[llvm] r197135 - [AArch64]Fix the problem that AArch64 backend fails to select scalar_to_vector of vector types having more than one element.

[Hao Liu]

Author: haoliu
Date: Thu Dec 12 01:36:26 2013
New Revision: 197135

URL: http://llvm.org/viewvc/llvm-project?rev=197135&view=rev
Log:
[AArch64]Fix the problem that AArch64 backend fails to select scalar_to_vector of vector types having more than one element.

Modified:
    llvm/trunk/lib/Target/AArch64/AArch64InstrNEON.td
    llvm/trunk/test/CodeGen/AArch64/neon-copy.ll

Modified: llvm/trunk/lib/Target/AArch64/AArch64InstrNEON.td
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/AArch64/AArch64InstrNEON.td?rev=197135&r1=197134&r2=197135&view=diff
==============================================================================
--- llvm/trunk/lib/Target/AArch64/AArch64InstrNEON.td (original)
+++ llvm/trunk/lib/Target/AArch64/AArch64InstrNEON.td Thu Dec 12 01:36:26 2013
@@ -6707,14 +6707,48 @@ def : Pat<(v1i32 (scalar_to_vector GPR32
 def : Pat<(v1i64 (scalar_to_vector GPR64:$src)),
           (FMOVdx $src)>;
 
+def : Pat<(v8i8 (scalar_to_vector GPR32:$Rn)),
+          (v8i8 (EXTRACT_SUBREG (v16i8
+            (INSbw (v16i8 (IMPLICIT_DEF)), $Rn, (i64 0))),
+            sub_64))>;
+
+def : Pat<(v4i16 (scalar_to_vector GPR32:$Rn)),
+          (v4i16 (EXTRACT_SUBREG (v8i16
+            (INShw (v8i16 (IMPLICIT_DEF)), $Rn, (i64 0))),
+            sub_64))>;
+
+def : Pat<(v2i32 (scalar_to_vector GPR32:$Rn)),
+          (v2i32 (EXTRACT_SUBREG (v16i8
+            (INSsw (v4i32 (IMPLICIT_DEF)), $Rn, (i64 0))),
+            sub_64))>;
+
+def : Pat<(v16i8 (scalar_to_vector GPR32:$Rn)),
+          (INSbw (v16i8 (IMPLICIT_DEF)), $Rn, (i64 0))>;
+
+def : Pat<(v8i16 (scalar_to_vector GPR32:$Rn)),
+          (INShw (v8i16 (IMPLICIT_DEF)), $Rn, (i64 0))>;
+
+def : Pat<(v4i32 (scalar_to_vector GPR32:$Rn)),
+          (INSsw (v4i32 (IMPLICIT_DEF)), $Rn, (i64 0))>;
+
+def : Pat<(v2i64 (scalar_to_vector GPR64:$Rn)),
+          (INSdx (v2i64 (IMPLICIT_DEF)), $Rn, (i64 0))>;
+
+def : Pat<(v2i32 (scalar_to_vector GPR32:$Rn)),
+          (v2i32 (EXTRACT_SUBREG (v16i8
+            (INSsw (v4i32 (IMPLICIT_DEF)), $Rn, (i64 0))),
+            sub_64))>;
+
+def : Pat<(v2i32 (scalar_to_vector GPR32:$Rn)),
+          (v2i32 (EXTRACT_SUBREG (v16i8
+            (INSsw (v4i32 (IMPLICIT_DEF)), $Rn, (i64 0))),
+            sub_64))>;
+
 def : Pat<(v1f32 (scalar_to_vector (f32 FPR32:$Rn))),
           (v1f32 FPR32:$Rn)>;
 def : Pat<(v1f64 (scalar_to_vector (f64 FPR64:$Rn))),
           (v1f64 FPR64:$Rn)>;
 
-def : Pat<(v1f64 (scalar_to_vector (f64 FPR64:$src))),
-          (FMOVdd $src)>;
-
 def : Pat<(v2f64 (scalar_to_vector (f64 FPR64:$src))),
           (INSERT_SUBREG (v2f64 (IMPLICIT_DEF)),
                          (f64 FPR64:$src), sub_64)>;

Modified: llvm/trunk/test/CodeGen/AArch64/neon-copy.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/AArch64/neon-copy.ll?rev=197135&r1=197134&r2=197135&view=diff
==============================================================================
--- llvm/trunk/test/CodeGen/AArch64/neon-copy.ll (original)
+++ llvm/trunk/test/CodeGen/AArch64/neon-copy.ll Thu Dec 12 01:36:26 2013
@@ -612,4 +612,54 @@ define <1 x double> @test_bitcasti64tov1
    %res = bitcast i64 %in to <1 x double>
 ; CHECK: fmov {{d[0-9]+}}, {{x[0-9]+}}
    ret <1 x double> %res
+}
+
+; Test insert element into an undef vector
+define <8 x i8> @scalar_to_vector.v8i8(i8 %a) {
+; CHECK-LABEL: scalar_to_vector.v8i8:
+; CHECK: ins {{v[0-9]+}}.b[0], {{w[0-9]+}}
+  %b = insertelement <8 x i8> undef, i8 %a, i32 0
+  ret <8 x i8> %b
+}
+
+define <16 x i8> @scalar_to_vector.v16i8(i8 %a) {
+; CHECK-LABEL: scalar_to_vector.v16i8:
+; CHECK: ins {{v[0-9]+}}.b[0], {{w[0-9]+}}
+  %b = insertelement <16 x i8> undef, i8 %a, i32 0
+  ret <16 x i8> %b
+}
+
+define <4 x i16> @scalar_to_vector.v4i16(i16 %a) {
+; CHECK-LABEL: scalar_to_vector.v4i16:
+; CHECK: ins {{v[0-9]+}}.h[0], {{w[0-9]+}}
+  %b = insertelement <4 x i16> undef, i16 %a, i32 0
+  ret <4 x i16> %b
+}
+
+define <8 x i16> @scalar_to_vector.v8i16(i16 %a) {
+; CHECK-LABEL: scalar_to_vector.v8i16:
+; CHECK: ins {{v[0-9]+}}.h[0], {{w[0-9]+}}
+  %b = insertelement <8 x i16> undef, i16 %a, i32 0
+  ret <8 x i16> %b
+}
+
+define <2 x i32> @scalar_to_vector.v2i32(i32 %a) {
+; CHECK-LABEL: scalar_to_vector.v2i32:
+; CHECK: ins {{v[0-9]+}}.s[0], {{w[0-9]+}}
+  %b = insertelement <2 x i32> undef, i32 %a, i32 0
+  ret <2 x i32> %b
+}
+
+define <4 x i32> @scalar_to_vector.v4i32(i32 %a) {
+; CHECK-LABEL: scalar_to_vector.v4i32:
+; CHECK: ins {{v[0-9]+}}.s[0], {{w[0-9]+}}
+  %b = insertelement <4 x i32> undef, i32 %a, i32 0
+  ret <4 x i32> %b
+}
+
+define <2 x i64> @scalar_to_vector.v2i64(i64 %a) {
+; CHECK-LABEL: scalar_to_vector.v2i64:
+; CHECK: ins {{v[0-9]+}}.d[0], {{x[0-9]+}}
+  %b = insertelement <2 x i64> undef, i64 %a, i32 0
+  ret <2 x i64> %b
 }
\ No newline at end of file