deepseek-ai/DeepEP
1
0
Fork 0
mirror of https://github.com/deepseek-ai/DeepEP.git synced 2026-11-05 00:47:32 +00:00
Code Issues 308 Projects Packages Wiki Activity

[PR #288] [MERGED] ibgda: support non-bond dual-port environments via multi-port config #505

New issue
Closed
opened 2026-10-27 21:03:19 +00:00 by bsakaguchi · 0 comments
bsakaguchi commented 2026-10-27 21:03:19 +00:00
Owner
Copy link

📋 Pull Request Information

Original PR: https://github.com/deepseek-ai/DeepEP/pull/288
Author: @liuhe-spec
Created: 7/9/2025
Status: ✅ Merged
Merged: 7/11/2025
Merged by: @sphish

Base: main ← Head: main

📝 Commits (2)

  • 3571a92 add DeepEP_multi_port_nobond ibgda support
  • 1cd5eea Increase MAX_NUM_HCAS from 16 to 32 to support more NICs in NVSHMEM

📊 Changes

1 file changed (+25 additions, -13 deletions)

View changed files

📝 csrc/kernels/ibgda_device.cuh (+25 -13)

📄 Description

Summary

This patch enables support for non-bonded dual-port InfiniBand environments when using the ibgda transport in NVSHMEM.

In certain HPC/AI networking deployments, multiple InfiniBand NIC ports may be connected without link bonding. This patch allows NVSHMEM to fully utilize such configurations by enabling multi-port awareness at runtime and compiling with appropriate NCCL support.

Changes

  • Added support for initializing multiple independent NIC ports in ibgda transport.

  • Introduced environment variable and build option:

    # Enable multi-port mode at runtime
export NVSHMEM_IBGDA_ENABLE_MULTI_PORT=1

# Compile NVSHMEM with NCCL support
NVSHMEM_USE_NCCL=1

🔄 This issue represents a GitHub Pull Request. It cannot be merged through Gitea due to API limitations.

## 📋 Pull Request Information **Original PR:** https://github.com/deepseek-ai/DeepEP/pull/288 **Author:** [@liuhe-spec](https://github.com/liuhe-spec) **Created:** 7/9/2025 **Status:** ✅ Merged **Merged:** 7/11/2025 **Merged by:** [@sphish](https://github.com/sphish) **Base:** `main` ← **Head:** `main` --- ### 📝 Commits (2) - [`3571a92`](https://github.com/deepseek-ai/DeepEP/commit/3571a927a342e1cb584f4cb6705a1396c4446f5e) add DeepEP_multi_port_nobond ibgda support - [`1cd5eea`](https://github.com/deepseek-ai/DeepEP/commit/1cd5eea6fdbf452733682e2970124b99ec54d44d) Increase MAX_NUM_HCAS from 16 to 32 to support more NICs in NVSHMEM ### 📊 Changes **1 file changed** (+25 additions, -13 deletions) <details> <summary>View changed files</summary> 📝 `csrc/kernels/ibgda_device.cuh` (+25 -13) </details> ### 📄 Description ### Summary This patch enables support for **non-bonded dual-port InfiniBand environments** when using the `ibgda` transport in NVSHMEM. In certain HPC/AI networking deployments, multiple InfiniBand NIC ports may be connected without link bonding. This patch allows NVSHMEM to fully utilize such configurations by enabling multi-port awareness at runtime and compiling with appropriate NCCL support. ### Changes - Added support for initializing multiple independent NIC ports in `ibgda` transport. - Introduced environment variable and build option: ```bash # Enable multi-port mode at runtime export NVSHMEM_IBGDA_ENABLE_MULTI_PORT=1 # Compile NVSHMEM with NCCL support NVSHMEM_USE_NCCL=1 --- <sub>🔄 This issue represents a GitHub Pull Request. It cannot be merged through Gitea due to API limitations.</sub>
bsakaguchi 2026-10-27 21:03:19 +00:00
Sign in to join this conversation.
No Branch/Tag specified
Branches Tags
main
hybrid-ep
chore/remove-clang-format-yapf
v2.5
nvDev
try-fix-cu128
mori-ep
antgroup-opt
compat-old-torch
add-linter
eager-rdma
v1.2.1
Labels
Clear labels   
pull-request
Mirrored from GitHub Pull Request

No labels
pull-request
Milestone
Clear milestone
No items
No milestone
Projects
Clear projects
No items
No project
Assignees
Clear assignees
No assignees
1 participant
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set.

Reference
deepseek-ai/DeepEP#505
No description provided.