I am trying to integrate libtorch into my C++ codebase. I do however use Intel MKL internally and libtorch_cpu.so also exports mkl*
symbols - linking them together causes the symbols to get mixed up and results in segfaults.
Is there a way to isolate symbols exported by libtorch *.so
s or some other supported way for me to proceed?
I’m not using CMake
for the builds and have to integrate the process into existing infrastructure.