NVIDIA nccl issues - Githubissues

NVIDIA / nccl

Optimized primitives for collective multi-GPU communication

Other

3.28k stars 829 forks source link

issues

Newest

Newest Most commented Recently updated Oldest Least commented Least recently updated

NvlsRegSupport limitation

#1377 zhang662817 opened 4 months ago
0
Modify 'netComms' to 'sharedNetComms' to void misunderstanding.

#1376 GeofferyGeng opened 4 months ago
0
Why needs double check in ncclIbIsend and ncclIbIrecv?

#1375 jiangxiaobin96 opened 4 months ago
0
Why NVLS allreduce performance is so weird？

#1374 yupatrick22 opened 4 months ago
6
NCCL panic with nccl-test with 2 GPUs inside kubevirt VM

#1373 winsopc opened 4 months ago
10
NCCL WARN Cuda failure 'named symbol not found'

#1372 casparvl opened 4 months ago
2
standardize code style

#1371 alpha-baby opened 4 months ago
0
Scheduling to Packet Sending Connection from ncclEnqueueCheck to finally sending the packet in misc/socket.cc::socketProgressOpt

#1370 y344shi opened 4 months ago
0
Use nsight system to profile nccl p2p, I find something confused ...

#1369 oliverYoung2001 opened 4 months ago
0
NCCL test, Tree is slower than Ring

#1368 wangdaw2023 opened 4 months ago
2
why intra-node ring only search one-direction bw?

#1367 echobinarybytes opened 4 months ago
0
NCCL Error on Multi-Node Training with Mixed GPU Setup

#1366 asdfry closed 4 months ago
2
inter-node nvls process when ib sharp not supported

#1365 echobinarybytes closed 4 months ago
1
Ring broadcast

#1364 tks2004 opened 4 months ago
0
Encountering Random Segmentation Fault During NCCL-Tests

#1363 tonyw1213 opened 4 months ago
15
Why does NVLSTree Allreduce perform worse than Ring Allreduce?

#1362 MoringKing opened 4 months ago
1
Missing header file

#1361 alpha-baby opened 4 months ago
7
Single or double ring

#1360 CatalinLucian opened 4 months ago
1
How can I test IB bandwidth when NCCL is running?

#1359 MonroeD opened 4 months ago
0
how does NCCL support peer-to-peer connections across NUMA nodes without the features of NICs and NVLinks?

#1358 themoonstone opened 4 months ago
3
NCCL Tree allreduce test cannot reach the theoretical bus bandwidth on 2 nodes with 4 nics

#1357 ProHuper opened 4 months ago
7
About sync in nvls algorithm

#1356 cyqmonkey opened 4 months ago
0
Is there someway to measure gpu i/o usage or allreduce waiting time?

#1355 MaCasK9 opened 4 months ago
1
Fix double free in ncclTopoRemovePathType

#1354 junxu closed 4 months ago
1
work request complete err: status 5 and vendor err 249

#1353 913871734 closed 3 months ago
7
NCCL Logs Communicator Query

#1352 gjit-juniper opened 4 months ago
1
Is it possible to swap the calling order of `initTransportsRank` and `ncclTunerPluginLoad`

#1351 jeseszhang1010 closed 4 months ago
1
How to tell nccl that those network communication is disabled?

#1350 gpzlx1 closed 4 months ago
2
Enabling read for P2p transport

#1349 NEWPLAN closed 4 months ago
1
Network IP setup and physical wiring

#1348 cold2stone opened 4 months ago
0
deadlock when using multiple communicators for Point-To-Point Communication within the same GPU Group

#1347 CtfGo opened 4 months ago
0
what does non-blocking communicator for？

#1346 CtfGo opened 4 months ago
4
question about a new single-node communication mode

#1345 PhdShi opened 4 months ago
0
Debug results for sendSetup() and recvSetup()

#1344 ZhiyiHu1999 opened 4 months ago
0
How to enable GDR for NIC?

#1343 gpzlx1 closed 4 months ago
2
Will the execution of multimem instructions be kept in same-location program order?

#1342 shixuan94 opened 5 months ago
0
Questions about PATH_PIX in src/graph/paths.cc

#1341 limuhu1994 opened 5 months ago
0
Network Topology awareness

#1340 liranschour opened 5 months ago
2
Where is internal implementation of isend/irecv defined?

#1339 ZhiyiHu1999 closed 1 month ago
1
Invalid Argument when running nccl-test on a single machine with multiple GPUs (H800)

#1338 Jason3900 closed 5 months ago
2
NCCL kernels participating in the same collective synchronize their termination?

#1337 taekyounghan opened 5 months ago
0
How to switch between run_tree_up_down and run_tree_split algorithm

#1336 ZhiyiHu1999 closed 5 months ago
0
Include the CUDA call that failed in errors.

#1335 tfogal opened 5 months ago
2
Questions about the latency numbers in src/graph/tuning.cc

#1334 ege-erdil opened 5 months ago
0
GID index change cause training to stop on ConnectX-7 400G Adapters when traing LLM

#1333 wangdaw2023 closed 5 months ago
2
The variable NCCL_IB_ADDR_RANGE did not work properly after being configured

#1332 riverzhang opened 5 months ago
3
Set the default for the AMD-x86 using the PATH_PXB p2p communication.

#1331 AlextangQ19 opened 5 months ago
2
get rid of some unbounded sprintfs

#1330 madeleineth closed 3 months ago
0
Why can't two GPUs in a virtual machine communicate using P2P？

#1329 qianxiaoliang opened 5 months ago
1
NCCL error "receiving 524288 bytes instead of 65536"

#1328 JingyuQian closed 4 months ago
2

Previous Next