Demo GEMM execution in multiple platforms

Introduction

This program executes the same GEMM function on a GPU and CPU, demonstrating SYCL's capability to execute the same code on different platforms.

It prints the first elements of the resulting matrix.

Building and running

From the source root directory:

sh scripts/build.sh
bin/gemm_demo

Building and running the SYCL version

Build and install oneMKL for CUDA (https://oneapi-src.github.io/oneMKL/building_the_project.html#building-with-cmake)

From the root directory:

The script will take the path to the SYCL mkl root path from the SYCL_MKL environmental variable or as an argument.

ssh scripts/build.sh [path to the sycl mkl root dir]
export LD_LIBRARY_PATH=<path to the sycl mkl lib>:$LD_LIBRARY_PATH
bin/gemm_usm_gpu_cpu