Unified Collective Communication (UCC)

UCC is a collective communication operations API and library that is flexible, complete, and feature-rich for current and emerging programming models and runtimes.

• Design Goals
• API
• Building
• Community
• Contributing
• License
• Publication

Design Goals

• Highly scalable and performant collectives for HPC, AI/ML and I/O workloads
• Nonblocking collective operations that cover a variety of programming models
• Flexible resource allocation model
• Support for relaxed ordering model
• Flexible synchronous model
• Repetitive collective operations (init once and invoke multiple times)
• Hardware collectives are a first-class citizen

UCC Component Architecture

Contributing

Thanks for your interest in contributing to UCC, please see our technical and legal guidelines in the contributing file.

All contributors have to comply with "Membership Voluntary Consensus Standard" and "Export Compliant Contribution Submissions" policies.

License

UCC is BSD-style licensed, as found in the LICENSE file.

Required packages

• UCX

  • UCC uses utilities provided by UCX's UCS component

• CUDA (optional)

  • UCC supports CUDA collectives. To compile with CUDA support, install NVIDIA CUDA 11.0 or above.

• HIP (optional)

  • UCC supports AMD GPUs using HIP. Instructions for installing ROCM/HIP can be found at AMD ROCM.

• Doxygen

  • UCC uses Doxygen for generating API documentation

Compiling and Installing

Developer's Build

$ ./autogen.sh
$ ./configure --prefix=<ucc-install-path> --with-ucx=<ucx-install-path>
$ make

Build Documentation

$ ./autogen.sh
$ ./configure --prefix=<ucc-install-path> --with-docs-only
$ make docs

Open MPI and UCC collectives

Compile UCX 

$ git clone https://github.com/openucx/ucx
$ cd ucx
$ ./autogen.sh; ./configure --prefix=<ucx-install-path>; make -j install

Compile UCC

$ git clone https://github.com/openucx/ucc
$ cd ucc
$ ./autogen.sh; ./configure --prefix=<ucc-install-path> --with-ucx=<ucx-install-path>; make -j install

Compile Open MPI 

$ git clone https://github.com/open-mpi/ompi
$ cd ompi
$ ./autogen.pl; ./configure --prefix=<ompi-install-path> --with-ucx=<ucx-install-path> --with-ucc=<ucc-install-path>; make -j install

Run MPI programs

$ mpirun -np 2 --mca coll_ucc_enable 1 --mca coll_ucc_priority 100 ./my_mpi_app

Run OpenSHMEM programs

$ mpirun -np 2 --mca scoll_ucc_enable 1 --mca scoll_ucc_priority 100 ./my_openshmem_app

SUPPORTED Transports

• UCX/UCP
  • InfiniBand, ROCE, Cray Gemini and Aries, Shared Memory
• SHARP
• CUDA
• NCCL
• RCCL
• MLX5

Publication

To cite UCC in a publication, please use the following BibTex entry:

@inproceedings{DBLP:conf/hoti/VenkataPLBALBDS24,
  author       = {Manjunath Gorentla Venkata and
                  Valentine Petrov and
                  Sergey Lebedev and
                  Devendar Bureddy and
                  Ferrol Aderholdt and
                  Joshua Ladd and
                  Gil Bloch and
                  Mike Dubman and
                  Gilad Shainer},
  title        = {Unified Collective Communication {(UCC):} An Unified Library for CPU,
                  GPU, and {DPU} Collectives},
  booktitle    = {{IEEE} Symposium on High-Performance Interconnects, {HOTI} 2024, Albuquerque,
                  NM, USA, August 21-23, 2024},
  pages        = {37--46},
  publisher    = {{IEEE}},
  year         = {2024},
  url          = {https://doi.org/10.1109/HOTI63208.2024.00018},
  doi          = {10.1109/HOTI63208.2024.00018},
  timestamp    = {Thu, 19 Sep 2024 11:00:54 +0200},
  biburl       = {https://dblp.org/rec/conf/hoti/VenkataPLBALBDS24.bib},
  bibsource    = {dblp computer science bibliography, https://dblp.org}
}