Files
FlaxEngine/Source/Engine/Graphics/Models/ModelData.Tool.cpp

951 lines
32 KiB
C++

// Copyright (c) Wojciech Figat. All rights reserved.
#if COMPILE_WITH_MODEL_TOOL
#include "ModelData.h"
#include "Engine/Core/Log.h"
#include "Engine/Core/Utilities.h"
#include "Engine/Core/Types/DateTime.h"
#include "Engine/Core/Types/Stopwatch.h"
#include "Engine/Core/Collections/BitArray.h"
#include "Engine/Tools/ModelTool/ModelTool.h"
#include "Engine/Tools/ModelTool/VertexTriangleAdjacency.h"
#include "Engine/Profiler/ProfilerCPU.h"
#include "Engine/Platform/Platform.h"
#define USE_MIKKTSPACE 1
#include "ThirdParty/MikkTSpace/mikktspace.h"
#if USE_ASSIMP
#define USE_SPATIAL_SORT 1
#define ASSIMP_BUILD_NO_EXPORT
#include "Engine/Tools/ModelTool/SpatialSort.h"
//#include <ThirdParty/assimp/SpatialSort.h>
#else
#define USE_SPATIAL_SORT 0
#endif
#include <stack>
#if PLATFORM_WINDOWS
// Import UVAtlas library
// Source: https://github.com/Microsoft/UVAtlas
#include <ThirdParty/UVAtlas/UVAtlas.h>
// Import DirectXMesh library
// Source: https://github.com/Microsoft/DirectXMesh
#include <ThirdParty/DirectXMesh/DirectXMesh.h>
HRESULT __cdecl UVAtlasCallback(float fPercentDone)
{
/*static ULONGLONG s_lastTick = 0;
ULONGLONG tick = GetTickCount64();
if ((tick - s_lastTick) > 1000)
{
wprintf(L"%.2f%% \r", fPercentDone * 100);
s_lastTick = tick;
}
*/
/*
if (_kbhit())
{
if (_getch() == 27)
{
wprintf(L"*** ABORT ***");
return E_ABORT;
}
}
*/
return S_OK;
}
#endif
template<typename T>
void RemapArrayHelper(Array<T>& target, const std::vector<uint32_t>& remap)
{
if (target.HasItems())
{
Array<T> tmp;
tmp.Swap(target);
int32 size = (int32)remap.size();
target.Resize(size, false);
for (int32 i = 0; i < size; i++)
target[i] = tmp.At(remap[i]);
}
}
void MeshData::SetLightmapUVsSource(ModelLightmapUVsSource source)
{
if (source == ModelLightmapUVsSource::Disable)
{
// No lightmap UVs
}
else if (source == ModelLightmapUVsSource::Generate)
{
// Generate lightmap UVs
if (GenerateLightmapUVs())
{
LOG(Error, "Failed to generate lightmap uvs");
}
}
else
{
// Select input channel index
const int32 inputChannelIndex = (int32)source - (int32)ModelLightmapUVsSource::Channel0;
if (inputChannelIndex >= 0 && inputChannelIndex < UVs.Count() && UVs[inputChannelIndex].HasItems())
{
LightmapUVsIndex = inputChannelIndex;
}
else
{
LOG(Warning, "Cannot import result lightmap uvs. Missing texcoords channel {0}.", inputChannelIndex);
}
}
}
bool MeshData::GenerateLightmapUVs()
{
if (Positions.IsEmpty() || Indices.IsEmpty())
return true;
PROFILE_CPU();
#if PLATFORM_WINDOWS
// Prepare
HRESULT hr;
int32 verticesCount = Positions.Count();
int32 facesCount = Indices.Count() / 3;
DirectX::XMFLOAT3* positions = (DirectX::XMFLOAT3*)Positions.Get();
LOG(Info, "Generating lightmaps UVs ({0} vertices, {1} triangles)...", verticesCount, facesCount);
Stopwatch stopwatch;
// Generate adjacency data
const float adjacencyEpsilon = 0.001f;
Array<uint32> adjacency;
adjacency.Resize(Indices.Count());
hr = GenerateAdjacencyAndPointReps(Indices.Get(), facesCount, positions, verticesCount, adjacencyEpsilon, nullptr, adjacency.Get());
if (FAILED(hr))
{
LOG(Warning, "Cannot generate lightmap uvs. Result: {0:x}:{1}", hr, 0);
return true;
}
// Generate UVs
std::vector<DirectX::UVAtlasVertex> vb;
std::vector<uint8_t> ib;
float outStretch = 0.0f;
size_t outCharts = 0;
std::vector<uint32_t> facePartitioning;
std::vector<uint32_t> vertexRemapArray;
int32 size = 1024;
float gutter = 1.0f;
hr = UVAtlasCreate(
positions, verticesCount,
Indices.Get(), DXGI_FORMAT_R32_UINT, facesCount,
0, 0.1f, size, size, gutter,
adjacency.Get(), nullptr, nullptr,
UVAtlasCallback, DirectX::UVATLAS_DEFAULT_CALLBACK_FREQUENCY,
DirectX::UVATLAS_GEODESIC_FAST, vb, ib,
&facePartitioning,
&vertexRemapArray,
&outStretch, &outCharts);
if (FAILED(hr))
{
LOG(Warning, "Cannot generate lightmap uvs. Result: {0:x}:{1}", hr, 1);
return true;
}
// Log info
stopwatch.Stop();
LOG(Info, "Lightmap UVs generated! Charts: {0}, stretching: {1}, {2} vertices. Time: {3}ms", outCharts, outStretch, (int32)vb.size(), stopwatch.GetMilliseconds());
// Update mesh data (remap vertices due to vertex buffer and index buffer change)
RemapArrayHelper(Positions, vertexRemapArray);
LightmapUVsIndex = Math::Min(UVs.Count(), MODEL_MAX_UV - 1);
for (int32 channel = 0; channel < LightmapUVsIndex; channel++)
RemapArrayHelper(UVs[channel], vertexRemapArray);
RemapArrayHelper(Normals, vertexRemapArray);
RemapArrayHelper(Tangents, vertexRemapArray);
RemapArrayHelper(Colors, vertexRemapArray);
RemapArrayHelper(BlendIndices, vertexRemapArray);
RemapArrayHelper(BlendWeights, vertexRemapArray);
uint32* ibP = (uint32*)ib.data();
for (int32 i = 0; i < Indices.Count(); i++)
Indices[i] = *ibP++;
// Add generated data
UVs.Resize(LightmapUVsIndex + 1);
auto& lightmapChannel = UVs[LightmapUVsIndex];
lightmapChannel.Resize((int32)vb.size(), false);
for (int32 i = 0; i < (int32)vb.size(); i++)
lightmapChannel.Get()[i] = *(Float2*)&vb[i].uv;
#else
LOG(Error, "Model lightmap UVs generation is not supported on this platform.");
#endif
return false;
}
int32 FindVertex(const MeshData& mesh, int32 vertexIndex, int32 startIndex, int32 searchRange, const Array<int32>& mapping
#if USE_SPATIAL_SORT
, const Assimp::SpatialSort& spatialSort
, std::vector<unsigned int>& spatialSortCache
#endif
)
{
const float uvEpsSqr = (1.0f / 250.0f) * (1.0f / 250.0f);
const Float2* uv0 = mesh.UVs.Count() > 0 && mesh.UVs[0].HasItems() ? mesh.UVs[0].Get() : nullptr;
const Float2* uv1 = mesh.UVs.Count() > 1 && mesh.UVs[1].HasItems() ? mesh.UVs[1].Get() : nullptr;
const Float2* uv2 = mesh.UVs.Count() > 2 && mesh.UVs[2].HasItems() ? mesh.UVs[2].Get() : nullptr;
const Float2* uv3 = mesh.UVs.Count() > 3 && mesh.UVs[3].HasItems() ? mesh.UVs[3].Get() : nullptr;
#if USE_SPATIAL_SORT
const Float3 vPosition = mesh.Positions[vertexIndex];
spatialSort.FindPositions(*(aiVector3D*)&vPosition, 1e-5f, spatialSortCache);
if (spatialSortCache.empty())
return INVALID_INDEX;
const Float2 vUV0 = uv0 ? uv0[vertexIndex] : Float2::Zero;
const Float2 vUV1 = uv1 ? uv1[vertexIndex] : Float2::Zero;
const Float2 vUV2 = uv2 ? uv2[vertexIndex] : Float2::Zero;
const Float2 vUV3 = uv3 ? uv3[vertexIndex] : Float2::Zero;
const Float3 vNormal = mesh.Normals.HasItems() ? mesh.Normals[vertexIndex] : Float3::Zero;
const Float3 vTangent = mesh.Tangents.HasItems() ? mesh.Tangents[vertexIndex] : Float3::Zero;
const Color vColor = mesh.Colors.HasItems() ? mesh.Colors[vertexIndex] : Color::Black; // Assuming Color::Black as a default color
const int32 end = startIndex + searchRange;
for (size_t i = 0; i < spatialSortCache.size(); i++)
{
const int32 v = spatialSortCache[i];
if (v < startIndex || v >= end)
continue;
#else
const Float3 vPosition = mesh.Positions[vertexIndex];
const Float2 vUV0 = uv0 ? uv0[vertexIndex] : Float2::Zero;
const Float2 vUV1 = uv1 ? uv1[vertexIndex] : Float2::Zero;
const Float2 vUV2 = uv2 ? uv2[vertexIndex] : Float2::Zero;
const Float2 vUV3 = uv3 ? uv3[vertexIndex] : Float2::Zero;
const Float3 vNormal = mesh.Normals.HasItems() ? mesh.Normals[vertexIndex] : Float3::Zero;
const Float3 vTangent = mesh.Tangents.HasItems() ? mesh.Tangents[vertexIndex] : Float3::Zero;
const Color vColor = mesh.Colors.HasItems() ? mesh.Colors[vertexIndex] : Color::Black; // Assuming Color::Black as a default color
const int32 end = startIndex + searchRange;
for (int32 v = startIndex; v < end; v++)
{
if (!Float3::NearEqual(vPosition, mesh.Positions[v]))
continue;
#endif
if (mapping[v] == INVALID_INDEX)
continue;
if (uv0 && (vUV0 - uv0[v]).LengthSquared() > uvEpsSqr)
continue;
if (uv1 && (vUV1 - uv1[v]).LengthSquared() > uvEpsSqr)
continue;
if (uv2 && (vUV2 - uv2[v]).LengthSquared() > uvEpsSqr)
continue;
if (uv3 && (vUV3 - uv3[v]).LengthSquared() > uvEpsSqr)
continue;
if (mesh.Normals.HasItems() && Float3::Dot(vNormal, mesh.Normals[v]) < 0.98f)
continue;
if (mesh.Tangents.HasItems() && Float3::Dot(vTangent, mesh.Tangents[v]) < 0.98f)
continue;
if (mesh.Colors.HasItems() && vColor != mesh.Colors[v])
continue;
return v;
}
return INVALID_INDEX;
}
template<typename T>
void RemapBuffer(Array<T>& src, Array<T>& dst, const Array<int32>& mapping, int32 newSize)
{
if (src.IsEmpty())
return;
dst.Resize(newSize, false);
for (int32 i = 0, j = 0; i < src.Count(); i++)
{
const auto idx = mapping[i];
if (idx != INVALID_INDEX)
{
dst[j++] = src[i];
}
}
}
void MeshData::BuildIndexBuffer()
{
PROFILE_CPU();
Stopwatch stopwatch;
const int32 vertexCount = Positions.Count();
MeshData newMesh;
newMesh.Indices.EnsureCapacity(vertexCount);
Array<int32> mapping;
mapping.Resize(vertexCount);
int32 newVertexCounter = 0;
#if USE_SPATIAL_SORT
// Set up a SpatialSort to quickly find all vertices close to a given position
Assimp::SpatialSort vertexFinder;
vertexFinder.Fill((const aiVector3D*)Positions.Get(), vertexCount, sizeof(Float3));
std::vector<unsigned int> spatialSortCache;
#endif
// Build index buffer
for (int32 vertexIndex = 0; vertexIndex < vertexCount; vertexIndex++)
{
// Find duplicated vertex before the current one
const int32 reuseVertexIndex = FindVertex(*this, vertexIndex, 0, vertexIndex, mapping
#if USE_SPATIAL_SORT
, vertexFinder, spatialSortCache
#endif
);
if (reuseVertexIndex == INVALID_INDEX)
{
newMesh.Indices.Add(newVertexCounter);
mapping[vertexIndex] = newVertexCounter;
newVertexCounter++;
}
else
{
// Remove vertex
const int32 mapped = mapping[reuseVertexIndex];
ASSERT(mapped != INVALID_INDEX && mapped < vertexIndex);
newMesh.Indices.Add(mapped);
mapping[vertexIndex] = INVALID_INDEX;
}
}
// Skip if no change
if (vertexCount == newVertexCounter)
return;
// Remove unused vertices
newMesh.SwapBuffers(*this);
#define REMAP_BUFFER(name) RemapBuffer(newMesh.name, name, mapping, newVertexCounter)
REMAP_BUFFER(Positions);
REMAP_BUFFER(Normals);
REMAP_BUFFER(Tangents);
REMAP_BUFFER(BitangentSigns);
REMAP_BUFFER(Colors);
REMAP_BUFFER(BlendIndices);
REMAP_BUFFER(BlendWeights);
UVs.Resize(newMesh.UVs.Count());
for (int32 channelIdx = 0; channelIdx < UVs.Count(); channelIdx++)
REMAP_BUFFER(UVs[channelIdx]);
#undef REMAP_BUFFER
BlendShapes.Resize(newMesh.BlendShapes.Count());
for (int32 blendShapeIndex = 0; blendShapeIndex < newMesh.BlendShapes.Count(); blendShapeIndex++)
{
auto& srcBlendShape = newMesh.BlendShapes[blendShapeIndex];
auto& dstBlendShape = BlendShapes[blendShapeIndex];
dstBlendShape.Name = srcBlendShape.Name;
dstBlendShape.Weight = srcBlendShape.Weight;
dstBlendShape.Vertices.EnsureCapacity(srcBlendShape.Vertices.Count());
for (int32 i = 0; i < srcBlendShape.Vertices.Count(); i++)
{
auto& v = srcBlendShape.Vertices[i];
int32 newVertexIndex = v.VertexIndex < (uint32)vertexCount ? mapping[v.VertexIndex] : INVALID_INDEX;
if (newVertexIndex != INVALID_INDEX)
{
v.VertexIndex = newVertexIndex;
dstBlendShape.Vertices.Add(v);
}
}
}
stopwatch.Stop();
const double time = Utilities::RoundTo2DecimalPlaces(stopwatch.GetTotalSeconds());
if (time > 0.5f) // Don't log if generation was fast enough
LOG(Info, "Generated {3} for mesh in {0}s ({1} vertices, {2} indices)", time, vertexCount, Indices.Count(), TEXT("indices"));
}
void MeshData::FindPositions(const Float3& position, float epsilon, Array<int32>& result)
{
result.Clear();
for (int32 i = 0; i < Positions.Count(); i++)
{
if (Float3::NearEqual(position, Positions[i], epsilon))
result.Add(i);
}
}
bool MeshData::GenerateNormals(float smoothingAngle)
{
if (Positions.IsEmpty() || Indices.IsEmpty())
{
LOG(Warning, "Missing vertex or index data to generate normals.");
return true;
}
PROFILE_CPU();
const auto startTime = Platform::GetTimeSeconds();
const int32 vertexCount = Positions.Count();
const int32 indexCount = Indices.Count();
Normals.Resize(vertexCount, false);
smoothingAngle = Math::Clamp(smoothingAngle, 0.0f, 175.0f);
// Compute per-face normals but store them per-vertex
Float3 min, max;
min = max = Positions[0];
for (int32 i = 0; i < indexCount; i += 3)
{
const Float3 v1 = Positions[Indices[i + 0]];
const Float3 v2 = Positions[Indices[i + 1]];
const Float3 v3 = Positions[Indices[i + 2]];
const Float3 n = ((v2 - v1) ^ (v3 - v1));
Normals[Indices[i + 0]] = n;
Normals[Indices[i + 1]] = n;
Normals[Indices[i + 2]] = n;
Float3::Min(min, v1, min);
Float3::Min(min, v2, min);
Float3::Min(min, v3, min);
Float3::Max(max, v1, max);
Float3::Max(max, v2, max);
Float3::Max(max, v3, max);
}
#if USE_SPATIAL_SORT
// Set up a SpatialSort to quickly find all vertices close to a given position
Assimp::SpatialSort vertexFinder;
vertexFinder.Fill((const aiVector3D*)Positions.Get(), vertexCount, sizeof(Float3));
std::vector<uint32> verticesFound(16);
#else
Array<int32> verticesFound(16);
#endif
const float posEpsilon = (max - min).Length() * 1e-4f;
// Check if use the angle limit (then use the faster path)
if (smoothingAngle >= 175.0f)
{
BitArray<> used;
used.Resize(vertexCount);
used.SetAll(false);
for (int32 i = 0; i < vertexCount; i++)
{
if (used[i])
continue;
// Get all vertices that share this one
#if USE_SPATIAL_SORT
vertexFinder.FindPositions(*(aiVector3D*)&Positions[i], posEpsilon, verticesFound);
const int32 verticesFoundCount = (int32)verticesFound.size();
#else
FindPositions(Positions[i], posEpsilon, verticesFound);
const int32 verticesFoundCount = verticesFound.Count();
#endif
// Get the smooth normal
Float3 n;
for (int32 a = 0; a < verticesFoundCount; a++)
n += Normals[verticesFound[a]];
n.Normalize();
// Write the smoothed normal back to all affected normals
for (int32 a = 0; a < verticesFoundCount; a++)
{
const auto vtx = verticesFound[a];
Normals[vtx] = n;
used.Set(vtx, true);
}
}
}
else
{
const float limit = cosf(smoothingAngle * DegreesToRadians);
for (int32 i = 0; i < vertexCount; i++)
{
// Get all vertices that share this one
#if USE_SPATIAL_SORT
vertexFinder.FindPositions(*(aiVector3D*)&Positions[i], posEpsilon, verticesFound);
const int32 verticesFoundCount = (int32)verticesFound.size();
#else
FindPositions(Positions[i], posEpsilon, verticesFound);
const int32 verticesFoundCount = verticesFound.Count();
#endif
// Get the smooth normal
Float3 vr = Normals[i];
const float vrlen = vr.Length();
Float3 n;
for (int32 a = 0; a < verticesFoundCount; a++)
{
Float3 v = Normals[verticesFound[a]];
// Check whether the angle between the two normals is not too large
// TODO: use length squared?
if (v * vr >= limit * vrlen * v.Length())
n += v;
}
Normals[i] = Float3::Normalize(n);
}
}
const auto endTime = Platform::GetTimeSeconds();
const double time = Utilities::RoundTo2DecimalPlaces(endTime - startTime);
if (time > 0.5f) // Don't log if generation was fast enough
LOG(Info, "Generated {3} for mesh in {0}s ({1} vertices, {2} indices)", time, vertexCount, indexCount, TEXT("normals"));
return false;
}
#if USE_MIKKTSPACE
namespace
{
int GetNumFaces(const SMikkTSpaceContext* pContext)
{
const auto meshData = (MeshData*)pContext->m_pUserData;
return meshData->Indices.Count() / 3;
}
int GetNumVerticesOfFace(const SMikkTSpaceContext* pContext, const int iFace)
{
return 3;
}
void GetPosition(const SMikkTSpaceContext* pContext, float fvPosOut[], const int iFace, const int iVert)
{
const auto meshData = (MeshData*)pContext->m_pUserData;
const auto e = meshData->Positions[meshData->Indices[iFace * 3 + iVert]];
fvPosOut[0] = e.X;
fvPosOut[1] = e.Y;
fvPosOut[2] = e.Z;
}
void GetNormal(const SMikkTSpaceContext* pContext, float fvNormOut[], const int iFace, const int iVert)
{
const auto meshData = (MeshData*)pContext->m_pUserData;
const auto e = meshData->Normals[meshData->Indices[iFace * 3 + iVert]];
fvNormOut[0] = e.X;
fvNormOut[1] = e.Y;
fvNormOut[2] = e.Z;
}
void GetTexCoord(const SMikkTSpaceContext* pContext, float fvTexcOut[], const int iFace, const int iVert)
{
const auto meshData = (MeshData*)pContext->m_pUserData;
const auto e = meshData->UVs[0][meshData->Indices[iFace * 3 + iVert]];
fvTexcOut[0] = e.X;
fvTexcOut[1] = e.Y;
}
void SetTSpaceBasic(const SMikkTSpaceContext* pContext, const float fvTangent[], const float fSign, const int iFace, const int iVert)
{
const auto meshData = (MeshData*)pContext->m_pUserData;
const auto v = meshData->Indices[iFace * 3 + iVert];
meshData->Tangents[v] = Float3(fvTangent);
meshData->BitangentSigns[v] = fSign;
}
}
#endif
bool MeshData::GenerateTangents(float smoothingAngle)
{
if (Positions.IsEmpty() || Indices.IsEmpty())
{
LOG(Warning, "Missing vertex or index data to generate tangents.");
return true;
}
if (Normals.IsEmpty() || UVs.IsEmpty())
{
LOG(Warning, "Missing normals or texcoords data to generate tangents.");
return true;
}
PROFILE_CPU();
const auto startTime = Platform::GetTimeSeconds();
const int32 vertexCount = Positions.Count();
const int32 indexCount = Indices.Count();
Tangents.Resize(vertexCount, false);
smoothingAngle = Math::Clamp(smoothingAngle, 0.0f, 45.0f);
#if USE_MIKKTSPACE
SMikkTSpaceInterface callbacks = {
GetNumFaces,
GetNumVerticesOfFace,
GetPosition,
GetNormal,
GetTexCoord,
SetTSpaceBasic,
nullptr
};
const SMikkTSpaceContext context = { &callbacks, this };
BitangentSigns.Resize(vertexCount, false);
genTangSpace(&context, 180.0f - smoothingAngle);
#else
const float angleEpsilon = 0.9999f;
BitArray<> vertexDone;
vertexDone.Resize(vertexCount);
vertexDone.SetAll(false);
const Float3* meshNorm = Normals.Get();
const Float2* meshTex = UVs[0].Get();
Float3* meshTang = Tangents.Get();
// Calculate the tangent per-triangle
Float3 min, max;
min = max = Positions[Indices[0]];
for (int32 i = 0; i < indexCount; i += 3)
{
const int32 p0 = Indices[i + 0], p1 = Indices[i + 1], p2 = Indices[i + 2];
const Float3 v0 = Positions[p0];
const Float3 v1 = Positions[p1];
const Float3 v2 = Positions[p2];
Float3::Min(min, v0, min);
Float3::Min(min, v1, min);
Float3::Min(min, v2, min);
Float3::Max(max, v0, max);
Float3::Max(max, v1, max);
Float3::Max(max, v2, max);
// Position differences p1->p2 and p1->p3
Float3 v = v1 - v0, w = v2 - v0;
// Texture offset p1->p2 and p1->p3
float sx = meshTex[p1].X - meshTex[p0].X, sy = meshTex[p1].Y - meshTex[p0].Y;
float tx = meshTex[p2].X - meshTex[p0].X, ty = meshTex[p2].Y - meshTex[p0].Y;
const float dir = (tx * sy - ty * sx) < 0.0f ? -1.0f : 1.0f;
if (sx * ty == sy * tx)
{
// Use default UV direction for invalid case
sx = 0.0;
sy = 1.0;
tx = 1.0;
ty = 0.0;
}
// Tangent points in the direction where to positive X axis of the texture coord's would point in model space
// Bitangent's points along the positive Y axis of the texture coord's, respectively
Float3 tangent, bitangent;
tangent.X = (w.X * sy - v.X * ty) * dir;
tangent.Y = (w.Y * sy - v.Y * ty) * dir;
tangent.Z = (w.Z * sy - v.Z * ty) * dir;
bitangent.X = (w.X * sx - v.X * tx) * dir;
bitangent.Y = (w.Y * sx - v.Y * tx) * dir;
bitangent.Z = (w.Z * sx - v.Z * tx) * dir;
// Set tangent frame for every vertex in that triangle
for (int32 b = 0; b < 3; b++)
{
const int32 p = Indices[i + b];
// Project tangent and bitangent into the plane formed by the normal
Float3 localTangent = tangent - meshNorm[p] * (tangent * meshNorm[p]);
Float3 localBitangent = bitangent - meshNorm[p] * (bitangent * meshNorm[p]);
localTangent.Normalize();
localBitangent.Normalize();
// Reconstruct tangent according to normal and bitangent when it's infinite or NaN
if (localTangent.IsNanOrInfinity())
{
localTangent = meshNorm[p] ^ localBitangent;
localTangent.Normalize();
}
// Write data into the mesh
meshTang[p] = localTangent;
}
}
#if USE_SPATIAL_SORT
// Set up a SpatialSort to quickly find all vertices close to a given position
Assimp::SpatialSort vertexFinder;
vertexFinder.Fill((const aiVector3D*)Positions.Get(), vertexCount, sizeof(Float3));
std::vector<uint32> verticesFound(16);
#else
Array<int32> verticesFound(16);
#endif
const float posEpsilon = (max - min).Length() * 1e-4f;
const float limit = cosf(smoothingAngle * DegreesToRadians);
Array<int32> closeVertices;
// In the second pass we now smooth out all tangents at the same local position if they are not too far off
for (int32 a = 0; a < vertexCount; a++)
{
if (vertexDone[a])
continue;
const Float3 origPos = Positions[a];
const Float3 origNorm = Normals[a];
const Float3 origTang = Tangents[a];
closeVertices.Clear();
// Find all vertices close to that position
#if USE_SPATIAL_SORT
vertexFinder.FindPositions(*(aiVector3D*)&origPos, posEpsilon, verticesFound);
const int32 verticesFoundCount = (int32)verticesFound.size();
#else
FindPositions(origPos, posEpsilon, verticesFound);
const int32 verticesFoundCount = verticesFound.Count();
#endif
closeVertices.EnsureCapacity(verticesFoundCount + 5);
closeVertices.Add(a);
// Look among them for other vertices sharing the same normal and a close-enough tangent
for (int32 b = 0; b < verticesFoundCount; b++)
{
const int32 idx = verticesFound[b];
if (vertexDone[idx])
continue;
if (meshNorm[idx] * origNorm < angleEpsilon)
continue;
if (meshTang[idx] * origTang < limit)
continue;
// It's similar enough -> add it to the smoothing group
closeVertices.Add(idx);
vertexDone.Set(idx, true);
}
// Smooth the tangents of all vertices that were found to be close enough
Float3 smoothTangent(0, 0, 0), smoothBitangent(0, 0, 0);
for (int32 b = 0; b < closeVertices.Count(); b++)
{
auto p = closeVertices[b];
smoothTangent += meshTang[p];
}
smoothTangent.Normalize();
// Write it back into all affected tangents
for (int32 b = 0; b < closeVertices.Count(); b++)
{
meshTang[closeVertices[b]] = smoothTangent;
}
}
#endif
const auto endTime = Platform::GetTimeSeconds();
const double time = Utilities::RoundTo2DecimalPlaces(endTime - startTime);
if (time > 0.5f) // Don't log if generation was fast enough
LOG(Info, "Generated {3} for mesh in {0}s ({1} vertices, {2} indices)", time, vertexCount, indexCount, TEXT("tangents"));
return false;
}
void MeshData::ImproveCacheLocality()
{
// The algorithm is roughly basing on this paper (except without overdraw reduction):
// http://www.cs.princeton.edu/gfx/pubs/Sander_2007_%3ETR/tipsy.pdf
// Configured size of the cache for the vertex buffer used by the algorithm (size is in vertices count, it's not a stride)
const int32 VertexCacheSize = 12;
if (Positions.IsEmpty() || Indices.IsEmpty() || Positions.Count() <= VertexCacheSize)
return;
PROFILE_CPU();
const auto startTime = Platform::GetTimeSeconds();
const auto vertexCount = Positions.Count();
const auto indexCount = Indices.Count();
const uint32* const pcEnd = Indices.Get() + Indices.Count();
// First we need to build a vertex-triangle adjacency list
VertexTriangleAdjacency adj(Indices.Get(), indexCount, vertexCount, true);
// Build a list to store per-vertex caching time stamps
uint32* const piCachingStamps = NewArray<uint32>(vertexCount);
Platform::MemoryClear(piCachingStamps, vertexCount * sizeof(uint32));
// Allocate an empty output index buffer. We store the output indices in one large array.
// Since the number of triangles won't change the input triangles can be reused. This is how
const uint32 iIdxCnt = indexCount;
uint32* const piIBOutput = NewArray<uint32>(iIdxCnt);
uint32* piCSIter = piIBOutput;
// Allocate the flag array to hold the information
// Whether a triangle has already been emitted or not
std::vector<bool> abEmitted(indexCount / 3, false);
// Dead-end vertex index stack
std::stack<uint32, std::vector<uint32>> sDeadEndVStack;
// Create a copy of the piNumTriPtr buffer
uint32* const piNumTriPtr = adj.LiveTriangles;
const std::vector<uint32> piNumTriPtrNoModify(piNumTriPtr, piNumTriPtr + vertexCount);
// Get the largest number of referenced triangles and allocate the "candidate buffer"
uint32 iMaxRefTris = 0;
{
const uint32* piCur = adj.LiveTriangles;
const uint32* const piCurEnd = adj.LiveTriangles + vertexCount;
for (; piCur != piCurEnd; piCur++)
{
iMaxRefTris = Math::Max(iMaxRefTris, *piCur);
}
}
ASSERT(iMaxRefTris > 0);
uint32* piCandidates = NewArray<uint32>(iMaxRefTris * 3);
uint32 iCacheMisses = 0;
int ivdx = 0;
int ics = 1;
int iStampCnt = VertexCacheSize + 1;
while (ivdx >= 0)
{
uint32 icnt = piNumTriPtrNoModify[ivdx];
uint32* piList = adj.GetAdjacentTriangles(ivdx);
uint32* piCurCandidate = piCandidates;
// Get all triangles in the neighborhood
for (uint32 tri = 0; tri < icnt; tri++)
{
// If they have not yet been emitted, add them to the output IB
const uint32 fidx = *piList++;
if (!abEmitted[fidx])
{
// So iterate through all vertices of the current triangle
uint32* pcFace = &Indices[fidx * 3];
for (uint32 *p = pcFace, *p2 = pcFace + 3; p != p2; p++)
{
const uint32 dp = *p;
// The current vertex won't have any free triangles after this step
if (ivdx != (int)dp)
{
// Append the vertex to the dead-end stack
sDeadEndVStack.push(dp);
// Register as candidate for the next step
*piCurCandidate++ = dp;
// Decrease the per-vertex triangle counts
piNumTriPtr[dp]--;
}
// Append the vertex to the output index buffer
*piCSIter++ = dp;
// If the vertex is not yet in cache, set its cache count
if (iStampCnt - piCachingStamps[dp] > VertexCacheSize)
{
piCachingStamps[dp] = iStampCnt++;
iCacheMisses++;
}
}
// Flag triangle as emitted
abEmitted[fidx] = true;
}
}
// The vertex has now no living adjacent triangles anymore
piNumTriPtr[ivdx] = 0;
// Get next fanning vertex
ivdx = -1;
int max_priority = -1;
for (uint32* piCur = piCandidates; piCur != piCurCandidate; ++piCur)
{
const uint32 dp = *piCur;
// Must have live triangles
if (piNumTriPtr[dp] > 0)
{
int priority = 0;
// Will the vertex be in cache, even after fanning occurs?
uint32 tmp;
if ((tmp = iStampCnt - piCachingStamps[dp]) + 2 * piNumTriPtr[dp] <= VertexCacheSize)
{
priority = tmp;
}
// Keep best candidate
if (priority > max_priority)
{
max_priority = priority;
ivdx = dp;
}
}
}
// Did we reach a dead end?
if (-1 == ivdx)
{
// Need to get a non-local vertex for which we have a good chance that it is still in the cache
while (!sDeadEndVStack.empty())
{
uint32 iCachedIdx = sDeadEndVStack.top();
sDeadEndVStack.pop();
if (piNumTriPtr[iCachedIdx] > 0)
{
ivdx = iCachedIdx;
break;
}
}
if (-1 == ivdx)
{
// Well, there isn't such a vertex. Simply get the next vertex in input order and hope it is not too bad
while (ics < (int)vertexCount)
{
ics++;
if (piNumTriPtr[ics] > 0)
{
ivdx = ics;
break;
}
}
}
}
}
// Sort the output index buffer back to the input array
piCSIter = piIBOutput;
for (uint32* pcFace = Indices.Get(); pcFace != pcEnd; pcFace += 3)
{
pcFace[0] = *piCSIter++;
pcFace[1] = *piCSIter++;
pcFace[2] = *piCSIter++;
}
// Delete temporary storage
Allocator::Free(piCachingStamps);
Allocator::Free(piIBOutput);
Allocator::Free(piCandidates);
const auto endTime = Platform::GetTimeSeconds();
const double time = Utilities::RoundTo2DecimalPlaces(endTime - startTime);
if (time > 0.5f) // Don't log if generation was fast enough
LOG(Info, "Generated {3} for mesh in {0}s ({1} vertices, {2} indices)", time, vertexCount, indexCount, TEXT("optimized indices"));
}
float MeshData::CalculateTrianglesArea() const
{
PROFILE_CPU();
float sum = 0;
// TODO: use SIMD
for (int32 i = 0; i + 2 < Indices.Count(); i += 3)
{
const Float3 v1 = Positions[Indices[i + 0]];
const Float3 v2 = Positions[Indices[i + 1]];
const Float3 v3 = Positions[Indices[i + 2]];
sum += Float3::TriangleArea(v1, v2, v3);
}
return sum;
}
#endif