Chapter_02 of COD

Instruction Set(指令集)

The repertoire(全体) of instruction of a computer

Arithmetic Operations

• Add and subtract, three operands
  • Two sources and one destination
  • add a, b, c // a gets b + c
• All arithmetic operations have this form

Register Operands

• Arithmetic instructions use register operands
• RISC-V has a 32 * 32-bit register file
  • use for frequently accessed data
  • 32-bit data is called a “word”
    • 32 * 32-bit general purpose registers x0 to x31
    • 64-bit data is called a “doubleword”
    • 16-bit data is called a “halfword”

Immediate Operands

• Constant data specified in an instruction
  addi x22, x22, 4
• Make the Common case fast
  • small constants are common
  • Immediate operand avoids a load instruction
    立即操作数可以避免load指令操作

The Constant Zero

• RISC-V register x0 is the constant 0
  • Cannot be overwritten
• Useful for common operatons
  • E.g. negate a value in a register by sub zero by the value
    使用该寄存器进行取反操作
  • sub x9, x0, x8

RISC-V Registers

• x0: the constant value 0 （存常数0）
• x1: return address（返回地址）
• x2: stack pointer (栈指针)
• x3: global pointer（全局指针）
• x4: thread pointer
• x5 - x7, x28 - x31: temporaries（临时变量）
• x8: frame pointer (帧指针)
• x9, x18 - x27: saved registers（保存寄存器）
• x10 - x11: function arguments/results（函数返回值）
• x12 - x17: function arguments（函数参数）

Design Principle

• Simplicity favours regularity
  简单源于规整
• Smaller is faster
  小则快
• Good design demands good compromises

Endianness (端序)

小端字节序，高字节存放于内存高地址，低字节存于内存低地址，大端字节反之

Sign Extension

简单来说就是将一个数扩展到更高的位数，而为了保持本来这个数的值，则按照最高位向前扩展。

• Representing a number using more bits
  • Preserve the numeric value
• Replicate the sign bit to the left
  • c.f. unsigned values: extend with 0s
  • e.g. addi x11, x0, 0x8f5
    • Immediate is always sign-extended to 32-bits before use in an arithmetic operation
  • Examples: 8-bit to 16-bit
    • +2：0000 0010 => 0000 0000 0000 0010
    • -2： 1111 1110=> 1111 1111 1111 1110
• In RISC-V instruction set
  • lb: sign-extend loaded byte (符号扩展)
  • lbu: zero-extend loaded byte (零扩展)

RISC-V Instruction Formats

#Instruction_Formats

R-format

同时具有rs1,rs2,rd
#Instruction_Formats/R-format

• Instruction field
  • opcode: operation code
  • rd: destination register number
  • funct3: 3-bit function code(additional opcode)
  • rs1: first source register number
  • rs2: second source register number
  • funct7: 7-bit function code(additional opcode)

I-format

#Instruction_Formats/I-format
立即数为12位字段，即[11:0],同时注意符号位扩展，从12位扩展到32位
代表指令：addi,load,slli,slri,jalr

• Immediate arithmetic and load instructions

  • rs1: source or base address register number
  • immediate: constant operand, or offset added to base address

• addi x15, x1, -50

  • 
  • Immediate is always sign-extended to 32-bits before use in an arithmetic operation

• Load Instruction

  • lw x9, 240(x10)
  • 
  • The 12-bit signed immediate is added to the base address in register rs1 to form the memory address（由rs1作为基地址，然后加上立即数进行跳转）
    • This is very similar to add-immediate operation but used to create address not to create final result
  • The value loaded from memory is stored in register rd

S-format

#Instruction_Formats/S-format
有rs1, rs2，rd变为立即数的[4:0]字段，立即数仍然为12位[11:0]，但是是隔断的
代表指令：store

• rs1: base address register number（基地址寄存器，根据这个里面的地址再加上立即数进行跳转）
• rs2: source operand register number
• immediate: offset added to base address
  • Split so that rs1 and rs2 field always in the same place
    • ![[Chapter_2 Instructions in different format#Design Principle]]

U-format

#Instruction_Formats/U-format
没有rs1, rs2，只有rd，也没有func字段，立即数为20位[31:12]（低位字段清零），（常与addi连用，注意符号位扩展）
代表指令:lui

32-bit Constants (大立即数)

• For the occasional 32-bit constant
  • lui rd, constant
  • Copies 20-bit constant to bits [31:12] of rd
  • Clear bits [11:0] of rd to 0

U-format for “Upper Immediate” instructions

• Has 20-bit immediate in upper 20 bits of 32-bit instruction word
• One destination resgister, rd
• Used for two instructions
  • LUI - Load Upper Immediate
  • AUIPC - Add Upper Immediate to PC
• LUI to create long immediates
• LUI Copies 20-bit constant to bits[31:12] of rd, Clear bits [11:0] of rd to 0.（高20位加载，低12位置0）
• Together with an addi to set low 12 bits, can create any 32-bit value in a register using two instructions (lui, addi)（使用lui和addi两条指令来构建一个完整的加载大立即数的操作）

#Create_long_immediates

思考：如何得到0xDEADBEEF ？

错误解答（简单的拼接）：

lui x10, 0xDEADB # x10 = 0xDEADB000
addi x10, x10, 0xEEF # x10 = 0xDEADAEEF

Point:
addi 12-bit immediate is always sign-extended, if top bit is set, will subtract -1 from upper 20 bits
一定要注意低12位的符号位，addi指令总是进行符号扩展，如果符号位为1，会从高20位中减去1

正确解答：

lui x10, 0xDEADC
addi x10, x10, 0xEEF

SB-format

有rs1, rs2 (进行算术运算),没有rd，12位立即数字段被分为[12],[10:5],[4:1],[11]，所以立即数为[12:1]，以半字为单位，这点要非常注意
代表指令：beq

• Branch instructions specify
  • Opcode, two registers, target address
• Most branch targets are near branch
  • Forward or backward
• SB format, mostly same as S-format

#Instruction_Formats/SB-format

• But now immediate represents values -4096 to +4094 in 2-byte increments
• The 12 immediate bits encode even 13-bit signed byte offsets(lowest bit of offset is always zero, so no need to store it)

#Instruction_Formats/SB-format/Branch_Example

UJ-format

#Instruction_Formats/R-format/UJ-format
有rd，没有rs1, rs2,立即数字段被拆分为[20],[10:1],[11],[19:12]，主要用于跳转到指定地址。（注意也是半字为单位）可以向前后各跳转2^18个指令

如果要进行大范围跳转，那使用lui和jalr的组合指令（注意jalr就不是半字为单位了）

代表指令：jal

• jal target uses 20-bit immediate for larger range
• For long jumps, eg, to 32-bit absolute address
  • lui: load address [31:12] to temp register
  • jalr: add address [11:0] and jump to target
    Uses of jal:
• j pseudo-instruction
  • j Label = jal x0, Lable #Discard return address
• Call function within +-2^18 instructions of PC
  • jal ra, FuncName
  • jal的imm是2^(20+1)个byte
  • 由于一个指令对应4个byte，所以2^19对应+-2^18条指令

jal 与 jalr 对比

• jal为UJ-format
• jalr为I-format
  

四种不同的寻址方式

• 立即数寻址
• 寄存器寻址
• 基地址寻址
• PC相对寻址
  • beq，bne，jal

过程调用的步骤

1. 将参数放到x10 - x17
2. 交换过程的控制权
3. 为过程分配存储空间
4. 执行过程操作
5. 将结果放到寄存器中给调用者
6. 回到 x1 中的保存地址

#logic_operation/and

AND Operation

• 选择性置零
• and 0

#logic_operation/or

OR Operation

• 选择性置1
• or 1

#logic_operation/xor

XOR Operation

• 选择性翻转
• xor 1

边界检查

#边界检查

• 当使用角标大于等于数组长度或为负数发生数组下标越界
• 将有符号数当作无符号数处理
  • 最高有效位在有符号数中表示符号位，在无符号数中表示数的很大一部分
  • 无符号比较在检测x<y时，也检测了x是否为负数
  • bgeu x, y, IndexOutofBounds
    • case 1: x >= y (索引号>数组长度)
    • case 2: x < 0(索引号为负)

The “ABI”(Application Binary Interface)

A contract between functions

Caller-saved register and Callee-saved register

The difference between two strategies

The RISC-V Register and Convention

To Summerize:
caller 保护：

• x1 返回地址
• x10 - x17 参数
• x5 - x7, x28 - x31 临时变量

why identify the caller/callee saved

example:

int a, b, c, d;
a = b - c + d;

对于上面的例子，a,b,c,d 是声明出来的变量（存储在r0 - r3中），而在计算a 的值的时候，无法通过一条指令完成，需要借助临时变量，比如tmp（分配t0寄存器）

首先计算 tmp = b - c
sub r1, r2, t0
然后计算a = tmp + d
add t0, r3, r0
tmp的生命周期到此为止，它的生命周期非常短，但又有专门的寄存器来保存这个临时变量，这种寄存器里面的值总是易变的，用完即废

所以在函数调用过程中，对于两类寄存器有着不同的责任保护方，第一类寄存器（r）如果在callee里面改了，不做保护的话，caller里面声明的变量值都改变了，这类寄存器，callee用了一个就要保护一个；
对于第二类寄存器（t），临时变量用完即废，callee破坏了也可能没啥影响，callee没有义务保护，但如果caller确实希望保存该临时变量，也必须由caller声明进行保护。

Change 64-bit to 32-bit

cut all the labeled red half:

• doubleword to word
• 16 to 8
• 8 to 4