What I don’t quite understand in the discussions here is the hardware focus. No offense, but the libraries Intel is putting out there (IPP, MKL) for efficient

What I don’t quite understand in the discussions here is the hardware focus. No offense, but the libraries Intel is putting out there (IPP, MKL) for efficient parallel computing are really outstanding. It has been a fee years for me in the end-consumer high-performance market, but AMD chips would easily run 20% slower if you optimized the code w/ the Intel Libraries and OpenCL would Not get anywhere near. A similar thing happened w/ Nvidia and Apple many more years back; E.g., Apple dropping Nvidia; fair amount of rumors went around that Apple did that so Adobe would be less competitive if they had to rewrite their rendering engine (which they just ported to CUDA) and Apple could sell their video editing solutions with an edge (they have been hand tuned for a while of course...). Apple dropping Nvidia has been such a “non-customer-focused” bullshit decision. Not because the AMD hardware is bad - but OpenCL has simply been nowhere near CUDA. Same with Intel IPPs and OpenCL. But maybe that changed. The differences in performance, ease of use etc. have been mind-blowing back in the days for CUDA and IPP. Hardware alone isn’t gonna cut it. The machines are built to run software after all... Maybe that has changed or maybe nobody needs any parallelised compute intensive applications on Mac (research anyone, image and video editing anyone?). The fact aside that mayor software companies like Adobe et al. will probably have to staff entire departments for rewrites...