2026-09-07 20:08:08

Signed-off-by: lizzie <lizzie@eden-emu.dev>

src/video_core/host_shaders/fxaa.frag

@@ -7,69 +7,65 @@
 #version 460
 
 #ifdef VULKAN
-
 #define BINDING_COLOR_TEXTURE 1
-
 #else // ^^^ Vulkan ^^^ // vvv OpenGL vvv
-
 #define BINDING_COLOR_TEXTURE 0
-
 #endif
 
 layout (location = 0) in vec4 posPos;
-
 layout (location = 0) out vec4 frag_color;
-
 layout (binding = BINDING_COLOR_TEXTURE) uniform sampler2D input_texture;
 
 const float FXAA_SPAN_MAX = 8.0;
 const float FXAA_REDUCE_MUL = 1.0 / 8.0;
 const float FXAA_REDUCE_MIN = 1.0 / 128.0;
 
-#define FxaaTexLod0(t, p) textureLod(t, p, 0.0)
-#define FxaaTexOff(t, p, o) textureLodOffset(t, p, 0.0, o)
-
 vec3 FxaaPixelShader(vec4 posPos, sampler2D tex) {
-
-    vec3 rgbNW = FxaaTexLod0(tex, posPos.zw).xyz;
-    vec3 rgbNE = FxaaTexOff(tex, posPos.zw, ivec2(1,0)).xyz;
-    vec3 rgbSW = FxaaTexOff(tex, posPos.zw, ivec2(0,1)).xyz;
-    vec3 rgbSE = FxaaTexOff(tex, posPos.zw, ivec2(1,1)).xyz;
-    vec3 rgbM  = FxaaTexLod0(tex, posPos.xy).xyz;
-/*---------------------------------------------------------*/
+    mat4x3 rgb_matrix = mat4x3(
+        textureLod(tex, posPos.zw, 0.0).xyz,
+        textureLodOffset(tex, posPos.zw, 0.0, ivec2(1, 0)).xyz,
+        textureLodOffset(tex, posPos.zw, 0.0, ivec2(0, 1)).xyz,
+        textureLodOffset(tex, posPos.zw, 0.0, ivec2(1, 1)).xyz
+    );
+    vec3 rgbM  = textureLod(tex, posPos.xy, 0.0).xyz;
     vec3 luma = vec3(0.299, 0.587, 0.114);
-    float lumaNW = dot(rgbNW, luma);
-    float lumaNE = dot(rgbNE, luma);
-    float lumaSW = dot(rgbSW, luma);
-    float lumaSE = dot(rgbSE, luma);
+    // vec4(dot(m1, l), dot(m2, l), ...) => m * l
+    vec4 lume_per_rgb = luma * rgb_matrix;
     float lumaM  = dot(rgbM,  luma);
-/*---------------------------------------------------------*/
-    float lumaMin = min(lumaM, min(min(lumaNW, lumaNE), min(lumaSW, lumaSE)));
-    float lumaMax = max(lumaM, max(max(lumaNW, lumaNE), max(lumaSW, lumaSE)));
-/*---------------------------------------------------------*/
-    vec2 dir;
-    dir.x = -((lumaNW + lumaNE) - (lumaSW + lumaSE));
-    dir.y =  ((lumaNW + lumaSW) - (lumaNE + lumaSE));
-/*---------------------------------------------------------*/
-    float dirReduce = max(
-        (lumaNW + lumaNE + lumaSW + lumaSE) * (0.25 * FXAA_REDUCE_MUL),
-        FXAA_REDUCE_MIN);
-    float rcpDirMin = 1.0/(min(abs(dir.x), abs(dir.y)) + dirReduce);
-    dir = min(vec2( FXAA_SPAN_MAX,  FXAA_SPAN_MAX),
-          max(vec2(-FXAA_SPAN_MAX, -FXAA_SPAN_MAX),
-          dir * rcpDirMin)) / textureSize(tex, 0);
-/*--------------------------------------------------------*/
+    float lumaMin = min(lumaM, min(min(lume_per_rgb.x, lume_per_rgb.y), min(lume_per_rgb.z, lume_per_rgb.w)));
+    float lumaMax = max(lumaM, max(max(lume_per_rgb.x, lume_per_rgb.y), max(lume_per_rgb.z, lume_per_rgb.w)));
+    vec2 dir = vec2(
+        -((lume_per_rgb.x + lume_per_rgb.y) - (lume_per_rgb.z + lume_per_rgb.w)),
+        +((lume_per_rgb.x + lume_per_rgb.z) - (lume_per_rgb.y + lume_per_rgb.w))
+    );
+    float rcp_dir_min = 1.0 / (min(abs(dir.x), abs(dir.y)) + max(
+        (lume_per_rgb.x + lume_per_rgb.y + lume_per_rgb.z + lume_per_rgb.w) * (0.25 * FXAA_REDUCE_MUL),
+        FXAA_REDUCE_MIN
+    ));
+    dir = min(
+        vec2(FXAA_SPAN_MAX),
+        max(vec2(-FXAA_SPAN_MAX), dir * rcp_dir_min)
+    );
+    // Constants (calculating via division is faster)
+    mat4x2 const_dir = mat4x2(
+        (1.0 / 3.0 - 0.5) / textureSize(tex, 0),
+        (2.0 / 3.0 - 0.5) / textureSize(tex, 0),
+        (0.0 / 3.0 - 0.5) / textureSize(tex, 0),
+        (3.0 / 3.0 - 0.5) / textureSize(tex, 0)
+    );
+    // dir * const_dir[i] => dir * const_dir => vec4
     vec3 rgbA = (1.0 / 2.0) * (
-        FxaaTexLod0(tex, posPos.xy + dir * (1.0 / 3.0 - 0.5)).xyz +
-        FxaaTexLod0(tex, posPos.xy + dir * (2.0 / 3.0 - 0.5)).xyz);
-    vec3 rgbB = rgbA * (1.0 / 2.0) + (1.0 / 4.0) * (
-        FxaaTexLod0(tex, posPos.xy + dir * (0.0 / 3.0 - 0.5)).xyz +
-        FxaaTexLod0(tex, posPos.xy + dir * (3.0 / 3.0 - 0.5)).xyz);
+        textureLod(tex, fma(dir, const_dir[0], posPos.xy), 0.0).xyz +
+        textureLod(tex, fma(dir, const_dir[1], posPos.xy), 0.0).xyz
+    );
+    vec3 rgbB = fma(rgbA, vec3(1.0 / 2.0), (1.0 / 4.0) * (
+        textureLod(tex, fma(dir, const_dir[2], posPos.xy), 0.0).xyz +
+        textureLod(tex, fma(dir, const_dir[3], posPos.xy), 0.0).xyz
+    ));
     float lumaB = dot(rgbB, luma);
-    if((lumaB < lumaMin) || (lumaB > lumaMax)) return rgbA;
-    return rgbB;
+    return ((lumaB < lumaMin) || (lumaB > lumaMax)) ? rgbA : rgbB;
 }
 
 void main() {
-  frag_color = vec4(FxaaPixelShader(posPos, input_texture), texture(input_texture, posPos.xy).a);
+    frag_color = vec4(FxaaPixelShader(posPos, input_texture), texture(input_texture, posPos.xy).a);
 }

src/video_core/host_shaders/fxaa.vert

@@ -7,31 +7,35 @@ out gl_PerVertex {
     vec4 gl_Position;
 };
 
-const vec2 vertices[3] =
-    vec2[3](vec2(-1,-1), vec2(3,-1), vec2(-1, 3));
-
 layout (location = 0) out vec4 posPos;
 
 #ifdef VULKAN
-
 #define BINDING_COLOR_TEXTURE 0
 #define VERTEX_ID gl_VertexIndex
-
 #else // ^^^ Vulkan ^^^ // vvv OpenGL vvv
-
 #define BINDING_COLOR_TEXTURE 0
 #define VERTEX_ID gl_VertexID
-
 #endif
 
 layout (binding = BINDING_COLOR_TEXTURE) uniform sampler2D input_texture;
 
 const float FXAA_SUBPIX_SHIFT = 0;
-
 void main() {
-  vec2 vertex = vertices[VERTEX_ID];
-  gl_Position = vec4(vertex, 0.0, 1.0);
-  vec2 vert_tex_coord = (vertex + 1.0) / 2.0;
-  posPos.xy = vert_tex_coord;
-  posPos.zw = vert_tex_coord - (0.5 + FXAA_SUBPIX_SHIFT) / textureSize(input_texture, 0);
+    // 0b00 -> (-1, -1) 0b01 -> (-1, 3) 0b10 -> (3, -1)
+    // ((x * 4) - 1) => 3
+    // :: (v + 1) / 2 => ((x * 4) + 1) / 2
+    // => (x * 4 + 1) / 2 => x * (4 / 2) + 1 / 2
+    // :: (v+1)/2 => v/2 + 1/2 => v*(1/2) + (1/2)
+    gl_Position = vec4(vec2(
+        float(((VERTEX_ID & 1) << 2) - 1),
+        float(((VERTEX_ID & 2) << 1) - 1)
+    ), 0.0, 1.0);
+    posPos = vec2(
+        float((VERTEX_ID & 1) << 1),
+        float((VERTEX_ID & 2) << 0)
+    ).xyxy - vec4(
+        0.0,
+        0.0,
+        (0.5 + FXAA_SUBPIX_SHIFT) / textureSize(input_texture, 0)
+    );
 }