Comment by zero_iq

In general no, but the provided example depends on parallel memory accesses at the cache level, so cache effects can indeed come into play with instruction-level parallelism. Did you just miss this detail in the article, or are you suggesting it's wrong?