Still working on atomics

Author: AEspinosaDev

examples/resources/scenes/sponza.xml

@@ -58,7 +58,7 @@
                     <albedo r="0.5" g="0.5" b="0.5" />
                     <Textures>
                         <albedo path="textures/Columns/Column_C_Diffuse.png" />
-                        <normals path="textures/Columns/Column_C_Normal.png" />
+                        <!-- <normals path="textures/Columns/Column_C_Normal.png" /> -->
                         <!-- <mask path="textures/Columns/Column_C_MaskMap.png" type="UnityHDRP" /> -->
                     </Textures>
                 </Material>
@@ -281,7 +281,7 @@
 
     <Enviroment type="skybox">
         <Filename value="textures/cloudy.hdr" />
-        <intensity value="0.01" />
+        <intensity value="0.4" />
     </Enviroment>
 
 

include/engine/common.h

@@ -286,6 +286,8 @@ typedef enum ColorFormatTypeFlagBits
     SRGB_8    = VK_FORMAT_R8G8B8_SRGB,   // RGB
     SRGBA_8   = VK_FORMAT_R8G8B8A8_SRGB, // RGB with Alpha
     SBGRA_8   = VK_FORMAT_B8G8R8A8_SRGB, // Other order
+    SR_16F    = VK_FORMAT_R16_SFLOAT,
+    SR_32F    = VK_FORMAT_R32_SFLOAT,
     SRG_16F   = VK_FORMAT_R16G16_SFLOAT,
     SRG_32F   = VK_FORMAT_R32G32_SFLOAT,
     SRGB_16F  = VK_FORMAT_R16G16B16_SFLOAT,
@@ -296,6 +298,7 @@ typedef enum ColorFormatTypeFlagBits
     RG_8U     = VK_FORMAT_R8G8_UNORM,
     RGB_8U    = VK_FORMAT_R8G8B8_UNORM,
     RGBA_8U   = VK_FORMAT_R8G8B8A8_UNORM,
+    R_32_UINT = VK_FORMAT_R32_UINT,
     DEPTH_16F = VK_FORMAT_D16_UNORM,
     DEPTH_32F = VK_FORMAT_D32_SFLOAT
 } ColorFormatType;
@@ -403,8 +406,8 @@ typedef enum TextureFormatTypeFlagBits
 {
     TEXTURE_FORMAT_SRGB  = 0x00000000,
     TEXTURE_FORMAT_UNORM = 0x00000001,
-    TEXTURE_FORMAT_DEPTH  = 0x00000002,
-    TEXTURE_FORMAT_HDR    = 0x00000003
+    TEXTURE_FORMAT_DEPTH = 0x00000002,
+    TEXTURE_FORMAT_HDR   = 0x00000003
 } TextureFormatType;
 typedef enum BindingTypeFlagBits
 {

include/engine/core/passes/composition_pass.h

@@ -44,6 +44,7 @@ struct VXGI { // Settings for Voxel Based GI
     uint32_t resolution        = 256;
     uint32_t samples           = 8;
     uint32_t enabled           = 1;
+    uint32_t updateMode        = 0;
 };
 
 class CompositionPass : public GraphicPass

include/engine/core/passes/voxelization_pass.h

@@ -11,13 +11,19 @@
 #include <engine/core/passes/pass.h>
 #include <engine/core/resource_manager.h>
 
+#define USE_IMG_ATOMIC_OPERATION
+
 VULKAN_ENGINE_NAMESPACE_BEGIN
 
 namespace Core {
 
 class VoxelizationPass : public GraphicPass
 {
+#ifdef USE_IMG_ATOMIC_OPERATION
     const uint16_t RESOURCE_IMAGES = 4;
+#else
+    const uint16_t RESOURCE_IMAGES = 1;
+#endif
 
     /*Descriptors*/
     struct FrameDescriptors {
@@ -32,7 +38,7 @@ class VoxelizationPass : public GraphicPass
   public:
     VoxelizationPass(Graphics::Device* ctx, uint32_t resolution)
         : BasePass(ctx, {resolution, resolution}, 1, 1, false, "VOXELIZATION") {
-            m_resourceImages.resize(RESOURCE_IMAGES);
+        m_resourceImages.resize(RESOURCE_IMAGES);
     }
 
     void setup_attachments(std::vector<Graphics::AttachmentInfo>&    attachments,

resources/shaders/VXGI/merge_intermediates.glsl

@@ -0,0 +1,24 @@
+#shader compute
+#version 460 core
+
+// In this case the voxelizer atomically writes to 3 seperate r32ui textures
+// instead of only one rgba16f texture. Here we merge them together into one.
+
+layout(local_size_x = 4, local_size_y = 4, local_size_z = 4) in;
+
+layout(set = 0,  binding =  6, r32f)    uniform image3D finalVoxelImage;
+layout(set = 0,  binding =  7, r32ui)   uniform uimage3D interVoxelImages[3];
+
+void main()
+{
+    ivec3 imgCoord = ivec3(gl_GlobalInvocationID);
+
+    float a = imageLoad(finalVoxelImage, imgCoord).a;
+    if (a > 0.0)
+    {
+        float r = imageLoad(interVoxelImages[0], imgCoord).r;
+        float g = imageLoad(interVoxelImages[1], imgCoord).r;
+        float b = imageLoad(interVoxelImages[2], imgCoord).r;
+        imageStore(finalVoxelImage, imgCoord, vec4(r, g, b, a));
+    }
+}

resources/shaders/VXGI/voxelization.glsl

@@ -43,6 +43,8 @@ void main() {
 }
 #shader geometry
 #version 460
+#extension GL_NV_geometry_shader_passthrough : enable
+
 
 layout(triangles) in;
 layout(triangle_strip, max_vertices = 3) out;
@@ -80,12 +82,15 @@ void main(){
 
 #shader fragment
 #version 460
+
+#define USE_IMG_ATOMIC_OPERATION
+
 #extension GL_EXT_ray_tracing : enable
 #extension GL_EXT_ray_query : enable
-#extension GL_EXT_shader_atomic_float2 : enable
-// #extension GL_EXT_shader_16bit_storage : require
+
+
+#extension GL_EXT_shader_atomic_float2 : require
 // #extension GL_EXT_gpu_shader5 : require
-// #extension GL_EXT_shader_explicit_arithmetic_types_float16 : require
 // #extension GL_EXT_shader_atomic_fp16_vector : require
 #include light.glsl
 #include scene.glsl
@@ -95,6 +100,9 @@ void main(){
 #include raytracing.glsl
 #include material_defines.glsl
 
+
+
+
 layout(location = 0) in vec3 _pos;
 layout(location = 1) in vec3 _normal;
 layout(location = 2) in vec2 _uv;
@@ -103,7 +111,11 @@ layout(set = 0, binding =   2) uniform sampler2DArray              shadowMap;
 layout(set = 0, binding =   3) uniform samplerCube                 irradianceMap;
 layout(set = 0,  binding =  4) uniform accelerationStructureEXT    TLAS;
 layout(set = 0,  binding =  5) uniform sampler2D                   samplerMap;
+
 layout(set = 0,  binding =  6, r32f) uniform image3D               voxelImage;
+#ifdef USE_IMG_ATOMIC_OPERATION
+layout(set = 0,  binding =  7, r32ui) uniform uimage3D              auxVoxelImages[3];
+#endif
 
 layout(set = 1, binding = 1) uniform MaterialUniforms {
     vec4 slot1; 
@@ -217,9 +229,16 @@ void main() {
 
 	vec4 result = g_opacity * vec4(vec3(color), 1);
     ivec3 voxelPos = worldSpaceToVoxelSpace(_pos);
-    
-    // imageAtomicMax(voxelImage, voxelPos, 1);
-    //  imageAtomicMax(voxelImage, voxelPos, f16vec4(color, 1.0));
+
+#ifdef USE_IMG_ATOMIC_OPERATION
+    imageAtomicMax(auxVoxelImages[0], voxelPos, floatBitsToUint(result.r));
+    imageAtomicMax(auxVoxelImages[1], voxelPos, floatBitsToUint(result.g));
+    imageAtomicMax(auxVoxelImages[2], voxelPos, floatBitsToUint(result.b));
+    // imageAtomicMax(voxelImage, voxelPos, f16vec4(vec4(result, 1.0)));
+    // imageStore(voxelImage, voxelPos, vec4(0.0, 0.0, 0.0, 1.0));
+    imageStore(voxelImage, voxelPos, result);
+#else
     imageStore(voxelImage, voxelPos, result);
+#endif
 
 }

resources/shaders/deferred/composition.glsl

@@ -124,8 +124,8 @@ void main()
 
         vec3 direct     = vec3(0.0);
         vec3 ambient    = vec3(0.0);
-        vec3 indirect   = vec3(0.0);
-        vec3 indirect2   = vec3(0.0);
+        vec4 indirect   = vec4(0.0);
+        vec3 indirectSpecular   = vec3(0.0);
 
         vec3 modelPos = (camera.invView * vec4(g_pos.xyz, 1.0)).xyz;
         vec3 modelNormal = (camera.invView  * vec4(g_normal.xyz, 0.0)).xyz;
@@ -150,16 +150,20 @@ void main()
             // Indirect Component ____________________________
             if(settings.vxgi.enabled == 1){
 
-                vec4 diffuseIndirect = diffuseVoxelGI(voxelMap, modelPos, modelNormal, settings.vxgi, scene.maxCoord.x-scene.minCoord.x);
+                // Diffuse
+                indirect = diffuseVoxelGI2(voxelMap, modelPos, modelNormal, settings.vxgi, scene.maxCoord.x-scene.minCoord.x);
+                indirect.rgb *= g_albedo;
+                indirect.rgb *= settings.enableAO == 1 ? (brdf.ao * SSAO) : brdf.ao; //Add ambient occlusion
 
-                indirect = diffuseIndirect.rgb * g_albedo;
-                indirect*= settings.enableAO == 1 ? (brdf.ao * SSAO) : brdf.ao; //Add ambient occlusion
+                // Specular
+                vec3 specularConeDirection = reflect(-normalize(camera.position.xyz-modelPos), modelNormal);
+                float voxelWorldSize =  (scene.maxCoord.x-scene.minCoord.x) / float(settings.vxgi.resolution);
+	            vec3 startPos = modelPos + modelNormal * voxelWorldSize * settings.vxgi.offset;
 
-                // vec3 specularConeDirection = reflect(normalize(camera.position.xyz-modelPos), modelNormal);
-                // vec3 specularIndirect = vec3(0.0);
-                //  specularIndirect =   traceVoxelCone(voxelMap, startPos, specularConeDirection, max(brdf.roughness, 0.05) , VOXEL_SIZE );
-                // // specularIndirect = castCone(startPos, specularConeDirection, max(roughness, MIN_SPECULAR_APERTURE), MAX_TRACE_DISTANCE, minLevel).rgb * specColor.rgb * u_indirectSpecularIntensity;
-                // indirect2+=specularIndirect;
+                const float CONE_SPREAD = mix(0.005, settings.vxgi.diffuseConeSpread, brdf.roughness);
+                indirectSpecular = traceCone(voxelMap, startPos, specularConeDirection, settings.vxgi.maxDistance, CONE_SPREAD, voxelWorldSize ).rgb;
+                
+                // indirect+=specularIndirect.rgb;
             }
             //Direct Component ________________________
             for(int i = 0; i < scene.numLights; i++) {
@@ -262,7 +266,8 @@ void main()
 
 
 
-    color = direct + indirect + ambient + reflectedColor;
+    color = direct  + indirect.rgb + ambient + reflectedColor;
+    // color = direct + indirect.rgb + ambient + reflectedColor;
 
     //////////////////////////////////////
     // IF UNLIT MATERIAL

resources/shaders/include/VXGI.glsl

@@ -18,6 +18,7 @@ struct VXGI {
 	uint 	resolution;
 	uint 	samples;
 	uint 	enabled;
+	uint 	updateMode;
 };
 
 vec4 traceCone(sampler3D voxelization, vec3 origin, vec3 direction, const float MAX_DISTANCE ,const float CONE_SPREAD, const float VOXEL_WORLD_SIZE)