It turns out 256bit V[SZ]EXT nodes are still
generated by the new shuffle lowering, so this
is here to stay!
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@231422
91177308-0d34-0410-b5e6-
96231b3b80d8
(!cast<I>(OpcPrefix#DQYrr) VR128:$src)>;
// On AVX2, we also support 256bit inputs.
- // FIXME: remove these patterns when the old shuffle lowering goes away.
def : Pat<(v16i16 (ExtOp (v32i8 VR256:$src))),
(!cast<I>(OpcPrefix#BWYrr) (EXTRACT_SUBREG VR256:$src, sub_xmm))>;
def : Pat<(v8i32 (ExtOp (v32i8 VR256:$src))),