I just don’t understand - Apple decided to implement a more capable SIMD instruction set and just … didn’t document it? SIMD is hugely important in anything remotely HPC related. My first gen apple silicon laptop still favorably competes against Intel’s current best, years later. Why would they hide such an advantage?
My guess is that it's not complete or stable or qualified enough for them to publish it as a supported feature. Maybe the power draw is too high and it triggers thermal throttling too quickly, similar to early Intel AVX-512? Maybe they are still sorting out how to have each process choose SVE vs NEON mode, so libraries using the other one don't cause a fault? (Note that this seems limited to M4 and newer CPUs.)
I agree this is an important feature. I have very often wished that my Mac had wider vectors.
Comments
I just don’t understand - Apple decided to implement a more capable SIMD instruction set and just … didn’t document it? SIMD is hugely important in anything remotely HPC related. My first gen apple silicon laptop still favorably competes against Intel’s current best, years later. Why would they hide such an advantage?
My guess is that it's not complete or stable or qualified enough for them to publish it as a supported feature. Maybe the power draw is too high and it triggers thermal throttling too quickly, similar to early Intel AVX-512? Maybe they are still sorting out how to have each process choose SVE vs NEON mode, so libraries using the other one don't cause a fault? (Note that this seems limited to M4 and newer CPUs.)
I agree this is an important feature. I have very often wished that my Mac had wider vectors.
Maybe they were waiting to announce it until it's been supported for a while?
Besides Apple who is doing hpc on Apple silicon
Apple shipped their “Neural Engine” in lots of iPhones before publicizing or fully documenting it, could be part of their playbook.