msp430 cycle time problem

[steve.hunch]

Hi ,
I use eval board "LaunchPad" I set CLK to 1MHz.
When I try to check the maximum on-off rate on one of the pins I get 10usec - 100KHz.
Doing just:

while(1)
{
P1OUT = 0;
P1OUT = 0xff;
}


The problem is that system clock is 1MHz, but this code gives my only 100kHz.
What can be a problem?

Changing clock to 10MHz, I get 1MHz at the I/O.

 

[andre_luis]

If you disassemble this C program into Assembly code, maybe will obtain the answer.
Each instruction and statement of code execute specific funcion and consumes instruction time.

+++

 

[steve.hunch]

I did this, but apparently each each of those instructions disassembles to one assembler instruction. This means that I should get, at least for logical '0' approximate 1usec time (with 1MHz system clock).
But I still get 10usec. There is some factor of 10 in instruction execution...

 

[andre_luis]

while(1) -> Instruction that compare, and skip
{
P1OUT = 0; --> instruction that set output
P1OUT = 0xff; -> instruction that set output
} -> instruction that jump

Maybe there are other operations.
Try change optimization options.

+++

 

[Sink0]

No .. actually you are getting a good toggle speed on you output, but for sure with no 50% duty-cycle. Toggling a port on MSP430 does not take 1 instruction cycle. It can take 3 to 5 cycles, dpending from where your data is coming, and you got the added jump cycle from your while (i think that takes 3 cycles). Some other architectures have some special pin manipulation hardware. Always read the manual before chosing your architecture.. i always tell that to people that thinks that one architecture is perfect (as many does with ARM as an example). With a 50% duty cycle you would have to use xor instruction, and that woud take a total of 14 cycles for a full positive and negative edge. 3 from jump, and 4 from xor. There is no optimization that can handle hardware limitation.

Cya and GL

 

[steve.hunch]

Thank you!