[llvm] [AArch64] Optimization of repeated constant loads (#51483) (PR #86249)

[Eli Friedman via llvm-commits]

efriedma-quic wrote:

> orr(x, x, lsl n) is usually as cheap as a movk.

On which processors?  It looks like on Cortex-X2 and newer, "orr(x, x, lsl n)" has the same cost as movk, but looking through various sources, I can't find any other chip that does.  Granted, it's probably still better than generating two MOVK.

https://github.com/llvm/llvm-project/pull/86249