Objects

The object scene element is used to encapsulate geometric details of various types of shapes, like freeform surfaces or curves, polygons, triangle or subdivision base meshes, and even hair primitives. All these object types share a set of common properties, referred to as object flags. The object description merely provides the template for the construction of the shape, but only an instance of the object will actually generate a renderable copy that is placed in the virtual world giving it a designated size and orientation.

All geometric information is typically specified relative to its own internal origin, in other words, in object space. In this mode, which is the default, the camera location is determined by its instance, all objects are defined in local object coordinates and positioned in the scene with their object instances. Every object, camera, or light requires an instance to be established in the scene for rendering.

Deprecated In camera space mode, enabled in the scene options, the camera is fixed and assumed to sit at the coordinate origin and point down the negative Z axis, and all objects are defined using camera space coordinates. Camera space mode is supported only for backwards compatibility, and is not recommended.

The appearance of the object, such as color and transparency, is determined by referencing materials in the object definition. Before a material can be used in an object, it must be defined. Assigning a material determines all aspects of the object's appearance. No further parameters, textures, or lights need to be specified; they are all part of the material definition.

The most common approaches to materials and objects are: to define all materials upfront followed by all objects, or material and object definitions may be interspersed. The first approach is preferable when materials are authored separately, and should be included as a single .mi file. Either way, each material definition precedes its first use.

All polygonal and free-form surface objects have the same common format in the .mi file:

object "object_name"
    [ visible [on|off] ]
    [ shadow [on|off|mode] ]
    [ shadowmap [on|off] ]
    [ trace [on|off] ]
    [ reflection mode ]
    [ refraction mode ]
    [ transparency mode ]
    [ caustic on|off ]
    [ caustic [mode] ]
    [ globillum on|off ]
    [ globillum [mode] ]
    [ photonmap file ["file_name"|off] ]
    [ finalgather on|off ]
    [ finalgather mode ]
    [ finalgather file ["file_name"|[file_list]|off] ]
    [ finalgather force on|off ]
    [ select [on|off] ]
    [ tagged [on|off] ]
    [ face front|back|both ]
    [ box [x_min y_min z_min x_max y_max z_max] ]
    [ motion box [x_min y_min z_min x_max y_max z_max] ]
    [ max displace value ]
    [ ray offset value ]
    [ samples min max ]
    [ shading samples samples_scalar ]
    [ data "data_name"|null ]
    [ tag label_int ]

    [ file "file_name" ]

    [ basis_list ]
    group
        vector_list
        vertex_list
        geometry_list
        approximation_list
    end group
end object

The individual parameters are:

object "object_name"

The object_name character string is used to uniquely identify the object.

The name is not used by mental ray in any form except for progress reports and error messages. Object names should be enclosed in double quotes to disambiguate them from reserved words, even though this is not mandatory.

visible [ on|off ]

Determines the primary visibility of the object.

Most objects will have this flag set. Not setting it, or setting it to off will make the object invisible to primary rays, which are rays originating from the camera. This means that the object will disappear from the rendered image.

The default is off, for not visible.
Given without a boolean argument sets it to on.

Note Transparency rays will also hit objects marked visible, if the transparency mode is not touched. In this case, transparency is not considered a special case of refraction, but as a scanline rendering feature. This apparent inconsistency permits usingtransparency effects even if ray tracing is turned off altogether. More importantly, it also keeps objects that have the visible flag but not the trace flag out of the BSP tree that is used to accelerate ray tracing, which has a significant performance impact. This tradeoff can become very significant for large scenes.

shadow [ on|off|mode_int ]

Controls contribution of the object to shadows.

When using the form with the boolean argument on or off it enables or disables just shadow casting - shadows thrown from this object onto other objects, but the object will always receive - or show - shadows from other objects. The mode allows to control shadowing details of the object explicitly. It specifies whether the object casts shadows (1), receives shadows (2), does both (3), or has no impact on any shadows (0). The shadow flags are ignored if shadows are globally disabled, like in the scene options.

The default is off, for not casting but only receiving shadows.
Given without any argument sets it to on.

Note If an object is very complex, it may be desirable to set only the visible flag but not the shadow flag, and create a second object that resembles the first one but is much simpler, and set its shadow but not the visible flag. The effect is that the object appears unchanged, but shadow calculations see a much simpler shadow object that casts approximately the same shadow as the primary visible object would.

shadowmap [ on|off ]

Determines if the object contributes to a shadowmap.

For an object to cast shadows from a shadowmap, the shadow mode must have casting enabled as well. This flag may be useful to control shadows from shadowmaps differently than ray tracing shadows. For example, for volume bounding geometry called hull objects, such as light cones, disable to cast any shadows from a shadowmap but keep shadow tracing enabled.

The default is on.
Given without any argument sets it to on.

Note This flag is rarely used because the regular shadow flag impacts shadowmap shadows as well.

trace [ on|off ]

Deprecated This flag is supported for backwards compatibility. It is superseded by reflection, refraction, and finalgather flags.

Controls contribution of the object to secondary ray tracing effects.

Determines whether the object is visible to secondary rays for reflection, refraction and finalgather. When using the form with the boolean argument on or off it enables or disables casting of reflections, refractions, and finalgather onto other objects, but the object will always receive - or show - these effects from other objects. If given without any argument it defaults to on (cast and receive) for the respective effects. This flag has no effect if ray tracing is globally disabled in the scene options with trace off.

The default is off, for not casting but only receiving reflections, refractions, and final gathering.
Given without any argument sets it to on.

Note If the reflecting or refracting objects are fuzzy or only slightly reflective or refractive, it may result in a considerable speedup to let the reflection and refraction rays see a much simpler object than primary rays would. Like with the shadow flag, this is achieved by not setting the trace flag in the primary, high-definition object, and create a second one that roughly resembles the primary object that has the trace flag (and perhaps the shadow flag) but not the visible flag set.

reflection mode_int

Controls if an object is visible in reflections on other objects, and if it shows reflections of other objects.

A value of 1 enables reflection casting, in other words, the object becomes visible in reflections on other objects. A value of 2 enables reflection receiving, thus will show reflections of other objects on this object. A value of 3 enables both, and 0 disables both.

Default is 2, for not casting but receiving reflections.

Note The phrase casting reflections has an opposite meaning to casting reflection rays: in order to cast reflections or become visible in reflections on other objects an object must receive reflection rays, and to receive reflections or show reflections of other objects an object has to cast reflection rays. The mode controls casting and receiving of the shading effect rather than rays, similar to the meaning of casting shadows.

refraction mode_int

Controls if the object is visible in refractions on other objects, and if it shows other objects in refractions.

A value of 1 enables refraction casting, in other words, the object becomes visible in refractions on other objects. A value of 2 enables refraction receiving, thus will show other objects in refractions. A value of 3 enables both, and 0 disables both.

Default is 2, for not casting but receiving refractions.

Note The phrase casting means casting a refraction, which has an opposite meaning to casting a refraction ray. See the note about reflection mode for an explanation.

transparency mode_int

Controls if an object is visible through transparent parts of other objects, and if it shows other objects in transparent parts.

A value of 1 enables transparency casting, in other words, the object becomes visible through transparent regions of other objects. A value of 2 enables transparency receiving, thus will show other objects in transparent parts. A value of 3 enables both, and 0 disables both.

Default is 3, for both cast and receive transparency.

caustic on|off

Determines if the object participates in generation of caustics at all.

Setting this flag to off means that the object is completely hidden from caustics effects. In other words, the object is considered as not existing for caustic photons, and they will pass the space covered by the object perfectly unaffected. Setting the flag to on generally enables contribution of this object to caustics, and further details of how caustics are treated can be controlled with the caustic mode flag below.

The default is on, for enabled for caustics.

caustic [ mode_int ]

Controls details of the interaction of caustics with the object.

A value of 1 enables caustic casting, 2 enables caustic receiving, 3 enables both, and 0 disables both. This mode has an effect only if caustic interaction for this object has not been disabled, like with caustic off.

The default is 0, for not casting nor receiving caustics.
Given without a mode argument sets it to 3, for both cast and receive caustics.

Note This setting can be used to optimize performance by focusing the effect on the desired scene elements. The cast flag impacts the emission of caustic photons into the scene; they will be shot towards casting objects only and, hence, avoid to waste shooting photons in other directions. The receive flag marks objects for photon storage, hence has influence on the size of the photon map.

Example
To focus caustic effects on specific objects, the global options should disable caustic interaction by default, but the involved objects should have the respective caustic flags set. To render caustics in a water pool, the flags may be set like this:

options "options"
    caustic 0
end options

object "water"
    caustic on
    caustic 1
end object

object "ground"
    caustic on
    caustic 2
end object

globillum on|off

Determines if the object participates in generation of global illumination from photons at all.

Setting this flag to off means that the object is completely hidden from global illumination photon effects. In other words, the object is considered as not existing for such photons, and they will pass the space covered by the object perfectly unaffected. Setting the flag to on generally enables contribution of this object to global illumination, and further details can be controlled with the globillum mode flag below.

The default is on, for enabled for global illumination photons.

Note The globillum on flag marks specific objects to participate in optimized generation of global illumination. Global illumination is a lighting effect caused by diffuse interreflection, such as a red table bleeding red color onto an adjacent white wall. To simulate this effect, the material of both the table and the wall must provide photon shaders, the light source must emit globillum photons and contain appropriate photon energy settings, and the material shader of the wall must pick up global illumination light information.

Note If an object is very complex it may be desirable to set only the visible flag but not the globillum flag, and create a second object with simplified geometry that resembles the first one with the globillum but not the visible flag set. The effect is that the rendered object appears in full detail but simulation of global illumination is faster since a simpler object representation is used.

globillum [ mode_int ]

Controls details of the interaction of global illumination photons with the object.

A value of 1 enables global illumination casting, 2 enables receiving, 3 enables both, and 0 disables both. This mode has an effect only if global illumination interaction for this object has not been disabled, like with globillum off.

The default is 0, for not casting nor receiving global illumination photons.
Given without a mode argument sets it to 3, for both cast and receive global illumination photons.

Note This setting can be used to optimize performance by focusing the effect on the desired scene elements. The cast flag impacts the emission of global illumination photons into the scene; they will be shot towards casting objects only and, hence, avoid to waste shooting photons in other directions. The receive flag marks objects for photon storage, hence has influence on the size of the photon map.

Example
To enable global illumination effects only on specific objects, the global options should disable global illumination interaction by default, but the involved objects should have the respective globillum flags set. To render global illumination in a white room with a red table, the flags may be set like this:

options "options"
    globillum 0
end options

object "room"
    globillum on
    globillum 3
end object

object "table"
    globillum on
    globillum 3
end object

finalgather on|off

Determines if the object participates in final gathering at all.

Setting this flag to off means that the object is completely hidden from final gathering. In other words, the object is considered as not existing for finalgather rays, and they will pass the space covered by the object perfectly unaffected. Setting the flag to on generally enables contribution of this object to final gathering, and further details of how finalgather rays are treated can be controlled with the finalgather mode flag below.

The default is on, for enabled for final gathering.

finalgather mode_int

Controls details of final gathering interaction with the object.

A value of 1 enables finalgather casting, 2 enables receiving, 3 enables both, and 0 disables both. This mode has an effect only if final gathering interaction for this object has not been disabled, like with finalgather off.

The default is 2, for not casting but receiving final gathering effects.

Note If the final gathering cast bit is not set, like with finalgather 1 or finalgather 3, but

finalgather
on

is given, then finalgather rays would be terminated by the object and return a black color.

finalgather file [ "file_name" | [file_list] |off ]

Controls final gathering cache per object.

The finalgather points computed for this object are stored in the given disk file(s). This helps to reduce the memory requirements compared to using a single global map. If multiple files are given as file_list, a comma-separated list of file names within square brackets, then all finalgather maps are merged upon reading but the first file in the list is used to write the final map, similar to the behavior of the global finalgather file option. If identical file names or file name lists are used for several objects, only one copy of those finalgather maps will be kept in memory. The values off or "" can be used to explicitly disable this feature, for example in the next frame of an animation.

The default is off.

finalgather force on|off

Controls final gathering mode for this object.

When the value of the flag is on, then final gathering for this object will be computed in force mode instead of using interpolation from nearby finalgather points. This can be used to tune for higher final gathering rendering quality just on selected areas of the scene without impacting the overall performance too much. The mode may also be explicitly switched to off on this object but enabled just for a particular instance of the object.

The default is off.

select

Deprecated This flag is kept for backwards compatibility. It has no effect.

tagged [ on|off ]

Controls how materials are referenced, directly or by index.

Every polygon, surface, and triangle can get a material assigned directly. If an object does not specify any material, it will be inherited from the instance automatically. In contrast, objects marked tagged do not store any material directly at all but always rely on the inherited instance material, or list of materials. In this case, a mandatory label integer value is provided in place of the material, in polygon and surface definitions. If the instance specifies a list of materials then the integer value is used as an index into that list to find the related material. The label can also be accessed by shaders during rendering using the miQ_GEO_LABEL query. Thus, if a shader is shared on multiple objects it can distinguish tagged parts of the object by label.

The default is off, for not tagged.
Given without any argument sets it to on.

face front|back|both

Controls face culling for the object.

Determines if either the front or the back side of the object faces are rendered, or both sides. This flag overrides the global default given in the scene options for this object only. Shaders may overrule the behavior during rendering.

The default is unspecified, for determined by the global option.

box [ x_min y_min z_min x_max y_max z_max ]

Specifies the bounding box of the static object.

The bounds are given as two 3D vectors in object space providing the minimum and maximum coordinates in x,y,z of all surface points. For polygons, this is easy to compute: the box components are simply the smallest (min) and largest (max) components of all points contained in the vectors. For free-form surfaces, the box can in principle be derived in the same way from the control points; however, such a box would often be much too large, especially for NURBS. Boxes that are too large are legal but may be inefficient. If the box statement is omitted, mental ray will compute one from all vertices, which may not be very efficient as well. The object bounds are needed for demand loading using placeholder geometry.

The default is unspecified, for automatically computed.
Given without any argument sets it to unspecified, for no box.

NoteThis box should not include enlargement caused by displacement or motion.

motion box [ x_min y_min z_min x_max y_max z_max ]

Specifies the motion deformation bounds of the object.

This bounding box contains the minimum and maximum components of all motion vectors for this object. Note, that this is not the same as providing the smallest and largest distance vector; but these vectors are put together from the individual smallest and largest x,y,z components of all motion vectors. A motion box must be specified if the object has both a bounding box and motion vectors, otherwise loading on demand may fail for deforming objects.

The default is unspecified.
Given without any argument sets it to unspecified, for no motion box.

max displace value

Specifies maximum displacement bounds of the object.

This setting should provide the maximum displacement applied to object control points in local object space in normal direction. If a displacement shader returns values grater than value, the rendered geometry may appear clipped and a warning is printed. Displaced objects should always specify this value to accelerate displacement calculation.

The default is 0.0, for truncate to zero or disable displacement.

Note A max displace value that is too large generates correct images but puts more pressure on memory and performance, since larger areas of an object may be displaced, thus creating much more triangles in longer time. In particular, mental ray may suffer serious performance losses, even by an order of magnitude, if the displaced object uses fine approximations. If the max displace value is too small, mental ray will limit the absolute displacement to this value.

ray offset value

Offset the ray origin when tracing rays from this object.

This setting allows to enforce an offset for the origin of secondary rays cast from the geometry, in world space units. When a ray is traced, any intersections that are closer to the emanating surface than the given offset value are ignored. This can be useful to avoid self-intersections, especially in situations where multiple, slightly different, tessellations of the same geometry are used, like rasterized triangles for primary visible rendering and approximate tessellations for ray tracing.

samples min_int max_int

Controls primary sampling for this object.

This statement allows to tune adaptive oversampling per object. For all pixels that touch this object, the global samples range specified in the scene options is constrained to the min max range, such that at least 2^min and at most 2^max samples are taken. Constraints that exceed the min and max global limits are ignored, so that the options can be used to force quick preview rendering without having to change all object sampling specifiers. If the pixel touches multiple objects, perhaps due to reflection or refraction, the highest min and the highest max are used. If the pixel touches no objects at all, the default global limits defmin and defmax defined in the scene options are used.

The defaults are -128 127, for disable per-object sampling limits.

This setting is not used by the rasterizer.

shading samples value

Controls rasterizer sampling for this object.

This will override the global shading samples setting for this object. It is also possible to specify this setting per-instance, if needed, where the instance setting will override the object setting, when both are present. Typically, this is used to keep the global default value low for overall performance, but boost the setting for selected objects or instances only, to achieve the desired quality in the final rendering.

The default is 0.0, disable per-object shading samples.

This setting is only used by the rasterizer.

tag number_int

Assigns a label number to identify this object.

The number is an arbitrary 32-bit integer value. The term label is normally used in mental ray for this identifier, the keyword tag is retained for backwards compatibility (a tag is used in shaders to identify entries in the scene database). Using an appropriate framebuffer or output statement in the render cameras it is possible to let mental ray write a label file that, for each pixel, contains the label code of the object that the camera ray hits first. Which label exactly is stored is under control of the material shader. Using the label of the foremost object regardless of reflections and refractions is merely the default behavior.

The default is 0, for no label.

data "data_name"|null

Attach user data to this object.

The argument must be the name of a previously defined data element in the scene. If the keyword null is specified instead of a data element name, a previously assigned data reference is removed. Several user data blocks may be chained together using multiple data statements.

The default is null.

file "file_name"

Specifies a file for demand loading geometry for this placeholder object.

This statement is used to enable lazy loading of the final object from another file at render time, when mental ray hits the bounding box of the object, also known as a placeholder object. Such an object needs to specify valid bounds for the static, moving, and displaced geometry using box, motion box and max displace settings. The actual geometry normally given inside the object group can be omitted. The specified file must contain an exact copy of the object, with the file statement there is replaced with the complete object group block containing the actual geometry. The incremental keyword needs to be specified for that object. Such an object may have only a single object group. Demand loading objects are useful to avoid parsing and permanently storing large objects that may or may not be required for rendering.

The default is unspecified, for no placeholder.

NoteThe standalone mental ray supports the -echo explode command line option to automatically create demand-loaded object files. On the other hand, the -reload command line option allows to utilize demand loading without generating extra files.

NoteThe assembly mechanism allows to load larger parts of a scene on demand, like a full character or a complete building.

Example
The object definitions in the master scene and the placeholder file look like this:
placeholder object:

object "my_object"
    # flags ...
    box ...
    motion box ...
    max displace ...
    file "my_object.mi"
end object

object file my_object.mi:

incremental
object "my_object"
    # flags ...
    group
        # geometry ...
    end group
end object

basis_list

Specifies the bases for free-form surface geometry.

Only free-form surfaces and curves use bases. The defined bases can be used in all groups that follow, until the end of the object.

group

This starts a section that contains the actual geometric representation. Multiple object groups are deprecated but supported for backwards compatibility; mental ray will automatically convert a multi-group object to an instance group containing multiple sub-objects. See object group for details.

Note The visible, shadow, trace, reflection, refraction, transparency, caustic, globillum, finalgather and face flags can be overridden by the instance using the standard inheritance mechanism. Instances can specify that a flag in the instanced object is turned on or off, or that it is left unchanged. The object flags are used only if all instances from the root of the scene graph down to the object all leave the flag unchanged.

Object Group

Object groups contain the actual geometry. Common to all geometry are vector lists and vertex lists.

Vector List

The vector list contains 3D vectors which describe points in space, normals, texture vertices, basis vectors, motion vectors, and others. Vectors are anonymous, they are triples of floating-point numbers separated by whitespace without inherent meaning. They are numbered beginning with 0. Numbering restarts at 0 whenever a new object group starts.

mental ray also accepts a compressed binary format for vectors. Instead of three floating-point numbers, a sequence of 12 bytes enclosed in backquotes is accepted. These 12 bytes are the memory image of three floats in IEEE 854 format, using big-endian byte order. This format is intended to increase translation and parsing speed when ray is connected to a native translator; it should not be used in files to be modified with text editors or filters, since they may refuse to accept files containing binary data, or corrupt them.

Vertex List

Vertices build on vectors. In the .mi format, there is no syntactical difference between polygon vertices and control points vertices for free-form surfaces; both are collectively referred to as vertices in this discussion. All vertices define a point in space and optional vertex normals, motion vectors, derivatives, zero or more textures and basis vectors, and optional user vectors:

v index_int  
[ n index_int ]  
[ d index_int index_int [ index_int [ index_int index_int ] ] ]  
[ t index_int [ index_int index_int ] ]  
[ m index_int ]  
[ u index_int  
...

Polygon vertices may use all of these. Free-form surface control points may use v and m only; the others are either computed analytically or are specified in other ways as part of the surface definition.

v: specifies the point in space,
n: specifies the vertex normal vector (ignored when the vertex is used as a curve or surface control point). A normal vector defines the orientation (vertical) of the surface at this vertex.
m: specifies the motion vector. mental ray permits up to 15 m statements per v statement for nonlinear motion paths. If multiple motion vectors are specified, the shutter time interval 0...1 is equally subdivided, and each motion vector specifies the motion from the origin specified by the v vector to the corresponding point on the path. In other words, multiple motion vectors do not form a connected chain but a bundle of "rays" originating at the same point v.
t: specifies a texture vertex with an optional X/Y basis vector pair for bump map calculation. There can be up to 64 t lines for any given v vertex. The texture and basis vectors are ignored when the vertex is used as a curve or surface control point. Texture and basis information for surfaces is defined using a "texture surface", described below.
d: specifies a first and/or second surface derivative. First derivatives describe the UV parametric gradient of a surface; second derivatives describe the curvature. mental ray can compute surface derivatives analytically for free-form surfaces but not for polygons because polygons have no inherent UV coordinate space. Therefore, the d keyword must be used to define explicit surface derivatives for polygons. If d is followed by two indices they are taken to reference the first derivative dP du and dP dv (with P being the point in space); if three indices follow they are taken to reference the second derivative d²P du², d²P dv², and d²P du dv; and if five indices follow the first two describe the first derivatives and the next three the second derivatives. Derivatives are not used by mental ray, they are made available to shaders only.
u: specifies a user vector. No constraints are imposed on user vectors. mental ray does not operate on them in any way, they are passed through with the vertex and can be picked up by the shader.

Every vertex begins with a v statement and ends with the next v statement or with the start of the geometry description. All occurrences of index above reference the vector list; 0 is the first vector in this group. References of different types (for example, v and n) may not reference the same vector. As stated before, all vectors are 3D. If the third coordinate is not used (as is the case for 2D texture vertices, for 2D curve control points, and for 2D surface special points) it should be set to 0.0 by convention. If both the second and third coordinates are unused (as is the case for 1D curve special points), they should both be set to 0.0.

Vertices themselves are numbered independently of vectors. The first vertex in every group is numbered 0. The geometry description is referencing vertices by vertex index, just like vertices are referencing vectors by vector index. This results in a three-stage definition of geometry:

List of vectors
List of vertices
List of geometry

The reason for this three-stage process is that it allows both sharing vectors and sharing vertices. This is best illustrated with the following examples.

Example

The examples will use the simplest form of polygon syntax that will be described later in this section.

Consider two triangles ABC and ABD sharing an edge AB. The simplest definition of this two-triangle object is:

object "twotri"
    visible
    group
        0.0  0.0  0.0
        1.0  0.0  0.0
        0.0  1.0  0.0
        1.0  0.0  0.0
        1.0  1.0  0.0
        0.0  1.0  0.0
 
        v 0
        v 1
        v 2
        v 3
        v 4
        v 5
 
        p "material_name" 0 1 2
        p 3 4 5
    end group
end object

The first three vectors are used to build the first three vertices, which are used in the first triangle. The remaining three vectors build the next three vertices, which are used for the second triangle. Two vectors are listed twice and can be shared:

object "twotri"
    visible
    group
        0.0  0.0  0.0
        1.0  0.0  0.0
        0.0  1.0  0.0
        1.0  1.0  0.0
 
        v 0
        v 1
        v 2
        v 1
        v 3
        v 2
 
        p "material_name" 0 1 2
        p 3 4 5
    end group
end object

The order of vector references is noncontiguous to ensure that the second triangle is in counter-clockwise order. Two vertices are redundant and can also be removed by sharing:

object "twotri"
    visible
    group
        0.0  0.0  0.0
        1.0  0.0  0.0
        0.0  1.0  0.0
        1.0  1.0  0.0
 
        v 0
        v 1
        v 2
        v 3
 
        p "material_name" 0 1 2
        p 1 3 2
    end group
end object

The need for sharing both vectors and vertices can be shown by specifying vertex normals:

object "twotri"
    visible
    group
        0.0  0.0  0.0
        1.0  0.0  0.0
        0.0  1.0  0.0
        1.0  1.0  0.0
        0.0  0.0  1.0
 
        v 0  n 4
        v 1  n 4
        v 2  n 4
        v 3  n 4
 
        p "material_name" 0 1 2
        p 1 3 2
    end group
end object

In this last example, both vector sharing and vertex sharing takes place. The normal in this example is actually redundant: if no normal is specified, mental ray uses the polygon normal. Defaulting to the polygon normal is slightly more efficient than interpolating vertex normals, if vertex normals are explicitly specified.

Geometry List

Various types of geometry can be contained in the geometry_list, like polygonal geometry, triangle lists, free-from and subdivision surfaces. In the next sections the syntax of these definitions is described in detail.

Approximation List

At the end of each object group several approximation statements may be given which control the tessellation of the geometry. In polygonal object groups, the approximation is used only for polygons whose material contains a displacement shader. Free-form and subdivision surfaces are always controlled by their approximations. See approximations for details.

An object group permits only one type of geometry, either polygons or surfaces but not both. It is recommended that objects contain only a single object group, so normally objects are either polygonal or surface objects but not both at the same time. Also, vector sharing is supported only for vectors of similar types (point in space, normal, motion, texture, basis vector, derivative, or user vector. A vector may not be referenced by vertices once as a point in space and once as a normal, for example.