SkinningAttribute.kl¶
Types¶
SkinningAttributeData (struct)¶
A simple struct that holds the data for the SkinningAttribute. This struct can be moved to the GPU wholesale, and its methods will also continue to work on the GPU. This is a workaround untill objects are supported on the GPU. currently only structs are supported, so we move the data and important fucntions to the struct, and add the struct to objct as a member.
Members¶
Size | count | |
UInt32ArrayAllocator | weightsAllocator | |
UInt16[] | ids | |
UInt32[] | offsets | |
UInt16[] | counts |
Methods¶
SkinningAttributeData ( in SkinningAttributeData other ) | |
SkinningAttributeData () | |
SkinningAttributeData | clone ? () |
cloneMembersTo ? ( io SkinningAttributeData that ) | |
copyFrom ! ( in SkinningAttributeData other ) | |
copyValue ! ( in Size sourceIndex, in Size targetIndex ) | |
copyValue ! ( in SkinningAttributeData source, in Size sourceIndex, in Size targetIndex ) | |
Boolean | equalValues ? ( in Size index1, in Size index2 ) |
getLinearCombination ? ( in LocalL16UInt32Array sourceIndices, in LocalL16ScalarArray sourceWeights, out LocalL16UInt32Array skinIds, out LocalL16ScalarArray skinWeights ) | |
getPair ? ( in UInt32 index, in UInt32 offset, out UInt16 id, out Scalar weight ) | |
UInt16 | getPairCount ? ( in UInt32 index ) |
getPairs ? ( in UInt32 index, out LocalL16UInt32Array skinIds, out LocalL16ScalarArray skinWeights ) | |
linearCombine ! ( in Ref<Object> sourceAttr, in LocalL16UInt32Array sourceIndices, in LocalL16ScalarArray sourceWeights, in Size targetIndex ) | |
Size | memUsage ? () |
normalize ! ( in Size index ) | |
resize ! ( in Size count ) | |
setPair ! ( in UInt32 index, in UInt32 offset, in UInt16 id, in Scalar weight ) | |
setPairCount ! ( in UInt32 index, in UInt16 pairCount ) | |
setPairs ! ( in UInt32 index, in LocalL16UInt32Array skinIds, in LocalL16ScalarArray skinWeights ) | |
Size | size ? () |
GeometryAttributes (object)¶
The GeometryAttributes object is a generic container that contains a named set of GeometryAttribute. Each GeometryAttribute is an array containing values of a specific type. For example, the “positions” attribute is typically a Vec3Attribute containing Vec3 elements.
All Geometry types, such as the PolygonMesh, contain a GeometryAttributes container which can be retrieved using the Geometry.getAttributes interface method. However, it is the geometry that controls the order and meaning of the attribute elements within the GeometryAttribute array. For example, the PolygonMesh maintains information about how the values are shared.
注釈
Attributes needs to be uniquely identified by a name. The GeometryAttributes cannot contain multiple attributes with the same name, even if the type differs.
Attribute versioning¶
In order to allow for optimal performance and multithreading, attributes contain a version number (UInt32) which has to be explicitly incremented after value modifications. However, if the size of the attribute array changes, the version will be automatically incremented. Attribute clients, such as InlineDrawing, can then detect value changes efficiently in order to update their state.
注釈
This versioning strategy favors performance over simplicity. Another choice could have been to automatically increase the version when detecting a value difference. Another could have been to let the clients compare values to their last state using a copy or a hash of the values. However, these would have caused performance overhead that could be relatively important in some situations.
Similarly, the GeometryAttributes.keysVersion member defines a version for the set of attributes, which will get incremented only if attributes get added or removed from the container, but not upon attribute resize or value changes.
Members¶
UInt32 | keysVersion | Version of the attribute set. This is incremented only if attributes are added or removed, and not if values are changed or values array is resized. |
UInt32 | attributesValueVersion | |
Ref<GeometryAttribute> | positionsAttribute | Fast access to the positions attribute, if it exists. Note: might be Vec3Attribute or Vec3_dAttribute depending on the precision. |
Ref<GeometryAttribute> | normalsAttribute | Fast access to normals attribute, if it exists. |
String | debugName | Name to be used by descriptions |
/*
** Example: GeometryAttributes
*/
require Geometry;
function reportAttributes( Geometry geom ) {
// Get the attributes container from the geometry
Ref<GeometryAttributes> attributes = geom.getAttributes();
report('Number of attributes: ' + attributes.attributeCount());
report('Number of elements: ' + attributes.size());
// Iterate over the attributes
GeometryAttributesIterator iter = attributes.getAttributesIterator();
Ref<GeometryAttribute> attribute = null;
while( attribute = iter.getNext() )
report( 'Attribute ' + attribute.getName() + ' of type ' + attribute.type() );
}
operator entry() {
PolygonMesh p();
p.addSphere( Xfo(), 1.0, 3, true, true );
reportAttributes(p);
}
/*
** Output:
Number of attributes: 3
Number of elements: 77
Attribute positions of type Vec3Attribute
Attribute normals of type Vec3Attribute
Attribute uvs0 of type Vec2Attribute
*/
Methods¶
SkinningAttribute (object)¶
brief The SkinningAttribute
contains multiple pairs of id and weight per index.
For an index, id-weight pairs can be set one by one or in batch. It provides, too, helpers such as
SkinningAttribute.normalize to set weight sum to 1.0. By implementing methods such as
GeometryAttribute.LinearCombination, it can properly support modeling operations.
Various strategies are used to lower memory consumption and increase performance:
- contiguous memory storage in a few buffers to lower memory fragmentation and minimize memory allocations
counts
per index is implicit if values are appended in order (0 to N-1)- ids are stored
UInt16
, which lowers memory usage and still allows for up to 64K deformer ids
Members¶
UInt32 | version | |
String | name | |
UInt32 | containerIndex | |
StatisticsAutoRegisterMember | autoStats | |
Size | count | |
Ref<GeometryAttributes> | container | |
SkinningAttributeData | internalData | |
UInt32 | memoryLocation |
/*
** Example: SkinningAttribute
*/
require Geometry;
operator entry() {
SkinningAttribute skinningAttr = SkinningAttribute();
//Create 2 skinned items
skinningAttr.resize(2);
//Item 0: 2 id-weight pairs:
// pair 0 = [20:0.25]
// pair 1 = [21:0.75]
skinningAttr.setPairCount( 0, 2 );
skinningAttr.setPair( 0, 0, 20, 0.25 );
skinningAttr.setPair( 0, 1, 21, 0.75 );
//Item 1: set 3 id-weight pairs in batch = [30:0.5, 31:0.5, 32:0.25]
LocalL16UInt32Array indices;
indices.push(30);
indices.push(31);
indices.push(32);
LocalL16ScalarArray weights;
weights.push(0.5);
weights.push(0.5);
weights.push(0.25);
skinningAttr.setPairs(1, indices, weights);
Scalar weight;
UInt16 id;
for( Size i = 0; i < skinningAttr.getPairCount(0); ++i ) {
skinningAttr.getPair( 0, i, id, weight );
report("Item0: pair" + i + ": id="+id+" weight="+weight);
}
skinningAttr.getPairs(1, indices, weights);
report("Item1: ids="+indices.getString()+" weights="+weights.getString());
skinningAttr.normalize(0);
skinningAttr.normalize(1);
report("Values after normalize:");
report( skinningAttr.getDesc( true) );
}
/*
** Output:
Item0: pair0: id=20 weight=+0.25
Item0: pair1: id=21 weight=+0.75
Item1: ids=[30, 31, 32] weights=[+0.5, +0.5, +0.25]
Values after normalize:
Attribute: nb elements = 2
values:[
[20:+0.25, 21:+0.75]
[30:+0.400024, 31:+0.400024, 32:+0.200012]
]
*/
Methods¶
SkinningAttribute ( in SkinningAttribute other ) | |
SkinningAttribute () | |
Boolean | checkConsistency ? () |
SkinningAttribute | clone ? () |
cloneMembersTo ? ( io SkinningAttribute that ) | |
convertToCPU ! () | |
convertToGPU ! () | |
copyValue ! ( in Ref<Object> sourceAttr, in Size sourceIndex, in Size targetIndex ) | |
copyValue ! ( in Size sourceIndex, in Size targetIndex ) | |
UInt16[] | counts ? () |
Data | data ? () |
Boolean | equalValues ? ( in Size index1, in Size index2 ) |
UInt32 | getElementsMemType ? () |
getLinearCombination ? ( in LocalL16UInt32Array sourceIndices, in LocalL16ScalarArray sourceWeights, out LocalL16UInt32Array skinIds, out LocalL16ScalarArray skinWeights ) | |
getPair ? ( in UInt32 index, in UInt32 offset, out UInt16 id, out Scalar weight ) | |
UInt16 | getPairCount ? ( in UInt32 index ) |
getPairs ? ( in UInt32 index, out LocalL16UInt32Array skinIds, out LocalL16ScalarArray skinWeights ) | |
Size | getScalarValueSize ? () |
UInt16[] | ids ? () |
linearCombine ! ( in Ref<Object> sourceAttr, in LocalL16UInt32Array sourceIndices, in LocalL16ScalarArray sourceWeights, in Size targetIndex ) | |
normalize ! ( in Size index ) | |
UInt32[] | offsets ? () |
resize ! ( in Size count ) | |
setFromScalar ! ( in Size index, in LocalL16ScalarArray value ) | |
setPair ! ( in UInt32 index, in UInt32 offset, in UInt16 id, in Scalar weight ) | |
setPairCount ! ( in UInt32 index, in UInt16 pairCount ) | |
setPairs ! ( in UInt32 index, in LocalL16UInt32Array skinIds, in LocalL16ScalarArray skinWeights ) | |
String | truncateDecimalsPrint ? ( in Size index, in UInt32 n ) |
String | unitTestPrint ? ( in Size index ) |
String | valueDesc ? ( in Size index ) |
UInt32ArrayAllocator | weightsAllocator ? () |