class PThread |
This class defines a thread of execution in the system.
Run Time Type functions
I/O functions
Miscellaneous functions
Comparison functions
This class defines a thread of execution in the system. A thread is an independent flow of processor instructions. This differs from a process which also embodies a program address space and resource allocation. So threads can share memory and resources as they run in the context of a given process. A process always contains at least one thread. This is reflected in this library by the PProcess class being descended from the PThread class.The implementation of a thread is platform dependent. Not all platforms support concurrent threads within a process or even concurrent processes! For example, MS-DOS has no form of multi-threading or multi-processing, Microsoft Windows has a cooperative multi-processing but no multi-threading. Unix has full pre-emptive multi-processing but most cannot do multiple threads within that process while some Unix systems and Windows NT have full preemptive proceses and threads.
If a platform does not directly support multiple threads, the library will them using a cooperative co-routine technique. This requires that each thread of execution within a process, voluntarily yields control to other threads. This will occur if the thread is blocked inside an I/O function on a PChannel or when the PThread::Yield() function is explicitly called.
Note that this is cooperative. An endless loop will stop all threads in a process, possibly all processes on some platforms. If a lengthy operation is to take place that does not involve blocking I/O, eg pure computation or disk file I/O, then it is the responsiblity of the programmer to assure enough yielding for background threads to execute.
Note that the exact timing of the execution of code in threads can never be predicted. Thus you you can get a race condition on intialising a descendent class. To avoid this problem a thread is always started suspended. You must call the Resume() function after your descendent class construction is complete.
If synchronisation is required between threads then the use of semaphores is essential.
If the deletion is set to AutoDeleteThread then the PThread is assumed to be allocated with the new operator and may be freed using the delete operator as soon as the thread is terminated or executes to completion (usually the latter).
The stack size specified is not simply in bytes. It is a value that is multiplied by a factor into bytes depending on the target platform. For example a Unix system with a RISC processor may use significantly more stack than an MS-DOS platform. These sizes are normalised to the "stack factor" provided here. For some platforms, eg Windows NT, the stack size is only an initial size and the stack will automatically be increased as required.
Note that the correct way for a thread to terminate is to return from the Main() function.
If the thread is still running then this function is ignored.
Note that the correct way for a thread to terminate is to return from the Main() function or self terminate by calling Terminate() within the context of the thread which can then assure that all resources are cleaned up.
If susp is TRUE this increments an internal count of suspensions that must be matched by an equal number of calls to Resume() or Suspend(FALSE) before the thread actually executes again.
If susp is FALSE then this decrements the internal count of suspensions. If the count is <= 0 then the thread will run. Note that the thread will not be suspended until an equal number of Suspend(TRUE) calls are made.
Note that the correct way for a thread to terminate is to return from this function.
If the platform directly supports multiple threads then this function will do nothing.
Alphabetic index HTML hierarchy of classes or Java