TensorRT-10.5.0.18 nvinfer_10.dll possibly corrupted or not fully Windows made?

AI & Data Science Deep Learning (Training & Inference) TensorRT
1

Description

When running a very simple inference C++ API test with TensorRT-10.5.0.18 having a crash even before starting main(), just on nvinfer_10.dll initialization.
The error is
Exception thrown at 0x00007FFEE3EF3F18 (nvinfer_10.dll) in Test2D.exe: 0xC0000005: Access violation reading location 0xFFFFFFFFFFFFFFFF.
If say run Profile in Dependency Walker then the error is “The application was unable to start correctly (0xc0000142)” which is again an obvious dll initialization error.

PS, Running as an Administrator doesn’t help, still the same issue.

Environment

TensorRT-10.5.0.18:
RTX 3080 Laptop:
Driver 556.12:
Cuda 11.8:
cudnn-8.6.0.163, cudnn-9.5.0.50:
Windows 10, 11:
Python Version (if applicable):
TensorFlow Version (if applicable):
PyTorch Version (if applicable):
Baremetal or Container (if container which image + tag):

Relevant Files

Please attach or include links to any models, data, files, or scripts necessary to reproduce your issue. (Github repo, Google Drive, Dropbox, etc.)

Steps To Reproduce

Please include:

  • Exact steps/commands to build your repro
  • Exact steps/commands to run your repro
  • Full traceback of errors encountered
5

Hi @PavloTytarenko ,
Can you pls share the onnx model.

Thanks

6

Oh, thank you for the answer! Didn’t hope already for it.))
The matter is that the problem is not in the model. It is not reaching any model stage. The crash is before main() starts on the stage of dll loading.
It’s crashing in exe_common.inl on line 256: _initterm(__xc_a, __xc_z); This is if to run from Visual Studio and look in debugger. If just to run .exe then there will be what I described in the starting post.

Thanks!

9

The problem solved. One of the 3rdParty libraries were built for lower cuda version. So guys - always check that everything is built with the actual TRT version.))