Labels

Branch Instructions and Labels

Do we want to have to compute the value of the offset for a branch instruction? No.

too much work

subject to error

may need to change if instructions are added or removed

better way:

beq  $r1, $r2, L1   # (0) If R[1] == R[2] goto L1

addi $r1, $r1, 1    # (1) R[1]++

addi $r2, $r2, 1    # (2) R[2]++

L1: add  $r1, $r1, $r2  # (3) R[1] = R[1] + R[2]

L1 is a label referring to the address of the instruction where it is located

The assembler computes the offset

target instruction - (branch instruction + 1).

The same rule applies even if you branch backwards

Why compute branch instruction + 1?

Once the instruction is fetched, PC is incremented to PC + 4,

next instruction in memory

This kind of address computation is called PC-relative

Jump Instructions and Labels

Jump instructions also jump to labels (at least, j and jal)

Assembler figures out the appropriate addresses using pseudo-direct addressing.

Disadvantage compared to PC-relative addressing: It's harder to relocate the code