Clustered Mesh

ClusteredMesh
aka CollisionMeshData
Track unit 203's clustered mesh.
Resource names	Unknown
Type ID	0x24
Category	Generic
Memory distribution	Main Memory only
Imports	Unknown
Imported by	ID List
Editor available?	Viewer only Use Blender addons

Clustered Mesh resources were used for world collision (WORLDCOL) prior to the introduction of the Polygon Soup List resource type.

Development Information on versions of the Clustered Mesh resource type used during development.

Structures

CgsPhysics::CollisionMeshData

32-bit

Offset	Length	Type	Name	Description	Comments
0x0	0x4	Volume*	mpCollisionVolume
0x4	0xC			Padding
0x10	0x10	Vector3Plus	mBoundingSphere

rw::collision::VolRef::Volume

32-bit

Offset	Length	Type	Name	Description	Comments
0x0	0x40	Matrix44Affine	transform
0x40	0x4	Volume::VTable*	vTable		When written to file, this actually represents an rw::collision::VolumeType instead rather than a pointer. This is always VOLUMETYPEAGGREGATE, which has a value of 6.
0x44	0x4	ClusteredMesh*			This is actually implemented as a 12-byte union accepting multiple types, including three floats representing triangles or boxes. The union type used is defined by the VolumeType in the vTable. The Clustered Mesh resource type however only accepts a ClusteredMesh pointer here (represented by the aggregate volume type), and will cause undefined behaviour if other types are used instead.
0x48	0x8			Union padding
0x50	0x4	float32_t	radius
0x54	0x4	uint32_t	groupID
0x58	0x4	uint32_t	surfaceID
0x5C	0x4	uint32_t	m_flags		List of flags is unknown

rw::collision::Aggregate

32-bit

Offset	Length	Type	Name	Description	Comments
0x0	0x20	AABBox	m_AABB
0x20	0x4	Aggregate::VTable*	m_vTable		Garbage data in the file - always replaced on read
0x24	0x4	uint32_t	m_numTagBits
0x28	0x4	uint32_t	m_numVolumes
0x2C	0x4	uint32_t	pad

rw::collision::Procedural

32-bit

Offset	Length	Type	Name	Description	Comments
0x0	0x30	Aggregate		Base class

rw::collision::ClusteredMesh

The base pointer resets here - all pointer offsets here are relative to the start of this struct.

32-bit

Offset	Length	Type	Name	Description	Comments
0x0	0x30	Procedural		Base class
0x30	0x4	KDTree*	mKDTree
0x34	0x4	uint32_t*	mCluster		Effectively a ClusteredMeshCluster**
0x38	0x8	ClusterParams	mClusterParams
0x40	0x4	uint32_t	mNumClusters
0x44	0x4	uint32_t	mMaxClusters
0x48	0x4	uint32_t	mNumUnits
0x4C	0x4	uint32_t	mMaxUnits
0x50	0x4	uint32_t	mSizeOfThis
0x54	0x2	uint16_t	mDefaultGroupId
0x56	0x2	uint16_t	mDefaultSurfaceId
0x58	0x1	uint8_t	mDefaultEdgeAngle
0x59	0x7			Padding

rw::collision::AALineClipper::AABBox

Offset	Length	Type	Name	Description	Comments
0x0	0x10	vpu::Vector3	m_min
0x10	0x10	vpu::Vector3	m_max

rw::collision::ClusterParams

Offset	Length	Type	Name	Description	Comments
0x0	0x4	float32_t	mVertexCompressionGranularity		Relevant for cluster compression modes 1 and 2
0x4	0x2	uint16_t	mFlags		List of flags is unknown
0x6	0x1	uint8_t	mGroupIdSize	Number of bytes	1 or 2
0x7	0x1	uint8_t	mSurfaceIdSize	Number of bytes	1 or 2

rw::collision::KDTree

The base pointer resets here - all pointer offsets here are relative to the start of this struct.

32-bit

Offset	Length	Type	Name	Description	Comments
0x0	0x4	BranchNode*	m_branchNodes
0x4	0x4	uint32_t	m_numBranchNodes
0x8	0x4	uint32_t	m_numEntries
0xC	0x4			Padding
0x10	0x20	AABBox	m_bbox

rw::collision::KDTree::BranchNode

Offset	Length	Type	Name	Description	Comments
0x0	0x4	uint32_t	m_parent
0x4	0x4	uint32_t	m_axis
0x8	0x10	NodeRef[2]	m_childRefs
0x18	0x8	float32_t[2]	m_extents

rw::collision::KDTree::NodeRef

Offset	Length	Type	Name	Description	Comments
0x0	0x4	uint32_t	m_content
0x4	0x4	uint32_t	m_index

rw::collision::ClusteredMeshCluster

Offset	Length	Type	Name	Description	Comments
0x0	0x2	uint16_t	unitCount
0x2	0x2	uint16_t	unitDataSize
0x4	0x2	uint16_t	unitDataStart	The index of the vertex array where unit data starts
0x6	0x2	uint16_t	normalStart	The index of the vertex array where normal data starts
0x8	0x2	uint16_t	totalSize
0xA	0x1	uint8_t	vertexCount
0xB	0x1	uint8_t	normalCount
0xC	0x1	uint8_t	compressionMode		0, 1 or 2
0xD	0x3	uint8_t[3]	Padding
0x10	0x10	vpu::Vector3[1]	vertexArray	Variable size data	Hmmm... To do: Normals? Vertex data Unit data

Vertex data

Compression mode 0 (uncompressed)

Offset	Length	Type	Name	Description	Comments
0x0 (repeating)	0x10	vpu::Vector3	?		Raw vector (mVertexCompressionGranularity unused)

Compression mode 1

Offset	Length	Type	Name	Description	Comments
0x0	0xC	int32_t[3]	?	Component base
0xC (repeating)	0x6	uint16_t[3]	?		Add to initial int32_t components and then multiply components by mVertexCompressionGranularity to get coordinate

Compression mode 2

Offset	Length	Type	Name	Description	Comments
0x0 (repeating)	0xC	int32_t[3]	?		Multiply components by mVertexCompressionGranularity to get coordinate

Unit data

This is a byte stream. Exact offsets are variable and depends on several factors, as detailed below.

Length	Type	Description	Comments
0x1	uint8_t	Lower nibble (type) 1 - Triangles 2 - Quads 3 - Variable? Upper nibble Flag 1 - ? Flag 2 - Has edge cosine? data Flag 4 - Has a group ID Flag 8 - Has a surface ID
0x1	uint8_t	Only present if type == 3 The number of points minus two
Point count	uint8_t	Vertex index	3 points for type 1, 4 points for type 2, and specified by the previous byte plus 2 for type 3
Point count	uint8_t	Only present with flag 2 Edge cosines? Mask with 0x1F and flags above that?
Group ID size (see ClusterParams)	uint8_t or uint16_t	Only present with flag 4 Group ID	This is always little endian.
Surface ID size (see ClusterParams)	uint8_t or uint16_t	Only present with flag 8 Surface ID	This is always little endian.