1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
|
/*
* Copyright (c) 2008-2017, NVIDIA CORPORATION. All rights reserved.
*
* NVIDIA CORPORATION and its licensors retain all intellectual property
* and proprietary rights in and to this software, related documentation
* and any modifications thereto. Any use, reproduction, disclosure or
* distribution of this software and related documentation without an express
* license agreement from NVIDIA CORPORATION is strictly prohibited.
*/
#ifndef _THREAD_POOL_H_
#define _THREAD_POOL_H_
#include "PxTask.h"
#include "PxTaskManager.h"
#include "PxCpuDispatcher.h"
#include "PsSList.h"
#include "PsSync.h"
#include "PsThread.h"
#include "ApexUsingNamespace.h"
namespace physx
{
namespace pvdsdk
{
class ApexPvdClient;
}
}
namespace nvidia
{
namespace apex
{
PxCpuDispatcher* createDefaultThreadPool(unsigned int numThreads);
class SharedQueueEntry : public shdfnd::SListEntry
{
public:
SharedQueueEntry(void* objectRef) : mObjectRef(objectRef), mPooledEntry(false) {}
SharedQueueEntry() : mObjectRef(NULL), mPooledEntry(true) {}
public:
void* mObjectRef;
bool mPooledEntry; // True if the entry was preallocated in a pool
};
template<class Alloc = typename shdfnd::AllocatorTraits<SharedQueueEntry>::Type >
class SharedQueueEntryPool : private Alloc
{
public:
SharedQueueEntryPool(uint32_t poolSize, const Alloc& alloc = Alloc());
~SharedQueueEntryPool();
SharedQueueEntry* getEntry(void* objectRef);
void putEntry(SharedQueueEntry& entry);
private:
SharedQueueEntry* mTaskEntryPool;
shdfnd::SList mTaskEntryPtrPool;
};
template <class Alloc>
SharedQueueEntryPool<Alloc>::SharedQueueEntryPool(uint32_t poolSize, const Alloc& alloc)
: Alloc(alloc)
{
shdfnd::AlignedAllocator<PX_SLIST_ALIGNMENT, Alloc> alignedAlloc;
mTaskEntryPool = poolSize ? (SharedQueueEntry*)alignedAlloc.allocate(sizeof(SharedQueueEntry) * poolSize, __FILE__, __LINE__) : NULL;
if (mTaskEntryPool)
{
for (uint32_t i = 0; i < poolSize; i++)
{
PX_ASSERT((size_t(&mTaskEntryPool[i]) & (PX_SLIST_ALIGNMENT - 1)) == 0); // The SList entry must be aligned according to PX_SLIST_ALIGNMENT
PX_PLACEMENT_NEW(&mTaskEntryPool[i], SharedQueueEntry)();
PX_ASSERT(mTaskEntryPool[i].mPooledEntry == true);
mTaskEntryPtrPool.push(mTaskEntryPool[i]);
}
}
}
template <class Alloc>
SharedQueueEntryPool<Alloc>::~SharedQueueEntryPool()
{
if (mTaskEntryPool)
{
shdfnd::AlignedAllocator<PX_SLIST_ALIGNMENT, Alloc> alignedAlloc;
alignedAlloc.deallocate(mTaskEntryPool);
}
}
template <class Alloc>
SharedQueueEntry* SharedQueueEntryPool<Alloc>::getEntry(void* objectRef)
{
SharedQueueEntry* e = static_cast<SharedQueueEntry*>(mTaskEntryPtrPool.pop());
if (e)
{
PX_ASSERT(e->mPooledEntry == true);
e->mObjectRef = objectRef;
return e;
}
else
{
shdfnd::AlignedAllocator<PX_SLIST_ALIGNMENT, Alloc> alignedAlloc;
e = (SharedQueueEntry*)alignedAlloc.allocate(sizeof(SharedQueueEntry), __FILE__, __LINE__);
if (e)
{
PX_PLACEMENT_NEW(e, SharedQueueEntry)(objectRef);
PX_ASSERT(e->mPooledEntry == false);
}
return e;
}
}
template <class Alloc>
void SharedQueueEntryPool<Alloc>::putEntry(SharedQueueEntry& entry)
{
if (entry.mPooledEntry)
{
entry.mObjectRef = NULL;
mTaskEntryPtrPool.push(entry);
}
else
{
shdfnd::AlignedAllocator<PX_SLIST_ALIGNMENT, Alloc> alignedAlloc;
alignedAlloc.deallocate(&entry);
}
}
#define TASK_QUEUE_ENTRY_POOL_SIZE 128
class TaskQueueHelper
{
public:
static PxBaseTask* fetchTask(shdfnd::SList& taskQueue, SharedQueueEntryPool<>& entryPool)
{
SharedQueueEntry* entry = static_cast<SharedQueueEntry*>(taskQueue.pop());
if (entry)
{
PxBaseTask* task = reinterpret_cast<PxBaseTask*>(entry->mObjectRef);
entryPool.putEntry(*entry);
return task;
}
else
{
return NULL;
}
}
};
class DefaultCpuDispatcher;
class CpuWorkerThread : public shdfnd::Thread
{
public:
CpuWorkerThread();
~CpuWorkerThread();
void initialize(DefaultCpuDispatcher* ownerDispatcher);
void execute();
bool tryAcceptJobToLocalQueue(PxBaseTask& task, shdfnd::Thread::Id taskSubmitionThread);
PxBaseTask* giveUpJob();
protected:
SharedQueueEntryPool<> mQueueEntryPool;
DefaultCpuDispatcher* mOwner;
shdfnd::SList mLocalJobList;
shdfnd::Thread::Id mThreadId;
pvdsdk::ApexPvdClient *mApexPvdClient;
};
/*
* Default CpuDispatcher implementation, if none is provided
*/
class DefaultCpuDispatcher : public PxCpuDispatcher, public shdfnd::UserAllocated
{
friend class TaskQueueHelper;
private:
DefaultCpuDispatcher() : mQueueEntryPool(0) {}
~DefaultCpuDispatcher();
public:
DefaultCpuDispatcher(uint32_t numThreads, uint32_t* affinityMasks);
void submitTask(PxBaseTask& task);
void flush( PxBaseTask& task, int32_t targetRef);
uint32_t getWorkerCount() const;
void release();
PxBaseTask* getJob();
PxBaseTask* stealJob();
void waitForWork()
{
mWorkReady.wait();
}
void resetWakeSignal();
static uint32_t getAffinityMask(uint32_t affinityMask);
protected:
CpuWorkerThread* mWorkerThreads;
SharedQueueEntryPool<> mQueueEntryPool;
shdfnd::SList mJobList;
shdfnd::Sync mWorkReady;
uint32_t mNumThreads;
bool mShuttingDown;
pvdsdk::ApexPvdClient *mApexPvdClient;
};
} // end pxtask namespace
} // end physx namespace
#endif // _THREAD_POOL_H_
|