[llvm] [AArch64] merge index address with large offset into base address (PR #72187)

[via llvm-commits]

vfdff wrote:

> I still seem to see some performance issues from this, in the TSVC benchmarks in the llvm test suite. Do you see the same thing?
hi @davemgreen : 
  I don't see any changes from this patch with command `clang -march=armv8-a+sve -O3 -ffp-contract=fast -mllvm -scalable-vectorization=preferred -ffast-math -fvectorize  -S  tsvc.c`, would you show your command, thanks.

https://github.com/llvm/llvm-project/pull/72187