Index: llvm/lib/Target/X86/X86InstrInfo.cpp
===================================================================
--- llvm/lib/Target/X86/X86InstrInfo.cpp
+++ llvm/lib/Target/X86/X86InstrInfo.cpp
@@ -9223,6 +9223,32 @@
   if (Load)
     BeforeOps.push_back(SDValue(Load, 0));
   BeforeOps.insert(BeforeOps.end(), AfterOps.begin(), AfterOps.end());
+  // Change CMP32ri r, 0 back to TEST32rr r, r, etc.
+  switch (Opc) {
+    default: break;
+    case X86::CMP64ri32:
+    case X86::CMP64ri8:
+    case X86::CMP32ri:
+    case X86::CMP32ri8:
+    case X86::CMP16ri:
+    case X86::CMP16ri8:
+    case X86::CMP8ri:
+      if (auto *C = dyn_cast<ConstantSDNode>(BeforeOps[1].getNode())) {
+        if(C->getConstantIntValue()->isZeroValue()) {
+          switch (Opc) {
+            default: llvm_unreachable("Unreachable!");
+            case X86::CMP64ri8:
+            case X86::CMP64ri32: Opc = X86::TEST64rr; break;
+            case X86::CMP32ri8:
+            case X86::CMP32ri:   Opc = X86::TEST32rr; break;
+            case X86::CMP16ri8:
+            case X86::CMP16ri:   Opc = X86::TEST16rr; break;
+            case X86::CMP8ri:    Opc = X86::TEST8rr; break;
+          }
+          BeforeOps[1] = BeforeOps[0];
+        }
+      }
+  }
   SDNode *NewNode= DAG.getMachineNode(Opc, dl, VTs, BeforeOps);
   NewNodes.push_back(NewNode);
 
Index: llvm/test/CodeGen/X86/2009-10-19-atomic-cmp-eflags.ll
===================================================================
--- llvm/test/CodeGen/X86/2009-10-19-atomic-cmp-eflags.ll
+++ llvm/test/CodeGen/X86/2009-10-19-atomic-cmp-eflags.ll
@@ -1,6 +1,6 @@
 ; RUN: llvm-as <%s | llc | FileCheck %s
 ; PR 5247
-; check that cmp is not scheduled before the add
+; check that cmp/test is not scheduled before the add
 target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128"
 target triple = "x86_64-unknown-linux-gnu"
 
@@ -38,7 +38,7 @@
   %5 = sub i64 %4, %2                             ; <i64> [#uses=1]
   %6 = atomicrmw add i64* getelementptr inbounds ([1216 x i64], [1216 x i64]* @__profiling_callsite_timestamps_live, i32 0, i32 51), i64 %5 monotonic
 ;CHECK: lock {{xadd|addq}} %rdx, __profiling_callsite_timestamps_live
-;CHECK-NEXT: cmpl $0,
+;CHECK-NEXT: testl [[REG:%e[a-z]+]], [[REG]]
 ;CHECK-NEXT: jne
   %cmp = icmp eq i32 %3, 0                        ; <i1> [#uses=1]
   br i1 %cmp, label %if.then, label %if.end
Index: llvm/test/CodeGen/X86/2012-01-16-mfence-nosse-flags.ll
===================================================================
--- llvm/test/CodeGen/X86/2012-01-16-mfence-nosse-flags.ll
+++ llvm/test/CodeGen/X86/2012-01-16-mfence-nosse-flags.ll
@@ -14,9 +14,7 @@
 ; clobbers EFLAGS.
 
 ; CHECK: lock orl {{.*}}, (%esp)
-; CHECK-NEXT: cmpl $0, [[REG:%e[a-z]+]]
-; unfoldMemoryOperand should to convert this back to testl. 
-; FIXME-NEXT: testl [[REG:%e[a-z]+]], [[REG]]
+; CHECK-NEXT: testl [[REG:%e[a-z]+]], [[REG]]
 
 if.then:                                          ; preds = %entry
   tail call void bitcast (void (...)* @foo to void ()*)() nounwind