Couple of small optimisation

Add option to build for gprof
Do not build by default with the renderstat (they have a reasonable impact on performances)
Separated created ray and castedray in the stats
Trying to force some simple function to be inlined

CMakeLists.txt

@@ -12,11 +12,17 @@ set(CMAKE_MODULE_PATH ${CMAKE_MODULE_PATH} ${PROJECT_SOURCE_DIR}/external/covera
 option(PACKAGE_TESTS "Build the tests" ON)
 option(ENABLE_COVERAGE "Build for code coverage" OFF)
 
-option(SHOW_STATS "Show rendering stat" ON)
+option(SHOW_STATS "Show rendering stat" OFF)
 if (SHOW_STATS)
     add_compile_options(-DRENDER_STATS)
 endif()
 
+option(USE_GPROF "Enable profiling" OFF)
+if (USE_GPROF)
+    add_compile_options(-pg)
+    add_link_options(-pg)
+endif()
+
 option(USE_LUA "Enable the use of Lua" ON)
 if (USE_LUA)
     add_compile_options(-DENABLE_LUA_SUPPORT)

source/camera.cpp

@@ -54,6 +54,7 @@ Ray Camera::rayForPixel(uint32_t pixelX, uint32_t pixelY)
     Tuple origin = this->inverseTransform * Point(0, 0, 0);
     Tuple direction = (pixel - origin).normalise();
 
+    stats.addCastedRay();
     return Ray(origin, direction);
 }
 

source/include/renderstat.h

@@ -27,7 +27,8 @@ private:
     uint64_t csgCount;                   /* Total number of CSG             */
 
     uint64_t pixelCount;                 /* Total number of rendered pixels */
-    uint64_t rayCount;                   /* Total number of rays */
+    uint64_t rayCount;                   /* Total number of rays object created */
+    uint64_t rayCasted;                  /* Total number of rays actually casted */
     uint64_t lightRayEmitedCount;        /* Total number of ray launched for light tests */
     uint64_t reflectionRayCount;         /* Total number of reflection ray launched */
     uint64_t refractedRayCount;          /* Total number of refracted ray launched */
@@ -43,7 +44,8 @@ public:
     RenderStats() : coneCount(0), cylinderCount(0), cubeCount(0), groupCount(0), lightCount(0), planeCount(0), sphereCount(0), triangleCount(0),
                     pixelCount(0), rayCount(0), lightRayEmitedCount(0), reflectionRayCount(0), refractedRayCount(0),
                     intersectCount(0), intersectionCount(0), reallocCallCount(0), mallocCallCount(0), smoothTriangleCount(0),
-                    discardedIntersectCount(0), maxDepthAttained(UINT64_MAX), maxIntersectOnARay(0), objfileCount(0), csgCount(0) {};
+                    discardedIntersectCount(0), maxDepthAttained(UINT64_MAX), maxIntersectOnARay(0), objfileCount(0),
+                    csgCount(0), rayCasted(0) {};
 #ifdef RENDER_STATS
     void addCone();
     void addCylinder();
@@ -59,6 +61,7 @@ public:
     void printStats();
     void addPixel();
     void addRay();
+    void addCastedRay();
     void addLightRay();
     void addReflectRay();
     void addRefractRay();
@@ -82,6 +85,7 @@ public:
     static void printStats() {};
     static void addPixel() {};
     static void addRay() {};
+    static void addCastedRay() {};
     static void addLightRay() {};
     static void addReflectRay() {};
     static void addRefractRay() {};

source/include/shape.h

@@ -65,7 +65,7 @@ public:
 
     ShapeType getType() { return this->type; };
 
-    virtual void intersect(Ray &r, Intersect &xs);
+    virtual void intersect(Ray &r, Intersect &xs) { this->localIntersect(this->invTransform(r), xs); };
     Tuple normalAt(Tuple point, Intersection *hit = nullptr);
 
     uint64_t getObjectId() { return this->objectId; };

source/renderstat.cpp

@@ -85,20 +85,32 @@ void RenderStats::addRay()
     this->rayCount++;
 };
 
+void RenderStats::addCastedRay()
+{
+#pragma omp atomic
+    this->rayCasted++;
+};
+
 void RenderStats::addLightRay()
 {
+    this->addCastedRay();
+
 #pragma omp atomic
     this->lightRayEmitedCount++;
 };
 
 void RenderStats::addReflectRay()
 {
+    this->addCastedRay();
+
 #pragma omp atomic
     this->reflectionRayCount++;
 };
 
 void RenderStats::addRefractRay()
 {
+    this->addCastedRay();
+
 #pragma omp atomic
     this->refractedRayCount++;
 };
@@ -172,7 +184,8 @@ void RenderStats::printStats()
     printf("CSG                     : %lld\n", this->csgCount);
     printf("==================================================\n");
     printf("Pixel rendered          : %lld\n", this->pixelCount);
-    printf("Ray casted              : %lld\n", this->rayCount);
+    printf("Ray created             : %lld\n", this->rayCount);
+    printf("Ray casted              : %lld\n", this->rayCasted);
     printf("Light Ray casted        : %lld\n", this->lightRayEmitedCount);
     printf("Reflection ray casted   : %lld\n", this->reflectionRayCount);
     printf("Refraction ray casted   : %lld\n", this->refractedRayCount);

source/shapes/shape.cpp

@@ -36,12 +36,6 @@ uint64_t Shape::newObjectId()
     return ret;
 }
 
-
-void Shape::intersect(Ray &r, Intersect &xs)
-{
-    this->localIntersect(this->invTransform(r), xs);
-};
-
 Tuple Shape::normalToWorld(Tuple normalVector)
 {
     Tuple world_normal = this->transposedInverseTransform * normalVector;

tests/christmasball_render.cpp

@@ -229,13 +229,14 @@ int main()
     w.addObject(s);
 
     /* ----------------------------- */
-
+    /*
     FILE *fpOut = fopen("christmas_worlddump.json", "wt");
     if (fpOut)
     {
         w.dumpMe(fpOut);
         fclose(fpOut);
     }
+    */
     /* ----------------------------- */
 
 

tests/dragon_scene.cpp

@@ -117,14 +117,7 @@ int main()
     w.addObject(dragon);
 
     /* ----------------------------- */
-/*
+
-    FILE *fpOut = fopen("dragon_worlddump.json", "wt");
-    if (fpOut)
-    {
-        w.dumpMe(fpOut);
-        fclose(fpOut);
-    }
-*/
     OctreeOptimisation opt;
     w.finalise(opt);