Hello,

I'm using Allegro 4.2. The readkey() routine uses 100% of the CPU, it seems like it's doing active waiting for a key press. Is there any way to avoid it (different routine?). Maybe I'm doing something wrong?

Thanks,
Michal

Makes sense. readkey() is a blocking function and Allegro 4 doesn't ever natively yield by itself, IIRC. If you don't want to eat the CPU but still block on input, just add a loop checking if there's a key press before calling readkey():

while (!keypressed()) {
   sleep(0); // yield the CPU however you want.
}
int keycode = readkey();

The readkey() function apparently asks the OS over and over for a keypress, so naturally it hogs the CPU.

Let's say this runs on a 1Ghz cpu.

int keycodes[BUFFSIZE];
int i;

for(i=0;i<BUFFSIZE;i++)
{
     while (!keypressed()) {    //Wild guess, this takes 100 machine cycles
        usleep(1000); // 1/1000 of a second means 1Ghz rests for 1000000 machine cycles
     }
     keycodes[i] = readkey();   //Another wild guess, this takes 1000 cycles
} //end of i loop

So readkey actually fetches a key 10 times instead of spinning without resting between keys and then getting a key, the keypressed() and usleep "spin", but with a rest in there.

When Allegro was first conceived back in the middle of the '90's, just about all programs hogged the CPU, MS-DOS didn't multitask, and those 50Mhz 486's didn't get all that hot. There was a company that made electronically cooled CPU's sometime around there (think of a thermocouple in reverse) and when somebody ran Linux on it, which slept when all the tasks were done, it actually formed frost on it, and when a program warmed the CPU up again, the frost melted and shorted out the motherboard!

Wha? No way! I'm not buying that story!
...Please tell me it's true, pleeeeease.

But yeah, without having a cycle that's always asking, and resting for a few milliseconds to not hog the CPU, there's no other way to do it without events, which is why Allegro 5 is so highly recommended.