HyGrid: A CPU-GPU Hybrid Convolution-Based Gridding Algorithm in Radio Astronomy

New-generation radio telescopes have been producing an unprecedented scale of data every day and requiring fast algorithms to speedup their data processing work flow urgently. The most data intensive computing phase during the entire work flow is gridding, which converts original data from irregular sampling space to regular grid space. Current methods are mainly focused on interferometers or have limitations on the resolutions due to the memory wall. Here we propose a CPU-GPU hybrid algorithm which accelerates the process of gridding. It employs multi-CPU to perform pre-ordering and GPU to speed up convolution-based gridding. Several optimization strategies are further proposed for reducing unnecessary memory access and maximizing the utilization of the heterogeneous architecture. Testing results demonstrate that the proposal is especially suitable for gridding large-scale data and can improve performance by up to 71.25 times compared to the traditional multi-thread CPU-based approach.

[1]  Justin P. Haldar,et al.  Accelerating MR image reconstruction on GPUs , 2009, 2009 IEEE International Symposium on Biomedical Imaging: From Nano to Macro.

[2]  R. Sinnott.  Virtues of the Haversine , 1984 .

[3]  E. al.,et al.  The Arecibo Legacy Fast ALFA Survey. I. Science Goals, Survey Design, and Strategy , 2005, astro-ph/0508301.

[4]  Andrew Chi-Sing Leung,et al.  HEALPIX DCT technique for compressing PCA-based illumination adjustable images , 2013, Neural Computing and Applications.

[5]  J. D. O'Sullivan,et al.  A Fast Sinc Function Gridding Algorithm for Fourier Inversion in Computer Tomography , 1985, IEEE Transactions on Medical Imaging.

[6]  Hanspeter Pfister,et al.  Enabling a high throughput real time data pipeline for a large radio telescope array with GPUs , 2010, Comput. Phys. Commun..

[7]  Daniel Muscat,et al.  High-Performance Image Synthesis for Radio Interferometry , 2014, 1403.4209.

[8]  T. Boch,et al.  HEALPix Based Cross-Correlation in Astronomy , 2013 .

[9]  D Rosenfeld,et al.  An optimal and efficient new gridding algorithm using singular value decomposition , 1998, Magnetic resonance in medicine.

[10]  John D. Monnier,et al.  Toward 5D image reconstruction for optical interferometry , 2012, Other Conferences.

[11]  Mark R. Calabretta,et al.  Mapping on the HEALPix grid , 2004 .

[12]  R. Nan.  Five hundred meter aperture spherical radio telescope (FAST) , 2006 .

[13]  Z. Paragi,et al.  Pipeline Processing of VLBI Data , 2002 .

[14]  Jason Sanders,et al.  CUDA by example: an introduction to general purpose GPU programming , 2010 .

[15]  A. Maeda,et al.  Reconstruction by weighted correlation for MRI with time-varying gradients. , 1988, IEEE transactions on medical imaging.

[16]  Nathan Bell,et al.  Thrust: A Productivity-Oriented Library for CUDA , 2012 .

[17]  Rob van Nieuwpoort,et al.  Evaluating multi-core platforms for HPC data-intensive kernels , 2009, CF '09.

[18]  J. G. Mangum,et al.  The On The Fly Imaging Technique , 2007, 0709.0553.

[19]  B. D. Jeffs,et al.  The Australian Square Kilometre Array Pathfinder: System Architecture and Specifications of the Boolardy Engineering Test Array , 2014, Publications of the Astronomical Society of Australia.

[20]  T. Vincenty.  DIRECT AND INVERSE SOLUTIONS OF GEODESICS ON THE ELLIPSOID WITH APPLICATION OF NESTED EQUATIONS , 1975 .

[21]  Tim J. Cornwell,et al.  The Noncoplanar Baselines Effect in Radio Interferometry: The W-Projection Algorithm , 2008, IEEE Journal of Selected Topics in Signal Processing.

[22]  K. Golap,et al.  W Projection: A New Algorithm for Wide Field Imaging with Radio Synthesis Arrays , 2005 .

[23]  A. Macovski,et al.  Selection of a convolution function for Fourier inversion using gridding [computerised tomography application]. , 1991, IEEE transactions on medical imaging.

[24]  K. Gorski,et al.  HEALPix: A Framework for High-Resolution Discretization and Fast Analysis of Data Distributed on the Sphere , 2004, astro-ph/0409513.

[25]  R. K. Sinnott.  Design '82 : Institution of Chemical Engineers Symposium Series 76; published by the Institution of Chemical Engineers, Rugby, 1982; 381 pp.; price, £21 , 1984 .

[26]  B. Winkel,et al.  Cygrid: A fast Cython-powered convolution-based gridding module for Python , 2016, 1604.06667.

[27]  Douglas J. Mink.  WCSTools 4.0: Building Astrometry and Catalogs into Pipelines , 2006 .

[28]  Jizhou Sun,et al.  AQUAdexIM: highly efficient in-memory indexing and querying of astronomy time series images , 2016 .

[29]  Justin P. Haldar,et al.  More IMPATIENT: A gridding-accelerated Toeplitz-based strategy for non-Cartesian high-resolution 3D MRI on GPUs , 2013, J. Parallel Distributed Comput..

[30]  Di Li,et al.  The Five-hundred-meter Aperture Spherical radio Telescope (FAST) project , 2011, 2015 International Topical Meeting on Microwave Photonics (MWP).

[31]  Arch D. Robison,et al.  Structured Parallel Programming: Patterns for Efficient Computation , 2012 .

[32]  A. R. Whitney,et al.  The Murchison Widefield Array: The Square Kilometre Array Precursor at Low Radio Frequencies , 2012, Publications of the Astronomical Society of Australia.

[33]  Yashwant Gupta,et al.  A real-time coherent dedispersion pipeline for the giant metrewave radio telescope , 2015, 1509.00186.

[34]  Bruce Merry,et al.  Faster GPU-based convolutional gridding via thread coarsening , 2016, Astron. Comput..

[35]  John W. Romein,et al.  An efficient work-distribution strategy for gridding radio-telescope data on GPUs , 2012, ICS '12.