4.1. BranchΒΆ

The unconditional branch instruction allows us to skip to a new location to continue executing code from. When the branch executes, the program counter is not increased by 4 bytes as normal, but instead its value is replaced with the address of a new instruction.

Blabel

Branch. Change the program counter to the address of the line identified by label.

In this code sample, line number 3 has a branch that will cause execution to skip over the code on lines 4 and 5 and execute the next line after the label target, which is line 7.

1
2
3
4
5
6
7
   MOV   r1, #1    @random work
   MOV   r2, #2
   B     target    @branch to target
   MOV   r3, #3    @skipped
   MOV   r4, #4    @skipped
target:
   MOV   r5, #5
Try sample

Note

The assembler calculates the destination address and stores it in the instruction as +/- value to add to the PC register when the branch executes. This offset is stored as 24-bits in the instruction:

313029282726252423222120191817161514131211109876543210
cond 101 L 24-bit offset

Since instruction addresses in ARM are always multiples of 4, the last two bits of any instruction address are always 00. This fact is used to compress the stored offsets - the value stored in the instruction has been right shifted 2 bits to remove those 0s. To use the stored address, the hardware first shifts it to the left 2 bits and then sign extends it. For example, in the code sample above, the instruction looks like:

313029282726252423222120191817161514131211109876543210
11101010000000000000000000000001
EA000001

If we mask the last 24 bits and then shift them left 2 and sign extend, we get:

313029282726252423222120191817161514131211109876543210
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0

The final offset value is 4. To that value, we have to add 8. (This add 8 is due to the historical design of the ARM architecture where the PC will already have advanced 2 instructions or 8 bytes before the branch address is actually used). So the final amount for the branch is 12 bytes, or at 4 bytes per instructions, 3 instructions ahead. Sure enough, the shown branch skips over 2 instructions and runs the third one after the B.

You have attempted of activities on this page