[PATCH] D82560: AMDGPU: Add llvm.amdgcn.sqrt intrinsic

[Jay Foad via Phabricator via llvm-commits]

foad added a comment.

Thanks! I assume the intention is that it should map directly to a v_sqrt_f* instruction, even if llvm.sqrt does something more sophisticated?



================
Comment at: llvm/lib/Target/AMDGPU/AMDGPUInstructions.td:819
+
+// FIXME: fsqrt shoulud not select directly
+def any_amdgcn_sqrt : PatFrags<(ops node:$src0),
----------------
Typo "shoulud".


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D82560/new/

https://reviews.llvm.org/D82560