support compound versus compound collision shape acceleration on GPU, using aabb tree versus aabb tree.
Remove constructor from b3Vector3, to make it a POD type, so it can go into a union (and more compatible with OpenCL float4) Use b3MakeVector3 instead of constructor Share some code between C++ and GPU in a shared file: see b3TransformAabb2 in src/Bullet3Collision/BroadPhaseCollision/shared/b3Aabb.h Improve PairBench a bit, show timings and #overlapping pairs. Increase shadowmap default size to 8192x8192 (hope the GPU supports it)
This commit is contained in:
erwincoumans
@@ -6,16 +6,90 @@ static const char* satKernelsCL= \
|
||||
"#define SHAPE_CONCAVE_TRIMESH 5\n"
|
||||
"#define TRIANGLE_NUM_CONVEX_FACES 5\n"
|
||||
"#define SHAPE_COMPOUND_OF_CONVEX_HULLS 6\n"
|
||||
"#define B3_MAX_STACK_DEPTH 256\n"
|
||||
"typedef unsigned int u32;\n"
|
||||
"///keep this in sync with btCollidable.h\n"
|
||||
"typedef struct\n"
|
||||
"{\n"
|
||||
" int m_numChildShapes;\n"
|
||||
" int blaat2;\n"
|
||||
" union {\n"
|
||||
" int m_numChildShapes;\n"
|
||||
" int m_bvhIndex;\n"
|
||||
" };\n"
|
||||
" union\n"
|
||||
" {\n"
|
||||
" float m_radius;\n"
|
||||
" int m_compoundBvhIndex;\n"
|
||||
" };\n"
|
||||
" \n"
|
||||
" int m_shapeType;\n"
|
||||
" int m_shapeIndex;\n"
|
||||
" \n"
|
||||
"} btCollidableGpu;\n"
|
||||
"#define MAX_NUM_PARTS_IN_BITS 10\n"
|
||||
"///b3QuantizedBvhNode is a compressed aabb node, 16 bytes.\n"
|
||||
"///Node can be used for leafnode or internal node. Leafnodes can point to 32-bit triangle index (non-negative range).\n"
|
||||
"typedef struct\n"
|
||||
"{\n"
|
||||
" //12 bytes\n"
|
||||
" unsigned short int m_quantizedAabbMin[3];\n"
|
||||
" unsigned short int m_quantizedAabbMax[3];\n"
|
||||
" //4 bytes\n"
|
||||
" int m_escapeIndexOrTriangleIndex;\n"
|
||||
"} b3QuantizedBvhNode;\n"
|
||||
"typedef struct\n"
|
||||
"{\n"
|
||||
" float4 m_aabbMin;\n"
|
||||
" float4 m_aabbMax;\n"
|
||||
" float4 m_quantization;\n"
|
||||
" int m_numNodes;\n"
|
||||
" int m_numSubTrees;\n"
|
||||
" int m_nodeOffset;\n"
|
||||
" int m_subTreeOffset;\n"
|
||||
"} b3BvhInfo;\n"
|
||||
"int getTriangleIndex(const b3QuantizedBvhNode* rootNode)\n"
|
||||
"{\n"
|
||||
" unsigned int x=0;\n"
|
||||
" unsigned int y = (~(x&0))<<(31-MAX_NUM_PARTS_IN_BITS);\n"
|
||||
" // Get only the lower bits where the triangle index is stored\n"
|
||||
" return (rootNode->m_escapeIndexOrTriangleIndex&~(y));\n"
|
||||
"}\n"
|
||||
"int getTriangleIndexGlobal(__global const b3QuantizedBvhNode* rootNode)\n"
|
||||
"{\n"
|
||||
" unsigned int x=0;\n"
|
||||
" unsigned int y = (~(x&0))<<(31-MAX_NUM_PARTS_IN_BITS);\n"
|
||||
" // Get only the lower bits where the triangle index is stored\n"
|
||||
" return (rootNode->m_escapeIndexOrTriangleIndex&~(y));\n"
|
||||
"}\n"
|
||||
"int isLeafNode(const b3QuantizedBvhNode* rootNode)\n"
|
||||
"{\n"
|
||||
" //skipindex is negative (internal node), triangleindex >=0 (leafnode)\n"
|
||||
" return (rootNode->m_escapeIndexOrTriangleIndex >= 0)? 1 : 0;\n"
|
||||
"}\n"
|
||||
"int isLeafNodeGlobal(__global const b3QuantizedBvhNode* rootNode)\n"
|
||||
"{\n"
|
||||
" //skipindex is negative (internal node), triangleindex >=0 (leafnode)\n"
|
||||
" return (rootNode->m_escapeIndexOrTriangleIndex >= 0)? 1 : 0;\n"
|
||||
"}\n"
|
||||
" \n"
|
||||
"int getEscapeIndex(const b3QuantizedBvhNode* rootNode)\n"
|
||||
"{\n"
|
||||
" return -rootNode->m_escapeIndexOrTriangleIndex;\n"
|
||||
"}\n"
|
||||
"int getEscapeIndexGlobal(__global const b3QuantizedBvhNode* rootNode)\n"
|
||||
"{\n"
|
||||
" return -rootNode->m_escapeIndexOrTriangleIndex;\n"
|
||||
"}\n"
|
||||
"typedef struct\n"
|
||||
"{\n"
|
||||
" //12 bytes\n"
|
||||
" unsigned short int m_quantizedAabbMin[3];\n"
|
||||
" unsigned short int m_quantizedAabbMax[3];\n"
|
||||
" //4 bytes, points to the root of the subtree\n"
|
||||
" int m_rootNodeIndex;\n"
|
||||
" //4 bytes\n"
|
||||
" int m_subtreeSize;\n"
|
||||
" int m_padding[3];\n"
|
||||
"} b3BvhSubtreeInfo;\n"
|
||||
"typedef struct\n"
|
||||
"{\n"
|
||||
" float4 m_childPosition;\n"
|
||||
@@ -67,6 +141,210 @@ static const char* satKernelsCL= \
|
||||
" int m_maxIndices[4];\n"
|
||||
" };\n"
|
||||
"} btAabbCL;\n"
|
||||
"#ifndef B3_AABB_H\n"
|
||||
"#define B3_AABB_H\n"
|
||||
"#ifndef B3_FLOAT4_H\n"
|
||||
"#define B3_FLOAT4_H\n"
|
||||
"#ifndef B3_PLATFORM_DEFINITIONS_H\n"
|
||||
"#define B3_PLATFORM_DEFINITIONS_H\n"
|
||||
"struct MyTest\n"
|
||||
"{\n"
|
||||
" int bla;\n"
|
||||
"};\n"
|
||||
"#ifdef __cplusplus\n"
|
||||
"#else\n"
|
||||
"#define b3AtomicInc atomic_inc\n"
|
||||
"#endif\n"
|
||||
"#endif\n"
|
||||
"#ifdef __cplusplus\n"
|
||||
"#else\n"
|
||||
" typedef float4 b3Float4;\n"
|
||||
" #define b3Float4ConstArg const b3Float4\n"
|
||||
" #define b3MakeFloat4 (float4)\n"
|
||||
" float b3Dot3F4(b3Float4ConstArg v0,b3Float4ConstArg v1)\n"
|
||||
" {\n"
|
||||
" float4 a1 = b3MakeFloat4(v0.xyz,0.f);\n"
|
||||
" float4 b1 = b3MakeFloat4(v1.xyz,0.f);\n"
|
||||
" return dot(a1, b1);\n"
|
||||
" }\n"
|
||||
" b3Float4 b3Cross3(b3Float4ConstArg v0,b3Float4ConstArg v1)\n"
|
||||
" {\n"
|
||||
" float4 a1 = b3MakeFloat4(v0.xyz,0.f);\n"
|
||||
" float4 b1 = b3MakeFloat4(v1.xyz,0.f);\n"
|
||||
" return cross(a1, b1);\n"
|
||||
" }\n"
|
||||
"#endif \n"
|
||||
"#endif //B3_FLOAT4_H\n"
|
||||
"#ifndef B3_MAT3x3_H\n"
|
||||
"#define B3_MAT3x3_H\n"
|
||||
"#ifndef B3_QUAT_H\n"
|
||||
"#define B3_QUAT_H\n"
|
||||
"#ifndef B3_PLATFORM_DEFINITIONS_H\n"
|
||||
"#ifdef __cplusplus\n"
|
||||
"#else\n"
|
||||
"#endif\n"
|
||||
"#endif\n"
|
||||
"#ifndef B3_FLOAT4_H\n"
|
||||
"#ifdef __cplusplus\n"
|
||||
"#else\n"
|
||||
"#endif \n"
|
||||
"#endif //B3_FLOAT4_H\n"
|
||||
"#ifdef __cplusplus\n"
|
||||
"#else\n"
|
||||
" typedef float4 b3Quat;\n"
|
||||
" #define b3QuatConstArg const b3Quat\n"
|
||||
" \n"
|
||||
" \n"
|
||||
"inline float4 b3FastNormalize4(float4 v)\n"
|
||||
"{\n"
|
||||
" v = (float4)(v.xyz,0.f);\n"
|
||||
" return fast_normalize(v);\n"
|
||||
"}\n"
|
||||
" \n"
|
||||
"inline b3Quat b3QuatMul(b3Quat a, b3Quat b);\n"
|
||||
"inline b3Quat b3QuatNormalize(b3QuatConstArg in);\n"
|
||||
"inline b3Quat b3QuatRotate(b3QuatConstArg q, b3QuatConstArg vec);\n"
|
||||
"inline b3Quat b3QuatInvert(b3QuatConstArg q);\n"
|
||||
"inline b3Quat b3QuatMul(b3QuatConstArg a, b3QuatConstArg b)\n"
|
||||
"{\n"
|
||||
" b3Quat ans;\n"
|
||||
" ans = b3Cross3( a, b );\n"
|
||||
" ans += a.w*b+b.w*a;\n"
|
||||
"// ans.w = a.w*b.w - (a.x*b.x+a.y*b.y+a.z*b.z);\n"
|
||||
" ans.w = a.w*b.w - b3Dot3F4(a, b);\n"
|
||||
" return ans;\n"
|
||||
"}\n"
|
||||
"inline b3Quat b3QuatNormalize(b3QuatConstArg in)\n"
|
||||
"{\n"
|
||||
" return b3FastNormalize4(in);\n"
|
||||
"}\n"
|
||||
"inline float4 b3QuatRotate(b3QuatConstArg q, b3QuatConstArg vec)\n"
|
||||
"{\n"
|
||||
" b3Quat qInv = b3QuatInvert( q );\n"
|
||||
" float4 vcpy = vec;\n"
|
||||
" vcpy.w = 0.f;\n"
|
||||
" float4 out = b3QuatMul(b3QuatMul(q,vcpy),qInv);\n"
|
||||
" return out;\n"
|
||||
"}\n"
|
||||
"inline b3Quat b3QuatInvert(b3QuatConstArg q)\n"
|
||||
"{\n"
|
||||
" return (b3Quat)(-q.xyz, q.w);\n"
|
||||
"}\n"
|
||||
"inline float4 b3QuatInvRotate(b3QuatConstArg q, b3QuatConstArg vec)\n"
|
||||
"{\n"
|
||||
" return b3QuatRotate( b3QuatInvert( q ), vec );\n"
|
||||
"}\n"
|
||||
"inline b3Float4 b3TransformPoint(b3Float4ConstArg point, b3Float4ConstArg translation, b3QuatConstArg orientation)\n"
|
||||
"{\n"
|
||||
" return b3QuatRotate( orientation, point ) + (translation);\n"
|
||||
"}\n"
|
||||
" \n"
|
||||
"#endif \n"
|
||||
"#endif //B3_QUAT_H\n"
|
||||
"#ifdef __cplusplus\n"
|
||||
"#else\n"
|
||||
"typedef struct\n"
|
||||
"{\n"
|
||||
" float4 m_row[3];\n"
|
||||
"}b3Mat3x3;\n"
|
||||
"#define b3Mat3x3ConstArg const b3Mat3x3\n"
|
||||
"#define b3GetRow(m,row) (m.m_row[row])\n"
|
||||
"inline b3Mat3x3 b3QuatGetRotationMatrix(b3Quat quat)\n"
|
||||
"{\n"
|
||||
" float4 quat2 = (float4)(quat.x*quat.x, quat.y*quat.y, quat.z*quat.z, 0.f);\n"
|
||||
" b3Mat3x3 out;\n"
|
||||
" out.m_row[0].x=1-2*quat2.y-2*quat2.z;\n"
|
||||
" out.m_row[0].y=2*quat.x*quat.y-2*quat.w*quat.z;\n"
|
||||
" out.m_row[0].z=2*quat.x*quat.z+2*quat.w*quat.y;\n"
|
||||
" out.m_row[0].w = 0.f;\n"
|
||||
" out.m_row[1].x=2*quat.x*quat.y+2*quat.w*quat.z;\n"
|
||||
" out.m_row[1].y=1-2*quat2.x-2*quat2.z;\n"
|
||||
" out.m_row[1].z=2*quat.y*quat.z-2*quat.w*quat.x;\n"
|
||||
" out.m_row[1].w = 0.f;\n"
|
||||
" out.m_row[2].x=2*quat.x*quat.z-2*quat.w*quat.y;\n"
|
||||
" out.m_row[2].y=2*quat.y*quat.z+2*quat.w*quat.x;\n"
|
||||
" out.m_row[2].z=1-2*quat2.x-2*quat2.y;\n"
|
||||
" out.m_row[2].w = 0.f;\n"
|
||||
" return out;\n"
|
||||
"}\n"
|
||||
"inline b3Mat3x3 b3AbsoluteMat3x3(b3Mat3x3ConstArg matIn)\n"
|
||||
"{\n"
|
||||
" b3Mat3x3 out;\n"
|
||||
" out.m_row[0] = fabs(matIn.m_row[0]);\n"
|
||||
" out.m_row[1] = fabs(matIn.m_row[1]);\n"
|
||||
" out.m_row[2] = fabs(matIn.m_row[2]);\n"
|
||||
" return out;\n"
|
||||
"}\n"
|
||||
"#endif\n"
|
||||
"#endif //B3_MAT3x3_H\n"
|
||||
"typedef struct b3Aabb b3Aabb_t;\n"
|
||||
"struct b3Aabb\n"
|
||||
"{\n"
|
||||
" union\n"
|
||||
" {\n"
|
||||
" float m_min[4];\n"
|
||||
" b3Float4 m_minVec;\n"
|
||||
" int m_minIndices[4];\n"
|
||||
" };\n"
|
||||
" union\n"
|
||||
" {\n"
|
||||
" float m_max[4];\n"
|
||||
" b3Float4 m_maxVec;\n"
|
||||
" int m_signedMaxIndices[4];\n"
|
||||
" };\n"
|
||||
"};\n"
|
||||
"inline void b3TransformAabb2(b3Float4ConstArg localAabbMin,b3Float4ConstArg localAabbMax, float margin,\n"
|
||||
" b3Float4ConstArg pos,\n"
|
||||
" b3QuatConstArg orn,\n"
|
||||
" b3Float4* aabbMinOut,b3Float4* aabbMaxOut)\n"
|
||||
"{\n"
|
||||
" b3Float4 localHalfExtents = 0.5f*(localAabbMax-localAabbMin);\n"
|
||||
" localHalfExtents+=b3MakeFloat4(margin,margin,margin,0.f);\n"
|
||||
" b3Float4 localCenter = 0.5f*(localAabbMax+localAabbMin);\n"
|
||||
" b3Mat3x3 m;\n"
|
||||
" m = b3QuatGetRotationMatrix(orn);\n"
|
||||
" b3Mat3x3 abs_b = b3AbsoluteMat3x3(m);\n"
|
||||
" b3Float4 center = b3TransformPoint(localCenter,pos,orn);\n"
|
||||
" \n"
|
||||
" b3Float4 extent = b3MakeFloat4(b3Dot3F4(localHalfExtents,b3GetRow(abs_b,0)),\n"
|
||||
" b3Dot3F4(localHalfExtents,b3GetRow(abs_b,1)),\n"
|
||||
" b3Dot3F4(localHalfExtents,b3GetRow(abs_b,2)),\n"
|
||||
" 0.f);\n"
|
||||
" *aabbMinOut = center-extent;\n"
|
||||
" *aabbMaxOut = center+extent;\n"
|
||||
"}\n"
|
||||
"/// conservative test for overlap between two aabbs\n"
|
||||
"inline bool b3TestAabbAgainstAabb(b3Float4ConstArg aabbMin1,b3Float4ConstArg aabbMax1,\n"
|
||||
" b3Float4ConstArg aabbMin2, b3Float4ConstArg aabbMax2)\n"
|
||||
"{\n"
|
||||
" bool overlap = true;\n"
|
||||
" overlap = (aabbMin1.x > aabbMax2.x || aabbMax1.x < aabbMin2.x) ? false : overlap;\n"
|
||||
" overlap = (aabbMin1.z > aabbMax2.z || aabbMax1.z < aabbMin2.z) ? false : overlap;\n"
|
||||
" overlap = (aabbMin1.y > aabbMax2.y || aabbMax1.y < aabbMin2.y) ? false : overlap;\n"
|
||||
" return overlap;\n"
|
||||
"}\n"
|
||||
"#endif //B3_AABB_H\n"
|
||||
"/*\n"
|
||||
"Bullet Continuous Collision Detection and Physics Library\n"
|
||||
"Copyright (c) 2003-2013 Erwin Coumans http://bulletphysics.org\n"
|
||||
"This software is provided 'as-is', without any express or implied warranty.\n"
|
||||
"In no event will the authors be held liable for any damages arising from the use of this software.\n"
|
||||
"Permission is granted to anyone to use this software for any purpose,\n"
|
||||
"including commercial applications, and to alter it and redistribute it freely,\n"
|
||||
"subject to the following restrictions:\n"
|
||||
"1. The origin of this software must not be misrepresented; you must not claim that you wrote the original software. If you use this software in a product, an acknowledgment in the product documentation would be appreciated but is not required.\n"
|
||||
"2. Altered source versions must be plainly marked as such, and must not be misrepresented as being the original software.\n"
|
||||
"3. This notice may not be removed or altered from any source distribution.\n"
|
||||
"*/\n"
|
||||
"#ifndef B3_INT2_H\n"
|
||||
"#define B3_INT2_H\n"
|
||||
"#ifdef __cplusplus\n"
|
||||
"#else\n"
|
||||
"#define b3UnsignedInt2 uint2\n"
|
||||
"#define b3Int2 int2\n"
|
||||
"#define b3MakeInt2 (int2)\n"
|
||||
"#endif //__cplusplus\n"
|
||||
"#endif\n"
|
||||
"typedef struct\n"
|
||||
"{\n"
|
||||
" float4 m_plane;\n"
|
||||
@@ -650,6 +928,28 @@ static const char* satKernelsCL= \
|
||||
" }\n"
|
||||
" \n"
|
||||
"}\n"
|
||||
"inline b3Float4 MyUnQuantize(const unsigned short* vecIn, b3Float4 quantization, b3Float4 bvhAabbMin)\n"
|
||||
"{\n"
|
||||
" b3Float4 vecOut;\n"
|
||||
" vecOut = b3MakeFloat4(\n"
|
||||
" (float)(vecIn[0]) / (quantization.x),\n"
|
||||
" (float)(vecIn[1]) / (quantization.y),\n"
|
||||
" (float)(vecIn[2]) / (quantization.z),\n"
|
||||
" 0.f);\n"
|
||||
" vecOut += bvhAabbMin;\n"
|
||||
" return vecOut;\n"
|
||||
"}\n"
|
||||
"inline b3Float4 MyUnQuantizeGlobal(__global const unsigned short* vecIn, b3Float4 quantization, b3Float4 bvhAabbMin)\n"
|
||||
"{\n"
|
||||
" b3Float4 vecOut;\n"
|
||||
" vecOut = b3MakeFloat4(\n"
|
||||
" (float)(vecIn[0]) / (quantization.x),\n"
|
||||
" (float)(vecIn[1]) / (quantization.y),\n"
|
||||
" (float)(vecIn[2]) / (quantization.z),\n"
|
||||
" 0.f);\n"
|
||||
" vecOut += bvhAabbMin;\n"
|
||||
" return vecOut;\n"
|
||||
"}\n"
|
||||
"// work-in-progress\n"
|
||||
"__kernel void findCompoundPairsKernel( __global const int4* pairs, \n"
|
||||
" __global const BodyData* rigidBodies, \n"
|
||||
@@ -659,10 +959,13 @@ static const char* satKernelsCL= \
|
||||
" __global const float4* uniqueEdges,\n"
|
||||
" __global const btGpuFace* faces,\n"
|
||||
" __global const int* indices,\n"
|
||||
" __global btAabbCL* aabbs,\n"
|
||||
" __global b3Aabb_t* aabbLocalSpace,\n"
|
||||
" __global const btGpuChildShape* gpuChildShapes,\n"
|
||||
" __global volatile int4* gpuCompoundPairsOut,\n"
|
||||
" __global volatile int* numCompoundPairsOut,\n"
|
||||
" __global const b3BvhSubtreeInfo* subtrees,\n"
|
||||
" __global const b3QuantizedBvhNode* quantizedNodes,\n"
|
||||
" __global const b3BvhInfo* bvhInfos,\n"
|
||||
" int numPairs,\n"
|
||||
" int maxNumCompoundPairsCapacity\n"
|
||||
" )\n"
|
||||
@@ -681,6 +984,131 @@ static const char* satKernelsCL= \
|
||||
" {\n"
|
||||
" return;\n"
|
||||
" }\n"
|
||||
" if ((collidables[collidableIndexA].m_shapeType==SHAPE_COMPOUND_OF_CONVEX_HULLS) &&(collidables[collidableIndexB].m_shapeType==SHAPE_COMPOUND_OF_CONVEX_HULLS))\n"
|
||||
" {\n"
|
||||
" int bvhA = collidables[collidableIndexA].m_compoundBvhIndex;\n"
|
||||
" int bvhB = collidables[collidableIndexB].m_compoundBvhIndex;\n"
|
||||
" int numSubTreesA = bvhInfos[bvhA].m_numSubTrees;\n"
|
||||
" int subTreesOffsetA = bvhInfos[bvhA].m_subTreeOffset;\n"
|
||||
" int subTreesOffsetB = bvhInfos[bvhB].m_subTreeOffset;\n"
|
||||
" int numSubTreesB = bvhInfos[bvhB].m_numSubTrees;\n"
|
||||
" \n"
|
||||
" float4 posA = rigidBodies[bodyIndexA].m_pos;\n"
|
||||
" b3Quat ornA = rigidBodies[bodyIndexA].m_quat;\n"
|
||||
" b3Quat ornB = rigidBodies[bodyIndexB].m_quat;\n"
|
||||
" float4 posB = rigidBodies[bodyIndexB].m_pos;\n"
|
||||
" \n"
|
||||
" for (int p=0;p<numSubTreesA;p++)\n"
|
||||
" {\n"
|
||||
" b3BvhSubtreeInfo subtreeA = subtrees[subTreesOffsetA+p];\n"
|
||||
" //bvhInfos[bvhA].m_quantization\n"
|
||||
" b3Float4 treeAminLocal = MyUnQuantize(subtreeA.m_quantizedAabbMin,bvhInfos[bvhA].m_quantization,bvhInfos[bvhA].m_aabbMin);\n"
|
||||
" b3Float4 treeAmaxLocal = MyUnQuantize(subtreeA.m_quantizedAabbMax,bvhInfos[bvhA].m_quantization,bvhInfos[bvhA].m_aabbMin);\n"
|
||||
" b3Float4 aabbAMinOut,aabbAMaxOut;\n"
|
||||
" float margin=0.f;\n"
|
||||
" b3TransformAabb2(treeAminLocal,treeAmaxLocal, margin,posA,ornA,&aabbAMinOut,&aabbAMaxOut);\n"
|
||||
" \n"
|
||||
" for (int q=0;q<numSubTreesB;q++)\n"
|
||||
" {\n"
|
||||
" b3BvhSubtreeInfo subtreeB = subtrees[subTreesOffsetB+q];\n"
|
||||
" b3Float4 treeBminLocal = MyUnQuantize(subtreeB.m_quantizedAabbMin,bvhInfos[bvhB].m_quantization,bvhInfos[bvhB].m_aabbMin);\n"
|
||||
" b3Float4 treeBmaxLocal = MyUnQuantize(subtreeB.m_quantizedAabbMax,bvhInfos[bvhB].m_quantization,bvhInfos[bvhB].m_aabbMin);\n"
|
||||
" b3Float4 aabbBMinOut,aabbBMaxOut;\n"
|
||||
" float margin=0.f;\n"
|
||||
" b3TransformAabb2(treeBminLocal,treeBmaxLocal, margin,posB,ornB,&aabbBMinOut,&aabbBMaxOut);\n"
|
||||
" \n"
|
||||
" \n"
|
||||
" bool aabbOverlap = b3TestAabbAgainstAabb(aabbAMinOut,aabbAMaxOut,aabbBMinOut,aabbBMaxOut);\n"
|
||||
" if (aabbOverlap)\n"
|
||||
" {\n"
|
||||
" \n"
|
||||
" int startNodeIndexA = subtreeA.m_rootNodeIndex+bvhInfos[bvhA].m_nodeOffset;\n"
|
||||
" int endNodeIndexA = startNodeIndexA+subtreeA.m_subtreeSize;\n"
|
||||
" int startNodeIndexB = subtreeB.m_rootNodeIndex+bvhInfos[bvhB].m_nodeOffset;\n"
|
||||
" int endNodeIndexB = startNodeIndexB+subtreeB.m_subtreeSize;\n"
|
||||
" b3Int2 nodeStack[B3_MAX_STACK_DEPTH];\n"
|
||||
" b3Int2 node0;\n"
|
||||
" node0.x = startNodeIndexA;\n"
|
||||
" node0.y = startNodeIndexB;\n"
|
||||
" int maxStackDepth = B3_MAX_STACK_DEPTH;\n"
|
||||
" int depth=0;\n"
|
||||
" nodeStack[depth++]=node0;\n"
|
||||
" do\n"
|
||||
" {\n"
|
||||
" b3Int2 node = nodeStack[--depth];\n"
|
||||
" b3Float4 aMinLocal = MyUnQuantizeGlobal(quantizedNodes[node.x].m_quantizedAabbMin,bvhInfos[bvhA].m_quantization,bvhInfos[bvhA].m_aabbMin);\n"
|
||||
" b3Float4 aMaxLocal = MyUnQuantizeGlobal(quantizedNodes[node.x].m_quantizedAabbMax,bvhInfos[bvhA].m_quantization,bvhInfos[bvhA].m_aabbMin);\n"
|
||||
" b3Float4 bMinLocal = MyUnQuantizeGlobal(quantizedNodes[node.y].m_quantizedAabbMin,bvhInfos[bvhB].m_quantization,bvhInfos[bvhB].m_aabbMin);\n"
|
||||
" b3Float4 bMaxLocal = MyUnQuantizeGlobal(quantizedNodes[node.y].m_quantizedAabbMax,bvhInfos[bvhB].m_quantization,bvhInfos[bvhB].m_aabbMin);\n"
|
||||
" float margin=0.f;\n"
|
||||
" b3Float4 aabbAMinOut,aabbAMaxOut;\n"
|
||||
" b3TransformAabb2(aMinLocal,aMaxLocal, margin,posA,ornA,&aabbAMinOut,&aabbAMaxOut);\n"
|
||||
" b3Float4 aabbBMinOut,aabbBMaxOut;\n"
|
||||
" b3TransformAabb2(bMinLocal,bMaxLocal, margin,posB,ornB,&aabbBMinOut,&aabbBMaxOut);\n"
|
||||
" \n"
|
||||
" bool nodeOverlap = b3TestAabbAgainstAabb(aabbAMinOut,aabbAMaxOut,aabbBMinOut,aabbBMaxOut);\n"
|
||||
" if (nodeOverlap)\n"
|
||||
" {\n"
|
||||
" bool isLeafA = isLeafNodeGlobal(&quantizedNodes[node.x]);\n"
|
||||
" bool isLeafB = isLeafNodeGlobal(&quantizedNodes[node.y]);\n"
|
||||
" bool isInternalA = !isLeafA;\n"
|
||||
" bool isInternalB = !isLeafB;\n"
|
||||
" //fail, even though it might hit two leaf nodes\n"
|
||||
" if (depth+4>maxStackDepth && !(isLeafA && isLeafB))\n"
|
||||
" {\n"
|
||||
" //printf(\"Error: traversal exceeded maxStackDepth\");\n"
|
||||
" continue;\n"
|
||||
" }\n"
|
||||
" if(isInternalA)\n"
|
||||
" {\n"
|
||||
" int nodeAleftChild = node.x+1;\n"
|
||||
" bool isNodeALeftChildLeaf = isLeafNodeGlobal(&quantizedNodes[node.x+1]);\n"
|
||||
" int nodeArightChild = isNodeALeftChildLeaf? node.x+2 : node.x+1 + getEscapeIndexGlobal(&quantizedNodes[node.x+1]);\n"
|
||||
" if(isInternalB)\n"
|
||||
" { \n"
|
||||
" int nodeBleftChild = node.y+1;\n"
|
||||
" bool isNodeBLeftChildLeaf = isLeafNodeGlobal(&quantizedNodes[node.y+1]);\n"
|
||||
" int nodeBrightChild = isNodeBLeftChildLeaf? node.y+2 : node.y+1 + getEscapeIndexGlobal(&quantizedNodes[node.y+1]);\n"
|
||||
" nodeStack[depth++] = b3MakeInt2(nodeAleftChild, nodeBleftChild);\n"
|
||||
" nodeStack[depth++] = b3MakeInt2(nodeArightChild, nodeBleftChild);\n"
|
||||
" nodeStack[depth++] = b3MakeInt2(nodeAleftChild, nodeBrightChild);\n"
|
||||
" nodeStack[depth++] = b3MakeInt2(nodeArightChild, nodeBrightChild);\n"
|
||||
" }\n"
|
||||
" else\n"
|
||||
" {\n"
|
||||
" nodeStack[depth++] = b3MakeInt2(nodeAleftChild,node.y);\n"
|
||||
" nodeStack[depth++] = b3MakeInt2(nodeArightChild,node.y);\n"
|
||||
" }\n"
|
||||
" }\n"
|
||||
" else\n"
|
||||
" {\n"
|
||||
" if(isInternalB)\n"
|
||||
" {\n"
|
||||
" int nodeBleftChild = node.y+1;\n"
|
||||
" bool isNodeBLeftChildLeaf = isLeafNodeGlobal(&quantizedNodes[node.y+1]);\n"
|
||||
" int nodeBrightChild = isNodeBLeftChildLeaf? node.y+2 : node.y+1 + getEscapeIndexGlobal(&quantizedNodes[node.y+1]);\n"
|
||||
" nodeStack[depth++] = b3MakeInt2(node.x,nodeBleftChild);\n"
|
||||
" nodeStack[depth++] = b3MakeInt2(node.x,nodeBrightChild);\n"
|
||||
" }\n"
|
||||
" else\n"
|
||||
" {\n"
|
||||
" int compoundPairIdx = atomic_inc(numCompoundPairsOut);\n"
|
||||
" if (compoundPairIdx<maxNumCompoundPairsCapacity)\n"
|
||||
" {\n"
|
||||
" int childShapeIndexA = getTriangleIndexGlobal(&quantizedNodes[node.x]);\n"
|
||||
" int childShapeIndexB = getTriangleIndexGlobal(&quantizedNodes[node.y]);\n"
|
||||
" gpuCompoundPairsOut[compoundPairIdx] = (int4)(bodyIndexA,bodyIndexB,childShapeIndexA,childShapeIndexB);\n"
|
||||
" }\n"
|
||||
" }\n"
|
||||
" }\n"
|
||||
" }\n"
|
||||
" } while (depth);\n"
|
||||
" }\n"
|
||||
" }\n"
|
||||
" }\n"
|
||||
" \n"
|
||||
" return;\n"
|
||||
" }\n"
|
||||
" if ((collidables[collidableIndexA].m_shapeType==SHAPE_COMPOUND_OF_CONVEX_HULLS) ||(collidables[collidableIndexB].m_shapeType==SHAPE_COMPOUND_OF_CONVEX_HULLS))\n"
|
||||
" {\n"
|
||||
" if (collidables[collidableIndexA].m_shapeType==SHAPE_COMPOUND_OF_CONVEX_HULLS) \n"
|
||||
@@ -697,6 +1125,18 @@ static const char* satKernelsCL= \
|
||||
" float4 newPosA = qtRotate(ornA,childPosA)+posA;\n"
|
||||
" float4 newOrnA = qtMul(ornA,childOrnA);\n"
|
||||
" int shapeIndexA = collidables[childColIndexA].m_shapeIndex;\n"
|
||||
" b3Aabb_t aabbAlocal = aabbLocalSpace[shapeIndexA];\n"
|
||||
" float margin = 0.f;\n"
|
||||
" \n"
|
||||
" b3Float4 aabbAMinWS;\n"
|
||||
" b3Float4 aabbAMaxWS;\n"
|
||||
" \n"
|
||||
" b3TransformAabb2(aabbAlocal.m_minVec,aabbAlocal.m_maxVec,margin,\n"
|
||||
" newPosA,\n"
|
||||
" newOrnA,\n"
|
||||
" &aabbAMinWS,&aabbAMaxWS);\n"
|
||||
" \n"
|
||||
" \n"
|
||||
" if (collidables[collidableIndexB].m_shapeType==SHAPE_COMPOUND_OF_CONVEX_HULLS)\n"
|
||||
" {\n"
|
||||
" int numChildrenB = collidables[collidableIndexB].m_numChildShapes;\n"
|
||||
@@ -711,7 +1151,20 @@ static const char* satKernelsCL= \
|
||||
" float4 newPosB = transform(&childPosB,&posB,&ornB);\n"
|
||||
" float4 newOrnB = qtMul(ornB,childOrnB);\n"
|
||||
" int shapeIndexB = collidables[childColIndexB].m_shapeIndex;\n"
|
||||
" if (1)\n"
|
||||
" b3Aabb_t aabbBlocal = aabbLocalSpace[shapeIndexB];\n"
|
||||
" \n"
|
||||
" b3Float4 aabbBMinWS;\n"
|
||||
" b3Float4 aabbBMaxWS;\n"
|
||||
" \n"
|
||||
" b3TransformAabb2(aabbBlocal.m_minVec,aabbBlocal.m_maxVec,margin,\n"
|
||||
" newPosB,\n"
|
||||
" newOrnB,\n"
|
||||
" &aabbBMinWS,&aabbBMaxWS);\n"
|
||||
" \n"
|
||||
" \n"
|
||||
" \n"
|
||||
" bool aabbOverlap = b3TestAabbAgainstAabb(aabbAMinWS,aabbAMaxWS,aabbBMinWS,aabbBMaxWS);\n"
|
||||
" if (aabbOverlap)\n"
|
||||
" {\n"
|
||||
" int numFacesA = convexShapes[shapeIndexA].m_numFaces;\n"
|
||||
" float dmin = FLT_MAX;\n"
|
||||
|
||||
Reference in New Issue
Block a user