How it works

When a NEW or GETMEM statement is invoked, the allocation routine searches a block in this order of priorities:

• A block that matches exactly the required size.
• A block that is greater, in this case it will be split as required (see below).
• After heap defragmentation, the two first steps are tried again; if they don't match again, the ERROR procedure is called and the processor is stopped in an infinite loop.

When DISPOSE or FREEMEM is invoked, the given block is simply linked into the heap; it becomes the first free block.

To minimize the heap fragmentation, the memory blocks are allocated by chunks with a granularity of 4 bytes. The actual block size is the required size plus the memory needed for maintaining a “block size” at the beginning of the block (one byte for PIC16, two bytes for PIC18). If a block is greater than the required size rounded to the next 4 bytes boundary, it is split and the remaining bytes return to the heap.

• Heap is initialized once at main program start-up.

Heap block memory format:

• <block-size>: one or two bytes.
• <next-block-address>: two bytes.
• <data>: n bytes of free memory.

<block-size> is maintained in memory; the pointer returned by NEW or GETMEM is the address next to <block-size>.

When the memory block is freed by DISPOSE or FREEMEM, the given pointer is decremented to skip back to <block-size> and a consistency check is made between the size argument and the current block size; an error is generated if they don't match.