sycl-microbench

This is a micro-benchmark for testing the overhead of SYCL features, the following features are tested in this benchmark

• memory allocation
• parallelization
• atomics
• barriers
• vectorization

Building

CMake is used to build this benchmark.

mkdir build && cd build
cmake .. -DSYCL_COMPILE= DPCPP|HIPSYCL -DOMP_COMPILE=true|false
make

if HIPSYCL is chosen as a SYCL implementation then -DHIPSYCL_INSTALL_DIR need to be specified.

For optimal performance OMP_PROC_BIND environment variable is set to true.

Example

./binary " [-s size |-b blocksize <optional>|\n
   	 --mat-mul : to run matrix multiplication \n
	 --vec-add : to run vector addition \n
	   can run only mat-mul or vec-add at a time, can't run both simultaneously \n
	 --mem-alloc : to alloc memory using SYCL and standard malloc \n
	 --reduction : to test reduction using atomics and sycl reduction construct
	 --range : to test sycl range construct
	 --ndrange : to test sycl nd_range construct
	 -i : for different routines in vectorization benchmark\n
	       1 - range with USM\n
	       2 - range with Buffer and Accessors\n
	       3 - nd_range with USM\n
	       4 - nd_range with Buffer and Accessor\n