Changelog#

Version 0.4.0#

Released on: 08/03/2026

Added pylops_mpi.Distributed.DistributedMixIn class with communicator-agnostic calls to communication methods.
Added pylops_mpi.utils._mpi with implementations of MPI communication methods.
Added kind=”summa” implementation in pylops_mpi.basicoperators.MPIMatrixMult operator.
Added kind paramter to all operators in pylops_mpi.basicoperators.MPILaplacian
Added cp.cuda.Device().synchronize() before any MPI call when using Cuda-Aware MPI.
Modified pylops_mpi.utils._nccl.initialize_nccl_comm to handle nodes with more GPUs than ranks
Fixed bug in pylops_mpi.DistributedArray.__neg__ by explicitely passing base_comm_nccl during internal creation of distributed array

Released on: 05/08/2025

Added pylops_mpi.basicoperators.MPIMatrixMult operator.
Added NCCL support to all operators in pylops_mpi.basicoperators, and pylops_mpi.signalprocessing.
Added base_comm_nccl in constructor of pylops_mpi.DistributedArray, to enable NCCL communication backend.
Added pylops_mpi.utils.benchmark subpackage providing methods to decorate and mark functions / class methods to measure their execution time.
Added pylops_mpi.utils._nccl subpackage implementing methods for NCCL communication backend.
Added pylops_mpi.utils.deps subpackage to safely import nccl
Fixed partition in the creation of the output distributed array in pylops_mpi.signalprocessing.MPIFredholm1.

Released on: 16/12/2024

Added support for using CuPy arrays with PyLops-MPI.
Introduced the pylops_mpi.signalprocessing.MPIFredholm1 and pylops_mpi.waveeqprocessing.MPIMDC operators.
Allowed the UNSAFE_BROADCAST partition to give users an option to handle overflow in broadcast scenarios.
Added a dottest function to perform dot tests on PyLops-MPI operators.
Created a tutorial for Multi-Dimensional Deconvolution (MDD).

Released on: 13/04/2024

Released on: 28/08/2023