[PATCH] D117765: [AMDGPU][GlobalISel] Select source modifiers for VOP3Opsel

[Mirko Brkusanin via Phabricator via llvm-commits]

mbrkusanin added inline comments.


================
Comment at: llvm/test/CodeGen/AMDGPU/GlobalISel/vop3-op-sel.ll:100-105
+  %z1 = fneg <2 x half> %z
+  %z2 = shufflevector <2 x half> %z1, <2 x half> undef, <2 x i32> <i32 1, i32 0>
+  %z3 = fneg <2 x half> %z2
+  %z4 = shufflevector <2 x half> %z3, <2 x half> undef, <2 x i32> <i32 1, i32 0>
+  %z5 = extractelement <2 x half> %z4, i32 0
+  %res = call half @llvm.fma.f16(half %x, half %y, half %z5)
----------------
We should make combines for eliminating these. selectVOP3NoMods currently prevents this from being vop2 but other instructions would not have been eliminated otherwise.


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D117765/new/

https://reviews.llvm.org/D117765