TensorRT make error


when I build “sample_FasterRCNN” based on TensorRT3.0.1 on TitanGPU

and error occurs:

TensorRT-3.0.1/lib/libnvinfer_plugin.so: undefined reference to `std::__throw_out_of_range_fmt(char const*, …)@GLIBCXX_3.4.20

TensorRT-3.0.1/lib/libnvcaffe_parser.so: undefined reference to `__cxa_throw_bad_array_new_length@CXXABI_1.3.8’
collect2: error: ld returned 1 exit status

but TensorRT3.0.0 has no such error when build despite couldn’t be runned after build for not supported in Titan

with ver3.0.0

thanks for your help!


can anybody give some advice?

Probably related to your libc and compiler versions. Did you make sure you downloaded the correct one? 14 vs 16 ubuntu?

Thx! Error is really caused by incorrect version for ubuntu

this time, I have checked my cuda and cudnn version already

However, I face another error:

cudnnLayerUtils.cpp(288)-Cuda Error in smVersion: 35

in IBuilder* builder = createInferBuilder(gLogger);

Should I revise anything in makefile?

for avoiding error:

cudnnLayerUtils.cpp(288)-Cuda Error in smVersion: 35

I have met this issue too.
This issue may cause by set wrong environment variable “CUDA_VISIBLE_DEVICES”.

There is 2 way to figure out:

  1. export CUDA_VISIBLE_DEVIDES=;make clean;make

  2. reboot the machine and make clean and make again.