I have a very simple test case in a 1x1 2D square domain, with a prescribed horizontal velocity. I have 3 individual lagrangian detectors -- (0.5,0.0), (0.5,0.5), (0.5,1.0).
If I run this case in serial, the positions of these detectors, after each timestep, are output into the .detector file correctly. However, if I run this simulation in parallel, on only 2 cores, this is not the case - the position of the first detector is output correctly, but all other written detector positions are erroneuous.
I have followed individual detector positions through the code, and their positions are updated correctly, implying that this is strictly an issue with parallel detector IO. In other words, a write statement of detector positions at line 2787 of subroutine write_mpi_out in Diagnostic_variables.F90 (I am on the latest trunk version) suggests that the individual detectors have been advected to the correct positions. However, when writting to the .detector file, this information seems to get lost/mangled.
I have attached everything that is required to reproduce this bug (prescribed.flml is the serial case, parallel_prescribed.flml is the 2 processor case). If something is unclear please let me know.
Hi Rhodri,
Thanks for reporting this bug and providing the test case. The problem was that the names of lagrangian detectors were being overwritten during initialisation, causing I/O to fail, which I have fixed in the attached branch (lp:~fluidity-core/fluidity/fix-lagr-detectors).
Hope this helps
Michael