AnotherFoxGuy/polyvox
1
0
Fork 0
Code Issues Pull Requests Activity

Memory improvements for surface extractor.

Browse Source
This commit is contained in:
David Williams 2010-10-20 22:14:03 +00:00
parent 695c15cecd
commit 0b410eaed3
2 changed files with 35 additions and 20 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

11
library/PolyVoxCore/include/CubicSurfaceExtractor.h
View File

@ -45,7 +45,7 @@ namespace PolyVox
void execute(); void execute();
int32_t addVertex(float fX, float fY, float fZ, uint8_t uMaterial); int32_t addVertex(float fX, float fY, float fZ, uint8_t uMaterial, Array<3, IndexAndMaterial>& existingVertices);
private: private:
//The volume data and a sampler to access it. //The volume data and a sampler to access it.
@ -59,7 +59,14 @@ namespace PolyVox
Region m_regSizeInVoxels; Region m_regSizeInVoxels;
Region m_regSizeInCells; Region m_regSizeInCells;
Array<4, IndexAndMaterial> m_vertices; //Array<4, IndexAndMaterial> m_vertices;
Array<3, IndexAndMaterial> m_previousSliceVertices;
Array<3, IndexAndMaterial> m_currentSliceVertices;
//Although we try to avoid creating multiple vertices at the same location, sometimes this is unavoidable
//if they have different materials. For example, four different materials next to each other would mean
//four quads (though more triangles) sharing the vertex. As far as I can tell, four is the worst case scenario.
static const uint32_t MaxQuadsSharingVertex;
}; };
} }

44
library/PolyVoxCore/include/CubicSurfaceExtractor.inl
View File

@ -29,6 +29,9 @@ freely, subject to the following restrictions:
namespace PolyVox namespace PolyVox
{ {
template <typename VoxelType>
const uint32_t CubicSurfaceExtractor<VoxelType>::MaxQuadsSharingVertex = 4;
template <typename VoxelType> template <typename VoxelType>
CubicSurfaceExtractor<VoxelType>::CubicSurfaceExtractor(Volume<VoxelType>* volData, Region region, SurfaceMesh<PositionMaterial>* result) CubicSurfaceExtractor<VoxelType>::CubicSurfaceExtractor(Volume<VoxelType>* volData, Region region, SurfaceMesh<PositionMaterial>* result)
:m_volData(volData) :m_volData(volData)
@ -46,9 +49,11 @@ namespace PolyVox
template <typename VoxelType> template <typename VoxelType>
void CubicSurfaceExtractor<VoxelType>::execute() void CubicSurfaceExtractor<VoxelType>::execute()
{ {
uint32_t arraySize[4]= {m_regSizeInVoxels.width()+2, m_regSizeInVoxels.height()+2, m_regSizeInVoxels.depth()+2, 4}; uint32_t arraySize[3]= {m_regSizeInVoxels.width()+2, m_regSizeInVoxels.height()+2, MaxQuadsSharingVertex};
m_vertices.resize(arraySize); m_previousSliceVertices.resize(arraySize);
memset(m_vertices.getRawData(), 0xff, m_vertices.getNoOfElements() * sizeof(IndexAndMaterial)); m_currentSliceVertices.resize(arraySize);
memset(m_previousSliceVertices.getRawData(), 0xff, m_previousSliceVertices.getNoOfElements() * sizeof(IndexAndMaterial));
memset(m_currentSliceVertices.getRawData(), 0xff, m_currentSliceVertices.getNoOfElements() * sizeof(IndexAndMaterial));
for(uint16_t z = m_regSizeInVoxels.getLowerCorner().getZ(); z <= m_regSizeInVoxels.getUpperCorner().getZ() + 1; z++) for(uint16_t z = m_regSizeInVoxels.getLowerCorner().getZ(); z <= m_regSizeInVoxels.getUpperCorner().getZ() + 1; z++)
{ {
@ -79,10 +84,10 @@ namespace PolyVox
uint32_t v2 = m_meshCurrent->addVertex(PositionMaterial(Vector3DFloat(regX - 0.5f, regY + 0.5f, regZ - 0.5f), material)); uint32_t v2 = m_meshCurrent->addVertex(PositionMaterial(Vector3DFloat(regX - 0.5f, regY + 0.5f, regZ - 0.5f), material));
uint32_t v3 = m_meshCurrent->addVertex(PositionMaterial(Vector3DFloat(regX - 0.5f, regY + 0.5f, regZ + 0.5f), material));*/ uint32_t v3 = m_meshCurrent->addVertex(PositionMaterial(Vector3DFloat(regX - 0.5f, regY + 0.5f, regZ + 0.5f), material));*/
uint32_t v0 = addVertex(regX - 0.5f, regY - 0.5f, regZ - 0.5f, material); uint32_t v0 = addVertex(regX - 0.5f, regY - 0.5f, regZ - 0.5f, material, m_previousSliceVertices);
uint32_t v1 = addVertex(regX - 0.5f, regY - 0.5f, regZ + 0.5f, material); uint32_t v1 = addVertex(regX - 0.5f, regY - 0.5f, regZ + 0.5f, material, m_currentSliceVertices);
uint32_t v2 = addVertex(regX - 0.5f, regY + 0.5f, regZ - 0.5f, material); uint32_t v2 = addVertex(regX - 0.5f, regY + 0.5f, regZ - 0.5f, material, m_previousSliceVertices);
uint32_t v3 = addVertex(regX - 0.5f, regY + 0.5f, regZ + 0.5f, material); uint32_t v3 = addVertex(regX - 0.5f, regY + 0.5f, regZ + 0.5f, material, m_currentSliceVertices);
if(currentVoxelIsSolid > negXVoxelIsSolid) if(currentVoxelIsSolid > negXVoxelIsSolid)
{ {
@ -103,10 +108,10 @@ namespace PolyVox
{ {
int material = std::max(currentVoxel.getMaterial(),negYVoxel.getMaterial()); int material = std::max(currentVoxel.getMaterial(),negYVoxel.getMaterial());
uint32_t v0 = addVertex(regX - 0.5f, regY - 0.5f, regZ - 0.5f, material); uint32_t v0 = addVertex(regX - 0.5f, regY - 0.5f, regZ - 0.5f, material, m_previousSliceVertices);
uint32_t v1 = addVertex(regX - 0.5f, regY - 0.5f, regZ + 0.5f, material); uint32_t v1 = addVertex(regX - 0.5f, regY - 0.5f, regZ + 0.5f, material, m_currentSliceVertices);
uint32_t v2 = addVertex(regX + 0.5f, regY - 0.5f, regZ - 0.5f, material); uint32_t v2 = addVertex(regX + 0.5f, regY - 0.5f, regZ - 0.5f, material, m_previousSliceVertices);
uint32_t v3 = addVertex(regX + 0.5f, regY - 0.5f, regZ + 0.5f, material); uint32_t v3 = addVertex(regX + 0.5f, regY - 0.5f, regZ + 0.5f, material, m_currentSliceVertices);
if(currentVoxelIsSolid > negYVoxelIsSolid) if(currentVoxelIsSolid > negYVoxelIsSolid)
{ {
@ -127,10 +132,10 @@ namespace PolyVox
{ {
int material = std::max(currentVoxel.getMaterial(), negZVoxel.getMaterial()); int material = std::max(currentVoxel.getMaterial(), negZVoxel.getMaterial());
uint32_t v0 = addVertex(regX - 0.5f, regY - 0.5f, regZ - 0.5f, material); uint32_t v0 = addVertex(regX - 0.5f, regY - 0.5f, regZ - 0.5f, material, m_previousSliceVertices);
uint32_t v1 = addVertex(regX - 0.5f, regY + 0.5f, regZ - 0.5f, material); uint32_t v1 = addVertex(regX - 0.5f, regY + 0.5f, regZ - 0.5f, material, m_previousSliceVertices);
uint32_t v2 = addVertex(regX + 0.5f, regY - 0.5f, regZ - 0.5f, material); uint32_t v2 = addVertex(regX + 0.5f, regY - 0.5f, regZ - 0.5f, material, m_previousSliceVertices);
uint32_t v3 = addVertex(regX + 0.5f, regY + 0.5f, regZ - 0.5f, material); uint32_t v3 = addVertex(regX + 0.5f, regY + 0.5f, regZ - 0.5f, material, m_previousSliceVertices);
if(currentVoxelIsSolid > negZVoxelIsSolid) if(currentVoxelIsSolid > negZVoxelIsSolid)
{ {
@ -145,6 +150,9 @@ namespace PolyVox
} }
} }
} }
m_previousSliceVertices.swap(m_currentSliceVertices);
memset(m_currentSliceVertices.getRawData(), 0xff, m_currentSliceVertices.getNoOfElements() * sizeof(IndexAndMaterial));
} }
m_meshCurrent->m_Region = m_regSizeInVoxels; m_meshCurrent->m_Region = m_regSizeInVoxels;
@ -157,15 +165,15 @@ namespace PolyVox
} }
template <typename VoxelType> template <typename VoxelType>
int32_t CubicSurfaceExtractor<VoxelType>::addVertex(float fX, float fY, float fZ, uint8_t uMaterialIn) int32_t CubicSurfaceExtractor<VoxelType>::addVertex(float fX, float fY, float fZ, uint8_t uMaterialIn, Array<3, IndexAndMaterial>& existingVertices)
{ {
uint16_t uX = static_cast<uint16_t>(fX + 0.75f); uint16_t uX = static_cast<uint16_t>(fX + 0.75f);
uint16_t uY = static_cast<uint16_t>(fY + 0.75f); uint16_t uY = static_cast<uint16_t>(fY + 0.75f);
uint16_t uZ = static_cast<uint16_t>(fZ + 0.75f); uint16_t uZ = static_cast<uint16_t>(fZ + 0.75f);
for(int ct = 0; ct < 16; ct++) for(int ct = 0; ct < MaxQuadsSharingVertex; ct++)
{ {
IndexAndMaterial& rEntry = m_vertices[uX][uY][uZ][ct]; IndexAndMaterial& rEntry = existingVertices[uX][uY][ct];
int32_t iIndex = static_cast<int32_t>(rEntry.iIndex); int32_t iIndex = static_cast<int32_t>(rEntry.iIndex);
uint8_t uMaterial = static_cast<uint8_t>(rEntry.uMaterial); uint8_t uMaterial = static_cast<uint8_t>(rEntry.uMaterial);