Fast NonRecursiveMutex support by Yakov Markovitch, markovitch@iso.ru,
who wrote:
Here's the new version of thread_nt.h. More particular, there is a
new version of thread lock that uses kernel object (e.g. semaphore)
only in case of contention; in other case it simply uses interlocked
functions, which are faster by the order of magnitude. It doesn't
make much difference without threads present, but as soon as thread
machinery initialised and (mostly) the interpreter global lock is on,
difference becomes tremendous. I've included a small script, which
initialises threads and launches pystone. With original thread_nt.h,
Pystone results with initialised threads are twofold worse then w/o
threads. With the new version, only 10% worse. I have used this
patch for about 6 months (with threaded and non-threaded
applications). It works remarkably well (though I'd desperately
prefer Python was free-threaded; I hope, it will soon).
26 years ago Fast NonRecursiveMutex support by Yakov Markovitch, markovitch@iso.ru,
who wrote:
Here's the new version of thread_nt.h. More particular, there is a
new version of thread lock that uses kernel object (e.g. semaphore)
only in case of contention; in other case it simply uses interlocked
functions, which are faster by the order of magnitude. It doesn't
make much difference without threads present, but as soon as thread
machinery initialised and (mostly) the interpreter global lock is on,
difference becomes tremendous. I've included a small script, which
initialises threads and launches pystone. With original thread_nt.h,
Pystone results with initialised threads are twofold worse then w/o
threads. With the new version, only 10% worse. I have used this
patch for about 6 months (with threaded and non-threaded
applications). It works remarkably well (though I'd desperately
prefer Python was free-threaded; I hope, it will soon).
26 years ago Fast NonRecursiveMutex support by Yakov Markovitch, markovitch@iso.ru,
who wrote:
Here's the new version of thread_nt.h. More particular, there is a
new version of thread lock that uses kernel object (e.g. semaphore)
only in case of contention; in other case it simply uses interlocked
functions, which are faster by the order of magnitude. It doesn't
make much difference without threads present, but as soon as thread
machinery initialised and (mostly) the interpreter global lock is on,
difference becomes tremendous. I've included a small script, which
initialises threads and launches pystone. With original thread_nt.h,
Pystone results with initialised threads are twofold worse then w/o
threads. With the new version, only 10% worse. I have used this
patch for about 6 months (with threaded and non-threaded
applications). It works remarkably well (though I'd desperately
prefer Python was free-threaded; I hope, it will soon).
26 years ago Fast NonRecursiveMutex support by Yakov Markovitch, markovitch@iso.ru,
who wrote:
Here's the new version of thread_nt.h. More particular, there is a
new version of thread lock that uses kernel object (e.g. semaphore)
only in case of contention; in other case it simply uses interlocked
functions, which are faster by the order of magnitude. It doesn't
make much difference without threads present, but as soon as thread
machinery initialised and (mostly) the interpreter global lock is on,
difference becomes tremendous. I've included a small script, which
initialises threads and launches pystone. With original thread_nt.h,
Pystone results with initialised threads are twofold worse then w/o
threads. With the new version, only 10% worse. I have used this
patch for about 6 months (with threaded and non-threaded
applications). It works remarkably well (though I'd desperately
prefer Python was free-threaded; I hope, it will soon).
26 years ago Fast NonRecursiveMutex support by Yakov Markovitch, markovitch@iso.ru,
who wrote:
Here's the new version of thread_nt.h. More particular, there is a
new version of thread lock that uses kernel object (e.g. semaphore)
only in case of contention; in other case it simply uses interlocked
functions, which are faster by the order of magnitude. It doesn't
make much difference without threads present, but as soon as thread
machinery initialised and (mostly) the interpreter global lock is on,
difference becomes tremendous. I've included a small script, which
initialises threads and launches pystone. With original thread_nt.h,
Pystone results with initialised threads are twofold worse then w/o
threads. With the new version, only 10% worse. I have used this
patch for about 6 months (with threaded and non-threaded
applications). It works remarkably well (though I'd desperately
prefer Python was free-threaded; I hope, it will soon).
26 years ago Fast NonRecursiveMutex support by Yakov Markovitch, markovitch@iso.ru,
who wrote:
Here's the new version of thread_nt.h. More particular, there is a
new version of thread lock that uses kernel object (e.g. semaphore)
only in case of contention; in other case it simply uses interlocked
functions, which are faster by the order of magnitude. It doesn't
make much difference without threads present, but as soon as thread
machinery initialised and (mostly) the interpreter global lock is on,
difference becomes tremendous. I've included a small script, which
initialises threads and launches pystone. With original thread_nt.h,
Pystone results with initialised threads are twofold worse then w/o
threads. With the new version, only 10% worse. I have used this
patch for about 6 months (with threaded and non-threaded
applications). It works remarkably well (though I'd desperately
prefer Python was free-threaded; I hope, it will soon).
26 years ago Fast NonRecursiveMutex support by Yakov Markovitch, markovitch@iso.ru,
who wrote:
Here's the new version of thread_nt.h. More particular, there is a
new version of thread lock that uses kernel object (e.g. semaphore)
only in case of contention; in other case it simply uses interlocked
functions, which are faster by the order of magnitude. It doesn't
make much difference without threads present, but as soon as thread
machinery initialised and (mostly) the interpreter global lock is on,
difference becomes tremendous. I've included a small script, which
initialises threads and launches pystone. With original thread_nt.h,
Pystone results with initialised threads are twofold worse then w/o
threads. With the new version, only 10% worse. I have used this
patch for about 6 months (with threaded and non-threaded
applications). It works remarkably well (though I'd desperately
prefer Python was free-threaded; I hope, it will soon).
26 years ago Fast NonRecursiveMutex support by Yakov Markovitch, markovitch@iso.ru,
who wrote:
Here's the new version of thread_nt.h. More particular, there is a
new version of thread lock that uses kernel object (e.g. semaphore)
only in case of contention; in other case it simply uses interlocked
functions, which are faster by the order of magnitude. It doesn't
make much difference without threads present, but as soon as thread
machinery initialised and (mostly) the interpreter global lock is on,
difference becomes tremendous. I've included a small script, which
initialises threads and launches pystone. With original thread_nt.h,
Pystone results with initialised threads are twofold worse then w/o
threads. With the new version, only 10% worse. I have used this
patch for about 6 months (with threaded and non-threaded
applications). It works remarkably well (though I'd desperately
prefer Python was free-threaded; I hope, it will soon).
26 years ago Fast NonRecursiveMutex support by Yakov Markovitch, markovitch@iso.ru,
who wrote:
Here's the new version of thread_nt.h. More particular, there is a
new version of thread lock that uses kernel object (e.g. semaphore)
only in case of contention; in other case it simply uses interlocked
functions, which are faster by the order of magnitude. It doesn't
make much difference without threads present, but as soon as thread
machinery initialised and (mostly) the interpreter global lock is on,
difference becomes tremendous. I've included a small script, which
initialises threads and launches pystone. With original thread_nt.h,
Pystone results with initialised threads are twofold worse then w/o
threads. With the new version, only 10% worse. I have used this
patch for about 6 months (with threaded and non-threaded
applications). It works remarkably well (though I'd desperately
prefer Python was free-threaded; I hope, it will soon).
26 years ago Fast NonRecursiveMutex support by Yakov Markovitch, markovitch@iso.ru,
who wrote:
Here's the new version of thread_nt.h. More particular, there is a
new version of thread lock that uses kernel object (e.g. semaphore)
only in case of contention; in other case it simply uses interlocked
functions, which are faster by the order of magnitude. It doesn't
make much difference without threads present, but as soon as thread
machinery initialised and (mostly) the interpreter global lock is on,
difference becomes tremendous. I've included a small script, which
initialises threads and launches pystone. With original thread_nt.h,
Pystone results with initialised threads are twofold worse then w/o
threads. With the new version, only 10% worse. I have used this
patch for about 6 months (with threaded and non-threaded
applications). It works remarkably well (though I'd desperately
prefer Python was free-threaded; I hope, it will soon).
26 years ago Fast NonRecursiveMutex support by Yakov Markovitch, markovitch@iso.ru,
who wrote:
Here's the new version of thread_nt.h. More particular, there is a
new version of thread lock that uses kernel object (e.g. semaphore)
only in case of contention; in other case it simply uses interlocked
functions, which are faster by the order of magnitude. It doesn't
make much difference without threads present, but as soon as thread
machinery initialised and (mostly) the interpreter global lock is on,
difference becomes tremendous. I've included a small script, which
initialises threads and launches pystone. With original thread_nt.h,
Pystone results with initialised threads are twofold worse then w/o
threads. With the new version, only 10% worse. I have used this
patch for about 6 months (with threaded and non-threaded
applications). It works remarkably well (though I'd desperately
prefer Python was free-threaded; I hope, it will soon).
26 years ago Fast NonRecursiveMutex support by Yakov Markovitch, markovitch@iso.ru,
who wrote:
Here's the new version of thread_nt.h. More particular, there is a
new version of thread lock that uses kernel object (e.g. semaphore)
only in case of contention; in other case it simply uses interlocked
functions, which are faster by the order of magnitude. It doesn't
make much difference without threads present, but as soon as thread
machinery initialised and (mostly) the interpreter global lock is on,
difference becomes tremendous. I've included a small script, which
initialises threads and launches pystone. With original thread_nt.h,
Pystone results with initialised threads are twofold worse then w/o
threads. With the new version, only 10% worse. I have used this
patch for about 6 months (with threaded and non-threaded
applications). It works remarkably well (though I'd desperately
prefer Python was free-threaded; I hope, it will soon).
26 years ago Fast NonRecursiveMutex support by Yakov Markovitch, markovitch@iso.ru,
who wrote:
Here's the new version of thread_nt.h. More particular, there is a
new version of thread lock that uses kernel object (e.g. semaphore)
only in case of contention; in other case it simply uses interlocked
functions, which are faster by the order of magnitude. It doesn't
make much difference without threads present, but as soon as thread
machinery initialised and (mostly) the interpreter global lock is on,
difference becomes tremendous. I've included a small script, which
initialises threads and launches pystone. With original thread_nt.h,
Pystone results with initialised threads are twofold worse then w/o
threads. With the new version, only 10% worse. I have used this
patch for about 6 months (with threaded and non-threaded
applications). It works remarkably well (though I'd desperately
prefer Python was free-threaded; I hope, it will soon).
26 years ago Fast NonRecursiveMutex support by Yakov Markovitch, markovitch@iso.ru,
who wrote:
Here's the new version of thread_nt.h. More particular, there is a
new version of thread lock that uses kernel object (e.g. semaphore)
only in case of contention; in other case it simply uses interlocked
functions, which are faster by the order of magnitude. It doesn't
make much difference without threads present, but as soon as thread
machinery initialised and (mostly) the interpreter global lock is on,
difference becomes tremendous. I've included a small script, which
initialises threads and launches pystone. With original thread_nt.h,
Pystone results with initialised threads are twofold worse then w/o
threads. With the new version, only 10% worse. I have used this
patch for about 6 months (with threaded and non-threaded
applications). It works remarkably well (though I'd desperately
prefer Python was free-threaded; I hope, it will soon).
26 years ago Fast NonRecursiveMutex support by Yakov Markovitch, markovitch@iso.ru,
who wrote:
Here's the new version of thread_nt.h. More particular, there is a
new version of thread lock that uses kernel object (e.g. semaphore)
only in case of contention; in other case it simply uses interlocked
functions, which are faster by the order of magnitude. It doesn't
make much difference without threads present, but as soon as thread
machinery initialised and (mostly) the interpreter global lock is on,
difference becomes tremendous. I've included a small script, which
initialises threads and launches pystone. With original thread_nt.h,
Pystone results with initialised threads are twofold worse then w/o
threads. With the new version, only 10% worse. I have used this
patch for about 6 months (with threaded and non-threaded
applications). It works remarkably well (though I'd desperately
prefer Python was free-threaded; I hope, it will soon).
26 years ago  An Anonymous Coward on c.l.py posted a little program with bizarre
behavior, creating many threads very quickly. A long debugging session
revealed that the Windows implementation of PyThread_start_new_thread()
was choked with "laziness" errors:
1. It checked MS _beginthread() for a failure return, but when that
happened it returned heap trash as the function result, instead of
an id of -1 (the proper error-return value).
2. It didn't consider that the Win32 CreateSemaphore() can fail.
3. When creating a great many threads very quickly, it's quite possible
that any particular bootstrap call can take virtually any amount of
time to return. But the code waited for a maximum of 5 seconds, and
didn't check to see whether the semaphore it was waiting for got
signaled. If it in fact timed out, the function could again return
heap trash as the function result. This is actually what confused
the test program, as the heap trash usually turned out to be 0, and
then multiple threads all got id 0 simultaneously, confusing the
hell out of threading.py's _active dict (mapping id to thread
object). A variety of baffling behaviors followed from that.
WRT #1 and #2, error returns are checked now, and "thread.error: can't
start new thread" gets raised now if a new thread (or new semaphore)
can't be created. WRT #3, we now wait for the semaphore without a
timeout.
Also removed useless local vrbls, folded long lines, and changed callobj
to a stack auto (it was going thru malloc/free instead, for no discernible
reason).
Bugfix candidate.
23 years ago Fast NonRecursiveMutex support by Yakov Markovitch, markovitch@iso.ru,
who wrote:
Here's the new version of thread_nt.h. More particular, there is a
new version of thread lock that uses kernel object (e.g. semaphore)
only in case of contention; in other case it simply uses interlocked
functions, which are faster by the order of magnitude. It doesn't
make much difference without threads present, but as soon as thread
machinery initialised and (mostly) the interpreter global lock is on,
difference becomes tremendous. I've included a small script, which
initialises threads and launches pystone. With original thread_nt.h,
Pystone results with initialised threads are twofold worse then w/o
threads. With the new version, only 10% worse. I have used this
patch for about 6 months (with threaded and non-threaded
applications). It works remarkably well (though I'd desperately
prefer Python was free-threaded; I hope, it will soon).
26 years ago |
|
/* This code implemented by Dag.Gruneau@elsa.preseco.comm.se *//* Fast NonRecursiveMutex support by Yakov Markovitch, markovitch@iso.ru *//* Eliminated some memory leaks, gsw@agere.com */
#include <windows.h>
#include <limits.h>
#ifdef HAVE_PROCESS_H
#include <process.h>
#endif
typedef struct NRMUTEX { LONG owned ; DWORD thread_id ; HANDLE hevent ;} NRMUTEX, *PNRMUTEX ;
BOOLInitializeNonRecursiveMutex(PNRMUTEX mutex){ mutex->owned = -1 ; /* No threads have entered NonRecursiveMutex */ mutex->thread_id = 0 ; mutex->hevent = CreateEvent(NULL, FALSE, FALSE, NULL) ; return mutex->hevent != NULL ; /* TRUE if the mutex is created */}
VOIDDeleteNonRecursiveMutex(PNRMUTEX mutex){ /* No in-use check */ CloseHandle(mutex->hevent) ; mutex->hevent = NULL ; /* Just in case */}
DWORDEnterNonRecursiveMutex(PNRMUTEX mutex, BOOL wait){ /* Assume that the thread waits successfully */ DWORD ret ;
/* InterlockedIncrement(&mutex->owned) == 0 means that no thread currently owns the mutex */ if (!wait) { if (InterlockedCompareExchange(&mutex->owned, 0, -1) != -1) return WAIT_TIMEOUT ; ret = WAIT_OBJECT_0 ; } else ret = InterlockedIncrement(&mutex->owned) ? /* Some thread owns the mutex, let's wait... */ WaitForSingleObject(mutex->hevent, INFINITE) : WAIT_OBJECT_0 ;
mutex->thread_id = GetCurrentThreadId() ; /* We own it */ return ret ;}
BOOLLeaveNonRecursiveMutex(PNRMUTEX mutex){ /* We don't own the mutex */ mutex->thread_id = 0 ; return InterlockedDecrement(&mutex->owned) < 0 || SetEvent(mutex->hevent) ; /* Other threads are waiting, wake one on them up */}
PNRMUTEXAllocNonRecursiveMutex(void){ PNRMUTEX mutex = (PNRMUTEX)malloc(sizeof(NRMUTEX)) ; if (mutex && !InitializeNonRecursiveMutex(mutex)) { free(mutex) ; mutex = NULL ; } return mutex ;}
voidFreeNonRecursiveMutex(PNRMUTEX mutex){ if (mutex) { DeleteNonRecursiveMutex(mutex) ; free(mutex) ; }}
long PyThread_get_thread_ident(void);
/*
* Initialization of the C package, should not be needed. */static voidPyThread__init_thread(void){}
/*
* Thread support. */
typedef struct { void (*func)(void*); void *arg;} callobj;
/* thunker to call adapt between the function type used by the system's
thread start function and the internally used one. */#if defined(MS_WINCE)
static DWORD WINAPI#else
static unsigned __stdcall#endif
bootstrap(void *call){ callobj *obj = (callobj*)call; void (*func)(void*) = obj->func; void *arg = obj->arg; HeapFree(GetProcessHeap(), 0, obj); func(arg); return 0;}
longPyThread_start_new_thread(void (*func)(void *), void *arg){ HANDLE hThread; unsigned threadID; callobj *obj;
dprintf(("%ld: PyThread_start_new_thread called\n", PyThread_get_thread_ident())); if (!initialized) PyThread_init_thread();
obj = (callobj*)HeapAlloc(GetProcessHeap(), 0, sizeof(*obj)); if (!obj) return -1; obj->func = func; obj->arg = arg;#if defined(MS_WINCE)
hThread = CreateThread(NULL, Py_SAFE_DOWNCAST(_pythread_stacksize, Py_ssize_t, SIZE_T), bootstrap, obj, 0, &threadID);#else
hThread = (HANDLE)_beginthreadex(0, Py_SAFE_DOWNCAST(_pythread_stacksize, Py_ssize_t, unsigned int), bootstrap, obj, 0, &threadID);#endif
if (hThread == 0) {#if defined(MS_WINCE)
/* Save error in variable, to prevent PyThread_get_thread_ident
from clobbering it. */ unsigned e = GetLastError(); dprintf(("%ld: PyThread_start_new_thread failed, win32 error code %u\n", PyThread_get_thread_ident(), e));#else
/* I've seen errno == EAGAIN here, which means "there are
* too many threads". */ int e = errno; dprintf(("%ld: PyThread_start_new_thread failed, errno %d\n", PyThread_get_thread_ident(), e));#endif
threadID = (unsigned)-1; HeapFree(GetProcessHeap(), 0, obj); } else { dprintf(("%ld: PyThread_start_new_thread succeeded: %p\n", PyThread_get_thread_ident(), (void*)hThread)); CloseHandle(hThread); } return (long) threadID;}
/*
* Return the thread Id instead of an handle. The Id is said to uniquely identify the * thread in the system */longPyThread_get_thread_ident(void){ if (!initialized) PyThread_init_thread();
return GetCurrentThreadId();}
voidPyThread_exit_thread(void){ dprintf(("%ld: PyThread_exit_thread called\n", PyThread_get_thread_ident())); if (!initialized) exit(0);#if defined(MS_WINCE)
ExitThread(0);#else
_endthreadex(0);#endif
}
/*
* Lock support. It has too be implemented as semaphores. * I [Dag] tried to implement it with mutex but I could find a way to * tell whether a thread already own the lock or not. */PyThread_type_lockPyThread_allocate_lock(void){ PNRMUTEX aLock;
dprintf(("PyThread_allocate_lock called\n")); if (!initialized) PyThread_init_thread();
aLock = AllocNonRecursiveMutex() ;
dprintf(("%ld: PyThread_allocate_lock() -> %p\n", PyThread_get_thread_ident(), aLock));
return (PyThread_type_lock) aLock;}
voidPyThread_free_lock(PyThread_type_lock aLock){ dprintf(("%ld: PyThread_free_lock(%p) called\n", PyThread_get_thread_ident(),aLock));
FreeNonRecursiveMutex(aLock) ;}
/*
* Return 1 on success if the lock was acquired * * and 0 if the lock was not acquired. This means a 0 is returned * if the lock has already been acquired by this thread! */intPyThread_acquire_lock(PyThread_type_lock aLock, int waitflag){ int success ;
dprintf(("%ld: PyThread_acquire_lock(%p, %d) called\n", PyThread_get_thread_ident(),aLock, waitflag));
success = aLock && EnterNonRecursiveMutex((PNRMUTEX) aLock, (waitflag ? INFINITE : 0)) == WAIT_OBJECT_0 ;
dprintf(("%ld: PyThread_acquire_lock(%p, %d) -> %d\n", PyThread_get_thread_ident(),aLock, waitflag, success));
return success;}
voidPyThread_release_lock(PyThread_type_lock aLock){ dprintf(("%ld: PyThread_release_lock(%p) called\n", PyThread_get_thread_ident(),aLock));
if (!(aLock && LeaveNonRecursiveMutex((PNRMUTEX) aLock))) dprintf(("%ld: Could not PyThread_release_lock(%p) error: %ld\n", PyThread_get_thread_ident(), aLock, GetLastError()));}
/* minimum/maximum thread stack sizes supported */#define THREAD_MIN_STACKSIZE 0x8000 /* 32kB */
#define THREAD_MAX_STACKSIZE 0x10000000 /* 256MB */
/* set the thread stack size.
* Return 0 if size is valid, -1 otherwise. */static int_pythread_nt_set_stacksize(size_t size){ /* set to default */ if (size == 0) { _pythread_stacksize = 0; return 0; }
/* valid range? */ if (size >= THREAD_MIN_STACKSIZE && size < THREAD_MAX_STACKSIZE) { _pythread_stacksize = size; return 0; }
return -1;}
#define THREAD_SET_STACKSIZE(x) _pythread_nt_set_stacksize(x)
/* use native Windows TLS functions */#define Py_HAVE_NATIVE_TLS
#ifdef Py_HAVE_NATIVE_TLS
intPyThread_create_key(void){ return (int) TlsAlloc();}
voidPyThread_delete_key(int key){ TlsFree(key);}
/* We must be careful to emulate the strange semantics implemented in thread.c,
* where the value is only set if it hasn't been set before. */intPyThread_set_key_value(int key, void *value){ BOOL ok; void *oldvalue;
assert(value != NULL); oldvalue = TlsGetValue(key); if (oldvalue != NULL) /* ignore value if already set */ return 0; ok = TlsSetValue(key, value); if (!ok) return -1; return 0;}
void *PyThread_get_key_value(int key){ /* because TLS is used in the Py_END_ALLOW_THREAD macro,
* it is necessary to preserve the windows error state, because * it is assumed to be preserved across the call to the macro. * Ideally, the macro should be fixed, but it is simpler to * do it here. */ DWORD error = GetLastError(); void *result = TlsGetValue(key); SetLastError(error); return result;}
voidPyThread_delete_key_value(int key){ /* NULL is used as "key missing", and it is also the default
* given by TlsGetValue() if nothing has been set yet. */ TlsSetValue(key, NULL);}
/* reinitialization of TLS is not necessary after fork when using
* the native TLS functions. And forking isn't supported on Windows either. */voidPyThread_ReInitTLS(void){}
#endif
|