openucx / ucc

Unified Collective Communication Library

https://openucx.github.io/ucc/

Repository from Github https://github.comopenucx/ucc

openucx/ucc Issues

The question of testing the onesided alltoall algorithm
Closed 2 months ago5
How to link and use CL and TL library?
Closed 2 months ago1
Add option to create teams with only specific TLs
Updated 3 months ago
Disable unsupported TLs when CUDA Graph capture is in progress
Updated 3 months ago
Performance degradation on TL/CUDA/ALLGATHER
Updated 4 months ago2
raise min UCX version in anticipation of release
Updated 4 months ago
Failing PyTorch Collective Operations on Non-Homogeneous Ring (NVIDIA and AMD GPUs)
Closed 4 months ago3
Question about failing collective allreduce across on non-homogenous ring (NVIDIA and AMD GPU)
Closed 5 months ago6
Error and hang on v100
Updated 5 months ago1
Build Errors with CL/DOCA_UROM
Updated 5 months ago3
ucc compilation is getting failed on master branch
Updated 5 months ago5
Question about communication between Nvidia and AMD GPUs
Updated 5 months ago3
Some puzzled problems about algorithm
Updated 5 months ago1
Question about ucc topology
Updated 5 months ago1
MPI-style INPLACE operations
Updated 6 months ago2
Some puzzled problems about algorithms
Closed 6 months ago
Do TL/nccl and rccl only support allgatherv?
Closed 7 months ago3
Mixed usage with TL/UCP and TL/MLX5：create tl_mlx5 ctx failed
Closed 7 months ago2
`--with-nvcc-gencode` help text differs slightly from expected format
Updated 8 months ago
--with-tlcp=(cp1,cp2) Enable build of TL collectives plugins
Updated 8 months ago1
add new alltoallv alg field
Closed 9 months ago
Correct cleanup of internal resources associated with communicators
Updated 10 months ago3
[RFE] EC/CUDA: mode to avoid using persistent kernel
Closed 10 months ago1
segfault when nranks>=18
Closed 10 months ago7
Building with ROCm/HIP fails on a system without GPU
Updated 10 months ago5
Wrong data received when sharing a team with multiple threads
Updated 10 months ago22
Configurable OOB/service collectives
Closed a year ago
API/ABI compatibility of UCC
Closed a year ago2
1.3.0 release tag
Closed a year ago2
Fortran OpenMP+ target offload: TL_CUDA DEBUG cannot create CUDA TL context without active CUDA context
Closed a year ago1
Create a conda-forge package for the UCC library.
Updated a year ago
Include #847 in v1.3.0 RC
Closed a year ago2
Duplicate function declaration in src/ucc/api/ucc.h file
Closed a year ago1
Issues regarding ucc_coll_args::tag
Updated a year ago6
Compilation of GPU kernel code generates warnings
Updated a year ago5
Minor build issue at ucc_ep_map_is_identity()
Closed a year ago1
ucc_mem_map limitations
Updated a year ago4
`ucc_context_create` seems to take a constant time
Updated a year ago1
Provide a NCCL-like initialization mechanism
Updated a year ago5
Segfault when running `ucc_info`
Closed a year ago2
Segmentation fault in scatter using Pytorch's UCC backend
Closed a year ago2
sm_52 compilation error, test for cuda version > 11.6
Updated a year ago1
Version 1.2.0 Release (April 30th)
Closed a year ago3
ucc_lib_get_attr doesn't produce output
Updated a year ago1
make install failed
Closed a year ago6
Reduce-Scatter in TL/SHARP
Closed a year ago3
[Question] Relationship between CLs and TLs
Closed a year ago2
Clarify relation between ucc_thread_mode_t and ucs_thread_mode_t
Closed 2 years ago1
Updating ucc.conf usages in Wiki->FAQ
Closed 2 years ago4
Use collective active_set to send data to self
Updated 2 years ago1