My research requires expertise in several facets of computing, including numerical analysis, information processing, algorithm development, and visualization. My favorite language for general purpose use, including scripting and task automation is Perl. Since being awarded a private ten node (80 CPUs, 160GB RAM) cluster as research startup at Sam Houston State, I have taken an interest in high-performance and high-throughput computing, becoming proficient in use of the Message Passing Interface (MPI) protocols for inter-process communication during massively parallel computations, and also in the use of batch schedulers such as Platform LSF. Both techniques have proven valuable to my study of collider phenomenology.
AEACuS (Algorithmic Event Arbiter and Cut Selection) is a fully-featured package for implementing generic event selection cuts on Monte-Carlo collider-detector simulated event specification files in the standardized .lhco format, which is designed for ease of use, automation, and lightweight installation, targeting high energy theorists and phenomenologists. It tidily facilitates the expression of hierarchical lepton and jet classifications, and a wealth of original algorithms are readily accessible, including all of the standard dimensionful transverse statistics, sophisticated jet, lepton and dilepton object filtering, jet reclustering, most specialized discovery statistics and event shape variables employed by CMS and ATLAS, arbitrary object recombination, and user-defined functions of computable statistics, with detailed cut-flow reporting. Its MT2 implementation appears to be the only public, safe, fully generalized one- and two-step computer, and Christopher Lester (Cambridge), co-developer (with Summers) of the original MT2 variable, has called it "complete in all respects". A presentation from the CERN (Re)interpreting New Physics Searches workshop is available here.
RHADAManTHUS (Recursively Heuristic Analysis, Display and Manipulation: The Histogram Utility Suite) is a companion package for the one- and two-dimensional plotting of collider event statistics, which leverages the graphical functionality of the Python MatPlotLib backend. Any function of variables computed by AEACuS may be used as a histogram key or for secondary event selection. Histogram channels may be arbitrarily merged or transformed bin-by-bin, for example in visualization of signal-to-background significance versus cut threshold. Weighting and recombination of distinct or multiply sampled data sets is handled transparently. A simple card file control syntax facilitates automation and reuse. SHSU students Fantahun and Fernando were additionally helpful in certain aspects of developing and debugging RHADAManTHUS, and they will be co-authors of the forthcoming documentation. A presentation is available here.
Brazos is a single-site online monitoring solution designed for CMS Tier 3 computing centers. It was funded in part by a Norman Hackerman Advanced Research Program grant "Discovery of Dark Matter using High Performance Computing and LHC Data at Texas A&M" in the amount of $100,000, which was awarded to Co-PI's David Toback and Guy Almes of Texas A&M University. My role in this project was the development of new web-based tools for the monitoring of CMS Tier 3 Clusters involved in the massive job of distributed GRID data analysis for the Large Hadron Collider. My students Jacob Hill and Micael Kowalczyk assisted in that work, which was additionally supported by a month-long fellowship at the Fermilab LHC Physics Center. A summary presentation is available here. The monitor remains in use here, although others have now taken over active development. Historical code is available on GitHub or as a tar-zipped download. Contact Dave Toback for the current software version.
Corlim.F95 is code for establishing cross-section expected confidence limits with correlated errors. It is an extension of work by John Conway.