As the ARM chips proliferate we will see NEON or GPU/neural engine APIs adopted in lieu of the SIMD from the x86 side. Until Now ARM just wasn’t relevant with the ubiquity of x86 and that the x86 family of chips have been the desktop standard since the 80s of course there will be hardware that takes advantage of this.
I wonder what final cut and other software suites tailored for the new Macs are doing under the covers as they seem to perform very quickly. Also this isn’t really all the relevant as cryptographic validation is required more and more on the server side of things whereas Apple doesn’t have a SKU for the server market.
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).
Ahh yeah sorry API was used wrong here. I knew it’s a type of CPU instruction set. Of course NEON is specific to ARM so x86 won’t get it. I didn’t know the neural engine is basically SIMD though.
I wonder what final cut and other software suites tailored for the new Macs are doing under the covers as they seem to perform very quickly. Also this isn’t really all the relevant as cryptographic validation is required more and more on the server side of things whereas Apple doesn’t have a SKU for the server market.