Change the peep for EXTRACT_VECTOR_ELT of BUILD_PAIR to look for

the new CONCAT_VECTORS node type instead, as that's what legalize
uses now. And add a peep for EXTRACT_VECTOR_ELT of INSERT_VECTOR_ELT.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@38503 91177308-0d34-0410-b5e6-96231b3b80d8

diff --git a/lib/CodeGen/SelectionDAG/SelectionDAG.cpp b/lib/CodeGen/SelectionDAG/SelectionDAG.cpp
index 9687a52cdab6d4f694224d0777b51c987139db9c..dc23151bddd1f2c9307970c413deee88eaf58623 100644 (file)
--- a/lib/CodeGen/SelectionDAG/SelectionDAG.cpp
+++ b/lib/CodeGen/SelectionDAG/SelectionDAG.cpp
@@ -1968,19 +1968,31 @@ SDOperand SelectionDAG::getNode(unsigned Opcode, MVT::ValueType VT,
   case ISD::EXTRACT_VECTOR_ELT:
     assert(N2C && "Bad EXTRACT_VECTOR_ELT!");
 
-    // EXTRACT_VECTOR_ELT of BUILD_PAIR is often formed while lowering is
+    // EXTRACT_VECTOR_ELT of CONCAT_VECTORS is often formed while lowering is
     // expanding copies of large vectors from registers.
-    if (N1.getOpcode() == ISD::BUILD_PAIR) {
-      unsigned NewNumElts = MVT::getVectorNumElements(N1.getValueType()) / 2;
-      bool Low = N2C->getValue() < NewNumElts;
-      return getNode(ISD::EXTRACT_VECTOR_ELT, VT, N1.getOperand(!Low),
-                     Low ? N2 : getConstant(N2C->getValue() - NewNumElts,
-                                            N2.getValueType()));
+    if (N1.getOpcode() == ISD::CONCAT_VECTORS &&
+        N1.getNumOperands() > 0) {
+      unsigned Factor =
+        MVT::getVectorNumElements(N1.getOperand(0).getValueType());
+      return getNode(ISD::EXTRACT_VECTOR_ELT, VT,
+                     N1.getOperand(N2C->getValue() / Factor),
+                     getConstant(N2C->getValue() % Factor, N2.getValueType()));
     }
+
     // EXTRACT_VECTOR_ELT of BUILD_VECTOR is often formed while lowering is
     // expanding large vector constants.
     if (N1.getOpcode() == ISD::BUILD_VECTOR)
       return N1.getOperand(N2C->getValue());
+
+    // EXTRACT_VECTOR_ELT of INSERT_VECTOR_ELT is often formed when vector
+    // operations are lowered to scalars.
+    if (N1.getOpcode() == ISD::INSERT_VECTOR_ELT)
+      if (ConstantSDNode *IEC = dyn_cast<ConstantSDNode>(N1.getOperand(2))) {
+        if (IEC == N2C)
+          return N1.getOperand(1);
+        else
+          return getNode(ISD::EXTRACT_VECTOR_ELT, VT, N1.getOperand(0), N2);
+      }
     break;
   case ISD::EXTRACT_ELEMENT:
     assert(N2C && (unsigned)N2C->getValue() < 2 && "Bad EXTRACT_ELEMENT!");

diff --git a/test/CodeGen/X86/peep-vector-extract-concat.ll b/test/CodeGen/X86/peep-vector-extract-concat.ll
new file mode 100644 (file)
index 0000000..6880fc3
--- /dev/null
+++ b/test/CodeGen/X86/peep-vector-extract-concat.ll
@@ -0,0 +1,6 @@
+; RUN: llvm-as < %s | llc -march=x86-64 | grep {shufps \$3, %xmm0, %xmm0}
+
+define float @foo(<8 x float> %a) {
+  %c = extractelement <8 x float> %a, i32 3
+  ret float %c
+}

diff --git a/test/CodeGen/X86/peep-vector-extract-insert.ll b/test/CodeGen/X86/peep-vector-extract-insert.ll
new file mode 100644 (file)
index 0000000..9aeb4f1
--- /dev/null
+++ b/test/CodeGen/X86/peep-vector-extract-insert.ll
@@ -0,0 +1,12 @@
+; RUN: llvm-as < %s | llc -march=x86-64 | grep {pxor %xmm0, %xmm0} | wc -l | grep 2
+
+define float @foo(<4 x float> %a) {
+  %b = insertelement <4 x float> %a, float 0.0, i32 3
+  %c = extractelement <4 x float> %b, i32 3
+  ret float %c
+}
+define float @bar(float %a) {
+  %b = insertelement <4 x float> <float 3.4, float 4.5, float 0.0, float 9.2>, float %a, i32 3
+  %c = extractelement <4 x float> %b, i32 2
+  ret float %c
+}