slang-shaders/bezel/koko-aio/shaders-ng/shift_and_bleed.slang
2024-03-05 09:38:54 +01:00

231 lines
8.4 KiB
Plaintext

#version 450
#include "config.inc"
#define RGB_SHIFT_RANGE 20
#pragma stage vertex
layout(location = 0) in vec4 Position;
layout(location = 1) in vec2 TexCoord;
layout(location = 0) out vec2 vTexCoord;
layout(location = 1) out vec2 vR_offset;
layout(location = 2) out vec2 vG_offset;
layout(location = 3) out vec2 vB_offset;
layout(location = 4) out float vDo_shadow_mode;
layout(location = 5) out float vDecon_or_JustShadow;
layout(location = 6) out float vShow_artifact_mask;
#include "includes/functions.include.slang"
void main()
{
gl_Position = global.MVP * Position;
vTexCoord = TexCoord ;
//...to tell fragment shader if dot matrix feature requests shadows.
vDo_shadow_mode = float(DOT_M_SHADOW_STR + DO_DOT_MATRIX > 1 + eps);
//Shadow mode disables deconvergence, they do not coexist in real life.
if (vDo_shadow_mode == 1.0) {
vR_offset = vec2(DOT_M_SHADOW_OFF, abs(DOT_M_SHADOW_OFF));
vG_offset = vR_offset;
vB_offset = vR_offset;
} else if (DO_SHIFT_RGB == 1.0) {
vR_offset=offsets_from_float(SHIFT_R+210.0,RGB_SHIFT_RANGE);
vG_offset=offsets_from_float(SHIFT_G+210.0,RGB_SHIFT_RANGE);
vB_offset=offsets_from_float(SHIFT_B+210.0,RGB_SHIFT_RANGE);
}
vec2 d = -params.OutputSize.zw * 0.5;
vR_offset *= d;
vG_offset *= d;
vB_offset *= d;
//Precalc some conditions:
bool do_shadow = (DO_DOT_MATRIX == 1.0 && DOT_M_SHADOW_STR > 0.0);
vDecon_or_JustShadow = float(DO_SHIFT_RGB > 0.0 || do_shadow );
vShow_artifact_mask = float(NTSC_SHOW_ARTF_MASK + DO_NTSC_ARTIFACTS == 2.0);
}
#pragma stage fragment
#include "includes/functions.include.slang"
layout(location = 0) in vec2 vTexCoord;
layout(location = 1) in vec2 vR_offset;
layout(location = 2) in vec2 vG_offset;
layout(location = 3) in vec2 vB_offset;
layout(location = 4) in float vDo_shadow_mode;
layout(location = 5) in float vDecon_or_JustShadow;
layout(location = 6) in float vShow_artifact_mask;
layout(location = 0) out vec4 FragColor;
layout(set = 0, binding = 3) uniform sampler2D FXAA_pass;
layout(set = 0, binding = 4) uniform sampler2D flick_and_noise_pass;
layout(set = 0, binding = 5) uniform sampler2D colortools_and_ntsc_pass;
#define bandwidth_mhz_Y_ntsc 4.2
#define bandwidth_mhz_I 1.5
#define bandwidth_mhz_Q 0.5
const mat3 mat3_RGB2YIQ = mat3(
0.2989, 0.5959, 0.2115,
0.5870, -0.2744, -0.5229,
0.1140, -0.3216, 0.3114);
const mat3 mat3_YIQ2RGB = mat3(
1.0, 1.0, 1.0,
0.956, -0.2720, -1.1060,
0.6210, -0.6474, 1.7046);
//https://www.sciencedirect.com/topics/computer-science/color-subcarrier
#define bandwidth_mhz_Y_pal 5.0
#define bandwidth_mhz_U 1.3
#define bandwidth_mhz_V 1.3
const mat3 mat3_RGB2YUV = mat3(
0.299, 0.587, 0.114,
-0.14713, -0.28886, 0.436,
0.615, -0.514991, -0.10001);
const mat3 mat3_YUV2RGB = mat3(
1.000, 0.000, 1.13983,
1.000,-0.39465,-0.58060,
1.000, 2.03211, 0.00000);
vec4 deconvergence_shadow(vec2 coords, sampler2D in_texture, vec4 source_pixel) {
//Emulates deconvergence or shadowing for dot matrix screens like Gameboy.
//Since both effects should not coexist in real life, we use a singe function for both.
//The function stays more or less the same, but rgb offsets calculated in vertex shader
//will be the same for shadow mode. Hopefully shader texture cache will take care of
//multiple sampling for the same tex coords.
vec3 pixel_offset;
//Since we are sampling from a "clamp-to-border" texture, ensure we've no black bleeding
//by clamping the texture coords
coords = clamp(coords,params.SourceSize.zw*1 ,1.0-params.SourceSize.zw);
pixel_offset.r=texture(in_texture,coords + vR_offset).r;
pixel_offset.g=texture(in_texture,coords + vG_offset).g;
pixel_offset.b=texture(in_texture,coords + vB_offset).b;
vec3 deconvergence_mode = mix(source_pixel.rgb, pixel_offset, OFFSET_STRENGTH);
vec3 shadow_mode = source_pixel.rgb - max( (source_pixel.rgb - pixel_offset), vec3(0.0) ) * DOT_M_SHADOW_STR;
return vec4( mix_step(deconvergence_mode, shadow_mode, vDo_shadow_mode),
source_pixel.a );
}
vec4 deconvergence_shadow_wrap(vec2 coord, vec4 source_pixel) {
if ( DO_FXAA == 1.0)
return deconvergence_shadow(coord, FXAA_pass, source_pixel);
else
return deconvergence_shadow(coord, flick_and_noise_pass, source_pixel);
}
#define SIDE_RIGHT 1
#define SIDE_LEFT -1
vec3 pixel_bleed_side_NTSC(vec4 pixel_in, vec2 co, float size, sampler2D in_texture, vec4 sourcesize2) {
//apply strength modifier to blur ntsc artifacts more.
float strength_modifier = mix(1.0, pixel_in.a, DO_NTSC_ARTIFACTS);
float w = SAT_BLEED_STRENGTH * clamp (strength_modifier, (1-NTSC_ARTF_NOBLEED), 1.0 );
vec3 blur_YIQ_l = pixel_in.rgb * mat3_RGB2YIQ; //Work in YIQ space
vec3 blur_YIQ_r = blur_YIQ_l;
if (vShow_artifact_mask == 1.0) return vec3(strength_modifier);
vec2 off_l = vec2(SIDE_LEFT * sourcesize2.z,0.0);
vec2 off_r = vec2(SIDE_RIGHT * sourcesize2.z,0.0);
for ( float i=1 ; i <= size ; i++ ){
w=w/SAT_BLEED_FALLOFF; //w = w * exp(i*i*(1-SAT_BLEED_FALLOFF)*0.1);
vec3 smp_YIQ = texture(in_texture, co - i*off_l ).rgb * mat3_RGB2YIQ;
//Blur Y, I and Q
blur_YIQ_l = mix(blur_YIQ_l.xyz, smp_YIQ.xyz, w/vec3( bandwidth_mhz_Y_ntsc, bandwidth_mhz_I, bandwidth_mhz_Q));
smp_YIQ = texture(in_texture, co - i*off_r ).rgb * mat3_RGB2YIQ;
//Blur Y, I and Q
blur_YIQ_r = mix(blur_YIQ_r.xyz, smp_YIQ.xyz, w/vec3( bandwidth_mhz_Y_ntsc, bandwidth_mhz_I, bandwidth_mhz_Q));
}
vec3 blur_RGB_l = blur_YIQ_l * mat3_YIQ2RGB; //return to RGB colorspace
vec3 blur_RGB_r = blur_YIQ_r * mat3_YIQ2RGB; //return to RGB colorspace
// Clamping min to 0.0 is needed for nvidia to avoid bad graphical glitches, why?
return max(mix(blur_RGB_l,blur_RGB_r,0.5), 0.0);
}
vec3 pixel_bleed_side_PAL(vec3 pixel_in, vec2 co, float size, sampler2D in_texture, vec4 sourcesize2) {
float w = SAT_BLEED_STRENGTH;
vec3 blur_YUV_l = pixel_in * mat3_RGB2YUV; //Work in YIQ space
vec3 blur_YUV_r = blur_YUV_l;
vec2 off_l = vec2(SIDE_LEFT * sourcesize2.z,0.0);
vec2 off_r = vec2(SIDE_RIGHT * sourcesize2.z,0.0);
for ( float i=1 ; i <= size ; i++ ){
w=w/SAT_BLEED_FALLOFF;
vec2 off = vec2(sourcesize2.z*i,0.0);
vec3 smp_YUV = texture(in_texture, co - i*off_l ).rgb * mat3_RGB2YUV;
//Blur Y, U, V
blur_YUV_l = mix(blur_YUV_l.xyz, smp_YUV.xyz, w/vec3( bandwidth_mhz_Y_pal, bandwidth_mhz_U, bandwidth_mhz_V));
smp_YUV = texture(in_texture, co - i*off_r ).rgb * mat3_RGB2YUV;
//Blur Y, U, V
blur_YUV_r = mix(blur_YUV_r.xyz, smp_YUV.xyz, w/vec3( bandwidth_mhz_Y_pal, bandwidth_mhz_U, bandwidth_mhz_V));
}
vec3 blur_RGB_l = blur_YUV_l * mat3_YUV2RGB; //return to RGB colorspace
vec3 blur_RGB_r = blur_YUV_r * mat3_YUV2RGB; //return to RGB colorspace
// Clamping min to 0.0 is needed for nvidia to avoid bad graphical glitches, why?
return max(mix(blur_RGB_l,blur_RGB_r,0.5), 0.0);
}
vec4 pixel_bleed(vec4 pixel_in, vec2 co, sampler2D in_texture, vec4 sourcesize2) {
float size = SAT_BLEED_SIZE;
#ifdef D3D_WORKAROUND
// d3d11 compiler complains it is unable to unroll the loop because it is
// clueless about the maximum pragma value; d3d12 hangs somewhere.
size = min(5.0,SAT_BLEED_SIZE) ;
#else
size = min(20.0,SAT_BLEED_SIZE) ;
#endif
if (SAT_BLEED_PAL == 1.0) {
return vec4 ( pixel_bleed_side_PAL(pixel_in.rgb, co, size, in_texture, sourcesize2), pixel_in.a);
} else {
return vec4 ( pixel_bleed_side_NTSC(pixel_in.rgba, co, size, in_texture, sourcesize2), pixel_in.a);
}
}
void main() {
float pixel_alpha_ntsc_artifacts; // <- this carries ntsc artifacts needed by glow to modulate blur there.
vec4 pixel_out;
if ( DO_FXAA == 1.0) {
pixel_out = texture(FXAA_pass, vTexCoord );
} else {
pixel_out = texture(flick_and_noise_pass, vTexCoord);
}
//shift
if (vDecon_or_JustShadow > 0.0 )
pixel_out = deconvergence_shadow_wrap(vTexCoord, pixel_out);
//..and bleed
if (DO_SAT_BLEED > 0.0)
pixel_out = pixel_bleed(pixel_out, vTexCoord, flick_and_noise_pass, global.flick_and_noise_passSize);
//pre-gamma if needed by glow.
if (DO_CCORRECTION > 0.5)
pixel_out.rgb = pow(pixel_out.rgb, vec3(IN_GLOW_GAMMA));
FragColor = pixel_out;
}