[SelectionDAG][WIP] Add support for evaluating SetCC based on knownbits

[goldstein.w.n]

Summary

In some cases folds done through ISel/DAGCombining result in SetCC
conditions that we can easily simplify using known bits.

This patch adds support for that.

NB: The medium-term goal is to get this patch in, then use comparisons
against zero for testing future improvements to isKnownNeverZero as
there are issues using cttz/ctlz for that.

[goldstein.w.n]

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

615

Why do we emit any code here?

llvm/test/CodeGen/X86/fold-rmw-ops.ll

1359

These movb $1, %al; testb %al, %al's (here and in many other cases) are unnecessary. I assume its because SelectionDAG only has BB view, so even if we can rule out some BBs (based on known true/false br-cond), there is no pass for that. Is there anything we can/should do about that?

Also NB, we really should never emit movb $1, %al; testb %al, %al just grabbing any gpr (I guess least recently used to minimize potential latency) and do cmpb %gpr8, gpr8 then jne/je depending if we want it to be always true/false.

[goldstein.w.n]

Fix some typos

[craig.topper]

llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp

2472 ↗

(On Diff #517720)

I would not do this in FoldSetCC. FoldSetCC is called from getNode and I don't think we want to call to computeKnownBits hidden in that.

You you can do it in SimplifySetCC in TargetLowering.cpp I think.

[craig.topper]

llvm/test/CodeGen/X86/fold-rmw-ops.ll

1359

Hopefully most of these are just tests that should have been folded by InstCombine or other passes earlier and not really cases that originate in SelectionDAG.

[craig.topper]

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

615

I'm guesting we started with a conditional branch and it got optimized out?

[goldstein.w.n]

llvm/test/CodeGen/X86/fold-rmw-ops.ll

1359

Do you think these regressions are something to worry about? Or acceptable as cases we would never expect to get from the middle-end.

[RKSimon]

This is very similar to a patch I tried a while ago: D86578 - one of the blockers for which was all the over-reduced test cases that it broke

[goldstein.w.n]

In D149383#4305324, @RKSimon wrote:

This is very similar to a patch I tried a while ago: D86578 - one of the blockers for which was all the over-reduced test cases that it broke

What do you mean over-reduces?
But nikic has a point that realistic backend code shouldn't be affected by this much.
I think there are some cases that benefit like cttz/ctlz cases if zero is no poison AND target doesn't have tzcnt/lzcnt
but thats niche.

It may also be able to help with intermediate select statement introduced during lowering that are
actually constant foldable.

Maybe a different way to test isKnownNeverZero is better?

[RKSimon]

llvm/test/CodeGen/AArch64/cmp-const-max.ll

1 ↗

(On Diff #517720)

regenerate + commit this file and rebase to show the diffs from the patch

llvm/test/CodeGen/ARM/sub-cmp-peephole.ll

1 ↗

(On Diff #517720)

regenerate this file first so you show any diffs

llvm/test/CodeGen/X86/fold-rmw-ops.ll

1372

comparing against zero in all these or-with-imm tests just seems to be a copy+paste from the other logic ops in this file - maybe change it to something that isn't constant foldable (test for -ve?)

[goldstein.w.n]

Rebase + move impl to simplifysetcc

[goldstein.w.n]

llvm/test/CodeGen/X86/fold-rmw-ops.ll

1372

re: 'test for -ve?' hmm?

But would generally prefer to add new tests than modify existing.

[RKSimon]

llvm/test/CodeGen/X86/fold-rmw-ops.ll

1372

'test for -ve' === 'test for negative'

Adding additional tests would be fine.

[goldstein.w.n]

Rebase

[foad]

llvm/lib/CodeGen/SelectionDAG/TargetLowering.cpp

4281–4282

Just curious: does "strengthening" conditions like this actually generate better code? Is it something we do elsewhere in the compiler?

[RKSimon]

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

4708

Whats going on here?

[RKSimon]

llvm/lib/CodeGen/SelectionDAG/TargetLowering.cpp

4295

I'm not sure this is always false?

[goldstein.w.n]

llvm/lib/CodeGen/SelectionDAG/TargetLowering.cpp

4281–4282

No worse in fact. This was really motivated by wanting to improve the knownbits analysis in selectiondag but having no good way to test it.

I think at the moment, since we don't have the infrastructure to fold basic blocks when icmp; br conditions are known true/false we end up with regressions.

Not sure exactly where this is going to go unless we add that infrastructure.

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

4708

Its what you thought was a bad merge in the knownbits impl.

if (computeKnownBits(Op.getOperand(0)).One[0], Depth + 1)
  return true;

See the bug?

Should be:

if (computeKnownBits(Op.getOperand(0), Depth + 1).One[0])
  return true;

Suprised the former is accepted as an expression (not a clang warning/error). Its

if (expr_A, expr_B)

llvm/test/CodeGen/X86/fold-rmw-ops.ll

1372

As in add a new test file? This isn't a new file for the series, its just affected by the change.

[goldstein.w.n]

Rebase (fixed bug in KnownBitsIsNeverZero)

[goldstein.w.n]

rebase

[RKSimon]

There's still a lot of tests in here that need adjusting so that they still test what we want them to test

llvm/test/CodeGen/X86/fold-rmw-ops.ll

1372

I'd prefer that these tests were adjusted, the icmp_eq vs 0 was just a dumb copy + paste - but if you don't want to do that, duplicating these OR tests immediately below with a icmp_sgt 0 would be OK

llvm/test/CodeGen/X86/hoist-and-by-const-from-shl-in-eqcmp-zero.ll

801

this needs adjusting

llvm/test/CodeGen/X86/or-with-overflow.ll

6

All these tests need adjusting

llvm/test/CodeGen/X86/pr16031.ll

9

not sure what to do with this test - either we try to fix it so it still matches what the original bug was about, or we delete it

llvm/test/CodeGen/X86/shrink-compare.ll

128

This is no longer a shrink-compare test

[foad]

llvm/lib/CodeGen/SelectionDAG/TargetLowering.cpp

4281–4282

You're adding code to the compiler to change these setcc conditions, but not actually make anything better? I do not think that is a good idea.

4295

It's confusing but I think it's correct. From this point onwards, Res represents whether LHS and RHS are equal, irrespective of Cond.

[goldstein.w.n]

llvm/lib/CodeGen/SelectionDAG/TargetLowering.cpp

4281–4282

So worse might be a bit of a exageration. It's worse for some IR that the middle-end would never pass to us.

I was thinking maybe we could do this for SETCC not used by br. That case we see improvement. What I need to do I think though, do a little work seeing if any of the cases this changes are cases generated by the backend (i.e cases that wouldn't normally be cleaned up by the middle-end).

But I agree, in the current state its not submitable.

4295

I'll make it a new variable.

llvm/test/CodeGen/X86/fold-rmw-ops.ll

1372

Ah, okay. Will do.

[goldstein.w.n]

Abandoning this revision. Still doesn't seem feasible to add this in the backend.

llvm/lib/CodeGen/SelectionDAG/TargetLowering.cpp

SDValue TargetLowering::SimplifySetCC(EVT VT, SDValue N0, SDValue N1,
   // instruction on some targets.
   if (!N0ConstOrSplat && !N1ConstOrSplat &&
       (DCI.isBeforeLegalizeOps() ||
        isCondCodeLegal(SwappedCC, N0.getSimpleValueType())) &&
       DAG.doesNodeExist(ISD::SUB, DAG.getVTList(OpVT), {N1, N0}) &&
       !DAG.doesNodeExist(ISD::SUB, DAG.getVTList(OpVT), {N0, N1}))
     return DAG.getSetCC(dl, VT, N1, N0, SwappedCC);
 
+  // Try to constant fold SetCC.
+  if (OpVT.isInteger()) {
+    KnownBits KnownRHS = DAG.computeKnownBits(N1);
+    if (!KnownRHS.isUnknown()) {
+      KnownBits KnownLHS = DAG.computeKnownBits(N0);
+      std::optional<bool> Res;
+      // Check if we can constant fold this with knownbits.
+      switch (Cond) {
+      case ISD::SETEQ:
+        Res = KnownBits::eq(KnownLHS, KnownRHS);
+        break;
+      case ISD::SETNE:
+        Res = KnownBits::ne(KnownLHS, KnownRHS);
+        break;
+      case ISD::SETLT:
+        Res = KnownBits::slt(KnownLHS, KnownRHS);
+        break;
+      case ISD::SETULT:
+        Res = KnownBits::ult(KnownLHS, KnownRHS);
+        break;
+      case ISD::SETGT:
+        Res = KnownBits::sgt(KnownLHS, KnownRHS);
+        break;
+      case ISD::SETUGT:
+        Res = KnownBits::ugt(KnownLHS, KnownRHS);
+        break;
+      case ISD::SETLE:
+        Res = KnownBits::sle(KnownLHS, KnownRHS);
+        break;
+      case ISD::SETULE:
+        Res = KnownBits::ule(KnownLHS, KnownRHS);
+        break;
+      case ISD::SETGE:
+        Res = KnownBits::sge(KnownLHS, KnownRHS);
+        break;
+      case ISD::SETUGE:
+        Res = KnownBits::uge(KnownLHS, KnownRHS);
+        break;
+      default:
+        break;
+      }
+
+      if (Res)
+        return DAG.getBoolConstant(*Res, dl, VT, OpVT);
+
+      // We aren't able to constant fold with known bits but can either 1) make
+      // conditions stronger (i.e ule -> ult) or 2) simplify with
+      // isKnownNeverZero if RHS is zero.
+      switch (Cond) {
+      case ISD::SETLE:
+      case ISD::SETULE:
+      case ISD::SETGE:
+      case ISD::SETUGE:
+        Res = KnownBits::eq(KnownLHS, KnownRHS);
+        [[fallthrough]];
+      case ISD::SETEQ:
+      case ISD::SETNE:
+        // isKnownNeverZero is able to prove cases computeKnownBits can't.
+        if (!Res && KnownRHS.isZero() && DAG.isKnownNeverZero(N0))
+          Res = false;
+        break;
+      default:
+        break;
+      }
+
+      if (Res) {
+        assert(*Res == false &&
+               "There is a bug in KnownBits::{sge,uge,sle,ule}");
+        ISD::CondCode NewCond = Cond;
+        // NB: We could remove this switch and just do `Cond ^ ISD::SETEQ` for
+        // the new opcode.
+        switch (Cond) {
+          // Remove the or eq portion of the condition.
+        case ISD::SETULE:
+          NewCond = ISD::SETULT;
+          break;
+        case ISD::SETLE:
+          NewCond = ISD::SETLT;
+          break;
+        case ISD::SETUGE:
+          NewCond = ISD::SETUGT;
+          break;
+        case ISD::SETGE:
+          NewCond = ISD::SETGT;
+          break;
+          // Evaluate to true/false.
+        case ISD::SETNE:
+          return DAG.getBoolConstant(true, dl, VT, OpVT);
+        case ISD::SETEQ:
+          return DAG.getBoolConstant(false, dl, VT, OpVT);
+        default:
+          break;
+        }
+        if (Cond != NewCond)
+          return DAG.getSetCC(dl, VT, N0, N1, NewCond);
+      }
+    }
+  }
+
   if (SDValue V = foldSetCCWithRotate(VT, N0, N1, Cond, dl, DAG))
     return V;
 
   if (SDValue V = foldSetCCWithFunnelShift(VT, N0, N1, Cond, dl, DAG))
     return V;
 
   if (auto *N1C = isConstOrConstSplat(N1)) {
     const APInt &C1 = N1C->getAPIntValue();

llvm/test/CodeGen/AArch64/aarch64-split-and-bitmask-immediate.ll

entry:
   %conv = zext i1 %cmp to i8
   ret i8 %conv
 }
 
 ; This constant should not be split because it can be handled by one mov.
 define i8 @test2(i32 %a) {
 ; CHECK-LABEL: test2:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    mov w8, #135
-; CHECK-NEXT:    and w8, w0, w8
-; CHECK-NEXT:    cmp w8, #1024
-; CHECK-NEXT:    cset w0, eq
+; CHECK-NEXT:    mov w0, wzr
 ; CHECK-NEXT:    ret
 entry:
   %and = and i32 %a, 135
   %cmp = icmp eq i32 %and, 1024
   %conv = zext i1 %cmp to i8
   ret i8 %conv
 }
 
 ; This constant should not be split because the split immediate is not valid
 ; bitmask immediate.
 define i8 @test3(i32 %a) {
 ; CHECK-LABEL: test3:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    mov w8, #1024
+; CHECK-NEXT:    mov w8, #1024 // =0x400
 ; CHECK-NEXT:    movk w8, #33, lsl #16
 ; CHECK-NEXT:    and w8, w0, w8
 ; CHECK-NEXT:    cmp w8, #1024
 ; CHECK-NEXT:    cset w0, eq
 ; CHECK-NEXT:    ret
 entry:
   %and = and i32 %a, 2163712
   %cmp = icmp eq i32 %and, 1024
entry:
   %cmp = icmp eq i64 %and, 1024
   %conv = zext i1 %cmp to i8
   ret i8 %conv
 }
 
 define i8 @test5(i64 %a) {
 ; CHECK-LABEL: test5:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    and x8, x0, #0x3ffffc000
-; CHECK-NEXT:    and x8, x8, #0xfffffffe00007fff
-; CHECK-NEXT:    cmp x8, #1024
-; CHECK-NEXT:    cset w0, eq
+; CHECK-NEXT:    mov w0, wzr
 ; CHECK-NEXT:    ret
 entry:
   %and = and i64 %a, 8589950976
   %cmp = icmp eq i64 %and, 1024
   %conv = zext i1 %cmp to i8
   ret i8 %conv
 }
 
 ; This constant should not be split because it can be handled by one mov.
 define i8 @test6(i64 %a) {
 ; CHECK-LABEL: test6:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    mov w8, #135
-; CHECK-NEXT:    and x8, x0, x8
-; CHECK-NEXT:    cmp x8, #1024
-; CHECK-NEXT:    cset w0, eq
+; CHECK-NEXT:    mov w0, wzr
 ; CHECK-NEXT:    ret
 entry:
   %and = and i64 %a, 135
   %cmp = icmp eq i64 %and, 1024
   %conv = zext i1 %cmp to i8
   ret i8 %conv
 }
 
 ; This constant should not be split because the split immediate is not valid
 ; bitmask immediate.
 define i8 @test7(i64 %a) {
 ; CHECK-LABEL: test7:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    mov w8, #1024
+; CHECK-NEXT:    mov w8, #1024 // =0x400
 ; CHECK-NEXT:    movk w8, #33, lsl #16
 ; CHECK-NEXT:    and x8, x0, x8
 ; CHECK-NEXT:    cmp x8, #1024
 ; CHECK-NEXT:    cset w0, eq
 ; CHECK-NEXT:    ret
 entry:
   %and = and i64 %a, 2163712
   %cmp = icmp eq i64 %and, 1024

 
 ; This constant should not be split because the `and` is not loop invariant.
 define i32 @test9(ptr nocapture %x, ptr nocapture readonly %y, i32 %n) {
 ; CHECK-LABEL: test9:
 ; CHECK:       // %bb.0: // %entry
 ; CHECK-NEXT:    cmp w2, #1
 ; CHECK-NEXT:    b.lt .LBB8_3
 ; CHECK-NEXT:  // %bb.1: // %for.body.preheader
-; CHECK-NEXT:    mov w9, #1024
+; CHECK-NEXT:    mov w9, #1024 // =0x400
 ; CHECK-NEXT:    mov w8, w2
 ; CHECK-NEXT:    movk w9, #32, lsl #16
 ; CHECK-NEXT:  .LBB8_2: // %for.body
 ; CHECK-NEXT:    // =>This Inner Loop Header: Depth=1
 ; CHECK-NEXT:    ldr w10, [x1], #4
 ; CHECK-NEXT:    subs x8, x8, #1
 ; CHECK-NEXT:    and w10, w10, w9
 ; CHECK-NEXT:    str w10, [x0], #4

 ; %7:gpr32 = ORRWrr killed %6:gpr32, %4:gpr32
 ;
 ; In this case, the constant should not be split because it causes more
 ; instructions.
 define void @test10(ptr nocapture %x, ptr nocapture readonly %y, ptr nocapture %z) {
 ; CHECK-LABEL: test10:
 ; CHECK:       // %bb.0: // %entry
 ; CHECK-NEXT:    ldr w8, [x1]
-; CHECK-NEXT:    mov w9, #1024
+; CHECK-NEXT:    mov w9, #1024 // =0x400
 ; CHECK-NEXT:    movk w9, #32, lsl #16
 ; CHECK-NEXT:    and w8, w8, w9
 ; CHECK-NEXT:    str w8, [x0]
 ; CHECK-NEXT:    ldr w8, [x1]
 ; CHECK-NEXT:    orr w8, w8, w9
 ; CHECK-NEXT:    str w8, [x2]
 ; CHECK-NEXT:    ret
 entry:

 ;
 ; MOVi32imm -1610612736
 ; SUBREG_TO_REG
 ;
 ; The constant should be zero-extended to 64 bit and it should not be split.
 define i8 @test11(i64 %a) {
 ; CHECK-LABEL: test11:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    mov w8, #-1610612736
-; CHECK-NEXT:    and x8, x0, x8
-; CHECK-NEXT:    cmp x8, #1024
-; CHECK-NEXT:    cset w0, eq
+; CHECK-NEXT:    mov w0, wzr
 ; CHECK-NEXT:    ret
 entry:
   %and = and i64 %a, 2684354560
   %cmp = icmp eq i64 %and, 1024
   %conv = zext i1 %cmp to i8
   ret i8 %conv
 }

llvm/test/CodeGen/AArch64/andcompare.ll

 ; SDISEL-LABEL: cmp_to_ands3:
 ; SDISEL:       // %bb.0:
 ; SDISEL-NEXT:    tst w0, #0x10
 ; SDISEL-NEXT:    csel w0, w1, wzr, ne
 ; SDISEL-NEXT:    ret
 ;
 ; GISEL-LABEL: cmp_to_ands3:
 ; GISEL:       // %bb.0:
-; GISEL-NEXT:    mov w8, #23
+; GISEL-NEXT:    mov w8, #23 // =0x17
 ; GISEL-NEXT:    and w8, w0, w8
 ; GISEL-NEXT:    cmp w8, #7
 ; GISEL-NEXT:    csel w0, w1, wzr, hi
 ; GISEL-NEXT:    ret
   %and = and i32 %num, 23
   %cmp = icmp ugt i32 %and, 7
   %r = select i1 %cmp, i32 %a, i32 0
   ret i32 %r
 }
 
 define i32 @cmp_to_ands4(i32 %num, i32 %a) {
 ; SDISEL-LABEL: cmp_to_ands4:
 ; SDISEL:       // %bb.0:
 ; SDISEL-NEXT:    and w8, w0, #0x30
-; SDISEL-NEXT:    tst w0, #0x20
-; SDISEL-NEXT:    csel w0, w8, w1, eq
+; SDISEL-NEXT:    cmp w8, #31
+; SDISEL-NEXT:    csel w0, w8, w1, lo
 ; SDISEL-NEXT:    ret
 ;
 ; GISEL-LABEL: cmp_to_ands4:
 ; GISEL:       // %bb.0:
 ; GISEL-NEXT:    and w8, w0, #0x30
 ; GISEL-NEXT:    cmp w8, #31
 ; GISEL-NEXT:    csel w0, w8, w1, ls
 ; GISEL-NEXT:    ret

llvm/test/CodeGen/AArch64/hoist-and-by-const-from-shl-in-eqcmp-zero.ll

 
 ;------------------------------------------------------------------------------;
 ; What if X is a constant too?
 ;------------------------------------------------------------------------------;
 
 define i1 @scalar_i32_x_is_const_eq(i32 %y) nounwind {
 ; CHECK-LABEL: scalar_i32_x_is_const_eq:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    mov w8, #43605
+; CHECK-NEXT:    mov w8, #43605 // =0xaa55
 ; CHECK-NEXT:    movk w8, #43605, lsl #16
 ; CHECK-NEXT:    lsl w8, w8, w0
 ; CHECK-NEXT:    tst w8, #0x1
 ; CHECK-NEXT:    cset w0, eq
 ; CHECK-NEXT:    ret
   %t0 = shl i32 2857740885, %y
   %t1 = and i32 %t0, 1
   %res = icmp eq i32 %t1, 0
   ret i1 %res
 }
 define i1 @scalar_i32_x_is_const2_eq(i32 %y) nounwind {
 ; CHECK-LABEL: scalar_i32_x_is_const2_eq:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    mov w8, #1
-; CHECK-NEXT:    mov w9, #43605
+; CHECK-NEXT:    mov w8, #1 // =0x1
+; CHECK-NEXT:    mov w9, #43605 // =0xaa55
 ; CHECK-NEXT:    lsl w8, w8, w0
 ; CHECK-NEXT:    movk w9, #43605, lsl #16
 ; CHECK-NEXT:    tst w8, w9
 ; CHECK-NEXT:    cset w0, eq
 ; CHECK-NEXT:    ret
   %t0 = shl i32 1, %y
   %t1 = and i32 %t0, 2857740885
   %res = icmp eq i32 %t1, 0
   ret i1 %res
 }
 
 define i1 @scalar_i8_bitsinmiddle_slt(i8 %x, i8 %y) nounwind {
 ; CHECK-LABEL: scalar_i8_bitsinmiddle_slt:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    mov w8, #24
+; CHECK-NEXT:    mov w8, #24 // =0x18
 ; CHECK-NEXT:    // kill: def $w1 killed $w1 def $x1
 ; CHECK-NEXT:    lsl w8, w8, w1
 ; CHECK-NEXT:    and w8, w8, w0
 ; CHECK-NEXT:    ubfx w0, w8, #7, #1
 ; CHECK-NEXT:    ret
   %t0 = shl i8 24, %y
   %t1 = and i8 %t0, %x
   %res = icmp slt i8 %t1, 0
   ret i1 %res
 }
 
 define i1 @scalar_i8_signbit_eq_with_nonzero(i8 %x, i8 %y) nounwind {
 ; CHECK-LABEL: scalar_i8_signbit_eq_with_nonzero:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    mov w8, #-128
-; CHECK-NEXT:    // kill: def $w1 killed $w1 def $x1
-; CHECK-NEXT:    lsl w8, w8, w1
-; CHECK-NEXT:    and w8, w8, w0
-; CHECK-NEXT:    and w8, w8, #0x80
-; CHECK-NEXT:    cmp w8, #1
-; CHECK-NEXT:    cset w0, eq
+; CHECK-NEXT:    mov w0, wzr
 ; CHECK-NEXT:    ret
   %t0 = shl i8 128, %y
   %t1 = and i8 %t0, %x
   %res = icmp eq i8 %t1, 1 ; should be comparing with 0
   ret i1 %res
 }

llvm/test/CodeGen/AArch64/pr59902.ll

 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc < %s -mtriple=aarch64-none-linux-gnu | FileCheck %s
 
 ; This used to miscompile because foldCSELOfCSEL function
 ; doesn't check const x != y
 define i1 @test() {
 ; CHECK-LABEL: test:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    mov x8, #9007199254740990
-; CHECK-NEXT:    movk x8, #65503, lsl #16
-; CHECK-NEXT:    movk x8, #65407, lsl #32
-; CHECK-NEXT:    cmp x8, x8
-; CHECK-NEXT:    csel x9, x8, x8, gt
-; CHECK-NEXT:    cmp x9, x8
-; CHECK-NEXT:    cset w0, eq
+; CHECK-NEXT:    mov w0, #1 // =0x1
 ; CHECK-NEXT:    ret
   %1 = select i1 false, i64 0, i64 9006649496829950
   %2 = call i64 @llvm.smax.i64(i64 %1, i64 9006649496829950)
   %3 = icmp eq i64 %2, 9006649496829950
   ret i1 %3
 }
 
 declare i64 @llvm.smax.i64(i64, i64)

llvm/test/CodeGen/AArch64/urem-seteq-vec-tautological.ll

; CHECK-NEXT:    ret
   %urem = urem <4 x i32> %X, <i32 1, i32 1, i32 2, i32 2>
   %cmp = icmp eq <4 x i32> %urem, <i32 0, i32 1, i32 2, i32 3>
   ret <4 x i1> %cmp
 }
 
 define <4 x i1> @t1_all_odd_eq(<4 x i32> %X) nounwind {
 ; CHECK-LABEL: t1_all_odd_eq:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    mov w8, #43691
+; CHECK-NEXT:    mov w8, #43691 // =0xaaab
 ; CHECK-NEXT:    movk w8, #43690, lsl #16
 ; CHECK-NEXT:    dup v1.4s, w8
 ; CHECK-NEXT:    adrp x8, .LCPI1_0
 ; CHECK-NEXT:    mul v0.4s, v0.4s, v1.4s
 ; CHECK-NEXT:    ldr q1, [x8, :lo12:.LCPI1_0]
 ; CHECK-NEXT:    cmhs v0.4s, v1.4s, v0.4s
 ; CHECK-NEXT:    movi d1, #0xffff0000ffff0000
 ; CHECK-NEXT:    xtn v0.4h, v0.4s
 ; CHECK-NEXT:    eor v0.8b, v0.8b, v1.8b
 ; CHECK-NEXT:    ret
   %urem = urem <4 x i32> %X, <i32 3, i32 1, i32 1, i32 9>
   %cmp = icmp eq <4 x i32> %urem, <i32 0, i32 42, i32 0, i32 42>
   ret <4 x i1> %cmp
 }
 
 define <4 x i1> @t1_all_odd_ne(<4 x i32> %X) nounwind {
 ; CHECK-LABEL: t1_all_odd_ne:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    mov w8, #43691
+; CHECK-NEXT:    mov w8, #43691 // =0xaaab
 ; CHECK-NEXT:    movk w8, #43690, lsl #16
 ; CHECK-NEXT:    dup v1.4s, w8
 ; CHECK-NEXT:    adrp x8, .LCPI2_0
 ; CHECK-NEXT:    mul v0.4s, v0.4s, v1.4s
 ; CHECK-NEXT:    ldr q1, [x8, :lo12:.LCPI2_0]
 ; CHECK-NEXT:    cmhi v0.4s, v0.4s, v1.4s
 ; CHECK-NEXT:    movi d1, #0xffff0000ffff0000
 ; CHECK-NEXT:    xtn v0.4h, v0.4s
 ; CHECK-NEXT:    eor v0.8b, v0.8b, v1.8b
 ; CHECK-NEXT:    ret
   %urem = urem <4 x i32> %X, <i32 3, i32 1, i32 1, i32 9>
   %cmp = icmp ne <4 x i32> %urem, <i32 0, i32 42, i32 0, i32 42>
   ret <4 x i1> %cmp
 }
 
 define <8 x i1> @t2_narrow(<8 x i16> %X) nounwind {
 ; CHECK-LABEL: t2_narrow:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    mov w8, #43691
+; CHECK-NEXT:    mov w8, #43691 // =0xaaab
 ; CHECK-NEXT:    dup v1.8h, w8
 ; CHECK-NEXT:    adrp x8, .LCPI3_0
 ; CHECK-NEXT:    mul v0.8h, v0.8h, v1.8h
 ; CHECK-NEXT:    ldr q1, [x8, :lo12:.LCPI3_0]
 ; CHECK-NEXT:    cmhs v0.8h, v1.8h, v0.8h
 ; CHECK-NEXT:    movi d1, #0xffff0000ffff0000
 ; CHECK-NEXT:    xtn v0.8b, v0.8h
 ; CHECK-NEXT:    eor v0.8b, v0.8b, v1.8b
 ; CHECK-NEXT:    ret
   %urem = urem <8 x i16> %X, <i16 3, i16 1, i16 1, i16 9, i16 3, i16 1, i16 1, i16 9>
   %cmp = icmp eq <8 x i16> %urem, <i16 0, i16 0, i16 42, i16 42, i16 0, i16 0, i16 42, i16 42>
   ret <8 x i1> %cmp
 }
 
 define <2 x i1> @t3_wide(<2 x i64> %X) nounwind {
 ; CHECK-LABEL: t3_wide:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    mov x8, #-6148914691236517206
+; CHECK-NEXT:    mov x8, #-6148914691236517206 // =0xaaaaaaaaaaaaaaaa
 ; CHECK-NEXT:    fmov x9, d0
 ; CHECK-NEXT:    movk x8, #43691
 ; CHECK-NEXT:    mov x10, v0.d[1]
 ; CHECK-NEXT:    mul x9, x9, x8
 ; CHECK-NEXT:    mul x8, x10, x8
 ; CHECK-NEXT:    fmov d0, x9
 ; CHECK-NEXT:    adrp x9, .LCPI4_0
 ; CHECK-NEXT:    mov v0.d[1], x8

llvm/test/CodeGen/ARM/bfi.ll

; CHECK-NEXT:    bx lr
   %cmp = icmp eq i32 %and, 0
   %sel = select i1 %cmp, i32 %y2, i32 %or
   ret i32 %sel
 }
 
 define i32 @f13(i32 %x, i32 %y) {
 ; CHECK-LABEL: f13:
 ; CHECK:       @ %bb.0:
-; CHECK-NEXT:    and r2, r0, #4
-; CHECK-NEXT:    bic r0, r1, #255
-; CHECK-NEXT:    cmp r2, #42
-; CHECK-NEXT:    orrne r0, r0, #16
+; CHECK-NEXT:    mov r0, r1
+; CHECK-NEXT:    mov r1, #16
+; CHECK-NEXT:    bfi r0, r1, #0, #8
 ; CHECK-NEXT:    bx lr
   %y2 = and i32 %y, 4294967040 ; 0xFFFFFF00
   %and = and i32 %x, 4
   %or = or i32 %y2, 16
   %cmp = icmp eq i32 %and, 42 ; Not comparing against zero!
   %sel = select i1 %cmp, i32 %y2, i32 %or
   ret i32 %sel
 }

llvm/test/CodeGen/ARM/cmp-peephole.ll

; THUMB2-NEXT:    bx lr
   %or = or i32 %a, %b
   %res = icmp ne i32 %or, 0
   ret i1 %res
 }
 
 define i1 @cmp_ne_zero_or_ri(i32 %a) {
 ; ARM-LABEL: cmp_ne_zero_or_ri:
 ; ARM:       @ %bb.0:
-; ARM-NEXT:    orrs r0, r0, #42
-; ARM-NEXT:    movwne r0, #1
+; ARM-NEXT:    mov r0, #1
 ; ARM-NEXT:    bx lr
 ;
 ; THUMB-LABEL: cmp_ne_zero_or_ri:
 ; THUMB:       @ %bb.0:
-; THUMB-NEXT:    movs r1, #42
-; THUMB-NEXT:    orrs r0, r1
-; THUMB-NEXT:    subs r1, r0, #1
-; THUMB-NEXT:    sbcs r0, r1
+; THUMB-NEXT:    movs r0, #1
 ; THUMB-NEXT:    bx lr
 ;
 ; THUMB2-LABEL: cmp_ne_zero_or_ri:
 ; THUMB2:       @ %bb.0:
-; THUMB2-NEXT:    orrs r0, r0, #42
-; THUMB2-NEXT:    it ne
-; THUMB2-NEXT:    movne r0, #1
+; THUMB2-NEXT:    movs r0, #1
 ; THUMB2-NEXT:    bx lr
   %or = or i32 %a, 42
   %res = icmp ne i32 %or, 0
   ret i1 %res
 }
 
 define i1 @cmp_ne_zero_or_rsr(i32 %a, i32 %b, i32 %c) {
 ; ARM-LABEL: cmp_ne_zero_or_rsr:

 define i1 @cmp_eq_zero_or_ri(i32 %a) {
 ; ARM-LABEL: cmp_eq_zero_or_ri:
 ; ARM:       @ %bb.0:
 ; ARM-NEXT:    mov r0, #0
 ; ARM-NEXT:    bx lr
 ;
 ; THUMB-LABEL: cmp_eq_zero_or_ri:
 ; THUMB:       @ %bb.0:
-; THUMB-NEXT:    movs r1, #42
-; THUMB-NEXT:    orrs r0, r1
-; THUMB-NEXT:    rsbs r1, r0, #0
-; THUMB-NEXT:    adcs r0, r1
+; THUMB-NEXT:    movs r0, #0
 ; THUMB-NEXT:    bx lr
 ;
 ; THUMB2-LABEL: cmp_eq_zero_or_ri:
 ; THUMB2:       @ %bb.0:
 ; THUMB2-NEXT:    movs r0, #0
 ; THUMB2-NEXT:    bx lr
   %or = or i32 %a, 42
   %res = icmp eq i32 %or, 0

llvm/test/CodeGen/ARM/hoist-and-by-const-from-lshr-in-eqcmp-zero.ll

; THUMB78-NEXT:    bx lr
   ret i1 %res
 }
 
 ;------------------------------------------------------------------------------;
 ; A few negative tests
 ;------------------------------------------------------------------------------;
 
 define i1 @negative_scalar_i8_bitsinmiddle_slt(i8 %x, i8 %y) nounwind {
-; ARM6-LABEL: negative_scalar_i8_bitsinmiddle_slt:
-; ARM6:       @ %bb.0:
-; ARM6-NEXT:    uxtb r1, r1
-; ARM6-NEXT:    mov r2, #24
-; ARM6-NEXT:    ands r0, r0, r2, lsr r1
-; ARM6-NEXT:    mov r0, #0
-; ARM6-NEXT:    movmi r0, #1
-; ARM6-NEXT:    bx lr
-;
-; ARM78-LABEL: negative_scalar_i8_bitsinmiddle_slt:
-; ARM78:       @ %bb.0:
-; ARM78-NEXT:    uxtb r1, r1
-; ARM78-NEXT:    mov r2, #24
-; ARM78-NEXT:    ands r0, r0, r2, lsr r1
-; ARM78-NEXT:    mov r0, #0
-; ARM78-NEXT:    movwmi r0, #1
-; ARM78-NEXT:    bx lr
-;
-; THUMB6-LABEL: negative_scalar_i8_bitsinmiddle_slt:
-; THUMB6:       @ %bb.0:
-; THUMB6-NEXT:    uxtb r1, r1
-; THUMB6-NEXT:    movs r2, #24
-; THUMB6-NEXT:    lsrs r2, r1
-; THUMB6-NEXT:    ands r2, r0
-; THUMB6-NEXT:    bmi .LBB20_2
-; THUMB6-NEXT:  @ %bb.1:
-; THUMB6-NEXT:    movs r0, #0
-; THUMB6-NEXT:    bx lr
-; THUMB6-NEXT:  .LBB20_2:
-; THUMB6-NEXT:    movs r0, #1
-; THUMB6-NEXT:    bx lr
+; ARM-LABEL: negative_scalar_i8_bitsinmiddle_slt:
+; ARM:       @ %bb.0:
+; ARM-NEXT:    mov r0, #0
+; ARM-NEXT:    bx lr
 ;
-; THUMB78-LABEL: negative_scalar_i8_bitsinmiddle_slt:
-; THUMB78:       @ %bb.0:
-; THUMB78-NEXT:    uxtb r1, r1
-; THUMB78-NEXT:    movs r2, #24
-; THUMB78-NEXT:    lsr.w r1, r2, r1
-; THUMB78-NEXT:    ands r0, r1
-; THUMB78-NEXT:    mov.w r0, #0
-; THUMB78-NEXT:    it mi
-; THUMB78-NEXT:    movmi r0, #1
-; THUMB78-NEXT:    bx lr
+; THUMB-LABEL: negative_scalar_i8_bitsinmiddle_slt:
+; THUMB:       @ %bb.0:
+; THUMB-NEXT:    movs r0, #0
+; THUMB-NEXT:    bx lr
   %t0 = lshr i8 24, %y
   %t1 = and i8 %t0, %x
   %res = icmp slt i8 %t1, 0
   ret i1 %res
 }
 
 define i1 @scalar_i8_signbit_eq_with_nonzero(i8 %x, i8 %y) nounwind {
 ; ARM-LABEL: scalar_i8_signbit_eq_with_nonzero:

llvm/test/CodeGen/ARM/hoist-and-by-const-from-shl-in-eqcmp-zero.ll

 }
 
 define i1 @scalar_i8_signbit_eq_with_nonzero(i8 %x, i8 %y) nounwind {
 ; ARM-LABEL: scalar_i8_signbit_eq_with_nonzero:
 ; ARM:       @ %bb.0:
 ; ARM-NEXT:    mov r0, #0
 ; ARM-NEXT:    bx lr
 ;
-; THUMB6-LABEL: scalar_i8_signbit_eq_with_nonzero:
-; THUMB6:       @ %bb.0:
-; THUMB6-NEXT:    uxtb r1, r1
-; THUMB6-NEXT:    movs r2, #127
-; THUMB6-NEXT:    mvns r2, r2
-; THUMB6-NEXT:    lsls r2, r1
-; THUMB6-NEXT:    ands r2, r0
-; THUMB6-NEXT:    uxtb r0, r2
-; THUMB6-NEXT:    subs r1, r0, #1
-; THUMB6-NEXT:    rsbs r0, r1, #0
-; THUMB6-NEXT:    adcs r0, r1
-; THUMB6-NEXT:    bx lr
-;
-; THUMB78-LABEL: scalar_i8_signbit_eq_with_nonzero:
-; THUMB78:       @ %bb.0:
-; THUMB78-NEXT:    movs r0, #0
-; THUMB78-NEXT:    bx lr
+; THUMB-LABEL: scalar_i8_signbit_eq_with_nonzero:
+; THUMB:       @ %bb.0:
+; THUMB-NEXT:    movs r0, #0
+; THUMB-NEXT:    bx lr
   %t0 = shl i8 128, %y
   %t1 = and i8 %t0, %x
   %res = icmp eq i8 %t1, 1 ; should be comparing with 0
   ret i1 %res
 }

llvm/test/CodeGen/Hexagon/vect/zext-v4i1.ll

 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc -march=hexagon -hexagon-instsimplify=0 < %s | FileCheck %s
 
 ; Check that this compiles successfully.
 
 target datalayout = "e-m:e-p:32:32:32-a:0-n16:32-i64:64:64-i32:32:32-i16:16:16-i1:8:8-f32:32:32-f64:64:64-v32:32:32-v64:64:64-v512:512:512-v1024:1024:1024-v2048:2048:2048"
 target triple = "hexagon"
 
 define i32 @fred(ptr %a0) #0 {
 ; CHECK-LABEL: fred:
 ; CHECK:       // %bb.0: // %b0
 ; CHECK-NEXT:    {
-; CHECK-NEXT:     if (p0) jump:nt .LBB0_2
-; CHECK-NEXT:    }
-; CHECK-NEXT:  // %bb.1: // %b2
-; CHECK-NEXT:    {
-; CHECK-NEXT:     r3:2 = combine(#0,#0)
-; CHECK-NEXT:     r1:0 = memd(r0+#0)
-; CHECK-NEXT:    }
-; CHECK-NEXT:    {
-; CHECK-NEXT:     p0 = vcmph.eq(r1:0,r3:2)
-; CHECK-NEXT:    }
-; CHECK-NEXT:    {
-; CHECK-NEXT:     r1:0 = mask(p0)
-; CHECK-NEXT:    }
-; CHECK-NEXT:    {
-; CHECK-NEXT:     r0 = and(r0,#1)
-; CHECK-NEXT:    }
-; CHECK-NEXT:    {
-; CHECK-NEXT:     p0 = cmp.eq(r0,#11)
-; CHECK-NEXT:     r0 = #1
-; CHECK-NEXT:    }
-; CHECK-NEXT:    {
+; CHECK-NEXT:     if (!p0) r0 = #1
 ; CHECK-NEXT:     if (p0) r0 = #0
 ; CHECK-NEXT:     jumpr r31
 ; CHECK-NEXT:    }
-; CHECK-NEXT:  .LBB0_2: // %b14
-; CHECK-NEXT:    {
-; CHECK-NEXT:     r0 = #0
-; CHECK-NEXT:     jumpr r31
-; CHECK-NEXT:    }
 b0:
   switch i32 undef, label %b14 [
     i32 5, label %b2
     i32 3, label %b1
   ]
 
 b1:                                               ; preds = %b0
   br label %b14

llvm/test/CodeGen/RISCV/sextw-removal.ll

 
 declare i64 @llvm.ctpop.i64(i64)
 
 define void @test8(i32 signext %arg, i32 signext %arg1) nounwind {
 ; CHECK-LABEL: test8:
 ; CHECK:       # %bb.0: # %bb
 ; CHECK-NEXT:    addi sp, sp, -16
 ; CHECK-NEXT:    sd ra, 8(sp) # 8-byte Folded Spill
+; CHECK-NEXT:    sd s0, 0(sp) # 8-byte Folded Spill
 ; CHECK-NEXT:    sraw a0, a0, a1
+; CHECK-NEXT:    li s0, 1
 ; CHECK-NEXT:  .LBB7_1: # %bb2
 ; CHECK-NEXT:    # =>This Inner Loop Header: Depth=1
 ; CHECK-NEXT:    call foo@plt
 ; CHECK-NEXT:    ori a0, a0, -256
-; CHECK-NEXT:    bnez a0, .LBB7_1
+; CHECK-NEXT:    bnez s0, .LBB7_1
 ; CHECK-NEXT:  # %bb.2: # %bb7
 ; CHECK-NEXT:    ld ra, 8(sp) # 8-byte Folded Reload
+; CHECK-NEXT:    ld s0, 0(sp) # 8-byte Folded Reload
 ; CHECK-NEXT:    addi sp, sp, 16
 ; CHECK-NEXT:    ret
 ;
 ; NOREMOVAL-LABEL: test8:
 ; NOREMOVAL:       # %bb.0: # %bb
 ; NOREMOVAL-NEXT:    addi sp, sp, -16
 ; NOREMOVAL-NEXT:    sd ra, 8(sp) # 8-byte Folded Spill
+; NOREMOVAL-NEXT:    sd s0, 0(sp) # 8-byte Folded Spill
 ; NOREMOVAL-NEXT:    sraw a0, a0, a1
+; NOREMOVAL-NEXT:    li s0, 1
 ; NOREMOVAL-NEXT:  .LBB7_1: # %bb2
 ; NOREMOVAL-NEXT:    # =>This Inner Loop Header: Depth=1
 ; NOREMOVAL-NEXT:    sext.w a0, a0
 ; NOREMOVAL-NEXT:    call foo@plt
 ; NOREMOVAL-NEXT:    ori a0, a0, -256
-; NOREMOVAL-NEXT:    bnez a0, .LBB7_1
+; NOREMOVAL-NEXT:    bnez s0, .LBB7_1
 ; NOREMOVAL-NEXT:  # %bb.2: # %bb7
 ; NOREMOVAL-NEXT:    ld ra, 8(sp) # 8-byte Folded Reload
+; NOREMOVAL-NEXT:    ld s0, 0(sp) # 8-byte Folded Reload
 ; NOREMOVAL-NEXT:    addi sp, sp, 16
 ; NOREMOVAL-NEXT:    ret
 bb:
   %i = ashr i32 %arg, %arg1
   br label %bb2
 
 bb2:                                              ; preds = %bb2, %bb
   %i3 = phi i32 [ %i, %bb ], [ %i6, %bb2 ]

llvm/test/CodeGen/X86/2007-10-12-CoalesceExtSubReg.ll

 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc < %s -mtriple=i686-- | FileCheck %s
 
 define signext i16 @f(ptr %bp, ptr %ss)   {
 ; CHECK-LABEL: f:
 ; CHECK:       # %bb.0: # %entry
-; CHECK-NEXT:    pushl %esi
-; CHECK-NEXT:    .cfi_def_cfa_offset 8
-; CHECK-NEXT:    .cfi_offset %esi, -8
 ; CHECK-NEXT:    movl {{[0-9]+}}(%esp), %eax
-; CHECK-NEXT:    movl {{[0-9]+}}(%esp), %ecx
+; CHECK-NEXT:    movb $1, %cl
 ; CHECK-NEXT:    .p2align 4, 0x90
 ; CHECK-NEXT:  .LBB0_1: # %cond_next127
 ; CHECK-NEXT:    # =>This Inner Loop Header: Depth=1
 ; CHECK-NEXT:    movl (%eax), %edx
-; CHECK-NEXT:    movl (%ecx), %esi
 ; CHECK-NEXT:    andl $15, %edx
-; CHECK-NEXT:    andl $15, %esi
-; CHECK-NEXT:    addl %esi, (%ecx)
-; CHECK-NEXT:    cmpl $63, %edx
-; CHECK-NEXT:    jb .LBB0_1
+; CHECK-NEXT:    addl %edx, (%eax)
+; CHECK-NEXT:    testb %cl, %cl
+; CHECK-NEXT:    jne .LBB0_1
 ; CHECK-NEXT:  # %bb.2: # %UnifiedReturnBlock
 ; CHECK-NEXT:    xorl %eax, %eax
-; CHECK-NEXT:    popl %esi
-; CHECK-NEXT:    .cfi_def_cfa_offset 4
 ; CHECK-NEXT:    retl
 entry:
   br label %cond_next127
 
 cond_next127:   ; preds = %cond_next391, %entry
   %v.1 = phi i32 [ undef, %entry ], [ %tmp411, %cond_next391 ]    ; <i32> [#uses=1]
   %tmp149 = mul i32 0, %v.1   ; <i32> [#uses=0]
   %tmpss = load i32, ptr %ss, align 4    ; <i32> [#uses=1]

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

 true:
   ret void
 
 false:
   ret void
 }
 
 define void @test7(<8 x i1> %mask)  {
-; KNL-LABEL: test7:
-; KNL:       ## %bb.0: ## %allocas
-; KNL-NEXT:    vpmovsxwq %xmm0, %zmm0
-; KNL-NEXT:    vpsllq $63, %zmm0, %zmm0
-; KNL-NEXT:    vptestmq %zmm0, %zmm0, %k0
-; KNL-NEXT:    kmovw %k0, %eax
-; KNL-NEXT:    orb $85, %al
-; KNL-NEXT:    vzeroupper
-; KNL-NEXT:    retq
-;
-; SKX-LABEL: test7:
-; SKX:       ## %bb.0: ## %allocas
-; SKX-NEXT:    vpsllw $15, %xmm0, %xmm0
-; SKX-NEXT:    vpmovw2m %xmm0, %k0
-; SKX-NEXT:    kmovd %k0, %eax
-; SKX-NEXT:    orb $85, %al
-; SKX-NEXT:    retq
-;
-; AVX512BW-LABEL: test7:
-; AVX512BW:       ## %bb.0: ## %allocas
-; AVX512BW-NEXT:    vpsllw $15, %xmm0, %xmm0
-; AVX512BW-NEXT:    vpmovw2m %zmm0, %k0
-; AVX512BW-NEXT:    kmovd %k0, %eax
-; AVX512BW-NEXT:    orb $85, %al
-; AVX512BW-NEXT:    vzeroupper
-; AVX512BW-NEXT:    retq
-;
-; AVX512DQ-LABEL: test7:
-; AVX512DQ:       ## %bb.0: ## %allocas
-; AVX512DQ-NEXT:    vpmovsxwq %xmm0, %zmm0
-; AVX512DQ-NEXT:    vpsllq $63, %zmm0, %zmm0
-; AVX512DQ-NEXT:    vpmovq2m %zmm0, %k0
-; AVX512DQ-NEXT:    kmovw %k0, %eax
-; AVX512DQ-NEXT:    orb $85, %al
-; AVX512DQ-NEXT:    vzeroupper
-; AVX512DQ-NEXT:    retq
+; CHECK-LABEL: test7:
+; CHECK:       ## %bb.0: ## %allocas
+; CHECK-NEXT:    movb $1, %al
+; CHECK-NEXT:    testb %al, %al
+; CHECK-NEXT:    retq
 ;
 ; X86-LABEL: test7:
 ; X86:       ## %bb.0: ## %allocas
-; X86-NEXT:    vpsllw $15, %xmm0, %xmm0
-; X86-NEXT:    vpmovw2m %xmm0, %k0
-; X86-NEXT:    kmovd %k0, %eax
-; X86-NEXT:    orb $85, %al
+; X86-NEXT:    movb $1, %al
+; X86-NEXT:    testb %al, %al
 ; X86-NEXT:    retl
 allocas:
   %a= or <8 x i1> %mask, <i1 true, i1 false, i1 true, i1 false, i1 true, i1 false, i1 true, i1 false>
   %b = bitcast <8 x i1> %a to i8
   %c = icmp eq i8 %b, 0
   br i1 %c, label %true, label %false
 
 true:

 ; KNL-NEXT:    vinserti64x4 $1, %ymm2, %zmm3, %zmm2
 ; KNL-NEXT:    vpternlogq $200, %zmm1, %zmm0, %zmm2
 ; KNL-NEXT:    vextracti64x4 $1, %zmm2, %ymm0
 ; KNL-NEXT:    vpor %ymm0, %ymm2, %ymm0
 ; KNL-NEXT:    vpmovsxwd %ymm0, %zmm0
 ; KNL-NEXT:    vpslld $31, %zmm0, %zmm0
 ; KNL-NEXT:    vptestmd %zmm0, %zmm0, %k0
 ; KNL-NEXT:    kortestw %k0, %k0
 ; KNL-NEXT:    je LBB77_1
 ; KNL-NEXT:  ## %bb.2: ## %exit
 ; KNL-NEXT:    vzeroupper
 ; KNL-NEXT:    retq
 ; KNL-NEXT:  LBB77_1: ## %bar
 ; KNL-NEXT:    pushq %rax
 ; KNL-NEXT:    .cfi_def_cfa_offset 16
 ; KNL-NEXT:    vzeroupper
 ; KNL-NEXT:    callq _foo

llvm/test/CodeGen/X86/cmp.ll

; CHECK-NEXT:    retq # encoding: [0xc3]
   %tobool = icmp ne i32 %and, 0
   %cond = select i1 %tobool, i32 %intra, i32 %base
   ret i32 %cond
 }
 
 define i32 @test14(i32 %mask, i32 %base, i32 %intra) {
 ; CHECK-LABEL: test14:
 ; CHECK:       # %bb.0:
-; CHECK-NEXT:    movl %esi, %eax # encoding: [0x89,0xf0]
-; CHECK-NEXT:    shrl $7, %edi # encoding: [0xc1,0xef,0x07]
-; CHECK-NEXT:    cmovnsl %edx, %eax # encoding: [0x0f,0x49,0xc2]
+; CHECK-NEXT:    movl %edx, %eax # encoding: [0x89,0xd0]
 ; CHECK-NEXT:    retq # encoding: [0xc3]
   %s = lshr i32 %mask, 7
   %tobool = icmp sgt i32 %s, -1
   %cond = select i1 %tobool, i32 %intra, i32 %base
   ret i32 %cond
 }
 
 ; PR19964

llvm/test/CodeGen/X86/fold-rmw-ops.ll

 b:
   tail call void @b()
   ret void
 }
 
 define void @or64_imm32_br() nounwind {
 ; CHECK-LABEL: or64_imm32_br:
 ; CHECK:       # %bb.0: # %entry
-; CHECK-NEXT:    orq $16777215, g64(%rip) # encoding: [0x48,0x81,0x0d,A,A,A,A,0xff,0xff,0xff,0x00]
-; CHECK-NEXT:    # fixup A - offset: 3, value: g64-8, kind: reloc_riprel_4byte
+; CHECK-NEXT:    orl $16777215, g64(%rip) # encoding: [0x81,0x0d,A,A,A,A,0xff,0xff,0xff,0x00]
+; CHECK-NEXT:    # fixup A - offset: 2, value: g64-8, kind: reloc_riprel_4byte
 ; CHECK-NEXT:    # imm = 0xFFFFFF
+; CHECK-NEXT:    movb $1, %al # encoding: [0xb0,0x01]
+; CHECK-NEXT:    testb %al, %al # encoding: [0x84,0xc0]
 ; CHECK-NEXT:    jne b # TAILCALL
 ; CHECK-NEXT:    # encoding: [0x75,A]
 ; CHECK-NEXT:    # fixup A - offset: 1, value: b-1, kind: FK_PCRel_1
 ; CHECK-NEXT:  # %bb.1: # %a
 ; CHECK-NEXT:    jmp a # TAILCALL
 ; CHECK-NEXT:    # encoding: [0xeb,A]
 ; CHECK-NEXT:    # fixup A - offset: 1, value: a-1, kind: FK_PCRel_1
 entry:
   %load1 = load i64, ptr @g64
   ; Or 0x00FFFFFF, a positive immediate requiring 24-bits.
   %or = or i64 %load1, 16777215
   store i64 %or, ptr @g64
   %cond = icmp eq i64 %or, 0
   br i1 %cond, label %a, label %b
 
 a:
   tail call void @a()
   ret void
 
 b:
   tail call void @b()
   ret void
 }
 
 define void @or64_sext_imm32_br() nounwind {
 ; CHECK-LABEL: or64_sext_imm32_br:
 ; CHECK:       # %bb.0: # %entry
 ; CHECK-NEXT:    orq $-2147483648, g64(%rip) # encoding: [0x48,0x81,0x0d,A,A,A,A,0x00,0x00,0x00,0x80]
 ; CHECK-NEXT:    # fixup A - offset: 3, value: g64-8, kind: reloc_riprel_4byte
 ; CHECK-NEXT:    # imm = 0x80000000
+; CHECK-NEXT:    movb $1, %al # encoding: [0xb0,0x01]
+; CHECK-NEXT:    testb %al, %al # encoding: [0x84,0xc0]
 ; CHECK-NEXT:    jne b # TAILCALL
 ; CHECK-NEXT:    # encoding: [0x75,A]
 ; CHECK-NEXT:    # fixup A - offset: 1, value: b-1, kind: FK_PCRel_1
 ; CHECK-NEXT:  # %bb.1: # %a
 ; CHECK-NEXT:    jmp a # TAILCALL
 ; CHECK-NEXT:    # encoding: [0xeb,A]
 ; CHECK-NEXT:    # fixup A - offset: 1, value: a-1, kind: FK_PCRel_1
 entry:

 b:
   tail call void @b()
   ret void
 }
 
 define void @or64_imm8_br() nounwind {
 ; CHECK-LABEL: or64_imm8_br:
 ; CHECK:       # %bb.0: # %entry
-; CHECK-NEXT:    orq $15, g64(%rip) # encoding: [0x48,0x83,0x0d,A,A,A,A,0x0f]
-; CHECK-NEXT:    # fixup A - offset: 3, value: g64-5, kind: reloc_riprel_4byte
+; CHECK-NEXT:    orb $15, g64(%rip) # encoding: [0x80,0x0d,A,A,A,A,0x0f]
+; CHECK-NEXT:    # fixup A - offset: 2, value: g64-5, kind: reloc_riprel_4byte
+; CHECK-NEXT:    movb $1, %al # encoding: [0xb0,0x01]
+; CHECK-NEXT:    testb %al, %al # encoding: [0x84,0xc0]
 ; CHECK-NEXT:    jne b # TAILCALL
 ; CHECK-NEXT:    # encoding: [0x75,A]
 ; CHECK-NEXT:    # fixup A - offset: 1, value: b-1, kind: FK_PCRel_1
 ; CHECK-NEXT:  # %bb.1: # %a
 ; CHECK-NEXT:    jmp a # TAILCALL
 ; CHECK-NEXT:    # encoding: [0xeb,A]
 ; CHECK-NEXT:    # fixup A - offset: 1, value: a-1, kind: FK_PCRel_1
 entry:
b:
   ret void
 }
 
 define void @or64_imm8_neg_br() nounwind {
 ; CHECK-LABEL: or64_imm8_neg_br:
 ; CHECK:       # %bb.0: # %entry
 ; CHECK-NEXT:    orq $-4, g64(%rip) # encoding: [0x48,0x83,0x0d,A,A,A,A,0xfc]
 ; CHECK-NEXT:    # fixup A - offset: 3, value: g64-5, kind: reloc_riprel_4byte
+; CHECK-NEXT:    movb $1, %al # encoding: [0xb0,0x01]
+; CHECK-NEXT:    testb %al, %al # encoding: [0x84,0xc0]
 ; CHECK-NEXT:    jne b # TAILCALL
 ; CHECK-NEXT:    # encoding: [0x75,A]
 ; CHECK-NEXT:    # fixup A - offset: 1, value: b-1, kind: FK_PCRel_1
 ; CHECK-NEXT:  # %bb.1: # %a
 ; CHECK-NEXT:    jmp a # TAILCALL
 ; CHECK-NEXT:    # encoding: [0xeb,A]
 ; CHECK-NEXT:    # fixup A - offset: 1, value: a-1, kind: FK_PCRel_1
 entry:

 b:
   tail call void @b()
   ret void
 }
 
 define void @or32_imm_br() nounwind {
 ; CHECK-LABEL: or32_imm_br:
 ; CHECK:       # %bb.0: # %entry
-; CHECK-NEXT:    orl $-2147483648, g32(%rip) # encoding: [0x81,0x0d,A,A,A,A,0x00,0x00,0x00,0x80]
-; CHECK-NEXT:    # fixup A - offset: 2, value: g32-8, kind: reloc_riprel_4byte
-; CHECK-NEXT:    # imm = 0x80000000
+; CHECK-NEXT:    orb $-128, g32+3(%rip) # encoding: [0x80,0x0d,A,A,A,A,0x80]
+; CHECK-NEXT:    # fixup A - offset: 2, value: (g32+3)-5, kind: reloc_riprel_4byte
+; CHECK-NEXT:    movb $1, %al # encoding: [0xb0,0x01]
+; CHECK-NEXT:    testb %al, %al # encoding: [0x84,0xc0]
 ; CHECK-NEXT:    jne b # TAILCALL
 ; CHECK-NEXT:    # encoding: [0x75,A]
 ; CHECK-NEXT:    # fixup A - offset: 1, value: b-1, kind: FK_PCRel_1
 ; CHECK-NEXT:  # %bb.1: # %a
 ; CHECK-NEXT:    jmp a # TAILCALL
 ; CHECK-NEXT:    # encoding: [0xeb,A]
 ; CHECK-NEXT:    # fixup A - offset: 1, value: a-1, kind: FK_PCRel_1
 entry:

 b:
   tail call void @b()
   ret void
 }
 
 define void @or32_imm8_br() nounwind {
 ; CHECK-LABEL: or32_imm8_br:
 ; CHECK:       # %bb.0: # %entry
-; CHECK-NEXT:    orl $15, g32(%rip) # encoding: [0x83,0x0d,A,A,A,A,0x0f]
+; CHECK-NEXT:    orb $15, g32(%rip) # encoding: [0x80,0x0d,A,A,A,A,0x0f]
 ; CHECK-NEXT:    # fixup A - offset: 2, value: g32-5, kind: reloc_riprel_4byte
+; CHECK-NEXT:    movb $1, %al # encoding: [0xb0,0x01]
+; CHECK-NEXT:    testb %al, %al # encoding: [0x84,0xc0]
 ; CHECK-NEXT:    jne b # TAILCALL
 ; CHECK-NEXT:    # encoding: [0x75,A]
 ; CHECK-NEXT:    # fixup A - offset: 1, value: b-1, kind: FK_PCRel_1
 ; CHECK-NEXT:  # %bb.1: # %a
 ; CHECK-NEXT:    jmp a # TAILCALL
 ; CHECK-NEXT:    # encoding: [0xeb,A]
 ; CHECK-NEXT:    # fixup A - offset: 1, value: a-1, kind: FK_PCRel_1
 entry:
b:
   ret void
 }
 
 define void @or32_imm8_neg_br() nounwind {
 ; CHECK-LABEL: or32_imm8_neg_br:
 ; CHECK:       # %bb.0: # %entry
 ; CHECK-NEXT:    orl $-4, g32(%rip) # encoding: [0x83,0x0d,A,A,A,A,0xfc]
 ; CHECK-NEXT:    # fixup A - offset: 2, value: g32-5, kind: reloc_riprel_4byte
+; CHECK-NEXT:    movb $1, %al # encoding: [0xb0,0x01]
+; CHECK-NEXT:    testb %al, %al # encoding: [0x84,0xc0]
 ; CHECK-NEXT:    jne b # TAILCALL
 ; CHECK-NEXT:    # encoding: [0x75,A]
 ; CHECK-NEXT:    # fixup A - offset: 1, value: b-1, kind: FK_PCRel_1
 ; CHECK-NEXT:  # %bb.1: # %a
 ; CHECK-NEXT:    jmp a # TAILCALL
 ; CHECK-NEXT:    # encoding: [0xeb,A]
 ; CHECK-NEXT:    # fixup A - offset: 1, value: a-1, kind: FK_PCRel_1
 entry:

 b:
   tail call void @b()
   ret void
 }
 
 define void @or16_imm_br() nounwind {
 ; CHECK-LABEL: or16_imm_br:
 ; CHECK:       # %bb.0: # %entry
-; CHECK-NEXT:    orw $-32768, g16(%rip) # encoding: [0x66,0x81,0x0d,A,A,A,A,0x00,0x80]
-; CHECK-NEXT:    # fixup A - offset: 3, value: g16-6, kind: reloc_riprel_4byte
-; CHECK-NEXT:    # imm = 0x8000
+; CHECK-NEXT:    orb $-128, g16+1(%rip) # encoding: [0x80,0x0d,A,A,A,A,0x80]
+; CHECK-NEXT:    # fixup A - offset: 2, value: (g16+1)-5, kind: reloc_riprel_4byte
+; CHECK-NEXT:    movb $1, %al # encoding: [0xb0,0x01]
+; CHECK-NEXT:    testb %al, %al # encoding: [0x84,0xc0]
 ; CHECK-NEXT:    jne b # TAILCALL
 ; CHECK-NEXT:    # encoding: [0x75,A]
 ; CHECK-NEXT:    # fixup A - offset: 1, value: b-1, kind: FK_PCRel_1
 ; CHECK-NEXT:  # %bb.1: # %a
 ; CHECK-NEXT:    jmp a # TAILCALL
 ; CHECK-NEXT:    # encoding: [0xeb,A]
 ; CHECK-NEXT:    # fixup A - offset: 1, value: a-1, kind: FK_PCRel_1
 entry:

 b:
   tail call void @b()
   ret void
 }
 
 define void @or16_imm8_br() nounwind {
 ; CHECK-LABEL: or16_imm8_br:
 ; CHECK:       # %bb.0: # %entry
-; CHECK-NEXT:    orw $15, g16(%rip) # encoding: [0x66,0x83,0x0d,A,A,A,A,0x0f]
-; CHECK-NEXT:    # fixup A - offset: 3, value: g16-5, kind: reloc_riprel_4byte
+; CHECK-NEXT:    orb $15, g16(%rip) # encoding: [0x80,0x0d,A,A,A,A,0x0f]
+; CHECK-NEXT:    # fixup A - offset: 2, value: g16-5, kind: reloc_riprel_4byte
+; CHECK-NEXT:    movb $1, %al # encoding: [0xb0,0x01]
+; CHECK-NEXT:    testb %al, %al # encoding: [0x84,0xc0]
 ; CHECK-NEXT:    jne b # TAILCALL
 ; CHECK-NEXT:    # encoding: [0x75,A]
 ; CHECK-NEXT:    # fixup A - offset: 1, value: b-1, kind: FK_PCRel_1
 ; CHECK-NEXT:  # %bb.1: # %a
 ; CHECK-NEXT:    jmp a # TAILCALL
 ; CHECK-NEXT:    # encoding: [0xeb,A]
 ; CHECK-NEXT:    # fixup A - offset: 1, value: a-1, kind: FK_PCRel_1
 entry:
b:
   ret void
 }
 
 define void @or16_imm8_neg_br() nounwind {
 ; CHECK-LABEL: or16_imm8_neg_br:
 ; CHECK:       # %bb.0: # %entry
 ; CHECK-NEXT:    orw $-4, g16(%rip) # encoding: [0x66,0x83,0x0d,A,A,A,A,0xfc]
 ; CHECK-NEXT:    # fixup A - offset: 3, value: g16-5, kind: reloc_riprel_4byte
+; CHECK-NEXT:    movb $1, %al # encoding: [0xb0,0x01]
+; CHECK-NEXT:    testb %al, %al # encoding: [0x84,0xc0]
 ; CHECK-NEXT:    jne b # TAILCALL
 ; CHECK-NEXT:    # encoding: [0x75,A]
 ; CHECK-NEXT:    # fixup A - offset: 1, value: b-1, kind: FK_PCRel_1
 ; CHECK-NEXT:  # %bb.1: # %a
 ; CHECK-NEXT:    jmp a # TAILCALL
 ; CHECK-NEXT:    # encoding: [0xeb,A]
 ; CHECK-NEXT:    # fixup A - offset: 1, value: a-1, kind: FK_PCRel_1
 entry:
b:
   ret void
 }
 
 define void @or8_imm_br() nounwind {
 ; CHECK-LABEL: or8_imm_br:
 ; CHECK:       # %bb.0: # %entry
 ; CHECK-NEXT:    orb $-4, g8(%rip) # encoding: [0x80,0x0d,A,A,A,A,0xfc]
 ; CHECK-NEXT:    # fixup A - offset: 2, value: g8-5, kind: reloc_riprel_4byte
+; CHECK-NEXT:    movb $1, %al # encoding: [0xb0,0x01]
+; CHECK-NEXT:    testb %al, %al # encoding: [0x84,0xc0]
 ; CHECK-NEXT:    jne b # TAILCALL
 ; CHECK-NEXT:    # encoding: [0x75,A]
 ; CHECK-NEXT:    # fixup A - offset: 1, value: b-1, kind: FK_PCRel_1
 ; CHECK-NEXT:  # %bb.1: # %a
 ; CHECK-NEXT:    jmp a # TAILCALL
 ; CHECK-NEXT:    # encoding: [0xeb,A]
 ; CHECK-NEXT:    # fixup A - offset: 1, value: a-1, kind: FK_PCRel_1
 entry:

llvm/test/CodeGen/X86/hoist-and-by-const-from-shl-in-eqcmp-zero.ll

; X64-NEXT:    retq
   %t1 = and i8 %t0, %x
   %res = icmp slt i8 %t1, 0
   ret i1 %res
 }
 
 define i1 @scalar_i8_signbit_eq_with_nonzero(i8 %x, i8 %y) nounwind {
 ; X86-LABEL: scalar_i8_signbit_eq_with_nonzero:
 ; X86:       # %bb.0:
-; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %ecx
-; X86-NEXT:    movb $-128, %al
-; X86-NEXT:    shlb %cl, %al
-; X86-NEXT:    andb {{[0-9]+}}(%esp), %al
-; X86-NEXT:    cmpb $1, %al
-; X86-NEXT:    sete %al
+; X86-NEXT:    xorl %eax, %eax
 ; X86-NEXT:    retl
 ;
 ; X64-LABEL: scalar_i8_signbit_eq_with_nonzero:
 ; X64:       # %bb.0:
-; X64-NEXT:    movl %esi, %ecx
-; X64-NEXT:    movb $-128, %al
-; X64-NEXT:    # kill: def $cl killed $cl killed $ecx
-; X64-NEXT:    shlb %cl, %al
-; X64-NEXT:    andb %dil, %al
-; X64-NEXT:    cmpb $1, %al
-; X64-NEXT:    sete %al
+; X64-NEXT:    xorl %eax, %eax
 ; X64-NEXT:    retq
   %t0 = shl i8 128, %y
   %t1 = and i8 %t0, %x
   %res = icmp eq i8 %t1, 1 ; should be comparing with 0
   ret i1 %res
 }

llvm/test/CodeGen/X86/omit-urem-of-power-of-two-or-zero-when-comparing-with-zero.ll

 define i1 @p3_scalar_shifted2_urem_by_const(i32 %x, i32 %y) {
 ; CHECK-LABEL: p3_scalar_shifted2_urem_by_const:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    movl %esi, %ecx
 ; CHECK-NEXT:    andl $2, %edi
 ; CHECK-NEXT:    # kill: def $cl killed $cl killed $ecx
 ; CHECK-NEXT:    shll %cl, %edi
 ; CHECK-NEXT:    imull $-1431655765, %edi, %eax # imm = 0xAAAAAAAB
-; CHECK-NEXT:    cmpl $1431655766, %eax # imm = 0x55555556
+; CHECK-NEXT:    cmpl $1431655765, %eax # imm = 0x55555555
 ; CHECK-NEXT:    setb %al
 ; CHECK-NEXT:    retq
   %t0 = and i32 %x, 2 ; clearly a power-of-two or zero
   %t1 = shl i32 %t0, %y ; will still be a power-of-two or zero with any %y
   %t2 = urem i32 %t1, 3 ; '3' is clearly not a power of two
   %t3 = icmp eq i32 %t2, 0
   ret i1 %t3
 }

llvm/test/CodeGen/X86/or-with-overflow.ll

 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc < %s -mtriple=i686-unknown-unknown | FileCheck %s --check-prefix=X86
 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+cmov | FileCheck %s --check-prefix=X64
 
 ;
 ; PR48768 - 'or' clears the overflow flag, so we don't need a separate 'test'.
 ;
 
 define i8 @or_i8_ri(i8 zeroext %0, i8 zeroext %1) {
 ; X86-LABEL: or_i8_ri:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
-; X86-NEXT:    movl %eax, %ecx
-; X86-NEXT:    orb $-17, %cl
-; X86-NEXT:    je .LBB0_2
-; X86-NEXT:  # %bb.1:
-; X86-NEXT:    movl %ecx, %eax
-; X86-NEXT:  .LBB0_2:
+; X86-NEXT:    orb $-17, %al
 ; X86-NEXT:    retl
 ;
 ; X64-LABEL: or_i8_ri:
 ; X64:       # %bb.0:
 ; X64-NEXT:    movl %edi, %eax
 ; X64-NEXT:    orb $-17, %al
-; X64-NEXT:    movzbl %al, %eax
-; X64-NEXT:    cmovel %edi, %eax
 ; X64-NEXT:    # kill: def $al killed $al killed $eax
 ; X64-NEXT:    retq
   %3 = or i8 %0, -17
   %4 = icmp eq i8 %3, 0
   %5 = select i1 %4, i8 %0, i8 %3
   ret i8 %5
 }
 
; X64-NEXT:    retq
   %4 = icmp eq i8 %3, 0
   %5 = select i1 %4, i8 %0, i8 %3
   ret i8 %5
 }
 
 define i16 @or_i16_ri(i16 zeroext %0, i16 zeroext %1) {
 ; X86-LABEL: or_i16_ri:
 ; X86:       # %bb.0:
-; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
-; X86-NEXT:    movl %eax, %ecx
-; X86-NEXT:    orl $65519, %ecx # imm = 0xFFEF
-; X86-NEXT:    testw %cx, %cx
-; X86-NEXT:    je .LBB2_2
-; X86-NEXT:  # %bb.1:
-; X86-NEXT:    movl %ecx, %eax
-; X86-NEXT:  .LBB2_2:
+; X86-NEXT:    movl $65519, %eax # imm = 0xFFEF
+; X86-NEXT:    orl {{[0-9]+}}(%esp), %eax
 ; X86-NEXT:    # kill: def $ax killed $ax killed $eax
 ; X86-NEXT:    retl
 ;
 ; X64-LABEL: or_i16_ri:
 ; X64:       # %bb.0:
 ; X64-NEXT:    movl %edi, %eax
 ; X64-NEXT:    orl $65519, %eax # imm = 0xFFEF
-; X64-NEXT:    cmovel %edi, %eax
 ; X64-NEXT:    # kill: def $ax killed $ax killed $eax
 ; X64-NEXT:    retq
   %3 = or i16 %0, -17
   %4 = icmp eq i16 %3, 0
   %5 = select i1 %4, i16 %0, i16 %3
   ret i16 %5
 }
 
; X64-NEXT:    retq
   %5 = select i1 %4, i16 %0, i16 %3
   ret i16 %5
 }
 
 define i32 @or_i32_ri(i32 %0, i32 %1) {
 ; X86-LABEL: or_i32_ri:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
-; X86-NEXT:    movl %eax, %ecx
-; X86-NEXT:    orl $-17, %ecx
-; X86-NEXT:    jle .LBB4_2
-; X86-NEXT:  # %bb.1:
-; X86-NEXT:    movl %ecx, %eax
-; X86-NEXT:  .LBB4_2:
 ; X86-NEXT:    retl
 ;
 ; X64-LABEL: or_i32_ri:
 ; X64:       # %bb.0:
 ; X64-NEXT:    movl %edi, %eax
-; X64-NEXT:    orl $-17, %eax
-; X64-NEXT:    cmovlel %edi, %eax
 ; X64-NEXT:    retq
   %3 = or i32 %0, -17
   %4 = icmp slt i32 %3, 1
   %5 = select i1 %4, i32 %0, i32 %3
   ret i32 %5
 }
 
 define i32 @or_i32_rr(i32 %0, i32 %1) {
; X64-NEXT:    retq
   %4 = icmp slt i32 %3, 1
   %5 = select i1 %4, i32 %0, i32 %3
   ret i32 %5
 }
 
 define i64 @or_i64_ri(i64 %0, i64 %1) nounwind {
 ; X86-LABEL: or_i64_ri:
 ; X86:       # %bb.0:
-; X86-NEXT:    pushl %esi
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
-; X86-NEXT:    movl %eax, %ecx
-; X86-NEXT:    orl $-17, %ecx
-; X86-NEXT:    cmpl $1, %ecx
-; X86-NEXT:    movl $-1, %edx
-; X86-NEXT:    movl $-1, %esi
-; X86-NEXT:    sbbl $0, %esi
-; X86-NEXT:    jl .LBB6_1
-; X86-NEXT:  # %bb.2:
-; X86-NEXT:    movl %ecx, %eax
-; X86-NEXT:    popl %esi
-; X86-NEXT:    retl
-; X86-NEXT:  .LBB6_1:
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %edx
-; X86-NEXT:    popl %esi
 ; X86-NEXT:    retl
 ;
 ; X64-LABEL: or_i64_ri:
 ; X64:       # %bb.0:
 ; X64-NEXT:    movq %rdi, %rax
-; X64-NEXT:    orq $-17, %rax
-; X64-NEXT:    cmovleq %rdi, %rax
 ; X64-NEXT:    retq
   %3 = or i64 %0, -17
   %4 = icmp slt i64 %3, 1
   %5 = select i1 %4, i64 %0, i64 %3
   ret i64 %5
 }
 
 define i64 @or_i64_rr(i64 %0, i64 %1) nounwind {

llvm/test/CodeGen/X86/pr16031.ll

 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc < %s -mtriple=i386-unknown-linux-gnu -mcpu=corei7-avx -enable-misched=false | FileCheck %s
 
 define i64 @main(i1 %tobool1) nounwind {
 ; CHECK-LABEL: main:
 ; CHECK:       # %bb.0: # %entry
-; CHECK-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
-; CHECK-NEXT:    andl $1, %eax
-; CHECK-NEXT:    decl %eax
-; CHECK-NEXT:    orl $-12, %eax
-; CHECK-NEXT:    xorl %ecx, %ecx
-; CHECK-NEXT:    movl %eax, %edx
-; CHECK-NEXT:    addl $-1, %edx
-; CHECK-NEXT:    movl $0, %edx
-; CHECK-NEXT:    adcl $-2, %edx
-; CHECK-NEXT:    cmovsl %ecx, %eax
+; CHECK-NEXT:    xorl %eax, %eax
 ; CHECK-NEXT:    xorl %edx, %edx
 ; CHECK-NEXT:    retl
 entry:
   %0 = zext i1 %tobool1 to i32
   %. = xor i32 %0, 1
   %.21 = select i1 %tobool1, i32 -12, i32 -1
   %conv = sext i32 %.21 to i64
   %1 = add i64 %conv, -1
   %cmp10 = icmp slt i64 %1, 0
   %sub17 = select i1 %cmp10, i64 0, i64 %conv
   ret i64 %sub17
 }

llvm/test/CodeGen/X86/select.ll

 ; GENERIC-LABEL: test2:
 ; GENERIC:       ## %bb.0: ## %entry
 ; GENERIC-NEXT:    pushq %rax
 ; GENERIC-NEXT:    callq _return_false
 ; GENERIC-NEXT:    xorl %ecx, %ecx
 ; GENERIC-NEXT:    testb $1, %al
 ; GENERIC-NEXT:    movl $-3840, %eax ## imm = 0xF100
 ; GENERIC-NEXT:    cmovnel %ecx, %eax
-; GENERIC-NEXT:    cmpl $32768, %eax ## imm = 0x8000
+; GENERIC-NEXT:    cmpl $32767, %eax ## imm = 0x7FFF
 ; GENERIC-NEXT:    jge LBB1_1
 ; GENERIC-NEXT:  ## %bb.2: ## %bb91
 ; GENERIC-NEXT:    xorl %eax, %eax
 ; GENERIC-NEXT:    popq %rcx
 ; GENERIC-NEXT:    retq
 ; GENERIC-NEXT:  LBB1_1: ## %bb90
 ; GENERIC-NEXT:    ud2
 ;
 ; ATOM-LABEL: test2:
 ; ATOM:       ## %bb.0: ## %entry
 ; ATOM-NEXT:    pushq %rax
 ; ATOM-NEXT:    callq _return_false
 ; ATOM-NEXT:    xorl %ecx, %ecx
 ; ATOM-NEXT:    movl $-3840, %edx ## imm = 0xF100
 ; ATOM-NEXT:    testb $1, %al
 ; ATOM-NEXT:    cmovnel %ecx, %edx
-; ATOM-NEXT:    cmpl $32768, %edx ## imm = 0x8000
+; ATOM-NEXT:    cmpl $32767, %edx ## imm = 0x7FFF
 ; ATOM-NEXT:    jge LBB1_1
 ; ATOM-NEXT:  ## %bb.2: ## %bb91
 ; ATOM-NEXT:    xorl %eax, %eax
 ; ATOM-NEXT:    popq %rcx
 ; ATOM-NEXT:    retq
 ; ATOM-NEXT:  LBB1_1: ## %bb90
 ; ATOM-NEXT:    ud2
 ;
 ; ATHLON-LABEL: test2:
 ; ATHLON:       ## %bb.0: ## %entry
 ; ATHLON-NEXT:    subl $12, %esp
 ; ATHLON-NEXT:    calll _return_false
 ; ATHLON-NEXT:    xorl %ecx, %ecx
 ; ATHLON-NEXT:    testb $1, %al
 ; ATHLON-NEXT:    movl $-3840, %eax ## imm = 0xF100
 ; ATHLON-NEXT:    cmovnel %ecx, %eax
-; ATHLON-NEXT:    cmpl $32768, %eax ## imm = 0x8000
+; ATHLON-NEXT:    cmpl $32767, %eax ## imm = 0x7FFF
 ; ATHLON-NEXT:    jge LBB1_1
 ; ATHLON-NEXT:  ## %bb.2: ## %bb91
 ; ATHLON-NEXT:    xorl %eax, %eax
 ; ATHLON-NEXT:    addl $12, %esp
 ; ATHLON-NEXT:    retl
 ; ATHLON-NEXT:  LBB1_1: ## %bb90
 ; ATHLON-NEXT:    ud2
 ;
 ; MCU-LABEL: test2:
 ; MCU:       # %bb.0: # %entry
 ; MCU-NEXT:    calll return_false@PLT
 ; MCU-NEXT:    xorl %ecx, %ecx
 ; MCU-NEXT:    testb $1, %al
 ; MCU-NEXT:    jne .LBB1_2
 ; MCU-NEXT:  # %bb.1: # %entry
 ; MCU-NEXT:    movl $-3840, %ecx # imm = 0xF100
 ; MCU-NEXT:  .LBB1_2: # %entry
-; MCU-NEXT:    cmpl $32768, %ecx # imm = 0x8000
+; MCU-NEXT:    cmpl $32767, %ecx # imm = 0x7FFF
 ; MCU-NEXT:    jge .LBB1_3
 ; MCU-NEXT:  # %bb.4: # %bb91
 ; MCU-NEXT:    xorl %eax, %eax
 ; MCU-NEXT:    retl
 ; MCU-NEXT:  .LBB1_3: # %bb90
 entry:
   %tmp73 = tail call i1 @return_false()
   %g.0 = select i1 %tmp73, i16 0, i16 -480

llvm/test/CodeGen/X86/shrink-compare-pgso.ll

 
 if.end:
   ret void
 }
 
 define dso_local void @test2_1(i32 %X) nounwind !prof !14 {
 ; CHECK-LABEL: test2_1:
 ; CHECK:       # %bb.0: # %entry
-; CHECK-NEXT:    movzbl %dil, %eax
-; CHECK-NEXT:    cmpl $256, %eax # imm = 0x100
+; CHECK-NEXT:    movb $1, %al
+; CHECK-NEXT:    testb %al, %al
 ; CHECK-NEXT:    je bar # TAILCALL
 ; CHECK-NEXT:  # %bb.1: # %if.end
 ; CHECK-NEXT:    retq
 entry:
   %and = and i32 %X, 255
   %cmp = icmp eq i32 %and, 256
   br i1 %cmp, label %if.then, label %if.end
 

llvm/test/CodeGen/X86/shrink-compare.ll

 
 if.end:
   ret void
 }
 
 define dso_local void @test2_1(i32 %X) nounwind minsize {
 ; CHECK-LABEL: test2_1:
 ; CHECK:       # %bb.0: # %entry
-; CHECK-NEXT:    movzbl %dil, %eax
-; CHECK-NEXT:    cmpl $256, %eax # imm = 0x100
+; CHECK-NEXT:    movb $1, %al
+; CHECK-NEXT:    testb %al, %al
 ; CHECK-NEXT:    je bar # TAILCALL
 ; CHECK-NEXT:  # %bb.1: # %if.end
 ; CHECK-NEXT:    retq
 entry:
   %and = and i32 %X, 255
   %cmp = icmp eq i32 %and, 256
   br i1 %cmp, label %if.then, label %if.end