Jia Zhao

C2FD

A 3D toolkit for computational complex-fluid dynamics

Introduction

C2FD solves hydrodynamic models of complex fluids in two and three dimensions on hybrid CPU–GPU hardware. The project started in 2008 in Qi Wang's group at the University of South Carolina; I am its maintainer and main active developer. The established models ship as tutorial examples, each with an energy-stable scheme implemented on GPUs, and physical boundary conditions are selected through parameter files.

It contains solvers in 2D and 3D for

  • Navier–Stokes equations with variable density and viscosity
  • Viscoelastic constitutive models: Oldroyd-B, Giesekus, PTT, FENE-P and Rolie-Poly
  • Cahn–Hilliard and Allen–Cahn phase-field equations
  • Convection–diffusion–reaction equations with non-homogeneous diffusion tensors

Project Information

Project leaders
Qi Wang (University of South Carolina) and Jia Zhao
Developers
X. P. Zhao, Y. Z. Gong and Y. C. Zhao
Contributors
P. Seeluangsawat, M. M. Ren and C. Chen

Architecture

Layered diagram of C2FD: solver level on top, kernels and libraries in the middle, programming protocols and CPU–GPU hardware at the bottom
Software stack of C2FD, from the model-level solvers down to the CUDA, MPI and OpenMP layers on hybrid CPU–GPU hardware.
Solvers
Navier–Stokes, gradient-flow, constitutive-equation and general tensor-equation solvers
Kernels
Preconditioned linear solver; Newton–Raphson nonlinear solver
Libraries
CUSP, Thrust, HDF5, cuFFT, cuBLAS
Protocols
OpenMP, MPI, CUDA, C++
Hardware
Hybrid CPU + GPU architecture
Extensions
Fluid–structure interaction by the immersed boundary method; the deal.II finite-element library

Applications

Collage of cell simulations: a wrinkled membrane, a blebbing cell, a dumbbell shape, a dividing cell, and a rounding cell
Applications to cell mechanics: membrane wrinkling, blebbing, a two-lobed cell, division, and mitotic rounding.
Experimental images of F-actin contractile rings above two rows of simulated ring contraction and cell division at four times
Cytokinesis, with the F-actin contractile ring observed in experiments (top) and the simulated ring and dividing cell (below).

Papers with GPU Implementations

Papers whose numerical schemes were implemented on GPUs.

  1. Numerical approximations for the molecular beam epitaxial growth model based on the invariant energy quadratization methodX. Yang*, J. Zhao and Q. Wang · Journal of Computational Physics, 2017 · DOI
  2. A novel linear, second order, unconditionally energy-stable scheme for a hydrodynamic Q-tensor model for liquid crystalsJ. Zhao, X. Yang, Y. Gong and Q. Wang* · Computer Methods in Applied Mechanics and Engineering, 2017 · DOI
  3. A decoupled energy stable scheme for a hydrodynamic phase-field model of mixtures of nematic liquid crystals and viscous fluidsJ. Zhao, X. Yang, J. Shen and Q. Wang* · Journal of Computational Physics, 2016 · DOI
  4. Modeling cytokinesis of eukaryotic cells driven by the actomyosin contractile ringJ. Zhao and Q. Wang* · International Journal for Numerical Methods in Biomedical Engineering, 2016 · DOI
  5. A 3D numerical study of antimicrobial persistence in heterogeneous multi-species biofilmsJ. Zhao, Y. Shen, M. Haapasalo, Z. Wang and Q. Wang* · Journal of Theoretical Biology, 2016 · DOI
  6. Modeling antimicrobial tolerance and treatment of heterogeneous biofilmsJ. Zhao*, P. Seeluangsawat and Q. Wang · Mathematical Biosciences, 2016 · DOI