Commit ·
ffff475
1
Parent(s): dd867d5
Added nvshmem.
Browse files
PyTorchConference2025_GithubRepos.json
CHANGED
|
@@ -1177,5 +1177,11 @@
|
|
| 1177 |
"repo_name": "kraken",
|
| 1178 |
"repo_link": "https://github.com/meta-pytorch/kraken",
|
| 1179 |
"github_about_section": "Triton-based Symmetric Memory operators and examples"
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1180 |
}
|
| 1181 |
]
|
|
|
|
| 1177 |
"repo_name": "kraken",
|
| 1178 |
"repo_link": "https://github.com/meta-pytorch/kraken",
|
| 1179 |
"github_about_section": "Triton-based Symmetric Memory operators and examples"
|
| 1180 |
+
},
|
| 1181 |
+
{
|
| 1182 |
+
"repo_name": "nvshmem",
|
| 1183 |
+
"repo_link": "https://github.com/NVIDIA/nvshmem",
|
| 1184 |
+
"github_about_section": "NVIDIA NVSHMEM is a parallel programming interface for NVIDIA GPUs based on OpenSHMEM. NVSHMEM can significantly reduce multi-process communication and coordination overheads by allowing programmers to perform one-sided communication from within CUDA kernels and on CUDA streams.",
|
| 1185 |
+
"homepage_link": "https://docs.nvidia.com/nvshmem/api/index.html"
|
| 1186 |
}
|
| 1187 |
]
|