TorqueGameEngines/Torque3D
1
0
Fork 0
mirror of https://github.com/TorqueGameEngines/Torque3D.git synced 2026-07-13 23:54:35 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

Bullet Library v2.81

Browse source
This commit is contained in:
LuisAntonRebollo 2013-07-04 20:50:16 +02:00
parent 64fef8b2ad
commit 1eb94f4828
462 changed files with 59613 additions and 8036 deletions
Download patch file Download diff file Expand all files Collapse all files

66
Engine/lib/bullet/src/MiniCL/CMakeLists.txt Normal file
View file

@ -0,0 +1,66 @@
#MiniCL provides a small subset of OpenCL
INCLUDE_DIRECTORIES(
${BULLET_PHYSICS_SOURCE_DIR}/src
${VECTOR_MATH_INCLUDE}
)
SET(MiniCL_SRCS
MiniCL.cpp
MiniCLTaskScheduler.cpp
MiniCLTask/MiniCLTask.cpp
)
SET(Root_HDRS
MiniCLTaskScheduler.h
cl.h
cl_gl.h
cl_platform.h
cl_MiniCL_Defs.h
)
SET(MiniCLTask_HDRS
MiniCLTask/MiniCLTask.h
)
SET(MiniCL_HDRS
${Root_HDRS}
${MiniCLTask_HDRS}
)
ADD_LIBRARY(MiniCL ${MiniCL_SRCS} ${MiniCL_HDRS} )
SET_TARGET_PROPERTIES(MiniCL PROPERTIES VERSION ${BULLET_VERSION})
SET_TARGET_PROPERTIES(MiniCL PROPERTIES SOVERSION ${BULLET_VERSION})
IF (BUILD_SHARED_LIBS)
TARGET_LINK_LIBRARIES(MiniCL BulletMultiThreaded BulletDynamics BulletCollision)
ENDIF (BUILD_SHARED_LIBS)
IF (INSTALL_LIBS)
IF (NOT INTERNAL_CREATE_DISTRIBUTABLE_MSVC_PROJECTFILES)
#INSTALL of other files requires CMake 2.6
IF (${CMAKE_MAJOR_VERSION}.${CMAKE_MINOR_VERSION} GREATER 2.5)
# IF(INSTALL_EXTRA_LIBS)
IF (APPLE AND BUILD_SHARED_LIBS AND FRAMEWORK)
INSTALL(TARGETS MiniCL DESTINATION .)
ELSE (APPLE AND BUILD_SHARED_LIBS AND FRAMEWORK)
INSTALL(TARGETS MiniCL DESTINATION lib${LIB_SUFFIX})
INSTALL(DIRECTORY
${CMAKE_CURRENT_SOURCE_DIR} DESTINATION ${INCLUDE_INSTALL_DIR} FILES_MATCHING
PATTERN "*.h" PATTERN ".svn" EXCLUDE PATTERN "CMakeFiles" EXCLUDE)
ENDIF (APPLE AND BUILD_SHARED_LIBS AND FRAMEWORK)
# ENDIF (INSTALL_EXTRA_LIBS)
ENDIF (${CMAKE_MAJOR_VERSION}.${CMAKE_MINOR_VERSION} GREATER 2.5)
IF (APPLE AND BUILD_SHARED_LIBS AND FRAMEWORK)
SET_TARGET_PROPERTIES(MiniCL PROPERTIES FRAMEWORK true)
SET_TARGET_PROPERTIES(MiniCL PROPERTIES PUBLIC_HEADER "${Root_HDRS}")
# Have to list out sub-directories manually:
SET_PROPERTY(SOURCE ${MiniCLTask_HDRS} PROPERTY MACOSX_PACKAGE_LOCATION Headers/MiniCLTask)
ENDIF (APPLE AND BUILD_SHARED_LIBS AND FRAMEWORK)
ENDIF (NOT INTERNAL_CREATE_DISTRIBUTABLE_MSVC_PROJECTFILES)
ENDIF (INSTALL_LIBS)

788
Engine/lib/bullet/src/MiniCL/MiniCL.cpp Normal file
View file

@ -0,0 +1,788 @@
/*
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 "MiniCL/cl.h"
#define __PHYSICS_COMMON_H__ 1
#ifdef _WIN32
#include "BulletMultiThreaded/Win32ThreadSupport.h"
#endif
#include "BulletMultiThreaded/PlatformDefinitions.h"
#ifdef USE_PTHREADS
#include "BulletMultiThreaded/PosixThreadSupport.h"
#endif
#include "BulletMultiThreaded/SequentialThreadSupport.h"
#include "MiniCLTaskScheduler.h"
#include "MiniCLTask/MiniCLTask.h"
#include "LinearMath/btMinMax.h"
#include <stdio.h>
#include <stddef.h>
//#define DEBUG_MINICL_KERNELS 1
static const char* spPlatformID = "MiniCL, SCEA";
static const char* spDriverVersion= "1.0";
CL_API_ENTRY cl_int CL_API_CALL clGetPlatformIDs(
cl_uint num_entries,
cl_platform_id * platforms,
cl_uint * num_platforms ) CL_API_SUFFIX__VERSION_1_0
{
if(platforms != NULL)
{
if(num_entries <= 0)
{
return CL_INVALID_VALUE;
}
*((const char**)platforms) = spPlatformID;
}
if(num_platforms != NULL)
{
*num_platforms = 1;
}
return CL_SUCCESS;
}
CL_API_ENTRY cl_int CL_API_CALL clGetPlatformInfo(
cl_platform_id platform,
cl_platform_info param_name,
size_t param_value_size,
void * param_value,
size_t * param_value_size_ret) CL_API_SUFFIX__VERSION_1_0
{
char* pId = (char*)platform;
if(strcmp(pId, spPlatformID))
{
return CL_INVALID_PLATFORM;
}
switch(param_name)
{
case CL_PLATFORM_VERSION:
{
if(param_value_size < (strlen(spDriverVersion) + 1))
{
return CL_INVALID_VALUE;
}
strcpy((char*)param_value, spDriverVersion);
if(param_value_size_ret != NULL)
{
*param_value_size_ret = strlen(spDriverVersion) + 1;
}
break;
}
case CL_PLATFORM_NAME:
case CL_PLATFORM_VENDOR :
if(param_value_size < (strlen(spPlatformID) + 1))
{
return CL_INVALID_VALUE;
}
strcpy((char*)param_value, spPlatformID);
if(param_value_size_ret != NULL)
{
*param_value_size_ret = strlen(spPlatformID) + 1;
}
break;
default :
return CL_INVALID_VALUE;
}
return CL_SUCCESS;
}
CL_API_ENTRY cl_int CL_API_CALL clGetDeviceInfo(
cl_device_id device ,
cl_device_info param_name ,
size_t param_value_size ,
void * param_value ,
size_t * param_value_size_ret) CL_API_SUFFIX__VERSION_1_0
{
switch (param_name)
{
case CL_DEVICE_NAME:
{
char deviceName[] = "MiniCL CPU";
unsigned int nameLen = (unsigned int)strlen(deviceName)+1;
btAssert(param_value_size>strlen(deviceName));
if (nameLen < param_value_size)
{
const char* cpuName = "MiniCL CPU";
sprintf((char*)param_value,"%s",cpuName);
} else
{
printf("error: param_value_size should be at least %d, but it is %zu\n",nameLen,param_value_size);
return CL_INVALID_VALUE;
}
break;
}
case CL_DEVICE_TYPE:
{
if (param_value_size>=sizeof(cl_device_type))
{
cl_device_type* deviceType = (cl_device_type*)param_value;
*deviceType = CL_DEVICE_TYPE_CPU;
} else
{
printf("error: param_value_size should be at least %zu\n",sizeof(cl_device_type));
return CL_INVALID_VALUE;
}
break;
}
case CL_DEVICE_MAX_COMPUTE_UNITS:
{
if (param_value_size>=sizeof(cl_uint))
{
cl_uint* numUnits = (cl_uint*)param_value;
*numUnits= 4;
} else
{
printf("error: param_value_size should be at least %zu\n",sizeof(cl_uint));
return CL_INVALID_VALUE;
}
break;
}
case CL_DEVICE_MAX_WORK_ITEM_SIZES:
{
size_t workitem_size[3];
if (param_value_size>=sizeof(workitem_size))
{
size_t* workItemSize = (size_t*)param_value;
workItemSize[0] = 64;
workItemSize[1] = 24;
workItemSize[2] = 16;
} else
{
printf("error: param_value_size should be at least %zu\n",sizeof(cl_uint));
return CL_INVALID_VALUE;
}
break;
}
case CL_DEVICE_MAX_CLOCK_FREQUENCY:
{
cl_uint* clock_frequency = (cl_uint*)param_value;
*clock_frequency = 3*1024;
break;
}
case CL_DEVICE_VENDOR :
{
if(param_value_size < (strlen(spPlatformID) + 1))
{
return CL_INVALID_VALUE;
}
strcpy((char*)param_value, spPlatformID);
if(param_value_size_ret != NULL)
{
*param_value_size_ret = strlen(spPlatformID) + 1;
}
break;
}
case CL_DRIVER_VERSION:
{
if(param_value_size < (strlen(spDriverVersion) + 1))
{
return CL_INVALID_VALUE;
}
strcpy((char*)param_value, spDriverVersion);
if(param_value_size_ret != NULL)
{
*param_value_size_ret = strlen(spDriverVersion) + 1;
}
break;
}
case CL_DEVICE_MAX_WORK_ITEM_DIMENSIONS:
{
cl_uint* maxDimensions = (cl_uint*)param_value;
*maxDimensions = 1;
break;
}
case CL_DEVICE_MAX_WORK_GROUP_SIZE:
{
cl_uint* maxWorkGroupSize = (cl_uint*)param_value;
*maxWorkGroupSize = 128;//1;
break;
}
case CL_DEVICE_ADDRESS_BITS:
{
cl_uint* addressBits = (cl_uint*)param_value;
*addressBits= 32; //@todo: should this be 64 for 64bit builds?
break;
}
case CL_DEVICE_MAX_MEM_ALLOC_SIZE:
{
cl_ulong* maxMemAlloc = (cl_ulong*)param_value;
*maxMemAlloc= 512*1024*1024; //this "should be enough for everyone" ?
break;
}
case CL_DEVICE_GLOBAL_MEM_SIZE:
{
cl_ulong* maxMemAlloc = (cl_ulong*)param_value;
*maxMemAlloc= 1024*1024*1024; //this "should be enough for everyone" ?
break;
}
case CL_DEVICE_ERROR_CORRECTION_SUPPORT:
{
cl_bool* error_correction_support = (cl_bool*)param_value;
*error_correction_support = CL_FALSE;
break;
}
case CL_DEVICE_LOCAL_MEM_TYPE:
{
cl_device_local_mem_type* local_mem_type = (cl_device_local_mem_type*)param_value;
*local_mem_type = CL_GLOBAL;
break;
}
case CL_DEVICE_LOCAL_MEM_SIZE:
{
cl_ulong* localmem = (cl_ulong*) param_value;
*localmem = 32*1024;
break;
}
case CL_DEVICE_MAX_CONSTANT_BUFFER_SIZE:
{
cl_ulong* localmem = (cl_ulong*) param_value;
*localmem = 64*1024;
break;
}
case CL_DEVICE_QUEUE_PROPERTIES:
{
cl_command_queue_properties* queueProp = (cl_command_queue_properties*) param_value;
memset(queueProp,0,param_value_size);
break;
}
case CL_DEVICE_IMAGE_SUPPORT:
{
cl_bool* imageSupport = (cl_bool*) param_value;
*imageSupport = CL_FALSE;
break;
}
case CL_DEVICE_MAX_WRITE_IMAGE_ARGS:
case CL_DEVICE_MAX_READ_IMAGE_ARGS:
{
cl_uint* imageArgs = (cl_uint*) param_value;
*imageArgs = 0;
break;
}
case CL_DEVICE_IMAGE3D_MAX_DEPTH:
case CL_DEVICE_IMAGE3D_MAX_HEIGHT:
case CL_DEVICE_IMAGE2D_MAX_HEIGHT:
case CL_DEVICE_IMAGE3D_MAX_WIDTH:
case CL_DEVICE_IMAGE2D_MAX_WIDTH:
{
size_t* maxSize = (size_t*) param_value;
*maxSize = 0;
break;
}
case CL_DEVICE_EXTENSIONS:
{
char* extensions = (char*) param_value;
*extensions = 0;
break;
}
case CL_DEVICE_PREFERRED_VECTOR_WIDTH_DOUBLE:
case CL_DEVICE_PREFERRED_VECTOR_WIDTH_FLOAT:
case CL_DEVICE_PREFERRED_VECTOR_WIDTH_LONG:
case CL_DEVICE_PREFERRED_VECTOR_WIDTH_INT:
case CL_DEVICE_PREFERRED_VECTOR_WIDTH_SHORT:
case CL_DEVICE_PREFERRED_VECTOR_WIDTH_CHAR:
{
cl_uint* width = (cl_uint*) param_value;
*width = 1;
break;
}
default:
{
printf("error: unsupported param_name:%d\n",param_name);
}
}
return 0;
}
CL_API_ENTRY cl_int CL_API_CALL clReleaseMemObject(cl_mem /* memobj */) CL_API_SUFFIX__VERSION_1_0
{
return 0;
}
CL_API_ENTRY cl_int CL_API_CALL clReleaseCommandQueue(cl_command_queue /* command_queue */) CL_API_SUFFIX__VERSION_1_0
{
return 0;
}
CL_API_ENTRY cl_int CL_API_CALL clReleaseProgram(cl_program /* program */) CL_API_SUFFIX__VERSION_1_0
{
return 0;
}
CL_API_ENTRY cl_int CL_API_CALL clReleaseKernel(cl_kernel /* kernel */) CL_API_SUFFIX__VERSION_1_0
{
return 0;
}
// Enqueued Commands APIs
CL_API_ENTRY cl_int CL_API_CALL clEnqueueReadBuffer(cl_command_queue command_queue ,
cl_mem buffer ,
cl_bool /* blocking_read */,
size_t offset ,
size_t cb ,
void * ptr ,
cl_uint /* num_events_in_wait_list */,
const cl_event * /* event_wait_list */,
cl_event * /* event */) CL_API_SUFFIX__VERSION_1_0
{
MiniCLTaskScheduler* scheduler = (MiniCLTaskScheduler*) command_queue;
///wait for all work items to be completed
scheduler->flush();
memcpy(ptr,(char*)buffer + offset,cb);
return 0;
}
CL_API_ENTRY cl_int clGetProgramBuildInfo(cl_program /* program */,
cl_device_id /* device */,
cl_program_build_info /* param_name */,
size_t /* param_value_size */,
void * /* param_value */,
size_t * /* param_value_size_ret */) CL_API_SUFFIX__VERSION_1_0
{
return 0;
}
// Program Object APIs
CL_API_ENTRY cl_program
clCreateProgramWithSource(cl_context context ,
cl_uint /* count */,
const char ** /* strings */,
const size_t * /* lengths */,
cl_int * errcode_ret ) CL_API_SUFFIX__VERSION_1_0
{
*errcode_ret = CL_SUCCESS;
return (cl_program)context;
}
CL_API_ENTRY cl_int CL_API_CALL clEnqueueWriteBuffer(cl_command_queue command_queue ,
cl_mem buffer ,
cl_bool /* blocking_read */,
size_t offset,
size_t cb ,
const void * ptr ,
cl_uint /* num_events_in_wait_list */,
const cl_event * /* event_wait_list */,
cl_event * /* event */) CL_API_SUFFIX__VERSION_1_0
{
MiniCLTaskScheduler* scheduler = (MiniCLTaskScheduler*) command_queue;
///wait for all work items to be completed
scheduler->flush();
memcpy((char*)buffer + offset, ptr,cb);
return 0;
}
CL_API_ENTRY cl_int CL_API_CALL clFlush(cl_command_queue command_queue)
{
MiniCLTaskScheduler* scheduler = (MiniCLTaskScheduler*) command_queue;
///wait for all work items to be completed
scheduler->flush();
return 0;
}
CL_API_ENTRY cl_int CL_API_CALL clEnqueueNDRangeKernel(cl_command_queue /* command_queue */,
cl_kernel clKernel ,
cl_uint work_dim ,
const size_t * /* global_work_offset */,
const size_t * global_work_size ,
const size_t * /* local_work_size */,
cl_uint /* num_events_in_wait_list */,
const cl_event * /* event_wait_list */,
cl_event * /* event */) CL_API_SUFFIX__VERSION_1_0
{
MiniCLKernel* kernel = (MiniCLKernel*) clKernel;
for (unsigned int ii=0;ii<work_dim;ii++)
{
int maxTask = kernel->m_scheduler->getMaxNumOutstandingTasks();
int numWorkItems = global_work_size[ii];
// //at minimum 64 work items per task
// int numWorkItemsPerTask = btMax(64,numWorkItems / maxTask);
int numWorkItemsPerTask = numWorkItems / maxTask;
if (!numWorkItemsPerTask) numWorkItemsPerTask = 1;
for (int t=0;t<numWorkItems;)
{
//Performance Hint: tweak this number during benchmarking
int endIndex = (t+numWorkItemsPerTask) < numWorkItems ? t+numWorkItemsPerTask : numWorkItems;
kernel->m_scheduler->issueTask(t, endIndex, kernel);
t = endIndex;
}
}
/*
void* bla = 0;
scheduler->issueTask(bla,2,3);
scheduler->flush();
*/
return 0;
}
#define LOCAL_BUF_SIZE 32768
static int sLocalMemBuf[LOCAL_BUF_SIZE * 4 + 16];
static int* spLocalBufCurr = NULL;
static int sLocalBufUsed = LOCAL_BUF_SIZE; // so it will be reset at the first call
static void* localBufMalloc(int size)
{
int size16 = (size + 15) >> 4; // in 16-byte units
if((sLocalBufUsed + size16) > LOCAL_BUF_SIZE)
{ // reset
spLocalBufCurr = sLocalMemBuf;
while((size_t)spLocalBufCurr & 0x0F) spLocalBufCurr++; // align to 16 bytes
sLocalBufUsed = 0;
}
void* ret = spLocalBufCurr;
spLocalBufCurr += size16 * 4;
sLocalBufUsed += size;
return ret;
}
CL_API_ENTRY cl_int CL_API_CALL clSetKernelArg(cl_kernel clKernel ,
cl_uint arg_index ,
size_t arg_size ,
const void * arg_value ) CL_API_SUFFIX__VERSION_1_0
{
MiniCLKernel* kernel = (MiniCLKernel* ) clKernel;
btAssert(arg_size <= MINICL_MAX_ARGLENGTH);
if (arg_index>MINI_CL_MAX_ARG)
{
printf("error: clSetKernelArg arg_index (%u) exceeds %u\n",arg_index,MINI_CL_MAX_ARG);
} else
{
if (arg_size>MINICL_MAX_ARGLENGTH)
//if (arg_size != MINICL_MAX_ARGLENGTH)
{
printf("error: clSetKernelArg argdata too large: %zu (maximum is %zu)\n",arg_size,MINICL_MAX_ARGLENGTH);
}
else
{
if(arg_value == NULL)
{ // this is only for __local memory qualifier
void* ptr = localBufMalloc(arg_size);
kernel->m_argData[arg_index] = ptr;
}
else
{
memcpy(&(kernel->m_argData[arg_index]), arg_value, arg_size);
}
kernel->m_argSizes[arg_index] = arg_size;
if(arg_index >= kernel->m_numArgs)
{
kernel->m_numArgs = arg_index + 1;
kernel->updateLauncher();
}
}
}
return 0;
}
// Kernel Object APIs
CL_API_ENTRY cl_kernel CL_API_CALL clCreateKernel(cl_program program ,
const char * kernel_name ,
cl_int * errcode_ret ) CL_API_SUFFIX__VERSION_1_0
{
MiniCLTaskScheduler* scheduler = (MiniCLTaskScheduler*) program;
int nameLen = strlen(kernel_name);
if(nameLen >= MINI_CL_MAX_KERNEL_NAME)
{
*errcode_ret = CL_INVALID_KERNEL_NAME;
return NULL;
}
MiniCLKernel* kernel = new MiniCLKernel();
strcpy(kernel->m_name, kernel_name);
kernel->m_numArgs = 0;
//kernel->m_kernelProgramCommandId = scheduler->findProgramCommandIdByName(kernel_name);
//if (kernel->m_kernelProgramCommandId>=0)
//{
// *errcode_ret = CL_SUCCESS;
//} else
//{
// *errcode_ret = CL_INVALID_KERNEL_NAME;
//}
kernel->m_scheduler = scheduler;
if(kernel->registerSelf() == NULL)
{
*errcode_ret = CL_INVALID_KERNEL_NAME;
delete kernel;
return NULL;
}
else
{
*errcode_ret = CL_SUCCESS;
}
return (cl_kernel)kernel;
}
CL_API_ENTRY cl_int CL_API_CALL clBuildProgram(cl_program /* program */,
cl_uint /* num_devices */,
const cl_device_id * /* device_list */,
const char * /* options */,
void (*pfn_notify)(cl_program /* program */, void * /* user_data */),
void * /* user_data */) CL_API_SUFFIX__VERSION_1_0
{
return CL_SUCCESS;
}
CL_API_ENTRY cl_program CL_API_CALL clCreateProgramWithBinary(cl_context context ,
cl_uint /* num_devices */,
const cl_device_id * /* device_list */,
const size_t * /* lengths */,
const unsigned char ** /* binaries */,
cl_int * /* binary_status */,
cl_int * /* errcode_ret */) CL_API_SUFFIX__VERSION_1_0
{
return (cl_program)context;
}
// Memory Object APIs
CL_API_ENTRY cl_mem CL_API_CALL clCreateBuffer(cl_context /* context */,
cl_mem_flags flags ,
size_t size,
void * host_ptr ,
cl_int * errcode_ret ) CL_API_SUFFIX__VERSION_1_0
{
cl_mem buf = (cl_mem)malloc(size);
if ((flags&CL_MEM_COPY_HOST_PTR) && host_ptr)
{
memcpy(buf,host_ptr,size);
}
*errcode_ret = 0;
return buf;
}
// Command Queue APIs
CL_API_ENTRY cl_command_queue CL_API_CALL clCreateCommandQueue(cl_context context ,
cl_device_id /* device */,
cl_command_queue_properties /* properties */,
cl_int * errcode_ret ) CL_API_SUFFIX__VERSION_1_0
{
*errcode_ret = 0;
return (cl_command_queue) context;
}
extern CL_API_ENTRY cl_int CL_API_CALL clGetContextInfo(cl_context /* context */,
cl_context_info param_name ,
size_t param_value_size ,
void * param_value,
size_t * param_value_size_ret ) CL_API_SUFFIX__VERSION_1_0
{
switch (param_name)
{
case CL_CONTEXT_DEVICES:
{
if (!param_value_size)
{
*param_value_size_ret = 13;
} else
{
const char* testName = "MiniCL_Test.";
sprintf((char*)param_value,"%s",testName);
}
break;
};
default:
{
printf("unsupported\n");
}
}
return 0;
}
CL_API_ENTRY cl_context CL_API_CALL clCreateContextFromType(const cl_context_properties * /* properties */,
cl_device_type device_type ,
void (*pfn_notify)(const char *, const void *, size_t, void *) /* pfn_notify */,
void * /* user_data */,
cl_int * errcode_ret ) CL_API_SUFFIX__VERSION_1_0
{
int maxNumOutstandingTasks = 4;
// int maxNumOutstandingTasks = 2;
// int maxNumOutstandingTasks = 1;
gMiniCLNumOutstandingTasks = maxNumOutstandingTasks;
const int maxNumOfThreadSupports = 8;
static int sUniqueThreadSupportIndex = 0;
static const char* sUniqueThreadSupportName[maxNumOfThreadSupports] =
{
"MiniCL_0", "MiniCL_1", "MiniCL_2", "MiniCL_3", "MiniCL_4", "MiniCL_5", "MiniCL_6", "MiniCL_7"
};
btThreadSupportInterface* threadSupport = 0;
if (device_type==CL_DEVICE_TYPE_DEBUG)
{
SequentialThreadSupport::SequentialThreadConstructionInfo stc("MiniCL",processMiniCLTask,createMiniCLLocalStoreMemory);
threadSupport = new SequentialThreadSupport(stc);
} else
{
#if _WIN32
btAssert(sUniqueThreadSupportIndex < maxNumOfThreadSupports);
const char* bla = "MiniCL";
threadSupport = new Win32ThreadSupport(Win32ThreadSupport::Win32ThreadConstructionInfo(
// bla,
sUniqueThreadSupportName[sUniqueThreadSupportIndex++],
processMiniCLTask, //processCollisionTask,
createMiniCLLocalStoreMemory,//createCollisionLocalStoreMemory,
maxNumOutstandingTasks));
#else
#ifdef USE_PTHREADS
PosixThreadSupport::ThreadConstructionInfo constructionInfo("PosixThreads",
processMiniCLTask,
createMiniCLLocalStoreMemory,
maxNumOutstandingTasks);
threadSupport = new PosixThreadSupport(constructionInfo);
#else
///todo: add posix thread support for other platforms
SequentialThreadSupport::SequentialThreadConstructionInfo stc("MiniCL",processMiniCLTask,createMiniCLLocalStoreMemory);
threadSupport = new SequentialThreadSupport(stc);
#endif //USE_PTHREADS
#endif
}
MiniCLTaskScheduler* scheduler = new MiniCLTaskScheduler(threadSupport,maxNumOutstandingTasks);
*errcode_ret = 0;
return (cl_context)scheduler;
}
CL_API_ENTRY cl_int CL_API_CALL
clGetDeviceIDs(cl_platform_id /* platform */,
cl_device_type /* device_type */,
cl_uint /* num_entries */,
cl_device_id * /* devices */,
cl_uint * /* num_devices */) CL_API_SUFFIX__VERSION_1_0
{
return 0;
}
CL_API_ENTRY cl_context CL_API_CALL
clCreateContext(const cl_context_properties * properties ,
cl_uint num_devices ,
const cl_device_id * devices ,
void (*pfn_notify)(const char *, const void *, size_t, void *),
void * user_data ,
cl_int * errcode_ret ) CL_API_SUFFIX__VERSION_1_0
{
return clCreateContextFromType(properties,CL_DEVICE_TYPE_ALL,pfn_notify,user_data,errcode_ret);
}
CL_API_ENTRY cl_int CL_API_CALL clReleaseContext(cl_context context ) CL_API_SUFFIX__VERSION_1_0
{
MiniCLTaskScheduler* scheduler = (MiniCLTaskScheduler*) context;
btThreadSupportInterface* threadSupport = scheduler->getThreadSupportInterface();
delete scheduler;
delete threadSupport;
return 0;
}
extern CL_API_ENTRY cl_int CL_API_CALL
clFinish(cl_command_queue command_queue ) CL_API_SUFFIX__VERSION_1_0
{
MiniCLTaskScheduler* scheduler = (MiniCLTaskScheduler*) command_queue;
///wait for all work items to be completed
scheduler->flush();
return CL_SUCCESS;
}
extern CL_API_ENTRY cl_int CL_API_CALL
clGetProgramInfo(cl_program /* program */,
cl_program_info /* param_name */,
size_t /* param_value_size */,
void * /* param_value */,
size_t * /* param_value_size_ret */) CL_API_SUFFIX__VERSION_1_0
{
return 0;
}
extern CL_API_ENTRY cl_int CL_API_CALL
clGetKernelWorkGroupInfo(cl_kernel kernel ,
cl_device_id /* device */,
cl_kernel_work_group_info wgi/* param_name */,
size_t sz /* param_value_size */,
void * ptr /* param_value */,
size_t * /* param_value_size_ret */) CL_API_SUFFIX__VERSION_1_0
{
if((wgi == CL_KERNEL_WORK_GROUP_SIZE)
&&(sz == sizeof(size_t))
&&(ptr != NULL))
{
MiniCLKernel* miniCLKernel = (MiniCLKernel*)kernel;
MiniCLTaskScheduler* scheduler = miniCLKernel->m_scheduler;
*((size_t*)ptr) = scheduler->getMaxNumOutstandingTasks();
return CL_SUCCESS;
}
else
{
return CL_INVALID_VALUE;
}
}

74
Engine/lib/bullet/src/MiniCL/MiniCLTask/MiniCLTask.cpp Normal file
View file

@ -0,0 +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 "MiniCL/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

62
Engine/lib/bullet/src/MiniCL/MiniCLTask/MiniCLTask.h Normal file
View file

@ -0,0 +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

519
Engine/lib/bullet/src/MiniCL/MiniCLTaskScheduler.cpp Normal file
View file

@ -0,0 +1,519 @@
/*
Bullet Continuous Collision Detection and Physics Library
Copyright (c) 2003-2007 Erwin Coumans http://bulletphysics.com
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.
*/
//#define __CELLOS_LV2__ 1
#define __BT_SKIP_UINT64_H 1
#define USE_SAMPLE_PROCESS 1
#ifdef USE_SAMPLE_PROCESS
#include "MiniCLTaskScheduler.h"
#include <stdio.h>
#ifdef __SPU__
void SampleThreadFunc(void* userPtr,void* lsMemory)
{
//do nothing
printf("hello world\n");
}
void* SamplelsMemoryFunc()
{
//don't create local store memory, just return 0
return 0;
}
#else
#include "BulletMultiThreaded/btThreadSupportInterface.h"
//# include "SPUAssert.h"
#include <string.h>
#include "MiniCL/cl_platform.h"
extern "C" {
extern char SPU_SAMPLE_ELF_SYMBOL[];
}
MiniCLTaskScheduler::MiniCLTaskScheduler(btThreadSupportInterface* threadInterface, int maxNumOutstandingTasks)
:m_threadInterface(threadInterface),
m_maxNumOutstandingTasks(maxNumOutstandingTasks)
{
m_taskBusy.resize(m_maxNumOutstandingTasks);
m_spuSampleTaskDesc.resize(m_maxNumOutstandingTasks);
m_kernels.resize(0);
for (int i = 0; i < m_maxNumOutstandingTasks; i++)
{
m_taskBusy[i] = false;
}
m_numBusyTasks = 0;
m_currentTask = 0;
m_initialized = false;
m_threadInterface->startSPU();
}
MiniCLTaskScheduler::~MiniCLTaskScheduler()
{
m_threadInterface->stopSPU();
}
void MiniCLTaskScheduler::initialize()
{
#ifdef DEBUG_SPU_TASK_SCHEDULING
printf("MiniCLTaskScheduler::initialize()\n");
#endif //DEBUG_SPU_TASK_SCHEDULING
for (int i = 0; i < m_maxNumOutstandingTasks; i++)
{
m_taskBusy[i] = false;
}
m_numBusyTasks = 0;
m_currentTask = 0;
m_initialized = true;
}
void MiniCLTaskScheduler::issueTask(int firstWorkUnit, int lastWorkUnit, MiniCLKernel* kernel)
{
#ifdef DEBUG_SPU_TASK_SCHEDULING
printf("MiniCLTaskScheduler::issueTask (m_currentTask= %d\)n", m_currentTask);
#endif //DEBUG_SPU_TASK_SCHEDULING
m_taskBusy[m_currentTask] = true;
m_numBusyTasks++;
MiniCLTaskDesc& taskDesc = m_spuSampleTaskDesc[m_currentTask];
{
// send task description in event message
taskDesc.m_firstWorkUnit = firstWorkUnit;
taskDesc.m_lastWorkUnit = lastWorkUnit;
taskDesc.m_kernel = kernel;
//some bookkeeping to recognize finished tasks
taskDesc.m_taskId = m_currentTask;
// for (int i=0;i<MINI_CL_MAX_ARG;i++)
for (unsigned int i=0; i < kernel->m_numArgs; i++)
{
taskDesc.m_argSizes[i] = kernel->m_argSizes[i];
if (taskDesc.m_argSizes[i])
{
taskDesc.m_argData[i] = kernel->m_argData[i];
// memcpy(&taskDesc.m_argData[i],&argData[MINICL_MAX_ARGLENGTH*i],taskDesc.m_argSizes[i]);
}
}
}
m_threadInterface->sendRequest(1, (ppu_address_t) &taskDesc, m_currentTask);
// if all tasks busy, wait for spu event to clear the task.
if (m_numBusyTasks >= m_maxNumOutstandingTasks)
{
unsigned int taskId;
unsigned int outputSize;
for (int i=0;i<m_maxNumOutstandingTasks;i++)
{
if (m_taskBusy[i])
{
taskId = i;
break;
}
}
m_threadInterface->waitForResponse(&taskId, &outputSize);
//printf("PPU: after issue, received event: %u %d\n", taskId, outputSize);
postProcess(taskId, outputSize);
m_taskBusy[taskId] = false;
m_numBusyTasks--;
}
// find new task buffer
for (int i = 0; i < m_maxNumOutstandingTasks; i++)
{
if (!m_taskBusy[i])
{
m_currentTask = i;
break;
}
}
}
///Optional PPU-size post processing for each task
void MiniCLTaskScheduler::postProcess(int taskId, int outputSize)
{
}
void MiniCLTaskScheduler::flush()
{
#ifdef DEBUG_SPU_TASK_SCHEDULING
printf("\nSpuCollisionTaskProcess::flush()\n");
#endif //DEBUG_SPU_TASK_SCHEDULING
// all tasks are issued, wait for all tasks to be complete
while(m_numBusyTasks > 0)
{
// Consolidating SPU code
unsigned int taskId;
unsigned int outputSize;
for (int i=0;i<m_maxNumOutstandingTasks;i++)
{
if (m_taskBusy[i])
{
taskId = i;
break;
}
}
{
m_threadInterface->waitForResponse(&taskId, &outputSize);
}
//printf("PPU: flushing, received event: %u %d\n", taskId, outputSize);
postProcess(taskId, outputSize);
m_taskBusy[taskId] = false;
m_numBusyTasks--;
}
}
typedef void (*MiniCLKernelLauncher0)(int);
typedef void (*MiniCLKernelLauncher1)(void*, int);
typedef void (*MiniCLKernelLauncher2)(void*, void*, int);
typedef void (*MiniCLKernelLauncher3)(void*, void*, void*, int);
typedef void (*MiniCLKernelLauncher4)(void*, void*, void*, void*, int);
typedef void (*MiniCLKernelLauncher5)(void*, void*, void*, void*, void*, int);
typedef void (*MiniCLKernelLauncher6)(void*, void*, void*, void*, void*, void*, int);
typedef void (*MiniCLKernelLauncher7)(void*, void*, void*, void*, void*, void*, void*, int);
typedef void (*MiniCLKernelLauncher8)(void*, void*, void*, void*, void*, void*, void*, void*, int);
typedef void (*MiniCLKernelLauncher9)(void*, void*, void*, void*, void*, void*, void*, void*, void*, int);
typedef void (*MiniCLKernelLauncher10)(void*, void*, void*, void*, void*, void*, void*, void*, void*, void*, int);
typedef void (*MiniCLKernelLauncher11)(void*, void*, void*, void*, void*, void*, void*, void*, void*, void*, void*, int);
typedef void (*MiniCLKernelLauncher12)(void*, void*, void*, void*, void*, void*, void*, void*, void*, void*, void*, void*, int);
typedef void (*MiniCLKernelLauncher13)(void*, void*, void*, void*, void*, void*, void*, void*, void*, void*, void*, void*, void*, int);
typedef void (*MiniCLKernelLauncher14)(void*, void*, void*, void*, void*, void*, void*, void*, void*, void*, void*, void*, void*, void*, int);
typedef void (*MiniCLKernelLauncher15)(void*, void*, void*, void*, void*, void*, void*, void*, void*, void*, void*, void*, void*, void*, void*, int);
typedef void (*MiniCLKernelLauncher16)(void*, void*, void*, void*, void*, void*, void*, void*, void*, void*, void*, void*, void*, void*, void*, void*, int);
static void kernelLauncher0(MiniCLTaskDesc* taskDesc, int guid)
{
((MiniCLKernelLauncher0)(taskDesc->m_kernel->m_launcher))(guid);
}
static void kernelLauncher1(MiniCLTaskDesc* taskDesc, int guid)
{
((MiniCLKernelLauncher1)(taskDesc->m_kernel->m_pCode))( taskDesc->m_argData[0],
guid);
}
static void kernelLauncher2(MiniCLTaskDesc* taskDesc, int guid)
{
((MiniCLKernelLauncher2)(taskDesc->m_kernel->m_pCode))( taskDesc->m_argData[0],
taskDesc->m_argData[1],
guid);
}
static void kernelLauncher3(MiniCLTaskDesc* taskDesc, int guid)
{
((MiniCLKernelLauncher3)(taskDesc->m_kernel->m_pCode))( taskDesc->m_argData[0],
taskDesc->m_argData[1],
taskDesc->m_argData[2],
guid);
}
static void kernelLauncher4(MiniCLTaskDesc* taskDesc, int guid)
{
((MiniCLKernelLauncher4)(taskDesc->m_kernel->m_pCode))( taskDesc->m_argData[0],
taskDesc->m_argData[1],
taskDesc->m_argData[2],
taskDesc->m_argData[3],
guid);
}
static void kernelLauncher5(MiniCLTaskDesc* taskDesc, int guid)
{
((MiniCLKernelLauncher5)(taskDesc->m_kernel->m_pCode))( taskDesc->m_argData[0],
taskDesc->m_argData[1],
taskDesc->m_argData[2],
taskDesc->m_argData[3],
taskDesc->m_argData[4],
guid);
}
static void kernelLauncher6(MiniCLTaskDesc* taskDesc, int guid)
{
((MiniCLKernelLauncher6)(taskDesc->m_kernel->m_pCode))( taskDesc->m_argData[0],
taskDesc->m_argData[1],
taskDesc->m_argData[2],
taskDesc->m_argData[3],
taskDesc->m_argData[4],
taskDesc->m_argData[5],
guid);
}
static void kernelLauncher7(MiniCLTaskDesc* taskDesc, int guid)
{
((MiniCLKernelLauncher7)(taskDesc->m_kernel->m_pCode))( taskDesc->m_argData[0],
taskDesc->m_argData[1],
taskDesc->m_argData[2],
taskDesc->m_argData[3],
taskDesc->m_argData[4],
taskDesc->m_argData[5],
taskDesc->m_argData[6],
guid);
}
static void kernelLauncher8(MiniCLTaskDesc* taskDesc, int guid)
{
((MiniCLKernelLauncher8)(taskDesc->m_kernel->m_pCode))( taskDesc->m_argData[0],
taskDesc->m_argData[1],
taskDesc->m_argData[2],
taskDesc->m_argData[3],
taskDesc->m_argData[4],
taskDesc->m_argData[5],
taskDesc->m_argData[6],
taskDesc->m_argData[7],
guid);
}
static void kernelLauncher9(MiniCLTaskDesc* taskDesc, int guid)
{
((MiniCLKernelLauncher9)(taskDesc->m_kernel->m_pCode))( taskDesc->m_argData[0],
taskDesc->m_argData[1],
taskDesc->m_argData[2],
taskDesc->m_argData[3],
taskDesc->m_argData[4],
taskDesc->m_argData[5],
taskDesc->m_argData[6],
taskDesc->m_argData[7],
taskDesc->m_argData[8],
guid);
}
static void kernelLauncher10(MiniCLTaskDesc* taskDesc, int guid)
{
((MiniCLKernelLauncher10)(taskDesc->m_kernel->m_pCode))(taskDesc->m_argData[0],
taskDesc->m_argData[1],
taskDesc->m_argData[2],
taskDesc->m_argData[3],
taskDesc->m_argData[4],
taskDesc->m_argData[5],
taskDesc->m_argData[6],
taskDesc->m_argData[7],
taskDesc->m_argData[8],
taskDesc->m_argData[9],
guid);
}
static void kernelLauncher11(MiniCLTaskDesc* taskDesc, int guid)
{
((MiniCLKernelLauncher11)(taskDesc->m_kernel->m_pCode))(taskDesc->m_argData[0],
taskDesc->m_argData[1],
taskDesc->m_argData[2],
taskDesc->m_argData[3],
taskDesc->m_argData[4],
taskDesc->m_argData[5],
taskDesc->m_argData[6],
taskDesc->m_argData[7],
taskDesc->m_argData[8],
taskDesc->m_argData[9],
taskDesc->m_argData[10],
guid);
}
static void kernelLauncher12(MiniCLTaskDesc* taskDesc, int guid)
{
((MiniCLKernelLauncher12)(taskDesc->m_kernel->m_pCode))(taskDesc->m_argData[0],
taskDesc->m_argData[1],
taskDesc->m_argData[2],
taskDesc->m_argData[3],
taskDesc->m_argData[4],
taskDesc->m_argData[5],
taskDesc->m_argData[6],
taskDesc->m_argData[7],
taskDesc->m_argData[8],
taskDesc->m_argData[9],
taskDesc->m_argData[10],
taskDesc->m_argData[11],
guid);
}
static void kernelLauncher13(MiniCLTaskDesc* taskDesc, int guid)
{
((MiniCLKernelLauncher13)(taskDesc->m_kernel->m_pCode))(taskDesc->m_argData[0],
taskDesc->m_argData[1],
taskDesc->m_argData[2],
taskDesc->m_argData[3],
taskDesc->m_argData[4],
taskDesc->m_argData[5],
taskDesc->m_argData[6],
taskDesc->m_argData[7],
taskDesc->m_argData[8],
taskDesc->m_argData[9],
taskDesc->m_argData[10],
taskDesc->m_argData[11],
taskDesc->m_argData[12],
guid);
}
static void kernelLauncher14(MiniCLTaskDesc* taskDesc, int guid)
{
((MiniCLKernelLauncher14)(taskDesc->m_kernel->m_pCode))(taskDesc->m_argData[0],
taskDesc->m_argData[1],
taskDesc->m_argData[2],
taskDesc->m_argData[3],
taskDesc->m_argData[4],
taskDesc->m_argData[5],
taskDesc->m_argData[6],
taskDesc->m_argData[7],
taskDesc->m_argData[8],
taskDesc->m_argData[9],
taskDesc->m_argData[10],
taskDesc->m_argData[11],
taskDesc->m_argData[12],
taskDesc->m_argData[13],
guid);
}
static void kernelLauncher15(MiniCLTaskDesc* taskDesc, int guid)
{
((MiniCLKernelLauncher15)(taskDesc->m_kernel->m_pCode))(taskDesc->m_argData[0],
taskDesc->m_argData[1],
taskDesc->m_argData[2],
taskDesc->m_argData[3],
taskDesc->m_argData[4],
taskDesc->m_argData[5],
taskDesc->m_argData[6],
taskDesc->m_argData[7],
taskDesc->m_argData[8],
taskDesc->m_argData[9],
taskDesc->m_argData[10],
taskDesc->m_argData[11],
taskDesc->m_argData[12],
taskDesc->m_argData[13],
taskDesc->m_argData[14],
guid);
}
static void kernelLauncher16(MiniCLTaskDesc* taskDesc, int guid)
{
((MiniCLKernelLauncher16)(taskDesc->m_kernel->m_pCode))(taskDesc->m_argData[0],
taskDesc->m_argData[1],
taskDesc->m_argData[2],
taskDesc->m_argData[3],
taskDesc->m_argData[4],
taskDesc->m_argData[5],
taskDesc->m_argData[6],
taskDesc->m_argData[7],
taskDesc->m_argData[8],
taskDesc->m_argData[9],
taskDesc->m_argData[10],
taskDesc->m_argData[11],
taskDesc->m_argData[12],
taskDesc->m_argData[13],
taskDesc->m_argData[14],
taskDesc->m_argData[15],
guid);
}
static kernelLauncherCB spLauncherList[MINI_CL_MAX_ARG+1] =
{
kernelLauncher0,
kernelLauncher1,
kernelLauncher2,
kernelLauncher3,
kernelLauncher4,
kernelLauncher5,
kernelLauncher6,
kernelLauncher7,
kernelLauncher8,
kernelLauncher9,
kernelLauncher10,
kernelLauncher11,
kernelLauncher12,
kernelLauncher13,
kernelLauncher14,
kernelLauncher15,
kernelLauncher16
};
void MiniCLKernel::updateLauncher()
{
m_launcher = spLauncherList[m_numArgs];
}
struct MiniCLKernelDescEntry
{
void* pCode;
const char* pName;
};
static MiniCLKernelDescEntry spKernelDesc[256];
static int sNumKernelDesc = 0;
MiniCLKernelDesc::MiniCLKernelDesc(void* pCode, const char* pName)
{
for(int i = 0; i < sNumKernelDesc; i++)
{
if(!strcmp(pName, spKernelDesc[i].pName))
{ // already registered
btAssert(spKernelDesc[i].pCode == pCode);
return;
}
}
spKernelDesc[sNumKernelDesc].pCode = pCode;
spKernelDesc[sNumKernelDesc].pName = pName;
sNumKernelDesc++;
}
MiniCLKernel* MiniCLKernel::registerSelf()
{
m_scheduler->registerKernel(this);
for(int i = 0; i < sNumKernelDesc; i++)
{
if(!strcmp(m_name, spKernelDesc[i].pName))
{
m_pCode = spKernelDesc[i].pCode;
return this;
}
}
return NULL;
}
#endif
#endif //USE_SAMPLE_PROCESS

194
Engine/lib/bullet/src/MiniCL/MiniCLTaskScheduler.h Normal file
View file

@ -0,0 +1,194 @@
/*
Bullet Continuous Collision Detection and Physics Library
Copyright (c) 2003-2007 Erwin Coumans http://bulletphysics.com
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_SCHEDULER_H
#define MINICL_TASK_SCHEDULER_H
#include <assert.h>
#include "BulletMultiThreaded/PlatformDefinitions.h"
#include <stdlib.h>
#include "LinearMath/btAlignedObjectArray.h"
#include "MiniCLTask/MiniCLTask.h"
//just add your commands here, try to keep them globally unique for debugging purposes
#define CMD_SAMPLE_TASK_COMMAND 10
struct MiniCLKernel;
/// MiniCLTaskScheduler handles SPU processing of collision pairs.
/// When PPU issues a task, it will look for completed task buffers
/// PPU will do postprocessing, dependent on workunit output (not likely)
class MiniCLTaskScheduler
{
// track task buffers that are being used, and total busy tasks
btAlignedObjectArray<bool> m_taskBusy;
btAlignedObjectArray<MiniCLTaskDesc> m_spuSampleTaskDesc;
btAlignedObjectArray<const MiniCLKernel*> m_kernels;
int m_numBusyTasks;
// the current task and the current entry to insert a new work unit
int m_currentTask;
bool m_initialized;
void postProcess(int taskId, int outputSize);
class btThreadSupportInterface* m_threadInterface;
int m_maxNumOutstandingTasks;
public:
MiniCLTaskScheduler(btThreadSupportInterface* threadInterface, int maxNumOutstandingTasks);
~MiniCLTaskScheduler();
///call initialize in the beginning of the frame, before addCollisionPairToTask
void initialize();
void issueTask(int firstWorkUnit, int lastWorkUnit, MiniCLKernel* kernel);
///call flush to submit potential outstanding work to SPUs and wait for all involved SPUs to be finished
void flush();
class btThreadSupportInterface* getThreadSupportInterface()
{
return m_threadInterface;
}
int findProgramCommandIdByName(const char* programName) const;
int getMaxNumOutstandingTasks() const
{
return m_maxNumOutstandingTasks;
}
void registerKernel(MiniCLKernel* kernel)
{
m_kernels.push_back(kernel);
}
};
typedef void (*kernelLauncherCB)(MiniCLTaskDesc* taskDesc, int guid);
struct MiniCLKernel
{
MiniCLTaskScheduler* m_scheduler;
// int m_kernelProgramCommandId;
char m_name[MINI_CL_MAX_KERNEL_NAME];
unsigned int m_numArgs;
kernelLauncherCB m_launcher;
void* m_pCode;
void updateLauncher();
MiniCLKernel* registerSelf();
void* m_argData[MINI_CL_MAX_ARG];
int m_argSizes[MINI_CL_MAX_ARG];
};
#if defined(USE_LIBSPE2) && defined(__SPU__)
////////////////////MAIN/////////////////////////////
#include "../SpuLibspe2Support.h"
#include <spu_intrinsics.h>
#include <spu_mfcio.h>
#include <SpuFakeDma.h>
void * SamplelsMemoryFunc();
void SampleThreadFunc(void* userPtr,void* lsMemory);
//#define DEBUG_LIBSPE2_MAINLOOP
int main(unsigned long long speid, addr64 argp, addr64 envp)
{
printf("SPU is up \n");
ATTRIBUTE_ALIGNED128(btSpuStatus status);
ATTRIBUTE_ALIGNED16( SpuSampleTaskDesc taskDesc ) ;
unsigned int received_message = Spu_Mailbox_Event_Nothing;
bool shutdown = false;
cellDmaGet(&status, argp.ull, sizeof(btSpuStatus), DMA_TAG(3), 0, 0);
cellDmaWaitTagStatusAll(DMA_MASK(3));
status.m_status = Spu_Status_Free;
status.m_lsMemory.p = SamplelsMemoryFunc();
cellDmaLargePut(&status, argp.ull, sizeof(btSpuStatus), DMA_TAG(3), 0, 0);
cellDmaWaitTagStatusAll(DMA_MASK(3));
while (!shutdown)
{
received_message = spu_read_in_mbox();
switch(received_message)
{
case Spu_Mailbox_Event_Shutdown:
shutdown = true;
break;
case Spu_Mailbox_Event_Task:
// refresh the status
#ifdef DEBUG_LIBSPE2_MAINLOOP
printf("SPU recieved Task \n");
#endif //DEBUG_LIBSPE2_MAINLOOP
cellDmaGet(&status, argp.ull, sizeof(btSpuStatus), DMA_TAG(3), 0, 0);
cellDmaWaitTagStatusAll(DMA_MASK(3));
btAssert(status.m_status==Spu_Status_Occupied);
cellDmaGet(&taskDesc, status.m_taskDesc.p, sizeof(SpuSampleTaskDesc), DMA_TAG(3), 0, 0);
cellDmaWaitTagStatusAll(DMA_MASK(3));
SampleThreadFunc((void*)&taskDesc, reinterpret_cast<void*> (taskDesc.m_mainMemoryPtr) );
break;
case Spu_Mailbox_Event_Nothing:
default:
break;
}
// set to status free and wait for next task
status.m_status = Spu_Status_Free;
cellDmaLargePut(&status, argp.ull, sizeof(btSpuStatus), DMA_TAG(3), 0, 0);
cellDmaWaitTagStatusAll(DMA_MASK(3));
}
return 0;
}
//////////////////////////////////////////////////////
#endif
#endif // MINICL_TASK_SCHEDULER_H

6
Engine/lib/bullet/src/MiniCL/cl.h
View file

@ -155,8 +155,10 @@ typedef struct _cl_image_format {
#define CL_DEVICE_TYPE_CPU (1 << 1)
#define CL_DEVICE_TYPE_GPU (1 << 2)
#define CL_DEVICE_TYPE_ACCELERATOR (1 << 3)
#define CL_DEVICE_TYPE_DEBUG (1 << 4)
#define CL_DEVICE_TYPE_ALL 0xFFFFFFFF
// cl_device_info
#define CL_DEVICE_TYPE 0x1000
#define CL_DEVICE_VENDOR_ID 0x1001
@ -435,7 +437,7 @@ clGetDeviceInfo(cl_device_id /* device */,
// Context APIs
extern CL_API_ENTRY cl_context CL_API_CALL
clCreateContext(cl_context_properties * /* properties */,
clCreateContext(const cl_context_properties * /* properties */,
cl_uint /* num_devices */,
const cl_device_id * /* devices */,
void (*pfn_notify)(const char *, const void *, size_t, void *) /* pfn_notify */,
@ -443,7 +445,7 @@ clCreateContext(cl_context_properties * /* properties */,
cl_int * /* errcode_ret */) CL_API_SUFFIX__VERSION_1_0;
extern CL_API_ENTRY cl_context CL_API_CALL
clCreateContextFromType(cl_context_properties * /* properties */,
clCreateContextFromType(const cl_context_properties * /* properties */,
cl_device_type /* device_type */,
void (*pfn_notify)(const char *, const void *, size_t, void *) /* pfn_notify */,
void * /* user_data */,

439
Engine/lib/bullet/src/MiniCL/cl_MiniCL_Defs.h Normal file
View file

@ -0,0 +1,439 @@
/*
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 <float.h>
#include <math.h>
#include "LinearMath/btScalar.h"
#include "MiniCL/cl.h"
#define __kernel
#define __global
#define __local
#define get_global_id(a) __guid_arg
#define get_local_id(a) ((__guid_arg) % gMiniCLNumOutstandingTasks)
#define get_local_size(a) (gMiniCLNumOutstandingTasks)
#define get_group_id(a) ((__guid_arg) / gMiniCLNumOutstandingTasks)
//static unsigned int as_uint(float val) { return *((unsigned int*)&val); }
#define CLK_LOCAL_MEM_FENCE 0x01
#define CLK_GLOBAL_MEM_FENCE 0x02
static void barrier(unsigned int a)
{
// TODO : implement
}
//ATTRIBUTE_ALIGNED16(struct) float8
struct float8
{
float s0;
float s1;
float s2;
float s3;
float s4;
float s5;
float s6;
float s7;
float8(float scalar)
{
s0=s1=s2=s3=s4=s5=s6=s7=scalar;
}
};
float select( float arg0, float arg1, bool select)
{
if (select)
return arg0;
return arg1;
}
#define __constant
struct float3
{
float x,y,z;
float3& operator+=(const float3& other)
{
x += other.x;
y += other.y;
z += other.z;
return *this;
}
float3& operator-=(const float3& other)
{
x -= other.x;
y -= other.y;
z -= other.z;
return *this;
}
};
static float dot(const float3&a ,const float3& b)
{
float3 tmp;
tmp.x = a.x*b.x;
tmp.y = a.y*b.y;
tmp.z = a.z*b.z;
return tmp.x+tmp.y+tmp.z;
}
static float3 operator-(const float3& a,const float3& b)
{
float3 tmp;
tmp.x = a.x - b.x;
tmp.y = a.y - b.y;
tmp.z = a.z - b.z;
return tmp;
}
static float3 operator*(const float& scalar,const float3& b)
{
float3 tmp;
tmp.x = scalar * b.x;
tmp.y = scalar * b.y;
tmp.z = scalar * b.z;
return tmp;
}
static float3 operator*(const float3& a,const float& scalar)
{
float3 tmp;
tmp.x = a.x * scalar;
tmp.y = a.y * scalar;
tmp.z = a.z * scalar;
return tmp;
}
static float3 operator*(const float3& a,const float3& b)
{
float3 tmp;
tmp.x = a.x * b.x;
tmp.y = a.y * b.y;
tmp.z = a.z * b.z;
return tmp;
}
//ATTRIBUTE_ALIGNED16(struct) float4
struct float4
{
union
{
struct {
float x;
float y;
float z;
};
float3 xyz;
};
float w;
float4() {}
float4(float v0, float v1, float v2, float v3)
{
x=v0;
y=v1;
z=v2;
w=v3;
}
float4(float3 xyz, float scalarW)
{
x = xyz.x;
y = xyz.y;
z = xyz.z;
w = scalarW;
}
float4(float v)
{
x = y = z = w = v;
}
float4 operator*(const float4& other)
{
float4 tmp;
tmp.x = x*other.x;
tmp.y = y*other.y;
tmp.z = z*other.z;
tmp.w = w*other.w;
return tmp;
}
float4 operator*(const float& other)
{
float4 tmp;
tmp.x = x*other;
tmp.y = y*other;
tmp.z = z*other;
tmp.w = w*other;
return tmp;
}
float4& operator+=(const float4& other)
{
x += other.x;
y += other.y;
z += other.z;
w += other.w;
return *this;
}
float4& operator-=(const float4& other)
{
x -= other.x;
y -= other.y;
z -= other.z;
w -= other.w;
return *this;
}
float4& operator *=(float scalar)
{
x *= scalar;
y *= scalar;
z *= scalar;
w *= scalar;
return (*this);
}
};
static float4 fabs(const float4& a)
{
float4 tmp;
tmp.x = a.x < 0.f ? 0.f : a.x;
tmp.y = a.y < 0.f ? 0.f : a.y;
tmp.z = a.z < 0.f ? 0.f : a.z;
tmp.w = a.w < 0.f ? 0.f : a.w;
return tmp;
}
static float4 operator+(const float4& a,const float4& b)
{
float4 tmp;
tmp.x = a.x + b.x;
tmp.y = a.y + b.y;
tmp.z = a.z + b.z;
tmp.w = a.w + b.w;
return tmp;
}
static float8 operator+(const float8& a,const float8& b)
{
float8 tmp(0);
tmp.s0 = a.s0 + b.s0;
tmp.s1 = a.s1 + b.s1;
tmp.s2 = a.s2 + b.s2;
tmp.s3 = a.s3 + b.s3;
tmp.s4 = a.s4 + b.s4;
tmp.s5 = a.s5 + b.s5;
tmp.s6 = a.s6 + b.s6;
tmp.s7 = a.s7 + b.s7;
return tmp;
}
static float4 operator-(const float4& a,const float4& b)
{
float4 tmp;
tmp.x = a.x - b.x;
tmp.y = a.y - b.y;
tmp.z = a.z - b.z;
tmp.w = a.w - b.w;
return tmp;
}
static float8 operator-(const float8& a,const float8& b)
{
float8 tmp(0);
tmp.s0 = a.s0 - b.s0;
tmp.s1 = a.s1 - b.s1;
tmp.s2 = a.s2 - b.s2;
tmp.s3 = a.s3 - b.s3;
tmp.s4 = a.s4 - b.s4;
tmp.s5 = a.s5 - b.s5;
tmp.s6 = a.s6 - b.s6;
tmp.s7 = a.s7 - b.s7;
return tmp;
}
static float4 operator*(float a,const float4& b)
{
float4 tmp;
tmp.x = a * b.x;
tmp.y = a * b.y;
tmp.z = a * b.z;
tmp.w = a * b.w;
return tmp;
}
static float4 operator/(const float4& b,float a)
{
float4 tmp;
tmp.x = b.x/a;
tmp.y = b.y/a;
tmp.z = b.z/a;
tmp.w = b.w/a;
return tmp;
}
static float dot(const float4&a ,const float4& b)
{
float4 tmp;
tmp.x = a.x*b.x;
tmp.y = a.y*b.y;
tmp.z = a.z*b.z;
tmp.w = a.w*b.w;
return tmp.x+tmp.y+tmp.z+tmp.w;
}
static float length(const float4&a)
{
float l = sqrtf(a.x*a.x+a.y*a.y+a.z*a.z);
return l;
}
static float4 normalize(const float4&a)
{
float4 tmp;
float l = length(a);
tmp = 1.f/l*a;
return tmp;
}
static float4 cross(const float4&a ,const float4& b)
{
float4 tmp;
tmp.x = a.y*b.z - a.z*b.y;
tmp.y = -a.x*b.z + a.z*b.x;
tmp.z = a.x*b.y - a.y*b.x;
tmp.w = 0.f;
return tmp;
}
static float max(float a, float b)
{
return (a >= b) ? a : b;
}
static float min(float a, float b)
{
return (a <= b) ? a : b;
}
static float fmax(float a, float b)
{
return (a >= b) ? a : b;
}
static float fmin(float a, float b)
{
return (a <= b) ? a : b;
}
struct int2
{
int x,y;
};
struct uint2
{
unsigned int x,y;
};
//typedef int2 uint2;
typedef unsigned int uint;
struct int4
{
int x,y,z,w;
};
struct uint4
{
unsigned int x,y,z,w;
uint4() {}
uint4(uint val) { x = y = z = w = val; }
uint4& operator+=(const uint4& other)
{
x += other.x;
y += other.y;
z += other.z;
w += other.w;
return *this;
}
};
static uint4 operator+(const uint4& a,const uint4& b)
{
uint4 tmp;
tmp.x = a.x + b.x;
tmp.y = a.y + b.y;
tmp.z = a.z + b.z;
tmp.w = a.w + b.w;
return tmp;
}
static uint4 operator-(const uint4& a,const uint4& b)
{
uint4 tmp;
tmp.x = a.x - b.x;
tmp.y = a.y - b.y;
tmp.z = a.z - b.z;
tmp.w = a.w - b.w;
return tmp;
}
#define native_sqrt sqrtf
#define native_sin sinf
#define native_cos cosf
#define native_powr powf
#define GUID_ARG ,int __guid_arg
#define GUID_ARG_VAL ,__guid_arg
#define as_int(a) (*((int*)&(a)))
extern "C" int gMiniCLNumOutstandingTasks;
// extern "C" void __kernel_func();

2
Engine/lib/bullet/src/MiniCL/cl_gl.h
View file

@ -27,7 +27,7 @@
#ifdef __APPLE__
#include <OpenCL/cl_platform.h>
#else
#include <CL/cl_platform.h>
#include <MiniCL/cl_platform.h>
#endif
#ifdef __cplusplus

12
Engine/lib/bullet/src/MiniCL/cl_platform.h
View file

@ -24,6 +24,16 @@
#ifndef __CL_PLATFORM_H
#define __CL_PLATFORM_H
#define CL_PLATFORM_MINI_CL 0x12345
struct MiniCLKernelDesc
{
MiniCLKernelDesc(void* pCode, const char* pName);
};
#define MINICL_REGISTER(__kernel_func) static MiniCLKernelDesc __kernel_func##Desc((void*)__kernel_func, #__kernel_func);
#ifdef __APPLE__
/* Contains #defines for AVAILABLE_MAC_OS_X_VERSION_10_6_AND_LATER below */
#include <AvailabilityMacros.h>
@ -43,7 +53,7 @@ extern "C" {
#define CL_EXTENSION_WEAK_LINK
#endif
#ifdef WIN32
#if defined (_WIN32) && ! defined (__MINGW32__)
typedef signed __int8 int8_t;
typedef unsigned __int8 uint8_t;
typedef signed __int16 int16_t;