我正在关注C代码的这个片段，并且我试图将其翻译为MIPS汇编语言：
for(int i = 0; i < DIM; i++){ /* Row counter */
for(int j = 0; j < DIM; j++){ /* Column counter */
ri = i * DIM + j; /* Calculate R index */
r[ri] = 0; /* Initialize the entry to zero */
for (int k = 0; k < DIM; k++){ /* Now perform the inner product */
ai = i * DIM + k; /* Calculate the A index */
bi = k * DIM + j; /* Calculate the B index */
r[ri] = r[ri] + a[ai] * b[bi]; /* In the lab I used r[ri] += a[ai] * b[bi] */
}
}
}
这是MIPS代码。
在fork之前我需要将条目初始化为零，就像在C代码中一样：r[ri] = 0;但我不知道如何，我尝试使用la $t4, $zero但它不起作用。它说＆＃34;操作数太少或格式不正确＆＃34;。
.text # Indicate that everything below is code
.globl main # The entry point is called main
main: la $t0, DIMSQ # Load the address of DIM*DIM
lbu $t0, 0($t0) # Load the value of [DIM*DIM], re-use $t0
la $t1, DIM # Load the address of DIM
lbu $t1, 0($t1) # Load the value of [DIM], re-use $t1
# Capture the first matrix element by element
la $a0, msg1 # Load the string address
la $a1, MATA # Load the address of A[0]
add $a2, $a1, $t0 # Check dimension + base address
jal input # Call the capture subroutine
# Capture the second matrix element by element
la $a0, msg2 # Load the string address
la $a1, MATB # Load the address of B[0]
add $a2, $a1, $t0 # Check dimension + base address
jal input # Call the capture subroutine
# Perform the sum of Matrix A * Matrix B
xor $t2, $t2, $t2 # i = 0
fori: xor $t3, $t3, $t3 # j = 0
forj: mul $t4, $t1, $t3 # temp = DIM * j
add $t4, $t4, $t2 # temp = j * DIM + i
fork: xor $t9, $t9, $t9 # k = 0
la $t5, MATA # Load the address of A[0]
add $t5, $t5, $t4 # temp += Address of A[0]
lbu $t6, 0($t5) # Load A[temp]
lbu $t7, 9($t5) # Load B[temp]
mul $t6, $t2, $t1
add $t6, $t6, $t9
mul $t7, $t9, $t1
add $t7, $t7, $t3
mul $t6, $t6, $t7 # C[temp] = A[temp] * B[temp]
sb $t6, 18($t5) # Store C[temp]
addi $t3, $t3, 1 # j++
bne $t3, $t1, forj # j < DIM? goto forj
addi $t2, $t2, 1 # i++
bne $t2, $t1, fori # i < DIM? goto fori
addi $t9, $t9, 1
bne $t9, $t9, fork
# Now display the result
la $t1, MATC # Load the address of C[0]
add $t2, $t0, $t1 # Add dimension to the base address
out: li $v0,4 # mesg1 asking for Matrix B element
la $a0, msg3
syscall
lbu $a0, 0($t1) # load the result
li $v0,1 # system call that prints an integer
syscall
li $v0,4 # prints a '\n'
la $a0, cr
syscall
addi $t1, $t1, 1 # increment index pointer
bne $t2, $t1, out # i < DIM?
li $v0,10 # system call 10, exit
syscall
#
#
#
input: move $s0, $a0 # Store the address of the original string
li $v0,4 # Call the print string system call
syscall # $a0 is already loaded
li $v0,5 # system call that reads an integer
syscall
bgtz $v0, validA # check if the input is greater than zero
li $v0,4 #let user know something was not OK
la $a0, error
syscall
move $a0, $s0 # restore the original string
j input # not valid ask for the number again
validA:
sb $v0, 0($a1) # store value for matrix A
addi $a1, $a1, 1 # increment index pointer
bne $a2, $a1, input
jr $ra # return to the caller
#
# These are the space allocations for the three matrices (9 elements each)
# and their dimensions (3 x 3)
#
.data
MATA: .space 9
MATB: .space 9
MATC: .space 9
DIMSQ: .byte 09
DIM: .byte 03
msg1: .asciiz "Enter Matrix A element (positive number): "
msg2: .asciiz "Enter Matrix B element (positive number): "
msg3: .asciiz "Result element: "
error: .asciiz "Error, the number must be positive\n"
cr: .asciiz "\n"