Multi-Directed Gravitational Wave Search

We just started to distribute the first tasks for the next search for continuous gravitational waves on Einstein@Home. It works similar to the recent all-sky search but also has some new features. Here is an overview:

• We are using a refined dataset based on the data gathered during the first Observational Run (O1) of the LIGO detectors. So it's new datafiles that your computers are downloading. Old datafiles from the all-sky search (and the tuning run) should automatically get deleted from your computer.
• This time we are looking at three specific targets (positions in the sky) namely Cassiopeia A (CasA), Vela Junior (VelaJr) and G347.3 which are all supernova remnants and most likely emitting continuous gravitational waves.
• The search is again divided in two parts because the runtime of a task is again depending on the type of the CPU. Since the all-sky search ended we refined our internal timing and memory model and are now able to better predict runtimes based on CPU model. So we are now sending out CasA and VelaJr tasks to fast computers and tasks for G347.3 to all other computers. But we tailored the tasks so that they run for 8h on average regardless of the CPU model (with exceptions, see below).
• Before starting the real scientific search we are doing a tuning run which will give us valuable information about the data refinement that we did and is only covering a selected frequency range for all targets. This tuning run is expected to run for 2-3 weeks and will have a 7 day deadline. The scientific run thereafter will use a new dataset spanning the whole frequency range from 20 to 1500 Hz.
• You will experience a varying runtime for tasks in the lower frequency range (20 to 100 Hz) which is a new feature of this kind of search. The explanation is simple. The parameter space that we have to search is frequency dependent which means it is more computing intensive to search all parameters in higher frequencies than in lower. So we don't have enough parameters to check in the lower frequencies which means that they are shorter. But that shouldn't affect the runtime estimation of your client as we scale the FLOPS count of these tasks.