ARM64 architecture support in RyuJIT is close to preview quality in .NET Core 2.1, and was built from the start of its implementation on the RyuJIT architecture. We’ve actually been working on ARM64 support in RyuJIT for over four years, and this work was pushed forward more recently with significant work by Qualcomm contributors.
Overall, our RyuJIT investments have been focused on evolving the code base towards enabling better support for:
Multiple code generation targets (instruction sets and operating systems),
Improved optimizations,
Better and more flexible code generation, and
Open, flexible, and robust design and implementation.
We believe that the new RyuJIT compiler architecture is a major improvement over the (now-removed) legacy code generator for achieving these goals.
We have recently invested in new code generation technology solely in the RyuJIT code generator, for example, SIMD support, architecture-specific hardware intrinisics, and support for the Linux software conventions.
It’s very satisfying to get to this point, and we can already see how removing all this old code will free us to be more agile going forward.
Thanks to everyone who contributed to this long effort!