Chaining off a divide
You've seen divw rD, rA, rB already. It's divide word, a signed rA / rB that keeps no remainder, with rA holding the dividend and rB the divisor. Nothing gets swapped the way subf swaps its operands, so it reads in the obvious order.
A divide is slow. Twenty-plus cycles on PowerPC is typical, far more than a multiply, so a divw always stands out in disassembly. The result behaves like any other, though. It lands in a scratch register, ready for the next instruction.
Here's scaled_div(p, q, r), which multiplies first and then divides.
mullw r0, r3, r4 # r0 = p * q
divw r3, r0, r5 # r3 = r0 / r5 = (p * q) / r
blr
The product p * q lands in r0, and divw divides it by r5. Both operands are already in registers by then. r0 holds the intermediate, and r5 still holds r from entry.
So for your target, spot which argument registers reach divw directly, then check the second instruction for what happens to the quotient.
Your task
Write divadd3, taking three ints, to reproduce the assembly above.