Bart/bullet3
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity

replace 'copy' step of large/small Aabbs by indices

Browse Source
This commit is contained in:
erwincoumans
2014-01-29 12:52:31 -08:00
parent 33ebebd1c9
commit ff051f87aa
5 changed files with 114 additions and 175 deletions
Download Patch File Download Diff File Expand all files Collapse all files

31
src/Bullet3OpenCL/BroadphaseCollision/kernels/sap.cl
View File

@@ -63,7 +63,7 @@ bool TestAabbAgainstAabb2Global(const btAabbCL* aabb1, __global const btAabbCL*
}
__kernel void computePairsKernelTwoArrays( __global const btAabbCL* unsortedAabbs, __global const btAabbCL* sortedAabbs, volatile __global int4* pairsOut,volatile __global int* pairCount, int numUnsortedAabbs, int numSortedAabbs, int axis, int maxPairs)
__kernel void computePairsKernelTwoArrays( __global const btAabbCL* unsortedAabbs, __global const int* unsortedAabbMapping, __global const btAabbCL* sortedAabbs, volatile __global int4* pairsOut,volatile __global int* pairCount, int numUnsortedAabbs, int numSortedAabbs, int axis, int maxPairs)
{
int i = get_global_id(0);
if (i>=numUnsortedAabbs)
@@ -73,11 +73,14 @@ __kernel void computePairsKernelTwoArrays( __global const btAabbCL* unsortedAa
if (j>=numSortedAabbs)
return;
if (TestAabbAgainstAabb2GlobalGlobal(&unsortedAabbs[i],&sortedAabbs[j]))
__global const btAabbCL* unsortedAabbPtr = &unsortedAabbs[unsortedAabbMapping[i]];
if (TestAabbAgainstAabb2GlobalGlobal(unsortedAabbPtr,&sortedAabbs[j]))
{
int4 myPair;
int xIndex = unsortedAabbs[i].m_minIndices[3];
int xIndex = unsortedAabbPtr[0].m_minIndices[3];
int yIndex = sortedAabbs[j].m_minIndices[3];
if (xIndex>yIndex)
{
@@ -346,36 +349,40 @@ __kernel void copyAabbsKernel( __global const btAabbCL* allAabbs, __global btA
}
__kernel void flipFloatKernel( __global const btAabbCL* aabbs, volatile __global int2* sortData, int numObjects, int axis)
__kernel void flipFloatKernel( __global const btAabbCL* allAabbs, __global const int* smallAabbMapping, volatile __global int2* sortData, int numObjects, int axis)
{
int i = get_global_id(0);
if (i>=numObjects)
return;
sortData[i].x = FloatFlip(aabbs[i].m_minElems[axis]);
sortData[i].y = i;
sortData[i].x = FloatFlip(allAabbs[smallAabbMapping[i]].m_minElems[axis]);
sortData[i].y = i;
}
__kernel void scatterKernel( __global const btAabbCL* aabbs, volatile __global const int2* sortData, __global btAabbCL* sortedAabbs, int numObjects)
__kernel void scatterKernel( __global const btAabbCL* allAabbs, __global const int* smallAabbMapping, volatile __global const int2* sortData, __global btAabbCL* sortedAabbs, int numObjects)
{
int i = get_global_id(0);
if (i>=numObjects)
return;
sortedAabbs[i] = aabbs[sortData[i].y];
sortedAabbs[i] = allAabbs[smallAabbMapping[sortData[i].y]];
}
__kernel void prepareSumVarianceKernel( __global const btAabbCL* aabbs, __global float4* sum, __global float4* sum2,int numAabbs)
__kernel void prepareSumVarianceKernel( __global const btAabbCL* allAabbs, __global const int* smallAabbMapping, __global float4* sum, __global float4* sum2,int numAabbs)
{
int i = get_global_id(0);
if (i>numAabbs)
return;
btAabbCL smallAabb = allAabbs[smallAabbMapping[i]];
float4 s;
s = (aabbs[i].m_max+aabbs[i].m_min)*0.5f;
s = (smallAabb.m_max+smallAabb.m_min)*0.5f;
sum[i]=s;
sum2[i]=s*s;
}

28
src/Bullet3OpenCL/BroadphaseCollision/kernels/sapKernels.h
View File

@@ -56,7 +56,7 @@ static const char* sapCL= \
" overlap = (aabb1->m_min.y > aabb2->m_max.y || aabb1->m_max.y < aabb2->m_min.y) ? false : overlap;\n"
" return overlap;\n"
"}\n"
"__kernel void computePairsKernelTwoArrays( __global const btAabbCL* unsortedAabbs, __global const btAabbCL* sortedAabbs, volatile __global int4* pairsOut,volatile __global int* pairCount, int numUnsortedAabbs, int numSortedAabbs, int axis, int maxPairs)\n"
"__kernel void computePairsKernelTwoArrays( __global const btAabbCL* unsortedAabbs, __global const int* unsortedAabbMapping, __global const btAabbCL* sortedAabbs, volatile __global int4* pairsOut,volatile __global int* pairCount, int numUnsortedAabbs, int numSortedAabbs, int axis, int maxPairs)\n"
"{\n"
" int i = get_global_id(0);\n"
" if (i>=numUnsortedAabbs)\n"
@@ -64,11 +64,12 @@ static const char* sapCL= \
" int j = get_global_id(1);\n"
" if (j>=numSortedAabbs)\n"
" return;\n"
" if (TestAabbAgainstAabb2GlobalGlobal(&unsortedAabbs[i],&sortedAabbs[j]))\n"
" __global const btAabbCL* unsortedAabbPtr = &unsortedAabbs[unsortedAabbMapping[i]];\n"
" if (TestAabbAgainstAabb2GlobalGlobal(unsortedAabbPtr,&sortedAabbs[j]))\n"
" {\n"
" int4 myPair;\n"
" \n"
" int xIndex = unsortedAabbs[i].m_minIndices[3];\n"
" int xIndex = unsortedAabbPtr[0].m_minIndices[3];\n"
" int yIndex = sortedAabbs[j].m_minIndices[3];\n"
" if (xIndex>yIndex)\n"
" {\n"
@@ -305,30 +306,35 @@ static const char* sapCL= \
" destAabbs[i] = allAabbs[src];\n"
" destAabbs[i].m_maxIndices[3] = src;\n"
"}\n"
"__kernel void flipFloatKernel( __global const btAabbCL* aabbs, volatile __global int2* sortData, int numObjects, int axis)\n"
"__kernel void flipFloatKernel( __global const btAabbCL* allAabbs, __global const int* smallAabbMapping, volatile __global int2* sortData, int numObjects, int axis)\n"
"{\n"
" int i = get_global_id(0);\n"
" if (i>=numObjects)\n"
" return;\n"
" \n"
" sortData[i].x = FloatFlip(aabbs[i].m_minElems[axis]);\n"
" sortData[i].y = i;\n"
" \n"
" \n"
" sortData[i].x = FloatFlip(allAabbs[smallAabbMapping[i]].m_minElems[axis]);\n"
" sortData[i].y = i;\n"
" \n"
"}\n"
"__kernel void scatterKernel( __global const btAabbCL* aabbs, volatile __global const int2* sortData, __global btAabbCL* sortedAabbs, int numObjects)\n"
"__kernel void scatterKernel( __global const btAabbCL* allAabbs, __global const int* smallAabbMapping, volatile __global const int2* sortData, __global btAabbCL* sortedAabbs, int numObjects)\n"
"{\n"
" int i = get_global_id(0);\n"
" if (i>=numObjects)\n"
" return;\n"
" sortedAabbs[i] = aabbs[sortData[i].y];\n"
" \n"
" sortedAabbs[i] = allAabbs[smallAabbMapping[sortData[i].y]];\n"
"}\n"
"__kernel void prepareSumVarianceKernel( __global const btAabbCL* aabbs, __global float4* sum, __global float4* sum2,int numAabbs)\n"
"__kernel void prepareSumVarianceKernel( __global const btAabbCL* allAabbs, __global const int* smallAabbMapping, __global float4* sum, __global float4* sum2,int numAabbs)\n"
"{\n"
" int i = get_global_id(0);\n"
" if (i>numAabbs)\n"
" return;\n"
" \n"
" btAabbCL smallAabb = allAabbs[smallAabbMapping[i]];\n"
" \n"
" float4 s;\n"
" s = (aabbs[i].m_max+aabbs[i].m_min)*0.5f;\n"
" s = (smallAabb.m_max+smallAabb.m_min)*0.5f;\n"
" sum[i]=s;\n"
" sum2[i]=s*s; \n"
"}\n"