Upgrading from mental ray 3.4 to 3.5

Upgrading from mental ray 3.4 to 3.5

What's new in Version 3.5
Scene Description Language
Shader Writing and Integration
Incompatible Changes

What's new in Version 3.5

Here is a summary of some of the new features and feature improvements in version 3.5 of mental ray:

New final gathering modes

mental ray now has four final gathering modes: "automatic", "multiframe", "3.4", and "strict 3.4". The automatic and multiframe modes have been greatly simplified and do not require control by radii, while usually giving superious image quality. The multiframe mode is specifically designed to avoid flickering artifacts in camera animations. The 3.4 and strict 3.4 modes are provided for backwards compatibility.

Light lists in instances

Light lists have been added to geometric instances to offer a general, more flexible ways to specify lights for illumination, and for finer control over shadow casting. The same material may now use different lights on different geometries. Light lists provide a new, standard method to pass lights to shaders for illumination; they require, however, support in the shaders by using the new light iterators.

Catmull-Clark meshes

Catmull-Clark polygon meshes ("ccmeshes") have been introduced. This subdivision surface type uses Catmull-Clark rules for smoothening and supports polygons with arbitrary numbers of vertices.

Reloading of .mi files for reduced memory consumption

A command line option -reload has been introduced. If this is specified then mental ray attempts to avoid keeping geometric objects read from the mi file in memory permanently, but re-read them from disk as needed. This optimization is not applicable to all scenes and workflows but may save large amounts of memory on certain complex scenes.

Per-face data

Per-face data have been introduced for polygonal geometric objects. Per-face data make it possible to attach arbitrary floating point data to the primitives. An object may also contain per-primitive data information to identify the data attached to the primitives.

String options

Up to version 3.4, options are hardcoded in the .mi syntax and options data structures. It is now also possible to add new string options in a more extensible way as arbitrary name - value pairs, where the name of the option is an arbitrary string, and the value can be of any of a few basic types.

Photon merging

A new, optional optimization of global illumination using photon mapping called photon merging has been implemented. This feature reduces memory consumption in scenes with complex indirect illumination where were many photons must be shot but only few of them contribute to the rendered image, by merging photons in memory which are closer to each other than a user specified distance threshold. Some scenes which previously could not be rendered in sufficient quality due to memory size of the shot photons may be rendered now using this option.

Built-in shaders

As of version 3.5, mental ray has a small set of simple built-in shaders. These are intended for simple visualizations, where speed and simplicity of scene setup are more important than customization of shading elements. The shaders are designed to be flexible, but to offer only little control over shading details.

Paint shader library

A paint shader library has been added with shaders for metallic paint.

Other improvements

Support for spectral rendering has been improved.
Support for color spaces and color profiles has been extended.
The splitting of large objects with boundaries has been improved.
The handling of out-of-memory errors has been improved.
The rasterizer rendering method has been optimized for better performance.
In the hardware rendering mode, larger images and higher supersampling for antialiasing can now be rendered by automatic tiling of the image into smaller regions.

Scene Description Language

The following changes were made in the .mi scene description syntax:

String options have been introduced in the options syntax.
Instances may contain light lists for illumination and shadow casting.
An option finalgather secondary scale (-finalgather_secondary_scale) has been added to specify a scaling color for secondary bounces.
An option photonvol scale (-photonvol_scale) has been added to scale volume global illumination by a color.
A string option "finalgather contrast" has been added.
A photon map may be specified in an object: photonmap file "filename".
Shader parameters can now be prefixed with colorprofile name to specify the colorspace of the parameter.
String options "globillum merge", "caustics merge", and "photonvol merge" have been added for photon merging.
For lightmapping ("baking"), the file type of writable textures can now be set explicitly in the syntax of writable textures.
Light energies may now contain a color profile or a spectrum: energy colorprofile name color or energy spectrum name.
Lights may now contain a light profile (lightprofile name).
The scalar values in color spectra are no longer separated by commas: spectrum symbol scalar... end spectrum.
Per face data have been added to primitive lists.
A registry variable _MI_REG_MEMORY_LIMIT may now be set to specify the memory limit for mental ray which is otherwise specified with the -memory command line option.

Shader Writing and Integration

The following changes were made in the shader and integration interfaces.

Note: All shaders and integrations should be recompiled for mental ray version 3.5. Many data structures in the shader interface and integration API have been changed. Shaders compiled for an earlier version of mental ray may crash without error indication if they access changed data structures in the shader interface.

The C-style shader and integration interface has been extended with a new C++ interface for shaders and integrations. This interface introduces more modern software interface approaches, simplifies linking and access to functions in mental ray, and supports versioning for future interface changes while maintaining backwards compatibility.
It uses styles and methodologies slightly different from the ones used in the interfaces so far, and requires compilation of the shader or integration as C++. The C interfaces will be continued to be supported but some new features may only be available in the C++ interface. The C++ interface currently provides light iterators for access to light lists in instances, and access to string options.
The camera and options entries in the state structure are now marked as const, meaning that state->camera and state->options can not be modified in shaders. This enforces a restriction which is very important for the correctness of shaders, since the camera and options are shared among multiple threads executing in parallel and can therefore not be modified without undesired and dangerous side effects. A recommended and safer way is to make a copy of the camera or options, modify the copy, and install it in the state; please see the release notes for details.
Two new functions int mi_img_get_width(const miImg_image *) and int mi_img_get_height(const miImg_image *) have been added which return the true, full width of tiled images and textures.
Two new query modes miQ_LIGHT_NORMAL and miQ_LIGHT_TANGENT_SPACE were introduced. They return for area light sources the tangent space of the current sample on the light surface.

Incompatible Changes

The following changes were made in mental ray 3.5 which are not backwards compatible or which may cause different results than previous versions of mental ray:

The default final gathering mode is automatic; use the 3.4 or strict 3.4 mode to get the same behaviour as in version 3.4.
Color clipping is now performed after the accumulation of image samples for motion blur, while it used to be performed before the accumulation. This has the effect of rendering super-white motion blurred objects brighter than white motion blurred objects, and may cause motion blurred objects to be rendered significantly brighter.
Extra samples in the generation of shadow maps have been eliminated. This may result in different images of some scenes using shadow maps, in particular in combination with motion blur, since shadows previously may have been too dark.
The memory limit (specified e.g. with the -memory option) is now by default set to 1 gigabyte on 32 bit architectures. This may affect the memory behaviour of mental ray on some scenes.
The handling of lens shader with the rasterizer has been improved, lens shader are now called in the end like volume shaders. This could break some rare lens shaders which exchange data with material shaders through the state.
The default low samples settings of area lights have been changed from 2 x 2 to 1 x 1. This is more appropriate for the low samples main use in final gathering.
Texture file paths are substituted and replaced by absolute paths now on all machines contributing in network parallel rendering, not just on the master machine.
Final gathering in combination with bump mapping now computes final gathering by default on the "unbumped" surface normals.
The output and exit code of the imf_diff utility has been altered.
If the min samples option is set to -2 or a lower number then mental ray no longer fills coverage buffers.
The bias computations for shadow maps without the Woo trick have been improved.
A limit for the number of auxiliary helper threads which mental ray uses for network rendering, has been removed. This may theoretically increase the number of threads created by mental ray but is not expected to cause a noticable difference.
Some platform configurations have been changed. The platform nt-x86 is now compiled with the Microsoft Visual Studio .net 2003 (VC 7.1) compiler, and no longer with the VC 6 compiler.