Translate the following into IA64 code, making use of predication.

for (i = 0; i < 100: i++)

{ if (X[i] = Y[i])

Y[i] = Y[i] + 1;

else X[i] = X[i] – 1;

}

---

mov ar.lc = 100 //set up loop count in lc register

ld4 r2 = [r10],4 ;; //get x[i] in r2

ld4 r3 = [r11],4 ;; //get y[i] in r3

cmp.eq p1,p2 = r2,r3 ;; //compare x[i] and y[i]

(p1) add r3 = r3,1 //if part

(p2) sub r2 = r2,1 ;; //else part

(p1) st4 [r11] = r3,4 //copy to destination (if)

(p2) [r10] = r2,4 //copy to destination (else)

br.cloop next ;; //decrement lc register and branch

Note that we can do the if and else parts in parallel by using the true/false predication. Note also that there
is little room for parallel operation exception the two if/else parts which are mutually exclusive.