[AMDGPU][True16][GlobalISel] Fix v2*16 build_vector patterns

- Pattern with IMPLICIT_DEF failed to generate an entry in MatchTable and
did not an report error, just silently failed. This is fixed by casting
IMPLICIT_DEF to appropriate type. This also fixes selecting
"build_vector s16, undef" for GlobalISel with True16.
- Add pattern for "build_vector undef, s16" that will work for GlobalISel.
True16 GlobalISel has a G_TRUNC that it needs to deal with.
- Use REG_SEQUENCE for Real16 patterns instead of V_LSHLREV_B32_e64 to
generate more optimal code.

Author: mbrkusanin

llvm/lib/Target/AMDGPU/SIInstructions.td

@@ -3543,15 +3543,29 @@ def : GCNPat <
   (vecTy (UniformBinFrag<build_vector> (Ty undef), (Ty SReg_32:$src1))),
   (S_LSHL_B32 SReg_32:$src1, (i32 16))
 >;
-}
 
 def : GCNPat <
   (vecTy (DivergentBinFrag<build_vector> (Ty undef), (Ty VGPR_32:$src1))),
   (vecTy (V_LSHLREV_B32_e64 (i32 16), VGPR_32:$src1))
 >;
-} // End foreach Ty = ...
 }
 
+let True16Predicate = UseRealTrue16Insts in
+def : GCNPat <
+  (vecTy (DivergentBinFrag<build_vector> (Ty undef), (Ty VGPR_32:$src1))),
+  (REG_SEQUENCE VGPR_32, (Ty (IMPLICIT_DEF)), lo16, (Ty VGPR_32:$src1), hi16)
+>;
+
+} // End foreach Ty = ...
+} // End AddedComplexity = 1
+
+let True16Predicate = UseRealTrue16Insts in
+def : GCNPat <
+  (v2i16 (DivergentBinFrag<build_vector> (i16 undef), (i16 (trunc i32:$src1)))),
+  (REG_SEQUENCE VGPR_32, (i16 (IMPLICIT_DEF)), lo16,
+                         (i16 (EXTRACT_SUBREG VGPR_32:$src1, lo16)), hi16)
+>;
+
 let SubtargetPredicate = HasVOP3PInsts in {
 foreach p = [NotHasTrue16BitInsts, UseFakeTrue16Insts] in
 let True16Predicate = p in
@@ -3599,7 +3613,7 @@ def : GCNPat <
 >;
 def : GCNPat <
   (vecTy (DivergentBinFrag<build_vector> (Ty VGPR_16:$src0), (Ty undef))),
-  (REG_SEQUENCE VGPR_32, $src0, lo16, (IMPLICIT_DEF), hi16)
+  (REG_SEQUENCE VGPR_32, $src0, lo16, (Ty (IMPLICIT_DEF)), hi16)
 >;
 }