Problem $6$

Consider the following segment of a program. Assume all registers were initialized to $0$ before

execution.

ADD R$1,$ R$1,$ #$3$

MUL R$2,$ R$1,$ #$4$

DIV R$2,$ R$2,$ #$6$

ADD R$3,$ R$3,$ #$9$

SUB R$2,$ R$3,$ R$1$

This code is run on a pipelined, out$-$of$-$order machine with $4$ functional units $($add$,$ subtract,

multiply, divide$).$ The add and subtract functional units take $1$ cycle to execute. The multiply

functional unit takes $4$ cycles to execute, and the divide functional unit takes $5$ cycles to

execute.

Depict a data$-$flow diagram of the program segment.

How many cycles would this program segment take to run, if the machine were utilizing

the original Tomasulo algorithm, i$.$e$.$ no Reorder Buffer.

How many cycles would this program segment take to run, if the machine were utilizing

the improved Tomasulo algorithm, i$.$e$.$ there is a Reorder Buffer.

Compare your answers to questions $2$ and $3.$ Is the original version faster? Is it closer in

execution to your data$-$flow diagram from question $1?$ If so$,$ why do we prefer the

improved version? Is the problem of the original algorithm showcased in this program

segment?