Shared Memory-based Algorithms for MPI Irregular Collective Operations

Algorithms for MPI_Scatterv, MPI_Gatherv and MPI_Allgatherv collective operations using the shared memory of SMP/NUMA machine are proposed. The algorithms use an approach based on copying message fragments through queues in a shared memory segment (Copy-In-Copy-Out). The implementation is based on O...

Celý popis

Uloženo v:
Podrobná bibliografie
Vydáno v:2022 IEEE International Multi-Conference on Engineering, Computer and Information Sciences (SIBIRCON) s. 1 - 4
Hlavní autoři: Romanyuta, Alexey, Kurnosov, Mikhail
Médium: Konferenční příspěvek
Jazyk:angličtina
Vydáno: IEEE 11.11.2022
Témata:
Allgatherv
collective operations
computer systems
Gatherv
Libraries
MPI
Scatterv
Servers
Software
Software algorithms
On-line přístup:Získat plný text
Tagy: Přidat tag
Žádné tagy, Buďte první, kdo vytvoří štítek k tomuto záznamu!
Popis
Shrnutí:Algorithms for MPI_Scatterv, MPI_Gatherv and MPI_Allgatherv collective operations using the shared memory of SMP/NUMA machine are proposed. The algorithms use an approach based on copying message fragments through queues in a shared memory segment (Copy-In-Copy-Out). The implementation is based on Open MPI and on average reduces the execution time by 20-40% for the MPI_Scatterv, MPI_Gatherv, MPI_Allgatherv operations compared to the coll/tuned component of the Open MPI library.
DOI:10.1109/SIBIRCON56155.2022.10017005