SME is a crucial part of ARM's vision for AI
Arm's hardware advancements are bolstered by a sophisticated software ecosystem designed to exploit the full potential of its processors. At the heart of this ecosystem are the new Kleidi libraries, which play a crucial role in optimizing artificial intelligence (AI) and computer-based applications. These libraries provide developers with tools tailored to maximize the performance and efficiency of Arm's latest cores
KleidiAI is a key component that focuses on accelerating AI workloads. It includes a comprehensive set of computational kernels optimized for Arm's architecture, enabling efficient execution of various AI tasks such as machine learning, natural language processing, and data analytics. By offering highly optimized routines for common AI operations, KleidiAI allows developers to achieve significant performance gains while maintaining energy efficiency. This is increasingly important as AI applications become more prevalent in mobile devices, smart home systems, and industrial automation.
KleidiAI relies on two technologies: SVE2 and SME.
One could argue that, for the purpose of AI, SME is more important than SVE, since it deals with matrices.
Thing is they don't even have SVE 😞 And it's not trivial to implement (though it is more or less a prerequisite for SME).
M4 has SME but no SVE/SVE2.
Technically apple-m4 is ARMv9.2a, but a quirk of LLVM defines v9.0 as requiring SVE, which is optional according to the Arm ARM and not supported by the core. ARMv8.7a is the next closest choice
But beyond that if those features aren't already in Phoenix in some form then they aren't going to just get added so easily as it's effectively a very major optional feature.
I agree, it's not certain that Oryon V2/Pegasus/X Elite G2 will have SME. After all, even Cortex X925 (ARM's 2024 IP) doesn't support SME, as far as I am aware.
But in the long term, I think it's reasonable yo assume that Qualcomm will add SME sometime in the future.
Intel has already boldly proclaimed a "multi-engine AI strategy". That means having the ability to run AI on multiple blocks in the SoC: CPU, GPU and NPU.
Apple has made no such loud proclamations, but it's clear they also embrace the multi-engine AI strategy.
It seems ARM also believes in this vision, and hence they laid the groundwork with extensions such as SVE2, SME, SME2 and libraries/toolkits such as KleidiAI and KleidiCV.