You wouldn't typically use this when you need lots of random accesses, but when you process the data sequentially.

It also simplifies and speeds up SIMD loads and stores, because you can load/store entire SIMD registers with a continuous, non-strided memory access.