#include <cstdint>
#include <cmath>
+#include <limits>
-// shader translated from SuperTuxKart data/shaders/rh.frag
-// https://github.com/supertuxkart/stk-code/blob/20ea7ca2711f0cbe5320b4877a5d332b3b935893/data/shaders/rh.frag
+namespace shader
+{
+constexpr float max(float a, float b) noexcept
+{
+ return a > b ? a : b;
+}
-// From http://graphics.cs.aueb.gr/graphics/research_illumination.html
-// "Real-Time Diffuse Global Illumination Using Radiance Hints"
-// paper and shader code
+struct vec2
+{
+ float x;
+ float y;
+ vec2() = default;
+ constexpr vec2(float v) noexcept : x(v), y(v)
+ {
+ }
+ constexpr vec2(float x, float y) noexcept : x(x), y(y)
+ {
+ }
+ friend constexpr vec2 operator *(vec2 a, float b) noexcept
+ {
+ return vec2(a.x * b, a.y * b);
+ }
+ friend constexpr vec2 operator *(float a, vec2 b) noexcept
+ {
+ return vec2(a * b.x, a * b.y);
+ }
+ friend constexpr vec2 operator +(vec2 a, vec2 b) noexcept
+ {
+ return vec2(a.x + b.x, a.y + b.y);
+ }
+ friend constexpr vec2 operator -(vec2 a, vec2 b) noexcept
+ {
+ return vec2(a.x - b.x, a.y - b.y);
+ }
+ friend constexpr vec2 operator /(vec2 a, vec2 b) noexcept
+ {
+ return vec2(a.x / b.x, a.y / b.y);
+ }
+};
struct vec3
{
constexpr vec3(float v) noexcept : x(v), y(v), z(v)
{
}
+ constexpr vec3(float x, float y, float z) noexcept : x(x), y(y), z(z)
+ {
+ }
+ constexpr vec3(vec2 xy, float z) noexcept : x(xy.x), y(xy.y), z(z)
+ {
+ }
+ constexpr vec3 xzy() const noexcept
+ {
+ return vec3(x, z, y);
+ }
+ constexpr vec3 xyz() const noexcept
+ {
+ return vec3(x, y, z);
+ }
+ constexpr vec3 rgb() const noexcept
+ {
+ return vec3(x, y, z);
+ }
+ constexpr float r() const noexcept
+ {
+ return x;
+ }
+ constexpr float g() const noexcept
+ {
+ return y;
+ }
+ constexpr float b() const noexcept
+ {
+ return z;
+ }
+ friend constexpr vec3 operator /(vec3 a, vec3 b) noexcept
+ {
+ return vec3(a.x / b.x, a.y / b.y, a.z / b.z);
+ }
+ friend constexpr vec3 operator *(vec3 a, vec3 b) noexcept
+ {
+ return vec3(a.x * b.x, a.y * b.y, a.z * b.z);
+ }
+ friend constexpr vec3 operator +(vec3 a, vec3 b) noexcept
+ {
+ return vec3(a.x + b.x, a.y + b.y, a.z + b.z);
+ }
+ friend constexpr vec3 operator -(vec3 a, vec3 b) noexcept
+ {
+ return vec3(a.x - b.x, a.y - b.y, a.z - b.z);
+ }
};
+struct vec4
+{
+ float x;
+ float y;
+ float z;
+ float w;
+ vec4() = default;
+ constexpr vec4(float v) noexcept : x(v), y(v), z(v), w(v)
+ {
+ }
+ constexpr vec4(float x, float y, float z, float w) noexcept : x(x), y(y), z(z), w(w)
+ {
+ }
+ constexpr vec4(vec2 xy, float z, float w) noexcept : x(xy.x), y(xy.y), z(z), w(w)
+ {
+ }
+ constexpr vec4(vec3 xyz, float w) noexcept : x(xyz.x), y(xyz.y), z(xyz.z), w(w)
+ {
+ }
+ constexpr vec3 xyz() const noexcept
+ {
+ return {x, y, z};
+ }
+ constexpr vec2 xy() const noexcept
+ {
+ return {x, y};
+ }
+ friend constexpr vec4 operator *(vec4 a, float b) noexcept
+ {
+ return vec4(a.x * b, a.y * b, a.z * b, a.w * b);
+ }
+ friend constexpr vec4 operator *(float a, vec4 b) noexcept
+ {
+ return vec4(a * b.x, a * b.y, a * b.z, a * b.w);
+ }
+ friend constexpr vec4 operator /(vec4 a, float b) noexcept
+ {
+ return vec4(a.x / b, a.y / b, a.z / b, a.w / b);
+ }
+ constexpr vec4 &operator /=(float v) noexcept
+ {
+ return *this = *this / v;
+ }
+ friend constexpr vec4 operator +(vec4 a, vec4 b) noexcept
+ {
+ return vec4(a.x + b.x, a.y + b.y, a.z + b.z, a.w + b.w);
+ }
+ friend constexpr vec4 operator -(vec4 a, vec4 b) noexcept
+ {
+ return vec4(a.x - b.x, a.y - b.y, a.z - b.z, a.w - b.w);
+ }
+ constexpr vec4 &operator +=(vec4 v) noexcept
+ {
+ return *this = *this + v;
+ }
+};
+
+constexpr float dot(vec3 a, vec3 b) noexcept
+{
+ return a.x * b.x + a.y * b.y + a.z * b.z;
+}
+
+inline float length(vec3 v) noexcept
+{
+ return std::sqrt(dot(v, v));
+}
+
+inline float distance(vec3 a, vec3 b) noexcept
+{
+ return length(a - b);
+}
+
+inline vec3 normalize(vec3 v) noexcept
+{
+ return v / length(v);
+}
+
+struct mat4
+{
+ float values[4][4] = {{1, 0, 0, 0}, {0, 1, 0, 0}, {0, 0, 1, 0}, {0, 0, 0, 1}};
+ constexpr mat4() noexcept {}
+ constexpr mat4(float value_0_0, float value_0_1, float value_0_2, float value_0_3,
+ float value_1_0, float value_1_1, float value_1_2, float value_1_3,
+ float value_2_0, float value_2_1, float value_2_2, float value_2_3,
+ float value_3_0, float value_3_1, float value_3_2, float value_3_3) noexcept
+ : values{
+ {value_0_0, value_0_1, value_0_2, value_0_3},
+ {value_1_0, value_1_1, value_1_2, value_1_3},
+ {value_2_0, value_2_1, value_2_2, value_2_3},
+ {value_3_0, value_3_1, value_3_2, value_3_3},
+ }
+ {
+ }
+ friend constexpr mat4 operator *(float a, const mat4 &b) noexcept
+ {
+ return mat4(a * b.values[0][0], a * b.values[0][1], a * b.values[0][2], a * b.values[0][3],
+ a * b.values[1][0], a * b.values[1][1], a * b.values[1][2], a * b.values[1][3],
+ a * b.values[2][0], a * b.values[2][1], a * b.values[2][2], a * b.values[2][3],
+ a * b.values[3][0], a * b.values[3][1], a * b.values[3][2], a * b.values[3][3]);
+ }
+ friend constexpr mat4 operator *(const mat4 &a, float b) noexcept
+ {
+ return mat4(a.values[0][0] * b, a.values[0][1] * b, a.values[0][2] * b, a.values[0][3] * b,
+ a.values[1][0] * b, a.values[1][1] * b, a.values[1][2] * b, a.values[1][3] * b,
+ a.values[2][0] * b, a.values[2][1] * b, a.values[2][2] * b, a.values[2][3] * b,
+ a.values[3][0] * b, a.values[3][1] * b, a.values[3][2] * b, a.values[3][3] * b);
+ }
+ friend constexpr vec4 operator *(const mat4 &m, vec4 v) noexcept
+ {
+ return vec4(m.values[0][2]*v.z+m.values[0][1]*v.y+m.values[0][0]*v.x
+ +m.values[0][3]*v.w,
+ m.values[1][2]*v.z+m.values[1][1]*v.y+m.values[1][0]*v.x
+ +m.values[1][3]*v.w,
+ m.values[2][2]*v.z+m.values[2][1]*v.y+m.values[2][0]*v.x
+ +m.values[2][3]*v.w,
+ m.values[3][2]*v.z+m.values[3][1]*v.y+m.values[3][0]*v.x
+ +m.values[3][3]*v.w);
+ }
+};
+
+constexpr float determinant(const mat4 &m) noexcept
+{
+ return ((m.values[0][1]*m.values[1][2]
+ -m.values[0][2]*m.values[1][1])
+ *m.values[2][0]
+ +(m.values[0][2]*m.values[1][0]
+ -m.values[0][0]*m.values[1][2])
+ *m.values[2][1]
+ +(m.values[0][0]*m.values[1][1]
+ -m.values[0][1]*m.values[1][0])
+ *m.values[2][2])
+ *m.values[3][3]
+ +((m.values[0][3]*m.values[1][1]
+ -m.values[0][1]*m.values[1][3])
+ *m.values[2][0]
+ +(m.values[0][0]*m.values[1][3]
+ -m.values[0][3]*m.values[1][0])
+ *m.values[2][1]
+ +(m.values[0][1]*m.values[1][0]
+ -m.values[0][0]*m.values[1][1])
+ *m.values[2][3])
+ *m.values[3][2]
+ +((m.values[0][2]*m.values[1][3]
+ -m.values[0][3]*m.values[1][2])
+ *m.values[2][0]
+ +(m.values[0][3]*m.values[1][0]
+ -m.values[0][0]*m.values[1][3])
+ *m.values[2][2]
+ +(m.values[0][0]*m.values[1][2]
+ -m.values[0][2]*m.values[1][0])
+ *m.values[2][3])
+ *m.values[3][1]
+ +((m.values[0][3]*m.values[1][2]
+ -m.values[0][2]*m.values[1][3])
+ *m.values[2][1]
+ +(m.values[0][1]*m.values[1][3]
+ -m.values[0][3]*m.values[1][1])
+ *m.values[2][2]
+ +(m.values[0][2]*m.values[1][1]
+ -m.values[0][1]*m.values[1][2])
+ *m.values[2][3])
+ *m.values[3][0];
+}
+
+constexpr mat4 inverse(const mat4 &m) noexcept
+{
+ return 1.0f / determinant(m) * mat4((m.values[1][1]*m.values[2][2]
+ -m.values[1][2]*m.values[2][1])
+ *m.values[3][3]
+ +(m.values[1][3]*m.values[2][1]
+ -m.values[1][1]*m.values[2][3])
+ *m.values[3][2]
+ +(m.values[1][2]*m.values[2][3]
+ -m.values[1][3]*m.values[2][2])
+ *m.values[3][1],
+ (m.values[0][2]*m.values[2][1]
+ -m.values[0][1]*m.values[2][2])
+ *m.values[3][3]
+ +(m.values[0][1]*m.values[2][3]
+ -m.values[0][3]*m.values[2][1])
+ *m.values[3][2]
+ +(m.values[0][3]*m.values[2][2]
+ -m.values[0][2]*m.values[2][3])
+ *m.values[3][1],
+ (m.values[0][1]*m.values[1][2]
+ -m.values[0][2]*m.values[1][1])
+ *m.values[3][3]
+ +(m.values[0][3]*m.values[1][1]
+ -m.values[0][1]*m.values[1][3])
+ *m.values[3][2]
+ +(m.values[0][2]*m.values[1][3]
+ -m.values[0][3]*m.values[1][2])
+ *m.values[3][1],
+ (m.values[0][2]*m.values[1][1]
+ -m.values[0][1]*m.values[1][2])
+ *m.values[2][3]
+ +(m.values[0][1]*m.values[1][3]
+ -m.values[0][3]*m.values[1][1])
+ *m.values[2][2]
+ +(m.values[0][3]*m.values[1][2]
+ -m.values[0][2]*m.values[1][3])
+ *m.values[2][1],
+ (m.values[1][2]*m.values[2][0]
+ -m.values[1][0]*m.values[2][2])
+ *m.values[3][3]
+ +(m.values[1][0]*m.values[2][3]
+ -m.values[1][3]*m.values[2][0])
+ *m.values[3][2]
+ +(m.values[1][3]*m.values[2][2]
+ -m.values[1][2]*m.values[2][3])
+ *m.values[3][0],
+ (m.values[0][0]*m.values[2][2]
+ -m.values[0][2]*m.values[2][0])
+ *m.values[3][3]
+ +(m.values[0][3]*m.values[2][0]
+ -m.values[0][0]*m.values[2][3])
+ *m.values[3][2]
+ +(m.values[0][2]*m.values[2][3]
+ -m.values[0][3]*m.values[2][2])
+ *m.values[3][0],
+ (m.values[0][2]*m.values[1][0]
+ -m.values[0][0]*m.values[1][2])
+ *m.values[3][3]
+ +(m.values[0][0]*m.values[1][3]
+ -m.values[0][3]*m.values[1][0])
+ *m.values[3][2]
+ +(m.values[0][3]*m.values[1][2]
+ -m.values[0][2]*m.values[1][3])
+ *m.values[3][0],
+ (m.values[0][0]*m.values[1][2]
+ -m.values[0][2]*m.values[1][0])
+ *m.values[2][3]
+ +(m.values[0][3]*m.values[1][0]
+ -m.values[0][0]*m.values[1][3])
+ *m.values[2][2]
+ +(m.values[0][2]*m.values[1][3]
+ -m.values[0][3]*m.values[1][2])
+ *m.values[2][0],
+ (m.values[1][0]*m.values[2][1]
+ -m.values[1][1]*m.values[2][0])
+ *m.values[3][3]
+ +(m.values[1][3]*m.values[2][0]
+ -m.values[1][0]*m.values[2][3])
+ *m.values[3][1]
+ +(m.values[1][1]*m.values[2][3]
+ -m.values[1][3]*m.values[2][1])
+ *m.values[3][0],
+ (m.values[0][1]*m.values[2][0]
+ -m.values[0][0]*m.values[2][1])
+ *m.values[3][3]
+ +(m.values[0][0]*m.values[2][3]
+ -m.values[0][3]*m.values[2][0])
+ *m.values[3][1]
+ +(m.values[0][3]*m.values[2][1]
+ -m.values[0][1]*m.values[2][3])
+ *m.values[3][0],
+ (m.values[0][0]*m.values[1][1]
+ -m.values[0][1]*m.values[1][0])
+ *m.values[3][3]
+ +(m.values[0][3]*m.values[1][0]
+ -m.values[0][0]*m.values[1][3])
+ *m.values[3][1]
+ +(m.values[0][1]*m.values[1][3]
+ -m.values[0][3]*m.values[1][1])
+ *m.values[3][0],
+ (m.values[0][1]*m.values[1][0]
+ -m.values[0][0]*m.values[1][1])
+ *m.values[2][3]
+ +(m.values[0][0]*m.values[1][3]
+ -m.values[0][3]*m.values[1][0])
+ *m.values[2][1]
+ +(m.values[0][3]*m.values[1][1]
+ -m.values[0][1]*m.values[1][3])
+ *m.values[2][0],
+ (m.values[1][1]*m.values[2][0]
+ -m.values[1][0]*m.values[2][1])
+ *m.values[3][2]
+ +(m.values[1][0]*m.values[2][2]
+ -m.values[1][2]*m.values[2][0])
+ *m.values[3][1]
+ +(m.values[1][2]*m.values[2][1]
+ -m.values[1][1]*m.values[2][2])
+ *m.values[3][0],
+ (m.values[0][0]*m.values[2][1]
+ -m.values[0][1]*m.values[2][0])
+ *m.values[3][2]
+ +(m.values[0][2]*m.values[2][0]
+ -m.values[0][0]*m.values[2][2])
+ *m.values[3][1]
+ +(m.values[0][1]*m.values[2][2]
+ -m.values[0][2]*m.values[2][1])
+ *m.values[3][0],
+ (m.values[0][1]*m.values[1][0]
+ -m.values[0][0]*m.values[1][1])
+ *m.values[3][2]
+ +(m.values[0][0]*m.values[1][2]
+ -m.values[0][2]*m.values[1][0])
+ *m.values[3][1]
+ +(m.values[0][2]*m.values[1][1]
+ -m.values[0][1]*m.values[1][2])
+ *m.values[3][0],
+ (m.values[0][0]*m.values[1][1]
+ -m.values[0][1]*m.values[1][0])
+ *m.values[2][2]
+ +(m.values[0][2]*m.values[1][0]
+ -m.values[0][0]*m.values[1][2])
+ *m.values[2][1]
+ +(m.values[0][1]*m.values[1][2]
+ -m.values[0][2]*m.values[1][1])
+ *m.values[2][0]);
+}
+
+struct pixel
+{
+ std::uint8_t r, g, b, a;
+ constexpr operator vec4() const noexcept
+ {
+ constexpr float scale_factor = 1.0 / std::numeric_limits<std::uint8_t>::max();
+ return vec4(r, g, b, a) * scale_factor;
+ }
+};
+
+struct sampler2D
+{
+ const pixel *pixels;
+ std::size_t width;
+ std::size_t height;
+ vec4 get_pixel_int(int x, int y) const noexcept
+ {
+ if(x < 0)
+ x = 0;
+ else if(static_cast<std::size_t>(x) > width - 1)
+ x = width - 1;
+ if(y < 0)
+ y = 0;
+ else if(static_cast<std::size_t>(y) > height - 1)
+ y = height - 1;
+ return pixels[static_cast<std::size_t>(x) + width * static_cast<std::size_t>(y)];
+ }
+ vec4 get_pixel(vec2 position) const noexcept
+ {
+ // bilinear interpolation
+ int min_x = position.x; // works if position.x >= 0
+ int max_x = min_x + 1;
+ position.x -= min_x;
+ int min_y = position.y; // works if position.y >= 0
+ int max_y = min_y + 1;
+ position.y -= min_y;
+ vec4 min_min_value = get_pixel_int(min_x, min_y);
+ vec4 max_min_value = get_pixel_int(max_x, min_y);
+ vec4 min_max_value = get_pixel_int(min_x, max_y);
+ vec4 max_max_value = get_pixel_int(max_x, max_y);
+ vec4 min_interp_value = min_min_value + position.y * (min_max_value - min_min_value);
+ vec4 max_interp_value = max_min_value + position.y * (max_max_value - max_min_value);
+ return min_interp_value + position.x * (max_interp_value - min_interp_value);
+ }
+};
+
+vec4 texture(const sampler2D &sampler, vec2 uv) noexcept
+{
+ return sampler.get_pixel(uv);
+}
+
+extern vec4 gl_FragCoord;
+
+// shader translated from SuperTuxKart data/shaders/rh.frag
+// https://github.com/supertuxkart/stk-code/blob/20ea7ca2711f0cbe5320b4877a5d332b3b935893/data/shaders/rh.frag
+
+// From http://graphics.cs.aueb.gr/graphics/research_illumination.html
+// "Real-Time Diffuse Global Illumination Using Radiance Hints"
+// paper and shader code
+
float R_wcs = 10.f; // Rmax: maximum sampling distance (in WCS units)
vec3 extents;
mat4 RHMatrix;
mat4 RSMMatrix;
-sampler2D dtex;
-sampler2D ctex;
-sampler2D ntex;
+extern sampler2D dtex;
+extern sampler2D ctex;
+extern sampler2D ntex;
vec3 suncol;
int slice;
float depth = texture(dtex, uv).x;
vec4 RSMPos = inverse(RSMMatrix) * (2.f * vec4(uv, depth, 1.f) - 1.f);
RSMPos /= RSMPos.w;
- vec3 RSMAlbedo = texture(ctex, uv).xyz;
- vec3 normal = normalize(2.f * texture(ntex, uv).xyz - 1.f);
+ vec3 RSMAlbedo = texture(ctex, uv).xyz();
+ vec3 normal = normalize(2.f * texture(ntex, uv).xyz() - 1.f);
// Sampled location inside the RH cell
vec3 offset3d = vec3(uv, 0);
- vec3 SamplePos = RHcenter + .5f * offset3d.xzy * RHCellSize;
+ vec3 SamplePos = RHcenter + .5f * offset3d.xzy() * RHCellSize;
// Normalize distance to RSM sample
- float dist = distance(SamplePos, RSMPos.xyz) / R_wcs;
+ float dist = distance(SamplePos, RSMPos.xyz()) / R_wcs;
// Determine the incident direction.
// Avoid very close samples (and numerical instability problems)
- vec3 RSM_to_RH_dir = (dist <= 0.1f) ? vec3(0.) : normalize(SamplePos - RSMPos.xyz);
- float dotprod = max(dot(RSM_to_RH_dir, normal.xyz), 0.f);
+ vec3 RSM_to_RH_dir = (dist <= 0.1f) ? vec3(0.) : normalize(SamplePos - RSMPos.xyz());
+ float dotprod = max(dot(RSM_to_RH_dir, normal.xyz()), 0.f);
float factor = dotprod / (0.1f + dist * dist);
- vec3 color = RSMAlbedo.rgb * factor * suncol.rgb;
+ vec3 color = RSMAlbedo.rgb() * factor * suncol.rgb();
- SHr += DirToSh(RSM_to_RH_dir, color.r);
- SHg += DirToSh(RSM_to_RH_dir, color.g);
- SHb += DirToSh(RSM_to_RH_dir, color.b);
+ SHr += DirToSh(RSM_to_RH_dir, color.r());
+ SHg += DirToSh(RSM_to_RH_dir, color.g());
+ SHb += DirToSh(RSM_to_RH_dir, color.b());
}
-void shader_main(void) noexcept
+extern "C" void shader_main(void) noexcept
{
- vec3 normalizedRHCenter = 2.f * vec3(gl_FragCoord.xy, slice) / resolution - 1.f;
- vec3 RHcenter = (RHMatrix * vec4(normalizedRHCenter * extents, 1.f)).xyz;
+ vec3 normalizedRHCenter = 2.f * vec3(gl_FragCoord.xy(), slice) / resolution - 1.f;
+ vec3 RHcenter = (RHMatrix * vec4(normalizedRHCenter * extents, 1.f)).xyz();
vec4 ShadowProjectedRH = RSMMatrix * vec4(RHcenter, 1.f);
vec3 RHCellSize = extents / resolution;
- vec2 RHuv = .5f * ShadowProjectedRH.xy / ShadowProjectedRH.w + .5f;
+ vec2 RHuv = .5f * ShadowProjectedRH.xy() / ShadowProjectedRH.w + .5f;
float RHdepth = .5f * ShadowProjectedRH.z / ShadowProjectedRH.w + .5f;
vec4 SHr = vec4(0.f);
SHGreen = SHg;
SHBlue = SHb;
}
+}
projects / kazan.git / commitdiff