#version 450 #include "config.inc" #include "includes/functions.include.slang" #define AA_MODE_FXAA // 110 fps // on fastest preset (good for 2x upscale) //#define AA_MODE_SCALE2XPLUS // 111.2 fps //good general purpose, very light, but blurs edges too much? or not?; unsharp mask does not help. //#define AA_MODE_2XSCALEHQ // 109.5 //used to blur too much shades, tweaked better now. rounder look than ddtxbrlv1, but rounds almost any corner. #define AA_MODE_DDTXBRLV1 // 108.4 // a bit blurry, parameters do help, yes, done. //#define AA_MODE_SUPER2XSAI // 106.5 // less Dirty. #ifndef D3D_WORKAROUND #define FPS_ESTIMATE_PASS halo_pre_gamma_passFeedback #else #define FPS_ESTIMATE_PASS Original #endif #pragma stage vertex layout(location = 0) in vec4 Position; layout(location = 1) in vec2 TexCoord; layout(location = 0) noperspective out vec2 vTexCoord; layout(location = 3) noperspective out vec2 vDynamicSeed; layout(location = 5) noperspective out vec4 vpt1; layout(location = 6) noperspective out vec4 vpt2; layout(location = 7) noperspective out vec4 vhqt1; layout(location = 8) noperspective out vec4 vhqt2; layout(location = 9) noperspective out vec4 vhqt3; layout(location = 10) noperspective out vec4 vhqt4; layout(location = 11) noperspective out vec4 vxt1; layout(location = 12) flat out float vDo_flickering; layout(location = 13) noperspective out vec2 vFlickerCoords; void main() { gl_Position = global.MVP * Position; vTexCoord = TexCoord; //noise vDynamicSeed = mod(params.FrameCount, 120.0001) * vTexCoord; //flickering helper vDo_flickering = float ( scanline_have_to_flicker(is_interlaced()) ) ; #ifdef FLICKER_IN_MOTION vDo_flickering = 1.0; #endif if (vDo_flickering == 1.0) { //Flicker one over 3 frames: float ModFlicker = params.FrameCount % 3; //Compute y flicker offset: float FlickerOffset = 0.0; float line_tick = is_interlaced() ? 1 : 2 ; if (ModFlicker == 1.0 ) FlickerOffset = params.OriginalSize.w/line_tick; else if (ModFlicker == 2.0) FlickerOffset = -params.OriginalSize.w/line_tick; //Export flicker coords for Fragment shader here: vFlickerCoords = vec2(TexCoord.x, TexCoord.y + FlickerOffset); } #ifdef AA_MODE_SCALE2XPLUS vec2 pd = params.SourceSize.zw; vpt1 = vTexCoord.xyxy + vec4( 0, -pd.y, -pd.x, 0); // B, D vpt2 = vTexCoord.xyxy + vec4( pd.x, 0, 0, pd.y); // F, H #endif #ifdef AA_MODE_2XSCALEHQ vec2 s = params.SourceSize.zw * 0.5; vec2 dg1 = vec2( s.x, s.y); vec2 dg2 = vec2(-s.x, s.y); vec2 dx = vec2(s.x, 0.0); vec2 dy = vec2(0.0, s.y); vhqt1 = vec4(vTexCoord - dg1, vTexCoord - dy); vhqt2 = vec4(vTexCoord - dg2, vTexCoord + dx); vhqt3 = vec4(vTexCoord + dg1, vTexCoord + dy); vhqt4 = vec4(vTexCoord + dg2, vTexCoord - dx); #endif #ifdef AA_MODE_DDTXBRLV1 vec2 xd = params.SourceSize.zw; vxt1.xy = vec2( xd.x, 0); // F vxt1.zw = vec2( 0, xd.y); // H #endif } #pragma stage fragment layout(location = 0) noperspective in vec2 vTexCoord; layout(location = 3) noperspective in vec2 vDynamicSeed; layout(location = 5) noperspective in vec4 vpt1; layout(location = 6) noperspective in vec4 vpt2; layout(location = 7) noperspective in vec4 vhqt1; layout(location = 8) noperspective in vec4 vhqt2; layout(location = 9) noperspective in vec4 vhqt3; layout(location = 10) noperspective in vec4 vhqt4; layout(location = 11) noperspective in vec4 vxt1; layout(location = 12) flat in float vDo_flickering; layout(location = 13) noperspective in vec2 vFlickerCoords; layout(location = 0) out vec4 FragColor; layout(set = 0, binding = 4) uniform sampler2D colortools_and_ntsc_pass; layout(set = 0, binding = 5) uniform sampler2D colortools_and_ntsc_passFeedback; layout(set = 0, binding = 6) uniform sampler2D FPS_ESTIMATE_PASS; //#define _KOKO_AA_PRELINEARIZE #define _KOKO_AA_PRELINEARIZE_IN (2.0) #define _KOKO_AA_PRELINEARIZE_OUT (1/_KOKO_AA_PRELINEARIZE_IN) vec4 texture_ovr_nearest( sampler2D tex, vec2 co) { //return texture(tex, co); vec4 c = texture_NEAREST(tex, co, params.SourceSize); #ifdef _KOKO_AA_PRELINEARIZE c.rgb = pow(c.rgb, vec3(_KOKO_AA_PRELINEARIZE_IN)); #endif return c; } vec4 textureLod_ovr(sampler2D tex, vec2 co, float lod) { vec4 c = textureLod(tex, co, lod); #ifdef _KOKO_AA_PRELINEARIZE c.rgb = pow(c.rgb, vec3(_KOKO_AA_PRELINEARIZE_IN)); #endif return c; } vec4 texture_ovr_nearest_lod0( sampler2D tex, vec2 co) { //return texture(tex, co); vec4 c = texture_NEAREST_lod(tex, co, params.SourceSize, 0.0); #ifdef _KOKO_AA_PRELINEARIZE c.rgb=pow(c.rgb, vec3(_KOKO_AA_PRELINEARIZE_IN)); #endif return c; } #ifdef AA_MODE_DDTXBRLV1 /* Hyllian's DDT-xBR-lv1 Shader Copyright (C) 2011-2022 Hyllian/Jararaca - sergiogdb@gmail.com Sharpness control - Copyright (c) 2022 Filippo Scognamiglio Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions: The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software. THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ // // set to 1.0 to use dynamic sharpening // #pragma parameter USE_DYNAMIC_SHARPNESS "Dynamic Sharpness [ 0FF | ON ]" 1.0 0.0 1.0 1.0 // // // Set to 1.0 to bias the interpolation towards sharpening // #pragma parameter USE_SHARPENING_BIAS "Sharpness Bias [ 0FF | ON ]" 1.0 0.0 1.0 1.0 // // // Minimum amount of sharpening in range [0.0, 1.0] // #pragma parameter DYNAMIC_SHARPNESS_MIN "Dynamic Sharpness Min" 0.0 0.0 1.0 0.1 // // // Maximum amount of sharpening in range [0.0, 1.0] // #pragma parameter DYNAMIC_SHARPNESS_MAX "Dynamic Sharpness Max" 0.3 0.0 1.0 0.1 // // // If USE_DYNAMIC_SHARPNESS is 0 apply this static sharpness // #pragma parameter STATIC_SHARPNESS "Static Sharpness" 0.5 0.0 1.0 0.1 // // #pragma parameter DDT_THRESHOLD "DDT Diagonal Threshold" 2.6 1.0 6.0 0.2 // // #define USE_DYNAMIC_SHARPNESS params.USE_DYNAMIC_SHARPNESS // #define USE_SHARPENING_BIAS params.USE_SHARPENING_BIAS // #define DYNAMIC_SHARPNESS_MIN (params.DYNAMIC_SHARPNESS_MIN * 0.5) // #define DYNAMIC_SHARPNESS_MAX (params.DYNAMIC_SHARPNESS_MAX * 0.5) // #define STATIC_SHARPNESS (params.STATIC_SHARPNESS * 0.5) // #define DDT_THRESHOLD params.DDT_THRESHOLD // #define WP1 4.0 #define WP2 1.0 #define WP3 -1.0 #define USE_DYNAMIC_SHARPNESS 1.0 #define USE_SHARPENING_BIAS 1.0 #define DYNAMIC_SHARPNESS_MIN 0.1 // default: (0.0 * 0.5) #define DYNAMIC_SHARPNESS_MAX 0.25 // default: (0.3 * 0.5) #define STATIC_SHARPNESS (0.5 * 0.5) #define DDT_THRESHOLD 2.6 const vec3 Y = vec3( 0.299, 0.587, 0.114); float luma(vec3 color) { return dot(color, Y); } float linearStep(float edge0, float edge1, float t) { return clamp((t - edge0) / (edge1 - edge0), 0.0, 1.0); } float sharpSmooth(float t, float sharpness) { return linearStep(sharpness, 1.0 - sharpness, t); } vec3 quadBilinear(vec3 a, vec3 b, vec3 c, vec3 d, vec2 p, float sharpness){ float x = sharpSmooth(p.x, sharpness); float y = sharpSmooth(p.y, sharpness); return mix(mix(a, b, x), mix(c, d, x), y); } // Fast computation of barycentric coordinates only in the sub-triangle 1 2 4 vec3 fastBarycentric(vec2 p, float sharpness) { float l0 = sharpSmooth(1.0 - p.x - p.y, sharpness); float l1 = sharpSmooth(p.x, sharpness); return vec3(l0, l1, 1.0 - l0 - l1); } vec3 triangleInterpolate(vec3 t1, vec3 t2, vec3 t3, vec3 t4, vec2 c, float sharpness) { // Alter colors and coordinates to compute the other triangle. bool altTriangle = 1.0 - c.x < c.y; vec3 cornerColor = altTriangle ? t3 : t1; //vec2 triangleCoords = altTriangle ? vec2(1.0 - c.y, 1.0 - c.x) : c; vec2 triangleCoords; if (altTriangle) { triangleCoords = vec2(1.0 - c.y, 1.0 - c.x); } else { triangleCoords = c; } vec3 weights = fastBarycentric(triangleCoords, sharpness); return weights.x * cornerColor + weights.y * t2 + weights.z * t4; } vec3 aa_ddtxbrlv1(sampler2D tex, vec2 co, vec4 pt1) { vec2 pos = fract(co*params.SourceSize.xy)-vec2(0.5); // pos = pixel position vec2 dir = sign(pos); // dir = pixel direction float lmax, lmin, contrast; float sharpness = STATIC_SHARPNESS; vec2 g1 = dir*pt1.xy; vec2 g2 = dir*pt1.zw; vec3 A = texture_ovr_nearest_lod0(tex, vTexCoord ).xyz; vec3 B = texture_ovr_nearest_lod0(tex, vTexCoord +g1 ).xyz; vec3 C = texture_ovr_nearest_lod0(tex, vTexCoord +g2).xyz; vec3 D = texture_ovr_nearest_lod0(tex, vTexCoord +g1+g2).xyz; vec3 A1 = texture_ovr_nearest_lod0(tex, vTexCoord -g2).xyz; vec3 B1 = texture_ovr_nearest_lod0(tex, vTexCoord +g1-g2).xyz; vec3 A0 = texture_ovr_nearest_lod0(tex, vTexCoord -g1 ).xyz; vec3 C0 = texture_ovr_nearest_lod0(tex, vTexCoord -g1+g2).xyz; vec3 B2 = texture_ovr_nearest_lod0(tex, vTexCoord +2*g1 ).xyz; vec3 D2 = texture_ovr_nearest_lod0(tex, vTexCoord +2*g1+ g2).xyz; vec3 C3 = texture_ovr_nearest_lod0(tex, vTexCoord +2*g2).xyz; vec3 D3 = texture_ovr_nearest_lod0(tex, vTexCoord +g1+2*g2).xyz; float a = luma(A); float b = luma(B); float c = luma(C); float d = luma(D); if (USE_DYNAMIC_SHARPNESS == 1.0) { lmax = max(max(a, b), max(c, d)); lmin = min(min(a, b), min(c, d)); contrast = (lmax - lmin) / (lmax + lmin + 0.05); if (USE_SHARPENING_BIAS == 1.0) contrast = sqrt(contrast); sharpness = mix(DYNAMIC_SHARPNESS_MIN, DYNAMIC_SHARPNESS_MAX, contrast); } float a1 = luma(A1); float b1 = luma(B1); float a0 = luma(A0); float c0 = luma(C0); float b2 = luma(B2); float d2 = luma(D2); float c3 = luma(C3); float d3 = luma(D3); float p = abs(pos.x); float q = abs(pos.y); // A1 B1 // A0 A B B2 // C0 C D D2 // C3 D3 float wd1 = (WP1*abs(a-d) + WP2*(abs(b-a1) + abs(b-d2) + abs(c-a0) + abs(c-d3)) + WP3*(abs(a1-d2) + abs(a0-d3))); float wd2 = (WP1*abs(b-c) + WP2*(abs(a-b1) + abs(a-c0) + abs(d-b2) + abs(d-c3)) + WP3*(abs(b1-c0) + abs(b2-c3))); float irlv1 = (abs(a-b)+abs(a-c)+abs(d-c)+abs(d-b)); vec3 color; if ( ((wd1+0.1*DDT_THRESHOLD)*DDT_THRESHOLD < wd2) && (irlv1 > 0.0) ) { color = triangleInterpolate(B, D, C, A, vec2(q, 1-p), sharpness); } else if ( (wd1 > (wd2+0.1*DDT_THRESHOLD)*DDT_THRESHOLD) && (irlv1 > 0.0)) { color = triangleInterpolate(A, B, D, C, vec2(p, q), sharpness); } else color = quadBilinear(A, B, C, D, vec2(p, q), sharpness); return color; } #endif #ifdef AA_MODE_SUPER2XSAI /* Super2xSaI code */ /* Copied from the Dosbox source code */ /* Copyright (C) 2002-2007 The DOSBox Team */ /* License: GNU-GPL */ /* Adapted by guest(r) on 16.4.2007 */ const vec3 dt = vec3(65536.0,256.0,1.0); float GET_RESULT(float A, float B, float C, float D) { return (sign(abs(A-C)+abs(A-D)) - sign(abs(B-C)+abs(B-D))); } float reduce(vec3 color) { return dot(color,dt); } vec3 aa_super2x_sai(sampler2D tex, vec2 co) { vec2 OGL2Size = params.SourceSize.xy; vec2 OGL2InvSize = params.SourceSize.zw; // Calculating texel coordinates vec2 OGL2Pos = co * OGL2Size.xy; vec2 fp = fract(OGL2Pos); vec2 dx = vec2(OGL2InvSize.x,0.0); vec2 dy = vec2(0.0,OGL2InvSize.y); vec2 g1 = vec2( OGL2InvSize.x,OGL2InvSize.y); vec2 g2 = vec2(-OGL2InvSize.x,OGL2InvSize.y); vec2 pC4 = floor(OGL2Pos)/OGL2Size.xy + 0.5*OGL2InvSize; vec2 pC8 = pC4 + g1; // Reading the texels vec3 C0 = textureLod_ovr(tex,pC4-g1, 0.0).xyz; vec3 C1 = textureLod_ovr(tex,pC4-dy, 0.0).xyz; vec3 C2 = textureLod_ovr(tex,pC4-g2, 0.0).xyz; vec3 D3 = textureLod_ovr(tex,pC4-g2+dx, 0.0).xyz; vec3 C3 = textureLod_ovr(tex,pC4-dx, 0.0).xyz; vec3 C4 = textureLod_ovr(tex,pC4, 0.0).xyz; vec3 C5 = textureLod_ovr(tex,pC4+dx, 0.0).xyz; vec3 D4 = textureLod_ovr(tex,pC8-g2, 0.0).xyz; vec3 C6 = textureLod_ovr(tex,pC4+g2, 0.0).xyz; vec3 C7 = textureLod_ovr(tex,pC4+dy, 0.0).xyz; vec3 C8 = textureLod_ovr(tex,pC8, 0.0).xyz; vec3 D5 = textureLod_ovr(tex,pC8+dx, 0.0).xyz; vec3 D0 = textureLod_ovr(tex,pC4+g2+dy, 0.0).xyz; vec3 D1 = textureLod_ovr(tex,pC8+g2, 0.0).xyz; vec3 D2 = textureLod_ovr(tex,pC8+dy, 0.0).xyz; vec3 D6 = textureLod_ovr(tex,pC8+g1, 0.0).xyz; vec3 p00,p10,p01,p11; float c0 = reduce(C0);float c1 = reduce(C1); float c2 = reduce(C2);float c3 = reduce(C3); float c4 = reduce(C4);float c5 = reduce(C5); float c6 = reduce(C6);float c7 = reduce(C7); float c8 = reduce(C8);float d0 = reduce(D0); float d1 = reduce(D1);float d2 = reduce(D2); float d3 = reduce(D3);float d4 = reduce(D4); float d5 = reduce(D5);float d6 = reduce(D6); if (c7 == c5 && c4 != c8) { p11 = p01 = C7; } else if (c4 == c8 && c7 != c5) { p11 = p01 = C4; } else if (c4 == c8 && c7 == c5) { float r; r = GET_RESULT(c5,c4,c6,d1); r+= GET_RESULT(c5,c4,c3,c1); r+= GET_RESULT(c5,c4,d2,d5); r+= GET_RESULT(c5,c4,c2,d4); if (r > 0.0) p11 = p01 = C5; else if (r < 0.0) p11 = p01 = C4; else { p11 = p01 = 0.5*(C4+C5); } } else { if (c5 == c8 && c8 == d1 && c7 != d2 && c8 != d0) p11 = 0.25*(3.0*C8+C7); else if (c4 == c7 && c7 == d2 && d1 != c8 && c7 != d6) p11 = 0.25*(3.0*C7+C8); else p11 = 0.5*(C7+C8); if (c5 == c8 && c5 == c1 && c4 != c2 && c5 != c0) p01 = 0.25*(3.0*C5+C4); else if (c4 == c7 && c4 == c2 && c1 != c5 && c4 != d3) p01 = 0.25*(3.0*C4+C5); else p01 = 0.5*(C4+C5); } if (c4 == c8 && c7 != c5 && c3 == c4 && c4 != d2) p10 = 0.5*(C7+C4); else if (c4 == c6 && c5 == c4 && c3 != c7 && c4 != d0) p10 = 0.5*(C7+C4); else p10 = C7; if (c7 == c5 && c4 != c8 && c6 == c7 && c7 != c2) p00 = 0.5*(C7+C4); else if (c3 == c7 && c8 == c7 && c6 != c4 && c7 != c0) p00 = 0.5*(C7+C4); else p00 = C4; // Distributing the final products vec3 color = 0.0.xxx; if (fp.x < 0.50) { if (fp.y < 0.50) color = p00; else color = p10;} else { if (fp.y < 0.50) color = p01; else color = p11;} return color; } #endif #ifdef AA_MODE_SCALE2XPLUS vec3 aa_scale2xplus(sampler2D tex, vec2 co, vec4 pt1, vec4 pt2) { /* Scale2xPlus shader - Copyright (C) 2007 guest(r) - guest.r@gmail.com - License: GNU-GPL The Scale2x algorithm: - Scale2x Homepage: http://scale2x.sourceforge.net/ - Copyright (C) 2001, 2002, 2003, 2004 Andrea Mazzoleni - License: GNU-GPL */ vec2 fp = fract(co*params.SourceSize.xy); // Reading the texels vec3 B = texture_ovr_nearest_lod0(tex, pt1.xy).xyz; vec3 D = texture_ovr_nearest_lod0(tex, pt1.zw).xyz; vec3 E = texture_ovr_nearest_lod0(tex, co).xyz; vec3 F = texture_ovr_nearest_lod0(tex, pt2.xy).xyz; vec3 H = texture_ovr_nearest_lod0(tex, pt2.zw).xyz; vec3 E0 = D == B && B != H && D != F ? D : E; vec3 E1 = B == F && B != H && D != F ? F : E; vec3 E2 = D == H && B != H && D != F ? D : E; vec3 E3 = H == F && B != H && D != F ? F : E; // Product interpolation return (E3*fp.x+E2*(1-fp.x))*fp.y+(E1*fp.x+E0*(1-fp.x))*(1-fp.y); } #endif #ifdef AA_MODE_2XSCALEHQ vec3 aa_2xscalehq(sampler2D tex, vec2 co, vec4 pt1, vec4 pt2, vec4 pt3, vec4 pt4) { float mx = 0.325; // start smoothing wt. float k = -0.250; // wt. decrease factor //does not seem to affect anything float max_w = 0.25 * 0.0; // max filter weigth // the lower, the less "debanding" float min_w =-0.05; // min filter weigth float lum_add = 0.25; // affects smoothing vec3 c00 = texture_ovr_nearest_lod0(tex, pt1.xy).xyz; vec3 c10 = texture_ovr_nearest_lod0(tex, pt1.zw).xyz; vec3 c20 = texture_ovr_nearest_lod0(tex, pt2.xy).xyz; vec3 c01 = texture_ovr_nearest_lod0(tex, pt4.zw).xyz; vec3 c11 = texture_ovr_nearest_lod0(tex, co).xyz; vec3 c21 = texture_ovr_nearest_lod0(tex, pt2.zw).xyz; vec3 c02 = texture_ovr_nearest_lod0(tex, pt4.xy).xyz; vec3 c12 = texture_ovr_nearest_lod0(tex, pt3.zw).xyz; vec3 c22 = texture_ovr_nearest_lod0(tex, pt3.xy).xyz; float md1 = dot(abs(c00 - c22), vec3(1.0)); float md2 = dot(abs(c02 - c20), vec3(1.0)); float w1 = dot(abs(c22 - c11), vec3(1.0)) * md2; float w2 = dot(abs(c02 - c11), vec3(1.0)) * md1; float w3 = dot(abs(c00 - c11), vec3(1.0)) * md2; float w4 = dot(abs(c20 - c11), vec3(1.0)) * md1; float t1 = w1 + w3; float t2 = w2 + w4; float ww = max(t1, t2) + 0.0001; c11 = (w1 * c00 + w2 * c20 + w3 * c22 + w4 * c02 + ww * c11) / (t1 + t2 + ww); float lc1 = k / (0.12 * dot(c10 + c12 + c11, vec3(1.0)) + lum_add); float lc2 = k / (0.12 * dot(c01 + c21 + c11, vec3(1.0)) + lum_add); w1 = clamp(lc1 * dot(abs(c11 - c10), vec3(1.0)) + mx, min_w, max_w); w2 = clamp(lc2 * dot(abs(c11 - c21), vec3(1.0)) + mx, min_w, max_w); w3 = clamp(lc1 * dot(abs(c11 - c12), vec3(1.0)) + mx, min_w, max_w); w4 = clamp(lc2 * dot(abs(c11 - c01), vec3(1.0)) + mx, min_w, max_w); return (w1 * c10) + (w2 * c21) + (w3 * c12) + (w4 * c01) + (1.0 - w1 - w2 - w3 - w4) * c11; } #endif vec3 pixel_flickering(vec3 pixel_cur) { /* Simulates the flickering effect of the interlaced screens. * As I remember, it was visible when a line and the next had high * luminosity differences. * So we need to sample the current line and the previous one * (eventually applying color corrections to both). * * Repeating the following: * On frame 0, return the "clean" pixel * On frame 1, mix the upper pixel with the current one * On frame 2, mix the lower pixel with the current one * * The effect of the mix is the flickering itself, and we modulate * the mix according to the luminance difference between the current * pixel and the mixed one. * * We choose to alternate on a period of 3, * (thus considering the upper pixel and the lower one) * or else the high pixel persistance of lcd displays wont allow * to see the effect (the lcd panel would just mix the pixels by itself (meh). */ vec3 flickline = texture(colortools_and_ntsc_pass,vFlickerCoords).rgb; //float lumdiff = abs( flickline.r + flickline.g + flickline.b - pixel_cur.r - pixel_cur.g - pixel_cur.b); //float lumdiff = abs( (flickline.r + flickline.g) + (flickline.b - pixel_cur.r) - (pixel_cur.g + pixel_cur.b) ); //1687 1754 2946 ASM PROOF: FASTER float lumdiff = abs( dot(flickline.rgb, vec3(1.0)) - dot(pixel_cur.rgb, vec3(1.0)));// 1685 1750 2948 //ASM PROOF: Faster on SIMD8 and SIMD16, slower on SIMD32 // disable flickering on 25 and 30fps content: if (low_fps(FPS_ESTIMATE_PASS)) lumdiff = 0.0; lumdiff = min(lumdiff * PIXELGRID_INTR_FLICK_POWR, 1.0); #ifdef FLICKER_IN_MOTION lumdiff = 0.4; vec3 prev_frame = texture(colortools_and_ntsc_passFeedback, vTexCoord).rgb; vec3 cur_frame = texture(colortools_and_ntsc_pass, vTexCoord).rgb; float is_moving = float(prev_frame != cur_frame); lumdiff *= is_moving; #endif return mix(pixel_cur.rgb,flickline.rgb,lumdiff); } #ifndef FXAA_PRESET #define FXAA_PRESET 2 #endif #if (FXAA_PRESET == 1) //Dumb preset, looks like others for 2x scaling, but skip search. #define FXAA_EDGE_THRESHOLD (1.0/8.0) #define FXAA_EDGE_THRESHOLD_MIN (1.0/24.0) #define FXAA_SEARCH_STEPS 0 #define FXAA_SEARCH_THRESHOLD (1.0/4.0) #define FXAA_SUBPIX_CAP (3.0/4.0) #define FXAA_SUBPIX_TRIM (1.0/8.0) #endif #if (FXAA_PRESET == 2) // Limited search, rounder #define FXAA_EDGE_THRESHOLD (1.0/8.0) #define FXAA_EDGE_THRESHOLD_MIN (1.0/24.0) #define FXAA_SEARCH_STEPS 4 #define FXAA_SEARCH_THRESHOLD (1.0/4.0) #undef FXAA_SUBPIX_TRIM //koko: brighter edges #define FXAA_SUBPIX_CAP (3.0/4.0) //koko: unused when FXAA_SUBPIX_TRIM is not defined #endif #if (FXAA_PRESET == 3) #define FXAA_EDGE_THRESHOLD (1.0/8.0) #define FXAA_EDGE_THRESHOLD_MIN (1.0/16.0) #define FXAA_SEARCH_STEPS 16 #define FXAA_SEARCH_THRESHOLD (1.0/4.0) #define FXAA_SUBPIX_CAP (3.0/4.0) #define FXAA_SUBPIX_TRIM (1.0/4.0) #endif #if (FXAA_PRESET == 4) #define FXAA_EDGE_THRESHOLD (1.0/8.0) #define FXAA_EDGE_THRESHOLD_MIN (1.0/24.0) #define FXAA_SEARCH_STEPS 24 #define FXAA_SEARCH_THRESHOLD (1.0/4.0) #define FXAA_SUBPIX_CAP (3.0/4.0) #define FXAA_SUBPIX_TRIM (1.0/4.0) #endif #if (FXAA_PRESET == 5) #define FXAA_EDGE_THRESHOLD (1.0/8.0) #define FXAA_EDGE_THRESHOLD_MIN (1.0/24.0) #define FXAA_SEARCH_STEPS 32 #define FXAA_SEARCH_THRESHOLD (1.0/4.0) #define FXAA_SUBPIX_CAP (3.0/4.0) #define FXAA_SUBPIX_TRIM (1.0/4.0) #endif #ifdef FXAA_SUBPIX_TRIM #define FXAA_SUBPIX_TRIM_SCALE (1.0/(1.0 - FXAA_SUBPIX_TRIM)) #endif // Return the luma, the estimation of luminance from rgb inputs. // This approximates luma using one FMA instruction, // skipping normalization and tossing out blue. // FxaaLuma() will range 0.0 to 2.963210702. float FxaaLuma(vec3 rgb) { //return rgb.y * (0.587/0.299) + rgb.x; return fma( rgb.y, (0.587/0.299), rgb.x ); //ASM_PROOF: SAME } vec3 FxaaLerp3(vec3 a, vec3 b, float amountOfA) { return (vec3(-amountOfA) * b) + ((a * vec3(amountOfA)) + b); } vec4 FxaaTexOff(sampler2D tex, vec2 pos, ivec2 off, vec4 sourcesize, vec4 outputsize) { float x = pos.x + float(off.x) * sourcesize.z; float y = pos.y + float(off.y) * sourcesize.w; vec2 fragmentShift = off * outputsize.zw; vec2 shiftedUV = pos + fragmentShift; vec2 nearestUV = floor(shiftedUV * sourcesize.xy) * sourcesize.zw + sourcesize.zw * 0.5; return textureLod_ovr(tex, nearestUV, 0.0); } // pos is the output of FxaaVertexShader interpolated across screen. // xy -> actual texture position {0.0 to 1.0} // sourcesize.zw should be a uniform equal to {1.0/frameWidth, 1.0/frameHeight} vec3 FxaaPixelShader(vec2 pos, sampler2D tex, vec4 sourcesize, vec4 outputsize) { vec3 rgbN = FxaaTexOff(tex, pos.xy, ivec2( 0,-1), sourcesize, outputsize).xyz; vec3 rgbW = FxaaTexOff(tex, pos.xy, ivec2(-1, 0), sourcesize, outputsize).xyz; vec3 rgbM = FxaaTexOff(tex, pos.xy, ivec2( 0, 0), sourcesize, outputsize).xyz; vec3 rgbE = FxaaTexOff(tex, pos.xy, ivec2( 1, 0), sourcesize, outputsize).xyz; vec3 rgbS = FxaaTexOff(tex, pos.xy, ivec2( 0, 1), sourcesize, outputsize).xyz; float lumaN = FxaaLuma(rgbN); float lumaW = FxaaLuma(rgbW); float lumaM = FxaaLuma(rgbM); float lumaE = FxaaLuma(rgbE); float lumaS = FxaaLuma(rgbS); float rangeMin = min(lumaM, min(min(lumaN, lumaW), min(lumaS, lumaE))); float rangeMax = max(lumaM, max(max(lumaN, lumaW), max(lumaS, lumaE))); float range = rangeMax - rangeMin; if(range < max(FXAA_EDGE_THRESHOLD_MIN, rangeMax * FXAA_EDGE_THRESHOLD)) { return rgbM; } vec3 rgbL = rgbN + rgbW + rgbM + rgbE + rgbS; //ASM_PROOF: USELESS parenthesis float lumaL = (lumaN + lumaW + lumaE + lumaS) * 0.25; float rangeL = abs(lumaL - lumaM); #ifdef FXAA_SUBPIX_TRIM float blendL = max(0.0, (rangeL / range) - FXAA_SUBPIX_TRIM) * FXAA_SUBPIX_TRIM_SCALE; blendL = min(FXAA_SUBPIX_CAP, blendL); #else float blendL = 0.0; #endif vec3 rgbNW = FxaaTexOff(tex, pos.xy, ivec2(-1,-1), sourcesize, outputsize).xyz; vec3 rgbNE = FxaaTexOff(tex, pos.xy, ivec2( 1,-1), sourcesize, outputsize).xyz; vec3 rgbSW = FxaaTexOff(tex, pos.xy, ivec2(-1, 1), sourcesize, outputsize).xyz; vec3 rgbSE = FxaaTexOff(tex, pos.xy, ivec2( 1, 1), sourcesize, outputsize).xyz; rgbL += (rgbNW + rgbNE + rgbSW + rgbSE); rgbL *= vec3(1.0/9.0); float lumaNW = FxaaLuma(rgbNW); float lumaNE = FxaaLuma(rgbNE); float lumaSW = FxaaLuma(rgbSW); float lumaSE = FxaaLuma(rgbSE); float edgeVert = abs((0.25 * lumaNW) + (-0.5 * lumaN) + (0.25 * lumaNE)) + abs((0.50 * lumaW ) + (-1.0 * lumaM) + (0.50 * lumaE )) + abs((0.25 * lumaSW) + (-0.5 * lumaS) + (0.25 * lumaSE)); float edgeHorz = abs((0.25 * lumaNW) + (-0.5 * lumaW) + (0.25 * lumaSW)) + abs((0.50 * lumaN ) + (-1.0 * lumaM) + (0.50 * lumaS )) + abs((0.25 * lumaNE) + (-0.5 * lumaE) + (0.25 * lumaSE)); bool horzSpan = edgeHorz >= edgeVert; vec2 texel_size_scaled = sourcesize.zw * 0.5; //<-- KOKO, hack. float lengthSign = horzSpan ? -texel_size_scaled.y : -texel_size_scaled.x; if (!horzSpan) { lumaN = lumaW; lumaS = lumaE; } float gradientN = abs(lumaN - lumaM); float gradientS = abs(lumaS - lumaM); lumaN = (lumaN + lumaM) * 0.5; lumaS = (lumaS + lumaM) * 0.5; if (gradientN < gradientS) { lumaN = lumaS; lumaN = lumaS; gradientN = gradientS; lengthSign *= -1.0; } vec2 posN; posN.x = pos.x + (horzSpan ? 0.0 : lengthSign * 0.5); posN.y = pos.y + (horzSpan ? lengthSign * 0.5 : 0.0); gradientN *= FXAA_SEARCH_THRESHOLD; vec2 posP = posN; vec2 offNP = horzSpan ? vec2(texel_size_scaled.x, 0.0) : vec2(0.0, texel_size_scaled.y); float lumaEndN = lumaN; float lumaEndP = lumaN; bool doneN = false; bool doneP = false; posN += offNP * vec2(-1.0, -1.0); posP += offNP * vec2( 1.0, 1.0); for(int i = 0; i < FXAA_SEARCH_STEPS; i++) { vec2 posN_adjusted; vec2 posP_adjusted; if (horzSpan) { posN_adjusted.x = coords_NEAREST(posN, params.SourceSize).x; posN_adjusted.y = posN.y; posP_adjusted.x = coords_NEAREST(posP, params.SourceSize).x; posP_adjusted.y = posP.y; } else { posN_adjusted.x = posN.x; posN_adjusted.y = coords_NEAREST(posN, params.SourceSize).y; posP_adjusted.x = posP.x; posP_adjusted.y = coords_NEAREST(posP, params.SourceSize).y; } if(!doneN) { lumaEndN = FxaaLuma(textureLod_ovr(tex, posN_adjusted.xy,0.0).xyz); } if(!doneP) { lumaEndP = FxaaLuma(textureLod_ovr(tex, posP_adjusted.xy,0.0).xyz); } doneN = doneN || (abs(lumaEndN - lumaN) >= gradientN); doneP = doneP || (abs(lumaEndP - lumaN) >= gradientN); if (doneN && doneP) { break; } if (!doneN) { posN -= offNP; } if (!doneP) { posP += offNP; } } float dstN = horzSpan ? pos.x - posN.x : pos.y - posN.y; float dstP = horzSpan ? posP.x - pos.x : posP.y - pos.y; bool directionN = dstN < dstP; lumaEndN = directionN ? lumaEndN : lumaEndP; if ( ((lumaM - lumaN) < 0.0) == ((lumaEndN - lumaN) < 0.0) ) { lengthSign = 0.0; } float spanLength = (dstP + dstN); dstN = directionN ? dstN : dstP; float subPixelOffset = (0.5 + (dstN * (-1.0/spanLength))) * lengthSign; subPixelOffset *= 2.0; //<-- KOKO, hack. vec2 smpco; if (horzSpan) { smpco.x = pos.x + 0.0; smpco.y = coords_NEAREST( pos, params.SourceSize ).y + subPixelOffset; } else { smpco.x = coords_NEAREST( pos, params.SourceSize ).x + subPixelOffset; smpco.y = pos.y + 0.0; } vec3 rgbF = textureLod_ovr(tex, smpco, 0.0).rgb; return FxaaLerp3(rgbL, rgbF, blendL); } void main() { #ifndef D3D_WORKAROUND if (is_skipped_frame()) return; #endif vec4 pixel_out; pixel_out.rgb = texture_NEAREST(colortools_and_ntsc_pass, vTexCoord, params.SourceSize).rgb; pixel_out.a = texture(colortools_and_ntsc_pass, vTexCoord).a;//, params.SourceSize).a; // If we need to smooth waterfalls, it is good to smooth // their detection mask too, maybe using an higher lod ist good tho, // but we still require linear sampling. pixel_out.a = 0.0; if (DO_WTFL > 0.5) { pixel_out.a = texture(colortools_and_ntsc_pass, vTexCoord).a; } /* if (DO_FXAA == 1.0) { //pixel_out.rgb = FxaaPixelShader(vTexCoord, Source, params.SourceSize, params.OutputSize); // yes linearization pixel_out.rgb = aa_ddtxbrlv1(Source, vTexCoord, vxt1); // yes linearization //pixel_out.rgb = aa_2xscalehq(Source, vTexCoord , vhqt1, vhqt2, vhqt3, vhqt4); //no linearization //pixel_out.rgb = aa_scale2xplus(Source, vTexCoord, vpt1, vpt2); // no linearization //pixel_out.rgb = aa_super2x_sai(Source, vTexCoord); ยง// ?? linearization #ifdef _KOKO_AA_PRELINEARIZE pixel_out.rgb = pow(pixel_out.rgb, vec3(_KOKO_AA_PRELINEARIZE_OUT) ); #endif } */ if (DO_FXAA == 1.0) { pixel_out.rgb = FxaaPixelShader(vTexCoord, colortools_and_ntsc_pass, params.SourceSize, params.OutputSize); #ifdef _KOKO_AA_PRELINEARIZE pixel_out.rgb = pow(pixel_out.rgb, vec3(_KOKO_AA_PRELINEARIZE_OUT) ); #endif } else if (DO_FXAA == 2.0) { pixel_out.rgb = aa_ddtxbrlv1(colortools_and_ntsc_pass, vTexCoord, vxt1); #ifdef _KOKO_AA_PRELINEARIZE pixel_out.rgb = pow(pixel_out.rgb, vec3(_KOKO_AA_PRELINEARIZE_OUT) ); #endif } /* if (DO_FXAA == 1.0) { pixel_out.rgb = FxaaPixelShader(vTexCoord, colortools_and_ntsc_pass, params.SourceSize, params.OutputSize); } else if (DO_FXAA == 2.0) { pixel_out.rgb = aa_ddtxbrlv1(colortools_and_ntsc_pass, vTexCoord, vxt1); } else if (DO_FXAA == 3.0) { pixel_out.rgb = aa_2xscalehq(colortools_and_ntsc_pass, vTexCoord , vhqt1, vhqt2, vhqt3, vhqt4); } else if (DO_FXAA == 4.0) { pixel_out.rgb = aa_scale2xplus(colortools_and_ntsc_pass, vTexCoord, vpt1, vpt2); } else if (DO_FXAA == 5.0) { pixel_out.rgb = aa_super2x_sai(colortools_and_ntsc_pass, vTexCoord); } #ifdef _KOKO_AA_PRELINEARIZE pixel_out.rgb = pow(pixel_ozut.rgb, vec3(_KOKO_AA_PRELINEARIZE_OUT) ); #endif */ if ( !bdelta_render_enabled() ) { if (vDo_flickering==1.0) pixel_out.rgb = pixel_flickering(pixel_out.rgb).rgb; if (DO_RF_NOISE > 0.01 ) { //Unfortunately we need to clamp to min 0.0 again after applying noise //to avoid glitches on amd (and probably nvidia too). //It should be possible to change the noise function to just emit positive //numbers and pre-lowering the brightness of pixel_out in the very first pass tho. float rfnoise = get_rf_noise(vDynamicSeed); pixel_out.rgb += rfnoise; pixel_out.rgb = max(pixel_out.rgb, vec3(0.0)); } } //pixel_out.rgb = max(vec3(0.0), pixel_out.rgb); FragColor = pixel_out; }