DLB can now turn off, when slower
Under certain conditions, especially with (shared) GPUs, DLB can
decrease the performance. We now measure the cycles per step before
turning on DLB. When the running average of cycles per step with DLB
gets above the average without DLB, we turn off DLB. We then measure
again without DLB. DLB can then turn on again. If we turn on DLB of
DLB multiple times in close succesion and we measure performance loss,
we keep DLB off for the remainder of the run. This procedure ensures
that the performance will never deteriorate due to DLB.
Updated and expanded the DLB section in the manual.
Change-Id: I6e0291c1a41adf6da94fae46d36e0fcb95585a02