Merge pull request #18 from tippesi/feature/rt-fx

Feature/rt fx

README.md

@@ -8,11 +8,9 @@
 [code-quality-image]: https://api.codiga.io/project/10000/score/svg
 [code-quality-url]: https://app.codiga.io/public/project/10000/Atlas-Engine/dashboard
 
-![GI scene](wiki/images/gi.gif) <br/>
-*Realtime GI demo scene* <br/>
-#### The current version is a WIP which means many changes and less stability.
+![GI scene](wiki/images/intel_sponza.gif) <br/>*Realtime Sponza scene with software raytraced GI, AO and reflections (model from [Intel Graphics Research Sample Library](https://www.intel.com/content/www/us/en/developer/topic-technology/graphics-research/samples.html))* <br/>
 ## Introduction
-This is a cross platform toy engine that is available on Linux and Windows.
+This is a cross platform toy engine developed in my spare time that is available on Linux and Windows.
 ## Requirements
 - OpenGL 4.6
 - C++17 compatible compiler
@@ -72,5 +70,3 @@ For a code example have a look at the [demo application](https://github.com/tipp
 *Path traced scene*
 ![Island scene](wiki/images/island.gif) <br/>
 *Island demo scene using the terrain and ocean systems* <br/>
-![Gun scene](wiki/images/gun.gif) <br/>
-*PBR rendering* <br/>

data/shader/ao/rtao.csh

@@ -6,6 +6,7 @@
 #include <../common/random.hsh>
 #include <../common/flatten.hsh>
 #include <../common/convert.hsh>
+#include <../brdf/brdfSample.hsh>
 
 layout (local_size_x = 8, local_size_y = 4) in;
 
@@ -14,92 +15,91 @@ layout (binding = 3, r16f) writeonly uniform image2D rtaoImage;
 layout(binding = 0) uniform sampler2D normalTexture;
 layout(binding = 1) uniform sampler2D shadowMap;
 layout(binding = 2) uniform sampler2D randomTexture;
+layout(binding = 3) uniform isampler2D offsetTexture;
 
-uniform uint sampleCount;
+const ivec2 offsets[4] = ivec2[4](
+    ivec2(0, 0),
+    ivec2(1, 0),
+    ivec2(0, 1),
+    ivec2(1, 1)
+);
+
+uniform int sampleCount;
 uniform float radius;
 
 uniform mat4 pMatrix;
 uniform mat4 ivMatrix;
 
 uniform ivec2 resolution;
+uniform uint frameSeed;
 
-/*
-Spherical Fibonacci Mapping
-http://lgdv.cs.fau.de/publications/publication/Pub.2015.tech.IMMD.IMMD9.spheri/
-Authors: Benjamin Keinert, Matthias Innmann, Michael Sänger, Marc Stamminger
-*/
-float madfrac(float a, float b) { return a * b - floor(a * b); }
-
-vec3 fibonacciSphere(float i, float n) {
-	const float PHI = 1.6180339887498948482045868343656;
-	float phi = 2.0 * PI * madfrac(i, PHI);
-	float cosTheta = 1.0 - (2.0 * i + 1.0) * (1.0 / n);
-	float sinTheta = sqrt(saturate(1.0 - sqr(cosTheta)));
-	return vec3(cos(phi) * sinTheta, sin(phi) * sinTheta, cosTheta);
-}
-
-ivec2 FindNearest2x2(ivec2 pixel, out float depth) {
-    float depths[4];
-    depths[0] = texelFetch(shadowMap, pixel, 0).r;
-    depths[1] = texelFetch(shadowMap, pixel + ivec2(1, 0), 0).r;
-    depths[2] = texelFetch(shadowMap, pixel + ivec2(0, 1), 0).r;
-    depths[3] = texelFetch(shadowMap, pixel + ivec2(1, 1), 0).r;
-
-    ivec2 depthOffset = ivec2(0, 0);
-    depth = depths[0];
-    for (uint i = 1; i < 4; i++) {
-        if (depths[i] < depth) {
-            switch(i) {
-                case 1: depthOffset = ivec2(1, 0); break;
-                case 2: depthOffset = ivec2(0, 1); break;
-                case 3: depthOffset = ivec2(1, 1); break;
-            }
-            depth = depths[i];
-        }
-    }
+void main() {
 
-    return depthOffset;
-}
+    int verticalGroupCount = resolution.y / int(gl_WorkGroupSize.y);
+    verticalGroupCount += ((verticalGroupCount * int(gl_WorkGroupSize.y) == resolution.y) ? 0 : 1);
 
-void main() {
+    int tileSize = 4;
+    int groupsPerTile = tileSize * verticalGroupCount;
+
+    int tileGroupIdx = int(gl_WorkGroupID.x) % groupsPerTile;
+    int tileIdx = int(gl_WorkGroupID.x) / groupsPerTile;
+
+    ivec2 tileBaseOffset = ivec2(tileIdx * tileSize, 0);
+    ivec2 tileGroupOffset = ivec2(tileGroupIdx % tileSize, tileGroupIdx / tileSize);
 
-	if (int(gl_GlobalInvocationID.x) < resolution.x &&
-		int(gl_GlobalInvocationID.y) < resolution.y) {
+    ivec2 groupOffset = (tileBaseOffset + tileGroupOffset) * ivec2(gl_WorkGroupSize);
+    ivec2 pixel = ivec2(gl_LocalInvocationID.xy) + groupOffset;
+
+	if (int(pixel.x) < resolution.x &&
+		int(pixel.y) < resolution.y) {
 
-		ivec2 pixel = ivec2(gl_GlobalInvocationID.xy);
 		vec2 texCoord = (vec2(pixel) + vec2(0.5)) / vec2(resolution);
 
-        float depth;
-        ivec2 offset  = FindNearest2x2(pixel * 2, depth);
+        float depth = texelFetch(shadowMap, pixel, 0).r;
 
-	    vec3 worldPos = vec3(ivMatrix * vec4(ConvertDepthToViewSpace(depth, texCoord), 1.0));
-        vec3 worldNorm = normalize(vec3(ivMatrix * vec4(2.0 * texelFetch(normalTexture, pixel * 2 + offset, 0).rgb - 1.0, 0.0)));
+        int offsetIdx = texelFetch(offsetTexture, pixel, 0).r;
+        ivec2 offset = offsets[offsetIdx];
+
+        vec2 recontructTexCoord = (2.0 * vec2(pixel) + offset + vec2(0.5)) / (2.0 * vec2(resolution));
+	    vec3 worldPos = vec3(ivMatrix * vec4(ConvertDepthToViewSpace(depth, recontructTexCoord), 1.0));
+        vec3 worldNorm = normalize(vec3(ivMatrix * vec4(2.0 * textureLod(normalTexture, texCoord, 0).rgb - 1.0, 0.0)));
+        float seed = texelFetch(randomTexture, pixel % ivec2(4), 0).r;
 
         float ao = 0.0;
-
-        vec3 up = abs(worldNorm.z) < 0.999 ? vec3(0.0, 0.0, 1.0) : vec3(1.0, 0.0, 0.0);
-        vec3 tangent = normalize(cross(up, worldNorm));
-        vec3 bitangent = cross(worldNorm, tangent);
 
-        mat3 TBN = mat3(tangent, bitangent, worldNorm);
+        float raySeed = float(seed);
+        float curSeed = float(0);
+
+        ivec2 noiseOffset = Unflatten2D(int(frameSeed), ivec2(16)) * ivec2(8);
+        vec2 blueNoiseVec = texelFetch(randomTexture, ((pixel % ivec2(8)) + noiseOffset) % ivec2(128), 0).xy * 256.0;
+		blueNoiseVec = clamp(blueNoiseVec, 0.0, 255.0);
+		blueNoiseVec = (blueNoiseVec + 0.5) / 256.0;
 
-        const int sampleCount = 32;
+        const int sampleCount = 1;
         for (uint i = 0; i < sampleCount; i++) {
-            float fibIdx = round(float(sampleCount) * random(vec3(texCoord, float(i))));
             Ray ray;
-            ray.origin = worldPos + worldNorm * EPSILON;
-            ray.direction = normalize(worldNorm + normalize(TBN * fibonacciSphere(fibIdx, float(sampleCount))));
+            Surface surface;
+
+            float u0 = random(raySeed, curSeed);
+    	    float u1 = random(raySeed, curSeed);
+
+            surface.N = worldNorm;
+            surface.P = worldPos;
+            BRDFSample brdfSample = SampleDiffuseBRDF(surface, blueNoiseVec);
+
+            ray.direction = brdfSample.L;
             ray.inverseDirection = 1.0 / ray.direction;
+            ray.origin = worldPos + ray.direction * EPSILON + worldNorm * EPSILON;
 
             ray.hitID = -1;
             ray.hitDistance = 0.0;
 
-            HitClosest(ray, 0.0, radius);
+            bool hit = HitAny(ray, 0.0, radius);
 
-            ao += ray.hitID > 0 ? saturate(radius - sqr(ray.hitDistance)) : 0.0;
+            ao += hit ? 1.0 : 0.0;
         }
 
-        float result = pow(1.0 - (ao / float(sampleCount)), 2.0);
+        float result = 1.0 - (ao / float(sampleCount));
 
         imageStore(rtaoImage, pixel, vec4(result, 0.0, 0.0, 0.0));
 	}

data/shader/ao/ssao.csh

@@ -14,7 +14,7 @@ layout(binding = 3) uniform sampler2D randomTexture;
 uniform vec3 samples[64];
 uniform vec2 resolution;
 
-uniform uint sampleCount;
+uniform int sampleCount;
 uniform float radius;
 uniform float strength;
 uniform int frameCount;
@@ -43,7 +43,7 @@ void main() {
 
     //Create TBN matrix
     vec3 tang = normalize(randomVec - norm * dot(randomVec, norm));
-    vec3 bitang = cross(norm, tang);
+    vec3 bitang = normalize(cross(norm, tang));
     mat3 TBN = mat3(tang, bitang, norm);
 
     //Calculate occlusion factor
@@ -70,7 +70,7 @@ void main() {
 
     }
 
-    float result = pow(1.0 - (occlusion / float(sampleCount)), strength);
+    float result = pow(1.0 - (occlusion / float(sampleCount)), 2.0);
     result = depth == 1.0 ? 0.0 : result;
     imageStore(textureOut, pixel, vec4(result, 0.0, 0.0, 0.0));