Using the Protein MD Setup tutorial using BioExcel Building Blocks (biobb) --AmberTools package version-- with Amber installed frome sources

BioBB
1

Hi,
I’ve used the Protein MD Setup tutorial using BioExcel Building Blocks (biobb)
–AmberTools package version–
following the instructions:
git clone https://github.com/bioexcel/biobb_wf_amber_md_setup.git
cd biobb_wf_amber_md_setup
conda env create -f conda_env/environment.yml
conda activate biobb_AMBER_MDsetup_tutorials
jupyter-notebook biobb_wf_amber_md_setup/notebooks/mdsetup/biobb_amber_setup_notebook.ipynb

I have also installed AmberTools23 and Amber22 on the same system ( UBUNTU 22.04 ) compiling both from the sources with -DCUDA=TRUE -DOPENMP=TRUE.

It is possible to run the notebook using the sander and pmemd that have been compiled from the
sources so to use all threads and the gpu?

I have run the notebook and analyzing htop I have noticed that only one thread has been used during all the processing and I guess that the gpu was never used.

Thanks.

Saverio

2

Hi Saverio,

yes, for this cases you can use the binary_path property, which lets you use a locally installed software instead of the one packaged with the BioBB module. You’ll need to first identify the corresponding binaries (e.g. pmemd.CUDA or pmemd.MPI).

You should also change the sander_mdrun block for the pmemd_mdrun block if using pmemd binary instead of the sander one.

Hope this helps!

3

Ok.
Thanks.

Saverio