GPU-based autocorrelator to run on VLBI disk based Data Recorders. Written in CUDA
Features
- The GPU autocorrelator, G-AUTO, is a GPU implementation of an autocorrelator designed orginally for the Long Baseline Array (LBA), the Australian VLBI Network. It allows the formation of a autocorrelation computed on a GPU installed in the LBA Data Recorder (LBADR), thereby limiting the load on the CPU. The LBADR is a computer-based disk-based data recorder, which is fully loaded whilst capturing and storaging the VLBI data. By transferring the computation to the GPU we are able to form both the autocorrelations and the cross-correlations for dual polarisation data collected at the maximum data rate of the LBA, 512Mbps, using a nVIDIA GT-9800. G-AUTO can read LBA and Mk5B data files or e-VLBI port connections, can form autocorrelations of upto 0.5 million channels, can write out the data as a FITS file and form all four stokes products all within real time.
Follow GPU Autocorrelator for VLBI
You Might Also Like
Rate This Project
Login To Rate This Project
User Reviews
-
Its a good first step. We have tested it by sending data from a Mk5B using "jive5ab" and the command "in2net" to stream the data real time over a research network between New Zealand and Australia. 4 channels with 16MHz bandwidth.