NEON is not API, it is literally the SIMD instruction set for ARM. https://developer.arm.com/architectures/instruction-sets/simd-isas/neon https://developer.ar

NEON is not API, it is literally the SIMD instruction set for ARM. https://developer.arm.com/architectures/instruction-sets/simd-isas/neon https://developer.arm.com/architectures/instruction-sets/sim... So x86 is certainly not going to adopt it. The GPU/neural stuff is basically nothing else but SIMD, just in slightly different guise and running on different hardware. On PC it is the difference between MMX/AVX (CPU) and shaders/CUDA/"compute" (GPU), on M1 which is all-in-one SoC it is just a different part of the CPU being used, optimized for somewhat different tasks. SIMD is only a technical term for one type of parallel computation - single instruction, multiple data. It is not some sort of Intel-specific magic technology (and the term far predates Intel's support for it, going back to vector processing on Crays and such).