sdl3gpu

data/shaders/crtpi_frag.glsl

@@ -0,0 +1,152 @@
+#version 450
+
+// Vulkan GLSL fragment shader — CRT-Pi PostFX
+// Algoritmo de scanlines continuas con pesos gaussianos, bloom y máscara de fósforo.
+// Basado en el shader CRT-Pi original (GLSL 3.3), portado a GLSL 4.50 con parámetros uniformes.
+//
+// Compile: glslc -fshader-stage=frag --target-env=vulkan1.0 crtpi_frag.glsl -o crtpi_frag.spv
+//          xxd -i crtpi_frag.spv > ../../source/core/rendering/sdl3gpu/crtpi_frag_spv.h
+
+layout(location = 0) in  vec2 v_uv;
+layout(location = 0) out vec4 out_color;
+
+layout(set = 2, binding = 0) uniform sampler2D Texture;
+
+layout(set = 3, binding = 0) uniform CrtPiBlock {
+    // vec4 #0
+    float scanline_weight;          // Ajuste gaussiano de scanlines (default 6.0)
+    float scanline_gap_brightness;  // Brillo mínimo entre scanlines (default 0.12)
+    float bloom_factor;             // Factor de brillo en zonas iluminadas (default 3.5)
+    float input_gamma;              // Gamma de entrada — linealización (default 2.4)
+    // vec4 #1
+    float output_gamma;             // Gamma de salida — codificación (default 2.2)
+    float mask_brightness;          // Brillo sub-píxeles de la máscara (default 0.80)
+    float curvature_x;              // Distorsión barrel eje X (default 0.05)
+    float curvature_y;              // Distorsión barrel eje Y (default 0.10)
+    // vec4 #2
+    int mask_type;                  // 0=ninguna, 1=verde/magenta, 2=RGB fósforo
+    int enable_scanlines;           // 0 = off, 1 = on
+    int enable_multisample;         // 0 = off, 1 = on (antialiasing analítico de scanlines)
+    int enable_gamma;               // 0 = off, 1 = on
+    // vec4 #3
+    int enable_curvature;           // 0 = off, 1 = on
+    int enable_sharper;             // 0 = off, 1 = on
+    float texture_width;            // Ancho del canvas lógico en píxeles
+    float texture_height;           // Alto del canvas lógico en píxeles
+} u;
+
+// Distorsión barrel CRT
+vec2 distort(vec2 coord, vec2 screen_scale) {
+    vec2 curvature = vec2(u.curvature_x, u.curvature_y);
+    vec2 barrel_scale = 1.0 - (0.23 * curvature);
+    coord *= screen_scale;
+    coord -= vec2(0.5);
+    float rsq = coord.x * coord.x + coord.y * coord.y;
+    coord += coord * (curvature * rsq);
+    coord *= barrel_scale;
+    if (abs(coord.x) >= 0.5 || abs(coord.y) >= 0.5) {
+        return vec2(-1.0);  // fuera de pantalla
+    }
+    coord += vec2(0.5);
+    coord /= screen_scale;
+    return coord;
+}
+
+float calcScanLineWeight(float dist) {
+    return max(1.0 - dist * dist * u.scanline_weight, u.scanline_gap_brightness);
+}
+
+float calcScanLine(float dy, float filter_width) {
+    float weight = calcScanLineWeight(dy);
+    if (u.enable_multisample != 0) {
+        weight += calcScanLineWeight(dy - filter_width);
+        weight += calcScanLineWeight(dy + filter_width);
+        weight *= 0.3333333;
+    }
+    return weight;
+}
+
+void main() {
+    vec2 tex_size = vec2(u.texture_width, u.texture_height);
+
+    // filterWidth: equivalente al original (768.0 / TextureSize.y) / 3.0
+    float filter_width = (768.0 / u.texture_height) / 3.0;
+
+    vec2 texcoord = v_uv;
+
+    // Curvatura barrel opcional
+    if (u.enable_curvature != 0) {
+        texcoord = distort(texcoord, vec2(1.0, 1.0));
+        if (texcoord.x < 0.0) {
+            out_color = vec4(0.0, 0.0, 0.0, 1.0);
+            return;
+        }
+    }
+
+    vec2 texcoord_in_pixels = texcoord * tex_size;
+    vec2 tc;
+    float scan_line_weight;
+
+    if (u.enable_sharper != 0) {
+        // Modo SHARPER: filtrado bicúbico-like con subpixel sharpen
+        vec2 temp_coord = floor(texcoord_in_pixels) + 0.5;
+        tc = temp_coord / tex_size;
+        vec2 deltas = texcoord_in_pixels - temp_coord;
+        scan_line_weight = calcScanLine(deltas.y, filter_width);
+        vec2 signs = sign(deltas);
+        deltas.x *= 2.0;
+        deltas = deltas * deltas;
+        deltas.y = deltas.y * deltas.y;
+        deltas.x *= 0.5;
+        deltas.y *= 8.0;
+        deltas /= tex_size;
+        deltas *= signs;
+        tc = tc + deltas;
+    } else {
+        // Modo estándar
+        float temp_y     = floor(texcoord_in_pixels.y) + 0.5;
+        float y_coord    = temp_y / tex_size.y;
+        float dy         = texcoord_in_pixels.y - temp_y;
+        scan_line_weight = calcScanLine(dy, filter_width);
+        float sign_y     = sign(dy);
+        dy = dy * dy;
+        dy = dy * dy;
+        dy *= 8.0;
+        dy /= tex_size.y;
+        dy *= sign_y;
+        tc = vec2(texcoord.x, y_coord + dy);
+    }
+
+    vec3 colour = texture(Texture, tc).rgb;
+
+    if (u.enable_scanlines != 0) {
+        if (u.enable_gamma != 0) {
+            colour = pow(colour, vec3(u.input_gamma));
+        }
+        colour *= scan_line_weight * u.bloom_factor;
+        if (u.enable_gamma != 0) {
+            colour = pow(colour, vec3(1.0 / u.output_gamma));
+        }
+    }
+
+    // Máscara de fósforo
+    if (u.mask_type == 1) {
+        float which_mask = fract(gl_FragCoord.x * 0.5);
+        vec3 mask = (which_mask < 0.5)
+            ? vec3(u.mask_brightness, 1.0, u.mask_brightness)
+            : vec3(1.0, u.mask_brightness, 1.0);
+        colour *= mask;
+    } else if (u.mask_type == 2) {
+        float which_mask = fract(gl_FragCoord.x * 0.3333333);
+        vec3 mask = vec3(u.mask_brightness);
+        if (which_mask < 0.3333333)
+            mask.x = 1.0;
+        else if (which_mask < 0.6666666)
+            mask.y = 1.0;
+        else
+            mask.z = 1.0;
+        colour *= mask;
+    }
+
+    out_color = vec4(colour, 1.0);
+}

data/shaders/downscale.frag

@@ -0,0 +1,48 @@
+#version 450
+layout(location = 0) in  vec2 v_uv;
+layout(location = 0) out vec4 out_color;
+
+layout(set = 2, binding = 0) uniform sampler2D source;
+
+layout(set = 3, binding = 0) uniform DownscaleUniforms {
+    int algorithm;  // 0 = Lanczos2 (ventana 2, ±2 taps), 1 = Lanczos3 (ventana 3, ±3 taps)
+    float pad0;
+    float pad1;
+    float pad2;
+} u;
+
+// Kernel Lanczos normalizado: sinc(t) * sinc(t/a) para |t| < a, 0 fuera.
+float lanczos(float t, float a) {
+    t = abs(t);
+    if (t < 0.0001) { return 1.0; }
+    if (t >= a) { return 0.0; }
+    const float PI = 3.14159265358979;
+    float pt = PI * t;
+    return (a * sin(pt) * sin(pt / a)) / (pt * pt);
+}
+
+void main() {
+    vec2 src_size = vec2(textureSize(source, 0));
+    // Posición en coordenadas de texel (centros de texel en N+0.5)
+    vec2 p = v_uv * src_size;
+    vec2 p_floor = floor(p);
+
+    float a = (u.algorithm == 0) ? 2.0 : 3.0;
+    int win = int(a);
+
+    vec4 color = vec4(0.0);
+    float weight_sum = 0.0;
+
+    for (int j = -win; j <= win; j++) {
+        for (int i = -win; i <= win; i++) {
+            // Centro del texel (i,j) relativo a p_floor
+            vec2 tap_center = p_floor + vec2(float(i), float(j)) + 0.5;
+            vec2 offset = tap_center - p;
+            float w = lanczos(offset.x, a) * lanczos(offset.y, a);
+            color += texture(source, tap_center / src_size) * w;
+            weight_sum += w;
+        }
+    }
+
+    out_color = (weight_sum > 0.0) ? (color / weight_sum) : vec4(0.0, 0.0, 0.0, 1.0);
+}

data/shaders/postfx.frag

@@ -0,0 +1,142 @@
+#version 450
+
+// Vulkan GLSL fragment shader — PostFX effects
+// Used for SDL3 GPU API (SPIR-V path, Win/Linux).
+// Compile: glslc postfx.frag -o postfx.frag.spv
+//          xxd -i postfx.frag.spv > ../../source/core/rendering/sdl3gpu/postfx_frag_spv.h
+//
+// PostFXUniforms must match exactly the C++ struct in sdl3gpu_shader.hpp
+// (8 floats, 32 bytes, std140/scalar layout).
+
+layout(location = 0) in  vec2 v_uv;
+layout(location = 0) out vec4 out_color;
+
+layout(set = 2, binding = 0) uniform sampler2D scene;
+
+layout(set = 3, binding = 0) uniform PostFXUniforms {
+    float vignette_strength;
+    float chroma_strength;
+    float scanline_strength;
+    float screen_height;
+    float mask_strength;
+    float gamma_strength;
+    float curvature;
+    float bleeding;
+    float pixel_scale;   // physical pixels per logical pixel (vh / tex_height_)
+    float time;          // seconds since SDL init
+    float oversample;    // supersampling factor (1.0 = off, 3.0 = 3×SS)
+    float flicker;       // 0 = off, 1 = phosphor flicker ~50 Hz — 48 bytes total (3 × 16)
+} u;
+
+// YCbCr helpers for NTSC bleeding
+vec3 rgb_to_ycc(vec3 rgb) {
+    return vec3(
+         0.299*rgb.r + 0.587*rgb.g + 0.114*rgb.b,
+        -0.169*rgb.r - 0.331*rgb.g + 0.500*rgb.b + 0.5,
+         0.500*rgb.r - 0.419*rgb.g - 0.081*rgb.b + 0.5
+    );
+}
+vec3 ycc_to_rgb(vec3 ycc) {
+    float y  = ycc.x;
+    float cb = ycc.y - 0.5;
+    float cr = ycc.z - 0.5;
+    return clamp(vec3(
+        y + 1.402*cr,
+        y - 0.344*cb - 0.714*cr,
+        y + 1.772*cb
+    ), 0.0, 1.0);
+}
+
+void main() {
+    vec2 uv = v_uv;
+
+    // Curvatura barrel CRT
+    if (u.curvature > 0.0) {
+        vec2 c = uv - 0.5;
+        float rsq = dot(c, c);
+        vec2 dist = vec2(0.05, 0.1) * u.curvature;
+        vec2 barrelScale = vec2(1.0) - 0.23 * dist;
+        c += c * (dist * rsq);
+        c *= barrelScale;
+        if (abs(c.x) >= 0.5 || abs(c.y) >= 0.5) {
+            out_color = vec4(0.0, 0.0, 0.0, 1.0);
+            return;
+        }
+        uv = c + 0.5;
+    }
+
+    // Muestra base
+    vec3 base = texture(scene, uv).rgb;
+
+    // Sangrado NTSC — difuminado horizontal de crominancia.
+    // step = 1 pixel lógico de juego en UV (corrige SS: textureSize.x = game_w * oversample).
+    vec3 colour;
+    if (u.bleeding > 0.0) {
+        float tw   = float(textureSize(scene, 0).x);
+        float step = u.oversample / tw;  // 1 pixel lógico en UV
+        vec3 ycc    = rgb_to_ycc(base);
+        vec3 ycc_l2 = rgb_to_ycc(texture(scene, uv - vec2(2.0*step, 0.0)).rgb);
+        vec3 ycc_l1 = rgb_to_ycc(texture(scene, uv - vec2(1.0*step, 0.0)).rgb);
+        vec3 ycc_r1 = rgb_to_ycc(texture(scene, uv + vec2(1.0*step, 0.0)).rgb);
+        vec3 ycc_r2 = rgb_to_ycc(texture(scene, uv + vec2(2.0*step, 0.0)).rgb);
+        ycc.yz = (ycc_l2.yz + ycc_l1.yz*2.0 + ycc.yz*2.0 + ycc_r1.yz*2.0 + ycc_r2.yz) / 8.0;
+        colour = mix(base, ycc_to_rgb(ycc), u.bleeding);
+    } else {
+        colour = base;
+    }
+
+    // Aberración cromática (drift animado con time para efecto NTSC real)
+    float ca = u.chroma_strength * 0.005 * (1.0 + 0.15 * sin(u.time * 7.3));
+    colour.r = texture(scene, uv + vec2(ca, 0.0)).r;
+    colour.b = texture(scene, uv - vec2(ca, 0.0)).b;
+
+    // Corrección gamma (linealizar antes de scanlines, codificar después)
+    if (u.gamma_strength > 0.0) {
+        vec3 lin = pow(colour, vec3(2.4));
+        colour = mix(colour, lin, u.gamma_strength);
+    }
+
+    // Scanlines — 1 pixel físico oscuro por fila lógica.
+    // Modelo sustractivo: las filas de scanline se oscurecen, las demás no cambian.
+    // Esto evita el efecto de sobrebrillo en contenido con colores vivos.
+    if (u.scanline_strength > 0.0) {
+        float ps          = max(1.0, round(u.pixel_scale));
+        float frac_in_row = fract(uv.y * u.screen_height);
+        float row_pos     = floor(frac_in_row * ps);
+        float is_dark     = step(ps - 1.0, row_pos);
+        float scan        = mix(1.0, 0.0, is_dark);
+        colour           *= mix(1.0, scan, u.scanline_strength);
+    }
+
+    if (u.gamma_strength > 0.0) {
+        vec3 enc = pow(colour, vec3(1.0 / 2.2));
+        colour = mix(colour, enc, u.gamma_strength);
+    }
+
+    // Viñeta
+    vec2 d = uv - 0.5;
+    float vignette = 1.0 - dot(d, d) * u.vignette_strength;
+    colour *= clamp(vignette, 0.0, 1.0);
+
+    // Máscara de fósforo RGB — después de scanlines (orden original):
+    // filas brillantes saturadas → máscara invisible, filas oscuras → RGB visible.
+    if (u.mask_strength > 0.0) {
+        float whichMask = fract(gl_FragCoord.x * 0.3333333);
+        vec3 mask = vec3(0.80);
+        if (whichMask < 0.3333333)
+            mask.x = 1.0;
+        else if (whichMask < 0.6666666)
+            mask.y = 1.0;
+        else
+            mask.z = 1.0;
+        colour = mix(colour, colour * mask, u.mask_strength);
+    }
+
+    // Parpadeo de fósforo CRT (~50 Hz)
+    if (u.flicker > 0.0) {
+        float flicker_wave = sin(u.time * 100.0) * 0.5 + 0.5;
+        colour *= 1.0 - u.flicker * 0.04 * flicker_wave;
+    }
+
+    out_color = vec4(colour, 1.0);
+}

data/shaders/postfx.vert

@@ -0,0 +1,24 @@
+#version 450
+
+// Vulkan GLSL vertex shader — postfx full-screen triangle
+// Used for SDL3 GPU API (SPIR-V path, Win/Linux).
+// Compile: glslc postfx.vert -o postfx.vert.spv
+//          xxd -i postfx.vert.spv > ../../source/core/rendering/sdl3gpu/postfx_vert_spv.h
+
+layout(location = 0) out vec2 v_uv;
+
+void main() {
+    // Full-screen triangle (no vertex buffer needed)
+    const vec2 positions[3] = vec2[3](
+        vec2(-1.0, -1.0),
+        vec2( 3.0, -1.0),
+        vec2(-1.0,  3.0)
+    );
+    const vec2 uvs[3] = vec2[3](
+        vec2(0.0, 1.0),
+        vec2(2.0, 1.0),
+        vec2(0.0,-1.0)
+    );
+    gl_Position = vec4(positions[gl_VertexIndex], 0.0, 1.0);
+    v_uv        = uvs[gl_VertexIndex];
+}

data/shaders/upscale.frag

@@ -0,0 +1,15 @@
+#version 450
+
+// Vulkan GLSL fragment shader — Nearest-neighbour upscale pass
+// Used as the first render pass when supersampling is active.
+// Compile: glslc upscale.frag -o upscale.frag.spv
+//          xxd -i upscale.frag.spv > ../../source/core/rendering/sdl3gpu/upscale_frag_spv.h
+
+layout(location = 0) in  vec2 v_uv;
+layout(location = 0) out vec4 out_color;
+
+layout(set = 2, binding = 0) uniform sampler2D scene;
+
+void main() {
+    out_color = texture(scene, v_uv);
+}