[llvm] [GlobalISel] LegalizationArtifactCombiner for merge-unmerge->unmerge where src and dst of result would be different scalar/vector types (PR #213086)

[Petar Avramovic via llvm-commits]

================
@@ -1611,3 +1611,132 @@ body: |
     $vgpr0 = COPY %4
     $vgpr1 = COPY %5
 ...
+
+# Vector to scalar unmerge + merge combine tests.
+# G_UNMERGE_VALUES of a vector feeding G_MERGE_VALUES into a scalar
+# should fold to direct G_UNMERGE_VALUES of the source vector.
+
+---
+name: vector_to_scalar_merge_values
+body:             |
+  bb.0.entry:
+    liveins: $vgpr0_vgpr1
+
+    ; CHECK-LABEL: name: vector_to_scalar_merge_values
+    ; CHECK: liveins: $vgpr0_vgpr1
+    ; CHECK-NEXT: {{  $}}
+    ; CHECK-NEXT: [[COPY:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr0_vgpr1
+    ; CHECK-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](<4 x s16>)
+    ; CHECK-NEXT: [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](<4 x s16>)
+    ; CHECK-NEXT: $vgpr0 = COPY [[UV]](s32)
+    ; CHECK-NEXT: $vgpr1 = COPY [[UV3]](s32)
+    %0:_(<4 x s16>) = COPY $vgpr0_vgpr1
+    %1:_(s16), %2:_(s16), %3:_(s16), %4:_(s16) = G_UNMERGE_VALUES %0:_(<4 x s16>)
----------------
petar-avramovic wrote:

don't use scalar here, use integer (i16, <4 x i16> etc)

https://github.com/llvm/llvm-project/pull/213086