{"id":18421958,"url":"https://github.com/spcl/muliticast-based-allgather","last_synced_at":"2025-04-14T20:21:45.836Z","repository":{"id":254119561,"uuid":"845550627","full_name":"spcl/muliticast-based-allgather","owner":"spcl","description":null,"archived":false,"fork":false,"pushed_at":"2025-02-12T14:43:25.000Z","size":3480,"stargazers_count":12,"open_issues_count":0,"forks_count":3,"subscribers_count":7,"default_branch":"master","last_synced_at":"2025-03-28T08:40:47.900Z","etag":null,"topics":[],"latest_commit_sha":null,"homepage":null,"language":"C","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":null,"status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/spcl.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":null,"funding":null,"license":null,"code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":null,"support":null,"governance":null,"roadmap":null,"authors":null,"dei":null,"publiccode":null,"codemeta":null}},"created_at":"2024-08-21T13:18:11.000Z","updated_at":"2025-02-24T02:59:05.000Z","dependencies_parsed_at":"2024-08-21T14:59:31.399Z","dependency_job_id":null,"html_url":"https://github.com/spcl/muliticast-based-allgather","commit_stats":null,"previous_names":["spcl/muliticast-based-allgather"],"tags_count":2,"template":false,"template_full_name":null,"repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/spcl%2Fmuliticast-based-allgather","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/spcl%2Fmuliticast-based-allgather/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/spcl%2Fmuliticast-based-allgather/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/spcl%2Fmuliticast-based-allgather/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/spcl","download_url":"https://codeload.github.com/spcl/muliticast-based-allgather/tar.gz/refs/heads/master","host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":248952408,"owners_count":21188441,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2022-07-04T15:15:14.044Z","host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":[],"created_at":"2024-11-06T04:27:25.647Z","updated_at":"2025-04-14T20:21:45.816Z","avatar_url":"https://github.com/spcl.png","language":"C","funding_links":[],"categories":[],"sub_categories":[],"readme":"# Bandwidth-Optimal, Fully-Offloaded Collectives\n\nThe paper will be presented at the SC '24 (The International Conference for High Performance Computing, Networking, Storage, and Analysis).\n\n## Artifact structure\n\n - CPU-driven Multicast-based Allgather/Broadcast integrated within the UCC library:\n    - `./ucc/src/components/tl/spin` - collectives backend source-code.\n    - `./osu-micro-benchmarks-7.3` - OSU MPI benchmarks supporting per-iteration timing for Allgather/Broadcast collectives\n    - `./ompi/` and `./ucx`- subrepos with project dependencies \n - DPA-offloaded collective receive datapath PoC:\n    - `./coll-offloading/*`\n    - `./coll-offloading/README.md` - detailed manual for building/running PoC.\n - Multicast-based Allgather traffic model:\n    - `./sim/estimate_allgather_cost.py`\n\n## Building and running UCC-based collectives backend\n\n**Prerequisites**: InfiniBand-based cluster and user permissions to create multicast-groups through OpenSM.\n\n1) `$ git submodule update --init --recursive`\n2) `$ bash ./build.sh`\n3) `$ source ./sourceme.sh`\n4) `$ cd ./osu-micro-benchmarks-7.3/ \u0026\u0026 ./configure CC=mpicc CXX=mpicxx \u0026\u0026 make -j`\n\nOpenMPI 5.0.3 compiled with the support of UCC multicast backend will be stored in `./ompi/build`. Multicast-based backend can be tested by running:\n\n`mpirun -x LD_LIBRARY_PATH -x PATH --hostfile \u003cmachine_hostfile\u003e -np \u003cnprocs\u003e --mca coll_ucc_enable 1 --mca coll_ucc_priority 100 ./osu-micro-benchmarks-7.3/c/mpi/collective/blocking/osu_allgather -m 4096:8388608`\n\n## DPA offloading\n\n**Prerequisites**: Two servers with BlueField-3 NIC and DOCA v2.2.0.\n\nSee `./coll-offloading/README.md` for building and running instructions instructions.\n\n## Traffic reduction with Multicast\n\nRun `python3 ./sim/estimate_allgather_cost.py` to get .csv output with traffic estimations for different Allgather algorithms.\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fspcl%2Fmuliticast-based-allgather","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fspcl%2Fmuliticast-based-allgather","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fspcl%2Fmuliticast-based-allgather/lists"}