In order to catch possible performance regressions and to optimize parsing we need to gather timings.