[all-commits] [llvm/llvm-project] 798358: [AMDGPU] Fix lowerFCOPYSIGN dropping the sign bit ...

[Arseniy Obolenskiy via All-commits]

  Branch: refs/heads/main
  Home:   https://github.com/llvm/llvm-project
  Commit: 798358faf4501d90e20b6ad45f138c5b660db60d
      https://github.com/llvm/llvm-project/commit/798358faf4501d90e20b6ad45f138c5b660db60d
  Author: Arseniy Obolenskiy <arseniy.obolenskiy at amd.com>
  Date:   2026-06-16 (Tue, 16 Jun 2026)

  Changed paths:
    M llvm/lib/Target/AMDGPU/SIISelLowering.cpp
    M llvm/test/CodeGen/AMDGPU/fcopysign.bf16.ll
    M llvm/test/CodeGen/AMDGPU/fcopysign.f16.ll

  Log Message:
  -----------
  [AMDGPU] Fix lowerFCOPYSIGN dropping the sign bit when narrowing the sign operand (#203492)

TRUNCATE of the v2i32-bitcast sign kept the low 16 bits of each lane but
dropped f32 sign bit at bit 31

Shift right by 16 first so the sign bit lands in the f16 sign position

---------

Co-authored-by: Jay Foad <jay.foad at gmail.com>



To unsubscribe from these emails, change your notification settings at https://github.com/llvm/llvm-project/settings/notifications