Add a missing pattern for X86ISD::MOVLPD. rdar://10436044

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@144566 91177308-0d34-0410-b5e6-96231b3b80d8

diff --git a/lib/Target/X86/X86InstrSSE.td b/lib/Target/X86/X86InstrSSE.td
index caaf54490c33feaa51036e264d7f0f03dcf60fa0..6deee4f68ef3e96b05c137c4694816fc9187bffe 100644 (file)
--- a/lib/Target/X86/X86InstrSSE.td
+++ b/lib/Target/X86/X86InstrSSE.td
@@ -520,6 +520,8 @@ let Predicates = [HasSSE2] in {
   // is during lowering, where it's not possible to recognize the fold cause
   // it has two uses through a bitcast. One use disappears at isel time and the
   // fold opportunity reappears.
+  def : Pat<(v2f64 (X86Movlpd VR128:$src1, VR128:$src2)),
+            (MOVSDrr VR128:$src1, (EXTRACT_SUBREG (v2f64 VR128:$src2),sub_sd))>;
   def : Pat<(v4f32 (X86Movlps VR128:$src1, VR128:$src2)),
             (MOVSDrr VR128:$src1, (EXTRACT_SUBREG (v4f32 VR128:$src2),sub_sd))>;
   def : Pat<(v4i32 (X86Movlps VR128:$src1, VR128:$src2)),
@@ -647,6 +649,9 @@ let Predicates = [HasAVX] in {
   // is during lowering, where it's not possible to recognize the fold cause
   // it has two uses through a bitcast. One use disappears at isel time and the
   // fold opportunity reappears.
+  def : Pat<(v2f64 (X86Movlpd VR128:$src1, VR128:$src2)),
+            (VMOVSDrr VR128:$src1, (EXTRACT_SUBREG (v2f64 VR128:$src2),
+                                                   sub_sd))>;
   def : Pat<(v4f32 (X86Movlps VR128:$src1, VR128:$src2)),
             (VMOVSDrr VR128:$src1, (EXTRACT_SUBREG (v4f32 VR128:$src2),
                                                    sub_sd))>;

diff --git a/test/CodeGen/X86/vec_shuffle-39.ll b/test/CodeGen/X86/vec_shuffle-39.ll
index f9c00eb0c8924ea3e7ad9e925f33621628a93714..64508b529d6816d020dc5273cb76f2a46163eeb6 100644 (file)
--- a/test/CodeGen/X86/vec_shuffle-39.ll
+++ b/test/CodeGen/X86/vec_shuffle-39.ll
@@ -49,3 +49,21 @@ entry:
   store double %2, double* %0
   ret void
 }
+
+; rdar://10436044
+define <2 x double> @t3() nounwind readonly {
+bb:
+; CHECK: t3:
+; CHECK: punpcklqdq %xmm1, %xmm0
+; CHECK: movq (%rax), %xmm1
+; CHECK: movsd %xmm1, %xmm0
+  %tmp0 = load i128* null, align 1
+  %tmp1 = load <2 x i32>* undef, align 8
+  %tmp2 = bitcast i128 %tmp0 to <16 x i8>
+  %tmp3 = bitcast <2 x i32> %tmp1 to i64
+  %tmp4 = insertelement <2 x i64> undef, i64 %tmp3, i32 0
+  %tmp5 = bitcast <16 x i8> %tmp2 to <2 x double>
+  %tmp6 = bitcast <2 x i64> %tmp4 to <2 x double>
+  %tmp7 = shufflevector <2 x double> %tmp5, <2 x double> %tmp6, <2 x i32> <i32 2, i32 1>
+  ret <2 x double> %tmp7
+}