All Questions
261 questions
3
votes
0
answers
42
views
OpenMpi5 Partitioned Communication with multiple neighbors
I've been trying out the partitioned communication that was added in openmpi 5. I am aware that "The current implementation is an early prototype and is not fully compliant with the MPI-4.0 ...
- 31
1
vote
0
answers
65
views
Overlaying openMP onto MPI program causes slow down of the region parallelised with openMP
I have a particle simulation in C which is split over 4 MPI processes and running fast (compared to serial). However, one region of my implementation is N^2 complexity, where I need to compare each ...
- 11
0
votes
1
answer
291
views
Xcode problem while executing C OpenMPI project
On an M3 Pro, in Xcode, I want to run a simple C source file:
#include <mpi.h>
#include <stdio.h>
int main(int argc, char** argv) {
MPI_Init(&argc, &argv);
int world_size;...
1
vote
0
answers
37
views
Custom MPI_Datatype inside OpenMPI MCA module
I'm implementing an Allreduce algorithm inside the mca/coll framework.
The algorithm I'm implementing needs each node to send at each step of the computation only a part of the vector (like a ring ...
0
votes
0
answers
23
views
MPI works with 4 processor but fails with more, problem with decomposition? Send/Recv? 2d Heat diffusion
I have to simulate 2D plate heat diffusion. Everything works fine with 4 tasks, but it doesn't when using N*4 tasks. I suspect the problem is in the logic of communications between the upper and lower ...
0
votes
0
answers
39
views
This MPI_Gatherv code not working and I don't know why
I wrote this code in C that multiplies a matrix by a vector, locally in each process(ylocal) and then the root process gathers the results in its local vector(y). I am using MPI_Gatherv because it has ...
- 1
1
vote
0
answers
90
views
Generate and write to file a random matrix in a distributed, balanced way with MPI
I am trying to generate a random 4x4 matrix in a parallel, "even" way: given the 1D array (representing a matrix) dimension N (e.g. 16 in this case) and the number of processes NP, each ...
- 149
1
vote
0
answers
89
views
C: MPI_ERR_BUFFER: invalid buffer pointer Open MPI
I'm studying for an exam at university about parallel programming with Open MPI.
I'm trying to read a file with the master process (rank 0) and send some data to all workers processes, but I get a ...
2
votes
1
answer
1k
views
How to set openmpi C compiler when installed with conda?
I installed openmpi v4.1.6 (and dependencies) to a clean environment using anaconda conda-forge channel. After installing and attempting to compile, I get the error
------------------------------------...
0
votes
0
answers
94
views
Segmentation Fault When Sending Arrays Over a Certain Size with Open MPI
I am writing a program to run with an ifiniband and intel-based cluster using openmpi, pmix, and SLURM scheduling.
When I run my program on the cluster with an input matrix over 38x38 on each node, I ...
1
vote
1
answer
54
views
Is MPI_Ibcast buffered?
I'm using OpenMPI for a little project, and I've got a doubt.
When I use MPI_Ibcast to send an array to all the other processes, can I start computing on the same array?
Example:
MPI_Ibcast(array, ... ...
- 13
0
votes
1
answer
79
views
MPI Broadcast on subset of MPI_COMM_WORLD results in deadlock
I don't get why the program deadlocks on rank=3 in first iteration. (Possibly different for you).
I want to do some batch processing and if the batch size isn't a multiply of processes I have to ...
- 23
0
votes
0
answers
56
views
mpi application performace get worse when adding more process
i write parallel binary search algorithm with MPI it works as expected in term of searching for a value but when the -n is 1 (serial) the total time is much lower than any value above that like 2, 4, ...
- 13
0
votes
0
answers
108
views
OpenMPI, how can multiple clients connected to one server be one communication?
I want all the clients connected to the server can be one communication.
I have tried merge, but it just merge the intra-comm and one client, and the output is inter-comm which can't be the param of ...
- 21
2
votes
1
answer
69
views
Better way than loop over MPI_Send/MPI_Recv if data not divisable my COMM_SIZE?
I use an MPI program for parallelize batches.
Imaging multidimensional MRI images (3 spatial dimensions, coil data, ...) and we have several of them aggregated on a batch dimension.
The program should ...
- 23