diff options
Diffstat (limited to 'servers/rendering/renderer_rd/shaders')
11 files changed, 116 insertions, 104 deletions
diff --git a/servers/rendering/renderer_rd/shaders/canvas.glsl b/servers/rendering/renderer_rd/shaders/canvas.glsl index 8593e6b265..eb5f68849e 100644 --- a/servers/rendering/renderer_rd/shaders/canvas.glsl +++ b/servers/rendering/renderer_rd/shaders/canvas.glsl @@ -191,48 +191,6 @@ void main() { uv += 1e-5; } -#ifdef USE_ATTRIBUTES -#if 0 - if (bool(draw_data.flags & FLAGS_USE_SKELETON) && bone_weights != vec4(0.0)) { //must be a valid bone - //skeleton transform - ivec4 bone_indicesi = ivec4(bone_indices); - - uvec2 tex_ofs = bone_indicesi.x * 2; - - mat2x4 m; - m = mat2x4( - texelFetch(skeleton_buffer, tex_ofs + 0), - texelFetch(skeleton_buffer, tex_ofs + 1)) * - bone_weights.x; - - tex_ofs = bone_indicesi.y * 2; - - m += mat2x4( - texelFetch(skeleton_buffer, tex_ofs + 0), - texelFetch(skeleton_buffer, tex_ofs + 1)) * - bone_weights.y; - - tex_ofs = bone_indicesi.z * 2; - - m += mat2x4( - texelFetch(skeleton_buffer, tex_ofs + 0), - texelFetch(skeleton_buffer, tex_ofs + 1)) * - bone_weights.z; - - tex_ofs = bone_indicesi.w * 2; - - m += mat2x4( - texelFetch(skeleton_buffer, tex_ofs + 0), - texelFetch(skeleton_buffer, tex_ofs + 1)) * - bone_weights.w; - - mat4 bone_matrix = skeleton_data.skeleton_transform * transpose(mat4(m[0], m[1], vec4(0.0, 0.0, 1.0, 0.0), vec4(0.0, 0.0, 0.0, 1.0))) * skeleton_data.skeleton_transform_inverse; - - //outvec = bone_matrix * outvec; - } -#endif -#endif - vertex = (canvas_data.canvas_transform * vec4(vertex, 0.0, 1.0)).xy; vertex_interp = vertex; diff --git a/servers/rendering/renderer_rd/shaders/cluster_render.glsl b/servers/rendering/renderer_rd/shaders/cluster_render.glsl index 932312de82..8c26a67926 100644 --- a/servers/rendering/renderer_rd/shaders/cluster_render.glsl +++ b/servers/rendering/renderer_rd/shaders/cluster_render.glsl @@ -142,7 +142,11 @@ void main() { } } #else - if (!gl_HelperInvocation) { +// MoltenVK/Metal fails to compile shaders using gl_HelperInvocation for some GPUs +#ifndef MOLTENVK_USED + if (!gl_HelperInvocation) +#endif + { atomicOr(cluster_render.data[usage_write_offset], usage_write_bit); } #endif @@ -162,7 +166,11 @@ void main() { } } #else - if (!gl_HelperInvocation) { +// MoltenVK/Metal fails to compile shaders using gl_HelperInvocation for some GPUs +#ifndef MOLTENVK_USED + if (!gl_HelperInvocation) +#endif + { atomicOr(cluster_render.data[z_write_offset], z_write_bit); } #endif diff --git a/servers/rendering/renderer_rd/shaders/effects/blur_raster.glsl b/servers/rendering/renderer_rd/shaders/effects/blur_raster.glsl index cb06250cf2..31aabbe9d2 100644 --- a/servers/rendering/renderer_rd/shaders/effects/blur_raster.glsl +++ b/servers/rendering/renderer_rd/shaders/effects/blur_raster.glsl @@ -53,30 +53,31 @@ void main() { #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; - } + // For Gaussian Blur we use 13 taps in a single pass instead of 12 taps over 2 passes. + // This minimizes the number of times we change framebuffers which is very important for mobile. + // Source: http://www.iryoku.com/next-generation-post-processing-in-call-of-duty-advanced-warfare + vec4 A = texture(source_color, uv_interp + blur.pixel_size * vec2(-1.0, -1.0)); + vec4 B = texture(source_color, uv_interp + blur.pixel_size * vec2(0.0, -1.0)); + vec4 C = texture(source_color, uv_interp + blur.pixel_size * vec2(1.0, -1.0)); + vec4 D = texture(source_color, uv_interp + blur.pixel_size * vec2(-0.5, -0.5)); + vec4 E = texture(source_color, uv_interp + blur.pixel_size * vec2(0.5, -0.5)); + vec4 F = texture(source_color, uv_interp + blur.pixel_size * vec2(-1.0, 0.0)); + vec4 G = texture(source_color, uv_interp); + vec4 H = texture(source_color, uv_interp + blur.pixel_size * vec2(1.0, 0.0)); + vec4 I = texture(source_color, uv_interp + blur.pixel_size * vec2(-0.5, 0.5)); + vec4 J = texture(source_color, uv_interp + blur.pixel_size * vec2(0.5, 0.5)); + vec4 K = texture(source_color, uv_interp + blur.pixel_size * vec2(-1.0, 1.0)); + vec4 L = texture(source_color, uv_interp + blur.pixel_size * vec2(0.0, 1.0)); + vec4 M = texture(source_color, uv_interp + blur.pixel_size * vec2(1.0, 1.0)); + + float base_weight = 0.5 / 4.0; + float lesser_weight = 0.125 / 4.0; + + frag_color = (D + E + I + J) * base_weight; + frag_color += (A + B + G + F) * lesser_weight; + frag_color += (B + C + H + G) * lesser_weight; + frag_color += (F + G + L + K) * lesser_weight; + frag_color += (G + H + M + L) * lesser_weight; #endif #ifdef MODE_GAUSSIAN_GLOW diff --git a/servers/rendering/renderer_rd/shaders/effects/copy.glsl b/servers/rendering/renderer_rd/shaders/effects/copy.glsl index bfe329b8ec..5cc2ed7622 100644 --- a/servers/rendering/renderer_rd/shaders/effects/copy.glsl +++ b/servers/rendering/renderer_rd/shaders/effects/copy.glsl @@ -194,10 +194,10 @@ void main() { color = min(color * feedback, vec4(params.glow_luminance_cap)); } -#endif +#endif // MODE_GLOW imageStore(dest_buffer, pos + params.target, color); -#endif +#endif // MODE_GAUSSIAN_BLUR #ifdef MODE_SIMPLE_COPY @@ -227,7 +227,7 @@ void main() { imageStore(dest_buffer, pos + params.target, color); -#endif +#endif // MODE_SIMPLE_COPY #ifdef MODE_SIMPLE_COPY_DEPTH @@ -239,7 +239,7 @@ void main() { imageStore(dest_buffer, pos + params.target, vec4(color.r)); -#endif +#endif // MODE_SIMPLE_COPY_DEPTH #ifdef MODE_LINEARIZE_DEPTH_COPY @@ -253,7 +253,7 @@ void main() { } imageStore(dest_buffer, pos + params.target, color); -#endif +#endif // MODE_LINEARIZE_DEPTH_COPY #if defined(MODE_CUBEMAP_TO_PANORAMA) || defined(MODE_CUBEMAP_ARRAY_TO_PANORAMA) @@ -276,7 +276,7 @@ void main() { 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 +#endif // defined(MODE_CUBEMAP_TO_PANORAMA) || defined(MODE_CUBEMAP_ARRAY_TO_PANORAMA) #ifdef MODE_SET_COLOR imageStore(dest_buffer, pos + params.target, params.set_color); diff --git a/servers/rendering/renderer_rd/shaders/effects/copy_to_fb.glsl b/servers/rendering/renderer_rd/shaders/effects/copy_to_fb.glsl index 1c17eabb56..46bb99794d 100644 --- a/servers/rendering/renderer_rd/shaders/effects/copy_to_fb.glsl +++ b/servers/rendering/renderer_rd/shaders/effects/copy_to_fb.glsl @@ -13,6 +13,14 @@ #endif // has_VK_KHR_multiview #endif //MULTIVIEW +#define FLAG_FLIP_Y (1 << 0) +#define FLAG_USE_SECTION (1 << 1) +#define FLAG_FORCE_LUMINANCE (1 << 2) +#define FLAG_ALPHA_TO_ZERO (1 << 3) +#define FLAG_SRGB (1 << 4) +#define FLAG_ALPHA_TO_ONE (1 << 5) +#define FLAG_LINEAR (1 << 6) + #ifdef MULTIVIEW layout(location = 0) out vec3 uv_interp; #else @@ -22,11 +30,10 @@ layout(location = 0) out vec2 uv_interp; layout(push_constant, std430) uniform Params { vec4 section; vec2 pixel_size; - bool flip_y; - bool use_section; + float luminance_multiplier; + uint flags; - bool force_luminance; - uint pad[3]; + vec4 color; } params; @@ -37,13 +44,13 @@ void main() { uv_interp.z = ViewIndex; #endif vec2 vpos = uv_interp.xy; - if (params.use_section) { + if (bool(params.flags & FLAG_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) { + if (bool(params.flags & FLAG_FLIP_Y)) { uv_interp.y = 1.0 - uv_interp.y; } } @@ -63,16 +70,21 @@ void main() { #endif // has_VK_KHR_multiview #endif //MULTIVIEW +#define FLAG_FLIP_Y (1 << 0) +#define FLAG_USE_SECTION (1 << 1) +#define FLAG_FORCE_LUMINANCE (1 << 2) +#define FLAG_ALPHA_TO_ZERO (1 << 3) +#define FLAG_SRGB (1 << 4) +#define FLAG_ALPHA_TO_ONE (1 << 5) +#define FLAG_LINEAR (1 << 6) + layout(push_constant, std430) uniform Params { vec4 section; vec2 pixel_size; - bool flip_y; - bool use_section; + float luminance_multiplier; + uint flags; - bool force_luminance; - bool alpha_to_zero; - bool srgb; - uint pad; + vec4 color; } params; @@ -104,7 +116,15 @@ vec3 linear_to_srgb(vec3 color) { 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))); } +vec3 srgb_to_linear(vec3 color) { + return mix(pow((color.rgb + vec3(0.055)) * (1.0 / (1.0 + 0.055)), vec3(2.4)), color.rgb * (1.0 / 12.92), lessThan(color.rgb, vec3(0.04045))); +} + void main() { +#ifdef MODE_SET_COLOR + frag_color = params.color; +#else + #ifdef MULTIVIEW vec3 uv = uv_interp; #else @@ -155,15 +175,22 @@ void main() { #endif /* MODE_TWO_SOURCES */ #endif /* MULTIVIEW */ - if (params.force_luminance) { + if (bool(params.flags & FLAG_FORCE_LUMINANCE)) { color.rgb = vec3(max(max(color.r, color.g), color.b)); } - if (params.alpha_to_zero) { + if (bool(params.flags & FLAG_ALPHA_TO_ZERO)) { color.rgb *= color.a; } - if (params.srgb) { + if (bool(params.flags & FLAG_SRGB)) { color.rgb = linear_to_srgb(color.rgb); } + if (bool(params.flags & FLAG_ALPHA_TO_ONE)) { + color.a = 1.0; + } + if (bool(params.flags & FLAG_LINEAR)) { + color.rgb = srgb_to_linear(color.rgb); + } - frag_color = color; + frag_color = color / params.luminance_multiplier; +#endif // MODE_SET_COLOR } diff --git a/servers/rendering/renderer_rd/shaders/effects/vrs.glsl b/servers/rendering/renderer_rd/shaders/effects/vrs.glsl index 5ef83c0b44..b450bb9fe9 100644 --- a/servers/rendering/renderer_rd/shaders/effects/vrs.glsl +++ b/servers/rendering/renderer_rd/shaders/effects/vrs.glsl @@ -63,10 +63,18 @@ void main() { #ifdef MULTIVIEW vec4 color = textureLod(source_color, uv, 0.0); + frag_color = uint(color.r * 255.0); #else /* MULTIVIEW */ vec4 color = textureLod(source_color, uv, 0.0); -#endif /* MULTIVIEW */ - // See if we can change the sampler to one that returns int... - frag_color = uint(color.r * 256.0); + // for user supplied VRS map we do a color mapping + color.r *= 3.0; + frag_color = int(color.r) << 2; + + color.g *= 3.0; + frag_color += int(color.g); + + // note 1x4, 4x1, 1x8, 8x1, 2x8 and 8x2 are not supported + // 4x8, 8x4 and 8x8 are only available on some GPUs +#endif /* MULTIVIEW */ } diff --git a/servers/rendering/renderer_rd/shaders/forward_clustered/scene_forward_clustered.glsl b/servers/rendering/renderer_rd/shaders/forward_clustered/scene_forward_clustered.glsl index 0bdf0e50aa..896f51ca01 100644 --- a/servers/rendering/renderer_rd/shaders/forward_clustered/scene_forward_clustered.glsl +++ b/servers/rendering/renderer_rd/shaders/forward_clustered/scene_forward_clustered.glsl @@ -62,7 +62,7 @@ vec3 oct_to_vec3(vec2 e) { vec3 v = vec3(e.xy, 1.0 - abs(e.x) - abs(e.y)); float t = max(-v.z, 0.0); v.xy += t * -sign(v.xy); - return v; + return normalize(v); } /* Varyings */ @@ -826,7 +826,8 @@ void fragment_shader(in SceneData scene_data) { // alpha hash can be used in unison with alpha antialiasing #ifdef ALPHA_HASH_USED - if (alpha < compute_alpha_hash_threshold(vertex, alpha_hash_scale)) { + vec3 object_pos = (inverse(read_model_matrix) * inv_view_matrix * vec4(vertex, 1.0)).xyz; + if (alpha < compute_alpha_hash_threshold(object_pos, alpha_hash_scale)) { discard; } #endif // ALPHA_HASH_USED diff --git a/servers/rendering/renderer_rd/shaders/forward_mobile/scene_forward_mobile.glsl b/servers/rendering/renderer_rd/shaders/forward_mobile/scene_forward_mobile.glsl index 9aeaa6d978..d50749306e 100644 --- a/servers/rendering/renderer_rd/shaders/forward_mobile/scene_forward_mobile.glsl +++ b/servers/rendering/renderer_rd/shaders/forward_mobile/scene_forward_mobile.glsl @@ -63,7 +63,7 @@ vec3 oct_to_vec3(vec2 e) { vec3 v = vec3(e.xy, 1.0 - abs(e.x) - abs(e.y)); float t = max(-v.z, 0.0); v.xy += t * -sign(v.xy); - return v; + return normalize(v); } /* Varyings */ @@ -779,7 +779,8 @@ void main() { // alpha hash can be used in unison with alpha antialiasing #ifdef ALPHA_HASH_USED - if (alpha < compute_alpha_hash_threshold(vertex, alpha_hash_scale)) { + vec3 object_pos = (inverse(read_model_matrix) * inv_view_matrix * vec4(vertex, 1.0)).xyz; + if (alpha < compute_alpha_hash_threshold(object_pos, alpha_hash_scale)) { discard; } #endif // ALPHA_HASH_USED diff --git a/servers/rendering/renderer_rd/shaders/particles.glsl b/servers/rendering/renderer_rd/shaders/particles.glsl index 3a6dd579b9..9f6aa7adc0 100644 --- a/servers/rendering/renderer_rd/shaders/particles.glsl +++ b/servers/rendering/renderer_rd/shaders/particles.glsl @@ -462,7 +462,7 @@ void main() { if (any(lessThan(uvw_pos, vec3(0.0))) || any(greaterThan(uvw_pos, vec3(1.0)))) { continue; } - vec3 s = texture(sampler3D(sdf_vec_textures[FRAME.attractors[i].texture_index], material_samplers[SAMPLER_LINEAR_CLAMP]), uvw_pos).xyz * 2.0 - 1.0; + vec3 s = texture(sampler3D(sdf_vec_textures[FRAME.attractors[i].texture_index], material_samplers[SAMPLER_LINEAR_CLAMP]), uvw_pos).xyz * -2.0 + 1.0; dir = mat3(FRAME.attractors[i].transform) * safe_normalize(s); //revert direction amount = length(s); diff --git a/servers/rendering/renderer_rd/shaders/scene_forward_aa_inc.glsl b/servers/rendering/renderer_rd/shaders/scene_forward_aa_inc.glsl index 97c913d489..71510ee06a 100644 --- a/servers/rendering/renderer_rd/shaders/scene_forward_aa_inc.glsl +++ b/servers/rendering/renderer_rd/shaders/scene_forward_aa_inc.glsl @@ -11,7 +11,8 @@ float hash_3d(vec3 p) { float compute_alpha_hash_threshold(vec3 pos, float hash_scale) { vec3 dx = dFdx(pos); - vec3 dy = dFdx(pos); + vec3 dy = dFdy(pos); + float delta_max_sqr = max(length(dx), length(dy)); float pix_scale = 1.0 / (hash_scale * delta_max_sqr); @@ -32,9 +33,9 @@ float compute_alpha_hash_threshold(vec3 pos, float hash_scale) { 1.0 - ((1.0 - a_interp) * (1.0 - a_interp) / (2.0 * min_lerp * (1.0 - min_lerp)))); float alpha_hash_threshold = - (lerp_factor < (1.0 - min_lerp)) ? ((lerp_factor < min_lerp) ? cases.x : cases.y) : cases.z; + (a_interp < (1.0 - min_lerp)) ? ((a_interp < min_lerp) ? cases.x : cases.y) : cases.z; - return clamp(alpha_hash_threshold, 0.0, 1.0); + return clamp(alpha_hash_threshold, 0.00001, 1.0); } #endif // ALPHA_HASH_USED diff --git a/servers/rendering/renderer_rd/shaders/skeleton.glsl b/servers/rendering/renderer_rd/shaders/skeleton.glsl index 75bea9300b..f5b233cca0 100644 --- a/servers/rendering/renderer_rd/shaders/skeleton.glsl +++ b/servers/rendering/renderer_rd/shaders/skeleton.glsl @@ -63,7 +63,7 @@ vec3 oct_to_vec3(vec2 oct) { vec3 v = vec3(oct.xy, 1.0 - abs(oct.x) - abs(oct.y)); float t = max(-v.z, 0.0); v.xy += t * -sign(v.xy); - return v; + return normalize(v); } vec3 decode_uint_oct_to_norm(uint base) { @@ -143,8 +143,8 @@ void main() { uint skin_offset = params.skin_stride * index; uvec2 bones = uvec2(src_bone_weights.data[skin_offset + 0], src_bone_weights.data[skin_offset + 1]); - uvec2 bones_01 = uvec2(bones.x & 0xFFFF, bones.x >> 16) * 3; //pre-add xform offset - uvec2 bones_23 = uvec2(bones.y & 0xFFFF, bones.y >> 16) * 3; + uvec2 bones_01 = uvec2(bones.x & 0xFFFF, bones.x >> 16) * 2; //pre-add xform offset + uvec2 bones_23 = uvec2(bones.y & 0xFFFF, bones.y >> 16) * 2; skin_offset += params.skin_weight_offset; @@ -161,6 +161,13 @@ void main() { //reverse order because its transposed vertex = (vec4(vertex, 0.0, 1.0) * m).xy; } + + uint dst_offset = index * params.vertex_stride; + + uvec2 uvertex = floatBitsToUint(vertex); + dst_vertices.data[dst_offset + 0] = uvertex.x; + dst_vertices.data[dst_offset + 1] = uvertex.y; + #else vec3 vertex; vec3 normal; |