reorder files, in preparation for Bullet 3 -> Bullet 2 merge
This commit is contained in:
erwincoumans
320
src/Bullet3OpenCL/BroadphaseCollision/kernels/sap.cl
Normal file
320
src/Bullet3OpenCL/BroadphaseCollision/kernels/sap.cl
Normal file
@@ -0,0 +1,320 @@
|
||||
/*
|
||||
Copyright (c) 2012 Advanced Micro Devices, Inc.
|
||||
|
||||
This software is provided 'as-is', without any express or implied warranty.
|
||||
In no event will the authors be held liable for any damages arising from the use of this software.
|
||||
Permission is granted to anyone to use this software for any purpose,
|
||||
including commercial applications, and to alter it and redistribute it freely,
|
||||
subject to the following restrictions:
|
||||
|
||||
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.
|
||||
2. Altered source versions must be plainly marked as such, and must not be misrepresented as being the original software.
|
||||
3. This notice may not be removed or altered from any source distribution.
|
||||
*/
|
||||
//Originally written by Erwin Coumans
|
||||
|
||||
|
||||
typedef struct
|
||||
{
|
||||
union
|
||||
{
|
||||
float4 m_min;
|
||||
float m_minElems[4];
|
||||
int m_minIndices[4];
|
||||
};
|
||||
union
|
||||
{
|
||||
float4 m_max;
|
||||
float m_maxElems[4];
|
||||
int m_maxIndices[4];
|
||||
};
|
||||
} btAabbCL;
|
||||
|
||||
|
||||
/// conservative test for overlap between two aabbs
|
||||
bool TestAabbAgainstAabb2(const btAabbCL* aabb1, __local const btAabbCL* aabb2);
|
||||
bool TestAabbAgainstAabb2(const btAabbCL* aabb1, __local const btAabbCL* aabb2)
|
||||
{
|
||||
bool overlap = true;
|
||||
overlap = (aabb1->m_min.x > aabb2->m_max.x || aabb1->m_max.x < aabb2->m_min.x) ? false : overlap;
|
||||
overlap = (aabb1->m_min.z > aabb2->m_max.z || aabb1->m_max.z < aabb2->m_min.z) ? false : overlap;
|
||||
overlap = (aabb1->m_min.y > aabb2->m_max.y || aabb1->m_max.y < aabb2->m_min.y) ? false : overlap;
|
||||
return overlap;
|
||||
}
|
||||
bool TestAabbAgainstAabb2GlobalGlobal(__global const btAabbCL* aabb1, __global const btAabbCL* aabb2);
|
||||
bool TestAabbAgainstAabb2GlobalGlobal(__global const btAabbCL* aabb1, __global const btAabbCL* aabb2)
|
||||
{
|
||||
bool overlap = true;
|
||||
overlap = (aabb1->m_min.x > aabb2->m_max.x || aabb1->m_max.x < aabb2->m_min.x) ? false : overlap;
|
||||
overlap = (aabb1->m_min.z > aabb2->m_max.z || aabb1->m_max.z < aabb2->m_min.z) ? false : overlap;
|
||||
overlap = (aabb1->m_min.y > aabb2->m_max.y || aabb1->m_max.y < aabb2->m_min.y) ? false : overlap;
|
||||
return overlap;
|
||||
}
|
||||
|
||||
bool TestAabbAgainstAabb2Global(const btAabbCL* aabb1, __global const btAabbCL* aabb2);
|
||||
bool TestAabbAgainstAabb2Global(const btAabbCL* aabb1, __global const btAabbCL* aabb2)
|
||||
{
|
||||
bool overlap = true;
|
||||
overlap = (aabb1->m_min.x > aabb2->m_max.x || aabb1->m_max.x < aabb2->m_min.x) ? false : overlap;
|
||||
overlap = (aabb1->m_min.z > aabb2->m_max.z || aabb1->m_max.z < aabb2->m_min.z) ? false : overlap;
|
||||
overlap = (aabb1->m_min.y > aabb2->m_max.y || aabb1->m_max.y < aabb2->m_min.y) ? false : overlap;
|
||||
return overlap;
|
||||
}
|
||||
|
||||
|
||||
__kernel void computePairsKernelTwoArrays( __global const btAabbCL* unsortedAabbs, __global const btAabbCL* sortedAabbs, volatile __global int2* pairsOut,volatile __global int* pairCount, int numUnsortedAabbs, int numSortedAabbs, int axis, int maxPairs)
|
||||
{
|
||||
int i = get_global_id(0);
|
||||
if (i>=numUnsortedAabbs)
|
||||
return;
|
||||
|
||||
int j = get_global_id(1);
|
||||
if (j>=numSortedAabbs)
|
||||
return;
|
||||
|
||||
if (TestAabbAgainstAabb2GlobalGlobal(&unsortedAabbs[i],&sortedAabbs[j]))
|
||||
{
|
||||
int2 myPair;
|
||||
|
||||
myPair.x = unsortedAabbs[i].m_minIndices[3];
|
||||
myPair.y = sortedAabbs[j].m_minIndices[3];
|
||||
|
||||
int curPair = atomic_inc (pairCount);
|
||||
if (curPair<maxPairs)
|
||||
{
|
||||
pairsOut[curPair] = myPair; //flush to main memory
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
__kernel void computePairsKernelOriginal( __global const btAabbCL* aabbs, volatile __global int2* pairsOut,volatile __global int* pairCount, int numObjects, int axis, int maxPairs)
|
||||
{
|
||||
int i = get_global_id(0);
|
||||
if (i>=numObjects)
|
||||
return;
|
||||
for (int j=i+1;j<numObjects;j++)
|
||||
{
|
||||
if(aabbs[i].m_maxElems[axis] < (aabbs[j].m_minElems[axis]))
|
||||
{
|
||||
break;
|
||||
}
|
||||
if (TestAabbAgainstAabb2GlobalGlobal(&aabbs[i],&aabbs[j]))
|
||||
{
|
||||
int2 myPair;
|
||||
myPair.x = aabbs[i].m_minIndices[3];
|
||||
myPair.y = aabbs[j].m_minIndices[3];
|
||||
int curPair = atomic_inc (pairCount);
|
||||
if (curPair<maxPairs)
|
||||
{
|
||||
pairsOut[curPair] = myPair; //flush to main memory
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
|
||||
|
||||
|
||||
__kernel void computePairsKernelBarrier( __global const btAabbCL* aabbs, volatile __global int2* pairsOut,volatile __global int* pairCount, int numObjects, int axis, int maxPairs)
|
||||
{
|
||||
int i = get_global_id(0);
|
||||
int localId = get_local_id(0);
|
||||
|
||||
__local int numActiveWgItems[1];
|
||||
__local int breakRequest[1];
|
||||
|
||||
if (localId==0)
|
||||
{
|
||||
numActiveWgItems[0] = 0;
|
||||
breakRequest[0] = 0;
|
||||
}
|
||||
barrier(CLK_LOCAL_MEM_FENCE);
|
||||
atomic_inc(numActiveWgItems);
|
||||
barrier(CLK_LOCAL_MEM_FENCE);
|
||||
int localBreak = 0;
|
||||
|
||||
int j=i+1;
|
||||
do
|
||||
{
|
||||
barrier(CLK_LOCAL_MEM_FENCE);
|
||||
|
||||
if (j<numObjects)
|
||||
{
|
||||
if(aabbs[i].m_maxElems[axis] < (aabbs[j].m_minElems[axis]))
|
||||
{
|
||||
if (!localBreak)
|
||||
{
|
||||
atomic_inc(breakRequest);
|
||||
localBreak = 1;
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
barrier(CLK_LOCAL_MEM_FENCE);
|
||||
|
||||
if (j>=numObjects && !localBreak)
|
||||
{
|
||||
atomic_inc(breakRequest);
|
||||
localBreak = 1;
|
||||
}
|
||||
barrier(CLK_LOCAL_MEM_FENCE);
|
||||
|
||||
if (!localBreak)
|
||||
{
|
||||
if (TestAabbAgainstAabb2GlobalGlobal(&aabbs[i],&aabbs[j]))
|
||||
{
|
||||
int2 myPair;
|
||||
myPair.x = aabbs[i].m_minIndices[3];
|
||||
myPair.y = aabbs[j].m_minIndices[3];
|
||||
int curPair = atomic_inc (pairCount);
|
||||
if (curPair<maxPairs)
|
||||
{
|
||||
pairsOut[curPair] = myPair; //flush to main memory
|
||||
}
|
||||
}
|
||||
}
|
||||
j++;
|
||||
|
||||
} while (breakRequest[0]<numActiveWgItems[0]);
|
||||
}
|
||||
|
||||
|
||||
__kernel void computePairsKernelLocalSharedMemory( __global const btAabbCL* aabbs, volatile __global int2* pairsOut,volatile __global int* pairCount, int numObjects, int axis, int maxPairs)
|
||||
{
|
||||
int i = get_global_id(0);
|
||||
int localId = get_local_id(0);
|
||||
|
||||
__local int numActiveWgItems[1];
|
||||
__local int breakRequest[1];
|
||||
__local btAabbCL localAabbs[128];// = aabbs[i];
|
||||
|
||||
btAabbCL myAabb;
|
||||
|
||||
myAabb = (i<numObjects)? aabbs[i]:aabbs[0];
|
||||
float testValue = myAabb.m_maxElems[axis];
|
||||
|
||||
if (localId==0)
|
||||
{
|
||||
numActiveWgItems[0] = 0;
|
||||
breakRequest[0] = 0;
|
||||
}
|
||||
int localCount=0;
|
||||
int block=0;
|
||||
localAabbs[localId] = (i+block)<numObjects? aabbs[i+block] : aabbs[0];
|
||||
localAabbs[localId+64] = (i+block+64)<numObjects? aabbs[i+block+64]: aabbs[0];
|
||||
|
||||
barrier(CLK_LOCAL_MEM_FENCE);
|
||||
atomic_inc(numActiveWgItems);
|
||||
barrier(CLK_LOCAL_MEM_FENCE);
|
||||
int localBreak = 0;
|
||||
|
||||
int j=i+1;
|
||||
do
|
||||
{
|
||||
barrier(CLK_LOCAL_MEM_FENCE);
|
||||
|
||||
if (j<numObjects)
|
||||
{
|
||||
if(testValue < (localAabbs[localCount+localId+1].m_minElems[axis]))
|
||||
{
|
||||
if (!localBreak)
|
||||
{
|
||||
atomic_inc(breakRequest);
|
||||
localBreak = 1;
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
barrier(CLK_LOCAL_MEM_FENCE);
|
||||
|
||||
if (j>=numObjects && !localBreak)
|
||||
{
|
||||
atomic_inc(breakRequest);
|
||||
localBreak = 1;
|
||||
}
|
||||
barrier(CLK_LOCAL_MEM_FENCE);
|
||||
|
||||
if (!localBreak)
|
||||
{
|
||||
if (TestAabbAgainstAabb2(&myAabb,&localAabbs[localCount+localId+1]))
|
||||
{
|
||||
int2 myPair;
|
||||
myPair.x = myAabb.m_minIndices[3];
|
||||
myPair.y = localAabbs[localCount+localId+1].m_minIndices[3];
|
||||
int curPair = atomic_inc (pairCount);
|
||||
if (curPair<maxPairs)
|
||||
{
|
||||
pairsOut[curPair] = myPair; //flush to main memory
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
barrier(CLK_LOCAL_MEM_FENCE);
|
||||
|
||||
localCount++;
|
||||
if (localCount==64)
|
||||
{
|
||||
localCount = 0;
|
||||
block+=64;
|
||||
localAabbs[localId] = ((i+block)<numObjects) ? aabbs[i+block] : aabbs[0];
|
||||
localAabbs[localId+64] = ((i+64+block)<numObjects) ? aabbs[i+block+64] : aabbs[0];
|
||||
}
|
||||
j++;
|
||||
|
||||
} while (breakRequest[0]<numActiveWgItems[0]);
|
||||
|
||||
}
|
||||
|
||||
|
||||
|
||||
|
||||
//http://stereopsis.com/radix.html
|
||||
unsigned int FloatFlip(float fl);
|
||||
unsigned int FloatFlip(float fl)
|
||||
{
|
||||
unsigned int f = *(unsigned int*)&fl;
|
||||
unsigned int mask = -(int)(f >> 31) | 0x80000000;
|
||||
return f ^ mask;
|
||||
}
|
||||
float IFloatFlip(unsigned int f);
|
||||
float IFloatFlip(unsigned int f)
|
||||
{
|
||||
unsigned int mask = ((f >> 31) - 1) | 0x80000000;
|
||||
unsigned int fl = f ^ mask;
|
||||
return *(float*)&fl;
|
||||
}
|
||||
|
||||
|
||||
|
||||
|
||||
__kernel void copyAabbsKernel( __global const btAabbCL* allAabbs, __global btAabbCL* destAabbs, int numObjects)
|
||||
{
|
||||
int i = get_global_id(0);
|
||||
if (i>=numObjects)
|
||||
return;
|
||||
int src = destAabbs[i].m_maxIndices[3];
|
||||
destAabbs[i] = allAabbs[src];
|
||||
destAabbs[i].m_maxIndices[3] = src;
|
||||
}
|
||||
|
||||
|
||||
__kernel void flipFloatKernel( __global const btAabbCL* aabbs, 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;
|
||||
|
||||
}
|
||||
|
||||
|
||||
__kernel void scatterKernel( __global const btAabbCL* aabbs, 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];
|
||||
}
|
||||
161
src/Bullet3OpenCL/BroadphaseCollision/kernels/sapFast.cl
Normal file
161
src/Bullet3OpenCL/BroadphaseCollision/kernels/sapFast.cl
Normal file
@@ -0,0 +1,161 @@
|
||||
/*
|
||||
Copyright (c) 2012 Advanced Micro Devices, Inc.
|
||||
|
||||
This software is provided 'as-is', without any express or implied warranty.
|
||||
In no event will the authors be held liable for any damages arising from the use of this software.
|
||||
Permission is granted to anyone to use this software for any purpose,
|
||||
including commercial applications, and to alter it and redistribute it freely,
|
||||
subject to the following restrictions:
|
||||
|
||||
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.
|
||||
2. Altered source versions must be plainly marked as such, and must not be misrepresented as being the original software.
|
||||
3. This notice may not be removed or altered from any source distribution.
|
||||
*/
|
||||
//Originally written by Erwin Coumans
|
||||
|
||||
|
||||
typedef struct
|
||||
{
|
||||
union
|
||||
{
|
||||
float4 m_min;
|
||||
float m_minElems[4];
|
||||
int m_minIndices[4];
|
||||
};
|
||||
union
|
||||
{
|
||||
float4 m_max;
|
||||
float m_maxElems[4];
|
||||
int m_maxIndices[4];
|
||||
};
|
||||
} btAabbCL;
|
||||
|
||||
|
||||
/// conservative test for overlap between two aabbs
|
||||
bool TestAabbAgainstAabb2(const btAabbCL* aabb1, __local const btAabbCL* aabb2);
|
||||
bool TestAabbAgainstAabb2(const btAabbCL* aabb1, __local const btAabbCL* aabb2)
|
||||
{
|
||||
//skip pairs between static (mass=0) objects
|
||||
if ((aabb1->m_maxIndices[3]==0) && (aabb2->m_maxIndices[3] == 0))
|
||||
return false;
|
||||
|
||||
bool overlap = true;
|
||||
overlap = (aabb1->m_min.x > aabb2->m_max.x || aabb1->m_max.x < aabb2->m_min.x) ? false : overlap;
|
||||
overlap = (aabb1->m_min.z > aabb2->m_max.z || aabb1->m_max.z < aabb2->m_min.z) ? false : overlap;
|
||||
overlap = (aabb1->m_min.y > aabb2->m_max.y || aabb1->m_max.y < aabb2->m_min.y) ? false : overlap;
|
||||
return overlap;
|
||||
}
|
||||
|
||||
|
||||
//computePairsKernelBatchWrite
|
||||
__kernel void computePairsKernel( __global const btAabbCL* aabbs, volatile __global int2* pairsOut,volatile __global int* pairCount, int numObjects, int axis, int maxPairs)
|
||||
{
|
||||
int i = get_global_id(0);
|
||||
int localId = get_local_id(0);
|
||||
|
||||
__local int numActiveWgItems[1];
|
||||
__local int breakRequest[1];
|
||||
__local btAabbCL localAabbs[128];// = aabbs[i];
|
||||
|
||||
int2 myPairs[64];
|
||||
|
||||
btAabbCL myAabb;
|
||||
|
||||
myAabb = (i<numObjects)? aabbs[i]:aabbs[0];
|
||||
float testValue = myAabb.m_maxElems[axis];
|
||||
|
||||
if (localId==0)
|
||||
{
|
||||
numActiveWgItems[0] = 0;
|
||||
breakRequest[0] = 0;
|
||||
}
|
||||
int localCount=0;
|
||||
int block=0;
|
||||
localAabbs[localId] = (i+block)<numObjects? aabbs[i+block] : aabbs[0];
|
||||
localAabbs[localId+64] = (i+block+64)<numObjects? aabbs[i+block+64]: aabbs[0];
|
||||
|
||||
barrier(CLK_LOCAL_MEM_FENCE);
|
||||
atomic_inc(numActiveWgItems);
|
||||
barrier(CLK_LOCAL_MEM_FENCE);
|
||||
int localBreak = 0;
|
||||
int curNumPairs = 0;
|
||||
|
||||
int j=i+1;
|
||||
do
|
||||
{
|
||||
barrier(CLK_LOCAL_MEM_FENCE);
|
||||
|
||||
if (j<numObjects)
|
||||
{
|
||||
if(testValue < (localAabbs[localCount+localId+1].m_minElems[axis]))
|
||||
{
|
||||
if (!localBreak)
|
||||
{
|
||||
atomic_inc(breakRequest);
|
||||
localBreak = 1;
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
barrier(CLK_LOCAL_MEM_FENCE);
|
||||
|
||||
if (j>=numObjects && !localBreak)
|
||||
{
|
||||
atomic_inc(breakRequest);
|
||||
localBreak = 1;
|
||||
}
|
||||
barrier(CLK_LOCAL_MEM_FENCE);
|
||||
|
||||
if (!localBreak)
|
||||
{
|
||||
if (TestAabbAgainstAabb2(&myAabb,&localAabbs[localCount+localId+1]))
|
||||
{
|
||||
int2 myPair;
|
||||
myPair.x = myAabb.m_minIndices[3];
|
||||
myPair.y = localAabbs[localCount+localId+1].m_minIndices[3];
|
||||
myPairs[curNumPairs] = myPair;
|
||||
curNumPairs++;
|
||||
if (curNumPairs==64)
|
||||
{
|
||||
int curPair = atomic_add(pairCount,curNumPairs);
|
||||
//avoid a buffer overrun
|
||||
if ((curPair+curNumPairs)<maxPairs)
|
||||
{
|
||||
for (int p=0;p<curNumPairs;p++)
|
||||
{
|
||||
pairsOut[curPair+p] = myPairs[p]; //flush to main memory
|
||||
}
|
||||
}
|
||||
curNumPairs = 0;
|
||||
}
|
||||
}
|
||||
}
|
||||
barrier(CLK_LOCAL_MEM_FENCE);
|
||||
|
||||
localCount++;
|
||||
if (localCount==64)
|
||||
{
|
||||
localCount = 0;
|
||||
block+=64;
|
||||
localAabbs[localId] = ((i+block)<numObjects) ? aabbs[i+block] : aabbs[0];
|
||||
localAabbs[localId+64] = ((i+64+block)<numObjects) ? aabbs[i+block+64] : aabbs[0];
|
||||
}
|
||||
j++;
|
||||
|
||||
} while (breakRequest[0]<numActiveWgItems[0]);
|
||||
|
||||
|
||||
if (curNumPairs>0)
|
||||
{
|
||||
//avoid a buffer overrun
|
||||
int curPair = atomic_add(pairCount,curNumPairs);
|
||||
if ((curPair+curNumPairs)<maxPairs)
|
||||
{
|
||||
for (int p=0;p<curNumPairs;p++)
|
||||
{
|
||||
pairsOut[curPair+p] = myPairs[p]; //flush to main memory
|
||||
}
|
||||
}
|
||||
curNumPairs = 0;
|
||||
}
|
||||
}
|
||||
164
src/Bullet3OpenCL/BroadphaseCollision/kernels/sapFastKernels.h
Normal file
164
src/Bullet3OpenCL/BroadphaseCollision/kernels/sapFastKernels.h
Normal file
@@ -0,0 +1,164 @@
|
||||
//this file is autogenerated using stringify.bat (premake --stringify) in the build folder of this project
|
||||
static const char* sapFastCL= \
|
||||
"/*\n"
|
||||
"Copyright (c) 2012 Advanced Micro Devices, Inc. \n"
|
||||
"\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"
|
||||
"\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"
|
||||
"//Originally written by Erwin Coumans\n"
|
||||
"\n"
|
||||
"\n"
|
||||
"typedef struct \n"
|
||||
"{\n"
|
||||
" union\n"
|
||||
" {\n"
|
||||
" float4 m_min;\n"
|
||||
" float m_minElems[4];\n"
|
||||
" int m_minIndices[4];\n"
|
||||
" };\n"
|
||||
" union\n"
|
||||
" {\n"
|
||||
" float4 m_max;\n"
|
||||
" float m_maxElems[4];\n"
|
||||
" int m_maxIndices[4];\n"
|
||||
" };\n"
|
||||
"} b3AabbCL;\n"
|
||||
"\n"
|
||||
"\n"
|
||||
"/// conservative test for overlap between two aabbs\n"
|
||||
"bool TestAabbAgainstAabb2(const b3AabbCL* aabb1, __local const b3AabbCL* aabb2);\n"
|
||||
"bool TestAabbAgainstAabb2(const b3AabbCL* aabb1, __local const b3AabbCL* aabb2)\n"
|
||||
"{\n"
|
||||
"//skip pairs between static (mass=0) objects\n"
|
||||
" if ((aabb1->m_maxIndices[3]==0) && (aabb2->m_maxIndices[3] == 0))\n"
|
||||
" return false;\n"
|
||||
" \n"
|
||||
" bool overlap = true;\n"
|
||||
" overlap = (aabb1->m_min.x > aabb2->m_max.x || aabb1->m_max.x < aabb2->m_min.x) ? false : overlap;\n"
|
||||
" overlap = (aabb1->m_min.z > aabb2->m_max.z || aabb1->m_max.z < aabb2->m_min.z) ? false : overlap;\n"
|
||||
" overlap = (aabb1->m_min.y > aabb2->m_max.y || aabb1->m_max.y < aabb2->m_min.y) ? false : overlap;\n"
|
||||
" return overlap;\n"
|
||||
"}\n"
|
||||
"\n"
|
||||
"\n"
|
||||
"//computePairsKernelBatchWrite\n"
|
||||
"__kernel void computePairsKernel( __global const b3AabbCL* aabbs, volatile __global int2* pairsOut,volatile __global int* pairCount, int numObjects, int axis, int maxPairs)\n"
|
||||
"{\n"
|
||||
" int i = get_global_id(0);\n"
|
||||
" int localId = get_local_id(0);\n"
|
||||
"\n"
|
||||
" __local int numActiveWgItems[1];\n"
|
||||
" __local int breakRequest[1];\n"
|
||||
" __local b3AabbCL localAabbs[128];// = aabbs[i];\n"
|
||||
" \n"
|
||||
" int2 myPairs[64];\n"
|
||||
" \n"
|
||||
" b3AabbCL myAabb;\n"
|
||||
" \n"
|
||||
" myAabb = (i<numObjects)? aabbs[i]:aabbs[0];\n"
|
||||
" float testValue = myAabb.m_maxElems[axis];\n"
|
||||
" \n"
|
||||
" if (localId==0)\n"
|
||||
" {\n"
|
||||
" numActiveWgItems[0] = 0;\n"
|
||||
" breakRequest[0] = 0;\n"
|
||||
" }\n"
|
||||
" int localCount=0;\n"
|
||||
" int block=0;\n"
|
||||
" localAabbs[localId] = (i+block)<numObjects? aabbs[i+block] : aabbs[0];\n"
|
||||
" localAabbs[localId+64] = (i+block+64)<numObjects? aabbs[i+block+64]: aabbs[0];\n"
|
||||
" \n"
|
||||
" barrier(CLK_LOCAL_MEM_FENCE);\n"
|
||||
" atomic_inc(numActiveWgItems);\n"
|
||||
" barrier(CLK_LOCAL_MEM_FENCE);\n"
|
||||
" int localBreak = 0;\n"
|
||||
" int curNumPairs = 0;\n"
|
||||
" \n"
|
||||
" int j=i+1;\n"
|
||||
" do\n"
|
||||
" {\n"
|
||||
" barrier(CLK_LOCAL_MEM_FENCE);\n"
|
||||
" \n"
|
||||
" if (j<numObjects)\n"
|
||||
" {\n"
|
||||
" if(testValue < (localAabbs[localCount+localId+1].m_minElems[axis])) \n"
|
||||
" {\n"
|
||||
" if (!localBreak)\n"
|
||||
" {\n"
|
||||
" atomic_inc(breakRequest);\n"
|
||||
" localBreak = 1;\n"
|
||||
" }\n"
|
||||
" }\n"
|
||||
" }\n"
|
||||
" \n"
|
||||
" barrier(CLK_LOCAL_MEM_FENCE);\n"
|
||||
" \n"
|
||||
" if (j>=numObjects && !localBreak)\n"
|
||||
" {\n"
|
||||
" atomic_inc(breakRequest);\n"
|
||||
" localBreak = 1;\n"
|
||||
" }\n"
|
||||
" barrier(CLK_LOCAL_MEM_FENCE);\n"
|
||||
" \n"
|
||||
" if (!localBreak)\n"
|
||||
" {\n"
|
||||
" if (TestAabbAgainstAabb2(&myAabb,&localAabbs[localCount+localId+1]))\n"
|
||||
" {\n"
|
||||
" int2 myPair;\n"
|
||||
" myPair.x = myAabb.m_minIndices[3];\n"
|
||||
" myPair.y = localAabbs[localCount+localId+1].m_minIndices[3];\n"
|
||||
" myPairs[curNumPairs] = myPair;\n"
|
||||
" curNumPairs++;\n"
|
||||
" if (curNumPairs==64)\n"
|
||||
" {\n"
|
||||
" int curPair = atomic_add(pairCount,curNumPairs);\n"
|
||||
" //avoid a buffer overrun\n"
|
||||
" if ((curPair+curNumPairs)<maxPairs)\n"
|
||||
" {\n"
|
||||
" for (int p=0;p<curNumPairs;p++)\n"
|
||||
" {\n"
|
||||
" pairsOut[curPair+p] = myPairs[p]; //flush to main memory\n"
|
||||
" }\n"
|
||||
" }\n"
|
||||
" curNumPairs = 0;\n"
|
||||
" }\n"
|
||||
" }\n"
|
||||
" }\n"
|
||||
" barrier(CLK_LOCAL_MEM_FENCE);\n"
|
||||
" \n"
|
||||
" localCount++;\n"
|
||||
" if (localCount==64)\n"
|
||||
" {\n"
|
||||
" localCount = 0;\n"
|
||||
" block+=64; \n"
|
||||
" localAabbs[localId] = ((i+block)<numObjects) ? aabbs[i+block] : aabbs[0];\n"
|
||||
" localAabbs[localId+64] = ((i+64+block)<numObjects) ? aabbs[i+block+64] : aabbs[0];\n"
|
||||
" }\n"
|
||||
" j++;\n"
|
||||
" \n"
|
||||
" } while (breakRequest[0]<numActiveWgItems[0]);\n"
|
||||
" \n"
|
||||
" \n"
|
||||
" if (curNumPairs>0)\n"
|
||||
" {\n"
|
||||
" //avoid a buffer overrun\n"
|
||||
" int curPair = atomic_add(pairCount,curNumPairs);\n"
|
||||
" if ((curPair+curNumPairs)<maxPairs)\n"
|
||||
" {\n"
|
||||
" for (int p=0;p<curNumPairs;p++)\n"
|
||||
" {\n"
|
||||
" pairsOut[curPair+p] = myPairs[p]; //flush to main memory\n"
|
||||
" }\n"
|
||||
" }\n"
|
||||
" curNumPairs = 0;\n"
|
||||
" }\n"
|
||||
"}\n"
|
||||
;
|
||||
324
src/Bullet3OpenCL/BroadphaseCollision/kernels/sapKernels.h
Normal file
324
src/Bullet3OpenCL/BroadphaseCollision/kernels/sapKernels.h
Normal file
@@ -0,0 +1,324 @@
|
||||
//this file is autogenerated using stringify.bat (premake --stringify) in the build folder of this project
|
||||
static const char* sapCL= \
|
||||
"/*\n"
|
||||
"Copyright (c) 2012 Advanced Micro Devices, Inc. \n"
|
||||
"\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"
|
||||
"\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"
|
||||
"//Originally written by Erwin Coumans\n"
|
||||
"\n"
|
||||
"\n"
|
||||
"typedef struct \n"
|
||||
"{\n"
|
||||
" union\n"
|
||||
" {\n"
|
||||
" float4 m_min;\n"
|
||||
" float m_minElems[4];\n"
|
||||
" int m_minIndices[4];\n"
|
||||
" };\n"
|
||||
" union\n"
|
||||
" {\n"
|
||||
" float4 m_max;\n"
|
||||
" float m_maxElems[4];\n"
|
||||
" int m_maxIndices[4];\n"
|
||||
" };\n"
|
||||
"} b3AabbCL;\n"
|
||||
"\n"
|
||||
"\n"
|
||||
"/// conservative test for overlap between two aabbs\n"
|
||||
"bool TestAabbAgainstAabb2(const b3AabbCL* aabb1, __local const b3AabbCL* aabb2);\n"
|
||||
"bool TestAabbAgainstAabb2(const b3AabbCL* aabb1, __local const b3AabbCL* aabb2)\n"
|
||||
"{\n"
|
||||
" bool overlap = true;\n"
|
||||
" overlap = (aabb1->m_min.x > aabb2->m_max.x || aabb1->m_max.x < aabb2->m_min.x) ? false : overlap;\n"
|
||||
" overlap = (aabb1->m_min.z > aabb2->m_max.z || aabb1->m_max.z < aabb2->m_min.z) ? false : overlap;\n"
|
||||
" overlap = (aabb1->m_min.y > aabb2->m_max.y || aabb1->m_max.y < aabb2->m_min.y) ? false : overlap;\n"
|
||||
" return overlap;\n"
|
||||
"}\n"
|
||||
"bool TestAabbAgainstAabb2GlobalGlobal(__global const b3AabbCL* aabb1, __global const b3AabbCL* aabb2);\n"
|
||||
"bool TestAabbAgainstAabb2GlobalGlobal(__global const b3AabbCL* aabb1, __global const b3AabbCL* aabb2)\n"
|
||||
"{\n"
|
||||
" bool overlap = true;\n"
|
||||
" overlap = (aabb1->m_min.x > aabb2->m_max.x || aabb1->m_max.x < aabb2->m_min.x) ? false : overlap;\n"
|
||||
" overlap = (aabb1->m_min.z > aabb2->m_max.z || aabb1->m_max.z < aabb2->m_min.z) ? false : overlap;\n"
|
||||
" overlap = (aabb1->m_min.y > aabb2->m_max.y || aabb1->m_max.y < aabb2->m_min.y) ? false : overlap;\n"
|
||||
" return overlap;\n"
|
||||
"}\n"
|
||||
"\n"
|
||||
"bool TestAabbAgainstAabb2Global(const b3AabbCL* aabb1, __global const b3AabbCL* aabb2);\n"
|
||||
"bool TestAabbAgainstAabb2Global(const b3AabbCL* aabb1, __global const b3AabbCL* aabb2)\n"
|
||||
"{\n"
|
||||
" bool overlap = true;\n"
|
||||
" overlap = (aabb1->m_min.x > aabb2->m_max.x || aabb1->m_max.x < aabb2->m_min.x) ? false : overlap;\n"
|
||||
" overlap = (aabb1->m_min.z > aabb2->m_max.z || aabb1->m_max.z < aabb2->m_min.z) ? false : overlap;\n"
|
||||
" overlap = (aabb1->m_min.y > aabb2->m_max.y || aabb1->m_max.y < aabb2->m_min.y) ? false : overlap;\n"
|
||||
" return overlap;\n"
|
||||
"}\n"
|
||||
"\n"
|
||||
"\n"
|
||||
"__kernel void computePairsKernelTwoArrays( __global const b3AabbCL* unsortedAabbs, __global const b3AabbCL* sortedAabbs, volatile __global int2* 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"
|
||||
" return;\n"
|
||||
"\n"
|
||||
" int j = get_global_id(1);\n"
|
||||
" if (j>=numSortedAabbs)\n"
|
||||
" return;\n"
|
||||
"\n"
|
||||
" if (TestAabbAgainstAabb2GlobalGlobal(&unsortedAabbs[i],&sortedAabbs[j]))\n"
|
||||
" {\n"
|
||||
" int2 myPair;\n"
|
||||
" \n"
|
||||
" myPair.x = unsortedAabbs[i].m_minIndices[3];\n"
|
||||
" myPair.y = sortedAabbs[j].m_minIndices[3];\n"
|
||||
"\n"
|
||||
" int curPair = atomic_inc (pairCount);\n"
|
||||
" if (curPair<maxPairs)\n"
|
||||
" {\n"
|
||||
" pairsOut[curPair] = myPair; //flush to main memory\n"
|
||||
" }\n"
|
||||
" }\n"
|
||||
"}\n"
|
||||
"\n"
|
||||
"__kernel void computePairsKernelOriginal( __global const b3AabbCL* aabbs, volatile __global int2* pairsOut,volatile __global int* pairCount, int numObjects, int axis, int maxPairs)\n"
|
||||
"{\n"
|
||||
" int i = get_global_id(0);\n"
|
||||
" if (i>=numObjects)\n"
|
||||
" return;\n"
|
||||
" for (int j=i+1;j<numObjects;j++)\n"
|
||||
" {\n"
|
||||
" if(aabbs[i].m_maxElems[axis] < (aabbs[j].m_minElems[axis])) \n"
|
||||
" {\n"
|
||||
" break;\n"
|
||||
" }\n"
|
||||
" if (TestAabbAgainstAabb2GlobalGlobal(&aabbs[i],&aabbs[j]))\n"
|
||||
" {\n"
|
||||
" int2 myPair;\n"
|
||||
" myPair.x = aabbs[i].m_minIndices[3];\n"
|
||||
" myPair.y = aabbs[j].m_minIndices[3];\n"
|
||||
" int curPair = atomic_inc (pairCount);\n"
|
||||
" if (curPair<maxPairs)\n"
|
||||
" {\n"
|
||||
" pairsOut[curPair] = myPair; //flush to main memory\n"
|
||||
" }\n"
|
||||
" }\n"
|
||||
" }\n"
|
||||
"}\n"
|
||||
"\n"
|
||||
"\n"
|
||||
"\n"
|
||||
"\n"
|
||||
"__kernel void computePairsKernelBarrier( __global const b3AabbCL* aabbs, volatile __global int2* pairsOut,volatile __global int* pairCount, int numObjects, int axis, int maxPairs)\n"
|
||||
"{\n"
|
||||
" int i = get_global_id(0);\n"
|
||||
" int localId = get_local_id(0);\n"
|
||||
"\n"
|
||||
" __local int numActiveWgItems[1];\n"
|
||||
" __local int breakRequest[1];\n"
|
||||
"\n"
|
||||
" if (localId==0)\n"
|
||||
" {\n"
|
||||
" numActiveWgItems[0] = 0;\n"
|
||||
" breakRequest[0] = 0;\n"
|
||||
" }\n"
|
||||
" barrier(CLK_LOCAL_MEM_FENCE);\n"
|
||||
" atomic_inc(numActiveWgItems);\n"
|
||||
" barrier(CLK_LOCAL_MEM_FENCE);\n"
|
||||
" int localBreak = 0;\n"
|
||||
"\n"
|
||||
" int j=i+1;\n"
|
||||
" do\n"
|
||||
" {\n"
|
||||
" barrier(CLK_LOCAL_MEM_FENCE);\n"
|
||||
" \n"
|
||||
" if (j<numObjects)\n"
|
||||
" {\n"
|
||||
" if(aabbs[i].m_maxElems[axis] < (aabbs[j].m_minElems[axis])) \n"
|
||||
" {\n"
|
||||
" if (!localBreak)\n"
|
||||
" {\n"
|
||||
" atomic_inc(breakRequest);\n"
|
||||
" localBreak = 1;\n"
|
||||
" }\n"
|
||||
" }\n"
|
||||
" }\n"
|
||||
" \n"
|
||||
" barrier(CLK_LOCAL_MEM_FENCE);\n"
|
||||
" \n"
|
||||
" if (j>=numObjects && !localBreak)\n"
|
||||
" {\n"
|
||||
" atomic_inc(breakRequest);\n"
|
||||
" localBreak = 1;\n"
|
||||
" }\n"
|
||||
" barrier(CLK_LOCAL_MEM_FENCE);\n"
|
||||
" \n"
|
||||
" if (!localBreak)\n"
|
||||
" {\n"
|
||||
" if (TestAabbAgainstAabb2GlobalGlobal(&aabbs[i],&aabbs[j]))\n"
|
||||
" {\n"
|
||||
" int2 myPair;\n"
|
||||
" myPair.x = aabbs[i].m_minIndices[3];\n"
|
||||
" myPair.y = aabbs[j].m_minIndices[3];\n"
|
||||
" int curPair = atomic_inc (pairCount);\n"
|
||||
" if (curPair<maxPairs)\n"
|
||||
" {\n"
|
||||
" pairsOut[curPair] = myPair; //flush to main memory\n"
|
||||
" }\n"
|
||||
" }\n"
|
||||
" }\n"
|
||||
" j++;\n"
|
||||
"\n"
|
||||
" } while (breakRequest[0]<numActiveWgItems[0]);\n"
|
||||
"}\n"
|
||||
"\n"
|
||||
"\n"
|
||||
"__kernel void computePairsKernelLocalSharedMemory( __global const b3AabbCL* aabbs, volatile __global int2* pairsOut,volatile __global int* pairCount, int numObjects, int axis, int maxPairs)\n"
|
||||
"{\n"
|
||||
" int i = get_global_id(0);\n"
|
||||
" int localId = get_local_id(0);\n"
|
||||
"\n"
|
||||
" __local int numActiveWgItems[1];\n"
|
||||
" __local int breakRequest[1];\n"
|
||||
" __local b3AabbCL localAabbs[128];// = aabbs[i];\n"
|
||||
" \n"
|
||||
" b3AabbCL myAabb;\n"
|
||||
" \n"
|
||||
" myAabb = (i<numObjects)? aabbs[i]:aabbs[0];\n"
|
||||
" float testValue = myAabb.m_maxElems[axis];\n"
|
||||
" \n"
|
||||
" if (localId==0)\n"
|
||||
" {\n"
|
||||
" numActiveWgItems[0] = 0;\n"
|
||||
" breakRequest[0] = 0;\n"
|
||||
" }\n"
|
||||
" int localCount=0;\n"
|
||||
" int block=0;\n"
|
||||
" localAabbs[localId] = (i+block)<numObjects? aabbs[i+block] : aabbs[0];\n"
|
||||
" localAabbs[localId+64] = (i+block+64)<numObjects? aabbs[i+block+64]: aabbs[0];\n"
|
||||
" \n"
|
||||
" barrier(CLK_LOCAL_MEM_FENCE);\n"
|
||||
" atomic_inc(numActiveWgItems);\n"
|
||||
" barrier(CLK_LOCAL_MEM_FENCE);\n"
|
||||
" int localBreak = 0;\n"
|
||||
" \n"
|
||||
" int j=i+1;\n"
|
||||
" do\n"
|
||||
" {\n"
|
||||
" barrier(CLK_LOCAL_MEM_FENCE);\n"
|
||||
" \n"
|
||||
" if (j<numObjects)\n"
|
||||
" {\n"
|
||||
" if(testValue < (localAabbs[localCount+localId+1].m_minElems[axis])) \n"
|
||||
" {\n"
|
||||
" if (!localBreak)\n"
|
||||
" {\n"
|
||||
" atomic_inc(breakRequest);\n"
|
||||
" localBreak = 1;\n"
|
||||
" }\n"
|
||||
" }\n"
|
||||
" }\n"
|
||||
" \n"
|
||||
" barrier(CLK_LOCAL_MEM_FENCE);\n"
|
||||
" \n"
|
||||
" if (j>=numObjects && !localBreak)\n"
|
||||
" {\n"
|
||||
" atomic_inc(breakRequest);\n"
|
||||
" localBreak = 1;\n"
|
||||
" }\n"
|
||||
" barrier(CLK_LOCAL_MEM_FENCE);\n"
|
||||
" \n"
|
||||
" if (!localBreak)\n"
|
||||
" {\n"
|
||||
" if (TestAabbAgainstAabb2(&myAabb,&localAabbs[localCount+localId+1]))\n"
|
||||
" {\n"
|
||||
" int2 myPair;\n"
|
||||
" myPair.x = myAabb.m_minIndices[3];\n"
|
||||
" myPair.y = localAabbs[localCount+localId+1].m_minIndices[3];\n"
|
||||
" int curPair = atomic_inc (pairCount);\n"
|
||||
" if (curPair<maxPairs)\n"
|
||||
" {\n"
|
||||
" pairsOut[curPair] = myPair; //flush to main memory\n"
|
||||
" }\n"
|
||||
" }\n"
|
||||
" }\n"
|
||||
" \n"
|
||||
" barrier(CLK_LOCAL_MEM_FENCE);\n"
|
||||
"\n"
|
||||
" localCount++;\n"
|
||||
" if (localCount==64)\n"
|
||||
" {\n"
|
||||
" localCount = 0;\n"
|
||||
" block+=64; \n"
|
||||
" localAabbs[localId] = ((i+block)<numObjects) ? aabbs[i+block] : aabbs[0];\n"
|
||||
" localAabbs[localId+64] = ((i+64+block)<numObjects) ? aabbs[i+block+64] : aabbs[0];\n"
|
||||
" }\n"
|
||||
" j++;\n"
|
||||
" \n"
|
||||
" } while (breakRequest[0]<numActiveWgItems[0]);\n"
|
||||
" \n"
|
||||
"}\n"
|
||||
"\n"
|
||||
"\n"
|
||||
"\n"
|
||||
"\n"
|
||||
"//http://stereopsis.com/radix.html\n"
|
||||
"unsigned int FloatFlip(float fl);\n"
|
||||
"unsigned int FloatFlip(float fl)\n"
|
||||
"{\n"
|
||||
" unsigned int f = *(unsigned int*)&fl;\n"
|
||||
" unsigned int mask = -(int)(f >> 31) | 0x80000000;\n"
|
||||
" return f ^ mask;\n"
|
||||
"}\n"
|
||||
"float IFloatFlip(unsigned int f);\n"
|
||||
"float IFloatFlip(unsigned int f)\n"
|
||||
"{\n"
|
||||
" unsigned int mask = ((f >> 31) - 1) | 0x80000000;\n"
|
||||
" unsigned int fl = f ^ mask;\n"
|
||||
" return *(float*)&fl;\n"
|
||||
"}\n"
|
||||
"\n"
|
||||
"\n"
|
||||
"\n"
|
||||
"\n"
|
||||
"__kernel void copyAabbsKernel( __global const b3AabbCL* allAabbs, __global b3AabbCL* destAabbs, int numObjects)\n"
|
||||
"{\n"
|
||||
" int i = get_global_id(0);\n"
|
||||
" if (i>=numObjects)\n"
|
||||
" return;\n"
|
||||
" int src = destAabbs[i].m_maxIndices[3];\n"
|
||||
" destAabbs[i] = allAabbs[src];\n"
|
||||
" destAabbs[i].m_maxIndices[3] = src;\n"
|
||||
"}\n"
|
||||
"\n"
|
||||
"\n"
|
||||
"__kernel void flipFloatKernel( __global const b3AabbCL* aabbs, 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"
|
||||
"\n"
|
||||
"\n"
|
||||
"__kernel void scatterKernel( __global const b3AabbCL* aabbs, volatile __global const int2* sortData, __global b3AabbCL* sortedAabbs, int numObjects)\n"
|
||||
"{\n"
|
||||
" int i = get_global_id(0);\n"
|
||||
" if (i>=numObjects)\n"
|
||||
" return;\n"
|
||||
"\n"
|
||||
" sortedAabbs[i] = aabbs[sortData[i].y];\n"
|
||||
"}\n"
|
||||
"\n"
|
||||
;
|
||||
Reference in New Issue
Block a user