Using the calloc function on an STM32F4

[rajpaged]

Posted on July 15, 2014 at 17:12

I'm trying to understand how this function should be used on an F405 chip. The code that calls calloc is all generated from a simulink fuzzy controller model so I'm confident it's correct in terms of code.

The program seems to be ending up in the default handler as soon as it hits a calloc function. Do I need to sort out a heap for this or something? Not really sure how it works #calloc-stm32f4

[Tesla DeLorean]

Posted on July 15, 2014 at 17:52

Yes, you'd need to address the size of the heap, however your particular tool chain deals with it. RTM

[rajpaged]

Posted on July 15, 2014 at 19:00

I think I got past that issue, I added this into the arm-gcc-link.ld

/* heap section */    .co_heap (NOLOAD):    {      __cs3_heap_start = . ;         . = ALIGN(8);        *(.co_heap .co_heap.*)     __cs3_heap_end = . ;     } > ramThis into startup_stm324fxx.c#define HEAP_SIZE       0x00000200__attribute__ ((section(''.co_heap'')))unsigned long pulHeap[HEAP_SIZE];extern unsigned long __cs3_heap_start;extern unsigned long __cs3_heap_end;and this into my syscalls.c__attribute__ ((used))caddr_t _sbrk (int size){   extern char __cs3_heap_start;   extern char __cs3_heap_end ;   static char *current_heap_end = &__cs3_heap_start;   char *previous_heap_end;   previous_heap_end = current_heap_end;   if (current_heap_end + size > &__cs3_heap_end )   {      errno = ENOMEM;   // don't forget to include <errno.h>      return (caddr_t) -1;   }   current_heap_end += size;   return (caddr_t) previous_heap_end;}Now my code ends up in the default handler later on when assigning a value to one of the matlab generated structures. Is there a way I can determine how much memory it's all using from within GDB?

[Tesla DeLorean]

Posted on July 15, 2014 at 19:51

Can you not gauge the size and number of allocations the code might be expecting to use?

512 bytes won't go far if you have a 1024 array of 64-bit doubles You could instrument calloc(), or a proxy, and quantify what's going on. Is the machine generated code not looking for NULL returns from calloc()/malloc() ? You could supply a Hard Fault handler, or other handlers, to more specifically identify the cause of the errors/faults.