Hi,
I have a couple of executables [ while(1) {...} loop programs within a main() function ] generated ( hex or bin or s19 ; for simplicity lets say hex ) and I want to allocate each hex file at a specific starting address in the flash memory, so that in a main program I can jump to a specific starting address depending on the input to execute the present code.
1. Any help to solve this problem ? 2. Is there any way to merge these hex files together and then I can can jump to their specific location ?
Thank you, Walid Farid
-------------------------------------- More about the problem ------------------------------
So, I've 3 developed projects where each project interfaces with specific peripherals on the development board. And, I want to load all 3 hex files ( or any other form of executable ) corresponding to each project to the flash memory and execute one of these codes at a time from the flash. I'm not sure why did you mention executing the code from the RAM, if I can execute it directly from the flash.
I thought, if I'm able to load the 3 hex files to the flash memory, I can load a 4th hex file, which given a specific input will run one of the 3 hex files at a time directly from the flash.
I'm using KEIL-MDK evaluation version and looking at the hex file, I can see that the starting/entry point of the code is 0x8000_00ED. So, I'm sure that all hex files will have the same entry point.
In KEIL-MDK, If I changed the code memory area to start at a different location than 0x8000_0000 ( say 0x8000_0100 ), will this change the starting/entry point ? Will the bootloader start executing the code at 0x8000_0100 which maps to the RESET routine? Or does the bootloader, regardless of the entry point in the hex file, places the code starting at 0x8000_0000 ?
I hope I clarified what I want to do ... so, regardless of or with regard to using hex files, how can I solve this problem ?
----------------------------------------------------------------------------------------------
P.S. : This question has already been posted on the STM32 forums
You shouldn't just mention the STM32 board but should post a link to the relevant thread.
Sorry, but it sounds meaningless to store multiple small applications like that.
Just have a small main() that looks at some GPIO signals to decide what to do:
function my_app1() { init1(); for (;;) { do_something1(); } } function my_app2() { init2(); for (;;) { do_something2(); } } function my_app3() { init3(); for (;;) { do_something3(); } } void main(void) { unsigned pins; pins = check_gpio_stimuli(); switch (pins & 0x03) { case 0: my_app1(); break; case 1: my_app2(); break; case 2: my_app3(); break; default: init_serial(); printf("Sorry - no such application\n"); } for (;;) ; }
The compilation time to build everything is no much longer than to compile one tiny app. The download time time is not much longer. The time saved from not having to jump between applications is well worth it. The ability to switch and run without reprogramming is well worth it. Too simple solution? Well, why make it more complicated than needed?
Well, the projects are not that simple to be all merged in one big project. We came across this idea, but loading multiple executables seemed to be more valid for what we wanted to do.
As I mentioned in my post, each project interfaces with specific peripherals on the development board and more often projects use the same sensors and same peripheral interrupt service routines. I think the one problem is due to the interrupt service routines and exception handlers which use the same address locations for each project. Can I change the vector addresses for each project ?
Is what I'm trying to do in general valid ?
The link for the post on STM32 forum is :
my.st.com/.../Flat.aspx
Thank you, Walid
The link is :
ttps://my.st.com/public/STe2ecommunities/mcu/Lists/ARM%20CortexM3%20STM32/Flat.aspx?RootFolder=/public/STe2ecommunities/mcu/Lists/ARM%20CortexM3%20STM32/Multiple%20Executables%20on%20the%20STM32%20flash%20memory&FolderCTID=0x01200200770978C69A1141439FE559EB459D758000626BE2B829C32145B9EB5739142DC17E¤tviews=50
Take a deep breath. You are trying to solve multiple problems with the wrong approach. You still did not make clear why you need separate binaries. If you do, you must probably use a bootloader of some kind. My project(s) use up to 4 components, but it has a good reason! Check out the ST material (application notes/user manuals) about remapping interrupt service routines. It sounds like you people have a number of parts that are foreign to one another, and you hope to glue them together by placing them on flash. That not how solid software is written and you skipped several stages anyway - how will these software components interact? You do realize that the last components to load will run in "user mode", while the previous ones will have to be executed by an interrupt service routine, unless you intend to explicitly jump around? Much better to have one application instead!
Probably not!
The whole point of a small embedded system is that it is dedicated to a single, specific, task.
If you have multiple, diverse tasks, it's probably better to just reprogram the target for each one.
Maybe you could adapt Per's suggestion, so that each of your projects is a Library - that way, each Library could have its own Project, and it would only be a very simple project to build them all into a single application...
Hard to read with posts in the middle. This is a threaded forum without the tree. Not sure why they introduced just one half of the minimum requirements for a threaded forum.
Anyway - some processors have a great VIC that allows a vector (address of ISR handler) to be programmed directly into the interrupt controller. Some does not.
But it is possible to create stub interrupt handlers that calls a function pointer to the real implementation. A bit slower but in many situations, the requirements does not need 100% maximum performance out of the processor. In some situations, the interrupt routing can be done in the assembler startup file to minimize the amount of double-saves from first saving a large state before entering an ISR and then saving even more registers before entering a standard C function.
The ARM compiler supports C++. This means that the functions and variables can be hidden in namespaces to avoid colliding names. This concept obviously only works if the target has enough RAM for all global variables for all sub-applications. If it doesn't, then some poor soul may have to move the global variables for each application into a struct, and then place the multiple structs into a single union (or maybe allocate dynamically on startup), so the code in main supplies a pointer to a struct of the correct type.
Life is definitely easier if the globals fits at the same time - maybe just manually create optimized declarations for larger buffers to let them overlay each other depending on what target sub-application that gets started. In real life, a few variables tends to consume the majority of the RAM while all other variables are tiny char, short and int variables.
If the applciations aren't huge, and you are skilled in the use of one of the better programming editors, you could convert even non-trivial applications into becomming sub-applications with reasonable amount of work.
One interesting thing with multiple applets run from a common main - if one produces output data that needs to be further processed, it could even be possible to run the applets in sequence as a form of pipelined filter.
View all questions in Keil forum