[PATCH] Convert a vselect into a concat_vector if possible

[hfinkel at anl.gov]

Aside from fixing up the test, this LGTM.

================
Comment at: test/CodeGen/X86/vselect.ll:269
@@ +268,3 @@
+; CHECK: movlhps %xmm{{[23]}}, %xmm{{[01]}}
+; CHECK: movlhps %xmm{{[23]}}, %xmm{{[01]}}
+; CHECK: addps %xmm{{[01]}}, %xmm{{[01]}}
----------------
I think can be made a little more specific. You know what the input registers are (because they're dictated by the calling convention), so you can do something like this:

  ; CHECK-DAG: movlhps %xmm2, %xmm{{[01]}}
  ; CHECK-DAG: movlhps %xmm3, %xmm{{[01]}}

(where the CHECK-DAGs can match in either order). Because the add is commutative, it is hard to pattern match the resulting registers in a stable way (because they could be flipped). If you make the test use a subtract (so that the correct order is fixed), then we can be even more specific:

  ; CHECK-DAG: movlhps %xmm2, %[[REG1:[xmm[01]]]]
  ; CHECK-DAG: movlhps %xmm3, %[[REG2:[xmm[01]]]]
  ; CHECK-NEXT: subps [[REG1]], [[REG2]]
(or something like that).

http://reviews.llvm.org/D3916