Skip to content

README.md

Latest commit

 

History

History
43 lines (27 loc) · 1.22 KB

README.md

File metadata and controls

43 lines (27 loc) · 1.22 KB

Template Generator

This example shows how to use our script to generate a interception template for the CUDA driver library.

Prerequisites

  • CUDA driver library (libcuda.so)
  • CUDA header file (cuda.h)
  • Python 3.8+
  • libclang (install using pip install libclang)

Step 1: Generate the intercept template

export GEN_SCRIPT_DIR=../../scripts # path to the gen.py script
export CUDA_HEADER_DIR=/usr/local/cuda/include # path to the CUDA header file
export CUDA_LIB_PATH=/usr/lib/x86_64-linux-gnu/libcuda.so # path to the CUDA library
export GCC_HEADER_PATH=/usr/lib/gcc/x86_64-linux-gnu/11/include # path to the GCC header file

python3 ${GEN_SCRIPT_DIR}/gen.py -s ${CUDA_HEADER_DIR}/cuda.h --platform cuda --lib ${CUDA_LIB_PATH} --prefix cu -I ${GCC_HEADER_PATH}

For convenience, we'd better copy the CUDA header file to the current directory.

cp ${CUDA_HEADER_DIR}/cuda.h .

Step 2: Generate cuGetProcAddress related code

python3 ${GEN_SCRIPT_DIR}/cuda_func_map.py -f ${CUDA_HEADER_DIR}/cudaTypedefs.h -I ${GCC_HEADER_PATH} -I ${CUDA_HEADER_DIR} >> intercept.cpp

Step 3: Process cuGetProcAddress related code

python3 ${GEN_SCRIPT_DIR}/process_proc_addr.py -f intercept.cpp