Printable View
I have created a non blocking socket recv checker and keyboard checker. As you could imagain its using a loop to keep testing. But the problem is it bugs the CPU down majorly - how could I avoid this?
Using windows XP, visual C++ 6 enterprise, winsock2.h, conio.h
Thank you for your help!
My first thoughts are perhaps you are just running a loop that checks what is actually a blocking socket recv() instead of letting windows queue events in your program. My second thoughts are, may we see your code please?
haha sure - it might be a little strange. Its a small chat relay that I have devoloped. If you need more code let me know...
how would you setup events in console programs?Code:while(recent != 1)
{
if(_kbhit())
{
sendbuf[sendx] = _getch();
cout << sendbuf[sendx];
if(sendbuf[sendx] == 13)
{
sendbuf[sendx+1] = '\0';
TBS += send(m_socket, sendbuf, strlen(sendbuf), 0);
for(x = 0; x < strlen(name); x++)
{
cout << name[x];
}
cout << ": ";
for(x = 0; x < (strlen(sendbuf) - 1); x++)
{
cout << sendbuf[x];
}
cout << endl;
sendx = -1;
}
sendx++;
}
//FD_ZERO(&ReadFDs);
FD_SET(m_socket, &ReadFDs);
if(select(0, &ReadFDs, 0, 0, &count) > 0)
{
if(FD_ISSET(m_socket, &ReadFDs))
{
TBR += bytesRecv = recv( m_socket, &recent, 1, 0 );
recvbuf[recvx] = recent;
if(recvbuf[recvx] == 13)
{
for(x = 0; x < 80; x++)
cout << '\b';
for(x = 0; x < strlen(cname); x++)
{
cout << cname[x];
}
cout << ": ";
recvbuf[recvx] = '\0';
for(x = 0; x < strlen(recvbuf); x++)
{
cout << recvbuf[x];
}
if(recvx + strlen(cname) + 2 < 80)
{
for(x = 0; x < 80 - recvx - strlen(cname) - 2; x++)
cout << " ";
}
for(x = 0; x < sendx; x++)
cout << sendbuf[x];
recvx = -1;
}
recvx++;
}
if(bytesRecv == SOCKET_ERROR)
break;
}
}
One idea would be to add a sleep() function in your loop, and specify just a few milliseconds as the pause. It would slow down your program, but then so would any other option to stop using the CPU so much. Do a bit of research on the different options (there are many implementations of a pause function out there), and make sure you use one that uses the free time to give to other CPU uses. IIRC the windows.h version does that.
Yea I have seen many posts on this board saying to use the sleep() funciton. But I never really liked the idea of my program sleeping. Really I just want it to wait for certain incoming data without bashing my CPU. I mean windows is able to function without the CPU being used a lot.
>>to wait for certain incoming data without bashing my CPU
Well then, specify a timeout other than 0 in select() ;)
Just like to point out, since m_socket is the only socket in the FD_SET, as long as select() returns >0 then the socket that is readable will be m_socket. So the nested if is somewhat redundant in this case.Code:if(select(0, &ReadFDs, 0, 0, &count) > 0)
{
if(FD_ISSET(m_socket, &ReadFDs))
Actually I had code like the for while - but then it slows down keyboard input time - haha.
I'm such a problem aint I :-p.
Ohh and the redundant FD_SET is necessary because if I hadnt set it - then FD_ISSET will ALWAYS return true - even if there is no data wait. I think that is what your are pointing out about my redundancy...
No I mean:
Since there is only 1 socket in the FD_SET, if m_socket is readable then select() will return 1; if m_socket is not readable, select() will return <1. So you don't need the if(FD_ISSET...) test at all in this particular case.Code:if(select(0, &ReadFDs, 0, 0, &count) > 0)
{
//do something with m_socket
>>but then it slows down keyboard input time
Well then, that changes things a little doesn't it ;) The only thing I can tell you, then, is Sleep() from <windows.h> or in your case <winsock2.h>, coupled with the nonstandard functions kbhit() or _kbhit() and getch(), will be your best bet. The Sleep() function (even specify just 1ms for the sleep duration) will take your CPU usage down from 100% to the 0-4% range, kbhit() lets you know when getch() won't block, so that you can keep updating if you receive a message.
If you were writing this as a Windows application on the other hand, there would be a much nicer solution: Since all input/output/etc. goes through the message loop anyway, use asynchronous winsock with WSAAsyncSelect() and the world will be all hunky dory ;) But in a console application there's no message loop, so that doesn't work for you.
Ohh my goodness - it worked this time. Holy crap - last time I tried Sleep it did didily squat.
Ill work on that select statement a lil - still a little confused - but if I reread it a few times it might sink in. Thanks a bunch! :-D
You could do it manually, using loops and flags and constant 'if' statements, but judging by your current problem, I doubt that's the route you want to take. You could also use SDL. It's typically thought of as a media library for gaming, but it does event handling and a number of other jobs too. Check it out at www.libsdl.orgQuote:
how would you setup events in console programs?
You could use something like this for maximum efficiency.
Code:HANDLE hEvents[2];
HANDLE hEvents[0] = CreateEvent(NULL, FALSE, FALSE, NULL);
HANDLE hEvents[1] = GetStdHandle(STD_INPUT_HANDLE);
WSAEventSelect(m_socket, hEvents[0], FD_READ);
/* At this point hEvents[0] will be signalled when socket data is available
* and hEvents[1] will be signalled when console input is available. */
while (recent != -1)
{
dwEvent = WaitForMultipleObjects(2, hEvents, FALSE, INFINITE);
if (dwEvent == WAIT_OBJECT_0)
{
// Socket data has arrived, call recv()...
}
else if (dwEvent == WAIT_OBJECT_0 + 1)
{
// Console input arrived...
}
}
Do you keep your engine revved-up to redline all the time? ...when you’re at a stoplight? OK, a bad analogy, ‘cause your processor is running full-speed all the time (unless it’s in a power-save mode).Quote:
But I never really liked the idea of my program sleeping.
But, you wouldn’t like it if Internet Explorer, or MS Word were sitting there doing nothing except taking CPU time away from your program.
Note that this is only because we’re in a multitasking environment. Your program should play-nice with others. In a non-multitasking environment, you’re free to use the CPU however you wish. In fact, in simple (i.e. embedded) systems, you may not even have an operating system. In this case, your program is required to manage (i.e. hog) the CPU full-time!
Thats exactly what I needed! I was looking on the msdn library about that function but microsoft always makes things seem much more complicated with documentation. Blah. I understood the timeframe and I wanted to use INFINITE because it will just hang there waiting for input. Thank you so much!! :-D
Hey dougdbug that anology was very well put actually. I can see your point. I suppose that the CPU is always querying hardware and filling buffers. What I ment was when you look at the processor usage that windows gives it usually 0% (if your not doing anything) - but then again thats only saying usage from processes. Sorry my bad - I should have clarified what I was trying to say. :-p.
>>You could use something like this for maximum efficiency.
Cool, I thought *maybe* something like that could be done, but figured it was just wishful thinking :D
Quote:
It's like bathrooms in the house. Most of the time they're empty, but everybody wants to use them at the same time.
Actually, this stuff doesn't need much CPU "power"... If your CPU monitor says 0%, then all that background stuff is taking less than 1%.Quote:
I suppose that the CPU is always querying hardware and filling buffers.
That automobile anology was bad, because the CPU never slows down, or stops execuiting instructions. When it says 0%, then your processor is "revving" full speed in a loop that has a bunch of conditional-branch statements (machine language if-statements). 99% (or more) of the conditional statements return "false", so it is running lots of do-nothing loops... ready to jump into action when needed.
The operating system will allocate CPU time to all running applications. If you are running only one application, then it's not so important for your program to "play-nice". But, you shouldn't use CPU time just to sit in a loop. By using Sleep(), you are telling the operating system that you aren't doing anything important, and other programs can have 100% for a while.
Oh no! Another analogy! :D
Say you go to a party, and there are three cookies for everyone. You're not hungry and you only eat one... You can either take 3 and throw 2 away, or leave the extra two for somebody who wants a couple extra. If you are the only one who shows up to the party, you can eat all you want, and throw some away without affecting anyone else.
Hahahaha nice illistration. (speaking of cookies im going to have some brownies soon!)
I never really realized that the CPU is runing at max. I came with my own theory that when data was available the hardware that got it (keyboards, mice, network cards, ect) send info to processor to be processed. So in essence it just sits there waiting for orders. But then again that is wrong since it really is the brain that governs many things...