AMD GPU OpenCL Benchmarking

The code in this repository allows benchmarking an AMD GPU's computing power using the OpenCL framework against a CPU.

Installing prerequisites

The followings need to be installed in order to have an OpenCL setup working with an AMD GPU:

• The amdgpu-pro driver from AMD

• The OpenCL-Headers from KhronosGroup (The benchmark Makefile assumes that these headers are present in the ~/src/OpenCL-Headers/ directory. This can be customized.)

Running benchmarks

Provided that the OpenCL library has been properly installed (and the Makefile possibly adapted), make builds the following binaries:

• cl-info provides information about the OpenCL hardware on the machine.

• main performs the actual GPU benchmark using the OpenCL framework. It generates the bench_gpu_loops.txt and bench_gpu_workers.txt files. They contain the time elapsed in nanoseconds (right column) for computing a given number of operations (left column), respectively for a single worker and for many workers.

• main-cpu performs the same computation on the computer's CPU than the GPU benchmark tool. This allows comparing the CPU's computational power against the GPU's. It generates both bench_cpu_loops.txt and bench_cpu_workers.txt files, which kind of translate the benchmark performed on the GPU, though on the CPU. There is almost no difference between these 2 files, since the computation is performed serialy anyway on the CPU.

• gnuplot display.plot displays a graph and prints some statistics from results generated by main and main-cpu.