metaforce/DataSpec/DNACommon/AROTBuilder.cpp

395 lines
12 KiB
C++

#include "AROTBuilder.hpp"
#include "hecl/Blender/Connection.hpp"
namespace DataSpec
{
logvisor::Module Log("AROTBuilder");
#define AROT_MAX_LEVEL 6
#define COLLISION_MIN_NODE_TRIANGLES 16
static zeus::CAABox SplitAABB(const zeus::CAABox& aabb, int i)
{
zeus::CAABox pos, neg;
aabb.splitZ(neg, pos);
if (i & 4)
{
zeus::CAABox(pos).splitY(neg, pos);
if (i & 2)
{
zeus::CAABox(pos).splitX(neg, pos);
if (i & 1)
return pos;
else
return neg;
}
else
{
zeus::CAABox(neg).splitX(neg, pos);
if (i & 1)
return pos;
else
return neg;
}
}
else
{
zeus::CAABox(neg).splitY(neg, pos);
if (i & 2)
{
zeus::CAABox(pos).splitX(neg, pos);
if (i & 1)
return pos;
else
return neg;
}
else
{
zeus::CAABox(neg).splitX(neg, pos);
if (i & 1)
return pos;
else
return neg;
}
}
}
void AROTBuilder::Node::addChild(int level, int minChildren, const std::vector<zeus::CAABox>& triBoxes,
const zeus::CAABox& curAABB, BspNodeType& typeOut)
{
/* Gather intersecting faces */
for (int i=0 ; i<triBoxes.size() ; ++i)
if (triBoxes[i].intersects(curAABB))
childIndices.insert(i);
/* Return early if empty, triangle intersection below performance threshold, or at max level */
if (childIndices.empty())
{
typeOut = BspNodeType::Invalid;
return;
}
else if (childIndices.size() < minChildren || level == AROT_MAX_LEVEL)
{
typeOut = BspNodeType::Leaf;
return;
}
/* Subdivide */
typeOut = BspNodeType::Branch;
childNodes.resize(8);
for (int i=0 ; i<8 ; ++i)
{
BspNodeType chType;
childNodes[i].addChild(level + 1, minChildren, triBoxes, SplitAABB(curAABB, i), chType);
flags |= int(chType) << (i * 2);
}
/* Unsubdivide */
compSubdivs = 0;
if (childNodes[0].childIndices != childNodes[1].childIndices ||
childNodes[4].childIndices != childNodes[5].childIndices ||
childNodes[2].childIndices != childNodes[3].childIndices ||
childNodes[6].childIndices != childNodes[7].childIndices)
compSubdivs |= 0x4;
if (childNodes[0].childIndices != childNodes[2].childIndices ||
childNodes[1].childIndices != childNodes[3].childIndices ||
childNodes[4].childIndices != childNodes[6].childIndices ||
childNodes[5].childIndices != childNodes[7].childIndices)
compSubdivs |= 0x2;
if (childNodes[0].childIndices != childNodes[4].childIndices ||
childNodes[1].childIndices != childNodes[5].childIndices ||
childNodes[2].childIndices != childNodes[6].childIndices ||
childNodes[3].childIndices != childNodes[7].childIndices)
compSubdivs |= 0x1;
if (!compSubdivs)
{
typeOut = BspNodeType::Leaf;
childNodes = std::vector<Node>();
flags = 0;
}
}
size_t AROTBuilder::BitmapPool::addIndices(const std::set<int>& indices)
{
for (size_t i=0 ; i<m_pool.size() ; ++i)
if (m_pool[i] == indices)
return i;
m_pool.push_back(indices);
return m_pool.size() - 1;
}
static const uint32_t AROTChildCounts[] = { 0, 2, 2, 4, 2, 4, 4, 8 };
void AROTBuilder::Node::nodeCount(size_t& sz, size_t& idxRefs, BitmapPool& bmpPool, size_t& curOff)
{
sz += 1;
poolIdx = bmpPool.addIndices(childIndices);
if (poolIdx > 65535)
Log.report(logvisor::Fatal, "AROT bitmap exceeds 16-bit node addressing; area too complex");
uint32_t childCount = AROTChildCounts[compSubdivs];
nodeOff = curOff;
nodeSz = childCount * 2 + 4;
curOff += nodeSz;
if (childNodes.size())
{
for (int k=0 ; k < 1 + ((compSubdivs & 0x1) != 0) ; ++k)
{
for (int j=0 ; j < 1 + ((compSubdivs & 0x2) != 0) ; ++j)
{
for (int i=0 ; i < 1 + ((compSubdivs & 0x4) != 0) ; ++i)
{
int idx = k*4 + j*2 + i;
childNodes[idx].nodeCount(sz, idxRefs, bmpPool, curOff);
}
}
}
idxRefs += childCount;
}
}
void AROTBuilder::Node::writeIndirectionTable(athena::io::MemoryWriter& w)
{
w.writeUint32Big(nodeOff);
if (childNodes.size())
{
for (int k=0 ; k < 1 + ((compSubdivs & 0x1) != 0) ; ++k)
{
for (int j=0 ; j < 1 + ((compSubdivs & 0x2) != 0) ; ++j)
{
for (int i=0 ; i < 1 + ((compSubdivs & 0x4) != 0) ; ++i)
{
int idx = k*4 + j*2 + i;
childNodes[idx].writeIndirectionTable(w);
}
}
}
}
}
void AROTBuilder::Node::writeNodes(athena::io::MemoryWriter& w, int nodeIdx)
{
w.writeUint16Big(poolIdx);
w.writeUint16Big(compSubdivs);
if (childNodes.size())
{
int curIdx = nodeIdx + 1;
if (curIdx > 65535)
Log.report(logvisor::Fatal, "AROT node exceeds 16-bit node addressing; area too complex");
int childIndices[8];
for (int k=0 ; k < 1 + ((compSubdivs & 0x1) != 0) ; ++k)
{
for (int j=0 ; j < 1 + ((compSubdivs & 0x2) != 0) ; ++j)
{
for (int i=0 ; i < 1 + ((compSubdivs & 0x4) != 0) ; ++i)
{
int idx = k*4 + j*2 + i;
w.writeUint16Big(curIdx);
childIndices[idx] = curIdx;
childNodes[idx].advanceIndex(curIdx);
}
}
}
for (int k=0 ; k < 1 + ((compSubdivs & 0x1) != 0) ; ++k)
{
for (int j=0 ; j < 1 + ((compSubdivs & 0x2) != 0) ; ++j)
{
for (int i=0 ; i < 1 + ((compSubdivs & 0x4) != 0) ; ++i)
{
int idx = k*4 + j*2 + i;
childNodes[idx].writeNodes(w, childIndices[idx]);
}
}
}
}
}
void AROTBuilder::Node::advanceIndex(int& nodeIdx)
{
++nodeIdx;
if (childNodes.size())
{
for (int k=0 ; k < 1 + ((compSubdivs & 0x1) != 0) ; ++k)
{
for (int j=0 ; j < 1 + ((compSubdivs & 0x2) != 0) ; ++j)
{
for (int i=0 ; i < 1 + ((compSubdivs & 0x4) != 0) ; ++i)
{
int idx = k*4 + j*2 + i;
childNodes[idx].advanceIndex(nodeIdx);
}
}
}
}
}
void AROTBuilder::Node::colSize(size_t& totalSz)
{
if (childIndices.size())
{
nodeOff = totalSz;
if (childNodes.empty())
{
totalSz += 26 + childIndices.size() * 2;
}
else
{
totalSz += 36;
for (int i=0 ; i<8 ; ++i)
childNodes[i].colSize(totalSz);
}
}
}
void AROTBuilder::Node::writeColNodes(uint8_t*& ptr, const zeus::CAABox& curAABB)
{
if (childIndices.size())
{
if (childNodes.empty())
{
float* aabbOut = reinterpret_cast<float*>(ptr);
aabbOut[0] = hecl::SBig(curAABB.min[0]);
aabbOut[1] = hecl::SBig(curAABB.min[1]);
aabbOut[2] = hecl::SBig(curAABB.min[2]);
aabbOut[3] = hecl::SBig(curAABB.max[0]);
aabbOut[4] = hecl::SBig(curAABB.max[1]);
aabbOut[5] = hecl::SBig(curAABB.max[2]);
athena::io::MemoryWriter w(ptr + 24, INT32_MAX);
w.writeUint16Big(childIndices.size());
for (int idx : childIndices)
w.writeUint16Big(idx);
ptr += 26 + childIndices.size() * 2;
}
else
{
uint16_t* pflags = reinterpret_cast<uint16_t*>(ptr);
uint32_t* offsets = reinterpret_cast<uint32_t*>(ptr + 4);
memset(pflags, 0, sizeof(uint32_t) * 9);
for (int i=0 ; i<8 ; ++i)
{
const Node& chNode = childNodes[i];
BspNodeType type = BspNodeType((flags >> (i * 2)) & 0x3);
if (type != BspNodeType::Invalid)
offsets[i] = hecl::SBig(uint32_t(chNode.nodeOff - nodeOff - 36));
}
*pflags = hecl::SBig(flags);
ptr += 36;
for (int i=0 ; i<8 ; ++i)
childNodes[i].writeColNodes(ptr, SplitAABB(curAABB, i));
}
}
}
void AROTBuilder::build(std::vector<std::vector<uint8_t>>& secs, const zeus::CAABox& fullAabb,
const std::vector<zeus::CAABox>& meshAabbs, const std::vector<DNACMDL::Mesh>& meshes)
{
/* Recursively split */
BspNodeType rootType;
rootNode.addChild(0, 1, meshAabbs, fullAabb, rootType);
/* Calculate indexing metrics */
size_t totalNodeCount = 0;
size_t idxRefCount = 0;
size_t curOff = 0;
rootNode.nodeCount(totalNodeCount, idxRefCount, bmpPool, curOff);
size_t bmpWordCount = ROUND_UP_32(meshes.size()) / 32;
size_t arotSz = 64 + bmpWordCount * bmpPool.m_pool.size() * 4 + totalNodeCount * 8 + idxRefCount * 2;
/* Write header */
secs.emplace_back(arotSz, 0);
athena::io::MemoryWriter w(secs.back().data(), secs.back().size());
w.writeUint32Big('AROT');
w.writeUint32Big(1);
w.writeUint32Big(bmpPool.m_pool.size());
w.writeUint32Big(meshes.size());
w.writeUint32Big(totalNodeCount);
w.writeVec3fBig(fullAabb.min);
w.writeVec3fBig(fullAabb.max);
w.seekAlign32();
/* Write bitmap */
std::vector<uint32_t> bmpWords;
bmpWords.reserve(bmpWordCount);
for (const std::set<int>& bmp : bmpPool.m_pool)
{
bmpWords.clear();
bmpWords.resize(bmpWordCount);
auto bmpIt = bmp.cbegin();
if (bmpIt != bmp.cend())
{
int curIdx = 0;
for (int w=0 ; w<bmpWordCount ; ++w)
{
for (int b=0 ; b<32 ; ++b)
{
if (*bmpIt == curIdx)
{
bmpWords[w] |= 1 << b;
++bmpIt;
if (bmpIt == bmp.cend())
break;
}
++curIdx;
}
if (bmpIt == bmp.cend())
break;
}
}
for (uint32_t word : bmpWords)
w.writeUint32Big(word);
}
/* Write the rest */
rootNode.writeIndirectionTable(w);
rootNode.writeNodes(w, 0);
}
std::pair<std::unique_ptr<uint8_t[]>, uint32_t> AROTBuilder::buildCol(const ColMesh& mesh, BspNodeType& rootOut)
{
/* Accumulate total AABB */
zeus::CAABox fullAABB;
for (const auto& vert : mesh.verts)
fullAABB.accumulateBounds(zeus::CVector3f(vert));
/* Predetermine triangle AABBs */
std::vector<zeus::CAABox> triBoxes;
triBoxes.reserve(mesh.trianges.size());
for (const ColMesh::Triangle& tri : mesh.trianges)
{
triBoxes.emplace_back();
zeus::CAABox& aabb = triBoxes.back();
for (int e=0 ; e<3 ; ++e)
{
const ColMesh::Edge& edge = mesh.edges[tri.edges[e]];
for (int v=0 ; v<2 ; ++v)
{
const auto& vert = mesh.verts[edge.verts[v]];
aabb.accumulateBounds(zeus::CVector3f(vert));
}
}
}
/* Recursively split */
rootNode.addChild(0, COLLISION_MIN_NODE_TRIANGLES, triBoxes, fullAABB, rootOut);
/* Calculate offsets and write out */
size_t totalSize = 0;
rootNode.colSize(totalSize);
std::unique_ptr<uint8_t[]> ret(new uint8_t[totalSize]);
uint8_t* ptr = ret.get();
rootNode.writeColNodes(ptr, fullAABB);
return {std::move(ret), totalSize};
}
}