Setting up shared memory between two cores

Kernel
1

1/ I have ordered the Discovery kit with STM32H745XI MCU. I do not have it yet.

2/ I am looking at the CORTEX_M7_M4_AMP_STM32H745I_Discovery_IAR example and can follow the instructions and have built both executables.

3/ I would like to define an unmanaged COMMON block that’s shared between the two processes, core 1 writes it, core 2 reads it; and I will use the message queue to send the ready flag.

Q1: In MessageBufferLocations.h I see the static declaration of the message buffers at #pragma location 0x380000200, so my first instinct would be to compute the offset of my shared memory and set it up as a static uint8_t big_buffer[1024]. Does this seem reasonable?

Q2: I have read thru other questions, so I’ve learned I should do a cache flush on the writer, but other than that is it as simple as this?

Q3: Is it in the HAL that the memory map is defined such that memory allocations “know” to avoid reserved blocks in the SDRAM?

Q4: I’m a 10+ year VxWorks user, so I’m used to a different way of doing things. So bear with me.

Thanks to anyone who helps!!

(Again, no hardware til Thursday, but I’m trying to understand)

2

You will need to edit the linker script so the block you want to share between the cores is not known to either core - that way the linker won’t use it for anything and you can have free access to it. As to whether that is done in the HAL or not - if the linker script is in the HAL then the answer would be yes, but I’m not sure where the linker script is. Sometimes IAR projects use a single central linker script for the device as the device is known to IAR - however in that project the linker script has already been messed with to place the message buffers in memory so I assume the linker script is already part of the project.

3

Thank you.
I am pouring over the project files; and there are a LOT of files!
I may be tackling to much… multi-core MPU, new RTOS, new IDE…