斐波那契数列64位示例
.data
msg_input: .asciz "Please type a number: "
scanf_fmt : .asciz "%d"
msg_output: .asciz "Fibonacci number %d is %ld\n"
.text
.global _start
_start:
stp x19, x30, [sp, #-16]! // Keep x19 and x30 (link register)
sub sp, sp, #16 // Grow the stack because for a local
// variable used by scanf.
/*
堆栈如下：
Contents Address
| var | [sp] We will use the first 4 bytes for scanf
| | [sp, #8]
| x19 | [sp, #16]
| x30 | [sp, #24]
*/
// 调用printf
ldr x0, addr_msg_input // x0 ← &msg_input [64-bit]
bl printf // call printf
// 调用call
// scanf("%d", &var);
mov x1, sp // x1 ← sp
ldr x0, addr_scanf_fmt // x0 ← &scanf_fmt [64-bit]
bl scanf // call scanf
// 调用fibonacci
// res = fibonacci(var);
ldr w0, [sp] // w0 ← *sp [32-bit]
bl fibonacci // call fibonacci
// Setup call to printf
// printf("Fibonacci number %d is %ld\n", var, res);
mov x2, x0 // x2 ← x0
ldr w1, [sp] // w1 ← *sp [32-bit]
ldr x0, addr_msg_output // x0 ← &msg_output [64-bit]
bl printf // call printf
add sp, sp, #16 // Shrink the stack.
ldp x19, x30, [sp], #16 // Restore x19 and x30 (link register)
mov w0, #0 // w0 ← 0
mov x8, 93
svc 0
fibonacci:
// fibonacci(n) -> result
// n 是 32位 通过w0传递
// 结果是 64位 ，通过 x0传递
stp x19, x30, [sp, #-16]! // Keep x19 and x30 (link register)
stp x20, x21, [sp, #-16]! // Keep x20 and x21
/*
堆栈如下：
| x20 | [sp]
| x21 | [sp, #8]
| x19 | [sp, #16]
| x30 | [sp, #24]
*/
cmp w0, #1 // Compare w0 with 1 and update the flags
ble simple_case // if w0 <= 1 branch to simple_case
// (otherwise continue to recursive_case)
recursive_case: // recursive case
// (this label is not used, added for clarity)
mov w19, w0 // w19 ← w0
// Set up call to fibonacci
// fibonacci(n-1);
sub w0, w0, #1 // w0 ← w0 - 1
bl fibonacci // call fibonacci
mov x20, x0 // x20 ← x0
sub w0, w19, #2 // w0 ← w19 - 2
bl fibonacci // call fibonacci
mov x21, x0 // x21 ← x0
add x0, x20, x21 // x0 ← x20 + x21
b end // (unconditional) branch to end
simple_case:
sxtw x0, w0 // x0 ← ExtendSigned32To64(w0)
end:
ldp x20, x21, [sp], #16 // Restore x20 and x21
ldp x19, x30, [sp], #16 // Restore x19 and x30 (link register)
ret
addr_msg_input: .dword msg_input
addr_msg_output: .dword msg_output
addr_scanf_fmt: .dword scanf_fmt
as -g -o fabo.o fabo.s
ld -g -o fabo fabo.o -lc -I /lib64/ld-linux-aarch64.so.1
斐波那契数列优化单循环64位示例
在上个例子中，对斐波那契数列函数进行优化。
- F0 = 0
- F1 = 1
- Fn = Fn-1 + Fn-2, where n > 1
要计算当前元素，只需要前面的两个元素，并不需要严格遵循数学定义来进行递归函数调用。
可以通过重复计算n-1个先前的斐波那契数来计算任何其他斐波那（n）。
.data
msg_input: .asciz "Please type a number: "
scanf_fmt : .asciz "%d"
msg_output: .asciz "Fibonacci number %d is %ld\n"
.text
.global _start
_start:
stp x19, x30, [sp, #-16]! // Keep x19 and x30 (link register)
sub sp, sp, #16 // Grow the stack because for a local
// variable used by scanf.
/*
堆栈如下：
Contents Address
| var | [sp] We will use the first 4 bytes for scanf
| | [sp, #8]
| x19 | [sp, #16]
| x30 | [sp, #24]
*/
// 调用printf
ldr x0, addr_msg_input // x0 ← &msg_input [64-bit]
bl printf // call printf
// 调用call
// scanf("%d", &var);
mov x1, sp // x1 ← sp
ldr x0, addr_scanf_fmt // x0 ← &scanf_fmt [64-bit]
bl scanf // call scanf
// 调用fibonacci
// res = fibonacci(var);
ldr w0, [sp] // w0 ← *sp [32-bit]
bl fibonacci // call fibonacci
// Setup call to printf
// printf("Fibonacci number %d is %ld\n", var, res);
mov x2, x0 // x2 ← x0
ldr w1, [sp] // w1 ← *sp [32-bit]
ldr x0, addr_msg_output // x0 ← &msg_output [64-bit]
bl printf // call printf
add sp, sp, #16 // Shrink the stack.
ldp x19, x30, [sp], #16 // Restore x19 and x30 (link register)
mov w0, #0 // w0 ← 0
mov x8, 93
svc 0
fibonacci:
// fibonacci(n) -> result
// n is 32-bit and will be passed in w0
// result is 64-bit and will be returned in x0
mov w3, w0 // w3 ← w0
cmp w3, #1 // compare w3 and 1 and update the flags
b.le simple_case // 如果w3小于等于1，则到simple_case
sub w3, w3, #1 // 否则w3 ← w3 - 1
mov x1, #0 // x1 ← 0
mov x2, #1 // x1 ← 1
b loop_check // branch to loop_check
loop_body:
add x0, x1, x2 // x0 ← x1 + x2
mov x1, x2 // x1 ← x2
mov x2, x0 // x2 ← x0
sub w3, w3, #1 // w3 ← w3 - 1
loop_check:
cbnz w3, loop_body // 如果w3不等于0，则到loop_body
b fibonacci_end // branch to fibonacci_end
simple_case:
sxtw x0, w0 // x0 ← ExtendSigned32To64(w0)
fibonacci_end:
ret
addr_msg_input: .dword msg_input
addr_msg_output: .dword msg_output
addr_scanf_fmt: .dword scanf_fmt
as -g -o fabo.o fabo.s
ld -g -o fabo fabo.o -lc -I /lib64/ld-linux-aarch64.so.1
优化后减少了对栈的入侵。
