xenia-canary/third_party/crunch/crnlib/crn_threading_win32.cpp

429 lines
11 KiB
C++

// File: crn_win32_threading.cpp
// See Copyright Notice and license at the end of inc/crnlib.h
#include "crn_core.h"
#include "crn_threading_win32.h"
#include "crn_winhdr.h"
#include <process.h>
namespace crnlib
{
uint g_number_of_processors = 1;
void crn_threading_init()
{
SYSTEM_INFO g_system_info;
GetSystemInfo(&g_system_info);
g_number_of_processors = math::maximum<uint>(1U, g_system_info.dwNumberOfProcessors);
}
crn_thread_id_t crn_get_current_thread_id()
{
return static_cast<crn_thread_id_t>(GetCurrentThreadId());
}
void crn_sleep(unsigned int milliseconds)
{
Sleep(milliseconds);
}
uint crn_get_max_helper_threads()
{
if (g_number_of_processors > 1)
{
// use all CPU's
return CRNLIB_MIN((int)task_pool::cMaxThreads, (int)g_number_of_processors - 1);
}
return 0;
}
mutex::mutex(unsigned int spin_count)
{
CRNLIB_ASSUME(sizeof(mutex) >= sizeof(CRITICAL_SECTION));
void *p = m_buf;
CRITICAL_SECTION &m_cs = *static_cast<CRITICAL_SECTION *>(p);
BOOL status = true;
status = InitializeCriticalSectionAndSpinCount(&m_cs, spin_count);
if (!status)
crnlib_fail("mutex::mutex: InitializeCriticalSectionAndSpinCount failed", __FILE__, __LINE__);
#ifdef CRNLIB_BUILD_DEBUG
m_lock_count = 0;
#endif
}
mutex::~mutex()
{
void *p = m_buf;
CRITICAL_SECTION &m_cs = *static_cast<CRITICAL_SECTION *>(p);
#ifdef CRNLIB_BUILD_DEBUG
if (m_lock_count)
crnlib_assert("mutex::~mutex: mutex is still locked", __FILE__, __LINE__);
#endif
DeleteCriticalSection(&m_cs);
}
void mutex::lock()
{
void *p = m_buf;
CRITICAL_SECTION &m_cs = *static_cast<CRITICAL_SECTION *>(p);
EnterCriticalSection(&m_cs);
#ifdef CRNLIB_BUILD_DEBUG
m_lock_count++;
#endif
}
void mutex::unlock()
{
void *p = m_buf;
CRITICAL_SECTION &m_cs = *static_cast<CRITICAL_SECTION *>(p);
#ifdef CRNLIB_BUILD_DEBUG
if (!m_lock_count)
crnlib_assert("mutex::unlock: mutex is not locked", __FILE__, __LINE__);
m_lock_count--;
#endif
LeaveCriticalSection(&m_cs);
}
void mutex::set_spin_count(unsigned int count)
{
void *p = m_buf;
CRITICAL_SECTION &m_cs = *static_cast<CRITICAL_SECTION *>(p);
SetCriticalSectionSpinCount(&m_cs, count);
}
void spinlock::lock(uint32 max_spins, bool yielding)
{
if (g_number_of_processors <= 1)
max_spins = 1;
uint32 spinCount = 0;
uint32 yieldCount = 0;
for ( ; ; )
{
CRNLIB_ASSUME(sizeof(long) == sizeof(int32));
if (!InterlockedExchange((volatile long*)&m_flag, TRUE))
break;
YieldProcessor();
YieldProcessor();
YieldProcessor();
YieldProcessor();
YieldProcessor();
YieldProcessor();
YieldProcessor();
YieldProcessor();
spinCount++;
if ((yielding) && (spinCount >= max_spins))
{
switch (yieldCount)
{
case 0:
{
spinCount = 0;
Sleep(0);
yieldCount++;
break;
}
case 1:
{
if (g_number_of_processors <= 1)
spinCount = 0;
else
spinCount = max_spins / 2;
Sleep(1);
yieldCount++;
break;
}
case 2:
{
if (g_number_of_processors <= 1)
spinCount = 0;
else
spinCount = max_spins;
Sleep(2);
break;
}
}
}
}
CRNLIB_MEMORY_IMPORT_BARRIER
}
void spinlock::unlock()
{
CRNLIB_MEMORY_EXPORT_BARRIER
InterlockedExchange((volatile long*)&m_flag, FALSE);
}
semaphore::semaphore(int32 initialCount, int32 maximumCount, const char* pName)
{
m_handle = CreateSemaphoreA(NULL, initialCount, maximumCount, pName);
if (NULL == m_handle)
{
CRNLIB_FAIL("semaphore: CreateSemaphore() failed");
}
}
semaphore::~semaphore()
{
if (m_handle)
{
CloseHandle(m_handle);
m_handle = NULL;
}
}
void semaphore::release(int32 releaseCount, int32 *pPreviousCount)
{
CRNLIB_ASSUME(sizeof(LONG) == sizeof(int32));
if (0 == ReleaseSemaphore(m_handle, releaseCount, (LPLONG)pPreviousCount))
{
CRNLIB_FAIL("semaphore: ReleaseSemaphore() failed");
}
}
bool semaphore::try_release(int32 releaseCount, int32 *pPreviousCount)
{
CRNLIB_ASSUME(sizeof(LONG) == sizeof(int32));
return ReleaseSemaphore(m_handle, releaseCount, (LPLONG)pPreviousCount) != 0;
}
bool semaphore::wait(uint32 milliseconds)
{
uint32 result = WaitForSingleObject(m_handle, milliseconds);
if (WAIT_FAILED == result)
{
CRNLIB_FAIL("semaphore: WaitForSingleObject() failed");
}
return WAIT_OBJECT_0 == result;
}
task_pool::task_pool() :
m_pTask_stack(crnlib_new<ts_task_stack_t>()),
m_num_threads(0),
m_tasks_available(0, 32767),
m_all_tasks_completed(0, 1),
m_total_submitted_tasks(0),
m_total_completed_tasks(0),
m_exit_flag(false)
{
utils::zero_object(m_threads);
}
task_pool::task_pool(uint num_threads) :
m_pTask_stack(crnlib_new<ts_task_stack_t>()),
m_num_threads(0),
m_tasks_available(0, 32767),
m_all_tasks_completed(0, 1),
m_total_submitted_tasks(0),
m_total_completed_tasks(0),
m_exit_flag(false)
{
utils::zero_object(m_threads);
bool status = init(num_threads);
CRNLIB_VERIFY(status);
}
task_pool::~task_pool()
{
deinit();
crnlib_delete(m_pTask_stack);
}
bool task_pool::init(uint num_threads)
{
CRNLIB_ASSERT(num_threads <= cMaxThreads);
num_threads = math::minimum<uint>(num_threads, cMaxThreads);
deinit();
bool succeeded = true;
m_num_threads = 0;
while (m_num_threads < num_threads)
{
m_threads[m_num_threads] = (HANDLE)_beginthreadex(NULL, 32768, thread_func, this, 0, NULL);
CRNLIB_ASSERT(m_threads[m_num_threads] != 0);
if (!m_threads[m_num_threads])
{
succeeded = false;
break;
}
m_num_threads++;
}
if (!succeeded)
{
deinit();
return false;
}
return true;
}
void task_pool::deinit()
{
if (m_num_threads)
{
join();
// Set exit flag, then release all threads. Each should wakeup and exit.
atomic_exchange32(&m_exit_flag, true);
m_tasks_available.release(m_num_threads);
// Now wait for each thread to exit.
for (uint i = 0; i < m_num_threads; i++)
{
if (m_threads[i])
{
for ( ; ; )
{
// Can be an INFINITE delay, but set at 30 seconds so this function always provably exits.
DWORD result = WaitForSingleObject(m_threads[i], 30000);
if ((result == WAIT_OBJECT_0) || (result == WAIT_ABANDONED))
break;
}
CloseHandle(m_threads[i]);
m_threads[i] = NULL;
}
}
m_num_threads = 0;
atomic_exchange32(&m_exit_flag, false);
}
if (m_pTask_stack)
m_pTask_stack->clear();
m_total_submitted_tasks = 0;
m_total_completed_tasks = 0;
}
bool task_pool::queue_task(task_callback_func pFunc, uint64 data, void* pData_ptr)
{
CRNLIB_ASSERT(pFunc);
task tsk;
tsk.m_callback = pFunc;
tsk.m_data = data;
tsk.m_pData_ptr = pData_ptr;
tsk.m_flags = 0;
atomic_increment32(&m_total_submitted_tasks);
if (!m_pTask_stack->try_push(tsk))
{
atomic_increment32(&m_total_completed_tasks);
return false;
}
m_tasks_available.release(1);
return true;
}
// It's the object's responsibility to delete pObj within the execute_task() method, if needed!
bool task_pool::queue_task(executable_task* pObj, uint64 data, void* pData_ptr)
{
CRNLIB_ASSERT(pObj);
task tsk;
tsk.m_pObj = pObj;
tsk.m_data = data;
tsk.m_pData_ptr = pData_ptr;
tsk.m_flags = cTaskFlagObject;
atomic_increment32(&m_total_submitted_tasks);
if (!m_pTask_stack->try_push(tsk))
{
atomic_increment32(&m_total_completed_tasks);
return false;
}
m_tasks_available.release(1);
return true;
}
void task_pool::process_task(task& tsk)
{
if (tsk.m_flags & cTaskFlagObject)
tsk.m_pObj->execute_task(tsk.m_data, tsk.m_pData_ptr);
else
tsk.m_callback(tsk.m_data, tsk.m_pData_ptr);
if (atomic_increment32(&m_total_completed_tasks) == m_total_submitted_tasks)
{
// Try to signal the semaphore (the max count is 1 so this may actually fail).
m_all_tasks_completed.try_release();
}
}
void task_pool::join()
{
// Try to steal any outstanding tasks. This could cause one or more worker threads to wake up and immediately go back to sleep, which is wasteful but should be harmless.
task tsk;
while (m_pTask_stack->pop(tsk))
process_task(tsk);
// At this point the task stack is empty.
// Now wait for all concurrent tasks to complete. The m_all_tasks_completed semaphore has a max count of 1, so it's possible it could have saturated to 1 as the tasks
// where issued and asynchronously completed, so this loop may iterate a few times.
const int total_submitted_tasks = atomic_add32(&m_total_submitted_tasks, 0);
while (m_total_completed_tasks != total_submitted_tasks)
{
// If the previous (m_total_completed_tasks != total_submitted_tasks) check failed the semaphore MUST be eventually signalled once the last task completes.
// So I think this can actually be an INFINITE delay, but it shouldn't really matter if it's 1ms.
m_all_tasks_completed.wait(1);
}
}
unsigned __stdcall task_pool::thread_func(void* pContext)
{
task_pool* pPool = static_cast<task_pool*>(pContext);
for ( ; ; )
{
if (!pPool->m_tasks_available.wait())
break;
if (pPool->m_exit_flag)
break;
task tsk;
if (pPool->m_pTask_stack->pop(tsk))
pPool->process_task(tsk);
}
_endthreadex(0);
return 0;
}
} // namespace crnlib