In this lab, you will implement your own dynamic memory allocator (heap manager) in C, using a free list with rst- t block selection.

Implementation

You must follow these implementation guidelines:

1. De ne a struct to represent an allocation block.

struct Block { int block_size; // # of bytes in the data section

struct Block *next_block; // in C, you have to use struct Block as the type

};

2. Determine the size of a Block value using sizeof(struct Block) and assign it to a global const variable. Well refer to this value as the overhead size.
3. Determine the size of a void* and save it in another const global. Well refer to this value as the pointer size .
4. Create a global pointer struct Block *free_head, which will always point to the rst free block in the free list.
5. Create a function void my_initialize_heap(int size), which uses malloc to initialize a bu er of the given size to use in your custom allocator. (This is the only time you can use malloc in the entire program.) Your global free_head should point to this bu er, and you should initialize the header with appropriate values for block_size and next_block.
6. Create a function void* my_alloc(int size), which lls an allocation request of size bytes and returns a pointer to the data portion of the block used to satisfy the request.
   • size can be any positive integer value, but any block you use must have a data size that is a multiple of your pointer size. So if a pointer is 4 bytes, and the function is told to allocate a 2 byte block, you would actually nd a block with 4 bytes of data and use that, with 2 bytes being fragmentation.
   • Walk the free list starting at free_head, looking for a block with a large enough size to t the request. If no blocks can be found, return 0. (null) Use the rst t heuristic.
   • Once you have found a block to t the data size, decide whether you need to split that block.
     1. A block needs to be split if its data portion is large enough to t the (rounded-up) size being allocated, AND the excess space in the data portion is su cient to t another block with overhead and a minimum block size of the pointer size. Example: if overhead size is 8 and pointer size is 4, we should split if the remaining space (after the allocation size is accounted for) is at least 12 bytes.
     2. If you cannot split the block, then you need to redirect pointers to the block to point to the block that follows it, as if you are removing a node from a singly linked list.
     3. WARNING: the logic for removing a node in a linked list is di erent depending on whether or not the node is the head of the list. Draw it out and convince yourself why you need to account for this.

• If you can split the block, then nd the byte location of where the new block will start, based on the location of the block you are splitting and the size of the allocation request. Initialize a new struct Block* pointer to that location and assign its new block_size. The new blocks next_block pointer needs to point to the same block as the next_pointer of the block you are splitting. Reduce the size of the original block to match the allocation request.

• Return a pointer to the data region of the block, which is overhead size bytes past the start of the block. Use pointer arithmetic.

7. Create a function void my_free(void *data), which deallocates a value that was allocated on the heap. The pointer will be to the data portion of a block; move backwards in memory to nd the blocks overhead information, and then link it into the free list.

Pointer Arithmetic

In C, one can add integer values to pointers to move that pointer forward or backward in memory from its current position. You will need to do this when splitting blocks: to determine the start of the new block, you will start at the current block, then move forward by enough bytes to skip the old blocks data (the size of the allocation request) plus the old blocks overhead. However, we must note that adding an integer x to a pointer does not move the pointer by x bytes; it moves it by x multiples of the data type the pointer points to.

Example: suppose int *p currently points to memory address 1000. The expression p+1 would point to address 1004, not 1001, assuming int is 4 bytes. Example 2: suppose struct Block *temp points to address 1000. p+2 would point to address 1032, assuming the size of struct Block is 16 bytes.

This is most likely not what you want; you know an exact number of bytes that you want to move forward in memory by, not the number of multiples of X . To take a pointer to an arbitrary type and move it forward by x bytes, we temporarily convert that pointer to a char*, then move the pointer, then cast it back to the expected type. Since a char is always 1 byte in C, adding x to a char pointer will move the pointer forward by x 1=x bytes.

Example: to move int *p forward by 11 bytes, we would do p = (int*)((char*)p + 11). In other words: treat p like a char pointer, increment it by 11 multiples of a character (11 bytes), then reinterpret it as an int* as intended. We could then dereference p to retrieve a 4 byte value at the destination address. ^[1]

Testing Your Code

Test your code thoroughly by allocating values of various types. You should write (and turn in) your own testing main, which at least includes the following tests, each a separate branch of main so that only one runs per execution of the program:

1. Allocate an int; print the address of the returned pointer. Free the int, then allocate another int and print its address. The addresses should be the same.
2. Allocate two ints and print their addresses; they should be exactly the size of your overhead plus the larger of (the size of an integer; the minimum block size) apart.
3. Allocate three ints and print their addresses, then free the second of the three. Allocate an array of 2 double values and print its address (to allocate an array in C, allocate (2 * sizeof(double)); verify that the address is correct. Allocate another int and print its address; verify that the address is the same as the int that you freed.
4. Allocate one char, then allocate one int, and print their addresses. They should be exactly the same distance apart as in test #2.
5. Allocate space for a 80-element int array, then for one more int value. Verify that the address of the int value is 80 * sizeof(int) + the size of your header after the arrays address. Free the array. Verify that the ints address and value has not changed.

[1] Dont actually do this. Moving an int pointer by 11 bytes will cause it to no longer be aligned to a multiple of 4 bytes, and derefencing it will give a super-fun segmentation fault error.