Optimize Qwen3 decode scope2 SV path

[high-cloud]

Summary

• Fuse the scope2 softmax and SV matmul loops in Qwen3 decode.
• Remove the intermediate all_exp_padded tensor and its extra write/read path.
• Keep A2/A3-friendly Q_HEAD_PAD alignment by computing padded softmax rows directly.

Validation

• python -m py_compile examples/models/qwen3/qwen3_32b_decode.py
• python modules/pypto-lib/examples/models/qwen3/qwen3_32b_decode.py -p a2a3 -d 10 --max-seq --runtime-profiling
  • out: PASS (131072/131072 elements matched)
  • Total Test Time: 2992.18 us

[coderabbitai]

Important

Review skipped

Draft detected.

Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: Organization UI

Review profile: CHILL

Plan: Pro

Run ID: 26df5072-affa-48a7-9324-a6de1b5736a0

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

[gemini-code-assist]

Code Review

This pull request refactors the Qwen3 decode implementation by fusing the softmax and SV matmul stages. It removes the intermediate all_exp_padded tensor and updates tensor shapes and slicing logic to use Q_HEAD_PAD for improved alignment. The online softmax accumulation is also adjusted to handle the updated padding strides. I have no feedback to provide as the review comments were purely explanatory or validated the existing changes.