feat: updated engine version to 4.4-rc1

This commit is contained in:
Sara 2025-02-23 14:38:14 +01:00
parent ee00efde1f
commit 21ba8e33af
5459 changed files with 1128836 additions and 198305 deletions

View file

@ -1,4 +1,5 @@
#!/usr/bin/env python
from misc.utility.scons_hints import *
Import("env")
@ -16,6 +17,7 @@ if "GLES3_GLSL" in env["BUILDERS"]:
# as we have a few, not yet, converted files we name the ones we want to include:
env.GLES3_GLSL("canvas.glsl")
env.GLES3_GLSL("feed.glsl")
env.GLES3_GLSL("scene.glsl")
env.GLES3_GLSL("sky.glsl")
env.GLES3_GLSL("canvas_occlusion.glsl")

View file

@ -1,17 +1,16 @@
/* clang-format off */
#[modes]
mode_quad =
mode_ninepatch = #define USE_NINEPATCH
mode_primitive = #define USE_PRIMITIVE
mode_attributes = #define USE_ATTRIBUTES
mode_instanced = #define USE_ATTRIBUTES \n#define USE_INSTANCING
mode_default =
#[specializations]
DISABLE_LIGHTING = true
USE_RGBA_SHADOWS = false
SINGLE_INSTANCE = false
USE_NINEPATCH = false
USE_PRIMITIVE = false
USE_ATTRIBUTES = false
USE_INSTANCING = false
#[vertex]
@ -84,7 +83,7 @@ layout(location = 15) in highp uvec4 attrib_H;
#endif
#define read_draw_data_flags attrib_G.z
#define read_draw_data_specular_shininess attrib_G.w
#define read_draw_data_instance_offset attrib_G.w
#define read_draw_data_lights attrib_H
// Varyings so the per-instance info can be used in the fragment shader
@ -111,6 +110,9 @@ layout(std140) uniform MaterialUniforms{ //ubo:4
};
#endif
uniform mediump uint batch_flags;
/* clang-format on */
#include "canvas_uniforms_inc.glsl"
@ -140,7 +142,7 @@ void main() {
#endif // !USE_ATTRIBUTES
#endif // USE_PRIMITIVE
varying_F = uvec2(read_draw_data_flags, read_draw_data_specular_shininess);
varying_F = uvec2(read_draw_data_flags, read_draw_data_instance_offset);
varying_G = read_draw_data_lights;
vec4 instance_custom = vec4(0.0);
@ -180,13 +182,13 @@ void main() {
vec2 uv = uv_attrib;
#ifdef USE_INSTANCING
if (bool(read_draw_data_flags & FLAGS_INSTANCING_HAS_COLORS)) {
if (bool(batch_flags & BATCH_FLAGS_INSTANCING_HAS_COLORS)) {
vec4 instance_color;
instance_color.xy = unpackHalf2x16(uint(instance_color_custom_data.x));
instance_color.zw = unpackHalf2x16(uint(instance_color_custom_data.y));
color *= instance_color;
}
if (bool(read_draw_data_flags & FLAGS_INSTANCING_HAS_CUSTOM_DATA)) {
if (bool(batch_flags & BATCH_FLAGS_INSTANCING_HAS_CUSTOM_DATA)) {
instance_custom.xy = unpackHalf2x16(instance_color_custom_data.z);
instance_custom.zw = unpackHalf2x16(instance_color_custom_data.w);
}
@ -206,20 +208,21 @@ void main() {
// no crash or freeze on all Adreno 3xx with 'if / else if' and slightly faster!
int vertex_id = gl_VertexID % 6;
vec2 vertex_base;
if (vertex_id == 0)
if (vertex_id == 0) {
vertex_base = vec2(0.0, 0.0);
else if (vertex_id == 1)
} else if (vertex_id == 1) {
vertex_base = vec2(0.0, 1.0);
else if (vertex_id == 2)
} else if (vertex_id == 2) {
vertex_base = vec2(1.0, 1.0);
else if (vertex_id == 3)
} else if (vertex_id == 3) {
vertex_base = vec2(1.0, 0.0);
else if (vertex_id == 4)
} else if (vertex_id == 4) {
vertex_base = vec2(0.0, 0.0);
else if (vertex_id == 5)
} else if (vertex_id == 5) {
vertex_base = vec2(1.0, 1.0);
}
vec2 uv = read_draw_data_src_rect.xy + abs(read_draw_data_src_rect.zw) * ((read_draw_data_flags & FLAGS_TRANSPOSE_RECT) != uint(0) ? vertex_base.yx : vertex_base.xy);
vec2 uv = read_draw_data_src_rect.xy + abs(read_draw_data_src_rect.zw) * ((read_draw_data_flags & INSTANCE_FLAGS_TRANSPOSE_RECT) != uint(0) ? vertex_base.yx : vertex_base.xy);
vec4 color = read_draw_data_modulation;
vec2 vertex = read_draw_data_dst_rect.xy + abs(read_draw_data_dst_rect.zw) * mix(vertex_base, vec2(1.0, 1.0) - vertex_base, lessThan(read_draw_data_src_rect.zw, vec2(0.0, 0.0)));
@ -262,6 +265,8 @@ void main() {
color_interp = color;
vertex = (canvas_transform * vec4(vertex, 0.0, 1.0)).xy;
if (use_pixel_snap) {
vertex = floor(vertex + 0.5);
// precision issue on some hardware creates artifacts within texture
@ -269,8 +274,6 @@ void main() {
uv += 1e-5;
}
vertex = (canvas_transform * vec4(vertex, 0.0, 1.0)).xy;
vertex_interp = vertex;
uv_interp = uv;
@ -323,7 +326,7 @@ flat in vec4 varying_E;
flat in uvec2 varying_F;
flat in uvec4 varying_G;
#define read_draw_data_flags varying_F.x
#define read_draw_data_specular_shininess varying_F.y
#define read_draw_data_instance_offset varying_F.y
#define read_draw_data_lights varying_G
#ifndef DISABLE_LIGHTING
@ -337,6 +340,9 @@ uniform sampler2D specular_texture; //texunit:-7
uniform sampler2D color_texture; //texunit:0
uniform mediump uint batch_flags;
uniform highp uint specular_shininess_in;
layout(location = 0) out vec4 frag_color;
/* clang-format off */
@ -520,7 +526,7 @@ float map_ninepatch_axis(float pixel, float draw_size, float tex_pixel_size, flo
} else if (pixel >= draw_size - margin_end) {
return (tex_size - (draw_size - pixel)) * tex_pixel_size;
} else {
if (!bool(read_draw_data_flags & FLAGS_NINEPACH_DRAW_CENTER)) {
if (!bool(read_draw_data_flags & INSTANCE_FLAGS_NINEPATCH_DRAW_CENTER)) {
draw_center--;
}
@ -568,8 +574,8 @@ void main() {
int draw_center = 2;
uv = vec2(
map_ninepatch_axis(pixel_size_interp.x, abs(read_draw_data_dst_rect_z), read_draw_data_color_texture_pixel_size.x, read_draw_data_ninepatch_margins.x, read_draw_data_ninepatch_margins.z, int(read_draw_data_flags >> FLAGS_NINEPATCH_H_MODE_SHIFT) & 0x3, draw_center),
map_ninepatch_axis(pixel_size_interp.y, abs(read_draw_data_dst_rect_w), read_draw_data_color_texture_pixel_size.y, read_draw_data_ninepatch_margins.y, read_draw_data_ninepatch_margins.w, int(read_draw_data_flags >> FLAGS_NINEPATCH_V_MODE_SHIFT) & 0x3, draw_center));
map_ninepatch_axis(pixel_size_interp.x, abs(read_draw_data_dst_rect_z), read_draw_data_color_texture_pixel_size.x, read_draw_data_ninepatch_margins.x, read_draw_data_ninepatch_margins.z, int(read_draw_data_flags >> INSTANCE_FLAGS_NINEPATCH_H_MODE_SHIFT) & 0x3, draw_center),
map_ninepatch_axis(pixel_size_interp.y, abs(read_draw_data_dst_rect_w), read_draw_data_color_texture_pixel_size.y, read_draw_data_ninepatch_margins.y, read_draw_data_ninepatch_margins.w, int(read_draw_data_flags >> INSTANCE_FLAGS_NINEPATCH_V_MODE_SHIFT) & 0x3, draw_center));
if (draw_center == 0) {
color.a = 0.0;
@ -578,14 +584,15 @@ void main() {
uv = uv * read_draw_data_src_rect.zw + read_draw_data_src_rect.xy; //apply region if needed
#endif
if (bool(read_draw_data_flags & FLAGS_CLIP_RECT_UV)) {
uv = clamp(uv, read_draw_data_src_rect.xy, read_draw_data_src_rect.xy + abs(read_draw_data_src_rect.zw));
if (bool(read_draw_data_flags & INSTANCE_FLAGS_CLIP_RECT_UV)) {
vec2 half_texpixel = read_draw_data_color_texture_pixel_size * 0.5;
uv = clamp(uv, read_draw_data_src_rect.xy + half_texpixel, read_draw_data_src_rect.xy + abs(read_draw_data_src_rect.zw) - half_texpixel);
}
#endif
#ifndef USE_PRIMITIVE
if (bool(read_draw_data_flags & FLAGS_USE_MSDF)) {
if (bool(read_draw_data_flags & INSTANCE_FLAGS_USE_MSDF)) {
float px_range = read_draw_data_ninepatch_margins.x;
float outline_thickness = read_draw_data_ninepatch_margins.y;
@ -603,7 +610,7 @@ void main() {
float a = clamp(d * px_size + 0.5, 0.0, 1.0);
color.a = a * color.a;
}
} else if (bool(read_draw_data_flags & FLAGS_USE_LCD)) {
} else if (bool(read_draw_data_flags & INSTANCE_FLAGS_USE_LCD)) {
vec4 lcd_sample = texture(color_texture, uv);
if (lcd_sample.a == 1.0) {
color.rgb = lcd_sample.rgb * color.a;
@ -617,7 +624,7 @@ void main() {
color *= texture(color_texture, uv);
}
uint light_count = (read_draw_data_flags >> uint(FLAGS_LIGHT_COUNT_SHIFT)) & uint(0xF); //max 16 lights
uint light_count = read_draw_data_flags & uint(0xF); // Max 16 lights.
bool using_light = light_count > 0u || directional_light_count > 0u;
vec3 normal;
@ -628,17 +635,16 @@ void main() {
bool normal_used = false;
#endif
if (normal_used || (using_light && bool(read_draw_data_flags & FLAGS_DEFAULT_NORMAL_MAP_USED))) {
if (normal_used || (using_light && bool(batch_flags & BATCH_FLAGS_DEFAULT_NORMAL_MAP_USED))) {
normal.xy = texture(normal_texture, uv).xy * vec2(2.0, -2.0) - vec2(1.0, -1.0);
if (bool(read_draw_data_flags & FLAGS_TRANSPOSE_RECT)) {
#if !defined(USE_ATTRIBUTES) && !defined(USE_PRIMITIVE)
if (bool(read_draw_data_flags & INSTANCE_FLAGS_TRANSPOSE_RECT)) {
normal.xy = normal.yx;
}
if (bool(read_draw_data_flags & FLAGS_FLIP_H)) {
normal.x = -normal.x;
}
if (bool(read_draw_data_flags & FLAGS_FLIP_V)) {
normal.y = -normal.y;
}
normal.xy *= sign(read_draw_data_src_rect.zw);
#endif
normal.z = sqrt(max(0.0, 1.0 - dot(normal.xy, normal.xy)));
normal_used = true;
} else {
@ -654,9 +660,9 @@ void main() {
bool specular_shininess_used = false;
#endif
if (specular_shininess_used || (using_light && normal_used && bool(read_draw_data_flags & FLAGS_DEFAULT_SPECULAR_MAP_USED))) {
if (specular_shininess_used || (using_light && normal_used && bool(batch_flags & BATCH_FLAGS_DEFAULT_SPECULAR_MAP_USED))) {
specular_shininess = texture(specular_texture, uv);
specular_shininess *= godot_unpackUnorm4x8(read_draw_data_specular_shininess);
specular_shininess *= godot_unpackUnorm4x8(specular_shininess_in);
specular_shininess_used = true;
} else {
specular_shininess = vec4(1.0);
@ -802,7 +808,7 @@ void main() {
}
#endif
if (bool(light_array[light_base].flags & LIGHT_FLAGS_HAS_SHADOW)) {
if (bool(light_array[light_base].flags & LIGHT_FLAGS_HAS_SHADOW) && bool(read_draw_data_flags & uint(INSTANCE_FLAGS_SHADOW_MASKED << i))) {
vec2 shadow_pos = (vec4(shadow_vertex, 0.0, 1.0) * mat4(light_array[light_base].shadow_matrix[0], light_array[light_base].shadow_matrix[1], vec4(0.0, 0.0, 1.0, 0.0), vec4(0.0, 0.0, 0.0, 1.0))).xy; //multiply inverse given its transposed. Optimizer removes useless operations.
vec2 pos_norm = normalize(shadow_pos);

View file

@ -1,33 +1,32 @@
#define MAX_LIGHTS_PER_ITEM uint(16)
#define M_PI 3.14159265359
#define SDF_MAX_LENGTH 16384.0
//1 means enabled, 2+ means trails in use
#define FLAGS_INSTANCING_MASK uint(0x7F)
#define FLAGS_INSTANCING_HAS_COLORS uint(1 << 7)
#define FLAGS_INSTANCING_HAS_CUSTOM_DATA uint(1 << 8)
#define INSTANCE_FLAGS_LIGHT_COUNT_SHIFT 0 // 4 bits.
#define FLAGS_CLIP_RECT_UV uint(1 << 9)
#define FLAGS_TRANSPOSE_RECT uint(1 << 10)
// (1 << 11) is for FLAGS_CONVERT_ATTRIBUTES_TO_LINEAR in RD backends, unused here.
#define FLAGS_NINEPACH_DRAW_CENTER uint(1 << 12)
#define INSTANCE_FLAGS_CLIP_RECT_UV uint(1 << 4)
#define INSTANCE_FLAGS_TRANSPOSE_RECT uint(1 << 5)
#define INSTANCE_FLAGS_USE_MSDF uint(1 << 6)
#define INSTANCE_FLAGS_USE_LCD uint(1 << 7)
#define FLAGS_NINEPATCH_H_MODE_SHIFT 16
#define FLAGS_NINEPATCH_V_MODE_SHIFT 18
#define INSTANCE_FLAGS_NINEPATCH_DRAW_CENTER uint(1 << 8)
#define INSTANCE_FLAGS_NINEPATCH_H_MODE_SHIFT 9
#define INSTANCE_FLAGS_NINEPATCH_V_MODE_SHIFT 11
#define FLAGS_LIGHT_COUNT_SHIFT 20
#define INSTANCE_FLAGS_SHADOW_MASKED_SHIFT 13u // 16 bits.
#define INSTANCE_FLAGS_SHADOW_MASKED uint(1 << INSTANCE_FLAGS_SHADOW_MASKED_SHIFT)
#define FLAGS_DEFAULT_NORMAL_MAP_USED uint(1 << 26)
#define FLAGS_DEFAULT_SPECULAR_MAP_USED uint(1 << 27)
// 1 means enabled, 2+ means trails in use
#define BATCH_FLAGS_INSTANCING_MASK uint(0x7F)
#define BATCH_FLAGS_INSTANCING_HAS_COLORS_SHIFT 7
#define BATCH_FLAGS_INSTANCING_HAS_COLORS uint(1 << BATCH_FLAGS_INSTANCING_HAS_COLORS_SHIFT)
#define BATCH_FLAGS_INSTANCING_HAS_CUSTOM_DATA_SHIFT 8
#define BATCH_FLAGS_INSTANCING_HAS_CUSTOM_DATA uint(1 << BATCH_FLAGS_INSTANCING_HAS_CUSTOM_DATA_SHIFT)
#define FLAGS_USE_MSDF uint(1 << 28)
#define FLAGS_USE_LCD uint(1 << 29)
#define FLAGS_FLIP_H uint(1 << 30)
#define FLAGS_FLIP_V uint(1 << 31)
#define BATCH_FLAGS_DEFAULT_NORMAL_MAP_USED uint(1 << 9)
#define BATCH_FLAGS_DEFAULT_SPECULAR_MAP_USED uint(1 << 10)
layout(std140) uniform GlobalShaderUniformData { //ubo:1
vec4 global_shader_uniforms[MAX_GLOBAL_SHADER_UNIFORMS];

View file

@ -1,4 +1,5 @@
#!/usr/bin/env python
from misc.utility.scons_hints import *
Import("env")

View file

@ -0,0 +1,39 @@
/* clang-format off */
#[modes]
mode_default =
#[specializations]
USE_EXTERNAL_SAMPLER = false
#[vertex]
layout(location = 0) in vec2 vertex_attrib;
out vec2 uv_interp;
void main() {
uv_interp = vertex_attrib * 0.5 + 0.5;
gl_Position = vec4(vertex_attrib, 1.0, 1.0);
}
/* clang-format off */
#[fragment]
layout(location = 0) out vec4 frag_color;
in vec2 uv_interp;
/* clang-format on */
#ifdef USE_EXTERNAL_SAMPLER
uniform samplerExternalOES sourceFeed; // texunit:0
#else
uniform sampler2D sourceFeed; // texunit:0
#endif
void main() {
vec4 color = texture(sourceFeed, uv_interp);
frag_color = color;
}

View file

@ -342,7 +342,7 @@ void main() {
mediump float attractor_attenuation = attractors[i].attenuation;
amount = pow(amount, attractor_attenuation);
dir = safe_normalize(mix(dir, attractors[i].transform[2].xyz, attractors[i].directionality));
attractor_force -= amount * dir * attractors[i].strength;
attractor_force -= mass * amount * dir * attractors[i].strength;
}
float particle_size = particle_size;
@ -453,14 +453,14 @@ void main() {
vec3 uvw_pos = vec3(local_pos_bottom / colliders[i].extents.xyz) * 0.5 + 0.5;
float y = 1.0 - texture(height_field_texture, uvw_pos.xz).r;
float y = texture(height_field_texture, uvw_pos.xz).r;
if (y + EPSILON > uvw_pos.y) {
//inside heightfield
vec3 pos1 = (vec3(uvw_pos.x, y, uvw_pos.z) * 2.0 - 1.0) * colliders[i].extents.xyz;
vec3 pos2 = (vec3(uvw_pos.x + DELTA, 1.0 - texture(height_field_texture, uvw_pos.xz + vec2(DELTA, 0)).r, uvw_pos.z) * 2.0 - 1.0) * colliders[i].extents.xyz;
vec3 pos3 = (vec3(uvw_pos.x, 1.0 - texture(height_field_texture, uvw_pos.xz + vec2(0, DELTA)).r, uvw_pos.z + DELTA) * 2.0 - 1.0) * colliders[i].extents.xyz;
vec3 pos2 = (vec3(uvw_pos.x + DELTA, texture(height_field_texture, uvw_pos.xz + vec2(DELTA, 0)).r, uvw_pos.z) * 2.0 - 1.0) * colliders[i].extents.xyz;
vec3 pos3 = (vec3(uvw_pos.x, texture(height_field_texture, uvw_pos.xz + vec2(0, DELTA)).r, uvw_pos.z + DELTA) * 2.0 - 1.0) * colliders[i].extents.xyz;
normal = normalize(cross(pos1 - pos2, pos1 - pos3));
float local_y = (vec3(local_pos / colliders[i].extents.xyz) * 0.5 + 0.5).y;

File diff suppressed because it is too large Load diff

View file

@ -59,7 +59,7 @@ layout(location = 10) in highp uvec4 in_bone_attrib;
layout(location = 11) in mediump vec4 in_weight_attrib;
#endif
uniform mediump sampler2D skeleton_texture; // texunit:0
uniform highp sampler2D skeleton_texture; // texunit:0
#endif
/* clang-format on */

View file

@ -2,17 +2,15 @@
#[modes]
mode_background =
mode_half_res = #define USE_HALF_RES_PASS
mode_quarter_res = #define USE_QUARTER_RES_PASS
mode_cubemap = #define USE_CUBEMAP_PASS
mode_cubemap_half_res = #define USE_CUBEMAP_PASS \n#define USE_HALF_RES_PASS
mode_cubemap_quarter_res = #define USE_CUBEMAP_PASS \n#define USE_QUARTER_RES_PASS
#[specializations]
USE_MULTIVIEW = false
USE_INVERTED_Y = true
APPLY_TONEMAPPING = true
USE_QUARTER_RES_PASS = false
USE_HALF_RES_PASS = false
#[vertex]
@ -108,11 +106,11 @@ uniform float sky_energy_multiplier;
uniform float luminance_multiplier;
uniform float fog_aerial_perspective;
uniform vec3 fog_light_color;
uniform vec4 fog_light_color;
uniform float fog_sun_scatter;
uniform bool fog_enabled;
uniform float fog_density;
uniform float z_far;
uniform float fog_sky_affect;
uniform uint directional_light_count;
#ifdef USE_MULTIVIEW
@ -135,6 +133,24 @@ vec3 interleaved_gradient_noise(vec2 pos) {
}
#endif
#if !defined(DISABLE_FOG)
vec4 fog_process(vec3 view, vec3 sky_color) {
vec3 fog_color = mix(fog_light_color.rgb, sky_color, fog_aerial_perspective);
if (fog_sun_scatter > 0.001) {
vec4 sun_scatter = vec4(0.0);
float sun_total = 0.0;
for (uint i = 0u; i < directional_light_count; i++) {
vec3 light_color = directional_lights.data[i].color_size.xyz * directional_lights.data[i].direction_energy.w;
float light_amount = pow(max(dot(view, directional_lights.data[i].direction_energy.xyz), 0.0), 8.0);
fog_color += light_color * light_amount * fog_sun_scatter;
}
}
return vec4(fog_color, 1.0);
}
#endif // !DISABLE_FOG
void main() {
vec3 cube_normal;
#ifdef USE_MULTIVIEW
@ -194,15 +210,28 @@ void main() {
#endif
{
#CODE : SKY
}
color *= sky_energy_multiplier;
// Convert to Linear for tonemapping so color matches scene shader better
color = srgb_to_linear(color);
#if !defined(DISABLE_FOG) && !defined(USE_CUBEMAP_PASS)
// Draw "fixed" fog before volumetric fog to ensure volumetric fog can appear in front of the sky.
if (fog_enabled) {
vec4 fog = fog_process(cube_normal, color.rgb);
color.rgb = mix(color.rgb, fog.rgb, fog.a * fog_sky_affect);
}
if (custom_fog.a > 0.0) {
color.rgb = mix(color.rgb, custom_fog.rgb, custom_fog.a);
}
#endif // DISABLE_FOG
color *= exposure;
#ifdef APPLY_TONEMAPPING
color = apply_tonemapping(color, white);

View file

@ -1,13 +1,14 @@
// Compatibility renames. These are exposed with the "godot_" prefix
// to work around two distinct Adreno bugs:
// 1. Some Adreno devices expose ES310 functions in ES300 shaders.
// Internally, we must use the "godot_" prefix, but user shaders
// will be mapped automatically.
// 2. Adreno 3XX devices have poor implementations of the other packing
// functions, so we just use our own everywhere to keep it simple.
// functions, so we just use our own there to keep it simple.
#ifdef USE_HALF2FLOAT
// Floating point pack/unpack functions are part of the GLSL ES 300 specification used by web and mobile.
// It appears to be safe to expose these on mobile, but when running through ANGLE this appears to break.
uint float2half(uint f) {
uint e = f & uint(0x7f800000);
if (e <= uint(0x38000000)) {
@ -52,6 +53,17 @@ vec2 godot_unpackSnorm2x16(uint p) {
return clamp((v - 32767.0) * vec2(0.00003051851), vec2(-1.0), vec2(1.0));
}
#define packHalf2x16 godot_packHalf2x16
#define unpackHalf2x16 godot_unpackHalf2x16
#define packUnorm2x16 godot_packUnorm2x16
#define unpackUnorm2x16 godot_unpackUnorm2x16
#define packSnorm2x16 godot_packSnorm2x16
#define unpackSnorm2x16 godot_unpackSnorm2x16
#endif // USE_HALF2FLOAT
// Always expose these as they are ES310 functions and not available in ES300 or GLSL 330.
uint godot_packUnorm4x8(vec4 v) {
uvec4 uv = uvec4(round(clamp(v, vec4(0.0), vec4(1.0)) * 255.0));
return uv.x | (uv.y << uint(8)) | (uv.z << uint(16)) | (uv.w << uint(24));
@ -75,9 +87,3 @@ vec4 godot_unpackSnorm4x8(uint p) {
#define unpackUnorm4x8 godot_unpackUnorm4x8
#define packSnorm4x8 godot_packSnorm4x8
#define unpackSnorm4x8 godot_unpackSnorm4x8
#define packHalf2x16 godot_packHalf2x16
#define unpackHalf2x16 godot_unpackHalf2x16
#define packUnorm2x16 godot_packUnorm2x16
#define unpackUnorm2x16 godot_unpackUnorm2x16
#define packSnorm2x16 godot_packSnorm2x16
#define unpackSnorm2x16 godot_unpackSnorm2x16

View file

@ -27,6 +27,14 @@ vec3 srgb_to_linear(vec3 color) {
#ifdef APPLY_TONEMAPPING
// Based on Reinhard's extended formula, see equation 4 in https://doi.org/cjbgrt
vec3 tonemap_reinhard(vec3 color, float p_white) {
float white_squared = p_white * p_white;
vec3 white_squared_color = white_squared * color;
// Equivalent to color * (1 + color / white_squared) / (1 + color)
return (white_squared_color + color * color) / (white_squared_color + white_squared);
}
vec3 tonemap_filmic(vec3 color, float p_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)
@ -76,16 +84,86 @@ vec3 tonemap_aces(vec3 color, float p_white) {
return color_tonemapped / p_white_tonemapped;
}
vec3 tonemap_reinhard(vec3 color, float p_white) {
return (p_white * color + color) / (color * p_white + p_white);
// Polynomial approximation of EaryChow's AgX sigmoid curve.
// x must be within the range [0.0, 1.0]
vec3 agx_contrast_approx(vec3 x) {
// Generated with Excel trendline
// Input data: Generated using python sigmoid with EaryChow's configuration and 57 steps
// Additional padding values were added to give correct intersections at 0.0 and 1.0
// 6th order, intercept of 0.0 to remove an operation and ensure intersection at 0.0
vec3 x2 = x * x;
vec3 x4 = x2 * x2;
return 0.021 * x + 4.0111 * x2 - 25.682 * x2 * x + 70.359 * x4 - 74.778 * x4 * x + 27.069 * x4 * x2;
}
// This is an approximation and simplification of EaryChow's AgX implementation that is used by Blender.
// This code is based off of the script that generates the AgX_Base_sRGB.cube LUT that Blender uses.
// Source: https://github.com/EaryChow/AgX_LUT_Gen/blob/main/AgXBasesRGB.py
vec3 tonemap_agx(vec3 color) {
// Combined linear sRGB to linear Rec 2020 and Blender AgX inset matrices:
const mat3 srgb_to_rec2020_agx_inset_matrix = mat3(
0.54490813676363087053, 0.14044005884001287035, 0.088827411851915368603,
0.37377945959812267119, 0.75410959864013760045, 0.17887712465043811023,
0.081384976686407536266, 0.10543358536857773485, 0.73224999956948382528);
// Combined inverse AgX outset matrix and linear Rec 2020 to linear sRGB matrices.
const mat3 agx_outset_rec2020_to_srgb_matrix = mat3(
1.9645509602733325934, -0.29932243390911083839, -0.16436833806080403409,
-0.85585845117807513559, 1.3264510741502356555, -0.23822464068860595117,
-0.10886710826831608324, -0.027084020983874825605, 1.402665347143271889);
// LOG2_MIN = -10.0
// LOG2_MAX = +6.5
// MIDDLE_GRAY = 0.18
const float min_ev = -12.4739311883324; // log2(pow(2, LOG2_MIN) * MIDDLE_GRAY)
const float max_ev = 4.02606881166759; // log2(pow(2, LOG2_MAX) * MIDDLE_GRAY)
// Large negative values in one channel and large positive values in other
// channels can result in a colour that appears darker and more saturated than
// desired after passing it through the inset matrix. For this reason, it is
// best to prevent negative input values.
// This is done before the Rec. 2020 transform to allow the Rec. 2020
// transform to be combined with the AgX inset matrix. This results in a loss
// of color information that could be correctly interpreted within the
// Rec. 2020 color space as positive RGB values, but it is less common for Godot
// to provide this function with negative sRGB values and therefore not worth
// the performance cost of an additional matrix multiplication.
// A value of 2e-10 intentionally introduces insignificant error to prevent
// log2(0.0) after the inset matrix is applied; color will be >= 1e-10 after
// the matrix transform.
color = max(color, 2e-10);
// Do AGX in rec2020 to match Blender and then apply inset matrix.
color = srgb_to_rec2020_agx_inset_matrix * color;
// Log2 space encoding.
// Must be clamped because agx_contrast_approx may not work
// well with values outside of the range [0.0, 1.0]
color = clamp(log2(color), min_ev, max_ev);
color = (color - min_ev) / (max_ev - min_ev);
// Apply sigmoid function approximation.
color = agx_contrast_approx(color);
// Convert back to linear before applying outset matrix.
color = pow(color, vec3(2.4));
// Apply outset to make the result more chroma-laden and then go back to linear sRGB.
color = agx_outset_rec2020_to_srgb_matrix * color;
// Blender's lusRGB.compensate_low_side is too complex for this shader, so
// simply return the color, even if it has negative components. These negative
// components may be useful for subsequent color adjustments.
return color;
}
#define TONEMAPPER_LINEAR 0
#define TONEMAPPER_REINHARD 1
#define TONEMAPPER_FILMIC 2
#define TONEMAPPER_ACES 3
#define TONEMAPPER_AGX 4
vec3 apply_tonemapping(vec3 color, float p_white) { // inputs are LINEAR, always outputs clamped [0;1] color
vec3 apply_tonemapping(vec3 color, float p_white) { // inputs are LINEAR
// Ensure color values passed to tonemappers are positive.
// They can be negative in the case of negative lights, which leads to undesired behavior.
if (tonemapper == TONEMAPPER_LINEAR) {
@ -94,8 +172,10 @@ vec3 apply_tonemapping(vec3 color, float p_white) { // inputs are LINEAR, always
return tonemap_reinhard(max(vec3(0.0f), color), p_white);
} else if (tonemapper == TONEMAPPER_FILMIC) {
return tonemap_filmic(max(vec3(0.0f), color), p_white);
} else { // TONEMAPPER_ACES
} else if (tonemapper == TONEMAPPER_ACES) {
return tonemap_aces(max(vec3(0.0f), color), p_white);
} else { // TONEMAPPER_AGX
return tonemap_agx(color);
}
}