gkiar/fuzzafq

Procedure to follow:

0. Download or build singularity container

   $ module load singularity $ $ singularity pull docker://gkiar/fuzzy:pyafq $ $ # OR, if this doesn't work on your system, follow a guide for direct-builds $ # from downloaded Docker images, such as the one shown here: $ # https://docs.computecanada.ca/wiki/Singularity

1. Run the setup script on a head-node (it involves downlaoding data)

   $ module load singularity $ $ # Obviously, update paths as needed for your system $ cd /project/6049200/afq-pytracer/ $ $ singularity exec -B ${PWD} -B ${HOME} fuzzyafq.sif python3 gkiar-fuzzafq/code/afq.py --setup

2. Get a resource allocation (or script the rest and submit it to a queue)

   $ # Similarly, change resource request based on your affiliations, the system, etc.. $ salloc --account rrg-glatard --mem=12000 --time=2:0:0

3. Enter singularity environment

   $ module load singularity $ $ singularity exec -B ${PWD} -B ${HOME} fuzzyafq.sif /bin/bash

4. Configure the MCA backend as you wish

   $ export VFC_BACKENDS="libinterflop_ieee.so" # OR $ export VFC_BACKENDS="libinterflop_mca.so -m rr"

5. Run the script with unique params to avoid overwriting results

   $ # In the case of IEEE backend use 'ref', or just a numeric ID for MCA $ run_id="ref" $ $ # This will hold symlinks to the original dataset and the pipeline outputs $ dpath="/project/6049200/afq-pytracer/data/" $ $ python3 gkiar-fuzzafq/code/afq.py -p ${dpath} -i ${run_id}