Control Suite GPIO_WritePin functions

Hi Benjamin,

I don't think there's any specific reason TI was trying to avoid using macros for this. However, functions are a little easier to read and debug, so perhaps for demonstration/teaching purposes, macros weren't the preferred option.

There's no doubt that writing it as you have is much faster. Thanks for sharing it with the forum.

Whitney

As Kris mentioned, "GPIOx" is actually a single bit in a register, not a memory location with a unique address. In your case GpioDataRegs.GPASET is actually a 32-bit register. You can access this register by dereferencing its associated address. The GPIOs, however, are the 32-bits that make up the register itself. So the reason you cannot dereference them individually is because they all share the same memory address. If you are trying to set the GPIOs quickly, you can set any combination of GPIOs in the same register simultaneously (e.g. GpioDataRegs.GPASET.all = 0xA34F9384). If you need this to execute in a for loop with an index, then I would suggest using macros.

Exactly what Benjamin said. When you do a ".bit.GPIOx" write what happens under the hood is the compiler turns this into a Read Modify Write instruction at the GPASET address (for example). This involves generating a mask for the bit you are trying to write ( based on the bit position ) and writing it back to the whole GPASET register. The most efficient method is to write the whole GPASET register as Benjamin demonstrated. The difficulty with this in your application is that the GPIOs are in different ports (GPASET, GPBSET, etc) so you would need to maintain a value for each of these. It's certainly possible- just not trivial. I suggested the driver method just for simplicity, but if it's time critical it is probably better to develop a more complex routine.

Just thinking through this for a few minutes, one thing you could probably do is an array of GPxSET registers. I promise this syntax isn't correct, but just for the general idea:

my_array[] {&GpioDataRegs.GPASET, &GpioDataRegs.GPBSET, &GpioDataRegs.GPCSET, etc....}

Then in your code you could divide your GPIO number by 32 to figure out the port (0 -> A, 1->B, 2->C) and get the array index. Then you could use the remainder of the GPIO number divided by 32 (this will tell you the bit position of your desired GPIO, so shift a '1' that number of places) to calculate a value to write to the my_array[index].

That sounds complicated, but it's a pretty small number of instructions. *Disclaimer, this is just a theory :)

Regards.

Kris

Hi Kris, Hi Benjamin,

you are great, thanks!

Both your posts helped me further to understand the information Ti provides (which is excellent, if you can see the things you are looking for).

Since I am new in this task of programming micro-controllers, every step is a kind of space trip :)

Following your remarks, I will for the time being simplify my use of the ports. In this way, I could solve the issue by creating an array with the pin numbers in use as hex values and talking to the GPIOSET.all setting  the pins as needed, or clearing them with GPIOCLEAR.all. The experiment you suggested, Kris, I will sleep long nights over it, and will come to you later to tell you how it went.

Just to make sure, my code would do this, using pins GPIO0 to GPIO5, GPIO16 and GPIO24, all in the A section:

Uint16 myGPIOpins[8] = {0x1,0x2,0x4,0x8,0x10,0x20,0x10000,0x1000000};

// or like this, might be ?
// Uint16 myGPIOpins[8] = {1b,10b,100b,1000b,10000b,100000b,10000000000000000b,1000000000000000000000000b};

void SetAllPins()   // to set the pins, and the same in another function for GPACLEAR
{

    Uint16 j;

    for(j=0; j<8; j++)

    {

        GpioDataRegs.GPASET.all = myGPIOpins[j];

    }

}

I know, this is far from inventing the Wheel, but my father always told me not to do it.

Many, many thanks and a very nice Sunday for you!
Gustavo