This is RenderMan 21 Documentation you are viewing. The Rmanwiki home page will redirect you to the current documentation.

Page tree

Versions Compared


  • This line was added.
  • This line was removed.
  • Formatting was changed.


  • Volume rendering improvements can increase speeds on scenes with sparse volumes using single scatter up to 4x and improve scenes with multiscatter volumes up to 25%.
  • DeepEXR's storage changed from "tiled" to "scanline". Note that this may dramatically increase memory requirements to cache the frames as they are being rendered. We recommend Horizontal or Zigzag-X bucket order when rendering to help ameliorate the cost. Nuke requires scanline storage for deepEXR files.
  • PxrSurface's Specular and Clearcoat energy compensation now applies to the physical fresnel mode.
  • Volumes in deep images rendered using the PxrPathTracer integrator have been improved.  In particular, single-scatter volumes now work with deep images and developers of custom integrators can now register splats at distinct multiple depths along a single ray for deep images using the new RixDisplayServices::AddSample() method. Note that this may require developers to recompile their plugins, but no source changes are needed."

  • The RixDisplayChannel::ChannelType::k_IdChannel enum has been renamed to k_IntegerChannel.  The old name remains available as an alias but will be removed in the next major release.
  • The "float rawId" builtin AOV is now deprecated. Use the equivalent "int id" AOV instead.

  • Made optimizations to the core raytracing speed.
  •  An incremental render that successfully uses recover mode to pick up from a previous render now does at least 'minextrasamples' new samples (Hider "raytrace" "int minextrasamples").

  • Attribute "identifier" "id" of -1 (old behavior was < 0) signals the master ID wins, not the instance (i.e. no instance override desired).

  • Speed improvements to txmake when dealing with large input images have been made. Large textures should use the OpenEXR format option.
  • Reduced noise and improved convergence in the PxrOcclusion integrator.

  • Texture cache memory statistics are now more accurate, particularly the texture entry in the overall memory summary.
  • Custom integrator plugins are now responsible for splatting builtins; PxrGeoAovs.h has been added to help with this.

  • Improved performance of OSL access to primvars and inputs from C++ shaders.