[PR #86] [MERGED] [TransferEngine] feature: registerLocalMemory support the "*" location. #1316

New issue

Closed

opened 2026-10-29 21:26:19 +00:00 by bsakaguchi · 0 comments

bsakaguchi commented

2026-10-29 21:26:19 +00:00

Owner

📋 Pull Request Information

Original PR: https://github.com/kvcache-ai/Mooncake/pull/86
Author: @doujiang24
Created: 1/21/2025
Status: ✅ Merged
Merged: 1/23/2025
Merged by: @stmatengss

Base: main ← Head: auto-numa

📝 Commits (2)

02ba3ed [TransferEngine] feature: registerLocalMemory support the "*" location.
ab9f457 fix test when there is only one numa node

📊 Changes

11 files changed (+342 additions, -67 deletions)

View changed files

📝 doc/en/transfer-engine.md (+1 -1)
📝 doc/zh/transfer-engine.md (+1 -1)
📝 mooncake-transfer-engine/include/common.h (+1 -2)
➕ mooncake-transfer-engine/include/memory_location.h (+40 -0)
➕ mooncake-transfer-engine/src/memory_location.cpp (+67 -0)
📝 mooncake-transfer-engine/src/topology.cpp (+8 -1)
📝 mooncake-transfer-engine/src/transport/rdma_transport/rdma_transport.cpp (+22 -5)
📝 mooncake-transfer-engine/tests/CMakeLists.txt (+4 -0)
➕ mooncake-transfer-engine/tests/memory_location_test.cpp (+123 -0)
📝 mooncake-transfer-engine/tests/rdma_transport_test.cpp (+1 -1)
📝 mooncake-transfer-engine/tests/topology_test.cpp (+74 -56)

📄 Description

try best to recognize the cpu numa node for now,
use all nic when failed to get the numa node,
will support cuda memory in the feature.

_{🔄 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/kvcache-ai/Mooncake/pull/86 **Author:** [@doujiang24](https://github.com/doujiang24) **Created:** 1/21/2025 **Status:** ✅ Merged **Merged:** 1/23/2025 **Merged by:** [@stmatengss](https://github.com/stmatengss) **Base:** `main` ← **Head:** `auto-numa` --- ### 📝 Commits (2) - [`02ba3ed`](https://github.com/kvcache-ai/Mooncake/commit/02ba3eda46b3f6e024205e083aad3aceb2bd1ce7) [TransferEngine] feature: registerLocalMemory support the "*" location. - [`ab9f457`](https://github.com/kvcache-ai/Mooncake/commit/ab9f457c693b97c571951b8bb9c1c78d0cf75a2e) fix test when there is only one numa node ### 📊 Changes **11 files changed** (+342 additions, -67 deletions) <details> <summary>View changed files</summary> 📝 `doc/en/transfer-engine.md` (+1 -1) 📝 `doc/zh/transfer-engine.md` (+1 -1) 📝 `mooncake-transfer-engine/include/common.h` (+1 -2) ➕ `mooncake-transfer-engine/include/memory_location.h` (+40 -0) ➕ `mooncake-transfer-engine/src/memory_location.cpp` (+67 -0) 📝 `mooncake-transfer-engine/src/topology.cpp` (+8 -1) 📝 `mooncake-transfer-engine/src/transport/rdma_transport/rdma_transport.cpp` (+22 -5) 📝 `mooncake-transfer-engine/tests/CMakeLists.txt` (+4 -0) ➕ `mooncake-transfer-engine/tests/memory_location_test.cpp` (+123 -0) 📝 `mooncake-transfer-engine/tests/rdma_transport_test.cpp` (+1 -1) 📝 `mooncake-transfer-engine/tests/topology_test.cpp` (+74 -56) </details> ### 📄 Description 1. try best to recognize the cpu numa node for now, 2. use all nic when failed to get the numa node, 3. will support cuda memory in the feature. --- <sub>🔄 This issue represents a GitHub Pull Request. It cannot be merged through Gitea due to API limitations.</sub>

bsakaguchi

2026-10-29 21:26:19 +00:00