OpenCL-Benchmark v1.6

• automatically use zero-copy buffers on CPUs/iGPUs to reduce memory footprint
• bandwidth kernels now write non-zero data, to avoid hardware optimizations for zero-initialized buffers