[llvm] r318202 - [CodeGen] Peel off the dominant case in switch statement in lowering

Dan Gohman via llvm-commits llvm-commits at lists.llvm.org
Wed Nov 15 10:04:45 PST 2017


Hello,

wasm's cfg-stackify.ll is a particularly complicated testcase. I'm happy to
update this testcase for you if you'd like.

Dan


On Wed, Nov 15, 2017 at 9:57 AM, Rong Xu <xur at google.com> wrote:

> Thanks for the work-around.
>
> This test is marked as unsupported in my Linux box. I'm not familiar with
> wasm target. Could you send me the instructions to reproduce the error?
>
> Thanks!
>
> -Rong
>
> On Wed, Nov 15, 2017 at 2:57 AM, Ilya Biryukov <ibiryukov at google.com>
> wrote:
>
>> Commited a workaround in r318288.
>>
>> +djg, who authored the original cfg-stackify.ll.
>>
>> On Wed, Nov 15, 2017 at 11:44 AM, Ilya Biryukov <ibiryukov at google.com>
>> wrote:
>>
>>> It seems this change broke "test/CodeGen/WebAssembly/cfg-stackify.ll"
>>> The test obviously passes for me if I set -switch-peel-threshold=101
>>> and disable the optimization. I'm gonna commit this as a quick workaround.
>>> Could you take a look at the test, though? Maybe it's better to update
>>> the CHECK: statements accordingly?
>>>
>>> On Tue, Nov 14, 2017 at 10:44 PM, Rong Xu via llvm-commits <
>>> llvm-commits at lists.llvm.org> wrote:
>>>
>>>> Author: xur
>>>> Date: Tue Nov 14 13:44:09 2017
>>>> New Revision: 318202
>>>>
>>>> URL: http://llvm.org/viewvc/llvm-project?rev=318202&view=rev
>>>> Log:
>>>> [CodeGen] Peel off the dominant case in switch statement in lowering
>>>>
>>>> This patch peels off the top case in switch statement into a branch if
>>>> the
>>>> probability exceeds a threshold. This will help the branch prediction
>>>> and
>>>> avoids the extra compares when lowering into chain of branches.
>>>>
>>>> Differential Revision: http://reviews.llvm.org/D39262
>>>>
>>>> Added:
>>>>     llvm/trunk/test/CodeGen/X86/switch-lower-peel-top-case.ll
>>>> Modified:
>>>>     llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp
>>>>     llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.h
>>>>     llvm/trunk/test/CodeGen/Generic/MachineBranchProb.ll
>>>>     llvm/trunk/test/CodeGen/SystemZ/loop-03.ll
>>>>     llvm/trunk/test/CodeGen/X86/switch-bt.ll
>>>>     llvm/trunk/test/CodeGen/X86/switch.ll
>>>>
>>>> Modified: llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp
>>>> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/S
>>>> electionDAG/SelectionDAGBuilder.cpp?rev=318202&r1=318201&r2=
>>>> 318202&view=diff
>>>> ============================================================
>>>> ==================
>>>> --- llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp
>>>> (original)
>>>> +++ llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp Tue
>>>> Nov 14 13:44:09 2017
>>>> @@ -134,6 +134,12 @@ LimitFPPrecision("limit-float-precision"
>>>>                   cl::location(LimitFloatPrecision),
>>>>                   cl::init(0));
>>>>
>>>> +static cl::opt<unsigned> SwitchPeelThreshold(
>>>> +    "switch-peel-threshold", cl::Hidden, cl::init(66),
>>>> +    cl::desc("Set the case probability threshold for peeling the case
>>>> from a "
>>>> +             "switch statement. A value greater than 100 will void
>>>> this "
>>>> +             "optimization"));
>>>> +
>>>>  // Limit the width of DAG chains. This is important in general to
>>>> prevent
>>>>  // DAG-based analysis from blowing up. For example, alias analysis and
>>>>  // load clustering may not complete in reasonable time. It is
>>>> difficult to
>>>> @@ -9834,6 +9840,74 @@ void SelectionDAGBuilder::splitWorkItem(
>>>>      SwitchCases.push_back(CB);
>>>>  }
>>>>
>>>> +// Scale CaseProb after peeling a case with the probablity of
>>>> PeeledCaseProb
>>>> +// from the swith statement.
>>>> +static BranchProbability scaleCaseProbality(BranchProbability
>>>> CaseProb,
>>>> +                                            BranchProbability
>>>> PeeledCaseProb) {
>>>> +  if (PeeledCaseProb == BranchProbability::getOne())
>>>> +    return BranchProbability::getZero();
>>>> +  BranchProbability SwitchProb = PeeledCaseProb.getCompl();
>>>> +  return BranchProbability(CaseProb.getNumerator(),
>>>> +                           SwitchProb.scale(CaseProb.get
>>>> Denominator()));
>>>> +}
>>>> +
>>>> +// Try to peel the top probability case if it exceeds the threshold.
>>>> +// Return current MachineBasicBlock for the switch statement if the
>>>> peeling
>>>> +// does not occur.
>>>> +// If the peeling is performed, return the newly created
>>>> MachineBasicBlock
>>>> +// for the peeled switch statement. Also update Clusters to remove the
>>>> peeled
>>>> +// case. PeeledCaseProb is the BranchProbability for the peeled case.
>>>> +MachineBasicBlock *SelectionDAGBuilder::peelDominantCaseCluster(
>>>> +    const SwitchInst &SI, CaseClusterVector &Clusters,
>>>> +    BranchProbability &PeeledCaseProb) {
>>>> +  MachineBasicBlock *SwitchMBB = FuncInfo.MBB;
>>>> +  // Don't perform if there is only one cluster or optimizing for size.
>>>> +  if (SwitchPeelThreshold > 100 || !FuncInfo.BPI || Clusters.size() <
>>>> 2 ||
>>>> +      TM.getOptLevel() == CodeGenOpt::None ||
>>>> +      SwitchMBB->getParent()->getFunction()->optForMinSize())
>>>> +    return SwitchMBB;
>>>> +
>>>> +  BranchProbability TopCaseProb = BranchProbability(SwitchPeelThreshold,
>>>> 100);
>>>> +  unsigned PeeledCaseIndex = 0;
>>>> +  bool SwitchPeeled = false;
>>>> +  for (unsigned Index = 0; Index < Clusters.size(); ++Index) {
>>>> +    CaseCluster &CC = Clusters[Index];
>>>> +    if (CC.Prob < TopCaseProb)
>>>> +      continue;
>>>> +    TopCaseProb = CC.Prob;
>>>> +    PeeledCaseIndex = Index;
>>>> +    SwitchPeeled = true;
>>>> +  }
>>>> +  if (!SwitchPeeled)
>>>> +    return SwitchMBB;
>>>> +
>>>> +  DEBUG(dbgs() << "Peeled one top case in switch stmt, prob: " <<
>>>> TopCaseProb
>>>> +               << "\n");
>>>> +
>>>> +  // Record the MBB for the peeled switch statement.
>>>> +  MachineFunction::iterator BBI(SwitchMBB);
>>>> +  ++BBI;
>>>> +  MachineBasicBlock *PeeledSwitchMBB =
>>>> +      FuncInfo.MF->CreateMachineBasicBlock(SwitchMBB->getBasicBloc
>>>> k());
>>>> +  FuncInfo.MF->insert(BBI, PeeledSwitchMBB);
>>>> +
>>>> +  ExportFromCurrentBlock(SI.getCondition());
>>>> +  auto PeeledCaseIt = Clusters.begin() + PeeledCaseIndex;
>>>> +  SwitchWorkListItem W = {SwitchMBB, PeeledCaseIt, PeeledCaseIt,
>>>> +                          nullptr,   nullptr,
>>>> TopCaseProb.getCompl()};
>>>> +  lowerWorkItem(W, SI.getCondition(), SwitchMBB, PeeledSwitchMBB);
>>>> +
>>>> +  Clusters.erase(PeeledCaseIt);
>>>> +  for (CaseCluster &CC : Clusters) {
>>>> +    DEBUG(dbgs() << "Scale the probablity for one cluster, before
>>>> scaling: "
>>>> +                 << CC.Prob << "\n");
>>>> +    CC.Prob = scaleCaseProbality(CC.Prob, TopCaseProb);
>>>> +    DEBUG(dbgs() << "After scaling: " << CC.Prob << "\n");
>>>> +  }
>>>> +  PeeledCaseProb = TopCaseProb;
>>>> +  return PeeledSwitchMBB;
>>>> +}
>>>> +
>>>>  void SelectionDAGBuilder::visitSwitch(const SwitchInst &SI) {
>>>>    // Extract cases from the switch.
>>>>    BranchProbabilityInfo *BPI = FuncInfo.BPI;
>>>> @@ -9887,9 +9961,15 @@ void SelectionDAGBuilder::visitSwitch(co
>>>>      }
>>>>    }
>>>>
>>>> +  // The branch probablity of the peeled case.
>>>> +  BranchProbability PeeledCaseProb = BranchProbability::getZero();
>>>> +  MachineBasicBlock *PeeledSwitchMBB =
>>>> +      peelDominantCaseCluster(SI, Clusters, PeeledCaseProb);
>>>> +
>>>>    // If there is only the default destination, jump there directly.
>>>>    MachineBasicBlock *SwitchMBB = FuncInfo.MBB;
>>>>    if (Clusters.empty()) {
>>>> +    assert(PeeledSwitchMBB == SwitchMBB);
>>>>      SwitchMBB->addSuccessor(DefaultMBB);
>>>>      if (DefaultMBB != NextBlock(SwitchMBB)) {
>>>>        DAG.setRoot(DAG.getNode(ISD::BR, getCurSDLoc(), MVT::Other,
>>>> @@ -9921,8 +10001,14 @@ void SelectionDAGBuilder::visitSwitch(co
>>>>    SwitchWorkList WorkList;
>>>>    CaseClusterIt First = Clusters.begin();
>>>>    CaseClusterIt Last = Clusters.end() - 1;
>>>> -  auto DefaultProb = getEdgeProbability(SwitchMBB, DefaultMBB);
>>>> -  WorkList.push_back({SwitchMBB, First, Last, nullptr, nullptr,
>>>> DefaultProb});
>>>> +  auto DefaultProb = getEdgeProbability(PeeledSwitchMBB, DefaultMBB);
>>>> +  // Scale the branchprobability for DefaultMBB if the peel occurs and
>>>> +  // DefaultMBB is not replaced.
>>>> +  if (PeeledCaseProb != BranchProbability::getZero() &&
>>>> +      DefaultMBB == FuncInfo.MBBMap[SI.getDefaultDest()])
>>>> +    DefaultProb = scaleCaseProbality(DefaultProb, PeeledCaseProb);
>>>> +  WorkList.push_back(
>>>> +      {PeeledSwitchMBB, First, Last, nullptr, nullptr, DefaultProb});
>>>>
>>>>    while (!WorkList.empty()) {
>>>>      SwitchWorkListItem W = WorkList.back();
>>>>
>>>> Modified: llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.h
>>>> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/S
>>>> electionDAG/SelectionDAGBuilder.h?rev=318202&r1=318201&r2=31
>>>> 8202&view=diff
>>>> ============================================================
>>>> ==================
>>>> --- llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.h
>>>> (original)
>>>> +++ llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.h Tue Nov
>>>> 14 13:44:09 2017
>>>> @@ -369,6 +369,10 @@ private:
>>>>                       MachineBasicBlock *SwitchMBB,
>>>>                       MachineBasicBlock *DefaultMBB);
>>>>
>>>> +  /// Peel the top probability case if it exceeds the threshold
>>>> +  MachineBasicBlock *peelDominantCaseCluster(const SwitchInst &SI,
>>>> +                                             CaseClusterVector
>>>> &Clusters,
>>>> +                                             BranchProbability
>>>> &PeeledCaseProb);
>>>>
>>>>    /// A class which encapsulates all of the information needed to
>>>> generate a
>>>>    /// stack protector check and signals to isel via its state being
>>>> initialized
>>>>
>>>> Modified: llvm/trunk/test/CodeGen/Generic/MachineBranchProb.ll
>>>> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/
>>>> Generic/MachineBranchProb.ll?rev=318202&r1=318201&r2=318202&view=diff
>>>> ============================================================
>>>> ==================
>>>> --- llvm/trunk/test/CodeGen/Generic/MachineBranchProb.ll (original)
>>>> +++ llvm/trunk/test/CodeGen/Generic/MachineBranchProb.ll Tue Nov 14
>>>> 13:44:09 2017
>>>> @@ -19,12 +19,15 @@ entry:
>>>>      i64 1, label %sw.bb
>>>>      i64 4, label %sw.bb
>>>>      i64 5, label %sw.bb1
>>>> +    i64 15, label %sw.bb
>>>>    ], !prof !0
>>>>  ; CHECK: BB#0: derived from LLVM BB %entry
>>>> -; CHECK: Successors according to CFG: BB#2({{[0-9a-fx/= ]+}}75.29%)
>>>> BB#4({{[0-9a-fx/= ]+}}24.71%)
>>>> +; CHECK: Successors according to CFG: BB#1({{[0-9a-fx/= ]+}}92.17%)
>>>> BB#4({{[0-9a-fx/= ]+}}7.83%)
>>>>  ; CHECK: BB#4: derived from LLVM BB %entry
>>>> -; CHECK: Successors according to CFG: BB#1({{[0-9a-fx/= ]+}}47.62%)
>>>> BB#5({{[0-9a-fx/= ]+}}52.38%)
>>>> +; CHECK: Successors according to CFG: BB#2({{[0-9a-fx/= ]+}}75.29%)
>>>> BB#5({{[0-9a-fx/= ]+}}24.71%)
>>>>  ; CHECK: BB#5: derived from LLVM BB %entry
>>>> +; CHECK: Successors according to CFG: BB#1({{[0-9a-fx/= ]+}}47.62%)
>>>> BB#6({{[0-9a-fx/= ]+}}52.38%)
>>>> +; CHECK: BB#6: derived from LLVM BB %entry
>>>>  ; CHECK: Successors according to CFG: BB#1({{[0-9a-fx/= ]+}}36.36%)
>>>> BB#3({{[0-9a-fx/= ]+}}63.64%)
>>>>
>>>>  sw.bb:
>>>> @@ -40,7 +43,7 @@ return:
>>>>    ret i32 %retval.0
>>>>  }
>>>>
>>>> -!0 = !{!"branch_weights", i32 7, i32 6, i32 4, i32 4, i32 64}
>>>> +!0 = !{!"branch_weights", i32 7, i32 6, i32 4, i32 4, i32 64, i21 1000}
>>>>
>>>>
>>>>  declare void @g(i32)
>>>>
>>>> Modified: llvm/trunk/test/CodeGen/SystemZ/loop-03.ll
>>>> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/
>>>> SystemZ/loop-03.ll?rev=318202&r1=318201&r2=318202&view=diff
>>>> ============================================================
>>>> ==================
>>>> --- llvm/trunk/test/CodeGen/SystemZ/loop-03.ll (original)
>>>> +++ llvm/trunk/test/CodeGen/SystemZ/loop-03.ll Tue Nov 14 13:44:09 2017
>>>> @@ -3,7 +3,7 @@
>>>>  ; FP128 registers part of the callee saved registers list in order to
>>>> avoid
>>>>  ; spilling / reloading.
>>>>  ;
>>>> -; RUN: llc < %s -mtriple=s390x-linux-gnu -mcpu=z13 | FileCheck %s
>>>> +; RUN: llc -switch-peel-threshold=101 < %s -mtriple=s390x-linux-gnu
>>>> -mcpu=z13 | FileCheck %s
>>>>
>>>>  %0 = type { %0*, %0*, %0*, i32, %1*, i64, i64, i64, i64, i64, i64, %2,
>>>> %5, %7 }
>>>>  %1 = type { i32, i32, i32 (%1*, i64, i32)*, i32 (%1*, i64, i64, i32,
>>>> i8**)*, i32 (%1*, i64, i64, i64, i32)*, i32 (%1*)*, void (i8*)*, i8*, i8* }
>>>>
>>>> Modified: llvm/trunk/test/CodeGen/X86/switch-bt.ll
>>>> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/
>>>> X86/switch-bt.ll?rev=318202&r1=318201&r2=318202&view=diff
>>>> ============================================================
>>>> ==================
>>>> --- llvm/trunk/test/CodeGen/X86/switch-bt.ll (original)
>>>> +++ llvm/trunk/test/CodeGen/X86/switch-bt.ll Tue Nov 14 13:44:09 2017
>>>> @@ -1,4 +1,4 @@
>>>> -; RUN: llc -mtriple=x86_64-- -asm-verbose=false < %s
>>>> -jump-table-density=40 | FileCheck %s
>>>> +; RUN: llc -mtriple=x86_64-- -asm-verbose=false < %s
>>>> -jump-table-density=40 -switch-peel-threshold=101 | FileCheck %s
>>>>
>>>>  ; This switch should use bit tests, and the third bit test case is just
>>>>  ; testing for one possible value, so it doesn't need a bt.
>>>>
>>>> Added: llvm/trunk/test/CodeGen/X86/switch-lower-peel-top-case.ll
>>>> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/
>>>> X86/switch-lower-peel-top-case.ll?rev=318202&view=auto
>>>> ============================================================
>>>> ==================
>>>> --- llvm/trunk/test/CodeGen/X86/switch-lower-peel-top-case.ll (added)
>>>> +++ llvm/trunk/test/CodeGen/X86/switch-lower-peel-top-case.ll Tue Nov
>>>> 14 13:44:09 2017
>>>> @@ -0,0 +1,135 @@
>>>> +; RUN: llc -stop-after=isel < %s  | FileCheck %s
>>>> +
>>>> +define i32 @foo(i32 %n) !prof !1 {
>>>> +entry:
>>>> +  switch i32 %n, label %bb_default [
>>>> +    i32 8, label %bb1
>>>> +    i32 -8826, label %bb2
>>>> +    i32 18312, label %bb3
>>>> +    i32 18568, label %bb4
>>>> +    i32 129, label %bb5
>>>> +  ], !prof !2
>>>> +
>>>> +; CHECK: successors: %[[PEELED_CASE_LABEL:.*]](0x5999999a),
>>>> %[[PEELED_SWITCH_LABEL:.*]](0x26666666)
>>>> +; CHECK:    %[[VAL:[0-9]+]]:gr32 = COPY %edi
>>>> +; CHECK:    %{{[0-9]+}}:gr32 = SUB32ri %[[VAL]], 18568, implicit-def
>>>> %eflags
>>>> +; CHECK:    JE_1 %[[PEELED_CASE_LABEL]], implicit %eflags
>>>> +; CHECK:    JMP_1 %[[PEELED_SWITCH_LABEL]]
>>>> +; CHECK:  [[PEELED_SWITCH_LABEL]]:
>>>> +; CHECK:    successors: %[[BB1_LABEL:.*]](0x0206d3a0),
>>>> %[[BB2_LABEL:.*]](0x7df92c60)
>>>> +; CHECK:    %{{[0-9]+}}:gr32 = SUB32ri %[[VAL]], 18311, implicit-def
>>>> %eflags
>>>> +; CHECK:    JG_1 %[[BB2_LABEL]], implicit %eflags
>>>> +; CHECK:    JMP_1 %[[BB1_LABEL]]
>>>> +; CHECK:  [[BB1_LABEL]]:
>>>> +; CHECK:    successors: %[[CASE2_LABEL:.*]](0x35e50d5b),
>>>> %[[BB3_LABEL:.*]](0x4a1af2a5)
>>>> +; CHECK:    %{{[0-9]+}}:gr32 = SUB32ri %[[VAL]], -8826, implicit-def
>>>> %eflags
>>>> +; CHECK:    JE_1 %[[CASE2_LABEL]], implicit %eflags
>>>> +; CHECK:    JMP_1 %[[BB3_LABEL]]
>>>> +; CHECK:  [[BB3_LABEL]]
>>>> +; CHECK:    successors: %[[CASE5_LABEL:.*]](0x45d173c8),
>>>> %[[BB4_LABEL:.*]](0x3a2e8c38)
>>>> +; CHECK:    %{{[0-9]+}}:gr32 = SUB32ri %[[VAL]], 129, implicit-def
>>>> %eflags
>>>> +; CHECK:    JE_1 %[[CASE5_LABEL]], implicit %eflags
>>>> +; CHECK:    JMP_1 %[[BB4_LABEL]]
>>>> +; CHECK:  [[BB4_LABEL:.*]]:
>>>> +; CHECK:    successors: %[[CASE1_LABEL:.*]](0x66666666),
>>>> %[[DEFAULT_BB_LABEL:.*]](0x1999999a)
>>>> +; CHECK:    %{{[0-9]+}}:gr32 = SUB32ri8 %[[VAL]], 8, implicit-def
>>>> %eflags
>>>> +; CHECK:    JE_1 %[[CASE1_LABEL]], implicit %eflags
>>>> +; CHECK:    JMP_1 %[[DEFAULT_BB_LABEL]]
>>>> +; CHECK:  [[BB2_LABEL]]:
>>>> +; CHECK:    successors: %[[CASE3_LABEL:.*]](0x7fe44107),
>>>> %[[DEFAULT_BB_LABEL]](0x001bbef9)
>>>> +; CHECK:    %{{[0-9]+}}:gr32 = SUB32ri %[[VAL]], 18312, implicit-def
>>>> %eflags
>>>> +; CHECK:    JE_1 %[[CASE3_LABEL]], implicit %eflags
>>>> +; CHECK:    JMP_1 %[[DEFAULT_BB_LABEL]]
>>>> +
>>>> +bb1:
>>>> +  br label %return
>>>> +bb2:
>>>> +  br label %return
>>>> +bb3:
>>>> +  br label %return
>>>> +bb4:
>>>> +  br label %return
>>>> +bb5:
>>>> +  br label %return
>>>> +bb_default:
>>>> +  br label %return
>>>> +
>>>> +return:
>>>> +  %retval = phi i32 [ 0, %bb_default ], [ 5, %bb5 ], [ 4, %bb4 ], [ 3,
>>>> %bb3 ], [ 2, %bb2 ], [ 1, %bb1 ]
>>>> +  ret i32 %retval
>>>> +}
>>>> +
>>>> +; Test the peeling of the merged cases value 85 and 86.
>>>> +define i32 @foo1(i32 %n) !prof !1 {
>>>> +entry:
>>>> +  switch i32 %n, label %bb_default [
>>>> +    i32 -40, label %bb1
>>>> +    i32 86, label %bb2
>>>> +    i32 85, label %bb2
>>>> +    i32 1, label %bb3
>>>> +    i32 5, label %bb4
>>>> +    i32 7, label %bb5
>>>> +    i32 49, label %bb6
>>>> +  ], !prof !3
>>>> +
>>>> +; CHECK:   successors: %[[PEELED_CASE_LABEL:.*]](0x59999999),
>>>> %[[PEELED_SWITCH_LABEL:.*]](0x26666667)
>>>> +; CHECK:   %[[VAL:[0-9]+]]:gr32 = COPY %edi
>>>> +; CHECK:   %{{[0-9]+}}:gr32 = ADD32ri8 %{{[0-9]+}}, -85, implicit-def
>>>> dead %eflags
>>>> +; CHECK:   %{{[0-9]+}}:gr32 = SUB32ri8 %{{[0-9]+}}, 2, implicit-def
>>>> %eflags
>>>> +; CHECK:   JB_1 %[[PEELED_CASE_LABEL]], implicit %eflags
>>>> +; CHECK:   JMP_1 %[[PEELED_SWITCH_LABEL]]
>>>> +; CHECK: [[PEELED_SWITCH_LABEL]]:
>>>> +; CHECK:    successors: %[[BB1_LABEL:.*]](0x0088888a),
>>>> %[[BB2_LABEL:.*]](0x7f777776)
>>>> +; CHECK:    %{{[0-9]+}}:gr32 = SUB32ri8 %[[VAL]], 4, implicit-def
>>>> %eflags
>>>> +; CHECK:    JG_1 %[[BB2_LABEL]], implicit %eflags
>>>> +; CHECK:    JMP_1 %[[BB1_LABEL]]
>>>> +; CHECK:  [[BB1_LABEL]]:
>>>> +; CHECK:    successors: %[[CASE4_LABEL:.*]](0x7f775a4f),
>>>> %[[BB3_LABEL:.*]](0x0088a5b1)
>>>> +; CHECK:    %{{[0-9]+}}:gr32 = SUB32ri8 %[[VAL]], 1, implicit-def
>>>> %eflags
>>>> +; CHECK:    JE_1 %[[CASE4_LABEL]], implicit %eflags
>>>> +; CHECK:    JMP_1 %[[BB3_LABEL]]
>>>> +; CHECK:  [[BB3_LABEL]]:
>>>> +; CHECK:    successors: %[[CASE1_LABEL:.*]](0x66666666),
>>>> %[[DEFAULT_BB_LABEL:.*]](0x1999999a)
>>>> +; CHECK:    %{{[0-9]+}}:gr32 = SUB32ri8 %[[VAL]], -40, implicit-def
>>>> %eflags
>>>> +; CHECK:    JE_1 %[[CASE1_LABEL]], implicit %eflags
>>>> +; CHECK:    JMP_1 %[[DEFAULT_BB_LABEL]]
>>>> +; CHECK:  [[BB2_LABEL]]:
>>>> +; CHECK:    successors: %[[CASE5_LABEL:.*]](0x00000000),
>>>> %[[BB4_LABEL:.*]](0x80000000)
>>>> +; CHECK:    %{{[0-9]+}}:gr32 = SUB32ri8 %[[VAL]], 5, implicit-def
>>>> %eflags
>>>> +; CHECK:    JE_1 %[[CASE5_LABEL]], implicit %eflags
>>>> +; CHECK:    JMP_1 %[[BB4_LABEL]]
>>>> +; CHECK:  [[BB4_LABEL]]:
>>>> +; CHECK:    successors: %[[CASE6_LABEL:.*]](0x00000000),
>>>> %[[BB5_LABEL:.*]](0x80000000)
>>>> +; CHECK:    %{{[0-9]+}}:gr32 = SUB32ri8 %[[VAL]], 7, implicit-def
>>>> %eflags
>>>> +; CHECK:    JE_1 %[[CASE6_LABEL]], implicit %eflags
>>>> +; CHECK:    JMP_1 %[[BB5_LABEL]]
>>>> +; CHECK:  [[BB5_LABEL]]:
>>>> +; CHECK:    successors: %[[CASE7_LABEL:.*]](0x00000000),
>>>> %[[DEFAULT_BB_LABEL]](0x80000000)
>>>> +; CHECK:    %{{[0-9]+}}:gr32 = SUB32ri8 %[[VAL]], 49, implicit-def
>>>> %eflags
>>>> +; CHECK:    JE_1 %[[CASE7_LABEL]], implicit %eflags
>>>> +; CHECK:    JMP_1 %[[DEFAULT_BB_LABEL]]
>>>> +
>>>> +
>>>> +bb1:
>>>> +  br label %return
>>>> +bb2:
>>>> +  br label %return
>>>> +bb3:
>>>> +  br label %return
>>>> +bb4:
>>>> +  br label %return
>>>> +bb5:
>>>> +  br label %return
>>>> +bb6:
>>>> +  br label %return
>>>> +bb_default:
>>>> +  br label %return
>>>> +
>>>> +return:
>>>> +  %retval = phi i32 [ 0, %bb_default ], [ 6, %bb6 ], [ 5, %bb5 ], [ 4,
>>>> %bb4 ], [ 3, %bb3 ], [ 2, %bb2 ], [ 1, %bb1 ]
>>>> +  ret i32 %retval
>>>> +}
>>>> +!1 = !{!"function_entry_count", i64 100000}
>>>> +!2 = !{!"branch_weights", i32 50, i32 100, i32 200, i32 29500, i32
>>>> 70000, i32 150}
>>>> +!3 = !{!"branch_weights", i32 50, i32 100, i32 500, i32 69500, i32
>>>> 29850, i32 0, i32 0, i32 0}
>>>> +
>>>>
>>>> Modified: llvm/trunk/test/CodeGen/X86/switch.ll
>>>> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/
>>>> X86/switch.ll?rev=318202&r1=318201&r2=318202&view=diff
>>>> ============================================================
>>>> ==================
>>>> --- llvm/trunk/test/CodeGen/X86/switch.ll (original)
>>>> +++ llvm/trunk/test/CodeGen/X86/switch.ll Tue Nov 14 13:44:09 2017
>>>> @@ -1,4 +1,4 @@
>>>> -; RUN: llc -mtriple=x86_64-linux-gnu %s -o - -jump-table-density=40
>>>> -verify-machineinstrs | FileCheck %s
>>>> +; RUN: llc -mtriple=x86_64-linux-gnu %s -o - -jump-table-density=40
>>>> -switch-peel-threshold=101 -verify-machineinstrs | FileCheck %s
>>>>  ; RUN: llc -mtriple=x86_64-linux-gnu %s -o - -O0
>>>> -jump-table-density=40 -verify-machineinstrs | FileCheck
>>>> --check-prefix=NOOPT %s
>>>>
>>>>  declare void @g(i32)
>>>>
>>>>
>>>> _______________________________________________
>>>> llvm-commits mailing list
>>>> llvm-commits at lists.llvm.org
>>>> http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-commits
>>>>
>>>
>>>
>>>
>>> --
>>> Regards,
>>> Ilya Biryukov
>>>
>>
>>
>>
>> --
>> Regards,
>> Ilya Biryukov
>>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20171115/f4fcb74b/attachment-0001.html>


More information about the llvm-commits mailing list