Update crt-pi.slang

Fix crt-pi.

Also did some MAD optimizations, combine some lines, massage the formatting a bit so it's more readable
This commit is contained in:
rz5 2016-07-30 21:45:00 +01:00 committed by GitHub
parent 915132eb93
commit 3c1dfb6fe5

View file

@ -2,26 +2,31 @@
layout(std140, set = 0, binding = 0) uniform UBO
{
mat4 MVP;
vec4 OutputSize;
vec4 OriginalSize;
vec4 SourceSize;
mat4 MVP;
vec4 OutputSize;
vec4 OriginalSize;
vec4 SourceSize;
} global;
#define CURVATURE_X 0.10
#define CURVATURE_Y 0.25
#define MASK_BRIGHTNESS 0.70
#define SCANLINE_WEIGHT 6.0
#define CURVATURE_X 0.10
#define CURVATURE_Y 0.25
#define MASK_BRIGHTNESS 0.70
#define SCANLINE_WEIGHT 6.0
#define SCANLINE_GAP_BRIGHTNESS 0.12
#define BLOOM_FACTOR 1.5
#define INPUT_GAMMA 2.4
#define OUTPUT_GAMMA 2.2
#define BLOOM_FACTOR 1.5
#define INPUT_GAMMA 2.4
#define OUTPUT_GAMMA 2.2
/* MASK_TYPE: 0 = none, 1 = green/magenta, 2 = trinitron(ish) */
#define MASK_TYPE 2
#define SCANLINES
#define CURVATURE
//#define FAKE_GAMMA
//#define GAMMA
//#define SHARPER
#define MULTISAMPLE
#define MASK_TYPE 2
/*
crt-pi - A Raspberry Pi friendly CRT shader.
@ -30,14 +35,37 @@ layout(std140, set = 0, binding = 0) uniform UBO
under the terms of the GNU General Public License as published by the Free
Software Foundation; either version 2 of the License, or (at your option)
any later version.
Notes:
This shader is designed to work well on Raspberry Pi GPUs (i.e. 1080P @ 60Hz on a game with a 4:3 aspect ratio). It pushes the Pi's GPU hard and enabling some features will slow it down so that it is no longer able to match 1080P @ 60Hz. You will need to overclock your Pi to the fastest setting in raspi-config to get the best results from this shader: 'Pi2' for Pi2 and 'Turbo' for original Pi and Pi Zero. Note: Pi2s are slower at running the shader than other Pis, this seems to be down to Pi2s lower maximum memory speed. Pi2s don't quite manage 1080P @ 60Hz - they drop about 1 in 1000 frames. You probably won't notice this, but if you do, try enabling FAKE_GAMMA.
SCANLINES enables scanlines. You'll almost certainly want to use it with MULTISAMPLE to reduce moire effects. SCANLINE_WEIGHT defines how wide scanlines are (it is an inverse value so a higher number = thinner lines). SCANLINE_GAP_BRIGHTNESS defines how dark the gaps between the scan lines are. Darker gaps between scan lines make moire effects more likely.
GAMMA enables gamma correction using the values in INPUT_GAMMA and OUTPUT_GAMMA. FAKE_GAMMA causes it to ignore the values in INPUT_GAMMA and OUTPUT_GAMMA and approximate gamma correction in a way which is faster than true gamma whilst still looking better than having none. You must have GAMMA defined to enable FAKE_GAMMA.
CURVATURE distorts the screen by CURVATURE_X and CURVATURE_Y. Curvature slows things down a lot.
By default the shader uses linear blending horizontally. If you find this too blury, enable SHARPER.
This shader is designed to work well on Raspberry Pi GPUs (i.e. 1080P @ 60Hz on
a game with a 4:3 aspect ratio).
It pushes the Pi's GPU hard and enabling some features will slow it down so that
it is no longer able to match 1080P @ 60Hz.
You will need to overclock your Pi to the fastest setting in raspi-config to get
the best results from this shader: 'Pi2' for Pi2 and 'Turbo' for original Pi and
Pi Zero.
Note: Pi2s are slower at running the shader than other Pis, this seems to be
down to Pi2s lower maximum memory speed.
Pi2s don't quite manage 1080P @ 60Hz - they drop about 1 in 1000 frames.
You probably won't notice this, but if you do, try enabling FAKE_GAMMA.
SCANLINES enables scanlines.
You'll almost certainly want to use it with MULTISAMPLE to reduce moire effects.
SCANLINE_WEIGHT defines how wide scanlines are (it is an inverse value so a
higher number = thinner lines).
SCANLINE_GAP_BRIGHTNESS defines how dark the gaps between the scan lines are.
Darker gaps between scan lines make moire effects more likely.
GAMMA enables gamma correction using the values in INPUT_GAMMA and OUTPUT_GAMMA.
FAKE_GAMMA causes it to ignore the values in INPUT_GAMMA and OUTPUT_GAMMA and
approximate gamma correction in a way which is faster than true gamma whilst
still looking better than having none.
You must have GAMMA defined to enable FAKE_GAMMA.
CURVATURE distorts the screen by CURVATURE_X and CURVATURE_Y.
Curvature slows things down a lot.
By default the shader uses linear blending horizontally. If you find this too
blury, enable SHARPER.
BLOOM_FACTOR controls the increase in width for bright scanlines.
MASK_TYPE defines what, if any, shadow mask to use. MASK_BRIGHTNESS defines how much the mask type darkens the screen.
MASK_TYPE defines what, if any, shadow mask to use. MASK_BRIGHTNESS defines how
much the mask type darkens the screen.
*/
#pragma stage vertex
@ -48,9 +76,9 @@ layout(location = 1) out float filterWidth;
void main()
{
gl_Position = global.MVP * Position;
vTexCoord = TexCoord;
filterWidth = (global.SourceSize.y / global.OutputSize.y) / 3.0;
gl_Position = global.MVP * Position;
vTexCoord = TexCoord;
filterWidth = (global.SourceSize.y * global.OutputSize.w) * 0.333333333;
}
#pragma stage fragment
@ -66,122 +94,125 @@ vec2 barrelScale = 1.0 - (0.23 * CURVATURE_DISTORTION);
vec2 Distort(vec2 coord)
{
// coord *= screenScale; // not necessary in slang
coord -= vec2(0.5);
float rsq = coord.x * coord.x + coord.y * coord.y;
coord += coord * (CURVATURE_DISTORTION * rsq);
coord *= barrelScale;
if (abs(coord.x) >= 0.5 || abs(coord.y) >= 0.5)
coord = vec2(-1.0); // If out of bounds, return an invalid value.
else
{
coord += vec2(0.5);
// coord /= screenScale; // not necessary in slang
}
// coord *= screenScale; // not necessary in slang
coord -= vec2(0.5);
float rsq = coord.x * coord.x + coord.y * coord.y;
coord += coord * (CURVATURE_DISTORTION * rsq);
coord *= barrelScale;
if (abs(coord.x) >= 0.5 || abs(coord.y) >= 0.5)
coord = vec2(-1.0); // If out of bounds, return an invalid value.
else
{
coord += vec2(0.5);
// coord /= screenScale; // not necessary in slang
}
return coord;
return coord;
}
#endif
float CalcScanLineWeight(float dist)
{
return max(1.0-dist*dist*SCANLINE_WEIGHT, SCANLINE_GAP_BRIGHTNESS);
return max(1.0-dist*dist*SCANLINE_WEIGHT, SCANLINE_GAP_BRIGHTNESS);
}
float CalcScanLine(float dy)
{
float scanLineWeight = CalcScanLineWeight(dy);
float scanLineWeight = CalcScanLineWeight(dy);
#if defined(MULTISAMPLE)
scanLineWeight += CalcScanLineWeight(dy-filterWidth);
scanLineWeight += CalcScanLineWeight(dy+filterWidth);
scanLineWeight *= 0.3333333;
scanLineWeight += CalcScanLineWeight(dy - filterWidth);
scanLineWeight += CalcScanLineWeight(dy + filterWidth);
scanLineWeight *= 0.3333333;
#endif
return scanLineWeight;
return scanLineWeight;
}
void main()
{
vec2 texcoord = vTexCoord;
#if defined(CURVATURE)
vec2 texcoord = Distort(vTexCoord);
if (texcoord.x < 0.0)
FragColor = vec4(0.0);
else
#else
vec2 texcoord = vTexCoord;
texcoord = Distort(texcoord);
if (texcoord.x < 0.0)
{
FragColor = vec4(0.0);
return;
}
#endif
{
vec2 texcoordInPixels = texcoord * global.SourceSize.xy;
vec2 texcoordInPixels = texcoord * global.SourceSize.xy;
#if defined(SHARPER)
vec2 tempCoord = floor(texcoordInPixels) + 0.5;
vec2 coord = tempCoord / global.SourceSize.xy;
vec2 deltas = texcoordInPixels - tempCoord;
float scanLineWeight = CalcScanLine(deltas.y);
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 /= global.SourceSize.xy;
deltas *= signs;
vec2 tc = coord + deltas;
vec2 tempCoord = floor(texcoordInPixels) + 0.5;
vec2 coord = tempCoord * global.SourceSize.zw;
vec2 deltas = texcoordInPixels - tempCoord;
float scanLineWeight = CalcScanLine(deltas.y);
vec2 signs = sign(deltas);
deltas = abs(deltas) * 2.0;
deltas.x = deltas.x * deltas.x;
deltas.y = deltas.y * deltas.y * deltas.y;
deltas *= 0.5 * global.SourceSize.zw * signs;
vec2 tc = coord + deltas;
#else
float tempY = floor(texcoordInPixels.y) + 0.5;
float yCoord = tempY / global.SourceSize.y;
float dy = texcoordInPixels.y - tempY;
float scanLineWeight = CalcScanLine(dy);
float signY = sign(dy);
dy = dy * dy;
dy = dy * dy;
dy = dy * 8.0;
dy = dy / global.SourceSize.y;
dy = dy * signY;
vec2 tc = vec2(texcoord.x, yCoord + dy);
float tempCoord = floor(texcoordInPixels.y) + 0.5;
float coord = tempCoord * global.SourceSize.w;
float deltas = texcoordInPixels.y - tempCoord;
float scanLineWeight = CalcScanLine(deltas);
float signs = sign(deltas);
deltas = abs(deltas) * 2.0;
deltas = deltas * deltas * deltas;
deltas *= 0.5 * global.SourceSize.w * signs;
vec2 tc = vec2(texcoord.x, coord + deltas);
#endif
vec3 colour = texture(Source, tc).rgb;
vec3 colour = texture(Source, tc).rgb;
#if defined(SCANLINES)
#if defined(GAMMA)
#if defined(FAKE_GAMMA)
colour = colour * colour;
#else
colour = pow(colour, vec3(INPUT_GAMMA));
#endif
#endif
scanLineWeight *= BLOOM_FACTOR;
colour *= scanLineWeight;
#if defined(GAMMA)
#if defined(FAKE_GAMMA)
colour = sqrt(colour);
#else
colour = pow(colour, vec3(1.0/OUTPUT_GAMMA));
#endif
#if defined(GAMMA) && defined(FAKE_GAMMA)
colour = colour * colour;
#elif defined(GAMMA)
colour = pow(colour, vec3(INPUT_GAMMA));
#endif
/* Apply scanlines */
scanLineWeight *= BLOOM_FACTOR;
colour *= scanLineWeight;
#if defined(GAMMA) && defined(FAKE_GAMMA)
colour = sqrt(colour);
#elif defined(GAMMA)
colour = pow(colour, vec3(1.0/OUTPUT_GAMMA));
#endif
#if MASK_TYPE == 0
FragColor = vec4(colour, 1.0);
#else
#endif /* SCANLINES */
#if MASK_TYPE == 1
float whichMask = fract(vTexCoord.x * 0.5);
vec3 mask;
if (whichMask < 0.5)
mask = vec3(MASK_BRIGHTNESS, 1.0, MASK_BRIGHTNESS);
else
mask = vec3(1.0, MASK_BRIGHTNESS, 1.0);
float whichMask = fract(gl_FragCoord.x * 0.5);
vec3 mask = vec3(1.0);
if (whichMask < 0.5) mask.rb = vec2(MASK_BRIGHTNESS);
else mask.g = MASK_BRIGHTNESS;
colour *= mask;
#elif MASK_TYPE == 2
float whichMask = fract(vTexCoord.x * 0.3333333);
vec3 mask = vec3(MASK_BRIGHTNESS, MASK_BRIGHTNESS, MASK_BRIGHTNESS);
if (whichMask < 0.3333333)
mask.x = 1.0;
else if (whichMask < 0.6666666)
mask.y = 1.0;
else
mask.z = 1.0;
float whichMask = fract(gl_FragCoord.x * 0.3333333);
vec3 mask = vec3(MASK_BRIGHTNESS);
if (whichMask < 0.3333333) mask.r = 1.0;
else if (whichMask < 0.6666666) mask.g = 1.0;
else mask.b = 1.0;
colour *= mask;
#endif
FragColor = vec4(colour * mask, 1.0);
#endif
}
FragColor = vec4(colour, 1.0);
}