-
This sequence (trivial modification of `testrd_par.c`):
```
exoid = ex_open_par("test.exo", /* filename path */
EX_READ, /* access mode = READ */
…
-
Related: https://github.com/kubeflow/training-operator/issues/2040
As we discussed multiple times, Kubeflow community are looking for examples on how to use MPI Operator and [DeepSpeed](https://git…
-
Work on the post processing scripts for Vernier.
Script ideas/examples:
Merge script - Merge several Vernier files together into one (combining sub region times from each MPI rank).
Diff script -…
-
In `read_inp.F` there is a parameter `max_fname` which defaults to 256. However, when using MPI, the following `MPI_Bcast` call cuts this to 64, which I am exceeding by using absolute paths.
In my lo…
-
At the moment, we do not include MPI in the wheels since it would require the user to also install MPI on the system.
Since it is not trivial to provide wheels with and without MPI, we could think …
-
### Problem Description
I ran my code on Frontier for scaling on AMD GPUS. It scaled fine with MPI . But as soon as i replace the MPI_Alltoall call with nccl_Alltoall, it is behaving way worse than M…
-
I'm seeing some strange performance issues with the GROMACS test on our system. I.e. occasionally, it just runs 10 times slower. Looking at `htop`, I see individual cores not being used - even though …
-
Hello,
I am trying to install PeTar on my MacBook Pro M2 2022 (macOS Sonoma 14.4.1), but I am having issue when I try to 'make'.
I did : `./configure --with-fdps-prefix=$HOME/Desktop/PBH/code/FD…
-
Such a module could export functionality through which parallel programs could switch parallel programming models at compile-time based on generic interfaces:
```
module parallelism_m
!! Use comp…
-
Hello,
I've been using specfem2d with MPI and until now it has worked correctly. Now that I trying to move to other softwares specialized in inversion like adjtomo/seisflows that uses specfem2d, I h…