Imported from Bitbucket
updated with the correct chemfp 3.3 timings (had re-used chemfp 1.5 numbers)
re-did the timings using the average time in milliseconds rather than total time in seconds. Updated the plots to use the correct units and change label text.
updated the equations for the fits based on the correct intepretation of the fit parameters

heads

tip
browse log

clone

read-only
https://hg.sr.ht/~dalke/chemfp_paper_reproducibles
read/write
ssh://hg@hg.sr.ht/~dalke/chemfp_paper_reproducibles
This repository contains the details of how to reproduce the counts,
timing numbers, and other part of my chemfp paper.

It is not an automated suite. You will need to run the steps yourself,
and may need to change configurations.

This depends on:
  ~/chemfp-3.3.tar.gz - the current commercial version of chemfp
  ~/chemfp-1.5.tar.gz - the current no-cost/open source version of chemfp
  ~/chemfp_benchmark - the location of the chemfp benchmark

Chemfp is available from chemfp.com .

The benchmark is available from
https://bitbucket.org/dalke/chemfp_benchmark/src . If the benchmark is
not in ~/chemfp_benchmark then you must set the "CFP_BENCHMARK_DIR"
environment variable to point to the correct directory.

Note: chemfp 3.3 contains a bug that causes the build to fail if
OpenMP is not available. The patch files for that version contain a
fix.

The specific details are:

1) Compute the number of intersection popcounts for the t=0.7
benchmark suite.

2) Compute the timings for different popcounts