userMESO 2.5 is a highly optimized hybrid CPU/GPU particle simulation package. This package has a code structure design that aligns with LAMMPS, and is highly customizable. The package's programming language is CUDA C/C++ with MPI and OpenMP.
userMESO 2.5
(Dated Aug 2018) userMESO 2.5 (https://github.com/AnselGitAccount/USERMESO-2.0-mdpd) is an updated version of userMESO 2.0. This version was developed by Ansel Blumers (then intern at Idaho National Laboratory and PhD student at Brown University) and Yidong Xia (then at Idaho National Laboratory). Major updates in this version include the implementation of 1) the many-body DPD (mDPD) model for multi-fluid/phase flow [1] and 2) a no-slip and no-penetration boundary condition for fluid flow in an arbitrarily complex geometry [2] for the mDPD model.
Reference
[1] https://aip.scitation.org/doi/abs/10.1063/1.4981136 [2] https://www.sciencedirect.com/science/article/pii/S0021999117308525
userMESO 2.0
(Dated Apr 2017) userMESO 2.0 (https://github.com/AnselGitAccount/USERMESO-2.0) is an updated version of the original userMESO code. This version was developed by Ansel Blumers and Yu-Hang Tang (then at Brown University). Now it is possible to simulate advection, diffusion, and reaction processes with dissipative particle dynamics (tDPD). Another major upgrade is the ability to simulate red blood cells. Combining tDPD and the red blood cell model, the simulation of the chemical-releasing process from the red blood cells becomes realizable.
The details regarding code implementation can be found at https://www.sciencedirect.com/science/article/pii/S0010465517301042
userMESO
(Dated Oct 2016) userMESO (https://github.com/yhtang/MESO) was developed by Yu-Hang Tang (then at Brown University) to simulate molecular dynamics, classic dissipative particle dynamics, and smoothed particle dynamics. It integrates several algorithmic innovations that take advantage of CUDA devices:
- An atomic-free warp-synchronous neighbor list construction algorithm;
- A 2-level particle reordering scheme, which aligns with the cell list lattice boundaries for generating strictly monotonic neighbor list;
- A locally transposed neighbor list;
- Redesigned non-branching transcendental functions (
$\sin$ ,$\cos$ , pow,$\log$ ,$\exp$ , etc.); - An overlapped pairwise force evaluation and halo exchange using CUDA streams for hiding the communication and the kernel launch latency;
- Radix sorting with GPU stream support;
- Pairwise random number generation based on per-timestep binary particle signatures and the prepriority Tiny Encryption Algorithm.
The details regarding code implementation can be found at https://www.sciencedirect.com/science/article/pii/S0010465514002203
==Important: do not install Nvidia GPU drivers with your Linux distro's software installer ==
If you happen to have done so, we suggest you remote them. Follow Nvidia's official instruction to install CUDA Toolkit (e.g. version 9.2) including the GPU driver. Then check if the GPU(s) can be detected.
cd ~/NVIDIA_CUDA-9.2_Samples/1_Utilities/deviceQuery
make
./deviceQuery
Set the required environment variales.
export PATH=/usr/local/cuda/bin:${PATH}
export LD_LIBRARY_PATH=/usr/local/cuda/lib64:$LD_LIBRARY_PATH
export LIBRARY_PATH=/usr/local/cuda/lib64:$LIBRARY_PATH
Compile the source code as follows. To know the compute capability version of your GPU, check https://en.wikipedia.org/wiki/CUDA
cd <code_repository>/src
make yes-molecule
make yes-user-meso
make meso ARCH=[sm_30|sm_35|sm_52|sm_60|sm_61|sm_70|...]
Simulation of fluid flow in a quasi-2D nanoporous domain (e.g. using 2 GPUs and 2 MPI ranks with 1 GPU/rank and 2 threads/rank)
cd <code_repository>/examples/v2.5/quasi_2d_pores
OMP_NUM_THREADS=2 mpirun -np 2 ../../../src/lmp_meso -i in.quasi_2d_pores
Simulation of a red blood cell in fluid.
cd <working_copy>/exmaple/simple
../../src/lmp_meso -in tDPD_RBC_spec1_Single_GPU.in
Benchmark of RBC suspension in a single node. The simulations of different system volumes are timed.
cd <working_copy>/example/single_node_benchmark
./run_file.sh
| Chemical-release of Red Blood Cells in a Microfluidic Device |
|---|
 |
In addition to the source files in userMESO (program summary URL: http://cpc.cs.qub.ac.uk/summaries/AETN_v1_0.html), the following source files are included in userMESO 2.0 package.
The files in this section are essential to the simulation of transport dissipative particle dynamics. atom_vec_tdpd_atomic_meso.cu/.h
These files contain the tdpd class declaration and implementation.
pair_tdpd_meso.cu/.h
These files compute the pairwise interactions which includes forces and concentration fluxes.
fix_nve_tdpd_meso.cu/.h
These files performs constant energy and volume integration to update position, velocity, and concentration for atoms in each timestep.
The files in this section are solely needed to compute red blood cell dynamics. atom_vec_tdpd_rbc_meso.cu/.h
These files contain the tdpd class declaration and implementation that are specifically designed for red blood cells.
angle_area_volume_meso.cu/.h
These files compute the angle term in the red blood cell model.
bond_wlc_pow_all_visc_meso.cu/.h
These files compute the bond term in the red blood cell model.
dihedral_bend_meso.cu/.h
These files compute the dihedral term in the red blood cell model.
compute_concent_tdpd_meso.ch/.h
These files sum the concentration of each species over all particles.
dump_tdpd_meso.cu/.h
These files prints the coordinates and concentrations to an output file.
fix_addconf_tdpd_meso.cu/.h
These files increases the concentration by a constant.
fix_resetconc_tdpd_meso.cu/.h
These files reset the concentration to a value.