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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301
|
/*++
Module Name:
ParallelTask.h
Abstract:
Simple parallel task manager
Authors:
Ravi Pandya, May 2012
Environment:
User mode service.
Revision History:
--*/
#pragma once
#include "stdafx.h"
#include "Compat.h"
#include "exit.h"
#include "Error.h"
/*++
Simple class to handle parallelized algorithms.
TContext should extend TContextBase, and provide the following methods:
void initializeThread()
Called once on main thread after TContext has been assigned from common,
and threadNum set.
void runThread()
Called to run the thread's work until termination.
May use something like RangeSplitter to get work.
void finishThread(TContext* common)
Called once on main thread after all threads have finished,
to write results back to common.
--*/
template <class TContext>
class ParallelTask
{
public:
inline TContext* getCommonContext() { return common; }
// i_common should have totalThreads & bindToProcessors set
ParallelTask(TContext* i_common);
// run all threads until completion, gather results in common
void run();
// run all tasks on a separate thread
void fork();
private:
// initial & final context
TContext* common;
// array of per-thread contexts
TContext* contexts;
static void threadWorker(void* threadContext);
static void forkWorker(void* threadContext);
};
/*++
Base for type parameter to parallel task
--*/
struct TaskContextBase
{
// should be set before passing to ParallelTask constructor
int totalThreads;
bool bindToProcessors;
// time taken to run in millis
_int64 time;
// for internal use:
int threadNum; // current thread number, 0...totalThreads-1
SingleWaiterObject *doneWaiter; // Gets notified when the last thread ends.
volatile int runningThreads;
volatile int *pRunningThreads;
#ifdef _MSC_VER
volatile int *nThreadsAllocatingMemory;
EventObject *memoryAllocationCompleteBarrier;
bool useTimingBarrier;
#endif // _MSC_VER
};
template <class TContext>
ParallelTask<TContext>::ParallelTask(
TContext* i_common)
: common(i_common), contexts(NULL)
{
_ASSERT(i_common->totalThreads > 0);
}
template <class TContext>
void
ParallelTask<TContext>::run()
{
_int64 start = timeInMillis();
SingleWaiterObject doneWaiter;
if (!CreateSingleWaiterObject(&doneWaiter)) {
WriteErrorMessage( "Failed to create single waiter object for thread completion.\n");
soft_exit(1);
}
#ifdef _MSC_VER
int nThreadsAllocatingMemory = common->totalThreads;
EventObject memoryAllocationCompleteBarrier;
CreateEventObject(&memoryAllocationCompleteBarrier);
common->nThreadsAllocatingMemory = &nThreadsAllocatingMemory;
common->memoryAllocationCompleteBarrier = &memoryAllocationCompleteBarrier;
#endif // _MSC_VER
common->doneWaiter = &doneWaiter;
common->runningThreads = common->totalThreads;
common->pRunningThreads = &common->runningThreads;
contexts = new TContext[common->totalThreads];
for (int i = 0; i < common->totalThreads; i++) {
contexts[i] = *common;
contexts[i].threadNum = i;
contexts[i].initializeThread();
if (!StartNewThread(ParallelTask<TContext>::threadWorker, &contexts[i])) {
WriteErrorMessage( "Unable to start worker thread.\n");
soft_exit(1);
}
}
#ifdef _MSC_VER
if (common->useTimingBarrier) {
WaitForEvent(&memoryAllocationCompleteBarrier);
WriteStatusMessage("Cleared timing barrier.\n");
start = timeInMillis();
}
#endif // _MSC_VER
if (!WaitForSingleWaiterObject(&doneWaiter)) {
WriteErrorMessage( "Waiting for all threads to finish failed\n");
soft_exit(1);
}
DestroySingleWaiterObject(&doneWaiter);
#ifdef _MSC_VER
DestroyEventObject(&memoryAllocationCompleteBarrier);
#endif // _MSC_VER
for (int i = 0; i < common->totalThreads; i++) {
contexts[i].finishThread(common);
}
common->time = timeInMillis() - start;
}
template <class TContext>
void
ParallelTask<TContext>::fork()
{
if (!StartNewThread(ParallelTask<TContext>::forkWorker, this)) {
WriteErrorMessage( "Unable to fork task thread.\n");
soft_exit(1);
}
}
template <class TContext>
void
ParallelTask<TContext>::forkWorker(
void* forkArg)
{
((ParallelTask<TContext>*) forkArg)->run();
}
template <class TContext>
void
ParallelTask<TContext>::threadWorker(
void* threadArg)
{
TContext* context = (TContext*) threadArg;
if (context->bindToProcessors && !DoesThreadHaveProcessorAffinitySet()) {
BindThreadToProcessor(context->threadNum);
}
context->runThread();
// Decrement the running thread count and wake up the waiter if it hits 0.
if (0 == InterlockedDecrementAndReturnNewValue(context->pRunningThreads)) {
SignalSingleWaiterObject(context->doneWaiter);
}
}
struct WorkerContext;
class ParallelWorker;
class ParallelWorkerManager;
// coroutined parallel workers
// does code inline if numThreads = 0
// can either callback when done, or synchronously wait for all to complete
class ParallelCoworker
{
public:
typedef void (*Callback)(void*);
ParallelCoworker(int i_numThreads, bool i_bindToProcessors, ParallelWorkerManager* supplier, Callback callback = NULL, void* parameter = NULL);
~ParallelCoworker();
// start forked thread running
void start();
// do one unit of work, asynchronously if callback, else synchronously
void step();
// stop everything, waits until all threads exit
void stop();
ParallelWorkerManager* getManager() { return manager; }
private:
EventObject *workReady; // One per worker thread
EventObject *workDone; // One per worker thread
ParallelWorker** workers; // one per worker thread
ParallelWorkerManager* manager;
volatile bool stopped;
const int numThreads;
const bool bindToProcessors;
Callback callback;
void* parameter;
WorkerContext* context;
ParallelTask<WorkerContext>* task;
SingleWaiterObject finished;
friend struct WorkerContext;
};
// abstract classes for specifying the actual work
// creates new per-thread workers
class ParallelWorker;
class ParallelWorkerManager
{
public:
// todo: using void* context pointers to avoid pain of templates but should really be made typesafe
virtual void initialize(void* context) {}
virtual ParallelWorker* createWorker() = 0;
virtual ~ParallelWorkerManager() {}
virtual void beginStep() {}
virtual void finishStep() {}
void configure(ParallelWorker* worker, int threadNum, int totalThreads); // special case
};
// per-thread worker
class ParallelWorker
{
public:
ParallelWorker() {}
virtual ~ParallelWorker() {}
virtual void initialize() {}
virtual void step() = 0;
protected:
ParallelWorkerManager* getManager() { return manager; }
int getThreadNum() { return threadNum; }
int getNumThreads() { return numThreads; }
private:
friend class ParallelCoworker;
friend class ParallelWorkerManager;
void configure(ParallelWorkerManager* i_manager, int i_threadNum, int i_numThreads)
{ manager = i_manager; threadNum = i_threadNum; numThreads = i_numThreads; }
ParallelWorkerManager* manager;
int threadNum;
int numThreads;
};
struct WorkerContext : public TaskContextBase
{
ParallelCoworker* shared;
void initializeThread();
void runThread();
void finishThread(WorkerContext* common);
};
|