[llvm] [SVE][InstCombine] Fold ld1d and splice into ld1ro (PR #69565)

[Paul Walker via llvm-commits]

https://github.com/paulwalker-arm commented:

I cannot say I'm that happy with this patch because I think such fixed-length ACLE optimisations can quickly get out of control.  Also, by their nature they're affectively CPU specific whereas the VLS ACLE extension is a solution to embedding SVE types within C/C++ classes and arrays rather than a mechanism for low-level optimisation.

That's not to say I'm blocking such optimisations but I do wonder if there's an alternative.  For example, is it the case that you know within the code that you're wanting to splat a NEON ACLE type? If so do you know about `arm_neon_sve_bridge.h`? which was added a few releases ago to make it easier to transition between the SVE and NEON ACLE types.

https://github.com/llvm/llvm-project/pull/69565