fix voxel gi

Shaders/deferred_light/deferred_light.frag.glsl

@@ -283,7 +283,7 @@ void main() {
 #endif
 
 #ifdef _VoxelAOvar
-	envl.rgb *= 1.0 - textureLod(voxels_ao, texCoord, 0.0).r;
+	envl.rgb *= textureLod(voxels_ao, texCoord, 0.0).r;
 #endif
 
 #ifndef _VoxelGI

Shaders/std/conetrace.glsl

@@ -33,7 +33,7 @@ THE SOFTWARE.
 // http://www.seas.upenn.edu/%7Epcozzi/OpenGLInsights/OpenGLInsights-SparseVoxelization.pdf
 // https://research.nvidia.com/sites/default/files/publications/GIVoxels-pg2011-authors.pdf
 
-const float MAX_DISTANCE = voxelgiRange * 100.0;
+const float MAX_DISTANCE = voxelgiRange;
 
 #ifdef _VoxelGI
 uniform sampler3D dummy;
@@ -157,7 +157,7 @@ vec4 traceDiffuse(const vec3 origin, const vec3 normal, const sampler3D voxels,
 		if (cosTheta <= 0)
 			continue;
 		int precomputed_direction = 6 + i;
-		amount += traceCone(voxels, dummy, origin, normal, coneDir, precomputed_direction, false, DIFFUSE_CONE_APERTURE, 1, clipmaps) * cosTheta;
+		amount += traceCone(voxels, dummy, origin, normal, coneDir, precomputed_direction, false, DIFFUSE_CONE_APERTURE, 1.0, clipmaps) * cosTheta;
 		sum += cosTheta;
 	}
 	amount /= sum;
@@ -167,7 +167,7 @@ vec4 traceDiffuse(const vec3 origin, const vec3 normal, const sampler3D voxels,
 }
 
 vec4 traceSpecular(const vec3 origin, const vec3 normal, const sampler3D voxels, const sampler3D voxelsSDF, const vec3 viewDir, const float roughness, const float clipmaps[voxelgiClipmapCount * 10], const vec2 pixel) {
-	vec3 specularDir = reflect(-viewDir, normal);
+	vec3 specularDir = normalize(reflect(-viewDir, normal));
 	vec3 P = origin;// + specularDir * (BayerMatrix8[int(pixel.x) % 8][int(pixel.y) % 8] - 0.5) * voxelgiStep;
 	vec4 amount = traceCone(voxels, voxelsSDF, P, normal, specularDir, 0, true, roughness, voxelgiStep, clipmaps);
 
@@ -186,8 +186,8 @@ vec3 traceRefraction(const vec3 origin, const vec3 normal, sampler3D voxels, sam
 #endif
 
 #ifdef _VoxelAOvar
-float traceConeAO(const sampler3D voxels, const vec3 origin, const vec3 n, const vec3 dir, const int precomputed_direction, const float aperture, const float clipmaps[voxelgiClipmapCount * 10]) {
-	float opacity = 0.0;
+float traceConeAO(const sampler3D voxels, const vec3 origin, const vec3 n, const vec3 dir, const int precomputed_direction, const float aperture, const float step_size, const float clipmaps[voxelgiClipmapCount * 10]) {
+	float sampleCol = 0.0;
 	float voxelSize0 = float(clipmaps[0]) * 2.0;
 	float dist = voxelSize0;
 	float step_dist = dist;
@@ -205,7 +205,7 @@ float traceConeAO(const sampler3D voxels, const vec3 origin, const vec3 n, const
 
 	float coneCoefficient = 2.0 * tan(aperture * 0.5);
 
-    while (opacity < 1.0 && dist < MAX_DISTANCE && clipmap_index0 < voxelgiClipmapCount) {
+    while (sampleCol < 1.0 && dist < MAX_DISTANCE && clipmap_index0 < voxelgiClipmapCount) {
 		float mipSample = 0.0;
 		float diam = max(voxelSize0, dist * coneCoefficient);
         float lod = clamp(log2(diam / voxelSize0), clipmap_index0, voxelgiClipmapCount - 1);
@@ -228,13 +228,12 @@ float traceConeAO(const sampler3D voxels, const vec3 origin, const vec3 n, const
 			mipSample = mix(mipSample, mipSampleNext, clipmap_blend);
 		}
 
-		float a = 1.0 - opacity;
-		opacity += a * mipSample;
+		sampleCol += (1.0 - sampleCol) * mipSample;
 
-		step_dist = diam;
+		step_dist = diam * step_size;
 		dist += step_dist;
 	}
-    return opacity;
+    return sampleCol;
 }
 
 
@@ -247,19 +246,19 @@ float traceAO(const vec3 origin, const vec3 normal, const sampler3D voxels, cons
 		const float cosTheta = dot(normal, coneDir);
 		if (cosTheta <= 0)
 			continue;
-		amount += traceConeAO(voxels, origin, normal, coneDir, precomputed_direction, DIFFUSE_CONE_APERTURE, clipmaps) * cosTheta;
+		amount += traceConeAO(voxels, origin, normal, coneDir, precomputed_direction, DIFFUSE_CONE_APERTURE, voxelgiStep, clipmaps) * cosTheta;
 		sum += cosTheta;
 	}
 	amount /= sum;
-	amount = max(0.0, amount);
+	amount = clamp(amount, 0.0, 1.0);
 	return amount * voxelgiOcc;
 }
 #endif
 
 
 #ifdef _VoxelShadow
 float traceConeShadow(const sampler3D voxels, const sampler3D voxelsSDF, const vec3 origin, const vec3 n, const vec3 dir, const float aperture, const float step_size, const float clipmaps[voxelgiClipmapCount * 10]) {
-    float opacity = 0.0;
+    float sampleCol = 0.0;
 	float voxelSize0 = float(clipmaps[0]) * 2.0;
 	float dist = voxelSize0;
 	float step_dist = dist;
@@ -275,7 +274,7 @@ float traceConeShadow(const sampler3D voxels, const sampler3D voxelsSDF, const v
 	) / (6 + DIFFUSE_CONE_COUNT);
 	vec3 direction_weight = abs(dir);
 
-    while (opacity < 1.0 && dist < MAX_DISTANCE && clipmap_index0 < voxelgiClipmapCount) {
+    while (sampleCol < 1.0 && dist < MAX_DISTANCE * 100 && clipmap_index0 < voxelgiClipmapCount) {
 		float mipSample = 0.0;
 		float diam = max(voxelSize0, dist * 2.0 * tan(aperture * 0.5));
         float lod = clamp(log2(diam / voxelSize0), clipmap_index0, voxelgiClipmapCount - 1);
@@ -306,12 +305,11 @@ float traceConeShadow(const sampler3D voxels, const sampler3D voxelsSDF, const v
 			mipSample = mix(mipSample, mipSampleNext, clipmap_blend);
 		}
 
-		float a = 1.0 - opacity;
-		opacity += a * mipSample;
+		sampleCol += (1.0 - sampleCol) * mipSample;
 
 		float stepSizeCurrent = step_size;
 
-		if (true) {
+		if (false) {
 			// half texel correction is applied to avoid sampling over current clipmap:
 			const vec3 half_texel = vec3(0.5) / voxelgiResolution;
 			vec3 tc0 = clamp(samplePos, half_texel, 1 - half_texel);
@@ -322,14 +320,14 @@ float traceConeShadow(const sampler3D voxels, const sampler3D voxelsSDF, const v
 		step_dist = diam * stepSizeCurrent;
 		dist += step_dist;
 	}
-	return opacity;
+	return sampleCol;
 }
 
 
 float traceShadow(const vec3 origin, const vec3 normal, const sampler3D voxels, const sampler3D voxelsSDF, const vec3 dir, const float clipmaps[voxelgiClipmapCount * 10]) {
 	float amount = traceConeShadow(voxels, voxelsSDF, origin, normal, dir, DIFFUSE_CONE_APERTURE, voxelgiStep, clipmaps);
-	amount = max(0.0, amount);
-	return amount * voxelgiOcc;
+	amount = clamp(amount, 0.0, 1.0);
+	return amount;
 }
 #endif
 #endif // _CONETRACE_GLSL_

Shaders/std/light.glsl

@@ -75,7 +75,7 @@ uniform sampler2D sltcMag;
 	#endif
 	#ifdef _Clusters
 		uniform sampler2DShadow shadowMapSpot[maxLightsCluster];
-		uniform mat4 LWVPSpotArray[maxLightsCluster];
+		uniform mat4 LWVPSpot[maxLightsCluster];
 	#endif
 	#endif
 #endif

Shaders/voxel_light/voxel_light.comp.glsl

@@ -27,14 +27,19 @@ uniform float clipmaps[voxelgiClipmapCount * 10];
 uniform int clipmapLevel;
 
 uniform layout(r32ui) uimage3D voxelsLight;
+uniform layout(r32ui) uimage3D voxels;
 
 #ifdef _ShadowMap
 uniform sampler2DShadow shadowMap;
+uniform sampler2D shadowMapTransparent;
 uniform sampler2DShadow shadowMapSpot;
+uniform sampler2D shadowMapSpotTransparent;
 #ifdef _ShadowMapAtlas
 uniform sampler2DShadow shadowMapPoint;
+uniform sampler2D shadowMapPointTransparent;
 #else
 uniform samplerCubeShadow shadowMapPoint;
+uniform samplerCube shadowMapPointTransparent;
 #endif
 #endif
 
@@ -83,7 +88,6 @@ float lpToDepth(vec3 lp, const vec2 lightProj) {
 
 void main() {
 	int res = voxelgiResolution.x;
-
 	ivec3 dst = ivec3(gl_GlobalInvocationID.xyz);
 
 	vec3 P = (gl_GlobalInvocationID.xyz + 0.5) / voxelgiResolution;
@@ -92,13 +96,11 @@ void main() {
 	P *= voxelgiResolution;
 	P += vec3(clipmaps[int(clipmapLevel * 10 + 4)], clipmaps[int(clipmapLevel * 10 + 5)], clipmaps[int(clipmapLevel * 10 + 6)]);
 
-	vec4 light = vec4(0.0);
-
-	float visibility;
+	vec3 visibility;
 	vec3 lp = lightPos - P;
 	vec3 l;
-	if (lightType == 0) { l = lightDir; visibility = 1.0; }
-	else { l = normalize(lp); visibility = attenuate(distance(P, lightPos)); }
+	if (lightType == 0) { l = lightDir; visibility = vec3(1.0); }
+	else { l = normalize(lp); visibility = vec3(attenuate(distance(P, lightPos))); }
 
 	// float dotNL = max(dot(wnormal, l), 0.0);
 	// if (dotNL == 0.0) return;
@@ -107,7 +109,7 @@ void main() {
 	if (lightShadow == 1) {
 		vec4 lightPosition = LVP * vec4(P, 1.0);
 		vec3 lPos = lightPosition.xyz / lightPosition.w;
-		visibility = texture(shadowMap, vec3(lPos.xy, lPos.z - shadowsBias)).r;
+		visibility = texture(shadowMap, vec3(lPos.xy, lPos.z - shadowsBias)).rrr;
 	}
 	else if (lightShadow == 2) {
 		vec4 lightPosition = LVP * vec4(P, 1.0);
@@ -138,8 +140,45 @@ void main() {
 		}
 	}
 
-	light.rgb += visibility * lightColor;
-	light = clamp(light, vec4(0.0), vec4(1.0));
+	vec3 light = visibility * lightColor;
+
+	for (int i = 0; i < 6; i++) {
+		ivec3 src = ivec3(gl_GlobalInvocationID.xyz);
+		src.x += i * res;
+		vec3 N = vec3(0.0);
+		N.r = float(imageLoad(voxels, src + ivec3(0, 0, voxelgiResolution.x * 7))) / 255;
+		N.g = float(imageLoad(voxels, src + ivec3(0, 0, voxelgiResolution.x * 8))) / 255;
+		N /= 2;
+		vec3 wnormal = decode_oct(N.rg * 2 - 1);
+
+		vec3 coneDirection = wnormal;
+		vec3 aniso_direction = coneDirection;
+		uvec3 face_offsets = uvec3(
+			aniso_direction.x > 0 ? 0 : 1,
+			aniso_direction.y > 0 ? 2 : 3,
+			aniso_direction.z > 0 ? 4 : 5
+		);
+		vec3 direction_weights = abs(coneDirection);
+
+		if (direction_weights.x > 0) {
+			vec3 light_weight = light * direction_weights.x;
+			imageAtomicAdd(voxelsLight, src + ivec3(face_offsets.x, 0, 0), uint(light_weight.r * 255));
+			imageAtomicAdd(voxelsLight, src + ivec3(face_offsets.x, 0, voxelgiResolution.x), uint(light_weight.g * 255));
+			imageAtomicAdd(voxelsLight, src + ivec3(face_offsets.x, 0, voxelgiResolution.x * 2), uint(light_weight.b * 255));
+		}
 
-	imageAtomicMax(voxelsLight, dst, convVec4ToRGBA8(light));
+		if (direction_weights.y > 0) {
+			vec3 light_weight = light * direction_weights.y;
+			imageAtomicAdd(voxelsLight, src + ivec3(face_offsets.y, 0, 0), uint(light_weight.r * 255));
+			imageAtomicAdd(voxelsLight, src + ivec3(face_offsets.y, 0, voxelgiResolution.x), uint(light_weight.g * 255));
+			imageAtomicAdd(voxelsLight, src + ivec3(face_offsets.y, 0, voxelgiResolution.x * 2), uint(light_weight.b * 255));
+		}
+
+		if (direction_weights.z > 0) {
+			vec3 light_weight = light * direction_weights.z;
+			imageAtomicAdd(voxelsLight, src + ivec3(face_offsets.z, 0, 0), uint(light_weight.r * 255));
+			imageAtomicAdd(voxelsLight, src + ivec3(face_offsets.z, 0, voxelgiResolution.x), uint(light_weight.g * 255));
+			imageAtomicAdd(voxelsLight, src + ivec3(face_offsets.z, 0, voxelgiResolution.x * 2), uint(light_weight.b * 255));
+		}
+	}
 }

Shaders/voxel_offsetprev/voxel_offsetprev.comp.glsl

@@ -30,11 +30,11 @@ layout (local_size_x = 8, local_size_y = 8, local_size_z = 8) in;
 #include "std/voxels_constants.glsl"
 
 #ifdef _VoxelGI
-uniform layout(rgba8) image3D voxelsB;
-uniform layout(rgba8) image3D voxelsOut;
+uniform layout(rgba16) image3D voxelsB;
+uniform layout(rgba16) image3D voxelsOut;
 #else
-uniform layout(r8) image3D voxelsB;
-uniform layout(r8) image3D voxelsOut;
+uniform layout(r16) image3D voxelsB;
+uniform layout(r16) image3D voxelsOut;
 #endif
 
 uniform int clipmapLevel;

Shaders/voxel_sdf_jumpflood/voxel_sdf_jumpflood.comp.glsl

@@ -23,8 +23,8 @@ THE SOFTWARE.
 
 #include "compiled.inc"
 
-uniform layout(r8) image3D input_sdf;
-uniform layout(r8) image3D output_sdf;
+uniform layout(r16) image3D input_sdf;
+uniform layout(r16) image3D output_sdf;
 
 uniform float jump_size;
 uniform int clipmapLevel;