Rearrangement and refactoring and optimizations and more accuracy

1 files changed, 202 insertions, 0 deletions

diff --git a/shaders/gl_crt_fragment.glsl b/shaders/gl_crt_fragment.glsl
new file mode 100644
index 0000000..73dcfec
--- /dev/null
+++ b/shaders/gl_crt_fragment.glsl
@@ -0,0 +1,202 @@
+out vec4 outcolor;
+in vec2 frag_texture_coord;
+
+uniform vec2 resolution;
+uniform vec2 src_image_size;
+uniform float brightness;
+uniform vec4 tone_data;
+uniform bool crt_emulation;
+uniform bool apply_mask;
+uniform sampler2D iChannel0;
+
+//==============================================================
+//                  CRTS SHADER PORTABILITY
+//==============================================================
+#define CrtsF1 float
+#define CrtsF2 vec2
+#define CrtsF3 vec3
+#define CrtsF4 vec4
+#define CrtsFractF1 fract
+#define CrtsRcpF1(x) (1.0/(x))
+#define CrtsSatF1(x) clamp((x),0.0,1.0)
+
+CrtsF1 CrtsMax3F1(CrtsF1 a, CrtsF1 b, CrtsF1 c) {
+	return max(a, max(b, c));
+}
+
+//==============================================================
+//                    FETCH FUNCTION
+//==============================================================
+CrtsF3 CrtsFetch(CrtsF2 uv) {
+	const float bias = 0.002533333;
+	return max(texture(iChannel0, uv, -16.0).rgb, vec3(bias));
+}
+
+//==============================================================
+//                      PHOSPHOR MASK
+//==============================================================
+CrtsF3 CrtsMask(CrtsF2 pos, CrtsF1 dark) {
+#ifdef CRTS_MASK_GRILLE
+	CrtsF3 m = CrtsF3(dark, dark, dark);
+	CrtsF1 x = CrtsFractF1(pos.x * (1.0/3.0));
+	if(x < (1.0/3.0)) m.r = 1.0;
+	else if(x < (2.0/3.0)) m.g = 1.0;
+	else m.b = 1.0;
+	return m;
+#endif
+
+#ifdef CRTS_MASK_GRILLE_LITE
+	CrtsF3 m = CrtsF3(1.0, 1.0, 1.0);
+	CrtsF1 x = CrtsFractF1(pos.x * (1.0/3.0));
+	if(x < (1.0/3.0)) m.r = dark;
+	else if(x < (2.0/3.0)) m.g = dark;
+	else m.b = dark;
+	return m;
+#endif
+
+#ifdef CRTS_MASK_NONE
+	return CrtsF3(1.0, 1.0, 1.0);
+#endif
+
+#ifdef CRTS_MASK_SHADOW
+	pos.x += pos.y * 3.0;
+	CrtsF3 m = CrtsF3(dark, dark, dark);
+	CrtsF1 x = CrtsFractF1(pos.x * (1.0/6.0));
+	if(x < (1.0/3.0)) m.r = 1.0;
+	else if(x < (2.0/3.0)) m.g = 1.0;
+	else m.b = 1.0;
+	return m;
+#endif
+}
+
+//==============================================================
+//                      CRTS FILTER
+//==============================================================
+CrtsF3 CrtsFilter(CrtsF2 ipos, CrtsF2 inputSizeDivOutputSize, CrtsF2 halfInputSize, CrtsF2 rcpInputSize, CrtsF2 rcpOutputSize, CrtsF2 twoDivOutputSize, CrtsF1 inputHeight, CrtsF2 warp, CrtsF1 thin, CrtsF1 blur, CrtsF1 mask, CrtsF4 tone) {
+	// Apply warp
+	// Convert to {-1 to 1} range
+	CrtsF2 pos = ipos * twoDivOutputSize - CrtsF2(1.0, 1.0);
+	// Distort pushes image outside {-1 to 1} range
+	pos *= CrtsF2(
+		1.0 + (pos.y * pos.y) * warp.x,
+		1.0 + (pos.x * pos.x) * warp.y);
+	// Vignette disabled - use rounded corners in composite shader instead
+	CrtsF1 vin = 1.0;
+	// Leave in {0 to inputSize}
+	pos = pos * halfInputSize + halfInputSize;
+
+	// Snap to center of first scanline
+	CrtsF1 y0 = floor(pos.y - 0.5) + 0.5;
+
+	// Snap to center of one of four pixels
+	CrtsF1 x0 = floor(pos.x - 1.5) + 0.5;
+	// Initial UV position
+	CrtsF2 p = CrtsF2(x0 * rcpInputSize.x, y0 * rcpInputSize.y);
+	// Fetch 4 nearest texels from 2 nearest scanlines
+	CrtsF3 colA0 = CrtsFetch(p);
+	p.x += rcpInputSize.x;
+	CrtsF3 colA1 = CrtsFetch(p);
+	p.x += rcpInputSize.x;
+	CrtsF3 colA2 = CrtsFetch(p);
+	p.x += rcpInputSize.x;
+	CrtsF3 colA3 = CrtsFetch(p);
+	p.y += rcpInputSize.y;
+	CrtsF3 colB3 = CrtsFetch(p);
+	p.x -= rcpInputSize.x;
+	CrtsF3 colB2 = CrtsFetch(p);
+	p.x -= rcpInputSize.x;
+	CrtsF3 colB1 = CrtsFetch(p);
+	p.x -= rcpInputSize.x;
+	CrtsF3 colB0 = CrtsFetch(p);
+
+	// Vertical filter - Scanline intensity using cosine wave
+	CrtsF1 off = pos.y - y0;
+	CrtsF1 pi2 = 6.28318530717958;
+	CrtsF1 hlf = 0.5;
+	CrtsF1 scanA = cos(min(0.5,   off  * thin     ) * pi2) * hlf + hlf;
+	CrtsF1 scanB = cos(min(0.5, (-off) * thin + thin) * pi2) * hlf + hlf;
+
+	// Horizontal kernel is simple gaussian filter
+	CrtsF1 off0 = pos.x - x0;
+	CrtsF1 off1 = off0 - 1.0;
+	CrtsF1 off2 = off0 - 2.0;
+	CrtsF1 off3 = off0 - 3.0;
+	CrtsF1 pix0 = exp2(blur * off0 * off0);
+	CrtsF1 pix1 = exp2(blur * off1 * off1);
+	CrtsF1 pix2 = exp2(blur * off2 * off2);
+	CrtsF1 pix3 = exp2(blur * off3 * off3);
+	CrtsF1 pixT = CrtsRcpF1(pix0 + pix1 + pix2 + pix3);
+	// Get rid of wrong pixels on edge
+	pixT *= vin;
+	scanA *= pixT;
+	scanB *= pixT;
+	// Apply horizontal and vertical filters
+	CrtsF3 color =
+		(colA0 * pix0 + colA1 * pix1 + colA2 * pix2 + colA3 * pix3) * scanA +
+		(colB0 * pix0 + colB1 * pix1 + colB2 * pix2 + colB3 * pix3) * scanB;
+
+	// Apply phosphor mask
+	color *= CrtsMask(ipos, mask);
+
+	// Tonal control, start by protecting from /0
+	CrtsF1 peak = max(1.0 / (256.0 * 65536.0),
+		CrtsMax3F1(color.r, color.g, color.b));
+	// Compute the ratios of {R,G,B}
+	CrtsF3 ratio = color * CrtsRcpF1(peak);
+	// Apply tonal curve to peak value
+	peak = pow(peak, tone.x);
+	peak = peak * CrtsRcpF1(peak * tone.y + tone.z);
+	// Apply saturation
+	ratio = pow(ratio, CrtsF3(tone.w, tone.w, tone.w));
+	// Reconstruct color
+	return ratio * peak;
+}
+
+//==============================================================
+//                         MAIN
+//==============================================================
+void main() {
+	// Add half_pixel offset to reduce aliasing from warp
+	vec2 half_pixel = 0.5 / src_image_size;
+	vec2 fragCoord = vec2(frag_texture_coord.x, 1.0 - frag_texture_coord.y) + half_pixel;
+
+	if(crt_emulation) {
+		outcolor.rgb = CrtsFilter(
+			fragCoord.xy * resolution + vec2(0.5),
+			src_image_size / resolution,
+			src_image_size * vec2(0.5),
+			1.0 / src_image_size,
+			1.0 / resolution,
+			2.0 / resolution,
+			src_image_size.y,
+			vec2(1.0 / 24.0, 1.0 / 16.0),
+			INPUT_THIN,
+			INPUT_BLUR,
+			INPUT_MASK,
+			tone_data
+		);
+		outcolor.rgb *= brightness;
+	} else {
+		outcolor.rgb = texture(iChannel0, fragCoord, -16.0).rgb;
+	}
+
+	// Apply rounded corner mask (matching composite shader) only when bloom is disabled
+	if(apply_mask) {
+		vec2 ipos = frag_texture_coord * resolution;
+		vec2 pos = ipos * (2.0 / resolution) - vec2(1.0);
+		vec2 warp = vec2(1.0 / 24.0, 1.0 / 16.0);
+		pos *= vec2(1.0 + (pos.y * pos.y) * warp.x, 1.0 + (pos.x * pos.x) * warp.y);
+		vec2 uv = (pos + vec2(1.0)) * 0.5;
+
+		float corner_radius = 0.05;
+		vec2 edge_distance = abs(pos) - vec2(1.0 - corner_radius);
+		float dist = length(max(edge_distance, 0.0));
+		float edge_softness = 0.003;
+		float mask = smoothstep(corner_radius + edge_softness, corner_radius - edge_softness, dist);
+		mask *= (uv.x >= 0.0 && uv.x <= 1.0 && uv.y >= 0.0 && uv.y <= 1.0) ? 1.0 : 0.0;
+
+		outcolor = vec4(outcolor.rgb * mask, 1.0);
+	} else {
+		outcolor = vec4(outcolor.rgb, 1.0);
+	}
+}