Skip to content

Chapter 4 - Data Movement Instructions

Data movement instructions do not affect flags.

4-1 MOV Revisited

三种操作数模式

模式 默认地址大小 默认操作数大小
16-bit模式 16-bit 16-bit
32-bit保护模式 32-bit 32-bit
64-bit模式 64-bit 32-bit

在 code segment descriptor 中, L-bit and D/B-bit indicate the operation mode:

  • L=1 for 64-bit instruction mode
  • L=0 and D/B =1 for 32-bit instruction mode
  • L=0 and D/B =0 for 16-bit instruction mode

指令不能超过 15B，否则发生 general-protection exception

如何分辨 - 操作数前缀

指令前缀

REX 前缀

alt text

Legacy Prefixes - 4 组，同组只能存在一个

  • Group 1
    • LOCK - 原子操作，只有某些指令能用，否则 undefined opcode exception (#UD) occurs
    • REPNE/REPNZ - Repeat 相应 string instruction rCX 次
    • REP or REPE/REPZ
  • Group 2: segment override prefix
    • CS/ES/SS/FS/DS/GS segment override
    • 有默认值 Instructions: CS, Local Data: DS, Stack: SS, Destination Strings: ES
    • alt text
  • Group 3: Operand-size override prefix
    • REX (REX.W) prefix - 64 位操作数
    • 66H prefix - 16 位
    • alt text
  • Group 4: Address-size override prefix
    • alt text

Escape sequence/Opcode

  • 增加操作码长度来提供其他操作码（1B to 2B）

4–2 LOAD EFFECTIVE ADDRESS

LEA: Loads a near pointer (offset)

alt text

  • load 一个偏移地址到 16/32 位寄存器中
  • LEA BX,[DI] loads the offset address specified by [DI] (contents of DI) into the BX register
  • SEG and OFFSET 操作返回一个内存位置的段地址和偏移量
  • LEA SI, DATA1 相当于 MOV SI, OFFSET DATA1 (DATA1 是一个 label)
  • OFFSET 一般比 LEA 快

LDS, LES, LFS, LGS, and LSS: Loads a far pointer (segment selector and offset)

不是基本指令，先跳了

4–3 STRING DATA TRANSFERS

字符串数据传输指令

  • LODS
    • implicit operands (AL, AX, EAX)
  • STOS - 连续内存的初始化
    • implicit operands (AL, AX, EAX)
  • MOVS - 内存数据之间的传输（x86 中唯一的一条）
    • only memory-to-memory transfer
    • SI and DI 指向的内存的数据交换
    • A suffix (B, W or D) indicates the data size to operate on

alt text

字符串比较指令

  • SCAS
  • CMPS

alt text

用到的寄存器和 flag

  • DI - ES 段的偏移地址，cannot be overridden
  • SI - DS 段的偏移地址，can be overridden
  • direction flag (D, located in the flag register) - 通过 CLD SLD 清零/设定
    • D=0, auto-increment
    • D=1, auto-decrement
  • REP and CX/ECX - REP 前缀可以让指令重复执行 n 次，n 存在 CX/ECX 中
  • Permissible forms with suffix: B - byte, W - word, D - double word
    • e.g. MOVSB, byte-sized MOVS

INS - 从设备端口（如 disk drives）取数据到内存

  • explicit-operands form - INS WORD PTR [DI], DX（必须要 WORD PTR，因为 DX 代表端口，长度不确定）
  • no-operands form - INSB，INSW，INSD（目的地址必须放在 DI/EDI，端口一定放在 DX，可省略，需要用后缀指定长度）
  • 可以用 REP prefix 重复执行

inputs 50 bytes of data from an I/O device： alt text

OUTS - 从内存取数据送到设备端口

  • explicit-operands form - OUTS DX, WORD PTR [SI]
  • no-operands form - 类似 INS

4-4 MISCELLANEOUS DATA TRANSFER INSTRUCTIONS

XCHG

  • 把一个寄存器和另一个寄存器/内存位置的数据交换
  • 不能交换 segment registers or memory-to-memory data
  • 用来实现 semaphores

alt text

LAHF and SAHF

什么东西，懒得看

XLAT (Table Look-up Translation)

  • implicit operands (AL, BX)
  • works like MOV AL, [seg:BX + AL]
  • XLAT writes AL without changing EAX[31:8]

IN & OUT

  • port alt text alt text
  • IN - IN AL, 19H: 把 19H 接口处的数据放入 AL
  • OUT - OUT 32H, AX: 把 AX 的数据放到输出端口 32H
  • Fixed-port addressing: 输入输出端口是一个长度为 8bit 的地址
  • Variable-port addressing: IN AL, DX, OUT DX, AX (16-bit)

MOVSX & MOVZX - move and sign-extend/move and zero-extend

BSWAP

  • reverses the byte order
  • 大小端转换

CMOV

  • CMOVcc - 当 condition code (cc) 被满足时执行 MOV
  • CMOVZ - ZF = 1 时 MOV
  • alt text
  • The purpose of CMOV is to avoid a branchalt text
  • 把控制依赖转换成数据依赖，extends instruction scheduling space

alt text

4–5 ASSEMBLER DETAIL

Directives vs Instructions

  • Directives: tell assembler how to do
  • Instructions: tell CPU what to do

Directives in MASM

  • Data Allocation – DB, DW, DD, DQ, DT
    • DUP 可以初始化一个数据多次
    • DB 100 DUP(6) - reserves 100 bytes of 6
  • Structure – STRUCT, RECORD
  • Code Labels – ALIGN, ORG
    • ALIGN directive aligns the next data element or instruction on an address that is a multiple of its parameter (必须是 2 的整数次方)
  • Segment – SEGMENT, ENDS, ASSUME
  • Simplified Segment – .CODE, .DATA, .STACK, .MODEL, .EXIT
  • Procedures – PROC, ENDP
    • 定义函数 
      name PROC   [near/far] // must be followed with a NEAR or FAR (local & global)
    statements 
    ret 
name ENDP
  • Macros – MACRO, ENDM
    • inline expansion alt text alt text
  • Miscellaneous – EQU, INCLUDE
    • CONSTANT_NAME EQU expression - e.g. TEN EQU 10, TEN 会被替换成 10
    • THIS BYTE, THIS WORD, THIS DWORD, or THIS QWORD
    • ORG - change the starting offset address
    • ASSUME - tells the assembler what names have been chosen for the code, data, extra, and stack segments

alt text

MACRO and ENDM - 宏

Memory Organization