[X86] Don't emit KTEST instructions unless only the Z flag is being used

[craig.topper]

Summary

KTEST has weird flag behavior. The Z flag is set for all bits in the AND of the k-registers being 0, and the C flag is set for all bits being 1. All other flags are cleared.

We currently emit this instruction in EmitTEST and don't check the condition code. This can lead to strange things like using the S flag after a KTEST for a signed compare.

The domain reassignment pass can also transform TEST instructions into KTEST and is not protected against the flag usage either. For now I've disabled this part of the domain reassignment pass. I tried to comment out the checks in the mir test so that we could recover them later, but I couldn't figure out how to get that to work.

This patch moves the KTEST handling into LowerSETCC and now creates a ktest+x86setcc. I've chosen this approach because I'd like to add support for the C flag for all ones in a followup patch. To do that requires that I can rewrite the condition code going in the x86setcc to be different than the original SETCC condition code.

This fixes PR36182. I'll file a PR to fix domain reassignment once this goes in. Should this be merged to 6.0?

[craig.topper]

Simplify some code by checking VT against MVT enum values directly instead of using isVector, getVectorElementType and getSizeInBits.

[zvi]

lib/Target/X86/X86ISelLowering.cpp

18133 ↗

(On Diff #132729)

Which tests cover this?

[craig.topper]

lib/Target/X86/X86ISelLowering.cpp

18133 ↗

(On Diff #132729)

The tests that switched from ktest to kortest in D42772

[guyblank]

maybe use the update_mir_test_checks.py script for the mir test?

[craig.topper]

Re-generate test with update_mir_test_checks.py. Leave the TEST instructions in the MIR test but commented out with a FIXME

llvm/trunk/lib/Target/X86/X86DomainReassignment.cpp

if (STI->hasBWI()) {
     createReplacer(X86::AND64rr, X86::KANDQrr);
 
     createReplacer(X86::ANDN32rr, X86::KANDNDrr);
     createReplacer(X86::ANDN64rr, X86::KANDNQrr);
 
     createReplacer(X86::XOR32rr, X86::KXORDrr);
     createReplacer(X86::XOR64rr, X86::KXORQrr);
 
-    createReplacer(X86::TEST32rr, X86::KTESTDrr);
-    createReplacer(X86::TEST64rr, X86::KTESTQrr);
+    // TODO: KTEST is not a replacement for TEST due to flag differences. Need
+    // to prove only Z flag is used.
+    //createReplacer(X86::TEST32rr, X86::KTESTDrr);
+    //createReplacer(X86::TEST64rr, X86::KTESTQrr);
   }
 
   if (STI->hasDQI()) {
     createReplacer(X86::ADD8rr, X86::KADDBrr);
     createReplacer(X86::ADD16rr, X86::KADDWrr);
 
     createReplacer(X86::AND8rr, X86::KANDBrr);
 
     createReplacer(X86::MOV8rm, X86::KMOVBkm);
     createReplacer(X86::MOV8mr, X86::KMOVBmk);
     createReplacer(X86::MOV8rr, X86::KMOVBkk);
 
     createReplacer(X86::NOT8r, X86::KNOTBrr);
 
     createReplacer(X86::OR8rr, X86::KORBrr);
 
     createReplacer(X86::SHR8ri, X86::KSHIFTRBri);
     createReplacer(X86::SHL8ri, X86::KSHIFTLBri);
 
-    createReplacer(X86::TEST8rr, X86::KTESTBrr);
-    createReplacer(X86::TEST16rr, X86::KTESTWrr);
+    // TODO: KTEST is not a replacement for TEST due to flag differences. Need
+    // to prove only Z flag is used.
+    //createReplacer(X86::TEST8rr, X86::KTESTBrr);
+    //createReplacer(X86::TEST16rr, X86::KTESTWrr);
 
     createReplacer(X86::XOR8rr, X86::KXORBrr);
   }
 }
 
 bool X86DomainReassignment::runOnMachineFunction(MachineFunction &MF) {
   if (skipFunction(MF.getFunction()))
     return false;

llvm/trunk/lib/Target/X86/X86ISelLowering.cpp

for (SDNode::use_iterator UI = Op->use_begin(), UE = Op->use_end(); UI != UE;
 
     if (User->getOpcode() != ISD::BRCOND && User->getOpcode() != ISD::SETCC &&
         !(User->getOpcode() == ISD::SELECT && UOpNo == 0))
       return true;
   }
   return false;
 }
 
-// Emit KTEST instruction for bit vectors on AVX-512
-static SDValue EmitKTEST(SDValue Op, SelectionDAG &DAG,
-                         const X86Subtarget &Subtarget) {
-  if (Op.getOpcode() == ISD::BITCAST) {
-    auto hasKTEST = [&](MVT VT) {
-      unsigned SizeInBits = VT.getSizeInBits();
-      return (Subtarget.hasDQI() && (SizeInBits == 8 || SizeInBits == 16)) ||
-        (Subtarget.hasBWI() && (SizeInBits == 32 || SizeInBits == 64));
-    };
-    SDValue Op0 = Op.getOperand(0);
-    MVT Op0VT = Op0.getValueType().getSimpleVT();
-    if (Op0VT.isVector() && Op0VT.getVectorElementType() == MVT::i1 &&
-        hasKTEST(Op0VT))
-      return DAG.getNode(X86ISD::KTEST, SDLoc(Op), Op0VT, Op0, Op0);
-  }
-  return SDValue();
-}
-
 /// Emit nodes that will be selected as "test Op0,Op0", or something
 /// equivalent.
 SDValue X86TargetLowering::EmitTest(SDValue Op, unsigned X86CC, const SDLoc &dl,
                                     SelectionDAG &DAG) const {
   // CF and OF aren't always set the way we want. Determine which
   // of these we need.
   bool NeedCF = false;
   bool NeedOF = false;
case X86::COND_O: case X86::COND_NO: {
     }
     break;
   }
   }
   // See if we can use the EFLAGS value from the operand instead of
   // doing a separate TEST. TEST always sets OF and CF to 0, so unless
   // we prove that the arithmetic won't overflow, we can't use OF or CF.
   if (Op.getResNo() != 0 || NeedOF || NeedCF) {
-    // Emit KTEST for bit vectors
-    if (auto Node = EmitKTEST(Op, DAG, Subtarget))
-      return Node;
     // Emit a CMP with 0, which is the TEST pattern.
     return DAG.getNode(X86ISD::CMP, dl, MVT::i32, Op,
                        DAG.getConstant(0, dl, Op.getValueType()));
   }
   unsigned Opcode = 0;
   unsigned NumOperands = 0;
 
   // Truncate operations may prevent the merge of the SETCC instruction
if (ConvertedOp) {
         SDValue V1 = DAG.getNode(ISD::TRUNCATE, dl, VT, WideVal.getOperand(1));
         SDVTList VTs = DAG.getVTList(Op.getValueType(), MVT::i32);
         Op = DAG.getNode(ConvertedOp, dl, VTs, V0, V1);
       }
     }
   }
 
   if (Opcode == 0) {
-    // Emit KTEST for bit vectors
-    if (auto Node = EmitKTEST(Op, DAG, Subtarget))
-      return Node;
-
     // Emit a CMP with 0, which is the TEST pattern.
     return DAG.getNode(X86ISD::CMP, dl, MVT::i32, Op,
                        DAG.getConstant(0, dl, Op.getValueType()));
   }
   SDVTList VTs = DAG.getVTList(Op.getValueType(), MVT::i32);
   SmallVector<SDValue, 4> Ops(Op->op_begin(), Op->op_begin() + NumOperands);
 
   SDValue New = DAG.getNode(Opcode, dl, VTs, Ops);
#endif
 
   if (Subus)
     Result = DAG.getNode(X86ISD::PCMPEQ, dl, VT, Result,
                          getZeroVector(VT, Subtarget, DAG, dl));
 
   return Result;
 }
 
+// Try to select this as a KTEST+SETCC if possible.
+static SDValue EmitKTEST(SDValue Op0, SDValue Op1, ISD::CondCode CC,
+                         const SDLoc &dl, SelectionDAG &DAG,
+                         const X86Subtarget &Subtarget) {
+  // Only support equality comparisons.
+  if (CC != ISD::SETEQ && CC != ISD::SETNE)
+    return SDValue();
+
+  // Must be a bitcast from vXi1.
+  if (Op0.getOpcode() != ISD::BITCAST)
+    return SDValue();
+
+  Op0 = Op0.getOperand(0);
+  MVT VT = Op0.getSimpleValueType();
+  if (!(Subtarget.hasDQI() && (VT == MVT::v8i1  || VT == MVT::v16i1)) &&
+      !(Subtarget.hasBWI() && (VT == MVT::v32i1 || VT == MVT::v64i1)))
+    return SDValue();
+
+  X86::CondCode X86CC;
+  if (isNullConstant(Op1)) {
+    X86CC = CC == ISD::SETEQ ? X86::COND_E : X86::COND_NE;
+  } else
+    return SDValue();
+
+  SDValue KTEST = DAG.getNode(X86ISD::KTEST, dl, MVT::i32, Op0, Op0);
+  return getSETCC(X86CC, KTEST, dl, DAG);
+}
+
 SDValue X86TargetLowering::LowerSETCC(SDValue Op, SelectionDAG &DAG) const {
 
   MVT VT = Op.getSimpleValueType();
 
   if (VT.isVector()) return LowerVSETCC(Op, Subtarget, DAG);
 
   assert(VT == MVT::i8 && "SetCC type must be 8-bit integer");
   SDValue Op0 = Op.getOperand(0);
SDValue X86TargetLowering::LowerSETCC(SDValue Op, SelectionDAG &DAG) const {
   // Try to use PTEST for a tree ORs equality compared with 0.
   // TODO: We could do AND tree with all 1s as well by using the C flag.
   if (Op0.getOpcode() == ISD::OR && isNullConstant(Op1) &&
       (CC == ISD::SETEQ || CC == ISD::SETNE)) {
     if (SDValue NewSetCC = LowerVectorAllZeroTest(Op0, CC, Subtarget, DAG))
       return NewSetCC;
   }
 
+  // Try to lower using KTEST.
+  if (SDValue NewSetCC = EmitKTEST(Op0, Op1, CC, dl, DAG, Subtarget))
+    return NewSetCC;
+
   // Look for X == 0, X == 1, X != 0, or X != 1.  We can simplify some forms of
   // these.
   if ((isOneConstant(Op1) || isNullConstant(Op1)) &&
       (CC == ISD::SETEQ || CC == ISD::SETNE)) {
 
     // If the input is a setcc, then reuse the input setcc or use a new one with
     // the inverted condition.
     if (Op0.getOpcode() == X86ISD::SETCC) {

llvm/trunk/test/CodeGen/X86/avx512-mask-op.ll

 ; AVX512DQ-NEXT:    ## kill: def $al killed $al killed $eax
 ; AVX512DQ-NEXT:    retq
   %m0 = bitcast i8 %x to <8 x i1>
   %m1 = bitcast i8 %y to <8 x i1>
   %m2 = mul <8 x i1> %m0,  %m1
   %ret = bitcast <8 x i1> %m2 to i8
   ret i8 %ret
 }
+
+; Make sure we don't emit a ktest for signed comparisons.
+define void @ktest_signed(<16 x i32> %x, <16 x i32> %y) {
+; KNL-LABEL: ktest_signed:
+; KNL:       ## %bb.0:
+; KNL-NEXT:    pushq %rax
+; KNL-NEXT:    .cfi_def_cfa_offset 16
+; KNL-NEXT:    vporq %zmm1, %zmm0, %zmm0
+; KNL-NEXT:    vptestnmd %zmm0, %zmm0, %k0
+; KNL-NEXT:    kmovw %k0, %eax
+; KNL-NEXT:    testw %ax, %ax
+; KNL-NEXT:    jle LBB64_1
+; KNL-NEXT:  ## %bb.2: ## %bb.2
+; KNL-NEXT:    popq %rax
+; KNL-NEXT:    vzeroupper
+; KNL-NEXT:    retq
+; KNL-NEXT:  LBB64_1: ## %bb.1
+; KNL-NEXT:    vzeroupper
+; KNL-NEXT:    callq _foo
+; KNL-NEXT:    popq %rax
+; KNL-NEXT:    retq
+;
+; SKX-LABEL: ktest_signed:
+; SKX:       ## %bb.0:
+; SKX-NEXT:    pushq %rax
+; SKX-NEXT:    .cfi_def_cfa_offset 16
+; SKX-NEXT:    vporq %zmm1, %zmm0, %zmm0
+; SKX-NEXT:    vptestnmd %zmm0, %zmm0, %k0
+; SKX-NEXT:    kmovd %k0, %eax
+; SKX-NEXT:    testw %ax, %ax
+; SKX-NEXT:    jle LBB64_1
+; SKX-NEXT:  ## %bb.2: ## %bb.2
+; SKX-NEXT:    popq %rax
+; SKX-NEXT:    vzeroupper
+; SKX-NEXT:    retq
+; SKX-NEXT:  LBB64_1: ## %bb.1
+; SKX-NEXT:    vzeroupper
+; SKX-NEXT:    callq _foo
+; SKX-NEXT:    popq %rax
+; SKX-NEXT:    retq
+;
+; AVX512BW-LABEL: ktest_signed:
+; AVX512BW:       ## %bb.0:
+; AVX512BW-NEXT:    pushq %rax
+; AVX512BW-NEXT:    .cfi_def_cfa_offset 16
+; AVX512BW-NEXT:    vporq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vptestnmd %zmm0, %zmm0, %k0
+; AVX512BW-NEXT:    kmovd %k0, %eax
+; AVX512BW-NEXT:    testw %ax, %ax
+; AVX512BW-NEXT:    jle LBB64_1
+; AVX512BW-NEXT:  ## %bb.2: ## %bb.2
+; AVX512BW-NEXT:    popq %rax
+; AVX512BW-NEXT:    vzeroupper
+; AVX512BW-NEXT:    retq
+; AVX512BW-NEXT:  LBB64_1: ## %bb.1
+; AVX512BW-NEXT:    vzeroupper
+; AVX512BW-NEXT:    callq _foo
+; AVX512BW-NEXT:    popq %rax
+; AVX512BW-NEXT:    retq
+;
+; AVX512DQ-LABEL: ktest_signed:
+; AVX512DQ:       ## %bb.0:
+; AVX512DQ-NEXT:    pushq %rax
+; AVX512DQ-NEXT:    .cfi_def_cfa_offset 16
+; AVX512DQ-NEXT:    vporq %zmm1, %zmm0, %zmm0
+; AVX512DQ-NEXT:    vptestnmd %zmm0, %zmm0, %k0
+; AVX512DQ-NEXT:    kmovw %k0, %eax
+; AVX512DQ-NEXT:    testw %ax, %ax
+; AVX512DQ-NEXT:    jle LBB64_1
+; AVX512DQ-NEXT:  ## %bb.2: ## %bb.2
+; AVX512DQ-NEXT:    popq %rax
+; AVX512DQ-NEXT:    vzeroupper
+; AVX512DQ-NEXT:    retq
+; AVX512DQ-NEXT:  LBB64_1: ## %bb.1
+; AVX512DQ-NEXT:    vzeroupper
+; AVX512DQ-NEXT:    callq _foo
+; AVX512DQ-NEXT:    popq %rax
+; AVX512DQ-NEXT:    retq
+  %a = icmp eq <16 x i32> %x, zeroinitializer
+  %b = icmp eq <16 x i32> %y, zeroinitializer
+  %c = and <16 x i1> %a, %b
+  %d = bitcast <16 x i1> %c to i16
+  %e = icmp sgt i16 %d, 0
+  br i1 %e, label %bb.2, label %bb.1
+bb.1:
+  call void @foo()
+  br label %bb.2
+bb.2:
+  ret void
+}
+declare void @foo()
+

llvm/trunk/test/CodeGen/X86/domain-reassignment.mir

frameInfo:
   savePoint:       ''
   restorePoint:    ''
 fixedStack:
 stack:
 constants:
 body:             |
   ; CHECK-LABEL: name: test_8bitops
   ; CHECK: bb.0:
-  ; CHECK:   successors: %bb.1(0x40000000), %bb.2(0x40000000)
+  ; CHECK:   successors: %bb.1(0x80000000)
   ; CHECK:   liveins: $rdi, $zmm0, $zmm1, $zmm2, $zmm3
   ; CHECK:   [[COPY:%[0-9]+]]:gr64 = COPY $rdi
   ; CHECK:   [[COPY1:%[0-9]+]]:vr512 = COPY $zmm0
   ; CHECK:   [[COPY2:%[0-9]+]]:vr512 = COPY $zmm1
   ; CHECK:   [[COPY3:%[0-9]+]]:vr512 = COPY $zmm2
   ; CHECK:   [[COPY4:%[0-9]+]]:vr512 = COPY $zmm3
   ; CHECK:   [[VCMPPDZrri:%[0-9]+]]:vk8 = VCMPPDZrri [[COPY3]], [[COPY4]], 0
   ; CHECK:   [[COPY5:%[0-9]+]]:vk32 = COPY [[VCMPPDZrri]]
   ; CHECK:   [[COPY6:%[0-9]+]]:vk8 = COPY [[COPY5]]
   ; CHECK:   [[KSHIFTRBri:%[0-9]+]]:vk8 = KSHIFTRBri [[COPY6]], 2
   ; CHECK:   [[KSHIFTLBri:%[0-9]+]]:vk8 = KSHIFTLBri [[KSHIFTRBri]], 1
   ; CHECK:   [[KNOTBrr:%[0-9]+]]:vk8 = KNOTBrr [[KSHIFTLBri]]
   ; CHECK:   [[KORBrr:%[0-9]+]]:vk8 = KORBrr [[KNOTBrr]], [[KSHIFTRBri]]
   ; CHECK:   [[KANDBrr:%[0-9]+]]:vk8 = KANDBrr [[KORBrr]], [[KSHIFTLBri]]
   ; CHECK:   [[KXORBrr:%[0-9]+]]:vk8 = KXORBrr [[KANDBrr]], [[KSHIFTRBri]]
   ; CHECK:   [[KADDBrr:%[0-9]+]]:vk8 = KADDBrr [[KXORBrr]], [[KNOTBrr]]
   ; CHECK:   [[COPY7:%[0-9]+]]:vk32 = COPY [[KADDBrr]]
   ; CHECK:   [[COPY8:%[0-9]+]]:vk8wm = COPY [[COPY7]]
   ; CHECK:   [[VMOVAPDZrrk:%[0-9]+]]:vr512 = VMOVAPDZrrk [[COPY2]], killed [[COPY8]], [[COPY1]]
   ; CHECK:   VMOVAPDZmr [[COPY]], 1, $noreg, 0, $noreg, killed [[VMOVAPDZrrk]]
-  ; CHECK:   KTESTBrr [[KADDBrr]], [[KADDBrr]], implicit-def $eflags
-  ; CHECK:   JE_1 %bb.1, implicit $eflags
-  ; CHECK:   JMP_1 %bb.2
   ; CHECK: bb.1:
   ; CHECK:   successors: %bb.2(0x80000000)
   ; CHECK: bb.2:
   ; CHECK:   RET 0
   bb.0:
     liveins: $rdi, $zmm0, $zmm1, $zmm2, $zmm3
 
     %0 = COPY $rdi
bb.0:
     %18 = ADD8rr %17, %14, implicit-def dead $eflags
 
     %8 = IMPLICIT_DEF
     %9 = INSERT_SUBREG %8, %18, 1
     %10 = COPY %9
     %11 = VMOVAPDZrrk %2, killed %10, %1
     VMOVAPDZmr %0, 1, $noreg, 0, $noreg, killed %11
 
-    TEST8rr %18, %18, implicit-def $eflags
-    JE_1 %bb.1, implicit $eflags
-    JMP_1 %bb.2
+    ; FIXME We can't replace TEST with KTEST due to flag differences
+    ; TEST8rr %18, %18, implicit-def $eflags
+    ; JE_1 %bb.1, implicit $eflags
+    ; JMP_1 %bb.2
 
   bb.1:
 
   bb.2:
     RET 0
 
 ...
 ---
frameInfo:
   savePoint:       ''
   restorePoint:    ''
 fixedStack:
 stack:
 constants:
 body:             |
   ; CHECK-LABEL: name: test_16bitops
   ; CHECK: bb.0:
-  ; CHECK:   successors: %bb.1(0x40000000), %bb.2(0x40000000)
+  ; CHECK:   successors: %bb.1(0x80000000)
   ; CHECK:   liveins: $rdi, $zmm0, $zmm1, $zmm2, $zmm3
   ; CHECK:   [[COPY:%[0-9]+]]:gr64 = COPY $rdi
   ; CHECK:   [[COPY1:%[0-9]+]]:vr512 = COPY $zmm0
   ; CHECK:   [[COPY2:%[0-9]+]]:vr512 = COPY $zmm1
   ; CHECK:   [[COPY3:%[0-9]+]]:vr512 = COPY $zmm2
   ; CHECK:   [[COPY4:%[0-9]+]]:vr512 = COPY $zmm3
   ; CHECK:   [[VCMPPSZrri:%[0-9]+]]:vk16 = VCMPPSZrri [[COPY3]], [[COPY4]], 0
   ; CHECK:   [[COPY5:%[0-9]+]]:vk32 = COPY [[VCMPPSZrri]]
   ; CHECK:   [[COPY6:%[0-9]+]]:vk16 = COPY [[COPY5]]
   ; CHECK:   [[KSHIFTRWri:%[0-9]+]]:vk16 = KSHIFTRWri [[COPY6]], 2
   ; CHECK:   [[KSHIFTLWri:%[0-9]+]]:vk16 = KSHIFTLWri [[KSHIFTRWri]], 1
   ; CHECK:   [[KNOTWrr:%[0-9]+]]:vk16 = KNOTWrr [[KSHIFTLWri]]
   ; CHECK:   [[KORWrr:%[0-9]+]]:vk16 = KORWrr [[KNOTWrr]], [[KSHIFTRWri]]
   ; CHECK:   [[KANDWrr:%[0-9]+]]:vk16 = KANDWrr [[KORWrr]], [[KSHIFTLWri]]
   ; CHECK:   [[KXORWrr:%[0-9]+]]:vk16 = KXORWrr [[KANDWrr]], [[KSHIFTRWri]]
   ; CHECK:   [[COPY7:%[0-9]+]]:vk32 = COPY [[KXORWrr]]
   ; CHECK:   [[COPY8:%[0-9]+]]:vk16wm = COPY [[COPY7]]
   ; CHECK:   [[VMOVAPSZrrk:%[0-9]+]]:vr512 = VMOVAPSZrrk [[COPY2]], killed [[COPY8]], [[COPY1]]
   ; CHECK:   VMOVAPSZmr [[COPY]], 1, $noreg, 0, $noreg, killed [[VMOVAPSZrrk]]
-  ; CHECK:   KTESTWrr [[KXORWrr]], [[KXORWrr]], implicit-def $eflags
-  ; CHECK:   JE_1 %bb.1, implicit $eflags
-  ; CHECK:   JMP_1 %bb.2
   ; CHECK: bb.1:
   ; CHECK:   successors: %bb.2(0x80000000)
   ; CHECK: bb.2:
   ; CHECK:   RET 0
   bb.0:
     liveins: $rdi, $zmm0, $zmm1, $zmm2, $zmm3
 
     %0 = COPY $rdi
bb.0:
     %17 = XOR16rr %16, %12, implicit-def dead $eflags
 
     %8 = IMPLICIT_DEF
     %9 = INSERT_SUBREG %8, %17, 3
     %10 = COPY %9
     %11 = VMOVAPSZrrk %2, killed %10, %1
     VMOVAPSZmr %0, 1, $noreg, 0, $noreg, killed %11
 
-    TEST16rr %17, %17, implicit-def $eflags
-    JE_1 %bb.1, implicit $eflags
-    JMP_1 %bb.2
+    ; FIXME We can't replace TEST with KTEST due to flag differences
+    ; FIXME TEST16rr %17, %17, implicit-def $eflags
+    ; FIXME JE_1 %bb.1, implicit $eflags
+    ; FIXME JMP_1 %bb.2
 
   bb.1:
 
   bb.2:
     RET 0
 
 ...
 ---
frameInfo:
   savePoint:       ''
   restorePoint:    ''
 fixedStack:
 stack:
 constants:
 body:             |
   ; CHECK-LABEL: name: test_32bitops
   ; CHECK: bb.0:
-  ; CHECK:   successors: %bb.1(0x40000000), %bb.2(0x40000000)
+  ; CHECK:   successors: %bb.1(0x80000000)
   ; CHECK:   liveins: $rdi, $zmm0, $zmm1
   ; CHECK:   [[COPY:%[0-9]+]]:gr64 = COPY $rdi
   ; CHECK:   [[COPY1:%[0-9]+]]:vr512 = COPY $zmm0
   ; CHECK:   [[COPY2:%[0-9]+]]:vr512 = COPY $zmm1
   ; CHECK:   [[KMOVDkm:%[0-9]+]]:vk32 = KMOVDkm [[COPY]], 1, $noreg, 0, $noreg
   ; CHECK:   [[KSHIFTRDri:%[0-9]+]]:vk32 = KSHIFTRDri [[KMOVDkm]], 2
   ; CHECK:   [[KSHIFTLDri:%[0-9]+]]:vk32 = KSHIFTLDri [[KSHIFTRDri]], 1
   ; CHECK:   [[KNOTDrr:%[0-9]+]]:vk32 = KNOTDrr [[KSHIFTLDri]]
   ; CHECK:   [[KORDrr:%[0-9]+]]:vk32 = KORDrr [[KNOTDrr]], [[KSHIFTRDri]]
   ; CHECK:   [[KANDDrr:%[0-9]+]]:vk32 = KANDDrr [[KORDrr]], [[KSHIFTLDri]]
   ; CHECK:   [[KXORDrr:%[0-9]+]]:vk32 = KXORDrr [[KANDDrr]], [[KSHIFTRDri]]
   ; CHECK:   [[KANDNDrr:%[0-9]+]]:vk32 = KANDNDrr [[KXORDrr]], [[KORDrr]]
   ; CHECK:   [[KADDDrr:%[0-9]+]]:vk32 = KADDDrr [[KANDNDrr]], [[KXORDrr]]
   ; CHECK:   [[COPY3:%[0-9]+]]:vk32wm = COPY [[KADDDrr]]
   ; CHECK:   [[VMOVDQU16Zrrk:%[0-9]+]]:vr512 = VMOVDQU16Zrrk [[COPY2]], killed [[COPY3]], [[COPY1]]
   ; CHECK:   VMOVDQA32Zmr [[COPY]], 1, $noreg, 0, $noreg, killed [[VMOVDQU16Zrrk]]
-  ; CHECK:   KTESTDrr [[KADDDrr]], [[KADDDrr]], implicit-def $eflags
-  ; CHECK:   JE_1 %bb.1, implicit $eflags
-  ; CHECK:   JMP_1 %bb.2
   ; CHECK: bb.1:
   ; CHECK:   successors: %bb.2(0x80000000)
   ; CHECK: bb.2:
   ; CHECK:   RET 0
   bb.0:
     liveins: $rdi, $zmm0, $zmm1
 
     %0 = COPY $rdi
bb.0:
     %11 = XOR32rr %10, %6, implicit-def dead $eflags
     %12 = ANDN32rr %11, %9, implicit-def dead $eflags
     %13 = ADD32rr %12, %11, implicit-def dead $eflags
 
     %3 = COPY %13
     %4 = VMOVDQU16Zrrk %2, killed %3, %1
     VMOVDQA32Zmr %0, 1, $noreg, 0, $noreg, killed %4
 
-    TEST32rr %13, %13, implicit-def $eflags
-    JE_1 %bb.1, implicit $eflags
-    JMP_1 %bb.2
+    ; FIXME We can't replace TEST with KTEST due to flag differences
+    ; FIXME TEST32rr %13, %13, implicit-def $eflags
+    ; FIXME JE_1 %bb.1, implicit $eflags
+    ; FIXME JMP_1 %bb.2
 
   bb.1:
 
   bb.2:
     RET 0
 
 ...
 ---
frameInfo:
   savePoint:       ''
   restorePoint:    ''
 fixedStack:
 stack:
 constants:
 body:             |
   ; CHECK-LABEL: name: test_64bitops
   ; CHECK: bb.0:
-  ; CHECK:   successors: %bb.1(0x40000000), %bb.2(0x40000000)
+  ; CHECK:   successors: %bb.1(0x80000000)
   ; CHECK:   liveins: $rdi, $zmm0, $zmm1
   ; CHECK:   [[COPY:%[0-9]+]]:gr64 = COPY $rdi
   ; CHECK:   [[COPY1:%[0-9]+]]:vr512 = COPY $zmm0
   ; CHECK:   [[COPY2:%[0-9]+]]:vr512 = COPY $zmm1
   ; CHECK:   [[KMOVQkm:%[0-9]+]]:vk64 = KMOVQkm [[COPY]], 1, $noreg, 0, $noreg
   ; CHECK:   [[KSHIFTRQri:%[0-9]+]]:vk64 = KSHIFTRQri [[KMOVQkm]], 2
   ; CHECK:   [[KSHIFTLQri:%[0-9]+]]:vk64 = KSHIFTLQri [[KSHIFTRQri]], 1
   ; CHECK:   [[KNOTQrr:%[0-9]+]]:vk64 = KNOTQrr [[KSHIFTLQri]]
   ; CHECK:   [[KORQrr:%[0-9]+]]:vk64 = KORQrr [[KNOTQrr]], [[KSHIFTRQri]]
   ; CHECK:   [[KANDQrr:%[0-9]+]]:vk64 = KANDQrr [[KORQrr]], [[KSHIFTLQri]]
   ; CHECK:   [[KXORQrr:%[0-9]+]]:vk64 = KXORQrr [[KANDQrr]], [[KSHIFTRQri]]
   ; CHECK:   [[KANDNQrr:%[0-9]+]]:vk64 = KANDNQrr [[KXORQrr]], [[KORQrr]]
   ; CHECK:   [[KADDQrr:%[0-9]+]]:vk64 = KADDQrr [[KANDNQrr]], [[KXORQrr]]
   ; CHECK:   [[COPY3:%[0-9]+]]:vk64wm = COPY [[KADDQrr]]
   ; CHECK:   [[VMOVDQU8Zrrk:%[0-9]+]]:vr512 = VMOVDQU8Zrrk [[COPY2]], killed [[COPY3]], [[COPY1]]
   ; CHECK:   VMOVDQA32Zmr [[COPY]], 1, $noreg, 0, $noreg, killed [[VMOVDQU8Zrrk]]
-  ; CHECK:   KTESTQrr [[KADDQrr]], [[KADDQrr]], implicit-def $eflags
-  ; CHECK:   JE_1 %bb.1, implicit $eflags
-  ; CHECK:   JMP_1 %bb.2
   ; CHECK: bb.1:
   ; CHECK:   successors: %bb.2(0x80000000)
   ; CHECK: bb.2:
   ; CHECK:   RET 0
   bb.0:
     liveins: $rdi, $zmm0, $zmm1
 
     %0 = COPY $rdi
bb.0:
     %11 = XOR64rr %10, %6, implicit-def dead $eflags
     %12 = ANDN64rr %11, %9, implicit-def dead $eflags
     %13 = ADD64rr %12, %11, implicit-def dead $eflags
 
     %3 = COPY %13
     %4 = VMOVDQU8Zrrk %2, killed %3, %1
     VMOVDQA32Zmr %0, 1, $noreg, 0, $noreg, killed %4
 
-    TEST64rr %13, %13, implicit-def $eflags
-    JE_1 %bb.1, implicit $eflags
-    JMP_1 %bb.2
+    ; FIXME We can't replace TEST with KTEST due to flag differences
+    ; FIXME TEST64rr %13, %13, implicit-def $eflags
+    ; FIXME JE_1 %bb.1, implicit $eflags
+    ; FIXME JMP_1 %bb.2
 
   bb.1:
 
   bb.2:
     RET 0
 
 ...
 ---