[llvm-branch-commits] [llvm] release/21.x: [AArch64] Keep floating-point conversion in SIMD (#147707) (PR #151317)

[Guy David via llvm-branch-commits]

guy-david wrote:

This optimization is important for us to land in this release, since it affects internal workloads we care about. It's rather conservative and concerns a very specific pattern around floating-point conversions in the AArch64 backend. Safety is above all, so we should definitely wait for the the more expensive bots to complete their verification.

https://github.com/llvm/llvm-project/pull/151317