Gromacs on Theta

What is Gromacs?

GROMACS is a versatile package to perform molecular dynamics, i.e. simulate the Newtonian equations of motion for systems with hundreds to millions of particles. It is primarily designed for biochemical molecules like proteins, lipids, and nucleic acids that have a lot of complicated bonded interactions, but since GROMACS is extremely fast at calculating the nonbonded interactions (that usually dominate simulations) many groups are also using it for research on non-biological systems, e.g. polymers.

Using GROMACS at ALCF

ALCF offers assistance with building binaries and compiling instructions for GROMACS. For questions, contact us at support@alcf.anl.gov.

Building Gromacs

1. Download latest source code: http://manual.gromacs.org/documentation/2018.5/download.html
2. Tar xzf gromacs-2018.5.tar.gz
3. cd gromacs-2018.5
4. Mkdir build
5. cd ../
6. Make sure use Intel compiler suite
7. CmakeCC=cc CXX=CC CFLAGS=” -xMIC-AVX512 -g -static-intel” CXXFLAGS=” -xMIC-AVX512 -g -static-intel” - DBUILD_SHARED_LIBS=OFF -DGMX_FFT_LIBRARY=mkl DCMAKE_INSTALL_PREFIX=/PATH-to-Gromacs-Build-directory/ - DGMX_MPI=ON DGMX_OPENMP=ON -DGMX_CYCLE_SUBCOUNTERS=ON -DGMX_GPU=OFF DGMX_BUILD_HELP=OFF -DGMX_HWLOC=OFF -DGMX_SIMD=AVX_512_KNL DGMX_OPENMP_MAX_THREADS=256
8. Make –j 16
9. Make install
10. cd build
11. The Gromacs binary is exe/gmx_mpi

Running Gromacs

A prebuilt Gromacs binary can be found in directory /soft/applications/gromacs

A sample qsub script:

```

!/bin/bash

COBALT -t 30

COBALT --project catalyst

COBALT --attrs mcdram=cache:numa=quad

COBALT -n 1&

COBALT -q debug-cache-quad

export I_MPI_DEBUG=5

export I_MPI_FABRICS=shm

export I_MPI_PIN_MODE=lib

export KMP_AFFINITY=verbose,compact,1

exportGMX_MAXBACKUP=-1

aprun -n64-N64--envOMP_NUM_THREADS=2-ccdepth -d2-j2 /lus/theta-fs0/home/user_name/gromacs-2018.5/build/bin/gmx_mpi\ mdrun -dlbyes-resethway-pinon -v-deffnmstep5_1 -gtest.log ```

We strongly suggest that users try combinations of different numbers of nodes, MPI ranks per node, and OMP threads per rank to find the optimal throughput for the application.

This is a benchmark for a 30,000 atoms system generated on 1 single Theta node with above qprun:

	Core time(sec)	Wall time(sec)
Time	8497.278(ns/day)	12800.0(hour/ns)
Performance	65.078	0.369