Moving data to/from memory, endianness

Notes

• Most instructions in intel assembly are in the form
  • [label:] operand opcode, opcode
  • Operand is a mnemonic like mov, add, ...
  • The operands are normally dest, src
  • some are single operand instructions (push, pop)
  • Some are no operand instructions (ret)
• Operands can be
  • a register
    • Jorgensen uses <reg>
    • And also <regn>, where n ∈ {8, 16, 32, 64}
  • An immediate
    • <imm>
    • And also <immn>, where n ∈ {8, 16, 32, 64}
  • Memory
    • <mem>
  • He specifies <src
    • This can be mem, imm or reg
  • And <dest
    • mem or reg,
  • But src and dest can not both be mem.
• mov dest, src
  • The src and dest must be the same size.
  • The assembler decides which of the mov instructions to execute
  • See this page.
• There is some strange behavior
  • quadword instructions are fine
  • double word instructions zero out the top of the register
  • word and byte instructions do not change the value of the register.
• More on memory indexing later but
  • mov dest, label is actually an immediate
    • It loads the address into memory
  • mov dest,[label]
  • mov [label], src
    • deals with the memory location
• x/8bx (char*)&storage+1
• Look at play.asm
  • Examine memory with x/1gx &storage
  • Step through each of the moves, looking at what is in memory after the move.
• After the last mov
  • Examine storage as a
    • big integer: x /1gx &storage
    • string: x /1s (char*)&storage
    • As an individual characters: x /8bx &storage
  • Double check rax, as hex.
  • What is going on?
  • Examine the data file in the code directory, that is even worse!
• Little Endian vs Big Endian
  • Assume you want to implment your own number type.
    • You might use an array to hold the digits.
    • So the number 42601
      • Is in a 5 element array.
      • But where do you put the 1?
        • digit[0]?
        • digit[4]?
    • If you select 0, you are little-endian
    • If you select the 4, you are big-endian.
    • See (from Wikipedia)
    • Look the article on Endinness on wikipedia.
  • Intel is little endian
    • The least sugnificant byte goes at address+0
    • The next sugnificant byte goes at address+1
  • What we see on the file system however, is words as they are stored
    • Two characters per word
    • First character shows up "last" in the word
    • But od -c changes the order.
    • Try od -cx !