3 files changed, 28 insertions, 4 deletions
diff --git a/llvm/lib/Target/X86/X86ISelDAGToDAG.cpp b/llvm/lib/Target/X86/X86ISelDAGToDAG.cpp
index c42a05f46ba..4386028e9f5 100644
--- a/llvm/lib/Target/X86/X86ISelDAGToDAG.cpp
+++ b/llvm/lib/Target/X86/X86ISelDAGToDAG.cpp
@@ -2218,6 +2218,25 @@ SDNode *X86DAGToDAGISel::Select(SDNode *Node) {
     return CurDAG->SelectNodeTo(Node, ShlOp, NVT, SDValue(New, 0),
                                 getI8Imm(ShlVal));
   }
+  case X86ISD::UMUL8:
+  case X86ISD::SMUL8: {
+    SDValue N0 = Node->getOperand(0);
+    SDValue N1 = Node->getOperand(1);
+
+    Opc = (Opcode == X86ISD::SMUL8 ? X86::IMUL8r : X86::MUL8r);
+
+    SDValue InFlag = CurDAG->getCopyToReg(CurDAG->getEntryNode(), dl, X86::AL,
+                                          N0, SDValue()).getValue(1);
+
+    SDVTList VTs = CurDAG->getVTList(NVT, MVT::i32);
+    SDValue Ops[] = {N1, InFlag};
+    SDNode *CNode = CurDAG->getMachineNode(Opc, dl, VTs, Ops);
+
+    ReplaceUses(SDValue(Node, 0), SDValue(CNode, 0));
+    ReplaceUses(SDValue(Node, 1), SDValue(CNode, 1));
+    return nullptr;
+  }
+
   case X86ISD::UMUL: {
     SDValue N0 = Node->getOperand(0);
     SDValue N1 = Node->getOperand(1);
diff --git a/llvm/lib/Target/X86/X86ISelLowering.cpp b/llvm/lib/Target/X86/X86ISelLowering.cpp
index 73cbd563f5a..dbe3c4aee1c 100644
--- a/llvm/lib/Target/X86/X86ISelLowering.cpp
+++ b/llvm/lib/Target/X86/X86ISelLowering.cpp
@@ -1597,9 +1597,6 @@ void X86TargetLowering::resetOperationActions() {
     setOperationAction(ISD::UMULO, VT, Custom);
   }
 
-  // There are no 8-bit 3-address imul/mul instructions
-  setOperationAction(ISD::SMULO, MVT::i8, Expand);
-  setOperationAction(ISD::UMULO, MVT::i8, Expand);
 
   if (!Subtarget->is64Bit()) {
     // These libcalls are not available in 32-bit.
@@ -18190,10 +18187,15 @@ static SDValue LowerXALUO(SDValue Op, SelectionDAG &DAG) {
     Cond = X86::COND_B;
     break;
   case ISD::SMULO:
-    BaseOp = X86ISD::SMUL;
+    BaseOp = N->getValueType(0) == MVT::i8 ? X86ISD::SMUL8 : X86ISD::SMUL;
     Cond = X86::COND_O;
     break;
   case ISD::UMULO: { // i64, i8 = umulo lhs, rhs --> i64, i64, i32 umul lhs,rhs
+    if (N->getValueType(0) == MVT::i8) {
+      BaseOp = X86ISD::UMUL8;
+      Cond = X86::COND_O;
+      break;
+    }
     SDVTList VTs = DAG.getVTList(N->getValueType(0), N->getValueType(0),
                                  MVT::i32);
     SDValue Sum = DAG.getNode(X86ISD::UMUL, DL, VTs, LHS, RHS);
diff --git a/llvm/lib/Target/X86/X86ISelLowering.h b/llvm/lib/Target/X86/X86ISelLowering.h
index f63b821ce6a..e8e611d0719 100644
--- a/llvm/lib/Target/X86/X86ISelLowering.h
+++ b/llvm/lib/Target/X86/X86ISelLowering.h
@@ -301,6 +301,9 @@ namespace llvm {
 
       UMUL, // LOW, HI, FLAGS = umul LHS, RHS
 
+      // 8-bit SMUL/UMUL - AX, FLAGS = smul8/umul8 AL, RHS
+      SMUL8, UMUL8,
+
       // MUL_IMM - X86 specific multiply by immediate.
       MUL_IMM,