[PR #78] [MERGED] Implement NVML pressure and async pre-free clamping #106

New issue

Closed

opened 2026-10-26 22:24:36 +00:00 by bsakaguchi · 0 comments

bsakaguchi commented

2026-10-26 22:24:36 +00:00

Owner

📋 Pull Request Information

Original PR: https://github.com/Comfy-Org/comfy-aimdo/pull/78
Author: @rattus128
Created: 8/2/2026
Status: ✅ Merged
Merged: 8/2/2026
Merged by: @rattus128

Base: master ← Head: dev/async-clamp

📝 Commits (2)

640e9cf prioritize: add optional clamp on async pre-emptive free
d423a70 prioritize NVML pressure

📊 Changes

12 files changed (+209 additions, -19 deletions)

View changed files

📝 .github/workflows/build-wheels.yml (+1 -1)
📝 comfy_aimdo/control.py (+6 -1)
📝 comfy_aimdo/model_vbar.py (+5 -3)
📝 scripts/build-linux-aimdo.sh (+1 -0)
📝 src-cuda/dispatch.c (+129 -0)
📝 src-win/shmem-detect.c (+7 -3)
📝 src/control.c (+38 -9)
📝 src/control.h (+8 -0)
📝 src/gpu_abi.h (+4 -0)
📝 src/gpu_dispatch.h (+7 -0)
📝 src/model-vbar.c (+2 -1)
📝 src/pyt-cu-plug-alloc-async.c (+1 -1)

📄 Description

Contribution Agreement

I agree that my contributions are licensed under the GPLv3.
I grant Comfy Org the rights to relicense these contributions as outlined in CONTRIBUTING.md.

2 changes.

Add an optional clamp to async memory allocation vbars_free.

Implement NVML based VRAM pressure - initial tests indicate this is better when its available.

_{🔄 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/Comfy-Org/comfy-aimdo/pull/78 **Author:** [@rattus128](https://github.com/rattus128) **Created:** 8/2/2026 **Status:** ✅ Merged **Merged:** 8/2/2026 **Merged by:** [@rattus128](https://github.com/rattus128) **Base:** `master` ← **Head:** `dev/async-clamp` --- ### 📝 Commits (2) - [`640e9cf`](https://github.com/Comfy-Org/comfy-aimdo/commit/640e9cfd0aff496d9b9e1e4f2cd2fd860a82d80b) prioritize: add optional clamp on async pre-emptive free - [`d423a70`](https://github.com/Comfy-Org/comfy-aimdo/commit/d423a70625a6dacdb73bd163925d5803dd3c636f) prioritize NVML pressure ### 📊 Changes **12 files changed** (+209 additions, -19 deletions) <details> <summary>View changed files</summary> 📝 `.github/workflows/build-wheels.yml` (+1 -1) 📝 `comfy_aimdo/control.py` (+6 -1) 📝 `comfy_aimdo/model_vbar.py` (+5 -3) 📝 `scripts/build-linux-aimdo.sh` (+1 -0) 📝 `src-cuda/dispatch.c` (+129 -0) 📝 `src-win/shmem-detect.c` (+7 -3) 📝 `src/control.c` (+38 -9) 📝 `src/control.h` (+8 -0) 📝 `src/gpu_abi.h` (+4 -0) 📝 `src/gpu_dispatch.h` (+7 -0) 📝 `src/model-vbar.c` (+2 -1) 📝 `src/pyt-cu-plug-alloc-async.c` (+1 -1) </details> ### 📄 Description ### Contribution Agreement - [X] I agree that my contributions are licensed under the GPLv3. - [X] I grant **Comfy Org** the rights to relicense these contributions as outlined in [CONTRIBUTING.md](./CONTRIBUTING.md). 2 changes. Add an optional clamp to async memory allocation vbars_free. Implement NVML based VRAM pressure - initial tests indicate this is better when its available. --- <sub>🔄 This issue represents a GitHub Pull Request. It cannot be merged through Gitea due to API limitations.</sub>

bsakaguchi

2026-10-26 22:24:36 +00:00