I failed to find the definition of 16_MB despite of precompiler of sampleMNIST.cpp, is there anyone kowns its header file?
I don’t think that macro is exposed.
However, if you want to change the value of the workspace size, it is just an integer number of bytes. 16_MiB is just for convenience and readability, but you can specify custom values like 1 << 20 (~1MB), 1 << 24 (~16MB), etc.
See the example here: https://docs.nvidia.com/deeplearning/sdk/tensorrt-developer-guide/index.html#build_engine_c ( config->setMaxWorkspaceSize(1 << 20) )
NVES_R, thank you for reply, i just confused how to make 16_MB unexposed, by the way, 16_MB maybe not a macro, because it’s not identifier.