Fix/improve raycast performance for btBvhTriangleMeshShape (btQuantizedBvh, btOptimizedBvh)

Add raycast acceleration for btAxisSweep3, using btDbvtBroadphase, providing large speedup. Can be disabled in constructor to safe memory.
This commit is contained in:
erwin.coumans
2008-10-14 20:24:28 +00:00
parent fffca75e9f
commit e6c850f13b
7 changed files with 207 additions and 61 deletions

View File

@@ -18,14 +18,16 @@ subject to the following restrictions:
#include "LinearMath/btAabbUtil2.h"
#include "LinearMath/btIDebugDraw.h"
#define RAYAABB2
btQuantizedBvh::btQuantizedBvh() : m_useQuantization(false),
//m_traversalMode(TRAVERSAL_STACKLESS_CACHE_FRIENDLY)
m_traversalMode(TRAVERSAL_STACKLESS)
//m_traversalMode(TRAVERSAL_RECURSIVE)
,m_subtreeHeaderCount(0) //PCK: add this line
{
{
m_bvhAabbMin.setValue(-SIMD_INFINITY,-SIMD_INFINITY,-SIMD_INFINITY);
m_bvhAabbMax.setValue(SIMD_INFINITY,SIMD_INFINITY,SIMD_INFINITY);
}
@@ -119,7 +121,7 @@ void btQuantizedBvh::buildTree (int startIndex,int endIndex)
int numIndices =endIndex-startIndex;
int curIndex = m_curNodeIndex;
assert(numIndices>0);
btAssert(numIndices>0);
if (numIndices==1)
{
@@ -140,8 +142,11 @@ void btQuantizedBvh::buildTree (int startIndex,int endIndex)
int internalNodeIndex = m_curNodeIndex;
setInternalNodeAabbMax(m_curNodeIndex,m_bvhAabbMin);
setInternalNodeAabbMin(m_curNodeIndex,m_bvhAabbMax);
//set the min aabb to 'inf' or a max value, and set the max aabb to a -inf/minimum value.
//the aabb will be expanded during buildTree/mergeInternalNodeAabb with actual node values
setInternalNodeAabbMin(m_curNodeIndex,m_bvhAabbMax);//can't use btVector3(SIMD_INFINITY,SIMD_INFINITY,SIMD_INFINITY)) because of quantization
setInternalNodeAabbMax(m_curNodeIndex,m_bvhAabbMin);//can't use btVector3(-SIMD_INFINITY,-SIMD_INFINITY,-SIMD_INFINITY)) because of quantization
for (i=startIndex;i<endIndex;i++)
{
@@ -177,6 +182,9 @@ void btQuantizedBvh::buildTree (int startIndex,int endIndex)
{
updateSubtreeHeaders(leftChildNodexIndex,rightChildNodexIndex);
}
} else
{
}
setInternalNodeEscapeIndex(internalNodeIndex,escapeIndex);
@@ -338,6 +346,7 @@ void btQuantizedBvh::reportAabbOverlappingNodex(btNodeOverlapCallback* nodeCallb
int maxIterations = 0;
void btQuantizedBvh::walkStacklessTree(btNodeOverlapCallback* nodeCallback,const btVector3& aabbMin,const btVector3& aabbMax) const
{
btAssert(!m_useQuantization);
@@ -352,7 +361,7 @@ void btQuantizedBvh::walkStacklessTree(btNodeOverlapCallback* nodeCallback,const
while (curIndex < m_curNodeIndex)
{
//catch bugs in tree data
assert (walkIterations < m_curNodeIndex);
btAssert (walkIterations < m_curNodeIndex);
walkIterations++;
aabbOverlap = TestAabbAgainstAabb2(aabbMin,aabbMax,rootNode->m_aabbMinOrg,rootNode->m_aabbMaxOrg);
@@ -434,6 +443,96 @@ void btQuantizedBvh::walkRecursiveQuantizedTreeAgainstQueryAabb(const btQuantize
void btQuantizedBvh::walkStacklessTreeAgainstRay(btNodeOverlapCallback* nodeCallback, const btVector3& raySource, const btVector3& rayTarget, const btVector3& aabbMin, const btVector3& aabbMax, int startNodeIndex,int endNodeIndex) const
{
btAssert(!m_useQuantization);
const btOptimizedBvhNode* rootNode = &m_contiguousNodes[0];
int escapeIndex, curIndex = 0;
int walkIterations = 0;
bool isLeafNode;
//PCK: unsigned instead of bool
unsigned aabbOverlap=0;
unsigned rayBoxOverlap=0;
btScalar lambda_max = 1.0;
/* Quick pruning by quantized box */
btVector3 rayAabbMin = raySource;
btVector3 rayAabbMax = raySource;
rayAabbMin.setMin(rayTarget);
rayAabbMax.setMax(rayTarget);
/* Add box cast extents to bounding box */
rayAabbMin += aabbMin;
rayAabbMax += aabbMax;
#ifdef RAYAABB2
btVector3 rayFrom = raySource;
btVector3 rayDirection = (rayTarget-raySource);
rayDirection.normalize ();
lambda_max = rayDirection.dot(rayTarget-raySource);
///what about division by zero? --> just set rayDirection[i] to 1.0
rayDirection[0] = rayDirection[0] == btScalar(0.0) ? btScalar(1e30) : btScalar(1.0) / rayDirection[0];
rayDirection[1] = rayDirection[1] == btScalar(0.0) ? btScalar(1e30) : btScalar(1.0) / rayDirection[1];
rayDirection[2] = rayDirection[2] == btScalar(0.0) ? btScalar(1e30) : btScalar(1.0) / rayDirection[2];
unsigned int sign[3] = { rayDirection[0] < 0.0, rayDirection[1] < 0.0, rayDirection[2] < 0.0};
#endif
btVector3 bounds[2];
while (curIndex < m_curNodeIndex)
{
btScalar param = 1.0;
//catch bugs in tree data
btAssert (walkIterations < m_curNodeIndex);
walkIterations++;
bounds[0] = rootNode->m_aabbMinOrg;
bounds[1] = rootNode->m_aabbMaxOrg;
/* Add box cast extents */
bounds[0] += aabbMin;
bounds[1] += aabbMax;
aabbOverlap = TestAabbAgainstAabb2(rayAabbMin,rayAabbMax,rootNode->m_aabbMinOrg,rootNode->m_aabbMaxOrg);
//perhaps profile if it is worth doing the aabbOverlap test first
#ifdef RAYAABB2
///careful with this check: need to check division by zero (above) and fix the unQuantize method
///thanks Joerg/hiker for the reproduction case!
///http://www.bulletphysics.com/Bullet/phpBB3/viewtopic.php?f=9&t=1858
rayBoxOverlap = aabbOverlap ? btRayAabb2 (raySource, rayDirection, sign, bounds, param, 0.0f, lambda_max) : false;
#else
btVector3 normal;
rayBoxOverlap = btRayAabb(raySource, rayTarget,bounds[0],bounds[1],param, normal);
#endif
isLeafNode = rootNode->m_escapeIndex == -1;
//PCK: unsigned instead of bool
if (isLeafNode && (rayBoxOverlap != 0))
{
nodeCallback->processNode(rootNode->m_subPart,rootNode->m_triangleIndex);
}
//PCK: unsigned instead of bool
if ((rayBoxOverlap != 0) || isLeafNode)
{
rootNode++;
curIndex++;
} else
{
escapeIndex = rootNode->m_escapeIndex;
rootNode += escapeIndex;
curIndex += escapeIndex;
}
}
if (maxIterations < walkIterations)
maxIterations = walkIterations;
}
void btQuantizedBvh::walkStacklessQuantizedTreeAgainstRay(btNodeOverlapCallback* nodeCallback, const btVector3& raySource, const btVector3& rayTarget, const btVector3& aabbMin, const btVector3& aabbMax, int startNodeIndex,int endNodeIndex) const
@@ -454,7 +553,7 @@ void btQuantizedBvh::walkStacklessQuantizedTreeAgainstRay(btNodeOverlapCallback*
unsigned rayBoxOverlap = 0;
btScalar lambda_max = 1.0;
#define RAYAABB2
#ifdef RAYAABB2
btVector3 rayFrom = raySource;
btVector3 rayDirection = (rayTarget-raySource);
@@ -502,7 +601,7 @@ void btQuantizedBvh::walkStacklessQuantizedTreeAgainstRay(btNodeOverlapCallback*
#endif//VISUALLY_ANALYZE_BVH
//catch bugs in tree data
assert (walkIterations < subTreeSize);
btAssert (walkIterations < subTreeSize);
walkIterations++;
//PCK: unsigned instead of bool
@@ -533,7 +632,9 @@ void btQuantizedBvh::walkStacklessQuantizedTreeAgainstRay(btNodeOverlapCallback*
///thanks Joerg/hiker for the reproduction case!
///http://www.bulletphysics.com/Bullet/phpBB3/viewtopic.php?f=9&t=1858
//BT_PROFILE("btRayAabb2");
rayBoxOverlap = btRayAabb2 (raySource, rayDirection, sign, bounds, param, 0.0f, lambda_max);
#else
rayBoxOverlap = true;//btRayAabb(raySource, rayTarget, bounds[0], bounds[1], param, normal);
#endif
@@ -597,7 +698,7 @@ void btQuantizedBvh::walkStacklessQuantizedTree(btNodeOverlapCallback* nodeCallb
#endif//VISUALLY_ANALYZE_BVH
//catch bugs in tree data
assert (walkIterations < subTreeSize);
btAssert (walkIterations < subTreeSize);
walkIterations++;
//PCK: unsigned instead of bool
@@ -652,30 +753,25 @@ void btQuantizedBvh::walkStacklessQuantizedTreeCacheFriendly(btNodeOverlapCallba
void btQuantizedBvh::reportRayOverlappingNodex (btNodeOverlapCallback* nodeCallback, const btVector3& raySource, const btVector3& rayTarget) const
{
bool fast_path = m_useQuantization && m_traversalMode == TRAVERSAL_STACKLESS;
if (fast_path)
{
walkStacklessQuantizedTreeAgainstRay(nodeCallback, raySource, rayTarget, btVector3(0, 0, 0), btVector3(0, 0, 0), 0, m_curNodeIndex);
} else {
/* Otherwise fallback to AABB overlap test */
btVector3 aabbMin = raySource;
btVector3 aabbMax = raySource;
aabbMin.setMin(rayTarget);
aabbMax.setMax(rayTarget);
reportAabbOverlappingNodex(nodeCallback,aabbMin,aabbMax);
}
reportBoxCastOverlappingNodex(nodeCallback,raySource,rayTarget,btVector3(0,0,0),btVector3(0,0,0));
}
void btQuantizedBvh::reportBoxCastOverlappingNodex(btNodeOverlapCallback* nodeCallback, const btVector3& raySource, const btVector3& rayTarget, const btVector3& aabbMin,const btVector3& aabbMax) const
{
bool fast_path = m_useQuantization && m_traversalMode == TRAVERSAL_STACKLESS;
if (fast_path)
//always use stackless
if (m_useQuantization)
{
walkStacklessQuantizedTreeAgainstRay(nodeCallback, raySource, rayTarget, aabbMin, aabbMax, 0, m_curNodeIndex);
} else {
/* Slow path:
Construct the bounding box for the entire box cast and send that down the tree */
}
else
{
walkStacklessTreeAgainstRay(nodeCallback, raySource, rayTarget, aabbMin, aabbMax, 0, m_curNodeIndex);
}
/*
{
//recursive traversal
btVector3 qaabbMin = raySource;
btVector3 qaabbMax = raySource;
qaabbMin.setMin(rayTarget);
@@ -684,6 +780,8 @@ void btQuantizedBvh::reportBoxCastOverlappingNodex(btNodeOverlapCallback* nodeCa
qaabbMax += aabbMax;
reportAabbOverlappingNodex(nodeCallback,qaabbMin,qaabbMax);
}
*/
}
@@ -743,7 +841,7 @@ unsigned btQuantizedBvh::calculateSerializeBufferSize()
bool btQuantizedBvh::serialize(void *o_alignedDataBuffer, unsigned /*i_dataBufferSize */, bool i_swapEndian)
{
assert(m_subtreeHeaderCount == m_SubtreeHeaders.size());
btAssert(m_subtreeHeaderCount == m_SubtreeHeaders.size());
m_subtreeHeaderCount = m_SubtreeHeaders.size();
/* if (i_dataBufferSize < calculateSerializeBufferSize() || o_alignedDataBuffer == NULL || (((unsigned)o_alignedDataBuffer & BVH_ALIGNMENT_MASK) != 0))