[PR #90] [MERGED] Indivisible TMA #242

New issue

Closed

opened 2026-10-26 21:15:13 +00:00 by bsakaguchi · 0 comments

bsakaguchi commented

2026-10-26 21:15:13 +00:00

Owner

📋 Pull Request Information

Original PR: https://github.com/deepseek-ai/DeepGEMM/pull/90
Author: @yukuai26
Created: 4/22/2025
Status: ✅ Merged
Merged: 4/23/2025
Merged by: @LyricZhao

Base: main ← Head: indivisible-tma

📝 Commits (10+)

ee4204a tma support indivisible num_n_blocks/num_m_blocks
6947703 fix typo
d2369e8 fix typo2
bfb2bcc add notes
8a92750 fix typo
80f1cfc add notes2
902208a Several code lints
f401495 Several code lints x2
5988421 Simplify
07ef809 Optimize performance

📊 Changes

5 files changed (+73 additions, -53 deletions)

View changed files

📝 README.md (+1 -1)
📝 deep_gemm/include/deep_gemm/fp8_gemm.cuh (+17 -18)
📝 deep_gemm/include/deep_gemm/scheduler.cuh (+45 -24)
📝 deep_gemm/include/deep_gemm/tma_utils.cuh (+3 -4)
📝 deep_gemm/jit_kernels/gemm.py (+7 -6)

📄 Description

Now GEMM and grouped contiguous GEMM support indivisible TMA.

_{🔄 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/DeepGEMM/pull/90 **Author:** [@yukuai26](https://github.com/yukuai26) **Created:** 4/22/2025 **Status:** ✅ Merged **Merged:** 4/23/2025 **Merged by:** [@LyricZhao](https://github.com/LyricZhao) **Base:** `main` ← **Head:** `indivisible-tma` --- ### 📝 Commits (10+) - [`ee4204a`](https://github.com/deepseek-ai/DeepGEMM/commit/ee4204ad986a7b1b869eadd8864a1816989e2675) tma support indivisible num_n_blocks/num_m_blocks - [`6947703`](https://github.com/deepseek-ai/DeepGEMM/commit/69477036c093dae593dc717a3f39ab93f2949dab) fix typo - [`d2369e8`](https://github.com/deepseek-ai/DeepGEMM/commit/d2369e8f30e48d95efb21087cdc9416297d0a5cd) fix typo2 - [`bfb2bcc`](https://github.com/deepseek-ai/DeepGEMM/commit/bfb2bcc04dfc69a784d3090e2f1458b866440e4a) add notes - [`8a92750`](https://github.com/deepseek-ai/DeepGEMM/commit/8a92750027a89c054ad6ac98031482d22bd31756) fix typo - [`80f1cfc`](https://github.com/deepseek-ai/DeepGEMM/commit/80f1cfc6304b26ca2b38245de355cf792678aa94) add notes2 - [`902208a`](https://github.com/deepseek-ai/DeepGEMM/commit/902208a17ede9e231e9e7554d182d025e154b723) Several code lints - [`f401495`](https://github.com/deepseek-ai/DeepGEMM/commit/f4014953ada2d2e1db61a935258ec842a103e6d6) Several code lints x2 - [`5988421`](https://github.com/deepseek-ai/DeepGEMM/commit/59884211ea81d06d18b1915a11cff669bb9e2dfe) Simplify - [`07ef809`](https://github.com/deepseek-ai/DeepGEMM/commit/07ef809d82d6dfe727323267ace516403e580db2) Optimize performance ### 📊 Changes **5 files changed** (+73 additions, -53 deletions) <details> <summary>View changed files</summary> 📝 `README.md` (+1 -1) 📝 `deep_gemm/include/deep_gemm/fp8_gemm.cuh` (+17 -18) 📝 `deep_gemm/include/deep_gemm/scheduler.cuh` (+45 -24) 📝 `deep_gemm/include/deep_gemm/tma_utils.cuh` (+3 -4) 📝 `deep_gemm/jit_kernels/gemm.py` (+7 -6) </details> ### 📄 Description Now GEMM and grouped contiguous GEMM support indivisible TMA. --- <sub>🔄 This issue represents a GitHub Pull Request. It cannot be merged through Gitea due to API limitations.</sub>

bsakaguchi

2026-10-26 21:15:13 +00:00