diff options
Diffstat (limited to 'servers/rendering/renderer_rd/shaders/effects')
9 files changed, 1649 insertions, 0 deletions
diff --git a/servers/rendering/renderer_rd/shaders/effects/SCsub b/servers/rendering/renderer_rd/shaders/effects/SCsub new file mode 100644 index 0000000000..fc513d3fb9 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/effects/SCsub @@ -0,0 +1,17 @@ +#!/usr/bin/env python + +Import("env") + +if "RD_GLSL" in env["BUILDERS"]: + # find all include files + gl_include_files = [str(f) for f in Glob("*_inc.glsl")] + + # find all shader code(all glsl files excluding our include files) + glsl_files = [str(f) for f in Glob("*.glsl") if str(f) not in gl_include_files] + + # make sure we recompile shaders if include files change + env.Depends([f + ".gen.h" for f in glsl_files], gl_include_files) + + # compile shaders + for glsl_file in glsl_files: + env.RD_GLSL(glsl_file) diff --git a/servers/rendering/renderer_rd/shaders/effects/blur_raster.glsl b/servers/rendering/renderer_rd/shaders/effects/blur_raster.glsl new file mode 100644 index 0000000000..96f5c3e9f2 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/effects/blur_raster.glsl @@ -0,0 +1,148 @@ +/* clang-format off */ +#[vertex] + +#version 450 + +#VERSION_DEFINES + +#include "blur_raster_inc.glsl" + +layout(location = 0) out vec2 uv_interp; +/* clang-format on */ + +void main() { + vec2 base_arr[4] = vec2[](vec2(0.0, 0.0), vec2(0.0, 1.0), vec2(1.0, 1.0), vec2(1.0, 0.0)); + uv_interp = base_arr[gl_VertexIndex]; + + gl_Position = vec4(uv_interp * 2.0 - 1.0, 0.0, 1.0); +} + +/* clang-format off */ +#[fragment] + +#version 450 + +#VERSION_DEFINES + +#include "blur_raster_inc.glsl" + +layout(location = 0) in vec2 uv_interp; +/* clang-format on */ + +layout(set = 0, binding = 0) uniform sampler2D source_color; + +#ifdef GLOW_USE_AUTO_EXPOSURE +layout(set = 1, binding = 0) uniform sampler2D source_auto_exposure; +#endif + +layout(location = 0) out vec4 frag_color; + +void main() { + // We do not apply our color scale for our mobile renderer here, we'll leave our colors at half brightness and apply scale in the tonemap raster. + +#ifdef MODE_MIPMAP + + vec2 pix_size = blur.pixel_size; + vec4 color = texture(source_color, uv_interp + vec2(-0.5, -0.5) * pix_size); + color += texture(source_color, uv_interp + vec2(0.5, -0.5) * pix_size); + color += texture(source_color, uv_interp + vec2(0.5, 0.5) * pix_size); + color += texture(source_color, uv_interp + vec2(-0.5, 0.5) * pix_size); + frag_color = color / 4.0; + +#endif + +#ifdef MODE_GAUSSIAN_BLUR + + // Simpler blur uses SIGMA2 for the gaussian kernel for a stronger effect + + // note, for blur blur.luminance_multiplier is irrelavant, we would be multiplying and then dividing by this amount. + + if (bool(blur.flags & FLAG_HORIZONTAL)) { + vec2 pix_size = blur.pixel_size; + pix_size *= 0.5; //reading from larger buffer, so use more samples + vec4 color = texture(source_color, uv_interp + vec2(0.0, 0.0) * pix_size) * 0.214607; + color += texture(source_color, uv_interp + vec2(1.0, 0.0) * pix_size) * 0.189879; + color += texture(source_color, uv_interp + vec2(2.0, 0.0) * pix_size) * 0.131514; + color += texture(source_color, uv_interp + vec2(3.0, 0.0) * pix_size) * 0.071303; + color += texture(source_color, uv_interp + vec2(-1.0, 0.0) * pix_size) * 0.189879; + color += texture(source_color, uv_interp + vec2(-2.0, 0.0) * pix_size) * 0.131514; + color += texture(source_color, uv_interp + vec2(-3.0, 0.0) * pix_size) * 0.071303; + frag_color = color; + } else { + vec2 pix_size = blur.pixel_size; + vec4 color = texture(source_color, uv_interp + vec2(0.0, 0.0) * pix_size) * 0.38774; + color += texture(source_color, uv_interp + vec2(0.0, 1.0) * pix_size) * 0.24477; + color += texture(source_color, uv_interp + vec2(0.0, 2.0) * pix_size) * 0.06136; + color += texture(source_color, uv_interp + vec2(0.0, -1.0) * pix_size) * 0.24477; + color += texture(source_color, uv_interp + vec2(0.0, -2.0) * pix_size) * 0.06136; + frag_color = color; + } +#endif + +#ifdef MODE_GAUSSIAN_GLOW + + //Glow uses larger sigma 1 for a more rounded blur effect + +#define GLOW_ADD(m_ofs, m_mult) \ + { \ + vec2 ofs = uv_interp + m_ofs * pix_size; \ + vec4 c = texture(source_color, ofs) * m_mult; \ + if (any(lessThan(ofs, vec2(0.0))) || any(greaterThan(ofs, vec2(1.0)))) { \ + c *= 0.0; \ + } \ + color += c; \ + } + + if (bool(blur.flags & FLAG_HORIZONTAL)) { + vec2 pix_size = blur.pixel_size; + pix_size *= 0.5; //reading from larger buffer, so use more samples + + vec4 color = texture(source_color, uv_interp + vec2(0.0, 0.0) * pix_size) * 0.174938; + GLOW_ADD(vec2(1.0, 0.0), 0.165569); + GLOW_ADD(vec2(2.0, 0.0), 0.140367); + GLOW_ADD(vec2(3.0, 0.0), 0.106595); + GLOW_ADD(vec2(-1.0, 0.0), 0.165569); + GLOW_ADD(vec2(-2.0, 0.0), 0.140367); + GLOW_ADD(vec2(-3.0, 0.0), 0.106595); + + // only do this in the horizontal pass, if we also do this in the vertical pass we're doubling up. + color *= blur.glow_strength; + + frag_color = color; + } else { + vec2 pix_size = blur.pixel_size; + vec4 color = texture(source_color, uv_interp + vec2(0.0, 0.0) * pix_size) * 0.288713; + GLOW_ADD(vec2(0.0, 1.0), 0.233062); + GLOW_ADD(vec2(0.0, 2.0), 0.122581); + GLOW_ADD(vec2(0.0, -1.0), 0.233062); + GLOW_ADD(vec2(0.0, -2.0), 0.122581); + + frag_color = color; + } + +#undef GLOW_ADD + + if (bool(blur.flags & FLAG_GLOW_FIRST_PASS)) { + // In the first pass bring back to correct color range else we're applying the wrong threshold + // in subsequent passes we can use it as is as we'd just be undoing it right after. + frag_color *= blur.luminance_multiplier; + +#ifdef GLOW_USE_AUTO_EXPOSURE + + frag_color /= texelFetch(source_auto_exposure, ivec2(0, 0), 0).r / blur.glow_auto_exposure_grey; +#endif + frag_color *= blur.glow_exposure; + + float luminance = max(frag_color.r, max(frag_color.g, frag_color.b)); + float feedback = max(smoothstep(blur.glow_hdr_threshold, blur.glow_hdr_threshold + blur.glow_hdr_scale, luminance), blur.glow_bloom); + + frag_color = min(frag_color * feedback, vec4(blur.glow_luminance_cap)) / blur.luminance_multiplier; + } + +#endif // MODE_GAUSSIAN_GLOW + +#ifdef MODE_COPY + vec4 color = textureLod(source_color, uv_interp, 0.0); + frag_color = color; +#endif +} diff --git a/servers/rendering/renderer_rd/shaders/effects/blur_raster_inc.glsl b/servers/rendering/renderer_rd/shaders/effects/blur_raster_inc.glsl new file mode 100644 index 0000000000..730504571a --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/effects/blur_raster_inc.glsl @@ -0,0 +1,26 @@ +#define FLAG_HORIZONTAL (1 << 0) +#define FLAG_USE_ORTHOGONAL_PROJECTION (1 << 1) +#define FLAG_GLOW_FIRST_PASS (1 << 2) + +layout(push_constant, std430) uniform Blur { + vec2 pixel_size; // 08 - 08 + uint flags; // 04 - 12 + uint pad; // 04 - 16 + + // Glow. + float glow_strength; // 04 - 20 + float glow_bloom; // 04 - 24 + float glow_hdr_threshold; // 04 - 28 + float glow_hdr_scale; // 04 - 32 + + float glow_exposure; // 04 - 36 + float glow_white; // 04 - 40 + float glow_luminance_cap; // 04 - 44 + float glow_auto_exposure_grey; // 04 - 48 + + float luminance_multiplier; // 04 - 52 + float res1; // 04 - 56 + float res2; // 04 - 60 + float res3; // 04 - 64 +} +blur; diff --git a/servers/rendering/renderer_rd/shaders/effects/bokeh_dof.glsl b/servers/rendering/renderer_rd/shaders/effects/bokeh_dof.glsl new file mode 100644 index 0000000000..0438671dd2 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/effects/bokeh_dof.glsl @@ -0,0 +1,215 @@ +#[compute] + +#version 450 + +#VERSION_DEFINES + +#define BLOCK_SIZE 8 + +layout(local_size_x = BLOCK_SIZE, local_size_y = BLOCK_SIZE, local_size_z = 1) in; + +#ifdef MODE_GEN_BLUR_SIZE +layout(rgba16f, set = 0, binding = 0) uniform restrict image2D color_image; +layout(set = 1, binding = 0) uniform sampler2D source_depth; +#endif + +#if defined(MODE_BOKEH_BOX) || defined(MODE_BOKEH_HEXAGONAL) || defined(MODE_BOKEH_CIRCULAR) +layout(set = 1, binding = 0) uniform sampler2D color_texture; +layout(rgba16f, set = 0, binding = 0) uniform restrict writeonly image2D bokeh_image; +#endif + +#ifdef MODE_COMPOSITE_BOKEH +layout(rgba16f, set = 0, binding = 0) uniform restrict image2D color_image; +layout(set = 1, binding = 0) uniform sampler2D source_bokeh; +#endif + +// based on https://www.shadertoy.com/view/Xd3GDl + +#include "bokeh_dof_inc.glsl" + +#ifdef MODE_GEN_BLUR_SIZE + +float get_depth_at_pos(vec2 uv) { + float depth = textureLod(source_depth, uv, 0.0).x; + if (params.orthogonal) { + depth = ((depth + (params.z_far + params.z_near) / (params.z_far - params.z_near)) * (params.z_far - params.z_near)) / 2.0; + } else { + depth = 2.0 * params.z_near * params.z_far / (params.z_far + params.z_near - depth * (params.z_far - params.z_near)); + } + return depth; +} + +float get_blur_size(float depth) { + if (params.blur_near_active && depth < params.blur_near_begin) { + return -(1.0 - smoothstep(params.blur_near_end, params.blur_near_begin, depth)) * params.blur_size - DEPTH_GAP; //near blur is negative + } + + if (params.blur_far_active && depth > params.blur_far_begin) { + return smoothstep(params.blur_far_begin, params.blur_far_end, depth) * params.blur_size + DEPTH_GAP; + } + + return 0.0; +} + +#endif + +#if defined(MODE_BOKEH_BOX) || defined(MODE_BOKEH_HEXAGONAL) + +vec4 weighted_filter_dir(vec2 dir, vec2 uv, vec2 pixel_size) { + dir *= pixel_size; + vec4 color = texture(color_texture, uv); + + vec4 accum = color; + float total = 1.0; + + float blur_scale = params.blur_size / float(params.blur_steps); + + if (params.use_jitter) { + uv += dir * (hash12n(uv + params.jitter_seed) - 0.5); + } + + for (int i = -params.blur_steps; i <= params.blur_steps; i++) { + if (i == 0) { + continue; + } + float radius = float(i) * blur_scale; + vec2 suv = uv + dir * radius; + radius = abs(radius); + + vec4 sample_color = texture(color_texture, suv); + float limit; + + if (sample_color.a < color.a) { + limit = abs(sample_color.a); + } else { + limit = abs(color.a); + } + + limit -= DEPTH_GAP; + + float m = smoothstep(radius - 0.5, radius + 0.5, limit); + + accum += mix(color, sample_color, m); + + total += 1.0; + } + + return accum / total; +} + +#endif + +void main() { + ivec2 pos = ivec2(gl_GlobalInvocationID.xy); + + if (any(greaterThan(pos, params.size))) { //too large, do nothing + return; + } + + vec2 pixel_size = 1.0 / vec2(params.size); + vec2 uv = vec2(pos) / vec2(params.size); + +#ifdef MODE_GEN_BLUR_SIZE + uv += pixel_size * 0.5; + //precompute size in alpha channel + float depth = get_depth_at_pos(uv); + float size = get_blur_size(depth); + + vec4 color = imageLoad(color_image, pos); + color.a = size; + imageStore(color_image, pos, color); +#endif + +#ifdef MODE_BOKEH_BOX + + //pixel_size*=0.5; //resolution is doubled + if (params.second_pass || !params.half_size) { + uv += pixel_size * 0.5; //half pixel to read centers + } else { + uv += pixel_size * 0.25; //half pixel to read centers from full res + } + + vec2 dir = (params.second_pass ? vec2(0.0, 1.0) : vec2(1.0, 0.0)); + + vec4 color = weighted_filter_dir(dir, uv, pixel_size); + + imageStore(bokeh_image, pos, color); + +#endif + +#ifdef MODE_BOKEH_HEXAGONAL + + //pixel_size*=0.5; //resolution is doubled + if (params.second_pass || !params.half_size) { + uv += pixel_size * 0.5; //half pixel to read centers + } else { + uv += pixel_size * 0.25; //half pixel to read centers from full res + } + + vec2 dir = (params.second_pass ? normalize(vec2(1.0, 0.577350269189626)) : vec2(0.0, 1.0)); + + vec4 color = weighted_filter_dir(dir, uv, pixel_size); + + if (params.second_pass) { + dir = normalize(vec2(-1.0, 0.577350269189626)); + + vec4 color2 = weighted_filter_dir(dir, uv, pixel_size); + + color.rgb = min(color.rgb, color2.rgb); + color.a = (color.a + color2.a) * 0.5; + } + + imageStore(bokeh_image, pos, color); + +#endif + +#ifdef MODE_BOKEH_CIRCULAR + + if (params.half_size) { + pixel_size *= 0.5; //resolution is doubled + } + + uv += pixel_size * 0.5; //half pixel to read centers + + vec4 color = texture(color_texture, uv); + float accum = 1.0; + float radius = params.blur_scale; + + for (float ang = 0.0; radius < params.blur_size; ang += GOLDEN_ANGLE) { + vec2 suv = uv + vec2(cos(ang), sin(ang)) * pixel_size * radius; + vec4 sample_color = texture(color_texture, suv); + float sample_size = abs(sample_color.a); + if (sample_color.a > color.a) { + sample_size = clamp(sample_size, 0.0, abs(color.a) * 2.0); + } + + float m = smoothstep(radius - 0.5, radius + 0.5, sample_size); + color += mix(color / accum, sample_color, m); + accum += 1.0; + radius += params.blur_scale / radius; + } + + color /= accum; + + imageStore(bokeh_image, pos, color); +#endif + +#ifdef MODE_COMPOSITE_BOKEH + + uv += pixel_size * 0.5; + vec4 color = imageLoad(color_image, pos); + vec4 bokeh = texture(source_bokeh, uv); + + float mix_amount; + if (bokeh.a < color.a) { + mix_amount = min(1.0, max(0.0, max(abs(color.a), abs(bokeh.a)) - DEPTH_GAP)); + } else { + mix_amount = min(1.0, max(0.0, abs(color.a) - DEPTH_GAP)); + } + + color.rgb = mix(color.rgb, bokeh.rgb, mix_amount); //blend between hires and lowres + + color.a = 0; //reset alpha + imageStore(color_image, pos, color); +#endif +} diff --git a/servers/rendering/renderer_rd/shaders/effects/bokeh_dof_inc.glsl b/servers/rendering/renderer_rd/shaders/effects/bokeh_dof_inc.glsl new file mode 100644 index 0000000000..b90a527554 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/effects/bokeh_dof_inc.glsl @@ -0,0 +1,37 @@ +layout(push_constant, std430) uniform Params { + ivec2 size; + float z_far; + float z_near; + + bool orthogonal; + float blur_size; + float blur_scale; + int blur_steps; + + bool blur_near_active; + float blur_near_begin; + float blur_near_end; + bool blur_far_active; + + float blur_far_begin; + float blur_far_end; + bool second_pass; + bool half_size; + + bool use_jitter; + float jitter_seed; + uint pad[2]; +} +params; + +//used to work around downsampling filter +#define DEPTH_GAP 0.0 + +const float GOLDEN_ANGLE = 2.39996323; + +//note: uniform pdf rand [0;1[ +float hash12n(vec2 p) { + p = fract(p * vec2(5.3987, 5.4421)); + p += dot(p.yx, p.xy + vec2(21.5351, 14.3137)); + return fract(p.x * p.y * 95.4307); +} diff --git a/servers/rendering/renderer_rd/shaders/effects/bokeh_dof_raster.glsl b/servers/rendering/renderer_rd/shaders/effects/bokeh_dof_raster.glsl new file mode 100644 index 0000000000..a3b3938ee9 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/effects/bokeh_dof_raster.glsl @@ -0,0 +1,253 @@ +/* clang-format off */ +#[vertex] + +#version 450 + +#VERSION_DEFINES + +#include "bokeh_dof_inc.glsl" + +layout(location = 0) out vec2 uv_interp; +/* clang-format on */ + +void main() { + vec2 base_arr[4] = vec2[](vec2(0.0, 0.0), vec2(0.0, 1.0), vec2(1.0, 1.0), vec2(1.0, 0.0)); + uv_interp = base_arr[gl_VertexIndex]; + + gl_Position = vec4(uv_interp * 2.0 - 1.0, 0.0, 1.0); +} + +/* clang-format off */ +#[fragment] + +#version 450 + +#VERSION_DEFINES + +#include "bokeh_dof_inc.glsl" + +layout(location = 0) in vec2 uv_interp; +/* clang-format on */ + +#ifdef MODE_GEN_BLUR_SIZE +layout(location = 0) out float weight; + +layout(set = 0, binding = 0) uniform sampler2D source_depth; +#else +layout(location = 0) out vec4 frag_color; +#ifdef OUTPUT_WEIGHT +layout(location = 1) out float weight; +#endif + +layout(set = 0, binding = 0) uniform sampler2D source_color; +layout(set = 1, binding = 0) uniform sampler2D source_weight; +#ifdef MODE_COMPOSITE_BOKEH +layout(set = 2, binding = 0) uniform sampler2D original_weight; +#endif +#endif + +//DOF +// Bokeh single pass implementation based on https://tuxedolabs.blogspot.com/2018/05/bokeh-depth-of-field-in-single-pass.html + +#ifdef MODE_GEN_BLUR_SIZE + +float get_depth_at_pos(vec2 uv) { + float depth = textureLod(source_depth, uv, 0.0).x; + if (params.orthogonal) { + depth = ((depth + (params.z_far + params.z_near) / (params.z_far - params.z_near)) * (params.z_far - params.z_near)) / 2.0; + } else { + depth = 2.0 * params.z_near * params.z_far / (params.z_far + params.z_near - depth * (params.z_far - params.z_near)); + } + return depth; +} + +float get_blur_size(float depth) { + if (params.blur_near_active && depth < params.blur_near_begin) { + return -(1.0 - smoothstep(params.blur_near_end, params.blur_near_begin, depth)) * params.blur_size - DEPTH_GAP; //near blur is negative + } + + if (params.blur_far_active && depth > params.blur_far_begin) { + return smoothstep(params.blur_far_begin, params.blur_far_end, depth) * params.blur_size + DEPTH_GAP; + } + + return 0.0; +} + +#endif + +#if defined(MODE_BOKEH_BOX) || defined(MODE_BOKEH_HEXAGONAL) + +vec4 weighted_filter_dir(vec2 dir, vec2 uv, vec2 pixel_size) { + dir *= pixel_size; + vec4 color = texture(source_color, uv); + color.a = texture(source_weight, uv).r; + + vec4 accum = color; + float total = 1.0; + + float blur_scale = params.blur_size / float(params.blur_steps); + + if (params.use_jitter) { + uv += dir * (hash12n(uv + params.jitter_seed) - 0.5); + } + + for (int i = -params.blur_steps; i <= params.blur_steps; i++) { + if (i == 0) { + continue; + } + float radius = float(i) * blur_scale; + vec2 suv = uv + dir * radius; + radius = abs(radius); + + vec4 sample_color = texture(source_color, suv); + sample_color.a = texture(source_weight, suv).r; + float limit; + + if (sample_color.a < color.a) { + limit = abs(sample_color.a); + } else { + limit = abs(color.a); + } + + limit -= DEPTH_GAP; + + float m = smoothstep(radius - 0.5, radius + 0.5, limit); + + accum += mix(color, sample_color, m); + + total += 1.0; + } + + return accum / total; +} + +#endif + +void main() { + vec2 pixel_size = 1.0 / vec2(params.size); + vec2 uv = uv_interp; + +#ifdef MODE_GEN_BLUR_SIZE + uv += pixel_size * 0.5; + float center_depth = get_depth_at_pos(uv); + weight = get_blur_size(center_depth); +#endif + +#ifdef MODE_BOKEH_BOX + //pixel_size*=0.5; //resolution is doubled + if (params.second_pass || !params.half_size) { + uv += pixel_size * 0.5; //half pixel to read centers + } else { + uv += pixel_size * 0.25; //half pixel to read centers from full res + } + + float alpha = texture(source_color, uv).a; // retain this + vec2 dir = (params.second_pass ? vec2(0.0, 1.0) : vec2(1.0, 0.0)); + + vec4 color = weighted_filter_dir(dir, uv, pixel_size); + + frag_color = color; + frag_color.a = alpha; // attempt to retain this in case we have a transparent background, ignored if half_size +#ifdef OUTPUT_WEIGHT + weight = color.a; +#endif + +#endif + +#ifdef MODE_BOKEH_HEXAGONAL + + //pixel_size*=0.5; //resolution is doubled + if (params.second_pass || !params.half_size) { + uv += pixel_size * 0.5; //half pixel to read centers + } else { + uv += pixel_size * 0.25; //half pixel to read centers from full res + } + + float alpha = texture(source_color, uv).a; // retain this + + vec2 dir = (params.second_pass ? normalize(vec2(1.0, 0.577350269189626)) : vec2(0.0, 1.0)); + + vec4 color = weighted_filter_dir(dir, uv, pixel_size); + + if (params.second_pass) { + dir = normalize(vec2(-1.0, 0.577350269189626)); + + vec4 color2 = weighted_filter_dir(dir, uv, pixel_size); + + color.rgb = min(color.rgb, color2.rgb); + color.a = (color.a + color2.a) * 0.5; + } + + frag_color = color; + frag_color.a = alpha; // attempt to retain this in case we have a transparent background, ignored if half_size +#ifdef OUTPUT_WEIGHT + weight = color.a; +#endif + +#endif + +#ifdef MODE_BOKEH_CIRCULAR + if (params.half_size) { + pixel_size *= 0.5; //resolution is doubled + } + + uv += pixel_size * 0.5; //half pixel to read centers + + vec4 color = texture(source_color, uv); + float alpha = color.a; // retain this + color.a = texture(source_weight, uv).r; + + vec4 color_accum = color; + float accum = 1.0; + + float radius = params.blur_scale; + for (float ang = 0.0; radius < params.blur_size; ang += GOLDEN_ANGLE) { + vec2 uv_adj = uv + vec2(cos(ang), sin(ang)) * pixel_size * radius; + + vec4 sample_color = texture(source_color, uv_adj); + sample_color.a = texture(source_weight, uv_adj).r; + + float limit; + + if (sample_color.a < color.a) { + limit = abs(sample_color.a); + } else { + limit = abs(color.a); + } + + limit -= DEPTH_GAP; + + float m = smoothstep(radius - 0.5, radius + 0.5, limit); + color_accum += mix(color_accum / accum, sample_color, m); + accum += 1.0; + + radius += params.blur_scale / radius; + } + + color_accum = color_accum / accum; + + frag_color.rgb = color_accum.rgb; + frag_color.a = alpha; // attempt to retain this in case we have a transparent background, ignored if half_size +#ifdef OUTPUT_WEIGHT + weight = color_accum.a; +#endif + +#endif + +#ifdef MODE_COMPOSITE_BOKEH + frag_color.rgb = texture(source_color, uv).rgb; + + float center_weigth = texture(source_weight, uv).r; + float sample_weight = texture(original_weight, uv).r; + + float mix_amount; + if (sample_weight < center_weigth) { + mix_amount = min(1.0, max(0.0, max(abs(center_weigth), abs(sample_weight)) - DEPTH_GAP)); + } else { + mix_amount = min(1.0, max(0.0, abs(center_weigth) - DEPTH_GAP)); + } + + // let alpha blending take care of mixing + frag_color.a = mix_amount; +#endif +} diff --git a/servers/rendering/renderer_rd/shaders/effects/copy.glsl b/servers/rendering/renderer_rd/shaders/effects/copy.glsl new file mode 100644 index 0000000000..3a4ef86ef0 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/effects/copy.glsl @@ -0,0 +1,284 @@ +#[compute] + +#version 450 + +#VERSION_DEFINES + +layout(local_size_x = 8, local_size_y = 8, local_size_z = 1) in; + +#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) +#define FLAG_HIGH_QUALITY_GLOW (1 << 8) +#define FLAG_ALPHA_TO_ONE (1 << 9) + +layout(push_constant, 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]; + + vec4 set_color; +} +params; + +#ifdef MODE_CUBEMAP_ARRAY_TO_PANORAMA +layout(set = 0, binding = 0) uniform samplerCubeArray source_color; +#elif defined(MODE_CUBEMAP_TO_PANORAMA) +layout(set = 0, binding = 0) uniform samplerCube source_color; +#elif !defined(MODE_SET_COLOR) +layout(set = 0, binding = 0) uniform sampler2D source_color; +#endif + +#ifdef GLOW_USE_AUTO_EXPOSURE +layout(set = 1, binding = 0) uniform sampler2D source_auto_exposure; +#endif + +#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 + +#ifdef MODE_GAUSSIAN_BLUR +shared vec4 local_cache[256]; +shared vec4 temp_cache[128]; +#endif + +void main() { + // Pixel being shaded + ivec2 pos = ivec2(gl_GlobalInvocationID.xy); + +#ifndef MODE_GAUSSIAN_BLUR // Gaussian blur needs the extra threads + if (any(greaterThanEqual(pos, params.section.zw))) { //too large, do nothing + return; + } +#endif + +#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; + color = mix(color, vec4(100.0, 100.0, 100.0, 1.0), isinf(color)); + color = mix(color, vec4(100.0, 100.0, 100.0, 1.0), isnan(color)); + + imageStore(dest_buffer, pos + params.target, color); +#endif + +#ifdef MODE_GAUSSIAN_BLUR + + // First pass copy texture into 16x16 local memory for every 8x8 thread block + vec2 quad_center_uv = clamp(vec2(gl_GlobalInvocationID.xy + gl_LocalInvocationID.xy - 3.5) / params.section.zw, vec2(0.5 / params.section.zw), vec2(1.0 - 1.5 / params.section.zw)); + uint dest_index = gl_LocalInvocationID.x * 2 + gl_LocalInvocationID.y * 2 * 16; + +#ifdef MODE_GLOW + if (bool(params.flags & FLAG_HIGH_QUALITY_GLOW)) { + vec2 quad_offset_uv = clamp((vec2(gl_GlobalInvocationID.xy + gl_LocalInvocationID.xy - 3.0)) / params.section.zw, vec2(0.5 / params.section.zw), vec2(1.0 - 1.5 / params.section.zw)); + + local_cache[dest_index] = (textureLod(source_color, quad_center_uv, 0) + textureLod(source_color, quad_offset_uv, 0)) * 0.5; + local_cache[dest_index + 1] = (textureLod(source_color, quad_center_uv + vec2(1.0 / params.section.z, 0.0), 0) + textureLod(source_color, quad_offset_uv + vec2(1.0 / params.section.z, 0.0), 0)) * 0.5; + local_cache[dest_index + 16] = (textureLod(source_color, quad_center_uv + vec2(0.0, 1.0 / params.section.w), 0) + textureLod(source_color, quad_offset_uv + vec2(0.0, 1.0 / params.section.w), 0)) * 0.5; + local_cache[dest_index + 16 + 1] = (textureLod(source_color, quad_center_uv + vec2(1.0 / params.section.zw), 0) + textureLod(source_color, quad_offset_uv + vec2(1.0 / params.section.zw), 0)) * 0.5; + } else +#endif + { + local_cache[dest_index] = textureLod(source_color, quad_center_uv, 0); + local_cache[dest_index + 1] = textureLod(source_color, quad_center_uv + vec2(1.0 / params.section.z, 0.0), 0); + local_cache[dest_index + 16] = textureLod(source_color, quad_center_uv + vec2(0.0, 1.0 / params.section.w), 0); + local_cache[dest_index + 16 + 1] = textureLod(source_color, quad_center_uv + vec2(1.0 / params.section.zw), 0); + } +#ifdef MODE_GLOW + if (bool(params.flags & FLAG_GLOW_FIRST_PASS)) { + // Tonemap initial samples to reduce weight of fireflies: https://graphicrants.blogspot.com/2013/12/tone-mapping.html + local_cache[dest_index] /= 1.0 + dot(local_cache[dest_index].rgb, vec3(0.299, 0.587, 0.114)); + local_cache[dest_index + 1] /= 1.0 + dot(local_cache[dest_index + 1].rgb, vec3(0.299, 0.587, 0.114)); + local_cache[dest_index + 16] /= 1.0 + dot(local_cache[dest_index + 16].rgb, vec3(0.299, 0.587, 0.114)); + local_cache[dest_index + 16 + 1] /= 1.0 + dot(local_cache[dest_index + 16 + 1].rgb, vec3(0.299, 0.587, 0.114)); + } + const float kernel[4] = { 0.174938, 0.165569, 0.140367, 0.106595 }; +#else + // Simpler blur uses SIGMA2 for the gaussian kernel for a stronger effect. + const float kernel[4] = { 0.214607, 0.189879, 0.131514, 0.071303 }; +#endif + memoryBarrierShared(); + barrier(); + + // Horizontal pass. Needs to copy into 8x16 chunk of local memory so vertical pass has full resolution + uint read_index = gl_LocalInvocationID.x + gl_LocalInvocationID.y * 32 + 4; + vec4 color_top = vec4(0.0); + color_top += local_cache[read_index] * kernel[0]; + color_top += local_cache[read_index + 1] * kernel[1]; + color_top += local_cache[read_index + 2] * kernel[2]; + color_top += local_cache[read_index + 3] * kernel[3]; + color_top += local_cache[read_index - 1] * kernel[1]; + color_top += local_cache[read_index - 2] * kernel[2]; + color_top += local_cache[read_index - 3] * kernel[3]; + + vec4 color_bottom = vec4(0.0); + color_bottom += local_cache[read_index + 16] * kernel[0]; + color_bottom += local_cache[read_index + 1 + 16] * kernel[1]; + color_bottom += local_cache[read_index + 2 + 16] * kernel[2]; + color_bottom += local_cache[read_index + 3 + 16] * kernel[3]; + color_bottom += local_cache[read_index - 1 + 16] * kernel[1]; + color_bottom += local_cache[read_index - 2 + 16] * kernel[2]; + color_bottom += local_cache[read_index - 3 + 16] * kernel[3]; + + // rotate samples to take advantage of cache coherency + uint write_index = gl_LocalInvocationID.y * 2 + gl_LocalInvocationID.x * 16; + + temp_cache[write_index] = color_top; + temp_cache[write_index + 1] = color_bottom; + + memoryBarrierShared(); + barrier(); + + // If destination outside of texture, can stop doing work now + if (any(greaterThanEqual(pos, params.section.zw))) { + return; + } + + // Vertical pass + uint index = gl_LocalInvocationID.y + gl_LocalInvocationID.x * 16 + 4; + vec4 color = vec4(0.0); + + color += temp_cache[index] * kernel[0]; + color += temp_cache[index + 1] * kernel[1]; + color += temp_cache[index + 2] * kernel[2]; + color += temp_cache[index + 3] * kernel[3]; + color += temp_cache[index - 1] * kernel[1]; + color += temp_cache[index - 2] * kernel[2]; + color += temp_cache[index - 3] * kernel[3]; + +#ifdef MODE_GLOW + if (bool(params.flags & FLAG_GLOW_FIRST_PASS)) { + // Undo tonemap to restore range: https://graphicrants.blogspot.com/2013/12/tone-mapping.html + color /= 1.0 - dot(color.rgb, vec3(0.299, 0.587, 0.114)); + } + + color *= params.glow_strength; + + 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)); + } +#endif + 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); + + } else { + color = texelFetch(source_color, pos + params.section.xy, 0); + + if (bool(params.flags & FLAG_FLIP_Y)) { + pos.y = params.section.w - pos.y - 1; + } + } + + if (bool(params.flags & FLAG_FORCE_LUMINANCE)) { + color.rgb = vec3(max(max(color.r, color.g), color.b)); + } + + if (bool(params.flags & FLAG_ALPHA_TO_ONE)) { + color.a = 1.0; + } + + 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; + } + + 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); + + if (bool(params.flags & FLAG_FLIP_Y)) { + pos.y = params.section.w - pos.y - 1; + } + + imageStore(dest_buffer, pos + params.target, color); +#endif + +#if defined(MODE_CUBEMAP_TO_PANORAMA) || defined(MODE_CUBEMAP_ARRAY_TO_PANORAMA) + + const float PI = 3.14159265359; + vec2 uv = vec2(pos) / vec2(params.section.zw); + if (bool(params.flags & FLAG_FLIP_Y)) { + uv.y = 1.0 - uv.y; + } + float phi = uv.x * 2.0 * PI; + float theta = uv.y * PI; + + vec3 normal; + normal.x = sin(phi) * sin(theta) * -1.0; + normal.y = cos(theta); + normal.z = cos(phi) * sin(theta) * -1.0; + +#ifdef MODE_CUBEMAP_TO_PANORAMA + vec4 color = textureLod(source_color, normal, params.camera_z_far); //the biggest the lod the least the acne +#else + vec4 color = textureLod(source_color, vec4(normal, params.camera_z_far), 0.0); //the biggest the lod the least the acne +#endif + imageStore(dest_buffer, pos + params.target, color); +#endif + +#ifdef MODE_SET_COLOR + imageStore(dest_buffer, pos + params.target, params.set_color); +#endif +} diff --git a/servers/rendering/renderer_rd/shaders/effects/copy_to_fb.glsl b/servers/rendering/renderer_rd/shaders/effects/copy_to_fb.glsl new file mode 100644 index 0000000000..9787c9879d --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/effects/copy_to_fb.glsl @@ -0,0 +1,167 @@ +#[vertex] + +#version 450 + +#VERSION_DEFINES + +#ifdef MULTIVIEW +#ifdef has_VK_KHR_multiview +#extension GL_EXT_multiview : enable +#define ViewIndex gl_ViewIndex +#else // has_VK_KHR_multiview +#define ViewIndex 0 +#endif // has_VK_KHR_multiview +#endif //MULTIVIEW + +#ifdef MULTIVIEW +layout(location = 0) out vec3 uv_interp; +#else +layout(location = 0) out vec2 uv_interp; +#endif + +layout(push_constant, 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.xy = base_arr[gl_VertexIndex]; +#ifdef MULTIVIEW + uv_interp.z = ViewIndex; +#endif + vec2 vpos = uv_interp.xy; + if (params.use_section) { + vpos = params.section.xy + vpos * params.section.zw; + } + + gl_Position = vec4(vpos * 2.0 - 1.0, 0.0, 1.0); + + if (params.flip_y) { + uv_interp.y = 1.0 - uv_interp.y; + } +} + +#[fragment] + +#version 450 + +#VERSION_DEFINES + +#ifdef MULTIVIEW +#ifdef has_VK_KHR_multiview +#extension GL_EXT_multiview : enable +#define ViewIndex gl_ViewIndex +#else // has_VK_KHR_multiview +#define ViewIndex 0 +#endif // has_VK_KHR_multiview +#endif //MULTIVIEW + +layout(push_constant, std430) uniform Params { + vec4 section; + vec2 pixel_size; + bool flip_y; + bool use_section; + + bool force_luminance; + bool alpha_to_zero; + bool srgb; + uint pad; +} +params; + +#ifdef MULTIVIEW +layout(location = 0) in vec3 uv_interp; +#else +layout(location = 0) in vec2 uv_interp; +#endif + +#ifdef MULTIVIEW +layout(set = 0, binding = 0) uniform sampler2DArray source_color; +#ifdef MODE_TWO_SOURCES +layout(set = 1, binding = 0) uniform sampler2DArray source_depth; +layout(location = 1) out float depth; +#endif /* MODE_TWO_SOURCES */ +#else +layout(set = 0, binding = 0) uniform sampler2D source_color; +#ifdef MODE_TWO_SOURCES +layout(set = 1, binding = 0) uniform sampler2D source_color2; +#endif /* MODE_TWO_SOURCES */ +#endif /* MULTIVIEW */ + +layout(location = 0) out vec4 frag_color; + +vec3 linear_to_srgb(vec3 color) { + //if going to srgb, clamp from 0 to 1. + color = clamp(color, vec3(0.0), vec3(1.0)); + const vec3 a = vec3(0.055f); + return mix((vec3(1.0f) + a) * pow(color.rgb, vec3(1.0f / 2.4f)) - a, 12.92f * color.rgb, lessThan(color.rgb, vec3(0.0031308f))); +} + +void main() { +#ifdef MULTIVIEW + vec3 uv = uv_interp; +#else + vec2 uv = uv_interp; +#endif + +#ifdef MODE_PANORAMA_TO_DP + // Note, multiview and panorama should not be mixed at this time + + //obtain normal from dual paraboloid uv +#define M_PI 3.14159265359 + + 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 + +#ifdef MULTIVIEW + vec4 color = textureLod(source_color, uv, 0.0); +#ifdef MODE_TWO_SOURCES + // In multiview our 2nd input will be our depth map + depth = textureLod(source_depth, uv, 0.0).r; +#endif /* MODE_TWO_SOURCES */ + +#else + vec4 color = textureLod(source_color, uv, 0.0); +#ifdef MODE_TWO_SOURCES + color += textureLod(source_color2, uv, 0.0); +#endif /* MODE_TWO_SOURCES */ +#endif /* MULTIVIEW */ + if (params.force_luminance) { + color.rgb = vec3(max(max(color.r, color.g), color.b)); + } + if (params.alpha_to_zero) { + color.rgb *= color.a; + } + if (params.srgb) { + color.rgb = linear_to_srgb(color.rgb); + } + frag_color = color; +} diff --git a/servers/rendering/renderer_rd/shaders/effects/tonemap.glsl b/servers/rendering/renderer_rd/shaders/effects/tonemap.glsl new file mode 100644 index 0000000000..62a7b0e7d7 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/effects/tonemap.glsl @@ -0,0 +1,502 @@ +#[vertex] + +#version 450 + +#VERSION_DEFINES + +#ifdef MULTIVIEW +#ifdef has_VK_KHR_multiview +#extension GL_EXT_multiview : enable +#endif +#endif + +layout(location = 0) out vec2 uv_interp; + +void main() { + 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); +} + +#[fragment] + +#version 450 + +#VERSION_DEFINES + +#ifdef MULTIVIEW +#ifdef has_VK_KHR_multiview +#extension GL_EXT_multiview : enable +#define ViewIndex gl_ViewIndex +#else // has_VK_KHR_multiview +#define ViewIndex 0 +#endif // has_VK_KHR_multiview +#endif //MULTIVIEW + +layout(location = 0) in vec2 uv_interp; + +#ifdef SUBPASS +layout(input_attachment_index = 0, set = 0, binding = 0) uniform subpassInput input_color; +#elif defined(MULTIVIEW) +layout(set = 0, binding = 0) uniform sampler2DArray source_color; +#else +layout(set = 0, binding = 0) uniform sampler2D source_color; +#endif + +layout(set = 1, binding = 0) uniform sampler2D source_auto_exposure; +#ifdef MULTIVIEW +layout(set = 2, binding = 0) uniform sampler2DArray source_glow; +#else +layout(set = 2, binding = 0) uniform sampler2D source_glow; +#endif +layout(set = 2, binding = 1) uniform sampler2D glow_map; + +#ifdef USE_1D_LUT +layout(set = 3, binding = 0) uniform sampler2D source_color_correction; +#else +layout(set = 3, binding = 0) uniform sampler3D source_color_correction; +#endif + +layout(push_constant, std430) uniform Params { + vec3 bcs; + bool use_bcs; + + bool use_glow; + bool use_auto_exposure; + bool use_color_correction; + uint tonemapper; + + uvec2 glow_texture_size; + float glow_intensity; + float glow_map_strength; + + uint glow_mode; + float glow_levels[7]; + + float exposure; + float white; + float auto_exposure_grey; + float luminance_multiplier; + + vec2 pixel_size; + bool use_fxaa; + bool use_debanding; +} +params; + +layout(location = 0) out vec4 frag_color; + +#ifdef USE_GLOW_FILTER_BICUBIC +// w0, w1, w2, and w3 are the four cubic B-spline basis functions +float w0(float a) { + return (1.0f / 6.0f) * (a * (a * (-a + 3.0f) - 3.0f) + 1.0f); +} + +float w1(float a) { + return (1.0f / 6.0f) * (a * a * (3.0f * a - 6.0f) + 4.0f); +} + +float w2(float a) { + return (1.0f / 6.0f) * (a * (a * (-3.0f * a + 3.0f) + 3.0f) + 1.0f); +} + +float w3(float a) { + return (1.0f / 6.0f) * (a * a * a); +} + +// g0 and g1 are the two amplitude functions +float g0(float a) { + return w0(a) + w1(a); +} + +float g1(float a) { + return w2(a) + w3(a); +} + +// h0 and h1 are the two offset functions +float h0(float a) { + return -1.0f + w1(a) / (w0(a) + w1(a)); +} + +float h1(float a) { + return 1.0f + w3(a) / (w2(a) + w3(a)); +} + +#ifdef MULTIVIEW +vec4 texture2D_bicubic(sampler2DArray tex, vec2 uv, int p_lod) { + float lod = float(p_lod); + vec2 tex_size = vec2(params.glow_texture_size >> p_lod); + vec2 pixel_size = vec2(1.0f) / tex_size; + + uv = uv * tex_size + vec2(0.5f); + + vec2 iuv = floor(uv); + vec2 fuv = fract(uv); + + float g0x = g0(fuv.x); + float g1x = g1(fuv.x); + float h0x = h0(fuv.x); + float h1x = h1(fuv.x); + float h0y = h0(fuv.y); + float h1y = h1(fuv.y); + + vec3 p0 = vec3((vec2(iuv.x + h0x, iuv.y + h0y) - vec2(0.5f)) * pixel_size, ViewIndex); + vec3 p1 = vec3((vec2(iuv.x + h1x, iuv.y + h0y) - vec2(0.5f)) * pixel_size, ViewIndex); + vec3 p2 = vec3((vec2(iuv.x + h0x, iuv.y + h1y) - vec2(0.5f)) * pixel_size, ViewIndex); + vec3 p3 = vec3((vec2(iuv.x + h1x, iuv.y + h1y) - vec2(0.5f)) * pixel_size, ViewIndex); + + return (g0(fuv.y) * (g0x * textureLod(tex, p0, lod) + g1x * textureLod(tex, p1, lod))) + + (g1(fuv.y) * (g0x * textureLod(tex, p2, lod) + g1x * textureLod(tex, p3, lod))); +} + +#define GLOW_TEXTURE_SAMPLE(m_tex, m_uv, m_lod) texture2D_bicubic(m_tex, m_uv, m_lod) +#else // MULTIVIEW + +vec4 texture2D_bicubic(sampler2D tex, vec2 uv, int p_lod) { + float lod = float(p_lod); + vec2 tex_size = vec2(params.glow_texture_size >> p_lod); + vec2 pixel_size = vec2(1.0f) / tex_size; + + uv = uv * tex_size + vec2(0.5f); + + vec2 iuv = floor(uv); + vec2 fuv = fract(uv); + + float g0x = g0(fuv.x); + float g1x = g1(fuv.x); + float h0x = h0(fuv.x); + float h1x = h1(fuv.x); + float h0y = h0(fuv.y); + float h1y = h1(fuv.y); + + vec2 p0 = (vec2(iuv.x + h0x, iuv.y + h0y) - vec2(0.5f)) * pixel_size; + vec2 p1 = (vec2(iuv.x + h1x, iuv.y + h0y) - vec2(0.5f)) * pixel_size; + vec2 p2 = (vec2(iuv.x + h0x, iuv.y + h1y) - vec2(0.5f)) * pixel_size; + vec2 p3 = (vec2(iuv.x + h1x, iuv.y + h1y) - vec2(0.5f)) * pixel_size; + + return (g0(fuv.y) * (g0x * textureLod(tex, p0, lod) + g1x * textureLod(tex, p1, lod))) + + (g1(fuv.y) * (g0x * textureLod(tex, p2, lod) + g1x * textureLod(tex, p3, lod))); +} + +#define GLOW_TEXTURE_SAMPLE(m_tex, m_uv, m_lod) texture2D_bicubic(m_tex, m_uv, m_lod) +#endif // !MULTIVIEW + +#else // USE_GLOW_FILTER_BICUBIC + +#ifdef MULTIVIEW +#define GLOW_TEXTURE_SAMPLE(m_tex, m_uv, m_lod) textureLod(m_tex, vec3(m_uv, ViewIndex), float(m_lod)) +#else // MULTIVIEW +#define GLOW_TEXTURE_SAMPLE(m_tex, m_uv, m_lod) textureLod(m_tex, m_uv, float(m_lod)) +#endif // !MULTIVIEW + +#endif // !USE_GLOW_FILTER_BICUBIC + +vec3 tonemap_filmic(vec3 color, float white) { + // exposure bias: input scale (color *= bias, white *= bias) to make the brightness consistent with other tonemappers + // also useful to scale the input to the range that the tonemapper is designed for (some require very high input values) + // has no effect on the curve's general shape or visual properties + const float exposure_bias = 2.0f; + const float A = 0.22f * exposure_bias * exposure_bias; // bias baked into constants for performance + const float B = 0.30f * exposure_bias; + const float C = 0.10f; + const float D = 0.20f; + const float E = 0.01f; + const float F = 0.30f; + + vec3 color_tonemapped = ((color * (A * color + C * B) + D * E) / (color * (A * color + B) + D * F)) - E / F; + float white_tonemapped = ((white * (A * white + C * B) + D * E) / (white * (A * white + B) + D * F)) - E / F; + + return color_tonemapped / white_tonemapped; +} + +// Adapted from https://github.com/TheRealMJP/BakingLab/blob/master/BakingLab/ACES.hlsl +// (MIT License). +vec3 tonemap_aces(vec3 color, float white) { + const float exposure_bias = 1.8f; + const float A = 0.0245786f; + const float B = 0.000090537f; + const float C = 0.983729f; + const float D = 0.432951f; + const float E = 0.238081f; + + // Exposure bias baked into transform to save shader instructions. Equivalent to `color *= exposure_bias` + const mat3 rgb_to_rrt = mat3( + vec3(0.59719f * exposure_bias, 0.35458f * exposure_bias, 0.04823f * exposure_bias), + vec3(0.07600f * exposure_bias, 0.90834f * exposure_bias, 0.01566f * exposure_bias), + vec3(0.02840f * exposure_bias, 0.13383f * exposure_bias, 0.83777f * exposure_bias)); + + const mat3 odt_to_rgb = mat3( + vec3(1.60475f, -0.53108f, -0.07367f), + vec3(-0.10208f, 1.10813f, -0.00605f), + vec3(-0.00327f, -0.07276f, 1.07602f)); + + color *= rgb_to_rrt; + vec3 color_tonemapped = (color * (color + A) - B) / (color * (C * color + D) + E); + color_tonemapped *= odt_to_rgb; + + white *= exposure_bias; + float white_tonemapped = (white * (white + A) - B) / (white * (C * white + D) + E); + + return color_tonemapped / white_tonemapped; +} + +vec3 tonemap_reinhard(vec3 color, float white) { + return (white * color + color) / (color * white + white); +} + +vec3 linear_to_srgb(vec3 color) { + //if going to srgb, clamp from 0 to 1. + color = clamp(color, vec3(0.0), vec3(1.0)); + const vec3 a = vec3(0.055f); + return mix((vec3(1.0f) + a) * pow(color.rgb, vec3(1.0f / 2.4f)) - a, 12.92f * color.rgb, lessThan(color.rgb, vec3(0.0031308f))); +} + +#define TONEMAPPER_LINEAR 0 +#define TONEMAPPER_REINHARD 1 +#define TONEMAPPER_FILMIC 2 +#define TONEMAPPER_ACES 3 + +vec3 apply_tonemapping(vec3 color, float white) { // inputs are LINEAR, always outputs clamped [0;1] color + // Ensure color values passed to tonemappers are positive. + // They can be negative in the case of negative lights, which leads to undesired behavior. + if (params.tonemapper == TONEMAPPER_LINEAR) { + return color; + } else if (params.tonemapper == TONEMAPPER_REINHARD) { + return tonemap_reinhard(max(vec3(0.0f), color), white); + } else if (params.tonemapper == TONEMAPPER_FILMIC) { + return tonemap_filmic(max(vec3(0.0f), color), white); + } else { // TONEMAPPER_ACES + return tonemap_aces(max(vec3(0.0f), color), white); + } +} + +#ifdef MULTIVIEW +vec3 gather_glow(sampler2DArray tex, vec2 uv) { // sample all selected glow levels, view is added to uv later +#else +vec3 gather_glow(sampler2D tex, vec2 uv) { // sample all selected glow levels +#endif // defined(MULTIVIEW) + vec3 glow = vec3(0.0f); + + if (params.glow_levels[0] > 0.0001) { + glow += GLOW_TEXTURE_SAMPLE(tex, uv, 0).rgb * params.glow_levels[0]; + } + + if (params.glow_levels[1] > 0.0001) { + glow += GLOW_TEXTURE_SAMPLE(tex, uv, 1).rgb * params.glow_levels[1]; + } + + if (params.glow_levels[2] > 0.0001) { + glow += GLOW_TEXTURE_SAMPLE(tex, uv, 2).rgb * params.glow_levels[2]; + } + + if (params.glow_levels[3] > 0.0001) { + glow += GLOW_TEXTURE_SAMPLE(tex, uv, 3).rgb * params.glow_levels[3]; + } + + if (params.glow_levels[4] > 0.0001) { + glow += GLOW_TEXTURE_SAMPLE(tex, uv, 4).rgb * params.glow_levels[4]; + } + + if (params.glow_levels[5] > 0.0001) { + glow += GLOW_TEXTURE_SAMPLE(tex, uv, 5).rgb * params.glow_levels[5]; + } + + if (params.glow_levels[6] > 0.0001) { + glow += GLOW_TEXTURE_SAMPLE(tex, uv, 6).rgb * params.glow_levels[6]; + } + + return glow; +} + +#define GLOW_MODE_ADD 0 +#define GLOW_MODE_SCREEN 1 +#define GLOW_MODE_SOFTLIGHT 2 +#define GLOW_MODE_REPLACE 3 +#define GLOW_MODE_MIX 4 + +vec3 apply_glow(vec3 color, vec3 glow) { // apply glow using the selected blending mode + if (params.glow_mode == GLOW_MODE_ADD) { + return color + glow; + } else if (params.glow_mode == GLOW_MODE_SCREEN) { + //need color clamping + return max((color + glow) - (color * glow), vec3(0.0)); + } else if (params.glow_mode == GLOW_MODE_SOFTLIGHT) { + //need color clamping + glow = glow * vec3(0.5f) + vec3(0.5f); + + color.r = (glow.r <= 0.5f) ? (color.r - (1.0f - 2.0f * glow.r) * color.r * (1.0f - color.r)) : (((glow.r > 0.5f) && (color.r <= 0.25f)) ? (color.r + (2.0f * glow.r - 1.0f) * (4.0f * color.r * (4.0f * color.r + 1.0f) * (color.r - 1.0f) + 7.0f * color.r)) : (color.r + (2.0f * glow.r - 1.0f) * (sqrt(color.r) - color.r))); + color.g = (glow.g <= 0.5f) ? (color.g - (1.0f - 2.0f * glow.g) * color.g * (1.0f - color.g)) : (((glow.g > 0.5f) && (color.g <= 0.25f)) ? (color.g + (2.0f * glow.g - 1.0f) * (4.0f * color.g * (4.0f * color.g + 1.0f) * (color.g - 1.0f) + 7.0f * color.g)) : (color.g + (2.0f * glow.g - 1.0f) * (sqrt(color.g) - color.g))); + color.b = (glow.b <= 0.5f) ? (color.b - (1.0f - 2.0f * glow.b) * color.b * (1.0f - color.b)) : (((glow.b > 0.5f) && (color.b <= 0.25f)) ? (color.b + (2.0f * glow.b - 1.0f) * (4.0f * color.b * (4.0f * color.b + 1.0f) * (color.b - 1.0f) + 7.0f * color.b)) : (color.b + (2.0f * glow.b - 1.0f) * (sqrt(color.b) - color.b))); + return color; + } else { //replace + return glow; + } +} + +vec3 apply_bcs(vec3 color, vec3 bcs) { + color = mix(vec3(0.0f), color, bcs.x); + color = mix(vec3(0.5f), color, bcs.y); + color = mix(vec3(dot(vec3(1.0f), color) * 0.33333f), color, bcs.z); + + return color; +} +#ifdef USE_1D_LUT +vec3 apply_color_correction(vec3 color) { + color.r = texture(source_color_correction, vec2(color.r, 0.0f)).r; + color.g = texture(source_color_correction, vec2(color.g, 0.0f)).g; + color.b = texture(source_color_correction, vec2(color.b, 0.0f)).b; + return color; +} +#else +vec3 apply_color_correction(vec3 color) { + return textureLod(source_color_correction, color, 0.0).rgb; +} +#endif + +#ifndef SUBPASS +vec3 do_fxaa(vec3 color, float exposure, vec2 uv_interp) { + const float FXAA_REDUCE_MIN = (1.0 / 128.0); + const float FXAA_REDUCE_MUL = (1.0 / 8.0); + const float FXAA_SPAN_MAX = 8.0; + +#ifdef MULTIVIEW + vec3 rgbNW = textureLod(source_color, vec3(uv_interp + vec2(-1.0, -1.0) * params.pixel_size, ViewIndex), 0.0).xyz * exposure * params.luminance_multiplier; + vec3 rgbNE = textureLod(source_color, vec3(uv_interp + vec2(1.0, -1.0) * params.pixel_size, ViewIndex), 0.0).xyz * exposure * params.luminance_multiplier; + vec3 rgbSW = textureLod(source_color, vec3(uv_interp + vec2(-1.0, 1.0) * params.pixel_size, ViewIndex), 0.0).xyz * exposure * params.luminance_multiplier; + vec3 rgbSE = textureLod(source_color, vec3(uv_interp + vec2(1.0, 1.0) * params.pixel_size, ViewIndex), 0.0).xyz * exposure * params.luminance_multiplier; +#else + vec3 rgbNW = textureLod(source_color, uv_interp + vec2(-1.0, -1.0) * params.pixel_size, 0.0).xyz * exposure * params.luminance_multiplier; + vec3 rgbNE = textureLod(source_color, uv_interp + vec2(1.0, -1.0) * params.pixel_size, 0.0).xyz * exposure * params.luminance_multiplier; + vec3 rgbSW = textureLod(source_color, uv_interp + vec2(-1.0, 1.0) * params.pixel_size, 0.0).xyz * exposure * params.luminance_multiplier; + vec3 rgbSE = textureLod(source_color, uv_interp + vec2(1.0, 1.0) * params.pixel_size, 0.0).xyz * exposure * params.luminance_multiplier; +#endif + vec3 rgbM = color; + vec3 luma = vec3(0.299, 0.587, 0.114); + float lumaNW = dot(rgbNW, luma); + 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; + +#ifdef MULTIVIEW + vec3 rgbA = 0.5 * exposure * (textureLod(source_color, vec3(uv_interp + dir * (1.0 / 3.0 - 0.5), ViewIndex), 0.0).xyz + textureLod(source_color, vec3(uv_interp + dir * (2.0 / 3.0 - 0.5), ViewIndex), 0.0).xyz) * params.luminance_multiplier; + vec3 rgbB = rgbA * 0.5 + 0.25 * exposure * (textureLod(source_color, vec3(uv_interp + dir * -0.5, ViewIndex), 0.0).xyz + textureLod(source_color, vec3(uv_interp + dir * 0.5, ViewIndex), 0.0).xyz) * params.luminance_multiplier; +#else + vec3 rgbA = 0.5 * exposure * (textureLod(source_color, uv_interp + dir * (1.0 / 3.0 - 0.5), 0.0).xyz + textureLod(source_color, uv_interp + dir * (2.0 / 3.0 - 0.5), 0.0).xyz) * params.luminance_multiplier; + vec3 rgbB = rgbA * 0.5 + 0.25 * exposure * (textureLod(source_color, uv_interp + dir * -0.5, 0.0).xyz + textureLod(source_color, uv_interp + dir * 0.5, 0.0).xyz) * params.luminance_multiplier; +#endif + + float lumaB = dot(rgbB, luma); + if ((lumaB < lumaMin) || (lumaB > lumaMax)) { + return rgbA; + } else { + return rgbB; + } +} +#endif // !SUBPASS + +// From https://alex.vlachos.com/graphics/Alex_Vlachos_Advanced_VR_Rendering_GDC2015.pdf +// and https://www.shadertoy.com/view/MslGR8 (5th one starting from the bottom) +// NOTE: `frag_coord` is in pixels (i.e. not normalized UV). +vec3 screen_space_dither(vec2 frag_coord) { + // Iestyn's RGB dither (7 asm instructions) from Portal 2 X360, slightly modified for VR. + vec3 dither = vec3(dot(vec2(171.0, 231.0), frag_coord)); + dither.rgb = fract(dither.rgb / vec3(103.0, 71.0, 97.0)); + + // Subtract 0.5 to avoid slightly brightening the whole viewport. + return (dither.rgb - 0.5) / 255.0; +} + +void main() { +#ifdef SUBPASS + // SUBPASS and MULTIVIEW can be combined but in that case we're already reading from the correct layer + vec4 color = subpassLoad(input_color); +#elif defined(MULTIVIEW) + vec4 color = textureLod(source_color, vec3(uv_interp, ViewIndex), 0.0f); +#else + vec4 color = textureLod(source_color, uv_interp, 0.0f); +#endif + color.rgb *= params.luminance_multiplier; + + // Exposure + + float exposure = params.exposure; + +#ifndef SUBPASS + if (params.use_auto_exposure) { + exposure *= 1.0 / (texelFetch(source_auto_exposure, ivec2(0, 0), 0).r * params.luminance_multiplier / params.auto_exposure_grey); + } +#endif + + color.rgb *= exposure; + + // Early Tonemap & SRGB Conversion +#ifndef SUBPASS + if (params.use_fxaa) { + // FXAA must be performed before glow to preserve the "bleed" effect of glow. + color.rgb = do_fxaa(color.rgb, exposure, uv_interp); + } + + if (params.use_glow && params.glow_mode == GLOW_MODE_MIX) { + vec3 glow = gather_glow(source_glow, uv_interp) * params.luminance_multiplier; + if (params.glow_map_strength > 0.001) { + glow = mix(glow, texture(glow_map, uv_interp).rgb * glow, params.glow_map_strength); + } + color.rgb = mix(color.rgb, glow, params.glow_intensity); + } +#endif + + if (params.use_debanding) { + // For best results, debanding should be done before tonemapping. + // Otherwise, we're adding noise to an already-quantized image. + color.rgb += screen_space_dither(gl_FragCoord.xy); + } + + color.rgb = apply_tonemapping(color.rgb, params.white); + + color.rgb = linear_to_srgb(color.rgb); // regular linear -> SRGB conversion + +#ifndef SUBPASS + // Glow + if (params.use_glow && params.glow_mode != GLOW_MODE_MIX) { + vec3 glow = gather_glow(source_glow, uv_interp) * params.glow_intensity * params.luminance_multiplier; + if (params.glow_map_strength > 0.001) { + glow = mix(glow, texture(glow_map, uv_interp).rgb * glow, params.glow_map_strength); + } + + // high dynamic range -> SRGB + glow = apply_tonemapping(glow, params.white); + glow = linear_to_srgb(glow); + + color.rgb = apply_glow(color.rgb, glow); + } +#endif + + // Additional effects + + if (params.use_bcs) { + color.rgb = apply_bcs(color.rgb, params.bcs); + } + + if (params.use_color_correction) { + color.rgb = apply_color_correction(color.rgb); + } + + frag_color = color; +} |