Thread: cpu clock cycles

this may not be the best forum to post this in, but i have had great results posting comments here....thank you everyone for your help

what i am trying to figure out is how many clock cycles are in a second. My power settings for the device (the PSoC micro) is 5v at 24 mega herts but the cpu_clock is sysclk/1 which is just a setting i have it set to over clock it slightly

so in a for loop if i had

Code:

for(i=0; i<1000;i++)
{....
}

how many seconds would be in that 1000 cpu clock cycles?

A megahertz is 1,000,000 clock cycles per second...

Thus 24mhz is 24,000,000 clocks per second.

But here's the bugaboo in your idea... a for loop does not increment on clock cycles, it can take many clock cycles just to do the i++ part since it has to fetch the variable from the stack, increment it, put it back... then there's the comparison i<1000 which is not done in a single clock cycle either... Then of course there's whatever's going on inside the loop... Each turn through the loop could take dozens or hundreds of clock cycles and there's no way of being certain how many...

I suggest you check your uC documentation to see if there's a precision counter/interrupt you can use instead of relying upon software for timing...

Obviously if you have a 24MHz CPU with 1:1 CPU clock, there are 24 million cycles in one second. That's the definition of hertz.

It's impossible to tell how many cycles the loop will take. You can only do instruction level timing in assembly, and even then only on non-superscalar architectures.

The compiler will compile your code into any sequence of instructions that will achieve the effect. It's all up to the compiler.

NO ASSEMBLY FOR THIS GUY i took a class that using the HC112 micro in all asembly and about shot my self. i was just thinking there was going to be an easy way to tell how many cpu clock cycles it would be

what i am really doing is controlling different leds with the rising and falling edge of the clock cycle and i need to know in my code how many are on per second so that i can calculate the power output of them

Originally Posted by brewbuck

Why not just loop some huge number of times (say, 500 million iterations) and time it? With a stopwatch, if necessary. I've timed many a thing with a stopwatch.

And probably gotten well within 20% accuracy...

(just kidding)

i hate to go on a tangent about this, but here is what i am thinking, both of your points make tremendous sense but here is some of my code and though process...please let me know if i am on the right track about thinking about this....i am studying electrical engineering, and the programming half of it just don't not make any sense...so like i said before i am controlling leds on the rising edge and falling edge of the clock(i need to figure out power of the leds, so i need to know how long they are on for, which is the point of the thread) so my for lop runs though the pattern and inside the for loop i have the leds on separate counters that interrupt once, twice, four and eight times a second

so for this for loop

for (int i; i<1000; i++)
{
if(G_Flag1)
{
led1_on;
led2_off;
}
else
{
led1_off;
led2_on;
}

void Counter16_ISR(void)
{
G_Flag1= !G_Flag1;
num1=rand( );

}

this interrupt interrupts four times a second so would led1 and led 2 both be on for four times a second, depending if the clock was on a rising edge or falling edge right, or could they be on longer because the cpu clock is still less then 1000

Originally Posted by begginer

i hate to go on a tangent about this, but here is what i am thinking, both of your points make tremendous sense but here is some of my code and though process...please let me know if i am on the right track about thinking about this....i am studying electrical engineering, and the programming half of it just don't not make any sense...so like i said before i am controlling leds on the rising edge and falling edge of the clock(i need to figure out power of the leds, so i need to know how long they are on for, which is the point of the thread) so my for lop runs though the pattern and inside the for loop i have the leds on separate counters that interrupt once, twice, four and eight times a second

Provide more details:
- what oscillator frequency is input to the microcontroller.
- are the leds activated on both falling / rising edges.
- width of hardware interrupt timers (8/16/32 bit).
- and post the schematic if you have one.

Originally Posted by begginer

so for this for loop

this interrupt interrupts four times a second so would led1 and led 2 both be on for four times a second, depending if the clock was on a rising edge or falling edge right, or could they be on longer because the cpu clock is still less then 1000

Depends on what the interrupt is setup to do, does it activate LEDs whenever an interrupt is generated?

Originally Posted by CommonTater

Look at his logic, there are 2 leds... The width of the pulses is irrelevent, the time between pulses doesn't matter

It matters because the timer overflow is related to the frequency of the internal clock.

Originally Posted by CommonTater

how fast his loop runs makes no difference... Even the schematic is irrelevent (so long as it actually works).

Correct.

Originally Posted by CommonTater

One LED is on while the other is off... thus there is always 1 led on... determining the power usage is as simple as calculating the load of 1 led...

Plenty of assumptions that need to be clarifed with more data, imo.

Originally Posted by CommonTater

Determining how long each is on for is equally simple... 50% of the time... because the flag he uses toggles on each interrupt.

We're on the same page regarding that.

Originally Posted by CommonTater

Matter of fact he doesn't even need the for() loop...

Depends on how the whole thing's been coded.

Originally Posted by cyberfish

The frequency IS important. LEDs have a maximum frequency because the depletion region at the PN junction needs time to establish and collapse. IIRC, the frequency limit is actually quite low, unless the LED is properly biased (always kept around the switching voltage). Especially true if you are driving with a high impedance source (like microcontroller output).

From the op...

separate counters that interrupt once, twice, four and eight times a second

I don't think it's much of an issue... most LEDs are good to at least a kilohertz.

The question is how fast do you need to blink the LEDs? Slower than 18 Hz? Systems have a regular interrupt happening... on most machines something like 55ms. I read "The Win95/98 timer ticks every 55 ms and the Win NT's every 10 ms." On a micro controller it's configurable some place.
So you need to tap into that interval and use it as a basis for timing.