STM32H743 FMC with ST7789 doesn't work without delay

[UKatz.1]

I have an STM32H743 connected to a screen that uses ST7789 via FMC (i.e. parallel connection) using 8-bit bandwidth and A18 to select between command and data.

The current code has an issue that the screen output is messed up (see pictures below) if I don't add a sleep to my write functions that uses FMC.

The code is working (no need for delay) under STM32CubeIDE but I ported it to work as part of MicroPython.

Things that I already considered:

1. Python speed - The code in C is a nested for loop and calls writeData so it isn't a python speed thing.
2. System clock/FMC - I copied the whole code of FMC initialization (including Msp), GPIO and systemclock to MicroPYthon without any fix.
3. HAL version - MicroPython uses a different stm32 HAL version but I copied the SRAM/FMC files and it still didn't work

Any pointer on what might be the issue? Is it a FMC clock configuration? some configuration on external memory location?

Any config that slows down the FMC as RAM access

Appendix:

#define COMMAND_LOCATION 0xC0000000
 
#define DATA_LOCATION 0xC0040000
 
static uint8_t* fmc_command_pointer = (uint8_t*) COMMAND_LOCATION;
 
static uint8_t* fmc_data_pointer = (uint8_t*) DATA_LOCATION;
 
 
 
void writeCommand(uint8_t command)
 
{
 
    *fmc_command_pointer = command;
 
    HAL_Delay(100);
 
}
 
//*****************************************************************************
 
void writeData(uint8_t data)
 
{
 
    *fmc_data_pointer = data;
 
    HAL_Delay(1);
 
}

Filling the screen with the color blue without the delay in writeData produces: 

 

With the delay (after a while since it takes 1ms per pixel) fills it correctly: 

[UKatz.1 · Accepted solution]

Found the issue, it was a missing volatile on the pointers to the FMC memory or as in STM32HAL syntax __IO - the code should be:

static __IO uint8_t* fmc_command_pointer = (__IO uint8_t*) COMMAND_LOCATION;
static __IO uint8_t* fmc_data_pointer = (__IO uint8_t*) DATA_LOCATION;

Now why this works in my STM32CubeIDE is a mystery, probably a gcc compile flag