[PATCH] D142682: AMDGPU: Combine down fcopysign f64 magnitude

[Matt Arsenault via Phabricator via llvm-commits]

arsenm added a comment.

In D142682#4085325 <https://reviews.llvm.org/D142682#4085325>, @foad wrote:

>> This mostly improves the cases involving constants.
>
> It would be nice to see a test case that demonstrates that.

See s_test_copysign_f64_0_mag, it let a constant fold into an operand and avoided a BFI (I'm adding this test in a precommit)


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D142682/new/

https://reviews.llvm.org/D142682