[PR #293] support low latency dispatch tma #506

New issue

Open

opened 2026-10-27 21:03:22 +00:00 by bsakaguchi · 0 comments

bsakaguchi commented

2026-10-27 21:03:22 +00:00

Owner

📋 Pull Request Information

Original PR: https://github.com/deepseek-ai/DeepEP/pull/293
Author: @ayrnb
Created: 7/10/2025
Status: 🔄 Open

Base: main ← Head: dispatch_tma

📝 Commits (4)

49445f9 ll dispatch tma
faf4507 refine code
868675f refine code
482c887 refine

📊 Changes

1 file changed (+44 additions, -8 deletions)

View changed files

📝 csrc/kernels/internode_ll.cu (+44 -8)

📄 Description

A practice to learn the use of TMA. If there are any errors here, hope to get your guidance and feedback. 🙏🙏🙏🙏

_{🔄 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/293 **Author:** [@ayrnb](https://github.com/ayrnb) **Created:** 7/10/2025 **Status:** 🔄 Open **Base:** `main` ← **Head:** `dispatch_tma` --- ### 📝 Commits (4) - [`49445f9`](https://github.com/deepseek-ai/DeepEP/commit/49445f9dd4a8d856a2fdd4505f7152b94e8a044f) ll dispatch tma - [`faf4507`](https://github.com/deepseek-ai/DeepEP/commit/faf4507a88285f36612cd2069f971b6a95c6e963) refine code - [`868675f`](https://github.com/deepseek-ai/DeepEP/commit/868675f1dca60457dc818e1bec2e77f230ef29ee) refine code - [`482c887`](https://github.com/deepseek-ai/DeepEP/commit/482c88790c32c4b9a694de6ff5008341d6f33fda) refine ### 📊 Changes **1 file changed** (+44 additions, -8 deletions) <details> <summary>View changed files</summary> 📝 `csrc/kernels/internode_ll.cu` (+44 -8) </details> ### 📄 Description A practice to learn the use of TMA. If there are any errors here, hope to get your guidance and feedback. :pray::pray::pray::pray: --- <sub>🔄 This issue represents a GitHub Pull Request. It cannot be merged through Gitea due to API limitations.</sub>

bsakaguchi added the

pull-request

label

2026-10-27 21:03:22 +00:00

bsakaguchi referenced this issue

2026-10-27 21:04:33 +00:00

[PR #532] Fix TMA assertion for non-32-divisible hidden sizes in combine_token #615