What’s new in Version 3.9
Here is a summary of some of the new features and feature improvements in version 3.9 of mental ray. Please refer to the release notes for more details and for other changes which are not mentioned here.
Introduce a new sampling mechanism that unifies the different sampling controls of the ray tracing, rasterizer, and progressive rendering engines. It offers direct control of sampling quality coupled with perceptual adaptivity to detail in the image. It is disabled by default, and gets enabled and controlled with string options. The existing sampling controls are kept for backwards compatibility and work as in previous versions of mental ray when the feature is not turned on.
The ray tracing acceleration algorithm BSP2 in mental ray has been revised and dramatically improved both in terms of memory usage and speed when rendering scenes with motion blur. Especially, scenes with lots of instanced geometry render ray tracing motion blur using just a small fraction of the memory compared to previous versions. In many cases with extreme instance motion this new approach is able to outperform the rasterizer.
In addition, the quality of motion blur has been enhanced by shading with time-accurate normals and shadow tolerances. This may lead to visible differences in the rendered images caused by, now exactly computed, light reflections on animated objects, or similar. The effect of this improvement becomes most obvious in rotating objects.
Hair rendering, especially with ray tracing, is noticeably accelerated and memory consumption is reduced. The rarely used per-vertex normal feature is deprecated to improve performance. A new option allows to control if hair caps should be created for ray tracing.
The most common stereo mode offaxis support a parallax option for convenient control of the zero parallax plane distance. Additional configuration registry variables allow to customize the file names of the stereo images.
The updated version 1.2 of the iray rendering engine is built into mental ray. It provides numerous improvements in performance and quality, as well as added support for existing common shaders. ling support for new CUDA architecture Fermi
The updated version of MetaSL 1.2 is supported by mental ray, providing new features like material definitions and techniques. The LLVM back-end performance has been improved noticeably.
Native Particle Rendering
mental ray can now render particles. Currently, particles with certain pre-defined shapes are supported that are stored in native 3d Map files. With the help of shaders external point cloud and particle formats can be used with mental ray.
The finalgather map has been switched to utilize the 3d Map primitive and file format for internal caching and disk storage. This offers benefits like better handling of large maps, and publication and exchange of finalgather map data for display and use with other software.
New helpful tools are delivered with mental ray. The imf_comp utility is a small and simple compositing tool that allows to combine images rendered with mental ray easily. The dsm_copy utility can be used to convert detail shadow map files generated with mental ray to regular image files for display and investigation.
New Shader Package
A new shader package userdata is now part of mental ray. It supports special workflows with large scene, where dynamic attributes attached to scene elements can indirectly drive parameters of the assigned shader(s).
Scene Description Language
The following changes were made in the .mi scene description syntax:
[li] The unified sampling mode can be controlled with new string options: “unified sampling” on|off “samples min” float “samples max” float “samples quality” color “samples error cutoff” color Note, that min/max samples specifies the actual number of samples to take, not a sample level or exponent as in traditional sampling. Otherwise, these parameters have similar meaning of determining the lower and upper limits of overall samples. Adaptive sampling and oversampling rate is controlled with the quality and error parameters. A higher quality value will lead to better quality. The “soft” valid range for the quality color components is [0,1], however, larger values are accepted. The default value is 0.5. The error cutoff defines a threshold where further improvements are considered negligible. The default value is zero (black). Setting it to higher values will lead to less samples taken in dark areas of the image.[/li]> [li] The hair endings may be flat or rounded caps. A global option allows to control the creation of caps for all hair objects in a scene in case of ray tracing. “hair caps” on|off[/li]> [/ul]Shader Writing and Integration
[li] New query modes have been introduced to return per-vertex data of geometric primitives for triangles and hair. miQ_HAIR_NUM_TEXTURES miQ_HAIR_NUM_USERS miQ_HAIR_USERS miQ_PRI_NUM_VERTICES miQ_PRI_VERTICES_POINTS miQ_PRI_VERTICES_NORMALS miQ_PRI_VERTICES_MOTION miQ_PRI_VERTICES_TEX miQ_PRI_VERTICES_USER miQ_PRI_VERTICES_BUMPS_U miQ_PRI_VERTICES_BUMPS_V miQ_PRI_VERTICES_DERIVS_U miQ_PRI_VERTICES_DERIVS_V miQ_PRI_VERTICES_DERIVS_UU miQ_PRI_VERTICES_DERIVS_VV miQ_PRI_VERTICES_DERIVS_UV miQ_PRI_VERTICES_DERIVS_VU miQ_PRI_VERTICES_DERIVS_UV They offer more reliable ways to retrieve such data in shaders, in comparison to the deprecated function mi_tri_vectors. At the same time, queries of hair per-vertex user data are now supported. See mi_query for more details.[/li]> [/ul]