Fix STM32H7 dual-core device files

[chris-durand]

• [x] Remove OPENAMP and RESMGR_UTILITY software IPs from device files for H7 dual core devices.
• [x] Fix Cortex-M4 core type (must be cortex-m4f, not cortex-m4fd)
• [ ] Fix memory layout for size g devices

[chris-durand]

I am trying to compile a blink example for a Nucleo H755ZI-Q. Compilation fails without the fix because the Peripheral enum has the entry OpenampCortex-m4 which is not a valid identifier.

[chris-durand]

Please don't merge yet. I have found another issue. There is no Cortex-M4 with a double-precision FPU but the device file has cortex-m4fd.

[chris-durand]

The memory layout is not correct for size g devices. The flash address space is not contiguous on those devices. Flash bank 0 ends at 0x807F FFFF whereas bank 1 starts at 0x8100 0000. The reason for this design choice is to have the same boot address for the M4 core at 0x8100 0000 on both 1MB and 2MB flash devices. It is intended to use one bank for each core. The boot addresses are configurable by writing option bytes, though.

[salkinium]

Then the flash memories should be separated and given to each core separately, the merging will then tag it per core.

[chris-durand]

Then the flash memories should be separated and given to each core separately, the merging will then tag it per core.

You mean we should assign a core to each half in the device file? Also for size i devices?

This would hard-code the default usage in the device file. There is nothing that prevents you from reconfiguring the core boot addresses in the option bytes or only using a part of a bank for one core. It's a user choice and not a property of the hardware.

I am fine with that limitation for modm but a potential external user of the device files might not. (Is there actually someone using modm-devices for another purpose that is not modm?).

[salkinium]

Good points. Then we need to split the Flash into two areas, not sure what will break in modm if anything.