🚀 go-pugleaf

Keith Bennett (kbennett@cpcug.org) wrote:
: This question concerns the 4th parameter passed to _beginthread().
: This is the void * that can point to anything the programmer wants
: it to.

: I have a function that loops and starts 'n' separate threads (pls see
: example below).  In order to be safe, do I need to guarantee that the
: memory pointed to by that pointer I pass is active and dedicated
: exclusively to that thread for the thread's lifetime?  (Assuming I
: don't *want* any interference from other threads.)

: For example, I assume that the code below will not work properly because
: the contents pointed to by &n may change before the newly started thread
: can read it:


: void _Optlink my_thread_func( void *args)
: {
: 	int the_number_passed = * (int *) args;  /* ... */
: }


: void my_thread_starting_func()
: {
: 	for( int n = 0; n < 4; n++)
: 		_beginthread( my_thread_func, 0, 32 * 1024, &n);
: }


: But how about if I change this to:

: void my_thread_starting_func()
: {
: 	int n[4];

: 	n[0] = 0; n[1] = 1; n[2] = 2; n[3] = 3;

: 	for( int i = 0; i < 4; i++)
: 		_beginthread( my_thread_func, 0, 32 * 1024, &n[i]);
: }

: What if 'my_thread_starting_func()' finishes before one or more of my
: 'my_thread_func()''s reads the value, and that stack area is already
: overwritten by another function?

: So, am I correct in assuming that in this situation I must absolutely
: guarantee that this memory will be used exclusively by that thread
: for as long as the thread might need it?  And therefore, that I should
: probably use a static int array?

I haven't tried it so I may have this all wrong, but Dorfman and 
Neuberger (Effective Multithreading in OS/2) list a function called 
_threadstore() which is to be used in conjunction with _beginthread and 
_endthread for "per thread data area".  the prototype is

#include <process.h>
void **_threadstore(void);

They are using the CSet++ compiler.

Hope this helps,

Hawley Rising
rising@crl.com