[ARM][MVE] Enable VMOVN for tail predication

These instructions also don't exchange lanes, so make them legal.

Differential Revision: https://reviews.llvm.org/D75669

Author: sparker-arm

llvm/lib/Target/ARM/ARMInstrMVE.td

@@ -4421,6 +4421,7 @@ class MVE_VxMOVxN<string iname, string suffix, bit bit_28, bit bit_17,
   let Inst{8} = 0b0;
   let Inst{7} = !if(!eq(bit_17, 0), 1, 0);
   let Inst{0} = 0b1;
+  let validForTailPredication = 1;
 }
 
 multiclass MVE_VxMOVxN_halves<string iname, string suffix,

llvm/unittests/Target/ARM/MachineInstrTest.cpp

@@ -256,6 +256,10 @@ TEST(MachineInstrValidTailPredication, IsCorrect) {
     case MVE_VMOVimmi32:	
     case MVE_VMOVimmi64:
     case MVE_VMOVimmi8:	
+    case MVE_VMOVNi16bh:
+    case MVE_VMOVNi16th:
+    case MVE_VMOVNi32bh:
+    case MVE_VMOVNi32th:
     case MVE_VMULLBp16:
     case MVE_VMULLBp8:
     case MVE_VMULLBs16:
@@ -321,6 +325,18 @@ TEST(MachineInstrValidTailPredication, IsCorrect) {
     case MVE_VQNEGs16:
     case MVE_VQNEGs32:
     case MVE_VQNEGs8:
+    case MVE_VQMOVNs16bh:
+    case MVE_VQMOVNs16th:
+    case MVE_VQMOVNs32bh:
+    case MVE_VQMOVNs32th:
+    case MVE_VQMOVNu16bh:
+    case MVE_VQMOVNu16th:
+    case MVE_VQMOVNu32bh:
+    case MVE_VQMOVNu32th:
+    case MVE_VQMOVUNs16bh:
+    case MVE_VQMOVUNs16th:
+    case MVE_VQMOVUNs32bh:
+    case MVE_VQMOVUNs32th:
     case MVE_VQRSHL_by_vecs16:
     case MVE_VQRSHL_by_vecs32:
     case MVE_VQRSHL_by_vecs8: