https://www.cnblogs.com/thoupin/p/4018455.html
无意中找到一篇十分好用,而且篇幅也不是很大的入门教程,通篇阅后,再把“栗子”敲一遍,基本可以有一个比较理性的认识,从而方便更好地进一步深入学习。
废话不多说,上干货(英语好的直接跳过本人的渣翻译了哈——!纯本人手打原创,有错请指教,要转载请声明出处,谢~~):
MIPS Architecture and Assembly Language Overview
MIPS架构及其汇编初步
(开始之前稍微再提下,整体分为4个结构:)
1: 寄存器种类;
2: 算术及寻址指令
3: 程序结构
4: 系统调用
Data Types and Literals
数据类型
- 所有MIPS指令都是32位长的
- 各单位:1字节=8位,半字长=2个字节,1字长=4个字节
- 一个字符空间=1个字节
- 一个整型=一个字长=4个字节
- 单个字符用单引号,例如:'b'
- 字符串用双引号,例如:"A string"
Registers
寄存器
-
-
- MIPS下一共有32个通用寄存器
- 在汇编中,寄存器标志由$符开头
- 寄存器表示可以有两种方式
-
-
-
- 直接使用该寄存器对应的编号,例如:从$0到$31
- 使用对应的寄存器名称,例如:$t1, $sp(详细含义,下文有表格
- 对于乘法和除法分别有对应的两个寄存器$lo, $hi
- 对于以上二者,不存在直接寻址;必须要通过mfhi("move from hi")以及mflo("move from lo")分别来进行访问对应的内容
- 栈的走向是从高地址到低地址
-
MIPS下各个寄存器编号及描述:
Register 寄存器编号 | Alternative 寄存器名 | Description 寄存器用途 |
---|---|---|
0 | zero | the value 0 永远返回零 |
1 | $at | (assembler temporary) reserved by the assembler 汇编保留寄存器(不可做其他用途) |
2-3 | $v0 - $v1 | (values) from expression evaluation and function results (Value简写)存储表达式或者是函数的返回值 |
4-7 | $a0 - $a3 | (arguments) First four parameters for subroutine. (Argument简写)存储子程序的前4个参数,在子程序调用过程中释放 |
8-15 | $t0 - $t7 | (temporaries) Caller saved if needed. Subroutines can use w/out saving. (Temp简写)临时变量,同上调用时不保存 |
16-23 | $s0 - $s7 | (saved values) - Callee saved. (Saved or Static简写?)静态变量?调用时保存 |
24-25 | $t8 - $t9 | (temporaries) Caller saved if needed. Subroutines can use w/out saving. (Temp简写)算是前面$0~$7的一个继续,属性同$t0~$t7 |
26-27 | $k0 - $k1 | reserved for use by the interrupt/trap handler (breaK off简写?)中断函数返回值,不可做其他用途 |
28 | $gp | global pointer. (Global Pointer简写)指向64k(2^16)大小的静态数据块的中间地址(字面上好像就是这个意思,块的中间) |
29 | $sp | stack pointer (Stack Pointer简写)栈指针,指向的是栈顶 |
30 | $s8/$fp | saved value / frame pointer (Saved/Frame Pointer简写)帧指针 |
31 | $ra | return address 返回地址,目测也是不可做其他用途 |
Program Structure
程序结构
- 本质其实就只是数据声明+普通文本+程序编码(文件后缀为.s,或者.asm也行)
- 数据声明在代码段之后(其实在其之前也没啥问题,也更符合高级程序设计的习惯)
Data Declarations
数据声明
- 数据段以 .data为开始标志
- 声明变量后,即在主存中分配空间。
Code
代码
- 代码段以 .text为开始标志
- 其实就是各项指令操作
- 程序入口为main:标志(这个都一样啦)
- 程序结束标志(详见下文)
Comments
注释
- 同C系语言
-
-
- MIPS程序的基本模板如下:
# Comment giving name of program and description of function # 说明下程序的目的和作用(其实和高级语言都差不多了) # Template.s #Bare-bones outline of MIPS assembly language program .data # variable declarations follow this line # 数据变量声明 # ... .text # instructions follow this line # 代码段部分 main: # indicates start of code (first instruction to execute) # 主程序 # ... # End of program, leave a blank line afterwards to make SPIM happy # 必须多给你一行,你才欢?
- MIPS程序的基本模板如下:
-
Data Declarations
数据声明
format for declarations:
声明的格式:
name: storage_type value(s) 变量名:(冒号别少了) 数据类型 变量值
-
- create storage for variable of specified type with given name and specified value
- value(s) usually gives initial value(s); for storage type .space, gives number of spaces to be allocated
- 通常给变量赋一个初始值;对于.space,需要指明需要多少大小空间(bytes)
Note: labels always followed by colon ( : )
example var1: .word 3 # create a single integer variable with initial value 3 # 声明一个 word 类型的变量 var1, 同时给其赋值为 3 array1: .byte 'a','b' # create a 2-element character array with elements initialized # to a and b # 声明一个存储2个字符的数组 array1,并赋值 'a', 'b' array2: .space 40 # allocate 40 consecutive bytes, with storage uninitialized # could be used as a 40-element character array, or a # 10-element integer array; a comment should indicate which! # 为变量 array2 分配 40字节(bytes)未使用的连续空间,当然,对于这个变量 # 到底要存放什么类型的值, 最好事先声明注释下!
Load / Store Instructions
加载/保存(也许这里写成读取/写入 可能更易理解一点) 指令集
- 如果要访问内存,不好意思,你只能用 load 或者 store 指令
- 其他的只能都一律是寄存器操作
load:
lw register_destination, RAM_source
#copy word (4 bytes) at source RAM location to destination register.
从内存中 复制 RAM_source 的内容到 对应的寄存器中
(lw中的'w'意为'word',即该数据大小为4个字节)
lb register_destination, RAM_source
#copy byte at source RAM location to low-order byte of destination register,
# and sign-e.g.tend to higher-order bytes同上, lb 意为 load byte
store word:
sw register_source, RAM_destination
#store word in source register into RAM destination
#将指定寄存器中的数据 写入 到指定的内存中
sb register_source, RAM_destination
#store byte (low-order) in source register into RAM destination
load immediate:
li register_destination, value
#load immediate value into destination register
顾名思义,这里的 li 意为 load immediate
example: .data var1: .word 23 # declare storage for var1; initial value is 23 # 先声明一个 word 型的变量 var1 = 3; .text __start: lw $t0, var1 # load contents of RAM location into register $t0: $t0 = var1 # 令寄存器 $t0 = var1 = 3; li $t1, 5 # $t1 = 5 ("load immediate") # 令寄存器 $t1 = 5; sw $t1, var1 # store contents of register $t1 into RAM: var1 = $t1 # 将var1的值修改为$t1中的值: var1 = $t1 = 5; done
Indirect and Based Addressing
立即与间接寻址
load address:
直接给地址
la $t0, var1
- copy RAM address of var1 (presumably a label defined in the program) into register $t0
indirect addressing:
地址是寄存器的内容(可以理解为指针)
lw $t2, ($t0)
- load word at RAM address contained in $t0 into $t2
sw $t2, ($t0)
- store word in register $t2 into RAM at address contained in $t0
based or indexed addressing:
+偏移量
lw $t2, 4($t0)
- load word at RAM address ($t0+4) into register $t2
- "4" gives offset from address in register $t0
sw $t2, -12($t0)
- store word in register $t2 into RAM at address ($t0 - 12)
- negative offsets are fine
Note: based addressing is especially useful for:
不必多说,要用到偏移量的寻址,基本上使用最多的场景无非两种:数组,栈。
- arrays; access elements as offset from base address
- stacks; easy to access elements at offset from stack pointer or frame pointer
example: 栗子: .data array1: .space 12 # declare 12 bytes of storage to hold array of 3 integers # 定义一个 12字节 长度的数组 array1, 容纳 3个整型 .text __start: la $t0, array1 # load base address of array into register $t0 # 让 $t0 = 数组首地址 li $t1, 5 # $t1 = 5 ("load immediate") sw $t1, ($t0) # first array element set to 5; indirect addressing # 对于 数组第一个元素赋值 array[0] = $1 = 5 li $t1, 13 # $t1 = 13 sw $t1, 4($t0) # second array element set to 13 # 对于 数组第二个元素赋值 array[1] = $1 = 13 # (该数组中每个元素地址相距长度就是自身数据类型长度,即4字节, 所以对于array+4就是array[1]) li $t1, -7 # $t1 = -7 sw $t1, 8($t0) # third array element set to -7 # 同上, array+8 = (address[array[0])+4)+ 4 = address(array[1]) + 4 = address(array[2]) done
Arithmetic Instructions
算术指令集
- 最多3个操作数
- 再说一遍,在这里,操作数只能是寄存器,绝对不允许出现地址
- 所有指令统一是32位 = 4 * 8 bit = 4bytes = 1 word
add $t0,$t1,$t2 # $t0 = $t1 + $t2; add as signed (2's complement) integers
sub $t2,$t3,$t4 # $t2 = $t3 Ð $t4 addi $t2,$t3, 5 # $t2 = $t3 + 5; "add immediate" (no sub immediate) addu $t1,$t6,$t7 # $t1 = $t6 + $t7; add as unsigned integers subu $t1,$t6,$t7 # $t1 = $t6 + $t7; subtract as unsigned integers mult $t3,$t4 # multiply 32-bit quantities in $t3 and $t4, and store 64-bit # result in special registers Lo and Hi: (Hi,Lo) = $t3 * $t4 运算结果存储在hi,lo(hi高位数据, lo地位数据) div $t5,$t6 # Lo = $t5 / $t6 (integer quotient) # Hi = $t5 mod $t6 (remainder) 商数存放在 lo, 余数存放在 hi mfhi $t0 # move quantity in special register Hi to $t0: $t0 = Hi 不能直接获取 hi 或 lo中的值, 需要mfhi, mflo指令传值给寄存器 mflo $t1 # move quantity in special register Lo to $t1: $t1 = Lo # used to get at result of product or quotient move $t2,$t3 # $t2 = $t3
Control Structures
控制流
Branches
分支(if else系列)
- comparison for conditional branches is built into instruction
b target # unconditional branch to program label target beq $t0,$t1,target # branch to target if $t0 = $t1 blt $t0,$t1,target # branch to target if $t0 < $t1 ble $t0,$t1,target # branch to target if $t0 <= $t1 bgt $t0,$t1,target # branch to target if $t0 > $t1 bge $t0,$t1,target # branch to target if $t0 >= $t1 bne $t0,$t1,target # branch to target if $t0 <> $t1Jumps
跳转(while, for, goto系列)
j target # unconditional jump to program label target 看到就跳, 不用考虑任何条件 jr $t3 # jump to address contained in $t3 ("jump register") 类似相对寻址,跳到该寄存器给出的地址处Subroutine Calls
子程序调用
subroutine call: "jump and link" instruction
jal sub_label # "jump and link"
- copy program counter (return address) to register $ra (return address register)
- 将当前的程序计数器保存到 $ra 中
- jump to program statement at sub_label
subroutine return: "jump register" instruction
jr $ra # "jump register"
- jump to return address in $ra (stored by jal instruction)
- 通过上面保存在 $ra 中的计数器返回调用前
Note: return address stored in register $ra; if subroutine will call other subroutines, or is recursive, return address should be copied from $ra onto stack to preserve it, since jal always places return address in this register and hence will overwrite previous value
如果说调用的子程序中有调用了其他子程序,如此往复, 则返回地址的标记就用 栈(stack) 来存储, 毕竟 $ra 只有一个, (哥哥我分身乏术啊~~)。
System Calls and I/O (SPIM Simulator)
系统调用 与 输入/输出(主要针对SPIM模拟器)
(本人使用的是Mars 4.4,也通用--!)
- 通过系统调用实现终端的输入输出,以及声明程序结束
- 学会使用 syscall
- 参数所使用的寄存器:$v0, $a0, $a1
- 返回值使用: $v0
下表给出了系统调用中对应功能,代码,参数机返回值
Service | Code 对应功能的调用码 | Arguments 所需参数 | Results 返回值 |
---|---|---|---|
print_int 打印一个整型 | $v0 = 1 | $a0 = integer to be printed 将要打印的整型赋值给 $a0 | |
print_float 打印一个浮点 | $v0 = 2 | $f12 = float to be printed 将要打印的浮点赋值给 $f12 | |
print_double 打印双精度 | $v0 = 3 | $f12 = double to be printed 将要打印的双精度赋值给 $f12 | |
print_string | $v0 = 4 | $a0 = address of string in memory 将要打印的字符串的地址赋值给 $a0 | |
read_int | $v0 = 5 | integer returned in $v0 将读取的整型赋值给 $v0 | |
read_float 读取浮点 | $v0 = 6 | float returned in $v0 将读取的浮点赋值给 $v0 | |
read_double 读取双精度 | $v0 = 7 | double returned in $v0 将读取的双精度赋值给 $v0 | |
read_string 读取字符串 | $v0 = 8 | $a0 = memory address of string input buffer 将读取的字符串地址赋值给 $a0 将读取的字符串长度赋值给 $a1 | |
sbrk 应该同C中的sbrk()函数 动态分配内存 | $v0 = 9 | $a0 = amount 需要分配的空间大小(单位目测是字节 bytes) | address in $v0 将分配好的空间首地址给 $v0 |
exit 退出 | $v0 =10 | 你懂得![]() |
-
- 大概意思是要打印的字符串应该有一个终止符,估计类似C中的'\0', 在这里我们只要声明字符串为 .asciiz 类型即可。下面给个我用Mars4.4的提示:
- .ascii 与 .asciiz唯一区别就是 后者会在字符串最后自动加上一个终止符, 仅此而已
- The read_int, read_float and read_double services read an entire line of input up to and including the newline character.
- 对于读取整型, 浮点型,双精度的数据操作, 系统会读取一整行,(也就是说以换行符为标志 '\n')
- The read_string service has the same semantices as the UNIX library routine fgets.
- It reads up to n-1 characters into a buffer and terminates the string with a null character.
- If fewer than n-1 characters are in the current line, it reads up to and including the newline and terminates the string with a null character.
- 这个不多说了,反正就是输入过长就截取,过短就这样,最后都要加一个终止符。
- The sbrk service returns the address to a block of memory containing n additional bytes. This would be used for dynamic memory allocation.
- 上边的表里已经说得很清楚了。
- The exit service stops a program from running.
- 你懂得。。。
e.g. Print out integer value contained in register $t2 栗子: 打印一个存储在寄存器 $2 里的整型 li $v0, 1 # load appropriate system call code into register $v0; 声明需要调用的操作代码为 1 (print_int) 并赋值给 $v0 # code for printing integer is 1 move $a0, $t2 # move integer to be printed into $a0: $a0 = $t2 将要打印的整型赋值给 $a0 syscall # call operating system to perform operation e.g. Read integer value, store in RAM location with label int_value (presumably declared in data section) 栗子: 读取一个数,并且存储到内存中的 int_value 变量中 li $v0, 5 # load appropriate system call code into register $v0; # code for reading integer is 5 声明需要调用的操作代码为 5 (read_int) 并赋值给 $v0 syscall # call operating system to perform operation、 经过读取操作后, $v0 的值已经变成了 输入的 5 sw $v0, int_value # value read from keyboard returned in register $v0; # store this in desired location 通过写入(store_word)指令 将 $v0的值(5) 存入 内存中 e.g. Print out string (useful for prompts) 栗子: 打印一个字符串(这是完整的,其实上面栗子都可以直接替换main: 部分,都能直接运行) .data string1 .asciiz "Print this.\n" # declaration for string variable, # .asciiz directive makes string null terminated .text main: li $v0, 4 # load appropriate system call code into register $v0; # code for printing string is 4 打印字符串, 赋值对应的操作代码 $v0 = 4 la $a0, string1 # load address of string to be printed into $a0 将要打印的字符串地址赋值 $a0 = address(string1) syscall # call operating system to perform print operation e.g. To indicate end of program, use exit system call; thus last lines of program should be: 执行到这里, 程序结束, 立马走人, 管他后边洪水滔天~~ li $v0, 10 # system call code for exit = 10 syscall # call operating sys -------------------------------------------------我是那个分呀分呀分割线--------------------------------------------------------------------------OK, 十分轻松又愉快的MIPS入门之旅到此告一段落, 下面我把用到的一些软件和这篇文章的原文链接贴到下边,有需要的, 各位客官自取哈~~~ 1.Mars4.4 2.PCSpim Simulator 3.《MIPS Qucik Tutorial》