[PATCH] D142641: AMDGPU: Force sign operand of f64 fcopysign to f32

[Jay Foad via Phabricator via llvm-commits]

foad accepted this revision.
foad added inline comments.
This revision is now accepted and ready to land.


================
Comment at: llvm/lib/Target/AMDGPU/SIISelLowering.cpp:9445
+  //   fcopysign f64:x,
+  //       (bitcast (extract_vector_elt (bitcast f64:y to v2i32), 1))
+  // TODO: In some cases it might make sense to go all the way to f16.
----------------
arsenm wrote:
> foad wrote:
> > Can't you bitcast f64 to v2f32, to avoid the second bitcast?
> Yes, but surprisingly this loses the load width reduction optimization
Ugh.


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D142641/new/

https://reviews.llvm.org/D142641