*****Contact Akhil Langer at alanger@illinois.edu (or akhilanger@gmail.com) for any questions****
1. Changes in the Makefile
a) set the CHARMHOME variable to the path of charm directory
Compiling the code:
1. make
2. for debugging: make OPTS=-g -O0
Running the code:
./charmrun +p8 ++local ./advection arg1 arg2 arg3
where arg1 = max_depth of the mesh
arg2 = #zones in one dimension per chare
arg3 = #iterations for which the code is to be run
arg4 = load balancing frequency (should be a multiple of 3, 3 is the refinement frequency)
arg5(optional) = intial mesh size (set to 128 by default)
READING ORDER
1. advection.ci - read the entry methods - iterate, remeshing, doComputationStep
2. QuadIndex.h - contains the bit Vector based 3D Octtree implementaion
provides interface to get neighbor indices, child indices, etc.
3. Advection.C - contains implementation of boundary exchange functions, iterpolation,
finite state machine for granularity decision consensus, refinement and
coarsening functions, etc.
Citation:
@inproceedings{langer2012scalable,
title={Scalable Algorithms for Distributed-Memory Adaptive Mesh Refinement},
author={Langer, Akhil and Lifflander, Jonathan and Miller, Phil and Pan, Kuo-Chuan and Kale, Laxmikant V and Ricker, Paul},
booktitle={Computer Architecture and High Performance Computing (SBAC-PAD), 2012 IEEE 24th International Symposium on},
pages={100--107},
year={2012},
organization={IEEE}
}
*****Contact Akhil Langer at alanger@illinois.edu (or akhilanger@gmail.com) for any questions****