Inconsistent Output Shape of Parameters

[Rachel Bertaud]

Hello all!

We are using VisIt files (post-processed with a custom MATLAB script) to extract quantities such as Div_U, U_x, U_y, and pressure for our simulation. We have noticed that the grid layout of these quantities differs depending on the IBAMR version used.

Seemingly, the older version outputs velocities and divergence defined on a cell-center grid of size N_Finest. In newer versions (built with autoIBAMR), it seems that the divergence is defined on a cell-centered grid of size N_finest x N_finest, whereas other quantities (e.g., velocity and pressure) are defined on a nodal grid of size (N_finest + 1) x (N_finest + 1).

We would like to confirm whether this change in staggering/layout is expected behavior across IBAMR versions, or if it may be related to build configuration, VisIt output settings, or SAMRAI changes. Any insight into whether this behavior is expected, or guidance on how to ensure consistent grid layouts across versions, would be greatly appreciated!

For reference, here are the build environments:

Older Version
1. Built and Run On Colorado School of Mines' HPC Wendian.

• IBAMR-0.11.0
• cmake-3.20.5
• cmake-3.28.4
• hdf5-1.12.1
• hdf5-1.12.2
• libmesh-1.6.2
• numdiff-5.9.0
• petsc-3.17.2
• samrai-2.4.4
• samrai-2024.06.28
• silo-4.11-bsd

Newer Versions
1. Built and Run on a 2022 Macbook Pro:

• IBAMR-0.16.0
• cmake-3.30.6
• hdf5-1.12.2
• libmesh-1.7.8
• numdiff-5.9.0
•  petsc-3.23.3
• samrai-2025.01.09
• silo-4.11-bsd

2. Built and run on Purdue's HPC Anvil:

• IBAMR-0.14.0
• cmake-3.20.5
• cmake-3.28.4
• hdf5-1.12.2
• petsc-3.17.5
• samrai-2024.01.24
• silo-4.11-bsd

Here is the shell script used to setup and run autoibamr on Anvil:

#!/bin/bash 

module load anaconda/2024.02-py311 

module load gcc/11.2.0 openmpi/4.1.6 

export WORK="/anvil/projects/x-mth250056" 

port PROJECT=IBAMR-toolchain 

VERSION=0.14.0 

echo $WORK 

INSTALL_DIR=$WORK/sw/libs 

export TMPDIR=$INSTALL_DIR/tmp 

cd $TMPDIR 

mkdir -p $TMPDIR 

 

 mkdir -p $INSTALL_DIR 

wget https://vault.centos.org/centos/8/BaseOS/x86_64/os/Packages/libtirpc-devel-1.1.4-5.el8.x86_64.rpm 

 

rpm2cpio libtirpc-devel-1.1.4-5.el8.x86_64.rpm | cpio -D $INSTALL_DIR/libtirpc-devel -idmv 

export CXXFLAGS="-I${INSTALL_DIR}/libtirpc-devel/usr/include" 

export CFLAGS="-I${INSTALL_DIR}/libtirpc-devel/lib64" 

export CMAKE_CXX_FLAGS=$CXXFLAGS 

export CMAKE_C_FLAGS=$CFLAGS 

export CMAKE_VERBOSE_MAKEFILE=ON 

 

 INSTALL_DIR=$WORK/sw/apps 

 mkdir -p $INSTALL_DIR 

 cd $INSTALL_DIR 

 

 pwd 

 rm -rf ${INSTALL_DIR}/ibamr 

 cd $TMPDIR 

 git clone https://github.com/IBAMR/autoibamr.git 

 cd autoibamr 

 git checkout ibamr-0.14 

 ./autoibamr.sh --disable-libmesh --prefix=${INSTALL_DIR}

[Boyce Griffith]

We switched the default output for velocity and pressure from cell-centered to node-centered since 0.12.1 (https://github.com/IBAMR/IBAMR/releases/tag/v0.12.1). Using nodal data centerings makes it easier to visualize things like streamlines and isocontours across grid patches and, especially, coarse-fine interfaces.

For the velocity, the actual internal velocity field that is computed internally uses a staggered-grid discretization, and so either node-centered or cell-centered outputs corresponds to an interpolation of the true velocity field. The nodal interpolation gives a representation that plays more nicely with VisIt.

For the pressure, the actual internal pressure field is computed using a cell-centered discretization, and so the nodal representation is different from the internal representation. However, like the velocity, using a nodal representation just results in cleaner visualizations in VisIt.

— Boyce

--
You received this message because you are subscribed to the Google Groups "IBAMR Users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to ibamr-users...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/ibamr-users/3997a229-b422-4460-84e8-679477c0aa9bn%40googlegroups.com.

[Rachel Bertaud]

Dr. Griffith, 

Ah, thank you very much for the direction, information, and your time! 

All the best, 
Rachel