Starlink Project
Starlink System Notes 79.1

Gaelen Marsden

8th January 2015

Copyright © 2015 University of British Columbia

Further parallelization of iteratemap with MPI


This document describes how to extend the parallelization of the iterative makemap method for use on a distributed-memory cluster using the MPI. A prototype demonstrating the algorithm can be found at


1 Introduction
2 Parallel Algorithm
3 Parallelizing the Models
 3.1 High-pass Filter
 3.2 Commom-mode Model
 3.3 Astronomical Sky Model
 3.4 Other Models
4 Implementation Details
5 Acknowledgments


[1]   Chapin E., et al., Starlink User Note 258, 2013.

[2]   Patanchon G. P., et al., 2008, ApJ, 681, 708, doi:10.1086/588543, arXiv:0711.3462

[3]   Stacey G., et al., 2014, Proc. SPIE 9153, 91530L, doi:10.1117/12.2057101

[4]   Jenness T., et al., 2014, Proc. SPIE, 9152, 91522W, doi:10.1117/12.2056516, arXiv:1406.1515

[5]   Marsden G., et al., 2014, Astronomical Data Analysis Software and Systems XXIII, ASP Conf. Ser., 485, 399, arXiv:1405.0482

[6]   Marsden G., et al., 2014, Astronomical Data Analysis Software and Systems XXIV, ASP Conf. Ser., in press, arXiv:1410.8416

[7]   Chapin E. L., et al., 2013, MNRAS, 430, 2545, doi:10.1093/mnras/stt052, arXiv:1301.3652