MF-LBM 1 2 is a high-performance lattice Boltzmann (LB) code for direct numerical simulation (DNS) of flow in porous media, primarily developed by Dr. Yu Chen (LANL), under the supervision of Prof. Albert Valocchi (UIUC), Dr. Qinjun Kang (LANL) and Dr. Hari Viswananthan (LANL). 'MF' refers to microfluidics or 'Magic Find'. The code was first developed at University of Illinois at Urbana-Champaign based on a mainstream LB color-gradient multiphase model and further improved at Los Alamos National Laboratory by implementing the Continuum-Surface-Force and geometrical wetting models to reduce spurious currents so that the inertial effects in scCO2 and brine displacement in porous media can be accounted for 2. Modern manycore processors/coprocessors, such as GPUs and Intel Xeon Phi processors, are developing rapidly and greatly boost computing power. These processors not only provide much higher FLOPS (floating-point operations per second) but also much higher memory bandwidth compared with traditional CPU. One of the most attractive features of the lattice Boltzmann method (LBM) is that it is explicit in time, has nearest neighbor communication, and the computational effort is in the collision step, which is localized at a grid node. For these reasons, the LBM is well suited for manycore processors which require a higher degree of explicit parallelism. The data movement in the LBM is much more intensive than for traditional CFD considering that the D3Q19 lattice model has 19 lattice velocities. Given the current state of computational hardware, in particular the relative speed and capacity of processors and memory, the LBM is a memory-bandwidth-bound numerical method. The high memory bandwidth provided by GPUs or Intel Xeon Phi processors greatly benefits the LBM.