树莓派上关于ARM指令的验证
made by Rk
本文由浙江大学《嵌入式系统》课程提供强力支持。
感谢翁恺老师 @翁恺BA5AG
/*************************************************************/
通过C代码和反汇编工具研究以下几点：
生成了Thumb指令还是ARM指令，如何通过编译参数改变；
对于ARM指令，能否产生条件执行的指令；
设计C的代码场景，观察是否产生了寄存器移位寻址；
设计C的代码场景,观察一个复杂的32位数是如何装载到寄存器的;
写一个C的多重函数调用的程序，观察和分析:
调用时的返回地址在哪里？
传入的参数在哪里？
本地变量的堆栈分配是如何做的？
寄存器是caller保存还是callee保存？是全体保存还是部分保存？
MLA是带累加的乘法，尝试要如何写C的表达式能编译得到MLA指令。
为了实验代码的完整可用，新建一个文件夹arm。
生成了Thumb指令还是ARM指令，如何通过编译参数改变
1、编写下列指令：
#include
int main(){
int t = 0;
t++;
return 0;
} gcc -c arm_1.c
objdump -d arm_1.o可以看到结果如下：
pi@raspberrypi ~/code/arm $ objdump -d arm_1.o
arm_1.o: file format elf32-littlearm
Disassembly of section .text:
00000000 :
0: e52db004 push {fp} ; (str fp, [sp, #-4]!)
4: e28db000 add fp, sp, #0
8: e24dd00c sub sp, sp, #12
c: e3a03000 mov r3, #0
10: e50b3008 str r3, [fp, #-8]
14: e51b3008 ldr r3, [fp, #-8]
18: e2833001 add r3, r3, #1
1c: e50b3008 str r3, [fp, #-8]
20: e3a03000 mov r3, #0
24: e1a00003 mov r0, r3
28: e28bd000 add sp, fp, #0
2c: e8bd0800 pop {fp}
30: e12fff1e bx lr
指令程度为32位，gcc默认使用arm指令集。
3、使用Thumb编译
gcc -c -mthumb -msoft-float arm_1.c pi@raspberrypi ~/code/arm $ objdump -d arm_1.o
arm_1.o: file format elf32-littlearm
Disassembly of section .text:
00000000 :
0: b580 push {r7, lr}
2: b082 sub sp, #8
4: af00 add r7, sp, #0
6: 2300 movs r3, #0
8: 607b str r3, [r7, #4]
a: 687b ldr r3, [r7, #4]
c: 3301 adds r3, #1
e: 607b str r3, [r7, #4]
10: 2300 movs r3, #0
12: 1c18 adds r0, r3, #0
14: 46bd mov sp, r7
16: b002 add sp, #8
18: bd80 pop {r7, pc}
1a: 46c0 nop ; (mov r8, r8) 指令长度是16位（最左一列以2Byte递增）
对于ARM指令，能否产生条件执行的指令；
1、编写arm_branch.c源文件：
#include
int f(int a, int b) {
int t;
if (a > b)
t = a - b--;
if (a == b - 10)
t = a + b++;
return t;
}
int main() {
f(10, 20);
return 0;
} pi@raspberrypi ~/code/arm $ objdump -d arm_branch.o
arm_branch.o: file format elf32-littlearm
Disassembly of section .text:
00000000 :
0: e52db004 push {fp} ; (str fp, [sp, #-4]!)
4: e28db000 add fp, sp, #0
8: e24dd014 sub sp, sp, #20
c: e50b0010 str r0, [fp, #-16]
10: e50b1014 str r1, [fp, #-20]
14: e51b2010 ldr r2, [fp, #-16]
18: e51b3014 ldr r3, [fp, #-20]
1c: e1520003 cmp r2, r3
20: da000006 ble 40
24: e51b2010 ldr r2, [fp, #-16]
28: e51b3014 ldr r3, [fp, #-20]
2c: e0633002 rsb r3, r3, r2
30: e50b3008 str r3, [fp, #-8]
34: e51b3014 ldr r3, [fp, #-20]
38: e2433001 sub r3, r3, #1
3c: e50b3014 str r3, [fp, #-20]
40: e51b3014 ldr r3, [fp, #-20]
44: e243200a sub r2, r3, #10
48: e51b3010 ldr r3, [fp, #-16]
4c: e1520003 cmp r2, r3
50: 1a000006 bne 70
54: e51b2014 ldr r2, [fp, #-20]
58: e51b3010 ldr r3, [fp, #-16]
5c: e0823003 add r3, r2, r3
60: e50b3008 str r3, [fp, #-8]
64: e51b3014 ldr r3, [fp, #-20]
68: e2833001 add r3, r3, #1
6c: e50b3014 str r3, [fp, #-20]
70: e51b3008 ldr r3, [fp, #-8]
74: e1a00003 mov r0, r3
78: e28bd000 add sp, fp, #0
7c: e8bd0800 pop {fp}
80: e12fff1e bx lr
00000084 :
84: e92d4800 push {fp, lr}
88: e28db004 add fp, sp, #4
8c: e3a0000a mov r0, #10
90: e3a01014 mov r1, #20
94: ebfffffe bl 0
98: e3a03000 mov r3, #0
9c: e1a00003 mov r0, r3
a0: e8bd8800 pop {fp, pc} 可以看到bne等条件分支指令的出现，可见ARM是支持的条件执行指令的。
设计C的代码场景，观察是否产生了寄存器移位寻址；
1、编译生成测试代码：
#include
int main() {
int t[10];
int i = 0;
int j = 0;
for(j=0; j<4; j++) {
t[8] = t[i*2];
}
printf("%d", t[8]);
return 0;
}
2、查看编译结果：
pi@raspberrypi ~/code/arm $ objdump -d arm_shift.o
arm_shift.o: file format elf32-littlearm
Disassembly of section .text:
00000000 :
0: e92d4800 push {fp, lr}
4: e28db004 add fp, sp, #4
8: e24dd030 sub sp, sp, #48 ; 0x30
c: e3a03000 mov r3, #0
10: e50b300c str r3, [fp, #-12]
14: e3a03000 mov r3, #0
18: e50b3008 str r3, [fp, #-8]
1c: e3a03000 mov r3, #0
20: e50b3008 str r3, [fp, #-8]
24: ea00000b b 58
28: e51b300c ldr r3, [fp, #-12]
2c: e1a02083 lsl r2, r3, #1
30: e3e0302f mvn r3, #47 ; 0x2f
34: e1a02102 lsl r2, r2, #2
38: e24b1004 sub r1, fp, #4
3c: e0812002 add r2, r1, r2
40: e0823003 add r3, r2, r3
44: e5933000 ldr r3, [r3]
48: e50b3014 str r3, [fp, #-20]
4c: e51b3008 ldr r3, [fp, #-8]
50: e2833001 add r3, r3, #1
54: e50b3008 str r3, [fp, #-8]
58: e51b3008 ldr r3, [fp, #-8]
5c: e3530003 cmp r3, #3
60: dafffff0 ble 28
64: e59f201c ldr r2, [pc, #28] ; 88
68: e51b3014 ldr r3, [fp, #-20]
6c: e1a00002 mov r0, r2
70: e1a01003 mov r1, r3
74: ebfffffe bl 0
78: e3a03000 mov r3, #0
7c: e1a00003 mov r0, r3
80: e24bd004 sub sp, fp, #4
84: e8bd8800 pop {fp, pc}
88: 00000000 .word 0x00000000 lsl r2, r3, #1设计C的代码场景,观察一个复杂的32位数是如何装载到寄存器的;
1、编写测试代码
#include
int main() {
unsigned int a = 0x12345678;
a++;
return 0;
}
2、编译查看结果：
pi@raspberrypi ~/code/arm $ objdump -d arm_big_num.o
arm_big_num.o: file format elf32-littlearm
Disassembly of section .text:
00000000 :
0: e52db004 push {fp} ; (str fp, [sp, #-4]!)
4: e28db000 add fp, sp, #0
8: e24dd00c sub sp, sp, #12
c: e59f3020 ldr r3, [pc, #32] ; 34
10: e50b3008 str r3, [fp, #-8]
14: e51b3008 ldr r3, [fp, #-8]
18: e2833001 add r3, r3, #1
1c: e50b3008 str r3, [fp, #-8]
20: e3a03000 mov r3, #0
24: e1a00003 mov r0, r3
28: e28bd000 add sp, fp, #0
2c: e8bd0800 pop {fp}
30: e12fff1e bx lr
34: 12345678 .word 0x12345678
可以看到ARM指令将大数字存在命令段，通过ldr指令去加载，而非MIPS等指令集使用lui和ori来实现32位大数字的实现。
写一个C的多重函数调用的程序，观察和分析:
调用时的返回地址在哪里？
传入的参数在哪里？
本地变量的堆栈分配是如何做的？
寄存器是caller保存还是callee保存？是全体保存还是部分保存？
1、编写测试代码
#include
int f2(int a, int b, int c, int x, int y, int z, int p, int q, int r) {
int temp = 0;
temp = a*x + b*y + c*z + p*q +r;
return temp;
}
int f3(int a,int b,int c,int d,int e,int f,int g,int h,int i,int j,int k)
{
int t1=a+b;
int t2=c+d;
int t3=e+f;
int t4=g+h;
int t5=i+j;
f2(1,2,3,4,5,6,7,8,9);
int t6=t1*t2;
int t7=t3*t4;
int t8=t6-t7;
int t9=t8*t5*k;
return t9;
}
int f1(int a, int b, int c) {
int temp = 0;
temp = f2(a, b, c, 4, 5, 6, 7, 8, 9);
return temp;
}
int main() {
int a,b,c;
a = 0;
b = 1;
c = 2;
printf("%d", f1(a,b,c));
f3(11,10,9,8,7,6,5,4,3,2,1);
return 0;
}
2、编译查看结果
pi@raspberrypi ~/code/arm $ gcc -c arm_multi_call.c
pi@raspberrypi ~/code/arm $ objdump -d arm_multi_call.o
arm_multi_call.o: file format elf32-littlearm
Disassembly of section .text:
00000000 :
0: e52db004 push {fp} ; (str fp, [sp, #-4]!)
4: e28db000 add fp, sp, #0
8: e24dd01c sub sp, sp, #28
c: e50b0010 str r0, [fp, #-16]
10: e50b1014 str r1, [fp, #-20]
14: e50b2018 str r2, [fp, #-24]
18: e50b301c str r3, [fp, #-28]
1c: e3a03000 mov r3, #0
20: e50b3008 str r3, [fp, #-8]
24: e51b3010 ldr r3, [fp, #-16]
28: e51b201c ldr r2, [fp, #-28]
2c: e0020392 mul r2, r2, r3
30: e51b3014 ldr r3, [fp, #-20]
34: e59b1004 ldr r1, [fp, #4]
38: e0030391 mul r3, r1, r3
3c: e0822003 add r2, r2, r3
40: e59b300c ldr r3, [fp, #12]
44: e59b1010 ldr r1, [fp, #16]
48: e0010391 mul r1, r1, r3
4c: e51b3018 ldr r3, [fp, #-24]
50: e59b0008 ldr r0, [fp, #8]
54: e0030390 mul r3, r0, r3
58: e0813003 add r3, r1, r3
5c: e0822003 add r2, r2, r3
60: e59b3014 ldr r3, [fp, #20]
64: e0823003 add r3, r2, r3
68: e50b3008 str r3, [fp, #-8]
6c: e51b3008 ldr r3, [fp, #-8]
70: e1a00003 mov r0, r3
74: e28bd000 add sp, fp, #0
78: e8bd0800 pop {fp}
7c: e12fff1e bx lr
00000080 :
80: e92d4800 push {fp, lr}
84: e28db004 add fp, sp, #4
88: e24dd050 sub sp, sp, #80 ; 0x50
8c: e50b0030 str r0, [fp, #-48] ; 0x30
90: e50b1034 str r1, [fp, #-52] ; 0x34
94: e50b2038 str r2, [fp, #-56] ; 0x38
98: e50b303c str r3, [fp, #-60] ; 0x3c
9c: e51b2030 ldr r2, [fp, #-48] ; 0x30
a0: e51b3034 ldr r3, [fp, #-52] ; 0x34
a4: e0823003 add r3, r2, r3
a8: e50b3008 str r3, [fp, #-8]
ac: e51b2038 ldr r2, [fp, #-56] ; 0x38
b0: e51b303c ldr r3, [fp, #-60] ; 0x3c
b4: e0823003 add r3, r2, r3
b8: e50b300c str r3, [fp, #-12]
bc: e59b2004 ldr r2, [fp, #4]
c0: e59b3008 ldr r3, [fp, #8]
c4: e0823003 add r3, r2, r3
c8: e50b3010 str r3, [fp, #-16]
cc: e59b200c ldr r2, [fp, #12]
d0: e59b3010 ldr r3, [fp, #16]
d4: e0823003 add r3, r2, r3
d8: e50b3014 str r3, [fp, #-20]
dc: e59b2014 ldr r2, [fp, #20]
e0: e59b3018 ldr r3, [fp, #24]
e4: e0823003 add r3, r2, r3
e8: e50b3018 str r3, [fp, #-24]
ec: e3a03005 mov r3, #5
f0: e58d3000 str r3, [sp]
f4: e3a03006 mov r3, #6
f8: e58d3004 str r3, [sp, #4]
fc: e3a03007 mov r3, #7
100: e58d3008 str r3, [sp, #8]
104: e3a03008 mov r3, #8
108: e58d300c str r3, [sp, #12]
10c: e3a03009 mov r3, #9
110: e58d3010 str r3, [sp, #16]
114: e3a00001 mov r0, #1
118: e3a01002 mov r1, #2
11c: e3a02003 mov r2, #3
120: e3a03004 mov r3, #4
124: ebfffffe bl 0
128: e51b3008 ldr r3, [fp, #-8]
12c: e51b200c ldr r2, [fp, #-12]
130: e0030392 mul r3, r2, r3
134: e50b301c str r3, [fp, #-28]
138: e51b3010 ldr r3, [fp, #-16]
13c: e51b2014 ldr r2, [fp, #-20]
140: e0030392 mul r3, r2, r3
144: e50b3020 str r3, [fp, #-32]
148: e51b201c ldr r2, [fp, #-28]
14c: e51b3020 ldr r3, [fp, #-32]
150: e0633002 rsb r3, r3, r2
154: e50b3024 str r3, [fp, #-36] ; 0x24
158: e51b3024 ldr r3, [fp, #-36] ; 0x24
15c: e51b2018 ldr r2, [fp, #-24]
160: e0030392 mul r3, r2, r3
164: e59b201c ldr r2, [fp, #28]
168: e0030392 mul r3, r2, r3
16c: e50b3028 str r3, [fp, #-40] ; 0x28
170: e51b3028 ldr r3, [fp, #-40] ; 0x28
174: e1a00003 mov r0, r3
178: e24bd004 sub sp, fp, #4
17c: e8bd8800 pop {fp, pc}
00000180 :
180: e92d4800 push {fp, lr}
184: e28db004 add fp, sp, #4
188: e24dd030 sub sp, sp, #48 ; 0x30
18c: e50b0010 str r0, [fp, #-16]
190: e50b1014 str r1, [fp, #-20]
194: e50b2018 str r2, [fp, #-24]
198: e3a03000 mov r3, #0
19c: e50b3008 str r3, [fp, #-8]
1a0: e3a03005 mov r3, #5
1a4: e58d3000 str r3, [sp]
1a8: e3a03006 mov r3, #6
1ac: e58d3004 str r3, [sp, #4]
1b0: e3a03007 mov r3, #7
1b4: e58d3008 str r3, [sp, #8]
1b8: e3a03008 mov r3, #8
1bc: e58d300c str r3, [sp, #12]
1c0: e3a03009 mov r3, #9
1c4: e58d3010 str r3, [sp, #16]
1c8: e51b0010 ldr r0, [fp, #-16]
1cc: e51b1014 ldr r1, [fp, #-20]
1d0: e51b2018 ldr r2, [fp, #-24]
1d4: e3a03004 mov r3, #4
1d8: ebfffffe bl 0
1dc: e50b0008 str r0, [fp, #-8]
1e0: e51b3008 ldr r3, [fp, #-8]
1e4: e1a00003 mov r0, r3
1e8: e24bd004 sub sp, fp, #4
1ec: e8bd8800 pop {fp, pc}
000001f0 :
1f0: e92d4810 push {r4, fp, lr}
1f4: e28db008 add fp, sp, #8
1f8: e24dd034 sub sp, sp, #52 ; 0x34
1fc: e3a03000 mov r3, #0
200: e50b3010 str r3, [fp, #-16]
204: e3a03001 mov r3, #1
208: e50b3014 str r3, [fp, #-20]
20c: e3a03002 mov r3, #2
210: e50b3018 str r3, [fp, #-24]
214: e59f4078 ldr r4, [pc, #120] ; 294
218: e51b0010 ldr r0, [fp, #-16]
21c: e51b1014 ldr r1, [fp, #-20]
220: e51b2018 ldr r2, [fp, #-24]
224: ebfffffe bl 180
228: e1a03000 mov r3, r0
22c: e1a00004 mov r0, r4
230: e1a01003 mov r1, r3
234: ebfffffe bl 0
238: e3a03007 mov r3, #7
23c: e58d3000 str r3, [sp]
240: e3a03006 mov r3, #6
244: e58d3004 str r3, [sp, #4]
248: e3a03005 mov r3, #5
24c: e58d3008 str r3, [sp, #8]
250: e3a03004 mov r3, #4
254: e58d300c str r3, [sp, #12]
258: e3a03003 mov r3, #3
25c: e58d3010 str r3, [sp, #16]
260: e3a03002 mov r3, #2
264: e58d3014 str r3, [sp, #20]
268: e3a03001 mov r3, #1
26c: e58d3018 str r3, [sp, #24]
270: e3a0000b mov r0, #11
274: e3a0100a mov r1, #10
278: e3a02009 mov r2, #9
27c: e3a03008 mov r3, #8
280: ebfffffe bl 80
284: e3a03000 mov r3, #0
288: e1a00003 mov r0, r3
28c: e24bd008 sub sp, fp, #8
290: e8bd8810 pop {r4, fp, pc}
294: 00000000 .word 0x00000000 可以发现：
函数调用后的返回地址存放在LR寄存器中。
传入的参数存放在R0 R1 R2 R3四个寄存器中，多余的参数放在堆栈中。
本地变量存放在堆栈高地址，传进来的参数存放在堆栈低地址。
R0到R3由caller保存，R4以上由callee保存。
MLA是带累加的乘法，尝试要如何写C的表达式能编译得到MLA指令。
1、编写函数如下：
#include
int f1(int a, int b, int c) {
return a*b + c;
}
int main() {
f1(1,2,3);
return 0;
}
2、编译查看结果：
可以看到MLA命令得到了使用。