Posted by bugmenot on 2017-05-24 12:41

In STM32 you rarely need to run code from RAM because the pros are very limited and there are cons. In more than 10 projects I used running code from core-coupled (CCM) RAM ( http://www.openstm32.org/Using+CCM+Memory?structure=Documentation ) just once.

ARM Cortex-M uses the Harvard architecture with separate code and data buses, when you are running code from RAM you are losing the speed advantage of this architecture as now you have both your code and data in the RAM.

The core provides other means to workaround the slow flash access time: You should carefully read all the available documentation (programmer’s manual, datasheet, reference manual) then configure the bits for pre-fetch, pre-read, wait-time, buffers, caches, etc. according your core frequency and voltage scale to have maximum possible performance or energy efficiency (or something between).

Posted by pandreuzza on 2017-05-25 08:29

This is the first time for me on this architecture.
Thank you for your advices, now it is more clear...
I’ll go through the documentation to get maximum performance running code from flash memory.