Write the modified algrothm for EREW PRAM semigroup

Q : Semigroup computation on a PRAM: In order to avoid the final broadcasting phase, which essentially doubles the execution time of the semigroup computation algorithm in the EREW PRAM, it has been suggested that we replace the indices j + s in the algorithm with j + s mod p (i.e., allow the computations to wrap around). Under what conditions would this method yield correct results in all processors?

(Hint: what else should change in the algorithm? Check All to all broadcasting algorithm)