这段汇编代码有什么问题？

Question

这是在用 Verilog 编写的 5 级流水线 MIPS 上运行的机器代码（mem 是其指令存储器）：

mem[0]  <= 32'h3c040000;  // lui  $4, 0

mem[1]  <= 32'h3409000a;  // ori  $9, $0, 10 

mem[2]  <= 32'h340a0000;  // ori  $10, $0, 0 

mem[3]  <= 32'h8c880000;  // lw   $8, 0($4) 

mem[4]  <= 32'h208b0028;  // addi $11, $4, 40 

mem[5]  <= 32'h208c002c;  // addi $12, $4, 44 

mem[6]  <= 32'had680000;  // sw   $8, 0($11) 

mem[7]  <= 32'h8c8d0000;  // lw   $13, 0($4) 

mem[8]  <= 32'h010d082a;  // slt  $1, $8, $13 

mem[9]  <= 32'h10200004;  // beq  $1, $0, 16

mem[10] <= 32'h000d4021;  // addu $8, $0, $13 

mem[11] <= 32'had6d0000;  // sw   $13, 0($11) 

mem[12] <= 32'had8a0000;  // sw   $10, 0($12) 

mem[13] <= 32'h20840004;  // addi $4, $4, 4 

mem[14] <= 32'h214a0001;  // addi $10, $10, 1 

mem[15] <= 32'h2129ffff;  // addi $9, $9, -1 

mem[16] <= 32'h1520fff7;  // bne  $9, $0, -36

这段代码应该找到 10 个元素数组中的最大数字及其索引，并将它们存储在紧邻数组之后的数据存储器中。这是我的数据存储器：

Memory[0] <= 32'h5;

Memory[1] <= 32'h1;

Memory[2] <= 32'hc;

Memory[3] <= 32'h7;

Memory[4] <= 32'h2;

Memory[5] <= 32'h8;

Memory[6] <= 32'hb;

Memory[7] <= 32'h4;

Memory[8] <= 32'h6;

Memory[9] <= 32'h3;

// end of array

Memory[10] <= 32'h0; // the greatest number in data memory

Memory[11] <= 32'h0; // index of the greatest number in memory 

为什么这段汇编代码不起作用？

运行后，数据内存一定是这样的：

// Memory[10] is equal to 32'hc  (the greatest number in array)
// Memory[11] is equal to 32'h2  (the array index of the greatest number)

但是我得到了：

// Memory[10] is equal to 32'h8  (the greatest number in array)
// Memory[11] is equal to 32'h2  (the array index of the greatest number)

这是写入序列（控制台日志）：

WriteData = 00000005 (greatest number)
WriteData = 00000005 (greatest number)
WriteData = 00000001 (greatest number)
WriteData = 0000000c (greatest number)
WriteData = 00000002 (index of greatest number)
WriteData = 00000007 (greatest number)
WriteData = 00000002 (greatest number)
WriteData = 00000008 (greatest number)

Answer 1

在下面，您能看到与机器代码的区别吗？分支的立即数减少了 1？

从这段代码开始：

    .set noat
    .text
    lui $4, 0
    ori $9, $0, 10
    ori $10, $0, 0
    lw $8, 0($4)
    addi $11, $4, 40
    addi $12, $4, 44
    sw $8, 0($11)
Loop1:
    lw $13, 0($4)
    slt $1, $8, $13
    beq $1, $0, EndIf1
    addu $8, $0, $13
    sw $13, 0($11)
    sw $10, 0($12)
EndIf1:
    addi $4, $4, 4
    addi $10, $10, 1
    addi $9, $9, -1 
    bne $9, $0, Loop1

MARS 生成以下内容：

3c040000
3409000a
340a0000
8c880000
208b0028
208c002c
ad680000
8c8d0000
010d082a
10200003   <----- this is the forward branch for the if-then statement
000d4021
ad6d0000
ad8a0000
20840004
214a0001
2129ffff
1520fff6   <----- this is the backward branch for the do/while loop

火星截图：

---

这是使用延迟分支并默认关闭/未选中设置的 QtSpim：