Problem:: Running the examples using the parallel version

[mo7ammedmostafa]

Hello my name Kamra,
I came across your code while looking a mpi parallel research that I can use to test new schemes and algorithm.
I downloaded the code and managed to compile it but it would seem that there is a problem with mesh reader when reading the boundary file for examples cavity and pitzDaily
I get the following error
`At line 436 of file mesh_geometry_and_topology.f90 (unit = 7, file = 'processor0/constant/polyMesh/boundary')
Fortran runtime error: Bad integer for item 2 in list input

Error termination. Backtrace:
At line 436 of file mesh_geometry_and_topology.f90 (unit = 7, file = 'processor1/constant/polyMesh/boundary')
Fortran runtime error: Bad integer for item 2 in list input

Error termination. Backtrace:
#0 0x7ff209cd131a
#1 0x7ff209cd1ec5
#2 0x7ff209cd268d
#3 0x7ff209e3e924
#4 0x7ff209e41c1a
#5 0x7ff209e430f9
#6 0x55a6efbdfb0b
#7 0x55a6efbd61e8
#8 0x55a6efbb006e
#9 0x7ff209101b96
#10 0x55a6efbb00a9
#11 0xffffffffffffffff
#0 0x7f504480731a
#1 0x7f5044807ec5
#2 0x7f504480868d
#3 0x7f5044974924
#4 0x7f5044977c1a
#5 0x7f50449790f9
#6 0x559a1359eb0b
#7 0x559a135951e8
#8 0x559a1356f06e
#9 0x7f5043c37b96
#10 0x559a1356f0a9
#11 0xffffffffffffffff

Primary job terminated normally, but 1 process returned
a non-zero exit code.. Per user-direction, the job has been aborted.

---

mpiexec detected that one or more processes exited with non-zero status, thus causing
the job to be terminated. The first process to do so was:

Process name: [[49949,1],1]
Exit code: 2
--------------------------------------------------------------------------`

So do I just need to modify the boundary file from OpenFOAM format to you native format?

THANKS
Kamra

[nikola-m]

Hi Kamra. Well yes, we use a bit different layout for boundary files, as in serial version. If you don't manage to make it work I will send you correct ones.on monday,ok? I am working on a new version of the code,so I didn't update the github repo in a while. I will keep you updated. What are you interested to investigate in general?

…

On 31 May 2019 09:56, "mo7ammedmostafa" ***@***.***> wrote: Hello my name Kamra, I came across your code while looking a mpi parallel research that I can use to test new schemes and algorithm. I downloaded the code and managed to compile it but it would seem that there is a problem with mesh reader when reading the boundary file for examples cavity and pitzDaily I get the following error `At line 436 of file mesh_geometry_and_topology.f90 (unit = 7, file = 'processor0/constant/polyMesh/boundary') Fortran runtime error: Bad integer for item 2 in list input Error termination. Backtrace: At line 436 of file mesh_geometry_and_topology.f90 (unit = 7, file = 'processor1/constant/polyMesh/boundary') Fortran runtime error: Bad integer for item 2 in list input Error termination. Backtrace: #0 0x7ff209cd131a #1 <#1> 0x7ff209cd1ec5 #2 <#2> 0x7ff209cd268d #3 <#3> 0x7ff209e3e924 #4 0x7ff209e41c1a #5 0x7ff209e430f9 #6 0x55a6efbdfb0b #7 0x55a6efbd61e8 #8 0x55a6efbb006e #9 0x7ff209101b96 #10 0x55a6efbb00a9 #11 0xffffffffffffffff #0 0x7f504480731a #1 <#1> 0x7f5044807ec5 #2 <#2> 0x7f504480868d #3 <#3> 0x7f5044974924 #4 0x7f5044977c1a #5 0x7f50449790f9 #6 0x559a1359eb0b #7 0x559a135951e8 #8 0x559a1356f06e #9 0x7f5043c37b96 #10 0x559a1356f0a9 #11 0xffffffffffffffff Primary job terminated normally, but 1 process returned a non-zero exit code.. Per user-direction, the job has been aborted. ------------------------------ mpiexec detected that one or more processes exited with non-zero status, thus causing the job to be terminated. The first process to do so was: Process name: [[49949,1],1] Exit code: 2 ------------------------------------------------------------ --------------` So do I just need to modify the boundary file from OpenFOAM format to you native format? THANKS Kamra — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#3?email_source=notifications&email_token=AAD7LDYQLKSINQXA3ICG4QTPYDK23A5CNFSM4HRVWEIKYY3PNVWWK3TUL52HS4DFUVEXG43VMWVGG33NNVSW45C7NFSM4GW4UAIA>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AAD7LD5RPMXL2LZGEV3N2YLPYDK23ANCNFSM4HRVWEIA> .

[mo7ammedmostafa]

I am planning to implement a new scheme for high order flux Reconstruction
Based on a quick read of the code it looks like the necessary data structure will work but I wanted to evaluate the parallel efficiency because most of my test problems are 3d
I already wrote my own code for icnomp navier stokes on unstructured grid but its parallel efficiency is terrible, so it will take a long time to optimize it so i was looking for code as similar to my own as possible but with better performance to use and learn from

[nikola-m]

I will be able to help you out. But algorithm layout will be different for your case. It will probably be explicit code, based on Runge-Kutta. The starting point should maybe be simple second order 3D explicit Euler or NS code based on sequence reconstruction (gradient evaluation)- fluxes-evolution. I am interested in higher order methods but didn't have time to do anything about it. I will send you dev version of the code soon as well.

…

On 1 Jun 2019 15:13, "mo7ammedmostafa" ***@***.***> wrote: I am planning to implement a new scheme for high order flux Reconstruction Based on a quick read of the code it looks like the necessary data structure will work but I wanted to evaluate the parallel efficiency because most of my test problems are 3d — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#3?email_source=notifications&email_token=AAD7LDY26V3O2TBAJMWN7XLPYJYYHA5CNFSM4HRVWEIKYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGODWXANAA#issuecomment-497944192>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AAD7LD63ADA6ROMSMI2L33TPYJYYHANCNFSM4HRVWEIA> .

[mo7ammedmostafa]

I understand the difference between the two algorithms, but what I am interested in is the parallel efficiency of the implementation and data structure.

[nikola-m]

I didn't forget about you, at the moment I'm changing the mesh reading function that will result in a bit different spec of boundary data, and the way they are structured. The way boundary faces are looped is changed too. I don't have access to a bigger machine at the monent, but I will get some cpu hours somewhere to test the code. As with many other codes the main bottleneck in scalability are dot products and residul calculation in solution of linear systems (mpi_allreduce). This is what I got so far, but further testing will be necessary. I want to do hybrid mpi/openmp during the year, where the openmp parallelisation won't be on loop level, but also of spmd domain decomposition type, just focused on shared storage on the node. Do you have any thoughts on this? Are you at some university or in industry? Regards

…

On 14 Jun 2019 14:56, "mo7ammedmostafa" ***@***.***> wrote: I understand the difference between the two algorithms, but what I am interested in is the parallel efficiency of the implementation and data structure. — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#3?email_source=notifications&email_token=AAD7LD2PP3YUPAM2LMADKWLP2OIRHA5CNFSM4HRVWEIKYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGODXWWYUA#issuecomment-502099024>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AAD7LD3THRWBQJQWJ2YCUXDP2OIRHANCNFSM4HRVWEIA> .