This document uses some microbenchmarks to choose which data structures to use as candidates for detailed performance studies, to replace std::map<int,int> in the SimPhotonsLite data product.

This is the second in my series of physics validation documents for my work on the PDFastSimPAR module. This document looks at the effects of changing the fast_acos algorithm.

This document provides the start of analysis of a naive parallel global minimization based upon random start locations for BFGS local minimization.

This document shows some performance results of using the dlib library’s find_min_global function on the Rastrigin function in 2-5 dimensions.

The purpose of this document is to verify that our timing measurements are well-understood.

This document explores the code generation and performance of expression templates in the context of [*dual numbers*](https://en.wikipedia.org/wiki/Dual_number). In particular, we are concerned with artithmetic operations on dual numbers that occur in the field of automatic differentiation.

A brief and informal description of CosmoSIS integration modules.

This is a brief look at the timing data from ICARUS workflow runs using HEPnOS on the csresearch machines.

This is an example containing a few visualizations of the output of CosmoSIS, using the Metropolis-Hastings sampler with multiple MPI ranks.

The goal of this presentation is to give the flavor of GPU programming, as opposed to both CPU (“normal”) programming and the use of GPU-accelerated libraries or tools (such as are common in the machine learning field). It will not be sufficient for you to go out and start programming. I hope it is sufficient to help you decide whether you are interested in further study of GPU programming to support your own work.

This document contains preliminary analysis of the error estimates returned by the version of the m-Cubes algorithm used in our paper submitted to ISC HP22.

Preliminary analysis of detailed Phase 1 region information, using the DES integrand.

This is a preliminary analysis of the scaling of the eventselection program, reading from HEPnOS.

This is a continued analysis of the June 30 eventselection run, using HEPnOS.

Preliminary analysis of reading speed of PandAna, and the effect of both HDF5 compression and striping of files on the Cori filesystem.

Continued analysis of PandAna performance, this time with a larger dataset and more MPI ranks.

This is a (not yet complete) analysis of the parallel performance of PandAna.

Initial results of snapshot runs.

This document shows the results of using Mathematica 12.1.1 to numerically integrate the Genz_1 (absolute value) function in 5.

Preliminary analysis of timing data from the June 30 HEPnOS run.

Using the integrand Genz_1abs_5d|, this document contains a very preliminary analysis of the subregions produced by the parallel CUHRE algorithm.

This document shows a comparison of the speed of the VEGAS and CUHRE algorithms, as implemented in the CUBA (http://www.feynarts.de/cuba/) library, and wrapped by cubacpp (https://bitbucket.org/mpaterno/cubacpp).

This document shows a performance comparison between the serial and parallel implementations of the CUHRE algorithm for a non-positive-definite integrand.

This document shows where the CUHRE algorithm evaluates the function it is integrating. Unlike a Monte Carlo algorithm (such as VEGAS), CUHRE evaluates the function at a set of determinalistically chosen points.

This document shows a comparison of the speed of the VEGAS and CUHRE algorithms, as implemented in the CUBA (http://www.feynarts.de/cuba/) library, and wrapped by cubacpp (https://bitbucket.org/mpaterno/cubacpp).

Preliminary results for integration of non-negative function in 5d.

Preliminary analysis of reduction data.

This is a preliminary analysis of an eventselection run on Theta using 128 nodes.

This is a public scratchpad for measurements of the evolving performance of PandAna

This document describes measurements of the scaling of PandAna code, using as an example the analysis of NOvA data. It is VERY PRELIMINARY.

We compare the speed of execution of DES Y3 galaxy cluster analysis code, as built two different ways.

A brief study of timing results from the Mu2e Geant4 simulation.

Looking to the cause of very slow running of some MOF jobs.

Demonstrates the use of the output of the _art_ framework's MemoryTracker to diagnose a problem of excessive memory usage.

rgallery is a package for reading HDF5 ntuple files into R.