Montblanc: GPU accelerated Radio Interferometer Measurement

Discussion topics for individual codes
Post Reply
Ada Coda
ASCL Robot
Posts: 1487
Joined: Thu May 08, 2014 5:37 am

Montblanc: GPU accelerated Radio Interferometer Measurement Equations in support of Bayesian Inference for Radio Observa

Post by Ada Coda » Tue Feb 03, 2015 2:46 am

Montblanc: GPU accelerated Radio Interferometer Measurement Equations in support of Bayesian Inference for Radio Observations

Abstract: Montblanc, written in Python, is a GPU implementation of the Radio interferometer measurement equation (RIME) in support of the Bayesian inference for radio observations (BIRO) technique. The parameter space that BIRO explores results in tens of thousands of computationally expensive RIME evaluations before reduction to a single X2 value. The RIME is calculated over four dimensions, time, baseline, channel and source and the values in this 4D space can be independently calculated; therefore, the RIME is particularly amenable to a parallel implementation accelerated by Graphics Programming Units (GPUs). Montblanc is implemented for NVIDIA's CUDA architecture and outperforms MeqTrees (ascl:1209.010) and OSKAR.

Credit: Perkins, Simon; Marais, Patrick; Zwart, Jonathan; Natarajan, Iniyan; Tasse, Cyril; Smirnov, Oleg

Site: https://github.com/ska-sa/montblanc
http://adsabs.harvard.edu/abs/2015A%26C....12...73P

Bibcode: 2015ascl.soft02006P

Preferred citation method: http://adsabs.harvard.edu/abs/2015A%26C....12...73P

ID: ascl:1502.006
Last edited by Ada Coda on Wed Feb 06, 2019 7:41 am, edited 1 time in total.
Reason: Updated code entry.

Post Reply