Lecture 004

The Basics About Assembly

Architecture: Instruction Set Architecture (ISA). Parts of processor design for writing assembly. (e.g. registers)

• Example ISA:
  • Intel: x86, IA32, Itanium, x86-64
  • ARM: in mobile phones
  • RISC V: New open-source ISA

Micro-architecture: implementation of architecture (eg. cache size and core frequency)

Programmer Visible States

• PC: Program counter

  • address of next instruction
  • called "RIP" (x86-64)

• Register File

  • place to store program data

• Condition code

  • place to store status information about recent arithmetic
  • used in conditional branching

• Memory

  • byte addressable array
  • store user data
  • store code

Assembly data type

• integer or address of 1, 2, 4, 8 bytes

• float of 4, 8, 10 bytes

• SIMD vector data of 8, 16, 32, 64 bytes

• code instructions (encoded by sequence of bytes)

Registers

• where %esp(stack pointer) and %ebp(base pointer) are special

• has x86 capability (IA32 registers)

• contain

  

  • which also contain 16-bit very early registers
  • which also contain lowest byte reference

    

3 kinds operations in assembly

• transfer data between memory and register

• arithmetic function on register or memory

• control flow and jumps, branches

%rdi: first argument %rsi: second argument %rax: third argument

Basic Instructions

movq

movq [Source] [Dest]

• operand types:
  • immediate (constant integer)
    • e.g.: $0x400, $-533
  • register
    • e.g.: %rax, %r13 (but %rsp is reserved for special use)
  • memory: 8 bytes memory
    • e.g.: (%rax)

Unix: [Source] [Dest] Win: [Dest] [Source]

Possible Combinations:

• Note: cannot do memory-to-memory transfer in one instruction

Addressing Modes

D(Rb, Ri, S) = (Rb + Ri * S) + D

• D: displacement of 8, 16, or 32 bits, default 0

• Rb: base register (has to be one of 16 integer register)

• Ri: index register (any except %%rsp), default 0

• S: scale of 1, 2, 4, or 8, default 1

If above not specified, they are identity (0 or 1 depending on arithmetics)

leaq

leaq [Src], [Dst]

• move but without memory access

Other Arithmetic Operations

• note: no distinction between signed and unsigned

Compilation

• assembler: translate .s into .o

• linker: resolve reference between files; run-time libraries; (some libraries are dynamically linked in run-time)

Disassemble: objdump -d sum

• examining object code and bit patterns

• can be run on a.out or .o files