-
Notifications
You must be signed in to change notification settings - Fork 1.2k
Pull requests: NVIDIA/nccl
Author
Label
Projects
Milestones
Reviews
Assignee
Sort
Pull requests list
[Graph] Improve AMD Turin no-XML ring graph selection on AMD Turin
#2127
opened Apr 22, 2026 by
voipmonitor
•
Draft
Fix uninitialized comm->netDeviceType leading to CUDA illegal memory access
#2121
opened Apr 20, 2026 by
david-abr
Loading…
Fix cuMulticastUnbind silent swallow in ncclNvlsDeregBuffer
#2116
opened Apr 20, 2026 by
shubh3794
Loading…
Correct crossing multi-sockets bandwith of AMD Zen5/Turin CPUs
#2113
opened Apr 17, 2026 by
hz-cheng
Loading…
GIN: implement reset-without-zeroing for signals and counters
#2093
opened Apr 6, 2026 by
bhasunit
Loading…
[nccl-ep] Add ncclEpUpdateHandle to rebind topk_idx without reallocating
#2085
opened Apr 2, 2026 by
kwen2501
Collaborator
Loading…
[nccl-ep] Remove redundant
cudaStreamSynchronize in ncclEpCreateHandle
#2082
opened Apr 1, 2026 by
kwen2501
Collaborator
Loading…
Enable P2P transport for AMD systems with >2 GPUs at PHB level
#2080
opened Mar 31, 2026 by
voipmonitor
Loading…
5 of 6 tasks
Enable strict C++ compiler warnings and -Werror by default
#2061
opened Mar 22, 2026 by
maxwbuckley
Loading…
4 tasks done
graph/tuning: disable LL128 on PATH_P2C inter-node paths (GH200 corruption fix)
#2053
opened Mar 13, 2026 by
ryanhankins
Loading…
Fix AMD inter-CPU bandwidth and LL128 protocol for multi-socket EPYC
#2036
opened Mar 8, 2026 by
voipmonitor
Loading…
7 tasks
[BUG] fix opCount not self-increasing in single node communicator
#2021
opened Feb 10, 2026 by
alpha-baby
Loading…
Previous Next
ProTip!
Filter pull requests by the default branch with base:master.