Compiler when gets a statement where a variable is divided by a constant number then it uses multiplication not division instruction along with bit shift.
Example:
#include <stdio.h>
int main()
{
unsigned int a = 13;
unsigned int b;
unsigned int c;
b = a/5;
printf("b = %u \n", b);
c = 13/5;
printf("c = %u \n", c);
return 0;
}
Save the above code as div_const.c
Compile and disassemble for X86 architecture
$ gcc -o div_const div_const.c
$ objdump -d div_const
On x86_64 linux, gcc generate the following assembly.
000000000040052d <main>:
40052d: 55 push %rbp
40052e: 48 89 e5 mov %rsp,%rbp
400531: 48 83 ec 10 sub $0x10,%rsp
400535: c7 45 f4 0d 00 00 00 movl $0xd,-0xc(%rbp)
40053c: 8b 45 f4 mov -0xc(%rbp),%eax
40053f: ba cd cc cc cc mov $0xcccccccd,%edx \
400544: f7 e2 mul %edx | multiplication
400546: 89 d0 mov %edx,%eax | and bit shift
400548: c1 e8 02 shr $0x2,%eax /
40054b: 89 45 f8 mov %eax,-0x8(%rbp)
40054e: 8b 45 f8 mov -0x8(%rbp),%eax
400551: 89 c6 mov %eax,%esi
400553: bf 14 06 40 00 mov $0x400614,%edi
400558: b8 00 00 00 00 mov $0x0,%eax
40055d: e8 ae fe ff ff callq 400410 <printf@plt>
400562: c7 45 fc 02 00 00 00 movl $0x2,-0x4(%rbp) <- 13/5 = 2 placed in c
400569: 8b 45 fc mov -0x4(%rbp),%eax
40056c: 89 c6 mov %eax,%esi
40056e: bf 1d 06 40 00 mov $0x40061d,%edi
400573: b8 00 00 00 00 mov $0x0,%eax
400578: e8 93 fe ff ff callq 400410 <printf@plt>
40057d: b8 00 00 00 00 mov $0x0,%eax
400582: c9 leaveq
Compiler generates instruction at address 0x40053f to 0x400548 for b = a /5. There is no division instruction, only multiplication with constant and bit shift. When compiler finds a statement where a constant is divided with another constant it optimizes the code even further by directly replacing the result without evaluation. For c = 13/5 compiler generates instructions from address 0x400562 to 0x400569. Here direct result of 13/5 is placed in variable c.
For ARM architecture
Compile and disassemble using using cross tool chain
$ arm-linux-gnueabihf-gcc -o div_const_arm div_const.c
$ arm-linux-gnueabihf-objdump -d div_const_arm
000083f0 <main>:
83f0: b580 push {r7, lr}
83f2: b084 sub sp, #16
83f4: af00 add r7, sp, #0
83f6: 230d movs r3, #13
83f8: 607b str r3, [r7, #4]
83fa: 687a ldr r2, [r7, #4]
83fc: f64c 43cd movw r3, #52429 ; 0xcccd \
8400: f6cc 43cc movt r3, #52428 ; 0xcccc | multiplication
8404: fba3 1302 umull r1, r3, r3, r2 | and bit shift
8408: 089b lsrs r3, r3, #2 /
840a: 60bb str r3, [r7, #8]
840c: f248 4088 movw r0, #33928 ; 0x8488
8410: f2c0 0000 movt r0, #0
8414: 68b9 ldr r1, [r7, #8]
8416: f7ff ef68 blx 82e8 <_init+0x20>
841a: 2302 movs r3, #2 <- 13/5=2 placed in c
841c: 60fb str r3, [r7, #12]
841e: f248 4094 movw r0, #33940 ; 0x8494
8422: f2c0 0000 movt r0, #0
8426: 68f9 ldr r1, [r7, #12]
8428: f7ff ef5e blx 82e8 <_init+0x20>
842c: 2300 movs r3, #0
842e: 4618 mov r0, r3
8430: f107 0710 add.w r7, r7, #16
8434: 46bd mov sp, r7
8436: bd80 pop {r7, pc}