[PR #59] [ROCm] revert chunk size to 16MB #93

New issue

Open

opened 2026-10-26 22:24:29 +00:00 by bsakaguchi · 0 comments

bsakaguchi commented

2026-10-26 22:24:29 +00:00

Owner

📋 Pull Request Information

Original PR: https://github.com/Comfy-Org/comfy-aimdo/pull/59
Author: @Apophis3158
Created: 6/4/2026
Status: 🔄 Open

Base: master ← Head: dev/rocm-revert-chunk

📝 Commits (2)

d803e6f revert ROCm chunk size to 16MB
ec63efe Merge branch 'Comfy-Org:master' into dev/rocm-revert-chunk

📊 Changes

2 files changed (+2 additions, -12 deletions)

View changed files

📝 comfy_aimdo/vram_buffer.py (+1 -7)
📝 src/vrambuf.c (+1 -5)

📄 Description

Was intended to fix ROCm Windows' chunk size in vram_buffer.py, but it seems fine to revert Linux's as well, since we had this workaround to fix VMM bug:
github.com/Comfy-Org/comfy-aimdo@7bbb4d4ac0/src/plat.h (L14-L20)

Wheel for testing: https://github.com/Comfy-Org/comfy-aimdo/actions/runs/26936503053

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.

_{🔄 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/59 **Author:** [@Apophis3158](https://github.com/Apophis3158) **Created:** 6/4/2026 **Status:** 🔄 Open **Base:** `master` ← **Head:** `dev/rocm-revert-chunk` --- ### 📝 Commits (2) - [`d803e6f`](https://github.com/Comfy-Org/comfy-aimdo/commit/d803e6f4fc88c341a7648fc0eb096861d183d8d1) revert ROCm chunk size to 16MB - [`ec63efe`](https://github.com/Comfy-Org/comfy-aimdo/commit/ec63efedfacf4b6f68daae5afc135388d9761122) Merge branch 'Comfy-Org:master' into dev/rocm-revert-chunk ### 📊 Changes **2 files changed** (+2 additions, -12 deletions) <details> <summary>View changed files</summary> 📝 `comfy_aimdo/vram_buffer.py` (+1 -7) 📝 `src/vrambuf.c` (+1 -5) </details> ### 📄 Description Was intended to fix ROCm Windows' chunk size in `vram_buffer.py`, but it seems fine to revert Linux's as well, since we had this workaround to fix VMM bug: https://github.com/Comfy-Org/comfy-aimdo/blob/7bbb4d4ac07aa27eb4aeef4fddd0410d82357870/src/plat.h#L14-L20 Wheel for testing: https://github.com/Comfy-Org/comfy-aimdo/actions/runs/26936503053 ### 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). --- <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

2026-10-26 22:24:29 +00:00

bsakaguchi referenced this issue

2026-10-26 22:24:40 +00:00

[PR #93] [MERGED] Implement a Memory Compiler #112