TMS320F28388D: Problem Clearing Bits with Bit Mask in the GPADAT Register

Part Number: TMS320F28388D

Tool/software:

I use bit masks (my preference) to set bits in the TMS320F28388D microcontroller registers which works fine everywhere except that I am having problems clearing bits in the GPADAT data register; specifically, I can easily set specific bits using GPADAT  |=  aValue  but when I try and clear specific bits using GPADAT &= clearMask bits with 1 will not clear to 0.

I think these values are correct with:

GPADAT   defined as  #define GPADAT (*((volatile Uint32 *)(GPIO_DAT_BASE_ADDRESS + GPADAT_OFFSET)))

with #define GPIO_DAT_BASE_ADDRESS 0x00007F00 and #define GPADAT_OFFSET  0x00000000 (just at the base address, but I use this form to facilitate setting up for other registers)

and with the value (aValue) being 4 bits in GPADAT bits 24 - 27 set up as aValue = aValue << 24

and with the mask (clearMask) to clear the positions before the OR'ing operation = 0xF0FFFFFF

so that I

aValue = aValue << 24 ; // To position the bits correctly

GPADAT &= clearMask;  // To clear the respective bit positions in GPADAT

GPADAT |= aValue ; // To write the bit values to the respective positions in the data register

What happens is that when I start out with a zeroed GPADAT register, the ORing operation correctly writes aValue into the GPADAT register.

But after that, bit positions with a value of 1 don't get reset, by the AND operation but bit positions with 1 get writen by the OR operation so that eventually the 4-bit value is just 1111  (0xF).

I heard something about a clear-modify-set problem with the data register, so I tried substituting GPADAT &= clearMask; with GPACLEAR = clearMask2

where GPACLEAR  is  #define GPACLEAR (*((volatile Uint32 *)(GPIO_DAT_BASE_ADDRESS + GPACLEAR_OFFSET))) with #define GPACLEAR_OFFSET  0x00000004

and that did not work either (did not clear the 4 bits).

Finally, I got this to work by clearing each bit individually with this:

mask = 1;

GPADAT &= ~(mask << position) ;

But I would prefer to be able to just do it with a single mask in a single operation as i tried to do above.

Anybody see the problem?   Thanks.