Releases v0.3.0  
 
  
  
        
  
  
                    
  
      
    Compare 
      
          
     
       
   
 
    
            
        
          
            
                
  
                  
                    
                    
                        
                          
     
                          Sorry, something went wrong. 
                         
                     
 
  
              
            
              
No results found 
            
          
       
    
 
   
 
  
        Added 
Light tree for hierarchical NEE and direct intersections 
PMJ sampling 
Ray type visibility masks (diffuse, glossy etc.) 
Gauss and Blackman-Harris image filters 
Multithreaded interface for scene Finalize 
Texture compression for CPU backends 
Direct loading of BC-compressed images 
TSAN tests on CI 
Path-space regularization 
Option to flip normalmap Y channel (DirectX convention) 
Ability to create single-sided mesh lights 
Physical sky sample 
Build version string 
 
Fixed 
Indirect clamp not affecting unshadowed lights 
Clamping not preserving color hue 
Glossy material desaturation an glazing angles 
Flipped front/back materials on GPU 
Not implemented RemoveMeshInstance, RemoveLight, RemoveMesh 
Incorrect scaled mesh lights intensity 
Incorrect mix node resolution in SIMD mode 
Incorrect NEE for textured triangle lights 
Incorrect MIS at the last bounce 
Incorrect far clip plane 
Incorrect transmission below 1.0 
Incorrect pitch black values with filmic tonemap 
Incorrect clamping of direct light intersections 
Triangular artifacts on AMD 
Crash on out of memory fallback to CPU RAM 
Fireflies with some HDRI images 
Flipped tangent basis on flipped UV islands 
Crash on hitting transparency limit (CPU only) 
Broken DOF RNG (GPU only) 
Incorrect alpha of 0.8 in samples 
Meshlights triangles memory leak 
Adreno issues with new driver 
Non-reproduceable GPU test runs 
 
Changed 
Improved physical sky (multiple scattering, clouds, moon, stars) 
Improved area light sampling 
Improved adaptive sampling 
BC4/BC5 compression uses SSE acceleration 
Texture compression uses NEON acceleration 
AUX buffers output is enabled permanently 
Bundled GPU shaders compressed using deflate 
More flexible interface for specifying vertex attributes 
Sphere light is allowed to have zero radius 
Bounded sampling is used for VNDF 
Mix node texture is allowed to be SRGB 
HWRT BLAS build happens per mesh instead of all at once 
15-seconds time limit is used for README images 
 
Removed 
LinearAlloc 
Tent image filter 
  
  
  
 
   
 
     
   
   
          
    
    
     
    
      
     
     
    You can’t perform that action at this time.