Copy a function from SRAM to flash and execute from flash

You may have to settle for a pointer table to your functions.

Position-independent code can be moved. But moved as a complete block - not move a single function.

Function calls are either relative - in which case both the source and target functions must be moved the same number of bytes to make sure the relative jump is correct.

Function calls can be indexed relative a pointer - but same there. The assumption is that the pointer points to the start of the block, and both source and destination function have a fixed offset within the block.

Function calls can be absolute - in which case the linker script specifies the memory region the functions should use. Note that the linker script can have one address range where the code is stored. And a different address range where the functions should be used. So the linker can link the functions for use in RAM, but store the code in flash. The instructions can then be copied into RAM before the functions are called.

If the code is built with absolute addressing, then you can't have the function usable from both flash and RAM, unless you have a processor capable of remapping memory pages so flash or RAM can be mapped on top of each other.

If the code is built with relative addressing, then the full block must be copied flash->RAM.

You may have to settle for a pointer table to your functions.

Position-independent code can be moved. But moved as a complete block - not move a single function.

Function calls are either relative - in which case both the source and target functions must be moved the same number of bytes to make sure the relative jump is correct.

Function calls can be indexed relative a pointer - but same there. The assumption is that the pointer points to the start of the block, and both source and destination function have a fixed offset within the block.

Function calls can be absolute - in which case the linker script specifies the memory region the functions should use. Note that the linker script can have one address range where the code is stored. And a different address range where the functions should be used. So the linker can link the functions for use in RAM, but store the code in flash. The instructions can then be copied into RAM before the functions are called.

If the code is built with absolute addressing, then you can't have the function usable from both flash and RAM, unless you have a processor capable of remapping memory pages so flash or RAM can be mapped on top of each other.

If the code is built with relative addressing, then the full block must be copied flash->RAM.

I forgot to ask - but exactly _why_ do you want main() to decide at runtime if it should call a specific function in RAM or in flash? What problem do you hope to solve? You are introducing lots of problems with this specific requirement.

Hi,

Thank you very much for your quick answer:-)
So to better explain lets' say I have the function f calling functions g and l:

void f (void)
{
 call g;
 call l;
}

all these functions are in SRAM, then I copied f in flash but I want that g and l are still in SRAM when I execute the code from flash. How can I do that?

Alex

No, that doesn't address the question at all:

"exactly _why_ do you want main() to decide at runtime if it should call a specific function in RAM or in flash? What problem do you hope to solve?. You are introducing lots of problems with this specific requirement."

www.catb.org/.../smart-questions.html

Hello,

This is a customer application and this requiremets was given to me by my customer.
The customer wants the possibility to have a function in sram and the same in flash and to decide when to use when or the other.

Alex

Easy - make two copies of the function. One linked for RAM and one linked for flash.

Good solution? Not at all. But the customer gets what he wants - and gets to pay the price from wanting what he shouldn't.