{"id":19439456,"url":"https://github.com/openucx/ucc","last_synced_at":"2025-05-16T06:07:43.909Z","repository":{"id":37075669,"uuid":"282490868","full_name":"openucx/ucc","owner":"openucx","description":"Unified Collective Communication Library","archived":false,"fork":false,"pushed_at":"2025-05-06T08:46:19.000Z","size":18264,"stargazers_count":251,"open_issues_count":79,"forks_count":109,"subscribers_count":24,"default_branch":"master","last_synced_at":"2025-05-06T09:47:47.639Z","etag":null,"topics":["collectives","cuda","deep-learning","hpc","infiniband","mpi","openshmem","pgas","pytorch","roce","sharp"],"latest_commit_sha":null,"homepage":"https://openucx.github.io/ucc/","language":"C","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":"bsd-3-clause","status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/openucx.png","metadata":{"files":{"readme":"README.md","changelog":"NEWS","contributing":"CONTRIBUTING.md","funding":null,"license":"LICENSE","code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":null,"support":null,"governance":null,"roadmap":null,"authors":"AUTHORS","dei":null,"publiccode":null,"codemeta":null,"zenodo":null}},"created_at":"2020-07-25T17:18:51.000Z","updated_at":"2025-05-06T08:46:24.000Z","dependencies_parsed_at":"2023-09-26T14:08:28.071Z","dependency_job_id":"caacc095-1c91-4cd5-84e1-1bc2200a6065","html_url":"https://github.com/openucx/ucc","commit_stats":null,"previous_names":[],"tags_count":14,"template":false,"template_full_name":null,"repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/openucx%2Fucc","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/openucx%2Fucc/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/openucx%2Fucc/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/openucx%2Fucc/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/openucx","download_url":"https://codeload.github.com/openucx/ucc/tar.gz/refs/heads/master","host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":254478193,"owners_count":22077676,"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":["collectives","cuda","deep-learning","hpc","infiniband","mpi","openshmem","pgas","pytorch","roce","sharp"],"created_at":"2024-11-10T15:23:05.090Z","updated_at":"2025-05-16T06:07:38.892Z","avatar_url":"https://github.com/openucx.png","language":"C","readme":"# Unified Collective Communication (UCC)\n\n\u003cimg src=\"docs/images/ucc_logo.png\" width=\"75%\" height=\"75%\"\u003e\n\nUCC is a collective communication operations API and library that is flexible, complete, and feature-rich for current and emerging programming models and runtimes.\n\n- [Design Goals](#design-goals)\n- [API](https://openucx.github.io/ucc/)\n- [Building](#compiling-and-installing)\n- [Community](#community)\n- [Contributing](#contributing)\n- [License](#license)\n- [Publication](#publication)\n\n## Design Goals\n* Highly scalable and performant collectives for HPC, AI/ML and I/O workloads\n* Nonblocking collective operations that cover a variety of programming models\n* Flexible resource allocation model\n* Support for relaxed ordering model\n* Flexible synchronous model\n* Repetitive collective operations (init once and invoke multiple times)\n* Hardware collectives are a first-class citizen\n\n### UCC Component Architecture\n![](docs/images/ucc_components.png)\n\n## Contributing\nThanks for your interest in contributing to UCC, please see our technical and\nlegal guidelines in the [contributing](CONTRIBUTING.md) file.\n\nAll contributors have to comply with [\"Membership Voluntary\nConsensus Standard\"](https://ucfconsortium.org/policy/)  and [\"Export Compliant\nContribution Submissions\"](https://ucfconsortium.org/policy/) policies.\n\n## License\nUCC is BSD-style licensed, as found in the [LICENSE](LICENSE) file.\n\n## Required packages\n\n* [UCX](https://github.com/openucx/ucx)\n   * UCC uses utilities provided by UCX's UCS component\n\n* [CUDA](https://developer.nvidia.com/cuda-toolkit) (optional)\n   * UCC supports CUDA collectives. To compile with CUDA support, install\n     [NVIDIA CUDA](https://developer.nvidia.com/cuda-downloads) 11.0 or above.\n\n* [HIP](https://rocmdocs.amd.com/en/latest/Programming_Guides/HIP-GUIDE.html) (optional)\n   * UCC supports AMD GPUs using HIP. Instructions for installing ROCM/HIP can be found at\n     [AMD ROCM](https://rocmdocs.amd.com/en/latest/Installation_Guide/Installation_new.html).\n\n* Doxygen\n   * UCC uses Doxygen for generating API documentation\n\n## Compiling and Installing\n\n### Developer's Build\n```sh\n$ ./autogen.sh\n$ ./configure --prefix=\u003cucc-install-path\u003e --with-ucx=\u003cucx-install-path\u003e\n$ make\n```\n\n### Build Documentation\n```sh\n$ ./autogen.sh\n$ ./configure --prefix=\u003cucc-install-path\u003e --with-docs-only\n$ make docs\n```\n\n### Open MPI and UCC collectives\n\n#### Compile UCX \n```sh\n$ git clone https://github.com/openucx/ucx\n$ cd ucx\n$ ./autogen.sh; ./configure --prefix=\u003cucx-install-path\u003e; make -j install\n```\n#### Compile UCC\n\n```sh\n$ git clone https://github.com/openucx/ucc\n$ cd ucc\n$ ./autogen.sh; ./configure --prefix=\u003cucc-install-path\u003e --with-ucx=\u003cucx-install-path\u003e; make -j install\n```\n\n#### Compile Open MPI \n\n```sh\n$ git clone https://github.com/open-mpi/ompi\n$ cd ompi\n$ ./autogen.pl; ./configure --prefix=\u003compi-install-path\u003e --with-ucx=\u003cucx-install-path\u003e --with-ucc=\u003cucc-install-path\u003e; make -j install\n```\n\n#### Run MPI programs\n\n```sh\n$ mpirun -np 2 --mca coll_ucc_enable 1 --mca coll_ucc_priority 100 ./my_mpi_app\n```\n\n#### Run OpenSHMEM programs\n\n```sh\n$ mpirun -np 2 --mca scoll_ucc_enable 1 --mca scoll_ucc_priority 100 ./my_openshmem_app\n```\n\n\n### SUPPORTED Transports\n* UCX/UCP\n  - InfiniBand, ROCE, Cray Gemini and Aries, Shared Memory\n* SHARP\n* CUDA\n* NCCL\n* RCCL\n* MLX5\n\n\n### Publication\n\nTo cite UCC in a publication, please use the following BibTex entry:\n\n```\n@inproceedings{DBLP:conf/hoti/VenkataPLBALBDS24,\n  author       = {Manjunath Gorentla Venkata and\n                  Valentine Petrov and\n                  Sergey Lebedev and\n                  Devendar Bureddy and\n                  Ferrol Aderholdt and\n                  Joshua Ladd and\n                  Gil Bloch and\n                  Mike Dubman and\n                  Gilad Shainer},\n  title        = {Unified Collective Communication {(UCC):} An Unified Library for CPU,\n                  GPU, and {DPU} Collectives},\n  booktitle    = {{IEEE} Symposium on High-Performance Interconnects, {HOTI} 2024, Albuquerque,\n                  NM, USA, August 21-23, 2024},\n  pages        = {37--46},\n  publisher    = {{IEEE}},\n  year         = {2024},\n  url          = {https://doi.org/10.1109/HOTI63208.2024.00018},\n  doi          = {10.1109/HOTI63208.2024.00018},\n  timestamp    = {Thu, 19 Sep 2024 11:00:54 +0200},\n  biburl       = {https://dblp.org/rec/conf/hoti/VenkataPLBALBDS24.bib},\n  bibsource    = {dblp computer science bibliography, https://dblp.org}\n}\n```\n","funding_links":[],"categories":[],"sub_categories":[],"project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fopenucx%2Fucc","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fopenucx%2Fucc","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fopenucx%2Fucc/lists"}