diff options
Diffstat (limited to 'servers/rendering/renderer_rd/shaders')
47 files changed, 3814 insertions, 2261 deletions
diff --git a/servers/rendering/renderer_rd/shaders/blit.glsl b/servers/rendering/renderer_rd/shaders/blit.glsl index 967da1e6e4..8051f96738 100644 --- a/servers/rendering/renderer_rd/shaders/blit.glsl +++ b/servers/rendering/renderer_rd/shaders/blit.glsl @@ -5,6 +5,7 @@ #VERSION_DEFINES layout(push_constant, binding = 0, std140) uniform Pos { + vec4 src_rect; vec4 dst_rect; vec2 eye_center; @@ -22,8 +23,8 @@ 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 = base_arr[gl_VertexIndex]; - vec2 vtx = data.dst_rect.xy + uv * data.dst_rect.zw; + 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); } @@ -34,6 +35,7 @@ void main() { #VERSION_DEFINES layout(push_constant, binding = 0, std140) uniform Pos { + vec4 src_rect; vec4 dst_rect; vec2 eye_center; diff --git a/servers/rendering/renderer_rd/shaders/blur_raster.glsl b/servers/rendering/renderer_rd/shaders/blur_raster.glsl new file mode 100644 index 0000000000..f8b4e3f610 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/blur_raster.glsl @@ -0,0 +1,138 @@ +/* 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 + + 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); + 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); + color *= blur.glow_strength; + frag_color = color; + } + +#undef GLOW_ADD + + if (bool(blur.flags & FLAG_GLOW_FIRST_PASS)) { +#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)); + } + +#endif + +#ifdef MODE_COPY + vec4 color = textureLod(source_color, uv_interp, 0.0); + frag_color = color; +#endif +} diff --git a/servers/rendering/renderer_rd/shaders/blur_raster_inc.glsl b/servers/rendering/renderer_rd/shaders/blur_raster_inc.glsl new file mode 100644 index 0000000000..52bf2886b5 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/blur_raster_inc.glsl @@ -0,0 +1,21 @@ +#define FLAG_HORIZONTAL (1 << 0) +#define FLAG_USE_ORTHOGONAL_PROJECTION (1 << 1) +#define FLAG_GLOW_FIRST_PASS (1 << 2) + +layout(push_constant, binding = 1, std430) uniform Blur { + vec2 pixel_size; + uint flags; + uint pad; + + // Glow. + float glow_strength; + float glow_bloom; + float glow_hdr_threshold; + float glow_hdr_scale; + + float glow_exposure; + float glow_white; + float glow_luminance_cap; + float glow_auto_exposure_grey; +} +blur; diff --git a/servers/rendering/renderer_rd/shaders/bokeh_dof.glsl b/servers/rendering/renderer_rd/shaders/bokeh_dof.glsl index b70e0b6bd5..0438671dd2 100644 --- a/servers/rendering/renderer_rd/shaders/bokeh_dof.glsl +++ b/servers/rendering/renderer_rd/shaders/bokeh_dof.glsl @@ -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/bokeh_dof_inc.glsl b/servers/rendering/renderer_rd/shaders/bokeh_dof_inc.glsl new file mode 100644 index 0000000000..fadea1631c --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/bokeh_dof_inc.glsl @@ -0,0 +1,37 @@ +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 + +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/bokeh_dof_raster.glsl b/servers/rendering/renderer_rd/shaders/bokeh_dof_raster.glsl new file mode 100644 index 0000000000..a3b3938ee9 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/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/canvas.glsl b/servers/rendering/renderer_rd/shaders/canvas.glsl index cf4c77db0d..65a621b203 100644 --- a/servers/rendering/renderer_rd/shaders/canvas.glsl +++ b/servers/rendering/renderer_rd/shaders/canvas.glsl @@ -65,7 +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; @@ -91,7 +91,6 @@ void main() { uint instancing = draw_data.flags & FLAGS_INSTANCING_MASK; #ifdef USE_ATTRIBUTES - if (instancing > 1) { // trails @@ -101,66 +100,64 @@ void main() { uint offset = trail_size * stride * gl_InstanceIndex; - mat4 matrix; vec4 pcolor; + vec2 new_vertex; { uint boffset = offset + bone_attrib.x * stride; - matrix = 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)) * weight_attrib.x; - pcolor = transforms.data[boffset + 3] * weight_attrib.x; + 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; - matrix += 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)) * weight_attrib.y; - pcolor += transforms.data[boffset + 3] * weight_attrib.y; + 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; - matrix += 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)) * weight_attrib.z; - pcolor += transforms.data[boffset + 3] * weight_attrib.z; + 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; - matrix += 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)) * weight_attrib.w; - pcolor += transforms.data[boffset + 3] * weight_attrib.w; + 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 + 4]; + instance_custom = transforms.data[offset + 3]; + vertex = new_vertex; color *= pcolor; - - matrix = transpose(matrix); - world_matrix = world_matrix * matrix; - } 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 (instancing == 1) { - uint stride = 2; - { if (bool(draw_data.flags & FLAGS_INSTANCING_HAS_COLORS)) { - stride += 1; + color *= transforms.data[offset]; + offset += 1; } + if (bool(draw_data.flags & FLAGS_INSTANCING_HAS_CUSTOM_DATA)) { - stride += 1; + instance_custom = transforms.data[offset]; } - } - 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; + matrix = transpose(matrix); + world_matrix = world_matrix * matrix; } - - if (bool(draw_data.flags & FLAGS_INSTANCING_HAS_CUSTOM_DATA)) { - instance_custom = transforms.data[offset]; - } - - matrix = transpose(matrix); - world_matrix = world_matrix * matrix; } #if !defined(USE_ATTRIBUTES) && !defined(USE_PRIMITIVE) @@ -283,7 +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; } @@ -460,6 +457,14 @@ 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); +} + +vec2 msdf_map(vec2 value, vec2 in_min, vec2 in_max, vec2 out_min, vec2 out_max) { + return out_min + (out_max - out_min) * (value - in_min) / (in_max - in_min); +} + void main() { vec4 color = color_interp; vec2 uv = uv_interp; @@ -487,7 +492,34 @@ 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; diff --git a/servers/rendering/renderer_rd/shaders/canvas_sdf.glsl b/servers/rendering/renderer_rd/shaders/canvas_sdf.glsl index 65a554e839..2bdfbabfcf 100644 --- a/servers/rendering/renderer_rd/shaders/canvas_sdf.glsl +++ b/servers/rendering/renderer_rd/shaders/canvas_sdf.glsl @@ -7,7 +7,7 @@ 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; @@ -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 451f9b0089..0cff505cae 100644 --- a/servers/rendering/renderer_rd/shaders/canvas_uniforms_inc.glsl +++ b/servers/rendering/renderer_rd/shaders/canvas_uniforms_inc.glsl @@ -24,6 +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 diff --git a/servers/rendering/renderer_rd/shaders/cluster_render.glsl b/servers/rendering/renderer_rd/shaders/cluster_render.glsl index da7d189281..6d95722a57 100644 --- a/servers/rendering/renderer_rd/shaders/cluster_render.glsl +++ b/servers/rendering/renderer_rd/shaders/cluster_render.glsl @@ -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/cube_to_dp.glsl b/servers/rendering/renderer_rd/shaders/cube_to_dp.glsl index dfbce29119..69b895ed29 100644 --- a/servers/rendering/renderer_rd/shaders/cube_to_dp.glsl +++ b/servers/rendering/renderer_rd/shaders/cube_to_dp.glsl @@ -7,8 +7,7 @@ layout(push_constant, binding = 1, std430) uniform Params { float z_far; float z_near; - bool z_flip; - uint pad; + vec2 texel_size; vec4 screen_rect; } params; @@ -35,22 +34,23 @@ layout(set = 0, binding = 0) uniform samplerCube source_cube; layout(push_constant, binding = 1, 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 9fa84657d1..63f0ce690e 100644 --- a/servers/rendering/renderer_rd/shaders/cubemap_downsampler.glsl +++ b/servers/rendering/renderer_rd/shaders/cubemap_downsampler.glsl @@ -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..b329e67293 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/cubemap_downsampler_inc.glsl @@ -0,0 +1,48 @@ +layout(push_constant, binding = 1, 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_raster.glsl b/servers/rendering/renderer_rd/shaders/cubemap_filter_raster.glsl new file mode 100644 index 0000000000..324d306218 --- /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, binding = 1, 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, binding = 1, 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 ce7c03c1d4..28f4dc59ec 100644 --- a/servers/rendering/renderer_rd/shaders/cubemap_roughness.glsl +++ b/servers/rendering/renderer_rd/shaders/cubemap_roughness.glsl @@ -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; 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..be12be5dec --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/cubemap_roughness_inc.glsl @@ -0,0 +1,94 @@ +#define M_PI 3.14159265359 + +layout(push_constant, binding = 1, 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 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; +} + +// 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..2570308816 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/cubemap_roughness_raster.glsl @@ -0,0 +1,63 @@ +/* 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); + + 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); + + float ndotl = clamp(dot(N, L), 0.0, 1.0); + + if (ndotl > 0.0) { + sum.rgb += textureLod(source_cube, L, 0.0).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 index ccaad13311..158096d3c7 100644 --- a/servers/rendering/renderer_rd/shaders/decal_data_inc.glsl +++ b/servers/rendering/renderer_rd/shaders/decal_data_inc.glsl @@ -1,18 +1,18 @@ 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; + 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; - float upper_fade; - float lower_fade; - mat3x4 normal_xform; - vec3 normal; - float normal_fade; + 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/fsr_upscale.glsl b/servers/rendering/renderer_rd/shaders/fsr_upscale.glsl new file mode 100644 index 0000000000..4e2ba84033 --- /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) 2007-2021 Juan Linietsky, Ariel Manzur. */ +/* Copyright (c) 2014-2021 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, binding = 1, 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 bfd5c4c88d..5528ea3659 100644 --- a/servers/rendering/renderer_rd/shaders/gi.glsl +++ b/servers/rendering/renderer_rd/shaders/gi.glsl @@ -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; @@ -65,9 +65,9 @@ layout(set = 0, binding = 15, std140) uniform SDFGI { } sdfgi; -#define MAX_GI_PROBES 8 +#define MAX_VOXEL_GI_INSTANCES 8 -struct GIProbeData { +struct VoxelGIData { mat4 xform; vec3 bounds; float dynamic_range; @@ -77,18 +77,18 @@ struct GIProbeData { bool blend_ambient; uint texture_slot; - float anisotropy_strength; - float ambient_occlusion; - float ambient_occlusion_size; + uint pad0; + uint pad1; + uint pad2; uint mipmaps; }; -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 { ivec2 screen_size; @@ -97,12 +97,10 @@ layout(push_constant, binding = 0, std430) uniform Params { 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 +153,7 @@ 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 +291,7 @@ 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 +321,7 @@ 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 +492,26 @@ 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 +532,7 @@ 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 +545,21 @@ 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; + 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); } - - 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; - } - - 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 +591,9 @@ 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 +605,9 @@ 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 5dc2d08a3b..25d87ca45d 100644 --- a/servers/rendering/renderer_rd/shaders/giprobe_write.glsl +++ b/servers/rendering/renderer_rd/shaders/giprobe_write.glsl @@ -202,12 +202,7 @@ 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 +237,35 @@ 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 index 2fce258cff..fdc7729338 100644 --- a/servers/rendering/renderer_rd/shaders/light_data_inc.glsl +++ b/servers/rendering/renderer_rd/shaders/light_data_inc.glsl @@ -3,31 +3,31 @@ #define LIGHT_BAKE_STATIC 2 struct LightData { //this structure needs to be as packed as possible - vec3 position; - float inv_radius; + highp vec3 position; + highp float inv_radius; - vec3 direction; - float size; + mediump vec3 direction; + highp float size; - vec3 color; - float attenuation; + mediump vec3 color; + mediump float attenuation; - float cone_attenuation; - float cone_angle; - float specular_amount; + mediump float cone_attenuation; + mediump float cone_angle; + mediump 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 + 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; - float shadow_volumetric_fog_fade; + mediump float shadow_volumetric_fog_fade; uint bake_mode; - vec4 projector_rect; //projector rect in srgb decal atlas + highp vec4 projector_rect; //projector rect in srgb decal atlas }; #define REFLECTION_AMBIENT_DISABLED 0 @@ -35,53 +35,53 @@ struct LightData { //this structure needs to be as packed as possible #define REFLECTION_AMBIENT_COLOR 2 struct ReflectionData { - vec3 box_extents; - float index; - vec3 box_offset; + highp vec3 box_extents; + mediump float index; + highp vec3 box_offset; uint mask; - vec3 ambient; // ambient color - float intensity; + 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 - mat4 local_matrix; // up to here for spot and omni, rest is for directional + 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 { - vec3 direction; - float energy; - vec3 color; - float size; - float specular; + mediump vec3 direction; + mediump float energy; + mediump vec3 color; + mediump float size; + mediump float specular; uint mask; - float softshadow_angle; - float soft_shadow_scale; + highp float softshadow_angle; + highp float soft_shadow_scale; bool blend_splits; bool shadow_enabled; - float fade_from; - float fade_to; + highp float fade_from; + highp float fade_to; uvec2 pad; uint bake_mode; - 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; + 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; + mediump vec4 shadow_color1; + mediump vec4 shadow_color2; + mediump vec4 shadow_color3; + mediump vec4 shadow_color4; + highp vec2 uv_scale1; + highp vec2 uv_scale2; + highp vec2 uv_scale3; + highp vec2 uv_scale4; }; 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..3cde9923fa --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/luminance_reduce_raster_inc.glsl @@ -0,0 +1,11 @@ + +layout(push_constant, binding = 1, 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 beaff10793..328becbc20 100644 --- a/servers/rendering/renderer_rd/shaders/particles.glsl +++ b/servers/rendering/renderer_rd/shaders/particles.glsl @@ -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; @@ -452,128 +455,167 @@ void main() { #endif - for (uint i = 0; i < FRAME.collider_count; i++) { - vec3 normal; - float depth; - bool col = false; + if (FRAME.collider_count == 1 && FRAME.colliders[0].type == COLLIDER_TYPE_2D_SDF) { + //2D collision - vec3 rel_vec = PARTICLE.xform[3].xyz - FRAME.colliders[i].transform[3].xyz; - vec3 local_pos = rel_vec * mat3(FRAME.colliders[i].transform); + 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)); - switch (FRAME.colliders[i].type) { - case COLLIDER_TYPE_SPHERE: { - float d = length(rel_vec) - (particle_size + FRAME.colliders[i].extents.x); + vec4 sdf_to_screen = vec4(FRAME.colliders[0].extents, FRAME.colliders[0].scale); - if (d < 0.0) { - col = true; - depth = -d; - normal = normalize(rel_vec); - } + vec2 uv_pos = sdf_pos * sdf_to_screen.xy + sdf_to_screen.zw; - } break; - case COLLIDER_TYPE_BOX: { - vec3 abs_pos = abs(local_pos); - vec3 sgn_pos = sign(local_pos); + 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; - if (any(greaterThan(abs_pos, FRAME.colliders[i].extents))) { - //point outside box + vec3 rel_vec = PARTICLE.xform[3].xyz - FRAME.colliders[i].transform[3].xyz; + vec3 local_pos = rel_vec * mat3(FRAME.colliders[i].transform); - vec3 closest = min(abs_pos, FRAME.colliders[i].extents); - vec3 rel = abs_pos - closest; - depth = length(rel) - particle_size; - if (depth < 0.0) { + 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; - normal = mat3(FRAME.colliders[i].transform) * (normalize(rel) * sgn_pos); - depth = -depth; + depth = -d; + normal = normalize(rel_vec); } - } 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); + + } 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 { - normal = vec3(0, 0, 1); + //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); } - 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); + } - } 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; + } - 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)); + } - 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; - } 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; - if (any(greaterThan(abs(local_pos_bottom), FRAME.colliders[i].extents))) { - continue; - } + vec3 uvw_pos = vec3(local_pos_bottom / FRAME.colliders[i].extents) * 0.5 + 0.5; - const float DELTA = 1.0 / 8192.0; + float y = 1.0 - texture(sampler2D(height_field_texture, material_samplers[SAMPLER_LINEAR_CLAMP]), uvw_pos.xz).r; - vec3 uvw_pos = vec3(local_pos_bottom / FRAME.colliders[i].extents) * 0.5 + 0.5; + if (y > uvw_pos.y) { + //inside heightfield - float y = 1.0 - texture(sampler2D(height_field_texture, material_samplers[SAMPLER_LINEAR_CLAMP]), uvw_pos.xz).r; + 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; - if (y > uvw_pos.y) { - //inside heightfield + normal = normalize(cross(pos1 - pos2, pos1 - pos3)); + float local_y = (vec3(local_pos / FRAME.colliders[i].extents) * 0.5 + 0.5).y; - 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; + col = true; + depth = dot(normal, pos1) - dot(normal, local_pos_bottom); + } - normal = normalize(cross(pos1 - pos2, pos1 - pos3)); - float local_y = (vec3(local_pos / FRAME.colliders[i].extents) * 0.5 + 0.5).y; + } break; + } - col = true; - depth = dot(normal, pos1) - dot(normal, local_pos_bottom); + 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); } - - } 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); } } } diff --git a/servers/rendering/renderer_rd/shaders/particles_copy.glsl b/servers/rendering/renderer_rd/shaders/particles_copy.glsl index e2bebadf1a..e88e68b511 100644 --- a/servers/rendering/renderer_rd/shaders/particles_copy.glsl +++ b/servers/rendering/renderer_rd/shaders/particles_copy.glsl @@ -53,6 +53,11 @@ layout(push_constant, binding = 0, std430) uniform Params { vec3 align_up; uint align_mode; + + bool order_by_lifetime; + uint lifetime_split; + bool lifetime_reverse; + uint pad; } params; @@ -80,7 +85,6 @@ void main() { #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 @@ -93,14 +97,48 @@ void main() { } else { particle = uint(sort_buffer.data[particle].y); //use index from sort buffer } -#endif +#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 (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 dont fit precisely in the history frames, must do a tiny bit of + // 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; @@ -165,12 +203,17 @@ void main() { #ifdef 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]; diff --git a/servers/rendering/renderer_rd/shaders/resolve.glsl b/servers/rendering/renderer_rd/shaders/resolve.glsl index 2286a26485..fecf812a8c 100644 --- a/servers/rendering/renderer_rd/shaders/resolve.glsl +++ b/servers/rendering/renderer_rd/shaders/resolve.glsl @@ -6,6 +6,11 @@ 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; @@ -13,9 +18,9 @@ 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 @@ -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/scene_forward_aa_inc.glsl b/servers/rendering/renderer_rd/shaders/scene_forward_aa_inc.glsl index 99714b4504..97c913d489 100644 --- a/servers/rendering/renderer_rd/shaders/scene_forward_aa_inc.glsl +++ b/servers/rendering/renderer_rd/shaders/scene_forward_aa_inc.glsl @@ -2,7 +2,7 @@ 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)))); + (0.1 + abs(sin(13.0 * p.y + p.x)))); } float hash_3d(vec3 p) { @@ -29,8 +29,7 @@ float compute_alpha_hash_threshold(vec3 pos, float hash_scale) { 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)))); + 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; diff --git a/servers/rendering/renderer_rd/shaders/scene_forward_clustered.glsl b/servers/rendering/renderer_rd/shaders/scene_forward_clustered.glsl index 1d67a3f1df..e4628b2d5a 100644 --- a/servers/rendering/renderer_rd/shaders/scene_forward_clustered.glsl +++ b/servers/rendering/renderer_rd/shaders/scene_forward_clustered.glsl @@ -6,6 +6,8 @@ #include "scene_forward_clustered_inc.glsl" +#define SHADER_IS_SRGB false + /* INPUT ATTRIBS */ layout(location = 0) in vec3 vertex_attrib; @@ -95,7 +97,7 @@ layout(location = 8) out float dp_clip; #endif -layout(location = 9) out flat uint instance_index; +layout(location = 9) out flat uint instance_index_interp; invariant gl_Position; @@ -107,18 +109,20 @@ void main() { color_interp = color_attrib; #endif - instance_index = draw_call.instance_index; + uint instance_index = draw_call.instance_index; bool is_multimesh = bool(instances.data[instance_index].flags & INSTANCE_FLAGS_MULTIMESH); if (!is_multimesh) { instance_index += gl_InstanceIndex; } + instance_index_interp = instance_index; + mat4 world_matrix = instances.data[instance_index].transform; mat3 world_normal_matrix; if (bool(instances.data[instance_index].flags & INSTANCE_FLAGS_NON_UNIFORM_SCALE)) { - world_normal_matrix = inverse(mat3(world_matrix)); + world_normal_matrix = transpose(inverse(mat3(world_matrix))); } else { world_normal_matrix = mat3(world_matrix); } @@ -356,6 +360,29 @@ void main() { #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" /* Varyings */ @@ -389,7 +416,7 @@ layout(location = 8) in float dp_clip; #endif -layout(location = 9) in flat uint instance_index; +layout(location = 9) in flat uint instance_index_interp; //defines to keep compatibility with vertex @@ -426,8 +453,8 @@ layout(location = 4) out float depth_output_buffer; #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 @@ -448,14 +475,15 @@ layout(location = 0) out vec4 frag_color; #if !defined(MODE_RENDER_DEPTH) && !defined(MODE_UNSHADED) -#include "scene_forward_lights_inc.glsl" +/* Make a default specular mode SPECULAR_SCHLICK_GGX. */ +#if !defined(SPECULAR_DISABLED) && !defined(SPECULAR_SCHLICK_GGX) && !defined(SPECULAR_BLINN) && !defined(SPECULAR_PHONG) && !defined(SPECULAR_TOON) +#define SPECULAR_SCHLICK_GGX +#endif -#ifdef USE_FORWARD_GI +#include "scene_forward_lights_inc.glsl" #include "scene_forward_gi_inc.glsl" -#endif //USE_FORWARD_GI - #endif //!defined(MODE_RENDER_DEPTH) && !defined(MODE_UNSHADED) #ifndef MODE_RENDER_DEPTH @@ -502,14 +530,14 @@ vec4 fog_process(vec3 vertex) { } } - 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.fog_density)); - if (abs(scene_data.fog_height_density) > 0.001) { + if (abs(scene_data.fog_height_density) >= 0.0001) { float y = (scene_data.camera_matrix * vec4(vertex, 1.0)).y; - float y_dist = scene_data.fog_height - y; + float y_dist = y - scene_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.fog_height_density)); fog_amount = max(vfog_amount, fog_amount); } @@ -542,14 +570,15 @@ void main() { discard; #endif + uint instance_index = instance_index_interp; + //lay out everything, whathever is unused is optimized away anyway vec3 vertex = vertex_interp; vec3 view = -normalize(vertex_interp); 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; @@ -572,7 +601,7 @@ 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); @@ -634,12 +663,8 @@ void main() { } #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 @@ -662,7 +687,7 @@ void main() { #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 @@ -670,7 +695,7 @@ void main() { #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 @@ -798,25 +823,35 @@ void main() { 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 @@ -826,7 +861,12 @@ void main() { } 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); @@ -835,7 +875,11 @@ void main() { 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; + } } } } @@ -848,9 +892,9 @@ void main() { #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); @@ -867,6 +911,7 @@ void main() { if (scene_data.use_reflection_cubemap) { vec3 ref_vec = reflect(-view, normal); + 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 @@ -879,6 +924,7 @@ void main() { specular_light = textureLod(samplerCube(radiance_cubemap, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), ref_vec, roughness * MAX_ROUGHNESS_LOD).rgb; #endif //USE_RADIANCE_CUBEMAP_ARRAY + specular_light *= horizon * horizon; specular_light *= scene_data.ambient_light_color_energy.a; } @@ -904,7 +950,7 @@ void main() { } #endif // USE_LIGHTMAP #if defined(CUSTOM_IRRADIANCE_USED) - ambient_light = mix(specular_light, custom_irradiance.rgb, custom_irradiance.a); + ambient_light = mix(ambient_light, custom_irradiance.rgb, custom_irradiance.a); #endif #endif //!defined(MODE_RENDER_DEPTH) && !defined(MODE_UNSHADED) @@ -926,15 +972,15 @@ void main() { const float c4 = 0.886227; const float c5 = 0.247708; ambient_light += (c1 * lightmap_captures.data[index].sh[8].rgb * (wnormal.x * wnormal.x - wnormal.y * wnormal.y) + - c3 * lightmap_captures.data[index].sh[6].rgb * wnormal.z * wnormal.z + - c4 * lightmap_captures.data[index].sh[0].rgb - - c5 * lightmap_captures.data[index].sh[6].rgb + - 2.0 * c1 * lightmap_captures.data[index].sh[4].rgb * wnormal.x * wnormal.y + - 2.0 * c1 * lightmap_captures.data[index].sh[7].rgb * wnormal.x * wnormal.z + - 2.0 * c1 * lightmap_captures.data[index].sh[5].rgb * wnormal.y * wnormal.z + - 2.0 * c2 * lightmap_captures.data[index].sh[3].rgb * wnormal.x + - 2.0 * c2 * lightmap_captures.data[index].sh[1].rgb * wnormal.y + - 2.0 * c2 * lightmap_captures.data[index].sh[2].rgb * wnormal.z); + c3 * lightmap_captures.data[index].sh[6].rgb * wnormal.z * wnormal.z + + c4 * lightmap_captures.data[index].sh[0].rgb - + c5 * lightmap_captures.data[index].sh[6].rgb + + 2.0 * c1 * lightmap_captures.data[index].sh[4].rgb * wnormal.x * wnormal.y + + 2.0 * c1 * lightmap_captures.data[index].sh[7].rgb * wnormal.x * wnormal.z + + 2.0 * c1 * lightmap_captures.data[index].sh[5].rgb * wnormal.y * wnormal.z + + 2.0 * c2 * lightmap_captures.data[index].sh[3].rgb * wnormal.x + + 2.0 * c2 * lightmap_captures.data[index].sh[1].rgb * wnormal.y + + 2.0 * c2 * lightmap_captures.data[index].sh[2].rgb * wnormal.z); } 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); @@ -968,9 +1014,9 @@ void main() { 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; @@ -1042,7 +1088,7 @@ void main() { } } - 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)); @@ -1054,12 +1100,12 @@ void main() { 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) { @@ -1073,9 +1119,8 @@ void main() { specular_light = spec_accum.rgb; ambient_light = amb_accum.rgb; } -#else - 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; @@ -1106,7 +1151,7 @@ void main() { 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 if (scene_data.ssao_enabled) { float ssao = texture(sampler2D(ao_buffer, material_samplers[SAMPLER_LINEAR_CLAMP]), screen_uv).r; @@ -1187,7 +1232,7 @@ void main() { 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 @@ -1212,9 +1257,10 @@ void main() { // 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; @@ -1233,306 +1279,178 @@ void main() { 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); - shadow_color = directional_lights.data[i].shadow_color3.rgb; + 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); + } - } else { - vec4 v = vec4(vertex, 1.0); + blend_count++; + } - BIAS_FUNC(v, 3) + if (blend_count < blend_max) { + vec4 v = vec4(vertex, 1.0); - pssm_coord = (directional_lights.data[i].shadow_matrix4 * v); - pssm_coord /= pssm_coord.w; + BIAS_FUNC(v, 3) + + 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); - shadow_color = directional_lights.data[i].shadow_color4.rgb; - } + 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); + } + } - if (directional_lights.data[i].blend_splits) { - vec3 shadow_color_blend = vec3(0.0); - float pssm_blend; - float shadow2; + } else { //no soft shadows + vec4 pssm_coord; 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); } 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); } 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, 2) - BIAS_FUNC(v, 3) + pssm_coord = (directional_lights.data[i].shadow_matrix3 * v); - 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; + } else { + vec4 v = vec4(vertex, 1.0); - 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; + BIAS_FUNC(v, 3) - transmittance_z = z - shadow_z; + pssm_coord = (directional_lights.data[i].shadow_matrix4 * v); } -#endif - } - pssm_coord /= pssm_coord.w; + pssm_coord /= pssm_coord.w; - shadow = sample_directional_pcf_shadow(directional_shadow_atlas, scene_data.directional_shadow_pixel_size * directional_lights.data[i].soft_shadow_scale, pssm_coord); + shadow = sample_directional_pcf_shadow(directional_shadow_atlas, scene_data.directional_shadow_pixel_size * directional_lights.data[i].soft_shadow_scale, 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) { + 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) + } - if (directional_lights.data[i].blend_splits) { - float pssm_blend; + pssm_coord /= pssm_coord.w; - 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) { - 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) + 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); } - - 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 = 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); @@ -1540,6 +1458,7 @@ void main() { 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) { @@ -1562,8 +1481,8 @@ void main() { 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) { @@ -1572,8 +1491,8 @@ void main() { 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) { @@ -1582,8 +1501,8 @@ void main() { 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; @@ -1593,221 +1512,208 @@ void main() { 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_gloss, #endif #ifdef LIGHT_ANISOTROPY_USED - binormal, tangent, anisotropy, + binormal, tangent, anisotropy, #endif -#ifdef USE_SOFT_SHADOW - directional_lights.data[i].size, -#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 - } + 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_gloss, #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 - } + 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_gloss, #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 @@ -1819,126 +1725,130 @@ void main() { #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.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); } + } - 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 @@ -1991,9 +1901,8 @@ void main() { // 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_RENDER_DEPTH - } +} 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 ca75d6300e..be29cf4f58 100644 --- a/servers/rendering/renderer_rd/shaders/scene_forward_clustered_inc.glsl +++ b/servers/rendering/renderer_rd/shaders/scene_forward_clustered_inc.glsl @@ -1,7 +1,7 @@ #define M_PI 3.14159265359 #define ROUGHNESS_MAX_LOD 5 -#define MAX_GI_PROBES 8 +#define MAX_VOXEL_GI_INSTANCES 8 #if defined(has_GL_KHR_shader_subgroup_ballot) && defined(has_GL_KHR_shader_subgroup_arithmetic) @@ -15,7 +15,7 @@ #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) #ifndef NORMAL_USED #define NORMAL_USED #endif @@ -52,38 +52,42 @@ 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_PARTICLE_TRAIL_SHIFT 16 +#define INSTANCE_FLAGS_FADE_SHIFT 24 //3 bits of stride #define INSTANCE_FLAGS_PARTICLE_TRAIL_MASK 0xFF -#define INSTANCE_FLAGS_NON_UNIFORM_SCALE (1 << 24) - -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, std430) restrict readonly buffer 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 +99,7 @@ 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; @@ -104,32 +108,32 @@ 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; -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; @@ -153,7 +157,7 @@ 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; @@ -174,17 +178,12 @@ 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; @@ -209,9 +208,8 @@ layout(set = 1, binding = 0, std140) uniform SceneData { 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; @@ -275,7 +273,7 @@ layout(set = 1, binding = 5) uniform texture2D directional_shadow_atlas; layout(set = 1, binding = 6) uniform texture2DArray lightmap_textures[MAX_LIGHTMAP_TEXTURES]; -layout(set = 1, binding = 7) uniform texture3D gi_probe_textures[MAX_GI_PROBES]; +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[]; @@ -306,7 +304,7 @@ 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 { +struct VoxelGIData { mat4 xform; vec3 bounds; float dynamic_range; @@ -322,10 +320,10 @@ struct GIProbeData { uint mipmaps; }; -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; diff --git a/servers/rendering/renderer_rd/shaders/scene_forward_gi_inc.glsl b/servers/rendering/renderer_rd/shaders/scene_forward_gi_inc.glsl index b41f16cbe7..c88bd0a14b 100644 --- a/servers/rendering/renderer_rd/shaders/scene_forward_gi_inc.glsl +++ b/servers/rendering/renderer_rd/shaders/scene_forward_gi_inc.glsl @@ -48,24 +48,24 @@ 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, 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); +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 * 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; } - 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; //radiance @@ -83,26 +83,26 @@ void gi_probe_compute(uint index, vec3 position, vec3 normal, vec3 ref_vec, mat3 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); + 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(gi_probe_textures[index], cell_size, position, dir, cone_angle_tan, max_distance, gi_probes.data[index].bias); + 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 (gi_probes.data[index].blend_ambient) { + 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 *= gi_probes.data[index].dynamic_range; + light *= voxel_gi_instances.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) { + 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 *= gi_probes.data[index].dynamic_range; + irr_light.rgb *= voxel_gi_instances.data[index].dynamic_range; //irr_light=vec3(0.0); out_spec += vec4(irr_light.rgb * blend, blend); diff --git a/servers/rendering/renderer_rd/shaders/scene_forward_lights_inc.glsl b/servers/rendering/renderer_rd/shaders/scene_forward_lights_inc.glsl index 32a86cb166..d22f936a35 100644 --- a/servers/rendering/renderer_rd/shaders/scene_forward_lights_inc.glsl +++ b/servers/rendering/renderer_rd/shaders/scene_forward_lights_inc.glsl @@ -73,19 +73,18 @@ vec3 F0(float metallic, float specular, vec3 albedo) { 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, +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_curve, float transmittance_boost, float transmittance_z, #endif #ifdef LIGHT_RIM_USED - float rim, float rim_tint, vec3 rim_color, + float rim, float rim_tint, #endif #ifdef LIGHT_CLEARCOAT_USED float clearcoat, float clearcoat_gloss, @@ -93,14 +92,13 @@ void light_compute(vec3 N, vec3 L, vec3 V, vec3 light_color, float attenuation, #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) { + 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 @@ -112,11 +110,7 @@ void light_compute(vec3 N, vec3 L, vec3 V, vec3 light_color, float attenuation, #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); @@ -126,30 +120,15 @@ void light_compute(vec3 N, vec3 L, vec3 V, vec3 light_color, float attenuation, #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 @@ -189,43 +168,37 @@ void light_compute(vec3 N, vec3 L, vec3 V, vec3 light_color, float attenuation, #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; + diffuse_light += rim_light * rim * mix(vec3(1.0), albedo, rim_tint) * light_color; #endif #ifdef LIGHT_TRANSMITTANCE_USED -#ifdef SSS_MODE_SKIN - { +#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); + 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; + 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 } - -#endif //SSS_MODE_SKIN +#else #endif //LIGHT_TRANSMITTANCE_USED } - float roughness = unpackUnorm4x8(orms).y; if (roughness > 0.0) { // FIXME: roughness == 0 should not disable specular light entirely // D @@ -234,11 +207,10 @@ void light_compute(vec3 N, vec3 L, vec3 V, vec3 light_color, float attenuation, //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; + float blinn = pow(cNdotH, shininess); + blinn *= (shininess + 2.0) * (1.0 / (8.0 * M_PI)); - specular_light += light_color * intensity * attenuation * specular_amount; + specular_light += light_color * attenuation * specular_amount * blinn * f0 * orms_unpacked.w; #elif defined(SPECULAR_PHONG) @@ -246,10 +218,9 @@ void light_compute(vec3 N, vec3 L, vec3 V, vec3 light_color, float attenuation, 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); + phong *= (shininess + 1.0) * (1.0 / (8.0 * M_PI)); - specular_light += light_color * intensity * attenuation * specular_amount; + specular_light += light_color * attenuation * specular_amount * phong * f0 * orms_unpacked.w; #elif defined(SPECULAR_TOON) @@ -307,16 +278,16 @@ void light_compute(vec3 N, vec3 L, vec3 V, vec3 light_color, float attenuation, } #ifdef USE_SHADOW_TO_OPACITY - alpha = min(alpha, clamp(1.0 - attenuation), 0.0, 1.0)); + alpha = min(alpha, clamp(1.0 - attenuation, 0.0, 1.0)); #endif #endif //defined(LIGHT_CODE_USED) } -#ifndef USE_NO_SHADOWS +#ifndef SHADOWS_DISABLED // Interleaved Gradient Noise -// http://www.iryoku.com/next-generation-post-processing-in-call-of-duty-advanced-warfare +// 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))); @@ -327,7 +298,7 @@ float sample_directional_pcf_shadow(texture2D shadow, vec2 shadow_pixel_size, ve float depth = coord.z; //if only one sample is taken, take it from the center - if (scene_data.directional_soft_shadow_samples == 1) { + if (sc_directional_soft_shadow_samples == 0) { return textureProj(sampler2DShadow(shadow, shadow_sampler), vec4(pos, depth, 1.0)); } @@ -341,19 +312,19 @@ float sample_directional_pcf_shadow(texture2D shadow, vec2 shadow_pixel_size, ve float avg = 0.0; - for (uint i = 0; i < scene_data.directional_soft_shadow_samples; i++) { + for (uint i = 0; i < sc_directional_soft_shadow_samples; i++) { avg += textureProj(sampler2DShadow(shadow, shadow_sampler), vec4(pos + shadow_pixel_size * (disk_rotation * scene_data.directional_soft_shadow_kernel[i].xy), depth, 1.0)); } - return avg * (1.0 / float(scene_data.directional_soft_shadow_samples)); + return avg * (1.0 / float(sc_directional_soft_shadow_samples)); } -float sample_pcf_shadow(texture2D shadow, vec2 shadow_pixel_size, vec4 coord) { +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 (scene_data.soft_shadow_samples == 1) { + if (sc_soft_shadow_samples == 0) { return textureProj(sampler2DShadow(shadow, shadow_sampler), vec4(pos, depth, 1.0)); } @@ -367,11 +338,54 @@ float sample_pcf_shadow(texture2D shadow, vec2 shadow_pixel_size, vec4 coord) { float avg = 0.0; - for (uint i = 0; i < scene_data.soft_shadow_samples; i++) { + for (uint i = 0; i < sc_soft_shadow_samples; i++) { avg += textureProj(sampler2DShadow(shadow, shadow_sampler), vec4(pos + shadow_pixel_size * (disk_rotation * scene_data.soft_shadow_kernel[i].xy), depth, 1.0)); } - return avg * (1.0 / float(scene_data.soft_shadow_samples)); + 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.shadow_atlas_pixel_size / uv_rect.zw; + + for (uint i = 0; i < sc_soft_shadow_samples; i++) { + vec2 offset = offset_scale * (disk_rotation * scene_data.soft_shadow_kernel[i].xy); + vec2 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) { @@ -387,7 +401,7 @@ float sample_directional_soft_shadow(texture2D shadow, vec3 pssm_coord, vec2 tex disk_rotation = mat2(vec2(cr, -sr), vec2(sr, cr)); } - for (uint i = 0; i < scene_data.directional_penumbra_shadow_samples; i++) { + for (uint i = 0; i < sc_directional_penumbra_shadow_samples; i++) { vec2 suv = pssm_coord.xy + (disk_rotation * scene_data.directional_penumbra_shadow_kernel[i].xy) * tex_scale; float d = textureLod(sampler2D(shadow, material_samplers[SAMPLER_LINEAR_CLAMP]), suv, 0.0).r; if (d < pssm_coord.z) { @@ -403,12 +417,12 @@ float sample_directional_soft_shadow(texture2D shadow, vec3 pssm_coord, vec2 tex tex_scale *= penumbra; float s = 0.0; - for (uint i = 0; i < scene_data.directional_penumbra_shadow_samples; i++) { + for (uint i = 0; i < sc_directional_penumbra_shadow_samples; i++) { vec2 suv = pssm_coord.xy + (disk_rotation * scene_data.directional_penumbra_shadow_kernel[i].xy) * tex_scale; s += textureProj(sampler2DShadow(shadow, shadow_sampler), vec4(suv, pssm_coord.z, 1.0)); } - return s / float(scene_data.directional_penumbra_shadow_samples); + return s / float(sc_directional_penumbra_shadow_samples); } else { //no blockers found, so no shadow @@ -416,7 +430,7 @@ float sample_directional_soft_shadow(texture2D shadow, vec3 pssm_coord, vec2 tex } } -#endif //USE_NO_SHADOWS +#endif // SHADOWS_DISABLED float get_omni_attenuation(float distance, float inv_range, float decay) { float nd = distance * inv_range; @@ -428,29 +442,28 @@ float get_omni_attenuation(float distance, float inv_range, float decay) { } float light_process_omni_shadow(uint idx, vec3 vertex, vec3 normal) { -#ifndef USE_NO_SHADOWS +#ifndef SHADOWS_DISABLED if (omni_lights.data[idx].shadow_enabled) { // there is a shadowmap + vec2 texel_size = scene_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; - vec3 light_rel_vec = omni_lights.data[idx].position - vertex; - float light_length = length(light_rel_vec); + // 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; - vec4 v = vec4(vertex, 1.0); + vec3 local_vert = (omni_lights.data[idx].shadow_matrix * vec4(vertex, 1.0)).xyz; - vec4 splane = (omni_lights.data[idx].shadow_matrix * v); - float shadow_len = length(splane.xyz); //need to remember shadow len from here + float shadow_len = length(local_vert); //need to remember shadow len from here + vec3 shadow_dir = normalize(local_vert); - { - 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); - } + 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; -#ifdef USE_SOFT_SHADOWS - if (omni_lights.data[idx].soft_shadow_size > 0.0) { + if (sc_use_light_soft_shadows && omni_lights.data[idx].soft_shadow_size > 0.0) { //soft shadow //find blocker @@ -466,30 +479,29 @@ float light_process_omni_shadow(uint idx, vec3 vertex, vec3 normal) { 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)); + 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 < scene_data.penumbra_shadow_samples; i++) { + for (uint i = 0; i < sc_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; + vec3 pos = local_vert + tangent * disk.x + bitangent * disk.y; pos = normalize(pos); - vec4 uv_rect = omni_lights.data[idx].atlas_rect; + + vec4 uv_rect = base_uv_rect; if (pos.z >= 0.0) { - pos.z += 1.0; - uv_rect.y += uv_rect.w; - } else { - pos.z = 1.0 - pos.z; + uv_rect.xy += flip_offset; } + pos.z = 1.0 + abs(pos.z); pos.xy /= pos.z; pos.xy = pos.xy * 0.5 + 0.5; @@ -512,20 +524,20 @@ float light_process_omni_shadow(uint idx, vec3 vertex, vec3 normal) { 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++) { + for (uint i = 0; i < sc_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; + vec3 pos = local_vert + tangent * disk.x + bitangent * disk.y; pos = normalize(pos); - vec4 uv_rect = omni_lights.data[idx].atlas_rect; + pos = normalize(pos + normal_bias); + + vec4 uv_rect = base_uv_rect; if (pos.z >= 0.0) { - pos.z += 1.0; - uv_rect.y += uv_rect.w; - } else { - pos.z = 1.0 - pos.z; + uv_rect.xy += flip_offset; } + pos.z = 1.0 + abs(pos.z); pos.xy /= pos.z; pos.xy = pos.xy * 0.5 + 0.5; @@ -533,36 +545,27 @@ float light_process_omni_shadow(uint idx, vec3 vertex, vec3 normal) { shadow += textureProj(sampler2DShadow(shadow_atlas, shadow_sampler), vec4(pos.xy, z_norm, 1.0)); } - shadow /= float(scene_data.penumbra_shadow_samples); + shadow /= float(sc_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; + vec4 uv_rect = base_uv_rect; - clamp_rect.y += clamp_rect.w; - - } else { - splane.z = 1.0 - splane.z; + vec3 shadow_sample = normalize(shadow_dir + normal_bias); + if (shadow_sample.z >= 0.0) { + uv_rect.xy += flip_offset; + flip_offset *= -1.0; } - 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 + 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); } -#endif return shadow; } @@ -571,18 +574,17 @@ float light_process_omni_shadow(uint idx, vec3 vertex, vec3 normal) { 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, +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_curve, float transmittance_boost, #endif #ifdef LIGHT_RIM_USED - float rim, float rim_tint, vec3 rim_color, + float rim, float rim_tint, #endif #ifdef LIGHT_CLEARCOAT_USED float clearcoat, float clearcoat_gloss, @@ -590,9 +592,6 @@ void light_process_omni(uint idx, vec3 vertex, vec3 eye_vec, vec3 normal, vec3 v #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); @@ -600,14 +599,12 @@ void light_process_omni(uint idx, vec3 vertex, vec3 eye_vec, vec3 normal, vec3 v 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) { + 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)); } -#endif #ifdef LIGHT_TRANSMITTANCE_USED float transmittance_z = transmittance_depth; //no transmittance by default @@ -618,20 +615,22 @@ void light_process_omni(uint idx, vec3 vertex, vec3 eye_vec, vec3 normal, vec3 v //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); + 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.shadow_atlas_pixel_size,clamp_rect.xy + clamp_rect.zw - scene_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; @@ -639,78 +638,77 @@ void light_process_omni(uint idx, vec3 vertex, vec3 eye_vec, vec3 normal, vec3 v } #endif -#if 0 - - if (omni_lights.data[idx].projector_rect != vec4(0.0)) { + 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) { - local_v.z += 1.0; atlas_rect.y += atlas_rect.w; - - } else { - local_v.z = 1.0 - local_v.z; } + 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; - 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 (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; - } + 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; + 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; + 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); + 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; - } + 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; + 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; - } + 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); + 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; + } } -#endif light_attenuation *= shadow; - light_compute(normal, normalize(light_rel_vec), eye_vec, color, light_attenuation, f0, orms, omni_lights.data[idx].specular_amount, + 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_curve, transmittance_boost, transmittance_z, #endif #ifdef LIGHT_RIM_USED - rim * omni_attenuation, rim_tint, rim_color, + rim * omni_attenuation, rim_tint, #endif #ifdef LIGHT_CLEARCOAT_USED clearcoat, clearcoat_gloss, @@ -718,47 +716,33 @@ void light_process_omni(uint idx, vec3 vertex, vec3 eye_vec, vec3 normal, vec3 v #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 +#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; - //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; + 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))); - float shadow; + //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; -#ifdef USE_SOFT_SHADOWS - if (spot_lights.data[idx].soft_shadow_size > 0.0) { + 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; @@ -775,11 +759,11 @@ float light_process_spot_shadow(uint idx, vec3 vertex, vec3 normal) { float uv_size = spot_lights.data[idx].soft_shadow_size * z_norm * spot_lights.data[idx].soft_shadow_scale; vec2 clamp_max = spot_lights.data[idx].atlas_rect.xy + spot_lights.data[idx].atlas_rect.zw; - for (uint i = 0; i < scene_data.penumbra_shadow_samples; i++) { + for (uint i = 0; i < sc_penumbra_shadow_samples; i++) { vec2 suv = shadow_uv + (disk_rotation * scene_data.penumbra_shadow_kernel[i].xy) * uv_size; 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) { + if (d < splane.z) { blocker_average += d; blocker_count += 1.0; } @@ -792,49 +776,55 @@ float light_process_spot_shadow(uint idx, vec3 vertex, vec3 normal) { uv_size *= penumbra; shadow = 0.0; - for (uint i = 0; i < scene_data.penumbra_shadow_samples; i++) { + for (uint i = 0; i < sc_penumbra_shadow_samples; i++) { vec2 suv = shadow_uv + (disk_rotation * scene_data.penumbra_shadow_kernel[i].xy) * uv_size; 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 += textureProj(sampler2DShadow(shadow_atlas, shadow_sampler), vec4(suv, splane.z, 1.0)); } - shadow /= float(scene_data.penumbra_shadow_samples); + shadow /= float(sc_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); - + vec3 shadow_uv = vec3(splane.xy * spot_lights.data[idx].atlas_rect.zw + spot_lights.data[idx].atlas_rect.xy, splane.z); shadow = sample_pcf_shadow(shadow_atlas, spot_lights.data[idx].soft_shadow_scale * scene_data.shadow_atlas_pixel_size, shadow_uv); -#ifdef USE_SOFT_SHADOWS } -#endif return shadow; } -#endif //USE_NO_SHADOWS +#endif // SHADOWS_DISABLED 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, +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_curve, float transmittance_boost, #endif #ifdef LIGHT_RIM_USED - float rim, float rim_tint, vec3 rim_color, + float rim, float rim_tint, #endif #ifdef LIGHT_CLEARCOAT_USED float clearcoat, float clearcoat_gloss, @@ -842,9 +832,6 @@ void light_process_spot(uint idx, vec3 vertex, vec3 eye_vec, vec3 normal, vec3 v #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; @@ -858,53 +845,71 @@ void light_process_spot(uint idx, vec3 vertex, vec3 eye_vec, vec3 normal, vec3 v 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) { + 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)); } -#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)); + 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; - //reconstruct depth - shadow_z /= spot_lights.data[idx].inv_radius; + + 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 = normal_to_panorama(splane.xyz) * 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 = normal_to_panorama(splane_ddx.xyz) * 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 = normal_to_panorama(splane_ddy.xyz) * 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, color, light_attenuation, f0, orms, spot_lights.data[idx].specular_amount, + 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_curve, transmittance_boost, transmittance_z, #endif #ifdef LIGHT_RIM_USED - rim * spot_attenuation, rim_tint, rim_color, + rim * spot_attenuation, rim_tint, #endif #ifdef LIGHT_CLEARCOAT_USED clearcoat, clearcoat_gloss, @@ -912,12 +917,6 @@ void light_process_spot(uint idx, vec3 vertex, vec3 eye_vec, vec3 normal, vec3 v #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); } @@ -957,7 +956,7 @@ void reflection_process(uint ref_index, vec3 vertex, vec3 normal, float roughnes 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; + 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); diff --git a/servers/rendering/renderer_rd/shaders/scene_forward_mobile.glsl b/servers/rendering/renderer_rd/shaders/scene_forward_mobile.glsl index b38b8d803d..e92fbecfd0 100644 --- a/servers/rendering/renderer_rd/shaders/scene_forward_mobile.glsl +++ b/servers/rendering/renderer_rd/shaders/scene_forward_mobile.glsl @@ -7,6 +7,8 @@ /* 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; @@ -59,27 +61,27 @@ layout(location = 11) in vec4 weight_attrib; /* Varyings */ -layout(location = 0) out vec3 vertex_interp; +layout(location = 0) highp out vec3 vertex_interp; #ifdef NORMAL_USED -layout(location = 1) out vec3 normal_interp; +layout(location = 1) mediump out vec3 normal_interp; #endif #if defined(COLOR_USED) -layout(location = 2) out vec4 color_interp; +layout(location = 2) mediump out vec4 color_interp; #endif #ifdef UV_USED -layout(location = 3) out vec2 uv_interp; +layout(location = 3) mediump out vec2 uv_interp; #endif #if defined(UV2_USED) || defined(USE_LIGHTMAP) -layout(location = 4) out vec2 uv2_interp; +layout(location = 4) mediump out vec2 uv2_interp; #endif #if defined(TANGENT_USED) || defined(NORMAL_MAP_USED) || defined(LIGHT_ANISOTROPY_USED) -layout(location = 5) out vec3 tangent_interp; -layout(location = 6) out vec3 binormal_interp; +layout(location = 5) mediump out vec3 tangent_interp; +layout(location = 6) mediump out vec3 binormal_interp; #endif #ifdef MATERIAL_UNIFORMS_USED @@ -92,9 +94,21 @@ layout(set = MATERIAL_UNIFORM_SET, binding = 0, std140) uniform MaterialUniforms #ifdef MODE_DUAL_PARABOLOID -layout(location = 8) out float dp_clip; +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; @@ -112,7 +126,7 @@ void main() { mat3 world_normal_matrix; if (bool(draw_call.flags & INSTANCE_FLAGS_NON_UNIFORM_SCALE)) { - world_normal_matrix = inverse(mat3(world_matrix)); + world_normal_matrix = transpose(inverse(mat3(world_matrix))); } else { world_normal_matrix = mat3(world_matrix); } @@ -234,7 +248,13 @@ 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) @@ -352,44 +372,95 @@ void main() { #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) in vec3 vertex_interp; +layout(location = 0) highp in vec3 vertex_interp; #ifdef NORMAL_USED -layout(location = 1) in vec3 normal_interp; +layout(location = 1) mediump in vec3 normal_interp; #endif #if defined(COLOR_USED) -layout(location = 2) in vec4 color_interp; +layout(location = 2) mediump in vec4 color_interp; #endif #ifdef UV_USED -layout(location = 3) in vec2 uv_interp; +layout(location = 3) mediump in vec2 uv_interp; #endif #if defined(UV2_USED) || defined(USE_LIGHTMAP) -layout(location = 4) in vec2 uv2_interp; +layout(location = 4) mediump in vec2 uv2_interp; #endif #if defined(TANGENT_USED) || defined(NORMAL_MAP_USED) || defined(LIGHT_ANISOTROPY_USED) -layout(location = 5) in vec3 tangent_interp; -layout(location = 6) in vec3 binormal_interp; +layout(location = 5) mediump in vec3 tangent_interp; +layout(location = 6) mediump in vec3 binormal_interp; #endif #ifdef MODE_DUAL_PARABOLOID -layout(location = 8) in float dp_clip; +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 world_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 @@ -428,7 +499,7 @@ 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; +layout(location = 0) out mediump vec4 frag_color; #endif // MODE_MULTIPLE_RENDER_TARGETS #endif // RENDER DEPTH @@ -437,6 +508,11 @@ layout(location = 0) out vec4 frag_color; #if !defined(MODE_RENDER_DEPTH) && !defined(MODE_UNSHADED) +/* Make a default specular mode SPECULAR_SCHLICK_GGX. */ +#if !defined(SPECULAR_DISABLED) && !defined(SPECULAR_SCHLICK_GGX) && !defined(SPECULAR_BLINN) && !defined(SPECULAR_PHONG) && !defined(SPECULAR_TOON) +#define SPECULAR_SCHLICK_GGX +#endif + #include "scene_forward_lights_inc.glsl" #endif //!defined(MODE_RENDER_DEPTH) && !defined(MODE_UNSHADED) @@ -478,14 +554,14 @@ vec4 fog_process(vec3 vertex) { } } - 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.fog_density)); - if (abs(scene_data.fog_height_density) > 0.001) { + if (abs(scene_data.fog_height_density) >= 0.0001) { float y = (scene_data.camera_matrix * vec4(vertex, 1.0)).y; - float y_dist = scene_data.fog_height - y; + float y_dist = y - scene_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.fog_height_density)); fog_amount = max(vfog_amount, fog_amount); } @@ -509,7 +585,6 @@ void main() { vec3 backlight = vec3(0.0); vec4 transmittance_color = vec4(0.0); float transmittance_depth = 0.0; - float transmittance_curve = 1.0; float transmittance_boost = 0.0; float metallic = 0.0; float specular = 0.5; @@ -622,7 +697,7 @@ void main() { #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 @@ -630,7 +705,7 @@ void main() { #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 @@ -673,7 +748,7 @@ void main() { // to maximize VGPR usage // Draw "fixed" fog before volumetric fog to ensure volumetric fog can appear in front of the sky. - if (scene_data.fog_enabled) { + if (!sc_disable_fog && scene_data.fog_enabled) { fog = fog_process(vertex); } @@ -691,7 +766,7 @@ void main() { vec3 vertex_ddx = dFdx(vertex); vec3 vertex_ddy = dFdy(vertex); - { //Decals + if (!sc_disable_decals) { //Decals // must implement uint decal_indices = draw_call.decals.x; @@ -712,25 +787,35 @@ void main() { 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 @@ -740,7 +825,12 @@ void main() { } 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); @@ -749,7 +839,11 @@ void main() { 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; + } } } } //Decals @@ -759,9 +853,9 @@ void main() { #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); @@ -778,6 +872,7 @@ void main() { if (scene_data.use_reflection_cubemap) { vec3 ref_vec = reflect(-view, normal); + 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 @@ -790,6 +885,7 @@ void main() { specular_light = textureLod(samplerCube(radiance_cubemap, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), ref_vec, roughness * MAX_ROUGHNESS_LOD).rgb; #endif //USE_RADIANCE_CUBEMAP_ARRAY + specular_light *= horizon * horizon; specular_light *= scene_data.ambient_light_color_energy.a; } @@ -838,15 +934,15 @@ void main() { const float c4 = 0.886227; const float c5 = 0.247708; ambient_light += (c1 * lightmap_captures.data[index].sh[8].rgb * (wnormal.x * wnormal.x - wnormal.y * wnormal.y) + - c3 * lightmap_captures.data[index].sh[6].rgb * wnormal.z * wnormal.z + - c4 * lightmap_captures.data[index].sh[0].rgb - - c5 * lightmap_captures.data[index].sh[6].rgb + - 2.0 * c1 * lightmap_captures.data[index].sh[4].rgb * wnormal.x * wnormal.y + - 2.0 * c1 * lightmap_captures.data[index].sh[7].rgb * wnormal.x * wnormal.z + - 2.0 * c1 * lightmap_captures.data[index].sh[5].rgb * wnormal.y * wnormal.z + - 2.0 * c2 * lightmap_captures.data[index].sh[3].rgb * wnormal.x + - 2.0 * c2 * lightmap_captures.data[index].sh[1].rgb * wnormal.y + - 2.0 * c2 * lightmap_captures.data[index].sh[2].rgb * wnormal.z); + c3 * lightmap_captures.data[index].sh[6].rgb * wnormal.z * wnormal.z + + c4 * lightmap_captures.data[index].sh[0].rgb - + c5 * lightmap_captures.data[index].sh[6].rgb + + 2.0 * c1 * lightmap_captures.data[index].sh[4].rgb * wnormal.x * wnormal.y + + 2.0 * c1 * lightmap_captures.data[index].sh[7].rgb * wnormal.x * wnormal.z + + 2.0 * c1 * lightmap_captures.data[index].sh[5].rgb * wnormal.y * wnormal.z + + 2.0 * c2 * lightmap_captures.data[index].sh[3].rgb * wnormal.x + + 2.0 * c2 * lightmap_captures.data[index].sh[1].rgb * wnormal.y + + 2.0 * c2 * lightmap_captures.data[index].sh[2].rgb * wnormal.z); } 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); @@ -887,7 +983,7 @@ void main() { // skipping ssao, do we remove ssao totally? - { //Reflection probes + 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); @@ -928,7 +1024,7 @@ void main() { 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 @@ -953,8 +1049,8 @@ void main() { // LIGHTING #if !defined(MODE_RENDER_DEPTH) && !defined(MODE_UNSHADED) - { //directional light - + 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; @@ -1230,6 +1326,8 @@ void main() { } } +#endif // SHADOWS_DISABLED + for (uint i = 0; i < 8; i++) { if (i >= scene_data.directional_light_count) { break; @@ -1242,16 +1340,16 @@ void main() { // 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), 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), 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 @@ -1259,13 +1357,12 @@ void main() { #ifdef LIGHT_TRANSMITTANCE_USED transmittance_color, transmittance_depth, - transmittance_curve, transmittance_boost, transmittance_z, #endif */ #ifdef LIGHT_RIM_USED - rim, rim_tint, albedo, + rim, rim_tint, #endif #ifdef LIGHT_CLEARCOAT_USED clearcoat, clearcoat_gloss, @@ -1276,15 +1373,12 @@ void main() { #ifdef USE_SOFT_SHADOW directional_lights.data[i].size, #endif -#ifdef USE_SHADOW_TO_OPACITY - alpha, -#endif diffuse_light, specular_light); } } //directional light - { //omni lights + 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; @@ -1298,11 +1392,11 @@ void main() { break; } - float shadow = light_process_omni_shadow(light_index, vertex, view); + 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, + light_process_omni(light_index, vertex, view, normal, vertex_ddx, vertex_ddy, f0, orms, shadow, albedo, alpha, #ifdef LIGHT_BACKLIGHT_USED backlight, #endif @@ -1310,14 +1404,12 @@ void main() { #ifdef LIGHT_TRANSMITTANCE_USED transmittance_color, transmittance_depth, - transmittance_curve, transmittance_boost, #endif */ #ifdef LIGHT_RIM_USED rim, rim_tint, - albedo, #endif #ifdef LIGHT_CLEARCOAT_USED clearcoat, clearcoat_gloss, @@ -1325,14 +1417,11 @@ void main() { #ifdef LIGHT_ANISOTROPY_USED tangent, binormal, anisotropy, #endif -#ifdef USE_SHADOW_TO_OPACITY - alpha, -#endif diffuse_light, specular_light); } } //omni lights - { //spot lights + if (!sc_disable_spot_lights) { //spot lights uint light_indices = draw_call.spot_lights.x; for (uint i = 0; i < 8; i++) { @@ -1347,11 +1436,11 @@ void main() { break; } - float shadow = light_process_spot_shadow(light_index, vertex, view); + 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, + light_process_spot(light_index, vertex, view, normal, vertex_ddx, vertex_ddy, f0, orms, shadow, albedo, alpha, #ifdef LIGHT_BACKLIGHT_USED backlight, #endif @@ -1359,14 +1448,12 @@ void main() { #ifdef LIGHT_TRANSMITTANCE_USED transmittance_color, transmittance_depth, - transmittance_curve, transmittance_boost, #endif */ #ifdef LIGHT_RIM_USED rim, rim_tint, - albedo, #endif #ifdef LIGHT_CLEARCOAT_USED clearcoat, clearcoat_gloss, @@ -1374,9 +1461,6 @@ void main() { #ifdef LIGHT_ANISOTROPY_USED tangent, binormal, anisotropy, #endif -#ifdef USE_SHADOW_TO_OPACITY - alpha, -#endif diffuse_light, specular_light); } } //spot lights @@ -1392,7 +1476,7 @@ void main() { #ifdef USE_OPAQUE_PREPASS - if (alpha < opaque_prepass_threshold) { + if (alpha < scene_data.opaque_prepass_threshold) { discard; } @@ -1464,12 +1548,15 @@ void main() { frag_color = vec4(albedo, alpha); #else // MODE_UNSHADED frag_color = vec4(emission + ambient_light + diffuse_light + specular_light, alpha); - //frag_color = vec4(1.0); #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 index 0156b58574..a9a4fce82a 100644 --- a/servers/rendering/renderer_rd/shaders/scene_forward_mobile_inc.glsl +++ b/servers/rendering/renderer_rd/shaders/scene_forward_mobile_inc.glsl @@ -1,4 +1,9 @@ #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" @@ -11,12 +16,12 @@ /* don't exceed 128 bytes!! */ /* put instance data into our push content, not a array */ layout(push_constant, binding = 0, std430) uniform DrawCall { - mat4 transform; // 64 - 64 + 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 - vec4 lightmap_uv_scale; // 16 - 96 doubles as uv_offset when needed + highp vec4 lightmap_uv_scale; // 16 - 96 doubles as uv_offset when needed uvec2 reflection_probes; // 08 - 104 uvec2 omni_lights; // 08 - 112 @@ -46,12 +51,16 @@ 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) @@ -60,24 +69,22 @@ layout(set = 0, binding = 2) uniform sampler shadow_sampler; //3 bits of stride #define INSTANCE_FLAGS_PARTICLE_TRAIL_MASK 0xFF -#define INSTANCE_FLAGS_NON_UNIFORM_SCALE (1 << 24) - -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, std430) restrict readonly buffer 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; @@ -86,100 +93,97 @@ directional_lights; #define LIGHTMAP_FLAG_USE_SPECULAR_DIRECTION 2 struct Lightmap { - mat3 normal_xform; + mediump 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; struct LightmapCapture { - vec4 sh[9]; + mediump 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 mediump texture2D decal_atlas; +layout(set = 0, binding = 12) uniform mediump 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 { - vec4 data[]; +layout(set = 0, binding = 14, std430) restrict readonly buffer GlobalVariableData { + highp vec4 data[]; } global_variables; /* Set 1: Render Pass (changes per render pass) */ layout(set = 1, binding = 0, std140) uniform SceneData { - mat4 projection_matrix; - mat4 inv_projection_matrix; - - mat4 camera_matrix; - mat4 inv_camera_matrix; + highp mat4 projection_matrix; + highp mat4 inv_projection_matrix; + highp mat4 camera_matrix; + highp mat4 inv_camera_matrix; - vec2 viewport_size; - vec2 screen_pixel_size; + // only used for multiview + highp mat4 projection_matrix_view[MAX_VIEWS]; + highp mat4 inv_projection_matrix_view[MAX_VIEWS]; - //use vec4s because std140 doesnt 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]; + highp vec2 viewport_size; + highp vec2 screen_pixel_size; - uint directional_penumbra_shadow_samples; - uint directional_soft_shadow_samples; - uint penumbra_shadow_samples; - uint soft_shadow_samples; + // 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]; - vec4 ambient_light_color_energy; + mediump vec4 ambient_light_color_energy; - float ambient_color_sky_mix; + mediump float ambient_color_sky_mix; bool use_ambient_light; bool use_ambient_cubemap; bool use_reflection_cubemap; - mat3 radiance_inverse_xform; + mediump mat3 radiance_inverse_xform; - vec2 shadow_atlas_pixel_size; - vec2 directional_shadow_pixel_size; + highp vec2 shadow_atlas_pixel_size; + highp vec2 directional_shadow_pixel_size; uint directional_light_count; - float dual_paraboloid_side; - float z_far; - float z_near; + mediump float dual_paraboloid_side; + highp float z_far; + highp float z_near; bool ssao_enabled; - float ssao_light_affect; - float ssao_ao_affect; + mediump float ssao_light_affect; + mediump float ssao_ao_affect; bool roughness_limiter_enabled; - float roughness_limiter_amount; - float roughness_limiter_limit; - uvec2 roughness_limiter_pad; - - vec4 ao_color; + mediump float roughness_limiter_amount; + mediump float roughness_limiter_limit; + mediump float opaque_prepass_threshold; + uint roughness_limiter_pad; bool fog_enabled; - float fog_density; - float fog_height; - float fog_height_density; + highp float fog_density; + highp float fog_height; + highp float fog_height_density; - vec3 fog_light_color; - float fog_sun_scatter; + mediump vec3 fog_light_color; + mediump float fog_sun_scatter; - float fog_aerial_perspective; + mediump float fog_aerial_perspective; bool material_uv2_mode; - float time; - float reflection_multiplier; // one normally, zero when rendering reflections + highp float time; + mediump float reflection_multiplier; // one normally, zero when rendering reflections bool pancake_shadows; uint pad1; @@ -190,30 +194,30 @@ scene_data; #ifdef USE_RADIANCE_CUBEMAP_ARRAY -layout(set = 1, binding = 2) uniform textureCubeArray radiance_cubemap; +layout(set = 1, binding = 2) uniform mediump textureCubeArray radiance_cubemap; #else -layout(set = 1, binding = 2) uniform textureCube radiance_cubemap; +layout(set = 1, binding = 2) uniform mediump textureCube radiance_cubemap; #endif -layout(set = 1, binding = 3) uniform textureCubeArray reflection_atlas; +layout(set = 1, binding = 3) uniform mediump textureCubeArray reflection_atlas; -layout(set = 1, binding = 4) uniform texture2D shadow_atlas; +layout(set = 1, binding = 4) uniform highp texture2D shadow_atlas; -layout(set = 1, binding = 5) uniform texture2D directional_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 texture2D depth_buffer; -layout(set = 1, binding = 10) uniform texture2D color_buffer; +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 { - vec4 data[]; + highp vec4 data[]; } transforms; 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 7e06516d90..2328effe7b 100644 --- a/servers/rendering/renderer_rd/shaders/screen_space_reflection_scale.glsl +++ b/servers/rendering/renderer_rd/shaders/screen_space_reflection_scale.glsl @@ -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_probes.glsl b/servers/rendering/renderer_rd/shaders/sdfgi_debug_probes.glsl index 0eacbc5363..4290d5b869 100644 --- a/servers/rendering/renderer_rd/shaders/sdfgi_debug_probes.glsl +++ b/servers/rendering/renderer_rd/shaders/sdfgi_debug_probes.glsl @@ -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); diff --git a/servers/rendering/renderer_rd/shaders/sdfgi_direct_light.glsl b/servers/rendering/renderer_rd/shaders/sdfgi_direct_light.glsl index 99db35bb34..d6e5c6a92e 100644 --- a/servers/rendering/renderer_rd/shaders/sdfgi_direct_light.glsl +++ b/servers/rendering/renderer_rd/shaders/sdfgi_direct_light.glsl @@ -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 }; diff --git a/servers/rendering/renderer_rd/shaders/sdfgi_integrate.glsl b/servers/rendering/renderer_rd/shaders/sdfgi_integrate.glsl index bc376e9522..eedd28959c 100644 --- a/servers/rendering/renderer_rd/shaders/sdfgi_integrate.glsl +++ b/servers/rendering/renderer_rd/shaders/sdfgi_integrate.glsl @@ -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 aa4ded146f..4d9fa85a74 100644 --- a/servers/rendering/renderer_rd/shaders/sdfgi_preprocess.glsl +++ b/servers/rendering/renderer_rd/shaders/sdfgi_preprocess.glsl @@ -101,7 +101,7 @@ 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 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 neibhbours uint light; //rgbe8985 encoded total saved light, extra 2 bits for neighbours uint light_aniso; //55555 light anisotropy, extra 2 bits for neighbours @@ -134,7 +134,7 @@ 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 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 neibhbours uint light; //rgbe8985 encoded total saved light, extra 2 bits for neighbours uint light_aniso; //55555 light anisotropy, extra 2 bits for neighbours @@ -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/sky.glsl b/servers/rendering/renderer_rd/shaders/sky.glsl index 9924da37d5..d07a454ade 100644 --- a/servers/rendering/renderer_rd/shaders/sky.glsl +++ b/servers/rendering/renderer_rd/shaders/sky.glsl @@ -4,13 +4,21 @@ #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 { mat3 orientation; - vec4 proj; + vec4 projections[MAX_VIEWS]; vec4 position_multiplier; float time; + float luminance_multiplier; + float pad[2]; } params; @@ -26,15 +34,31 @@ void main() { #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 { 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; @@ -85,7 +109,6 @@ struct DirectionalLightData { layout(set = 0, binding = 3, std140) uniform DirectionalLights { DirectionalLightData data[MAX_DIRECTIONAL_LIGHT_DATA_STRUCTS]; } - directional_lights; #ifdef MATERIAL_UNIFORMS_USED @@ -154,8 +177,8 @@ 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); @@ -180,17 +203,17 @@ void main() { 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]), inverted_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]), inverted_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 @@ -227,4 +250,7 @@ void main() { 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/tonemap.glsl b/servers/rendering/renderer_rd/shaders/tonemap.glsl index 86b4da6b08..948c6e1e39 100644 --- a/servers/rendering/renderer_rd/shaders/tonemap.glsl +++ b/servers/rendering/renderer_rd/shaders/tonemap.glsl @@ -4,6 +4,12 @@ #VERSION_DEFINES +#ifdef MULTIVIEW +#ifdef has_VK_KHR_multiview +#extension GL_EXT_multiview : enable +#endif +#endif + layout(location = 0) out vec2 uv_interp; void main() { @@ -18,11 +24,28 @@ void main() { #VERSION_DEFINES +#ifdef MULTIVIEW +#ifdef has_VK_KHR_multiview +#extension GL_EXT_multiview : enable +#define ViewIndex gl_ViewIndex +#else // has_VK_KHR_multiview +#define ViewIndex 0 +#endif // has_VK_KHR_multiview +#endif //MULTIVIEW + layout(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; layout(set = 2, binding = 0) uniform sampler2D source_glow; + #ifdef USE_1D_LUT layout(set = 3, binding = 0) uniform sampler2D source_color_correction; #else @@ -48,7 +71,7 @@ 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; @@ -117,7 +140,7 @@ 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) @@ -146,25 +169,38 @@ 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); } @@ -181,15 +217,16 @@ 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); } } @@ -272,15 +309,23 @@ 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 +341,22 @@ 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 +365,9 @@ 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) { @@ -329,41 +380,54 @@ 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 + vec3 color = subpassLoad(input_color).rgb * params.luminance_multiplier; +#elif defined(MULTIVIEW) + vec3 color = textureLod(source_color, vec3(uv_interp, ViewIndex), 0.0f).rgb * params.luminance_multiplier; +#else + vec3 color = textureLod(source_color, uv_interp, 0.0f).rgb * params.luminance_multiplier; +#endif // 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; // Early Tonemap & SRGB Conversion - +#ifndef SUBPASS 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; color.rgb = mix(color.rgb, glow, params.glow_intensity); } if (params.use_fxaa) { color = do_fxaa(color, exposure, uv_interp); } +#endif + if (params.use_debanding) { // For best results, debanding should be done before tonemapping. // Otherwise, we're adding noise to an already-quantized image. color += screen_space_dither(gl_FragCoord.xy); } + color = apply_tonemapping(color, params.white); color = linear_to_srgb(color); // 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; // high dynamic range -> SRGB glow = apply_tonemapping(glow, params.white); @@ -371,6 +435,7 @@ void main() { color = apply_glow(color, glow); } +#endif // Additional effects diff --git a/servers/rendering/renderer_rd/shaders/volumetric_fog.glsl b/servers/rendering/renderer_rd/shaders/volumetric_fog.glsl index c793b6ebe1..181d3b272f 100644 --- a/servers/rendering/renderer_rd/shaders/volumetric_fog.glsl +++ b/servers/rendering/renderer_rd/shaders/volumetric_fog.glsl @@ -4,219 +4,103 @@ #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 -*/ - -#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) - 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, binding = 0, 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 @@ -240,464 +124,167 @@ 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) { + // 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 <= 1) { +#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..747f88960c --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/volumetric_fog_process.glsl @@ -0,0 +1,776 @@ +#[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; + vec3 bounds; + float dynamic_range; + + 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 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 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); + + 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 * 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 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 * 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 49a493cdc7..779f04ed35 100644 --- a/servers/rendering/renderer_rd/shaders/giprobe.glsl +++ b/servers/rendering/renderer_rd/shaders/voxel_gi.glsl @@ -71,11 +71,6 @@ 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 @@ -110,13 +105,6 @@ 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 @@ -170,13 +158,6 @@ 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 +355,7 @@ 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 +366,16 @@ 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 +385,8 @@ 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 +414,6 @@ 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 +425,18 @@ 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 +444,20 @@ 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 +465,7 @@ 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 +608,6 @@ 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 7d4d72967a..281c496df3 100644 --- a/servers/rendering/renderer_rd/shaders/giprobe_debug.glsl +++ b/servers/rendering/renderer_rd/shaders/voxel_gi_debug.glsl @@ -20,11 +20,6 @@ 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 { mat4 projection; uint cell_offset; diff --git a/servers/rendering/renderer_rd/shaders/giprobe_sdf.glsl b/servers/rendering/renderer_rd/shaders/voxel_gi_sdf.glsl index e20b3f680d..e20b3f680d 100644 --- a/servers/rendering/renderer_rd/shaders/giprobe_sdf.glsl +++ b/servers/rendering/renderer_rd/shaders/voxel_gi_sdf.glsl |