I made some documentation for #WebAssembly SIMD, sort of an intrinsics guide but more. There is documentation for instructions, intrinsics, and WAV functions, plus lots of timing information for different architectures.
Does anyone know of a project using NEON intrinsics which might be interested in a port to, well, everything other than Arm (e.g., WebAssembly)? We're looking for goals to help set priorities for implementing the remaining NEON functions in SIMDe…
The portable implementation of SVE in SIMDe is working out better than I expected. Lots of API to left cover, but the infrastructure in place now is pretty solid so that can proceed relatively quickly.
@CompileExplore link if anyone wants to play with it:
Fujitsu's A64FX Microarchitecture Manual has lots of interesting details:
raw.githubusercontent.com/fujitsu/A64FX/…
A64FX is, AFAIK, the only production CPU which supports SVE. I also don't usually see this much public detail about Arm CPUs.