I’m trying to do parallel simultaneous inference for multiple networks using TensorRT.
I tried to realize it using cudaStream, but when I use the cudaStreamCreate function, it freezes when I enter that function.
It happens when cuda is implemented in the dll. It does not happen when cuda is implemented in exe without using dll
TensorRT Version : 22.214.171.124
GPU Type : RTX2070 or RTX2080
CUDA Version : v11.1
CUDNN Version : 8.1.1
Operating System + Version : Windows10 64bit
Can cudaStream be implemented in a dll?? Or do you need a special method?? teach me please.