It would be great if the LoongArch LSX/LASX intrinsics were rewritten using intrinsics::simd. This would allow for using them in Miri and would help the Cranelift or GCC codegen backends by not relying on LLVM directly. Most Some of the x86, ARM, s390x and PowerPC intrinsics are already making use of this.
Some examples of this being done for aarch64 code: #1872, #1871
For people interested in this, there's a good unofficial documentation on equivalents in SSE and AVX2: https://jia.je/unofficial-loongarch-intrinsics-guide/
cc @heiher, @RalfJung