Let's start with an API for creating threads, refer to: https://msdn.microsoft.com/en-us/library/windows/desktop/ms682453%28v=vs.85%29.aspx
Creates a thread to execute within the virtual address space of the the calling process.
CreateThread ( _in_opt_ lpsecurity_attributes lpthreadattributes, _in_ size_t dwstacksize, _in_ lpthread_start_routine lpstartaddress, _in_opt_ lpvoid lpparameter, _in_ DWORD dwcreationflags, _out_opt_ lpdword lpthreadid);
Use the example reference msdn:https://msdn.microsoft.com/en-us/library/windows/desktop/ms682516%28v=vs.85%29.aspx
For example, the thread:
#include <iostream>#include<windows.h>using namespacestd;#defineMax_threads 20#defineBuf_size 255DWORD Dwaccessorder=0;D word WINAPI mythreadfunction (lpvoid lppara) {DWORD*dwpara = (dword*) Lppara; cout<<"This is the first"<< *dwpara <<"a thread that starts up! "<<Endl; return 0;}intMain () {DWORD dwthreadidarray[max_threads]; HANDLE Hthreadarray[max_threads]; LPVOID Lpparam=nullptr; for(inti =0; i < max_threads; i++) {Lpparam= &i;//The following conditions may occur: parameter has not been passed to the thread, it has already put the thread up, such as i=12, up a thread //The thread has not received the parameter 12 o'clock, entered the next loop, I was overwritten, was assigned to 13, so this time again//up a thread, it might be the same thread as the one that just got the parameter.//And then the thread that receives the parameter, it is possible to output to the window than the thread that receives the parameter firstHthreadarray[i] =CreateThread (NULL,0, MyThreadFunction, Lpparam,0, &Dwthreadidarray[i]); } cout<<"This is the main thread"<<Endl; for(inti =0; i < max_threads; i++) {CloseHandle (hthreadarray[i]); } System ("Pause"); return 0;}
Output, no order:
From This we can determine that the threads created and the main thread that created them are independent actuators. When they are created "quickly", they begin to "act independently" and "scramble" for the CPU. In fact, the thread is determined by the operating system according to a certain scheduling algorithm to determine who first who executes after. Multithreading is actually asynchronous, and its purpose is to quickly and alternately get the execution of the CPU, allowing the user to assume that these threads are executed in parallel .
Let's make a slight change to the example above:
#include <iostream>#include<windows.h>using namespacestd;#defineMax_threads 20#defineBuf_size 255DWORD Dwaccessorder=0;D word WINAPI mythreadfunction (lpvoid lpparam) {Dwaccessorder++; cout<<"This is the first"<< Dwaccessorder <<"a thread that starts up! "<<Endl; return 0;}intMain () {DWORD dwthreadidarray[max_threads]; HANDLE Hthreadarray[max_threads]; LPVOID Lpparam=nullptr; for(inti =0; i < max_threads; i++) {Hthreadarray[i]=CreateThread (NULL,0, MyThreadFunction, NULL,0, &Dwthreadidarray[i]); } for(inti =0; i < max_threads; i++) {CloseHandle (hthreadarray[i]); } System ("Pause"); return 0;}
The following results are output:
When the thread T1 and T2 both need to access the variable dwaccessorder, if T1 is modifying the value of Dwaccessorder but has not yet completed, the operating system switches the execution to T2 at this point, T2 is given an incorrect value to read Dwaccessorder. Therefore, we must use some means to enable data that needs to be shared by multiple threads to be accessed by only one thread at a time. This means, called thread synchronization .
Windows thread synchronization (not finished)