NVIDIA / nccl

Optimized primitives for collective multi-GPU communication

NVIDIA/nccl Issues

A question about sequences of functions called in nccl/src/transport /net.cc
Updated 7 days ago
P2P in non-blocking mode
Updated 8 days ago
nvlsAllocateMem report error
Updated 9 days ago4
Alternating rings cause bad performance (NIC sending PFC) in a cluster with mixed crossNic=0/1 nodes
Updated 9 days ago
NCCL Collective Log Query
Updated 12 days ago2
NCCL Logs Questions
Updated 12 days ago
1000x latency with `ncclSend` and `ncclRecv`
Closed 15 days ago10
Question about tree channel
Updated 13 days ago
Question about the topology of double binary tree
Updated 13 days ago
NCCL WARN socketProgress: Connection closed by remote peer
Updated 15 days ago2
The internal errors caused by host-side multi-threads
Updated 16 days ago
How to use the profiler plugin?
Updated 17 days ago1
what is the nvidia IMEX channels in Multi-Node NVLink (MNNVL)
Updated 17 days ago1
The reason for breakdown "NCCL WARN comm 0x55ec10690820 has already been destroyed"
Closed 18 days ago2
what will happen if i call ncclAbort after ncclSend
Updated 18 days ago
How to interpret NCCL's NVTX data
Closed 19 days ago1
The nsys profile will hang when NCCL_P2P_USE_CUDA_MEMCPY is enabled
Updated 19 days ago5
cuda memcpy instead of gpu kernel in p2p sendrecv operation
Updated 19 days ago11
Some question about NCCL_IB_SCA
Closed 19 days ago2
question about PXN
Updated 19 days ago
ncclCommSplit in non-blocking API mode
Updated 22 days ago6
Does FabricManager support isolation at GPU card granularity in MNNVL envrionmant?
Updated 23 days ago1
Question about hostStream, deviceStream and userStream
Updated 23 days ago
nvidia-peermem nv_get_p2p_free_callback:127 ERROR detected invalid context, skipping further processing
Closed 23 days ago2
Why tree algorithms are specifically targeted at All-Reduce?
Updated 25 days ago1
mpirun nccl capture fails in nccl2.22.3 but suceess in nccl2.19.3
Updated 25 days ago
Topology printing bug in `ncclTopoPrintGraph` in NCCL 2.23?
Updated a month ago4
Unstable busbw for data size 67108864B in nccl test
Updated a month ago
Enroot
Updated a month ago1
[BUG]: NCCL_SHM_DISABLE flag is not working
Updated a month ago1
A Question about network buffer
Closed 2 months ago3
NCCL GPU affinity (nvidia-smi topo -m) on VM with fail PIX,PHB, PXB GDRDMA. but good performance on BM.
Updated a month ago4
Build failure on nccl 2.23.4. Missing shmutils.h
Closed 2 months ago8
nccl-test An error occurs when multiple IB_Hcas are used
Updated a month ago
300node 8GPU 4 IB NCCL TEST
Updated a month ago4
Why are not all SMs active when NCCL kernel and compute kernel overlap?
Updated a month ago
NCCL Topology File, Rank Ordering, and Determinism
Updated a month ago
[SHARP] Error When Running nccl-tests with multi-GPUs per node using SHARP
Updated a month ago
Can we change the number of proxy threads
Updated a month ago
How to estimate the communication time of NCCL alltoallv?
Updated a month ago
Some questions about fifo buffer design
Updated a month ago
[Question] Why ncclSend is non-blocking?
Updated a month ago1
Poor NCCL allreduce performance
Updated a month ago4
[SHArP] about the intranode allreduce performance with SHArP
Updated a month ago
Unable to Specify CUDA Stream for Collective Operations Using with torch.cuda.stream() context
Closed 2 months ago
NCCL socket performance over multiple NICs
Updated 2 months ago2
[ext-net] is bundling headers still recommended?
Updated 2 months ago1
How can I see the algorithm chosen by NCCL?
Updated 2 months ago2
【the difference between NCCL and cudaMemcpyPeerAsync】
Updated 2 months ago
NVLink SHARP Performance on AWS P5
Updated 2 months ago