Teach DAG combine to fold (extract_subvec (concat v1, ..) i) to v_i

- If the extracted vector has the same type of all vectored being concatenated
  together, it should be simplified directly into v_i, where i is the index of
  the element being extracted.

llvm-svn: 166125

2 files changed, 34 insertions, 2 deletions

diff --git a/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp b/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
index 6e4a772a89d..4ac6d1b5163 100644
--- a/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
+++ b/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
@@ -8610,8 +8610,8 @@ SDValue DAGCombiner::visitEXTRACT_SUBVECTOR(SDNode* N) {
       return SDValue();
 
     // Only handle cases where both indexes are constants with the same type.
-    ConstantSDNode *InsIdx = dyn_cast<ConstantSDNode>(N->getOperand(1));
-    ConstantSDNode *ExtIdx = dyn_cast<ConstantSDNode>(V->getOperand(2));
+    ConstantSDNode *ExtIdx = dyn_cast<ConstantSDNode>(N->getOperand(1));
+    ConstantSDNode *InsIdx = dyn_cast<ConstantSDNode>(V->getOperand(2));
 
     if (InsIdx && ExtIdx &&
         InsIdx->getValueType(0).getSizeInBits() <= 64 &&
@@ -8628,6 +8628,21 @@ SDValue DAGCombiner::visitEXTRACT_SUBVECTOR(SDNode* N) {
     }
   }
 
+  if (V->getOpcode() == ISD::CONCAT_VECTORS) {
+    // Combine:
+    //    (extract_subvec (concat V1, V2, ...), i)
+    // Into:
+    //    Vi if possible
+    for (unsigned i = 0, e = V->getNumOperands(); i != e; ++i)
+      if (V->getOperand(i).getValueType() != NVT)
+        return SDValue();
+    unsigned Idx = dyn_cast<ConstantSDNode>(N->getOperand(1))->getZExtValue();
+    unsigned NumElems = NVT.getVectorNumElements();
+    assert((Idx % NumElems) == 0 &&
+           "IDX in concat is not a multiple of the result vector length.");
+    return V->getOperand(Idx / NumElems);
+  }
+
   return SDValue();
 }
 
diff --git a/llvm/test/CodeGen/X86/extract-concat.ll b/llvm/test/CodeGen/X86/extract-concat.ll
new file mode 100644
index 00000000000..5c8d20a651d
--- /dev/null
+++ b/llvm/test/CodeGen/X86/extract-concat.ll
@@ -0,0 +1,17 @@
+; RUN: llc < %s -mcpu=corei7 -mtriple=x86_64-unknown-linux-gnu | FileCheck %s
+
+define void @foo(<4 x float> %in, <4 x i8>* %out) {
+  %t0 = fptoui <4 x float> %in to <4 x i32>
+  %t1 = trunc <4 x i32> %t0 to <4 x i16>
+  %t2 = shufflevector <4 x i16> %t1, <4 x i16> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+  %t3 = trunc <8 x i16> %t2 to <8 x i8>
+  %t4 = shufflevector <8 x i8> %t3, <8 x i8> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+  %t5 = insertelement <4 x i8> %t4, i8 -1, i32 3
+  store <4 x i8> %t5, <4 x i8>* %out
+  ret void
+; CHECK: foo
+; CHECK: cvttps2dq
+; CHECK-NOT: pextrd
+; CHECK: pshufb
+; CHECK: ret
+}