diff options
Diffstat (limited to 'servers/rendering/rasterizer_rd/shaders')
16 files changed, 2181 insertions, 526 deletions
diff --git a/servers/rendering/rasterizer_rd/shaders/SCsub b/servers/rendering/rasterizer_rd/shaders/SCsub index 6e852e2dc5..a454d144aa 100644 --- a/servers/rendering/rasterizer_rd/shaders/SCsub +++ b/servers/rendering/rasterizer_rd/shaders/SCsub @@ -5,14 +5,15 @@ Import("env") if "RD_GLSL" in env["BUILDERS"]: env.RD_GLSL("canvas.glsl") env.RD_GLSL("canvas_occlusion.glsl") - env.RD_GLSL("blur.glsl") + env.RD_GLSL("copy.glsl") + env.RD_GLSL("copy_to_fb.glsl") env.RD_GLSL("cubemap_roughness.glsl") env.RD_GLSL("cubemap_downsampler.glsl") env.RD_GLSL("cubemap_filter.glsl") env.RD_GLSL("scene_high_end.glsl") env.RD_GLSL("sky.glsl") env.RD_GLSL("tonemap.glsl") - env.RD_GLSL("copy.glsl") + env.RD_GLSL("cube_to_dp.glsl") env.RD_GLSL("giprobe.glsl") env.RD_GLSL("giprobe_debug.glsl") env.RD_GLSL("giprobe_sdf.glsl") @@ -22,3 +23,8 @@ if "RD_GLSL" in env["BUILDERS"]: env.RD_GLSL("ssao_minify.glsl") env.RD_GLSL("ssao_blur.glsl") env.RD_GLSL("roughness_limiter.glsl") + env.RD_GLSL("screen_space_reflection.glsl") + env.RD_GLSL("screen_space_reflection_filter.glsl") + env.RD_GLSL("screen_space_reflection_scale.glsl") + env.RD_GLSL("subsurface_scattering.glsl") + env.RD_GLSL("specular_merge.glsl") diff --git a/servers/rendering/rasterizer_rd/shaders/blur.glsl b/servers/rendering/rasterizer_rd/shaders/blur.glsl deleted file mode 100644 index 87c20ebaef..0000000000 --- a/servers/rendering/rasterizer_rd/shaders/blur.glsl +++ /dev/null @@ -1,294 +0,0 @@ -/* clang-format off */ -[vertex] - -#version 450 - -VERSION_DEFINES - -#include "blur_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]; - - if (bool(blur.flags & FLAG_USE_BLUR_SECTION)) { - uv_interp = blur.section.xy + uv_interp * blur.section.zw; - } - - gl_Position = vec4(uv_interp * 2.0 - 1.0, 0.0, 1.0); - - if (bool(blur.flags & FLAG_FLIP_Y)) { - uv_interp.y = 1.0 - uv_interp.y; - } -} - -/* clang-format off */ -[fragment] - -#version 450 - -VERSION_DEFINES - -#include "blur_inc.glsl" - -layout(location = 0) in vec2 uv_interp; -/* clang-format on */ - -layout(set = 0, binding = 0) uniform sampler2D source_color; - -#ifdef MODE_SSAO_MERGE -layout(set = 1, binding = 0) uniform sampler2D source_ssao; -#endif - -#ifdef GLOW_USE_AUTO_EXPOSURE -layout(set = 1, binding = 0) uniform sampler2D source_auto_exposure; -#endif - -layout(location = 0) out vec4 frag_color; - -//DOF -#if defined(MODE_DOF_FAR_BLUR) || defined(MODE_DOF_NEAR_BLUR) - -layout(set = 1, binding = 0) uniform sampler2D dof_source_depth; - -#ifdef DOF_NEAR_BLUR_MERGE -layout(set = 2, binding = 0) uniform sampler2D source_dof_original; -#endif - -#ifdef DOF_QUALITY_LOW -const int dof_kernel_size = 5; -const int dof_kernel_from = 2; -const float dof_kernel[5] = float[](0.153388, 0.221461, 0.250301, 0.221461, 0.153388); -#endif - -#ifdef DOF_QUALITY_MEDIUM -const int dof_kernel_size = 11; -const int dof_kernel_from = 5; -const float dof_kernel[11] = float[](0.055037, 0.072806, 0.090506, 0.105726, 0.116061, 0.119726, 0.116061, 0.105726, 0.090506, 0.072806, 0.055037); - -#endif - -#ifdef DOF_QUALITY_HIGH -const int dof_kernel_size = 21; -const int dof_kernel_from = 10; -const float dof_kernel[21] = float[](0.028174, 0.032676, 0.037311, 0.041944, 0.046421, 0.050582, 0.054261, 0.057307, 0.059587, 0.060998, 0.061476, 0.060998, 0.059587, 0.057307, 0.054261, 0.050582, 0.046421, 0.041944, 0.037311, 0.032676, 0.028174); -#endif - -#endif - -void main() { - -#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_DOF_FAR_BLUR - - vec4 color_accum = vec4(0.0); - - float depth = texture(dof_source_depth, uv_interp, 0.0).r; - depth = depth * 2.0 - 1.0; - - if (bool(blur.flags & FLAG_USE_ORTHOGONAL_PROJECTION)) { - depth = ((depth + (blur.camera_z_far + blur.camera_z_near) / (blur.camera_z_far - blur.camera_z_near)) * (blur.camera_z_far - blur.camera_z_near)) / 2.0; - } else { - depth = 2.0 * blur.camera_z_near * blur.camera_z_far / (blur.camera_z_far + blur.camera_z_near - depth * (blur.camera_z_far - blur.camera_z_near)); - } - - float amount = smoothstep(blur.dof_begin, blur.dof_end, depth); - float k_accum = 0.0; - - for (int i = 0; i < dof_kernel_size; i++) { - - int int_ofs = i - dof_kernel_from; - vec2 tap_uv = uv_interp + blur.dof_dir * float(int_ofs) * amount * blur.dof_radius; - - float tap_k = dof_kernel[i]; - - float tap_depth = texture(dof_source_depth, tap_uv, 0.0).r; - tap_depth = tap_depth * 2.0 - 1.0; - - if (bool(blur.flags & FLAG_USE_ORTHOGONAL_PROJECTION)) { - - tap_depth = ((tap_depth + (blur.camera_z_far + blur.camera_z_near) / (blur.camera_z_far - blur.camera_z_near)) * (blur.camera_z_far - blur.camera_z_near)) / 2.0; - } else { - tap_depth = 2.0 * blur.camera_z_near * blur.camera_z_far / (blur.camera_z_far + blur.camera_z_near - tap_depth * (blur.camera_z_far - blur.camera_z_near)); - } - - float tap_amount = mix(smoothstep(blur.dof_begin, blur.dof_end, tap_depth), 1.0, int_ofs == 0); - tap_amount *= tap_amount * tap_amount; //prevent undesired glow effect - - vec4 tap_color = texture(source_color, tap_uv, 0.0) * tap_k; - - k_accum += tap_k * tap_amount; - color_accum += tap_color * tap_amount; - } - - if (k_accum > 0.0) { - color_accum /= k_accum; - } - - frag_color = color_accum; ///k_accum; - -#endif - -#ifdef MODE_DOF_NEAR_BLUR - - vec4 color_accum = vec4(0.0); - - float max_accum = 0.0; - - for (int i = 0; i < dof_kernel_size; i++) { - - int int_ofs = i - dof_kernel_from; - vec2 tap_uv = uv_interp + blur.dof_dir * float(int_ofs) * blur.dof_radius; - float ofs_influence = max(0.0, 1.0 - float(abs(int_ofs)) / float(dof_kernel_from)); - - float tap_k = dof_kernel[i]; - - vec4 tap_color = texture(source_color, tap_uv, 0.0); - - float tap_depth = texture(dof_source_depth, tap_uv, 0.0).r; - tap_depth = tap_depth * 2.0 - 1.0; - if (bool(blur.flags & FLAG_USE_ORTHOGONAL_PROJECTION)) { - - tap_depth = ((tap_depth + (blur.camera_z_far + blur.camera_z_near) / (blur.camera_z_far - blur.camera_z_near)) * (blur.camera_z_far - blur.camera_z_near)) / 2.0; - } else { - tap_depth = 2.0 * blur.camera_z_near * blur.camera_z_far / (blur.camera_z_far + blur.camera_z_near - tap_depth * (blur.camera_z_far - blur.camera_z_near)); - } - float tap_amount = 1.0 - smoothstep(blur.dof_end, blur.dof_begin, tap_depth); - tap_amount *= tap_amount * tap_amount; //prevent undesired glow effect - - if (bool(blur.flags & FLAG_DOF_NEAR_FIRST_TAP)) { - tap_color.a = 1.0 - smoothstep(blur.dof_end, blur.dof_begin, tap_depth); - } - - max_accum = max(max_accum, tap_amount * ofs_influence); - - color_accum += tap_color * tap_k; - } - - color_accum.a = max(color_accum.a, sqrt(max_accum)); - -#ifdef DOF_NEAR_BLUR_MERGE - { - vec4 original = texture(source_dof_original, uv_interp, 0.0); - color_accum = mix(original, color_accum, color_accum.a); - } -#endif - - if (bool(blur.flags & FLAG_DOF_NEAR_FIRST_TAP)) { - frag_color = color_accum; - } -#endif - -#ifdef MODE_SIMPLE_COPY - vec4 color = texture(source_color, uv_interp, 0.0); - if (bool(blur.flags & FLAG_COPY_FORCE_LUMINANCE)) { - color.rgb = vec3(max(max(color.r, color.g), color.b)); - } - frag_color = color; -#endif - -#ifdef MODE_SSAO_MERGE - vec4 color = texture(source_color, uv_interp, 0.0); - float ssao = texture(source_ssao, uv_interp, 0.0).r; - frag_color = vec4(mix(color.rgb, color.rgb * mix(blur.ssao_color.rgb, vec3(1.0), ssao), color.a), 1.0); - -#endif -} diff --git a/servers/rendering/rasterizer_rd/shaders/blur_inc.glsl b/servers/rendering/rasterizer_rd/shaders/blur_inc.glsl deleted file mode 100644 index 33ba9de7bb..0000000000 --- a/servers/rendering/rasterizer_rd/shaders/blur_inc.glsl +++ /dev/null @@ -1,35 +0,0 @@ -#define FLAG_HORIZONTAL (1 << 0) -#define FLAG_USE_BLUR_SECTION (1 << 1) -#define FLAG_USE_ORTHOGONAL_PROJECTION (1 << 2) -#define FLAG_DOF_NEAR_FIRST_TAP (1 << 3) -#define FLAG_GLOW_FIRST_PASS (1 << 4) -#define FLAG_FLIP_Y (1 << 5) -#define FLAG_COPY_FORCE_LUMINANCE (1 << 6) - -layout(push_constant, binding = 1, std430) uniform Blur { - vec4 section; - 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; - // DOF. - float dof_begin; - float dof_end; - float dof_radius; - float dof_pad; - - vec2 dof_dir; - float camera_z_far; - float camera_z_near; - - vec4 ssao_color; -} -blur; diff --git a/servers/rendering/rasterizer_rd/shaders/canvas_uniforms_inc.glsl b/servers/rendering/rasterizer_rd/shaders/canvas_uniforms_inc.glsl index 1ac43480cd..a39866004b 100644 --- a/servers/rendering/rasterizer_rd/shaders/canvas_uniforms_inc.glsl +++ b/servers/rendering/rasterizer_rd/shaders/canvas_uniforms_inc.glsl @@ -132,6 +132,11 @@ layout(set = 2, binding = 6) uniform sampler shadow_sampler; #endif +layout(set = 2, binding = 7, std430) restrict readonly buffer GlobalVariableData { + vec4 data[]; +} +global_variables; + /* SET3: Render Target Data */ #ifdef SCREEN_TEXTURE_USED diff --git a/servers/rendering/rasterizer_rd/shaders/copy.glsl b/servers/rendering/rasterizer_rd/shaders/copy.glsl index cbb9b546a3..2d7661f65f 100644 --- a/servers/rendering/rasterizer_rd/shaders/copy.glsl +++ b/servers/rendering/rasterizer_rd/shaders/copy.glsl @@ -1,86 +1,220 @@ /* clang-format off */ -[vertex] +[compute] #version 450 VERSION_DEFINES -layout(location = 0) out vec2 uv_interp; +layout(local_size_x = 8, local_size_y = 8, local_size_z = 1) in; /* clang-format on */ -void main() { +#define FLAG_HORIZONTAL (1 << 0) +#define FLAG_USE_BLUR_SECTION (1 << 1) +#define FLAG_USE_ORTHOGONAL_PROJECTION (1 << 2) +#define FLAG_DOF_NEAR_FIRST_TAP (1 << 3) +#define FLAG_GLOW_FIRST_PASS (1 << 4) +#define FLAG_FLIP_Y (1 << 5) +#define FLAG_FORCE_LUMINANCE (1 << 6) +#define FLAG_COPY_ALL_SOURCE (1 << 7) + +layout(push_constant, binding = 1, std430) uniform Params { + ivec4 section; + ivec2 target; + 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; + // DOF. + float camera_z_far; + float camera_z_near; + uint pad2[2]; +} +params; - 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]; +layout(set = 0, binding = 0) uniform sampler2D source_color; - gl_Position = vec4(uv_interp * 2.0 - 1.0, 0.0, 1.0); -} +#ifdef GLOW_USE_AUTO_EXPOSURE +layout(set = 1, binding = 0) uniform sampler2D source_auto_exposure; +#endif -/* clang-format off */ -[fragment] +#if defined(MODE_LINEARIZE_DEPTH_COPY) || defined(MODE_SIMPLE_COPY_DEPTH) +layout(r32f, set = 3, binding = 0) uniform restrict writeonly image2D dest_buffer; +#elif defined(DST_IMAGE_8BIT) +layout(rgba8, set = 3, binding = 0) uniform restrict writeonly image2D dest_buffer; +#else +layout(rgba32f, set = 3, binding = 0) uniform restrict writeonly image2D dest_buffer; +#endif -#version 450 +void main() { -VERSION_DEFINES + // Pixel being shaded + ivec2 pos = ivec2(gl_GlobalInvocationID.xy); + if (any(greaterThan(pos, params.section.zw))) { //too large, do nothing + return; + } -layout(location = 0) in vec2 uv_interp; -/* clang-format on */ +#ifdef MODE_MIPMAP + + ivec2 base_pos = (pos + params.section.xy) << 1; + vec4 color = texelFetch(source_color, base_pos, 0); + color += texelFetch(source_color, base_pos + ivec2(0, 1), 0); + color += texelFetch(source_color, base_pos + ivec2(1, 0), 0); + color += texelFetch(source_color, base_pos + ivec2(1, 1), 0); + color /= 4.0; -#ifdef MODE_CUBE_TO_DP + imageStore(dest_buffer, pos + params.target, color); +#endif -layout(set = 0, binding = 0) uniform samplerCube source_cube; +#ifdef MODE_GAUSSIAN_BLUR -layout(push_constant, binding = 0, std430) uniform Params { - float bias; - float z_far; - float z_near; - bool z_flip; -} -params; + //Simpler blur uses SIGMA2 for the gaussian kernel for a stronger effect -layout(location = 0) out float depth_buffer; + if (bool(params.flags & FLAG_HORIZONTAL)) { + ivec2 base_pos = (pos + params.section.xy) << 1; + vec4 color = texelFetch(source_color, base_pos + ivec2(0, 0), 0) * 0.214607; + color += texelFetch(source_color, base_pos + ivec2(1, 0), 0) * 0.189879; + color += texelFetch(source_color, base_pos + ivec2(2, 0), 0) * 0.131514; + color += texelFetch(source_color, base_pos + ivec2(3, 0), 0) * 0.071303; + color += texelFetch(source_color, base_pos + ivec2(-1, 0), 0) * 0.189879; + color += texelFetch(source_color, base_pos + ivec2(-2, 0), 0) * 0.131514; + color += texelFetch(source_color, base_pos + ivec2(-3, 0), 0) * 0.071303; + imageStore(dest_buffer, pos + params.target, color); + } else { + + ivec2 base_pos = (pos + params.section.xy); + vec4 color = texelFetch(source_color, base_pos + ivec2(0, 0), 0) * 0.38774; + color += texelFetch(source_color, base_pos + ivec2(0, 1), 0) * 0.24477; + color += texelFetch(source_color, base_pos + ivec2(0, 2), 0) * 0.06136; + color += texelFetch(source_color, base_pos + ivec2(0, -1), 0) * 0.24477; + color += texelFetch(source_color, base_pos + ivec2(0, -2), 0) * 0.06136; + imageStore(dest_buffer, pos + params.target, color); + } #endif -void main() { +#ifdef MODE_GAUSSIAN_GLOW -#ifdef MODE_CUBE_TO_DP + //Glow uses larger sigma 1 for a more rounded blur effect - vec3 normal = vec3(uv_interp * 2.0 - 1.0, 0.0); +#define GLOW_ADD(m_ofs, m_mult) \ + { \ + ivec2 ofs = base_pos + m_ofs; \ + if (all(greaterThanEqual(ofs, section_begin)) && all(lessThan(ofs, section_end))) { \ + color += texelFetch(source_color, ofs, 0) * m_mult; \ + } \ + } + + vec4 color = vec4(0.0); + + if (bool(params.flags & FLAG_HORIZONTAL)) { + + ivec2 base_pos = (pos + params.section.xy) << 1; + ivec2 section_begin = params.section.xy << 1; + ivec2 section_end = section_begin + (params.section.zw << 1); + + GLOW_ADD(ivec2(0, 0), 0.174938); + GLOW_ADD(ivec2(1, 0), 0.165569); + GLOW_ADD(ivec2(2, 0), 0.140367); + GLOW_ADD(ivec2(3, 0), 0.106595); + GLOW_ADD(ivec2(-1, 0), 0.165569); + GLOW_ADD(ivec2(-2, 0), 0.140367); + GLOW_ADD(ivec2(-3, 0), 0.106595); + color *= params.glow_strength; + } else { - normal.z = 0.5 - 0.5 * ((normal.x * normal.x) + (normal.y * normal.y)); - normal = normalize(normal); + ivec2 base_pos = pos + params.section.xy; + ivec2 section_begin = params.section.xy; + ivec2 section_end = section_begin + params.section.zw; - normal.y = -normal.y; //needs to be flipped to match projection matrix - if (!params.z_flip) { - normal.z = -normal.z; + GLOW_ADD(ivec2(0, 0), 0.288713); + GLOW_ADD(ivec2(0, 1), 0.233062); + GLOW_ADD(ivec2(0, 2), 0.122581); + GLOW_ADD(ivec2(0, -1), 0.233062); + GLOW_ADD(ivec2(0, -2), 0.122581); + color *= params.glow_strength; } - float depth = texture(source_cube, normal).r; +#undef GLOW_ADD - // absolute values for direction cosines, bigger value equals closer to basis axis - vec3 unorm = abs(normal); + if (bool(params.flags & FLAG_GLOW_FIRST_PASS)) { +#ifdef GLOW_USE_AUTO_EXPOSURE + + color /= texelFetch(source_auto_exposure, ivec2(0, 0), 0).r / params.glow_auto_exposure_grey; +#endif + color *= params.glow_exposure; + + float luminance = max(color.r, max(color.g, color.b)); + float feedback = max(smoothstep(params.glow_hdr_threshold, params.glow_hdr_threshold + params.glow_hdr_scale, luminance), params.glow_bloom); + + color = min(color * feedback, vec4(params.glow_luminance_cap)); + } + + imageStore(dest_buffer, pos + params.target, color); + +#endif + +#ifdef MODE_SIMPLE_COPY + + vec4 color; + if (bool(params.flags & FLAG_COPY_ALL_SOURCE)) { + vec2 uv = vec2(pos) / vec2(params.section.zw); + if (bool(params.flags & FLAG_FLIP_Y)) { + uv.y = 1.0 - uv.y; + } + color = textureLod(source_color, uv, 0.0); + + if (bool(params.flags & FLAG_FORCE_LUMINANCE)) { + color.rgb = vec3(max(max(color.r, color.g), color.b)); + } + imageStore(dest_buffer, pos + params.target, color); - if ((unorm.x >= unorm.y) && (unorm.x >= unorm.z)) { - // x code - unorm = normal.x > 0.0 ? vec3(1.0, 0.0, 0.0) : vec3(-1.0, 0.0, 0.0); - } else if ((unorm.y > unorm.x) && (unorm.y >= unorm.z)) { - // y code - unorm = normal.y > 0.0 ? vec3(0.0, 1.0, 0.0) : vec3(0.0, -1.0, 0.0); - } else if ((unorm.z > unorm.x) && (unorm.z > unorm.y)) { - // z code - unorm = normal.z > 0.0 ? vec3(0.0, 0.0, 1.0) : vec3(0.0, 0.0, -1.0); } else { - // oh-no we messed up code - // has to be - unorm = vec3(1.0, 0.0, 0.0); + color = texelFetch(source_color, pos + params.section.xy, 0); + + if (bool(params.flags & FLAG_FORCE_LUMINANCE)) { + color.rgb = vec3(max(max(color.r, color.g), color.b)); + } + + if (bool(params.flags & FLAG_FLIP_Y)) { + pos.y = params.section.w - pos.y - 1; + } + + imageStore(dest_buffer, pos + params.target, color); + } + +#endif + +#ifdef MODE_SIMPLE_COPY_DEPTH + + vec4 color = texelFetch(source_color, pos + params.section.xy, 0); + + if (bool(params.flags & FLAG_FLIP_Y)) { + pos.y = params.section.w - pos.y - 1; } - float depth_fix = 1.0 / dot(normal, unorm); + imageStore(dest_buffer, pos + params.target, vec4(color.r)); + +#endif + +#ifdef MODE_LINEARIZE_DEPTH_COPY + + float depth = texelFetch(source_color, pos + params.section.xy, 0).r; + depth = depth * 2.0 - 1.0; + depth = 2.0 * params.camera_z_near * params.camera_z_far / (params.camera_z_far + params.camera_z_near - depth * (params.camera_z_far - params.camera_z_near)); + vec4 color = vec4(depth / params.camera_z_far); - depth = 2.0 * depth - 1.0; - float linear_depth = 2.0 * params.z_near * params.z_far / (params.z_far + params.z_near - depth * (params.z_far - params.z_near)); - depth_buffer = (linear_depth * depth_fix + params.bias) / params.z_far; + if (bool(params.flags & FLAG_FLIP_Y)) { + pos.y = params.section.w - pos.y - 1; + } + imageStore(dest_buffer, pos + params.target, color); #endif } diff --git a/servers/rendering/rasterizer_rd/shaders/copy_to_fb.glsl b/servers/rendering/rasterizer_rd/shaders/copy_to_fb.glsl new file mode 100644 index 0000000000..07f8d09743 --- /dev/null +++ b/servers/rendering/rasterizer_rd/shaders/copy_to_fb.glsl @@ -0,0 +1,104 @@ +/* clang-format off */ +[vertex] + +#version 450 + +VERSION_DEFINES + +layout(location = 0) out vec2 uv_interp; +/* clang-format on */ + +layout(push_constant, binding = 1, std430) uniform Params { + vec4 section; + vec2 pixel_size; + bool flip_y; + bool use_section; + + bool force_luminance; + uint pad[3]; +} +params; + +void main() { + + vec2 base_arr[4] = vec2[](vec2(0.0, 0.0), vec2(0.0, 1.0), vec2(1.0, 1.0), vec2(1.0, 0.0)); + uv_interp = base_arr[gl_VertexIndex]; + + vec2 vpos = uv_interp; + if (params.use_section) { + vpos = params.section.xy + vpos * params.section.zw; + } + + gl_Position = vec4(vpos * 2.0 - 1.0, 0.0, 1.0); + + if (params.flip_y) { + uv_interp.y = 1.0 - uv_interp.y; + } +} + +/* clang-format off */ +[fragment] + +#version 450 + +VERSION_DEFINES + +layout(push_constant, binding = 1, std430) uniform Params { + vec4 section; + vec2 pixel_size; + bool flip_y; + bool use_section; + + bool force_luminance; + bool alpha_to_zero; + uint pad[2]; +} params; + + +layout(location = 0) in vec2 uv_interp; +/* clang-format on */ + +layout(set = 0, binding = 0) uniform sampler2D source_color; + +layout(location = 0) out vec4 frag_color; + +void main() { + + vec2 uv = uv_interp; + +#ifdef MODE_PANORAMA_TO_DP + + //obtain normal from dual paraboloid uv +#define M_PI 3.14159265359 + + float side; + uv.y = modf(uv.y * 2.0, side); + side = side * 2.0 - 1.0; + vec3 normal = vec3(uv * 2.0 - 1.0, 0.0); + normal.z = 0.5 - 0.5 * ((normal.x * normal.x) + (normal.y * normal.y)); + normal *= -side; + normal = normalize(normal); + + //now convert normal to panorama uv + + vec2 st = vec2(atan(normal.x, normal.z), acos(normal.y)); + + if (st.x < 0.0) + st.x += M_PI * 2.0; + + uv = st / vec2(M_PI * 2.0, M_PI); + + if (side < 0.0) { + //uv.y = 1.0 - uv.y; + uv = 1.0 - uv; + } +#endif + vec4 color = textureLod(source_color, uv, 0.0); + if (params.force_luminance) { + color.rgb = vec3(max(max(color.r, color.g), color.b)); + } + if (params.alpha_to_zero) { + color.rgb *= color.a; + } + frag_color = color; +} diff --git a/servers/rendering/rasterizer_rd/shaders/cube_to_dp.glsl b/servers/rendering/rasterizer_rd/shaders/cube_to_dp.glsl new file mode 100644 index 0000000000..02ebe1a53b --- /dev/null +++ b/servers/rendering/rasterizer_rd/shaders/cube_to_dp.glsl @@ -0,0 +1,72 @@ +/* clang-format off */ +[compute] + +#version 450 + +VERSION_DEFINES + +layout(local_size_x = 8, local_size_y = 8, local_size_z = 1) in; +/* clang-format on */ + +layout(set = 0, binding = 0) uniform samplerCube source_cube; + +layout(push_constant, binding = 1, std430) uniform Params { + ivec2 screen_size; + ivec2 offset; + float bias; + float z_far; + float z_near; + bool z_flip; +} +params; + +layout(r32f, set = 1, binding = 0) uniform restrict writeonly image2D depth_buffer; + +void main() { + + ivec2 pos = ivec2(gl_GlobalInvocationID.xy); + if (any(greaterThan(pos, params.screen_size))) { //too large, do nothing + return; + } + + vec2 pixel_size = 1.0 / vec2(params.screen_size); + vec2 uv = (vec2(pos) + 0.5) * pixel_size; + + vec3 normal = vec3(uv * 2.0 - 1.0, 0.0); + + normal.z = 0.5 - 0.5 * ((normal.x * normal.x) + (normal.y * normal.y)); + normal = normalize(normal); + + normal.y = -normal.y; //needs to be flipped to match projection matrix + if (!params.z_flip) { + normal.z = -normal.z; + } + + float depth = texture(source_cube, normal).r; + + // absolute values for direction cosines, bigger value equals closer to basis axis + vec3 unorm = abs(normal); + + if ((unorm.x >= unorm.y) && (unorm.x >= unorm.z)) { + // x code + unorm = normal.x > 0.0 ? vec3(1.0, 0.0, 0.0) : vec3(-1.0, 0.0, 0.0); + } else if ((unorm.y > unorm.x) && (unorm.y >= unorm.z)) { + // y code + unorm = normal.y > 0.0 ? vec3(0.0, 1.0, 0.0) : vec3(0.0, -1.0, 0.0); + } else if ((unorm.z > unorm.x) && (unorm.z > unorm.y)) { + // z code + unorm = normal.z > 0.0 ? vec3(0.0, 0.0, 1.0) : vec3(0.0, 0.0, -1.0); + } else { + // oh-no we messed up code + // has to be + unorm = vec3(1.0, 0.0, 0.0); + } + + float depth_fix = 1.0 / dot(normal, unorm); + + depth = 2.0 * depth - 1.0; + float linear_depth = 2.0 * params.z_near * params.z_far / (params.z_far + params.z_near - depth * (params.z_far - params.z_near)); + depth = (linear_depth * depth_fix) / params.z_far; + + imageStore(depth_buffer, pos + params.offset, vec4(depth)); +} diff --git a/servers/rendering/rasterizer_rd/shaders/scene_high_end.glsl b/servers/rendering/rasterizer_rd/shaders/scene_high_end.glsl index 07f4770b14..ec47887036 100644 --- a/servers/rendering/rasterizer_rd/shaders/scene_high_end.glsl +++ b/servers/rendering/rasterizer_rd/shaders/scene_high_end.glsl @@ -20,9 +20,7 @@ layout(location = 2) in vec4 tangent_attrib; layout(location = 3) in vec4 color_attrib; #endif -#if defined(UV_USED) layout(location = 4) in vec2 uv_attrib; -#endif #if defined(UV2_USED) || defined(USE_LIGHTMAP) layout(location = 5) in vec2 uv2_attrib; @@ -39,9 +37,7 @@ layout(location = 1) out vec3 normal_interp; layout(location = 2) out vec4 color_interp; #endif -#if defined(UV_USED) layout(location = 3) out vec2 uv_interp; -#endif #if defined(UV2_USED) || defined(USE_LIGHTMAP) layout(location = 4) out vec2 uv2_interp; @@ -157,9 +153,7 @@ void main() { #endif } -#if defined(UV_USED) uv_interp = uv_attrib; -#endif #if defined(UV2_USED) || defined(USE_LIGHTMAP) uv2_interp = uv2_attrib; @@ -244,19 +238,13 @@ VERTEX_SHADER_CODE //for dual paraboloid shadow mapping, this is the fastest but least correct way, as it curves straight edges - vec3 vtx = vertex_interp + normalize(vertex_interp) * scene_data.z_offset; + vec3 vtx = vertex_interp; float distance = length(vtx); vtx = normalize(vtx); vtx.xy /= 1.0 - vtx.z; vtx.z = (distance / scene_data.z_far); vtx.z = vtx.z * 2.0 - 1.0; - vertex_interp = vtx; -#else - - float z_ofs = scene_data.z_offset; - z_ofs += max(0.0, 1.0 - abs(normalize(normal_interp).z)) * scene_data.z_slope_scale; - vertex_interp.z -= z_ofs; #endif @@ -267,6 +255,14 @@ VERTEX_SHADER_CODE #else gl_Position = projection_matrix * vec4(vertex_interp, 1.0); #endif + +#ifdef MODE_RENDER_DEPTH + if (scene_data.pancake_shadows) { + if (gl_Position.z <= 0.00001) { + gl_Position.z = 0.00001; + } + } +#endif } /* clang-format off */ @@ -288,9 +284,7 @@ layout(location = 1) in vec3 normal_interp; layout(location = 2) in vec4 color_interp; #endif -#if defined(UV_USED) layout(location = 3) in vec2 uv_interp; -#endif #if defined(UV2_USED) || defined(USE_LIGHTMAP) layout(location = 4) in vec2 uv2_interp; @@ -315,6 +309,11 @@ layout(location = 8) in float dp_clip; #define world_normal_matrix instances.data[instance_index].normal_transform #define projection_matrix scene_data.projection_matrix +#if defined(ENABLE_SSS) && defined(ENABLE_TRANSMITTANCE) +//both required for transmittance to be enabled +#define LIGHT_TRANSMITTANCE_USED +#endif + #ifdef USE_MATERIAL_UNIFORMS layout(set = 5, binding = 0, std140) uniform MaterialUniforms{ /* clang-format off */ @@ -434,9 +433,16 @@ 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, vec3 attenuation, vec3 diffuse_color, float roughness, float metallic, float specular, float specular_blob_intensity, -#ifdef LIGHT_TRANSMISSION_USED - vec3 transmission, +void light_compute(vec3 N, vec3 L, vec3 V, float A, vec3 light_color, float attenuation, vec3 shadow_attenuation, vec3 diffuse_color, float roughness, float metallic, float specular, float specular_blob_intensity, +#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, @@ -467,7 +473,7 @@ LIGHT_SHADER_CODE /* clang-format on */ #else - float NdotL = dot(N, L); + float NdotL = min(A + dot(N, L), 1.0); float cNdotL = max(NdotL, 0.0); // clamped NdotL float NdotV = dot(N, V); float cNdotV = max(NdotV, 0.0); @@ -477,11 +483,11 @@ LIGHT_SHADER_CODE #endif #if defined(SPECULAR_BLINN) || defined(SPECULAR_SCHLICK_GGX) || defined(LIGHT_CLEARCOAT_USED) - float cNdotH = max(dot(N, H), 0.0); + float cNdotH = clamp(A + dot(N, H), 0.0, 1.0); #endif #if defined(DIFFUSE_BURLEY) || defined(SPECULAR_SCHLICK_GGX) || defined(LIGHT_CLEARCOAT_USED) - float cLdotH = max(dot(L, H), 0.0); + float cLdotH = clamp(A + dot(L, H), 0.0, 1.0); #endif if (metallic < 1.0) { @@ -538,16 +544,48 @@ LIGHT_SHADER_CODE diffuse_brdf_NL = cNdotL * (1.0 / M_PI); #endif - diffuse_light += light_color * diffuse_color * diffuse_brdf_NL * attenuation; + diffuse_light += light_color * diffuse_color * shadow_attenuation * diffuse_brdf_NL * attenuation; -#if defined(LIGHT_TRANSMISSION_USED) - diffuse_light += light_color * diffuse_color * (vec3(1.0 / M_PI) - diffuse_brdf_NL) * transmission * attenuation; +#if defined(LIGHT_BACKLIGHT_USED) + diffuse_light += light_color * diffuse_color * (vec3(1.0 / M_PI) - diffuse_brdf_NL) * backlight * attenuation; #endif #if defined(LIGHT_RIM_USED) float rim_light = pow(max(0.0, 1.0 - cNdotV), max(0.0, (1.0 - roughness) * 16.0)); diffuse_light += rim_light * rim * mix(vec3(1.0), diffuse_color, rim_tint) * light_color; #endif + +#ifdef LIGHT_TRANSMITTANCE_USED + +#ifdef SSS_MODE_SKIN + + { + float scale = 8.25 / transmittance_depth; + float d = scale * abs(transmittance_z); + float dd = -d * d; + vec3 profile = vec3(0.233, 0.455, 0.649) * exp(dd / 0.0064) + + vec3(0.1, 0.336, 0.344) * exp(dd / 0.0484) + + vec3(0.118, 0.198, 0.0) * exp(dd / 0.187) + + vec3(0.113, 0.007, 0.007) * exp(dd / 0.567) + + vec3(0.358, 0.004, 0.0) * exp(dd / 1.99) + + vec3(0.078, 0.0, 0.0) * exp(dd / 7.41); + + diffuse_light += profile * transmittance_color.a * diffuse_color * light_color * clamp(transmittance_boost - NdotL, 0.0, 1.0) * (1.0 / M_PI) * attenuation; + } +#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 += diffuse_color * transmittance_color.rgb * light_color * (1.0 / M_PI) * transmittance_color.a * fade * attenuation; + } + +#endif //SSS_MODE_SKIN + +#endif //LIGHT_TRANSMITTANCE_USED } if (roughness > 0.0) { // FIXME: roughness == 0 should not disable specular light entirely @@ -562,18 +600,18 @@ LIGHT_SHADER_CODE blinn *= (shininess + 8.0) * (1.0 / (8.0 * M_PI)); float intensity = blinn; - specular_light += light_color * intensity * specular_blob_intensity * attenuation; + specular_light += light_color * shadow_attenuation * intensity * specular_blob_intensity * attenuation; #elif defined(SPECULAR_PHONG) vec3 R = normalize(-reflect(L, N)); - float cRdotV = max(0.0, dot(R, V)); + float cRdotV = clamp(A + dot(R, V), 0.0, 1.0); float shininess = exp2(15.0 * (1.0 - roughness) + 1.0) * 0.25; float phong = pow(cRdotV, shininess); phong *= (shininess + 8.0) * (1.0 / (8.0 * M_PI)); float intensity = (phong) / max(4.0 * cNdotV * cNdotL, 0.75); - specular_light += light_color * intensity * specular_blob_intensity * attenuation; + specular_light += light_color * shadow_attenuation * intensity * specular_blob_intensity * attenuation; #elif defined(SPECULAR_TOON) @@ -582,7 +620,7 @@ LIGHT_SHADER_CODE float mid = 1.0 - roughness; mid *= mid; float intensity = smoothstep(mid - roughness * 0.5, mid + roughness * 0.5, RdotV) * mid; - diffuse_light += light_color * intensity * specular_blob_intensity * attenuation; // write to diffuse_light, as in toon shading you generally want no reflection + diffuse_light += light_color * shadow_attenuation * intensity * specular_blob_intensity * attenuation; // write to diffuse_light, as in toon shading you generally want no reflection #elif defined(SPECULAR_DISABLED) // none.. @@ -613,7 +651,7 @@ LIGHT_SHADER_CODE vec3 specular_brdf_NL = cNdotL * D * F * G; - specular_light += specular_brdf_NL * light_color * specular_blob_intensity * attenuation; + specular_light += specular_brdf_NL * light_color * shadow_attenuation * specular_blob_intensity * attenuation; #endif #if defined(LIGHT_CLEARCOAT_USED) @@ -627,12 +665,12 @@ LIGHT_SHADER_CODE float clearcoat_specular_brdf_NL = 0.25 * clearcoat * Gr * Fr * Dr * cNdotL; - specular_light += clearcoat_specular_brdf_NL * light_color * specular_blob_intensity * attenuation; + specular_light += clearcoat_specular_brdf_NL * light_color * shadow_attenuation * specular_blob_intensity * attenuation; #endif } #ifdef USE_SHADOW_TO_OPACITY - alpha = min(alpha, clamp(1.0 - length(attenuation), 0.0, 1.0)); + alpha = min(alpha, clamp(1.0 - length(shadow_attenuation * attenuation), 0.0, 1.0)); #endif #endif //defined(USE_LIGHT_SHADER_CODE) @@ -640,53 +678,121 @@ LIGHT_SHADER_CODE #ifndef USE_NO_SHADOWS -float sample_shadow(texture2D shadow, vec2 shadow_pixel_size, vec4 coord) { +// Produces cheap but low-quality white noise, nothing special +float quick_hash(vec2 pos) { + return fract(sin(dot(pos * 19.19, vec2(49.5791, 97.413))) * 49831.189237); +} + +float sample_directional_pcf_shadow(texture2D shadow, vec2 shadow_pixel_size, vec4 coord) { - //todo optimize vec2 pos = coord.xy; float depth = coord.z; -#ifdef SHADOW_MODE_PCF_13 + //if only one sample is taken, take it from the center + if (scene_data.directional_soft_shadow_samples == 1) { + return textureProj(sampler2DShadow(shadow, shadow_sampler), vec4(pos, depth, 1.0)); + } - float avg = textureProj(shadow, vec4(pos, depth, 1.0)); - avg += textureProj(sampler2DShadow(shadow, shadow_sampler), vec4(pos + vec2(shadow_pixel_size.x, 0.0), depth, 1.0)); - avg += textureProj(sampler2DShadow(shadow, shadow_sampler), vec4(pos + vec2(-shadow_pixel_size.x, 0.0), depth, 1.0)); - avg += textureProj(sampler2DShadow(shadow, shadow_sampler), vec4(pos + vec2(0.0, shadow_pixel_size.y), depth, 1.0)); - avg += textureProj(sampler2DShadow(shadow, shadow_sampler), vec4(pos + vec2(0.0, -shadow_pixel_size.y), depth, 1.0)); - avg += textureProj(sampler2DShadow(shadow, shadow_sampler), vec4(pos + vec2(shadow_pixel_size.x, shadow_pixel_size.y), depth, 1.0)); - avg += textureProj(sampler2DShadow(shadow, shadow_sampler), vec4(pos + vec2(-shadow_pixel_size.x, shadow_pixel_size.y), depth, 1.0)); - avg += textureProj(sampler2DShadow(shadow, shadow_sampler), vec4(pos + vec2(shadow_pixel_size.x, -shadow_pixel_size.y), depth, 1.0)); - avg += textureProj(sampler2DShadow(shadow, shadow_sampler), vec4(pos + vec2(-shadow_pixel_size.x, -shadow_pixel_size.y), depth, 1.0)); - avg += textureProj(sampler2DShadow(shadow, shadow_sampler), vec4(pos + vec2(shadow_pixel_size.x * 2.0, 0.0), depth, 1.0)); - avg += textureProj(sampler2DShadow(shadow, shadow_sampler), vec4(pos + vec2(-shadow_pixel_size.x * 2.0, 0.0), depth, 1.0)); - avg += textureProj(sampler2DShadow(shadow, shadow_sampler), vec4(pos + vec2(0.0, shadow_pixel_size.y * 2.0), depth, 1.0)); - avg += textureProj(sampler2DShadow(shadow, shadow_sampler), vec4(pos + vec2(0.0, -shadow_pixel_size.y * 2.0), depth, 1.0)); - return avg * (1.0 / 13.0); -#endif + 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)); + } -#ifdef SHADOW_MODE_PCF_5 + float avg = 0.0; - float avg = textureProj(sampler2DShadow(shadow, shadow_sampler), vec4(pos, depth, 1.0)); - avg += textureProj(sampler2DShadow(shadow, shadow_sampler), vec4(pos + vec2(shadow_pixel_size.x, 0.0), depth, 1.0)); - avg += textureProj(sampler2DShadow(shadow, shadow_sampler), vec4(pos + vec2(-shadow_pixel_size.x, 0.0), depth, 1.0)); - avg += textureProj(sampler2DShadow(shadow, shadow_sampler), vec4(pos + vec2(0.0, shadow_pixel_size.y), depth, 1.0)); - avg += textureProj(sampler2DShadow(shadow, shadow_sampler), vec4(pos + vec2(0.0, -shadow_pixel_size.y), depth, 1.0)); - return avg * (1.0 / 5.0); + for (uint i = 0; i < scene_data.directional_soft_shadow_samples; i++) { + avg += textureProj(sampler2DShadow(shadow, shadow_sampler), vec4(pos + shadow_pixel_size * (disk_rotation * scene_data.directional_soft_shadow_kernel[i].xy), depth, 1.0)); + } -#endif + return avg * (1.0 / float(scene_data.directional_soft_shadow_samples)); +} -#if !defined(SHADOW_MODE_PCF_5) || !defined(SHADOW_MODE_PCF_13) +float sample_pcf_shadow(texture2D shadow, vec2 shadow_pixel_size, vec4 coord) { - return textureProj(sampler2DShadow(shadow, shadow_sampler), vec4(pos, depth, 1.0)); + vec2 pos = coord.xy; + float depth = coord.z; -#endif + //if only one sample is taken, take it from the center + if (scene_data.soft_shadow_samples == 1) { + return textureProj(sampler2DShadow(shadow, shadow_sampler), vec4(pos, depth, 1.0)); + } + + mat2 disk_rotation; + { + float r = quick_hash(gl_FragCoord.xy) * 2.0 * M_PI; + float sr = sin(r); + float cr = cos(r); + disk_rotation = mat2(vec2(cr, -sr), vec2(sr, cr)); + } + + float avg = 0.0; + + for (uint i = 0; i < scene_data.soft_shadow_samples; i++) { + avg += textureProj(sampler2DShadow(shadow, shadow_sampler), vec4(pos + shadow_pixel_size * (disk_rotation * scene_data.soft_shadow_kernel[i].xy), depth, 1.0)); + } + + return avg * (1.0 / float(scene_data.soft_shadow_samples)); +} + +float sample_directional_soft_shadow(texture2D shadow, vec3 pssm_coord, vec2 tex_scale) { + + //find blocker + float blocker_count = 0.0; + float blocker_average = 0.0; + + mat2 disk_rotation; + { + float r = quick_hash(gl_FragCoord.xy) * 2.0 * M_PI; + float sr = sin(r); + float cr = cos(r); + disk_rotation = mat2(vec2(cr, -sr), vec2(sr, cr)); + } + + for (uint i = 0; i < scene_data.directional_penumbra_shadow_samples; i++) { + + vec2 suv = pssm_coord.xy + (disk_rotation * scene_data.directional_penumbra_shadow_kernel[i].xy) * tex_scale; + float d = textureLod(sampler2D(shadow, material_samplers[SAMPLER_LINEAR_CLAMP]), suv, 0.0).r; + if (d < pssm_coord.z) { + blocker_average += d; + blocker_count += 1.0; + } + } + + if (blocker_count > 0.0) { + + //blockers found, do soft shadow + blocker_average /= blocker_count; + float penumbra = (pssm_coord.z - blocker_average) / blocker_average; + tex_scale *= penumbra; + + float s = 0.0; + for (uint i = 0; i < scene_data.directional_penumbra_shadow_samples; i++) { + vec2 suv = pssm_coord.xy + (disk_rotation * scene_data.directional_penumbra_shadow_kernel[i].xy) * tex_scale; + s += textureProj(sampler2DShadow(shadow, shadow_sampler), vec4(suv, pssm_coord.z, 1.0)); + } + + return s / float(scene_data.directional_penumbra_shadow_samples); + + } else { + //no blockers found, so no shadow + return 1.0; + } } #endif //USE_NO_SHADOWS -void light_process_omni(uint idx, vec3 vertex, vec3 eye_vec, vec3 normal, vec3 albedo, float roughness, float metallic, float specular, float p_blob_intensity, -#ifdef LIGHT_TRANSMISSION_USED - vec3 transmission, +void light_process_omni(uint idx, vec3 vertex, vec3 eye_vec, vec3 normal, vec3 vertex_ddx, vec3 vertex_ddy, vec3 albedo, float roughness, float metallic, float specular, float p_blob_intensity, +#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, @@ -707,45 +813,269 @@ void light_process_omni(uint idx, vec3 vertex, vec3 eye_vec, vec3 normal, vec3 a float normalized_distance = light_length * lights.data[idx].inv_radius; vec2 attenuation_energy = unpackHalf2x16(lights.data[idx].attenuation_energy); float omni_attenuation = pow(max(1.0 - normalized_distance, 0.0), attenuation_energy.x); - vec3 light_attenuation = vec3(omni_attenuation); + float light_attenuation = omni_attenuation; + vec3 shadow_attenuation = vec3(1.0); vec4 color_specular = unpackUnorm4x8(lights.data[idx].color_specular); color_specular.rgb *= attenuation_energy.y; + float size_A = 0.0; + + if (lights.data[idx].size > 0.0) { + + float t = lights.data[idx].size / max(0.001, light_length); + size_A = max(0.0, 1.0 - 1 / sqrt(1 + t * t)); + } + +#ifdef LIGHT_TRANSMITTANCE_USED + float transmittance_z = transmittance_depth; //no transmittance by default +#endif #ifndef USE_NO_SHADOWS vec4 shadow_color_enabled = unpackUnorm4x8(lights.data[idx].shadow_color_enabled); if (shadow_color_enabled.w > 0.5) { // there is a shadowmap - vec4 splane = (lights.data[idx].shadow_matrix * vec4(vertex, 1.0)); - float shadow_len = length(splane); - splane = normalize(splane); - vec4 clamp_rect = lights.data[idx].atlas_rect; + vec4 v = vec4(vertex, 1.0); + + vec4 splane = (lights.data[idx].shadow_matrix * v); + float shadow_len = length(splane.xyz); //need to remember shadow len from here + + { + vec3 nofs = normal_interp * lights.data[idx].shadow_normal_bias / lights.data[idx].inv_radius; + nofs *= (1.0 - max(0.0, dot(normalize(light_rel_vec), normalize(normal_interp)))); + v.xyz += nofs; + splane = (lights.data[idx].shadow_matrix * v); + } + + float shadow; + + if (lights.data[idx].soft_shadow_size > 0.0) { + //soft shadow + + //find blocker - if (splane.z >= 0.0) { + float blocker_count = 0.0; + float blocker_average = 0.0; - splane.z += 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)); + } + + vec3 normal = normalize(splane.xyz); + vec3 v0 = abs(normal.z) < 0.999 ? vec3(0.0, 0.0, 1.0) : vec3(0.0, 1.0, 0.0); + vec3 tangent = normalize(cross(v0, normal)); + vec3 bitangent = normalize(cross(tangent, normal)); + float z_norm = shadow_len * lights.data[idx].inv_radius; + + tangent *= lights.data[idx].soft_shadow_size * lights.data[idx].soft_shadow_scale; + bitangent *= lights.data[idx].soft_shadow_size * lights.data[idx].soft_shadow_scale; + + for (uint i = 0; i < scene_data.penumbra_shadow_samples; i++) { + + vec2 disk = disk_rotation * scene_data.penumbra_shadow_kernel[i].xy; + + vec3 pos = splane.xyz + tangent * disk.x + bitangent * disk.y; + + pos = normalize(pos); + vec4 uv_rect = lights.data[idx].atlas_rect; + + if (pos.z >= 0.0) { + + pos.z += 1.0; + uv_rect.y += uv_rect.w; + } else { + + pos.z = 1.0 - pos.z; + } + + pos.xy /= pos.z; + + pos.xy = pos.xy * 0.5 + 0.5; + pos.xy = uv_rect.xy + pos.xy * uv_rect.zw; + + float d = textureLod(sampler2D(shadow_atlas, material_samplers[SAMPLER_LINEAR_CLAMP]), pos.xy, 0.0).r; + if (d < z_norm) { + blocker_average += d; + blocker_count += 1.0; + } + } - clamp_rect.y += clamp_rect.w; + if (blocker_count > 0.0) { + //blockers found, do soft shadow + blocker_average /= blocker_count; + float penumbra = (z_norm - blocker_average) / blocker_average; + tangent *= penumbra; + bitangent *= penumbra; + + z_norm -= lights.data[idx].inv_radius * lights.data[idx].shadow_bias; + + shadow = 0.0; + for (uint i = 0; i < scene_data.penumbra_shadow_samples; i++) { + + vec2 disk = disk_rotation * scene_data.penumbra_shadow_kernel[i].xy; + vec3 pos = splane.xyz + tangent * disk.x + bitangent * disk.y; + + pos = normalize(pos); + vec4 uv_rect = lights.data[idx].atlas_rect; + + if (pos.z >= 0.0) { + + pos.z += 1.0; + uv_rect.y += uv_rect.w; + } else { + + pos.z = 1.0 - pos.z; + } + + pos.xy /= pos.z; + + pos.xy = pos.xy * 0.5 + 0.5; + pos.xy = uv_rect.xy + pos.xy * uv_rect.zw; + shadow += textureProj(sampler2DShadow(shadow_atlas, shadow_sampler), vec4(pos.xy, z_norm, 1.0)); + } + + shadow /= float(scene_data.penumbra_shadow_samples); + + } else { + //no blockers found, so no shadow + shadow = 1.0; + } } else { - splane.z = 1.0 - splane.z; + splane.xyz = normalize(splane.xyz); + vec4 clamp_rect = lights.data[idx].atlas_rect; + + if (splane.z >= 0.0) { + + splane.z += 1.0; + + clamp_rect.y += clamp_rect.w; + + } else { + splane.z = 1.0 - splane.z; + } + + splane.xy /= splane.z; + + splane.xy = splane.xy * 0.5 + 0.5; + splane.z = (shadow_len - lights.data[idx].shadow_bias) * 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, lights.data[idx].soft_shadow_scale * scene_data.shadow_atlas_pixel_size, splane); + } + +#ifdef LIGHT_TRANSMITTANCE_USED + { + + vec4 clamp_rect = lights.data[idx].atlas_rect; + + //redo shadowmapping, but shrink the model a bit to avoid arctifacts + splane = (lights.data[idx].shadow_matrix * vec4(vertex - normalize(normal_interp) * lights.data[idx].transmittance_bias, 1.0)); + + shadow_len = length(splane.xyz); + splane = normalize(splane.xyz); + + if (splane.z >= 0.0) { + + splane.z += 1.0; + + } else { + + splane.z = 1.0 - splane.z; + } + + splane.xy /= splane.z; + splane.xy = splane.xy * 0.5 + 0.5; + splane.z = shadow_len * lights.data[idx].inv_radius; + splane.xy = clamp_rect.xy + splane.xy * clamp_rect.zw; + splane.w = 1.0; //needed? i think it should be 1 already + + float shadow_z = textureLod(sampler2D(shadow_atlas, material_samplers[SAMPLER_LINEAR_CLAMP]), splane.xy, 0.0).r; + transmittance_z = (splane.z - shadow_z) / lights.data[idx].inv_radius; } +#endif + + vec3 no_shadow = vec3(1.0); + + if (lights.data[idx].projector_rect != vec4(0.0)) { + + vec3 local_v = (lights.data[idx].shadow_matrix * vec4(vertex, 1.0)).xyz; + local_v = normalize(local_v); + + vec4 atlas_rect = lights.data[idx].projector_rect; + + if (local_v.z >= 0.0) { + + local_v.z += 1.0; + atlas_rect.y += atlas_rect.w; + + } else { + + local_v.z = 1.0 - local_v.z; + } + + local_v.xy /= local_v.z; + local_v.xy = local_v.xy * 0.5 + 0.5; + vec2 proj_uv = local_v.xy * atlas_rect.zw; + + vec2 proj_uv_ddx; + vec2 proj_uv_ddy; + { + vec3 local_v_ddx = (lights.data[idx].shadow_matrix * vec4(vertex + vertex_ddx, 1.0)).xyz; + local_v_ddx = normalize(local_v_ddx); + + if (local_v_ddx.z >= 0.0) { + + local_v_ddx.z += 1.0; + } else { + + local_v_ddx.z = 1.0 - local_v_ddx.z; + } + + local_v_ddx.xy /= local_v_ddx.z; + local_v_ddx.xy = local_v_ddx.xy * 0.5 + 0.5; + + proj_uv_ddx = local_v_ddx.xy * atlas_rect.zw - proj_uv; + + vec3 local_v_ddy = (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 { - splane.xy /= splane.z; - splane.xy = splane.xy * 0.5 + 0.5; - splane.z = shadow_len * lights.data[idx].inv_radius; - splane.xy = clamp_rect.xy + splane.xy * clamp_rect.zw; - splane.w = 1.0; //needed? i think it should be 1 already - float shadow = sample_shadow(shadow_atlas, scene_data.shadow_atlas_pixel_size, splane); + local_v_ddy.z = 1.0 - local_v_ddy.z; + } + + local_v_ddy.xy /= local_v_ddy.z; + local_v_ddy.xy = local_v_ddy.xy * 0.5 + 0.5; + + proj_uv_ddy = local_v_ddy.xy * atlas_rect.zw - proj_uv; + } + + vec4 proj = textureGrad(sampler2D(decal_atlas_srgb, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), proj_uv + atlas_rect.xy, proj_uv_ddx, proj_uv_ddy); + no_shadow = mix(no_shadow, proj.rgb, proj.a); + } - light_attenuation *= mix(shadow_color_enabled.rgb, vec3(1.0), shadow); + shadow_attenuation = mix(shadow_color_enabled.rgb, no_shadow, shadow); } #endif //USE_NO_SHADOWS - light_compute(normal, normalize(light_rel_vec), eye_vec, color_specular.rgb, light_attenuation, albedo, roughness, metallic, specular, color_specular.a * p_blob_intensity, -#ifdef LIGHT_TRANSMISSION_USED - transmission, + light_compute(normal, normalize(light_rel_vec), eye_vec, size_A, color_specular.rgb, light_attenuation, shadow_attenuation, albedo, roughness, metallic, specular, color_specular.a * p_blob_intensity, +#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, @@ -763,9 +1093,15 @@ void light_process_omni(uint idx, vec3 vertex, vec3 eye_vec, vec3 normal, vec3 a specular_light); } -void light_process_spot(uint idx, vec3 vertex, vec3 eye_vec, vec3 normal, vec3 albedo, float roughness, float metallic, float specular, float p_blob_intensity, -#ifdef LIGHT_TRANSMISSION_USED - vec3 transmission, +void light_process_spot(uint idx, vec3 vertex, vec3 eye_vec, vec3 normal, vec3 vertex_ddx, vec3 vertex_ddy, vec3 albedo, float roughness, float metallic, float specular, float p_blob_intensity, +#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, @@ -792,31 +1128,162 @@ void light_process_spot(uint idx, vec3 vertex, vec3 eye_vec, vec3 normal, vec3 a float scos = max(dot(-normalize(light_rel_vec), spot_dir), spot_att_angle.y); float spot_rim = max(0.0001, (1.0 - scos) / (1.0 - spot_att_angle.y)); spot_attenuation *= 1.0 - pow(spot_rim, spot_att_angle.x); - vec3 light_attenuation = vec3(spot_attenuation); + float light_attenuation = spot_attenuation; + vec3 shadow_attenuation = vec3(1.0); vec4 color_specular = unpackUnorm4x8(lights.data[idx].color_specular); color_specular.rgb *= attenuation_energy.y; + float size_A = 0.0; + + if (lights.data[idx].size > 0.0) { + + float t = lights.data[idx].size / max(0.001, light_length); + size_A = max(0.0, 1.0 - 1 / sqrt(1 + t * t)); + } /* if (lights.data[idx].atlas_rect!=vec4(0.0)) { //use projector texture } */ +#ifdef LIGHT_TRANSMITTANCE_USED + float transmittance_z = transmittance_depth; +#endif + #ifndef USE_NO_SHADOWS vec4 shadow_color_enabled = unpackUnorm4x8(lights.data[idx].shadow_color_enabled); if (shadow_color_enabled.w > 0.5) { //there is a shadowmap - vec4 splane = (lights.data[idx].shadow_matrix * vec4(vertex, 1.0)); + vec4 v = vec4(vertex, 1.0); + + v.xyz -= spot_dir * lights.data[idx].shadow_bias; + + float z_norm = dot(spot_dir, -light_rel_vec) * 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)))) * 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 - lights.data[idx].position) * lights.data[idx].inv_radius; + + float shadow; + + vec4 splane = (lights.data[idx].shadow_matrix * v); splane /= splane.w; - float shadow = sample_shadow(shadow_atlas, scene_data.shadow_atlas_pixel_size, splane); - light_attenuation *= mix(shadow_color_enabled.rgb, vec3(1.0), shadow); + if (lights.data[idx].soft_shadow_size > 0.0) { + //soft shadow + + //find blocker + + vec2 shadow_uv = splane.xy * lights.data[idx].atlas_rect.zw + lights.data[idx].atlas_rect.xy; + + float blocker_count = 0.0; + float blocker_average = 0.0; + + mat2 disk_rotation; + { + float r = quick_hash(gl_FragCoord.xy) * 2.0 * M_PI; + float sr = sin(r); + float cr = cos(r); + disk_rotation = mat2(vec2(cr, -sr), vec2(sr, cr)); + } + + float uv_size = lights.data[idx].soft_shadow_size * z_norm * lights.data[idx].soft_shadow_scale; + vec2 clamp_max = lights.data[idx].atlas_rect.xy + lights.data[idx].atlas_rect.zw; + for (uint i = 0; i < scene_data.penumbra_shadow_samples; i++) { + + vec2 suv = shadow_uv + (disk_rotation * scene_data.penumbra_shadow_kernel[i].xy) * uv_size; + suv = clamp(suv, lights.data[idx].atlas_rect.xy, clamp_max); + float d = textureLod(sampler2D(shadow_atlas, material_samplers[SAMPLER_LINEAR_CLAMP]), suv, 0.0).r; + if (d < z_norm) { + blocker_average += d; + blocker_count += 1.0; + } + } + + if (blocker_count > 0.0) { + + //blockers found, do soft shadow + blocker_average /= blocker_count; + float penumbra = (z_norm - blocker_average) / blocker_average; + uv_size *= penumbra; + + shadow = 0.0; + for (uint i = 0; i < scene_data.penumbra_shadow_samples; i++) { + vec2 suv = shadow_uv + (disk_rotation * scene_data.penumbra_shadow_kernel[i].xy) * uv_size; + suv = clamp(suv, lights.data[idx].atlas_rect.xy, clamp_max); + shadow += textureProj(sampler2DShadow(shadow_atlas, shadow_sampler), vec4(suv, z_norm, 1.0)); + } + + shadow /= float(scene_data.penumbra_shadow_samples); + + } else { + //no blockers found, so no shadow + shadow = 1.0; + } + + } else { + //hard shadow + vec4 shadow_uv = vec4(splane.xy * lights.data[idx].atlas_rect.zw + lights.data[idx].atlas_rect.xy, z_norm, 1.0); + + shadow = sample_pcf_shadow(shadow_atlas, lights.data[idx].soft_shadow_scale * scene_data.shadow_atlas_pixel_size, shadow_uv); + } + + vec3 no_shadow = vec3(1.0); + + if (lights.data[idx].projector_rect != vec4(0.0)) { + + splane = (lights.data[idx].shadow_matrix * vec4(vertex, 1.0)); + splane /= splane.w; + + vec2 proj_uv = splane.xy * lights.data[idx].projector_rect.zw; + + //ensure we have proper mipmaps + vec4 splane_ddx = (lights.data[idx].shadow_matrix * vec4(vertex + vertex_ddx, 1.0)); + splane_ddx /= splane_ddx.w; + vec2 proj_uv_ddx = splane_ddx.xy * lights.data[idx].projector_rect.zw - proj_uv; + + vec4 splane_ddy = (lights.data[idx].shadow_matrix * vec4(vertex + vertex_ddy, 1.0)); + splane_ddy /= splane_ddy.w; + vec2 proj_uv_ddy = splane_ddy.xy * lights.data[idx].projector_rect.zw - proj_uv; + + vec4 proj = textureGrad(sampler2D(decal_atlas_srgb, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), proj_uv + lights.data[idx].projector_rect.xy, proj_uv_ddx, proj_uv_ddy); + no_shadow = mix(no_shadow, proj.rgb, proj.a); + } + + shadow_attenuation = mix(shadow_color_enabled.rgb, no_shadow, shadow); + +#ifdef LIGHT_TRANSMITTANCE_USED + { + + splane = (lights.data[idx].shadow_matrix * vec4(vertex - normalize(normal_interp) * lights.data[idx].transmittance_bias, 1.0)); + splane /= splane.w; + splane.xy = splane.xy * lights.data[idx].atlas_rect.zw + 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 / lights.data[idx].inv_radius; + //distance to light plane + float z = dot(spot_dir, -light_rel_vec); + transmittance_z = z - shadow_z; + } +#endif //LIGHT_TRANSMITTANCE_USED } #endif //USE_NO_SHADOWS - light_compute(normal, normalize(light_rel_vec), eye_vec, color_specular.rgb, light_attenuation, albedo, roughness, metallic, specular, color_specular.a * p_blob_intensity, -#ifdef LIGHT_TRANSMISSION_USED - transmission, + light_compute(normal, normalize(light_rel_vec), eye_vec, size_A, color_specular.rgb, light_attenuation, shadow_attenuation, albedo, roughness, metallic, specular, color_specular.a * p_blob_intensity, +#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, @@ -1185,7 +1652,11 @@ void main() { vec3 vertex = vertex_interp; vec3 view = -normalize(vertex_interp); vec3 albedo = vec3(1.0); - vec3 transmission = vec3(0.0); + 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; vec3 emission = vec3(0.0); @@ -1223,9 +1694,7 @@ void main() { } #endif -#if defined(UV_USED) vec2 uv = uv_interp; -#endif #if defined(UV2_USED) || defined(USE_LIGHTMAP) vec2 uv2 = uv2_interp; @@ -1254,6 +1723,14 @@ FRAGMENT_SHADER_CODE /* clang-format on */ } +#if defined(LIGHT_TRANSMITTANCE_USED) +#ifdef SSS_MODE_SKIN + transmittance_color.a = sss_strength; +#else + transmittance_color.a *= sss_strength; +#endif +#endif + #if !defined(USE_SHADOW_TO_OPACITY) #if defined(ALPHA_SCISSOR_USED) @@ -1299,7 +1776,81 @@ FRAGMENT_SHADER_CODE discard; } #endif + /////////////////////// DECALS //////////////////////////////// + +#ifndef MODE_RENDER_DEPTH + + uvec4 cluster_cell = texture(usampler3D(cluster_texture, material_samplers[SAMPLER_NEAREST_CLAMP]), vec3(screen_uv, (abs(vertex.z) - scene_data.z_near) / (scene_data.z_far - scene_data.z_near))); + //used for interpolating anything cluster related + vec3 vertex_ddx = dFdx(vertex); + vec3 vertex_ddy = dFdy(vertex); + + { // process decals + + uint decal_count = cluster_cell.w >> CLUSTER_COUNTER_SHIFT; + uint decal_pointer = cluster_cell.w & CLUSTER_POINTER_MASK; + + //do outside for performance and avoiding arctifacts + + for (uint i = 0; i < decal_count; i++) { + + uint decal_index = cluster_data.indices[decal_pointer + i]; + if (!bool(decals.data[decal_index].mask & instances.data[instance_index].layer_mask)) { + continue; //not masked + } + + vec3 uv_local = (decals.data[decal_index].xform * vec4(vertex, 1.0)).xyz; + if (any(lessThan(uv_local, vec3(0.0, -1.0, 0.0))) || any(greaterThan(uv_local, vec3(1.0)))) { + continue; //out of decal + } + + //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); + } + + 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); + 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; + decal_normal.xy = decal_normal.xy * vec2(2.0, -2.0) - vec2(1.0, -1.0); //users prefer flipped y normal maps in most authoring software + decal_normal.z = sqrt(max(0.0, 1.0 - dot(decal_normal.xy, decal_normal.xy))); + //convert to view space, use xzy because y is up + decal_normal = (decals.data[decal_index].normal_xform * decal_normal.xzy).xyz; + + normal = normalize(mix(normal, decal_normal, decal_albedo.a)); + } + + if (decals.data[decal_index].orm_rect != vec4(0.0)) { + + vec3 decal_orm = 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; +#if defined(AO_USED) + ao = mix(ao, decal_orm.r, decal_albedo.a); +#endif + roughness = mix(roughness, decal_orm.g, decal_albedo.a); + metallic = mix(metallic, decal_orm.b, decal_albedo.a); + } + } + + if (decals.data[decal_index].emission_rect != vec4(0.0)) { + //emission is additive, so its independent from albedo + 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; + } + } + } +#endif //not render depth /////////////////////// LIGHTING ////////////////////////////// //apply energy conservation @@ -1404,8 +1955,6 @@ FRAGMENT_SHADER_CODE } #endif - uvec4 cluster_cell = texture(usampler3D(cluster_texture, material_samplers[SAMPLER_NEAREST_CLAMP]), vec3(screen_uv, (abs(vertex.z) - scene_data.z_near) / (scene_data.z_far - scene_data.z_near))); - { // process reflections vec4 reflection_accum = vec4(0.0, 0.0, 0.0, 0.0); @@ -1462,58 +2011,250 @@ FRAGMENT_SHADER_CODE continue; //not masked } - vec3 light_attenuation = vec3(1.0); + vec3 shadow_attenuation = vec3(1.0); + +#ifdef LIGHT_TRANSMITTANCE_USED + float transmittance_z = transmittance_depth; +#endif 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; + +#define BIAS_FUNC(m_var, m_idx) \ + m_var.xyz += light_dir * directional_lights.data[i].shadow_bias[m_idx]; \ + vec3 normal_bias = normalize(normal_interp) * (1.0 - max(0.0, dot(light_dir, -normalize(normal_interp)))) * directional_lights.data[i].shadow_normal_bias[m_idx]; \ + normal_bias -= light_dir * dot(light_dir, normal_bias); \ + m_var.xyz += normal_bias; + + float shadow = 0.0; if (depth_z < directional_lights.data[i].shadow_split_offsets.x) { - pssm_coord = (directional_lights.data[i].shadow_matrix1 * vec4(vertex, 1.0)); + vec4 v = vec4(vertex, 1.0); + + BIAS_FUNC(v, 0) + + pssm_coord = (directional_lights.data[i].shadow_matrix1 * v); + pssm_coord /= pssm_coord.w; + + if (directional_lights.data[i].softshadow_angle > 0) { + float range_pos = dot(directional_lights.data[i].direction, v.xyz); + float range_begin = directional_lights.data[i].shadow_range_begin.x; + float test_radius = (range_pos - range_begin) * directional_lights.data[i].softshadow_angle; + vec2 tex_scale = directional_lights.data[i].uv_scale1 * test_radius; + shadow = sample_directional_soft_shadow(directional_shadow_atlas, pssm_coord.xyz, tex_scale * directional_lights.data[i].soft_shadow_scale); + } else { + shadow = sample_directional_pcf_shadow(directional_shadow_atlas, scene_data.directional_shadow_pixel_size * directional_lights.data[i].soft_shadow_scale, pssm_coord); + } + + shadow_color = directional_lights.data[i].shadow_color1.rgb; + +#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) { - pssm_coord = (directional_lights.data[i].shadow_matrix2 * vec4(vertex, 1.0)); + + vec4 v = vec4(vertex, 1.0); + + BIAS_FUNC(v, 1) + + pssm_coord = (directional_lights.data[i].shadow_matrix2 * v); + pssm_coord /= pssm_coord.w; + + if (directional_lights.data[i].softshadow_angle > 0) { + float range_pos = dot(directional_lights.data[i].direction, v.xyz); + float range_begin = directional_lights.data[i].shadow_range_begin.y; + float test_radius = (range_pos - range_begin) * directional_lights.data[i].softshadow_angle; + vec2 tex_scale = directional_lights.data[i].uv_scale2 * test_radius; + shadow = sample_directional_soft_shadow(directional_shadow_atlas, pssm_coord.xyz, tex_scale * directional_lights.data[i].soft_shadow_scale); + } else { + shadow = sample_directional_pcf_shadow(directional_shadow_atlas, scene_data.directional_shadow_pixel_size * directional_lights.data[i].soft_shadow_scale, pssm_coord); + } + + shadow_color = directional_lights.data[i].shadow_color2.rgb; +#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) { - pssm_coord = (directional_lights.data[i].shadow_matrix3 * vec4(vertex, 1.0)); + + vec4 v = vec4(vertex, 1.0); + + BIAS_FUNC(v, 2) + + pssm_coord = (directional_lights.data[i].shadow_matrix3 * v); + pssm_coord /= pssm_coord.w; + + if (directional_lights.data[i].softshadow_angle > 0) { + float range_pos = dot(directional_lights.data[i].direction, v.xyz); + float range_begin = directional_lights.data[i].shadow_range_begin.z; + float test_radius = (range_pos - range_begin) * directional_lights.data[i].softshadow_angle; + vec2 tex_scale = directional_lights.data[i].uv_scale3 * test_radius; + shadow = sample_directional_soft_shadow(directional_shadow_atlas, pssm_coord.xyz, tex_scale * directional_lights.data[i].soft_shadow_scale); + } else { + shadow = sample_directional_pcf_shadow(directional_shadow_atlas, scene_data.directional_shadow_pixel_size * directional_lights.data[i].soft_shadow_scale, pssm_coord); + } + + shadow_color = directional_lights.data[i].shadow_color3.rgb; +#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 { - pssm_coord = (directional_lights.data[i].shadow_matrix4 * vec4(vertex, 1.0)); - } - pssm_coord /= pssm_coord.w; + vec4 v = vec4(vertex, 1.0); - float shadow = sample_shadow(directional_shadow_atlas, scene_data.directional_shadow_pixel_size, pssm_coord); + BIAS_FUNC(v, 3) + + pssm_coord = (directional_lights.data[i].shadow_matrix4 * v); + pssm_coord /= pssm_coord.w; + + if (directional_lights.data[i].softshadow_angle > 0) { + float range_pos = dot(directional_lights.data[i].direction, v.xyz); + float range_begin = directional_lights.data[i].shadow_range_begin.w; + float test_radius = (range_pos - range_begin) * directional_lights.data[i].softshadow_angle; + vec2 tex_scale = directional_lights.data[i].uv_scale4 * test_radius; + shadow = sample_directional_soft_shadow(directional_shadow_atlas, pssm_coord.xyz, tex_scale * directional_lights.data[i].soft_shadow_scale); + } else { + shadow = sample_directional_pcf_shadow(directional_shadow_atlas, scene_data.directional_shadow_pixel_size * directional_lights.data[i].soft_shadow_scale, pssm_coord); + } + + shadow_color = directional_lights.data[i].shadow_color4.rgb; + +#ifdef LIGHT_TRANSMITTANCE_USED + { + vec4 trans_vertex = vec4(vertex - normalize(normal_interp) * directional_lights.data[i].shadow_transmittance_bias.w, 1.0); + vec4 trans_coord = directional_lights.data[i].shadow_matrix4 * trans_vertex; + trans_coord /= trans_coord.w; + + float shadow_z = textureLod(sampler2D(directional_shadow_atlas, material_samplers[SAMPLER_LINEAR_CLAMP]), trans_coord.xy, 0.0).r; + shadow_z *= directional_lights.data[i].shadow_transmittance_z_scale.w; + float z = trans_coord.z * directional_lights.data[i].shadow_transmittance_z_scale.w; + + transmittance_z = z - shadow_z; + } +#endif + } if (directional_lights.data[i].blend_splits) { + vec3 shadow_color_blend = vec3(0.0); float pssm_blend; + float shadow2; if (depth_z < directional_lights.data[i].shadow_split_offsets.x) { - pssm_coord = (directional_lights.data[i].shadow_matrix2 * vec4(vertex, 1.0)); + vec4 v = vec4(vertex, 1.0); + BIAS_FUNC(v, 1) + pssm_coord = (directional_lights.data[i].shadow_matrix2 * v); + pssm_coord /= pssm_coord.w; + + if (directional_lights.data[i].softshadow_angle > 0) { + float range_pos = dot(directional_lights.data[i].direction, v.xyz); + float range_begin = directional_lights.data[i].shadow_range_begin.y; + float test_radius = (range_pos - range_begin) * directional_lights.data[i].softshadow_angle; + vec2 tex_scale = directional_lights.data[i].uv_scale2 * test_radius; + shadow2 = sample_directional_soft_shadow(directional_shadow_atlas, pssm_coord.xyz, tex_scale * directional_lights.data[i].soft_shadow_scale); + } else { + shadow2 = sample_directional_pcf_shadow(directional_shadow_atlas, scene_data.directional_shadow_pixel_size * directional_lights.data[i].soft_shadow_scale, pssm_coord); + } + pssm_blend = smoothstep(0.0, directional_lights.data[i].shadow_split_offsets.x, depth_z); + shadow_color_blend = directional_lights.data[i].shadow_color2.rgb; } else if (depth_z < directional_lights.data[i].shadow_split_offsets.y) { - pssm_coord = (directional_lights.data[i].shadow_matrix3 * vec4(vertex, 1.0)); + 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); + + shadow_color_blend = directional_lights.data[i].shadow_color3.rgb; } else if (depth_z < directional_lights.data[i].shadow_split_offsets.z) { - pssm_coord = (directional_lights.data[i].shadow_matrix4 * vec4(vertex, 1.0)); + 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_coord /= pssm_coord.w; + pssm_blend = sqrt(pssm_blend); - float shadow2 = sample_shadow(directional_shadow_atlas, scene_data.directional_shadow_pixel_size, pssm_coord); 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 - light_attenuation = mix(directional_lights.data[i].shadow_color, vec3(1.0), shadow); + shadow_attenuation = mix(shadow_color, vec3(1.0), shadow); + +#undef BIAS_FUNC } - light_compute(normal, directional_lights.data[i].direction, normalize(view), directional_lights.data[i].color * directional_lights.data[i].energy, light_attenuation, albedo, roughness, metallic, specular, directional_lights.data[i].specular * specular_blob_intensity, -#ifdef LIGHT_TRANSMISSION_USED - transmission, + light_compute(normal, directional_lights.data[i].direction, normalize(view), directional_lights.data[i].size, directional_lights.data[i].color * directional_lights.data[i].energy, 1.0, shadow_attenuation, albedo, roughness, metallic, specular, directional_lights.data[i].specular * specular_blob_intensity, +#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, rim_tint, @@ -1545,9 +2286,15 @@ FRAGMENT_SHADER_CODE continue; //not masked } - light_process_omni(light_index, vertex, view, normal, albedo, roughness, metallic, specular, specular_blob_intensity, -#ifdef LIGHT_TRANSMISSION_USED - transmission, + light_process_omni(light_index, vertex, view, normal, vertex_ddx, vertex_ddy, albedo, roughness, metallic, specular, specular_blob_intensity, +#ifdef LIGHT_BACKLIGHT_USED + backlight, +#endif +#ifdef LIGHT_TRANSMITTANCE_USED + transmittance_color, + transmittance_depth, + transmittance_curve, + transmittance_boost, #endif #ifdef LIGHT_RIM_USED rim, @@ -1578,9 +2325,15 @@ FRAGMENT_SHADER_CODE continue; //not masked } - light_process_spot(light_index, vertex, view, normal, albedo, roughness, metallic, specular, specular_blob_intensity, -#ifdef LIGHT_TRANSMISSION_USED - transmission, + light_process_spot(light_index, vertex, view, normal, vertex_ddx, vertex_ddy, albedo, roughness, metallic, specular, specular_blob_intensity, +#ifdef LIGHT_BACKLIGHT_USED + backlight, +#endif +#ifdef LIGHT_TRANSMITTANCE_USED + transmittance_color, + transmittance_depth, + transmittance_curve, + transmittance_boost, #endif #ifdef LIGHT_RIM_USED rim, @@ -1697,6 +2450,9 @@ FRAGMENT_SHADER_CODE #else +#ifdef SSS_MODE_SKIN + sss_strength = -sss_strength; +#endif diffuse_buffer = vec4(emission + diffuse_light + ambient_light, sss_strength); specular_buffer = vec4(specular_light, metallic); diff --git a/servers/rendering/rasterizer_rd/shaders/scene_high_end_inc.glsl b/servers/rendering/rasterizer_rd/shaders/scene_high_end_inc.glsl index baef1e060f..ce4fabf9f2 100644 --- a/servers/rendering/rasterizer_rd/shaders/scene_high_end_inc.glsl +++ b/servers/rendering/rasterizer_rd/shaders/scene_high_end_inc.glsl @@ -37,13 +37,23 @@ layout(set = 0, binding = 3, std140) uniform SceneData { vec2 viewport_size; vec2 screen_pixel_size; - //used for shadow mapping only - float z_offset; - float z_slope_scale; - float time; float reflection_multiplier; // one normally, zero when rendering reflections + bool pancake_shadows; + uint pad; + + //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]; + + 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; @@ -124,31 +134,39 @@ struct InstanceData { mat4 transform; mat4 normal_transform; uint flags; - uint instance_ofs; //instance_offset in instancing/skeleton buffer + uint instance_uniforms_ofs; //base offset in global buffer for instance variables uint gi_offset; //GI information when using lightmapping (VCT or lightmap) uint layer_mask; }; -layout(set = 0, binding = 4, std430) buffer Instances { +layout(set = 0, binding = 4, std430) restrict readonly buffer Instances { InstanceData data[]; } instances; -struct LightData { //this structure needs to be 128 bits +struct LightData { //this structure needs to be as packed as possible vec3 position; float inv_radius; vec3 direction; + float size; uint attenuation_energy; //attenuation uint color_specular; //rgb color, a specular (8 bit unorm) uint cone_attenuation_angle; // attenuation and angle, (16bit float) - uint mask; uint shadow_color_enabled; //shadow rgb color, a>0.5 enabled (8bit unorm) - vec4 atlas_rect; //used for shadow atlas uv on omni, and for projection atlas on spot + vec4 atlas_rect; // rect in the shadow atlas mat4 shadow_matrix; + float shadow_bias; + float shadow_normal_bias; + float transmittance_bias; + float soft_shadow_size; // for spot, it's the size in uv coordinates of the light, for omni it's the span angle + float soft_shadow_scale; // scales the shadow kernel for blurrier shadows + uint mask; + uint pad[2]; + vec4 projector_rect; //projector rect in srgb decal atlas }; -layout(set = 0, binding = 5, std140) uniform Lights { - LightData data[MAX_LIGHT_DATA_STRUCTS]; +layout(set = 0, binding = 5, std430) restrict readonly buffer Lights { + LightData data[]; } lights; @@ -173,18 +191,33 @@ struct DirectionalLightData { vec3 direction; float energy; vec3 color; + float size; float specular; - vec3 shadow_color; uint mask; + float softshadow_angle; + float soft_shadow_scale; bool blend_splits; bool shadow_enabled; float fade_from; float fade_to; + vec4 shadow_bias; + vec4 shadow_normal_bias; + vec4 shadow_transmittance_bias; + vec4 shadow_transmittance_z_scale; + 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; }; layout(set = 0, binding = 7, std140) uniform DirectionalLights { @@ -219,14 +252,45 @@ layout(set = 0, binding = 9) uniform texture3D gi_probe_textures[MAX_GI_PROBE_TE #define CLUSTER_POINTER_MASK ((1 << CLUSTER_COUNTER_SHIFT) - 1) #define CLUSTER_COUNTER_MASK 0xfff -layout(set = 0, binding = 10) uniform utexture3D cluster_texture; +layout(set = 0, binding = 10) uniform texture2D decal_atlas; +layout(set = 0, binding = 11) uniform texture2D decal_atlas_srgb; + +struct DecalData { + mat4 xform; //to decal transform + vec3 inv_extents; + float albedo_mix; + vec4 albedo_rect; + vec4 normal_rect; + vec4 orm_rect; + vec4 emission_rect; + vec4 modulate; + float emission_energy; + uint mask; + float upper_fade; + float lower_fade; + mat3x4 normal_xform; + vec3 normal; + float normal_fade; +}; + +layout(set = 0, binding = 12, std430) restrict readonly buffer Decals { + DecalData data[]; +} +decals; + +layout(set = 0, binding = 13) uniform utexture3D cluster_texture; -layout(set = 0, binding = 11, std430) buffer ClusterData { +layout(set = 0, binding = 14, std430) restrict readonly buffer ClusterData { uint indices[]; } cluster_data; -layout(set = 0, binding = 12) uniform texture2D directional_shadow_atlas; +layout(set = 0, binding = 15) uniform texture2D directional_shadow_atlas; + +layout(set = 0, binding = 16, std430) restrict readonly buffer GlobalVariableData { + vec4 data[]; +} +global_variables; // decal atlas @@ -258,7 +322,7 @@ layout(set = 3, binding = 4) uniform texture2D ao_buffer; /* Set 4 Skeleton & Instancing (Multimesh) */ -layout(set = 4, binding = 0, std430) buffer Transforms { +layout(set = 4, binding = 0, std430) restrict readonly buffer Transforms { vec4 data[]; } transforms; diff --git a/servers/rendering/rasterizer_rd/shaders/screen_space_reflection.glsl b/servers/rendering/rasterizer_rd/shaders/screen_space_reflection.glsl new file mode 100644 index 0000000000..e3c26c9b72 --- /dev/null +++ b/servers/rendering/rasterizer_rd/shaders/screen_space_reflection.glsl @@ -0,0 +1,262 @@ +/* clang-format off */ +[compute] + +#version 450 + +VERSION_DEFINES + + + +layout(local_size_x = 8, local_size_y = 8, local_size_z = 1) in; + +/* clang-format on */ + +layout(rgba16f, set = 0, binding = 0) uniform restrict readonly image2D source_diffuse; +layout(r32f, set = 0, binding = 1) uniform restrict readonly image2D source_depth; +layout(rgba16f, set = 1, binding = 0) uniform restrict writeonly image2D ssr_image; +#ifdef MODE_ROUGH +layout(r8, set = 1, binding = 1) uniform restrict writeonly image2D blur_radius_image; +#endif +layout(rgba8, set = 2, binding = 0) uniform restrict readonly image2D source_normal; +layout(set = 3, binding = 0) uniform sampler2D source_metallic; +#ifdef MODE_ROUGH +layout(set = 3, binding = 1) uniform sampler2D source_roughness; +#endif + +layout(push_constant, binding = 2, std430) uniform Params { + + vec4 proj_info; + + ivec2 screen_size; + float camera_z_near; + float camera_z_far; + + int num_steps; + float depth_tolerance; + float distance_fade; + float curve_fade_in; + + bool orthogonal; + float filter_mipmap_levels; + bool use_half_res; + uint metallic_mask; + + mat4 projection; +} +params; + +vec2 view_to_screen(vec3 view_pos, out float w) { + vec4 projected = params.projection * vec4(view_pos, 1.0); + projected.xyz /= projected.w; + projected.xy = projected.xy * 0.5 + 0.5; + w = projected.w; + return projected.xy; +} + +#define M_PI 3.14159265359 + +vec3 reconstructCSPosition(vec2 S, float z) { + if (params.orthogonal) { + return vec3((S.xy * params.proj_info.xy + params.proj_info.zw), z); + } else { + return vec3((S.xy * params.proj_info.xy + params.proj_info.zw) * z, z); + } +} + +void main() { + + // Pixel being shaded + ivec2 ssC = ivec2(gl_GlobalInvocationID.xy); + + if (any(greaterThan(ssC, params.screen_size))) { //too large, do nothing + return; + } + + vec2 pixel_size = 1.0 / vec2(params.screen_size); + vec2 uv = vec2(ssC) * pixel_size; + + uv += pixel_size * 0.5; + + float base_depth = imageLoad(source_depth, ssC).r; + + // World space point being shaded + vec3 vertex = reconstructCSPosition(uv * vec2(params.screen_size), base_depth); + + vec3 normal = imageLoad(source_normal, ssC).xyz * 2.0 - 1.0; + normal = normalize(normal); + normal.y = -normal.y; //because this code reads flipped + + vec3 view_dir = normalize(vertex); + vec3 ray_dir = normalize(reflect(view_dir, normal)); + + if (dot(ray_dir, normal) < 0.001) { + imageStore(ssr_image, ssC, vec4(0.0)); + return; + } + //ray_dir = normalize(view_dir - normal * dot(normal,view_dir) * 2.0); + //ray_dir = normalize(vec3(1.0, 1.0, -1.0)); + + //////////////// + + // make ray length and clip it against the near plane (don't want to trace beyond visible) + float ray_len = (vertex.z + ray_dir.z * params.camera_z_far) > -params.camera_z_near ? (-params.camera_z_near - vertex.z) / ray_dir.z : params.camera_z_far; + vec3 ray_end = vertex + ray_dir * ray_len; + + float w_begin; + vec2 vp_line_begin = view_to_screen(vertex, w_begin); + float w_end; + vec2 vp_line_end = view_to_screen(ray_end, w_end); + vec2 vp_line_dir = vp_line_end - vp_line_begin; + + // we need to interpolate w along the ray, to generate perspective correct reflections + w_begin = 1.0 / w_begin; + w_end = 1.0 / w_end; + + float z_begin = vertex.z * w_begin; + float z_end = ray_end.z * w_end; + + vec2 line_begin = vp_line_begin / pixel_size; + vec2 line_dir = vp_line_dir / pixel_size; + float z_dir = z_end - z_begin; + float w_dir = w_end - w_begin; + + // clip the line to the viewport edges + + float scale_max_x = min(1.0, 0.99 * (1.0 - vp_line_begin.x) / max(1e-5, vp_line_dir.x)); + float scale_max_y = min(1.0, 0.99 * (1.0 - vp_line_begin.y) / max(1e-5, vp_line_dir.y)); + float scale_min_x = min(1.0, 0.99 * vp_line_begin.x / max(1e-5, -vp_line_dir.x)); + float scale_min_y = min(1.0, 0.99 * vp_line_begin.y / max(1e-5, -vp_line_dir.y)); + float line_clip = min(scale_max_x, scale_max_y) * min(scale_min_x, scale_min_y); + line_dir *= line_clip; + z_dir *= line_clip; + w_dir *= line_clip; + + // clip z and w advance to line advance + vec2 line_advance = normalize(line_dir); // down to pixel + float step_size = length(line_advance) / length(line_dir); + float z_advance = z_dir * step_size; // adapt z advance to line advance + float w_advance = w_dir * step_size; // adapt w advance to line advance + + // make line advance faster if direction is closer to pixel edges (this avoids sampling the same pixel twice) + float advance_angle_adj = 1.0 / max(abs(line_advance.x), abs(line_advance.y)); + line_advance *= advance_angle_adj; // adapt z advance to line advance + z_advance *= advance_angle_adj; + w_advance *= advance_angle_adj; + + vec2 pos = line_begin; + float z = z_begin; + float w = w_begin; + float z_from = z / w; + float z_to = z_from; + float depth; + vec2 prev_pos = pos; + + bool found = false; + + float steps_taken = 0.0; + + for (int i = 0; i < params.num_steps; i++) { + + pos += line_advance; + z += z_advance; + w += w_advance; + + // convert to linear depth + + depth = imageLoad(source_depth, ivec2(pos - 0.5)).r; + + if (-depth >= params.camera_z_far) { //went beyond camera + break; + } + + z_from = z_to; + z_to = z / w; + + if (depth > z_to) { + // if depth was surpassed + if (depth <= max(z_to, z_from) + params.depth_tolerance) { + // check the depth tolerance + //check that normal is valid + found = true; + } + break; + } + + steps_taken += 1.0; + prev_pos = pos; + } + + if (found) { + + float margin_blend = 1.0; + + vec2 margin = vec2((params.screen_size.x + params.screen_size.y) * 0.5 * 0.05); // make a uniform margin + if (any(bvec4(lessThan(pos, -margin), greaterThan(pos, params.screen_size + margin)))) { + // clip outside screen + margin + imageStore(ssr_image, ssC, vec4(0.0)); + return; + } + + { + //blend fading out towards external margin + vec2 margin_grad = mix(pos - params.screen_size, -pos, lessThan(pos, vec2(0.0))); + margin_blend = 1.0 - smoothstep(0.0, margin.x, max(margin_grad.x, margin_grad.y)); + //margin_blend = 1.0; + } + + vec2 final_pos; + float grad; + grad = steps_taken / float(params.num_steps); + float initial_fade = params.curve_fade_in == 0.0 ? 1.0 : pow(clamp(grad, 0.0, 1.0), params.curve_fade_in); + float fade = pow(clamp(1.0 - grad, 0.0, 1.0), params.distance_fade) * initial_fade; + final_pos = pos; + + vec4 final_color; + +#ifdef MODE_ROUGH + + // if roughness is enabled, do screen space cone tracing + float blur_radius = 0.0; + float roughness = texelFetch(source_roughness, ssC << 1, 0).r; + + if (roughness > 0.001) { + + float cone_angle = min(roughness, 0.999) * M_PI * 0.5; + float cone_len = length(final_pos - line_begin); + float op_len = 2.0 * tan(cone_angle) * cone_len; // opposite side of iso triangle + { + // fit to sphere inside cone (sphere ends at end of cone), something like this: + // ___ + // \O/ + // V + // + // as it avoids bleeding from beyond the reflection as much as possible. As a plus + // it also makes the rough reflection more elongated. + float a = op_len; + float h = cone_len; + float a2 = a * a; + float fh2 = 4.0f * h * h; + blur_radius = (a * (sqrt(a2 + fh2) - a)) / (4.0f * h); + } + } + + final_color = imageLoad(source_diffuse, ivec2((final_pos - 0.5) * pixel_size)); + + imageStore(blur_radius_image, ssC, vec4(blur_radius / 255.0)); //stored in r8 + +#endif + + final_color = vec4(imageLoad(source_diffuse, ivec2(final_pos - 0.5)).rgb, fade * margin_blend); + //change blend by metallic + vec4 metallic_mask = unpackUnorm4x8(params.metallic_mask); + final_color.a *= dot(metallic_mask, texelFetch(source_metallic, ssC << 1, 0)); + + imageStore(ssr_image, ssC, final_color); + + } else { +#ifdef MODE_ROUGH + imageStore(blur_radius_image, ssC, vec4(0.0)); +#endif + imageStore(ssr_image, ssC, vec4(0.0)); + } +} diff --git a/servers/rendering/rasterizer_rd/shaders/screen_space_reflection_filter.glsl b/servers/rendering/rasterizer_rd/shaders/screen_space_reflection_filter.glsl new file mode 100644 index 0000000000..1a5dd5ab55 --- /dev/null +++ b/servers/rendering/rasterizer_rd/shaders/screen_space_reflection_filter.glsl @@ -0,0 +1,164 @@ +/* clang-format off */ +[compute] + +#version 450 + +VERSION_DEFINES + + + +layout(local_size_x = 8, local_size_y = 8, local_size_z = 1) in; + +/* clang-format on */ + +layout(rgba16f, set = 0, binding = 0) uniform restrict readonly image2D source_ssr; +layout(r8, set = 0, binding = 1) uniform restrict readonly image2D source_radius; +layout(rgba8, set = 1, binding = 0) uniform restrict readonly image2D source_normal; + +layout(rgba16f, set = 2, binding = 0) uniform restrict writeonly image2D dest_ssr; +#ifndef VERTICAL_PASS +layout(r8, set = 2, binding = 1) uniform restrict writeonly image2D dest_radius; +#endif +layout(r32f, set = 3, binding = 0) uniform restrict readonly image2D source_depth; + +layout(push_constant, binding = 2, std430) uniform Params { + + vec4 proj_info; + + bool orthogonal; + float edge_tolerance; + int increment; + uint pad; + + ivec2 screen_size; + bool vertical; + uint steps; +} +params; + +#define GAUSS_TABLE_SIZE 15 + +const float gauss_table[GAUSS_TABLE_SIZE + 1] = float[]( + 0.1847392078702266, + 0.16595854345772326, + 0.12031364177766891, + 0.07038755277896766, + 0.03322925565155569, + 0.012657819729901945, + 0.0038903040680094217, + 0.0009646503390864025, + 0.00019297087402915717, + 0.000031139936308099136, + 0.000004053309048174758, + 4.255228059965837e-7, + 3.602517634249573e-8, + 2.4592560765896795e-9, + 1.3534945386863618e-10, + 0.0 //one more for interpolation +); + +float gauss_weight(float p_val) { + + float idxf; + float c = modf(max(0.0, p_val * float(GAUSS_TABLE_SIZE)), idxf); + int idx = int(idxf); + if (idx >= GAUSS_TABLE_SIZE + 1) { + return 0.0; + } + + return mix(gauss_table[idx], gauss_table[idx + 1], c); +} + +#define M_PI 3.14159265359 + +vec3 reconstructCSPosition(vec2 S, float z) { + if (params.orthogonal) { + return vec3((S.xy * params.proj_info.xy + params.proj_info.zw), z); + } else { + return vec3((S.xy * params.proj_info.xy + params.proj_info.zw) * z, z); + } +} + +void do_filter(inout vec4 accum, inout float accum_radius, inout float divisor, ivec2 texcoord, ivec2 increment, vec3 p_pos, vec3 normal, float p_limit_radius) { + + for (int i = 1; i < params.steps; i++) { + float d = float(i * params.increment); + ivec2 tc = texcoord + increment * i; + float depth = imageLoad(source_depth, tc).r; + vec3 view_pos = reconstructCSPosition(vec2(tc) + 0.5, depth); + vec3 view_normal = normalize(imageLoad(source_normal, tc).rgb * 2.0 - 1.0); + view_normal.y = -view_normal.y; + + float r = imageLoad(source_radius, tc).r; + float radius = round(r * 255.0); + + float angle_n = 1.0 - abs(dot(normal, view_normal)); + if (angle_n > params.edge_tolerance) { + break; + } + + float angle = abs(dot(normal, normalize(view_pos - p_pos))); + + if (angle > params.edge_tolerance) { + break; + } + + if (d < radius) { + + float w = gauss_weight(d / radius); + accum += imageLoad(source_ssr, tc) * w; +#ifndef VERTICAL_PASS + accum_radius += r * w; +#endif + divisor += w; + } + } +} + +void main() { + + // Pixel being shaded + ivec2 ssC = ivec2(gl_GlobalInvocationID.xy); + + if (any(greaterThan(ssC, params.screen_size))) { //too large, do nothing + return; + } + + float base_contrib = gauss_table[0]; + + vec4 accum = imageLoad(source_ssr, ssC); + + float accum_radius = imageLoad(source_radius, ssC).r; + float radius = accum_radius * 255.0; + + float divisor = gauss_table[0]; + accum *= divisor; + accum_radius *= divisor; +#ifdef VERTICAL_PASS + ivec2 direction = ivec2(0, params.increment); +#else + ivec2 direction = ivec2(params.increment, 0); +#endif + float depth = imageLoad(source_depth, ssC).r; + vec3 pos = reconstructCSPosition(vec2(ssC) + 0.5, depth); + vec3 normal = imageLoad(source_normal, ssC).xyz * 2.0 - 1.0; + normal = normalize(normal); + normal.y = -normal.y; + + do_filter(accum, accum_radius, divisor, ssC, direction, pos, normal, radius); + do_filter(accum, accum_radius, divisor, ssC, -direction, pos, normal, radius); + + if (divisor > 0.0) { + accum /= divisor; + accum_radius /= divisor; + } else { + accum = vec4(0.0); + accum_radius = 0.0; + } + + imageStore(dest_ssr, ssC, accum); + +#ifndef VERTICAL_PASS + imageStore(dest_radius, ssC, vec4(accum_radius)); +#endif +} diff --git a/servers/rendering/rasterizer_rd/shaders/screen_space_reflection_scale.glsl b/servers/rendering/rasterizer_rd/shaders/screen_space_reflection_scale.glsl new file mode 100644 index 0000000000..cec6c14c76 --- /dev/null +++ b/servers/rendering/rasterizer_rd/shaders/screen_space_reflection_scale.glsl @@ -0,0 +1,96 @@ +/* clang-format off */ +[compute] + +#version 450 + +VERSION_DEFINES + + +layout(local_size_x = 8, local_size_y = 8, local_size_z = 1) in; + +/* clang-format on */ + +layout(set = 0, binding = 0) uniform sampler2D source_ssr; +layout(set = 1, binding = 0) uniform sampler2D source_depth; +layout(set = 1, binding = 1) uniform sampler2D source_normal; +layout(rgba16f, set = 2, binding = 0) uniform restrict writeonly image2D dest_ssr; +layout(r32f, set = 3, binding = 0) uniform restrict writeonly image2D dest_depth; +layout(rgba8, set = 3, binding = 1) uniform restrict writeonly image2D dest_normal; + +layout(push_constant, binding = 1, std430) uniform Params { + + ivec2 screen_size; + float camera_z_near; + float camera_z_far; + + bool orthogonal; + bool filtered; + uint pad[2]; +} +params; + +void main() { + + // Pixel being shaded + ivec2 ssC = ivec2(gl_GlobalInvocationID.xy); + + if (any(greaterThan(ssC, params.screen_size))) { //too large, do nothing + return; + } + //do not filter, SSR will generate arctifacts if this is done + + float divisor = 0.0; + vec4 color; + float depth; + vec3 normal; + + if (params.filtered) { + + color = vec4(0.0); + depth = 0.0; + normal = vec3(0.0); + + for (int i = 0; i < 4; i++) { + + ivec2 ofs = ssC << 1; + if (bool(i & 1)) { + ofs.x += 1; + } + if (bool(i & 2)) { + ofs.y += 1; + } + 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; + + d = d * 2.0 - 1.0; + if (params.orthogonal) { + d = ((d + (params.camera_z_far + params.camera_z_near) / (params.camera_z_far - params.camera_z_near)) * (params.camera_z_far - params.camera_z_near)) / 2.0; + } else { + d = 2.0 * params.camera_z_near * params.camera_z_far / (params.camera_z_far + params.camera_z_near - d * (params.camera_z_far - params.camera_z_near)); + } + depth += -d; + } + + color /= 4.0; + depth /= 4.0; + normal = normalize(normal / 4.0) * 0.5 + 0.5; + + } else { + color = texelFetch(source_ssr, ssC << 1, 0); + depth = texelFetch(source_depth, ssC << 1, 0).r; + normal = texelFetch(source_normal, ssC << 1, 0).xyz; + + depth = depth * 2.0 - 1.0; + if (params.orthogonal) { + depth = ((depth + (params.camera_z_far + params.camera_z_near) / (params.camera_z_far - params.camera_z_near)) * (params.camera_z_far - params.camera_z_near)) / 2.0; + } else { + depth = 2.0 * params.camera_z_near * params.camera_z_far / (params.camera_z_far + params.camera_z_near - depth * (params.camera_z_far - params.camera_z_near)); + } + depth = -depth; + } + + imageStore(dest_ssr, ssC, color); + imageStore(dest_depth, ssC, vec4(depth)); + imageStore(dest_normal, ssC, vec4(normal, 0.0)); +} diff --git a/servers/rendering/rasterizer_rd/shaders/sky.glsl b/servers/rendering/rasterizer_rd/shaders/sky.glsl index 3f433eb2ee..536077980d 100644 --- a/servers/rendering/rasterizer_rd/shaders/sky.glsl +++ b/servers/rendering/rasterizer_rd/shaders/sky.glsl @@ -58,6 +58,11 @@ params; layout(set = 0, binding = 0) uniform sampler material_samplers[12]; +layout(set = 0, binding = 1, std430) restrict readonly buffer GlobalVariableData { + vec4 data[]; +} +global_variables; + #ifdef USE_MATERIAL_UNIFORMS layout(set = 1, binding = 0, std140) uniform MaterialUniforms{ /* clang-format off */ @@ -96,9 +101,8 @@ layout(set = 2, binding = 2) uniform texture2D quarter_res; #endif struct DirectionalLightData { - vec3 direction; - float energy; - vec3 color; + vec4 direction_energy; + vec4 color_size; bool enabled; }; @@ -141,15 +145,15 @@ void main() { vec4 quarter_res_color = vec4(1.0); #ifdef USE_CUBEMAP_PASS - float using_cubemap = 1.0; + 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]), cube_normal); + half_res_color = texture(samplerCube(half_res, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), inverted_cube_normal); #endif #ifdef USES_QUARTER_RES_COLOR - quarter_res_color = texture(samplerCube(quarter_res, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), cube_normal); + quarter_res_color = texture(samplerCube(quarter_res, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), inverted_cube_normal); #endif #else - float using_cubemap = 0.0; #ifdef USES_HALF_RES_COLOR half_res_color = textureLod(sampler2D(half_res, material_samplers[SAMPLER_LINEAR_CLAMP]), uv, 0.0); #endif @@ -178,4 +182,10 @@ FRAGMENT_SHADER_CODE frag_color.rgb = color * params.position_multiplier.w; frag_color.a = alpha; + + // Blending is disabled for Sky, so alpha doesn't blend + // alpha is used for subsurface scattering so make sure it doesn't get applied to Sky + if (!AT_CUBEMAP_PASS && !AT_HALF_RES_PASS && !AT_QUARTER_RES_PASS) { + frag_color.a = 0.0; + } } diff --git a/servers/rendering/rasterizer_rd/shaders/specular_merge.glsl b/servers/rendering/rasterizer_rd/shaders/specular_merge.glsl new file mode 100644 index 0000000000..b24f7dccc7 --- /dev/null +++ b/servers/rendering/rasterizer_rd/shaders/specular_merge.glsl @@ -0,0 +1,59 @@ +/* clang-format off */ +[vertex] + +#version 450 + +VERSION_DEFINES + +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 + +layout(location = 0) in vec2 uv_interp; +/* clang-format on */ + +layout(set = 0, binding = 0) uniform sampler2D specular; + +#ifdef MODE_SSR + +layout(set = 1, binding = 0) uniform sampler2D ssr; + +#endif + +#ifdef MODE_MERGE + +layout(set = 2, binding = 0) uniform sampler2D diffuse; + +#endif + +layout(location = 0) out vec4 frag_color; + +void main() { + + frag_color.rgb = texture(specular, uv_interp).rgb; + frag_color.a = 0.0; +#ifdef MODE_SSR + + vec4 ssr_color = texture(ssr, uv_interp); + frag_color.rgb = mix(frag_color.rgb, ssr_color.rgb, ssr_color.a); +#endif + +#ifdef MODE_MERGE + frag_color += texture(diffuse, uv_interp); +#endif + //added using additive blend +} diff --git a/servers/rendering/rasterizer_rd/shaders/subsurface_scattering.glsl b/servers/rendering/rasterizer_rd/shaders/subsurface_scattering.glsl new file mode 100644 index 0000000000..41f8fde3ca --- /dev/null +++ b/servers/rendering/rasterizer_rd/shaders/subsurface_scattering.glsl @@ -0,0 +1,198 @@ +/* clang-format off */ +[compute] + +#version 450 + +VERSION_DEFINES + + + +layout(local_size_x = 8, local_size_y = 8, local_size_z = 1) in; + +/* clang-format on */ + +#ifdef USE_25_SAMPLES +const int kernel_size = 13; + +const vec2 kernel[kernel_size] = vec2[]( + vec2(0.530605, 0.0), + vec2(0.0211412, 0.0208333), + vec2(0.0402784, 0.0833333), + vec2(0.0493588, 0.1875), + vec2(0.0410172, 0.333333), + vec2(0.0263642, 0.520833), + vec2(0.017924, 0.75), + vec2(0.0128496, 1.02083), + vec2(0.0094389, 1.33333), + vec2(0.00700976, 1.6875), + vec2(0.00500364, 2.08333), + vec2(0.00333804, 2.52083), + vec2(0.000973794, 3.0)); + +const vec4 skin_kernel[kernel_size] = vec4[]( + vec4(0.530605, 0.613514, 0.739601, 0), + vec4(0.0211412, 0.0459286, 0.0378196, 0.0208333), + vec4(0.0402784, 0.0657244, 0.04631, 0.0833333), + vec4(0.0493588, 0.0367726, 0.0219485, 0.1875), + vec4(0.0410172, 0.0199899, 0.0118481, 0.333333), + vec4(0.0263642, 0.0119715, 0.00684598, 0.520833), + vec4(0.017924, 0.00711691, 0.00347194, 0.75), + vec4(0.0128496, 0.00356329, 0.00132016, 1.02083), + vec4(0.0094389, 0.00139119, 0.000416598, 1.33333), + vec4(0.00700976, 0.00049366, 0.000151938, 1.6875), + vec4(0.00500364, 0.00020094, 5.28848e-005, 2.08333), + vec4(0.00333804, 7.85443e-005, 1.2945e-005, 2.52083), + vec4(0.000973794, 1.11862e-005, 9.43437e-007, 3)); + +#endif //USE_25_SAMPLES + +#ifdef USE_17_SAMPLES +const int kernel_size = 9; +const vec2 kernel[kernel_size] = vec2[]( + vec2(0.536343, 0.0), + vec2(0.0324462, 0.03125), + vec2(0.0582416, 0.125), + vec2(0.0571056, 0.28125), + vec2(0.0347317, 0.5), + vec2(0.0216301, 0.78125), + vec2(0.0144609, 1.125), + vec2(0.0100386, 1.53125), + vec2(0.00317394, 2.0)); + +const vec4 skin_kernel[kernel_size] = vec4[]( + vec4(0.536343, 0.624624, 0.748867, 0), + vec4(0.0324462, 0.0656718, 0.0532821, 0.03125), + vec4(0.0582416, 0.0659959, 0.0411329, 0.125), + vec4(0.0571056, 0.0287432, 0.0172844, 0.28125), + vec4(0.0347317, 0.0151085, 0.00871983, 0.5), + vec4(0.0216301, 0.00794618, 0.00376991, 0.78125), + vec4(0.0144609, 0.00317269, 0.00106399, 1.125), + vec4(0.0100386, 0.000914679, 0.000275702, 1.53125), + vec4(0.00317394, 0.000134823, 3.77269e-005, 2)); +#endif //USE_17_SAMPLES + +#ifdef USE_11_SAMPLES +const int kernel_size = 6; +const vec2 kernel[kernel_size] = vec2[]( + vec2(0.560479, 0.0), + vec2(0.0771802, 0.08), + vec2(0.0821904, 0.32), + vec2(0.03639, 0.72), + vec2(0.0192831, 1.28), + vec2(0.00471691, 2.0)); + +const vec4 skin_kernel[kernel_size] = vec4[]( + + vec4(0.560479, 0.669086, 0.784728, 0), + vec4(0.0771802, 0.113491, 0.0793803, 0.08), + vec4(0.0821904, 0.0358608, 0.0209261, 0.32), + vec4(0.03639, 0.0130999, 0.00643685, 0.72), + vec4(0.0192831, 0.00282018, 0.00084214, 1.28), + vec4(0.00471691, 0.000184771, 5.07565e-005, 2)); + +#endif //USE_11_SAMPLES + +layout(push_constant, binding = 1, std430) uniform Params { + + ivec2 screen_size; + float camera_z_far; + float camera_z_near; + + bool vertical; + bool orthogonal; + float unit_size; + float scale; + + float depth_scale; + uint pad[3]; +} +params; + +layout(set = 0, binding = 0) uniform sampler2D source_image; +layout(rgba16f, set = 1, binding = 0) uniform restrict writeonly image2D dest_image; +layout(set = 2, binding = 0) uniform sampler2D source_depth; + +void do_filter(inout vec3 color_accum, inout vec3 divisor, vec2 uv, vec2 step, bool p_skin) { + + // Accumulate the other samples: + for (int i = 1; i < kernel_size; i++) { + // Fetch color and depth for current sample: + vec2 offset = uv + kernel[i].y * step; + vec4 color = texture(source_image, offset); + + if (abs(color.a) < 0.001) { + break; //mix no more + } + + vec3 w; + if (p_skin) { + //skin + w = skin_kernel[i].rgb; + } else { + w = vec3(kernel[i].x); + } + + color_accum += color.rgb * w; + divisor += w; + } +} + +void main() { + + // Pixel being shaded + ivec2 ssC = ivec2(gl_GlobalInvocationID.xy); + + if (any(greaterThan(ssC, params.screen_size))) { //too large, do nothing + return; + } + + vec2 uv = (vec2(ssC) + 0.5) / vec2(params.screen_size); + + // Fetch color of current pixel: + vec4 base_color = texture(source_image, uv); + float strength = abs(base_color.a); + + if (strength > 0.0) { + + vec2 dir = params.vertical ? vec2(0.0, 1.0) : vec2(1.0, 0.0); + + // Fetch linear depth of current pixel: + float depth = texture(source_depth, uv).r * 2.0 - 1.0; + float depth_scale; + + if (params.orthogonal) { + depth = ((depth + (params.camera_z_far + params.camera_z_near) / (params.camera_z_far - params.camera_z_near)) * (params.camera_z_far - params.camera_z_near)) / 2.0; + depth_scale = params.unit_size; //remember depth is negative by default in OpenGL + } else { + depth = 2.0 * params.camera_z_near * params.camera_z_far / (params.camera_z_far + params.camera_z_near - depth * (params.camera_z_far - params.camera_z_near)); + depth_scale = params.unit_size / depth; //remember depth is negative by default in OpenGL + } + + float scale = mix(params.scale, depth_scale, params.depth_scale); + + // Calculate the final step to fetch the surrounding pixels: + vec2 step = scale * dir; + step *= strength; + step /= 3.0; + // Accumulate the center sample: + + vec3 divisor; + bool skin = bool(base_color.a < 0.0); + + if (skin) { + //skin + divisor = skin_kernel[0].rgb; + } else { + divisor = vec3(kernel[0].x); + } + + vec3 color = base_color.rgb * divisor; + + do_filter(color, divisor, uv, step, skin); + do_filter(color, divisor, uv, -step, skin); + + base_color.rgb = color / divisor; + } + + imageStore(dest_image, ssC, base_color); +} diff --git a/servers/rendering/rasterizer_rd/shaders/tonemap.glsl b/servers/rendering/rasterizer_rd/shaders/tonemap.glsl index 524ca5e2ea..a142d263e2 100644 --- a/servers/rendering/rasterizer_rd/shaders/tonemap.glsl +++ b/servers/rendering/rasterizer_rd/shaders/tonemap.glsl @@ -48,6 +48,10 @@ layout(push_constant, binding = 1, std430) uniform Params { float exposure; float white; float auto_exposure_grey; + + vec2 pixel_size; + bool use_fxaa; + uint pad; } params; @@ -255,16 +259,63 @@ vec3 apply_color_correction(vec3 color, sampler3D correction_tex) { return texture(correction_tex, color).rgb; } +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; + vec3 rgbM = color; + vec3 luma = vec3(0.299, 0.587, 0.114); + float lumaNW = dot(rgbNW, luma); + float lumaNE = dot(rgbNE, luma); + float lumaSW = dot(rgbSW, luma); + float lumaSE = dot(rgbSE, luma); + float lumaM = dot(rgbM, luma); + float lumaMin = min(lumaM, min(min(lumaNW, lumaNE), min(lumaSW, lumaSE))); + float lumaMax = max(lumaM, max(max(lumaNW, lumaNE), max(lumaSW, lumaSE))); + + vec2 dir; + dir.x = -((lumaNW + lumaNE) - (lumaSW + lumaSE)); + dir.y = ((lumaNW + lumaSW) - (lumaNE + lumaSE)); + + float dirReduce = max((lumaNW + lumaNE + lumaSW + lumaSE) * + (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; + + vec3 rgbA = 0.5 * (textureLod(source_color, uv_interp + dir * (1.0 / 3.0 - 0.5), 0.0).xyz * exposure + textureLod(source_color, uv_interp + dir * (2.0 / 3.0 - 0.5), 0.0).xyz) * exposure; + vec3 rgbB = rgbA * 0.5 + 0.25 * (textureLod(source_color, uv_interp + dir * -0.5, 0.0).xyz * exposure + + textureLod(source_color, uv_interp + dir * 0.5, 0.0).xyz * exposure); + + float lumaB = dot(rgbB, luma); + if ((lumaB < lumaMin) || (lumaB > lumaMax)) + return rgbA; + else + return rgbB; +} + void main() { vec3 color = textureLod(source_color, uv_interp, 0.0f).rgb; // Exposure + float exposure = params.exposure; + if (params.use_auto_exposure) { - color /= 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.auto_exposure_grey); } - color *= params.exposure; + color *= exposure; // Early Tonemap & SRGB Conversion @@ -274,6 +325,9 @@ void main() { color.rgb = mix(color.rgb, glow, params.glow_intensity); } + if (params.use_fxaa) { + color = do_fxaa(color, exposure, uv_interp); + } color = apply_tonemapping(color, params.white); color = linear_to_srgb(color); // regular linear -> SRGB conversion |