dp4a now is also supported on AMD RDNA3+ GPUs

Updated OpenCL-Wrapper, added hardware-supported dp4a for AMD/Intel G…

…PUs too, updated benchmark examples in Readme

Minor cosmetics

Automatically use zero-copy buffers on CPUs/iGPUs, bandwidth kernels …

…now write non-zero data

Removed wait() call at the end of benchmark on Linux

Cosmetics

Fixed terrible performance on ARM GPUs by macro-replacing fused-multi…

…ply-add (fma) with a*b+c, added automatic OS detection in make.sh

Added operating system info to OpenCL device driver version printout

Fixed several issues with macOS

Minor cosmetics in Readme