어셈블리 언어와 주소 지정 방식

Thirty-one instructions on a row of switches

• In 1949 the EDSAC computer in Cambridge started up from 31 instructions set on switches, written by David Wheeler and called the "initial orders".
• Their whole job was to read letters such as A and S from paper tape and turn them into the numbers the machine actually understood: the first assembler 汇编器.
• Programmers have written mnemonics instead of bit patterns ever since, and the exam's own instruction set is a direct descendant.
• This lesson is about that translation, tracing a program by hand, and the five ways an instruction can say where its data is.

Machine code and assembly language

• The processor runs machine code 机器码: bit patterns, specific to one architecture, each one an operation code and an operand.
• Assembly language 汇编语言 is the readable form. Each instruction is written with a mnemonic 助记符 such as LDD, ADD or JMP, and each assembly instruction becomes exactly one machine-code instruction.
• The assembler translates it. The relationship is one to one, which is what makes assembly different from a high-level language, where one statement becomes many instructions.

Each mnemonic becomes one opcode; each symbolic address becomes a number

The two-pass assembler

• Pass 1 reads the source and builds a symbol table 符号表: every time a label 标签 such as LOOP: appears, its address is recorded. No code is produced.
• Pass 2 reads the source again and generates the code, replacing each mnemonic by its opcode 操作码 and each symbolic address by the number from the symbol table.
• Two passes are needed because of forward references 前向引用: JMP LOOP may appear before the line LOOP: has been seen, so its address is not known on the first pass.

Worked example: applying the two passes

100        LDD  COUNT
101 LOOP:  DEC  ACC
102        CMP  #0
103        JPN  LOOP
104        END
105 COUNT: 5

The instruction set

• Cambridge's instruction set 指令集 has one general-purpose register, the accumulator ACC, and an index register IX. An operand #n is denary, Bn binary, &n hexadecimal, and <address> a location number or a label.
• Data movement: LDM #n, LDD <address>, LDI <address>, LDX <address>, LDR #n, MOV IX, STO <address>. Input and output: IN, OUT.
• Arithmetic: ADD, SUB, INC, DEC. Compare: CMP, CMI. Jumps: JMP unconditional, JPE and JPN conditional. Bit manipulation: AND, OR, XOR, LSL, LSR. Then END.
• "Instructions are grouped": name the groups and give one instruction from each.

Worked example: tracing a program

100        LDM  #0
101        STO  TOTAL
102        LDD  TOTAL
103        ADD  #5
104        STO  TOTAL
105        LDD  COUNT
106        DEC  ACC
107        STO  COUNT
108        CMP  #0
109        JPN  102
110        END
111 TOTAL: 0
112 COUNT: 3

Addressing modes

• The addressing mode 寻址方式 says how the processor finds the operand.

The same operand field, four different meanings

Worked example: one operand, four answers

Why indexed addressing exists

• An array is a run of consecutive locations. LDX 200 with IX = 0, 1, 2, … reads element 0, 1, 2, … without changing the instruction.
• The loop pattern: LDR #0 to zero the index, LDX ARRAY to fetch an element, do something, INC IX, compare with the length, JPN back.
• Direct addressing would need a separate instruction for every element; indexed addressing needs one.

Compare and jump

• CMP #n or CMP <address> compares the ACC with a value and sets the flag. It does not change the ACC.
• JPE <address> jumps if the last compare found the values equal; JPN <address> jumps if they were not equal; JMP <address> always jumps.
• An IF in a high-level language becomes a CMP followed by a conditional jump; a loop becomes a compare and a jump back to a label.

Marks that slip away

You've got it