/lib/include/Bullet/Bullet3OpenCL/ParallelPrimitives/b3OpenCLArray.h
File deleted
Property changes:
Deleted: svn:executable
## -1 +0,0 ##
-*
\ No newline at end of property
Index: include/Bullet/Bullet3OpenCL/ParallelPrimitives/b3LauncherCL.h
===================================================================
--- include/Bullet/Bullet3OpenCL/ParallelPrimitives/b3LauncherCL.h (revision 22)
+++ include/Bullet/Bullet3OpenCL/ParallelPrimitives/b3LauncherCL.h (nonexistent)
@@ -1,135 +0,0 @@
-
-#ifndef B3_LAUNCHER_CL_H
-#define B3_LAUNCHER_CL_H
-
-#include "b3BufferInfoCL.h"
-#include "Bullet3Common/b3MinMax.h"
-#include "b3OpenCLArray.h"
-#include <stdio.h>
-
-#define B3_DEBUG_SERIALIZE_CL
-
-
-#ifdef _WIN32
-#pragma warning(disable :4996)
-#endif
-#define B3_CL_MAX_ARG_SIZE 16
-B3_ATTRIBUTE_ALIGNED16(struct) b3KernelArgData
-{
- int m_isBuffer;
- int m_argIndex;
- int m_argSizeInBytes;
- int m_unusedPadding;
- union
- {
- cl_mem m_clBuffer;
- unsigned char m_argData[B3_CL_MAX_ARG_SIZE];
- };
-
-};
-
-class b3LauncherCL
-{
-
- cl_command_queue m_commandQueue;
- cl_kernel m_kernel;
- int m_idx;
-
- b3AlignedObjectArray<b3KernelArgData> m_kernelArguments;
- int m_serializationSizeInBytes;
- bool m_enableSerialization;
-
- const char* m_name;
- public:
-
- b3AlignedObjectArray<b3OpenCLArray<unsigned char>* > m_arrays;
-
- b3LauncherCL(cl_command_queue queue, cl_kernel kernel, const char* name);
-
- virtual ~b3LauncherCL();
-
- void setBuffer( cl_mem clBuffer);
-
- void setBuffers( b3BufferInfoCL* buffInfo, int n );
-
- int getSerializationBufferSize() const
- {
- return m_serializationSizeInBytes;
- }
-
- int deserializeArgs(unsigned char* buf, int bufSize, cl_context ctx);
-
- inline int validateResults(unsigned char* goldBuffer, int goldBufferCapacity, cl_context ctx);
-
- int serializeArguments(unsigned char* destBuffer, int destBufferCapacity);
-
- int getNumArguments() const
- {
- return m_kernelArguments.size();
- }
-
- b3KernelArgData getArgument(int index)
- {
- return m_kernelArguments[index];
- }
-
- void serializeToFile(const char* fileName, int numWorkItems);
-
- template<typename T>
- inline void setConst( const T& consts )
- {
- int sz=sizeof(T);
- b3Assert(sz<=B3_CL_MAX_ARG_SIZE);
-
- if (m_enableSerialization)
- {
- b3KernelArgData kernelArg;
- kernelArg.m_argIndex = m_idx;
- kernelArg.m_isBuffer = 0;
- T* destArg = (T*)kernelArg.m_argData;
- *destArg = consts;
- kernelArg.m_argSizeInBytes = sizeof(T);
- m_kernelArguments.push_back(kernelArg);
- m_serializationSizeInBytes+=sizeof(b3KernelArgData);
- }
-
- cl_int status = clSetKernelArg( m_kernel, m_idx++, sz, &consts );
- b3Assert( status == CL_SUCCESS );
- }
-
- inline void launch1D( int numThreads, int localSize = 64)
- {
- launch2D( numThreads, 1, localSize, 1 );
- }
-
- inline void launch2D( int numThreadsX, int numThreadsY, int localSizeX, int localSizeY )
- {
- size_t gRange[3] = {1,1,1};
- size_t lRange[3] = {1,1,1};
- lRange[0] = localSizeX;
- lRange[1] = localSizeY;
- gRange[0] = b3Max((size_t)1, (numThreadsX/lRange[0])+(!(numThreadsX%lRange[0])?0:1));
- gRange[0] *= lRange[0];
- gRange[1] = b3Max((size_t)1, (numThreadsY/lRange[1])+(!(numThreadsY%lRange[1])?0:1));
- gRange[1] *= lRange[1];
-
- cl_int status = clEnqueueNDRangeKernel( m_commandQueue,
- m_kernel, 2, NULL, gRange, lRange, 0,0,0 );
- if (status != CL_SUCCESS)
- {
- printf("Error: OpenCL status = %d\n",status);
- }
- b3Assert( status == CL_SUCCESS );
-
- }
-
- void enableSerialization(bool serialize)
- {
- m_enableSerialization = serialize;
- }
-
-};
-
-
-
-#endif //B3_LAUNCHER_CL_H
/include/Bullet/Bullet3OpenCL/ParallelPrimitives/b3LauncherCL.h
Property changes:
Deleted: svn:executable
## -1 +0,0 ##
-*
\ No newline at end of property
Index: include/Bullet/Bullet3OpenCL/ParallelPrimitives/b3PrefixScanCL.h
===================================================================
--- include/Bullet/Bullet3OpenCL/ParallelPrimitives/b3PrefixScanCL.h (revision 22)
+++ include/Bullet/Bullet3OpenCL/ParallelPrimitives/b3PrefixScanCL.h (nonexistent)
@@ -1,37 +0,0 @@
-
-#ifndef B3_PREFIX_SCAN_CL_H
-#define B3_PREFIX_SCAN_CL_H
-
-#include "b3OpenCLArray.h"
-#include "b3BufferInfoCL.h"
-#include "Bullet3Common/b3AlignedObjectArray.h"
-
-class b3PrefixScanCL
-{
- enum
- {
- BLOCK_SIZE = 128
- };
-
-// Option m_option;
-
- cl_command_queue m_commandQueue;
-
- cl_kernel m_localScanKernel;
- cl_kernel m_blockSumKernel;
- cl_kernel m_propagationKernel;
-
- b3OpenCLArray<unsigned int>* m_workBuffer;
-
-
- public:
-
- b3PrefixScanCL(cl_context ctx, cl_device_id device, cl_command_queue queue,int size=0);
-
- virtual ~b3PrefixScanCL();
-
- void execute(b3OpenCLArray<unsigned int>& src, b3OpenCLArray<unsigned int>& dst, int n, unsigned int* sum = 0);
- void executeHost(b3AlignedObjectArray<unsigned int>& src, b3AlignedObjectArray<unsigned int>& dst, int n, unsigned int* sum=0);
-};
-
-#endif //B3_PREFIX_SCAN_CL_H
/include/Bullet/Bullet3OpenCL/ParallelPrimitives/b3PrefixScanCL.h
Property changes:
Deleted: svn:executable
## -1 +0,0 ##
-*
\ No newline at end of property
Index: include/Bullet/Bullet3OpenCL/ParallelPrimitives/b3RadixSort32CL.h
===================================================================
--- include/Bullet/Bullet3OpenCL/ParallelPrimitives/b3RadixSort32CL.h (revision 22)
+++ include/Bullet/Bullet3OpenCL/ParallelPrimitives/b3RadixSort32CL.h (nonexistent)
@@ -1,95 +0,0 @@
-
-#ifndef B3_RADIXSORT32_H
-#define B3_RADIXSORT32_H
-
-#include "b3OpenCLArray.h"
-
-struct b3SortData
-{
- union
- {
- unsigned int m_key;
- unsigned int x;
- };
-
- union
- {
- unsigned int m_value;
- unsigned int y;
-
- };
-};
-#include "b3BufferInfoCL.h"
-
-class b3RadixSort32CL
-{
-
- b3OpenCLArray<unsigned int>* m_workBuffer1;
- b3OpenCLArray<unsigned int>* m_workBuffer2;
-
- b3OpenCLArray<b3SortData>* m_workBuffer3;
- b3OpenCLArray<b3SortData>* m_workBuffer4;
-
- b3OpenCLArray<unsigned int>* m_workBuffer3a;
- b3OpenCLArray<unsigned int>* m_workBuffer4a;
-
- cl_command_queue m_commandQueue;
-
- cl_kernel m_streamCountSortDataKernel;
- cl_kernel m_streamCountKernel;
-
- cl_kernel m_prefixScanKernel;
- cl_kernel m_sortAndScatterSortDataKernel;
- cl_kernel m_sortAndScatterKernel;
-
-
- bool m_deviceCPU;
-
- class b3PrefixScanCL* m_scan;
- class b3FillCL* m_fill;
-
-public:
- struct b3ConstData
- {
- int m_n;
- int m_nWGs;
- int m_startBit;
- int m_nBlocksPerWG;
- };
- enum
- {
- DATA_ALIGNMENT = 256,
- WG_SIZE = 64,
- BLOCK_SIZE = 256,
- ELEMENTS_PER_WORK_ITEM = (BLOCK_SIZE/WG_SIZE),
- BITS_PER_PASS = 4,
- NUM_BUCKET=(1<<BITS_PER_PASS),
- // if you change this, change nPerWI in kernel as well
- NUM_WGS = 20*6, // cypress
-// NUM_WGS = 24*6, // cayman
-// NUM_WGS = 32*4, // nv
- };
-
-
-private:
-
-
-public:
-
- b3RadixSort32CL(cl_context ctx, cl_device_id device, cl_command_queue queue, int initialCapacity =0);
-
- virtual ~b3RadixSort32CL();
-
- void execute(b3OpenCLArray<unsigned int>& keysIn, b3OpenCLArray<unsigned int>& keysOut, b3OpenCLArray<unsigned int>& valuesIn,
- b3OpenCLArray<unsigned int>& valuesOut, int n, int sortBits = 32);
-
- ///keys only
- void execute(b3OpenCLArray<unsigned int>& keysInOut, int sortBits = 32 );
-
- void execute(b3OpenCLArray<b3SortData>& keyValuesInOut, int sortBits = 32 );
- void executeHost(b3OpenCLArray<b3SortData>& keyValuesInOut, int sortBits = 32);
- void executeHost(b3AlignedObjectArray<b3SortData>& keyValuesInOut, int sortBits = 32);
-
-};
-#endif //B3_RADIXSORT32_H
-
/include/Bullet/Bullet3OpenCL/ParallelPrimitives/b3RadixSort32CL.h
Property changes:
Deleted: svn:executable
## -1 +0,0 ##
-*
\ No newline at end of property
Index: include/Bullet/Bullet3OpenCL/ParallelPrimitives/b3BufferInfoCL.h
===================================================================
--- include/Bullet/Bullet3OpenCL/ParallelPrimitives/b3BufferInfoCL.h (revision 22)
+++ include/Bullet/Bullet3OpenCL/ParallelPrimitives/b3BufferInfoCL.h (nonexistent)
@@ -1,19 +0,0 @@
-
-#ifndef B3_BUFFER_INFO_CL_H
-#define B3_BUFFER_INFO_CL_H
-
-#include "b3OpenCLArray.h"
-
-
-struct b3BufferInfoCL
-{
- //b3BufferInfoCL(){}
-
-// template<typename T>
- b3BufferInfoCL(cl_mem buff, bool isReadOnly = false): m_clBuffer(buff), m_isReadOnly(isReadOnly){}
-
- cl_mem m_clBuffer;
- bool m_isReadOnly;
-};
-
-#endif //B3_BUFFER_INFO_CL_H
/include/Bullet/Bullet3OpenCL/ParallelPrimitives/b3BufferInfoCL.h
Property changes:
Deleted: svn:executable
## -1 +0,0 ##
-*
\ No newline at end of property
Index: include/Bullet/Bullet3OpenCL/ParallelPrimitives/b3FillCL.h
===================================================================
--- include/Bullet/Bullet3OpenCL/ParallelPrimitives/b3FillCL.h (revision 22)
+++ include/Bullet/Bullet3OpenCL/ParallelPrimitives/b3FillCL.h (nonexistent)
@@ -1,63 +0,0 @@
-#ifndef B3_FILL_CL_H
-#define B3_FILL_CL_H
-
-#include "b3OpenCLArray.h"
-#include "Bullet3Common/b3Scalar.h"
-
-#include "Bullet3Common/shared/b3Int2.h"
-#include "Bullet3Common/shared/b3Int4.h"
-
-
-class b3FillCL
-{
-
- cl_command_queue m_commandQueue;
-
- cl_kernel m_fillKernelInt2;
- cl_kernel m_fillIntKernel;
- cl_kernel m_fillUnsignedIntKernel;
- cl_kernel m_fillFloatKernel;
-
- public:
-
- struct b3ConstData
- {
- union
- {
- b3Int4 m_data;
- b3UnsignedInt4 m_UnsignedData;
- };
- int m_offset;
- int m_n;
- int m_padding[2];
- };
-
-protected:
-
-public:
-
- b3FillCL(cl_context ctx, cl_device_id device, cl_command_queue queue);
-
- virtual ~b3FillCL();
-
- void execute(b3OpenCLArray<unsigned int>& src, const unsigned int value, int n, int offset = 0);
-
- void execute(b3OpenCLArray<int>& src, const int value, int n, int offset = 0);
-
- void execute(b3OpenCLArray<float>& src, const float value, int n, int offset = 0);
-
- void execute(b3OpenCLArray<b3Int2>& src, const b3Int2& value, int n, int offset = 0);
-
- void executeHost(b3AlignedObjectArray<b3Int2> &src, const b3Int2 &value, int n, int offset);
-
- void executeHost(b3AlignedObjectArray<int> &src, const int value, int n, int offset);
-
- // void execute(b3OpenCLArray<b3Int4>& src, const b3Int4& value, int n, int offset = 0);
-
-};
-
-
-
-
-
-#endif //B3_FILL_CL_H
/include/Bullet/Bullet3OpenCL/ParallelPrimitives/b3FillCL.h
Property changes:
Deleted: svn:executable
## -1 +0,0 ##
-*
\ No newline at end of property
Index: include/Bullet/Bullet3OpenCL/ParallelPrimitives/kernels/PrefixScanKernelsCL.h
===================================================================
--- include/Bullet/Bullet3OpenCL/ParallelPrimitives/kernels/PrefixScanKernelsCL.h (revision 22)
+++ include/Bullet/Bullet3OpenCL/ParallelPrimitives/kernels/PrefixScanKernelsCL.h (nonexistent)
@@ -1,129 +0,0 @@
-//this file is autogenerated using stringify.bat (premake --stringify) in the build folder of this project
-static const char* prefixScanKernelsCL= \
-"/*\n"
-"Copyright (c) 2012 Advanced Micro Devices, Inc. \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"
-"//Originally written by Takahiro Harada\n"
-"typedef unsigned int u32;\n"
-"#define GET_GROUP_IDX get_group_id(0)\n"
-"#define GET_LOCAL_IDX get_local_id(0)\n"
-"#define GET_GLOBAL_IDX get_global_id(0)\n"
-"#define GET_GROUP_SIZE get_local_size(0)\n"
-"#define GROUP_LDS_BARRIER barrier(CLK_LOCAL_MEM_FENCE)\n"
-"// takahiro end\n"
-"#define WG_SIZE 128 \n"
-"#define m_numElems x\n"
-"#define m_numBlocks y\n"
-"#define m_numScanBlocks z\n"
-"/*typedef struct\n"
-"{\n"
-" uint m_numElems;\n"
-" uint m_numBlocks;\n"
-" uint m_numScanBlocks;\n"
-" uint m_padding[1];\n"
-"} ConstBuffer;\n"
-"*/\n"
-"u32 ScanExclusive(__local u32* data, u32 n, int lIdx, int lSize)\n"
-"{\n"
-" u32 blocksum;\n"
-" int offset = 1;\n"
-" for(int nActive=n>>1; nActive>0; nActive>>=1, offset<<=1)\n"
-" {\n"
-" GROUP_LDS_BARRIER;\n"
-" for(int iIdx=lIdx; iIdx<nActive; iIdx+=lSize)\n"
-" {\n"
-" int ai = offset*(2*iIdx+1)-1;\n"
-" int bi = offset*(2*iIdx+2)-1;\n"
-" data[bi] += data[ai];\n"
-" }\n"
-" }\n"
-" GROUP_LDS_BARRIER;\n"
-" if( lIdx == 0 )\n"
-" {\n"
-" blocksum = data[ n-1 ];\n"
-" data[ n-1 ] = 0;\n"
-" }\n"
-" GROUP_LDS_BARRIER;\n"
-" offset >>= 1;\n"
-" for(int nActive=1; nActive<n; nActive<<=1, offset>>=1 )\n"
-" {\n"
-" GROUP_LDS_BARRIER;\n"
-" for( int iIdx = lIdx; iIdx<nActive; iIdx += lSize )\n"
-" {\n"
-" int ai = offset*(2*iIdx+1)-1;\n"
-" int bi = offset*(2*iIdx+2)-1;\n"
-" u32 temp = data[ai];\n"
-" data[ai] = data[bi];\n"
-" data[bi] += temp;\n"
-" }\n"
-" }\n"
-" GROUP_LDS_BARRIER;\n"
-" return blocksum;\n"
-"}\n"
-"__attribute__((reqd_work_group_size(WG_SIZE,1,1)))\n"
-"__kernel\n"
-"void LocalScanKernel(__global u32* dst, __global u32 *src, __global u32 *sumBuffer,\n"
-" uint4 cb)\n"
-"{\n"
-" __local u32 ldsData[WG_SIZE*2];\n"
-" int gIdx = GET_GLOBAL_IDX;\n"
-" int lIdx = GET_LOCAL_IDX;\n"
-" ldsData[2*lIdx] = ( 2*gIdx < cb.m_numElems )? src[2*gIdx]: 0;\n"
-" ldsData[2*lIdx + 1] = ( 2*gIdx+1 < cb.m_numElems )? src[2*gIdx + 1]: 0;\n"
-" u32 sum = ScanExclusive(ldsData, WG_SIZE*2, GET_LOCAL_IDX, GET_GROUP_SIZE);\n"
-" if( lIdx == 0 ) sumBuffer[GET_GROUP_IDX] = sum;\n"
-" if( (2*gIdx) < cb.m_numElems )\n"
-" {\n"
-" dst[2*gIdx] = ldsData[2*lIdx];\n"
-" }\n"
-" if( (2*gIdx + 1) < cb.m_numElems )\n"
-" {\n"
-" dst[2*gIdx + 1] = ldsData[2*lIdx + 1];\n"
-" }\n"
-"}\n"
-"__attribute__((reqd_work_group_size(WG_SIZE,1,1)))\n"
-"__kernel\n"
-"void AddOffsetKernel(__global u32 *dst, __global u32 *blockSum, uint4 cb)\n"
-"{\n"
-" const u32 blockSize = WG_SIZE*2;\n"
-" int myIdx = GET_GROUP_IDX+1;\n"
-" int lIdx = GET_LOCAL_IDX;\n"
-" u32 iBlockSum = blockSum[myIdx];\n"
-" int endValue = min((myIdx+1)*(blockSize), cb.m_numElems);\n"
-" for(int i=myIdx*blockSize+lIdx; i<endValue; i+=GET_GROUP_SIZE)\n"
-" {\n"
-" dst[i] += iBlockSum;\n"
-" }\n"
-"}\n"
-"__attribute__((reqd_work_group_size(WG_SIZE,1,1)))\n"
-"__kernel\n"
-"void TopLevelScanKernel(__global u32* dst, uint4 cb)\n"
-"{\n"
-" __local u32 ldsData[2048];\n"
-" int gIdx = GET_GLOBAL_IDX;\n"
-" int lIdx = GET_LOCAL_IDX;\n"
-" int lSize = GET_GROUP_SIZE;\n"
-" for(int i=lIdx; i<cb.m_numScanBlocks; i+=lSize )\n"
-" {\n"
-" ldsData[i] = (i<cb.m_numBlocks)? dst[i]:0;\n"
-" }\n"
-" GROUP_LDS_BARRIER;\n"
-" u32 sum = ScanExclusive(ldsData, cb.m_numScanBlocks, GET_LOCAL_IDX, GET_GROUP_SIZE);\n"
-" for(int i=lIdx; i<cb.m_numBlocks; i+=lSize )\n"
-" {\n"
-" dst[i] = ldsData[i];\n"
-" }\n"
-" if( gIdx == 0 )\n"
-" {\n"
-" dst[cb.m_numBlocks] = sum;\n"
-" }\n"
-"}\n"
-;
/include/Bullet/Bullet3OpenCL/ParallelPrimitives/kernels/PrefixScanKernelsCL.h
Property changes:
Deleted: svn:executable
## -1 +0,0 ##
-*
\ No newline at end of property
Index: include/Bullet/Bullet3OpenCL/ParallelPrimitives/kernels/RadixSort32KernelsCL.h
===================================================================
--- include/Bullet/Bullet3OpenCL/ParallelPrimitives/kernels/RadixSort32KernelsCL.h (revision 22)
+++ include/Bullet/Bullet3OpenCL/ParallelPrimitives/kernels/RadixSort32KernelsCL.h (nonexistent)
@@ -1,910 +0,0 @@
-//this file is autogenerated using stringify.bat (premake --stringify) in the build folder of this project
-static const char* radixSort32KernelsCL= \
-"/*\n"
-"Bullet Continuous Collision Detection and Physics Library\n"
-"Copyright (c) 2011 Advanced Micro Devices, Inc. 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"
-"//Author Takahiro Harada\n"
-"//#pragma OPENCL EXTENSION cl_amd_printf : enable\n"
-"#pragma OPENCL EXTENSION cl_khr_local_int32_base_atomics : enable\n"
-"#pragma OPENCL EXTENSION cl_khr_global_int32_base_atomics : enable\n"
-"typedef unsigned int u32;\n"
-"#define GET_GROUP_IDX get_group_id(0)\n"
-"#define GET_LOCAL_IDX get_local_id(0)\n"
-"#define GET_GLOBAL_IDX get_global_id(0)\n"
-"#define GET_GROUP_SIZE get_local_size(0)\n"
-"#define GROUP_LDS_BARRIER barrier(CLK_LOCAL_MEM_FENCE)\n"
-"#define GROUP_MEM_FENCE mem_fence(CLK_LOCAL_MEM_FENCE)\n"
-"#define AtomInc(x) atom_inc(&(x))\n"
-"#define AtomInc1(x, out) out = atom_inc(&(x))\n"
-"#define AtomAdd(x, value) atom_add(&(x), value)\n"
-"#define SELECT_UINT4( b, a, condition ) select( b,a,condition )\n"
-"#define make_uint4 (uint4)\n"
-"#define make_uint2 (uint2)\n"
-"#define make_int2 (int2)\n"
-"#define WG_SIZE 64\n"
-"#define ELEMENTS_PER_WORK_ITEM (256/WG_SIZE)\n"
-"#define BITS_PER_PASS 4\n"
-"#define NUM_BUCKET (1<<BITS_PER_PASS)\n"
-"typedef uchar u8;\n"
-"// this isn't optimization for VLIW. But just reducing writes. \n"
-"#define USE_2LEVEL_REDUCE 1\n"
-"//#define CHECK_BOUNDARY 1\n"
-"//#define NV_GPU 1\n"
-"// Cypress\n"
-"#define nPerWI 16\n"
-"// Cayman\n"
-"//#define nPerWI 20\n"
-"#define m_n x\n"
-"#define m_nWGs y\n"
-"#define m_startBit z\n"
-"#define m_nBlocksPerWG w\n"
-"/*\n"
-"typedef struct\n"
-"{\n"
-" int m_n;\n"
-" int m_nWGs;\n"
-" int m_startBit;\n"
-" int m_nBlocksPerWG;\n"
-"} ConstBuffer;\n"
-"*/\n"
-"typedef struct\n"
-"{\n"
-" unsigned int m_key;\n"
-" unsigned int m_value;\n"
-"} SortDataCL;\n"
-"uint prefixScanVectorEx( uint4* data )\n"
-"{\n"
-" u32 sum = 0;\n"
-" u32 tmp = data[0].x;\n"
-" data[0].x = sum;\n"
-" sum += tmp;\n"
-" tmp = data[0].y;\n"
-" data[0].y = sum;\n"
-" sum += tmp;\n"
-" tmp = data[0].z;\n"
-" data[0].z = sum;\n"
-" sum += tmp;\n"
-" tmp = data[0].w;\n"
-" data[0].w = sum;\n"
-" sum += tmp;\n"
-" return sum;\n"
-"}\n"
-"u32 localPrefixSum( u32 pData, uint lIdx, uint* totalSum, __local u32* sorterSharedMemory, int wgSize /*64 or 128*/ )\n"
-"{\n"
-" { // Set data\n"
-" sorterSharedMemory[lIdx] = 0;\n"
-" sorterSharedMemory[lIdx+wgSize] = pData;\n"
-" }\n"
-" GROUP_LDS_BARRIER;\n"
-" { // Prefix sum\n"
-" int idx = 2*lIdx + (wgSize+1);\n"
-"#if defined(USE_2LEVEL_REDUCE)\n"
-" if( lIdx < 64 )\n"
-" {\n"
-" u32 u0, u1, u2;\n"
-" u0 = sorterSharedMemory[idx-3];\n"
-" u1 = sorterSharedMemory[idx-2];\n"
-" u2 = sorterSharedMemory[idx-1];\n"
-" AtomAdd( sorterSharedMemory[idx], u0+u1+u2 ); \n"
-" GROUP_MEM_FENCE;\n"
-" u0 = sorterSharedMemory[idx-12];\n"
-" u1 = sorterSharedMemory[idx-8];\n"
-" u2 = sorterSharedMemory[idx-4];\n"
-" AtomAdd( sorterSharedMemory[idx], u0+u1+u2 ); \n"
-" GROUP_MEM_FENCE;\n"
-" u0 = sorterSharedMemory[idx-48];\n"
-" u1 = sorterSharedMemory[idx-32];\n"
-" u2 = sorterSharedMemory[idx-16];\n"
-" AtomAdd( sorterSharedMemory[idx], u0+u1+u2 ); \n"
-" GROUP_MEM_FENCE;\n"
-" if( wgSize > 64 )\n"
-" {\n"
-" sorterSharedMemory[idx] += sorterSharedMemory[idx-64];\n"
-" GROUP_MEM_FENCE;\n"
-" }\n"
-" sorterSharedMemory[idx-1] += sorterSharedMemory[idx-2];\n"
-" GROUP_MEM_FENCE;\n"
-" }\n"
-"#else\n"
-" if( lIdx < 64 )\n"
-" {\n"
-" sorterSharedMemory[idx] += sorterSharedMemory[idx-1];\n"
-" GROUP_MEM_FENCE;\n"
-" sorterSharedMemory[idx] += sorterSharedMemory[idx-2]; \n"
-" GROUP_MEM_FENCE;\n"
-" sorterSharedMemory[idx] += sorterSharedMemory[idx-4];\n"
-" GROUP_MEM_FENCE;\n"
-" sorterSharedMemory[idx] += sorterSharedMemory[idx-8];\n"
-" GROUP_MEM_FENCE;\n"
-" sorterSharedMemory[idx] += sorterSharedMemory[idx-16];\n"
-" GROUP_MEM_FENCE;\n"
-" sorterSharedMemory[idx] += sorterSharedMemory[idx-32];\n"
-" GROUP_MEM_FENCE;\n"
-" if( wgSize > 64 )\n"
-" {\n"
-" sorterSharedMemory[idx] += sorterSharedMemory[idx-64];\n"
-" GROUP_MEM_FENCE;\n"
-" }\n"
-" sorterSharedMemory[idx-1] += sorterSharedMemory[idx-2];\n"
-" GROUP_MEM_FENCE;\n"
-" }\n"
-"#endif\n"
-" }\n"
-" GROUP_LDS_BARRIER;\n"
-" *totalSum = sorterSharedMemory[wgSize*2-1];\n"
-" u32 addValue = sorterSharedMemory[lIdx+wgSize-1];\n"
-" return addValue;\n"
-"}\n"
-"//__attribute__((reqd_work_group_size(128,1,1)))\n"
-"uint4 localPrefixSum128V( uint4 pData, uint lIdx, uint* totalSum, __local u32* sorterSharedMemory )\n"
-"{\n"
-" u32 s4 = prefixScanVectorEx( &pData );\n"
-" u32 rank = localPrefixSum( s4, lIdx, totalSum, sorterSharedMemory, 128 );\n"
-" return pData + make_uint4( rank, rank, rank, rank );\n"
-"}\n"
-"//__attribute__((reqd_work_group_size(64,1,1)))\n"
-"uint4 localPrefixSum64V( uint4 pData, uint lIdx, uint* totalSum, __local u32* sorterSharedMemory )\n"
-"{\n"
-" u32 s4 = prefixScanVectorEx( &pData );\n"
-" u32 rank = localPrefixSum( s4, lIdx, totalSum, sorterSharedMemory, 64 );\n"
-" return pData + make_uint4( rank, rank, rank, rank );\n"
-"}\n"
-"u32 unpack4Key( u32 key, int keyIdx ){ return (key>>(keyIdx*8)) & 0xff;}\n"
-"u32 bit8Scan(u32 v)\n"
-"{\n"
-" return (v<<8) + (v<<16) + (v<<24);\n"
-"}\n"
-"//===\n"
-"#define MY_HISTOGRAM(idx) localHistogramMat[(idx)*WG_SIZE+lIdx]\n"
-"__kernel\n"
-"__attribute__((reqd_work_group_size(WG_SIZE,1,1)))\n"
-"void StreamCountKernel( __global u32* gSrc, __global u32* histogramOut, int4 cb )\n"
-"{\n"
-" __local u32 localHistogramMat[NUM_BUCKET*WG_SIZE];\n"
-" u32 gIdx = GET_GLOBAL_IDX;\n"
-" u32 lIdx = GET_LOCAL_IDX;\n"
-" u32 wgIdx = GET_GROUP_IDX;\n"
-" u32 wgSize = GET_GROUP_SIZE;\n"
-" const int startBit = cb.m_startBit;\n"
-" const int n = cb.m_n;\n"
-" const int nWGs = cb.m_nWGs;\n"
-" const int nBlocksPerWG = cb.m_nBlocksPerWG;\n"
-" for(int i=0; i<NUM_BUCKET; i++)\n"
-" {\n"
-" MY_HISTOGRAM(i) = 0;\n"
-" }\n"
-" GROUP_LDS_BARRIER;\n"
-" const int blockSize = ELEMENTS_PER_WORK_ITEM*WG_SIZE;\n"
-" u32 localKey;\n"
-" int nBlocks = (n)/blockSize - nBlocksPerWG*wgIdx;\n"
-" int addr = blockSize*nBlocksPerWG*wgIdx + ELEMENTS_PER_WORK_ITEM*lIdx;\n"
-" for(int iblock=0; iblock<min(nBlocksPerWG, nBlocks); iblock++, addr+=blockSize)\n"
-" {\n"
-" // MY_HISTOGRAM( localKeys.x ) ++ is much expensive than atomic add as it requires read and write while atomics can just add on AMD\n"
-" // Using registers didn't perform well. It seems like use localKeys to address requires a lot of alu ops\n"
-" // AMD: AtomInc performs better while NV prefers ++\n"
-" for(int i=0; i<ELEMENTS_PER_WORK_ITEM; i++)\n"
-" {\n"
-"#if defined(CHECK_BOUNDARY)\n"
-" if( addr+i < n )\n"
-"#endif\n"
-" {\n"
-" localKey = (gSrc[addr+i]>>startBit) & 0xf;\n"
-"#if defined(NV_GPU)\n"
-" MY_HISTOGRAM( localKey )++;\n"
-"#else\n"
-" AtomInc( MY_HISTOGRAM( localKey ) );\n"
-"#endif\n"
-" }\n"
-" }\n"
-" }\n"
-" GROUP_LDS_BARRIER;\n"
-" \n"
-" if( lIdx < NUM_BUCKET )\n"
-" {\n"
-" u32 sum = 0;\n"
-" for(int i=0; i<GET_GROUP_SIZE; i++)\n"
-" {\n"
-" sum += localHistogramMat[lIdx*WG_SIZE+(i+lIdx)%GET_GROUP_SIZE];\n"
-" }\n"
-" histogramOut[lIdx*nWGs+wgIdx] = sum;\n"
-" }\n"
-"}\n"
-"__kernel\n"
-"__attribute__((reqd_work_group_size(WG_SIZE,1,1)))\n"
-"void StreamCountSortDataKernel( __global SortDataCL* gSrc, __global u32* histogramOut, int4 cb )\n"
-"{\n"
-" __local u32 localHistogramMat[NUM_BUCKET*WG_SIZE];\n"
-" u32 gIdx = GET_GLOBAL_IDX;\n"
-" u32 lIdx = GET_LOCAL_IDX;\n"
-" u32 wgIdx = GET_GROUP_IDX;\n"
-" u32 wgSize = GET_GROUP_SIZE;\n"
-" const int startBit = cb.m_startBit;\n"
-" const int n = cb.m_n;\n"
-" const int nWGs = cb.m_nWGs;\n"
-" const int nBlocksPerWG = cb.m_nBlocksPerWG;\n"
-" for(int i=0; i<NUM_BUCKET; i++)\n"
-" {\n"
-" MY_HISTOGRAM(i) = 0;\n"
-" }\n"
-" GROUP_LDS_BARRIER;\n"
-" const int blockSize = ELEMENTS_PER_WORK_ITEM*WG_SIZE;\n"
-" u32 localKey;\n"
-" int nBlocks = (n)/blockSize - nBlocksPerWG*wgIdx;\n"
-" int addr = blockSize*nBlocksPerWG*wgIdx + ELEMENTS_PER_WORK_ITEM*lIdx;\n"
-" for(int iblock=0; iblock<min(nBlocksPerWG, nBlocks); iblock++, addr+=blockSize)\n"
-" {\n"
-" // MY_HISTOGRAM( localKeys.x ) ++ is much expensive than atomic add as it requires read and write while atomics can just add on AMD\n"
-" // Using registers didn't perform well. It seems like use localKeys to address requires a lot of alu ops\n"
-" // AMD: AtomInc performs better while NV prefers ++\n"
-" for(int i=0; i<ELEMENTS_PER_WORK_ITEM; i++)\n"
-" {\n"
-"#if defined(CHECK_BOUNDARY)\n"
-" if( addr+i < n )\n"
-"#endif\n"
-" {\n"
-" localKey = (gSrc[addr+i].m_key>>startBit) & 0xf;\n"
-"#if defined(NV_GPU)\n"
-" MY_HISTOGRAM( localKey )++;\n"
-"#else\n"
-" AtomInc( MY_HISTOGRAM( localKey ) );\n"
-"#endif\n"
-" }\n"
-" }\n"
-" }\n"
-" GROUP_LDS_BARRIER;\n"
-" \n"
-" if( lIdx < NUM_BUCKET )\n"
-" {\n"
-" u32 sum = 0;\n"
-" for(int i=0; i<GET_GROUP_SIZE; i++)\n"
-" {\n"
-" sum += localHistogramMat[lIdx*WG_SIZE+(i+lIdx)%GET_GROUP_SIZE];\n"
-" }\n"
-" histogramOut[lIdx*nWGs+wgIdx] = sum;\n"
-" }\n"
-"}\n"
-"#define nPerLane (nPerWI/4)\n"
-"// NUM_BUCKET*nWGs < 128*nPerWI\n"
-"__kernel\n"
-"__attribute__((reqd_work_group_size(128,1,1)))\n"
-"void PrefixScanKernel( __global u32* wHistogram1, int4 cb )\n"
-"{\n"
-" __local u32 ldsTopScanData[128*2];\n"
-" u32 lIdx = GET_LOCAL_IDX;\n"
-" u32 wgIdx = GET_GROUP_IDX;\n"
-" const int nWGs = cb.m_nWGs;\n"
-" u32 data[nPerWI];\n"
-" for(int i=0; i<nPerWI; i++)\n"
-" {\n"
-" data[i] = 0;\n"
-" if( (nPerWI*lIdx+i) < NUM_BUCKET*nWGs )\n"
-" data[i] = wHistogram1[nPerWI*lIdx+i];\n"
-" }\n"
-" uint4 myData = make_uint4(0,0,0,0);\n"
-" for(int i=0; i<nPerLane; i++)\n"
-" {\n"
-" myData.x += data[nPerLane*0+i];\n"
-" myData.y += data[nPerLane*1+i];\n"
-" myData.z += data[nPerLane*2+i];\n"
-" myData.w += data[nPerLane*3+i];\n"
-" }\n"
-" uint totalSum;\n"
-" uint4 scanned = localPrefixSum128V( myData, lIdx, &totalSum, ldsTopScanData );\n"
-"// for(int j=0; j<4; j++) // somehow it introduces a lot of branches\n"
-" { int j = 0;\n"
-" u32 sum = 0;\n"
-" for(int i=0; i<nPerLane; i++)\n"
-" {\n"
-" u32 tmp = data[nPerLane*j+i];\n"
-" data[nPerLane*j+i] = sum;\n"
-" sum += tmp;\n"
-" }\n"
-" }\n"
-" { int j = 1;\n"
-" u32 sum = 0;\n"
-" for(int i=0; i<nPerLane; i++)\n"
-" {\n"
-" u32 tmp = data[nPerLane*j+i];\n"
-" data[nPerLane*j+i] = sum;\n"
-" sum += tmp;\n"
-" }\n"
-" }\n"
-" { int j = 2;\n"
-" u32 sum = 0;\n"
-" for(int i=0; i<nPerLane; i++)\n"
-" {\n"
-" u32 tmp = data[nPerLane*j+i];\n"
-" data[nPerLane*j+i] = sum;\n"
-" sum += tmp;\n"
-" }\n"
-" }\n"
-" { int j = 3;\n"
-" u32 sum = 0;\n"
-" for(int i=0; i<nPerLane; i++)\n"
-" {\n"
-" u32 tmp = data[nPerLane*j+i];\n"
-" data[nPerLane*j+i] = sum;\n"
-" sum += tmp;\n"
-" }\n"
-" }\n"
-" for(int i=0; i<nPerLane; i++)\n"
-" {\n"
-" data[nPerLane*0+i] += scanned.x;\n"
-" data[nPerLane*1+i] += scanned.y;\n"
-" data[nPerLane*2+i] += scanned.z;\n"
-" data[nPerLane*3+i] += scanned.w;\n"
-" }\n"
-" for(int i=0; i<nPerWI; i++)\n"
-" {\n"
-" int index = nPerWI*lIdx+i;\n"
-" if (index < NUM_BUCKET*nWGs)\n"
-" wHistogram1[nPerWI*lIdx+i] = data[i];\n"
-" }\n"
-"}\n"
-"// 4 scan, 4 exchange\n"
-"void sort4Bits(u32 sortData[4], int startBit, int lIdx, __local u32* ldsSortData)\n"
-"{\n"
-" for(int bitIdx=0; bitIdx<BITS_PER_PASS; bitIdx++)\n"
-" {\n"
-" u32 mask = (1<<bitIdx);\n"
-" uint4 cmpResult = make_uint4( (sortData[0]>>startBit) & mask, (sortData[1]>>startBit) & mask, (sortData[2]>>startBit) & mask, (sortData[3]>>startBit) & mask );\n"
-" uint4 prefixSum = SELECT_UINT4( make_uint4(1,1,1,1), make_uint4(0,0,0,0), cmpResult != make_uint4(0,0,0,0) );\n"
-" u32 total;\n"
-" prefixSum = localPrefixSum64V( prefixSum, lIdx, &total, ldsSortData );\n"
-" {\n"
-" uint4 localAddr = make_uint4(lIdx*4+0,lIdx*4+1,lIdx*4+2,lIdx*4+3);\n"
-" uint4 dstAddr = localAddr - prefixSum + make_uint4( total, total, total, total );\n"
-" dstAddr = SELECT_UINT4( prefixSum, dstAddr, cmpResult != make_uint4(0, 0, 0, 0) );\n"
-" GROUP_LDS_BARRIER;\n"
-" ldsSortData[dstAddr.x] = sortData[0];\n"
-" ldsSortData[dstAddr.y] = sortData[1];\n"
-" ldsSortData[dstAddr.z] = sortData[2];\n"
-" ldsSortData[dstAddr.w] = sortData[3];\n"
-" GROUP_LDS_BARRIER;\n"
-" sortData[0] = ldsSortData[localAddr.x];\n"
-" sortData[1] = ldsSortData[localAddr.y];\n"
-" sortData[2] = ldsSortData[localAddr.z];\n"
-" sortData[3] = ldsSortData[localAddr.w];\n"
-" GROUP_LDS_BARRIER;\n"
-" }\n"
-" }\n"
-"}\n"
-"// 2 scan, 2 exchange\n"
-"void sort4Bits1(u32 sortData[4], int startBit, int lIdx, __local u32* ldsSortData)\n"
-"{\n"
-" for(uint ibit=0; ibit<BITS_PER_PASS; ibit+=2)\n"
-" {\n"
-" uint4 b = make_uint4((sortData[0]>>(startBit+ibit)) & 0x3, \n"
-" (sortData[1]>>(startBit+ibit)) & 0x3, \n"
-" (sortData[2]>>(startBit+ibit)) & 0x3, \n"
-" (sortData[3]>>(startBit+ibit)) & 0x3);\n"
-" u32 key4;\n"
-" u32 sKeyPacked[4] = { 0, 0, 0, 0 };\n"
-" {\n"
-" sKeyPacked[0] |= 1<<(8*b.x);\n"
-" sKeyPacked[1] |= 1<<(8*b.y);\n"
-" sKeyPacked[2] |= 1<<(8*b.z);\n"
-" sKeyPacked[3] |= 1<<(8*b.w);\n"
-" key4 = sKeyPacked[0] + sKeyPacked[1] + sKeyPacked[2] + sKeyPacked[3];\n"
-" }\n"
-" u32 rankPacked;\n"
-" u32 sumPacked;\n"
-" {\n"
-" rankPacked = localPrefixSum( key4, lIdx, &sumPacked, ldsSortData, WG_SIZE );\n"
-" }\n"
-" GROUP_LDS_BARRIER;\n"
-" u32 newOffset[4] = { 0,0,0,0 };\n"
-" {\n"
-" u32 sumScanned = bit8Scan( sumPacked );\n"
-" u32 scannedKeys[4];\n"
-" scannedKeys[0] = 1<<(8*b.x);\n"
-" scannedKeys[1] = 1<<(8*b.y);\n"
-" scannedKeys[2] = 1<<(8*b.z);\n"
-" scannedKeys[3] = 1<<(8*b.w);\n"
-" { // 4 scans at once\n"
-" u32 sum4 = 0;\n"
-" for(int ie=0; ie<4; ie++)\n"
-" {\n"
-" u32 tmp = scannedKeys[ie];\n"
-" scannedKeys[ie] = sum4;\n"
-" sum4 += tmp;\n"
-" }\n"
-" }\n"
-" {\n"
-" u32 sumPlusRank = sumScanned + rankPacked;\n"
-" { u32 ie = b.x;\n"
-" scannedKeys[0] += sumPlusRank;\n"
-" newOffset[0] = unpack4Key( scannedKeys[0], ie );\n"
-" }\n"
-" { u32 ie = b.y;\n"
-" scannedKeys[1] += sumPlusRank;\n"
-" newOffset[1] = unpack4Key( scannedKeys[1], ie );\n"
-" }\n"
-" { u32 ie = b.z;\n"
-" scannedKeys[2] += sumPlusRank;\n"
-" newOffset[2] = unpack4Key( scannedKeys[2], ie );\n"
-" }\n"
-" { u32 ie = b.w;\n"
-" scannedKeys[3] += sumPlusRank;\n"
-" newOffset[3] = unpack4Key( scannedKeys[3], ie );\n"
-" }\n"
-" }\n"
-" }\n"
-" GROUP_LDS_BARRIER;\n"
-" {\n"
-" ldsSortData[newOffset[0]] = sortData[0];\n"
-" ldsSortData[newOffset[1]] = sortData[1];\n"
-" ldsSortData[newOffset[2]] = sortData[2];\n"
-" ldsSortData[newOffset[3]] = sortData[3];\n"
-" GROUP_LDS_BARRIER;\n"
-" u32 dstAddr = 4*lIdx;\n"
-" sortData[0] = ldsSortData[dstAddr+0];\n"
-" sortData[1] = ldsSortData[dstAddr+1];\n"
-" sortData[2] = ldsSortData[dstAddr+2];\n"
-" sortData[3] = ldsSortData[dstAddr+3];\n"
-" GROUP_LDS_BARRIER;\n"
-" }\n"
-" }\n"
-"}\n"
-"#define SET_HISTOGRAM(setIdx, key) ldsSortData[(setIdx)*NUM_BUCKET+key]\n"
-"__kernel\n"
-"__attribute__((reqd_work_group_size(WG_SIZE,1,1)))\n"
-"void SortAndScatterKernel( __global const u32* restrict gSrc, __global const u32* rHistogram, __global u32* restrict gDst, int4 cb )\n"
-"{\n"
-" __local u32 ldsSortData[WG_SIZE*ELEMENTS_PER_WORK_ITEM+16];\n"
-" __local u32 localHistogramToCarry[NUM_BUCKET];\n"
-" __local u32 localHistogram[NUM_BUCKET*2];\n"
-" u32 gIdx = GET_GLOBAL_IDX;\n"
-" u32 lIdx = GET_LOCAL_IDX;\n"
-" u32 wgIdx = GET_GROUP_IDX;\n"
-" u32 wgSize = GET_GROUP_SIZE;\n"
-" const int n = cb.m_n;\n"
-" const int nWGs = cb.m_nWGs;\n"
-" const int startBit = cb.m_startBit;\n"
-" const int nBlocksPerWG = cb.m_nBlocksPerWG;\n"
-" if( lIdx < (NUM_BUCKET) )\n"
-" {\n"
-" localHistogramToCarry[lIdx] = rHistogram[lIdx*nWGs + wgIdx];\n"
-" }\n"
-" GROUP_LDS_BARRIER;\n"
-" const int blockSize = ELEMENTS_PER_WORK_ITEM*WG_SIZE;\n"
-" int nBlocks = n/blockSize - nBlocksPerWG*wgIdx;\n"
-" int addr = blockSize*nBlocksPerWG*wgIdx + ELEMENTS_PER_WORK_ITEM*lIdx;\n"
-" for(int iblock=0; iblock<min(nBlocksPerWG, nBlocks); iblock++, addr+=blockSize)\n"
-" {\n"
-" u32 myHistogram = 0;\n"
-" u32 sortData[ELEMENTS_PER_WORK_ITEM];\n"
-" for(int i=0; i<ELEMENTS_PER_WORK_ITEM; i++)\n"
-"#if defined(CHECK_BOUNDARY)\n"
-" sortData[i] = ( addr+i < n )? gSrc[ addr+i ] : 0xffffffff;\n"
-"#else\n"
-" sortData[i] = gSrc[ addr+i ];\n"
-"#endif\n"
-" sort4Bits(sortData, startBit, lIdx, ldsSortData);\n"
-" u32 keys[ELEMENTS_PER_WORK_ITEM];\n"
-" for(int i=0; i<ELEMENTS_PER_WORK_ITEM; i++)\n"
-" keys[i] = (sortData[i]>>startBit) & 0xf;\n"
-" { // create histogram\n"
-" u32 setIdx = lIdx/16;\n"
-" if( lIdx < NUM_BUCKET )\n"
-" {\n"
-" localHistogram[lIdx] = 0;\n"
-" }\n"
-" ldsSortData[lIdx] = 0;\n"
-" GROUP_LDS_BARRIER;\n"
-" for(int i=0; i<ELEMENTS_PER_WORK_ITEM; i++)\n"
-"#if defined(CHECK_BOUNDARY)\n"
-" if( addr+i < n )\n"
-"#endif\n"
-"#if defined(NV_GPU)\n"
-" SET_HISTOGRAM( setIdx, keys[i] )++;\n"
-"#else\n"
-" AtomInc( SET_HISTOGRAM( setIdx, keys[i] ) );\n"
-"#endif\n"
-" \n"
-" GROUP_LDS_BARRIER;\n"
-" \n"
-" uint hIdx = NUM_BUCKET+lIdx;\n"
-" if( lIdx < NUM_BUCKET )\n"
-" {\n"
-" u32 sum = 0;\n"
-" for(int i=0; i<WG_SIZE/16; i++)\n"
-" {\n"
-" sum += SET_HISTOGRAM( i, lIdx );\n"
-" }\n"
-" myHistogram = sum;\n"
-" localHistogram[hIdx] = sum;\n"
-" }\n"
-" GROUP_LDS_BARRIER;\n"
-"#if defined(USE_2LEVEL_REDUCE)\n"
-" if( lIdx < NUM_BUCKET )\n"
-" {\n"
-" localHistogram[hIdx] = localHistogram[hIdx-1];\n"
-" GROUP_MEM_FENCE;\n"
-" u32 u0, u1, u2;\n"
-" u0 = localHistogram[hIdx-3];\n"
-" u1 = localHistogram[hIdx-2];\n"
-" u2 = localHistogram[hIdx-1];\n"
-" AtomAdd( localHistogram[hIdx], u0 + u1 + u2 );\n"
-" GROUP_MEM_FENCE;\n"
-" u0 = localHistogram[hIdx-12];\n"
-" u1 = localHistogram[hIdx-8];\n"
-" u2 = localHistogram[hIdx-4];\n"
-" AtomAdd( localHistogram[hIdx], u0 + u1 + u2 );\n"
-" GROUP_MEM_FENCE;\n"
-" }\n"
-"#else\n"
-" if( lIdx < NUM_BUCKET )\n"
-" {\n"
-" localHistogram[hIdx] = localHistogram[hIdx-1];\n"
-" GROUP_MEM_FENCE;\n"
-" localHistogram[hIdx] += localHistogram[hIdx-1];\n"
-" GROUP_MEM_FENCE;\n"
-" localHistogram[hIdx] += localHistogram[hIdx-2];\n"
-" GROUP_MEM_FENCE;\n"
-" localHistogram[hIdx] += localHistogram[hIdx-4];\n"
-" GROUP_MEM_FENCE;\n"
-" localHistogram[hIdx] += localHistogram[hIdx-8];\n"
-" GROUP_MEM_FENCE;\n"
-" }\n"
-"#endif\n"
-" GROUP_LDS_BARRIER;\n"
-" }\n"
-" {\n"
-" for(int ie=0; ie<ELEMENTS_PER_WORK_ITEM; ie++)\n"
-" {\n"
-" int dataIdx = ELEMENTS_PER_WORK_ITEM*lIdx+ie;\n"
-" int binIdx = keys[ie];\n"
-" int groupOffset = localHistogramToCarry[binIdx];\n"
-" int myIdx = dataIdx - localHistogram[NUM_BUCKET+binIdx];\n"
-"#if defined(CHECK_BOUNDARY)\n"
-" if( addr+ie < n )\n"
-"#endif\n"
-" gDst[ groupOffset + myIdx ] = sortData[ie];\n"
-" }\n"
-" }\n"
-" GROUP_LDS_BARRIER;\n"
-" if( lIdx < NUM_BUCKET )\n"
-" {\n"
-" localHistogramToCarry[lIdx] += myHistogram;\n"
-" }\n"
-" GROUP_LDS_BARRIER;\n"
-" }\n"
-"}\n"
-"// 2 scan, 2 exchange\n"
-"void sort4Bits1KeyValue(u32 sortData[4], int sortVal[4], int startBit, int lIdx, __local u32* ldsSortData, __local int *ldsSortVal)\n"
-"{\n"
-" for(uint ibit=0; ibit<BITS_PER_PASS; ibit+=2)\n"
-" {\n"
-" uint4 b = make_uint4((sortData[0]>>(startBit+ibit)) & 0x3, \n"
-" (sortData[1]>>(startBit+ibit)) & 0x3, \n"
-" (sortData[2]>>(startBit+ibit)) & 0x3, \n"
-" (sortData[3]>>(startBit+ibit)) & 0x3);\n"
-" u32 key4;\n"
-" u32 sKeyPacked[4] = { 0, 0, 0, 0 };\n"
-" {\n"
-" sKeyPacked[0] |= 1<<(8*b.x);\n"
-" sKeyPacked[1] |= 1<<(8*b.y);\n"
-" sKeyPacked[2] |= 1<<(8*b.z);\n"
-" sKeyPacked[3] |= 1<<(8*b.w);\n"
-" key4 = sKeyPacked[0] + sKeyPacked[1] + sKeyPacked[2] + sKeyPacked[3];\n"
-" }\n"
-" u32 rankPacked;\n"
-" u32 sumPacked;\n"
-" {\n"
-" rankPacked = localPrefixSum( key4, lIdx, &sumPacked, ldsSortData, WG_SIZE );\n"
-" }\n"
-" GROUP_LDS_BARRIER;\n"
-" u32 newOffset[4] = { 0,0,0,0 };\n"
-" {\n"
-" u32 sumScanned = bit8Scan( sumPacked );\n"
-" u32 scannedKeys[4];\n"
-" scannedKeys[0] = 1<<(8*b.x);\n"
-" scannedKeys[1] = 1<<(8*b.y);\n"
-" scannedKeys[2] = 1<<(8*b.z);\n"
-" scannedKeys[3] = 1<<(8*b.w);\n"
-" { // 4 scans at once\n"
-" u32 sum4 = 0;\n"
-" for(int ie=0; ie<4; ie++)\n"
-" {\n"
-" u32 tmp = scannedKeys[ie];\n"
-" scannedKeys[ie] = sum4;\n"
-" sum4 += tmp;\n"
-" }\n"
-" }\n"
-" {\n"
-" u32 sumPlusRank = sumScanned + rankPacked;\n"
-" { u32 ie = b.x;\n"
-" scannedKeys[0] += sumPlusRank;\n"
-" newOffset[0] = unpack4Key( scannedKeys[0], ie );\n"
-" }\n"
-" { u32 ie = b.y;\n"
-" scannedKeys[1] += sumPlusRank;\n"
-" newOffset[1] = unpack4Key( scannedKeys[1], ie );\n"
-" }\n"
-" { u32 ie = b.z;\n"
-" scannedKeys[2] += sumPlusRank;\n"
-" newOffset[2] = unpack4Key( scannedKeys[2], ie );\n"
-" }\n"
-" { u32 ie = b.w;\n"
-" scannedKeys[3] += sumPlusRank;\n"
-" newOffset[3] = unpack4Key( scannedKeys[3], ie );\n"
-" }\n"
-" }\n"
-" }\n"
-" GROUP_LDS_BARRIER;\n"
-" {\n"
-" ldsSortData[newOffset[0]] = sortData[0];\n"
-" ldsSortData[newOffset[1]] = sortData[1];\n"
-" ldsSortData[newOffset[2]] = sortData[2];\n"
-" ldsSortData[newOffset[3]] = sortData[3];\n"
-" ldsSortVal[newOffset[0]] = sortVal[0];\n"
-" ldsSortVal[newOffset[1]] = sortVal[1];\n"
-" ldsSortVal[newOffset[2]] = sortVal[2];\n"
-" ldsSortVal[newOffset[3]] = sortVal[3];\n"
-" GROUP_LDS_BARRIER;\n"
-" u32 dstAddr = 4*lIdx;\n"
-" sortData[0] = ldsSortData[dstAddr+0];\n"
-" sortData[1] = ldsSortData[dstAddr+1];\n"
-" sortData[2] = ldsSortData[dstAddr+2];\n"
-" sortData[3] = ldsSortData[dstAddr+3];\n"
-" sortVal[0] = ldsSortVal[dstAddr+0];\n"
-" sortVal[1] = ldsSortVal[dstAddr+1];\n"
-" sortVal[2] = ldsSortVal[dstAddr+2];\n"
-" sortVal[3] = ldsSortVal[dstAddr+3];\n"
-" GROUP_LDS_BARRIER;\n"
-" }\n"
-" }\n"
-"}\n"
-"__kernel\n"
-"__attribute__((reqd_work_group_size(WG_SIZE,1,1)))\n"
-"void SortAndScatterSortDataKernel( __global const SortDataCL* restrict gSrc, __global const u32* rHistogram, __global SortDataCL* restrict gDst, int4 cb)\n"
-"{\n"
-" __local int ldsSortData[WG_SIZE*ELEMENTS_PER_WORK_ITEM+16];\n"
-" __local int ldsSortVal[WG_SIZE*ELEMENTS_PER_WORK_ITEM+16];\n"
-" __local u32 localHistogramToCarry[NUM_BUCKET];\n"
-" __local u32 localHistogram[NUM_BUCKET*2];\n"
-" u32 gIdx = GET_GLOBAL_IDX;\n"
-" u32 lIdx = GET_LOCAL_IDX;\n"
-" u32 wgIdx = GET_GROUP_IDX;\n"
-" u32 wgSize = GET_GROUP_SIZE;\n"
-" const int n = cb.m_n;\n"
-" const int nWGs = cb.m_nWGs;\n"
-" const int startBit = cb.m_startBit;\n"
-" const int nBlocksPerWG = cb.m_nBlocksPerWG;\n"
-" if( lIdx < (NUM_BUCKET) )\n"
-" {\n"
-" localHistogramToCarry[lIdx] = rHistogram[lIdx*nWGs + wgIdx];\n"
-" }\n"
-" GROUP_LDS_BARRIER;\n"
-" \n"
-" const int blockSize = ELEMENTS_PER_WORK_ITEM*WG_SIZE;\n"
-" int nBlocks = n/blockSize - nBlocksPerWG*wgIdx;\n"
-" int addr = blockSize*nBlocksPerWG*wgIdx + ELEMENTS_PER_WORK_ITEM*lIdx;\n"
-" for(int iblock=0; iblock<min(nBlocksPerWG, nBlocks); iblock++, addr+=blockSize)\n"
-" {\n"
-" u32 myHistogram = 0;\n"
-" int sortData[ELEMENTS_PER_WORK_ITEM];\n"
-" int sortVal[ELEMENTS_PER_WORK_ITEM];\n"
-" for(int i=0; i<ELEMENTS_PER_WORK_ITEM; i++)\n"
-"#if defined(CHECK_BOUNDARY)\n"
-" {\n"
-" sortData[i] = ( addr+i < n )? gSrc[ addr+i ].m_key : 0xffffffff;\n"
-" sortVal[i] = ( addr+i < n )? gSrc[ addr+i ].m_value : 0xffffffff;\n"
-" }\n"
-"#else\n"
-" {\n"
-" sortData[i] = gSrc[ addr+i ].m_key;\n"
-" sortVal[i] = gSrc[ addr+i ].m_value;\n"
-" }\n"
-"#endif\n"
-" sort4Bits1KeyValue(sortData, sortVal, startBit, lIdx, ldsSortData, ldsSortVal);\n"
-" u32 keys[ELEMENTS_PER_WORK_ITEM];\n"
-" for(int i=0; i<ELEMENTS_PER_WORK_ITEM; i++)\n"
-" keys[i] = (sortData[i]>>startBit) & 0xf;\n"
-" { // create histogram\n"
-" u32 setIdx = lIdx/16;\n"
-" if( lIdx < NUM_BUCKET )\n"
-" {\n"
-" localHistogram[lIdx] = 0;\n"
-" }\n"
-" ldsSortData[lIdx] = 0;\n"
-" GROUP_LDS_BARRIER;\n"
-" for(int i=0; i<ELEMENTS_PER_WORK_ITEM; i++)\n"
-"#if defined(CHECK_BOUNDARY)\n"
-" if( addr+i < n )\n"
-"#endif\n"
-"#if defined(NV_GPU)\n"
-" SET_HISTOGRAM( setIdx, keys[i] )++;\n"
-"#else\n"
-" AtomInc( SET_HISTOGRAM( setIdx, keys[i] ) );\n"
-"#endif\n"
-" \n"
-" GROUP_LDS_BARRIER;\n"
-" \n"
-" uint hIdx = NUM_BUCKET+lIdx;\n"
-" if( lIdx < NUM_BUCKET )\n"
-" {\n"
-" u32 sum = 0;\n"
-" for(int i=0; i<WG_SIZE/16; i++)\n"
-" {\n"
-" sum += SET_HISTOGRAM( i, lIdx );\n"
-" }\n"
-" myHistogram = sum;\n"
-" localHistogram[hIdx] = sum;\n"
-" }\n"
-" GROUP_LDS_BARRIER;\n"
-"#if defined(USE_2LEVEL_REDUCE)\n"
-" if( lIdx < NUM_BUCKET )\n"
-" {\n"
-" localHistogram[hIdx] = localHistogram[hIdx-1];\n"
-" GROUP_MEM_FENCE;\n"
-" u32 u0, u1, u2;\n"
-" u0 = localHistogram[hIdx-3];\n"
-" u1 = localHistogram[hIdx-2];\n"
-" u2 = localHistogram[hIdx-1];\n"
-" AtomAdd( localHistogram[hIdx], u0 + u1 + u2 );\n"
-" GROUP_MEM_FENCE;\n"
-" u0 = localHistogram[hIdx-12];\n"
-" u1 = localHistogram[hIdx-8];\n"
-" u2 = localHistogram[hIdx-4];\n"
-" AtomAdd( localHistogram[hIdx], u0 + u1 + u2 );\n"
-" GROUP_MEM_FENCE;\n"
-" }\n"
-"#else\n"
-" if( lIdx < NUM_BUCKET )\n"
-" {\n"
-" localHistogram[hIdx] = localHistogram[hIdx-1];\n"
-" GROUP_MEM_FENCE;\n"
-" localHistogram[hIdx] += localHistogram[hIdx-1];\n"
-" GROUP_MEM_FENCE;\n"
-" localHistogram[hIdx] += localHistogram[hIdx-2];\n"
-" GROUP_MEM_FENCE;\n"
-" localHistogram[hIdx] += localHistogram[hIdx-4];\n"
-" GROUP_MEM_FENCE;\n"
-" localHistogram[hIdx] += localHistogram[hIdx-8];\n"
-" GROUP_MEM_FENCE;\n"
-" }\n"
-"#endif\n"
-" GROUP_LDS_BARRIER;\n"
-" }\n"
-" {\n"
-" for(int ie=0; ie<ELEMENTS_PER_WORK_ITEM; ie++)\n"
-" {\n"
-" int dataIdx = ELEMENTS_PER_WORK_ITEM*lIdx+ie;\n"
-" int binIdx = keys[ie];\n"
-" int groupOffset = localHistogramToCarry[binIdx];\n"
-" int myIdx = dataIdx - localHistogram[NUM_BUCKET+binIdx];\n"
-"#if defined(CHECK_BOUNDARY)\n"
-" if( addr+ie < n )\n"
-" {\n"
-" if ((groupOffset + myIdx)<n)\n"
-" {\n"
-" if (sortData[ie]==sortVal[ie])\n"
-" {\n"
-" \n"
-" SortDataCL tmp;\n"
-" tmp.m_key = sortData[ie];\n"
-" tmp.m_value = sortVal[ie];\n"
-" if (tmp.m_key == tmp.m_value)\n"
-" gDst[groupOffset + myIdx ] = tmp;\n"
-" }\n"
-" \n"
-" }\n"
-" }\n"
-"#else\n"
-" if ((groupOffset + myIdx)<n)\n"
-" {\n"
-" gDst[ groupOffset + myIdx ].m_key = sortData[ie];\n"
-" gDst[ groupOffset + myIdx ].m_value = sortVal[ie];\n"
-" }\n"
-"#endif\n"
-" }\n"
-" }\n"
-" GROUP_LDS_BARRIER;\n"
-" if( lIdx < NUM_BUCKET )\n"
-" {\n"
-" localHistogramToCarry[lIdx] += myHistogram;\n"
-" }\n"
-" GROUP_LDS_BARRIER;\n"
-" }\n"
-"}\n"
-"__kernel\n"
-"__attribute__((reqd_work_group_size(WG_SIZE,1,1)))\n"
-"void SortAndScatterSortDataKernelSerial( __global const SortDataCL* restrict gSrc, __global const u32* rHistogram, __global SortDataCL* restrict gDst, int4 cb)\n"
-"{\n"
-" \n"
-" u32 gIdx = GET_GLOBAL_IDX;\n"
-" u32 realLocalIdx = GET_LOCAL_IDX;\n"
-" u32 wgIdx = GET_GROUP_IDX;\n"
-" u32 wgSize = GET_GROUP_SIZE;\n"
-" const int startBit = cb.m_startBit;\n"
-" const int n = cb.m_n;\n"
-" const int nWGs = cb.m_nWGs;\n"
-" const int nBlocksPerWG = cb.m_nBlocksPerWG;\n"
-" int counter[NUM_BUCKET];\n"
-" \n"
-" if (realLocalIdx>0)\n"
-" return;\n"
-" \n"
-" for (int c=0;c<NUM_BUCKET;c++)\n"
-" counter[c]=0;\n"
-" const int blockSize = ELEMENTS_PER_WORK_ITEM*WG_SIZE;\n"
-" \n"
-" int nBlocks = (n)/blockSize - nBlocksPerWG*wgIdx;\n"
-" for(int iblock=0; iblock<min(nBlocksPerWG, nBlocks); iblock++)\n"
-" {\n"
-" for (int lIdx=0;lIdx<WG_SIZE;lIdx++)\n"
-" {\n"
-" int addr2 = iblock*blockSize + blockSize*nBlocksPerWG*wgIdx + ELEMENTS_PER_WORK_ITEM*lIdx;\n"
-" \n"
-" for(int j=0; j<ELEMENTS_PER_WORK_ITEM; j++)\n"
-" {\n"
-" int i = addr2+j;\n"
-" if( i < n )\n"
-" {\n"
-" int tableIdx;\n"
-" tableIdx = (gSrc[i].m_key>>startBit) & 0xf;//0xf = NUM_TABLES-1\n"
-" gDst[rHistogram[tableIdx*nWGs+wgIdx] + counter[tableIdx]] = gSrc[i];\n"
-" counter[tableIdx] ++;\n"
-" }\n"
-" }\n"
-" }\n"
-" }\n"
-" \n"
-"}\n"
-"__kernel\n"
-"__attribute__((reqd_work_group_size(WG_SIZE,1,1)))\n"
-"void SortAndScatterKernelSerial( __global const u32* restrict gSrc, __global const u32* rHistogram, __global u32* restrict gDst, int4 cb )\n"
-"{\n"
-" \n"
-" u32 gIdx = GET_GLOBAL_IDX;\n"
-" u32 realLocalIdx = GET_LOCAL_IDX;\n"
-" u32 wgIdx = GET_GROUP_IDX;\n"
-" u32 wgSize = GET_GROUP_SIZE;\n"
-" const int startBit = cb.m_startBit;\n"
-" const int n = cb.m_n;\n"
-" const int nWGs = cb.m_nWGs;\n"
-" const int nBlocksPerWG = cb.m_nBlocksPerWG;\n"
-" int counter[NUM_BUCKET];\n"
-" \n"
-" if (realLocalIdx>0)\n"
-" return;\n"
-" \n"
-" for (int c=0;c<NUM_BUCKET;c++)\n"
-" counter[c]=0;\n"
-" const int blockSize = ELEMENTS_PER_WORK_ITEM*WG_SIZE;\n"
-" \n"
-" int nBlocks = (n)/blockSize - nBlocksPerWG*wgIdx;\n"
-" for(int iblock=0; iblock<min(nBlocksPerWG, nBlocks); iblock++)\n"
-" {\n"
-" for (int lIdx=0;lIdx<WG_SIZE;lIdx++)\n"
-" {\n"
-" int addr2 = iblock*blockSize + blockSize*nBlocksPerWG*wgIdx + ELEMENTS_PER_WORK_ITEM*lIdx;\n"
-" \n"
-" for(int j=0; j<ELEMENTS_PER_WORK_ITEM; j++)\n"
-" {\n"
-" int i = addr2+j;\n"
-" if( i < n )\n"
-" {\n"
-" int tableIdx;\n"
-" tableIdx = (gSrc[i]>>startBit) & 0xf;//0xf = NUM_TABLES-1\n"
-" gDst[rHistogram[tableIdx*nWGs+wgIdx] + counter[tableIdx]] = gSrc[i];\n"
-" counter[tableIdx] ++;\n"
-" }\n"
-" }\n"
-" }\n"
-" }\n"
-" \n"
-"}\n"
-;
/include/Bullet/Bullet3OpenCL/ParallelPrimitives/kernels/RadixSort32KernelsCL.h
Property changes:
Deleted: svn:executable
## -1 +0,0 ##
-*
\ No newline at end of property
Index: include/Bullet/Bullet3OpenCL/ParallelPrimitives/kernels/FillKernelsCL.h
===================================================================
--- include/Bullet/Bullet3OpenCL/ParallelPrimitives/kernels/FillKernelsCL.h (revision 22)
+++ include/Bullet/Bullet3OpenCL/ParallelPrimitives/kernels/FillKernelsCL.h (nonexistent)
@@ -1,91 +0,0 @@
-//this file is autogenerated using stringify.bat (premake --stringify) in the build folder of this project
-static const char* fillKernelsCL= \
-"/*\n"
-"Copyright (c) 2012 Advanced Micro Devices, Inc. \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"
-"//Originally written by Takahiro Harada\n"
-"#pragma OPENCL EXTENSION cl_amd_printf : enable\n"
-"#pragma OPENCL EXTENSION cl_khr_local_int32_base_atomics : enable\n"
-"typedef unsigned int u32;\n"
-"#define GET_GROUP_IDX get_group_id(0)\n"
-"#define GET_LOCAL_IDX get_local_id(0)\n"
-"#define GET_GLOBAL_IDX get_global_id(0)\n"
-"#define GET_GROUP_SIZE get_local_size(0)\n"
-"#define GROUP_LDS_BARRIER barrier(CLK_LOCAL_MEM_FENCE)\n"
-"#define GROUP_MEM_FENCE mem_fence(CLK_LOCAL_MEM_FENCE)\n"
-"#define AtomInc(x) atom_inc(&(x))\n"
-"#define AtomInc1(x, out) out = atom_inc(&(x))\n"
-"#define make_uint4 (uint4)\n"
-"#define make_uint2 (uint2)\n"
-"#define make_int2 (int2)\n"
-"typedef struct\n"
-"{\n"
-" union\n"
-" {\n"
-" int4 m_data;\n"
-" uint4 m_unsignedData;\n"
-" float m_floatData;\n"
-" };\n"
-" int m_offset;\n"
-" int m_n;\n"
-" int m_padding[2];\n"
-"} ConstBuffer;\n"
-"__kernel\n"
-"__attribute__((reqd_work_group_size(64,1,1)))\n"
-"void FillIntKernel(__global int* dstInt, int num_elements, int value, const int offset)\n"
-"{\n"
-" int gIdx = GET_GLOBAL_IDX;\n"
-" if( gIdx < num_elements )\n"
-" {\n"
-" dstInt[ offset+gIdx ] = value;\n"
-" }\n"
-"}\n"
-"__kernel\n"
-"__attribute__((reqd_work_group_size(64,1,1)))\n"
-"void FillFloatKernel(__global float* dstFloat, int num_elements, float value, const int offset)\n"
-"{\n"
-" int gIdx = GET_GLOBAL_IDX;\n"
-" if( gIdx < num_elements )\n"
-" {\n"
-" dstFloat[ offset+gIdx ] = value;\n"
-" }\n"
-"}\n"
-"__kernel\n"
-"__attribute__((reqd_work_group_size(64,1,1)))\n"
-"void FillUnsignedIntKernel(__global unsigned int* dstInt, const int num, const unsigned int value, const int offset)\n"
-"{\n"
-" int gIdx = GET_GLOBAL_IDX;\n"
-" if( gIdx < num )\n"
-" {\n"
-" dstInt[ offset+gIdx ] = value;\n"
-" }\n"
-"}\n"
-"__kernel\n"
-"__attribute__((reqd_work_group_size(64,1,1)))\n"
-"void FillInt2Kernel(__global int2* dstInt2, const int num, const int2 value, const int offset)\n"
-"{\n"
-" int gIdx = GET_GLOBAL_IDX;\n"
-" if( gIdx < num )\n"
-" {\n"
-" dstInt2[ gIdx + offset] = make_int2( value.x, value.y );\n"
-" }\n"
-"}\n"
-"__kernel\n"
-"__attribute__((reqd_work_group_size(64,1,1)))\n"
-"void FillInt4Kernel(__global int4* dstInt4, const int num, const int4 value, const int offset)\n"
-"{\n"
-" int gIdx = GET_GLOBAL_IDX;\n"
-" if( gIdx < num )\n"
-" {\n"
-" dstInt4[ offset+gIdx ] = value;\n"
-" }\n"
-"}\n"
-;
/include/Bullet/Bullet3OpenCL/ParallelPrimitives/kernels/FillKernelsCL.h
Property changes:
Deleted: svn:executable
## -1 +0,0 ##
-*
\ No newline at end of property
Index: include/Bullet/Bullet3OpenCL/ParallelPrimitives/kernels/CopyKernelsCL.h
===================================================================
--- include/Bullet/Bullet3OpenCL/ParallelPrimitives/kernels/CopyKernelsCL.h (revision 22)
+++ include/Bullet/Bullet3OpenCL/ParallelPrimitives/kernels/CopyKernelsCL.h (nonexistent)
@@ -1,132 +0,0 @@
-//this file is autogenerated using stringify.bat (premake --stringify) in the build folder of this project
-static const char* copyKernelsCL= \
-"/*\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 Takahiro Harada\n"
-"\n"
-"#pragma OPENCL EXTENSION cl_amd_printf : enable\n"
-"#pragma OPENCL EXTENSION cl_khr_local_int32_base_atomics : enable\n"
-"\n"
-"typedef unsigned int u32;\n"
-"#define GET_GROUP_IDX get_group_id(0)\n"
-"#define GET_LOCAL_IDX get_local_id(0)\n"
-"#define GET_GLOBAL_IDX get_global_id(0)\n"
-"#define GET_GROUP_SIZE get_local_size(0)\n"
-"#define GROUP_LDS_BARRIER barrier(CLK_LOCAL_MEM_FENCE)\n"
-"#define GROUP_MEM_FENCE mem_fence(CLK_LOCAL_MEM_FENCE)\n"
-"#define AtomInc(x) atom_inc(&(x))\n"
-"#define AtomInc1(x, out) out = atom_inc(&(x))\n"
-"\n"
-"#define make_uint4 (uint4)\n"
-"#define make_uint2 (uint2)\n"
-"#define make_int2 (int2)\n"
-"\n"
-"typedef struct\n"
-"{\n"
-" int m_n;\n"
-" int m_padding[3];\n"
-"} ConstBuffer;\n"
-"\n"
-"\n"
-"\n"
-"__kernel\n"
-"__attribute__((reqd_work_group_size(64,1,1)))\n"
-"void Copy1F4Kernel(__global float4* dst, __global float4* src, \n"
-" ConstBuffer cb)\n"
-"{\n"
-" int gIdx = GET_GLOBAL_IDX;\n"
-"\n"
-" if( gIdx < cb.m_n )\n"
-" {\n"
-" float4 a0 = src[gIdx];\n"
-"\n"
-" dst[ gIdx ] = a0;\n"
-" }\n"
-"}\n"
-"\n"
-"__kernel\n"
-"__attribute__((reqd_work_group_size(64,1,1)))\n"
-"void Copy2F4Kernel(__global float4* dst, __global float4* src, \n"
-" ConstBuffer cb)\n"
-"{\n"
-" int gIdx = GET_GLOBAL_IDX;\n"
-"\n"
-" if( 2*gIdx <= cb.m_n )\n"
-" {\n"
-" float4 a0 = src[gIdx*2+0];\n"
-" float4 a1 = src[gIdx*2+1];\n"
-"\n"
-" dst[ gIdx*2+0 ] = a0;\n"
-" dst[ gIdx*2+1 ] = a1;\n"
-" }\n"
-"}\n"
-"\n"
-"__kernel\n"
-"__attribute__((reqd_work_group_size(64,1,1)))\n"
-"void Copy4F4Kernel(__global float4* dst, __global float4* src, \n"
-" ConstBuffer cb)\n"
-"{\n"
-" int gIdx = GET_GLOBAL_IDX;\n"
-"\n"
-" if( 4*gIdx <= cb.m_n )\n"
-" {\n"
-" int idx0 = gIdx*4+0;\n"
-" int idx1 = gIdx*4+1;\n"
-" int idx2 = gIdx*4+2;\n"
-" int idx3 = gIdx*4+3;\n"
-"\n"
-" float4 a0 = src[idx0];\n"
-" float4 a1 = src[idx1];\n"
-" float4 a2 = src[idx2];\n"
-" float4 a3 = src[idx3];\n"
-"\n"
-" dst[ idx0 ] = a0;\n"
-" dst[ idx1 ] = a1;\n"
-" dst[ idx2 ] = a2;\n"
-" dst[ idx3 ] = a3;\n"
-" }\n"
-"}\n"
-"\n"
-"__kernel\n"
-"__attribute__((reqd_work_group_size(64,1,1)))\n"
-"void CopyF1Kernel(__global float* dstF1, __global float* srcF1, \n"
-" ConstBuffer cb)\n"
-"{\n"
-" int gIdx = GET_GLOBAL_IDX;\n"
-"\n"
-" if( gIdx < cb.m_n )\n"
-" {\n"
-" float a0 = srcF1[gIdx];\n"
-"\n"
-" dstF1[ gIdx ] = a0;\n"
-" }\n"
-"}\n"
-"\n"
-"__kernel\n"
-"__attribute__((reqd_work_group_size(64,1,1)))\n"
-"void CopyF2Kernel(__global float2* dstF2, __global float2* srcF2, \n"
-" ConstBuffer cb)\n"
-"{\n"
-" int gIdx = GET_GLOBAL_IDX;\n"
-"\n"
-" if( gIdx < cb.m_n )\n"
-" {\n"
-" float2 a0 = srcF2[gIdx];\n"
-"\n"
-" dstF2[ gIdx ] = a0;\n"
-" }\n"
-"}\n"
-"\n"
-"\n"
-;
/include/Bullet/Bullet3OpenCL/ParallelPrimitives/kernels/CopyKernelsCL.h
Property changes:
Deleted: svn:executable
## -1 +0,0 ##
-*
\ No newline at end of property
Index: include/Bullet/Bullet3OpenCL/ParallelPrimitives/kernels/PrefixScanKernelsFloat4CL.h
===================================================================
--- include/Bullet/Bullet3OpenCL/ParallelPrimitives/kernels/PrefixScanKernelsFloat4CL.h (revision 22)
+++ include/Bullet/Bullet3OpenCL/ParallelPrimitives/kernels/PrefixScanKernelsFloat4CL.h (nonexistent)
@@ -1,129 +0,0 @@
-//this file is autogenerated using stringify.bat (premake --stringify) in the build folder of this project
-static const char* prefixScanKernelsFloat4CL= \
-"/*\n"
-"Copyright (c) 2012 Advanced Micro Devices, Inc. \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"
-"//Originally written by Takahiro Harada\n"
-"typedef unsigned int u32;\n"
-"#define GET_GROUP_IDX get_group_id(0)\n"
-"#define GET_LOCAL_IDX get_local_id(0)\n"
-"#define GET_GLOBAL_IDX get_global_id(0)\n"
-"#define GET_GROUP_SIZE get_local_size(0)\n"
-"#define GROUP_LDS_BARRIER barrier(CLK_LOCAL_MEM_FENCE)\n"
-"// takahiro end\n"
-"#define WG_SIZE 128 \n"
-"#define m_numElems x\n"
-"#define m_numBlocks y\n"
-"#define m_numScanBlocks z\n"
-"/*typedef struct\n"
-"{\n"
-" uint m_numElems;\n"
-" uint m_numBlocks;\n"
-" uint m_numScanBlocks;\n"
-" uint m_padding[1];\n"
-"} ConstBuffer;\n"
-"*/\n"
-"float4 ScanExclusiveFloat4(__local float4* data, u32 n, int lIdx, int lSize)\n"
-"{\n"
-" float4 blocksum;\n"
-" int offset = 1;\n"
-" for(int nActive=n>>1; nActive>0; nActive>>=1, offset<<=1)\n"
-" {\n"
-" GROUP_LDS_BARRIER;\n"
-" for(int iIdx=lIdx; iIdx<nActive; iIdx+=lSize)\n"
-" {\n"
-" int ai = offset*(2*iIdx+1)-1;\n"
-" int bi = offset*(2*iIdx+2)-1;\n"
-" data[bi] += data[ai];\n"
-" }\n"
-" }\n"
-" GROUP_LDS_BARRIER;\n"
-" if( lIdx == 0 )\n"
-" {\n"
-" blocksum = data[ n-1 ];\n"
-" data[ n-1 ] = 0;\n"
-" }\n"
-" GROUP_LDS_BARRIER;\n"
-" offset >>= 1;\n"
-" for(int nActive=1; nActive<n; nActive<<=1, offset>>=1 )\n"
-" {\n"
-" GROUP_LDS_BARRIER;\n"
-" for( int iIdx = lIdx; iIdx<nActive; iIdx += lSize )\n"
-" {\n"
-" int ai = offset*(2*iIdx+1)-1;\n"
-" int bi = offset*(2*iIdx+2)-1;\n"
-" float4 temp = data[ai];\n"
-" data[ai] = data[bi];\n"
-" data[bi] += temp;\n"
-" }\n"
-" }\n"
-" GROUP_LDS_BARRIER;\n"
-" return blocksum;\n"
-"}\n"
-"__attribute__((reqd_work_group_size(WG_SIZE,1,1)))\n"
-"__kernel\n"
-"void LocalScanKernel(__global float4* dst, __global float4* src, __global float4* sumBuffer, uint4 cb)\n"
-"{\n"
-" __local float4 ldsData[WG_SIZE*2];\n"
-" int gIdx = GET_GLOBAL_IDX;\n"
-" int lIdx = GET_LOCAL_IDX;\n"
-" ldsData[2*lIdx] = ( 2*gIdx < cb.m_numElems )? src[2*gIdx]: 0;\n"
-" ldsData[2*lIdx + 1] = ( 2*gIdx+1 < cb.m_numElems )? src[2*gIdx + 1]: 0;\n"
-" float4 sum = ScanExclusiveFloat4(ldsData, WG_SIZE*2, GET_LOCAL_IDX, GET_GROUP_SIZE);\n"
-" if( lIdx == 0 ) \n"
-" sumBuffer[GET_GROUP_IDX] = sum;\n"
-" if( (2*gIdx) < cb.m_numElems )\n"
-" {\n"
-" dst[2*gIdx] = ldsData[2*lIdx];\n"
-" }\n"
-" if( (2*gIdx + 1) < cb.m_numElems )\n"
-" {\n"
-" dst[2*gIdx + 1] = ldsData[2*lIdx + 1];\n"
-" }\n"
-"}\n"
-"__attribute__((reqd_work_group_size(WG_SIZE,1,1)))\n"
-"__kernel\n"
-"void AddOffsetKernel(__global float4* dst, __global float4* blockSum, uint4 cb)\n"
-"{\n"
-" const u32 blockSize = WG_SIZE*2;\n"
-" int myIdx = GET_GROUP_IDX+1;\n"
-" int lIdx = GET_LOCAL_IDX;\n"
-" float4 iBlockSum = blockSum[myIdx];\n"
-" int endValue = min((myIdx+1)*(blockSize), cb.m_numElems);\n"
-" for(int i=myIdx*blockSize+lIdx; i<endValue; i+=GET_GROUP_SIZE)\n"
-" {\n"
-" dst[i] += iBlockSum;\n"
-" }\n"
-"}\n"
-"__attribute__((reqd_work_group_size(WG_SIZE,1,1)))\n"
-"__kernel\n"
-"void TopLevelScanKernel(__global float4* dst, uint4 cb)\n"
-"{\n"
-" __local float4 ldsData[2048];\n"
-" int gIdx = GET_GLOBAL_IDX;\n"
-" int lIdx = GET_LOCAL_IDX;\n"
-" int lSize = GET_GROUP_SIZE;\n"
-" for(int i=lIdx; i<cb.m_numScanBlocks; i+=lSize )\n"
-" {\n"
-" ldsData[i] = (i<cb.m_numBlocks)? dst[i]:0;\n"
-" }\n"
-" GROUP_LDS_BARRIER;\n"
-" float4 sum = ScanExclusiveFloat4(ldsData, cb.m_numScanBlocks, GET_LOCAL_IDX, GET_GROUP_SIZE);\n"
-" for(int i=lIdx; i<cb.m_numBlocks; i+=lSize )\n"
-" {\n"
-" dst[i] = ldsData[i];\n"
-" }\n"
-" if( gIdx == 0 )\n"
-" {\n"
-" dst[cb.m_numBlocks] = sum;\n"
-" }\n"
-"}\n"
-;
/include/Bullet/Bullet3OpenCL/ParallelPrimitives/kernels/PrefixScanKernelsFloat4CL.h
Property changes:
Deleted: svn:executable
## -1 +0,0 ##
-*
\ No newline at end of property
Index: include/Bullet/Bullet3OpenCL/ParallelPrimitives/kernels/BoundSearchKernelsCL.h
===================================================================
--- include/Bullet/Bullet3OpenCL/ParallelPrimitives/kernels/BoundSearchKernelsCL.h (revision 22)
+++ include/Bullet/Bullet3OpenCL/ParallelPrimitives/kernels/BoundSearchKernelsCL.h (nonexistent)
@@ -1,87 +0,0 @@
-//this file is autogenerated using stringify.bat (premake --stringify) in the build folder of this project
-static const char* boundSearchKernelsCL= \
-"/*\n"
-"Copyright (c) 2012 Advanced Micro Devices, Inc. \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"
-"//Originally written by Takahiro Harada\n"
-"typedef unsigned int u32;\n"
-"#define GET_GROUP_IDX get_group_id(0)\n"
-"#define GET_LOCAL_IDX get_local_id(0)\n"
-"#define GET_GLOBAL_IDX get_global_id(0)\n"
-"#define GET_GROUP_SIZE get_local_size(0)\n"
-"#define GROUP_LDS_BARRIER barrier(CLK_LOCAL_MEM_FENCE)\n"
-"typedef struct\n"
-"{\n"
-" u32 m_key; \n"
-" u32 m_value;\n"
-"}SortData;\n"
-"typedef struct\n"
-"{\n"
-" u32 m_nSrc;\n"
-" u32 m_nDst;\n"
-" u32 m_padding[2];\n"
-"} ConstBuffer;\n"
-"__attribute__((reqd_work_group_size(64,1,1)))\n"
-"__kernel\n"
-"void SearchSortDataLowerKernel(__global SortData* src, __global u32 *dst, \n"
-" unsigned int nSrc, unsigned int nDst)\n"
-"{\n"
-" int gIdx = GET_GLOBAL_IDX;\n"
-" if( gIdx < nSrc )\n"
-" {\n"
-" SortData first; first.m_key = (u32)(-1); first.m_value = (u32)(-1);\n"
-" SortData end; end.m_key = nDst; end.m_value = nDst;\n"
-" SortData iData = (gIdx==0)? first: src[gIdx-1];\n"
-" SortData jData = (gIdx==nSrc)? end: src[gIdx];\n"
-" if( iData.m_key != jData.m_key )\n"
-" {\n"
-"// for(u32 k=iData.m_key+1; k<=min(jData.m_key, nDst-1); k++)\n"
-" u32 k = jData.m_key;\n"
-" {\n"
-" dst[k] = gIdx;\n"
-" }\n"
-" }\n"
-" }\n"
-"}\n"
-"__attribute__((reqd_work_group_size(64,1,1)))\n"
-"__kernel\n"
-"void SearchSortDataUpperKernel(__global SortData* src, __global u32 *dst, \n"
-" unsigned int nSrc, unsigned int nDst)\n"
-"{\n"
-" int gIdx = GET_GLOBAL_IDX+1;\n"
-" if( gIdx < nSrc+1 )\n"
-" {\n"
-" SortData first; first.m_key = 0; first.m_value = 0;\n"
-" SortData end; end.m_key = nDst; end.m_value = nDst;\n"
-" SortData iData = src[gIdx-1];\n"
-" SortData jData = (gIdx==nSrc)? end: src[gIdx];\n"
-" if( iData.m_key != jData.m_key )\n"
-" {\n"
-" u32 k = iData.m_key;\n"
-" {\n"
-" dst[k] = gIdx;\n"
-" }\n"
-" }\n"
-" }\n"
-"}\n"
-"__attribute__((reqd_work_group_size(64,1,1)))\n"
-"__kernel\n"
-"void SubtractKernel(__global u32* A, __global u32 *B, __global u32 *C, \n"
-" unsigned int nSrc, unsigned int nDst)\n"
-"{\n"
-" int gIdx = GET_GLOBAL_IDX;\n"
-" \n"
-" if( gIdx < nDst )\n"
-" {\n"
-" C[gIdx] = A[gIdx] - B[gIdx];\n"
-" }\n"
-"}\n"
-;
/include/Bullet/Bullet3OpenCL/ParallelPrimitives/kernels/BoundSearchKernelsCL.h
Property changes:
Deleted: svn:executable
## -1 +0,0 ##
-*
\ No newline at end of property
Index: include/Bullet/Bullet3OpenCL/ParallelPrimitives/b3PrefixScanFloat4CL.h
===================================================================
--- include/Bullet/Bullet3OpenCL/ParallelPrimitives/b3PrefixScanFloat4CL.h (revision 22)
+++ include/Bullet/Bullet3OpenCL/ParallelPrimitives/b3PrefixScanFloat4CL.h (nonexistent)
@@ -1,38 +0,0 @@
-
-#ifndef B3_PREFIX_SCAN_CL_H
-#define B3_PREFIX_SCAN_CL_H
-
-#include "b3OpenCLArray.h"
-#include "b3BufferInfoCL.h"
-#include "Bullet3Common/b3AlignedObjectArray.h"
-#include "Bullet3Common/b3Vector3.h"
-
-class b3PrefixScanFloat4CL
-{
- enum
- {
- BLOCK_SIZE = 128
- };
-
-// Option m_option;
-
- cl_command_queue m_commandQueue;
-
- cl_kernel m_localScanKernel;
- cl_kernel m_blockSumKernel;
- cl_kernel m_propagationKernel;
-
- b3OpenCLArray<b3Vector3>* m_workBuffer;
-
-
- public:
-
- b3PrefixScanFloat4CL(cl_context ctx, cl_device_id device, cl_command_queue queue,int size=0);
-
- virtual ~b3PrefixScanFloat4CL();
-
- void execute(b3OpenCLArray<b3Vector3>& src, b3OpenCLArray<b3Vector3>& dst, int n, b3Vector3* sum = 0);
- void executeHost(b3AlignedObjectArray<b3Vector3>& src, b3AlignedObjectArray<b3Vector3>& dst, int n, b3Vector3* sum);
-};
-
-#endif //B3_PREFIX_SCAN_CL_H
/include/Bullet/Bullet3OpenCL/ParallelPrimitives/b3PrefixScanFloat4CL.h
Property changes:
Deleted: svn:executable
## -1 +0,0 ##
-*
\ No newline at end of property
Index: include/Bullet/Bullet3OpenCL/ParallelPrimitives/b3BoundSearchCL.h
===================================================================
--- include/Bullet/Bullet3OpenCL/ParallelPrimitives/b3BoundSearchCL.h (revision 22)
+++ include/Bullet/Bullet3OpenCL/ParallelPrimitives/b3BoundSearchCL.h (nonexistent)
@@ -1,67 +0,0 @@
-/*
-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 Takahiro Harada
-
-#ifndef B3_BOUNDSEARCH_H
-#define B3_BOUNDSEARCH_H
-
-#pragma once
-
-/*#include <Adl/Adl.h>
-#include <AdlPrimitives/Math/Math.h>
-#include <AdlPrimitives/Sort/SortData.h>
-#include <AdlPrimitives/Fill/Fill.h>
-*/
-
-#include "b3OpenCLArray.h"
-#include "b3FillCL.h"
-#include "b3RadixSort32CL.h" //for b3SortData (perhaps move it?)
-class b3BoundSearchCL
-{
- public:
-
- enum Option
- {
- BOUND_LOWER,
- BOUND_UPPER,
- COUNT,
- };
-
- cl_context m_context;
- cl_device_id m_device;
- cl_command_queue m_queue;
-
-
- cl_kernel m_lowerSortDataKernel;
- cl_kernel m_upperSortDataKernel;
- cl_kernel m_subtractKernel;
-
- b3OpenCLArray<b3Int4>* m_constbtOpenCLArray;
- b3OpenCLArray<unsigned int>* m_lower;
- b3OpenCLArray<unsigned int>* m_upper;
-
- b3FillCL* m_filler;
-
- b3BoundSearchCL(cl_context context, cl_device_id device, cl_command_queue queue, int size);
-
- virtual ~b3BoundSearchCL();
-
- // src has to be src[i].m_key <= src[i+1].m_key
- void execute( b3OpenCLArray<b3SortData>& src, int nSrc, b3OpenCLArray<unsigned int>& dst, int nDst, Option option = BOUND_LOWER );
-
- void executeHost( b3AlignedObjectArray<b3SortData>& src, int nSrc, b3AlignedObjectArray<unsigned int>& dst, int nDst, Option option = BOUND_LOWER);
-};
-
-
-#endif //B3_BOUNDSEARCH_H
/include/Bullet/Bullet3OpenCL/ParallelPrimitives/b3BoundSearchCL.h
Property changes:
Deleted: svn:executable
## -1 +0,0 ##
-*
\ No newline at end of property
Index: include/Bullet/Bullet3OpenCL/RigidBody/b3GpuNarrowPhaseInternalData.h
===================================================================
--- include/Bullet/Bullet3OpenCL/RigidBody/b3GpuNarrowPhaseInternalData.h (revision 22)
+++ include/Bullet/Bullet3OpenCL/RigidBody/b3GpuNarrowPhaseInternalData.h (nonexistent)
@@ -1,95 +0,0 @@
-
-#ifndef B3_GPU_NARROWPHASE_INTERNAL_DATA_H
-#define B3_GPU_NARROWPHASE_INTERNAL_DATA_H
-
-#include "Bullet3OpenCL/ParallelPrimitives/b3OpenCLArray.h"
-#include "Bullet3Collision/NarrowPhaseCollision/shared/b3ConvexPolyhedronData.h"
-#include "Bullet3Collision/NarrowPhaseCollision/b3Config.h"
-#include "Bullet3Collision/NarrowPhaseCollision/shared/b3Collidable.h"
-
-#include "Bullet3OpenCL/Initialize/b3OpenCLInclude.h"
-#include "Bullet3Common/b3AlignedObjectArray.h"
-#include "Bullet3Common/b3Vector3.h"
-
-#include "Bullet3Collision/NarrowPhaseCollision/shared/b3RigidBodyData.h"
-#include "Bullet3Collision/NarrowPhaseCollision/b3Contact4.h"
-#include "Bullet3OpenCL/BroadphaseCollision/b3SapAabb.h"
-
-#include "Bullet3OpenCL/NarrowphaseCollision/b3QuantizedBvh.h"
-#include "Bullet3OpenCL/NarrowphaseCollision/b3BvhInfo.h"
-#include "Bullet3Common/shared/b3Int4.h"
-#include "Bullet3Common/shared/b3Int2.h"
-
-
-class b3ConvexUtility;
-
-struct b3GpuNarrowPhaseInternalData
-{
- b3AlignedObjectArray<b3ConvexUtility*>* m_convexData;
-
- b3AlignedObjectArray<b3ConvexPolyhedronData> m_convexPolyhedra;
- b3AlignedObjectArray<b3Vector3> m_uniqueEdges;
- b3AlignedObjectArray<b3Vector3> m_convexVertices;
- b3AlignedObjectArray<int> m_convexIndices;
-
- b3OpenCLArray<b3ConvexPolyhedronData>* m_convexPolyhedraGPU;
- b3OpenCLArray<b3Vector3>* m_uniqueEdgesGPU;
- b3OpenCLArray<b3Vector3>* m_convexVerticesGPU;
- b3OpenCLArray<int>* m_convexIndicesGPU;
-
- b3OpenCLArray<b3Vector3>* m_worldVertsB1GPU;
- b3OpenCLArray<b3Int4>* m_clippingFacesOutGPU;
- b3OpenCLArray<b3Vector3>* m_worldNormalsAGPU;
- b3OpenCLArray<b3Vector3>* m_worldVertsA1GPU;
- b3OpenCLArray<b3Vector3>* m_worldVertsB2GPU;
-
- b3AlignedObjectArray<b3GpuChildShape> m_cpuChildShapes;
- b3OpenCLArray<b3GpuChildShape>* m_gpuChildShapes;
-
- b3AlignedObjectArray<b3GpuFace> m_convexFaces;
- b3OpenCLArray<b3GpuFace>* m_convexFacesGPU;
-
- struct GpuSatCollision* m_gpuSatCollision;
-
-
- b3OpenCLArray<b3Int4>* m_triangleConvexPairs;
-
-
- b3OpenCLArray<b3Contact4>* m_pBufContactBuffersGPU[2];
- int m_currentContactBuffer;
- b3AlignedObjectArray<b3Contact4>* m_pBufContactOutCPU;
-
-
- b3AlignedObjectArray<b3RigidBodyData>* m_bodyBufferCPU;
- b3OpenCLArray<b3RigidBodyData>* m_bodyBufferGPU;
-
- b3AlignedObjectArray<b3InertiaData>* m_inertiaBufferCPU;
- b3OpenCLArray<b3InertiaData>* m_inertiaBufferGPU;
-
- int m_numAcceleratedShapes;
- int m_numAcceleratedRigidBodies;
-
- b3AlignedObjectArray<b3Collidable> m_collidablesCPU;
- b3OpenCLArray<b3Collidable>* m_collidablesGPU;
-
- b3OpenCLArray<b3SapAabb>* m_localShapeAABBGPU;
- b3AlignedObjectArray<b3SapAabb>* m_localShapeAABBCPU;
-
- b3AlignedObjectArray<class b3OptimizedBvh*> m_bvhData;
- b3AlignedObjectArray<class b3TriangleIndexVertexArray*> m_meshInterfaces;
-
- b3AlignedObjectArray<b3QuantizedBvhNode> m_treeNodesCPU;
- b3AlignedObjectArray<b3BvhSubtreeInfo> m_subTreesCPU;
-
- b3AlignedObjectArray<b3BvhInfo> m_bvhInfoCPU;
- b3OpenCLArray<b3BvhInfo>* m_bvhInfoGPU;
-
- b3OpenCLArray<b3QuantizedBvhNode>* m_treeNodesGPU;
- b3OpenCLArray<b3BvhSubtreeInfo>* m_subTreesGPU;
-
-
- b3Config m_config;
-
-};
-
-#endif //B3_GPU_NARROWPHASE_INTERNAL_DATA_H
/include/Bullet/Bullet3OpenCL/RigidBody/b3GpuNarrowPhaseInternalData.h
Property changes:
Deleted: svn:executable
## -1 +0,0 ##
-*
\ No newline at end of property
Index: include/Bullet/Bullet3OpenCL/RigidBody/b3GpuPgsContactSolver.h
===================================================================
--- include/Bullet/Bullet3OpenCL/RigidBody/b3GpuPgsContactSolver.h (revision 22)
+++ include/Bullet/Bullet3OpenCL/RigidBody/b3GpuPgsContactSolver.h (nonexistent)
@@ -1,43 +0,0 @@
-
-#ifndef B3_GPU_BATCHING_PGS_SOLVER_H
-#define B3_GPU_BATCHING_PGS_SOLVER_H
-
-#include "Bullet3OpenCL/Initialize/b3OpenCLInclude.h"
-#include "Bullet3OpenCL/ParallelPrimitives/b3OpenCLArray.h"
-#include "Bullet3Collision/NarrowPhaseCollision/shared/b3RigidBodyData.h"
-#include "Bullet3Collision/NarrowPhaseCollision/b3Contact4.h"
-#include "b3GpuConstraint4.h"
-
-class b3GpuPgsContactSolver
-{
-protected:
-
- int m_debugOutput;
-
- struct b3GpuBatchingPgsSolverInternalData* m_data;
-
- void batchContacts( b3OpenCLArray<b3Contact4>* contacts, int nContacts, b3OpenCLArray<unsigned int>* n, b3OpenCLArray<unsigned int>* offsets, int staticIdx );
-
- inline int sortConstraintByBatch( b3Contact4* cs, int n, int simdWidth , int staticIdx, int numBodies);
- inline int sortConstraintByBatch2( b3Contact4* cs, int n, int simdWidth , int staticIdx, int numBodies);
- inline int sortConstraintByBatch3( b3Contact4* cs, int n, int simdWidth , int staticIdx, int numBodies, int* batchSizes);
-
-
-
- void solveContactConstraintBatchSizes( const b3OpenCLArray<b3RigidBodyData>* bodyBuf, const b3OpenCLArray<b3InertiaData>* shapeBuf,
- b3OpenCLArray<b3GpuConstraint4>* constraint, void* additionalData, int n ,int maxNumBatches, int numIterations, const b3AlignedObjectArray<int>* batchSizes);//const b3OpenCLArray<int>* gpuBatchSizes);
-
- void solveContactConstraint( const b3OpenCLArray<b3RigidBodyData>* bodyBuf, const b3OpenCLArray<b3InertiaData>* shapeBuf,
- b3OpenCLArray<b3GpuConstraint4>* constraint, void* additionalData, int n ,int maxNumBatches, int numIterations, const b3AlignedObjectArray<int>* batchSizes);//const b3OpenCLArray<int>* gpuBatchSizes);
-
-public:
-
- b3GpuPgsContactSolver(cl_context ctx,cl_device_id device, cl_command_queue q,int pairCapacity);
- virtual ~b3GpuPgsContactSolver();
-
- void solveContacts(int numBodies, cl_mem bodyBuf, cl_mem inertiaBuf, int numContacts, cl_mem contactBuf, const struct b3Config& config, int static0Index);
-
-};
-
-#endif //B3_GPU_BATCHING_PGS_SOLVER_H
-
/include/Bullet/Bullet3OpenCL/RigidBody/b3GpuPgsContactSolver.h
Property changes:
Deleted: svn:executable
## -1 +0,0 ##
-*
\ No newline at end of property
Index: include/Bullet/Bullet3OpenCL/RigidBody/b3GpuGenericConstraint.h
===================================================================
--- include/Bullet/Bullet3OpenCL/RigidBody/b3GpuGenericConstraint.h (revision 22)
+++ include/Bullet/Bullet3OpenCL/RigidBody/b3GpuGenericConstraint.h (nonexistent)
@@ -1,132 +0,0 @@
-/*
-Copyright (c) 2013 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
-
-#ifndef B3_GPU_GENERIC_CONSTRAINT_H
-#define B3_GPU_GENERIC_CONSTRAINT_H
-
-#include "Bullet3Common/b3Quaternion.h"
-struct b3RigidBodyData;
-enum B3_CONSTRAINT_FLAGS
-{
- B3_CONSTRAINT_FLAG_ENABLED=1,
-};
-
-enum b3GpuGenericConstraintType
-{
- B3_GPU_POINT2POINT_CONSTRAINT_TYPE=3,
- B3_GPU_FIXED_CONSTRAINT_TYPE=4,
-// B3_HINGE_CONSTRAINT_TYPE,
-// B3_CONETWIST_CONSTRAINT_TYPE,
-// B3_D6_CONSTRAINT_TYPE,
-// B3_SLIDER_CONSTRAINT_TYPE,
-// B3_CONTACT_CONSTRAINT_TYPE,
-// B3_D6_SPRING_CONSTRAINT_TYPE,
-// B3_GEAR_CONSTRAINT_TYPE,
-
- B3_GPU_MAX_CONSTRAINT_TYPE
-};
-
-
-
-struct b3GpuConstraintInfo2
-{
- // integrator parameters: frames per second (1/stepsize), default error
- // reduction parameter (0..1).
- b3Scalar fps,erp;
-
- // for the first and second body, pointers to two (linear and angular)
- // n*3 jacobian sub matrices, stored by rows. these matrices will have
- // been initialized to 0 on entry. if the second body is zero then the
- // J2xx pointers may be 0.
- b3Scalar *m_J1linearAxis,*m_J1angularAxis,*m_J2linearAxis,*m_J2angularAxis;
-
- // elements to jump from one row to the next in J's
- int rowskip;
-
- // right hand sides of the equation J*v = c + cfm * lambda. cfm is the
- // "constraint force mixing" vector. c is set to zero on entry, cfm is
- // set to a constant value (typically very small or zero) value on entry.
- b3Scalar *m_constraintError,*cfm;
-
- // lo and hi limits for variables (set to -/+ infinity on entry).
- b3Scalar *m_lowerLimit,*m_upperLimit;
-
- // findex vector for variables. see the LCP solver interface for a
- // description of what this does. this is set to -1 on entry.
- // note that the returned indexes are relative to the first index of
- // the constraint.
- int *findex;
- // number of solver iterations
- int m_numIterations;
-
- //damping of the velocity
- b3Scalar m_damping;
-};
-
-
-B3_ATTRIBUTE_ALIGNED16(struct) b3GpuGenericConstraint
-{
- int m_constraintType;
- int m_rbA;
- int m_rbB;
- float m_breakingImpulseThreshold;
-
- b3Vector3 m_pivotInA;
- b3Vector3 m_pivotInB;
- b3Quaternion m_relTargetAB;
-
- int m_flags;
- int m_uid;
- int m_padding[2];
-
- int getRigidBodyA() const
- {
- return m_rbA;
- }
- int getRigidBodyB() const
- {
- return m_rbB;
- }
-
- const b3Vector3& getPivotInA() const
- {
- return m_pivotInA;
- }
-
- const b3Vector3& getPivotInB() const
- {
- return m_pivotInB;
- }
-
- int isEnabled() const
- {
- return m_flags & B3_CONSTRAINT_FLAG_ENABLED;
- }
-
- float getBreakingImpulseThreshold() const
- {
- return m_breakingImpulseThreshold;
- }
-
- ///internal method used by the constraint solver, don't use them directly
- void getInfo1 (unsigned int* info,const b3RigidBodyData* bodies);
-
- ///internal method used by the constraint solver, don't use them directly
- void getInfo2 (b3GpuConstraintInfo2* info, const b3RigidBodyData* bodies);
-
-
-};
-
-#endif //B3_GPU_GENERIC_CONSTRAINT_H
\ No newline at end of file
/include/Bullet/Bullet3OpenCL/RigidBody/b3GpuGenericConstraint.h
Property changes:
Deleted: svn:executable
## -1 +0,0 ##
-*
\ No newline at end of property
Index: include/Bullet/Bullet3OpenCL/RigidBody/b3GpuRigidBodyPipeline.h
===================================================================
--- include/Bullet/Bullet3OpenCL/RigidBody/b3GpuRigidBodyPipeline.h (revision 22)
+++ include/Bullet/Bullet3OpenCL/RigidBody/b3GpuRigidBodyPipeline.h (nonexistent)
@@ -1,74 +0,0 @@
-/*
-Copyright (c) 2013 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
-
-#ifndef B3_GPU_RIGIDBODY_PIPELINE_H
-#define B3_GPU_RIGIDBODY_PIPELINE_H
-
-#include "Bullet3OpenCL/Initialize/b3OpenCLInclude.h"
-#include "Bullet3Collision/NarrowPhaseCollision/b3Config.h"
-
-#include "Bullet3Common/b3AlignedObjectArray.h"
-#include "Bullet3Collision/NarrowPhaseCollision/b3RaycastInfo.h"
-
-class b3GpuRigidBodyPipeline
-{
-protected:
- struct b3GpuRigidBodyPipelineInternalData* m_data;
-
- int allocateCollidable();
-
-public:
-
-
- b3GpuRigidBodyPipeline(cl_context ctx,cl_device_id device, cl_command_queue q , class b3GpuNarrowPhase* narrowphase, class b3GpuBroadphaseInterface* broadphaseSap, struct b3DynamicBvhBroadphase* broadphaseDbvt, const b3Config& config);
- virtual ~b3GpuRigidBodyPipeline();
-
- void stepSimulation(float deltaTime);
- void integrate(float timeStep);
- void setupGpuAabbsFull();
-
- int registerConvexPolyhedron(class b3ConvexUtility* convex);
-
- //int registerConvexPolyhedron(const float* vertices, int strideInBytes, int numVertices, const float* scaling);
- //int registerSphereShape(float radius);
- //int registerPlaneShape(const b3Vector3& planeNormal, float planeConstant);
-
- //int registerConcaveMesh(b3AlignedObjectArray<b3Vector3>* vertices, b3AlignedObjectArray<int>* indices, const float* scaling);
- //int registerCompoundShape(b3AlignedObjectArray<b3GpuChildShape>* childShapes);
-
-
- int registerPhysicsInstance(float mass, const float* position, const float* orientation, int collisionShapeIndex, int userData, bool writeInstanceToGpu);
- //if you passed "writeInstanceToGpu" false in the registerPhysicsInstance method (for performance) you need to call writeAllInstancesToGpu after all instances are registered
- void writeAllInstancesToGpu();
- void copyConstraintsToHost();
- void setGravity(const float* grav);
- void reset();
-
- int createPoint2PointConstraint(int bodyA, int bodyB, const float* pivotInA, const float* pivotInB,float breakingThreshold);
- int createFixedConstraint(int bodyA, int bodyB, const float* pivotInA, const float* pivotInB, const float* relTargetAB, float breakingThreshold);
- void removeConstraintByUid(int uid);
-
- void addConstraint(class b3TypedConstraint* constraint);
- void removeConstraint(b3TypedConstraint* constraint);
-
- void castRays(const b3AlignedObjectArray<b3RayInfo>& rays, b3AlignedObjectArray<b3RayHit>& hitResults);
-
- cl_mem getBodyBuffer();
-
- int getNumBodies() const;
-
-};
-
-#endif //B3_GPU_RIGIDBODY_PIPELINE_H
\ No newline at end of file
/include/Bullet/Bullet3OpenCL/RigidBody/b3GpuRigidBodyPipeline.h
Property changes:
Deleted: svn:executable
## -1 +0,0 ##
-*
\ No newline at end of property
Index: include/Bullet/Bullet3OpenCL/RigidBody/b3GpuSolverBody.h
===================================================================
--- include/Bullet/Bullet3OpenCL/RigidBody/b3GpuSolverBody.h (revision 22)
+++ include/Bullet/Bullet3OpenCL/RigidBody/b3GpuSolverBody.h (nonexistent)
@@ -1,228 +0,0 @@
-/*
-Copyright (c) 2013 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
-
-
-#ifndef B3_GPU_SOLVER_BODY_H
-#define B3_GPU_SOLVER_BODY_H
-
-
-#include "Bullet3Common/b3Vector3.h"
-#include "Bullet3Common/b3Matrix3x3.h"
-
-#include "Bullet3Common/b3AlignedAllocator.h"
-#include "Bullet3Common/b3TransformUtil.h"
-
-///Until we get other contributions, only use SIMD on Windows, when using Visual Studio 2008 or later, and not double precision
-#ifdef B3_USE_SSE
-#define USE_SIMD 1
-#endif //
-
-
-
-///The b3SolverBody is an internal datastructure for the constraint solver. Only necessary data is packed to increase cache coherence/performance.
-B3_ATTRIBUTE_ALIGNED16 (struct) b3GpuSolverBody
-{
- B3_DECLARE_ALIGNED_ALLOCATOR();
-// b3Transform m_worldTransformUnused;
- b3Vector3 m_deltaLinearVelocity;
- b3Vector3 m_deltaAngularVelocity;
- b3Vector3 m_angularFactor;
- b3Vector3 m_linearFactor;
- b3Vector3 m_invMass;
- b3Vector3 m_pushVelocity;
- b3Vector3 m_turnVelocity;
- b3Vector3 m_linearVelocity;
- b3Vector3 m_angularVelocity;
-
- union
- {
- void* m_originalBody;
- int m_originalBodyIndex;
- };
-
- int padding[3];
-
- /*
- void setWorldTransform(const b3Transform& worldTransform)
- {
- m_worldTransform = worldTransform;
- }
-
- const b3Transform& getWorldTransform() const
- {
- return m_worldTransform;
- }
- */
- B3_FORCE_INLINE void getVelocityInLocalPointObsolete(const b3Vector3& rel_pos, b3Vector3& velocity ) const
- {
- if (m_originalBody)
- velocity = m_linearVelocity+m_deltaLinearVelocity + (m_angularVelocity+m_deltaAngularVelocity).cross(rel_pos);
- else
- velocity.setValue(0,0,0);
- }
-
- B3_FORCE_INLINE void getAngularVelocity(b3Vector3& angVel) const
- {
- if (m_originalBody)
- angVel =m_angularVelocity+m_deltaAngularVelocity;
- else
- angVel.setValue(0,0,0);
- }
-
-
- //Optimization for the iterative solver: avoid calculating constant terms involving inertia, normal, relative position
- B3_FORCE_INLINE void applyImpulse(const b3Vector3& linearComponent, const b3Vector3& angularComponent,const b3Scalar impulseMagnitude)
- {
- if (m_originalBody)
- {
- m_deltaLinearVelocity += linearComponent*impulseMagnitude*m_linearFactor;
- m_deltaAngularVelocity += angularComponent*(impulseMagnitude*m_angularFactor);
- }
- }
-
- B3_FORCE_INLINE void internalApplyPushImpulse(const b3Vector3& linearComponent, const b3Vector3& angularComponent,b3Scalar impulseMagnitude)
- {
- if (m_originalBody)
- {
- m_pushVelocity += linearComponent*impulseMagnitude*m_linearFactor;
- m_turnVelocity += angularComponent*(impulseMagnitude*m_angularFactor);
- }
- }
-
-
-
- const b3Vector3& getDeltaLinearVelocity() const
- {
- return m_deltaLinearVelocity;
- }
-
- const b3Vector3& getDeltaAngularVelocity() const
- {
- return m_deltaAngularVelocity;
- }
-
- const b3Vector3& getPushVelocity() const
- {
- return m_pushVelocity;
- }
-
- const b3Vector3& getTurnVelocity() const
- {
- return m_turnVelocity;
- }
-
-
- ////////////////////////////////////////////////
- ///some internal methods, don't use them
-
- b3Vector3& internalGetDeltaLinearVelocity()
- {
- return m_deltaLinearVelocity;
- }
-
- b3Vector3& internalGetDeltaAngularVelocity()
- {
- return m_deltaAngularVelocity;
- }
-
- const b3Vector3& internalGetAngularFactor() const
- {
- return m_angularFactor;
- }
-
- const b3Vector3& internalGetInvMass() const
- {
- return m_invMass;
- }
-
- void internalSetInvMass(const b3Vector3& invMass)
- {
- m_invMass = invMass;
- }
-
- b3Vector3& internalGetPushVelocity()
- {
- return m_pushVelocity;
- }
-
- b3Vector3& internalGetTurnVelocity()
- {
- return m_turnVelocity;
- }
-
- B3_FORCE_INLINE void internalGetVelocityInLocalPointObsolete(const b3Vector3& rel_pos, b3Vector3& velocity ) const
- {
- velocity = m_linearVelocity+m_deltaLinearVelocity + (m_angularVelocity+m_deltaAngularVelocity).cross(rel_pos);
- }
-
- B3_FORCE_INLINE void internalGetAngularVelocity(b3Vector3& angVel) const
- {
- angVel = m_angularVelocity+m_deltaAngularVelocity;
- }
-
-
- //Optimization for the iterative solver: avoid calculating constant terms involving inertia, normal, relative position
- B3_FORCE_INLINE void internalApplyImpulse(const b3Vector3& linearComponent, const b3Vector3& angularComponent,const b3Scalar impulseMagnitude)
- {
- //if (m_originalBody)
- {
- m_deltaLinearVelocity += linearComponent*impulseMagnitude*m_linearFactor;
- m_deltaAngularVelocity += angularComponent*(impulseMagnitude*m_angularFactor);
- }
- }
-
-
-
-
- void writebackVelocity()
- {
- //if (m_originalBody>=0)
- {
- m_linearVelocity +=m_deltaLinearVelocity;
- m_angularVelocity += m_deltaAngularVelocity;
-
- //m_originalBody->setCompanionId(-1);
- }
- }
-
-
- void writebackVelocityAndTransform(b3Scalar timeStep, b3Scalar splitImpulseTurnErp)
- {
- (void) timeStep;
- if (m_originalBody)
- {
- m_linearVelocity += m_deltaLinearVelocity;
- m_angularVelocity += m_deltaAngularVelocity;
-
- //correct the position/orientation based on push/turn recovery
- b3Transform newTransform;
- if (m_pushVelocity[0]!=0.f || m_pushVelocity[1]!=0 || m_pushVelocity[2]!=0 || m_turnVelocity[0]!=0.f || m_turnVelocity[1]!=0 || m_turnVelocity[2]!=0)
- {
- // b3Quaternion orn = m_worldTransform.getRotation();
-// b3TransformUtil::integrateTransform(m_worldTransform,m_pushVelocity,m_turnVelocity*splitImpulseTurnErp,timeStep,newTransform);
-// m_worldTransform = newTransform;
- }
- //m_worldTransform.setRotation(orn);
- //m_originalBody->setCompanionId(-1);
- }
- }
-
-
-
-};
-
-#endif //B3_SOLVER_BODY_H
-
-
/include/Bullet/Bullet3OpenCL/RigidBody/b3GpuSolverBody.h
Property changes:
Deleted: svn:executable
## -1 +0,0 ##
-*
\ No newline at end of property
Index: include/Bullet/Bullet3OpenCL/RigidBody/b3GpuConstraint4.h
===================================================================
--- include/Bullet/Bullet3OpenCL/RigidBody/b3GpuConstraint4.h (revision 22)
+++ include/Bullet/Bullet3OpenCL/RigidBody/b3GpuConstraint4.h (nonexistent)
@@ -1,18 +0,0 @@
-
-#ifndef B3_CONSTRAINT4_h
-#define B3_CONSTRAINT4_h
-#include "Bullet3Common/b3Vector3.h"
-
-#include "Bullet3Dynamics/shared/b3ContactConstraint4.h"
-
-
-B3_ATTRIBUTE_ALIGNED16(struct) b3GpuConstraint4 : public b3ContactConstraint4
-{
- B3_DECLARE_ALIGNED_ALLOCATOR();
-
- inline void setFrictionCoeff(float value) { m_linear[3] = value; }
- inline float getFrictionCoeff() const { return m_linear[3]; }
-};
-
-#endif //B3_CONSTRAINT4_h
-
/include/Bullet/Bullet3OpenCL/RigidBody/b3GpuConstraint4.h
Property changes:
Deleted: svn:executable
## -1 +0,0 ##
-*
\ No newline at end of property
Index: include/Bullet/Bullet3OpenCL/RigidBody/b3GpuPgsConstraintSolver.h
===================================================================
--- include/Bullet/Bullet3OpenCL/RigidBody/b3GpuPgsConstraintSolver.h (revision 22)
+++ include/Bullet/Bullet3OpenCL/RigidBody/b3GpuPgsConstraintSolver.h (nonexistent)
@@ -1,78 +0,0 @@
-/*
-Copyright (c) 2013 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
-
-#ifndef B3_GPU_PGS_CONSTRAINT_SOLVER_H
-#define B3_GPU_PGS_CONSTRAINT_SOLVER_H
-
-struct b3Contact4;
-struct b3ContactPoint;
-
-
-class b3Dispatcher;
-
-#include "Bullet3Dynamics/ConstraintSolver/b3TypedConstraint.h"
-#include "Bullet3Dynamics/ConstraintSolver/b3ContactSolverInfo.h"
-#include "b3GpuSolverBody.h"
-#include "b3GpuSolverConstraint.h"
-#include "Bullet3OpenCL/ParallelPrimitives/b3OpenCLArray.h"
-struct b3RigidBodyData;
-struct b3InertiaData;
-
-#include "Bullet3OpenCL/Initialize/b3OpenCLInclude.h"
-#include "b3GpuGenericConstraint.h"
-
-class b3GpuPgsConstraintSolver
-{
-protected:
- int m_staticIdx;
- struct b3GpuPgsJacobiSolverInternalData* m_gpuData;
- protected:
- b3AlignedObjectArray<b3GpuSolverBody> m_tmpSolverBodyPool;
- b3GpuConstraintArray m_tmpSolverContactConstraintPool;
- b3GpuConstraintArray m_tmpSolverNonContactConstraintPool;
- b3GpuConstraintArray m_tmpSolverContactFrictionConstraintPool;
- b3GpuConstraintArray m_tmpSolverContactRollingFrictionConstraintPool;
-
- b3AlignedObjectArray<unsigned int> m_tmpConstraintSizesPool;
-
-
- bool m_usePgs;
- void averageVelocities();
-
- int m_maxOverrideNumSolverIterations;
-
- int m_numSplitImpulseRecoveries;
-
-// int getOrInitSolverBody(int bodyIndex, b3RigidBodyData* bodies,b3InertiaData* inertias);
- void initSolverBody(int bodyIndex, b3GpuSolverBody* solverBody, b3RigidBodyData* rb);
-
-public:
- b3GpuPgsConstraintSolver (cl_context ctx, cl_device_id device, cl_command_queue queue,bool usePgs);
- virtual~b3GpuPgsConstraintSolver ();
-
- virtual b3Scalar solveGroupCacheFriendlyIterations(b3OpenCLArray<b3GpuGenericConstraint>* gpuConstraints1,int numConstraints,const b3ContactSolverInfo& infoGlobal);
- virtual b3Scalar solveGroupCacheFriendlySetup(b3OpenCLArray<b3RigidBodyData>* gpuBodies, b3OpenCLArray<b3InertiaData>* gpuInertias, int numBodies,b3OpenCLArray<b3GpuGenericConstraint>* gpuConstraints,int numConstraints,const b3ContactSolverInfo& infoGlobal);
- b3Scalar solveGroupCacheFriendlyFinish(b3OpenCLArray<b3RigidBodyData>* gpuBodies,b3OpenCLArray<b3InertiaData>* gpuInertias,int numBodies,b3OpenCLArray<b3GpuGenericConstraint>* gpuConstraints,int numConstraints,const b3ContactSolverInfo& infoGlobal);
-
-
- b3Scalar solveGroup(b3OpenCLArray<b3RigidBodyData>* gpuBodies,b3OpenCLArray<b3InertiaData>* gpuInertias, int numBodies,b3OpenCLArray<b3GpuGenericConstraint>* gpuConstraints,int numConstraints,const b3ContactSolverInfo& infoGlobal);
- void solveJoints(int numBodies, b3OpenCLArray<b3RigidBodyData>* gpuBodies, b3OpenCLArray<b3InertiaData>* gpuInertias,
- int numConstraints, b3OpenCLArray<b3GpuGenericConstraint>* gpuConstraints);
-
- int sortConstraintByBatch3( struct b3BatchConstraint* cs, int numConstraints, int simdWidth , int staticIdx, int numBodies);
- void recomputeBatches();
-};
-
-#endif //B3_GPU_PGS_CONSTRAINT_SOLVER_H
/include/Bullet/Bullet3OpenCL/RigidBody/b3GpuPgsConstraintSolver.h
Property changes:
Deleted: svn:executable
## -1 +0,0 ##
-*
\ No newline at end of property
Index: include/Bullet/Bullet3OpenCL/RigidBody/b3GpuRigidBodyPipelineInternalData.h
===================================================================
--- include/Bullet/Bullet3OpenCL/RigidBody/b3GpuRigidBodyPipelineInternalData.h (revision 22)
+++ include/Bullet/Bullet3OpenCL/RigidBody/b3GpuRigidBodyPipelineInternalData.h (nonexistent)
@@ -1,73 +0,0 @@
-/*
-Copyright (c) 2013 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
-
-#ifndef B3_GPU_RIGIDBODY_PIPELINE_INTERNAL_DATA_H
-#define B3_GPU_RIGIDBODY_PIPELINE_INTERNAL_DATA_H
-
-#include "Bullet3OpenCL/Initialize/b3OpenCLInclude.h"
-#include "Bullet3Common/b3AlignedObjectArray.h"
-
-#include "Bullet3OpenCL/ParallelPrimitives/b3OpenCLArray.h"
-#include "Bullet3Collision/NarrowPhaseCollision/shared/b3Collidable.h"
-
-
-#include "Bullet3OpenCL/BroadphaseCollision/b3SapAabb.h"
-#include "Bullet3Dynamics/ConstraintSolver/b3TypedConstraint.h"
-#include "Bullet3Collision/NarrowPhaseCollision/b3Config.h"
-
-
-
-#include "Bullet3Collision/BroadPhaseCollision/b3OverlappingPair.h"
-#include "Bullet3OpenCL/RigidBody/b3GpuGenericConstraint.h"
-
-struct b3GpuRigidBodyPipelineInternalData
-{
-
- cl_context m_context;
- cl_device_id m_device;
- cl_command_queue m_queue;
-
- cl_kernel m_integrateTransformsKernel;
- cl_kernel m_updateAabbsKernel;
- cl_kernel m_clearOverlappingPairsKernel;
-
- class b3PgsJacobiSolver* m_solver;
-
- class b3GpuPgsConstraintSolver* m_gpuSolver;
-
- class b3GpuPgsContactSolver* m_solver2;
- class b3GpuJacobiContactSolver* m_solver3;
- class b3GpuRaycast* m_raycaster;
-
- class b3GpuBroadphaseInterface* m_broadphaseSap;
-
- struct b3DynamicBvhBroadphase* m_broadphaseDbvt;
- b3OpenCLArray<b3SapAabb>* m_allAabbsGPU;
- b3AlignedObjectArray<b3SapAabb> m_allAabbsCPU;
- b3OpenCLArray<b3BroadphasePair>* m_overlappingPairsGPU;
-
- b3OpenCLArray<b3GpuGenericConstraint>* m_gpuConstraints;
- b3AlignedObjectArray<b3GpuGenericConstraint> m_cpuConstraints;
-
- b3AlignedObjectArray<b3TypedConstraint*> m_joints;
- int m_constraintUid;
- class b3GpuNarrowPhase* m_narrowphase;
- b3Vector3 m_gravity;
-
- b3Config m_config;
-};
-
-#endif //B3_GPU_RIGIDBODY_PIPELINE_INTERNAL_DATA_H
-
/include/Bullet/Bullet3OpenCL/RigidBody/b3GpuRigidBodyPipelineInternalData.h
Property changes:
Deleted: svn:executable
## -1 +0,0 ##
-*
\ No newline at end of property
Index: include/Bullet/Bullet3OpenCL/RigidBody/b3GpuNarrowPhase.h
===================================================================
--- include/Bullet/Bullet3OpenCL/RigidBody/b3GpuNarrowPhase.h (revision 22)
+++ include/Bullet/Bullet3OpenCL/RigidBody/b3GpuNarrowPhase.h (nonexistent)
@@ -1,109 +0,0 @@
-#ifndef B3_GPU_NARROWPHASE_H
-#define B3_GPU_NARROWPHASE_H
-
-#include "Bullet3Collision/NarrowPhaseCollision/shared/b3Collidable.h"
-#include "Bullet3OpenCL/Initialize/b3OpenCLInclude.h"
-#include "Bullet3Common/b3AlignedObjectArray.h"
-#include "Bullet3Common/b3Vector3.h"
-
-class b3GpuNarrowPhase
-{
-protected:
-
- struct b3GpuNarrowPhaseInternalData* m_data;
- int m_acceleratedCompanionShapeIndex;
- int m_planeBodyIndex;
- int m_static0Index;
-
- cl_context m_context;
- cl_device_id m_device;
- cl_command_queue m_queue;
-
- int registerConvexHullShapeInternal(class b3ConvexUtility* convexPtr, b3Collidable& col);
- int registerConcaveMeshShape(b3AlignedObjectArray<b3Vector3>* vertices, b3AlignedObjectArray<int>* indices, b3Collidable& col, const float* scaling);
-
-public:
-
-
-
-
- b3GpuNarrowPhase(cl_context vtx, cl_device_id dev, cl_command_queue q, const struct b3Config& config);
-
- virtual ~b3GpuNarrowPhase(void);
-
- int registerSphereShape(float radius);
- int registerPlaneShape(const b3Vector3& planeNormal, float planeConstant);
-
- int registerCompoundShape(b3AlignedObjectArray<b3GpuChildShape>* childShapes);
- int registerFace(const b3Vector3& faceNormal, float faceConstant);
-
- int registerConcaveMesh(b3AlignedObjectArray<b3Vector3>* vertices, b3AlignedObjectArray<int>* indices,const float* scaling);
-
- //do they need to be merged?
-
- int registerConvexHullShape(b3ConvexUtility* utilPtr);
- int registerConvexHullShape(const float* vertices, int strideInBytes, int numVertices, const float* scaling);
-
- int registerRigidBody(int collidableIndex, float mass, const float* position, const float* orientation, const float* aabbMin, const float* aabbMax,bool writeToGpu);
- void setObjectTransform(const float* position, const float* orientation , int bodyIndex);
-
- void writeAllBodiesToGpu();
- void reset();
- void readbackAllBodiesToCpu();
- bool getObjectTransformFromCpu(float* position, float* orientation , int bodyIndex) const;
-
- void setObjectTransformCpu(float* position, float* orientation , int bodyIndex);
- void setObjectVelocityCpu(float* linVel, float* angVel, int bodyIndex);
-
-
- virtual void computeContacts(cl_mem broadphasePairs, int numBroadphasePairs, cl_mem aabbsWorldSpace, int numObjects);
-
-
- cl_mem getBodiesGpu();
- const struct b3RigidBodyData* getBodiesCpu() const;
- //struct b3RigidBodyData* getBodiesCpu();
-
- int getNumBodiesGpu() const;
-
- cl_mem getBodyInertiasGpu();
- int getNumBodyInertiasGpu() const;
-
- cl_mem getCollidablesGpu();
- const struct b3Collidable* getCollidablesCpu() const;
- int getNumCollidablesGpu() const;
-
- const struct b3SapAabb* getLocalSpaceAabbsCpu() const;
-
- const struct b3Contact4* getContactsCPU() const;
-
- cl_mem getContactsGpu();
- int getNumContactsGpu() const;
-
- cl_mem getAabbLocalSpaceBufferGpu();
-
- int getNumRigidBodies() const;
-
- int allocateCollidable();
-
- int getStatic0Index() const
- {
- return m_static0Index;
- }
- b3Collidable& getCollidableCpu(int collidableIndex);
- const b3Collidable& getCollidableCpu(int collidableIndex) const;
-
- const b3GpuNarrowPhaseInternalData* getInternalData() const
- {
- return m_data;
- }
-
- b3GpuNarrowPhaseInternalData* getInternalData()
- {
- return m_data;
- }
-
- const struct b3SapAabb& getLocalSpaceAabb(int collidableIndex) const;
-};
-
-#endif //B3_GPU_NARROWPHASE_H
-
/include/Bullet/Bullet3OpenCL/RigidBody/b3GpuNarrowPhase.h
Property changes:
Deleted: svn:executable
## -1 +0,0 ##
-*
\ No newline at end of property
Index: include/Bullet/Bullet3OpenCL/RigidBody/kernels/batchingKernelsNew.h
===================================================================
--- include/Bullet/Bullet3OpenCL/RigidBody/kernels/batchingKernelsNew.h (revision 22)
+++ include/Bullet/Bullet3OpenCL/RigidBody/kernels/batchingKernelsNew.h (nonexistent)
@@ -1,290 +0,0 @@
-//this file is autogenerated using stringify.bat (premake --stringify) in the build folder of this project
-static const char* batchingKernelsNewCL= \
-"/*\n"
-"Copyright (c) 2012 Advanced Micro Devices, Inc. \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"
-"//Originally written by Erwin Coumans\n"
-"#ifndef B3_CONTACT4DATA_H\n"
-"#define B3_CONTACT4DATA_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"
-"//keep B3_LARGE_FLOAT*B3_LARGE_FLOAT < FLT_MAX\n"
-"#define B3_LARGE_FLOAT 1e18f\n"
-"#define B3_INFINITY 1e18f\n"
-"#define b3Assert(a)\n"
-"#define b3ConstArray(a) __global const a*\n"
-"#define b3AtomicInc atomic_inc\n"
-"#define b3AtomicAdd atomic_add\n"
-"#define b3Fabs fabs\n"
-"#define b3Sqrt native_sqrt\n"
-"#define b3Sin native_sin\n"
-"#define b3Cos native_cos\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"
-" #define b3MinFloat4 min\n"
-" #define b3MaxFloat4 max\n"
-" #define b3Normalized(a) normalize(a)\n"
-"#endif \n"
-" \n"
-"inline bool b3IsAlmostZero(b3Float4ConstArg v)\n"
-"{\n"
-" if(b3Fabs(v.x)>1e-6 || b3Fabs(v.y)>1e-6 || b3Fabs(v.z)>1e-6) \n"
-" return false;\n"
-" return true;\n"
-"}\n"
-"inline int b3MaxDot( b3Float4ConstArg vec, __global const b3Float4* vecArray, int vecLen, float* dotOut )\n"
-"{\n"
-" float maxDot = -B3_INFINITY;\n"
-" int i = 0;\n"
-" int ptIndex = -1;\n"
-" for( i = 0; i < vecLen; i++ )\n"
-" {\n"
-" float dot = b3Dot3F4(vecArray[i],vec);\n"
-" \n"
-" if( dot > maxDot )\n"
-" {\n"
-" maxDot = dot;\n"
-" ptIndex = i;\n"
-" }\n"
-" }\n"
-" b3Assert(ptIndex>=0);\n"
-" if (ptIndex<0)\n"
-" {\n"
-" ptIndex = 0;\n"
-" }\n"
-" *dotOut = maxDot;\n"
-" return ptIndex;\n"
-"}\n"
-"#endif //B3_FLOAT4_H\n"
-"typedef struct b3Contact4Data b3Contact4Data_t;\n"
-"struct b3Contact4Data\n"
-"{\n"
-" b3Float4 m_worldPosB[4];\n"
-"// b3Float4 m_localPosA[4];\n"
-"// b3Float4 m_localPosB[4];\n"
-" b3Float4 m_worldNormalOnB; // w: m_nPoints\n"
-" unsigned short m_restituitionCoeffCmp;\n"
-" unsigned short m_frictionCoeffCmp;\n"
-" int m_batchIdx;\n"
-" int m_bodyAPtrAndSignBit;//x:m_bodyAPtr, y:m_bodyBPtr\n"
-" int m_bodyBPtrAndSignBit;\n"
-" int m_childIndexA;\n"
-" int m_childIndexB;\n"
-" int m_unused1;\n"
-" int m_unused2;\n"
-"};\n"
-"inline int b3Contact4Data_getNumPoints(const struct b3Contact4Data* contact)\n"
-"{\n"
-" return (int)contact->m_worldNormalOnB.w;\n"
-"};\n"
-"inline void b3Contact4Data_setNumPoints(struct b3Contact4Data* contact, int numPoints)\n"
-"{\n"
-" contact->m_worldNormalOnB.w = (float)numPoints;\n"
-"};\n"
-"#endif //B3_CONTACT4DATA_H\n"
-"#pragma OPENCL EXTENSION cl_amd_printf : enable\n"
-"#pragma OPENCL EXTENSION cl_khr_local_int32_base_atomics : enable\n"
-"#pragma OPENCL EXTENSION cl_khr_global_int32_base_atomics : enable\n"
-"#pragma OPENCL EXTENSION cl_khr_local_int32_extended_atomics : enable\n"
-"#pragma OPENCL EXTENSION cl_khr_global_int32_extended_atomics : enable\n"
-"#ifdef cl_ext_atomic_counters_32\n"
-"#pragma OPENCL EXTENSION cl_ext_atomic_counters_32 : enable\n"
-"#else\n"
-"#define counter32_t volatile __global int*\n"
-"#endif\n"
-"#define SIMD_WIDTH 64\n"
-"typedef unsigned int u32;\n"
-"typedef unsigned short u16;\n"
-"typedef unsigned char u8;\n"
-"#define GET_GROUP_IDX get_group_id(0)\n"
-"#define GET_LOCAL_IDX get_local_id(0)\n"
-"#define GET_GLOBAL_IDX get_global_id(0)\n"
-"#define GET_GROUP_SIZE get_local_size(0)\n"
-"#define GET_NUM_GROUPS get_num_groups(0)\n"
-"#define GROUP_LDS_BARRIER barrier(CLK_LOCAL_MEM_FENCE)\n"
-"#define GROUP_MEM_FENCE mem_fence(CLK_LOCAL_MEM_FENCE)\n"
-"#define AtomInc(x) atom_inc(&(x))\n"
-"#define AtomInc1(x, out) out = atom_inc(&(x))\n"
-"#define AppendInc(x, out) out = atomic_inc(x)\n"
-"#define AtomAdd(x, value) atom_add(&(x), value)\n"
-"#define AtomCmpxhg(x, cmp, value) atom_cmpxchg( &(x), cmp, value )\n"
-"#define AtomXhg(x, value) atom_xchg ( &(x), value )\n"
-"#define SELECT_UINT4( b, a, condition ) select( b,a,condition )\n"
-"#define make_float4 (float4)\n"
-"#define make_float2 (float2)\n"
-"#define make_uint4 (uint4)\n"
-"#define make_int4 (int4)\n"
-"#define make_uint2 (uint2)\n"
-"#define make_int2 (int2)\n"
-"#define max2 max\n"
-"#define min2 min\n"
-"#define WG_SIZE 64\n"
-"typedef struct \n"
-"{\n"
-" int m_n;\n"
-" int m_start;\n"
-" int m_staticIdx;\n"
-" int m_paddings[1];\n"
-"} ConstBuffer;\n"
-"typedef struct \n"
-"{\n"
-" int m_a;\n"
-" int m_b;\n"
-" u32 m_idx;\n"
-"}Elem;\n"
-"// batching on the GPU\n"
-"__kernel void CreateBatchesBruteForce( __global struct b3Contact4Data* gConstraints, __global const u32* gN, __global const u32* gStart, int m_staticIdx )\n"
-"{\n"
-" int wgIdx = GET_GROUP_IDX;\n"
-" int lIdx = GET_LOCAL_IDX;\n"
-" \n"
-" const int m_n = gN[wgIdx];\n"
-" const int m_start = gStart[wgIdx];\n"
-" \n"
-" if( lIdx == 0 )\n"
-" {\n"
-" for (int i=0;i<m_n;i++)\n"
-" {\n"
-" int srcIdx = i+m_start;\n"
-" int batchIndex = i;\n"
-" gConstraints[ srcIdx ].m_batchIdx = batchIndex; \n"
-" }\n"
-" }\n"
-"}\n"
-"#define CHECK_SIZE (WG_SIZE)\n"
-"u32 readBuf(__local u32* buff, int idx)\n"
-"{\n"
-" idx = idx % (32*CHECK_SIZE);\n"
-" int bitIdx = idx%32;\n"
-" int bufIdx = idx/32;\n"
-" return buff[bufIdx] & (1<<bitIdx);\n"
-"}\n"
-"void writeBuf(__local u32* buff, int idx)\n"
-"{\n"
-" idx = idx % (32*CHECK_SIZE);\n"
-" int bitIdx = idx%32;\n"
-" int bufIdx = idx/32;\n"
-" buff[bufIdx] |= (1<<bitIdx);\n"
-" //atom_or( &buff[bufIdx], (1<<bitIdx) );\n"
-"}\n"
-"u32 tryWrite(__local u32* buff, int idx)\n"
-"{\n"
-" idx = idx % (32*CHECK_SIZE);\n"
-" int bitIdx = idx%32;\n"
-" int bufIdx = idx/32;\n"
-" u32 ans = (u32)atom_or( &buff[bufIdx], (1<<bitIdx) );\n"
-" return ((ans >> bitIdx)&1) == 0;\n"
-"}\n"
-"// batching on the GPU\n"
-"__kernel void CreateBatchesNew( __global struct b3Contact4Data* gConstraints, __global const u32* gN, __global const u32* gStart, __global int* batchSizes, int staticIdx )\n"
-"{\n"
-" int wgIdx = GET_GROUP_IDX;\n"
-" int lIdx = GET_LOCAL_IDX;\n"
-" const int numConstraints = gN[wgIdx];\n"
-" const int m_start = gStart[wgIdx];\n"
-" b3Contact4Data_t tmp;\n"
-" \n"
-" __local u32 ldsFixedBuffer[CHECK_SIZE];\n"
-" \n"
-" \n"
-" \n"
-" \n"
-" \n"
-" if( lIdx == 0 )\n"
-" {\n"
-" \n"
-" \n"
-" __global struct b3Contact4Data* cs = &gConstraints[m_start]; \n"
-" \n"
-" \n"
-" int numValidConstraints = 0;\n"
-" int batchIdx = 0;\n"
-" while( numValidConstraints < numConstraints)\n"
-" {\n"
-" int nCurrentBatch = 0;\n"
-" // clear flag\n"
-" \n"
-" for(int i=0; i<CHECK_SIZE; i++) \n"
-" ldsFixedBuffer[i] = 0; \n"
-" for(int i=numValidConstraints; i<numConstraints; i++)\n"
-" {\n"
-" int bodyAS = cs[i].m_bodyAPtrAndSignBit;\n"
-" int bodyBS = cs[i].m_bodyBPtrAndSignBit;\n"
-" int bodyA = abs(bodyAS);\n"
-" int bodyB = abs(bodyBS);\n"
-" bool aIsStatic = (bodyAS<0) || bodyAS==staticIdx;\n"
-" bool bIsStatic = (bodyBS<0) || bodyBS==staticIdx;\n"
-" int aUnavailable = aIsStatic ? 0 : readBuf( ldsFixedBuffer, bodyA);\n"
-" int bUnavailable = bIsStatic ? 0 : readBuf( ldsFixedBuffer, bodyB);\n"
-" \n"
-" if( aUnavailable==0 && bUnavailable==0 ) // ok\n"
-" {\n"
-" if (!aIsStatic)\n"
-" {\n"
-" writeBuf( ldsFixedBuffer, bodyA );\n"
-" }\n"
-" if (!bIsStatic)\n"
-" {\n"
-" writeBuf( ldsFixedBuffer, bodyB );\n"
-" }\n"
-" cs[i].m_batchIdx = batchIdx;\n"
-" if (i!=numValidConstraints)\n"
-" {\n"
-" tmp = cs[i];\n"
-" cs[i] = cs[numValidConstraints];\n"
-" cs[numValidConstraints] = tmp;\n"
-" }\n"
-" numValidConstraints++;\n"
-" \n"
-" nCurrentBatch++;\n"
-" if( nCurrentBatch == SIMD_WIDTH)\n"
-" {\n"
-" nCurrentBatch = 0;\n"
-" for(int i=0; i<CHECK_SIZE; i++) \n"
-" ldsFixedBuffer[i] = 0;\n"
-" \n"
-" }\n"
-" }\n"
-" }//for\n"
-" batchIdx ++;\n"
-" }//while\n"
-" \n"
-" batchSizes[wgIdx] = batchIdx;\n"
-" }//if( lIdx == 0 )\n"
-" \n"
-" //return batchIdx;\n"
-"}\n"
-;
/include/Bullet/Bullet3OpenCL/RigidBody/kernels/batchingKernelsNew.h
Property changes:
Deleted: svn:executable
## -1 +0,0 ##
-*
\ No newline at end of property
Index: include/Bullet/Bullet3OpenCL/RigidBody/kernels/solverSetup2.h
===================================================================
--- include/Bullet/Bullet3OpenCL/RigidBody/kernels/solverSetup2.h (revision 22)
+++ include/Bullet/Bullet3OpenCL/RigidBody/kernels/solverSetup2.h (nonexistent)
@@ -1,600 +0,0 @@
-//this file is autogenerated using stringify.bat (premake --stringify) in the build folder of this project
-static const char* solverSetup2CL= \
-"/*\n"
-"Copyright (c) 2012 Advanced Micro Devices, Inc. \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"
-"//Originally written by Takahiro Harada\n"
-"#ifndef B3_CONTACT4DATA_H\n"
-"#define B3_CONTACT4DATA_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"
-"//keep B3_LARGE_FLOAT*B3_LARGE_FLOAT < FLT_MAX\n"
-"#define B3_LARGE_FLOAT 1e18f\n"
-"#define B3_INFINITY 1e18f\n"
-"#define b3Assert(a)\n"
-"#define b3ConstArray(a) __global const a*\n"
-"#define b3AtomicInc atomic_inc\n"
-"#define b3AtomicAdd atomic_add\n"
-"#define b3Fabs fabs\n"
-"#define b3Sqrt native_sqrt\n"
-"#define b3Sin native_sin\n"
-"#define b3Cos native_cos\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"
-" #define b3MinFloat4 min\n"
-" #define b3MaxFloat4 max\n"
-" #define b3Normalized(a) normalize(a)\n"
-"#endif \n"
-" \n"
-"inline bool b3IsAlmostZero(b3Float4ConstArg v)\n"
-"{\n"
-" if(b3Fabs(v.x)>1e-6 || b3Fabs(v.y)>1e-6 || b3Fabs(v.z)>1e-6) \n"
-" return false;\n"
-" return true;\n"
-"}\n"
-"inline int b3MaxDot( b3Float4ConstArg vec, __global const b3Float4* vecArray, int vecLen, float* dotOut )\n"
-"{\n"
-" float maxDot = -B3_INFINITY;\n"
-" int i = 0;\n"
-" int ptIndex = -1;\n"
-" for( i = 0; i < vecLen; i++ )\n"
-" {\n"
-" float dot = b3Dot3F4(vecArray[i],vec);\n"
-" \n"
-" if( dot > maxDot )\n"
-" {\n"
-" maxDot = dot;\n"
-" ptIndex = i;\n"
-" }\n"
-" }\n"
-" b3Assert(ptIndex>=0);\n"
-" if (ptIndex<0)\n"
-" {\n"
-" ptIndex = 0;\n"
-" }\n"
-" *dotOut = maxDot;\n"
-" return ptIndex;\n"
-"}\n"
-"#endif //B3_FLOAT4_H\n"
-"typedef struct b3Contact4Data b3Contact4Data_t;\n"
-"struct b3Contact4Data\n"
-"{\n"
-" b3Float4 m_worldPosB[4];\n"
-"// b3Float4 m_localPosA[4];\n"
-"// b3Float4 m_localPosB[4];\n"
-" b3Float4 m_worldNormalOnB; // w: m_nPoints\n"
-" unsigned short m_restituitionCoeffCmp;\n"
-" unsigned short m_frictionCoeffCmp;\n"
-" int m_batchIdx;\n"
-" int m_bodyAPtrAndSignBit;//x:m_bodyAPtr, y:m_bodyBPtr\n"
-" int m_bodyBPtrAndSignBit;\n"
-" int m_childIndexA;\n"
-" int m_childIndexB;\n"
-" int m_unused1;\n"
-" int m_unused2;\n"
-"};\n"
-"inline int b3Contact4Data_getNumPoints(const struct b3Contact4Data* contact)\n"
-"{\n"
-" return (int)contact->m_worldNormalOnB.w;\n"
-"};\n"
-"inline void b3Contact4Data_setNumPoints(struct b3Contact4Data* contact, int numPoints)\n"
-"{\n"
-" contact->m_worldNormalOnB.w = (float)numPoints;\n"
-"};\n"
-"#endif //B3_CONTACT4DATA_H\n"
-"#pragma OPENCL EXTENSION cl_amd_printf : enable\n"
-"#pragma OPENCL EXTENSION cl_khr_local_int32_base_atomics : enable\n"
-"#pragma OPENCL EXTENSION cl_khr_global_int32_base_atomics : enable\n"
-"#pragma OPENCL EXTENSION cl_khr_local_int32_extended_atomics : enable\n"
-"#pragma OPENCL EXTENSION cl_khr_global_int32_extended_atomics : enable\n"
-"#ifdef cl_ext_atomic_counters_32\n"
-"#pragma OPENCL EXTENSION cl_ext_atomic_counters_32 : enable\n"
-"#else\n"
-"#define counter32_t volatile global int*\n"
-"#endif\n"
-"typedef unsigned int u32;\n"
-"typedef unsigned short u16;\n"
-"typedef unsigned char u8;\n"
-"#define GET_GROUP_IDX get_group_id(0)\n"
-"#define GET_LOCAL_IDX get_local_id(0)\n"
-"#define GET_GLOBAL_IDX get_global_id(0)\n"
-"#define GET_GROUP_SIZE get_local_size(0)\n"
-"#define GET_NUM_GROUPS get_num_groups(0)\n"
-"#define GROUP_LDS_BARRIER barrier(CLK_LOCAL_MEM_FENCE)\n"
-"#define GROUP_MEM_FENCE mem_fence(CLK_LOCAL_MEM_FENCE)\n"
-"#define AtomInc(x) atom_inc(&(x))\n"
-"#define AtomInc1(x, out) out = atom_inc(&(x))\n"
-"#define AppendInc(x, out) out = atomic_inc(x)\n"
-"#define AtomAdd(x, value) atom_add(&(x), value)\n"
-"#define AtomCmpxhg(x, cmp, value) atom_cmpxchg( &(x), cmp, value )\n"
-"#define AtomXhg(x, value) atom_xchg ( &(x), value )\n"
-"#define SELECT_UINT4( b, a, condition ) select( b,a,condition )\n"
-"#define make_float4 (float4)\n"
-"#define make_float2 (float2)\n"
-"#define make_uint4 (uint4)\n"
-"#define make_int4 (int4)\n"
-"#define make_uint2 (uint2)\n"
-"#define make_int2 (int2)\n"
-"#define max2 max\n"
-"#define min2 min\n"
-"///////////////////////////////////////\n"
-"// Vector\n"
-"///////////////////////////////////////\n"
-"__inline\n"
-"float fastDiv(float numerator, float denominator)\n"
-"{\n"
-" return native_divide(numerator, denominator); \n"
-"// return numerator/denominator; \n"
-"}\n"
-"__inline\n"
-"float4 fastDiv4(float4 numerator, float4 denominator)\n"
-"{\n"
-" return native_divide(numerator, denominator); \n"
-"}\n"
-"__inline\n"
-"float fastSqrtf(float f2)\n"
-"{\n"
-" return native_sqrt(f2);\n"
-"// return sqrt(f2);\n"
-"}\n"
-"__inline\n"
-"float fastRSqrt(float f2)\n"
-"{\n"
-" return native_rsqrt(f2);\n"
-"}\n"
-"__inline\n"
-"float fastLength4(float4 v)\n"
-"{\n"
-" return fast_length(v);\n"
-"}\n"
-"__inline\n"
-"float4 fastNormalize4(float4 v)\n"
-"{\n"
-" return fast_normalize(v);\n"
-"}\n"
-"__inline\n"
-"float sqrtf(float a)\n"
-"{\n"
-"// return sqrt(a);\n"
-" return native_sqrt(a);\n"
-"}\n"
-"__inline\n"
-"float4 cross3(float4 a, float4 b)\n"
-"{\n"
-" return cross(a,b);\n"
-"}\n"
-"__inline\n"
-"float dot3F4(float4 a, float4 b)\n"
-"{\n"
-" float4 a1 = make_float4(a.xyz,0.f);\n"
-" float4 b1 = make_float4(b.xyz,0.f);\n"
-" return dot(a1, b1);\n"
-"}\n"
-"__inline\n"
-"float length3(const float4 a)\n"
-"{\n"
-" return sqrtf(dot3F4(a,a));\n"
-"}\n"
-"__inline\n"
-"float dot4(const float4 a, const float4 b)\n"
-"{\n"
-" return dot( a, b );\n"
-"}\n"
-"// for height\n"
-"__inline\n"
-"float dot3w1(const float4 point, const float4 eqn)\n"
-"{\n"
-" return dot3F4(point,eqn) + eqn.w;\n"
-"}\n"
-"__inline\n"
-"float4 normalize3(const float4 a)\n"
-"{\n"
-" float4 n = make_float4(a.x, a.y, a.z, 0.f);\n"
-" return fastNormalize4( n );\n"
-"// float length = sqrtf(dot3F4(a, a));\n"
-"// return 1.f/length * a;\n"
-"}\n"
-"__inline\n"
-"float4 normalize4(const float4 a)\n"
-"{\n"
-" float length = sqrtf(dot4(a, a));\n"
-" return 1.f/length * a;\n"
-"}\n"
-"__inline\n"
-"float4 createEquation(const float4 a, const float4 b, const float4 c)\n"
-"{\n"
-" float4 eqn;\n"
-" float4 ab = b-a;\n"
-" float4 ac = c-a;\n"
-" eqn = normalize3( cross3(ab, ac) );\n"
-" eqn.w = -dot3F4(eqn,a);\n"
-" return eqn;\n"
-"}\n"
-"///////////////////////////////////////\n"
-"// Matrix3x3\n"
-"///////////////////////////////////////\n"
-"typedef struct\n"
-"{\n"
-" float4 m_row[3];\n"
-"}Matrix3x3;\n"
-"__inline\n"
-"Matrix3x3 mtZero();\n"
-"__inline\n"
-"Matrix3x3 mtIdentity();\n"
-"__inline\n"
-"Matrix3x3 mtTranspose(Matrix3x3 m);\n"
-"__inline\n"
-"Matrix3x3 mtMul(Matrix3x3 a, Matrix3x3 b);\n"
-"__inline\n"
-"float4 mtMul1(Matrix3x3 a, float4 b);\n"
-"__inline\n"
-"float4 mtMul3(float4 a, Matrix3x3 b);\n"
-"__inline\n"
-"Matrix3x3 mtZero()\n"
-"{\n"
-" Matrix3x3 m;\n"
-" m.m_row[0] = (float4)(0.f);\n"
-" m.m_row[1] = (float4)(0.f);\n"
-" m.m_row[2] = (float4)(0.f);\n"
-" return m;\n"
-"}\n"
-"__inline\n"
-"Matrix3x3 mtIdentity()\n"
-"{\n"
-" Matrix3x3 m;\n"
-" m.m_row[0] = (float4)(1,0,0,0);\n"
-" m.m_row[1] = (float4)(0,1,0,0);\n"
-" m.m_row[2] = (float4)(0,0,1,0);\n"
-" return m;\n"
-"}\n"
-"__inline\n"
-"Matrix3x3 mtTranspose(Matrix3x3 m)\n"
-"{\n"
-" Matrix3x3 out;\n"
-" out.m_row[0] = (float4)(m.m_row[0].x, m.m_row[1].x, m.m_row[2].x, 0.f);\n"
-" out.m_row[1] = (float4)(m.m_row[0].y, m.m_row[1].y, m.m_row[2].y, 0.f);\n"
-" out.m_row[2] = (float4)(m.m_row[0].z, m.m_row[1].z, m.m_row[2].z, 0.f);\n"
-" return out;\n"
-"}\n"
-"__inline\n"
-"Matrix3x3 mtMul(Matrix3x3 a, Matrix3x3 b)\n"
-"{\n"
-" Matrix3x3 transB;\n"
-" transB = mtTranspose( b );\n"
-" Matrix3x3 ans;\n"
-" // why this doesn't run when 0ing in the for{}\n"
-" a.m_row[0].w = 0.f;\n"
-" a.m_row[1].w = 0.f;\n"
-" a.m_row[2].w = 0.f;\n"
-" for(int i=0; i<3; i++)\n"
-" {\n"
-"// a.m_row[i].w = 0.f;\n"
-" ans.m_row[i].x = dot3F4(a.m_row[i],transB.m_row[0]);\n"
-" ans.m_row[i].y = dot3F4(a.m_row[i],transB.m_row[1]);\n"
-" ans.m_row[i].z = dot3F4(a.m_row[i],transB.m_row[2]);\n"
-" ans.m_row[i].w = 0.f;\n"
-" }\n"
-" return ans;\n"
-"}\n"
-"__inline\n"
-"float4 mtMul1(Matrix3x3 a, float4 b)\n"
-"{\n"
-" float4 ans;\n"
-" ans.x = dot3F4( a.m_row[0], b );\n"
-" ans.y = dot3F4( a.m_row[1], b );\n"
-" ans.z = dot3F4( a.m_row[2], b );\n"
-" ans.w = 0.f;\n"
-" return ans;\n"
-"}\n"
-"__inline\n"
-"float4 mtMul3(float4 a, Matrix3x3 b)\n"
-"{\n"
-" float4 colx = make_float4(b.m_row[0].x, b.m_row[1].x, b.m_row[2].x, 0);\n"
-" float4 coly = make_float4(b.m_row[0].y, b.m_row[1].y, b.m_row[2].y, 0);\n"
-" float4 colz = make_float4(b.m_row[0].z, b.m_row[1].z, b.m_row[2].z, 0);\n"
-" float4 ans;\n"
-" ans.x = dot3F4( a, colx );\n"
-" ans.y = dot3F4( a, coly );\n"
-" ans.z = dot3F4( a, colz );\n"
-" return ans;\n"
-"}\n"
-"///////////////////////////////////////\n"
-"// Quaternion\n"
-"///////////////////////////////////////\n"
-"typedef float4 Quaternion;\n"
-"__inline\n"
-"Quaternion qtMul(Quaternion a, Quaternion b);\n"
-"__inline\n"
-"Quaternion qtNormalize(Quaternion in);\n"
-"__inline\n"
-"float4 qtRotate(Quaternion q, float4 vec);\n"
-"__inline\n"
-"Quaternion qtInvert(Quaternion q);\n"
-"__inline\n"
-"Quaternion qtMul(Quaternion a, Quaternion b)\n"
-"{\n"
-" Quaternion ans;\n"
-" ans = cross3( 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 - dot3F4(a, b);\n"
-" return ans;\n"
-"}\n"
-"__inline\n"
-"Quaternion qtNormalize(Quaternion in)\n"
-"{\n"
-" return fastNormalize4(in);\n"
-"// in /= length( in );\n"
-"// return in;\n"
-"}\n"
-"__inline\n"
-"float4 qtRotate(Quaternion q, float4 vec)\n"
-"{\n"
-" Quaternion qInv = qtInvert( q );\n"
-" float4 vcpy = vec;\n"
-" vcpy.w = 0.f;\n"
-" float4 out = qtMul(qtMul(q,vcpy),qInv);\n"
-" return out;\n"
-"}\n"
-"__inline\n"
-"Quaternion qtInvert(Quaternion q)\n"
-"{\n"
-" return (Quaternion)(-q.xyz, q.w);\n"
-"}\n"
-"__inline\n"
-"float4 qtInvRotate(const Quaternion q, float4 vec)\n"
-"{\n"
-" return qtRotate( qtInvert( q ), vec );\n"
-"}\n"
-"#define WG_SIZE 64\n"
-"typedef struct\n"
-"{\n"
-" float4 m_pos;\n"
-" Quaternion m_quat;\n"
-" float4 m_linVel;\n"
-" float4 m_angVel;\n"
-" u32 m_shapeIdx;\n"
-" float m_invMass;\n"
-" float m_restituitionCoeff;\n"
-" float m_frictionCoeff;\n"
-"} Body;\n"
-"typedef struct\n"
-"{\n"
-" Matrix3x3 m_invInertia;\n"
-" Matrix3x3 m_initInvInertia;\n"
-"} Shape;\n"
-"typedef struct\n"
-"{\n"
-" float4 m_linear;\n"
-" float4 m_worldPos[4];\n"
-" float4 m_center; \n"
-" float m_jacCoeffInv[4];\n"
-" float m_b[4];\n"
-" float m_appliedRambdaDt[4];\n"
-" float m_fJacCoeffInv[2]; \n"
-" float m_fAppliedRambdaDt[2]; \n"
-" u32 m_bodyA;\n"
-" u32 m_bodyB;\n"
-" int m_batchIdx;\n"
-" u32 m_paddings[1];\n"
-"} Constraint4;\n"
-"typedef struct\n"
-"{\n"
-" int m_nConstraints;\n"
-" int m_start;\n"
-" int m_batchIdx;\n"
-" int m_nSplit;\n"
-"// int m_paddings[1];\n"
-"} ConstBuffer;\n"
-"typedef struct\n"
-"{\n"
-" int m_solveFriction;\n"
-" int m_maxBatch; // long batch really kills the performance\n"
-" int m_batchIdx;\n"
-" int m_nSplit;\n"
-"// int m_paddings[1];\n"
-"} ConstBufferBatchSolve;\n"
-" \n"
-"typedef struct \n"
-"{\n"
-" int m_valInt0;\n"
-" int m_valInt1;\n"
-" int m_valInt2;\n"
-" int m_valInt3;\n"
-" float m_val0;\n"
-" float m_val1;\n"
-" float m_val2;\n"
-" float m_val3;\n"
-"} SolverDebugInfo;\n"
-"// others\n"
-"__kernel\n"
-"__attribute__((reqd_work_group_size(WG_SIZE,1,1)))\n"
-"void ReorderContactKernel(__global struct b3Contact4Data* in, __global struct b3Contact4Data* out, __global int2* sortData, int4 cb )\n"
-"{\n"
-" int nContacts = cb.x;\n"
-" int gIdx = GET_GLOBAL_IDX;\n"
-" if( gIdx < nContacts )\n"
-" {\n"
-" int srcIdx = sortData[gIdx].y;\n"
-" out[gIdx] = in[srcIdx];\n"
-" }\n"
-"}\n"
-"__kernel __attribute__((reqd_work_group_size(WG_SIZE,1,1)))\n"
-"void SetDeterminismSortDataChildShapeB(__global struct b3Contact4Data* contactsIn, __global int2* sortDataOut, int nContacts)\n"
-"{\n"
-" int gIdx = GET_GLOBAL_IDX;\n"
-" if( gIdx < nContacts )\n"
-" {\n"
-" int2 sd;\n"
-" sd.x = contactsIn[gIdx].m_childIndexB;\n"
-" sd.y = gIdx;\n"
-" sortDataOut[gIdx] = sd;\n"
-" }\n"
-"}\n"
-"__kernel __attribute__((reqd_work_group_size(WG_SIZE,1,1)))\n"
-"void SetDeterminismSortDataChildShapeA(__global struct b3Contact4Data* contactsIn, __global int2* sortDataInOut, int nContacts)\n"
-"{\n"
-" int gIdx = GET_GLOBAL_IDX;\n"
-" if( gIdx < nContacts )\n"
-" {\n"
-" int2 sdIn;\n"
-" sdIn = sortDataInOut[gIdx];\n"
-" int2 sdOut;\n"
-" sdOut.x = contactsIn[sdIn.y].m_childIndexA;\n"
-" sdOut.y = sdIn.y;\n"
-" sortDataInOut[gIdx] = sdOut;\n"
-" }\n"
-"}\n"
-"__kernel __attribute__((reqd_work_group_size(WG_SIZE,1,1)))\n"
-"void SetDeterminismSortDataBodyA(__global struct b3Contact4Data* contactsIn, __global int2* sortDataInOut, int nContacts)\n"
-"{\n"
-" int gIdx = GET_GLOBAL_IDX;\n"
-" if( gIdx < nContacts )\n"
-" {\n"
-" int2 sdIn;\n"
-" sdIn = sortDataInOut[gIdx];\n"
-" int2 sdOut;\n"
-" sdOut.x = contactsIn[sdIn.y].m_bodyAPtrAndSignBit;\n"
-" sdOut.y = sdIn.y;\n"
-" sortDataInOut[gIdx] = sdOut;\n"
-" }\n"
-"}\n"
-"__kernel\n"
-"__attribute__((reqd_work_group_size(WG_SIZE,1,1)))\n"
-"void SetDeterminismSortDataBodyB(__global struct b3Contact4Data* contactsIn, __global int2* sortDataInOut, int nContacts)\n"
-"{\n"
-" int gIdx = GET_GLOBAL_IDX;\n"
-" if( gIdx < nContacts )\n"
-" {\n"
-" int2 sdIn;\n"
-" sdIn = sortDataInOut[gIdx];\n"
-" int2 sdOut;\n"
-" sdOut.x = contactsIn[sdIn.y].m_bodyBPtrAndSignBit;\n"
-" sdOut.y = sdIn.y;\n"
-" sortDataInOut[gIdx] = sdOut;\n"
-" }\n"
-"}\n"
-"typedef struct\n"
-"{\n"
-" int m_nContacts;\n"
-" int m_staticIdx;\n"
-" float m_scale;\n"
-" int m_nSplit;\n"
-"} ConstBufferSSD;\n"
-"static __constant const int gridTable4x4[] = \n"
-"{\n"
-" 0,1,17,16,\n"
-" 1,2,18,19,\n"
-" 17,18,32,3,\n"
-" 16,19,3,34\n"
-"};\n"
-"static __constant const int gridTable8x8[] = \n"
-"{\n"
-" 0, 2, 3, 16, 17, 18, 19, 1,\n"
-" 66, 64, 80, 67, 82, 81, 65, 83,\n"
-" 131,144,128,130,147,129,145,146,\n"
-" 208,195,194,192,193,211,210,209,\n"
-" 21, 22, 23, 5, 4, 6, 7, 20,\n"
-" 86, 85, 69, 87, 70, 68, 84, 71,\n"
-" 151,133,149,150,135,148,132,134,\n"
-" 197,27,214,213,212,199,198,196\n"
-" \n"
-"};\n"
-"#define USE_SPATIAL_BATCHING 1\n"
-"#define USE_4x4_GRID 1\n"
-"__kernel\n"
-"__attribute__((reqd_work_group_size(WG_SIZE,1,1)))\n"
-"void SetSortDataKernel(__global struct b3Contact4Data* gContact, __global Body* gBodies, __global int2* gSortDataOut, \n"
-"int nContacts,float scale,int4 nSplit,int staticIdx)\n"
-"{\n"
-" int gIdx = GET_GLOBAL_IDX;\n"
-" \n"
-" if( gIdx < nContacts )\n"
-" {\n"
-" int aPtrAndSignBit = gContact[gIdx].m_bodyAPtrAndSignBit;\n"
-" int bPtrAndSignBit = gContact[gIdx].m_bodyBPtrAndSignBit;\n"
-" int aIdx = abs(aPtrAndSignBit );\n"
-" int bIdx = abs(bPtrAndSignBit);\n"
-" bool aStatic = (aPtrAndSignBit<0) ||(aPtrAndSignBit==staticIdx);\n"
-" bool bStatic = (bPtrAndSignBit<0) ||(bPtrAndSignBit==staticIdx);\n"
-"#if USE_SPATIAL_BATCHING \n"
-" int idx = (aStatic)? bIdx: aIdx;\n"
-" float4 p = gBodies[idx].m_pos;\n"
-" int xIdx = (int)((p.x-((p.x<0.f)?1.f:0.f))*scale) & (nSplit.x-1);\n"
-" int yIdx = (int)((p.y-((p.y<0.f)?1.f:0.f))*scale) & (nSplit.y-1);\n"
-" int zIdx = (int)((p.z-((p.z<0.f)?1.f:0.f))*scale) & (nSplit.z-1);\n"
-" int newIndex = (xIdx+yIdx*nSplit.x+zIdx*nSplit.x*nSplit.y);\n"
-" \n"
-"#else//USE_SPATIAL_BATCHING\n"
-" #if USE_4x4_GRID\n"
-" int aa = aIdx&3;\n"
-" int bb = bIdx&3;\n"
-" if (aStatic)\n"
-" aa = bb;\n"
-" if (bStatic)\n"
-" bb = aa;\n"
-" int gridIndex = aa + bb*4;\n"
-" int newIndex = gridTable4x4[gridIndex];\n"
-" #else//USE_4x4_GRID\n"
-" int aa = aIdx&7;\n"
-" int bb = bIdx&7;\n"
-" if (aStatic)\n"
-" aa = bb;\n"
-" if (bStatic)\n"
-" bb = aa;\n"
-" int gridIndex = aa + bb*8;\n"
-" int newIndex = gridTable8x8[gridIndex];\n"
-" #endif//USE_4x4_GRID\n"
-"#endif//USE_SPATIAL_BATCHING\n"
-" gSortDataOut[gIdx].x = newIndex;\n"
-" gSortDataOut[gIdx].y = gIdx;\n"
-" }\n"
-" else\n"
-" {\n"
-" gSortDataOut[gIdx].x = 0xffffffff;\n"
-" }\n"
-"}\n"
-"__kernel\n"
-"__attribute__((reqd_work_group_size(WG_SIZE,1,1)))\n"
-"void CopyConstraintKernel(__global struct b3Contact4Data* gIn, __global struct b3Contact4Data* gOut, int4 cb )\n"
-"{\n"
-" int gIdx = GET_GLOBAL_IDX;\n"
-" if( gIdx < cb.x )\n"
-" {\n"
-" gOut[gIdx] = gIn[gIdx];\n"
-" }\n"
-"}\n"
-;
/include/Bullet/Bullet3OpenCL/RigidBody/kernels/solverSetup2.h
Property changes:
Deleted: svn:executable
## -1 +0,0 ##
-*
\ No newline at end of property
Index: include/Bullet/Bullet3OpenCL/RigidBody/kernels/jointSolver.h
===================================================================
--- include/Bullet/Bullet3OpenCL/RigidBody/kernels/jointSolver.h (revision 22)
+++ include/Bullet/Bullet3OpenCL/RigidBody/kernels/jointSolver.h (nonexistent)
@@ -1,721 +0,0 @@
-//this file is autogenerated using stringify.bat (premake --stringify) in the build folder of this project
-static const char* solveConstraintRowsCL= \
-"/*\n"
-"Copyright (c) 2013 Advanced Micro Devices, Inc. \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"
-"//Originally written by Erwin Coumans\n"
-"#define B3_CONSTRAINT_FLAG_ENABLED 1\n"
-"#define B3_GPU_POINT2POINT_CONSTRAINT_TYPE 3\n"
-"#define B3_GPU_FIXED_CONSTRAINT_TYPE 4\n"
-"#define MOTIONCLAMP 100000 //unused, for debugging/safety in case constraint solver fails\n"
-"#define B3_INFINITY 1e30f\n"
-"#define mymake_float4 (float4)\n"
-"__inline float dot3F4(float4 a, float4 b)\n"
-"{\n"
-" float4 a1 = mymake_float4(a.xyz,0.f);\n"
-" float4 b1 = mymake_float4(b.xyz,0.f);\n"
-" return dot(a1, b1);\n"
-"}\n"
-"typedef float4 Quaternion;\n"
-"typedef struct\n"
-"{\n"
-" float4 m_row[3];\n"
-"}Matrix3x3;\n"
-"__inline\n"
-"float4 mtMul1(Matrix3x3 a, float4 b);\n"
-"__inline\n"
-"float4 mtMul3(float4 a, Matrix3x3 b);\n"
-"__inline\n"
-"float4 mtMul1(Matrix3x3 a, float4 b)\n"
-"{\n"
-" float4 ans;\n"
-" ans.x = dot3F4( a.m_row[0], b );\n"
-" ans.y = dot3F4( a.m_row[1], b );\n"
-" ans.z = dot3F4( a.m_row[2], b );\n"
-" ans.w = 0.f;\n"
-" return ans;\n"
-"}\n"
-"__inline\n"
-"float4 mtMul3(float4 a, Matrix3x3 b)\n"
-"{\n"
-" float4 colx = mymake_float4(b.m_row[0].x, b.m_row[1].x, b.m_row[2].x, 0);\n"
-" float4 coly = mymake_float4(b.m_row[0].y, b.m_row[1].y, b.m_row[2].y, 0);\n"
-" float4 colz = mymake_float4(b.m_row[0].z, b.m_row[1].z, b.m_row[2].z, 0);\n"
-" float4 ans;\n"
-" ans.x = dot3F4( a, colx );\n"
-" ans.y = dot3F4( a, coly );\n"
-" ans.z = dot3F4( a, colz );\n"
-" return ans;\n"
-"}\n"
-"typedef struct\n"
-"{\n"
-" Matrix3x3 m_invInertiaWorld;\n"
-" Matrix3x3 m_initInvInertia;\n"
-"} BodyInertia;\n"
-"typedef struct\n"
-"{\n"
-" Matrix3x3 m_basis;//orientation\n"
-" float4 m_origin;//transform\n"
-"}b3Transform;\n"
-"typedef struct\n"
-"{\n"
-"// b3Transform m_worldTransformUnused;\n"
-" float4 m_deltaLinearVelocity;\n"
-" float4 m_deltaAngularVelocity;\n"
-" float4 m_angularFactor;\n"
-" float4 m_linearFactor;\n"
-" float4 m_invMass;\n"
-" float4 m_pushVelocity;\n"
-" float4 m_turnVelocity;\n"
-" float4 m_linearVelocity;\n"
-" float4 m_angularVelocity;\n"
-" union \n"
-" {\n"
-" void* m_originalBody;\n"
-" int m_originalBodyIndex;\n"
-" };\n"
-" int padding[3];\n"
-"} b3GpuSolverBody;\n"
-"typedef struct\n"
-"{\n"
-" float4 m_pos;\n"
-" Quaternion m_quat;\n"
-" float4 m_linVel;\n"
-" float4 m_angVel;\n"
-" unsigned int m_shapeIdx;\n"
-" float m_invMass;\n"
-" float m_restituitionCoeff;\n"
-" float m_frictionCoeff;\n"
-"} b3RigidBodyCL;\n"
-"typedef struct\n"
-"{\n"
-" float4 m_relpos1CrossNormal;\n"
-" float4 m_contactNormal;\n"
-" float4 m_relpos2CrossNormal;\n"
-" //float4 m_contactNormal2;//usually m_contactNormal2 == -m_contactNormal\n"
-" float4 m_angularComponentA;\n"
-" float4 m_angularComponentB;\n"
-" \n"
-" float m_appliedPushImpulse;\n"
-" float m_appliedImpulse;\n"
-" int m_padding1;\n"
-" int m_padding2;\n"
-" float m_friction;\n"
-" float m_jacDiagABInv;\n"
-" float m_rhs;\n"
-" float m_cfm;\n"
-" \n"
-" float m_lowerLimit;\n"
-" float m_upperLimit;\n"
-" float m_rhsPenetration;\n"
-" int m_originalConstraint;\n"
-" int m_overrideNumSolverIterations;\n"
-" int m_frictionIndex;\n"
-" int m_solverBodyIdA;\n"
-" int m_solverBodyIdB;\n"
-"} b3SolverConstraint;\n"
-"typedef struct \n"
-"{\n"
-" int m_bodyAPtrAndSignBit;\n"
-" int m_bodyBPtrAndSignBit;\n"
-" int m_originalConstraintIndex;\n"
-" int m_batchId;\n"
-"} b3BatchConstraint;\n"
-"typedef struct \n"
-"{\n"
-" int m_constraintType;\n"
-" int m_rbA;\n"
-" int m_rbB;\n"
-" float m_breakingImpulseThreshold;\n"
-" float4 m_pivotInA;\n"
-" float4 m_pivotInB;\n"
-" Quaternion m_relTargetAB;\n"
-" int m_flags;\n"
-" int m_padding[3];\n"
-"} b3GpuGenericConstraint;\n"
-"/*b3Transform getWorldTransform(b3RigidBodyCL* rb)\n"
-"{\n"
-" b3Transform newTrans;\n"
-" newTrans.setOrigin(rb->m_pos);\n"
-" newTrans.setRotation(rb->m_quat);\n"
-" return newTrans;\n"
-"}*/\n"
-"__inline\n"
-"float4 cross3(float4 a, float4 b)\n"
-"{\n"
-" return cross(a,b);\n"
-"}\n"
-"__inline\n"
-"float4 fastNormalize4(float4 v)\n"
-"{\n"
-" v = mymake_float4(v.xyz,0.f);\n"
-" return fast_normalize(v);\n"
-"}\n"
-"__inline\n"
-"Quaternion qtMul(Quaternion a, Quaternion b);\n"
-"__inline\n"
-"Quaternion qtNormalize(Quaternion in);\n"
-"__inline\n"
-"float4 qtRotate(Quaternion q, float4 vec);\n"
-"__inline\n"
-"Quaternion qtInvert(Quaternion q);\n"
-"__inline\n"
-"Quaternion qtMul(Quaternion a, Quaternion b)\n"
-"{\n"
-" Quaternion ans;\n"
-" ans = cross3( 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 - dot3F4(a, b);\n"
-" return ans;\n"
-"}\n"
-"__inline\n"
-"Quaternion qtNormalize(Quaternion in)\n"
-"{\n"
-" return fastNormalize4(in);\n"
-"// in /= length( in );\n"
-"// return in;\n"
-"}\n"
-"__inline\n"
-"float4 qtRotate(Quaternion q, float4 vec)\n"
-"{\n"
-" Quaternion qInv = qtInvert( q );\n"
-" float4 vcpy = vec;\n"
-" vcpy.w = 0.f;\n"
-" float4 out = qtMul(qtMul(q,vcpy),qInv);\n"
-" return out;\n"
-"}\n"
-"__inline\n"
-"Quaternion qtInvert(Quaternion q)\n"
-"{\n"
-" return (Quaternion)(-q.xyz, q.w);\n"
-"}\n"
-"__inline void internalApplyImpulse(__global b3GpuSolverBody* body, float4 linearComponent, float4 angularComponent,float impulseMagnitude)\n"
-"{\n"
-" body->m_deltaLinearVelocity += linearComponent*impulseMagnitude*body->m_linearFactor;\n"
-" body->m_deltaAngularVelocity += angularComponent*(impulseMagnitude*body->m_angularFactor);\n"
-"}\n"
-"void resolveSingleConstraintRowGeneric(__global b3GpuSolverBody* body1, __global b3GpuSolverBody* body2, __global b3SolverConstraint* c)\n"
-"{\n"
-" float deltaImpulse = c->m_rhs-c->m_appliedImpulse*c->m_cfm;\n"
-" float deltaVel1Dotn = dot3F4(c->m_contactNormal,body1->m_deltaLinearVelocity) + dot3F4(c->m_relpos1CrossNormal,body1->m_deltaAngularVelocity);\n"
-" float deltaVel2Dotn = -dot3F4(c->m_contactNormal,body2->m_deltaLinearVelocity) + dot3F4(c->m_relpos2CrossNormal,body2->m_deltaAngularVelocity);\n"
-" deltaImpulse -= deltaVel1Dotn*c->m_jacDiagABInv;\n"
-" deltaImpulse -= deltaVel2Dotn*c->m_jacDiagABInv;\n"
-" float sum = c->m_appliedImpulse + deltaImpulse;\n"
-" if (sum < c->m_lowerLimit)\n"
-" {\n"
-" deltaImpulse = c->m_lowerLimit-c->m_appliedImpulse;\n"
-" c->m_appliedImpulse = c->m_lowerLimit;\n"
-" }\n"
-" else if (sum > c->m_upperLimit) \n"
-" {\n"
-" deltaImpulse = c->m_upperLimit-c->m_appliedImpulse;\n"
-" c->m_appliedImpulse = c->m_upperLimit;\n"
-" }\n"
-" else\n"
-" {\n"
-" c->m_appliedImpulse = sum;\n"
-" }\n"
-" internalApplyImpulse(body1,c->m_contactNormal*body1->m_invMass,c->m_angularComponentA,deltaImpulse);\n"
-" internalApplyImpulse(body2,-c->m_contactNormal*body2->m_invMass,c->m_angularComponentB,deltaImpulse);\n"
-"}\n"
-"__kernel void solveJointConstraintRows(__global b3GpuSolverBody* solverBodies,\n"
-" __global b3BatchConstraint* batchConstraints,\n"
-" __global b3SolverConstraint* rows,\n"
-" __global unsigned int* numConstraintRowsInfo1, \n"
-" __global unsigned int* rowOffsets,\n"
-" __global b3GpuGenericConstraint* constraints,\n"
-" int batchOffset,\n"
-" int numConstraintsInBatch\n"
-" )\n"
-"{\n"
-" int b = get_global_id(0);\n"
-" if (b>=numConstraintsInBatch)\n"
-" return;\n"
-" __global b3BatchConstraint* c = &batchConstraints[b+batchOffset];\n"
-" int originalConstraintIndex = c->m_originalConstraintIndex;\n"
-" if (constraints[originalConstraintIndex].m_flags&B3_CONSTRAINT_FLAG_ENABLED)\n"
-" {\n"
-" int numConstraintRows = numConstraintRowsInfo1[originalConstraintIndex];\n"
-" int rowOffset = rowOffsets[originalConstraintIndex];\n"
-" for (int jj=0;jj<numConstraintRows;jj++)\n"
-" {\n"
-" __global b3SolverConstraint* constraint = &rows[rowOffset+jj];\n"
-" resolveSingleConstraintRowGeneric(&solverBodies[constraint->m_solverBodyIdA],&solverBodies[constraint->m_solverBodyIdB],constraint);\n"
-" }\n"
-" }\n"
-"};\n"
-"__kernel void initSolverBodies(__global b3GpuSolverBody* solverBodies,__global b3RigidBodyCL* bodiesCL, int numBodies)\n"
-"{\n"
-" int i = get_global_id(0);\n"
-" if (i>=numBodies)\n"
-" return;\n"
-" __global b3GpuSolverBody* solverBody = &solverBodies[i];\n"
-" __global b3RigidBodyCL* bodyCL = &bodiesCL[i];\n"
-" solverBody->m_deltaLinearVelocity = (float4)(0.f,0.f,0.f,0.f);\n"
-" solverBody->m_deltaAngularVelocity = (float4)(0.f,0.f,0.f,0.f);\n"
-" solverBody->m_pushVelocity = (float4)(0.f,0.f,0.f,0.f);\n"
-" solverBody->m_pushVelocity = (float4)(0.f,0.f,0.f,0.f);\n"
-" solverBody->m_invMass = (float4)(bodyCL->m_invMass,bodyCL->m_invMass,bodyCL->m_invMass,0.f);\n"
-" solverBody->m_originalBodyIndex = i;\n"
-" solverBody->m_angularFactor = (float4)(1,1,1,0);\n"
-" solverBody->m_linearFactor = (float4) (1,1,1,0);\n"
-" solverBody->m_linearVelocity = bodyCL->m_linVel;\n"
-" solverBody->m_angularVelocity = bodyCL->m_angVel;\n"
-"}\n"
-"__kernel void breakViolatedConstraintsKernel(__global b3GpuGenericConstraint* constraints, __global unsigned int* numConstraintRows, __global unsigned int* rowOffsets, __global b3SolverConstraint* rows, int numConstraints)\n"
-"{\n"
-" int cid = get_global_id(0);\n"
-" if (cid>=numConstraints)\n"
-" return;\n"
-" int numRows = numConstraintRows[cid];\n"
-" if (numRows)\n"
-" {\n"
-" for (int i=0;i<numRows;i++)\n"
-" {\n"
-" int rowIndex = rowOffsets[cid]+i;\n"
-" float breakingThreshold = constraints[cid].m_breakingImpulseThreshold;\n"
-" if (fabs(rows[rowIndex].m_appliedImpulse) >= breakingThreshold)\n"
-" {\n"
-" constraints[cid].m_flags =0;//&= ~B3_CONSTRAINT_FLAG_ENABLED;\n"
-" }\n"
-" }\n"
-" }\n"
-"}\n"
-"__kernel void getInfo1Kernel(__global unsigned int* infos, __global b3GpuGenericConstraint* constraints, int numConstraints)\n"
-"{\n"
-" int i = get_global_id(0);\n"
-" if (i>=numConstraints)\n"
-" return;\n"
-" __global b3GpuGenericConstraint* constraint = &constraints[i];\n"
-" switch (constraint->m_constraintType)\n"
-" {\n"
-" case B3_GPU_POINT2POINT_CONSTRAINT_TYPE:\n"
-" {\n"
-" infos[i] = 3;\n"
-" break;\n"
-" }\n"
-" case B3_GPU_FIXED_CONSTRAINT_TYPE:\n"
-" {\n"
-" infos[i] = 6;\n"
-" break;\n"
-" }\n"
-" default:\n"
-" {\n"
-" }\n"
-" }\n"
-"}\n"
-"__kernel void initBatchConstraintsKernel(__global unsigned int* numConstraintRows, __global unsigned int* rowOffsets, \n"
-" __global b3BatchConstraint* batchConstraints, \n"
-" __global b3GpuGenericConstraint* constraints,\n"
-" __global b3RigidBodyCL* bodies,\n"
-" int numConstraints)\n"
-"{\n"
-" int i = get_global_id(0);\n"
-" if (i>=numConstraints)\n"
-" return;\n"
-" int rbA = constraints[i].m_rbA;\n"
-" int rbB = constraints[i].m_rbB;\n"
-" batchConstraints[i].m_bodyAPtrAndSignBit = bodies[rbA].m_invMass? rbA : -rbA;\n"
-" batchConstraints[i].m_bodyBPtrAndSignBit = bodies[rbB].m_invMass? rbB : -rbB;\n"
-" batchConstraints[i].m_batchId = -1;\n"
-" batchConstraints[i].m_originalConstraintIndex = i;\n"
-"}\n"
-"typedef struct\n"
-"{\n"
-" // integrator parameters: frames per second (1/stepsize), default error\n"
-" // reduction parameter (0..1).\n"
-" float fps,erp;\n"
-" // for the first and second body, pointers to two (linear and angular)\n"
-" // n*3 jacobian sub matrices, stored by rows. these matrices will have\n"
-" // been initialized to 0 on entry. if the second body is zero then the\n"
-" // J2xx pointers may be 0.\n"
-" union \n"
-" {\n"
-" __global float4* m_J1linearAxisFloat4;\n"
-" __global float* m_J1linearAxis;\n"
-" };\n"
-" union\n"
-" {\n"
-" __global float4* m_J1angularAxisFloat4;\n"
-" __global float* m_J1angularAxis;\n"
-" };\n"
-" union\n"
-" {\n"
-" __global float4* m_J2linearAxisFloat4;\n"
-" __global float* m_J2linearAxis;\n"
-" };\n"
-" union\n"
-" {\n"
-" __global float4* m_J2angularAxisFloat4;\n"
-" __global float* m_J2angularAxis;\n"
-" };\n"
-" // elements to jump from one row to the next in J's\n"
-" int rowskip;\n"
-" // right hand sides of the equation J*v = c + cfm * lambda. cfm is the\n"
-" // \"constraint force mixing\" vector. c is set to zero on entry, cfm is\n"
-" // set to a constant value (typically very small or zero) value on entry.\n"
-" __global float* m_constraintError;\n"
-" __global float* cfm;\n"
-" // lo and hi limits for variables (set to -/+ infinity on entry).\n"
-" __global float* m_lowerLimit;\n"
-" __global float* m_upperLimit;\n"
-" // findex vector for variables. see the LCP solver interface for a\n"
-" // description of what this does. this is set to -1 on entry.\n"
-" // note that the returned indexes are relative to the first index of\n"
-" // the constraint.\n"
-" __global int *findex;\n"
-" // number of solver iterations\n"
-" int m_numIterations;\n"
-" //damping of the velocity\n"
-" float m_damping;\n"
-"} b3GpuConstraintInfo2;\n"
-"void getSkewSymmetricMatrix(float4 vecIn, __global float4* v0,__global float4* v1,__global float4* v2)\n"
-"{\n"
-" *v0 = (float4)(0. ,-vecIn.z ,vecIn.y,0.f);\n"
-" *v1 = (float4)(vecIn.z ,0. ,-vecIn.x,0.f);\n"
-" *v2 = (float4)(-vecIn.y ,vecIn.x ,0.f,0.f);\n"
-"}\n"
-"void getInfo2Point2Point(__global b3GpuGenericConstraint* constraint,b3GpuConstraintInfo2* info,__global b3RigidBodyCL* bodies)\n"
-"{\n"
-" float4 posA = bodies[constraint->m_rbA].m_pos;\n"
-" Quaternion rotA = bodies[constraint->m_rbA].m_quat;\n"
-" float4 posB = bodies[constraint->m_rbB].m_pos;\n"
-" Quaternion rotB = bodies[constraint->m_rbB].m_quat;\n"
-" // anchor points in global coordinates with respect to body PORs.\n"
-" \n"
-" // set jacobian\n"
-" info->m_J1linearAxis[0] = 1;\n"
-" info->m_J1linearAxis[info->rowskip+1] = 1;\n"
-" info->m_J1linearAxis[2*info->rowskip+2] = 1;\n"
-" float4 a1 = qtRotate(rotA,constraint->m_pivotInA);\n"
-" {\n"
-" __global float4* angular0 = (__global float4*)(info->m_J1angularAxis);\n"
-" __global float4* angular1 = (__global float4*)(info->m_J1angularAxis+info->rowskip);\n"
-" __global float4* angular2 = (__global float4*)(info->m_J1angularAxis+2*info->rowskip);\n"
-" float4 a1neg = -a1;\n"
-" getSkewSymmetricMatrix(a1neg,angular0,angular1,angular2);\n"
-" }\n"
-" if (info->m_J2linearAxis)\n"
-" {\n"
-" info->m_J2linearAxis[0] = -1;\n"
-" info->m_J2linearAxis[info->rowskip+1] = -1;\n"
-" info->m_J2linearAxis[2*info->rowskip+2] = -1;\n"
-" }\n"
-" \n"
-" float4 a2 = qtRotate(rotB,constraint->m_pivotInB);\n"
-" \n"
-" {\n"
-" // float4 a2n = -a2;\n"
-" __global float4* angular0 = (__global float4*)(info->m_J2angularAxis);\n"
-" __global float4* angular1 = (__global float4*)(info->m_J2angularAxis+info->rowskip);\n"
-" __global float4* angular2 = (__global float4*)(info->m_J2angularAxis+2*info->rowskip);\n"
-" getSkewSymmetricMatrix(a2,angular0,angular1,angular2);\n"
-" }\n"
-" \n"
-" // set right hand side\n"
-"// float currERP = (m_flags & B3_P2P_FLAGS_ERP) ? m_erp : info->erp;\n"
-" float currERP = info->erp;\n"
-" float k = info->fps * currERP;\n"
-" int j;\n"
-" float4 result = a2 + posB - a1 - posA;\n"
-" float* resultPtr = &result;\n"
-" for (j=0; j<3; j++)\n"
-" {\n"
-" info->m_constraintError[j*info->rowskip] = k * (resultPtr[j]);\n"
-" }\n"
-"}\n"
-"Quaternion nearest( Quaternion first, Quaternion qd)\n"
-"{\n"
-" Quaternion diff,sum;\n"
-" diff = first- qd;\n"
-" sum = first + qd;\n"
-" \n"
-" if( dot(diff,diff) < dot(sum,sum) )\n"
-" return qd;\n"
-" return (-qd);\n"
-"}\n"
-"float b3Acos(float x) \n"
-"{ \n"
-" if (x<-1) \n"
-" x=-1; \n"
-" if (x>1) \n"
-" x=1;\n"
-" return acos(x); \n"
-"}\n"
-"float getAngle(Quaternion orn)\n"
-"{\n"
-" if (orn.w>=1.f)\n"
-" orn.w=1.f;\n"
-" float s = 2.f * b3Acos(orn.w);\n"
-" return s;\n"
-"}\n"
-"void calculateDiffAxisAngleQuaternion( Quaternion orn0,Quaternion orn1a,float4* axis,float* angle)\n"
-"{\n"
-" Quaternion orn1 = nearest(orn0,orn1a);\n"
-" \n"
-" Quaternion dorn = qtMul(orn1,qtInvert(orn0));\n"
-" *angle = getAngle(dorn);\n"
-" *axis = (float4)(dorn.x,dorn.y,dorn.z,0.f);\n"
-" \n"
-" //check for axis length\n"
-" float len = dot3F4(*axis,*axis);\n"
-" if (len < FLT_EPSILON*FLT_EPSILON)\n"
-" *axis = (float4)(1,0,0,0);\n"
-" else\n"
-" *axis /= sqrt(len);\n"
-"}\n"
-"void getInfo2FixedOrientation(__global b3GpuGenericConstraint* constraint,b3GpuConstraintInfo2* info,__global b3RigidBodyCL* bodies, int start_row)\n"
-"{\n"
-" Quaternion worldOrnA = bodies[constraint->m_rbA].m_quat;\n"
-" Quaternion worldOrnB = bodies[constraint->m_rbB].m_quat;\n"
-" int s = info->rowskip;\n"
-" int start_index = start_row * s;\n"
-" // 3 rows to make body rotations equal\n"
-" info->m_J1angularAxis[start_index] = 1;\n"
-" info->m_J1angularAxis[start_index + s + 1] = 1;\n"
-" info->m_J1angularAxis[start_index + s*2+2] = 1;\n"
-" if ( info->m_J2angularAxis)\n"
-" {\n"
-" info->m_J2angularAxis[start_index] = -1;\n"
-" info->m_J2angularAxis[start_index + s+1] = -1;\n"
-" info->m_J2angularAxis[start_index + s*2+2] = -1;\n"
-" }\n"
-" \n"
-" float currERP = info->erp;\n"
-" float k = info->fps * currERP;\n"
-" float4 diff;\n"
-" float angle;\n"
-" float4 qrelCur = qtMul(worldOrnA,qtInvert(worldOrnB));\n"
-" \n"
-" calculateDiffAxisAngleQuaternion(constraint->m_relTargetAB,qrelCur,&diff,&angle);\n"
-" diff*=-angle;\n"
-" \n"
-" float* resultPtr = &diff;\n"
-" \n"
-" for (int j=0; j<3; j++)\n"
-" {\n"
-" info->m_constraintError[(3+j)*info->rowskip] = k * resultPtr[j];\n"
-" }\n"
-" \n"
-"}\n"
-"__kernel void writeBackVelocitiesKernel(__global b3RigidBodyCL* bodies,__global b3GpuSolverBody* solverBodies,int numBodies)\n"
-"{\n"
-" int i = get_global_id(0);\n"
-" if (i>=numBodies)\n"
-" return;\n"
-" if (bodies[i].m_invMass)\n"
-" {\n"
-"// if (length(solverBodies[i].m_deltaLinearVelocity)<MOTIONCLAMP)\n"
-" {\n"
-" bodies[i].m_linVel += solverBodies[i].m_deltaLinearVelocity;\n"
-" }\n"
-"// if (length(solverBodies[i].m_deltaAngularVelocity)<MOTIONCLAMP)\n"
-" {\n"
-" bodies[i].m_angVel += solverBodies[i].m_deltaAngularVelocity;\n"
-" } \n"
-" }\n"
-"}\n"
-"__kernel void getInfo2Kernel(__global b3SolverConstraint* solverConstraintRows, \n"
-" __global unsigned int* infos, \n"
-" __global unsigned int* constraintRowOffsets, \n"
-" __global b3GpuGenericConstraint* constraints, \n"
-" __global b3BatchConstraint* batchConstraints, \n"
-" __global b3RigidBodyCL* bodies,\n"
-" __global BodyInertia* inertias,\n"
-" __global b3GpuSolverBody* solverBodies,\n"
-" float timeStep,\n"
-" float globalErp,\n"
-" float globalCfm,\n"
-" float globalDamping,\n"
-" int globalNumIterations,\n"
-" int numConstraints)\n"
-"{\n"
-" int i = get_global_id(0);\n"
-" if (i>=numConstraints)\n"
-" return;\n"
-" \n"
-" //for now, always initialize the batch info\n"
-" int info1 = infos[i];\n"
-" \n"
-" __global b3SolverConstraint* currentConstraintRow = &solverConstraintRows[constraintRowOffsets[i]];\n"
-" __global b3GpuGenericConstraint* constraint = &constraints[i];\n"
-" __global b3RigidBodyCL* rbA = &bodies[ constraint->m_rbA];\n"
-" __global b3RigidBodyCL* rbB = &bodies[ constraint->m_rbB];\n"
-" int solverBodyIdA = constraint->m_rbA;\n"
-" int solverBodyIdB = constraint->m_rbB;\n"
-" __global b3GpuSolverBody* bodyAPtr = &solverBodies[solverBodyIdA];\n"
-" __global b3GpuSolverBody* bodyBPtr = &solverBodies[solverBodyIdB];\n"
-" if (rbA->m_invMass)\n"
-" {\n"
-" batchConstraints[i].m_bodyAPtrAndSignBit = solverBodyIdA;\n"
-" } else\n"
-" {\n"
-"// if (!solverBodyIdA)\n"
-"// m_staticIdx = 0;\n"
-" batchConstraints[i].m_bodyAPtrAndSignBit = -solverBodyIdA;\n"
-" }\n"
-" if (rbB->m_invMass)\n"
-" {\n"
-" batchConstraints[i].m_bodyBPtrAndSignBit = solverBodyIdB;\n"
-" } else\n"
-" {\n"
-"// if (!solverBodyIdB)\n"
-"// m_staticIdx = 0;\n"
-" batchConstraints[i].m_bodyBPtrAndSignBit = -solverBodyIdB;\n"
-" }\n"
-" if (info1)\n"
-" {\n"
-" int overrideNumSolverIterations = 0;//constraint->getOverrideNumSolverIterations() > 0 ? constraint->getOverrideNumSolverIterations() : infoGlobal.m_numIterations;\n"
-"// if (overrideNumSolverIterations>m_maxOverrideNumSolverIterations)\n"
-" // m_maxOverrideNumSolverIterations = overrideNumSolverIterations;\n"
-" int j;\n"
-" for ( j=0;j<info1;j++)\n"
-" {\n"
-"// memset(&currentConstraintRow[j],0,sizeof(b3SolverConstraint));\n"
-" currentConstraintRow[j].m_angularComponentA = (float4)(0,0,0,0);\n"
-" currentConstraintRow[j].m_angularComponentB = (float4)(0,0,0,0);\n"
-" currentConstraintRow[j].m_appliedImpulse = 0.f;\n"
-" currentConstraintRow[j].m_appliedPushImpulse = 0.f;\n"
-" currentConstraintRow[j].m_cfm = 0.f;\n"
-" currentConstraintRow[j].m_contactNormal = (float4)(0,0,0,0);\n"
-" currentConstraintRow[j].m_friction = 0.f;\n"
-" currentConstraintRow[j].m_frictionIndex = 0;\n"
-" currentConstraintRow[j].m_jacDiagABInv = 0.f;\n"
-" currentConstraintRow[j].m_lowerLimit = 0.f;\n"
-" currentConstraintRow[j].m_upperLimit = 0.f;\n"
-" currentConstraintRow[j].m_originalConstraint = i;\n"
-" currentConstraintRow[j].m_overrideNumSolverIterations = 0;\n"
-" currentConstraintRow[j].m_relpos1CrossNormal = (float4)(0,0,0,0);\n"
-" currentConstraintRow[j].m_relpos2CrossNormal = (float4)(0,0,0,0);\n"
-" currentConstraintRow[j].m_rhs = 0.f;\n"
-" currentConstraintRow[j].m_rhsPenetration = 0.f;\n"
-" currentConstraintRow[j].m_solverBodyIdA = 0;\n"
-" currentConstraintRow[j].m_solverBodyIdB = 0;\n"
-" \n"
-" currentConstraintRow[j].m_lowerLimit = -B3_INFINITY;\n"
-" currentConstraintRow[j].m_upperLimit = B3_INFINITY;\n"
-" currentConstraintRow[j].m_appliedImpulse = 0.f;\n"
-" currentConstraintRow[j].m_appliedPushImpulse = 0.f;\n"
-" currentConstraintRow[j].m_solverBodyIdA = solverBodyIdA;\n"
-" currentConstraintRow[j].m_solverBodyIdB = solverBodyIdB;\n"
-" currentConstraintRow[j].m_overrideNumSolverIterations = overrideNumSolverIterations; \n"
-" }\n"
-" bodyAPtr->m_deltaLinearVelocity = (float4)(0,0,0,0);\n"
-" bodyAPtr->m_deltaAngularVelocity = (float4)(0,0,0,0);\n"
-" bodyAPtr->m_pushVelocity = (float4)(0,0,0,0);\n"
-" bodyAPtr->m_turnVelocity = (float4)(0,0,0,0);\n"
-" bodyBPtr->m_deltaLinearVelocity = (float4)(0,0,0,0);\n"
-" bodyBPtr->m_deltaAngularVelocity = (float4)(0,0,0,0);\n"
-" bodyBPtr->m_pushVelocity = (float4)(0,0,0,0);\n"
-" bodyBPtr->m_turnVelocity = (float4)(0,0,0,0);\n"
-" int rowskip = sizeof(b3SolverConstraint)/sizeof(float);//check this\n"
-" \n"
-" b3GpuConstraintInfo2 info2;\n"
-" info2.fps = 1.f/timeStep;\n"
-" info2.erp = globalErp;\n"
-" info2.m_J1linearAxisFloat4 = &currentConstraintRow->m_contactNormal;\n"
-" info2.m_J1angularAxisFloat4 = &currentConstraintRow->m_relpos1CrossNormal;\n"
-" info2.m_J2linearAxisFloat4 = 0;\n"
-" info2.m_J2angularAxisFloat4 = &currentConstraintRow->m_relpos2CrossNormal;\n"
-" info2.rowskip = sizeof(b3SolverConstraint)/sizeof(float);//check this\n"
-" ///the size of b3SolverConstraint needs be a multiple of float\n"
-"// b3Assert(info2.rowskip*sizeof(float)== sizeof(b3SolverConstraint));\n"
-" info2.m_constraintError = &currentConstraintRow->m_rhs;\n"
-" currentConstraintRow->m_cfm = globalCfm;\n"
-" info2.m_damping = globalDamping;\n"
-" info2.cfm = &currentConstraintRow->m_cfm;\n"
-" info2.m_lowerLimit = &currentConstraintRow->m_lowerLimit;\n"
-" info2.m_upperLimit = &currentConstraintRow->m_upperLimit;\n"
-" info2.m_numIterations = globalNumIterations;\n"
-" switch (constraint->m_constraintType)\n"
-" {\n"
-" case B3_GPU_POINT2POINT_CONSTRAINT_TYPE:\n"
-" {\n"
-" getInfo2Point2Point(constraint,&info2,bodies);\n"
-" break;\n"
-" }\n"
-" case B3_GPU_FIXED_CONSTRAINT_TYPE:\n"
-" {\n"
-" getInfo2Point2Point(constraint,&info2,bodies);\n"
-" getInfo2FixedOrientation(constraint,&info2,bodies,3);\n"
-" break;\n"
-" }\n"
-" default:\n"
-" {\n"
-" }\n"
-" }\n"
-" ///finalize the constraint setup\n"
-" for ( j=0;j<info1;j++)\n"
-" {\n"
-" __global b3SolverConstraint* solverConstraint = &currentConstraintRow[j];\n"
-" if (solverConstraint->m_upperLimit>=constraint->m_breakingImpulseThreshold)\n"
-" {\n"
-" solverConstraint->m_upperLimit = constraint->m_breakingImpulseThreshold;\n"
-" }\n"
-" if (solverConstraint->m_lowerLimit<=-constraint->m_breakingImpulseThreshold)\n"
-" {\n"
-" solverConstraint->m_lowerLimit = -constraint->m_breakingImpulseThreshold;\n"
-" }\n"
-"// solverConstraint->m_originalContactPoint = constraint;\n"
-" \n"
-" Matrix3x3 invInertiaWorldA= inertias[constraint->m_rbA].m_invInertiaWorld;\n"
-" {\n"
-" //float4 angularFactorA(1,1,1);\n"
-" float4 ftorqueAxis1 = solverConstraint->m_relpos1CrossNormal;\n"
-" solverConstraint->m_angularComponentA = mtMul1(invInertiaWorldA,ftorqueAxis1);//*angularFactorA;\n"
-" }\n"
-" \n"
-" Matrix3x3 invInertiaWorldB= inertias[constraint->m_rbB].m_invInertiaWorld;\n"
-" {\n"
-" float4 ftorqueAxis2 = solverConstraint->m_relpos2CrossNormal;\n"
-" solverConstraint->m_angularComponentB = mtMul1(invInertiaWorldB,ftorqueAxis2);//*constraint->m_rbB.getAngularFactor();\n"
-" }\n"
-" {\n"
-" //it is ok to use solverConstraint->m_contactNormal instead of -solverConstraint->m_contactNormal\n"
-" //because it gets multiplied iMJlB\n"
-" float4 iMJlA = solverConstraint->m_contactNormal*rbA->m_invMass;\n"
-" float4 iMJaA = mtMul3(solverConstraint->m_relpos1CrossNormal,invInertiaWorldA);\n"
-" float4 iMJlB = solverConstraint->m_contactNormal*rbB->m_invMass;//sign of normal?\n"
-" float4 iMJaB = mtMul3(solverConstraint->m_relpos2CrossNormal,invInertiaWorldB);\n"
-" float sum = dot3F4(iMJlA,solverConstraint->m_contactNormal);\n"
-" sum += dot3F4(iMJaA,solverConstraint->m_relpos1CrossNormal);\n"
-" sum += dot3F4(iMJlB,solverConstraint->m_contactNormal);\n"
-" sum += dot3F4(iMJaB,solverConstraint->m_relpos2CrossNormal);\n"
-" float fsum = fabs(sum);\n"
-" if (fsum>FLT_EPSILON)\n"
-" {\n"
-" solverConstraint->m_jacDiagABInv = 1.f/sum;\n"
-" } else\n"
-" {\n"
-" solverConstraint->m_jacDiagABInv = 0.f;\n"
-" }\n"
-" }\n"
-" ///fix rhs\n"
-" ///todo: add force/torque accelerators\n"
-" {\n"
-" float rel_vel;\n"
-" float vel1Dotn = dot3F4(solverConstraint->m_contactNormal,rbA->m_linVel) + dot3F4(solverConstraint->m_relpos1CrossNormal,rbA->m_angVel);\n"
-" float vel2Dotn = -dot3F4(solverConstraint->m_contactNormal,rbB->m_linVel) + dot3F4(solverConstraint->m_relpos2CrossNormal,rbB->m_angVel);\n"
-" rel_vel = vel1Dotn+vel2Dotn;\n"
-" float restitution = 0.f;\n"
-" float positionalError = solverConstraint->m_rhs;//already filled in by getConstraintInfo2\n"
-" float velocityError = restitution - rel_vel * info2.m_damping;\n"
-" float penetrationImpulse = positionalError*solverConstraint->m_jacDiagABInv;\n"
-" float velocityImpulse = velocityError *solverConstraint->m_jacDiagABInv;\n"
-" solverConstraint->m_rhs = penetrationImpulse+velocityImpulse;\n"
-" solverConstraint->m_appliedImpulse = 0.f;\n"
-" }\n"
-" }\n"
-" }\n"
-"}\n"
-;
/include/Bullet/Bullet3OpenCL/RigidBody/kernels/jointSolver.h
Property changes:
Deleted: svn:executable
## -1 +0,0 ##
-*
\ No newline at end of property
Index: include/Bullet/Bullet3OpenCL/RigidBody/kernels/batchingKernels.h
===================================================================
--- include/Bullet/Bullet3OpenCL/RigidBody/kernels/batchingKernels.h (revision 22)
+++ include/Bullet/Bullet3OpenCL/RigidBody/kernels/batchingKernels.h (nonexistent)
@@ -1,387 +0,0 @@
-//this file is autogenerated using stringify.bat (premake --stringify) in the build folder of this project
-static const char* batchingKernelsCL= \
-"/*\n"
-"Copyright (c) 2012 Advanced Micro Devices, Inc. \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"
-"//Originally written by Takahiro Harada\n"
-"#ifndef B3_CONTACT4DATA_H\n"
-"#define B3_CONTACT4DATA_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"
-"//keep B3_LARGE_FLOAT*B3_LARGE_FLOAT < FLT_MAX\n"
-"#define B3_LARGE_FLOAT 1e18f\n"
-"#define B3_INFINITY 1e18f\n"
-"#define b3Assert(a)\n"
-"#define b3ConstArray(a) __global const a*\n"
-"#define b3AtomicInc atomic_inc\n"
-"#define b3AtomicAdd atomic_add\n"
-"#define b3Fabs fabs\n"
-"#define b3Sqrt native_sqrt\n"
-"#define b3Sin native_sin\n"
-"#define b3Cos native_cos\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"
-" #define b3MinFloat4 min\n"
-" #define b3MaxFloat4 max\n"
-" #define b3Normalized(a) normalize(a)\n"
-"#endif \n"
-" \n"
-"inline bool b3IsAlmostZero(b3Float4ConstArg v)\n"
-"{\n"
-" if(b3Fabs(v.x)>1e-6 || b3Fabs(v.y)>1e-6 || b3Fabs(v.z)>1e-6) \n"
-" return false;\n"
-" return true;\n"
-"}\n"
-"inline int b3MaxDot( b3Float4ConstArg vec, __global const b3Float4* vecArray, int vecLen, float* dotOut )\n"
-"{\n"
-" float maxDot = -B3_INFINITY;\n"
-" int i = 0;\n"
-" int ptIndex = -1;\n"
-" for( i = 0; i < vecLen; i++ )\n"
-" {\n"
-" float dot = b3Dot3F4(vecArray[i],vec);\n"
-" \n"
-" if( dot > maxDot )\n"
-" {\n"
-" maxDot = dot;\n"
-" ptIndex = i;\n"
-" }\n"
-" }\n"
-" b3Assert(ptIndex>=0);\n"
-" if (ptIndex<0)\n"
-" {\n"
-" ptIndex = 0;\n"
-" }\n"
-" *dotOut = maxDot;\n"
-" return ptIndex;\n"
-"}\n"
-"#endif //B3_FLOAT4_H\n"
-"typedef struct b3Contact4Data b3Contact4Data_t;\n"
-"struct b3Contact4Data\n"
-"{\n"
-" b3Float4 m_worldPosB[4];\n"
-"// b3Float4 m_localPosA[4];\n"
-"// b3Float4 m_localPosB[4];\n"
-" b3Float4 m_worldNormalOnB; // w: m_nPoints\n"
-" unsigned short m_restituitionCoeffCmp;\n"
-" unsigned short m_frictionCoeffCmp;\n"
-" int m_batchIdx;\n"
-" int m_bodyAPtrAndSignBit;//x:m_bodyAPtr, y:m_bodyBPtr\n"
-" int m_bodyBPtrAndSignBit;\n"
-" int m_childIndexA;\n"
-" int m_childIndexB;\n"
-" int m_unused1;\n"
-" int m_unused2;\n"
-"};\n"
-"inline int b3Contact4Data_getNumPoints(const struct b3Contact4Data* contact)\n"
-"{\n"
-" return (int)contact->m_worldNormalOnB.w;\n"
-"};\n"
-"inline void b3Contact4Data_setNumPoints(struct b3Contact4Data* contact, int numPoints)\n"
-"{\n"
-" contact->m_worldNormalOnB.w = (float)numPoints;\n"
-"};\n"
-"#endif //B3_CONTACT4DATA_H\n"
-"#pragma OPENCL EXTENSION cl_amd_printf : enable\n"
-"#pragma OPENCL EXTENSION cl_khr_local_int32_base_atomics : enable\n"
-"#pragma OPENCL EXTENSION cl_khr_global_int32_base_atomics : enable\n"
-"#pragma OPENCL EXTENSION cl_khr_local_int32_extended_atomics : enable\n"
-"#pragma OPENCL EXTENSION cl_khr_global_int32_extended_atomics : enable\n"
-"#ifdef cl_ext_atomic_counters_32\n"
-"#pragma OPENCL EXTENSION cl_ext_atomic_counters_32 : enable\n"
-"#else\n"
-"#define counter32_t volatile __global int*\n"
-"#endif\n"
-"typedef unsigned int u32;\n"
-"typedef unsigned short u16;\n"
-"typedef unsigned char u8;\n"
-"#define GET_GROUP_IDX get_group_id(0)\n"
-"#define GET_LOCAL_IDX get_local_id(0)\n"
-"#define GET_GLOBAL_IDX get_global_id(0)\n"
-"#define GET_GROUP_SIZE get_local_size(0)\n"
-"#define GET_NUM_GROUPS get_num_groups(0)\n"
-"#define GROUP_LDS_BARRIER barrier(CLK_LOCAL_MEM_FENCE)\n"
-"#define GROUP_MEM_FENCE mem_fence(CLK_LOCAL_MEM_FENCE)\n"
-"#define AtomInc(x) atom_inc(&(x))\n"
-"#define AtomInc1(x, out) out = atom_inc(&(x))\n"
-"#define AppendInc(x, out) out = atomic_inc(x)\n"
-"#define AtomAdd(x, value) atom_add(&(x), value)\n"
-"#define AtomCmpxhg(x, cmp, value) atom_cmpxchg( &(x), cmp, value )\n"
-"#define AtomXhg(x, value) atom_xchg ( &(x), value )\n"
-"#define SELECT_UINT4( b, a, condition ) select( b,a,condition )\n"
-"#define make_float4 (float4)\n"
-"#define make_float2 (float2)\n"
-"#define make_uint4 (uint4)\n"
-"#define make_int4 (int4)\n"
-"#define make_uint2 (uint2)\n"
-"#define make_int2 (int2)\n"
-"#define max2 max\n"
-"#define min2 min\n"
-"#define WG_SIZE 64\n"
-"typedef struct \n"
-"{\n"
-" int m_n;\n"
-" int m_start;\n"
-" int m_staticIdx;\n"
-" int m_paddings[1];\n"
-"} ConstBuffer;\n"
-"typedef struct \n"
-"{\n"
-" int m_a;\n"
-" int m_b;\n"
-" u32 m_idx;\n"
-"}Elem;\n"
-"#define STACK_SIZE (WG_SIZE*10)\n"
-"//#define STACK_SIZE (WG_SIZE)\n"
-"#define RING_SIZE 1024\n"
-"#define RING_SIZE_MASK (RING_SIZE-1)\n"
-"#define CHECK_SIZE (WG_SIZE)\n"
-"#define GET_RING_CAPACITY (RING_SIZE - ldsRingEnd)\n"
-"#define RING_END ldsTmp\n"
-"u32 readBuf(__local u32* buff, int idx)\n"
-"{\n"
-" idx = idx % (32*CHECK_SIZE);\n"
-" int bitIdx = idx%32;\n"
-" int bufIdx = idx/32;\n"
-" return buff[bufIdx] & (1<<bitIdx);\n"
-"}\n"
-"void writeBuf(__local u32* buff, int idx)\n"
-"{\n"
-" idx = idx % (32*CHECK_SIZE);\n"
-" int bitIdx = idx%32;\n"
-" int bufIdx = idx/32;\n"
-"// buff[bufIdx] |= (1<<bitIdx);\n"
-" atom_or( &buff[bufIdx], (1<<bitIdx) );\n"
-"}\n"
-"u32 tryWrite(__local u32* buff, int idx)\n"
-"{\n"
-" idx = idx % (32*CHECK_SIZE);\n"
-" int bitIdx = idx%32;\n"
-" int bufIdx = idx/32;\n"
-" u32 ans = (u32)atom_or( &buff[bufIdx], (1<<bitIdx) );\n"
-" return ((ans >> bitIdx)&1) == 0;\n"
-"}\n"
-"// batching on the GPU\n"
-"__kernel void CreateBatches( __global const struct b3Contact4Data* gConstraints, __global struct b3Contact4Data* gConstraintsOut,\n"
-" __global const u32* gN, __global const u32* gStart, __global int* batchSizes, \n"
-" int m_staticIdx )\n"
-"{\n"
-" __local u32 ldsStackIdx[STACK_SIZE];\n"
-" __local u32 ldsStackEnd;\n"
-" __local Elem ldsRingElem[RING_SIZE];\n"
-" __local u32 ldsRingEnd;\n"
-" __local u32 ldsTmp;\n"
-" __local u32 ldsCheckBuffer[CHECK_SIZE];\n"
-" __local u32 ldsFixedBuffer[CHECK_SIZE];\n"
-" __local u32 ldsGEnd;\n"
-" __local u32 ldsDstEnd;\n"
-" int wgIdx = GET_GROUP_IDX;\n"
-" int lIdx = GET_LOCAL_IDX;\n"
-" \n"
-" const int m_n = gN[wgIdx];\n"
-" const int m_start = gStart[wgIdx];\n"
-" \n"
-" if( lIdx == 0 )\n"
-" {\n"
-" ldsRingEnd = 0;\n"
-" ldsGEnd = 0;\n"
-" ldsStackEnd = 0;\n"
-" ldsDstEnd = m_start;\n"
-" }\n"
-" \n"
-" \n"
-" \n"
-"// while(1)\n"
-"//was 250\n"
-" int ie=0;\n"
-" int maxBatch = 0;\n"
-" for(ie=0; ie<50; ie++)\n"
-" {\n"
-" ldsFixedBuffer[lIdx] = 0;\n"
-" for(int giter=0; giter<4; giter++)\n"
-" {\n"
-" int ringCap = GET_RING_CAPACITY;\n"
-" \n"
-" // 1. fill ring\n"
-" if( ldsGEnd < m_n )\n"
-" {\n"
-" while( ringCap > WG_SIZE )\n"
-" {\n"
-" if( ldsGEnd >= m_n ) break;\n"
-" if( lIdx < ringCap - WG_SIZE )\n"
-" {\n"
-" int srcIdx;\n"
-" AtomInc1( ldsGEnd, srcIdx );\n"
-" if( srcIdx < m_n )\n"
-" {\n"
-" int dstIdx;\n"
-" AtomInc1( ldsRingEnd, dstIdx );\n"
-" \n"
-" int a = gConstraints[m_start+srcIdx].m_bodyAPtrAndSignBit;\n"
-" int b = gConstraints[m_start+srcIdx].m_bodyBPtrAndSignBit;\n"
-" ldsRingElem[dstIdx].m_a = (a>b)? b:a;\n"
-" ldsRingElem[dstIdx].m_b = (a>b)? a:b;\n"
-" ldsRingElem[dstIdx].m_idx = srcIdx;\n"
-" }\n"
-" }\n"
-" ringCap = GET_RING_CAPACITY;\n"
-" }\n"
-" }\n"
-" GROUP_LDS_BARRIER;\n"
-" \n"
-" // 2. fill stack\n"
-" __local Elem* dst = ldsRingElem;\n"
-" if( lIdx == 0 ) RING_END = 0;\n"
-" int srcIdx=lIdx;\n"
-" int end = ldsRingEnd;\n"
-" {\n"
-" for(int ii=0; ii<end; ii+=WG_SIZE, srcIdx+=WG_SIZE)\n"
-" {\n"
-" Elem e;\n"
-" if(srcIdx<end) e = ldsRingElem[srcIdx];\n"
-" bool done = (srcIdx<end)?false:true;\n"
-" for(int i=lIdx; i<CHECK_SIZE; i+=WG_SIZE) ldsCheckBuffer[lIdx] = 0;\n"
-" \n"
-" if( !done )\n"
-" {\n"
-" int aUsed = readBuf( ldsFixedBuffer, abs(e.m_a));\n"
-" int bUsed = readBuf( ldsFixedBuffer, abs(e.m_b));\n"
-" if( aUsed==0 && bUsed==0 )\n"
-" {\n"
-" int aAvailable=1;\n"
-" int bAvailable=1;\n"
-" int ea = abs(e.m_a);\n"
-" int eb = abs(e.m_b);\n"
-" bool aStatic = (e.m_a<0) ||(ea==m_staticIdx);\n"
-" bool bStatic = (e.m_b<0) ||(eb==m_staticIdx);\n"
-" \n"
-" if (!aStatic)\n"
-" aAvailable = tryWrite( ldsCheckBuffer, ea );\n"
-" if (!bStatic)\n"
-" bAvailable = tryWrite( ldsCheckBuffer, eb );\n"
-" \n"
-" //aAvailable = aStatic? 1: aAvailable;\n"
-" //bAvailable = bStatic? 1: bAvailable;\n"
-" bool success = (aAvailable && bAvailable);\n"
-" if(success)\n"
-" {\n"
-" \n"
-" if (!aStatic)\n"
-" writeBuf( ldsFixedBuffer, ea );\n"
-" if (!bStatic)\n"
-" writeBuf( ldsFixedBuffer, eb );\n"
-" }\n"
-" done = success;\n"
-" }\n"
-" }\n"
-" // put it aside\n"
-" if(srcIdx<end)\n"
-" {\n"
-" if( done )\n"
-" {\n"
-" int dstIdx; AtomInc1( ldsStackEnd, dstIdx );\n"
-" if( dstIdx < STACK_SIZE )\n"
-" ldsStackIdx[dstIdx] = e.m_idx;\n"
-" else{\n"
-" done = false;\n"
-" AtomAdd( ldsStackEnd, -1 );\n"
-" }\n"
-" }\n"
-" if( !done )\n"
-" {\n"
-" int dstIdx; AtomInc1( RING_END, dstIdx );\n"
-" dst[dstIdx] = e;\n"
-" }\n"
-" }\n"
-" // if filled, flush\n"
-" if( ldsStackEnd == STACK_SIZE )\n"
-" {\n"
-" for(int i=lIdx; i<STACK_SIZE; i+=WG_SIZE)\n"
-" {\n"
-" int idx = m_start + ldsStackIdx[i];\n"
-" int dstIdx; AtomInc1( ldsDstEnd, dstIdx );\n"
-" gConstraintsOut[ dstIdx ] = gConstraints[ idx ];\n"
-" gConstraintsOut[ dstIdx ].m_batchIdx = ie;\n"
-" }\n"
-" if( lIdx == 0 ) ldsStackEnd = 0;\n"
-" //for(int i=lIdx; i<CHECK_SIZE; i+=WG_SIZE) \n"
-" ldsFixedBuffer[lIdx] = 0;\n"
-" }\n"
-" }\n"
-" }\n"
-" if( lIdx == 0 ) ldsRingEnd = RING_END;\n"
-" }\n"
-" GROUP_LDS_BARRIER;\n"
-" for(int i=lIdx; i<ldsStackEnd; i+=WG_SIZE)\n"
-" {\n"
-" int idx = m_start + ldsStackIdx[i];\n"
-" int dstIdx; AtomInc1( ldsDstEnd, dstIdx );\n"
-" gConstraintsOut[ dstIdx ] = gConstraints[ idx ];\n"
-" gConstraintsOut[ dstIdx ].m_batchIdx = ie;\n"
-" }\n"
-" // in case it couldn't consume any pair. Flush them\n"
-" // todo. Serial batch worth while?\n"
-" if( ldsStackEnd == 0 )\n"
-" {\n"
-" for(int i=lIdx; i<ldsRingEnd; i+=WG_SIZE)\n"
-" {\n"
-" int idx = m_start + ldsRingElem[i].m_idx;\n"
-" int dstIdx; AtomInc1( ldsDstEnd, dstIdx );\n"
-" gConstraintsOut[ dstIdx ] = gConstraints[ idx ];\n"
-" int curBatch = 100+i;\n"
-" if (maxBatch < curBatch)\n"
-" maxBatch = curBatch;\n"
-" \n"
-" gConstraintsOut[ dstIdx ].m_batchIdx = curBatch;\n"
-" \n"
-" }\n"
-" GROUP_LDS_BARRIER;\n"
-" if( lIdx == 0 ) ldsRingEnd = 0;\n"
-" }\n"
-" if( lIdx == 0 ) ldsStackEnd = 0;\n"
-" GROUP_LDS_BARRIER;\n"
-" // termination\n"
-" if( ldsGEnd == m_n && ldsRingEnd == 0 )\n"
-" break;\n"
-" }\n"
-" if( lIdx == 0 )\n"
-" {\n"
-" if (maxBatch < ie)\n"
-" maxBatch=ie;\n"
-" batchSizes[wgIdx]=maxBatch;\n"
-" }\n"
-"}\n"
-;
/include/Bullet/Bullet3OpenCL/RigidBody/kernels/batchingKernels.h
Property changes:
Deleted: svn:executable
## -1 +0,0 ##
-*
\ No newline at end of property
Index: include/Bullet/Bullet3OpenCL/RigidBody/kernels/integrateKernel.h
===================================================================
--- include/Bullet/Bullet3OpenCL/RigidBody/kernels/integrateKernel.h (revision 22)
+++ include/Bullet/Bullet3OpenCL/RigidBody/kernels/integrateKernel.h (nonexistent)
@@ -1,432 +0,0 @@
-//this file is autogenerated using stringify.bat (premake --stringify) in the build folder of this project
-static const char* integrateKernelCL= \
-"/*\n"
-"Copyright (c) 2013 Advanced Micro Devices, Inc. \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"
-"//Originally written by Erwin Coumans\n"
-"#ifndef B3_RIGIDBODY_DATA_H\n"
-"#define B3_RIGIDBODY_DATA_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"
-"//keep B3_LARGE_FLOAT*B3_LARGE_FLOAT < FLT_MAX\n"
-"#define B3_LARGE_FLOAT 1e18f\n"
-"#define B3_INFINITY 1e18f\n"
-"#define b3Assert(a)\n"
-"#define b3ConstArray(a) __global const a*\n"
-"#define b3AtomicInc atomic_inc\n"
-"#define b3AtomicAdd atomic_add\n"
-"#define b3Fabs fabs\n"
-"#define b3Sqrt native_sqrt\n"
-"#define b3Sin native_sin\n"
-"#define b3Cos native_cos\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"
-" #define b3MinFloat4 min\n"
-" #define b3MaxFloat4 max\n"
-" #define b3Normalized(a) normalize(a)\n"
-"#endif \n"
-" \n"
-"inline bool b3IsAlmostZero(b3Float4ConstArg v)\n"
-"{\n"
-" if(b3Fabs(v.x)>1e-6 || b3Fabs(v.y)>1e-6 || b3Fabs(v.z)>1e-6) \n"
-" return false;\n"
-" return true;\n"
-"}\n"
-"inline int b3MaxDot( b3Float4ConstArg vec, __global const b3Float4* vecArray, int vecLen, float* dotOut )\n"
-"{\n"
-" float maxDot = -B3_INFINITY;\n"
-" int i = 0;\n"
-" int ptIndex = -1;\n"
-" for( i = 0; i < vecLen; i++ )\n"
-" {\n"
-" float dot = b3Dot3F4(vecArray[i],vec);\n"
-" \n"
-" if( dot > maxDot )\n"
-" {\n"
-" maxDot = dot;\n"
-" ptIndex = i;\n"
-" }\n"
-" }\n"
-" b3Assert(ptIndex>=0);\n"
-" if (ptIndex<0)\n"
-" {\n"
-" ptIndex = 0;\n"
-" }\n"
-" *dotOut = maxDot;\n"
-" return ptIndex;\n"
-"}\n"
-"#endif //B3_FLOAT4_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 b3QuatNormalized(b3QuatConstArg in);\n"
-"inline b3Quat b3QuatRotate(b3QuatConstArg q, b3QuatConstArg vec);\n"
-"inline b3Quat b3QuatInvert(b3QuatConstArg q);\n"
-"inline b3Quat b3QuatInverse(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 b3QuatNormalized(b3QuatConstArg in)\n"
-"{\n"
-" b3Quat q;\n"
-" q=in;\n"
-" //return b3FastNormalize4(in);\n"
-" float len = native_sqrt(dot(q, q));\n"
-" if(len > 0.f)\n"
-" {\n"
-" q *= 1.f / len;\n"
-" }\n"
-" else\n"
-" {\n"
-" q.x = q.y = q.z = 0.f;\n"
-" q.w = 1.f;\n"
-" }\n"
-" return q;\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 b3QuatInverse(b3QuatConstArg q)\n"
-"{\n"
-" return (b3Quat)(-q.xyz, q.w);\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"
-"#ifndef B3_MAT3x3_H\n"
-"#define B3_MAT3x3_H\n"
-"#ifndef B3_QUAT_H\n"
-"#ifdef __cplusplus\n"
-"#else\n"
-"#endif \n"
-"#endif //B3_QUAT_H\n"
-"#ifdef __cplusplus\n"
-"#else\n"
-"typedef struct\n"
-"{\n"
-" b3Float4 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"
-" b3Float4 quat2 = (b3Float4)(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"
-"__inline\n"
-"b3Mat3x3 mtZero();\n"
-"__inline\n"
-"b3Mat3x3 mtIdentity();\n"
-"__inline\n"
-"b3Mat3x3 mtTranspose(b3Mat3x3 m);\n"
-"__inline\n"
-"b3Mat3x3 mtMul(b3Mat3x3 a, b3Mat3x3 b);\n"
-"__inline\n"
-"b3Float4 mtMul1(b3Mat3x3 a, b3Float4 b);\n"
-"__inline\n"
-"b3Float4 mtMul3(b3Float4 a, b3Mat3x3 b);\n"
-"__inline\n"
-"b3Mat3x3 mtZero()\n"
-"{\n"
-" b3Mat3x3 m;\n"
-" m.m_row[0] = (b3Float4)(0.f);\n"
-" m.m_row[1] = (b3Float4)(0.f);\n"
-" m.m_row[2] = (b3Float4)(0.f);\n"
-" return m;\n"
-"}\n"
-"__inline\n"
-"b3Mat3x3 mtIdentity()\n"
-"{\n"
-" b3Mat3x3 m;\n"
-" m.m_row[0] = (b3Float4)(1,0,0,0);\n"
-" m.m_row[1] = (b3Float4)(0,1,0,0);\n"
-" m.m_row[2] = (b3Float4)(0,0,1,0);\n"
-" return m;\n"
-"}\n"
-"__inline\n"
-"b3Mat3x3 mtTranspose(b3Mat3x3 m)\n"
-"{\n"
-" b3Mat3x3 out;\n"
-" out.m_row[0] = (b3Float4)(m.m_row[0].x, m.m_row[1].x, m.m_row[2].x, 0.f);\n"
-" out.m_row[1] = (b3Float4)(m.m_row[0].y, m.m_row[1].y, m.m_row[2].y, 0.f);\n"
-" out.m_row[2] = (b3Float4)(m.m_row[0].z, m.m_row[1].z, m.m_row[2].z, 0.f);\n"
-" return out;\n"
-"}\n"
-"__inline\n"
-"b3Mat3x3 mtMul(b3Mat3x3 a, b3Mat3x3 b)\n"
-"{\n"
-" b3Mat3x3 transB;\n"
-" transB = mtTranspose( b );\n"
-" b3Mat3x3 ans;\n"
-" // why this doesn't run when 0ing in the for{}\n"
-" a.m_row[0].w = 0.f;\n"
-" a.m_row[1].w = 0.f;\n"
-" a.m_row[2].w = 0.f;\n"
-" for(int i=0; i<3; i++)\n"
-" {\n"
-"// a.m_row[i].w = 0.f;\n"
-" ans.m_row[i].x = b3Dot3F4(a.m_row[i],transB.m_row[0]);\n"
-" ans.m_row[i].y = b3Dot3F4(a.m_row[i],transB.m_row[1]);\n"
-" ans.m_row[i].z = b3Dot3F4(a.m_row[i],transB.m_row[2]);\n"
-" ans.m_row[i].w = 0.f;\n"
-" }\n"
-" return ans;\n"
-"}\n"
-"__inline\n"
-"b3Float4 mtMul1(b3Mat3x3 a, b3Float4 b)\n"
-"{\n"
-" b3Float4 ans;\n"
-" ans.x = b3Dot3F4( a.m_row[0], b );\n"
-" ans.y = b3Dot3F4( a.m_row[1], b );\n"
-" ans.z = b3Dot3F4( a.m_row[2], b );\n"
-" ans.w = 0.f;\n"
-" return ans;\n"
-"}\n"
-"__inline\n"
-"b3Float4 mtMul3(b3Float4 a, b3Mat3x3 b)\n"
-"{\n"
-" b3Float4 colx = b3MakeFloat4(b.m_row[0].x, b.m_row[1].x, b.m_row[2].x, 0);\n"
-" b3Float4 coly = b3MakeFloat4(b.m_row[0].y, b.m_row[1].y, b.m_row[2].y, 0);\n"
-" b3Float4 colz = b3MakeFloat4(b.m_row[0].z, b.m_row[1].z, b.m_row[2].z, 0);\n"
-" b3Float4 ans;\n"
-" ans.x = b3Dot3F4( a, colx );\n"
-" ans.y = b3Dot3F4( a, coly );\n"
-" ans.z = b3Dot3F4( a, colz );\n"
-" return ans;\n"
-"}\n"
-"#endif\n"
-"#endif //B3_MAT3x3_H\n"
-"typedef struct b3RigidBodyData b3RigidBodyData_t;\n"
-"struct b3RigidBodyData\n"
-"{\n"
-" b3Float4 m_pos;\n"
-" b3Quat m_quat;\n"
-" b3Float4 m_linVel;\n"
-" b3Float4 m_angVel;\n"
-" int m_collidableIdx;\n"
-" float m_invMass;\n"
-" float m_restituitionCoeff;\n"
-" float m_frictionCoeff;\n"
-"};\n"
-"typedef struct b3InertiaData b3InertiaData_t;\n"
-"struct b3InertiaData\n"
-"{\n"
-" b3Mat3x3 m_invInertiaWorld;\n"
-" b3Mat3x3 m_initInvInertia;\n"
-"};\n"
-"#endif //B3_RIGIDBODY_DATA_H\n"
-" \n"
-"#ifndef B3_RIGIDBODY_DATA_H\n"
-"#endif //B3_RIGIDBODY_DATA_H\n"
-" \n"
-"inline void integrateSingleTransform( __global b3RigidBodyData_t* bodies,int nodeID, float timeStep, float angularDamping, b3Float4ConstArg gravityAcceleration)\n"
-"{\n"
-" \n"
-" if (bodies[nodeID].m_invMass != 0.f)\n"
-" {\n"
-" float BT_GPU_ANGULAR_MOTION_THRESHOLD = (0.25f * 3.14159254f);\n"
-" //angular velocity\n"
-" {\n"
-" b3Float4 axis;\n"
-" //add some hardcoded angular damping\n"
-" bodies[nodeID].m_angVel.x *= angularDamping;\n"
-" bodies[nodeID].m_angVel.y *= angularDamping;\n"
-" bodies[nodeID].m_angVel.z *= angularDamping;\n"
-" \n"
-" b3Float4 angvel = bodies[nodeID].m_angVel;\n"
-" float fAngle = b3Sqrt(b3Dot3F4(angvel, angvel));\n"
-" \n"
-" //limit the angular motion\n"
-" if(fAngle*timeStep > BT_GPU_ANGULAR_MOTION_THRESHOLD)\n"
-" {\n"
-" fAngle = BT_GPU_ANGULAR_MOTION_THRESHOLD / timeStep;\n"
-" }\n"
-" if(fAngle < 0.001f)\n"
-" {\n"
-" // use Taylor's expansions of sync function\n"
-" axis = angvel * (0.5f*timeStep-(timeStep*timeStep*timeStep)*0.020833333333f * fAngle * fAngle);\n"
-" }\n"
-" else\n"
-" {\n"
-" // sync(fAngle) = sin(c*fAngle)/t\n"
-" axis = angvel * ( b3Sin(0.5f * fAngle * timeStep) / fAngle);\n"
-" }\n"
-" \n"
-" b3Quat dorn;\n"
-" dorn.x = axis.x;\n"
-" dorn.y = axis.y;\n"
-" dorn.z = axis.z;\n"
-" dorn.w = b3Cos(fAngle * timeStep * 0.5f);\n"
-" b3Quat orn0 = bodies[nodeID].m_quat;\n"
-" b3Quat predictedOrn = b3QuatMul(dorn, orn0);\n"
-" predictedOrn = b3QuatNormalized(predictedOrn);\n"
-" bodies[nodeID].m_quat=predictedOrn;\n"
-" }\n"
-" //linear velocity \n"
-" bodies[nodeID].m_pos += bodies[nodeID].m_linVel * timeStep;\n"
-" \n"
-" //apply gravity\n"
-" bodies[nodeID].m_linVel += gravityAcceleration * timeStep;\n"
-" \n"
-" }\n"
-" \n"
-"}\n"
-"inline void b3IntegrateTransform( __global b3RigidBodyData_t* body, float timeStep, float angularDamping, b3Float4ConstArg gravityAcceleration)\n"
-"{\n"
-" float BT_GPU_ANGULAR_MOTION_THRESHOLD = (0.25f * 3.14159254f);\n"
-" \n"
-" if( (body->m_invMass != 0.f))\n"
-" {\n"
-" //angular velocity\n"
-" {\n"
-" b3Float4 axis;\n"
-" //add some hardcoded angular damping\n"
-" body->m_angVel.x *= angularDamping;\n"
-" body->m_angVel.y *= angularDamping;\n"
-" body->m_angVel.z *= angularDamping;\n"
-" \n"
-" b3Float4 angvel = body->m_angVel;\n"
-" float fAngle = b3Sqrt(b3Dot3F4(angvel, angvel));\n"
-" //limit the angular motion\n"
-" if(fAngle*timeStep > BT_GPU_ANGULAR_MOTION_THRESHOLD)\n"
-" {\n"
-" fAngle = BT_GPU_ANGULAR_MOTION_THRESHOLD / timeStep;\n"
-" }\n"
-" if(fAngle < 0.001f)\n"
-" {\n"
-" // use Taylor's expansions of sync function\n"
-" axis = angvel * (0.5f*timeStep-(timeStep*timeStep*timeStep)*0.020833333333f * fAngle * fAngle);\n"
-" }\n"
-" else\n"
-" {\n"
-" // sync(fAngle) = sin(c*fAngle)/t\n"
-" axis = angvel * ( b3Sin(0.5f * fAngle * timeStep) / fAngle);\n"
-" }\n"
-" b3Quat dorn;\n"
-" dorn.x = axis.x;\n"
-" dorn.y = axis.y;\n"
-" dorn.z = axis.z;\n"
-" dorn.w = b3Cos(fAngle * timeStep * 0.5f);\n"
-" b3Quat orn0 = body->m_quat;\n"
-" b3Quat predictedOrn = b3QuatMul(dorn, orn0);\n"
-" predictedOrn = b3QuatNormalized(predictedOrn);\n"
-" body->m_quat=predictedOrn;\n"
-" }\n"
-" //apply gravity\n"
-" body->m_linVel += gravityAcceleration * timeStep;\n"
-" //linear velocity \n"
-" body->m_pos += body->m_linVel * timeStep;\n"
-" \n"
-" }\n"
-" \n"
-"}\n"
-"__kernel void \n"
-" integrateTransformsKernel( __global b3RigidBodyData_t* bodies,const int numNodes, float timeStep, float angularDamping, float4 gravityAcceleration)\n"
-"{\n"
-" int nodeID = get_global_id(0);\n"
-" \n"
-" if( nodeID < numNodes)\n"
-" {\n"
-" integrateSingleTransform(bodies,nodeID, timeStep, angularDamping,gravityAcceleration);\n"
-" }\n"
-"}\n"
-;
/include/Bullet/Bullet3OpenCL/RigidBody/kernels/integrateKernel.h
Property changes:
Deleted: svn:executable
## -1 +0,0 ##
-*
\ No newline at end of property
Index: include/Bullet/Bullet3OpenCL/RigidBody/kernels/solveContact.h
===================================================================
--- include/Bullet/Bullet3OpenCL/RigidBody/kernels/solveContact.h (revision 22)
+++ include/Bullet/Bullet3OpenCL/RigidBody/kernels/solveContact.h (nonexistent)
@@ -1,393 +0,0 @@
-//this file is autogenerated using stringify.bat (premake --stringify) in the build folder of this project
-static const char* solveContactCL= \
-"/*\n"
-"Copyright (c) 2012 Advanced Micro Devices, Inc. \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"
-"//Originally written by Takahiro Harada\n"
-"//#pragma OPENCL EXTENSION cl_amd_printf : enable\n"
-"#pragma OPENCL EXTENSION cl_khr_local_int32_base_atomics : enable\n"
-"#pragma OPENCL EXTENSION cl_khr_global_int32_base_atomics : enable\n"
-"#pragma OPENCL EXTENSION cl_khr_local_int32_extended_atomics : enable\n"
-"#pragma OPENCL EXTENSION cl_khr_global_int32_extended_atomics : enable\n"
-"#ifdef cl_ext_atomic_counters_32\n"
-"#pragma OPENCL EXTENSION cl_ext_atomic_counters_32 : enable\n"
-"#else\n"
-"#define counter32_t volatile global int*\n"
-"#endif\n"
-"typedef unsigned int u32;\n"
-"typedef unsigned short u16;\n"
-"typedef unsigned char u8;\n"
-"#define GET_GROUP_IDX get_group_id(0)\n"
-"#define GET_LOCAL_IDX get_local_id(0)\n"
-"#define GET_GLOBAL_IDX get_global_id(0)\n"
-"#define GET_GROUP_SIZE get_local_size(0)\n"
-"#define GET_NUM_GROUPS get_num_groups(0)\n"
-"#define GROUP_LDS_BARRIER barrier(CLK_LOCAL_MEM_FENCE)\n"
-"#define GROUP_MEM_FENCE mem_fence(CLK_LOCAL_MEM_FENCE)\n"
-"#define AtomInc(x) atom_inc(&(x))\n"
-"#define AtomInc1(x, out) out = atom_inc(&(x))\n"
-"#define AppendInc(x, out) out = atomic_inc(x)\n"
-"#define AtomAdd(x, value) atom_add(&(x), value)\n"
-"#define AtomCmpxhg(x, cmp, value) atom_cmpxchg( &(x), cmp, value )\n"
-"#define AtomXhg(x, value) atom_xchg ( &(x), value )\n"
-"#define SELECT_UINT4( b, a, condition ) select( b,a,condition )\n"
-"#define mymake_float4 (float4)\n"
-"//#define make_float2 (float2)\n"
-"//#define make_uint4 (uint4)\n"
-"//#define make_int4 (int4)\n"
-"//#define make_uint2 (uint2)\n"
-"//#define make_int2 (int2)\n"
-"#define max2 max\n"
-"#define min2 min\n"
-"///////////////////////////////////////\n"
-"// Vector\n"
-"///////////////////////////////////////\n"
-"__inline\n"
-"float4 fastNormalize4(float4 v)\n"
-"{\n"
-" return fast_normalize(v);\n"
-"}\n"
-"__inline\n"
-"float4 cross3(float4 a, float4 b)\n"
-"{\n"
-" return cross(a,b);\n"
-"}\n"
-"__inline\n"
-"float dot3F4(float4 a, float4 b)\n"
-"{\n"
-" float4 a1 = mymake_float4(a.xyz,0.f);\n"
-" float4 b1 = mymake_float4(b.xyz,0.f);\n"
-" return dot(a1, b1);\n"
-"}\n"
-"__inline\n"
-"float4 normalize3(const float4 a)\n"
-"{\n"
-" float4 n = mymake_float4(a.x, a.y, a.z, 0.f);\n"
-" return fastNormalize4( n );\n"
-"// float length = sqrtf(dot3F4(a, a));\n"
-"// return 1.f/length * a;\n"
-"}\n"
-"///////////////////////////////////////\n"
-"// Matrix3x3\n"
-"///////////////////////////////////////\n"
-"typedef struct\n"
-"{\n"
-" float4 m_row[3];\n"
-"}Matrix3x3;\n"
-"__inline\n"
-"float4 mtMul1(Matrix3x3 a, float4 b);\n"
-"__inline\n"
-"float4 mtMul3(float4 a, Matrix3x3 b);\n"
-"__inline\n"
-"float4 mtMul1(Matrix3x3 a, float4 b)\n"
-"{\n"
-" float4 ans;\n"
-" ans.x = dot3F4( a.m_row[0], b );\n"
-" ans.y = dot3F4( a.m_row[1], b );\n"
-" ans.z = dot3F4( a.m_row[2], b );\n"
-" ans.w = 0.f;\n"
-" return ans;\n"
-"}\n"
-"__inline\n"
-"float4 mtMul3(float4 a, Matrix3x3 b)\n"
-"{\n"
-" float4 colx = mymake_float4(b.m_row[0].x, b.m_row[1].x, b.m_row[2].x, 0);\n"
-" float4 coly = mymake_float4(b.m_row[0].y, b.m_row[1].y, b.m_row[2].y, 0);\n"
-" float4 colz = mymake_float4(b.m_row[0].z, b.m_row[1].z, b.m_row[2].z, 0);\n"
-" float4 ans;\n"
-" ans.x = dot3F4( a, colx );\n"
-" ans.y = dot3F4( a, coly );\n"
-" ans.z = dot3F4( a, colz );\n"
-" return ans;\n"
-"}\n"
-"///////////////////////////////////////\n"
-"// Quaternion\n"
-"///////////////////////////////////////\n"
-"typedef float4 Quaternion;\n"
-"#define WG_SIZE 64\n"
-"typedef struct\n"
-"{\n"
-" float4 m_pos;\n"
-" Quaternion m_quat;\n"
-" float4 m_linVel;\n"
-" float4 m_angVel;\n"
-" u32 m_shapeIdx;\n"
-" float m_invMass;\n"
-" float m_restituitionCoeff;\n"
-" float m_frictionCoeff;\n"
-"} Body;\n"
-"typedef struct\n"
-"{\n"
-" Matrix3x3 m_invInertia;\n"
-" Matrix3x3 m_initInvInertia;\n"
-"} Shape;\n"
-"typedef struct\n"
-"{\n"
-" float4 m_linear;\n"
-" float4 m_worldPos[4];\n"
-" float4 m_center; \n"
-" float m_jacCoeffInv[4];\n"
-" float m_b[4];\n"
-" float m_appliedRambdaDt[4];\n"
-" float m_fJacCoeffInv[2]; \n"
-" float m_fAppliedRambdaDt[2]; \n"
-" u32 m_bodyA;\n"
-" u32 m_bodyB;\n"
-" int m_batchIdx;\n"
-" u32 m_paddings[1];\n"
-"} Constraint4;\n"
-"typedef struct\n"
-"{\n"
-" int m_nConstraints;\n"
-" int m_start;\n"
-" int m_batchIdx;\n"
-" int m_nSplit;\n"
-"// int m_paddings[1];\n"
-"} ConstBuffer;\n"
-"typedef struct\n"
-"{\n"
-" int m_solveFriction;\n"
-" int m_maxBatch; // long batch really kills the performance\n"
-" int m_batchIdx;\n"
-" int m_nSplit;\n"
-"// int m_paddings[1];\n"
-"} ConstBufferBatchSolve;\n"
-"void setLinearAndAngular( float4 n, float4 r0, float4 r1, float4* linear, float4* angular0, float4* angular1);\n"
-"void setLinearAndAngular( float4 n, float4 r0, float4 r1, float4* linear, float4* angular0, float4* angular1)\n"
-"{\n"
-" *linear = mymake_float4(-n.xyz,0.f);\n"
-" *angular0 = -cross3(r0, n);\n"
-" *angular1 = cross3(r1, n);\n"
-"}\n"
-"float calcRelVel( float4 l0, float4 l1, float4 a0, float4 a1, float4 linVel0, float4 angVel0, float4 linVel1, float4 angVel1 );\n"
-"float calcRelVel( float4 l0, float4 l1, float4 a0, float4 a1, float4 linVel0, float4 angVel0, float4 linVel1, float4 angVel1 )\n"
-"{\n"
-" return dot3F4(l0, linVel0) + dot3F4(a0, angVel0) + dot3F4(l1, linVel1) + dot3F4(a1, angVel1);\n"
-"}\n"
-"float calcJacCoeff(const float4 linear0, const float4 linear1, const float4 angular0, const float4 angular1,\n"
-" float invMass0, const Matrix3x3* invInertia0, float invMass1, const Matrix3x3* invInertia1);\n"
-"float calcJacCoeff(const float4 linear0, const float4 linear1, const float4 angular0, const float4 angular1,\n"
-" float invMass0, const Matrix3x3* invInertia0, float invMass1, const Matrix3x3* invInertia1)\n"
-"{\n"
-" // linear0,1 are normlized\n"
-" float jmj0 = invMass0;//dot3F4(linear0, linear0)*invMass0;\n"
-" float jmj1 = dot3F4(mtMul3(angular0,*invInertia0), angular0);\n"
-" float jmj2 = invMass1;//dot3F4(linear1, linear1)*invMass1;\n"
-" float jmj3 = dot3F4(mtMul3(angular1,*invInertia1), angular1);\n"
-" return -1.f/(jmj0+jmj1+jmj2+jmj3);\n"
-"}\n"
-"void solveContact(__global Constraint4* cs,\n"
-" float4 posA, float4* linVelA, float4* angVelA, float invMassA, Matrix3x3 invInertiaA,\n"
-" float4 posB, float4* linVelB, float4* angVelB, float invMassB, Matrix3x3 invInertiaB);\n"
-"void solveContact(__global Constraint4* cs,\n"
-" float4 posA, float4* linVelA, float4* angVelA, float invMassA, Matrix3x3 invInertiaA,\n"
-" float4 posB, float4* linVelB, float4* angVelB, float invMassB, Matrix3x3 invInertiaB)\n"
-"{\n"
-" float minRambdaDt = 0;\n"
-" float maxRambdaDt = FLT_MAX;\n"
-" for(int ic=0; ic<4; ic++)\n"
-" {\n"
-" if( cs->m_jacCoeffInv[ic] == 0.f ) continue;\n"
-" float4 angular0, angular1, linear;\n"
-" float4 r0 = cs->m_worldPos[ic] - posA;\n"
-" float4 r1 = cs->m_worldPos[ic] - posB;\n"
-" setLinearAndAngular( -cs->m_linear, r0, r1, &linear, &angular0, &angular1 );\n"
-" float rambdaDt = calcRelVel( cs->m_linear, -cs->m_linear, angular0, angular1, \n"
-" *linVelA, *angVelA, *linVelB, *angVelB ) + cs->m_b[ic];\n"
-" rambdaDt *= cs->m_jacCoeffInv[ic];\n"
-" {\n"
-" float prevSum = cs->m_appliedRambdaDt[ic];\n"
-" float updated = prevSum;\n"
-" updated += rambdaDt;\n"
-" updated = max2( updated, minRambdaDt );\n"
-" updated = min2( updated, maxRambdaDt );\n"
-" rambdaDt = updated - prevSum;\n"
-" cs->m_appliedRambdaDt[ic] = updated;\n"
-" }\n"
-" float4 linImp0 = invMassA*linear*rambdaDt;\n"
-" float4 linImp1 = invMassB*(-linear)*rambdaDt;\n"
-" float4 angImp0 = mtMul1(invInertiaA, angular0)*rambdaDt;\n"
-" float4 angImp1 = mtMul1(invInertiaB, angular1)*rambdaDt;\n"
-" *linVelA += linImp0;\n"
-" *angVelA += angImp0;\n"
-" *linVelB += linImp1;\n"
-" *angVelB += angImp1;\n"
-" }\n"
-"}\n"
-"void btPlaneSpace1 (const float4* n, float4* p, float4* q);\n"
-" void btPlaneSpace1 (const float4* n, float4* p, float4* q)\n"
-"{\n"
-" if (fabs(n[0].z) > 0.70710678f) {\n"
-" // choose p in y-z plane\n"
-" float a = n[0].y*n[0].y + n[0].z*n[0].z;\n"
-" float k = 1.f/sqrt(a);\n"
-" p[0].x = 0;\n"
-" p[0].y = -n[0].z*k;\n"
-" p[0].z = n[0].y*k;\n"
-" // set q = n x p\n"
-" q[0].x = a*k;\n"
-" q[0].y = -n[0].x*p[0].z;\n"
-" q[0].z = n[0].x*p[0].y;\n"
-" }\n"
-" else {\n"
-" // choose p in x-y plane\n"
-" float a = n[0].x*n[0].x + n[0].y*n[0].y;\n"
-" float k = 1.f/sqrt(a);\n"
-" p[0].x = -n[0].y*k;\n"
-" p[0].y = n[0].x*k;\n"
-" p[0].z = 0;\n"
-" // set q = n x p\n"
-" q[0].x = -n[0].z*p[0].y;\n"
-" q[0].y = n[0].z*p[0].x;\n"
-" q[0].z = a*k;\n"
-" }\n"
-"}\n"
-"void solveContactConstraint(__global Body* gBodies, __global Shape* gShapes, __global Constraint4* ldsCs);\n"
-"void solveContactConstraint(__global Body* gBodies, __global Shape* gShapes, __global Constraint4* ldsCs)\n"
-"{\n"
-" //float frictionCoeff = ldsCs[0].m_linear.w;\n"
-" int aIdx = ldsCs[0].m_bodyA;\n"
-" int bIdx = ldsCs[0].m_bodyB;\n"
-" float4 posA = gBodies[aIdx].m_pos;\n"
-" float4 linVelA = gBodies[aIdx].m_linVel;\n"
-" float4 angVelA = gBodies[aIdx].m_angVel;\n"
-" float invMassA = gBodies[aIdx].m_invMass;\n"
-" Matrix3x3 invInertiaA = gShapes[aIdx].m_invInertia;\n"
-" float4 posB = gBodies[bIdx].m_pos;\n"
-" float4 linVelB = gBodies[bIdx].m_linVel;\n"
-" float4 angVelB = gBodies[bIdx].m_angVel;\n"
-" float invMassB = gBodies[bIdx].m_invMass;\n"
-" Matrix3x3 invInertiaB = gShapes[bIdx].m_invInertia;\n"
-" solveContact( ldsCs, posA, &linVelA, &angVelA, invMassA, invInertiaA,\n"
-" posB, &linVelB, &angVelB, invMassB, invInertiaB );\n"
-" if (gBodies[aIdx].m_invMass)\n"
-" {\n"
-" gBodies[aIdx].m_linVel = linVelA;\n"
-" gBodies[aIdx].m_angVel = angVelA;\n"
-" } else\n"
-" {\n"
-" gBodies[aIdx].m_linVel = mymake_float4(0,0,0,0);\n"
-" gBodies[aIdx].m_angVel = mymake_float4(0,0,0,0);\n"
-" \n"
-" }\n"
-" if (gBodies[bIdx].m_invMass)\n"
-" {\n"
-" gBodies[bIdx].m_linVel = linVelB;\n"
-" gBodies[bIdx].m_angVel = angVelB;\n"
-" } else\n"
-" {\n"
-" gBodies[bIdx].m_linVel = mymake_float4(0,0,0,0);\n"
-" gBodies[bIdx].m_angVel = mymake_float4(0,0,0,0);\n"
-" \n"
-" }\n"
-"}\n"
-"typedef struct \n"
-"{\n"
-" int m_valInt0;\n"
-" int m_valInt1;\n"
-" int m_valInt2;\n"
-" int m_valInt3;\n"
-" float m_val0;\n"
-" float m_val1;\n"
-" float m_val2;\n"
-" float m_val3;\n"
-"} SolverDebugInfo;\n"
-"__kernel\n"
-"__attribute__((reqd_work_group_size(WG_SIZE,1,1)))\n"
-"void BatchSolveKernelContact(__global Body* gBodies,\n"
-" __global Shape* gShapes,\n"
-" __global Constraint4* gConstraints,\n"
-" __global int* gN,\n"
-" __global int* gOffsets,\n"
-" __global int* batchSizes,\n"
-" int maxBatch1,\n"
-" int cellBatch,\n"
-" int4 nSplit\n"
-" )\n"
-"{\n"
-" //__local int ldsBatchIdx[WG_SIZE+1];\n"
-" __local int ldsCurBatch;\n"
-" __local int ldsNextBatch;\n"
-" __local int ldsStart;\n"
-" int lIdx = GET_LOCAL_IDX;\n"
-" int wgIdx = GET_GROUP_IDX;\n"
-"// int gIdx = GET_GLOBAL_IDX;\n"
-"// debugInfo[gIdx].m_valInt0 = gIdx;\n"
-" //debugInfo[gIdx].m_valInt1 = GET_GROUP_SIZE;\n"
-" \n"
-" \n"
-" int zIdx = (wgIdx/((nSplit.x*nSplit.y)/4))*2+((cellBatch&4)>>2);\n"
-" int remain= (wgIdx%((nSplit.x*nSplit.y)/4));\n"
-" int yIdx = (remain/(nSplit.x/2))*2 + ((cellBatch&2)>>1);\n"
-" int xIdx = (remain%(nSplit.x/2))*2 + (cellBatch&1);\n"
-" int cellIdx = xIdx+yIdx*nSplit.x+zIdx*(nSplit.x*nSplit.y);\n"
-" //int xIdx = (wgIdx/(nSplit/2))*2 + (bIdx&1);\n"
-" //int yIdx = (wgIdx%(nSplit/2))*2 + (bIdx>>1);\n"
-" //int cellIdx = xIdx+yIdx*nSplit;\n"
-" \n"
-" if( gN[cellIdx] == 0 ) \n"
-" return;\n"
-" int maxBatch = batchSizes[cellIdx];\n"
-" \n"
-" \n"
-" const int start = gOffsets[cellIdx];\n"
-" const int end = start + gN[cellIdx];\n"
-" \n"
-" \n"
-" \n"
-" if( lIdx == 0 )\n"
-" {\n"
-" ldsCurBatch = 0;\n"
-" ldsNextBatch = 0;\n"
-" ldsStart = start;\n"
-" }\n"
-" GROUP_LDS_BARRIER;\n"
-" int idx=ldsStart+lIdx;\n"
-" while (ldsCurBatch < maxBatch)\n"
-" {\n"
-" for(; idx<end; )\n"
-" {\n"
-" if (gConstraints[idx].m_batchIdx == ldsCurBatch)\n"
-" {\n"
-" solveContactConstraint( gBodies, gShapes, &gConstraints[idx] );\n"
-" idx+=64;\n"
-" } else\n"
-" {\n"
-" break;\n"
-" }\n"
-" }\n"
-" GROUP_LDS_BARRIER;\n"
-" \n"
-" if( lIdx == 0 )\n"
-" {\n"
-" ldsCurBatch++;\n"
-" }\n"
-" GROUP_LDS_BARRIER;\n"
-" }\n"
-" \n"
-" \n"
-"}\n"
-"__kernel void solveSingleContactKernel(__global Body* gBodies,\n"
-" __global Shape* gShapes,\n"
-" __global Constraint4* gConstraints,\n"
-" int cellIdx,\n"
-" int batchOffset,\n"
-" int numConstraintsInBatch\n"
-" )\n"
-"{\n"
-" int index = get_global_id(0);\n"
-" if (index < numConstraintsInBatch)\n"
-" {\n"
-" int idx=batchOffset+index;\n"
-" solveContactConstraint( gBodies, gShapes, &gConstraints[idx] );\n"
-" } \n"
-"}\n"
-;
/include/Bullet/Bullet3OpenCL/RigidBody/kernels/solveContact.h
Property changes:
Deleted: svn:executable
## -1 +0,0 ##
-*
\ No newline at end of property
Index: include/Bullet/Bullet3OpenCL/RigidBody/kernels/solveFriction.h
===================================================================
--- include/Bullet/Bullet3OpenCL/RigidBody/kernels/solveFriction.h (revision 22)
+++ include/Bullet/Bullet3OpenCL/RigidBody/kernels/solveFriction.h (nonexistent)
@@ -1,421 +0,0 @@
-//this file is autogenerated using stringify.bat (premake --stringify) in the build folder of this project
-static const char* solveFrictionCL= \
-"/*\n"
-"Copyright (c) 2012 Advanced Micro Devices, Inc. \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"
-"//Originally written by Takahiro Harada\n"
-"//#pragma OPENCL EXTENSION cl_amd_printf : enable\n"
-"#pragma OPENCL EXTENSION cl_khr_local_int32_base_atomics : enable\n"
-"#pragma OPENCL EXTENSION cl_khr_global_int32_base_atomics : enable\n"
-"#pragma OPENCL EXTENSION cl_khr_local_int32_extended_atomics : enable\n"
-"#pragma OPENCL EXTENSION cl_khr_global_int32_extended_atomics : enable\n"
-"#ifdef cl_ext_atomic_counters_32\n"
-"#pragma OPENCL EXTENSION cl_ext_atomic_counters_32 : enable\n"
-"#else\n"
-"#define counter32_t volatile global int*\n"
-"#endif\n"
-"typedef unsigned int u32;\n"
-"typedef unsigned short u16;\n"
-"typedef unsigned char u8;\n"
-"#define GET_GROUP_IDX get_group_id(0)\n"
-"#define GET_LOCAL_IDX get_local_id(0)\n"
-"#define GET_GLOBAL_IDX get_global_id(0)\n"
-"#define GET_GROUP_SIZE get_local_size(0)\n"
-"#define GET_NUM_GROUPS get_num_groups(0)\n"
-"#define GROUP_LDS_BARRIER barrier(CLK_LOCAL_MEM_FENCE)\n"
-"#define GROUP_MEM_FENCE mem_fence(CLK_LOCAL_MEM_FENCE)\n"
-"#define AtomInc(x) atom_inc(&(x))\n"
-"#define AtomInc1(x, out) out = atom_inc(&(x))\n"
-"#define AppendInc(x, out) out = atomic_inc(x)\n"
-"#define AtomAdd(x, value) atom_add(&(x), value)\n"
-"#define AtomCmpxhg(x, cmp, value) atom_cmpxchg( &(x), cmp, value )\n"
-"#define AtomXhg(x, value) atom_xchg ( &(x), value )\n"
-"#define SELECT_UINT4( b, a, condition ) select( b,a,condition )\n"
-"#define mymake_float4 (float4)\n"
-"//#define make_float2 (float2)\n"
-"//#define make_uint4 (uint4)\n"
-"//#define make_int4 (int4)\n"
-"//#define make_uint2 (uint2)\n"
-"//#define make_int2 (int2)\n"
-"#define max2 max\n"
-"#define min2 min\n"
-"///////////////////////////////////////\n"
-"// Vector\n"
-"///////////////////////////////////////\n"
-"__inline\n"
-"float4 fastNormalize4(float4 v)\n"
-"{\n"
-" return fast_normalize(v);\n"
-"}\n"
-"__inline\n"
-"float4 cross3(float4 a, float4 b)\n"
-"{\n"
-" return cross(a,b);\n"
-"}\n"
-"__inline\n"
-"float dot3F4(float4 a, float4 b)\n"
-"{\n"
-" float4 a1 = mymake_float4(a.xyz,0.f);\n"
-" float4 b1 = mymake_float4(b.xyz,0.f);\n"
-" return dot(a1, b1);\n"
-"}\n"
-"__inline\n"
-"float4 normalize3(const float4 a)\n"
-"{\n"
-" float4 n = mymake_float4(a.x, a.y, a.z, 0.f);\n"
-" return fastNormalize4( n );\n"
-"// float length = sqrtf(dot3F4(a, a));\n"
-"// return 1.f/length * a;\n"
-"}\n"
-"///////////////////////////////////////\n"
-"// Matrix3x3\n"
-"///////////////////////////////////////\n"
-"typedef struct\n"
-"{\n"
-" float4 m_row[3];\n"
-"}Matrix3x3;\n"
-"__inline\n"
-"float4 mtMul1(Matrix3x3 a, float4 b);\n"
-"__inline\n"
-"float4 mtMul3(float4 a, Matrix3x3 b);\n"
-"__inline\n"
-"float4 mtMul1(Matrix3x3 a, float4 b)\n"
-"{\n"
-" float4 ans;\n"
-" ans.x = dot3F4( a.m_row[0], b );\n"
-" ans.y = dot3F4( a.m_row[1], b );\n"
-" ans.z = dot3F4( a.m_row[2], b );\n"
-" ans.w = 0.f;\n"
-" return ans;\n"
-"}\n"
-"__inline\n"
-"float4 mtMul3(float4 a, Matrix3x3 b)\n"
-"{\n"
-" float4 colx = mymake_float4(b.m_row[0].x, b.m_row[1].x, b.m_row[2].x, 0);\n"
-" float4 coly = mymake_float4(b.m_row[0].y, b.m_row[1].y, b.m_row[2].y, 0);\n"
-" float4 colz = mymake_float4(b.m_row[0].z, b.m_row[1].z, b.m_row[2].z, 0);\n"
-" float4 ans;\n"
-" ans.x = dot3F4( a, colx );\n"
-" ans.y = dot3F4( a, coly );\n"
-" ans.z = dot3F4( a, colz );\n"
-" return ans;\n"
-"}\n"
-"///////////////////////////////////////\n"
-"// Quaternion\n"
-"///////////////////////////////////////\n"
-"typedef float4 Quaternion;\n"
-"#define WG_SIZE 64\n"
-"typedef struct\n"
-"{\n"
-" float4 m_pos;\n"
-" Quaternion m_quat;\n"
-" float4 m_linVel;\n"
-" float4 m_angVel;\n"
-" u32 m_shapeIdx;\n"
-" float m_invMass;\n"
-" float m_restituitionCoeff;\n"
-" float m_frictionCoeff;\n"
-"} Body;\n"
-"typedef struct\n"
-"{\n"
-" Matrix3x3 m_invInertia;\n"
-" Matrix3x3 m_initInvInertia;\n"
-"} Shape;\n"
-"typedef struct\n"
-"{\n"
-" float4 m_linear;\n"
-" float4 m_worldPos[4];\n"
-" float4 m_center; \n"
-" float m_jacCoeffInv[4];\n"
-" float m_b[4];\n"
-" float m_appliedRambdaDt[4];\n"
-" float m_fJacCoeffInv[2]; \n"
-" float m_fAppliedRambdaDt[2]; \n"
-" u32 m_bodyA;\n"
-" u32 m_bodyB;\n"
-" int m_batchIdx;\n"
-" u32 m_paddings[1];\n"
-"} Constraint4;\n"
-"typedef struct\n"
-"{\n"
-" int m_nConstraints;\n"
-" int m_start;\n"
-" int m_batchIdx;\n"
-" int m_nSplit;\n"
-"// int m_paddings[1];\n"
-"} ConstBuffer;\n"
-"typedef struct\n"
-"{\n"
-" int m_solveFriction;\n"
-" int m_maxBatch; // long batch really kills the performance\n"
-" int m_batchIdx;\n"
-" int m_nSplit;\n"
-"// int m_paddings[1];\n"
-"} ConstBufferBatchSolve;\n"
-"void setLinearAndAngular( float4 n, float4 r0, float4 r1, float4* linear, float4* angular0, float4* angular1);\n"
-"void setLinearAndAngular( float4 n, float4 r0, float4 r1, float4* linear, float4* angular0, float4* angular1)\n"
-"{\n"
-" *linear = mymake_float4(-n.xyz,0.f);\n"
-" *angular0 = -cross3(r0, n);\n"
-" *angular1 = cross3(r1, n);\n"
-"}\n"
-"float calcRelVel( float4 l0, float4 l1, float4 a0, float4 a1, float4 linVel0, float4 angVel0, float4 linVel1, float4 angVel1 );\n"
-"float calcRelVel( float4 l0, float4 l1, float4 a0, float4 a1, float4 linVel0, float4 angVel0, float4 linVel1, float4 angVel1 )\n"
-"{\n"
-" return dot3F4(l0, linVel0) + dot3F4(a0, angVel0) + dot3F4(l1, linVel1) + dot3F4(a1, angVel1);\n"
-"}\n"
-"float calcJacCoeff(const float4 linear0, const float4 linear1, const float4 angular0, const float4 angular1,\n"
-" float invMass0, const Matrix3x3* invInertia0, float invMass1, const Matrix3x3* invInertia1);\n"
-"float calcJacCoeff(const float4 linear0, const float4 linear1, const float4 angular0, const float4 angular1,\n"
-" float invMass0, const Matrix3x3* invInertia0, float invMass1, const Matrix3x3* invInertia1)\n"
-"{\n"
-" // linear0,1 are normlized\n"
-" float jmj0 = invMass0;//dot3F4(linear0, linear0)*invMass0;\n"
-" float jmj1 = dot3F4(mtMul3(angular0,*invInertia0), angular0);\n"
-" float jmj2 = invMass1;//dot3F4(linear1, linear1)*invMass1;\n"
-" float jmj3 = dot3F4(mtMul3(angular1,*invInertia1), angular1);\n"
-" return -1.f/(jmj0+jmj1+jmj2+jmj3);\n"
-"}\n"
-"void btPlaneSpace1 (const float4* n, float4* p, float4* q);\n"
-" void btPlaneSpace1 (const float4* n, float4* p, float4* q)\n"
-"{\n"
-" if (fabs(n[0].z) > 0.70710678f) {\n"
-" // choose p in y-z plane\n"
-" float a = n[0].y*n[0].y + n[0].z*n[0].z;\n"
-" float k = 1.f/sqrt(a);\n"
-" p[0].x = 0;\n"
-" p[0].y = -n[0].z*k;\n"
-" p[0].z = n[0].y*k;\n"
-" // set q = n x p\n"
-" q[0].x = a*k;\n"
-" q[0].y = -n[0].x*p[0].z;\n"
-" q[0].z = n[0].x*p[0].y;\n"
-" }\n"
-" else {\n"
-" // choose p in x-y plane\n"
-" float a = n[0].x*n[0].x + n[0].y*n[0].y;\n"
-" float k = 1.f/sqrt(a);\n"
-" p[0].x = -n[0].y*k;\n"
-" p[0].y = n[0].x*k;\n"
-" p[0].z = 0;\n"
-" // set q = n x p\n"
-" q[0].x = -n[0].z*p[0].y;\n"
-" q[0].y = n[0].z*p[0].x;\n"
-" q[0].z = a*k;\n"
-" }\n"
-"}\n"
-"void solveFrictionConstraint(__global Body* gBodies, __global Shape* gShapes, __global Constraint4* ldsCs);\n"
-"void solveFrictionConstraint(__global Body* gBodies, __global Shape* gShapes, __global Constraint4* ldsCs)\n"
-"{\n"
-" float frictionCoeff = ldsCs[0].m_linear.w;\n"
-" int aIdx = ldsCs[0].m_bodyA;\n"
-" int bIdx = ldsCs[0].m_bodyB;\n"
-" float4 posA = gBodies[aIdx].m_pos;\n"
-" float4 linVelA = gBodies[aIdx].m_linVel;\n"
-" float4 angVelA = gBodies[aIdx].m_angVel;\n"
-" float invMassA = gBodies[aIdx].m_invMass;\n"
-" Matrix3x3 invInertiaA = gShapes[aIdx].m_invInertia;\n"
-" float4 posB = gBodies[bIdx].m_pos;\n"
-" float4 linVelB = gBodies[bIdx].m_linVel;\n"
-" float4 angVelB = gBodies[bIdx].m_angVel;\n"
-" float invMassB = gBodies[bIdx].m_invMass;\n"
-" Matrix3x3 invInertiaB = gShapes[bIdx].m_invInertia;\n"
-" \n"
-" {\n"
-" float maxRambdaDt[4] = {FLT_MAX,FLT_MAX,FLT_MAX,FLT_MAX};\n"
-" float minRambdaDt[4] = {0.f,0.f,0.f,0.f};\n"
-" float sum = 0;\n"
-" for(int j=0; j<4; j++)\n"
-" {\n"
-" sum +=ldsCs[0].m_appliedRambdaDt[j];\n"
-" }\n"
-" frictionCoeff = 0.7f;\n"
-" for(int j=0; j<4; j++)\n"
-" {\n"
-" maxRambdaDt[j] = frictionCoeff*sum;\n"
-" minRambdaDt[j] = -maxRambdaDt[j];\n"
-" }\n"
-" \n"
-"// solveFriction( ldsCs, posA, &linVelA, &angVelA, invMassA, invInertiaA,\n"
-"// posB, &linVelB, &angVelB, invMassB, invInertiaB, maxRambdaDt, minRambdaDt );\n"
-" \n"
-" \n"
-" {\n"
-" \n"
-" __global Constraint4* cs = ldsCs;\n"
-" \n"
-" if( cs->m_fJacCoeffInv[0] == 0 && cs->m_fJacCoeffInv[0] == 0 ) return;\n"
-" const float4 center = cs->m_center;\n"
-" \n"
-" float4 n = -cs->m_linear;\n"
-" \n"
-" float4 tangent[2];\n"
-" btPlaneSpace1(&n,&tangent[0],&tangent[1]);\n"
-" float4 angular0, angular1, linear;\n"
-" float4 r0 = center - posA;\n"
-" float4 r1 = center - posB;\n"
-" for(int i=0; i<2; i++)\n"
-" {\n"
-" setLinearAndAngular( tangent[i], r0, r1, &linear, &angular0, &angular1 );\n"
-" float rambdaDt = calcRelVel(linear, -linear, angular0, angular1,\n"
-" linVelA, angVelA, linVelB, angVelB );\n"
-" rambdaDt *= cs->m_fJacCoeffInv[i];\n"
-" \n"
-" {\n"
-" float prevSum = cs->m_fAppliedRambdaDt[i];\n"
-" float updated = prevSum;\n"
-" updated += rambdaDt;\n"
-" updated = max2( updated, minRambdaDt[i] );\n"
-" updated = min2( updated, maxRambdaDt[i] );\n"
-" rambdaDt = updated - prevSum;\n"
-" cs->m_fAppliedRambdaDt[i] = updated;\n"
-" }\n"
-" \n"
-" float4 linImp0 = invMassA*linear*rambdaDt;\n"
-" float4 linImp1 = invMassB*(-linear)*rambdaDt;\n"
-" float4 angImp0 = mtMul1(invInertiaA, angular0)*rambdaDt;\n"
-" float4 angImp1 = mtMul1(invInertiaB, angular1)*rambdaDt;\n"
-" \n"
-" linVelA += linImp0;\n"
-" angVelA += angImp0;\n"
-" linVelB += linImp1;\n"
-" angVelB += angImp1;\n"
-" }\n"
-" { // angular damping for point constraint\n"
-" float4 ab = normalize3( posB - posA );\n"
-" float4 ac = normalize3( center - posA );\n"
-" if( dot3F4( ab, ac ) > 0.95f || (invMassA == 0.f || invMassB == 0.f))\n"
-" {\n"
-" float angNA = dot3F4( n, angVelA );\n"
-" float angNB = dot3F4( n, angVelB );\n"
-" \n"
-" angVelA -= (angNA*0.1f)*n;\n"
-" angVelB -= (angNB*0.1f)*n;\n"
-" }\n"
-" }\n"
-" }\n"
-" \n"
-" \n"
-" }\n"
-" if (gBodies[aIdx].m_invMass)\n"
-" {\n"
-" gBodies[aIdx].m_linVel = linVelA;\n"
-" gBodies[aIdx].m_angVel = angVelA;\n"
-" } else\n"
-" {\n"
-" gBodies[aIdx].m_linVel = mymake_float4(0,0,0,0);\n"
-" gBodies[aIdx].m_angVel = mymake_float4(0,0,0,0);\n"
-" }\n"
-" if (gBodies[bIdx].m_invMass)\n"
-" {\n"
-" gBodies[bIdx].m_linVel = linVelB;\n"
-" gBodies[bIdx].m_angVel = angVelB;\n"
-" } else\n"
-" {\n"
-" gBodies[bIdx].m_linVel = mymake_float4(0,0,0,0);\n"
-" gBodies[bIdx].m_angVel = mymake_float4(0,0,0,0);\n"
-" }\n"
-" \n"
-"}\n"
-"typedef struct \n"
-"{\n"
-" int m_valInt0;\n"
-" int m_valInt1;\n"
-" int m_valInt2;\n"
-" int m_valInt3;\n"
-" float m_val0;\n"
-" float m_val1;\n"
-" float m_val2;\n"
-" float m_val3;\n"
-"} SolverDebugInfo;\n"
-"__kernel\n"
-"__attribute__((reqd_work_group_size(WG_SIZE,1,1)))\n"
-"void BatchSolveKernelFriction(__global Body* gBodies,\n"
-" __global Shape* gShapes,\n"
-" __global Constraint4* gConstraints,\n"
-" __global int* gN,\n"
-" __global int* gOffsets,\n"
-" __global int* batchSizes,\n"
-" int maxBatch1,\n"
-" int cellBatch,\n"
-" int4 nSplit\n"
-" )\n"
-"{\n"
-" //__local int ldsBatchIdx[WG_SIZE+1];\n"
-" __local int ldsCurBatch;\n"
-" __local int ldsNextBatch;\n"
-" __local int ldsStart;\n"
-" int lIdx = GET_LOCAL_IDX;\n"
-" int wgIdx = GET_GROUP_IDX;\n"
-"// int gIdx = GET_GLOBAL_IDX;\n"
-"// debugInfo[gIdx].m_valInt0 = gIdx;\n"
-" //debugInfo[gIdx].m_valInt1 = GET_GROUP_SIZE;\n"
-" int zIdx = (wgIdx/((nSplit.x*nSplit.y)/4))*2+((cellBatch&4)>>2);\n"
-" int remain= (wgIdx%((nSplit.x*nSplit.y)/4));\n"
-" int yIdx = (remain/(nSplit.x/2))*2 + ((cellBatch&2)>>1);\n"
-" int xIdx = (remain%(nSplit.x/2))*2 + (cellBatch&1);\n"
-" int cellIdx = xIdx+yIdx*nSplit.x+zIdx*(nSplit.x*nSplit.y);\n"
-" \n"
-" if( gN[cellIdx] == 0 ) \n"
-" return;\n"
-" int maxBatch = batchSizes[cellIdx];\n"
-" const int start = gOffsets[cellIdx];\n"
-" const int end = start + gN[cellIdx];\n"
-" \n"
-" if( lIdx == 0 )\n"
-" {\n"
-" ldsCurBatch = 0;\n"
-" ldsNextBatch = 0;\n"
-" ldsStart = start;\n"
-" }\n"
-" GROUP_LDS_BARRIER;\n"
-" int idx=ldsStart+lIdx;\n"
-" while (ldsCurBatch < maxBatch)\n"
-" {\n"
-" for(; idx<end; )\n"
-" {\n"
-" if (gConstraints[idx].m_batchIdx == ldsCurBatch)\n"
-" {\n"
-" solveFrictionConstraint( gBodies, gShapes, &gConstraints[idx] );\n"
-" idx+=64;\n"
-" } else\n"
-" {\n"
-" break;\n"
-" }\n"
-" }\n"
-" GROUP_LDS_BARRIER;\n"
-" if( lIdx == 0 )\n"
-" {\n"
-" ldsCurBatch++;\n"
-" }\n"
-" GROUP_LDS_BARRIER;\n"
-" }\n"
-" \n"
-" \n"
-"}\n"
-"__kernel void solveSingleFrictionKernel(__global Body* gBodies,\n"
-" __global Shape* gShapes,\n"
-" __global Constraint4* gConstraints,\n"
-" int cellIdx,\n"
-" int batchOffset,\n"
-" int numConstraintsInBatch\n"
-" )\n"
-"{\n"
-" int index = get_global_id(0);\n"
-" if (index < numConstraintsInBatch)\n"
-" {\n"
-" \n"
-" int idx=batchOffset+index;\n"
-" \n"
-" solveFrictionConstraint( gBodies, gShapes, &gConstraints[idx] );\n"
-" } \n"
-"}\n"
-;
/include/Bullet/Bullet3OpenCL/RigidBody/kernels/solveFriction.h
Property changes:
Deleted: svn:executable
## -1 +0,0 ##
-*
\ No newline at end of property
Index: include/Bullet/Bullet3OpenCL/RigidBody/kernels/solverUtils.h
===================================================================
--- include/Bullet/Bullet3OpenCL/RigidBody/kernels/solverUtils.h (revision 22)
+++ include/Bullet/Bullet3OpenCL/RigidBody/kernels/solverUtils.h (nonexistent)
@@ -1,908 +0,0 @@
-//this file is autogenerated using stringify.bat (premake --stringify) in the build folder of this project
-static const char* solverUtilsCL= \
-"/*\n"
-"Copyright (c) 2013 Advanced Micro Devices, Inc. \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"
-"//Originally written by Erwin Coumans\n"
-"#ifndef B3_CONTACT4DATA_H\n"
-"#define B3_CONTACT4DATA_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"
-"//keep B3_LARGE_FLOAT*B3_LARGE_FLOAT < FLT_MAX\n"
-"#define B3_LARGE_FLOAT 1e18f\n"
-"#define B3_INFINITY 1e18f\n"
-"#define b3Assert(a)\n"
-"#define b3ConstArray(a) __global const a*\n"
-"#define b3AtomicInc atomic_inc\n"
-"#define b3AtomicAdd atomic_add\n"
-"#define b3Fabs fabs\n"
-"#define b3Sqrt native_sqrt\n"
-"#define b3Sin native_sin\n"
-"#define b3Cos native_cos\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"
-" #define b3MinFloat4 min\n"
-" #define b3MaxFloat4 max\n"
-" #define b3Normalized(a) normalize(a)\n"
-"#endif \n"
-" \n"
-"inline bool b3IsAlmostZero(b3Float4ConstArg v)\n"
-"{\n"
-" if(b3Fabs(v.x)>1e-6 || b3Fabs(v.y)>1e-6 || b3Fabs(v.z)>1e-6) \n"
-" return false;\n"
-" return true;\n"
-"}\n"
-"inline int b3MaxDot( b3Float4ConstArg vec, __global const b3Float4* vecArray, int vecLen, float* dotOut )\n"
-"{\n"
-" float maxDot = -B3_INFINITY;\n"
-" int i = 0;\n"
-" int ptIndex = -1;\n"
-" for( i = 0; i < vecLen; i++ )\n"
-" {\n"
-" float dot = b3Dot3F4(vecArray[i],vec);\n"
-" \n"
-" if( dot > maxDot )\n"
-" {\n"
-" maxDot = dot;\n"
-" ptIndex = i;\n"
-" }\n"
-" }\n"
-" b3Assert(ptIndex>=0);\n"
-" if (ptIndex<0)\n"
-" {\n"
-" ptIndex = 0;\n"
-" }\n"
-" *dotOut = maxDot;\n"
-" return ptIndex;\n"
-"}\n"
-"#endif //B3_FLOAT4_H\n"
-"typedef struct b3Contact4Data b3Contact4Data_t;\n"
-"struct b3Contact4Data\n"
-"{\n"
-" b3Float4 m_worldPosB[4];\n"
-"// b3Float4 m_localPosA[4];\n"
-"// b3Float4 m_localPosB[4];\n"
-" b3Float4 m_worldNormalOnB; // w: m_nPoints\n"
-" unsigned short m_restituitionCoeffCmp;\n"
-" unsigned short m_frictionCoeffCmp;\n"
-" int m_batchIdx;\n"
-" int m_bodyAPtrAndSignBit;//x:m_bodyAPtr, y:m_bodyBPtr\n"
-" int m_bodyBPtrAndSignBit;\n"
-" int m_childIndexA;\n"
-" int m_childIndexB;\n"
-" int m_unused1;\n"
-" int m_unused2;\n"
-"};\n"
-"inline int b3Contact4Data_getNumPoints(const struct b3Contact4Data* contact)\n"
-"{\n"
-" return (int)contact->m_worldNormalOnB.w;\n"
-"};\n"
-"inline void b3Contact4Data_setNumPoints(struct b3Contact4Data* contact, int numPoints)\n"
-"{\n"
-" contact->m_worldNormalOnB.w = (float)numPoints;\n"
-"};\n"
-"#endif //B3_CONTACT4DATA_H\n"
-"#pragma OPENCL EXTENSION cl_amd_printf : enable\n"
-"#pragma OPENCL EXTENSION cl_khr_local_int32_base_atomics : enable\n"
-"#pragma OPENCL EXTENSION cl_khr_global_int32_base_atomics : enable\n"
-"#pragma OPENCL EXTENSION cl_khr_local_int32_extended_atomics : enable\n"
-"#pragma OPENCL EXTENSION cl_khr_global_int32_extended_atomics : enable\n"
-"#ifdef cl_ext_atomic_counters_32\n"
-"#pragma OPENCL EXTENSION cl_ext_atomic_counters_32 : enable\n"
-"#else\n"
-"#define counter32_t volatile global int*\n"
-"#endif\n"
-"typedef unsigned int u32;\n"
-"typedef unsigned short u16;\n"
-"typedef unsigned char u8;\n"
-"#define GET_GROUP_IDX get_group_id(0)\n"
-"#define GET_LOCAL_IDX get_local_id(0)\n"
-"#define GET_GLOBAL_IDX get_global_id(0)\n"
-"#define GET_GROUP_SIZE get_local_size(0)\n"
-"#define GET_NUM_GROUPS get_num_groups(0)\n"
-"#define GROUP_LDS_BARRIER barrier(CLK_LOCAL_MEM_FENCE)\n"
-"#define GROUP_MEM_FENCE mem_fence(CLK_LOCAL_MEM_FENCE)\n"
-"#define AtomInc(x) atom_inc(&(x))\n"
-"#define AtomInc1(x, out) out = atom_inc(&(x))\n"
-"#define AppendInc(x, out) out = atomic_inc(x)\n"
-"#define AtomAdd(x, value) atom_add(&(x), value)\n"
-"#define AtomCmpxhg(x, cmp, value) atom_cmpxchg( &(x), cmp, value )\n"
-"#define AtomXhg(x, value) atom_xchg ( &(x), value )\n"
-"#define SELECT_UINT4( b, a, condition ) select( b,a,condition )\n"
-"#define make_float4 (float4)\n"
-"#define make_float2 (float2)\n"
-"#define make_uint4 (uint4)\n"
-"#define make_int4 (int4)\n"
-"#define make_uint2 (uint2)\n"
-"#define make_int2 (int2)\n"
-"#define max2 max\n"
-"#define min2 min\n"
-"///////////////////////////////////////\n"
-"// Vector\n"
-"///////////////////////////////////////\n"
-"__inline\n"
-"float fastDiv(float numerator, float denominator)\n"
-"{\n"
-" return native_divide(numerator, denominator); \n"
-"// return numerator/denominator; \n"
-"}\n"
-"__inline\n"
-"float4 fastDiv4(float4 numerator, float4 denominator)\n"
-"{\n"
-" return native_divide(numerator, denominator); \n"
-"}\n"
-"__inline\n"
-"float fastSqrtf(float f2)\n"
-"{\n"
-" return native_sqrt(f2);\n"
-"// return sqrt(f2);\n"
-"}\n"
-"__inline\n"
-"float fastRSqrt(float f2)\n"
-"{\n"
-" return native_rsqrt(f2);\n"
-"}\n"
-"__inline\n"
-"float fastLength4(float4 v)\n"
-"{\n"
-" return fast_length(v);\n"
-"}\n"
-"__inline\n"
-"float4 fastNormalize4(float4 v)\n"
-"{\n"
-" return fast_normalize(v);\n"
-"}\n"
-"__inline\n"
-"float sqrtf(float a)\n"
-"{\n"
-"// return sqrt(a);\n"
-" return native_sqrt(a);\n"
-"}\n"
-"__inline\n"
-"float4 cross3(float4 a1, float4 b1)\n"
-"{\n"
-" float4 a=make_float4(a1.xyz,0.f);\n"
-" float4 b=make_float4(b1.xyz,0.f);\n"
-" //float4 a=a1;\n"
-" //float4 b=b1;\n"
-" return cross(a,b);\n"
-"}\n"
-"__inline\n"
-"float dot3F4(float4 a, float4 b)\n"
-"{\n"
-" float4 a1 = make_float4(a.xyz,0.f);\n"
-" float4 b1 = make_float4(b.xyz,0.f);\n"
-" return dot(a1, b1);\n"
-"}\n"
-"__inline\n"
-"float length3(const float4 a)\n"
-"{\n"
-" return sqrtf(dot3F4(a,a));\n"
-"}\n"
-"__inline\n"
-"float dot4(const float4 a, const float4 b)\n"
-"{\n"
-" return dot( a, b );\n"
-"}\n"
-"// for height\n"
-"__inline\n"
-"float dot3w1(const float4 point, const float4 eqn)\n"
-"{\n"
-" return dot3F4(point,eqn) + eqn.w;\n"
-"}\n"
-"__inline\n"
-"float4 normalize3(const float4 a)\n"
-"{\n"
-" float4 n = make_float4(a.x, a.y, a.z, 0.f);\n"
-" return fastNormalize4( n );\n"
-"// float length = sqrtf(dot3F4(a, a));\n"
-"// return 1.f/length * a;\n"
-"}\n"
-"__inline\n"
-"float4 normalize4(const float4 a)\n"
-"{\n"
-" float length = sqrtf(dot4(a, a));\n"
-" return 1.f/length * a;\n"
-"}\n"
-"__inline\n"
-"float4 createEquation(const float4 a, const float4 b, const float4 c)\n"
-"{\n"
-" float4 eqn;\n"
-" float4 ab = b-a;\n"
-" float4 ac = c-a;\n"
-" eqn = normalize3( cross3(ab, ac) );\n"
-" eqn.w = -dot3F4(eqn,a);\n"
-" return eqn;\n"
-"}\n"
-"///////////////////////////////////////\n"
-"// Matrix3x3\n"
-"///////////////////////////////////////\n"
-"typedef struct\n"
-"{\n"
-" float4 m_row[3];\n"
-"}Matrix3x3;\n"
-"__inline\n"
-"Matrix3x3 mtZero();\n"
-"__inline\n"
-"Matrix3x3 mtIdentity();\n"
-"__inline\n"
-"Matrix3x3 mtTranspose(Matrix3x3 m);\n"
-"__inline\n"
-"Matrix3x3 mtMul(Matrix3x3 a, Matrix3x3 b);\n"
-"__inline\n"
-"float4 mtMul1(Matrix3x3 a, float4 b);\n"
-"__inline\n"
-"float4 mtMul3(float4 a, Matrix3x3 b);\n"
-"__inline\n"
-"Matrix3x3 mtZero()\n"
-"{\n"
-" Matrix3x3 m;\n"
-" m.m_row[0] = (float4)(0.f);\n"
-" m.m_row[1] = (float4)(0.f);\n"
-" m.m_row[2] = (float4)(0.f);\n"
-" return m;\n"
-"}\n"
-"__inline\n"
-"Matrix3x3 mtIdentity()\n"
-"{\n"
-" Matrix3x3 m;\n"
-" m.m_row[0] = (float4)(1,0,0,0);\n"
-" m.m_row[1] = (float4)(0,1,0,0);\n"
-" m.m_row[2] = (float4)(0,0,1,0);\n"
-" return m;\n"
-"}\n"
-"__inline\n"
-"Matrix3x3 mtTranspose(Matrix3x3 m)\n"
-"{\n"
-" Matrix3x3 out;\n"
-" out.m_row[0] = (float4)(m.m_row[0].x, m.m_row[1].x, m.m_row[2].x, 0.f);\n"
-" out.m_row[1] = (float4)(m.m_row[0].y, m.m_row[1].y, m.m_row[2].y, 0.f);\n"
-" out.m_row[2] = (float4)(m.m_row[0].z, m.m_row[1].z, m.m_row[2].z, 0.f);\n"
-" return out;\n"
-"}\n"
-"__inline\n"
-"Matrix3x3 mtMul(Matrix3x3 a, Matrix3x3 b)\n"
-"{\n"
-" Matrix3x3 transB;\n"
-" transB = mtTranspose( b );\n"
-" Matrix3x3 ans;\n"
-" // why this doesn't run when 0ing in the for{}\n"
-" a.m_row[0].w = 0.f;\n"
-" a.m_row[1].w = 0.f;\n"
-" a.m_row[2].w = 0.f;\n"
-" for(int i=0; i<3; i++)\n"
-" {\n"
-"// a.m_row[i].w = 0.f;\n"
-" ans.m_row[i].x = dot3F4(a.m_row[i],transB.m_row[0]);\n"
-" ans.m_row[i].y = dot3F4(a.m_row[i],transB.m_row[1]);\n"
-" ans.m_row[i].z = dot3F4(a.m_row[i],transB.m_row[2]);\n"
-" ans.m_row[i].w = 0.f;\n"
-" }\n"
-" return ans;\n"
-"}\n"
-"__inline\n"
-"float4 mtMul1(Matrix3x3 a, float4 b)\n"
-"{\n"
-" float4 ans;\n"
-" ans.x = dot3F4( a.m_row[0], b );\n"
-" ans.y = dot3F4( a.m_row[1], b );\n"
-" ans.z = dot3F4( a.m_row[2], b );\n"
-" ans.w = 0.f;\n"
-" return ans;\n"
-"}\n"
-"__inline\n"
-"float4 mtMul3(float4 a, Matrix3x3 b)\n"
-"{\n"
-" float4 colx = make_float4(b.m_row[0].x, b.m_row[1].x, b.m_row[2].x, 0);\n"
-" float4 coly = make_float4(b.m_row[0].y, b.m_row[1].y, b.m_row[2].y, 0);\n"
-" float4 colz = make_float4(b.m_row[0].z, b.m_row[1].z, b.m_row[2].z, 0);\n"
-" float4 ans;\n"
-" ans.x = dot3F4( a, colx );\n"
-" ans.y = dot3F4( a, coly );\n"
-" ans.z = dot3F4( a, colz );\n"
-" return ans;\n"
-"}\n"
-"///////////////////////////////////////\n"
-"// Quaternion\n"
-"///////////////////////////////////////\n"
-"typedef float4 Quaternion;\n"
-"__inline\n"
-"Quaternion qtMul(Quaternion a, Quaternion b);\n"
-"__inline\n"
-"Quaternion qtNormalize(Quaternion in);\n"
-"__inline\n"
-"float4 qtRotate(Quaternion q, float4 vec);\n"
-"__inline\n"
-"Quaternion qtInvert(Quaternion q);\n"
-"__inline\n"
-"Quaternion qtMul(Quaternion a, Quaternion b)\n"
-"{\n"
-" Quaternion ans;\n"
-" ans = cross3( 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 - dot3F4(a, b);\n"
-" return ans;\n"
-"}\n"
-"__inline\n"
-"Quaternion qtNormalize(Quaternion in)\n"
-"{\n"
-" return fastNormalize4(in);\n"
-"// in /= length( in );\n"
-"// return in;\n"
-"}\n"
-"__inline\n"
-"float4 qtRotate(Quaternion q, float4 vec)\n"
-"{\n"
-" Quaternion qInv = qtInvert( q );\n"
-" float4 vcpy = vec;\n"
-" vcpy.w = 0.f;\n"
-" float4 out = qtMul(qtMul(q,vcpy),qInv);\n"
-" return out;\n"
-"}\n"
-"__inline\n"
-"Quaternion qtInvert(Quaternion q)\n"
-"{\n"
-" return (Quaternion)(-q.xyz, q.w);\n"
-"}\n"
-"__inline\n"
-"float4 qtInvRotate(const Quaternion q, float4 vec)\n"
-"{\n"
-" return qtRotate( qtInvert( q ), vec );\n"
-"}\n"
-"#define WG_SIZE 64\n"
-"typedef struct\n"
-"{\n"
-" float4 m_pos;\n"
-" Quaternion m_quat;\n"
-" float4 m_linVel;\n"
-" float4 m_angVel;\n"
-" u32 m_shapeIdx;\n"
-" float m_invMass;\n"
-" float m_restituitionCoeff;\n"
-" float m_frictionCoeff;\n"
-"} Body;\n"
-"typedef struct\n"
-"{\n"
-" Matrix3x3 m_invInertia;\n"
-" Matrix3x3 m_initInvInertia;\n"
-"} Shape;\n"
-"typedef struct\n"
-"{\n"
-" float4 m_linear;\n"
-" float4 m_worldPos[4];\n"
-" float4 m_center; \n"
-" float m_jacCoeffInv[4];\n"
-" float m_b[4];\n"
-" float m_appliedRambdaDt[4];\n"
-" float m_fJacCoeffInv[2]; \n"
-" float m_fAppliedRambdaDt[2]; \n"
-" u32 m_bodyA;\n"
-" u32 m_bodyB;\n"
-" int m_batchIdx;\n"
-" u32 m_paddings;\n"
-"} Constraint4;\n"
-"__kernel void CountBodiesKernel(__global struct b3Contact4Data* manifoldPtr, __global unsigned int* bodyCount, __global int2* contactConstraintOffsets, int numContactManifolds, int fixedBodyIndex)\n"
-"{\n"
-" int i = GET_GLOBAL_IDX;\n"
-" \n"
-" if( i < numContactManifolds)\n"
-" {\n"
-" int pa = manifoldPtr[i].m_bodyAPtrAndSignBit;\n"
-" bool isFixedA = (pa <0) || (pa == fixedBodyIndex);\n"
-" int bodyIndexA = abs(pa);\n"
-" if (!isFixedA)\n"
-" {\n"
-" AtomInc1(bodyCount[bodyIndexA],contactConstraintOffsets[i].x);\n"
-" }\n"
-" barrier(CLK_GLOBAL_MEM_FENCE);\n"
-" int pb = manifoldPtr[i].m_bodyBPtrAndSignBit;\n"
-" bool isFixedB = (pb <0) || (pb == fixedBodyIndex);\n"
-" int bodyIndexB = abs(pb);\n"
-" if (!isFixedB)\n"
-" {\n"
-" AtomInc1(bodyCount[bodyIndexB],contactConstraintOffsets[i].y);\n"
-" } \n"
-" }\n"
-"}\n"
-"__kernel void ClearVelocitiesKernel(__global float4* linearVelocities,__global float4* angularVelocities, int numSplitBodies)\n"
-"{\n"
-" int i = GET_GLOBAL_IDX;\n"
-" \n"
-" if( i < numSplitBodies)\n"
-" {\n"
-" linearVelocities[i] = make_float4(0);\n"
-" angularVelocities[i] = make_float4(0);\n"
-" }\n"
-"}\n"
-"__kernel void AverageVelocitiesKernel(__global Body* gBodies,__global int* offsetSplitBodies,__global const unsigned int* bodyCount,\n"
-"__global float4* deltaLinearVelocities, __global float4* deltaAngularVelocities, int numBodies)\n"
-"{\n"
-" int i = GET_GLOBAL_IDX;\n"
-" if (i<numBodies)\n"
-" {\n"
-" if (gBodies[i].m_invMass)\n"
-" {\n"
-" int bodyOffset = offsetSplitBodies[i];\n"
-" int count = bodyCount[i];\n"
-" float factor = 1.f/((float)count);\n"
-" float4 averageLinVel = make_float4(0.f);\n"
-" float4 averageAngVel = make_float4(0.f);\n"
-" \n"
-" for (int j=0;j<count;j++)\n"
-" {\n"
-" averageLinVel += deltaLinearVelocities[bodyOffset+j]*factor;\n"
-" averageAngVel += deltaAngularVelocities[bodyOffset+j]*factor;\n"
-" }\n"
-" \n"
-" for (int j=0;j<count;j++)\n"
-" {\n"
-" deltaLinearVelocities[bodyOffset+j] = averageLinVel;\n"
-" deltaAngularVelocities[bodyOffset+j] = averageAngVel;\n"
-" }\n"
-" \n"
-" }//bodies[i].m_invMass\n"
-" }//i<numBodies\n"
-"}\n"
-"void setLinearAndAngular( float4 n, float4 r0, float4 r1, float4* linear, float4* angular0, float4* angular1)\n"
-"{\n"
-" *linear = make_float4(n.xyz,0.f);\n"
-" *angular0 = cross3(r0, n);\n"
-" *angular1 = -cross3(r1, n);\n"
-"}\n"
-"float calcRelVel( float4 l0, float4 l1, float4 a0, float4 a1, float4 linVel0, float4 angVel0, float4 linVel1, float4 angVel1 )\n"
-"{\n"
-" return dot3F4(l0, linVel0) + dot3F4(a0, angVel0) + dot3F4(l1, linVel1) + dot3F4(a1, angVel1);\n"
-"}\n"
-"float calcJacCoeff(const float4 linear0, const float4 linear1, const float4 angular0, const float4 angular1,\n"
-" float invMass0, const Matrix3x3* invInertia0, float invMass1, const Matrix3x3* invInertia1, float countA, float countB)\n"
-"{\n"
-" // linear0,1 are normlized\n"
-" float jmj0 = invMass0;//dot3F4(linear0, linear0)*invMass0;\n"
-" float jmj1 = dot3F4(mtMul3(angular0,*invInertia0), angular0);\n"
-" float jmj2 = invMass1;//dot3F4(linear1, linear1)*invMass1;\n"
-" float jmj3 = dot3F4(mtMul3(angular1,*invInertia1), angular1);\n"
-" return -1.f/((jmj0+jmj1)*countA+(jmj2+jmj3)*countB);\n"
-"}\n"
-"void btPlaneSpace1 (float4 n, float4* p, float4* q);\n"
-" void btPlaneSpace1 (float4 n, float4* p, float4* q)\n"
-"{\n"
-" if (fabs(n.z) > 0.70710678f) {\n"
-" // choose p in y-z plane\n"
-" float a = n.y*n.y + n.z*n.z;\n"
-" float k = 1.f/sqrt(a);\n"
-" p[0].x = 0;\n"
-" p[0].y = -n.z*k;\n"
-" p[0].z = n.y*k;\n"
-" // set q = n x p\n"
-" q[0].x = a*k;\n"
-" q[0].y = -n.x*p[0].z;\n"
-" q[0].z = n.x*p[0].y;\n"
-" }\n"
-" else {\n"
-" // choose p in x-y plane\n"
-" float a = n.x*n.x + n.y*n.y;\n"
-" float k = 1.f/sqrt(a);\n"
-" p[0].x = -n.y*k;\n"
-" p[0].y = n.x*k;\n"
-" p[0].z = 0;\n"
-" // set q = n x p\n"
-" q[0].x = -n.z*p[0].y;\n"
-" q[0].y = n.z*p[0].x;\n"
-" q[0].z = a*k;\n"
-" }\n"
-"}\n"
-"void solveContact(__global Constraint4* cs,\n"
-" float4 posA, float4* linVelA, float4* angVelA, float invMassA, Matrix3x3 invInertiaA,\n"
-" float4 posB, float4* linVelB, float4* angVelB, float invMassB, Matrix3x3 invInertiaB,\n"
-" float4* dLinVelA, float4* dAngVelA, float4* dLinVelB, float4* dAngVelB)\n"
-"{\n"
-" float minRambdaDt = 0;\n"
-" float maxRambdaDt = FLT_MAX;\n"
-" for(int ic=0; ic<4; ic++)\n"
-" {\n"
-" if( cs->m_jacCoeffInv[ic] == 0.f ) continue;\n"
-" float4 angular0, angular1, linear;\n"
-" float4 r0 = cs->m_worldPos[ic] - posA;\n"
-" float4 r1 = cs->m_worldPos[ic] - posB;\n"
-" setLinearAndAngular( cs->m_linear, r0, r1, &linear, &angular0, &angular1 );\n"
-" \n"
-" float rambdaDt = calcRelVel( cs->m_linear, -cs->m_linear, angular0, angular1, \n"
-" *linVelA+*dLinVelA, *angVelA+*dAngVelA, *linVelB+*dLinVelB, *angVelB+*dAngVelB ) + cs->m_b[ic];\n"
-" rambdaDt *= cs->m_jacCoeffInv[ic];\n"
-" \n"
-" {\n"
-" float prevSum = cs->m_appliedRambdaDt[ic];\n"
-" float updated = prevSum;\n"
-" updated += rambdaDt;\n"
-" updated = max2( updated, minRambdaDt );\n"
-" updated = min2( updated, maxRambdaDt );\n"
-" rambdaDt = updated - prevSum;\n"
-" cs->m_appliedRambdaDt[ic] = updated;\n"
-" }\n"
-" \n"
-" float4 linImp0 = invMassA*linear*rambdaDt;\n"
-" float4 linImp1 = invMassB*(-linear)*rambdaDt;\n"
-" float4 angImp0 = mtMul1(invInertiaA, angular0)*rambdaDt;\n"
-" float4 angImp1 = mtMul1(invInertiaB, angular1)*rambdaDt;\n"
-" \n"
-" if (invMassA)\n"
-" {\n"
-" *dLinVelA += linImp0;\n"
-" *dAngVelA += angImp0;\n"
-" }\n"
-" if (invMassB)\n"
-" {\n"
-" *dLinVelB += linImp1;\n"
-" *dAngVelB += angImp1;\n"
-" }\n"
-" }\n"
-"}\n"
-"// solveContactConstraint( gBodies, gShapes, &gConstraints[i] ,contactConstraintOffsets,offsetSplitBodies, deltaLinearVelocities, deltaAngularVelocities);\n"
-"void solveContactConstraint(__global Body* gBodies, __global Shape* gShapes, __global Constraint4* ldsCs, \n"
-"__global int2* contactConstraintOffsets,__global unsigned int* offsetSplitBodies,\n"
-"__global float4* deltaLinearVelocities, __global float4* deltaAngularVelocities)\n"
-"{\n"
-" //float frictionCoeff = ldsCs[0].m_linear.w;\n"
-" int aIdx = ldsCs[0].m_bodyA;\n"
-" int bIdx = ldsCs[0].m_bodyB;\n"
-" float4 posA = gBodies[aIdx].m_pos;\n"
-" float4 linVelA = gBodies[aIdx].m_linVel;\n"
-" float4 angVelA = gBodies[aIdx].m_angVel;\n"
-" float invMassA = gBodies[aIdx].m_invMass;\n"
-" Matrix3x3 invInertiaA = gShapes[aIdx].m_invInertia;\n"
-" float4 posB = gBodies[bIdx].m_pos;\n"
-" float4 linVelB = gBodies[bIdx].m_linVel;\n"
-" float4 angVelB = gBodies[bIdx].m_angVel;\n"
-" float invMassB = gBodies[bIdx].m_invMass;\n"
-" Matrix3x3 invInertiaB = gShapes[bIdx].m_invInertia;\n"
-" \n"
-" float4 dLinVelA = make_float4(0,0,0,0);\n"
-" float4 dAngVelA = make_float4(0,0,0,0);\n"
-" float4 dLinVelB = make_float4(0,0,0,0);\n"
-" float4 dAngVelB = make_float4(0,0,0,0);\n"
-" \n"
-" int bodyOffsetA = offsetSplitBodies[aIdx];\n"
-" int constraintOffsetA = contactConstraintOffsets[0].x;\n"
-" int splitIndexA = bodyOffsetA+constraintOffsetA;\n"
-" \n"
-" if (invMassA)\n"
-" {\n"
-" dLinVelA = deltaLinearVelocities[splitIndexA];\n"
-" dAngVelA = deltaAngularVelocities[splitIndexA];\n"
-" }\n"
-" int bodyOffsetB = offsetSplitBodies[bIdx];\n"
-" int constraintOffsetB = contactConstraintOffsets[0].y;\n"
-" int splitIndexB= bodyOffsetB+constraintOffsetB;\n"
-" if (invMassB)\n"
-" {\n"
-" dLinVelB = deltaLinearVelocities[splitIndexB];\n"
-" dAngVelB = deltaAngularVelocities[splitIndexB];\n"
-" }\n"
-" solveContact( ldsCs, posA, &linVelA, &angVelA, invMassA, invInertiaA,\n"
-" posB, &linVelB, &angVelB, invMassB, invInertiaB ,&dLinVelA, &dAngVelA, &dLinVelB, &dAngVelB);\n"
-" if (invMassA)\n"
-" {\n"
-" deltaLinearVelocities[splitIndexA] = dLinVelA;\n"
-" deltaAngularVelocities[splitIndexA] = dAngVelA;\n"
-" } \n"
-" if (invMassB)\n"
-" {\n"
-" deltaLinearVelocities[splitIndexB] = dLinVelB;\n"
-" deltaAngularVelocities[splitIndexB] = dAngVelB;\n"
-" }\n"
-"}\n"
-"__kernel void SolveContactJacobiKernel(__global Constraint4* gConstraints, __global Body* gBodies, __global Shape* gShapes ,\n"
-"__global int2* contactConstraintOffsets,__global unsigned int* offsetSplitBodies,__global float4* deltaLinearVelocities, __global float4* deltaAngularVelocities,\n"
-"float deltaTime, float positionDrift, float positionConstraintCoeff, int fixedBodyIndex, int numManifolds\n"
-")\n"
-"{\n"
-" int i = GET_GLOBAL_IDX;\n"
-" if (i<numManifolds)\n"
-" {\n"
-" solveContactConstraint( gBodies, gShapes, &gConstraints[i] ,&contactConstraintOffsets[i],offsetSplitBodies, deltaLinearVelocities, deltaAngularVelocities);\n"
-" }\n"
-"}\n"
-"void solveFrictionConstraint(__global Body* gBodies, __global Shape* gShapes, __global Constraint4* ldsCs,\n"
-" __global int2* contactConstraintOffsets,__global unsigned int* offsetSplitBodies,\n"
-" __global float4* deltaLinearVelocities, __global float4* deltaAngularVelocities)\n"
-"{\n"
-" float frictionCoeff = 0.7f;//ldsCs[0].m_linear.w;\n"
-" int aIdx = ldsCs[0].m_bodyA;\n"
-" int bIdx = ldsCs[0].m_bodyB;\n"
-" float4 posA = gBodies[aIdx].m_pos;\n"
-" float4 linVelA = gBodies[aIdx].m_linVel;\n"
-" float4 angVelA = gBodies[aIdx].m_angVel;\n"
-" float invMassA = gBodies[aIdx].m_invMass;\n"
-" Matrix3x3 invInertiaA = gShapes[aIdx].m_invInertia;\n"
-" float4 posB = gBodies[bIdx].m_pos;\n"
-" float4 linVelB = gBodies[bIdx].m_linVel;\n"
-" float4 angVelB = gBodies[bIdx].m_angVel;\n"
-" float invMassB = gBodies[bIdx].m_invMass;\n"
-" Matrix3x3 invInertiaB = gShapes[bIdx].m_invInertia;\n"
-" \n"
-" float4 dLinVelA = make_float4(0,0,0,0);\n"
-" float4 dAngVelA = make_float4(0,0,0,0);\n"
-" float4 dLinVelB = make_float4(0,0,0,0);\n"
-" float4 dAngVelB = make_float4(0,0,0,0);\n"
-" \n"
-" int bodyOffsetA = offsetSplitBodies[aIdx];\n"
-" int constraintOffsetA = contactConstraintOffsets[0].x;\n"
-" int splitIndexA = bodyOffsetA+constraintOffsetA;\n"
-" \n"
-" if (invMassA)\n"
-" {\n"
-" dLinVelA = deltaLinearVelocities[splitIndexA];\n"
-" dAngVelA = deltaAngularVelocities[splitIndexA];\n"
-" }\n"
-" int bodyOffsetB = offsetSplitBodies[bIdx];\n"
-" int constraintOffsetB = contactConstraintOffsets[0].y;\n"
-" int splitIndexB= bodyOffsetB+constraintOffsetB;\n"
-" if (invMassB)\n"
-" {\n"
-" dLinVelB = deltaLinearVelocities[splitIndexB];\n"
-" dAngVelB = deltaAngularVelocities[splitIndexB];\n"
-" }\n"
-" {\n"
-" float maxRambdaDt[4] = {FLT_MAX,FLT_MAX,FLT_MAX,FLT_MAX};\n"
-" float minRambdaDt[4] = {0.f,0.f,0.f,0.f};\n"
-" float sum = 0;\n"
-" for(int j=0; j<4; j++)\n"
-" {\n"
-" sum +=ldsCs[0].m_appliedRambdaDt[j];\n"
-" }\n"
-" frictionCoeff = 0.7f;\n"
-" for(int j=0; j<4; j++)\n"
-" {\n"
-" maxRambdaDt[j] = frictionCoeff*sum;\n"
-" minRambdaDt[j] = -maxRambdaDt[j];\n"
-" }\n"
-" \n"
-"// solveFriction( ldsCs, posA, &linVelA, &angVelA, invMassA, invInertiaA,\n"
-"// posB, &linVelB, &angVelB, invMassB, invInertiaB, maxRambdaDt, minRambdaDt );\n"
-" \n"
-" \n"
-" {\n"
-" \n"
-" __global Constraint4* cs = ldsCs;\n"
-" \n"
-" if( cs->m_fJacCoeffInv[0] == 0 && cs->m_fJacCoeffInv[0] == 0 ) return;\n"
-" const float4 center = cs->m_center;\n"
-" \n"
-" float4 n = -cs->m_linear;\n"
-" \n"
-" float4 tangent[2];\n"
-" btPlaneSpace1(n,&tangent[0],&tangent[1]);\n"
-" float4 angular0, angular1, linear;\n"
-" float4 r0 = center - posA;\n"
-" float4 r1 = center - posB;\n"
-" for(int i=0; i<2; i++)\n"
-" {\n"
-" setLinearAndAngular( tangent[i], r0, r1, &linear, &angular0, &angular1 );\n"
-" float rambdaDt = calcRelVel(linear, -linear, angular0, angular1,\n"
-" linVelA+dLinVelA, angVelA+dAngVelA, linVelB+dLinVelB, angVelB+dAngVelB );\n"
-" rambdaDt *= cs->m_fJacCoeffInv[i];\n"
-" \n"
-" {\n"
-" float prevSum = cs->m_fAppliedRambdaDt[i];\n"
-" float updated = prevSum;\n"
-" updated += rambdaDt;\n"
-" updated = max2( updated, minRambdaDt[i] );\n"
-" updated = min2( updated, maxRambdaDt[i] );\n"
-" rambdaDt = updated - prevSum;\n"
-" cs->m_fAppliedRambdaDt[i] = updated;\n"
-" }\n"
-" \n"
-" float4 linImp0 = invMassA*linear*rambdaDt;\n"
-" float4 linImp1 = invMassB*(-linear)*rambdaDt;\n"
-" float4 angImp0 = mtMul1(invInertiaA, angular0)*rambdaDt;\n"
-" float4 angImp1 = mtMul1(invInertiaB, angular1)*rambdaDt;\n"
-" \n"
-" dLinVelA += linImp0;\n"
-" dAngVelA += angImp0;\n"
-" dLinVelB += linImp1;\n"
-" dAngVelB += angImp1;\n"
-" }\n"
-" { // angular damping for point constraint\n"
-" float4 ab = normalize3( posB - posA );\n"
-" float4 ac = normalize3( center - posA );\n"
-" if( dot3F4( ab, ac ) > 0.95f || (invMassA == 0.f || invMassB == 0.f))\n"
-" {\n"
-" float angNA = dot3F4( n, angVelA );\n"
-" float angNB = dot3F4( n, angVelB );\n"
-" \n"
-" dAngVelA -= (angNA*0.1f)*n;\n"
-" dAngVelB -= (angNB*0.1f)*n;\n"
-" }\n"
-" }\n"
-" }\n"
-" \n"
-" \n"
-" }\n"
-" if (invMassA)\n"
-" {\n"
-" deltaLinearVelocities[splitIndexA] = dLinVelA;\n"
-" deltaAngularVelocities[splitIndexA] = dAngVelA;\n"
-" } \n"
-" if (invMassB)\n"
-" {\n"
-" deltaLinearVelocities[splitIndexB] = dLinVelB;\n"
-" deltaAngularVelocities[splitIndexB] = dAngVelB;\n"
-" }\n"
-" \n"
-"}\n"
-"__kernel void SolveFrictionJacobiKernel(__global Constraint4* gConstraints, __global Body* gBodies, __global Shape* gShapes ,\n"
-" __global int2* contactConstraintOffsets,__global unsigned int* offsetSplitBodies,\n"
-" __global float4* deltaLinearVelocities, __global float4* deltaAngularVelocities,\n"
-" float deltaTime, float positionDrift, float positionConstraintCoeff, int fixedBodyIndex, int numManifolds\n"
-")\n"
-"{\n"
-" int i = GET_GLOBAL_IDX;\n"
-" if (i<numManifolds)\n"
-" {\n"
-" solveFrictionConstraint( gBodies, gShapes, &gConstraints[i] ,&contactConstraintOffsets[i],offsetSplitBodies, deltaLinearVelocities, deltaAngularVelocities);\n"
-" }\n"
-"}\n"
-"__kernel void UpdateBodyVelocitiesKernel(__global Body* gBodies,__global int* offsetSplitBodies,__global const unsigned int* bodyCount,\n"
-" __global float4* deltaLinearVelocities, __global float4* deltaAngularVelocities, int numBodies)\n"
-"{\n"
-" int i = GET_GLOBAL_IDX;\n"
-" if (i<numBodies)\n"
-" {\n"
-" if (gBodies[i].m_invMass)\n"
-" {\n"
-" int bodyOffset = offsetSplitBodies[i];\n"
-" int count = bodyCount[i];\n"
-" if (count)\n"
-" {\n"
-" gBodies[i].m_linVel += deltaLinearVelocities[bodyOffset];\n"
-" gBodies[i].m_angVel += deltaAngularVelocities[bodyOffset];\n"
-" }\n"
-" }\n"
-" }\n"
-"}\n"
-"void setConstraint4( const float4 posA, const float4 linVelA, const float4 angVelA, float invMassA, const Matrix3x3 invInertiaA,\n"
-" const float4 posB, const float4 linVelB, const float4 angVelB, float invMassB, const Matrix3x3 invInertiaB, \n"
-" __global struct b3Contact4Data* src, float dt, float positionDrift, float positionConstraintCoeff,float countA, float countB,\n"
-" Constraint4* dstC )\n"
-"{\n"
-" dstC->m_bodyA = abs(src->m_bodyAPtrAndSignBit);\n"
-" dstC->m_bodyB = abs(src->m_bodyBPtrAndSignBit);\n"
-" float dtInv = 1.f/dt;\n"
-" for(int ic=0; ic<4; ic++)\n"
-" {\n"
-" dstC->m_appliedRambdaDt[ic] = 0.f;\n"
-" }\n"
-" dstC->m_fJacCoeffInv[0] = dstC->m_fJacCoeffInv[1] = 0.f;\n"
-" dstC->m_linear = src->m_worldNormalOnB;\n"
-" dstC->m_linear.w = 0.7f ;//src->getFrictionCoeff() );\n"
-" for(int ic=0; ic<4; ic++)\n"
-" {\n"
-" float4 r0 = src->m_worldPosB[ic] - posA;\n"
-" float4 r1 = src->m_worldPosB[ic] - posB;\n"
-" if( ic >= src->m_worldNormalOnB.w )//npoints\n"
-" {\n"
-" dstC->m_jacCoeffInv[ic] = 0.f;\n"
-" continue;\n"
-" }\n"
-" float relVelN;\n"
-" {\n"
-" float4 linear, angular0, angular1;\n"
-" setLinearAndAngular(src->m_worldNormalOnB, r0, r1, &linear, &angular0, &angular1);\n"
-" dstC->m_jacCoeffInv[ic] = calcJacCoeff(linear, -linear, angular0, angular1,\n"
-" invMassA, &invInertiaA, invMassB, &invInertiaB , countA, countB);\n"
-" relVelN = calcRelVel(linear, -linear, angular0, angular1,\n"
-" linVelA, angVelA, linVelB, angVelB);\n"
-" float e = 0.f;//src->getRestituitionCoeff();\n"
-" if( relVelN*relVelN < 0.004f ) e = 0.f;\n"
-" dstC->m_b[ic] = e*relVelN;\n"
-" //float penetration = src->m_worldPosB[ic].w;\n"
-" dstC->m_b[ic] += (src->m_worldPosB[ic].w + positionDrift)*positionConstraintCoeff*dtInv;\n"
-" dstC->m_appliedRambdaDt[ic] = 0.f;\n"
-" }\n"
-" }\n"
-" if( src->m_worldNormalOnB.w > 0 )//npoints\n"
-" { // prepare friction\n"
-" float4 center = make_float4(0.f);\n"
-" for(int i=0; i<src->m_worldNormalOnB.w; i++) \n"
-" center += src->m_worldPosB[i];\n"
-" center /= (float)src->m_worldNormalOnB.w;\n"
-" float4 tangent[2];\n"
-" btPlaneSpace1(-src->m_worldNormalOnB,&tangent[0],&tangent[1]);\n"
-" \n"
-" float4 r[2];\n"
-" r[0] = center - posA;\n"
-" r[1] = center - posB;\n"
-" for(int i=0; i<2; i++)\n"
-" {\n"
-" float4 linear, angular0, angular1;\n"
-" setLinearAndAngular(tangent[i], r[0], r[1], &linear, &angular0, &angular1);\n"
-" dstC->m_fJacCoeffInv[i] = calcJacCoeff(linear, -linear, angular0, angular1,\n"
-" invMassA, &invInertiaA, invMassB, &invInertiaB ,countA, countB);\n"
-" dstC->m_fAppliedRambdaDt[i] = 0.f;\n"
-" }\n"
-" dstC->m_center = center;\n"
-" }\n"
-" for(int i=0; i<4; i++)\n"
-" {\n"
-" if( i<src->m_worldNormalOnB.w )\n"
-" {\n"
-" dstC->m_worldPos[i] = src->m_worldPosB[i];\n"
-" }\n"
-" else\n"
-" {\n"
-" dstC->m_worldPos[i] = make_float4(0.f);\n"
-" }\n"
-" }\n"
-"}\n"
-"__kernel\n"
-"__attribute__((reqd_work_group_size(WG_SIZE,1,1)))\n"
-"void ContactToConstraintSplitKernel(__global const struct b3Contact4Data* gContact, __global const Body* gBodies, __global const Shape* gShapes, __global Constraint4* gConstraintOut, \n"
-"__global const unsigned int* bodyCount,\n"
-"int nContacts,\n"
-"float dt,\n"
-"float positionDrift,\n"
-"float positionConstraintCoeff\n"
-")\n"
-"{\n"
-" int gIdx = GET_GLOBAL_IDX;\n"
-" \n"
-" if( gIdx < nContacts )\n"
-" {\n"
-" int aIdx = abs(gContact[gIdx].m_bodyAPtrAndSignBit);\n"
-" int bIdx = abs(gContact[gIdx].m_bodyBPtrAndSignBit);\n"
-" float4 posA = gBodies[aIdx].m_pos;\n"
-" float4 linVelA = gBodies[aIdx].m_linVel;\n"
-" float4 angVelA = gBodies[aIdx].m_angVel;\n"
-" float invMassA = gBodies[aIdx].m_invMass;\n"
-" Matrix3x3 invInertiaA = gShapes[aIdx].m_invInertia;\n"
-" float4 posB = gBodies[bIdx].m_pos;\n"
-" float4 linVelB = gBodies[bIdx].m_linVel;\n"
-" float4 angVelB = gBodies[bIdx].m_angVel;\n"
-" float invMassB = gBodies[bIdx].m_invMass;\n"
-" Matrix3x3 invInertiaB = gShapes[bIdx].m_invInertia;\n"
-" Constraint4 cs;\n"
-" float countA = invMassA ? (float)bodyCount[aIdx] : 1;\n"
-" float countB = invMassB ? (float)bodyCount[bIdx] : 1;\n"
-" setConstraint4( posA, linVelA, angVelA, invMassA, invInertiaA, posB, linVelB, angVelB, invMassB, invInertiaB,\n"
-" &gContact[gIdx], dt, positionDrift, positionConstraintCoeff,countA,countB,\n"
-" &cs );\n"
-" \n"
-" cs.m_batchIdx = gContact[gIdx].m_batchIdx;\n"
-" gConstraintOut[gIdx] = cs;\n"
-" }\n"
-"}\n"
-;
/include/Bullet/Bullet3OpenCL/RigidBody/kernels/solverUtils.h
Property changes:
Deleted: svn:executable
## -1 +0,0 ##
-*
\ No newline at end of property
Index: include/Bullet/Bullet3OpenCL/RigidBody/kernels/solverSetup.h
===================================================================
--- include/Bullet/Bullet3OpenCL/RigidBody/kernels/solverSetup.h (revision 22)
+++ include/Bullet/Bullet3OpenCL/RigidBody/kernels/solverSetup.h (nonexistent)
@@ -1,702 +0,0 @@
-//this file is autogenerated using stringify.bat (premake --stringify) in the build folder of this project
-static const char* solverSetupCL= \
-"/*\n"
-"Copyright (c) 2012 Advanced Micro Devices, Inc. \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"
-"//Originally written by Takahiro Harada\n"
-"#ifndef B3_CONTACT4DATA_H\n"
-"#define B3_CONTACT4DATA_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"
-"//keep B3_LARGE_FLOAT*B3_LARGE_FLOAT < FLT_MAX\n"
-"#define B3_LARGE_FLOAT 1e18f\n"
-"#define B3_INFINITY 1e18f\n"
-"#define b3Assert(a)\n"
-"#define b3ConstArray(a) __global const a*\n"
-"#define b3AtomicInc atomic_inc\n"
-"#define b3AtomicAdd atomic_add\n"
-"#define b3Fabs fabs\n"
-"#define b3Sqrt native_sqrt\n"
-"#define b3Sin native_sin\n"
-"#define b3Cos native_cos\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"
-" #define b3MinFloat4 min\n"
-" #define b3MaxFloat4 max\n"
-" #define b3Normalized(a) normalize(a)\n"
-"#endif \n"
-" \n"
-"inline bool b3IsAlmostZero(b3Float4ConstArg v)\n"
-"{\n"
-" if(b3Fabs(v.x)>1e-6 || b3Fabs(v.y)>1e-6 || b3Fabs(v.z)>1e-6) \n"
-" return false;\n"
-" return true;\n"
-"}\n"
-"inline int b3MaxDot( b3Float4ConstArg vec, __global const b3Float4* vecArray, int vecLen, float* dotOut )\n"
-"{\n"
-" float maxDot = -B3_INFINITY;\n"
-" int i = 0;\n"
-" int ptIndex = -1;\n"
-" for( i = 0; i < vecLen; i++ )\n"
-" {\n"
-" float dot = b3Dot3F4(vecArray[i],vec);\n"
-" \n"
-" if( dot > maxDot )\n"
-" {\n"
-" maxDot = dot;\n"
-" ptIndex = i;\n"
-" }\n"
-" }\n"
-" b3Assert(ptIndex>=0);\n"
-" if (ptIndex<0)\n"
-" {\n"
-" ptIndex = 0;\n"
-" }\n"
-" *dotOut = maxDot;\n"
-" return ptIndex;\n"
-"}\n"
-"#endif //B3_FLOAT4_H\n"
-"typedef struct b3Contact4Data b3Contact4Data_t;\n"
-"struct b3Contact4Data\n"
-"{\n"
-" b3Float4 m_worldPosB[4];\n"
-"// b3Float4 m_localPosA[4];\n"
-"// b3Float4 m_localPosB[4];\n"
-" b3Float4 m_worldNormalOnB; // w: m_nPoints\n"
-" unsigned short m_restituitionCoeffCmp;\n"
-" unsigned short m_frictionCoeffCmp;\n"
-" int m_batchIdx;\n"
-" int m_bodyAPtrAndSignBit;//x:m_bodyAPtr, y:m_bodyBPtr\n"
-" int m_bodyBPtrAndSignBit;\n"
-" int m_childIndexA;\n"
-" int m_childIndexB;\n"
-" int m_unused1;\n"
-" int m_unused2;\n"
-"};\n"
-"inline int b3Contact4Data_getNumPoints(const struct b3Contact4Data* contact)\n"
-"{\n"
-" return (int)contact->m_worldNormalOnB.w;\n"
-"};\n"
-"inline void b3Contact4Data_setNumPoints(struct b3Contact4Data* contact, int numPoints)\n"
-"{\n"
-" contact->m_worldNormalOnB.w = (float)numPoints;\n"
-"};\n"
-"#endif //B3_CONTACT4DATA_H\n"
-"#ifndef B3_CONTACT_CONSTRAINT5_H\n"
-"#define B3_CONTACT_CONSTRAINT5_H\n"
-"#ifndef B3_FLOAT4_H\n"
-"#ifdef __cplusplus\n"
-"#else\n"
-"#endif \n"
-"#endif //B3_FLOAT4_H\n"
-"typedef struct b3ContactConstraint4 b3ContactConstraint4_t;\n"
-"struct b3ContactConstraint4\n"
-"{\n"
-" b3Float4 m_linear;//normal?\n"
-" b3Float4 m_worldPos[4];\n"
-" b3Float4 m_center; // friction\n"
-" float m_jacCoeffInv[4];\n"
-" float m_b[4];\n"
-" float m_appliedRambdaDt[4];\n"
-" float m_fJacCoeffInv[2]; // friction\n"
-" float m_fAppliedRambdaDt[2]; // friction\n"
-" unsigned int m_bodyA;\n"
-" unsigned int m_bodyB;\n"
-" int m_batchIdx;\n"
-" unsigned int m_paddings;\n"
-"};\n"
-"//inline void setFrictionCoeff(float value) { m_linear[3] = value; }\n"
-"inline float b3GetFrictionCoeff(b3ContactConstraint4_t* constraint) \n"
-"{\n"
-" return constraint->m_linear.w; \n"
-"}\n"
-"#endif //B3_CONTACT_CONSTRAINT5_H\n"
-"#ifndef B3_RIGIDBODY_DATA_H\n"
-"#define B3_RIGIDBODY_DATA_H\n"
-"#ifndef B3_FLOAT4_H\n"
-"#ifdef __cplusplus\n"
-"#else\n"
-"#endif \n"
-"#endif //B3_FLOAT4_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 b3QuatNormalized(b3QuatConstArg in);\n"
-"inline b3Quat b3QuatRotate(b3QuatConstArg q, b3QuatConstArg vec);\n"
-"inline b3Quat b3QuatInvert(b3QuatConstArg q);\n"
-"inline b3Quat b3QuatInverse(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 b3QuatNormalized(b3QuatConstArg in)\n"
-"{\n"
-" b3Quat q;\n"
-" q=in;\n"
-" //return b3FastNormalize4(in);\n"
-" float len = native_sqrt(dot(q, q));\n"
-" if(len > 0.f)\n"
-" {\n"
-" q *= 1.f / len;\n"
-" }\n"
-" else\n"
-" {\n"
-" q.x = q.y = q.z = 0.f;\n"
-" q.w = 1.f;\n"
-" }\n"
-" return q;\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 b3QuatInverse(b3QuatConstArg q)\n"
-"{\n"
-" return (b3Quat)(-q.xyz, q.w);\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"
-"#ifndef B3_MAT3x3_H\n"
-"#define B3_MAT3x3_H\n"
-"#ifndef B3_QUAT_H\n"
-"#ifdef __cplusplus\n"
-"#else\n"
-"#endif \n"
-"#endif //B3_QUAT_H\n"
-"#ifdef __cplusplus\n"
-"#else\n"
-"typedef struct\n"
-"{\n"
-" b3Float4 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"
-" b3Float4 quat2 = (b3Float4)(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"
-"__inline\n"
-"b3Mat3x3 mtZero();\n"
-"__inline\n"
-"b3Mat3x3 mtIdentity();\n"
-"__inline\n"
-"b3Mat3x3 mtTranspose(b3Mat3x3 m);\n"
-"__inline\n"
-"b3Mat3x3 mtMul(b3Mat3x3 a, b3Mat3x3 b);\n"
-"__inline\n"
-"b3Float4 mtMul1(b3Mat3x3 a, b3Float4 b);\n"
-"__inline\n"
-"b3Float4 mtMul3(b3Float4 a, b3Mat3x3 b);\n"
-"__inline\n"
-"b3Mat3x3 mtZero()\n"
-"{\n"
-" b3Mat3x3 m;\n"
-" m.m_row[0] = (b3Float4)(0.f);\n"
-" m.m_row[1] = (b3Float4)(0.f);\n"
-" m.m_row[2] = (b3Float4)(0.f);\n"
-" return m;\n"
-"}\n"
-"__inline\n"
-"b3Mat3x3 mtIdentity()\n"
-"{\n"
-" b3Mat3x3 m;\n"
-" m.m_row[0] = (b3Float4)(1,0,0,0);\n"
-" m.m_row[1] = (b3Float4)(0,1,0,0);\n"
-" m.m_row[2] = (b3Float4)(0,0,1,0);\n"
-" return m;\n"
-"}\n"
-"__inline\n"
-"b3Mat3x3 mtTranspose(b3Mat3x3 m)\n"
-"{\n"
-" b3Mat3x3 out;\n"
-" out.m_row[0] = (b3Float4)(m.m_row[0].x, m.m_row[1].x, m.m_row[2].x, 0.f);\n"
-" out.m_row[1] = (b3Float4)(m.m_row[0].y, m.m_row[1].y, m.m_row[2].y, 0.f);\n"
-" out.m_row[2] = (b3Float4)(m.m_row[0].z, m.m_row[1].z, m.m_row[2].z, 0.f);\n"
-" return out;\n"
-"}\n"
-"__inline\n"
-"b3Mat3x3 mtMul(b3Mat3x3 a, b3Mat3x3 b)\n"
-"{\n"
-" b3Mat3x3 transB;\n"
-" transB = mtTranspose( b );\n"
-" b3Mat3x3 ans;\n"
-" // why this doesn't run when 0ing in the for{}\n"
-" a.m_row[0].w = 0.f;\n"
-" a.m_row[1].w = 0.f;\n"
-" a.m_row[2].w = 0.f;\n"
-" for(int i=0; i<3; i++)\n"
-" {\n"
-"// a.m_row[i].w = 0.f;\n"
-" ans.m_row[i].x = b3Dot3F4(a.m_row[i],transB.m_row[0]);\n"
-" ans.m_row[i].y = b3Dot3F4(a.m_row[i],transB.m_row[1]);\n"
-" ans.m_row[i].z = b3Dot3F4(a.m_row[i],transB.m_row[2]);\n"
-" ans.m_row[i].w = 0.f;\n"
-" }\n"
-" return ans;\n"
-"}\n"
-"__inline\n"
-"b3Float4 mtMul1(b3Mat3x3 a, b3Float4 b)\n"
-"{\n"
-" b3Float4 ans;\n"
-" ans.x = b3Dot3F4( a.m_row[0], b );\n"
-" ans.y = b3Dot3F4( a.m_row[1], b );\n"
-" ans.z = b3Dot3F4( a.m_row[2], b );\n"
-" ans.w = 0.f;\n"
-" return ans;\n"
-"}\n"
-"__inline\n"
-"b3Float4 mtMul3(b3Float4 a, b3Mat3x3 b)\n"
-"{\n"
-" b3Float4 colx = b3MakeFloat4(b.m_row[0].x, b.m_row[1].x, b.m_row[2].x, 0);\n"
-" b3Float4 coly = b3MakeFloat4(b.m_row[0].y, b.m_row[1].y, b.m_row[2].y, 0);\n"
-" b3Float4 colz = b3MakeFloat4(b.m_row[0].z, b.m_row[1].z, b.m_row[2].z, 0);\n"
-" b3Float4 ans;\n"
-" ans.x = b3Dot3F4( a, colx );\n"
-" ans.y = b3Dot3F4( a, coly );\n"
-" ans.z = b3Dot3F4( a, colz );\n"
-" return ans;\n"
-"}\n"
-"#endif\n"
-"#endif //B3_MAT3x3_H\n"
-"typedef struct b3RigidBodyData b3RigidBodyData_t;\n"
-"struct b3RigidBodyData\n"
-"{\n"
-" b3Float4 m_pos;\n"
-" b3Quat m_quat;\n"
-" b3Float4 m_linVel;\n"
-" b3Float4 m_angVel;\n"
-" int m_collidableIdx;\n"
-" float m_invMass;\n"
-" float m_restituitionCoeff;\n"
-" float m_frictionCoeff;\n"
-"};\n"
-"typedef struct b3InertiaData b3InertiaData_t;\n"
-"struct b3InertiaData\n"
-"{\n"
-" b3Mat3x3 m_invInertiaWorld;\n"
-" b3Mat3x3 m_initInvInertia;\n"
-"};\n"
-"#endif //B3_RIGIDBODY_DATA_H\n"
-" \n"
-"void b3PlaneSpace1 (b3Float4ConstArg n, b3Float4* p, b3Float4* q);\n"
-" void b3PlaneSpace1 (b3Float4ConstArg n, b3Float4* p, b3Float4* q)\n"
-"{\n"
-" if (b3Fabs(n.z) > 0.70710678f) {\n"
-" // choose p in y-z plane\n"
-" float a = n.y*n.y + n.z*n.z;\n"
-" float k = 1.f/sqrt(a);\n"
-" p[0].x = 0;\n"
-" p[0].y = -n.z*k;\n"
-" p[0].z = n.y*k;\n"
-" // set q = n x p\n"
-" q[0].x = a*k;\n"
-" q[0].y = -n.x*p[0].z;\n"
-" q[0].z = n.x*p[0].y;\n"
-" }\n"
-" else {\n"
-" // choose p in x-y plane\n"
-" float a = n.x*n.x + n.y*n.y;\n"
-" float k = 1.f/sqrt(a);\n"
-" p[0].x = -n.y*k;\n"
-" p[0].y = n.x*k;\n"
-" p[0].z = 0;\n"
-" // set q = n x p\n"
-" q[0].x = -n.z*p[0].y;\n"
-" q[0].y = n.z*p[0].x;\n"
-" q[0].z = a*k;\n"
-" }\n"
-"}\n"
-" \n"
-"void setLinearAndAngular( b3Float4ConstArg n, b3Float4ConstArg r0, b3Float4ConstArg r1, b3Float4* linear, b3Float4* angular0, b3Float4* angular1)\n"
-"{\n"
-" *linear = b3MakeFloat4(n.x,n.y,n.z,0.f);\n"
-" *angular0 = b3Cross3(r0, n);\n"
-" *angular1 = -b3Cross3(r1, n);\n"
-"}\n"
-"float calcRelVel( b3Float4ConstArg l0, b3Float4ConstArg l1, b3Float4ConstArg a0, b3Float4ConstArg a1, b3Float4ConstArg linVel0,\n"
-" b3Float4ConstArg angVel0, b3Float4ConstArg linVel1, b3Float4ConstArg angVel1 )\n"
-"{\n"
-" return b3Dot3F4(l0, linVel0) + b3Dot3F4(a0, angVel0) + b3Dot3F4(l1, linVel1) + b3Dot3F4(a1, angVel1);\n"
-"}\n"
-"float calcJacCoeff(b3Float4ConstArg linear0, b3Float4ConstArg linear1, b3Float4ConstArg angular0, b3Float4ConstArg angular1,\n"
-" float invMass0, const b3Mat3x3* invInertia0, float invMass1, const b3Mat3x3* invInertia1)\n"
-"{\n"
-" // linear0,1 are normlized\n"
-" float jmj0 = invMass0;//b3Dot3F4(linear0, linear0)*invMass0;\n"
-" float jmj1 = b3Dot3F4(mtMul3(angular0,*invInertia0), angular0);\n"
-" float jmj2 = invMass1;//b3Dot3F4(linear1, linear1)*invMass1;\n"
-" float jmj3 = b3Dot3F4(mtMul3(angular1,*invInertia1), angular1);\n"
-" return -1.f/(jmj0+jmj1+jmj2+jmj3);\n"
-"}\n"
-"void setConstraint4( b3Float4ConstArg posA, b3Float4ConstArg linVelA, b3Float4ConstArg angVelA, float invMassA, b3Mat3x3ConstArg invInertiaA,\n"
-" b3Float4ConstArg posB, b3Float4ConstArg linVelB, b3Float4ConstArg angVelB, float invMassB, b3Mat3x3ConstArg invInertiaB, \n"
-" __global struct b3Contact4Data* src, float dt, float positionDrift, float positionConstraintCoeff,\n"
-" b3ContactConstraint4_t* dstC )\n"
-"{\n"
-" dstC->m_bodyA = abs(src->m_bodyAPtrAndSignBit);\n"
-" dstC->m_bodyB = abs(src->m_bodyBPtrAndSignBit);\n"
-" float dtInv = 1.f/dt;\n"
-" for(int ic=0; ic<4; ic++)\n"
-" {\n"
-" dstC->m_appliedRambdaDt[ic] = 0.f;\n"
-" }\n"
-" dstC->m_fJacCoeffInv[0] = dstC->m_fJacCoeffInv[1] = 0.f;\n"
-" dstC->m_linear = src->m_worldNormalOnB;\n"
-" dstC->m_linear.w = 0.7f ;//src->getFrictionCoeff() );\n"
-" for(int ic=0; ic<4; ic++)\n"
-" {\n"
-" b3Float4 r0 = src->m_worldPosB[ic] - posA;\n"
-" b3Float4 r1 = src->m_worldPosB[ic] - posB;\n"
-" if( ic >= src->m_worldNormalOnB.w )//npoints\n"
-" {\n"
-" dstC->m_jacCoeffInv[ic] = 0.f;\n"
-" continue;\n"
-" }\n"
-" float relVelN;\n"
-" {\n"
-" b3Float4 linear, angular0, angular1;\n"
-" setLinearAndAngular(src->m_worldNormalOnB, r0, r1, &linear, &angular0, &angular1);\n"
-" dstC->m_jacCoeffInv[ic] = calcJacCoeff(linear, -linear, angular0, angular1,\n"
-" invMassA, &invInertiaA, invMassB, &invInertiaB );\n"
-" relVelN = calcRelVel(linear, -linear, angular0, angular1,\n"
-" linVelA, angVelA, linVelB, angVelB);\n"
-" float e = 0.f;//src->getRestituitionCoeff();\n"
-" if( relVelN*relVelN < 0.004f ) e = 0.f;\n"
-" dstC->m_b[ic] = e*relVelN;\n"
-" //float penetration = src->m_worldPosB[ic].w;\n"
-" dstC->m_b[ic] += (src->m_worldPosB[ic].w + positionDrift)*positionConstraintCoeff*dtInv;\n"
-" dstC->m_appliedRambdaDt[ic] = 0.f;\n"
-" }\n"
-" }\n"
-" if( src->m_worldNormalOnB.w > 0 )//npoints\n"
-" { // prepare friction\n"
-" b3Float4 center = b3MakeFloat4(0.f,0.f,0.f,0.f);\n"
-" for(int i=0; i<src->m_worldNormalOnB.w; i++) \n"
-" center += src->m_worldPosB[i];\n"
-" center /= (float)src->m_worldNormalOnB.w;\n"
-" b3Float4 tangent[2];\n"
-" b3PlaneSpace1(src->m_worldNormalOnB,&tangent[0],&tangent[1]);\n"
-" \n"
-" b3Float4 r[2];\n"
-" r[0] = center - posA;\n"
-" r[1] = center - posB;\n"
-" for(int i=0; i<2; i++)\n"
-" {\n"
-" b3Float4 linear, angular0, angular1;\n"
-" setLinearAndAngular(tangent[i], r[0], r[1], &linear, &angular0, &angular1);\n"
-" dstC->m_fJacCoeffInv[i] = calcJacCoeff(linear, -linear, angular0, angular1,\n"
-" invMassA, &invInertiaA, invMassB, &invInertiaB );\n"
-" dstC->m_fAppliedRambdaDt[i] = 0.f;\n"
-" }\n"
-" dstC->m_center = center;\n"
-" }\n"
-" for(int i=0; i<4; i++)\n"
-" {\n"
-" if( i<src->m_worldNormalOnB.w )\n"
-" {\n"
-" dstC->m_worldPos[i] = src->m_worldPosB[i];\n"
-" }\n"
-" else\n"
-" {\n"
-" dstC->m_worldPos[i] = b3MakeFloat4(0.f,0.f,0.f,0.f);\n"
-" }\n"
-" }\n"
-"}\n"
-"#pragma OPENCL EXTENSION cl_amd_printf : enable\n"
-"#pragma OPENCL EXTENSION cl_khr_local_int32_base_atomics : enable\n"
-"#pragma OPENCL EXTENSION cl_khr_global_int32_base_atomics : enable\n"
-"#pragma OPENCL EXTENSION cl_khr_local_int32_extended_atomics : enable\n"
-"#pragma OPENCL EXTENSION cl_khr_global_int32_extended_atomics : enable\n"
-"#ifdef cl_ext_atomic_counters_32\n"
-"#pragma OPENCL EXTENSION cl_ext_atomic_counters_32 : enable\n"
-"#else\n"
-"#define counter32_t volatile global int*\n"
-"#endif\n"
-"typedef unsigned int u32;\n"
-"typedef unsigned short u16;\n"
-"typedef unsigned char u8;\n"
-"#define GET_GROUP_IDX get_group_id(0)\n"
-"#define GET_LOCAL_IDX get_local_id(0)\n"
-"#define GET_GLOBAL_IDX get_global_id(0)\n"
-"#define GET_GROUP_SIZE get_local_size(0)\n"
-"#define GET_NUM_GROUPS get_num_groups(0)\n"
-"#define GROUP_LDS_BARRIER barrier(CLK_LOCAL_MEM_FENCE)\n"
-"#define GROUP_MEM_FENCE mem_fence(CLK_LOCAL_MEM_FENCE)\n"
-"#define AtomInc(x) atom_inc(&(x))\n"
-"#define AtomInc1(x, out) out = atom_inc(&(x))\n"
-"#define AppendInc(x, out) out = atomic_inc(x)\n"
-"#define AtomAdd(x, value) atom_add(&(x), value)\n"
-"#define AtomCmpxhg(x, cmp, value) atom_cmpxchg( &(x), cmp, value )\n"
-"#define AtomXhg(x, value) atom_xchg ( &(x), value )\n"
-"#define SELECT_UINT4( b, a, condition ) select( b,a,condition )\n"
-"#define make_float4 (float4)\n"
-"#define make_float2 (float2)\n"
-"#define make_uint4 (uint4)\n"
-"#define make_int4 (int4)\n"
-"#define make_uint2 (uint2)\n"
-"#define make_int2 (int2)\n"
-"#define max2 max\n"
-"#define min2 min\n"
-"///////////////////////////////////////\n"
-"// Vector\n"
-"///////////////////////////////////////\n"
-"__inline\n"
-"float fastDiv(float numerator, float denominator)\n"
-"{\n"
-" return native_divide(numerator, denominator); \n"
-"// return numerator/denominator; \n"
-"}\n"
-"__inline\n"
-"float4 fastDiv4(float4 numerator, float4 denominator)\n"
-"{\n"
-" return native_divide(numerator, denominator); \n"
-"}\n"
-"__inline\n"
-"float fastSqrtf(float f2)\n"
-"{\n"
-" return native_sqrt(f2);\n"
-"// return sqrt(f2);\n"
-"}\n"
-"__inline\n"
-"float fastRSqrt(float f2)\n"
-"{\n"
-" return native_rsqrt(f2);\n"
-"}\n"
-"__inline\n"
-"float fastLength4(float4 v)\n"
-"{\n"
-" return fast_length(v);\n"
-"}\n"
-"__inline\n"
-"float4 fastNormalize4(float4 v)\n"
-"{\n"
-" return fast_normalize(v);\n"
-"}\n"
-"__inline\n"
-"float sqrtf(float a)\n"
-"{\n"
-"// return sqrt(a);\n"
-" return native_sqrt(a);\n"
-"}\n"
-"__inline\n"
-"float4 cross3(float4 a, float4 b)\n"
-"{\n"
-" return cross(a,b);\n"
-"}\n"
-"__inline\n"
-"float dot3F4(float4 a, float4 b)\n"
-"{\n"
-" float4 a1 = make_float4(a.xyz,0.f);\n"
-" float4 b1 = make_float4(b.xyz,0.f);\n"
-" return dot(a1, b1);\n"
-"}\n"
-"__inline\n"
-"float length3(const float4 a)\n"
-"{\n"
-" return sqrtf(dot3F4(a,a));\n"
-"}\n"
-"__inline\n"
-"float dot4(const float4 a, const float4 b)\n"
-"{\n"
-" return dot( a, b );\n"
-"}\n"
-"// for height\n"
-"__inline\n"
-"float dot3w1(const float4 point, const float4 eqn)\n"
-"{\n"
-" return dot3F4(point,eqn) + eqn.w;\n"
-"}\n"
-"__inline\n"
-"float4 normalize3(const float4 a)\n"
-"{\n"
-" float4 n = make_float4(a.x, a.y, a.z, 0.f);\n"
-" return fastNormalize4( n );\n"
-"// float length = sqrtf(dot3F4(a, a));\n"
-"// return 1.f/length * a;\n"
-"}\n"
-"__inline\n"
-"float4 normalize4(const float4 a)\n"
-"{\n"
-" float length = sqrtf(dot4(a, a));\n"
-" return 1.f/length * a;\n"
-"}\n"
-"__inline\n"
-"float4 createEquation(const float4 a, const float4 b, const float4 c)\n"
-"{\n"
-" float4 eqn;\n"
-" float4 ab = b-a;\n"
-" float4 ac = c-a;\n"
-" eqn = normalize3( cross3(ab, ac) );\n"
-" eqn.w = -dot3F4(eqn,a);\n"
-" return eqn;\n"
-"}\n"
-"#define WG_SIZE 64\n"
-"typedef struct\n"
-"{\n"
-" int m_nConstraints;\n"
-" int m_start;\n"
-" int m_batchIdx;\n"
-" int m_nSplit;\n"
-"// int m_paddings[1];\n"
-"} ConstBuffer;\n"
-"typedef struct\n"
-"{\n"
-" int m_solveFriction;\n"
-" int m_maxBatch; // long batch really kills the performance\n"
-" int m_batchIdx;\n"
-" int m_nSplit;\n"
-"// int m_paddings[1];\n"
-"} ConstBufferBatchSolve;\n"
-" \n"
-"typedef struct \n"
-"{\n"
-" int m_valInt0;\n"
-" int m_valInt1;\n"
-" int m_valInt2;\n"
-" int m_valInt3;\n"
-" float m_val0;\n"
-" float m_val1;\n"
-" float m_val2;\n"
-" float m_val3;\n"
-"} SolverDebugInfo;\n"
-"typedef struct\n"
-"{\n"
-" int m_nContacts;\n"
-" float m_dt;\n"
-" float m_positionDrift;\n"
-" float m_positionConstraintCoeff;\n"
-"} ConstBufferCTC;\n"
-"__kernel\n"
-"__attribute__((reqd_work_group_size(WG_SIZE,1,1)))\n"
-"void ContactToConstraintKernel(__global struct b3Contact4Data* gContact, __global b3RigidBodyData_t* gBodies, __global b3InertiaData_t* gShapes, __global b3ContactConstraint4_t* gConstraintOut, \n"
-"int nContacts,\n"
-"float dt,\n"
-"float positionDrift,\n"
-"float positionConstraintCoeff\n"
-")\n"
-"{\n"
-" int gIdx = GET_GLOBAL_IDX;\n"
-" \n"
-" if( gIdx < nContacts )\n"
-" {\n"
-" int aIdx = abs(gContact[gIdx].m_bodyAPtrAndSignBit);\n"
-" int bIdx = abs(gContact[gIdx].m_bodyBPtrAndSignBit);\n"
-" float4 posA = gBodies[aIdx].m_pos;\n"
-" float4 linVelA = gBodies[aIdx].m_linVel;\n"
-" float4 angVelA = gBodies[aIdx].m_angVel;\n"
-" float invMassA = gBodies[aIdx].m_invMass;\n"
-" b3Mat3x3 invInertiaA = gShapes[aIdx].m_initInvInertia;\n"
-" float4 posB = gBodies[bIdx].m_pos;\n"
-" float4 linVelB = gBodies[bIdx].m_linVel;\n"
-" float4 angVelB = gBodies[bIdx].m_angVel;\n"
-" float invMassB = gBodies[bIdx].m_invMass;\n"
-" b3Mat3x3 invInertiaB = gShapes[bIdx].m_initInvInertia;\n"
-" b3ContactConstraint4_t cs;\n"
-" setConstraint4( posA, linVelA, angVelA, invMassA, invInertiaA, posB, linVelB, angVelB, invMassB, invInertiaB,\n"
-" &gContact[gIdx], dt, positionDrift, positionConstraintCoeff,\n"
-" &cs );\n"
-" \n"
-" cs.m_batchIdx = gContact[gIdx].m_batchIdx;\n"
-" gConstraintOut[gIdx] = cs;\n"
-" }\n"
-"}\n"
-;
/include/Bullet/Bullet3OpenCL/RigidBody/kernels/solverSetup.h
Property changes:
Deleted: svn:executable
## -1 +0,0 ##
-*
\ No newline at end of property
Index: include/Bullet/Bullet3OpenCL/RigidBody/kernels/updateAabbsKernel.h
===================================================================
--- include/Bullet/Bullet3OpenCL/RigidBody/kernels/updateAabbsKernel.h (revision 22)
+++ include/Bullet/Bullet3OpenCL/RigidBody/kernels/updateAabbsKernel.h (nonexistent)
@@ -1,466 +0,0 @@
-//this file is autogenerated using stringify.bat (premake --stringify) in the build folder of this project
-static const char* updateAabbsKernelCL= \
-"#ifndef B3_UPDATE_AABBS_H\n"
-"#define B3_UPDATE_AABBS_H\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"
-"//keep B3_LARGE_FLOAT*B3_LARGE_FLOAT < FLT_MAX\n"
-"#define B3_LARGE_FLOAT 1e18f\n"
-"#define B3_INFINITY 1e18f\n"
-"#define b3Assert(a)\n"
-"#define b3ConstArray(a) __global const a*\n"
-"#define b3AtomicInc atomic_inc\n"
-"#define b3AtomicAdd atomic_add\n"
-"#define b3Fabs fabs\n"
-"#define b3Sqrt native_sqrt\n"
-"#define b3Sin native_sin\n"
-"#define b3Cos native_cos\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"
-" #define b3MinFloat4 min\n"
-" #define b3MaxFloat4 max\n"
-" #define b3Normalized(a) normalize(a)\n"
-"#endif \n"
-" \n"
-"inline bool b3IsAlmostZero(b3Float4ConstArg v)\n"
-"{\n"
-" if(b3Fabs(v.x)>1e-6 || b3Fabs(v.y)>1e-6 || b3Fabs(v.z)>1e-6) \n"
-" return false;\n"
-" return true;\n"
-"}\n"
-"inline int b3MaxDot( b3Float4ConstArg vec, __global const b3Float4* vecArray, int vecLen, float* dotOut )\n"
-"{\n"
-" float maxDot = -B3_INFINITY;\n"
-" int i = 0;\n"
-" int ptIndex = -1;\n"
-" for( i = 0; i < vecLen; i++ )\n"
-" {\n"
-" float dot = b3Dot3F4(vecArray[i],vec);\n"
-" \n"
-" if( dot > maxDot )\n"
-" {\n"
-" maxDot = dot;\n"
-" ptIndex = i;\n"
-" }\n"
-" }\n"
-" b3Assert(ptIndex>=0);\n"
-" if (ptIndex<0)\n"
-" {\n"
-" ptIndex = 0;\n"
-" }\n"
-" *dotOut = maxDot;\n"
-" return ptIndex;\n"
-"}\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 b3QuatNormalized(b3QuatConstArg in);\n"
-"inline b3Quat b3QuatRotate(b3QuatConstArg q, b3QuatConstArg vec);\n"
-"inline b3Quat b3QuatInvert(b3QuatConstArg q);\n"
-"inline b3Quat b3QuatInverse(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 b3QuatNormalized(b3QuatConstArg in)\n"
-"{\n"
-" b3Quat q;\n"
-" q=in;\n"
-" //return b3FastNormalize4(in);\n"
-" float len = native_sqrt(dot(q, q));\n"
-" if(len > 0.f)\n"
-" {\n"
-" q *= 1.f / len;\n"
-" }\n"
-" else\n"
-" {\n"
-" q.x = q.y = q.z = 0.f;\n"
-" q.w = 1.f;\n"
-" }\n"
-" return q;\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 b3QuatInverse(b3QuatConstArg q)\n"
-"{\n"
-" return (b3Quat)(-q.xyz, q.w);\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"
-" b3Float4 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"
-" b3Float4 quat2 = (b3Float4)(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"
-"__inline\n"
-"b3Mat3x3 mtZero();\n"
-"__inline\n"
-"b3Mat3x3 mtIdentity();\n"
-"__inline\n"
-"b3Mat3x3 mtTranspose(b3Mat3x3 m);\n"
-"__inline\n"
-"b3Mat3x3 mtMul(b3Mat3x3 a, b3Mat3x3 b);\n"
-"__inline\n"
-"b3Float4 mtMul1(b3Mat3x3 a, b3Float4 b);\n"
-"__inline\n"
-"b3Float4 mtMul3(b3Float4 a, b3Mat3x3 b);\n"
-"__inline\n"
-"b3Mat3x3 mtZero()\n"
-"{\n"
-" b3Mat3x3 m;\n"
-" m.m_row[0] = (b3Float4)(0.f);\n"
-" m.m_row[1] = (b3Float4)(0.f);\n"
-" m.m_row[2] = (b3Float4)(0.f);\n"
-" return m;\n"
-"}\n"
-"__inline\n"
-"b3Mat3x3 mtIdentity()\n"
-"{\n"
-" b3Mat3x3 m;\n"
-" m.m_row[0] = (b3Float4)(1,0,0,0);\n"
-" m.m_row[1] = (b3Float4)(0,1,0,0);\n"
-" m.m_row[2] = (b3Float4)(0,0,1,0);\n"
-" return m;\n"
-"}\n"
-"__inline\n"
-"b3Mat3x3 mtTranspose(b3Mat3x3 m)\n"
-"{\n"
-" b3Mat3x3 out;\n"
-" out.m_row[0] = (b3Float4)(m.m_row[0].x, m.m_row[1].x, m.m_row[2].x, 0.f);\n"
-" out.m_row[1] = (b3Float4)(m.m_row[0].y, m.m_row[1].y, m.m_row[2].y, 0.f);\n"
-" out.m_row[2] = (b3Float4)(m.m_row[0].z, m.m_row[1].z, m.m_row[2].z, 0.f);\n"
-" return out;\n"
-"}\n"
-"__inline\n"
-"b3Mat3x3 mtMul(b3Mat3x3 a, b3Mat3x3 b)\n"
-"{\n"
-" b3Mat3x3 transB;\n"
-" transB = mtTranspose( b );\n"
-" b3Mat3x3 ans;\n"
-" // why this doesn't run when 0ing in the for{}\n"
-" a.m_row[0].w = 0.f;\n"
-" a.m_row[1].w = 0.f;\n"
-" a.m_row[2].w = 0.f;\n"
-" for(int i=0; i<3; i++)\n"
-" {\n"
-"// a.m_row[i].w = 0.f;\n"
-" ans.m_row[i].x = b3Dot3F4(a.m_row[i],transB.m_row[0]);\n"
-" ans.m_row[i].y = b3Dot3F4(a.m_row[i],transB.m_row[1]);\n"
-" ans.m_row[i].z = b3Dot3F4(a.m_row[i],transB.m_row[2]);\n"
-" ans.m_row[i].w = 0.f;\n"
-" }\n"
-" return ans;\n"
-"}\n"
-"__inline\n"
-"b3Float4 mtMul1(b3Mat3x3 a, b3Float4 b)\n"
-"{\n"
-" b3Float4 ans;\n"
-" ans.x = b3Dot3F4( a.m_row[0], b );\n"
-" ans.y = b3Dot3F4( a.m_row[1], b );\n"
-" ans.z = b3Dot3F4( a.m_row[2], b );\n"
-" ans.w = 0.f;\n"
-" return ans;\n"
-"}\n"
-"__inline\n"
-"b3Float4 mtMul3(b3Float4 a, b3Mat3x3 b)\n"
-"{\n"
-" b3Float4 colx = b3MakeFloat4(b.m_row[0].x, b.m_row[1].x, b.m_row[2].x, 0);\n"
-" b3Float4 coly = b3MakeFloat4(b.m_row[0].y, b.m_row[1].y, b.m_row[2].y, 0);\n"
-" b3Float4 colz = b3MakeFloat4(b.m_row[0].z, b.m_row[1].z, b.m_row[2].z, 0);\n"
-" b3Float4 ans;\n"
-" ans.x = b3Dot3F4( a, colx );\n"
-" ans.y = b3Dot3F4( a, coly );\n"
-" ans.z = b3Dot3F4( a, colz );\n"
-" return ans;\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"
-"#ifndef B3_COLLIDABLE_H\n"
-"#define B3_COLLIDABLE_H\n"
-"#ifndef B3_FLOAT4_H\n"
-"#ifdef __cplusplus\n"
-"#else\n"
-"#endif \n"
-"#endif //B3_FLOAT4_H\n"
-"#ifndef B3_QUAT_H\n"
-"#ifdef __cplusplus\n"
-"#else\n"
-"#endif \n"
-"#endif //B3_QUAT_H\n"
-"enum b3ShapeTypes\n"
-"{\n"
-" SHAPE_HEIGHT_FIELD=1,\n"
-" SHAPE_CONVEX_HULL=3,\n"
-" SHAPE_PLANE=4,\n"
-" SHAPE_CONCAVE_TRIMESH=5,\n"
-" SHAPE_COMPOUND_OF_CONVEX_HULLS=6,\n"
-" SHAPE_SPHERE=7,\n"
-" MAX_NUM_SHAPE_TYPES,\n"
-"};\n"
-"typedef struct b3Collidable b3Collidable_t;\n"
-"struct b3Collidable\n"
-"{\n"
-" union {\n"
-" int m_numChildShapes;\n"
-" int m_bvhIndex;\n"
-" };\n"
-" union\n"
-" {\n"
-" float m_radius;\n"
-" int m_compoundBvhIndex;\n"
-" };\n"
-" int m_shapeType;\n"
-" int m_shapeIndex;\n"
-"};\n"
-"typedef struct b3GpuChildShape b3GpuChildShape_t;\n"
-"struct b3GpuChildShape\n"
-"{\n"
-" b3Float4 m_childPosition;\n"
-" b3Quat m_childOrientation;\n"
-" int m_shapeIndex;\n"
-" int m_unused0;\n"
-" int m_unused1;\n"
-" int m_unused2;\n"
-"};\n"
-"struct b3CompoundOverlappingPair\n"
-"{\n"
-" int m_bodyIndexA;\n"
-" int m_bodyIndexB;\n"
-"// int m_pairType;\n"
-" int m_childShapeIndexA;\n"
-" int m_childShapeIndexB;\n"
-"};\n"
-"#endif //B3_COLLIDABLE_H\n"
-"#ifndef B3_RIGIDBODY_DATA_H\n"
-"#define B3_RIGIDBODY_DATA_H\n"
-"#ifndef B3_FLOAT4_H\n"
-"#ifdef __cplusplus\n"
-"#else\n"
-"#endif \n"
-"#endif //B3_FLOAT4_H\n"
-"#ifndef B3_QUAT_H\n"
-"#ifdef __cplusplus\n"
-"#else\n"
-"#endif \n"
-"#endif //B3_QUAT_H\n"
-"#ifndef B3_MAT3x3_H\n"
-"#ifdef __cplusplus\n"
-"#else\n"
-"#endif\n"
-"#endif //B3_MAT3x3_H\n"
-"typedef struct b3RigidBodyData b3RigidBodyData_t;\n"
-"struct b3RigidBodyData\n"
-"{\n"
-" b3Float4 m_pos;\n"
-" b3Quat m_quat;\n"
-" b3Float4 m_linVel;\n"
-" b3Float4 m_angVel;\n"
-" int m_collidableIdx;\n"
-" float m_invMass;\n"
-" float m_restituitionCoeff;\n"
-" float m_frictionCoeff;\n"
-"};\n"
-"typedef struct b3InertiaData b3InertiaData_t;\n"
-"struct b3InertiaData\n"
-"{\n"
-" b3Mat3x3 m_invInertiaWorld;\n"
-" b3Mat3x3 m_initInvInertia;\n"
-"};\n"
-"#endif //B3_RIGIDBODY_DATA_H\n"
-" \n"
-"void b3ComputeWorldAabb( int bodyId, __global const b3RigidBodyData_t* bodies, __global const b3Collidable_t* collidables, __global const b3Aabb_t* localShapeAABB, __global b3Aabb_t* worldAabbs)\n"
-"{\n"
-" __global const b3RigidBodyData_t* body = &bodies[bodyId];\n"
-" b3Float4 position = body->m_pos;\n"
-" b3Quat orientation = body->m_quat;\n"
-" \n"
-" int collidableIndex = body->m_collidableIdx;\n"
-" int shapeIndex = collidables[collidableIndex].m_shapeIndex;\n"
-" \n"
-" if (shapeIndex>=0)\n"
-" {\n"
-" \n"
-" b3Aabb_t localAabb = localShapeAABB[collidableIndex];\n"
-" b3Aabb_t worldAabb;\n"
-" \n"
-" b3Float4 aabbAMinOut,aabbAMaxOut; \n"
-" float margin = 0.f;\n"
-" b3TransformAabb2(localAabb.m_minVec,localAabb.m_maxVec,margin,position,orientation,&aabbAMinOut,&aabbAMaxOut);\n"
-" \n"
-" worldAabb.m_minVec =aabbAMinOut;\n"
-" worldAabb.m_minIndices[3] = bodyId;\n"
-" worldAabb.m_maxVec = aabbAMaxOut;\n"
-" worldAabb.m_signedMaxIndices[3] = body[bodyId].m_invMass==0.f? 0 : 1;\n"
-" worldAabbs[bodyId] = worldAabb;\n"
-" }\n"
-"}\n"
-"#endif //B3_UPDATE_AABBS_H\n"
-"__kernel void initializeGpuAabbsFull( const int numNodes, __global b3RigidBodyData_t* gBodies,__global b3Collidable_t* collidables, __global b3Aabb_t* plocalShapeAABB, __global b3Aabb_t* pAABB)\n"
-"{\n"
-" int nodeID = get_global_id(0);\n"
-" if( nodeID < numNodes )\n"
-" {\n"
-" b3ComputeWorldAabb(nodeID, gBodies, collidables, plocalShapeAABB,pAABB);\n"
-" }\n"
-"}\n"
-"__kernel void clearOverlappingPairsKernel( __global int4* pairs, int numPairs)\n"
-"{\n"
-" int pairId = get_global_id(0);\n"
-" if( pairId< numPairs )\n"
-" {\n"
-" pairs[pairId].z = 0xffffffff;\n"
-" }\n"
-"}\n"
-;
/include/Bullet/Bullet3OpenCL/RigidBody/kernels/updateAabbsKernel.h
Property changes:
Deleted: svn:executable
## -1 +0,0 ##
-*
\ No newline at end of property
Index: include/Bullet/Bullet3OpenCL/RigidBody/b3GpuJacobiContactSolver.h
===================================================================
--- include/Bullet/Bullet3OpenCL/RigidBody/b3GpuJacobiContactSolver.h (revision 22)
+++ include/Bullet/Bullet3OpenCL/RigidBody/b3GpuJacobiContactSolver.h (nonexistent)
@@ -1,62 +0,0 @@
-
-#ifndef B3_GPU_JACOBI_CONTACT_SOLVER_H
-#define B3_GPU_JACOBI_CONTACT_SOLVER_H
-#include "Bullet3OpenCL/Initialize/b3OpenCLInclude.h"
-//#include "Bullet3Collision/NarrowPhaseCollision/shared/b3RigidBodyData.h"
-#include "Bullet3Collision/NarrowPhaseCollision/shared/b3RigidBodyData.h"
-
-#include "Bullet3Collision/NarrowPhaseCollision/shared/b3Contact4Data.h"
-#include "Bullet3OpenCL/ParallelPrimitives/b3OpenCLArray.h"
-
-
-//struct b3InertiaData;
-//b3InertiaData
-
-class b3TypedConstraint;
-
-struct b3JacobiSolverInfo
-{
- int m_fixedBodyIndex;
-
- float m_deltaTime;
- float m_positionDrift;
- float m_positionConstraintCoeff;
- int m_numIterations;
-
- b3JacobiSolverInfo()
- :m_fixedBodyIndex(0),
- m_deltaTime(1./60.f),
- m_positionDrift( 0.005f ),
- m_positionConstraintCoeff( 0.99f ),
- m_numIterations(7)
- {
- }
-};
-class b3GpuJacobiContactSolver
-{
-protected:
-
- struct b3GpuJacobiSolverInternalData* m_data;
-
- cl_context m_context;
- cl_device_id m_device;
- cl_command_queue m_queue;
-
-public:
-
- b3GpuJacobiContactSolver(cl_context ctx, cl_device_id device, cl_command_queue queue, int pairCapacity);
- virtual ~b3GpuJacobiContactSolver();
-
-
- void solveContacts(int numBodies, cl_mem bodyBuf, cl_mem inertiaBuf, int numContacts, cl_mem contactBuf, const struct b3Config& config, int static0Index);
- void solveGroupHost(b3RigidBodyData* bodies,b3InertiaData* inertias,int numBodies,struct b3Contact4* manifoldPtr, int numManifolds,const b3JacobiSolverInfo& solverInfo);
- //void solveGroupHost(btRigidBodyCL* bodies,b3InertiaData* inertias,int numBodies,btContact4* manifoldPtr, int numManifolds,btTypedConstraint** constraints,int numConstraints,const btJacobiSolverInfo& solverInfo);
-
- //b3Scalar solveGroup(b3OpenCLArray<b3RigidBodyData>* gpuBodies,b3OpenCLArray<b3InertiaData>* gpuInertias, int numBodies,b3OpenCLArray<b3GpuGenericConstraint>* gpuConstraints,int numConstraints,const b3ContactSolverInfo& infoGlobal);
-
- //void solveGroup(btOpenCLArray<btRigidBodyCL>* bodies,btOpenCLArray<btInertiaCL>* inertias,btOpenCLArray<btContact4>* manifoldPtr,const btJacobiSolverInfo& solverInfo);
- //void solveGroupMixed(btOpenCLArray<btRigidBodyCL>* bodies,btOpenCLArray<btInertiaCL>* inertias,btOpenCLArray<btContact4>* manifoldPtr,const btJacobiSolverInfo& solverInfo);
-
-};
-#endif //B3_GPU_JACOBI_CONTACT_SOLVER_H
-
/include/Bullet/Bullet3OpenCL/RigidBody/b3GpuJacobiContactSolver.h
Property changes:
Deleted: svn:executable
## -1 +0,0 ##
-*
\ No newline at end of property
Index: include/Bullet/Bullet3OpenCL/RigidBody/b3Solver.h
===================================================================
--- include/Bullet/Bullet3OpenCL/RigidBody/b3Solver.h (revision 22)
+++ include/Bullet/Bullet3OpenCL/RigidBody/b3Solver.h (nonexistent)
@@ -1,126 +0,0 @@
-/*
-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 Takahiro Harada
-
-
-#ifndef __ADL_SOLVER_H
-#define __ADL_SOLVER_H
-
-#include "Bullet3OpenCL/ParallelPrimitives/b3OpenCLArray.h"
-#include "b3GpuConstraint4.h"
-
-#include "Bullet3Collision/NarrowPhaseCollision/shared/b3RigidBodyData.h"
-#include "Bullet3Collision/NarrowPhaseCollision/b3Contact4.h"
-
-#include "Bullet3OpenCL/ParallelPrimitives/b3PrefixScanCL.h"
-#include "Bullet3OpenCL/ParallelPrimitives/b3RadixSort32CL.h"
-#include "Bullet3OpenCL/ParallelPrimitives/b3BoundSearchCL.h"
-
-#include "Bullet3OpenCL/Initialize/b3OpenCLUtils.h"
-
-
-#define B3NEXTMULTIPLEOF(num, alignment) (((num)/(alignment) + (((num)%(alignment)==0)?0:1))*(alignment))
-
-enum
-{
- B3_SOLVER_N_SPLIT_X = 8,//16,//4,
- B3_SOLVER_N_SPLIT_Y = 4,//16,//4,
- B3_SOLVER_N_SPLIT_Z = 8,//,
- B3_SOLVER_N_CELLS = B3_SOLVER_N_SPLIT_X*B3_SOLVER_N_SPLIT_Y*B3_SOLVER_N_SPLIT_Z,
- B3_SOLVER_N_BATCHES = 8,//4,//8,//4,
- B3_MAX_NUM_BATCHES = 128,
-};
-
-class b3SolverBase
-{
- public:
-
-
- struct ConstraintCfg
- {
- ConstraintCfg( float dt = 0.f ): m_positionDrift( 0.005f ), m_positionConstraintCoeff( 0.2f ), m_dt(dt), m_staticIdx(-1) {}
-
- float m_positionDrift;
- float m_positionConstraintCoeff;
- float m_dt;
- bool m_enableParallelSolve;
- float m_batchCellSize;
- int m_staticIdx;
- };
-
-};
-
-class b3Solver : public b3SolverBase
-{
- public:
-
- cl_context m_context;
- cl_device_id m_device;
- cl_command_queue m_queue;
-
-
- b3OpenCLArray<unsigned int>* m_numConstraints;
- b3OpenCLArray<unsigned int>* m_offsets;
- b3OpenCLArray<int> m_batchSizes;
-
-
- int m_nIterations;
- cl_kernel m_batchingKernel;
- cl_kernel m_batchingKernelNew;
- cl_kernel m_solveContactKernel;
- cl_kernel m_solveFrictionKernel;
- cl_kernel m_contactToConstraintKernel;
- cl_kernel m_setSortDataKernel;
- cl_kernel m_reorderContactKernel;
- cl_kernel m_copyConstraintKernel;
-
- class b3RadixSort32CL* m_sort32;
- class b3BoundSearchCL* m_search;
- class b3PrefixScanCL* m_scan;
-
- b3OpenCLArray<b3SortData>* m_sortDataBuffer;
- b3OpenCLArray<b3Contact4>* m_contactBuffer2;
-
- enum
- {
- DYNAMIC_CONTACT_ALLOCATION_THRESHOLD = 2000000,
- };
-
-
-
-
- b3Solver(cl_context ctx, cl_device_id device, cl_command_queue queue, int pairCapacity);
-
- virtual ~b3Solver();
-
- void solveContactConstraint( const b3OpenCLArray<b3RigidBodyData>* bodyBuf, const b3OpenCLArray<b3InertiaData>* inertiaBuf,
- b3OpenCLArray<b3GpuConstraint4>* constraint, void* additionalData, int n ,int maxNumBatches);
-
- void solveContactConstraintHost( b3OpenCLArray<b3RigidBodyData>* bodyBuf, b3OpenCLArray<b3InertiaData>* shapeBuf,
- b3OpenCLArray<b3GpuConstraint4>* constraint, void* additionalData, int n ,int maxNumBatches, b3AlignedObjectArray<int>* batchSizes);
-
-
- void convertToConstraints( const b3OpenCLArray<b3RigidBodyData>* bodyBuf,
- const b3OpenCLArray<b3InertiaData>* shapeBuf,
- b3OpenCLArray<b3Contact4>* contactsIn, b3OpenCLArray<b3GpuConstraint4>* contactCOut, void* additionalData,
- int nContacts, const ConstraintCfg& cfg );
-
- void batchContacts( b3OpenCLArray<b3Contact4>* contacts, int nContacts, b3OpenCLArray<unsigned int>* n, b3OpenCLArray<unsigned int>* offsets, int staticIdx );
-
-};
-
-
-
-
-#endif //__ADL_SOLVER_H
/include/Bullet/Bullet3OpenCL/RigidBody/b3Solver.h
Property changes:
Deleted: svn:executable
## -1 +0,0 ##
-*
\ No newline at end of property
Index: include/Bullet/Bullet3OpenCL/RigidBody/b3GpuSolverConstraint.h
===================================================================
--- include/Bullet/Bullet3OpenCL/RigidBody/b3GpuSolverConstraint.h (revision 22)
+++ include/Bullet/Bullet3OpenCL/RigidBody/b3GpuSolverConstraint.h (nonexistent)
@@ -1,82 +0,0 @@
-/*
-Bullet Continuous Collision Detection and Physics Library
-Copyright (c) 2013 Erwin Coumans http://github.com/erwincoumans/bullet3
-
-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.
-*/
-
-
-#ifndef B3_GPU_SOLVER_CONSTRAINT_H
-#define B3_GPU_SOLVER_CONSTRAINT_H
-
-
-#include "Bullet3Common/b3Vector3.h"
-#include "Bullet3Common/b3Matrix3x3.h"
-//#include "b3JacobianEntry.h"
-#include "Bullet3Common/b3AlignedObjectArray.h"
-
-//#define NO_FRICTION_TANGENTIALS 1
-
-
-
-///1D constraint along a normal axis between bodyA and bodyB. It can be combined to solve contact and friction constraints.
-B3_ATTRIBUTE_ALIGNED16 (struct) b3GpuSolverConstraint
-{
- B3_DECLARE_ALIGNED_ALLOCATOR();
-
- b3Vector3 m_relpos1CrossNormal;
- b3Vector3 m_contactNormal;
-
- b3Vector3 m_relpos2CrossNormal;
- //b3Vector3 m_contactNormal2;//usually m_contactNormal2 == -m_contactNormal
-
- b3Vector3 m_angularComponentA;
- b3Vector3 m_angularComponentB;
-
- mutable b3Scalar m_appliedPushImpulse;
- mutable b3Scalar m_appliedImpulse;
- int m_padding1;
- int m_padding2;
- b3Scalar m_friction;
- b3Scalar m_jacDiagABInv;
- b3Scalar m_rhs;
- b3Scalar m_cfm;
-
- b3Scalar m_lowerLimit;
- b3Scalar m_upperLimit;
- b3Scalar m_rhsPenetration;
- union
- {
- void* m_originalContactPoint;
- int m_originalConstraintIndex;
- b3Scalar m_unusedPadding4;
- };
-
- int m_overrideNumSolverIterations;
- int m_frictionIndex;
- int m_solverBodyIdA;
- int m_solverBodyIdB;
-
-
- enum b3SolverConstraintType
- {
- B3_SOLVER_CONTACT_1D = 0,
- B3_SOLVER_FRICTION_1D
- };
-};
-
-typedef b3AlignedObjectArray<b3GpuSolverConstraint> b3GpuConstraintArray;
-
-
-#endif //B3_GPU_SOLVER_CONSTRAINT_H
-
-
-
/include/Bullet/Bullet3OpenCL/RigidBody/b3GpuSolverConstraint.h
Property changes:
Deleted: svn:executable
## -1 +0,0 ##
-*
\ No newline at end of property
Index: include/Bullet/Bullet3OpenCL/Initialize/b3OpenCLInclude.h
===================================================================
--- include/Bullet/Bullet3OpenCL/Initialize/b3OpenCLInclude.h (revision 22)
+++ include/Bullet/Bullet3OpenCL/Initialize/b3OpenCLInclude.h (nonexistent)
@@ -1,48 +0,0 @@
-/*
-Bullet Continuous Collision Detection and Physics Library
-Copyright (c) 2011 Advanced Micro Devices, Inc. http://bulletphysics.org
-
-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.
-*/
-
-#ifndef B3_OPENCL_INCLUDE_H
-#define B3_OPENCL_INCLUDE_H
-
-#ifdef B3_USE_CLEW
- #include "clew/clew.h"
-#else
-
-#ifdef __APPLE__
-#ifdef USE_MINICL
-#include <MiniCL/cl.h>
-#else
-#include <OpenCL/cl.h>
-#include <OpenCL/cl_ext.h> //clLogMessagesToStderrAPPLE
-#endif
-#else
-#ifdef USE_MINICL
-#include <MiniCL/cl.h>
-#else
-#include <CL/cl.h>
-#ifdef _WIN32
-#include "CL/cl_gl.h"
-#endif //_WIN32
-#endif
-#endif //__APPLE__
-#endif //B3_USE_CLEW
-
-#include <assert.h>
-#include <stdio.h>
-#define oclCHECKERROR(a, b) if((a)!=(b)) { printf("OCL Error : %d\n", (a)); assert((a) == (b)); }
-
-
-#endif //B3_OPENCL_INCLUDE_H
-
/include/Bullet/Bullet3OpenCL/Initialize/b3OpenCLInclude.h
Property changes:
Deleted: svn:executable
## -1 +0,0 ##
-*
\ No newline at end of property
Index: include/Bullet/Bullet3OpenCL/Initialize/b3OpenCLUtils.h
===================================================================
--- include/Bullet/Bullet3OpenCL/Initialize/b3OpenCLUtils.h (revision 22)
+++ include/Bullet/Bullet3OpenCL/Initialize/b3OpenCLUtils.h (nonexistent)
@@ -1,194 +0,0 @@
-/*
-Bullet Continuous Collision Detection and Physics Library, http://bulletphysics.org
-Copyright (C) 2006 - 2011 Sony Computer Entertainment 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.
-*/
-
-//original author: Roman Ponomarev
-//cleanup by Erwin Coumans
-
-#ifndef B3_OPENCL_UTILS_H
-#define B3_OPENCL_UTILS_H
-
-#include "b3OpenCLInclude.h"
-
-#ifdef __cplusplus
-extern "C" {
-#endif
-
-
-///C API for OpenCL utilities: convenience functions, see below for C++ API
-
-/// CL Context optionally takes a GL context. This is a generic type because we don't really want this code
-/// to have to understand GL types. It is a HGLRC in _WIN32 or a GLXContext otherwise.
-cl_context b3OpenCLUtils_createContextFromType(cl_device_type deviceType, cl_int* pErrNum, void* pGLCtx , void* pGLDC , int preferredDeviceIndex , int preferredPlatformIndex, cl_platform_id* platformId);
-
-int b3OpenCLUtils_getNumDevices(cl_context cxMainContext);
-
-cl_device_id b3OpenCLUtils_getDevice(cl_context cxMainContext, int nr);
-
-void b3OpenCLUtils_printDeviceInfo(cl_device_id device);
-
-cl_kernel b3OpenCLUtils_compileCLKernelFromString( cl_context clContext,cl_device_id device, const char* kernelSource, const char* kernelName, cl_int* pErrNum, cl_program prog,const char* additionalMacros);
-
-//optional
-cl_program b3OpenCLUtils_compileCLProgramFromString( cl_context clContext,cl_device_id device, const char* kernelSource, cl_int* pErrNum,const char* additionalMacros , const char* srcFileNameForCaching, bool disableBinaryCaching);
-
-//the following optional APIs provide access using specific platform information
-int b3OpenCLUtils_getNumPlatforms(cl_int* pErrNum);
-
-///get the nr'th platform, where nr is in the range [0..getNumPlatforms)
-cl_platform_id b3OpenCLUtils_getPlatform(int nr, cl_int* pErrNum);
-
-
-void b3OpenCLUtils_printPlatformInfo(cl_platform_id platform);
-
-const char* b3OpenCLUtils_getSdkVendorName();
-
-///set the path (directory/folder) where the compiled OpenCL kernel are stored
-void b3OpenCLUtils_setCachePath(const char* path);
-
-cl_context b3OpenCLUtils_createContextFromPlatform(cl_platform_id platform, cl_device_type deviceType, cl_int* pErrNum, void* pGLCtx , void* pGLDC ,int preferredDeviceIndex , int preferredPlatformIndex);
-
-#ifdef __cplusplus
-}
-
-#define B3_MAX_STRING_LENGTH 1024
-
-typedef struct
-{
- char m_deviceName[B3_MAX_STRING_LENGTH];
- char m_deviceVendor[B3_MAX_STRING_LENGTH];
- char m_driverVersion[B3_MAX_STRING_LENGTH];
- char m_deviceExtensions[B3_MAX_STRING_LENGTH];
-
- cl_device_type m_deviceType;
- cl_uint m_computeUnits;
- size_t m_workitemDims;
- size_t m_workItemSize[3];
- size_t m_image2dMaxWidth;
- size_t m_image2dMaxHeight;
- size_t m_image3dMaxWidth;
- size_t m_image3dMaxHeight;
- size_t m_image3dMaxDepth;
- size_t m_workgroupSize;
- cl_uint m_clockFrequency;
- cl_ulong m_constantBufferSize;
- cl_ulong m_localMemSize;
- cl_ulong m_globalMemSize;
- cl_bool m_errorCorrectionSupport;
- cl_device_local_mem_type m_localMemType;
- cl_uint m_maxReadImageArgs;
- cl_uint m_maxWriteImageArgs;
-
-
-
- cl_uint m_addressBits;
- cl_ulong m_maxMemAllocSize;
- cl_command_queue_properties m_queueProperties;
- cl_bool m_imageSupport;
- cl_uint m_vecWidthChar;
- cl_uint m_vecWidthShort;
- cl_uint m_vecWidthInt;
- cl_uint m_vecWidthLong;
- cl_uint m_vecWidthFloat;
- cl_uint m_vecWidthDouble;
-
-} b3OpenCLDeviceInfo;
-
-struct b3OpenCLPlatformInfo
-{
- char m_platformVendor[B3_MAX_STRING_LENGTH];
- char m_platformName[B3_MAX_STRING_LENGTH];
- char m_platformVersion[B3_MAX_STRING_LENGTH];
-
- b3OpenCLPlatformInfo()
- {
- m_platformVendor[0]=0;
- m_platformName[0]=0;
- m_platformVersion[0]=0;
- }
-};
-
-
-///C++ API for OpenCL utilities: convenience functions
-struct b3OpenCLUtils
-{
- /// CL Context optionally takes a GL context. This is a generic type because we don't really want this code
- /// to have to understand GL types. It is a HGLRC in _WIN32 or a GLXContext otherwise.
- static inline cl_context createContextFromType(cl_device_type deviceType, cl_int* pErrNum, void* pGLCtx = 0, void* pGLDC = 0, int preferredDeviceIndex = -1, int preferredPlatformIndex= - 1, cl_platform_id* platformId=0)
- {
- return b3OpenCLUtils_createContextFromType(deviceType, pErrNum, pGLCtx , pGLDC , preferredDeviceIndex, preferredPlatformIndex, platformId);
- }
-
- static inline int getNumDevices(cl_context cxMainContext)
- {
- return b3OpenCLUtils_getNumDevices(cxMainContext);
- }
- static inline cl_device_id getDevice(cl_context cxMainContext, int nr)
- {
- return b3OpenCLUtils_getDevice(cxMainContext,nr);
- }
-
- static void getDeviceInfo(cl_device_id device, b3OpenCLDeviceInfo* info);
-
- static inline void printDeviceInfo(cl_device_id device)
- {
- b3OpenCLUtils_printDeviceInfo(device);
- }
-
- static inline cl_kernel compileCLKernelFromString( cl_context clContext,cl_device_id device, const char* kernelSource, const char* kernelName, cl_int* pErrNum=0, cl_program prog=0,const char* additionalMacros = "" )
- {
- return b3OpenCLUtils_compileCLKernelFromString(clContext,device, kernelSource, kernelName, pErrNum, prog,additionalMacros);
- }
-
- //optional
- static inline cl_program compileCLProgramFromString( cl_context clContext,cl_device_id device, const char* kernelSource, cl_int* pErrNum=0,const char* additionalMacros = "" , const char* srcFileNameForCaching=0, bool disableBinaryCaching=false)
- {
- return b3OpenCLUtils_compileCLProgramFromString(clContext,device, kernelSource, pErrNum,additionalMacros, srcFileNameForCaching, disableBinaryCaching);
- }
-
- //the following optional APIs provide access using specific platform information
- static inline int getNumPlatforms(cl_int* pErrNum=0)
- {
- return b3OpenCLUtils_getNumPlatforms(pErrNum);
- }
- ///get the nr'th platform, where nr is in the range [0..getNumPlatforms)
- static inline cl_platform_id getPlatform(int nr, cl_int* pErrNum=0)
- {
- return b3OpenCLUtils_getPlatform(nr,pErrNum);
- }
-
- static void getPlatformInfo(cl_platform_id platform, b3OpenCLPlatformInfo* platformInfo);
-
- static inline void printPlatformInfo(cl_platform_id platform)
- {
- b3OpenCLUtils_printPlatformInfo(platform);
- }
-
- static inline const char* getSdkVendorName()
- {
- return b3OpenCLUtils_getSdkVendorName();
- }
- static inline cl_context createContextFromPlatform(cl_platform_id platform, cl_device_type deviceType, cl_int* pErrNum, void* pGLCtx = 0, void* pGLDC = 0,int preferredDeviceIndex = -1, int preferredPlatformIndex= -1)
- {
- return b3OpenCLUtils_createContextFromPlatform(platform, deviceType, pErrNum, pGLCtx,pGLDC,preferredDeviceIndex, preferredPlatformIndex);
- }
- static void setCachePath(const char* path)
- {
- b3OpenCLUtils_setCachePath(path);
- }
-};
-
-#endif //__cplusplus
-
-#endif // B3_OPENCL_UTILS_H
/include/Bullet/Bullet3OpenCL/Initialize/b3OpenCLUtils.h
Property changes:
Deleted: svn:executable
## -1 +0,0 ##
-*
\ No newline at end of property
Index: include/Bullet/Bullet3OpenCL/BroadphaseCollision/b3GpuSapBroadphase.h
===================================================================
--- include/Bullet/Bullet3OpenCL/BroadphaseCollision/b3GpuSapBroadphase.h (revision 22)
+++ include/Bullet/Bullet3OpenCL/BroadphaseCollision/b3GpuSapBroadphase.h (nonexistent)
@@ -1,151 +0,0 @@
-#ifndef B3_GPU_SAP_BROADPHASE_H
-#define B3_GPU_SAP_BROADPHASE_H
-
-#include "Bullet3OpenCL/ParallelPrimitives/b3OpenCLArray.h"
-#include "Bullet3OpenCL/ParallelPrimitives/b3FillCL.h" //b3Int2
-class b3Vector3;
-#include "Bullet3OpenCL/ParallelPrimitives/b3RadixSort32CL.h"
-
-#include "b3SapAabb.h"
-#include "Bullet3Common/shared/b3Int2.h"
-
-#include "b3GpuBroadphaseInterface.h"
-
-
-class b3GpuSapBroadphase : public b3GpuBroadphaseInterface
-{
-
- cl_context m_context;
- cl_device_id m_device;
- cl_command_queue m_queue;
- cl_kernel m_flipFloatKernel;
- cl_kernel m_scatterKernel ;
- cl_kernel m_copyAabbsKernel;
- cl_kernel m_sapKernel;
- cl_kernel m_sap2Kernel;
- cl_kernel m_prepareSumVarianceKernel;
-
-
- class b3RadixSort32CL* m_sorter;
-
- ///test for 3d SAP
- b3AlignedObjectArray<b3SortData> m_sortedAxisCPU[3][2];
- b3AlignedObjectArray<b3UnsignedInt2> m_objectMinMaxIndexCPU[3][2];
- b3OpenCLArray<b3UnsignedInt2> m_objectMinMaxIndexGPUaxis0;
- b3OpenCLArray<b3UnsignedInt2> m_objectMinMaxIndexGPUaxis1;
- b3OpenCLArray<b3UnsignedInt2> m_objectMinMaxIndexGPUaxis2;
- b3OpenCLArray<b3UnsignedInt2> m_objectMinMaxIndexGPUaxis0prev;
- b3OpenCLArray<b3UnsignedInt2> m_objectMinMaxIndexGPUaxis1prev;
- b3OpenCLArray<b3UnsignedInt2> m_objectMinMaxIndexGPUaxis2prev;
-
- b3OpenCLArray<b3SortData> m_sortedAxisGPU0;
- b3OpenCLArray<b3SortData> m_sortedAxisGPU1;
- b3OpenCLArray<b3SortData> m_sortedAxisGPU2;
- b3OpenCLArray<b3SortData> m_sortedAxisGPU0prev;
- b3OpenCLArray<b3SortData> m_sortedAxisGPU1prev;
- b3OpenCLArray<b3SortData> m_sortedAxisGPU2prev;
-
-
- b3OpenCLArray<b3Int4> m_addedHostPairsGPU;
- b3OpenCLArray<b3Int4> m_removedHostPairsGPU;
- b3OpenCLArray<int> m_addedCountGPU;
- b3OpenCLArray<int> m_removedCountGPU;
-
- int m_currentBuffer;
-
-public:
-
- b3OpenCLArray<int> m_pairCount;
-
-
- b3OpenCLArray<b3SapAabb> m_allAabbsGPU;
- b3AlignedObjectArray<b3SapAabb> m_allAabbsCPU;
-
- virtual b3OpenCLArray<b3SapAabb>& getAllAabbsGPU()
- {
- return m_allAabbsGPU;
- }
- virtual b3AlignedObjectArray<b3SapAabb>& getAllAabbsCPU()
- {
- return m_allAabbsCPU;
- }
-
- b3OpenCLArray<b3Vector3> m_sum;
- b3OpenCLArray<b3Vector3> m_sum2;
- b3OpenCLArray<b3Vector3> m_dst;
-
- b3OpenCLArray<int> m_smallAabbsMappingGPU;
- b3AlignedObjectArray<int> m_smallAabbsMappingCPU;
-
- b3OpenCLArray<int> m_largeAabbsMappingGPU;
- b3AlignedObjectArray<int> m_largeAabbsMappingCPU;
-
-
- b3OpenCLArray<b3Int4> m_overlappingPairs;
-
- //temporary gpu work memory
- b3OpenCLArray<b3SortData> m_gpuSmallSortData;
- b3OpenCLArray<b3SapAabb> m_gpuSmallSortedAabbs;
-
- class b3PrefixScanFloat4CL* m_prefixScanFloat4;
-
- enum b3GpuSapKernelType
- {
- B3_GPU_SAP_KERNEL_BRUTE_FORCE_CPU=1,
- B3_GPU_SAP_KERNEL_BRUTE_FORCE_GPU,
- B3_GPU_SAP_KERNEL_ORIGINAL,
- B3_GPU_SAP_KERNEL_BARRIER,
- B3_GPU_SAP_KERNEL_LOCAL_SHARED_MEMORY
- };
-
- b3GpuSapBroadphase(cl_context ctx,cl_device_id device, cl_command_queue q , b3GpuSapKernelType kernelType=B3_GPU_SAP_KERNEL_LOCAL_SHARED_MEMORY);
- virtual ~b3GpuSapBroadphase();
-
- static b3GpuBroadphaseInterface* CreateFuncBruteForceCpu(cl_context ctx,cl_device_id device, cl_command_queue q)
- {
- return new b3GpuSapBroadphase(ctx,device,q,B3_GPU_SAP_KERNEL_BRUTE_FORCE_CPU);
- }
-
- static b3GpuBroadphaseInterface* CreateFuncBruteForceGpu(cl_context ctx,cl_device_id device, cl_command_queue q)
- {
- return new b3GpuSapBroadphase(ctx,device,q,B3_GPU_SAP_KERNEL_BRUTE_FORCE_GPU);
- }
-
- static b3GpuBroadphaseInterface* CreateFuncOriginal(cl_context ctx,cl_device_id device, cl_command_queue q)
- {
- return new b3GpuSapBroadphase(ctx,device,q,B3_GPU_SAP_KERNEL_ORIGINAL);
- }
- static b3GpuBroadphaseInterface* CreateFuncBarrier(cl_context ctx,cl_device_id device, cl_command_queue q)
- {
- return new b3GpuSapBroadphase(ctx,device,q,B3_GPU_SAP_KERNEL_BARRIER);
- }
- static b3GpuBroadphaseInterface* CreateFuncLocalMemory(cl_context ctx,cl_device_id device, cl_command_queue q)
- {
- return new b3GpuSapBroadphase(ctx,device,q,B3_GPU_SAP_KERNEL_LOCAL_SHARED_MEMORY);
- }
-
-
- virtual void calculateOverlappingPairs(int maxPairs);
- virtual void calculateOverlappingPairsHost(int maxPairs);
-
- void reset();
-
- void init3dSap();
- virtual void calculateOverlappingPairsHostIncremental3Sap();
-
- virtual void createProxy(const b3Vector3& aabbMin, const b3Vector3& aabbMax, int userPtr ,short int collisionFilterGroup,short int collisionFilterMask);
- virtual void createLargeProxy(const b3Vector3& aabbMin, const b3Vector3& aabbMax, int userPtr ,short int collisionFilterGroup,short int collisionFilterMask);
-
- //call writeAabbsToGpu after done making all changes (createProxy etc)
- virtual void writeAabbsToGpu();
-
- virtual cl_mem getAabbBufferWS();
- virtual int getNumOverlap();
- virtual cl_mem getOverlappingPairBuffer();
-
- virtual b3OpenCLArray<b3Int4>& getOverlappingPairsGPU();
- virtual b3OpenCLArray<int>& getSmallAabbIndicesGPU();
- virtual b3OpenCLArray<int>& getLargeAabbIndicesGPU();
-};
-
-#endif //B3_GPU_SAP_BROADPHASE_H
\ No newline at end of file
/include/Bullet/Bullet3OpenCL/BroadphaseCollision/b3GpuSapBroadphase.h
Property changes:
Deleted: svn:executable
## -1 +0,0 ##
-*
\ No newline at end of property
Index: include/Bullet/Bullet3OpenCL/BroadphaseCollision/kernels/parallelLinearBvhKernels.h
===================================================================
--- include/Bullet/Bullet3OpenCL/BroadphaseCollision/kernels/parallelLinearBvhKernels.h (revision 22)
+++ include/Bullet/Bullet3OpenCL/BroadphaseCollision/kernels/parallelLinearBvhKernels.h (nonexistent)
@@ -1,729 +0,0 @@
-//this file is autogenerated using stringify.bat (premake --stringify) in the build folder of this project
-static const char* parallelLinearBvhCL= \
-"/*\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"
-"//Initial Author Jackson Lee, 2014\n"
-"typedef float b3Scalar;\n"
-"typedef float4 b3Vector3;\n"
-"#define b3Max max\n"
-"#define b3Min min\n"
-"#define b3Sqrt sqrt\n"
-"typedef struct\n"
-"{\n"
-" unsigned int m_key;\n"
-" unsigned int m_value;\n"
-"} SortDataCL;\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"
-"unsigned int interleaveBits(unsigned int x)\n"
-"{\n"
-" //........ ........ ......12 3456789A //x\n"
-" //....1..2 ..3..4.. 5..6..7. .8..9..A //x after interleaving bits\n"
-" \n"
-" //......12 3456789A ......12 3456789A //x ^ (x << 16)\n"
-" //11111111 ........ ........ 11111111 //0x FF 00 00 FF\n"
-" //......12 ........ ........ 3456789A //x = (x ^ (x << 16)) & 0xFF0000FF;\n"
-" \n"
-" //......12 ........ 3456789A 3456789A //x ^ (x << 8)\n"
-" //......11 ........ 1111.... ....1111 //0x 03 00 F0 0F\n"
-" //......12 ........ 3456.... ....789A //x = (x ^ (x << 8)) & 0x0300F00F;\n"
-" \n"
-" //..12..12 ....3456 3456.... 789A789A //x ^ (x << 4)\n"
-" //......11 ....11.. ..11.... 11....11 //0x 03 0C 30 C3\n"
-" //......12 ....34.. ..56.... 78....9A //x = (x ^ (x << 4)) & 0x030C30C3;\n"
-" \n"
-" //....1212 ..3434.. 5656..78 78..9A9A //x ^ (x << 2)\n"
-" //....1..1 ..1..1.. 1..1..1. .1..1..1 //0x 09 24 92 49\n"
-" //....1..2 ..3..4.. 5..6..7. .8..9..A //x = (x ^ (x << 2)) & 0x09249249;\n"
-" \n"
-" //........ ........ ......11 11111111 //0x000003FF\n"
-" x &= 0x000003FF; //Clear all bits above bit 10\n"
-" \n"
-" x = (x ^ (x << 16)) & 0xFF0000FF;\n"
-" x = (x ^ (x << 8)) & 0x0300F00F;\n"
-" x = (x ^ (x << 4)) & 0x030C30C3;\n"
-" x = (x ^ (x << 2)) & 0x09249249;\n"
-" \n"
-" return x;\n"
-"}\n"
-"unsigned int getMortonCode(unsigned int x, unsigned int y, unsigned int z)\n"
-"{\n"
-" return interleaveBits(x) << 0 | interleaveBits(y) << 1 | interleaveBits(z) << 2;\n"
-"}\n"
-"__kernel void separateAabbs(__global b3AabbCL* unseparatedAabbs, __global int* aabbIndices, __global b3AabbCL* out_aabbs, int numAabbsToSeparate)\n"
-"{\n"
-" int separatedAabbIndex = get_global_id(0);\n"
-" if(separatedAabbIndex >= numAabbsToSeparate) return;\n"
-" int unseparatedAabbIndex = aabbIndices[separatedAabbIndex];\n"
-" out_aabbs[separatedAabbIndex] = unseparatedAabbs[unseparatedAabbIndex];\n"
-"}\n"
-"//Should replace with an optimized parallel reduction\n"
-"__kernel void findAllNodesMergedAabb(__global b3AabbCL* out_mergedAabb, int numAabbsNeedingMerge)\n"
-"{\n"
-" //Each time this kernel is added to the command queue, \n"
-" //the number of AABBs needing to be merged is halved\n"
-" //\n"
-" //Example with 159 AABBs:\n"
-" // numRemainingAabbs == 159 / 2 + 159 % 2 == 80\n"
-" // numMergedAabbs == 159 - 80 == 79\n"
-" //So, indices [0, 78] are merged with [0 + 80, 78 + 80]\n"
-" \n"
-" int numRemainingAabbs = numAabbsNeedingMerge / 2 + numAabbsNeedingMerge % 2;\n"
-" int numMergedAabbs = numAabbsNeedingMerge - numRemainingAabbs;\n"
-" \n"
-" int aabbIndex = get_global_id(0);\n"
-" if(aabbIndex >= numMergedAabbs) return;\n"
-" \n"
-" int otherAabbIndex = aabbIndex + numRemainingAabbs;\n"
-" \n"
-" b3AabbCL aabb = out_mergedAabb[aabbIndex];\n"
-" b3AabbCL otherAabb = out_mergedAabb[otherAabbIndex];\n"
-" \n"
-" b3AabbCL mergedAabb;\n"
-" mergedAabb.m_min = b3Min(aabb.m_min, otherAabb.m_min);\n"
-" mergedAabb.m_max = b3Max(aabb.m_max, otherAabb.m_max);\n"
-" out_mergedAabb[aabbIndex] = mergedAabb;\n"
-"}\n"
-"__kernel void assignMortonCodesAndAabbIndicies(__global b3AabbCL* worldSpaceAabbs, __global b3AabbCL* mergedAabbOfAllNodes, \n"
-" __global SortDataCL* out_mortonCodesAndAabbIndices, int numAabbs)\n"
-"{\n"
-" int leafNodeIndex = get_global_id(0); //Leaf node index == AABB index\n"
-" if(leafNodeIndex >= numAabbs) return;\n"
-" \n"
-" b3AabbCL mergedAabb = mergedAabbOfAllNodes[0];\n"
-" b3Vector3 gridCenter = (mergedAabb.m_min + mergedAabb.m_max) * 0.5f;\n"
-" b3Vector3 gridCellSize = (mergedAabb.m_max - mergedAabb.m_min) / (float)1024;\n"
-" \n"
-" b3AabbCL aabb = worldSpaceAabbs[leafNodeIndex];\n"
-" b3Vector3 aabbCenter = (aabb.m_min + aabb.m_max) * 0.5f;\n"
-" b3Vector3 aabbCenterRelativeToGrid = aabbCenter - gridCenter;\n"
-" \n"
-" //Quantize into integer coordinates\n"
-" //floor() is needed to prevent the center cell, at (0,0,0) from being twice the size\n"
-" b3Vector3 gridPosition = aabbCenterRelativeToGrid / gridCellSize;\n"
-" \n"
-" int4 discretePosition;\n"
-" discretePosition.x = (int)( (gridPosition.x >= 0.0f) ? gridPosition.x : floor(gridPosition.x) );\n"
-" discretePosition.y = (int)( (gridPosition.y >= 0.0f) ? gridPosition.y : floor(gridPosition.y) );\n"
-" discretePosition.z = (int)( (gridPosition.z >= 0.0f) ? gridPosition.z : floor(gridPosition.z) );\n"
-" \n"
-" //Clamp coordinates into [-512, 511], then convert range from [-512, 511] to [0, 1023]\n"
-" discretePosition = b3Max( -512, b3Min(discretePosition, 511) );\n"
-" discretePosition += 512;\n"
-" \n"
-" //Interleave bits(assign a morton code, also known as a z-curve)\n"
-" unsigned int mortonCode = getMortonCode(discretePosition.x, discretePosition.y, discretePosition.z);\n"
-" \n"
-" //\n"
-" SortDataCL mortonCodeIndexPair;\n"
-" mortonCodeIndexPair.m_key = mortonCode;\n"
-" mortonCodeIndexPair.m_value = leafNodeIndex;\n"
-" \n"
-" out_mortonCodesAndAabbIndices[leafNodeIndex] = mortonCodeIndexPair;\n"
-"}\n"
-"#define B3_PLVBH_TRAVERSE_MAX_STACK_SIZE 128\n"
-"//The most significant bit(0x80000000) of a int32 is used to distinguish between leaf and internal nodes.\n"
-"//If it is set, then the index is for an internal node; otherwise, it is a leaf node. \n"
-"//In both cases, the bit should be cleared to access the actual node index.\n"
-"int isLeafNode(int index) { return (index >> 31 == 0); }\n"
-"int getIndexWithInternalNodeMarkerRemoved(int index) { return index & (~0x80000000); }\n"
-"int getIndexWithInternalNodeMarkerSet(int isLeaf, int index) { return (isLeaf) ? index : (index | 0x80000000); }\n"
-"//From sap.cl\n"
-"#define NEW_PAIR_MARKER -1\n"
-"bool TestAabbAgainstAabb2(const b3AabbCL* aabb1, 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"
-"//From sap.cl\n"
-"__kernel void plbvhCalculateOverlappingPairs(__global b3AabbCL* rigidAabbs, \n"
-" __global int* rootNodeIndex, \n"
-" __global int2* internalNodeChildIndices, \n"
-" __global b3AabbCL* internalNodeAabbs,\n"
-" __global int2* internalNodeLeafIndexRanges,\n"
-" \n"
-" __global SortDataCL* mortonCodesAndAabbIndices,\n"
-" __global int* out_numPairs, __global int4* out_overlappingPairs, \n"
-" int maxPairs, int numQueryAabbs)\n"
-"{\n"
-" //Using get_group_id()/get_local_id() is Faster than get_global_id(0) since\n"
-" //mortonCodesAndAabbIndices[] contains rigid body indices sorted along the z-curve (more spatially coherent)\n"
-" int queryBvhNodeIndex = get_group_id(0) * get_local_size(0) + get_local_id(0);\n"
-" if(queryBvhNodeIndex >= numQueryAabbs) return;\n"
-" \n"
-" int queryRigidIndex = mortonCodesAndAabbIndices[queryBvhNodeIndex].m_value;\n"
-" b3AabbCL queryAabb = rigidAabbs[queryRigidIndex];\n"
-" \n"
-" int stack[B3_PLVBH_TRAVERSE_MAX_STACK_SIZE];\n"
-" \n"
-" int stackSize = 1;\n"
-" stack[0] = *rootNodeIndex;\n"
-" \n"
-" while(stackSize)\n"
-" {\n"
-" int internalOrLeafNodeIndex = stack[ stackSize - 1 ];\n"
-" --stackSize;\n"
-" \n"
-" int isLeaf = isLeafNode(internalOrLeafNodeIndex); //Internal node if false\n"
-" int bvhNodeIndex = getIndexWithInternalNodeMarkerRemoved(internalOrLeafNodeIndex);\n"
-" \n"
-" //Optimization - if the BVH is structured as a binary radix tree, then\n"
-" //each internal node corresponds to a contiguous range of leaf nodes(internalNodeLeafIndexRanges[]).\n"
-" //This can be used to avoid testing each AABB-AABB pair twice, including preventing each node from colliding with itself.\n"
-" {\n"
-" int highestLeafIndex = (isLeaf) ? bvhNodeIndex : internalNodeLeafIndexRanges[bvhNodeIndex].y;\n"
-" if(highestLeafIndex <= queryBvhNodeIndex) continue;\n"
-" }\n"
-" \n"
-" //bvhRigidIndex is not used if internal node\n"
-" int bvhRigidIndex = (isLeaf) ? mortonCodesAndAabbIndices[bvhNodeIndex].m_value : -1;\n"
-" \n"
-" b3AabbCL bvhNodeAabb = (isLeaf) ? rigidAabbs[bvhRigidIndex] : internalNodeAabbs[bvhNodeIndex];\n"
-" if( TestAabbAgainstAabb2(&queryAabb, &bvhNodeAabb) )\n"
-" {\n"
-" if(isLeaf)\n"
-" {\n"
-" int4 pair;\n"
-" pair.x = rigidAabbs[queryRigidIndex].m_minIndices[3];\n"
-" pair.y = rigidAabbs[bvhRigidIndex].m_minIndices[3];\n"
-" pair.z = NEW_PAIR_MARKER;\n"
-" pair.w = NEW_PAIR_MARKER;\n"
-" \n"
-" int pairIndex = atomic_inc(out_numPairs);\n"
-" if(pairIndex < maxPairs) out_overlappingPairs[pairIndex] = pair;\n"
-" }\n"
-" \n"
-" if(!isLeaf) //Internal node\n"
-" {\n"
-" if(stackSize + 2 > B3_PLVBH_TRAVERSE_MAX_STACK_SIZE)\n"
-" {\n"
-" //Error\n"
-" }\n"
-" else\n"
-" {\n"
-" stack[ stackSize++ ] = internalNodeChildIndices[bvhNodeIndex].x;\n"
-" stack[ stackSize++ ] = internalNodeChildIndices[bvhNodeIndex].y;\n"
-" }\n"
-" }\n"
-" }\n"
-" \n"
-" }\n"
-"}\n"
-"//From rayCastKernels.cl\n"
-"typedef struct\n"
-"{\n"
-" float4 m_from;\n"
-" float4 m_to;\n"
-"} b3RayInfo;\n"
-"//From rayCastKernels.cl\n"
-"b3Vector3 b3Vector3_normalize(b3Vector3 v)\n"
-"{\n"
-" b3Vector3 normal = (b3Vector3){v.x, v.y, v.z, 0.f};\n"
-" return normalize(normal); //OpenCL normalize == vector4 normalize\n"
-"}\n"
-"b3Scalar b3Vector3_length2(b3Vector3 v) { return v.x*v.x + v.y*v.y + v.z*v.z; }\n"
-"b3Scalar b3Vector3_dot(b3Vector3 a, b3Vector3 b) { return a.x*b.x + a.y*b.y + a.z*b.z; }\n"
-"int rayIntersectsAabb(b3Vector3 rayOrigin, b3Scalar rayLength, b3Vector3 rayNormalizedDirection, b3AabbCL aabb)\n"
-"{\n"
-" //AABB is considered as 3 pairs of 2 planes( {x_min, x_max}, {y_min, y_max}, {z_min, z_max} ).\n"
-" //t_min is the point of intersection with the closer plane, t_max is the point of intersection with the farther plane.\n"
-" //\n"
-" //if (rayNormalizedDirection.x < 0.0f), then max.x will be the near plane \n"
-" //and min.x will be the far plane; otherwise, it is reversed.\n"
-" //\n"
-" //In order for there to be a collision, the t_min and t_max of each pair must overlap.\n"
-" //This can be tested for by selecting the highest t_min and lowest t_max and comparing them.\n"
-" \n"
-" int4 isNegative = isless( rayNormalizedDirection, ((b3Vector3){0.0f, 0.0f, 0.0f, 0.0f}) ); //isless(x,y) returns (x < y)\n"
-" \n"
-" //When using vector types, the select() function checks the most signficant bit, \n"
-" //but isless() sets the least significant bit.\n"
-" isNegative <<= 31;\n"
-" //select(b, a, condition) == condition ? a : b\n"
-" //When using select() with vector types, (condition[i]) is true if its most significant bit is 1\n"
-" b3Vector3 t_min = ( select(aabb.m_min, aabb.m_max, isNegative) - rayOrigin ) / rayNormalizedDirection;\n"
-" b3Vector3 t_max = ( select(aabb.m_max, aabb.m_min, isNegative) - rayOrigin ) / rayNormalizedDirection;\n"
-" \n"
-" b3Scalar t_min_final = 0.0f;\n"
-" b3Scalar t_max_final = rayLength;\n"
-" \n"
-" //Must use fmin()/fmax(); if one of the parameters is NaN, then the parameter that is not NaN is returned. \n"
-" //Behavior of min()/max() with NaNs is undefined. (See OpenCL Specification 1.2 [6.12.2] and [6.12.4])\n"
-" //Since the innermost fmin()/fmax() is always not NaN, this should never return NaN.\n"
-" t_min_final = fmax( t_min.z, fmax(t_min.y, fmax(t_min.x, t_min_final)) );\n"
-" t_max_final = fmin( t_max.z, fmin(t_max.y, fmin(t_max.x, t_max_final)) );\n"
-" \n"
-" return (t_min_final <= t_max_final);\n"
-"}\n"
-"__kernel void plbvhRayTraverse(__global b3AabbCL* rigidAabbs,\n"
-" __global int* rootNodeIndex, \n"
-" __global int2* internalNodeChildIndices, \n"
-" __global b3AabbCL* internalNodeAabbs,\n"
-" __global int2* internalNodeLeafIndexRanges,\n"
-" __global SortDataCL* mortonCodesAndAabbIndices,\n"
-" \n"
-" __global b3RayInfo* rays,\n"
-" \n"
-" __global int* out_numRayRigidPairs, \n"
-" __global int2* out_rayRigidPairs,\n"
-" int maxRayRigidPairs, int numRays)\n"
-"{\n"
-" int rayIndex = get_global_id(0);\n"
-" if(rayIndex >= numRays) return;\n"
-" \n"
-" //\n"
-" b3Vector3 rayFrom = rays[rayIndex].m_from;\n"
-" b3Vector3 rayTo = rays[rayIndex].m_to;\n"
-" b3Vector3 rayNormalizedDirection = b3Vector3_normalize(rayTo - rayFrom);\n"
-" b3Scalar rayLength = b3Sqrt( b3Vector3_length2(rayTo - rayFrom) );\n"
-" \n"
-" //\n"
-" int stack[B3_PLVBH_TRAVERSE_MAX_STACK_SIZE];\n"
-" \n"
-" int stackSize = 1;\n"
-" stack[0] = *rootNodeIndex;\n"
-" \n"
-" while(stackSize)\n"
-" {\n"
-" int internalOrLeafNodeIndex = stack[ stackSize - 1 ];\n"
-" --stackSize;\n"
-" \n"
-" int isLeaf = isLeafNode(internalOrLeafNodeIndex); //Internal node if false\n"
-" int bvhNodeIndex = getIndexWithInternalNodeMarkerRemoved(internalOrLeafNodeIndex);\n"
-" \n"
-" //bvhRigidIndex is not used if internal node\n"
-" int bvhRigidIndex = (isLeaf) ? mortonCodesAndAabbIndices[bvhNodeIndex].m_value : -1;\n"
-" \n"
-" b3AabbCL bvhNodeAabb = (isLeaf) ? rigidAabbs[bvhRigidIndex] : internalNodeAabbs[bvhNodeIndex];\n"
-" if( rayIntersectsAabb(rayFrom, rayLength, rayNormalizedDirection, bvhNodeAabb) )\n"
-" {\n"
-" if(isLeaf)\n"
-" {\n"
-" int2 rayRigidPair;\n"
-" rayRigidPair.x = rayIndex;\n"
-" rayRigidPair.y = rigidAabbs[bvhRigidIndex].m_minIndices[3];\n"
-" \n"
-" int pairIndex = atomic_inc(out_numRayRigidPairs);\n"
-" if(pairIndex < maxRayRigidPairs) out_rayRigidPairs[pairIndex] = rayRigidPair;\n"
-" }\n"
-" \n"
-" if(!isLeaf) //Internal node\n"
-" {\n"
-" if(stackSize + 2 > B3_PLVBH_TRAVERSE_MAX_STACK_SIZE)\n"
-" {\n"
-" //Error\n"
-" }\n"
-" else\n"
-" {\n"
-" stack[ stackSize++ ] = internalNodeChildIndices[bvhNodeIndex].x;\n"
-" stack[ stackSize++ ] = internalNodeChildIndices[bvhNodeIndex].y;\n"
-" }\n"
-" }\n"
-" }\n"
-" }\n"
-"}\n"
-"__kernel void plbvhLargeAabbAabbTest(__global b3AabbCL* smallAabbs, __global b3AabbCL* largeAabbs, \n"
-" __global int* out_numPairs, __global int4* out_overlappingPairs, \n"
-" int maxPairs, int numLargeAabbRigids, int numSmallAabbRigids)\n"
-"{\n"
-" int smallAabbIndex = get_global_id(0);\n"
-" if(smallAabbIndex >= numSmallAabbRigids) return;\n"
-" \n"
-" b3AabbCL smallAabb = smallAabbs[smallAabbIndex];\n"
-" for(int i = 0; i < numLargeAabbRigids; ++i)\n"
-" {\n"
-" b3AabbCL largeAabb = largeAabbs[i];\n"
-" if( TestAabbAgainstAabb2(&smallAabb, &largeAabb) )\n"
-" {\n"
-" int4 pair;\n"
-" pair.x = largeAabb.m_minIndices[3];\n"
-" pair.y = smallAabb.m_minIndices[3];\n"
-" pair.z = NEW_PAIR_MARKER;\n"
-" pair.w = NEW_PAIR_MARKER;\n"
-" \n"
-" int pairIndex = atomic_inc(out_numPairs);\n"
-" if(pairIndex < maxPairs) out_overlappingPairs[pairIndex] = pair;\n"
-" }\n"
-" }\n"
-"}\n"
-"__kernel void plbvhLargeAabbRayTest(__global b3AabbCL* largeRigidAabbs, __global b3RayInfo* rays,\n"
-" __global int* out_numRayRigidPairs, __global int2* out_rayRigidPairs,\n"
-" int numLargeAabbRigids, int maxRayRigidPairs, int numRays)\n"
-"{\n"
-" int rayIndex = get_global_id(0);\n"
-" if(rayIndex >= numRays) return;\n"
-" \n"
-" b3Vector3 rayFrom = rays[rayIndex].m_from;\n"
-" b3Vector3 rayTo = rays[rayIndex].m_to;\n"
-" b3Vector3 rayNormalizedDirection = b3Vector3_normalize(rayTo - rayFrom);\n"
-" b3Scalar rayLength = b3Sqrt( b3Vector3_length2(rayTo - rayFrom) );\n"
-" \n"
-" for(int i = 0; i < numLargeAabbRigids; ++i)\n"
-" {\n"
-" b3AabbCL rigidAabb = largeRigidAabbs[i];\n"
-" if( rayIntersectsAabb(rayFrom, rayLength, rayNormalizedDirection, rigidAabb) )\n"
-" {\n"
-" int2 rayRigidPair;\n"
-" rayRigidPair.x = rayIndex;\n"
-" rayRigidPair.y = rigidAabb.m_minIndices[3];\n"
-" \n"
-" int pairIndex = atomic_inc(out_numRayRigidPairs);\n"
-" if(pairIndex < maxRayRigidPairs) out_rayRigidPairs[pairIndex] = rayRigidPair;\n"
-" }\n"
-" }\n"
-"}\n"
-"//Set so that it is always greater than the actual common prefixes, and never selected as a parent node.\n"
-"//If there are no duplicates, then the highest common prefix is 32 or 64, depending on the number of bits used for the z-curve.\n"
-"//Duplicate common prefixes increase the highest common prefix at most by the number of bits used to index the leaf node.\n"
-"//Since 32 bit ints are used to index leaf nodes, the max prefix is 64(32 + 32 bit z-curve) or 96(32 + 64 bit z-curve).\n"
-"#define B3_PLBVH_INVALID_COMMON_PREFIX 128\n"
-"#define B3_PLBVH_ROOT_NODE_MARKER -1\n"
-"#define b3Int64 long\n"
-"int computeCommonPrefixLength(b3Int64 i, b3Int64 j) { return (int)clz(i ^ j); }\n"
-"b3Int64 computeCommonPrefix(b3Int64 i, b3Int64 j) \n"
-"{\n"
-" //This function only needs to return (i & j) in order for the algorithm to work,\n"
-" //but it may help with debugging to mask out the lower bits.\n"
-" b3Int64 commonPrefixLength = (b3Int64)computeCommonPrefixLength(i, j);\n"
-" b3Int64 sharedBits = i & j;\n"
-" b3Int64 bitmask = ((b3Int64)(~0)) << (64 - commonPrefixLength); //Set all bits after the common prefix to 0\n"
-" \n"
-" return sharedBits & bitmask;\n"
-"}\n"
-"//Same as computeCommonPrefixLength(), but allows for prefixes with different lengths\n"
-"int getSharedPrefixLength(b3Int64 prefixA, int prefixLengthA, b3Int64 prefixB, int prefixLengthB)\n"
-"{\n"
-" return b3Min( computeCommonPrefixLength(prefixA, prefixB), b3Min(prefixLengthA, prefixLengthB) );\n"
-"}\n"
-"__kernel void computeAdjacentPairCommonPrefix(__global SortDataCL* mortonCodesAndAabbIndices,\n"
-" __global b3Int64* out_commonPrefixes,\n"
-" __global int* out_commonPrefixLengths,\n"
-" int numInternalNodes)\n"
-"{\n"
-" int internalNodeIndex = get_global_id(0);\n"
-" if (internalNodeIndex >= numInternalNodes) return;\n"
-" \n"
-" //Here, (internalNodeIndex + 1) is never out of bounds since it is a leaf node index,\n"
-" //and the number of internal nodes is always numLeafNodes - 1\n"
-" int leftLeafIndex = internalNodeIndex;\n"
-" int rightLeafIndex = internalNodeIndex + 1;\n"
-" \n"
-" int leftLeafMortonCode = mortonCodesAndAabbIndices[leftLeafIndex].m_key;\n"
-" int rightLeafMortonCode = mortonCodesAndAabbIndices[rightLeafIndex].m_key;\n"
-" \n"
-" //Binary radix tree construction algorithm does not work if there are duplicate morton codes.\n"
-" //Append the index of each leaf node to each morton code so that there are no duplicates.\n"
-" //The algorithm also requires that the morton codes are sorted in ascending order; this requirement\n"
-" //is also satisfied with this method, as (leftLeafIndex < rightLeafIndex) is always true.\n"
-" //\n"
-" //upsample(a, b) == ( ((b3Int64)a) << 32) | b\n"
-" b3Int64 nonduplicateLeftMortonCode = upsample(leftLeafMortonCode, leftLeafIndex);\n"
-" b3Int64 nonduplicateRightMortonCode = upsample(rightLeafMortonCode, rightLeafIndex);\n"
-" \n"
-" out_commonPrefixes[internalNodeIndex] = computeCommonPrefix(nonduplicateLeftMortonCode, nonduplicateRightMortonCode);\n"
-" out_commonPrefixLengths[internalNodeIndex] = computeCommonPrefixLength(nonduplicateLeftMortonCode, nonduplicateRightMortonCode);\n"
-"}\n"
-"__kernel void buildBinaryRadixTreeLeafNodes(__global int* commonPrefixLengths, __global int* out_leafNodeParentNodes,\n"
-" __global int2* out_childNodes, int numLeafNodes)\n"
-"{\n"
-" int leafNodeIndex = get_global_id(0);\n"
-" if (leafNodeIndex >= numLeafNodes) return;\n"
-" \n"
-" int numInternalNodes = numLeafNodes - 1;\n"
-" \n"
-" int leftSplitIndex = leafNodeIndex - 1;\n"
-" int rightSplitIndex = leafNodeIndex;\n"
-" \n"
-" int leftCommonPrefix = (leftSplitIndex >= 0) ? commonPrefixLengths[leftSplitIndex] : B3_PLBVH_INVALID_COMMON_PREFIX;\n"
-" int rightCommonPrefix = (rightSplitIndex < numInternalNodes) ? commonPrefixLengths[rightSplitIndex] : B3_PLBVH_INVALID_COMMON_PREFIX;\n"
-" \n"
-" //Parent node is the highest adjacent common prefix that is lower than the node's common prefix\n"
-" //Leaf nodes are considered as having the highest common prefix\n"
-" int isLeftHigherCommonPrefix = (leftCommonPrefix > rightCommonPrefix);\n"
-" \n"
-" //Handle cases for the edge nodes; the first and last node\n"
-" //For leaf nodes, leftCommonPrefix and rightCommonPrefix should never both be B3_PLBVH_INVALID_COMMON_PREFIX\n"
-" if(leftCommonPrefix == B3_PLBVH_INVALID_COMMON_PREFIX) isLeftHigherCommonPrefix = false;\n"
-" if(rightCommonPrefix == B3_PLBVH_INVALID_COMMON_PREFIX) isLeftHigherCommonPrefix = true;\n"
-" \n"
-" int parentNodeIndex = (isLeftHigherCommonPrefix) ? leftSplitIndex : rightSplitIndex;\n"
-" out_leafNodeParentNodes[leafNodeIndex] = parentNodeIndex;\n"
-" \n"
-" int isRightChild = (isLeftHigherCommonPrefix); //If the left node is the parent, then this node is its right child and vice versa\n"
-" \n"
-" //out_childNodesAsInt[0] == int2.x == left child\n"
-" //out_childNodesAsInt[1] == int2.y == right child\n"
-" int isLeaf = 1;\n"
-" __global int* out_childNodesAsInt = (__global int*)(&out_childNodes[parentNodeIndex]);\n"
-" out_childNodesAsInt[isRightChild] = getIndexWithInternalNodeMarkerSet(isLeaf, leafNodeIndex);\n"
-"}\n"
-"__kernel void buildBinaryRadixTreeInternalNodes(__global b3Int64* commonPrefixes, __global int* commonPrefixLengths,\n"
-" __global int2* out_childNodes,\n"
-" __global int* out_internalNodeParentNodes, __global int* out_rootNodeIndex,\n"
-" int numInternalNodes)\n"
-"{\n"
-" int internalNodeIndex = get_group_id(0) * get_local_size(0) + get_local_id(0);\n"
-" if(internalNodeIndex >= numInternalNodes) return;\n"
-" \n"
-" b3Int64 nodePrefix = commonPrefixes[internalNodeIndex];\n"
-" int nodePrefixLength = commonPrefixLengths[internalNodeIndex];\n"
-" \n"
-"//#define USE_LINEAR_SEARCH\n"
-"#ifdef USE_LINEAR_SEARCH\n"
-" int leftIndex = -1;\n"
-" int rightIndex = -1;\n"
-" \n"
-" //Find nearest element to left with a lower common prefix\n"
-" for(int i = internalNodeIndex - 1; i >= 0; --i)\n"
-" {\n"
-" int nodeLeftSharedPrefixLength = getSharedPrefixLength(nodePrefix, nodePrefixLength, commonPrefixes[i], commonPrefixLengths[i]);\n"
-" if(nodeLeftSharedPrefixLength < nodePrefixLength)\n"
-" {\n"
-" leftIndex = i;\n"
-" break;\n"
-" }\n"
-" }\n"
-" \n"
-" //Find nearest element to right with a lower common prefix\n"
-" for(int i = internalNodeIndex + 1; i < numInternalNodes; ++i)\n"
-" {\n"
-" int nodeRightSharedPrefixLength = getSharedPrefixLength(nodePrefix, nodePrefixLength, commonPrefixes[i], commonPrefixLengths[i]);\n"
-" if(nodeRightSharedPrefixLength < nodePrefixLength)\n"
-" {\n"
-" rightIndex = i;\n"
-" break;\n"
-" }\n"
-" }\n"
-" \n"
-"#else //Use binary search\n"
-" //Find nearest element to left with a lower common prefix\n"
-" int leftIndex = -1;\n"
-" {\n"
-" int lower = 0;\n"
-" int upper = internalNodeIndex - 1;\n"
-" \n"
-" while(lower <= upper)\n"
-" {\n"
-" int mid = (lower + upper) / 2;\n"
-" b3Int64 midPrefix = commonPrefixes[mid];\n"
-" int midPrefixLength = commonPrefixLengths[mid];\n"
-" \n"
-" int nodeMidSharedPrefixLength = getSharedPrefixLength(nodePrefix, nodePrefixLength, midPrefix, midPrefixLength);\n"
-" if(nodeMidSharedPrefixLength < nodePrefixLength) \n"
-" {\n"
-" int right = mid + 1;\n"
-" if(right < internalNodeIndex)\n"
-" {\n"
-" b3Int64 rightPrefix = commonPrefixes[right];\n"
-" int rightPrefixLength = commonPrefixLengths[right];\n"
-" \n"
-" int nodeRightSharedPrefixLength = getSharedPrefixLength(nodePrefix, nodePrefixLength, rightPrefix, rightPrefixLength);\n"
-" if(nodeRightSharedPrefixLength < nodePrefixLength) \n"
-" {\n"
-" lower = right;\n"
-" leftIndex = right;\n"
-" }\n"
-" else \n"
-" {\n"
-" leftIndex = mid;\n"
-" break;\n"
-" }\n"
-" }\n"
-" else \n"
-" {\n"
-" leftIndex = mid;\n"
-" break;\n"
-" }\n"
-" }\n"
-" else upper = mid - 1;\n"
-" }\n"
-" }\n"
-" \n"
-" //Find nearest element to right with a lower common prefix\n"
-" int rightIndex = -1;\n"
-" {\n"
-" int lower = internalNodeIndex + 1;\n"
-" int upper = numInternalNodes - 1;\n"
-" \n"
-" while(lower <= upper)\n"
-" {\n"
-" int mid = (lower + upper) / 2;\n"
-" b3Int64 midPrefix = commonPrefixes[mid];\n"
-" int midPrefixLength = commonPrefixLengths[mid];\n"
-" \n"
-" int nodeMidSharedPrefixLength = getSharedPrefixLength(nodePrefix, nodePrefixLength, midPrefix, midPrefixLength);\n"
-" if(nodeMidSharedPrefixLength < nodePrefixLength) \n"
-" {\n"
-" int left = mid - 1;\n"
-" if(left > internalNodeIndex)\n"
-" {\n"
-" b3Int64 leftPrefix = commonPrefixes[left];\n"
-" int leftPrefixLength = commonPrefixLengths[left];\n"
-" \n"
-" int nodeLeftSharedPrefixLength = getSharedPrefixLength(nodePrefix, nodePrefixLength, leftPrefix, leftPrefixLength);\n"
-" if(nodeLeftSharedPrefixLength < nodePrefixLength) \n"
-" {\n"
-" upper = left;\n"
-" rightIndex = left;\n"
-" }\n"
-" else \n"
-" {\n"
-" rightIndex = mid;\n"
-" break;\n"
-" }\n"
-" }\n"
-" else \n"
-" {\n"
-" rightIndex = mid;\n"
-" break;\n"
-" }\n"
-" }\n"
-" else lower = mid + 1;\n"
-" }\n"
-" }\n"
-"#endif\n"
-" \n"
-" //Select parent\n"
-" {\n"
-" int leftPrefixLength = (leftIndex != -1) ? commonPrefixLengths[leftIndex] : B3_PLBVH_INVALID_COMMON_PREFIX;\n"
-" int rightPrefixLength = (rightIndex != -1) ? commonPrefixLengths[rightIndex] : B3_PLBVH_INVALID_COMMON_PREFIX;\n"
-" \n"
-" int isLeftHigherPrefixLength = (leftPrefixLength > rightPrefixLength);\n"
-" \n"
-" if(leftPrefixLength == B3_PLBVH_INVALID_COMMON_PREFIX) isLeftHigherPrefixLength = false;\n"
-" else if(rightPrefixLength == B3_PLBVH_INVALID_COMMON_PREFIX) isLeftHigherPrefixLength = true;\n"
-" \n"
-" int parentNodeIndex = (isLeftHigherPrefixLength) ? leftIndex : rightIndex;\n"
-" \n"
-" int isRootNode = (leftIndex == -1 && rightIndex == -1);\n"
-" out_internalNodeParentNodes[internalNodeIndex] = (!isRootNode) ? parentNodeIndex : B3_PLBVH_ROOT_NODE_MARKER;\n"
-" \n"
-" int isLeaf = 0;\n"
-" if(!isRootNode)\n"
-" {\n"
-" int isRightChild = (isLeftHigherPrefixLength); //If the left node is the parent, then this node is its right child and vice versa\n"
-" \n"
-" //out_childNodesAsInt[0] == int2.x == left child\n"
-" //out_childNodesAsInt[1] == int2.y == right child\n"
-" __global int* out_childNodesAsInt = (__global int*)(&out_childNodes[parentNodeIndex]);\n"
-" out_childNodesAsInt[isRightChild] = getIndexWithInternalNodeMarkerSet(isLeaf, internalNodeIndex);\n"
-" }\n"
-" else *out_rootNodeIndex = getIndexWithInternalNodeMarkerSet(isLeaf, internalNodeIndex);\n"
-" }\n"
-"}\n"
-"__kernel void findDistanceFromRoot(__global int* rootNodeIndex, __global int* internalNodeParentNodes,\n"
-" __global int* out_maxDistanceFromRoot, __global int* out_distanceFromRoot, int numInternalNodes)\n"
-"{\n"
-" if( get_global_id(0) == 0 ) atomic_xchg(out_maxDistanceFromRoot, 0);\n"
-" int internalNodeIndex = get_global_id(0);\n"
-" if(internalNodeIndex >= numInternalNodes) return;\n"
-" \n"
-" //\n"
-" int distanceFromRoot = 0;\n"
-" {\n"
-" int parentIndex = internalNodeParentNodes[internalNodeIndex];\n"
-" while(parentIndex != B3_PLBVH_ROOT_NODE_MARKER)\n"
-" {\n"
-" parentIndex = internalNodeParentNodes[parentIndex];\n"
-" ++distanceFromRoot;\n"
-" }\n"
-" }\n"
-" out_distanceFromRoot[internalNodeIndex] = distanceFromRoot;\n"
-" \n"
-" //\n"
-" __local int localMaxDistanceFromRoot;\n"
-" if( get_local_id(0) == 0 ) localMaxDistanceFromRoot = 0;\n"
-" barrier(CLK_LOCAL_MEM_FENCE);\n"
-" \n"
-" atomic_max(&localMaxDistanceFromRoot, distanceFromRoot);\n"
-" barrier(CLK_LOCAL_MEM_FENCE);\n"
-" \n"
-" if( get_local_id(0) == 0 ) atomic_max(out_maxDistanceFromRoot, localMaxDistanceFromRoot);\n"
-"}\n"
-"__kernel void buildBinaryRadixTreeAabbsRecursive(__global int* distanceFromRoot, __global SortDataCL* mortonCodesAndAabbIndices,\n"
-" __global int2* childNodes,\n"
-" __global b3AabbCL* leafNodeAabbs, __global b3AabbCL* internalNodeAabbs,\n"
-" int maxDistanceFromRoot, int processedDistance, int numInternalNodes)\n"
-"{\n"
-" int internalNodeIndex = get_global_id(0);\n"
-" if(internalNodeIndex >= numInternalNodes) return;\n"
-" \n"
-" int distance = distanceFromRoot[internalNodeIndex];\n"
-" \n"
-" if(distance == processedDistance)\n"
-" {\n"
-" int leftChildIndex = childNodes[internalNodeIndex].x;\n"
-" int rightChildIndex = childNodes[internalNodeIndex].y;\n"
-" \n"
-" int isLeftChildLeaf = isLeafNode(leftChildIndex);\n"
-" int isRightChildLeaf = isLeafNode(rightChildIndex);\n"
-" \n"
-" leftChildIndex = getIndexWithInternalNodeMarkerRemoved(leftChildIndex);\n"
-" rightChildIndex = getIndexWithInternalNodeMarkerRemoved(rightChildIndex);\n"
-" \n"
-" //leftRigidIndex/rightRigidIndex is not used if internal node\n"
-" int leftRigidIndex = (isLeftChildLeaf) ? mortonCodesAndAabbIndices[leftChildIndex].m_value : -1;\n"
-" int rightRigidIndex = (isRightChildLeaf) ? mortonCodesAndAabbIndices[rightChildIndex].m_value : -1;\n"
-" \n"
-" b3AabbCL leftChildAabb = (isLeftChildLeaf) ? leafNodeAabbs[leftRigidIndex] : internalNodeAabbs[leftChildIndex];\n"
-" b3AabbCL rightChildAabb = (isRightChildLeaf) ? leafNodeAabbs[rightRigidIndex] : internalNodeAabbs[rightChildIndex];\n"
-" \n"
-" b3AabbCL mergedAabb;\n"
-" mergedAabb.m_min = b3Min(leftChildAabb.m_min, rightChildAabb.m_min);\n"
-" mergedAabb.m_max = b3Max(leftChildAabb.m_max, rightChildAabb.m_max);\n"
-" internalNodeAabbs[internalNodeIndex] = mergedAabb;\n"
-" }\n"
-"}\n"
-"__kernel void findLeafIndexRanges(__global int2* internalNodeChildNodes, __global int2* out_leafIndexRanges, int numInternalNodes)\n"
-"{\n"
-" int internalNodeIndex = get_global_id(0);\n"
-" if(internalNodeIndex >= numInternalNodes) return;\n"
-" \n"
-" int numLeafNodes = numInternalNodes + 1;\n"
-" \n"
-" int2 childNodes = internalNodeChildNodes[internalNodeIndex];\n"
-" \n"
-" int2 leafIndexRange; //x == min leaf index, y == max leaf index\n"
-" \n"
-" //Find lowest leaf index covered by this internal node\n"
-" {\n"
-" int lowestIndex = childNodes.x; //childNodes.x == Left child\n"
-" while( !isLeafNode(lowestIndex) ) lowestIndex = internalNodeChildNodes[ getIndexWithInternalNodeMarkerRemoved(lowestIndex) ].x;\n"
-" leafIndexRange.x = lowestIndex;\n"
-" }\n"
-" \n"
-" //Find highest leaf index covered by this internal node\n"
-" {\n"
-" int highestIndex = childNodes.y; //childNodes.y == Right child\n"
-" while( !isLeafNode(highestIndex) ) highestIndex = internalNodeChildNodes[ getIndexWithInternalNodeMarkerRemoved(highestIndex) ].y;\n"
-" leafIndexRange.y = highestIndex;\n"
-" }\n"
-" \n"
-" //\n"
-" out_leafIndexRanges[internalNodeIndex] = leafIndexRange;\n"
-"}\n"
-;
/include/Bullet/Bullet3OpenCL/BroadphaseCollision/kernels/parallelLinearBvhKernels.h
Property changes:
Deleted: svn:executable
## -1 +0,0 ##
-*
\ No newline at end of property
Index: include/Bullet/Bullet3OpenCL/BroadphaseCollision/kernels/gridBroadphaseKernels.h
===================================================================
--- include/Bullet/Bullet3OpenCL/BroadphaseCollision/kernels/gridBroadphaseKernels.h (revision 22)
+++ include/Bullet/Bullet3OpenCL/BroadphaseCollision/kernels/gridBroadphaseKernels.h (nonexistent)
@@ -1,199 +0,0 @@
-//this file is autogenerated using stringify.bat (premake --stringify) in the build folder of this project
-static const char* gridBroadphaseCL= \
-"int getPosHash(int4 gridPos, __global float4* pParams)\n"
-"{\n"
-" int4 gridDim = *((__global int4*)(pParams + 1));\n"
-" gridPos.x &= gridDim.x - 1;\n"
-" gridPos.y &= gridDim.y - 1;\n"
-" gridPos.z &= gridDim.z - 1;\n"
-" int hash = gridPos.z * gridDim.y * gridDim.x + gridPos.y * gridDim.x + gridPos.x;\n"
-" return hash;\n"
-"} \n"
-"int4 getGridPos(float4 worldPos, __global float4* pParams)\n"
-"{\n"
-" int4 gridPos;\n"
-" int4 gridDim = *((__global int4*)(pParams + 1));\n"
-" gridPos.x = (int)floor(worldPos.x * pParams[0].x) & (gridDim.x - 1);\n"
-" gridPos.y = (int)floor(worldPos.y * pParams[0].y) & (gridDim.y - 1);\n"
-" gridPos.z = (int)floor(worldPos.z * pParams[0].z) & (gridDim.z - 1);\n"
-" return gridPos;\n"
-"}\n"
-"// calculate grid hash value for each body using its AABB\n"
-"__kernel void kCalcHashAABB(int numObjects, __global float4* allpAABB, __global const int* smallAabbMapping, __global int2* pHash, __global float4* pParams )\n"
-"{\n"
-" int index = get_global_id(0);\n"
-" if(index >= numObjects)\n"
-" {\n"
-" return;\n"
-" }\n"
-" float4 bbMin = allpAABB[smallAabbMapping[index]*2];\n"
-" float4 bbMax = allpAABB[smallAabbMapping[index]*2 + 1];\n"
-" float4 pos;\n"
-" pos.x = (bbMin.x + bbMax.x) * 0.5f;\n"
-" pos.y = (bbMin.y + bbMax.y) * 0.5f;\n"
-" pos.z = (bbMin.z + bbMax.z) * 0.5f;\n"
-" pos.w = 0.f;\n"
-" // get address in grid\n"
-" int4 gridPos = getGridPos(pos, pParams);\n"
-" int gridHash = getPosHash(gridPos, pParams);\n"
-" // store grid hash and body index\n"
-" int2 hashVal;\n"
-" hashVal.x = gridHash;\n"
-" hashVal.y = index;\n"
-" pHash[index] = hashVal;\n"
-"}\n"
-"__kernel void kClearCellStart( int numCells, \n"
-" __global int* pCellStart )\n"
-"{\n"
-" int index = get_global_id(0);\n"
-" if(index >= numCells)\n"
-" {\n"
-" return;\n"
-" }\n"
-" pCellStart[index] = -1;\n"
-"}\n"
-"__kernel void kFindCellStart(int numObjects, __global int2* pHash, __global int* cellStart )\n"
-"{\n"
-" __local int sharedHash[513];\n"
-" int index = get_global_id(0);\n"
-" int2 sortedData;\n"
-" if(index < numObjects)\n"
-" {\n"
-" sortedData = pHash[index];\n"
-" // Load hash data into shared memory so that we can look \n"
-" // at neighboring body's hash value without loading\n"
-" // two hash values per thread\n"
-" sharedHash[get_local_id(0) + 1] = sortedData.x;\n"
-" if((index > 0) && (get_local_id(0) == 0))\n"
-" {\n"
-" // first thread in block must load neighbor body hash\n"
-" sharedHash[0] = pHash[index-1].x;\n"
-" }\n"
-" }\n"
-" barrier(CLK_LOCAL_MEM_FENCE);\n"
-" if(index < numObjects)\n"
-" {\n"
-" if((index == 0) || (sortedData.x != sharedHash[get_local_id(0)]))\n"
-" {\n"
-" cellStart[sortedData.x] = index;\n"
-" }\n"
-" }\n"
-"}\n"
-"int testAABBOverlap(float4 min0, float4 max0, float4 min1, float4 max1)\n"
-"{\n"
-" return (min0.x <= max1.x)&& (min1.x <= max0.x) && \n"
-" (min0.y <= max1.y)&& (min1.y <= max0.y) && \n"
-" (min0.z <= max1.z)&& (min1.z <= max0.z); \n"
-"}\n"
-"//search for AABB 'index' against other AABBs' in this cell\n"
-"void findPairsInCell( int numObjects,\n"
-" int4 gridPos,\n"
-" int index,\n"
-" __global int2* pHash,\n"
-" __global int* pCellStart,\n"
-" __global float4* allpAABB, \n"
-" __global const int* smallAabbMapping,\n"
-" __global float4* pParams,\n"
-" volatile __global int* pairCount,\n"
-" __global int4* pPairBuff2,\n"
-" int maxPairs\n"
-" )\n"
-"{\n"
-" int4 pGridDim = *((__global int4*)(pParams + 1));\n"
-" int maxBodiesPerCell = pGridDim.w;\n"
-" int gridHash = getPosHash(gridPos, pParams);\n"
-" // get start of bucket for this cell\n"
-" int bucketStart = pCellStart[gridHash];\n"
-" if (bucketStart == -1)\n"
-" {\n"
-" return; // cell empty\n"
-" }\n"
-" // iterate over bodies in this cell\n"
-" int2 sortedData = pHash[index];\n"
-" int unsorted_indx = sortedData.y;\n"
-" float4 min0 = allpAABB[smallAabbMapping[unsorted_indx]*2 + 0]; \n"
-" float4 max0 = allpAABB[smallAabbMapping[unsorted_indx]*2 + 1];\n"
-" int handleIndex = as_int(min0.w);\n"
-" \n"
-" int bucketEnd = bucketStart + maxBodiesPerCell;\n"
-" bucketEnd = (bucketEnd > numObjects) ? numObjects : bucketEnd;\n"
-" for(int index2 = bucketStart; index2 < bucketEnd; index2++) \n"
-" {\n"
-" int2 cellData = pHash[index2];\n"
-" if (cellData.x != gridHash)\n"
-" {\n"
-" break; // no longer in same bucket\n"
-" }\n"
-" int unsorted_indx2 = cellData.y;\n"
-" //if (unsorted_indx2 < unsorted_indx) // check not colliding with self\n"
-" if (unsorted_indx2 != unsorted_indx) // check not colliding with self\n"
-" { \n"
-" float4 min1 = allpAABB[smallAabbMapping[unsorted_indx2]*2 + 0];\n"
-" float4 max1 = allpAABB[smallAabbMapping[unsorted_indx2]*2 + 1];\n"
-" if(testAABBOverlap(min0, max0, min1, max1))\n"
-" {\n"
-" if (pairCount)\n"
-" {\n"
-" int handleIndex2 = as_int(min1.w);\n"
-" if (handleIndex<handleIndex2)\n"
-" {\n"
-" int curPair = atomic_add(pairCount,1);\n"
-" if (curPair<maxPairs)\n"
-" {\n"
-" int4 newpair;\n"
-" newpair.x = handleIndex;\n"
-" newpair.y = handleIndex2;\n"
-" newpair.z = -1;\n"
-" newpair.w = -1;\n"
-" pPairBuff2[curPair] = newpair;\n"
-" }\n"
-" }\n"
-" \n"
-" }\n"
-" }\n"
-" }\n"
-" }\n"
-"}\n"
-"__kernel void kFindOverlappingPairs( int numObjects,\n"
-" __global float4* allpAABB, \n"
-" __global const int* smallAabbMapping,\n"
-" __global int2* pHash, \n"
-" __global int* pCellStart, \n"
-" __global float4* pParams ,\n"
-" volatile __global int* pairCount,\n"
-" __global int4* pPairBuff2,\n"
-" int maxPairs\n"
-" )\n"
-"{\n"
-" int index = get_global_id(0);\n"
-" if(index >= numObjects)\n"
-" {\n"
-" return;\n"
-" }\n"
-" int2 sortedData = pHash[index];\n"
-" int unsorted_indx = sortedData.y;\n"
-" float4 bbMin = allpAABB[smallAabbMapping[unsorted_indx]*2 + 0];\n"
-" float4 bbMax = allpAABB[smallAabbMapping[unsorted_indx]*2 + 1];\n"
-" float4 pos;\n"
-" pos.x = (bbMin.x + bbMax.x) * 0.5f;\n"
-" pos.y = (bbMin.y + bbMax.y) * 0.5f;\n"
-" pos.z = (bbMin.z + bbMax.z) * 0.5f;\n"
-" // get address in grid\n"
-" int4 gridPosA = getGridPos(pos, pParams);\n"
-" int4 gridPosB; \n"
-" // examine only neighbouring cells\n"
-" for(int z=-1; z<=1; z++) \n"
-" {\n"
-" gridPosB.z = gridPosA.z + z;\n"
-" for(int y=-1; y<=1; y++) \n"
-" {\n"
-" gridPosB.y = gridPosA.y + y;\n"
-" for(int x=-1; x<=1; x++) \n"
-" {\n"
-" gridPosB.x = gridPosA.x + x;\n"
-" findPairsInCell(numObjects, gridPosB, index, pHash, pCellStart, allpAABB,smallAabbMapping, pParams, pairCount,pPairBuff2, maxPairs);\n"
-" }\n"
-" }\n"
-" }\n"
-"}\n"
-;
/include/Bullet/Bullet3OpenCL/BroadphaseCollision/kernels/gridBroadphaseKernels.h
Property changes:
Deleted: svn:executable
## -1 +0,0 ##
-*
\ No newline at end of property
Index: include/Bullet/Bullet3OpenCL/BroadphaseCollision/kernels/sapKernels.h
===================================================================
--- include/Bullet/Bullet3OpenCL/BroadphaseCollision/kernels/sapKernels.h (revision 22)
+++ include/Bullet/Bullet3OpenCL/BroadphaseCollision/kernels/sapKernels.h (nonexistent)
@@ -1,342 +0,0 @@
-//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"
-"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"
-"//Originally written by Erwin Coumans\n"
-"#define NEW_PAIR_MARKER -1\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"
-"} btAabbCL;\n"
-"/// conservative test for overlap between two aabbs\n"
-"bool TestAabbAgainstAabb2(const btAabbCL* aabb1, __local const btAabbCL* aabb2);\n"
-"bool TestAabbAgainstAabb2(const btAabbCL* aabb1, __local const btAabbCL* 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 btAabbCL* aabb1, __global const btAabbCL* aabb2);\n"
-"bool TestAabbAgainstAabb2GlobalGlobal(__global const btAabbCL* aabb1, __global const btAabbCL* 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 TestAabbAgainstAabb2Global(const btAabbCL* aabb1, __global const btAabbCL* aabb2);\n"
-"bool TestAabbAgainstAabb2Global(const btAabbCL* aabb1, __global const btAabbCL* 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"
-"__kernel void computePairsKernelTwoArrays( __global const btAabbCL* unsortedAabbs, __global const int* unsortedAabbMapping, __global const int* unsortedAabbMapping2, volatile __global int4* pairsOut,volatile __global int* pairCount, int numUnsortedAabbs, int numUnSortedAabbs2, int axis, int maxPairs)\n"
-"{\n"
-" int i = get_global_id(0);\n"
-" if (i>=numUnsortedAabbs)\n"
-" return;\n"
-" int j = get_global_id(1);\n"
-" if (j>=numUnSortedAabbs2)\n"
-" return;\n"
-" __global const btAabbCL* unsortedAabbPtr = &unsortedAabbs[unsortedAabbMapping[i]];\n"
-" __global const btAabbCL* unsortedAabbPtr2 = &unsortedAabbs[unsortedAabbMapping2[j]];\n"
-" if (TestAabbAgainstAabb2GlobalGlobal(unsortedAabbPtr,unsortedAabbPtr2))\n"
-" {\n"
-" int4 myPair;\n"
-" \n"
-" int xIndex = unsortedAabbPtr[0].m_minIndices[3];\n"
-" int yIndex = unsortedAabbPtr2[0].m_minIndices[3];\n"
-" if (xIndex>yIndex)\n"
-" {\n"
-" int tmp = xIndex;\n"
-" xIndex=yIndex;\n"
-" yIndex=tmp;\n"
-" }\n"
-" \n"
-" myPair.x = xIndex;\n"
-" myPair.y = yIndex;\n"
-" myPair.z = NEW_PAIR_MARKER;\n"
-" myPair.w = NEW_PAIR_MARKER;\n"
-" int curPair = atomic_inc (pairCount);\n"
-" if (curPair<maxPairs)\n"
-" {\n"
-" pairsOut[curPair] = myPair; //flush to main memory\n"
-" }\n"
-" }\n"
-"}\n"
-"__kernel void computePairsKernelBruteForce( __global const btAabbCL* aabbs, volatile __global int4* 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 (TestAabbAgainstAabb2GlobalGlobal(&aabbs[i],&aabbs[j]))\n"
-" {\n"
-" int4 myPair;\n"
-" myPair.x = aabbs[i].m_minIndices[3];\n"
-" myPair.y = aabbs[j].m_minIndices[3];\n"
-" myPair.z = NEW_PAIR_MARKER;\n"
-" myPair.w = NEW_PAIR_MARKER;\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 btAabbCL* aabbs, volatile __global int4* 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"
-" int4 myPair;\n"
-" myPair.x = aabbs[i].m_minIndices[3];\n"
-" myPair.y = aabbs[j].m_minIndices[3];\n"
-" myPair.z = NEW_PAIR_MARKER;\n"
-" myPair.w = NEW_PAIR_MARKER;\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 computePairsKernelBarrier( __global const btAabbCL* aabbs, volatile __global int4* 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"
-" __local int numActiveWgItems[1];\n"
-" __local int breakRequest[1];\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"
-" 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"
-" int4 myPair;\n"
-" myPair.x = aabbs[i].m_minIndices[3];\n"
-" myPair.y = aabbs[j].m_minIndices[3];\n"
-" myPair.z = NEW_PAIR_MARKER;\n"
-" myPair.w = NEW_PAIR_MARKER;\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"
-" } while (breakRequest[0]<numActiveWgItems[0]);\n"
-"}\n"
-"__kernel void computePairsKernelLocalSharedMemory( __global const btAabbCL* aabbs, volatile __global int4* 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"
-" __local int numActiveWgItems[1];\n"
-" __local int breakRequest[1];\n"
-" __local btAabbCL localAabbs[128];// = aabbs[i];\n"
-" \n"
-" btAabbCL 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"
-" int4 myPair;\n"
-" myPair.x = myAabb.m_minIndices[3];\n"
-" myPair.y = localAabbs[localCount+localId+1].m_minIndices[3];\n"
-" myPair.z = NEW_PAIR_MARKER;\n"
-" myPair.w = NEW_PAIR_MARKER;\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"
-" 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"
-"//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"
-"__kernel void copyAabbsKernel( __global const btAabbCL* allAabbs, __global btAabbCL* 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"
-"__kernel void flipFloatKernel( __global const btAabbCL* allAabbs, __global const int* smallAabbMapping, __global int2* sortData, int numObjects, int axis)\n"
-"{\n"
-" int i = get_global_id(0);\n"
-" if (i>=numObjects)\n"
-" return;\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* 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"
-" \n"
-" sortedAabbs[i] = allAabbs[smallAabbMapping[sortData[i].y]];\n"
-"}\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 = (smallAabb.m_max+smallAabb.m_min)*0.5f;\n"
-" sum[i]=s;\n"
-" sum2[i]=s*s; \n"
-"}\n"
-;
/include/Bullet/Bullet3OpenCL/BroadphaseCollision/kernels/sapKernels.h
Property changes:
Deleted: svn:executable
## -1 +0,0 ##
-*
\ No newline at end of property
Index: include/Bullet/Bullet3OpenCL/BroadphaseCollision/b3SapAabb.h
===================================================================
--- include/Bullet/Bullet3OpenCL/BroadphaseCollision/b3SapAabb.h (revision 22)
+++ include/Bullet/Bullet3OpenCL/BroadphaseCollision/b3SapAabb.h (nonexistent)
@@ -1,14 +0,0 @@
-#ifndef B3_SAP_AABB_H
-#define B3_SAP_AABB_H
-
-#include "Bullet3Common/b3Scalar.h"
-#include "Bullet3Collision/BroadPhaseCollision/shared/b3Aabb.h"
-
-///just make sure that the b3Aabb is 16-byte aligned
-B3_ATTRIBUTE_ALIGNED16(struct) b3SapAabb : public b3Aabb
-{
-
-};
-
-
-#endif //B3_SAP_AABB_H
/include/Bullet/Bullet3OpenCL/BroadphaseCollision/b3SapAabb.h
Property changes:
Deleted: svn:executable
## -1 +0,0 ##
-*
\ No newline at end of property
Index: include/Bullet/Bullet3OpenCL/BroadphaseCollision/b3GpuGridBroadphase.h
===================================================================
--- include/Bullet/Bullet3OpenCL/BroadphaseCollision/b3GpuGridBroadphase.h (revision 22)
+++ include/Bullet/Bullet3OpenCL/BroadphaseCollision/b3GpuGridBroadphase.h (nonexistent)
@@ -1,88 +0,0 @@
-#ifndef B3_GPU_GRID_BROADPHASE_H
-#define B3_GPU_GRID_BROADPHASE_H
-
-#include "b3GpuBroadphaseInterface.h"
-#include "Bullet3OpenCL/ParallelPrimitives/b3RadixSort32CL.h"
-
-struct b3ParamsGridBroadphaseCL
-{
-
- float m_invCellSize[4];
- int m_gridSize[4];
-
- int getMaxBodiesPerCell() const
- {
- return m_gridSize[3];
- }
-
- void setMaxBodiesPerCell(int maxOverlap)
- {
- m_gridSize[3] = maxOverlap;
- }
-};
-
-
-class b3GpuGridBroadphase : public b3GpuBroadphaseInterface
-{
-protected:
- cl_context m_context;
- cl_device_id m_device;
- cl_command_queue m_queue;
-
- b3OpenCLArray<b3SapAabb> m_allAabbsGPU1;
- b3AlignedObjectArray<b3SapAabb> m_allAabbsCPU1;
-
- b3OpenCLArray<int> m_smallAabbsMappingGPU;
- b3AlignedObjectArray<int> m_smallAabbsMappingCPU;
-
- b3OpenCLArray<int> m_largeAabbsMappingGPU;
- b3AlignedObjectArray<int> m_largeAabbsMappingCPU;
-
- b3AlignedObjectArray<b3Int4> m_hostPairs;
- b3OpenCLArray<b3Int4> m_gpuPairs;
-
- b3OpenCLArray<b3SortData> m_hashGpu;
- b3OpenCLArray<int> m_cellStartGpu;
-
-
- b3ParamsGridBroadphaseCL m_paramsCPU;
- b3OpenCLArray<b3ParamsGridBroadphaseCL> m_paramsGPU;
-
- class b3RadixSort32CL* m_sorter;
-
-public:
-
- b3GpuGridBroadphase(cl_context ctx,cl_device_id device, cl_command_queue q );
- virtual ~b3GpuGridBroadphase();
-
- static b3GpuBroadphaseInterface* CreateFunc(cl_context ctx,cl_device_id device, cl_command_queue q)
- {
- return new b3GpuGridBroadphase(ctx,device,q);
- }
-
-
-
-
- virtual void createProxy(const b3Vector3& aabbMin, const b3Vector3& aabbMax, int userPtr ,short int collisionFilterGroup,short int collisionFilterMask);
- virtual void createLargeProxy(const b3Vector3& aabbMin, const b3Vector3& aabbMax, int userPtr ,short int collisionFilterGroup,short int collisionFilterMask);
-
- virtual void calculateOverlappingPairs(int maxPairs);
- virtual void calculateOverlappingPairsHost(int maxPairs);
-
- //call writeAabbsToGpu after done making all changes (createProxy etc)
- virtual void writeAabbsToGpu();
-
- virtual cl_mem getAabbBufferWS();
- virtual int getNumOverlap();
- virtual cl_mem getOverlappingPairBuffer();
-
- virtual b3OpenCLArray<b3SapAabb>& getAllAabbsGPU();
- virtual b3AlignedObjectArray<b3SapAabb>& getAllAabbsCPU();
-
- virtual b3OpenCLArray<b3Int4>& getOverlappingPairsGPU();
- virtual b3OpenCLArray<int>& getSmallAabbIndicesGPU();
- virtual b3OpenCLArray<int>& getLargeAabbIndicesGPU();
-
-};
-
-#endif //B3_GPU_GRID_BROADPHASE_H
\ No newline at end of file
/include/Bullet/Bullet3OpenCL/BroadphaseCollision/b3GpuGridBroadphase.h
Property changes:
Deleted: svn:executable
## -1 +0,0 ##
-*
\ No newline at end of property
Index: include/Bullet/Bullet3OpenCL/BroadphaseCollision/b3GpuParallelLinearBvhBroadphase.h
===================================================================
--- include/Bullet/Bullet3OpenCL/BroadphaseCollision/b3GpuParallelLinearBvhBroadphase.h (revision 22)
+++ include/Bullet/Bullet3OpenCL/BroadphaseCollision/b3GpuParallelLinearBvhBroadphase.h (nonexistent)
@@ -1,66 +0,0 @@
-/*
-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.
-*/
-//Initial Author Jackson Lee, 2014
-
-#ifndef B3_GPU_PARALLEL_LINEAR_BVH_BROADPHASE_H
-#define B3_GPU_PARALLEL_LINEAR_BVH_BROADPHASE_H
-
-#include "Bullet3OpenCL/BroadphaseCollision/b3GpuBroadphaseInterface.h"
-
-#include "b3GpuParallelLinearBvh.h"
-
-class b3GpuParallelLinearBvhBroadphase : public b3GpuBroadphaseInterface
-{
- b3GpuParallelLinearBvh m_plbvh;
-
- b3OpenCLArray<b3Int4> m_overlappingPairsGpu;
-
- b3OpenCLArray<b3SapAabb> m_aabbsGpu;
- b3OpenCLArray<int> m_smallAabbsMappingGpu;
- b3OpenCLArray<int> m_largeAabbsMappingGpu;
-
- b3AlignedObjectArray<b3SapAabb> m_aabbsCpu;
- b3AlignedObjectArray<int> m_smallAabbsMappingCpu;
- b3AlignedObjectArray<int> m_largeAabbsMappingCpu;
-
-public:
- b3GpuParallelLinearBvhBroadphase(cl_context context, cl_device_id device, cl_command_queue queue);
- virtual ~b3GpuParallelLinearBvhBroadphase() {}
-
- virtual void createProxy(const b3Vector3& aabbMin, const b3Vector3& aabbMax, int userPtr, short int collisionFilterGroup, short int collisionFilterMask);
- virtual void createLargeProxy(const b3Vector3& aabbMin, const b3Vector3& aabbMax, int userPtr, short int collisionFilterGroup, short int collisionFilterMask);
-
- virtual void calculateOverlappingPairs(int maxPairs);
- virtual void calculateOverlappingPairsHost(int maxPairs);
-
- //call writeAabbsToGpu after done making all changes (createProxy etc)
- virtual void writeAabbsToGpu();
-
- virtual int getNumOverlap() { return m_overlappingPairsGpu.size(); }
- virtual cl_mem getOverlappingPairBuffer() { return m_overlappingPairsGpu.getBufferCL(); }
-
- virtual cl_mem getAabbBufferWS() { return m_aabbsGpu.getBufferCL(); }
- virtual b3OpenCLArray<b3SapAabb>& getAllAabbsGPU() { return m_aabbsGpu; }
-
- virtual b3OpenCLArray<b3Int4>& getOverlappingPairsGPU() { return m_overlappingPairsGpu; }
- virtual b3OpenCLArray<int>& getSmallAabbIndicesGPU() { return m_smallAabbsMappingGpu; }
- virtual b3OpenCLArray<int>& getLargeAabbIndicesGPU() { return m_largeAabbsMappingGpu; }
-
- virtual b3AlignedObjectArray<b3SapAabb>& getAllAabbsCPU() { return m_aabbsCpu; }
-
- static b3GpuBroadphaseInterface* CreateFunc(cl_context context, cl_device_id device, cl_command_queue queue)
- {
- return new b3GpuParallelLinearBvhBroadphase(context, device, queue);
- }
-};
-
-#endif
/include/Bullet/Bullet3OpenCL/BroadphaseCollision/b3GpuParallelLinearBvhBroadphase.h
Property changes:
Deleted: svn:executable
## -1 +0,0 ##
-*
\ No newline at end of property
Index: include/Bullet/Bullet3OpenCL/BroadphaseCollision/b3GpuParallelLinearBvh.h
===================================================================
--- include/Bullet/Bullet3OpenCL/BroadphaseCollision/b3GpuParallelLinearBvh.h (revision 22)
+++ include/Bullet/Bullet3OpenCL/BroadphaseCollision/b3GpuParallelLinearBvh.h (nonexistent)
@@ -1,125 +0,0 @@
-/*
-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.
-*/
-//Initial Author Jackson Lee, 2014
-
-#ifndef B3_GPU_PARALLEL_LINEAR_BVH_H
-#define B3_GPU_PARALLEL_LINEAR_BVH_H
-
-//#include "Bullet3Collision/BroadPhaseCollision/shared/b3Aabb.h"
-#include "Bullet3OpenCL/BroadphaseCollision/b3SapAabb.h"
-#include "Bullet3Common/shared/b3Int2.h"
-#include "Bullet3Common/shared/b3Int4.h"
-#include "Bullet3Collision/NarrowPhaseCollision/b3RaycastInfo.h"
-
-#include "Bullet3OpenCL/ParallelPrimitives/b3FillCL.h"
-#include "Bullet3OpenCL/ParallelPrimitives/b3RadixSort32CL.h"
-#include "Bullet3OpenCL/ParallelPrimitives/b3PrefixScanCL.h"
-
-#include "Bullet3OpenCL/BroadphaseCollision/kernels/parallelLinearBvhKernels.h"
-
-#define b3Int64 cl_long
-
-///@brief GPU Parallel Linearized Bounding Volume Heirarchy(LBVH) that is reconstructed every frame
-///@remarks
-///See presentation in docs/b3GpuParallelLinearBvh.pdf for algorithm details.
-///@par
-///Related papers: \n
-///"Fast BVH Construction on GPUs" [Lauterbach et al. 2009] \n
-///"Maximizing Parallelism in the Construction of BVHs, Octrees, and k-d trees" [Karras 2012] \n
-///@par
-///The basic algorithm for building the BVH as presented in [Lauterbach et al. 2009] consists of 4 stages:
-/// - [fully parallel] Assign morton codes for each AABB using its center (after quantizing the AABB centers into a virtual grid)
-/// - [fully parallel] Sort morton codes
-/// - [somewhat parallel] Build binary radix tree (assign parent/child pointers for internal nodes of the BVH)
-/// - [somewhat parallel] Set internal node AABBs
-///@par
-///[Karras 2012] improves on the algorithm by introducing fully parallel methods for the last 2 stages.
-///The BVH implementation here shares many concepts with [Karras 2012], but a different method is used for constructing the tree.
-///Instead of searching for the child nodes of each internal node, we search for the parent node of each node.
-///Additionally, a non-atomic traversal that starts from the leaf nodes and moves towards the root node is used to set the AABBs.
-class b3GpuParallelLinearBvh
-{
- cl_command_queue m_queue;
-
- cl_program m_parallelLinearBvhProgram;
-
- cl_kernel m_separateAabbsKernel;
- cl_kernel m_findAllNodesMergedAabbKernel;
- cl_kernel m_assignMortonCodesAndAabbIndiciesKernel;
-
- //Binary radix tree construction kernels
- cl_kernel m_computeAdjacentPairCommonPrefixKernel;
- cl_kernel m_buildBinaryRadixTreeLeafNodesKernel;
- cl_kernel m_buildBinaryRadixTreeInternalNodesKernel;
- cl_kernel m_findDistanceFromRootKernel;
- cl_kernel m_buildBinaryRadixTreeAabbsRecursiveKernel;
-
- cl_kernel m_findLeafIndexRangesKernel;
-
- //Traversal kernels
- cl_kernel m_plbvhCalculateOverlappingPairsKernel;
- cl_kernel m_plbvhRayTraverseKernel;
- cl_kernel m_plbvhLargeAabbAabbTestKernel;
- cl_kernel m_plbvhLargeAabbRayTestKernel;
-
- b3RadixSort32CL m_radixSorter;
-
- //1 element
- b3OpenCLArray<int> m_rootNodeIndex; //Most significant bit(0x80000000) is set to indicate internal node
- b3OpenCLArray<int> m_maxDistanceFromRoot; //Max number of internal nodes between an internal node and the root node
- b3OpenCLArray<int> m_temp; //Used to hold the number of pairs in calculateOverlappingPairs()
-
- //1 element per internal node (number_of_internal_nodes == number_of_leaves - 1)
- b3OpenCLArray<b3SapAabb> m_internalNodeAabbs;
- b3OpenCLArray<b3Int2> m_internalNodeLeafIndexRanges; //x == min leaf index, y == max leaf index
- b3OpenCLArray<b3Int2> m_internalNodeChildNodes; //x == left child, y == right child; msb(0x80000000) is set to indicate internal node
- b3OpenCLArray<int> m_internalNodeParentNodes; //For parent node index, msb(0x80000000) is not set since it is always internal
-
- //1 element per internal node; for binary radix tree construction
- b3OpenCLArray<b3Int64> m_commonPrefixes;
- b3OpenCLArray<int> m_commonPrefixLengths;
- b3OpenCLArray<int> m_distanceFromRoot; //Number of internal nodes between this node and the root
-
- //1 element per leaf node (leaf nodes only include small AABBs)
- b3OpenCLArray<int> m_leafNodeParentNodes; //For parent node index, msb(0x80000000) is not set since it is always internal
- b3OpenCLArray<b3SortData> m_mortonCodesAndAabbIndicies; //m_key == morton code, m_value == aabb index in m_leafNodeAabbs
- b3OpenCLArray<b3SapAabb> m_mergedAabb; //m_mergedAabb[0] contains the merged AABB of all leaf nodes
- b3OpenCLArray<b3SapAabb> m_leafNodeAabbs; //Contains only small AABBs
-
- //1 element per large AABB, which is not stored in the BVH
- b3OpenCLArray<b3SapAabb> m_largeAabbs;
-
-public:
- b3GpuParallelLinearBvh(cl_context context, cl_device_id device, cl_command_queue queue);
- virtual ~b3GpuParallelLinearBvh();
-
- ///Must be called before any other function
- void build(const b3OpenCLArray<b3SapAabb>& worldSpaceAabbs, const b3OpenCLArray<int>& smallAabbIndices,
- const b3OpenCLArray<int>& largeAabbIndices);
-
- ///calculateOverlappingPairs() uses the worldSpaceAabbs parameter of b3GpuParallelLinearBvh::build() as the query AABBs.
- ///@param out_overlappingPairs The size() of this array is used to determine the max number of pairs.
- ///If the number of overlapping pairs is < out_overlappingPairs.size(), out_overlappingPairs is resized.
- void calculateOverlappingPairs(b3OpenCLArray<b3Int4>& out_overlappingPairs);
-
- ///@param out_numRigidRayPairs Array of length 1; contains the number of detected ray-rigid AABB intersections;
- ///this value may be greater than out_rayRigidPairs.size() if out_rayRigidPairs is not large enough.
- ///@param out_rayRigidPairs Contains an array of rays intersecting rigid AABBs; x == ray index, y == rigid body index.
- ///If the size of this array is insufficient to hold all ray-rigid AABB intersections, additional intersections are discarded.
- void testRaysAgainstBvhAabbs(const b3OpenCLArray<b3RayInfo>& rays,
- b3OpenCLArray<int>& out_numRayRigidPairs, b3OpenCLArray<b3Int2>& out_rayRigidPairs);
-
-private:
- void constructBinaryRadixTree();
-};
-
-#endif
/include/Bullet/Bullet3OpenCL/BroadphaseCollision/b3GpuParallelLinearBvh.h
Property changes:
Deleted: svn:executable
## -1 +0,0 ##
-*
\ No newline at end of property
Index: include/Bullet/Bullet3OpenCL/BroadphaseCollision/b3GpuBroadphaseInterface.h
===================================================================
--- include/Bullet/Bullet3OpenCL/BroadphaseCollision/b3GpuBroadphaseInterface.h (revision 22)
+++ include/Bullet/Bullet3OpenCL/BroadphaseCollision/b3GpuBroadphaseInterface.h (nonexistent)
@@ -1,44 +0,0 @@
-
-#ifndef B3_GPU_BROADPHASE_INTERFACE_H
-#define B3_GPU_BROADPHASE_INTERFACE_H
-
-#include "Bullet3OpenCL/Initialize/b3OpenCLInclude.h"
-#include "Bullet3Common/b3Vector3.h"
-#include "b3SapAabb.h"
-#include "Bullet3Common/shared/b3Int2.h"
-#include "Bullet3Common/shared/b3Int4.h"
-#include "Bullet3OpenCL/ParallelPrimitives/b3OpenCLArray.h"
-
-class b3GpuBroadphaseInterface
-{
-public:
-
- typedef class b3GpuBroadphaseInterface* (CreateFunc)(cl_context ctx,cl_device_id device, cl_command_queue q);
-
- virtual ~b3GpuBroadphaseInterface()
- {
- }
-
- virtual void createProxy(const b3Vector3& aabbMin, const b3Vector3& aabbMax, int userPtr ,short int collisionFilterGroup,short int collisionFilterMask)=0;
- virtual void createLargeProxy(const b3Vector3& aabbMin, const b3Vector3& aabbMax, int userPtr ,short int collisionFilterGroup,short int collisionFilterMask)=0;
-
- virtual void calculateOverlappingPairs(int maxPairs)=0;
- virtual void calculateOverlappingPairsHost(int maxPairs)=0;
-
- //call writeAabbsToGpu after done making all changes (createProxy etc)
- virtual void writeAabbsToGpu()=0;
-
- virtual cl_mem getAabbBufferWS()=0;
- virtual int getNumOverlap()=0;
- virtual cl_mem getOverlappingPairBuffer()=0;
-
- virtual b3OpenCLArray<b3SapAabb>& getAllAabbsGPU()=0;
- virtual b3AlignedObjectArray<b3SapAabb>& getAllAabbsCPU()=0;
-
- virtual b3OpenCLArray<b3Int4>& getOverlappingPairsGPU() = 0;
- virtual b3OpenCLArray<int>& getSmallAabbIndicesGPU() = 0;
- virtual b3OpenCLArray<int>& getLargeAabbIndicesGPU() = 0;
-
-};
-
-#endif //B3_GPU_BROADPHASE_INTERFACE_H
/include/Bullet/Bullet3OpenCL/BroadphaseCollision/b3GpuBroadphaseInterface.h
Property changes:
Deleted: svn:executable
## -1 +0,0 ##
-*
\ No newline at end of property
Index: include/Bullet/Bullet3OpenCL/NarrowphaseCollision/b3ContactCache.h
===================================================================
--- include/Bullet/Bullet3OpenCL/NarrowphaseCollision/b3ContactCache.h (revision 22)
+++ include/Bullet/Bullet3OpenCL/NarrowphaseCollision/b3ContactCache.h (nonexistent)
@@ -1,80 +0,0 @@
-
-/*
-Bullet Continuous Collision Detection and Physics Library
-Copyright (c) 2003-2013 Erwin Coumans http://bulletphysics.org
-
-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.
-*/
-
-#ifndef B3_CONTACT_CACHE_H
-#define B3_CONTACT_CACHE_H
-
-
-#include "Bullet3Common/b3Vector3.h"
-#include "Bullet3Common/b3Transform.h"
-#include "Bullet3Common/b3AlignedAllocator.h"
-
-
-///maximum contact breaking and merging threshold
-extern b3Scalar gContactBreakingThreshold;
-
-
-
-#define MANIFOLD_CACHE_SIZE 4
-
-///b3ContactCache is a contact point cache, it stays persistent as long as objects are overlapping in the broadphase.
-///Those contact points are created by the collision narrow phase.
-///The cache can be empty, or hold 1,2,3 or 4 points. Some collision algorithms (GJK) might only add one point at a time.
-///updates/refreshes old contact points, and throw them away if necessary (distance becomes too large)
-///reduces the cache to 4 points, when more then 4 points are added, using following rules:
-///the contact point with deepest penetration is always kept, and it tries to maximuze the area covered by the points
-///note that some pairs of objects might have more then one contact manifold.
-B3_ATTRIBUTE_ALIGNED16( class) b3ContactCache
-{
-
-
-
-
- /// sort cached points so most isolated points come first
- int sortCachedPoints(const b3Vector3& pt);
-
-
-
-public:
-
- B3_DECLARE_ALIGNED_ALLOCATOR();
-
-
-
- int addManifoldPoint( const b3Vector3& newPoint);
-
- /*void replaceContactPoint(const b3Vector3& newPoint,int insertIndex)
- {
- b3Assert(validContactDistance(newPoint));
- m_pointCache[insertIndex] = newPoint;
- }
- */
-
-
-
- static bool validContactDistance(const b3Vector3& pt);
-
- /// calculated new worldspace coordinates and depth, and reject points that exceed the collision margin
- static void refreshContactPoints( const b3Transform& trA,const b3Transform& trB, struct b3Contact4Data& newContactCache);
-
- static void removeContactPoint(struct b3Contact4Data& newContactCache,int i);
-
-
-};
-
-
-
-#endif //B3_CONTACT_CACHE_H
/include/Bullet/Bullet3OpenCL/NarrowphaseCollision/b3ContactCache.h
Property changes:
Deleted: svn:executable
## -1 +0,0 ##
-*
\ No newline at end of property
Index: include/Bullet/Bullet3OpenCL/NarrowphaseCollision/b3StridingMeshInterface.h
===================================================================
--- include/Bullet/Bullet3OpenCL/NarrowphaseCollision/b3StridingMeshInterface.h (revision 22)
+++ include/Bullet/Bullet3OpenCL/NarrowphaseCollision/b3StridingMeshInterface.h (nonexistent)
@@ -1,167 +0,0 @@
-/*
-Bullet Continuous Collision Detection and Physics Library
-Copyright (c) 2003-2009 Erwin Coumans http://bulletphysics.org
-
-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.
-*/
-
-#ifndef B3_STRIDING_MESHINTERFACE_H
-#define B3_STRIDING_MESHINTERFACE_H
-
-#include "Bullet3Common/b3Vector3.h"
-#include "b3TriangleCallback.h"
-//#include "b3ConcaveShape.h"
-
-
-enum PHY_ScalarType {
- PHY_FLOAT, PHY_DOUBLE, PHY_INTEGER, PHY_SHORT,
- PHY_FIXEDPOINT88, PHY_UCHAR
-};
-
-
-/// The b3StridingMeshInterface is the interface class for high performance generic access to triangle meshes, used in combination with b3BvhTriangleMeshShape and some other collision shapes.
-/// Using index striding of 3*sizeof(integer) it can use triangle arrays, using index striding of 1*sizeof(integer) it can handle triangle strips.
-/// It allows for sharing graphics and collision meshes. Also it provides locking/unlocking of graphics meshes that are in gpu memory.
-B3_ATTRIBUTE_ALIGNED16(class ) b3StridingMeshInterface
-{
- protected:
-
- b3Vector3 m_scaling;
-
- public:
- B3_DECLARE_ALIGNED_ALLOCATOR();
-
- b3StridingMeshInterface() :m_scaling(b3MakeVector3(b3Scalar(1.),b3Scalar(1.),b3Scalar(1.)))
- {
-
- }
-
- virtual ~b3StridingMeshInterface();
-
-
-
- virtual void InternalProcessAllTriangles(b3InternalTriangleIndexCallback* callback,const b3Vector3& aabbMin,const b3Vector3& aabbMax) const;
-
- ///brute force method to calculate aabb
- void calculateAabbBruteForce(b3Vector3& aabbMin,b3Vector3& aabbMax);
-
- /// get read and write access to a subpart of a triangle mesh
- /// this subpart has a continuous array of vertices and indices
- /// in this way the mesh can be handled as chunks of memory with striding
- /// very similar to OpenGL vertexarray support
- /// make a call to unLockVertexBase when the read and write access is finished
- virtual void getLockedVertexIndexBase(unsigned char **vertexbase, int& numverts,PHY_ScalarType& type, int& stride,unsigned char **indexbase,int & indexstride,int& numfaces,PHY_ScalarType& indicestype,int subpart=0)=0;
-
- virtual void getLockedReadOnlyVertexIndexBase(const unsigned char **vertexbase, int& numverts,PHY_ScalarType& type, int& stride,const unsigned char **indexbase,int & indexstride,int& numfaces,PHY_ScalarType& indicestype,int subpart=0) const=0;
-
- /// unLockVertexBase finishes the access to a subpart of the triangle mesh
- /// make a call to unLockVertexBase when the read and write access (using getLockedVertexIndexBase) is finished
- virtual void unLockVertexBase(int subpart)=0;
-
- virtual void unLockReadOnlyVertexBase(int subpart) const=0;
-
-
- /// getNumSubParts returns the number of seperate subparts
- /// each subpart has a continuous array of vertices and indices
- virtual int getNumSubParts() const=0;
-
- virtual void preallocateVertices(int numverts)=0;
- virtual void preallocateIndices(int numindices)=0;
-
- virtual bool hasPremadeAabb() const { return false; }
- virtual void setPremadeAabb(const b3Vector3& aabbMin, const b3Vector3& aabbMax ) const
- {
- (void) aabbMin;
- (void) aabbMax;
- }
- virtual void getPremadeAabb(b3Vector3* aabbMin, b3Vector3* aabbMax ) const
- {
- (void) aabbMin;
- (void) aabbMax;
- }
-
- const b3Vector3& getScaling() const {
- return m_scaling;
- }
- void setScaling(const b3Vector3& scaling)
- {
- m_scaling = scaling;
- }
-
- virtual int calculateSerializeBufferSize() const;
-
- ///fills the dataBuffer and returns the struct name (and 0 on failure)
- //virtual const char* serialize(void* dataBuffer, b3Serializer* serializer) const;
-
-
-};
-
-struct b3IntIndexData
-{
- int m_value;
-};
-
-struct b3ShortIntIndexData
-{
- short m_value;
- char m_pad[2];
-};
-
-struct b3ShortIntIndexTripletData
-{
- short m_values[3];
- char m_pad[2];
-};
-
-struct b3CharIndexTripletData
-{
- unsigned char m_values[3];
- char m_pad;
-};
-
-
-///do not change those serialization structures, it requires an updated sBulletDNAstr/sBulletDNAstr64
-struct b3MeshPartData
-{
- b3Vector3FloatData *m_vertices3f;
- b3Vector3DoubleData *m_vertices3d;
-
- b3IntIndexData *m_indices32;
- b3ShortIntIndexTripletData *m_3indices16;
- b3CharIndexTripletData *m_3indices8;
-
- b3ShortIntIndexData *m_indices16;//backwards compatibility
-
- int m_numTriangles;//length of m_indices = m_numTriangles
- int m_numVertices;
-};
-
-
-///do not change those serialization structures, it requires an updated sBulletDNAstr/sBulletDNAstr64
-struct b3StridingMeshInterfaceData
-{
- b3MeshPartData *m_meshPartsPtr;
- b3Vector3FloatData m_scaling;
- int m_numMeshParts;
- char m_padding[4];
-};
-
-
-
-
-B3_FORCE_INLINE int b3StridingMeshInterface::calculateSerializeBufferSize() const
-{
- return sizeof(b3StridingMeshInterfaceData);
-}
-
-
-
-#endif //B3_STRIDING_MESHINTERFACE_H
/include/Bullet/Bullet3OpenCL/NarrowphaseCollision/b3StridingMeshInterface.h
Property changes:
Deleted: svn:executable
## -1 +0,0 ##
-*
\ No newline at end of property
Index: include/Bullet/Bullet3OpenCL/NarrowphaseCollision/b3GjkPairDetector.h
===================================================================
--- include/Bullet/Bullet3OpenCL/NarrowphaseCollision/b3GjkPairDetector.h (revision 22)
+++ include/Bullet/Bullet3OpenCL/NarrowphaseCollision/b3GjkPairDetector.h (nonexistent)
@@ -1,84 +0,0 @@
-
-
-
-
-#ifndef B3_GJK_PAIR_DETECTOR_H
-#define B3_GJK_PAIR_DETECTOR_H
-
-
-#include "Bullet3Common/b3Vector3.h"
-#include "Bullet3Common/b3AlignedObjectArray.h"
-
-class b3Transform;
-struct b3GjkEpaSolver2;
-class b3VoronoiSimplexSolver;
-struct b3ConvexPolyhedronData;
-
-B3_ATTRIBUTE_ALIGNED16(struct) b3GjkPairDetector
-{
-
-
- b3Vector3 m_cachedSeparatingAxis;
- b3GjkEpaSolver2* m_penetrationDepthSolver;
- b3VoronoiSimplexSolver* m_simplexSolver;
-
-
- bool m_ignoreMargin;
- b3Scalar m_cachedSeparatingDistance;
-
-
-public:
-
- //some debugging to fix degeneracy problems
- int m_lastUsedMethod;
- int m_curIter;
- int m_degenerateSimplex;
- int m_catchDegeneracies;
- int m_fixContactNormalDirection;
-
- b3GjkPairDetector(b3VoronoiSimplexSolver* simplexSolver,b3GjkEpaSolver2* penetrationDepthSolver);
-
- virtual ~b3GjkPairDetector() {};
-
-
- //void getClosestPoints(,Result& output);
-
- void setCachedSeperatingAxis(const b3Vector3& seperatingAxis)
- {
- m_cachedSeparatingAxis = seperatingAxis;
- }
-
- const b3Vector3& getCachedSeparatingAxis() const
- {
- return m_cachedSeparatingAxis;
- }
- b3Scalar getCachedSeparatingDistance() const
- {
- return m_cachedSeparatingDistance;
- }
-
- void setPenetrationDepthSolver(b3GjkEpaSolver2* penetrationDepthSolver)
- {
- m_penetrationDepthSolver = penetrationDepthSolver;
- }
-
- ///don't use setIgnoreMargin, it's for Bullet's internal use
- void setIgnoreMargin(bool ignoreMargin)
- {
- m_ignoreMargin = ignoreMargin;
- }
-
-
-};
-
-
-bool getClosestPoints(b3GjkPairDetector* gjkDetector, const b3Transform& transA, const b3Transform& transB,
- const b3ConvexPolyhedronData& hullA, const b3ConvexPolyhedronData& hullB,
- const b3AlignedObjectArray<b3Vector3>& verticesA,
- const b3AlignedObjectArray<b3Vector3>& verticesB,
- b3Scalar maximumDistanceSquared,
- b3Vector3& resultSepNormal,
- float& resultSepDistance,
- b3Vector3& resultPointOnB);
-
-#endif //B3_GJK_PAIR_DETECTOR_H
/include/Bullet/Bullet3OpenCL/NarrowphaseCollision/b3GjkPairDetector.h
Property changes:
Deleted: svn:executable
## -1 +0,0 ##
-*
\ No newline at end of property
Index: include/Bullet/Bullet3OpenCL/NarrowphaseCollision/b3GjkEpa.h
===================================================================
--- include/Bullet/Bullet3OpenCL/NarrowphaseCollision/b3GjkEpa.h (revision 22)
+++ include/Bullet/Bullet3OpenCL/NarrowphaseCollision/b3GjkEpa.h (nonexistent)
@@ -1,82 +0,0 @@
-/*
-Bullet Continuous Collision Detection and Physics Library
-Copyright (c) 2003-2008 Erwin Coumans http://continuousphysics.com/Bullet/
-
-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.
-*/
-
-/*
-GJK-EPA collision solver by Nathanael Presson, 2008
-*/
-#ifndef B3_GJK_EPA2_H
-#define B3_GJK_EPA2_H
-
-#include "Bullet3Common/b3AlignedObjectArray.h"
-#include "Bullet3Common/b3Transform.h"
-#include "Bullet3Collision/NarrowPhaseCollision/shared/b3ConvexPolyhedronData.h"
-
-
-///btGjkEpaSolver contributed under zlib by Nathanael Presson
-struct b3GjkEpaSolver2
-{
-struct sResults
- {
- enum eStatus
- {
- Separated, /* Shapes doesnt penetrate */
- Penetrating, /* Shapes are penetrating */
- GJK_Failed, /* GJK phase fail, no big issue, shapes are probably just 'touching' */
- EPA_Failed /* EPA phase fail, bigger problem, need to save parameters, and debug */
- } status;
- b3Vector3 witnesses[2];
- b3Vector3 normal;
- b3Scalar distance;
- };
-
-static int StackSizeRequirement();
-
-static bool Distance( const b3Transform& transA, const b3Transform& transB,
- const b3ConvexPolyhedronData* hullA, const b3ConvexPolyhedronData* hullB,
- const b3AlignedObjectArray<b3Vector3>& verticesA,
- const b3AlignedObjectArray<b3Vector3>& verticesB,
- const b3Vector3& guess,
- sResults& results);
-
-static bool Penetration( const b3Transform& transA, const b3Transform& transB,
- const b3ConvexPolyhedronData* hullA, const b3ConvexPolyhedronData* hullB,
- const b3AlignedObjectArray<b3Vector3>& verticesA,
- const b3AlignedObjectArray<b3Vector3>& verticesB,
- const b3Vector3& guess,
- sResults& results,
- bool usemargins=true);
-#if 0
-static b3Scalar SignedDistance( const b3Vector3& position,
- b3Scalar margin,
- const btConvexShape* shape,
- const btTransform& wtrs,
- sResults& results);
-
-static bool SignedDistance( const btConvexShape* shape0,const btTransform& wtrs0,
- const btConvexShape* shape1,const btTransform& wtrs1,
- const b3Vector3& guess,
- sResults& results);
-#endif
-
-};
-
-#endif //B3_GJK_EPA2_H
-
/include/Bullet/Bullet3OpenCL/NarrowphaseCollision/b3GjkEpa.h
Property changes:
Deleted: svn:executable
## -1 +0,0 ##
-*
\ No newline at end of property
Index: include/Bullet/Bullet3OpenCL/NarrowphaseCollision/b3TriangleCallback.h
===================================================================
--- include/Bullet/Bullet3OpenCL/NarrowphaseCollision/b3TriangleCallback.h (revision 22)
+++ include/Bullet/Bullet3OpenCL/NarrowphaseCollision/b3TriangleCallback.h (nonexistent)
@@ -1,42 +0,0 @@
-/*
-Bullet Continuous Collision Detection and Physics Library
-Copyright (c) 2003-2009 Erwin Coumans http://bulletphysics.org
-
-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.
-*/
-
-#ifndef B3_TRIANGLE_CALLBACK_H
-#define B3_TRIANGLE_CALLBACK_H
-
-#include "Bullet3Common/b3Vector3.h"
-
-
-///The b3TriangleCallback provides a callback for each overlapping triangle when calling processAllTriangles.
-///This callback is called by processAllTriangles for all b3ConcaveShape derived class, such as b3BvhTriangleMeshShape, b3StaticPlaneShape and b3HeightfieldTerrainShape.
-class b3TriangleCallback
-{
-public:
-
- virtual ~b3TriangleCallback();
- virtual void processTriangle(b3Vector3* triangle, int partId, int triangleIndex) = 0;
-};
-
-class b3InternalTriangleIndexCallback
-{
-public:
-
- virtual ~b3InternalTriangleIndexCallback();
- virtual void internalProcessTriangleIndex(b3Vector3* triangle,int partId,int triangleIndex) = 0;
-};
-
-
-
-#endif //B3_TRIANGLE_CALLBACK_H
/include/Bullet/Bullet3OpenCL/NarrowphaseCollision/b3TriangleCallback.h
Property changes:
Deleted: svn:executable
## -1 +0,0 ##
-*
\ No newline at end of property
Index: include/Bullet/Bullet3OpenCL/NarrowphaseCollision/b3ConvexHullContact.h
===================================================================
--- include/Bullet/Bullet3OpenCL/NarrowphaseCollision/b3ConvexHullContact.h (revision 22)
+++ include/Bullet/Bullet3OpenCL/NarrowphaseCollision/b3ConvexHullContact.h (nonexistent)
@@ -1,118 +0,0 @@
-
-#ifndef _CONVEX_HULL_CONTACT_H
-#define _CONVEX_HULL_CONTACT_H
-
-#include "Bullet3OpenCL/ParallelPrimitives/b3OpenCLArray.h"
-#include "Bullet3Collision/NarrowPhaseCollision/shared/b3RigidBodyData.h"
-#include "Bullet3Common/b3AlignedObjectArray.h"
-
-#include "Bullet3Collision/NarrowPhaseCollision/shared/b3ConvexPolyhedronData.h"
-#include "Bullet3Collision/NarrowPhaseCollision/shared/b3Collidable.h"
-#include "Bullet3Collision/NarrowPhaseCollision/b3Contact4.h"
-#include "Bullet3Common/shared/b3Int2.h"
-#include "Bullet3Common/shared/b3Int4.h"
-#include "b3OptimizedBvh.h"
-#include "b3BvhInfo.h"
-#include "Bullet3Collision/BroadPhaseCollision/shared/b3Aabb.h"
-
-//#include "../../dynamics/basic_demo/Stubs/ChNarrowPhase.h"
-
-
-
-
-struct GpuSatCollision
-{
- cl_context m_context;
- cl_device_id m_device;
- cl_command_queue m_queue;
- cl_kernel m_findSeparatingAxisKernel;
- cl_kernel m_mprPenetrationKernel;
- cl_kernel m_findSeparatingAxisUnitSphereKernel;
-
-
- cl_kernel m_findSeparatingAxisVertexFaceKernel;
- cl_kernel m_findSeparatingAxisEdgeEdgeKernel;
-
- cl_kernel m_findConcaveSeparatingAxisKernel;
- cl_kernel m_findConcaveSeparatingAxisVertexFaceKernel;
- cl_kernel m_findConcaveSeparatingAxisEdgeEdgeKernel;
-
-
-
-
- cl_kernel m_findCompoundPairsKernel;
- cl_kernel m_processCompoundPairsKernel;
-
- cl_kernel m_clipHullHullKernel;
- cl_kernel m_clipCompoundsHullHullKernel;
-
- cl_kernel m_clipFacesAndFindContacts;
- cl_kernel m_findClippingFacesKernel;
-
- cl_kernel m_clipHullHullConcaveConvexKernel;
-// cl_kernel m_extractManifoldAndAddContactKernel;
- cl_kernel m_newContactReductionKernel;
-
- cl_kernel m_bvhTraversalKernel;
- cl_kernel m_primitiveContactsKernel;
- cl_kernel m_findConcaveSphereContactsKernel;
-
- cl_kernel m_processCompoundPairsPrimitivesKernel;
-
- b3OpenCLArray<b3Vector3> m_unitSphereDirections;
-
- b3OpenCLArray<int> m_totalContactsOut;
-
- b3OpenCLArray<b3Vector3> m_sepNormals;
- b3OpenCLArray<float> m_dmins;
-
- b3OpenCLArray<int> m_hasSeparatingNormals;
- b3OpenCLArray<b3Vector3> m_concaveSepNormals;
- b3OpenCLArray<int> m_concaveHasSeparatingNormals;
- b3OpenCLArray<int> m_numConcavePairsOut;
- b3OpenCLArray<b3CompoundOverlappingPair> m_gpuCompoundPairs;
- b3OpenCLArray<b3Vector3> m_gpuCompoundSepNormals;
- b3OpenCLArray<int> m_gpuHasCompoundSepNormals;
- b3OpenCLArray<int> m_numCompoundPairsOut;
-
-
- GpuSatCollision(cl_context ctx,cl_device_id device, cl_command_queue q );
- virtual ~GpuSatCollision();
-
-
- void computeConvexConvexContactsGPUSAT( b3OpenCLArray<b3Int4>* pairs, int nPairs,
- const b3OpenCLArray<b3RigidBodyData>* bodyBuf,
- b3OpenCLArray<b3Contact4>* contactOut, int& nContacts,
- const b3OpenCLArray<b3Contact4>* oldContacts,
- int maxContactCapacity,
- int compoundPairCapacity,
- const b3OpenCLArray<b3ConvexPolyhedronData>& hostConvexData,
- const b3OpenCLArray<b3Vector3>& vertices,
- const b3OpenCLArray<b3Vector3>& uniqueEdges,
- const b3OpenCLArray<b3GpuFace>& faces,
- const b3OpenCLArray<int>& indices,
- const b3OpenCLArray<b3Collidable>& gpuCollidables,
- const b3OpenCLArray<b3GpuChildShape>& gpuChildShapes,
-
- const b3OpenCLArray<b3Aabb>& clAabbsWorldSpace,
- const b3OpenCLArray<b3Aabb>& clAabbsLocalSpace,
-
- b3OpenCLArray<b3Vector3>& worldVertsB1GPU,
- b3OpenCLArray<b3Int4>& clippingFacesOutGPU,
- b3OpenCLArray<b3Vector3>& worldNormalsAGPU,
- b3OpenCLArray<b3Vector3>& worldVertsA1GPU,
- b3OpenCLArray<b3Vector3>& worldVertsB2GPU,
- b3AlignedObjectArray<class b3OptimizedBvh*>& bvhData,
- b3OpenCLArray<b3QuantizedBvhNode>* treeNodesGPU,
- b3OpenCLArray<b3BvhSubtreeInfo>* subTreesGPU,
- b3OpenCLArray<b3BvhInfo>* bvhInfo,
- int numObjects,
- int maxTriConvexPairCapacity,
- b3OpenCLArray<b3Int4>& triangleConvexPairs,
- int& numTriConvexPairsOut
- );
-
-
-};
-
-#endif //_CONVEX_HULL_CONTACT_H
/include/Bullet/Bullet3OpenCL/NarrowphaseCollision/b3ConvexHullContact.h
Property changes:
Deleted: svn:executable
## -1 +0,0 ##
-*
\ No newline at end of property
Index: include/Bullet/Bullet3OpenCL/NarrowphaseCollision/b3OptimizedBvh.h
===================================================================
--- include/Bullet/Bullet3OpenCL/NarrowphaseCollision/b3OptimizedBvh.h (revision 22)
+++ include/Bullet/Bullet3OpenCL/NarrowphaseCollision/b3OptimizedBvh.h (nonexistent)
@@ -1,65 +0,0 @@
-/*
-Bullet Continuous Collision Detection and Physics Library
-Copyright (c) 2003-2009 Erwin Coumans http://bulletphysics.org
-
-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.
-*/
-
-///Contains contributions from Disney Studio's
-
-#ifndef B3_OPTIMIZED_BVH_H
-#define B3_OPTIMIZED_BVH_H
-
-#include "b3QuantizedBvh.h"
-
-class b3StridingMeshInterface;
-
-
-///The b3OptimizedBvh extends the b3QuantizedBvh to create AABB tree for triangle meshes, through the b3StridingMeshInterface.
-B3_ATTRIBUTE_ALIGNED16(class) b3OptimizedBvh : public b3QuantizedBvh
-{
-
-public:
- B3_DECLARE_ALIGNED_ALLOCATOR();
-
-protected:
-
-public:
-
- b3OptimizedBvh();
-
- virtual ~b3OptimizedBvh();
-
- void build(b3StridingMeshInterface* triangles,bool useQuantizedAabbCompression, const b3Vector3& bvhAabbMin, const b3Vector3& bvhAabbMax);
-
- void refit(b3StridingMeshInterface* triangles,const b3Vector3& aabbMin,const b3Vector3& aabbMax);
-
- void refitPartial(b3StridingMeshInterface* triangles,const b3Vector3& aabbMin, const b3Vector3& aabbMax);
-
- void updateBvhNodes(b3StridingMeshInterface* meshInterface,int firstNode,int endNode,int index);
-
- /// Data buffer MUST be 16 byte aligned
- virtual bool serializeInPlace(void *o_alignedDataBuffer, unsigned i_dataBufferSize, bool i_swapEndian) const
- {
- return b3QuantizedBvh::serialize(o_alignedDataBuffer,i_dataBufferSize,i_swapEndian);
-
- }
-
- ///deSerializeInPlace loads and initializes a BVH from a buffer in memory 'in place'
- static b3OptimizedBvh *deSerializeInPlace(void *i_alignedDataBuffer, unsigned int i_dataBufferSize, bool i_swapEndian);
-
-
-};
-
-
-#endif //B3_OPTIMIZED_BVH_H
-
-
/include/Bullet/Bullet3OpenCL/NarrowphaseCollision/b3OptimizedBvh.h
Property changes:
Deleted: svn:executable
## -1 +0,0 ##
-*
\ No newline at end of property
Index: include/Bullet/Bullet3OpenCL/NarrowphaseCollision/b3QuantizedBvh.h
===================================================================
--- include/Bullet/Bullet3OpenCL/NarrowphaseCollision/b3QuantizedBvh.h (revision 22)
+++ include/Bullet/Bullet3OpenCL/NarrowphaseCollision/b3QuantizedBvh.h (nonexistent)
@@ -1,556 +0,0 @@
-/*
-Bullet Continuous Collision Detection and Physics Library
-Copyright (c) 2003-2006 Erwin Coumans http://continuousphysics.com/Bullet/
-
-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.
-*/
-
-#ifndef B3_QUANTIZED_BVH_H
-#define B3_QUANTIZED_BVH_H
-
-class b3Serializer;
-
-//#define DEBUG_CHECK_DEQUANTIZATION 1
-#ifdef DEBUG_CHECK_DEQUANTIZATION
-#ifdef __SPU__
-#define printf spu_printf
-#endif //__SPU__
-
-#include <stdio.h>
-#include <stdlib.h>
-#endif //DEBUG_CHECK_DEQUANTIZATION
-
-#include "Bullet3Common/b3Vector3.h"
-#include "Bullet3Common/b3AlignedAllocator.h"
-
-#ifdef B3_USE_DOUBLE_PRECISION
-#define b3QuantizedBvhData b3QuantizedBvhDoubleData
-#define b3OptimizedBvhNodeData b3OptimizedBvhNodeDoubleData
-#define b3QuantizedBvhDataName "b3QuantizedBvhDoubleData"
-#else
-#define b3QuantizedBvhData b3QuantizedBvhFloatData
-#define b3OptimizedBvhNodeData b3OptimizedBvhNodeFloatData
-#define b3QuantizedBvhDataName "b3QuantizedBvhFloatData"
-#endif
-
-#include "Bullet3Collision/NarrowPhaseCollision/shared/b3QuantizedBvhNodeData.h"
-#include "Bullet3Collision/NarrowPhaseCollision/shared/b3BvhSubtreeInfoData.h"
-
-
-
-//http://msdn.microsoft.com/library/default.asp?url=/library/en-us/vclang/html/vclrf__m128.asp
-
-
-//Note: currently we have 16 bytes per quantized node
-#define MAX_SUBTREE_SIZE_IN_BYTES 2048
-
-// 10 gives the potential for 1024 parts, with at most 2^21 (2097152) (minus one
-// actually) triangles each (since the sign bit is reserved
-#define MAX_NUM_PARTS_IN_BITS 10
-
-///b3QuantizedBvhNode is a compressed aabb node, 16 bytes.
-///Node can be used for leafnode or internal node. Leafnodes can point to 32-bit triangle index (non-negative range).
-B3_ATTRIBUTE_ALIGNED16 (struct) b3QuantizedBvhNode : public b3QuantizedBvhNodeData
-{
- B3_DECLARE_ALIGNED_ALLOCATOR();
-
- bool isLeafNode() const
- {
- //skipindex is negative (internal node), triangleindex >=0 (leafnode)
- return (m_escapeIndexOrTriangleIndex >= 0);
- }
- int getEscapeIndex() const
- {
- b3Assert(!isLeafNode());
- return -m_escapeIndexOrTriangleIndex;
- }
- int getTriangleIndex() const
- {
- b3Assert(isLeafNode());
- unsigned int x=0;
- unsigned int y = (~(x&0))<<(31-MAX_NUM_PARTS_IN_BITS);
- // Get only the lower bits where the triangle index is stored
- return (m_escapeIndexOrTriangleIndex&~(y));
- }
- int getPartId() const
- {
- b3Assert(isLeafNode());
- // Get only the highest bits where the part index is stored
- return (m_escapeIndexOrTriangleIndex>>(31-MAX_NUM_PARTS_IN_BITS));
- }
-}
-;
-
-/// b3OptimizedBvhNode contains both internal and leaf node information.
-/// Total node size is 44 bytes / node. You can use the compressed version of 16 bytes.
-B3_ATTRIBUTE_ALIGNED16 (struct) b3OptimizedBvhNode
-{
- B3_DECLARE_ALIGNED_ALLOCATOR();
-
- //32 bytes
- b3Vector3 m_aabbMinOrg;
- b3Vector3 m_aabbMaxOrg;
-
- //4
- int m_escapeIndex;
-
- //8
- //for child nodes
- int m_subPart;
- int m_triangleIndex;
-
-//pad the size to 64 bytes
- char m_padding[20];
-};
-
-
-///b3BvhSubtreeInfo provides info to gather a subtree of limited size
-B3_ATTRIBUTE_ALIGNED16(class) b3BvhSubtreeInfo : public b3BvhSubtreeInfoData
-{
-public:
- B3_DECLARE_ALIGNED_ALLOCATOR();
-
- b3BvhSubtreeInfo()
- {
- //memset(&m_padding[0], 0, sizeof(m_padding));
- }
-
-
- void setAabbFromQuantizeNode(const b3QuantizedBvhNode& quantizedNode)
- {
- m_quantizedAabbMin[0] = quantizedNode.m_quantizedAabbMin[0];
- m_quantizedAabbMin[1] = quantizedNode.m_quantizedAabbMin[1];
- m_quantizedAabbMin[2] = quantizedNode.m_quantizedAabbMin[2];
- m_quantizedAabbMax[0] = quantizedNode.m_quantizedAabbMax[0];
- m_quantizedAabbMax[1] = quantizedNode.m_quantizedAabbMax[1];
- m_quantizedAabbMax[2] = quantizedNode.m_quantizedAabbMax[2];
- }
-}
-;
-
-
-class b3NodeOverlapCallback
-{
-public:
- virtual ~b3NodeOverlapCallback() {};
-
- virtual void processNode(int subPart, int triangleIndex) = 0;
-};
-
-#include "Bullet3Common/b3AlignedAllocator.h"
-#include "Bullet3Common/b3AlignedObjectArray.h"
-
-
-
-///for code readability:
-typedef b3AlignedObjectArray<b3OptimizedBvhNode> NodeArray;
-typedef b3AlignedObjectArray<b3QuantizedBvhNode> QuantizedNodeArray;
-typedef b3AlignedObjectArray<b3BvhSubtreeInfo> BvhSubtreeInfoArray;
-
-
-///The b3QuantizedBvh class stores an AABB tree that can be quickly traversed on CPU and Cell SPU.
-///It is used by the b3BvhTriangleMeshShape as midphase, and by the b3MultiSapBroadphase.
-///It is recommended to use quantization for better performance and lower memory requirements.
-B3_ATTRIBUTE_ALIGNED16(class) b3QuantizedBvh
-{
-public:
- enum b3TraversalMode
- {
- TRAVERSAL_STACKLESS = 0,
- TRAVERSAL_STACKLESS_CACHE_FRIENDLY,
- TRAVERSAL_RECURSIVE
- };
-
-
-
-
- b3Vector3 m_bvhAabbMin;
- b3Vector3 m_bvhAabbMax;
- b3Vector3 m_bvhQuantization;
-
-protected:
- int m_bulletVersion; //for serialization versioning. It could also be used to detect endianess.
-
- int m_curNodeIndex;
- //quantization data
- bool m_useQuantization;
-
-
-
- NodeArray m_leafNodes;
- NodeArray m_contiguousNodes;
- QuantizedNodeArray m_quantizedLeafNodes;
- QuantizedNodeArray m_quantizedContiguousNodes;
-
- b3TraversalMode m_traversalMode;
- BvhSubtreeInfoArray m_SubtreeHeaders;
-
- //This is only used for serialization so we don't have to add serialization directly to b3AlignedObjectArray
- mutable int m_subtreeHeaderCount;
-
-
-
-
-
- ///two versions, one for quantized and normal nodes. This allows code-reuse while maintaining readability (no template/macro!)
- ///this might be refactored into a virtual, it is usually not calculated at run-time
- void setInternalNodeAabbMin(int nodeIndex, const b3Vector3& aabbMin)
- {
- if (m_useQuantization)
- {
- quantize(&m_quantizedContiguousNodes[nodeIndex].m_quantizedAabbMin[0] ,aabbMin,0);
- } else
- {
- m_contiguousNodes[nodeIndex].m_aabbMinOrg = aabbMin;
-
- }
- }
- void setInternalNodeAabbMax(int nodeIndex,const b3Vector3& aabbMax)
- {
- if (m_useQuantization)
- {
- quantize(&m_quantizedContiguousNodes[nodeIndex].m_quantizedAabbMax[0],aabbMax,1);
- } else
- {
- m_contiguousNodes[nodeIndex].m_aabbMaxOrg = aabbMax;
- }
- }
-
- b3Vector3 getAabbMin(int nodeIndex) const
- {
- if (m_useQuantization)
- {
- return unQuantize(&m_quantizedLeafNodes[nodeIndex].m_quantizedAabbMin[0]);
- }
- //non-quantized
- return m_leafNodes[nodeIndex].m_aabbMinOrg;
-
- }
- b3Vector3 getAabbMax(int nodeIndex) const
- {
- if (m_useQuantization)
- {
- return unQuantize(&m_quantizedLeafNodes[nodeIndex].m_quantizedAabbMax[0]);
- }
- //non-quantized
- return m_leafNodes[nodeIndex].m_aabbMaxOrg;
-
- }
-
-
- void setInternalNodeEscapeIndex(int nodeIndex, int escapeIndex)
- {
- if (m_useQuantization)
- {
- m_quantizedContiguousNodes[nodeIndex].m_escapeIndexOrTriangleIndex = -escapeIndex;
- }
- else
- {
- m_contiguousNodes[nodeIndex].m_escapeIndex = escapeIndex;
- }
-
- }
-
- void mergeInternalNodeAabb(int nodeIndex,const b3Vector3& newAabbMin,const b3Vector3& newAabbMax)
- {
- if (m_useQuantization)
- {
- unsigned short int quantizedAabbMin[3];
- unsigned short int quantizedAabbMax[3];
- quantize(quantizedAabbMin,newAabbMin,0);
- quantize(quantizedAabbMax,newAabbMax,1);
- for (int i=0;i<3;i++)
- {
- if (m_quantizedContiguousNodes[nodeIndex].m_quantizedAabbMin[i] > quantizedAabbMin[i])
- m_quantizedContiguousNodes[nodeIndex].m_quantizedAabbMin[i] = quantizedAabbMin[i];
-
- if (m_quantizedContiguousNodes[nodeIndex].m_quantizedAabbMax[i] < quantizedAabbMax[i])
- m_quantizedContiguousNodes[nodeIndex].m_quantizedAabbMax[i] = quantizedAabbMax[i];
-
- }
- } else
- {
- //non-quantized
- m_contiguousNodes[nodeIndex].m_aabbMinOrg.setMin(newAabbMin);
- m_contiguousNodes[nodeIndex].m_aabbMaxOrg.setMax(newAabbMax);
- }
- }
-
- void swapLeafNodes(int firstIndex,int secondIndex);
-
- void assignInternalNodeFromLeafNode(int internalNode,int leafNodeIndex);
-
-protected:
-
-
-
- void buildTree (int startIndex,int endIndex);
-
- int calcSplittingAxis(int startIndex,int endIndex);
-
- int sortAndCalcSplittingIndex(int startIndex,int endIndex,int splitAxis);
-
- void walkStacklessTree(b3NodeOverlapCallback* nodeCallback,const b3Vector3& aabbMin,const b3Vector3& aabbMax) const;
-
- void walkStacklessQuantizedTreeAgainstRay(b3NodeOverlapCallback* nodeCallback, const b3Vector3& raySource, const b3Vector3& rayTarget, const b3Vector3& aabbMin, const b3Vector3& aabbMax, int startNodeIndex,int endNodeIndex) const;
- void walkStacklessQuantizedTree(b3NodeOverlapCallback* nodeCallback,unsigned short int* quantizedQueryAabbMin,unsigned short int* quantizedQueryAabbMax,int startNodeIndex,int endNodeIndex) const;
- void walkStacklessTreeAgainstRay(b3NodeOverlapCallback* nodeCallback, const b3Vector3& raySource, const b3Vector3& rayTarget, const b3Vector3& aabbMin, const b3Vector3& aabbMax, int startNodeIndex,int endNodeIndex) const;
-
- ///tree traversal designed for small-memory processors like PS3 SPU
- void walkStacklessQuantizedTreeCacheFriendly(b3NodeOverlapCallback* nodeCallback,unsigned short int* quantizedQueryAabbMin,unsigned short int* quantizedQueryAabbMax) const;
-
- ///use the 16-byte stackless 'skipindex' node tree to do a recursive traversal
- void walkRecursiveQuantizedTreeAgainstQueryAabb(const b3QuantizedBvhNode* currentNode,b3NodeOverlapCallback* nodeCallback,unsigned short int* quantizedQueryAabbMin,unsigned short int* quantizedQueryAabbMax) const;
-
- ///use the 16-byte stackless 'skipindex' node tree to do a recursive traversal
- void walkRecursiveQuantizedTreeAgainstQuantizedTree(const b3QuantizedBvhNode* treeNodeA,const b3QuantizedBvhNode* treeNodeB,b3NodeOverlapCallback* nodeCallback) const;
-
-
-
-
- void updateSubtreeHeaders(int leftChildNodexIndex,int rightChildNodexIndex);
-
-public:
-
- B3_DECLARE_ALIGNED_ALLOCATOR();
-
- b3QuantizedBvh();
-
- virtual ~b3QuantizedBvh();
-
-
- ///***************************************** expert/internal use only *************************
- void setQuantizationValues(const b3Vector3& bvhAabbMin,const b3Vector3& bvhAabbMax,b3Scalar quantizationMargin=b3Scalar(1.0));
- QuantizedNodeArray& getLeafNodeArray() { return m_quantizedLeafNodes; }
- ///buildInternal is expert use only: assumes that setQuantizationValues and LeafNodeArray are initialized
- void buildInternal();
- ///***************************************** expert/internal use only *************************
-
- void reportAabbOverlappingNodex(b3NodeOverlapCallback* nodeCallback,const b3Vector3& aabbMin,const b3Vector3& aabbMax) const;
- void reportRayOverlappingNodex (b3NodeOverlapCallback* nodeCallback, const b3Vector3& raySource, const b3Vector3& rayTarget) const;
- void reportBoxCastOverlappingNodex(b3NodeOverlapCallback* nodeCallback, const b3Vector3& raySource, const b3Vector3& rayTarget, const b3Vector3& aabbMin,const b3Vector3& aabbMax) const;
-
- B3_FORCE_INLINE void quantize(unsigned short* out, const b3Vector3& point,int isMax) const
- {
-
- b3Assert(m_useQuantization);
-
- b3Assert(point.getX() <= m_bvhAabbMax.getX());
- b3Assert(point.getY() <= m_bvhAabbMax.getY());
- b3Assert(point.getZ() <= m_bvhAabbMax.getZ());
-
- b3Assert(point.getX() >= m_bvhAabbMin.getX());
- b3Assert(point.getY() >= m_bvhAabbMin.getY());
- b3Assert(point.getZ() >= m_bvhAabbMin.getZ());
-
- b3Vector3 v = (point - m_bvhAabbMin) * m_bvhQuantization;
- ///Make sure rounding is done in a way that unQuantize(quantizeWithClamp(...)) is conservative
- ///end-points always set the first bit, so that they are sorted properly (so that neighbouring AABBs overlap properly)
- ///@todo: double-check this
- if (isMax)
- {
- out[0] = (unsigned short) (((unsigned short)(v.getX()+b3Scalar(1.)) | 1));
- out[1] = (unsigned short) (((unsigned short)(v.getY()+b3Scalar(1.)) | 1));
- out[2] = (unsigned short) (((unsigned short)(v.getZ()+b3Scalar(1.)) | 1));
- } else
- {
- out[0] = (unsigned short) (((unsigned short)(v.getX()) & 0xfffe));
- out[1] = (unsigned short) (((unsigned short)(v.getY()) & 0xfffe));
- out[2] = (unsigned short) (((unsigned short)(v.getZ()) & 0xfffe));
- }
-
-
-#ifdef DEBUG_CHECK_DEQUANTIZATION
- b3Vector3 newPoint = unQuantize(out);
- if (isMax)
- {
- if (newPoint.getX() < point.getX())
- {
- printf("unconservative X, diffX = %f, oldX=%f,newX=%f\n",newPoint.getX()-point.getX(), newPoint.getX(),point.getX());
- }
- if (newPoint.getY() < point.getY())
- {
- printf("unconservative Y, diffY = %f, oldY=%f,newY=%f\n",newPoint.getY()-point.getY(), newPoint.getY(),point.getY());
- }
- if (newPoint.getZ() < point.getZ())
- {
-
- printf("unconservative Z, diffZ = %f, oldZ=%f,newZ=%f\n",newPoint.getZ()-point.getZ(), newPoint.getZ(),point.getZ());
- }
- } else
- {
- if (newPoint.getX() > point.getX())
- {
- printf("unconservative X, diffX = %f, oldX=%f,newX=%f\n",newPoint.getX()-point.getX(), newPoint.getX(),point.getX());
- }
- if (newPoint.getY() > point.getY())
- {
- printf("unconservative Y, diffY = %f, oldY=%f,newY=%f\n",newPoint.getY()-point.getY(), newPoint.getY(),point.getY());
- }
- if (newPoint.getZ() > point.getZ())
- {
- printf("unconservative Z, diffZ = %f, oldZ=%f,newZ=%f\n",newPoint.getZ()-point.getZ(), newPoint.getZ(),point.getZ());
- }
- }
-#endif //DEBUG_CHECK_DEQUANTIZATION
-
- }
-
-
- B3_FORCE_INLINE void quantizeWithClamp(unsigned short* out, const b3Vector3& point2,int isMax) const
- {
-
- b3Assert(m_useQuantization);
-
- b3Vector3 clampedPoint(point2);
- clampedPoint.setMax(m_bvhAabbMin);
- clampedPoint.setMin(m_bvhAabbMax);
-
- quantize(out,clampedPoint,isMax);
-
- }
-
- B3_FORCE_INLINE b3Vector3 unQuantize(const unsigned short* vecIn) const
- {
- b3Vector3 vecOut;
- vecOut.setValue(
- (b3Scalar)(vecIn[0]) / (m_bvhQuantization.getX()),
- (b3Scalar)(vecIn[1]) / (m_bvhQuantization.getY()),
- (b3Scalar)(vecIn[2]) / (m_bvhQuantization.getZ()));
- vecOut += m_bvhAabbMin;
- return vecOut;
- }
-
- ///setTraversalMode let's you choose between stackless, recursive or stackless cache friendly tree traversal. Note this is only implemented for quantized trees.
- void setTraversalMode(b3TraversalMode traversalMode)
- {
- m_traversalMode = traversalMode;
- }
-
-
- B3_FORCE_INLINE QuantizedNodeArray& getQuantizedNodeArray()
- {
- return m_quantizedContiguousNodes;
- }
-
-
- B3_FORCE_INLINE BvhSubtreeInfoArray& getSubtreeInfoArray()
- {
- return m_SubtreeHeaders;
- }
-
-////////////////////////////////////////////////////////////////////
-
- /////Calculate space needed to store BVH for serialization
- unsigned calculateSerializeBufferSize() const;
-
- /// Data buffer MUST be 16 byte aligned
- virtual bool serialize(void *o_alignedDataBuffer, unsigned i_dataBufferSize, bool i_swapEndian) const;
-
- ///deSerializeInPlace loads and initializes a BVH from a buffer in memory 'in place'
- static b3QuantizedBvh *deSerializeInPlace(void *i_alignedDataBuffer, unsigned int i_dataBufferSize, bool i_swapEndian);
-
- static unsigned int getAlignmentSerializationPadding();
-//////////////////////////////////////////////////////////////////////
-
-
- virtual int calculateSerializeBufferSizeNew() const;
-
- ///fills the dataBuffer and returns the struct name (and 0 on failure)
- virtual const char* serialize(void* dataBuffer, b3Serializer* serializer) const;
-
- virtual void deSerializeFloat(struct b3QuantizedBvhFloatData& quantizedBvhFloatData);
-
- virtual void deSerializeDouble(struct b3QuantizedBvhDoubleData& quantizedBvhDoubleData);
-
-
-////////////////////////////////////////////////////////////////////
-
- B3_FORCE_INLINE bool isQuantized()
- {
- return m_useQuantization;
- }
-
-private:
- // Special "copy" constructor that allows for in-place deserialization
- // Prevents b3Vector3's default constructor from being called, but doesn't inialize much else
- // ownsMemory should most likely be false if deserializing, and if you are not, don't call this (it also changes the function signature, which we need)
- b3QuantizedBvh(b3QuantizedBvh &other, bool ownsMemory);
-
-}
-;
-
-
-struct b3OptimizedBvhNodeFloatData
-{
- b3Vector3FloatData m_aabbMinOrg;
- b3Vector3FloatData m_aabbMaxOrg;
- int m_escapeIndex;
- int m_subPart;
- int m_triangleIndex;
- char m_pad[4];
-};
-
-struct b3OptimizedBvhNodeDoubleData
-{
- b3Vector3DoubleData m_aabbMinOrg;
- b3Vector3DoubleData m_aabbMaxOrg;
- int m_escapeIndex;
- int m_subPart;
- int m_triangleIndex;
- char m_pad[4];
-};
-
-
-
-struct b3QuantizedBvhFloatData
-{
- b3Vector3FloatData m_bvhAabbMin;
- b3Vector3FloatData m_bvhAabbMax;
- b3Vector3FloatData m_bvhQuantization;
- int m_curNodeIndex;
- int m_useQuantization;
- int m_numContiguousLeafNodes;
- int m_numQuantizedContiguousNodes;
- b3OptimizedBvhNodeFloatData *m_contiguousNodesPtr;
- b3QuantizedBvhNodeData *m_quantizedContiguousNodesPtr;
- b3BvhSubtreeInfoData *m_subTreeInfoPtr;
- int m_traversalMode;
- int m_numSubtreeHeaders;
-
-};
-
-struct b3QuantizedBvhDoubleData
-{
- b3Vector3DoubleData m_bvhAabbMin;
- b3Vector3DoubleData m_bvhAabbMax;
- b3Vector3DoubleData m_bvhQuantization;
- int m_curNodeIndex;
- int m_useQuantization;
- int m_numContiguousLeafNodes;
- int m_numQuantizedContiguousNodes;
- b3OptimizedBvhNodeDoubleData *m_contiguousNodesPtr;
- b3QuantizedBvhNodeData *m_quantizedContiguousNodesPtr;
-
- int m_traversalMode;
- int m_numSubtreeHeaders;
- b3BvhSubtreeInfoData *m_subTreeInfoPtr;
-};
-
-
-B3_FORCE_INLINE int b3QuantizedBvh::calculateSerializeBufferSizeNew() const
-{
- return sizeof(b3QuantizedBvhData);
-}
-
-
-
-#endif //B3_QUANTIZED_BVH_H
/include/Bullet/Bullet3OpenCL/NarrowphaseCollision/b3QuantizedBvh.h
Property changes:
Deleted: svn:executable
## -1 +0,0 ##
-*
\ No newline at end of property
Index: include/Bullet/Bullet3OpenCL/NarrowphaseCollision/b3ConvexPolyhedronCL.h
===================================================================
--- include/Bullet/Bullet3OpenCL/NarrowphaseCollision/b3ConvexPolyhedronCL.h (revision 22)
+++ include/Bullet/Bullet3OpenCL/NarrowphaseCollision/b3ConvexPolyhedronCL.h (nonexistent)
@@ -1,9 +0,0 @@
-#ifndef CONVEX_POLYHEDRON_CL
-#define CONVEX_POLYHEDRON_CL
-
-#include "Bullet3Common/b3Transform.h"
-#include "Bullet3Collision/NarrowPhaseCollision/shared/b3ConvexPolyhedronData.h"
-
-
-
-#endif //CONVEX_POLYHEDRON_CL
/include/Bullet/Bullet3OpenCL/NarrowphaseCollision/b3ConvexPolyhedronCL.h
Property changes:
Deleted: svn:executable
## -1 +0,0 ##
-*
\ No newline at end of property
Index: include/Bullet/Bullet3OpenCL/NarrowphaseCollision/kernels/mprKernels.h
===================================================================
--- include/Bullet/Bullet3OpenCL/NarrowphaseCollision/kernels/mprKernels.h (revision 22)
+++ include/Bullet/Bullet3OpenCL/NarrowphaseCollision/kernels/mprKernels.h (nonexistent)
@@ -1,1440 +0,0 @@
-//this file is autogenerated using stringify.bat (premake --stringify) in the build folder of this project
-static const char* mprKernelsCL= \
-"/***\n"
-" * ---------------------------------\n"
-" * Copyright (c)2012 Daniel Fiser <danfis@danfis.cz>\n"
-" *\n"
-" * This file was ported from mpr.c file, part of libccd.\n"
-" * The Minkoski Portal Refinement implementation was ported \n"
-" * to OpenCL by Erwin Coumans for the Bullet 3 Physics library.\n"
-" * at http://github.com/erwincoumans/bullet3\n"
-" *\n"
-" * Distributed under the OSI-approved BSD License (the \"License\");\n"
-" * see <http://www.opensource.org/licenses/bsd-license.php>.\n"
-" * This software is distributed WITHOUT ANY WARRANTY; without even the\n"
-" * implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.\n"
-" * See the License for more information.\n"
-" */\n"
-"#ifndef B3_MPR_PENETRATION_H\n"
-"#define B3_MPR_PENETRATION_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"
-"//keep B3_LARGE_FLOAT*B3_LARGE_FLOAT < FLT_MAX\n"
-"#define B3_LARGE_FLOAT 1e18f\n"
-"#define B3_INFINITY 1e18f\n"
-"#define b3Assert(a)\n"
-"#define b3ConstArray(a) __global const a*\n"
-"#define b3AtomicInc atomic_inc\n"
-"#define b3AtomicAdd atomic_add\n"
-"#define b3Fabs fabs\n"
-"#define b3Sqrt native_sqrt\n"
-"#define b3Sin native_sin\n"
-"#define b3Cos native_cos\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"
-" #define b3MinFloat4 min\n"
-" #define b3MaxFloat4 max\n"
-" #define b3Normalized(a) normalize(a)\n"
-"#endif \n"
-" \n"
-"inline bool b3IsAlmostZero(b3Float4ConstArg v)\n"
-"{\n"
-" if(b3Fabs(v.x)>1e-6 || b3Fabs(v.y)>1e-6 || b3Fabs(v.z)>1e-6) \n"
-" return false;\n"
-" return true;\n"
-"}\n"
-"inline int b3MaxDot( b3Float4ConstArg vec, __global const b3Float4* vecArray, int vecLen, float* dotOut )\n"
-"{\n"
-" float maxDot = -B3_INFINITY;\n"
-" int i = 0;\n"
-" int ptIndex = -1;\n"
-" for( i = 0; i < vecLen; i++ )\n"
-" {\n"
-" float dot = b3Dot3F4(vecArray[i],vec);\n"
-" \n"
-" if( dot > maxDot )\n"
-" {\n"
-" maxDot = dot;\n"
-" ptIndex = i;\n"
-" }\n"
-" }\n"
-" b3Assert(ptIndex>=0);\n"
-" if (ptIndex<0)\n"
-" {\n"
-" ptIndex = 0;\n"
-" }\n"
-" *dotOut = maxDot;\n"
-" return ptIndex;\n"
-"}\n"
-"#endif //B3_FLOAT4_H\n"
-"#ifndef B3_RIGIDBODY_DATA_H\n"
-"#define B3_RIGIDBODY_DATA_H\n"
-"#ifndef B3_FLOAT4_H\n"
-"#ifdef __cplusplus\n"
-"#else\n"
-"#endif \n"
-"#endif //B3_FLOAT4_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 b3QuatNormalized(b3QuatConstArg in);\n"
-"inline b3Quat b3QuatRotate(b3QuatConstArg q, b3QuatConstArg vec);\n"
-"inline b3Quat b3QuatInvert(b3QuatConstArg q);\n"
-"inline b3Quat b3QuatInverse(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 b3QuatNormalized(b3QuatConstArg in)\n"
-"{\n"
-" b3Quat q;\n"
-" q=in;\n"
-" //return b3FastNormalize4(in);\n"
-" float len = native_sqrt(dot(q, q));\n"
-" if(len > 0.f)\n"
-" {\n"
-" q *= 1.f / len;\n"
-" }\n"
-" else\n"
-" {\n"
-" q.x = q.y = q.z = 0.f;\n"
-" q.w = 1.f;\n"
-" }\n"
-" return q;\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 b3QuatInverse(b3QuatConstArg q)\n"
-"{\n"
-" return (b3Quat)(-q.xyz, q.w);\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"
-"#ifndef B3_MAT3x3_H\n"
-"#define B3_MAT3x3_H\n"
-"#ifndef B3_QUAT_H\n"
-"#ifdef __cplusplus\n"
-"#else\n"
-"#endif \n"
-"#endif //B3_QUAT_H\n"
-"#ifdef __cplusplus\n"
-"#else\n"
-"typedef struct\n"
-"{\n"
-" b3Float4 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"
-" b3Float4 quat2 = (b3Float4)(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"
-"__inline\n"
-"b3Mat3x3 mtZero();\n"
-"__inline\n"
-"b3Mat3x3 mtIdentity();\n"
-"__inline\n"
-"b3Mat3x3 mtTranspose(b3Mat3x3 m);\n"
-"__inline\n"
-"b3Mat3x3 mtMul(b3Mat3x3 a, b3Mat3x3 b);\n"
-"__inline\n"
-"b3Float4 mtMul1(b3Mat3x3 a, b3Float4 b);\n"
-"__inline\n"
-"b3Float4 mtMul3(b3Float4 a, b3Mat3x3 b);\n"
-"__inline\n"
-"b3Mat3x3 mtZero()\n"
-"{\n"
-" b3Mat3x3 m;\n"
-" m.m_row[0] = (b3Float4)(0.f);\n"
-" m.m_row[1] = (b3Float4)(0.f);\n"
-" m.m_row[2] = (b3Float4)(0.f);\n"
-" return m;\n"
-"}\n"
-"__inline\n"
-"b3Mat3x3 mtIdentity()\n"
-"{\n"
-" b3Mat3x3 m;\n"
-" m.m_row[0] = (b3Float4)(1,0,0,0);\n"
-" m.m_row[1] = (b3Float4)(0,1,0,0);\n"
-" m.m_row[2] = (b3Float4)(0,0,1,0);\n"
-" return m;\n"
-"}\n"
-"__inline\n"
-"b3Mat3x3 mtTranspose(b3Mat3x3 m)\n"
-"{\n"
-" b3Mat3x3 out;\n"
-" out.m_row[0] = (b3Float4)(m.m_row[0].x, m.m_row[1].x, m.m_row[2].x, 0.f);\n"
-" out.m_row[1] = (b3Float4)(m.m_row[0].y, m.m_row[1].y, m.m_row[2].y, 0.f);\n"
-" out.m_row[2] = (b3Float4)(m.m_row[0].z, m.m_row[1].z, m.m_row[2].z, 0.f);\n"
-" return out;\n"
-"}\n"
-"__inline\n"
-"b3Mat3x3 mtMul(b3Mat3x3 a, b3Mat3x3 b)\n"
-"{\n"
-" b3Mat3x3 transB;\n"
-" transB = mtTranspose( b );\n"
-" b3Mat3x3 ans;\n"
-" // why this doesn't run when 0ing in the for{}\n"
-" a.m_row[0].w = 0.f;\n"
-" a.m_row[1].w = 0.f;\n"
-" a.m_row[2].w = 0.f;\n"
-" for(int i=0; i<3; i++)\n"
-" {\n"
-"// a.m_row[i].w = 0.f;\n"
-" ans.m_row[i].x = b3Dot3F4(a.m_row[i],transB.m_row[0]);\n"
-" ans.m_row[i].y = b3Dot3F4(a.m_row[i],transB.m_row[1]);\n"
-" ans.m_row[i].z = b3Dot3F4(a.m_row[i],transB.m_row[2]);\n"
-" ans.m_row[i].w = 0.f;\n"
-" }\n"
-" return ans;\n"
-"}\n"
-"__inline\n"
-"b3Float4 mtMul1(b3Mat3x3 a, b3Float4 b)\n"
-"{\n"
-" b3Float4 ans;\n"
-" ans.x = b3Dot3F4( a.m_row[0], b );\n"
-" ans.y = b3Dot3F4( a.m_row[1], b );\n"
-" ans.z = b3Dot3F4( a.m_row[2], b );\n"
-" ans.w = 0.f;\n"
-" return ans;\n"
-"}\n"
-"__inline\n"
-"b3Float4 mtMul3(b3Float4 a, b3Mat3x3 b)\n"
-"{\n"
-" b3Float4 colx = b3MakeFloat4(b.m_row[0].x, b.m_row[1].x, b.m_row[2].x, 0);\n"
-" b3Float4 coly = b3MakeFloat4(b.m_row[0].y, b.m_row[1].y, b.m_row[2].y, 0);\n"
-" b3Float4 colz = b3MakeFloat4(b.m_row[0].z, b.m_row[1].z, b.m_row[2].z, 0);\n"
-" b3Float4 ans;\n"
-" ans.x = b3Dot3F4( a, colx );\n"
-" ans.y = b3Dot3F4( a, coly );\n"
-" ans.z = b3Dot3F4( a, colz );\n"
-" return ans;\n"
-"}\n"
-"#endif\n"
-"#endif //B3_MAT3x3_H\n"
-"typedef struct b3RigidBodyData b3RigidBodyData_t;\n"
-"struct b3RigidBodyData\n"
-"{\n"
-" b3Float4 m_pos;\n"
-" b3Quat m_quat;\n"
-" b3Float4 m_linVel;\n"
-" b3Float4 m_angVel;\n"
-" int m_collidableIdx;\n"
-" float m_invMass;\n"
-" float m_restituitionCoeff;\n"
-" float m_frictionCoeff;\n"
-"};\n"
-"typedef struct b3InertiaData b3InertiaData_t;\n"
-"struct b3InertiaData\n"
-"{\n"
-" b3Mat3x3 m_invInertiaWorld;\n"
-" b3Mat3x3 m_initInvInertia;\n"
-"};\n"
-"#endif //B3_RIGIDBODY_DATA_H\n"
-" \n"
-"#ifndef B3_CONVEX_POLYHEDRON_DATA_H\n"
-"#define B3_CONVEX_POLYHEDRON_DATA_H\n"
-"#ifndef B3_FLOAT4_H\n"
-"#ifdef __cplusplus\n"
-"#else\n"
-"#endif \n"
-"#endif //B3_FLOAT4_H\n"
-"#ifndef B3_QUAT_H\n"
-"#ifdef __cplusplus\n"
-"#else\n"
-"#endif \n"
-"#endif //B3_QUAT_H\n"
-"typedef struct b3GpuFace b3GpuFace_t;\n"
-"struct b3GpuFace\n"
-"{\n"
-" b3Float4 m_plane;\n"
-" int m_indexOffset;\n"
-" int m_numIndices;\n"
-" int m_unusedPadding1;\n"
-" int m_unusedPadding2;\n"
-"};\n"
-"typedef struct b3ConvexPolyhedronData b3ConvexPolyhedronData_t;\n"
-"struct b3ConvexPolyhedronData\n"
-"{\n"
-" b3Float4 m_localCenter;\n"
-" b3Float4 m_extents;\n"
-" b3Float4 mC;\n"
-" b3Float4 mE;\n"
-" float m_radius;\n"
-" int m_faceOffset;\n"
-" int m_numFaces;\n"
-" int m_numVertices;\n"
-" int m_vertexOffset;\n"
-" int m_uniqueEdgesOffset;\n"
-" int m_numUniqueEdges;\n"
-" int m_unused;\n"
-"};\n"
-"#endif //B3_CONVEX_POLYHEDRON_DATA_H\n"
-"#ifndef B3_COLLIDABLE_H\n"
-"#define B3_COLLIDABLE_H\n"
-"#ifndef B3_FLOAT4_H\n"
-"#ifdef __cplusplus\n"
-"#else\n"
-"#endif \n"
-"#endif //B3_FLOAT4_H\n"
-"#ifndef B3_QUAT_H\n"
-"#ifdef __cplusplus\n"
-"#else\n"
-"#endif \n"
-"#endif //B3_QUAT_H\n"
-"enum b3ShapeTypes\n"
-"{\n"
-" SHAPE_HEIGHT_FIELD=1,\n"
-" SHAPE_CONVEX_HULL=3,\n"
-" SHAPE_PLANE=4,\n"
-" SHAPE_CONCAVE_TRIMESH=5,\n"
-" SHAPE_COMPOUND_OF_CONVEX_HULLS=6,\n"
-" SHAPE_SPHERE=7,\n"
-" MAX_NUM_SHAPE_TYPES,\n"
-"};\n"
-"typedef struct b3Collidable b3Collidable_t;\n"
-"struct b3Collidable\n"
-"{\n"
-" union {\n"
-" int m_numChildShapes;\n"
-" int m_bvhIndex;\n"
-" };\n"
-" union\n"
-" {\n"
-" float m_radius;\n"
-" int m_compoundBvhIndex;\n"
-" };\n"
-" int m_shapeType;\n"
-" int m_shapeIndex;\n"
-"};\n"
-"typedef struct b3GpuChildShape b3GpuChildShape_t;\n"
-"struct b3GpuChildShape\n"
-"{\n"
-" b3Float4 m_childPosition;\n"
-" b3Quat m_childOrientation;\n"
-" int m_shapeIndex;\n"
-" int m_unused0;\n"
-" int m_unused1;\n"
-" int m_unused2;\n"
-"};\n"
-"struct b3CompoundOverlappingPair\n"
-"{\n"
-" int m_bodyIndexA;\n"
-" int m_bodyIndexB;\n"
-"// int m_pairType;\n"
-" int m_childShapeIndexA;\n"
-" int m_childShapeIndexB;\n"
-"};\n"
-"#endif //B3_COLLIDABLE_H\n"
-"#ifdef __cplusplus\n"
-"#else\n"
-"#define B3_MPR_SQRT sqrt\n"
-"#endif\n"
-"#define B3_MPR_FMIN(x, y) ((x) < (y) ? (x) : (y))\n"
-"#define B3_MPR_FABS fabs\n"
-"#define B3_MPR_TOLERANCE 1E-6f\n"
-"#define B3_MPR_MAX_ITERATIONS 1000\n"
-"struct _b3MprSupport_t \n"
-"{\n"
-" b3Float4 v; //!< Support point in minkowski sum\n"
-" b3Float4 v1; //!< Support point in obj1\n"
-" b3Float4 v2; //!< Support point in obj2\n"
-"};\n"
-"typedef struct _b3MprSupport_t b3MprSupport_t;\n"
-"struct _b3MprSimplex_t \n"
-"{\n"
-" b3MprSupport_t ps[4];\n"
-" int last; //!< index of last added point\n"
-"};\n"
-"typedef struct _b3MprSimplex_t b3MprSimplex_t;\n"
-"inline b3MprSupport_t* b3MprSimplexPointW(b3MprSimplex_t *s, int idx)\n"
-"{\n"
-" return &s->ps[idx];\n"
-"}\n"
-"inline void b3MprSimplexSetSize(b3MprSimplex_t *s, int size)\n"
-"{\n"
-" s->last = size - 1;\n"
-"}\n"
-"inline int b3MprSimplexSize(const b3MprSimplex_t *s)\n"
-"{\n"
-" return s->last + 1;\n"
-"}\n"
-"inline const b3MprSupport_t* b3MprSimplexPoint(const b3MprSimplex_t* s, int idx)\n"
-"{\n"
-" // here is no check on boundaries\n"
-" return &s->ps[idx];\n"
-"}\n"
-"inline void b3MprSupportCopy(b3MprSupport_t *d, const b3MprSupport_t *s)\n"
-"{\n"
-" *d = *s;\n"
-"}\n"
-"inline void b3MprSimplexSet(b3MprSimplex_t *s, size_t pos, const b3MprSupport_t *a)\n"
-"{\n"
-" b3MprSupportCopy(s->ps + pos, a);\n"
-"}\n"
-"inline void b3MprSimplexSwap(b3MprSimplex_t *s, size_t pos1, size_t pos2)\n"
-"{\n"
-" b3MprSupport_t supp;\n"
-" b3MprSupportCopy(&supp, &s->ps[pos1]);\n"
-" b3MprSupportCopy(&s->ps[pos1], &s->ps[pos2]);\n"
-" b3MprSupportCopy(&s->ps[pos2], &supp);\n"
-"}\n"
-"inline int b3MprIsZero(float val)\n"
-"{\n"
-" return B3_MPR_FABS(val) < FLT_EPSILON;\n"
-"}\n"
-"inline int b3MprEq(float _a, float _b)\n"
-"{\n"
-" float ab;\n"
-" float a, b;\n"
-" ab = B3_MPR_FABS(_a - _b);\n"
-" if (B3_MPR_FABS(ab) < FLT_EPSILON)\n"
-" return 1;\n"
-" a = B3_MPR_FABS(_a);\n"
-" b = B3_MPR_FABS(_b);\n"
-" if (b > a){\n"
-" return ab < FLT_EPSILON * b;\n"
-" }else{\n"
-" return ab < FLT_EPSILON * a;\n"
-" }\n"
-"}\n"
-"inline int b3MprVec3Eq(const b3Float4* a, const b3Float4 *b)\n"
-"{\n"
-" return b3MprEq((*a).x, (*b).x)\n"
-" && b3MprEq((*a).y, (*b).y)\n"
-" && b3MprEq((*a).z, (*b).z);\n"
-"}\n"
-"inline b3Float4 b3LocalGetSupportVertex(b3Float4ConstArg supportVec,__global const b3ConvexPolyhedronData_t* hull, b3ConstArray(b3Float4) verticesA)\n"
-"{\n"
-" b3Float4 supVec = b3MakeFloat4(0,0,0,0);\n"
-" float maxDot = -B3_LARGE_FLOAT;\n"
-" if( 0 < hull->m_numVertices )\n"
-" {\n"
-" const b3Float4 scaled = supportVec;\n"
-" int index = b3MaxDot(scaled, &verticesA[hull->m_vertexOffset], hull->m_numVertices, &maxDot);\n"
-" return verticesA[hull->m_vertexOffset+index];\n"
-" }\n"
-" return supVec;\n"