Lecture 021

Undefined Behavior for Numbers

8-bit	16-bits	32-bits	64-bits
256 bytes	64 KB	4GB	2^{64} bytes
+-128	+-32768

C standard: compiler define size of an int (implementation-defined behavior)

• programmer can find out size by INT_MIN, INT_MAX in limits.h

• but representation of pixel as int need specific size of integer

int's Undefined Behaviors

1. Safety violation in C0

   • division / modulus by 0
   • INT_MIN by -1
   • shifting more than size of int

2. Overflow!

   • compiler might not use two's complement
   • impossible to reason about int
   • gcc: -fwrapv force use of two's complement

long: 64 bits short: 16 bits char: 8 bits (1 byte) unsigned variations: always follow modular arithmetic

size_t: size of a pointer

• argument of malloc() calloc()

• array indices

• return type of sizeof()

Casting Integers

Literal numbers have type int

• compiler will do implicit casts if needed

• but this implicit cast can have

  

Casting between signed and unsigned:

• implementation defined: bit patter is preserved

• so no change in bits representation

Casting from small to big:

• the value is preserved

• casting unsigned small to unsigned big preserves both

• 

Casting from big to small

• the value is preserved if it fits

• so chop left bits

Casting between both size and sign/unsigned

• compiler may apply rules in either order

• (the order of operation is defined in a complicated way)

• 

• Solution:

  

Fixed-size Integers

Decimals

Float: 32 bits Double: 64 bits Calculation rounding error depend on compiler

Union and Enum

enum season {WINTER, SPRING, SUMMER, FALL};
emun season today = FALL;
if (today == WINTER) printf("snow!\n");
switch (today) {
  case WINTER:
    printf("snow!\n");
    break;
  default:
    printf("other\n");
}

Union Type allow using the same space in different ways:

• 

• 

• 

•