Optimized primitives for collective multi-GPU communication
active 2023-08-15 → 2026-09-04 (UTC)
Activity over time
Daily event counts in the loaded window
Line chart, 1117 days from 2023-08-15 to 2026-09-04. Pushes: 163 total, peak 4 in a day. Pull requests: 142 total, peak 3 in a day. Issues: 1,516 total, peak 69 in a day. Comments: 3,270 total, peak 73 in a day. Stars: 1,795 total, peak 11 in a day.
- Pushes
- Pull requests
- Issues
- Comments
- Stars
Stars, PRs, issues and forks are under-captured in the later part of this window. GH Archive progressively stopped capturing non-push events during 2026 — −95% or worse by the end of the window. Every series here except Pushes fades for that reason, so a decline above reflects the archive, not this repository. Pushes stay reliable throughout, so read them, and the contributor counts derived from them, as the real signal. Data health has the measurements.
Top contributors
Pushes, PRs, issues, reviews and comments — stars and forks excluded, so this is contribution rather than popularity
| Contributor | Contributions | Pushes | PRs | Comments |
|---|---|---|---|---|
| sjeaugey | 496 | 22 | 3 | 464 |
| stephenmsachs | 382 | 2 | 3 | 195 |
| kiskra-nvidia | 290 | 4 | 4 | 258 |
| github-actions[bot] | 283 | 0 | 0 | 144 |
| AddyLaddy | 125 | 2 | 2 | 120 |
| jskrobola | 109 | 101 | 2 | 6 |
| KaimingOuyang | 78 | 3 | 0 | 75 |
| xiaofanl-nvidia | 72 | 0 | 1 | 56 |
| shanleo1986 | 65 | 0 | 0 | 47 |
| GeofferyGeng | 49 | 0 | 4 | 30 |
| yanminjia | 48 | 0 | 0 | 35 |
| gcongiu | 45 | 0 | 0 | 45 |
| PhdShi | 37 | 0 | 0 | 25 |
| acphile | 33 | 0 | 0 | 32 |
| kwen2501 | 32 | 0 | 1 | 20 |
| shanleo2024 | 31 | 0 | 0 | 18 |
| gangxie112 | 30 | 0 | 0 | 22 |
| ZhiyiHu1999 | 26 | 0 | 0 | 5 |
| chenhongyu2048 | 26 | 0 | 0 | 15 |
| wangfakang | 25 | 0 | 4 | 16 |
Recent activity
Latest issues, pull requests and releases
- Issue comment#2359rishdas2026-08-21 15:05Add Inspector ProxyOp and ProxyStep tracking
- Issue comment#2332xiaofanl-nvidia2026-08-09 23:47[Issue]: `ncclCommRegister` fails with `CUDA_ERROR_NOT_FOUND` (500) on memory allocated with CUDA VMM.
- Issue#2241xiaofanl-nvidia2026-06-23 22:03[Issue]: Potential AllGather deadlock on B200 with ncclCommWindowRegister
- Issue comment#2193WizardKing-Asta2026-06-09 16:33[Question]: Disable NCCL CROSS NIC
- Issue#2216xiaofanl-nvidia2026-06-07 00:58[RFE]: Add opt-in auto-selection for IB NIC merge mode on two-node multi-HCA systems
- Issue comment#2216xiaofanl-nvidia2026-06-07 00:58[RFE]: Add opt-in auto-selection for IB NIC merge mode on two-node multi-HCA systems
- Issue#2204xiaofanl-nvidia2026-06-07 00:47[Question]: Potential `ncclIntruQueueTransfer` bug + memory leak?
- Issue comment#2192xiaofanl-nvidia2026-06-07 00:39[Question]: NIC resiliancy/failover and VNICs
- Issue#2190xiaofanl-nvidia2026-06-07 00:38[Issue]: NCCL WARN Error while attaching to shared memory segment
- Issue#2189xiaofanl-nvidia2026-06-07 00:36[Issue]: Potential memory leak in ncclIbInitDevices error paths after wrap_ibv_get_device_list()
- Issue comment#2183xiaofanl-nvidia2026-06-07 00:21Fix RoCE GID cache after port flap during P2P recovery
- Issue comment#2214xiaofanl-nvidia2026-06-07 00:18docs: clarify profiler kernel names
- Issue comment#2208xiaofanl-nvidia2026-06-07 00:16dev_runtime: preserve GIN resource requirements
- Issue comment#2206xiaofanl-nvidia2026-06-07 00:13Clean up GIN connect failures
- Issue#2216CyberSecurityErial2026-06-05 03:58[RFE]: Add opt-in auto-selection for IB NIC merge mode on two-node multi-HCA systems
- Pull request#2209bhramesh-nvidia2026-06-04 22:28
- Issue comment#2062AddyLaddy2026-05-29 20:16[Issue]: Potential BUG: Memory Visibility and Race Condition in bootstrapNetInit (Identified by RustCC)
- Issue comment#2179wanglei8752026-05-28 01:53Fix GIN init context leak
- Issue#2193kamal1123-git2026-05-24 22:10[Question]: Disable NCCL CROSS NIC
- Issue comment#2090functionstackx2026-05-24 00:20[Roadmap]: NCCL Roadmap Q2 2026
- Issue#2192andreistan262026-05-23 09:53[Question]: NIC resiliancy/failover and VNICs
- Issue comment#2174jskrobola2026-05-13 21:34bootstrap, socket: handle bad magic without blocking caller
- Issue comment#2021kiskra-nvidia2026-05-13 21:04[BUG] fix opCount not self-increasing in single node communicator
- Issue comment#2168xiaofanl-nvidia2026-05-13 05:01bootstrap: Fix data race in network init flag
- Issue comment#1792gangxie1122026-05-13 02:44[Question]: pool alltoall performance with 16 H20 nodes
Totals cover only the window loaded into ClickHouse and count events, not GitHub's lifetime totals — 1,795 stars here means stars gained during the window, not the repo's star count.