System Information

OpenCV version: 4.8.0 (current 4.x branch)
Windows
Compiler: msvc 2022

Running a onnx file (same as here: #22713 )
I noticed very high CPU usage (GPU also gets used but its bottlenecked by the CPU)
So we went and profiled with intel vtune

Turns out that most of the time is spent on opt_AVC2::fastGEMM1T
is this op only available for CPU? why doesn't the full model run on GPU

Cuda runtime is enabled I can run many other models without issue on GPU

Detailed description

Steps to reproduce

Very basic just use net->forward with the model linked in the other issue (https://drive.google.com/file/d/1UHkudK8wjHoE9UM81x9H07nVRz_xjT9i/view?usp=sharing)

Issue submission checklist

• I report the issue, it's not a question
• I checked the problem with documentation, FAQ, open issues, forum.opencv.org, Stack Overflow, etc and have not found any solution
• I updated to the latest OpenCV version and the issue is still there
• There is reproducer code and related data files (videos, images, onnx, etc)