Reproducing paper: Benchmarking the Performance of Irregular Computations in AutoDock-GPU Molecular Docking

This repository contains the scripts and additional files for reproducing the results presented in the paper accepted at Parallel Computing:

https://doi.org/10.1016/j.parco.2021.102861

Instructions

1. Program preparation

Clone the AutoDock-GPU v1.3 (used in this paper):

> git clone --depth 1 --branch v1.3 https://github.com/ccsb-scripps/AutoDock-GPU.git
> cd AutoDock-GPU

Set environmental variables (adapt the paths to your systems' installation):

> export GPU_INCLUDE_PATH=/usr/local/cuda/include
> export GPU_LIBRARY_PATH=/usr/local/cuda/lib64

Clone this repository:

> git clone https://github.com/L30nardoSV/reproduce-parcosi-moleculardocking.git

Compile AutoDock-GPU and rename the produced binaries conveniently (for the scripts to work):

> make DEVICE=OCLGPU NUMWI=32
> mv bin/autodock_gpu_32wi bin/autodock_gpu_32wi_ocl
... Do the same for NUMWI = {32, 64, 128, 256}

> make DEVICE=GPU NUMWI=256
> mv bin/autodock_gpu_256wi bin/autodock_gpu_256wi_cuda
... Do the same for NUMWI = {32, 64, 128, 256}

Move above binaries into the test folder:

> cp bin/autodock_gpu_64wi_ocl reproduce-parcosi-moleculardocking/
... Repeat for all OpenCL/CUDA cases above

2. Performance evaluation

Input dataset is provided as a git submodule. Clone that repository automatically:

> ./prepare_inputs.sh

Evaluate the performance for different OpenCL work group sizes:

> ./evaluate_numwi.sh

Evaluate the impact of enabling the autostop option:

> ./evaluate_autostop.sh

Evaluate the impact of the heuristic option:

> ./evaluate_heuristics.sh

Evaluate the impact of combining of both autostop and heuristic options:

> ./evaluate_auto_plus_heur.sh