OpenSubdiv/opensubdiv/hbr/mesh.h

1032 lines
32 KiB
C++

//
// Copyright (C) Pixar. All rights reserved.
//
// This license governs use of the accompanying software. If you
// use the software, you accept this license. If you do not accept
// the license, do not use the software.
//
// 1. Definitions
// The terms "reproduce," "reproduction," "derivative works," and
// "distribution" have the same meaning here as under U.S.
// copyright law. A "contribution" is the original software, or
// any additions or changes to the software.
// A "contributor" is any person or entity that distributes its
// contribution under this license.
// "Licensed patents" are a contributor's patent claims that read
// directly on its contribution.
//
// 2. Grant of Rights
// (A) Copyright Grant- Subject to the terms of this license,
// including the license conditions and limitations in section 3,
// each contributor grants you a non-exclusive, worldwide,
// royalty-free copyright license to reproduce its contribution,
// prepare derivative works of its contribution, and distribute
// its contribution or any derivative works that you create.
// (B) Patent Grant- Subject to the terms of this license,
// including the license conditions and limitations in section 3,
// each contributor grants you a non-exclusive, worldwide,
// royalty-free license under its licensed patents to make, have
// made, use, sell, offer for sale, import, and/or otherwise
// dispose of its contribution in the software or derivative works
// of the contribution in the software.
//
// 3. Conditions and Limitations
// (A) No Trademark License- This license does not grant you
// rights to use any contributor's name, logo, or trademarks.
// (B) If you bring a patent claim against any contributor over
// patents that you claim are infringed by the software, your
// patent license from such contributor to the software ends
// automatically.
// (C) If you distribute any portion of the software, you must
// retain all copyright, patent, trademark, and attribution
// notices that are present in the software.
// (D) If you distribute any portion of the software in source
// code form, you may do so only under this license by including a
// complete copy of this license with your distribution. If you
// distribute any portion of the software in compiled or object
// code form, you may only do so under a license that complies
// with this license.
// (E) The software is licensed "as-is." You bear the risk of
// using it. The contributors give no express warranties,
// guarantees or conditions. You may have additional consumer
// rights under your local laws which this license cannot change.
// To the extent permitted under your local laws, the contributors
// exclude the implied warranties of merchantability, fitness for
// a particular purpose and non-infringement.
//
#ifndef HBRMESH_H
#define HBRMESH_H
#ifdef PRMAN
#include "libtarget/TgMalloc.h" // only for alloca
#include "libtarget/TgThread.h"
#ifdef HBRSTITCH
#include "libtarget/TgHashMap.h"
#endif
#endif
#include <algorithm>
#include <cstring>
#include <iterator>
#include <vector>
#include <set>
#include <iostream>
#include "../hbr/vertex.h"
#include "../hbr/face.h"
#include "../hbr/hierarchicalEdit.h"
#include "../hbr/vertexEdit.h"
#include "../hbr/creaseEdit.h"
#include "../hbr/allocator.h"
#include "../version.h"
namespace OpenSubdiv {
namespace OPENSUBDIV_VERSION {
template <class T> class HbrSubdivision;
template <class T> class HbrHalfedge;
template <class T> class HbrMesh {
public:
HbrMesh(HbrSubdivision<T>* subdivision = 0, int fvarcount = 0, const int *fvarindices = 0, const int *fvarwidths = 0, int totalfvarwidth = 0
#ifdef HBRSTITCH
, int stitchCount = 0
#endif
);
~HbrMesh();
// Create vertex with the indicated ID and data
HbrVertex<T>* NewVertex(int id, const T &data);
// Create vertex with the indicated data. The ID will be assigned
// by the mesh.
HbrVertex<T>* NewVertex(const T &data);
// Create vertex without an ID - one will be assigned by the mesh,
// and the data implicitly created will share the same id
HbrVertex<T>* NewVertex();
// Ask for vertex with the indicated ID
HbrVertex<T>* GetVertex(int id) const {
if (id >= nvertices) {
return 0;
} else {
return vertices[id];
}
}
// Ask for client data associated with the vertex with the indicated ID
void* GetVertexClientData(int id) const {
if (id >= vertexClientData.size()) {
return 0;
} else {
return vertexClientData[id];
}
}
// Set client data associated with the vertex with the indicated ID
void SetVertexClientData(int id, void *data) {
if (id >= vertexClientData.size()) {
size_t oldsize = vertexClientData.size();
vertexClientData.resize(nvertices);
if (s_memStatsIncrement) {
s_memStatsIncrement((vertexClientData.size() - oldsize) * sizeof(void*));
}
}
vertexClientData[id] = data;
}
// Create face from a list of vertex IDs
HbrFace<T>* NewFace(int nvertices, const int *vtx, int uindex);
// Create face from a list of vertices
HbrFace<T>* NewFace(int nvertices, HbrVertex<T>** vtx, HbrFace<T>* parent, int childindex);
// "Create" a new uniform index
int NewUniformIndex() { return ++maxUniformIndex; }
// Finishes initialization of the mesh
void Finish();
// Remove the indicated face from the mesh
void DeleteFace(HbrFace<T>* face);
// Remove the indicated vertex from the mesh
void DeleteVertex(HbrVertex<T>* vertex);
// Returns number of vertices in the mesh
int GetNumVertices() const;
// Returns number of disconnected vertices in the mesh
int GetNumDisconnectedVertices() const;
// Returns number of faces in the mesh
int GetNumFaces() const;
// Returns number of coarse faces in the mesh
int GetNumCoarseFaces() const;
// Ask for face with the indicated ID
HbrFace<T>* GetFace(int id) const;
// Ask for client data associated with the face with the indicated ID
void* GetFaceClientData(int id) const {
if (id >= faceClientData.size()) {
return 0;
} else {
return faceClientData[id];
}
}
// Set client data associated with the face with the indicated ID
void SetFaceClientData(int id, void *data) {
if (id >= faceClientData.size()) {
size_t oldsize = faceClientData.size();
faceClientData.resize(nfaces);
if (s_memStatsIncrement) {
s_memStatsIncrement((faceClientData.size() - oldsize) * sizeof(void*));
}
}
faceClientData[id] = data;
}
// Returns a collection of all vertices in the mesh. This function
// requires an output iterator; to get the vertices into a
// std::vector, use GetVertices(std::back_inserter(myvector))
template <typename OutputIterator>
void GetVertices(OutputIterator vertices) const;
// Applies operator to all vertices
void ApplyOperatorAllVertices(HbrVertexOperator<T> &op) const;
// Returns a collection of all faces in the mesh. This function
// requires an output iterator; to get the faces into a
// std::vector, use GetFaces(std::back_inserter(myvector))
template <typename OutputIterator>
void GetFaces(OutputIterator faces) const;
// Returns the subdivision method
HbrSubdivision<T>* GetSubdivision() const { return subdivision; }
// Return the number of facevarying variables
int GetFVarCount() const { return fvarcount; }
// Return a table of the start index of each facevarying variable
const int *GetFVarIndices() const { return fvarindices; }
// Return a table of the size of each facevarying variable
const int *GetFVarWidths() const { return fvarwidths; }
// Return the sum size of facevarying variables per vertex
int GetTotalFVarWidth() const { return totalfvarwidth; }
#ifdef HBRSTITCH
int GetStitchCount() const { return stitchCount; }
#endif
void PrintStats(std::ostream& out);
// Returns memory statistics
size_t GetMemStats() const { return m_memory; }
// Interpolate boundary management
enum InterpolateBoundaryMethod {
k_InterpolateBoundaryNone,
k_InterpolateBoundaryEdgeOnly,
k_InterpolateBoundaryEdgeAndCorner,
k_InterpolateBoundaryAlwaysSharp
};
InterpolateBoundaryMethod GetInterpolateBoundaryMethod() const { return interpboundarymethod; }
void SetInterpolateBoundaryMethod(InterpolateBoundaryMethod method) { interpboundarymethod = method; }
InterpolateBoundaryMethod GetFVarInterpolateBoundaryMethod() const { return fvarinterpboundarymethod; }
void SetFVarInterpolateBoundaryMethod(InterpolateBoundaryMethod method) { fvarinterpboundarymethod = method; }
bool GetFVarPropagateCorners() const { return fvarpropagatecorners; }
void SetFVarPropagateCorners(bool p) { fvarpropagatecorners = p; }
// Register routines for keeping track of memory usage
void RegisterMemoryRoutines(void (*increment)(unsigned long bytes), void (*decrement)(unsigned long bytes)) {
m_faceAllocator.SetMemStatsIncrement(increment);
m_faceAllocator.SetMemStatsDecrement(decrement);
m_vertexAllocator.SetMemStatsIncrement(increment);
m_vertexAllocator.SetMemStatsDecrement(decrement);
s_memStatsIncrement = increment;
s_memStatsDecrement = decrement;
}
// Add a vertex to consider for garbage collection. All
// neighboring faces of that vertex will be examined to see if
// they can be deleted
void AddGarbageCollectableVertex(HbrVertex<T>* vertex) {
if (!m_transientMode) {
assert(vertex);
if (!vertex->IsCollected()) {
gcVertices.push_back(vertex); vertex->SetCollected();
}
}
}
// Apply garbage collection to the mesh
void GarbageCollect();
// Add a new hierarchical edit to the mesh
void AddHierarchicalEdit(HbrHierarchicalEdit<T>* edit);
// Return the hierarchical edits associated with the mesh
const std::vector<HbrHierarchicalEdit<T>*> &GetHierarchicalEdits() const {
return hierarchicalEdits;
}
// Return the hierarchical edits associated with the mesh at an
// offset
HbrHierarchicalEdit<T>** GetHierarchicalEditsAtOffset(int offset) {
return &hierarchicalEdits[offset];
}
// Whether the mesh has certain types of edits
bool HasVertexEdits() const { return hasVertexEdits; }
bool HasCreaseEdits() const { return hasCreaseEdits; }
void Unrefine(int numCoarseVerts, int numCoarseFaces) {
static int oldMaxFaceID = 0;
if(oldMaxFaceID == 0) {
oldMaxFaceID = numCoarseFaces;
}
for (int i = numCoarseFaces; i < maxFaceID; ++i) {
HbrFace<T>* f = GetFace(i);
if(f and not f->IsCoarse())
DeleteFace(f);
}
//oldMaxFaceID = maxFaceID;
maxFaceID = numCoarseFaces;
for(int i=numCoarseVerts; (int)vertices.size(); ++i ) {
HbrVertex<T>* v = GetVertex(i);
if(v and not v->IsReferenced())
DeleteVertex(v);
}
}
// When mode is true, the mesh is put in a "transient" mode,
// i.e. all subsequent intermediate vertices/faces that are
// created by subdivision are deemed temporary. This transient
// data can be entirely freed by a subsequent call to
// FreeTransientData(). Essentially, the mesh is checkpointed and
// restored. This is useful when space is at a premium and
// subdivided results are cached elsewhere. On the other hand,
// repeatedly putting the mesh in and out of transient mode and
// performing the same evaluations comes at a significant compute
// cost.
void SetTransientMode(bool mode) {
m_transientMode = mode;
}
// Frees transient subdivision data; returns the mesh to a
// checkpointed state prior to a call to SetTransientMode.
void FreeTransientData();
// Create new face children block for use by HbrFace
HbrFaceChildren<T>* NewFaceChildren() {
return m_faceChildrenAllocator.Allocate();
}
// Recycle face children block used by HbrFace
void DeleteFaceChildren(HbrFaceChildren<T>* facechildren) {
m_faceChildrenAllocator.Deallocate(facechildren);
}
#ifdef HBRSTITCH
void * GetStitchData(const HbrHalfedge<T>* edge) const {
typename TgHashMap<const HbrHalfedge<T>*, void *>::const_iterator i =
stitchData.find(edge);
if (i != stitchData.end()) {
return i->second;
} else {
return NULL;
}
}
void SetStitchData(const HbrHalfedge<T>* edge, void *data) {
stitchData[edge] = data;
}
#endif
private:
// Subdivision method used in this mesh
HbrSubdivision<T>* subdivision;
// Number of facevarying datums
int fvarcount;
// Start indices of the facevarying data we want to store
const int *fvarindices;
// Individual widths of the facevarying data we want to store
const int *fvarwidths;
// Total widths of the facevarying data
const int totalfvarwidth;
#ifdef HBRSTITCH
// Number of stitch edges per halfedge
const int stitchCount;
// Client (sparse) data used on some halfedges
TgHashMap<const HbrHalfedge<T>*, void *> stitchData;
#endif
// Vertices which comprise this mesh
std::vector<HbrVertex<T> *> vertices;
int nvertices;
// Client data associated with each face
std::vector<void *> vertexClientData;
// Faces which comprise this mesh
std::vector<HbrFace<T> *> faces;
int nfaces;
// Client data associated with each face
std::vector<void *> faceClientData;
// Maximum vertex ID - may be needed when generating a unique
// vertex ID
int maxVertexID;
// Maximum face ID - needed when generating a unique face ID
int maxFaceID;
// Maximum uniform index - needed to generate a new uniform index
int maxUniformIndex;
// Boundary interpolation method
InterpolateBoundaryMethod interpboundarymethod;
// Facevarying boundary interpolation method
InterpolateBoundaryMethod fvarinterpboundarymethod;
// Whether facevarying corners propagate their sharpness
bool fvarpropagatecorners;
// Memory statistics tracking routines
HbrMemStatFunction s_memStatsIncrement;
HbrMemStatFunction s_memStatsDecrement;
// Vertices which may be garbage collected
std::vector<HbrVertex<T>*> gcVertices;
// List of vertex IDs which may be recycled
std::set<int> recycleIDs;
// Hierarchical edits. This vector is left unsorted until Finish()
// is called, at which point it is sorted. After that point,
// HbrFaces have pointers directly into this array so manipulation
// of it should be avoided.
std::vector<HbrHierarchicalEdit<T>*> hierarchicalEdits;
// Size of faces (including 4 facevarying bits and stitch edges)
const size_t m_faceSize;
HbrAllocator<HbrFace<T> > m_faceAllocator;
// Size of vertices (includes storage for one piece of facevarying data)
const size_t m_vertexSize;
HbrAllocator<HbrVertex<T> > m_vertexAllocator;
// Allocator for face children blocks used by HbrFace
HbrAllocator<HbrFaceChildren<T> > m_faceChildrenAllocator;
// Memory used by this mesh alone, plus all its faces and vertices
size_t m_memory;
// Number of coarse faces. Initialized at Finish()
int m_numCoarseFaces;
// Flags which indicate whether the mesh has certain types of
// edits
unsigned hasVertexEdits:1;
unsigned hasCreaseEdits:1;
// True if the mesh is in "transient" mode, meaning all
// vertices/faces that are created via NewVertex/NewFace should be
// deemed temporary
bool m_transientMode;
// Vertices which are transient
std::vector<HbrVertex<T>*> m_transientVertices;
// Faces which are transient
std::vector<HbrFace<T>*> m_transientFaces;
};
} // end namespace OPENSUBDIV_VERSION
using namespace OPENSUBDIV_VERSION;
} // end namespace OpenSubdiv
#include <algorithm>
#include "../hbr/mesh.h"
#include "../hbr/halfedge.h"
namespace OpenSubdiv {
namespace OPENSUBDIV_VERSION {
template <class T>
HbrMesh<T>::HbrMesh(HbrSubdivision<T>* s, int _fvarcount, const int *_fvarindices, const int *_fvarwidths, int _totalfvarwidth
#ifdef HBRSTITCH
, int _stitchCount
#endif
)
: subdivision(s), fvarcount(_fvarcount), fvarindices(_fvarindices),
fvarwidths(_fvarwidths), totalfvarwidth(_totalfvarwidth),
#ifdef HBRSTITCH
stitchCount(_stitchCount),
#endif
nvertices(0), nfaces(0), maxVertexID(0), maxFaceID(0), maxUniformIndex(0),
interpboundarymethod(k_InterpolateBoundaryNone),
fvarinterpboundarymethod(k_InterpolateBoundaryNone),
fvarpropagatecorners(false),
s_memStatsIncrement(0), s_memStatsDecrement(0),
m_faceSize(sizeof(HbrFace<T>) + 4 *
((fvarcount + 15) / 16 * sizeof(unsigned int)
#ifdef HBRSTITCH
+ stitchCount * sizeof(StitchEdge*)
#endif
)),
m_faceAllocator(&m_memory, 512, 0, 0, m_faceSize),
m_vertexSize(sizeof(HbrVertex<T>) +
(totalfvarwidth ? (sizeof(HbrFVarData<T>) + (totalfvarwidth - 1) * sizeof(float)) : 0)),
m_vertexAllocator(&m_memory, 512, 0, 0, m_vertexSize),
m_faceChildrenAllocator(&m_memory, 512, 0, 0),
m_memory(0),
m_numCoarseFaces(-1),
hasVertexEdits(0),
hasCreaseEdits(0),
m_transientMode(false) {
}
template <class T>
HbrMesh<T>::~HbrMesh() {
GarbageCollect();
int i;
if (!faces.empty()) {
for (i = 0; i < nfaces; ++i) {
if (faces[i]) {
faces[i]->Destroy();
m_faceAllocator.Deallocate(faces[i]);
}
}
if (s_memStatsDecrement) {
s_memStatsDecrement(faces.size() * sizeof(HbrFace<T>*));
}
}
if (!vertices.empty()) {
for (i = 0; i < nvertices; ++i) {
if (vertices[i]) {
vertices[i]->Destroy(this);
m_vertexAllocator.Deallocate(vertices[i]);
}
}
if (s_memStatsDecrement) {
s_memStatsDecrement(vertices.size() * sizeof(HbrVertex<T>*));
}
}
if (!vertexClientData.empty() && s_memStatsDecrement) {
s_memStatsDecrement(vertexClientData.size() * sizeof(void*));
}
if (!faceClientData.empty() && s_memStatsDecrement) {
s_memStatsDecrement(faceClientData.size() * sizeof(void*));
}
for (typename std::vector<HbrHierarchicalEdit<T>* >::iterator hi =
hierarchicalEdits.begin(); hi != hierarchicalEdits.end(); ++hi) {
delete *hi;
}
}
template <class T>
HbrVertex<T>*
HbrMesh<T>::NewVertex(int id, const T &data) {
HbrVertex<T>* v = 0;
if (nvertices <= id) {
while (nvertices <= maxVertexID) {
nvertices *= 2;
if (nvertices < 1) nvertices = 1;
}
size_t oldsize = vertices.size();
vertices.resize(nvertices);
if (s_memStatsIncrement) {
s_memStatsIncrement((vertices.size() - oldsize) * sizeof(HbrVertex<T>*));
}
}
v = vertices[id];
if (v) {
v->Destroy(this);
} else {
v = m_vertexAllocator.Allocate();
}
v->Initialize(id, data, GetTotalFVarWidth());
vertices[id] = v;
if (id >= maxVertexID) {
maxVertexID = id + 1;
}
// Newly created vertices are always candidates for garbage
// collection, until they get "owned" by someone who
// IncrementsUsage on the vertex.
AddGarbageCollectableVertex(v);
// If mesh is in transient mode, add vertex to transient list
if (m_transientMode) {
m_transientVertices.push_back(v);
}
return v;
}
template <class T>
HbrVertex<T>*
HbrMesh<T>::NewVertex(const T &data) {
// Pick an ID - either the maximum vertex ID or a recycled ID if
// we can
int id = maxVertexID;
if (!recycleIDs.empty()) {
id = *recycleIDs.begin();
recycleIDs.erase(recycleIDs.begin());
}
if (id >= maxVertexID) {
maxVertexID = id + 1;
}
return NewVertex(id, data);
}
template <class T>
HbrVertex<T>*
HbrMesh<T>::NewVertex() {
// Pick an ID - either the maximum vertex ID or a recycled ID if
// we can
int id = maxVertexID;
if (!recycleIDs.empty()) {
id = *recycleIDs.begin();
recycleIDs.erase(recycleIDs.begin());
}
if (id >= maxVertexID) {
maxVertexID = id + 1;
}
T data(id);
data.Clear();
return NewVertex(id, data);
}
template <class T>
HbrFace<T>*
HbrMesh<T>::NewFace(int nv, const int *vtx, int uindex) {
HbrVertex<T>** facevertices = reinterpret_cast<HbrVertex<T>**>(alloca(sizeof(HbrVertex<T>*) * nv));
int i;
for (i = 0; i < nv; ++i) {
facevertices[i] = GetVertex(vtx[i]);
if (!facevertices[i]) {
return 0;
}
}
HbrFace<T> *f = 0;
// Resize if needed
if (nfaces <= maxFaceID) {
while (nfaces <= maxFaceID) {
nfaces *= 2;
if (nfaces < 1) nfaces = 1;
}
size_t oldsize = faces.size();
faces.resize(nfaces);
if (s_memStatsIncrement) {
s_memStatsIncrement((faces.size() - oldsize) * sizeof(HbrVertex<T>*));
}
}
f = faces[maxFaceID];
if (f) {
f->Destroy();
} else {
f = m_faceAllocator.Allocate();
}
f->Initialize(this, NULL, -1, maxFaceID, uindex, nv, facevertices, totalfvarwidth, 0);
faces[maxFaceID] = f;
maxFaceID++;
// Update the maximum encountered uniform index
if (uindex > maxUniformIndex) maxUniformIndex = uindex;
// If mesh is in transient mode, add face to transient list
if (m_transientMode) {
m_transientFaces.push_back(f);
}
return f;
}
template <class T>
HbrFace<T>*
HbrMesh<T>::NewFace(int nv, HbrVertex<T> **vtx, HbrFace<T>* parent, int childindex) {
HbrFace<T> *f = 0;
// Resize if needed
if (nfaces <= maxFaceID) {
while (nfaces <= maxFaceID) {
nfaces *= 2;
if (nfaces < 1) nfaces = 1;
}
size_t oldsize = faces.size();
faces.resize(nfaces);
if (s_memStatsIncrement) {
s_memStatsIncrement((faces.size() - oldsize) * sizeof(HbrVertex<T>*));
}
}
f = faces[maxFaceID];
if (f) {
f->Destroy();
} else {
f = m_faceAllocator.Allocate();
}
f->Initialize(this, parent, childindex, maxFaceID, parent ? parent->GetUniformIndex() : 0, nv, vtx, totalfvarwidth, parent ? parent->GetDepth() + 1 : 0);
if (parent) {
f->SetPtexIndex(parent->GetPtexIndex());
}
faces[maxFaceID] = f;
maxFaceID++;
// If mesh is in transient mode, add face to transient list
if (m_transientMode) {
m_transientFaces.push_back(f);
}
return f;
}
template <class T>
void
HbrMesh<T>::Finish() {
int i, j;
m_numCoarseFaces = 0;
for (i = 0; i < nfaces; ++i) {
if (faces[i]) {
faces[i]->SetCoarse();
m_numCoarseFaces++;
}
}
std::vector<HbrVertex<T>*> vertexlist;
GetVertices(std::back_inserter(vertexlist));
for (typename std::vector<HbrVertex<T>*>::iterator vi = vertexlist.begin();
vi != vertexlist.end(); ++vi) {
HbrVertex<T>* vertex = *vi;
if (vertex->IsConnected()) vertex->Finish();
}
// Finish may have added new vertices
vertexlist.clear();
GetVertices(std::back_inserter(vertexlist));
// If interpolateboundary is on, process boundary edges
if (interpboundarymethod == k_InterpolateBoundaryEdgeOnly || interpboundarymethod == k_InterpolateBoundaryEdgeAndCorner) {
for (i = 0; i < nfaces; ++i) {
if (HbrFace<T>* face = faces[i]) {
int nv = face->GetNumVertices();
for (int k = 0; k < nv; ++k) {
HbrHalfedge<T>* edge = face->GetEdge(k);
if (edge->IsBoundary()) {
edge->SetSharpness(HbrHalfedge<T>::k_InfinitelySharp);
}
}
}
}
}
// Process corners
if (interpboundarymethod == k_InterpolateBoundaryEdgeAndCorner) {
for (typename std::vector<HbrVertex<T>*>::iterator vi = vertexlist.begin();
vi != vertexlist.end(); ++vi) {
HbrVertex<T>* vertex = *vi;
if (vertex && vertex->IsConnected() && vertex->OnBoundary() && vertex->GetCoarseValence() == 2) {
vertex->SetSharpness(HbrVertex<T>::k_InfinitelySharp);
}
}
}
// Sort the hierarchical edits
if (!hierarchicalEdits.empty()) {
HbrHierarchicalEditComparator<T> cmp;
int nHierarchicalEdits = (int)hierarchicalEdits.size();
std::sort(hierarchicalEdits.begin(), hierarchicalEdits.end(), cmp);
// Push a sentinel null value - we rely upon this sentinel to
// ensure face->GetHierarchicalEdits knows when to terminate
hierarchicalEdits.push_back(0);
j = 0;
// Link faces to hierarchical edits
for (i = 0; i < nfaces; ++i) {
if (faces[i]) {
while (j < nHierarchicalEdits && hierarchicalEdits[j]->GetFaceID() < i) {
++j;
}
if (j < nHierarchicalEdits && hierarchicalEdits[j]->GetFaceID() == i) {
faces[i]->SetHierarchicalEdits(&hierarchicalEdits[j]);
}
}
}
}
}
template <class T>
void
HbrMesh<T>::DeleteFace(HbrFace<T>* face) {
if (face->GetID() < nfaces) {
HbrFace<T>* f = faces[face->GetID()];
if (f == face) {
faces[face->GetID()] = 0;
face->Destroy();
m_faceAllocator.Deallocate(face);
}
}
}
template <class T>
void
HbrMesh<T>::DeleteVertex(HbrVertex<T>* vertex) {
HbrVertex<T> *v = GetVertex(vertex->GetID());
if (v == vertex) {
recycleIDs.insert(vertex->GetID());
int id = vertex->GetID();
vertices[id] = 0;
vertex->Destroy(this);
m_vertexAllocator.Deallocate(vertex);
}
}
template <class T>
int
HbrMesh<T>::GetNumVertices() const {
int count = 0;
for (int i = 0; i < nvertices; ++i) {
if (vertices[i]) count++;
}
return count;
}
template <class T>
int
HbrMesh<T>::GetNumDisconnectedVertices() const {
int disconnected = 0;
for (int i = 0; i < nvertices; ++i) {
if (HbrVertex<T>* v = vertices[i]) {
if (!v->IsConnected()) {
disconnected++;
}
}
}
return disconnected;
}
template <class T>
int
HbrMesh<T>::GetNumFaces() const {
int count = 0;
for (int i = 0; i < nfaces; ++i) {
if (faces[i]) count++;
}
return count;
}
template <class T>
int
HbrMesh<T>::GetNumCoarseFaces() const {
// Use the value computed by Finish() if it exists
if (m_numCoarseFaces >= 0) return m_numCoarseFaces;
// Otherwise we have to just count it up now
int count = 0;
for (int i = 0; i < nfaces; ++i) {
if (faces[i] && faces[i]->IsCoarse()) count++;
}
return count;
}
template <class T>
HbrFace<T>*
HbrMesh<T>::GetFace(int id) const {
if (id < nfaces) {
return faces[id];
}
return 0;
}
template <class T>
template <typename OutputIterator>
void
HbrMesh<T>::GetVertices(OutputIterator lvertices) const {
for (int i = 0; i < nvertices; ++i) {
if (vertices[i]) *lvertices++ = vertices[i];
}
}
template <class T>
void
HbrMesh<T>::ApplyOperatorAllVertices(HbrVertexOperator<T> &op) const {
for (int i = 0; i < nvertices; ++i) {
if (vertices[i]) op(*vertices[i]);
}
}
template <class T>
template <typename OutputIterator>
void
HbrMesh<T>::GetFaces(OutputIterator lfaces) const {
for (int i = 0; i < nfaces; ++i) {
if (faces[i]) *lfaces++ = faces[i];
}
}
template <class T>
void
HbrMesh<T>::PrintStats(std::ostream &out) {
int singular = 0;
int sumvalence = 0;
int i, nv = 0;
int disconnected = 0;
int extraordinary = 0;
for (i = 0; i < nvertices; ++i) {
if (HbrVertex<T>* v = vertices[i]) {
nv++;
if (v->IsSingular()) {
out << " singular: " << *v << "\n";
singular++;
} else if (!v->IsConnected()) {
out << " disconnected: " << *v << "\n";
disconnected++;
} else {
if (v->IsExtraordinary()) {
extraordinary++;
}
sumvalence += v->GetValence();
}
}
}
out << "Mesh has " << nv << " vertices\n";
out << "Total singular vertices " << singular << "\n";
out << "Total disconnected vertices " << disconnected << "\n";
out << "Total extraordinary vertices " << extraordinary << "\n";
out << "Average valence " << (float) sumvalence / nv << "\n";
int sumsides = 0;
int numfaces = 0;
for (i = 0; i < nfaces; ++i) {
if (HbrFace<T>* f = faces[i]) {
numfaces++;
sumsides += f->GetNumVertices();
}
}
out << "Mesh has " << nfaces << " faces\n";
out << "Average sidedness " << (float) sumsides / nfaces << "\n";
}
template <class T>
void
HbrMesh<T>::GarbageCollect() {
if (gcVertices.empty()) return;
static const size_t gcthreshold = 4096;
if (gcVertices.size() <= gcthreshold) return;
// Go through the list of garbage collectable vertices and gather
// up the neighboring faces of those vertices which can be garbage
// collected.
std::vector<HbrFace<T>*> killlist;
std::vector<HbrVertex<T>*> vlist;
// Process the vertices in the same order as they were collected
// (gcVertices used to be declared as a std::deque, but that was
// causing unnecessary heap traffic).
int numprocessed = (int)gcVertices.size() - gcthreshold / 2;
for (int i = 0; i < numprocessed; ++i) {
HbrVertex<T>* v = gcVertices[i];
v->ClearCollected();
if (v->IsUsed()) continue;
vlist.push_back(v);
HbrHalfedge<T>* start = v->GetIncidentEdge(), *edge;
edge = start;
while (edge) {
HbrFace<T>* f = edge->GetLeftFace();
if (!f->IsCollected()) {
f->SetCollected();
killlist.push_back(f);
}
edge = v->GetNextEdge(edge);
if (edge == start) break;
}
}
gcVertices.erase(gcVertices.begin(), gcVertices.begin() + numprocessed);
// Delete those faces
for (typename std::vector<HbrFace<T>*>::iterator fi = killlist.begin(); fi != killlist.end(); ++fi) {
if ((*fi)->GarbageCollectable()) {
DeleteFace(*fi);
} else {
(*fi)->ClearCollected();
}
}
// Delete as many vertices as we can
for (typename std::vector<HbrVertex<T>*>::iterator vi = vlist.begin(); vi != vlist.end(); ++vi) {
HbrVertex<T>* v = *vi;
if (!v->IsReferenced()) {
DeleteVertex(v);
}
}
}
template <class T>
void
HbrMesh<T>::AddHierarchicalEdit(HbrHierarchicalEdit<T>* edit) {
hierarchicalEdits.push_back(edit);
if (dynamic_cast<HbrVertexEdit<T>*>(edit) ||
dynamic_cast<HbrMovingVertexEdit<T>*>(edit)) {
hasVertexEdits = 1;
} else if (dynamic_cast<HbrCreaseEdit<T>*>(edit)) {
hasCreaseEdits = 1;
}
}
template <class T>
void
HbrMesh<T>::FreeTransientData() {
// When purging transient data, we must clear the faces first
for (typename std::vector<HbrFace<T>*>::iterator fi = m_transientFaces.begin();
fi != m_transientFaces.end(); ++fi) {
DeleteFace(*fi);
}
// The vertices should now be trivial to purge after the transient
// faces have been cleared
for (typename std::vector<HbrVertex<T>*>::iterator vi = m_transientVertices.begin();
vi != m_transientVertices.end(); ++vi) {
DeleteVertex(*vi);
}
m_transientVertices.clear();
m_transientFaces.clear();
// Reset max face ID
int i;
for (i = nfaces - 1; i >= 0; --i) {
if (faces[i]) {
maxFaceID = i + 1;
break;
}
}
// Reset max vertex ID
for (i = nvertices - 1; i >= 0; --i) {
if (vertices[i]) {
maxVertexID = i + 1;
break;
}
}
}
} // end namespace OPENSUBDIV_VERSION
using namespace OPENSUBDIV_VERSION;
} // end namespace OpenSubdiv
#endif /* HBRMESH_H */