The result type of EXTRACT_VECTOR_ELT doesn't have to match the element type of

the input vector, it can be bigger (this is helpful for powerpc where <2 x i16>
is a legal vector type but i16 isn't a legal type, IIRC).  However this wasn't
being taken into account by ExpandRes_EXTRACT_VECTOR_ELT, causing PR13220.
Lightly tweaked version of a patch by Michael Liao.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@160116 91177308-0d34-0410-b5e6-96231b3b80d8

diff --git a/lib/CodeGen/SelectionDAG/LegalizeTypesGeneric.cpp b/lib/CodeGen/SelectionDAG/LegalizeTypesGeneric.cpp
index a8ff7c65abdec56508fbb61650fa864a42b86da6..06f6bd63b671b6b303a00db0d4dcbceefa9e3438 100644 (file)
--- a/lib/CodeGen/SelectionDAG/LegalizeTypesGeneric.cpp
+++ b/lib/CodeGen/SelectionDAG/LegalizeTypesGeneric.cpp
@@ -168,6 +168,7 @@ void DAGTypeLegalizer::ExpandRes_EXTRACT_VECTOR_ELT(SDNode *N, SDValue &Lo,
                                                     SDValue &Hi) {
   SDValue OldVec = N->getOperand(0);
   unsigned OldElts = OldVec.getValueType().getVectorNumElements();
+  EVT OldEltVT = OldVec.getValueType().getVectorElementType();
   DebugLoc dl = N->getDebugLoc();
 
   // Convert to a vector of the expanded element type, for example
@@ -175,6 +176,15 @@ void DAGTypeLegalizer::ExpandRes_EXTRACT_VECTOR_ELT(SDNode *N, SDValue &Lo,
   EVT OldVT = N->getValueType(0);
   EVT NewVT = TLI.getTypeToTransformTo(*DAG.getContext(), OldVT);
 
+  if (OldVT != OldEltVT) {
+    // The result of EXTRACT_VECTOR_ELT may be larger than the element type of
+    // the input vector.  If so, extend the elements of the input vector to the
+    // same bitwidth as the result before expanding.
+    assert(OldEltVT.bitsLT(OldVT) && "Result type smaller then element type!");
+    EVT NVecVT = EVT::getVectorVT(*DAG.getContext(), OldVT, OldElts);
+    OldVec = DAG.getNode(ISD::ANY_EXTEND, dl, NVecVT, N->getOperand(0));
+  }
+
   SDValue NewVec = DAG.getNode(ISD::BITCAST, dl,
                                EVT::getVectorVT(*DAG.getContext(),
                                                 NewVT, 2*OldElts),

diff --git a/test/CodeGen/X86/pr13220.ll b/test/CodeGen/X86/pr13220.ll
new file mode 100644 (file)
index 0000000..aed6957
--- /dev/null
+++ b/test/CodeGen/X86/pr13220.ll
@@ -0,0 +1,20 @@
+; RUN: llc < %s
+; PR13220
+
+define <8 x i32> @foo(<8 x i96> %x) {
+  %a = lshr <8 x i96> %x, <i96 1, i96 1, i96 1, i96 1, i96 1, i96 1, i96 1, i96 1>
+  %b = trunc <8 x i96> %a to <8 x i32>
+  ret <8 x i32> %b
+}
+
+define <8 x i32> @bar(<8 x i97> %x) {
+  %a = lshr <8 x i97> %x, <i97 1, i97 1, i97 1, i97 1, i97 1, i97 1, i97 1, i97 1>
+  %b = trunc <8 x i97> %a to <8 x i32>
+  ret <8 x i32> %b
+}
+
+define <8 x i32> @bax() {
+  %a = lshr <8 x i96> <i96 4, i96 4, i96 4, i96 4, i96 4, i96 4, i96 4, i96 4>, <i96 1, i96 1, i96 1, i96 1, i96 1, i96 1, i96 1, i96 1>
+  %b = trunc <8 x i96> %a to <8 x i32>
+  ret <8 x i32> %b
+}