[GH-ISSUE #57] aimdo.dll crashes with ACCESS_VIOLATION on CUDA 13.x #16

Closed
opened 2026-10-26 22:23:38 +00:00 by bsakaguchi · 1 comment
Owner

Originally created by @hahajaja on GitHub (May 31, 2026).
Original GitHub issue: https://github.com/Comfy-Org/comfy-aimdo/issues/57

## Description

imdo.dll crashes with OSError: exception: access violation writing 0x0000000000000000 when lib.init() is called on a system with CUDA 13.x driver.

Error

`
Windows fatal exception: access violation

Current thread 0x0000e430 (most recent call first):
File "...\comfy_aimdo\control.py", line 98 in init_devices
if lib.init(device_array, len(requested)):
OSError: exception: access violation writing 0x0000000000000000
`

Environment

  • OS: Windows 10
  • Python: 3.11.9
  • PyTorch: 2.10.0+cu130
  • CUDA: 13.0 (PyTorch) / 13.1 (Driver 591.59)
  • GPU: NVIDIA GeForce RTX 3070 Laptop GPU
  • comfy-aimdo: 0.4.7

Steps to Reproduce

  1. Install PyTorch 2.10+cu130 with CUDA 13.x driver
  2. Install comfy-aimdo 0.4.7
  3. Run software that calls comfy_aimdo.control.init_devices
  4. Crash in lib.init(device_array, len(requested))

Root Cause

Access violation writing to address x0 (NULL dereference) inside imdo.dll's init function. The DLL was likely compiled against CUDA 12.x. A CUDA API call returns NULL on CUDA 13.x and the return value is dereferenced without checking.

README states Cuda 12.8+ support, but it is incompatible with CUDA 13.x.

Workaround

Use --disable-dynamic-vram flag to bypass comfy_aimdo entirely.

Originally created by @hahajaja on GitHub (May 31, 2026). Original GitHub issue: https://github.com/Comfy-Org/comfy-aimdo/issues/57 ## Description imdo.dll crashes with OSError: exception: access violation writing 0x0000000000000000 when lib.init() is called on a system with CUDA 13.x driver. ## Error ` Windows fatal exception: access violation Current thread 0x0000e430 (most recent call first): File "...\comfy_aimdo\control.py", line 98 in init_devices if lib.init(device_array, len(requested)): OSError: exception: access violation writing 0x0000000000000000 ` ## Environment - **OS**: Windows 10 - **Python**: 3.11.9 - **PyTorch**: 2.10.0+cu130 - **CUDA**: 13.0 (PyTorch) / 13.1 (Driver 591.59) - **GPU**: NVIDIA GeForce RTX 3070 Laptop GPU - **comfy-aimdo**: 0.4.7 ## Steps to Reproduce 1. Install PyTorch 2.10+cu130 with CUDA 13.x driver 2. Install comfy-aimdo 0.4.7 3. Run software that calls comfy_aimdo.control.init_devices 4. Crash in lib.init(device_array, len(requested)) ## Root Cause Access violation writing to address x0 (NULL dereference) inside imdo.dll's init function. The DLL was likely compiled against CUDA 12.x. A CUDA API call returns NULL on CUDA 13.x and the return value is dereferenced without checking. README states Cuda 12.8+ support, but it is incompatible with CUDA 13.x. ## Workaround Use --disable-dynamic-vram flag to bypass comfy_aimdo entirely.
Author
Owner

@rattus128 commented on GitHub (Jun 11, 2026):

bad facts from AI

<!-- gh-comment-id:4677752896 --> @rattus128 commented on GitHub (Jun 11, 2026): bad facts from AI
Sign in to join this conversation.
No labels
pull-request
No milestone
No project
No assignees
1 participant
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set.

Reference
Comfy-Org/comfy-aimdo#16
No description provided.