Luckily the world seems to be so that the things that have to be fast also happen to be relatively easy to parallelize (most tasks using huge data sets, renderi

Luckily the world seems to be so that the things that have to be fast also happen to be relatively easy to parallelize (most tasks using huge data sets, rendering, matrix operations, ...). What are applications that have to be fast but are not easy to parallelize?