Atmel CAVR-4 Manual De Usuario

CAVR-4

Part 1. Using the compiler

Memory for objects allocated on the heap will live until the objects are explicitly 
released. This type of memory storage is very useful for applications where the amount 
of data is not known until runtime.

In C, memory is allocated using the standard library function 

malloc

, or one of the 

related functions 

calloc

 and 

realloc

. The memory is released again using 

free

.

In C++, there is a special keyword, 

new

, designed to allocate memory and run 

constructors. Memory allocated with 

new

 must be released using the keyword 

delete

.

The AVR IAR C/C++ Compiler supports having heaps in tiny, near, far, and huge 
memory. For more information about this, see The return address stack, page 43.

Applications that are using heap-allocated objects must be designed very carefully, 
because it is easy to end up in a situation where it is not possible to allocate objects on 
the heap.

The heap can become exhausted because your application simply uses too much 
memory. It can also become full if memory that no longer is in use has not been released.

For each allocated memory block, a few bytes of data for administrative purposes is 
required. For applications that allocate a large number of small blocks, this 
administrative overhead can be substantial.

There is also the matter of fragmentation; this means a heap where small sections of free 
memory is separated by memory used by allocated objects. It is not possible to allocate 
a new object if there is no piece of free memory that is large enough for the object, even 
though the sum of the sizes of the free memory exceeds the size of the object.

Unfortunately, fragmentation tends to increase as memory is allocated and released. 
Hence, applications that are designed to run for a long time should try to avoid using 
memory allocated on the heap.