Search results

What RDMA verbs does NCCL GIN use?

Status: Open.

#2233 In NVIDIA/nccl;

· baiwei0427 opened on Jun 11, 2026

[Issue]: ncclCommInitAll() leaves a CUDA driver context on the stack

Status: Open.

#2231 In NVIDIA/nccl;

· Jacobfaib opened on Jun 10, 2026

[Question]: [GIN] GPI vs GDAKI: who selects the QP, and what NIC support does GPI need?

question

Status: Open.

#2226 In NVIDIA/nccl;

· QizhouZhang97 opened on Jun 9, 2026

[RFE]: Include qp_num in detailed RoCE completion error log

enhancement

Status: Open.

#2219 In NVIDIA/nccl;

· zrss opened on Jun 7, 2026

[Question]: Why NCCL allgather performance drop with D2H background traffic?

question

Status: Open.

#2218 In NVIDIA/nccl;

· ssb233 opened on Jun 6, 2026

[RFE]: Add opt-in auto-selection for IB NIC merge mode on two-node multi-HCA systems

Status: Open.

#2216 In NVIDIA/nccl;

· CyberSecurityErial opened on Jun 5, 2026

[Bug] SIGSEGV in ncclLocalOpAppend after ~1.1M collectives on CPU-path — dual DGX Spark (aarch64, GB10, NCCL 2.28.9+cuda13.0)

Status: Open.

#2213 In NVIDIA/nccl;

· parallelArchitect opened on Jun 4, 2026

[Question]: why window flag NCCL_WIN_STRICT_ORDERING is needed when using a VA signal?

question

Status: Open.

#2210 In NVIDIA/nccl;

· baymaxhuang opened on Jun 4, 2026

[Question]: Potential ncclIntruQueueTransfer bug + memory leak?

question

Status: Open.

#2204 In NVIDIA/nccl;

· miEsMar opened on Jun 3, 2026

[Issue]: NCCL-GIN hang during the initial AllGather

Status: Open.

#2203 In NVIDIA/nccl;

· sheepx86 opened on Jun 3, 2026

[RFE]: MXFP8 / MXFP4 / NVFP4 AllReduce support and implementation considerations

enhancement

Status: Open.

#2199 In NVIDIA/nccl;

· YanyunDuan opened on Jun 2, 2026

[RFE]: Misleading RING_LL suffix in AllGather/ReduceScatter kernel names

enhancement

Status: Open.

#2196 In NVIDIA/nccl;

· baptiste-bt opened on May 26, 2026