Upgrading from mental ray 3.11 to 3.12

What's new in Version 3.12
Scene Description Language
Shader Writing and Integration
Incompatible Changes

What's new in Version 3.12

Here is a summary of some of the new features and feature improvements in version 3.12 of mental ray. Please refer to the release notes for more details and for other changes which are not mentioned here.

Global Importance Sampling

The Global Importance Sampling is a new advanced optimization technique that allows to render modern and complex lighting setups much more efficiently without the need to adjust the scene. It combines the various separate importance sampling techniques into a single feature, to accelerate lighting and rendering in general.

The new approach is superior to previous attempts because it is able to distribute samples across multiple lights based on "importance". This will significantly reduce noise and rendering time, especially for scenes with a larger number of lights, especially "area" lights. The optimization is implemented in the rendering kernel so that it works seamlessly with existing shaders, no code changes are required.

The feature is currently disabled by default. It can be enabled and controlled with string options and on the command line of standalone mental ray.

Unified Sampling Flickering Filter

The Unified Sampling filtering scheme has been improved to avoid flickering artifacts. It improves filtering quality for scenes with large min/max samples variations, and helps to reduce artifacts along object edges, very bright but small highlights, and flickering in animations.

The feature is currently disabled by default, since it imposes a small memory overhead. It can be enabled and controlled with string options and on the command line of standalone mental ray.

Global Illumination GPU

This version of mental ray offers a GPU accelerated implementation of global illumination effects. In this initial form, the GI GPU mode speeds up indirect illumination through diffuse bounces of light, similar to what final gathering typically computes.

The GI GPU acceleration can be enabled and controlled with string options and on the command line of standalone mental ray.

OpenEXR 2.0

This version of mental ray has integrated OpenEXR 2.0, which provides new features like "multi-part" file storage, and support for "deep image" formats, among others.

By default, mental ray now writes multiple layers into separate OpenEXR Parts. By convention, the names of OpenEXR Parts match the names of the mental ray frame buffers.

The OpenEXR multi-part support can be controlled with a registry setting, to retain strict backwards compatibility with previous .exr files.

iray

The iray version 3.5 is a major update of this rendering mode built into mental ray. It adds support for the most recent NVIDIA GPU architectures, while retaining compatibility to previous generations of GPUs. This version also uses the latest CUDA runtime for best performance. The iray rendering core has been further optimized towards stability and scalability in multi-GPU and CPU/GPU setups.

Human Hair Shader

The new shader package basehair provides an implementation of a hair shading model to render human hair effects very efficiently. Even though it uses approximations of the complex lighting and scattering effects, the rendered results prove to be much more realistic than previous solutions for mental ray.

Improved Layering Shaders

The new shader package layering was introduced in the previous version of mental ray. The current version has been partially re-worked and greatly improved towards ease of use, and seamless integration with mental ray's frame buffer system.
The layering shader package introduces a new way of material compositing by layering base components, like illumination models and reflection properties. Commonly used mental ray shaders like architectural material or subsurface scattering are also provided as standard base layers in this package.

Scene Description Language

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

To enable and tune Global Importance Sampling, these string options with new values are supported:
```
"light importance sampling" "on"|"all"|"local"|"off"
```
The values are:

"on"
Enable light importance sampling across all area lights in the scene.
"all"
Enable light importance sampling across all area and point/spot lights in the scene. This is useful for scenes with large number of point lights. However, for scenes with one or few point lights, the value "on" is recommended.
"local"
Importance sampling is applied to each light individually, with no cross-balancing among the lights. This corresponds to the light importance sampling performed in the previous version of mental ray 3.11.
"off"
Light importance sampling is disabled. This is the default.
The following options are available for further control:
```
"light importance sampling quality" num_scalar[num_scalar]
"light importance sampling precomp" on|off
"light importance sampling resolution" num_int
```
On the command line of standalone mental ray these arguments can be used to override of the scene settings:
```
-light_is on/off/all/local
-light_is_quality Q [F]
-light_is_precomp on/off 
```
Please see the release notes for more details.
To enable and tune the new Unified Sampling Filter scheme, a new string option is supported:
```
"unified sampling subfilter" on|off
```
On the command line of standalone mental ray this argument can be used:
```
-unified_sampling_subfilter on/off
```
For Global Illumination GPU rendering, these new string options are supported:
```
"gpu gi" "on"|"diffuse"|"off"
```
The values are:

"on"
for general GPU acceleration enabled. The diffuse paths are computed on the GPU, specular sub-paths are computed on the CPU. The performance gain compared to CPU rendering is moderate, however the rendered images are of superior quality.
"diffuse"
GPU acceleration is enabled, but only diffuse paths are computed. Performance gain is dramatic, however some scenes may be darker due to missing illumination from light paths through specular bounces.
"off"
GPU acceleration is disabled.
On the command line of standalone mental ray the following argument can be used:
```
-finalgather gpu
```
For OpenEXR 2.0 multi-part support, the following registry setting is supported:
```
registry "{_MI_OPENEXR_MULTI_PART}" value "on"|"off"
```
If the registry value is set to "off", mental ray creates single-part multi-layer .exr files, which are compatible with OpenEXR 1.7 (unless other new OpenEXR 2.0 features are used).

Shader Writing and Integration

For progressive rendering and iray interactive rendering, the integration API has been extended with a callback mechanism. Applications that register the callback will get notified of progressive rendering events asynchronously, rather than polling the status using the existing API functions.
```
typedef miBoolean (*miRc_progressive_callback)  (
    void        *data,      /* callback data */
    miUint      subframe,   /* subframe count */
    miUint      what);      /* event type */

miBoolean mi_rc_progressive_set_callback(
    miRc_progressive_callback   callback,
    void                        *callback_data);
```
Please see the release notes for more details.
For Alembic support in the abcimport shader, the binary is now based on the Alembic version 1.5, which adds support for the Ogawa file backend giving better performance, among others.

Incompatible Changes

The new OpenEXR 2.0 multi-part files are not backwards compatible with tools based on OpenEXR 1.7 or earlier library. Unless the number of parts is equal to one, such tools would report error on attempt to open the file. In particular, this applies to previous versions of mental ray 3.11 and earlier, including the related imf tools coming with those versions.