Skip to content

Pull requests: NVIDIA/nccl

New pull request New
126 Open 269 Closed
126 Open 269 Closed
Author
Filter by author
Loading
Label
Filter by label
Loading
Use alt + click/return to exclude labels
or ⇧ + click/return for logical OR
Projects
Filter by project
Loading
Milestones
Filter by milestone
Loading
Reviews
Filter by reviews
No reviews Review required Approved review Changes requested
Assignee
Filter by who’s assigned
Assigned to nobody Loading
Sort
Sort by
Newest Oldest Most commented Least commented Recently updated Least recently updated Best match
Most reactions
👍 👎 😄 🎉 😕 ❤️ 🚀 👀

Pull requests list

Fix device list leak in ncclIbInitDevices error paths
#2271 opened Jul 5, 2026 by nileshpatil6 Loading…
1
Add GIN GPU/NIC peer affinity mapping
#2269 opened Jul 5, 2026 by hzt123123 Loading…
net: close DMA-BUF fd on the regMrDmaBuf-failure fallback path
#2267 opened Jul 2, 2026 by EylonKrause Loading…
net_ib: add missing break in printIbWcStatusHint switch
#2265 opened Jun 29, 2026 by EylonKrause Loading…
net_ib: use uint32 truncation (not % UINT32_MAX) for BY_ID imm-data request id
#2264 opened Jun 29, 2026 by EylonKrause Loading…
ras: include ncclTimeout in the client error breakdown
#2263 opened Jun 29, 2026 by EylonKrause Loading…
collnet: size shared collNet comm arrays by NCCL_MAX_NETDEVS
#2262 opened Jun 29, 2026 by EylonKrause Loading…
plugin: set tunerPluginLoaded on the already-loaded path so refcount balances
#2261 opened Jun 29, 2026 by EylonKrause Loading…
net: compute recv sameDevice against self, not the PXN proxyRank
#2260 opened Jun 29, 2026 by EylonKrause Loading…
register: restore caller CUDA device on all deregister exit paths
#2259 opened Jun 29, 2026 by EylonKrause Loading…
graph: pass netId (not netDev) to ncclTopoGetIntermediateRank in PXN level 1
#2258 opened Jun 26, 2026 by EylonKrause Loading…
2
Adding FI_MORE to the Gin Proxy Path
#2254 opened Jun 25, 2026 by hershys-aws Loading…
12
Avoid RMA proxy when cuMem host is unavailable
#2253 opened Jun 25, 2026 by jm99k56 Loading…
Fix Python ep_test.py HT mode (under-sized recv buffers → ncclInvalidUsage)
#2252 opened Jun 25, 2026 by ybenvidia Loading…
2
fix: null comm->groupJob after abort to avoid use-after-free
#2251 opened Jun 24, 2026 by EylonKrause Loading…
fix: stack buffer overflow in IPv6 address parsing
#2250 opened Jun 24, 2026 by EylonKrause Loading…
fix: propagate error from ncclCudaContextTrack
#2249 opened Jun 24, 2026 by EylonKrause Loading…
fix: correct NVML P2P WRITE status warning to say INDEX_WRITE
#2248 opened Jun 24, 2026 by EylonKrause Loading…
fix: close XML file on parse-error paths to avoid FILE* leak
#2247 opened Jun 24, 2026 by EylonKrause Loading…
fix gin proxy load balance problem when nic is a lag port
#2230 opened Jun 9, 2026 by YeSho-cpp Loading…
2
bound remote device count in ib connect and accept
#2228 opened Jun 9, 2026 by jmestwa-coder Loading…
2
net_ib: free GIN barrier buffer on failure
#2227 opened Jun 9, 2026 by wanglei875 Contributor Loading…
3
Remove misleading NCCL kernel protocol suffix
#2224 opened Jun 8, 2026 by sanrise Loading…
1
Add NCCL task protocol trace metadata
#2223 opened Jun 8, 2026 by sanrise Loading…
3
Fix KernelCh completion ordering in profiler proxy
#2221 opened Jun 8, 2026 by usernamehaha2022 Loading…
2
ProTip! Follow long discussions with comments:>50.