, 0.587, 0.114)); + local_cache[dest_index + 16] /= 1.0 + dot(local_cache[dest_index + 16].rgb, vec3(0.299, 0.587, 0.114)); + local_cache[dest_index + 16 + 1] /= 1.0 + dot(local_cache[dest_index + 16 + 1].rgb, vec3(0.299, 0.587, 0.114)); + } + const float kernel[4] = { 0.174938, 0.165569, 0.140367, 0.106595 }; +#else + // Simpler blur uses SIGMA2 for the gaussian kernel for a stronger effect. + const float kernel[4] = { 0.214607, 0.189879, 0.131514, 0.071303 }; +#endif memoryBarrierShared(); barrier(); // Horizontal pass. Needs to copy into 8x16 chunk of local memory so vertical pass has full resolution uint read_index = gl_LocalInvocationID.x + gl_LocalInvocationID.y * 32 + 4; vec4 color_top = vec4(0.0); - color_top += local_cache[read_index] * 0.174938; - color_top += local_cache[read_index + 1] * 0.165569; - color_top += local_cache[read_index + 2] * 0.140367; - color_top += local_cache[read_index + 3] * 0.106595; - color_top += local_cache[read_index - 1] * 0.165569; - color_top += local_cache[read_index - 2] * 0.140367; - color_top += local_cache[read_index - 3] * 0.106595; + color_top += local_cache[read_index] * kernel[0]; + color_top += local_cache[read_index + 1] * kernel[1]; + color_top += local_cache[read_index + 2] * kernel[2]; + color_top += local_cache[read_index + 3] * kernel[3]; + color_top += local_cache[read_index - 1] * kernel[1]; + color_top += local_cache[read_index - 2] * kernel[2]; + color_top += local_cache[read_index - 3] * kernel[3]; vec4 color_bottom = vec4(0.0); - color_bottom += local_cache[read_index + 16] * 0.174938; - color_bottom += local_cache[read_index + 1 + 16] * 0.165569; - color_bottom += local_cache[read_index + 2 + 16] * 0.140367; - color_bottom += local_cache[read_index + 3 + 16] * 0.106595; - color_bottom += local_cache[read_index - 1 + 16] * 0.165569; - color_bottom += local_cache[read_index - 2 + 16] * 0.140367; - color_bottom += local_cache[read_index - 3 + 16] * 0.106595; + color_bottom += local_cache[read_index + 16] * kernel[0]; + color_bottom += local_cache[read_index + 1 + 16] * kernel[1]; + color_bottom += local_cache[read_index + 2 + 16] * kernel[2]; + color_bottom += local_cache[read_index + 3 + 16] * kernel[3]; + color_bottom += local_cache[read_index - 1 + 16] * kernel[1]; + color_bottom += local_cache[read_index - 2 + 16] * kernel[2]; + color_bottom += local_cache[read_index - 3 + 16] * kernel[3]; // rotate samples to take advantage of cache coherency uint write_index = gl_LocalInvocationID.y * 2 + gl_LocalInvocationID.x * 16; @@ -165,17 +157,28 @@ void main() { memoryBarrierShared(); barrier(); + // If destination outside of texture, can stop doing work now + if (any(greaterThanEqual(pos, params.section.zw))) { + return; + } + // Vertical pass uint index = gl_LocalInvocationID.y + gl_LocalInvocationID.x * 16 + 4; vec4 color = vec4(0.0); - color += temp_cache[index] * 0.174938; - color += temp_cache[index + 1] * 0.165569; - color += temp_cache[index + 2] * 0.140367; - color += temp_cache[index + 3] * 0.106595; - color += temp_cache[index - 1] * 0.165569; - color += temp_cache[index - 2] * 0.140367; - color += temp_cache[index - 3] * 0.106595; + color += temp_cache[index] * kernel[0]; + color += temp_cache[index + 1] * kernel[1]; + color += temp_cache[index + 2] * kernel[2]; + color += temp_cache[index + 3] * kernel[3]; + color += temp_cache[index - 1] * kernel[1]; + color += temp_cache[index - 2] * kernel[2]; + color += temp_cache[index - 3] * kernel[3]; + +#ifdef MODE_GLOW + if (bool(params.flags & FLAG_GLOW_FIRST_PASS)) { + // Undo tonemap to restore range: https://graphicrants.blogspot.com/2013/12/tone-mapping.html + color /= 1.0 - dot(color.rgb, vec3(0.299, 0.587, 0.114)); + } color *= params.glow_strength; @@ -191,7 +194,7 @@ void main() { color = min(color * feedback, vec4(params.glow_luminance_cap)); } - +#endif imageStore(dest_buffer, pos + params.target, color); #endif @@ -256,7 +259,9 @@ void main() { const float PI = 3.14159265359; vec2 uv = vec2(pos) / vec2(params.section.zw); - uv.y = 1.0 - uv.y; + if (bool(params.flags & FLAG_FLIP_Y)) { + uv.y = 1.0 - uv.y; + } float phi = uv.x * 2.0 * PI; float theta = uv.y * PI; diff --git a/servers/rendering/renderer_rd/shaders/copy_to_fb.glsl b/servers/rendering/renderer_rd/shaders/effects/copy_to_fb.glsl index 9751e13b4e..9787c9879d 100644 --- a/servers/rendering/renderer_rd/shaders/copy_to_fb.glsl +++ b/servers/rendering/renderer_rd/shaders/effects/copy_to_fb.glsl @@ -2,11 +2,24 @@ #version 450 -VERSION_DEFINES - +#VERSION_DEFINES + +#ifdef MULTIVIEW +#ifdef has_VK_KHR_multiview +#extension GL_EXT_multiview : enable +#define ViewIndex gl_ViewIndex +#else // has_VK_KHR_multiview +#define ViewIndex 0 +#endif // has_VK_KHR_multiview +#endif //MULTIVIEW + +#ifdef MULTIVIEW +layout(location = 0) out vec3 uv_interp; +#else layout(location = 0) out vec2 uv_interp; +#endif -layout(push_constant, binding = 1, std430) uniform Params { +layout(push_constant, std430) uniform Params { vec4 section; vec2 pixel_size; bool flip_y; @@ -19,9 +32,11 @@ params; void main() { vec2 base_arr[4] = vec2[](vec2(0.0, 0.0), vec2(0.0, 1.0), vec2(1.0, 1.0), vec2(1.0, 0.0)); - uv_interp = base_arr[gl_VertexIndex]; - - vec2 vpos = uv_interp; + uv_interp.xy = base_arr[gl_VertexIndex]; +#ifdef MULTIVIEW + uv_interp.z = ViewIndex; +#endif + vec2 vpos = uv_interp.xy; if (params.use_section) { vpos = params.section.xy + vpos * params.section.zw; } @@ -37,9 +52,18 @@ void main() { #version 450 -VERSION_DEFINES +#VERSION_DEFINES -layout(push_constant, binding = 1, std430) uniform Params { +#ifdef MULTIVIEW +#ifdef has_VK_KHR_multiview +#extension GL_EXT_multiview : enable +#define ViewIndex gl_ViewIndex +#else // has_VK_KHR_multiview +#define ViewIndex 0 +#endif // has_VK_KHR_multiview +#endif //MULTIVIEW + +layout(push_constant, std430) uniform Params { vec4 section; vec2 pixel_size; bool flip_y; @@ -52,12 +76,25 @@ layout(push_constant, binding = 1, std430) uniform Params { } params; +#ifdef MULTIVIEW +layout(location = 0) in vec3 uv_interp; +#else layout(location = 0) in vec2 uv_interp; +#endif +#ifdef MULTIVIEW +layout(set = 0, binding = 0) uniform sampler2DArray source_color; +#ifdef MODE_TWO_SOURCES +layout(set = 1, binding = 0) uniform sampler2DArray source_depth; +layout(location = 1) out float depth; +#endif /* MODE_TWO_SOURCES */ +#else layout(set = 0, binding = 0) uniform sampler2D source_color; #ifdef MODE_TWO_SOURCES layout(set = 1, binding = 0) uniform sampler2D source_color2; -#endif +#endif /* MODE_TWO_SOURCES */ +#endif /* MULTIVIEW */ + layout(location = 0) out vec4 frag_color; vec3 linear_to_srgb(vec3 color) { @@ -68,9 +105,14 @@ vec3 linear_to_srgb(vec3 color) { } void main() { +#ifdef MULTIVIEW + vec3 uv = uv_interp; +#else vec2 uv = uv_interp; +#endif #ifdef MODE_PANORAMA_TO_DP + // Note, multiview and panorama should not be mixed at this time //obtain normal from dual paraboloid uv #define M_PI 3.14159265359 @@ -98,10 +140,20 @@ void main() { uv = 1.0 - uv; } #endif + +#ifdef MULTIVIEW + vec4 color = textureLod(source_color, uv, 0.0); +#ifdef MODE_TWO_SOURCES + // In multiview our 2nd input will be our depth map + depth = textureLod(source_depth, uv, 0.0).r; +#endif /* MODE_TWO_SOURCES */ + +#else vec4 color = textureLod(source_color, uv, 0.0); #ifdef MODE_TWO_SOURCES color += textureLod(source_color2, uv, 0.0); -#endif +#endif /* MODE_TWO_SOURCES */ +#endif /* MULTIVIEW */ if (params.force_luminance) { color.rgb = vec3(max(max(color.r, color.g), color.b)); } diff --git a/servers/rendering/renderer_rd/shaders/tonemap.glsl b/servers/rendering/renderer_rd/shaders/effects/tonemap.glsl index 7de91fd541..62a7b0e7d7 100644 --- a/servers/rendering/renderer_rd/shaders/tonemap.glsl +++ b/servers/rendering/renderer_rd/shaders/effects/tonemap.glsl @@ -2,7 +2,13 @@ #version 450 -VERSION_DEFINES +#VERSION_DEFINES + +#ifdef MULTIVIEW +#ifdef has_VK_KHR_multiview +#extension GL_EXT_multiview : enable +#endif +#endif layout(location = 0) out vec2 uv_interp; @@ -16,20 +22,42 @@ void main() { #version 450 -VERSION_DEFINES +#VERSION_DEFINES + +#ifdef MULTIVIEW +#ifdef has_VK_KHR_multiview +#extension GL_EXT_multiview : enable +#define ViewIndex gl_ViewIndex +#else // has_VK_KHR_multiview +#define ViewIndex 0 +#endif // has_VK_KHR_multiview +#endif //MULTIVIEW layout(location = 0) in vec2 uv_interp; +#ifdef SUBPASS +layout(input_attachment_index = 0, set = 0, binding = 0) uniform subpassInput input_color; +#elif defined(MULTIVIEW) +layout(set = 0, binding = 0) uniform sampler2DArray source_color; +#else layout(set = 0, binding = 0) uniform sampler2D source_color; +#endif + layout(set = 1, binding = 0) uniform sampler2D source_auto_exposure; +#ifdef MULTIVIEW +layout(set = 2, binding = 0) uniform sampler2DArray source_glow; +#else layout(set = 2, binding = 0) uniform sampler2D source_glow; +#endif +layout(set = 2, binding = 1) uniform sampler2D glow_map; + #ifdef USE_1D_LUT layout(set = 3, binding = 0) uniform sampler2D source_color_correction; #else layout(set = 3, binding = 0) uniform sampler3D source_color_correction; #endif -layout(push_constant, binding = 1, std430) uniform Params { +layout(push_constant, std430) uniform Params { vec3 bcs; bool use_bcs; @@ -40,7 +68,7 @@ layout(push_constant, binding = 1, std430) uniform Params { uvec2 glow_texture_size; float glow_intensity; - uint pad3; + float glow_map_strength; uint glow_mode; float glow_levels[7]; @@ -48,7 +76,7 @@ layout(push_constant, binding = 1, std430) uniform Params { float exposure; float white; float auto_exposure_grey; - uint pad2; + float luminance_multiplier; vec2 pixel_size; bool use_fxaa; @@ -94,6 +122,36 @@ float h1(float a) { return 1.0f + w3(a) / (w2(a) + w3(a)); } +#ifdef MULTIVIEW +vec4 texture2D_bicubic(sampler2DArray tex, vec2 uv, int p_lod) { + float lod = float(p_lod); + vec2 tex_size = vec2(params.glow_texture_size >> p_lod); + vec2 pixel_size = vec2(1.0f) / tex_size; + + uv = uv * tex_size + vec2(0.5f); + + vec2 iuv = floor(uv); + vec2 fuv = fract(uv); + + float g0x = g0(fuv.x); + float g1x = g1(fuv.x); + float h0x = h0(fuv.x); + float h1x = h1(fuv.x); + float h0y = h0(fuv.y); + float h1y = h1(fuv.y); + + vec3 p0 = vec3((vec2(iuv.x + h0x, iuv.y + h0y) - vec2(0.5f)) * pixel_size, ViewIndex); + vec3 p1 = vec3((vec2(iuv.x + h1x, iuv.y + h0y) - vec2(0.5f)) * pixel_size, ViewIndex); + vec3 p2 = vec3((vec2(iuv.x + h0x, iuv.y + h1y) - vec2(0.5f)) * pixel_size, ViewIndex); + vec3 p3 = vec3((vec2(iuv.x + h1x, iuv.y + h1y) - vec2(0.5f)) * pixel_size, ViewIndex); + + return (g0(fuv.y) * (g0x * textureLod(tex, p0, lod) + g1x * textureLod(tex, p1, lod))) + + (g1(fuv.y) * (g0x * textureLod(tex, p2, lod) + g1x * textureLod(tex, p3, lod))); +} + +#define GLOW_TEXTURE_SAMPLE(m_tex, m_uv, m_lod) texture2D_bicubic(m_tex, m_uv, m_lod) +#else // MULTIVIEW + vec4 texture2D_bicubic(sampler2D tex, vec2 uv, int p_lod) { float lod = float(p_lod); vec2 tex_size = vec2(params.glow_texture_size >> p_lod); @@ -117,16 +175,21 @@ vec4 texture2D_bicubic(sampler2D tex, vec2 uv, int p_lod) { vec2 p3 = (vec2(iuv.x + h1x, iuv.y + h1y) - vec2(0.5f)) * pixel_size; return (g0(fuv.y) * (g0x * textureLod(tex, p0, lod) + g1x * textureLod(tex, p1, lod))) + - (g1(fuv.y) * (g0x * textureLod(tex, p2, lod) + g1x * textureLod(tex, p3, lod))); + (g1(fuv.y) * (g0x * textureLod(tex, p2, lod) + g1x * textureLod(tex, p3, lod))); } #define GLOW_TEXTURE_SAMPLE(m_tex, m_uv, m_lod) texture2D_bicubic(m_tex, m_uv, m_lod) +#endif // !MULTIVIEW -#else +#else // USE_GLOW_FILTER_BICUBIC +#ifdef MULTIVIEW +#define GLOW_TEXTURE_SAMPLE(m_tex, m_uv, m_lod) textureLod(m_tex, vec3(m_uv, ViewIndex), float(m_lod)) +#else // MULTIVIEW #define GLOW_TEXTURE_SAMPLE(m_tex, m_uv, m_lod) textureLod(m_tex, m_uv, float(m_lod)) +#endif // !MULTIVIEW -#endif +#endif // !USE_GLOW_FILTER_BICUBIC vec3 tonemap_filmic(vec3 color, float white) { // exposure bias: input scale (color *= bias, white *= bias) to make the brightness consistent with other tonemappers @@ -146,25 +209,38 @@ vec3 tonemap_filmic(vec3 color, float white) { return color_tonemapped / white_tonemapped; } +// Adapted from https://github.com/TheRealMJP/BakingLab/blob/master/BakingLab/ACES.hlsl +// (MIT License). vec3 tonemap_aces(vec3 color, float white) { - const float exposure_bias = 0.85f; - const float A = 2.51f * exposure_bias * exposure_bias; - const float B = 0.03f * exposure_bias; - const float C = 2.43f * exposure_bias * exposure_bias; - const float D = 0.59f * exposure_bias; - const float E = 0.14f; - - vec3 color_tonemapped = (color * (A * color + B)) / (color * (C * color + D) + E); - float white_tonemapped = (white * (A * white + B)) / (white * (C * white + D) + E); + const float exposure_bias = 1.8f; + const float A = 0.0245786f; + const float B = 0.000090537f; + const float C = 0.983729f; + const float D = 0.432951f; + const float E = 0.238081f; + + // Exposure bias baked into transform to save shader instructions. Equivalent to `color *= exposure_bias` + const mat3 rgb_to_rrt = mat3( + vec3(0.59719f * exposure_bias, 0.35458f * exposure_bias, 0.04823f * exposure_bias), + vec3(0.07600f * exposure_bias, 0.90834f * exposure_bias, 0.01566f * exposure_bias), + vec3(0.02840f * exposure_bias, 0.13383f * exposure_bias, 0.83777f * exposure_bias)); + + const mat3 odt_to_rgb = mat3( + vec3(1.60475f, -0.53108f, -0.07367f), + vec3(-0.10208f, 1.10813f, -0.00605f), + vec3(-0.00327f, -0.07276f, 1.07602f)); + + color *= rgb_to_rrt; + vec3 color_tonemapped = (color * (color + A) - B) / (color * (C * color + D) + E); + color_tonemapped *= odt_to_rgb; + + white *= exposure_bias; + float white_tonemapped = (white * (white + A) - B) / (white * (C * white + D) + E); return color_tonemapped / white_tonemapped; } vec3 tonemap_reinhard(vec3 color, float white) { - // Ensure color values are positive. - // They can be negative in the case of negative lights, which leads to undesired behavior. - color = max(vec3(0.0), color); - return (white * color + color) / (color * white + white); } @@ -181,19 +257,24 @@ vec3 linear_to_srgb(vec3 color) { #define TONEMAPPER_ACES 3 vec3 apply_tonemapping(vec3 color, float white) { // inputs are LINEAR, always outputs clamped [0;1] color - + // Ensure color values passed to tonemappers are positive. + // They can be negative in the case of negative lights, which leads to undesired behavior. if (params.tonemapper == TONEMAPPER_LINEAR) { return color; } else if (params.tonemapper == TONEMAPPER_REINHARD) { - return tonemap_reinhard(color, white); + return tonemap_reinhard(max(vec3(0.0f), color), white); } else if (params.tonemapper == TONEMAPPER_FILMIC) { - return tonemap_filmic(color, white); - } else { //aces - return tonemap_aces(color, white); + return tonemap_filmic(max(vec3(0.0f), color), white); + } else { // TONEMAPPER_ACES + return tonemap_aces(max(vec3(0.0f), color), white); } } +#ifdef MULTIVIEW +vec3 gather_glow(sampler2DArray tex, vec2 uv) { // sample all selected glow levels, view is added to uv later +#else vec3 gather_glow(sampler2D tex, vec2 uv) { // sample all selected glow levels +#endif // defined(MULTIVIEW) vec3 glow = vec3(0.0f); if (params.glow_levels[0] > 0.0001) { @@ -272,15 +353,23 @@ vec3 apply_color_correction(vec3 color) { } #endif +#ifndef SUBPASS vec3 do_fxaa(vec3 color, float exposure, vec2 uv_interp) { const float FXAA_REDUCE_MIN = (1.0 / 128.0); const float FXAA_REDUCE_MUL = (1.0 / 8.0); const float FXAA_SPAN_MAX = 8.0; - vec3 rgbNW = textureLod(source_color, uv_interp + vec2(-1.0, -1.0) * params.pixel_size, 0.0).xyz * exposure; - vec3 rgbNE = textureLod(source_color, uv_interp + vec2(1.0, -1.0) * params.pixel_size, 0.0).xyz * exposure; - vec3 rgbSW = textureLod(source_color, uv_interp + vec2(-1.0, 1.0) * params.pixel_size, 0.0).xyz * exposure; - vec3 rgbSE = textureLod(source_color, uv_interp + vec2(1.0, 1.0) * params.pixel_size, 0.0).xyz * exposure; +#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; +#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; +#endif vec3 rgbM = color; vec3 luma = vec3(0.299, 0.587, 0.114); float lumaNW = dot(rgbNW, luma); @@ -296,17 +385,22 @@ vec3 do_fxaa(vec3 color, float exposure, vec2 uv_interp) { dir.y = ((lumaNW + lumaSW) - (lumaNE + lumaSE)); float dirReduce = max((lumaNW + lumaNE + lumaSW + lumaSE) * - (0.25 * FXAA_REDUCE_MUL), + (0.25 * FXAA_REDUCE_MUL), FXAA_REDUCE_MIN); float rcpDirMin = 1.0 / (min(abs(dir.x), abs(dir.y)) + dirReduce); dir = min(vec2(FXAA_SPAN_MAX, FXAA_SPAN_MAX), max(vec2(-FXAA_SPAN_MAX, -FXAA_SPAN_MAX), dir * rcpDirMin)) * - params.pixel_size; + params.pixel_size; - vec3 rgbA = 0.5 * exposure * (textureLod(source_color, uv_interp + dir * (1.0 / 3.0 - 0.5), 0.0).xyz + textureLod(source_color, uv_interp + dir * (2.0 / 3.0 - 0.5), 0.0).xyz); - vec3 rgbB = rgbA * 0.5 + 0.25 * exposure * (textureLod(source_color, uv_interp + dir * -0.5, 0.0).xyz + textureLod(source_color, uv_interp + dir * 0.5, 0.0).xyz); +#ifdef MULTIVIEW + vec3 rgbA = 0.5 * exposure * (textureLod(source_color, vec3(uv_interp + dir * (1.0 / 3.0 - 0.5), ViewIndex), 0.0).xyz + textureLod(source_color, vec3(uv_interp + dir * (2.0 / 3.0 - 0.5), ViewIndex), 0.0).xyz) * params.luminance_multiplier; + vec3 rgbB = rgbA * 0.5 + 0.25 * exposure * (textureLod(source_color, vec3(uv_interp + dir * -0.5, ViewIndex), 0.0).xyz + textureLod(source_color, vec3(uv_interp + dir * 0.5, ViewIndex), 0.0).xyz) * params.luminance_multiplier; +#else + vec3 rgbA = 0.5 * exposure * (textureLod(source_color, uv_interp + dir * (1.0 / 3.0 - 0.5), 0.0).xyz + textureLod(source_color, uv_interp + dir * (2.0 / 3.0 - 0.5), 0.0).xyz) * params.luminance_multiplier; + vec3 rgbB = rgbA * 0.5 + 0.25 * exposure * (textureLod(source_color, uv_interp + dir * -0.5, 0.0).xyz + textureLod(source_color, uv_interp + dir * 0.5, 0.0).xyz) * params.luminance_multiplier; +#endif float lumaB = dot(rgbB, luma); if ((lumaB < lumaMin) || (lumaB > lumaMax)) { @@ -315,8 +409,9 @@ vec3 do_fxaa(vec3 color, float exposure, vec2 uv_interp) { return rgbB; } } +#endif // !SUBPASS -// From http://alex.vlachos.com/graphics/Alex_Vlachos_Advanced_VR_Rendering_GDC2015.pdf +// From https://alex.vlachos.com/graphics/Alex_Vlachos_Advanced_VR_Rendering_GDC2015.pdf // and https://www.shadertoy.com/view/MslGR8 (5th one starting from the bottom) // NOTE: `frag_coord` is in pixels (i.e. not normalized UV). vec3 screen_space_dither(vec2 frag_coord) { @@ -329,58 +424,79 @@ vec3 screen_space_dither(vec2 frag_coord) { } void main() { - vec3 color = textureLod(source_color, uv_interp, 0.0f).rgb; +#ifdef SUBPASS + // SUBPASS and MULTIVIEW can be combined but in that case we're already reading from the correct layer + vec4 color = subpassLoad(input_color); +#elif defined(MULTIVIEW) + vec4 color = textureLod(source_color, vec3(uv_interp, ViewIndex), 0.0f); +#else + vec4 color = textureLod(source_color, uv_interp, 0.0f); +#endif + color.rgb *= params.luminance_multiplier; // Exposure float exposure = params.exposure; +#ifndef SUBPASS if (params.use_auto_exposure) { - exposure *= 1.0 / (texelFetch(source_auto_exposure, ivec2(0, 0), 0).r / params.auto_exposure_grey); + exposure *= 1.0 / (texelFetch(source_auto_exposure, ivec2(0, 0), 0).r * params.luminance_multiplier / params.auto_exposure_grey); } +#endif - color *= exposure; + color.rgb *= exposure; // Early Tonemap & SRGB Conversion +#ifndef SUBPASS + if (params.use_fxaa) { + // FXAA must be performed before glow to preserve the "bleed" effect of glow. + color.rgb = do_fxaa(color.rgb, exposure, uv_interp); + } if (params.use_glow && params.glow_mode == GLOW_MODE_MIX) { - vec3 glow = gather_glow(source_glow, uv_interp); + vec3 glow = gather_glow(source_glow, uv_interp) * params.luminance_multiplier; + if (params.glow_map_strength > 0.001) { + glow = mix(glow, texture(glow_map, uv_interp).rgb * glow, params.glow_map_strength); + } color.rgb = mix(color.rgb, glow, params.glow_intensity); } +#endif - if (params.use_fxaa) { - color = do_fxaa(color, exposure, uv_interp); - } if (params.use_debanding) { // For best results, debanding should be done before tonemapping. // Otherwise, we're adding noise to an already-quantized image. - color += screen_space_dither(gl_FragCoord.xy); + color.rgb += screen_space_dither(gl_FragCoord.xy); } - color = apply_tonemapping(color, params.white); - color = linear_to_srgb(color); // regular linear -> SRGB conversion + color.rgb = apply_tonemapping(color.rgb, params.white); - // Glow + color.rgb = linear_to_srgb(color.rgb); // regular linear -> SRGB conversion +#ifndef SUBPASS + // Glow if (params.use_glow && params.glow_mode != GLOW_MODE_MIX) { - vec3 glow = gather_glow(source_glow, uv_interp) * params.glow_intensity; + vec3 glow = gather_glow(source_glow, uv_interp) * params.glow_intensity * params.luminance_multiplier; + if (params.glow_map_strength > 0.001) { + glow = mix(glow, texture(glow_map, uv_interp).rgb * glow, params.glow_map_strength); + } // high dynamic range -> SRGB glow = apply_tonemapping(glow, params.white); glow = linear_to_srgb(glow); - color = apply_glow(color, glow); + color.rgb = apply_glow(color.rgb, glow); } +#endif // Additional effects if (params.use_bcs) { - color = apply_bcs(color, params.bcs); + color.rgb = apply_bcs(color.rgb, params.bcs); } if (params.use_color_correction) { - color = apply_color_correction(color); + color.rgb = apply_color_correction(color.rgb); } - frag_color = vec4(color, 1.0f); + frag_color = color; } diff --git a/servers/rendering/renderer_rd/shaders/fsr_upscale.glsl b/servers/rendering/renderer_rd/shaders/fsr_upscale.glsl new file mode 100644 index 0000000000..c8eb78a2f0 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/fsr_upscale.glsl @@ -0,0 +1,173 @@ +/*************************************************************************/ +/* fsr_upscale.glsl */ +/*************************************************************************/ +/* This file is part of: */ +/* GODOT ENGINE */ +/* https://godotengine.org */ +/*************************************************************************/ +/* Copyright (c) 2007-2022 Juan Linietsky, Ariel Manzur. */ +/* Copyright (c) 2014-2022 Godot Engine contributors (cf. AUTHORS.md). */ +/* */ +/* Permission is hereby granted, free of charge, to any person obtaining */ +/* a copy of this software and associated documentation files (the */ +/* "Software"), to deal in the Software without restriction, including */ +/* without limitation the rights to use, copy, modify, merge, publish, */ +/* distribute, sublicense, and/or sell copies of the Software, and to */ +/* permit persons to whom the Software is furnished to do so, subject to */ +/* the following conditions: */ +/* */ +/* The above copyright notice and this permission notice shall be */ +/* included in all copies or substantial portions of the Software. */ +/* */ +/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, */ +/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF */ +/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.*/ +/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY */ +/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, */ +/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE */ +/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ +/*************************************************************************/ + +#[compute] + +#version 450 + +#VERSION_DEFINES + +#define A_GPU +#define A_GLSL + +#ifdef MODE_FSR_UPSCALE_NORMAL + +#define A_HALF + +#endif + +#include "thirdparty/amd-fsr/ffx_a.h" + +layout(local_size_x = 64, local_size_y = 1, local_size_z = 1) in; + +layout(rgba16f, set = 1, binding = 0) uniform restrict writeonly image2D fsr_image; +layout(set = 0, binding = 0) uniform sampler2D source_image; + +#define FSR_UPSCALE_PASS_TYPE_EASU 0 +#define FSR_UPSCALE_PASS_TYPE_RCAS 1 + +layout(push_constant, std430) uniform Params { + float resolution_width; + float resolution_height; + float upscaled_width; + float upscaled_height; + float sharpness; + int pass; +} +params; + +AU4 Const0, Const1, Const2, Const3; + +#ifdef MODE_FSR_UPSCALE_FALLBACK + +#define FSR_EASU_F +AF4 FsrEasuRF(AF2 p) { + AF4 res = textureGather(source_image, p, 0); + return res; +} +AF4 FsrEasuGF(AF2 p) { + AF4 res = textureGather(source_image, p, 1); + return res; +} +AF4 FsrEasuBF(AF2 p) { + AF4 res = textureGather(source_image, p, 2); + return res; +} + +#define FSR_RCAS_F +AF4 FsrRcasLoadF(ASU2 p) { + return AF4(texelFetch(source_image, ASU2(p), 0)); +} +void FsrRcasInputF(inout AF1 r, inout AF1 g, inout AF1 b) {} + +#else + +#define FSR_EASU_H +AH4 FsrEasuRH(AF2 p) { + AH4 res = AH4(textureGather(source_image, p, 0)); + return res; +} +AH4 FsrEasuGH(AF2 p) { + AH4 res = AH4(textureGather(source_image, p, 1)); + return res; +} +AH4 FsrEasuBH(AF2 p) { + AH4 res = AH4(textureGather(source_image, p, 2)); + return res; +} + +#define FSR_RCAS_H +AH4 FsrRcasLoadH(ASW2 p) { + return AH4(texelFetch(source_image, ASU2(p), 0)); +} +void FsrRcasInputH(inout AH1 r, inout AH1 g, inout AH1 b) {} + +#endif + +#include "thirdparty/amd-fsr/ffx_fsr1.h" + +void fsr_easu_pass(AU2 pos) { +#ifdef MODE_FSR_UPSCALE_NORMAL + + AH3 Gamma2Color = AH3(0, 0, 0); + FsrEasuH(Gamma2Color, pos, Const0, Const1, Const2, Const3); + imageStore(fsr_image, ASU2(pos), AH4(Gamma2Color, 1)); + +#else + + AF3 Gamma2Color = AF3(0, 0, 0); + FsrEasuF(Gamma2Color, pos, Const0, Const1, Const2, Const3); + imageStore(fsr_image, ASU2(pos), AF4(Gamma2Color, 1)); + +#endif +} + +void fsr_rcas_pass(AU2 pos) { +#ifdef MODE_FSR_UPSCALE_NORMAL + + AH3 Gamma2Color = AH3(0, 0, 0); + FsrRcasH(Gamma2Color.r, Gamma2Color.g, Gamma2Color.b, pos, Const0); + imageStore(fsr_image, ASU2(pos), AH4(Gamma2Color, 1)); + +#else + + AF3 Gamma2Color = AF3(0, 0, 0); + FsrRcasF(Gamma2Color.r, Gamma2Color.g, Gamma2Color.b, pos, Const0); + imageStore(fsr_image, ASU2(pos), AF4(Gamma2Color, 1)); + +#endif +} + +void fsr_pass(AU2 pos) { + if (params.pass == FSR_UPSCALE_PASS_TYPE_EASU) { + fsr_easu_pass(pos); + } else if (params.pass == FSR_UPSCALE_PASS_TYPE_RCAS) { + fsr_rcas_pass(pos); + } +} + +void main() { + // Clang does not like unused functions. If ffx_a.h is included in the binary, clang will throw a fit and not compile so we must configure FSR in this shader + if (params.pass == FSR_UPSCALE_PASS_TYPE_EASU) { + FsrEasuCon(Const0, Const1, Const2, Const3, params.resolution_width, params.resolution_height, params.resolution_width, params.resolution_height, params.upscaled_width, params.upscaled_height); + } else if (params.pass == FSR_UPSCALE_PASS_TYPE_RCAS) { + FsrRcasCon(Const0, params.sharpness); + } + + AU2 gxy = ARmp8x8(gl_LocalInvocationID.x) + AU2(gl_WorkGroupID.x << 4u, gl_WorkGroupID.y << 4u); + + fsr_pass(gxy); + gxy.x += 8u; + fsr_pass(gxy); + gxy.y += 8u; + fsr_pass(gxy); + gxy.x -= 8u; + fsr_pass(gxy); +} diff --git a/servers/rendering/renderer_rd/shaders/gi.glsl b/servers/rendering/renderer_rd/shaders/gi.glsl index 92a5682572..0c7f08813b 100644 --- a/servers/rendering/renderer_rd/shaders/gi.glsl +++ b/servers/rendering/renderer_rd/shaders/gi.glsl @@ -2,7 +2,7 @@ #version 450 -VERSION_DEFINES +#VERSION_DEFINES layout(local_size_x = 8, local_size_y = 8, local_size_z = 1) in; @@ -35,7 +35,7 @@ layout(set = 0, binding = 11) uniform texture2DArray lightprobe_texture; layout(set = 0, binding = 12) uniform texture2D depth_buffer; layout(set = 0, binding = 13) uniform texture2D normal_roughness_buffer; -layout(set = 0, binding = 14) uniform utexture2D giprobe_buffer; +layout(set = 0, binding = 14) uniform utexture2D voxel_gi_buffer; layout(set = 0, binding = 15, std140) uniform SDFGI { vec3 grid_size; @@ -65,44 +65,38 @@ layout(set = 0, binding = 15, std140) uniform SDFGI { } sdfgi; -#define MAX_GI_PROBES 8 +#define MAX_VOXEL_GI_INSTANCES 8 -struct GIProbeData { - mat4 xform; - vec3 bounds; - float dynamic_range; +struct VoxelGIData { + mat4 xform; // 64 - 64 - float bias; - float normal_bias; - bool blend_ambient; - uint texture_slot; + vec3 bounds; // 12 - 76 + float dynamic_range; // 4 - 80 - float anisotropy_strength; - float ambient_occlusion; - float ambient_occlusion_size; - uint mipmaps; + float bias; // 4 - 84 + float normal_bias; // 4 - 88 + bool blend_ambient; // 4 - 92 + uint mipmaps; // 4 - 96 }; -layout(set = 0, binding = 16, std140) uniform GIProbes { - GIProbeData data[MAX_GI_PROBES]; +layout(set = 0, binding = 16, std140) uniform VoxelGIs { + VoxelGIData data[MAX_VOXEL_GI_INSTANCES]; } -gi_probes; +voxel_gi_instances; -layout(set = 0, binding = 17) uniform texture3D gi_probe_textures[MAX_GI_PROBES]; +layout(set = 0, binding = 17) uniform texture3D voxel_gi_textures[MAX_VOXEL_GI_INSTANCES]; -layout(push_constant, binding = 0, std430) uniform Params { +layout(push_constant, std430) uniform Params { ivec2 screen_size; float z_near; float z_far; vec4 proj_info; - vec3 ao_color; - uint max_giprobes; - + uint max_voxel_gi_instances; bool high_quality_vct; bool orthogonal; - uint pad[2]; + uint pad; mat3x4 cam_rotation; } @@ -155,7 +149,7 @@ vec3 reconstruct_position(ivec2 screen_pos) { return pos; } -void sdfgi_probe_process(uint cascade, vec3 cascade_pos, vec3 cam_pos, vec3 cam_normal, vec3 cam_specular_normal, float roughness, out vec3 diffuse_light, out vec3 specular_light) { +void sdfvoxel_gi_process(uint cascade, vec3 cascade_pos, vec3 cam_pos, vec3 cam_normal, vec3 cam_specular_normal, float roughness, out vec3 diffuse_light, out vec3 specular_light) { cascade_pos += cam_normal * sdfgi.normal_bias; vec3 base_pos = floor(cascade_pos); @@ -293,7 +287,7 @@ void sdfgi_process(vec3 vertex, vec3 normal, vec3 reflection, float roughness, o float blend; vec3 diffuse, specular; - sdfgi_probe_process(cascade, cascade_pos, cam_pos, cam_normal, reflection, roughness, diffuse, specular); + sdfvoxel_gi_process(cascade, cascade_pos, cam_pos, cam_normal, reflection, roughness, diffuse, specular); { //process blend @@ -323,7 +317,7 @@ void sdfgi_process(vec3 vertex, vec3 normal, vec3 reflection, float roughness, o } else { vec3 diffuse2, specular2; cascade_pos = (cam_pos - sdfgi.cascades[cascade + 1].position) * sdfgi.cascades[cascade + 1].to_probe; - sdfgi_probe_process(cascade + 1, cascade_pos, cam_pos, cam_normal, reflection, roughness, diffuse2, specular2); + sdfvoxel_gi_process(cascade + 1, cascade_pos, cam_pos, cam_normal, reflection, roughness, diffuse2, specular2); diffuse = mix(diffuse, diffuse2, blend); specular = mix(specular, specular2, blend); } @@ -494,26 +488,26 @@ vec4 voxel_cone_trace_45_degrees(texture3D probe, vec3 cell_size, vec3 pos, vec3 return color; } -void gi_probe_compute(uint index, vec3 position, vec3 normal, vec3 ref_vec, mat3 normal_xform, float roughness, inout vec4 out_spec, inout vec4 out_diff, inout float out_blend) { - position = (gi_probes.data[index].xform * vec4(position, 1.0)).xyz; - ref_vec = normalize((gi_probes.data[index].xform * vec4(ref_vec, 0.0)).xyz); - normal = normalize((gi_probes.data[index].xform * vec4(normal, 0.0)).xyz); +void voxel_gi_compute(uint index, vec3 position, vec3 normal, vec3 ref_vec, mat3 normal_xform, float roughness, inout vec4 out_spec, inout vec4 out_diff, inout float out_blend) { + position = (voxel_gi_instances.data[index].xform * vec4(position, 1.0)).xyz; + ref_vec = normalize((voxel_gi_instances.data[index].xform * vec4(ref_vec, 0.0)).xyz); + normal = normalize((voxel_gi_instances.data[index].xform * vec4(normal, 0.0)).xyz); - position += normal * gi_probes.data[index].normal_bias; + position += normal * voxel_gi_instances.data[index].normal_bias; //this causes corrupted pixels, i have no idea why.. - if (any(bvec2(any(lessThan(position, vec3(0.0))), any(greaterThan(position, gi_probes.data[index].bounds))))) { + if (any(bvec2(any(lessThan(position, vec3(0.0))), any(greaterThan(position, voxel_gi_instances.data[index].bounds))))) { return; } - mat3 dir_xform = mat3(gi_probes.data[index].xform) * normal_xform; + mat3 dir_xform = mat3(voxel_gi_instances.data[index].xform) * normal_xform; - vec3 blendv = abs(position / gi_probes.data[index].bounds * 2.0 - 1.0); + vec3 blendv = abs(position / voxel_gi_instances.data[index].bounds * 2.0 - 1.0); float blend = clamp(1.0 - max(blendv.x, max(blendv.y, blendv.z)), 0.0, 1.0); //float blend=1.0; - float max_distance = length(gi_probes.data[index].bounds); - vec3 cell_size = 1.0 / gi_probes.data[index].bounds; + float max_distance = length(voxel_gi_instances.data[index].bounds); + vec3 cell_size = 1.0 / voxel_gi_instances.data[index].bounds; //irradiance @@ -534,7 +528,7 @@ void gi_probe_compute(uint index, vec3 position, vec3 normal, vec3 ref_vec, mat3 for (uint i = 0; i < cone_dir_count; i++) { vec3 dir = normalize(dir_xform * cone_dirs[i]); - light += cone_weights[i] * voxel_cone_trace(gi_probe_textures[index], cell_size, position, dir, cone_angle_tan, max_distance, gi_probes.data[index].bias); + light += cone_weights[i] * voxel_cone_trace(voxel_gi_textures[index], cell_size, position, dir, cone_angle_tan, max_distance, voxel_gi_instances.data[index].bias); } } else { const uint cone_dir_count = 4; @@ -547,42 +541,21 @@ void gi_probe_compute(uint index, vec3 position, vec3 normal, vec3 ref_vec, mat3 float cone_weights[cone_dir_count] = float[](0.25, 0.25, 0.25, 0.25); for (int i = 0; i < cone_dir_count; i++) { vec3 dir = normalize(dir_xform * cone_dirs[i]); - light += cone_weights[i] * voxel_cone_trace_45_degrees(gi_probe_textures[index], cell_size, position, dir, max_distance, gi_probes.data[index].bias); - } - } - - if (gi_probes.data[index].ambient_occlusion > 0.001) { - float size = 1.0 + gi_probes.data[index].ambient_occlusion_size * 7.0; - - float taps, blend; - blend = modf(size, taps); - float ao = 0.0; - for (float i = 1.0; i <= taps; i++) { - vec3 ofs = (position + normal * (i * 0.5 + 1.0)) * cell_size; - ao += textureLod(sampler3D(gi_probe_textures[index], linear_sampler_with_mipmaps), ofs, i - 1.0).a * i; - } - - if (blend > 0.001) { - vec3 ofs = (position + normal * ((taps + 1.0) * 0.5 + 1.0)) * cell_size; - ao += textureLod(sampler3D(gi_probe_textures[index], linear_sampler_with_mipmaps), ofs, taps).a * (taps + 1.0) * blend; + light += cone_weights[i] * voxel_cone_trace_45_degrees(voxel_gi_textures[index], cell_size, position, dir, max_distance, voxel_gi_instances.data[index].bias); } - - ao = 1.0 - min(1.0, ao); - - light.rgb = mix(params.ao_color, light.rgb, mix(1.0, ao, gi_probes.data[index].ambient_occlusion)); } - light.rgb *= gi_probes.data[index].dynamic_range; - if (!gi_probes.data[index].blend_ambient) { + light.rgb *= voxel_gi_instances.data[index].dynamic_range; + if (!voxel_gi_instances.data[index].blend_ambient) { light.a = 1.0; } out_diff += light * blend; //radiance - vec4 irr_light = voxel_cone_trace(gi_probe_textures[index], cell_size, position, ref_vec, tan(roughness * 0.5 * M_PI * 0.99), max_distance, gi_probes.data[index].bias); - irr_light.rgb *= gi_probes.data[index].dynamic_range; - if (!gi_probes.data[index].blend_ambient) { + 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; + if (!voxel_gi_instances.data[index].blend_ambient) { irr_light.a = 1.0; } @@ -614,9 +587,9 @@ void process_gi(ivec2 pos, vec3 vertex, inout vec4 ambient_light, inout vec4 ref sdfgi_process(vertex, normal, reflection, roughness, ambient_light, reflection_light); #endif -#ifdef USE_GIPROBES +#ifdef USE_VOXEL_GI_INSTANCES { - uvec2 giprobe_tex = texelFetch(usampler2D(giprobe_buffer, linear_sampler), pos, 0).rg; + uvec2 voxel_gi_tex = texelFetch(usampler2D(voxel_gi_buffer, linear_sampler), pos, 0).rg; 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); @@ -628,9 +601,9 @@ void process_gi(ivec2 pos, vec3 vertex, inout vec4 ambient_light, inout vec4 ref vec4 spec_accum = vec4(0.0); float blend_accum = 0.0; - for (uint i = 0; i < params.max_giprobes; i++) { - if (any(equal(uvec2(i), giprobe_tex))) { - gi_probe_compute(i, vertex, normal, reflection, normal_mat, roughness, spec_accum, amb_accum, blend_accum); + for (uint i = 0; i < params.max_voxel_gi_instances; i++) { + if (any(equal(uvec2(i), voxel_gi_tex))) { + voxel_gi_compute(i, vertex, normal, reflection, normal_mat, roughness, spec_accum, amb_accum, blend_accum); } } if (blend_accum > 0.0) { diff --git a/servers/rendering/renderer_rd/shaders/giprobe_write.glsl b/servers/rendering/renderer_rd/shaders/giprobe_write.glsl index 56b3b7ccb4..6c73864bf6 100644 --- a/servers/rendering/renderer_rd/shaders/giprobe_write.glsl +++ b/servers/rendering/renderer_rd/shaders/giprobe_write.glsl @@ -2,12 +2,11 @@ #version 450 -VERSION_DEFINES +#VERSION_DEFINES layout(local_size_x = 64, local_size_y = 1, local_size_z = 1) in; #define NO_CHILDREN 0xFFFFFFFF -#define GREY_VEC vec3(0.33333, 0.33333, 0.33333) struct CellChildren { uint children[8]; @@ -59,7 +58,7 @@ lights; #endif -layout(push_constant, binding = 0, std430) uniform Params { +layout(push_constant, std430) uniform Params { ivec3 limits; uint stack_size; @@ -202,12 +201,7 @@ void main() { vec3 emission = vec3(ivec3(cell_data.data[cell_index].emission & 0x3FF, (cell_data.data[cell_index].emission >> 10) & 0x7FF, cell_data.data[cell_index].emission >> 21)) * params.emission_scale; vec4 normal = unpackSnorm4x8(cell_data.data[cell_index].normal); -#ifdef MODE_ANISOTROPIC - vec3 accum[6] = vec3[](vec3(0.0), vec3(0.0), vec3(0.0), vec3(0.0), vec3(0.0), vec3(0.0)); - const vec3 accum_dirs[6] = vec3[](vec3(1.0, 0.0, 0.0), vec3(-1.0, 0.0, 0.0), vec3(0.0, 1.0, 0.0), vec3(0.0, -1.0, 0.0), vec3(0.0, 0.0, 1.0), vec3(0.0, 0.0, -1.0)); -#else vec3 accum = vec3(0.0); -#endif for (uint i = 0; i < params.light_count; i++) { float attenuation; @@ -242,77 +236,35 @@ void main() { vec3 light = lights.data[i].color * albedo.rgb * attenuation * lights.data[i].energy; -#ifdef MODE_ANISOTROPIC - for (uint j = 0; j < 6; j++) { - accum[j] += max(0.0, dot(accum_dir, -light_dir)) * light + emission; - } -#else if (length(normal.xyz) > 0.2) { accum += max(0.0, dot(normal.xyz, -light_dir)) * light + emission; } else { //all directions accum += light + emission; } -#endif } -#ifdef MODE_ANISOTROPIC - - output.data[cell_index * 6 + 0] = vec4(accum[0], 0.0); - output.data[cell_index * 6 + 1] = vec4(accum[1], 0.0); - output.data[cell_index * 6 + 2] = vec4(accum[2], 0.0); - output.data[cell_index * 6 + 3] = vec4(accum[3], 0.0); - output.data[cell_index * 6 + 4] = vec4(accum[4], 0.0); - output.data[cell_index * 6 + 5] = vec4(accum[5], 0.0); -#else output.data[cell_index] = vec4(accum, 0.0); -#endif - #endif //MODE_COMPUTE_LIGHT #ifdef MODE_UPDATE_MIPMAPS { -#ifdef MODE_ANISOTROPIC - vec3 light_accum[6] = vec3[](vec3(0.0), vec3(0.0), vec3(0.0), vec3(0.0), vec3(0.0), vec3(0.0)); -#else vec3 light_accum = vec3(0.0); -#endif float count = 0.0; for (uint i = 0; i < 8; i++) { uint child_index = cell_children.data[cell_index].children[i]; if (child_index == NO_CHILDREN) { continue; } -#ifdef MODE_ANISOTROPIC - light_accum[1] += output.data[child_index * 6 + 0].rgb; - light_accum[2] += output.data[child_index * 6 + 1].rgb; - light_accum[3] += output.data[child_index * 6 + 2].rgb; - light_accum[4] += output.data[child_index * 6 + 3].rgb; - light_accum[5] += output.data[child_index * 6 + 4].rgb; - light_accum[6] += output.data[child_index * 6 + 5].rgb; - -#else light_accum += output.data[child_index].rgb; -#endif - count += 1.0; } float divisor = mix(8.0, count, params.propagation); -#ifdef MODE_ANISOTROPIC - output.data[cell_index * 6 + 0] = vec4(light_accum[0] / divisor, 0.0); - output.data[cell_index * 6 + 1] = vec4(light_accum[1] / divisor, 0.0); - output.data[cell_index * 6 + 2] = vec4(light_accum[2] / divisor, 0.0); - output.data[cell_index * 6 + 3] = vec4(light_accum[3] / divisor, 0.0); - output.data[cell_index * 6 + 4] = vec4(light_accum[4] / divisor, 0.0); - output.data[cell_index * 6 + 5] = vec4(light_accum[5] / divisor, 0.0); - -#else output.data[cell_index] = vec4(light_accum / divisor, 0.0); -#endif } #endif diff --git a/servers/rendering/renderer_rd/shaders/light_data_inc.glsl b/servers/rendering/renderer_rd/shaders/light_data_inc.glsl new file mode 100644 index 0000000000..61c8488a05 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/light_data_inc.glsl @@ -0,0 +1,83 @@ +#define LIGHT_BAKE_DISABLED 0 +#define LIGHT_BAKE_STATIC 1 +#define LIGHT_BAKE_DYNAMIC 2 + +struct LightData { //this structure needs to be as packed as possible + highp vec3 position; + highp float inv_radius; + + mediump vec3 direction; + highp float size; + + mediump vec3 color; + mediump float attenuation; + + mediump float cone_attenuation; + mediump float cone_angle; + mediump float specular_amount; + bool shadow_enabled; + + highp vec4 atlas_rect; // rect in the shadow atlas + highp mat4 shadow_matrix; + highp float shadow_bias; + highp float shadow_normal_bias; + highp float transmittance_bias; + 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; + uint bake_mode; + highp vec4 projector_rect; //projector rect in srgb decal atlas +}; + +#define REFLECTION_AMBIENT_DISABLED 0 +#define REFLECTION_AMBIENT_ENVIRONMENT 1 +#define REFLECTION_AMBIENT_COLOR 2 + +struct ReflectionData { + highp vec3 box_extents; + mediump float index; + highp vec3 box_offset; + uint mask; + mediump vec3 ambient; // ambient color + mediump float intensity; + bool exterior; + bool box_project; + uint ambient_mode; + uint pad; + //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 +}; + +struct DirectionalLightData { + mediump vec3 direction; + mediump float energy; + mediump vec3 color; + mediump float size; + mediump float specular; + uint mask; + highp float softshadow_angle; + highp float soft_shadow_scale; + bool blend_splits; + bool shadow_enabled; + highp float fade_from; + highp float fade_to; + uvec2 pad; + uint bake_mode; + mediump float shadow_volumetric_fog_fade; + highp vec4 shadow_bias; + highp vec4 shadow_normal_bias; + highp vec4 shadow_transmittance_bias; + highp vec4 shadow_z_range; + highp vec4 shadow_range_begin; + highp vec4 shadow_split_offsets; + highp mat4 shadow_matrix1; + highp mat4 shadow_matrix2; + highp mat4 shadow_matrix3; + highp mat4 shadow_matrix4; + highp vec2 uv_scale1; + highp vec2 uv_scale2; + highp vec2 uv_scale3; + highp vec2 uv_scale4; +}; diff --git a/servers/rendering/renderer_rd/shaders/luminance_reduce.glsl b/servers/rendering/renderer_rd/shaders/luminance_reduce.glsl index 8a11c35b78..0ee4cf6e31 100644 --- a/servers/rendering/renderer_rd/shaders/luminance_reduce.glsl +++ b/servers/rendering/renderer_rd/shaders/luminance_reduce.glsl @@ -2,7 +2,7 @@ #version 450 -VERSION_DEFINES +#VERSION_DEFINES #define BLOCK_SIZE 8 @@ -28,7 +28,7 @@ layout(r32f, set = 1, binding = 0) uniform restrict writeonly image2D dest_lumin layout(set = 2, binding = 0) uniform sampler2D prev_luminance; #endif -layout(push_constant, binding = 1, std430) uniform Params { +layout(push_constant, std430) uniform Params { ivec2 source_size; float max_luminance; float min_luminance; diff --git a/servers/rendering/renderer_rd/shaders/luminance_reduce_raster.glsl b/servers/rendering/renderer_rd/shaders/luminance_reduce_raster.glsl new file mode 100644 index 0000000000..29ebd74a90 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/luminance_reduce_raster.glsl @@ -0,0 +1,74 @@ +/* clang-format off */ +#[vertex] + +#version 450 + +#VERSION_DEFINES + +#include "luminance_reduce_raster_inc.glsl" + +layout(location = 0) out vec2 uv_interp; +/* clang-format on */ + +void main() { + vec2 base_arr[4] = vec2[](vec2(0.0, 0.0), vec2(0.0, 1.0), vec2(1.0, 1.0), vec2(1.0, 0.0)); + uv_interp = base_arr[gl_VertexIndex]; + + gl_Position = vec4(uv_interp * 2.0 - 1.0, 0.0, 1.0); +} + +/* clang-format off */ +#[fragment] + +#version 450 + +#VERSION_DEFINES + +#include "luminance_reduce_raster_inc.glsl" + +layout(location = 0) in vec2 uv_interp; +/* clang-format on */ + +layout(set = 0, binding = 0) uniform sampler2D source_exposure; + +#ifdef FINAL_PASS +layout(set = 1, binding = 0) uniform sampler2D prev_luminance; +#endif + +layout(location = 0) out highp float luminance; + +void main() { + ivec2 dest_pos = ivec2(uv_interp * settings.dest_size); + ivec2 src_pos = ivec2(uv_interp * settings.source_size); + + ivec2 next_pos = (dest_pos + ivec2(1)) * settings.source_size / settings.dest_size; + next_pos = max(next_pos, src_pos + ivec2(1)); //so it at least reads one pixel + + highp vec3 source_color = vec3(0.0); + for (int i = src_pos.x; i < next_pos.x; i++) { + for (int j = src_pos.y; j < next_pos.y; j++) { + source_color += texelFetch(source_exposure, ivec2(i, j), 0).rgb; + } + } + + source_color /= float((next_pos.x - src_pos.x) * (next_pos.y - src_pos.y)); + +#ifdef FIRST_PASS + luminance = max(source_color.r, max(source_color.g, source_color.b)); + + // This formula should be more "accurate" but gave an overexposed result when testing. + // Leaving it here so we can revisit it if we want. + // luminance = source_color.r * 0.21 + source_color.g * 0.71 + source_color.b * 0.07; +#else + luminance = source_color.r; +#endif + +#ifdef FINAL_PASS + // Obtain our target luminance + luminance = clamp(luminance, settings.min_luminance, settings.max_luminance); + + // Now smooth to our transition + highp float prev_lum = texelFetch(prev_luminance, ivec2(0, 0), 0).r; //1 pixel previous luminance + luminance = prev_lum + (luminance - prev_lum) * clamp(settings.exposure_adjust, 0.0, 1.0); +#endif +} diff --git a/servers/rendering/renderer_rd/shaders/luminance_reduce_raster_inc.glsl b/servers/rendering/renderer_rd/shaders/luminance_reduce_raster_inc.glsl new file mode 100644 index 0000000000..b8860f6518 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/luminance_reduce_raster_inc.glsl @@ -0,0 +1,11 @@ + +layout(push_constant, std430) uniform PushConstant { + ivec2 source_size; + ivec2 dest_size; + + float exposure_adjust; + float min_luminance; + float max_luminance; + uint pad1; +} +settings; diff --git a/servers/rendering/renderer_rd/shaders/particles.glsl b/servers/rendering/renderer_rd/shaders/particles.glsl index cb6d8dc7f6..acb62b812e 100644 --- a/servers/rendering/renderer_rd/shaders/particles.glsl +++ b/servers/rendering/renderer_rd/shaders/particles.glsl @@ -2,7 +2,7 @@ #version 450 -VERSION_DEFINES +#VERSION_DEFINES layout(local_size_x = 64, local_size_y = 1, local_size_z = 1) in; @@ -19,6 +19,8 @@ layout(local_size_x = 64, local_size_y = 1, local_size_z = 1) in; #define SAMPLER_NEAREST_WITH_MIPMAPS_ANISOTROPIC_REPEAT 10 #define SAMPLER_LINEAR_WITH_MIPMAPS_ANISOTROPIC_REPEAT 11 +#define SDF_MAX_LENGTH 16384.0 + /* SET 0: GLOBAL DATA */ layout(set = 0, binding = 1) uniform sampler material_samplers[12]; @@ -54,6 +56,7 @@ struct Attractor { #define COLLIDER_TYPE_BOX 1 #define COLLIDER_TYPE_SDF 2 #define COLLIDER_TYPE_HEIGHT_FIELD 3 +#define COLLIDER_TYPE_2D_SDF 4 struct Collider { mat4 transform; @@ -76,6 +79,11 @@ struct FrameParams { float time; float delta; + uint frame; + uint pad0; + uint pad1; + uint pad2; + uint random_seed; uint attractor_count; uint collider_count; @@ -92,12 +100,36 @@ layout(set = 1, binding = 0, std430) restrict buffer FrameHistory { } frame_history; +#define PARTICLE_FLAG_ACTIVE uint(1) +#define PARTICLE_FLAG_STARTED uint(2) +#define PARTICLE_FLAG_TRAILED uint(4) +#define PARTICLE_FRAME_MASK uint(0xFFFF) +#define PARTICLE_FRAME_SHIFT uint(16) + struct ParticleData { mat4 xform; vec3 velocity; - bool is_active; + uint flags; vec4 color; vec4 custom; +#ifdef USERDATA1_USED + vec4 userdata1; +#endif +#ifdef USERDATA2_USED + vec4 userdata2; +#endif +#ifdef USERDATA3_USED + vec4 userdata3; +#endif +#ifdef USERDATA4_USED + vec4 userdata4; +#endif +#ifdef USERDATA5_USED + vec4 userdata5; +#endif +#ifdef USERDATA6_USED + vec4 userdata6; +#endif }; layout(set = 1, binding = 1, std430) restrict buffer Particles { @@ -146,15 +178,15 @@ layout(set = 2, binding = 1) uniform texture2D height_field_texture; /* SET 3: MATERIAL */ -#ifdef USE_MATERIAL_UNIFORMS +#ifdef MATERIAL_UNIFORMS_USED layout(set = 3, binding = 0, std140) uniform MaterialUniforms{ - /* clang-format off */ -MATERIAL_UNIFORMS - /* clang-format on */ + +#MATERIAL_UNIFORMS + } material; #endif -layout(push_constant, binding = 0, std430) uniform Params { +layout(push_constant, std430) uniform Params { float lifetime; bool clear; uint total_particles; @@ -162,7 +194,7 @@ layout(push_constant, binding = 0, std430) uniform Params { bool use_fractional_delta; bool sub_emitter_mode; bool can_emit; - uint pad; + bool trail_pass; } params; @@ -196,15 +228,27 @@ bool emit_subparticle(mat4 p_xform, vec3 p_velocity, vec4 p_color, vec4 p_custom return true; } -/* clang-format off */ - -COMPUTE_SHADER_GLOBALS +vec3 safe_normalize(vec3 direction) { + const float EPSILON = 0.001; + if (length(direction) < EPSILON) { + return vec3(0.0); + } + return normalize(direction); +} -/* clang-format on */ +#GLOBALS void main() { uint particle = gl_GlobalInvocationID.x; + if (params.trail_size > 1) { + if (params.trail_pass) { + particle += (particle / (params.trail_size - 1)) + 1; + } else { + particle *= params.trail_size; + } + } + if (particle >= params.total_particles * params.trail_size) { return; //discard } @@ -233,7 +277,7 @@ void main() { PARTICLE.color = vec4(1.0); PARTICLE.custom = vec4(0.0); PARTICLE.velocity = vec3(0.0); - PARTICLE.is_active = false; + PARTICLE.flags = 0; PARTICLE.xform = mat4( vec4(1.0, 0.0, 0.0, 0.0), vec4(0.0, 1.0, 0.0, 0.0), @@ -241,6 +285,29 @@ void main() { vec4(0.0, 0.0, 0.0, 1.0)); } + //clear started flag if set + + if (params.trail_pass) { + //trail started + uint src_idx = index * params.trail_size; + if (bool(particles.data[src_idx].flags & PARTICLE_FLAG_STARTED)) { + //save start conditions for trails + PARTICLE.color = particles.data[src_idx].color; + PARTICLE.custom = particles.data[src_idx].custom; + PARTICLE.velocity = particles.data[src_idx].velocity; + PARTICLE.flags = PARTICLE_FLAG_TRAILED | ((frame_history.data[0].frame & PARTICLE_FRAME_MASK) << PARTICLE_FRAME_SHIFT); //mark it as trailed, save in which frame it will start + PARTICLE.xform = particles.data[src_idx].xform; + } + + if (bool(PARTICLE.flags & PARTICLE_FLAG_TRAILED) && ((PARTICLE.flags >> PARTICLE_FRAME_SHIFT) == (FRAME.frame & PARTICLE_FRAME_MASK))) { //check this is trailed and see if it should start now + // we just assume that this is the first frame of the particle, the rest is deterministic + PARTICLE.flags = PARTICLE_FLAG_ACTIVE | (particles.data[src_idx].flags & (PARTICLE_FRAME_MASK << PARTICLE_FRAME_SHIFT)); + return; //- this appears like it should be correct, but it seems not to be.. wonder why. + } + } else { + PARTICLE.flags &= ~PARTICLE_FLAG_STARTED; + } + bool collided = false; vec3 collision_normal = vec3(0.0); float collision_depth = 0.0; @@ -249,197 +316,17 @@ void main() { #if !defined(DISABLE_VELOCITY) - if (PARTICLE.is_active) { + if (bool(PARTICLE.flags & PARTICLE_FLAG_ACTIVE)) { PARTICLE.xform[3].xyz += PARTICLE.velocity * local_delta; } #endif - /* Process physics if active */ - - if (PARTICLE.is_active) { - for (uint i = 0; i < FRAME.attractor_count; i++) { - vec3 dir; - float amount; - vec3 rel_vec = PARTICLE.xform[3].xyz - FRAME.attractors[i].transform[3].xyz; - vec3 local_pos = rel_vec * mat3(FRAME.attractors[i].transform); - - switch (FRAME.attractors[i].type) { - case ATTRACTOR_TYPE_SPHERE: { - dir = normalize(rel_vec); - float d = length(local_pos) / FRAME.attractors[i].extents.x; - if (d > 1.0) { - continue; - } - amount = max(0.0, 1.0 - d); - } break; - case ATTRACTOR_TYPE_BOX: { - dir = normalize(rel_vec); - - vec3 abs_pos = abs(local_pos / FRAME.attractors[i].extents); - float d = max(abs_pos.x, max(abs_pos.y, abs_pos.z)); - if (d > 1.0) { - continue; - } - amount = max(0.0, 1.0 - d); - - } break; - case ATTRACTOR_TYPE_VECTOR_FIELD: { - vec3 uvw_pos = (local_pos / FRAME.attractors[i].extents) * 2.0 - 1.0; - 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; - dir = mat3(FRAME.attractors[i].transform) * normalize(s); //revert direction - amount = length(s); - - } break; - } - amount = pow(amount, FRAME.attractors[i].attenuation); - dir = normalize(mix(dir, FRAME.attractors[i].transform[2].xyz, FRAME.attractors[i].directionality)); - attractor_force -= amount * dir * FRAME.attractors[i].strength; - } - - float particle_size = FRAME.particle_size; - -#ifdef USE_COLLISON_SCALE - - particle_size *= dot(vec3(length(PARTICLE.xform[0].xyz), length(PARTICLE.xform[1].xyz), length(PARTICLE.xform[2].xyz)), vec3(0.33333333333)); - -#endif - - for (uint i = 0; i < FRAME.collider_count; i++) { - vec3 normal; - float depth; - bool col = false; - - vec3 rel_vec = PARTICLE.xform[3].xyz - FRAME.colliders[i].transform[3].xyz; - vec3 local_pos = rel_vec * mat3(FRAME.colliders[i].transform); - - switch (FRAME.colliders[i].type) { - case COLLIDER_TYPE_SPHERE: { - float d = length(rel_vec) - (particle_size + FRAME.colliders[i].extents.x); - - if (d < 0.0) { - col = true; - depth = -d; - normal = normalize(rel_vec); - } - - } break; - case COLLIDER_TYPE_BOX: { - vec3 abs_pos = abs(local_pos); - vec3 sgn_pos = sign(local_pos); - - if (any(greaterThan(abs_pos, FRAME.colliders[i].extents))) { - //point outside box - - vec3 closest = min(abs_pos, FRAME.colliders[i].extents); - vec3 rel = abs_pos - closest; - depth = length(rel) - particle_size; - if (depth < 0.0) { - col = true; - normal = mat3(FRAME.colliders[i].transform) * (normalize(rel) * sgn_pos); - depth = -depth; - } - } else { - //point inside box - vec3 axis_len = FRAME.colliders[i].extents - abs_pos; - // there has to be a faster way to do this? - if (all(lessThan(axis_len.xx, axis_len.yz))) { - normal = vec3(1, 0, 0); - } else if (all(lessThan(axis_len.yy, axis_len.xz))) { - normal = vec3(0, 1, 0); - } else { - normal = vec3(0, 0, 1); - } - - col = true; - depth = dot(normal * axis_len, vec3(1)) + particle_size; - normal = mat3(FRAME.colliders[i].transform) * (normal * sgn_pos); - } - - } break; - case COLLIDER_TYPE_SDF: { - vec3 apos = abs(local_pos); - float extra_dist = 0.0; - if (any(greaterThan(apos, FRAME.colliders[i].extents))) { //outside - vec3 mpos = min(apos, FRAME.colliders[i].extents); - extra_dist = distance(mpos, apos); - } - - if (extra_dist > particle_size) { - continue; - } - - vec3 uvw_pos = (local_pos / FRAME.colliders[i].extents) * 0.5 + 0.5; - float s = texture(sampler3D(sdf_vec_textures[FRAME.colliders[i].texture_index], material_samplers[SAMPLER_LINEAR_CLAMP]), uvw_pos).r; - s *= FRAME.colliders[i].scale; - s += extra_dist; - if (s < particle_size) { - col = true; - depth = particle_size - s; - const float EPSILON = 0.001; - normal = mat3(FRAME.colliders[i].transform) * - normalize( - vec3( - texture(sampler3D(sdf_vec_textures[FRAME.colliders[i].texture_index], material_samplers[SAMPLER_LINEAR_CLAMP]), uvw_pos + vec3(EPSILON, 0.0, 0.0)).r - texture(sampler3D(sdf_vec_textures[FRAME.colliders[i].texture_index], material_samplers[SAMPLER_LINEAR_CLAMP]), uvw_pos - vec3(EPSILON, 0.0, 0.0)).r, - texture(sampler3D(sdf_vec_textures[FRAME.colliders[i].texture_index], material_samplers[SAMPLER_LINEAR_CLAMP]), uvw_pos + vec3(0.0, EPSILON, 0.0)).r - texture(sampler3D(sdf_vec_textures[FRAME.colliders[i].texture_index], material_samplers[SAMPLER_LINEAR_CLAMP]), uvw_pos - vec3(0.0, EPSILON, 0.0)).r, - texture(sampler3D(sdf_vec_textures[FRAME.colliders[i].texture_index], material_samplers[SAMPLER_LINEAR_CLAMP]), uvw_pos + vec3(0.0, 0.0, EPSILON)).r - texture(sampler3D(sdf_vec_textures[FRAME.colliders[i].texture_index], material_samplers[SAMPLER_LINEAR_CLAMP]), uvw_pos - vec3(0.0, 0.0, EPSILON)).r)); - } - - } break; - case COLLIDER_TYPE_HEIGHT_FIELD: { - vec3 local_pos_bottom = local_pos; - local_pos_bottom.y -= particle_size; - - if (any(greaterThan(abs(local_pos_bottom), FRAME.colliders[i].extents))) { - continue; - } - - const float DELTA = 1.0 / 8192.0; - - vec3 uvw_pos = vec3(local_pos_bottom / FRAME.colliders[i].extents) * 0.5 + 0.5; - - float y = 1.0 - texture(sampler2D(height_field_texture, material_samplers[SAMPLER_LINEAR_CLAMP]), uvw_pos.xz).r; - - if (y > uvw_pos.y) { - //inside heightfield - - vec3 pos1 = (vec3(uvw_pos.x, y, uvw_pos.z) * 2.0 - 1.0) * FRAME.colliders[i].extents; - vec3 pos2 = (vec3(uvw_pos.x + DELTA, 1.0 - texture(sampler2D(height_field_texture, material_samplers[SAMPLER_LINEAR_CLAMP]), uvw_pos.xz + vec2(DELTA, 0)).r, uvw_pos.z) * 2.0 - 1.0) * FRAME.colliders[i].extents; - vec3 pos3 = (vec3(uvw_pos.x, 1.0 - texture(sampler2D(height_field_texture, material_samplers[SAMPLER_LINEAR_CLAMP]), uvw_pos.xz + vec2(0, DELTA)).r, uvw_pos.z + DELTA) * 2.0 - 1.0) * FRAME.colliders[i].extents; - - normal = normalize(cross(pos1 - pos2, pos1 - pos3)); - float local_y = (vec3(local_pos / FRAME.colliders[i].extents) * 0.5 + 0.5).y; - - col = true; - depth = dot(normal, pos1) - dot(normal, local_pos_bottom); - } - - } break; - } - - if (col) { - if (!collided) { - collided = true; - collision_normal = normal; - collision_depth = depth; - } else { - vec3 c = collision_normal * collision_depth; - c += normal * max(0.0, depth - dot(normal, c)); - collision_normal = normalize(c); - collision_depth = length(c); - } - } - } - } - - if (params.sub_emitter_mode) { - if (!PARTICLE.is_active) { + if (!params.trail_pass && params.sub_emitter_mode) { + if (!bool(PARTICLE.flags & PARTICLE_FLAG_ACTIVE)) { int src_index = atomicAdd(src_particles.particle_count, -1) - 1; if (src_index >= 0) { - PARTICLE.is_active = true; + PARTICLE.flags = (PARTICLE_FLAG_ACTIVE | PARTICLE_FLAG_STARTED | (FRAME.cycle << PARTICLE_FRAME_SHIFT)); restart = true; if (bool(src_particles.data[src_index].flags & EMISSION_FLAG_HAS_POSITION)) { @@ -521,16 +408,12 @@ void main() { } } - uint current_cycle = FRAME.cycle; - - if (FRAME.system_phase < restart_phase) { - current_cycle -= uint(1); + if (params.trail_pass) { + restart = false; } - uint particle_number = current_cycle * uint(params.total_particles) + particle; - if (restart) { - PARTICLE.is_active = FRAME.emitting; + PARTICLE.flags = FRAME.emitting ? (PARTICLE_FLAG_ACTIVE | PARTICLE_FLAG_STARTED | (FRAME.cycle << PARTICLE_FRAME_SHIFT)) : 0; restart_position = true; restart_rotation_scale = true; restart_velocity = true; @@ -539,11 +422,237 @@ void main() { } } - if (PARTICLE.is_active) { - /* clang-format off */ + bool particle_active = bool(PARTICLE.flags & PARTICLE_FLAG_ACTIVE); + + uint particle_number = (PARTICLE.flags >> PARTICLE_FRAME_SHIFT) * uint(params.total_particles) + index; + + if (restart && particle_active) { +#CODE : START + } + + if (particle_active) { + for (uint i = 0; i < FRAME.attractor_count; i++) { + vec3 dir; + float amount; + vec3 rel_vec = PARTICLE.xform[3].xyz - FRAME.attractors[i].transform[3].xyz; + vec3 local_pos = rel_vec * mat3(FRAME.attractors[i].transform); + + switch (FRAME.attractors[i].type) { + case ATTRACTOR_TYPE_SPHERE: { + dir = safe_normalize(rel_vec); + float d = length(local_pos) / FRAME.attractors[i].extents.x; + if (d > 1.0) { + continue; + } + amount = max(0.0, 1.0 - d); + } break; + case ATTRACTOR_TYPE_BOX: { + dir = safe_normalize(rel_vec); + + vec3 abs_pos = abs(local_pos / FRAME.attractors[i].extents); + float d = max(abs_pos.x, max(abs_pos.y, abs_pos.z)); + if (d > 1.0) { + continue; + } + amount = max(0.0, 1.0 - d); + + } break; + case ATTRACTOR_TYPE_VECTOR_FIELD: { + vec3 uvw_pos = (local_pos / FRAME.attractors[i].extents) * 2.0 - 1.0; + 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; + dir = mat3(FRAME.attractors[i].transform) * safe_normalize(s); //revert direction + amount = length(s); + + } break; + } + amount = pow(amount, FRAME.attractors[i].attenuation); + dir = safe_normalize(mix(dir, FRAME.attractors[i].transform[2].xyz, FRAME.attractors[i].directionality)); + attractor_force -= amount * dir * FRAME.attractors[i].strength; + } + + float particle_size = FRAME.particle_size; + +#ifdef USE_COLLISON_SCALE + + particle_size *= dot(vec3(length(PARTICLE.xform[0].xyz), length(PARTICLE.xform[1].xyz), length(PARTICLE.xform[2].xyz)), vec3(0.33333333333)); + +#endif + + if (FRAME.collider_count == 1 && FRAME.colliders[0].type == COLLIDER_TYPE_2D_SDF) { + //2D collision + + vec2 pos = PARTICLE.xform[3].xy; + vec4 to_sdf_x = FRAME.colliders[0].transform[0]; + vec4 to_sdf_y = FRAME.colliders[0].transform[1]; + vec2 sdf_pos = vec2(dot(vec4(pos, 0, 1), to_sdf_x), dot(vec4(pos, 0, 1), to_sdf_y)); + + vec4 sdf_to_screen = vec4(FRAME.colliders[0].extents, FRAME.colliders[0].scale); + + vec2 uv_pos = sdf_pos * sdf_to_screen.xy + sdf_to_screen.zw; + + if (all(greaterThan(uv_pos, vec2(0.0))) && all(lessThan(uv_pos, vec2(1.0)))) { + vec2 pos2 = pos + vec2(0, particle_size); + vec2 sdf_pos2 = vec2(dot(vec4(pos2, 0, 1), to_sdf_x), dot(vec4(pos2, 0, 1), to_sdf_y)); + float sdf_particle_size = distance(sdf_pos, sdf_pos2); + + float d = texture(sampler2D(height_field_texture, material_samplers[SAMPLER_LINEAR_CLAMP]), uv_pos).r * SDF_MAX_LENGTH; + + d -= sdf_particle_size; + + if (d < 0.0) { + const float EPSILON = 0.001; + vec2 n = normalize(vec2( + texture(sampler2D(height_field_texture, material_samplers[SAMPLER_LINEAR_CLAMP]), uv_pos + vec2(EPSILON, 0.0)).r - texture(sampler2D(height_field_texture, material_samplers[SAMPLER_LINEAR_CLAMP]), uv_pos - vec2(EPSILON, 0.0)).r, + texture(sampler2D(height_field_texture, material_samplers[SAMPLER_LINEAR_CLAMP]), uv_pos + vec2(0.0, EPSILON)).r - texture(sampler2D(height_field_texture, material_samplers[SAMPLER_LINEAR_CLAMP]), uv_pos - vec2(0.0, EPSILON)).r)); + + collided = true; + sdf_pos2 = sdf_pos + n * d; + pos2 = vec2(dot(vec4(sdf_pos2, 0, 1), FRAME.colliders[0].transform[2]), dot(vec4(sdf_pos2, 0, 1), FRAME.colliders[0].transform[3])); + + n = pos - pos2; + + collision_normal = normalize(vec3(n, 0.0)); + collision_depth = length(n); + } + } + + } else { + for (uint i = 0; i < FRAME.collider_count; i++) { + vec3 normal; + float depth; + bool col = false; + + vec3 rel_vec = PARTICLE.xform[3].xyz - FRAME.colliders[i].transform[3].xyz; + vec3 local_pos = rel_vec * mat3(FRAME.colliders[i].transform); + + switch (FRAME.colliders[i].type) { + case COLLIDER_TYPE_SPHERE: { + float d = length(rel_vec) - (particle_size + FRAME.colliders[i].extents.x); + + if (d < 0.0) { + col = true; + depth = -d; + normal = normalize(rel_vec); + } + + } break; + case COLLIDER_TYPE_BOX: { + vec3 abs_pos = abs(local_pos); + vec3 sgn_pos = sign(local_pos); + + if (any(greaterThan(abs_pos, FRAME.colliders[i].extents))) { + //point outside box + + vec3 closest = min(abs_pos, FRAME.colliders[i].extents); + vec3 rel = abs_pos - closest; + depth = length(rel) - particle_size; + if (depth < 0.0) { + col = true; + normal = mat3(FRAME.colliders[i].transform) * (normalize(rel) * sgn_pos); + depth = -depth; + } + } else { + //point inside box + vec3 axis_len = FRAME.colliders[i].extents - abs_pos; + // there has to be a faster way to do this? + if (all(lessThan(axis_len.xx, axis_len.yz))) { + normal = vec3(1, 0, 0); + } else if (all(lessThan(axis_len.yy, axis_len.xz))) { + normal = vec3(0, 1, 0); + } else { + normal = vec3(0, 0, 1); + } + + col = true; + depth = dot(normal * axis_len, vec3(1)) + particle_size; + normal = mat3(FRAME.colliders[i].transform) * (normal * sgn_pos); + } -COMPUTE_SHADER_CODE + } break; + case COLLIDER_TYPE_SDF: { + vec3 apos = abs(local_pos); + float extra_dist = 0.0; + if (any(greaterThan(apos, FRAME.colliders[i].extents))) { //outside + vec3 mpos = min(apos, FRAME.colliders[i].extents); + extra_dist = distance(mpos, apos); + } + + if (extra_dist > particle_size) { + continue; + } + + vec3 uvw_pos = (local_pos / FRAME.colliders[i].extents) * 0.5 + 0.5; + float s = texture(sampler3D(sdf_vec_textures[FRAME.colliders[i].texture_index], material_samplers[SAMPLER_LINEAR_CLAMP]), uvw_pos).r; + s *= FRAME.colliders[i].scale; + s += extra_dist; + if (s < particle_size) { + col = true; + depth = particle_size - s; + const float EPSILON = 0.001; + normal = mat3(FRAME.colliders[i].transform) * + normalize( + vec3( + texture(sampler3D(sdf_vec_textures[FRAME.colliders[i].texture_index], material_samplers[SAMPLER_LINEAR_CLAMP]), uvw_pos + vec3(EPSILON, 0.0, 0.0)).r - texture(sampler3D(sdf_vec_textures[FRAME.colliders[i].texture_index], material_samplers[SAMPLER_LINEAR_CLAMP]), uvw_pos - vec3(EPSILON, 0.0, 0.0)).r, + texture(sampler3D(sdf_vec_textures[FRAME.colliders[i].texture_index], material_samplers[SAMPLER_LINEAR_CLAMP]), uvw_pos + vec3(0.0, EPSILON, 0.0)).r - texture(sampler3D(sdf_vec_textures[FRAME.colliders[i].texture_index], material_samplers[SAMPLER_LINEAR_CLAMP]), uvw_pos - vec3(0.0, EPSILON, 0.0)).r, + texture(sampler3D(sdf_vec_textures[FRAME.colliders[i].texture_index], material_samplers[SAMPLER_LINEAR_CLAMP]), uvw_pos + vec3(0.0, 0.0, EPSILON)).r - texture(sampler3D(sdf_vec_textures[FRAME.colliders[i].texture_index], material_samplers[SAMPLER_LINEAR_CLAMP]), uvw_pos - vec3(0.0, 0.0, EPSILON)).r)); + } + + } break; + case COLLIDER_TYPE_HEIGHT_FIELD: { + vec3 local_pos_bottom = local_pos; + local_pos_bottom.y -= particle_size; + + if (any(greaterThan(abs(local_pos_bottom), FRAME.colliders[i].extents))) { + continue; + } + const float DELTA = 1.0 / 8192.0; + + vec3 uvw_pos = vec3(local_pos_bottom / FRAME.colliders[i].extents) * 0.5 + 0.5; + + float y = 1.0 - texture(sampler2D(height_field_texture, material_samplers[SAMPLER_LINEAR_CLAMP]), uvw_pos.xz).r; + + if (y > uvw_pos.y) { + //inside heightfield + + vec3 pos1 = (vec3(uvw_pos.x, y, uvw_pos.z) * 2.0 - 1.0) * FRAME.colliders[i].extents; + vec3 pos2 = (vec3(uvw_pos.x + DELTA, 1.0 - texture(sampler2D(height_field_texture, material_samplers[SAMPLER_LINEAR_CLAMP]), uvw_pos.xz + vec2(DELTA, 0)).r, uvw_pos.z) * 2.0 - 1.0) * FRAME.colliders[i].extents; + vec3 pos3 = (vec3(uvw_pos.x, 1.0 - texture(sampler2D(height_field_texture, material_samplers[SAMPLER_LINEAR_CLAMP]), uvw_pos.xz + vec2(0, DELTA)).r, uvw_pos.z + DELTA) * 2.0 - 1.0) * FRAME.colliders[i].extents; + + normal = normalize(cross(pos1 - pos2, pos1 - pos3)); + float local_y = (vec3(local_pos / FRAME.colliders[i].extents) * 0.5 + 0.5).y; + + col = true; + depth = dot(normal, pos1) - dot(normal, local_pos_bottom); + } + + } break; + } + + if (col) { + if (!collided) { + collided = true; + collision_normal = normal; + collision_depth = depth; + } else { + vec3 c = collision_normal * collision_depth; + c += normal * max(0.0, depth - dot(normal, c)); + collision_normal = normalize(c); + collision_depth = length(c); + } + } + } + } + } + + if (particle_active) { +#CODE : PROCESS + } - /* clang-format on */ + PARTICLE.flags &= ~PARTICLE_FLAG_ACTIVE; + if (particle_active) { + PARTICLE.flags |= PARTICLE_FLAG_ACTIVE; } } diff --git a/servers/rendering/renderer_rd/shaders/particles_copy.glsl b/servers/rendering/renderer_rd/shaders/particles_copy.glsl index 6c782b6045..afbd5a9caa 100644 --- a/servers/rendering/renderer_rd/shaders/particles_copy.glsl +++ b/servers/rendering/renderer_rd/shaders/particles_copy.glsl @@ -2,16 +2,23 @@ #version 450 -VERSION_DEFINES +#VERSION_DEFINES layout(local_size_x = 64, local_size_y = 1, local_size_z = 1) in; +#define PARTICLE_FLAG_ACTIVE uint(1) +#define PARTICLE_FLAG_STARTED uint(2) +#define PARTICLE_FLAG_TRAILED uint(4) + struct ParticleData { mat4 xform; vec3 velocity; - bool is_active; + uint flags; vec4 color; vec4 custom; +#ifdef USERDATA_COUNT + vec4 userdata[USERDATA_COUNT]; +#endif }; layout(set = 0, binding = 1, std430) restrict readonly buffer Particles { @@ -33,12 +40,37 @@ sort_buffer; #endif // USE_SORT_BUFFER -layout(push_constant, binding = 0, std430) uniform Params { +layout(set = 2, binding = 0, std430) restrict readonly buffer TrailBindPoses { + mat4 data[]; +} +trail_bind_poses; + +layout(push_constant, std430) uniform Params { vec3 sort_direction; uint total_particles; + + uint trail_size; + uint trail_total; + float frame_delta; + float frame_remainder; + + vec3 align_up; + uint align_mode; + + bool order_by_lifetime; + uint lifetime_split; + bool lifetime_reverse; + bool copy_mode_2d; + + mat4 inv_emission_transform; } params; +#define TRANSFORM_ALIGN_DISABLED 0 +#define TRANSFORM_ALIGN_Z_BILLBOARD 1 +#define TRANSFORM_ALIGN_Y_TO_VELOCITY 2 +#define TRANSFORM_ALIGN_Z_BILLBOARD_Y_TO_VELOCITY 3 + void main() { #ifdef MODE_FILL_SORT_BUFFER @@ -47,36 +79,155 @@ void main() { return; //discard } - sort_buffer.data[particle].x = dot(params.sort_direction, particles.data[particle].xform[3].xyz); + uint src_particle = particle; + if (params.trail_size > 1) { + src_particle = src_particle * params.trail_size + params.trail_size / 2; //use trail center for sorting + } + sort_buffer.data[particle].x = dot(params.sort_direction, particles.data[src_particle].xform[3].xyz); sort_buffer.data[particle].y = float(particle); #endif #ifdef MODE_FILL_INSTANCES uint particle = gl_GlobalInvocationID.x; - uint write_offset = gl_GlobalInvocationID.x * (3 + 1 + 1); //xform + color + custom if (particle >= params.total_particles) { return; //discard } #ifdef USE_SORT_BUFFER - particle = uint(sort_buffer.data[particle].y); //use index from sort buffer -#endif + + if (params.trail_size > 1) { + particle = uint(sort_buffer.data[particle / params.trail_size].y) + (particle % params.trail_size); + } else { + particle = uint(sort_buffer.data[particle].y); //use index from sort buffer + } +#else + if (params.order_by_lifetime) { + if (params.trail_size > 1) { + uint limit = (params.total_particles / params.trail_size) - params.lifetime_split; + + uint base_index = particle / params.trail_size; + uint base_offset = particle % params.trail_size; + + if (params.lifetime_reverse) { + base_index = (params.total_particles / params.trail_size) - base_index - 1; + } + + if (base_index < limit) { + base_index = params.lifetime_split + base_index; + } else { + base_index -= limit; + } + + particle = base_index * params.trail_size + base_offset; + + } else { + uint limit = params.total_particles - params.lifetime_split; + + if (params.lifetime_reverse) { + particle = params.total_particles - particle - 1; + } + + if (particle < limit) { + particle = params.lifetime_split + particle; + } else { + particle -= limit; + } + } + } +#endif // USE_SORT_BUFFER mat4 txform; - if (particles.data[particle].is_active) { - txform = transpose(particles.data[particle].xform); + if (bool(particles.data[particle].flags & PARTICLE_FLAG_ACTIVE) || bool(particles.data[particle].flags & PARTICLE_FLAG_TRAILED)) { + txform = particles.data[particle].xform; + if (params.trail_size > 1) { + // Since the steps don't fit precisely in the history frames, must do a tiny bit of + // interpolation to get them close to their intended location. + uint part_ofs = particle % params.trail_size; + float natural_ofs = fract((float(part_ofs) / float(params.trail_size)) * float(params.trail_total)) * params.frame_delta; + + txform[3].xyz -= particles.data[particle].velocity * natural_ofs; + } + + switch (params.align_mode) { + case TRANSFORM_ALIGN_DISABLED: { + } break; //nothing + case TRANSFORM_ALIGN_Z_BILLBOARD: { + mat3 local = mat3(normalize(cross(params.align_up, params.sort_direction)), params.align_up, params.sort_direction); + local = local * mat3(txform); + txform[0].xyz = local[0]; + txform[1].xyz = local[1]; + txform[2].xyz = local[2]; + + } break; + case TRANSFORM_ALIGN_Y_TO_VELOCITY: { + vec3 v = particles.data[particle].velocity; + float s = (length(txform[0]) + length(txform[1]) + length(txform[2])) / 3.0; + if (length(v) > 0.0) { + txform[1].xyz = normalize(v); + } else { + txform[1].xyz = normalize(txform[1].xyz); + } + + txform[0].xyz = normalize(cross(txform[1].xyz, txform[2].xyz)); + txform[2].xyz = vec3(0.0, 0.0, 1.0) * s; + txform[0].xyz *= s; + txform[1].xyz *= s; + } break; + case TRANSFORM_ALIGN_Z_BILLBOARD_Y_TO_VELOCITY: { + vec3 v = particles.data[particle].velocity; + vec3 sv = v - params.sort_direction * dot(params.sort_direction, v); //screen velocity + float s = (length(txform[0]) + length(txform[1]) + length(txform[2])) / 3.0; + + if (length(sv) == 0) { + sv = params.align_up; + } + + sv = normalize(sv); + + txform[0].xyz = normalize(cross(sv, params.sort_direction)) * s; + txform[1].xyz = sv * s; + txform[2].xyz = params.sort_direction * s; + + } break; + } + + txform[3].xyz += particles.data[particle].velocity * params.frame_remainder; + + if (params.trail_size > 1) { + uint part_ofs = particle % params.trail_size; + txform = txform * trail_bind_poses.data[part_ofs]; + } + + if (params.copy_mode_2d) { + // In global mode, bring 2D particles to local coordinates + // as they will be drawn with the node position as origin. + txform = params.inv_emission_transform * txform; + } + + txform = transpose(txform); } else { txform = mat4(vec4(0.0), vec4(0.0), vec4(0.0), vec4(0.0)); //zero scale, becomes invisible } - instances.data[write_offset + 0] = txform[0]; - instances.data[write_offset + 1] = txform[1]; - instances.data[write_offset + 2] = txform[2]; - instances.data[write_offset + 3] = particles.data[particle].color; - instances.data[write_offset + 4] = particles.data[particle].custom; + if (params.copy_mode_2d) { + uint write_offset = gl_GlobalInvocationID.x * (2 + 1 + 1); //xform + color + custom + + instances.data[write_offset + 0] = txform[0]; + instances.data[write_offset + 1] = txform[1]; + instances.data[write_offset + 2] = particles.data[particle].color; + instances.data[write_offset + 3] = particles.data[particle].custom; + } else { + uint write_offset = gl_GlobalInvocationID.x * (3 + 1 + 1); //xform + color + custom + + instances.data[write_offset + 0] = txform[0]; + instances.data[write_offset + 1] = txform[1]; + instances.data[write_offset + 2] = txform[2]; + instances.data[write_offset + 3] = particles.data[particle].color; + instances.data[write_offset + 4] = particles.data[particle].custom; + } #endif } diff --git a/servers/rendering/renderer_rd/shaders/resolve.glsl b/servers/rendering/renderer_rd/shaders/resolve.glsl index e83c4ca93b..0e086331c0 100644 --- a/servers/rendering/renderer_rd/shaders/resolve.glsl +++ b/servers/rendering/renderer_rd/shaders/resolve.glsl @@ -2,10 +2,15 @@ #version 450 -VERSION_DEFINES +#VERSION_DEFINES layout(local_size_x = 8, local_size_y = 8, local_size_z = 1) in; +#ifdef MODE_RESOLVE_DEPTH +layout(set = 0, binding = 0) uniform sampler2DMS source_depth; +layout(r32f, set = 1, binding = 0) uniform restrict writeonly image2D dest_depth; +#endif + #ifdef MODE_RESOLVE_GI layout(set = 0, binding = 0) uniform sampler2DMS source_depth; layout(set = 0, binding = 1) uniform sampler2DMS source_normal_roughness; @@ -13,14 +18,14 @@ layout(set = 0, binding = 1) uniform sampler2DMS source_normal_roughness; layout(r32f, set = 1, binding = 0) uniform restrict writeonly image2D dest_depth; layout(rgba8, set = 1, binding = 1) uniform restrict writeonly image2D dest_normal_roughness; -#ifdef GIPROBE_RESOLVE -layout(set = 2, binding = 0) uniform usampler2DMS source_giprobe; -layout(rg8ui, set = 3, binding = 0) uniform restrict writeonly uimage2D dest_giprobe; +#ifdef VOXEL_GI_RESOLVE +layout(set = 2, binding = 0) uniform usampler2DMS source_voxel_gi; +layout(rg8ui, set = 3, binding = 0) uniform restrict writeonly uimage2D dest_voxel_gi; #endif #endif -layout(push_constant, binding = 16, std430) uniform Params { +layout(push_constant, std430) uniform Params { ivec2 screen_size; int sample_count; uint pad; @@ -34,12 +39,23 @@ void main() { return; } +#ifdef MODE_RESOLVE_DEPTH + + float depth_avg = 0.0; + for (int i = 0; i < params.sample_count; i++) { + depth_avg += texelFetch(source_depth, pos, i).r; + } + depth_avg /= float(params.sample_count); + imageStore(dest_depth, pos, vec4(depth_avg)); + +#endif + #ifdef MODE_RESOLVE_GI float best_depth = 1e20; vec4 best_normal_roughness = vec4(0.0); -#ifdef GIPROBE_RESOLVE - uvec2 best_giprobe; +#ifdef VOXEL_GI_RESOLVE + uvec2 best_voxel_gi; #endif #if 0 @@ -50,8 +66,8 @@ void main() { best_depth = depth; best_normal_roughness = texelFetch(source_normal_roughness,pos,i); -#ifdef GIPROBE_RESOLVE - best_giprobe = texelFetch(source_giprobe,pos,i).rg; +#ifdef VOXEL_GI_RESOLVE + best_voxel_gi = texelFetch(source_voxel_gi,pos,i).rg; #endif } } @@ -204,16 +220,16 @@ void main() { #endif best_depth = texelFetch(source_depth, pos, best_index).r; best_normal_roughness = texelFetch(source_normal_roughness, pos, best_index); -#ifdef GIPROBE_RESOLVE - best_giprobe = texelFetch(source_giprobe, pos, best_index).rg; +#ifdef VOXEL_GI_RESOLVE + best_voxel_gi = texelFetch(source_voxel_gi, pos, best_index).rg; #endif #endif imageStore(dest_depth, pos, vec4(best_depth)); imageStore(dest_normal_roughness, pos, vec4(best_normal_roughness)); -#ifdef GIPROBE_RESOLVE - imageStore(dest_giprobe, pos, uvec4(best_giprobe, 0, 0)); +#ifdef VOXEL_GI_RESOLVE + imageStore(dest_voxel_gi, pos, uvec4(best_voxel_gi, 0, 0)); #endif #endif diff --git a/servers/rendering/renderer_rd/shaders/roughness_limiter.glsl b/servers/rendering/renderer_rd/shaders/roughness_limiter.glsl index 464895928a..59027df8e9 100644 --- a/servers/rendering/renderer_rd/shaders/roughness_limiter.glsl +++ b/servers/rendering/renderer_rd/shaders/roughness_limiter.glsl @@ -2,14 +2,14 @@ #version 450 -VERSION_DEFINES +#VERSION_DEFINES layout(local_size_x = 8, local_size_y = 8, local_size_z = 1) in; layout(set = 0, binding = 0) uniform sampler2D source_normal; layout(r8, set = 1, binding = 0) uniform restrict writeonly image2D dest_roughness; -layout(push_constant, binding = 1, std430) uniform Params { +layout(push_constant, std430) uniform Params { ivec2 screen_size; float curve; uint pad; diff --git a/servers/rendering/renderer_rd/shaders/scene_forward_aa_inc.glsl b/servers/rendering/renderer_rd/shaders/scene_forward_aa_inc.glsl new file mode 100644 index 0000000000..97c913d489 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/scene_forward_aa_inc.glsl @@ -0,0 +1,57 @@ +#ifdef ALPHA_HASH_USED + +float hash_2d(vec2 p) { + return fract(1.0e4 * sin(17.0 * p.x + 0.1 * p.y) * + (0.1 + abs(sin(13.0 * p.y + p.x)))); +} + +float hash_3d(vec3 p) { + return hash_2d(vec2(hash_2d(p.xy), p.z)); +} + +float compute_alpha_hash_threshold(vec3 pos, float hash_scale) { + vec3 dx = dFdx(pos); + vec3 dy = dFdx(pos); + float delta_max_sqr = max(length(dx), length(dy)); + float pix_scale = 1.0 / (hash_scale * delta_max_sqr); + + vec2 pix_scales = + vec2(exp2(floor(log2(pix_scale))), exp2(ceil(log2(pix_scale)))); + + vec2 a_thresh = vec2(hash_3d(floor(pix_scales.x * pos.xyz)), + hash_3d(floor(pix_scales.y * pos.xyz))); + + float lerp_factor = fract(log2(pix_scale)); + + float a_interp = (1.0 - lerp_factor) * a_thresh.x + lerp_factor * a_thresh.y; + + float min_lerp = min(lerp_factor, 1.0 - lerp_factor); + + vec3 cases = vec3(a_interp * a_interp / (2.0 * min_lerp * (1.0 - min_lerp)), + (a_interp - 0.5 * min_lerp) / (1.0 - min_lerp), + 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; + + return clamp(alpha_hash_threshold, 0.0, 1.0); +} + +#endif // ALPHA_HASH_USED + +#ifdef ALPHA_ANTIALIASING_EDGE_USED + +float calc_mip_level(vec2 texture_coord) { + vec2 dx = dFdx(texture_coord); + vec2 dy = dFdy(texture_coord); + float delta_max_sqr = max(dot(dx, dx), dot(dy, dy)); + return max(0.0, 0.5 * log2(delta_max_sqr)); +} + +float compute_alpha_antialiasing_edge(float input_alpha, vec2 texture_coord, float alpha_edge) { + input_alpha *= 1.0 + max(0, calc_mip_level(texture_coord)) * 0.25; // 0.25 mip scale, magic number + input_alpha = (input_alpha - alpha_edge) / max(fwidth(input_alpha), 0.0001) + 0.5; + return clamp(input_alpha, 0.0, 1.0); +} + +#endif // ALPHA_ANTIALIASING_USED diff --git a/servers/rendering/renderer_rd/shaders/scene_forward_clustered.glsl b/servers/rendering/renderer_rd/shaders/scene_forward_clustered.glsl index 7b86dac143..268e1dd7d0 100644 --- a/servers/rendering/renderer_rd/shaders/scene_forward_clustered.glsl +++ b/servers/rendering/renderer_rd/shaders/scene_forward_clustered.glsl @@ -2,10 +2,12 @@ #version 450 -VERSION_DEFINES +#VERSION_DEFINES #include "scene_forward_clustered_inc.glsl" +#define SHADER_IS_SRGB false + /* INPUT ATTRIBS */ layout(location = 0) in vec3 vertex_attrib; @@ -48,11 +50,11 @@ layout(location = 8) in vec4 custom2_attrib; layout(location = 9) in vec4 custom3_attrib; #endif -#if defined(BONES_USED) +#if defined(BONES_USED) || defined(USE_PARTICLE_TRAILS) layout(location = 10) in uvec4 bone_attrib; #endif -#if defined(WEIGHTS_USED) +#if defined(WEIGHTS_USED) || defined(USE_PARTICLE_TRAILS) layout(location = 11) in vec4 weight_attrib; #endif @@ -81,58 +83,126 @@ layout(location = 5) out vec3 tangent_interp; layout(location = 6) out vec3 binormal_interp; #endif -#ifdef USE_MATERIAL_UNIFORMS +#ifdef MOTION_VECTORS +layout(location = 7) out vec4 screen_position; +layout(location = 8) out vec4 prev_screen_position; +#endif + +#ifdef MATERIAL_UNIFORMS_USED layout(set = MATERIAL_UNIFORM_SET, binding = 0, std140) uniform MaterialUniforms{ - /* clang-format off */ -MATERIAL_UNIFORMS - /* clang-format on */ + +#MATERIAL_UNIFORMS + } material; #endif -invariant gl_Position; +float global_time; #ifdef MODE_DUAL_PARABOLOID -layout(location = 8) out float dp_clip; +layout(location = 9) out float dp_clip; #endif -layout(location = 9) out flat uint instance_index; +layout(location = 10) out flat uint instance_index_interp; -/* clang-format off */ +#ifdef USE_MULTIVIEW +#ifdef has_VK_KHR_multiview +#define ViewIndex gl_ViewIndex +#else // has_VK_KHR_multiview +// !BAS! This needs to become an input once we implement our fallback! +#define ViewIndex 0 +#endif // has_VK_KHR_multiview +#else // USE_MULTIVIEW +// Set to zero, not supported in non stereo +#define ViewIndex 0 +#endif //USE_MULTIVIEW -VERTEX_SHADER_GLOBALS +invariant gl_Position; -/* clang-format on */ +#GLOBALS -void main() { +void vertex_shader(in uint instance_index, in bool is_multimesh, in SceneData scene_data, in mat4 model_matrix, out vec4 screen_pos) { vec4 instance_custom = vec4(0.0); #if defined(COLOR_USED) color_interp = color_attrib; #endif - instance_index = draw_call.instance_index; - - bool is_multimesh = bool(instances.data[instance_index].flags & INSTANCE_FLAGS_MULTIMESH); - if (!is_multimesh) { - instance_index += gl_InstanceIndex; - } - - mat4 world_matrix = instances.data[instance_index].transform; - - mat3 world_normal_matrix; + mat3 model_normal_matrix; if (bool(instances.data[instance_index].flags & INSTANCE_FLAGS_NON_UNIFORM_SCALE)) { - world_normal_matrix = inverse(mat3(world_matrix)); + model_normal_matrix = transpose(inverse(mat3(model_matrix))); } else { - world_normal_matrix = mat3(world_matrix); + model_normal_matrix = mat3(model_matrix); } if (is_multimesh) { //multimesh, instances are for it - uint offset = (instances.data[instance_index].flags >> INSTANCE_FLAGS_MULTIMESH_STRIDE_SHIFT) & INSTANCE_FLAGS_MULTIMESH_STRIDE_MASK; - offset *= gl_InstanceIndex; 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 + + uint offset = trail_size * stride * gl_InstanceIndex; + +#ifdef COLOR_USED + vec4 pcolor; +#endif + { + uint boffset = offset + bone_attrib.x * stride; + matrix = mat4(transforms.data[boffset + 0], transforms.data[boffset + 1], transforms.data[boffset + 2], vec4(0.0, 0.0, 0.0, 1.0)) * weight_attrib.x; +#ifdef COLOR_USED + pcolor = transforms.data[boffset + 3] * weight_attrib.x; +#endif + } + if (weight_attrib.y > 0.001) { + uint boffset = offset + bone_attrib.y * stride; + matrix += mat4(transforms.data[boffset + 0], transforms.data[boffset + 1], transforms.data[boffset + 2], vec4(0.0, 0.0, 0.0, 1.0)) * weight_attrib.y; +#ifdef COLOR_USED + pcolor += transforms.data[boffset + 3] * weight_attrib.y; +#endif + } + if (weight_attrib.z > 0.001) { + uint boffset = offset + bone_attrib.z * stride; + matrix += mat4(transforms.data[boffset + 0], transforms.data[boffset + 1], transforms.data[boffset + 2], vec4(0.0, 0.0, 0.0, 1.0)) * weight_attrib.z; +#ifdef COLOR_USED + pcolor += transforms.data[boffset + 3] * weight_attrib.z; +#endif + } + if (weight_attrib.w > 0.001) { + uint boffset = offset + bone_attrib.w * stride; + matrix += mat4(transforms.data[boffset + 0], transforms.data[boffset + 1], transforms.data[boffset + 2], vec4(0.0, 0.0, 0.0, 1.0)) * weight_attrib.w; +#ifdef COLOR_USED + pcolor += transforms.data[boffset + 3] * weight_attrib.w; +#endif + } + + instance_custom = transforms.data[offset + 4]; + +#ifdef COLOR_USED + color_interp *= pcolor; +#endif + +#else + uint stride = 0; + { + //TODO implement a small lookup table for the stride + if (bool(instances.data[instance_index].flags & INSTANCE_FLAGS_MULTIMESH_FORMAT_2D)) { + stride += 2; + } else { + stride += 3; + } + if (bool(instances.data[instance_index].flags & INSTANCE_FLAGS_MULTIMESH_HAS_COLOR)) { + stride += 1; + } + if (bool(instances.data[instance_index].flags & INSTANCE_FLAGS_MULTIMESH_HAS_CUSTOM_DATA)) { + stride += 1; + } + } + + uint offset = stride * gl_InstanceIndex; + 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)); offset += 2; @@ -152,10 +222,11 @@ void main() { instance_custom = transforms.data[offset]; } +#endif //transpose matrix = transpose(matrix); - world_matrix = world_matrix * matrix; - world_normal_matrix = world_normal_matrix * mat3(matrix); + model_matrix = model_matrix * matrix; + model_normal_matrix = model_normal_matrix * mat3(matrix); } vec3 vertex = vertex_attrib; @@ -169,32 +240,6 @@ void main() { vec3 binormal = normalize(cross(normal, tangent) * binormalf); #endif -#if 0 - if (bool(instances.data[instance_index].flags & INSTANCE_FLAGS_SKELETON)) { - //multimesh, instances are for it - - uvec2 bones_01 = uvec2(bone_attrib.x & 0xFFFF, bone_attrib.x >> 16) * 3; - uvec2 bones_23 = uvec2(bone_attrib.y & 0xFFFF, bone_attrib.y >> 16) * 3; - vec2 weights_01 = unpackUnorm2x16(bone_attrib.z); - vec2 weights_23 = unpackUnorm2x16(bone_attrib.w); - - mat4 m = mat4(transforms.data[bones_01.x], transforms.data[bones_01.x + 1], transforms.data[bones_01.x + 2], vec4(0.0, 0.0, 0.0, 1.0)) * weights_01.x; - m += mat4(transforms.data[bones_01.y], transforms.data[bones_01.y + 1], transforms.data[bones_01.y + 2], vec4(0.0, 0.0, 0.0, 1.0)) * weights_01.y; - m += mat4(transforms.data[bones_23.x], transforms.data[bones_23.x + 1], transforms.data[bones_23.x + 2], vec4(0.0, 0.0, 0.0, 1.0)) * weights_23.x; - m += mat4(transforms.data[bones_23.y], transforms.data[bones_23.y + 1], transforms.data[bones_23.y + 2], vec4(0.0, 0.0, 0.0, 1.0)) * weights_23.y; - - //reverse order because its transposed - vertex = (vec4(vertex, 1.0) * m).xyz; - normal = (vec4(normal, 0.0) * m).xyz; - -#if defined(TANGENT_USED) || defined(NORMAL_MAP_USED) || defined(LIGHT_ANISOTROPY_USED) - - tangent = (vec4(tangent, 0.0) * m).xyz; - binormal = (vec4(binormal, 0.0) * m).xyz; -#endif - } -#endif - #ifdef UV_USED uv_interp = uv_attrib; #endif @@ -207,34 +252,38 @@ void main() { vec4 position; #endif +#ifdef USE_MULTIVIEW + mat4 projection_matrix = scene_data.projection_matrix_view[ViewIndex]; + mat4 inv_projection_matrix = scene_data.inv_projection_matrix_view[ViewIndex]; +#else mat4 projection_matrix = scene_data.projection_matrix; + mat4 inv_projection_matrix = scene_data.inv_projection_matrix; +#endif //USE_MULTIVIEW //using world coordinates #if !defined(SKIP_TRANSFORM_USED) && defined(VERTEX_WORLD_COORDS_USED) - vertex = (world_matrix * vec4(vertex, 1.0)).xyz; + vertex = (model_matrix * vec4(vertex, 1.0)).xyz; - normal = world_normal_matrix * normal; +#ifdef NORMAL_USED + normal = model_normal_matrix * normal; +#endif #if defined(TANGENT_USED) || defined(NORMAL_MAP_USED) || defined(LIGHT_ANISOTROPY_USED) - tangent = world_normal_matrix * tangent; - binormal = world_normal_matrix * binormal; + tangent = model_normal_matrix * tangent; + binormal = model_normal_matrix * binormal; #endif #endif float roughness = 1.0; - mat4 modelview = scene_data.inv_camera_matrix * world_matrix; - mat3 modelview_normal = mat3(scene_data.inv_camera_matrix) * world_normal_matrix; + mat4 modelview = scene_data.view_matrix * model_matrix; + mat3 modelview_normal = mat3(scene_data.view_matrix) * model_normal_matrix; { - /* clang-format off */ - -VERTEX_SHADER_CODE - - /* clang-format on */ +#CODE : VERTEX } // using local coordinates (default) @@ -256,17 +305,23 @@ VERTEX_SHADER_CODE //using world coordinates #if !defined(SKIP_TRANSFORM_USED) && defined(VERTEX_WORLD_COORDS_USED) - vertex = (scene_data.inv_camera_matrix * vec4(vertex, 1.0)).xyz; - normal = mat3(scene_data.inverse_normal_matrix) * normal; + vertex = (scene_data.view_matrix * vec4(vertex, 1.0)).xyz; +#ifdef NORMAL_USED + normal = (scene_data.view_matrix * vec4(normal, 0.0)).xyz; +#endif #if defined(TANGENT_USED) || defined(NORMAL_MAP_USED) || defined(LIGHT_ANISOTROPY_USED) - - binormal = mat3(scene_data.camera_inverse_binormal_matrix) * binormal; - tangent = mat3(scene_data.camera_inverse_tangent_matrix) * tangent; + binormal = (scene_data.view_matrix * vec4(binormal, 0.0)).xyz; + tangent = (scene_data.view_matrix * vec4(tangent, 0.0)).xyz; #endif #endif vertex_interp = vertex; + +#ifdef MOTION_VECTORS + screen_pos = projection_matrix * vec4(vertex_interp, 1.0); +#endif + #ifdef NORMAL_USED normal_interp = normal; #endif @@ -321,11 +376,57 @@ VERTEX_SHADER_CODE #endif } +void main() { + uint instance_index = draw_call.instance_index; + + bool is_multimesh = bool(instances.data[instance_index].flags & INSTANCE_FLAGS_MULTIMESH); + if (!is_multimesh) { + instance_index += gl_InstanceIndex; + } + + instance_index_interp = instance_index; + + mat4 model_matrix = instances.data[instance_index].transform; +#if defined(MOTION_VECTORS) + global_time = scene_data_block.prev_data.time; + vertex_shader(instance_index, is_multimesh, scene_data_block.prev_data, instances.data[instance_index].prev_transform, prev_screen_position); + global_time = scene_data_block.data.time; + vertex_shader(instance_index, is_multimesh, scene_data_block.data, model_matrix, screen_position); +#else + global_time = scene_data_block.data.time; + vec4 screen_position; + vertex_shader(instance_index, is_multimesh, scene_data_block.data, model_matrix, screen_position); +#endif +} + #[fragment] #version 450 -VERSION_DEFINES +#VERSION_DEFINES + +#define SHADER_IS_SRGB false + +/* Specialization Constants (Toggles) */ + +layout(constant_id = 0) const bool sc_use_forward_gi = false; +layout(constant_id = 1) const bool sc_use_light_projector = false; +layout(constant_id = 2) const bool sc_use_light_soft_shadows = false; +layout(constant_id = 3) const bool sc_use_directional_soft_shadows = false; + +/* Specialization Constants (Values) */ + +layout(constant_id = 6) const uint sc_soft_shadow_samples = 4; +layout(constant_id = 7) const uint sc_penumbra_shadow_samples = 4; + +layout(constant_id = 8) const uint sc_directional_soft_shadow_samples = 4; +layout(constant_id = 9) const uint sc_directional_penumbra_shadow_samples = 4; + +layout(constant_id = 10) const bool sc_decal_use_mipmaps = true; +layout(constant_id = 11) const bool sc_projector_use_mipmaps = true; + +// not used in clustered renderer but we share some code with the mobile renderer that requires this. +const float sc_luminance_multiplier = 1.0; #include "scene_forward_clustered_inc.glsl" @@ -354,37 +455,58 @@ layout(location = 5) in vec3 tangent_interp; layout(location = 6) in vec3 binormal_interp; #endif +#ifdef MOTION_VECTORS +layout(location = 7) in vec4 screen_position; +layout(location = 8) in vec4 prev_screen_position; +#endif + #ifdef MODE_DUAL_PARABOLOID -layout(location = 8) in float dp_clip; +layout(location = 9) in float dp_clip; #endif -layout(location = 9) in flat uint instance_index; +layout(location = 10) in flat uint instance_index_interp; + +#ifdef USE_MULTIVIEW +#ifdef has_VK_KHR_multiview +#define ViewIndex gl_ViewIndex +#else // has_VK_KHR_multiview +// !BAS! This needs to become an input once we implement our fallback! +#define ViewIndex 0 +#endif // has_VK_KHR_multiview +#else // USE_MULTIVIEW +// Set to zero, not supported in non stereo +#define ViewIndex 0 +#endif //USE_MULTIVIEW //defines to keep compatibility with vertex -#define world_matrix instances.data[instance_index].transform +#define model_matrix instances.data[draw_call.instance_index].transform +#ifdef USE_MULTIVIEW +#define projection_matrix scene_data.projection_matrix_view[ViewIndex] +#define inv_projection_matrix scene_data.inv_projection_matrix_view[ViewIndex] +#else #define projection_matrix scene_data.projection_matrix +#define inv_projection_matrix scene_data.inv_projection_matrix +#endif + +#define global_time scene_data_block.data.time #if defined(ENABLE_SSS) && defined(ENABLE_TRANSMITTANCE) //both required for transmittance to be enabled #define LIGHT_TRANSMITTANCE_USED #endif -#ifdef USE_MATERIAL_UNIFORMS +#ifdef MATERIAL_UNIFORMS_USED layout(set = MATERIAL_UNIFORM_SET, binding = 0, std140) uniform MaterialUniforms{ - /* clang-format off */ -MATERIAL_UNIFORMS - /* clang-format on */ -} material; -#endif -/* clang-format off */ +#MATERIAL_UNIFORMS -FRAGMENT_SHADER_GLOBALS +} material; +#endif -/* clang-format on */ +#GLOBALS #ifdef MODE_RENDER_DEPTH @@ -396,1367 +518,69 @@ layout(location = 2) out vec4 orm_output_buffer; layout(location = 3) out vec4 emission_output_buffer; layout(location = 4) out float depth_output_buffer; -#endif +#endif // MODE_RENDER_MATERIAL #ifdef MODE_RENDER_NORMAL_ROUGHNESS layout(location = 0) out vec4 normal_roughness_output_buffer; -#ifdef MODE_RENDER_GIPROBE -layout(location = 1) out uvec2 giprobe_buffer; +#ifdef MODE_RENDER_VOXEL_GI +layout(location = 1) out uvec2 voxel_gi_buffer; #endif #endif //MODE_RENDER_NORMAL #else // RENDER DEPTH -#ifdef MODE_MULTIPLE_RENDER_TARGETS +#ifdef MODE_SEPARATE_SPECULAR layout(location = 0) out vec4 diffuse_buffer; //diffuse (rgb) and roughness layout(location = 1) out vec4 specular_buffer; //specular and SSS (subsurface scatter) #else layout(location = 0) out vec4 frag_color; -#endif +#endif // MODE_SEPARATE_SPECULAR #endif // RENDER DEPTH -#ifdef ALPHA_HASH_USED - -float hash_2d(vec2 p) { - return fract(1.0e4 * sin(17.0 * p.x + 0.1 * p.y) * - (0.1 + abs(sin(13.0 * p.y + p.x)))); -} - -float hash_3d(vec3 p) { - return hash_2d(vec2(hash_2d(p.xy), p.z)); -} - -float compute_alpha_hash_threshold(vec3 pos, float hash_scale) { - vec3 dx = dFdx(pos); - vec3 dy = dFdx(pos); - float delta_max_sqr = max(length(dx), length(dy)); - float pix_scale = 1.0 / (hash_scale * delta_max_sqr); - - vec2 pix_scales = - vec2(exp2(floor(log2(pix_scale))), exp2(ceil(log2(pix_scale)))); - - vec2 a_thresh = vec2(hash_3d(floor(pix_scales.x * pos.xyz)), - hash_3d(floor(pix_scales.y * pos.xyz))); - - float lerp_factor = fract(log2(pix_scale)); - - float a_interp = (1.0 - lerp_factor) * a_thresh.x + lerp_factor * a_thresh.y; - - float min_lerp = min(lerp_factor, 1.0 - lerp_factor); - - vec3 cases = vec3(a_interp * a_interp / (2.0 * min_lerp * (1.0 - min_lerp)), - (a_interp - 0.5 * min_lerp) / (1.0 - min_lerp), - 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; - - return clamp(alpha_hash_threshold, 0.0, 1.0); -} - -#endif // ALPHA_HASH_USED - -#ifdef ALPHA_ANTIALIASING_EDGE_USED - -float calc_mip_level(vec2 texture_coord) { - vec2 dx = dFdx(texture_coord); - vec2 dy = dFdy(texture_coord); - float delta_max_sqr = max(dot(dx, dx), dot(dy, dy)); - return max(0.0, 0.5 * log2(delta_max_sqr)); -} - -float compute_alpha_antialiasing_edge(float input_alpha, vec2 texture_coord, float alpha_edge) { - input_alpha *= 1.0 + max(0, calc_mip_level(texture_coord)) * 0.25; // 0.25 mip scale, magic number - input_alpha = (input_alpha - alpha_edge) / max(fwidth(input_alpha), 0.0001) + 0.5; - return clamp(input_alpha, 0.0, 1.0); -} - -#endif // ALPHA_ANTIALIASING_USED - -// This returns the G_GGX function divided by 2 cos_theta_m, where in practice cos_theta_m is either N.L or N.V. -// We're dividing this factor off because the overall term we'll end up looks like -// (see, for example, the first unnumbered equation in B. Burley, "Physically Based Shading at Disney", SIGGRAPH 2012): -// -// F(L.V) D(N.H) G(N.L) G(N.V) / (4 N.L N.V) -// -// We're basically regouping this as -// -// F(L.V) D(N.H) [G(N.L)/(2 N.L)] [G(N.V) / (2 N.V)] -// -// and thus, this function implements the [G(N.m)/(2 N.m)] part with m = L or V. -// -// The contents of the D and G (G1) functions (GGX) are taken from -// E. Heitz, "Understanding the Masking-Shadowing Function in Microfacet-Based BRDFs", J. Comp. Graph. Tech. 3 (2) (2014). -// Eqns 71-72 and 85-86 (see also Eqns 43 and 80). - -#if !defined(MODE_RENDER_DEPTH) && !defined(MODE_UNSHADED) - -float G_GGX_2cos(float cos_theta_m, float alpha) { - // Schlick's approximation - // C. Schlick, "An Inexpensive BRDF Model for Physically-based Rendering", Computer Graphics Forum. 13 (3): 233 (1994) - // Eq. (19), although see Heitz (2014) the about the problems with his derivation. - // It nevertheless approximates GGX well with k = alpha/2. - float k = 0.5 * alpha; - return 0.5 / (cos_theta_m * (1.0 - k) + k); - - // float cos2 = cos_theta_m * cos_theta_m; - // float sin2 = (1.0 - cos2); - // return 1.0 / (cos_theta_m + sqrt(cos2 + alpha * alpha * sin2)); -} - -float D_GGX(float cos_theta_m, float alpha) { - float alpha2 = alpha * alpha; - float d = 1.0 + (alpha2 - 1.0) * cos_theta_m * cos_theta_m; - return alpha2 / (M_PI * d * d); -} - -float G_GGX_anisotropic_2cos(float cos_theta_m, float alpha_x, float alpha_y, float cos_phi, float sin_phi) { - float cos2 = cos_theta_m * cos_theta_m; - float sin2 = (1.0 - cos2); - float s_x = alpha_x * cos_phi; - float s_y = alpha_y * sin_phi; - return 1.0 / max(cos_theta_m + sqrt(cos2 + (s_x * s_x + s_y * s_y) * sin2), 0.001); -} - -float D_GGX_anisotropic(float cos_theta_m, float alpha_x, float alpha_y, float cos_phi, float sin_phi) { - float cos2 = cos_theta_m * cos_theta_m; - float sin2 = (1.0 - cos2); - float r_x = cos_phi / alpha_x; - float r_y = sin_phi / alpha_y; - float d = cos2 + sin2 * (r_x * r_x + r_y * r_y); - return 1.0 / max(M_PI * alpha_x * alpha_y * d * d, 0.001); -} - -float SchlickFresnel(float u) { - float m = 1.0 - u; - float m2 = m * m; - return m2 * m2 * m; // pow(m,5) -} - -float GTR1(float NdotH, float a) { - if (a >= 1.0) - return 1.0 / M_PI; - float a2 = a * a; - float t = 1.0 + (a2 - 1.0) * NdotH * NdotH; - return (a2 - 1.0) / (M_PI * log(a2) * t); -} - -vec3 F0(float metallic, float specular, vec3 albedo) { - float dielectric = 0.16 * specular * specular; - // use albedo * metallic as colored specular reflectance at 0 angle for metallic materials; - // see https://google.github.io/filament/Filament.md.html - return mix(vec3(dielectric), albedo, vec3(metallic)); -} - -void light_compute(vec3 N, vec3 L, vec3 V, vec3 light_color, float attenuation, vec3 f0, uint orms, float specular_amount, -#ifdef LIGHT_BACKLIGHT_USED - vec3 backlight, -#endif -#ifdef LIGHT_TRANSMITTANCE_USED - vec4 transmittance_color, - float transmittance_depth, - float transmittance_curve, - float transmittance_boost, - float transmittance_z, -#endif -#ifdef LIGHT_RIM_USED - float rim, float rim_tint, vec3 rim_color, -#endif -#ifdef LIGHT_CLEARCOAT_USED - float clearcoat, float clearcoat_gloss, -#endif -#ifdef LIGHT_ANISOTROPY_USED - vec3 B, vec3 T, float anisotropy, -#endif -#ifdef USE_SOFT_SHADOWS - float A, -#endif -#ifdef USE_SHADOW_TO_OPACITY - inout float alpha, -#endif - inout vec3 diffuse_light, inout vec3 specular_light) { - -#if defined(USE_LIGHT_SHADER_CODE) - // light is written by the light shader - - vec3 normal = N; - vec3 light = L; - vec3 view = V; - - /* clang-format off */ - -LIGHT_SHADER_CODE - - /* clang-format on */ - -#else - -#ifdef USE_SOFT_SHADOWS - float NdotL = min(A + dot(N, L), 1.0); -#else - float NdotL = dot(N, L); -#endif - float cNdotL = max(NdotL, 0.0); // clamped NdotL - float NdotV = dot(N, V); - float cNdotV = max(NdotV, 0.0); - -#if defined(DIFFUSE_BURLEY) || defined(SPECULAR_BLINN) || defined(SPECULAR_SCHLICK_GGX) || defined(LIGHT_CLEARCOAT_USED) - vec3 H = normalize(V + L); -#endif - -#if defined(SPECULAR_BLINN) || defined(SPECULAR_SCHLICK_GGX) || defined(LIGHT_CLEARCOAT_USED) -#ifdef USE_SOFT_SHADOWS - float cNdotH = clamp(A + dot(N, H), 0.0, 1.0); -#else - float cNdotH = clamp(dot(N, H), 0.0, 1.0); -#endif -#endif - -#if defined(DIFFUSE_BURLEY) || defined(SPECULAR_SCHLICK_GGX) || defined(LIGHT_CLEARCOAT_USED) -#ifdef USE_SOFT_SHADOWS - float cLdotH = clamp(A + dot(L, H), 0.0, 1.0); -#else - float cLdotH = clamp(dot(L, H), 0.0, 1.0); -#endif -#endif - - float metallic = unpackUnorm4x8(orms).z; - if (metallic < 1.0) { - float roughness = unpackUnorm4x8(orms).y; - -#if defined(DIFFUSE_OREN_NAYAR) - vec3 diffuse_brdf_NL; -#else - float diffuse_brdf_NL; // BRDF times N.L for calculating diffuse radiance -#endif - -#if defined(DIFFUSE_LAMBERT_WRAP) - // energy conserving lambert wrap shader - diffuse_brdf_NL = max(0.0, (NdotL + roughness) / ((1.0 + roughness) * (1.0 + roughness))); -#elif defined(DIFFUSE_TOON) - - diffuse_brdf_NL = smoothstep(-roughness, max(roughness, 0.01), NdotL); - -#elif defined(DIFFUSE_BURLEY) - - { - float FD90_minus_1 = 2.0 * cLdotH * cLdotH * roughness - 0.5; - float FdV = 1.0 + FD90_minus_1 * SchlickFresnel(cNdotV); - float FdL = 1.0 + FD90_minus_1 * SchlickFresnel(cNdotL); - diffuse_brdf_NL = (1.0 / M_PI) * FdV * FdL * cNdotL; - /* - float energyBias = mix(roughness, 0.0, 0.5); - float energyFactor = mix(roughness, 1.0, 1.0 / 1.51); - float fd90 = energyBias + 2.0 * VoH * VoH * roughness; - float f0 = 1.0; - float lightScatter = f0 + (fd90 - f0) * pow(1.0 - cNdotL, 5.0); - float viewScatter = f0 + (fd90 - f0) * pow(1.0 - cNdotV, 5.0); - - diffuse_brdf_NL = lightScatter * viewScatter * energyFactor; - */ - } -#else - // lambert - diffuse_brdf_NL = cNdotL * (1.0 / M_PI); -#endif - - diffuse_light += light_color * diffuse_brdf_NL * attenuation; - -#if defined(LIGHT_BACKLIGHT_USED) - diffuse_light += light_color * (vec3(1.0 / M_PI) - diffuse_brdf_NL) * backlight * attenuation; -#endif - -#if defined(LIGHT_RIM_USED) - float rim_light = pow(max(0.0, 1.0 - cNdotV), max(0.0, (1.0 - roughness) * 16.0)); - diffuse_light += rim_light * rim * mix(vec3(1.0), rim_color, rim_tint) * light_color; -#endif - -#ifdef LIGHT_TRANSMITTANCE_USED - -#ifdef SSS_MODE_SKIN - - { - float scale = 8.25 / transmittance_depth; - float d = scale * abs(transmittance_z); - float dd = -d * d; - vec3 profile = vec3(0.233, 0.455, 0.649) * exp(dd / 0.0064) + - vec3(0.1, 0.336, 0.344) * exp(dd / 0.0484) + - vec3(0.118, 0.198, 0.0) * exp(dd / 0.187) + - vec3(0.113, 0.007, 0.007) * exp(dd / 0.567) + - vec3(0.358, 0.004, 0.0) * exp(dd / 1.99) + - vec3(0.078, 0.0, 0.0) * exp(dd / 7.41); - - diffuse_light += profile * transmittance_color.a * light_color * clamp(transmittance_boost - NdotL, 0.0, 1.0) * (1.0 / M_PI); - } -#else - - if (transmittance_depth > 0.0) { - float fade = clamp(abs(transmittance_z / transmittance_depth), 0.0, 1.0); - - fade = pow(max(0.0, 1.0 - fade), transmittance_curve); - fade *= clamp(transmittance_boost - NdotL, 0.0, 1.0); - - diffuse_light += transmittance_color.rgb * light_color * (1.0 / M_PI) * transmittance_color.a * fade; - } - -#endif //SSS_MODE_SKIN - -#endif //LIGHT_TRANSMITTANCE_USED - } - - float roughness = unpackUnorm4x8(orms).y; - if (roughness > 0.0) { // FIXME: roughness == 0 should not disable specular light entirely - - // D - -#if defined(SPECULAR_BLINN) - - //normalized blinn - float shininess = exp2(15.0 * (1.0 - roughness) + 1.0) * 0.25; - float blinn = pow(cNdotH, shininess) * cNdotL; - blinn *= (shininess + 8.0) * (1.0 / (8.0 * M_PI)); - float intensity = blinn; - - specular_light += light_color * intensity * attenuation * specular_amount; - -#elif defined(SPECULAR_PHONG) - - vec3 R = normalize(-reflect(L, N)); - float cRdotV = clamp(A + dot(R, V), 0.0, 1.0); - float shininess = exp2(15.0 * (1.0 - roughness) + 1.0) * 0.25; - float phong = pow(cRdotV, shininess); - phong *= (shininess + 8.0) * (1.0 / (8.0 * M_PI)); - float intensity = (phong) / max(4.0 * cNdotV * cNdotL, 0.75); - - specular_light += light_color * intensity * attenuation * specular_amount; - -#elif defined(SPECULAR_TOON) - - vec3 R = normalize(-reflect(L, N)); - float RdotV = dot(R, V); - float mid = 1.0 - roughness; - mid *= mid; - float intensity = smoothstep(mid - roughness * 0.5, mid + roughness * 0.5, RdotV) * mid; - diffuse_light += light_color * intensity * attenuation * specular_amount; // write to diffuse_light, as in toon shading you generally want no reflection - -#elif defined(SPECULAR_DISABLED) - // none.. - -#elif defined(SPECULAR_SCHLICK_GGX) - // shlick+ggx as default - -#if defined(LIGHT_ANISOTROPY_USED) - - float alpha_ggx = roughness * roughness; - float aspect = sqrt(1.0 - anisotropy * 0.9); - float ax = alpha_ggx / aspect; - float ay = alpha_ggx * aspect; - float XdotH = dot(T, H); - float YdotH = dot(B, H); - float D = D_GGX_anisotropic(cNdotH, ax, ay, XdotH, YdotH); - float G = G_GGX_anisotropic_2cos(cNdotL, ax, ay, XdotH, YdotH) * G_GGX_anisotropic_2cos(cNdotV, ax, ay, XdotH, YdotH); - -#else - float alpha_ggx = roughness * roughness; - float D = D_GGX(cNdotH, alpha_ggx); - float G = G_GGX_2cos(cNdotL, alpha_ggx) * G_GGX_2cos(cNdotV, alpha_ggx); -#endif - // F - float cLdotH5 = SchlickFresnel(cLdotH); - vec3 F = mix(vec3(cLdotH5), vec3(1.0), f0); - - vec3 specular_brdf_NL = cNdotL * D * F * G; - - specular_light += specular_brdf_NL * light_color * attenuation * specular_amount; -#endif - -#if defined(LIGHT_CLEARCOAT_USED) - -#if !defined(SPECULAR_SCHLICK_GGX) - float cLdotH5 = SchlickFresnel(cLdotH); +#ifdef MOTION_VECTORS +layout(location = 2) out vec2 motion_vector; #endif - float Dr = GTR1(cNdotH, mix(.1, .001, clearcoat_gloss)); - float Fr = mix(.04, 1.0, cLdotH5); - float Gr = G_GGX_2cos(cNdotL, .25) * G_GGX_2cos(cNdotV, .25); - float clearcoat_specular_brdf_NL = 0.25 * clearcoat * Gr * Fr * Dr * cNdotL; - - specular_light += clearcoat_specular_brdf_NL * light_color * attenuation * specular_amount; -#endif - } - -#ifdef USE_SHADOW_TO_OPACITY - alpha = min(alpha, clamp(1.0 - attenuation), 0.0, 1.0)); -#endif - -#endif //defined(USE_LIGHT_SHADER_CODE) -} - -#ifndef USE_NO_SHADOWS - -// Interleaved Gradient Noise -// http://www.iryoku.com/next-generation-post-processing-in-call-of-duty-advanced-warfare -float quick_hash(vec2 pos) { - const vec3 magic = vec3(0.06711056f, 0.00583715f, 52.9829189f); - return fract(magic.z * fract(dot(pos, magic.xy))); -} - -float sample_directional_pcf_shadow(texture2D shadow, vec2 shadow_pixel_size, vec4 coord) { - vec2 pos = coord.xy; - float depth = coord.z; - - //if only one sample is taken, take it from the center - if (scene_data.directional_soft_shadow_samples == 1) { - return textureProj(sampler2DShadow(shadow, shadow_sampler), vec4(pos, depth, 1.0)); - } - - mat2 disk_rotation; - { - float r = quick_hash(gl_FragCoord.xy) * 2.0 * M_PI; - float sr = sin(r); - float cr = cos(r); - disk_rotation = mat2(vec2(cr, -sr), vec2(sr, cr)); - } - - float avg = 0.0; - - for (uint i = 0; i < scene_data.directional_soft_shadow_samples; i++) { - avg += textureProj(sampler2DShadow(shadow, shadow_sampler), vec4(pos + shadow_pixel_size * (disk_rotation * scene_data.directional_soft_shadow_kernel[i].xy), depth, 1.0)); - } - - return avg * (1.0 / float(scene_data.directional_soft_shadow_samples)); -} - -float sample_pcf_shadow(texture2D shadow, vec2 shadow_pixel_size, vec4 coord) { - vec2 pos = coord.xy; - float depth = coord.z; - - //if only one sample is taken, take it from the center - if (scene_data.soft_shadow_samples == 1) { - return textureProj(sampler2DShadow(shadow, shadow_sampler), vec4(pos, depth, 1.0)); - } - - mat2 disk_rotation; - { - float r = quick_hash(gl_FragCoord.xy) * 2.0 * M_PI; - float sr = sin(r); - float cr = cos(r); - disk_rotation = mat2(vec2(cr, -sr), vec2(sr, cr)); - } - - float avg = 0.0; - - for (uint i = 0; i < scene_data.soft_shadow_samples; i++) { - avg += textureProj(sampler2DShadow(shadow, shadow_sampler), vec4(pos + shadow_pixel_size * (disk_rotation * scene_data.soft_shadow_kernel[i].xy), depth, 1.0)); - } - - return avg * (1.0 / float(scene_data.soft_shadow_samples)); -} - -float sample_directional_soft_shadow(texture2D shadow, vec3 pssm_coord, vec2 tex_scale) { - //find blocker - float blocker_count = 0.0; - float blocker_average = 0.0; - - mat2 disk_rotation; - { - float r = quick_hash(gl_FragCoord.xy) * 2.0 * M_PI; - float sr = sin(r); - float cr = cos(r); - disk_rotation = mat2(vec2(cr, -sr), vec2(sr, cr)); - } - - for (uint i = 0; i < scene_data.directional_penumbra_shadow_samples; i++) { - vec2 suv = pssm_coord.xy + (disk_rotation * scene_data.directional_penumbra_shadow_kernel[i].xy) * tex_scale; - float d = textureLod(sampler2D(shadow, material_samplers[SAMPLER_LINEAR_CLAMP]), suv, 0.0).r; - if (d < pssm_coord.z) { - blocker_average += d; - blocker_count += 1.0; - } - } - - if (blocker_count > 0.0) { - //blockers found, do soft shadow - blocker_average /= blocker_count; - float penumbra = (pssm_coord.z - blocker_average) / blocker_average; - tex_scale *= penumbra; - - float s = 0.0; - for (uint i = 0; i < scene_data.directional_penumbra_shadow_samples; i++) { - vec2 suv = pssm_coord.xy + (disk_rotation * scene_data.directional_penumbra_shadow_kernel[i].xy) * tex_scale; - s += textureProj(sampler2DShadow(shadow, shadow_sampler), vec4(suv, pssm_coord.z, 1.0)); - } - - return s / float(scene_data.directional_penumbra_shadow_samples); - - } else { - //no blockers found, so no shadow - return 1.0; - } -} - -#endif //USE_NO_SHADOWS - -float get_omni_attenuation(float distance, float inv_range, float decay) { - float nd = distance * inv_range; - nd *= nd; - nd *= nd; // nd^4 - nd = max(1.0 - nd, 0.0); - nd *= nd; // nd^2 - return nd * pow(max(distance, 0.0001), -decay); -} - -float light_process_omni_shadow(uint idx, vec3 vertex, vec3 normal) { -#ifndef USE_NO_SHADOWS - if (omni_lights.data[idx].shadow_enabled) { - // there is a shadowmap - - vec3 light_rel_vec = omni_lights.data[idx].position - vertex; - float light_length = length(light_rel_vec); - - vec4 v = vec4(vertex, 1.0); - - vec4 splane = (omni_lights.data[idx].shadow_matrix * v); - float shadow_len = length(splane.xyz); //need to remember shadow len from here - - { - vec3 nofs = normal_interp * omni_lights.data[idx].shadow_normal_bias / omni_lights.data[idx].inv_radius; - nofs *= (1.0 - max(0.0, dot(normalize(light_rel_vec), normalize(normal_interp)))); - v.xyz += nofs; - splane = (omni_lights.data[idx].shadow_matrix * v); - } - - float shadow; - -#ifdef USE_SOFT_SHADOWS - if (omni_lights.data[idx].soft_shadow_size > 0.0) { - //soft shadow - - //find blocker - - float blocker_count = 0.0; - float blocker_average = 0.0; - - mat2 disk_rotation; - { - float r = quick_hash(gl_FragCoord.xy) * 2.0 * M_PI; - float sr = sin(r); - float cr = cos(r); - disk_rotation = mat2(vec2(cr, -sr), vec2(sr, cr)); - } - - vec3 normal = normalize(splane.xyz); - 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)); - vec3 bitangent = normalize(cross(tangent, normal)); - float z_norm = shadow_len * omni_lights.data[idx].inv_radius; - - tangent *= omni_lights.data[idx].soft_shadow_size * omni_lights.data[idx].soft_shadow_scale; - bitangent *= omni_lights.data[idx].soft_shadow_size * omni_lights.data[idx].soft_shadow_scale; - - for (uint i = 0; i < scene_data.penumbra_shadow_samples; i++) { - vec2 disk = disk_rotation * scene_data.penumbra_shadow_kernel[i].xy; - - vec3 pos = splane.xyz + tangent * disk.x + bitangent * disk.y; - - pos = normalize(pos); - vec4 uv_rect = omni_lights.data[idx].atlas_rect; - - if (pos.z >= 0.0) { - pos.z += 1.0; - uv_rect.y += uv_rect.w; - } else { - pos.z = 1.0 - pos.z; - } - - pos.xy /= pos.z; - - pos.xy = pos.xy * 0.5 + 0.5; - pos.xy = uv_rect.xy + pos.xy * uv_rect.zw; - - float d = textureLod(sampler2D(shadow_atlas, material_samplers[SAMPLER_LINEAR_CLAMP]), pos.xy, 0.0).r; - if (d < z_norm) { - blocker_average += d; - blocker_count += 1.0; - } - } - - if (blocker_count > 0.0) { - //blockers found, do soft shadow - blocker_average /= blocker_count; - float penumbra = (z_norm - blocker_average) / blocker_average; - tangent *= penumbra; - bitangent *= penumbra; - - z_norm -= omni_lights.data[idx].inv_radius * omni_lights.data[idx].shadow_bias; - - shadow = 0.0; - for (uint i = 0; i < scene_data.penumbra_shadow_samples; i++) { - vec2 disk = disk_rotation * scene_data.penumbra_shadow_kernel[i].xy; - vec3 pos = splane.xyz + tangent * disk.x + bitangent * disk.y; - - pos = normalize(pos); - vec4 uv_rect = omni_lights.data[idx].atlas_rect; - - if (pos.z >= 0.0) { - pos.z += 1.0; - uv_rect.y += uv_rect.w; - } else { - pos.z = 1.0 - pos.z; - } - - pos.xy /= pos.z; - - pos.xy = pos.xy * 0.5 + 0.5; - pos.xy = uv_rect.xy + pos.xy * uv_rect.zw; - shadow += textureProj(sampler2DShadow(shadow_atlas, shadow_sampler), vec4(pos.xy, z_norm, 1.0)); - } - - shadow /= float(scene_data.penumbra_shadow_samples); - - } else { - //no blockers found, so no shadow - shadow = 1.0; - } - } else { -#endif - splane.xyz = normalize(splane.xyz); - vec4 clamp_rect = omni_lights.data[idx].atlas_rect; - - if (splane.z >= 0.0) { - splane.z += 1.0; - - clamp_rect.y += clamp_rect.w; - - } else { - splane.z = 1.0 - splane.z; - } - - splane.xy /= splane.z; - - splane.xy = splane.xy * 0.5 + 0.5; - splane.z = (shadow_len - omni_lights.data[idx].shadow_bias) * omni_lights.data[idx].inv_radius; - splane.xy = clamp_rect.xy + splane.xy * clamp_rect.zw; - splane.w = 1.0; //needed? i think it should be 1 already - shadow = sample_pcf_shadow(shadow_atlas, omni_lights.data[idx].soft_shadow_scale * scene_data.shadow_atlas_pixel_size, splane); -#ifdef USE_SOFT_SHADOWS - } -#endif - - return shadow; - } -#endif - - return 1.0; -} - -void light_process_omni(uint idx, vec3 vertex, vec3 eye_vec, vec3 normal, vec3 vertex_ddx, vec3 vertex_ddy, vec3 f0, uint orms, float shadow, -#ifdef LIGHT_BACKLIGHT_USED - vec3 backlight, -#endif -#ifdef LIGHT_TRANSMITTANCE_USED - vec4 transmittance_color, - float transmittance_depth, - float transmittance_curve, - float transmittance_boost, -#endif -#ifdef LIGHT_RIM_USED - float rim, float rim_tint, vec3 rim_color, -#endif -#ifdef LIGHT_CLEARCOAT_USED - float clearcoat, float clearcoat_gloss, -#endif -#ifdef LIGHT_ANISOTROPY_USED - vec3 binormal, vec3 tangent, float anisotropy, -#endif -#ifdef USE_SHADOW_TO_OPACITY - inout float alpha, -#endif - inout vec3 diffuse_light, inout vec3 specular_light) { - vec3 light_rel_vec = omni_lights.data[idx].position - vertex; - float light_length = length(light_rel_vec); - float omni_attenuation = get_omni_attenuation(light_length, omni_lights.data[idx].inv_radius, omni_lights.data[idx].attenuation); - float light_attenuation = omni_attenuation; - vec3 color = omni_lights.data[idx].color; - -#ifdef USE_SOFT_SHADOWS - float size_A = 0.0; - - if (omni_lights.data[idx].size > 0.0) { - float t = omni_lights.data[idx].size / max(0.001, light_length); - size_A = max(0.0, 1.0 - 1 / sqrt(1 + t * t)); - } -#endif - -#ifdef LIGHT_TRANSMITTANCE_USED - float transmittance_z = transmittance_depth; //no transmittance by default - transmittance_color.a *= light_attenuation; - { - vec4 clamp_rect = omni_lights.data[idx].atlas_rect; - - //redo shadowmapping, but shrink the model a bit to avoid arctifacts - vec4 splane = (omni_lights.data[idx].shadow_matrix * vec4(vertex - normalize(normal_interp) * omni_lights.data[idx].transmittance_bias, 1.0)); - - shadow_len = length(splane.xyz); - splane = normalize(splane.xyz); - - if (splane.z >= 0.0) { - splane.z += 1.0; - - } else { - splane.z = 1.0 - splane.z; - } - - splane.xy /= splane.z; - splane.xy = splane.xy * 0.5 + 0.5; - splane.z = shadow_len * omni_lights.data[idx].inv_radius; - splane.xy = clamp_rect.xy + splane.xy * clamp_rect.zw; - splane.w = 1.0; //needed? i think it should be 1 already - - float shadow_z = textureLod(sampler2D(shadow_atlas, material_samplers[SAMPLER_LINEAR_CLAMP]), splane.xy, 0.0).r; - transmittance_z = (splane.z - shadow_z) / omni_lights.data[idx].inv_radius; - } -#endif - -#if 0 - - if (omni_lights.data[idx].projector_rect != vec4(0.0)) { - vec3 local_v = (omni_lights.data[idx].shadow_matrix * vec4(vertex, 1.0)).xyz; - local_v = normalize(local_v); - - vec4 atlas_rect = omni_lights.data[idx].projector_rect; - - if (local_v.z >= 0.0) { - local_v.z += 1.0; - atlas_rect.y += atlas_rect.w; - - } else { - local_v.z = 1.0 - local_v.z; - } - - local_v.xy /= local_v.z; - local_v.xy = local_v.xy * 0.5 + 0.5; - vec2 proj_uv = local_v.xy * atlas_rect.zw; - - vec2 proj_uv_ddx; - vec2 proj_uv_ddy; - { - vec3 local_v_ddx = (omni_lights.data[idx].shadow_matrix * vec4(vertex + vertex_ddx, 1.0)).xyz; - local_v_ddx = normalize(local_v_ddx); - - if (local_v_ddx.z >= 0.0) { - local_v_ddx.z += 1.0; - } else { - local_v_ddx.z = 1.0 - local_v_ddx.z; - } - - local_v_ddx.xy /= local_v_ddx.z; - local_v_ddx.xy = local_v_ddx.xy * 0.5 + 0.5; - - proj_uv_ddx = local_v_ddx.xy * atlas_rect.zw - proj_uv; - - vec3 local_v_ddy = (omni_lights.data[idx].shadow_matrix * vec4(vertex + vertex_ddy, 1.0)).xyz; - local_v_ddy = normalize(local_v_ddy); - - if (local_v_ddy.z >= 0.0) { - local_v_ddy.z += 1.0; - } else { - local_v_ddy.z = 1.0 - local_v_ddy.z; - } - - local_v_ddy.xy /= local_v_ddy.z; - local_v_ddy.xy = local_v_ddy.xy * 0.5 + 0.5; - - proj_uv_ddy = local_v_ddy.xy * atlas_rect.zw - proj_uv; - } - - vec4 proj = textureGrad(sampler2D(decal_atlas_srgb, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), proj_uv + atlas_rect.xy, proj_uv_ddx, proj_uv_ddy); - no_shadow = mix(no_shadow, proj.rgb, proj.a); - } -#endif - - light_attenuation *= shadow; - - light_compute(normal, normalize(light_rel_vec), eye_vec, color, light_attenuation, f0, orms, omni_lights.data[idx].specular_amount, -#ifdef LIGHT_BACKLIGHT_USED - backlight, -#endif -#ifdef LIGHT_TRANSMITTANCE_USED - transmittance_color, - transmittance_depth, - transmittance_curve, - transmittance_boost, - transmittance_z, -#endif -#ifdef LIGHT_RIM_USED - rim * omni_attenuation, rim_tint, rim_color, -#endif -#ifdef LIGHT_CLEARCOAT_USED - clearcoat, clearcoat_gloss, -#endif -#ifdef LIGHT_ANISOTROPY_USED - binormal, tangent, anisotropy, -#endif -#ifdef USE_SOFT_SHADOWS - size_A, -#endif -#ifdef USE_SHADOW_TO_OPACITY - alpha, -#endif - diffuse_light, - specular_light); -} - -float light_process_spot_shadow(uint idx, vec3 vertex, vec3 normal) { -#ifndef USE_NO_SHADOWS - if (spot_lights.data[idx].shadow_enabled) { - 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; - //there is a shadowmap - vec4 v = vec4(vertex, 1.0); - - v.xyz -= spot_dir * spot_lights.data[idx].shadow_bias; - - float z_norm = dot(spot_dir, -light_rel_vec) * spot_lights.data[idx].inv_radius; - - float depth_bias_scale = 1.0 / (max(0.0001, z_norm)); //the closer to the light origin, the more you have to offset to reach 1px in the map - vec3 normal_bias = normalize(normal_interp) * (1.0 - max(0.0, dot(spot_dir, -normalize(normal_interp)))) * spot_lights.data[idx].shadow_normal_bias * depth_bias_scale; - normal_bias -= spot_dir * dot(spot_dir, normal_bias); //only XY, no Z - v.xyz += normal_bias; - - //adjust with bias - z_norm = dot(spot_dir, v.xyz - spot_lights.data[idx].position) * spot_lights.data[idx].inv_radius; - - float shadow; - - vec4 splane = (spot_lights.data[idx].shadow_matrix * v); - splane /= splane.w; - -#ifdef USE_SOFT_SHADOWS - if (spot_lights.data[idx].soft_shadow_size > 0.0) { - //soft shadow - - //find blocker - - vec2 shadow_uv = splane.xy * spot_lights.data[idx].atlas_rect.zw + spot_lights.data[idx].atlas_rect.xy; - - float blocker_count = 0.0; - float blocker_average = 0.0; - - mat2 disk_rotation; - { - float r = quick_hash(gl_FragCoord.xy) * 2.0 * M_PI; - float sr = sin(r); - float cr = cos(r); - disk_rotation = mat2(vec2(cr, -sr), vec2(sr, cr)); - } - - float uv_size = spot_lights.data[idx].soft_shadow_size * z_norm * spot_lights.data[idx].soft_shadow_scale; - vec2 clamp_max = spot_lights.data[idx].atlas_rect.xy + spot_lights.data[idx].atlas_rect.zw; - for (uint i = 0; i < scene_data.penumbra_shadow_samples; i++) { - vec2 suv = shadow_uv + (disk_rotation * scene_data.penumbra_shadow_kernel[i].xy) * uv_size; - suv = clamp(suv, spot_lights.data[idx].atlas_rect.xy, clamp_max); - float d = textureLod(sampler2D(shadow_atlas, material_samplers[SAMPLER_LINEAR_CLAMP]), suv, 0.0).r; - if (d < z_norm) { - blocker_average += d; - blocker_count += 1.0; - } - } - - if (blocker_count > 0.0) { - //blockers found, do soft shadow - blocker_average /= blocker_count; - float penumbra = (z_norm - blocker_average) / blocker_average; - uv_size *= penumbra; - - shadow = 0.0; - for (uint i = 0; i < scene_data.penumbra_shadow_samples; i++) { - vec2 suv = shadow_uv + (disk_rotation * scene_data.penumbra_shadow_kernel[i].xy) * uv_size; - suv = clamp(suv, spot_lights.data[idx].atlas_rect.xy, clamp_max); - shadow += textureProj(sampler2DShadow(shadow_atlas, shadow_sampler), vec4(suv, z_norm, 1.0)); - } - - shadow /= float(scene_data.penumbra_shadow_samples); - - } else { - //no blockers found, so no shadow - shadow = 1.0; - } - - } else { -#endif - //hard shadow - vec4 shadow_uv = vec4(splane.xy * spot_lights.data[idx].atlas_rect.zw + spot_lights.data[idx].atlas_rect.xy, splane.z, 1.0); - - shadow = sample_pcf_shadow(shadow_atlas, spot_lights.data[idx].soft_shadow_scale * scene_data.shadow_atlas_pixel_size, shadow_uv); -#ifdef USE_SOFT_SHADOWS - } -#endif - - return shadow; - } - -#endif //USE_NO_SHADOWS - - return 1.0; -} - -void light_process_spot(uint idx, vec3 vertex, vec3 eye_vec, vec3 normal, vec3 vertex_ddx, vec3 vertex_ddy, vec3 f0, uint orms, float shadow, -#ifdef LIGHT_BACKLIGHT_USED - vec3 backlight, -#endif -#ifdef LIGHT_TRANSMITTANCE_USED - vec4 transmittance_color, - float transmittance_depth, - float transmittance_curve, - float transmittance_boost, -#endif -#ifdef LIGHT_RIM_USED - float rim, float rim_tint, vec3 rim_color, -#endif -#ifdef LIGHT_CLEARCOAT_USED - float clearcoat, float clearcoat_gloss, -#endif -#ifdef LIGHT_ANISOTROPY_USED - vec3 binormal, vec3 tangent, float anisotropy, -#endif -#ifdef USE_SHADOW_TO_OPACITY - inout float alpha, -#endif - inout vec3 diffuse_light, - inout vec3 specular_light) { - vec3 light_rel_vec = spot_lights.data[idx].position - vertex; - float light_length = length(light_rel_vec); - float spot_attenuation = get_omni_attenuation(light_length, spot_lights.data[idx].inv_radius, spot_lights.data[idx].attenuation); - vec3 spot_dir = spot_lights.data[idx].direction; - float scos = max(dot(-normalize(light_rel_vec), spot_dir), spot_lights.data[idx].cone_angle); - float spot_rim = max(0.0001, (1.0 - scos) / (1.0 - spot_lights.data[idx].cone_angle)); - spot_attenuation *= 1.0 - pow(spot_rim, spot_lights.data[idx].cone_attenuation); - float light_attenuation = spot_attenuation; - vec3 color = spot_lights.data[idx].color; - float specular_amount = spot_lights.data[idx].specular_amount; - -#ifdef USE_SOFT_SHADOWS - float size_A = 0.0; - - if (spot_lights.data[idx].size > 0.0) { - float t = spot_lights.data[idx].size / max(0.001, light_length); - size_A = max(0.0, 1.0 - 1 / sqrt(1 + t * t)); - } -#endif - - /* - if (spot_lights.data[idx].atlas_rect!=vec4(0.0)) { - //use projector texture - } - */ - -#ifdef LIGHT_TRANSMITTANCE_USED - float transmittance_z = transmittance_depth; - transmittance_color.a *= light_attenuation; - { - splane = (spot_lights.data[idx].shadow_matrix * vec4(vertex - normalize(normal_interp) * spot_lights.data[idx].transmittance_bias, 1.0)); - splane /= splane.w; - splane.xy = splane.xy * spot_lights.data[idx].atlas_rect.zw + spot_lights.data[idx].atlas_rect.xy; - - float shadow_z = textureLod(sampler2D(shadow_atlas, material_samplers[SAMPLER_LINEAR_CLAMP]), splane.xy, 0.0).r; - //reconstruct depth - shadow_z /= spot_lights.data[idx].inv_radius; - //distance to light plane - float z = dot(spot_dir, -light_rel_vec); - transmittance_z = z - shadow_z; - } -#endif //LIGHT_TRANSMITTANCE_USED +#include "scene_forward_aa_inc.glsl" - light_attenuation *= shadow; +#if !defined(MODE_RENDER_DEPTH) && !defined(MODE_UNSHADED) - light_compute(normal, normalize(light_rel_vec), eye_vec, color, light_attenuation, f0, orms, spot_lights.data[idx].specular_amount, -#ifdef LIGHT_BACKLIGHT_USED - backlight, -#endif -#ifdef LIGHT_TRANSMITTANCE_USED - transmittance_color, - transmittance_depth, - transmittance_curve, - transmittance_boost, - transmittance_z, -#endif -#ifdef LIGHT_RIM_USED - rim * spot_attenuation, rim_tint, rim_color, +// Default to SPECULAR_SCHLICK_GGX. +#if !defined(SPECULAR_DISABLED) && !defined(SPECULAR_SCHLICK_GGX) && !defined(SPECULAR_TOON) +#define SPECULAR_SCHLICK_GGX #endif -#ifdef LIGHT_CLEARCOAT_USED - clearcoat, clearcoat_gloss, -#endif -#ifdef LIGHT_ANISOTROPY_USED - binormal, tangent, anisotropy, -#endif -#ifdef USE_SOFT_SHADOW - size_A, -#endif -#ifdef USE_SHADOW_TO_OPACITY - alpha, -#endif - diffuse_light, specular_light); -} - -void reflection_process(uint ref_index, vec3 vertex, vec3 normal, float roughness, vec3 ambient_light, vec3 specular_light, inout vec4 ambient_accum, inout vec4 reflection_accum) { - vec3 box_extents = reflections.data[ref_index].box_extents; - vec3 local_pos = (reflections.data[ref_index].local_matrix * vec4(vertex, 1.0)).xyz; - - if (any(greaterThan(abs(local_pos), box_extents))) { //out of the reflection box - return; - } - - vec3 ref_vec = normalize(reflect(vertex, 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 - blend = mix(length(inner_pos), blend, blend); - blend *= blend; - blend = max(0.0, 1.0 - blend); - - if (reflections.data[ref_index].intensity > 0.0) { // compute reflection - - vec3 local_ref_vec = (reflections.data[ref_index].local_matrix * vec4(ref_vec, 0.0)).xyz; - - if (reflections.data[ref_index].box_project) { //box project - - vec3 nrdir = normalize(local_ref_vec); - vec3 rbmax = (box_extents - local_pos) / nrdir; - vec3 rbmin = (-box_extents - local_pos) / nrdir; - - vec3 rbminmax = mix(rbmin, rbmax, greaterThan(nrdir, vec3(0.0, 0.0, 0.0))); - - float fa = min(min(rbminmax.x, rbminmax.y), rbminmax.z); - vec3 posonbox = local_pos + nrdir * fa; - local_ref_vec = posonbox - reflections.data[ref_index].box_offset; - } - - 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; - - if (reflections.data[ref_index].exterior) { - reflection.rgb = mix(specular_light, reflection.rgb, blend); - } - - reflection.rgb *= reflections.data[ref_index].intensity; //intensity - reflection.a = blend; - reflection.rgb *= reflection.a; - - reflection_accum += reflection; - } - - switch (reflections.data[ref_index].ambient_mode) { - case REFLECTION_AMBIENT_DISABLED: { - //do nothing - } break; - case REFLECTION_AMBIENT_ENVIRONMENT: { - //do nothing - vec3 local_amb_vec = (reflections.data[ref_index].local_matrix * vec4(normal, 0.0)).xyz; - - 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.a = blend; - if (reflections.data[ref_index].exterior) { - ambient_out.rgb = mix(ambient_light, ambient_out.rgb, blend); - } - - ambient_out.rgb *= ambient_out.a; - ambient_accum += ambient_out; - } break; - case REFLECTION_AMBIENT_COLOR: { - vec4 ambient_out; - ambient_out.a = blend; - ambient_out.rgb = reflections.data[ref_index].ambient; - if (reflections.data[ref_index].exterior) { - ambient_out.rgb = mix(ambient_light, ambient_out.rgb, blend); - } - ambient_out.rgb *= ambient_out.a; - ambient_accum += ambient_out; - } break; - } -} - -#ifdef USE_FORWARD_GI - -//standard voxel cone trace -vec4 voxel_cone_trace(texture3D probe, vec3 cell_size, vec3 pos, vec3 direction, float tan_half_angle, float max_distance, float p_bias) { - float dist = p_bias; - vec4 color = vec4(0.0); - - while (dist < max_distance && color.a < 0.95) { - float diameter = max(1.0, 2.0 * tan_half_angle * dist); - vec3 uvw_pos = (pos + dist * direction) * cell_size; - float half_diameter = diameter * 0.5; - //check if outside, then break - if (any(greaterThan(abs(uvw_pos - 0.5), vec3(0.5f + half_diameter * cell_size)))) { - break; - } - vec4 scolor = textureLod(sampler3D(probe, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), uvw_pos, log2(diameter)); - float a = (1.0 - color.a); - color += a * scolor; - dist += half_diameter; - } - - return color; -} - -vec4 voxel_cone_trace_45_degrees(texture3D probe, vec3 cell_size, vec3 pos, vec3 direction, float tan_half_angle, float max_distance, float p_bias) { - float dist = p_bias; - vec4 color = vec4(0.0); - float radius = max(0.5, tan_half_angle * dist); - float lod_level = log2(radius * 2.0); - - while (dist < max_distance && color.a < 0.95) { - vec3 uvw_pos = (pos + dist * direction) * cell_size; - - //check if outside, then break - if (any(greaterThan(abs(uvw_pos - 0.5), vec3(0.5f + radius * cell_size)))) { - break; - } - vec4 scolor = textureLod(sampler3D(probe, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), uvw_pos, lod_level); - lod_level += 1.0; - - float a = (1.0 - color.a); - scolor *= a; - color += scolor; - dist += radius; - radius = max(0.5, tan_half_angle * dist); - } - - return color; -} - -void gi_probe_compute(uint index, vec3 position, vec3 normal, vec3 ref_vec, mat3 normal_xform, float roughness, vec3 ambient, vec3 environment, inout vec4 out_spec, inout vec4 out_diff) { - position = (gi_probes.data[index].xform * vec4(position, 1.0)).xyz; - ref_vec = normalize((gi_probes.data[index].xform * vec4(ref_vec, 0.0)).xyz); - normal = normalize((gi_probes.data[index].xform * vec4(normal, 0.0)).xyz); - - position += normal * gi_probes.data[index].normal_bias; - - //this causes corrupted pixels, i have no idea why.. - if (any(bvec2(any(lessThan(position, vec3(0.0))), any(greaterThan(position, gi_probes.data[index].bounds))))) { - return; - } - - vec3 blendv = abs(position / gi_probes.data[index].bounds * 2.0 - 1.0); - float blend = clamp(1.0 - max(blendv.x, max(blendv.y, blendv.z)), 0.0, 1.0); - //float blend=1.0; - - float max_distance = length(gi_probes.data[index].bounds); - vec3 cell_size = 1.0 / gi_probes.data[index].bounds; - - //radiance - -#define MAX_CONE_DIRS 4 - - vec3 cone_dirs[MAX_CONE_DIRS] = vec3[]( - vec3(0.707107, 0.0, 0.707107), - vec3(0.0, 0.707107, 0.707107), - vec3(-0.707107, 0.0, 0.707107), - vec3(0.0, -0.707107, 0.707107)); - - float cone_weights[MAX_CONE_DIRS] = float[](0.25, 0.25, 0.25, 0.25); - float cone_angle_tan = 0.98269; - - vec3 light = vec3(0.0); - - for (int i = 0; i < MAX_CONE_DIRS; i++) { - vec3 dir = normalize((gi_probes.data[index].xform * vec4(normal_xform * cone_dirs[i], 0.0)).xyz); - - vec4 cone_light = voxel_cone_trace_45_degrees(gi_probe_textures[index], cell_size, position, dir, cone_angle_tan, max_distance, gi_probes.data[index].bias); - - if (gi_probes.data[index].blend_ambient) { - cone_light.rgb = mix(ambient, cone_light.rgb, min(1.0, cone_light.a / 0.95)); - } - - light += cone_weights[i] * cone_light.rgb; - } - - light *= gi_probes.data[index].dynamic_range; - out_diff += vec4(light * blend, blend); - - //irradiance - vec4 irr_light = voxel_cone_trace(gi_probe_textures[index], cell_size, position, ref_vec, tan(roughness * 0.5 * M_PI * 0.99), max_distance, gi_probes.data[index].bias); - if (gi_probes.data[index].blend_ambient) { - irr_light.rgb = mix(environment, irr_light.rgb, min(1.0, irr_light.a / 0.95)); - } - irr_light.rgb *= gi_probes.data[index].dynamic_range; - //irr_light=vec3(0.0); - - out_spec += vec4(irr_light.rgb * blend, blend); -} - -vec2 octahedron_wrap(vec2 v) { - vec2 signVal; - signVal.x = v.x >= 0.0 ? 1.0 : -1.0; - signVal.y = v.y >= 0.0 ? 1.0 : -1.0; - return (1.0 - abs(v.yx)) * signVal; -} - -vec2 octahedron_encode(vec3 n) { - // https://twitter.com/Stubbesaurus/status/937994790553227264 - n /= (abs(n.x) + abs(n.y) + abs(n.z)); - n.xy = n.z >= 0.0 ? n.xy : octahedron_wrap(n.xy); - n.xy = n.xy * 0.5 + 0.5; - return n.xy; -} - -void sdfgi_process(uint cascade, vec3 cascade_pos, vec3 cam_pos, vec3 cam_normal, vec3 cam_specular_normal, bool use_specular, float roughness, out vec3 diffuse_light, out vec3 specular_light, out float blend) { - cascade_pos += cam_normal * sdfgi.normal_bias; - - vec3 base_pos = floor(cascade_pos); - //cascade_pos += mix(vec3(0.0),vec3(0.01),lessThan(abs(cascade_pos-base_pos),vec3(0.01))) * cam_normal; - ivec3 probe_base_pos = ivec3(base_pos); - - vec4 diffuse_accum = vec4(0.0); - vec3 specular_accum; - - ivec3 tex_pos = ivec3(probe_base_pos.xy, int(cascade)); - tex_pos.x += probe_base_pos.z * sdfgi.probe_axis_size; - tex_pos.xy = tex_pos.xy * (SDFGI_OCT_SIZE + 2) + ivec2(1); - - vec3 diffuse_posf = (vec3(tex_pos) + vec3(octahedron_encode(cam_normal) * float(SDFGI_OCT_SIZE), 0.0)) * sdfgi.lightprobe_tex_pixel_size; - - vec3 specular_posf; - - if (use_specular) { - specular_accum = vec3(0.0); - specular_posf = (vec3(tex_pos) + vec3(octahedron_encode(cam_specular_normal) * float(SDFGI_OCT_SIZE), 0.0)) * sdfgi.lightprobe_tex_pixel_size; - } - - vec4 light_accum = vec4(0.0); - float weight_accum = 0.0; - - for (uint j = 0; j < 8; j++) { - ivec3 offset = (ivec3(j) >> ivec3(0, 1, 2)) & ivec3(1, 1, 1); - ivec3 probe_posi = probe_base_pos; - probe_posi += offset; - - // Compute weight - - vec3 probe_pos = vec3(probe_posi); - vec3 probe_to_pos = cascade_pos - probe_pos; - vec3 probe_dir = normalize(-probe_to_pos); - - vec3 trilinear = vec3(1.0) - abs(probe_to_pos); - float weight = trilinear.x * trilinear.y * trilinear.z * max(0.005, dot(cam_normal, probe_dir)); - - // Compute lightprobe occlusion - - if (sdfgi.use_occlusion) { - ivec3 occ_indexv = abs((sdfgi.cascades[cascade].probe_world_offset + probe_posi) & ivec3(1, 1, 1)) * ivec3(1, 2, 4); - vec4 occ_mask = mix(vec4(0.0), vec4(1.0), equal(ivec4(occ_indexv.x | occ_indexv.y), ivec4(0, 1, 2, 3))); - - vec3 occ_pos = clamp(cascade_pos, probe_pos - sdfgi.occlusion_clamp, probe_pos + sdfgi.occlusion_clamp) * sdfgi.probe_to_uvw; - occ_pos.z += float(cascade); - if (occ_indexv.z != 0) { //z bit is on, means index is >=4, so make it switch to the other half of textures - occ_pos.x += 1.0; - } - - occ_pos *= sdfgi.occlusion_renormalize; - float occlusion = dot(textureLod(sampler3D(sdfgi_occlusion_cascades, material_samplers[SAMPLER_LINEAR_CLAMP]), occ_pos, 0.0), occ_mask); - - weight *= max(occlusion, 0.01); - } - - // Compute lightprobe texture position - - vec3 diffuse; - vec3 pos_uvw = diffuse_posf; - pos_uvw.xy += vec2(offset.xy) * sdfgi.lightprobe_uv_offset.xy; - 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); - - if (use_specular) { - vec3 specular = vec3(0.0); - vec3 pos_uvw = specular_posf; - pos_uvw.xy += vec2(offset.xy) * sdfgi.lightprobe_uv_offset.xy; - pos_uvw.x += float(offset.z) * sdfgi.lightprobe_uv_offset.z; - if (roughness < 0.99) { - specular = textureLod(sampler2DArray(sdfgi_lightprobe_texture, material_samplers[SAMPLER_LINEAR_CLAMP]), pos_uvw + vec3(0, 0, float(sdfgi.max_cascades)), 0.0).rgb; - } - if (roughness > 0.5) { - 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; - } - } - - if (diffuse_accum.a > 0.0) { - diffuse_accum.rgb /= diffuse_accum.a; - } - - diffuse_light = diffuse_accum.rgb; - - if (use_specular) { - if (diffuse_accum.a > 0.0) { - specular_accum /= diffuse_accum.a; - } - - specular_light = specular_accum; - } - - { - //process blend - float blend_from = (float(sdfgi.probe_axis_size - 1) / 2.0) - 2.5; - float blend_to = blend_from + 2.0; - vec3 inner_pos = cam_pos * sdfgi.cascades[cascade].to_probe; - - float len = length(inner_pos); - - inner_pos = abs(normalize(inner_pos)); - len *= max(inner_pos.x, max(inner_pos.y, inner_pos.z)); - - if (len >= blend_from) { - blend = smoothstep(blend_from, blend_to, len); - } else { - blend = 0.0; - } - } -} +#include "scene_forward_lights_inc.glsl" -#endif //USE_FORWARD_GI +#include "scene_forward_gi_inc.glsl" #endif //!defined(MODE_RENDER_DEPTH) && !defined(MODE_UNSHADED) #ifndef MODE_RENDER_DEPTH -#ifndef LOW_END_MODE - vec4 volumetric_fog_process(vec2 screen_uv, float z) { - vec3 fog_pos = vec3(screen_uv, z * scene_data.volumetric_fog_inv_length); + vec3 fog_pos = vec3(screen_uv, z * scene_data_block.data.volumetric_fog_inv_length); if (fog_pos.z < 0.0) { return vec4(0.0); } else if (fog_pos.z < 1.0) { - fog_pos.z = pow(fog_pos.z, scene_data.volumetric_fog_detail_spread); + fog_pos.z = pow(fog_pos.z, scene_data_block.data.volumetric_fog_detail_spread); } return texture(sampler3D(volumetric_fog_texture, material_samplers[SAMPLER_LINEAR_CLAMP]), fog_pos); } -#endif vec4 fog_process(vec3 vertex) { - vec3 fog_color = scene_data.fog_light_color; + vec3 fog_color = scene_data_block.data.fog_light_color; - if (scene_data.fog_aerial_perspective > 0.0) { + if (scene_data_block.data.fog_aerial_perspective > 0.0) { vec3 sky_fog_color = vec3(0.0); - vec3 cube_view = scene_data.radiance_inverse_xform * vertex; + vec3 cube_view = scene_data_block.data.radiance_inverse_xform * vertex; // mip_level always reads from the second mipmap and higher so the fog is always slightly blurred - float mip_level = mix(1.0 / MAX_ROUGHNESS_LOD, 1.0, 1.0 - (abs(vertex.z) - scene_data.z_near) / (scene_data.z_far - scene_data.z_near)); + float mip_level = mix(1.0 / MAX_ROUGHNESS_LOD, 1.0, 1.0 - (abs(vertex.z) - scene_data_block.data.z_near) / (scene_data_block.data.z_far - scene_data_block.data.z_near)); #ifdef USE_RADIANCE_CUBEMAP_ARRAY float lod, blend; blend = modf(mip_level * MAX_ROUGHNESS_LOD, lod); @@ -1765,29 +589,29 @@ vec4 fog_process(vec3 vertex) { #else sky_fog_color = textureLod(samplerCube(radiance_cubemap, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), cube_view, mip_level * MAX_ROUGHNESS_LOD).rgb; #endif //USE_RADIANCE_CUBEMAP_ARRAY - fog_color = mix(fog_color, sky_fog_color, scene_data.fog_aerial_perspective); + fog_color = mix(fog_color, sky_fog_color, scene_data_block.data.fog_aerial_perspective); } - if (scene_data.fog_sun_scatter > 0.001) { + if (scene_data_block.data.fog_sun_scatter > 0.001) { vec4 sun_scatter = vec4(0.0); float sun_total = 0.0; vec3 view = normalize(vertex); - for (uint i = 0; i < scene_data.directional_light_count; i++) { + for (uint i = 0; i < scene_data_block.data.directional_light_count; i++) { vec3 light_color = directional_lights.data[i].color * directional_lights.data[i].energy; float light_amount = pow(max(dot(view, directional_lights.data[i].direction), 0.0), 8.0); - fog_color += light_color * light_amount * scene_data.fog_sun_scatter; + fog_color += light_color * light_amount * scene_data_block.data.fog_sun_scatter; } } - float fog_amount = 1.0 - exp(min(0.0, vertex.z * scene_data.fog_density)); + float fog_amount = 1.0 - exp(min(0.0, -length(vertex) * scene_data_block.data.fog_density)); - if (abs(scene_data.fog_height_density) > 0.001) { - float y = (scene_data.camera_matrix * vec4(vertex, 1.0)).y; + if (abs(scene_data_block.data.fog_height_density) >= 0.0001) { + float y = (scene_data_block.data.inv_view_matrix * vec4(vertex, 1.0)).y; - float y_dist = scene_data.fog_height - y; + float y_dist = y - scene_data_block.data.fog_height; - float vfog_amount = clamp(exp(y_dist * scene_data.fog_height_density), 0.0, 1.0); + float vfog_amount = 1.0 - exp(min(0.0, y_dist * scene_data_block.data.fog_height_density)); fog_amount = max(vfog_amount, fog_amount); } @@ -1799,7 +623,6 @@ void cluster_get_item_range(uint p_offset, out uint item_min, out uint item_max, uint item_min_max = cluster_buffer.data[p_offset]; item_min = item_min_max & 0xFFFF; item_max = item_min_max >> 16; - ; item_from = item_min >> 5; item_to = (item_max == 0) ? 0 : ((item_max - 1) >> 5) + 1; //side effect of how it is stored, as item_max 0 means no elements @@ -1811,43 +634,22 @@ uint cluster_get_range_clip_mask(uint i, uint z_min, uint z_max) { return bitfieldInsert(uint(0), uint(0xFFFFFFFF), local_min, mask_width); } -float blur_shadow(float shadow) { - return shadow; -#if 0 - //disabling for now, will investigate later - float interp_shadow = shadow; - if (gl_HelperInvocation) { - interp_shadow = -4.0; // technically anything below -4 will do but just to make sure - } - - uvec2 fc2 = uvec2(gl_FragCoord.xy); - interp_shadow -= dFdx(interp_shadow) * (float(fc2.x & 1) - 0.5); - interp_shadow -= dFdy(interp_shadow) * (float(fc2.y & 1) - 0.5); - - if (interp_shadow >= 0.0) { - shadow = interp_shadow; - } - return shadow; -#endif -} - #endif //!MODE_RENDER DEPTH -void main() { -#ifdef MODE_DUAL_PARABOLOID - - if (dp_clip > 0.0) - discard; -#endif +void fragment_shader(in SceneData scene_data) { + uint instance_index = instance_index_interp; - //lay out everything, whathever is unused is optimized away anyway + //lay out everything, whatever is unused is optimized away anyway vec3 vertex = vertex_interp; +#ifdef USE_MULTIVIEW + vec3 view = -normalize(vertex_interp - scene_data.eye_offset[ViewIndex].xyz); +#else vec3 view = -normalize(vertex_interp); +#endif vec3 albedo = vec3(1.0); vec3 backlight = vec3(0.0); - vec4 transmittance_color = vec4(0.0); + vec4 transmittance_color = vec4(0.0, 0.0, 0.0, 1.0); float transmittance_depth = 0.0; - float transmittance_curve = 1.0; float transmittance_boost = 0.0; float metallic = 0.0; float specular = 0.5; @@ -1856,7 +658,7 @@ void main() { float rim = 0.0; float rim_tint = 0.0; float clearcoat = 0.0; - float clearcoat_gloss = 0.0; + float clearcoat_roughness = 0.0; float anisotropy = 0.0; vec2 anisotropy_flow = vec2(1.0, 0.0); vec4 fog = vec4(0.0); @@ -1870,7 +672,7 @@ void main() { float ao = 1.0; float ao_light_affect = 0.0; - float alpha = 1.0; + float alpha = float(instances.data[instance_index].flags >> INSTANCE_FLAGS_FADE_SHIFT) / float(255.0); #if defined(TANGENT_USED) || defined(NORMAL_MAP_USED) || defined(LIGHT_ANISOTROPY_USED) vec3 binormal = normalize(binormal_interp); @@ -1910,7 +712,7 @@ void main() { float normal_map_depth = 1.0; - vec2 screen_uv = gl_FragCoord.xy * scene_data.screen_pixel_size + scene_data.screen_pixel_size * 0.5; //account for center + vec2 screen_uv = gl_FragCoord.xy * scene_data.screen_pixel_size; float sss_strength = 0.0; @@ -1928,20 +730,12 @@ void main() { #endif // ALPHA_ANTIALIASING_EDGE_USED { - /* clang-format off */ - -FRAGMENT_SHADER_CODE - - /* clang-format on */ +#CODE : FRAGMENT } #ifdef LIGHT_TRANSMITTANCE_USED -#ifdef SSS_MODE_SKIN - transmittance_color.a = sss_strength; -#else transmittance_color.a *= sss_strength; #endif -#endif #ifndef USE_SHADOW_TO_OPACITY @@ -1964,7 +758,7 @@ FRAGMENT_SHADER_CODE #endif #ifdef ALPHA_ANTIALIASING_EDGE_USED -// If alpha scissor is used, we must further the edge threshold, otherwise we wont get any edge feather +// If alpha scissor is used, we must further the edge threshold, otherwise we won't get any edge feather #ifdef ALPHA_SCISSOR_USED alpha_antialiasing_edge = clamp(alpha_scissor_threshold + alpha_antialiasing_edge, 0.0, 1.0); #endif @@ -1972,7 +766,7 @@ FRAGMENT_SHADER_CODE #endif // ALPHA_ANTIALIASING_EDGE_USED #ifdef USE_OPAQUE_PREPASS - if (alpha < opaque_prepass_threshold) { + if (alpha < scene_data.opaque_prepass_threshold) { discard; } #endif // USE_OPAQUE_PREPASS @@ -2019,7 +813,6 @@ FRAGMENT_SHADER_CODE fog = fog_process(vertex); } -#ifndef LOW_END_MODE if (scene_data.volumetric_fog_enabled) { vec4 volumetric_fog = volumetric_fog_process(screen_uv, -vertex.z); if (scene_data.fog_enabled) { @@ -2037,7 +830,6 @@ FRAGMENT_SHADER_CODE fog = volumetric_fog; } } -#endif //!LOW_END_MODE #endif //!CUSTOM_FOG_USED uint fog_rg = packHalf2x16(fog.rg); @@ -2102,25 +894,35 @@ FRAGMENT_SHADER_CODE continue; //out of decal } - //we need ddx/ddy for mipmaps, so simulate them - vec2 ddx = (decals.data[decal_index].xform * vec4(vertex_ddx, 0.0)).xz; - vec2 ddy = (decals.data[decal_index].xform * vec4(vertex_ddy, 0.0)).xz; - float fade = pow(1.0 - (uv_local.y > 0.0 ? uv_local.y : -uv_local.y), uv_local.y > 0.0 ? decals.data[decal_index].upper_fade : decals.data[decal_index].lower_fade); if (decals.data[decal_index].normal_fade > 0.0) { fade *= smoothstep(decals.data[decal_index].normal_fade, 1.0, dot(normal_interp, decals.data[decal_index].normal) * 0.5 + 0.5); } + //we need ddx/ddy for mipmaps, so simulate them + vec2 ddx = (decals.data[decal_index].xform * vec4(vertex_ddx, 0.0)).xz; + vec2 ddy = (decals.data[decal_index].xform * vec4(vertex_ddy, 0.0)).xz; + if (decals.data[decal_index].albedo_rect != vec4(0.0)) { //has albedo - vec4 decal_albedo = textureGrad(sampler2D(decal_atlas_srgb, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), uv_local.xz * decals.data[decal_index].albedo_rect.zw + decals.data[decal_index].albedo_rect.xy, ddx * decals.data[decal_index].albedo_rect.zw, ddy * decals.data[decal_index].albedo_rect.zw); + vec4 decal_albedo; + if (sc_decal_use_mipmaps) { + decal_albedo = textureGrad(sampler2D(decal_atlas_srgb, decal_sampler), uv_local.xz * decals.data[decal_index].albedo_rect.zw + decals.data[decal_index].albedo_rect.xy, ddx * decals.data[decal_index].albedo_rect.zw, ddy * decals.data[decal_index].albedo_rect.zw); + } else { + decal_albedo = textureLod(sampler2D(decal_atlas_srgb, decal_sampler), uv_local.xz * decals.data[decal_index].albedo_rect.zw + decals.data[decal_index].albedo_rect.xy, 0.0); + } decal_albedo *= decals.data[decal_index].modulate; decal_albedo.a *= fade; albedo = mix(albedo, decal_albedo.rgb, decal_albedo.a * decals.data[decal_index].albedo_mix); if (decals.data[decal_index].normal_rect != vec4(0.0)) { - vec3 decal_normal = textureGrad(sampler2D(decal_atlas, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), uv_local.xz * decals.data[decal_index].normal_rect.zw + decals.data[decal_index].normal_rect.xy, ddx * decals.data[decal_index].normal_rect.zw, ddy * decals.data[decal_index].normal_rect.zw).xyz; + vec3 decal_normal; + if (sc_decal_use_mipmaps) { + decal_normal = textureGrad(sampler2D(decal_atlas, decal_sampler), uv_local.xz * decals.data[decal_index].normal_rect.zw + decals.data[decal_index].normal_rect.xy, ddx * decals.data[decal_index].normal_rect.zw, ddy * decals.data[decal_index].normal_rect.zw).xyz; + } else { + decal_normal = textureLod(sampler2D(decal_atlas, decal_sampler), uv_local.xz * decals.data[decal_index].normal_rect.zw + decals.data[decal_index].normal_rect.xy, 0.0).xyz; + } decal_normal.xy = decal_normal.xy * vec2(2.0, -2.0) - vec2(1.0, -1.0); //users prefer flipped y normal maps in most authoring software decal_normal.z = sqrt(max(0.0, 1.0 - dot(decal_normal.xy, decal_normal.xy))); //convert to view space, use xzy because y is up @@ -2130,7 +932,12 @@ FRAGMENT_SHADER_CODE } if (decals.data[decal_index].orm_rect != vec4(0.0)) { - vec3 decal_orm = textureGrad(sampler2D(decal_atlas, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), uv_local.xz * decals.data[decal_index].orm_rect.zw + decals.data[decal_index].orm_rect.xy, ddx * decals.data[decal_index].orm_rect.zw, ddy * decals.data[decal_index].orm_rect.zw).xyz; + vec3 decal_orm; + if (sc_decal_use_mipmaps) { + decal_orm = textureGrad(sampler2D(decal_atlas, decal_sampler), uv_local.xz * decals.data[decal_index].orm_rect.zw + decals.data[decal_index].orm_rect.xy, ddx * decals.data[decal_index].orm_rect.zw, ddy * decals.data[decal_index].orm_rect.zw).xyz; + } else { + decal_orm = textureLod(sampler2D(decal_atlas, decal_sampler), uv_local.xz * decals.data[decal_index].orm_rect.zw + decals.data[decal_index].orm_rect.xy, 0.0).xyz; + } ao = mix(ao, decal_orm.r, decal_albedo.a); roughness = mix(roughness, decal_orm.g, decal_albedo.a); metallic = mix(metallic, decal_orm.b, decal_albedo.a); @@ -2139,7 +946,11 @@ FRAGMENT_SHADER_CODE if (decals.data[decal_index].emission_rect != vec4(0.0)) { //emission is additive, so its independent from albedo - emission += textureGrad(sampler2D(decal_atlas_srgb, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), 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; + 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; + } 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; + } } } } @@ -2152,9 +963,9 @@ FRAGMENT_SHADER_CODE #ifdef NORMAL_USED if (scene_data.roughness_limiter_enabled) { - //http://www.jp.square-enix.com/tech/library/pdf/ImprovedGeometricSpecularAA.pdf + //https://www.jp.square-enix.com/tech/library/pdf/ImprovedGeometricSpecularAA.pdf float roughness2 = roughness * roughness; - vec3 dndu = dFdx(normal), dndv = dFdx(normal); + vec3 dndu = dFdx(normal), dndv = dFdy(normal); float variance = scene_data.roughness_limiter_amount * (dot(dndu, dndu) + dot(dndv, dndv)); float kernelRoughness2 = min(2.0 * variance, scene_data.roughness_limiter_limit); //limit effect float filteredRoughness2 = min(1.0, roughness2 + kernelRoughness2); @@ -2170,7 +981,18 @@ FRAGMENT_SHADER_CODE #if !defined(MODE_RENDER_DEPTH) && !defined(MODE_UNSHADED) if (scene_data.use_reflection_cubemap) { +#ifdef LIGHT_ANISOTROPY_USED + // https://google.github.io/filament/Filament.html#lighting/imagebasedlights/anisotropy + vec3 anisotropic_direction = anisotropy >= 0.0 ? binormal : tangent; + vec3 anisotropic_tangent = cross(anisotropic_direction, view); + vec3 anisotropic_normal = cross(anisotropic_tangent, anisotropic_direction); + vec3 bent_normal = normalize(mix(normal, anisotropic_normal, abs(anisotropy) * clamp(5.0 * roughness, 0.0, 1.0))); + vec3 ref_vec = reflect(-view, bent_normal); +#else vec3 ref_vec = reflect(-view, normal); +#endif + + float horizon = min(1.0 + dot(ref_vec, normal), 1.0); ref_vec = scene_data.radiance_inverse_xform * ref_vec; #ifdef USE_RADIANCE_CUBEMAP_ARRAY @@ -2183,6 +1005,7 @@ FRAGMENT_SHADER_CODE 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 *= horizon * horizon; specular_light *= scene_data.ambient_light_color_energy.a; } @@ -2208,7 +1031,37 @@ FRAGMENT_SHADER_CODE } #endif // USE_LIGHTMAP #if defined(CUSTOM_IRRADIANCE_USED) - ambient_light = mix(specular_light, custom_irradiance.rgb, custom_irradiance.a); + ambient_light = mix(ambient_light, custom_irradiance.rgb, custom_irradiance.a); +#endif + +#ifdef LIGHT_CLEARCOAT_USED + + if (scene_data.use_reflection_cubemap) { + vec3 n = normalize(normal_interp); // We want to use geometric normal, not normal_map + float NoV = max(dot(n, view), 0.0001); + vec3 ref_vec = reflect(-view, n); + // The clear coat layer assumes an IOR of 1.5 (4% reflectance) + float Fc = clearcoat * (0.04 + 0.96 * SchlickFresnel(NoV)); + float attenuation = 1.0 - Fc; + ambient_light *= attenuation; + specular_light *= attenuation; + + float horizon = min(1.0 + dot(ref_vec, normal), 1.0); + ref_vec = scene_data.radiance_inverse_xform * ref_vec; + float roughness_lod = mix(0.001, 0.1, clearcoat_roughness) * MAX_ROUGHNESS_LOD; +#ifdef USE_RADIANCE_CUBEMAP_ARRAY + + float lod, blend; + blend = modf(roughness_lod, lod); + vec3 clearcoat_light = texture(samplerCubeArray(radiance_cubemap, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), vec4(ref_vec, lod)).rgb; + clearcoat_light = mix(clearcoat_light, texture(samplerCubeArray(radiance_cubemap, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), vec4(ref_vec, lod + 1)).rgb, blend); + +#else + vec3 clearcoat_light = textureLod(samplerCube(radiance_cubemap, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), ref_vec, roughness_lod).rgb; + +#endif //USE_RADIANCE_CUBEMAP_ARRAY + specular_light += clearcoat_light * horizon * horizon * Fc * scene_data.ambient_light_color_energy.a; + } #endif #endif //!defined(MODE_RENDER_DEPTH) && !defined(MODE_UNSHADED) @@ -2223,22 +1076,22 @@ FRAGMENT_SHADER_CODE if (bool(instances.data[instance_index].flags & INSTANCE_FLAGS_USE_LIGHTMAP_CAPTURE)) { //has lightmap capture uint index = instances.data[instance_index].gi_offset; - vec3 wnormal = mat3(scene_data.camera_matrix) * normal; + vec3 wnormal = mat3(scene_data.inv_view_matrix) * normal; const float c1 = 0.429043; const float c2 = 0.511664; const float c3 = 0.743125; 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); } 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); @@ -2272,14 +1125,14 @@ FRAGMENT_SHADER_CODE ambient_light += textureLod(sampler2DArray(lightmap_textures[ofs], material_samplers[SAMPLER_LINEAR_CLAMP]), uvw, 0.0).rgb; } } -#elif defined(USE_FORWARD_GI) +#else - if (bool(instances.data[instance_index].flags & INSTANCE_FLAGS_USE_SDFGI)) { //has lightmap capture + if (sc_use_forward_gi && bool(instances.data[instance_index].flags & INSTANCE_FLAGS_USE_SDFGI)) { //has lightmap capture //make vertex orientation the world one, but still align to camera - vec3 cam_pos = mat3(scene_data.camera_matrix) * vertex; - vec3 cam_normal = mat3(scene_data.camera_matrix) * normal; - vec3 cam_reflection = mat3(scene_data.camera_matrix) * reflect(-view, normal); + vec3 cam_pos = mat3(scene_data.inv_view_matrix) * vertex; + vec3 cam_normal = mat3(scene_data.inv_view_matrix) * normal; + vec3 cam_reflection = mat3(scene_data.inv_view_matrix) * reflect(-view, normal); //apply y-mult cam_pos.y *= sdfgi.y_mult; @@ -2346,10 +1199,10 @@ FRAGMENT_SHADER_CODE } } - if (bool(instances.data[instance_index].flags & INSTANCE_FLAGS_USE_GIPROBE)) { // process giprobes + if (sc_use_forward_gi && bool(instances.data[instance_index].flags & INSTANCE_FLAGS_USE_VOXEL_GI)) { // process voxel_gi_instances uint index1 = instances.data[instance_index].gi_offset & 0xFFFF; - vec3 ref_vec = normalize(reflect(normalize(vertex), normal)); + vec3 ref_vec = normalize(reflect(-view, normal)); //find arbitrary tangent and bitangent, then build a matrix vec3 v0 = abs(normal.z) < 0.999 ? vec3(0.0, 0.0, 1.0) : vec3(0.0, 1.0, 0.0); vec3 tangent = normalize(cross(v0, normal)); @@ -2358,12 +1211,12 @@ FRAGMENT_SHADER_CODE vec4 amb_accum = vec4(0.0); vec4 spec_accum = vec4(0.0); - gi_probe_compute(index1, vertex, normal, ref_vec, normal_mat, roughness * roughness, ambient_light, specular_light, spec_accum, amb_accum); + voxel_gi_compute(index1, vertex, normal, ref_vec, normal_mat, roughness * roughness, ambient_light, specular_light, spec_accum, amb_accum); uint index2 = instances.data[instance_index].gi_offset >> 16; if (index2 != 0xFFFF) { - gi_probe_compute(index2, vertex, normal, ref_vec, normal_mat, roughness * roughness, ambient_light, specular_light, spec_accum, amb_accum); + voxel_gi_compute(index2, vertex, normal, ref_vec, normal_mat, roughness * roughness, ambient_light, specular_light, spec_accum, amb_accum); } if (amb_accum.a > 0.0) { @@ -2377,9 +1230,8 @@ FRAGMENT_SHADER_CODE specular_light = spec_accum.rgb; ambient_light = amb_accum.rgb; } -#elif !defined(LOW_END_MODE) - if (bool(instances.data[instance_index].flags & INSTANCE_FLAGS_USE_GI_BUFFERS)) { //use GI buffers + if (!sc_use_forward_gi && bool(instances.data[instance_index].flags & INSTANCE_FLAGS_USE_GI_BUFFERS)) { //use GI buffers vec2 coord; @@ -2410,15 +1262,13 @@ FRAGMENT_SHADER_CODE ambient_light = mix(ambient_light, buffer_ambient.rgb, buffer_ambient.a); specular_light = mix(specular_light, buffer_reflection.rgb, buffer_reflection.a); } -#endif +#endif // !USE_LIGHTMAP -#ifndef LOW_END_MODE - if (scene_data.ssao_enabled) { + if (bool(scene_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); } -#endif //LOW_END_MODE { // process reflections @@ -2461,8 +1311,16 @@ FRAGMENT_SHADER_CODE if (!bool(reflections.data[reflection_index].mask & instances.data[instance_index].layer_mask)) { continue; //not masked } - - reflection_process(reflection_index, vertex, normal, roughness, ambient_light, specular_light, ambient_accum, reflection_accum); +#ifdef LIGHT_ANISOTROPY_USED + // https://google.github.io/filament/Filament.html#lighting/imagebasedlights/anisotropy + vec3 anisotropic_direction = anisotropy >= 0.0 ? binormal : tangent; + vec3 anisotropic_tangent = cross(anisotropic_direction, view); + vec3 anisotropic_normal = cross(anisotropic_tangent, anisotropic_direction); + vec3 bent_normal = normalize(mix(normal, anisotropic_normal, abs(anisotropy) * clamp(5.0 * roughness, 0.0, 1.0))); +#else + vec3 bent_normal = normal; +#endif + reflection_process(reflection_index, view, vertex, bent_normal, roughness, ambient_light, specular_light, ambient_accum, reflection_accum); } } @@ -2484,6 +1342,12 @@ FRAGMENT_SHADER_CODE // 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); @@ -2493,7 +1357,7 @@ FRAGMENT_SHADER_CODE specular_light *= specular * metallic * albedo * 2.0; #else - // scales the specular reflections, needs to be be computed before lighting happens, + // scales the specular reflections, needs to be computed before lighting happens, // but after environment, GI, and reflection probes are added // Environment brdf approximation (Lazarov 2013) // see https://www.unrealengine.com/en-US/blog/physically-based-shading-on-mobile @@ -2518,9 +1382,10 @@ FRAGMENT_SHADER_CODE // LIGHTING #if !defined(MODE_RENDER_DEPTH) && !defined(MODE_UNSHADED) - { //directional light + { // Directional light. - // Do shadow and lighting in two passes to reduce register pressure + // Do shadow and lighting in two passes to reduce register pressure. +#ifndef SHADOWS_DISABLED uint shadow0 = 0; uint shadow1 = 0; @@ -2533,308 +1398,200 @@ FRAGMENT_SHADER_CODE continue; //not masked } + if (directional_lights.data[i].bake_mode == LIGHT_BAKE_STATIC && bool(instances.data[instance_index].flags & INSTANCE_FLAGS_USE_LIGHTMAP)) { + continue; // Statically baked light and object uses lightmap, skip + } + float shadow = 1.0; -#ifdef USE_SOFT_SHADOWS - //version with soft shadows, more expensive if (directional_lights.data[i].shadow_enabled) { float depth_z = -vertex.z; - - vec4 pssm_coord; - vec3 shadow_color = vec3(0.0); 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)))); -#define BIAS_FUNC(m_var, m_idx) \ - m_var.xyz += light_dir * directional_lights.data[i].shadow_bias[m_idx]; \ - vec3 normal_bias = normalize(normal_interp) * (1.0 - max(0.0, dot(light_dir, -normalize(normal_interp)))) * directional_lights.data[i].shadow_normal_bias[m_idx]; \ - normal_bias -= light_dir * dot(light_dir, normal_bias); \ +#define BIAS_FUNC(m_var, m_idx) \ + m_var.xyz += light_dir * directional_lights.data[i].shadow_bias[m_idx]; \ + vec3 normal_bias = base_normal_bias * directional_lights.data[i].shadow_normal_bias[m_idx]; \ + normal_bias -= light_dir * dot(light_dir, normal_bias); \ m_var.xyz += normal_bias; - if (depth_z < directional_lights.data[i].shadow_split_offsets.x) { - vec4 v = vec4(vertex, 1.0); + //version with soft shadows, more expensive + if (sc_use_directional_soft_shadows && directional_lights.data[i].softshadow_angle > 0) { + uint blend_count = 0; + const uint blend_max = directional_lights.data[i].blend_splits ? 2 : 1; - BIAS_FUNC(v, 0) + if (depth_z < directional_lights.data[i].shadow_split_offsets.x) { + vec4 v = vec4(vertex, 1.0); - pssm_coord = (directional_lights.data[i].shadow_matrix1 * v); - pssm_coord /= pssm_coord.w; + BIAS_FUNC(v, 0) + + vec4 pssm_coord = (directional_lights.data[i].shadow_matrix1 * v); + pssm_coord /= pssm_coord.w; - if (directional_lights.data[i].softshadow_angle > 0) { float range_pos = dot(directional_lights.data[i].direction, v.xyz); float range_begin = directional_lights.data[i].shadow_range_begin.x; float test_radius = (range_pos - range_begin) * directional_lights.data[i].softshadow_angle; vec2 tex_scale = directional_lights.data[i].uv_scale1 * test_radius; shadow = sample_directional_soft_shadow(directional_shadow_atlas, pssm_coord.xyz, tex_scale * directional_lights.data[i].soft_shadow_scale); - } else { - shadow = sample_directional_pcf_shadow(directional_shadow_atlas, scene_data.directional_shadow_pixel_size * directional_lights.data[i].soft_shadow_scale, pssm_coord); + blend_count++; } - shadow_color = directional_lights.data[i].shadow_color1.rgb; - - } else if (depth_z < directional_lights.data[i].shadow_split_offsets.y) { - vec4 v = vec4(vertex, 1.0); + if (blend_count < blend_max && depth_z < directional_lights.data[i].shadow_split_offsets.y) { + vec4 v = vec4(vertex, 1.0); - BIAS_FUNC(v, 1) + BIAS_FUNC(v, 1) - pssm_coord = (directional_lights.data[i].shadow_matrix2 * v); - pssm_coord /= pssm_coord.w; + vec4 pssm_coord = (directional_lights.data[i].shadow_matrix2 * v); + pssm_coord /= pssm_coord.w; - if (directional_lights.data[i].softshadow_angle > 0) { float range_pos = dot(directional_lights.data[i].direction, v.xyz); float range_begin = directional_lights.data[i].shadow_range_begin.y; float test_radius = (range_pos - range_begin) * directional_lights.data[i].softshadow_angle; vec2 tex_scale = directional_lights.data[i].uv_scale2 * test_radius; - shadow = sample_directional_soft_shadow(directional_shadow_atlas, pssm_coord.xyz, tex_scale * directional_lights.data[i].soft_shadow_scale); - } else { - shadow = sample_directional_pcf_shadow(directional_shadow_atlas, scene_data.directional_shadow_pixel_size * directional_lights.data[i].soft_shadow_scale, pssm_coord); + float s = sample_directional_soft_shadow(directional_shadow_atlas, pssm_coord.xyz, tex_scale * directional_lights.data[i].soft_shadow_scale); + + if (blend_count == 0) { + shadow = s; + } else { + //blend + float blend = smoothstep(0.0, directional_lights.data[i].shadow_split_offsets.x, depth_z); + shadow = mix(shadow, s, blend); + } + + blend_count++; } - shadow_color = directional_lights.data[i].shadow_color2.rgb; - } else if (depth_z < directional_lights.data[i].shadow_split_offsets.z) { - vec4 v = vec4(vertex, 1.0); + if (blend_count < blend_max && depth_z < directional_lights.data[i].shadow_split_offsets.z) { + vec4 v = vec4(vertex, 1.0); - BIAS_FUNC(v, 2) + BIAS_FUNC(v, 2) - pssm_coord = (directional_lights.data[i].shadow_matrix3 * v); - pssm_coord /= pssm_coord.w; + vec4 pssm_coord = (directional_lights.data[i].shadow_matrix3 * v); + pssm_coord /= pssm_coord.w; - if (directional_lights.data[i].softshadow_angle > 0) { float range_pos = dot(directional_lights.data[i].direction, v.xyz); float range_begin = directional_lights.data[i].shadow_range_begin.z; float test_radius = (range_pos - range_begin) * directional_lights.data[i].softshadow_angle; vec2 tex_scale = directional_lights.data[i].uv_scale3 * test_radius; - shadow = sample_directional_soft_shadow(directional_shadow_atlas, pssm_coord.xyz, tex_scale * directional_lights.data[i].soft_shadow_scale); - } else { - shadow = sample_directional_pcf_shadow(directional_shadow_atlas, scene_data.directional_shadow_pixel_size * directional_lights.data[i].soft_shadow_scale, pssm_coord); - } + float s = sample_directional_soft_shadow(directional_shadow_atlas, pssm_coord.xyz, tex_scale * directional_lights.data[i].soft_shadow_scale); + + if (blend_count == 0) { + shadow = s; + } else { + //blend + float blend = smoothstep(directional_lights.data[i].shadow_split_offsets.x, directional_lights.data[i].shadow_split_offsets.y, depth_z); + shadow = mix(shadow, s, blend); + } - shadow_color = directional_lights.data[i].shadow_color3.rgb; + blend_count++; + } - } else { - vec4 v = vec4(vertex, 1.0); + if (blend_count < blend_max) { + vec4 v = vec4(vertex, 1.0); - BIAS_FUNC(v, 3) + BIAS_FUNC(v, 3) - pssm_coord = (directional_lights.data[i].shadow_matrix4 * v); - pssm_coord /= pssm_coord.w; + vec4 pssm_coord = (directional_lights.data[i].shadow_matrix4 * v); + pssm_coord /= pssm_coord.w; - if (directional_lights.data[i].softshadow_angle > 0) { float range_pos = dot(directional_lights.data[i].direction, v.xyz); float range_begin = directional_lights.data[i].shadow_range_begin.w; float test_radius = (range_pos - range_begin) * directional_lights.data[i].softshadow_angle; vec2 tex_scale = directional_lights.data[i].uv_scale4 * test_radius; - shadow = sample_directional_soft_shadow(directional_shadow_atlas, pssm_coord.xyz, tex_scale * directional_lights.data[i].soft_shadow_scale); - } else { - shadow = sample_directional_pcf_shadow(directional_shadow_atlas, scene_data.directional_shadow_pixel_size * directional_lights.data[i].soft_shadow_scale, pssm_coord); + float s = sample_directional_soft_shadow(directional_shadow_atlas, pssm_coord.xyz, tex_scale * directional_lights.data[i].soft_shadow_scale); + + if (blend_count == 0) { + shadow = s; + } else { + //blend + float blend = smoothstep(directional_lights.data[i].shadow_split_offsets.y, directional_lights.data[i].shadow_split_offsets.z, depth_z); + shadow = mix(shadow, s, blend); + } } - shadow_color = directional_lights.data[i].shadow_color4.rgb; - } + } else { //no soft shadows - if (directional_lights.data[i].blend_splits) { - vec3 shadow_color_blend = vec3(0.0); - float pssm_blend; - float shadow2; + vec4 pssm_coord; + float blur_factor; if (depth_z < directional_lights.data[i].shadow_split_offsets.x) { vec4 v = vec4(vertex, 1.0); - BIAS_FUNC(v, 1) - pssm_coord = (directional_lights.data[i].shadow_matrix2 * v); - pssm_coord /= pssm_coord.w; - if (directional_lights.data[i].softshadow_angle > 0) { - float range_pos = dot(directional_lights.data[i].direction, v.xyz); - float range_begin = directional_lights.data[i].shadow_range_begin.y; - float test_radius = (range_pos - range_begin) * directional_lights.data[i].softshadow_angle; - vec2 tex_scale = directional_lights.data[i].uv_scale2 * test_radius; - shadow2 = sample_directional_soft_shadow(directional_shadow_atlas, pssm_coord.xyz, tex_scale * directional_lights.data[i].soft_shadow_scale); - } else { - shadow2 = sample_directional_pcf_shadow(directional_shadow_atlas, scene_data.directional_shadow_pixel_size * directional_lights.data[i].soft_shadow_scale, pssm_coord); - } + BIAS_FUNC(v, 0) - pssm_blend = smoothstep(0.0, directional_lights.data[i].shadow_split_offsets.x, depth_z); - shadow_color_blend = directional_lights.data[i].shadow_color2.rgb; + pssm_coord = (directional_lights.data[i].shadow_matrix1 * v); + blur_factor = 1.0; } else if (depth_z < directional_lights.data[i].shadow_split_offsets.y) { vec4 v = vec4(vertex, 1.0); - BIAS_FUNC(v, 2) - pssm_coord = (directional_lights.data[i].shadow_matrix3 * v); - pssm_coord /= pssm_coord.w; - if (directional_lights.data[i].softshadow_angle > 0) { - float range_pos = dot(directional_lights.data[i].direction, v.xyz); - float range_begin = directional_lights.data[i].shadow_range_begin.z; - float test_radius = (range_pos - range_begin) * directional_lights.data[i].softshadow_angle; - vec2 tex_scale = directional_lights.data[i].uv_scale3 * test_radius; - shadow2 = sample_directional_soft_shadow(directional_shadow_atlas, pssm_coord.xyz, tex_scale * directional_lights.data[i].soft_shadow_scale); - } else { - shadow2 = sample_directional_pcf_shadow(directional_shadow_atlas, scene_data.directional_shadow_pixel_size * directional_lights.data[i].soft_shadow_scale, pssm_coord); - } - - pssm_blend = smoothstep(directional_lights.data[i].shadow_split_offsets.x, directional_lights.data[i].shadow_split_offsets.y, depth_z); + BIAS_FUNC(v, 1) - shadow_color_blend = directional_lights.data[i].shadow_color3.rgb; + pssm_coord = (directional_lights.data[i].shadow_matrix2 * v); + // Adjust shadow blur with reference to the first split to reduce discrepancy between shadow splits. + blur_factor = directional_lights.data[i].shadow_split_offsets.x / directional_lights.data[i].shadow_split_offsets.y; } else if (depth_z < directional_lights.data[i].shadow_split_offsets.z) { vec4 v = vec4(vertex, 1.0); - BIAS_FUNC(v, 3) - pssm_coord = (directional_lights.data[i].shadow_matrix4 * v); - pssm_coord /= pssm_coord.w; - if (directional_lights.data[i].softshadow_angle > 0) { - float range_pos = dot(directional_lights.data[i].direction, v.xyz); - float range_begin = directional_lights.data[i].shadow_range_begin.w; - float test_radius = (range_pos - range_begin) * directional_lights.data[i].softshadow_angle; - vec2 tex_scale = directional_lights.data[i].uv_scale4 * test_radius; - shadow2 = sample_directional_soft_shadow(directional_shadow_atlas, pssm_coord.xyz, tex_scale * directional_lights.data[i].soft_shadow_scale); - } else { - shadow2 = sample_directional_pcf_shadow(directional_shadow_atlas, scene_data.directional_shadow_pixel_size * directional_lights.data[i].soft_shadow_scale, pssm_coord); - } - - pssm_blend = smoothstep(directional_lights.data[i].shadow_split_offsets.y, directional_lights.data[i].shadow_split_offsets.z, depth_z); - shadow_color_blend = directional_lights.data[i].shadow_color4.rgb; - } else { - pssm_blend = 0.0; //if no blend, same coord will be used (divide by z will result in same value, and already cached) - } - - pssm_blend = sqrt(pssm_blend); - - shadow = mix(shadow, shadow2, pssm_blend); - shadow_color = mix(shadow_color, shadow_color_blend, pssm_blend); - } - - shadow = mix(shadow, 1.0, smoothstep(directional_lights.data[i].fade_from, directional_lights.data[i].fade_to, vertex.z)); //done with negative values for performance - -#undef BIAS_FUNC - } -#else - // Soft shadow disabled version - - if (directional_lights.data[i].shadow_enabled) { - float depth_z = -vertex.z; - - vec4 pssm_coord; - 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)))); - -#define BIAS_FUNC(m_var, m_idx) \ - m_var.xyz += light_dir * directional_lights.data[i].shadow_bias[m_idx]; \ - vec3 normal_bias = base_normal_bias * directional_lights.data[i].shadow_normal_bias[m_idx]; \ - normal_bias -= light_dir * dot(light_dir, normal_bias); \ - m_var.xyz += normal_bias; - - if (depth_z < directional_lights.data[i].shadow_split_offsets.x) { - vec4 v = vec4(vertex, 1.0); - BIAS_FUNC(v, 0) - - pssm_coord = (directional_lights.data[i].shadow_matrix1 * v); -#ifdef LIGHT_TRANSMITTANCE_USED - { - vec4 trans_vertex = vec4(vertex - normalize(normal_interp) * directional_lights.data[i].shadow_transmittance_bias.x, 1.0); - vec4 trans_coord = directional_lights.data[i].shadow_matrix1 * trans_vertex; - trans_coord /= trans_coord.w; - - float shadow_z = textureLod(sampler2D(directional_shadow_atlas, material_samplers[SAMPLER_LINEAR_CLAMP]), trans_coord.xy, 0.0).r; - shadow_z *= directional_lights.data[i].shadow_transmittance_z_scale.x; - float z = trans_coord.z * directional_lights.data[i].shadow_transmittance_z_scale.x; - - transmittance_z = z - shadow_z; - } -#endif - } else if (depth_z < directional_lights.data[i].shadow_split_offsets.y) { - vec4 v = vec4(vertex, 1.0); - - BIAS_FUNC(v, 1) - - pssm_coord = (directional_lights.data[i].shadow_matrix2 * v); -#ifdef LIGHT_TRANSMITTANCE_USED - { - vec4 trans_vertex = vec4(vertex - normalize(normal_interp) * directional_lights.data[i].shadow_transmittance_bias.y, 1.0); - vec4 trans_coord = directional_lights.data[i].shadow_matrix2 * trans_vertex; - trans_coord /= trans_coord.w; - - float shadow_z = textureLod(sampler2D(directional_shadow_atlas, material_samplers[SAMPLER_LINEAR_CLAMP]), trans_coord.xy, 0.0).r; - shadow_z *= directional_lights.data[i].shadow_transmittance_z_scale.y; - float z = trans_coord.z * directional_lights.data[i].shadow_transmittance_z_scale.y; - - transmittance_z = z - shadow_z; - } -#endif - } else if (depth_z < directional_lights.data[i].shadow_split_offsets.z) { - vec4 v = vec4(vertex, 1.0); - - BIAS_FUNC(v, 2) - - pssm_coord = (directional_lights.data[i].shadow_matrix3 * v); -#ifdef LIGHT_TRANSMITTANCE_USED - { - vec4 trans_vertex = vec4(vertex - normalize(normal_interp) * directional_lights.data[i].shadow_transmittance_bias.z, 1.0); - vec4 trans_coord = directional_lights.data[i].shadow_matrix3 * trans_vertex; - trans_coord /= trans_coord.w; - - float shadow_z = textureLod(sampler2D(directional_shadow_atlas, material_samplers[SAMPLER_LINEAR_CLAMP]), trans_coord.xy, 0.0).r; - shadow_z *= directional_lights.data[i].shadow_transmittance_z_scale.z; - float z = trans_coord.z * directional_lights.data[i].shadow_transmittance_z_scale.z; - - transmittance_z = z - shadow_z; - } -#endif - - } else { - vec4 v = vec4(vertex, 1.0); - - BIAS_FUNC(v, 3) - - pssm_coord = (directional_lights.data[i].shadow_matrix4 * v); -#ifdef LIGHT_TRANSMITTANCE_USED - { - vec4 trans_vertex = vec4(vertex - normalize(normal_interp) * directional_lights.data[i].shadow_transmittance_bias.w, 1.0); - vec4 trans_coord = directional_lights.data[i].shadow_matrix4 * trans_vertex; - trans_coord /= trans_coord.w; - - float shadow_z = textureLod(sampler2D(directional_shadow_atlas, material_samplers[SAMPLER_LINEAR_CLAMP]), trans_coord.xy, 0.0).r; - shadow_z *= directional_lights.data[i].shadow_transmittance_z_scale.w; - float z = trans_coord.z * directional_lights.data[i].shadow_transmittance_z_scale.w; - - transmittance_z = z - shadow_z; - } -#endif - } - - pssm_coord /= pssm_coord.w; - - shadow = sample_directional_pcf_shadow(directional_shadow_atlas, scene_data.directional_shadow_pixel_size * directional_lights.data[i].soft_shadow_scale, pssm_coord); - - if (directional_lights.data[i].blend_splits) { - float pssm_blend; - - if (depth_z < directional_lights.data[i].shadow_split_offsets.x) { - vec4 v = vec4(vertex, 1.0); - BIAS_FUNC(v, 1) - pssm_coord = (directional_lights.data[i].shadow_matrix2 * v); - pssm_blend = smoothstep(0.0, directional_lights.data[i].shadow_split_offsets.x, depth_z); - } else if (depth_z < directional_lights.data[i].shadow_split_offsets.y) { - vec4 v = vec4(vertex, 1.0); BIAS_FUNC(v, 2) + pssm_coord = (directional_lights.data[i].shadow_matrix3 * v); - pssm_blend = smoothstep(directional_lights.data[i].shadow_split_offsets.x, directional_lights.data[i].shadow_split_offsets.y, depth_z); - } else if (depth_z < directional_lights.data[i].shadow_split_offsets.z) { + // Adjust shadow blur with reference to the first split to reduce discrepancy between shadow splits. + blur_factor = directional_lights.data[i].shadow_split_offsets.x / directional_lights.data[i].shadow_split_offsets.z; + } else { vec4 v = vec4(vertex, 1.0); + BIAS_FUNC(v, 3) + pssm_coord = (directional_lights.data[i].shadow_matrix4 * v); - pssm_blend = smoothstep(directional_lights.data[i].shadow_split_offsets.y, directional_lights.data[i].shadow_split_offsets.z, depth_z); - } else { - pssm_blend = 0.0; //if no blend, same coord will be used (divide by z will result in same value, and already cached) + // Adjust shadow blur with reference to the first split to reduce discrepancy between shadow splits. + blur_factor = directional_lights.data[i].shadow_split_offsets.x / directional_lights.data[i].shadow_split_offsets.w; } pssm_coord /= pssm_coord.w; - float shadow2 = sample_directional_pcf_shadow(directional_shadow_atlas, scene_data.directional_shadow_pixel_size * directional_lights.data[i].soft_shadow_scale, pssm_coord); - shadow = mix(shadow, shadow2, pssm_blend); + shadow = sample_directional_pcf_shadow(directional_shadow_atlas, scene_data.directional_shadow_pixel_size * directional_lights.data[i].soft_shadow_scale * blur_factor, pssm_coord); + + if (directional_lights.data[i].blend_splits) { + float pssm_blend; + float blur_factor2; + + if (depth_z < directional_lights.data[i].shadow_split_offsets.x) { + vec4 v = vec4(vertex, 1.0); + BIAS_FUNC(v, 1) + pssm_coord = (directional_lights.data[i].shadow_matrix2 * v); + pssm_blend = smoothstep(0.0, directional_lights.data[i].shadow_split_offsets.x, depth_z); + // Adjust shadow blur with reference to the first split to reduce discrepancy between shadow splits. + blur_factor2 = directional_lights.data[i].shadow_split_offsets.x / directional_lights.data[i].shadow_split_offsets.y; + } else if (depth_z < directional_lights.data[i].shadow_split_offsets.y) { + vec4 v = vec4(vertex, 1.0); + BIAS_FUNC(v, 2) + pssm_coord = (directional_lights.data[i].shadow_matrix3 * v); + pssm_blend = smoothstep(directional_lights.data[i].shadow_split_offsets.x, directional_lights.data[i].shadow_split_offsets.y, depth_z); + // Adjust shadow blur with reference to the first split to reduce discrepancy between shadow splits. + blur_factor2 = directional_lights.data[i].shadow_split_offsets.x / directional_lights.data[i].shadow_split_offsets.z; + } else if (depth_z < directional_lights.data[i].shadow_split_offsets.z) { + vec4 v = vec4(vertex, 1.0); + BIAS_FUNC(v, 3) + pssm_coord = (directional_lights.data[i].shadow_matrix4 * v); + pssm_blend = smoothstep(directional_lights.data[i].shadow_split_offsets.y, directional_lights.data[i].shadow_split_offsets.z, depth_z); + // Adjust shadow blur with reference to the first split to reduce discrepancy between shadow splits. + blur_factor2 = directional_lights.data[i].shadow_split_offsets.x / directional_lights.data[i].shadow_split_offsets.w; + } else { + pssm_blend = 0.0; //if no blend, same coord will be used (divide by z will result in same value, and already cached) + blur_factor2 = 1.0; + } + + pssm_coord /= pssm_coord.w; + + float shadow2 = sample_directional_pcf_shadow(directional_shadow_atlas, scene_data.directional_shadow_pixel_size * directional_lights.data[i].soft_shadow_scale * blur_factor2, pssm_coord); + shadow = mix(shadow, shadow2, pssm_blend); + } } shadow = mix(shadow, 1.0, smoothstep(directional_lights.data[i].fade_from, directional_lights.data[i].fade_to, vertex.z)); //done with negative values for performance #undef BIAS_FUNC - } -#endif + } // shadows if (i < 4) { shadow0 |= uint(clamp(shadow * 255.0, 0.0, 255.0)) << (i * 8); @@ -2842,6 +1599,7 @@ FRAGMENT_SHADER_CODE shadow1 |= uint(clamp(shadow * 255.0, 0.0, 255.0)) << ((i - 4) * 8); } } +#endif // SHADOWS_DISABLED for (uint i = 0; i < 8; i++) { if (i >= scene_data.directional_light_count) { @@ -2864,8 +1622,8 @@ FRAGMENT_SHADER_CODE trans_coord /= trans_coord.w; float shadow_z = textureLod(sampler2D(directional_shadow_atlas, material_samplers[SAMPLER_LINEAR_CLAMP]), trans_coord.xy, 0.0).r; - shadow_z *= directional_lights.data[i].shadow_transmittance_z_scale.x; - float z = trans_coord.z * directional_lights.data[i].shadow_transmittance_z_scale.x; + shadow_z *= directional_lights.data[i].shadow_z_range.x; + float z = trans_coord.z * directional_lights.data[i].shadow_z_range.x; transmittance_z = z - shadow_z; } else if (depth_z < directional_lights.data[i].shadow_split_offsets.y) { @@ -2874,8 +1632,8 @@ FRAGMENT_SHADER_CODE trans_coord /= trans_coord.w; float shadow_z = textureLod(sampler2D(directional_shadow_atlas, material_samplers[SAMPLER_LINEAR_CLAMP]), trans_coord.xy, 0.0).r; - shadow_z *= directional_lights.data[i].shadow_transmittance_z_scale.y; - float z = trans_coord.z * directional_lights.data[i].shadow_transmittance_z_scale.y; + shadow_z *= directional_lights.data[i].shadow_z_range.y; + float z = trans_coord.z * directional_lights.data[i].shadow_z_range.y; transmittance_z = z - shadow_z; } else if (depth_z < directional_lights.data[i].shadow_split_offsets.z) { @@ -2884,8 +1642,8 @@ FRAGMENT_SHADER_CODE trans_coord /= trans_coord.w; float shadow_z = textureLod(sampler2D(directional_shadow_atlas, material_samplers[SAMPLER_LINEAR_CLAMP]), trans_coord.xy, 0.0).r; - shadow_z *= directional_lights.data[i].shadow_transmittance_z_scale.z; - float z = trans_coord.z * directional_lights.data[i].shadow_transmittance_z_scale.z; + shadow_z *= directional_lights.data[i].shadow_z_range.z; + float z = trans_coord.z * directional_lights.data[i].shadow_z_range.z; transmittance_z = z - shadow_z; @@ -2895,213 +1653,210 @@ FRAGMENT_SHADER_CODE trans_coord /= trans_coord.w; float shadow_z = textureLod(sampler2D(directional_shadow_atlas, material_samplers[SAMPLER_LINEAR_CLAMP]), trans_coord.xy, 0.0).r; - shadow_z *= directional_lights.data[i].shadow_transmittance_z_scale.w; - float z = trans_coord.z * directional_lights.data[i].shadow_transmittance_z_scale.w; + shadow_z *= directional_lights.data[i].shadow_z_range.w; + float z = trans_coord.z * directional_lights.data[i].shadow_z_range.w; transmittance_z = z - shadow_z; } + } #endif - float shadow = 1.0; + float shadow = 1.0; +#ifndef SHADOWS_DISABLED + if (i < 4) { + shadow = float(shadow0 >> (i * 8) & 0xFF) / 255.0; + } else { + shadow = float(shadow1 >> ((i - 4) * 8) & 0xFF) / 255.0; + } +#endif - if (i < 4) { - shadow = float(shadow0 >> (i * 8) & 0xFF) / 255.0; - } else { - shadow = float(shadow1 >> ((i - 4) * 8) & 0xFF) / 255.0; - } + blur_shadow(shadow); - blur_shadow(shadow); + float size_A = sc_use_light_soft_shadows ? directional_lights.data[i].size : 0.0; - light_compute(normal, directional_lights.data[i].direction, normalize(view), directional_lights.data[i].color * directional_lights.data[i].energy, shadow, f0, orms, 1.0, + light_compute(normal, directional_lights.data[i].direction, normalize(view), size_A, directional_lights.data[i].color * directional_lights.data[i].energy, shadow, f0, orms, 1.0, albedo, alpha, #ifdef LIGHT_BACKLIGHT_USED - backlight, + backlight, #endif #ifdef LIGHT_TRANSMITTANCE_USED - transmittance_color, - transmittance_depth, - transmittance_curve, - transmittance_boost, - transmittance_z, + transmittance_color, + transmittance_depth, + transmittance_boost, + transmittance_z, #endif #ifdef LIGHT_RIM_USED - rim, rim_tint, albedo, + rim, rim_tint, #endif #ifdef LIGHT_CLEARCOAT_USED - clearcoat, clearcoat_gloss, + clearcoat, clearcoat_roughness, normalize(normal_interp), #endif #ifdef LIGHT_ANISOTROPY_USED - binormal, tangent, anisotropy, -#endif -#ifdef USE_SOFT_SHADOW - directional_lights.data[i].size, + binormal, + tangent, anisotropy, #endif -#ifdef USE_SHADOW_TO_OPACITY - alpha, -#endif - diffuse_light, - specular_light); - } + diffuse_light, + specular_light); } + } - { //omni lights + { //omni lights - uint cluster_omni_offset = cluster_offset; + uint cluster_omni_offset = cluster_offset; - uint item_min; - uint item_max; - uint item_from; - uint item_to; + uint item_min; + uint item_max; + 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 + scene_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)); + item_from = subgroupBroadcastFirst(subgroupMin(item_from)); + item_to = subgroupBroadcastFirst(subgroupMax(item_to)); #endif - for (uint i = item_from; i < item_to; i++) { - uint mask = cluster_buffer.data[cluster_omni_offset + i]; - mask &= cluster_get_range_clip_mask(i, item_min, item_max); + for (uint i = item_from; i < item_to; i++) { + uint mask = cluster_buffer.data[cluster_omni_offset + i]; + mask &= cluster_get_range_clip_mask(i, item_min, item_max); #ifdef USE_SUBGROUPS - uint merged_mask = subgroupBroadcastFirst(subgroupOr(mask)); + uint merged_mask = subgroupBroadcastFirst(subgroupOr(mask)); #else uint merged_mask = mask; #endif - while (merged_mask != 0) { - uint bit = findMSB(merged_mask); - merged_mask &= ~(1 << bit); + while (merged_mask != 0) { + uint bit = findMSB(merged_mask); + merged_mask &= ~(1 << bit); #ifdef USE_SUBGROUPS - if (((1 << bit) & mask) == 0) { //do not process if not originally here - continue; - } + if (((1 << bit) & mask) == 0) { //do not process if not originally here + continue; + } #endif - uint light_index = 32 * i + bit; + uint light_index = 32 * i + bit; - if (!bool(omni_lights.data[light_index].mask & instances.data[instance_index].layer_mask)) { - continue; //not masked - } + if (!bool(omni_lights.data[light_index].mask & instances.data[instance_index].layer_mask)) { + continue; //not masked + } + + if (omni_lights.data[light_index].bake_mode == LIGHT_BAKE_STATIC && bool(instances.data[instance_index].flags & INSTANCE_FLAGS_USE_LIGHTMAP)) { + continue; // Statically baked light and object uses lightmap, skip + } - float shadow = light_process_omni_shadow(light_index, vertex, view); + float shadow = light_process_omni_shadow(light_index, vertex, normal); - shadow = blur_shadow(shadow); + shadow = blur_shadow(shadow); - light_process_omni(light_index, vertex, view, normal, vertex_ddx, vertex_ddy, f0, orms, shadow, + light_process_omni(light_index, vertex, view, normal, vertex_ddx, vertex_ddy, f0, orms, shadow, albedo, alpha, #ifdef LIGHT_BACKLIGHT_USED - backlight, + backlight, #endif #ifdef LIGHT_TRANSMITTANCE_USED - transmittance_color, - transmittance_depth, - transmittance_curve, - transmittance_boost, + transmittance_color, + transmittance_depth, + transmittance_boost, #endif #ifdef LIGHT_RIM_USED - rim, - rim_tint, - albedo, + rim, + rim_tint, #endif #ifdef LIGHT_CLEARCOAT_USED - clearcoat, clearcoat_gloss, + clearcoat, clearcoat_roughness, normalize(normal_interp), #endif #ifdef LIGHT_ANISOTROPY_USED - tangent, binormal, anisotropy, + tangent, binormal, anisotropy, #endif -#ifdef USE_SHADOW_TO_OPACITY - alpha, -#endif - diffuse_light, specular_light); - } + diffuse_light, specular_light); } } + } - { //spot lights + { //spot lights - uint cluster_spot_offset = cluster_offset + scene_data.cluster_type_size; + uint cluster_spot_offset = cluster_offset + scene_data.cluster_type_size; - uint item_min; - uint item_max; - uint item_from; - uint item_to; + 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 + scene_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)); + item_from = subgroupBroadcastFirst(subgroupMin(item_from)); + item_to = subgroupBroadcastFirst(subgroupMax(item_to)); #endif - for (uint i = item_from; i < item_to; i++) { - uint mask = cluster_buffer.data[cluster_spot_offset + i]; - mask &= cluster_get_range_clip_mask(i, item_min, item_max); + for (uint i = item_from; i < item_to; i++) { + uint mask = cluster_buffer.data[cluster_spot_offset + i]; + mask &= cluster_get_range_clip_mask(i, item_min, item_max); #ifdef USE_SUBGROUPS - uint merged_mask = subgroupBroadcastFirst(subgroupOr(mask)); + uint merged_mask = subgroupBroadcastFirst(subgroupOr(mask)); #else uint merged_mask = mask; #endif - while (merged_mask != 0) { - uint bit = findMSB(merged_mask); - merged_mask &= ~(1 << bit); + while (merged_mask != 0) { + uint bit = findMSB(merged_mask); + merged_mask &= ~(1 << bit); #ifdef USE_SUBGROUPS - if (((1 << bit) & mask) == 0) { //do not process if not originally here - continue; - } + if (((1 << bit) & mask) == 0) { //do not process if not originally here + continue; + } #endif - uint light_index = 32 * i + bit; + uint light_index = 32 * i + bit; - if (!bool(spot_lights.data[light_index].mask & instances.data[instance_index].layer_mask)) { - continue; //not masked - } + if (!bool(spot_lights.data[light_index].mask & instances.data[instance_index].layer_mask)) { + continue; //not masked + } + + if (spot_lights.data[light_index].bake_mode == LIGHT_BAKE_STATIC && bool(instances.data[instance_index].flags & INSTANCE_FLAGS_USE_LIGHTMAP)) { + continue; // Statically baked light and object uses lightmap, skip + } - float shadow = light_process_spot_shadow(light_index, vertex, view); + float shadow = light_process_spot_shadow(light_index, vertex, normal); - shadow = blur_shadow(shadow); + shadow = blur_shadow(shadow); - light_process_spot(light_index, vertex, view, normal, vertex_ddx, vertex_ddy, f0, orms, shadow, + light_process_spot(light_index, vertex, view, normal, vertex_ddx, vertex_ddy, f0, orms, shadow, albedo, alpha, #ifdef LIGHT_BACKLIGHT_USED - backlight, + backlight, #endif #ifdef LIGHT_TRANSMITTANCE_USED - transmittance_color, - transmittance_depth, - transmittance_curve, - transmittance_boost, + transmittance_color, + transmittance_depth, + transmittance_boost, #endif #ifdef LIGHT_RIM_USED - rim, - rim_tint, - albedo, + rim, + rim_tint, #endif #ifdef LIGHT_CLEARCOAT_USED - clearcoat, clearcoat_gloss, + clearcoat, clearcoat_roughness, normalize(normal_interp), #endif #ifdef LIGHT_ANISOTROPY_USED - tangent, binormal, anisotropy, + tangent, + binormal, anisotropy, #endif -#ifdef USE_SHADOW_TO_OPACITY - alpha, -#endif - diffuse_light, specular_light); - } + diffuse_light, specular_light); } } + } #ifdef USE_SHADOW_TO_OPACITY - alpha = min(alpha, clamp(length(ambient_light), 0.0, 1.0)); + alpha = min(alpha, clamp(length(ambient_light), 0.0, 1.0)); #if defined(ALPHA_SCISSOR_USED) - if (alpha < alpha_scissor) { - discard; - } + if (alpha < alpha_scissor) { + discard; + } #endif // ALPHA_SCISSOR_USED #ifdef USE_OPAQUE_PREPASS - if (alpha < opaque_prepass_threshold) { - discard; - } + if (alpha < scene_data.opaque_prepass_threshold) { + discard; + } #endif // USE_OPAQUE_PREPASS @@ -3113,126 +1868,130 @@ FRAGMENT_SHADER_CODE #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)); - - uint albedo16 = 0x1; //solid flag - albedo16 |= clamp(uint(albedo.r * 31.0), 0, 31) << 11; - albedo16 |= clamp(uint(albedo.g * 31.0), 0, 31) << 6; - albedo16 |= clamp(uint(albedo.b * 31.0), 0, 31) << 1; - - imageStore(albedo_volume_grid, grid_pos, uvec4(albedo16)); - - uint facing_bits = 0; - const vec3 aniso_dir[6] = vec3[]( - vec3(1, 0, 0), - vec3(0, 1, 0), - vec3(0, 0, 1), - vec3(-1, 0, 0), - vec3(0, -1, 0), - vec3(0, 0, -1)); - - vec3 cam_normal = mat3(scene_data.camera_matrix) * normalize(normal_interp); - - float closest_dist = -1e20; - - for (uint i = 0; i < 6; i++) { - float d = dot(cam_normal, aniso_dir[i]); - if (d > closest_dist) { - closest_dist = d; - facing_bits = (1 << i); - } + { + 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)); + + uint albedo16 = 0x1; //solid flag + albedo16 |= clamp(uint(albedo.r * 31.0), 0, 31) << 11; + albedo16 |= clamp(uint(albedo.g * 31.0), 0, 31) << 6; + albedo16 |= clamp(uint(albedo.b * 31.0), 0, 31) << 1; + + imageStore(albedo_volume_grid, grid_pos, uvec4(albedo16)); + + uint facing_bits = 0; + const vec3 aniso_dir[6] = vec3[]( + vec3(1, 0, 0), + vec3(0, 1, 0), + vec3(0, 0, 1), + vec3(-1, 0, 0), + vec3(0, -1, 0), + vec3(0, 0, -1)); + + vec3 cam_normal = mat3(scene_data.inv_view_matrix) * normalize(normal_interp); + + float closest_dist = -1e20; + + for (uint i = 0; i < 6; i++) { + float d = dot(cam_normal, aniso_dir[i]); + if (d > closest_dist) { + closest_dist = d; + facing_bits = (1 << i); } + } - imageAtomicOr(geom_facing_grid, grid_pos, facing_bits); //store facing bits +#ifdef MOLTENVK_USED + imageStore(geom_facing_grid, grid_pos, uvec4(imageLoad(geom_facing_grid, grid_pos).r | facing_bits)); //store facing bits +#else + imageAtomicOr(geom_facing_grid, grid_pos, facing_bits); //store facing bits +#endif - if (length(emission) > 0.001) { - float lumas[6]; - vec3 light_total = vec3(0); + if (length(emission) > 0.001) { + float lumas[6]; + vec3 light_total = vec3(0); - for (int i = 0; i < 6; i++) { - float strength = max(0.0, dot(cam_normal, aniso_dir[i])); - vec3 light = emission * strength; - light_total += light; - lumas[i] = max(light.r, max(light.g, light.b)); - } + for (int i = 0; i < 6; i++) { + float strength = max(0.0, dot(cam_normal, aniso_dir[i])); + vec3 light = emission * strength; + light_total += light; + lumas[i] = max(light.r, max(light.g, light.b)); + } - float luma_total = max(light_total.r, max(light_total.g, light_total.b)); + float luma_total = max(light_total.r, max(light_total.g, light_total.b)); - uint light_aniso = 0; + uint light_aniso = 0; - for (int i = 0; i < 6; i++) { - light_aniso |= min(31, uint((lumas[i] / luma_total) * 31.0)) << (i * 5); - } + for (int i = 0; i < 6; i++) { + light_aniso |= min(31, uint((lumas[i] / luma_total) * 31.0)) << (i * 5); + } - //compress to RGBE9995 to save space + //compress to RGBE9995 to save space - const float pow2to9 = 512.0f; - const float B = 15.0f; - const float N = 9.0f; - const float LN2 = 0.6931471805599453094172321215; + const float pow2to9 = 512.0f; + const float B = 15.0f; + const float N = 9.0f; + const float LN2 = 0.6931471805599453094172321215; - float cRed = clamp(light_total.r, 0.0, 65408.0); - float cGreen = clamp(light_total.g, 0.0, 65408.0); - float cBlue = clamp(light_total.b, 0.0, 65408.0); + float cRed = clamp(light_total.r, 0.0, 65408.0); + float cGreen = clamp(light_total.g, 0.0, 65408.0); + float cBlue = clamp(light_total.b, 0.0, 65408.0); - float cMax = max(cRed, max(cGreen, cBlue)); + float cMax = max(cRed, max(cGreen, cBlue)); - float expp = max(-B - 1.0f, floor(log(cMax) / LN2)) + 1.0f + B; + float expp = max(-B - 1.0f, floor(log(cMax) / LN2)) + 1.0f + B; - float sMax = floor((cMax / pow(2.0f, expp - B - N)) + 0.5f); + float sMax = floor((cMax / pow(2.0f, expp - B - N)) + 0.5f); - float exps = expp + 1.0f; + float exps = expp + 1.0f; - if (0.0 <= sMax && sMax < pow2to9) { - exps = expp; - } + if (0.0 <= sMax && sMax < pow2to9) { + exps = expp; + } - float sRed = floor((cRed / pow(2.0f, exps - B - N)) + 0.5f); - 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); + float sRed = floor((cRed / pow(2.0f, exps - B - N)) + 0.5f); + 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); - imageStore(emission_grid, grid_pos, uvec4(light_rgbe)); - imageStore(emission_aniso_grid, grid_pos, uvec4(light_aniso)); - } + imageStore(emission_grid, grid_pos, uvec4(light_rgbe)); + imageStore(emission_aniso_grid, grid_pos, uvec4(light_aniso)); } + } #endif #ifdef MODE_RENDER_MATERIAL - albedo_output_buffer.rgb = albedo; - albedo_output_buffer.a = alpha; + albedo_output_buffer.rgb = albedo; + albedo_output_buffer.a = alpha; - normal_output_buffer.rgb = normal * 0.5 + 0.5; - normal_output_buffer.a = 0.0; - depth_output_buffer.r = -vertex.z; + normal_output_buffer.rgb = normal * 0.5 + 0.5; + normal_output_buffer.a = 0.0; + depth_output_buffer.r = -vertex.z; - orm_output_buffer.r = ao; - orm_output_buffer.g = roughness; - orm_output_buffer.b = metallic; - orm_output_buffer.a = sss_strength; + orm_output_buffer.r = ao; + orm_output_buffer.g = roughness; + orm_output_buffer.b = metallic; + orm_output_buffer.a = sss_strength; - emission_output_buffer.rgb = emission; - emission_output_buffer.a = 0.0; + emission_output_buffer.rgb = emission; + emission_output_buffer.a = 0.0; #endif #ifdef MODE_RENDER_NORMAL_ROUGHNESS - normal_roughness_output_buffer = vec4(normal * 0.5 + 0.5, roughness); - -#ifdef MODE_RENDER_GIPROBE - if (bool(instances.data[instance_index].flags & INSTANCE_FLAGS_USE_GIPROBE)) { // process giprobes - uint index1 = instances.data[instance_index].gi_offset & 0xFFFF; - uint index2 = instances.data[instance_index].gi_offset >> 16; - giprobe_buffer.x = index1 & 0xFF; - giprobe_buffer.y = index2 & 0xFF; - } else { - giprobe_buffer.x = 0xFF; - giprobe_buffer.y = 0xFF; - } + normal_roughness_output_buffer = vec4(normal * 0.5 + 0.5, roughness); + +#ifdef MODE_RENDER_VOXEL_GI + 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; + } else { + voxel_gi_buffer.x = 0xFF; + voxel_gi_buffer.y = 0xFF; + } #endif #endif //MODE_RENDER_NORMAL_ROUGHNESS @@ -3256,7 +2015,7 @@ FRAGMENT_SHADER_CODE //restore fog fog = vec4(unpackHalf2x16(fog_rg), unpackHalf2x16(fog_ba)); -#ifdef MODE_MULTIPLE_RENDER_TARGETS +#ifdef MODE_SEPARATE_SPECULAR #ifdef MODE_UNSHADED diffuse_buffer = vec4(albedo.rgb, 0.0); @@ -3274,20 +2033,38 @@ FRAGMENT_SHADER_CODE diffuse_buffer.rgb = mix(diffuse_buffer.rgb, fog.rgb, fog.a); specular_buffer.rgb = mix(specular_buffer.rgb, vec3(0.0), fog.a); -#else //MODE_MULTIPLE_RENDER_TARGETS +#else //MODE_SEPARATE_SPECULAR #ifdef MODE_UNSHADED frag_color = vec4(albedo, alpha); #else frag_color = vec4(emission + ambient_light + diffuse_light + specular_light, alpha); - //frag_color = vec4(1.0); +//frag_color = vec4(1.0); #endif //USE_NO_SHADING // Draw "fixed" fog before volumetric fog to ensure volumetric fog can appear in front of the sky. frag_color.rgb = mix(frag_color.rgb, fog.rgb, fog.a); - ; -#endif //MODE_MULTIPLE_RENDER_TARGETS +#endif //MODE_SEPARATE_SPECULAR #endif //MODE_RENDER_DEPTH - } +#ifdef MOTION_VECTORS + vec2 position_clip = (screen_position.xy / screen_position.w) - scene_data.taa_jitter; + vec2 prev_position_clip = (prev_screen_position.xy / prev_screen_position.w) - scene_data_block.prev_data.taa_jitter; + + vec2 position_uv = position_clip * vec2(0.5, 0.5); + vec2 prev_position_uv = prev_position_clip * vec2(0.5, 0.5); + + motion_vector = position_uv - prev_position_uv; +#endif +} + +void main() { +#ifdef MODE_DUAL_PARABOLOID + + if (dp_clip > 0.0) + discard; +#endif + + fragment_shader(scene_data_block.data); +} diff --git a/servers/rendering/renderer_rd/shaders/scene_forward_clustered_inc.glsl b/servers/rendering/renderer_rd/shaders/scene_forward_clustered_inc.glsl index d78890fa9e..32ea83397a 100644 --- a/servers/rendering/renderer_rd/shaders/scene_forward_clustered_inc.glsl +++ b/servers/rendering/renderer_rd/shaders/scene_forward_clustered_inc.glsl @@ -1,9 +1,10 @@ #define M_PI 3.14159265359 #define ROUGHNESS_MAX_LOD 5 -#define MAX_GI_PROBES 8 +#define MAX_VOXEL_GI_INSTANCES 8 +#define MAX_VIEWS 2 -#if defined(GL_KHR_shader_subgroup_ballot) && defined(GL_KHR_shader_subgroup_arithmetic) +#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 @@ -12,15 +13,20 @@ #endif +#if defined(USE_MULTIVIEW) && defined(has_VK_KHR_multiview) +#extension GL_EXT_multiview : enable +#endif + #include "cluster_data_inc.glsl" +#include "decal_data_inc.glsl" -#if !defined(MODE_RENDER_DEPTH) || defined(MODE_RENDER_MATERIAL) || defined(MODE_RENDER_SDF) || defined(MODE_RENDER_NORMAL_ROUGHNESS) || defined(MODE_RENDER_GIPROBE) || defined(TANGENT_USED) || defined(NORMAL_MAP_USED) +#if !defined(MODE_RENDER_DEPTH) || defined(MODE_RENDER_MATERIAL) || defined(MODE_RENDER_SDF) || defined(MODE_RENDER_NORMAL_ROUGHNESS) || defined(MODE_RENDER_VOXEL_GI) || defined(TANGENT_USED) || defined(NORMAL_MAP_USED) || defined(LIGHT_ANISOTROPY_USED) #ifndef NORMAL_USED #define NORMAL_USED #endif #endif -layout(push_constant, binding = 0, std430) uniform DrawCall { +layout(push_constant, std430) uniform DrawCall { uint instance_index; uint uv_offset; uint pad0; @@ -28,7 +34,11 @@ layout(push_constant, binding = 0, std430) uniform DrawCall { } draw_call; -/* Set 0 Scene data that never changes, ever */ +#define SDFGI_MAX_CASCADES 8 + +/* Set 0: Base Pass (never changes) */ + +#include "light_data_inc.glsl" #define SAMPLER_NEAREST_CLAMP 0 #define SAMPLER_LINEAR_CLAMP 1 @@ -43,47 +53,49 @@ draw_call; #define SAMPLER_NEAREST_WITH_MIPMAPS_ANISOTROPIC_REPEAT 10 #define SAMPLER_LINEAR_WITH_MIPMAPS_ANISOTROPIC_REPEAT 11 -#define SDFGI_MAX_CASCADES 8 - -/* Set 1: Base Pass (never changes) */ - layout(set = 0, binding = 1) uniform sampler material_samplers[12]; 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_GIPROBE (1 << 11) +#define INSTANCE_FLAGS_USE_VOXEL_GI (1 << 11) #define INSTANCE_FLAGS_MULTIMESH (1 << 12) #define INSTANCE_FLAGS_MULTIMESH_FORMAT_2D (1 << 13) #define INSTANCE_FLAGS_MULTIMESH_HAS_COLOR (1 << 14) #define INSTANCE_FLAGS_MULTIMESH_HAS_CUSTOM_DATA (1 << 15) -#define INSTANCE_FLAGS_MULTIMESH_STRIDE_SHIFT 16 +#define INSTANCE_FLAGS_PARTICLE_TRAIL_SHIFT 16 +#define INSTANCE_FLAGS_FADE_SHIFT 24 //3 bits of stride -#define INSTANCE_FLAGS_MULTIMESH_STRIDE_MASK 0x7 +#define INSTANCE_FLAGS_PARTICLE_TRAIL_MASK 0xFF -#define INSTANCE_FLAGS_SKELETON (1 << 19) -#define INSTANCE_FLAGS_NON_UNIFORM_SCALE (1 << 20) +#define SCREEN_SPACE_EFFECTS_FLAGS_USE_SSAO 1 +#define SCREEN_SPACE_EFFECTS_FLAGS_USE_SSIL 2 -layout(set = 0, binding = 3, std430) restrict readonly buffer OmniLights { +layout(set = 0, binding = 5, std430) restrict readonly buffer OmniLights { LightData data[]; } omni_lights; -layout(set = 0, binding = 4, std430) restrict readonly buffer SpotLights { +layout(set = 0, binding = 6, std430) restrict readonly buffer SpotLights { LightData data[]; } spot_lights; -layout(set = 0, binding = 5) buffer restrict readonly ReflectionProbeData { +layout(set = 0, binding = 7, std430) restrict readonly buffer ReflectionProbeData { ReflectionData data[]; } reflections; -layout(set = 0, binding = 6, std140) uniform DirectionalLights { +layout(set = 0, binding = 8, std140) uniform DirectionalLights { DirectionalLightData data[MAX_DIRECTIONAL_LIGHT_DATA_STRUCTS]; } directional_lights; @@ -95,7 +107,7 @@ struct Lightmap { mat3 normal_xform; }; -layout(set = 0, binding = 7, std140) restrict readonly buffer Lightmaps { +layout(set = 0, binding = 9, std140) restrict readonly buffer Lightmaps { Lightmap data[]; } lightmaps; @@ -104,34 +116,32 @@ struct LightmapCapture { vec4 sh[9]; }; -layout(set = 0, binding = 8, std140) restrict readonly buffer LightmapCaptures { +layout(set = 0, binding = 10, std140) restrict readonly buffer LightmapCaptures { LightmapCapture data[]; } lightmap_captures; -layout(set = 0, binding = 9) uniform texture2D decal_atlas; -layout(set = 0, binding = 10) uniform texture2D decal_atlas_srgb; +layout(set = 0, binding = 11) uniform texture2D decal_atlas; +layout(set = 0, binding = 12) uniform texture2D decal_atlas_srgb; -layout(set = 0, binding = 11, std430) restrict readonly buffer Decals { +layout(set = 0, binding = 13, std430) restrict readonly buffer Decals { DecalData data[]; } decals; -layout(set = 0, binding = 12, std430) restrict readonly buffer GlobalVariableData { +layout(set = 0, binding = 14, std430) restrict readonly buffer GlobalVariableData { vec4 data[]; } global_variables; -#ifndef LOW_END_MODE - -struct SDFGIProbeCascadeData { +struct SDFVoxelGICascadeData { vec3 position; float to_probe; ivec3 probe_world_offset; float to_cell; // 1/bounds * grid_size }; -layout(set = 0, binding = 13, std140) uniform SDFGI { +layout(set = 0, binding = 15, std140) uniform SDFGI { vec3 grid_size; uint max_cascades; @@ -155,20 +165,22 @@ layout(set = 0, binding = 13, std140) uniform SDFGI { vec3 cascade_probe_size; uint pad5; - SDFGIProbeCascadeData cascades[SDFGI_MAX_CASCADES]; + SDFVoxelGICascadeData cascades[SDFGI_MAX_CASCADES]; } sdfgi; -#endif //LOW_END_MODE - -/* Set 2: Render Pass (changes per render pass) */ +/* Set 1: Render Pass (changes per render pass) */ -layout(set = 1, binding = 0, std140) uniform SceneData { +struct SceneData { mat4 projection_matrix; mat4 inv_projection_matrix; + mat4 inv_view_matrix; + mat4 view_matrix; - mat4 camera_matrix; - mat4 inv_camera_matrix; + // only used for multiview + mat4 projection_matrix_view[MAX_VIEWS]; + mat4 inv_projection_matrix_view[MAX_VIEWS]; + vec4 eye_offset[MAX_VIEWS]; vec2 viewport_size; vec2 screen_pixel_size; @@ -178,17 +190,12 @@ layout(set = 1, binding = 0, std140) uniform SceneData { uint cluster_type_size; uint max_cluster_element_count_div_32; - //use vec4s because std140 doesnt play nice with vec2s, z and w are wasted + // 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]; - uint directional_penumbra_shadow_samples; - uint directional_soft_shadow_samples; - uint penumbra_shadow_samples; - uint soft_shadow_samples; - vec4 ambient_light_color_energy; float ambient_color_sky_mix; @@ -206,16 +213,15 @@ layout(set = 1, binding = 0, std140) uniform SceneData { float z_far; float z_near; - bool ssao_enabled; + uint ss_effects_flags; float ssao_light_affect; float ssao_ao_affect; bool roughness_limiter_enabled; float roughness_limiter_amount; float roughness_limiter_limit; - uvec2 roughness_limiter_pad; - - vec4 ao_color; + float opaque_prepass_threshold; + uint roughness_limiter_pad; mat4 sdf_to_bounds; @@ -244,12 +250,19 @@ layout(set = 1, binding = 0, std140) uniform SceneData { float reflection_multiplier; // one normally, zero when rendering reflections bool pancake_shadows; -} + vec2 taa_jitter; + uvec2 pad2; +}; -scene_data; +layout(set = 1, binding = 0, std140) uniform SceneDataBlock { + SceneData data; + SceneData prev_data; +} +scene_data_block; struct InstanceData { mat4 transform; + mat4 prev_transform; uint flags; uint instance_uniforms_ofs; //base offset in global buffer for instance variables uint gi_offset; //GI information when using lightmapping (VCT or lightmap index) @@ -280,9 +293,7 @@ layout(set = 1, binding = 5) uniform texture2D directional_shadow_atlas; layout(set = 1, binding = 6) uniform texture2DArray lightmap_textures[MAX_LIGHTMAP_TEXTURES]; -#ifndef LOW_END_MOD -layout(set = 1, binding = 7) uniform texture3D gi_probe_textures[MAX_GI_PROBES]; -#endif +layout(set = 1, binding = 7) uniform texture3D voxel_gi_textures[MAX_VOXEL_GI_INSTANCES]; layout(set = 1, binding = 8, std430) buffer restrict readonly ClusterBuffer { uint data[]; @@ -306,8 +317,6 @@ layout(r32ui, set = 1, binding = 12) uniform restrict uimage3D geom_facing_grid; layout(set = 1, binding = 9) uniform texture2D depth_buffer; layout(set = 1, binding = 10) uniform texture2D color_buffer; -#ifndef LOW_END_MODE - layout(set = 1, binding = 11) uniform texture2D normal_roughness_buffer; layout(set = 1, binding = 12) uniform texture2D ao_buffer; layout(set = 1, binding = 13) uniform texture2D ambient_buffer; @@ -315,30 +324,26 @@ layout(set = 1, binding = 14) uniform texture2D reflection_buffer; layout(set = 1, binding = 15) uniform texture2DArray sdfgi_lightprobe_texture; layout(set = 1, binding = 16) uniform texture3D sdfgi_occlusion_cascades; -struct GIProbeData { - mat4 xform; - vec3 bounds; - float dynamic_range; +struct VoxelGIData { + mat4 xform; // 64 - 64 - float bias; - float normal_bias; - bool blend_ambient; - uint texture_slot; + vec3 bounds; // 12 - 76 + float dynamic_range; // 4 - 80 - float anisotropy_strength; - float ambient_occlusion; - float ambient_occlusion_size; - uint mipmaps; + float bias; // 4 - 84 + float normal_bias; // 4 - 88 + bool blend_ambient; // 4 - 92 + uint mipmaps; // 4 - 96 }; -layout(set = 1, binding = 17, std140) uniform GIProbes { - GIProbeData data[MAX_GI_PROBES]; +layout(set = 1, binding = 17, std140) uniform VoxelGIs { + VoxelGIData data[MAX_VOXEL_GI_INSTANCES]; } -gi_probes; +voxel_gi_instances; layout(set = 1, binding = 18) uniform texture3D volumetric_fog_texture; -#endif // LOW_END_MODE +layout(set = 1, binding = 19) uniform texture2D ssil_buffer; #endif diff --git a/servers/rendering/renderer_rd/shaders/scene_forward_gi_inc.glsl b/servers/rendering/renderer_rd/shaders/scene_forward_gi_inc.glsl new file mode 100644 index 0000000000..c88bd0a14b --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/scene_forward_gi_inc.glsl @@ -0,0 +1,242 @@ +// Functions related to gi/sdfgi for our forward renderer + +//standard voxel cone trace +vec4 voxel_cone_trace(texture3D probe, vec3 cell_size, vec3 pos, vec3 direction, float tan_half_angle, float max_distance, float p_bias) { + float dist = p_bias; + vec4 color = vec4(0.0); + + while (dist < max_distance && color.a < 0.95) { + float diameter = max(1.0, 2.0 * tan_half_angle * dist); + vec3 uvw_pos = (pos + dist * direction) * cell_size; + float half_diameter = diameter * 0.5; + //check if outside, then break + if (any(greaterThan(abs(uvw_pos - 0.5), vec3(0.5f + half_diameter * cell_size)))) { + break; + } + vec4 scolor = textureLod(sampler3D(probe, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), uvw_pos, log2(diameter)); + float a = (1.0 - color.a); + color += a * scolor; + dist += half_diameter; + } + + return color; +} + +vec4 voxel_cone_trace_45_degrees(texture3D probe, vec3 cell_size, vec3 pos, vec3 direction, float tan_half_angle, float max_distance, float p_bias) { + float dist = p_bias; + vec4 color = vec4(0.0); + float radius = max(0.5, tan_half_angle * dist); + float lod_level = log2(radius * 2.0); + + while (dist < max_distance && color.a < 0.95) { + vec3 uvw_pos = (pos + dist * direction) * cell_size; + + //check if outside, then break + if (any(greaterThan(abs(uvw_pos - 0.5), vec3(0.5f + radius * cell_size)))) { + break; + } + vec4 scolor = textureLod(sampler3D(probe, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), uvw_pos, lod_level); + lod_level += 1.0; + + float a = (1.0 - color.a); + scolor *= a; + color += scolor; + dist += radius; + radius = max(0.5, tan_half_angle * dist); + } + + return color; +} + +void voxel_gi_compute(uint index, vec3 position, vec3 normal, vec3 ref_vec, mat3 normal_xform, float roughness, vec3 ambient, vec3 environment, inout vec4 out_spec, inout vec4 out_diff) { + position = (voxel_gi_instances.data[index].xform * vec4(position, 1.0)).xyz; + ref_vec = normalize((voxel_gi_instances.data[index].xform * vec4(ref_vec, 0.0)).xyz); + normal = normalize((voxel_gi_instances.data[index].xform * vec4(normal, 0.0)).xyz); + + position += normal * voxel_gi_instances.data[index].normal_bias; + + //this causes corrupted pixels, i have no idea why.. + if (any(bvec2(any(lessThan(position, vec3(0.0))), any(greaterThan(position, voxel_gi_instances.data[index].bounds))))) { + return; + } + + vec3 blendv = abs(position / voxel_gi_instances.data[index].bounds * 2.0 - 1.0); + float blend = clamp(1.0 - max(blendv.x, max(blendv.y, blendv.z)), 0.0, 1.0); + //float blend=1.0; + + float max_distance = length(voxel_gi_instances.data[index].bounds); + vec3 cell_size = 1.0 / voxel_gi_instances.data[index].bounds; + + //radiance + +#define MAX_CONE_DIRS 4 + + vec3 cone_dirs[MAX_CONE_DIRS] = vec3[]( + vec3(0.707107, 0.0, 0.707107), + vec3(0.0, 0.707107, 0.707107), + vec3(-0.707107, 0.0, 0.707107), + vec3(0.0, -0.707107, 0.707107)); + + float cone_weights[MAX_CONE_DIRS] = float[](0.25, 0.25, 0.25, 0.25); + float cone_angle_tan = 0.98269; + + vec3 light = vec3(0.0); + + for (int i = 0; i < MAX_CONE_DIRS; i++) { + vec3 dir = normalize((voxel_gi_instances.data[index].xform * vec4(normal_xform * cone_dirs[i], 0.0)).xyz); + + vec4 cone_light = voxel_cone_trace_45_degrees(voxel_gi_textures[index], cell_size, position, dir, cone_angle_tan, max_distance, voxel_gi_instances.data[index].bias); + + if (voxel_gi_instances.data[index].blend_ambient) { + cone_light.rgb = mix(ambient, cone_light.rgb, min(1.0, cone_light.a / 0.95)); + } + + light += cone_weights[i] * cone_light.rgb; + } + + light *= voxel_gi_instances.data[index].dynamic_range; + out_diff += vec4(light * blend, blend); + + //irradiance + 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); + 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=vec3(0.0); + + out_spec += vec4(irr_light.rgb * blend, blend); +} + +vec2 octahedron_wrap(vec2 v) { + vec2 signVal; + signVal.x = v.x >= 0.0 ? 1.0 : -1.0; + signVal.y = v.y >= 0.0 ? 1.0 : -1.0; + return (1.0 - abs(v.yx)) * signVal; +} + +vec2 octahedron_encode(vec3 n) { + // https://twitter.com/Stubbesaurus/status/937994790553227264 + n /= (abs(n.x) + abs(n.y) + abs(n.z)); + n.xy = n.z >= 0.0 ? n.xy : octahedron_wrap(n.xy); + n.xy = n.xy * 0.5 + 0.5; + return n.xy; +} + +void sdfgi_process(uint cascade, vec3 cascade_pos, vec3 cam_pos, vec3 cam_normal, vec3 cam_specular_normal, bool use_specular, float roughness, out vec3 diffuse_light, out vec3 specular_light, out float blend) { + cascade_pos += cam_normal * sdfgi.normal_bias; + + vec3 base_pos = floor(cascade_pos); + //cascade_pos += mix(vec3(0.0),vec3(0.01),lessThan(abs(cascade_pos-base_pos),vec3(0.01))) * cam_normal; + ivec3 probe_base_pos = ivec3(base_pos); + + vec4 diffuse_accum = vec4(0.0); + vec3 specular_accum; + + ivec3 tex_pos = ivec3(probe_base_pos.xy, int(cascade)); + tex_pos.x += probe_base_pos.z * sdfgi.probe_axis_size; + tex_pos.xy = tex_pos.xy * (SDFGI_OCT_SIZE + 2) + ivec2(1); + + vec3 diffuse_posf = (vec3(tex_pos) + vec3(octahedron_encode(cam_normal) * float(SDFGI_OCT_SIZE), 0.0)) * sdfgi.lightprobe_tex_pixel_size; + + vec3 specular_posf; + + if (use_specular) { + specular_accum = vec3(0.0); + specular_posf = (vec3(tex_pos) + vec3(octahedron_encode(cam_specular_normal) * float(SDFGI_OCT_SIZE), 0.0)) * sdfgi.lightprobe_tex_pixel_size; + } + + vec4 light_accum = vec4(0.0); + float weight_accum = 0.0; + + for (uint j = 0; j < 8; j++) { + ivec3 offset = (ivec3(j) >> ivec3(0, 1, 2)) & ivec3(1, 1, 1); + ivec3 probe_posi = probe_base_pos; + probe_posi += offset; + + // Compute weight + + vec3 probe_pos = vec3(probe_posi); + vec3 probe_to_pos = cascade_pos - probe_pos; + vec3 probe_dir = normalize(-probe_to_pos); + + vec3 trilinear = vec3(1.0) - abs(probe_to_pos); + float weight = trilinear.x * trilinear.y * trilinear.z * max(0.005, dot(cam_normal, probe_dir)); + + // Compute lightprobe occlusion + + if (sdfgi.use_occlusion) { + ivec3 occ_indexv = abs((sdfgi.cascades[cascade].probe_world_offset + probe_posi) & ivec3(1, 1, 1)) * ivec3(1, 2, 4); + vec4 occ_mask = mix(vec4(0.0), vec4(1.0), equal(ivec4(occ_indexv.x | occ_indexv.y), ivec4(0, 1, 2, 3))); + + vec3 occ_pos = clamp(cascade_pos, probe_pos - sdfgi.occlusion_clamp, probe_pos + sdfgi.occlusion_clamp) * sdfgi.probe_to_uvw; + occ_pos.z += float(cascade); + if (occ_indexv.z != 0) { //z bit is on, means index is >=4, so make it switch to the other half of textures + occ_pos.x += 1.0; + } + + occ_pos *= sdfgi.occlusion_renormalize; + float occlusion = dot(textureLod(sampler3D(sdfgi_occlusion_cascades, material_samplers[SAMPLER_LINEAR_CLAMP]), occ_pos, 0.0), occ_mask); + + weight *= max(occlusion, 0.01); + } + + // Compute lightprobe texture position + + vec3 diffuse; + vec3 pos_uvw = diffuse_posf; + pos_uvw.xy += vec2(offset.xy) * sdfgi.lightprobe_uv_offset.xy; + 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); + + if (use_specular) { + vec3 specular = vec3(0.0); + vec3 pos_uvw = specular_posf; + pos_uvw.xy += vec2(offset.xy) * sdfgi.lightprobe_uv_offset.xy; + pos_uvw.x += float(offset.z) * sdfgi.lightprobe_uv_offset.z; + if (roughness < 0.99) { + specular = textureLod(sampler2DArray(sdfgi_lightprobe_texture, material_samplers[SAMPLER_LINEAR_CLAMP]), pos_uvw + vec3(0, 0, float(sdfgi.max_cascades)), 0.0).rgb; + } + if (roughness > 0.5) { + 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; + } + } + + if (diffuse_accum.a > 0.0) { + diffuse_accum.rgb /= diffuse_accum.a; + } + + diffuse_light = diffuse_accum.rgb; + + if (use_specular) { + if (diffuse_accum.a > 0.0) { + specular_accum /= diffuse_accum.a; + } + + specular_light = specular_accum; + } + + { + //process blend + float blend_from = (float(sdfgi.probe_axis_size - 1) / 2.0) - 2.5; + float blend_to = blend_from + 2.0; + + vec3 inner_pos = cam_pos * sdfgi.cascades[cascade].to_probe; + + float len = length(inner_pos); + + inner_pos = abs(normalize(inner_pos)); + len *= max(inner_pos.x, max(inner_pos.y, inner_pos.z)); + + if (len >= blend_from) { + blend = smoothstep(blend_from, blend_to, len); + } else { + blend = 0.0; + } + } +} diff --git a/servers/rendering/renderer_rd/shaders/scene_forward_lights_inc.glsl b/servers/rendering/renderer_rd/shaders/scene_forward_lights_inc.glsl new file mode 100644 index 0000000000..c92b29b14a --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/scene_forward_lights_inc.glsl @@ -0,0 +1,971 @@ +// Functions related to lighting + +float D_GGX(float cos_theta_m, float alpha) { + float a = cos_theta_m * alpha; + float k = alpha / (1.0 - cos_theta_m * cos_theta_m + a * a); + return k * k * (1.0 / M_PI); +} + +// From Earl Hammon, Jr. "PBR Diffuse Lighting for GGX+Smith Microsurfaces" https://www.gdcvault.com/play/1024478/PBR-Diffuse-Lighting-for-GGX +float V_GGX(float NdotL, float NdotV, float alpha) { + return 0.5 / mix(2.0 * NdotL * NdotV, NdotL + NdotV, alpha); +} + +float D_GGX_anisotropic(float cos_theta_m, float alpha_x, float alpha_y, float cos_phi, float sin_phi) { + float alpha2 = alpha_x * alpha_y; + highp vec3 v = vec3(alpha_y * cos_phi, alpha_x * sin_phi, alpha2 * cos_theta_m); + highp float v2 = dot(v, v); + float w2 = alpha2 / v2; + float D = alpha2 * w2 * w2 * (1.0 / M_PI); + return D; +} + +float V_GGX_anisotropic(float alpha_x, float alpha_y, float TdotV, float TdotL, float BdotV, float BdotL, float NdotV, float NdotL) { + float Lambda_V = NdotL * length(vec3(alpha_x * TdotV, alpha_y * BdotV, NdotV)); + float Lambda_L = NdotV * length(vec3(alpha_x * TdotL, alpha_y * BdotL, NdotL)); + return 0.5 / (Lambda_V + Lambda_L); +} + +float SchlickFresnel(float u) { + float m = 1.0 - u; + float m2 = m * m; + return m2 * m2 * m; // pow(m,5) +} + +vec3 F0(float metallic, float specular, vec3 albedo) { + float dielectric = 0.16 * specular * specular; + // use albedo * metallic as colored specular reflectance at 0 angle for metallic materials; + // see https://google.github.io/filament/Filament.md.html + return mix(vec3(dielectric), albedo, vec3(metallic)); +} + +void light_compute(vec3 N, vec3 L, vec3 V, float A, vec3 light_color, float attenuation, vec3 f0, uint orms, float specular_amount, vec3 albedo, inout float alpha, +#ifdef LIGHT_BACKLIGHT_USED + vec3 backlight, +#endif +#ifdef LIGHT_TRANSMITTANCE_USED + vec4 transmittance_color, + float transmittance_depth, + float transmittance_boost, + float transmittance_z, +#endif +#ifdef LIGHT_RIM_USED + float rim, float rim_tint, +#endif +#ifdef LIGHT_CLEARCOAT_USED + float clearcoat, float clearcoat_roughness, vec3 vertex_normal, +#endif +#ifdef LIGHT_ANISOTROPY_USED + vec3 B, vec3 T, float anisotropy, +#endif + inout vec3 diffuse_light, inout vec3 specular_light) { + + vec4 orms_unpacked = unpackUnorm4x8(orms); + + float roughness = orms_unpacked.y; + float metallic = orms_unpacked.z; + +#if defined(LIGHT_CODE_USED) + // light is written by the light shader + + vec3 normal = N; + vec3 light = L; + vec3 view = V; + +#CODE : LIGHT + +#else + + float NdotL = min(A + dot(N, L), 1.0); + float cNdotL = max(NdotL, 0.0); // clamped NdotL + float NdotV = dot(N, V); + float cNdotV = max(NdotV, 1e-4); + +#if defined(DIFFUSE_BURLEY) || defined(SPECULAR_SCHLICK_GGX) || defined(LIGHT_CLEARCOAT_USED) + vec3 H = normalize(V + L); +#endif + +#if defined(SPECULAR_SCHLICK_GGX) + float cNdotH = clamp(A + dot(N, H), 0.0, 1.0); +#endif + +#if defined(DIFFUSE_BURLEY) || defined(SPECULAR_SCHLICK_GGX) || defined(LIGHT_CLEARCOAT_USED) + float cLdotH = clamp(A + dot(L, H), 0.0, 1.0); +#endif + + if (metallic < 1.0) { + 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))); +#elif defined(DIFFUSE_TOON) + + diffuse_brdf_NL = smoothstep(-roughness, max(roughness, 0.01), NdotL); + +#elif defined(DIFFUSE_BURLEY) + + { + float FD90_minus_1 = 2.0 * cLdotH * cLdotH * roughness - 0.5; + float FdV = 1.0 + FD90_minus_1 * SchlickFresnel(cNdotV); + float FdL = 1.0 + FD90_minus_1 * SchlickFresnel(cNdotL); + diffuse_brdf_NL = (1.0 / M_PI) * FdV * FdL * cNdotL; + /* + float energyBias = mix(roughness, 0.0, 0.5); + float energyFactor = mix(roughness, 1.0, 1.0 / 1.51); + float fd90 = energyBias + 2.0 * VoH * VoH * roughness; + float f0 = 1.0; + float lightScatter = f0 + (fd90 - f0) * pow(1.0 - cNdotL, 5.0); + float viewScatter = f0 + (fd90 - f0) * pow(1.0 - cNdotV, 5.0); + + diffuse_brdf_NL = lightScatter * viewScatter * energyFactor; + */ + } +#else + // lambert + diffuse_brdf_NL = cNdotL * (1.0 / M_PI); +#endif + + diffuse_light += light_color * diffuse_brdf_NL * attenuation; + +#if defined(LIGHT_BACKLIGHT_USED) + diffuse_light += light_color * (vec3(1.0 / M_PI) - diffuse_brdf_NL) * backlight * attenuation; +#endif + +#if defined(LIGHT_RIM_USED) + float rim_light = pow(max(0.0, 1.0 - cNdotV), max(0.0, (1.0 - roughness) * 16.0)); + diffuse_light += rim_light * rim * mix(vec3(1.0), albedo, rim_tint) * light_color; +#endif + +#ifdef LIGHT_TRANSMITTANCE_USED + + { +#ifdef SSS_MODE_SKIN + float scale = 8.25 / transmittance_depth; + float d = scale * abs(transmittance_z); + float dd = -d * d; + vec3 profile = vec3(0.233, 0.455, 0.649) * exp(dd / 0.0064) + + vec3(0.1, 0.336, 0.344) * exp(dd / 0.0484) + + vec3(0.118, 0.198, 0.0) * exp(dd / 0.187) + + vec3(0.113, 0.007, 0.007) * exp(dd / 0.567) + + vec3(0.358, 0.004, 0.0) * exp(dd / 1.99) + + vec3(0.078, 0.0, 0.0) * exp(dd / 7.41); + + diffuse_light += profile * transmittance_color.a * light_color * clamp(transmittance_boost - NdotL, 0.0, 1.0) * (1.0 / M_PI); +#else + + float scale = 8.25 / transmittance_depth; + float d = scale * abs(transmittance_z); + float dd = -d * d; + diffuse_light += exp(dd) * transmittance_color.rgb * transmittance_color.a * light_color * clamp(transmittance_boost - NdotL, 0.0, 1.0) * (1.0 / M_PI); +#endif + } +#else + +#endif //LIGHT_TRANSMITTANCE_USED + } + + if (roughness > 0.0) { // FIXME: roughness == 0 should not disable specular light entirely + + // D + +#if defined(SPECULAR_TOON) + + vec3 R = normalize(-reflect(L, N)); + float RdotV = dot(R, V); + float mid = 1.0 - roughness; + mid *= mid; + float intensity = smoothstep(mid - roughness * 0.5, mid + roughness * 0.5, RdotV) * mid; + diffuse_light += light_color * intensity * attenuation * specular_amount; // write to diffuse_light, as in toon shading you generally want no reflection + +#elif defined(SPECULAR_DISABLED) + // none.. + +#elif defined(SPECULAR_SCHLICK_GGX) + // shlick+ggx as default + float alpha_ggx = roughness * roughness; +#if defined(LIGHT_ANISOTROPY_USED) + + float aspect = sqrt(1.0 - anisotropy * 0.9); + float ax = alpha_ggx / aspect; + float ay = alpha_ggx * aspect; + float XdotH = dot(T, H); + float YdotH = dot(B, H); + float D = D_GGX_anisotropic(cNdotH, ax, ay, XdotH, YdotH); + float G = V_GGX_anisotropic(ax, ay, dot(T, V), dot(T, L), dot(B, V), dot(B, L), cNdotV, cNdotL); +#else // LIGHT_ANISOTROPY_USED + float D = D_GGX(cNdotH, alpha_ggx); + float G = V_GGX(cNdotL, cNdotV, alpha_ggx); +#endif // LIGHT_ANISOTROPY_USED + // F + float cLdotH5 = SchlickFresnel(cLdotH); + vec3 F = mix(vec3(cLdotH5), vec3(1.0), f0); + + vec3 specular_brdf_NL = cNdotL * D * F * G; + + specular_light += specular_brdf_NL * light_color * attenuation * specular_amount; +#endif + +#if defined(LIGHT_CLEARCOAT_USED) + // Clearcoat ignores normal_map, use vertex normal instead + float ccNdotL = max(min(A + dot(vertex_normal, L), 1.0), 0.0); + float ccNdotH = clamp(A + dot(vertex_normal, H), 0.0, 1.0); + float ccNdotV = max(dot(vertex_normal, V), 1e-4); + +#if !defined(SPECULAR_SCHLICK_GGX) + float cLdotH5 = SchlickFresnel(cLdotH); +#endif + float Dr = D_GGX(ccNdotH, mix(0.001, 0.1, clearcoat_roughness)); + float Gr = 0.25 / (cLdotH * cLdotH); + float Fr = mix(.04, 1.0, cLdotH5); + float clearcoat_specular_brdf_NL = clearcoat * Gr * Fr * Dr * cNdotL; + + specular_light += clearcoat_specular_brdf_NL * light_color * attenuation * specular_amount; + // TODO: Clearcoat adds light to the scene right now (it is non-energy conserving), both diffuse and specular need to be scaled by (1.0 - FR) + // but to do so we need to rearrange this entire function +#endif // LIGHT_CLEARCOAT_USED + } + +#ifdef USE_SHADOW_TO_OPACITY + alpha = min(alpha, clamp(1.0 - attenuation, 0.0, 1.0)); +#endif + +#endif //defined(LIGHT_CODE_USED) +} + +#ifndef SHADOWS_DISABLED + +// Interleaved Gradient Noise +// https://www.iryoku.com/next-generation-post-processing-in-call-of-duty-advanced-warfare +float quick_hash(vec2 pos) { + const vec3 magic = vec3(0.06711056f, 0.00583715f, 52.9829189f); + return fract(magic.z * fract(dot(pos, magic.xy))); +} + +float sample_directional_pcf_shadow(texture2D shadow, vec2 shadow_pixel_size, vec4 coord) { + vec2 pos = coord.xy; + float depth = coord.z; + + //if only one sample is taken, take it from the center + if (sc_directional_soft_shadow_samples == 0) { + return textureProj(sampler2DShadow(shadow, shadow_sampler), vec4(pos, depth, 1.0)); + } + + mat2 disk_rotation; + { + float r = quick_hash(gl_FragCoord.xy) * 2.0 * M_PI; + float sr = sin(r); + float cr = cos(r); + disk_rotation = mat2(vec2(cr, -sr), vec2(sr, cr)); + } + + float avg = 0.0; + + for (uint i = 0; i < sc_directional_soft_shadow_samples; i++) { + avg += textureProj(sampler2DShadow(shadow, shadow_sampler), vec4(pos + shadow_pixel_size * (disk_rotation * scene_data_block.data.directional_soft_shadow_kernel[i].xy), depth, 1.0)); + } + + return avg * (1.0 / float(sc_directional_soft_shadow_samples)); +} + +float sample_pcf_shadow(texture2D shadow, vec2 shadow_pixel_size, vec3 coord) { + vec2 pos = coord.xy; + float depth = coord.z; + + //if only one sample is taken, take it from the center + if (sc_soft_shadow_samples == 0) { + return textureProj(sampler2DShadow(shadow, shadow_sampler), vec4(pos, depth, 1.0)); + } + + mat2 disk_rotation; + { + float r = quick_hash(gl_FragCoord.xy) * 2.0 * M_PI; + float sr = sin(r); + float cr = cos(r); + disk_rotation = mat2(vec2(cr, -sr), vec2(sr, cr)); + } + + float avg = 0.0; + + for (uint i = 0; i < sc_soft_shadow_samples; i++) { + avg += textureProj(sampler2DShadow(shadow, shadow_sampler), vec4(pos + shadow_pixel_size * (disk_rotation * scene_data_block.data.soft_shadow_kernel[i].xy), depth, 1.0)); + } + + return avg * (1.0 / float(sc_soft_shadow_samples)); +} + +float sample_omni_pcf_shadow(texture2D shadow, float blur_scale, vec2 coord, vec4 uv_rect, vec2 flip_offset, float depth) { + //if only one sample is taken, take it from the center + if (sc_soft_shadow_samples == 0) { + vec2 pos = coord * 0.5 + 0.5; + pos = uv_rect.xy + pos * uv_rect.zw; + return textureProj(sampler2DShadow(shadow, shadow_sampler), vec4(pos, depth, 1.0)); + } + + mat2 disk_rotation; + { + float r = quick_hash(gl_FragCoord.xy) * 2.0 * M_PI; + float sr = sin(r); + float cr = cos(r); + disk_rotation = mat2(vec2(cr, -sr), vec2(sr, cr)); + } + + float avg = 0.0; + vec2 offset_scale = blur_scale * 2.0 * scene_data_block.data.shadow_atlas_pixel_size / uv_rect.zw; + + for (uint i = 0; i < sc_soft_shadow_samples; i++) { + vec2 offset = offset_scale * (disk_rotation * scene_data_block.data.soft_shadow_kernel[i].xy); + vec2 sample_coord = coord + offset; + + float sample_coord_length_sqaured = dot(sample_coord, sample_coord); + bool do_flip = sample_coord_length_sqaured > 1.0; + + if (do_flip) { + float len = sqrt(sample_coord_length_sqaured); + sample_coord = sample_coord * (2.0 / len - 1.0); + } + + sample_coord = sample_coord * 0.5 + 0.5; + sample_coord = uv_rect.xy + sample_coord * uv_rect.zw; + + if (do_flip) { + sample_coord += flip_offset; + } + avg += textureProj(sampler2DShadow(shadow, shadow_sampler), vec4(sample_coord, depth, 1.0)); + } + + return avg * (1.0 / float(sc_soft_shadow_samples)); +} + +float sample_directional_soft_shadow(texture2D shadow, vec3 pssm_coord, vec2 tex_scale) { + //find blocker + float blocker_count = 0.0; + float blocker_average = 0.0; + + mat2 disk_rotation; + { + float r = quick_hash(gl_FragCoord.xy) * 2.0 * M_PI; + float sr = sin(r); + float cr = cos(r); + disk_rotation = mat2(vec2(cr, -sr), vec2(sr, cr)); + } + + for (uint i = 0; i < sc_directional_penumbra_shadow_samples; i++) { + vec2 suv = pssm_coord.xy + (disk_rotation * scene_data_block.data.directional_penumbra_shadow_kernel[i].xy) * tex_scale; + float d = textureLod(sampler2D(shadow, material_samplers[SAMPLER_LINEAR_CLAMP]), suv, 0.0).r; + if (d < pssm_coord.z) { + blocker_average += d; + blocker_count += 1.0; + } + } + + if (blocker_count > 0.0) { + //blockers found, do soft shadow + blocker_average /= blocker_count; + float penumbra = (pssm_coord.z - blocker_average) / blocker_average; + tex_scale *= penumbra; + + float s = 0.0; + for (uint i = 0; i < sc_directional_penumbra_shadow_samples; i++) { + vec2 suv = pssm_coord.xy + (disk_rotation * scene_data_block.data.directional_penumbra_shadow_kernel[i].xy) * tex_scale; + s += textureProj(sampler2DShadow(shadow, shadow_sampler), vec4(suv, pssm_coord.z, 1.0)); + } + + return s / float(sc_directional_penumbra_shadow_samples); + + } else { + //no blockers found, so no shadow + return 1.0; + } +} + +#endif // SHADOWS_DISABLED + +float get_omni_attenuation(float distance, float inv_range, float decay) { + float nd = distance * inv_range; + nd *= nd; + nd *= nd; // nd^4 + nd = max(1.0 - nd, 0.0); + nd *= nd; // nd^2 + return nd * pow(max(distance, 0.0001), -decay); +} + +float light_process_omni_shadow(uint idx, vec3 vertex, vec3 normal) { +#ifndef SHADOWS_DISABLED + if (omni_lights.data[idx].shadow_enabled) { + // there is a shadowmap + vec2 texel_size = scene_data_block.data.shadow_atlas_pixel_size; + vec4 base_uv_rect = omni_lights.data[idx].atlas_rect; + base_uv_rect.xy += texel_size; + base_uv_rect.zw -= texel_size * 2.0; + + // Omni lights use direction.xy to store to store the offset between the two paraboloid regions + vec2 flip_offset = omni_lights.data[idx].direction.xy; + + vec3 local_vert = (omni_lights.data[idx].shadow_matrix * vec4(vertex, 1.0)).xyz; + + float shadow_len = length(local_vert); //need to remember shadow len from here + vec3 shadow_dir = normalize(local_vert); + + vec3 local_normal = normalize(mat3(omni_lights.data[idx].shadow_matrix) * normal); + vec3 normal_bias = local_normal * omni_lights.data[idx].shadow_normal_bias * (1.0 - abs(dot(local_normal, shadow_dir))); + + float shadow; + + if (sc_use_light_soft_shadows && omni_lights.data[idx].soft_shadow_size > 0.0) { + //soft shadow + + //find blocker + + float blocker_count = 0.0; + float blocker_average = 0.0; + + mat2 disk_rotation; + { + float r = quick_hash(gl_FragCoord.xy) * 2.0 * M_PI; + float sr = sin(r); + float cr = cos(r); + disk_rotation = mat2(vec2(cr, -sr), vec2(sr, cr)); + } + + vec3 basis_normal = shadow_dir; + vec3 v0 = abs(basis_normal.z) < 0.999 ? vec3(0.0, 0.0, 1.0) : vec3(0.0, 1.0, 0.0); + vec3 tangent = normalize(cross(v0, basis_normal)); + vec3 bitangent = normalize(cross(tangent, basis_normal)); + float z_norm = shadow_len * omni_lights.data[idx].inv_radius; + + tangent *= omni_lights.data[idx].soft_shadow_size * omni_lights.data[idx].soft_shadow_scale; + bitangent *= omni_lights.data[idx].soft_shadow_size * omni_lights.data[idx].soft_shadow_scale; + + for (uint i = 0; i < sc_penumbra_shadow_samples; i++) { + vec2 disk = disk_rotation * scene_data_block.data.penumbra_shadow_kernel[i].xy; + + vec3 pos = local_vert + tangent * disk.x + bitangent * disk.y; + + pos = normalize(pos); + + vec4 uv_rect = base_uv_rect; + + if (pos.z >= 0.0) { + uv_rect.xy += flip_offset; + } + + pos.z = 1.0 + abs(pos.z); + pos.xy /= pos.z; + + pos.xy = pos.xy * 0.5 + 0.5; + pos.xy = uv_rect.xy + pos.xy * uv_rect.zw; + + float d = textureLod(sampler2D(shadow_atlas, material_samplers[SAMPLER_LINEAR_CLAMP]), pos.xy, 0.0).r; + if (d < z_norm) { + blocker_average += d; + blocker_count += 1.0; + } + } + + if (blocker_count > 0.0) { + //blockers found, do soft shadow + blocker_average /= blocker_count; + float penumbra = (z_norm - blocker_average) / blocker_average; + tangent *= penumbra; + bitangent *= penumbra; + + z_norm -= omni_lights.data[idx].inv_radius * omni_lights.data[idx].shadow_bias; + + shadow = 0.0; + for (uint i = 0; i < sc_penumbra_shadow_samples; i++) { + vec2 disk = disk_rotation * scene_data_block.data.penumbra_shadow_kernel[i].xy; + vec3 pos = local_vert + tangent * disk.x + bitangent * disk.y; + + pos = normalize(pos); + pos = normalize(pos + normal_bias); + + vec4 uv_rect = base_uv_rect; + + if (pos.z >= 0.0) { + uv_rect.xy += flip_offset; + } + + pos.z = 1.0 + abs(pos.z); + pos.xy /= pos.z; + + pos.xy = pos.xy * 0.5 + 0.5; + pos.xy = uv_rect.xy + pos.xy * uv_rect.zw; + shadow += textureProj(sampler2DShadow(shadow_atlas, shadow_sampler), vec4(pos.xy, z_norm, 1.0)); + } + + shadow /= float(sc_penumbra_shadow_samples); + + } else { + //no blockers found, so no shadow + shadow = 1.0; + } + } else { + vec4 uv_rect = base_uv_rect; + + vec3 shadow_sample = normalize(shadow_dir + normal_bias); + if (shadow_sample.z >= 0.0) { + uv_rect.xy += flip_offset; + flip_offset *= -1.0; + } + + shadow_sample.z = 1.0 + abs(shadow_sample.z); + 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); + } + + return shadow; + } +#endif + + return 1.0; +} + +void light_process_omni(uint idx, vec3 vertex, vec3 eye_vec, vec3 normal, vec3 vertex_ddx, vec3 vertex_ddy, vec3 f0, uint orms, float shadow, vec3 albedo, inout float alpha, +#ifdef LIGHT_BACKLIGHT_USED + vec3 backlight, +#endif +#ifdef LIGHT_TRANSMITTANCE_USED + vec4 transmittance_color, + float transmittance_depth, + float transmittance_boost, +#endif +#ifdef LIGHT_RIM_USED + float rim, float rim_tint, +#endif +#ifdef LIGHT_CLEARCOAT_USED + float clearcoat, float clearcoat_roughness, vec3 vertex_normal, +#endif +#ifdef LIGHT_ANISOTROPY_USED + vec3 binormal, vec3 tangent, float anisotropy, +#endif + inout vec3 diffuse_light, inout vec3 specular_light) { + vec3 light_rel_vec = omni_lights.data[idx].position - vertex; + float light_length = length(light_rel_vec); + float omni_attenuation = get_omni_attenuation(light_length, omni_lights.data[idx].inv_radius, omni_lights.data[idx].attenuation); + float light_attenuation = omni_attenuation; + vec3 color = omni_lights.data[idx].color; + + float size_A = 0.0; + + if (sc_use_light_soft_shadows && omni_lights.data[idx].size > 0.0) { + float t = omni_lights.data[idx].size / max(0.001, light_length); + size_A = max(0.0, 1.0 - 1 / sqrt(1 + t * t)); + } + +#ifdef LIGHT_TRANSMITTANCE_USED + float transmittance_z = transmittance_depth; //no transmittance by default + transmittance_color.a *= light_attenuation; + { + vec4 clamp_rect = omni_lights.data[idx].atlas_rect; + + //redo shadowmapping, but shrink the model a bit to avoid arctifacts + vec4 splane = (omni_lights.data[idx].shadow_matrix * vec4(vertex - normalize(normal_interp) * omni_lights.data[idx].transmittance_bias, 1.0)); + + float shadow_len = length(splane.xyz); + splane.xyz = normalize(splane.xyz); + + if (splane.z >= 0.0) { + splane.z += 1.0; + clamp_rect.y += clamp_rect.w; + } else { + splane.z = 1.0 - splane.z; + } + + splane.xy /= splane.z; + + splane.xy = splane.xy * 0.5 + 0.5; + splane.z = shadow_len * omni_lights.data[idx].inv_radius; + splane.xy = clamp_rect.xy + splane.xy * clamp_rect.zw; + // splane.xy = clamp(splane.xy,clamp_rect.xy + scene_data_block.data.shadow_atlas_pixel_size,clamp_rect.xy + clamp_rect.zw - scene_data_block.data.shadow_atlas_pixel_size ); + splane.w = 1.0; //needed? i think it should be 1 already + + float shadow_z = textureLod(sampler2D(shadow_atlas, material_samplers[SAMPLER_LINEAR_CLAMP]), splane.xy, 0.0).r; + transmittance_z = (splane.z - shadow_z) / omni_lights.data[idx].inv_radius; + } +#endif + + if (sc_use_light_projector && omni_lights.data[idx].projector_rect != vec4(0.0)) { + vec3 local_v = (omni_lights.data[idx].shadow_matrix * vec4(vertex, 1.0)).xyz; + local_v = normalize(local_v); + + vec4 atlas_rect = omni_lights.data[idx].projector_rect; + + if (local_v.z >= 0.0) { + atlas_rect.y += atlas_rect.w; + } + + local_v.z = 1.0 + abs(local_v.z); + + local_v.xy /= local_v.z; + local_v.xy = local_v.xy * 0.5 + 0.5; + vec2 proj_uv = local_v.xy * atlas_rect.zw; + + if (sc_projector_use_mipmaps) { + vec2 proj_uv_ddx; + vec2 proj_uv_ddy; + { + vec3 local_v_ddx = (omni_lights.data[idx].shadow_matrix * vec4(vertex + vertex_ddx, 1.0)).xyz; + local_v_ddx = normalize(local_v_ddx); + + if (local_v_ddx.z >= 0.0) { + local_v_ddx.z += 1.0; + } else { + local_v_ddx.z = 1.0 - local_v_ddx.z; + } + + local_v_ddx.xy /= local_v_ddx.z; + local_v_ddx.xy = local_v_ddx.xy * 0.5 + 0.5; + + proj_uv_ddx = local_v_ddx.xy * atlas_rect.zw - proj_uv; + + vec3 local_v_ddy = (omni_lights.data[idx].shadow_matrix * vec4(vertex + vertex_ddy, 1.0)).xyz; + local_v_ddy = normalize(local_v_ddy); + + if (local_v_ddy.z >= 0.0) { + local_v_ddy.z += 1.0; + } else { + local_v_ddy.z = 1.0 - local_v_ddy.z; + } + + local_v_ddy.xy /= local_v_ddy.z; + local_v_ddy.xy = local_v_ddy.xy * 0.5 + 0.5; + + proj_uv_ddy = local_v_ddy.xy * atlas_rect.zw - proj_uv; + } + + vec4 proj = textureGrad(sampler2D(decal_atlas_srgb, light_projector_sampler), proj_uv + atlas_rect.xy, proj_uv_ddx, proj_uv_ddy); + color *= proj.rgb * proj.a; + } else { + vec4 proj = textureLod(sampler2D(decal_atlas_srgb, light_projector_sampler), proj_uv + atlas_rect.xy, 0.0); + color *= proj.rgb * proj.a; + } + } + + light_attenuation *= shadow; + + light_compute(normal, normalize(light_rel_vec), eye_vec, size_A, color, light_attenuation, f0, orms, omni_lights.data[idx].specular_amount, albedo, alpha, +#ifdef LIGHT_BACKLIGHT_USED + backlight, +#endif +#ifdef LIGHT_TRANSMITTANCE_USED + transmittance_color, + transmittance_depth, + transmittance_boost, + transmittance_z, +#endif +#ifdef LIGHT_RIM_USED + rim * omni_attenuation, rim_tint, +#endif +#ifdef LIGHT_CLEARCOAT_USED + clearcoat, clearcoat_roughness, vertex_normal, +#endif +#ifdef LIGHT_ANISOTROPY_USED + binormal, tangent, anisotropy, +#endif + diffuse_light, + specular_light); +} + +float light_process_spot_shadow(uint idx, vec3 vertex, vec3 normal) { +#ifndef SHADOWS_DISABLED + if (spot_lights.data[idx].shadow_enabled) { + 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; + + vec3 shadow_dir = light_rel_vec / light_length; + vec3 normal_bias = normal * light_length * spot_lights.data[idx].shadow_normal_bias * (1.0 - abs(dot(normal, shadow_dir))); + + //there is a shadowmap + vec4 v = vec4(vertex + normal_bias, 1.0); + + vec4 splane = (spot_lights.data[idx].shadow_matrix * v); + splane.z -= spot_lights.data[idx].shadow_bias / (light_length * spot_lights.data[idx].inv_radius); + splane /= splane.w; + + float shadow; + if (sc_use_light_soft_shadows && spot_lights.data[idx].soft_shadow_size > 0.0) { + //soft shadow + + //find blocker + float z_norm = dot(spot_dir, -light_rel_vec) * spot_lights.data[idx].inv_radius; + + vec2 shadow_uv = splane.xy * spot_lights.data[idx].atlas_rect.zw + spot_lights.data[idx].atlas_rect.xy; + + float blocker_count = 0.0; + float blocker_average = 0.0; + + mat2 disk_rotation; + { + float r = quick_hash(gl_FragCoord.xy) * 2.0 * M_PI; + float sr = sin(r); + float cr = cos(r); + disk_rotation = mat2(vec2(cr, -sr), vec2(sr, cr)); + } + + float uv_size = spot_lights.data[idx].soft_shadow_size * z_norm * spot_lights.data[idx].soft_shadow_scale; + vec2 clamp_max = spot_lights.data[idx].atlas_rect.xy + spot_lights.data[idx].atlas_rect.zw; + for (uint i = 0; i < sc_penumbra_shadow_samples; i++) { + vec2 suv = shadow_uv + (disk_rotation * scene_data_block.data.penumbra_shadow_kernel[i].xy) * uv_size; + suv = clamp(suv, spot_lights.data[idx].atlas_rect.xy, clamp_max); + float d = textureLod(sampler2D(shadow_atlas, material_samplers[SAMPLER_LINEAR_CLAMP]), suv, 0.0).r; + if (d < splane.z) { + blocker_average += d; + blocker_count += 1.0; + } + } + + if (blocker_count > 0.0) { + //blockers found, do soft shadow + blocker_average /= blocker_count; + float penumbra = (z_norm - blocker_average) / blocker_average; + uv_size *= penumbra; + + shadow = 0.0; + for (uint i = 0; i < sc_penumbra_shadow_samples; i++) { + vec2 suv = shadow_uv + (disk_rotation * scene_data_block.data.penumbra_shadow_kernel[i].xy) * uv_size; + suv = clamp(suv, spot_lights.data[idx].atlas_rect.xy, clamp_max); + shadow += textureProj(sampler2DShadow(shadow_atlas, shadow_sampler), vec4(suv, splane.z, 1.0)); + } + + shadow /= float(sc_penumbra_shadow_samples); + + } else { + //no blockers found, so no shadow + shadow = 1.0; + } + } 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); + } + + return shadow; + } + +#endif // SHADOWS_DISABLED + + return 1.0; +} + +vec2 normal_to_panorama(vec3 n) { + n = normalize(n); + vec2 panorama_coords = vec2(atan(n.x, n.z), acos(-n.y)); + + if (panorama_coords.x < 0.0) { + panorama_coords.x += M_PI * 2.0; + } + + panorama_coords /= vec2(M_PI * 2.0, M_PI); + return panorama_coords; +} + +void light_process_spot(uint idx, vec3 vertex, vec3 eye_vec, vec3 normal, vec3 vertex_ddx, vec3 vertex_ddy, vec3 f0, uint orms, float shadow, vec3 albedo, inout float alpha, +#ifdef LIGHT_BACKLIGHT_USED + vec3 backlight, +#endif +#ifdef LIGHT_TRANSMITTANCE_USED + vec4 transmittance_color, + float transmittance_depth, + float transmittance_boost, +#endif +#ifdef LIGHT_RIM_USED + float rim, float rim_tint, +#endif +#ifdef LIGHT_CLEARCOAT_USED + float clearcoat, float clearcoat_roughness, vec3 vertex_normal, +#endif +#ifdef LIGHT_ANISOTROPY_USED + vec3 binormal, vec3 tangent, float anisotropy, +#endif + inout vec3 diffuse_light, + inout vec3 specular_light) { + vec3 light_rel_vec = spot_lights.data[idx].position - vertex; + float light_length = length(light_rel_vec); + float spot_attenuation = get_omni_attenuation(light_length, spot_lights.data[idx].inv_radius, spot_lights.data[idx].attenuation); + vec3 spot_dir = spot_lights.data[idx].direction; + float scos = max(dot(-normalize(light_rel_vec), spot_dir), spot_lights.data[idx].cone_angle); + float spot_rim = max(0.0001, (1.0 - scos) / (1.0 - spot_lights.data[idx].cone_angle)); + spot_attenuation *= 1.0 - pow(spot_rim, spot_lights.data[idx].cone_attenuation); + float light_attenuation = spot_attenuation; + vec3 color = spot_lights.data[idx].color; + float specular_amount = spot_lights.data[idx].specular_amount; + + float size_A = 0.0; + + if (sc_use_light_soft_shadows && spot_lights.data[idx].size > 0.0) { + float t = spot_lights.data[idx].size / max(0.001, light_length); + size_A = max(0.0, 1.0 - 1 / sqrt(1 + t * t)); + } + +#ifdef LIGHT_TRANSMITTANCE_USED + float transmittance_z = transmittance_depth; + transmittance_color.a *= light_attenuation; + { + vec4 splane = (spot_lights.data[idx].shadow_matrix * vec4(vertex - normalize(normal_interp) * spot_lights.data[idx].transmittance_bias, 1.0)); + splane /= splane.w; + splane.xy = splane.xy * spot_lights.data[idx].atlas_rect.zw + spot_lights.data[idx].atlas_rect.xy; + + float shadow_z = textureLod(sampler2D(shadow_atlas, material_samplers[SAMPLER_LINEAR_CLAMP]), splane.xy, 0.0).r; + + shadow_z = shadow_z * 2.0 - 1.0; + float z_far = 1.0 / spot_lights.data[idx].inv_radius; + float z_near = 0.01; + shadow_z = 2.0 * z_near * z_far / (z_far + z_near - shadow_z * (z_far - z_near)); + + //distance to light plane + float z = dot(spot_dir, -light_rel_vec); + transmittance_z = z - shadow_z; + } +#endif //LIGHT_TRANSMITTANCE_USED + + if (sc_use_light_projector && spot_lights.data[idx].projector_rect != vec4(0.0)) { + vec4 splane = (spot_lights.data[idx].shadow_matrix * vec4(vertex, 1.0)); + splane /= splane.w; + + vec2 proj_uv = splane.xy * spot_lights.data[idx].projector_rect.zw; + + if (sc_projector_use_mipmaps) { + //ensure we have proper mipmaps + vec4 splane_ddx = (spot_lights.data[idx].shadow_matrix * vec4(vertex + vertex_ddx, 1.0)); + splane_ddx /= splane_ddx.w; + vec2 proj_uv_ddx = splane_ddx.xy * spot_lights.data[idx].projector_rect.zw - proj_uv; + + vec4 splane_ddy = (spot_lights.data[idx].shadow_matrix * vec4(vertex + vertex_ddy, 1.0)); + splane_ddy /= splane_ddy.w; + vec2 proj_uv_ddy = splane_ddy.xy * spot_lights.data[idx].projector_rect.zw - proj_uv; + + vec4 proj = textureGrad(sampler2D(decal_atlas_srgb, light_projector_sampler), proj_uv + spot_lights.data[idx].projector_rect.xy, proj_uv_ddx, proj_uv_ddy); + color *= proj.rgb * proj.a; + } else { + vec4 proj = textureLod(sampler2D(decal_atlas_srgb, light_projector_sampler), proj_uv + spot_lights.data[idx].projector_rect.xy, 0.0); + color *= proj.rgb * proj.a; + } + } + light_attenuation *= shadow; + + light_compute(normal, normalize(light_rel_vec), eye_vec, size_A, color, light_attenuation, f0, orms, spot_lights.data[idx].specular_amount, albedo, alpha, +#ifdef LIGHT_BACKLIGHT_USED + backlight, +#endif +#ifdef LIGHT_TRANSMITTANCE_USED + transmittance_color, + transmittance_depth, + transmittance_boost, + transmittance_z, +#endif +#ifdef LIGHT_RIM_USED + rim * spot_attenuation, rim_tint, +#endif +#ifdef LIGHT_CLEARCOAT_USED + clearcoat, clearcoat_roughness, vertex_normal, +#endif +#ifdef LIGHT_ANISOTROPY_USED + binormal, tangent, anisotropy, +#endif + 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) { + vec3 box_extents = reflections.data[ref_index].box_extents; + vec3 local_pos = (reflections.data[ref_index].local_matrix * vec4(vertex, 1.0)).xyz; + + if (any(greaterThan(abs(local_pos), box_extents))) { //out of the reflection box + 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 + blend = mix(length(inner_pos), blend, blend); + blend *= blend; + blend = max(0.0, 1.0 - blend); + + if (reflections.data[ref_index].intensity > 0.0) { // compute reflection + + vec3 local_ref_vec = (reflections.data[ref_index].local_matrix * vec4(ref_vec, 0.0)).xyz; + + if (reflections.data[ref_index].box_project) { //box project + + vec3 nrdir = normalize(local_ref_vec); + vec3 rbmax = (box_extents - local_pos) / nrdir; + vec3 rbmin = (-box_extents - local_pos) / nrdir; + + vec3 rbminmax = mix(rbmin, rbmax, greaterThan(nrdir, vec3(0.0, 0.0, 0.0))); + + float fa = min(min(rbminmax.x, rbminmax.y), rbminmax.z); + vec3 posonbox = local_pos + nrdir * fa; + local_ref_vec = posonbox - reflections.data[ref_index].box_offset; + } + + 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; + + if (reflections.data[ref_index].exterior) { + reflection.rgb = mix(specular_light, reflection.rgb, blend); + } + + reflection.rgb *= reflections.data[ref_index].intensity; //intensity + reflection.a = blend; + reflection.rgb *= reflection.a; + + reflection_accum += reflection; + } + + switch (reflections.data[ref_index].ambient_mode) { + case REFLECTION_AMBIENT_DISABLED: { + //do nothing + } break; + case REFLECTION_AMBIENT_ENVIRONMENT: { + //do nothing + vec3 local_amb_vec = (reflections.data[ref_index].local_matrix * vec4(normal, 0.0)).xyz; + + 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.a = blend; + if (reflections.data[ref_index].exterior) { + ambient_out.rgb = mix(ambient_light, ambient_out.rgb, blend); + } + + ambient_out.rgb *= ambient_out.a; + ambient_accum += ambient_out; + } break; + case REFLECTION_AMBIENT_COLOR: { + vec4 ambient_out; + ambient_out.a = blend; + ambient_out.rgb = reflections.data[ref_index].ambient; + if (reflections.data[ref_index].exterior) { + ambient_out.rgb = mix(ambient_light, ambient_out.rgb, blend); + } + ambient_out.rgb *= ambient_out.a; + ambient_accum += ambient_out; + } break; + } +} + +float blur_shadow(float shadow) { + return shadow; +#if 0 + //disabling for now, will investigate later + float interp_shadow = shadow; + if (gl_HelperInvocation) { + interp_shadow = -4.0; // technically anything below -4 will do but just to make sure + } + + uvec2 fc2 = uvec2(gl_FragCoord.xy); + interp_shadow -= dFdx(interp_shadow) * (float(fc2.x & 1) - 0.5); + interp_shadow -= dFdy(interp_shadow) * (float(fc2.y & 1) - 0.5); + + if (interp_shadow >= 0.0) { + shadow = interp_shadow; + } + return shadow; +#endif +} diff --git a/servers/rendering/renderer_rd/shaders/scene_forward_mobile.glsl b/servers/rendering/renderer_rd/shaders/scene_forward_mobile.glsl new file mode 100644 index 0000000000..26d0de46c2 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/scene_forward_mobile.glsl @@ -0,0 +1,1626 @@ +#[vertex] + +#version 450 + +#VERSION_DEFINES + +/* Include our forward mobile UBOs definitions etc. */ +#include "scene_forward_mobile_inc.glsl" + +#define SHADER_IS_SRGB false + +/* INPUT ATTRIBS */ + +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; +#endif + +#if defined(TANGENT_USED) || defined(NORMAL_MAP_USED) || defined(LIGHT_ANISOTROPY_USED) +layout(location = 2) in vec4 tangent_attrib; +#endif + +#if defined(COLOR_USED) +layout(location = 3) in vec4 color_attrib; +#endif + +#ifdef UV_USED +layout(location = 4) in vec2 uv_attrib; +#endif + +#if defined(UV2_USED) || defined(USE_LIGHTMAP) || defined(MODE_RENDER_MATERIAL) +layout(location = 5) in vec2 uv2_attrib; +#endif // MODE_RENDER_MATERIAL + +#if defined(CUSTOM0_USED) +layout(location = 6) in vec4 custom0_attrib; +#endif + +#if defined(CUSTOM1_USED) +layout(location = 7) in vec4 custom1_attrib; +#endif + +#if defined(CUSTOM2_USED) +layout(location = 8) in vec4 custom2_attrib; +#endif + +#if defined(CUSTOM3_USED) +layout(location = 9) in vec4 custom3_attrib; +#endif + +#if defined(BONES_USED) || defined(USE_PARTICLE_TRAILS) +layout(location = 10) in uvec4 bone_attrib; +#endif + +#if defined(WEIGHTS_USED) || defined(USE_PARTICLE_TRAILS) +layout(location = 11) in vec4 weight_attrib; +#endif + +/* Varyings */ + +layout(location = 0) highp out vec3 vertex_interp; + +#ifdef NORMAL_USED +layout(location = 1) mediump out vec3 normal_interp; +#endif + +#if defined(COLOR_USED) +layout(location = 2) mediump out vec4 color_interp; +#endif + +#ifdef UV_USED +layout(location = 3) mediump out vec2 uv_interp; +#endif + +#if defined(UV2_USED) || defined(USE_LIGHTMAP) +layout(location = 4) mediump out vec2 uv2_interp; +#endif + +#if defined(TANGENT_USED) || defined(NORMAL_MAP_USED) || defined(LIGHT_ANISOTROPY_USED) +layout(location = 5) mediump out vec3 tangent_interp; +layout(location = 6) mediump out vec3 binormal_interp; +#endif + +#ifdef MATERIAL_UNIFORMS_USED +layout(set = MATERIAL_UNIFORM_SET, binding = 0, std140) uniform MaterialUniforms{ + +#MATERIAL_UNIFORMS + +} material; +#endif + +#ifdef MODE_DUAL_PARABOLOID + +layout(location = 8) out highp float dp_clip; + +#endif + +#ifdef USE_MULTIVIEW +#ifdef has_VK_KHR_multiview +#define ViewIndex gl_ViewIndex +#else +// !BAS! This needs to become an input once we implement our fallback! +#define ViewIndex 0 +#endif +#else +// Set to zero, not supported in non stereo +#define ViewIndex 0 +#endif //USE_MULTIVIEW + +invariant gl_Position; + +#GLOBALS + +#define scene_data scene_data_block.data + +void main() { + vec4 instance_custom = vec4(0.0); +#if defined(COLOR_USED) + color_interp = color_attrib; +#endif + + bool is_multimesh = bool(draw_call.flags & INSTANCE_FLAGS_MULTIMESH); + + mat4 model_matrix = draw_call.transform; + + mat3 model_normal_matrix; + if (bool(draw_call.flags & INSTANCE_FLAGS_NON_UNIFORM_SCALE)) { + model_normal_matrix = transpose(inverse(mat3(model_matrix))); + } else { + model_normal_matrix = mat3(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 + + uint offset = trail_size * stride * gl_InstanceIndex; + +#ifdef COLOR_USED + vec4 pcolor; +#endif + { + uint boffset = offset + bone_attrib.x * stride; + matrix = mat4(transforms.data[boffset + 0], transforms.data[boffset + 1], transforms.data[boffset + 2], vec4(0.0, 0.0, 0.0, 1.0)) * weight_attrib.x; +#ifdef COLOR_USED + pcolor = transforms.data[boffset + 3] * weight_attrib.x; +#endif + } + if (weight_attrib.y > 0.001) { + uint boffset = offset + bone_attrib.y * stride; + matrix += mat4(transforms.data[boffset + 0], transforms.data[boffset + 1], transforms.data[boffset + 2], vec4(0.0, 0.0, 0.0, 1.0)) * weight_attrib.y; +#ifdef COLOR_USED + pcolor += transforms.data[boffset + 3] * weight_attrib.y; +#endif + } + if (weight_attrib.z > 0.001) { + uint boffset = offset + bone_attrib.z * stride; + matrix += mat4(transforms.data[boffset + 0], transforms.data[boffset + 1], transforms.data[boffset + 2], vec4(0.0, 0.0, 0.0, 1.0)) * weight_attrib.z; +#ifdef COLOR_USED + pcolor += transforms.data[boffset + 3] * weight_attrib.z; +#endif + } + if (weight_attrib.w > 0.001) { + uint boffset = offset + bone_attrib.w * stride; + matrix += mat4(transforms.data[boffset + 0], transforms.data[boffset + 1], transforms.data[boffset + 2], vec4(0.0, 0.0, 0.0, 1.0)) * weight_attrib.w; +#ifdef COLOR_USED + pcolor += transforms.data[boffset + 3] * weight_attrib.w; +#endif + } + + instance_custom = transforms.data[offset + 4]; + +#ifdef COLOR_USED + color_interp *= pcolor; +#endif + +#else + uint stride = 0; + { + //TODO implement a small lookup table for the stride + if (bool(draw_call.flags & INSTANCE_FLAGS_MULTIMESH_FORMAT_2D)) { + stride += 2; + } else { + stride += 3; + } + if (bool(draw_call.flags & INSTANCE_FLAGS_MULTIMESH_HAS_COLOR)) { + stride += 1; + } + if (bool(draw_call.flags & INSTANCE_FLAGS_MULTIMESH_HAS_CUSTOM_DATA)) { + stride += 1; + } + } + + uint offset = stride * gl_InstanceIndex; + + if (bool(draw_call.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)); + offset += 2; + } else { + matrix = mat4(transforms.data[offset + 0], transforms.data[offset + 1], transforms.data[offset + 2], vec4(0.0, 0.0, 0.0, 1.0)); + offset += 3; + } + + if (bool(draw_call.flags & INSTANCE_FLAGS_MULTIMESH_HAS_COLOR)) { +#ifdef COLOR_USED + color_interp *= transforms.data[offset]; +#endif + offset += 1; + } + + if (bool(draw_call.flags & INSTANCE_FLAGS_MULTIMESH_HAS_CUSTOM_DATA)) { + instance_custom = transforms.data[offset]; + } + +#endif + //transpose + matrix = transpose(matrix); + model_matrix = model_matrix * matrix; + model_normal_matrix = model_normal_matrix * mat3(matrix); + } + + vec3 vertex = vertex_attrib; +#ifdef NORMAL_USED + vec3 normal = 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; + vec3 binormal = normalize(cross(normal, tangent) * binormalf); +#endif + +#ifdef UV_USED + uv_interp = uv_attrib; +#endif + +#if defined(UV2_USED) || defined(USE_LIGHTMAP) + uv2_interp = uv2_attrib; +#endif + +#ifdef OVERRIDE_POSITION + vec4 position; +#endif + +#ifdef USE_MULTIVIEW + mat4 projection_matrix = scene_data.projection_matrix_view[ViewIndex]; + mat4 inv_projection_matrix = scene_data.inv_projection_matrix_view[ViewIndex]; +#else + mat4 projection_matrix = scene_data.projection_matrix; + mat4 inv_projection_matrix = scene_data.inv_projection_matrix; +#endif //USE_MULTIVIEW + +//using world coordinates +#if !defined(SKIP_TRANSFORM_USED) && defined(VERTEX_WORLD_COORDS_USED) + + vertex = (model_matrix * vec4(vertex, 1.0)).xyz; + +#ifdef NORMAL_USED + normal = model_normal_matrix * normal; +#endif + +#if defined(TANGENT_USED) || defined(NORMAL_MAP_USED) || defined(LIGHT_ANISOTROPY_USED) + + tangent = model_normal_matrix * tangent; + binormal = model_normal_matrix * binormal; + +#endif +#endif + + float roughness = 1.0; + + mat4 modelview = scene_data.view_matrix * model_matrix; + mat3 modelview_normal = mat3(scene_data.view_matrix) * model_normal_matrix; + + { +#CODE : VERTEX + } + + /* output */ + +// using local coordinates (default) +#if !defined(SKIP_TRANSFORM_USED) && !defined(VERTEX_WORLD_COORDS_USED) + + vertex = (modelview * vec4(vertex, 1.0)).xyz; +#ifdef NORMAL_USED + normal = modelview_normal * normal; +#endif + +#endif + +#if defined(TANGENT_USED) || defined(NORMAL_MAP_USED) || defined(LIGHT_ANISOTROPY_USED) + + binormal = modelview_normal * binormal; + tangent = modelview_normal * tangent; +#endif + +//using world coordinates +#if !defined(SKIP_TRANSFORM_USED) && defined(VERTEX_WORLD_COORDS_USED) + + vertex = (scene_data.view_matrix * vec4(vertex, 1.0)).xyz; +#ifdef NORMAL_USED + normal = (scene_data.view_matrix * vec4(normal, 0.0)).xyz; +#endif + +#if defined(TANGENT_USED) || defined(NORMAL_MAP_USED) || defined(LIGHT_ANISOTROPY_USED) + binormal = (scene_data.view_matrix * vec4(binormal, 0.0)).xyz; + tangent = (scene_data.view_matrix * vec4(tangent, 0.0)).xyz; +#endif +#endif + + vertex_interp = vertex; +#ifdef NORMAL_USED + normal_interp = normal; +#endif + +#if defined(TANGENT_USED) || defined(NORMAL_MAP_USED) || defined(LIGHT_ANISOTROPY_USED) + tangent_interp = tangent; + binormal_interp = binormal; +#endif + +#ifdef MODE_RENDER_DEPTH + +#ifdef MODE_DUAL_PARABOLOID + + vertex_interp.z *= scene_data.dual_paraboloid_side; + + dp_clip = vertex_interp.z; //this attempts to avoid noise caused by objects sent to the other parabolloid side due to bias + + //for dual paraboloid shadow mapping, this is the fastest but least correct way, as it curves straight edges + + vec3 vtx = vertex_interp; + float distance = length(vtx); + vtx = normalize(vtx); + vtx.xy /= 1.0 - vtx.z; + vtx.z = (distance / scene_data.z_far); + vtx.z = vtx.z * 2.0 - 1.0; + vertex_interp = vtx; + +#endif + +#endif //MODE_RENDER_DEPTH + +#ifdef OVERRIDE_POSITION + gl_Position = position; +#else + gl_Position = projection_matrix * vec4(vertex_interp, 1.0); +#endif // OVERRIDE_POSITION + +#ifdef MODE_RENDER_DEPTH + if (scene_data.pancake_shadows) { + if (gl_Position.z <= 0.00001) { + gl_Position.z = 0.00001; + } + } +#endif // MODE_RENDER_DEPTH +#ifdef MODE_RENDER_MATERIAL + if (scene_data.material_uv2_mode) { + vec2 uv_offset = draw_call.lightmap_uv_scale.xy; // we are abusing lightmap_uv_scale here, we shouldn't have a lightmap during a depth pass... + gl_Position.xy = (uv2_attrib.xy + uv_offset) * 2.0 - 1.0; + gl_Position.z = 0.00001; + gl_Position.w = 1.0; + } +#endif // MODE_RENDER_MATERIAL +} + +#[fragment] + +#version 450 + +#VERSION_DEFINES + +#define SHADER_IS_SRGB false + +/* Specialization Constants */ + +#if !defined(MODE_RENDER_DEPTH) + +#if !defined(MODE_UNSHADED) + +layout(constant_id = 0) const bool sc_use_light_projector = false; +layout(constant_id = 1) const bool sc_use_light_soft_shadows = false; +layout(constant_id = 2) const bool sc_use_directional_soft_shadows = false; + +layout(constant_id = 3) const uint sc_soft_shadow_samples = 4; +layout(constant_id = 4) const uint sc_penumbra_shadow_samples = 4; + +layout(constant_id = 5) const uint sc_directional_soft_shadow_samples = 4; +layout(constant_id = 6) const uint sc_directional_penumbra_shadow_samples = 4; + +layout(constant_id = 8) const bool sc_projector_use_mipmaps = true; + +layout(constant_id = 9) const bool sc_disable_omni_lights = false; +layout(constant_id = 10) const bool sc_disable_spot_lights = false; +layout(constant_id = 11) const bool sc_disable_reflection_probes = false; +layout(constant_id = 12) const bool sc_disable_directional_lights = false; + +#endif //!MODE_UNSHADED + +layout(constant_id = 7) const bool sc_decal_use_mipmaps = true; +layout(constant_id = 13) const bool sc_disable_decals = false; +layout(constant_id = 14) const bool sc_disable_fog = false; + +#endif //!MODE_RENDER_DEPTH + +layout(constant_id = 15) const float sc_luminance_multiplier = 2.0; + +/* Include our forward mobile UBOs definitions etc. */ +#include "scene_forward_mobile_inc.glsl" + +/* Varyings */ + +layout(location = 0) highp in vec3 vertex_interp; + +#ifdef NORMAL_USED +layout(location = 1) mediump in vec3 normal_interp; +#endif + +#if defined(COLOR_USED) +layout(location = 2) mediump in vec4 color_interp; +#endif + +#ifdef UV_USED +layout(location = 3) mediump in vec2 uv_interp; +#endif + +#if defined(UV2_USED) || defined(USE_LIGHTMAP) +layout(location = 4) mediump in vec2 uv2_interp; +#endif + +#if defined(TANGENT_USED) || defined(NORMAL_MAP_USED) || defined(LIGHT_ANISOTROPY_USED) +layout(location = 5) mediump in vec3 tangent_interp; +layout(location = 6) mediump in vec3 binormal_interp; +#endif + +#ifdef MODE_DUAL_PARABOLOID + +layout(location = 8) highp in float dp_clip; + +#endif + +#ifdef USE_MULTIVIEW +#ifdef has_VK_KHR_multiview +#define ViewIndex gl_ViewIndex +#else +// !BAS! This needs to become an input once we implement our fallback! +#define ViewIndex 0 +#endif +#else +// Set to zero, not supported in non stereo +#define ViewIndex 0 +#endif //USE_MULTIVIEW + +//defines to keep compatibility with vertex + +#define model_matrix draw_call.transform +#ifdef USE_MULTIVIEW +#define projection_matrix scene_data.projection_matrix_view[ViewIndex] +#else +#define projection_matrix scene_data.projection_matrix +#endif + +#if defined(ENABLE_SSS) && defined(ENABLE_TRANSMITTANCE) +//both required for transmittance to be enabled +#define LIGHT_TRANSMITTANCE_USED +#endif + +#ifdef MATERIAL_UNIFORMS_USED +layout(set = MATERIAL_UNIFORM_SET, binding = 0, std140) uniform MaterialUniforms{ + +#MATERIAL_UNIFORMS + +} material; +#endif + +#GLOBALS + +/* clang-format on */ + +#ifdef MODE_RENDER_DEPTH + +#ifdef MODE_RENDER_MATERIAL + +layout(location = 0) out vec4 albedo_output_buffer; +layout(location = 1) out vec4 normal_output_buffer; +layout(location = 2) out vec4 orm_output_buffer; +layout(location = 3) out vec4 emission_output_buffer; +layout(location = 4) out float depth_output_buffer; + +#endif // MODE_RENDER_MATERIAL + +#else // RENDER DEPTH + +#ifdef MODE_MULTIPLE_RENDER_TARGETS + +layout(location = 0) out vec4 diffuse_buffer; //diffuse (rgb) and roughness +layout(location = 1) out vec4 specular_buffer; //specular and SSS (subsurface scatter) +#else + +layout(location = 0) out mediump vec4 frag_color; +#endif // MODE_MULTIPLE_RENDER_TARGETS + +#endif // RENDER DEPTH + +#include "scene_forward_aa_inc.glsl" + +#if !defined(MODE_RENDER_DEPTH) && !defined(MODE_UNSHADED) + +// Default to SPECULAR_SCHLICK_GGX. +#if !defined(SPECULAR_DISABLED) && !defined(SPECULAR_SCHLICK_GGX) && !defined(SPECULAR_TOON) +#define SPECULAR_SCHLICK_GGX +#endif + +#include "scene_forward_lights_inc.glsl" + +#endif //!defined(MODE_RENDER_DEPTH) && !defined(MODE_UNSHADED) + +#ifndef MODE_RENDER_DEPTH + +/* + Only supporting normal fog here. +*/ + +vec4 fog_process(vec3 vertex) { + vec3 fog_color = scene_data_block.data.fog_light_color; + + if (scene_data_block.data.fog_aerial_perspective > 0.0) { + vec3 sky_fog_color = vec3(0.0); + vec3 cube_view = scene_data_block.data.radiance_inverse_xform * vertex; + // mip_level always reads from the second mipmap and higher so the fog is always slightly blurred + float mip_level = mix(1.0 / MAX_ROUGHNESS_LOD, 1.0, 1.0 - (abs(vertex.z) - scene_data_block.data.z_near) / (scene_data_block.data.z_far - scene_data_block.data.z_near)); +#ifdef USE_RADIANCE_CUBEMAP_ARRAY + float lod, blend; + blend = modf(mip_level * MAX_ROUGHNESS_LOD, lod); + sky_fog_color = texture(samplerCubeArray(radiance_cubemap, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), vec4(cube_view, lod)).rgb; + sky_fog_color = mix(sky_fog_color, texture(samplerCubeArray(radiance_cubemap, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), vec4(cube_view, lod + 1)).rgb, blend); +#else + sky_fog_color = textureLod(samplerCube(radiance_cubemap, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), cube_view, mip_level * MAX_ROUGHNESS_LOD).rgb; +#endif //USE_RADIANCE_CUBEMAP_ARRAY + fog_color = mix(fog_color, sky_fog_color, scene_data_block.data.fog_aerial_perspective); + } + + if (scene_data_block.data.fog_sun_scatter > 0.001) { + vec4 sun_scatter = vec4(0.0); + float sun_total = 0.0; + vec3 view = normalize(vertex); + + for (uint i = 0; i < scene_data_block.data.directional_light_count; i++) { + vec3 light_color = directional_lights.data[i].color * directional_lights.data[i].energy; + float light_amount = pow(max(dot(view, directional_lights.data[i].direction), 0.0), 8.0); + fog_color += light_color * light_amount * scene_data_block.data.fog_sun_scatter; + } + } + + float fog_amount = 1.0 - exp(min(0.0, -length(vertex) * scene_data_block.data.fog_density)); + + if (abs(scene_data_block.data.fog_height_density) >= 0.0001) { + float y = (scene_data_block.data.inv_view_matrix * vec4(vertex, 1.0)).y; + + float y_dist = y - scene_data_block.data.fog_height; + + float vfog_amount = 1.0 - exp(min(0.0, y_dist * scene_data_block.data.fog_height_density)); + + fog_amount = max(vfog_amount, fog_amount); + } + + return vec4(fog_color, fog_amount); +} + +#endif //!MODE_RENDER DEPTH + +#define scene_data scene_data_block.data + +void main() { +#ifdef MODE_DUAL_PARABOLOID + + if (dp_clip > 0.0) + discard; +#endif + + //lay out everything, whatever is unused is optimized away anyway + vec3 vertex = vertex_interp; +#ifdef USE_MULTIVIEW + vec3 view = -normalize(vertex_interp - scene_data.eye_offset[ViewIndex].xyz); +#else + vec3 view = -normalize(vertex_interp); +#endif + vec3 albedo = vec3(1.0); + vec3 backlight = vec3(0.0); + vec4 transmittance_color = vec4(0.0); + float transmittance_depth = 0.0; + float transmittance_boost = 0.0; + float metallic = 0.0; + float specular = 0.5; + vec3 emission = vec3(0.0); + float roughness = 1.0; + float rim = 0.0; + float rim_tint = 0.0; + float clearcoat = 0.0; + float clearcoat_roughness = 0.0; + float anisotropy = 0.0; + vec2 anisotropy_flow = vec2(1.0, 0.0); + vec4 fog = vec4(0.0); +#if defined(CUSTOM_RADIANCE_USED) + vec4 custom_radiance = vec4(0.0); +#endif +#if defined(CUSTOM_IRRADIANCE_USED) + vec4 custom_irradiance = vec4(0.0); +#endif + + float ao = 1.0; + float ao_light_affect = 0.0; + + float alpha = 1.0; + +#if defined(TANGENT_USED) || defined(NORMAL_MAP_USED) || defined(LIGHT_ANISOTROPY_USED) + vec3 binormal = normalize(binormal_interp); + vec3 tangent = normalize(tangent_interp); +#else + vec3 binormal = vec3(0.0); + vec3 tangent = vec3(0.0); +#endif + +#ifdef NORMAL_USED + vec3 normal = normalize(normal_interp); + +#if defined(DO_SIDE_CHECK) + if (!gl_FrontFacing) { + normal = -normal; + } +#endif + +#endif //NORMAL_USED + +#ifdef UV_USED + vec2 uv = uv_interp; +#endif + +#if defined(UV2_USED) || defined(USE_LIGHTMAP) + vec2 uv2 = uv2_interp; +#endif + +#if defined(COLOR_USED) + vec4 color = color_interp; +#endif + +#if defined(NORMAL_MAP_USED) + + vec3 normal_map = vec3(0.5); +#endif + + float normal_map_depth = 1.0; + + vec2 screen_uv = gl_FragCoord.xy * scene_data.screen_pixel_size; + + float sss_strength = 0.0; + +#ifdef ALPHA_SCISSOR_USED + float alpha_scissor_threshold = 1.0; +#endif // ALPHA_SCISSOR_USED + +#ifdef ALPHA_HASH_USED + float alpha_hash_scale = 1.0; +#endif // ALPHA_HASH_USED + +#ifdef ALPHA_ANTIALIASING_EDGE_USED + float alpha_antialiasing_edge = 0.0; + vec2 alpha_texture_coordinate = vec2(0.0, 0.0); +#endif // ALPHA_ANTIALIASING_EDGE_USED + + { +#CODE : FRAGMENT + } + +#ifdef LIGHT_TRANSMITTANCE_USED +#ifdef SSS_MODE_SKIN + transmittance_color.a = sss_strength; +#else + transmittance_color.a *= sss_strength; +#endif +#endif + +#ifndef USE_SHADOW_TO_OPACITY + +#ifdef ALPHA_SCISSOR_USED + if (alpha < alpha_scissor_threshold) { + discard; + } +#endif // ALPHA_SCISSOR_USED + +// alpha hash can be used in unison with alpha antialiasing +#ifdef ALPHA_HASH_USED + if (alpha < compute_alpha_hash_threshold(vertex, alpha_hash_scale)) { + discard; + } +#endif // ALPHA_HASH_USED + +// If we are not edge antialiasing, we need to remove the output alpha channel from scissor and hash +#if (defined(ALPHA_SCISSOR_USED) || defined(ALPHA_HASH_USED)) && !defined(ALPHA_ANTIALIASING_EDGE_USED) + alpha = 1.0; +#endif + +#ifdef ALPHA_ANTIALIASING_EDGE_USED +// If alpha scissor is used, we must further the edge threshold, otherwise we won't get any edge feather +#ifdef ALPHA_SCISSOR_USED + alpha_antialiasing_edge = clamp(alpha_scissor_threshold + alpha_antialiasing_edge, 0.0, 1.0); +#endif + alpha = compute_alpha_antialiasing_edge(alpha, alpha_texture_coordinate, alpha_antialiasing_edge); +#endif // ALPHA_ANTIALIASING_EDGE_USED + +#ifdef USE_OPAQUE_PREPASS + if (alpha < scene_data.opaque_prepass_threshold) { + discard; + } +#endif // USE_OPAQUE_PREPASS + +#endif // !USE_SHADOW_TO_OPACITY + +#ifdef NORMAL_MAP_USED + + normal_map.xy = normal_map.xy * 2.0 - 1.0; + normal_map.z = sqrt(max(0.0, 1.0 - dot(normal_map.xy, normal_map.xy))); //always ignore Z, as it can be RG packed, Z may be pos/neg, etc. + + normal = normalize(mix(normal, tangent * normal_map.x + binormal * normal_map.y + normal * normal_map.z, normal_map_depth)); + +#endif + +#ifdef LIGHT_ANISOTROPY_USED + + if (anisotropy > 0.01) { + //rotation matrix + mat3 rot = mat3(tangent, binormal, normal); + //make local to space + tangent = normalize(rot * vec3(anisotropy_flow.x, anisotropy_flow.y, 0.0)); + binormal = normalize(rot * vec3(-anisotropy_flow.y, anisotropy_flow.x, 0.0)); + } + +#endif + +#ifdef ENABLE_CLIP_ALPHA + if (albedo.a < 0.99) { + //used for doublepass and shadowmapping + discard; + } +#endif + + /////////////////////// FOG ////////////////////// +#ifndef MODE_RENDER_DEPTH + +#ifndef CUSTOM_FOG_USED + // fog must be processed as early as possible and then packed. + // to maximize VGPR usage + // Draw "fixed" fog before volumetric fog to ensure volumetric fog can appear in front of the sky. + + if (!sc_disable_fog && scene_data.fog_enabled) { + fog = fog_process(vertex); + } + +#endif //!CUSTOM_FOG_USED + + uint fog_rg = packHalf2x16(fog.rg); + uint fog_ba = packHalf2x16(fog.ba); + +#endif //!MODE_RENDER_DEPTH + + /////////////////////// DECALS //////////////////////////////// + +#ifndef MODE_RENDER_DEPTH + + vec3 vertex_ddx = dFdx(vertex); + vec3 vertex_ddy = dFdy(vertex); + + if (!sc_disable_decals) { //Decals + // must implement + + uint decal_indices = draw_call.decals.x; + for (uint i = 0; i < 8; i++) { + uint decal_index = decal_indices & 0xFF; + if (i == 4) { + decal_indices = draw_call.decals.y; + } else { + decal_indices = decal_indices >> 8; + } + + if (decal_index == 0xFF) { + break; + } + + vec3 uv_local = (decals.data[decal_index].xform * vec4(vertex, 1.0)).xyz; + if (any(lessThan(uv_local, vec3(0.0, -1.0, 0.0))) || any(greaterThan(uv_local, vec3(1.0)))) { + continue; //out of decal + } + + float fade = pow(1.0 - (uv_local.y > 0.0 ? uv_local.y : -uv_local.y), uv_local.y > 0.0 ? decals.data[decal_index].upper_fade : decals.data[decal_index].lower_fade); + + if (decals.data[decal_index].normal_fade > 0.0) { + fade *= smoothstep(decals.data[decal_index].normal_fade, 1.0, dot(normal_interp, decals.data[decal_index].normal) * 0.5 + 0.5); + } + + //we need ddx/ddy for mipmaps, so simulate them + vec2 ddx = (decals.data[decal_index].xform * vec4(vertex_ddx, 0.0)).xz; + vec2 ddy = (decals.data[decal_index].xform * vec4(vertex_ddy, 0.0)).xz; + + if (decals.data[decal_index].albedo_rect != vec4(0.0)) { + //has albedo + vec4 decal_albedo; + if (sc_decal_use_mipmaps) { + decal_albedo = textureGrad(sampler2D(decal_atlas_srgb, decal_sampler), uv_local.xz * decals.data[decal_index].albedo_rect.zw + decals.data[decal_index].albedo_rect.xy, ddx * decals.data[decal_index].albedo_rect.zw, ddy * decals.data[decal_index].albedo_rect.zw); + } else { + decal_albedo = textureLod(sampler2D(decal_atlas_srgb, decal_sampler), uv_local.xz * decals.data[decal_index].albedo_rect.zw + decals.data[decal_index].albedo_rect.xy, 0.0); + } + decal_albedo *= decals.data[decal_index].modulate; + decal_albedo.a *= fade; + albedo = mix(albedo, decal_albedo.rgb, decal_albedo.a * decals.data[decal_index].albedo_mix); + + if (decals.data[decal_index].normal_rect != vec4(0.0)) { + vec3 decal_normal; + if (sc_decal_use_mipmaps) { + decal_normal = textureGrad(sampler2D(decal_atlas, decal_sampler), uv_local.xz * decals.data[decal_index].normal_rect.zw + decals.data[decal_index].normal_rect.xy, ddx * decals.data[decal_index].normal_rect.zw, ddy * decals.data[decal_index].normal_rect.zw).xyz; + } else { + decal_normal = textureLod(sampler2D(decal_atlas, decal_sampler), uv_local.xz * decals.data[decal_index].normal_rect.zw + decals.data[decal_index].normal_rect.xy, 0.0).xyz; + } + decal_normal.xy = decal_normal.xy * vec2(2.0, -2.0) - vec2(1.0, -1.0); //users prefer flipped y normal maps in most authoring software + decal_normal.z = sqrt(max(0.0, 1.0 - dot(decal_normal.xy, decal_normal.xy))); + //convert to view space, use xzy because y is up + decal_normal = (decals.data[decal_index].normal_xform * decal_normal.xzy).xyz; + + normal = normalize(mix(normal, decal_normal, decal_albedo.a)); + } + + if (decals.data[decal_index].orm_rect != vec4(0.0)) { + vec3 decal_orm; + if (sc_decal_use_mipmaps) { + decal_orm = textureGrad(sampler2D(decal_atlas, decal_sampler), uv_local.xz * decals.data[decal_index].orm_rect.zw + decals.data[decal_index].orm_rect.xy, ddx * decals.data[decal_index].orm_rect.zw, ddy * decals.data[decal_index].orm_rect.zw).xyz; + } else { + decal_orm = textureLod(sampler2D(decal_atlas, decal_sampler), uv_local.xz * decals.data[decal_index].orm_rect.zw + decals.data[decal_index].orm_rect.xy, 0.0).xyz; + } + ao = mix(ao, decal_orm.r, decal_albedo.a); + roughness = mix(roughness, decal_orm.g, decal_albedo.a); + metallic = mix(metallic, decal_orm.b, decal_albedo.a); + } + } + + 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; + } 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; + } + } + } + } //Decals +#endif //!MODE_RENDER_DEPTH + + /////////////////////// LIGHTING ////////////////////////////// + +#ifdef NORMAL_USED + if (scene_data.roughness_limiter_enabled) { + //https://www.jp.square-enix.com/tech/library/pdf/ImprovedGeometricSpecularAA.pdf + float roughness2 = roughness * roughness; + vec3 dndu = dFdx(normal), dndv = dFdy(normal); + float variance = scene_data.roughness_limiter_amount * (dot(dndu, dndu) + dot(dndv, dndv)); + float kernelRoughness2 = min(2.0 * variance, scene_data.roughness_limiter_limit); //limit effect + float filteredRoughness2 = min(1.0, roughness2 + kernelRoughness2); + roughness = sqrt(filteredRoughness2); + } +#endif // NORMAL_USED + //apply energy conservation + + vec3 specular_light = vec3(0.0, 0.0, 0.0); + vec3 diffuse_light = vec3(0.0, 0.0, 0.0); + vec3 ambient_light = vec3(0.0, 0.0, 0.0); + +#if !defined(MODE_RENDER_DEPTH) && !defined(MODE_UNSHADED) + + if (scene_data.use_reflection_cubemap) { +#ifdef LIGHT_ANISOTROPY_USED + // https://google.github.io/filament/Filament.html#lighting/imagebasedlights/anisotropy + vec3 anisotropic_direction = anisotropy >= 0.0 ? binormal : tangent; + vec3 anisotropic_tangent = cross(anisotropic_direction, view); + vec3 anisotropic_normal = cross(anisotropic_tangent, anisotropic_direction); + vec3 bent_normal = normalize(mix(normal, anisotropic_normal, abs(anisotropy) * clamp(5.0 * roughness, 0.0, 1.0))); + vec3 ref_vec = reflect(-view, bent_normal); +#else + vec3 ref_vec = reflect(-view, normal); +#endif + float horizon = min(1.0 + dot(ref_vec, normal), 1.0); + ref_vec = scene_data.radiance_inverse_xform * ref_vec; +#ifdef USE_RADIANCE_CUBEMAP_ARRAY + + float lod, blend; + blend = modf(roughness * MAX_ROUGHNESS_LOD, lod); + specular_light = texture(samplerCubeArray(radiance_cubemap, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), vec4(ref_vec, lod)).rgb; + specular_light = mix(specular_light, texture(samplerCubeArray(radiance_cubemap, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), vec4(ref_vec, lod + 1)).rgb, blend); + +#else // USE_RADIANCE_CUBEMAP_ARRAY + 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 *= horizon * horizon; + specular_light *= scene_data.ambient_light_color_energy.a; + } + +#if defined(CUSTOM_RADIANCE_USED) + specular_light = mix(specular_light, custom_radiance.rgb, custom_radiance.a); +#endif // CUSTOM_RADIANCE_USED + +#ifndef USE_LIGHTMAP + //lightmap overrides everything + if (scene_data.use_ambient_light) { + ambient_light = scene_data.ambient_light_color_energy.rgb; + + if (scene_data.use_ambient_cubemap) { + vec3 ambient_dir = scene_data.radiance_inverse_xform * normal; +#ifdef USE_RADIANCE_CUBEMAP_ARRAY + vec3 cubemap_ambient = texture(samplerCubeArray(radiance_cubemap, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), vec4(ambient_dir, MAX_ROUGHNESS_LOD)).rgb; +#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 + + ambient_light = mix(ambient_light, cubemap_ambient * scene_data.ambient_light_color_energy.a, scene_data.ambient_color_sky_mix); + } + } +#endif // !USE_LIGHTMAP + +#if defined(CUSTOM_IRRADIANCE_USED) + ambient_light = mix(ambient_light, custom_irradiance.rgb, custom_irradiance.a); +#endif // CUSTOM_IRRADIANCE_USED +#ifdef LIGHT_CLEARCOAT_USED + + if (scene_data.use_reflection_cubemap) { + vec3 n = normalize(normal_interp); // We want to use geometric normal, not normal_map + float NoV = max(dot(n, view), 0.0001); + vec3 ref_vec = reflect(-view, n); + // The clear coat layer assumes an IOR of 1.5 (4% reflectance) + float Fc = clearcoat * (0.04 + 0.96 * SchlickFresnel(NoV)); + float attenuation = 1.0 - Fc; + ambient_light *= attenuation; + specular_light *= attenuation; + + float horizon = min(1.0 + dot(ref_vec, normal), 1.0); + ref_vec = scene_data.radiance_inverse_xform * ref_vec; + float roughness_lod = mix(0.001, 0.1, clearcoat_roughness) * MAX_ROUGHNESS_LOD; +#ifdef USE_RADIANCE_CUBEMAP_ARRAY + + float lod, blend; + blend = modf(roughness_lod, lod); + vec3 clearcoat_light = texture(samplerCubeArray(radiance_cubemap, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), vec4(ref_vec, lod)).rgb; + clearcoat_light = mix(clearcoat_light, texture(samplerCubeArray(radiance_cubemap, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), vec4(ref_vec, lod + 1)).rgb, blend); + +#else + vec3 clearcoat_light = textureLod(samplerCube(radiance_cubemap, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), ref_vec, roughness_lod).rgb; + +#endif //USE_RADIANCE_CUBEMAP_ARRAY + specular_light += clearcoat_light * horizon * horizon * Fc * scene_data.ambient_light_color_energy.a; + } +#endif +#endif //!defined(MODE_RENDER_DEPTH) && !defined(MODE_UNSHADED) + + //radiance + +#if !defined(MODE_RENDER_DEPTH) && !defined(MODE_UNSHADED) + +#ifdef USE_LIGHTMAP + + //lightmap + if (bool(draw_call.flags & INSTANCE_FLAGS_USE_LIGHTMAP_CAPTURE)) { //has lightmap capture + uint index = draw_call.gi_offset; + + vec3 wnormal = mat3(scene_data.inv_view_matrix) * normal; + const float c1 = 0.429043; + const float c2 = 0.511664; + const float c3 = 0.743125; + 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); + + } 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); + uint ofs = draw_call.gi_offset & 0xFFFF; + vec3 uvw; + uvw.xy = uv2 * draw_call.lightmap_uv_scale.zw + draw_call.lightmap_uv_scale.xy; + uvw.z = float((draw_call.gi_offset >> 16) & 0xFFFF); + + 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; + vec3 lm_light_l1n1 = textureLod(sampler2DArray(lightmap_textures[ofs], material_samplers[SAMPLER_LINEAR_CLAMP]), uvw + vec3(0.0, 0.0, 1.0), 0.0).rgb; + 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); + + 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; + 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; + } + + } else { + ambient_light += textureLod(sampler2DArray(lightmap_textures[ofs], material_samplers[SAMPLER_LINEAR_CLAMP]), uvw, 0.0).rgb; + } + } + + // No GI nor non low end mode... + +#endif // USE_LIGHTMAP + + // skipping ssao, do we remove ssao totally? + + if (!sc_disable_reflection_probes) { //Reflection probes + 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 reflection_indices = draw_call.reflection_probes.x; + for (uint i = 0; i < 8; i++) { + uint reflection_index = reflection_indices & 0xFF; + if (i == 4) { + reflection_indices = draw_call.reflection_probes.y; + } else { + reflection_indices = reflection_indices >> 8; + } + + 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); + } + + if (reflection_accum.a > 0.0) { + specular_light = reflection_accum.rgb / reflection_accum.a; + } + +#if !defined(USE_LIGHTMAP) + if (ambient_accum.a > 0.0) { + ambient_light = ambient_accum.rgb / ambient_accum.a; + } +#endif + } //Reflection probes + + // finalize ambient light here + ambient_light *= albedo.rgb; + ambient_light *= ao; + + // convert ao to direct light ao + ao = mix(1.0, ao, ao_light_affect); + + //this saves some VGPRs + vec3 f0 = F0(metallic, specular, albedo); + + { +#if defined(DIFFUSE_TOON) + //simplify for toon, as + specular_light *= specular * metallic * albedo * 2.0; +#else + + // scales the specular reflections, needs to be computed before lighting happens, + // but after environment, GI, and reflection probes are added + // Environment brdf approximation (Lazarov 2013) + // see https://www.unrealengine.com/en-US/blog/physically-based-shading-on-mobile + const vec4 c0 = vec4(-1.0, -0.0275, -0.572, 0.022); + const vec4 c1 = vec4(1.0, 0.0425, 1.04, -0.04); + vec4 r = roughness * c0 + c1; + float ndotv = clamp(dot(normal, view), 0.0, 1.0); + 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; +#endif + } + +#endif // !defined(MODE_RENDER_DEPTH) && !defined(MODE_UNSHADED) + +#if !defined(MODE_RENDER_DEPTH) + //this saves some VGPRs + uint orms = packUnorm4x8(vec4(ao, roughness, metallic, specular)); +#endif + +// LIGHTING +#if !defined(MODE_RENDER_DEPTH) && !defined(MODE_UNSHADED) + + if (!sc_disable_directional_lights) { //directional light +#ifndef SHADOWS_DISABLED + // Do shadow and lighting in two passes to reduce register pressure + uint shadow0 = 0; + uint shadow1 = 0; + + for (uint i = 0; i < 8; i++) { + if (i >= scene_data.directional_light_count) { + break; + } + + if (!bool(directional_lights.data[i].mask & draw_call.layer_mask)) { + continue; //not masked + } + + float shadow = 1.0; + + // Directional light shadow code is basically the same as forward clustered at this point in time minus `LIGHT_TRANSMITTANCE_USED` support. + // Not sure if there is a reason to change this seeing directional lights are part of our global data + // Should think about whether we may want to move this code into an include file or function?? + +#ifdef USE_SOFT_SHADOWS + //version with soft shadows, more expensive + if (directional_lights.data[i].shadow_enabled) { + float depth_z = -vertex.z; + + vec4 pssm_coord; + vec3 light_dir = directional_lights.data[i].direction; + +#define BIAS_FUNC(m_var, m_idx) \ + m_var.xyz += light_dir * directional_lights.data[i].shadow_bias[m_idx]; \ + vec3 normal_bias = normalize(normal_interp) * (1.0 - max(0.0, dot(light_dir, -normalize(normal_interp)))) * directional_lights.data[i].shadow_normal_bias[m_idx]; \ + normal_bias -= light_dir * dot(light_dir, normal_bias); \ + m_var.xyz += normal_bias; + + if (depth_z < directional_lights.data[i].shadow_split_offsets.x) { + vec4 v = vec4(vertex, 1.0); + + BIAS_FUNC(v, 0) + + pssm_coord = (directional_lights.data[i].shadow_matrix1 * v); + pssm_coord /= pssm_coord.w; + + if (directional_lights.data[i].softshadow_angle > 0) { + float range_pos = dot(directional_lights.data[i].direction, v.xyz); + float range_begin = directional_lights.data[i].shadow_range_begin.x; + float test_radius = (range_pos - range_begin) * directional_lights.data[i].softshadow_angle; + vec2 tex_scale = directional_lights.data[i].uv_scale1 * test_radius; + shadow = sample_directional_soft_shadow(directional_shadow_atlas, pssm_coord.xyz, tex_scale * directional_lights.data[i].soft_shadow_scale); + } else { + shadow = sample_directional_pcf_shadow(directional_shadow_atlas, scene_data.directional_shadow_pixel_size * directional_lights.data[i].soft_shadow_scale, pssm_coord); + } + } else if (depth_z < directional_lights.data[i].shadow_split_offsets.y) { + vec4 v = vec4(vertex, 1.0); + + BIAS_FUNC(v, 1) + + pssm_coord = (directional_lights.data[i].shadow_matrix2 * v); + pssm_coord /= pssm_coord.w; + + if (directional_lights.data[i].softshadow_angle > 0) { + float range_pos = dot(directional_lights.data[i].direction, v.xyz); + float range_begin = directional_lights.data[i].shadow_range_begin.y; + float test_radius = (range_pos - range_begin) * directional_lights.data[i].softshadow_angle; + vec2 tex_scale = directional_lights.data[i].uv_scale2 * test_radius; + shadow = sample_directional_soft_shadow(directional_shadow_atlas, pssm_coord.xyz, tex_scale * directional_lights.data[i].soft_shadow_scale); + } else { + shadow = sample_directional_pcf_shadow(directional_shadow_atlas, scene_data.directional_shadow_pixel_size * directional_lights.data[i].soft_shadow_scale, pssm_coord); + } + } else if (depth_z < directional_lights.data[i].shadow_split_offsets.z) { + vec4 v = vec4(vertex, 1.0); + + BIAS_FUNC(v, 2) + + pssm_coord = (directional_lights.data[i].shadow_matrix3 * v); + pssm_coord /= pssm_coord.w; + + if (directional_lights.data[i].softshadow_angle > 0) { + float range_pos = dot(directional_lights.data[i].direction, v.xyz); + float range_begin = directional_lights.data[i].shadow_range_begin.z; + float test_radius = (range_pos - range_begin) * directional_lights.data[i].softshadow_angle; + vec2 tex_scale = directional_lights.data[i].uv_scale3 * test_radius; + shadow = sample_directional_soft_shadow(directional_shadow_atlas, pssm_coord.xyz, tex_scale * directional_lights.data[i].soft_shadow_scale); + } else { + shadow = sample_directional_pcf_shadow(directional_shadow_atlas, scene_data.directional_shadow_pixel_size * directional_lights.data[i].soft_shadow_scale, pssm_coord); + } + } else { + vec4 v = vec4(vertex, 1.0); + + BIAS_FUNC(v, 3) + + pssm_coord = (directional_lights.data[i].shadow_matrix4 * v); + pssm_coord /= pssm_coord.w; + + if (directional_lights.data[i].softshadow_angle > 0) { + float range_pos = dot(directional_lights.data[i].direction, v.xyz); + float range_begin = directional_lights.data[i].shadow_range_begin.w; + float test_radius = (range_pos - range_begin) * directional_lights.data[i].softshadow_angle; + vec2 tex_scale = directional_lights.data[i].uv_scale4 * test_radius; + shadow = sample_directional_soft_shadow(directional_shadow_atlas, pssm_coord.xyz, tex_scale * directional_lights.data[i].soft_shadow_scale); + } else { + shadow = sample_directional_pcf_shadow(directional_shadow_atlas, scene_data.directional_shadow_pixel_size * directional_lights.data[i].soft_shadow_scale, pssm_coord); + } + } + + if (directional_lights.data[i].blend_splits) { + float pssm_blend; + float shadow2; + + if (depth_z < directional_lights.data[i].shadow_split_offsets.x) { + vec4 v = vec4(vertex, 1.0); + BIAS_FUNC(v, 1) + pssm_coord = (directional_lights.data[i].shadow_matrix2 * v); + pssm_coord /= pssm_coord.w; + + if (directional_lights.data[i].softshadow_angle > 0) { + float range_pos = dot(directional_lights.data[i].direction, v.xyz); + float range_begin = directional_lights.data[i].shadow_range_begin.y; + float test_radius = (range_pos - range_begin) * directional_lights.data[i].softshadow_angle; + vec2 tex_scale = directional_lights.data[i].uv_scale2 * test_radius; + shadow2 = sample_directional_soft_shadow(directional_shadow_atlas, pssm_coord.xyz, tex_scale * directional_lights.data[i].soft_shadow_scale); + } else { + shadow2 = sample_directional_pcf_shadow(directional_shadow_atlas, scene_data.directional_shadow_pixel_size * directional_lights.data[i].soft_shadow_scale, pssm_coord); + } + + pssm_blend = smoothstep(0.0, directional_lights.data[i].shadow_split_offsets.x, depth_z); + } else if (depth_z < directional_lights.data[i].shadow_split_offsets.y) { + vec4 v = vec4(vertex, 1.0); + BIAS_FUNC(v, 2) + pssm_coord = (directional_lights.data[i].shadow_matrix3 * v); + pssm_coord /= pssm_coord.w; + + if (directional_lights.data[i].softshadow_angle > 0) { + float range_pos = dot(directional_lights.data[i].direction, v.xyz); + float range_begin = directional_lights.data[i].shadow_range_begin.z; + float test_radius = (range_pos - range_begin) * directional_lights.data[i].softshadow_angle; + vec2 tex_scale = directional_lights.data[i].uv_scale3 * test_radius; + shadow2 = sample_directional_soft_shadow(directional_shadow_atlas, pssm_coord.xyz, tex_scale * directional_lights.data[i].soft_shadow_scale); + } else { + shadow2 = sample_directional_pcf_shadow(directional_shadow_atlas, scene_data.directional_shadow_pixel_size * directional_lights.data[i].soft_shadow_scale, pssm_coord); + } + + pssm_blend = smoothstep(directional_lights.data[i].shadow_split_offsets.x, directional_lights.data[i].shadow_split_offsets.y, depth_z); + } else if (depth_z < directional_lights.data[i].shadow_split_offsets.z) { + vec4 v = vec4(vertex, 1.0); + BIAS_FUNC(v, 3) + pssm_coord = (directional_lights.data[i].shadow_matrix4 * v); + pssm_coord /= pssm_coord.w; + if (directional_lights.data[i].softshadow_angle > 0) { + float range_pos = dot(directional_lights.data[i].direction, v.xyz); + float range_begin = directional_lights.data[i].shadow_range_begin.w; + float test_radius = (range_pos - range_begin) * directional_lights.data[i].softshadow_angle; + vec2 tex_scale = directional_lights.data[i].uv_scale4 * test_radius; + shadow2 = sample_directional_soft_shadow(directional_shadow_atlas, pssm_coord.xyz, tex_scale * directional_lights.data[i].soft_shadow_scale); + } else { + shadow2 = sample_directional_pcf_shadow(directional_shadow_atlas, scene_data.directional_shadow_pixel_size * directional_lights.data[i].soft_shadow_scale, pssm_coord); + } + + pssm_blend = smoothstep(directional_lights.data[i].shadow_split_offsets.y, directional_lights.data[i].shadow_split_offsets.z, depth_z); + } else { + pssm_blend = 0.0; //if no blend, same coord will be used (divide by z will result in same value, and already cached) + } + + pssm_blend = sqrt(pssm_blend); + + shadow = mix(shadow, shadow2, pssm_blend); + } + + shadow = mix(shadow, 1.0, smoothstep(directional_lights.data[i].fade_from, directional_lights.data[i].fade_to, vertex.z)); //done with negative values for performance + +#undef BIAS_FUNC + } +#else + // Soft shadow disabled version + + if (directional_lights.data[i].shadow_enabled) { + float depth_z = -vertex.z; + + vec4 pssm_coord; + float blur_factor; + vec3 light_dir = directional_lights.data[i].direction; + vec3 base_normal_bias = normalize(normal_interp) * (1.0 - max(0.0, dot(light_dir, -normalize(normal_interp)))); + +#define BIAS_FUNC(m_var, m_idx) \ + m_var.xyz += light_dir * directional_lights.data[i].shadow_bias[m_idx]; \ + vec3 normal_bias = base_normal_bias * directional_lights.data[i].shadow_normal_bias[m_idx]; \ + normal_bias -= light_dir * dot(light_dir, normal_bias); \ + m_var.xyz += normal_bias; + + if (depth_z < directional_lights.data[i].shadow_split_offsets.x) { + vec4 v = vec4(vertex, 1.0); + + BIAS_FUNC(v, 0) + + pssm_coord = (directional_lights.data[i].shadow_matrix1 * v); + blur_factor = 1.0; + } else if (depth_z < directional_lights.data[i].shadow_split_offsets.y) { + vec4 v = vec4(vertex, 1.0); + + BIAS_FUNC(v, 1) + + pssm_coord = (directional_lights.data[i].shadow_matrix2 * v); + // Adjust shadow blur with reference to the first split to reduce discrepancy between shadow splits. + blur_factor = directional_lights.data[i].shadow_split_offsets.x / directional_lights.data[i].shadow_split_offsets.y; + ; + } else if (depth_z < directional_lights.data[i].shadow_split_offsets.z) { + vec4 v = vec4(vertex, 1.0); + + BIAS_FUNC(v, 2) + + pssm_coord = (directional_lights.data[i].shadow_matrix3 * v); + // Adjust shadow blur with reference to the first split to reduce discrepancy between shadow splits. + blur_factor = directional_lights.data[i].shadow_split_offsets.x / directional_lights.data[i].shadow_split_offsets.z; + } else { + vec4 v = vec4(vertex, 1.0); + + BIAS_FUNC(v, 3) + + pssm_coord = (directional_lights.data[i].shadow_matrix4 * v); + // Adjust shadow blur with reference to the first split to reduce discrepancy between shadow splits. + blur_factor = directional_lights.data[i].shadow_split_offsets.x / directional_lights.data[i].shadow_split_offsets.w; + } + + pssm_coord /= pssm_coord.w; + + shadow = sample_directional_pcf_shadow(directional_shadow_atlas, scene_data.directional_shadow_pixel_size * directional_lights.data[i].soft_shadow_scale * blur_factor, pssm_coord); + + if (directional_lights.data[i].blend_splits) { + float pssm_blend; + float blur_factor2; + + if (depth_z < directional_lights.data[i].shadow_split_offsets.x) { + vec4 v = vec4(vertex, 1.0); + BIAS_FUNC(v, 1) + pssm_coord = (directional_lights.data[i].shadow_matrix2 * v); + pssm_blend = smoothstep(0.0, directional_lights.data[i].shadow_split_offsets.x, depth_z); + // Adjust shadow blur with reference to the first split to reduce discrepancy between shadow splits. + blur_factor2 = directional_lights.data[i].shadow_split_offsets.x / directional_lights.data[i].shadow_split_offsets.y; + } else if (depth_z < directional_lights.data[i].shadow_split_offsets.y) { + vec4 v = vec4(vertex, 1.0); + BIAS_FUNC(v, 2) + pssm_coord = (directional_lights.data[i].shadow_matrix3 * v); + pssm_blend = smoothstep(directional_lights.data[i].shadow_split_offsets.x, directional_lights.data[i].shadow_split_offsets.y, depth_z); + // Adjust shadow blur with reference to the first split to reduce discrepancy between shadow splits. + blur_factor2 = directional_lights.data[i].shadow_split_offsets.x / directional_lights.data[i].shadow_split_offsets.z; + } else if (depth_z < directional_lights.data[i].shadow_split_offsets.z) { + vec4 v = vec4(vertex, 1.0); + BIAS_FUNC(v, 3) + pssm_coord = (directional_lights.data[i].shadow_matrix4 * v); + pssm_blend = smoothstep(directional_lights.data[i].shadow_split_offsets.y, directional_lights.data[i].shadow_split_offsets.z, depth_z); + // Adjust shadow blur with reference to the first split to reduce discrepancy between shadow splits. + blur_factor2 = directional_lights.data[i].shadow_split_offsets.x / directional_lights.data[i].shadow_split_offsets.w; + } else { + pssm_blend = 0.0; //if no blend, same coord will be used (divide by z will result in same value, and already cached) + blur_factor2 = 1.0; + } + + pssm_coord /= pssm_coord.w; + + float shadow2 = sample_directional_pcf_shadow(directional_shadow_atlas, scene_data.directional_shadow_pixel_size * directional_lights.data[i].soft_shadow_scale * blur_factor2, pssm_coord); + shadow = mix(shadow, shadow2, pssm_blend); + } + + shadow = mix(shadow, 1.0, smoothstep(directional_lights.data[i].fade_from, directional_lights.data[i].fade_to, vertex.z)); //done with negative values for performance + +#undef BIAS_FUNC + } +#endif + + if (i < 4) { + shadow0 |= uint(clamp(shadow * 255.0, 0.0, 255.0)) << (i * 8); + } else { + shadow1 |= uint(clamp(shadow * 255.0, 0.0, 255.0)) << ((i - 4) * 8); + } + } + +#endif // SHADOWS_DISABLED + + for (uint i = 0; i < 8; i++) { + if (i >= scene_data.directional_light_count) { + break; + } + + if (!bool(directional_lights.data[i].mask & draw_call.layer_mask)) { + continue; //not masked + } + + // We're not doing light transmittence + + float shadow = 1.0; +#ifndef SHADOWS_DISABLED + if (i < 4) { + shadow = float(shadow0 >> (i * 8) & 0xFF) / 255.0; + } else { + shadow = float(shadow1 >> ((i - 4) * 8) & 0xFF) / 255.0; + } +#endif + blur_shadow(shadow); + + light_compute(normal, directional_lights.data[i].direction, normalize(view), 0.0, directional_lights.data[i].color * directional_lights.data[i].energy, shadow, f0, orms, 1.0, albedo, alpha, +#ifdef LIGHT_BACKLIGHT_USED + backlight, +#endif +/* not supported here +#ifdef LIGHT_TRANSMITTANCE_USED + transmittance_color, + transmittance_depth, + transmittance_boost, + transmittance_z, +#endif +*/ +#ifdef LIGHT_RIM_USED + rim, rim_tint, +#endif +#ifdef LIGHT_CLEARCOAT_USED + clearcoat, clearcoat_roughness, normalize(normal_interp), +#endif +#ifdef LIGHT_ANISOTROPY_USED + binormal, tangent, anisotropy, +#endif +#ifdef USE_SOFT_SHADOW + directional_lights.data[i].size, +#endif + diffuse_light, + specular_light); + } + } //directional light + + if (!sc_disable_omni_lights) { //omni lights + uint light_indices = draw_call.omni_lights.x; + for (uint i = 0; i < 8; i++) { + uint light_index = light_indices & 0xFF; + if (i == 4) { + light_indices = draw_call.omni_lights.y; + } else { + light_indices = light_indices >> 8; + } + + if (light_index == 0xFF) { + break; + } + + float shadow = light_process_omni_shadow(light_index, vertex, normal); + + shadow = blur_shadow(shadow); + + light_process_omni(light_index, vertex, view, normal, vertex_ddx, vertex_ddy, f0, orms, shadow, albedo, alpha, +#ifdef LIGHT_BACKLIGHT_USED + backlight, +#endif +/* +#ifdef LIGHT_TRANSMITTANCE_USED + transmittance_color, + transmittance_depth, + transmittance_boost, +#endif +*/ +#ifdef LIGHT_RIM_USED + rim, + rim_tint, +#endif +#ifdef LIGHT_CLEARCOAT_USED + clearcoat, clearcoat_roughness, normalize(normal_interp), +#endif +#ifdef LIGHT_ANISOTROPY_USED + tangent, + binormal, anisotropy, +#endif + diffuse_light, specular_light); + } + } //omni lights + + if (!sc_disable_spot_lights) { //spot lights + + uint light_indices = draw_call.spot_lights.x; + for (uint i = 0; i < 8; i++) { + uint light_index = light_indices & 0xFF; + if (i == 4) { + light_indices = draw_call.spot_lights.y; + } else { + light_indices = light_indices >> 8; + } + + if (light_index == 0xFF) { + break; + } + + float shadow = light_process_spot_shadow(light_index, vertex, normal); + + shadow = blur_shadow(shadow); + + light_process_spot(light_index, vertex, view, normal, vertex_ddx, vertex_ddy, f0, orms, shadow, albedo, alpha, +#ifdef LIGHT_BACKLIGHT_USED + backlight, +#endif +/* +#ifdef LIGHT_TRANSMITTANCE_USED + transmittance_color, + transmittance_depth, + transmittance_boost, +#endif +*/ +#ifdef LIGHT_RIM_USED + rim, + rim_tint, +#endif +#ifdef LIGHT_CLEARCOAT_USED + clearcoat, clearcoat_roughness, normalize(normal_interp), +#endif +#ifdef LIGHT_ANISOTROPY_USED + tangent, + binormal, anisotropy, +#endif + diffuse_light, specular_light); + } + } //spot lights + +#ifdef USE_SHADOW_TO_OPACITY + alpha = min(alpha, clamp(length(ambient_light), 0.0, 1.0)); + +#if defined(ALPHA_SCISSOR_USED) + if (alpha < alpha_scissor) { + discard; + } +#endif // ALPHA_SCISSOR_USED + +#ifdef USE_OPAQUE_PREPASS + + if (alpha < scene_data.opaque_prepass_threshold) { + discard; + } + +#endif // USE_OPAQUE_PREPASS + +#endif // USE_SHADOW_TO_OPACITY + +#endif //!defined(MODE_RENDER_DEPTH) && !defined(MODE_UNSHADED) + +#ifdef MODE_RENDER_DEPTH + +#ifdef MODE_RENDER_MATERIAL + + albedo_output_buffer.rgb = albedo; + albedo_output_buffer.a = alpha; + + normal_output_buffer.rgb = normal * 0.5 + 0.5; + normal_output_buffer.a = 0.0; + depth_output_buffer.r = -vertex.z; + + orm_output_buffer.r = ao; + orm_output_buffer.g = roughness; + orm_output_buffer.b = metallic; + orm_output_buffer.a = sss_strength; + + emission_output_buffer.rgb = emission; + emission_output_buffer.a = 0.0; +#endif // MODE_RENDER_MATERIAL + +#else // MODE_RENDER_DEPTH + + // multiply by albedo + diffuse_light *= albedo; // ambient must be multiplied by albedo at the end + + // apply direct light AO + ao = unpackUnorm4x8(orms).x; + specular_light *= ao; + diffuse_light *= ao; + + // apply metallic + metallic = unpackUnorm4x8(orms).z; + diffuse_light *= 1.0 - metallic; + ambient_light *= 1.0 - metallic; + + //restore fog + fog = vec4(unpackHalf2x16(fog_rg), unpackHalf2x16(fog_ba)); + +#ifdef MODE_MULTIPLE_RENDER_TARGETS + +#ifdef MODE_UNSHADED + diffuse_buffer = vec4(albedo.rgb, 0.0); + specular_buffer = vec4(0.0); + +#else // MODE_UNSHADED + +#ifdef SSS_MODE_SKIN + sss_strength = -sss_strength; +#endif // SSS_MODE_SKIN + diffuse_buffer = vec4(emission + diffuse_light + ambient_light, sss_strength); + specular_buffer = vec4(specular_light, metallic); +#endif // MODE_UNSHADED + + diffuse_buffer.rgb = mix(diffuse_buffer.rgb, fog.rgb, fog.a); + specular_buffer.rgb = mix(specular_buffer.rgb, vec3(0.0), fog.a); + +#else //MODE_MULTIPLE_RENDER_TARGETS + +#ifdef MODE_UNSHADED + frag_color = vec4(albedo, alpha); +#else // MODE_UNSHADED + frag_color = vec4(emission + ambient_light + diffuse_light + specular_light, alpha); +#endif // MODE_UNSHADED + + // Draw "fixed" fog before volumetric fog to ensure volumetric fog can appear in front of the sky. + frag_color.rgb = mix(frag_color.rgb, fog.rgb, fog.a); + + // On mobile we use a UNORM buffer with 10bpp which results in a range from 0.0 - 1.0 resulting in HDR breaking + // We divide by sc_luminance_multiplier to support a range from 0.0 - 2.0 both increasing precision on bright and darker images + frag_color.rgb = frag_color.rgb / sc_luminance_multiplier; + +#endif //MODE_MULTIPLE_RENDER_TARGETS + +#endif //MODE_RENDER_DEPTH +} diff --git a/servers/rendering/renderer_rd/shaders/scene_forward_mobile_inc.glsl b/servers/rendering/renderer_rd/shaders/scene_forward_mobile_inc.glsl new file mode 100644 index 0000000000..7413d8730a --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/scene_forward_mobile_inc.glsl @@ -0,0 +1,229 @@ +#define M_PI 3.14159265359 +#define MAX_VIEWS 2 + +#if defined(USE_MULTIVIEW) && defined(has_VK_KHR_multiview) +#extension GL_EXT_multiview : enable +#endif + +#include "decal_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 +#define NORMAL_USED +#endif +#endif + +/* don't exceed 128 bytes!! */ +/* put instance data into our push content, not a array */ +layout(push_constant, std430) uniform DrawCall { + highp mat4 transform; // 64 - 64 + uint flags; // 04 - 68 + uint instance_uniforms_ofs; //base offset in global buffer for instance variables // 04 - 72 + uint gi_offset; //GI information when using lightmapping (VCT or lightmap index) // 04 - 76 + uint layer_mask; // 04 - 80 + highp vec4 lightmap_uv_scale; // 16 - 96 doubles as uv_offset when needed + + uvec2 reflection_probes; // 08 - 104 + uvec2 omni_lights; // 08 - 112 + uvec2 spot_lights; // 08 - 120 + uvec2 decals; // 08 - 128 +} +draw_call; + +/* Set 0: Base Pass (never changes) */ + +#include "light_data_inc.glsl" + +#define SAMPLER_NEAREST_CLAMP 0 +#define SAMPLER_LINEAR_CLAMP 1 +#define SAMPLER_NEAREST_WITH_MIPMAPS_CLAMP 2 +#define SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP 3 +#define SAMPLER_NEAREST_WITH_MIPMAPS_ANISOTROPIC_CLAMP 4 +#define SAMPLER_LINEAR_WITH_MIPMAPS_ANISOTROPIC_CLAMP 5 +#define SAMPLER_NEAREST_REPEAT 6 +#define SAMPLER_LINEAR_REPEAT 7 +#define SAMPLER_NEAREST_WITH_MIPMAPS_REPEAT 8 +#define SAMPLER_LINEAR_WITH_MIPMAPS_REPEAT 9 +#define SAMPLER_NEAREST_WITH_MIPMAPS_ANISOTROPIC_REPEAT 10 +#define SAMPLER_LINEAR_WITH_MIPMAPS_ANISOTROPIC_REPEAT 11 + +layout(set = 0, binding = 1) uniform sampler material_samplers[12]; + +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_MULTIMESH (1 << 12) +#define INSTANCE_FLAGS_MULTIMESH_FORMAT_2D (1 << 13) +#define INSTANCE_FLAGS_MULTIMESH_HAS_COLOR (1 << 14) +#define INSTANCE_FLAGS_MULTIMESH_HAS_CUSTOM_DATA (1 << 15) +#define INSTANCE_FLAGS_PARTICLE_TRAIL_SHIFT 16 +//3 bits of stride +#define INSTANCE_FLAGS_PARTICLE_TRAIL_MASK 0xFF + +layout(set = 0, binding = 5, std430) restrict readonly buffer OmniLights { + LightData data[]; +} +omni_lights; + +layout(set = 0, binding = 6, std430) restrict readonly buffer SpotLights { + LightData data[]; +} +spot_lights; + +layout(set = 0, binding = 7, std430) restrict readonly buffer ReflectionProbeData { + ReflectionData data[]; +} +reflections; + +layout(set = 0, binding = 8, std140) uniform DirectionalLights { + DirectionalLightData data[MAX_DIRECTIONAL_LIGHT_DATA_STRUCTS]; +} +directional_lights; + +#define LIGHTMAP_FLAG_USE_DIRECTION 1 +#define LIGHTMAP_FLAG_USE_SPECULAR_DIRECTION 2 + +struct Lightmap { + mediump mat3 normal_xform; +}; + +layout(set = 0, binding = 9, std140) restrict readonly buffer Lightmaps { + Lightmap data[]; +} +lightmaps; + +struct LightmapCapture { + mediump vec4 sh[9]; +}; + +layout(set = 0, binding = 10, std140) restrict readonly buffer LightmapCaptures { + LightmapCapture data[]; +} +lightmap_captures; + +layout(set = 0, binding = 11) uniform mediump texture2D decal_atlas; +layout(set = 0, binding = 12) uniform mediump texture2D decal_atlas_srgb; + +layout(set = 0, binding = 13, std430) restrict readonly buffer Decals { + DecalData data[]; +} +decals; + +layout(set = 0, binding = 14, std430) restrict readonly buffer GlobalVariableData { + highp vec4 data[]; +} +global_variables; + +/* 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; +} +scene_data_block; + +#ifdef USE_RADIANCE_CUBEMAP_ARRAY + +layout(set = 1, binding = 2) uniform mediump textureCubeArray radiance_cubemap; + +#else + +layout(set = 1, binding = 2) uniform mediump textureCube radiance_cubemap; + +#endif + +layout(set = 1, binding = 3) uniform mediump textureCubeArray reflection_atlas; + +layout(set = 1, binding = 4) uniform highp texture2D shadow_atlas; + +layout(set = 1, binding = 5) uniform highp texture2D directional_shadow_atlas; + +// this needs to change to providing just the lightmap we're using.. +layout(set = 1, binding = 6) uniform texture2DArray lightmap_textures[MAX_LIGHTMAP_TEXTURES]; + +layout(set = 1, binding = 9) uniform highp texture2D depth_buffer; +layout(set = 1, binding = 10) uniform mediump texture2D color_buffer; + +/* Set 2 Skeleton & Instancing (can change per item) */ + +layout(set = 2, binding = 0, std430) restrict readonly buffer Transforms { + highp vec4 data[]; +} +transforms; + +/* Set 3 User Material */ diff --git a/servers/rendering/renderer_rd/shaders/screen_space_reflection.glsl b/servers/rendering/renderer_rd/shaders/screen_space_reflection.glsl index 06dc4b13de..a416891ff2 100644 --- a/servers/rendering/renderer_rd/shaders/screen_space_reflection.glsl +++ b/servers/rendering/renderer_rd/shaders/screen_space_reflection.glsl @@ -2,7 +2,7 @@ #version 450 -VERSION_DEFINES +#VERSION_DEFINES layout(local_size_x = 8, local_size_y = 8, local_size_z = 1) in; @@ -15,7 +15,7 @@ layout(r8, set = 1, binding = 1) uniform restrict writeonly image2D blur_radius_ layout(rgba8, set = 2, binding = 0) uniform restrict readonly image2D source_normal_roughness; layout(set = 3, binding = 0) uniform sampler2D source_metallic; -layout(push_constant, binding = 2, std430) uniform Params { +layout(push_constant, std430) uniform Params { vec4 proj_info; ivec2 screen_size; @@ -190,8 +190,7 @@ void main() { } vec2 final_pos; - float grad; - grad = steps_taken / float(params.num_steps); + float grad = (steps_taken + 1.0) / float(params.num_steps); float initial_fade = params.curve_fade_in == 0.0 ? 1.0 : pow(clamp(grad, 0.0, 1.0), params.curve_fade_in); float fade = pow(clamp(1.0 - grad, 0.0, 1.0), params.distance_fade) * initial_fade; final_pos = pos; @@ -223,7 +222,6 @@ void main() { blur_radius = (a * (sqrt(a2 + fh2) - a)) / (4.0f * h); } } - 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 diff --git a/servers/rendering/renderer_rd/shaders/screen_space_reflection_filter.glsl b/servers/rendering/renderer_rd/shaders/screen_space_reflection_filter.glsl index a5afe74cb2..20e1712496 100644 --- a/servers/rendering/renderer_rd/shaders/screen_space_reflection_filter.glsl +++ b/servers/rendering/renderer_rd/shaders/screen_space_reflection_filter.glsl @@ -2,7 +2,7 @@ #version 450 -VERSION_DEFINES +#VERSION_DEFINES layout(local_size_x = 8, local_size_y = 8, local_size_z = 1) in; @@ -16,7 +16,7 @@ layout(r8, set = 2, binding = 1) uniform restrict writeonly image2D dest_radius; #endif layout(r32f, set = 3, binding = 0) uniform restrict readonly image2D source_depth; -layout(push_constant, binding = 2, std430) uniform Params { +layout(push_constant, std430) uniform Params { vec4 proj_info; bool orthogonal; diff --git a/servers/rendering/renderer_rd/shaders/screen_space_reflection_scale.glsl b/servers/rendering/renderer_rd/shaders/screen_space_reflection_scale.glsl index 218605a962..3f537e273a 100644 --- a/servers/rendering/renderer_rd/shaders/screen_space_reflection_scale.glsl +++ b/servers/rendering/renderer_rd/shaders/screen_space_reflection_scale.glsl @@ -2,7 +2,7 @@ #version 450 -VERSION_DEFINES +#VERSION_DEFINES layout(local_size_x = 8, local_size_y = 8, local_size_z = 1) in; @@ -13,7 +13,7 @@ layout(rgba16f, set = 2, binding = 0) uniform restrict writeonly image2D dest_ss layout(r32f, set = 3, binding = 0) uniform restrict writeonly image2D dest_depth; layout(rgba8, set = 3, binding = 1) uniform restrict writeonly image2D dest_normal; -layout(push_constant, binding = 1, std430) uniform Params { +layout(push_constant, std430) uniform Params { ivec2 screen_size; float camera_z_near; float camera_z_far; @@ -36,12 +36,12 @@ void main() { float divisor = 0.0; vec4 color; float depth; - vec3 normal; + vec4 normal; if (params.filtered) { color = vec4(0.0); depth = 0.0; - normal = vec3(0.0); + normal = vec4(0.0); for (int i = 0; i < 4; i++) { ivec2 ofs = ssC << 1; @@ -53,7 +53,9 @@ void main() { } color += texelFetch(source_ssr, ofs, 0); float d = texelFetch(source_depth, ofs, 0).r; - normal += texelFetch(source_normal, ofs, 0).xyz * 2.0 - 1.0; + vec4 nr = texelFetch(source_normal, ofs, 0); + normal.xyz += nr.xyz * 2.0 - 1.0; + normal.w += nr.w; d = d * 2.0 - 1.0; if (params.orthogonal) { @@ -66,11 +68,12 @@ void main() { color /= 4.0; depth /= 4.0; - normal = normalize(normal / 4.0) * 0.5 + 0.5; + normal.xyz = normalize(normal.xyz / 4.0) * 0.5 + 0.5; + normal.w /= 4.0; } else { color = texelFetch(source_ssr, ssC << 1, 0); depth = texelFetch(source_depth, ssC << 1, 0).r; - normal = texelFetch(source_normal, ssC << 1, 0).xyz; + normal = texelFetch(source_normal, ssC << 1, 0); depth = depth * 2.0 - 1.0; if (params.orthogonal) { @@ -83,5 +86,5 @@ void main() { imageStore(dest_ssr, ssC, color); imageStore(dest_depth, ssC, vec4(depth)); - imageStore(dest_normal, ssC, vec4(normal, 0.0)); + imageStore(dest_normal, ssC, normal); } diff --git a/servers/rendering/renderer_rd/shaders/sdfgi_debug.glsl b/servers/rendering/renderer_rd/shaders/sdfgi_debug.glsl index e4c3f3a84b..802a410825 100644 --- a/servers/rendering/renderer_rd/shaders/sdfgi_debug.glsl +++ b/servers/rendering/renderer_rd/shaders/sdfgi_debug.glsl @@ -2,7 +2,7 @@ #version 450 -VERSION_DEFINES +#VERSION_DEFINES layout(local_size_x = 8, local_size_y = 8, local_size_z = 1) in; @@ -32,7 +32,7 @@ layout(rgba16f, set = 0, binding = 10) uniform restrict writeonly image2D screen layout(set = 0, binding = 11) uniform texture2DArray lightprobe_texture; -layout(push_constant, binding = 0, std430) uniform Params { +layout(push_constant, std430) uniform Params { vec3 grid_size; uint max_cascades; diff --git a/servers/rendering/renderer_rd/shaders/sdfgi_debug_probes.glsl b/servers/rendering/renderer_rd/shaders/sdfgi_debug_probes.glsl index 08da283dad..e0be0bca12 100644 --- a/servers/rendering/renderer_rd/shaders/sdfgi_debug_probes.glsl +++ b/servers/rendering/renderer_rd/shaders/sdfgi_debug_probes.glsl @@ -2,11 +2,11 @@ #version 450 -VERSION_DEFINES +#VERSION_DEFINES #define MAX_CASCADES 8 -layout(push_constant, binding = 0, std430) uniform Params { +layout(push_constant, std430) uniform Params { mat4 projection; uint band_power; @@ -24,7 +24,7 @@ layout(push_constant, binding = 0, std430) uniform Params { } params; -// http://in4k.untergrund.net/html_articles/hugi_27_-_coding_corner_polaris_sphere_tessellation_101.htm +// https://in4k.untergrund.net/html_articles/hugi_27_-_coding_corner_polaris_sphere_tessellation_101.htm vec3 get_sphere_vertex(uint p_vertex_id) { float x_angle = float(p_vertex_id & 1u) + (p_vertex_id >> params.band_power); @@ -153,14 +153,14 @@ void main() { #version 450 -VERSION_DEFINES +#VERSION_DEFINES layout(location = 0) out vec4 frag_color; layout(set = 0, binding = 2) uniform texture2DArray lightprobe_texture; layout(set = 0, binding = 3) uniform sampler linear_sampler; -layout(push_constant, binding = 0, std430) uniform Params { +layout(push_constant, std430) uniform Params { mat4 projection; uint band_power; diff --git a/servers/rendering/renderer_rd/shaders/sdfgi_direct_light.glsl b/servers/rendering/renderer_rd/shaders/sdfgi_direct_light.glsl index dc7238abed..b95fad650e 100644 --- a/servers/rendering/renderer_rd/shaders/sdfgi_direct_light.glsl +++ b/servers/rendering/renderer_rd/shaders/sdfgi_direct_light.glsl @@ -2,7 +2,7 @@ #version 450 -VERSION_DEFINES +#VERSION_DEFINES layout(local_size_x = 64, local_size_y = 1, local_size_z = 1) in; @@ -20,10 +20,10 @@ layout(set = 0, binding = 3, std430) restrict readonly buffer DispatchData { dispatch_data; struct ProcessVoxel { - uint position; //xyz 7 bit packed, extra 11 bits for neigbours - uint albedo; //rgb bits 0-15 albedo, bits 16-21 are normal bits (set if geometry exists toward that side), extra 11 bits for neibhbours - uint light; //rgbe8985 encoded total saved light, extra 2 bits for neighbours - uint light_aniso; //55555 light anisotropy, extra 2 bits for neighbours + uint position; // xyz 7 bit packed, extra 11 bits for neighbors. + uint albedo; // rgb bits 0-15 albedo, bits 16-21 are normal bits (set if geometry exists toward that side), extra 11 bits for neighbors. + uint light; // rgbe8985 encoded total saved light, extra 2 bits for neighbors. + uint light_aniso; // 55555 light anisotropy, extra 2 bits for neighbors. //total neighbours: 26 }; @@ -70,8 +70,6 @@ struct Light { float cos_spot_angle; float inv_spot_attenuation; float radius; - - vec4 shadow_color; }; layout(set = 0, binding = 9, std140) buffer restrict readonly Lights { @@ -82,7 +80,7 @@ lights; layout(set = 0, binding = 10) uniform texture2DArray lightprobe_texture; layout(set = 0, binding = 11) uniform texture3D occlusion_texture; -layout(push_constant, binding = 0, std430) uniform Params { +layout(push_constant, std430) uniform Params { vec3 grid_size; uint max_cascades; diff --git a/servers/rendering/renderer_rd/shaders/sdfgi_fields.glsl b/servers/rendering/renderer_rd/shaders/sdfgi_fields.glsl deleted file mode 100644 index 69d8824d8a..0000000000 --- a/servers/rendering/renderer_rd/shaders/sdfgi_fields.glsl +++ /dev/null @@ -1,182 +0,0 @@ -/* clang-format off */ -[compute] - -#version 450 - -VERSION_DEFINES - -layout(local_size_x = OCT_RES, local_size_y = OCT_RES, local_size_z = 1) in; - -/* clang-format on */ - -#define MAX_CASCADES 8 - -layout(rgba16f, set = 0, binding = 1) uniform restrict image2DArray irradiance_texture; -layout(rg16f, set = 0, binding = 2) uniform restrict image2DArray depth_texture; - -layout(rgba32ui, set = 0, binding = 3) uniform restrict uimage2DArray irradiance_history_texture; -layout(rg32ui, set = 0, binding = 4) uniform restrict uimage2DArray depth_history_texture; - -struct CascadeData { - vec3 offset; //offset of (0,0,0) in world coordinates - float to_cell; // 1/bounds * grid_size -}; - -layout(set = 0, binding = 5, std140) uniform Cascades { - CascadeData data[MAX_CASCADES]; -} -cascades; - -#define DEPTH_HISTORY_BITS 24 -#define IRRADIANCE_HISTORY_BITS 16 - -layout(push_constant, binding = 0, std430) uniform Params { - vec3 grid_size; - uint max_cascades; - - uint probe_axis_size; - uint cascade; - uint history_size; - uint pad0; - - ivec3 scroll; //scroll in probes - uint pad1; -} -params; - -void main() { - ivec2 local = ivec2(gl_LocalInvocationID.xy); - ivec2 probe = ivec2(gl_WorkGroupID.xy); - - ivec3 probe_cell; - probe_cell.x = probe.x % int(params.probe_axis_size); - probe_cell.y = probe.y; - probe_cell.z = probe.x / int(params.probe_axis_size); - -#ifdef MODE_SCROLL_BEGIN - - ivec3 read_cell = probe_cell - params.scroll; - - uint src_layer = (params.history_size + 1) * params.cascade; - uint dst_layer = (params.history_size + 1) * params.max_cascades; - - for (uint i = 0; i <= params.history_size; i++) { - ivec3 write_pos = ivec3(probe * OCT_RES + local, int(i)); - - if (any(lessThan(read_pos, ivec3(0))) || any(greaterThanEqual(read_pos, ivec3(params.probe_axis_size)))) { - // nowhere to read from for scrolling, try finding the value from upper probes - -#ifdef MODE_IRRADIANCE - imageStore(irradiance_history_texture, write_pos, uvec4(0)); -#endif -#ifdef MODE_DEPTH - imageStore(depth_history_texture, write_pos, uvec4(0)); -#endif - } else { - ivec3 read_pos; - read_pos.xy = read_cell.xy; - read_pos.x += read_cell.z * params.probe_axis_size; - read_pos.xy = read_pos.xy * OCT_RES + local; - read_pos.z = int(i); - -#ifdef MODE_IRRADIANCE - uvec4 value = imageLoad(irradiance_history_texture, read_pos); - imageStore(irradiance_history_texture, write_pos, value); -#endif -#ifdef MODE_DEPTH - uvec2 value = imageLoad(depth_history_texture, read_pos); - imageStore(depth_history_texture, write_pos, value); -#endif - } - } - -#endif // MODE_SCROLL_BEGIN - -#ifdef MODE_SCROLL_END - - uint src_layer = (params.history_size + 1) * params.max_cascades; - uint dst_layer = (params.history_size + 1) * params.cascade; - - for (uint i = 0; i <= params.history_size; i++) { - ivec3 pos = ivec3(probe * OCT_RES + local, int(i)); - -#ifdef MODE_IRRADIANCE - uvec4 value = imageLoad(irradiance_history_texture, read_pos); - imageStore(irradiance_history_texture, write_pos, value); -#endif -#ifdef MODE_DEPTH - uvec2 value = imageLoad(depth_history_texture, read_pos); - imageStore(depth_history_texture, write_pos, value); -#endif - } - -#endif //MODE_SCROLL_END - -#ifdef MODE_STORE - - uint src_layer = (params.history_size + 1) * params.cascade + params.history_size; - ivec3 read_pos = ivec3(probe * OCT_RES + local, int(src_layer)); - - ivec3 write_pos = ivec3(probe * (OCT_RES + 2) + ivec2(1), int(params.cascade)); - - ivec3 copy_to[4] = ivec3[](write_pos, ivec3(-2, -2, -2), ivec3(-2, -2, -2), ivec3(-2, -2, -2)); - -#ifdef MODE_IRRADIANCE - uvec4 average = imageLoad(irradiance_history_texture, read_pos); - vec4 light_accum = vec4(average / params.history_size) / float(1 << IRRADIANCE_HISTORY_BITS); - -#endif -#ifdef MODE_DEPTH - uvec2 value = imageLoad(depth_history_texture, read_pos); - vec2 depth_accum = vec4(average / params.history_size) / float(1 << IRRADIANCE_HISTORY_BITS); - - float probe_cell_size = float(params.grid_size / float(params.probe_axis_size - 1)) / cascades.data[params.cascade].to_cell; - float max_depth = length(params.grid_size / cascades.data[params.max_cascades - 1].to_cell); - max_depth /= probe_cell_size; - - depth_value = (vec2(average / params.history_size) / float(1 << DEPTH_HISTORY_BITS)) * vec2(max_depth, max_depth * max_depth); - -#endif - - /* Fill the border if required */ - - if (local == ivec2(0, 0)) { - copy_to[1] = texture_pos + ivec3(OCT_RES - 1, -1, 0); - copy_to[2] = texture_pos + ivec3(-1, OCT_RES - 1, 0); - copy_to[3] = texture_pos + ivec3(OCT_RES, OCT_RES, 0); - } else if (local == ivec2(OCT_RES - 1, 0)) { - copy_to[1] = texture_pos + ivec3(0, -1, 0); - copy_to[2] = texture_pos + ivec3(OCT_RES, OCT_RES - 1, 0); - copy_to[3] = texture_pos + ivec3(-1, OCT_RES, 0); - } else if (local == ivec2(0, OCT_RES - 1)) { - copy_to[1] = texture_pos + ivec3(-1, 0, 0); - copy_to[2] = texture_pos + ivec3(OCT_RES - 1, OCT_RES, 0); - copy_to[3] = texture_pos + ivec3(OCT_RES, -1, 0); - } else if (local == ivec2(OCT_RES - 1, OCT_RES - 1)) { - copy_to[1] = texture_pos + ivec3(0, OCT_RES, 0); - copy_to[2] = texture_pos + ivec3(OCT_RES, 0, 0); - copy_to[3] = texture_pos + ivec3(-1, -1, 0); - } else if (local.y == 0) { - copy_to[1] = texture_pos + ivec3(OCT_RES - local.x - 1, local.y - 1, 0); - } else if (local.x == 0) { - copy_to[1] = texture_pos + ivec3(local.x - 1, OCT_RES - local.y - 1, 0); - } else if (local.y == OCT_RES - 1) { - copy_to[1] = texture_pos + ivec3(OCT_RES - local.x - 1, local.y + 1, 0); - } else if (local.x == OCT_RES - 1) { - copy_to[1] = texture_pos + ivec3(local.x + 1, OCT_RES - local.y - 1, 0); - } - - for (int i = 0; i < 4; i++) { - if (copy_to[i] == ivec3(-2, -2, -2)) { - continue; - } -#ifdef MODE_IRRADIANCE - imageStore(irradiance_texture, copy_to[i], light_accum); -#endif -#ifdef MODE_DEPTH - imageStore(depth_texture, copy_to[i], vec4(depth_value, 0.0, 0.0)); -#endif - } - -#endif // MODE_STORE -} diff --git a/servers/rendering/renderer_rd/shaders/sdfgi_integrate.glsl b/servers/rendering/renderer_rd/shaders/sdfgi_integrate.glsl index 007e4c113a..9c03297f5c 100644 --- a/servers/rendering/renderer_rd/shaders/sdfgi_integrate.glsl +++ b/servers/rendering/renderer_rd/shaders/sdfgi_integrate.glsl @@ -2,7 +2,7 @@ #version 450 -VERSION_DEFINES +#VERSION_DEFINES layout(local_size_x = 8, local_size_y = 8, local_size_z = 1) in; @@ -52,7 +52,7 @@ layout(set = 1, binding = 1) uniform sampler linear_sampler_mipmaps; #define SKY_MODE_COLOR 1 #define SKY_MODE_SKY 2 -layout(push_constant, binding = 0, std430) uniform Params { +layout(push_constant, std430) uniform Params { vec3 grid_size; uint max_cascades; @@ -266,9 +266,9 @@ void main() { } else if (params.sky_mode == SKY_MODE_SKY) { #ifdef USE_CUBEMAP_ARRAY - light.rgb = textureLod(samplerCubeArray(sky_irradiance, linear_sampler_mipmaps), vec4(ray_dir, 0.0), 2.0).rgb; //use second mipmap because we dont usually throw a lot of rays, so this compensates + light.rgb = textureLod(samplerCubeArray(sky_irradiance, linear_sampler_mipmaps), vec4(ray_dir, 0.0), 2.0).rgb; // Use second mipmap because we don't usually throw a lot of rays, so this compensates. #else - light.rgb = textureLod(samplerCube(sky_irradiance, linear_sampler_mipmaps), ray_dir, 2.0).rgb; //use second mipmap because we dont usually throw a lot of rays, so this compensates + light.rgb = textureLod(samplerCube(sky_irradiance, linear_sampler_mipmaps), ray_dir, 2.0).rgb; // Use second mipmap because we don't usually throw a lot of rays, so this compensates. #endif light.rgb *= params.sky_energy; light.a = 0.0; diff --git a/servers/rendering/renderer_rd/shaders/sdfgi_preprocess.glsl b/servers/rendering/renderer_rd/shaders/sdfgi_preprocess.glsl index 916c60ac89..bce98f4054 100644 --- a/servers/rendering/renderer_rd/shaders/sdfgi_preprocess.glsl +++ b/servers/rendering/renderer_rd/shaders/sdfgi_preprocess.glsl @@ -2,7 +2,7 @@ #version 450 -VERSION_DEFINES +#VERSION_DEFINES #ifdef MODE_JUMPFLOOD_OPTIMIZED #define GROUP_SIZE 8 @@ -101,8 +101,8 @@ layout(set = 0, binding = 10, std430) restrict buffer DispatchData { dispatch_data; struct ProcessVoxel { - uint position; //xyz 7 bit packed, extra 11 bits for neigbours - uint albedo; //rgb bits 0-15 albedo, bits 16-21 are normal bits (set if geometry exists toward that side), extra 11 bits for neibhbours + uint position; // xyz 7 bit packed, extra 11 bits for neighbors. + uint albedo; //rgb bits 0-15 albedo, bits 16-21 are normal bits (set if geometry exists toward that side), extra 11 bits for neighbours uint light; //rgbe8985 encoded total saved light, extra 2 bits for neighbours uint light_aniso; //55555 light anisotropy, extra 2 bits for neighbours //total neighbours: 26 @@ -134,8 +134,8 @@ layout(set = 0, binding = 5, std430) restrict buffer readonly DispatchData { dispatch_data; struct ProcessVoxel { - uint position; //xyz 7 bit packed, extra 11 bits for neigbours - uint albedo; //rgb bits 0-15 albedo, bits 16-21 are normal bits (set if geometry exists toward that side), extra 11 bits for neibhbours + uint position; // xyz 7 bit packed, extra 11 bits for neighbors. + uint albedo; //rgb bits 0-15 albedo, bits 16-21 are normal bits (set if geometry exists toward that side), extra 11 bits for neighbours uint light; //rgbe8985 encoded total saved light, extra 2 bits for neighbours uint light_aniso; //55555 light anisotropy, extra 2 bits for neighbours //total neighbours: 26 @@ -155,7 +155,7 @@ layout(r16ui, set = 0, binding = 2) uniform restrict readonly uimage3D src_occlu #endif -layout(push_constant, binding = 0, std430) uniform Params { +layout(push_constant, std430) uniform Params { ivec3 scroll; int grid_size; @@ -183,7 +183,7 @@ void main() { ivec3 write_pos = read_pos + params.scroll; if (any(lessThan(write_pos, ivec3(0))) || any(greaterThanEqual(write_pos, ivec3(params.grid_size)))) { - return; //fits outside the 3D texture, dont do anything + return; // Fits outside the 3D texture, don't do anything. } uint albedo = ((src_process_voxels.data[index].albedo & 0x7FFF) << 1) | 1; //add solid bit diff --git a/servers/rendering/renderer_rd/shaders/skeleton.glsl b/servers/rendering/renderer_rd/shaders/skeleton.glsl index 680d1045cd..a893a66c94 100644 --- a/servers/rendering/renderer_rd/shaders/skeleton.glsl +++ b/servers/rendering/renderer_rd/shaders/skeleton.glsl @@ -2,7 +2,7 @@ #version 450 -VERSION_DEFINES +#VERSION_DEFINES layout(local_size_x = 64, local_size_y = 1, local_size_z = 1) in; @@ -36,7 +36,7 @@ layout(set = 2, binding = 0, std430) buffer restrict readonly SkeletonData { } bone_transforms; -layout(push_constant, binding = 0, std430) uniform Params { +layout(push_constant, std430) uniform Params { bool has_normal; bool has_tangent; bool has_skeleton; @@ -74,6 +74,53 @@ void main() { #ifdef MODE_2D vec2 vertex = uintBitsToFloat(uvec2(src_vertices.data[src_offset + 0], src_vertices.data[src_offset + 1])); + + if (params.has_blend_shape) { + float blend_total = 0.0; + vec2 blend_vertex = vec2(0.0); + + for (uint i = 0; i < params.blend_shape_count; i++) { + float w = blend_shape_weights.data[i]; + if (abs(w) > 0.0001) { + uint base_offset = (params.vertex_count * i + index) * params.vertex_stride; + + blend_vertex += uintBitsToFloat(uvec2(src_blend_shapes.data[base_offset + 0], src_blend_shapes.data[base_offset + 1])) * w; + + base_offset += 2; + + blend_total += w; + } + } + + if (params.normalized_blend_shapes) { + vertex = (1.0 - blend_total) * vertex; + } + + vertex += blend_vertex; + } + + if (params.has_skeleton) { + 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; + + skin_offset += params.skin_weight_offset; + + uvec2 weights = uvec2(src_bone_weights.data[skin_offset + 0], src_bone_weights.data[skin_offset + 1]); + + vec2 weights_01 = unpackUnorm2x16(weights.x); + vec2 weights_23 = unpackUnorm2x16(weights.y); + + mat4 m = mat4(bone_transforms.data[bones_01.x], bone_transforms.data[bones_01.x + 1], vec4(0.0, 0.0, 1.0, 0.0), vec4(0.0, 0.0, 0.0, 1.0)) * weights_01.x; + m += mat4(bone_transforms.data[bones_01.y], bone_transforms.data[bones_01.y + 1], vec4(0.0, 0.0, 1.0, 0.0), vec4(0.0, 0.0, 0.0, 1.0)) * weights_01.y; + m += mat4(bone_transforms.data[bones_23.x], bone_transforms.data[bones_23.x + 1], vec4(0.0, 0.0, 1.0, 0.0), vec4(0.0, 0.0, 0.0, 1.0)) * weights_23.x; + m += mat4(bone_transforms.data[bones_23.y], bone_transforms.data[bones_23.y + 1], vec4(0.0, 0.0, 1.0, 0.0), vec4(0.0, 0.0, 0.0, 1.0)) * weights_23.y; + + //reverse order because its transposed + vertex = (vec4(vertex, 0.0, 1.0) * m).xy; + } #else vec3 vertex; vec3 normal; @@ -113,7 +160,7 @@ void main() { } if (params.has_tangent) { - blend_tangent += decode_abgr_2_10_10_10(src_blend_shapes.data[base_offset]).rgb; + blend_tangent += decode_abgr_2_10_10_10(src_blend_shapes.data[base_offset]).rgb * w; } blend_total += w; @@ -127,8 +174,8 @@ void main() { } vertex += blend_vertex; - normal += normalize(normal + blend_normal); - tangent.rgb += normalize(tangent.rgb + blend_tangent); + normal = normalize(normal + blend_normal); + tangent.rgb = normalize(tangent.rgb + blend_tangent); } if (params.has_skeleton) { diff --git a/servers/rendering/renderer_rd/shaders/sky.glsl b/servers/rendering/renderer_rd/shaders/sky.glsl index 6c985e1f5c..5b4594da99 100644 --- a/servers/rendering/renderer_rd/shaders/sky.glsl +++ b/servers/rendering/renderer_rd/shaders/sky.glsl @@ -2,15 +2,23 @@ #version 450 -VERSION_DEFINES +#VERSION_DEFINES + +#define MAX_VIEWS 2 + +#if defined(USE_MULTIVIEW) && defined(has_VK_KHR_multiview) +#extension GL_EXT_multiview : enable +#endif layout(location = 0) out vec2 uv_interp; -layout(push_constant, binding = 1, std430) uniform Params { +layout(push_constant, std430) uniform Params { mat3 orientation; - vec4 proj; + vec4 projections[MAX_VIEWS]; vec4 position_multiplier; float time; + float luminance_multiplier; + float pad[2]; } params; @@ -24,17 +32,33 @@ void main() { #version 450 -VERSION_DEFINES +#VERSION_DEFINES + +#ifdef USE_MULTIVIEW +#ifdef has_VK_KHR_multiview +#extension GL_EXT_multiview : enable +#define ViewIndex gl_ViewIndex +#else // has_VK_KHR_multiview +// !BAS! This needs to become an input once we implement our fallback! +#define ViewIndex 0 +#endif // has_VK_KHR_multiview +#else // USE_MULTIVIEW +// Set to zero, not supported in non stereo +#define ViewIndex 0 +#endif //USE_MULTIVIEW #define M_PI 3.14159265359 +#define MAX_VIEWS 2 layout(location = 0) in vec2 uv_interp; -layout(push_constant, binding = 1, std430) uniform Params { +layout(push_constant, std430) uniform Params { mat3 orientation; - vec4 proj; + vec4 projections[MAX_VIEWS]; vec4 position_multiplier; - float time; //TODO consider adding vec2 screen res, and float radiance size + float time; + float luminance_multiplier; + float pad[2]; } params; @@ -85,16 +109,11 @@ struct DirectionalLightData { layout(set = 0, binding = 3, std140) uniform DirectionalLights { DirectionalLightData data[MAX_DIRECTIONAL_LIGHT_DATA_STRUCTS]; } - directional_lights; -#ifdef USE_MATERIAL_UNIFORMS +#ifdef MATERIAL_UNIFORMS_USED layout(set = 1, binding = 0, std140) uniform MaterialUniforms{ - /* clang-format off */ - -MATERIAL_UNIFORMS - - /* clang-format on */ +#MATERIAL_UNIFORMS } material; #endif @@ -127,11 +146,7 @@ layout(set = 3, binding = 0) uniform texture3D volumetric_fog_texture; #define AT_QUARTER_RES_PASS false #endif -/* clang-format off */ - -FRAGMENT_SHADER_GLOBALS - -/* clang-format on */ +#GLOBALS layout(location = 0) out vec4 frag_color; @@ -162,15 +177,14 @@ vec4 fog_process(vec3 view, vec3 sky_color) { void main() { vec3 cube_normal; cube_normal.z = -1.0; - cube_normal.x = (cube_normal.z * (-uv_interp.x - params.proj.x)) / params.proj.y; - cube_normal.y = -(cube_normal.z * (-uv_interp.y - params.proj.z)) / params.proj.w; + cube_normal.x = (cube_normal.z * (-uv_interp.x - params.projections[ViewIndex].x)) / params.projections[ViewIndex].y; + cube_normal.y = -(cube_normal.z * (-uv_interp.y - params.projections[ViewIndex].z)) / params.projections[ViewIndex].w; cube_normal = mat3(params.orientation) * cube_normal; - cube_normal.z = -cube_normal.z; cube_normal = normalize(cube_normal); vec2 uv = uv_interp * 0.5 + 0.5; - vec2 panorama_coords = vec2(atan(cube_normal.x, cube_normal.z), acos(cube_normal.y)); + vec2 panorama_coords = vec2(atan(cube_normal.x, -cube_normal.z), acos(cube_normal.y)); if (panorama_coords.x < 0.0) { panorama_coords.x += M_PI * 2.0; @@ -185,39 +199,25 @@ void main() { vec4 custom_fog = vec4(0.0); #ifdef USE_CUBEMAP_PASS - vec3 inverted_cube_normal = cube_normal; - inverted_cube_normal.z *= -1.0; #ifdef USES_HALF_RES_COLOR - half_res_color = texture(samplerCube(half_res, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), inverted_cube_normal); + half_res_color = texture(samplerCube(half_res, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), cube_normal) * params.luminance_multiplier; #endif #ifdef USES_QUARTER_RES_COLOR - quarter_res_color = texture(samplerCube(quarter_res, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), inverted_cube_normal); + 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); + 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); + quarter_res_color = textureLod(sampler2D(quarter_res, material_samplers[SAMPLER_LINEAR_CLAMP]), uv, 0.0) * params.luminance_multiplier; #endif #endif -// unused, just here to make our compiler happy, make sure we don't execute any light code the user adds in.. -#ifndef REALLYINCLUDETHIS { - /* clang-format off */ -LIGHT_SHADER_CODE +#CODE : SKY - /* clang-format on */ - } -#endif - { - /* clang-format off */ - -FRAGMENT_SHADER_CODE - - /* clang-format on */ } frag_color.rgb = color * params.position_multiplier.w; @@ -247,4 +247,7 @@ FRAGMENT_SHADER_CODE 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; } diff --git a/servers/rendering/renderer_rd/shaders/sort.glsl b/servers/rendering/renderer_rd/shaders/sort.glsl index e5ebb9c64b..48cf69012a 100644 --- a/servers/rendering/renderer_rd/shaders/sort.glsl +++ b/servers/rendering/renderer_rd/shaders/sort.glsl @@ -2,7 +2,7 @@ #version 450 -VERSION_DEFINES +#VERSION_DEFINES // Original version here: // https://github.com/GPUOpen-LibrariesAndSDKs/GPUParticles11/blob/master/gpuparticles11/src/Shaders @@ -47,7 +47,7 @@ layout(set = 1, binding = 0, std430) restrict buffer SortBuffer { } sort_buffer; -layout(push_constant, binding = 0, std430) uniform Params { +layout(push_constant, std430) uniform Params { uint total_elements; uint pad[3]; ivec4 job_params; diff --git a/servers/rendering/renderer_rd/shaders/specular_merge.glsl b/servers/rendering/renderer_rd/shaders/specular_merge.glsl index 0b8f406213..3579c35cce 100644 --- a/servers/rendering/renderer_rd/shaders/specular_merge.glsl +++ b/servers/rendering/renderer_rd/shaders/specular_merge.glsl @@ -2,7 +2,7 @@ #version 450 -VERSION_DEFINES +#VERSION_DEFINES layout(location = 0) out vec2 uv_interp; @@ -17,7 +17,7 @@ void main() { #version 450 -VERSION_DEFINES +#VERSION_DEFINES layout(location = 0) in vec2 uv_interp; diff --git a/servers/rendering/renderer_rd/shaders/ssao_downsample.glsl b/servers/rendering/renderer_rd/shaders/ss_effects_downsample.glsl index cb2d31f70d..134aae5ce7 100644 --- a/servers/rendering/renderer_rd/shaders/ssao_downsample.glsl +++ b/servers/rendering/renderer_rd/shaders/ss_effects_downsample.glsl @@ -21,11 +21,11 @@ #version 450 -VERSION_DEFINES +#VERSION_DEFINES layout(local_size_x = 8, local_size_y = 8, local_size_z = 1) in; -layout(push_constant, binding = 1, std430) uniform Params { +layout(push_constant, std430) uniform Params { vec2 pixel_size; float z_far; float z_near; @@ -42,6 +42,9 @@ layout(r16f, set = 1, binding = 0) uniform restrict writeonly image2DArray dest_ layout(r16f, set = 2, binding = 0) uniform restrict writeonly image2DArray dest_image1; layout(r16f, set = 2, binding = 1) uniform restrict writeonly image2DArray dest_image2; layout(r16f, set = 2, binding = 2) uniform restrict writeonly image2DArray dest_image3; +#ifdef GENERATE_FULL_MIPS +layout(r16f, set = 2, binding = 3) uniform restrict writeonly image2DArray dest_image4; +#endif #endif vec4 screen_space_to_view_space_depth(vec4 p_depth) { @@ -150,7 +153,27 @@ void prepare_depths_and_mips(vec4 p_samples, uvec2 p_output_coord, uvec2 p_gtid) float avg = mip_smart_average(vec4(sample_00, sample_01, sample_10, sample_11)); imageStore(dest_image3, ivec3(p_output_coord.x, p_output_coord.y, depth_array_index), vec4(avg)); +#ifndef GENERATE_FULL_MIPS + } +#else + depth_buffer[depth_array_index][buffer_coord.x][buffer_coord.y] = avg; } + still_alive = p_gtid.x % 16 == depth_array_offset.x && depth_array_offset.y % 16 == depth_array_offset.y; + + p_output_coord /= 2; + groupMemoryBarrier(); + barrier(); + + if (still_alive) { + float sample_00 = depth_buffer[depth_array_index][buffer_coord.x + 0][buffer_coord.y + 0]; + float sample_01 = depth_buffer[depth_array_index][buffer_coord.x + 0][buffer_coord.y + 8]; + float sample_10 = depth_buffer[depth_array_index][buffer_coord.x + 8][buffer_coord.y + 0]; + float sample_11 = depth_buffer[depth_array_index][buffer_coord.x + 8][buffer_coord.y + 8]; + + float avg = mip_smart_average(vec4(sample_00, sample_01, sample_10, sample_11)); + imageStore(dest_image4, ivec3(p_output_coord.x, p_output_coord.y, depth_array_index), vec4(avg)); + } +#endif } #else #ifndef USE_HALF_BUFFERS diff --git a/servers/rendering/renderer_rd/shaders/ssao.glsl b/servers/rendering/renderer_rd/shaders/ssao.glsl index 231f8f91ec..2a87e273bc 100644 --- a/servers/rendering/renderer_rd/shaders/ssao.glsl +++ b/servers/rendering/renderer_rd/shaders/ssao.glsl @@ -21,9 +21,7 @@ #version 450 -VERSION_DEFINES - -#define SSAO_ADAPTIVE_TAP_BASE_COUNT 5 +#VERSION_DEFINES #define INTELSSAO_MAIN_DISK_SAMPLE_COUNT (32) const vec4 sample_pattern[INTELSSAO_MAIN_DISK_SAMPLE_COUNT] = { @@ -62,7 +60,6 @@ const int num_taps[5] = { 3, 5, 12, 0, 0 }; #define SSAO_REDUCE_RADIUS_NEAR_SCREEN_BORDER_ENABLE_AT_QUALITY_PRESET (1) #define SSAO_MAX_TAPS 32 -#define SSAO_MAX_REF_TAPS 512 #define SSAO_ADAPTIVE_TAP_BASE_COUNT 5 #define SSAO_ADAPTIVE_TAP_FLEXIBLE_COUNT (SSAO_MAX_TAPS - SSAO_ADAPTIVE_TAP_BASE_COUNT) #define SSAO_DEPTH_MIP_LEVELS 4 @@ -88,7 +85,7 @@ counter; layout(rg8, set = 2, binding = 0) uniform restrict writeonly image2D dest_image; // This push_constant is full - 128 bytes - if you need to add more data, consider adding to the uniform buffer instead -layout(push_constant, binding = 3, std430) uniform Params { +layout(push_constant, std430) uniform Params { ivec2 screen_size; int pass; int quality; diff --git a/servers/rendering/renderer_rd/shaders/ssao_blur.glsl b/servers/rendering/renderer_rd/shaders/ssao_blur.glsl index 510a777048..f42734c46d 100644 --- a/servers/rendering/renderer_rd/shaders/ssao_blur.glsl +++ b/servers/rendering/renderer_rd/shaders/ssao_blur.glsl @@ -21,7 +21,7 @@ #version 450 -VERSION_DEFINES +#VERSION_DEFINES layout(local_size_x = 8, local_size_y = 8, local_size_z = 1) in; @@ -29,7 +29,7 @@ layout(set = 0, binding = 0) uniform sampler2D source_ssao; layout(rg8, set = 1, binding = 0) uniform restrict writeonly image2D dest_image; -layout(push_constant, binding = 1, std430) uniform Params { +layout(push_constant, std430) uniform Params { float edge_sharpness; float pad; vec2 half_screen_pixel_size; @@ -128,19 +128,19 @@ void main() { #ifdef MODE_NON_SMART - vec2 halfPixel = params.half_screen_pixel_size * 0.5f; + vec2 half_pixel = params.half_screen_pixel_size * 0.5; vec2 uv = (vec2(gl_GlobalInvocationID.xy) + vec2(0.5, 0.5)) * params.half_screen_pixel_size; - vec2 centre = textureLod(source_ssao, vec2(uv), 0.0).xy; + vec2 center = textureLod(source_ssao, vec2(uv), 0.0).xy; vec4 vals; - vals.x = textureLod(source_ssao, vec2(uv + vec2(-halfPixel.x * 3, -halfPixel.y)), 0.0).x; - vals.y = textureLod(source_ssao, vec2(uv + vec2(+halfPixel.x, -halfPixel.y * 3)), 0.0).x; - vals.z = textureLod(source_ssao, vec2(uv + vec2(-halfPixel.x, +halfPixel.y * 3)), 0.0).x; - vals.w = textureLod(source_ssao, vec2(uv + vec2(+halfPixel.x * 3, +halfPixel.y)), 0.0).x; + vals.x = textureLod(source_ssao, vec2(uv + vec2(-half_pixel.x * 3, -half_pixel.y)), 0.0).x; + vals.y = textureLod(source_ssao, vec2(uv + vec2(+half_pixel.x, -half_pixel.y * 3)), 0.0).x; + vals.z = textureLod(source_ssao, vec2(uv + vec2(-half_pixel.x, +half_pixel.y * 3)), 0.0).x; + vals.w = textureLod(source_ssao, vec2(uv + vec2(+half_pixel.x * 3, +half_pixel.y)), 0.0).x; - vec2 sampled = vec2(dot(vals, vec4(0.2)) + centre.x * 0.2, centre.y); + vec2 sampled = vec2(dot(vals, vec4(0.2)) + center.x * 0.2, center.y); #else #ifdef MODE_SMART diff --git a/servers/rendering/renderer_rd/shaders/ssao_importance_map.glsl b/servers/rendering/renderer_rd/shaders/ssao_importance_map.glsl index 6aa7624261..04f98964e8 100644 --- a/servers/rendering/renderer_rd/shaders/ssao_importance_map.glsl +++ b/servers/rendering/renderer_rd/shaders/ssao_importance_map.glsl @@ -21,12 +21,12 @@ #version 450 -VERSION_DEFINES +#VERSION_DEFINES layout(local_size_x = 8, local_size_y = 8, local_size_z = 1) in; #ifdef GENERATE_MAP -layout(set = 0, binding = 0) uniform sampler2DArray source_ssao; +layout(set = 0, binding = 0) uniform sampler2DArray source_texture; #else layout(set = 0, binding = 0) uniform sampler2D source_importance; #endif @@ -39,7 +39,7 @@ layout(set = 2, binding = 0, std430) buffer Counter { counter; #endif -layout(push_constant, binding = 1, std430) uniform Params { +layout(push_constant, std430) uniform Params { vec2 half_screen_pixel_size; float intensity; float power; @@ -56,11 +56,10 @@ void main() { vec2 base_uv = (vec2(base_position) + vec2(0.5f, 0.5f)) * params.half_screen_pixel_size; - float avg = 0.0; float minV = 1.0; float maxV = 0.0; for (int i = 0; i < 4; i++) { - vec4 vals = textureGather(source_ssao, vec3(base_uv, i)); + vec4 vals = textureGather(source_texture, vec3(base_uv, i)); // apply the same modifications that would have been applied in the main shader vals = params.intensity * vals; @@ -69,8 +68,6 @@ void main() { vals = pow(clamp(vals, 0.0, 1.0), vec4(params.power)); - avg += dot(vec4(vals.x, vals.y, vals.z, vals.w), vec4(1.0 / 16.0, 1.0 / 16.0, 1.0 / 16.0, 1.0 / 16.0)); - maxV = max(maxV, max(max(vals.x, vals.y), max(vals.z, vals.w))); minV = min(minV, min(min(vals.x, vals.y), min(vals.z, vals.w))); } diff --git a/servers/rendering/renderer_rd/shaders/ssao_interleave.glsl b/servers/rendering/renderer_rd/shaders/ssao_interleave.glsl index 4fdf334aa5..f6a9a92fac 100644 --- a/servers/rendering/renderer_rd/shaders/ssao_interleave.glsl +++ b/servers/rendering/renderer_rd/shaders/ssao_interleave.glsl @@ -20,14 +20,14 @@ #version 450 -VERSION_DEFINES +#VERSION_DEFINES layout(local_size_x = 8, local_size_y = 8, local_size_z = 1) in; layout(rgba8, set = 0, binding = 0) uniform restrict writeonly image2D dest_image; layout(set = 1, binding = 0) uniform sampler2DArray source_texture; -layout(push_constant, binding = 1, std430) uniform Params { +layout(push_constant, std430) uniform Params { float inv_sharpness; uint size_modifier; vec2 pixel_size; diff --git a/servers/rendering/renderer_rd/shaders/ssil.glsl b/servers/rendering/renderer_rd/shaders/ssil.glsl new file mode 100644 index 0000000000..513791dfbf --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/ssil.glsl @@ -0,0 +1,444 @@ +/////////////////////////////////////////////////////////////////////////////////////////////////////////////////////// +// Copyright (c) 2016, Intel Corporation +// Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated +// documentation files (the "Software"), to deal in the Software without restriction, including without limitation +// the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to +// permit persons to whom the Software is furnished to do so, subject to the following conditions: +// The above copyright notice and this permission notice shall be included in all copies or substantial portions of +// the Software. +// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO +// THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE +// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, +// TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE +// SOFTWARE. +/////////////////////////////////////////////////////////////////////////////////////////////////////////////////////// +// File changes (yyyy-mm-dd) +// 2016-09-07: filip.strugar@intel.com: first commit +// 2020-12-05: clayjohn: convert to Vulkan and Godot +// 2021-05-27: clayjohn: convert SSAO to SSIL +/////////////////////////////////////////////////////////////////////////////////////////////////////////////////////// + +#[compute] + +#version 450 + +#VERSION_DEFINES + +#define SSIL_MAIN_DISK_SAMPLE_COUNT (32) +const vec4 sample_pattern[SSIL_MAIN_DISK_SAMPLE_COUNT] = { + vec4(0.78488064, 0.56661671, 1.500000, -0.126083), vec4(0.26022232, -0.29575172, 1.500000, -1.064030), vec4(0.10459357, 0.08372527, 1.110000, -2.730563), vec4(-0.68286800, 0.04963045, 1.090000, -0.498827), + vec4(-0.13570161, -0.64190155, 1.250000, -0.532765), vec4(-0.26193795, -0.08205118, 0.670000, -1.783245), vec4(-0.61177456, 0.66664219, 0.710000, -0.044234), vec4(0.43675563, 0.25119025, 0.610000, -1.167283), + vec4(0.07884444, 0.86618668, 0.640000, -0.459002), vec4(-0.12790935, -0.29869005, 0.600000, -1.729424), vec4(-0.04031125, 0.02413622, 0.600000, -4.792042), vec4(0.16201244, -0.52851415, 0.790000, -1.067055), + vec4(-0.70991218, 0.47301072, 0.640000, -0.335236), vec4(0.03277707, -0.22349690, 0.600000, -1.982384), vec4(0.68921727, 0.36800742, 0.630000, -0.266718), vec4(0.29251814, 0.37775412, 0.610000, -1.422520), + vec4(-0.12224089, 0.96582592, 0.600000, -0.426142), vec4(0.11071457, -0.16131058, 0.600000, -2.165947), vec4(0.46562141, -0.59747696, 0.600000, -0.189760), vec4(-0.51548797, 0.11804193, 0.600000, -1.246800), + vec4(0.89141309, -0.42090443, 0.600000, 0.028192), vec4(-0.32402530, -0.01591529, 0.600000, -1.543018), vec4(0.60771245, 0.41635221, 0.600000, -0.605411), vec4(0.02379565, -0.08239821, 0.600000, -3.809046), + vec4(0.48951152, -0.23657045, 0.600000, -1.189011), vec4(-0.17611565, -0.81696892, 0.600000, -0.513724), vec4(-0.33930185, -0.20732205, 0.600000, -1.698047), vec4(-0.91974425, 0.05403209, 0.600000, 0.062246), + vec4(-0.15064627, -0.14949332, 0.600000, -1.896062), vec4(0.53180975, -0.35210401, 0.600000, -0.758838), vec4(0.41487166, 0.81442589, 0.600000, -0.505648), vec4(-0.24106961, -0.32721516, 0.600000, -1.665244) +}; + +// these values can be changed (up to SSIL_MAX_TAPS) with no changes required elsewhere; values for 4th and 5th preset are ignored but array needed to avoid compilation errors +// the actual number of texture samples is two times this value (each "tap" has two symmetrical depth texture samples) +const int num_taps[5] = { 3, 5, 12, 0, 0 }; + +#define SSIL_TILT_SAMPLES_ENABLE_AT_QUALITY_PRESET (99) // to disable simply set to 99 or similar +#define SSIL_TILT_SAMPLES_AMOUNT (0.4) +// +#define SSIL_HALOING_REDUCTION_ENABLE_AT_QUALITY_PRESET (1) // to disable simply set to 99 or similar +#define SSIL_HALOING_REDUCTION_AMOUNT (0.8) // values from 0.0 - 1.0, 1.0 means max weighting (will cause artifacts, 0.8 is more reasonable) +// +#define SSIL_DEPTH_MIPS_ENABLE_AT_QUALITY_PRESET (2) +#define SSIL_DEPTH_MIPS_GLOBAL_OFFSET (-4.3) // best noise/quality/performance tradeoff, found empirically +// +// !!warning!! the edge handling is hard-coded to 'disabled' on quality level 0, and enabled above, on the C++ side; while toggling it here will work for +// testing purposes, it will not yield performance gains (or correct results) +#define SSIL_DEPTH_BASED_EDGES_ENABLE_AT_QUALITY_PRESET (1) +// +#define SSIL_REDUCE_RADIUS_NEAR_SCREEN_BORDER_ENABLE_AT_QUALITY_PRESET (1) + +#define SSIL_MAX_TAPS 32 +#define SSIL_ADAPTIVE_TAP_BASE_COUNT 5 +#define SSIL_ADAPTIVE_TAP_FLEXIBLE_COUNT (SSIL_MAX_TAPS - SSIL_ADAPTIVE_TAP_BASE_COUNT) +#define SSIL_DEPTH_MIP_LEVELS 4 + +layout(local_size_x = 8, local_size_y = 8, local_size_z = 1) in; + +layout(set = 0, binding = 0) uniform sampler2DArray source_depth_mipmaps; +layout(rgba8, set = 0, binding = 1) uniform restrict readonly image2D source_normal; +layout(set = 0, binding = 2) uniform Constants { //get into a lower set + vec4 rotation_matrices[20]; +} +constants; + +#ifdef ADAPTIVE +layout(rgba16, set = 1, binding = 0) uniform restrict readonly image2DArray source_ssil; +layout(set = 1, binding = 1) uniform sampler2D source_importance; +layout(set = 1, binding = 2, std430) buffer Counter { + uint sum; +} +counter; +#endif + +layout(rgba16, set = 2, binding = 0) uniform restrict writeonly image2D dest_image; +layout(r8, set = 2, binding = 1) uniform image2D edges_weights_image; + +layout(set = 3, binding = 0) uniform sampler2D last_frame; +layout(set = 3, binding = 1) uniform ProjectionConstants { + mat4 reprojection; +} +projection_constants; + +layout(push_constant, std430) uniform Params { + ivec2 screen_size; + int pass; + int quality; + + vec2 half_screen_pixel_size; + vec2 half_screen_pixel_size_x025; + + vec2 NDC_to_view_mul; + vec2 NDC_to_view_add; + + vec2 pad2; + float z_near; + float z_far; + + float radius; + float intensity; + int size_multiplier; + int pad; + + float fade_out_mul; + float fade_out_add; + float normal_rejection_amount; + float inv_radius_near_limit; + + bool is_orthogonal; + float neg_inv_radius; + float load_counter_avg_div; + float adaptive_sample_limit; + + ivec2 pass_coord_offset; + vec2 pass_uv_offset; +} +params; + +float pack_edges(vec4 p_edgesLRTB) { + p_edgesLRTB = round(clamp(p_edgesLRTB, 0.0, 1.0) * 3.05); + return dot(p_edgesLRTB, vec4(64.0 / 255.0, 16.0 / 255.0, 4.0 / 255.0, 1.0 / 255.0)); +} + +vec3 NDC_to_view_space(vec2 p_pos, float p_viewspace_depth) { + if (params.is_orthogonal) { + return vec3((params.NDC_to_view_mul * p_pos.xy + params.NDC_to_view_add), p_viewspace_depth); + } else { + return vec3((params.NDC_to_view_mul * p_pos.xy + params.NDC_to_view_add) * p_viewspace_depth, p_viewspace_depth); + } +} + +// calculate effect radius and fit our screen sampling pattern inside it +void calculate_radius_parameters(const float p_pix_center_length, const vec2 p_pixel_size_at_center, out float r_lookup_radius, out float r_radius, out float r_fallof_sq) { + r_radius = params.radius; + + // when too close, on-screen sampling disk will grow beyond screen size; limit this to avoid closeup temporal artifacts + const float too_close_limit = clamp(p_pix_center_length * params.inv_radius_near_limit, 0.0, 1.0) * 0.8 + 0.2; + + r_radius *= too_close_limit; + + // 0.85 is to reduce the radius to allow for more samples on a slope to still stay within influence + r_lookup_radius = (0.85 * r_radius) / p_pixel_size_at_center.x; + + // used to calculate falloff (both for AO samples and per-sample weights) + r_fallof_sq = -1.0 / (r_radius * r_radius); +} + +vec4 calculate_edges(const float p_center_z, const float p_left_z, const float p_right_z, const float p_top_z, const float p_bottom_z) { + // slope-sensitive depth-based edge detection + vec4 edgesLRTB = vec4(p_left_z, p_right_z, p_top_z, p_bottom_z) - p_center_z; + vec4 edgesLRTB_slope_adjusted = edgesLRTB + edgesLRTB.yxwz; + edgesLRTB = min(abs(edgesLRTB), abs(edgesLRTB_slope_adjusted)); + return clamp((1.3 - edgesLRTB / (p_center_z * 0.040)), 0.0, 1.0); +} + +vec3 decode_normal(vec3 p_encoded_normal) { + vec3 normal = p_encoded_normal * 2.0 - 1.0; + return normal; +} + +vec3 load_normal(ivec2 p_pos) { + vec3 encoded_normal = imageLoad(source_normal, p_pos).xyz; + encoded_normal.z = 1.0 - encoded_normal.z; + return decode_normal(encoded_normal); +} + +vec3 load_normal(ivec2 p_pos, ivec2 p_offset) { + vec3 encoded_normal = imageLoad(source_normal, p_pos + p_offset).xyz; + encoded_normal.z = 1.0 - encoded_normal.z; + return decode_normal(encoded_normal); +} + +// all vectors in viewspace +float calculate_pixel_obscurance(vec3 p_pixel_normal, vec3 p_hit_delta, float p_fallof_sq) { + float length_sq = dot(p_hit_delta, p_hit_delta); + float NdotD = dot(p_pixel_normal, p_hit_delta) / sqrt(length_sq); + + float falloff_mult = max(0.0, length_sq * p_fallof_sq + 1.0); + + return max(0, NdotD - 0.05) * falloff_mult; +} + +void SSIL_tap_inner(const int p_quality_level, inout vec3 r_color_sum, inout float r_obscurance_sum, inout float r_weight_sum, const vec2 p_sampling_uv, const float p_mip_level, const vec3 p_pix_center_pos, vec3 p_pixel_normal, const float p_fallof_sq, const float p_weight_mod) { + // get depth at sample + float viewspace_sample_z = textureLod(source_depth_mipmaps, vec3(p_sampling_uv, params.pass), p_mip_level).x; + vec3 sample_normal = load_normal(ivec2(p_sampling_uv * vec2(params.screen_size))); + + // convert to viewspace + vec3 hit_pos = NDC_to_view_space(p_sampling_uv.xy, viewspace_sample_z); + vec3 hit_delta = hit_pos - p_pix_center_pos; + + float obscurance = calculate_pixel_obscurance(p_pixel_normal, hit_delta, p_fallof_sq); + float weight = 1.0; + + if (p_quality_level >= SSIL_HALOING_REDUCTION_ENABLE_AT_QUALITY_PRESET) { + float reduct = max(0, -hit_delta.z); + reduct = clamp(reduct * params.neg_inv_radius + 2.0, 0.0, 1.0); + weight = SSIL_HALOING_REDUCTION_AMOUNT * reduct + (1.0 - SSIL_HALOING_REDUCTION_AMOUNT); + } + + // Translate sampling_uv to last screen's coordinates + const vec4 sample_pos = projection_constants.reprojection * vec4(p_sampling_uv * 2.0 - 1.0, (viewspace_sample_z - params.z_near) / (params.z_far - params.z_near) * 2.0 - 1.0, 1.0); + vec2 reprojected_sampling_uv = (sample_pos.xy / sample_pos.w) * 0.5 + 0.5; + + weight *= p_weight_mod; + + r_obscurance_sum += obscurance * weight; + + vec3 sample_color = textureLod(last_frame, reprojected_sampling_uv, 5.0).rgb; + // Reduce impact of fireflies by tonemapping before averaging: http://graphicrants.blogspot.com/2013/12/tone-mapping.html + sample_color /= (1.0 + dot(sample_color, vec3(0.299, 0.587, 0.114))); + r_color_sum += sample_color * obscurance * weight * mix(1.0, smoothstep(0.0, 0.1, -dot(sample_normal, normalize(hit_delta))), params.normal_rejection_amount); + r_weight_sum += weight; +} + +void SSILTap(const int p_quality_level, inout vec3 r_color_sum, inout float r_obscurance_sum, inout float r_weight_sum, const int p_tap_index, const mat2 p_rot_scale, const vec3 p_pix_center_pos, vec3 p_pixel_normal, const vec2 p_normalized_screen_pos, const float p_mip_offset, const float p_fallof_sq, float p_weight_mod, vec2 p_norm_xy, float p_norm_xy_length) { + vec2 sample_offset; + float sample_pow_2_len; + + // patterns + { + vec4 new_sample = sample_pattern[p_tap_index]; + sample_offset = new_sample.xy * p_rot_scale; + sample_pow_2_len = new_sample.w; // precalculated, same as: sample_pow_2_len = log2( length( new_sample.xy ) ); + p_weight_mod *= new_sample.z; + } + + // snap to pixel center (more correct obscurance math, avoids artifacts) + sample_offset = round(sample_offset); + + // calculate MIP based on the sample distance from the centre, similar to as described + // in http://graphics.cs.williams.edu/papers/SAOHPG12/. + float mip_level = (p_quality_level < SSIL_DEPTH_MIPS_ENABLE_AT_QUALITY_PRESET) ? (0) : (sample_pow_2_len + p_mip_offset); + + vec2 sampling_uv = sample_offset * params.half_screen_pixel_size + p_normalized_screen_pos; + + SSIL_tap_inner(p_quality_level, r_color_sum, r_obscurance_sum, r_weight_sum, sampling_uv, mip_level, p_pix_center_pos, p_pixel_normal, p_fallof_sq, p_weight_mod); + + // for the second tap, just use the mirrored offset + vec2 sample_offset_mirrored_uv = -sample_offset; + + // tilt the second set of samples so that the disk is effectively rotated by the normal + // effective at removing one set of artifacts, but too expensive for lower quality settings + if (p_quality_level >= SSIL_TILT_SAMPLES_ENABLE_AT_QUALITY_PRESET) { + float dot_norm = dot(sample_offset_mirrored_uv, p_norm_xy); + sample_offset_mirrored_uv -= dot_norm * p_norm_xy_length * p_norm_xy; + sample_offset_mirrored_uv = round(sample_offset_mirrored_uv); + } + + // snap to pixel center (more correct obscurance math, avoids artifacts) + vec2 sampling_mirrored_uv = sample_offset_mirrored_uv * params.half_screen_pixel_size + p_normalized_screen_pos; + + SSIL_tap_inner(p_quality_level, r_color_sum, r_obscurance_sum, r_weight_sum, sampling_mirrored_uv, mip_level, p_pix_center_pos, p_pixel_normal, p_fallof_sq, p_weight_mod); +} + +void generate_SSIL(out vec3 r_color, out vec4 r_edges, out float r_obscurance, out float r_weight, const vec2 p_pos, int p_quality_level, bool p_adaptive_base) { + vec2 pos_rounded = trunc(p_pos); + uvec2 upos = uvec2(pos_rounded); + + const int number_of_taps = (p_adaptive_base) ? (SSIL_ADAPTIVE_TAP_BASE_COUNT) : (num_taps[p_quality_level]); + float pix_z, pix_left_z, pix_top_z, pix_right_z, pix_bottom_z; + + vec4 valuesUL = textureGather(source_depth_mipmaps, vec3(pos_rounded * params.half_screen_pixel_size, params.pass)); + vec4 valuesBR = textureGather(source_depth_mipmaps, vec3((pos_rounded + vec2(1.0)) * params.half_screen_pixel_size, params.pass)); + + // get this pixel's viewspace depth + pix_z = valuesUL.y; + + // get left right top bottom neighbouring pixels for edge detection (gets compiled out on quality_level == 0) + pix_left_z = valuesUL.x; + pix_top_z = valuesUL.z; + pix_right_z = valuesBR.z; + pix_bottom_z = valuesBR.x; + + vec2 normalized_screen_pos = pos_rounded * params.half_screen_pixel_size + params.half_screen_pixel_size_x025; + vec3 pix_center_pos = NDC_to_view_space(normalized_screen_pos, pix_z); + + // Load this pixel's viewspace normal + uvec2 full_res_coord = upos * 2 * params.size_multiplier + params.pass_coord_offset.xy; + vec3 pixel_normal = load_normal(ivec2(full_res_coord)); + + const vec2 pixel_size_at_center = NDC_to_view_space(normalized_screen_pos.xy + params.half_screen_pixel_size, pix_center_pos.z).xy - pix_center_pos.xy; + + float pixel_lookup_radius; + float fallof_sq; + + // calculate effect radius and fit our screen sampling pattern inside it + float viewspace_radius; + calculate_radius_parameters(length(pix_center_pos), pixel_size_at_center, pixel_lookup_radius, viewspace_radius, fallof_sq); + + // calculate samples rotation/scaling + mat2 rot_scale_matrix; + uint pseudo_random_index; + + { + vec4 rotation_scale; + // reduce effect radius near the screen edges slightly; ideally, one would render a larger depth buffer (5% on each side) instead + if (!p_adaptive_base && (p_quality_level >= SSIL_REDUCE_RADIUS_NEAR_SCREEN_BORDER_ENABLE_AT_QUALITY_PRESET)) { + float near_screen_border = min(min(normalized_screen_pos.x, 1.0 - normalized_screen_pos.x), min(normalized_screen_pos.y, 1.0 - normalized_screen_pos.y)); + near_screen_border = clamp(10.0 * near_screen_border + 0.6, 0.0, 1.0); + pixel_lookup_radius *= near_screen_border; + } + + // load & update pseudo-random rotation matrix + pseudo_random_index = uint(pos_rounded.y * 2 + pos_rounded.x) % 5; + rotation_scale = constants.rotation_matrices[params.pass * 5 + pseudo_random_index]; + rot_scale_matrix = mat2(rotation_scale.x * pixel_lookup_radius, rotation_scale.y * pixel_lookup_radius, rotation_scale.z * pixel_lookup_radius, rotation_scale.w * pixel_lookup_radius); + } + + // the main obscurance & sample weight storage + vec3 color_sum = vec3(0.0); + float obscurance_sum = 0.0; + float weight_sum = 0.0; + + // edge mask for between this and left/right/top/bottom neighbour pixels - not used in quality level 0 so initialize to "no edge" (1 is no edge, 0 is edge) + vec4 edgesLRTB = vec4(1.0, 1.0, 1.0, 1.0); + + // Move center pixel slightly towards camera to avoid imprecision artifacts due to using of 16bit depth buffer; a lot smaller offsets needed when using 32bit floats + pix_center_pos *= 0.9992; + + if (!p_adaptive_base && (p_quality_level >= SSIL_DEPTH_BASED_EDGES_ENABLE_AT_QUALITY_PRESET)) { + edgesLRTB = calculate_edges(pix_z, pix_left_z, pix_right_z, pix_top_z, pix_bottom_z); + } + + const float global_mip_offset = SSIL_DEPTH_MIPS_GLOBAL_OFFSET; + float mip_offset = (p_quality_level < SSIL_DEPTH_MIPS_ENABLE_AT_QUALITY_PRESET) ? (0) : (log2(pixel_lookup_radius) + global_mip_offset); + + // Used to tilt the second set of samples so that the disk is effectively rotated by the normal + // effective at removing one set of artifacts, but too expensive for lower quality settings + vec2 norm_xy = vec2(pixel_normal.x, pixel_normal.y); + float norm_xy_length = length(norm_xy); + norm_xy /= vec2(norm_xy_length, -norm_xy_length); + norm_xy_length *= SSIL_TILT_SAMPLES_AMOUNT; + + // standard, non-adaptive approach + if ((p_quality_level != 3) || p_adaptive_base) { + for (int i = 0; i < number_of_taps; i++) { + SSILTap(p_quality_level, color_sum, obscurance_sum, weight_sum, i, rot_scale_matrix, pix_center_pos, pixel_normal, normalized_screen_pos, mip_offset, fallof_sq, 1.0, norm_xy, norm_xy_length); + } + } +#ifdef ADAPTIVE + else { + // add new ones if needed + vec2 full_res_uv = normalized_screen_pos + params.pass_uv_offset.xy; + float importance = textureLod(source_importance, full_res_uv, 0.0).x; + + //Need to store obscurance from base pass + // load existing base values + vec4 base_values = imageLoad(source_ssil, ivec3(upos, params.pass)); + weight_sum += imageLoad(edges_weights_image, ivec2(upos)).r * float(SSIL_ADAPTIVE_TAP_BASE_COUNT * 4.0); + color_sum += (base_values.rgb) * weight_sum; + obscurance_sum += (base_values.a) * weight_sum; + + // increase importance around edges + float edge_count = dot(1.0 - edgesLRTB, vec4(1.0, 1.0, 1.0, 1.0)); + + float avg_total_importance = float(counter.sum) * params.load_counter_avg_div; + + float importance_limiter = clamp(params.adaptive_sample_limit / avg_total_importance, 0.0, 1.0); + importance *= importance_limiter; + + float additional_sample_count = SSIL_ADAPTIVE_TAP_FLEXIBLE_COUNT * importance; + + const float blend_range = 3.0; + const float blend_range_inv = 1.0 / blend_range; + + additional_sample_count += 0.5; + uint additional_samples = uint(additional_sample_count); + uint additional_samples_to = min(SSIL_MAX_TAPS, additional_samples + SSIL_ADAPTIVE_TAP_BASE_COUNT); + + for (uint i = SSIL_ADAPTIVE_TAP_BASE_COUNT; i < additional_samples_to; i++) { + additional_sample_count -= 1.0f; + float weight_mod = clamp(additional_sample_count * blend_range_inv, 0.0, 1.0); + SSILTap(p_quality_level, color_sum, obscurance_sum, weight_sum, int(i), rot_scale_matrix, pix_center_pos, pixel_normal, normalized_screen_pos, mip_offset, fallof_sq, weight_mod, norm_xy, norm_xy_length); + } + } +#endif + + // Early out for adaptive base + if (p_adaptive_base) { + vec3 color = color_sum / weight_sum; + + r_color = color; + r_edges = vec4(0.0); + r_obscurance = obscurance_sum / weight_sum; + r_weight = weight_sum; + return; + } + + // Calculate weighted average + vec3 color = color_sum / weight_sum; + color /= 1.0 - dot(color, vec3(0.299, 0.587, 0.114)); + + // Calculate fadeout (1 close, gradient, 0 far) + float fade_out = clamp(pix_center_pos.z * params.fade_out_mul + params.fade_out_add, 0.0, 1.0); + + // Reduce the SSIL if we're on the edge to remove artifacts on edges (we don't care for the lower quality one) + if (!p_adaptive_base && (p_quality_level >= SSIL_DEPTH_BASED_EDGES_ENABLE_AT_QUALITY_PRESET)) { + // when there's more than 2 opposite edges, start fading out the occlusion to reduce aliasing artifacts + float edge_fadeout_factor = clamp((1.0 - edgesLRTB.x - edgesLRTB.y) * 0.35, 0.0, 1.0) + clamp((1.0 - edgesLRTB.z - edgesLRTB.w) * 0.35, 0.0, 1.0); + + fade_out *= clamp(1.0 - edge_fadeout_factor, 0.0, 1.0); + } + + color = params.intensity * color; + + color *= fade_out; + + // outputs! + r_color = color; + r_edges = edgesLRTB; // These are used to prevent blurring across edges, 1 means no edge, 0 means edge, 0.5 means half way there, etc. + r_obscurance = clamp((obscurance_sum / weight_sum) * params.intensity, 0.0, 1.0); + r_weight = weight_sum; +} + +void main() { + vec3 out_color; + float out_obscurance; + float out_weight; + vec4 out_edges; + ivec2 ssC = ivec2(gl_GlobalInvocationID.xy); + if (any(greaterThanEqual(ssC, params.screen_size))) { //too large, do nothing + return; + } + + vec2 uv = vec2(gl_GlobalInvocationID) + vec2(0.5); +#ifdef SSIL_BASE + generate_SSIL(out_color, out_edges, out_obscurance, out_weight, uv, params.quality, true); + + imageStore(dest_image, ssC, vec4(out_color, out_obscurance)); + imageStore(edges_weights_image, ssC, vec4(out_weight / (float(SSIL_ADAPTIVE_TAP_BASE_COUNT) * 4.0))); +#else + generate_SSIL(out_color, out_edges, out_obscurance, out_weight, uv, params.quality, false); // pass in quality levels + + imageStore(dest_image, ssC, vec4(out_color, out_obscurance)); + imageStore(edges_weights_image, ssC, vec4(pack_edges(out_edges))); +#endif +} diff --git a/servers/rendering/renderer_rd/shaders/ssil_blur.glsl b/servers/rendering/renderer_rd/shaders/ssil_blur.glsl new file mode 100644 index 0000000000..47c56571f6 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/ssil_blur.glsl @@ -0,0 +1,144 @@ +/////////////////////////////////////////////////////////////////////////////////////////////////////////////////////// +// Copyright (c) 2016, Intel Corporation +// Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated +// documentation files (the "Software"), to deal in the Software without restriction, including without limitation +// the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to +// permit persons to whom the Software is furnished to do so, subject to the following conditions: +// The above copyright notice and this permission notice shall be included in all copies or substantial portions of +// the Software. +// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO +// THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE +// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, +// TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE +// SOFTWARE. +/////////////////////////////////////////////////////////////////////////////////////////////////////////////////////// +// File changes (yyyy-mm-dd) +// 2016-09-07: filip.strugar@intel.com: first commit +// 2020-12-05: clayjohn: convert to Vulkan and Godot +// 2021-05-27: clayjohn: convert SSAO to SSIL +/////////////////////////////////////////////////////////////////////////////////////////////////////////////////////// + +#[compute] + +#version 450 + +#VERSION_DEFINES + +layout(local_size_x = 8, local_size_y = 8, local_size_z = 1) in; + +layout(set = 0, binding = 0) uniform sampler2D source_ssil; + +layout(rgba16, set = 1, binding = 0) uniform restrict writeonly image2D dest_image; + +layout(r8, set = 2, binding = 0) uniform restrict readonly image2D source_edges; + +layout(push_constant, std430) uniform Params { + float edge_sharpness; + float pad; + vec2 half_screen_pixel_size; +} +params; + +vec4 unpack_edges(float p_packed_val) { + uint packed_val = uint(p_packed_val * 255.5); + vec4 edgesLRTB; + edgesLRTB.x = float((packed_val >> 6) & 0x03) / 3.0; + edgesLRTB.y = float((packed_val >> 4) & 0x03) / 3.0; + edgesLRTB.z = float((packed_val >> 2) & 0x03) / 3.0; + edgesLRTB.w = float((packed_val >> 0) & 0x03) / 3.0; + + return clamp(edgesLRTB + params.edge_sharpness, 0.0, 1.0); +} + +void add_sample(vec4 p_ssil_value, float p_edge_value, inout vec4 r_sum, inout float r_sum_weight) { + float weight = p_edge_value; + + r_sum += (weight * p_ssil_value); + r_sum_weight += weight; +} + +#ifdef MODE_WIDE +vec4 sample_blurred_wide(ivec2 p_pos, vec2 p_coord) { + vec4 ssil_value = textureLodOffset(source_ssil, vec2(p_coord), 0.0, ivec2(0, 0)); + vec4 ssil_valueL = textureLodOffset(source_ssil, vec2(p_coord), 0.0, ivec2(-2, 0)); + vec4 ssil_valueT = textureLodOffset(source_ssil, vec2(p_coord), 0.0, ivec2(0, -2)); + vec4 ssil_valueR = textureLodOffset(source_ssil, vec2(p_coord), 0.0, ivec2(2, 0)); + vec4 ssil_valueB = textureLodOffset(source_ssil, vec2(p_coord), 0.0, ivec2(0, 2)); + + vec4 edgesLRTB = unpack_edges(imageLoad(source_edges, p_pos).r); + edgesLRTB.x *= unpack_edges(imageLoad(source_edges, p_pos + ivec2(-2, 0)).r).y; + edgesLRTB.z *= unpack_edges(imageLoad(source_edges, p_pos + ivec2(0, -2)).r).w; + edgesLRTB.y *= unpack_edges(imageLoad(source_edges, p_pos + ivec2(2, 0)).r).x; + edgesLRTB.w *= unpack_edges(imageLoad(source_edges, p_pos + ivec2(0, 2)).r).z; + + float sum_weight = 0.8; + vec4 sum = ssil_value * sum_weight; + + add_sample(ssil_valueL, edgesLRTB.x, sum, sum_weight); + add_sample(ssil_valueR, edgesLRTB.y, sum, sum_weight); + add_sample(ssil_valueT, edgesLRTB.z, sum, sum_weight); + add_sample(ssil_valueB, edgesLRTB.w, sum, sum_weight); + + vec4 ssil_avg = sum / sum_weight; + + ssil_value = ssil_avg; + + return ssil_value; +} +#endif + +#ifdef MODE_SMART +vec4 sample_blurred(ivec2 p_pos, vec2 p_coord) { + vec4 vC = textureLodOffset(source_ssil, vec2(p_coord), 0.0, ivec2(0, 0)); + vec4 vL = textureLodOffset(source_ssil, vec2(p_coord), 0.0, ivec2(-1, 0)); + vec4 vT = textureLodOffset(source_ssil, vec2(p_coord), 0.0, ivec2(0, -1)); + vec4 vR = textureLodOffset(source_ssil, vec2(p_coord), 0.0, ivec2(1, 0)); + vec4 vB = textureLodOffset(source_ssil, vec2(p_coord), 0.0, ivec2(0, 1)); + + float packed_edges = imageLoad(source_edges, p_pos).r; + vec4 edgesLRTB = unpack_edges(packed_edges); + + float sum_weight = 0.5; + vec4 sum = vC * sum_weight; + + add_sample(vL, edgesLRTB.x, sum, sum_weight); + add_sample(vR, edgesLRTB.y, sum, sum_weight); + add_sample(vT, edgesLRTB.z, sum, sum_weight); + add_sample(vB, edgesLRTB.w, sum, sum_weight); + + vec4 ssil_avg = sum / sum_weight; + + vec4 ssil_value = ssil_avg; + + return ssil_value; +} +#endif + +void main() { + // Pixel being shaded + ivec2 ssC = ivec2(gl_GlobalInvocationID.xy); + +#ifdef MODE_NON_SMART + + vec2 half_pixel = params.half_screen_pixel_size * 0.5; + + vec2 uv = (vec2(gl_GlobalInvocationID.xy) + vec2(0.5, 0.5)) * params.half_screen_pixel_size; + + vec4 centre = textureLod(source_ssil, uv, 0.0); + + vec4 value = textureLod(source_ssil, vec2(uv + vec2(-half_pixel.x * 3, -half_pixel.y)), 0.0) * 0.2; + value += textureLod(source_ssil, vec2(uv + vec2(+half_pixel.x, -half_pixel.y * 3)), 0.0) * 0.2; + value += textureLod(source_ssil, vec2(uv + vec2(-half_pixel.x, +half_pixel.y * 3)), 0.0) * 0.2; + value += textureLod(source_ssil, vec2(uv + vec2(+half_pixel.x * 3, +half_pixel.y)), 0.0) * 0.2; + + vec4 sampled = value + centre * 0.2; + +#else +#ifdef MODE_SMART + vec4 sampled = sample_blurred(ssC, (vec2(gl_GlobalInvocationID.xy) + vec2(0.5, 0.5)) * params.half_screen_pixel_size); +#else // MODE_WIDE + vec4 sampled = sample_blurred_wide(ssC, (vec2(gl_GlobalInvocationID.xy) + vec2(0.5, 0.5)) * params.half_screen_pixel_size); +#endif +#endif // MODE_NON_SMART + imageStore(dest_image, ssC, sampled); +} diff --git a/servers/rendering/renderer_rd/shaders/ssil_importance_map.glsl b/servers/rendering/renderer_rd/shaders/ssil_importance_map.glsl new file mode 100644 index 0000000000..6b6b02739d --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/ssil_importance_map.glsl @@ -0,0 +1,125 @@ +/////////////////////////////////////////////////////////////////////////////////////////////////////////////////////// +// Copyright (c) 2016, Intel Corporation +// Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated +// documentation files (the "Software"), to deal in the Software without restriction, including without limitation +// the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to +// permit persons to whom the Software is furnished to do so, subject to the following conditions: +// The above copyright notice and this permission notice shall be included in all copies or substantial portions of +// the Software. +// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO +// THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE +// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, +// TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE +// SOFTWARE. +/////////////////////////////////////////////////////////////////////////////////////////////////////////////////////// +// File changes (yyyy-mm-dd) +// 2016-09-07: filip.strugar@intel.com: first commit +// 2020-12-05: clayjohn: convert to Vulkan and Godot +// 2021-05-27: clayjohn: convert SSAO to SSIL +/////////////////////////////////////////////////////////////////////////////////////////////////////////////////////// + +#[compute] + +#version 450 + +#VERSION_DEFINES + +layout(local_size_x = 8, local_size_y = 8, local_size_z = 1) in; + +#ifdef GENERATE_MAP +layout(set = 0, binding = 0) uniform sampler2DArray source_texture; +#else +layout(set = 0, binding = 0) uniform sampler2D source_importance; +#endif +layout(r8, set = 1, binding = 0) uniform restrict writeonly image2D dest_image; + +#ifdef PROCESS_MAPB +layout(set = 2, binding = 0, std430) buffer Counter { + uint sum; +} +counter; +#endif + +layout(push_constant, std430) uniform Params { + vec2 half_screen_pixel_size; + float intensity; + float pad; +} +params; + +void main() { + // Pixel being shaded + ivec2 ssC = ivec2(gl_GlobalInvocationID.xy); + +#ifdef GENERATE_MAP + // importance map stuff + uvec2 base_position = ssC * 2; + + float avg = 0.0; + float minV = 1.0; + float maxV = 0.0; + for (int i = 0; i < 4; i++) { + vec3 value_a = texelFetch(source_texture, ivec3(base_position, i), 0).rgb * params.intensity; + vec3 value_b = texelFetch(source_texture, ivec3(base_position, i) + ivec3(0, 1, 0), 0).rgb * params.intensity; + vec3 value_c = texelFetch(source_texture, ivec3(base_position, i) + ivec3(1, 0, 0), 0).rgb * params.intensity; + vec3 value_d = texelFetch(source_texture, ivec3(base_position, i) + ivec3(1, 1, 0), 0).rgb * params.intensity; + + // Calculate luminance (black and white value) + float a = dot(value_a, vec3(0.2125, 0.7154, 0.0721)); + float b = dot(value_b, vec3(0.2125, 0.7154, 0.0721)); + float c = dot(value_c, vec3(0.2125, 0.7154, 0.0721)); + float d = dot(value_d, vec3(0.2125, 0.7154, 0.0721)); + + maxV = max(maxV, max(max(a, b), max(c, d))); + minV = min(minV, min(min(a, b), min(c, d))); + } + + float min_max_diff = maxV - minV; + + imageStore(dest_image, ssC, vec4(pow(clamp(min_max_diff * 2.0, 0.0, 1.0), 0.6))); +#endif + +#ifdef PROCESS_MAPA + vec2 uv = (vec2(ssC) + 0.5) * params.half_screen_pixel_size * 2.0; + + float centre = textureLod(source_importance, uv, 0.0).x; + + vec2 half_pixel = params.half_screen_pixel_size; + + vec4 vals; + vals.x = textureLod(source_importance, uv + vec2(-half_pixel.x * 3, -half_pixel.y), 0.0).x; + vals.y = textureLod(source_importance, uv + vec2(+half_pixel.x, -half_pixel.y * 3), 0.0).x; + vals.z = textureLod(source_importance, uv + vec2(+half_pixel.x * 3, +half_pixel.y), 0.0).x; + vals.w = textureLod(source_importance, uv + vec2(-half_pixel.x, +half_pixel.y * 3), 0.0).x; + + float avg = dot(vals, vec4(0.25, 0.25, 0.25, 0.25)); + + imageStore(dest_image, ssC, vec4(avg)); +#endif + +#ifdef PROCESS_MAPB + vec2 uv = (vec2(ssC) + 0.5f) * params.half_screen_pixel_size * 2.0; + + float centre = textureLod(source_importance, uv, 0.0).x; + + vec2 half_pixel = params.half_screen_pixel_size; + + vec4 vals; + vals.x = textureLod(source_importance, uv + vec2(-half_pixel.x, -half_pixel.y * 3), 0.0).x; + vals.y = textureLod(source_importance, uv + vec2(+half_pixel.x * 3, -half_pixel.y), 0.0).x; + vals.z = textureLod(source_importance, uv + vec2(+half_pixel.x, +half_pixel.y * 3), 0.0).x; + vals.w = textureLod(source_importance, uv + vec2(-half_pixel.x * 3, +half_pixel.y), 0.0).x; + + float avg = dot(vals, vec4(0.25, 0.25, 0.25, 0.25)); + + imageStore(dest_image, ssC, vec4(avg)); + + // sum the average; to avoid overflowing we assume max AO resolution is not bigger than 16384x16384; so quarter res (used here) will be 4096x4096, which leaves us with 8 bits per pixel + uint sum = uint(clamp(avg, 0.0, 1.0) * 255.0 + 0.5); + + // save every 9th to avoid InterlockedAdd congestion - since we're blurring, this is good enough; compensated by multiplying load_counter_avg_div by 9 + if (((ssC.x % 3) + (ssC.y % 3)) == 0) { + atomicAdd(counter.sum, sum); + } +#endif +} diff --git a/servers/rendering/renderer_rd/shaders/ssil_interleave.glsl b/servers/rendering/renderer_rd/shaders/ssil_interleave.glsl new file mode 100644 index 0000000000..9e86ac0cf0 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/ssil_interleave.glsl @@ -0,0 +1,122 @@ +/////////////////////////////////////////////////////////////////////////////////////////////////////////////////////// +// Copyright (c) 2016, Intel Corporation +// Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated +// documentation files (the "Software"), to deal in the Software without restriction, including without limitation +// the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to +// permit persons to whom the Software is furnished to do so, subject to the following conditions: +// The above copyright notice and this permission notice shall be included in all copies or substantial portions of +// the Software. +// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO +// THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE +// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, +// TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE +// SOFTWARE. +/////////////////////////////////////////////////////////////////////////////////////////////////////////////////////// +// File changes (yyyy-mm-dd) +// 2016-09-07: filip.strugar@intel.com: first commit +// 2020-12-05: clayjohn: convert to Vulkan and Godot +// 2021-05-27: clayjohn: convert SSAO to SSIL +/////////////////////////////////////////////////////////////////////////////////////////////////////////////////////// + +#[compute] + +#version 450 + +#VERSION_DEFINES + +layout(local_size_x = 8, local_size_y = 8, local_size_z = 1) in; + +layout(rgba16, set = 0, binding = 0) uniform restrict writeonly image2D dest_image; +layout(set = 1, binding = 0) uniform sampler2DArray source_texture; +layout(r8, set = 2, binding = 0) uniform restrict readonly image2DArray source_edges; + +layout(push_constant, std430) uniform Params { + float inv_sharpness; + uint size_modifier; + vec2 pixel_size; +} +params; + +vec4 unpack_edges(float p_packed_val) { + uint packed_val = uint(p_packed_val * 255.5); + vec4 edgesLRTB; + edgesLRTB.x = float((packed_val >> 6) & 0x03) / 3.0; + edgesLRTB.y = float((packed_val >> 4) & 0x03) / 3.0; + edgesLRTB.z = float((packed_val >> 2) & 0x03) / 3.0; + edgesLRTB.w = float((packed_val >> 0) & 0x03) / 3.0; + + return clamp(edgesLRTB + params.inv_sharpness, 0.0, 1.0); +} + +void main() { + ivec2 ssC = ivec2(gl_GlobalInvocationID.xy); + if (any(greaterThanEqual(ssC, ivec2(1.0 / params.pixel_size)))) { //too large, do nothing + return; + } + +#ifdef MODE_SMART + uvec2 pix_pos = uvec2(gl_GlobalInvocationID.xy); + vec2 uv = (gl_GlobalInvocationID.xy + vec2(0.5)) * params.pixel_size; + + // calculate index in the four deinterleaved source array texture + int mx = int(pix_pos.x % 2); + int my = int(pix_pos.y % 2); + int index_center = mx + my * 2; // center index + int index_horizontal = (1 - mx) + my * 2; // neighbouring, horizontal + int index_vertical = mx + (1 - my) * 2; // neighbouring, vertical + int index_diagonal = (1 - mx) + (1 - my) * 2; // diagonal + + vec4 color = texelFetch(source_texture, ivec3(pix_pos / uvec2(params.size_modifier), index_center), 0); + + vec4 edgesLRTB = unpack_edges(imageLoad(source_edges, ivec3(pix_pos / uvec2(params.size_modifier), index_center)).r); + + // convert index shifts to sampling offsets + float fmx = float(mx); + float fmy = float(my); + + // in case of an edge, push sampling offsets away from the edge (towards pixel center) + float fmxe = (edgesLRTB.y - edgesLRTB.x); + float fmye = (edgesLRTB.w - edgesLRTB.z); + + // calculate final sampling offsets and sample using bilinear filter + vec2 uv_horizontal = (gl_GlobalInvocationID.xy + vec2(0.5) + vec2(fmx + fmxe - 0.5, 0.5 - fmy)) * params.pixel_size; + vec4 color_horizontal = textureLod(source_texture, vec3(uv_horizontal, index_horizontal), 0.0); + vec2 uv_vertical = (gl_GlobalInvocationID.xy + vec2(0.5) + vec2(0.5 - fmx, fmy - 0.5 + fmye)) * params.pixel_size; + vec4 color_vertical = textureLod(source_texture, vec3(uv_vertical, index_vertical), 0.0); + vec2 uv_diagonal = (gl_GlobalInvocationID.xy + vec2(0.5) + vec2(fmx - 0.5 + fmxe, fmy - 0.5 + fmye)) * params.pixel_size; + vec4 color_diagonal = textureLod(source_texture, vec3(uv_diagonal, index_diagonal), 0.0); + + // reduce weight for samples near edge - if the edge is on both sides, weight goes to 0 + vec4 blendWeights; + blendWeights.x = 1.0; + blendWeights.y = (edgesLRTB.x + edgesLRTB.y) * 0.5; + blendWeights.z = (edgesLRTB.z + edgesLRTB.w) * 0.5; + blendWeights.w = (blendWeights.y + blendWeights.z) * 0.5; + + // calculate weighted average + float blendWeightsSum = dot(blendWeights, vec4(1.0, 1.0, 1.0, 1.0)); + color += color_horizontal * blendWeights.y; + color += color_vertical * blendWeights.z; + color += color_diagonal * blendWeights.w; + color /= blendWeightsSum; + + imageStore(dest_image, ivec2(gl_GlobalInvocationID.xy), color); +#else // !MODE_SMART + + vec2 uv = (gl_GlobalInvocationID.xy + vec2(0.5)) * params.pixel_size; +#ifdef MODE_HALF + vec4 a = textureLod(source_texture, vec3(uv, 0), 0.0); + vec4 d = textureLod(source_texture, vec3(uv, 3), 0.0); + vec4 avg = (a + d) * 0.5; + +#else + vec4 a = textureLod(source_texture, vec3(uv, 0), 0.0); + vec4 b = textureLod(source_texture, vec3(uv, 1), 0.0); + vec4 c = textureLod(source_texture, vec3(uv, 2), 0.0); + vec4 d = textureLod(source_texture, vec3(uv, 3), 0.0); + vec4 avg = (a + b + c + d) * 0.25; + +#endif + imageStore(dest_image, ivec2(gl_GlobalInvocationID.xy), avg); +#endif +} diff --git a/servers/rendering/renderer_rd/shaders/subsurface_scattering.glsl b/servers/rendering/renderer_rd/shaders/subsurface_scattering.glsl index 88a953562f..fb35d3cde6 100644 --- a/servers/rendering/renderer_rd/shaders/subsurface_scattering.glsl +++ b/servers/rendering/renderer_rd/shaders/subsurface_scattering.glsl @@ -2,7 +2,7 @@ #version 450 -VERSION_DEFINES +#VERSION_DEFINES layout(local_size_x = 8, local_size_y = 8, local_size_z = 1) in; @@ -87,7 +87,7 @@ const vec4 skin_kernel[kernel_size] = vec4[]( #endif //USE_11_SAMPLES -layout(push_constant, binding = 1, std430) uniform Params { +layout(push_constant, std430) uniform Params { ivec2 screen_size; float camera_z_far; float camera_z_near; diff --git a/servers/rendering/renderer_rd/shaders/taa_resolve.glsl b/servers/rendering/renderer_rd/shaders/taa_resolve.glsl new file mode 100644 index 0000000000..a1a77b95aa --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/taa_resolve.glsl @@ -0,0 +1,393 @@ +/////////////////////////////////////////////////////////////////////////////////// +// Copyright(c) 2016-2022 Panos Karabelas +// +// Permission is hereby granted, free of charge, to any person obtaining a copy +// of this software and associated documentation files (the "Software"), to deal +// in the Software without restriction, including without limitation the rights +// to use, copy, modify, merge, publish, distribute, sublicense, and / or sell +// copies of the Software, and to permit persons to whom the Software is furnished +// to do so, subject to the following conditions : +// +// The above copyright notice and this permission notice shall be included in +// all copies or substantial portions of the Software. +// +// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR +// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS +// FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.IN NO EVENT SHALL THE AUTHORS OR +// COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER +// IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN +// CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. +/////////////////////////////////////////////////////////////////////////////////// +// File changes (yyyy-mm-dd) +// 2022-05-06: Panos Karabelas: first commit +// 2020-12-05: Joan Fons: convert to Vulkan and Godot +/////////////////////////////////////////////////////////////////////////////////// + +#[compute] + +#version 450 + +#VERSION_DEFINES + +// Based on Spartan Engine's TAA implementation https://github.com/PanosK92/SpartanEngine/blob/master/Data/shaders/temporal_antialiasing.hlsl + +#define USE_SUBGROUPS + +#define GROUP_SIZE 8 +#define FLT_MIN 0.00000001 +#define FLT_MAX 32767.0 +#define RPC_9 0.11111111111 +#define RPC_16 0.0625 + +#ifdef USE_SUBGROUPS +layout(local_size_x = GROUP_SIZE, local_size_y = GROUP_SIZE, local_size_z = 1) in; +#endif + +layout(rgba16f, set = 0, binding = 0) uniform restrict readonly image2D color_buffer; +layout(set = 0, binding = 1) uniform sampler2D depth_buffer; +layout(rg16f, set = 0, binding = 2) uniform restrict readonly image2D velocity_buffer; +layout(rg16f, set = 0, binding = 3) uniform restrict readonly image2D last_velocity_buffer; +layout(set = 0, binding = 4) uniform sampler2D history_buffer; +layout(rgba16f, set = 0, binding = 5) uniform restrict writeonly image2D output_buffer; + +layout(push_constant, std430) uniform Params { + vec2 resolution; + float disocclusion_threshold; // 0.1 / max(params.resolution.x, params.resolution.y + float disocclusion_scale; +} +params; + +const ivec2 kOffsets3x3[9] = { + ivec2(-1, -1), + ivec2(0, -1), + ivec2(1, -1), + ivec2(-1, 0), + ivec2(0, 0), + ivec2(1, 0), + ivec2(-1, 1), + ivec2(0, 1), + ivec2(1, 1), +}; + +/*------------------------------------------------------------------------------ + THREAD GROUP SHARED MEMORY (LDS) +------------------------------------------------------------------------------*/ + +const int kBorderSize = 1; +const int kGroupSize = GROUP_SIZE; +const int kTileDimension = kGroupSize + kBorderSize * 2; +const int kTileDimension2 = kTileDimension * kTileDimension; + +vec3 reinhard(vec3 hdr) { + return hdr / (hdr + 1.0); +} +vec3 reinhard_inverse(vec3 sdr) { + return sdr / (1.0 - sdr); +} + +float get_depth(ivec2 thread_id) { + return texelFetch(depth_buffer, thread_id, 0).r; +} + +#ifdef USE_SUBGROUPS +shared vec3 tile_color[kTileDimension][kTileDimension]; +shared float tile_depth[kTileDimension][kTileDimension]; + +vec3 load_color(uvec2 group_thread_id) { + group_thread_id += kBorderSize; + return tile_color[group_thread_id.x][group_thread_id.y]; +} + +void store_color(uvec2 group_thread_id, vec3 color) { + tile_color[group_thread_id.x][group_thread_id.y] = color; +} + +float load_depth(uvec2 group_thread_id) { + group_thread_id += kBorderSize; + return tile_depth[group_thread_id.x][group_thread_id.y]; +} + +void store_depth(uvec2 group_thread_id, float depth) { + tile_depth[group_thread_id.x][group_thread_id.y] = depth; +} + +void store_color_depth(uvec2 group_thread_id, ivec2 thread_id) { + // out of bounds clamp + thread_id = clamp(thread_id, ivec2(0, 0), ivec2(params.resolution) - ivec2(1, 1)); + + store_color(group_thread_id, imageLoad(color_buffer, thread_id).rgb); + store_depth(group_thread_id, get_depth(thread_id)); +} + +void populate_group_shared_memory(uvec2 group_id, uint group_index) { + // Populate group shared memory + ivec2 group_top_left = ivec2(group_id) * kGroupSize - kBorderSize; + if (group_index < (kTileDimension2 >> 2)) { + ivec2 group_thread_id_1 = ivec2(group_index % kTileDimension, group_index / kTileDimension); + ivec2 group_thread_id_2 = ivec2((group_index + (kTileDimension2 >> 2)) % kTileDimension, (group_index + (kTileDimension2 >> 2)) / kTileDimension); + ivec2 group_thread_id_3 = ivec2((group_index + (kTileDimension2 >> 1)) % kTileDimension, (group_index + (kTileDimension2 >> 1)) / kTileDimension); + ivec2 group_thread_id_4 = ivec2((group_index + kTileDimension2 * 3 / 4) % kTileDimension, (group_index + kTileDimension2 * 3 / 4) / kTileDimension); + + store_color_depth(group_thread_id_1, group_top_left + group_thread_id_1); + store_color_depth(group_thread_id_2, group_top_left + group_thread_id_2); + store_color_depth(group_thread_id_3, group_top_left + group_thread_id_3); + store_color_depth(group_thread_id_4, group_top_left + group_thread_id_4); + } + + // Wait for group threads to load store data. + groupMemoryBarrier(); + barrier(); +} +#else +vec3 load_color(uvec2 screen_pos) { + return imageLoad(color_buffer, ivec2(screen_pos)).rgb; +} + +float load_depth(uvec2 screen_pos) { + return get_depth(ivec2(screen_pos)); +} +#endif + +/*------------------------------------------------------------------------------ + VELOCITY +------------------------------------------------------------------------------*/ + +void depth_test_min(uvec2 pos, inout float min_depth, inout uvec2 min_pos) { + float depth = load_depth(pos); + + if (depth < min_depth) { + min_depth = depth; + min_pos = pos; + } +} + +// Returns velocity with closest depth (3x3 neighborhood) +void get_closest_pixel_velocity_3x3(in uvec2 group_pos, uvec2 group_top_left, out vec2 velocity) { + float min_depth = 1.0; + uvec2 min_pos = group_pos; + + depth_test_min(group_pos + kOffsets3x3[0], min_depth, min_pos); + depth_test_min(group_pos + kOffsets3x3[1], min_depth, min_pos); + depth_test_min(group_pos + kOffsets3x3[2], min_depth, min_pos); + depth_test_min(group_pos + kOffsets3x3[3], min_depth, min_pos); + depth_test_min(group_pos + kOffsets3x3[4], min_depth, min_pos); + depth_test_min(group_pos + kOffsets3x3[5], min_depth, min_pos); + depth_test_min(group_pos + kOffsets3x3[6], min_depth, min_pos); + depth_test_min(group_pos + kOffsets3x3[7], min_depth, min_pos); + depth_test_min(group_pos + kOffsets3x3[8], min_depth, min_pos); + + // Velocity out + velocity = imageLoad(velocity_buffer, ivec2(group_top_left + min_pos)).xy; +} + +/*------------------------------------------------------------------------------ + HISTORY SAMPLING +------------------------------------------------------------------------------*/ + +vec3 sample_catmull_rom_9(sampler2D stex, vec2 uv, vec2 resolution) { + // Source: https://gist.github.com/TheRealMJP/c83b8c0f46b63f3a88a5986f4fa982b1 + // License: https://gist.github.com/TheRealMJP/bc503b0b87b643d3505d41eab8b332ae + + // We're going to sample a a 4x4 grid of texels surrounding the target UV coordinate. We'll do this by rounding + // down the sample location to get the exact center of our "starting" texel. The starting texel will be at + // location [1, 1] in the grid, where [0, 0] is the top left corner. + vec2 sample_pos = uv * resolution; + vec2 texPos1 = floor(sample_pos - 0.5f) + 0.5f; + + // Compute the fractional offset from our starting texel to our original sample location, which we'll + // feed into the Catmull-Rom spline function to get our filter weights. + vec2 f = sample_pos - texPos1; + + // Compute the Catmull-Rom weights using the fractional offset that we calculated earlier. + // These equations are pre-expanded based on our knowledge of where the texels will be located, + // which lets us avoid having to evaluate a piece-wise function. + vec2 w0 = f * (-0.5f + f * (1.0f - 0.5f * f)); + vec2 w1 = 1.0f + f * f * (-2.5f + 1.5f * f); + vec2 w2 = f * (0.5f + f * (2.0f - 1.5f * f)); + vec2 w3 = f * f * (-0.5f + 0.5f * f); + + // Work out weighting factors and sampling offsets that will let us use bilinear filtering to + // simultaneously evaluate the middle 2 samples from the 4x4 grid. + vec2 w12 = w1 + w2; + vec2 offset12 = w2 / (w1 + w2); + + // Compute the final UV coordinates we'll use for sampling the texture + vec2 texPos0 = texPos1 - 1.0f; + vec2 texPos3 = texPos1 + 2.0f; + vec2 texPos12 = texPos1 + offset12; + + texPos0 /= resolution; + texPos3 /= resolution; + texPos12 /= resolution; + + vec3 result = vec3(0.0f, 0.0f, 0.0f); + + result += textureLod(stex, vec2(texPos0.x, texPos0.y), 0.0).xyz * w0.x * w0.y; + result += textureLod(stex, vec2(texPos12.x, texPos0.y), 0.0).xyz * w12.x * w0.y; + result += textureLod(stex, vec2(texPos3.x, texPos0.y), 0.0).xyz * w3.x * w0.y; + + result += textureLod(stex, vec2(texPos0.x, texPos12.y), 0.0).xyz * w0.x * w12.y; + result += textureLod(stex, vec2(texPos12.x, texPos12.y), 0.0).xyz * w12.x * w12.y; + result += textureLod(stex, vec2(texPos3.x, texPos12.y), 0.0).xyz * w3.x * w12.y; + + result += textureLod(stex, vec2(texPos0.x, texPos3.y), 0.0).xyz * w0.x * w3.y; + result += textureLod(stex, vec2(texPos12.x, texPos3.y), 0.0).xyz * w12.x * w3.y; + result += textureLod(stex, vec2(texPos3.x, texPos3.y), 0.0).xyz * w3.x * w3.y; + + return max(result, 0.0f); +} + +/*------------------------------------------------------------------------------ + HISTORY CLIPPING +------------------------------------------------------------------------------*/ + +// Based on "Temporal Reprojection Anti-Aliasing" - https://github.com/playdeadgames/temporal +vec3 clip_aabb(vec3 aabb_min, vec3 aabb_max, vec3 p, vec3 q) { + vec3 r = q - p; + vec3 rmax = (aabb_max - p.xyz); + vec3 rmin = (aabb_min - p.xyz); + + if (r.x > rmax.x + FLT_MIN) + r *= (rmax.x / r.x); + if (r.y > rmax.y + FLT_MIN) + r *= (rmax.y / r.y); + if (r.z > rmax.z + FLT_MIN) + r *= (rmax.z / r.z); + + if (r.x < rmin.x - FLT_MIN) + r *= (rmin.x / r.x); + if (r.y < rmin.y - FLT_MIN) + r *= (rmin.y / r.y); + if (r.z < rmin.z - FLT_MIN) + r *= (rmin.z / r.z); + + return p + r; +} + +// Clip history to the neighbourhood of the current sample +vec3 clip_history_3x3(uvec2 group_pos, vec3 color_history, vec2 velocity_closest) { + // Sample a 3x3 neighbourhood + vec3 s1 = load_color(group_pos + kOffsets3x3[0]); + vec3 s2 = load_color(group_pos + kOffsets3x3[1]); + vec3 s3 = load_color(group_pos + kOffsets3x3[2]); + vec3 s4 = load_color(group_pos + kOffsets3x3[3]); + vec3 s5 = load_color(group_pos + kOffsets3x3[4]); + vec3 s6 = load_color(group_pos + kOffsets3x3[5]); + vec3 s7 = load_color(group_pos + kOffsets3x3[6]); + vec3 s8 = load_color(group_pos + kOffsets3x3[7]); + vec3 s9 = load_color(group_pos + kOffsets3x3[8]); + + // Compute min and max (with an adaptive box size, which greatly reduces ghosting) + vec3 color_avg = (s1 + s2 + s3 + s4 + s5 + s6 + s7 + s8 + s9) * RPC_9; + vec3 color_avg2 = ((s1 * s1) + (s2 * s2) + (s3 * s3) + (s4 * s4) + (s5 * s5) + (s6 * s6) + (s7 * s7) + (s8 * s8) + (s9 * s9)) * RPC_9; + float box_size = mix(0.0f, 2.5f, smoothstep(0.02f, 0.0f, length(velocity_closest))); + vec3 dev = sqrt(abs(color_avg2 - (color_avg * color_avg))) * box_size; + vec3 color_min = color_avg - dev; + vec3 color_max = color_avg + dev; + + // Variance clipping + vec3 color = clip_aabb(color_min, color_max, clamp(color_avg, color_min, color_max), color_history); + + // Clamp to prevent NaNs + color = clamp(color, FLT_MIN, FLT_MAX); + + return color; +} + +/*------------------------------------------------------------------------------ + TAA +------------------------------------------------------------------------------*/ + +const vec3 lumCoeff = vec3(0.299f, 0.587f, 0.114f); + +float luminance(vec3 color) { + return max(dot(color, lumCoeff), 0.0001f); +} + +float get_factor_disocclusion(vec2 uv_reprojected, vec2 velocity) { + vec2 velocity_previous = imageLoad(last_velocity_buffer, ivec2(uv_reprojected * params.resolution)).xy; + vec2 velocity_texels = velocity * params.resolution; + vec2 prev_velocity_texels = velocity_previous * params.resolution; + float disocclusion = length(prev_velocity_texels - velocity_texels) - params.disocclusion_threshold; + return clamp(disocclusion * params.disocclusion_scale, 0.0, 1.0); +} + +vec3 temporal_antialiasing(uvec2 pos_group_top_left, uvec2 pos_group, uvec2 pos_screen, vec2 uv, sampler2D tex_history) { + // Get the velocity of the current pixel + vec2 velocity = imageLoad(velocity_buffer, ivec2(pos_screen)).xy; + + // Get reprojected uv + vec2 uv_reprojected = uv - velocity; + + // Get input color + vec3 color_input = load_color(pos_group); + + // Get history color (catmull-rom reduces a lot of the blurring that you get under motion) + vec3 color_history = sample_catmull_rom_9(tex_history, uv_reprojected, params.resolution).rgb; + + // Clip history to the neighbourhood of the current sample (fixes a lot of the ghosting). + vec2 velocity_closest = vec2(0.0); // This is best done by using the velocity with the closest depth. + get_closest_pixel_velocity_3x3(pos_group, pos_group_top_left, velocity_closest); + color_history = clip_history_3x3(pos_group, color_history, velocity_closest); + + // Compute blend factor + float blend_factor = RPC_16; // We want to be able to accumulate as many jitter samples as we generated, that is, 16. + { + // If re-projected UV is out of screen, converge to current color immediatel + float factor_screen = any(lessThan(uv_reprojected, vec2(0.0))) || any(greaterThan(uv_reprojected, vec2(1.0))) ? 1.0 : 0.0; + + // Increase blend factor when there is disocclusion (fixes a lot of the remaining ghosting). + float factor_disocclusion = get_factor_disocclusion(uv_reprojected, velocity); + + // Add to the blend factor + blend_factor = clamp(blend_factor + factor_screen + factor_disocclusion, 0.0, 1.0); + } + + // Resolve + vec3 color_resolved = vec3(0.0); + { + // Tonemap + color_history = reinhard(color_history); + color_input = reinhard(color_input); + + // Reduce flickering + float lum_color = luminance(color_input); + float lum_history = luminance(color_history); + float diff = abs(lum_color - lum_history) / max(lum_color, max(lum_history, 1.001)); + diff = 1.0 - diff; + diff = diff * diff; + blend_factor = mix(0.0, blend_factor, diff); + + // Lerp/blend + color_resolved = mix(color_history, color_input, blend_factor); + + // Inverse tonemap + color_resolved = reinhard_inverse(color_resolved); + } + + return color_resolved; +} + +void main() { +#ifdef USE_SUBGROUPS + populate_group_shared_memory(gl_WorkGroupID.xy, gl_LocalInvocationIndex); +#endif + + // Out of bounds check + if (any(greaterThanEqual(vec2(gl_GlobalInvocationID.xy), params.resolution))) { + return; + } + +#ifdef USE_SUBGROUPS + const uvec2 pos_group = gl_LocalInvocationID.xy; + const uvec2 pos_group_top_left = gl_WorkGroupID.xy * kGroupSize - kBorderSize; +#else + const uvec2 pos_group = gl_GlobalInvocationID.xy; + const uvec2 pos_group_top_left = uvec2(0, 0); +#endif + const uvec2 pos_screen = gl_GlobalInvocationID.xy; + const vec2 uv = (gl_GlobalInvocationID.xy + 0.5f) / params.resolution; + + vec3 result = temporal_antialiasing(pos_group_top_left, pos_group, pos_screen, uv, history_buffer); + imageStore(output_buffer, ivec2(gl_GlobalInvocationID.xy), vec4(result, 1.0)); +} diff --git a/servers/rendering/renderer_rd/shaders/volumetric_fog.glsl b/servers/rendering/renderer_rd/shaders/volumetric_fog.glsl index e7ba8feb80..eee609fb48 100644 --- a/servers/rendering/renderer_rd/shaders/volumetric_fog.glsl +++ b/servers/rendering/renderer_rd/shaders/volumetric_fog.glsl @@ -2,220 +2,105 @@ #version 450 -VERSION_DEFINES - -/* Do not use subgroups here, seems there is not much advantage and causes glitches -#extension GL_KHR_shader_subgroup_ballot: enable -#extension GL_KHR_shader_subgroup_arithmetic: enable - -#if defined(GL_KHR_shader_subgroup_ballot) && defined(GL_KHR_shader_subgroup_arithmetic) -#define USE_SUBGROUPS -#endif -*/ - -#if defined(MODE_FOG) || defined(MODE_FILTER) - -layout(local_size_x = 8, local_size_y = 8, local_size_z = 1) in; - -#endif - -#if defined(MODE_DENSITY) +#VERSION_DEFINES layout(local_size_x = 4, local_size_y = 4, local_size_z = 4) in; -#endif +#define SAMPLER_NEAREST_CLAMP 0 +#define SAMPLER_LINEAR_CLAMP 1 +#define SAMPLER_NEAREST_WITH_MIPMAPS_CLAMP 2 +#define SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP 3 +#define SAMPLER_NEAREST_WITH_MIPMAPS_ANISOTROPIC_CLAMP 4 +#define SAMPLER_LINEAR_WITH_MIPMAPS_ANISOTROPIC_CLAMP 5 +#define SAMPLER_NEAREST_REPEAT 6 +#define SAMPLER_LINEAR_REPEAT 7 +#define SAMPLER_NEAREST_WITH_MIPMAPS_REPEAT 8 +#define SAMPLER_LINEAR_WITH_MIPMAPS_REPEAT 9 +#define SAMPLER_NEAREST_WITH_MIPMAPS_ANISOTROPIC_REPEAT 10 +#define SAMPLER_LINEAR_WITH_MIPMAPS_ANISOTROPIC_REPEAT 11 + +#define DENSITY_SCALE 1024.0 #include "cluster_data_inc.glsl" +#include "light_data_inc.glsl" #define M_PI 3.14159265359 -layout(set = 0, binding = 1) uniform texture2D shadow_atlas; -layout(set = 0, binding = 2) uniform texture2D directional_shadow_atlas; - -layout(set = 0, binding = 3, std430) restrict readonly buffer OmniLights { - LightData data[]; -} -omni_lights; - -layout(set = 0, binding = 4, std430) restrict readonly buffer SpotLights { - LightData data[]; -} -spot_lights; +layout(set = 0, binding = 1) uniform sampler material_samplers[12]; -layout(set = 0, binding = 5, std140) uniform DirectionalLights { - DirectionalLightData data[MAX_DIRECTIONAL_LIGHT_DATA_STRUCTS]; +layout(set = 0, binding = 2, std430) restrict readonly buffer GlobalVariableData { + vec4 data[]; } -directional_lights; - -layout(set = 0, binding = 6, std430) buffer restrict readonly ClusterBuffer { - uint data[]; -} -cluster_buffer; - -layout(set = 0, binding = 7) uniform sampler linear_sampler; - -#ifdef MODE_DENSITY -layout(rgba16f, set = 0, binding = 8) uniform restrict writeonly image3D density_map; -layout(rgba16f, set = 0, binding = 9) uniform restrict readonly image3D fog_map; //unused -#endif - -#ifdef MODE_FOG -layout(rgba16f, set = 0, binding = 8) uniform restrict readonly image3D density_map; -layout(rgba16f, set = 0, binding = 9) uniform restrict writeonly image3D fog_map; -#endif - -#ifdef MODE_FILTER -layout(rgba16f, set = 0, binding = 8) uniform restrict readonly image3D source_map; -layout(rgba16f, set = 0, binding = 9) uniform restrict writeonly image3D dest_map; -#endif - -layout(set = 0, binding = 10) uniform sampler shadow_sampler; - -#define MAX_GI_PROBES 8 - -struct GIProbeData { - mat4 xform; - vec3 bounds; - float dynamic_range; +global_variables; - float bias; - float normal_bias; - bool blend_ambient; - uint texture_slot; - - float anisotropy_strength; - float ambient_occlusion; - float ambient_occlusion_size; - uint mipmaps; -}; - -layout(set = 0, binding = 11, std140) uniform GIProbes { - GIProbeData data[MAX_GI_PROBES]; -} -gi_probes; - -layout(set = 0, binding = 12) uniform texture3D gi_probe_textures[MAX_GI_PROBES]; - -layout(set = 0, binding = 13) uniform sampler linear_sampler_with_mipmaps; - -#ifdef ENABLE_SDFGI - -// SDFGI Integration on set 1 -#define SDFGI_MAX_CASCADES 8 - -struct SDFGIProbeCascadeData { +layout(push_constant, std430) uniform Params { vec3 position; - float to_probe; - ivec3 probe_world_offset; - float to_cell; // 1/bounds * grid_size -}; - -layout(set = 1, binding = 0, std140) uniform SDFGI { - vec3 grid_size; - uint max_cascades; - - bool use_occlusion; - int probe_axis_size; - float probe_to_uvw; - float normal_bias; + float pad; - vec3 lightprobe_tex_pixel_size; - float energy; + vec3 extents; + float pad2; - vec3 lightprobe_uv_offset; - float y_mult; + ivec3 corner; + uint shape; - vec3 occlusion_clamp; - uint pad3; - - vec3 occlusion_renormalize; - uint pad4; - - vec3 cascade_probe_size; - uint pad5; - - SDFGIProbeCascadeData cascades[SDFGI_MAX_CASCADES]; + mat4 transform; } -sdfgi; - -layout(set = 1, binding = 1) uniform texture2DArray sdfgi_ambient_texture; - -layout(set = 1, binding = 2) uniform texture3D sdfgi_occlusion_texture; +params; -#endif //SDFGI +#ifdef MOLTENVK_USED +layout(set = 1, binding = 1) volatile buffer emissive_only_map_buffer { + uint emissive_only_map[]; +}; +#else +layout(r32ui, set = 1, binding = 1) uniform volatile uimage3D emissive_only_map; +#endif -layout(set = 0, binding = 14, std140) uniform Params { +layout(set = 1, binding = 2, std140) uniform SceneParams { vec2 fog_frustum_size_begin; vec2 fog_frustum_size_end; float fog_frustum_end; - float z_near; - float z_far; - int filter_axis; + float z_near; // + float z_far; // + float time; ivec3 fog_volume_size; - uint directional_light_count; - - vec3 light_color; - float base_density; - - float detail_spread; - float gi_inject; - uint max_gi_probes; - uint cluster_type_size; + uint directional_light_count; // - vec2 screen_size; - uint cluster_shift; - uint cluster_width; - - uint max_cluster_element_count_div_32; bool use_temporal_reprojection; uint temporal_frame; + float detail_spread; float temporal_blend; - mat3x4 cam_rotation; mat4 to_prev_view; + mat4 transform; } -params; - -layout(set = 0, binding = 15) uniform texture3D prev_density_texture; +scene_params; -float get_depth_at_pos(float cell_depth_size, int z) { - float d = float(z) * cell_depth_size + cell_depth_size * 0.5; //center of voxels - d = pow(d, params.detail_spread); - return params.fog_frustum_end * d; -} - -vec3 hash3f(uvec3 x) { - x = ((x >> 16) ^ x) * 0x45d9f3b; - x = ((x >> 16) ^ x) * 0x45d9f3b; - x = (x >> 16) ^ x; - return vec3(x & 0xFFFFF) / vec3(float(0xFFFFF)); -} - -float get_omni_attenuation(float distance, float inv_range, float decay) { - float nd = distance * inv_range; - nd *= nd; - nd *= nd; // nd^4 - nd = max(1.0 - nd, 0.0); - nd *= nd; // nd^2 - return nd * pow(max(distance, 0.0001), -decay); -} +#ifdef MOLTENVK_USED +layout(set = 1, binding = 3) volatile buffer density_only_map_buffer { + uint density_only_map[]; +}; +layout(set = 1, binding = 4) volatile buffer light_only_map_buffer { + uint light_only_map[]; +}; +#else +layout(r32ui, set = 1, binding = 3) uniform volatile uimage3D density_only_map; +layout(r32ui, set = 1, binding = 4) uniform volatile uimage3D light_only_map; +#endif -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_max = item_min_max >> 16; - ; +#ifdef MATERIAL_UNIFORMS_USED +layout(set = 2, binding = 0, std140) uniform MaterialUniforms{ +#MATERIAL_UNIFORMS +} material; +#endif - item_from = item_min >> 5; - item_to = (item_max == 0) ? 0 : ((item_max - 1) >> 5) + 1; //side effect of how it is stored, as item_max 0 means no elements -} +#GLOBALS -uint cluster_get_range_clip_mask(uint i, uint z_min, uint z_max) { - int local_min = clamp(int(z_min) - int(i) * 32, 0, 31); - int mask_width = min(int(z_max) - int(z_min), 32 - local_min); - return bitfieldInsert(uint(0), uint(0xFFFFFFFF), local_min, mask_width); +float get_depth_at_pos(float cell_depth_size, int z) { + float d = float(z) * cell_depth_size + cell_depth_size * 0.5; //center of voxels + d = pow(d, scene_params.detail_spread); + return scene_params.fog_frustum_end * d; } #define TEMPORAL_FRAMES 16 @@ -239,464 +124,186 @@ const vec3 halton_map[TEMPORAL_FRAMES] = vec3[]( vec3(0.03125, 0.59259259, 0.32)); void main() { - vec3 fog_cell_size = 1.0 / vec3(params.fog_volume_size); + vec3 fog_cell_size = 1.0 / vec3(scene_params.fog_volume_size); -#ifdef MODE_DENSITY - - ivec3 pos = ivec3(gl_GlobalInvocationID.xyz); - if (any(greaterThanEqual(pos, params.fog_volume_size))) { + ivec3 pos = ivec3(gl_GlobalInvocationID.xyz) + params.corner; + if (any(greaterThanEqual(pos, scene_params.fog_volume_size))) { return; //do not compute } +#ifdef MOLTENVK_USED + uint lpos = pos.z * scene_params.fog_volume_size.x * scene_params.fog_volume_size.y + pos.y * scene_params.fog_volume_size.x + pos.x; +#endif vec3 posf = vec3(pos); - //posf += mix(vec3(0.0),vec3(1.0),0.3) * hash3f(uvec3(pos)) * 2.0 - 1.0; - vec3 fog_unit_pos = posf * fog_cell_size + fog_cell_size * 0.5; //center of voxels - - uvec2 screen_pos = uvec2(fog_unit_pos.xy * params.screen_size); - uvec2 cluster_pos = screen_pos >> params.cluster_shift; - uint cluster_offset = (params.cluster_width * cluster_pos.y + cluster_pos.x) * (params.max_cluster_element_count_div_32 + 32); - //positions in screen are too spread apart, no hopes for optimizing with subgroups - - fog_unit_pos.z = pow(fog_unit_pos.z, params.detail_spread); + fog_unit_pos.z = pow(fog_unit_pos.z, scene_params.detail_spread); vec3 view_pos; - view_pos.xy = (fog_unit_pos.xy * 2.0 - 1.0) * mix(params.fog_frustum_size_begin, params.fog_frustum_size_end, vec2(fog_unit_pos.z)); - view_pos.z = -params.fog_frustum_end * fog_unit_pos.z; + view_pos.xy = (fog_unit_pos.xy * 2.0 - 1.0) * mix(scene_params.fog_frustum_size_begin, scene_params.fog_frustum_size_end, vec2(fog_unit_pos.z)); + view_pos.z = -scene_params.fog_frustum_end * fog_unit_pos.z; view_pos.y = -view_pos.y; - vec4 reprojected_density = vec4(0.0); - float reproject_amount = 0.0; - - if (params.use_temporal_reprojection) { - vec3 prev_view = (params.to_prev_view * vec4(view_pos, 1.0)).xyz; + if (scene_params.use_temporal_reprojection) { + vec3 prev_view = (scene_params.to_prev_view * vec4(view_pos, 1.0)).xyz; //undo transform into prev view prev_view.y = -prev_view.y; //z back to unit size - prev_view.z /= -params.fog_frustum_end; + prev_view.z /= -scene_params.fog_frustum_end; //xy back to unit size - prev_view.xy /= mix(params.fog_frustum_size_begin, params.fog_frustum_size_end, vec2(prev_view.z)); + prev_view.xy /= mix(scene_params.fog_frustum_size_begin, scene_params.fog_frustum_size_end, vec2(prev_view.z)); prev_view.xy = prev_view.xy * 0.5 + 0.5; //z back to unspread value - prev_view.z = pow(prev_view.z, 1.0 / params.detail_spread); + prev_view.z = pow(prev_view.z, 1.0 / scene_params.detail_spread); if (all(greaterThan(prev_view, vec3(0.0))) && all(lessThan(prev_view, vec3(1.0)))) { //reprojectinon fits - - reprojected_density = textureLod(sampler3D(prev_density_texture, linear_sampler), prev_view, 0.0); - reproject_amount = params.temporal_blend; - // Since we can reproject, now we must jitter the current view pos. // This is done here because cells that can't reproject should not jitter. - fog_unit_pos = posf * fog_cell_size + fog_cell_size * halton_map[params.temporal_frame]; //center of voxels, offset by halton table - - screen_pos = uvec2(fog_unit_pos.xy * params.screen_size); - cluster_pos = screen_pos >> params.cluster_shift; - cluster_offset = (params.cluster_width * cluster_pos.y + cluster_pos.x) * (params.max_cluster_element_count_div_32 + 32); - //positions in screen are too spread apart, no hopes for optimizing with subgroups + fog_unit_pos = posf * fog_cell_size + fog_cell_size * halton_map[scene_params.temporal_frame]; //center of voxels, offset by halton table + fog_unit_pos.z = pow(fog_unit_pos.z, scene_params.detail_spread); - fog_unit_pos.z = pow(fog_unit_pos.z, params.detail_spread); - - view_pos.xy = (fog_unit_pos.xy * 2.0 - 1.0) * mix(params.fog_frustum_size_begin, params.fog_frustum_size_end, vec2(fog_unit_pos.z)); - view_pos.z = -params.fog_frustum_end * fog_unit_pos.z; + view_pos.xy = (fog_unit_pos.xy * 2.0 - 1.0) * mix(scene_params.fog_frustum_size_begin, scene_params.fog_frustum_size_end, vec2(fog_unit_pos.z)); + view_pos.z = -scene_params.fog_frustum_end * fog_unit_pos.z; view_pos.y = -view_pos.y; } } - uint cluster_z = uint(clamp((abs(view_pos.z) / params.z_far) * 32.0, 0.0, 31.0)); - - vec3 total_light = params.light_color; + float density = 0.0; + vec3 emission = vec3(0.0); + vec3 albedo = vec3(0.0); - float total_density = params.base_density; float cell_depth_size = abs(view_pos.z - get_depth_at_pos(fog_cell_size.z, pos.z + 1)); - //compute directional lights - - 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 shadow_color = directional_lights.data[i].shadow_color1.rgb; - vec3 light_dir = directional_lights.data[i].direction; - vec4 v = vec4(view_pos, 1.0); - float z_range; - - 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 shadow = textureProj(sampler2DShadow(directional_shadow_atlas,shadow_sampler),pssm_coord); - float shadow = 0.0; - for(float xi=-1;xi<=1;xi++) { - for(float yi=-1;yi<=1;yi++) { - vec2 ofs = vec2(xi,yi) * 1.5 * params.directional_shadow_pixel_size; - shadow += textureProj(sampler2DShadow(directional_shadow_atlas,shadow_sampler),pssm_coord + vec4(ofs,0.0,0.0)); - } - - } - - shadow /= 3.0 * 3.0; -*/ - shadow = mix(shadow, 1.0, smoothstep(directional_lights.data[i].fade_from, directional_lights.data[i].fade_to, view_pos.z)); //done with negative values for performance - - shadow_attenuation = mix(shadow_color, vec3(1.0), shadow); - } - - total_light += shadow_attenuation * directional_lights.data[i].color * directional_lights.data[i].energy / M_PI; + vec4 world = scene_params.transform * vec4(view_pos, 1.0); + world.xyz /= world.w; + + vec3 uvw = fog_unit_pos; + + vec4 local_pos = params.transform * world; + local_pos.xyz /= local_pos.w; + + float sdf = -1.0; + if (params.shape == 0) { + // Ellipsoid + // https://www.shadertoy.com/view/tdS3DG + float k0 = length(local_pos.xyz / params.extents); + float k1 = length(local_pos.xyz / (params.extents * params.extents)); + sdf = k0 * (k0 - 1.0) / k1; + } else if (params.shape == 1) { + // Cone + // https://iquilezles.org/www/articles/distfunctions/distfunctions.htm + + // Compute the cone angle automatically to fit within the volume's extents. + float inv_height = 1.0 / max(0.001, params.extents.y); + float radius = 1.0 / max(0.001, (min(params.extents.x, params.extents.z) * 0.5)); + float hypotenuse = sqrt(radius * radius + inv_height * inv_height); + float rsin = radius / hypotenuse; + float rcos = inv_height / hypotenuse; + vec2 c = vec2(rsin, rcos); + + float q = length(local_pos.xz); + sdf = max(dot(c, vec2(q, local_pos.y - params.extents.y)), -params.extents.y - local_pos.y); + } else if (params.shape == 2) { + // Cylinder + // https://iquilezles.org/www/articles/distfunctions/distfunctions.htm + vec2 d = abs(vec2(length(local_pos.xz), local_pos.y)) - vec2(min(params.extents.x, params.extents.z), params.extents.y); + sdf = min(max(d.x, d.y), 0.0) + length(max(d, 0.0)); + } else if (params.shape == 3) { + // Box + // https://iquilezles.org/www/articles/distfunctions/distfunctions.htm + vec3 q = abs(local_pos.xyz) - params.extents; + sdf = length(max(q, 0.0)) + min(max(q.x, max(q.y, q.z)), 0.0); } - //compute lights from cluster - - { //omni lights - - uint cluster_omni_offset = cluster_offset; - - uint item_min; - uint item_max; - uint item_from; - uint item_to; - - cluster_get_item_range(cluster_omni_offset + params.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)); + float cull_mask = 1.0; //used to cull cells that do not contribute + if (params.shape <= 3) { +#ifndef SDF_USED + cull_mask = 1.0 - smoothstep(-0.1, 0.0, sdf); #endif + uvw = clamp((local_pos.xyz + params.extents) / (2.0 * params.extents), 0.0, 1.0); + } - for (uint i = item_from; i < item_to; i++) { - uint mask = cluster_buffer.data[cluster_omni_offset + i]; - mask &= cluster_get_range_clip_mask(i, item_min, item_max); -#ifdef USE_SUBGROUPS - uint merged_mask = subgroupBroadcastFirst(subgroupOr(mask)); + if (cull_mask > 0.0) { + { +#CODE : FOG + } + +#ifdef DENSITY_USED + density *= cull_mask; + if (abs(density) > 0.001) { + int final_density = int(density * DENSITY_SCALE); +#ifdef MOLTENVK_USED + atomicAdd(density_only_map[lpos], uint(final_density)); #else - uint merged_mask = mask; + imageAtomicAdd(density_only_map, pos, uint(final_density)); #endif - while (merged_mask != 0) { - uint bit = findMSB(merged_mask); - merged_mask &= ~(1 << bit); -#ifdef USE_SUBGROUPS - if (((1 << bit) & mask) == 0) { //do not process if not originally here - continue; - } +#ifdef EMISSION_USED + { + emission *= clamp(density, 0.0, 1.0); + emission = clamp(emission, vec3(0.0), vec3(4.0)); + // Scale to fit into R11G11B10 with a range of 0-4 + uvec3 emission_u = uvec3(emission.r * 511.0, emission.g * 511.0, emission.b * 255.0); + // R and G have 11 bits each and B has 10. Then pack them into a 32 bit uint + uint final_emission = emission_u.r << 21 | emission_u.g << 10 | emission_u.b; +#ifdef MOLTENVK_USED + uint prev_emission = atomicAdd(emissive_only_map[lpos], final_emission); +#else + uint prev_emission = imageAtomicAdd(emissive_only_map, pos, final_emission); #endif - uint light_index = 32 * i + bit; - - //if (!bool(omni_omni_lights.data[light_index].mask & draw_call.layer_mask)) { - // continue; //not masked - //} - - vec3 light_pos = omni_lights.data[light_index].position; - 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) { - 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 / M_PI; - - if (omni_lights.data[light_index].shadow_enabled) { - //has shadow - vec4 v = vec4(view_pos, 1.0); - - vec4 splane = (omni_lights.data[light_index].shadow_matrix * v); - float shadow_len = length(splane.xyz); //need to remember shadow len from here - splane.xyz = normalize(splane.xyz); - vec4 clamp_rect = omni_lights.data[light_index].atlas_rect; + // Adding can lead to colors overflowing, so validate + uvec3 prev_emission_u = uvec3(prev_emission >> 21, (prev_emission << 11) >> 21, prev_emission % 1024); + uint add_emission = final_emission + prev_emission; + uvec3 add_emission_u = uvec3(add_emission >> 21, (add_emission << 11) >> 21, add_emission % 1024); - if (splane.z >= 0.0) { - splane.z += 1.0; + bvec3 overflowing = lessThan(add_emission_u, prev_emission_u + emission_u); - clamp_rect.y += clamp_rect.w; - - } else { - splane.z = 1.0 - splane.z; - } - - splane.xy /= splane.z; - - splane.xy = splane.xy * 0.5 + 0.5; - splane.z = shadow_len * omni_lights.data[light_index].inv_radius; - splane.xy = clamp_rect.xy + splane.xy * clamp_rect.zw; - splane.w = 1.0; //needed? i think it should be 1 already - - float depth = texture(sampler2D(shadow_atlas, linear_sampler), splane.xy).r; - - shadow_attenuation = exp(min(0.0, (depth - splane.z)) / omni_lights.data[light_index].inv_radius * omni_lights.data[light_index].shadow_volumetric_fog_fade); - } - total_light += light * attenuation * shadow_attenuation; + if (any(overflowing)) { + uvec3 overflow_factor = mix(uvec3(0), uvec3(2047 << 21, 2047 << 10, 1023), overflowing); + uint force_max = overflow_factor.r | overflow_factor.g | overflow_factor.b; +#ifdef MOLTENVK_USED + atomicOr(emissive_only_map[lpos], force_max); +#else + imageAtomicOr(emissive_only_map, pos, force_max); +#endif } } - } - } - - { //spot lights - - uint cluster_spot_offset = cluster_offset + params.cluster_type_size; - - uint item_min; - uint item_max; - uint item_from; - uint item_to; - - cluster_get_item_range(cluster_spot_offset + params.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 - - for (uint i = item_from; i < item_to; i++) { - uint mask = cluster_buffer.data[cluster_spot_offset + i]; - mask &= cluster_get_range_clip_mask(i, item_min, item_max); -#ifdef USE_SUBGROUPS - uint merged_mask = subgroupBroadcastFirst(subgroupOr(mask)); +#ifdef ALBEDO_USED + { + vec3 scattering = albedo * clamp(density, 0.0, 1.0); + scattering = clamp(scattering, vec3(0.0), vec3(1.0)); + uvec3 scattering_u = uvec3(scattering.r * 2047.0, scattering.g * 2047.0, scattering.b * 1023.0); + // R and G have 11 bits each and B has 10. Then pack them into a 32 bit uint + uint final_scattering = scattering_u.r << 21 | scattering_u.g << 10 | scattering_u.b; +#ifdef MOLTENVK_USED + uint prev_scattering = atomicAdd(light_only_map[lpos], final_scattering); #else - uint merged_mask = mask; -#endif - - while (merged_mask != 0) { - uint bit = findMSB(merged_mask); - merged_mask &= ~(1 << bit); -#ifdef USE_SUBGROUPS - if (((1 << bit) & mask) == 0) { //do not process if not originally here - continue; - } + uint prev_scattering = imageAtomicAdd(light_only_map, pos, final_scattering); #endif - //if (!bool(omni_lights.data[light_index].mask & draw_call.layer_mask)) { - // continue; //not masked - //} - - uint light_index = 32 * i + bit; - - vec3 light_pos = spot_lights.data[light_index].position; - vec3 light_rel_vec = spot_lights.data[light_index].position - view_pos; - float d = length(light_rel_vec); - float shadow_attenuation = 1.0; - - if (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; - float scos = max(dot(-normalize(light_rel_vec), spot_dir), spot_lights.data[light_index].cone_angle); - float spot_rim = max(0.0001, (1.0 - scos) / (1.0 - spot_lights.data[light_index].cone_angle)); - attenuation *= 1.0 - pow(spot_rim, spot_lights.data[light_index].cone_attenuation); - - vec3 light = spot_lights.data[light_index].color / M_PI; - - if (spot_lights.data[light_index].shadow_enabled) { - //has shadow - vec4 v = vec4(view_pos, 1.0); - - vec4 splane = (spot_lights.data[light_index].shadow_matrix * v); - splane /= splane.w; - - float depth = texture(sampler2D(shadow_atlas, linear_sampler), splane.xy).r; - - shadow_attenuation = exp(min(0.0, (depth - splane.z)) / spot_lights.data[light_index].inv_radius * spot_lights.data[light_index].shadow_volumetric_fog_fade); - } - - total_light += light * attenuation * shadow_attenuation; - } - } - } - } - - vec3 world_pos = mat3(params.cam_rotation) * view_pos; - - for (uint i = 0; i < params.max_gi_probes; i++) { - vec3 position = (gi_probes.data[i].xform * vec4(world_pos, 1.0)).xyz; - - //this causes corrupted pixels, i have no idea why.. - if (all(bvec2(all(greaterThanEqual(position, vec3(0.0))), all(lessThan(position, gi_probes.data[i].bounds))))) { - position /= gi_probes.data[i].bounds; - - vec4 light = vec4(0.0); - for (uint j = 0; j < gi_probes.data[i].mipmaps; j++) { - vec4 slight = textureLod(sampler3D(gi_probe_textures[i], linear_sampler_with_mipmaps), position, float(j)); - float a = (1.0 - light.a); - light += a * slight; - } - - light.rgb *= gi_probes.data[i].dynamic_range * params.gi_inject; - - total_light += light.rgb; - } - } - - //sdfgi -#ifdef ENABLE_SDFGI + // Adding can lead to colors overflowing, so validate + uvec3 prev_scattering_u = uvec3(prev_scattering >> 21, (prev_scattering << 11) >> 21, prev_scattering % 1024); + uint add_scattering = final_scattering + prev_scattering; + uvec3 add_scattering_u = uvec3(add_scattering >> 21, (add_scattering << 11) >> 21, add_scattering % 1024); - { - float blend = -1.0; - vec3 ambient_total = vec3(0.0); + bvec3 overflowing = lessThan(add_scattering_u, prev_scattering_u + scattering_u); - for (uint i = 0; i < sdfgi.max_cascades; i++) { - vec3 cascade_pos = (world_pos - sdfgi.cascades[i].position) * sdfgi.cascades[i].to_probe; - - if (any(lessThan(cascade_pos, vec3(0.0))) || any(greaterThanEqual(cascade_pos, sdfgi.cascade_probe_size))) { - continue; //skip cascade - } - - vec3 base_pos = floor(cascade_pos); - ivec3 probe_base_pos = ivec3(base_pos); - - vec4 ambient_accum = vec4(0.0); - - ivec3 tex_pos = ivec3(probe_base_pos.xy, int(i)); - tex_pos.x += probe_base_pos.z * sdfgi.probe_axis_size; - - for (uint j = 0; j < 8; j++) { - ivec3 offset = (ivec3(j) >> ivec3(0, 1, 2)) & ivec3(1, 1, 1); - ivec3 probe_posi = probe_base_pos; - probe_posi += offset; - - // Compute weight - - vec3 probe_pos = vec3(probe_posi); - vec3 probe_to_pos = cascade_pos - probe_pos; - - vec3 trilinear = vec3(1.0) - abs(probe_to_pos); - float weight = trilinear.x * trilinear.y * trilinear.z; - - // Compute lightprobe occlusion - - if (sdfgi.use_occlusion) { - ivec3 occ_indexv = abs((sdfgi.cascades[i].probe_world_offset + probe_posi) & ivec3(1, 1, 1)) * ivec3(1, 2, 4); - vec4 occ_mask = mix(vec4(0.0), vec4(1.0), equal(ivec4(occ_indexv.x | occ_indexv.y), ivec4(0, 1, 2, 3))); - - vec3 occ_pos = clamp(cascade_pos, probe_pos - sdfgi.occlusion_clamp, probe_pos + sdfgi.occlusion_clamp) * sdfgi.probe_to_uvw; - occ_pos.z += float(i); - if (occ_indexv.z != 0) { //z bit is on, means index is >=4, so make it switch to the other half of textures - occ_pos.x += 1.0; - } - - occ_pos *= sdfgi.occlusion_renormalize; - float occlusion = dot(textureLod(sampler3D(sdfgi_occlusion_texture, linear_sampler), occ_pos, 0.0), occ_mask); - - weight *= max(occlusion, 0.01); + if (any(overflowing)) { + uvec3 overflow_factor = mix(uvec3(0), uvec3(2047 << 21, 2047 << 10, 1023), overflowing); + uint force_max = overflow_factor.r | overflow_factor.g | overflow_factor.b; +#ifdef MOLTENVK_USED + atomicOr(light_only_map[lpos], force_max); +#else + imageAtomicOr(light_only_map, pos, force_max); +#endif } - - // Compute ambient texture position - - ivec3 uvw = tex_pos; - uvw.xy += offset.xy; - uvw.x += offset.z * sdfgi.probe_axis_size; - - vec3 ambient = texelFetch(sampler2DArray(sdfgi_ambient_texture, linear_sampler), uvw, 0).rgb; - - ambient_accum.rgb += ambient * weight; - ambient_accum.a += weight; - } - - if (ambient_accum.a > 0) { - ambient_accum.rgb /= ambient_accum.a; } - ambient_total = ambient_accum.rgb; - break; +#endif // ALBEDO_USED } - - total_light += ambient_total * params.gi_inject; +#endif // DENSITY_USED } - -#endif - - vec4 final_density = vec4(total_light, total_density); - - final_density = mix(final_density, reprojected_density, reproject_amount); - - imageStore(density_map, pos, final_density); -#endif - -#ifdef MODE_FOG - - ivec3 pos = ivec3(gl_GlobalInvocationID.xy, 0); - - if (any(greaterThanEqual(pos, params.fog_volume_size))) { - return; //do not compute - } - - vec4 fog_accum = vec4(0.0); - float prev_z = 0.0; - - float t = 1.0; - - for (int i = 0; i < params.fog_volume_size.z; i++) { - //compute fog position - ivec3 fog_pos = pos + ivec3(0, 0, i); - //get fog value - vec4 fog = imageLoad(density_map, fog_pos); - - //get depth at cell pos - float z = get_depth_at_pos(fog_cell_size.z, i); - //get distance from previous pos - float d = abs(prev_z - z); - //compute exinction based on beer's - float extinction = t * exp(-d * fog.a); - //compute alpha based on different of extinctions - float alpha = t - extinction; - //update extinction - t = extinction; - - fog_accum += vec4(fog.rgb * alpha, alpha); - prev_z = z; - - vec4 fog_value; - - if (fog_accum.a > 0.0) { - fog_value = vec4(fog_accum.rgb / fog_accum.a, 1.0 - t); - } else { - fog_value = vec4(0.0); - } - - imageStore(fog_map, fog_pos, fog_value); - } - -#endif - -#ifdef MODE_FILTER - - ivec3 pos = ivec3(gl_GlobalInvocationID.xyz); - - const float gauss[7] = float[](0.071303, 0.131514, 0.189879, 0.214607, 0.189879, 0.131514, 0.071303); - - const ivec3 filter_dir[3] = ivec3[](ivec3(1, 0, 0), ivec3(0, 1, 0), ivec3(0, 0, 1)); - ivec3 offset = filter_dir[params.filter_axis]; - - vec4 accum = vec4(0.0); - for (int i = -3; i <= 3; i++) { - accum += imageLoad(source_map, clamp(pos + offset * i, ivec3(0), params.fog_volume_size - ivec3(1))) * gauss[i + 3]; - } - - imageStore(dest_map, pos, accum); - -#endif } diff --git a/servers/rendering/renderer_rd/shaders/volumetric_fog_process.glsl b/servers/rendering/renderer_rd/shaders/volumetric_fog_process.glsl new file mode 100644 index 0000000000..347fd13b28 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/volumetric_fog_process.glsl @@ -0,0 +1,783 @@ +#[compute] + +#version 450 + +#VERSION_DEFINES + +/* Do not use subgroups here, seems there is not much advantage and causes glitches +#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 +*/ + +#ifdef MODE_DENSITY +layout(local_size_x = 4, local_size_y = 4, local_size_z = 4) in; +#else +layout(local_size_x = 8, local_size_y = 8, local_size_z = 1) in; +#endif + +#include "cluster_data_inc.glsl" +#include "light_data_inc.glsl" + +#define M_PI 3.14159265359 + +#define DENSITY_SCALE 1024.0 + +layout(set = 0, binding = 1) uniform texture2D shadow_atlas; +layout(set = 0, binding = 2) uniform texture2D directional_shadow_atlas; + +layout(set = 0, binding = 3, std430) restrict readonly buffer OmniLights { + LightData data[]; +} +omni_lights; + +layout(set = 0, binding = 4, std430) restrict readonly buffer SpotLights { + LightData data[]; +} +spot_lights; + +layout(set = 0, binding = 5, std140) uniform DirectionalLights { + DirectionalLightData data[MAX_DIRECTIONAL_LIGHT_DATA_STRUCTS]; +} +directional_lights; + +layout(set = 0, binding = 6, std430) buffer restrict readonly ClusterBuffer { + uint data[]; +} +cluster_buffer; + +layout(set = 0, binding = 7) uniform sampler linear_sampler; + +#ifdef MODE_DENSITY +layout(rgba16f, set = 0, binding = 8) uniform restrict writeonly image3D density_map; +layout(rgba16f, set = 0, binding = 9) uniform restrict readonly image3D fog_map; //unused +#endif + +#ifdef MODE_FOG +layout(rgba16f, set = 0, binding = 8) uniform restrict readonly image3D density_map; +layout(rgba16f, set = 0, binding = 9) uniform restrict writeonly image3D fog_map; +#endif + +#ifdef MODE_COPY +layout(rgba16f, set = 0, binding = 8) uniform restrict readonly image3D source_map; +layout(rgba16f, set = 0, binding = 9) uniform restrict writeonly image3D dest_map; +#endif + +#ifdef MODE_FILTER +layout(rgba16f, set = 0, binding = 8) uniform restrict readonly image3D source_map; +layout(rgba16f, set = 0, binding = 9) uniform restrict writeonly image3D dest_map; +#endif + +layout(set = 0, binding = 10) uniform sampler shadow_sampler; + +#define MAX_VOXEL_GI_INSTANCES 8 + +struct VoxelGIData { + mat4 xform; // 64 - 64 + + vec3 bounds; // 12 - 76 + float dynamic_range; // 4 - 80 + + float bias; // 4 - 84 + float normal_bias; // 4 - 88 + bool blend_ambient; // 4 - 92 + uint mipmaps; // 4 - 96 +}; + +layout(set = 0, binding = 11, std140) uniform VoxelGIs { + VoxelGIData data[MAX_VOXEL_GI_INSTANCES]; +} +voxel_gi_instances; + +layout(set = 0, binding = 12) uniform texture3D voxel_gi_textures[MAX_VOXEL_GI_INSTANCES]; + +layout(set = 0, binding = 13) uniform sampler linear_sampler_with_mipmaps; + +#ifdef ENABLE_SDFGI + +// SDFGI Integration on set 1 +#define SDFGI_MAX_CASCADES 8 + +struct SDFVoxelGICascadeData { + vec3 position; + float to_probe; + ivec3 probe_world_offset; + float to_cell; // 1/bounds * grid_size +}; + +layout(set = 1, binding = 0, std140) uniform SDFGI { + vec3 grid_size; + uint max_cascades; + + bool use_occlusion; + int probe_axis_size; + float probe_to_uvw; + float normal_bias; + + vec3 lightprobe_tex_pixel_size; + float energy; + + vec3 lightprobe_uv_offset; + float y_mult; + + vec3 occlusion_clamp; + uint pad3; + + vec3 occlusion_renormalize; + uint pad4; + + vec3 cascade_probe_size; + uint pad5; + + SDFVoxelGICascadeData cascades[SDFGI_MAX_CASCADES]; +} +sdfgi; + +layout(set = 1, binding = 1) uniform texture2DArray sdfgi_ambient_texture; + +layout(set = 1, binding = 2) uniform texture3D sdfgi_occlusion_texture; + +#endif //SDFGI + +layout(set = 0, binding = 14, std140) uniform Params { + vec2 fog_frustum_size_begin; + vec2 fog_frustum_size_end; + + float fog_frustum_end; + float ambient_inject; + float z_far; + int filter_axis; + + vec3 ambient_color; + float sky_contribution; + + ivec3 fog_volume_size; + uint directional_light_count; + + vec3 base_emission; + float base_density; + + vec3 base_scattering; + float phase_g; + + float detail_spread; + float gi_inject; + uint max_voxel_gi_instances; + uint cluster_type_size; + + vec2 screen_size; + uint cluster_shift; + uint cluster_width; + + uint max_cluster_element_count_div_32; + bool use_temporal_reprojection; + uint temporal_frame; + float temporal_blend; + + mat3x4 cam_rotation; + mat4 to_prev_view; + + mat3 radiance_inverse_xform; +} +params; +#ifndef MODE_COPY +layout(set = 0, binding = 15) uniform texture3D prev_density_texture; + +#ifdef MOLTENVK_USED +layout(set = 0, binding = 16) buffer density_only_map_buffer { + uint density_only_map[]; +}; +layout(set = 0, binding = 17) buffer light_only_map_buffer { + uint light_only_map[]; +}; +layout(set = 0, binding = 18) buffer emissive_only_map_buffer { + uint emissive_only_map[]; +}; +#else +layout(r32ui, set = 0, binding = 16) uniform uimage3D density_only_map; +layout(r32ui, set = 0, binding = 17) uniform uimage3D light_only_map; +layout(r32ui, set = 0, binding = 18) uniform uimage3D emissive_only_map; +#endif + +#ifdef USE_RADIANCE_CUBEMAP_ARRAY +layout(set = 0, binding = 19) uniform textureCubeArray sky_texture; +#else +layout(set = 0, binding = 19) uniform textureCube sky_texture; +#endif +#endif // MODE_COPY + +float get_depth_at_pos(float cell_depth_size, int z) { + float d = float(z) * cell_depth_size + cell_depth_size * 0.5; //center of voxels + d = pow(d, params.detail_spread); + return params.fog_frustum_end * d; +} + +vec3 hash3f(uvec3 x) { + x = ((x >> 16) ^ x) * 0x45d9f3b; + x = ((x >> 16) ^ x) * 0x45d9f3b; + x = (x >> 16) ^ x; + return vec3(x & 0xFFFFF) / vec3(float(0xFFFFF)); +} + +float get_omni_attenuation(float dist, float inv_range, float decay) { + float nd = dist * inv_range; + nd *= nd; + nd *= nd; // nd^4 + nd = max(1.0 - nd, 0.0); + nd *= nd; // nd^2 + return nd * pow(max(dist, 0.0001), -decay); +} + +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_max = item_min_max >> 16; + + item_from = item_min >> 5; + item_to = (item_max == 0) ? 0 : ((item_max - 1) >> 5) + 1; //side effect of how it is stored, as item_max 0 means no elements +} + +uint cluster_get_range_clip_mask(uint i, uint z_min, uint z_max) { + int local_min = clamp(int(z_min) - int(i) * 32, 0, 31); + int mask_width = min(int(z_max) - int(z_min), 32 - local_min); + return bitfieldInsert(uint(0), uint(0xFFFFFFFF), local_min, mask_width); +} + +float henyey_greenstein(float cos_theta, float g) { + const float k = 0.0795774715459; // 1 / (4 * PI) + return k * (1.0 - g * g) / (pow(1.0 + g * g - 2.0 * g * cos_theta, 1.5)); +} + +#define TEMPORAL_FRAMES 16 + +const vec3 halton_map[TEMPORAL_FRAMES] = vec3[]( + vec3(0.5, 0.33333333, 0.2), + vec3(0.25, 0.66666667, 0.4), + vec3(0.75, 0.11111111, 0.6), + vec3(0.125, 0.44444444, 0.8), + vec3(0.625, 0.77777778, 0.04), + vec3(0.375, 0.22222222, 0.24), + vec3(0.875, 0.55555556, 0.44), + vec3(0.0625, 0.88888889, 0.64), + vec3(0.5625, 0.03703704, 0.84), + vec3(0.3125, 0.37037037, 0.08), + vec3(0.8125, 0.7037037, 0.28), + vec3(0.1875, 0.14814815, 0.48), + vec3(0.6875, 0.48148148, 0.68), + vec3(0.4375, 0.81481481, 0.88), + vec3(0.9375, 0.25925926, 0.12), + vec3(0.03125, 0.59259259, 0.32)); + +void main() { + vec3 fog_cell_size = 1.0 / vec3(params.fog_volume_size); + +#ifdef MODE_DENSITY + + ivec3 pos = ivec3(gl_GlobalInvocationID.xyz); + if (any(greaterThanEqual(pos, params.fog_volume_size))) { + return; //do not compute + } +#ifdef MOLTENVK_USED + uint lpos = pos.z * params.fog_volume_size.x * params.fog_volume_size.y + pos.y * params.fog_volume_size.x + pos.x; +#endif + + vec3 posf = vec3(pos); + + //posf += mix(vec3(0.0),vec3(1.0),0.3) * hash3f(uvec3(pos)) * 2.0 - 1.0; + + vec3 fog_unit_pos = posf * fog_cell_size + fog_cell_size * 0.5; //center of voxels + + uvec2 screen_pos = uvec2(fog_unit_pos.xy * params.screen_size); + uvec2 cluster_pos = screen_pos >> params.cluster_shift; + uint cluster_offset = (params.cluster_width * cluster_pos.y + cluster_pos.x) * (params.max_cluster_element_count_div_32 + 32); + //positions in screen are too spread apart, no hopes for optimizing with subgroups + + fog_unit_pos.z = pow(fog_unit_pos.z, params.detail_spread); + + vec3 view_pos; + view_pos.xy = (fog_unit_pos.xy * 2.0 - 1.0) * mix(params.fog_frustum_size_begin, params.fog_frustum_size_end, vec2(fog_unit_pos.z)); + view_pos.z = -params.fog_frustum_end * fog_unit_pos.z; + view_pos.y = -view_pos.y; + + vec4 reprojected_density = vec4(0.0); + float reproject_amount = 0.0; + + if (params.use_temporal_reprojection) { + vec3 prev_view = (params.to_prev_view * vec4(view_pos, 1.0)).xyz; + //undo transform into prev view + prev_view.y = -prev_view.y; + //z back to unit size + prev_view.z /= -params.fog_frustum_end; + //xy back to unit size + prev_view.xy /= mix(params.fog_frustum_size_begin, params.fog_frustum_size_end, vec2(prev_view.z)); + prev_view.xy = prev_view.xy * 0.5 + 0.5; + //z back to unspread value + prev_view.z = pow(prev_view.z, 1.0 / params.detail_spread); + + if (all(greaterThan(prev_view, vec3(0.0))) && all(lessThan(prev_view, vec3(1.0)))) { + //reprojectinon fits + + reprojected_density = textureLod(sampler3D(prev_density_texture, linear_sampler), prev_view, 0.0); + reproject_amount = params.temporal_blend; + + // Since we can reproject, now we must jitter the current view pos. + // This is done here because cells that can't reproject should not jitter. + + fog_unit_pos = posf * fog_cell_size + fog_cell_size * halton_map[params.temporal_frame]; //center of voxels, offset by halton table + + screen_pos = uvec2(fog_unit_pos.xy * params.screen_size); + cluster_pos = screen_pos >> params.cluster_shift; + cluster_offset = (params.cluster_width * cluster_pos.y + cluster_pos.x) * (params.max_cluster_element_count_div_32 + 32); + //positions in screen are too spread apart, no hopes for optimizing with subgroups + + fog_unit_pos.z = pow(fog_unit_pos.z, params.detail_spread); + + view_pos.xy = (fog_unit_pos.xy * 2.0 - 1.0) * mix(params.fog_frustum_size_begin, params.fog_frustum_size_end, vec2(fog_unit_pos.z)); + view_pos.z = -params.fog_frustum_end * fog_unit_pos.z; + view_pos.y = -view_pos.y; + } + } + + uint cluster_z = uint(clamp((abs(view_pos.z) / params.z_far) * 32.0, 0.0, 31.0)); + + vec3 total_light = vec3(0.0); + + float total_density = params.base_density; +#ifdef MOLTENVK_USED + uint local_density = density_only_map[lpos]; +#else + uint local_density = imageLoad(density_only_map, pos).x; +#endif + + total_density += float(int(local_density)) / DENSITY_SCALE; + total_density = max(0.0, total_density); + +#ifdef MOLTENVK_USED + uint scattering_u = light_only_map[lpos]; +#else + uint scattering_u = imageLoad(light_only_map, pos).x; +#endif + vec3 scattering = vec3(scattering_u >> 21, (scattering_u << 11) >> 21, scattering_u % 1024) / vec3(2047.0, 2047.0, 1023.0); + scattering += params.base_scattering * params.base_density; + +#ifdef MOLTENVK_USED + uint emission_u = emissive_only_map[lpos]; +#else + uint emission_u = imageLoad(emissive_only_map, pos).x; +#endif + vec3 emission = vec3(emission_u >> 21, (emission_u << 11) >> 21, emission_u % 1024) / vec3(511.0, 511.0, 255.0); + emission += params.base_emission * params.base_density; + + 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) { + 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; + } + + 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); + + 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); + } + + 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); + } + + // Compute light from sky + if (params.ambient_inject > 0.0) { + vec3 isotropic = vec3(0.0); + vec3 anisotropic = vec3(0.0); + if (params.sky_contribution > 0.0) { + float mip_bias = 2.0 + total_density * (MAX_SKY_LOD - 2.0); // Not physically based, but looks nice + vec3 scatter_direction = (params.radiance_inverse_xform * normalize(view_pos)) * sign(params.phase_g); +#ifdef USE_RADIANCE_CUBEMAP_ARRAY + isotropic = texture(samplerCubeArray(sky_texture, linear_sampler_with_mipmaps), vec4(0.0, 1.0, 0.0, mip_bias)).rgb; + anisotropic = texture(samplerCubeArray(sky_texture, linear_sampler_with_mipmaps), vec4(scatter_direction, mip_bias)).rgb; +#else + isotropic = textureLod(samplerCube(sky_texture, linear_sampler_with_mipmaps), vec3(0.0, 1.0, 0.0), mip_bias).rgb; + anisotropic = textureLod(samplerCube(sky_texture, linear_sampler_with_mipmaps), vec3(scatter_direction), mip_bias).rgb; +#endif //USE_RADIANCE_CUBEMAP_ARRAY + } + + total_light += mix(params.ambient_color, mix(isotropic, anisotropic, abs(params.phase_g)), params.sky_contribution) * params.ambient_inject; + } + + //compute lights from cluster + + { //omni lights + + uint cluster_omni_offset = cluster_offset; + + uint item_min; + uint item_max; + uint item_from; + uint item_to; + + cluster_get_item_range(cluster_omni_offset + params.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 + + for (uint i = item_from; i < item_to; i++) { + uint mask = cluster_buffer.data[cluster_omni_offset + i]; + mask &= cluster_get_range_clip_mask(i, item_min, item_max); +#ifdef USE_SUBGROUPS + uint merged_mask = subgroupBroadcastFirst(subgroupOr(mask)); +#else + uint merged_mask = mask; +#endif + + while (merged_mask != 0) { + uint bit = findMSB(merged_mask); + merged_mask &= ~(1 << bit); +#ifdef USE_SUBGROUPS + if (((1 << bit) & mask) == 0) { //do not process if not originally here + continue; + } +#endif + uint light_index = 32 * i + bit; + + //if (!bool(omni_omni_lights.data[light_index].mask & draw_call.layer_mask)) { + // continue; //not masked + //} + + vec3 light_pos = omni_lights.data[light_index].position; + 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) { + 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) { + //has shadow + vec4 uv_rect = omni_lights.data[light_index].atlas_rect; + vec2 flip_offset = omni_lights.data[light_index].direction.xy; + + vec3 local_vert = (omni_lights.data[light_index].shadow_matrix * vec4(view_pos, 1.0)).xyz; + + float shadow_len = length(local_vert); //need to remember shadow len from here + vec3 shadow_sample = normalize(local_vert); + + if (shadow_sample.z >= 0.0) { + uv_rect.xy += flip_offset; + } + + shadow_sample.z = 1.0 + abs(shadow_sample.z); + vec3 pos = vec3(shadow_sample.xy / shadow_sample.z, shadow_len - omni_lights.data[light_index].shadow_bias); + pos.z *= omni_lights.data[light_index].inv_radius; + + pos.xy = pos.xy * 0.5 + 0.5; + pos.xy = uv_rect.xy + pos.xy * uv_rect.zw; + + 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); + } + total_light += light * attenuation * shadow_attenuation * henyey_greenstein(dot(normalize(light_pos - view_pos), normalize(view_pos)), params.phase_g); + } + } + } + } + + { //spot lights + + uint cluster_spot_offset = cluster_offset + params.cluster_type_size; + + uint item_min; + uint item_max; + uint item_from; + uint item_to; + + cluster_get_item_range(cluster_spot_offset + params.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 + + for (uint i = item_from; i < item_to; i++) { + uint mask = cluster_buffer.data[cluster_spot_offset + i]; + mask &= cluster_get_range_clip_mask(i, item_min, item_max); +#ifdef USE_SUBGROUPS + uint merged_mask = subgroupBroadcastFirst(subgroupOr(mask)); +#else + uint merged_mask = mask; +#endif + + while (merged_mask != 0) { + uint bit = findMSB(merged_mask); + merged_mask &= ~(1 << bit); +#ifdef USE_SUBGROUPS + if (((1 << bit) & mask) == 0) { //do not process if not originally here + continue; + } +#endif + + //if (!bool(omni_lights.data[light_index].mask & draw_call.layer_mask)) { + // continue; //not masked + //} + + uint light_index = 32 * i + bit; + + vec3 light_pos = spot_lights.data[light_index].position; + vec3 light_rel_vec = spot_lights.data[light_index].position - view_pos; + float d = length(light_rel_vec); + float shadow_attenuation = 1.0; + + if (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; + float scos = max(dot(-normalize(light_rel_vec), spot_dir), spot_lights.data[light_index].cone_angle); + float spot_rim = max(0.0001, (1.0 - scos) / (1.0 - spot_lights.data[light_index].cone_angle)); + attenuation *= 1.0 - pow(spot_rim, spot_lights.data[light_index].cone_attenuation); + + vec3 light = spot_lights.data[light_index].color; + + if (spot_lights.data[light_index].shadow_enabled) { + //has shadow + vec4 uv_rect = spot_lights.data[light_index].atlas_rect; + vec2 flip_offset = spot_lights.data[light_index].direction.xy; + + vec3 local_vert = (spot_lights.data[light_index].shadow_matrix * vec4(view_pos, 1.0)).xyz; + + float shadow_len = length(local_vert); //need to remember shadow len from here + vec3 shadow_sample = normalize(local_vert); + + if (shadow_sample.z >= 0.0) { + uv_rect.xy += flip_offset; + } + + shadow_sample.z = 1.0 + abs(shadow_sample.z); + vec3 pos = vec3(shadow_sample.xy / shadow_sample.z, shadow_len - spot_lights.data[light_index].shadow_bias); + pos.z *= spot_lights.data[light_index].inv_radius; + + pos.xy = pos.xy * 0.5 + 0.5; + pos.xy = uv_rect.xy + pos.xy * uv_rect.zw; + + 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); + } + total_light += light * attenuation * shadow_attenuation * henyey_greenstein(dot(normalize(light_rel_vec), normalize(view_pos)), params.phase_g); + } + } + } + } + + vec3 world_pos = mat3(params.cam_rotation) * view_pos; + + for (uint i = 0; i < params.max_voxel_gi_instances; i++) { + vec3 position = (voxel_gi_instances.data[i].xform * vec4(world_pos, 1.0)).xyz; + + //this causes corrupted pixels, i have no idea why.. + if (all(bvec2(all(greaterThanEqual(position, vec3(0.0))), all(lessThan(position, voxel_gi_instances.data[i].bounds))))) { + position /= voxel_gi_instances.data[i].bounds; + + vec4 light = vec4(0.0); + for (uint j = 0; j < voxel_gi_instances.data[i].mipmaps; j++) { + vec4 slight = textureLod(sampler3D(voxel_gi_textures[i], linear_sampler_with_mipmaps), position, float(j)); + float a = (1.0 - light.a); + light += a * slight; + } + + light.rgb *= voxel_gi_instances.data[i].dynamic_range * params.gi_inject; + + total_light += light.rgb; + } + } + + //sdfgi +#ifdef ENABLE_SDFGI + + { + float blend = -1.0; + vec3 ambient_total = vec3(0.0); + + for (uint i = 0; i < sdfgi.max_cascades; i++) { + vec3 cascade_pos = (world_pos - sdfgi.cascades[i].position) * sdfgi.cascades[i].to_probe; + + if (any(lessThan(cascade_pos, vec3(0.0))) || any(greaterThanEqual(cascade_pos, sdfgi.cascade_probe_size))) { + continue; //skip cascade + } + + vec3 base_pos = floor(cascade_pos); + ivec3 probe_base_pos = ivec3(base_pos); + + vec4 ambient_accum = vec4(0.0); + + ivec3 tex_pos = ivec3(probe_base_pos.xy, int(i)); + tex_pos.x += probe_base_pos.z * sdfgi.probe_axis_size; + + for (uint j = 0; j < 8; j++) { + ivec3 offset = (ivec3(j) >> ivec3(0, 1, 2)) & ivec3(1, 1, 1); + ivec3 probe_posi = probe_base_pos; + probe_posi += offset; + + // Compute weight + + vec3 probe_pos = vec3(probe_posi); + vec3 probe_to_pos = cascade_pos - probe_pos; + + vec3 trilinear = vec3(1.0) - abs(probe_to_pos); + float weight = trilinear.x * trilinear.y * trilinear.z; + + // Compute lightprobe occlusion + + if (sdfgi.use_occlusion) { + ivec3 occ_indexv = abs((sdfgi.cascades[i].probe_world_offset + probe_posi) & ivec3(1, 1, 1)) * ivec3(1, 2, 4); + vec4 occ_mask = mix(vec4(0.0), vec4(1.0), equal(ivec4(occ_indexv.x | occ_indexv.y), ivec4(0, 1, 2, 3))); + + vec3 occ_pos = clamp(cascade_pos, probe_pos - sdfgi.occlusion_clamp, probe_pos + sdfgi.occlusion_clamp) * sdfgi.probe_to_uvw; + occ_pos.z += float(i); + if (occ_indexv.z != 0) { //z bit is on, means index is >=4, so make it switch to the other half of textures + occ_pos.x += 1.0; + } + + occ_pos *= sdfgi.occlusion_renormalize; + float occlusion = dot(textureLod(sampler3D(sdfgi_occlusion_texture, linear_sampler), occ_pos, 0.0), occ_mask); + + weight *= max(occlusion, 0.01); + } + + // Compute ambient texture position + + ivec3 uvw = tex_pos; + uvw.xy += offset.xy; + uvw.x += offset.z * sdfgi.probe_axis_size; + + vec3 ambient = texelFetch(sampler2DArray(sdfgi_ambient_texture, linear_sampler), uvw, 0).rgb; + + ambient_accum.rgb += ambient * weight; + ambient_accum.a += weight; + } + + if (ambient_accum.a > 0) { + ambient_accum.rgb /= ambient_accum.a; + } + ambient_total = ambient_accum.rgb; + break; + } + + total_light += ambient_total * params.gi_inject; + } + +#endif + } + + vec4 final_density = vec4(total_light * scattering + emission, total_density); + + final_density = mix(final_density, reprojected_density, reproject_amount); + + imageStore(density_map, pos, final_density); +#ifdef MOLTENVK_USED + density_only_map[lpos] = 0; + light_only_map[lpos] = 0; + emissive_only_map[lpos] = 0; +#else + imageStore(density_only_map, pos, uvec4(0)); + imageStore(light_only_map, pos, uvec4(0)); + imageStore(emissive_only_map, pos, uvec4(0)); +#endif +#endif + +#ifdef MODE_FOG + + ivec3 pos = ivec3(gl_GlobalInvocationID.xy, 0); + + if (any(greaterThanEqual(pos, params.fog_volume_size))) { + return; //do not compute + } + + vec4 fog_accum = vec4(0.0, 0.0, 0.0, 1.0); + float prev_z = 0.0; + + for (int i = 0; i < params.fog_volume_size.z; i++) { + //compute fog position + ivec3 fog_pos = pos + ivec3(0, 0, i); + //get fog value + vec4 fog = imageLoad(density_map, fog_pos); + + //get depth at cell pos + float z = get_depth_at_pos(fog_cell_size.z, i); + //get distance from previous pos + float d = abs(prev_z - z); + //compute transmittance using beer's law + float transmittance = exp(-d * fog.a); + + fog_accum.rgb += ((fog.rgb - fog.rgb * transmittance) / max(fog.a, 0.00001)) * fog_accum.a; + fog_accum.a *= transmittance; + + prev_z = z; + + imageStore(fog_map, fog_pos, vec4(fog_accum.rgb, 1.0 - fog_accum.a)); + } + +#endif + +#ifdef MODE_FILTER + + ivec3 pos = ivec3(gl_GlobalInvocationID.xyz); + + const float gauss[7] = float[](0.071303, 0.131514, 0.189879, 0.214607, 0.189879, 0.131514, 0.071303); + + const ivec3 filter_dir[3] = ivec3[](ivec3(1, 0, 0), ivec3(0, 1, 0), ivec3(0, 0, 1)); + ivec3 offset = filter_dir[params.filter_axis]; + + vec4 accum = vec4(0.0); + for (int i = -3; i <= 3; i++) { + accum += imageLoad(source_map, clamp(pos + offset * i, ivec3(0), params.fog_volume_size - ivec3(1))) * gauss[i + 3]; + } + + imageStore(dest_map, pos, accum); + +#endif +#ifdef MODE_COPY + ivec3 pos = ivec3(gl_GlobalInvocationID.xyz); + if (any(greaterThanEqual(pos, params.fog_volume_size))) { + return; //do not compute + } + + imageStore(dest_map, pos, imageLoad(source_map, pos)); + +#endif +} diff --git a/servers/rendering/renderer_rd/shaders/giprobe.glsl b/servers/rendering/renderer_rd/shaders/voxel_gi.glsl index b931461b31..577c6d0cd0 100644 --- a/servers/rendering/renderer_rd/shaders/giprobe.glsl +++ b/servers/rendering/renderer_rd/shaders/voxel_gi.glsl @@ -2,7 +2,7 @@ #version 450 -VERSION_DEFINES +#VERSION_DEFINES #ifdef MODE_DYNAMIC layout(local_size_x = 8, local_size_y = 8, local_size_z = 1) in; @@ -13,7 +13,6 @@ layout(local_size_x = 64, local_size_y = 1, local_size_z = 1) in; #ifndef MODE_DYNAMIC #define NO_CHILDREN 0xFFFFFFFF -#define GREY_VEC vec3(0.33333, 0.33333, 0.33333) struct CellChildren { uint children[8]; @@ -71,16 +70,11 @@ lights; layout(set = 0, binding = 5) uniform texture3D color_texture; -#ifdef MODE_ANISOTROPIC -layout(set = 0, binding = 7) uniform texture3D aniso_pos_texture; -layout(set = 0, binding = 8) uniform texture3D aniso_neg_texture; -#endif // MODE ANISOTROPIC - #endif // MODE_SECOND_BOUNCE #ifndef MODE_DYNAMIC -layout(push_constant, binding = 0, std430) uniform Params { +layout(push_constant, std430) uniform Params { ivec3 limits; uint stack_size; @@ -110,18 +104,11 @@ layout(set = 0, binding = 10) uniform sampler texture_sampler; layout(rgba8, set = 0, binding = 5) uniform restrict writeonly image3D color_tex; -#ifdef MODE_ANISOTROPIC - -layout(r16ui, set = 0, binding = 6) uniform restrict writeonly uimage3D aniso_pos_tex; -layout(r16ui, set = 0, binding = 7) uniform restrict writeonly uimage3D aniso_neg_tex; - -#endif - #endif #ifdef MODE_DYNAMIC -layout(push_constant, binding = 0, std430) uniform Params { +layout(push_constant, std430) uniform Params { ivec3 limits; uint light_count; //when not lighting ivec3 x_dir; @@ -170,13 +157,6 @@ layout(r32f, set = 0, binding = 8) uniform restrict writeonly image2D depth; layout(rgba8, set = 0, binding = 11) uniform restrict image3D color_texture; -#ifdef MODE_ANISOTROPIC - -layout(r16ui, set = 0, binding = 12) uniform restrict writeonly uimage3D aniso_pos_texture; -layout(r16ui, set = 0, binding = 13) uniform restrict writeonly uimage3D aniso_neg_texture; - -#endif // MODE ANISOTROPIC - #endif //MODE_DYNAMIC_SHRINK_PLOT #endif // MODE_DYNAMIC_SHRINK @@ -374,12 +354,7 @@ void main() { vec3 emission = vec3(uvec3(cell_data.data[cell_index].emission & 0x1ff, (cell_data.data[cell_index].emission >> 9) & 0x1ff, (cell_data.data[cell_index].emission >> 18) & 0x1ff)) * pow(2.0, float(cell_data.data[cell_index].emission >> 27) - 15.0 - 9.0); vec3 normal = unpackSnorm4x8(cell_data.data[cell_index].normal).xyz; -#ifdef MODE_ANISOTROPIC - vec3 accum[6] = vec3[](vec3(0.0), vec3(0.0), vec3(0.0), vec3(0.0), vec3(0.0), vec3(0.0)); - const vec3 accum_dirs[6] = vec3[](vec3(1.0, 0.0, 0.0), vec3(-1.0, 0.0, 0.0), vec3(0.0, 1.0, 0.0), vec3(0.0, -1.0, 0.0), vec3(0.0, 0.0, 1.0), vec3(0.0, 0.0, -1.0)); -#else vec3 accum = vec3(0.0); -#endif for (uint i = 0; i < params.light_count; i++) { vec3 light; @@ -390,38 +365,16 @@ void main() { light *= albedo.rgb; -#ifdef MODE_ANISOTROPIC - for (uint j = 0; j < 6; j++) { - accum[j] += max(0.0, dot(accum_dirs[j], -light_dir)) * light; - } -#else if (length(normal) > 0.2) { accum += max(0.0, dot(normal, -light_dir)) * light; } else { //all directions accum += light; } -#endif } -#ifdef MODE_ANISOTROPIC - - for (uint i = 0; i < 6; i++) { - vec3 light = accum[i]; - if (length(normal) > 0.2) { - light += max(0.0, dot(accum_dirs[i], -normal)) * emission; - } else { - light += emission; - } - - outputs.data[cell_index * 6 + i] = vec4(light, 0.0); - } - -#else outputs.data[cell_index] = vec4(accum + emission, 0.0); -#endif - #endif //MODE_COMPUTE_LIGHT /////////////////SECOND BOUNCE/////////////////////////////// @@ -431,32 +384,8 @@ void main() { ivec3 ipos = ivec3(posu); vec4 normal = unpackSnorm4x8(cell_data.data[cell_index].normal); -#ifdef MODE_ANISOTROPIC - vec3 accum[6]; - const vec3 accum_dirs[6] = vec3[](vec3(1.0, 0.0, 0.0), vec3(-1.0, 0.0, 0.0), vec3(0.0, 1.0, 0.0), vec3(0.0, -1.0, 0.0), vec3(0.0, 0.0, 1.0), vec3(0.0, 0.0, -1.0)); - - /*vec3 src_color = texelFetch(sampler3D(color_texture,texture_sampler),ipos,0).rgb * params.dynamic_range; - vec3 src_aniso_pos = texelFetch(sampler3D(aniso_pos_texture,texture_sampler),ipos,0).rgb; - vec3 src_anisp_neg = texelFetch(sampler3D(anisp_neg_texture,texture_sampler),ipos,0).rgb; - accum[0]=src_col * src_aniso_pos.x; - accum[1]=src_col * src_aniso_neg.x; - accum[2]=src_col * src_aniso_pos.y; - accum[3]=src_col * src_aniso_neg.y; - accum[4]=src_col * src_aniso_pos.z; - accum[5]=src_col * src_aniso_neg.z;*/ - - accum[0] = outputs.data[cell_index * 6 + 0].rgb; - accum[1] = outputs.data[cell_index * 6 + 1].rgb; - accum[2] = outputs.data[cell_index * 6 + 2].rgb; - accum[3] = outputs.data[cell_index * 6 + 3].rgb; - accum[4] = outputs.data[cell_index * 6 + 4].rgb; - accum[5] = outputs.data[cell_index * 6 + 5].rgb; - -#else vec3 accum = outputs.data[cell_index].rgb; -#endif - if (length(normal.xyz) > 0.2) { 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.xyz)); @@ -484,9 +413,6 @@ void main() { float max_distance = length(vec3(params.limits)); vec3 cell_size = 1.0 / vec3(params.limits); -#ifdef MODE_ANISOTROPIC - vec3 aniso_normal = mix(direction, normal.xyz, params.aniso_strength); -#endif while (dist < max_distance && color.a < 0.95) { float diameter = max(1.0, 2.0 * tan_half_angle * dist); vec3 uvw_pos = (pos + dist * direction) * cell_size; @@ -498,42 +424,18 @@ void main() { float log2_diameter = log2(diameter); vec4 scolor = textureLod(sampler3D(color_texture, texture_sampler), uvw_pos, log2_diameter); -#ifdef MODE_ANISOTROPIC - - vec3 aniso_neg = textureLod(sampler3D(aniso_neg_texture, texture_sampler), uvw_pos, log2_diameter).rgb; - vec3 aniso_pos = textureLod(sampler3D(aniso_pos_texture, texture_sampler), uvw_pos, log2_diameter).rgb; - - scolor.rgb *= dot(max(vec3(0.0), (aniso_normal * aniso_pos)), vec3(1.0)) + dot(max(vec3(0.0), (-aniso_normal * aniso_neg)), vec3(1.0)); -#endif float a = (1.0 - color.a); color += a * scolor; dist += half_diameter; } } color *= cone_weights[i] * vec4(albedo.rgb, 1.0) * params.dynamic_range; //restore range -#ifdef MODE_ANISOTROPIC - for (uint j = 0; j < 6; j++) { - accum[j] += max(0.0, dot(accum_dirs[j], direction)) * color.rgb; - } -#else accum += color.rgb; -#endif } } -#ifdef MODE_ANISOTROPIC - - outputs.data[cell_index * 6 + 0] = vec4(accum[0], 0.0); - outputs.data[cell_index * 6 + 1] = vec4(accum[1], 0.0); - outputs.data[cell_index * 6 + 2] = vec4(accum[2], 0.0); - outputs.data[cell_index * 6 + 3] = vec4(accum[3], 0.0); - outputs.data[cell_index * 6 + 4] = vec4(accum[4], 0.0); - outputs.data[cell_index * 6 + 5] = vec4(accum[5], 0.0); -#else outputs.data[cell_index] = vec4(accum, 0.0); -#endif - #endif // MODE_SECOND_BOUNCE /////////////////UPDATE MIPMAPS/////////////////////////////// @@ -541,45 +443,20 @@ void main() { #ifdef MODE_UPDATE_MIPMAPS { -#ifdef MODE_ANISOTROPIC - vec3 light_accum[6] = vec3[](vec3(0.0), vec3(0.0), vec3(0.0), vec3(0.0), vec3(0.0), vec3(0.0)); -#else vec3 light_accum = vec3(0.0); -#endif float count = 0.0; for (uint i = 0; i < 8; i++) { uint child_index = cell_children.data[cell_index].children[i]; if (child_index == NO_CHILDREN) { continue; } -#ifdef MODE_ANISOTROPIC - light_accum[0] += outputs.data[child_index * 6 + 0].rgb; - light_accum[1] += outputs.data[child_index * 6 + 1].rgb; - light_accum[2] += outputs.data[child_index * 6 + 2].rgb; - light_accum[3] += outputs.data[child_index * 6 + 3].rgb; - light_accum[4] += outputs.data[child_index * 6 + 4].rgb; - light_accum[5] += outputs.data[child_index * 6 + 5].rgb; - -#else light_accum += outputs.data[child_index].rgb; -#endif - count += 1.0; } float divisor = mix(8.0, count, params.propagation); -#ifdef MODE_ANISOTROPIC - outputs.data[cell_index * 6 + 0] = vec4(light_accum[0] / divisor, 0.0); - outputs.data[cell_index * 6 + 1] = vec4(light_accum[1] / divisor, 0.0); - outputs.data[cell_index * 6 + 2] = vec4(light_accum[2] / divisor, 0.0); - outputs.data[cell_index * 6 + 3] = vec4(light_accum[3] / divisor, 0.0); - outputs.data[cell_index * 6 + 4] = vec4(light_accum[4] / divisor, 0.0); - outputs.data[cell_index * 6 + 5] = vec4(light_accum[5] / divisor, 0.0); - -#else outputs.data[cell_index] = vec4(light_accum / divisor, 0.0); -#endif } #endif @@ -587,40 +464,7 @@ void main() { #ifdef MODE_WRITE_TEXTURE { -#ifdef MODE_ANISOTROPIC - vec3 accum_total = vec3(0.0); - accum_total += outputs.data[cell_index * 6 + 0].rgb; - accum_total += outputs.data[cell_index * 6 + 1].rgb; - accum_total += outputs.data[cell_index * 6 + 2].rgb; - accum_total += outputs.data[cell_index * 6 + 3].rgb; - accum_total += outputs.data[cell_index * 6 + 4].rgb; - accum_total += outputs.data[cell_index * 6 + 5].rgb; - - float accum_total_energy = max(dot(accum_total, GREY_VEC), 0.00001); - vec3 iso_positive = vec3(dot(outputs.data[cell_index * 6 + 0].rgb, GREY_VEC), dot(outputs.data[cell_index * 6 + 2].rgb, GREY_VEC), dot(outputs.data[cell_index * 6 + 4].rgb, GREY_VEC)) / vec3(accum_total_energy); - vec3 iso_negative = vec3(dot(outputs.data[cell_index * 6 + 1].rgb, GREY_VEC), dot(outputs.data[cell_index * 6 + 3].rgb, GREY_VEC), dot(outputs.data[cell_index * 6 + 5].rgb, GREY_VEC)) / vec3(accum_total_energy); - - { - uint aniso_pos = uint(clamp(iso_positive.b * 31.0, 0.0, 31.0)); - aniso_pos |= uint(clamp(iso_positive.g * 63.0, 0.0, 63.0)) << 5; - aniso_pos |= uint(clamp(iso_positive.r * 31.0, 0.0, 31.0)) << 11; - imageStore(aniso_pos_tex, ivec3(posu), uvec4(aniso_pos)); - } - - { - uint aniso_neg = uint(clamp(iso_negative.b * 31.0, 0.0, 31.0)); - aniso_neg |= uint(clamp(iso_negative.g * 63.0, 0.0, 63.0)) << 5; - aniso_neg |= uint(clamp(iso_negative.r * 31.0, 0.0, 31.0)) << 11; - imageStore(aniso_neg_tex, ivec3(posu), uvec4(aniso_neg)); - } - - imageStore(color_tex, ivec3(posu), vec4(accum_total / params.dynamic_range, albedo.a)); - -#else - imageStore(color_tex, ivec3(posu), vec4(outputs.data[cell_index].rgb / params.dynamic_range, albedo.a)); - -#endif } #endif @@ -763,13 +607,6 @@ void main() { color.rgb /= params.dynamic_range; imageStore(color_texture, pos3d, color); //imageStore(color_texture,pos3d,vec4(1,1,1,1)); - -#ifdef MODE_ANISOTROPIC - //do not care about anisotropy for dynamic objects, just store full lit in all directions - imageStore(aniso_pos_texture, pos3d, uvec4(0xFFFF)); - imageStore(aniso_neg_texture, pos3d, uvec4(0xFFFF)); - -#endif // ANISOTROPIC } #endif // MODE_DYNAMIC_SHRINK_PLOT } diff --git a/servers/rendering/renderer_rd/shaders/giprobe_debug.glsl b/servers/rendering/renderer_rd/shaders/voxel_gi_debug.glsl index 515cc35507..fd7a2bf8ad 100644 --- a/servers/rendering/renderer_rd/shaders/giprobe_debug.glsl +++ b/servers/rendering/renderer_rd/shaders/voxel_gi_debug.glsl @@ -2,7 +2,7 @@ #version 450 -VERSION_DEFINES +#VERSION_DEFINES struct CellData { uint position; // xyz 10 bits @@ -20,12 +20,7 @@ layout(set = 0, binding = 2) uniform texture3D color_tex; layout(set = 0, binding = 3) uniform sampler tex_sampler; -#ifdef USE_ANISOTROPY -layout(set = 0, binding = 4) uniform texture3D aniso_pos_tex; -layout(set = 0, binding = 5) uniform texture3D aniso_neg_tex; -#endif - -layout(push_constant, binding = 0, std430) uniform Params { +layout(push_constant, std430) uniform Params { mat4 projection; uint cell_offset; float dynamic_range; @@ -95,66 +90,10 @@ void main() { #endif #ifdef MODE_DEBUG_LIGHT - -#ifdef USE_ANISOTROPY - -#define POS_X 0 -#define POS_Y 1 -#define POS_Z 2 -#define NEG_X 3 -#define NEG_Y 4 -#define NEG_Z 5 - - const uint triangle_aniso[12] = uint[]( - NEG_X, - NEG_Z, - NEG_Y, - NEG_Z, - NEG_X, - NEG_Y, - POS_Z, - POS_X, - POS_X, - POS_Y, - POS_Y, - POS_Z); - - color_interp.xyz = texelFetch(sampler3D(color_tex, tex_sampler), ivec3(posu), int(params.level)).xyz * params.dynamic_range; - vec3 aniso_pos = texelFetch(sampler3D(aniso_pos_tex, tex_sampler), ivec3(posu), int(params.level)).xyz; - vec3 aniso_neg = texelFetch(sampler3D(aniso_neg_tex, tex_sampler), ivec3(posu), int(params.level)).xyz; - uint side = triangle_aniso[gl_VertexIndex / 3]; - - float strength = 0.0; - switch (side) { - case POS_X: - strength = aniso_pos.x; - break; - case POS_Y: - strength = aniso_pos.y; - break; - case POS_Z: - strength = aniso_pos.z; - break; - case NEG_X: - strength = aniso_neg.x; - break; - case NEG_Y: - strength = aniso_neg.y; - break; - case NEG_Z: - strength = aniso_neg.z; - break; - } - - color_interp.xyz *= strength; - -#else color_interp = texelFetch(sampler3D(color_tex, tex_sampler), ivec3(posu), int(params.level)); color_interp.xyz *params.dynamic_range; - #endif -#endif float scale = (1 << params.level); gl_Position = params.projection * vec4((vec3(posu) + vertex) * scale, 1.0); @@ -172,7 +111,7 @@ void main() { #version 450 -VERSION_DEFINES +#VERSION_DEFINES layout(location = 0) in vec4 color_interp; layout(location = 0) out vec4 frag_color; diff --git a/servers/rendering/renderer_rd/shaders/giprobe_sdf.glsl b/servers/rendering/renderer_rd/shaders/voxel_gi_sdf.glsl index 5b3dec0ee7..47a611a543 100644 --- a/servers/rendering/renderer_rd/shaders/giprobe_sdf.glsl +++ b/servers/rendering/renderer_rd/shaders/voxel_gi_sdf.glsl @@ -2,14 +2,13 @@ #version 450 -VERSION_DEFINES +#VERSION_DEFINES layout(local_size_x = 4, local_size_y = 4, local_size_z = 4) in; -#define MAX_DISTANCE 100000 +#define MAX_DISTANCE 100000.0 #define NO_CHILDREN 0xFFFFFFFF -#define GREY_VEC vec3(0.33333, 0.33333, 0.33333) struct CellChildren { uint children[8]; @@ -34,7 +33,7 @@ cell_data; layout(r8ui, set = 0, binding = 3) uniform restrict writeonly uimage3D sdf_tex; -layout(push_constant, binding = 0, std430) uniform Params { +layout(push_constant, std430) uniform Params { uint offset; uint end; uint pad0; @@ -44,7 +43,7 @@ params; void main() { vec3 pos = vec3(gl_GlobalInvocationID); - float closest_dist = 100000.0; + float closest_dist = MAX_DISTANCE; for (uint i = params.offset; i < params.end; i++) { vec3 posu = vec3(uvec3(cell_data.data[i].position & 0x7FF, (cell_data.data[i].position >> 11) & 0x3FF, cell_data.data[i].position >> 21)); @@ -67,7 +66,7 @@ void main() { } #if 0 -layout(push_constant, binding = 0, std430) uniform Params { +layout(push_constant, std430) uniform Params { ivec3 limits; uint stack_size; } |