[cleanup] Final fixes for the release

Author: sherholz-intel

CHANGELOG.md

@@ -6,8 +6,8 @@
 
   - Radiance Caching (RC):
     - If RC is enabled, the guiding structure (i.e., `Field`) learns an
-      approximation of multiple radiance quantities (in linear RGB)
-      ,such as outgoing and incoming radiance, irradiance, fluence, and
+      approximation of multiple radiance quantities (in linear RGB),
+      such as outgoing and incoming radiance, irradiance, fluence, and
       in-scattered radiance. These quantities can be queried using the
       `SurfaceSamplingDistribution` and `VolumeSamplingDistribution`
       classes. RC support can be enabled using the

README.md

@@ -56,8 +56,8 @@ The full version history can be found [here](./CHANGELOG.md)
 
   - Radiance Caching (RC):
     - If RC is enabled, the guiding structure (i.e., `Field`) learns an
-      approximation of multiple radiance quantities (in linear RGB)
-      ,such as outgoing and incoming radiance, irradiance, fluence, and
+      approximation of multiple radiance quantities (in linear RGB),
+      such as outgoing and incoming radiance, irradiance, fluence, and
       in-scattered radiance. These quantities can be queried using the
       `SurfaceSamplingDistribution` and `VolumeSamplingDistribution`
       classes. RC support can be enabled using the

doc/changelog_latest.md

@@ -2,7 +2,7 @@
 
 - New (**Experimental**) Features:
     - Radiance Caching (RC):
-        - If RC is enabled, the guiding structure (i.e., `Field`) learns an approximation of multiple radiance quantities (in linear RGB) ,such as outgoing and incoming radiance, irradiance, fluence, and in-scattered radiance. These quantities can be queried using the `SurfaceSamplingDistribution` and `VolumeSamplingDistribution` classes.
+        - If RC is enabled, the guiding structure (i.e., `Field`) learns an approximation of multiple radiance quantities (in linear RGB), such as outgoing and incoming radiance, irradiance, fluence, and in-scattered radiance. These quantities can be queried using the `SurfaceSamplingDistribution` and `VolumeSamplingDistribution` classes.
         RC support can be enabled using the `OPENPGL_EF_RADIANCE_CACHES` CMake option. **Note:** Since the RC quantities are Monte-Carlo estimates, zero-value samples (`ZeroValueSampleData`) that are generated during rendering/training have to be passed/stored in the `SampleStorage` as well. 
     - Guided/Adjoint-driven Russian Roulette (GRR):
         - The information stored in radiance caches can be used to optimize stochastic path termination decisions (a.k.a. Russian roulette) to avoid a significant increase in variance (i.e., noise) caused by early terminations, which can occur when using standard throughput-based RR strategies.

openpgl/data/PathSegmentDataStorage.h

@@ -261,7 +261,9 @@ struct PathSegmentDataStorage
                 }
 
                 bool directLightSample = false;
+#ifdef OPENPGL_RADIANCE_CACHES
                 float misWeight = 1.f;
+#endif
                 // evalaute the incident radiance the incident
                 openpgl::Vector3 throughput{1.0f};
                 openpgl::Vector3 contribution{0.0f};
@@ -310,7 +312,9 @@ struct PathSegmentDataStorage
                                 else
                                 {
                                     contribution += clampedThroughput * nextPathSegment.miWeight * directContribution;
+#ifdef OPENPGL_RADIANCE_CACHES
                                     misWeight = nextPathSegment.miWeight;
+#endif
                                     OPENPGL_ASSERT(embree::isvalid(contribution));
                                     OPENPGL_ASSERT(contribution[0] >= 0.f && contribution[1] >= 0.f && contribution[2] >= 0.f);
                                 }

openpgl/include/openpgl/common.h

@@ -267,8 +267,9 @@ struct pgl_spectrum
     uint32_t spectrum;
 #ifdef __cplusplus
     pgl_spectrum() {}
-    pgl_spectrum(const float r, const float g, const float b) {
-        spectrum = vec3f2rgbe({r,g,b});
+    pgl_spectrum(const float r, const float g, const float b)
+    {
+        spectrum = vec3f2rgbe({r, g, b});
     }
     pgl_spectrum(const pgl_vec3f rgb)
     {