[djsdiver]

Hi Arun,

Success!!!!, now working after a few changes. Here's what I did -

1. Upgraded my Windows WSL to WSL2 - don't know why this worked, but my failed attempts were on WSL1. Set the default WSL to WSL2

2. Fresh install of Ubuntu 20.04 LTS and Windows Terminal

3. Set up all the Prerequisites, then sudo update and upgrade

4. Realised I'd been using Eigen 3.4-rc1 on the failed attempts, used Eigen 3.3.9 and the M10 error went away

Here's my full build steps, maybe others who suffer a similar issue can try what worked for me. Thanks for you help, much appreciated.

'Upgrade the windows linux sub system to ver2'

'Goto Windows Powershell as Admin'
dism.exe /online /enable-feature /featurename:Microsoft-Windows-Subsystem-Linux /all /norestart
dism.exe /online /enable-feature /featurename:VirtualMachinePlatform /all /norestart

'download the latest WSL2 kernel and install' https://wslstorestorage.blob.core.wi...update_x64.msi
wsl --set-default-version 2

'Install Ubuntu 20.04 LTS'
https://www.microsoft.com/store/apps/9n6svws3rx71

'Install Windows terminal'
https://docs.microsoft.com/en-us/win...al/get-started

'Get Prerequisites installed and set up'
sudo apt-get update
sudo apt install build-essential
sudo apt install bison
sudo apt install flex
sudo apt install unzip
sudo apt install mpi
sudo apt install mpich
sudo apt install openmpi-bin
sudo apt install openmpi-doc
sudo apt-get update
sudo apt-get upgrade

'for REEF3D and DiveMESH'
mkdir libraries
cd libraries/
cp /mnt/c/Users/User/hypre-master.tar ./
tar -xvf hypre-master.tar
cd hypre-master/
cd src/
./configure --prefix=/usr/local/hypre
make all
sudo make install
cd
echo 'export PATH=/usr/local/hypre/lib:/usr/local/hypre/include:$PATH' >> ~/.bashrc
source .bashrc
cp /mnt/c/Users/User/eigen-3.3.9.tar.gz libraries/
cd libraries
tar -xvf eigen-3.3.9.tar.gz
rm -f eigen-3.3.9.tar.gz hypre-master.tar
cd
cp /mnt/c/Users/User/REEF3D-master.zip ./
cp /mnt/c/Users/User/DIVEMesh-master.zip ./
unzip DIVEMesh-master.zip
unzip REEF3D-master.zip
cd DIVEMesh-master/

'make for all the processors you have on your pc, mine has 12'
make -j 12
cd ..
cd REEF3D-master/
make -j 12
cd

'Test It'

mkdir test1
cp REEF3D-master/bin/reef3d test1/
cp DIVEMesh-master/bin/DiveMESH test1/
cd test1/
'copy the ctrl and control files from Tutorials, check the M 10 n line entry for the processors number. Make sure they're the same and within the available processors on your pc. The 2D Dam Break M 10 number is 4, so we'll use 4. Generates ~ 330MB of simulation results for 25000 steps'
./DiveMESH
mpirun -n 4 reef3d

[kamath]

That is great to know Dean!
Good luck ahead!

[jfbrazao]

Originally Posted by kamath

In addition,
1. also check if the grid-000..dat files are actually produced after you run Divemesh. (using the ls command)

2. Check the permissions and ownerships on the different files and executables (command: ll). your username (command whoami) should own the files, and there should be read, write execute access symbolised by -rwx- at the beginning of the output in front of the Divemesh and Reef3d executables, and atleast read and write -rw- for the grid files.

Hi Arun, after having deleted and reinstalled all the required resources, with latest versions, DiveMESH runs. But despite the M 10 6 being in line with the make of the apps, and with the number of actual processors of my Mac, only one grid-000..dat is created.

Any idea what the origin of this may be? I even did a "brew reinstall mpfr" without errors. The only error I could not address but also not find again, was the "configure: error: Could not determine if REAL*16 bit-matches (...)". See attached.

I have been running the 2017 version to go around this issue, but some properties of the 2021 look very interesting and I would really like to upgrade.

Thank you.

[kamath]

I am not entirely sure from your post which platform you are on. I will assume it is a mac as you use brew.

Then the question is when do you get that error that you state? Possibly when you compile reef3d?

You could add -
no-multibyte-chars

to the CXXFLAGS

if REEF3D is not compiled.
But that does not explain the issue you have with DiveMESH.
I would recommend though that you operate with the latest version.

[jfbrazao]

Hi Arun,

Sorry, should've reported my system:

Model Name: MacBook Pro
Model Identifier: MacBookPro15,1
Processor Name: 6-Core Intel Core i7
Processor Speed: 2.6 GHz
Number of Processors: 1
Total Number of Cores: 6
Memory: 16 GB

The error I previously showed has occurred during the installation of openMPI with Brew, which all-in-all, should end with a perfect installation of Reef3D/DiveMESH... but it doesn't, because I went back using the method described on the manual, and found the source of the error during compiling, after providing the compilers (& their locations or lack thereof):

FCLD libmpi_usempif08.la
ld: library not found for -lhwloc
collect2: error: ld returned 1 exit status
make[2]: *** [libmpi_usempif08.la] Error 1
make[1]: *** [all-recursive] Error 1
make: *** [all-recursive] Error 1

Seems it's a common issue with more recent Macs, as described by the FAQ 92 of openMPI:

https://www.open-mpi.org/faq/?catego...inking-fortran
(Likely to become a common issue, as more users adopt Reef3D in more recent Macs)

It is described that different sources of installation or suites, can cause the failure when looking for libraries, so I've used the same code as in the manual, but specifying the source installation to be used:

./configure --with-hwloc-libdir=/usr/local/lib CC=gcc CXX=g++ F77=gfortran FC=gfortran --prefix=/usr/local/openmpi

This method hasn't worked, the DiveMESH still generates only 1 grd.. gonna try the other methods now, but if you have any advise, it would be welcome.

Thank you.

[kamath]

I have recently installed REEF3D on a MacbookPro 16.1 so this problem should not be connected.

Good news:
I can reproduce your problem with the file you have uploaded.

Bad news: I am not entirely sure why.

Solution:
Firstly (unrelated) remove the capital letters SC at the top. DiveMESH looks for commands when it encounters capital letters and will not execute in their presence.

Finally, somehow removing all the comments associated with the domain, starting with // domain.. // no of processors resulted in the a 3 2 1 decomposition for 6 processors.

Why? Good question! Could have something to do with whitespace characters?

[jfbrazao]

Thank you! Indeed, I ran some tests on a previously compiled DiveMESH and removing the comments did solve the 6 grids.

I also replaced the spacing with TAB for a space - also solved. Then replaced again the space for 1 TAB, then several TABS and hasn't failed since then. There must be a an issue around the spacing characters, but not clear if it's the TAB format..

Lesson learned for now is to remove comments before running. Would you like me to test a bit further?

(this also means I had an issue with the hwloc libraries, but running the previous code fixed that)

[kamath]

Good to know about the hwloc. I have been including it in the Linux compilations and WSL compilations, but had not thought about it for Mac. Probably was available as native previously? I am unsure.

With respect to the spacing and comments: I would recommend not commenting, though it would be a little annoying/difficult when you start off. There seems to be some issue with white space characters but testing each one of them would be quite arduous.
I generally use a single space between letter and number and the following number and a return after a command.
If you really need to make a small note, just use simple text with small letters, no need to to use //

[Taufik S]

Hi Team,
i have problem to run DiveMESH, with notes:

taufik@TaufikS:~/test1$ ls
DiveMESH control.txt ctrl.txt reef3d
taufik@TaufikS:~/test1$ DiveMESH
DiveMESH: command not found

please advise
Thank You

[kamath]

Please use
./
before DiveMESH

The program does not exist in your path. Therefore you have to navigate to the folder where the program exists (which you have done) and then instruct to run the program from the current location (./)

[Taufik S]

Originally Posted by kamath

Please use
./
before DiveMESH

The program does not exist in your path. Therefore you have to navigate to the folder where the program exists (which you have done) and then instruct to run the program from the current location (./)

Hi Kamath,
Thank you for your response, it is works.

[jfbrazao]

Originally Posted by jfbrazao

Seems it's a common issue with more recent Macs, as described by the FAQ 92 of openMPI:

https://www.open-mpi.org/faq/?catego...inking-fortran
(Likely to become a common issue, as more users adopt Reef3D in more recent Macs)

It is described that different sources of installation or suites, can cause the failure when looking for libraries, so I've used the same code as in the manual, but specifying the source installation to be used:

./configure --with-hwloc-libdir=/usr/local/lib CC=gcc CXX=g++ F77=gfortran FC=gfortran --prefix=/usr/local/openmpi

Just a small update on the above. I was able to get a workable installation by using the following configurations:

For OpenMPI:
./configure --with-hwloc=internal --with-libevent=internal CC=gcc CXX=g++ F77=gfortran FC=gfortran --prefix=/usr/local/openmpi

For Hypre:
./configure --prefix=/usr/local/hypre F77=/usr/local/bin/gfortran

Deleting the folders created by previous attempts to install OpenMPI and Hypre was also important (in /usr/local/...). It seems some files are not overwritten, just by reinstalling.