mpi4py-fft: Parallel Fast Fourier Transforms with MPI for Python

Mortensen, Mikael; Dalcín, Lisandro; Keyes, David E.

doi:10.21105/joss.01340

Cited by 5 publications

(3 citation statements)

References 7 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Afterward, we present our support for the latest revision of the MPI-3 standard. 5 With the rise of GPU (graphic processing unit) programming and CUDA-aware MPI implementations, we discuss how this model is supported in mpi4py. Following that, we present some recent mpi4py features, such as a new package for asynchronous task execution and support for efficient picklebased communication of large-size buffer-like Python objects.…”

Section: The Message Passing Interface (Mpi)mentioning

confidence: 99%

“…MPI4Dask [4] is planned to replace dask_mpi and to be integrated into dask.distributed as a communication backend. mpi4py-fft [5] is a Python package for distributed, multi-dimensional Fast Fourier Transforms (FFTs) [6], which is used in shenfun [7] to solve PDEs using the spectral Galerkin method and in FluidFFT [8] as a backend option to support a unified API for parallel FFT computations.…”

Section: Applicationsmentioning

confidence: 99%

See 1 more Smart Citation

mpi4py: Status Update After 12 Years of Development

2021

Self Cite

View full text Add to dashboard Cite

MPI for Python (mpi4py) has evolved to become the most used Python binding for the Message Passing Interface (MPI). We report on various improvements and features that mpi4py gradually accumulated over the past decade, including support up to the MPI-3.1 specification, support for CUDA-aware MPI implementations, and other utilities at the intersection of MPI-based parallel distributed computing and Python application development.

show abstract

Section: The Message Passing Interface (Mpi)mentioning

confidence: 99%

Section: Applicationsmentioning

confidence: 99%

mpi4py: Status Update After 12 Years of Development

2021

Self Cite

View full text Add to dashboard Cite

show abstract

“…As a result, parallel FFTs are computed using sequential (linear) transformations over undivided axes and global array redistributions (using interprocess communication) that realign the arrays for further serial transforms. [13][14] MPI for Python was developed on top of the MPI-1 specification, which specified an object-oriented interface that closely matched the MPI-2 bindings and facilitated communications of general Python objects in its first version. This kit has been updated to support nearly all MPI-2 functionality as well as direct blocking/non-blocking communication with numeric arrays.…”

Section: Message Passing Interface (Mpi)mentioning

confidence: 99%

Effects of Combining Multi-Core Processors in Creating Super-Computers Using Raspberry Pis

2021

IJATCSE

View full text Add to dashboard Cite

The vast resources of old working computers are getting bigger and bigger. Companies that consider these old computers as obsolete but still working put these computers to the storage, or worse, junk. Different entities are more and more relying on supercomputers to replace their ordinary and old computers. Supercomputers have become a basic necessity in this generation. It has a wide variety of functions that make our daily work easier. Having a well-built supercomputer with excellent specifications is a must; however, it comes with a cost. The computers can be clustered to create a supercomputer. The computers used in these clustered supercomputers have the exact specifications, such as uniform models with the same number of cores per processor and the same memory capacity. This study focuses on creating a supercomputer using different processors with a different number of cores per processor, processor speed, and different memory capacity. This study can be a valuable system for entities that have old computers in their workplace. They can reuse their old computers and combine them with other computers to create a supercomputer, thus reducing waste and recycling old resources. This study will be using different types of Raspberry Pi (RPi) computers combining old and new models to simulate the scenario of combining multi-core processors in creating supercomputers.

show abstract