[all-commits] [llvm/llvm-project] dbe8de: [AArch64] Lower mathlib call ldexp into fscale whe...

[huhu233 via All-commits]

  Branch: refs/heads/main
  Home:   https://github.com/llvm/llvm-project
  Commit: dbe8def9ccd1b8d50e12c8ee66ee509a86e72510
      https://github.com/llvm/llvm-project/commit/dbe8def9ccd1b8d50e12c8ee66ee509a86e72510
  Author: huhu233 <zhangtiehu at huawei.com>
  Date:   2023-10-24 (Tue, 24 Oct 2023)

  Changed paths:
    M llvm/lib/Target/AArch64/AArch64ISelLowering.cpp
    A llvm/test/CodeGen/AArch64/ldexp.ll

  Log Message:
  -----------
  [AArch64] Lower mathlib call ldexp into fscale when sve is enabled (#67552)

The function of 'fscale' is equivalent to mathlib call ldexp, but has
better performance. This patch lowers ldexp into fscale when sve is
enabled.