Skip to content

V6 support v0.3.7.post2#1

Open
chaosv598 wants to merge 22 commits into
JiusiServe:v6_support_v0.3.7.post2from
chaosv598:v6_support_v0.3.7.post2
Open

V6 support v0.3.7.post2#1
chaosv598 wants to merge 22 commits into
JiusiServe:v6_support_v0.3.7.post2from
chaosv598:v6_support_v0.3.7.post2

Conversation

@chaosv598

Copy link
Copy Markdown

fix bug for ipv6

ykwd and others added 22 commits November 4, 2025 14:13
Co-authored-by: youxiao <youxiao@huawei.com>
* Add TE Separation Switch

* Fix store compile and link error, add TE installation headers

* Install libmooncake_store.so and libmooncake_common.so when BUILD_SHARED_LIBS=ON
- Implement startHandshakeDaemon method in TcpTransport class
- Call handshake daemon during transport installation process

Signed-off-by: staryxchen <staryxchen@tencent.com>
* add batch [put/get] tensor

Signed-off-by: Xuchun Shang <xuchun.shang@gmail.com>

* fix

Signed-off-by: Xuchun Shang <xuchun.shang@gmail.com>

* add ci

Signed-off-by: Xuchun Shang <xuchun.shang@gmail.com>

* fix

Signed-off-by: Xuchun Shang <xuchun.shang@gmail.com>

* fix

Signed-off-by: Xuchun Shang <xuchun.shang@gmail.com>

---------

Signed-off-by: Xuchun Shang <xuchun.shang@gmail.com>
* handling cudaMemcpy errors in tcp_transport.cpp

- when cuda memory is almost full, cudaMemcpy can fail
- without the error checking, the transfer engine swallow the error,
  which resulted in upper level applications to behave incorrectly.

* Update mooncake-transfer-engine/src/transport/tcp_transport/tcp_transport.cpp

Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com>

* Update mooncake-transfer-engine/src/transport/tcp_transport/tcp_transport.cpp

Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com>

* format

---------

Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com>
…rEngine init (kvcache-ai#1051)

* store: apply MC_MS_FILTERS before TransferEngine init to honor whitelist during auto-discovery

Problem:
MC_MS_FILTERS was set after TransferEngine::init(), so Topology::discover() ignored the whitelist and registered unintended devices (e.g., mlx5_bond_0).

Solution:
Set the whitelist filters on TransferEngine prior to init() so auto-discovery honors MC_MS_FILTERS.

* store: warn when MC_MS_FILTERS is set but auto-discovery is disabled; simplify filter helper

- Apply whitelist before TransferEngine::init so auto-discovery honors filters
- Emit a single warning if MC_MS_FILTERS is set while auto-discovery is disabled
- Remove dead code by making get_auto_discover_filters() param-less
- No behavior change when auto-discovery is off (filters are ignored as before)

* clang-format

---------

Co-authored-by: wangwenqi <wenqi.wangwwq@shopee.com>
* add web api doc

* Apply suggestion from @gemini-code-assist[bot]

Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com>

---------

Co-authored-by: JinYan Su <jinyansu792@gmail.com>
Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com>
This document outlines the Contributor Covenant Code of Conduct, including our pledge, standards of behavior, enforcement responsibilities, and consequences for violations.
…#1065)

* Initial plan

* Add pull request template

Co-authored-by: stmatengss <11641725+stmatengss@users.noreply.github.com>

---------

Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com>
Co-authored-by: stmatengss <11641725+stmatengss@users.noreply.github.com>
…ad of busy-polling (kvcache-ai#1053)

* (Event-driven completion) Provide an option to use condition-variable based completion instead of busy-polling

Summary
- Add USE_EVENT_DRIVEN_COMPLETION compile-time option in common.cmake (default OFF).
    - Store: TransferEngineOperationState waits on BatchDesc::completion_cv with timeout; falls back to original polling when the flag is OFF.
    - TransferEngine: add per-batch finished_task_count, has_failure, is_finished and completion_cv on BatchDesc; notify on last task completion; unify __atomic_* usage.
- No behavior change when USE_EVENT_DRIVEN_COMPLETION is OFF (default).



---------

Co-authored-by: wangwenqi <wenqi.wangwwq@shopee.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.