three-gpu-pathtracer 0.0.1 → 0.0.4

package/src/shader/shaderEnvMapSampling.js

@@ -0,0 +1,67 @@
+export const shaderEnvMapSampling = /* glsl */`
+
+vec3 sampleEquirectEnvMapColor( vec3 direction, sampler2D map ) {
+
+	return texture2D( map, equirectDirectionToUv( direction ) ).rgb;
+
+}
+
+float envMapDirectionPdf( vec3 direction ) {
+
+	vec2 uv = equirectDirectionToUv( direction );
+	float theta = uv.y * PI;
+	float sinTheta = sin( theta );
+	if ( sinTheta == 0.0 ) {
+
+		return 0.0;
+
+	}
+
+	return 1.0 / ( 2.0 * PI * PI * sinTheta );
+
+}
+
+float envMapSample( vec3 direction, EquirectHdrInfo info, out vec3 color ) {
+
+	vec2 uv = equirectDirectionToUv( direction );
+	color = texture2D( info.map, uv ).rgb;
+
+	float totalSum = texture2D( info.totalSum, vec2( 0.0 ) ).r;
+	float lum = colorToLuminance( color );
+	ivec2 resolution = textureSize( info.map, 0 );
+	float pdf = lum / totalSum;
+
+	return float( resolution.x * resolution.y ) * pdf * envMapDirectionPdf( direction );
+
+}
+
+float randomEnvMapSample( EquirectHdrInfo info, out vec3 color, out vec3 direction ) {
+
+	// sample env map cdf
+	vec2 r = rand2();
+	float v = texture2D( info.marginalWeights, vec2( r.x, 0.0 ) ).x;
+	float u = texture2D( info.conditionalWeights, vec2( r.y, v ) ).x;
+	vec2 uv = vec2( u, v );
+
+	vec3 derivedDirection = equirectUvToDirection( uv );
+	direction = derivedDirection;
+	color = texture2D( info.map, uv ).rgb;
+
+	float totalSum = texture2D( info.totalSum, vec2( 0.0 ) ).r;
+	float lum = colorToLuminance( color );
+	ivec2 resolution = textureSize( info.map, 0 );
+	float pdf = lum / totalSum;
+
+	return float( resolution.x * resolution.y ) * pdf * envMapDirectionPdf( direction );
+
+}
+
+float misHeuristic( float a, float b ) {
+
+	float aa = a * a;
+	float bb = a * b;
+	return aa / ( bb + aa );
+
+}
+
+`;

package/src/shader/shaderGGXFunctions.js

@@ -1,107 +1,108 @@
-export const shaderGGXFunctions = /* glsl */`
-// The GGX functions provide sampling and distribution information for normals as output so
-// in order to get probability of scatter direction the half vector must be computed and provided.
-// [0] https://www.cs.cornell.edu/~srm/publications/EGSR07-btdf.pdf
-// [1] https://hal.archives-ouvertes.fr/hal-01509746/document
-// [2] http://jcgt.org/published/0007/04/01/
-// [4] http://jcgt.org/published/0003/02/03/
-
-// trowbridge-reitz === GGX === GTR
-
-vec3 ggxDirection( vec3 incidentDir, float roughnessX, float roughnessY, float random1, float random2 ) {
-
-	// TODO: try GGXVNDF implementation from reference [2], here. Needs to update ggxDistribution
-	// function below, as well
-
-	// Implementation from reference [1]
-	// stretch view
-	vec3 V = normalize( vec3( roughnessX * incidentDir.x, roughnessY * incidentDir.y, incidentDir.z ) );
-
-	// orthonormal basis
-	vec3 T1 = ( V.z < 0.9999 ) ? normalize( cross( V, vec3( 0.0, 0.0, 1.0 ) ) ) : vec3( 1.0, 0.0, 0.0 );
-	vec3 T2 = cross( T1, V );
-
-	// sample point with polar coordinates (r, phi)
-	float a = 1.0 / ( 1.0 + V.z );
-	float r = sqrt( random1 );
-	float phi = ( random2 < a ) ? random2 / a * PI : PI + ( random2 - a ) / ( 1.0 - a ) * PI;
-	float P1 = r * cos( phi );
-	float P2 = r * sin( phi ) * ( ( random2 < a ) ? 1.0 : V.z );
-
-	// compute normal
-	vec3 N = P1 * T1 + P2 * T2 + V * sqrt( max( 0.0, 1.0 - P1 * P1 - P2 * P2 ) );
-
-	// unstretch
-	N = normalize( vec3( roughnessX * N.x, roughnessY * N.y, max( 0.0, N.z ) ) );
-
-	return N;
-
-}
-
-// Below are PDF and related functions for use in a Monte Carlo path tracer
-// as specified in Appendix B of the following paper
-// See equation (2) from reference [2]
-float ggxLamda( float theta, float roughness ) {
-
-	float tanTheta = tan( theta );
-	float tanTheta2 = tanTheta * tanTheta;
-	float alpha2 = roughness * roughness;
-
-	float numerator = - 1.0 + sqrt( 1.0 + alpha2 * tanTheta2 );
-	return numerator / 2.0;
-
-}
-
-// See equation (2) from reference [2]
-float ggxShadowMaskG1( float theta, float roughness ) {
-
-	return 1.0 / ( 1.0 + ggxLamda( theta, roughness ) );
-
-}
-
-// See equation (125) from reference [4]
-float ggxShadowMaskG2( vec3 wi, vec3 wo, float roughness ) {
-
-	float incidentTheta = acos( wi.z );
-	float scatterTheta = acos( wo.z );
-	return 1.0 / ( 1.0 + ggxLamda( incidentTheta, roughness ) + ggxLamda( scatterTheta, roughness ) );
-
-}
-
-float ggxDistribution( vec3 halfVector, float roughness ) {
-
-	// See equation (33) from reference [0]
-	float a2 = roughness * roughness;
-	float cosTheta = halfVector.z;
-	float cosTheta4 = pow( cosTheta, 4.0 );
-
-	if ( cosTheta == 0.0 ) return 0.0;
-
-	float theta = acos( halfVector.z );
-	float tanTheta = tan( theta );
-	float tanTheta2 = pow( tanTheta, 2.0 );
-
-	float denom = PI * cosTheta4 * pow( a2 + tanTheta2, 2.0 );
-	return a2 / denom;
-
-	// See equation (1) from reference [2]
-	// const { x, y, z } = halfVector;
-	// const a2 = roughness * roughness;
-	// const mult = x * x / a2 + y * y / a2 + z * z;
-	// const mult2 = mult * mult;
-
-	// return 1.0 / Math.PI * a2 * mult2;
-
-}
-
-// See equation (3) from reference [2]
-float ggxPDF( vec3 wi, vec3 halfVector, float roughness ) {
-
-	float incidentTheta = acos( wi.z );
-	float D = ggxDistribution( halfVector, roughness );
-	float G1 = ggxShadowMaskG1( incidentTheta, roughness );
-
-	return D * G1 * max( 0.0, dot( wi, halfVector ) ) / wi.z;
-
-}
-`;
+export const shaderGGXFunctions = /* glsl */`
+// The GGX functions provide sampling and distribution information for normals as output so
+// in order to get probability of scatter direction the half vector must be computed and provided.
+// [0] https://www.cs.cornell.edu/~srm/publications/EGSR07-btdf.pdf
+// [1] https://hal.archives-ouvertes.fr/hal-01509746/document
+// [2] http://jcgt.org/published/0007/04/01/
+// [4] http://jcgt.org/published/0003/02/03/
+
+// trowbridge-reitz === GGX === GTR
+
+vec3 ggxDirection( vec3 incidentDir, float roughnessX, float roughnessY, float random1, float random2 ) {
+
+	// TODO: try GGXVNDF implementation from reference [2], here. Needs to update ggxDistribution
+	// function below, as well
+
+	// Implementation from reference [1]
+	// stretch view
+	vec3 V = normalize( vec3( roughnessX * incidentDir.x, roughnessY * incidentDir.y, incidentDir.z ) );
+
+	// orthonormal basis
+	vec3 T1 = ( V.z < 0.9999 ) ? normalize( cross( V, vec3( 0.0, 0.0, 1.0 ) ) ) : vec3( 1.0, 0.0, 0.0 );
+	vec3 T2 = cross( T1, V );
+
+	// sample point with polar coordinates (r, phi)
+	float a = 1.0 / ( 1.0 + V.z );
+	float r = sqrt( random1 );
+	float phi = ( random2 < a ) ? random2 / a * PI : PI + ( random2 - a ) / ( 1.0 - a ) * PI;
+	float P1 = r * cos( phi );
+	float P2 = r * sin( phi ) * ( ( random2 < a ) ? 1.0 : V.z );
+
+	// compute normal
+	vec3 N = P1 * T1 + P2 * T2 + V * sqrt( max( 0.0, 1.0 - P1 * P1 - P2 * P2 ) );
+
+	// unstretch
+	N = normalize( vec3( roughnessX * N.x, roughnessY * N.y, max( 0.0, N.z ) ) );
+
+	return N;
+
+}
+
+// Below are PDF and related functions for use in a Monte Carlo path tracer
+// as specified in Appendix B of the following paper
+// See equation (2) from reference [2]
+float ggxLamda( float theta, float roughness ) {
+
+	float tanTheta = tan( theta );
+	float tanTheta2 = tanTheta * tanTheta;
+	float alpha2 = roughness * roughness;
+
+	float numerator = - 1.0 + sqrt( 1.0 + alpha2 * tanTheta2 );
+	return numerator / 2.0;
+
+}
+
+// See equation (2) from reference [2]
+float ggxShadowMaskG1( float theta, float roughness ) {
+
+	return 1.0 / ( 1.0 + ggxLamda( theta, roughness ) );
+
+}
+
+// See equation (125) from reference [4]
+float ggxShadowMaskG2( vec3 wi, vec3 wo, float roughness ) {
+
+	float incidentTheta = acos( wi.z );
+	float scatterTheta = acos( wo.z );
+	return 1.0 / ( 1.0 + ggxLamda( incidentTheta, roughness ) + ggxLamda( scatterTheta, roughness ) );
+
+}
+
+float ggxDistribution( vec3 halfVector, float roughness ) {
+
+	// See equation (33) from reference [0]
+	float a2 = roughness * roughness;
+	a2 = max( EPSILON, a2 );
+	float cosTheta = halfVector.z;
+	float cosTheta4 = pow( cosTheta, 4.0 );
+
+	if ( cosTheta == 0.0 ) return 0.0;
+
+	float theta = acosSafe( halfVector.z );
+	float tanTheta = tan( theta );
+	float tanTheta2 = pow( tanTheta, 2.0 );
+
+	float denom = PI * cosTheta4 * pow( a2 + tanTheta2, 2.0 );
+	return ( a2 / denom );
+
+	// See equation (1) from reference [2]
+	// const { x, y, z } = halfVector;
+	// const a2 = roughness * roughness;
+	// const mult = x * x / a2 + y * y / a2 + z * z;
+	// const mult2 = mult * mult;
+
+	// return 1.0 / Math.PI * a2 * mult2;
+
+}
+
+// See equation (3) from reference [2]
+float ggxPDF( vec3 wi, vec3 halfVector, float roughness ) {
+
+	float incidentTheta = acos( wi.z );
+	float D = ggxDistribution( halfVector, roughness );
+	float G1 = ggxShadowMaskG1( incidentTheta, roughness );
+
+	return D * G1 * max( 0.0, dot( wi, halfVector ) ) / wi.z;
+
+}
+`;