In reply to MayurKubavat:
There is no race. Your conversion to remove the intra-assignment delay is not equivalent to the original code and yours does have a race. The LRM section I mention above states that if the ordering of the assignments is determinate, the outcome is determinate.