pthread_mutex

[KIBO]

I read that when porting Windows code using EnterCriticalSection etc that you must use pthread_mutex_lock and friends but I'm wondering if there are any differences. In windows the advantage of using critical sections is that they require no switch to kernel mode. Is that the same for the linux pthread mutex?

[matsp]

I don't see why a pthread_mutex would need to enter the kernel for the locking itself - of course, if some other thread holds the lock, you probably will enter a "sleep" or "wait" function that in turn involves the scheduler.

--
Mats

[KIBO]

in windows the mutex functions do that because it is possible to synchronize processes and only the kernel knows about the processes. I thought maybe its the same in linux

[CornedBee]

Nope, the posix primitives are not inter-process.

[brewbuck]

I read that when porting Windows code using EnterCriticalSection etc that you must use pthread_mutex_lock and friends but I'm wondering if there are any differences. In windows the advantage of using critical sections is that they require no switch to kernel mode. Is that the same for the linux pthread mutex?

In theory there need not be a switch to kernel mode when locking/unlocking an uncontended mutex. Obviously, the kernel must be informed in the following cases:

1. A process tries to lock an already-locked mutex, which means the process attempting to acquire the lock needs to go to sleep.
2. A process unlocks a mutex which has other processes waiting on it. The kernel has to wake them up.