[PR #320] [MERGED] Use TMA to optimize combine forwarder. #519

New issue

Closed

opened 2026-10-27 21:03:35 +00:00 by bsakaguchi · 0 comments

bsakaguchi commented

2026-10-27 21:03:35 +00:00

Owner

📋 Pull Request Information

Original PR: https://github.com/deepseek-ai/DeepEP/pull/320
Author: @sphish
Created: 7/21/2025
Status: ✅ Merged
Merged: 7/21/2025
Merged by: @LyricZhao

Base: main ← Head: combine-opt

📝 Commits (5)

1f5d02c Remove an outdated todo
21586ef Increase the number of combine forward warps.
052b20f forwarder use TMA.
7a13298 Small fix
7d092b8 Code lint

📊 Changes

1 file changed (+142 additions, -61 deletions)

View changed files

📝 csrc/kernels/internode.cu (+142 -61)

📄 Description

No description provided

_{🔄 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/320 **Author:** [@sphish](https://github.com/sphish) **Created:** 7/21/2025 **Status:** ✅ Merged **Merged:** 7/21/2025 **Merged by:** [@LyricZhao](https://github.com/LyricZhao) **Base:** `main` ← **Head:** `combine-opt` --- ### 📝 Commits (5) - [`1f5d02c`](https://github.com/deepseek-ai/DeepEP/commit/1f5d02c49e53c6c2ade3e11674a98cdacebb100d) Remove an outdated todo - [`21586ef`](https://github.com/deepseek-ai/DeepEP/commit/21586ef285ef344a0bc437c11a6828d4acebf96a) Increase the number of combine forward warps. - [`052b20f`](https://github.com/deepseek-ai/DeepEP/commit/052b20f0d75e728ba365dad0b3bfa81fc08e4b89) forwarder use TMA. - [`7a13298`](https://github.com/deepseek-ai/DeepEP/commit/7a132986417be5a995d646423eaf6eb276e6a90a) Small fix - [`7d092b8`](https://github.com/deepseek-ai/DeepEP/commit/7d092b836221da391ab5900e3cad339e6e071d7b) Code lint ### 📊 Changes **1 file changed** (+142 additions, -61 deletions) <details> <summary>View changed files</summary> 📝 `csrc/kernels/internode.cu` (+142 -61) </details> ### 📄 Description _No description provided_ --- <sub>🔄 This issue represents a GitHub Pull Request. It cannot be merged through Gitea due to API limitations.</sub>

bsakaguchi

2026-10-27 21:03:35 +00:00