diff options
Diffstat (limited to 'servers/rendering/renderer_rd/shaders')
8 files changed, 257 insertions, 213 deletions
diff --git a/servers/rendering/renderer_rd/shaders/cube_to_dp.glsl b/servers/rendering/renderer_rd/shaders/cube_to_dp.glsl index 54d67db6c6..c3ac0bee57 100644 --- a/servers/rendering/renderer_rd/shaders/cube_to_dp.glsl +++ b/servers/rendering/renderer_rd/shaders/cube_to_dp.glsl @@ -1,33 +1,48 @@ -#[compute] +#[vertex] #version 450 VERSION_DEFINES -layout(local_size_x = 8, local_size_y = 8, local_size_z = 1) in; +layout(push_constant, binding = 1, std430) uniform Params { + float z_far; + float z_near; + bool z_flip; + uint pad; + vec4 screen_rect; +} +params; + +layout(location = 0) out vec2 uv_interp; + +void main() { + vec2 base_arr[4] = vec2[](vec2(0.0, 0.0), vec2(0.0, 1.0), vec2(1.0, 1.0), vec2(1.0, 0.0)); + uv_interp = base_arr[gl_VertexIndex]; + vec2 screen_pos = uv_interp * params.screen_rect.zw + params.screen_rect.xy; + gl_Position = vec4(screen_pos * 2.0 - 1.0, 0.0, 1.0); +} + +#[fragment] + +#version 450 + +VERSION_DEFINES + +layout(location = 0) in vec2 uv_interp; layout(set = 0, binding = 0) uniform samplerCube source_cube; layout(push_constant, binding = 1, std430) uniform Params { - ivec2 screen_size; - ivec2 offset; - float bias; float z_far; float z_near; bool z_flip; + uint pad; + vec4 screen_rect; } params; -layout(r32f, set = 1, binding = 0) uniform restrict writeonly image2D depth_buffer; - void main() { - ivec2 pos = ivec2(gl_GlobalInvocationID.xy); - if (any(greaterThan(pos, params.screen_size))) { //too large, do nothing - return; - } - - vec2 pixel_size = 1.0 / vec2(params.screen_size); - vec2 uv = (vec2(pos) + 0.5) * pixel_size; + vec2 uv = uv_interp; vec3 normal = vec3(uv * 2.0 - 1.0, 0.0); @@ -65,5 +80,5 @@ void main() { float linear_depth = 2.0 * params.z_near * params.z_far / (params.z_far + params.z_near - depth * (params.z_far - params.z_near)); depth = (linear_depth * depth_fix) / params.z_far; - imageStore(depth_buffer, pos + params.offset, vec4(depth)); + gl_FragDepth = depth; } diff --git a/servers/rendering/renderer_rd/shaders/gi.glsl b/servers/rendering/renderer_rd/shaders/gi.glsl index c2965f9874..35522103df 100644 --- a/servers/rendering/renderer_rd/shaders/gi.glsl +++ b/servers/rendering/renderer_rd/shaders/gi.glsl @@ -97,13 +97,12 @@ layout(push_constant, binding = 0, std430) uniform Params { vec4 proj_info; + vec3 ao_color; uint max_giprobes; + bool high_quality_vct; - uint pad2; bool orthogonal; - - vec3 ao_color; - uint pad; + uint pad[2]; mat3x4 cam_rotation; } @@ -594,28 +593,16 @@ vec4 fetch_normal_and_roughness(ivec2 pos) { return normal_roughness; } -void main() { - // Pixel being shaded - ivec2 pos = ivec2(gl_GlobalInvocationID.xy); - if (any(greaterThanEqual(pos, params.screen_size))) { //too large, do nothing - return; - } - - vec3 vertex = reconstruct_position(pos); - vertex.y = -vertex.y; - +void process_gi(ivec2 pos, vec3 vertex, inout vec4 ambient_light, inout vec4 reflection_light) { vec4 normal_roughness = fetch_normal_and_roughness(pos); - vec3 normal = normal_roughness.xyz; - vec4 ambient_light = vec4(0.0), reflection_light = vec4(0.0); + vec3 normal = normal_roughness.xyz; if (normal.length() > 0.5) { //valid normal, can do GI float roughness = normal_roughness.w; - vertex = mat3(params.cam_rotation) * vertex; normal = normalize(mat3(params.cam_rotation) * normal); - vec3 reflection = normalize(reflect(normalize(vertex), normal)); #ifdef USE_SDFGI @@ -646,16 +633,39 @@ void main() { spec_accum /= blend_accum; } - if (params.use_sdfgi) { - reflection_light = blend_color(spec_accum, reflection_light); - ambient_light = blend_color(amb_accum, ambient_light); - } else { - reflection_light = spec_accum; - ambient_light = amb_accum; - } +#ifdef USE_SDFGI + reflection_light = blend_color(spec_accum, reflection_light); + ambient_light = blend_color(amb_accum, ambient_light); +#else + reflection_light = spec_accum; + ambient_light = amb_accum; +#endif } #endif } +} + +void main() { + ivec2 pos = ivec2(gl_GlobalInvocationID.xy); + +#ifdef MODE_HALF_RES + pos <<= 1; +#endif + if (any(greaterThanEqual(pos, params.screen_size))) { //too large, do nothing + return; + } + + vec4 ambient_light = vec4(0.0); + vec4 reflection_light = vec4(0.0); + + vec3 vertex = reconstruct_position(pos); + vertex.y = -vertex.y; + + process_gi(pos, vertex, ambient_light, reflection_light); + +#ifdef MODE_HALF_RES + pos >>= 1; +#endif imageStore(ambient_buffer, pos, ambient_light); imageStore(reflection_buffer, pos, reflection_light); diff --git a/servers/rendering/renderer_rd/shaders/resolve.glsl b/servers/rendering/renderer_rd/shaders/resolve.glsl index 9429a66dc9..e83c4ca93b 100644 --- a/servers/rendering/renderer_rd/shaders/resolve.glsl +++ b/servers/rendering/renderer_rd/shaders/resolve.glsl @@ -58,6 +58,116 @@ void main() { #else +#if 1 + + vec4 group1; + vec4 group2; + vec4 group3; + vec4 group4; + int best_index = 0; + + //2X + group1.x = texelFetch(source_depth, pos, 0).r; + group1.y = texelFetch(source_depth, pos, 1).r; + + //4X + if (params.sample_count >= 4) { + group1.z = texelFetch(source_depth, pos, 2).r; + group1.w = texelFetch(source_depth, pos, 3).r; + } + //8X + if (params.sample_count >= 8) { + group2.x = texelFetch(source_depth, pos, 4).r; + group2.y = texelFetch(source_depth, pos, 5).r; + group2.z = texelFetch(source_depth, pos, 6).r; + group2.w = texelFetch(source_depth, pos, 7).r; + } + //16X + if (params.sample_count >= 16) { + group3.x = texelFetch(source_depth, pos, 8).r; + group3.y = texelFetch(source_depth, pos, 9).r; + group3.z = texelFetch(source_depth, pos, 10).r; + group3.w = texelFetch(source_depth, pos, 11).r; + + group4.x = texelFetch(source_depth, pos, 12).r; + group4.y = texelFetch(source_depth, pos, 13).r; + group4.z = texelFetch(source_depth, pos, 14).r; + group4.w = texelFetch(source_depth, pos, 15).r; + } + + if (params.sample_count == 2) { + best_index = (pos.x & 1) ^ ((pos.y >> 1) & 1); //not much can be done here + } else if (params.sample_count == 4) { + vec4 freq = vec4(equal(group1, vec4(group1.x))); + freq += vec4(equal(group1, vec4(group1.y))); + freq += vec4(equal(group1, vec4(group1.z))); + freq += vec4(equal(group1, vec4(group1.w))); + + float min_f = freq.x; + best_index = 0; + if (freq.y < min_f) { + best_index = 1; + min_f = freq.y; + } + if (freq.z < min_f) { + best_index = 2; + min_f = freq.z; + } + if (freq.w < min_f) { + best_index = 3; + } + } else if (params.sample_count == 8) { + vec4 freq0 = vec4(equal(group1, vec4(group1.x))); + vec4 freq1 = vec4(equal(group2, vec4(group1.x))); + freq0 += vec4(equal(group1, vec4(group1.y))); + freq1 += vec4(equal(group2, vec4(group1.y))); + freq0 += vec4(equal(group1, vec4(group1.z))); + freq1 += vec4(equal(group2, vec4(group1.z))); + freq0 += vec4(equal(group1, vec4(group1.w))); + freq1 += vec4(equal(group2, vec4(group1.w))); + freq0 += vec4(equal(group1, vec4(group2.x))); + freq1 += vec4(equal(group2, vec4(group2.x))); + freq0 += vec4(equal(group1, vec4(group2.y))); + freq1 += vec4(equal(group2, vec4(group2.y))); + freq0 += vec4(equal(group1, vec4(group2.z))); + freq1 += vec4(equal(group2, vec4(group2.z))); + freq0 += vec4(equal(group1, vec4(group2.w))); + freq1 += vec4(equal(group2, vec4(group2.w))); + + float min_f0 = freq0.x; + int best_index0 = 0; + if (freq0.y < min_f0) { + best_index0 = 1; + min_f0 = freq0.y; + } + if (freq0.z < min_f0) { + best_index0 = 2; + min_f0 = freq0.z; + } + if (freq0.w < min_f0) { + best_index0 = 3; + min_f0 = freq0.w; + } + + float min_f1 = freq1.x; + int best_index1 = 4; + if (freq1.y < min_f1) { + best_index1 = 5; + min_f1 = freq1.y; + } + if (freq1.z < min_f1) { + best_index1 = 6; + min_f1 = freq1.z; + } + if (freq1.w < min_f1) { + best_index1 = 7; + min_f1 = freq1.w; + } + + best_index = mix(best_index0, best_index1, min_f0 < min_f1); + } + +#else float depths[16]; int depth_indices[16]; int depth_amount[16]; @@ -91,7 +201,7 @@ void main() { depth_least = depth_amount[j]; } } - +#endif best_depth = texelFetch(source_depth, pos, best_index).r; best_normal_roughness = texelFetch(source_normal_roughness, pos, best_index); #ifdef GIPROBE_RESOLVE diff --git a/servers/rendering/renderer_rd/shaders/scene_forward.glsl b/servers/rendering/renderer_rd/shaders/scene_forward.glsl index c3e7e2acbf..ea203c8abe 100644 --- a/servers/rendering/renderer_rd/shaders/scene_forward.glsl +++ b/servers/rendering/renderer_rd/shaders/scene_forward.glsl @@ -1282,7 +1282,7 @@ float light_process_spot_shadow(uint idx, vec3 vertex, vec3 normal) { } else { #endif //hard shadow - vec4 shadow_uv = vec4(splane.xy * spot_lights.data[idx].atlas_rect.zw + spot_lights.data[idx].atlas_rect.xy, z_norm, 1.0); + 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 @@ -2371,17 +2371,17 @@ FRAGMENT_SHADER_CODE if (bool(draw_call.flags & INSTANCE_FLAGS_USE_GI_BUFFERS)) { //use GI buffers - ivec2 coord; + vec2 coord; if (scene_data.gi_upscale_for_msaa) { - ivec2 base_coord = ivec2(gl_FragCoord.xy); - ivec2 closest_coord = base_coord; - float closest_ang = dot(normal, texelFetch(sampler2D(normal_roughness_buffer, material_samplers[SAMPLER_LINEAR_CLAMP]), base_coord, 0).xyz * 2.0 - 1.0); + vec2 base_coord = screen_uv; + vec2 closest_coord = base_coord; + float closest_ang = dot(normal, textureLod(sampler2D(normal_roughness_buffer, material_samplers[SAMPLER_LINEAR_CLAMP]), base_coord, 0.0).xyz * 2.0 - 1.0); for (int i = 0; i < 4; i++) { - const ivec2 neighbours[4] = ivec2[](ivec2(-1, 0), ivec2(1, 0), ivec2(0, -1), ivec2(0, 1)); - ivec2 neighbour_coord = base_coord + neighbours[i]; - float neighbour_ang = dot(normal, texelFetch(sampler2D(normal_roughness_buffer, material_samplers[SAMPLER_LINEAR_CLAMP]), neighbour_coord, 0).xyz * 2.0 - 1.0); + const vec2 neighbours[4] = vec2[](vec2(-1, 0), vec2(1, 0), vec2(0, -1), vec2(0, 1)); + vec2 neighbour_coord = base_coord + neighbours[i] * scene_data.screen_pixel_size; + float neighbour_ang = dot(normal, textureLod(sampler2D(normal_roughness_buffer, material_samplers[SAMPLER_LINEAR_CLAMP]), neighbour_coord, 0.0).xyz * 2.0 - 1.0); if (neighbour_ang > closest_ang) { closest_ang = neighbour_ang; closest_coord = neighbour_coord; @@ -2391,11 +2391,11 @@ FRAGMENT_SHADER_CODE coord = closest_coord; } else { - coord = ivec2(gl_FragCoord.xy); + coord = screen_uv; } - vec4 buffer_ambient = texelFetch(sampler2D(ambient_buffer, material_samplers[SAMPLER_LINEAR_CLAMP]), coord, 0); - vec4 buffer_reflection = texelFetch(sampler2D(reflection_buffer, material_samplers[SAMPLER_LINEAR_CLAMP]), coord, 0); + vec4 buffer_ambient = textureLod(sampler2D(ambient_buffer, material_samplers[SAMPLER_LINEAR_CLAMP]), coord, 0.0); + vec4 buffer_reflection = textureLod(sampler2D(reflection_buffer, material_samplers[SAMPLER_LINEAR_CLAMP]), coord, 0.0); ambient_light = mix(ambient_light, buffer_ambient.rgb, buffer_ambient.a); specular_light = mix(specular_light, buffer_reflection.rgb, buffer_reflection.a); diff --git a/servers/rendering/renderer_rd/shaders/scene_forward_inc.glsl b/servers/rendering/renderer_rd/shaders/scene_forward_inc.glsl index a37e32e1fc..e9b79e1560 100644 --- a/servers/rendering/renderer_rd/shaders/scene_forward_inc.glsl +++ b/servers/rendering/renderer_rd/shaders/scene_forward_inc.glsl @@ -153,22 +153,22 @@ scene_data; #define INSTANCE_FLAGS_SKELETON (1 << 19) #define INSTANCE_FLAGS_NON_UNIFORM_SCALE (1 << 20) -layout(set = 0, binding = 5, std430) restrict readonly buffer OmniLights { +layout(set = 0, binding = 4, std430) restrict readonly buffer OmniLights { LightData data[]; } omni_lights; -layout(set = 0, binding = 6, std430) restrict readonly buffer SpotLights { +layout(set = 0, binding = 5, std430) restrict readonly buffer SpotLights { LightData data[]; } spot_lights; -layout(set = 0, binding = 7) buffer restrict readonly ReflectionProbeData { +layout(set = 0, binding = 6) buffer restrict readonly ReflectionProbeData { ReflectionData data[]; } reflections; -layout(set = 0, binding = 8, std140) uniform DirectionalLights { +layout(set = 0, binding = 7, std140) uniform DirectionalLights { DirectionalLightData data[MAX_DIRECTIONAL_LIGHT_DATA_STRUCTS]; } directional_lights; @@ -180,7 +180,7 @@ struct Lightmap { mat3 normal_xform; }; -layout(set = 0, binding = 10, std140) restrict readonly buffer Lightmaps { +layout(set = 0, binding = 8, std140) restrict readonly buffer Lightmaps { Lightmap data[]; } lightmaps; @@ -189,22 +189,20 @@ struct LightmapCapture { vec4 sh[9]; }; -layout(set = 0, binding = 11, std140) restrict readonly buffer LightmapCaptures { +layout(set = 0, binding = 9, std140) restrict readonly buffer LightmapCaptures { LightmapCapture data[]; } lightmap_captures; -layout(set = 0, binding = 12) uniform texture2D decal_atlas; -layout(set = 0, binding = 13) uniform texture2D decal_atlas_srgb; +layout(set = 0, binding = 10) uniform texture2D decal_atlas; +layout(set = 0, binding = 11) uniform texture2D decal_atlas_srgb; -layout(set = 0, binding = 14, std430) restrict readonly buffer Decals { +layout(set = 0, binding = 12, std430) restrict readonly buffer Decals { DecalData data[]; } decals; -layout(set = 0, binding = 15) uniform texture2D directional_shadow_atlas; - -layout(set = 0, binding = 16, std430) restrict readonly buffer GlobalVariableData { +layout(set = 0, binding = 13, std430) restrict readonly buffer GlobalVariableData { vec4 data[]; } global_variables; @@ -218,7 +216,7 @@ struct SDFGIProbeCascadeData { float to_cell; // 1/bounds * grid_size }; -layout(set = 0, binding = 17, std140) uniform SDFGI { +layout(set = 0, binding = 14, std140) uniform SDFGI { vec3 grid_size; uint max_cascades; @@ -268,13 +266,15 @@ layout(set = 1, binding = 1) uniform textureCubeArray reflection_atlas; layout(set = 1, binding = 2) uniform texture2D shadow_atlas; -layout(set = 1, binding = 3) uniform texture2DArray lightmap_textures[MAX_LIGHTMAP_TEXTURES]; +layout(set = 1, binding = 3) uniform texture2D directional_shadow_atlas; + +layout(set = 1, binding = 4) uniform texture2DArray lightmap_textures[MAX_LIGHTMAP_TEXTURES]; #ifndef LOW_END_MODE -layout(set = 1, binding = 4) uniform texture3D gi_probe_textures[MAX_GI_PROBES]; +layout(set = 1, binding = 5) uniform texture3D gi_probe_textures[MAX_GI_PROBES]; #endif -layout(set = 1, binding = 5, std430) buffer restrict readonly ClusterBuffer { +layout(set = 1, binding = 6, std430) buffer restrict readonly ClusterBuffer { uint data[]; } cluster_buffer; @@ -283,10 +283,10 @@ cluster_buffer; #ifdef MODE_RENDER_SDF -layout(r16ui, set = 1, binding = 6) uniform restrict writeonly uimage3D albedo_volume_grid; -layout(r32ui, set = 1, binding = 7) uniform restrict writeonly uimage3D emission_grid; -layout(r32ui, set = 1, binding = 8) uniform restrict writeonly uimage3D emission_aniso_grid; -layout(r32ui, set = 1, binding = 9) uniform restrict uimage3D geom_facing_grid; +layout(r16ui, set = 1, binding = 7) uniform restrict writeonly uimage3D albedo_volume_grid; +layout(r32ui, set = 1, binding = 8) uniform restrict writeonly uimage3D emission_grid; +layout(r32ui, set = 1, binding = 9) uniform restrict writeonly uimage3D emission_aniso_grid; +layout(r32ui, set = 1, binding = 10) uniform restrict uimage3D geom_facing_grid; //still need to be present for shaders that use it, so remap them to something #define depth_buffer shadow_atlas @@ -295,17 +295,17 @@ layout(r32ui, set = 1, binding = 9) uniform restrict uimage3D geom_facing_grid; #else -layout(set = 1, binding = 6) uniform texture2D depth_buffer; -layout(set = 1, binding = 7) uniform texture2D color_buffer; +layout(set = 1, binding = 7) uniform texture2D depth_buffer; +layout(set = 1, binding = 8) uniform texture2D color_buffer; #ifndef LOW_END_MODE -layout(set = 1, binding = 8) uniform texture2D normal_roughness_buffer; -layout(set = 1, binding = 9) uniform texture2D ao_buffer; -layout(set = 1, binding = 10) uniform texture2D ambient_buffer; -layout(set = 1, binding = 11) uniform texture2D reflection_buffer; -layout(set = 1, binding = 12) uniform texture2DArray sdfgi_lightprobe_texture; -layout(set = 1, binding = 13) uniform texture3D sdfgi_occlusion_cascades; +layout(set = 1, binding = 9) uniform texture2D normal_roughness_buffer; +layout(set = 1, binding = 10) uniform texture2D ao_buffer; +layout(set = 1, binding = 11) uniform texture2D ambient_buffer; +layout(set = 1, binding = 12) uniform texture2D reflection_buffer; +layout(set = 1, binding = 13) uniform texture2DArray sdfgi_lightprobe_texture; +layout(set = 1, binding = 14) uniform texture3D sdfgi_occlusion_cascades; struct GIProbeData { mat4 xform; @@ -323,12 +323,12 @@ struct GIProbeData { uint mipmaps; }; -layout(set = 1, binding = 14, std140) uniform GIProbes { +layout(set = 1, binding = 15, std140) uniform GIProbes { GIProbeData data[MAX_GI_PROBES]; } gi_probes; -layout(set = 1, binding = 15) uniform texture3D volumetric_fog_texture; +layout(set = 1, binding = 16) uniform texture3D volumetric_fog_texture; #endif // LOW_END_MODE diff --git a/servers/rendering/renderer_rd/shaders/sdfgi_debug.glsl b/servers/rendering/renderer_rd/shaders/sdfgi_debug.glsl index 813ea29fa1..e4c3f3a84b 100644 --- a/servers/rendering/renderer_rd/shaders/sdfgi_debug.glsl +++ b/servers/rendering/renderer_rd/shaders/sdfgi_debug.glsl @@ -97,6 +97,8 @@ void main() { float blend = 0.0; #if 1 + // No interpolation + vec3 inv_dir = 1.0 / ray_dir; float rough = 0.5; @@ -161,114 +163,11 @@ void main() { hit_light *= (dot(max(vec3(0.0), (hit_normal * hit_aniso0)), vec3(1.0)) + dot(max(vec3(0.0), (-hit_normal * hit_aniso1)), vec3(1.0))); - if (blend > 0.0) { - light = mix(light, hit_light, blend); - blend = 0.0; - } else { - light = hit_light; - - //process blend - float blend_from = (float(params.probe_axis_size - 1) / 2.0) - 2.5; - float blend_to = blend_from + 2.0; - - vec3 cam_pos = params.cam_transform[3].xyz - cascades.data[i].offset; - cam_pos *= cascades.data[i].to_cell; - - pos += ray_dir * min(advance, max_advance); - vec3 inner_pos = pos - cam_pos; - - inner_pos = inner_pos * float(params.probe_axis_size - 1) / params.grid_size.x; - - 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); - - pos /= cascades.data[i].to_cell; - pos += cascades.data[i].offset; - ray_pos = pos; - hit = false; //continue trace for blend - - continue; - } - } + light = hit_light; break; } - light = mix(light, vec3(0.0), blend); - -#else - - vec3 inv_dir = 1.0 / ray_dir; - - bool hit = false; - vec4 light_accum = vec4(0.0); - - float blend_size = (params.grid_size.x / float(params.probe_axis_size - 1)) * 0.5; - - float radius_sizes[MAX_CASCADES]; - for (uint i = 0; i < params.max_cascades; i++) { - radius_sizes[i] = (1.0 / cascades.data[i].to_cell) * (params.grid_size.x * 0.5 - blend_size); - } - - float max_distance = radius_sizes[params.max_cascades - 1]; - float advance = 0; - while (advance < max_distance) { - for (uint i = 0; i < params.max_cascades; i++) { - if (advance < radius_sizes[i]) { - vec3 pos = (ray_pos + ray_dir * advance) - cascades.data[i].offset; - pos *= cascades.data[i].to_cell * pos_to_uvw; - - float distance = texture(sampler3D(sdf_cascades[i], linear_sampler), pos).r * 255.0 - 1.0; - - vec4 hit_light = vec4(0.0); - if (distance < 1.0) { - hit_light.a = max(0.0, 1.0 - distance); - hit_light.rgb = texture(sampler3D(light_cascades[i], linear_sampler), pos).rgb; - hit_light.rgb *= hit_light.a; - } - - distance /= cascades.data[i].to_cell; - - if (i < (params.max_cascades - 1)) { - pos = (ray_pos + ray_dir * advance) - cascades.data[i + 1].offset; - pos *= cascades.data[i + 1].to_cell * pos_to_uvw; - - float distance2 = texture(sampler3D(sdf_cascades[i + 1], linear_sampler), pos).r * 255.0 - 1.0; - - vec4 hit_light2 = vec4(0.0); - if (distance2 < 1.0) { - hit_light2.a = max(0.0, 1.0 - distance2); - hit_light2.rgb = texture(sampler3D(light_cascades[i + 1], linear_sampler), pos).rgb; - hit_light2.rgb *= hit_light2.a; - } - - float prev_radius = i == 0 ? 0.0 : radius_sizes[i - 1]; - float blend = (advance - prev_radius) / (radius_sizes[i] - prev_radius); - - distance2 /= cascades.data[i + 1].to_cell; - - hit_light = mix(hit_light, hit_light2, blend); - distance = mix(distance, distance2, blend); - } - - light_accum += hit_light; - advance += distance; - break; - } - } - - if (light_accum.a > 0.98) { - break; - } - } - - light = light_accum.rgb / light_accum.a; - #endif imageStore(screen_buffer, screen_pos, vec4(linear_to_srgb(light), 1.0)); diff --git a/servers/rendering/renderer_rd/shaders/sdfgi_direct_light.glsl b/servers/rendering/renderer_rd/shaders/sdfgi_direct_light.glsl index ed0a8a4b86..bcdfe8cc85 100644 --- a/servers/rendering/renderer_rd/shaders/sdfgi_direct_light.glsl +++ b/servers/rendering/renderer_rd/shaders/sdfgi_direct_light.glsl @@ -125,7 +125,10 @@ void main() { uint voxel_index = uint(gl_GlobalInvocationID.x); //used for skipping voxels every N frames - voxel_index = params.process_offset + voxel_index * params.process_increment; + if (params.process_increment > 1) { + voxel_index *= params.process_increment; + voxel_index += params.process_offset; + } if (voxel_index >= dispatch_data.total_count) { return; diff --git a/servers/rendering/renderer_rd/shaders/sdfgi_integrate.glsl b/servers/rendering/renderer_rd/shaders/sdfgi_integrate.glsl index 67630a3aa1..d122e7a38a 100644 --- a/servers/rendering/renderer_rd/shaders/sdfgi_integrate.glsl +++ b/servers/rendering/renderer_rd/shaders/sdfgi_integrate.glsl @@ -189,14 +189,12 @@ void main() { vec3 inv_dir = 1.0 / ray_dir; bool hit = false; - vec3 hit_normal; - vec3 hit_light; - vec3 hit_aniso0; - vec3 hit_aniso1; + uint hit_cascade; float bias = params.ray_bias; vec3 abs_ray_dir = abs(ray_dir); ray_pos += ray_dir * 1.0 / max(abs_ray_dir.x, max(abs_ray_dir.y, abs_ray_dir.z)) * bias / cascades.data[params.cascade].to_cell; + vec3 uvw; for (uint j = params.cascade; j < params.max_cascades; j++) { //convert to local bounds @@ -215,14 +213,12 @@ void main() { float advance = 0.0; - vec3 uvw; - while (advance < max_advance) { //read how much to advance from SDF uvw = (pos + ray_dir * advance) * pos_to_uvw; float distance = texture(sampler3D(sdf_cascades[j], linear_sampler), uvw).r * 255.0 - 1.0; - if (distance < 0.001) { + if (distance < 0.05) { //consider hit hit = true; break; @@ -232,17 +228,7 @@ void main() { } if (hit) { - const float EPSILON = 0.001; - hit_normal = normalize(vec3( - texture(sampler3D(sdf_cascades[j], linear_sampler), uvw + vec3(EPSILON, 0.0, 0.0)).r - texture(sampler3D(sdf_cascades[j], linear_sampler), uvw - vec3(EPSILON, 0.0, 0.0)).r, - texture(sampler3D(sdf_cascades[j], linear_sampler), uvw + vec3(0.0, EPSILON, 0.0)).r - texture(sampler3D(sdf_cascades[j], linear_sampler), uvw - vec3(0.0, EPSILON, 0.0)).r, - texture(sampler3D(sdf_cascades[j], linear_sampler), uvw + vec3(0.0, 0.0, EPSILON)).r - texture(sampler3D(sdf_cascades[j], linear_sampler), uvw - vec3(0.0, 0.0, EPSILON)).r)); - - hit_light = texture(sampler3D(light_cascades[j], linear_sampler), uvw).rgb; - vec4 aniso0 = texture(sampler3D(aniso0_cascades[j], linear_sampler), uvw); - hit_aniso0 = aniso0.rgb; - hit_aniso1 = vec3(aniso0.a, texture(sampler3D(aniso1_cascades[j], linear_sampler), uvw).rg); - + hit_cascade = j; break; } @@ -255,6 +241,17 @@ void main() { vec4 light; if (hit) { + const float EPSILON = 0.001; + vec3 hit_normal = normalize(vec3( + texture(sampler3D(sdf_cascades[hit_cascade], linear_sampler), uvw + vec3(EPSILON, 0.0, 0.0)).r - texture(sampler3D(sdf_cascades[hit_cascade], linear_sampler), uvw - vec3(EPSILON, 0.0, 0.0)).r, + texture(sampler3D(sdf_cascades[hit_cascade], linear_sampler), uvw + vec3(0.0, EPSILON, 0.0)).r - texture(sampler3D(sdf_cascades[hit_cascade], linear_sampler), uvw - vec3(0.0, EPSILON, 0.0)).r, + texture(sampler3D(sdf_cascades[hit_cascade], linear_sampler), uvw + vec3(0.0, 0.0, EPSILON)).r - texture(sampler3D(sdf_cascades[hit_cascade], linear_sampler), uvw - vec3(0.0, 0.0, EPSILON)).r)); + + vec3 hit_light = texture(sampler3D(light_cascades[hit_cascade], linear_sampler), uvw).rgb; + vec4 aniso0 = texture(sampler3D(aniso0_cascades[hit_cascade], linear_sampler), uvw); + vec3 hit_aniso0 = aniso0.rgb; + vec3 hit_aniso1 = vec3(aniso0.a, texture(sampler3D(aniso1_cascades[hit_cascade], linear_sampler), uvw).rg); + //one liner magic light.rgb = hit_light * (dot(max(vec3(0.0), (hit_normal * hit_aniso0)), vec3(1.0)) + dot(max(vec3(0.0), (-hit_normal * hit_aniso1)), vec3(1.0))); light.a = 1.0; @@ -490,13 +487,15 @@ void main() { //can't scroll, must look for position in parent cascade //to global coords - float probe_cell_size = float(params.grid_size.x / float(params.probe_axis_size - 1)) / cascades.data[params.cascade].to_cell; + float cell_to_probe = float(params.grid_size.x / float(params.probe_axis_size - 1)); + + float probe_cell_size = cell_to_probe / cascades.data[params.cascade].to_cell; vec3 probe_pos = cascades.data[params.cascade].offset + vec3(probe_cell) * probe_cell_size; //to parent local coords + float probe_cell_size_next = cell_to_probe / cascades.data[params.cascade + 1].to_cell; probe_pos -= cascades.data[params.cascade + 1].offset; - probe_pos *= cascades.data[params.cascade + 1].to_cell; - probe_pos = probe_pos * float(params.probe_axis_size - 1) / float(params.grid_size.x); + probe_pos /= probe_cell_size_next; ivec3 probe_posi = ivec3(probe_pos); //add up all light, no need to use occlusion here, since occlusion will do its work afterwards @@ -549,20 +548,28 @@ void main() { } } else { - // clear and let it re-raytrace, only for the last cascade, which happens very un-often - //scroll + //scroll at the edge of the highest cascade, just copy what is there, + //since its the closest we have anyway + for (uint j = 0; j < params.history_size; j++) { + ivec2 tex_pos; + tex_pos = probe_cell.xy; + tex_pos.x += probe_cell.z * int(params.probe_axis_size); + for (int i = 0; i < SH_SIZE; i++) { // copy from history texture + ivec3 src_pos = ivec3(tex_pos.x, tex_pos.y * SH_SIZE + i, int(j)); ivec3 dst_pos = ivec3(pos.x, pos.y * SH_SIZE + i, int(j)); - imageStore(lightprobe_history_scroll_texture, dst_pos, ivec4(0)); + ivec4 value = imageLoad(lightprobe_history_texture, dst_pos); + imageStore(lightprobe_history_scroll_texture, dst_pos, value); } } for (int i = 0; i < SH_SIZE; i++) { // copy from average texture - ivec2 dst_pos = ivec2(pos.x, pos.y * SH_SIZE + i); - imageStore(lightprobe_average_scroll_texture, dst_pos, ivec4(0)); + ivec2 spos = ivec2(pos.x, pos.y * SH_SIZE + i); + ivec4 average = imageLoad(lightprobe_average_texture, spos); + imageStore(lightprobe_average_scroll_texture, spos, average); } } |