gpu resampling seems to resample twice

[Casey Law]

Running a search with rfgpu and downsampling by 2 will find candidates at integration/2 and integration/4.

[Casey Law]

from astropy import time
import rfpipe
t0 = time.Time.now().mjd
meta = rfpipe.metadata.mock_metadata(t0,t0+10/(24*3600), 27, 4, 32*4, 4, 5e3, datasource='sim', antconfig='D')
mock = [(0, 1923, 0.0, 0.005, 0.18851293482655737, -0.0016288062774844044, 0.0)]
st = rfpipe.state.State(inmeta = meta, inprefs={'dtarr': [1, 2], 'npix_max': 512, 'savecands': False,\
                                                'saveplots' : False,\
                                                'maxdm': 0, 'applyonlineflags': False, 'flaglist': [],\
                                                'clustercands': False, 'simulated_transient': mock, \
                                                'fftmode': 'cuda', 'nthread':25})
cc = rfpipe.pipeline.pipeline_scan(st, devicenum = 1)

[Casey Law]

Then:

integration_ind = cc.array['integration'].copy()
dt_ind = cc.array['dtind'].copy()
time_ind = np.multiply(integration_ind,np.power(2,dt_ind))

And time_ind should all be clustered at roughly same time.

[Casey Law]

This issue seems to be sensitive to the size of the visibility data (in integrations) and the location of the injected transient. It is not sensitive to the DM of the transient and a search with DM=0 and dtarr=[1,2] can reproduce the issue.

[Casey Law]

Fixed in rfgpu