Index: llvm/trunk/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
===================================================================
--- llvm/trunk/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
+++ llvm/trunk/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
@@ -7171,7 +7171,11 @@
       SDValue Trunc = DAG.getNode(ISD::TRUNCATE, SDLoc(N0),
                                   N0.getValueType(), ExtLoad);
       ExtendSetCCUses(SetCCs, Trunc, ExtLoad, DL, ISD::SIGN_EXTEND);
-      CombineTo(N0.getNode(), Trunc, ExtLoad.getValue(1));
+      // If the load value is used only by N, replace it via CombineTo N.
+      if (N0.getValue(0).hasOneUse())
+        DAG.ReplaceAllUsesOfValueWith(SDValue(LN0, 1), ExtLoad.getValue(1));
+      else
+        CombineTo(LN0, Trunc, ExtLoad.getValue(1));
       return CombineTo(N, ExtLoad); // Return N so it doesn't get rechecked!
     }
   }
@@ -7229,7 +7233,11 @@
                                     SDLoc(N0.getOperand(0)),
                                     N0.getOperand(0).getValueType(), ExtLoad);
         ExtendSetCCUses(SetCCs, Trunc, ExtLoad, DL, ISD::SIGN_EXTEND);
-        CombineTo(N0.getOperand(0).getNode(), Trunc, ExtLoad.getValue(1));
+        // If the load value is used only by N, replace it via CombineTo N.
+        if (SDValue(LN0, 0).hasOneUse())
+          DAG.ReplaceAllUsesOfValueWith(SDValue(LN0, 1), ExtLoad.getValue(1));
+        else
+          CombineTo(LN0, Trunc, ExtLoad.getValue(1));
         return CombineTo(N, And); // Return N so it doesn't get rechecked!
       }
     }
@@ -7470,7 +7478,11 @@
       SDValue Trunc = DAG.getNode(ISD::TRUNCATE, SDLoc(N0),
                                   N0.getValueType(), ExtLoad);
       ExtendSetCCUses(SetCCs, Trunc, ExtLoad, SDLoc(N), ISD::ZERO_EXTEND);
-      CombineTo(N0.getNode(), Trunc, ExtLoad.getValue(1));
+      // If the load value is used only by N, replace it via CombineTo N.
+      if (SDValue(LN0, 0).hasOneUse())
+        DAG.ReplaceAllUsesOfValueWith(SDValue(LN0, 1), ExtLoad.getValue(1));
+      else
+        CombineTo(LN0, Trunc, ExtLoad.getValue(1));
       return CombineTo(N, ExtLoad); // Return N so it doesn't get rechecked!
     }
   }
@@ -7522,7 +7534,11 @@
                                     SDLoc(N0.getOperand(0)),
                                     N0.getOperand(0).getValueType(), ExtLoad);
         ExtendSetCCUses(SetCCs, Trunc, ExtLoad, DL, ISD::ZERO_EXTEND);
-        CombineTo(N0.getOperand(0).getNode(), Trunc, ExtLoad.getValue(1));
+        // If the load value is used only by N, replace it via CombineTo N.
+        if (SDValue(LN0, 0).hasOneUse())
+          DAG.ReplaceAllUsesOfValueWith(SDValue(LN0, 1), ExtLoad.getValue(1));
+        else
+          CombineTo(LN0, Trunc, ExtLoad.getValue(1));
         return CombineTo(N, And); // Return N so it doesn't get rechecked!
       }
     }
@@ -7695,13 +7711,16 @@
                                        LN0->getChain(),
                                        LN0->getBasePtr(), N0.getValueType(),
                                        LN0->getMemOperand());
-      CombineTo(N, ExtLoad);
       SDValue Trunc = DAG.getNode(ISD::TRUNCATE, SDLoc(N0),
                                   N0.getValueType(), ExtLoad);
-      CombineTo(N0.getNode(), Trunc, ExtLoad.getValue(1));
       ExtendSetCCUses(SetCCs, Trunc, ExtLoad, SDLoc(N),
                       ISD::ANY_EXTEND);
-      return SDValue(N, 0);   // Return N so it doesn't get rechecked!
+      // If the load value is used only by N, replace it via CombineTo N.
+      if (N0.hasOneUse())
+        DAG.ReplaceAllUsesOfValueWith(SDValue(LN0, 1), ExtLoad.getValue(1));
+      else
+        CombineTo(LN0, Trunc, ExtLoad.getValue(1));
+      return CombineTo(N, ExtLoad); // Return N so it doesn't get rechecked!
     }
   }
 
Index: llvm/trunk/test/CodeGen/X86/avx512-mask-op.ll
===================================================================
--- llvm/trunk/test/CodeGen/X86/avx512-mask-op.ll
+++ llvm/trunk/test/CodeGen/X86/avx512-mask-op.ll
@@ -1630,8 +1630,9 @@
 ; CHECK-LABEL: f1:
 ; CHECK:       ## BB#0: ## %entry
 ; CHECK-NEXT:    movzbl {{.*}}(%rip), %edi
-; CHECK-NEXT:    movl %edi, %eax
-; CHECK-NEXT:    xorb $1, %al
+; CHECK-NEXT:    movb {{.*}}(%rip), %al
+; CHECK-NEXT:    notb %al
+; CHECK-NEXT:    andb $1, %al
 ; CHECK-NEXT:    movb %al, {{.*}}(%rip)
 ; CHECK-NEXT:    xorl $1, %edi
 ; CHECK-NEXT:    jmp _f2 ## TAILCALL
Index: llvm/trunk/test/CodeGen/X86/pr32515.ll
===================================================================
--- llvm/trunk/test/CodeGen/X86/pr32515.ll
+++ llvm/trunk/test/CodeGen/X86/pr32515.ll
@@ -0,0 +1,29 @@
+; RUN: llc -O0 -mtriple=x86_64-unknown -mcpu=skx -o - %s
+; RUN: llc     -mtriple=x86_64-unknown -mcpu=skx -o - %s
+; RUN: llc -O0 -mtriple=i686-unknown   -mcpu=skx -o - %s
+; RUN: llc     -mtriple=i686-unknown   -mcpu=skx -o - %s
+; REQUIRES: asserts
+
+@var_26 = external global i16, align 2
+
+define void @foo() #0 {
+ %1 = alloca i16, align 2
+ %2 = load i16, i16* @var_26, align 2
+ %3 = zext i16 %2 to i32
+ %4 = icmp ne i32 %3, 7
+ %5 = zext i1 %4 to i16
+ store i16 %5, i16* %1, align 2
+ %6 = load i16, i16* @var_26, align 2
+ %7 = zext i16 %6 to i32
+ %8 = and i32 1, %7
+ %9 = shl i32 %8, 0
+ %10 = load i16, i16* @var_26, align 2
+ %11 = zext i16 %10 to i32
+ %12 = icmp ne i32 %11, 7
+ %13 = zext i1 %12 to i32
+ %14 = and i32 %9, %13
+ %15 = icmp ne i32 %14, 0
+ %16 = zext i1 %15 to i8
+ store i8 %16, i8* undef, align 1
+ unreachable
+ }