Download CS232 Discussion 3: Pointers, Structures, and Endianness and more Study notes Object Oriented Programming in PDF only on Docsity!
1 Pointers and Structures
So far we have talked about arrays, which are a collection of objects of the same type stored in a contiguous chunk of memory. A structure is also a collection of objects stored in a contiguous chunk of memory, but these objects are not necessarily all of the same type. Consider the following structure in C:
typedef struct node_s { int * data; // pointer to an integer struct node_s * next; // pointer to another struct node_s } node_t; // node_t is shorthand for "struct node_s"
The above structure can be used to implement a singly-linked list of nodes (where each node has type node t) as follows: the head node is the first node of the list, its next field contains a pointer to the next node in the list, etc. The list is terminated with a pointer to NULL. Recall from lecture that pointers are just memory addresses at the machine level. Below is C code for a function that increments each element of a singly-linked list by a given value. The variable head contains a pointer to the first element of the list.
void increment(node_t * head, int value) { for(node_t * trav = head; trav != NULL; trav = trav->next) { *(trav->data) += value; } }
Write increment as a MIPS funtcion.
2 Endian-ness
There are two possible ways to store multi-byte data words (e.g., 32-bit integers) in memory; in both cases when a word is stored to address N , the bytes are stored in byte addresses N, N+1, N+2, and N+3. Little endian stores the least significant byte (LSB) in the first byte (N, i.e., little end first), while big endian stores the most significant byte (MSB) in the first byte (N, i.e., big end first). For example, if storing the value 0xdeadbeef to address 0x10000000, you get the following: Both conventions are in active use by popular microprocessors: Alpha and x86 are little endian,
ADDRESS: 0x10000000 0x10000001 0x10000002 0x Little Endian 0xEF 0xBE 0xAD 0xDE Big Endian 0xDE 0xAD 0xBE 0xEF
SPARC and PowerPC are big endian. MIPS is actually bi-endian (i.e., it can be run in either mode) and SPIM (somewhat sadistically) uses the endianness of its host. That is, SPIM is big endian on a SPARC machine and little endian on an x86. As long as you don’t cast one data type to another, you shouldn’t have to worry about whether you are running on a big-endian or little-endian machine. That is unless you need to commu- nicate over the network. When two computers communicate, they must ensure that numbers are correctly interpreted even if one machine is little endian and the other is big endian. For example, remote procedure call (RPC) defines big endian as the canonical form (not surprising since Sun defined it and SPARC is big endian). Thus, little endian machines need to “swizzle” their bytes to big endian before sending them across the network.
unsigned swizzle_word(unsigned in) { unsigned temp = in >> 24; temp |= ((in >> 16) & 0xff) << 8; temp |= ((in >> 8) & 0xff) << 16; temp |= (in & 0xff) << 24; return temp; }
For your reference, here is the same function in MIPS assembly:
swizzle: srl $t0, $a0, 24 # v0 = (in >> 24); and $v0, $t0, 0xff
srl $t0, $a0, 16 # v0 |= ((in >> 16) & 0xff) << 8; and $t0, $t0, 0xff sll $t0, $t0, 8 or $v0, $v0, $t
srl $t0, $a0, 8 # v0 |= ((in >> 8) & 0xff) << 16; and $t0, $t0, 0xff sll $t0, $t0, 16 or $v0, $v0, $t
