mirror of
https://github.com/kvcache-ai/Mooncake.git
synced 2026-11-05 01:17:55 +00:00
[PR #1308] [MERGED] [EP] Use NVLink in EP if possible #2149
Labels
No labels
ascend
auto-closed
bug
bug
CI/Build
Common
documentation
enhancement
feature request
good first issue
help wanted
high priority
MetaX
nightly-failure
pull-request
PyTorch Backend
question
RFC
Roadmap
stale
Store
Tests
Transfer Engine
No milestone
No project
No assignees
1 participant
Notifications
Due date
No due date set.
Dependencies
No dependencies set.
Reference
kvcache-ai/Mooncake#2149
Loading…
Add table
Add a link
Reference in a new issue
No description provided.
Delete branch "%!s()"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
📋 Pull Request Information
Original PR: https://github.com/kvcache-ai/Mooncake/pull/1308
Author: @UNIDY2002
Created: 12/30/2025
Status: ✅ Merged
Merged: 12/31/2025
Merged by: @UNIDY2002
Base:
main← Head:sunxun/ep-debug📝 Commits (6)
6724e2aUse NVLink in EP19d3c92Update mooncake-wheel/mooncake/mooncake_ep_buffer.py3ea7fccUpdate mooncake-ep/src/mooncake_ep_buffer.cpp5c54bceFix1bf1751Fix29a9553Fix📊 Changes
6 files changed (+264 additions, -82 deletions)
View changed files
📝
mooncake-ep/include/mooncake_ep_api.cuh(+7 -5)📝
mooncake-ep/include/mooncake_ep_buffer.h(+9 -0)📝
mooncake-ep/src/mooncake_ep_buffer.cpp(+127 -8)📝
mooncake-ep/src/mooncake_ep_kernel.cu(+72 -32)📝
mooncake-integration/ep/ep_py.cpp(+3 -0)📝
mooncake-wheel/mooncake/mooncake_ep_buffer.py(+46 -37)📄 Description
Description
Part of #1225
Use NVLink in EP if possible.
(Don't consider NPU as EP currently only runs on NVIDIA's hardware.)
No changes to the public APIs.
Type of Change
How Has This Been Tested?
test_mooncake_ep.py should observe higher performance (100~200 GB/s in bandwidth)
Checklist
🔄 This issue represents a GitHub Pull Request. It cannot be merged through Gitea due to API limitations.