TApplencourt / GPU_IPC_Handle

Using IPC handle for different programing model

Sycl + level 0

dpcpp -lze_loader syclo_ipc_copy_dma_buf.cpp -o syclo_ipc_copy_dma_buf
SYCL_DEVICE_FILTER=PI_LEVEL_ZERO ./syclo_ipc_copy_dma_buf

Level 0

From: https://github.com/intel/compute-runtime/blob/master/level_zero/core/test/black_box_tests/zello_ipc_copy_dma_buf.cpp

g++ -lze_loader zello_ipc_copy_dma_buf.cpp -o zello_ipc_copy_dma_buf
./zello_ipc_copy_dma_buf

NVIDIA

Inspired by: https://github.com/heterodb/toybox/blob/master/cuda_ipc_open/cuda_ipc_open.c

nvcc -lcuda culo_ipc_copy_dma.buf.cpp -o culo_ipc_copy_dma.buf