6.0.4.0

Enhancements

• Automatic file path remapping: String parameter values can now be automatically modified according to the OS Arnold is running on. For example, this is useful when the artist workstations are running Windows but the renderfarm nodes are running Linux. Configuration is done with a JSON file that contains the strings to search and replace for each OS. This configuration file can be selected through the ARNOLD_PATHMAP environment variable.  (#9311)  For example:
  

  {
      "windows":
      {
         "this_path/": "some/other_path/windows/"
      },
      
      "mac":
      {
        "this_path/": "some/other_path/mac/"
      },
      
      "linux":
      {
        "this_path/": "some/other_path/linux/"
      }
  }

• Rim light control for the Toon shader: A new rim_light_tint float parameter was added to the toon shader to tint the color of the rim light using the base color of the surface. See the full documentation. (#9003)

  0 (default)	0.5	1

• Performance improvements on Windows with many cores and texture-heavy scenes: Several performance improvements have been made, particularly for those running Windows, using many cores, tiff based textures, and/or reading many textures/sec. Expected speedups can be anywhere from a few percent to orders of magnitude faster on many-core Windows machines. (#9515, #9685, #9686, #9695, #9709, #9761) For a medium complexity scene that uses exr based .tx textures we get the following speedups on Windows :

  CPU	Arnold 6.0.3.0	Arnold 6.0.4.0	Speedup
  AMD 3970X (64 threads)	439s	397s	1.11x
  AMD 3990X (128 threads)	379s	267s	1.42x
  AMD 3995WX (128 threads)	351s	276s	1.27x
  Intel Xeon 8280 x 2 (112 threads)	356s	337s	1.06x
  Intel i9-9980XE (36 threads)	748s	720s	1.04x

  For a texture heavy stress scene (8GB of tiff based .tx texture tiles end up being read in) we get the following speedups on Windows:

  CPU	Arnold 6.0.3.0	Arnold 6.0.4.0	Speedup
  AMD 3970X (64 threads)	2002s	9.9s	202x
  AMD 3990X (128 threads)	2103s	10.1s	208x
  AMD 3995WX (128 threads)	2130s	9.0s	237x
  Intel Xeon 8280 x 2 (112 threads)	1585s	11.2s	142x
  Intel i9-9980XE (36 threads)	1145s	16.8s	68x

• String parameter path metadata: A new path metadata can be assigned to string parameters to give specific hints about how they will be handled. Currently available values are: file, folder and folderlist. Arnold will expand environment variables, apply the remapping, and handle backslashes on string parameters with the metadata. (#4969)
• AA_adaptive_threshold in log: The options.AA_adaptive_threshold used is now output in the log file. (#9841)
• Upgraded OSL: Arnold now uses OSL 1.11.6 bringing multiple optimizations and bug fixes, see the full list of changes. (#8917)

GPU enhancements

• Global AOV shaders: The aov_shader option is now supported on GPU. This option defines a list of shaders that the renderer will evaluate after the regular surface shader. With this, it's possible to add shaders to set specific AOVs without modifying the original shader tree. See the option documentation. (#9394)
• Initial support for attributes in OSL:  You can now retrieve user data bound to shapes via the getattribute function or via a lockgeom parameter attribute. This initial implementation still has some limitations when compared to the CPU implementation. Queryable attributes are currently restricted to those bound to the current shaded shape and attributes of the active render camera. Querying node parameters is also not yet supported. (#9666)
• Initial support for tracing in OSL:  You can now use the trace function to trace rays in OSL shaders along with the getmessage function to retrieve information regarding any intersected object. This initial implementation has the restriction that you cannot yet query parameters on an intersected object. (#9797)

USD enhancements

• UsdRender schema: UsdRenderSetting, UsdRenderProduct and UsdRenderVar are now supported in the procedural. (usd#453)

API additions

• Improved log callbacks : It is now possible to register multiple log callbacks using AiMsgRegisterCallback. The registered log callbacks are associated with a specific mask, and they use the new AtMsgExtendedCallBack function pointer. See the API documentation for the full set of functions. The following functions will be deprecated and eventually removed AiMsgSetCallback, AiMsgAddCallback, and AiMsgResetCallback. (#9598)

• AiResolveFilePath: This new API resolves a given file path, expanding environment variables and applying path remapping, before searching in the corresponding default search path. A custom search path or paths can optionally be given, that will be searched before the default ones. (#3546)
• AtMutex and AtRecursiveMutex: AtCritSec and associated functions have been deprecated and replaced with AtMutex and AtRecursiveMutex which can be used similarly to std::mutex and std::recursive_mutex and on Windows will perform better than AtCritSec. Using AtMutex/AtRecursiveMutex instead of directly using std::mutex/recursive_mutex will enable Arnold to record time spent blocked, which is used during profiling and stats output. Visual Studio 2015 update 2 or older produces a compile-time warning with std::mutex and so we've disabled these with those older compilers. Upgrade to at least VS 2015 Update 3 in order to use this. (#6862)

Incompatible changes

• AiMsg* changes: Changes have been made to how logging behaves. Now the log session is for the life of the Arnold process instead of just between AiBegin/AiEnd pairs. If messages are not showing up or too many messages show up, you might need to set the logging flags and other options before calling AiBegin. Another change is that you can now set the log filename at any point in time, each time AiMsgSetLogFileName is called the previous file will be closed and the new file will be created. See the API documentation for the full set of functions. (#9575)
• Increased default texture maximum memory: The options.texture_max_memory_MB default value has been raised from 2048MB to 4096MB. (#9846)

Bug Fixes