Welcome to RPS 20.12!

Welcome to RenderMan 20.12 This version builds on the initial release of Renderman 20. The notes below include numerous bug fixes and performance enhancements.

Changes in 20.12_1684284

New Features in 20.12_1684284

• Watertight dicing is now available in RenderMan 20.12 using RIS rendering. The watertight tessellation primarily aims to eliminate a major source of pinhole artifacts in objects by eliminating T-junctions that arise due to the use of regular rectangular dicing on adjacent faces.

Attribute "dice" "watertight"

Changes in 20.11_1645142

Changes in 20.10_1623687

New Features in 20.10_1623687

• Exposed an attribute that allows the user to assign a relative pixel variance to an object or collection of objects. Artists can use this to improve sampling where it's needed instead of relying on global controls. Please note that large differences may cause artifacts around objects and small changes are recommended.

Attribute "shade" "relativepixelvariance" [float]

• Added a new Display Driver for Deep Output for OpenEXRs. This supports arbitrary sets of channels, Zback for volumes, and both tiled and scanline formats.

"int asrgba" [0|1]
 Override first four channel names with RGBA. Default 0.
"string compression" ["none"|"rle"|"zips"]
 Type of compression to use.  Default "zips".
"string type" ["half"|"float"]
 Whether to store data as half or float. Z and ZBack always float. Default "float".
"string storage" ["tile"|"scanline"]
 Whether to write a random-order tiled EXR or an increasing-order scanline EXR. The former should require less memory to
 write, the later is directly usable in Nuke. Default "tile".
"int[2] tilesize"
 Size of the tiles if using tiled storage. Default [16 16].

Changes in 20.9_1602511

New Features in 20.9_1602511

• There is now a way to de-prioritize low luminance samples in the adaptive sampler. Before it was possible the adaptive sampler was spending too much time sampling dark areas of an image where it was less important. The user can now add a small boost to the pixels for sampling consideration using a hider control. Using higher values may speed up render times. We suggest beginning with small amounts, e.g. 0.05. The default maintains the old behavior at 0.0:

  Hider "raytrace" "float darkfalloff" [0.0]
  

Default Render Result

Falloff Off - Samples per Pixel

Falloff On -- Samples per Pixel, note the reduction.

• New AOV "float rawID" that passes through integer id values to the display driver unchanged. For example, using: Attribute "identifier" "int id" [4294699949] on an object will result in the display driver receiving a value that is -nan if the bits are interpreted as a float, but the original 4294699949 is interpreted as an unsigned int.

Changes in 20.8_1587495

New Features in 20.8_1587495

• PxrPathTracer now supports three new attributes. These new attributes allow for a group subset to be used when tracing shadow, reflection, and refraction rays, respectively.:

  Attribute "trace" "string shadowsubset" ["<group-membership>"]
  Attribute "trace" "string reflectsubset" ["<group-membership>"]
  Attribute "trace" "string transmitsubset" ["<group-membership>"]
  

• Implemented new per-path-depth timer for PxrPathTracer integrator. The new feature reports the time spent at each ray depth for the user to tune the performance impact of maxPathDepth. The time is reported in wallclock time.

• Logging is enabled from the command line (-loglevel/-logfile) The logger can print to stdout/stderr (default) or be redirected to file.:

  -logfile s        : override render log file
  -loglevel i       : override render log level, 0..4
  

• A timer is added to account for the time spent preparing/intersecting/batching primary rays. The two stats are combined together in the stats java scripts to plot a nice histogram in the "integration" tab of the artist friendly stats.

• It is now possible to maintain a checkpointed "final" image using either an .ini setting or a RIB option, with the later overriding the former. If neither is present it defaults to off. When set, this prevents stripping out the extra channels and the checkpoint tag when writing the file image for the render. The final image will essentially be just another checkpoint, rather than a slimmed down image:

  /prman/checkpoint/asfinal [bool]
  Option "checkpoint" "int asfinal" [0|1]
  

Changes in 20.7_1571626

New Features in 20.7_1571626

• Heterogeneous volumes (using a PxrVolume BXDF bound to an RiVolume) can be multiple times faster to render, depending on the density of the volume. In addition, sparse and overlapping volumes are less noisy to render.

  PxrVolume now defaults to 4 samples and adds the ability to use a primvar for density directly without having to wire a PxrPrimVar input. The maxDensity parameter can also be set higher without greatly impacting render times, as the renderer will also automatically compute density bounds.

  As part of the optimizations for heterogeneous volumes, RiVolumes in RIS that use ImplicitField DSOs will no longer perform filtered evaluations, as doing so will result in a biased (and incorrect) render. You can find more information in the documentation for PxrVolume.

  Please note that ShadingRate on RiVolumes should be increased to improve render times and decrease memory usage on scenes that may be affected. In some cases you can use high values, i.e. 25 or more. When rendering with RIS this will have little visual impact if carefully used. Experimentation may be required if memory usage has changed.

  

• Added new diagnostic built-in AOVs: "float cpuTime" and "float sampleCount". These AOVs provide time spent and samples per pixel respectively.

• Added new special filter option, "sum", for Display and DisplayChannel "string filter" parameter.

• Min, max, zmin, and zmax filters are now implemented for deep image output. Id channels default to zmin and are consequently now filtered in a reasonable (non-blended) way.

Changes in 20.6_1562369

New Features in 20.6_1562369

• PxrPathTracer can now track the diffusedepth and speculardepth separately from each other based on which lobe of the BSDF is sampled. Previously rays that have both diffuse and specular contributions increment both the diffuse and specular depths. The new behavior is more intuitive and produce expected results when adjusting the trace:maxdiffusedepth and trace:maxspeculardepth attributes. This comes at the cost of, in some cases, effectively doubling the number of indirect ray bounces which can cause look differences and/or increase render times. We still default to the old behavior in order to avoid image differences and no action is needed by the user to preserve their previous results. Use Option "trace" "string depthmode" ["separate"] to enable the new behavior, or Option "trace" "string depthmode" ["combined"] for the old behavior. Alternatively, the new or old behavior can be enabled/disabled via a rendermn.ini setting: /prman/trace/depthmode separate|combined

Left, the old behavior (Combined). Right, optional new behavior (Separate). Notice that the right image is brighter where the ceiling is indirectly lit.

Changes in 20.5_1555015

New Features in 20.5_1555015

• PxrVolume can now take multiple samples over single scatter volume rendering by setting the new samples parameter. Taking multiple samples in the volume is a faster variance reduction technique than raising the number of pixel samples. Typical speedups can be up to 4X compared to before. [Note multi-scatter volumes do not benefit from this setting.]

Original method with 64 pixel samples on the left. Increasing volume samples to 64 on the right with 64 pixel samples.

Changes in 20.4_1545505

Changes in 20.3_1534387

New Features in 20.3_1534387

• New feature that optionally disables merging of shading geometry in RIS if object IDs differ. Use Option "shading" "int mergedifferingobjectids" [0|1] to toggle this new option. By default mergedifferingobjectids = 1
• Lights, gobos, barndoors and blockers now support multi segment motion blur.

Changes in 20.2_1521997

New Features in 20.1

Simplified Standard AOVs

Production integrators can now output geometric AOVs (__Pworld, __Nworld, etc.) Previously this required the use of PxrGeometricAOVs pattern. The older workflow is deprecated in favor of simple global control.

Changes in 20.1_1509300

Welcome to RPS 20!

Welcome to RenderMan 20. This release introduces many new improvements to the previous release. There are many volume improvements, new holdout workflows, a production quality hair Bxdf, a new physicaly based camera projection plugin, a new visualizer integrator, and much more. In addition there are changes to APIs and new APIs for the developers. Please dive into the release notes below for more detailed information.

New Features

Denoise

PRMan now includes a render denoising utility. It has the ability to emit the special set of AOVs that this tool requires to produce denoised images.

Volume Improvements

• RIS now supports the rendering of overlapping volumes. There are no limitations on the physical properties of these volumes - volumes of arbitrary density, albedo, phase function, etc. may be overlapped with physically accurate results.
• Support for equiangular importance sampling has been added to volumes. This can significantly improve convergence for scenes with lights inside volumes with low and/or slowly varying density.
• Volumes which use complicated field functions (such as OpenVDB) are now up to twice as fast to render.
• Issues pertaining to camera-visible volumes (including the blocking of environment lights and requirements that such volumes also required indirect visibility) have been fixed.
• RiVolumes may now be coincident with each other. A subsequent patch release will further improve the robustness of RiVolumes coincident with other geometry as well.
• Cameras can now render from within multiple RiVolumes.

HoldOut

A new holdout LPE prefix is available to collect shadows, reflections, and other channels from the CG objects on a holdout object. This allows compositing CG shadows, reflections and other channels onto a background image or an image plane interactively at render time.

PxrMarschnerHair

A new hair Bxdf PxrMarschnerHair is available. This Bxdf was developed for production at Pixar Animation Studios.

User Defined Signal LPE

Additional LPE tokens (U1 to U8) are now available for Bxdfs to output user defined signal LPEs such as albedo. In addition, LPEs now support intersection with the '&' symbol.

New Improved AreaLight Source Shaders

New RIS-only AreaLight Source shaders: PxrStdAreaLight, PxrStdEnvDayLight, and PxrStdEnvMapLight.

New Light Filter API

There is a new C++ API which allows developers to create more complex, custom light filter shaders.

Improved Light localization

Improves noise with multiple light sources. The default for RIS, mode 0, has been modified to get improved results in situations that combine environment/distant and local lights. A new mode, mode 4 (RIS-only), has been added that dynamically adjusts the relative importance of lights as the scene is rendered, accounting for Bxdf response to illumination and occlusion. In addition, mode 3 has been optimised so that it is faster in the case of very many light sources. See lights documentation for details.

New Subsurface scattering features in RIS

The subsurface scattering Bxdfs can now request sss continuation rays and whether the scattering amount should take surface topology into account: "continuationRays" (int) and "followTopology" (float) parameters. The scattering distance in simple/PxrSubsurface Bxdf can now (optionally) be specified in an alternative way: "meanFreePath" (color) instead of "diffuseMeanFreePath" (color).

Visualizer Integrator

PxrVisualizer is a new integrator that can be used to navigate large scenes and inspect geometry during Interactive re-rendering. It allows different styles of viewing, including shaded, flat, normals, st, wireframe.

PxrCamera Plugin

PxrCamera is a camera plugin that approximates a number of real world physical effects. It supports all of the traditional prman perspective camera settings, including shaped motion blur and bokeh.

Important Differences

Additional Changes