diff --git a/llvm/lib/Target/AMDGPU/AMDGPUISelDAGToDAG.cpp b/llvm/lib/Target/AMDGPU/AMDGPUISelDAGToDAG.cpp
--- a/llvm/lib/Target/AMDGPU/AMDGPUISelDAGToDAG.cpp
+++ b/llvm/lib/Target/AMDGPU/AMDGPUISelDAGToDAG.cpp
@@ -1050,6 +1050,30 @@
         {LHS, RHS, CI,
          CurDAG->getTargetConstant(0, {}, MVT::i1) /*clamp bit*/});
   } else {
+    // The result of a boolean operation is represented as a 32-bit/64-bit sgpr
+    // with bits set potentially even for inactive lanes, so mask them out here.
+    // Masking is not needed for SETCC and class intrinsics. We also exclude
+    // trivial patterns with two SETCCs connected by a boolean operation.
+    if (CI->getValueType(0) == MVT::i1) {
+      unsigned Opc = CI->getOpcode();
+      bool SkipMasking = (Opc == ISD::SETCC || Opc == AMDGPUISD::FP_CLASS);
+      SkipMasking |= (Opc == ISD::AND || Opc == ISD::OR || Opc == ISD::XOR) &&
+                     (CI->getOperand(0).getOpcode() == ISD::SETCC &&
+                      CI->getOperand(1).getOpcode() == ISD::SETCC);
+
+      if (!SkipMasking) {
+        const auto *ST = static_cast<const GCNSubtarget *>(Subtarget);
+        CI = SDValue(
+            CurDAG->getMachineNode(
+                ST->isWave32() ? AMDGPU::S_AND_B32 : AMDGPU::S_AND_B64, DL,
+                MVT::i1,
+                CurDAG->getRegister(
+                    ST->isWave32() ? AMDGPU::EXEC_LO : AMDGPU::EXEC, MVT::i1),
+                CI),
+            0);
+      }
+    }
+
     unsigned Opc = N->getOpcode() == ISD::ADDCARRY ? AMDGPU::S_ADD_CO_PSEUDO
                                                    : AMDGPU::S_SUB_CO_PSEUDO;
     CurDAG->SelectNodeTo(N, Opc, N->getVTList(), {LHS, RHS, CI});
diff --git a/llvm/test/CodeGen/AMDGPU/combine-add-zext-xor.ll b/llvm/test/CodeGen/AMDGPU/combine-add-zext-xor.ll
new file mode 100644
--- /dev/null
+++ b/llvm/test/CodeGen/AMDGPU/combine-add-zext-xor.ll
@@ -0,0 +1,183 @@
+; RUN: llc -march=amdgcn -start-before=amdgpu-isel  -stop-after=amdgpu-isel -mtriple=amdgcn--amdpal -verify-machineinstrs < %s | FileCheck %s
+
+; Test that unused lanes are masked out in the s_xor result before it is used as condition code.
+
+; CHECK-LABEL: name:            combine_add_zext_xor
+; CHECK: S_XOR_B64
+; CHECK-NEXT: S_AND_B64 $exec
+; CHECK: S_ADD_CO_PSEUDO
+
+define void @combine_add_zext_xor() {
+.entry:
+  br label %.exit, !amdgpu.uniform !0
+
+.exit:                                            ; preds = %bb7, %.entry
+  %t = phi i32 [ 1050, %.entry ], [ 0, %bb7 ]
+  %t1 = phi i32 [ 0, %.entry ], [ %t7, %bb7 ]
+  br i1 undef, label %bb7, label %bb, !amdgpu.uniform !0
+
+bb:                                               ; preds = %.exit
+  %t2 = add i32 %t, 2
+  %t3 = icmp eq i32 %t2, 0
+  br label %bb7, !amdgpu.uniform !0
+
+bb7:                                              ; preds = %bb, %.exit
+  %t4 = phi i1 [ %t3, %bb ], [ undef, %.exit ]
+  %t5 = xor i1 %t4, true
+  %t6 = zext i1 %t5 to i32
+  %t7 = add i32 %t1, %t6
+  br label %.exit, !amdgpu.uniform !0
+}
+
+; Test that unused lanes are masked out in the s_and result before it is used as condition code.
+
+; CHECK-LABEL: name:            combine_add_zext_and
+; CHECK: S_AND_B64
+; CHECK-NEXT: S_AND_B64 $exec
+; CHECK: S_ADD_CO_PSEUDO
+
+define void @combine_add_zext_and() {
+.entry:
+  br label %.exit, !amdgpu.uniform !0
+
+.exit:                                            ; preds = %bb7, %.entry
+  %t = phi i32 [ 1050, %.entry ], [ 0, %bb7 ]
+  %t1 = phi i32 [ 0, %.entry ], [ %t7, %bb7 ]
+  br i1 undef, label %bb7, label %bb, !amdgpu.uniform !0
+
+bb:                                               ; preds = %.exit
+  %t2 = add i32 %t, 2
+  %t3 = icmp eq i32 %t2, 0
+  br label %bb7, !amdgpu.uniform !0
+
+bb7:                                              ; preds = %bb, %.exit
+  %t4 = phi i1 [ %t3, %bb ], [ undef, %.exit ]
+  %t4a = phi i1 [ %t3, %bb ], [ undef, %.exit ]
+  %t5 = and i1 %t4, %t4a
+  %t6 = zext i1 %t5 to i32
+  %t7 = add i32 %t1, %t6
+  br label %.exit, !amdgpu.uniform !0
+}
+
+; Test that unused lanes are masked out in the s_or result before it is used as condition code.
+
+; CHECK-LABEL: name:            combine_add_zext_or
+; CHECK: S_OR_B64
+; CHECK-NEXT: S_AND_B64 $exec
+; CHECK: S_ADD_CO_PSEUDO
+
+define void @combine_add_zext_or() {
+.entry:
+  br label %.exit, !amdgpu.uniform !0
+
+.exit:                                            ; preds = %bb7, %.entry
+  %t = phi i32 [ 1050, %.entry ], [ 0, %bb7 ]
+  %t1 = phi i32 [ 0, %.entry ], [ %t7, %bb7 ]
+  br i1 undef, label %bb7, label %bb, !amdgpu.uniform !0
+
+bb:                                               ; preds = %.exit
+  %t2 = add i32 %t, 2
+  %t3 = icmp eq i32 %t2, 0
+  br label %bb7, !amdgpu.uniform !0
+
+bb7:                                              ; preds = %bb, %.exit
+  %t4 = phi i1 [ %t3, %bb ], [ undef, %.exit ]
+  %t4a = phi i1 [ %t3, %bb ], [ undef, %.exit ]
+  %t5 = or i1 %t4, %t4a
+  %t6 = zext i1 %t5 to i32
+  %t7 = add i32 %t1, %t6
+  br label %.exit, !amdgpu.uniform !0
+}
+
+; Test that unused lanes are masked out in the s_xor result before it is used as condition code.
+
+; CHECK-LABEL: name:            combine_sub_zext_xor
+; CHECK: S_XOR_B64
+; CHECK-NEXT: S_AND_B64 $exec
+; CHECK: S_SUB_CO_PSEUDO
+
+define void @combine_sub_zext_xor() {
+.entry:
+  br label %.exit, !amdgpu.uniform !0
+
+.exit:                                            ; preds = %bb7, %.entry
+  %t = phi i32 [ 1050, %.entry ], [ 0, %bb7 ]
+  %t1 = phi i32 [ 0, %.entry ], [ %t7, %bb7 ]
+  br i1 undef, label %bb7, label %bb, !amdgpu.uniform !0
+
+bb:                                               ; preds = %.exit
+  %t2 = add i32 %t, 2
+  %t3 = icmp eq i32 %t2, 0
+  br label %bb7, !amdgpu.uniform !0
+
+bb7:                                              ; preds = %bb, %.exit
+  %t4 = phi i1 [ %t3, %bb ], [ undef, %.exit ]
+  %t5 = xor i1 %t4, true
+  %t6 = zext i1 %t5 to i32
+  %t7 = sub i32 %t1, %t6
+  br label %.exit, !amdgpu.uniform !0
+}
+
+; Test that unused lanes are masked out in the s_and result before it is used as condition code.
+
+; CHECK-LABEL: name:            combine_sub_zext_and
+; CHECK: S_AND_B64
+; CHECK-NEXT: S_AND_B64 $exec
+; CHECK: S_SUB_CO_PSEUDO
+
+define void @combine_sub_zext_and() {
+.entry:
+  br label %.exit, !amdgpu.uniform !0
+
+.exit:                                            ; preds = %bb7, %.entry
+  %t = phi i32 [ 1050, %.entry ], [ 0, %bb7 ]
+  %t1 = phi i32 [ 0, %.entry ], [ %t7, %bb7 ]
+  br i1 undef, label %bb7, label %bb, !amdgpu.uniform !0
+
+bb:                                               ; preds = %.exit
+  %t2 = add i32 %t, 2
+  %t3 = icmp eq i32 %t2, 0
+  br label %bb7, !amdgpu.uniform !0
+
+bb7:                                              ; preds = %bb, %.exit
+  %t4 = phi i1 [ %t3, %bb ], [ undef, %.exit ]
+  %t4a = phi i1 [ %t3, %bb ], [ undef, %.exit ]
+  %t5 = and i1 %t4, %t4a
+  %t6 = zext i1 %t5 to i32
+  %t7 = sub i32 %t1, %t6
+  br label %.exit, !amdgpu.uniform !0
+}
+
+; Test that unused lanes are masked out in the s_or result before it is used as condition code.
+
+; CHECK-LABEL: name:            combine_sub_zext_or
+; CHECK: S_OR_B64
+; CHECK-NEXT: S_AND_B64 $exec
+; CHECK: S_SUB_CO_PSEUDO
+
+define void @combine_sub_zext_or() {
+.entry:
+  br label %.exit, !amdgpu.uniform !0
+
+.exit:                                            ; preds = %bb7, %.entry
+  %t = phi i32 [ 1050, %.entry ], [ 0, %bb7 ]
+  %t1 = phi i32 [ 0, %.entry ], [ %t7, %bb7 ]
+  br i1 undef, label %bb7, label %bb, !amdgpu.uniform !0
+
+bb:                                               ; preds = %.exit
+  %t2 = add i32 %t, 2
+  %t3 = icmp eq i32 %t2, 0
+  br label %bb7, !amdgpu.uniform !0
+
+bb7:                                              ; preds = %bb, %.exit
+  %t4 = phi i1 [ %t3, %bb ], [ undef, %.exit ]
+  %t4a = phi i1 [ %t3, %bb ], [ undef, %.exit ]
+  %t5 = or i1 %t4, %t4a
+  %t6 = zext i1 %t5 to i32
+  %t7 = sub i32 %t1, %t6
+  br label %.exit, !amdgpu.uniform !0
+}
+
+attributes #0 = { nounwind readonly willreturn }
+
+!0 = !{}