Closed JLBLine closed 2 months ago
This is added in 5d23e7f7adf572c9c4168016aa54617d4746ef39. Tested directly in test_kern_calc_autos_multiants.c
and integrated test in test_calculate_visibilities_multibeams.c
where I call calculate_visibilities
with woden_settings->use_dipamps
and woden_settings->do_autos
. This checks whether source_component_common
works as well.
Actually think it makes more sense to make
kern_calc_autos_two_antennas
as this will avoid a(if use_dipamps == 1)
inside a kernel (if statements cost computation times, bad if happening on every thread). I think this will just need up to using a version ofd_ant1_to_baseline_map
andd_ant2_to_baseline_map
as identical arrays, which are just the indexes of the antennas, so 0-127 if we have 128 antennas.Testing: Make a new test
cmake_testing/CUDA_code/source_components/test_kern_calc_autos_multi_antennas.c
. Here we can give a list of beam gains for three antennas, and double check just having a range of 0-127 in the mapping arrays gives a sensible output.