fix: some file didn't have the svn:eol-style native yet

This commit is contained in:
erwin.coumans
2010-03-06 15:23:36 +00:00
parent 4fd48ac691
commit 81f04a4d48
641 changed files with 301123 additions and 301123 deletions

File diff suppressed because it is too large Load Diff

View File

@@ -1,74 +1,74 @@
/*
Bullet Continuous Collision Detection and Physics Library, Copyright (c) 2007 Erwin Coumans
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.
*/
#include "MiniCLTask.h"
#include "BulletMultiThreaded/PlatformDefinitions.h"
#include "BulletMultiThreaded/SpuFakeDma.h"
#include "LinearMath/btMinMax.h"
#include "MiniCLTask.h"
#include "BulletMultiThreaded/MiniCLTaskScheduler.h"
#ifdef __SPU__
#include <spu_printf.h>
#else
#include <stdio.h>
#define spu_printf printf
#endif
int gMiniCLNumOutstandingTasks = 0;
struct MiniCLTask_LocalStoreMemory
{
};
//-- MAIN METHOD
void processMiniCLTask(void* userPtr, void* lsMemory)
{
// BT_PROFILE("processSampleTask");
MiniCLTask_LocalStoreMemory* localMemory = (MiniCLTask_LocalStoreMemory*)lsMemory;
MiniCLTaskDesc* taskDescPtr = (MiniCLTaskDesc*)userPtr;
MiniCLTaskDesc& taskDesc = *taskDescPtr;
for (unsigned int i=taskDesc.m_firstWorkUnit;i<taskDesc.m_lastWorkUnit;i++)
{
taskDesc.m_kernel->m_launcher(&taskDesc, i);
}
// printf("Compute Unit[%d] executed kernel %d work items [%d..%d)\n",taskDesc.m_taskId,taskDesc.m_kernelProgramId,taskDesc.m_firstWorkUnit,taskDesc.m_lastWorkUnit);
}
#if defined(__CELLOS_LV2__) || defined (LIBSPE2)
ATTRIBUTE_ALIGNED16(MiniCLTask_LocalStoreMemory gLocalStoreMemory);
void* createMiniCLLocalStoreMemory()
{
return &gLocalStoreMemory;
}
#else
void* createMiniCLLocalStoreMemory()
{
return new MiniCLTask_LocalStoreMemory;
};
#endif
/*
Bullet Continuous Collision Detection and Physics Library, Copyright (c) 2007 Erwin Coumans
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.
*/
#include "MiniCLTask.h"
#include "BulletMultiThreaded/PlatformDefinitions.h"
#include "BulletMultiThreaded/SpuFakeDma.h"
#include "LinearMath/btMinMax.h"
#include "MiniCLTask.h"
#include "BulletMultiThreaded/MiniCLTaskScheduler.h"
#ifdef __SPU__
#include <spu_printf.h>
#else
#include <stdio.h>
#define spu_printf printf
#endif
int gMiniCLNumOutstandingTasks = 0;
struct MiniCLTask_LocalStoreMemory
{
};
//-- MAIN METHOD
void processMiniCLTask(void* userPtr, void* lsMemory)
{
// BT_PROFILE("processSampleTask");
MiniCLTask_LocalStoreMemory* localMemory = (MiniCLTask_LocalStoreMemory*)lsMemory;
MiniCLTaskDesc* taskDescPtr = (MiniCLTaskDesc*)userPtr;
MiniCLTaskDesc& taskDesc = *taskDescPtr;
for (unsigned int i=taskDesc.m_firstWorkUnit;i<taskDesc.m_lastWorkUnit;i++)
{
taskDesc.m_kernel->m_launcher(&taskDesc, i);
}
// printf("Compute Unit[%d] executed kernel %d work items [%d..%d)\n",taskDesc.m_taskId,taskDesc.m_kernelProgramId,taskDesc.m_firstWorkUnit,taskDesc.m_lastWorkUnit);
}
#if defined(__CELLOS_LV2__) || defined (LIBSPE2)
ATTRIBUTE_ALIGNED16(MiniCLTask_LocalStoreMemory gLocalStoreMemory);
void* createMiniCLLocalStoreMemory()
{
return &gLocalStoreMemory;
}
#else
void* createMiniCLLocalStoreMemory()
{
return new MiniCLTask_LocalStoreMemory;
};
#endif

View File

@@ -1,62 +1,62 @@
/*
Bullet Continuous Collision Detection and Physics Library, Copyright (c) 2007 Erwin Coumans
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 MINICL__TASK_H
#define MINICL__TASK_H
#include "BulletMultiThreaded/PlatformDefinitions.h"
#include "LinearMath/btScalar.h"
#include "LinearMath/btAlignedAllocator.h"
#define MINICL_MAX_ARGLENGTH (sizeof(void*))
#define MINI_CL_MAX_ARG 16
#define MINI_CL_MAX_KERNEL_NAME 256
struct MiniCLKernel;
ATTRIBUTE_ALIGNED16(struct) MiniCLTaskDesc
{
BT_DECLARE_ALIGNED_ALLOCATOR();
MiniCLTaskDesc()
{
for (int i=0;i<MINI_CL_MAX_ARG;i++)
{
m_argSizes[i]=0;
}
}
uint32_t m_taskId;
uint32_t m_firstWorkUnit;
uint32_t m_lastWorkUnit;
MiniCLKernel* m_kernel;
void* m_argData[MINI_CL_MAX_ARG];
int m_argSizes[MINI_CL_MAX_ARG];
};
extern "C" int gMiniCLNumOutstandingTasks;
void processMiniCLTask(void* userPtr, void* lsMemory);
void* createMiniCLLocalStoreMemory();
#endif //MINICL__TASK_H
/*
Bullet Continuous Collision Detection and Physics Library, Copyright (c) 2007 Erwin Coumans
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 MINICL__TASK_H
#define MINICL__TASK_H
#include "BulletMultiThreaded/PlatformDefinitions.h"
#include "LinearMath/btScalar.h"
#include "LinearMath/btAlignedAllocator.h"
#define MINICL_MAX_ARGLENGTH (sizeof(void*))
#define MINI_CL_MAX_ARG 16
#define MINI_CL_MAX_KERNEL_NAME 256
struct MiniCLKernel;
ATTRIBUTE_ALIGNED16(struct) MiniCLTaskDesc
{
BT_DECLARE_ALIGNED_ALLOCATOR();
MiniCLTaskDesc()
{
for (int i=0;i<MINI_CL_MAX_ARG;i++)
{
m_argSizes[i]=0;
}
}
uint32_t m_taskId;
uint32_t m_firstWorkUnit;
uint32_t m_lastWorkUnit;
MiniCLKernel* m_kernel;
void* m_argData[MINI_CL_MAX_ARG];
int m_argSizes[MINI_CL_MAX_ARG];
};
extern "C" int gMiniCLNumOutstandingTasks;
void processMiniCLTask(void* userPtr, void* lsMemory);
void* createMiniCLLocalStoreMemory();
#endif //MINICL__TASK_H

File diff suppressed because it is too large Load Diff

View File

@@ -1,135 +1,135 @@
/*
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 FAKE_DMA_H
#define FAKE_DMA_H
#include "PlatformDefinitions.h"
#include "LinearMath/btScalar.h"
#ifdef __SPU__
#ifndef USE_LIBSPE2
#include <cell/dma.h>
#include <stdint.h>
#define DMA_TAG(xfer) (xfer + 1)
#define DMA_MASK(xfer) (1 << DMA_TAG(xfer))
#else // !USE_LIBSPE2
#define DMA_TAG(xfer) (xfer + 1)
#define DMA_MASK(xfer) (1 << DMA_TAG(xfer))
#include <spu_mfcio.h>
#define DEBUG_DMA
#ifdef DEBUG_DMA
#define dUASSERT(a,b) if (!(a)) { printf(b);}
#define uintsize ppu_address_t
#define cellDmaLargeGet(ls, ea, size, tag, tid, rid) if ( (((uintsize)ls%16) != ((uintsize)ea%16)) || ((((uintsize)ea%16) || ((uintsize)ls%16)) && (( ((uintsize)ls%16) != ((uintsize)size%16) ) || ( ((uintsize)ea%16) != ((uintsize)size%16) ) ) ) || ( ((uintsize)size%16) && ((uintsize)size!=1) && ((uintsize)size!=2) && ((uintsize)size!=4) && ((uintsize)size!=8) ) || (size >= 16384) || !(uintsize)ls || !(uintsize)ea) { \
dUASSERT( (((uintsize)ea % 16) == 0) || (size < 16), "XDR Address not aligned: "); \
dUASSERT( (((uintsize)ls % 16) == 0) || (size < 16), "LS Address not aligned: "); \
dUASSERT( ((((uintsize)ls % size) == 0) && (((uintsize)ea % size) == 0)) || (size > 16), "Not naturally aligned: "); \
dUASSERT((size == 1) || (size == 2) || (size == 4) || (size == 8) || ((size % 16) == 0), "size not a multiple of 16byte: "); \
dUASSERT(size < 16384, "size too big: "); \
dUASSERT( ((uintsize)ea%16)==((uintsize)ls%16), "wrong Quadword alignment of LS and EA: "); \
dUASSERT(ea != 0, "Nullpointer EA: "); dUASSERT(ls != 0, "Nullpointer LS: ");\
printf("GET %s:%d from: 0x%x, to: 0x%x - %d bytes\n", __FILE__, __LINE__, (unsigned int)ea,(unsigned int)ls,(unsigned int)size);\
} \
mfc_get(ls, ea, size, tag, tid, rid)
#define cellDmaGet(ls, ea, size, tag, tid, rid) if ( (((uintsize)ls%16) != ((uintsize)ea%16)) || ((((uintsize)ea%16) || ((uintsize)ls%16)) && (( ((uintsize)ls%16) != ((uintsize)size%16) ) || ( ((uintsize)ea%16) != ((uintsize)size%16) ) ) ) || ( ((uintsize)size%16) && ((uintsize)size!=1) && ((uintsize)size!=2) && ((uintsize)size!=4) && ((uintsize)size!=8) ) || (size >= 16384) || !(uintsize)ls || !(uintsize)ea) { \
dUASSERT( (((uintsize)ea % 16) == 0) || (size < 16), "XDR Address not aligned: "); \
dUASSERT( (((uintsize)ls % 16) == 0) || (size < 16), "LS Address not aligned: "); \
dUASSERT( ((((uintsize)ls % size) == 0) && (((uintsize)ea % size) == 0)) || (size > 16), "Not naturally aligned: "); \
dUASSERT((size == 1) || (size == 2) || (size == 4) || (size == 8) || ((size % 16) == 0), "size not a multiple of 16byte: "); \
dUASSERT(size < 16384, "size too big: "); \
dUASSERT( ((uintsize)ea%16)==((uintsize)ls%16), "wrong Quadword alignment of LS and EA: "); \
dUASSERT(ea != 0, "Nullpointer EA: "); dUASSERT(ls != 0, "Nullpointer LS: ");\
printf("GET %s:%d from: 0x%x, to: 0x%x - %d bytes\n", __FILE__, __LINE__, (unsigned int)ea,(unsigned int)ls,(unsigned int)size);\
} \
mfc_get(ls, ea, size, tag, tid, rid)
#define cellDmaLargePut(ls, ea, size, tag, tid, rid) if ( (((uintsize)ls%16) != ((uintsize)ea%16)) || ((((uintsize)ea%16) || ((uintsize)ls%16)) && (( ((uintsize)ls%16) != ((uintsize)size%16) ) || ( ((uintsize)ea%16) != ((uintsize)size%16) ) ) ) || ( ((uintsize)size%16) && ((uintsize)size!=1) && ((uintsize)size!=2) && ((uintsize)size!=4) && ((uintsize)size!=8) ) || (size >= 16384) || !(uintsize)ls || !(uintsize)ea) { \
dUASSERT( (((uintsize)ea % 16) == 0) || (size < 16), "XDR Address not aligned: "); \
dUASSERT( (((uintsize)ls % 16) == 0) || (size < 16), "LS Address not aligned: "); \
dUASSERT( ((((uintsize)ls % size) == 0) && (((uintsize)ea % size) == 0)) || (size > 16), "Not naturally aligned: "); \
dUASSERT((size == 1) || (size == 2) || (size == 4) || (size == 8) || ((size % 16) == 0), "size not a multiple of 16byte: "); \
dUASSERT(size < 16384, "size too big: "); \
dUASSERT( ((uintsize)ea%16)==((uintsize)ls%16), "wrong Quadword alignment of LS and EA: "); \
dUASSERT(ea != 0, "Nullpointer EA: "); dUASSERT(ls != 0, "Nullpointer LS: ");\
printf("PUT %s:%d from: 0x%x, to: 0x%x - %d bytes\n", __FILE__, __LINE__, (unsigned int)ls,(unsigned int)ea,(unsigned int)size); \
} \
mfc_put(ls, ea, size, tag, tid, rid)
#define cellDmaSmallGet(ls, ea, size, tag, tid, rid) if ( (((uintsize)ls%16) != ((uintsize)ea%16)) || ((((uintsize)ea%16) || ((uintsize)ls%16)) && (( ((uintsize)ls%16) != ((uintsize)size%16) ) || ( ((uintsize)ea%16) != ((uintsize)size%16) ) ) ) || ( ((uintsize)size%16) && ((uintsize)size!=1) && ((uintsize)size!=2) && ((uintsize)size!=4) && ((uintsize)size!=8) ) || (size >= 16384) || !(uintsize)ls || !(uintsize)ea) { \
dUASSERT( (((uintsize)ea % 16) == 0) || (size < 16), "XDR Address not aligned: "); \
dUASSERT( (((uintsize)ls % 16) == 0) || (size < 16), "LS Address not aligned: "); \
dUASSERT( ((((uintsize)ls % size) == 0) && (((uintsize)ea % size) == 0)) || (size > 16), "Not naturally aligned: "); \
dUASSERT((size == 1) || (size == 2) || (size == 4) || (size == 8) || ((size % 16) == 0), "size not a multiple of 16byte: "); \
dUASSERT(size < 16384, "size too big: "); \
dUASSERT( ((uintsize)ea%16)==((uintsize)ls%16), "wrong Quadword alignment of LS and EA: "); \
dUASSERT(ea != 0, "Nullpointer EA: "); dUASSERT(ls != 0, "Nullpointer LS: ");\
printf("GET %s:%d from: 0x%x, to: 0x%x - %d bytes\n", __FILE__, __LINE__, (unsigned int)ea,(unsigned int)ls,(unsigned int)size);\
} \
mfc_get(ls, ea, size, tag, tid, rid)
#define cellDmaWaitTagStatusAll(ignore) mfc_write_tag_mask(ignore) ; mfc_read_tag_status_all()
#else
#define cellDmaLargeGet(ls, ea, size, tag, tid, rid) mfc_get(ls, ea, size, tag, tid, rid)
#define cellDmaGet(ls, ea, size, tag, tid, rid) mfc_get(ls, ea, size, tag, tid, rid)
#define cellDmaLargePut(ls, ea, size, tag, tid, rid) mfc_put(ls, ea, size, tag, tid, rid)
#define cellDmaSmallGet(ls, ea, size, tag, tid, rid) mfc_get(ls, ea, size, tag, tid, rid)
#define cellDmaWaitTagStatusAll(ignore) mfc_write_tag_mask(ignore) ; mfc_read_tag_status_all()
#endif // DEBUG_DMA
#endif // USE_LIBSPE2
#else // !__SPU__
//Simulate DMA using memcpy or direct access on non-CELL platforms that don't have DMAs and SPUs (Win32, Mac, Linux etc)
//Potential to add networked simulation using this interface
#define DMA_TAG(a) (a)
#define DMA_MASK(a) (a)
/// cellDmaLargeGet Win32 replacements for Cell DMA to allow simulating most of the SPU code (just memcpy)
int cellDmaLargeGet(void *ls, uint64_t ea, uint32_t size, uint32_t tag, uint32_t tid, uint32_t rid);
int cellDmaGet(void *ls, uint64_t ea, uint32_t size, uint32_t tag, uint32_t tid, uint32_t rid);
/// cellDmaLargePut Win32 replacements for Cell DMA to allow simulating most of the SPU code (just memcpy)
int cellDmaLargePut(const void *ls, uint64_t ea, uint32_t size, uint32_t tag, uint32_t tid, uint32_t rid);
/// cellDmaWaitTagStatusAll Win32 replacements for Cell DMA to allow simulating most of the SPU code (just memcpy)
void cellDmaWaitTagStatusAll(int ignore);
#endif //__CELLOS_LV2__
///stallingUnalignedDmaSmallGet internally uses DMA_TAG(1)
int stallingUnalignedDmaSmallGet(void *ls, uint64_t ea, uint32_t size);
void* cellDmaLargeGetReadOnly(void *ls, uint64_t ea, uint32_t size, uint32_t tag, uint32_t tid, uint32_t rid);
void* cellDmaGetReadOnly(void *ls, uint64_t ea, uint32_t size, uint32_t tag, uint32_t tid, uint32_t rid);
void* cellDmaSmallGetReadOnly(void *ls, uint64_t ea, uint32_t size, uint32_t tag, uint32_t tid, uint32_t rid);
#endif //FAKE_DMA_H
/*
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 FAKE_DMA_H
#define FAKE_DMA_H
#include "PlatformDefinitions.h"
#include "LinearMath/btScalar.h"
#ifdef __SPU__
#ifndef USE_LIBSPE2
#include <cell/dma.h>
#include <stdint.h>
#define DMA_TAG(xfer) (xfer + 1)
#define DMA_MASK(xfer) (1 << DMA_TAG(xfer))
#else // !USE_LIBSPE2
#define DMA_TAG(xfer) (xfer + 1)
#define DMA_MASK(xfer) (1 << DMA_TAG(xfer))
#include <spu_mfcio.h>
#define DEBUG_DMA
#ifdef DEBUG_DMA
#define dUASSERT(a,b) if (!(a)) { printf(b);}
#define uintsize ppu_address_t
#define cellDmaLargeGet(ls, ea, size, tag, tid, rid) if ( (((uintsize)ls%16) != ((uintsize)ea%16)) || ((((uintsize)ea%16) || ((uintsize)ls%16)) && (( ((uintsize)ls%16) != ((uintsize)size%16) ) || ( ((uintsize)ea%16) != ((uintsize)size%16) ) ) ) || ( ((uintsize)size%16) && ((uintsize)size!=1) && ((uintsize)size!=2) && ((uintsize)size!=4) && ((uintsize)size!=8) ) || (size >= 16384) || !(uintsize)ls || !(uintsize)ea) { \
dUASSERT( (((uintsize)ea % 16) == 0) || (size < 16), "XDR Address not aligned: "); \
dUASSERT( (((uintsize)ls % 16) == 0) || (size < 16), "LS Address not aligned: "); \
dUASSERT( ((((uintsize)ls % size) == 0) && (((uintsize)ea % size) == 0)) || (size > 16), "Not naturally aligned: "); \
dUASSERT((size == 1) || (size == 2) || (size == 4) || (size == 8) || ((size % 16) == 0), "size not a multiple of 16byte: "); \
dUASSERT(size < 16384, "size too big: "); \
dUASSERT( ((uintsize)ea%16)==((uintsize)ls%16), "wrong Quadword alignment of LS and EA: "); \
dUASSERT(ea != 0, "Nullpointer EA: "); dUASSERT(ls != 0, "Nullpointer LS: ");\
printf("GET %s:%d from: 0x%x, to: 0x%x - %d bytes\n", __FILE__, __LINE__, (unsigned int)ea,(unsigned int)ls,(unsigned int)size);\
} \
mfc_get(ls, ea, size, tag, tid, rid)
#define cellDmaGet(ls, ea, size, tag, tid, rid) if ( (((uintsize)ls%16) != ((uintsize)ea%16)) || ((((uintsize)ea%16) || ((uintsize)ls%16)) && (( ((uintsize)ls%16) != ((uintsize)size%16) ) || ( ((uintsize)ea%16) != ((uintsize)size%16) ) ) ) || ( ((uintsize)size%16) && ((uintsize)size!=1) && ((uintsize)size!=2) && ((uintsize)size!=4) && ((uintsize)size!=8) ) || (size >= 16384) || !(uintsize)ls || !(uintsize)ea) { \
dUASSERT( (((uintsize)ea % 16) == 0) || (size < 16), "XDR Address not aligned: "); \
dUASSERT( (((uintsize)ls % 16) == 0) || (size < 16), "LS Address not aligned: "); \
dUASSERT( ((((uintsize)ls % size) == 0) && (((uintsize)ea % size) == 0)) || (size > 16), "Not naturally aligned: "); \
dUASSERT((size == 1) || (size == 2) || (size == 4) || (size == 8) || ((size % 16) == 0), "size not a multiple of 16byte: "); \
dUASSERT(size < 16384, "size too big: "); \
dUASSERT( ((uintsize)ea%16)==((uintsize)ls%16), "wrong Quadword alignment of LS and EA: "); \
dUASSERT(ea != 0, "Nullpointer EA: "); dUASSERT(ls != 0, "Nullpointer LS: ");\
printf("GET %s:%d from: 0x%x, to: 0x%x - %d bytes\n", __FILE__, __LINE__, (unsigned int)ea,(unsigned int)ls,(unsigned int)size);\
} \
mfc_get(ls, ea, size, tag, tid, rid)
#define cellDmaLargePut(ls, ea, size, tag, tid, rid) if ( (((uintsize)ls%16) != ((uintsize)ea%16)) || ((((uintsize)ea%16) || ((uintsize)ls%16)) && (( ((uintsize)ls%16) != ((uintsize)size%16) ) || ( ((uintsize)ea%16) != ((uintsize)size%16) ) ) ) || ( ((uintsize)size%16) && ((uintsize)size!=1) && ((uintsize)size!=2) && ((uintsize)size!=4) && ((uintsize)size!=8) ) || (size >= 16384) || !(uintsize)ls || !(uintsize)ea) { \
dUASSERT( (((uintsize)ea % 16) == 0) || (size < 16), "XDR Address not aligned: "); \
dUASSERT( (((uintsize)ls % 16) == 0) || (size < 16), "LS Address not aligned: "); \
dUASSERT( ((((uintsize)ls % size) == 0) && (((uintsize)ea % size) == 0)) || (size > 16), "Not naturally aligned: "); \
dUASSERT((size == 1) || (size == 2) || (size == 4) || (size == 8) || ((size % 16) == 0), "size not a multiple of 16byte: "); \
dUASSERT(size < 16384, "size too big: "); \
dUASSERT( ((uintsize)ea%16)==((uintsize)ls%16), "wrong Quadword alignment of LS and EA: "); \
dUASSERT(ea != 0, "Nullpointer EA: "); dUASSERT(ls != 0, "Nullpointer LS: ");\
printf("PUT %s:%d from: 0x%x, to: 0x%x - %d bytes\n", __FILE__, __LINE__, (unsigned int)ls,(unsigned int)ea,(unsigned int)size); \
} \
mfc_put(ls, ea, size, tag, tid, rid)
#define cellDmaSmallGet(ls, ea, size, tag, tid, rid) if ( (((uintsize)ls%16) != ((uintsize)ea%16)) || ((((uintsize)ea%16) || ((uintsize)ls%16)) && (( ((uintsize)ls%16) != ((uintsize)size%16) ) || ( ((uintsize)ea%16) != ((uintsize)size%16) ) ) ) || ( ((uintsize)size%16) && ((uintsize)size!=1) && ((uintsize)size!=2) && ((uintsize)size!=4) && ((uintsize)size!=8) ) || (size >= 16384) || !(uintsize)ls || !(uintsize)ea) { \
dUASSERT( (((uintsize)ea % 16) == 0) || (size < 16), "XDR Address not aligned: "); \
dUASSERT( (((uintsize)ls % 16) == 0) || (size < 16), "LS Address not aligned: "); \
dUASSERT( ((((uintsize)ls % size) == 0) && (((uintsize)ea % size) == 0)) || (size > 16), "Not naturally aligned: "); \
dUASSERT((size == 1) || (size == 2) || (size == 4) || (size == 8) || ((size % 16) == 0), "size not a multiple of 16byte: "); \
dUASSERT(size < 16384, "size too big: "); \
dUASSERT( ((uintsize)ea%16)==((uintsize)ls%16), "wrong Quadword alignment of LS and EA: "); \
dUASSERT(ea != 0, "Nullpointer EA: "); dUASSERT(ls != 0, "Nullpointer LS: ");\
printf("GET %s:%d from: 0x%x, to: 0x%x - %d bytes\n", __FILE__, __LINE__, (unsigned int)ea,(unsigned int)ls,(unsigned int)size);\
} \
mfc_get(ls, ea, size, tag, tid, rid)
#define cellDmaWaitTagStatusAll(ignore) mfc_write_tag_mask(ignore) ; mfc_read_tag_status_all()
#else
#define cellDmaLargeGet(ls, ea, size, tag, tid, rid) mfc_get(ls, ea, size, tag, tid, rid)
#define cellDmaGet(ls, ea, size, tag, tid, rid) mfc_get(ls, ea, size, tag, tid, rid)
#define cellDmaLargePut(ls, ea, size, tag, tid, rid) mfc_put(ls, ea, size, tag, tid, rid)
#define cellDmaSmallGet(ls, ea, size, tag, tid, rid) mfc_get(ls, ea, size, tag, tid, rid)
#define cellDmaWaitTagStatusAll(ignore) mfc_write_tag_mask(ignore) ; mfc_read_tag_status_all()
#endif // DEBUG_DMA
#endif // USE_LIBSPE2
#else // !__SPU__
//Simulate DMA using memcpy or direct access on non-CELL platforms that don't have DMAs and SPUs (Win32, Mac, Linux etc)
//Potential to add networked simulation using this interface
#define DMA_TAG(a) (a)
#define DMA_MASK(a) (a)
/// cellDmaLargeGet Win32 replacements for Cell DMA to allow simulating most of the SPU code (just memcpy)
int cellDmaLargeGet(void *ls, uint64_t ea, uint32_t size, uint32_t tag, uint32_t tid, uint32_t rid);
int cellDmaGet(void *ls, uint64_t ea, uint32_t size, uint32_t tag, uint32_t tid, uint32_t rid);
/// cellDmaLargePut Win32 replacements for Cell DMA to allow simulating most of the SPU code (just memcpy)
int cellDmaLargePut(const void *ls, uint64_t ea, uint32_t size, uint32_t tag, uint32_t tid, uint32_t rid);
/// cellDmaWaitTagStatusAll Win32 replacements for Cell DMA to allow simulating most of the SPU code (just memcpy)
void cellDmaWaitTagStatusAll(int ignore);
#endif //__CELLOS_LV2__
///stallingUnalignedDmaSmallGet internally uses DMA_TAG(1)
int stallingUnalignedDmaSmallGet(void *ls, uint64_t ea, uint32_t size);
void* cellDmaLargeGetReadOnly(void *ls, uint64_t ea, uint32_t size, uint32_t tag, uint32_t tid, uint32_t rid);
void* cellDmaGetReadOnly(void *ls, uint64_t ea, uint32_t size, uint32_t tag, uint32_t tid, uint32_t rid);
void* cellDmaSmallGetReadOnly(void *ls, uint64_t ea, uint32_t size, uint32_t tag, uint32_t tid, uint32_t rid);
#endif //FAKE_DMA_H

File diff suppressed because it is too large Load Diff

View File

@@ -1,138 +1,138 @@
/*
Bullet Continuous Collision Detection and Physics Library, http://bulletphysics.org
Copyright (C) 2006, 2009 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.
*/
//----------------------------------------------------------------------------------------
#ifndef BTGPU3DGRIDBROADPHASE_H
#define BTGPU3DGRIDBROADPHASE_H
//----------------------------------------------------------------------------------------
#include "BulletCollision/BroadphaseCollision/btSimpleBroadphase.h"
#include "btGpu3DGridBroadphaseSharedTypes.h"
//----------------------------------------------------------------------------------------
///The btGpu3DGridBroadphase uses GPU-style code compiled for CPU to compute overlapping pairs
class btGpu3DGridBroadphase : public btSimpleBroadphase
{
protected:
bool m_bInitialized;
unsigned int m_numBodies;
unsigned int m_numCells;
unsigned int m_maxPairsPerBody;
btScalar m_cellFactorAABB;
unsigned int m_maxBodiesPerCell;
bt3DGridBroadphaseParams m_params;
btScalar m_maxRadius;
// CPU data
unsigned int* m_hBodiesHash;
unsigned int* m_hCellStart;
unsigned int* m_hPairBuffStartCurr;
bt3DGrid3F1U* m_hAABB;
unsigned int* m_hPairBuff;
unsigned int* m_hPairScan;
unsigned int* m_hPairOut;
// large proxies
int m_numLargeHandles;
int m_maxLargeHandles;
int m_LastLargeHandleIndex;
btSimpleBroadphaseProxy* m_pLargeHandles;
void* m_pLargeHandlesRawPtr;
int m_firstFreeLargeHandle;
int allocLargeHandle()
{
btAssert(m_numLargeHandles < m_maxLargeHandles);
int freeLargeHandle = m_firstFreeLargeHandle;
m_firstFreeLargeHandle = m_pLargeHandles[freeLargeHandle].GetNextFree();
m_numLargeHandles++;
if(freeLargeHandle > m_LastLargeHandleIndex)
{
m_LastLargeHandleIndex = freeLargeHandle;
}
return freeLargeHandle;
}
void freeLargeHandle(btSimpleBroadphaseProxy* proxy)
{
int handle = int(proxy - m_pLargeHandles);
btAssert((handle >= 0) && (handle < m_maxHandles));
if(handle == m_LastLargeHandleIndex)
{
m_LastLargeHandleIndex--;
}
proxy->SetNextFree(m_firstFreeLargeHandle);
m_firstFreeLargeHandle = handle;
proxy->m_clientObject = 0;
m_numLargeHandles--;
}
bool isLargeProxy(const btVector3& aabbMin, const btVector3& aabbMax);
bool isLargeProxy(btBroadphaseProxy* proxy);
// debug
unsigned int m_numPairsAdded;
unsigned int m_numPairsRemoved;
unsigned int m_numOverflows;
//
public:
btGpu3DGridBroadphase(const btVector3& worldAabbMin,const btVector3& worldAabbMax,
int gridSizeX, int gridSizeY, int gridSizeZ,
int maxSmallProxies, int maxLargeProxies, int maxPairsPerBody,
int maxBodiesPerCell = 8,
btScalar cellFactorAABB = btScalar(1.0f));
btGpu3DGridBroadphase( btOverlappingPairCache* overlappingPairCache,
const btVector3& worldAabbMin,const btVector3& worldAabbMax,
int gridSizeX, int gridSizeY, int gridSizeZ,
int maxSmallProxies, int maxLargeProxies, int maxPairsPerBody,
int maxBodiesPerCell = 8,
btScalar cellFactorAABB = btScalar(1.0f));
virtual ~btGpu3DGridBroadphase();
virtual void calculateOverlappingPairs(btDispatcher* dispatcher);
virtual btBroadphaseProxy* createProxy(const btVector3& aabbMin, const btVector3& aabbMax,int shapeType,void* userPtr ,short int collisionFilterGroup,short int collisionFilterMask, btDispatcher* dispatcher,void* multiSapProxy);
virtual void destroyProxy(btBroadphaseProxy* proxy,btDispatcher* dispatcher);
virtual void rayTest(const btVector3& rayFrom,const btVector3& rayTo, btBroadphaseRayCallback& rayCallback);
virtual void resetPool(btDispatcher* dispatcher);
protected:
void _initialize( const btVector3& worldAabbMin,const btVector3& worldAabbMax,
int gridSizeX, int gridSizeY, int gridSizeZ,
int maxSmallProxies, int maxLargeProxies, int maxPairsPerBody,
int maxBodiesPerCell = 8,
btScalar cellFactorAABB = btScalar(1.0f));
void _finalize();
void addPairsToCache(btDispatcher* dispatcher);
void addLarge2LargePairsToCache(btDispatcher* dispatcher);
// overrides for CPU version
virtual void setParameters(bt3DGridBroadphaseParams* hostParams);
virtual void prepareAABB();
virtual void calcHashAABB();
virtual void sortHash();
virtual void findCellStart();
virtual void findOverlappingPairs();
virtual void findPairsLarge();
virtual void computePairCacheChanges();
virtual void scanOverlappingPairBuff();
virtual void squeezeOverlappingPairBuff();
};
//----------------------------------------------------------------------------------------
#endif //BTGPU3DGRIDBROADPHASE_H
//----------------------------------------------------------------------------------------
//----------------------------------------------------------------------------------------
//----------------------------------------------------------------------------------------
/*
Bullet Continuous Collision Detection and Physics Library, http://bulletphysics.org
Copyright (C) 2006, 2009 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.
*/
//----------------------------------------------------------------------------------------
#ifndef BTGPU3DGRIDBROADPHASE_H
#define BTGPU3DGRIDBROADPHASE_H
//----------------------------------------------------------------------------------------
#include "BulletCollision/BroadphaseCollision/btSimpleBroadphase.h"
#include "btGpu3DGridBroadphaseSharedTypes.h"
//----------------------------------------------------------------------------------------
///The btGpu3DGridBroadphase uses GPU-style code compiled for CPU to compute overlapping pairs
class btGpu3DGridBroadphase : public btSimpleBroadphase
{
protected:
bool m_bInitialized;
unsigned int m_numBodies;
unsigned int m_numCells;
unsigned int m_maxPairsPerBody;
btScalar m_cellFactorAABB;
unsigned int m_maxBodiesPerCell;
bt3DGridBroadphaseParams m_params;
btScalar m_maxRadius;
// CPU data
unsigned int* m_hBodiesHash;
unsigned int* m_hCellStart;
unsigned int* m_hPairBuffStartCurr;
bt3DGrid3F1U* m_hAABB;
unsigned int* m_hPairBuff;
unsigned int* m_hPairScan;
unsigned int* m_hPairOut;
// large proxies
int m_numLargeHandles;
int m_maxLargeHandles;
int m_LastLargeHandleIndex;
btSimpleBroadphaseProxy* m_pLargeHandles;
void* m_pLargeHandlesRawPtr;
int m_firstFreeLargeHandle;
int allocLargeHandle()
{
btAssert(m_numLargeHandles < m_maxLargeHandles);
int freeLargeHandle = m_firstFreeLargeHandle;
m_firstFreeLargeHandle = m_pLargeHandles[freeLargeHandle].GetNextFree();
m_numLargeHandles++;
if(freeLargeHandle > m_LastLargeHandleIndex)
{
m_LastLargeHandleIndex = freeLargeHandle;
}
return freeLargeHandle;
}
void freeLargeHandle(btSimpleBroadphaseProxy* proxy)
{
int handle = int(proxy - m_pLargeHandles);
btAssert((handle >= 0) && (handle < m_maxHandles));
if(handle == m_LastLargeHandleIndex)
{
m_LastLargeHandleIndex--;
}
proxy->SetNextFree(m_firstFreeLargeHandle);
m_firstFreeLargeHandle = handle;
proxy->m_clientObject = 0;
m_numLargeHandles--;
}
bool isLargeProxy(const btVector3& aabbMin, const btVector3& aabbMax);
bool isLargeProxy(btBroadphaseProxy* proxy);
// debug
unsigned int m_numPairsAdded;
unsigned int m_numPairsRemoved;
unsigned int m_numOverflows;
//
public:
btGpu3DGridBroadphase(const btVector3& worldAabbMin,const btVector3& worldAabbMax,
int gridSizeX, int gridSizeY, int gridSizeZ,
int maxSmallProxies, int maxLargeProxies, int maxPairsPerBody,
int maxBodiesPerCell = 8,
btScalar cellFactorAABB = btScalar(1.0f));
btGpu3DGridBroadphase( btOverlappingPairCache* overlappingPairCache,
const btVector3& worldAabbMin,const btVector3& worldAabbMax,
int gridSizeX, int gridSizeY, int gridSizeZ,
int maxSmallProxies, int maxLargeProxies, int maxPairsPerBody,
int maxBodiesPerCell = 8,
btScalar cellFactorAABB = btScalar(1.0f));
virtual ~btGpu3DGridBroadphase();
virtual void calculateOverlappingPairs(btDispatcher* dispatcher);
virtual btBroadphaseProxy* createProxy(const btVector3& aabbMin, const btVector3& aabbMax,int shapeType,void* userPtr ,short int collisionFilterGroup,short int collisionFilterMask, btDispatcher* dispatcher,void* multiSapProxy);
virtual void destroyProxy(btBroadphaseProxy* proxy,btDispatcher* dispatcher);
virtual void rayTest(const btVector3& rayFrom,const btVector3& rayTo, btBroadphaseRayCallback& rayCallback);
virtual void resetPool(btDispatcher* dispatcher);
protected:
void _initialize( const btVector3& worldAabbMin,const btVector3& worldAabbMax,
int gridSizeX, int gridSizeY, int gridSizeZ,
int maxSmallProxies, int maxLargeProxies, int maxPairsPerBody,
int maxBodiesPerCell = 8,
btScalar cellFactorAABB = btScalar(1.0f));
void _finalize();
void addPairsToCache(btDispatcher* dispatcher);
void addLarge2LargePairsToCache(btDispatcher* dispatcher);
// overrides for CPU version
virtual void setParameters(bt3DGridBroadphaseParams* hostParams);
virtual void prepareAABB();
virtual void calcHashAABB();
virtual void sortHash();
virtual void findCellStart();
virtual void findOverlappingPairs();
virtual void findPairsLarge();
virtual void computePairCacheChanges();
virtual void scanOverlappingPairBuff();
virtual void squeezeOverlappingPairBuff();
};
//----------------------------------------------------------------------------------------
#endif //BTGPU3DGRIDBROADPHASE_H
//----------------------------------------------------------------------------------------
//----------------------------------------------------------------------------------------
//----------------------------------------------------------------------------------------

View File

@@ -1,430 +1,430 @@
/*
Bullet Continuous Collision Detection and Physics Library, http://bulletphysics.org
Copyright (C) 2006, 2009 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.
*/
//----------------------------------------------------------------------------------------
//----------------------------------------------------------------------------------------
//----------------------------------------------------------------------------------------
//----------------------------------------------------------------------------------------
//----------------------------------------------------------------------------------------
// K E R N E L F U N C T I O N S
//----------------------------------------------------------------------------------------
//----------------------------------------------------------------------------------------
//----------------------------------------------------------------------------------------
//----------------------------------------------------------------------------------------
//----------------------------------------------------------------------------------------
//----------------------------------------------------------------------------------------
// calculate position in uniform grid
BT_GPU___device__ int3 bt3DGrid_calcGridPos(float4 p)
{
int3 gridPos;
gridPos.x = (int)floor((p.x - BT_GPU_params.m_worldOriginX) / BT_GPU_params.m_cellSizeX);
gridPos.y = (int)floor((p.y - BT_GPU_params.m_worldOriginY) / BT_GPU_params.m_cellSizeY);
gridPos.z = (int)floor((p.z - BT_GPU_params.m_worldOriginZ) / BT_GPU_params.m_cellSizeZ);
return gridPos;
} // bt3DGrid_calcGridPos()
//----------------------------------------------------------------------------------------
// calculate address in grid from position (clamping to edges)
BT_GPU___device__ uint bt3DGrid_calcGridHash(int3 gridPos)
{
gridPos.x = BT_GPU_max(0, BT_GPU_min(gridPos.x, (int)BT_GPU_params.m_gridSizeX - 1));
gridPos.y = BT_GPU_max(0, BT_GPU_min(gridPos.y, (int)BT_GPU_params.m_gridSizeY - 1));
gridPos.z = BT_GPU_max(0, BT_GPU_min(gridPos.z, (int)BT_GPU_params.m_gridSizeZ - 1));
return BT_GPU___mul24(BT_GPU___mul24(gridPos.z, BT_GPU_params.m_gridSizeY), BT_GPU_params.m_gridSizeX) + BT_GPU___mul24(gridPos.y, BT_GPU_params.m_gridSizeX) + gridPos.x;
} // bt3DGrid_calcGridHash()
//----------------------------------------------------------------------------------------
// calculate grid hash value for each body using its AABB
BT_GPU___global__ void calcHashAABBD(bt3DGrid3F1U* pAABB, uint2* pHash, uint numBodies)
{
int index = BT_GPU___mul24(BT_GPU_blockIdx.x, BT_GPU_blockDim.x) + BT_GPU_threadIdx.x;
if(index >= (int)numBodies)
{
return;
}
bt3DGrid3F1U bbMin = pAABB[index*2];
bt3DGrid3F1U bbMax = pAABB[index*2 + 1];
float4 pos;
pos.x = (bbMin.fx + bbMax.fx) * 0.5f;
pos.y = (bbMin.fy + bbMax.fy) * 0.5f;
pos.z = (bbMin.fz + bbMax.fz) * 0.5f;
// get address in grid
int3 gridPos = bt3DGrid_calcGridPos(pos);
uint gridHash = bt3DGrid_calcGridHash(gridPos);
// store grid hash and body index
pHash[index] = BT_GPU_make_uint2(gridHash, index);
} // calcHashAABBD()
//----------------------------------------------------------------------------------------
BT_GPU___global__ void findCellStartD(uint2* pHash, uint* cellStart, uint numBodies)
{
int index = BT_GPU___mul24(BT_GPU_blockIdx.x, BT_GPU_blockDim.x) + BT_GPU_threadIdx.x;
if(index >= (int)numBodies)
{
return;
}
uint2 sortedData = pHash[index];
// Load hash data into shared memory so that we can look
// at neighboring body's hash value without loading
// two hash values per thread
BT_GPU___shared__ uint sharedHash[257];
sharedHash[BT_GPU_threadIdx.x+1] = sortedData.x;
if((index > 0) && (BT_GPU_threadIdx.x == 0))
{
// first thread in block must load neighbor body hash
volatile uint2 prevData = pHash[index-1];
sharedHash[0] = prevData.x;
}
BT_GPU___syncthreads();
if((index == 0) || (sortedData.x != sharedHash[BT_GPU_threadIdx.x]))
{
cellStart[sortedData.x] = index;
}
} // findCellStartD()
//----------------------------------------------------------------------------------------
BT_GPU___device__ uint cudaTestAABBOverlap(bt3DGrid3F1U min0, bt3DGrid3F1U max0, bt3DGrid3F1U min1, bt3DGrid3F1U max1)
{
return (min0.fx <= max1.fx)&& (min1.fx <= max0.fx) &&
(min0.fy <= max1.fy)&& (min1.fy <= max0.fy) &&
(min0.fz <= max1.fz)&& (min1.fz <= max0.fz);
} // cudaTestAABBOverlap()
//----------------------------------------------------------------------------------------
BT_GPU___device__ void findPairsInCell( int3 gridPos,
uint index,
uint2* pHash,
uint* pCellStart,
bt3DGrid3F1U* pAABB,
uint* pPairBuff,
uint2* pPairBuffStartCurr,
uint numBodies)
{
if ( (gridPos.x < 0) || (gridPos.x > (int)BT_GPU_params.m_gridSizeX - 1)
|| (gridPos.y < 0) || (gridPos.y > (int)BT_GPU_params.m_gridSizeY - 1)
|| (gridPos.z < 0) || (gridPos.z > (int)BT_GPU_params.m_gridSizeZ - 1))
{
return;
}
uint gridHash = bt3DGrid_calcGridHash(gridPos);
// get start of bucket for this cell
uint bucketStart = pCellStart[gridHash];
if (bucketStart == 0xffffffff)
{
return; // cell empty
}
// iterate over bodies in this cell
uint2 sortedData = pHash[index];
uint unsorted_indx = sortedData.y;
bt3DGrid3F1U min0 = BT_GPU_FETCH(pAABB, unsorted_indx*2);
bt3DGrid3F1U max0 = BT_GPU_FETCH(pAABB, unsorted_indx*2 + 1);
uint handleIndex = min0.uw;
uint2 start_curr = pPairBuffStartCurr[handleIndex];
uint start = start_curr.x;
uint curr = start_curr.y;
uint2 start_curr_next = pPairBuffStartCurr[handleIndex+1];
uint curr_max = start_curr_next.x - start - 1;
uint bucketEnd = bucketStart + BT_GPU_params.m_maxBodiesPerCell;
bucketEnd = (bucketEnd > numBodies) ? numBodies : bucketEnd;
for(uint index2 = bucketStart; index2 < bucketEnd; index2++)
{
uint2 cellData = pHash[index2];
if (cellData.x != gridHash)
{
break; // no longer in same bucket
}
uint unsorted_indx2 = cellData.y;
if (unsorted_indx2 < unsorted_indx) // check not colliding with self
{
bt3DGrid3F1U min1 = BT_GPU_FETCH(pAABB, unsorted_indx2*2);
bt3DGrid3F1U max1 = BT_GPU_FETCH(pAABB, unsorted_indx2*2 + 1);
if(cudaTestAABBOverlap(min0, max0, min1, max1))
{
uint handleIndex2 = min1.uw;
uint k;
for(k = 0; k < curr; k++)
{
uint old_pair = pPairBuff[start+k] & (~BT_3DGRID_PAIR_ANY_FLG);
if(old_pair == handleIndex2)
{
pPairBuff[start+k] |= BT_3DGRID_PAIR_FOUND_FLG;
break;
}
}
if(k == curr)
{
if(curr >= curr_max)
{ // not a good solution, but let's avoid crash
break;
}
pPairBuff[start+curr] = handleIndex2 | BT_3DGRID_PAIR_NEW_FLG;
curr++;
}
}
}
}
pPairBuffStartCurr[handleIndex] = BT_GPU_make_uint2(start, curr);
return;
} // findPairsInCell()
//----------------------------------------------------------------------------------------
BT_GPU___global__ void findOverlappingPairsD( bt3DGrid3F1U* pAABB, uint2* pHash, uint* pCellStart,
uint* pPairBuff, uint2* pPairBuffStartCurr, uint numBodies)
{
int index = BT_GPU___mul24(BT_GPU_blockIdx.x, BT_GPU_blockDim.x) + BT_GPU_threadIdx.x;
if(index >= (int)numBodies)
{
return;
}
uint2 sortedData = pHash[index];
uint unsorted_indx = sortedData.y;
bt3DGrid3F1U bbMin = BT_GPU_FETCH(pAABB, unsorted_indx*2);
bt3DGrid3F1U bbMax = BT_GPU_FETCH(pAABB, unsorted_indx*2 + 1);
float4 pos;
pos.x = (bbMin.fx + bbMax.fx) * 0.5f;
pos.y = (bbMin.fy + bbMax.fy) * 0.5f;
pos.z = (bbMin.fz + bbMax.fz) * 0.5f;
// get address in grid
int3 gridPos = bt3DGrid_calcGridPos(pos);
// examine only neighbouring cells
for(int z=-1; z<=1; z++) {
for(int y=-1; y<=1; y++) {
for(int x=-1; x<=1; x++) {
findPairsInCell(gridPos + BT_GPU_make_int3(x, y, z), index, pHash, pCellStart, pAABB, pPairBuff, pPairBuffStartCurr, numBodies);
}
}
}
} // findOverlappingPairsD()
//----------------------------------------------------------------------------------------
BT_GPU___global__ void findPairsLargeD( bt3DGrid3F1U* pAABB, uint2* pHash, uint* pCellStart, uint* pPairBuff,
uint2* pPairBuffStartCurr, uint numBodies, uint numLarge)
{
int index = BT_GPU___mul24(BT_GPU_blockIdx.x, BT_GPU_blockDim.x) + BT_GPU_threadIdx.x;
if(index >= (int)numBodies)
{
return;
}
uint2 sortedData = pHash[index];
uint unsorted_indx = sortedData.y;
bt3DGrid3F1U min0 = BT_GPU_FETCH(pAABB, unsorted_indx*2);
bt3DGrid3F1U max0 = BT_GPU_FETCH(pAABB, unsorted_indx*2 + 1);
uint handleIndex = min0.uw;
uint2 start_curr = pPairBuffStartCurr[handleIndex];
uint start = start_curr.x;
uint curr = start_curr.y;
uint2 start_curr_next = pPairBuffStartCurr[handleIndex+1];
uint curr_max = start_curr_next.x - start - 1;
for(uint i = 0; i < numLarge; i++)
{
uint indx2 = numBodies + i;
bt3DGrid3F1U min1 = BT_GPU_FETCH(pAABB, indx2*2);
bt3DGrid3F1U max1 = BT_GPU_FETCH(pAABB, indx2*2 + 1);
if(cudaTestAABBOverlap(min0, max0, min1, max1))
{
uint k;
uint handleIndex2 = min1.uw;
for(k = 0; k < curr; k++)
{
uint old_pair = pPairBuff[start+k] & (~BT_3DGRID_PAIR_ANY_FLG);
if(old_pair == handleIndex2)
{
pPairBuff[start+k] |= BT_3DGRID_PAIR_FOUND_FLG;
break;
}
}
if(k == curr)
{
pPairBuff[start+curr] = handleIndex2 | BT_3DGRID_PAIR_NEW_FLG;
if(curr >= curr_max)
{ // not a good solution, but let's avoid crash
break;
}
curr++;
}
}
}
pPairBuffStartCurr[handleIndex] = BT_GPU_make_uint2(start, curr);
return;
} // findPairsLargeD()
//----------------------------------------------------------------------------------------
BT_GPU___global__ void computePairCacheChangesD(uint* pPairBuff, uint2* pPairBuffStartCurr,
uint* pPairScan, bt3DGrid3F1U* pAABB, uint numBodies)
{
int index = BT_GPU___mul24(BT_GPU_blockIdx.x, BT_GPU_blockDim.x) + BT_GPU_threadIdx.x;
if(index >= (int)numBodies)
{
return;
}
bt3DGrid3F1U bbMin = pAABB[index * 2];
uint handleIndex = bbMin.uw;
uint2 start_curr = pPairBuffStartCurr[handleIndex];
uint start = start_curr.x;
uint curr = start_curr.y;
uint *pInp = pPairBuff + start;
uint num_changes = 0;
for(uint k = 0; k < curr; k++, pInp++)
{
if(!((*pInp) & BT_3DGRID_PAIR_FOUND_FLG))
{
num_changes++;
}
}
pPairScan[index+1] = num_changes;
} // computePairCacheChangesD()
//----------------------------------------------------------------------------------------
BT_GPU___global__ void squeezeOverlappingPairBuffD(uint* pPairBuff, uint2* pPairBuffStartCurr, uint* pPairScan,
uint* pPairOut, bt3DGrid3F1U* pAABB, uint numBodies)
{
int index = BT_GPU___mul24(BT_GPU_blockIdx.x, BT_GPU_blockDim.x) + BT_GPU_threadIdx.x;
if(index >= (int)numBodies)
{
return;
}
bt3DGrid3F1U bbMin = pAABB[index * 2];
uint handleIndex = bbMin.uw;
uint2 start_curr = pPairBuffStartCurr[handleIndex];
uint start = start_curr.x;
uint curr = start_curr.y;
uint* pInp = pPairBuff + start;
uint* pOut = pPairOut + pPairScan[index];
uint* pOut2 = pInp;
uint num = 0;
for(uint k = 0; k < curr; k++, pInp++)
{
if(!((*pInp) & BT_3DGRID_PAIR_FOUND_FLG))
{
*pOut = *pInp;
pOut++;
}
if((*pInp) & BT_3DGRID_PAIR_ANY_FLG)
{
*pOut2 = (*pInp) & (~BT_3DGRID_PAIR_ANY_FLG);
pOut2++;
num++;
}
}
pPairBuffStartCurr[handleIndex] = BT_GPU_make_uint2(start, num);
} // squeezeOverlappingPairBuffD()
//----------------------------------------------------------------------------------------
//----------------------------------------------------------------------------------------
//----------------------------------------------------------------------------------------
//----------------------------------------------------------------------------------------
// E N D O F K E R N E L F U N C T I O N S
//----------------------------------------------------------------------------------------
//----------------------------------------------------------------------------------------
//----------------------------------------------------------------------------------------
//----------------------------------------------------------------------------------------
extern "C"
{
//----------------------------------------------------------------------------------------
void BT_GPU_PREF(calcHashAABB)(bt3DGrid3F1U* pAABB, unsigned int* hash, unsigned int numBodies)
{
int numThreads, numBlocks;
BT_GPU_PREF(computeGridSize)(numBodies, 256, numBlocks, numThreads);
// execute the kernel
BT_GPU_EXECKERNEL(numBlocks, numThreads, calcHashAABBD, (pAABB, (uint2*)hash, numBodies));
// check if kernel invocation generated an error
BT_GPU_CHECK_ERROR("calcHashAABBD kernel execution failed");
} // calcHashAABB()
//----------------------------------------------------------------------------------------
void BT_GPU_PREF(findCellStart(unsigned int* hash, unsigned int* cellStart, unsigned int numBodies, unsigned int numCells))
{
int numThreads, numBlocks;
BT_GPU_PREF(computeGridSize)(numBodies, 256, numBlocks, numThreads);
BT_GPU_SAFE_CALL(BT_GPU_Memset(cellStart, 0xffffffff, numCells*sizeof(uint)));
BT_GPU_EXECKERNEL(numBlocks, numThreads, findCellStartD, ((uint2*)hash, (uint*)cellStart, numBodies));
BT_GPU_CHECK_ERROR("Kernel execution failed: findCellStartD");
} // findCellStart()
//----------------------------------------------------------------------------------------
void BT_GPU_PREF(findOverlappingPairs(bt3DGrid3F1U* pAABB, unsigned int* pHash, unsigned int* pCellStart, unsigned int* pPairBuff, unsigned int* pPairBuffStartCurr, unsigned int numBodies))
{
#if B_CUDA_USE_TEX
BT_GPU_SAFE_CALL(cudaBindTexture(0, pAABBTex, pAABB, numBodies * 2 * sizeof(bt3DGrid3F1U)));
#endif
int numThreads, numBlocks;
BT_GPU_PREF(computeGridSize)(numBodies, 64, numBlocks, numThreads);
BT_GPU_EXECKERNEL(numBlocks, numThreads, findOverlappingPairsD, (pAABB,(uint2*)pHash,(uint*)pCellStart,(uint*)pPairBuff,(uint2*)pPairBuffStartCurr,numBodies));
BT_GPU_CHECK_ERROR("Kernel execution failed: bt_CudaFindOverlappingPairsD");
#if B_CUDA_USE_TEX
BT_GPU_SAFE_CALL(cudaUnbindTexture(pAABBTex));
#endif
} // findOverlappingPairs()
//----------------------------------------------------------------------------------------
void BT_GPU_PREF(findPairsLarge(bt3DGrid3F1U* pAABB, unsigned int* pHash, unsigned int* pCellStart, unsigned int* pPairBuff, unsigned int* pPairBuffStartCurr, unsigned int numBodies, unsigned int numLarge))
{
#if B_CUDA_USE_TEX
BT_GPU_SAFE_CALL(cudaBindTexture(0, pAABBTex, pAABB, (numBodies+numLarge) * 2 * sizeof(bt3DGrid3F1U)));
#endif
int numThreads, numBlocks;
BT_GPU_PREF(computeGridSize)(numBodies, 64, numBlocks, numThreads);
BT_GPU_EXECKERNEL(numBlocks, numThreads, findPairsLargeD, (pAABB,(uint2*)pHash,(uint*)pCellStart,(uint*)pPairBuff,(uint2*)pPairBuffStartCurr,numBodies,numLarge));
BT_GPU_CHECK_ERROR("Kernel execution failed: btCuda_findPairsLargeD");
#if B_CUDA_USE_TEX
BT_GPU_SAFE_CALL(cudaUnbindTexture(pAABBTex));
#endif
} // findPairsLarge()
//----------------------------------------------------------------------------------------
void BT_GPU_PREF(computePairCacheChanges(unsigned int* pPairBuff, unsigned int* pPairBuffStartCurr, unsigned int* pPairScan, bt3DGrid3F1U* pAABB, unsigned int numBodies))
{
int numThreads, numBlocks;
BT_GPU_PREF(computeGridSize)(numBodies, 256, numBlocks, numThreads);
BT_GPU_EXECKERNEL(numBlocks, numThreads, computePairCacheChangesD, ((uint*)pPairBuff,(uint2*)pPairBuffStartCurr,(uint*)pPairScan,pAABB,numBodies));
BT_GPU_CHECK_ERROR("Kernel execution failed: btCudaComputePairCacheChangesD");
} // computePairCacheChanges()
//----------------------------------------------------------------------------------------
void BT_GPU_PREF(squeezeOverlappingPairBuff(unsigned int* pPairBuff, unsigned int* pPairBuffStartCurr, unsigned int* pPairScan, unsigned int* pPairOut, bt3DGrid3F1U* pAABB, unsigned int numBodies))
{
int numThreads, numBlocks;
BT_GPU_PREF(computeGridSize)(numBodies, 256, numBlocks, numThreads);
BT_GPU_EXECKERNEL(numBlocks, numThreads, squeezeOverlappingPairBuffD, ((uint*)pPairBuff,(uint2*)pPairBuffStartCurr,(uint*)pPairScan,(uint*)pPairOut,pAABB,numBodies));
BT_GPU_CHECK_ERROR("Kernel execution failed: btCudaSqueezeOverlappingPairBuffD");
} // btCuda_squeezeOverlappingPairBuff()
//------------------------------------------------------------------------------------------------
} // extern "C"
//------------------------------------------------------------------------------------------------
//------------------------------------------------------------------------------------------------
//------------------------------------------------------------------------------------------------
/*
Bullet Continuous Collision Detection and Physics Library, http://bulletphysics.org
Copyright (C) 2006, 2009 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.
*/
//----------------------------------------------------------------------------------------
//----------------------------------------------------------------------------------------
//----------------------------------------------------------------------------------------
//----------------------------------------------------------------------------------------
//----------------------------------------------------------------------------------------
// K E R N E L F U N C T I O N S
//----------------------------------------------------------------------------------------
//----------------------------------------------------------------------------------------
//----------------------------------------------------------------------------------------
//----------------------------------------------------------------------------------------
//----------------------------------------------------------------------------------------
//----------------------------------------------------------------------------------------
// calculate position in uniform grid
BT_GPU___device__ int3 bt3DGrid_calcGridPos(float4 p)
{
int3 gridPos;
gridPos.x = (int)floor((p.x - BT_GPU_params.m_worldOriginX) / BT_GPU_params.m_cellSizeX);
gridPos.y = (int)floor((p.y - BT_GPU_params.m_worldOriginY) / BT_GPU_params.m_cellSizeY);
gridPos.z = (int)floor((p.z - BT_GPU_params.m_worldOriginZ) / BT_GPU_params.m_cellSizeZ);
return gridPos;
} // bt3DGrid_calcGridPos()
//----------------------------------------------------------------------------------------
// calculate address in grid from position (clamping to edges)
BT_GPU___device__ uint bt3DGrid_calcGridHash(int3 gridPos)
{
gridPos.x = BT_GPU_max(0, BT_GPU_min(gridPos.x, (int)BT_GPU_params.m_gridSizeX - 1));
gridPos.y = BT_GPU_max(0, BT_GPU_min(gridPos.y, (int)BT_GPU_params.m_gridSizeY - 1));
gridPos.z = BT_GPU_max(0, BT_GPU_min(gridPos.z, (int)BT_GPU_params.m_gridSizeZ - 1));
return BT_GPU___mul24(BT_GPU___mul24(gridPos.z, BT_GPU_params.m_gridSizeY), BT_GPU_params.m_gridSizeX) + BT_GPU___mul24(gridPos.y, BT_GPU_params.m_gridSizeX) + gridPos.x;
} // bt3DGrid_calcGridHash()
//----------------------------------------------------------------------------------------
// calculate grid hash value for each body using its AABB
BT_GPU___global__ void calcHashAABBD(bt3DGrid3F1U* pAABB, uint2* pHash, uint numBodies)
{
int index = BT_GPU___mul24(BT_GPU_blockIdx.x, BT_GPU_blockDim.x) + BT_GPU_threadIdx.x;
if(index >= (int)numBodies)
{
return;
}
bt3DGrid3F1U bbMin = pAABB[index*2];
bt3DGrid3F1U bbMax = pAABB[index*2 + 1];
float4 pos;
pos.x = (bbMin.fx + bbMax.fx) * 0.5f;
pos.y = (bbMin.fy + bbMax.fy) * 0.5f;
pos.z = (bbMin.fz + bbMax.fz) * 0.5f;
// get address in grid
int3 gridPos = bt3DGrid_calcGridPos(pos);
uint gridHash = bt3DGrid_calcGridHash(gridPos);
// store grid hash and body index
pHash[index] = BT_GPU_make_uint2(gridHash, index);
} // calcHashAABBD()
//----------------------------------------------------------------------------------------
BT_GPU___global__ void findCellStartD(uint2* pHash, uint* cellStart, uint numBodies)
{
int index = BT_GPU___mul24(BT_GPU_blockIdx.x, BT_GPU_blockDim.x) + BT_GPU_threadIdx.x;
if(index >= (int)numBodies)
{
return;
}
uint2 sortedData = pHash[index];
// Load hash data into shared memory so that we can look
// at neighboring body's hash value without loading
// two hash values per thread
BT_GPU___shared__ uint sharedHash[257];
sharedHash[BT_GPU_threadIdx.x+1] = sortedData.x;
if((index > 0) && (BT_GPU_threadIdx.x == 0))
{
// first thread in block must load neighbor body hash
volatile uint2 prevData = pHash[index-1];
sharedHash[0] = prevData.x;
}
BT_GPU___syncthreads();
if((index == 0) || (sortedData.x != sharedHash[BT_GPU_threadIdx.x]))
{
cellStart[sortedData.x] = index;
}
} // findCellStartD()
//----------------------------------------------------------------------------------------
BT_GPU___device__ uint cudaTestAABBOverlap(bt3DGrid3F1U min0, bt3DGrid3F1U max0, bt3DGrid3F1U min1, bt3DGrid3F1U max1)
{
return (min0.fx <= max1.fx)&& (min1.fx <= max0.fx) &&
(min0.fy <= max1.fy)&& (min1.fy <= max0.fy) &&
(min0.fz <= max1.fz)&& (min1.fz <= max0.fz);
} // cudaTestAABBOverlap()
//----------------------------------------------------------------------------------------
BT_GPU___device__ void findPairsInCell( int3 gridPos,
uint index,
uint2* pHash,
uint* pCellStart,
bt3DGrid3F1U* pAABB,
uint* pPairBuff,
uint2* pPairBuffStartCurr,
uint numBodies)
{
if ( (gridPos.x < 0) || (gridPos.x > (int)BT_GPU_params.m_gridSizeX - 1)
|| (gridPos.y < 0) || (gridPos.y > (int)BT_GPU_params.m_gridSizeY - 1)
|| (gridPos.z < 0) || (gridPos.z > (int)BT_GPU_params.m_gridSizeZ - 1))
{
return;
}
uint gridHash = bt3DGrid_calcGridHash(gridPos);
// get start of bucket for this cell
uint bucketStart = pCellStart[gridHash];
if (bucketStart == 0xffffffff)
{
return; // cell empty
}
// iterate over bodies in this cell
uint2 sortedData = pHash[index];
uint unsorted_indx = sortedData.y;
bt3DGrid3F1U min0 = BT_GPU_FETCH(pAABB, unsorted_indx*2);
bt3DGrid3F1U max0 = BT_GPU_FETCH(pAABB, unsorted_indx*2 + 1);
uint handleIndex = min0.uw;
uint2 start_curr = pPairBuffStartCurr[handleIndex];
uint start = start_curr.x;
uint curr = start_curr.y;
uint2 start_curr_next = pPairBuffStartCurr[handleIndex+1];
uint curr_max = start_curr_next.x - start - 1;
uint bucketEnd = bucketStart + BT_GPU_params.m_maxBodiesPerCell;
bucketEnd = (bucketEnd > numBodies) ? numBodies : bucketEnd;
for(uint index2 = bucketStart; index2 < bucketEnd; index2++)
{
uint2 cellData = pHash[index2];
if (cellData.x != gridHash)
{
break; // no longer in same bucket
}
uint unsorted_indx2 = cellData.y;
if (unsorted_indx2 < unsorted_indx) // check not colliding with self
{
bt3DGrid3F1U min1 = BT_GPU_FETCH(pAABB, unsorted_indx2*2);
bt3DGrid3F1U max1 = BT_GPU_FETCH(pAABB, unsorted_indx2*2 + 1);
if(cudaTestAABBOverlap(min0, max0, min1, max1))
{
uint handleIndex2 = min1.uw;
uint k;
for(k = 0; k < curr; k++)
{
uint old_pair = pPairBuff[start+k] & (~BT_3DGRID_PAIR_ANY_FLG);
if(old_pair == handleIndex2)
{
pPairBuff[start+k] |= BT_3DGRID_PAIR_FOUND_FLG;
break;
}
}
if(k == curr)
{
if(curr >= curr_max)
{ // not a good solution, but let's avoid crash
break;
}
pPairBuff[start+curr] = handleIndex2 | BT_3DGRID_PAIR_NEW_FLG;
curr++;
}
}
}
}
pPairBuffStartCurr[handleIndex] = BT_GPU_make_uint2(start, curr);
return;
} // findPairsInCell()
//----------------------------------------------------------------------------------------
BT_GPU___global__ void findOverlappingPairsD( bt3DGrid3F1U* pAABB, uint2* pHash, uint* pCellStart,
uint* pPairBuff, uint2* pPairBuffStartCurr, uint numBodies)
{
int index = BT_GPU___mul24(BT_GPU_blockIdx.x, BT_GPU_blockDim.x) + BT_GPU_threadIdx.x;
if(index >= (int)numBodies)
{
return;
}
uint2 sortedData = pHash[index];
uint unsorted_indx = sortedData.y;
bt3DGrid3F1U bbMin = BT_GPU_FETCH(pAABB, unsorted_indx*2);
bt3DGrid3F1U bbMax = BT_GPU_FETCH(pAABB, unsorted_indx*2 + 1);
float4 pos;
pos.x = (bbMin.fx + bbMax.fx) * 0.5f;
pos.y = (bbMin.fy + bbMax.fy) * 0.5f;
pos.z = (bbMin.fz + bbMax.fz) * 0.5f;
// get address in grid
int3 gridPos = bt3DGrid_calcGridPos(pos);
// examine only neighbouring cells
for(int z=-1; z<=1; z++) {
for(int y=-1; y<=1; y++) {
for(int x=-1; x<=1; x++) {
findPairsInCell(gridPos + BT_GPU_make_int3(x, y, z), index, pHash, pCellStart, pAABB, pPairBuff, pPairBuffStartCurr, numBodies);
}
}
}
} // findOverlappingPairsD()
//----------------------------------------------------------------------------------------
BT_GPU___global__ void findPairsLargeD( bt3DGrid3F1U* pAABB, uint2* pHash, uint* pCellStart, uint* pPairBuff,
uint2* pPairBuffStartCurr, uint numBodies, uint numLarge)
{
int index = BT_GPU___mul24(BT_GPU_blockIdx.x, BT_GPU_blockDim.x) + BT_GPU_threadIdx.x;
if(index >= (int)numBodies)
{
return;
}
uint2 sortedData = pHash[index];
uint unsorted_indx = sortedData.y;
bt3DGrid3F1U min0 = BT_GPU_FETCH(pAABB, unsorted_indx*2);
bt3DGrid3F1U max0 = BT_GPU_FETCH(pAABB, unsorted_indx*2 + 1);
uint handleIndex = min0.uw;
uint2 start_curr = pPairBuffStartCurr[handleIndex];
uint start = start_curr.x;
uint curr = start_curr.y;
uint2 start_curr_next = pPairBuffStartCurr[handleIndex+1];
uint curr_max = start_curr_next.x - start - 1;
for(uint i = 0; i < numLarge; i++)
{
uint indx2 = numBodies + i;
bt3DGrid3F1U min1 = BT_GPU_FETCH(pAABB, indx2*2);
bt3DGrid3F1U max1 = BT_GPU_FETCH(pAABB, indx2*2 + 1);
if(cudaTestAABBOverlap(min0, max0, min1, max1))
{
uint k;
uint handleIndex2 = min1.uw;
for(k = 0; k < curr; k++)
{
uint old_pair = pPairBuff[start+k] & (~BT_3DGRID_PAIR_ANY_FLG);
if(old_pair == handleIndex2)
{
pPairBuff[start+k] |= BT_3DGRID_PAIR_FOUND_FLG;
break;
}
}
if(k == curr)
{
pPairBuff[start+curr] = handleIndex2 | BT_3DGRID_PAIR_NEW_FLG;
if(curr >= curr_max)
{ // not a good solution, but let's avoid crash
break;
}
curr++;
}
}
}
pPairBuffStartCurr[handleIndex] = BT_GPU_make_uint2(start, curr);
return;
} // findPairsLargeD()
//----------------------------------------------------------------------------------------
BT_GPU___global__ void computePairCacheChangesD(uint* pPairBuff, uint2* pPairBuffStartCurr,
uint* pPairScan, bt3DGrid3F1U* pAABB, uint numBodies)
{
int index = BT_GPU___mul24(BT_GPU_blockIdx.x, BT_GPU_blockDim.x) + BT_GPU_threadIdx.x;
if(index >= (int)numBodies)
{
return;
}
bt3DGrid3F1U bbMin = pAABB[index * 2];
uint handleIndex = bbMin.uw;
uint2 start_curr = pPairBuffStartCurr[handleIndex];
uint start = start_curr.x;
uint curr = start_curr.y;
uint *pInp = pPairBuff + start;
uint num_changes = 0;
for(uint k = 0; k < curr; k++, pInp++)
{
if(!((*pInp) & BT_3DGRID_PAIR_FOUND_FLG))
{
num_changes++;
}
}
pPairScan[index+1] = num_changes;
} // computePairCacheChangesD()
//----------------------------------------------------------------------------------------
BT_GPU___global__ void squeezeOverlappingPairBuffD(uint* pPairBuff, uint2* pPairBuffStartCurr, uint* pPairScan,
uint* pPairOut, bt3DGrid3F1U* pAABB, uint numBodies)
{
int index = BT_GPU___mul24(BT_GPU_blockIdx.x, BT_GPU_blockDim.x) + BT_GPU_threadIdx.x;
if(index >= (int)numBodies)
{
return;
}
bt3DGrid3F1U bbMin = pAABB[index * 2];
uint handleIndex = bbMin.uw;
uint2 start_curr = pPairBuffStartCurr[handleIndex];
uint start = start_curr.x;
uint curr = start_curr.y;
uint* pInp = pPairBuff + start;
uint* pOut = pPairOut + pPairScan[index];
uint* pOut2 = pInp;
uint num = 0;
for(uint k = 0; k < curr; k++, pInp++)
{
if(!((*pInp) & BT_3DGRID_PAIR_FOUND_FLG))
{
*pOut = *pInp;
pOut++;
}
if((*pInp) & BT_3DGRID_PAIR_ANY_FLG)
{
*pOut2 = (*pInp) & (~BT_3DGRID_PAIR_ANY_FLG);
pOut2++;
num++;
}
}
pPairBuffStartCurr[handleIndex] = BT_GPU_make_uint2(start, num);
} // squeezeOverlappingPairBuffD()
//----------------------------------------------------------------------------------------
//----------------------------------------------------------------------------------------
//----------------------------------------------------------------------------------------
//----------------------------------------------------------------------------------------
// E N D O F K E R N E L F U N C T I O N S
//----------------------------------------------------------------------------------------
//----------------------------------------------------------------------------------------
//----------------------------------------------------------------------------------------
//----------------------------------------------------------------------------------------
extern "C"
{
//----------------------------------------------------------------------------------------
void BT_GPU_PREF(calcHashAABB)(bt3DGrid3F1U* pAABB, unsigned int* hash, unsigned int numBodies)
{
int numThreads, numBlocks;
BT_GPU_PREF(computeGridSize)(numBodies, 256, numBlocks, numThreads);
// execute the kernel
BT_GPU_EXECKERNEL(numBlocks, numThreads, calcHashAABBD, (pAABB, (uint2*)hash, numBodies));
// check if kernel invocation generated an error
BT_GPU_CHECK_ERROR("calcHashAABBD kernel execution failed");
} // calcHashAABB()
//----------------------------------------------------------------------------------------
void BT_GPU_PREF(findCellStart(unsigned int* hash, unsigned int* cellStart, unsigned int numBodies, unsigned int numCells))
{
int numThreads, numBlocks;
BT_GPU_PREF(computeGridSize)(numBodies, 256, numBlocks, numThreads);
BT_GPU_SAFE_CALL(BT_GPU_Memset(cellStart, 0xffffffff, numCells*sizeof(uint)));
BT_GPU_EXECKERNEL(numBlocks, numThreads, findCellStartD, ((uint2*)hash, (uint*)cellStart, numBodies));
BT_GPU_CHECK_ERROR("Kernel execution failed: findCellStartD");
} // findCellStart()
//----------------------------------------------------------------------------------------
void BT_GPU_PREF(findOverlappingPairs(bt3DGrid3F1U* pAABB, unsigned int* pHash, unsigned int* pCellStart, unsigned int* pPairBuff, unsigned int* pPairBuffStartCurr, unsigned int numBodies))
{
#if B_CUDA_USE_TEX
BT_GPU_SAFE_CALL(cudaBindTexture(0, pAABBTex, pAABB, numBodies * 2 * sizeof(bt3DGrid3F1U)));
#endif
int numThreads, numBlocks;
BT_GPU_PREF(computeGridSize)(numBodies, 64, numBlocks, numThreads);
BT_GPU_EXECKERNEL(numBlocks, numThreads, findOverlappingPairsD, (pAABB,(uint2*)pHash,(uint*)pCellStart,(uint*)pPairBuff,(uint2*)pPairBuffStartCurr,numBodies));
BT_GPU_CHECK_ERROR("Kernel execution failed: bt_CudaFindOverlappingPairsD");
#if B_CUDA_USE_TEX
BT_GPU_SAFE_CALL(cudaUnbindTexture(pAABBTex));
#endif
} // findOverlappingPairs()
//----------------------------------------------------------------------------------------
void BT_GPU_PREF(findPairsLarge(bt3DGrid3F1U* pAABB, unsigned int* pHash, unsigned int* pCellStart, unsigned int* pPairBuff, unsigned int* pPairBuffStartCurr, unsigned int numBodies, unsigned int numLarge))
{
#if B_CUDA_USE_TEX
BT_GPU_SAFE_CALL(cudaBindTexture(0, pAABBTex, pAABB, (numBodies+numLarge) * 2 * sizeof(bt3DGrid3F1U)));
#endif
int numThreads, numBlocks;
BT_GPU_PREF(computeGridSize)(numBodies, 64, numBlocks, numThreads);
BT_GPU_EXECKERNEL(numBlocks, numThreads, findPairsLargeD, (pAABB,(uint2*)pHash,(uint*)pCellStart,(uint*)pPairBuff,(uint2*)pPairBuffStartCurr,numBodies,numLarge));
BT_GPU_CHECK_ERROR("Kernel execution failed: btCuda_findPairsLargeD");
#if B_CUDA_USE_TEX
BT_GPU_SAFE_CALL(cudaUnbindTexture(pAABBTex));
#endif
} // findPairsLarge()
//----------------------------------------------------------------------------------------
void BT_GPU_PREF(computePairCacheChanges(unsigned int* pPairBuff, unsigned int* pPairBuffStartCurr, unsigned int* pPairScan, bt3DGrid3F1U* pAABB, unsigned int numBodies))
{
int numThreads, numBlocks;
BT_GPU_PREF(computeGridSize)(numBodies, 256, numBlocks, numThreads);
BT_GPU_EXECKERNEL(numBlocks, numThreads, computePairCacheChangesD, ((uint*)pPairBuff,(uint2*)pPairBuffStartCurr,(uint*)pPairScan,pAABB,numBodies));
BT_GPU_CHECK_ERROR("Kernel execution failed: btCudaComputePairCacheChangesD");
} // computePairCacheChanges()
//----------------------------------------------------------------------------------------
void BT_GPU_PREF(squeezeOverlappingPairBuff(unsigned int* pPairBuff, unsigned int* pPairBuffStartCurr, unsigned int* pPairScan, unsigned int* pPairOut, bt3DGrid3F1U* pAABB, unsigned int numBodies))
{
int numThreads, numBlocks;
BT_GPU_PREF(computeGridSize)(numBodies, 256, numBlocks, numThreads);
BT_GPU_EXECKERNEL(numBlocks, numThreads, squeezeOverlappingPairBuffD, ((uint*)pPairBuff,(uint2*)pPairBuffStartCurr,(uint*)pPairScan,(uint*)pPairOut,pAABB,numBodies));
BT_GPU_CHECK_ERROR("Kernel execution failed: btCudaSqueezeOverlappingPairBuffD");
} // btCuda_squeezeOverlappingPairBuff()
//------------------------------------------------------------------------------------------------
} // extern "C"
//------------------------------------------------------------------------------------------------
//------------------------------------------------------------------------------------------------
//------------------------------------------------------------------------------------------------

View File

@@ -1,61 +1,61 @@
/*
Bullet Continuous Collision Detection and Physics Library, http://bulletphysics.org
Copyright (C) 2006, 2009 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.
*/
//----------------------------------------------------------------------------------------
// Shared definitions for GPU-based 3D Grid collision detection broadphase
//!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
// Keep this file free from Bullet headers
// it is included into both CUDA and CPU code
//!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
//----------------------------------------------------------------------------------------
#ifndef BTGPU3DGRIDBROADPHASESHAREDDEFS_H
#define BTGPU3DGRIDBROADPHASESHAREDDEFS_H
//----------------------------------------------------------------------------------------
#include "btGpu3DGridBroadphaseSharedTypes.h"
//----------------------------------------------------------------------------------------
extern "C"
{
//----------------------------------------------------------------------------------------
void BT_GPU_PREF(calcHashAABB)(bt3DGrid3F1U* pAABB, unsigned int* hash, unsigned int numBodies);
void BT_GPU_PREF(findCellStart)(unsigned int* hash, unsigned int* cellStart, unsigned int numBodies, unsigned int numCells);
void BT_GPU_PREF(findOverlappingPairs)(bt3DGrid3F1U* pAABB, unsigned int* pHash, unsigned int* pCellStart, unsigned int* pPairBuff, unsigned int* pPairBuffStartCurr, unsigned int numBodies);
void BT_GPU_PREF(findPairsLarge)(bt3DGrid3F1U* pAABB, unsigned int* pHash, unsigned int* pCellStart, unsigned int* pPairBuff, unsigned int* pPairBuffStartCurr, unsigned int numBodies, unsigned int numLarge);
void BT_GPU_PREF(computePairCacheChanges)(unsigned int* pPairBuff, unsigned int* pPairBuffStartCurr, unsigned int* pPairScan, bt3DGrid3F1U* pAABB, unsigned int numBodies);
void BT_GPU_PREF(squeezeOverlappingPairBuff)(unsigned int* pPairBuff, unsigned int* pPairBuffStartCurr, unsigned int* pPairScan, unsigned int* pPairOut, bt3DGrid3F1U* pAABB, unsigned int numBodies);
//----------------------------------------------------------------------------------------
} // extern "C"
//----------------------------------------------------------------------------------------
#endif // BTGPU3DGRIDBROADPHASESHAREDDEFS_H
/*
Bullet Continuous Collision Detection and Physics Library, http://bulletphysics.org
Copyright (C) 2006, 2009 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.
*/
//----------------------------------------------------------------------------------------
// Shared definitions for GPU-based 3D Grid collision detection broadphase
//!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
// Keep this file free from Bullet headers
// it is included into both CUDA and CPU code
//!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
//----------------------------------------------------------------------------------------
#ifndef BTGPU3DGRIDBROADPHASESHAREDDEFS_H
#define BTGPU3DGRIDBROADPHASESHAREDDEFS_H
//----------------------------------------------------------------------------------------
#include "btGpu3DGridBroadphaseSharedTypes.h"
//----------------------------------------------------------------------------------------
extern "C"
{
//----------------------------------------------------------------------------------------
void BT_GPU_PREF(calcHashAABB)(bt3DGrid3F1U* pAABB, unsigned int* hash, unsigned int numBodies);
void BT_GPU_PREF(findCellStart)(unsigned int* hash, unsigned int* cellStart, unsigned int numBodies, unsigned int numCells);
void BT_GPU_PREF(findOverlappingPairs)(bt3DGrid3F1U* pAABB, unsigned int* pHash, unsigned int* pCellStart, unsigned int* pPairBuff, unsigned int* pPairBuffStartCurr, unsigned int numBodies);
void BT_GPU_PREF(findPairsLarge)(bt3DGrid3F1U* pAABB, unsigned int* pHash, unsigned int* pCellStart, unsigned int* pPairBuff, unsigned int* pPairBuffStartCurr, unsigned int numBodies, unsigned int numLarge);
void BT_GPU_PREF(computePairCacheChanges)(unsigned int* pPairBuff, unsigned int* pPairBuffStartCurr, unsigned int* pPairScan, bt3DGrid3F1U* pAABB, unsigned int numBodies);
void BT_GPU_PREF(squeezeOverlappingPairBuff)(unsigned int* pPairBuff, unsigned int* pPairBuffStartCurr, unsigned int* pPairScan, unsigned int* pPairOut, bt3DGrid3F1U* pAABB, unsigned int numBodies);
//----------------------------------------------------------------------------------------
} // extern "C"
//----------------------------------------------------------------------------------------
#endif // BTGPU3DGRIDBROADPHASESHAREDDEFS_H

View File

@@ -1,67 +1,67 @@
/*
Bullet Continuous Collision Detection and Physics Library, http://bulletphysics.org
Copyright (C) 2006, 2009 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.
*/
//----------------------------------------------------------------------------------------
// Shared definitions for GPU-based 3D Grid collision detection broadphase
//!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
// Keep this file free from Bullet headers
// it is included into both CUDA and CPU code
//!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
//----------------------------------------------------------------------------------------
#ifndef BTGPU3DGRIDBROADPHASESHAREDTYPES_H
#define BTGPU3DGRIDBROADPHASESHAREDTYPES_H
//----------------------------------------------------------------------------------------
#define BT_3DGRID_PAIR_FOUND_FLG (0x40000000)
#define BT_3DGRID_PAIR_NEW_FLG (0x20000000)
#define BT_3DGRID_PAIR_ANY_FLG (BT_3DGRID_PAIR_FOUND_FLG | BT_3DGRID_PAIR_NEW_FLG)
//----------------------------------------------------------------------------------------
struct bt3DGridBroadphaseParams
{
unsigned int m_gridSizeX;
unsigned int m_gridSizeY;
unsigned int m_gridSizeZ;
unsigned int m_numCells;
float m_worldOriginX;
float m_worldOriginY;
float m_worldOriginZ;
float m_cellSizeX;
float m_cellSizeY;
float m_cellSizeZ;
unsigned int m_numBodies;
unsigned int m_maxBodiesPerCell;
};
//----------------------------------------------------------------------------------------
struct bt3DGrid3F1U
{
float fx;
float fy;
float fz;
unsigned int uw;
};
//----------------------------------------------------------------------------------------
#endif // BTGPU3DGRIDBROADPHASESHAREDTYPES_H
/*
Bullet Continuous Collision Detection and Physics Library, http://bulletphysics.org
Copyright (C) 2006, 2009 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.
*/
//----------------------------------------------------------------------------------------
// Shared definitions for GPU-based 3D Grid collision detection broadphase
//!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
// Keep this file free from Bullet headers
// it is included into both CUDA and CPU code
//!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
//----------------------------------------------------------------------------------------
#ifndef BTGPU3DGRIDBROADPHASESHAREDTYPES_H
#define BTGPU3DGRIDBROADPHASESHAREDTYPES_H
//----------------------------------------------------------------------------------------
#define BT_3DGRID_PAIR_FOUND_FLG (0x40000000)
#define BT_3DGRID_PAIR_NEW_FLG (0x20000000)
#define BT_3DGRID_PAIR_ANY_FLG (BT_3DGRID_PAIR_FOUND_FLG | BT_3DGRID_PAIR_NEW_FLG)
//----------------------------------------------------------------------------------------
struct bt3DGridBroadphaseParams
{
unsigned int m_gridSizeX;
unsigned int m_gridSizeY;
unsigned int m_gridSizeZ;
unsigned int m_numCells;
float m_worldOriginX;
float m_worldOriginY;
float m_worldOriginZ;
float m_cellSizeX;
float m_cellSizeY;
float m_cellSizeZ;
unsigned int m_numBodies;
unsigned int m_maxBodiesPerCell;
};
//----------------------------------------------------------------------------------------
struct bt3DGrid3F1U
{
float fx;
float fy;
float fz;
unsigned int uw;
};
//----------------------------------------------------------------------------------------
#endif // BTGPU3DGRIDBROADPHASESHAREDTYPES_H

View File

@@ -1,55 +1,55 @@
/*
Bullet Continuous Collision Detection and Physics Library, http://bulletphysics.org
Copyright (C) 2006, 2009 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.
*/
//----------------------------------------------------------------------------------------
// Shared code for GPU-based utilities
//!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
// Keep this file free from Bullet headers
// will be compiled by both CPU and CUDA compilers
// file with definitions of BT_GPU_xxx should be included first
//!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
//----------------------------------------------------------------------------------------
#include "btGpuUtilsSharedDefs.h"
//----------------------------------------------------------------------------------------
extern "C"
{
//----------------------------------------------------------------------------------------
//Round a / b to nearest higher integer value
int BT_GPU_PREF(iDivUp)(int a, int b)
{
return (a % b != 0) ? (a / b + 1) : (a / b);
} // iDivUp()
//----------------------------------------------------------------------------------------
// compute grid and thread block size for a given number of elements
void BT_GPU_PREF(computeGridSize)(int n, int blockSize, int &numBlocks, int &numThreads)
{
numThreads = BT_GPU_min(blockSize, n);
numBlocks = BT_GPU_PREF(iDivUp)(n, numThreads);
} // computeGridSize()
//----------------------------------------------------------------------------------------
} // extern "C"
/*
Bullet Continuous Collision Detection and Physics Library, http://bulletphysics.org
Copyright (C) 2006, 2009 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.
*/
//----------------------------------------------------------------------------------------
// Shared code for GPU-based utilities
//!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
// Keep this file free from Bullet headers
// will be compiled by both CPU and CUDA compilers
// file with definitions of BT_GPU_xxx should be included first
//!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
//----------------------------------------------------------------------------------------
#include "btGpuUtilsSharedDefs.h"
//----------------------------------------------------------------------------------------
extern "C"
{
//----------------------------------------------------------------------------------------
//Round a / b to nearest higher integer value
int BT_GPU_PREF(iDivUp)(int a, int b)
{
return (a % b != 0) ? (a / b + 1) : (a / b);
} // iDivUp()
//----------------------------------------------------------------------------------------
// compute grid and thread block size for a given number of elements
void BT_GPU_PREF(computeGridSize)(int n, int blockSize, int &numBlocks, int &numThreads)
{
numThreads = BT_GPU_min(blockSize, n);
numBlocks = BT_GPU_PREF(iDivUp)(n, numThreads);
} // computeGridSize()
//----------------------------------------------------------------------------------------
} // extern "C"

View File

@@ -1,52 +1,52 @@
/*
Bullet Continuous Collision Detection and Physics Library, http://bulletphysics.org
Copyright (C) 2006, 2007 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.
*/
// Shared definitions for GPU-based utilities
//!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
// Keep this file free from Bullet headers
// it is included into both CUDA and CPU code
// file with definitions of BT_GPU_xxx should be included first
//!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
#ifndef BTGPUUTILSDHAREDDEFS_H
#define BTGPUUTILSDHAREDDEFS_H
extern "C"
{
//Round a / b to nearest higher integer value
int BT_GPU_PREF(iDivUp)(int a, int b);
// compute grid and thread block size for a given number of elements
void BT_GPU_PREF(computeGridSize)(int n, int blockSize, int &numBlocks, int &numThreads);
void BT_GPU_PREF(allocateArray)(void** devPtr, unsigned int size);
void BT_GPU_PREF(freeArray)(void* devPtr);
void BT_GPU_PREF(copyArrayFromDevice)(void* host, const void* device, unsigned int size);
void BT_GPU_PREF(copyArrayToDevice)(void* device, const void* host, unsigned int size);
void BT_GPU_PREF(registerGLBufferObject(unsigned int vbo));
void* BT_GPU_PREF(mapGLBufferObject(unsigned int vbo));
void BT_GPU_PREF(unmapGLBufferObject(unsigned int vbo));
} // extern "C"
#endif // BTGPUUTILSDHAREDDEFS_H
/*
Bullet Continuous Collision Detection and Physics Library, http://bulletphysics.org
Copyright (C) 2006, 2007 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.
*/
// Shared definitions for GPU-based utilities
//!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
// Keep this file free from Bullet headers
// it is included into both CUDA and CPU code
// file with definitions of BT_GPU_xxx should be included first
//!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
#ifndef BTGPUUTILSDHAREDDEFS_H
#define BTGPUUTILSDHAREDDEFS_H
extern "C"
{
//Round a / b to nearest higher integer value
int BT_GPU_PREF(iDivUp)(int a, int b);
// compute grid and thread block size for a given number of elements
void BT_GPU_PREF(computeGridSize)(int n, int blockSize, int &numBlocks, int &numThreads);
void BT_GPU_PREF(allocateArray)(void** devPtr, unsigned int size);
void BT_GPU_PREF(freeArray)(void* devPtr);
void BT_GPU_PREF(copyArrayFromDevice)(void* host, const void* device, unsigned int size);
void BT_GPU_PREF(copyArrayToDevice)(void* device, const void* host, unsigned int size);
void BT_GPU_PREF(registerGLBufferObject(unsigned int vbo));
void* BT_GPU_PREF(mapGLBufferObject(unsigned int vbo));
void BT_GPU_PREF(unmapGLBufferObject(unsigned int vbo));
} // extern "C"
#endif // BTGPUUTILSDHAREDDEFS_H

View File

@@ -1,74 +1,74 @@
/*
Copyright (C) 2010 Sony Computer Entertainment Inc.
All rights reserved.
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.
*/
#include "btParallelConstraintSolver.h"
#include "BulletDynamics/ConstraintSolver/btContactSolverInfo.h"
btParallelConstraintSolver::btParallelConstraintSolver()
{
//initialize MiniCL here
}
btParallelConstraintSolver::~btParallelConstraintSolver()
{
//exit MiniCL
}
btScalar btParallelConstraintSolver::solveGroupCacheFriendlySetup(btCollisionObject** bodies,int numBodies,btPersistentManifold** manifoldPtr, int numManifolds,btTypedConstraint** constraints,int numConstraints,const btContactSolverInfo& infoGlobal,btIDebugDraw* debugDrawer,btStackAlloc* stackAlloc)
{
{
int i;
btPersistentManifold* manifold = 0;
// btCollisionObject* colObj0=0,*colObj1=0;
for (i=0;i<numManifolds;i++)
{
manifold = manifoldPtr[i];
convertContact(manifold,infoGlobal);
}
}
btContactSolverInfo info = infoGlobal;
int numConstraintPool = m_tmpSolverContactConstraintPool.size();
int numFrictionPool = m_tmpSolverContactFrictionConstraintPool.size();
///@todo: use stack allocator for such temporarily memory, same for solver bodies/constraints
m_orderTmpConstraintPool.resize(numConstraintPool);
m_orderFrictionConstraintPool.resize(numFrictionPool);
{
int i;
for (i=0;i<numConstraintPool;i++)
{
m_orderTmpConstraintPool[i] = i;
}
for (i=0;i<numFrictionPool;i++)
{
m_orderFrictionConstraintPool[i] = i;
}
}
return 0.f;
}
/*
Copyright (C) 2010 Sony Computer Entertainment Inc.
All rights reserved.
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.
*/
#include "btParallelConstraintSolver.h"
#include "BulletDynamics/ConstraintSolver/btContactSolverInfo.h"
btParallelConstraintSolver::btParallelConstraintSolver()
{
//initialize MiniCL here
}
btParallelConstraintSolver::~btParallelConstraintSolver()
{
//exit MiniCL
}
btScalar btParallelConstraintSolver::solveGroupCacheFriendlySetup(btCollisionObject** bodies,int numBodies,btPersistentManifold** manifoldPtr, int numManifolds,btTypedConstraint** constraints,int numConstraints,const btContactSolverInfo& infoGlobal,btIDebugDraw* debugDrawer,btStackAlloc* stackAlloc)
{
{
int i;
btPersistentManifold* manifold = 0;
// btCollisionObject* colObj0=0,*colObj1=0;
for (i=0;i<numManifolds;i++)
{
manifold = manifoldPtr[i];
convertContact(manifold,infoGlobal);
}
}
btContactSolverInfo info = infoGlobal;
int numConstraintPool = m_tmpSolverContactConstraintPool.size();
int numFrictionPool = m_tmpSolverContactFrictionConstraintPool.size();
///@todo: use stack allocator for such temporarily memory, same for solver bodies/constraints
m_orderTmpConstraintPool.resize(numConstraintPool);
m_orderFrictionConstraintPool.resize(numFrictionPool);
{
int i;
for (i=0;i<numConstraintPool;i++)
{
m_orderTmpConstraintPool[i] = i;
}
for (i=0;i<numFrictionPool;i++)
{
m_orderFrictionConstraintPool[i] = i;
}
}
return 0.f;
}

View File

@@ -1,42 +1,42 @@
/*
Copyright (C) 2010 Sony Computer Entertainment Inc.
All rights reserved.
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 __BT_PARALLEL_CONSTRAINT_SOLVER_H
#define __BT_PARALLEL_CONSTRAINT_SOLVER_H
#include "BulletDynamics/ConstraintSolver/btSequentialImpulseConstraintSolver.h"
class btParallelConstraintSolver : public btSequentialImpulseConstraintSolver
{
protected:
public:
btParallelConstraintSolver();
virtual ~btParallelConstraintSolver();
//virtual btScalar solveGroup(btCollisionObject** bodies,int numBodies,btPersistentManifold** manifold,int numManifolds,btTypedConstraint** constraints,int numConstraints,const btContactSolverInfo& info, btIDebugDraw* debugDrawer, btStackAlloc* stackAlloc,btDispatcher* dispatcher);
btScalar solveGroupCacheFriendlySetup(btCollisionObject** bodies,int numBodies,btPersistentManifold** manifoldPtr, int numManifolds,btTypedConstraint** constraints,int numConstraints,const btContactSolverInfo& infoGlobal,btIDebugDraw* debugDrawer,btStackAlloc* stackAlloc);
};
/*
Copyright (C) 2010 Sony Computer Entertainment Inc.
All rights reserved.
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 __BT_PARALLEL_CONSTRAINT_SOLVER_H
#define __BT_PARALLEL_CONSTRAINT_SOLVER_H
#include "BulletDynamics/ConstraintSolver/btSequentialImpulseConstraintSolver.h"
class btParallelConstraintSolver : public btSequentialImpulseConstraintSolver
{
protected:
public:
btParallelConstraintSolver();
virtual ~btParallelConstraintSolver();
//virtual btScalar solveGroup(btCollisionObject** bodies,int numBodies,btPersistentManifold** manifold,int numManifolds,btTypedConstraint** constraints,int numConstraints,const btContactSolverInfo& info, btIDebugDraw* debugDrawer, btStackAlloc* stackAlloc,btDispatcher* dispatcher);
btScalar solveGroupCacheFriendlySetup(btCollisionObject** bodies,int numBodies,btPersistentManifold** manifoldPtr, int numManifolds,btTypedConstraint** constraints,int numConstraints,const btContactSolverInfo& infoGlobal,btIDebugDraw* debugDrawer,btStackAlloc* stackAlloc);
};
#endif //__BT_PARALLEL_CONSTRAINT_SOLVER_H

View File

@@ -1,225 +1,225 @@
/*
Copyright (C) 2009 Sony Computer Entertainment Inc.
All rights reserved.
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 _BOOLINVEC_H
#define _BOOLINVEC_H
#include <math.h>
namespace Vectormath {
class floatInVec;
//--------------------------------------------------------------------------------------------------
// boolInVec class
//
class boolInVec
{
private:
unsigned int mData;
public:
// Default constructor; does no initialization
//
inline boolInVec( ) { };
// Construct from a value converted from float
//
inline boolInVec(floatInVec vec);
// Explicit cast from bool
//
explicit inline boolInVec(bool scalar);
// Explicit cast to bool
//
inline bool getAsBool() const;
#ifndef _VECTORMATH_NO_SCALAR_CAST
// Implicit cast to bool
//
inline operator bool() const;
#endif
// Boolean negation operator
//
inline const boolInVec operator ! () const;
// Assignment operator
//
inline boolInVec& operator = (boolInVec vec);
// Boolean and assignment operator
//
inline boolInVec& operator &= (boolInVec vec);
// Boolean exclusive or assignment operator
//
inline boolInVec& operator ^= (boolInVec vec);
// Boolean or assignment operator
//
inline boolInVec& operator |= (boolInVec vec);
};
// Equal operator
//
inline const boolInVec operator == (boolInVec vec0, boolInVec vec1);
// Not equal operator
//
inline const boolInVec operator != (boolInVec vec0, boolInVec vec1);
// And operator
//
inline const boolInVec operator & (boolInVec vec0, boolInVec vec1);
// Exclusive or operator
//
inline const boolInVec operator ^ (boolInVec vec0, boolInVec vec1);
// Or operator
//
inline const boolInVec operator | (boolInVec vec0, boolInVec vec1);
// Conditionally select between two values
//
inline const boolInVec select(boolInVec vec0, boolInVec vec1, boolInVec select_vec1);
} // namespace Vectormath
//--------------------------------------------------------------------------------------------------
// boolInVec implementation
//
#include "floatInVec.h"
namespace Vectormath {
inline
boolInVec::boolInVec(floatInVec vec)
{
*this = (vec != floatInVec(0.0f));
}
inline
boolInVec::boolInVec(bool scalar)
{
mData = -(int)scalar;
}
inline
bool
boolInVec::getAsBool() const
{
return (mData > 0);
}
#ifndef _VECTORMATH_NO_SCALAR_CAST
inline
boolInVec::operator bool() const
{
return getAsBool();
}
#endif
inline
const boolInVec
boolInVec::operator ! () const
{
return boolInVec(!mData);
}
inline
boolInVec&
boolInVec::operator = (boolInVec vec)
{
mData = vec.mData;
return *this;
}
inline
boolInVec&
boolInVec::operator &= (boolInVec vec)
{
*this = *this & vec;
return *this;
}
inline
boolInVec&
boolInVec::operator ^= (boolInVec vec)
{
*this = *this ^ vec;
return *this;
}
inline
boolInVec&
boolInVec::operator |= (boolInVec vec)
{
*this = *this | vec;
return *this;
}
inline
const boolInVec
operator == (boolInVec vec0, boolInVec vec1)
{
return boolInVec(vec0.getAsBool() == vec1.getAsBool());
}
inline
const boolInVec
operator != (boolInVec vec0, boolInVec vec1)
{
return !(vec0 == vec1);
}
inline
const boolInVec
operator & (boolInVec vec0, boolInVec vec1)
{
return boolInVec(vec0.getAsBool() & vec1.getAsBool());
}
inline
const boolInVec
operator | (boolInVec vec0, boolInVec vec1)
{
return boolInVec(vec0.getAsBool() | vec1.getAsBool());
}
inline
const boolInVec
operator ^ (boolInVec vec0, boolInVec vec1)
{
return boolInVec(vec0.getAsBool() ^ vec1.getAsBool());
}
inline
const boolInVec
select(boolInVec vec0, boolInVec vec1, boolInVec select_vec1)
{
return (select_vec1.getAsBool() == 0) ? vec0 : vec1;
}
} // namespace Vectormath
#endif // boolInVec_h
/*
Copyright (C) 2009 Sony Computer Entertainment Inc.
All rights reserved.
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 _BOOLINVEC_H
#define _BOOLINVEC_H
#include <math.h>
namespace Vectormath {
class floatInVec;
//--------------------------------------------------------------------------------------------------
// boolInVec class
//
class boolInVec
{
private:
unsigned int mData;
public:
// Default constructor; does no initialization
//
inline boolInVec( ) { };
// Construct from a value converted from float
//
inline boolInVec(floatInVec vec);
// Explicit cast from bool
//
explicit inline boolInVec(bool scalar);
// Explicit cast to bool
//
inline bool getAsBool() const;
#ifndef _VECTORMATH_NO_SCALAR_CAST
// Implicit cast to bool
//
inline operator bool() const;
#endif
// Boolean negation operator
//
inline const boolInVec operator ! () const;
// Assignment operator
//
inline boolInVec& operator = (boolInVec vec);
// Boolean and assignment operator
//
inline boolInVec& operator &= (boolInVec vec);
// Boolean exclusive or assignment operator
//
inline boolInVec& operator ^= (boolInVec vec);
// Boolean or assignment operator
//
inline boolInVec& operator |= (boolInVec vec);
};
// Equal operator
//
inline const boolInVec operator == (boolInVec vec0, boolInVec vec1);
// Not equal operator
//
inline const boolInVec operator != (boolInVec vec0, boolInVec vec1);
// And operator
//
inline const boolInVec operator & (boolInVec vec0, boolInVec vec1);
// Exclusive or operator
//
inline const boolInVec operator ^ (boolInVec vec0, boolInVec vec1);
// Or operator
//
inline const boolInVec operator | (boolInVec vec0, boolInVec vec1);
// Conditionally select between two values
//
inline const boolInVec select(boolInVec vec0, boolInVec vec1, boolInVec select_vec1);
} // namespace Vectormath
//--------------------------------------------------------------------------------------------------
// boolInVec implementation
//
#include "floatInVec.h"
namespace Vectormath {
inline
boolInVec::boolInVec(floatInVec vec)
{
*this = (vec != floatInVec(0.0f));
}
inline
boolInVec::boolInVec(bool scalar)
{
mData = -(int)scalar;
}
inline
bool
boolInVec::getAsBool() const
{
return (mData > 0);
}
#ifndef _VECTORMATH_NO_SCALAR_CAST
inline
boolInVec::operator bool() const
{
return getAsBool();
}
#endif
inline
const boolInVec
boolInVec::operator ! () const
{
return boolInVec(!mData);
}
inline
boolInVec&
boolInVec::operator = (boolInVec vec)
{
mData = vec.mData;
return *this;
}
inline
boolInVec&
boolInVec::operator &= (boolInVec vec)
{
*this = *this & vec;
return *this;
}
inline
boolInVec&
boolInVec::operator ^= (boolInVec vec)
{
*this = *this ^ vec;
return *this;
}
inline
boolInVec&
boolInVec::operator |= (boolInVec vec)
{
*this = *this | vec;
return *this;
}
inline
const boolInVec
operator == (boolInVec vec0, boolInVec vec1)
{
return boolInVec(vec0.getAsBool() == vec1.getAsBool());
}
inline
const boolInVec
operator != (boolInVec vec0, boolInVec vec1)
{
return !(vec0 == vec1);
}
inline
const boolInVec
operator & (boolInVec vec0, boolInVec vec1)
{
return boolInVec(vec0.getAsBool() & vec1.getAsBool());
}
inline
const boolInVec
operator | (boolInVec vec0, boolInVec vec1)
{
return boolInVec(vec0.getAsBool() | vec1.getAsBool());
}
inline
const boolInVec
operator ^ (boolInVec vec0, boolInVec vec1)
{
return boolInVec(vec0.getAsBool() ^ vec1.getAsBool());
}
inline
const boolInVec
select(boolInVec vec0, boolInVec vec1, boolInVec select_vec1)
{
return (select_vec1.getAsBool() == 0) ? vec0 : vec1;
}
} // namespace Vectormath
#endif // boolInVec_h

View File

@@ -1,343 +1,343 @@
/*
Copyright (C) 2009 Sony Computer Entertainment Inc.
All rights reserved.
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 _FLOATINVEC_H
#define _FLOATINVEC_H
#include <math.h>
namespace Vectormath {
class boolInVec;
//--------------------------------------------------------------------------------------------------
// floatInVec class
//
// A class representing a scalar float value contained in a vector register
// This class does not support fastmath
class floatInVec
{
private:
float mData;
public:
// Default constructor; does no initialization
//
inline floatInVec( ) { };
// Construct from a value converted from bool
//
inline floatInVec(boolInVec vec);
// Explicit cast from float
//
explicit inline floatInVec(float scalar);
// Explicit cast to float
//
inline float getAsFloat() const;
#ifndef _VECTORMATH_NO_SCALAR_CAST
// Implicit cast to float
//
inline operator float() const;
#endif
// Post increment (add 1.0f)
//
inline const floatInVec operator ++ (int);
// Post decrement (subtract 1.0f)
//
inline const floatInVec operator -- (int);
// Pre increment (add 1.0f)
//
inline floatInVec& operator ++ ();
// Pre decrement (subtract 1.0f)
//
inline floatInVec& operator -- ();
// Negation operator
//
inline const floatInVec operator - () const;
// Assignment operator
//
inline floatInVec& operator = (floatInVec vec);
// Multiplication assignment operator
//
inline floatInVec& operator *= (floatInVec vec);
// Division assignment operator
//
inline floatInVec& operator /= (floatInVec vec);
// Addition assignment operator
//
inline floatInVec& operator += (floatInVec vec);
// Subtraction assignment operator
//
inline floatInVec& operator -= (floatInVec vec);
};
// Multiplication operator
//
inline const floatInVec operator * (floatInVec vec0, floatInVec vec1);
// Division operator
//
inline const floatInVec operator / (floatInVec vec0, floatInVec vec1);
// Addition operator
//
inline const floatInVec operator + (floatInVec vec0, floatInVec vec1);
// Subtraction operator
//
inline const floatInVec operator - (floatInVec vec0, floatInVec vec1);
// Less than operator
//
inline const boolInVec operator < (floatInVec vec0, floatInVec vec1);
// Less than or equal operator
//
inline const boolInVec operator <= (floatInVec vec0, floatInVec vec1);
// Greater than operator
//
inline const boolInVec operator > (floatInVec vec0, floatInVec vec1);
// Greater than or equal operator
//
inline const boolInVec operator >= (floatInVec vec0, floatInVec vec1);
// Equal operator
//
inline const boolInVec operator == (floatInVec vec0, floatInVec vec1);
// Not equal operator
//
inline const boolInVec operator != (floatInVec vec0, floatInVec vec1);
// Conditionally select between two values
//
inline const floatInVec select(floatInVec vec0, floatInVec vec1, boolInVec select_vec1);
} // namespace Vectormath
//--------------------------------------------------------------------------------------------------
// floatInVec implementation
//
#include "boolInVec.h"
namespace Vectormath {
inline
floatInVec::floatInVec(boolInVec vec)
{
mData = float(vec.getAsBool());
}
inline
floatInVec::floatInVec(float scalar)
{
mData = scalar;
}
inline
float
floatInVec::getAsFloat() const
{
return mData;
}
#ifndef _VECTORMATH_NO_SCALAR_CAST
inline
floatInVec::operator float() const
{
return getAsFloat();
}
#endif
inline
const floatInVec
floatInVec::operator ++ (int)
{
float olddata = mData;
operator ++();
return floatInVec(olddata);
}
inline
const floatInVec
floatInVec::operator -- (int)
{
float olddata = mData;
operator --();
return floatInVec(olddata);
}
inline
floatInVec&
floatInVec::operator ++ ()
{
*this += floatInVec(1.0f);
return *this;
}
inline
floatInVec&
floatInVec::operator -- ()
{
*this -= floatInVec(1.0f);
return *this;
}
inline
const floatInVec
floatInVec::operator - () const
{
return floatInVec(-mData);
}
inline
floatInVec&
floatInVec::operator = (floatInVec vec)
{
mData = vec.mData;
return *this;
}
inline
floatInVec&
floatInVec::operator *= (floatInVec vec)
{
*this = *this * vec;
return *this;
}
inline
floatInVec&
floatInVec::operator /= (floatInVec vec)
{
*this = *this / vec;
return *this;
}
inline
floatInVec&
floatInVec::operator += (floatInVec vec)
{
*this = *this + vec;
return *this;
}
inline
floatInVec&
floatInVec::operator -= (floatInVec vec)
{
*this = *this - vec;
return *this;
}
inline
const floatInVec
operator * (floatInVec vec0, floatInVec vec1)
{
return floatInVec(vec0.getAsFloat() * vec1.getAsFloat());
}
inline
const floatInVec
operator / (floatInVec num, floatInVec den)
{
return floatInVec(num.getAsFloat() / den.getAsFloat());
}
inline
const floatInVec
operator + (floatInVec vec0, floatInVec vec1)
{
return floatInVec(vec0.getAsFloat() + vec1.getAsFloat());
}
inline
const floatInVec
operator - (floatInVec vec0, floatInVec vec1)
{
return floatInVec(vec0.getAsFloat() - vec1.getAsFloat());
}
inline
const boolInVec
operator < (floatInVec vec0, floatInVec vec1)
{
return boolInVec(vec0.getAsFloat() < vec1.getAsFloat());
}
inline
const boolInVec
operator <= (floatInVec vec0, floatInVec vec1)
{
return !(vec0 > vec1);
}
inline
const boolInVec
operator > (floatInVec vec0, floatInVec vec1)
{
return boolInVec(vec0.getAsFloat() > vec1.getAsFloat());
}
inline
const boolInVec
operator >= (floatInVec vec0, floatInVec vec1)
{
return !(vec0 < vec1);
}
inline
const boolInVec
operator == (floatInVec vec0, floatInVec vec1)
{
return boolInVec(vec0.getAsFloat() == vec1.getAsFloat());
}
inline
const boolInVec
operator != (floatInVec vec0, floatInVec vec1)
{
return !(vec0 == vec1);
}
inline
const floatInVec
select(floatInVec vec0, floatInVec vec1, boolInVec select_vec1)
{
return (select_vec1.getAsBool() == 0) ? vec0 : vec1;
}
} // namespace Vectormath
#endif // floatInVec_h
/*
Copyright (C) 2009 Sony Computer Entertainment Inc.
All rights reserved.
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 _FLOATINVEC_H
#define _FLOATINVEC_H
#include <math.h>
namespace Vectormath {
class boolInVec;
//--------------------------------------------------------------------------------------------------
// floatInVec class
//
// A class representing a scalar float value contained in a vector register
// This class does not support fastmath
class floatInVec
{
private:
float mData;
public:
// Default constructor; does no initialization
//
inline floatInVec( ) { };
// Construct from a value converted from bool
//
inline floatInVec(boolInVec vec);
// Explicit cast from float
//
explicit inline floatInVec(float scalar);
// Explicit cast to float
//
inline float getAsFloat() const;
#ifndef _VECTORMATH_NO_SCALAR_CAST
// Implicit cast to float
//
inline operator float() const;
#endif
// Post increment (add 1.0f)
//
inline const floatInVec operator ++ (int);
// Post decrement (subtract 1.0f)
//
inline const floatInVec operator -- (int);
// Pre increment (add 1.0f)
//
inline floatInVec& operator ++ ();
// Pre decrement (subtract 1.0f)
//
inline floatInVec& operator -- ();
// Negation operator
//
inline const floatInVec operator - () const;
// Assignment operator
//
inline floatInVec& operator = (floatInVec vec);
// Multiplication assignment operator
//
inline floatInVec& operator *= (floatInVec vec);
// Division assignment operator
//
inline floatInVec& operator /= (floatInVec vec);
// Addition assignment operator
//
inline floatInVec& operator += (floatInVec vec);
// Subtraction assignment operator
//
inline floatInVec& operator -= (floatInVec vec);
};
// Multiplication operator
//
inline const floatInVec operator * (floatInVec vec0, floatInVec vec1);
// Division operator
//
inline const floatInVec operator / (floatInVec vec0, floatInVec vec1);
// Addition operator
//
inline const floatInVec operator + (floatInVec vec0, floatInVec vec1);
// Subtraction operator
//
inline const floatInVec operator - (floatInVec vec0, floatInVec vec1);
// Less than operator
//
inline const boolInVec operator < (floatInVec vec0, floatInVec vec1);
// Less than or equal operator
//
inline const boolInVec operator <= (floatInVec vec0, floatInVec vec1);
// Greater than operator
//
inline const boolInVec operator > (floatInVec vec0, floatInVec vec1);
// Greater than or equal operator
//
inline const boolInVec operator >= (floatInVec vec0, floatInVec vec1);
// Equal operator
//
inline const boolInVec operator == (floatInVec vec0, floatInVec vec1);
// Not equal operator
//
inline const boolInVec operator != (floatInVec vec0, floatInVec vec1);
// Conditionally select between two values
//
inline const floatInVec select(floatInVec vec0, floatInVec vec1, boolInVec select_vec1);
} // namespace Vectormath
//--------------------------------------------------------------------------------------------------
// floatInVec implementation
//
#include "boolInVec.h"
namespace Vectormath {
inline
floatInVec::floatInVec(boolInVec vec)
{
mData = float(vec.getAsBool());
}
inline
floatInVec::floatInVec(float scalar)
{
mData = scalar;
}
inline
float
floatInVec::getAsFloat() const
{
return mData;
}
#ifndef _VECTORMATH_NO_SCALAR_CAST
inline
floatInVec::operator float() const
{
return getAsFloat();
}
#endif
inline
const floatInVec
floatInVec::operator ++ (int)
{
float olddata = mData;
operator ++();
return floatInVec(olddata);
}
inline
const floatInVec
floatInVec::operator -- (int)
{
float olddata = mData;
operator --();
return floatInVec(olddata);
}
inline
floatInVec&
floatInVec::operator ++ ()
{
*this += floatInVec(1.0f);
return *this;
}
inline
floatInVec&
floatInVec::operator -- ()
{
*this -= floatInVec(1.0f);
return *this;
}
inline
const floatInVec
floatInVec::operator - () const
{
return floatInVec(-mData);
}
inline
floatInVec&
floatInVec::operator = (floatInVec vec)
{
mData = vec.mData;
return *this;
}
inline
floatInVec&
floatInVec::operator *= (floatInVec vec)
{
*this = *this * vec;
return *this;
}
inline
floatInVec&
floatInVec::operator /= (floatInVec vec)
{
*this = *this / vec;
return *this;
}
inline
floatInVec&
floatInVec::operator += (floatInVec vec)
{
*this = *this + vec;
return *this;
}
inline
floatInVec&
floatInVec::operator -= (floatInVec vec)
{
*this = *this - vec;
return *this;
}
inline
const floatInVec
operator * (floatInVec vec0, floatInVec vec1)
{
return floatInVec(vec0.getAsFloat() * vec1.getAsFloat());
}
inline
const floatInVec
operator / (floatInVec num, floatInVec den)
{
return floatInVec(num.getAsFloat() / den.getAsFloat());
}
inline
const floatInVec
operator + (floatInVec vec0, floatInVec vec1)
{
return floatInVec(vec0.getAsFloat() + vec1.getAsFloat());
}
inline
const floatInVec
operator - (floatInVec vec0, floatInVec vec1)
{
return floatInVec(vec0.getAsFloat() - vec1.getAsFloat());
}
inline
const boolInVec
operator < (floatInVec vec0, floatInVec vec1)
{
return boolInVec(vec0.getAsFloat() < vec1.getAsFloat());
}
inline
const boolInVec
operator <= (floatInVec vec0, floatInVec vec1)
{
return !(vec0 > vec1);
}
inline
const boolInVec
operator > (floatInVec vec0, floatInVec vec1)
{
return boolInVec(vec0.getAsFloat() > vec1.getAsFloat());
}
inline
const boolInVec
operator >= (floatInVec vec0, floatInVec vec1)
{
return !(vec0 < vec1);
}
inline
const boolInVec
operator == (floatInVec vec0, floatInVec vec1)
{
return boolInVec(vec0.getAsFloat() == vec1.getAsFloat());
}
inline
const boolInVec
operator != (floatInVec vec0, floatInVec vec1)
{
return !(vec0 == vec1);
}
inline
const floatInVec
select(floatInVec vec0, floatInVec vec1, boolInVec select_vec1)
{
return (select_vec1.getAsBool() == 0) ? vec0 : vec1;
}
} // namespace Vectormath
#endif // floatInVec_h