[OpenMPIRBuilder] Remove ContinuationBB argument from Body callback.

[Meinersbur]

Summary

The callback is expected to create a branch to the ContinuationBB (sometimes called FiniBB in some lambdas) argument when finishing. This creates problems:

1. The InsertPoint used for CodeGenIP does not need to be the end of a block. If it is not, a naive callback will insert a branch instruction into the middle of the block.
2. The BasicBlock the CodeGenIP is pointing to may or may not have a terminator. There is an conflict where to branch to if the block already has a terminator.
3. Some API functions work only with block having a terminator. Some workarounds have been used to insert a temporary terminator that is removed again.
4. Some callbacks are sensitive to whether the BasicBlock has a terminator or not. This creates a callback ordering problem where different callback may have different behaviour depending on whether a previous callback created a terminator or not. The problem also exists for FinalizeCallbackTy where some callbacks do create branch to another "continue" block, but unlike BodyGenCallbackTy does not receive the target as argument. This is not addressed in this patch.

With this patch, the callback receives an CodeGenIP into a BasicBlock where to insert instructions. If it has to insert control flow, it can split the block at that position as needed but otherwise no separate ContinuationBB is needed. In particular, a callback can be empty without breaking the emitted IR. If the caller needs the control flow to branch to a specific target, it can insert the branch instruction itself and pass an InsertPoint before the terminator to the callback.

Certain frontends such as Clang may expect the current IRBuilder position to be at the end of a basic block. In this case its callbacks must split the block at CodeGenIP before setting the IRBuilder position such that the instructions after CodeGenIP are moved to another basic block and before returning create a new branch instruction to the split block.

Some utility functions such as splitBB are supporting correct splitting of BasicBlocks, independent of whether they have a terminator or not, returning/setting the InsertPoint of an IRBuilder to the end of split predecessor block, and optionally omitting creating a branch to the split successor block to be added later.

[Meinersbur]

ping

[Meinersbur]

ping

[kiranchandramohan]

Thanks @Meinersbur for this patch. Apologies for the delay in reviewing many of your patches.

The patch looks mostly LGTM. A few nits. The patch probably needs a rebase since there are no dependencies now and a few more construct lowering has landed in MLIR (single, simdloop).

clang/lib/CodeGen/CGStmtOpenMP.cpp

5603–5605

Why is it not possible to use OMPBuilderCBHelpers::EmitOMPInlinedRegionBody here?

clang/lib/CodeGen/CodeGenFunction.h

1825

Nit:finalize

llvm/include/llvm/Frontend/OpenMP/OMPIRBuilder.h

26–58

I guess these are already in from your other patch (https://reviews.llvm.org/D114413).

147

Nit:itself

llvm/lib/Frontend/OpenMP/OMPIRBuilder.cpp

1301–1307

Nit:Unexpected

mlir/lib/Target/LLVMIR/Dialect/OpenMP/OpenMPToLLVMIRTranslation.cpp

127

Would this "omp.region.cont" be a mostly empty block in most cases? I guess it is not a problem since llvm will remove these blocks.

The IR generated for

omp.parallel {
  omp.barrier
  omp.terminator
}

is the following. Notice the empty (only a branch) omp.region.cont block. Previously we had this only at the source side omp.par.region.

omp.par.entry:
  %tid.addr.local = alloca i32, align 4
  %0 = load i32, i32* %tid.addr, align 4
  store i32 %0, i32* %tid.addr.local, align 4
  %tid = load i32, i32* %tid.addr.local, align 4
  br label %omp.par.region

omp.par.region:                                   ; preds = %omp.par.entry
  br label %omp.par.region1

omp.par.region1:                                  ; preds = %omp.par.region
  %omp_global_thread_num2 = call i32 @__kmpc_global_thread_num(%struct.ident_t* @4)
  call void @__kmpc_barrier(%struct.ident_t* @3, i32 %omp_global_thread_num2)
  br label %omp.region.cont, !dbg !12

omp.region.cont:                                  ; preds = %omp.par.region1
  br label %omp.par.pre_finalize

omp.par.pre_finalize:                             ; preds = %omp.region.cont
  br label %omp.par.outlined.exit.exitStub

openmp/runtime/test/ompt/cancel/cancel_parallel.c

1 ↗

(On Diff #403787)

Nit: unrelated change?

[Meinersbur]

• Rebase
• Address review

clang/lib/CodeGen/CGStmtOpenMP.cpp

5603–5605

EmitOMPInlinedRegionBody calls splitBBWithSuffix, which has already been done in this lambda.

llvm/include/llvm/Frontend/OpenMP/OMPIRBuilder.h

26–58

Yes, now public because also used in CodegenOpenMP.

These utility functions avoid all the workarounds that splitBasicBlock/SplitBlock do not all degenerate blocks, such as inserting temporary terminators or ContinuationBB (There may be insertion points referencing those temporary terminators!!).

mlir/lib/Target/LLVMIR/Dialect/OpenMP/OpenMPToLLVMIRTranslation.cpp

127

Empty BBs should not be considered an issue since they are removed by SimplifyCFG anyway. Why would it be? Correctness should be the priority.

What makes the current code so fragile are the conditions that depend on the current insert point. Invoking it in a different manner causes an untested code path to be taken. For instance, an insert point becomes invalided that did not under any previous test because nobody yet inserted new BBs in a callback. One if condition inside the callback and everything breaks.

[kiranchandramohan]

LGTM. You may wait for a day in case there are comments from other reviewers.

I think the CI is complaining about some clang-format issue in clang/test/OpenMP/critical_codegen_attr.cpp

[Meinersbur]

In D118409#3474131, @kiranchandramohan wrote:

I think the CI is complaining about some clang-format issue in clang/test/OpenMP/critical_codegen_attr.cpp

clang-format should not be applied on Clang tests, many of them have significant whitespace (eg. the preprocessor tests). In this case, clang-suggests to indent the // CHECK: lines by 2 spaces (to match the indentation of the statements), which seems unnecessary.

[jdoerfert]

llvm/include/llvm/Frontend/OpenMP/OMPIRBuilder.h

70

Argh... Why do we need to copy llvm functionality instead of making it work for our use case?
Nit: no llvm::.

clang/lib/CodeGen/CGStmtOpenMP.cpp

std::string CodeGenFunction::OMPBuilderCBHelpers::getNameWithSeparators(
   llvm::raw_svector_ostream OS(Buffer);
   StringRef Sep = FirstSeparator;
   for (StringRef Part : Parts) {
     OS << Sep << Part;
     Sep = Separator;
   }
   return OS.str().str();
 }
+
+void CodeGenFunction::OMPBuilderCBHelpers::EmitOMPInlinedRegionBody(
+    CodeGenFunction &CGF, const Stmt *RegionBodyStmt, InsertPointTy AllocaIP,
+    InsertPointTy CodeGenIP, Twine RegionName) {
+  CGBuilderTy &Builder = CGF.Builder;
+  Builder.restoreIP(CodeGenIP);
+  llvm::BasicBlock *FiniBB = splitBBWithSuffix(Builder, /*CreateBranch=*/false,
+                                               "." + RegionName + ".after");
+
+  {
+    OMPBuilderCBHelpers::InlinedRegionBodyRAII IRB(CGF, AllocaIP, *FiniBB);
+    CGF.EmitStmt(RegionBodyStmt);
+  }
+
+  if (Builder.saveIP().isSet())
+    Builder.CreateBr(FiniBB);
+}
+
+void CodeGenFunction::OMPBuilderCBHelpers::EmitOMPOutlinedRegionBody(
+    CodeGenFunction &CGF, const Stmt *RegionBodyStmt, InsertPointTy AllocaIP,
+    InsertPointTy CodeGenIP, Twine RegionName) {
+  CGBuilderTy &Builder = CGF.Builder;
+  Builder.restoreIP(CodeGenIP);
+  llvm::BasicBlock *FiniBB = splitBBWithSuffix(Builder, /*CreateBranch=*/false,
+                                               "." + RegionName + ".after");
+
+  {
+    OMPBuilderCBHelpers::OutlinedRegionBodyRAII IRB(CGF, AllocaIP, *FiniBB);
+    CGF.EmitStmt(RegionBodyStmt);
+  }
+
+  if (Builder.saveIP().isSet())
+    Builder.CreateBr(FiniBB);
+}
+
 void CodeGenFunction::EmitOMPParallelDirective(const OMPParallelDirective &S) {
   if (CGM.getLangOpts().OpenMPIRBuilder) {
     llvm::OpenMPIRBuilder &OMPBuilder = CGM.getOpenMPRuntime().getOMPBuilder();
     // Check if we have any if clause associated with the directive.
     llvm::Value *IfCond = nullptr;
     if (const auto *C = S.getSingleClause<OMPIfClause>())
       IfCond = EmitScalarExpr(C->getCondition(),
                               /*IgnoreResultAssign=*/true);
auto PrivCB = [](InsertPointTy AllocaIP, InsertPointTy CodeGenIP,
       ReplVal = &Val;
 
       return CodeGenIP;
     };
 
     const CapturedStmt *CS = S.getCapturedStmt(OMPD_parallel);
     const Stmt *ParallelRegionBodyStmt = CS->getCapturedStmt();
 
-    auto BodyGenCB = [ParallelRegionBodyStmt,
-                      this](InsertPointTy AllocaIP, InsertPointTy CodeGenIP,
-                            llvm::BasicBlock &ContinuationBB) {
-      OMPBuilderCBHelpers::OutlinedRegionBodyRAII ORB(*this, AllocaIP,
-                                                      ContinuationBB);
-      OMPBuilderCBHelpers::EmitOMPRegionBody(*this, ParallelRegionBodyStmt,
-                                             CodeGenIP, ContinuationBB);
+    auto BodyGenCB = [&, this](InsertPointTy AllocaIP,
+                               InsertPointTy CodeGenIP) {
+      OMPBuilderCBHelpers::EmitOMPOutlinedRegionBody(
+          *this, ParallelRegionBodyStmt, AllocaIP, CodeGenIP, "parallel");
     };
 
     CGCapturedStmtInfo CGSI(*CS, CR_OpenMP);
     CodeGenFunction::CGCapturedStmtRAII CapInfoRAII(*this, &CGSI);
     llvm::OpenMPIRBuilder::InsertPointTy AllocaIP(
         AllocaInsertPt->getParent(), AllocaInsertPt->getIterator());
     Builder.restoreIP(
         OMPBuilder.createParallel(Builder, AllocaIP, BodyGenCB, PrivCB, FiniCB,
if (CGM.getLangOpts().OpenMPIRBuilder) {
 
     const CapturedStmt *ICS = S.getInnermostCapturedStmt();
     const Stmt *CapturedStmt = S.getInnermostCapturedStmt()->getCapturedStmt();
     const auto *CS = dyn_cast<CompoundStmt>(CapturedStmt);
     llvm::SmallVector<BodyGenCallbackTy, 4> SectionCBVector;
     if (CS) {
       for (const Stmt *SubStmt : CS->children()) {
         auto SectionCB = [this, SubStmt](InsertPointTy AllocaIP,
-                                         InsertPointTy CodeGenIP,
-                                         llvm::BasicBlock &FiniBB) {
-          OMPBuilderCBHelpers::InlinedRegionBodyRAII IRB(*this, AllocaIP,
-                                                         FiniBB);
-          OMPBuilderCBHelpers::EmitOMPRegionBody(*this, SubStmt, CodeGenIP,
-                                                 FiniBB);
+                                         InsertPointTy CodeGenIP) {
+          OMPBuilderCBHelpers::EmitOMPInlinedRegionBody(
+              *this, SubStmt, AllocaIP, CodeGenIP, "section");
         };
         SectionCBVector.push_back(SectionCB);
       }
     } else {
       auto SectionCB = [this, CapturedStmt](InsertPointTy AllocaIP,
-                                            InsertPointTy CodeGenIP,
-                                            llvm::BasicBlock &FiniBB) {
-        OMPBuilderCBHelpers::InlinedRegionBodyRAII IRB(*this, AllocaIP, FiniBB);
-        OMPBuilderCBHelpers::EmitOMPRegionBody(*this, CapturedStmt, CodeGenIP,
-                                               FiniBB);
+                                            InsertPointTy CodeGenIP) {
+        OMPBuilderCBHelpers::EmitOMPInlinedRegionBody(
+            *this, CapturedStmt, AllocaIP, CodeGenIP, "section");
       };
       SectionCBVector.push_back(SectionCB);
     }
 
     // Privatization callback that performs appropriate action for
     // shared/private/firstprivate/lastprivate/copyin/... variables.
     //
     // TODO: This defaults to shared right now.
if (CGM.getLangOpts().OpenMPIRBuilder) {
     using InsertPointTy = llvm::OpenMPIRBuilder::InsertPointTy;
 
     const Stmt *SectionRegionBodyStmt = S.getAssociatedStmt();
     auto FiniCB = [this](InsertPointTy IP) {
       OMPBuilderCBHelpers::FinalizeOMPRegion(*this, IP);
     };
 
     auto BodyGenCB = [SectionRegionBodyStmt, this](InsertPointTy AllocaIP,
-                                                   InsertPointTy CodeGenIP,
-                                                   llvm::BasicBlock &FiniBB) {
-      OMPBuilderCBHelpers::InlinedRegionBodyRAII IRB(*this, AllocaIP, FiniBB);
-      OMPBuilderCBHelpers::EmitOMPRegionBody(*this, SectionRegionBodyStmt,
-                                             CodeGenIP, FiniBB);
+                                                   InsertPointTy CodeGenIP) {
+      OMPBuilderCBHelpers::EmitOMPInlinedRegionBody(
+          *this, SectionRegionBodyStmt, AllocaIP, CodeGenIP, "section");
     };
 
     LexicalScope Scope(*this, S.getSourceRange());
     EmitStopPoint(&S);
     Builder.restoreIP(OMPBuilder.createSection(Builder, BodyGenCB, FiniCB));
 
     return;
   }
if (CGM.getLangOpts().OpenMPIRBuilder) {
 
     const Stmt *MasterRegionBodyStmt = S.getAssociatedStmt();
 
     auto FiniCB = [this](InsertPointTy IP) {
       OMPBuilderCBHelpers::FinalizeOMPRegion(*this, IP);
     };
 
     auto BodyGenCB = [MasterRegionBodyStmt, this](InsertPointTy AllocaIP,
-                                                  InsertPointTy CodeGenIP,
-                                                  llvm::BasicBlock &FiniBB) {
-      OMPBuilderCBHelpers::InlinedRegionBodyRAII IRB(*this, AllocaIP, FiniBB);
-      OMPBuilderCBHelpers::EmitOMPRegionBody(*this, MasterRegionBodyStmt,
-                                             CodeGenIP, FiniBB);
+                                                  InsertPointTy CodeGenIP) {
+      OMPBuilderCBHelpers::EmitOMPInlinedRegionBody(
+          *this, MasterRegionBodyStmt, AllocaIP, CodeGenIP, "master");
     };
 
     LexicalScope Scope(*this, S.getSourceRange());
     EmitStopPoint(&S);
     Builder.restoreIP(OMPBuilder.createMaster(Builder, BodyGenCB, FiniCB));
 
     return;
   }
llvm::Value *FilterVal = Filter
                                  ? EmitScalarExpr(Filter, CGM.Int32Ty)
                                  : llvm::ConstantInt::get(CGM.Int32Ty, /*V=*/0);
 
     auto FiniCB = [this](InsertPointTy IP) {
       OMPBuilderCBHelpers::FinalizeOMPRegion(*this, IP);
     };
 
     auto BodyGenCB = [MaskedRegionBodyStmt, this](InsertPointTy AllocaIP,
-                                                  InsertPointTy CodeGenIP,
-                                                  llvm::BasicBlock &FiniBB) {
-      OMPBuilderCBHelpers::InlinedRegionBodyRAII IRB(*this, AllocaIP, FiniBB);
-      OMPBuilderCBHelpers::EmitOMPRegionBody(*this, MaskedRegionBodyStmt,
-                                             CodeGenIP, FiniBB);
+                                                  InsertPointTy CodeGenIP) {
+      OMPBuilderCBHelpers::EmitOMPInlinedRegionBody(
+          *this, MaskedRegionBodyStmt, AllocaIP, CodeGenIP, "masked");
     };
 
     LexicalScope Scope(*this, S.getSourceRange());
     EmitStopPoint(&S);
     Builder.restoreIP(
         OMPBuilder.createMasked(Builder, BodyGenCB, FiniCB, FilterVal));
 
     return;
if (Hint)
       HintInst =
           Builder.CreateIntCast(EmitScalarExpr(Hint), CGM.Int32Ty, false);
 
     auto FiniCB = [this](InsertPointTy IP) {
       OMPBuilderCBHelpers::FinalizeOMPRegion(*this, IP);
     };
 
     auto BodyGenCB = [CriticalRegionBodyStmt, this](InsertPointTy AllocaIP,
-                                                    InsertPointTy CodeGenIP,
-                                                    llvm::BasicBlock &FiniBB) {
-      OMPBuilderCBHelpers::InlinedRegionBodyRAII IRB(*this, AllocaIP, FiniBB);
-      OMPBuilderCBHelpers::EmitOMPRegionBody(*this, CriticalRegionBodyStmt,
-                                             CodeGenIP, FiniBB);
+                                                    InsertPointTy CodeGenIP) {
+      OMPBuilderCBHelpers::EmitOMPInlinedRegionBody(
+          *this, CriticalRegionBodyStmt, AllocaIP, CodeGenIP, "critical");
     };
 
     LexicalScope Scope(*this, S.getSourceRange());
     EmitStopPoint(&S);
     Builder.restoreIP(OMPBuilder.createCritical(
         Builder, BodyGenCB, FiniCB, S.getDirectiveName().getAsString(),
         HintInst));
 
if (S.hasClausesOfKind<OMPDependClause>()) {
       // Without clause, it behaves as if the threads clause is specified.
       const auto *C = S.getSingleClause<OMPSIMDClause>();
 
       auto FiniCB = [this](InsertPointTy IP) {
         OMPBuilderCBHelpers::FinalizeOMPRegion(*this, IP);
       };
 
       auto BodyGenCB = [&S, C, this](InsertPointTy AllocaIP,
-                                     InsertPointTy CodeGenIP,
-                                     llvm::BasicBlock &FiniBB) {
+                                     InsertPointTy CodeGenIP) {
+        Builder.restoreIP(CodeGenIP);
+
         const CapturedStmt *CS = S.getInnermostCapturedStmt();
         if (C) {
+          llvm::BasicBlock *FiniBB = splitBBWithSuffix(
+              Builder, /*CreateBranch=*/false, ".ordered.after");
           llvm::SmallVector<llvm::Value *, 16> CapturedVars;
           GenerateOpenMPCapturedVars(*CS, CapturedVars);
           llvm::Function *OutlinedFn =
               emitOutlinedOrderedFunction(CGM, CS, S.getBeginLoc());
           assert(S.getBeginLoc().isValid() &&
                  "Outlined function call location must be valid.");
           ApplyDebugLocation::CreateDefaultArtificial(*this, S.getBeginLoc());
-          OMPBuilderCBHelpers::EmitCaptureStmt(*this, CodeGenIP, FiniBB,
+          OMPBuilderCBHelpers::EmitCaptureStmt(*this, CodeGenIP, *FiniBB,
                                                OutlinedFn, CapturedVars);
         } else {
-          OMPBuilderCBHelpers::InlinedRegionBodyRAII IRB(*this, AllocaIP,
-                                                         FiniBB);
-          OMPBuilderCBHelpers::EmitOMPRegionBody(*this, CS->getCapturedStmt(),
-                                                 CodeGenIP, FiniBB);
+          OMPBuilderCBHelpers::EmitOMPInlinedRegionBody(
+              *this, CS->getCapturedStmt(), AllocaIP, CodeGenIP, "ordered");
         }
       };
 
       OMPLexicalScope Scope(*this, S, OMPD_unknown);
       Builder.restoreIP(
           OMPBuilder.createOrderedThreadsSimd(Builder, BodyGenCB, FiniCB, !C));
     }
     return;
   }

clang/lib/CodeGen/CodeGenFunction.h

static void FinalizeOMPRegion(CodeGenFunction &CGF, InsertPointTy IP) {
       // erase and replace with cleanup branch.
       IPBB->getTerminator()->eraseFromParent();
       CGF.Builder.SetInsertPoint(IPBB);
       CodeGenFunction::JumpDest Dest = CGF.getJumpDestInCurrentScope(DestBB);
       CGF.EmitBranchThroughCleanup(Dest);
     }
 
     /// Emit the body of an OMP region
-    /// \param CGF  The Codegen function this belongs to
-    /// \param RegionBodyStmt The body statement for the OpenMP region being
-    ///        generated
-    /// \param CodeGenIP  Insertion point for generating the body code.
-    /// \param FiniBB The finalization basic block
-    static void EmitOMPRegionBody(CodeGenFunction &CGF,
+    /// \param CGF            The Codegen function this belongs to
+    /// \param RegionBodyStmt The body statement for the OpenMP region being
+    ///                       generated
+    /// \param AllocaIP       Where to insert alloca instructions
+    /// \param CodeGenIP      Where to insert the region code
+    /// \param RegionName     Name to be used for new blocks
+    static void EmitOMPInlinedRegionBody(CodeGenFunction &CGF,
                                          const Stmt *RegionBodyStmt,
+                                         InsertPointTy AllocaIP,
                                          InsertPointTy CodeGenIP,
-                                  llvm::BasicBlock &FiniBB) {
-      llvm::BasicBlock *CodeGenIPBB = CodeGenIP.getBlock();
-      if (llvm::Instruction *CodeGenIPBBTI = CodeGenIPBB->getTerminator())
-        CodeGenIPBBTI->eraseFromParent();
-
-      CGF.Builder.SetInsertPoint(CodeGenIPBB);
-
-      CGF.EmitStmt(RegionBodyStmt);
-
-      if (CGF.Builder.saveIP().isSet())
-        CGF.Builder.CreateBr(&FiniBB);
-    }
+                                         Twine RegionName);
 
     static void EmitCaptureStmt(CodeGenFunction &CGF, InsertPointTy CodeGenIP,
                                 llvm::BasicBlock &FiniBB, llvm::Function *Fn,
                                 ArrayRef<llvm::Value *> Args) {
       llvm::BasicBlock *CodeGenIPBB = CodeGenIP.getBlock();
       if (llvm::Instruction *CodeGenIPBBTI = CodeGenIPBB->getTerminator())
         CodeGenIPBBTI->eraseFromParent();
 
       CGF.Builder.SetInsertPoint(CodeGenIPBB);
 
       if (Fn->doesNotThrow())
         CGF.EmitNounwindRuntimeCall(Fn, Args);
       else
         CGF.EmitRuntimeCall(Fn, Args);
 
       if (CGF.Builder.saveIP().isSet())
         CGF.Builder.CreateBr(&FiniBB);
     }
 
+    /// Emit the body of an OMP region that will be outlined in
+    /// OpenMPIRBuilder::finalize().
+    /// \param CGF            The Codegen function this belongs to
+    /// \param RegionBodyStmt The body statement for the OpenMP region being
+    ///                       generated
+    /// \param AllocaIP       Where to insert alloca instructions
+    /// \param CodeGenIP      Where to insert the region code
+    /// \param RegionName     Name to be used for new blocks
+    static void EmitOMPOutlinedRegionBody(CodeGenFunction &CGF,
+                                          const Stmt *RegionBodyStmt,
+                                          InsertPointTy AllocaIP,
+                                          InsertPointTy CodeGenIP,
+                                          Twine RegionName);
+
     /// RAII for preserving necessary info during Outlined region body codegen.
     class OutlinedRegionBodyRAII {
 
       llvm::AssertingVH<llvm::Instruction> OldAllocaIP;
       CodeGenFunction::JumpDest OldReturnBlock;
-      CGBuilderTy::InsertPoint IP;
       CodeGenFunction &CGF;
 
     public:
       OutlinedRegionBodyRAII(CodeGenFunction &cgf, InsertPointTy &AllocaIP,
                              llvm::BasicBlock &RetBB)
           : CGF(cgf) {
         assert(AllocaIP.isSet() &&
                "Must specify Insertion point for allocas of outlined function");
         OldAllocaIP = CGF.AllocaInsertPt;
         CGF.AllocaInsertPt = &*AllocaIP.getPoint();
-        IP = CGF.Builder.saveIP();
 
         OldReturnBlock = CGF.ReturnBlock;
         CGF.ReturnBlock = CGF.getJumpDestInCurrentScope(&RetBB);
       }
 
       ~OutlinedRegionBodyRAII() {
         CGF.AllocaInsertPt = OldAllocaIP;
         CGF.ReturnBlock = OldReturnBlock;
-        CGF.Builder.restoreIP(IP);
       }
     };
 
     /// RAII for preserving necessary info during inlined region body codegen.
     class InlinedRegionBodyRAII {
 
       llvm::AssertingVH<llvm::Instruction> OldAllocaIP;
       CodeGenFunction &CGF;

clang/test/OpenMP/cancel_codegen.cpp

 //
 // CHECK3-LABEL: define {{[^@]+}}@main
 // CHECK3-SAME: (i32 noundef [[ARGC:%.*]], i8** noundef [[ARGV:%.*]]) #[[ATTR0:[0-9]+]] {
 // CHECK3-NEXT:  entry:
 // CHECK3-NEXT:    [[STRUCTARG:%.*]] = alloca { i32*, i8*** }, align 8
 // CHECK3-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[ARGV_ADDR:%.*]] = alloca i8**, align 8
+// CHECK3-NEXT:    [[P_LASTITER:%.*]] = alloca i32, align 4
+// CHECK3-NEXT:    [[P_LOWERBOUND:%.*]] = alloca i32, align 4
+// CHECK3-NEXT:    [[P_UPPERBOUND:%.*]] = alloca i32, align 4
+// CHECK3-NEXT:    [[P_STRIDE:%.*]] = alloca i32, align 4
+// CHECK3-NEXT:    [[P_LASTITER28:%.*]] = alloca i32, align 4
+// CHECK3-NEXT:    [[P_LOWERBOUND29:%.*]] = alloca i32, align 4
+// CHECK3-NEXT:    [[P_UPPERBOUND30:%.*]] = alloca i32, align 4
+// CHECK3-NEXT:    [[P_STRIDE31:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR_34:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[I36:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[AGG_CAPTURED:%.*]] = alloca [[STRUCT_ANON:%.*]], align 1
 // CHECK3-NEXT:    [[R:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    store i32 0, i32* [[RETVAL]], align 4
 // CHECK3-NEXT:    store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
 // CHECK3-NEXT:    store i8** [[ARGV]], i8*** [[ARGV_ADDR]], align 8
 // CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1:[0-9]+]])
-// CHECK3-NEXT:    [[P_LASTITER:%.*]] = alloca i32, align 4
-// CHECK3-NEXT:    [[P_LOWERBOUND:%.*]] = alloca i32, align 4
-// CHECK3-NEXT:    [[P_UPPERBOUND:%.*]] = alloca i32, align 4
-// CHECK3-NEXT:    [[P_STRIDE:%.*]] = alloca i32, align 4
-// CHECK3-NEXT:    [[P_LASTITER27:%.*]] = alloca i32, align 4
-// CHECK3-NEXT:    [[P_LOWERBOUND28:%.*]] = alloca i32, align 4
-// CHECK3-NEXT:    [[P_UPPERBOUND29:%.*]] = alloca i32, align 4
-// CHECK3-NEXT:    [[P_STRIDE30:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    br label [[OMP_PARALLEL:%.*]]
 // CHECK3:       omp_parallel:
 // CHECK3-NEXT:    [[GEP_ARGC_ADDR:%.*]] = getelementptr { i32*, i8*** }, { i32*, i8*** }* [[STRUCTARG]], i32 0, i32 0
 // CHECK3-NEXT:    store i32* [[ARGC_ADDR]], i32** [[GEP_ARGC_ADDR]], align 8
 // CHECK3-NEXT:    [[GEP_ARGV_ADDR:%.*]] = getelementptr { i32*, i8*** }, { i32*, i8*** }* [[STRUCTARG]], i32 0, i32 1
 // CHECK3-NEXT:    store i8*** [[ARGV_ADDR]], i8**** [[GEP_ARGV_ADDR]], align 8
 // CHECK3-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, { i32*, i8*** }*)* @main..omp_par to void (i32*, i32*, ...)*), { i32*, i8*** }* [[STRUCTARG]])
 // CHECK3-NEXT:    br label [[OMP_PAR_OUTLINED_EXIT:%.*]]

 // CHECK3-NEXT:    br label [[OMP_SECTION_LOOP_COND:%.*]]
 // CHECK3:       omp_section_loop.cond:
 // CHECK3-NEXT:    [[OMP_SECTION_LOOP_CMP:%.*]] = icmp ult i32 [[OMP_SECTION_LOOP_IV]], [[TMP3]]
 // CHECK3-NEXT:    br i1 [[OMP_SECTION_LOOP_CMP]], label [[OMP_SECTION_LOOP_BODY:%.*]], label [[OMP_SECTION_LOOP_EXIT:%.*]]
 // CHECK3:       omp_section_loop.body:
 // CHECK3-NEXT:    [[TMP4:%.*]] = add i32 [[OMP_SECTION_LOOP_IV]], [[TMP0]]
 // CHECK3-NEXT:    [[TMP5:%.*]] = mul i32 [[TMP4]], 1
 // CHECK3-NEXT:    [[TMP6:%.*]] = add i32 [[TMP5]], 0
-// CHECK3-NEXT:    switch i32 [[TMP6]], label [[OMP_SECTION_LOOP_INC]] [
+// CHECK3-NEXT:    switch i32 [[TMP6]], label [[OMP_SECTION_LOOP_BODY_SECTIONS_AFTER:%.*]] [
 // CHECK3-NEXT:    i32 0, label [[OMP_SECTION_LOOP_BODY_CASE:%.*]]
 // CHECK3-NEXT:    ]
+// CHECK3:       omp_section_loop.body.case:
+// CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM10:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
+// CHECK3-NEXT:    [[TMP7:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM10]], i32 3)
+// CHECK3-NEXT:    [[TMP8:%.*]] = icmp eq i32 [[TMP7]], 0
+// CHECK3-NEXT:    br i1 [[TMP8]], label [[OMP_SECTION_LOOP_BODY_CASE_SPLIT:%.*]], label [[OMP_SECTION_LOOP_BODY_CASE_CNCL:%.*]]
+// CHECK3:       omp_section_loop.body.case.split:
+// CHECK3-NEXT:    br label [[OMP_SECTION_LOOP_BODY_CASE_SECTION_AFTER:%.*]]
+// CHECK3:       omp_section_loop.body.case.section.after:
+// CHECK3-NEXT:    br label [[OMP_SECTION_LOOP_BODY_SECTIONS_AFTER]]
+// CHECK3:       omp_section_loop.body.sections.after:
+// CHECK3-NEXT:    br label [[OMP_SECTION_LOOP_INC]]
 // CHECK3:       omp_section_loop.inc:
 // CHECK3-NEXT:    [[OMP_SECTION_LOOP_NEXT]] = add nuw i32 [[OMP_SECTION_LOOP_IV]], 1
 // CHECK3-NEXT:    br label [[OMP_SECTION_LOOP_HEADER]]
 // CHECK3:       omp_section_loop.exit:
 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM11]])
 // CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM12:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
 // CHECK3-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[OMP_GLOBAL_THREAD_NUM12]])
 // CHECK3-NEXT:    br label [[OMP_SECTION_LOOP_AFTER:%.*]]
 // CHECK3:       omp_section_loop.after:
-// CHECK3-NEXT:    br label [[OMP_SECTIONS_END:%.*]]
-// CHECK3:       omp_sections.end:
+// CHECK3-NEXT:    br label [[OMP_SECTION_LOOP_AFTERSECTIONS_FINI:%.*]]
+// CHECK3:       omp_section_loop.aftersections.fini:
 // CHECK3-NEXT:    br label [[OMP_SECTION_LOOP_PREHEADER13:%.*]]
 // CHECK3:       omp_section_loop.preheader13:
-// CHECK3-NEXT:    store i32 0, i32* [[P_LOWERBOUND28]], align 4
-// CHECK3-NEXT:    store i32 1, i32* [[P_UPPERBOUND29]], align 4
-// CHECK3-NEXT:    store i32 1, i32* [[P_STRIDE30]], align 4
-// CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM31:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
-// CHECK3-NEXT:    call void @__kmpc_for_static_init_4u(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM31]], i32 34, i32* [[P_LASTITER27]], i32* [[P_LOWERBOUND28]], i32* [[P_UPPERBOUND29]], i32* [[P_STRIDE30]], i32 1, i32 0)
-// CHECK3-NEXT:    [[TMP7:%.*]] = load i32, i32* [[P_LOWERBOUND28]], align 4
-// CHECK3-NEXT:    [[TMP8:%.*]] = load i32, i32* [[P_UPPERBOUND29]], align 4
-// CHECK3-NEXT:    [[TMP9:%.*]] = sub i32 [[TMP8]], [[TMP7]]
-// CHECK3-NEXT:    [[TMP10:%.*]] = add i32 [[TMP9]], 1
+// CHECK3-NEXT:    store i32 0, i32* [[P_LOWERBOUND29]], align 4
+// CHECK3-NEXT:    store i32 1, i32* [[P_UPPERBOUND30]], align 4
+// CHECK3-NEXT:    store i32 1, i32* [[P_STRIDE31]], align 4
+// CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM32:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
+// CHECK3-NEXT:    call void @__kmpc_for_static_init_4u(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM32]], i32 34, i32* [[P_LASTITER28]], i32* [[P_LOWERBOUND29]], i32* [[P_UPPERBOUND30]], i32* [[P_STRIDE31]], i32 1, i32 0)
+// CHECK3-NEXT:    [[TMP9:%.*]] = load i32, i32* [[P_LOWERBOUND29]], align 4
+// CHECK3-NEXT:    [[TMP10:%.*]] = load i32, i32* [[P_UPPERBOUND30]], align 4
+// CHECK3-NEXT:    [[TMP11:%.*]] = sub i32 [[TMP10]], [[TMP9]]
+// CHECK3-NEXT:    [[TMP12:%.*]] = add i32 [[TMP11]], 1
 // CHECK3-NEXT:    br label [[OMP_SECTION_LOOP_HEADER14:%.*]]
 // CHECK3:       omp_section_loop.header14:
 // CHECK3-NEXT:    [[OMP_SECTION_LOOP_IV20:%.*]] = phi i32 [ 0, [[OMP_SECTION_LOOP_PREHEADER13]] ], [ [[OMP_SECTION_LOOP_NEXT22:%.*]], [[OMP_SECTION_LOOP_INC17:%.*]] ]
 // CHECK3-NEXT:    br label [[OMP_SECTION_LOOP_COND15:%.*]]
 // CHECK3:       omp_section_loop.cond15:
-// CHECK3-NEXT:    [[OMP_SECTION_LOOP_CMP21:%.*]] = icmp ult i32 [[OMP_SECTION_LOOP_IV20]], [[TMP10]]
+// CHECK3-NEXT:    [[OMP_SECTION_LOOP_CMP21:%.*]] = icmp ult i32 [[OMP_SECTION_LOOP_IV20]], [[TMP12]]
 // CHECK3-NEXT:    br i1 [[OMP_SECTION_LOOP_CMP21]], label [[OMP_SECTION_LOOP_BODY16:%.*]], label [[OMP_SECTION_LOOP_EXIT18:%.*]]
 // CHECK3:       omp_section_loop.body16:
-// CHECK3-NEXT:    [[TMP11:%.*]] = add i32 [[OMP_SECTION_LOOP_IV20]], [[TMP7]]
-// CHECK3-NEXT:    [[TMP12:%.*]] = mul i32 [[TMP11]], 1
-// CHECK3-NEXT:    [[TMP13:%.*]] = add i32 [[TMP12]], 0
-// CHECK3-NEXT:    switch i32 [[TMP13]], label [[OMP_SECTION_LOOP_INC17]] [
+// CHECK3-NEXT:    [[TMP13:%.*]] = add i32 [[OMP_SECTION_LOOP_IV20]], [[TMP9]]
+// CHECK3-NEXT:    [[TMP14:%.*]] = mul i32 [[TMP13]], 1
+// CHECK3-NEXT:    [[TMP15:%.*]] = add i32 [[TMP14]], 0
+// CHECK3-NEXT:    switch i32 [[TMP15]], label [[OMP_SECTION_LOOP_BODY16_SECTIONS_AFTER:%.*]] [
 // CHECK3-NEXT:    i32 0, label [[OMP_SECTION_LOOP_BODY_CASE23:%.*]]
 // CHECK3-NEXT:    i32 1, label [[OMP_SECTION_LOOP_BODY_CASE25:%.*]]
 // CHECK3-NEXT:    ]
+// CHECK3:       omp_section_loop.body.case23:
+// CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM24:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
+// CHECK3-NEXT:    [[TMP16:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM24]], i32 3)
+// CHECK3-NEXT:    [[TMP17:%.*]] = icmp eq i32 [[TMP16]], 0
+// CHECK3-NEXT:    br i1 [[TMP17]], label [[OMP_SECTION_LOOP_BODY_CASE23_SPLIT:%.*]], label [[OMP_SECTION_LOOP_BODY_CASE23_CNCL:%.*]]
+// CHECK3:       omp_section_loop.body.case23.split:
+// CHECK3-NEXT:    br label [[OMP_SECTION_LOOP_BODY_CASE23_SECTION_AFTER:%.*]]
+// CHECK3:       omp_section_loop.body.case23.section.after:
+// CHECK3-NEXT:    br label [[OMP_SECTION_LOOP_BODY16_SECTIONS_AFTER]]
+// CHECK3:       omp_section_loop.body.case25:
+// CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM27:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
+// CHECK3-NEXT:    [[TMP18:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM27]], i32 3)
+// CHECK3-NEXT:    [[TMP19:%.*]] = icmp eq i32 [[TMP18]], 0
+// CHECK3-NEXT:    br i1 [[TMP19]], label [[OMP_SECTION_LOOP_BODY_CASE25_SPLIT:%.*]], label [[OMP_SECTION_LOOP_BODY_CASE25_CNCL:%.*]]
+// CHECK3:       omp_section_loop.body.case25.split:
+// CHECK3-NEXT:    br label [[OMP_SECTION_LOOP_BODY_CASE25_SECTION_AFTER26:%.*]]
+// CHECK3:       omp_section_loop.body.case25.section.after26:
+// CHECK3-NEXT:    br label [[OMP_SECTION_LOOP_BODY_CASE25_SECTION_AFTER:%.*]]
+// CHECK3:       omp_section_loop.body.case25.section.after:
+// CHECK3-NEXT:    br label [[OMP_SECTION_LOOP_BODY16_SECTIONS_AFTER]]
+// CHECK3:       omp_section_loop.body16.sections.after:
+// CHECK3-NEXT:    br label [[OMP_SECTION_LOOP_INC17]]
 // CHECK3:       omp_section_loop.inc17:
 // CHECK3-NEXT:    [[OMP_SECTION_LOOP_NEXT22]] = add nuw i32 [[OMP_SECTION_LOOP_IV20]], 1
 // CHECK3-NEXT:    br label [[OMP_SECTION_LOOP_HEADER14]]
 // CHECK3:       omp_section_loop.exit18:
-// CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM31]])
-// CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM32:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
-// CHECK3-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB2]], i32 [[OMP_GLOBAL_THREAD_NUM32]])
+// CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM32]])
+// CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM33:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
+// CHECK3-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB2]], i32 [[OMP_GLOBAL_THREAD_NUM33]])
 // CHECK3-NEXT:    br label [[OMP_SECTION_LOOP_AFTER19:%.*]]
 // CHECK3:       omp_section_loop.after19:
-// CHECK3-NEXT:    br label [[OMP_SECTIONS_END33:%.*]]
-// CHECK3:       omp_sections.end33:
-// CHECK3-NEXT:    [[TMP14:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
-// CHECK3-NEXT:    store i32 [[TMP14]], i32* [[DOTCAPTURE_EXPR_]], align 4
-// CHECK3-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
-// CHECK3-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP15]], 0
+// CHECK3-NEXT:    br label [[OMP_SECTION_LOOP_AFTER19SECTIONS_FINI:%.*]]
+// CHECK3:       omp_section_loop.after19sections.fini:
+// CHECK3-NEXT:    [[TMP20:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
+// CHECK3-NEXT:    store i32 [[TMP20]], i32* [[DOTCAPTURE_EXPR_]], align 4
+// CHECK3-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
+// CHECK3-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP21]], 0
 // CHECK3-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
 // CHECK3-NEXT:    [[SUB35:%.*]] = sub nsw i32 [[DIV]], 1
 // CHECK3-NEXT:    store i32 [[SUB35]], i32* [[DOTCAPTURE_EXPR_34]], align 4
 // CHECK3-NEXT:    store i32 0, i32* [[I]], align 4
-// CHECK3-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
-// CHECK3-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP16]]
+// CHECK3-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
+// CHECK3-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP22]]
 // CHECK3-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
 // CHECK3:       omp.precond.then:
 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
-// CHECK3-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_34]], align 4
-// CHECK3-NEXT:    store i32 [[TMP17]], i32* [[DOTOMP_UB]], align 4
+// CHECK3-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_34]], align 4
+// CHECK3-NEXT:    store i32 [[TMP23]], i32* [[DOTOMP_UB]], align 4
 // CHECK3-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM37:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB6:[0-9]+]])
 // CHECK3-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB4:[0-9]+]], i32 [[OMP_GLOBAL_THREAD_NUM37]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
-// CHECK3-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK3-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_34]], align 4
-// CHECK3-NEXT:    [[CMP38:%.*]] = icmp sgt i32 [[TMP18]], [[TMP19]]
+// CHECK3-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
+// CHECK3-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_34]], align 4
+// CHECK3-NEXT:    [[CMP38:%.*]] = icmp sgt i32 [[TMP24]], [[TMP25]]
 // CHECK3-NEXT:    br i1 [[CMP38]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK3:       cond.true:
-// CHECK3-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_34]], align 4
+// CHECK3-NEXT:    [[TMP26:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_34]], align 4
 // CHECK3-NEXT:    br label [[COND_END:%.*]]
 // CHECK3:       cond.false:
-// CHECK3-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
+// CHECK3-NEXT:    [[TMP27:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK3-NEXT:    br label [[COND_END]]
 // CHECK3:       cond.end:
-// CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP20]], [[COND_TRUE]] ], [ [[TMP21]], [[COND_FALSE]] ]
+// CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP26]], [[COND_TRUE]] ], [ [[TMP27]], [[COND_FALSE]] ]
 // CHECK3-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
-// CHECK3-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
-// CHECK3-NEXT:    store i32 [[TMP22]], i32* [[DOTOMP_IV]], align 4
+// CHECK3-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
+// CHECK3-NEXT:    store i32 [[TMP28]], i32* [[DOTOMP_IV]], align 4
 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK3:       omp.inner.for.cond:
-// CHECK3-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
-// CHECK3-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK3-NEXT:    [[CMP39:%.*]] = icmp sle i32 [[TMP23]], [[TMP24]]
+// CHECK3-NEXT:    [[TMP29:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
+// CHECK3-NEXT:    [[TMP30:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
+// CHECK3-NEXT:    [[CMP39:%.*]] = icmp sle i32 [[TMP29]], [[TMP30]]
 // CHECK3-NEXT:    br i1 [[CMP39]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK3:       omp.inner.for.body:
-// CHECK3-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
-// CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP25]], 1
+// CHECK3-NEXT:    [[TMP31:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
+// CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP31]], 1
 // CHECK3-NEXT:    [[ADD40:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK3-NEXT:    store i32 [[ADD40]], i32* [[I36]], align 4
-// CHECK3-NEXT:    [[TMP26:%.*]] = load float, float* @flag, align 4
-// CHECK3-NEXT:    [[TOBOOL41:%.*]] = fcmp une float [[TMP26]], 0.000000e+00
+// CHECK3-NEXT:    [[TMP32:%.*]] = load float, float* @flag, align 4
+// CHECK3-NEXT:    [[TOBOOL41:%.*]] = fcmp une float [[TMP32]], 0.000000e+00
 // CHECK3-NEXT:    br i1 [[TOBOOL41]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
 // CHECK3:       omp_if.then:
 // CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM42:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB8:[0-9]+]])
-// CHECK3-NEXT:    [[TMP27:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM42]], i32 2)
-// CHECK3-NEXT:    [[TMP28:%.*]] = icmp ne i32 [[TMP27]], 0
-// CHECK3-NEXT:    br i1 [[TMP28]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
+// CHECK3-NEXT:    [[TMP33:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM42]], i32 2)
+// CHECK3-NEXT:    [[TMP34:%.*]] = icmp ne i32 [[TMP33]], 0
+// CHECK3-NEXT:    br i1 [[TMP34]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
 // CHECK3:       .cancel.exit:
 // CHECK3-NEXT:    br label [[CANCEL_EXIT:%.*]]
-// CHECK3:       omp_section_loop.body.case:
-// CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM10:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
-// CHECK3-NEXT:    [[TMP29:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM10]], i32 3)
-// CHECK3-NEXT:    [[TMP30:%.*]] = icmp eq i32 [[TMP29]], 0
-// CHECK3-NEXT:    br i1 [[TMP30]], label [[OMP_SECTION_LOOP_BODY_CASE_SPLIT:%.*]], label [[OMP_SECTION_LOOP_BODY_CASE_CNCL:%.*]]
-// CHECK3:       omp_section_loop.body.case.split:
-// CHECK3-NEXT:    br label [[OMP_SECTION_LOOP_EXIT]]
 // CHECK3:       omp_section_loop.body.case.cncl:
 // CHECK3-NEXT:    br label [[OMP_SECTION_LOOP_EXIT]]
-// CHECK3:       omp_section_loop.body.case23:
-// CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM24:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
-// CHECK3-NEXT:    [[TMP31:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM24]], i32 3)
-// CHECK3-NEXT:    [[TMP32:%.*]] = icmp eq i32 [[TMP31]], 0
-// CHECK3-NEXT:    br i1 [[TMP32]], label [[OMP_SECTION_LOOP_BODY_CASE23_SPLIT:%.*]], label [[OMP_SECTION_LOOP_BODY_CASE23_CNCL:%.*]]
-// CHECK3:       omp_section_loop.body.case23.split:
-// CHECK3-NEXT:    br label [[OMP_SECTION_LOOP_EXIT18]]
 // CHECK3:       omp_section_loop.body.case23.cncl:
 // CHECK3-NEXT:    br label [[OMP_SECTION_LOOP_EXIT18]]
-// CHECK3:       omp_section_loop.body.case25:
-// CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM26:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
-// CHECK3-NEXT:    [[TMP33:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM26]], i32 3)
-// CHECK3-NEXT:    [[TMP34:%.*]] = icmp eq i32 [[TMP33]], 0
-// CHECK3-NEXT:    br i1 [[TMP34]], label [[OMP_SECTION_LOOP_BODY_CASE25_SPLIT:%.*]], label [[OMP_SECTION_LOOP_BODY_CASE25_CNCL:%.*]]
-// CHECK3:       omp_section_loop.body.case25.split:
-// CHECK3-NEXT:    br label [[OMP_SECTION_LOOP_EXIT18]]
 // CHECK3:       omp_section_loop.body.case25.cncl:
 // CHECK3-NEXT:    br label [[OMP_SECTION_LOOP_EXIT18]]
 // CHECK3:       .cancel.continue:
 // CHECK3-NEXT:    br label [[OMP_IF_END:%.*]]
 // CHECK3:       omp_if.else:
 // CHECK3-NEXT:    br label [[OMP_IF_END]]
 // CHECK3:       omp_if.end:
 // CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]

 // CHECK3-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds i8*, i8** [[TMP11]], i64 0
 // CHECK3-NEXT:    [[TMP12:%.*]] = load i8*, i8** [[ARRAYIDX6]], align 8
 // CHECK3-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds i8, i8* [[TMP12]], i64 0
 // CHECK3-NEXT:    [[TMP13:%.*]] = load i8, i8* [[ARRAYIDX7]], align 1
 // CHECK3-NEXT:    [[CONV8:%.*]] = sext i8 [[TMP13]] to i32
 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV8]], [[TMP10]]
 // CHECK3-NEXT:    [[CONV9:%.*]] = trunc i32 [[ADD]] to i8
 // CHECK3-NEXT:    store i8 [[CONV9]], i8* [[ARRAYIDX7]], align 1
+// CHECK3-NEXT:    br label [[OMP_PAR_REGION_PARALLEL_AFTER:%.*]]
+// CHECK3:       omp.par.region.parallel.after:
 // CHECK3-NEXT:    br label [[OMP_PAR_PRE_FINALIZE:%.*]]
 // CHECK3:       omp.par.pre_finalize:
 // CHECK3-NEXT:    br label [[OMP_PAR_OUTLINED_EXIT_EXITSTUB]]
 // CHECK3:       14:
 // CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
 // CHECK3-NEXT:    [[TMP15:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM1]], i32 1)
 // CHECK3-NEXT:    [[TMP16:%.*]] = icmp eq i32 [[TMP15]], 0
 // CHECK3-NEXT:    br i1 [[TMP16]], label [[DOTSPLIT:%.*]], label [[DOTCNCL:%.*]]

 // CHECK3:       .omp.sections.case.cncl:
 // CHECK3-NEXT:    br label [[CANCEL_CONT:%.*]]
 // CHECK3:       .omp.sections.case2:
 // CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM3:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
 // CHECK3-NEXT:    [[TMP9:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM3]], i32 3)
 // CHECK3-NEXT:    [[TMP10:%.*]] = icmp eq i32 [[TMP9]], 0
 // CHECK3-NEXT:    br i1 [[TMP10]], label [[DOTOMP_SECTIONS_CASE2_SPLIT:%.*]], label [[DOTOMP_SECTIONS_CASE2_CNCL:%.*]]
 // CHECK3:       .omp.sections.case2.split:
+// CHECK3-NEXT:    br label [[DOTOMP_SECTIONS_CASE2_SECTION_AFTER:%.*]]
+// CHECK3:       .omp.sections.case2.section.after:
 // CHECK3-NEXT:    br label [[DOTOMP_SECTIONS_EXIT]]
 // CHECK3:       .omp.sections.case2.cncl:
 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_END]]
 // CHECK3:       .omp.sections.exit:
 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK3:       omp.inner.for.inc:
 // CHECK3-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
 // CHECK3-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP11]], 1

 //
 // CHECK4-LABEL: define {{[^@]+}}@main
 // CHECK4-SAME: (i32 noundef [[ARGC:%.*]], i8** noundef [[ARGV:%.*]]) #[[ATTR0:[0-9]+]] {
 // CHECK4-NEXT:  entry:
 // CHECK4-NEXT:    [[STRUCTARG:%.*]] = alloca { i32*, i8*** }, align 8
 // CHECK4-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
 // CHECK4-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32, align 4
 // CHECK4-NEXT:    [[ARGV_ADDR:%.*]] = alloca i8**, align 8
+// CHECK4-NEXT:    [[P_LASTITER:%.*]] = alloca i32, align 4
+// CHECK4-NEXT:    [[P_LOWERBOUND:%.*]] = alloca i32, align 4
+// CHECK4-NEXT:    [[P_UPPERBOUND:%.*]] = alloca i32, align 4
+// CHECK4-NEXT:    [[P_STRIDE:%.*]] = alloca i32, align 4
+// CHECK4-NEXT:    [[P_LASTITER28:%.*]] = alloca i32, align 4
+// CHECK4-NEXT:    [[P_LOWERBOUND29:%.*]] = alloca i32, align 4
+// CHECK4-NEXT:    [[P_UPPERBOUND30:%.*]] = alloca i32, align 4
+// CHECK4-NEXT:    [[P_STRIDE31:%.*]] = alloca i32, align 4
 // CHECK4-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 // CHECK4-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 // CHECK4-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
 // CHECK4-NEXT:    [[DOTCAPTURE_EXPR_34:%.*]] = alloca i32, align 4
 // CHECK4-NEXT:    [[I:%.*]] = alloca i32, align 4
 // CHECK4-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
 // CHECK4-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
 // CHECK4-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
 // CHECK4-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 // CHECK4-NEXT:    [[I36:%.*]] = alloca i32, align 4
 // CHECK4-NEXT:    [[AGG_CAPTURED:%.*]] = alloca [[STRUCT_ANON:%.*]], align 1
 // CHECK4-NEXT:    [[R:%.*]] = alloca i32, align 4
 // CHECK4-NEXT:    store i32 0, i32* [[RETVAL]], align 4
 // CHECK4-NEXT:    store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
 // CHECK4-NEXT:    store i8** [[ARGV]], i8*** [[ARGV_ADDR]], align 8
 // CHECK4-NEXT:    [[OMP_GLOBAL_THREAD_NUM:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1:[0-9]+]])
-// CHECK4-NEXT:    [[P_LASTITER:%.*]] = alloca i32, align 4
-// CHECK4-NEXT:    [[P_LOWERBOUND:%.*]] = alloca i32, align 4
-// CHECK4-NEXT:    [[P_UPPERBOUND:%.*]] = alloca i32, align 4
-// CHECK4-NEXT:    [[P_STRIDE:%.*]] = alloca i32, align 4
-// CHECK4-NEXT:    [[P_LASTITER27:%.*]] = alloca i32, align 4
-// CHECK4-NEXT:    [[P_LOWERBOUND28:%.*]] = alloca i32, align 4
-// CHECK4-NEXT:    [[P_UPPERBOUND29:%.*]] = alloca i32, align 4
-// CHECK4-NEXT:    [[P_STRIDE30:%.*]] = alloca i32, align 4
 // CHECK4-NEXT:    br label [[OMP_PARALLEL:%.*]]
 // CHECK4:       omp_parallel:
 // CHECK4-NEXT:    [[GEP_ARGC_ADDR:%.*]] = getelementptr { i32*, i8*** }, { i32*, i8*** }* [[STRUCTARG]], i32 0, i32 0
 // CHECK4-NEXT:    store i32* [[ARGC_ADDR]], i32** [[GEP_ARGC_ADDR]], align 8
 // CHECK4-NEXT:    [[GEP_ARGV_ADDR:%.*]] = getelementptr { i32*, i8*** }, { i32*, i8*** }* [[STRUCTARG]], i32 0, i32 1
 // CHECK4-NEXT:    store i8*** [[ARGV_ADDR]], i8**** [[GEP_ARGV_ADDR]], align 8
 // CHECK4-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, { i32*, i8*** }*)* @main..omp_par to void (i32*, i32*, ...)*), { i32*, i8*** }* [[STRUCTARG]])
 // CHECK4-NEXT:    br label [[OMP_PAR_OUTLINED_EXIT:%.*]]

 // CHECK4-NEXT:    br label [[OMP_SECTION_LOOP_COND:%.*]]
 // CHECK4:       omp_section_loop.cond:
 // CHECK4-NEXT:    [[OMP_SECTION_LOOP_CMP:%.*]] = icmp ult i32 [[OMP_SECTION_LOOP_IV]], [[TMP3]]
 // CHECK4-NEXT:    br i1 [[OMP_SECTION_LOOP_CMP]], label [[OMP_SECTION_LOOP_BODY:%.*]], label [[OMP_SECTION_LOOP_EXIT:%.*]]
 // CHECK4:       omp_section_loop.body:
 // CHECK4-NEXT:    [[TMP4:%.*]] = add i32 [[OMP_SECTION_LOOP_IV]], [[TMP0]]
 // CHECK4-NEXT:    [[TMP5:%.*]] = mul i32 [[TMP4]], 1
 // CHECK4-NEXT:    [[TMP6:%.*]] = add i32 [[TMP5]], 0
-// CHECK4-NEXT:    switch i32 [[TMP6]], label [[OMP_SECTION_LOOP_INC]] [
+// CHECK4-NEXT:    switch i32 [[TMP6]], label [[OMP_SECTION_LOOP_BODY_SECTIONS_AFTER:%.*]] [
 // CHECK4-NEXT:    i32 0, label [[OMP_SECTION_LOOP_BODY_CASE:%.*]]
 // CHECK4-NEXT:    ]
+// CHECK4:       omp_section_loop.body.case:
+// CHECK4-NEXT:    [[OMP_GLOBAL_THREAD_NUM10:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
+// CHECK4-NEXT:    [[TMP7:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM10]], i32 3)
+// CHECK4-NEXT:    [[TMP8:%.*]] = icmp eq i32 [[TMP7]], 0
+// CHECK4-NEXT:    br i1 [[TMP8]], label [[OMP_SECTION_LOOP_BODY_CASE_SPLIT:%.*]], label [[OMP_SECTION_LOOP_BODY_CASE_CNCL:%.*]]
+// CHECK4:       omp_section_loop.body.case.split:
+// CHECK4-NEXT:    br label [[OMP_SECTION_LOOP_BODY_CASE_SECTION_AFTER:%.*]]
+// CHECK4:       omp_section_loop.body.case.section.after:
+// CHECK4-NEXT:    br label [[OMP_SECTION_LOOP_BODY_SECTIONS_AFTER]]
+// CHECK4:       omp_section_loop.body.sections.after:
+// CHECK4-NEXT:    br label [[OMP_SECTION_LOOP_INC]]
 // CHECK4:       omp_section_loop.inc:
 // CHECK4-NEXT:    [[OMP_SECTION_LOOP_NEXT]] = add nuw i32 [[OMP_SECTION_LOOP_IV]], 1
 // CHECK4-NEXT:    br label [[OMP_SECTION_LOOP_HEADER]]
 // CHECK4:       omp_section_loop.exit:
 // CHECK4-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM11]])
 // CHECK4-NEXT:    [[OMP_GLOBAL_THREAD_NUM12:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
 // CHECK4-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[OMP_GLOBAL_THREAD_NUM12]])
 // CHECK4-NEXT:    br label [[OMP_SECTION_LOOP_AFTER:%.*]]
 // CHECK4:       omp_section_loop.after:
-// CHECK4-NEXT:    br label [[OMP_SECTIONS_END:%.*]]
-// CHECK4:       omp_sections.end:
+// CHECK4-NEXT:    br label [[OMP_SECTION_LOOP_AFTERSECTIONS_FINI:%.*]]
+// CHECK4:       omp_section_loop.aftersections.fini:
 // CHECK4-NEXT:    br label [[OMP_SECTION_LOOP_PREHEADER13:%.*]]
 // CHECK4:       omp_section_loop.preheader13:
-// CHECK4-NEXT:    store i32 0, i32* [[P_LOWERBOUND28]], align 4
-// CHECK4-NEXT:    store i32 1, i32* [[P_UPPERBOUND29]], align 4
-// CHECK4-NEXT:    store i32 1, i32* [[P_STRIDE30]], align 4
-// CHECK4-NEXT:    [[OMP_GLOBAL_THREAD_NUM31:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
-// CHECK4-NEXT:    call void @__kmpc_for_static_init_4u(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM31]], i32 34, i32* [[P_LASTITER27]], i32* [[P_LOWERBOUND28]], i32* [[P_UPPERBOUND29]], i32* [[P_STRIDE30]], i32 1, i32 0)
-// CHECK4-NEXT:    [[TMP7:%.*]] = load i32, i32* [[P_LOWERBOUND28]], align 4
-// CHECK4-NEXT:    [[TMP8:%.*]] = load i32, i32* [[P_UPPERBOUND29]], align 4
-// CHECK4-NEXT:    [[TMP9:%.*]] = sub i32 [[TMP8]], [[TMP7]]
-// CHECK4-NEXT:    [[TMP10:%.*]] = add i32 [[TMP9]], 1
+// CHECK4-NEXT:    store i32 0, i32* [[P_LOWERBOUND29]], align 4
+// CHECK4-NEXT:    store i32 1, i32* [[P_UPPERBOUND30]], align 4
+// CHECK4-NEXT:    store i32 1, i32* [[P_STRIDE31]], align 4
+// CHECK4-NEXT:    [[OMP_GLOBAL_THREAD_NUM32:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
+// CHECK4-NEXT:    call void @__kmpc_for_static_init_4u(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM32]], i32 34, i32* [[P_LASTITER28]], i32* [[P_LOWERBOUND29]], i32* [[P_UPPERBOUND30]], i32* [[P_STRIDE31]], i32 1, i32 0)
+// CHECK4-NEXT:    [[TMP9:%.*]] = load i32, i32* [[P_LOWERBOUND29]], align 4
+// CHECK4-NEXT:    [[TMP10:%.*]] = load i32, i32* [[P_UPPERBOUND30]], align 4
+// CHECK4-NEXT:    [[TMP11:%.*]] = sub i32 [[TMP10]], [[TMP9]]
+// CHECK4-NEXT:    [[TMP12:%.*]] = add i32 [[TMP11]], 1
 // CHECK4-NEXT:    br label [[OMP_SECTION_LOOP_HEADER14:%.*]]
 // CHECK4:       omp_section_loop.header14:
 // CHECK4-NEXT:    [[OMP_SECTION_LOOP_IV20:%.*]] = phi i32 [ 0, [[OMP_SECTION_LOOP_PREHEADER13]] ], [ [[OMP_SECTION_LOOP_NEXT22:%.*]], [[OMP_SECTION_LOOP_INC17:%.*]] ]
 // CHECK4-NEXT:    br label [[OMP_SECTION_LOOP_COND15:%.*]]
 // CHECK4:       omp_section_loop.cond15:
-// CHECK4-NEXT:    [[OMP_SECTION_LOOP_CMP21:%.*]] = icmp ult i32 [[OMP_SECTION_LOOP_IV20]], [[TMP10]]
+// CHECK4-NEXT:    [[OMP_SECTION_LOOP_CMP21:%.*]] = icmp ult i32 [[OMP_SECTION_LOOP_IV20]], [[TMP12]]
 // CHECK4-NEXT:    br i1 [[OMP_SECTION_LOOP_CMP21]], label [[OMP_SECTION_LOOP_BODY16:%.*]], label [[OMP_SECTION_LOOP_EXIT18:%.*]]
 // CHECK4:       omp_section_loop.body16:
-// CHECK4-NEXT:    [[TMP11:%.*]] = add i32 [[OMP_SECTION_LOOP_IV20]], [[TMP7]]
-// CHECK4-NEXT:    [[TMP12:%.*]] = mul i32 [[TMP11]], 1
-// CHECK4-NEXT:    [[TMP13:%.*]] = add i32 [[TMP12]], 0
-// CHECK4-NEXT:    switch i32 [[TMP13]], label [[OMP_SECTION_LOOP_INC17]] [
+// CHECK4-NEXT:    [[TMP13:%.*]] = add i32 [[OMP_SECTION_LOOP_IV20]], [[TMP9]]
+// CHECK4-NEXT:    [[TMP14:%.*]] = mul i32 [[TMP13]], 1
+// CHECK4-NEXT:    [[TMP15:%.*]] = add i32 [[TMP14]], 0
+// CHECK4-NEXT:    switch i32 [[TMP15]], label [[OMP_SECTION_LOOP_BODY16_SECTIONS_AFTER:%.*]] [
 // CHECK4-NEXT:    i32 0, label [[OMP_SECTION_LOOP_BODY_CASE23:%.*]]
 // CHECK4-NEXT:    i32 1, label [[OMP_SECTION_LOOP_BODY_CASE25:%.*]]
 // CHECK4-NEXT:    ]
+// CHECK4:       omp_section_loop.body.case23:
+// CHECK4-NEXT:    [[OMP_GLOBAL_THREAD_NUM24:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
+// CHECK4-NEXT:    [[TMP16:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM24]], i32 3)
+// CHECK4-NEXT:    [[TMP17:%.*]] = icmp eq i32 [[TMP16]], 0
+// CHECK4-NEXT:    br i1 [[TMP17]], label [[OMP_SECTION_LOOP_BODY_CASE23_SPLIT:%.*]], label [[OMP_SECTION_LOOP_BODY_CASE23_CNCL:%.*]]
+// CHECK4:       omp_section_loop.body.case23.split:
+// CHECK4-NEXT:    br label [[OMP_SECTION_LOOP_BODY_CASE23_SECTION_AFTER:%.*]]
+// CHECK4:       omp_section_loop.body.case23.section.after:
+// CHECK4-NEXT:    br label [[OMP_SECTION_LOOP_BODY16_SECTIONS_AFTER]]
+// CHECK4:       omp_section_loop.body.case25:
+// CHECK4-NEXT:    [[OMP_GLOBAL_THREAD_NUM27:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
+// CHECK4-NEXT:    [[TMP18:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM27]], i32 3)
+// CHECK4-NEXT:    [[TMP19:%.*]] = icmp eq i32 [[TMP18]], 0
+// CHECK4-NEXT:    br i1 [[TMP19]], label [[OMP_SECTION_LOOP_BODY_CASE25_SPLIT:%.*]], label [[OMP_SECTION_LOOP_BODY_CASE25_CNCL:%.*]]
+// CHECK4:       omp_section_loop.body.case25.split:
+// CHECK4-NEXT:    br label [[OMP_SECTION_LOOP_BODY_CASE25_SECTION_AFTER26:%.*]]
+// CHECK4:       omp_section_loop.body.case25.section.after26:
+// CHECK4-NEXT:    br label [[OMP_SECTION_LOOP_BODY_CASE25_SECTION_AFTER:%.*]]
+// CHECK4:       omp_section_loop.body.case25.section.after:
+// CHECK4-NEXT:    br label [[OMP_SECTION_LOOP_BODY16_SECTIONS_AFTER]]
+// CHECK4:       omp_section_loop.body16.sections.after:
+// CHECK4-NEXT:    br label [[OMP_SECTION_LOOP_INC17]]
 // CHECK4:       omp_section_loop.inc17:
 // CHECK4-NEXT:    [[OMP_SECTION_LOOP_NEXT22]] = add nuw i32 [[OMP_SECTION_LOOP_IV20]], 1
 // CHECK4-NEXT:    br label [[OMP_SECTION_LOOP_HEADER14]]
 // CHECK4:       omp_section_loop.exit18:
-// CHECK4-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM31]])
-// CHECK4-NEXT:    [[OMP_GLOBAL_THREAD_NUM32:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
-// CHECK4-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB2]], i32 [[OMP_GLOBAL_THREAD_NUM32]])
+// CHECK4-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM32]])
+// CHECK4-NEXT:    [[OMP_GLOBAL_THREAD_NUM33:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
+// CHECK4-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB2]], i32 [[OMP_GLOBAL_THREAD_NUM33]])
 // CHECK4-NEXT:    br label [[OMP_SECTION_LOOP_AFTER19:%.*]]
 // CHECK4:       omp_section_loop.after19:
-// CHECK4-NEXT:    br label [[OMP_SECTIONS_END33:%.*]]
-// CHECK4:       omp_sections.end33:
-// CHECK4-NEXT:    [[TMP14:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
-// CHECK4-NEXT:    store i32 [[TMP14]], i32* [[DOTCAPTURE_EXPR_]], align 4
-// CHECK4-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
-// CHECK4-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP15]], 0
+// CHECK4-NEXT:    br label [[OMP_SECTION_LOOP_AFTER19SECTIONS_FINI:%.*]]
+// CHECK4:       omp_section_loop.after19sections.fini:
+// CHECK4-NEXT:    [[TMP20:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
+// CHECK4-NEXT:    store i32 [[TMP20]], i32* [[DOTCAPTURE_EXPR_]], align 4
+// CHECK4-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
+// CHECK4-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP21]], 0
 // CHECK4-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
 // CHECK4-NEXT:    [[SUB35:%.*]] = sub nsw i32 [[DIV]], 1
 // CHECK4-NEXT:    store i32 [[SUB35]], i32* [[DOTCAPTURE_EXPR_34]], align 4
 // CHECK4-NEXT:    store i32 0, i32* [[I]], align 4
-// CHECK4-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
-// CHECK4-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP16]]
+// CHECK4-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
+// CHECK4-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP22]]
 // CHECK4-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
 // CHECK4:       omp.precond.then:
 // CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
-// CHECK4-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_34]], align 4
-// CHECK4-NEXT:    store i32 [[TMP17]], i32* [[DOTOMP_UB]], align 4
+// CHECK4-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_34]], align 4
+// CHECK4-NEXT:    store i32 [[TMP23]], i32* [[DOTOMP_UB]], align 4
 // CHECK4-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
 // CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK4-NEXT:    [[OMP_GLOBAL_THREAD_NUM37:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB6:[0-9]+]])
 // CHECK4-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB4:[0-9]+]], i32 [[OMP_GLOBAL_THREAD_NUM37]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
-// CHECK4-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK4-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_34]], align 4
-// CHECK4-NEXT:    [[CMP38:%.*]] = icmp sgt i32 [[TMP18]], [[TMP19]]
+// CHECK4-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
+// CHECK4-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_34]], align 4
+// CHECK4-NEXT:    [[CMP38:%.*]] = icmp sgt i32 [[TMP24]], [[TMP25]]
 // CHECK4-NEXT:    br i1 [[CMP38]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK4:       cond.true:
-// CHECK4-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_34]], align 4
+// CHECK4-NEXT:    [[TMP26:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_34]], align 4
 // CHECK4-NEXT:    br label [[COND_END:%.*]]
 // CHECK4:       cond.false:
-// CHECK4-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
+// CHECK4-NEXT:    [[TMP27:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK4-NEXT:    br label [[COND_END]]
 // CHECK4:       cond.end:
-// CHECK4-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP20]], [[COND_TRUE]] ], [ [[TMP21]], [[COND_FALSE]] ]
+// CHECK4-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP26]], [[COND_TRUE]] ], [ [[TMP27]], [[COND_FALSE]] ]
 // CHECK4-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
-// CHECK4-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
-// CHECK4-NEXT:    store i32 [[TMP22]], i32* [[DOTOMP_IV]], align 4
+// CHECK4-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
+// CHECK4-NEXT:    store i32 [[TMP28]], i32* [[DOTOMP_IV]], align 4
 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK4:       omp.inner.for.cond:
-// CHECK4-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
-// CHECK4-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK4-NEXT:    [[CMP39:%.*]] = icmp sle i32 [[TMP23]], [[TMP24]]
+// CHECK4-NEXT:    [[TMP29:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
+// CHECK4-NEXT:    [[TMP30:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
+// CHECK4-NEXT:    [[CMP39:%.*]] = icmp sle i32 [[TMP29]], [[TMP30]]
 // CHECK4-NEXT:    br i1 [[CMP39]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK4:       omp.inner.for.body:
-// CHECK4-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
-// CHECK4-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP25]], 1
+// CHECK4-NEXT:    [[TMP31:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
+// CHECK4-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP31]], 1
 // CHECK4-NEXT:    [[ADD40:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK4-NEXT:    store i32 [[ADD40]], i32* [[I36]], align 4
-// CHECK4-NEXT:    [[TMP26:%.*]] = load float, float* @flag, align 4
-// CHECK4-NEXT:    [[TOBOOL41:%.*]] = fcmp une float [[TMP26]], 0.000000e+00
+// CHECK4-NEXT:    [[TMP32:%.*]] = load float, float* @flag, align 4
+// CHECK4-NEXT:    [[TOBOOL41:%.*]] = fcmp une float [[TMP32]], 0.000000e+00
 // CHECK4-NEXT:    br i1 [[TOBOOL41]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
 // CHECK4:       omp_if.then:
 // CHECK4-NEXT:    [[OMP_GLOBAL_THREAD_NUM42:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB8:[0-9]+]])
-// CHECK4-NEXT:    [[TMP27:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM42]], i32 2)
-// CHECK4-NEXT:    [[TMP28:%.*]] = icmp ne i32 [[TMP27]], 0
-// CHECK4-NEXT:    br i1 [[TMP28]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
+// CHECK4-NEXT:    [[TMP33:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM42]], i32 2)
+// CHECK4-NEXT:    [[TMP34:%.*]] = icmp ne i32 [[TMP33]], 0
+// CHECK4-NEXT:    br i1 [[TMP34]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
 // CHECK4:       .cancel.exit:
 // CHECK4-NEXT:    br label [[CANCEL_EXIT:%.*]]
-// CHECK4:       omp_section_loop.body.case:
-// CHECK4-NEXT:    [[OMP_GLOBAL_THREAD_NUM10:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
-// CHECK4-NEXT:    [[TMP29:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM10]], i32 3)
-// CHECK4-NEXT:    [[TMP30:%.*]] = icmp eq i32 [[TMP29]], 0
-// CHECK4-NEXT:    br i1 [[TMP30]], label [[OMP_SECTION_LOOP_BODY_CASE_SPLIT:%.*]], label [[OMP_SECTION_LOOP_BODY_CASE_CNCL:%.*]]
-// CHECK4:       omp_section_loop.body.case.split:
-// CHECK4-NEXT:    br label [[OMP_SECTION_LOOP_EXIT]]
 // CHECK4:       omp_section_loop.body.case.cncl:
 // CHECK4-NEXT:    br label [[OMP_SECTION_LOOP_EXIT]]
-// CHECK4:       omp_section_loop.body.case23:
-// CHECK4-NEXT:    [[OMP_GLOBAL_THREAD_NUM24:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
-// CHECK4-NEXT:    [[TMP31:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM24]], i32 3)
-// CHECK4-NEXT:    [[TMP32:%.*]] = icmp eq i32 [[TMP31]], 0
-// CHECK4-NEXT:    br i1 [[TMP32]], label [[OMP_SECTION_LOOP_BODY_CASE23_SPLIT:%.*]], label [[OMP_SECTION_LOOP_BODY_CASE23_CNCL:%.*]]
-// CHECK4:       omp_section_loop.body.case23.split:
-// CHECK4-NEXT:    br label [[OMP_SECTION_LOOP_EXIT18]]
 // CHECK4:       omp_section_loop.body.case23.cncl:
 // CHECK4-NEXT:    br label [[OMP_SECTION_LOOP_EXIT18]]
-// CHECK4:       omp_section_loop.body.case25:
-// CHECK4-NEXT:    [[OMP_GLOBAL_THREAD_NUM26:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
-// CHECK4-NEXT:    [[TMP33:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM26]], i32 3)
-// CHECK4-NEXT:    [[TMP34:%.*]] = icmp eq i32 [[TMP33]], 0
-// CHECK4-NEXT:    br i1 [[TMP34]], label [[OMP_SECTION_LOOP_BODY_CASE25_SPLIT:%.*]], label [[OMP_SECTION_LOOP_BODY_CASE25_CNCL:%.*]]
-// CHECK4:       omp_section_loop.body.case25.split:
-// CHECK4-NEXT:    br label [[OMP_SECTION_LOOP_EXIT18]]
 // CHECK4:       omp_section_loop.body.case25.cncl:
 // CHECK4-NEXT:    br label [[OMP_SECTION_LOOP_EXIT18]]
 // CHECK4:       .cancel.continue:
 // CHECK4-NEXT:    br label [[OMP_IF_END:%.*]]
 // CHECK4:       omp_if.else:
 // CHECK4-NEXT:    br label [[OMP_IF_END]]
 // CHECK4:       omp_if.end:
 // CHECK4-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]

 // CHECK4-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds i8*, i8** [[TMP11]], i64 0
 // CHECK4-NEXT:    [[TMP12:%.*]] = load i8*, i8** [[ARRAYIDX6]], align 8
 // CHECK4-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds i8, i8* [[TMP12]], i64 0
 // CHECK4-NEXT:    [[TMP13:%.*]] = load i8, i8* [[ARRAYIDX7]], align 1
 // CHECK4-NEXT:    [[CONV8:%.*]] = sext i8 [[TMP13]] to i32
 // CHECK4-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV8]], [[TMP10]]
 // CHECK4-NEXT:    [[CONV9:%.*]] = trunc i32 [[ADD]] to i8
 // CHECK4-NEXT:    store i8 [[CONV9]], i8* [[ARRAYIDX7]], align 1
+// CHECK4-NEXT:    br label [[OMP_PAR_REGION_PARALLEL_AFTER:%.*]]
+// CHECK4:       omp.par.region.parallel.after:
 // CHECK4-NEXT:    br label [[OMP_PAR_PRE_FINALIZE:%.*]]
 // CHECK4:       omp.par.pre_finalize:
 // CHECK4-NEXT:    br label [[OMP_PAR_OUTLINED_EXIT_EXITSTUB]]
 // CHECK4:       14:
 // CHECK4-NEXT:    [[OMP_GLOBAL_THREAD_NUM1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
 // CHECK4-NEXT:    [[TMP15:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM1]], i32 1)
 // CHECK4-NEXT:    [[TMP16:%.*]] = icmp eq i32 [[TMP15]], 0
 // CHECK4-NEXT:    br i1 [[TMP16]], label [[DOTSPLIT:%.*]], label [[DOTCNCL:%.*]]

 // CHECK4:       .omp.sections.case.cncl:
 // CHECK4-NEXT:    br label [[CANCEL_CONT:%.*]]
 // CHECK4:       .omp.sections.case2:
 // CHECK4-NEXT:    [[OMP_GLOBAL_THREAD_NUM3:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
 // CHECK4-NEXT:    [[TMP9:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM3]], i32 3)
 // CHECK4-NEXT:    [[TMP10:%.*]] = icmp eq i32 [[TMP9]], 0
 // CHECK4-NEXT:    br i1 [[TMP10]], label [[DOTOMP_SECTIONS_CASE2_SPLIT:%.*]], label [[DOTOMP_SECTIONS_CASE2_CNCL:%.*]]
 // CHECK4:       .omp.sections.case2.split:
+// CHECK4-NEXT:    br label [[DOTOMP_SECTIONS_CASE2_SECTION_AFTER:%.*]]
+// CHECK4:       .omp.sections.case2.section.after:
 // CHECK4-NEXT:    br label [[DOTOMP_SECTIONS_EXIT]]
 // CHECK4:       .omp.sections.case2.cncl:
 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_END]]
 // CHECK4:       .omp.sections.exit:
 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK4:       omp.inner.for.inc:
 // CHECK4-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
 // CHECK4-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP11]], 1

 //
 // CHECK9-LABEL: define {{[^@]+}}@main
 // CHECK9-SAME: (i32 noundef [[ARGC:%.*]], i8** noundef [[ARGV:%.*]]) #[[ATTR0:[0-9]+]] {
 // CHECK9-NEXT:  entry:
 // CHECK9-NEXT:    [[STRUCTARG:%.*]] = alloca { i32*, i8*** }, align 8
 // CHECK9-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
 // CHECK9-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32, align 4
 // CHECK9-NEXT:    [[ARGV_ADDR:%.*]] = alloca i8**, align 8
+// CHECK9-NEXT:    [[P_LASTITER:%.*]] = alloca i32, align 4
+// CHECK9-NEXT:    [[P_LOWERBOUND:%.*]] = alloca i32, align 4
+// CHECK9-NEXT:    [[P_UPPERBOUND:%.*]] = alloca i32, align 4
+// CHECK9-NEXT:    [[P_STRIDE:%.*]] = alloca i32, align 4
+// CHECK9-NEXT:    [[P_LASTITER28:%.*]] = alloca i32, align 4
+// CHECK9-NEXT:    [[P_LOWERBOUND29:%.*]] = alloca i32, align 4
+// CHECK9-NEXT:    [[P_UPPERBOUND30:%.*]] = alloca i32, align 4
+// CHECK9-NEXT:    [[P_STRIDE31:%.*]] = alloca i32, align 4
 // CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 // CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 // CHECK9-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
 // CHECK9-NEXT:    [[DOTCAPTURE_EXPR_34:%.*]] = alloca i32, align 4
 // CHECK9-NEXT:    [[I:%.*]] = alloca i32, align 4
 // CHECK9-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
 // CHECK9-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
 // CHECK9-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
 // CHECK9-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 // CHECK9-NEXT:    [[I36:%.*]] = alloca i32, align 4
 // CHECK9-NEXT:    [[AGG_CAPTURED:%.*]] = alloca [[STRUCT_ANON:%.*]], align 1
 // CHECK9-NEXT:    [[R:%.*]] = alloca i32, align 4
 // CHECK9-NEXT:    store i32 0, i32* [[RETVAL]], align 4
 // CHECK9-NEXT:    store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
 // CHECK9-NEXT:    store i8** [[ARGV]], i8*** [[ARGV_ADDR]], align 8
 // CHECK9-NEXT:    [[OMP_GLOBAL_THREAD_NUM:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1:[0-9]+]])
-// CHECK9-NEXT:    [[P_LASTITER:%.*]] = alloca i32, align 4
-// CHECK9-NEXT:    [[P_LOWERBOUND:%.*]] = alloca i32, align 4
-// CHECK9-NEXT:    [[P_UPPERBOUND:%.*]] = alloca i32, align 4
-// CHECK9-NEXT:    [[P_STRIDE:%.*]] = alloca i32, align 4
-// CHECK9-NEXT:    [[P_LASTITER27:%.*]] = alloca i32, align 4
-// CHECK9-NEXT:    [[P_LOWERBOUND28:%.*]] = alloca i32, align 4
-// CHECK9-NEXT:    [[P_UPPERBOUND29:%.*]] = alloca i32, align 4
-// CHECK9-NEXT:    [[P_STRIDE30:%.*]] = alloca i32, align 4
 // CHECK9-NEXT:    br label [[OMP_PARALLEL:%.*]]
 // CHECK9:       omp_parallel:
 // CHECK9-NEXT:    [[GEP_ARGC_ADDR:%.*]] = getelementptr { i32*, i8*** }, { i32*, i8*** }* [[STRUCTARG]], i32 0, i32 0
 // CHECK9-NEXT:    store i32* [[ARGC_ADDR]], i32** [[GEP_ARGC_ADDR]], align 8
 // CHECK9-NEXT:    [[GEP_ARGV_ADDR:%.*]] = getelementptr { i32*, i8*** }, { i32*, i8*** }* [[STRUCTARG]], i32 0, i32 1
 // CHECK9-NEXT:    store i8*** [[ARGV_ADDR]], i8**** [[GEP_ARGV_ADDR]], align 8
 // CHECK9-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, { i32*, i8*** }*)* @main..omp_par to void (i32*, i32*, ...)*), { i32*, i8*** }* [[STRUCTARG]])
 // CHECK9-NEXT:    br label [[OMP_PAR_OUTLINED_EXIT:%.*]]

 // CHECK9-NEXT:    br label [[OMP_SECTION_LOOP_COND:%.*]]
 // CHECK9:       omp_section_loop.cond:
 // CHECK9-NEXT:    [[OMP_SECTION_LOOP_CMP:%.*]] = icmp ult i32 [[OMP_SECTION_LOOP_IV]], [[TMP3]]
 // CHECK9-NEXT:    br i1 [[OMP_SECTION_LOOP_CMP]], label [[OMP_SECTION_LOOP_BODY:%.*]], label [[OMP_SECTION_LOOP_EXIT:%.*]]
 // CHECK9:       omp_section_loop.body:
 // CHECK9-NEXT:    [[TMP4:%.*]] = add i32 [[OMP_SECTION_LOOP_IV]], [[TMP0]]
 // CHECK9-NEXT:    [[TMP5:%.*]] = mul i32 [[TMP4]], 1
 // CHECK9-NEXT:    [[TMP6:%.*]] = add i32 [[TMP5]], 0
-// CHECK9-NEXT:    switch i32 [[TMP6]], label [[OMP_SECTION_LOOP_INC]] [
+// CHECK9-NEXT:    switch i32 [[TMP6]], label [[OMP_SECTION_LOOP_BODY_SECTIONS_AFTER:%.*]] [
 // CHECK9-NEXT:    i32 0, label [[OMP_SECTION_LOOP_BODY_CASE:%.*]]
 // CHECK9-NEXT:    ]
+// CHECK9:       omp_section_loop.body.case:
+// CHECK9-NEXT:    [[OMP_GLOBAL_THREAD_NUM10:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
+// CHECK9-NEXT:    [[TMP7:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM10]], i32 3)
+// CHECK9-NEXT:    [[TMP8:%.*]] = icmp eq i32 [[TMP7]], 0
+// CHECK9-NEXT:    br i1 [[TMP8]], label [[OMP_SECTION_LOOP_BODY_CASE_SPLIT:%.*]], label [[OMP_SECTION_LOOP_BODY_CASE_CNCL:%.*]]
+// CHECK9:       omp_section_loop.body.case.split:
+// CHECK9-NEXT:    br label [[OMP_SECTION_LOOP_BODY_CASE_SECTION_AFTER:%.*]]
+// CHECK9:       omp_section_loop.body.case.section.after:
+// CHECK9-NEXT:    br label [[OMP_SECTION_LOOP_BODY_SECTIONS_AFTER]]
+// CHECK9:       omp_section_loop.body.sections.after:
+// CHECK9-NEXT:    br label [[OMP_SECTION_LOOP_INC]]
 // CHECK9:       omp_section_loop.inc:
 // CHECK9-NEXT:    [[OMP_SECTION_LOOP_NEXT]] = add nuw i32 [[OMP_SECTION_LOOP_IV]], 1
 // CHECK9-NEXT:    br label [[OMP_SECTION_LOOP_HEADER]]
 // CHECK9:       omp_section_loop.exit:
 // CHECK9-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM11]])
 // CHECK9-NEXT:    [[OMP_GLOBAL_THREAD_NUM12:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
 // CHECK9-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[OMP_GLOBAL_THREAD_NUM12]])
 // CHECK9-NEXT:    br label [[OMP_SECTION_LOOP_AFTER:%.*]]
 // CHECK9:       omp_section_loop.after:
-// CHECK9-NEXT:    br label [[OMP_SECTIONS_END:%.*]]
-// CHECK9:       omp_sections.end:
+// CHECK9-NEXT:    br label [[OMP_SECTION_LOOP_AFTERSECTIONS_FINI:%.*]]
+// CHECK9:       omp_section_loop.aftersections.fini:
 // CHECK9-NEXT:    br label [[OMP_SECTION_LOOP_PREHEADER13:%.*]]
 // CHECK9:       omp_section_loop.preheader13:
-// CHECK9-NEXT:    store i32 0, i32* [[P_LOWERBOUND28]], align 4
-// CHECK9-NEXT:    store i32 1, i32* [[P_UPPERBOUND29]], align 4
-// CHECK9-NEXT:    store i32 1, i32* [[P_STRIDE30]], align 4
-// CHECK9-NEXT:    [[OMP_GLOBAL_THREAD_NUM31:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
-// CHECK9-NEXT:    call void @__kmpc_for_static_init_4u(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM31]], i32 34, i32* [[P_LASTITER27]], i32* [[P_LOWERBOUND28]], i32* [[P_UPPERBOUND29]], i32* [[P_STRIDE30]], i32 1, i32 0)
-// CHECK9-NEXT:    [[TMP7:%.*]] = load i32, i32* [[P_LOWERBOUND28]], align 4
-// CHECK9-NEXT:    [[TMP8:%.*]] = load i32, i32* [[P_UPPERBOUND29]], align 4
-// CHECK9-NEXT:    [[TMP9:%.*]] = sub i32 [[TMP8]], [[TMP7]]
-// CHECK9-NEXT:    [[TMP10:%.*]] = add i32 [[TMP9]], 1
+// CHECK9-NEXT:    store i32 0, i32* [[P_LOWERBOUND29]], align 4
+// CHECK9-NEXT:    store i32 1, i32* [[P_UPPERBOUND30]], align 4
+// CHECK9-NEXT:    store i32 1, i32* [[P_STRIDE31]], align 4
+// CHECK9-NEXT:    [[OMP_GLOBAL_THREAD_NUM32:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
+// CHECK9-NEXT:    call void @__kmpc_for_static_init_4u(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM32]], i32 34, i32* [[P_LASTITER28]], i32* [[P_LOWERBOUND29]], i32* [[P_UPPERBOUND30]], i32* [[P_STRIDE31]], i32 1, i32 0)
+// CHECK9-NEXT:    [[TMP9:%.*]] = load i32, i32* [[P_LOWERBOUND29]], align 4
+// CHECK9-NEXT:    [[TMP10:%.*]] = load i32, i32* [[P_UPPERBOUND30]], align 4
+// CHECK9-NEXT:    [[TMP11:%.*]] = sub i32 [[TMP10]], [[TMP9]]
+// CHECK9-NEXT:    [[TMP12:%.*]] = add i32 [[TMP11]], 1
 // CHECK9-NEXT:    br label [[OMP_SECTION_LOOP_HEADER14:%.*]]
 // CHECK9:       omp_section_loop.header14:
 // CHECK9-NEXT:    [[OMP_SECTION_LOOP_IV20:%.*]] = phi i32 [ 0, [[OMP_SECTION_LOOP_PREHEADER13]] ], [ [[OMP_SECTION_LOOP_NEXT22:%.*]], [[OMP_SECTION_LOOP_INC17:%.*]] ]
 // CHECK9-NEXT:    br label [[OMP_SECTION_LOOP_COND15:%.*]]
 // CHECK9:       omp_section_loop.cond15:
-// CHECK9-NEXT:    [[OMP_SECTION_LOOP_CMP21:%.*]] = icmp ult i32 [[OMP_SECTION_LOOP_IV20]], [[TMP10]]
+// CHECK9-NEXT:    [[OMP_SECTION_LOOP_CMP21:%.*]] = icmp ult i32 [[OMP_SECTION_LOOP_IV20]], [[TMP12]]
 // CHECK9-NEXT:    br i1 [[OMP_SECTION_LOOP_CMP21]], label [[OMP_SECTION_LOOP_BODY16:%.*]], label [[OMP_SECTION_LOOP_EXIT18:%.*]]
 // CHECK9:       omp_section_loop.body16:
-// CHECK9-NEXT:    [[TMP11:%.*]] = add i32 [[OMP_SECTION_LOOP_IV20]], [[TMP7]]
-// CHECK9-NEXT:    [[TMP12:%.*]] = mul i32 [[TMP11]], 1
-// CHECK9-NEXT:    [[TMP13:%.*]] = add i32 [[TMP12]], 0
-// CHECK9-NEXT:    switch i32 [[TMP13]], label [[OMP_SECTION_LOOP_INC17]] [
+// CHECK9-NEXT:    [[TMP13:%.*]] = add i32 [[OMP_SECTION_LOOP_IV20]], [[TMP9]]
+// CHECK9-NEXT:    [[TMP14:%.*]] = mul i32 [[TMP13]], 1
+// CHECK9-NEXT:    [[TMP15:%.*]] = add i32 [[TMP14]], 0
+// CHECK9-NEXT:    switch i32 [[TMP15]], label [[OMP_SECTION_LOOP_BODY16_SECTIONS_AFTER:%.*]] [
 // CHECK9-NEXT:    i32 0, label [[OMP_SECTION_LOOP_BODY_CASE23:%.*]]
 // CHECK9-NEXT:    i32 1, label [[OMP_SECTION_LOOP_BODY_CASE25:%.*]]
 // CHECK9-NEXT:    ]
+// CHECK9:       omp_section_loop.body.case23:
+// CHECK9-NEXT:    [[OMP_GLOBAL_THREAD_NUM24:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
+// CHECK9-NEXT:    [[TMP16:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM24]], i32 3)
+// CHECK9-NEXT:    [[TMP17:%.*]] = icmp eq i32 [[TMP16]], 0
+// CHECK9-NEXT:    br i1 [[TMP17]], label [[OMP_SECTION_LOOP_BODY_CASE23_SPLIT:%.*]], label [[OMP_SECTION_LOOP_BODY_CASE23_CNCL:%.*]]
+// CHECK9:       omp_section_loop.body.case23.split:
+// CHECK9-NEXT:    br label [[OMP_SECTION_LOOP_BODY_CASE23_SECTION_AFTER:%.*]]
+// CHECK9:       omp_section_loop.body.case23.section.after:
+// CHECK9-NEXT:    br label [[OMP_SECTION_LOOP_BODY16_SECTIONS_AFTER]]
+// CHECK9:       omp_section_loop.body.case25:
+// CHECK9-NEXT:    [[OMP_GLOBAL_THREAD_NUM27:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
+// CHECK9-NEXT:    [[TMP18:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM27]], i32 3)
+// CHECK9-NEXT:    [[TMP19:%.*]] = icmp eq i32 [[TMP18]], 0
+// CHECK9-NEXT:    br i1 [[TMP19]], label [[OMP_SECTION_LOOP_BODY_CASE25_SPLIT:%.*]], label [[OMP_SECTION_LOOP_BODY_CASE25_CNCL:%.*]]
+// CHECK9:       omp_section_loop.body.case25.split:
+// CHECK9-NEXT:    br label [[OMP_SECTION_LOOP_BODY_CASE25_SECTION_AFTER26:%.*]]
+// CHECK9:       omp_section_loop.body.case25.section.after26:
+// CHECK9-NEXT:    br label [[OMP_SECTION_LOOP_BODY_CASE25_SECTION_AFTER:%.*]]
+// CHECK9:       omp_section_loop.body.case25.section.after:
+// CHECK9-NEXT:    br label [[OMP_SECTION_LOOP_BODY16_SECTIONS_AFTER]]
+// CHECK9:       omp_section_loop.body16.sections.after:
+// CHECK9-NEXT:    br label [[OMP_SECTION_LOOP_INC17]]
 // CHECK9:       omp_section_loop.inc17:
 // CHECK9-NEXT:    [[OMP_SECTION_LOOP_NEXT22]] = add nuw i32 [[OMP_SECTION_LOOP_IV20]], 1
 // CHECK9-NEXT:    br label [[OMP_SECTION_LOOP_HEADER14]]
 // CHECK9:       omp_section_loop.exit18:
-// CHECK9-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM31]])
-// CHECK9-NEXT:    [[OMP_GLOBAL_THREAD_NUM32:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
-// CHECK9-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB2]], i32 [[OMP_GLOBAL_THREAD_NUM32]])
+// CHECK9-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM32]])
+// CHECK9-NEXT:    [[OMP_GLOBAL_THREAD_NUM33:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
+// CHECK9-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB2]], i32 [[OMP_GLOBAL_THREAD_NUM33]])
 // CHECK9-NEXT:    br label [[OMP_SECTION_LOOP_AFTER19:%.*]]
 // CHECK9:       omp_section_loop.after19:
-// CHECK9-NEXT:    br label [[OMP_SECTIONS_END33:%.*]]
-// CHECK9:       omp_sections.end33:
-// CHECK9-NEXT:    [[TMP14:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
-// CHECK9-NEXT:    store i32 [[TMP14]], i32* [[DOTCAPTURE_EXPR_]], align 4
-// CHECK9-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
-// CHECK9-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP15]], 0
+// CHECK9-NEXT:    br label [[OMP_SECTION_LOOP_AFTER19SECTIONS_FINI:%.*]]
+// CHECK9:       omp_section_loop.after19sections.fini:
+// CHECK9-NEXT:    [[TMP20:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
+// CHECK9-NEXT:    store i32 [[TMP20]], i32* [[DOTCAPTURE_EXPR_]], align 4
+// CHECK9-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
+// CHECK9-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP21]], 0
 // CHECK9-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
 // CHECK9-NEXT:    [[SUB35:%.*]] = sub nsw i32 [[DIV]], 1
 // CHECK9-NEXT:    store i32 [[SUB35]], i32* [[DOTCAPTURE_EXPR_34]], align 4
 // CHECK9-NEXT:    store i32 0, i32* [[I]], align 4
-// CHECK9-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
-// CHECK9-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP16]]
+// CHECK9-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
+// CHECK9-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP22]]
 // CHECK9-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
 // CHECK9:       omp.precond.then:
 // CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
-// CHECK9-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_34]], align 4
-// CHECK9-NEXT:    store i32 [[TMP17]], i32* [[DOTOMP_UB]], align 4
+// CHECK9-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_34]], align 4
+// CHECK9-NEXT:    store i32 [[TMP23]], i32* [[DOTOMP_UB]], align 4
 // CHECK9-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
 // CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK9-NEXT:    [[OMP_GLOBAL_THREAD_NUM37:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB6:[0-9]+]])
 // CHECK9-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB4:[0-9]+]], i32 [[OMP_GLOBAL_THREAD_NUM37]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
-// CHECK9-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK9-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_34]], align 4
-// CHECK9-NEXT:    [[CMP38:%.*]] = icmp sgt i32 [[TMP18]], [[TMP19]]
+// CHECK9-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
+// CHECK9-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_34]], align 4
+// CHECK9-NEXT:    [[CMP38:%.*]] = icmp sgt i32 [[TMP24]], [[TMP25]]
 // CHECK9-NEXT:    br i1 [[CMP38]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK9:       cond.true:
-// CHECK9-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_34]], align 4
+// CHECK9-NEXT:    [[TMP26:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_34]], align 4
 // CHECK9-NEXT:    br label [[COND_END:%.*]]
 // CHECK9:       cond.false:
-// CHECK9-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
+// CHECK9-NEXT:    [[TMP27:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK9-NEXT:    br label [[COND_END]]
 // CHECK9:       cond.end:
-// CHECK9-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP20]], [[COND_TRUE]] ], [ [[TMP21]], [[COND_FALSE]] ]
+// CHECK9-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP26]], [[COND_TRUE]] ], [ [[TMP27]], [[COND_FALSE]] ]
 // CHECK9-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
-// CHECK9-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
-// CHECK9-NEXT:    store i32 [[TMP22]], i32* [[DOTOMP_IV]], align 4
+// CHECK9-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
+// CHECK9-NEXT:    store i32 [[TMP28]], i32* [[DOTOMP_IV]], align 4
 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK9:       omp.inner.for.cond:
-// CHECK9-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
-// CHECK9-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK9-NEXT:    [[CMP39:%.*]] = icmp sle i32 [[TMP23]], [[TMP24]]
+// CHECK9-NEXT:    [[TMP29:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
+// CHECK9-NEXT:    [[TMP30:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
+// CHECK9-NEXT:    [[CMP39:%.*]] = icmp sle i32 [[TMP29]], [[TMP30]]
 // CHECK9-NEXT:    br i1 [[CMP39]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK9:       omp.inner.for.body:
-// CHECK9-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
-// CHECK9-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP25]], 1
+// CHECK9-NEXT:    [[TMP31:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
+// CHECK9-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP31]], 1
 // CHECK9-NEXT:    [[ADD40:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK9-NEXT:    store i32 [[ADD40]], i32* [[I36]], align 4
-// CHECK9-NEXT:    [[TMP26:%.*]] = load float, float* @flag, align 4
-// CHECK9-NEXT:    [[TOBOOL41:%.*]] = fcmp une float [[TMP26]], 0.000000e+00
+// CHECK9-NEXT:    [[TMP32:%.*]] = load float, float* @flag, align 4
+// CHECK9-NEXT:    [[TOBOOL41:%.*]] = fcmp une float [[TMP32]], 0.000000e+00
 // CHECK9-NEXT:    br i1 [[TOBOOL41]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
 // CHECK9:       omp_if.then:
 // CHECK9-NEXT:    [[OMP_GLOBAL_THREAD_NUM42:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB8:[0-9]+]])
-// CHECK9-NEXT:    [[TMP27:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM42]], i32 2)
-// CHECK9-NEXT:    [[TMP28:%.*]] = icmp ne i32 [[TMP27]], 0
-// CHECK9-NEXT:    br i1 [[TMP28]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
+// CHECK9-NEXT:    [[TMP33:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM42]], i32 2)
+// CHECK9-NEXT:    [[TMP34:%.*]] = icmp ne i32 [[TMP33]], 0
+// CHECK9-NEXT:    br i1 [[TMP34]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
 // CHECK9:       .cancel.exit:
 // CHECK9-NEXT:    br label [[CANCEL_EXIT:%.*]]
-// CHECK9:       omp_section_loop.body.case:
-// CHECK9-NEXT:    [[OMP_GLOBAL_THREAD_NUM10:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
-// CHECK9-NEXT:    [[TMP29:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM10]], i32 3)
-// CHECK9-NEXT:    [[TMP30:%.*]] = icmp eq i32 [[TMP29]], 0
-// CHECK9-NEXT:    br i1 [[TMP30]], label [[OMP_SECTION_LOOP_BODY_CASE_SPLIT:%.*]], label [[OMP_SECTION_LOOP_BODY_CASE_CNCL:%.*]]
-// CHECK9:       omp_section_loop.body.case.split:
-// CHECK9-NEXT:    br label [[OMP_SECTION_LOOP_EXIT]]
 // CHECK9:       omp_section_loop.body.case.cncl:
 // CHECK9-NEXT:    br label [[OMP_SECTION_LOOP_EXIT]]
-// CHECK9:       omp_section_loop.body.case23:
-// CHECK9-NEXT:    [[OMP_GLOBAL_THREAD_NUM24:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
-// CHECK9-NEXT:    [[TMP31:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM24]], i32 3)
-// CHECK9-NEXT:    [[TMP32:%.*]] = icmp eq i32 [[TMP31]], 0
-// CHECK9-NEXT:    br i1 [[TMP32]], label [[OMP_SECTION_LOOP_BODY_CASE23_SPLIT:%.*]], label [[OMP_SECTION_LOOP_BODY_CASE23_CNCL:%.*]]
-// CHECK9:       omp_section_loop.body.case23.split:
-// CHECK9-NEXT:    br label [[OMP_SECTION_LOOP_EXIT18]]
 // CHECK9:       omp_section_loop.body.case23.cncl:
 // CHECK9-NEXT:    br label [[OMP_SECTION_LOOP_EXIT18]]
-// CHECK9:       omp_section_loop.body.case25:
-// CHECK9-NEXT:    [[OMP_GLOBAL_THREAD_NUM26:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
-// CHECK9-NEXT:    [[TMP33:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM26]], i32 3)
-// CHECK9-NEXT:    [[TMP34:%.*]] = icmp eq i32 [[TMP33]], 0
-// CHECK9-NEXT:    br i1 [[TMP34]], label [[OMP_SECTION_LOOP_BODY_CASE25_SPLIT:%.*]], label [[OMP_SECTION_LOOP_BODY_CASE25_CNCL:%.*]]
-// CHECK9:       omp_section_loop.body.case25.split:
-// CHECK9-NEXT:    br label [[OMP_SECTION_LOOP_EXIT18]]
 // CHECK9:       omp_section_loop.body.case25.cncl:
 // CHECK9-NEXT:    br label [[OMP_SECTION_LOOP_EXIT18]]
 // CHECK9:       .cancel.continue:
 // CHECK9-NEXT:    br label [[OMP_IF_END:%.*]]
 // CHECK9:       omp_if.else:
 // CHECK9-NEXT:    br label [[OMP_IF_END]]
 // CHECK9:       omp_if.end:
 // CHECK9-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]

 // CHECK9-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds i8*, i8** [[TMP11]], i64 0
 // CHECK9-NEXT:    [[TMP12:%.*]] = load i8*, i8** [[ARRAYIDX6]], align 8
 // CHECK9-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds i8, i8* [[TMP12]], i64 0
 // CHECK9-NEXT:    [[TMP13:%.*]] = load i8, i8* [[ARRAYIDX7]], align 1
 // CHECK9-NEXT:    [[CONV8:%.*]] = sext i8 [[TMP13]] to i32
 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV8]], [[TMP10]]
 // CHECK9-NEXT:    [[CONV9:%.*]] = trunc i32 [[ADD]] to i8
 // CHECK9-NEXT:    store i8 [[CONV9]], i8* [[ARRAYIDX7]], align 1
+// CHECK9-NEXT:    br label [[OMP_PAR_REGION_PARALLEL_AFTER:%.*]]
+// CHECK9:       omp.par.region.parallel.after:
 // CHECK9-NEXT:    br label [[OMP_PAR_PRE_FINALIZE:%.*]]
 // CHECK9:       omp.par.pre_finalize:
 // CHECK9-NEXT:    br label [[OMP_PAR_OUTLINED_EXIT_EXITSTUB]]
 // CHECK9:       14:
 // CHECK9-NEXT:    [[OMP_GLOBAL_THREAD_NUM1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
 // CHECK9-NEXT:    [[TMP15:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM1]], i32 1)
 // CHECK9-NEXT:    [[TMP16:%.*]] = icmp eq i32 [[TMP15]], 0
 // CHECK9-NEXT:    br i1 [[TMP16]], label [[DOTSPLIT:%.*]], label [[DOTCNCL:%.*]]

 // CHECK9:       .omp.sections.case.cncl:
 // CHECK9-NEXT:    br label [[CANCEL_CONT:%.*]]
 // CHECK9:       .omp.sections.case2:
 // CHECK9-NEXT:    [[OMP_GLOBAL_THREAD_NUM3:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
 // CHECK9-NEXT:    [[TMP9:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM3]], i32 3)
 // CHECK9-NEXT:    [[TMP10:%.*]] = icmp eq i32 [[TMP9]], 0
 // CHECK9-NEXT:    br i1 [[TMP10]], label [[DOTOMP_SECTIONS_CASE2_SPLIT:%.*]], label [[DOTOMP_SECTIONS_CASE2_CNCL:%.*]]
 // CHECK9:       .omp.sections.case2.split:
+// CHECK9-NEXT:    br label [[DOTOMP_SECTIONS_CASE2_SECTION_AFTER:%.*]]
+// CHECK9:       .omp.sections.case2.section.after:
 // CHECK9-NEXT:    br label [[DOTOMP_SECTIONS_EXIT]]
 // CHECK9:       .omp.sections.case2.cncl:
 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_END]]
 // CHECK9:       .omp.sections.exit:
 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK9:       omp.inner.for.inc:
 // CHECK9-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
 // CHECK9-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP11]], 1

 //
 // CHECK10-LABEL: define {{[^@]+}}@main
 // CHECK10-SAME: (i32 noundef [[ARGC:%.*]], i8** noundef [[ARGV:%.*]]) #[[ATTR0:[0-9]+]] {
 // CHECK10-NEXT:  entry:
 // CHECK10-NEXT:    [[STRUCTARG:%.*]] = alloca { i32*, i8*** }, align 8
 // CHECK10-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
 // CHECK10-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32, align 4
 // CHECK10-NEXT:    [[ARGV_ADDR:%.*]] = alloca i8**, align 8
+// CHECK10-NEXT:    [[P_LASTITER:%.*]] = alloca i32, align 4
+// CHECK10-NEXT:    [[P_LOWERBOUND:%.*]] = alloca i32, align 4
+// CHECK10-NEXT:    [[P_UPPERBOUND:%.*]] = alloca i32, align 4
+// CHECK10-NEXT:    [[P_STRIDE:%.*]] = alloca i32, align 4
+// CHECK10-NEXT:    [[P_LASTITER28:%.*]] = alloca i32, align 4
+// CHECK10-NEXT:    [[P_LOWERBOUND29:%.*]] = alloca i32, align 4
+// CHECK10-NEXT:    [[P_UPPERBOUND30:%.*]] = alloca i32, align 4
+// CHECK10-NEXT:    [[P_STRIDE31:%.*]] = alloca i32, align 4
 // CHECK10-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 // CHECK10-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 // CHECK10-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
 // CHECK10-NEXT:    [[DOTCAPTURE_EXPR_34:%.*]] = alloca i32, align 4
 // CHECK10-NEXT:    [[I:%.*]] = alloca i32, align 4
 // CHECK10-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
 // CHECK10-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
 // CHECK10-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
 // CHECK10-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 // CHECK10-NEXT:    [[I36:%.*]] = alloca i32, align 4
 // CHECK10-NEXT:    [[AGG_CAPTURED:%.*]] = alloca [[STRUCT_ANON:%.*]], align 1
 // CHECK10-NEXT:    [[R:%.*]] = alloca i32, align 4
 // CHECK10-NEXT:    store i32 0, i32* [[RETVAL]], align 4
 // CHECK10-NEXT:    store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
 // CHECK10-NEXT:    store i8** [[ARGV]], i8*** [[ARGV_ADDR]], align 8
 // CHECK10-NEXT:    [[OMP_GLOBAL_THREAD_NUM:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1:[0-9]+]])
-// CHECK10-NEXT:    [[P_LASTITER:%.*]] = alloca i32, align 4
-// CHECK10-NEXT:    [[P_LOWERBOUND:%.*]] = alloca i32, align 4
-// CHECK10-NEXT:    [[P_UPPERBOUND:%.*]] = alloca i32, align 4
-// CHECK10-NEXT:    [[P_STRIDE:%.*]] = alloca i32, align 4
-// CHECK10-NEXT:    [[P_LASTITER27:%.*]] = alloca i32, align 4
-// CHECK10-NEXT:    [[P_LOWERBOUND28:%.*]] = alloca i32, align 4
-// CHECK10-NEXT:    [[P_UPPERBOUND29:%.*]] = alloca i32, align 4
-// CHECK10-NEXT:    [[P_STRIDE30:%.*]] = alloca i32, align 4
 // CHECK10-NEXT:    br label [[OMP_PARALLEL:%.*]]
 // CHECK10:       omp_parallel:
 // CHECK10-NEXT:    [[GEP_ARGC_ADDR:%.*]] = getelementptr { i32*, i8*** }, { i32*, i8*** }* [[STRUCTARG]], i32 0, i32 0
 // CHECK10-NEXT:    store i32* [[ARGC_ADDR]], i32** [[GEP_ARGC_ADDR]], align 8
 // CHECK10-NEXT:    [[GEP_ARGV_ADDR:%.*]] = getelementptr { i32*, i8*** }, { i32*, i8*** }* [[STRUCTARG]], i32 0, i32 1
 // CHECK10-NEXT:    store i8*** [[ARGV_ADDR]], i8**** [[GEP_ARGV_ADDR]], align 8
 // CHECK10-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, { i32*, i8*** }*)* @main..omp_par to void (i32*, i32*, ...)*), { i32*, i8*** }* [[STRUCTARG]])
 // CHECK10-NEXT:    br label [[OMP_PAR_OUTLINED_EXIT:%.*]]

 // CHECK10-NEXT:    br label [[OMP_SECTION_LOOP_COND:%.*]]
 // CHECK10:       omp_section_loop.cond:
 // CHECK10-NEXT:    [[OMP_SECTION_LOOP_CMP:%.*]] = icmp ult i32 [[OMP_SECTION_LOOP_IV]], [[TMP3]]
 // CHECK10-NEXT:    br i1 [[OMP_SECTION_LOOP_CMP]], label [[OMP_SECTION_LOOP_BODY:%.*]], label [[OMP_SECTION_LOOP_EXIT:%.*]]
 // CHECK10:       omp_section_loop.body:
 // CHECK10-NEXT:    [[TMP4:%.*]] = add i32 [[OMP_SECTION_LOOP_IV]], [[TMP0]]
 // CHECK10-NEXT:    [[TMP5:%.*]] = mul i32 [[TMP4]], 1
 // CHECK10-NEXT:    [[TMP6:%.*]] = add i32 [[TMP5]], 0
-// CHECK10-NEXT:    switch i32 [[TMP6]], label [[OMP_SECTION_LOOP_INC]] [
+// CHECK10-NEXT:    switch i32 [[TMP6]], label [[OMP_SECTION_LOOP_BODY_SECTIONS_AFTER:%.*]] [
 // CHECK10-NEXT:    i32 0, label [[OMP_SECTION_LOOP_BODY_CASE:%.*]]
 // CHECK10-NEXT:    ]
+// CHECK10:       omp_section_loop.body.case:
+// CHECK10-NEXT:    [[OMP_GLOBAL_THREAD_NUM10:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
+// CHECK10-NEXT:    [[TMP7:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM10]], i32 3)
+// CHECK10-NEXT:    [[TMP8:%.*]] = icmp eq i32 [[TMP7]], 0
+// CHECK10-NEXT:    br i1 [[TMP8]], label [[OMP_SECTION_LOOP_BODY_CASE_SPLIT:%.*]], label [[OMP_SECTION_LOOP_BODY_CASE_CNCL:%.*]]
+// CHECK10:       omp_section_loop.body.case.split:
+// CHECK10-NEXT:    br label [[OMP_SECTION_LOOP_BODY_CASE_SECTION_AFTER:%.*]]
+// CHECK10:       omp_section_loop.body.case.section.after:
+// CHECK10-NEXT:    br label [[OMP_SECTION_LOOP_BODY_SECTIONS_AFTER]]
+// CHECK10:       omp_section_loop.body.sections.after:
+// CHECK10-NEXT:    br label [[OMP_SECTION_LOOP_INC]]
 // CHECK10:       omp_section_loop.inc:
 // CHECK10-NEXT:    [[OMP_SECTION_LOOP_NEXT]] = add nuw i32 [[OMP_SECTION_LOOP_IV]], 1
 // CHECK10-NEXT:    br label [[OMP_SECTION_LOOP_HEADER]]
 // CHECK10:       omp_section_loop.exit:
 // CHECK10-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM11]])
 // CHECK10-NEXT:    [[OMP_GLOBAL_THREAD_NUM12:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
 // CHECK10-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[OMP_GLOBAL_THREAD_NUM12]])
 // CHECK10-NEXT:    br label [[OMP_SECTION_LOOP_AFTER:%.*]]
 // CHECK10:       omp_section_loop.after:
-// CHECK10-NEXT:    br label [[OMP_SECTIONS_END:%.*]]
-// CHECK10:       omp_sections.end:
+// CHECK10-NEXT:    br label [[OMP_SECTION_LOOP_AFTERSECTIONS_FINI:%.*]]
+// CHECK10:       omp_section_loop.aftersections.fini:
 // CHECK10-NEXT:    br label [[OMP_SECTION_LOOP_PREHEADER13:%.*]]
 // CHECK10:       omp_section_loop.preheader13:
-// CHECK10-NEXT:    store i32 0, i32* [[P_LOWERBOUND28]], align 4
-// CHECK10-NEXT:    store i32 1, i32* [[P_UPPERBOUND29]], align 4
-// CHECK10-NEXT:    store i32 1, i32* [[P_STRIDE30]], align 4
-// CHECK10-NEXT:    [[OMP_GLOBAL_THREAD_NUM31:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
-// CHECK10-NEXT:    call void @__kmpc_for_static_init_4u(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM31]], i32 34, i32* [[P_LASTITER27]], i32* [[P_LOWERBOUND28]], i32* [[P_UPPERBOUND29]], i32* [[P_STRIDE30]], i32 1, i32 0)
-// CHECK10-NEXT:    [[TMP7:%.*]] = load i32, i32* [[P_LOWERBOUND28]], align 4
-// CHECK10-NEXT:    [[TMP8:%.*]] = load i32, i32* [[P_UPPERBOUND29]], align 4
-// CHECK10-NEXT:    [[TMP9:%.*]] = sub i32 [[TMP8]], [[TMP7]]
-// CHECK10-NEXT:    [[TMP10:%.*]] = add i32 [[TMP9]], 1
+// CHECK10-NEXT:    store i32 0, i32* [[P_LOWERBOUND29]], align 4
+// CHECK10-NEXT:    store i32 1, i32* [[P_UPPERBOUND30]], align 4
+// CHECK10-NEXT:    store i32 1, i32* [[P_STRIDE31]], align 4
+// CHECK10-NEXT:    [[OMP_GLOBAL_THREAD_NUM32:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
+// CHECK10-NEXT:    call void @__kmpc_for_static_init_4u(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM32]], i32 34, i32* [[P_LASTITER28]], i32* [[P_LOWERBOUND29]], i32* [[P_UPPERBOUND30]], i32* [[P_STRIDE31]], i32 1, i32 0)
+// CHECK10-NEXT:    [[TMP9:%.*]] = load i32, i32* [[P_LOWERBOUND29]], align 4
+// CHECK10-NEXT:    [[TMP10:%.*]] = load i32, i32* [[P_UPPERBOUND30]], align 4
+// CHECK10-NEXT:    [[TMP11:%.*]] = sub i32 [[TMP10]], [[TMP9]]
+// CHECK10-NEXT:    [[TMP12:%.*]] = add i32 [[TMP11]], 1
 // CHECK10-NEXT:    br label [[OMP_SECTION_LOOP_HEADER14:%.*]]
 // CHECK10:       omp_section_loop.header14:
 // CHECK10-NEXT:    [[OMP_SECTION_LOOP_IV20:%.*]] = phi i32 [ 0, [[OMP_SECTION_LOOP_PREHEADER13]] ], [ [[OMP_SECTION_LOOP_NEXT22:%.*]], [[OMP_SECTION_LOOP_INC17:%.*]] ]
 // CHECK10-NEXT:    br label [[OMP_SECTION_LOOP_COND15:%.*]]
 // CHECK10:       omp_section_loop.cond15:
-// CHECK10-NEXT:    [[OMP_SECTION_LOOP_CMP21:%.*]] = icmp ult i32 [[OMP_SECTION_LOOP_IV20]], [[TMP10]]
+// CHECK10-NEXT:    [[OMP_SECTION_LOOP_CMP21:%.*]] = icmp ult i32 [[OMP_SECTION_LOOP_IV20]], [[TMP12]]
 // CHECK10-NEXT:    br i1 [[OMP_SECTION_LOOP_CMP21]], label [[OMP_SECTION_LOOP_BODY16:%.*]], label [[OMP_SECTION_LOOP_EXIT18:%.*]]
 // CHECK10:       omp_section_loop.body16:
-// CHECK10-NEXT:    [[TMP11:%.*]] = add i32 [[OMP_SECTION_LOOP_IV20]], [[TMP7]]
-// CHECK10-NEXT:    [[TMP12:%.*]] = mul i32 [[TMP11]], 1
-// CHECK10-NEXT:    [[TMP13:%.*]] = add i32 [[TMP12]], 0
-// CHECK10-NEXT:    switch i32 [[TMP13]], label [[OMP_SECTION_LOOP_INC17]] [
+// CHECK10-NEXT:    [[TMP13:%.*]] = add i32 [[OMP_SECTION_LOOP_IV20]], [[TMP9]]
+// CHECK10-NEXT:    [[TMP14:%.*]] = mul i32 [[TMP13]], 1
+// CHECK10-NEXT:    [[TMP15:%.*]] = add i32 [[TMP14]], 0
+// CHECK10-NEXT:    switch i32 [[TMP15]], label [[OMP_SECTION_LOOP_BODY16_SECTIONS_AFTER:%.*]] [
 // CHECK10-NEXT:    i32 0, label [[OMP_SECTION_LOOP_BODY_CASE23:%.*]]
 // CHECK10-NEXT:    i32 1, label [[OMP_SECTION_LOOP_BODY_CASE25:%.*]]
 // CHECK10-NEXT:    ]
+// CHECK10:       omp_section_loop.body.case23:
+// CHECK10-NEXT:    [[OMP_GLOBAL_THREAD_NUM24:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
+// CHECK10-NEXT:    [[TMP16:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM24]], i32 3)
+// CHECK10-NEXT:    [[TMP17:%.*]] = icmp eq i32 [[TMP16]], 0
+// CHECK10-NEXT:    br i1 [[TMP17]], label [[OMP_SECTION_LOOP_BODY_CASE23_SPLIT:%.*]], label [[OMP_SECTION_LOOP_BODY_CASE23_CNCL:%.*]]
+// CHECK10:       omp_section_loop.body.case23.split:
+// CHECK10-NEXT:    br label [[OMP_SECTION_LOOP_BODY_CASE23_SECTION_AFTER:%.*]]
+// CHECK10:       omp_section_loop.body.case23.section.after:
+// CHECK10-NEXT:    br label [[OMP_SECTION_LOOP_BODY16_SECTIONS_AFTER]]
+// CHECK10:       omp_section_loop.body.case25:
+// CHECK10-NEXT:    [[OMP_GLOBAL_THREAD_NUM27:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
+// CHECK10-NEXT:    [[TMP18:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM27]], i32 3)
+// CHECK10-NEXT:    [[TMP19:%.*]] = icmp eq i32 [[TMP18]], 0
+// CHECK10-NEXT:    br i1 [[TMP19]], label [[OMP_SECTION_LOOP_BODY_CASE25_SPLIT:%.*]], label [[OMP_SECTION_LOOP_BODY_CASE25_CNCL:%.*]]
+// CHECK10:       omp_section_loop.body.case25.split:
+// CHECK10-NEXT:    br label [[OMP_SECTION_LOOP_BODY_CASE25_SECTION_AFTER26:%.*]]
+// CHECK10:       omp_section_loop.body.case25.section.after26:
+// CHECK10-NEXT:    br label [[OMP_SECTION_LOOP_BODY_CASE25_SECTION_AFTER:%.*]]
+// CHECK10:       omp_section_loop.body.case25.section.after:
+// CHECK10-NEXT:    br label [[OMP_SECTION_LOOP_BODY16_SECTIONS_AFTER]]
+// CHECK10:       omp_section_loop.body16.sections.after:
+// CHECK10-NEXT:    br label [[OMP_SECTION_LOOP_INC17]]
 // CHECK10:       omp_section_loop.inc17:
 // CHECK10-NEXT:    [[OMP_SECTION_LOOP_NEXT22]] = add nuw i32 [[OMP_SECTION_LOOP_IV20]], 1
 // CHECK10-NEXT:    br label [[OMP_SECTION_LOOP_HEADER14]]
 // CHECK10:       omp_section_loop.exit18:
-// CHECK10-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM31]])
-// CHECK10-NEXT:    [[OMP_GLOBAL_THREAD_NUM32:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
-// CHECK10-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB2]], i32 [[OMP_GLOBAL_THREAD_NUM32]])
+// CHECK10-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM32]])
+// CHECK10-NEXT:    [[OMP_GLOBAL_THREAD_NUM33:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
+// CHECK10-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB2]], i32 [[OMP_GLOBAL_THREAD_NUM33]])
 // CHECK10-NEXT:    br label [[OMP_SECTION_LOOP_AFTER19:%.*]]
 // CHECK10:       omp_section_loop.after19:
-// CHECK10-NEXT:    br label [[OMP_SECTIONS_END33:%.*]]
-// CHECK10:       omp_sections.end33:
-// CHECK10-NEXT:    [[TMP14:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
-// CHECK10-NEXT:    store i32 [[TMP14]], i32* [[DOTCAPTURE_EXPR_]], align 4
-// CHECK10-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
-// CHECK10-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP15]], 0
+// CHECK10-NEXT:    br label [[OMP_SECTION_LOOP_AFTER19SECTIONS_FINI:%.*]]
+// CHECK10:       omp_section_loop.after19sections.fini:
+// CHECK10-NEXT:    [[TMP20:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
+// CHECK10-NEXT:    store i32 [[TMP20]], i32* [[DOTCAPTURE_EXPR_]], align 4
+// CHECK10-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
+// CHECK10-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP21]], 0
 // CHECK10-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
 // CHECK10-NEXT:    [[SUB35:%.*]] = sub nsw i32 [[DIV]], 1
 // CHECK10-NEXT:    store i32 [[SUB35]], i32* [[DOTCAPTURE_EXPR_34]], align 4
 // CHECK10-NEXT:    store i32 0, i32* [[I]], align 4
-// CHECK10-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
-// CHECK10-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP16]]
+// CHECK10-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
+// CHECK10-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP22]]
 // CHECK10-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
 // CHECK10:       omp.precond.then:
 // CHECK10-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
-// CHECK10-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_34]], align 4
-// CHECK10-NEXT:    store i32 [[TMP17]], i32* [[DOTOMP_UB]], align 4
+// CHECK10-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_34]], align 4
+// CHECK10-NEXT:    store i32 [[TMP23]], i32* [[DOTOMP_UB]], align 4
 // CHECK10-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
 // CHECK10-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK10-NEXT:    [[OMP_GLOBAL_THREAD_NUM37:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB6:[0-9]+]])
 // CHECK10-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB4:[0-9]+]], i32 [[OMP_GLOBAL_THREAD_NUM37]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
-// CHECK10-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK10-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_34]], align 4
-// CHECK10-NEXT:    [[CMP38:%.*]] = icmp sgt i32 [[TMP18]], [[TMP19]]
+// CHECK10-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
+// CHECK10-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_34]], align 4
+// CHECK10-NEXT:    [[CMP38:%.*]] = icmp sgt i32 [[TMP24]], [[TMP25]]
 // CHECK10-NEXT:    br i1 [[CMP38]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK10:       cond.true:
-// CHECK10-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_34]], align 4
+// CHECK10-NEXT:    [[TMP26:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_34]], align 4
 // CHECK10-NEXT:    br label [[COND_END:%.*]]
 // CHECK10:       cond.false:
-// CHECK10-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
+// CHECK10-NEXT:    [[TMP27:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK10-NEXT:    br label [[COND_END]]
 // CHECK10:       cond.end:
-// CHECK10-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP20]], [[COND_TRUE]] ], [ [[TMP21]], [[COND_FALSE]] ]
+// CHECK10-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP26]], [[COND_TRUE]] ], [ [[TMP27]], [[COND_FALSE]] ]
 // CHECK10-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
-// CHECK10-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
-// CHECK10-NEXT:    store i32 [[TMP22]], i32* [[DOTOMP_IV]], align 4
+// CHECK10-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
+// CHECK10-NEXT:    store i32 [[TMP28]], i32* [[DOTOMP_IV]], align 4
 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK10:       omp.inner.for.cond:
-// CHECK10-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
-// CHECK10-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
-// CHECK10-NEXT:    [[CMP39:%.*]] = icmp sle i32 [[TMP23]], [[TMP24]]
+// CHECK10-NEXT:    [[TMP29:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
+// CHECK10-NEXT:    [[TMP30:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
+// CHECK10-NEXT:    [[CMP39:%.*]] = icmp sle i32 [[TMP29]], [[TMP30]]
 // CHECK10-NEXT:    br i1 [[CMP39]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK10:       omp.inner.for.body:
-// CHECK10-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
-// CHECK10-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP25]], 1
+// CHECK10-NEXT:    [[TMP31:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
+// CHECK10-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP31]], 1
 // CHECK10-NEXT:    [[ADD40:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK10-NEXT:    store i32 [[ADD40]], i32* [[I36]], align 4
-// CHECK10-NEXT:    [[TMP26:%.*]] = load float, float* @flag, align 4
-// CHECK10-NEXT:    [[TOBOOL41:%.*]] = fcmp une float [[TMP26]], 0.000000e+00
+// CHECK10-NEXT:    [[TMP32:%.*]] = load float, float* @flag, align 4
+// CHECK10-NEXT:    [[TOBOOL41:%.*]] = fcmp une float [[TMP32]], 0.000000e+00
 // CHECK10-NEXT:    br i1 [[TOBOOL41]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
 // CHECK10:       omp_if.then:
 // CHECK10-NEXT:    [[OMP_GLOBAL_THREAD_NUM42:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB8:[0-9]+]])
-// CHECK10-NEXT:    [[TMP27:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM42]], i32 2)
-// CHECK10-NEXT:    [[TMP28:%.*]] = icmp ne i32 [[TMP27]], 0
-// CHECK10-NEXT:    br i1 [[TMP28]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
+// CHECK10-NEXT:    [[TMP33:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM42]], i32 2)
+// CHECK10-NEXT:    [[TMP34:%.*]] = icmp ne i32 [[TMP33]], 0
+// CHECK10-NEXT:    br i1 [[TMP34]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
 // CHECK10:       .cancel.exit:
 // CHECK10-NEXT:    br label [[CANCEL_EXIT:%.*]]
-// CHECK10:       omp_section_loop.body.case:
-// CHECK10-NEXT:    [[OMP_GLOBAL_THREAD_NUM10:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
-// CHECK10-NEXT:    [[TMP29:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM10]], i32 3)
-// CHECK10-NEXT:    [[TMP30:%.*]] = icmp eq i32 [[TMP29]], 0
-// CHECK10-NEXT:    br i1 [[TMP30]], label [[OMP_SECTION_LOOP_BODY_CASE_SPLIT:%.*]], label [[OMP_SECTION_LOOP_BODY_CASE_CNCL:%.*]]
-// CHECK10:       omp_section_loop.body.case.split:
-// CHECK10-NEXT:    br label [[OMP_SECTION_LOOP_EXIT]]
 // CHECK10:       omp_section_loop.body.case.cncl:
 // CHECK10-NEXT:    br label [[OMP_SECTION_LOOP_EXIT]]
-// CHECK10:       omp_section_loop.body.case23:
-// CHECK10-NEXT:    [[OMP_GLOBAL_THREAD_NUM24:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
-// CHECK10-NEXT:    [[TMP31:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM24]], i32 3)
-// CHECK10-NEXT:    [[TMP32:%.*]] = icmp eq i32 [[TMP31]], 0
-// CHECK10-NEXT:    br i1 [[TMP32]], label [[OMP_SECTION_LOOP_BODY_CASE23_SPLIT:%.*]], label [[OMP_SECTION_LOOP_BODY_CASE23_CNCL:%.*]]
-// CHECK10:       omp_section_loop.body.case23.split:
-// CHECK10-NEXT:    br label [[OMP_SECTION_LOOP_EXIT18]]
 // CHECK10:       omp_section_loop.body.case23.cncl:
 // CHECK10-NEXT:    br label [[OMP_SECTION_LOOP_EXIT18]]
-// CHECK10:       omp_section_loop.body.case25:
-// CHECK10-NEXT:    [[OMP_GLOBAL_THREAD_NUM26:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
-// CHECK10-NEXT:    [[TMP33:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM26]], i32 3)
-// CHECK10-NEXT:    [[TMP34:%.*]] = icmp eq i32 [[TMP33]], 0
-// CHECK10-NEXT:    br i1 [[TMP34]], label [[OMP_SECTION_LOOP_BODY_CASE25_SPLIT:%.*]], label [[OMP_SECTION_LOOP_BODY_CASE25_CNCL:%.*]]
-// CHECK10:       omp_section_loop.body.case25.split:
-// CHECK10-NEXT:    br label [[OMP_SECTION_LOOP_EXIT18]]
 // CHECK10:       omp_section_loop.body.case25.cncl:
 // CHECK10-NEXT:    br label [[OMP_SECTION_LOOP_EXIT18]]
 // CHECK10:       .cancel.continue:
 // CHECK10-NEXT:    br label [[OMP_IF_END:%.*]]
 // CHECK10:       omp_if.else:
 // CHECK10-NEXT:    br label [[OMP_IF_END]]
 // CHECK10:       omp_if.end:
 // CHECK10-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]

 // CHECK10-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds i8*, i8** [[TMP11]], i64 0
 // CHECK10-NEXT:    [[TMP12:%.*]] = load i8*, i8** [[ARRAYIDX6]], align 8
 // CHECK10-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds i8, i8* [[TMP12]], i64 0
 // CHECK10-NEXT:    [[TMP13:%.*]] = load i8, i8* [[ARRAYIDX7]], align 1
 // CHECK10-NEXT:    [[CONV8:%.*]] = sext i8 [[TMP13]] to i32
 // CHECK10-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV8]], [[TMP10]]
 // CHECK10-NEXT:    [[CONV9:%.*]] = trunc i32 [[ADD]] to i8
 // CHECK10-NEXT:    store i8 [[CONV9]], i8* [[ARRAYIDX7]], align 1
+// CHECK10-NEXT:    br label [[OMP_PAR_REGION_PARALLEL_AFTER:%.*]]
+// CHECK10:       omp.par.region.parallel.after:
 // CHECK10-NEXT:    br label [[OMP_PAR_PRE_FINALIZE:%.*]]
 // CHECK10:       omp.par.pre_finalize:
 // CHECK10-NEXT:    br label [[OMP_PAR_OUTLINED_EXIT_EXITSTUB]]
 // CHECK10:       14:
 // CHECK10-NEXT:    [[OMP_GLOBAL_THREAD_NUM1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
 // CHECK10-NEXT:    [[TMP15:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM1]], i32 1)
 // CHECK10-NEXT:    [[TMP16:%.*]] = icmp eq i32 [[TMP15]], 0
 // CHECK10-NEXT:    br i1 [[TMP16]], label [[DOTSPLIT:%.*]], label [[DOTCNCL:%.*]]

 // CHECK10:       .omp.sections.case.cncl:
 // CHECK10-NEXT:    br label [[CANCEL_CONT:%.*]]
 // CHECK10:       .omp.sections.case2:
 // CHECK10-NEXT:    [[OMP_GLOBAL_THREAD_NUM3:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
 // CHECK10-NEXT:    [[TMP9:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM3]], i32 3)
 // CHECK10-NEXT:    [[TMP10:%.*]] = icmp eq i32 [[TMP9]], 0
 // CHECK10-NEXT:    br i1 [[TMP10]], label [[DOTOMP_SECTIONS_CASE2_SPLIT:%.*]], label [[DOTOMP_SECTIONS_CASE2_CNCL:%.*]]
 // CHECK10:       .omp.sections.case2.split:
+// CHECK10-NEXT:    br label [[DOTOMP_SECTIONS_CASE2_SECTION_AFTER:%.*]]
+// CHECK10:       .omp.sections.case2.section.after:
 // CHECK10-NEXT:    br label [[DOTOMP_SECTIONS_EXIT]]
 // CHECK10:       .omp.sections.case2.cncl:
 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_END]]
 // CHECK10:       .omp.sections.exit:
 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK10:       omp.inner.for.inc:
 // CHECK10-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
 // CHECK10-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP11]], 1

clang/test/OpenMP/critical_codegen.cpp

 // TERM_DEBUG-LABEL: @main
 int main() {
   // ALL:       [[A_ADDR:%.+]] = alloca i8
   char a;
 
 // ALL:             [[GTID:%.+]] = call {{.*}}i32 @__kmpc_global_thread_num([[IDENT_T_TY]]* [[DEFAULT_LOC:@.+]])
 // ALL:             call {{.*}}void @__kmpc_critical([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 [[GTID]], [8 x i32]* [[UNNAMED_LOCK]])
 // ALL-NEXT:        store i8 2, i8* [[A_ADDR]]
+// IRBUILDER-NEXT:    br label %[[AFTER:[^ ,]+]]
+// IRBUILDER:     [[AFTER]]
 // ALL-NEXT:        call {{.*}}void @__kmpc_end_critical([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 [[GTID]], [8 x i32]* [[UNNAMED_LOCK]])
 #pragma omp critical
   a = 2;
 // IRBUILDER:       [[GTID:%.+]] = call {{.*}}i32 @__kmpc_global_thread_num([[IDENT_T_TY]]* [[DEFAULT_LOC:@.+]])
 // ALL:             call {{.*}}void @__kmpc_critical([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 [[GTID]], [8 x i32]* [[THE_NAME_LOCK]])
 // IRBUILDER-NEXT:  call {{.*}}void [[FOO]]()
 // NORMAL-NEXT:     invoke {{.*}}void [[FOO]]()
+// IRBUILDER-NEXT:    br label %[[AFTER:[^ ,]+]]
+// IRBUILDER:     [[AFTER]]
 // ALL:             call {{.*}}void @__kmpc_end_critical([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 [[GTID]], [8 x i32]* [[THE_NAME_LOCK]])
 #pragma omp critical(the_name)
   foo();
 // IRBUILDER:       [[GTID:%.+]] = call {{.*}}i32 @__kmpc_global_thread_num([[IDENT_T_TY]]* [[DEFAULT_LOC:@.+]])
 // ALL:             call {{.*}}void @__kmpc_critical_with_hint([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 [[GTID]], [8 x i32]* [[THE_NAME_LOCK1]], i{{64|32}} 23)
 // IRBUILDER-NEXT:  call {{.*}}void [[FOO]]()
 // NORMAL-NEXT:     invoke {{.*}}void [[FOO]]()
+// IRBUILDER-NEXT:    br label %[[AFTER:[^ ,]+]]
+// IRBUILDER:     [[AFTER]]
 // ALL:             call {{.*}}void @__kmpc_end_critical([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 [[GTID]], [8 x i32]* [[THE_NAME_LOCK1]])
 #pragma omp critical(the_name1) hint(23)
   foo();
   // IRBUILDER:       [[GTID:%.+]] = call {{.*}}i32 @__kmpc_global_thread_num([[IDENT_T_TY]]* [[DEFAULT_LOC:@.+]])
   // ALL:       call {{.*}}void @__kmpc_critical([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 [[GTID]], [8 x i32]* [[THE_NAME_LOCK]])
-  // ALL:       br label
-  // ALL-NOT:   call {{.*}}void @__kmpc_end_critical(
-  // ALL:       br label
-  // ALL-NOT:   call {{.*}}void @__kmpc_end_critical(
+  // NORMAL:       br label
+  // NORMAL-NOT:   call {{.*}}void @__kmpc_end_critical(
+  // NORMAL:       br label
+  // NORMAL-NOT:   call {{.*}}void @__kmpc_end_critical(
   // NORMAL:       br label
   if (a)
 #pragma omp critical(the_name)
     while (1)
       ;
   // ALL:  call {{.*}}void [[FOO]]()
   foo();
   // ALL-NOT:   call void @__kmpc_critical

clang/test/OpenMP/critical_codegen_attr.cpp

 // TERM_DEBUG-LABEL: @main
 int main() {
   // ALL:       [[A_ADDR:%.+]] = alloca i8
   char a;
 
 // ALL:             [[GTID:%.+]] = call {{.*}}i32 @__kmpc_global_thread_num([[IDENT_T_TY]]* [[DEFAULT_LOC:@.+]])
 // ALL:             call {{.*}}void @__kmpc_critical([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 [[GTID]], [8 x i32]* [[UNNAMED_LOCK]])
 // ALL-NEXT:        store i8 2, i8* [[A_ADDR]]
+// IRBUILDER-NEXT:    br label %[[AFTER:[^ ,]+]]
+// IRBUILDER:     [[AFTER]]
 // ALL-NEXT:        call {{.*}}void @__kmpc_end_critical([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 [[GTID]], [8 x i32]* [[UNNAMED_LOCK]])
   [[omp::directive(critical)]]
   a = 2;
 // IRBUILDER:       [[GTID:%.+]] = call {{.*}}i32 @__kmpc_global_thread_num([[IDENT_T_TY]]* [[DEFAULT_LOC:@.+]])
 // ALL:             call {{.*}}void @__kmpc_critical([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 [[GTID]], [8 x i32]* [[THE_NAME_LOCK]])
 // IRBUILDER-NEXT:  call {{.*}}void [[FOO]]()
 // NORMAL-NEXT:     invoke {{.*}}void [[FOO]]()
+// IRBUILDER-NEXT:    br label %[[AFTER:[^ ,]+]]
+// IRBUILDER:     [[AFTER]]
 // ALL:             call {{.*}}void @__kmpc_end_critical([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 [[GTID]], [8 x i32]* [[THE_NAME_LOCK]])
   [[omp::directive(critical(the_name))]]
   foo();
 // IRBUILDER:       [[GTID:%.+]] = call {{.*}}i32 @__kmpc_global_thread_num([[IDENT_T_TY]]* [[DEFAULT_LOC:@.+]])
 // ALL:             call {{.*}}void @__kmpc_critical_with_hint([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 [[GTID]], [8 x i32]* [[THE_NAME_LOCK1]], i{{64|32}} 23)
 // IRBUILDER-NEXT:  call {{.*}}void [[FOO]]()
 // NORMAL-NEXT:     invoke {{.*}}void [[FOO]]()
+// IRBUILDER-NEXT:    br label %[[AFTER:[^ ,]+]]
+// IRBUILDER:     [[AFTER]]
 // ALL:             call {{.*}}void @__kmpc_end_critical([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 [[GTID]], [8 x i32]* [[THE_NAME_LOCK1]])
   [[omp::directive(critical(the_name1) hint(23))]]
   foo();
   // IRBUILDER:       [[GTID:%.+]] = call {{.*}}i32 @__kmpc_global_thread_num([[IDENT_T_TY]]* [[DEFAULT_LOC:@.+]])
   // ALL:       call {{.*}}void @__kmpc_critical([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 [[GTID]], [8 x i32]* [[THE_NAME_LOCK]])
-  // ALL:       br label
-  // ALL-NOT:   call {{.*}}void @__kmpc_end_critical(
-  // ALL:       br label
-  // ALL-NOT:   call {{.*}}void @__kmpc_end_critical(
+  // NORMAL:       br label
+  // NORMAL-NOT:   call {{.*}}void @__kmpc_end_critical(
+  // NORMAL:       br label
+  // NORMAL-NOT:   call {{.*}}void @__kmpc_end_critical(
   // NORMAL:       br label
   if (a)
     [[omp::directive(critical(the_name))]]
     while (1)
       ;
   // ALL:  call {{.*}}void [[FOO]]()
   foo();
   // ALL-NOT:   call void @__kmpc_critical

clang/test/OpenMP/masked_codegen.cpp

int main() {
   char a;
 
 // ALL:             [[GTID:%.+]] = call {{.*}}i32 @__kmpc_global_thread_num([[IDENT_T_TY]]* [[DEFAULT_LOC:@.+]])
 // ALL:             [[RES:%.+]] = call {{.*}}i32 @__kmpc_masked([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 [[GTID]], i32 0)
 // ALL-NEXT:        [[IS_MASKED:%.+]] = icmp ne i32 [[RES]], 0
 // ALL-NEXT:        br i1 [[IS_MASKED]], label {{%?}}[[THEN:.+]], label {{%?}}[[EXIT:.+]]
 // ALL:             [[THEN]]
 // ALL-NEXT:        store i8 2, i8* [[A_ADDR]]
+// IRBUILDER-NEXT:    br label %[[AFTER:[^ ,]+]]
+// IRBUILDER:     [[AFTER]]
 // ALL-NEXT:        call {{.*}}void @__kmpc_end_masked([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 [[GTID]])
 // ALL-NEXT:        br label {{%?}}[[EXIT]]
 // ALL:             [[EXIT]]
 #pragma omp masked
   a = 2;
 // IRBUILDER:       [[GTID:%.+]] = call {{.*}}i32 @__kmpc_global_thread_num([[IDENT_T_TY]]* [[DEFAULT_LOC:@.+]])
 // ALL:             [[RES:%.+]] = call {{.*}}i32 @__kmpc_masked([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 [[GTID]], i32 2)
 // ALL-NEXT:        [[IS_MASKED:%.+]] = icmp ne i32 [[RES]], 0

clang/test/OpenMP/master_codegen.cpp

int main() {
   char a;
 
 // ALL:             [[GTID:%.+]] = call {{.*}}i32 @__kmpc_global_thread_num([[IDENT_T_TY]]* [[DEFAULT_LOC:@.+]])
 // ALL:             [[RES:%.+]] = call {{.*}}i32 @__kmpc_master([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 [[GTID]])
 // ALL-NEXT:        [[IS_MASTER:%.+]] = icmp ne i32 [[RES]], 0
 // ALL-NEXT:        br i1 [[IS_MASTER]], label {{%?}}[[THEN:.+]], label {{%?}}[[EXIT:.+]]
 // ALL:             [[THEN]]
 // ALL-NEXT:        store i8 2, i8* [[A_ADDR]]
+// IRBUILDER-NEXT:    br label %[[AFTER:[^ ,]+]]
+// IRBUILDER:     [[AFTER]]
 // ALL-NEXT:        call {{.*}}void @__kmpc_end_master([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 [[GTID]])
 // ALL-NEXT:        br label {{%?}}[[EXIT]]
 // ALL:             [[EXIT]]
 #pragma omp master
   a = 2;
 // IRBUILDER:       [[GTID:%.+]] = call {{.*}}i32 @__kmpc_global_thread_num([[IDENT_T_TY]]* [[DEFAULT_LOC:@.+]])
 // ALL:             [[RES:%.+]] = call {{.*}}i32 @__kmpc_master([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 [[GTID]])
 // ALL-NEXT:        [[IS_MASTER:%.+]] = icmp ne i32 [[RES]], 0

clang/test/OpenMP/ordered_codegen.cpp

 // NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py UTC_ARGS: --function-signature --include-generated-funcs --replace-value-regex "__omp_offloading_[0-9a-z]+_[0-9a-z]+" "reduction_size[.].+[.]" "pl_cond[.].+[.|,]" --prefix-filecheck-ir-name _
-// RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp -x c++ -triple x86_64-unknown-unknown -emit-llvm %s -fexceptions -fcxx-exceptions -o - | FileCheck %s --check-prefixes=CHECK1,CHECK1-NORMAL
+// RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp -x c++ -triple x86_64-unknown-unknown -emit-llvm %s -fexceptions -fcxx-exceptions -o - | FileCheck %s --check-prefixes=CHECK1
 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp -x c++ -std=c++11 -triple x86_64-unknown-unknown -fexceptions -fcxx-exceptions -emit-pch -o %t %s
-// RUN: %clang_cc1 -no-opaque-pointers -fopenmp -x c++ -triple x86_64-unknown-unknown -fexceptions -fcxx-exceptions -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefixes=CHECK2,CHECK2-NORMAL
+// RUN: %clang_cc1 -no-opaque-pointers -fopenmp -x c++ -triple x86_64-unknown-unknown -fexceptions -fcxx-exceptions -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefixes=CHECK2
 
-// RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp -fopenmp-enable-irbuilder -x c++ -triple x86_64-unknown-unknown -emit-llvm %s -fexceptions -fcxx-exceptions -o - | FileCheck %s --check-prefixes=CHECK1,CHECK1-IRBUILDER
+// RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp -fopenmp-enable-irbuilder -x c++ -triple x86_64-unknown-unknown -emit-llvm %s -fexceptions -fcxx-exceptions -o - | FileCheck %s --check-prefixes=CHECK1-IRBUILDER
 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp -fopenmp-enable-irbuilder -x c++ -std=c++11 -triple x86_64-unknown-unknown -fexceptions -fcxx-exceptions -emit-pch -o %t %s
-// RUN: %clang_cc1 -no-opaque-pointers -fopenmp -fopenmp-enable-irbuilder -x c++ -triple x86_64-unknown-unknown -fexceptions -fcxx-exceptions -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefixes=CHECK2,CHECK2-IRBUILDER
+// RUN: %clang_cc1 -no-opaque-pointers -fopenmp -fopenmp-enable-irbuilder -x c++ -triple x86_64-unknown-unknown -fexceptions -fcxx-exceptions -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefixes=CHECK2-IRBUILDER
 
-// RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp -x c++ -triple x86_64-unknown-unknown -emit-llvm %s -fexceptions -fcxx-exceptions -fopenmp-version=45 -o - | FileCheck %s --check-prefixes=CHECK3,CHECK3-NORMAL
+// RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp -x c++ -triple x86_64-unknown-unknown -emit-llvm %s -fexceptions -fcxx-exceptions -fopenmp-version=45 -o - | FileCheck %s --check-prefixes=CHECK3
 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp -x c++ -std=c++11 -triple x86_64-unknown-unknown -fexceptions -fcxx-exceptions -emit-pch -fopenmp-version=45 -o %t %s
-// RUN: %clang_cc1 -no-opaque-pointers -fopenmp -fopenmp-version=45 -x c++ -triple x86_64-unknown-unknown -fexceptions -fcxx-exceptions -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefixes=CHECK4,CHECK4-NORMAL
+// RUN: %clang_cc1 -no-opaque-pointers -fopenmp -fopenmp-version=45 -x c++ -triple x86_64-unknown-unknown -fexceptions -fcxx-exceptions -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefixes=CHECK4
 
-// RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp -fopenmp-enable-irbuilder -x c++ -triple x86_64-unknown-unknown -emit-llvm %s -fexceptions -fcxx-exceptions -fopenmp-version=45 -o - | FileCheck %s --check-prefixes=CHECK3,CHECK3-IRBUILDER
+// RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp -fopenmp-enable-irbuilder -x c++ -triple x86_64-unknown-unknown -emit-llvm %s -fexceptions -fcxx-exceptions -fopenmp-version=45 -o - | FileCheck %s --check-prefixes=CHECK3-IRBUILDER
 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp -fopenmp-enable-irbuilder -x c++ -std=c++11 -triple x86_64-unknown-unknown -fexceptions -fcxx-exceptions -emit-pch -fopenmp-version=45 -o %t %s
-// RUN: %clang_cc1 -no-opaque-pointers -fopenmp -fopenmp-enable-irbuilder -fopenmp-version=45 -x c++ -triple x86_64-unknown-unknown -fexceptions -fcxx-exceptions -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefixes=CHECK4,CHECK4-IRBUILDER
+// RUN: %clang_cc1 -no-opaque-pointers -fopenmp -fopenmp-enable-irbuilder -fopenmp-version=45 -x c++ -triple x86_64-unknown-unknown -fexceptions -fcxx-exceptions -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefixes=CHECK4-IRBUILDER
 
 // RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp-simd -x c++ -triple x86_64-unknown-unknown -emit-llvm %s -fexceptions -fcxx-exceptions -o - | FileCheck %s --check-prefix=CHECK5
 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp-simd -x c++ -std=c++11 -triple x86_64-unknown-unknown -fexceptions -fcxx-exceptions -emit-pch -o %t %s
 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp-simd -x c++ -triple x86_64-unknown-unknown -fexceptions -fcxx-exceptions -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK6
 // expected-no-diagnostics
 #ifndef HEADER
 #define HEADER
 

 // CHECK1-NEXT:    [[D_ADDR:%.*]] = alloca float*, align 8
 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
-// CHECK1-NORMAL-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1:[0-9]+]])
+// CHECK1-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1:[0-9]+]])
 // CHECK1-NEXT:    store float* [[A]], float** [[A_ADDR]], align 8
 // CHECK1-NEXT:    store float* [[B]], float** [[B_ADDR]], align 8
 // CHECK1-NEXT:    store float* [[C]], float** [[C_ADDR]], align 8
 // CHECK1-NEXT:    store float* [[D]], float** [[D_ADDR]], align 8
 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
 // CHECK1-NEXT:    store i32 4571423, i32* [[DOTOMP_UB]], align 4
 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
-// CHECK1-IRBUILDER-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB3:[0-9]+]])
-// CHECK1-IRBUILDER-NEXT:    call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP0]], i32 66, i32 0, i32 4571423, i32 1, i32 1)
-// CHECK1-NORMAL-NEXT:    call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 66, i32 0, i32 4571423, i32 1, i32 1)
+// CHECK1-NEXT:    call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 66, i32 0, i32 4571423, i32 1, i32 1)
 // CHECK1-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK1:       omp.dispatch.cond:
-// CHECK1-IRBUILDER-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB3]])
 // CHECK1-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
 // CHECK1-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP1]], 0
 // CHECK1-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
 // CHECK1:       omp.dispatch.body:
 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK1-NEXT:    store i32 [[TMP2]], i32* [[DOTOMP_IV]], align 4
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK1:       omp.inner.for.cond:
 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP3]], [[TMP4]]
 // CHECK1-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK1:       omp.inner.for.body:
 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP5]], 7
 // CHECK1-NEXT:    [[SUB:%.*]] = sub nsw i32 32000000, [[MUL]]
 // CHECK1-NEXT:    store i32 [[SUB]], i32* [[I]], align 4
-// CHECK1-IRBUILDER-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
 // CHECK1-NEXT:    call void @__kmpc_ordered(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
 // CHECK1-NEXT:    [[TMP6:%.*]] = load float*, float** [[B_ADDR]], align 8
 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[I]], align 4
 // CHECK1-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP7]] to i64
 // CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, float* [[TMP6]], i64 [[IDXPROM]]
 // CHECK1-NEXT:    [[TMP8:%.*]] = load float, float* [[ARRAYIDX]], align 4
 // CHECK1-NEXT:    [[TMP9:%.*]] = load float*, float** [[C_ADDR]], align 8
 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[I]], align 4

 // CHECK1-NEXT:    call void @__kmpc_end_ordered(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK1:       omp.body.continue:
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK1:       omp.inner.for.inc:
 // CHECK1-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP17]], 1
 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
-// CHECK1-IRBUILDER-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB3]])
 // CHECK1-NEXT:    call void @__kmpc_dispatch_fini_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
 // CHECK1:       omp.inner.for.end:
 // CHECK1-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
 // CHECK1:       omp.dispatch.inc:
 // CHECK1-NEXT:    br label [[OMP_DISPATCH_COND]]
 // CHECK1:       omp.dispatch.end:
-// CHECK1-IRBUILDER-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
-// CHECK1-IRBUILDER-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB4:[0-9]+]], i32 [[TMP0]])
-// CHECK1-NORMAL-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP0]])
+// CHECK1-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP0]])
 // CHECK1-NEXT:    ret void
 //
 //
 // CHECK1-LABEL: define {{[^@]+}}@_Z8dynamic1PfS_S_S_
 // CHECK1-SAME: (float* noundef [[A:%.*]], float* noundef [[B:%.*]], float* noundef [[C:%.*]], float* noundef [[D:%.*]]) #[[ATTR0]] {
 // CHECK1-NEXT:  entry:
 // CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca float*, align 8
 // CHECK1-NEXT:    [[B_ADDR:%.*]] = alloca float*, align 8
 // CHECK1-NEXT:    [[C_ADDR:%.*]] = alloca float*, align 8
 // CHECK1-NEXT:    [[D_ADDR:%.*]] = alloca float*, align 8
 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i64, align 8
 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i64, align 8
 // CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i64, align 8
 // CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i64, align 8
 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i64, align 8
 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[I:%.*]] = alloca i64, align 8
-// CHECK1-NORMAL-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
+// CHECK1-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
 // CHECK1-NEXT:    store float* [[A]], float** [[A_ADDR]], align 8
 // CHECK1-NEXT:    store float* [[B]], float** [[B_ADDR]], align 8
 // CHECK1-NEXT:    store float* [[C]], float** [[C_ADDR]], align 8
 // CHECK1-NEXT:    store float* [[D]], float** [[D_ADDR]], align 8
 // CHECK1-NEXT:    store i64 0, i64* [[DOTOMP_LB]], align 8
 // CHECK1-NEXT:    store i64 16908287, i64* [[DOTOMP_UB]], align 8
 // CHECK1-NEXT:    store i64 1, i64* [[DOTOMP_STRIDE]], align 8
 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
-// CHECK1-IRBUILDER-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB6:[0-9]+]])
 // CHECK1-NEXT:    call void @__kmpc_dispatch_init_8u(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 1073741891, i64 0, i64 16908287, i64 1, i64 1)
 // CHECK1-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK1:       omp.dispatch.cond:
-// CHECK1-IRBUILDER-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB6]])
 // CHECK1-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_dispatch_next_8u(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32* [[DOTOMP_IS_LAST]], i64* [[DOTOMP_LB]], i64* [[DOTOMP_UB]], i64* [[DOTOMP_STRIDE]])
 // CHECK1-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP1]], 0
 // CHECK1-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
 // CHECK1:       omp.dispatch.body:
 // CHECK1-NEXT:    [[TMP2:%.*]] = load i64, i64* [[DOTOMP_LB]], align 8
 // CHECK1-NEXT:    store i64 [[TMP2]], i64* [[DOTOMP_IV]], align 8
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK1:       omp.inner.for.cond:
 // CHECK1-NEXT:    [[TMP3:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
 // CHECK1-NEXT:    [[TMP4:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
 // CHECK1-NEXT:    [[ADD:%.*]] = add i64 [[TMP4]], 1
 // CHECK1-NEXT:    [[CMP:%.*]] = icmp ult i64 [[TMP3]], [[ADD]]
 // CHECK1-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK1:       omp.inner.for.body:
 // CHECK1-NEXT:    [[TMP5:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
 // CHECK1-NEXT:    [[MUL:%.*]] = mul i64 [[TMP5]], 127
 // CHECK1-NEXT:    [[ADD1:%.*]] = add i64 131071, [[MUL]]
 // CHECK1-NEXT:    store i64 [[ADD1]], i64* [[I]], align 8
-// CHECK1-IRBUILDER-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
 // CHECK1-NEXT:    call void @__kmpc_ordered(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
 // CHECK1-NEXT:    [[TMP6:%.*]] = load float*, float** [[B_ADDR]], align 8
 // CHECK1-NEXT:    [[TMP7:%.*]] = load i64, i64* [[I]], align 8
 // CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, float* [[TMP6]], i64 [[TMP7]]
 // CHECK1-NEXT:    [[TMP8:%.*]] = load float, float* [[ARRAYIDX]], align 4
 // CHECK1-NEXT:    [[TMP9:%.*]] = load float*, float** [[C_ADDR]], align 8
 // CHECK1-NEXT:    [[TMP10:%.*]] = load i64, i64* [[I]], align 8
 // CHECK1-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds float, float* [[TMP9]], i64 [[TMP10]]

 // CHECK1-NEXT:    call void @__kmpc_end_ordered(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK1:       omp.body.continue:
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK1:       omp.inner.for.inc:
 // CHECK1-NEXT:    [[TMP17:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
 // CHECK1-NEXT:    [[ADD7:%.*]] = add i64 [[TMP17]], 1
 // CHECK1-NEXT:    store i64 [[ADD7]], i64* [[DOTOMP_IV]], align 8
-// CHECK1-IRBUILDER-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB6]])
 // CHECK1-NEXT:    call void @__kmpc_dispatch_fini_8u(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
 // CHECK1:       omp.inner.for.end:
 // CHECK1-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
 // CHECK1:       omp.dispatch.inc:
 // CHECK1-NEXT:    br label [[OMP_DISPATCH_COND]]
 // CHECK1:       omp.dispatch.end:
-// CHECK1-IRBUILDER-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
-// CHECK1-IRBUILDER-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB4]], i32 [[TMP0]])
-// CHECK1-NORMAL-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]])
+// CHECK1-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]])
 // CHECK1-NEXT:    ret void
 //
 //
 // CHECK1-LABEL: define {{[^@]+}}@_Z9test_autoPfS_S_S_
 // CHECK1-SAME: (float* noundef [[A:%.*]], float* noundef [[B:%.*]], float* noundef [[C:%.*]], float* noundef [[D:%.*]]) #[[ATTR0]] {
 // CHECK1-NEXT:  entry:
 // CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca float*, align 8
 // CHECK1-NEXT:    [[B_ADDR:%.*]] = alloca float*, align 8

 // CHECK1-NEXT:    [[I:%.*]] = alloca i8, align 1
 // CHECK1-NEXT:    [[X6:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i64, align 8
 // CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i64, align 8
 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i64, align 8
 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[I8:%.*]] = alloca i8, align 1
 // CHECK1-NEXT:    [[X9:%.*]] = alloca i32, align 4
-// CHECK1-NORMAL-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
+// CHECK1-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
 // CHECK1-NEXT:    store float* [[A]], float** [[A_ADDR]], align 8
 // CHECK1-NEXT:    store float* [[B]], float** [[B_ADDR]], align 8
 // CHECK1-NEXT:    store float* [[C]], float** [[C_ADDR]], align 8
 // CHECK1-NEXT:    store float* [[D]], float** [[D_ADDR]], align 8
 // CHECK1-NEXT:    store i32 0, i32* [[X]], align 4
 // CHECK1-NEXT:    store i32 0, i32* [[Y]], align 4
 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32, i32* [[Y]], align 4
 // CHECK1-NEXT:    [[CONV:%.*]] = trunc i32 [[TMP1]] to i8

 // CHECK1-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
 // CHECK1:       omp.precond.then:
 // CHECK1-NEXT:    store i64 0, i64* [[DOTOMP_LB]], align 8
 // CHECK1-NEXT:    [[TMP5:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR_2]], align 8
 // CHECK1-NEXT:    store i64 [[TMP5]], i64* [[DOTOMP_UB]], align 8
 // CHECK1-NEXT:    store i64 1, i64* [[DOTOMP_STRIDE]], align 8
 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK1-NEXT:    [[TMP6:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR_2]], align 8
-// CHECK1-IRBUILDER-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB8:[0-9]+]])
 // CHECK1-NEXT:    call void @__kmpc_dispatch_init_8(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 1073741894, i64 0, i64 [[TMP6]], i64 1, i64 1)
 // CHECK1-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK1:       omp.dispatch.cond:
-// CHECK1-IRBUILDER-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB8]])
 // CHECK1-NEXT:    [[TMP7:%.*]] = call i32 @__kmpc_dispatch_next_8(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32* [[DOTOMP_IS_LAST]], i64* [[DOTOMP_LB]], i64* [[DOTOMP_UB]], i64* [[DOTOMP_STRIDE]])
 // CHECK1-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP7]], 0
 // CHECK1-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
 // CHECK1:       omp.dispatch.body:
 // CHECK1-NEXT:    [[TMP8:%.*]] = load i64, i64* [[DOTOMP_LB]], align 8
 // CHECK1-NEXT:    store i64 [[TMP8]], i64* [[DOTOMP_IV]], align 8
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK1:       omp.inner.for.cond:

 // CHECK1-NEXT:    [[TMP14:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
 // CHECK1-NEXT:    [[DIV16:%.*]] = sdiv i64 [[TMP14]], 11
 // CHECK1-NEXT:    [[MUL17:%.*]] = mul nsw i64 [[DIV16]], 11
 // CHECK1-NEXT:    [[SUB18:%.*]] = sub nsw i64 [[TMP13]], [[MUL17]]
 // CHECK1-NEXT:    [[MUL19:%.*]] = mul nsw i64 [[SUB18]], 1
 // CHECK1-NEXT:    [[SUB20:%.*]] = sub nsw i64 11, [[MUL19]]
 // CHECK1-NEXT:    [[CONV21:%.*]] = trunc i64 [[SUB20]] to i32
 // CHECK1-NEXT:    store i32 [[CONV21]], i32* [[X9]], align 4
-// CHECK1-IRBUILDER-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
 // CHECK1-NEXT:    call void @__kmpc_ordered(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
 // CHECK1-NEXT:    [[TMP15:%.*]] = load float*, float** [[B_ADDR]], align 8
 // CHECK1-NEXT:    [[TMP16:%.*]] = load i8, i8* [[I8]], align 1
 // CHECK1-NEXT:    [[IDXPROM:%.*]] = sext i8 [[TMP16]] to i64
 // CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, float* [[TMP15]], i64 [[IDXPROM]]
 // CHECK1-NEXT:    [[TMP17:%.*]] = load float, float* [[ARRAYIDX]], align 4
 // CHECK1-NEXT:    [[TMP18:%.*]] = load float*, float** [[C_ADDR]], align 8
 // CHECK1-NEXT:    [[TMP19:%.*]] = load i8, i8* [[I8]], align 1

 // CHECK1-NEXT:    call void @__kmpc_end_ordered(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK1:       omp.body.continue:
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK1:       omp.inner.for.inc:
 // CHECK1-NEXT:    [[TMP26:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
 // CHECK1-NEXT:    [[ADD30:%.*]] = add nsw i64 [[TMP26]], 1
 // CHECK1-NEXT:    store i64 [[ADD30]], i64* [[DOTOMP_IV]], align 8
-// CHECK1-IRBUILDER-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB8]])
 // CHECK1-NEXT:    call void @__kmpc_dispatch_fini_8(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
 // CHECK1:       omp.inner.for.end:
 // CHECK1-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
 // CHECK1:       omp.dispatch.inc:
 // CHECK1-NEXT:    br label [[OMP_DISPATCH_COND]]
 // CHECK1:       omp.dispatch.end:
 // CHECK1-NEXT:    br label [[OMP_PRECOND_END]]
 // CHECK1:       omp.precond.end:
-// CHECK1-IRBUILDER-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
-// CHECK1-IRBUILDER-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB4]], i32 [[TMP0]])
-// CHECK1-NORMAL-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]])
+// CHECK1-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]])
 // CHECK1-NEXT:    ret void
 //
 //
 // CHECK1-LABEL: define {{[^@]+}}@_Z7runtimePfS_S_S_
 // CHECK1-SAME: (float* noundef [[A:%.*]], float* noundef [[B:%.*]], float* noundef [[C:%.*]], float* noundef [[D:%.*]]) #[[ATTR0]] {
 // CHECK1-NEXT:  entry:
 // CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca float*, align 8
 // CHECK1-NEXT:    [[B_ADDR:%.*]] = alloca float*, align 8
 // CHECK1-NEXT:    [[C_ADDR:%.*]] = alloca float*, align 8
 // CHECK1-NEXT:    [[D_ADDR:%.*]] = alloca float*, align 8
 // CHECK1-NEXT:    [[X:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i8, align 1
 // CHECK1-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[I:%.*]] = alloca i8, align 1
 // CHECK1-NEXT:    [[X2:%.*]] = alloca i32, align 4
-// CHECK1-NORMAL-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
+// CHECK1-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
 // CHECK1-NEXT:    store float* [[A]], float** [[A_ADDR]], align 8
 // CHECK1-NEXT:    store float* [[B]], float** [[B_ADDR]], align 8
 // CHECK1-NEXT:    store float* [[C]], float** [[C_ADDR]], align 8
 // CHECK1-NEXT:    store float* [[D]], float** [[D_ADDR]], align 8
 // CHECK1-NEXT:    store i32 0, i32* [[X]], align 4
 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
 // CHECK1-NEXT:    store i32 199, i32* [[DOTOMP_UB]], align 4
 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
-// CHECK1-IRBUILDER-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB10:[0-9]+]])
 // CHECK1-NEXT:    call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 1073741893, i32 0, i32 199, i32 1, i32 1)
 // CHECK1-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK1:       omp.dispatch.cond:
-// CHECK1-IRBUILDER-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB10]])
 // CHECK1-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
 // CHECK1-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP1]], 0
 // CHECK1-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
 // CHECK1:       omp.dispatch.body:
 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK1-NEXT:    store i32 [[TMP2]], i32* [[DOTOMP_IV]], align 4
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK1:       omp.inner.for.cond:

 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK1-NEXT:    [[DIV3:%.*]] = sdiv i32 [[TMP7]], 20
 // CHECK1-NEXT:    [[MUL4:%.*]] = mul nsw i32 [[DIV3]], 20
 // CHECK1-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP6]], [[MUL4]]
 // CHECK1-NEXT:    [[MUL5:%.*]] = mul nsw i32 [[SUB]], 1
 // CHECK1-NEXT:    [[ADD6:%.*]] = add nsw i32 -10, [[MUL5]]
 // CHECK1-NEXT:    store i32 [[ADD6]], i32* [[X2]], align 4
-// CHECK1-IRBUILDER-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
 // CHECK1-NEXT:    call void @__kmpc_ordered(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
 // CHECK1-NEXT:    [[TMP8:%.*]] = load float*, float** [[B_ADDR]], align 8
 // CHECK1-NEXT:    [[TMP9:%.*]] = load i8, i8* [[I]], align 1
 // CHECK1-NEXT:    [[IDXPROM:%.*]] = zext i8 [[TMP9]] to i64
 // CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, float* [[TMP8]], i64 [[IDXPROM]]
 // CHECK1-NEXT:    [[TMP10:%.*]] = load float, float* [[ARRAYIDX]], align 4
 // CHECK1-NEXT:    [[TMP11:%.*]] = load float*, float** [[C_ADDR]], align 8
 // CHECK1-NEXT:    [[TMP12:%.*]] = load i8, i8* [[I]], align 1

 // CHECK1-NEXT:    call void @__kmpc_end_ordered(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK1:       omp.body.continue:
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK1:       omp.inner.for.inc:
 // CHECK1-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK1-NEXT:    [[ADD15:%.*]] = add nsw i32 [[TMP19]], 1
 // CHECK1-NEXT:    store i32 [[ADD15]], i32* [[DOTOMP_IV]], align 4
-// CHECK1-IRBUILDER-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB10]])
 // CHECK1-NEXT:    call void @__kmpc_dispatch_fini_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
 // CHECK1:       omp.inner.for.end:
 // CHECK1-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
 // CHECK1:       omp.dispatch.inc:
 // CHECK1-NEXT:    br label [[OMP_DISPATCH_COND]]
 // CHECK1:       omp.dispatch.end:
-// CHECK1-IRBUILDER-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
-// CHECK1-IRBUILDER-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB4]], i32 [[TMP0]])
-// CHECK1-NORMAL-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]])
+// CHECK1-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]])
 // CHECK1-NEXT:    ret void
 //
 //
 // CHECK1-LABEL: define {{[^@]+}}@_Z8foo_simdii
 // CHECK1-SAME: (i32 noundef [[LOW:%.*]], i32 noundef [[UP:%.*]]) #[[ATTR0]] {
 // CHECK1-NEXT:  entry:
 // CHECK1-NEXT:    [[LOW_ADDR:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[UP_ADDR:%.*]] = alloca i32, align 4

 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR_19:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR_20:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[I26:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[I28:%.*]] = alloca i32, align 4
-// CHECK1-NORMAL-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
+// CHECK1-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
 // CHECK1-NEXT:    store i32 [[LOW]], i32* [[LOW_ADDR]], align 4
 // CHECK1-NEXT:    store i32 [[UP]], i32* [[UP_ADDR]], align 4
 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32, i32* [[LOW_ADDR]], align 4
 // CHECK1-NEXT:    store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[UP_ADDR]], align 4
 // CHECK1-NEXT:    store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4

 // CHECK1-NEXT:    br i1 [[CMP27]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
 // CHECK1:       omp.precond.then:
 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
 // CHECK1-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_20]], align 4
 // CHECK1-NEXT:    store i32 [[TMP24]], i32* [[DOTOMP_UB]], align 4
 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK1-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_20]], align 4
-// CHECK1-IRBUILDER-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB12:[0-9]+]])
 // CHECK1-NEXT:    call void @__kmpc_dispatch_init_4u(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 66, i32 0, i32 [[TMP25]], i32 1, i32 1)
 // CHECK1-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK1:       omp.dispatch.cond:
-// CHECK1-IRBUILDER-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB12]])
 // CHECK1-NEXT:    [[TMP26:%.*]] = call i32 @__kmpc_dispatch_next_4u(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
 // CHECK1-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP26]], 0
 // CHECK1-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
 // CHECK1:       omp.dispatch.body:
 // CHECK1-NEXT:    [[TMP27:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK1-NEXT:    store i32 [[TMP27]], i32* [[DOTOMP_IV16]], align 4
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND29:%.*]]
-// CHECK1-IRBUILDER:       omp.inner.for.cond30:
-// CHECK1-NORMAL:       omp.inner.for.cond29:
+// CHECK1:       omp.inner.for.cond29:
 // CHECK1-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTOMP_IV16]], align 4, !llvm.access.group !7
 // CHECK1-NEXT:    [[TMP29:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !7
 // CHECK1-NEXT:    [[ADD30:%.*]] = add i32 [[TMP29]], 1
 // CHECK1-NEXT:    [[CMP31:%.*]] = icmp ult i32 [[TMP28]], [[ADD30]]
 // CHECK1-NEXT:    br i1 [[CMP31]], label [[OMP_INNER_FOR_BODY32:%.*]], label [[OMP_INNER_FOR_END40:%.*]]
-// CHECK1-IRBUILDER:       omp.inner.for.body33:
-// CHECK1-NORMAL:       omp.inner.for.body32:
+// CHECK1:       omp.inner.for.body32:
 // CHECK1-NEXT:    [[TMP30:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_18]], align 4, !llvm.access.group !7
 // CHECK1-NEXT:    [[TMP31:%.*]] = load i32, i32* [[DOTOMP_IV16]], align 4, !llvm.access.group !7
 // CHECK1-NEXT:    [[MUL33:%.*]] = mul i32 [[TMP31]], 1
 // CHECK1-NEXT:    [[ADD34:%.*]] = add i32 [[TMP30]], [[MUL33]]
 // CHECK1-NEXT:    store i32 [[ADD34]], i32* [[I28]], align 4, !llvm.access.group !7
 // CHECK1-NEXT:    [[TMP32:%.*]] = load i32, i32* [[I28]], align 4, !llvm.access.group !7
 // CHECK1-NEXT:    [[IDXPROM35:%.*]] = sext i32 [[TMP32]] to i64
 // CHECK1-NEXT:    [[ARRAYIDX36:%.*]] = getelementptr inbounds [10 x float], [10 x float]* @f, i64 0, i64 [[IDXPROM35]]
 // CHECK1-NEXT:    store float 0.000000e+00, float* [[ARRAYIDX36]], align 4, !llvm.access.group !7
 // CHECK1-NEXT:    call void @__captured_stmt.1(i32* [[I28]]), !llvm.access.group !7
 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE37:%.*]]
-// CHECK1-IRBUILDER:       omp.body.continue38:
-// CHECK1-NORMAL:       omp.body.continue37:
+// CHECK1:       omp.body.continue37:
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC38:%.*]]
-// CHECK1-IRBUILDER:       omp.inner.for.inc39:
-// CHECK1-NORMAL:       omp.inner.for.inc38:
+// CHECK1:       omp.inner.for.inc38:
 // CHECK1-NEXT:    [[TMP33:%.*]] = load i32, i32* [[DOTOMP_IV16]], align 4, !llvm.access.group !7
 // CHECK1-NEXT:    [[ADD39:%.*]] = add i32 [[TMP33]], 1
 // CHECK1-NEXT:    store i32 [[ADD39]], i32* [[DOTOMP_IV16]], align 4, !llvm.access.group !7
-// CHECK1-IRBUILDER-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB12]])
 // CHECK1-NEXT:    call void @__kmpc_dispatch_fini_4u(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]]), !llvm.access.group !7
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND29]], !llvm.loop [[LOOP8:![0-9]+]]
-// CHECK1-IRBUILDER:       omp.inner.for.end42:
-// CHECK1-NORMAL:       omp.inner.for.end40:
+// CHECK1:       omp.inner.for.end40:
 // CHECK1-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
 // CHECK1:       omp.dispatch.inc:
 // CHECK1-NEXT:    br label [[OMP_DISPATCH_COND]]
 // CHECK1:       omp.dispatch.end:
 // CHECK1-NEXT:    [[TMP34:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK1-NEXT:    [[TMP35:%.*]] = icmp ne i32 [[TMP34]], 0
 // CHECK1-NEXT:    br i1 [[TMP35]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
 // CHECK1:       .omp.final.then:
 // CHECK1-NEXT:    [[TMP36:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_18]], align 4
 // CHECK1-NEXT:    [[TMP37:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_19]], align 4
 // CHECK1-NEXT:    [[TMP38:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_18]], align 4
 // CHECK1-NEXT:    [[SUB41:%.*]] = sub i32 [[TMP37]], [[TMP38]]
 // CHECK1-NEXT:    [[SUB42:%.*]] = sub i32 [[SUB41]], 1
 // CHECK1-NEXT:    [[ADD43:%.*]] = add i32 [[SUB42]], 1
 // CHECK1-NEXT:    [[DIV44:%.*]] = udiv i32 [[ADD43]], 1
 // CHECK1-NEXT:    [[MUL45:%.*]] = mul i32 [[DIV44]], 1
 // CHECK1-NEXT:    [[ADD46:%.*]] = add i32 [[TMP36]], [[MUL45]]
 // CHECK1-NEXT:    store i32 [[ADD46]], i32* [[I28]], align 4
 // CHECK1-NEXT:    br label [[DOTOMP_FINAL_DONE]]
 // CHECK1:       .omp.final.done:
 // CHECK1-NEXT:    br label [[OMP_PRECOND_END]]
 // CHECK1:       omp.precond.end:
-// CHECK1-IRBUILDER-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
-// CHECK1-IRBUILDER-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB4]], i32 [[TMP0]])
-// CHECK1-NORMAL-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]])
+// CHECK1-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]])
 // CHECK1-NEXT:    ret void
 //
 //
 // CHECK1-LABEL: define {{[^@]+}}@__captured_stmt
 // CHECK1-SAME: (i32* noundef nonnull align 4 dereferenceable(4) [[I:%.*]]) #[[ATTR3:[0-9]+]] {
 // CHECK1-NEXT:  entry:
 // CHECK1-NEXT:    [[I_ADDR:%.*]] = alloca i32*, align 8
 // CHECK1-NEXT:    store i32* [[I]], i32** [[I_ADDR]], align 8

 // CHECK2-NEXT:    [[D_ADDR:%.*]] = alloca float*, align 8
 // CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    [[I:%.*]] = alloca i32, align 4
-// CHECK2-NORMAL-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1:[0-9]+]])
+// CHECK2-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1:[0-9]+]])
 // CHECK2-NEXT:    store float* [[A]], float** [[A_ADDR]], align 8
 // CHECK2-NEXT:    store float* [[B]], float** [[B_ADDR]], align 8
 // CHECK2-NEXT:    store float* [[C]], float** [[C_ADDR]], align 8
 // CHECK2-NEXT:    store float* [[D]], float** [[D_ADDR]], align 8
 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
 // CHECK2-NEXT:    store i32 4571423, i32* [[DOTOMP_UB]], align 4
 // CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
-// CHECK2-IRBUILDER-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB3:[0-9]+]])
-// CHECK2-IRBUILDER-NEXT:    call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP0]], i32 66, i32 0, i32 4571423, i32 1, i32 1)
-// CHECK2-NORMAL-NEXT:    call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 66, i32 0, i32 4571423, i32 1, i32 1)
+// CHECK2-NEXT:    call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 66, i32 0, i32 4571423, i32 1, i32 1)
 // CHECK2-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK2:       omp.dispatch.cond:
-// CHECK2-IRBUILDER-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB3]])
 // CHECK2-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
 // CHECK2-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP1]], 0
 // CHECK2-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
 // CHECK2:       omp.dispatch.body:
 // CHECK2-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK2-NEXT:    store i32 [[TMP2]], i32* [[DOTOMP_IV]], align 4
 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK2:       omp.inner.for.cond:
 // CHECK2-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK2-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP3]], [[TMP4]]
 // CHECK2-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK2:       omp.inner.for.body:
 // CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK2-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP5]], 7
 // CHECK2-NEXT:    [[SUB:%.*]] = sub nsw i32 32000000, [[MUL]]
 // CHECK2-NEXT:    store i32 [[SUB]], i32* [[I]], align 4
-// CHECK2-IRBUILDER-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
 // CHECK2-NEXT:    call void @__kmpc_ordered(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
 // CHECK2-NEXT:    [[TMP6:%.*]] = load float*, float** [[B_ADDR]], align 8
 // CHECK2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[I]], align 4
 // CHECK2-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP7]] to i64
 // CHECK2-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, float* [[TMP6]], i64 [[IDXPROM]]
 // CHECK2-NEXT:    [[TMP8:%.*]] = load float, float* [[ARRAYIDX]], align 4
 // CHECK2-NEXT:    [[TMP9:%.*]] = load float*, float** [[C_ADDR]], align 8
 // CHECK2-NEXT:    [[TMP10:%.*]] = load i32, i32* [[I]], align 4

 // CHECK2-NEXT:    call void @__kmpc_end_ordered(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
 // CHECK2-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK2:       omp.body.continue:
 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK2:       omp.inner.for.inc:
 // CHECK2-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP17]], 1
 // CHECK2-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
-// CHECK2-IRBUILDER-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB3]])
 // CHECK2-NEXT:    call void @__kmpc_dispatch_fini_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]]
 // CHECK2:       omp.inner.for.end:
 // CHECK2-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
 // CHECK2:       omp.dispatch.inc:
 // CHECK2-NEXT:    br label [[OMP_DISPATCH_COND]]
 // CHECK2:       omp.dispatch.end:
-// CHECK2-IRBUILDER-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
-// CHECK2-IRBUILDER-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB4:[0-9]+]], i32 [[TMP0]])
-// CHECK2-NORMAL-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP0]])
+// CHECK2-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP0]])
 // CHECK2-NEXT:    ret void
 //
 //
 // CHECK2-LABEL: define {{[^@]+}}@_Z8dynamic1PfS_S_S_
 // CHECK2-SAME: (float* noundef [[A:%.*]], float* noundef [[B:%.*]], float* noundef [[C:%.*]], float* noundef [[D:%.*]]) #[[ATTR0]] {
 // CHECK2-NEXT:  entry:
 // CHECK2-NEXT:    [[A_ADDR:%.*]] = alloca float*, align 8
 // CHECK2-NEXT:    [[B_ADDR:%.*]] = alloca float*, align 8
 // CHECK2-NEXT:    [[C_ADDR:%.*]] = alloca float*, align 8
 // CHECK2-NEXT:    [[D_ADDR:%.*]] = alloca float*, align 8
 // CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i64, align 8
 // CHECK2-NEXT:    [[TMP:%.*]] = alloca i64, align 8
 // CHECK2-NEXT:    [[DOTOMP_LB:%.*]] = alloca i64, align 8
 // CHECK2-NEXT:    [[DOTOMP_UB:%.*]] = alloca i64, align 8
 // CHECK2-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i64, align 8
 // CHECK2-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    [[I:%.*]] = alloca i64, align 8
-// CHECK2-NORMAL-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
+// CHECK2-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
 // CHECK2-NEXT:    store float* [[A]], float** [[A_ADDR]], align 8
 // CHECK2-NEXT:    store float* [[B]], float** [[B_ADDR]], align 8
 // CHECK2-NEXT:    store float* [[C]], float** [[C_ADDR]], align 8
 // CHECK2-NEXT:    store float* [[D]], float** [[D_ADDR]], align 8
 // CHECK2-NEXT:    store i64 0, i64* [[DOTOMP_LB]], align 8
 // CHECK2-NEXT:    store i64 16908287, i64* [[DOTOMP_UB]], align 8
 // CHECK2-NEXT:    store i64 1, i64* [[DOTOMP_STRIDE]], align 8
 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
-// CHECK2-IRBUILDER-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB6:[0-9]+]])
 // CHECK2-NEXT:    call void @__kmpc_dispatch_init_8u(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 1073741891, i64 0, i64 16908287, i64 1, i64 1)
 // CHECK2-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK2:       omp.dispatch.cond:
-// CHECK2-IRBUILDER-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB6]])
 // CHECK2-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_dispatch_next_8u(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32* [[DOTOMP_IS_LAST]], i64* [[DOTOMP_LB]], i64* [[DOTOMP_UB]], i64* [[DOTOMP_STRIDE]])
 // CHECK2-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP1]], 0
 // CHECK2-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
 // CHECK2:       omp.dispatch.body:
 // CHECK2-NEXT:    [[TMP2:%.*]] = load i64, i64* [[DOTOMP_LB]], align 8
 // CHECK2-NEXT:    store i64 [[TMP2]], i64* [[DOTOMP_IV]], align 8
 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK2:       omp.inner.for.cond:
 // CHECK2-NEXT:    [[TMP3:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
 // CHECK2-NEXT:    [[TMP4:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
 // CHECK2-NEXT:    [[ADD:%.*]] = add i64 [[TMP4]], 1
 // CHECK2-NEXT:    [[CMP:%.*]] = icmp ult i64 [[TMP3]], [[ADD]]
 // CHECK2-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK2:       omp.inner.for.body:
 // CHECK2-NEXT:    [[TMP5:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
 // CHECK2-NEXT:    [[MUL:%.*]] = mul i64 [[TMP5]], 127
 // CHECK2-NEXT:    [[ADD1:%.*]] = add i64 131071, [[MUL]]
 // CHECK2-NEXT:    store i64 [[ADD1]], i64* [[I]], align 8
-// CHECK2-IRBUILDER-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
 // CHECK2-NEXT:    call void @__kmpc_ordered(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
 // CHECK2-NEXT:    [[TMP6:%.*]] = load float*, float** [[B_ADDR]], align 8
 // CHECK2-NEXT:    [[TMP7:%.*]] = load i64, i64* [[I]], align 8
 // CHECK2-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, float* [[TMP6]], i64 [[TMP7]]
 // CHECK2-NEXT:    [[TMP8:%.*]] = load float, float* [[ARRAYIDX]], align 4
 // CHECK2-NEXT:    [[TMP9:%.*]] = load float*, float** [[C_ADDR]], align 8
 // CHECK2-NEXT:    [[TMP10:%.*]] = load i64, i64* [[I]], align 8
 // CHECK2-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds float, float* [[TMP9]], i64 [[TMP10]]

 // CHECK2-NEXT:    call void @__kmpc_end_ordered(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
 // CHECK2-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK2:       omp.body.continue:
 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK2:       omp.inner.for.inc:
 // CHECK2-NEXT:    [[TMP17:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
 // CHECK2-NEXT:    [[ADD7:%.*]] = add i64 [[TMP17]], 1
 // CHECK2-NEXT:    store i64 [[ADD7]], i64* [[DOTOMP_IV]], align 8
-// CHECK2-IRBUILDER-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB6]])
 // CHECK2-NEXT:    call void @__kmpc_dispatch_fini_8u(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]]
 // CHECK2:       omp.inner.for.end:
 // CHECK2-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
 // CHECK2:       omp.dispatch.inc:
 // CHECK2-NEXT:    br label [[OMP_DISPATCH_COND]]
 // CHECK2:       omp.dispatch.end:
-// CHECK2-IRBUILDER-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
-// CHECK2-IRBUILDER-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB4]], i32 [[TMP0]])
-// CHECK2-NORMAL-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]])
+// CHECK2-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]])
 // CHECK2-NEXT:    ret void
 //
 //
 // CHECK2-LABEL: define {{[^@]+}}@_Z9test_autoPfS_S_S_
 // CHECK2-SAME: (float* noundef [[A:%.*]], float* noundef [[B:%.*]], float* noundef [[C:%.*]], float* noundef [[D:%.*]]) #[[ATTR0]] {
 // CHECK2-NEXT:  entry:
 // CHECK2-NEXT:    [[A_ADDR:%.*]] = alloca float*, align 8
 // CHECK2-NEXT:    [[B_ADDR:%.*]] = alloca float*, align 8

 // CHECK2-NEXT:    [[I:%.*]] = alloca i8, align 1
 // CHECK2-NEXT:    [[X6:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    [[DOTOMP_LB:%.*]] = alloca i64, align 8
 // CHECK2-NEXT:    [[DOTOMP_UB:%.*]] = alloca i64, align 8
 // CHECK2-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i64, align 8
 // CHECK2-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    [[I8:%.*]] = alloca i8, align 1
 // CHECK2-NEXT:    [[X9:%.*]] = alloca i32, align 4
-// CHECK2-NORMAL-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
+// CHECK2-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
 // CHECK2-NEXT:    store float* [[A]], float** [[A_ADDR]], align 8
 // CHECK2-NEXT:    store float* [[B]], float** [[B_ADDR]], align 8
 // CHECK2-NEXT:    store float* [[C]], float** [[C_ADDR]], align 8
 // CHECK2-NEXT:    store float* [[D]], float** [[D_ADDR]], align 8
 // CHECK2-NEXT:    store i32 0, i32* [[X]], align 4
 // CHECK2-NEXT:    store i32 0, i32* [[Y]], align 4
 // CHECK2-NEXT:    [[TMP1:%.*]] = load i32, i32* [[Y]], align 4
 // CHECK2-NEXT:    [[CONV:%.*]] = trunc i32 [[TMP1]] to i8

 // CHECK2-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
 // CHECK2:       omp.precond.then:
 // CHECK2-NEXT:    store i64 0, i64* [[DOTOMP_LB]], align 8
 // CHECK2-NEXT:    [[TMP5:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR_2]], align 8
 // CHECK2-NEXT:    store i64 [[TMP5]], i64* [[DOTOMP_UB]], align 8
 // CHECK2-NEXT:    store i64 1, i64* [[DOTOMP_STRIDE]], align 8
 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK2-NEXT:    [[TMP6:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR_2]], align 8
-// CHECK2-IRBUILDER-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB8:[0-9]+]])
 // CHECK2-NEXT:    call void @__kmpc_dispatch_init_8(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 1073741894, i64 0, i64 [[TMP6]], i64 1, i64 1)
 // CHECK2-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK2:       omp.dispatch.cond:
-// CHECK2-IRBUILDER-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB8]])
 // CHECK2-NEXT:    [[TMP7:%.*]] = call i32 @__kmpc_dispatch_next_8(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32* [[DOTOMP_IS_LAST]], i64* [[DOTOMP_LB]], i64* [[DOTOMP_UB]], i64* [[DOTOMP_STRIDE]])
 // CHECK2-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP7]], 0
 // CHECK2-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
 // CHECK2:       omp.dispatch.body:
 // CHECK2-NEXT:    [[TMP8:%.*]] = load i64, i64* [[DOTOMP_LB]], align 8
 // CHECK2-NEXT:    store i64 [[TMP8]], i64* [[DOTOMP_IV]], align 8
 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK2:       omp.inner.for.cond:

 // CHECK2-NEXT:    [[TMP14:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
 // CHECK2-NEXT:    [[DIV16:%.*]] = sdiv i64 [[TMP14]], 11
 // CHECK2-NEXT:    [[MUL17:%.*]] = mul nsw i64 [[DIV16]], 11
 // CHECK2-NEXT:    [[SUB18:%.*]] = sub nsw i64 [[TMP13]], [[MUL17]]
 // CHECK2-NEXT:    [[MUL19:%.*]] = mul nsw i64 [[SUB18]], 1
 // CHECK2-NEXT:    [[SUB20:%.*]] = sub nsw i64 11, [[MUL19]]
 // CHECK2-NEXT:    [[CONV21:%.*]] = trunc i64 [[SUB20]] to i32
 // CHECK2-NEXT:    store i32 [[CONV21]], i32* [[X9]], align 4
-// CHECK2-IRBUILDER-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
 // CHECK2-NEXT:    call void @__kmpc_ordered(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
 // CHECK2-NEXT:    [[TMP15:%.*]] = load float*, float** [[B_ADDR]], align 8
 // CHECK2-NEXT:    [[TMP16:%.*]] = load i8, i8* [[I8]], align 1
 // CHECK2-NEXT:    [[IDXPROM:%.*]] = sext i8 [[TMP16]] to i64
 // CHECK2-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, float* [[TMP15]], i64 [[IDXPROM]]
 // CHECK2-NEXT:    [[TMP17:%.*]] = load float, float* [[ARRAYIDX]], align 4
 // CHECK2-NEXT:    [[TMP18:%.*]] = load float*, float** [[C_ADDR]], align 8
 // CHECK2-NEXT:    [[TMP19:%.*]] = load i8, i8* [[I8]], align 1

 // CHECK2-NEXT:    call void @__kmpc_end_ordered(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
 // CHECK2-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK2:       omp.body.continue:
 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK2:       omp.inner.for.inc:
 // CHECK2-NEXT:    [[TMP26:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
 // CHECK2-NEXT:    [[ADD30:%.*]] = add nsw i64 [[TMP26]], 1
 // CHECK2-NEXT:    store i64 [[ADD30]], i64* [[DOTOMP_IV]], align 8
-// CHECK2-IRBUILDER-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB8]])
 // CHECK2-NEXT:    call void @__kmpc_dispatch_fini_8(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]]
 // CHECK2:       omp.inner.for.end:
 // CHECK2-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
 // CHECK2:       omp.dispatch.inc:
 // CHECK2-NEXT:    br label [[OMP_DISPATCH_COND]]
 // CHECK2:       omp.dispatch.end:
 // CHECK2-NEXT:    br label [[OMP_PRECOND_END]]
 // CHECK2:       omp.precond.end:
-// CHECK2-IRBUILDER-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
-// CHECK2-IRBUILDER-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB4]], i32 [[TMP0]])
-// CHECK2-NORMAL-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]])
+// CHECK2-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]])
 // CHECK2-NEXT:    ret void
 //
 //
 // CHECK2-LABEL: define {{[^@]+}}@_Z7runtimePfS_S_S_
 // CHECK2-SAME: (float* noundef [[A:%.*]], float* noundef [[B:%.*]], float* noundef [[C:%.*]], float* noundef [[D:%.*]]) #[[ATTR0]] {
 // CHECK2-NEXT:  entry:
 // CHECK2-NEXT:    [[A_ADDR:%.*]] = alloca float*, align 8
 // CHECK2-NEXT:    [[B_ADDR:%.*]] = alloca float*, align 8
 // CHECK2-NEXT:    [[C_ADDR:%.*]] = alloca float*, align 8
 // CHECK2-NEXT:    [[D_ADDR:%.*]] = alloca float*, align 8
 // CHECK2-NEXT:    [[X:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    [[TMP:%.*]] = alloca i8, align 1
 // CHECK2-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    [[I:%.*]] = alloca i8, align 1
 // CHECK2-NEXT:    [[X2:%.*]] = alloca i32, align 4
-// CHECK2-NORMAL-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
+// CHECK2-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
 // CHECK2-NEXT:    store float* [[A]], float** [[A_ADDR]], align 8
 // CHECK2-NEXT:    store float* [[B]], float** [[B_ADDR]], align 8
 // CHECK2-NEXT:    store float* [[C]], float** [[C_ADDR]], align 8
 // CHECK2-NEXT:    store float* [[D]], float** [[D_ADDR]], align 8
 // CHECK2-NEXT:    store i32 0, i32* [[X]], align 4
 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
 // CHECK2-NEXT:    store i32 199, i32* [[DOTOMP_UB]], align 4
 // CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
-// CHECK2-IRBUILDER-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB10:[0-9]+]])
 // CHECK2-NEXT:    call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 1073741893, i32 0, i32 199, i32 1, i32 1)
 // CHECK2-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK2:       omp.dispatch.cond:
-// CHECK2-IRBUILDER-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB10]])
 // CHECK2-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
 // CHECK2-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP1]], 0
 // CHECK2-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
 // CHECK2:       omp.dispatch.body:
 // CHECK2-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK2-NEXT:    store i32 [[TMP2]], i32* [[DOTOMP_IV]], align 4
 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK2:       omp.inner.for.cond:

 // CHECK2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK2-NEXT:    [[DIV3:%.*]] = sdiv i32 [[TMP7]], 20
 // CHECK2-NEXT:    [[MUL4:%.*]] = mul nsw i32 [[DIV3]], 20
 // CHECK2-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP6]], [[MUL4]]
 // CHECK2-NEXT:    [[MUL5:%.*]] = mul nsw i32 [[SUB]], 1
 // CHECK2-NEXT:    [[ADD6:%.*]] = add nsw i32 -10, [[MUL5]]
 // CHECK2-NEXT:    store i32 [[ADD6]], i32* [[X2]], align 4
-// CHECK2-IRBUILDER-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
 // CHECK2-NEXT:    call void @__kmpc_ordered(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
 // CHECK2-NEXT:    [[TMP8:%.*]] = load float*, float** [[B_ADDR]], align 8
 // CHECK2-NEXT:    [[TMP9:%.*]] = load i8, i8* [[I]], align 1
 // CHECK2-NEXT:    [[IDXPROM:%.*]] = zext i8 [[TMP9]] to i64
 // CHECK2-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, float* [[TMP8]], i64 [[IDXPROM]]
 // CHECK2-NEXT:    [[TMP10:%.*]] = load float, float* [[ARRAYIDX]], align 4
 // CHECK2-NEXT:    [[TMP11:%.*]] = load float*, float** [[C_ADDR]], align 8
 // CHECK2-NEXT:    [[TMP12:%.*]] = load i8, i8* [[I]], align 1

 // CHECK2-NEXT:    call void @__kmpc_end_ordered(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
 // CHECK2-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK2:       omp.body.continue:
 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK2:       omp.inner.for.inc:
 // CHECK2-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK2-NEXT:    [[ADD15:%.*]] = add nsw i32 [[TMP19]], 1
 // CHECK2-NEXT:    store i32 [[ADD15]], i32* [[DOTOMP_IV]], align 4
-// CHECK2-IRBUILDER-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB10]])
 // CHECK2-NEXT:    call void @__kmpc_dispatch_fini_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]]
 // CHECK2:       omp.inner.for.end:
 // CHECK2-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
 // CHECK2:       omp.dispatch.inc:
 // CHECK2-NEXT:    br label [[OMP_DISPATCH_COND]]
 // CHECK2:       omp.dispatch.end:
-// CHECK2-IRBUILDER-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
-// CHECK2-IRBUILDER-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB4]], i32 [[TMP0]])
-// CHECK2-NORMAL-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]])
+// CHECK2-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]])
 // CHECK2-NEXT:    ret void
 //
 //
 // CHECK2-LABEL: define {{[^@]+}}@_Z8foo_simdii
 // CHECK2-SAME: (i32 noundef [[LOW:%.*]], i32 noundef [[UP:%.*]]) #[[ATTR0]] {
 // CHECK2-NEXT:  entry:
 // CHECK2-NEXT:    [[LOW_ADDR:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    [[UP_ADDR:%.*]] = alloca i32, align 4

 // CHECK2-NEXT:    [[DOTCAPTURE_EXPR_19:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    [[DOTCAPTURE_EXPR_20:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    [[I26:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    [[I28:%.*]] = alloca i32, align 4
-// CHECK2-NORMAL-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
+// CHECK2-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
 // CHECK2-NEXT:    store i32 [[LOW]], i32* [[LOW_ADDR]], align 4
 // CHECK2-NEXT:    store i32 [[UP]], i32* [[UP_ADDR]], align 4
 // CHECK2-NEXT:    [[TMP1:%.*]] = load i32, i32* [[LOW_ADDR]], align 4
 // CHECK2-NEXT:    store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
 // CHECK2-NEXT:    [[TMP2:%.*]] = load i32, i32* [[UP_ADDR]], align 4
 // CHECK2-NEXT:    store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
 // CHECK2-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
 // CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4

 // CHECK2-NEXT:    br i1 [[CMP27]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
 // CHECK2:       omp.precond.then:
 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
 // CHECK2-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_20]], align 4
 // CHECK2-NEXT:    store i32 [[TMP24]], i32* [[DOTOMP_UB]], align 4
 // CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK2-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_20]], align 4
-// CHECK2-IRBUILDER-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB12:[0-9]+]])
 // CHECK2-NEXT:    call void @__kmpc_dispatch_init_4u(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 66, i32 0, i32 [[TMP25]], i32 1, i32 1)
 // CHECK2-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK2:       omp.dispatch.cond:
-// CHECK2-IRBUILDER-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB12]])
 // CHECK2-NEXT:    [[TMP26:%.*]] = call i32 @__kmpc_dispatch_next_4u(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
 // CHECK2-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP26]], 0
 // CHECK2-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
 // CHECK2:       omp.dispatch.body:
 // CHECK2-NEXT:    [[TMP27:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK2-NEXT:    store i32 [[TMP27]], i32* [[DOTOMP_IV16]], align 4
 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND29:%.*]]
-// CHECK2-IRBUILDER:       omp.inner.for.cond30:
-// CHECK2-NORMAL:       omp.inner.for.cond29:
+// CHECK2:       omp.inner.for.cond29:
 // CHECK2-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTOMP_IV16]], align 4, !llvm.access.group !7
 // CHECK2-NEXT:    [[TMP29:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !7
 // CHECK2-NEXT:    [[ADD30:%.*]] = add i32 [[TMP29]], 1
 // CHECK2-NEXT:    [[CMP31:%.*]] = icmp ult i32 [[TMP28]], [[ADD30]]
 // CHECK2-NEXT:    br i1 [[CMP31]], label [[OMP_INNER_FOR_BODY32:%.*]], label [[OMP_INNER_FOR_END40:%.*]]
-// CHECK2-IRBUILDER:       omp.inner.for.body33:
-// CHECK2-NORMAL:       omp.inner.for.body32:
+// CHECK2:       omp.inner.for.body32:
 // CHECK2-NEXT:    [[TMP30:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_18]], align 4, !llvm.access.group !7
 // CHECK2-NEXT:    [[TMP31:%.*]] = load i32, i32* [[DOTOMP_IV16]], align 4, !llvm.access.group !7
 // CHECK2-NEXT:    [[MUL33:%.*]] = mul i32 [[TMP31]], 1
 // CHECK2-NEXT:    [[ADD34:%.*]] = add i32 [[TMP30]], [[MUL33]]
 // CHECK2-NEXT:    store i32 [[ADD34]], i32* [[I28]], align 4, !llvm.access.group !7
 // CHECK2-NEXT:    [[TMP32:%.*]] = load i32, i32* [[I28]], align 4, !llvm.access.group !7
 // CHECK2-NEXT:    [[IDXPROM35:%.*]] = sext i32 [[TMP32]] to i64
 // CHECK2-NEXT:    [[ARRAYIDX36:%.*]] = getelementptr inbounds [10 x float], [10 x float]* @f, i64 0, i64 [[IDXPROM35]]
 // CHECK2-NEXT:    store float 0.000000e+00, float* [[ARRAYIDX36]], align 4, !llvm.access.group !7
 // CHECK2-NEXT:    call void @__captured_stmt.1(i32* [[I28]]), !llvm.access.group !7
 // CHECK2-NEXT:    br label [[OMP_BODY_CONTINUE37:%.*]]
-// CHECK2-IRBUILDER:       omp.body.continue38:
-// CHECK2-NORMAL:       omp.body.continue37:
+// CHECK2:       omp.body.continue37:
 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC38:%.*]]
-// CHECK2-IRBUILDER:       omp.inner.for.inc39:
-// CHECK2-NORMAL:       omp.inner.for.inc38:
+// CHECK2:       omp.inner.for.inc38:
 // CHECK2-NEXT:    [[TMP33:%.*]] = load i32, i32* [[DOTOMP_IV16]], align 4, !llvm.access.group !7
 // CHECK2-NEXT:    [[ADD39:%.*]] = add i32 [[TMP33]], 1
 // CHECK2-NEXT:    store i32 [[ADD39]], i32* [[DOTOMP_IV16]], align 4, !llvm.access.group !7
-// CHECK2-IRBUILDER-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB12]])
 // CHECK2-NEXT:    call void @__kmpc_dispatch_fini_4u(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]]), !llvm.access.group !7
 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND29]], !llvm.loop [[LOOP8:![0-9]+]]
-// CHECK2-IRBUILDER:       omp.inner.for.end42:
-// CHECK2-NORMAL:       omp.inner.for.end40:
+// CHECK2:       omp.inner.for.end40:
 // CHECK2-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
 // CHECK2:       omp.dispatch.inc:
 // CHECK2-NEXT:    br label [[OMP_DISPATCH_COND]]
 // CHECK2:       omp.dispatch.end:
 // CHECK2-NEXT:    [[TMP34:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK2-NEXT:    [[TMP35:%.*]] = icmp ne i32 [[TMP34]], 0
 // CHECK2-NEXT:    br i1 [[TMP35]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
 // CHECK2:       .omp.final.then:
 // CHECK2-NEXT:    [[TMP36:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_18]], align 4
 // CHECK2-NEXT:    [[TMP37:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_19]], align 4
 // CHECK2-NEXT:    [[TMP38:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_18]], align 4
 // CHECK2-NEXT:    [[SUB41:%.*]] = sub i32 [[TMP37]], [[TMP38]]
 // CHECK2-NEXT:    [[SUB42:%.*]] = sub i32 [[SUB41]], 1
 // CHECK2-NEXT:    [[ADD43:%.*]] = add i32 [[SUB42]], 1
 // CHECK2-NEXT:    [[DIV44:%.*]] = udiv i32 [[ADD43]], 1
 // CHECK2-NEXT:    [[MUL45:%.*]] = mul i32 [[DIV44]], 1
 // CHECK2-NEXT:    [[ADD46:%.*]] = add i32 [[TMP36]], [[MUL45]]
 // CHECK2-NEXT:    store i32 [[ADD46]], i32* [[I28]], align 4
 // CHECK2-NEXT:    br label [[DOTOMP_FINAL_DONE]]
 // CHECK2:       .omp.final.done:
 // CHECK2-NEXT:    br label [[OMP_PRECOND_END]]
 // CHECK2:       omp.precond.end:
-// CHECK2-IRBUILDER-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
-// CHECK2-IRBUILDER-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB4]], i32 [[TMP0]])
-// CHECK2-NORMAL-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]])
+// CHECK2-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]])
 // CHECK2-NEXT:    ret void
 //
 //
 // CHECK2-LABEL: define {{[^@]+}}@__captured_stmt
 // CHECK2-SAME: (i32* noundef nonnull align 4 dereferenceable(4) [[I:%.*]]) #[[ATTR3:[0-9]+]] {
 // CHECK2-NEXT:  entry:
 // CHECK2-NEXT:    [[I_ADDR:%.*]] = alloca i32*, align 8
 // CHECK2-NEXT:    store i32* [[I]], i32** [[I_ADDR]], align 8

 // CHECK2-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[I_ADDR]], align 8
 // CHECK2-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
 // CHECK2-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP1]] to i64
 // CHECK2-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x float], [10 x float]* @f, i64 0, i64 [[IDXPROM]]
 // CHECK2-NEXT:    store float 1.000000e+00, float* [[ARRAYIDX]], align 4
 // CHECK2-NEXT:    ret void
 //
 //
+// CHECK1-IRBUILDER-LABEL: define {{[^@]+}}@_Z18static_not_chunkedPfS_S_S_
+// CHECK1-IRBUILDER-SAME: (float* noundef [[A:%.*]], float* noundef [[B:%.*]], float* noundef [[C:%.*]], float* noundef [[D:%.*]]) #[[ATTR0:[0-9]+]] {
+// CHECK1-IRBUILDER-NEXT:  entry:
+// CHECK1-IRBUILDER-NEXT:    [[A_ADDR:%.*]] = alloca float*, align 8
+// CHECK1-IRBUILDER-NEXT:    [[B_ADDR:%.*]] = alloca float*, align 8
+// CHECK1-IRBUILDER-NEXT:    [[C_ADDR:%.*]] = alloca float*, align 8
+// CHECK1-IRBUILDER-NEXT:    [[D_ADDR:%.*]] = alloca float*, align 8
+// CHECK1-IRBUILDER-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
+// CHECK1-IRBUILDER-NEXT:    [[TMP:%.*]] = alloca i32, align 4
+// CHECK1-IRBUILDER-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
+// CHECK1-IRBUILDER-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
+// CHECK1-IRBUILDER-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
+// CHECK1-IRBUILDER-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
+// CHECK1-IRBUILDER-NEXT:    [[I:%.*]] = alloca i32, align 4
+// CHECK1-IRBUILDER-NEXT:    store float* [[A]], float** [[A_ADDR]], align 8
+// CHECK1-IRBUILDER-NEXT:    store float* [[B]], float** [[B_ADDR]], align 8
+// CHECK1-IRBUILDER-NEXT:    store float* [[C]], float** [[C_ADDR]], align 8
+// CHECK1-IRBUILDER-NEXT:    store float* [[D]], float** [[D_ADDR]], align 8
+// CHECK1-IRBUILDER-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
+// CHECK1-IRBUILDER-NEXT:    store i32 4571423, i32* [[DOTOMP_UB]], align 4
+// CHECK1-IRBUILDER-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
+// CHECK1-IRBUILDER-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
+// CHECK1-IRBUILDER-NEXT:    [[OMP_GLOBAL_THREAD_NUM:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB3:[0-9]+]])
+// CHECK1-IRBUILDER-NEXT:    call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[OMP_GLOBAL_THREAD_NUM]], i32 66, i32 0, i32 4571423, i32 1, i32 1)
+// CHECK1-IRBUILDER-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
+// CHECK1-IRBUILDER:       omp.dispatch.cond:
+// CHECK1-IRBUILDER-NEXT:    [[OMP_GLOBAL_THREAD_NUM1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB3]])
+// CHECK1-IRBUILDER-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM1]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
+// CHECK1-IRBUILDER-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP0]], 0
+// CHECK1-IRBUILDER-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
+// CHECK1-IRBUILDER:       omp.dispatch.body:
+// CHECK1-IRBUILDER-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
+// CHECK1-IRBUILDER-NEXT:    store i32 [[TMP1]], i32* [[DOTOMP_IV]], align 4
+// CHECK1-IRBUILDER-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
+// CHECK1-IRBUILDER:       omp.inner.for.cond:
+// CHECK1-IRBUILDER-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
+// CHECK1-IRBUILDER-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
+// CHECK1-IRBUILDER-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP2]], [[TMP3]]
+// CHECK1-IRBUILDER-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// CHECK1-IRBUILDER:       omp.inner.for.body:
+// CHECK1-IRBUILDER-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
+// CHECK1-IRBUILDER-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP4]], 7
+// CHECK1-IRBUILDER-NEXT:    [[SUB:%.*]] = sub nsw i32 32000000, [[MUL]]
+// CHECK1-IRBUILDER-NEXT:    store i32 [[SUB]], i32* [[I]], align 4
+// CHECK1-IRBUILDER-NEXT:    [[OMP_GLOBAL_THREAD_NUM2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
+// CHECK1-IRBUILDER-NEXT:    call void @__kmpc_ordered(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM2]])
+// CHECK1-IRBUILDER-NEXT:    [[TMP5:%.*]] = load float*, float** [[B_ADDR]], align 8
+// CHECK1-IRBUILDER-NEXT:    [[TMP6:%.*]] = load i32, i32* [[I]], align 4
+// CHECK1-IRBUILDER-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP6]] to i64
+// CHECK1-IRBUILDER-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, float* [[TMP5]], i64 [[IDXPROM]]
+// CHECK1-IRBUILDER-NEXT:    [[TMP7:%.*]] = load float, float* [[ARRAYIDX]], align 4
+// CHECK1-IRBUILDER-NEXT:    [[TMP8:%.*]] = load float*, float** [[C_ADDR]], align 8
+// CHECK1-IRBUILDER-NEXT:    [[TMP9:%.*]] = load i32, i32* [[I]], align 4
+// CHECK1-IRBUILDER-NEXT:    [[IDXPROM3:%.*]] = sext i32 [[TMP9]] to i64
+// CHECK1-IRBUILDER-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds float, float* [[TMP8]], i64 [[IDXPROM3]]
+// CHECK1-IRBUILDER-NEXT:    [[TMP10:%.*]] = load float, float* [[ARRAYIDX4]], align 4
+// CHECK1-IRBUILDER-NEXT:    [[MUL5:%.*]] = fmul float [[TMP7]], [[TMP10]]
+// CHECK1-IRBUILDER-NEXT:    [[TMP11:%.*]] = load float*, float** [[D_ADDR]], align 8
+// CHECK1-IRBUILDER-NEXT:    [[TMP12:%.*]] = load i32, i32* [[I]], align 4
+// CHECK1-IRBUILDER-NEXT:    [[IDXPROM6:%.*]] = sext i32 [[TMP12]] to i64
+// CHECK1-IRBUILDER-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds float, float* [[TMP11]], i64 [[IDXPROM6]]
+// CHECK1-IRBUILDER-NEXT:    [[TMP13:%.*]] = load float, float* [[ARRAYIDX7]], align 4
+// CHECK1-IRBUILDER-NEXT:    [[MUL8:%.*]] = fmul float [[MUL5]], [[TMP13]]
+// CHECK1-IRBUILDER-NEXT:    [[TMP14:%.*]] = load float*, float** [[A_ADDR]], align 8
+// CHECK1-IRBUILDER-NEXT:    [[TMP15:%.*]] = load i32, i32* [[I]], align 4
+// CHECK1-IRBUILDER-NEXT:    [[IDXPROM9:%.*]] = sext i32 [[TMP15]] to i64
+// CHECK1-IRBUILDER-NEXT:    [[ARRAYIDX10:%.*]] = getelementptr inbounds float, float* [[TMP14]], i64 [[IDXPROM9]]
+// CHECK1-IRBUILDER-NEXT:    store float [[MUL8]], float* [[ARRAYIDX10]], align 4
+// CHECK1-IRBUILDER-NEXT:    br label [[OMP_INNER_FOR_BODY_ORDERED_AFTER:%.*]]
+// CHECK1-IRBUILDER:       omp.inner.for.body.ordered.after:
+// CHECK1-IRBUILDER-NEXT:    call void @__kmpc_end_ordered(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM2]])
+// CHECK1-IRBUILDER-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
+// CHECK1-IRBUILDER:       omp.body.continue:
+// CHECK1-IRBUILDER-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
+// CHECK1-IRBUILDER:       omp.inner.for.inc:
+// CHECK1-IRBUILDER-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
+// CHECK1-IRBUILDER-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP16]], 1
+// CHECK1-IRBUILDER-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
+// CHECK1-IRBUILDER-NEXT:    [[OMP_GLOBAL_THREAD_NUM11:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB3]])
+// CHECK1-IRBUILDER-NEXT:    call void @__kmpc_dispatch_fini_4(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM11]])
+// CHECK1-IRBUILDER-NEXT:    br label [[OMP_INNER_FOR_COND]]
+// CHECK1-IRBUILDER:       omp.inner.for.end:
+// CHECK1-IRBUILDER-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
+// CHECK1-IRBUILDER:       omp.dispatch.inc:
+// CHECK1-IRBUILDER-NEXT:    br label [[OMP_DISPATCH_COND]]
+// CHECK1-IRBUILDER:       omp.dispatch.end:
+// CHECK1-IRBUILDER-NEXT:    [[OMP_GLOBAL_THREAD_NUM12:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
+// CHECK1-IRBUILDER-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB4:[0-9]+]], i32 [[OMP_GLOBAL_THREAD_NUM12]])
+// CHECK1-IRBUILDER-NEXT:    ret void
+//
+//
+// CHECK1-IRBUILDER-LABEL: define {{[^@]+}}@_Z8dynamic1PfS_S_S_
+// CHECK1-IRBUILDER-SAME: (float* noundef [[A:%.*]], float* noundef [[B:%.*]], float* noundef [[C:%.*]], float* noundef [[D:%.*]]) #[[ATTR0]] {
+// CHECK1-IRBUILDER-NEXT:  entry:
+// CHECK1-IRBUILDER-NEXT:    [[A_ADDR:%.*]] = alloca float*, align 8
+// CHECK1-IRBUILDER-NEXT:    [[B_ADDR:%.*]] = alloca float*, align 8
+// CHECK1-IRBUILDER-NEXT:    [[C_ADDR:%.*]] = alloca float*, align 8
+// CHECK1-IRBUILDER-NEXT:    [[D_ADDR:%.*]] = alloca float*, align 8
+// CHECK1-IRBUILDER-NEXT:    [[DOTOMP_IV:%.*]] = alloca i64, align 8
+// CHECK1-IRBUILDER-NEXT:    [[TMP:%.*]] = alloca i64, align 8
+// CHECK1-IRBUILDER-NEXT:    [[DOTOMP_LB:%.*]] = alloca i64, align 8
+// CHECK1-IRBUILDER-NEXT:    [[DOTOMP_UB:%.*]] = alloca i64, align 8
+// CHECK1-IRBUILDER-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i64, align 8
+// CHECK1-IRBUILDER-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
+// CHECK1-IRBUILDER-NEXT:    [[I:%.*]] = alloca i64, align 8
+// CHECK1-IRBUILDER-NEXT:    store float* [[A]], float** [[A_ADDR]], align 8
+// CHECK1-IRBUILDER-NEXT:    store float* [[B]], float** [[B_ADDR]], align 8
+// CHECK1-IRBUILDER-NEXT:    store float* [[C]], float** [[C_ADDR]], align 8
+// CHECK1-IRBUILDER-NEXT:    store float* [[D]], float** [[D_ADDR]], align 8
+// CHECK1-IRBUILDER-NEXT:    store i64 0, i64* [[DOTOMP_LB]], align 8
+// CHECK1-IRBUILDER-NEXT:    store i64 16908287, i64* [[DOTOMP_UB]], align 8
+// CHECK1-IRBUILDER-NEXT:    store i64 1, i64* [[DOTOMP_STRIDE]], align 8
+// CHECK1-IRBUILDER-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
+// CHECK1-IRBUILDER-NEXT:    [[OMP_GLOBAL_THREAD_NUM:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB6:[0-9]+]])
+// CHECK1-IRBUILDER-NEXT:    call void @__kmpc_dispatch_init_8u(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM]], i32 1073741891, i64 0, i64 16908287, i64 1, i64 1)
+// CHECK1-IRBUILDER-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
+// CHECK1-IRBUILDER:       omp.dispatch.cond:
+// CHECK1-IRBUILDER-NEXT:    [[OMP_GLOBAL_THREAD_NUM1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB6]])
+// CHECK1-IRBUILDER-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_dispatch_next_8u(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM1]], i32* [[DOTOMP_IS_LAST]], i64* [[DOTOMP_LB]], i64* [[DOTOMP_UB]], i64* [[DOTOMP_STRIDE]])
+// CHECK1-IRBUILDER-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP0]], 0
+// CHECK1-IRBUILDER-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
+// CHECK1-IRBUILDER:       omp.dispatch.body:
+// CHECK1-IRBUILDER-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTOMP_LB]], align 8
+// CHECK1-IRBUILDER-NEXT:    store i64 [[TMP1]], i64* [[DOTOMP_IV]], align 8
+// CHECK1-IRBUILDER-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
+// CHECK1-IRBUILDER:       omp.inner.for.cond:
+// CHECK1-IRBUILDER-NEXT:    [[TMP2:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
+// CHECK1-IRBUILDER-NEXT:    [[TMP3:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
+// CHECK1-IRBUILDER-NEXT:    [[ADD:%.*]] = add i64 [[TMP3]], 1
+// CHECK1-IRBUILDER-NEXT:    [[CMP:%.*]] = icmp ult i64 [[TMP2]], [[ADD]]
+// CHECK1-IRBUILDER-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// CHECK1-IRBUILDER:       omp.inner.for.body:
+// CHECK1-IRBUILDER-NEXT:    [[TMP4:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
+// CHECK1-IRBUILDER-NEXT:    [[MUL:%.*]] = mul i64 [[TMP4]], 127
+// CHECK1-IRBUILDER-NEXT:    [[ADD2:%.*]] = add i64 131071, [[MUL]]
+// CHECK1-IRBUILDER-NEXT:    store i64 [[ADD2]], i64* [[I]], align 8
+// CHECK1-IRBUILDER-NEXT:    [[OMP_GLOBAL_THREAD_NUM3:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
+// CHECK1-IRBUILDER-NEXT:    call void @__kmpc_ordered(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM3]])
+// CHECK1-IRBUILDER-NEXT:    [[TMP5:%.*]] = load float*, float** [[B_ADDR]], align 8
+// CHECK1-IRBUILDER-NEXT:    [[TMP6:%.*]] = load i64, i64* [[I]], align 8
+// CHECK1-IRBUILDER-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, float* [[TMP5]], i64 [[TMP6]]
+// CHECK1-IRBUILDER-NEXT:    [[TMP7:%.*]] = load float, float* [[ARRAYIDX]], align 4
+// CHECK1-IRBUILDER-NEXT:    [[TMP8:%.*]] = load float*, float** [[C_ADDR]], align 8
+// CHECK1-IRBUILDER-NEXT:    [[TMP9:%.*]] = load i64, i64* [[I]], align 8
+// CHECK1-IRBUILDER-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds float, float* [[TMP8]], i64 [[TMP9]]
+// CHECK1-IRBUILDER-NEXT:    [[TMP10:%.*]] = load float, float* [[ARRAYIDX4]], align 4
+// CHECK1-IRBUILDER-NEXT:    [[MUL5:%.*]] = fmul float [[TMP7]], [[TMP10]]
+// CHECK1-IRBUILDER-NEXT:    [[TMP11:%.*]] = load float*, float** [[D_ADDR]], align 8
+// CHECK1-IRBUILDER-NEXT:    [[TMP12:%.*]] = load i64, i64* [[I]], align 8
+// CHECK1-IRBUILDER-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds float, float* [[TMP11]], i64 [[TMP12]]
+// CHECK1-IRBUILDER-NEXT:    [[TMP13:%.*]] = load float, float* [[ARRAYIDX6]], align 4
+// CHECK1-IRBUILDER-NEXT:    [[MUL7:%.*]] = fmul float [[MUL5]], [[TMP13]]
+// CHECK1-IRBUILDER-NEXT:    [[TMP14:%.*]] = load float*, float** [[A_ADDR]], align 8
+// CHECK1-IRBUILDER-NEXT:    [[TMP15:%.*]] = load i64, i64* [[I]], align 8
+// CHECK1-IRBUILDER-NEXT:    [[ARRAYIDX8:%.*]] = getelementptr inbounds float, float* [[TMP14]], i64 [[TMP15]]
+// CHECK1-IRBUILDER-NEXT:    store float [[MUL7]], float* [[ARRAYIDX8]], align 4
+// CHECK1-IRBUILDER-NEXT:    br label [[OMP_INNER_FOR_BODY_ORDERED_AFTER:%.*]]
+// CHECK1-IRBUILDER:       omp.inner.for.body.ordered.after:
+// CHECK1-IRBUILDER-NEXT:    call void @__kmpc_end_ordered(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM3]])
+// CHECK1-IRBUILDER-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
+// CHECK1-IRBUILDER:       omp.body.continue:
+// CHECK1-IRBUILDER-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
+// CHECK1-IRBUILDER:       omp.inner.for.inc:
+// CHECK1-IRBUILDER-NEXT:    [[TMP16:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
+// CHECK1-IRBUILDER-NEXT:    [[ADD9:%.*]] = add i64 [[TMP16]], 1
+// CHECK1-IRBUILDER-NEXT:    store i64 [[ADD9]], i64* [[DOTOMP_IV]], align 8
+// CHECK1-IRBUILDER-NEXT:    [[OMP_GLOBAL_THREAD_NUM10:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB6]])
+// CHECK1-IRBUILDER-NEXT:    call void @__kmpc_dispatch_fini_8u(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM10]])
+// CHECK1-IRBUILDER-NEXT:    br label [[OMP_INNER_FOR_COND]]
+// CHECK1-IRBUILDER:       omp.inner.for.end:
+// CHECK1-IRBUILDER-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
+// CHECK1-IRBUILDER:       omp.dispatch.inc:
+// CHECK1-IRBUILDER-NEXT:    br label [[OMP_DISPATCH_COND]]
+// CHECK1-IRBUILDER:       omp.dispatch.end:
+// CHECK1-IRBUILDER-NEXT:    [[OMP_GLOBAL_THREAD_NUM11:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
+// CHECK1-IRBUILDER-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB4]], i32 [[OMP_GLOBAL_THREAD_NUM11]])
+// CHECK1-IRBUILDER-NEXT:    ret void
+//
+//
+// CHECK1-IRBUILDER-LABEL: define {{[^@]+}}@_Z9test_autoPfS_S_S_
+// CHECK1-IRBUILDER-SAME: (float* noundef [[A:%.*]], float* noundef [[B:%.*]], float* noundef [[C:%.*]], float* noundef [[D:%.*]]) #[[ATTR0]] {
+// CHECK1-IRBUILDER-NEXT:  entry:
+// CHECK1-IRBUILDER-NEXT:    [[A_ADDR:%.*]] = alloca float*, align 8
+// CHECK1-IRBUILDER-NEXT:    [[B_ADDR:%.*]] = alloca float*, align 8
+// CHECK1-IRBUILDER-NEXT:    [[C_ADDR:%.*]] = alloca float*, align 8
+// CHECK1-IRBUILDER-NEXT:    [[D_ADDR:%.*]] = alloca float*, align 8
+// CHECK1-IRBUILDER-NEXT:    [[X:%.*]] = alloca i32, align 4
+// CHECK1-IRBUILDER-NEXT:    [[Y:%.*]] = alloca i32, align 4
+// CHECK1-IRBUILDER-NEXT:    [[DOTOMP_IV:%.*]] = alloca i64, align 8
+// CHECK1-IRBUILDER-NEXT:    [[TMP:%.*]] = alloca i8, align 1
+// CHECK1-IRBUILDER-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
+// CHECK1-IRBUILDER-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i8, align 1
+// CHECK1-IRBUILDER-NEXT:    [[DOTCAPTURE_EXPR_2:%.*]] = alloca i64, align 8
+// CHECK1-IRBUILDER-NEXT:    [[I:%.*]] = alloca i8, align 1
+// CHECK1-IRBUILDER-NEXT:    [[X6:%.*]] = alloca i32, align 4
+// CHECK1-IRBUILDER-NEXT:    [[DOTOMP_LB:%.*]] = alloca i64, align 8
+// CHECK1-IRBUILDER-NEXT:    [[DOTOMP_UB:%.*]] = alloca i64, align 8
+// CHECK1-IRBUILDER-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i64, align 8
+// CHECK1-IRBUILDER-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
+// CHECK1-IRBUILDER-NEXT:    [[I8:%.*]] = alloca i8, align 1
+// CHECK1-IRBUILDER-NEXT:    [[X9:%.*]] = alloca i32, align 4
+// CHECK1-IRBUILDER-NEXT:    store float* [[A]], float** [[A_ADDR]], align 8
+// CHECK1-IRBUILDER-NEXT:    store float* [[B]], float** [[B_ADDR]], align 8
+// CHECK1-IRBUILDER-NEXT:    store float* [[C]], float** [[C_ADDR]], align 8
+// CHECK1-IRBUILDER-NEXT:    store float* [[D]], float** [[D_ADDR]], align 8
+// CHECK1-IRBUILDER-NEXT:    store i32 0, i32* [[X]], align 4
+// CHECK1-IRBUILDER-NEXT:    store i32 0, i32* [[Y]], align 4
+// CHECK1-IRBUILDER-NEXT:    [[TMP0:%.*]] = load i32, i32* [[Y]], align 4
+// CHECK1-IRBUILDER-NEXT:    [[CONV:%.*]] = trunc i32 [[TMP0]] to i8
+// CHECK1-IRBUILDER-NEXT:    store i8 [[CONV]], i8* [[DOTCAPTURE_EXPR_]], align 1
+// CHECK1-IRBUILDER-NEXT:    [[TMP1:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1
+// CHECK1-IRBUILDER-NEXT:    [[CONV3:%.*]] = sext i8 [[TMP1]] to i32
+// CHECK1-IRBUILDER-NEXT:    [[SUB:%.*]] = sub i32 57, [[CONV3]]
+// CHECK1-IRBUILDER-NEXT:    [[ADD:%.*]] = add i32 [[SUB]], 1
+// CHECK1-IRBUILDER-NEXT:    [[DIV:%.*]] = udiv i32 [[ADD]], 1
+// CHECK1-IRBUILDER-NEXT:    [[CONV4:%.*]] = zext i32 [[DIV]] to i64
+// CHECK1-IRBUILDER-NEXT:    [[MUL:%.*]] = mul nsw i64 [[CONV4]], 11
+// CHECK1-IRBUILDER-NEXT:    [[SUB5:%.*]] = sub nsw i64 [[MUL]], 1
+// CHECK1-IRBUILDER-NEXT:    store i64 [[SUB5]], i64* [[DOTCAPTURE_EXPR_2]], align 8
+// CHECK1-IRBUILDER-NEXT:    [[TMP2:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1
+// CHECK1-IRBUILDER-NEXT:    store i8 [[TMP2]], i8* [[I]], align 1
+// CHECK1-IRBUILDER-NEXT:    store i32 11, i32* [[X6]], align 4
+// CHECK1-IRBUILDER-NEXT:    [[TMP3:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1
+// CHECK1-IRBUILDER-NEXT:    [[CONV7:%.*]] = sext i8 [[TMP3]] to i32
+// CHECK1-IRBUILDER-NEXT:    [[CMP:%.*]] = icmp sle i32 [[CONV7]], 57
+// CHECK1-IRBUILDER-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
+// CHECK1-IRBUILDER:       omp.precond.then:
+// CHECK1-IRBUILDER-NEXT:    store i64 0, i64* [[DOTOMP_LB]], align 8
+// CHECK1-IRBUILDER-NEXT:    [[TMP4:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR_2]], align 8
+// CHECK1-IRBUILDER-NEXT:    store i64 [[TMP4]], i64* [[DOTOMP_UB]], align 8
+// CHECK1-IRBUILDER-NEXT:    store i64 1, i64* [[DOTOMP_STRIDE]], align 8
+// CHECK1-IRBUILDER-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
+// CHECK1-IRBUILDER-NEXT:    [[TMP5:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR_2]], align 8
+// CHECK1-IRBUILDER-NEXT:    [[OMP_GLOBAL_THREAD_NUM:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB8:[0-9]+]])
+// CHECK1-IRBUILDER-NEXT:    call void @__kmpc_dispatch_init_8(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM]], i32 1073741894, i64 0, i64 [[TMP5]], i64 1, i64 1)
+// CHECK1-IRBUILDER-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
+// CHECK1-IRBUILDER:       omp.dispatch.cond:
+// CHECK1-IRBUILDER-NEXT:    [[OMP_GLOBAL_THREAD_NUM10:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB8]])
+// CHECK1-IRBUILDER-NEXT:    [[TMP6:%.*]] = call i32 @__kmpc_dispatch_next_8(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM10]], i32* [[DOTOMP_IS_LAST]], i64* [[DOTOMP_LB]], i64* [[DOTOMP_UB]], i64* [[DOTOMP_STRIDE]])
+// CHECK1-IRBUILDER-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP6]], 0
+// CHECK1-IRBUILDER-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
+// CHECK1-IRBUILDER:       omp.dispatch.body:
+// CHECK1-IRBUILDER-NEXT:    [[TMP7:%.*]] = load i64, i64* [[DOTOMP_LB]], align 8
+// CHECK1-IRBUILDER-NEXT:    store i64 [[TMP7]], i64* [[DOTOMP_IV]], align 8
+// CHECK1-IRBUILDER-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
+// CHECK1-IRBUILDER:       omp.inner.for.cond:
+// CHECK1-IRBUILDER-NEXT:    [[TMP8:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
+// CHECK1-IRBUILDER-NEXT:    [[TMP9:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
+// CHECK1-IRBUILDER-NEXT:    [[CMP11:%.*]] = icmp sle i64 [[TMP8]], [[TMP9]]
+// CHECK1-IRBUILDER-NEXT:    br i1 [[CMP11]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// CHECK1-IRBUILDER:       omp.inner.for.body:
+// CHECK1-IRBUILDER-NEXT:    [[TMP10:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1
+// CHECK1-IRBUILDER-NEXT:    [[CONV12:%.*]] = sext i8 [[TMP10]] to i64
+// CHECK1-IRBUILDER-NEXT:    [[TMP11:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
+// CHECK1-IRBUILDER-NEXT:    [[DIV13:%.*]] = sdiv i64 [[TMP11]], 11
+// CHECK1-IRBUILDER-NEXT:    [[MUL14:%.*]] = mul nsw i64 [[DIV13]], 1
+// CHECK1-IRBUILDER-NEXT:    [[ADD15:%.*]] = add nsw i64 [[CONV12]], [[MUL14]]
+// CHECK1-IRBUILDER-NEXT:    [[CONV16:%.*]] = trunc i64 [[ADD15]] to i8
+// CHECK1-IRBUILDER-NEXT:    store i8 [[CONV16]], i8* [[I8]], align 1
+// CHECK1-IRBUILDER-NEXT:    [[TMP12:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
+// CHECK1-IRBUILDER-NEXT:    [[TMP13:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
+// CHECK1-IRBUILDER-NEXT:    [[DIV17:%.*]] = sdiv i64 [[TMP13]], 11
+// CHECK1-IRBUILDER-NEXT:    [[MUL18:%.*]] = mul nsw i64 [[DIV17]], 11
+// CHECK1-IRBUILDER-NEXT:    [[SUB19:%.*]] = sub nsw i64 [[TMP12]], [[MUL18]]
+// CHECK1-IRBUILDER-NEXT:    [[MUL20:%.*]] = mul nsw i64 [[SUB19]], 1
+// CHECK1-IRBUILDER-NEXT:    [[SUB21:%.*]] = sub nsw i64 11, [[MUL20]]
+// CHECK1-IRBUILDER-NEXT:    [[CONV22:%.*]] = trunc i64 [[SUB21]] to i32
+// CHECK1-IRBUILDER-NEXT:    store i32 [[CONV22]], i32* [[X9]], align 4
+// CHECK1-IRBUILDER-NEXT:    [[OMP_GLOBAL_THREAD_NUM23:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
+// CHECK1-IRBUILDER-NEXT:    call void @__kmpc_ordered(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM23]])
+// CHECK1-IRBUILDER-NEXT:    [[TMP14:%.*]] = load float*, float** [[B_ADDR]], align 8
+// CHECK1-IRBUILDER-NEXT:    [[TMP15:%.*]] = load i8, i8* [[I8]], align 1
+// CHECK1-IRBUILDER-NEXT:    [[IDXPROM:%.*]] = sext i8 [[TMP15]] to i64
+// CHECK1-IRBUILDER-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, float* [[TMP14]], i64 [[IDXPROM]]
+// CHECK1-IRBUILDER-NEXT:    [[TMP16:%.*]] = load float, float* [[ARRAYIDX]], align 4
+// CHECK1-IRBUILDER-NEXT:    [[TMP17:%.*]] = load float*, float** [[C_ADDR]], align 8
+// CHECK1-IRBUILDER-NEXT:    [[TMP18:%.*]] = load i8, i8* [[I8]], align 1
+// CHECK1-IRBUILDER-NEXT:    [[IDXPROM24:%.*]] = sext i8 [[TMP18]] to i64
+// CHECK1-IRBUILDER-NEXT:    [[ARRAYIDX25:%.*]] = getelementptr inbounds float, float* [[TMP17]], i64 [[IDXPROM24]]
+// CHECK1-IRBUILDER-NEXT:    [[TMP19:%.*]] = load float, float* [[ARRAYIDX25]], align 4
+// CHECK1-IRBUILDER-NEXT:    [[MUL26:%.*]] = fmul float [[TMP16]], [[TMP19]]
+// CHECK1-IRBUILDER-NEXT:    [[TMP20:%.*]] = load float*, float** [[D_ADDR]], align 8
+// CHECK1-IRBUILDER-NEXT:    [[TMP21:%.*]] = load i8, i8* [[I8]], align 1
+// CHECK1-IRBUILDER-NEXT:    [[IDXPROM27:%.*]] = sext i8 [[TMP21]] to i64
+// CHECK1-IRBUILDER-NEXT:    [[ARRAYIDX28:%.*]] = getelementptr inbounds float, float* [[TMP20]], i64 [[IDXPROM27]]
+// CHECK1-IRBUILDER-NEXT:    [[TMP22:%.*]] = load float, float* [[ARRAYIDX28]], align 4
+// CHECK1-IRBUILDER-NEXT:    [[MUL29:%.*]] = fmul float [[MUL26]], [[TMP22]]
+// CHECK1-IRBUILDER-NEXT:    [[TMP23:%.*]] = load float*, float** [[A_ADDR]], align 8
+// CHECK1-IRBUILDER-NEXT:    [[TMP24:%.*]] = load i8, i8* [[I8]], align 1
+// CHECK1-IRBUILDER-NEXT:    [[IDXPROM30:%.*]] = sext i8 [[TMP24]] to i64
+// CHECK1-IRBUILDER-NEXT:    [[ARRAYIDX31:%.*]] = getelementptr inbounds float, float* [[TMP23]], i64 [[IDXPROM30]]
+// CHECK1-IRBUILDER-NEXT:    store float [[MUL29]], float* [[ARRAYIDX31]], align 4
+// CHECK1-IRBUILDER-NEXT:    br label [[OMP_INNER_FOR_BODY_ORDERED_AFTER:%.*]]
+// CHECK1-IRBUILDER:       omp.inner.for.body.ordered.after:
+// CHECK1-IRBUILDER-NEXT:    call void @__kmpc_end_ordered(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM23]])
+// CHECK1-IRBUILDER-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
+// CHECK1-IRBUILDER:       omp.body.continue:
+// CHECK1-IRBUILDER-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
+// CHECK1-IRBUILDER:       omp.inner.for.inc:
+// CHECK1-IRBUILDER-NEXT:    [[TMP25:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
+// CHECK1-IRBUILDER-NEXT:    [[ADD32:%.*]] = add nsw i64 [[TMP25]], 1
+// CHECK1-IRBUILDER-NEXT:    store i64 [[ADD32]], i64* [[DOTOMP_IV]], align 8
+// CHECK1-IRBUILDER-NEXT:    [[OMP_GLOBAL_THREAD_NUM33:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB8]])
+// CHECK1-IRBUILDER-NEXT:    call void @__kmpc_dispatch_fini_8(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM33]])
+// CHECK1-IRBUILDER-NEXT:    br label [[OMP_INNER_FOR_COND]]
+// CHECK1-IRBUILDER:       omp.inner.for.end:
+// CHECK1-IRBUILDER-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
+// CHECK1-IRBUILDER:       omp.dispatch.inc:
+// CHECK1-IRBUILDER-NEXT:    br label [[OMP_DISPATCH_COND]]
+// CHECK1-IRBUILDER:       omp.dispatch.end:
+// CHECK1-IRBUILDER-NEXT:    br label [[OMP_PRECOND_END]]
+// CHECK1-IRBUILDER:       omp.precond.end:
+// CHECK1-IRBUILDER-NEXT:    [[OMP_GLOBAL_THREAD_NUM34:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
+// CHECK1-IRBUILDER-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB4]], i32 [[OMP_GLOBAL_THREAD_NUM34]])
+// CHECK1-IRBUILDER-NEXT:    ret void
+//
+//
+// CHECK1-IRBUILDER-LABEL: define {{[^@]+}}@_Z7runtimePfS_S_S_
+// CHECK1-IRBUILDER-SAME: (float* noundef [[A:%.*]], float* noundef [[B:%.*]], float* noundef [[C:%.*]], float* noundef [[D:%.*]]) #[[ATTR0]] {
+// CHECK1-IRBUILDER-NEXT:  entry:
+// CHECK1-IRBUILDER-NEXT:    [[A_ADDR:%.*]] = alloca float*, align 8
+// CHECK1-IRBUILDER-NEXT:    [[B_ADDR:%.*]] = alloca float*, align 8
+// CHECK1-IRBUILDER-NEXT:    [[C_ADDR:%.*]] = alloca float*, align 8
+// CHECK1-IRBUILDER-NEXT:    [[D_ADDR:%.*]] = alloca float*, align 8
+// CHECK1-IRBUILDER-NEXT:    [[X:%.*]] = alloca i32, align 4
+// CHECK1-IRBUILDER-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
+// CHECK1-IRBUILDER-NEXT:    [[TMP:%.*]] = alloca i8, align 1
+// CHECK1-IRBUILDER-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
+// CHECK1-IRBUILDER-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
+// CHECK1-IRBUILDER-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
+// CHECK1-IRBUILDER-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
+// CHECK1-IRBUILDER-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
+// CHECK1-IRBUILDER-NEXT:    [[I:%.*]] = alloca i8, align 1
+// CHECK1-IRBUILDER-NEXT:    [[X2:%.*]] = alloca i32, align 4
+// CHECK1-IRBUILDER-NEXT:    store float* [[A]], float** [[A_ADDR]], align 8
+// CHECK1-IRBUILDER-NEXT:    store float* [[B]], float** [[B_ADDR]], align 8
+// CHECK1-IRBUILDER-NEXT:    store float* [[C]], float** [[C_ADDR]], align 8
+// CHECK1-IRBUILDER-NEXT:    store float* [[D]], float** [[D_ADDR]], align 8
+// CHECK1-IRBUILDER-NEXT:    store i32 0, i32* [[X]], align 4
+// CHECK1-IRBUILDER-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
+// CHECK1-IRBUILDER-NEXT:    store i32 199, i32* [[DOTOMP_UB]], align 4
+// CHECK1-IRBUILDER-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
+// CHECK1-IRBUILDER-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
+// CHECK1-IRBUILDER-NEXT:    [[OMP_GLOBAL_THREAD_NUM:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB10:[0-9]+]])
+// CHECK1-IRBUILDER-NEXT:    call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM]], i32 1073741893, i32 0, i32 199, i32 1, i32 1)
+// CHECK1-IRBUILDER-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
+// CHECK1-IRBUILDER:       omp.dispatch.cond:
+// CHECK1-IRBUILDER-NEXT:    [[OMP_GLOBAL_THREAD_NUM3:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB10]])
+// CHECK1-IRBUILDER-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM3]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
+// CHECK1-IRBUILDER-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP0]], 0
+// CHECK1-IRBUILDER-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
+// CHECK1-IRBUILDER:       omp.dispatch.body:
+// CHECK1-IRBUILDER-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
+// CHECK1-IRBUILDER-NEXT:    store i32 [[TMP1]], i32* [[DOTOMP_IV]], align 4
+// CHECK1-IRBUILDER-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
+// CHECK1-IRBUILDER:       omp.inner.for.cond:
+// CHECK1-IRBUILDER-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
+// CHECK1-IRBUILDER-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
+// CHECK1-IRBUILDER-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP2]], [[TMP3]]
+// CHECK1-IRBUILDER-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// CHECK1-IRBUILDER:       omp.inner.for.body:
+// CHECK1-IRBUILDER-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
+// CHECK1-IRBUILDER-NEXT:    [[DIV:%.*]] = sdiv i32 [[TMP4]], 20
+// CHECK1-IRBUILDER-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV]], 1
+// CHECK1-IRBUILDER-NEXT:    [[ADD:%.*]] = add nsw i32 48, [[MUL]]
+// CHECK1-IRBUILDER-NEXT:    [[CONV:%.*]] = trunc i32 [[ADD]] to i8
+// CHECK1-IRBUILDER-NEXT:    store i8 [[CONV]], i8* [[I]], align 1
+// CHECK1-IRBUILDER-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
+// CHECK1-IRBUILDER-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
+// CHECK1-IRBUILDER-NEXT:    [[DIV4:%.*]] = sdiv i32 [[TMP6]], 20
+// CHECK1-IRBUILDER-NEXT:    [[MUL5:%.*]] = mul nsw i32 [[DIV4]], 20
+// CHECK1-IRBUILDER-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP5]], [[MUL5]]
+// CHECK1-IRBUILDER-NEXT:    [[MUL6:%.*]] = mul nsw i32 [[SUB]], 1
+// CHECK1-IRBUILDER-NEXT:    [[ADD7:%.*]] = add nsw i32 -10, [[MUL6]]
+// CHECK1-IRBUILDER-NEXT:    store i32 [[ADD7]], i32* [[X2]], align 4
+// CHECK1-IRBUILDER-NEXT:    [[OMP_GLOBAL_THREAD_NUM8:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
+// CHECK1-IRBUILDER-NEXT:    call void @__kmpc_ordered(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM8]])
+// CHECK1-IRBUILDER-NEXT:    [[TMP7:%.*]] = load float*, float** [[B_ADDR]], align 8
+// CHECK1-IRBUILDER-NEXT:    [[TMP8:%.*]] = load i8, i8* [[I]], align 1
+// CHECK1-IRBUILDER-NEXT:    [[IDXPROM:%.*]] = zext i8 [[TMP8]] to i64
+// CHECK1-IRBUILDER-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, float* [[TMP7]], i64 [[IDXPROM]]
+// CHECK1-IRBUILDER-NEXT:    [[TMP9:%.*]] = load float, float* [[ARRAYIDX]], align 4
+// CHECK1-IRBUILDER-NEXT:    [[TMP10:%.*]] = load float*, float** [[C_ADDR]], align 8
+// CHECK1-IRBUILDER-NEXT:    [[TMP11:%.*]] = load i8, i8* [[I]], align 1
+// CHECK1-IRBUILDER-NEXT:    [[IDXPROM9:%.*]] = zext i8 [[TMP11]] to i64
+// CHECK1-IRBUILDER-NEXT:    [[ARRAYIDX10:%.*]] = getelementptr inbounds float, float* [[TMP10]], i64 [[IDXPROM9]]
+// CHECK1-IRBUILDER-NEXT:    [[TMP12:%.*]] = load float, float* [[ARRAYIDX10]], align 4
+// CHECK1-IRBUILDER-NEXT:    [[MUL11:%.*]] = fmul float [[TMP9]], [[TMP12]]
+// CHECK1-IRBUILDER-NEXT:    [[TMP13:%.*]] = load float*, float** [[D_ADDR]], align 8
+// CHECK1-IRBUILDER-NEXT:    [[TMP14:%.*]] = load i8, i8* [[I]], align 1
+// CHECK1-IRBUILDER-NEXT:    [[IDXPROM12:%.*]] = zext i8 [[TMP14]] to i64
+// CHECK1-IRBUILDER-NEXT:    [[ARRAYIDX13:%.*]] = getelementptr inbounds float, float* [[TMP13]], i64 [[IDXPROM12]]
+// CHECK1-IRBUILDER-NEXT:    [[TMP15:%.*]] = load float, float* [[ARRAYIDX13]], align 4
+// CHECK1-IRBUILDER-NEXT:    [[MUL14:%.*]] = fmul float [[MUL11]], [[TMP15]]
+// CHECK1-IRBUILDER-NEXT:    [[TMP16:%.*]] = load float*, float** [[A_ADDR]], align 8
+// CHECK1-IRBUILDER-NEXT:    [[TMP17:%.*]] = load i8, i8* [[I]], align 1
+// CHECK1-IRBUILDER-NEXT:    [[IDXPROM15:%.*]] = zext i8 [[TMP17]] to i64
+// CHECK1-IRBUILDER-NEXT:    [[ARRAYIDX16:%.*]] = getelementptr inbounds float, float* [[TMP16]], i64 [[IDXPROM15]]
+// CHECK1-IRBUILDER-NEXT:    store float [[MUL14]], float* [[ARRAYIDX16]], align 4
+// CHECK1-IRBUILDER-NEXT:    br label [[OMP_INNER_FOR_BODY_ORDERED_AFTER:%.*]]
+// CHECK1-IRBUILDER:       omp.inner.for.body.ordered.after:
+// CHECK1-IRBUILDER-NEXT:    call void @__kmpc_end_ordered(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM8]])
+// CHECK1-IRBUILDER-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
+// CHECK1-IRBUILDER:       omp.body.continue:
+// CHECK1-IRBUILDER-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
+// CHECK1-IRBUILDER:       omp.inner.for.inc:
+// CHECK1-IRBUILDER-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
+// CHECK1-IRBUILDER-NEXT:    [[ADD17:%.*]] = add nsw i32 [[TMP18]], 1
+// CHECK1-IRBUILDER-NEXT:    store i32 [[ADD17]], i32* [[DOTOMP_IV]], align 4
+// CHECK1-IRBUILDER-NEXT:    [[OMP_GLOBAL_THREAD_NUM18:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB10]])
+// CHECK1-IRBUILDER-NEXT:    call void @__kmpc_dispatch_fini_4(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM18]])
+// CHECK1-IRBUILDER-NEXT:    br label [[OMP_INNER_FOR_COND]]
+// CHECK1-IRBUILDER:       omp.inner.for.end:
+// CHECK1-IRBUILDER-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
+// CHECK1-IRBUILDER:       omp.dispatch.inc:
+// CHECK1-IRBUILDER-NEXT:    br label [[OMP_DISPATCH_COND]]
+// CHECK1-IRBUILDER:       omp.dispatch.end:
+// CHECK1-IRBUILDER-NEXT:    [[OMP_GLOBAL_THREAD_NUM19:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
+// CHECK1-IRBUILDER-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB4]], i32 [[OMP_GLOBAL_THREAD_NUM19]])
+// CHECK1-IRBUILDER-NEXT:    ret void
+//
+//
+// CHECK1-IRBUILDER-LABEL: define {{[^@]+}}@_Z8foo_simdii
+// CHECK1-IRBUILDER-SAME: (i32 noundef [[LOW:%.*]], i32 noundef [[UP:%.*]]) #[[ATTR0]] {
+// CHECK1-IRBUILDER-NEXT:  entry:
+// CHECK1-IRBUILDER-NEXT:    [[LOW_ADDR:%.*]] = alloca i32, align 4
+// CHECK1-IRBUILDER-NEXT:    [[UP_ADDR:%.*]] = alloca i32, align 4
+// CHECK1-IRBUILDER-NEXT:    [[TMP:%.*]] = alloca i32, align 4
+// CHECK1-IRBUILDER-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
+// CHECK1-IRBUILDER-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
+// CHECK1-IRBUILDER-NEXT:    [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
+// CHECK1-IRBUILDER-NEXT:    [[I:%.*]] = alloca i32, align 4
+// CHECK1-IRBUILDER-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
+// CHECK1-IRBUILDER-NEXT:    [[I5:%.*]] = alloca i32, align 4
+// CHECK1-IRBUILDER-NEXT:    [[DOTOMP_IV16:%.*]] = alloca i32, align 4
+// CHECK1-IRBUILDER-NEXT:    [[_TMP17:%.*]] = alloca i32, align 4
+// CHECK1-IRBUILDER-NEXT:    [[DOTCAPTURE_EXPR_18:%.*]] = alloca i32, align 4
+// CHECK1-IRBUILDER-NEXT:    [[DOTCAPTURE_EXPR_19:%.*]] = alloca i32, align 4
+// CHECK1-IRBUILDER-NEXT:    [[DOTCAPTURE_EXPR_20:%.*]] = alloca i32, align 4
+// CHECK1-IRBUILDER-NEXT:    [[I26:%.*]] = alloca i32, align 4
+// CHECK1-IRBUILDER-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
+// CHECK1-IRBUILDER-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
+// CHECK1-IRBUILDER-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
+// CHECK1-IRBUILDER-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
+// CHECK1-IRBUILDER-NEXT:    [[I28:%.*]] = alloca i32, align 4
+// CHECK1-IRBUILDER-NEXT:    store i32 [[LOW]], i32* [[LOW_ADDR]], align 4
+// CHECK1-IRBUILDER-NEXT:    store i32 [[UP]], i32* [[UP_ADDR]], align 4
+// CHECK1-IRBUILDER-NEXT:    [[TMP0:%.*]] = load i32, i32* [[LOW_ADDR]], align 4
+// CHECK1-IRBUILDER-NEXT:    store i32 [[TMP0]], i32* [[DOTCAPTURE_EXPR_]], align 4
+// CHECK1-IRBUILDER-NEXT:    [[TMP1:%.*]] = load i32, i32* [[UP_ADDR]], align 4
+// CHECK1-IRBUILDER-NEXT:    store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_1]], align 4
+// CHECK1-IRBUILDER-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
+// CHECK1-IRBUILDER-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
+// CHECK1-IRBUILDER-NEXT:    [[SUB:%.*]] = sub i32 [[TMP2]], [[TMP3]]
+// CHECK1-IRBUILDER-NEXT:    [[SUB3:%.*]] = sub i32 [[SUB]], 1
+// CHECK1-IRBUILDER-NEXT:    [[ADD:%.*]] = add i32 [[SUB3]], 1
+// CHECK1-IRBUILDER-NEXT:    [[DIV:%.*]] = udiv i32 [[ADD]], 1
+// CHECK1-IRBUILDER-NEXT:    [[SUB4:%.*]] = sub i32 [[DIV]], 1
+// CHECK1-IRBUILDER-NEXT:    store i32 [[SUB4]], i32* [[DOTCAPTURE_EXPR_2]], align 4
+// CHECK1-IRBUILDER-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
+// CHECK1-IRBUILDER-NEXT:    store i32 [[TMP4]], i32* [[I]], align 4
+// CHECK1-IRBUILDER-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
+// CHECK1-IRBUILDER-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
+// CHECK1-IRBUILDER-NEXT:    [[CMP:%.*]] = icmp slt i32 [[TMP5]], [[TMP6]]
+// CHECK1-IRBUILDER-NEXT:    br i1 [[CMP]], label [[SIMD_IF_THEN:%.*]], label [[SIMD_IF_END:%.*]]
+// CHECK1-IRBUILDER:       simd.if.then:
+// CHECK1-IRBUILDER-NEXT:    store i32 0, i32* [[DOTOMP_IV]], align 4
+// CHECK1-IRBUILDER-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
+// CHECK1-IRBUILDER:       omp.inner.for.cond:
+// CHECK1-IRBUILDER-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !3
+// CHECK1-IRBUILDER-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4, !llvm.access.group !3
+// CHECK1-IRBUILDER-NEXT:    [[ADD6:%.*]] = add i32 [[TMP8]], 1
+// CHECK1-IRBUILDER-NEXT:    [[CMP7:%.*]] = icmp ult i32 [[TMP7]], [[ADD6]]
+// CHECK1-IRBUILDER-NEXT:    br i1 [[CMP7]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// CHECK1-IRBUILDER:       omp.inner.for.body:
+// CHECK1-IRBUILDER-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4, !llvm.access.group !3
+// CHECK1-IRBUILDER-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !3
+// CHECK1-IRBUILDER-NEXT:    [[MUL:%.*]] = mul i32 [[TMP10]], 1
+// CHECK1-IRBUILDER-NEXT:    [[ADD8:%.*]] = add i32 [[TMP9]], [[MUL]]
+// CHECK1-IRBUILDER-NEXT:    store i32 [[ADD8]], i32* [[I5]], align 4, !llvm.access.group !3
+// CHECK1-IRBUILDER-NEXT:    [[TMP11:%.*]] = load i32, i32* [[I5]], align 4, !llvm.access.group !3
+// CHECK1-IRBUILDER-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64
+// CHECK1-IRBUILDER-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x float], [10 x float]* @f, i64 0, i64 [[IDXPROM]]
+// CHECK1-IRBUILDER-NEXT:    store float 0.000000e+00, float* [[ARRAYIDX]], align 4, !llvm.access.group !3
+// CHECK1-IRBUILDER-NEXT:    call void @__captured_stmt(i32* [[I5]]), !llvm.access.group !3
+// CHECK1-IRBUILDER-NEXT:    br label [[OMP_INNER_FOR_BODY_ORDERED_AFTER:%.*]]
+// CHECK1-IRBUILDER:       omp.inner.for.body.ordered.after:
+// CHECK1-IRBUILDER-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
+// CHECK1-IRBUILDER:       omp.body.continue:
+// CHECK1-IRBUILDER-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
+// CHECK1-IRBUILDER:       omp.inner.for.inc:
+// CHECK1-IRBUILDER-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !3
+// CHECK1-IRBUILDER-NEXT:    [[ADD9:%.*]] = add i32 [[TMP12]], 1
+// CHECK1-IRBUILDER-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !3
+// CHECK1-IRBUILDER-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP4:![0-9]+]]
+// CHECK1-IRBUILDER:       omp.inner.for.end:
+// CHECK1-IRBUILDER-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
+// CHECK1-IRBUILDER-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
+// CHECK1-IRBUILDER-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
+// CHECK1-IRBUILDER-NEXT:    [[SUB10:%.*]] = sub i32 [[TMP14]], [[TMP15]]
+// CHECK1-IRBUILDER-NEXT:    [[SUB11:%.*]] = sub i32 [[SUB10]], 1
+// CHECK1-IRBUILDER-NEXT:    [[ADD12:%.*]] = add i32 [[SUB11]], 1
+// CHECK1-IRBUILDER-NEXT:    [[DIV13:%.*]] = udiv i32 [[ADD12]], 1
+// CHECK1-IRBUILDER-NEXT:    [[MUL14:%.*]] = mul i32 [[DIV13]], 1
+// CHECK1-IRBUILDER-NEXT:    [[ADD15:%.*]] = add i32 [[TMP13]], [[MUL14]]
+// CHECK1-IRBUILDER-NEXT:    store i32 [[ADD15]], i32* [[I5]], align 4
+// CHECK1-IRBUILDER-NEXT:    br label [[SIMD_IF_END]]
+// CHECK1-IRBUILDER:       simd.if.end:
+// CHECK1-IRBUILDER-NEXT:    [[TMP16:%.*]] = load i32, i32* [[LOW_ADDR]], align 4
+// CHECK1-IRBUILDER-NEXT:    store i32 [[TMP16]], i32* [[DOTCAPTURE_EXPR_18]], align 4
+// CHECK1-IRBUILDER-NEXT:    [[TMP17:%.*]] = load i32, i32* [[UP_ADDR]], align 4
+// CHECK1-IRBUILDER-NEXT:    store i32 [[TMP17]], i32* [[DOTCAPTURE_EXPR_19]], align 4
+// CHECK1-IRBUILDER-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_19]], align 4
+// CHECK1-IRBUILDER-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_18]], align 4
+// CHECK1-IRBUILDER-NEXT:    [[SUB21:%.*]] = sub i32 [[TMP18]], [[TMP19]]
+// CHECK1-IRBUILDER-NEXT:    [[SUB22:%.*]] = sub i32 [[SUB21]], 1
+// CHECK1-IRBUILDER-NEXT:    [[ADD23:%.*]] = add i32 [[SUB22]], 1
+// CHECK1-IRBUILDER-NEXT:    [[DIV24:%.*]] = udiv i32 [[ADD23]], 1
+// CHECK1-IRBUILDER-NEXT:    [[SUB25:%.*]] = sub i32 [[DIV24]], 1
+// CHECK1-IRBUILDER-NEXT:    store i32 [[SUB25]], i32* [[DOTCAPTURE_EXPR_20]], align 4
+// CHECK1-IRBUILDER-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_18]], align 4
+// CHECK1-IRBUILDER-NEXT:    store i32 [[TMP20]], i32* [[I26]], align 4
+// CHECK1-IRBUILDER-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_18]], align 4
+// CHECK1-IRBUILDER-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_19]], align 4
+// CHECK1-IRBUILDER-NEXT:    [[CMP27:%.*]] = icmp slt i32 [[TMP21]], [[TMP22]]
+// CHECK1-IRBUILDER-NEXT:    br i1 [[CMP27]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
+// CHECK1-IRBUILDER:       omp.precond.then:
+// CHECK1-IRBUILDER-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
+// CHECK1-IRBUILDER-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_20]], align 4
+// CHECK1-IRBUILDER-NEXT:    store i32 [[TMP23]], i32* [[DOTOMP_UB]], align 4
+// CHECK1-IRBUILDER-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
+// CHECK1-IRBUILDER-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
+// CHECK1-IRBUILDER-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_20]], align 4
+// CHECK1-IRBUILDER-NEXT:    [[OMP_GLOBAL_THREAD_NUM:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB12:[0-9]+]])
+// CHECK1-IRBUILDER-NEXT:    call void @__kmpc_dispatch_init_4u(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM]], i32 66, i32 0, i32 [[TMP24]], i32 1, i32 1)
+// CHECK1-IRBUILDER-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
+// CHECK1-IRBUILDER:       omp.dispatch.cond:
+// CHECK1-IRBUILDER-NEXT:    [[OMP_GLOBAL_THREAD_NUM29:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB12]])
+// CHECK1-IRBUILDER-NEXT:    [[TMP25:%.*]] = call i32 @__kmpc_dispatch_next_4u(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM29]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
+// CHECK1-IRBUILDER-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP25]], 0
+// CHECK1-IRBUILDER-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
+// CHECK1-IRBUILDER:       omp.dispatch.body:
+// CHECK1-IRBUILDER-NEXT:    [[TMP26:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
+// CHECK1-IRBUILDER-NEXT:    store i32 [[TMP26]], i32* [[DOTOMP_IV16]], align 4
+// CHECK1-IRBUILDER-NEXT:    br label [[OMP_INNER_FOR_COND30:%.*]]
+// CHECK1-IRBUILDER:       omp.inner.for.cond30:
+// CHECK1-IRBUILDER-NEXT:    [[TMP27:%.*]] = load i32, i32* [[DOTOMP_IV16]], align 4, !llvm.access.group !7
+// CHECK1-IRBUILDER-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !7
+// CHECK1-IRBUILDER-NEXT:    [[ADD31:%.*]] = add i32 [[TMP28]], 1
+// CHECK1-IRBUILDER-NEXT:    [[CMP32:%.*]] = icmp ult i32 [[TMP27]], [[ADD31]]
+// CHECK1-IRBUILDER-NEXT:    br i1 [[CMP32]], label [[OMP_INNER_FOR_BODY33:%.*]], label [[OMP_INNER_FOR_END42:%.*]]
+// CHECK1-IRBUILDER:       omp.inner.for.body33:
+// CHECK1-IRBUILDER-NEXT:    [[TMP29:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_18]], align 4, !llvm.access.group !7
+// CHECK1-IRBUILDER-NEXT:    [[TMP30:%.*]] = load i32, i32* [[DOTOMP_IV16]], align 4, !llvm.access.group !7
+// CHECK1-IRBUILDER-NEXT:    [[MUL34:%.*]] = mul i32 [[TMP30]], 1
+// CHECK1-IRBUILDER-NEXT:    [[ADD35:%.*]] = add i32 [[TMP29]], [[MUL34]]
+// CHECK1-IRBUILDER-NEXT:    store i32 [[ADD35]], i32* [[I28]], align 4, !llvm.access.group !7
+// CHECK1-IRBUILDER-NEXT:    [[TMP31:%.*]] = load i32, i32* [[I28]], align 4, !llvm.access.group !7
+// CHECK1-IRBUILDER-NEXT:    [[IDXPROM36:%.*]] = sext i32 [[TMP31]] to i64
+// CHECK1-IRBUILDER-NEXT:    [[ARRAYIDX37:%.*]] = getelementptr inbounds [10 x float], [10 x float]* @f, i64 0, i64 [[IDXPROM36]]
+// CHECK1-IRBUILDER-NEXT:    store float 0.000000e+00, float* [[ARRAYIDX37]], align 4, !llvm.access.group !7
+// CHECK1-IRBUILDER-NEXT:    call void @__captured_stmt.1(i32* [[I28]]), !llvm.access.group !7
+// CHECK1-IRBUILDER-NEXT:    br label [[OMP_INNER_FOR_BODY33_ORDERED_AFTER:%.*]]
+// CHECK1-IRBUILDER:       omp.inner.for.body33.ordered.after:
+// CHECK1-IRBUILDER-NEXT:    br label [[OMP_BODY_CONTINUE38:%.*]]
+// CHECK1-IRBUILDER:       omp.body.continue38:
+// CHECK1-IRBUILDER-NEXT:    br label [[OMP_INNER_FOR_INC39:%.*]]
+// CHECK1-IRBUILDER:       omp.inner.for.inc39:
+// CHECK1-IRBUILDER-NEXT:    [[TMP32:%.*]] = load i32, i32* [[DOTOMP_IV16]], align 4, !llvm.access.group !7
+// CHECK1-IRBUILDER-NEXT:    [[ADD40:%.*]] = add i32 [[TMP32]], 1
+// CHECK1-IRBUILDER-NEXT:    store i32 [[ADD40]], i32* [[DOTOMP_IV16]], align 4, !llvm.access.group !7
+// CHECK1-IRBUILDER-NEXT:    [[OMP_GLOBAL_THREAD_NUM41:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB12]])
+// CHECK1-IRBUILDER-NEXT:    call void @__kmpc_dispatch_fini_4u(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM41]]), !llvm.access.group !7
+// CHECK1-IRBUILDER-NEXT:    br label [[OMP_INNER_FOR_COND30]], !llvm.loop [[LOOP8:![0-9]+]]
+// CHECK1-IRBUILDER:       omp.inner.for.end42:
+// CHECK1-IRBUILDER-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
+// CHECK1-IRBUILDER:       omp.dispatch.inc:
+// CHECK1-IRBUILDER-NEXT:    br label [[OMP_DISPATCH_COND]]
+// CHECK1-IRBUILDER:       omp.dispatch.end:
+// CHECK1-IRBUILDER-NEXT:    [[TMP33:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
+// CHECK1-IRBUILDER-NEXT:    [[TMP34:%.*]] = icmp ne i32 [[TMP33]], 0
+// CHECK1-IRBUILDER-NEXT:    br i1 [[TMP34]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
+// CHECK1-IRBUILDER:       .omp.final.then:
+// CHECK1-IRBUILDER-NEXT:    [[TMP35:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_18]], align 4
+// CHECK1-IRBUILDER-NEXT:    [[TMP36:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_19]], align 4
+// CHECK1-IRBUILDER-NEXT:    [[TMP37:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_18]], align 4
+// CHECK1-IRBUILDER-NEXT:    [[SUB43:%.*]] = sub i32 [[TMP36]], [[TMP37]]
+// CHECK1-IRBUILDER-NEXT:    [[SUB44:%.*]] = sub i32 [[SUB43]], 1
+// CHECK1-IRBUILDER-NEXT:    [[ADD45:%.*]] = add i32 [[SUB44]], 1
+// CHECK1-IRBUILDER-NEXT:    [[DIV46:%.*]] = udiv i32 [[ADD45]], 1
+// CHECK1-IRBUILDER-NEXT:    [[MUL47:%.*]] = mul i32 [[DIV46]], 1
+// CHECK1-IRBUILDER-NEXT:    [[ADD48:%.*]] = add i32 [[TMP35]], [[MUL47]]
+// CHECK1-IRBUILDER-NEXT:    store i32 [[ADD48]], i32* [[I28]], align 4
+// CHECK1-IRBUILDER-NEXT:    br label [[DOTOMP_FINAL_DONE]]
+// CHECK1-IRBUILDER:       .omp.final.done:
+// CHECK1-IRBUILDER-NEXT:    br label [[OMP_PRECOND_END]]
+// CHECK1-IRBUILDER:       omp.precond.end:
+// CHECK1-IRBUILDER-NEXT:    [[OMP_GLOBAL_THREAD_NUM49:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
+// CHECK1-IRBUILDER-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB4]], i32 [[OMP_GLOBAL_THREAD_NUM49]])
+// CHECK1-IRBUILDER-NEXT:    ret void
+//
+//
+// CHECK1-IRBUILDER-LABEL: define {{[^@]+}}@__captured_stmt
+// CHECK1-IRBUILDER-SAME: (i32* noundef nonnull align 4 dereferenceable(4) [[I:%.*]]) #[[ATTR3:[0-9]+]] {
+// CHECK1-IRBUILDER-NEXT:  entry:
+// CHECK1-IRBUILDER-NEXT:    [[I_ADDR:%.*]] = alloca i32*, align 8
+// CHECK1-IRBUILDER-NEXT:    store i32* [[I]], i32** [[I_ADDR]], align 8
+// CHECK1-IRBUILDER-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[I_ADDR]], align 8
+// CHECK1-IRBUILDER-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
+// CHECK1-IRBUILDER-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP1]] to i64
+// CHECK1-IRBUILDER-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x float], [10 x float]* @f, i64 0, i64 [[IDXPROM]]
+// CHECK1-IRBUILDER-NEXT:    store float 1.000000e+00, float* [[ARRAYIDX]], align 4
+// CHECK1-IRBUILDER-NEXT:    ret void
+//
+//
+// CHECK1-IRBUILDER-LABEL: define {{[^@]+}}@__captured_stmt.1
+// CHECK1-IRBUILDER-SAME: (i32* noundef nonnull align 4 dereferenceable(4) [[I:%.*]]) #[[ATTR3]] {
+// CHECK1-IRBUILDER-NEXT:  entry:
+// CHECK1-IRBUILDER-NEXT:    [[I_ADDR:%.*]] = alloca i32*, align 8
+// CHECK1-IRBUILDER-NEXT:    store i32* [[I]], i32** [[I_ADDR]], align 8
+// CHECK1-IRBUILDER-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[I_ADDR]], align 8
+// CHECK1-IRBUILDER-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
+// CHECK1-IRBUILDER-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP1]] to i64
+// CHECK1-IRBUILDER-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x float], [10 x float]* @f, i64 0, i64 [[IDXPROM]]
+// CHECK1-IRBUILDER-NEXT:    store float 1.000000e+00, float* [[ARRAYIDX]], align 4
+// CHECK1-IRBUILDER-NEXT:    ret void
+//
+//
+// CHECK2-IRBUILDER-LABEL: define {{[^@]+}}@_Z18static_not_chunkedPfS_S_S_
+// CHECK2-IRBUILDER-SAME: (float* noundef [[A:%.*]], float* noundef [[B:%.*]], float* noundef [[C:%.*]], float* noundef [[D:%.*]]) #[[ATTR0:[0-9]+]] {
+// CHECK2-IRBUILDER-NEXT:  entry:
+// CHECK2-IRBUILDER-NEXT:    [[A_ADDR:%.*]] = alloca float*, align 8
+// CHECK2-IRBUILDER-NEXT:    [[B_ADDR:%.*]] = alloca float*, align 8
+// CHECK2-IRBUILDER-NEXT:    [[C_ADDR:%.*]] = alloca float*, align 8
+// CHECK2-IRBUILDER-NEXT:    [[D_ADDR:%.*]] = alloca float*, align 8
+// CHECK2-IRBUILDER-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
+// CHECK2-IRBUILDER-NEXT:    [[TMP:%.*]] = alloca i32, align 4
+// CHECK2-IRBUILDER-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
+// CHECK2-IRBUILDER-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
+// CHECK2-IRBUILDER-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
+// CHECK2-IRBUILDER-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
+// CHECK2-IRBUILDER-NEXT:    [[I:%.*]] = alloca i32, align 4
+// CHECK2-IRBUILDER-NEXT:    store float* [[A]], float** [[A_ADDR]], align 8
+// CHECK2-IRBUILDER-NEXT:    store float* [[B]], float** [[B_ADDR]], align 8
+// CHECK2-IRBUILDER-NEXT:    store float* [[C]], float** [[C_ADDR]], align 8
+// CHECK2-IRBUILDER-NEXT:    store float* [[D]], float** [[D_ADDR]], align 8
+// CHECK2-IRBUILDER-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
+// CHECK2-IRBUILDER-NEXT:    store i32 4571423, i32* [[DOTOMP_UB]], align 4
+// CHECK2-IRBUILDER-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
+// CHECK2-IRBUILDER-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
+// CHECK2-IRBUILDER-NEXT:    [[OMP_GLOBAL_THREAD_NUM:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB3:[0-9]+]])
+// CHECK2-IRBUILDER-NEXT:    call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[OMP_GLOBAL_THREAD_NUM]], i32 66, i32 0, i32 4571423, i32 1, i32 1)
+// CHECK2-IRBUILDER-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
+// CHECK2-IRBUILDER:       omp.dispatch.cond:
+// CHECK2-IRBUILDER-NEXT:    [[OMP_GLOBAL_THREAD_NUM1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB3]])
+// CHECK2-IRBUILDER-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM1]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
+// CHECK2-IRBUILDER-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP0]], 0
+// CHECK2-IRBUILDER-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
+// CHECK2-IRBUILDER:       omp.dispatch.body:
+// CHECK2-IRBUILDER-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
+// CHECK2-IRBUILDER-NEXT:    store i32 [[TMP1]], i32* [[DOTOMP_IV]], align 4
+// CHECK2-IRBUILDER-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
+// CHECK2-IRBUILDER:       omp.inner.for.cond:
+// CHECK2-IRBUILDER-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
+// CHECK2-IRBUILDER-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
+// CHECK2-IRBUILDER-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP2]], [[TMP3]]
+// CHECK2-IRBUILDER-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// CHECK2-IRBUILDER:       omp.inner.for.body:
+// CHECK2-IRBUILDER-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
+// CHECK2-IRBUILDER-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP4]], 7
+// CHECK2-IRBUILDER-NEXT:    [[SUB:%.*]] = sub nsw i32 32000000, [[MUL]]
+// CHECK2-IRBUILDER-NEXT:    store i32 [[SUB]], i32* [[I]], align 4
+// CHECK2-IRBUILDER-NEXT:    [[OMP_GLOBAL_THREAD_NUM2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
+// CHECK2-IRBUILDER-NEXT:    call void @__kmpc_ordered(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM2]])
+// CHECK2-IRBUILDER-NEXT:    [[TMP5:%.*]] = load float*, float** [[B_ADDR]], align 8
+// CHECK2-IRBUILDER-NEXT:    [[TMP6:%.*]] = load i32, i32* [[I]], align 4
+// CHECK2-IRBUILDER-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP6]] to i64
+// CHECK2-IRBUILDER-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, float* [[TMP5]], i64 [[IDXPROM]]
+// CHECK2-IRBUILDER-NEXT:    [[TMP7:%.*]] = load float, float* [[ARRAYIDX]], align 4
+// CHECK2-IRBUILDER-NEXT:    [[TMP8:%.*]] = load float*, float** [[C_ADDR]], align 8
+// CHECK2-IRBUILDER-NEXT:    [[TMP9:%.*]] = load i32, i32* [[I]], align 4
+// CHECK2-IRBUILDER-NEXT:    [[IDXPROM3:%.*]] = sext i32 [[TMP9]] to i64
+// CHECK2-IRBUILDER-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds float, float* [[TMP8]], i64 [[IDXPROM3]]
+// CHECK2-IRBUILDER-NEXT:    [[TMP10:%.*]] = load float, float* [[ARRAYIDX4]], align 4
+// CHECK2-IRBUILDER-NEXT:    [[MUL5:%.*]] = fmul float [[TMP7]], [[TMP10]]
+// CHECK2-IRBUILDER-NEXT:    [[TMP11:%.*]] = load float*, float** [[D_ADDR]], align 8
+// CHECK2-IRBUILDER-NEXT:    [[TMP12:%.*]] = load i32, i32* [[I]], align 4
+// CHECK2-IRBUILDER-NEXT:    [[IDXPROM6:%.*]] = sext i32 [[TMP12]] to i64
+// CHECK2-IRBUILDER-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds float, float* [[TMP11]], i64 [[IDXPROM6]]
+// CHECK2-IRBUILDER-NEXT:    [[TMP13:%.*]] = load float, float* [[ARRAYIDX7]], align 4
+// CHECK2-IRBUILDER-NEXT:    [[MUL8:%.*]] = fmul float [[MUL5]], [[TMP13]]
+// CHECK2-IRBUILDER-NEXT:    [[TMP14:%.*]] = load float*, float** [[A_ADDR]], align 8
+// CHECK2-IRBUILDER-NEXT:    [[TMP15:%.*]] = load i32, i32* [[I]], align 4
+// CHECK2-IRBUILDER-NEXT:    [[IDXPROM9:%.*]] = sext i32 [[TMP15]] to i64
+// CHECK2-IRBUILDER-NEXT:    [[ARRAYIDX10:%.*]] = getelementptr inbounds float, float* [[TMP14]], i64 [[IDXPROM9]]
+// CHECK2-IRBUILDER-NEXT:    store float [[MUL8]], float* [[ARRAYIDX10]], align 4
+// CHECK2-IRBUILDER-NEXT:    br label [[OMP_INNER_FOR_BODY_ORDERED_AFTER:%.*]]
+// CHECK2-IRBUILDER:       omp.inner.for.body.ordered.after:
+// CHECK2-IRBUILDER-NEXT:    call void @__kmpc_end_ordered(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM2]])
+// CHECK2-IRBUILDER-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
+// CHECK2-IRBUILDER:       omp.body.continue:
+// CHECK2-IRBUILDER-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
+// CHECK2-IRBUILDER:       omp.inner.for.inc:
+// CHECK2-IRBUILDER-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
+// CHECK2-IRBUILDER-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP16]], 1
+// CHECK2-IRBUILDER-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
+// CHECK2-IRBUILDER-NEXT:    [[OMP_GLOBAL_THREAD_NUM11:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB3]])
+// CHECK2-IRBUILDER-NEXT:    call void @__kmpc_dispatch_fini_4(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM11]])
+// CHECK2-IRBUILDER-NEXT:    br label [[OMP_INNER_FOR_COND]]
+// CHECK2-IRBUILDER:       omp.inner.for.end:
+// CHECK2-IRBUILDER-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
+// CHECK2-IRBUILDER:       omp.dispatch.inc:
+// CHECK2-IRBUILDER-NEXT:    br label [[OMP_DISPATCH_COND]]
+// CHECK2-IRBUILDER:       omp.dispatch.end:
+// CHECK2-IRBUILDER-NEXT:    [[OMP_GLOBAL_THREAD_NUM12:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
+// CHECK2-IRBUILDER-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB4:[0-9]+]], i32 [[OMP_GLOBAL_THREAD_NUM12]])
+// CHECK2-IRBUILDER-NEXT:    ret void
+//
+//
+// CHECK2-IRBUILDER-LABEL: define {{[^@]+}}@_Z8dynamic1PfS_S_S_
+// CHECK2-IRBUILDER-SAME: (float* noundef [[A:%.*]], float* noundef [[B:%.*]], float* noundef [[C:%.*]], float* noundef [[D:%.*]]) #[[ATTR0]] {
+// CHECK2-IRBUILDER-NEXT:  entry:
+// CHECK2-IRBUILDER-NEXT:    [[A_ADDR:%.*]] = alloca float*, align 8
+// CHECK2-IRBUILDER-NEXT:    [[B_ADDR:%.*]] = alloca float*, align 8
+// CHECK2-IRBUILDER-NEXT:    [[C_ADDR:%.*]] = alloca float*, align 8
+// CHECK2-IRBUILDER-NEXT:    [[D_ADDR:%.*]] = alloca float*, align 8
+// CHECK2-IRBUILDER-NEXT:    [[DOTOMP_IV:%.*]] = alloca i64, align 8
+// CHECK2-IRBUILDER-NEXT:    [[TMP:%.*]] = alloca i64, align 8
+// CHECK2-IRBUILDER-NEXT:    [[DOTOMP_LB:%.*]] = alloca i64, align 8
+// CHECK2-IRBUILDER-NEXT:    [[DOTOMP_UB:%.*]] = alloca i64, align 8
+// CHECK2-IRBUILDER-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i64, align 8
+// CHECK2-IRBUILDER-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
+// CHECK2-IRBUILDER-NEXT:    [[I:%.*]] = alloca i64, align 8
+// CHECK2-IRBUILDER-NEXT:    store float* [[A]], float** [[A_ADDR]], align 8
+// CHECK2-IRBUILDER-NEXT:    store float* [[B]], float** [[B_ADDR]], align 8
+// CHECK2-IRBUILDER-NEXT:    store float* [[C]], float** [[C_ADDR]], align 8
+// CHECK2-IRBUILDER-NEXT:    store float* [[D]], float** [[D_ADDR]], align 8
+// CHECK2-IRBUILDER-NEXT:    store i64 0, i64* [[DOTOMP_LB]], align 8
+// CHECK2-IRBUILDER-NEXT:    store i64 16908287, i64* [[DOTOMP_UB]], align 8
+// CHECK2-IRBUILDER-NEXT:    store i64 1, i64* [[DOTOMP_STRIDE]], align 8
+// CHECK2-IRBUILDER-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
+// CHECK2-IRBUILDER-NEXT:    [[OMP_GLOBAL_THREAD_NUM:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB6:[0-9]+]])
+// CHECK2-IRBUILDER-NEXT:    call void @__kmpc_dispatch_init_8u(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM]], i32 1073741891, i64 0, i64 16908287, i64 1, i64 1)
+// CHECK2-IRBUILDER-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
+// CHECK2-IRBUILDER:       omp.dispatch.cond:
+// CHECK2-IRBUILDER-NEXT:    [[OMP_GLOBAL_THREAD_NUM1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB6]])
+// CHECK2-IRBUILDER-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_dispatch_next_8u(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM1]], i32* [[DOTOMP_IS_LAST]], i64* [[DOTOMP_LB]], i64* [[DOTOMP_UB]], i64* [[DOTOMP_STRIDE]])
+// CHECK2-IRBUILDER-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP0]], 0
+// CHECK2-IRBUILDER-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
+// CHECK2-IRBUILDER:       omp.dispatch.body:
+// CHECK2-IRBUILDER-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTOMP_LB]], align 8
+// CHECK2-IRBUILDER-NEXT:    store i64 [[TMP1]], i64* [[DOTOMP_IV]], align 8
+// CHECK2-IRBUILDER-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
+// CHECK2-IRBUILDER:       omp.inner.for.cond:
+// CHECK2-IRBUILDER-NEXT:    [[TMP2:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
+// CHECK2-IRBUILDER-NEXT:    [[TMP3:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
+// CHECK2-IRBUILDER-NEXT:    [[ADD:%.*]] = add i64 [[TMP3]], 1
+// CHECK2-IRBUILDER-NEXT:    [[CMP:%.*]] = icmp ult i64 [[TMP2]], [[ADD]]
+// CHECK2-IRBUILDER-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// CHECK2-IRBUILDER:       omp.inner.for.body:
+// CHECK2-IRBUILDER-NEXT:    [[TMP4:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
+// CHECK2-IRBUILDER-NEXT:    [[MUL:%.*]] = mul i64 [[TMP4]], 127
+// CHECK2-IRBUILDER-NEXT:    [[ADD2:%.*]] = add i64 131071, [[MUL]]
+// CHECK2-IRBUILDER-NEXT:    store i64 [[ADD2]], i64* [[I]], align 8
+// CHECK2-IRBUILDER-NEXT:    [[OMP_GLOBAL_THREAD_NUM3:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
+// CHECK2-IRBUILDER-NEXT:    call void @__kmpc_ordered(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM3]])
+// CHECK2-IRBUILDER-NEXT:    [[TMP5:%.*]] = load float*, float** [[B_ADDR]], align 8
+// CHECK2-IRBUILDER-NEXT:    [[TMP6:%.*]] = load i64, i64* [[I]], align 8
+// CHECK2-IRBUILDER-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, float* [[TMP5]], i64 [[TMP6]]
+// CHECK2-IRBUILDER-NEXT:    [[TMP7:%.*]] = load float, float* [[ARRAYIDX]], align 4
+// CHECK2-IRBUILDER-NEXT:    [[TMP8:%.*]] = load float*, float** [[C_ADDR]], align 8
+// CHECK2-IRBUILDER-NEXT:    [[TMP9:%.*]] = load i64, i64* [[I]], align 8
+// CHECK2-IRBUILDER-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds float, float* [[TMP8]], i64 [[TMP9]]
+// CHECK2-IRBUILDER-NEXT:    [[TMP10:%.*]] = load float, float* [[ARRAYIDX4]], align 4
+// CHECK2-IRBUILDER-NEXT:    [[MUL5:%.*]] = fmul float [[TMP7]], [[TMP10]]
+// CHECK2-IRBUILDER-NEXT:    [[TMP11:%.*]] = load float*, float** [[D_ADDR]], align 8
+// CHECK2-IRBUILDER-NEXT:    [[TMP12:%.*]] = load i64, i64* [[I]], align 8
+// CHECK2-IRBUILDER-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds float, float* [[TMP11]], i64 [[TMP12]]
+// CHECK2-IRBUILDER-NEXT:    [[TMP13:%.*]] = load float, float* [[ARRAYIDX6]], align 4
+// CHECK2-IRBUILDER-NEXT:    [[MUL7:%.*]] = fmul float [[MUL5]], [[TMP13]]
+// CHECK2-IRBUILDER-NEXT:    [[TMP14:%.*]] = load float*, float** [[A_ADDR]], align 8
+// CHECK2-IRBUILDER-NEXT:    [[TMP15:%.*]] = load i64, i64* [[I]], align 8
+// CHECK2-IRBUILDER-NEXT:    [[ARRAYIDX8:%.*]] = getelementptr inbounds float, float* [[TMP14]], i64 [[TMP15]]
+// CHECK2-IRBUILDER-NEXT:    store float [[MUL7]], float* [[ARRAYIDX8]], align 4
+// CHECK2-IRBUILDER-NEXT:    br label [[OMP_INNER_FOR_BODY_ORDERED_AFTER:%.*]]
+// CHECK2-IRBUILDER:       omp.inner.for.body.ordered.after:
+// CHECK2-IRBUILDER-NEXT:    call void @__kmpc_end_ordered(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM3]])
+// CHECK2-IRBUILDER-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
+// CHECK2-IRBUILDER:       omp.body.continue:
+// CHECK2-IRBUILDER-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
+// CHECK2-IRBUILDER:       omp.inner.for.inc:
+// CHECK2-IRBUILDER-NEXT:    [[TMP16:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
+// CHECK2-IRBUILDER-NEXT:    [[ADD9:%.*]] = add i64 [[TMP16]], 1
+// CHECK2-IRBUILDER-NEXT:    store i64 [[ADD9]], i64* [[DOTOMP_IV]], align 8
+// CHECK2-IRBUILDER-NEXT:    [[OMP_GLOBAL_THREAD_NUM10:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB6]])
+// CHECK2-IRBUILDER-NEXT:    call void @__kmpc_dispatch_fini_8u(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM10]])
+// CHECK2-IRBUILDER-NEXT:    br label [[OMP_INNER_FOR_COND]]
+// CHECK2-IRBUILDER:       omp.inner.for.end:
+// CHECK2-IRBUILDER-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
+// CHECK2-IRBUILDER:       omp.dispatch.inc:
+// CHECK2-IRBUILDER-NEXT:    br label [[OMP_DISPATCH_COND]]
+// CHECK2-IRBUILDER:       omp.dispatch.end:
+// CHECK2-IRBUILDER-NEXT:    [[OMP_GLOBAL_THREAD_NUM11:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
+// CHECK2-IRBUILDER-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB4]], i32 [[OMP_GLOBAL_THREAD_NUM11]])
+// CHECK2-IRBUILDER-NEXT:    ret void
+//
+//
+// CHECK2-IRBUILDER-LABEL: define {{[^@]+}}@_Z9test_autoPfS_S_S_
+// CHECK2-IRBUILDER-SAME: (float* noundef [[A:%.*]], float* noundef [[B:%.*]], float* noundef [[C:%.*]], float* noundef [[D:%.*]]) #[[ATTR0]] {
+// CHECK2-IRBUILDER-NEXT:  entry:
+// CHECK2-IRBUILDER-NEXT:    [[A_ADDR:%.*]] = alloca float*, align 8
+// CHECK2-IRBUILDER-NEXT:    [[B_ADDR:%.*]] = alloca float*, align 8
+// CHECK2-IRBUILDER-NEXT:    [[C_ADDR:%.*]] = alloca float*, align 8
+// CHECK2-IRBUILDER-NEXT:    [[D_ADDR:%.*]] = alloca float*, align 8
+// CHECK2-IRBUILDER-NEXT:    [[X:%.*]] = alloca i32, align 4
+// CHECK2-IRBUILDER-NEXT:    [[Y:%.*]] = alloca i32, align 4
+// CHECK2-IRBUILDER-NEXT:    [[DOTOMP_IV:%.*]] = alloca i64, align 8
+// CHECK2-IRBUILDER-NEXT:    [[TMP:%.*]] = alloca i8, align 1
+// CHECK2-IRBUILDER-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
+// CHECK2-IRBUILDER-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i8, align 1
+// CHECK2-IRBUILDER-NEXT:    [[DOTCAPTURE_EXPR_2:%.*]] = alloca i64, align 8
+// CHECK2-IRBUILDER-NEXT:    [[I:%.*]] = alloca i8, align 1
+// CHECK2-IRBUILDER-NEXT:    [[X6:%.*]] = alloca i32, align 4
+// CHECK2-IRBUILDER-NEXT:    [[DOTOMP_LB:%.*]] = alloca i64, align 8
+// CHECK2-IRBUILDER-NEXT:    [[DOTOMP_UB:%.*]] = alloca i64, align 8
+// CHECK2-IRBUILDER-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i64, align 8
+// CHECK2-IRBUILDER-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
+// CHECK2-IRBUILDER-NEXT:    [[I8:%.*]] = alloca i8, align 1
+// CHECK2-IRBUILDER-NEXT:    [[X9:%.*]] = alloca i32, align 4
+// CHECK2-IRBUILDER-NEXT:    store float* [[A]], float** [[A_ADDR]], align 8
+// CHECK2-IRBUILDER-NEXT:    store float* [[B]], float** [[B_ADDR]], align 8
+// CHECK2-IRBUILDER-NEXT:    store float* [[C]], float** [[C_ADDR]], align 8
+// CHECK2-IRBUILDER-NEXT:    store float* [[D]], float** [[D_ADDR]], align 8
+// CHECK2-IRBUILDER-NEXT:    store i32 0, i32* [[X]], align 4
+// CHECK2-IRBUILDER-NEXT:    store i32 0, i32* [[Y]], align 4
+// CHECK2-IRBUILDER-NEXT:    [[TMP0:%.*]] = load i32, i32* [[Y]], align 4
+// CHECK2-IRBUILDER-NEXT:    [[CONV:%.*]] = trunc i32 [[TMP0]] to i8
+// CHECK2-IRBUILDER-NEXT:    store i8 [[CONV]], i8* [[DOTCAPTURE_EXPR_]], align 1
+// CHECK2-IRBUILDER-NEXT:    [[TMP1:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1
+// CHECK2-IRBUILDER-NEXT:    [[CONV3:%.*]] = sext i8 [[TMP1]] to i32
+// CHECK2-IRBUILDER-NEXT:    [[SUB:%.*]] = sub i32 57, [[CONV3]]
+// CHECK2-IRBUILDER-NEXT:    [[ADD:%.*]] = add i32 [[SUB]], 1
+// CHECK2-IRBUILDER-NEXT:    [[DIV:%.*]] = udiv i32 [[ADD]], 1
+// CHECK2-IRBUILDER-NEXT:    [[CONV4:%.*]] = zext i32 [[DIV]] to i64
+// CHECK2-IRBUILDER-NEXT:    [[MUL:%.*]] = mul nsw i64 [[CONV4]], 11
+// CHECK2-IRBUILDER-NEXT:    [[SUB5:%.*]] = sub nsw i64 [[MUL]], 1
+// CHECK2-IRBUILDER-NEXT:    store i64 [[SUB5]], i64* [[DOTCAPTURE_EXPR_2]], align 8
+// CHECK2-IRBUILDER-NEXT:    [[TMP2:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1
+// CHECK2-IRBUILDER-NEXT:    store i8 [[TMP2]], i8* [[I]], align 1
+// CHECK2-IRBUILDER-NEXT:    store i32 11, i32* [[X6]], align 4
+// CHECK2-IRBUILDER-NEXT:    [[TMP3:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1
+// CHECK2-IRBUILDER-NEXT:    [[CONV7:%.*]] = sext i8 [[TMP3]] to i32
+// CHECK2-IRBUILDER-NEXT:    [[CMP:%.*]] = icmp sle i32 [[CONV7]], 57
+// CHECK2-IRBUILDER-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
+// CHECK2-IRBUILDER:       omp.precond.then:
+// CHECK2-IRBUILDER-NEXT:    store i64 0, i64* [[DOTOMP_LB]], align 8
+// CHECK2-IRBUILDER-NEXT:    [[TMP4:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR_2]], align 8
+// CHECK2-IRBUILDER-NEXT:    store i64 [[TMP4]], i64* [[DOTOMP_UB]], align 8
+// CHECK2-IRBUILDER-NEXT:    store i64 1, i64* [[DOTOMP_STRIDE]], align 8
+// CHECK2-IRBUILDER-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
+// CHECK2-IRBUILDER-NEXT:    [[TMP5:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR_2]], align 8
+// CHECK2-IRBUILDER-NEXT:    [[OMP_GLOBAL_THREAD_NUM:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB8:[0-9]+]])
+// CHECK2-IRBUILDER-NEXT:    call void @__kmpc_dispatch_init_8(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM]], i32 1073741894, i64 0, i64 [[TMP5]], i64 1, i64 1)
+// CHECK2-IRBUILDER-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
+// CHECK2-IRBUILDER:       omp.dispatch.cond:
+// CHECK2-IRBUILDER-NEXT:    [[OMP_GLOBAL_THREAD_NUM10:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB8]])
+// CHECK2-IRBUILDER-NEXT:    [[TMP6:%.*]] = call i32 @__kmpc_dispatch_next_8(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM10]], i32* [[DOTOMP_IS_LAST]], i64* [[DOTOMP_LB]], i64* [[DOTOMP_UB]], i64* [[DOTOMP_STRIDE]])
+// CHECK2-IRBUILDER-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP6]], 0
+// CHECK2-IRBUILDER-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
+// CHECK2-IRBUILDER:       omp.dispatch.body:
+// CHECK2-IRBUILDER-NEXT:    [[TMP7:%.*]] = load i64, i64* [[DOTOMP_LB]], align 8
+// CHECK2-IRBUILDER-NEXT:    store i64 [[TMP7]], i64* [[DOTOMP_IV]], align 8
+// CHECK2-IRBUILDER-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
+// CHECK2-IRBUILDER:       omp.inner.for.cond:
+// CHECK2-IRBUILDER-NEXT:    [[TMP8:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
+// CHECK2-IRBUILDER-NEXT:    [[TMP9:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
+// CHECK2-IRBUILDER-NEXT:    [[CMP11:%.*]] = icmp sle i64 [[TMP8]], [[TMP9]]
+// CHECK2-IRBUILDER-NEXT:    br i1 [[CMP11]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// CHECK2-IRBUILDER:       omp.inner.for.body:
+// CHECK2-IRBUILDER-NEXT:    [[TMP10:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1
+// CHECK2-IRBUILDER-NEXT:    [[CONV12:%.*]] = sext i8 [[TMP10]] to i64
+// CHECK2-IRBUILDER-NEXT:    [[TMP11:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
+// CHECK2-IRBUILDER-NEXT:    [[DIV13:%.*]] = sdiv i64 [[TMP11]], 11
+// CHECK2-IRBUILDER-NEXT:    [[MUL14:%.*]] = mul nsw i64 [[DIV13]], 1
+// CHECK2-IRBUILDER-NEXT:    [[ADD15:%.*]] = add nsw i64 [[CONV12]], [[MUL14]]
+// CHECK2-IRBUILDER-NEXT:    [[CONV16:%.*]] = trunc i64 [[ADD15]] to i8
+// CHECK2-IRBUILDER-NEXT:    store i8 [[CONV16]], i8* [[I8]], align 1
+// CHECK2-IRBUILDER-NEXT:    [[TMP12:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
+// CHECK2-IRBUILDER-NEXT:    [[TMP13:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
+// CHECK2-IRBUILDER-NEXT:    [[DIV17:%.*]] = sdiv i64 [[TMP13]], 11
+// CHECK2-IRBUILDER-NEXT:    [[MUL18:%.*]] = mul nsw i64 [[DIV17]], 11
+// CHECK2-IRBUILDER-NEXT:    [[SUB19:%.*]] = sub nsw i64 [[TMP12]], [[MUL18]]
+// CHECK2-IRBUILDER-NEXT:    [[MUL20:%.*]] = mul nsw i64 [[SUB19]], 1
+// CHECK2-IRBUILDER-NEXT:    [[SUB21:%.*]] = sub nsw i64 11, [[MUL20]]
+// CHECK2-IRBUILDER-NEXT:    [[CONV22:%.*]] = trunc i64 [[SUB21]] to i32
+// CHECK2-IRBUILDER-NEXT:    store i32 [[CONV22]], i32* [[X9]], align 4
+// CHECK2-IRBUILDER-NEXT:    [[OMP_GLOBAL_THREAD_NUM23:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
+// CHECK2-IRBUILDER-NEXT:    call void @__kmpc_ordered(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM23]])
+// CHECK2-IRBUILDER-NEXT:    [[TMP14:%.*]] = load float*, float** [[B_ADDR]], align 8
+// CHECK2-IRBUILDER-NEXT:    [[TMP15:%.*]] = load i8, i8* [[I8]], align 1
+// CHECK2-IRBUILDER-NEXT:    [[IDXPROM:%.*]] = sext i8 [[TMP15]] to i64
+// CHECK2-IRBUILDER-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, float* [[TMP14]], i64 [[IDXPROM]]
+// CHECK2-IRBUILDER-NEXT:    [[TMP16:%.*]] = load float, float* [[ARRAYIDX]], align 4
+// CHECK2-IRBUILDER-NEXT:    [[TMP17:%.*]] = load float*, float** [[C_ADDR]], align 8
+// CHECK2-IRBUILDER-NEXT:    [[TMP18:%.*]] = load i8, i8* [[I8]], align 1
+// CHECK2-IRBUILDER-NEXT:    [[IDXPROM24:%.*]] = sext i8 [[TMP18]] to i64
+// CHECK2-IRBUILDER-NEXT:    [[ARRAYIDX25:%.*]] = getelementptr inbounds float, float* [[TMP17]], i64 [[IDXPROM24]]
+// CHECK2-IRBUILDER-NEXT:    [[TMP19:%.*]] = load float, float* [[ARRAYIDX25]], align 4
+// CHECK2-IRBUILDER-NEXT:    [[MUL26:%.*]] = fmul float [[TMP16]], [[TMP19]]
+// CHECK2-IRBUILDER-NEXT:    [[TMP20:%.*]] = load float*, float** [[D_ADDR]], align 8
+// CHECK2-IRBUILDER-NEXT:    [[TMP21:%.*]] = load i8, i8* [[I8]], align 1
+// CHECK2-IRBUILDER-NEXT:    [[IDXPROM27:%.*]] = sext i8 [[TMP21]] to i64
+// CHECK2-IRBUILDER-NEXT:    [[ARRAYIDX28:%.*]] = getelementptr inbounds float, float* [[TMP20]], i64 [[IDXPROM27]]
+// CHECK2-IRBUILDER-NEXT:    [[TMP22:%.*]] = load float, float* [[ARRAYIDX28]], align 4
+// CHECK2-IRBUILDER-NEXT:    [[MUL29:%.*]] = fmul float [[MUL26]], [[TMP22]]
+// CHECK2-IRBUILDER-NEXT:    [[TMP23:%.*]] = load float*, float** [[A_ADDR]], align 8
+// CHECK2-IRBUILDER-NEXT:    [[TMP24:%.*]] = load i8, i8* [[I8]], align 1
+// CHECK2-IRBUILDER-NEXT:    [[IDXPROM30:%.*]] = sext i8 [[TMP24]] to i64
+// CHECK2-IRBUILDER-NEXT:    [[ARRAYIDX31:%.*]] = getelementptr inbounds float, float* [[TMP23]], i64 [[IDXPROM30]]
+// CHECK2-IRBUILDER-NEXT:    store float [[MUL29]], float* [[ARRAYIDX31]], align 4
+// CHECK2-IRBUILDER-NEXT:    br label [[OMP_INNER_FOR_BODY_ORDERED_AFTER:%.*]]
+// CHECK2-IRBUILDER:       omp.inner.for.body.ordered.after:
+// CHECK2-IRBUILDER-NEXT:    call void @__kmpc_end_ordered(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM23]])
+// CHECK2-IRBUILDER-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
+// CHECK2-IRBUILDER:       omp.body.continue:
+// CHECK2-IRBUILDER-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
+// CHECK2-IRBUILDER:       omp.inner.for.inc:
+// CHECK2-IRBUILDER-NEXT:    [[TMP25:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
+// CHECK2-IRBUILDER-NEXT:    [[ADD32:%.*]] = add nsw i64 [[TMP25]], 1
+// CHECK2-IRBUILDER-NEXT:    store i64 [[ADD32]], i64* [[DOTOMP_IV]], align 8
+// CHECK2-IRBUILDER-NEXT:    [[OMP_GLOBAL_THREAD_NUM33:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB8]])
+// CHECK2-IRBUILDER-NEXT:    call void @__kmpc_dispatch_fini_8(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM33]])
+// CHECK2-IRBUILDER-NEXT:    br label [[OMP_INNER_FOR_COND]]
+// CHECK2-IRBUILDER:       omp.inner.for.end:
+// CHECK2-IRBUILDER-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
+// CHECK2-IRBUILDER:       omp.dispatch.inc:
+// CHECK2-IRBUILDER-NEXT:    br label [[OMP_DISPATCH_COND]]
+// CHECK2-IRBUILDER:       omp.dispatch.end:
+// CHECK2-IRBUILDER-NEXT:    br label [[OMP_PRECOND_END]]
+// CHECK2-IRBUILDER:       omp.precond.end:
+// CHECK2-IRBUILDER-NEXT:    [[OMP_GLOBAL_THREAD_NUM34:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
+// CHECK2-IRBUILDER-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB4]], i32 [[OMP_GLOBAL_THREAD_NUM34]])
+// CHECK2-IRBUILDER-NEXT:    ret void
+//
+//
+// CHECK2-IRBUILDER-LABEL: define {{[^@]+}}@_Z7runtimePfS_S_S_
+// CHECK2-IRBUILDER-SAME: (float* noundef [[A:%.*]], float* noundef [[B:%.*]], float* noundef [[C:%.*]], float* noundef [[D:%.*]]) #[[ATTR0]] {
+// CHECK2-IRBUILDER-NEXT:  entry:
+// CHECK2-IRBUILDER-NEXT:    [[A_ADDR:%.*]] = alloca float*, align 8
+// CHECK2-IRBUILDER-NEXT:    [[B_ADDR:%.*]] = alloca float*, align 8
+// CHECK2-IRBUILDER-NEXT:    [[C_ADDR:%.*]] = alloca float*, align 8
+// CHECK2-IRBUILDER-NEXT:    [[D_ADDR:%.*]] = alloca float*, align 8
+// CHECK2-IRBUILDER-NEXT:    [[X:%.*]] = alloca i32, align 4
+// CHECK2-IRBUILDER-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
+// CHECK2-IRBUILDER-NEXT:    [[TMP:%.*]] = alloca i8, align 1
+// CHECK2-IRBUILDER-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
+// CHECK2-IRBUILDER-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
+// CHECK2-IRBUILDER-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
+// CHECK2-IRBUILDER-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
+// CHECK2-IRBUILDER-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
+// CHECK2-IRBUILDER-NEXT:    [[I:%.*]] = alloca i8, align 1
+// CHECK2-IRBUILDER-NEXT:    [[X2:%.*]] = alloca i32, align 4
+// CHECK2-IRBUILDER-NEXT:    store float* [[A]], float** [[A_ADDR]], align 8
+// CHECK2-IRBUILDER-NEXT:    store float* [[B]], float** [[B_ADDR]], align 8
+// CHECK2-IRBUILDER-NEXT:    store float* [[C]], float** [[C_ADDR]], align 8
+// CHECK2-IRBUILDER-NEXT:    store float* [[D]], float** [[D_ADDR]], align 8
+// CHECK2-IRBUILDER-NEXT:    store i32 0, i32* [[X]], align 4
+// CHECK2-IRBUILDER-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
+// CHECK2-IRBUILDER-NEXT:    store i32 199, i32* [[DOTOMP_UB]], align 4
+// CHECK2-IRBUILDER-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
+// CHECK2-IRBUILDER-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
+// CHECK2-IRBUILDER-NEXT:    [[OMP_GLOBAL_THREAD_NUM:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB10:[0-9]+]])
+// CHECK2-IRBUILDER-NEXT:    call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM]], i32 1073741893, i32 0, i32 199, i32 1, i32 1)
+// CHECK2-IRBUILDER-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
+// CHECK2-IRBUILDER:       omp.dispatch.cond:
+// CHECK2-IRBUILDER-NEXT:    [[OMP_GLOBAL_THREAD_NUM3:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB10]])
+// CHECK2-IRBUILDER-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM3]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
+// CHECK2-IRBUILDER-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP0]], 0
+// CHECK2-IRBUILDER-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
+// CHECK2-IRBUILDER:       omp.dispatch.body:
+// CHECK2-IRBUILDER-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
+// CHECK2-IRBUILDER-NEXT:    store i32 [[TMP1]], i32* [[DOTOMP_IV]], align 4
+// CHECK2-IRBUILDER-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
+// CHECK2-IRBUILDER:       omp.inner.for.cond:
+// CHECK2-IRBUILDER-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
+// CHECK2-IRBUILDER-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
+// CHECK2-IRBUILDER-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP2]], [[TMP3]]
+// CHECK2-IRBUILDER-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// CHECK2-IRBUILDER:       omp.inner.for.body:
+// CHECK2-IRBUILDER-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
+// CHECK2-IRBUILDER-NEXT:    [[DIV:%.*]] = sdiv i32 [[TMP4]], 20
+// CHECK2-IRBUILDER-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV]], 1
+// CHECK2-IRBUILDER-NEXT:    [[ADD:%.*]] = add nsw i32 48, [[MUL]]
+// CHECK2-IRBUILDER-NEXT:    [[CONV:%.*]] = trunc i32 [[ADD]] to i8
+// CHECK2-IRBUILDER-NEXT:    store i8 [[CONV]], i8* [[I]], align 1
+// CHECK2-IRBUILDER-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
+// CHECK2-IRBUILDER-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
+// CHECK2-IRBUILDER-NEXT:    [[DIV4:%.*]] = sdiv i32 [[TMP6]], 20
+// CHECK2-IRBUILDER-NEXT:    [[MUL5:%.*]] = mul nsw i32 [[DIV4]], 20
+// CHECK2-IRBUILDER-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP5]], [[MUL5]]
+// CHECK2-IRBUILDER-NEXT:    [[MUL6:%.*]] = mul nsw i32 [[SUB]], 1
+// CHECK2-IRBUILDER-NEXT:    [[ADD7:%.*]] = add nsw i32 -10, [[MUL6]]
+// CHECK2-IRBUILDER-NEXT:    store i32 [[ADD7]], i32* [[X2]], align 4
+// CHECK2-IRBUILDER-NEXT:    [[OMP_GLOBAL_THREAD_NUM8:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
+// CHECK2-IRBUILDER-NEXT:    call void @__kmpc_ordered(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM8]])
+// CHECK2-IRBUILDER-NEXT:    [[TMP7:%.*]] = load float*, float** [[B_ADDR]], align 8
+// CHECK2-IRBUILDER-NEXT:    [[TMP8:%.*]] = load i8, i8* [[I]], align 1
+// CHECK2-IRBUILDER-NEXT:    [[IDXPROM:%.*]] = zext i8 [[TMP8]] to i64
+// CHECK2-IRBUILDER-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, float* [[TMP7]], i64 [[IDXPROM]]
+// CHECK2-IRBUILDER-NEXT:    [[TMP9:%.*]] = load float, float* [[ARRAYIDX]], align 4
+// CHECK2-IRBUILDER-NEXT:    [[TMP10:%.*]] = load float*, float** [[C_ADDR]], align 8
+// CHECK2-IRBUILDER-NEXT:    [[TMP11:%.*]] = load i8, i8* [[I]], align 1
+// CHECK2-IRBUILDER-NEXT:    [[IDXPROM9:%.*]] = zext i8 [[TMP11]] to i64
+// CHECK2-IRBUILDER-NEXT:    [[ARRAYIDX10:%.*]] = getelementptr inbounds float, float* [[TMP10]], i64 [[IDXPROM9]]
+// CHECK2-IRBUILDER-NEXT:    [[TMP12:%.*]] = load float, float* [[ARRAYIDX10]], align 4
+// CHECK2-IRBUILDER-NEXT:    [[MUL11:%.*]] = fmul float [[TMP9]], [[TMP12]]
+// CHECK2-IRBUILDER-NEXT:    [[TMP13:%.*]] = load float*, float** [[D_ADDR]], align 8
+// CHECK2-IRBUILDER-NEXT:    [[TMP14:%.*]] = load i8, i8* [[I]], align 1
+// CHECK2-IRBUILDER-NEXT:    [[IDXPROM12:%.*]] = zext i8 [[TMP14]] to i64
+// CHECK2-IRBUILDER-NEXT:    [[ARRAYIDX13:%.*]] = getelementptr inbounds float, float* [[TMP13]], i64 [[IDXPROM12]]
+// CHECK2-IRBUILDER-NEXT:    [[TMP15:%.*]] = load float, float* [[ARRAYIDX13]], align 4
+// CHECK2-IRBUILDER-NEXT:    [[MUL14:%.*]] = fmul float [[MUL11]], [[TMP15]]
+// CHECK2-IRBUILDER-NEXT:    [[TMP16:%.*]] = load float*, float** [[A_ADDR]], align 8
+// CHECK2-IRBUILDER-NEXT:    [[TMP17:%.*]] = load i8, i8* [[I]], align 1
+// CHECK2-IRBUILDER-NEXT:    [[IDXPROM15:%.*]] = zext i8 [[TMP17]] to i64
+// CHECK2-IRBUILDER-NEXT:    [[ARRAYIDX16:%.*]] = getelementptr inbounds float, float* [[TMP16]], i64 [[IDXPROM15]]
+// CHECK2-IRBUILDER-NEXT:    store float [[MUL14]], float* [[ARRAYIDX16]], align 4
+// CHECK2-IRBUILDER-NEXT:    br label [[OMP_INNER_FOR_BODY_ORDERED_AFTER:%.*]]
+// CHECK2-IRBUILDER:       omp.inner.for.body.ordered.after:
+// CHECK2-IRBUILDER-NEXT:    call void @__kmpc_end_ordered(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM8]])
+// CHECK2-IRBUILDER-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
+// CHECK2-IRBUILDER:       omp.body.continue:
+// CHECK2-IRBUILDER-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
+// CHECK2-IRBUILDER:       omp.inner.for.inc:
+// CHECK2-IRBUILDER-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
+// CHECK2-IRBUILDER-NEXT:    [[ADD17:%.*]] = add nsw i32 [[TMP18]], 1
+// CHECK2-IRBUILDER-NEXT:    store i32 [[ADD17]], i32* [[DOTOMP_IV]], align 4
+// CHECK2-IRBUILDER-NEXT:    [[OMP_GLOBAL_THREAD_NUM18:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB10]])
+// CHECK2-IRBUILDER-NEXT:    call void @__kmpc_dispatch_fini_4(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM18]])
+// CHECK2-IRBUILDER-NEXT:    br label [[OMP_INNER_FOR_COND]]
+// CHECK2-IRBUILDER:       omp.inner.for.end:
+// CHECK2-IRBUILDER-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
+// CHECK2-IRBUILDER:       omp.dispatch.inc:
+// CHECK2-IRBUILDER-NEXT:    br label [[OMP_DISPATCH_COND]]
+// CHECK2-IRBUILDER:       omp.dispatch.end:
+// CHECK2-IRBUILDER-NEXT:    [[OMP_GLOBAL_THREAD_NUM19:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
+// CHECK2-IRBUILDER-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB4]], i32 [[OMP_GLOBAL_THREAD_NUM19]])
+// CHECK2-IRBUILDER-NEXT:    ret void
+//
+//
+// CHECK2-IRBUILDER-LABEL: define {{[^@]+}}@_Z8foo_simdii
+// CHECK2-IRBUILDER-SAME: (i32 noundef [[LOW:%.*]], i32 noundef [[UP:%.*]]) #[[ATTR0]] {
+// CHECK2-IRBUILDER-NEXT:  entry:
+// CHECK2-IRBUILDER-NEXT:    [[LOW_ADDR:%.*]] = alloca i32, align 4
+// CHECK2-IRBUILDER-NEXT:    [[UP_ADDR:%.*]] = alloca i32, align 4
+// CHECK2-IRBUILDER-NEXT:    [[TMP:%.*]] = alloca i32, align 4
+// CHECK2-IRBUILDER-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
+// CHECK2-IRBUILDER-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
+// CHECK2-IRBUILDER-NEXT:    [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
+// CHECK2-IRBUILDER-NEXT:    [[I:%.*]] = alloca i32, align 4
+// CHECK2-IRBUILDER-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
+// CHECK2-IRBUILDER-NEXT:    [[I5:%.*]] = alloca i32, align 4
+// CHECK2-IRBUILDER-NEXT:    [[DOTOMP_IV16:%.*]] = alloca i32, align 4
+// CHECK2-IRBUILDER-NEXT:    [[_TMP17:%.*]] = alloca i32, align 4
+// CHECK2-IRBUILDER-NEXT:    [[DOTCAPTURE_EXPR_18:%.*]] = alloca i32, align 4
+// CHECK2-IRBUILDER-NEXT:    [[DOTCAPTURE_EXPR_19:%.*]] = alloca i32, align 4
+// CHECK2-IRBUILDER-NEXT:    [[DOTCAPTURE_EXPR_20:%.*]] = alloca i32, align 4
+// CHECK2-IRBUILDER-NEXT:    [[I26:%.*]] = alloca i32, align 4
+// CHECK2-IRBUILDER-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
+// CHECK2-IRBUILDER-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
+// CHECK2-IRBUILDER-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
+// CHECK2-IRBUILDER-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
+// CHECK2-IRBUILDER-NEXT:    [[I28:%.*]] = alloca i32, align 4
+// CHECK2-IRBUILDER-NEXT:    store i32 [[LOW]], i32* [[LOW_ADDR]], align 4
+// CHECK2-IRBUILDER-NEXT:    store i32 [[UP]], i32* [[UP_ADDR]], align 4
+// CHECK2-IRBUILDER-NEXT:    [[TMP0:%.*]] = load i32, i32* [[LOW_ADDR]], align 4
+// CHECK2-IRBUILDER-NEXT:    store i32 [[TMP0]], i32* [[DOTCAPTURE_EXPR_]], align 4
+// CHECK2-IRBUILDER-NEXT:    [[TMP1:%.*]] = load i32, i32* [[UP_ADDR]], align 4
+// CHECK2-IRBUILDER-NEXT:    store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_1]], align 4
+// CHECK2-IRBUILDER-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
+// CHECK2-IRBUILDER-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
+// CHECK2-IRBUILDER-NEXT:    [[SUB:%.*]] = sub i32 [[TMP2]], [[TMP3]]
+// CHECK2-IRBUILDER-NEXT:    [[SUB3:%.*]] = sub i32 [[SUB]], 1
+// CHECK2-IRBUILDER-NEXT:    [[ADD:%.*]] = add i32 [[SUB3]], 1
+// CHECK2-IRBUILDER-NEXT:    [[DIV:%.*]] = udiv i32 [[ADD]], 1
+// CHECK2-IRBUILDER-NEXT:    [[SUB4:%.*]] = sub i32 [[DIV]], 1
+// CHECK2-IRBUILDER-NEXT:    store i32 [[SUB4]], i32* [[DOTCAPTURE_EXPR_2]], align 4
+// CHECK2-IRBUILDER-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
+// CHECK2-IRBUILDER-NEXT:    store i32 [[TMP4]], i32* [[I]], align 4
+// CHECK2-IRBUILDER-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
+// CHECK2-IRBUILDER-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
+// CHECK2-IRBUILDER-NEXT:    [[CMP:%.*]] = icmp slt i32 [[TMP5]], [[TMP6]]
+// CHECK2-IRBUILDER-NEXT:    br i1 [[CMP]], label [[SIMD_IF_THEN:%.*]], label [[SIMD_IF_END:%.*]]
+// CHECK2-IRBUILDER:       simd.if.then:
+// CHECK2-IRBUILDER-NEXT:    store i32 0, i32* [[DOTOMP_IV]], align 4
+// CHECK2-IRBUILDER-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
+// CHECK2-IRBUILDER:       omp.inner.for.cond:
+// CHECK2-IRBUILDER-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !3
+// CHECK2-IRBUILDER-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4, !llvm.access.group !3
+// CHECK2-IRBUILDER-NEXT:    [[ADD6:%.*]] = add i32 [[TMP8]], 1
+// CHECK2-IRBUILDER-NEXT:    [[CMP7:%.*]] = icmp ult i32 [[TMP7]], [[ADD6]]
+// CHECK2-IRBUILDER-NEXT:    br i1 [[CMP7]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// CHECK2-IRBUILDER:       omp.inner.for.body:
+// CHECK2-IRBUILDER-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4, !llvm.access.group !3
+// CHECK2-IRBUILDER-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !3
+// CHECK2-IRBUILDER-NEXT:    [[MUL:%.*]] = mul i32 [[TMP10]], 1
+// CHECK2-IRBUILDER-NEXT:    [[ADD8:%.*]] = add i32 [[TMP9]], [[MUL]]
+// CHECK2-IRBUILDER-NEXT:    store i32 [[ADD8]], i32* [[I5]], align 4, !llvm.access.group !3
+// CHECK2-IRBUILDER-NEXT:    [[TMP11:%.*]] = load i32, i32* [[I5]], align 4, !llvm.access.group !3
+// CHECK2-IRBUILDER-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64
+// CHECK2-IRBUILDER-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x float], [10 x float]* @f, i64 0, i64 [[IDXPROM]]
+// CHECK2-IRBUILDER-NEXT:    store float 0.000000e+00, float* [[ARRAYIDX]], align 4, !llvm.access.group !3
+// CHECK2-IRBUILDER-NEXT:    call void @__captured_stmt(i32* [[I5]]), !llvm.access.group !3
+// CHECK2-IRBUILDER-NEXT:    br label [[OMP_INNER_FOR_BODY_ORDERED_AFTER:%.*]]
+// CHECK2-IRBUILDER:       omp.inner.for.body.ordered.after:
+// CHECK2-IRBUILDER-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
+// CHECK2-IRBUILDER:       omp.body.continue:
+// CHECK2-IRBUILDER-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
+// CHECK2-IRBUILDER:       omp.inner.for.inc:
+// CHECK2-IRBUILDER-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !3
+// CHECK2-IRBUILDER-NEXT:    [[ADD9:%.*]] = add i32 [[TMP12]], 1
+// CHECK2-IRBUILDER-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !3
+// CHECK2-IRBUILDER-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP4:![0-9]+]]
+// CHECK2-IRBUILDER:       omp.inner.for.end:
+// CHECK2-IRBUILDER-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
+// CHECK2-IRBUILDER-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
+// CHECK2-IRBUILDER-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
+// CHECK2-IRBUILDER-NEXT:    [[SUB10:%.*]] = sub i32 [[TMP14]], [[TMP15]]
+// CHECK2-IRBUILDER-NEXT:    [[SUB11:%.*]] = sub i32 [[SUB10]], 1
+// CHECK2-IRBUILDER-NEXT:    [[ADD12:%.*]] = add i32 [[SUB11]], 1
+// CHECK2-IRBUILDER-NEXT:    [[DIV13:%.*]] = udiv i32 [[ADD12]], 1
+// CHECK2-IRBUILDER-NEXT:    [[MUL14:%.*]] = mul i32 [[DIV13]], 1
+// CHECK2-IRBUILDER-NEXT:    [[ADD15:%.*]] = add i32 [[TMP13]], [[MUL14]]
+// CHECK2-IRBUILDER-NEXT:    store i32 [[ADD15]], i32* [[I5]], align 4
+// CHECK2-IRBUILDER-NEXT:    br label [[SIMD_IF_END]]
+// CHECK2-IRBUILDER:       simd.if.end:
+// CHECK2-IRBUILDER-NEXT:    [[TMP16:%.*]] = load i32, i32* [[LOW_ADDR]], align 4
+// CHECK2-IRBUILDER-NEXT:    store i32 [[TMP16]], i32* [[DOTCAPTURE_EXPR_18]], align 4
+// CHECK2-IRBUILDER-NEXT:    [[TMP17:%.*]] = load i32, i32* [[UP_ADDR]], align 4
+// CHECK2-IRBUILDER-NEXT:    store i32 [[TMP17]], i32* [[DOTCAPTURE_EXPR_19]], align 4
+// CHECK2-IRBUILDER-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_19]], align 4
+// CHECK2-IRBUILDER-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_18]], align 4
+// CHECK2-IRBUILDER-NEXT:    [[SUB21:%.*]] = sub i32 [[TMP18]], [[TMP19]]
+// CHECK2-IRBUILDER-NEXT:    [[SUB22:%.*]] = sub i32 [[SUB21]], 1
+// CHECK2-IRBUILDER-NEXT:    [[ADD23:%.*]] = add i32 [[SUB22]], 1
+// CHECK2-IRBUILDER-NEXT:    [[DIV24:%.*]] = udiv i32 [[ADD23]], 1
+// CHECK2-IRBUILDER-NEXT:    [[SUB25:%.*]] = sub i32 [[DIV24]], 1
+// CHECK2-IRBUILDER-NEXT:    store i32 [[SUB25]], i32* [[DOTCAPTURE_EXPR_20]], align 4
+// CHECK2-IRBUILDER-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_18]], align 4
+// CHECK2-IRBUILDER-NEXT:    store i32 [[TMP20]], i32* [[I26]], align 4
+// CHECK2-IRBUILDER-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_18]], align 4
+// CHECK2-IRBUILDER-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_19]], align 4
+// CHECK2-IRBUILDER-NEXT:    [[CMP27:%.*]] = icmp slt i32 [[TMP21]], [[TMP22]]
+// CHECK2-IRBUILDER-NEXT:    br i1 [[CMP27]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
+// CHECK2-IRBUILDER:       omp.precond.then:
+// CHECK2-IRBUILDER-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
+// CHECK2-IRBUILDER-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_20]], align 4
+// CHECK2-IRBUILDER-NEXT:    store i32 [[TMP23]], i32* [[DOTOMP_UB]], align 4
+// CHECK2-IRBUILDER-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
+// CHECK2-IRBUILDER-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
+// CHECK2-IRBUILDER-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_20]], align 4
+// CHECK2-IRBUILDER-NEXT:    [[OMP_GLOBAL_THREAD_NUM:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB12:[0-9]+]])
+// CHECK2-IRBUILDER-NEXT:    call void @__kmpc_dispatch_init_4u(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM]], i32 66, i32 0, i32 [[TMP24]], i32 1, i32 1)
+// CHECK2-IRBUILDER-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
+// CHECK2-IRBUILDER:       omp.dispatch.cond:
+// CHECK2-IRBUILDER-NEXT:    [[OMP_GLOBAL_THREAD_NUM29:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB12]])
+// CHECK2-IRBUILDER-NEXT:    [[TMP25:%.*]] = call i32 @__kmpc_dispatch_next_4u(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM29]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
+// CHECK2-IRBUILDER-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP25]], 0
+// CHECK2-IRBUILDER-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
+// CHECK2-IRBUILDER:       omp.dispatch.body:
+// CHECK2-IRBUILDER-NEXT:    [[TMP26:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
+// CHECK2-IRBUILDER-NEXT:    store i32 [[TMP26]], i32* [[DOTOMP_IV16]], align 4
+// CHECK2-IRBUILDER-NEXT:    br label [[OMP_INNER_FOR_COND30:%.*]]
+// CHECK2-IRBUILDER:       omp.inner.for.cond30:
+// CHECK2-IRBUILDER-NEXT:    [[TMP27:%.*]] = load i32, i32* [[DOTOMP_IV16]], align 4, !llvm.access.group !7
+// CHECK2-IRBUILDER-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !7
+// CHECK2-IRBUILDER-NEXT:    [[ADD31:%.*]] = add i32 [[TMP28]], 1
+// CHECK2-IRBUILDER-NEXT:    [[CMP32:%.*]] = icmp ult i32 [[TMP27]], [[ADD31]]
+// CHECK2-IRBUILDER-NEXT:    br i1 [[CMP32]], label [[OMP_INNER_FOR_BODY33:%.*]], label [[OMP_INNER_FOR_END42:%.*]]
+// CHECK2-IRBUILDER:       omp.inner.for.body33:
+// CHECK2-IRBUILDER-NEXT:    [[TMP29:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_18]], align 4, !llvm.access.group !7
+// CHECK2-IRBUILDER-NEXT:    [[TMP30:%.*]] = load i32, i32* [[DOTOMP_IV16]], align 4, !llvm.access.group !7
+// CHECK2-IRBUILDER-NEXT:    [[MUL34:%.*]] = mul i32 [[TMP30]], 1
+// CHECK2-IRBUILDER-NEXT:    [[ADD35:%.*]] = add i32 [[TMP29]], [[MUL34]]
+// CHECK2-IRBUILDER-NEXT:    store i32 [[ADD35]], i32* [[I28]], align 4, !llvm.access.group !7
+// CHECK2-IRBUILDER-NEXT:    [[TMP31:%.*]] = load i32, i32* [[I28]], align 4, !llvm.access.group !7
+// CHECK2-IRBUILDER-NEXT:    [[IDXPROM36:%.*]] = sext i32 [[TMP31]] to i64
+// CHECK2-IRBUILDER-NEXT:    [[ARRAYIDX37:%.*]] = getelementptr inbounds [10 x float], [10 x float]* @f, i64 0, i64 [[IDXPROM36]]
+// CHECK2-IRBUILDER-NEXT:    store float 0.000000e+00, float* [[ARRAYIDX37]], align 4, !llvm.access.group !7
+// CHECK2-IRBUILDER-NEXT:    call void @__captured_stmt.1(i32* [[I28]]), !llvm.access.group !7
+// CHECK2-IRBUILDER-NEXT:    br label [[OMP_INNER_FOR_BODY33_ORDERED_AFTER:%.*]]
+// CHECK2-IRBUILDER:       omp.inner.for.body33.ordered.after:
+// CHECK2-IRBUILDER-NEXT:    br label [[OMP_BODY_CONTINUE38:%.*]]
+// CHECK2-IRBUILDER:       omp.body.continue38:
+// CHECK2-IRBUILDER-NEXT:    br label [[OMP_INNER_FOR_INC39:%.*]]
+// CHECK2-IRBUILDER:       omp.inner.for.inc39:
+// CHECK2-IRBUILDER-NEXT:    [[TMP32:%.*]] = load i32, i32* [[DOTOMP_IV16]], align 4, !llvm.access.group !7
+// CHECK2-IRBUILDER-NEXT:    [[ADD40:%.*]] = add i32 [[TMP32]], 1
+// CHECK2-IRBUILDER-NEXT:    store i32 [[ADD40]], i32* [[DOTOMP_IV16]], align 4, !llvm.access.group !7
+// CHECK2-IRBUILDER-NEXT:    [[OMP_GLOBAL_THREAD_NUM41:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB12]])
+// CHECK2-IRBUILDER-NEXT:    call void @__kmpc_dispatch_fini_4u(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM41]]), !llvm.access.group !7
+// CHECK2-IRBUILDER-NEXT:    br label [[OMP_INNER_FOR_COND30]], !llvm.loop [[LOOP8:![0-9]+]]
+// CHECK2-IRBUILDER:       omp.inner.for.end42:
+// CHECK2-IRBUILDER-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
+// CHECK2-IRBUILDER:       omp.dispatch.inc:
+// CHECK2-IRBUILDER-NEXT:    br label [[OMP_DISPATCH_COND]]
+// CHECK2-IRBUILDER:       omp.dispatch.end:
+// CHECK2-IRBUILDER-NEXT:    [[TMP33:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
+// CHECK2-IRBUILDER-NEXT:    [[TMP34:%.*]] = icmp ne i32 [[TMP33]], 0
+// CHECK2-IRBUILDER-NEXT:    br i1 [[TMP34]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
+// CHECK2-IRBUILDER:       .omp.final.then:
+// CHECK2-IRBUILDER-NEXT:    [[TMP35:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_18]], align 4
+// CHECK2-IRBUILDER-NEXT:    [[TMP36:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_19]], align 4
+// CHECK2-IRBUILDER-NEXT:    [[TMP37:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_18]], align 4
+// CHECK2-IRBUILDER-NEXT:    [[SUB43:%.*]] = sub i32 [[TMP36]], [[TMP37]]
+// CHECK2-IRBUILDER-NEXT:    [[SUB44:%.*]] = sub i32 [[SUB43]], 1
+// CHECK2-IRBUILDER-NEXT:    [[ADD45:%.*]] = add i32 [[SUB44]], 1
+// CHECK2-IRBUILDER-NEXT:    [[DIV46:%.*]] = udiv i32 [[ADD45]], 1
+// CHECK2-IRBUILDER-NEXT:    [[MUL47:%.*]] = mul i32 [[DIV46]], 1
+// CHECK2-IRBUILDER-NEXT:    [[ADD48:%.*]] = add i32 [[TMP35]], [[MUL47]]
+// CHECK2-IRBUILDER-NEXT:    store i32 [[ADD48]], i32* [[I28]], align 4
+// CHECK2-IRBUILDER-NEXT:    br label [[DOTOMP_FINAL_DONE]]
+// CHECK2-IRBUILDER:       .omp.final.done:
+// CHECK2-IRBUILDER-NEXT:    br label [[OMP_PRECOND_END]]
+// CHECK2-IRBUILDER:       omp.precond.end:
+// CHECK2-IRBUILDER-NEXT:    [[OMP_GLOBAL_THREAD_NUM49:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
+// CHECK2-IRBUILDER-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB4]], i32 [[OMP_GLOBAL_THREAD_NUM49]])
+// CHECK2-IRBUILDER-NEXT:    ret void
+//
+//
+// CHECK2-IRBUILDER-LABEL: define {{[^@]+}}@__captured_stmt
+// CHECK2-IRBUILDER-SAME: (i32* noundef nonnull align 4 dereferenceable(4) [[I:%.*]]) #[[ATTR3:[0-9]+]] {
+// CHECK2-IRBUILDER-NEXT:  entry:
+// CHECK2-IRBUILDER-NEXT:    [[I_ADDR:%.*]] = alloca i32*, align 8
+// CHECK2-IRBUILDER-NEXT:    store i32* [[I]], i32** [[I_ADDR]], align 8
+// CHECK2-IRBUILDER-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[I_ADDR]], align 8
+// CHECK2-IRBUILDER-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
+// CHECK2-IRBUILDER-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP1]] to i64
+// CHECK2-IRBUILDER-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x float], [10 x float]* @f, i64 0, i64 [[IDXPROM]]
+// CHECK2-IRBUILDER-NEXT:    store float 1.000000e+00, float* [[ARRAYIDX]], align 4
+// CHECK2-IRBUILDER-NEXT:    ret void
+//
+//
+// CHECK2-IRBUILDER-LABEL: define {{[^@]+}}@__captured_stmt.1
+// CHECK2-IRBUILDER-SAME: (i32* noundef nonnull align 4 dereferenceable(4) [[I:%.*]]) #[[ATTR3]] {
+// CHECK2-IRBUILDER-NEXT:  entry:
+// CHECK2-IRBUILDER-NEXT:    [[I_ADDR:%.*]] = alloca i32*, align 8
+// CHECK2-IRBUILDER-NEXT:    store i32* [[I]], i32** [[I_ADDR]], align 8
+// CHECK2-IRBUILDER-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[I_ADDR]], align 8
+// CHECK2-IRBUILDER-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
+// CHECK2-IRBUILDER-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP1]] to i64
+// CHECK2-IRBUILDER-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x float], [10 x float]* @f, i64 0, i64 [[IDXPROM]]
+// CHECK2-IRBUILDER-NEXT:    store float 1.000000e+00, float* [[ARRAYIDX]], align 4
+// CHECK2-IRBUILDER-NEXT:    ret void
+//
+//
 // CHECK3-LABEL: define {{[^@]+}}@_Z18static_not_chunkedPfS_S_S_
 // CHECK3-SAME: (float* noundef [[A:%.*]], float* noundef [[B:%.*]], float* noundef [[C:%.*]], float* noundef [[D:%.*]]) #[[ATTR0:[0-9]+]] {
 // CHECK3-NEXT:  entry:
 // CHECK3-NEXT:    [[A_ADDR:%.*]] = alloca float*, align 8
 // CHECK3-NEXT:    [[B_ADDR:%.*]] = alloca float*, align 8
 // CHECK3-NEXT:    [[C_ADDR:%.*]] = alloca float*, align 8
 // CHECK3-NEXT:    [[D_ADDR:%.*]] = alloca float*, align 8
 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
-// CHECK3-NORMAL-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1:[0-9]+]])
+// CHECK3-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1:[0-9]+]])
 // CHECK3-NEXT:    store float* [[A]], float** [[A_ADDR]], align 8
 // CHECK3-NEXT:    store float* [[B]], float** [[B_ADDR]], align 8
 // CHECK3-NEXT:    store float* [[C]], float** [[C_ADDR]], align 8
 // CHECK3-NEXT:    store float* [[D]], float** [[D_ADDR]], align 8
 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
 // CHECK3-NEXT:    store i32 4571423, i32* [[DOTOMP_UB]], align 4
 // CHECK3-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
-// CHECK3-IRBUILDER-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB3:[0-9]+]])
-// CHECK3-IRBUILDER-NEXT:    call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP0]], i32 66, i32 0, i32 4571423, i32 1, i32 1)
-// CHECK3-NORMAL-NEXT:    call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 66, i32 0, i32 4571423, i32 1, i32 1)
+// CHECK3-NEXT:    call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 66, i32 0, i32 4571423, i32 1, i32 1)
 // CHECK3-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK3:       omp.dispatch.cond:
-// CHECK3-IRBUILDER-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB3]])
 // CHECK3-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
 // CHECK3-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP1]], 0
 // CHECK3-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
 // CHECK3:       omp.dispatch.body:
 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK3-NEXT:    store i32 [[TMP2]], i32* [[DOTOMP_IV]], align 4
 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK3:       omp.inner.for.cond:
 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP3]], [[TMP4]]
 // CHECK3-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK3:       omp.inner.for.body:
 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP5]], 7
 // CHECK3-NEXT:    [[SUB:%.*]] = sub nsw i32 32000000, [[MUL]]
 // CHECK3-NEXT:    store i32 [[SUB]], i32* [[I]], align 4
-// CHECK3-IRBUILDER-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
 // CHECK3-NEXT:    call void @__kmpc_ordered(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
 // CHECK3-NEXT:    [[TMP6:%.*]] = load float*, float** [[B_ADDR]], align 8
 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, i32* [[I]], align 4
 // CHECK3-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP7]] to i64
 // CHECK3-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, float* [[TMP6]], i64 [[IDXPROM]]
 // CHECK3-NEXT:    [[TMP8:%.*]] = load float, float* [[ARRAYIDX]], align 4
 // CHECK3-NEXT:    [[TMP9:%.*]] = load float*, float** [[C_ADDR]], align 8
 // CHECK3-NEXT:    [[TMP10:%.*]] = load i32, i32* [[I]], align 4

 // CHECK3-NEXT:    call void @__kmpc_end_ordered(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
 // CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK3:       omp.body.continue:
 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK3:       omp.inner.for.inc:
 // CHECK3-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP17]], 1
 // CHECK3-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
-// CHECK3-IRBUILDER-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB3]])
 // CHECK3-NEXT:    call void @__kmpc_dispatch_fini_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]]
 // CHECK3:       omp.inner.for.end:
 // CHECK3-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
 // CHECK3:       omp.dispatch.inc:
 // CHECK3-NEXT:    br label [[OMP_DISPATCH_COND]]
 // CHECK3:       omp.dispatch.end:
-// CHECK3-IRBUILDER-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
-// CHECK3-IRBUILDER-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB4:[0-9]+]], i32 [[TMP0]])
-// CHECK3-NORMAL-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP0]])
+// CHECK3-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP0]])
 // CHECK3-NEXT:    ret void
 //
 //
 // CHECK3-LABEL: define {{[^@]+}}@_Z8dynamic1PfS_S_S_
 // CHECK3-SAME: (float* noundef [[A:%.*]], float* noundef [[B:%.*]], float* noundef [[C:%.*]], float* noundef [[D:%.*]]) #[[ATTR0]] {
 // CHECK3-NEXT:  entry:
 // CHECK3-NEXT:    [[A_ADDR:%.*]] = alloca float*, align 8
 // CHECK3-NEXT:    [[B_ADDR:%.*]] = alloca float*, align 8
 // CHECK3-NEXT:    [[C_ADDR:%.*]] = alloca float*, align 8
 // CHECK3-NEXT:    [[D_ADDR:%.*]] = alloca float*, align 8
 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i64, align 8
 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i64, align 8
 // CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i64, align 8
 // CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i64, align 8
 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i64, align 8
 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[I:%.*]] = alloca i64, align 8
-// CHECK3-NORMAL-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
+// CHECK3-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
 // CHECK3-NEXT:    store float* [[A]], float** [[A_ADDR]], align 8
 // CHECK3-NEXT:    store float* [[B]], float** [[B_ADDR]], align 8
 // CHECK3-NEXT:    store float* [[C]], float** [[C_ADDR]], align 8
 // CHECK3-NEXT:    store float* [[D]], float** [[D_ADDR]], align 8
 // CHECK3-NEXT:    store i64 0, i64* [[DOTOMP_LB]], align 8
 // CHECK3-NEXT:    store i64 16908287, i64* [[DOTOMP_UB]], align 8
 // CHECK3-NEXT:    store i64 1, i64* [[DOTOMP_STRIDE]], align 8
 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
-// CHECK3-IRBUILDER-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB6:[0-9]+]])
 // CHECK3-NEXT:    call void @__kmpc_dispatch_init_8u(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 67, i64 0, i64 16908287, i64 1, i64 1)
 // CHECK3-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK3:       omp.dispatch.cond:
-// CHECK3-IRBUILDER-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB6]])
 // CHECK3-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_dispatch_next_8u(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32* [[DOTOMP_IS_LAST]], i64* [[DOTOMP_LB]], i64* [[DOTOMP_UB]], i64* [[DOTOMP_STRIDE]])
 // CHECK3-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP1]], 0
 // CHECK3-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
 // CHECK3:       omp.dispatch.body:
 // CHECK3-NEXT:    [[TMP2:%.*]] = load i64, i64* [[DOTOMP_LB]], align 8
 // CHECK3-NEXT:    store i64 [[TMP2]], i64* [[DOTOMP_IV]], align 8
 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK3:       omp.inner.for.cond:
 // CHECK3-NEXT:    [[TMP3:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
 // CHECK3-NEXT:    [[TMP4:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
 // CHECK3-NEXT:    [[ADD:%.*]] = add i64 [[TMP4]], 1
 // CHECK3-NEXT:    [[CMP:%.*]] = icmp ult i64 [[TMP3]], [[ADD]]
 // CHECK3-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK3:       omp.inner.for.body:
 // CHECK3-NEXT:    [[TMP5:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
 // CHECK3-NEXT:    [[MUL:%.*]] = mul i64 [[TMP5]], 127
 // CHECK3-NEXT:    [[ADD1:%.*]] = add i64 131071, [[MUL]]
 // CHECK3-NEXT:    store i64 [[ADD1]], i64* [[I]], align 8
-// CHECK3-IRBUILDER-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
 // CHECK3-NEXT:    call void @__kmpc_ordered(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
 // CHECK3-NEXT:    [[TMP6:%.*]] = load float*, float** [[B_ADDR]], align 8
 // CHECK3-NEXT:    [[TMP7:%.*]] = load i64, i64* [[I]], align 8
 // CHECK3-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, float* [[TMP6]], i64 [[TMP7]]
 // CHECK3-NEXT:    [[TMP8:%.*]] = load float, float* [[ARRAYIDX]], align 4
 // CHECK3-NEXT:    [[TMP9:%.*]] = load float*, float** [[C_ADDR]], align 8
 // CHECK3-NEXT:    [[TMP10:%.*]] = load i64, i64* [[I]], align 8
 // CHECK3-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds float, float* [[TMP9]], i64 [[TMP10]]

 // CHECK3-NEXT:    call void @__kmpc_end_ordered(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
 // CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK3:       omp.body.continue:
 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK3:       omp.inner.for.inc:
 // CHECK3-NEXT:    [[TMP17:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
 // CHECK3-NEXT:    [[ADD7:%.*]] = add i64 [[TMP17]], 1
 // CHECK3-NEXT:    store i64 [[ADD7]], i64* [[DOTOMP_IV]], align 8
-// CHECK3-IRBUILDER-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB6]])
 // CHECK3-NEXT:    call void @__kmpc_dispatch_fini_8u(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]]
 // CHECK3:       omp.inner.for.end:
 // CHECK3-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
 // CHECK3:       omp.dispatch.inc:
 // CHECK3-NEXT:    br label [[OMP_DISPATCH_COND]]
 // CHECK3:       omp.dispatch.end:
-// CHECK3-IRBUILDER-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
-// CHECK3-IRBUILDER-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB4]], i32 [[TMP0]])
-// CHECK3-NORMAL-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]])
+// CHECK3-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]])
 // CHECK3-NEXT:    ret void
 //
 //
 // CHECK3-LABEL: define {{[^@]+}}@_Z9test_autoPfS_S_S_
 // CHECK3-SAME: (float* noundef [[A:%.*]], float* noundef [[B:%.*]], float* noundef [[C:%.*]], float* noundef [[D:%.*]]) #[[ATTR0]] {
 // CHECK3-NEXT:  entry:
 // CHECK3-NEXT:    [[A_ADDR:%.*]] = alloca float*, align 8
 // CHECK3-NEXT:    [[B_ADDR:%.*]] = alloca float*, align 8

 // CHECK3-NEXT:    [[I:%.*]] = alloca i8, align 1
 // CHECK3-NEXT:    [[X6:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i64, align 8
 // CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i64, align 8
 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i64, align 8
 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[I8:%.*]] = alloca i8, align 1
 // CHECK3-NEXT:    [[X9:%.*]] = alloca i32, align 4
-// CHECK3-NORMAL-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
+// CHECK3-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
 // CHECK3-NEXT:    store float* [[A]], float** [[A_ADDR]], align 8
 // CHECK3-NEXT:    store float* [[B]], float** [[B_ADDR]], align 8
 // CHECK3-NEXT:    store float* [[C]], float** [[C_ADDR]], align 8
 // CHECK3-NEXT:    store float* [[D]], float** [[D_ADDR]], align 8
 // CHECK3-NEXT:    store i32 0, i32* [[X]], align 4
 // CHECK3-NEXT:    store i32 0, i32* [[Y]], align 4
 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[Y]], align 4
 // CHECK3-NEXT:    [[CONV:%.*]] = trunc i32 [[TMP1]] to i8

 // CHECK3-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
 // CHECK3:       omp.precond.then:
 // CHECK3-NEXT:    store i64 0, i64* [[DOTOMP_LB]], align 8
 // CHECK3-NEXT:    [[TMP5:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR_2]], align 8
 // CHECK3-NEXT:    store i64 [[TMP5]], i64* [[DOTOMP_UB]], align 8
 // CHECK3-NEXT:    store i64 1, i64* [[DOTOMP_STRIDE]], align 8
 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK3-NEXT:    [[TMP6:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR_2]], align 8
-// CHECK3-IRBUILDER-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB8:[0-9]+]])
 // CHECK3-NEXT:    call void @__kmpc_dispatch_init_8(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 70, i64 0, i64 [[TMP6]], i64 1, i64 1)
 // CHECK3-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK3:       omp.dispatch.cond:
-// CHECK3-IRBUILDER-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB8]])
 // CHECK3-NEXT:    [[TMP7:%.*]] = call i32 @__kmpc_dispatch_next_8(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32* [[DOTOMP_IS_LAST]], i64* [[DOTOMP_LB]], i64* [[DOTOMP_UB]], i64* [[DOTOMP_STRIDE]])
 // CHECK3-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP7]], 0
 // CHECK3-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
 // CHECK3:       omp.dispatch.body:
 // CHECK3-NEXT:    [[TMP8:%.*]] = load i64, i64* [[DOTOMP_LB]], align 8
 // CHECK3-NEXT:    store i64 [[TMP8]], i64* [[DOTOMP_IV]], align 8
 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK3:       omp.inner.for.cond:

 // CHECK3-NEXT:    [[TMP14:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
 // CHECK3-NEXT:    [[DIV16:%.*]] = sdiv i64 [[TMP14]], 11
 // CHECK3-NEXT:    [[MUL17:%.*]] = mul nsw i64 [[DIV16]], 11
 // CHECK3-NEXT:    [[SUB18:%.*]] = sub nsw i64 [[TMP13]], [[MUL17]]
 // CHECK3-NEXT:    [[MUL19:%.*]] = mul nsw i64 [[SUB18]], 1
 // CHECK3-NEXT:    [[SUB20:%.*]] = sub nsw i64 11, [[MUL19]]
 // CHECK3-NEXT:    [[CONV21:%.*]] = trunc i64 [[SUB20]] to i32
 // CHECK3-NEXT:    store i32 [[CONV21]], i32* [[X9]], align 4
-// CHECK3-IRBUILDER-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
 // CHECK3-NEXT:    call void @__kmpc_ordered(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
 // CHECK3-NEXT:    [[TMP15:%.*]] = load float*, float** [[B_ADDR]], align 8
 // CHECK3-NEXT:    [[TMP16:%.*]] = load i8, i8* [[I8]], align 1
 // CHECK3-NEXT:    [[IDXPROM:%.*]] = sext i8 [[TMP16]] to i64
 // CHECK3-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, float* [[TMP15]], i64 [[IDXPROM]]
 // CHECK3-NEXT:    [[TMP17:%.*]] = load float, float* [[ARRAYIDX]], align 4
 // CHECK3-NEXT:    [[TMP18:%.*]] = load float*, float** [[C_ADDR]], align 8
 // CHECK3-NEXT:    [[TMP19:%.*]] = load i8, i8* [[I8]], align 1

 // CHECK3-NEXT:    call void @__kmpc_end_ordered(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
 // CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK3:       omp.body.continue:
 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK3:       omp.inner.for.inc:
 // CHECK3-NEXT:    [[TMP26:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
 // CHECK3-NEXT:    [[ADD30:%.*]] = add nsw i64 [[TMP26]], 1
 // CHECK3-NEXT:    store i64 [[ADD30]], i64* [[DOTOMP_IV]], align 8
-// CHECK3-IRBUILDER-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB8]])
 // CHECK3-NEXT:    call void @__kmpc_dispatch_fini_8(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]]
 // CHECK3:       omp.inner.for.end:
 // CHECK3-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
 // CHECK3:       omp.dispatch.inc:
 // CHECK3-NEXT:    br label [[OMP_DISPATCH_COND]]
 // CHECK3:       omp.dispatch.end:
 // CHECK3-NEXT:    br label [[OMP_PRECOND_END]]
 // CHECK3:       omp.precond.end:
-// CHECK3-IRBUILDER-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
-// CHECK3-IRBUILDER-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB4]], i32 [[TMP0]])
-// CHECK3-NORMAL-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]])
+// CHECK3-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]])
 // CHECK3-NEXT:    ret void
 //
 //
 // CHECK3-LABEL: define {{[^@]+}}@_Z7runtimePfS_S_S_
 // CHECK3-SAME: (float* noundef [[A:%.*]], float* noundef [[B:%.*]], float* noundef [[C:%.*]], float* noundef [[D:%.*]]) #[[ATTR0]] {
 // CHECK3-NEXT:  entry:
 // CHECK3-NEXT:    [[A_ADDR:%.*]] = alloca float*, align 8
 // CHECK3-NEXT:    [[B_ADDR:%.*]] = alloca float*, align 8
 // CHECK3-NEXT:    [[C_ADDR:%.*]] = alloca float*, align 8
 // CHECK3-NEXT:    [[D_ADDR:%.*]] = alloca float*, align 8
 // CHECK3-NEXT:    [[X:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i8, align 1
 // CHECK3-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[I:%.*]] = alloca i8, align 1
 // CHECK3-NEXT:    [[X2:%.*]] = alloca i32, align 4
-// CHECK3-NORMAL-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
+// CHECK3-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
 // CHECK3-NEXT:    store float* [[A]], float** [[A_ADDR]], align 8
 // CHECK3-NEXT:    store float* [[B]], float** [[B_ADDR]], align 8
 // CHECK3-NEXT:    store float* [[C]], float** [[C_ADDR]], align 8
 // CHECK3-NEXT:    store float* [[D]], float** [[D_ADDR]], align 8
 // CHECK3-NEXT:    store i32 0, i32* [[X]], align 4
 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
 // CHECK3-NEXT:    store i32 199, i32* [[DOTOMP_UB]], align 4
 // CHECK3-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
-// CHECK3-IRBUILDER-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB10:[0-9]+]])
 // CHECK3-NEXT:    call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 69, i32 0, i32 199, i32 1, i32 1)
 // CHECK3-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK3:       omp.dispatch.cond:
-// CHECK3-IRBUILDER-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB10]])
 // CHECK3-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
 // CHECK3-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP1]], 0
 // CHECK3-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
 // CHECK3:       omp.dispatch.body:
 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK3-NEXT:    store i32 [[TMP2]], i32* [[DOTOMP_IV]], align 4
 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK3:       omp.inner.for.cond:

 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK3-NEXT:    [[DIV3:%.*]] = sdiv i32 [[TMP7]], 20
 // CHECK3-NEXT:    [[MUL4:%.*]] = mul nsw i32 [[DIV3]], 20
 // CHECK3-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP6]], [[MUL4]]
 // CHECK3-NEXT:    [[MUL5:%.*]] = mul nsw i32 [[SUB]], 1
 // CHECK3-NEXT:    [[ADD6:%.*]] = add nsw i32 -10, [[MUL5]]
 // CHECK3-NEXT:    store i32 [[ADD6]], i32* [[X2]], align 4
-// CHECK3-IRBUILDER-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
 // CHECK3-NEXT:    call void @__kmpc_ordered(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
 // CHECK3-NEXT:    [[TMP8:%.*]] = load float*, float** [[B_ADDR]], align 8
 // CHECK3-NEXT:    [[TMP9:%.*]] = load i8, i8* [[I]], align 1
 // CHECK3-NEXT:    [[IDXPROM:%.*]] = zext i8 [[TMP9]] to i64
 // CHECK3-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, float* [[TMP8]], i64 [[IDXPROM]]
 // CHECK3-NEXT:    [[TMP10:%.*]] = load float, float* [[ARRAYIDX]], align 4
 // CHECK3-NEXT:    [[TMP11:%.*]] = load float*, float** [[C_ADDR]], align 8
 // CHECK3-NEXT:    [[TMP12:%.*]] = load i8, i8* [[I]], align 1

 // CHECK3-NEXT:    call void @__kmpc_end_ordered(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
 // CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK3:       omp.body.continue:
 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK3:       omp.inner.for.inc:
 // CHECK3-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK3-NEXT:    [[ADD15:%.*]] = add nsw i32 [[TMP19]], 1
 // CHECK3-NEXT:    store i32 [[ADD15]], i32* [[DOTOMP_IV]], align 4
-// CHECK3-IRBUILDER-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB10]])
 // CHECK3-NEXT:    call void @__kmpc_dispatch_fini_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]]
 // CHECK3:       omp.inner.for.end:
 // CHECK3-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
 // CHECK3:       omp.dispatch.inc:
 // CHECK3-NEXT:    br label [[OMP_DISPATCH_COND]]
 // CHECK3:       omp.dispatch.end:
-// CHECK3-IRBUILDER-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
-// CHECK3-IRBUILDER-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB4]], i32 [[TMP0]])
-// CHECK3-NORMAL-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]])
+// CHECK3-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]])
 // CHECK3-NEXT:    ret void
 //
 //
 // CHECK3-LABEL: define {{[^@]+}}@_Z8foo_simdii
 // CHECK3-SAME: (i32 noundef [[LOW:%.*]], i32 noundef [[UP:%.*]]) #[[ATTR0]] {
 // CHECK3-NEXT:  entry:
 // CHECK3-NEXT:    [[LOW_ADDR:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[UP_ADDR:%.*]] = alloca i32, align 4

 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR_19:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR_20:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[I26:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[I28:%.*]] = alloca i32, align 4
-// CHECK3-NORMAL-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
+// CHECK3-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
 // CHECK3-NEXT:    store i32 [[LOW]], i32* [[LOW_ADDR]], align 4
 // CHECK3-NEXT:    store i32 [[UP]], i32* [[UP_ADDR]], align 4
 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[LOW_ADDR]], align 4
 // CHECK3-NEXT:    store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[UP_ADDR]], align 4
 // CHECK3-NEXT:    store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4

 // CHECK3-NEXT:    br i1 [[CMP27]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
 // CHECK3:       omp.precond.then:
 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
 // CHECK3-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_20]], align 4
 // CHECK3-NEXT:    store i32 [[TMP24]], i32* [[DOTOMP_UB]], align 4
 // CHECK3-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK3-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_20]], align 4
-// CHECK3-IRBUILDER-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB12:[0-9]+]])
 // CHECK3-NEXT:    call void @__kmpc_dispatch_init_4u(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 66, i32 0, i32 [[TMP25]], i32 1, i32 1)
 // CHECK3-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK3:       omp.dispatch.cond:
-// CHECK3-IRBUILDER-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB12]])
 // CHECK3-NEXT:    [[TMP26:%.*]] = call i32 @__kmpc_dispatch_next_4u(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
 // CHECK3-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP26]], 0
 // CHECK3-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
 // CHECK3:       omp.dispatch.body:
 // CHECK3-NEXT:    [[TMP27:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK3-NEXT:    store i32 [[TMP27]], i32* [[DOTOMP_IV16]], align 4
 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND29:%.*]]
-// CHECK3-IRBUILDER:       omp.inner.for.cond30:
-// CHECK3-NORMAL:       omp.inner.for.cond29:
+// CHECK3:       omp.inner.for.cond29:
 // CHECK3-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTOMP_IV16]], align 4, !llvm.access.group !7
 // CHECK3-NEXT:    [[TMP29:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !7
 // CHECK3-NEXT:    [[ADD30:%.*]] = add i32 [[TMP29]], 1
 // CHECK3-NEXT:    [[CMP31:%.*]] = icmp ult i32 [[TMP28]], [[ADD30]]
 // CHECK3-NEXT:    br i1 [[CMP31]], label [[OMP_INNER_FOR_BODY32:%.*]], label [[OMP_INNER_FOR_END40:%.*]]
-// CHECK3-IRBUILDER:       omp.inner.for.body33:
-// CHECK3-NORMAL:       omp.inner.for.body32:
+// CHECK3:       omp.inner.for.body32:
 // CHECK3-NEXT:    [[TMP30:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_18]], align 4, !llvm.access.group !7
 // CHECK3-NEXT:    [[TMP31:%.*]] = load i32, i32* [[DOTOMP_IV16]], align 4, !llvm.access.group !7
 // CHECK3-NEXT:    [[MUL33:%.*]] = mul i32 [[TMP31]], 1
 // CHECK3-NEXT:    [[ADD34:%.*]] = add i32 [[TMP30]], [[MUL33]]
 // CHECK3-NEXT:    store i32 [[ADD34]], i32* [[I28]], align 4, !llvm.access.group !7
 // CHECK3-NEXT:    [[TMP32:%.*]] = load i32, i32* [[I28]], align 4, !llvm.access.group !7
 // CHECK3-NEXT:    [[IDXPROM35:%.*]] = sext i32 [[TMP32]] to i64
 // CHECK3-NEXT:    [[ARRAYIDX36:%.*]] = getelementptr inbounds [10 x float], [10 x float]* @f, i64 0, i64 [[IDXPROM35]]
 // CHECK3-NEXT:    store float 0.000000e+00, float* [[ARRAYIDX36]], align 4, !llvm.access.group !7
 // CHECK3-NEXT:    call void @__captured_stmt.1(i32* [[I28]]), !llvm.access.group !7
 // CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE37:%.*]]
-// CHECK3-IRBUILDER:       omp.body.continue38:
-// CHECK3-NORMAL:       omp.body.continue37:
+// CHECK3:       omp.body.continue37:
 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC38:%.*]]
-// CHECK3-IRBUILDER:       omp.inner.for.inc39:
-// CHECK3-NORMAL:       omp.inner.for.inc38:
+// CHECK3:       omp.inner.for.inc38:
 // CHECK3-NEXT:    [[TMP33:%.*]] = load i32, i32* [[DOTOMP_IV16]], align 4, !llvm.access.group !7
 // CHECK3-NEXT:    [[ADD39:%.*]] = add i32 [[TMP33]], 1
 // CHECK3-NEXT:    store i32 [[ADD39]], i32* [[DOTOMP_IV16]], align 4, !llvm.access.group !7
-// CHECK3-IRBUILDER-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB12]])
 // CHECK3-NEXT:    call void @__kmpc_dispatch_fini_4u(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]]), !llvm.access.group !7
 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND29]], !llvm.loop [[LOOP8:![0-9]+]]
-// CHECK3-IRBUILDER:       omp.inner.for.end42:
-// CHECK3-NORMAL:       omp.inner.for.end40:
+// CHECK3:       omp.inner.for.end40:
 // CHECK3-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
 // CHECK3:       omp.dispatch.inc:
 // CHECK3-NEXT:    br label [[OMP_DISPATCH_COND]]
 // CHECK3:       omp.dispatch.end:
 // CHECK3-NEXT:    [[TMP34:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK3-NEXT:    [[TMP35:%.*]] = icmp ne i32 [[TMP34]], 0
 // CHECK3-NEXT:    br i1 [[TMP35]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
 // CHECK3:       .omp.final.then:
 // CHECK3-NEXT:    [[TMP36:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_18]], align 4
 // CHECK3-NEXT:    [[TMP37:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_19]], align 4
 // CHECK3-NEXT:    [[TMP38:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_18]], align 4
 // CHECK3-NEXT:    [[SUB41:%.*]] = sub i32 [[TMP37]], [[TMP38]]
 // CHECK3-NEXT:    [[SUB42:%.*]] = sub i32 [[SUB41]], 1
 // CHECK3-NEXT:    [[ADD43:%.*]] = add i32 [[SUB42]], 1
 // CHECK3-NEXT:    [[DIV44:%.*]] = udiv i32 [[ADD43]], 1
 // CHECK3-NEXT:    [[MUL45:%.*]] = mul i32 [[DIV44]], 1
 // CHECK3-NEXT:    [[ADD46:%.*]] = add i32 [[TMP36]], [[MUL45]]
 // CHECK3-NEXT:    store i32 [[ADD46]], i32* [[I28]], align 4
 // CHECK3-NEXT:    br label [[DOTOMP_FINAL_DONE]]
 // CHECK3:       .omp.final.done:
 // CHECK3-NEXT:    br label [[OMP_PRECOND_END]]
 // CHECK3:       omp.precond.end:
-// CHECK3-IRBUILDER-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
-// CHECK3-IRBUILDER-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB4]], i32 [[TMP0]])
-// CHECK3-NORMAL-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]])
+// CHECK3-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]])
 // CHECK3-NEXT:    ret void
 //
 //
 // CHECK3-LABEL: define {{[^@]+}}@__captured_stmt
 // CHECK3-SAME: (i32* noundef nonnull align 4 dereferenceable(4) [[I:%.*]]) #[[ATTR3:[0-9]+]] {
 // CHECK3-NEXT:  entry:
 // CHECK3-NEXT:    [[I_ADDR:%.*]] = alloca i32*, align 8
 // CHECK3-NEXT:    store i32* [[I]], i32** [[I_ADDR]], align 8

 // CHECK4-NEXT:    [[D_ADDR:%.*]] = alloca float*, align 8
 // CHECK4-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 // CHECK4-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 // CHECK4-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
 // CHECK4-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
 // CHECK4-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
 // CHECK4-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 // CHECK4-NEXT:    [[I:%.*]] = alloca i32, align 4
-// CHECK4-NORMAL-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1:[0-9]+]])
+// CHECK4-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1:[0-9]+]])
 // CHECK4-NEXT:    store float* [[A]], float** [[A_ADDR]], align 8
 // CHECK4-NEXT:    store float* [[B]], float** [[B_ADDR]], align 8
 // CHECK4-NEXT:    store float* [[C]], float** [[C_ADDR]], align 8
 // CHECK4-NEXT:    store float* [[D]], float** [[D_ADDR]], align 8
 // CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
 // CHECK4-NEXT:    store i32 4571423, i32* [[DOTOMP_UB]], align 4
 // CHECK4-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
 // CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
-// CHECK4-IRBUILDER-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB3:[0-9]+]])
-// CHECK4-IRBUILDER-NEXT:    call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP0]], i32 66, i32 0, i32 4571423, i32 1, i32 1)
-// CHECK4-NORMAL-NEXT:    call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 66, i32 0, i32 4571423, i32 1, i32 1)
+// CHECK4-NEXT:    call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 66, i32 0, i32 4571423, i32 1, i32 1)
 // CHECK4-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK4:       omp.dispatch.cond:
-// CHECK4-IRBUILDER-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB3]])
 // CHECK4-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
 // CHECK4-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP1]], 0
 // CHECK4-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
 // CHECK4:       omp.dispatch.body:
 // CHECK4-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK4-NEXT:    store i32 [[TMP2]], i32* [[DOTOMP_IV]], align 4
 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK4:       omp.inner.for.cond:
 // CHECK4-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK4-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK4-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP3]], [[TMP4]]
 // CHECK4-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK4:       omp.inner.for.body:
 // CHECK4-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK4-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP5]], 7
 // CHECK4-NEXT:    [[SUB:%.*]] = sub nsw i32 32000000, [[MUL]]
 // CHECK4-NEXT:    store i32 [[SUB]], i32* [[I]], align 4
-// CHECK4-IRBUILDER-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
 // CHECK4-NEXT:    call void @__kmpc_ordered(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
 // CHECK4-NEXT:    [[TMP6:%.*]] = load float*, float** [[B_ADDR]], align 8
 // CHECK4-NEXT:    [[TMP7:%.*]] = load i32, i32* [[I]], align 4
 // CHECK4-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP7]] to i64
 // CHECK4-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, float* [[TMP6]], i64 [[IDXPROM]]
 // CHECK4-NEXT:    [[TMP8:%.*]] = load float, float* [[ARRAYIDX]], align 4
 // CHECK4-NEXT:    [[TMP9:%.*]] = load float*, float** [[C_ADDR]], align 8
 // CHECK4-NEXT:    [[TMP10:%.*]] = load i32, i32* [[I]], align 4

 // CHECK4-NEXT:    call void @__kmpc_end_ordered(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
 // CHECK4-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK4:       omp.body.continue:
 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK4:       omp.inner.for.inc:
 // CHECK4-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK4-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP17]], 1
 // CHECK4-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
-// CHECK4-IRBUILDER-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB3]])
 // CHECK4-NEXT:    call void @__kmpc_dispatch_fini_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND]]
 // CHECK4:       omp.inner.for.end:
 // CHECK4-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
 // CHECK4:       omp.dispatch.inc:
 // CHECK4-NEXT:    br label [[OMP_DISPATCH_COND]]
 // CHECK4:       omp.dispatch.end:
-// CHECK4-IRBUILDER-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
-// CHECK4-IRBUILDER-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB4:[0-9]+]], i32 [[TMP0]])
-// CHECK4-NORMAL-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP0]])
+// CHECK4-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP0]])
 // CHECK4-NEXT:    ret void
 //
 //
 // CHECK4-LABEL: define {{[^@]+}}@_Z8dynamic1PfS_S_S_
 // CHECK4-SAME: (float* noundef [[A:%.*]], float* noundef [[B:%.*]], float* noundef [[C:%.*]], float* noundef [[D:%.*]]) #[[ATTR0]] {
 // CHECK4-NEXT:  entry:
 // CHECK4-NEXT:    [[A_ADDR:%.*]] = alloca float*, align 8
 // CHECK4-NEXT:    [[B_ADDR:%.*]] = alloca float*, align 8
 // CHECK4-NEXT:    [[C_ADDR:%.*]] = alloca float*, align 8
 // CHECK4-NEXT:    [[D_ADDR:%.*]] = alloca float*, align 8
 // CHECK4-NEXT:    [[DOTOMP_IV:%.*]] = alloca i64, align 8
 // CHECK4-NEXT:    [[TMP:%.*]] = alloca i64, align 8
 // CHECK4-NEXT:    [[DOTOMP_LB:%.*]] = alloca i64, align 8
 // CHECK4-NEXT:    [[DOTOMP_UB:%.*]] = alloca i64, align 8
 // CHECK4-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i64, align 8
 // CHECK4-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 // CHECK4-NEXT:    [[I:%.*]] = alloca i64, align 8
-// CHECK4-NORMAL-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
+// CHECK4-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
 // CHECK4-NEXT:    store float* [[A]], float** [[A_ADDR]], align 8
 // CHECK4-NEXT:    store float* [[B]], float** [[B_ADDR]], align 8
 // CHECK4-NEXT:    store float* [[C]], float** [[C_ADDR]], align 8
 // CHECK4-NEXT:    store float* [[D]], float** [[D_ADDR]], align 8
 // CHECK4-NEXT:    store i64 0, i64* [[DOTOMP_LB]], align 8
 // CHECK4-NEXT:    store i64 16908287, i64* [[DOTOMP_UB]], align 8
 // CHECK4-NEXT:    store i64 1, i64* [[DOTOMP_STRIDE]], align 8
 // CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
-// CHECK4-IRBUILDER-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB6:[0-9]+]])
 // CHECK4-NEXT:    call void @__kmpc_dispatch_init_8u(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 67, i64 0, i64 16908287, i64 1, i64 1)
 // CHECK4-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK4:       omp.dispatch.cond:
-// CHECK4-IRBUILDER-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB6]])
 // CHECK4-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_dispatch_next_8u(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32* [[DOTOMP_IS_LAST]], i64* [[DOTOMP_LB]], i64* [[DOTOMP_UB]], i64* [[DOTOMP_STRIDE]])
 // CHECK4-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP1]], 0
 // CHECK4-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
 // CHECK4:       omp.dispatch.body:
 // CHECK4-NEXT:    [[TMP2:%.*]] = load i64, i64* [[DOTOMP_LB]], align 8
 // CHECK4-NEXT:    store i64 [[TMP2]], i64* [[DOTOMP_IV]], align 8
 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK4:       omp.inner.for.cond:
 // CHECK4-NEXT:    [[TMP3:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
 // CHECK4-NEXT:    [[TMP4:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
 // CHECK4-NEXT:    [[ADD:%.*]] = add i64 [[TMP4]], 1
 // CHECK4-NEXT:    [[CMP:%.*]] = icmp ult i64 [[TMP3]], [[ADD]]
 // CHECK4-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK4:       omp.inner.for.body:
 // CHECK4-NEXT:    [[TMP5:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
 // CHECK4-NEXT:    [[MUL:%.*]] = mul i64 [[TMP5]], 127
 // CHECK4-NEXT:    [[ADD1:%.*]] = add i64 131071, [[MUL]]
 // CHECK4-NEXT:    store i64 [[ADD1]], i64* [[I]], align 8
-// CHECK4-IRBUILDER-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
 // CHECK4-NEXT:    call void @__kmpc_ordered(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
 // CHECK4-NEXT:    [[TMP6:%.*]] = load float*, float** [[B_ADDR]], align 8
 // CHECK4-NEXT:    [[TMP7:%.*]] = load i64, i64* [[I]], align 8
 // CHECK4-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, float* [[TMP6]], i64 [[TMP7]]
 // CHECK4-NEXT:    [[TMP8:%.*]] = load float, float* [[ARRAYIDX]], align 4
 // CHECK4-NEXT:    [[TMP9:%.*]] = load float*, float** [[C_ADDR]], align 8
 // CHECK4-NEXT:    [[TMP10:%.*]] = load i64, i64* [[I]], align 8
 // CHECK4-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds float, float* [[TMP9]], i64 [[TMP10]]

 // CHECK4-NEXT:    call void @__kmpc_end_ordered(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
 // CHECK4-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK4:       omp.body.continue:
 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK4:       omp.inner.for.inc:
 // CHECK4-NEXT:    [[TMP17:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
 // CHECK4-NEXT:    [[ADD7:%.*]] = add i64 [[TMP17]], 1
 // CHECK4-NEXT:    store i64 [[ADD7]], i64* [[DOTOMP_IV]], align 8
-// CHECK4-IRBUILDER-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB6]])
 // CHECK4-NEXT:    call void @__kmpc_dispatch_fini_8u(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND]]
 // CHECK4:       omp.inner.for.end:
 // CHECK4-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
 // CHECK4:       omp.dispatch.inc:
 // CHECK4-NEXT:    br label [[OMP_DISPATCH_COND]]
 // CHECK4:       omp.dispatch.end:
-// CHECK4-IRBUILDER-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
-// CHECK4-IRBUILDER-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB4]], i32 [[TMP0]])
-// CHECK4-NORMAL-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]])
+// CHECK4-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]])
 // CHECK4-NEXT:    ret void
 //
 //
 // CHECK4-LABEL: define {{[^@]+}}@_Z9test_autoPfS_S_S_
 // CHECK4-SAME: (float* noundef [[A:%.*]], float* noundef [[B:%.*]], float* noundef [[C:%.*]], float* noundef [[D:%.*]]) #[[ATTR0]] {
 // CHECK4-NEXT:  entry:
 // CHECK4-NEXT:    [[A_ADDR:%.*]] = alloca float*, align 8
 // CHECK4-NEXT:    [[B_ADDR:%.*]] = alloca float*, align 8

 // CHECK4-NEXT:    [[I:%.*]] = alloca i8, align 1
 // CHECK4-NEXT:    [[X6:%.*]] = alloca i32, align 4
 // CHECK4-NEXT:    [[DOTOMP_LB:%.*]] = alloca i64, align 8
 // CHECK4-NEXT:    [[DOTOMP_UB:%.*]] = alloca i64, align 8
 // CHECK4-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i64, align 8
 // CHECK4-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 // CHECK4-NEXT:    [[I8:%.*]] = alloca i8, align 1
 // CHECK4-NEXT:    [[X9:%.*]] = alloca i32, align 4
-// CHECK4-NORMAL-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
+// CHECK4-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
 // CHECK4-NEXT:    store float* [[A]], float** [[A_ADDR]], align 8
 // CHECK4-NEXT:    store float* [[B]], float** [[B_ADDR]], align 8
 // CHECK4-NEXT:    store float* [[C]], float** [[C_ADDR]], align 8
 // CHECK4-NEXT:    store float* [[D]], float** [[D_ADDR]], align 8
 // CHECK4-NEXT:    store i32 0, i32* [[X]], align 4
 // CHECK4-NEXT:    store i32 0, i32* [[Y]], align 4
 // CHECK4-NEXT:    [[TMP1:%.*]] = load i32, i32* [[Y]], align 4
 // CHECK4-NEXT:    [[CONV:%.*]] = trunc i32 [[TMP1]] to i8

 // CHECK4-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
 // CHECK4:       omp.precond.then:
 // CHECK4-NEXT:    store i64 0, i64* [[DOTOMP_LB]], align 8
 // CHECK4-NEXT:    [[TMP5:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR_2]], align 8
 // CHECK4-NEXT:    store i64 [[TMP5]], i64* [[DOTOMP_UB]], align 8
 // CHECK4-NEXT:    store i64 1, i64* [[DOTOMP_STRIDE]], align 8
 // CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK4-NEXT:    [[TMP6:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR_2]], align 8
-// CHECK4-IRBUILDER-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB8:[0-9]+]])
 // CHECK4-NEXT:    call void @__kmpc_dispatch_init_8(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 70, i64 0, i64 [[TMP6]], i64 1, i64 1)
 // CHECK4-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK4:       omp.dispatch.cond:
-// CHECK4-IRBUILDER-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB8]])
 // CHECK4-NEXT:    [[TMP7:%.*]] = call i32 @__kmpc_dispatch_next_8(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32* [[DOTOMP_IS_LAST]], i64* [[DOTOMP_LB]], i64* [[DOTOMP_UB]], i64* [[DOTOMP_STRIDE]])
 // CHECK4-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP7]], 0
 // CHECK4-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
 // CHECK4:       omp.dispatch.body:
 // CHECK4-NEXT:    [[TMP8:%.*]] = load i64, i64* [[DOTOMP_LB]], align 8
 // CHECK4-NEXT:    store i64 [[TMP8]], i64* [[DOTOMP_IV]], align 8
 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK4:       omp.inner.for.cond:

 // CHECK4-NEXT:    [[TMP14:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
 // CHECK4-NEXT:    [[DIV16:%.*]] = sdiv i64 [[TMP14]], 11
 // CHECK4-NEXT:    [[MUL17:%.*]] = mul nsw i64 [[DIV16]], 11
 // CHECK4-NEXT:    [[SUB18:%.*]] = sub nsw i64 [[TMP13]], [[MUL17]]
 // CHECK4-NEXT:    [[MUL19:%.*]] = mul nsw i64 [[SUB18]], 1
 // CHECK4-NEXT:    [[SUB20:%.*]] = sub nsw i64 11, [[MUL19]]
 // CHECK4-NEXT:    [[CONV21:%.*]] = trunc i64 [[SUB20]] to i32
 // CHECK4-NEXT:    store i32 [[CONV21]], i32* [[X9]], align 4
-// CHECK4-IRBUILDER-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
 // CHECK4-NEXT:    call void @__kmpc_ordered(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
 // CHECK4-NEXT:    [[TMP15:%.*]] = load float*, float** [[B_ADDR]], align 8
 // CHECK4-NEXT:    [[TMP16:%.*]] = load i8, i8* [[I8]], align 1
 // CHECK4-NEXT:    [[IDXPROM:%.*]] = sext i8 [[TMP16]] to i64
 // CHECK4-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, float* [[TMP15]], i64 [[IDXPROM]]
 // CHECK4-NEXT:    [[TMP17:%.*]] = load float, float* [[ARRAYIDX]], align 4
 // CHECK4-NEXT:    [[TMP18:%.*]] = load float*, float** [[C_ADDR]], align 8
 // CHECK4-NEXT:    [[TMP19:%.*]] = load i8, i8* [[I8]], align 1

 // CHECK4-NEXT:    call void @__kmpc_end_ordered(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
 // CHECK4-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK4:       omp.body.continue:
 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK4:       omp.inner.for.inc:
 // CHECK4-NEXT:    [[TMP26:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
 // CHECK4-NEXT:    [[ADD30:%.*]] = add nsw i64 [[TMP26]], 1
 // CHECK4-NEXT:    store i64 [[ADD30]], i64* [[DOTOMP_IV]], align 8
-// CHECK4-IRBUILDER-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB8]])
 // CHECK4-NEXT:    call void @__kmpc_dispatch_fini_8(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND]]
 // CHECK4:       omp.inner.for.end:
 // CHECK4-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
 // CHECK4:       omp.dispatch.inc:
 // CHECK4-NEXT:    br label [[OMP_DISPATCH_COND]]
 // CHECK4:       omp.dispatch.end:
 // CHECK4-NEXT:    br label [[OMP_PRECOND_END]]
 // CHECK4:       omp.precond.end:
-// CHECK4-IRBUILDER-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
-// CHECK4-IRBUILDER-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB4]], i32 [[TMP0]])
-// CHECK4-NORMAL-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]])
+// CHECK4-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]])
 // CHECK4-NEXT:    ret void
 //
 //
 // CHECK4-LABEL: define {{[^@]+}}@_Z7runtimePfS_S_S_
 // CHECK4-SAME: (float* noundef [[A:%.*]], float* noundef [[B:%.*]], float* noundef [[C:%.*]], float* noundef [[D:%.*]]) #[[ATTR0]] {
 // CHECK4-NEXT:  entry:
 // CHECK4-NEXT:    [[A_ADDR:%.*]] = alloca float*, align 8
 // CHECK4-NEXT:    [[B_ADDR:%.*]] = alloca float*, align 8
 // CHECK4-NEXT:    [[C_ADDR:%.*]] = alloca float*, align 8
 // CHECK4-NEXT:    [[D_ADDR:%.*]] = alloca float*, align 8
 // CHECK4-NEXT:    [[X:%.*]] = alloca i32, align 4
 // CHECK4-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 // CHECK4-NEXT:    [[TMP:%.*]] = alloca i8, align 1
 // CHECK4-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
 // CHECK4-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
 // CHECK4-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
 // CHECK4-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
 // CHECK4-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 // CHECK4-NEXT:    [[I:%.*]] = alloca i8, align 1
 // CHECK4-NEXT:    [[X2:%.*]] = alloca i32, align 4
-// CHECK4-NORMAL-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
+// CHECK4-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
 // CHECK4-NEXT:    store float* [[A]], float** [[A_ADDR]], align 8
 // CHECK4-NEXT:    store float* [[B]], float** [[B_ADDR]], align 8
 // CHECK4-NEXT:    store float* [[C]], float** [[C_ADDR]], align 8
 // CHECK4-NEXT:    store float* [[D]], float** [[D_ADDR]], align 8
 // CHECK4-NEXT:    store i32 0, i32* [[X]], align 4
 // CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
 // CHECK4-NEXT:    store i32 199, i32* [[DOTOMP_UB]], align 4
 // CHECK4-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
 // CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
-// CHECK4-IRBUILDER-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB10:[0-9]+]])
 // CHECK4-NEXT:    call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 69, i32 0, i32 199, i32 1, i32 1)
 // CHECK4-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK4:       omp.dispatch.cond:
-// CHECK4-IRBUILDER-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB10]])
 // CHECK4-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
 // CHECK4-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP1]], 0
 // CHECK4-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
 // CHECK4:       omp.dispatch.body:
 // CHECK4-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK4-NEXT:    store i32 [[TMP2]], i32* [[DOTOMP_IV]], align 4
 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK4:       omp.inner.for.cond:

 // CHECK4-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK4-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK4-NEXT:    [[DIV3:%.*]] = sdiv i32 [[TMP7]], 20
 // CHECK4-NEXT:    [[MUL4:%.*]] = mul nsw i32 [[DIV3]], 20
 // CHECK4-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP6]], [[MUL4]]
 // CHECK4-NEXT:    [[MUL5:%.*]] = mul nsw i32 [[SUB]], 1
 // CHECK4-NEXT:    [[ADD6:%.*]] = add nsw i32 -10, [[MUL5]]
 // CHECK4-NEXT:    store i32 [[ADD6]], i32* [[X2]], align 4
-// CHECK4-IRBUILDER-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
 // CHECK4-NEXT:    call void @__kmpc_ordered(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
 // CHECK4-NEXT:    [[TMP8:%.*]] = load float*, float** [[B_ADDR]], align 8
 // CHECK4-NEXT:    [[TMP9:%.*]] = load i8, i8* [[I]], align 1
 // CHECK4-NEXT:    [[IDXPROM:%.*]] = zext i8 [[TMP9]] to i64
 // CHECK4-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, float* [[TMP8]], i64 [[IDXPROM]]
 // CHECK4-NEXT:    [[TMP10:%.*]] = load float, float* [[ARRAYIDX]], align 4
 // CHECK4-NEXT:    [[TMP11:%.*]] = load float*, float** [[C_ADDR]], align 8
 // CHECK4-NEXT:    [[TMP12:%.*]] = load i8, i8* [[I]], align 1

 // CHECK4-NEXT:    call void @__kmpc_end_ordered(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
 // CHECK4-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK4:       omp.body.continue:
 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK4:       omp.inner.for.inc:
 // CHECK4-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
 // CHECK4-NEXT:    [[ADD15:%.*]] = add nsw i32 [[TMP19]], 1
 // CHECK4-NEXT:    store i32 [[ADD15]], i32* [[DOTOMP_IV]], align 4
-// CHECK4-IRBUILDER-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB10]])
 // CHECK4-NEXT:    call void @__kmpc_dispatch_fini_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND]]
 // CHECK4:       omp.inner.for.end:
 // CHECK4-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
 // CHECK4:       omp.dispatch.inc:
 // CHECK4-NEXT:    br label [[OMP_DISPATCH_COND]]
 // CHECK4:       omp.dispatch.end:
-// CHECK4-IRBUILDER-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
-// CHECK4-IRBUILDER-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB4]], i32 [[TMP0]])
-// CHECK4-NORMAL-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]])
+// CHECK4-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]])
 // CHECK4-NEXT:    ret void
 //
 //
 // CHECK4-LABEL: define {{[^@]+}}@_Z8foo_simdii
 // CHECK4-SAME: (i32 noundef [[LOW:%.*]], i32 noundef [[UP:%.*]]) #[[ATTR0]] {
 // CHECK4-NEXT:  entry:
 // CHECK4-NEXT:    [[LOW_ADDR:%.*]] = alloca i32, align 4
 // CHECK4-NEXT:    [[UP_ADDR:%.*]] = alloca i32, align 4

 // CHECK4-NEXT:    [[DOTCAPTURE_EXPR_19:%.*]] = alloca i32, align 4
 // CHECK4-NEXT:    [[DOTCAPTURE_EXPR_20:%.*]] = alloca i32, align 4
 // CHECK4-NEXT:    [[I26:%.*]] = alloca i32, align 4
 // CHECK4-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
 // CHECK4-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
 // CHECK4-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
 // CHECK4-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 // CHECK4-NEXT:    [[I28:%.*]] = alloca i32, align 4
-// CHECK4-NORMAL-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
+// CHECK4-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
 // CHECK4-NEXT:    store i32 [[LOW]], i32* [[LOW_ADDR]], align 4
 // CHECK4-NEXT:    store i32 [[UP]], i32* [[UP_ADDR]], align 4
 // CHECK4-NEXT:    [[TMP1:%.*]] = load i32, i32* [[LOW_ADDR]], align 4
 // CHECK4-NEXT:    store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
 // CHECK4-NEXT:    [[TMP2:%.*]] = load i32, i32* [[UP_ADDR]], align 4
 // CHECK4-NEXT:    store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
 // CHECK4-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
 // CHECK4-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4

 // CHECK4-NEXT:    br i1 [[CMP27]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
 // CHECK4:       omp.precond.then:
 // CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
 // CHECK4-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_20]], align 4
 // CHECK4-NEXT:    store i32 [[TMP24]], i32* [[DOTOMP_UB]], align 4
 // CHECK4-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
 // CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK4-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_20]], align 4
-// CHECK4-IRBUILDER-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB12:[0-9]+]])
 // CHECK4-NEXT:    call void @__kmpc_dispatch_init_4u(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 66, i32 0, i32 [[TMP25]], i32 1, i32 1)
 // CHECK4-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 // CHECK4:       omp.dispatch.cond:
-// CHECK4-IRBUILDER-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB12]])
 // CHECK4-NEXT:    [[TMP26:%.*]] = call i32 @__kmpc_dispatch_next_4u(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
 // CHECK4-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP26]], 0
 // CHECK4-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
 // CHECK4:       omp.dispatch.body:
 // CHECK4-NEXT:    [[TMP27:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
 // CHECK4-NEXT:    store i32 [[TMP27]], i32* [[DOTOMP_IV16]], align 4
 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND29:%.*]]
-// CHECK4-IRBUILDER:       omp.inner.for.cond30:
-// CHECK4-NORMAL:       omp.inner.for.cond29:
+// CHECK4:       omp.inner.for.cond29:
 // CHECK4-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTOMP_IV16]], align 4, !llvm.access.group !7
 // CHECK4-NEXT:    [[TMP29:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !7
 // CHECK4-NEXT:    [[ADD30:%.*]] = add i32 [[TMP29]], 1
 // CHECK4-NEXT:    [[CMP31:%.*]] = icmp ult i32 [[TMP28]], [[ADD30]]
 // CHECK4-NEXT:    br i1 [[CMP31]], label [[OMP_INNER_FOR_BODY32:%.*]], label [[OMP_INNER_FOR_END40:%.*]]
-// CHECK4-IRBUILDER:       omp.inner.for.body33:
-// CHECK4-NORMAL:       omp.inner.for.body32:
+// CHECK4:       omp.inner.for.body32:
 // CHECK4-NEXT:    [[TMP30:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_18]], align 4, !llvm.access.group !7
 // CHECK4-NEXT:    [[TMP31:%.*]] = load i32, i32* [[DOTOMP_IV16]], align 4, !llvm.access.group !7
 // CHECK4-NEXT:    [[MUL33:%.*]] = mul i32 [[TMP31]], 1
 // CHECK4-NEXT:    [[ADD34:%.*]] = add i32 [[TMP30]], [[MUL33]]
 // CHECK4-NEXT:    store i32 [[ADD34]], i32* [[I28]], align 4, !llvm.access.group !7
 // CHECK4-NEXT:    [[TMP32:%.*]] = load i32, i32* [[I28]], align 4, !llvm.access.group !7
 // CHECK4-NEXT:    [[IDXPROM35:%.*]] = sext i32 [[TMP32]] to i64
 // CHECK4-NEXT:    [[ARRAYIDX36:%.*]] = getelementptr inbounds [10 x float], [10 x float]* @f, i64 0, i64 [[IDXPROM35]]
 // CHECK4-NEXT:    store float 0.000000e+00, float* [[ARRAYIDX36]], align 4, !llvm.access.group !7
 // CHECK4-NEXT:    call void @__captured_stmt.1(i32* [[I28]]), !llvm.access.group !7
 // CHECK4-NEXT:    br label [[OMP_BODY_CONTINUE37:%.*]]
-// CHECK4-IRBUILDER:       omp.body.continue38:
-// CHECK4-NORMAL:       omp.body.continue37:
+// CHECK4:       omp.body.continue37:
 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_INC38:%.*]]
-// CHECK4-IRBUILDER:       omp.inner.for.inc39:
-// CHECK4-NORMAL:       omp.inner.for.inc38:
+// CHECK4:       omp.inner.for.inc38:
 // CHECK4-NEXT:    [[TMP33:%.*]] = load i32, i32* [[DOTOMP_IV16]], align 4, !llvm.access.group !7
 // CHECK4-NEXT:    [[ADD39:%.*]] = add i32 [[TMP33]], 1
 // CHECK4-NEXT:    store i32 [[ADD39]], i32* [[DOTOMP_IV16]], align 4, !llvm.access.group !7
-// CHECK4-IRBUILDER-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB12]])
 // CHECK4-NEXT:    call void @__kmpc_dispatch_fini_4u(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]]), !llvm.access.group !7
 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND29]], !llvm.loop [[LOOP8:![0-9]+]]
-// CHECK4-IRBUILDER:       omp.inner.for.end42:
-// CHECK4-NORMAL:       omp.inner.for.end40:
+// CHECK4:       omp.inner.for.end40:
 // CHECK4-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
 // CHECK4:       omp.dispatch.inc:
 // CHECK4-NEXT:    br label [[OMP_DISPATCH_COND]]
 // CHECK4:       omp.dispatch.end:
 // CHECK4-NEXT:    [[TMP34:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK4-NEXT:    [[TMP35:%.*]] = icmp ne i32 [[TMP34]], 0
 // CHECK4-NEXT:    br i1 [[TMP35]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
 // CHECK4:       .omp.final.then:
 // CHECK4-NEXT:    [[TMP36:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_18]], align 4
 // CHECK4-NEXT:    [[TMP37:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_19]], align 4
 // CHECK4-NEXT:    [[TMP38:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_18]], align 4
 // CHECK4-NEXT:    [[SUB41:%.*]] = sub i32 [[TMP37]], [[TMP38]]
 // CHECK4-NEXT:    [[SUB42:%.*]] = sub i32 [[SUB41]], 1
 // CHECK4-NEXT:    [[ADD43:%.*]] = add i32 [[SUB42]], 1
 // CHECK4-NEXT:    [[DIV44:%.*]] = udiv i32 [[ADD43]], 1
 // CHECK4-NEXT:    [[MUL45:%.*]] = mul i32 [[DIV44]], 1
 // CHECK4-NEXT:    [[ADD46:%.*]] = add i32 [[TMP36]], [[MUL45]]
 // CHECK4-NEXT:    store i32 [[ADD46]], i32* [[I28]], align 4
 // CHECK4-NEXT:    br label [[DOTOMP_FINAL_DONE]]
 // CHECK4:       .omp.final.done:
 // CHECK4-NEXT:    br label [[OMP_PRECOND_END]]
 // CHECK4:       omp.precond.end:
-// CHECK4-IRBUILDER-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
-// CHECK4-IRBUILDER-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB4]], i32 [[TMP0]])
-// CHECK4-NORMAL-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]])
+// CHECK4-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]])
 // CHECK4-NEXT:    ret void
 //
 //
 // CHECK4-LABEL: define {{[^@]+}}@__captured_stmt
 // CHECK4-SAME: (i32* noundef nonnull align 4 dereferenceable(4) [[I:%.*]]) #[[ATTR3:[0-9]+]] {
 // CHECK4-NEXT:  entry:
 // CHECK4-NEXT:    [[I_ADDR:%.*]] = alloca i32*, align 8
 // CHECK4-NEXT:    store i32* [[I]], i32** [[I_ADDR]], align 8

 // CHECK4-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[I_ADDR]], align 8
 // CHECK4-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
 // CHECK4-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP1]] to i64
 // CHECK4-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x float], [10 x float]* @f, i64 0, i64 [[IDXPROM]]
 // CHECK4-NEXT:    store float 1.000000e+00, float* [[ARRAYIDX]], align 4
 // CHECK4-NEXT:    ret void
 //
 //
+// CHECK3-IRBUILDER-LABEL: define {{[^@]+}}@_Z18static_not_chunkedPfS_S_S_
+// CHECK3-IRBUILDER-SAME: (float* noundef [[A:%.*]], float* noundef [[B:%.*]], float* noundef [[C:%.*]], float* noundef [[D:%.*]]) #[[ATTR0:[0-9]+]] {
+// CHECK3-IRBUILDER-NEXT:  entry:
+// CHECK3-IRBUILDER-NEXT:    [[A_ADDR:%.*]] = alloca float*, align 8
+// CHECK3-IRBUILDER-NEXT:    [[B_ADDR:%.*]] = alloca float*, align 8
+// CHECK3-IRBUILDER-NEXT:    [[C_ADDR:%.*]] = alloca float*, align 8
+// CHECK3-IRBUILDER-NEXT:    [[D_ADDR:%.*]] = alloca float*, align 8
+// CHECK3-IRBUILDER-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
+// CHECK3-IRBUILDER-NEXT:    [[TMP:%.*]] = alloca i32, align 4
+// CHECK3-IRBUILDER-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
+// CHECK3-IRBUILDER-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
+// CHECK3-IRBUILDER-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
+// CHECK3-IRBUILDER-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
+// CHECK3-IRBUILDER-NEXT:    [[I:%.*]] = alloca i32, align 4
+// CHECK3-IRBUILDER-NEXT:    store float* [[A]], float** [[A_ADDR]], align 8
+// CHECK3-IRBUILDER-NEXT:    store float* [[B]], float** [[B_ADDR]], align 8
+// CHECK3-IRBUILDER-NEXT:    store float* [[C]], float** [[C_ADDR]], align 8
+// CHECK3-IRBUILDER-NEXT:    store float* [[D]], float** [[D_ADDR]], align 8
+// CHECK3-IRBUILDER-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
+// CHECK3-IRBUILDER-NEXT:    store i32 4571423, i32* [[DOTOMP_UB]], align 4
+// CHECK3-IRBUILDER-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
+// CHECK3-IRBUILDER-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
+// CHECK3-IRBUILDER-NEXT:    [[OMP_GLOBAL_THREAD_NUM:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB3:[0-9]+]])
+// CHECK3-IRBUILDER-NEXT:    call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[OMP_GLOBAL_THREAD_NUM]], i32 66, i32 0, i32 4571423, i32 1, i32 1)
+// CHECK3-IRBUILDER-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
+// CHECK3-IRBUILDER:       omp.dispatch.cond:
+// CHECK3-IRBUILDER-NEXT:    [[OMP_GLOBAL_THREAD_NUM1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB3]])
+// CHECK3-IRBUILDER-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM1]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
+// CHECK3-IRBUILDER-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP0]], 0
+// CHECK3-IRBUILDER-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
+// CHECK3-IRBUILDER:       omp.dispatch.body:
+// CHECK3-IRBUILDER-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
+// CHECK3-IRBUILDER-NEXT:    store i32 [[TMP1]], i32* [[DOTOMP_IV]], align 4
+// CHECK3-IRBUILDER-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
+// CHECK3-IRBUILDER:       omp.inner.for.cond:
+// CHECK3-IRBUILDER-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
+// CHECK3-IRBUILDER-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
+// CHECK3-IRBUILDER-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP2]], [[TMP3]]
+// CHECK3-IRBUILDER-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// CHECK3-IRBUILDER:       omp.inner.for.body:
+// CHECK3-IRBUILDER-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
+// CHECK3-IRBUILDER-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP4]], 7
+// CHECK3-IRBUILDER-NEXT:    [[SUB:%.*]] = sub nsw i32 32000000, [[MUL]]
+// CHECK3-IRBUILDER-NEXT:    store i32 [[SUB]], i32* [[I]], align 4
+// CHECK3-IRBUILDER-NEXT:    [[OMP_GLOBAL_THREAD_NUM2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
+// CHECK3-IRBUILDER-NEXT:    call void @__kmpc_ordered(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM2]])
+// CHECK3-IRBUILDER-NEXT:    [[TMP5:%.*]] = load float*, float** [[B_ADDR]], align 8
+// CHECK3-IRBUILDER-NEXT:    [[TMP6:%.*]] = load i32, i32* [[I]], align 4
+// CHECK3-IRBUILDER-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP6]] to i64
+// CHECK3-IRBUILDER-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, float* [[TMP5]], i64 [[IDXPROM]]
+// CHECK3-IRBUILDER-NEXT:    [[TMP7:%.*]] = load float, float* [[ARRAYIDX]], align 4
+// CHECK3-IRBUILDER-NEXT:    [[TMP8:%.*]] = load float*, float** [[C_ADDR]], align 8
+// CHECK3-IRBUILDER-NEXT:    [[TMP9:%.*]] = load i32, i32* [[I]], align 4
+// CHECK3-IRBUILDER-NEXT:    [[IDXPROM3:%.*]] = sext i32 [[TMP9]] to i64
+// CHECK3-IRBUILDER-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds float, float* [[TMP8]], i64 [[IDXPROM3]]
+// CHECK3-IRBUILDER-NEXT:    [[TMP10:%.*]] = load float, float* [[ARRAYIDX4]], align 4
+// CHECK3-IRBUILDER-NEXT:    [[MUL5:%.*]] = fmul float [[TMP7]], [[TMP10]]
+// CHECK3-IRBUILDER-NEXT:    [[TMP11:%.*]] = load float*, float** [[D_ADDR]], align 8
+// CHECK3-IRBUILDER-NEXT:    [[TMP12:%.*]] = load i32, i32* [[I]], align 4
+// CHECK3-IRBUILDER-NEXT:    [[IDXPROM6:%.*]] = sext i32 [[TMP12]] to i64
+// CHECK3-IRBUILDER-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds float, float* [[TMP11]], i64 [[IDXPROM6]]
+// CHECK3-IRBUILDER-NEXT:    [[TMP13:%.*]] = load float, float* [[ARRAYIDX7]], align 4
+// CHECK3-IRBUILDER-NEXT:    [[MUL8:%.*]] = fmul float [[MUL5]], [[TMP13]]
+// CHECK3-IRBUILDER-NEXT:    [[TMP14:%.*]] = load float*, float** [[A_ADDR]], align 8
+// CHECK3-IRBUILDER-NEXT:    [[TMP15:%.*]] = load i32, i32* [[I]], align 4
+// CHECK3-IRBUILDER-NEXT:    [[IDXPROM9:%.*]] = sext i32 [[TMP15]] to i64
+// CHECK3-IRBUILDER-NEXT:    [[ARRAYIDX10:%.*]] = getelementptr inbounds float, float* [[TMP14]], i64 [[IDXPROM9]]
+// CHECK3-IRBUILDER-NEXT:    store float [[MUL8]], float* [[ARRAYIDX10]], align 4
+// CHECK3-IRBUILDER-NEXT:    br label [[OMP_INNER_FOR_BODY_ORDERED_AFTER:%.*]]
+// CHECK3-IRBUILDER:       omp.inner.for.body.ordered.after:
+// CHECK3-IRBUILDER-NEXT:    call void @__kmpc_end_ordered(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM2]])
+// CHECK3-IRBUILDER-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
+// CHECK3-IRBUILDER:       omp.body.continue:
+// CHECK3-IRBUILDER-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
+// CHECK3-IRBUILDER:       omp.inner.for.inc:
+// CHECK3-IRBUILDER-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
+// CHECK3-IRBUILDER-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP16]], 1
+// CHECK3-IRBUILDER-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
+// CHECK3-IRBUILDER-NEXT:    [[OMP_GLOBAL_THREAD_NUM11:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB3]])
+// CHECK3-IRBUILDER-NEXT:    call void @__kmpc_dispatch_fini_4(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM11]])
+// CHECK3-IRBUILDER-NEXT:    br label [[OMP_INNER_FOR_COND]]
+// CHECK3-IRBUILDER:       omp.inner.for.end:
+// CHECK3-IRBUILDER-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
+// CHECK3-IRBUILDER:       omp.dispatch.inc:
+// CHECK3-IRBUILDER-NEXT:    br label [[OMP_DISPATCH_COND]]
+// CHECK3-IRBUILDER:       omp.dispatch.end:
+// CHECK3-IRBUILDER-NEXT:    [[OMP_GLOBAL_THREAD_NUM12:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
+// CHECK3-IRBUILDER-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB4:[0-9]+]], i32 [[OMP_GLOBAL_THREAD_NUM12]])
+// CHECK3-IRBUILDER-NEXT:    ret void
+//
+//
+// CHECK3-IRBUILDER-LABEL: define {{[^@]+}}@_Z8dynamic1PfS_S_S_
+// CHECK3-IRBUILDER-SAME: (float* noundef [[A:%.*]], float* noundef [[B:%.*]], float* noundef [[C:%.*]], float* noundef [[D:%.*]]) #[[ATTR0]] {
+// CHECK3-IRBUILDER-NEXT:  entry:
+// CHECK3-IRBUILDER-NEXT:    [[A_ADDR:%.*]] = alloca float*, align 8
+// CHECK3-IRBUILDER-NEXT:    [[B_ADDR:%.*]] = alloca float*, align 8
+// CHECK3-IRBUILDER-NEXT:    [[C_ADDR:%.*]] = alloca float*, align 8
+// CHECK3-IRBUILDER-NEXT:    [[D_ADDR:%.*]] = alloca float*, align 8
+// CHECK3-IRBUILDER-NEXT:    [[DOTOMP_IV:%.*]] = alloca i64, align 8
+// CHECK3-IRBUILDER-NEXT:    [[TMP:%.*]] = alloca i64, align 8
+// CHECK3-IRBUILDER-NEXT:    [[DOTOMP_LB:%.*]] = alloca i64, align 8
+// CHECK3-IRBUILDER-NEXT:    [[DOTOMP_UB:%.*]] = alloca i64, align 8
+// CHECK3-IRBUILDER-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i64, align 8
+// CHECK3-IRBUILDER-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
+// CHECK3-IRBUILDER-NEXT:    [[I:%.*]] = alloca i64, align 8
+// CHECK3-IRBUILDER-NEXT:    store float* [[A]], float** [[A_ADDR]], align 8
+// CHECK3-IRBUILDER-NEXT:    store float* [[B]], float** [[B_ADDR]], align 8
+// CHECK3-IRBUILDER-NEXT:    store float* [[C]], float** [[C_ADDR]], align 8
+// CHECK3-IRBUILDER-NEXT:    store float* [[D]], float** [[D_ADDR]], align 8
+// CHECK3-IRBUILDER-NEXT:    store i64 0, i64* [[DOTOMP_LB]], align 8
+// CHECK3-IRBUILDER-NEXT:    store i64 16908287, i64* [[DOTOMP_UB]], align 8
+// CHECK3-IRBUILDER-NEXT:    store i64 1, i64* [[DOTOMP_STRIDE]], align 8
+// CHECK3-IRBUILDER-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
+// CHECK3-IRBUILDER-NEXT:    [[OMP_GLOBAL_THREAD_NUM:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB6:[0-9]+]])
+// CHECK3-IRBUILDER-NEXT:    call void @__kmpc_dispatch_init_8u(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM]], i32 67, i64 0, i64 16908287, i64 1, i64 1)
+// CHECK3-IRBUILDER-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
+// CHECK3-IRBUILDER:       omp.dispatch.cond:
+// CHECK3-IRBUILDER-NEXT:    [[OMP_GLOBAL_THREAD_NUM1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB6]])
+// CHECK3-IRBUILDER-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_dispatch_next_8u(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM1]], i32* [[DOTOMP_IS_LAST]], i64* [[DOTOMP_LB]], i64* [[DOTOMP_UB]], i64* [[DOTOMP_STRIDE]])
+// CHECK3-IRBUILDER-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP0]], 0
+// CHECK3-IRBUILDER-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
+// CHECK3-IRBUILDER:       omp.dispatch.body:
+// CHECK3-IRBUILDER-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTOMP_LB]], align 8
+// CHECK3-IRBUILDER-NEXT:    store i64 [[TMP1]], i64* [[DOTOMP_IV]], align 8
+// CHECK3-IRBUILDER-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
+// CHECK3-IRBUILDER:       omp.inner.for.cond:
+// CHECK3-IRBUILDER-NEXT:    [[TMP2:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
+// CHECK3-IRBUILDER-NEXT:    [[TMP3:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
+// CHECK3-IRBUILDER-NEXT:    [[ADD:%.*]] = add i64 [[TMP3]], 1
+// CHECK3-IRBUILDER-NEXT:    [[CMP:%.*]] = icmp ult i64 [[TMP2]], [[ADD]]
+// CHECK3-IRBUILDER-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// CHECK3-IRBUILDER:       omp.inner.for.body:
+// CHECK3-IRBUILDER-NEXT:    [[TMP4:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
+// CHECK3-IRBUILDER-NEXT:    [[MUL:%.*]] = mul i64 [[TMP4]], 127
+// CHECK3-IRBUILDER-NEXT:    [[ADD2:%.*]] = add i64 131071, [[MUL]]
+// CHECK3-IRBUILDER-NEXT:    store i64 [[ADD2]], i64* [[I]], align 8
+// CHECK3-IRBUILDER-NEXT:    [[OMP_GLOBAL_THREAD_NUM3:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
+// CHECK3-IRBUILDER-NEXT:    call void @__kmpc_ordered(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM3]])
+// CHECK3-IRBUILDER-NEXT:    [[TMP5:%.*]] = load float*, float** [[B_ADDR]], align 8
+// CHECK3-IRBUILDER-NEXT:    [[TMP6:%.*]] = load i64, i64* [[I]], align 8
+// CHECK3-IRBUILDER-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, float* [[TMP5]], i64 [[TMP6]]
+// CHECK3-IRBUILDER-NEXT:    [[TMP7:%.*]] = load float, float* [[ARRAYIDX]], align 4
+// CHECK3-IRBUILDER-NEXT:    [[TMP8:%.*]] = load float*, float** [[C_ADDR]], align 8
+// CHECK3-IRBUILDER-NEXT:    [[TMP9:%.*]] = load i64, i64* [[I]], align 8
+// CHECK3-IRBUILDER-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds float, float* [[TMP8]], i64 [[TMP9]]
+// CHECK3-IRBUILDER-NEXT:    [[TMP10:%.*]] = load float, float* [[ARRAYIDX4]], align 4
+// CHECK3-IRBUILDER-NEXT:    [[MUL5:%.*]] = fmul float [[TMP7]], [[TMP10]]
+// CHECK3-IRBUILDER-NEXT:    [[TMP11:%.*]] = load float*, float** [[D_ADDR]], align 8
+// CHECK3-IRBUILDER-NEXT:    [[TMP12:%.*]] = load i64, i64* [[I]], align 8
+// CHECK3-IRBUILDER-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds float, float* [[TMP11]], i64 [[TMP12]]
+// CHECK3-IRBUILDER-NEXT:    [[TMP13:%.*]] = load float, float* [[ARRAYIDX6]], align 4
+// CHECK3-IRBUILDER-NEXT:    [[MUL7:%.*]] = fmul float [[MUL5]], [[TMP13]]
+// CHECK3-IRBUILDER-NEXT:    [[TMP14:%.*]] = load float*, float** [[A_ADDR]], align 8
+// CHECK3-IRBUILDER-NEXT:    [[TMP15:%.*]] = load i64, i64* [[I]], align 8
+// CHECK3-IRBUILDER-NEXT:    [[ARRAYIDX8:%.*]] = getelementptr inbounds float, float* [[TMP14]], i64 [[TMP15]]
+// CHECK3-IRBUILDER-NEXT:    store float [[MUL7]], float* [[ARRAYIDX8]], align 4
+// CHECK3-IRBUILDER-NEXT:    br label [[OMP_INNER_FOR_BODY_ORDERED_AFTER:%.*]]
+// CHECK3-IRBUILDER:       omp.inner.for.body.ordered.after:
+// CHECK3-IRBUILDER-NEXT:    call void @__kmpc_end_ordered(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM3]])
+// CHECK3-IRBUILDER-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
+// CHECK3-IRBUILDER:       omp.body.continue:
+// CHECK3-IRBUILDER-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
+// CHECK3-IRBUILDER:       omp.inner.for.inc:
+// CHECK3-IRBUILDER-NEXT:    [[TMP16:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
+// CHECK3-IRBUILDER-NEXT:    [[ADD9:%.*]] = add i64 [[TMP16]], 1
+// CHECK3-IRBUILDER-NEXT:    store i64 [[ADD9]], i64* [[DOTOMP_IV]], align 8
+// CHECK3-IRBUILDER-NEXT:    [[OMP_GLOBAL_THREAD_NUM10:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB6]])
+// CHECK3-IRBUILDER-NEXT:    call void @__kmpc_dispatch_fini_8u(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM10]])
+// CHECK3-IRBUILDER-NEXT:    br label [[OMP_INNER_FOR_COND]]
+// CHECK3-IRBUILDER:       omp.inner.for.end:
+// CHECK3-IRBUILDER-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
+// CHECK3-IRBUILDER:       omp.dispatch.inc:
+// CHECK3-IRBUILDER-NEXT:    br label [[OMP_DISPATCH_COND]]
+// CHECK3-IRBUILDER:       omp.dispatch.end:
+// CHECK3-IRBUILDER-NEXT:    [[OMP_GLOBAL_THREAD_NUM11:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
+// CHECK3-IRBUILDER-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB4]], i32 [[OMP_GLOBAL_THREAD_NUM11]])
+// CHECK3-IRBUILDER-NEXT:    ret void
+//
+//
+// CHECK3-IRBUILDER-LABEL: define {{[^@]+}}@_Z9test_autoPfS_S_S_
+// CHECK3-IRBUILDER-SAME: (float* noundef [[A:%.*]], float* noundef [[B:%.*]], float* noundef [[C:%.*]], float* noundef [[D:%.*]]) #[[ATTR0]] {
+// CHECK3-IRBUILDER-NEXT:  entry:
+// CHECK3-IRBUILDER-NEXT:    [[A_ADDR:%.*]] = alloca float*, align 8
+// CHECK3-IRBUILDER-NEXT:    [[B_ADDR:%.*]] = alloca float*, align 8
+// CHECK3-IRBUILDER-NEXT:    [[C_ADDR:%.*]] = alloca float*, align 8
+// CHECK3-IRBUILDER-NEXT:    [[D_ADDR:%.*]] = alloca float*, align 8
+// CHECK3-IRBUILDER-NEXT:    [[X:%.*]] = alloca i32, align 4
+// CHECK3-IRBUILDER-NEXT:    [[Y:%.*]] = alloca i32, align 4
+// CHECK3-IRBUILDER-NEXT:    [[DOTOMP_IV:%.*]] = alloca i64, align 8
+// CHECK3-IRBUILDER-NEXT:    [[TMP:%.*]] = alloca i8, align 1
+// CHECK3-IRBUILDER-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
+// CHECK3-IRBUILDER-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i8, align 1
+// CHECK3-IRBUILDER-NEXT:    [[DOTCAPTURE_EXPR_2:%.*]] = alloca i64, align 8
+// CHECK3-IRBUILDER-NEXT:    [[I:%.*]] = alloca i8, align 1
+// CHECK3-IRBUILDER-NEXT:    [[X6:%.*]] = alloca i32, align 4
+// CHECK3-IRBUILDER-NEXT:    [[DOTOMP_LB:%.*]] = alloca i64, align 8
+// CHECK3-IRBUILDER-NEXT:    [[DOTOMP_UB:%.*]] = alloca i64, align 8
+// CHECK3-IRBUILDER-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i64, align 8
+// CHECK3-IRBUILDER-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
+// CHECK3-IRBUILDER-NEXT:    [[I8:%.*]] = alloca i8, align 1
+// CHECK3-IRBUILDER-NEXT:    [[X9:%.*]] = alloca i32, align 4
+// CHECK3-IRBUILDER-NEXT:    store float* [[A]], float** [[A_ADDR]], align 8
+// CHECK3-IRBUILDER-NEXT:    store float* [[B]], float** [[B_ADDR]], align 8
+// CHECK3-IRBUILDER-NEXT:    store float* [[C]], float** [[C_ADDR]], align 8
+// CHECK3-IRBUILDER-NEXT:    store float* [[D]], float** [[D_ADDR]], align 8
+// CHECK3-IRBUILDER-NEXT:    store i32 0, i32* [[X]], align 4
+// CHECK3-IRBUILDER-NEXT:    store i32 0, i32* [[Y]], align 4
+// CHECK3-IRBUILDER-NEXT:    [[TMP0:%.*]] = load i32, i32* [[Y]], align 4
+// CHECK3-IRBUILDER-NEXT:    [[CONV:%.*]] = trunc i32 [[TMP0]] to i8
+// CHECK3-IRBUILDER-NEXT:    store i8 [[CONV]], i8* [[DOTCAPTURE_EXPR_]], align 1
+// CHECK3-IRBUILDER-NEXT:    [[TMP1:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1
+// CHECK3-IRBUILDER-NEXT:    [[CONV3:%.*]] = sext i8 [[TMP1]] to i32
+// CHECK3-IRBUILDER-NEXT:    [[SUB:%.*]] = sub i32 57, [[CONV3]]
+// CHECK3-IRBUILDER-NEXT:    [[ADD:%.*]] = add i32 [[SUB]], 1
+// CHECK3-IRBUILDER-NEXT:    [[DIV:%.*]] = udiv i32 [[ADD]], 1
+// CHECK3-IRBUILDER-NEXT:    [[CONV4:%.*]] = zext i32 [[DIV]] to i64
+// CHECK3-IRBUILDER-NEXT:    [[MUL:%.*]] = mul nsw i64 [[CONV4]], 11
+// CHECK3-IRBUILDER-NEXT:    [[SUB5:%.*]] = sub nsw i64 [[MUL]], 1
+// CHECK3-IRBUILDER-NEXT:    store i64 [[SUB5]], i64* [[DOTCAPTURE_EXPR_2]], align 8
+// CHECK3-IRBUILDER-NEXT:    [[TMP2:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1
+// CHECK3-IRBUILDER-NEXT:    store i8 [[TMP2]], i8* [[I]], align 1
+// CHECK3-IRBUILDER-NEXT:    store i32 11, i32* [[X6]], align 4
+// CHECK3-IRBUILDER-NEXT:    [[TMP3:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1
+// CHECK3-IRBUILDER-NEXT:    [[CONV7:%.*]] = sext i8 [[TMP3]] to i32
+// CHECK3-IRBUILDER-NEXT:    [[CMP:%.*]] = icmp sle i32 [[CONV7]], 57
+// CHECK3-IRBUILDER-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
+// CHECK3-IRBUILDER:       omp.precond.then:
+// CHECK3-IRBUILDER-NEXT:    store i64 0, i64* [[DOTOMP_LB]], align 8
+// CHECK3-IRBUILDER-NEXT:    [[TMP4:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR_2]], align 8
+// CHECK3-IRBUILDER-NEXT:    store i64 [[TMP4]], i64* [[DOTOMP_UB]], align 8
+// CHECK3-IRBUILDER-NEXT:    store i64 1, i64* [[DOTOMP_STRIDE]], align 8
+// CHECK3-IRBUILDER-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
+// CHECK3-IRBUILDER-NEXT:    [[TMP5:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR_2]], align 8
+// CHECK3-IRBUILDER-NEXT:    [[OMP_GLOBAL_THREAD_NUM:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB8:[0-9]+]])
+// CHECK3-IRBUILDER-NEXT:    call void @__kmpc_dispatch_init_8(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM]], i32 70, i64 0, i64 [[TMP5]], i64 1, i64 1)
+// CHECK3-IRBUILDER-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
+// CHECK3-IRBUILDER:       omp.dispatch.cond:
+// CHECK3-IRBUILDER-NEXT:    [[OMP_GLOBAL_THREAD_NUM10:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB8]])
+// CHECK3-IRBUILDER-NEXT:    [[TMP6:%.*]] = call i32 @__kmpc_dispatch_next_8(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM10]], i32* [[DOTOMP_IS_LAST]], i64* [[DOTOMP_LB]], i64* [[DOTOMP_UB]], i64* [[DOTOMP_STRIDE]])
+// CHECK3-IRBUILDER-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP6]], 0
+// CHECK3-IRBUILDER-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
+// CHECK3-IRBUILDER:       omp.dispatch.body:
+// CHECK3-IRBUILDER-NEXT:    [[TMP7:%.*]] = load i64, i64* [[DOTOMP_LB]], align 8
+// CHECK3-IRBUILDER-NEXT:    store i64 [[TMP7]], i64* [[DOTOMP_IV]], align 8
+// CHECK3-IRBUILDER-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
+// CHECK3-IRBUILDER:       omp.inner.for.cond:
+// CHECK3-IRBUILDER-NEXT:    [[TMP8:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
+// CHECK3-IRBUILDER-NEXT:    [[TMP9:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
+// CHECK3-IRBUILDER-NEXT:    [[CMP11:%.*]] = icmp sle i64 [[TMP8]], [[TMP9]]
+// CHECK3-IRBUILDER-NEXT:    br i1 [[CMP11]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// CHECK3-IRBUILDER:       omp.inner.for.body:
+// CHECK3-IRBUILDER-NEXT:    [[TMP10:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1
+// CHECK3-IRBUILDER-NEXT:    [[CONV12:%.*]] = sext i8 [[TMP10]] to i64
+// CHECK3-IRBUILDER-NEXT:    [[TMP11:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
+// CHECK3-IRBUILDER-NEXT:    [[DIV13:%.*]] = sdiv i64 [[TMP11]], 11
+// CHECK3-IRBUILDER-NEXT:    [[MUL14:%.*]] = mul nsw i64 [[DIV13]], 1
+// CHECK3-IRBUILDER-NEXT:    [[ADD15:%.*]] = add nsw i64 [[CONV12]], [[MUL14]]
+// CHECK3-IRBUILDER-NEXT:    [[CONV16:%.*]] = trunc i64 [[ADD15]] to i8
+// CHECK3-IRBUILDER-NEXT:    store i8 [[CONV16]], i8* [[I8]], align 1
+// CHECK3-IRBUILDER-NEXT:    [[TMP12:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
+// CHECK3-IRBUILDER-NEXT:    [[TMP13:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
+// CHECK3-IRBUILDER-NEXT:    [[DIV17:%.*]] = sdiv i64 [[TMP13]], 11
+// CHECK3-IRBUILDER-NEXT:    [[MUL18:%.*]] = mul nsw i64 [[DIV17]], 11
+// CHECK3-IRBUILDER-NEXT:    [[SUB19:%.*]] = sub nsw i64 [[TMP12]], [[MUL18]]
+// CHECK3-IRBUILDER-NEXT:    [[MUL20:%.*]] = mul nsw i64 [[SUB19]], 1
+// CHECK3-IRBUILDER-NEXT:    [[SUB21:%.*]] = sub nsw i64 11, [[MUL20]]
+// CHECK3-IRBUILDER-NEXT:    [[CONV22:%.*]] = trunc i64 [[SUB21]] to i32
+// CHECK3-IRBUILDER-NEXT:    store i32 [[CONV22]], i32* [[X9]], align 4
+// CHECK3-IRBUILDER-NEXT:    [[OMP_GLOBAL_THREAD_NUM23:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
+// CHECK3-IRBUILDER-NEXT:    call void @__kmpc_ordered(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM23]])
+// CHECK3-IRBUILDER-NEXT:    [[TMP14:%.*]] = load float*, float** [[B_ADDR]], align 8
+// CHECK3-IRBUILDER-NEXT:    [[TMP15:%.*]] = load i8, i8* [[I8]], align 1
+// CHECK3-IRBUILDER-NEXT:    [[IDXPROM:%.*]] = sext i8 [[TMP15]] to i64
+// CHECK3-IRBUILDER-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, float* [[TMP14]], i64 [[IDXPROM]]
+// CHECK3-IRBUILDER-NEXT:    [[TMP16:%.*]] = load float, float* [[ARRAYIDX]], align 4
+// CHECK3-IRBUILDER-NEXT:    [[TMP17:%.*]] = load float*, float** [[C_ADDR]], align 8
+// CHECK3-IRBUILDER-NEXT:    [[TMP18:%.*]] = load i8, i8* [[I8]], align 1
+// CHECK3-IRBUILDER-NEXT:    [[IDXPROM24:%.*]] = sext i8 [[TMP18]] to i64
+// CHECK3-IRBUILDER-NEXT:    [[ARRAYIDX25:%.*]] = getelementptr inbounds float, float* [[TMP17]], i64 [[IDXPROM24]]
+// CHECK3-IRBUILDER-NEXT:    [[TMP19:%.*]] = load float, float* [[ARRAYIDX25]], align 4
+// CHECK3-IRBUILDER-NEXT:    [[MUL26:%.*]] = fmul float [[TMP16]], [[TMP19]]
+// CHECK3-IRBUILDER-NEXT:    [[TMP20:%.*]] = load float*, float** [[D_ADDR]], align 8
+// CHECK3-IRBUILDER-NEXT:    [[TMP21:%.*]] = load i8, i8* [[I8]], align 1
+// CHECK3-IRBUILDER-NEXT:    [[IDXPROM27:%.*]] = sext i8 [[TMP21]] to i64
+// CHECK3-IRBUILDER-NEXT:    [[ARRAYIDX28:%.*]] = getelementptr inbounds float, float* [[TMP20]], i64 [[IDXPROM27]]
+// CHECK3-IRBUILDER-NEXT:    [[TMP22:%.*]] = load float, float* [[ARRAYIDX28]], align 4
+// CHECK3-IRBUILDER-NEXT:    [[MUL29:%.*]] = fmul float [[MUL26]], [[TMP22]]
+// CHECK3-IRBUILDER-NEXT:    [[TMP23:%.*]] = load float*, float** [[A_ADDR]], align 8
+// CHECK3-IRBUILDER-NEXT:    [[TMP24:%.*]] = load i8, i8* [[I8]], align 1
+// CHECK3-IRBUILDER-NEXT:    [[IDXPROM30:%.*]] = sext i8 [[TMP24]] to i64
+// CHECK3-IRBUILDER-NEXT:    [[ARRAYIDX31:%.*]] = getelementptr inbounds float, float* [[TMP23]], i64 [[IDXPROM30]]
+// CHECK3-IRBUILDER-NEXT:    store float [[MUL29]], float* [[ARRAYIDX31]], align 4
+// CHECK3-IRBUILDER-NEXT:    br label [[OMP_INNER_FOR_BODY_ORDERED_AFTER:%.*]]
+// CHECK3-IRBUILDER:       omp.inner.for.body.ordered.after:
+// CHECK3-IRBUILDER-NEXT:    call void @__kmpc_end_ordered(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM23]])
+// CHECK3-IRBUILDER-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
+// CHECK3-IRBUILDER:       omp.body.continue:
+// CHECK3-IRBUILDER-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
+// CHECK3-IRBUILDER:       omp.inner.for.inc:
+// CHECK3-IRBUILDER-NEXT:    [[TMP25:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
+// CHECK3-IRBUILDER-NEXT:    [[ADD32:%.*]] = add nsw i64 [[TMP25]], 1
+// CHECK3-IRBUILDER-NEXT:    store i64 [[ADD32]], i64* [[DOTOMP_IV]], align 8
+// CHECK3-IRBUILDER-NEXT:    [[OMP_GLOBAL_THREAD_NUM33:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB8]])
+// CHECK3-IRBUILDER-NEXT:    call void @__kmpc_dispatch_fini_8(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM33]])
+// CHECK3-IRBUILDER-NEXT:    br label [[OMP_INNER_FOR_COND]]
+// CHECK3-IRBUILDER:       omp.inner.for.end:
+// CHECK3-IRBUILDER-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
+// CHECK3-IRBUILDER:       omp.dispatch.inc:
+// CHECK3-IRBUILDER-NEXT:    br label [[OMP_DISPATCH_COND]]
+// CHECK3-IRBUILDER:       omp.dispatch.end:
+// CHECK3-IRBUILDER-NEXT:    br label [[OMP_PRECOND_END]]
+// CHECK3-IRBUILDER:       omp.precond.end:
+// CHECK3-IRBUILDER-NEXT:    [[OMP_GLOBAL_THREAD_NUM34:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
+// CHECK3-IRBUILDER-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB4]], i32 [[OMP_GLOBAL_THREAD_NUM34]])
+// CHECK3-IRBUILDER-NEXT:    ret void
+//
+//
+// CHECK3-IRBUILDER-LABEL: define {{[^@]+}}@_Z7runtimePfS_S_S_
+// CHECK3-IRBUILDER-SAME: (float* noundef [[A:%.*]], float* noundef [[B:%.*]], float* noundef [[C:%.*]], float* noundef [[D:%.*]]) #[[ATTR0]] {
+// CHECK3-IRBUILDER-NEXT:  entry:
+// CHECK3-IRBUILDER-NEXT:    [[A_ADDR:%.*]] = alloca float*, align 8
+// CHECK3-IRBUILDER-NEXT:    [[B_ADDR:%.*]] = alloca float*, align 8
+// CHECK3-IRBUILDER-NEXT:    [[C_ADDR:%.*]] = alloca float*, align 8
+// CHECK3-IRBUILDER-NEXT:    [[D_ADDR:%.*]] = alloca float*, align 8
+// CHECK3-IRBUILDER-NEXT:    [[X:%.*]] = alloca i32, align 4
+// CHECK3-IRBUILDER-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
+// CHECK3-IRBUILDER-NEXT:    [[TMP:%.*]] = alloca i8, align 1
+// CHECK3-IRBUILDER-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
+// CHECK3-IRBUILDER-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
+// CHECK3-IRBUILDER-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
+// CHECK3-IRBUILDER-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
+// CHECK3-IRBUILDER-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
+// CHECK3-IRBUILDER-NEXT:    [[I:%.*]] = alloca i8, align 1
+// CHECK3-IRBUILDER-NEXT:    [[X2:%.*]] = alloca i32, align 4
+// CHECK3-IRBUILDER-NEXT:    store float* [[A]], float** [[A_ADDR]], align 8
+// CHECK3-IRBUILDER-NEXT:    store float* [[B]], float** [[B_ADDR]], align 8
+// CHECK3-IRBUILDER-NEXT:    store float* [[C]], float** [[C_ADDR]], align 8
+// CHECK3-IRBUILDER-NEXT:    store float* [[D]], float** [[D_ADDR]], align 8
+// CHECK3-IRBUILDER-NEXT:    store i32 0, i32* [[X]], align 4
+// CHECK3-IRBUILDER-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
+// CHECK3-IRBUILDER-NEXT:    store i32 199, i32* [[DOTOMP_UB]], align 4
+// CHECK3-IRBUILDER-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
+// CHECK3-IRBUILDER-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
+// CHECK3-IRBUILDER-NEXT:    [[OMP_GLOBAL_THREAD_NUM:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB10:[0-9]+]])
+// CHECK3-IRBUILDER-NEXT:    call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM]], i32 69, i32 0, i32 199, i32 1, i32 1)
+// CHECK3-IRBUILDER-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
+// CHECK3-IRBUILDER:       omp.dispatch.cond:
+// CHECK3-IRBUILDER-NEXT:    [[OMP_GLOBAL_THREAD_NUM3:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB10]])
+// CHECK3-IRBUILDER-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM3]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
+// CHECK3-IRBUILDER-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP0]], 0
+// CHECK3-IRBUILDER-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
+// CHECK3-IRBUILDER:       omp.dispatch.body:
+// CHECK3-IRBUILDER-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
+// CHECK3-IRBUILDER-NEXT:    store i32 [[TMP1]], i32* [[DOTOMP_IV]], align 4
+// CHECK3-IRBUILDER-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
+// CHECK3-IRBUILDER:       omp.inner.for.cond:
+// CHECK3-IRBUILDER-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
+// CHECK3-IRBUILDER-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
+// CHECK3-IRBUILDER-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP2]], [[TMP3]]
+// CHECK3-IRBUILDER-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// CHECK3-IRBUILDER:       omp.inner.for.body:
+// CHECK3-IRBUILDER-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
+// CHECK3-IRBUILDER-NEXT:    [[DIV:%.*]] = sdiv i32 [[TMP4]], 20
+// CHECK3-IRBUILDER-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV]], 1
+// CHECK3-IRBUILDER-NEXT:    [[ADD:%.*]] = add nsw i32 48, [[MUL]]
+// CHECK3-IRBUILDER-NEXT:    [[CONV:%.*]] = trunc i32 [[ADD]] to i8
+// CHECK3-IRBUILDER-NEXT:    store i8 [[CONV]], i8* [[I]], align 1
+// CHECK3-IRBUILDER-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
+// CHECK3-IRBUILDER-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
+// CHECK3-IRBUILDER-NEXT:    [[DIV4:%.*]] = sdiv i32 [[TMP6]], 20
+// CHECK3-IRBUILDER-NEXT:    [[MUL5:%.*]] = mul nsw i32 [[DIV4]], 20
+// CHECK3-IRBUILDER-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP5]], [[MUL5]]
+// CHECK3-IRBUILDER-NEXT:    [[MUL6:%.*]] = mul nsw i32 [[SUB]], 1
+// CHECK3-IRBUILDER-NEXT:    [[ADD7:%.*]] = add nsw i32 -10, [[MUL6]]
+// CHECK3-IRBUILDER-NEXT:    store i32 [[ADD7]], i32* [[X2]], align 4
+// CHECK3-IRBUILDER-NEXT:    [[OMP_GLOBAL_THREAD_NUM8:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
+// CHECK3-IRBUILDER-NEXT:    call void @__kmpc_ordered(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM8]])
+// CHECK3-IRBUILDER-NEXT:    [[TMP7:%.*]] = load float*, float** [[B_ADDR]], align 8
+// CHECK3-IRBUILDER-NEXT:    [[TMP8:%.*]] = load i8, i8* [[I]], align 1
+// CHECK3-IRBUILDER-NEXT:    [[IDXPROM:%.*]] = zext i8 [[TMP8]] to i64
+// CHECK3-IRBUILDER-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, float* [[TMP7]], i64 [[IDXPROM]]
+// CHECK3-IRBUILDER-NEXT:    [[TMP9:%.*]] = load float, float* [[ARRAYIDX]], align 4
+// CHECK3-IRBUILDER-NEXT:    [[TMP10:%.*]] = load float*, float** [[C_ADDR]], align 8
+// CHECK3-IRBUILDER-NEXT:    [[TMP11:%.*]] = load i8, i8* [[I]], align 1
+// CHECK3-IRBUILDER-NEXT:    [[IDXPROM9:%.*]] = zext i8 [[TMP11]] to i64
+// CHECK3-IRBUILDER-NEXT:    [[ARRAYIDX10:%.*]] = getelementptr inbounds float, float* [[TMP10]], i64 [[IDXPROM9]]
+// CHECK3-IRBUILDER-NEXT:    [[TMP12:%.*]] = load float, float* [[ARRAYIDX10]], align 4
+// CHECK3-IRBUILDER-NEXT:    [[MUL11:%.*]] = fmul float [[TMP9]], [[TMP12]]
+// CHECK3-IRBUILDER-NEXT:    [[TMP13:%.*]] = load float*, float** [[D_ADDR]], align 8
+// CHECK3-IRBUILDER-NEXT:    [[TMP14:%.*]] = load i8, i8* [[I]], align 1
+// CHECK3-IRBUILDER-NEXT:    [[IDXPROM12:%.*]] = zext i8 [[TMP14]] to i64
+// CHECK3-IRBUILDER-NEXT:    [[ARRAYIDX13:%.*]] = getelementptr inbounds float, float* [[TMP13]], i64 [[IDXPROM12]]
+// CHECK3-IRBUILDER-NEXT:    [[TMP15:%.*]] = load float, float* [[ARRAYIDX13]], align 4
+// CHECK3-IRBUILDER-NEXT:    [[MUL14:%.*]] = fmul float [[MUL11]], [[TMP15]]
+// CHECK3-IRBUILDER-NEXT:    [[TMP16:%.*]] = load float*, float** [[A_ADDR]], align 8
+// CHECK3-IRBUILDER-NEXT:    [[TMP17:%.*]] = load i8, i8* [[I]], align 1
+// CHECK3-IRBUILDER-NEXT:    [[IDXPROM15:%.*]] = zext i8 [[TMP17]] to i64
+// CHECK3-IRBUILDER-NEXT:    [[ARRAYIDX16:%.*]] = getelementptr inbounds float, float* [[TMP16]], i64 [[IDXPROM15]]
+// CHECK3-IRBUILDER-NEXT:    store float [[MUL14]], float* [[ARRAYIDX16]], align 4
+// CHECK3-IRBUILDER-NEXT:    br label [[OMP_INNER_FOR_BODY_ORDERED_AFTER:%.*]]
+// CHECK3-IRBUILDER:       omp.inner.for.body.ordered.after:
+// CHECK3-IRBUILDER-NEXT:    call void @__kmpc_end_ordered(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM8]])
+// CHECK3-IRBUILDER-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
+// CHECK3-IRBUILDER:       omp.body.continue:
+// CHECK3-IRBUILDER-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
+// CHECK3-IRBUILDER:       omp.inner.for.inc:
+// CHECK3-IRBUILDER-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
+// CHECK3-IRBUILDER-NEXT:    [[ADD17:%.*]] = add nsw i32 [[TMP18]], 1
+// CHECK3-IRBUILDER-NEXT:    store i32 [[ADD17]], i32* [[DOTOMP_IV]], align 4
+// CHECK3-IRBUILDER-NEXT:    [[OMP_GLOBAL_THREAD_NUM18:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB10]])
+// CHECK3-IRBUILDER-NEXT:    call void @__kmpc_dispatch_fini_4(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM18]])
+// CHECK3-IRBUILDER-NEXT:    br label [[OMP_INNER_FOR_COND]]
+// CHECK3-IRBUILDER:       omp.inner.for.end:
+// CHECK3-IRBUILDER-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
+// CHECK3-IRBUILDER:       omp.dispatch.inc:
+// CHECK3-IRBUILDER-NEXT:    br label [[OMP_DISPATCH_COND]]
+// CHECK3-IRBUILDER:       omp.dispatch.end:
+// CHECK3-IRBUILDER-NEXT:    [[OMP_GLOBAL_THREAD_NUM19:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
+// CHECK3-IRBUILDER-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB4]], i32 [[OMP_GLOBAL_THREAD_NUM19]])
+// CHECK3-IRBUILDER-NEXT:    ret void
+//
+//
+// CHECK3-IRBUILDER-LABEL: define {{[^@]+}}@_Z8foo_simdii
+// CHECK3-IRBUILDER-SAME: (i32 noundef [[LOW:%.*]], i32 noundef [[UP:%.*]]) #[[ATTR0]] {
+// CHECK3-IRBUILDER-NEXT:  entry:
+// CHECK3-IRBUILDER-NEXT:    [[LOW_ADDR:%.*]] = alloca i32, align 4
+// CHECK3-IRBUILDER-NEXT:    [[UP_ADDR:%.*]] = alloca i32, align 4
+// CHECK3-IRBUILDER-NEXT:    [[TMP:%.*]] = alloca i32, align 4
+// CHECK3-IRBUILDER-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
+// CHECK3-IRBUILDER-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
+// CHECK3-IRBUILDER-NEXT:    [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
+// CHECK3-IRBUILDER-NEXT:    [[I:%.*]] = alloca i32, align 4
+// CHECK3-IRBUILDER-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
+// CHECK3-IRBUILDER-NEXT:    [[I5:%.*]] = alloca i32, align 4
+// CHECK3-IRBUILDER-NEXT:    [[DOTOMP_IV16:%.*]] = alloca i32, align 4
+// CHECK3-IRBUILDER-NEXT:    [[_TMP17:%.*]] = alloca i32, align 4
+// CHECK3-IRBUILDER-NEXT:    [[DOTCAPTURE_EXPR_18:%.*]] = alloca i32, align 4
+// CHECK3-IRBUILDER-NEXT:    [[DOTCAPTURE_EXPR_19:%.*]] = alloca i32, align 4
+// CHECK3-IRBUILDER-NEXT:    [[DOTCAPTURE_EXPR_20:%.*]] = alloca i32, align 4
+// CHECK3-IRBUILDER-NEXT:    [[I26:%.*]] = alloca i32, align 4
+// CHECK3-IRBUILDER-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
+// CHECK3-IRBUILDER-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
+// CHECK3-IRBUILDER-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
+// CHECK3-IRBUILDER-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
+// CHECK3-IRBUILDER-NEXT:    [[I28:%.*]] = alloca i32, align 4
+// CHECK3-IRBUILDER-NEXT:    store i32 [[LOW]], i32* [[LOW_ADDR]], align 4
+// CHECK3-IRBUILDER-NEXT:    store i32 [[UP]], i32* [[UP_ADDR]], align 4
+// CHECK3-IRBUILDER-NEXT:    [[TMP0:%.*]] = load i32, i32* [[LOW_ADDR]], align 4
+// CHECK3-IRBUILDER-NEXT:    store i32 [[TMP0]], i32* [[DOTCAPTURE_EXPR_]], align 4
+// CHECK3-IRBUILDER-NEXT:    [[TMP1:%.*]] = load i32, i32* [[UP_ADDR]], align 4
+// CHECK3-IRBUILDER-NEXT:    store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_1]], align 4
+// CHECK3-IRBUILDER-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
+// CHECK3-IRBUILDER-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
+// CHECK3-IRBUILDER-NEXT:    [[SUB:%.*]] = sub i32 [[TMP2]], [[TMP3]]
+// CHECK3-IRBUILDER-NEXT:    [[SUB3:%.*]] = sub i32 [[SUB]], 1
+// CHECK3-IRBUILDER-NEXT:    [[ADD:%.*]] = add i32 [[SUB3]], 1
+// CHECK3-IRBUILDER-NEXT:    [[DIV:%.*]] = udiv i32 [[ADD]], 1
+// CHECK3-IRBUILDER-NEXT:    [[SUB4:%.*]] = sub i32 [[DIV]], 1
+// CHECK3-IRBUILDER-NEXT:    store i32 [[SUB4]], i32* [[DOTCAPTURE_EXPR_2]], align 4
+// CHECK3-IRBUILDER-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
+// CHECK3-IRBUILDER-NEXT:    store i32 [[TMP4]], i32* [[I]], align 4
+// CHECK3-IRBUILDER-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
+// CHECK3-IRBUILDER-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
+// CHECK3-IRBUILDER-NEXT:    [[CMP:%.*]] = icmp slt i32 [[TMP5]], [[TMP6]]
+// CHECK3-IRBUILDER-NEXT:    br i1 [[CMP]], label [[SIMD_IF_THEN:%.*]], label [[SIMD_IF_END:%.*]]
+// CHECK3-IRBUILDER:       simd.if.then:
+// CHECK3-IRBUILDER-NEXT:    store i32 0, i32* [[DOTOMP_IV]], align 4
+// CHECK3-IRBUILDER-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
+// CHECK3-IRBUILDER:       omp.inner.for.cond:
+// CHECK3-IRBUILDER-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !3
+// CHECK3-IRBUILDER-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4, !llvm.access.group !3
+// CHECK3-IRBUILDER-NEXT:    [[ADD6:%.*]] = add i32 [[TMP8]], 1
+// CHECK3-IRBUILDER-NEXT:    [[CMP7:%.*]] = icmp ult i32 [[TMP7]], [[ADD6]]
+// CHECK3-IRBUILDER-NEXT:    br i1 [[CMP7]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// CHECK3-IRBUILDER:       omp.inner.for.body:
+// CHECK3-IRBUILDER-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4, !llvm.access.group !3
+// CHECK3-IRBUILDER-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !3
+// CHECK3-IRBUILDER-NEXT:    [[MUL:%.*]] = mul i32 [[TMP10]], 1
+// CHECK3-IRBUILDER-NEXT:    [[ADD8:%.*]] = add i32 [[TMP9]], [[MUL]]
+// CHECK3-IRBUILDER-NEXT:    store i32 [[ADD8]], i32* [[I5]], align 4, !llvm.access.group !3
+// CHECK3-IRBUILDER-NEXT:    [[TMP11:%.*]] = load i32, i32* [[I5]], align 4, !llvm.access.group !3
+// CHECK3-IRBUILDER-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64
+// CHECK3-IRBUILDER-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x float], [10 x float]* @f, i64 0, i64 [[IDXPROM]]
+// CHECK3-IRBUILDER-NEXT:    store float 0.000000e+00, float* [[ARRAYIDX]], align 4, !llvm.access.group !3
+// CHECK3-IRBUILDER-NEXT:    call void @__captured_stmt(i32* [[I5]]), !llvm.access.group !3
+// CHECK3-IRBUILDER-NEXT:    br label [[OMP_INNER_FOR_BODY_ORDERED_AFTER:%.*]]
+// CHECK3-IRBUILDER:       omp.inner.for.body.ordered.after:
+// CHECK3-IRBUILDER-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
+// CHECK3-IRBUILDER:       omp.body.continue:
+// CHECK3-IRBUILDER-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
+// CHECK3-IRBUILDER:       omp.inner.for.inc:
+// CHECK3-IRBUILDER-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !3
+// CHECK3-IRBUILDER-NEXT:    [[ADD9:%.*]] = add i32 [[TMP12]], 1
+// CHECK3-IRBUILDER-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !3
+// CHECK3-IRBUILDER-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP4:![0-9]+]]
+// CHECK3-IRBUILDER:       omp.inner.for.end:
+// CHECK3-IRBUILDER-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
+// CHECK3-IRBUILDER-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
+// CHECK3-IRBUILDER-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
+// CHECK3-IRBUILDER-NEXT:    [[SUB10:%.*]] = sub i32 [[TMP14]], [[TMP15]]
+// CHECK3-IRBUILDER-NEXT:    [[SUB11:%.*]] = sub i32 [[SUB10]], 1
+// CHECK3-IRBUILDER-NEXT:    [[ADD12:%.*]] = add i32 [[SUB11]], 1
+// CHECK3-IRBUILDER-NEXT:    [[DIV13:%.*]] = udiv i32 [[ADD12]], 1
+// CHECK3-IRBUILDER-NEXT:    [[MUL14:%.*]] = mul i32 [[DIV13]], 1
+// CHECK3-IRBUILDER-NEXT:    [[ADD15:%.*]] = add i32 [[TMP13]], [[MUL14]]
+// CHECK3-IRBUILDER-NEXT:    store i32 [[ADD15]], i32* [[I5]], align 4
+// CHECK3-IRBUILDER-NEXT:    br label [[SIMD_IF_END]]
+// CHECK3-IRBUILDER:       simd.if.end:
+// CHECK3-IRBUILDER-NEXT:    [[TMP16:%.*]] = load i32, i32* [[LOW_ADDR]], align 4
+// CHECK3-IRBUILDER-NEXT:    store i32 [[TMP16]], i32* [[DOTCAPTURE_EXPR_18]], align 4
+// CHECK3-IRBUILDER-NEXT:    [[TMP17:%.*]] = load i32, i32* [[UP_ADDR]], align 4
+// CHECK3-IRBUILDER-NEXT:    store i32 [[TMP17]], i32* [[DOTCAPTURE_EXPR_19]], align 4
+// CHECK3-IRBUILDER-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_19]], align 4
+// CHECK3-IRBUILDER-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_18]], align 4
+// CHECK3-IRBUILDER-NEXT:    [[SUB21:%.*]] = sub i32 [[TMP18]], [[TMP19]]
+// CHECK3-IRBUILDER-NEXT:    [[SUB22:%.*]] = sub i32 [[SUB21]], 1
+// CHECK3-IRBUILDER-NEXT:    [[ADD23:%.*]] = add i32 [[SUB22]], 1
+// CHECK3-IRBUILDER-NEXT:    [[DIV24:%.*]] = udiv i32 [[ADD23]], 1
+// CHECK3-IRBUILDER-NEXT:    [[SUB25:%.*]] = sub i32 [[DIV24]], 1
+// CHECK3-IRBUILDER-NEXT:    store i32 [[SUB25]], i32* [[DOTCAPTURE_EXPR_20]], align 4
+// CHECK3-IRBUILDER-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_18]], align 4
+// CHECK3-IRBUILDER-NEXT:    store i32 [[TMP20]], i32* [[I26]], align 4
+// CHECK3-IRBUILDER-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_18]], align 4
+// CHECK3-IRBUILDER-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_19]], align 4
+// CHECK3-IRBUILDER-NEXT:    [[CMP27:%.*]] = icmp slt i32 [[TMP21]], [[TMP22]]
+// CHECK3-IRBUILDER-NEXT:    br i1 [[CMP27]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
+// CHECK3-IRBUILDER:       omp.precond.then:
+// CHECK3-IRBUILDER-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
+// CHECK3-IRBUILDER-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_20]], align 4
+// CHECK3-IRBUILDER-NEXT:    store i32 [[TMP23]], i32* [[DOTOMP_UB]], align 4
+// CHECK3-IRBUILDER-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
+// CHECK3-IRBUILDER-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
+// CHECK3-IRBUILDER-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_20]], align 4
+// CHECK3-IRBUILDER-NEXT:    [[OMP_GLOBAL_THREAD_NUM:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB12:[0-9]+]])
+// CHECK3-IRBUILDER-NEXT:    call void @__kmpc_dispatch_init_4u(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM]], i32 66, i32 0, i32 [[TMP24]], i32 1, i32 1)
+// CHECK3-IRBUILDER-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
+// CHECK3-IRBUILDER:       omp.dispatch.cond:
+// CHECK3-IRBUILDER-NEXT:    [[OMP_GLOBAL_THREAD_NUM29:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB12]])
+// CHECK3-IRBUILDER-NEXT:    [[TMP25:%.*]] = call i32 @__kmpc_dispatch_next_4u(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM29]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
+// CHECK3-IRBUILDER-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP25]], 0
+// CHECK3-IRBUILDER-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
+// CHECK3-IRBUILDER:       omp.dispatch.body:
+// CHECK3-IRBUILDER-NEXT:    [[TMP26:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
+// CHECK3-IRBUILDER-NEXT:    store i32 [[TMP26]], i32* [[DOTOMP_IV16]], align 4
+// CHECK3-IRBUILDER-NEXT:    br label [[OMP_INNER_FOR_COND30:%.*]]
+// CHECK3-IRBUILDER:       omp.inner.for.cond30:
+// CHECK3-IRBUILDER-NEXT:    [[TMP27:%.*]] = load i32, i32* [[DOTOMP_IV16]], align 4, !llvm.access.group !7
+// CHECK3-IRBUILDER-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !7
+// CHECK3-IRBUILDER-NEXT:    [[ADD31:%.*]] = add i32 [[TMP28]], 1
+// CHECK3-IRBUILDER-NEXT:    [[CMP32:%.*]] = icmp ult i32 [[TMP27]], [[ADD31]]
+// CHECK3-IRBUILDER-NEXT:    br i1 [[CMP32]], label [[OMP_INNER_FOR_BODY33:%.*]], label [[OMP_INNER_FOR_END42:%.*]]
+// CHECK3-IRBUILDER:       omp.inner.for.body33:
+// CHECK3-IRBUILDER-NEXT:    [[TMP29:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_18]], align 4, !llvm.access.group !7
+// CHECK3-IRBUILDER-NEXT:    [[TMP30:%.*]] = load i32, i32* [[DOTOMP_IV16]], align 4, !llvm.access.group !7
+// CHECK3-IRBUILDER-NEXT:    [[MUL34:%.*]] = mul i32 [[TMP30]], 1
+// CHECK3-IRBUILDER-NEXT:    [[ADD35:%.*]] = add i32 [[TMP29]], [[MUL34]]
+// CHECK3-IRBUILDER-NEXT:    store i32 [[ADD35]], i32* [[I28]], align 4, !llvm.access.group !7
+// CHECK3-IRBUILDER-NEXT:    [[TMP31:%.*]] = load i32, i32* [[I28]], align 4, !llvm.access.group !7
+// CHECK3-IRBUILDER-NEXT:    [[IDXPROM36:%.*]] = sext i32 [[TMP31]] to i64
+// CHECK3-IRBUILDER-NEXT:    [[ARRAYIDX37:%.*]] = getelementptr inbounds [10 x float], [10 x float]* @f, i64 0, i64 [[IDXPROM36]]
+// CHECK3-IRBUILDER-NEXT:    store float 0.000000e+00, float* [[ARRAYIDX37]], align 4, !llvm.access.group !7
+// CHECK3-IRBUILDER-NEXT:    call void @__captured_stmt.1(i32* [[I28]]), !llvm.access.group !7
+// CHECK3-IRBUILDER-NEXT:    br label [[OMP_INNER_FOR_BODY33_ORDERED_AFTER:%.*]]
+// CHECK3-IRBUILDER:       omp.inner.for.body33.ordered.after:
+// CHECK3-IRBUILDER-NEXT:    br label [[OMP_BODY_CONTINUE38:%.*]]
+// CHECK3-IRBUILDER:       omp.body.continue38:
+// CHECK3-IRBUILDER-NEXT:    br label [[OMP_INNER_FOR_INC39:%.*]]
+// CHECK3-IRBUILDER:       omp.inner.for.inc39:
+// CHECK3-IRBUILDER-NEXT:    [[TMP32:%.*]] = load i32, i32* [[DOTOMP_IV16]], align 4, !llvm.access.group !7
+// CHECK3-IRBUILDER-NEXT:    [[ADD40:%.*]] = add i32 [[TMP32]], 1
+// CHECK3-IRBUILDER-NEXT:    store i32 [[ADD40]], i32* [[DOTOMP_IV16]], align 4, !llvm.access.group !7
+// CHECK3-IRBUILDER-NEXT:    [[OMP_GLOBAL_THREAD_NUM41:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB12]])
+// CHECK3-IRBUILDER-NEXT:    call void @__kmpc_dispatch_fini_4u(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM41]]), !llvm.access.group !7
+// CHECK3-IRBUILDER-NEXT:    br label [[OMP_INNER_FOR_COND30]], !llvm.loop [[LOOP8:![0-9]+]]
+// CHECK3-IRBUILDER:       omp.inner.for.end42:
+// CHECK3-IRBUILDER-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
+// CHECK3-IRBUILDER:       omp.dispatch.inc:
+// CHECK3-IRBUILDER-NEXT:    br label [[OMP_DISPATCH_COND]]
+// CHECK3-IRBUILDER:       omp.dispatch.end:
+// CHECK3-IRBUILDER-NEXT:    [[TMP33:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
+// CHECK3-IRBUILDER-NEXT:    [[TMP34:%.*]] = icmp ne i32 [[TMP33]], 0
+// CHECK3-IRBUILDER-NEXT:    br i1 [[TMP34]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
+// CHECK3-IRBUILDER:       .omp.final.then:
+// CHECK3-IRBUILDER-NEXT:    [[TMP35:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_18]], align 4
+// CHECK3-IRBUILDER-NEXT:    [[TMP36:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_19]], align 4
+// CHECK3-IRBUILDER-NEXT:    [[TMP37:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_18]], align 4
+// CHECK3-IRBUILDER-NEXT:    [[SUB43:%.*]] = sub i32 [[TMP36]], [[TMP37]]
+// CHECK3-IRBUILDER-NEXT:    [[SUB44:%.*]] = sub i32 [[SUB43]], 1
+// CHECK3-IRBUILDER-NEXT:    [[ADD45:%.*]] = add i32 [[SUB44]], 1
+// CHECK3-IRBUILDER-NEXT:    [[DIV46:%.*]] = udiv i32 [[ADD45]], 1
+// CHECK3-IRBUILDER-NEXT:    [[MUL47:%.*]] = mul i32 [[DIV46]], 1
+// CHECK3-IRBUILDER-NEXT:    [[ADD48:%.*]] = add i32 [[TMP35]], [[MUL47]]
+// CHECK3-IRBUILDER-NEXT:    store i32 [[ADD48]], i32* [[I28]], align 4
+// CHECK3-IRBUILDER-NEXT:    br label [[DOTOMP_FINAL_DONE]]
+// CHECK3-IRBUILDER:       .omp.final.done:
+// CHECK3-IRBUILDER-NEXT:    br label [[OMP_PRECOND_END]]
+// CHECK3-IRBUILDER:       omp.precond.end:
+// CHECK3-IRBUILDER-NEXT:    [[OMP_GLOBAL_THREAD_NUM49:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
+// CHECK3-IRBUILDER-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB4]], i32 [[OMP_GLOBAL_THREAD_NUM49]])
+// CHECK3-IRBUILDER-NEXT:    ret void
+//
+//
+// CHECK3-IRBUILDER-LABEL: define {{[^@]+}}@__captured_stmt
+// CHECK3-IRBUILDER-SAME: (i32* noundef nonnull align 4 dereferenceable(4) [[I:%.*]]) #[[ATTR3:[0-9]+]] {
+// CHECK3-IRBUILDER-NEXT:  entry:
+// CHECK3-IRBUILDER-NEXT:    [[I_ADDR:%.*]] = alloca i32*, align 8
+// CHECK3-IRBUILDER-NEXT:    store i32* [[I]], i32** [[I_ADDR]], align 8
+// CHECK3-IRBUILDER-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[I_ADDR]], align 8
+// CHECK3-IRBUILDER-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
+// CHECK3-IRBUILDER-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP1]] to i64
+// CHECK3-IRBUILDER-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x float], [10 x float]* @f, i64 0, i64 [[IDXPROM]]
+// CHECK3-IRBUILDER-NEXT:    store float 1.000000e+00, float* [[ARRAYIDX]], align 4
+// CHECK3-IRBUILDER-NEXT:    ret void
+//
+//
+// CHECK3-IRBUILDER-LABEL: define {{[^@]+}}@__captured_stmt.1
+// CHECK3-IRBUILDER-SAME: (i32* noundef nonnull align 4 dereferenceable(4) [[I:%.*]]) #[[ATTR3]] {
+// CHECK3-IRBUILDER-NEXT:  entry:
+// CHECK3-IRBUILDER-NEXT:    [[I_ADDR:%.*]] = alloca i32*, align 8
+// CHECK3-IRBUILDER-NEXT:    store i32* [[I]], i32** [[I_ADDR]], align 8
+// CHECK3-IRBUILDER-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[I_ADDR]], align 8
+// CHECK3-IRBUILDER-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
+// CHECK3-IRBUILDER-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP1]] to i64
+// CHECK3-IRBUILDER-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x float], [10 x float]* @f, i64 0, i64 [[IDXPROM]]
+// CHECK3-IRBUILDER-NEXT:    store float 1.000000e+00, float* [[ARRAYIDX]], align 4
+// CHECK3-IRBUILDER-NEXT:    ret void
+//
+//
+// CHECK4-IRBUILDER-LABEL: define {{[^@]+}}@_Z18static_not_chunkedPfS_S_S_
+// CHECK4-IRBUILDER-SAME: (float* noundef [[A:%.*]], float* noundef [[B:%.*]], float* noundef [[C:%.*]], float* noundef [[D:%.*]]) #[[ATTR0:[0-9]+]] {
+// CHECK4-IRBUILDER-NEXT:  entry:
+// CHECK4-IRBUILDER-NEXT:    [[A_ADDR:%.*]] = alloca float*, align 8
+// CHECK4-IRBUILDER-NEXT:    [[B_ADDR:%.*]] = alloca float*, align 8
+// CHECK4-IRBUILDER-NEXT:    [[C_ADDR:%.*]] = alloca float*, align 8
+// CHECK4-IRBUILDER-NEXT:    [[D_ADDR:%.*]] = alloca float*, align 8
+// CHECK4-IRBUILDER-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
+// CHECK4-IRBUILDER-NEXT:    [[TMP:%.*]] = alloca i32, align 4
+// CHECK4-IRBUILDER-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
+// CHECK4-IRBUILDER-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
+// CHECK4-IRBUILDER-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
+// CHECK4-IRBUILDER-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
+// CHECK4-IRBUILDER-NEXT:    [[I:%.*]] = alloca i32, align 4
+// CHECK4-IRBUILDER-NEXT:    store float* [[A]], float** [[A_ADDR]], align 8
+// CHECK4-IRBUILDER-NEXT:    store float* [[B]], float** [[B_ADDR]], align 8
+// CHECK4-IRBUILDER-NEXT:    store float* [[C]], float** [[C_ADDR]], align 8
+// CHECK4-IRBUILDER-NEXT:    store float* [[D]], float** [[D_ADDR]], align 8
+// CHECK4-IRBUILDER-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
+// CHECK4-IRBUILDER-NEXT:    store i32 4571423, i32* [[DOTOMP_UB]], align 4
+// CHECK4-IRBUILDER-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
+// CHECK4-IRBUILDER-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
+// CHECK4-IRBUILDER-NEXT:    [[OMP_GLOBAL_THREAD_NUM:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB3:[0-9]+]])
+// CHECK4-IRBUILDER-NEXT:    call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[OMP_GLOBAL_THREAD_NUM]], i32 66, i32 0, i32 4571423, i32 1, i32 1)
+// CHECK4-IRBUILDER-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
+// CHECK4-IRBUILDER:       omp.dispatch.cond:
+// CHECK4-IRBUILDER-NEXT:    [[OMP_GLOBAL_THREAD_NUM1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB3]])
+// CHECK4-IRBUILDER-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM1]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
+// CHECK4-IRBUILDER-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP0]], 0
+// CHECK4-IRBUILDER-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
+// CHECK4-IRBUILDER:       omp.dispatch.body:
+// CHECK4-IRBUILDER-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
+// CHECK4-IRBUILDER-NEXT:    store i32 [[TMP1]], i32* [[DOTOMP_IV]], align 4
+// CHECK4-IRBUILDER-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
+// CHECK4-IRBUILDER:       omp.inner.for.cond:
+// CHECK4-IRBUILDER-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
+// CHECK4-IRBUILDER-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
+// CHECK4-IRBUILDER-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP2]], [[TMP3]]
+// CHECK4-IRBUILDER-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// CHECK4-IRBUILDER:       omp.inner.for.body:
+// CHECK4-IRBUILDER-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
+// CHECK4-IRBUILDER-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP4]], 7
+// CHECK4-IRBUILDER-NEXT:    [[SUB:%.*]] = sub nsw i32 32000000, [[MUL]]
+// CHECK4-IRBUILDER-NEXT:    store i32 [[SUB]], i32* [[I]], align 4
+// CHECK4-IRBUILDER-NEXT:    [[OMP_GLOBAL_THREAD_NUM2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
+// CHECK4-IRBUILDER-NEXT:    call void @__kmpc_ordered(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM2]])
+// CHECK4-IRBUILDER-NEXT:    [[TMP5:%.*]] = load float*, float** [[B_ADDR]], align 8
+// CHECK4-IRBUILDER-NEXT:    [[TMP6:%.*]] = load i32, i32* [[I]], align 4
+// CHECK4-IRBUILDER-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP6]] to i64
+// CHECK4-IRBUILDER-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, float* [[TMP5]], i64 [[IDXPROM]]
+// CHECK4-IRBUILDER-NEXT:    [[TMP7:%.*]] = load float, float* [[ARRAYIDX]], align 4
+// CHECK4-IRBUILDER-NEXT:    [[TMP8:%.*]] = load float*, float** [[C_ADDR]], align 8
+// CHECK4-IRBUILDER-NEXT:    [[TMP9:%.*]] = load i32, i32* [[I]], align 4
+// CHECK4-IRBUILDER-NEXT:    [[IDXPROM3:%.*]] = sext i32 [[TMP9]] to i64
+// CHECK4-IRBUILDER-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds float, float* [[TMP8]], i64 [[IDXPROM3]]
+// CHECK4-IRBUILDER-NEXT:    [[TMP10:%.*]] = load float, float* [[ARRAYIDX4]], align 4
+// CHECK4-IRBUILDER-NEXT:    [[MUL5:%.*]] = fmul float [[TMP7]], [[TMP10]]
+// CHECK4-IRBUILDER-NEXT:    [[TMP11:%.*]] = load float*, float** [[D_ADDR]], align 8
+// CHECK4-IRBUILDER-NEXT:    [[TMP12:%.*]] = load i32, i32* [[I]], align 4
+// CHECK4-IRBUILDER-NEXT:    [[IDXPROM6:%.*]] = sext i32 [[TMP12]] to i64
+// CHECK4-IRBUILDER-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds float, float* [[TMP11]], i64 [[IDXPROM6]]
+// CHECK4-IRBUILDER-NEXT:    [[TMP13:%.*]] = load float, float* [[ARRAYIDX7]], align 4
+// CHECK4-IRBUILDER-NEXT:    [[MUL8:%.*]] = fmul float [[MUL5]], [[TMP13]]
+// CHECK4-IRBUILDER-NEXT:    [[TMP14:%.*]] = load float*, float** [[A_ADDR]], align 8
+// CHECK4-IRBUILDER-NEXT:    [[TMP15:%.*]] = load i32, i32* [[I]], align 4
+// CHECK4-IRBUILDER-NEXT:    [[IDXPROM9:%.*]] = sext i32 [[TMP15]] to i64
+// CHECK4-IRBUILDER-NEXT:    [[ARRAYIDX10:%.*]] = getelementptr inbounds float, float* [[TMP14]], i64 [[IDXPROM9]]
+// CHECK4-IRBUILDER-NEXT:    store float [[MUL8]], float* [[ARRAYIDX10]], align 4
+// CHECK4-IRBUILDER-NEXT:    br label [[OMP_INNER_FOR_BODY_ORDERED_AFTER:%.*]]
+// CHECK4-IRBUILDER:       omp.inner.for.body.ordered.after:
+// CHECK4-IRBUILDER-NEXT:    call void @__kmpc_end_ordered(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM2]])
+// CHECK4-IRBUILDER-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
+// CHECK4-IRBUILDER:       omp.body.continue:
+// CHECK4-IRBUILDER-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
+// CHECK4-IRBUILDER:       omp.inner.for.inc:
+// CHECK4-IRBUILDER-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
+// CHECK4-IRBUILDER-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP16]], 1
+// CHECK4-IRBUILDER-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
+// CHECK4-IRBUILDER-NEXT:    [[OMP_GLOBAL_THREAD_NUM11:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB3]])
+// CHECK4-IRBUILDER-NEXT:    call void @__kmpc_dispatch_fini_4(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM11]])
+// CHECK4-IRBUILDER-NEXT:    br label [[OMP_INNER_FOR_COND]]
+// CHECK4-IRBUILDER:       omp.inner.for.end:
+// CHECK4-IRBUILDER-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
+// CHECK4-IRBUILDER:       omp.dispatch.inc:
+// CHECK4-IRBUILDER-NEXT:    br label [[OMP_DISPATCH_COND]]
+// CHECK4-IRBUILDER:       omp.dispatch.end:
+// CHECK4-IRBUILDER-NEXT:    [[OMP_GLOBAL_THREAD_NUM12:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
+// CHECK4-IRBUILDER-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB4:[0-9]+]], i32 [[OMP_GLOBAL_THREAD_NUM12]])
+// CHECK4-IRBUILDER-NEXT:    ret void
+//
+//
+// CHECK4-IRBUILDER-LABEL: define {{[^@]+}}@_Z8dynamic1PfS_S_S_
+// CHECK4-IRBUILDER-SAME: (float* noundef [[A:%.*]], float* noundef [[B:%.*]], float* noundef [[C:%.*]], float* noundef [[D:%.*]]) #[[ATTR0]] {
+// CHECK4-IRBUILDER-NEXT:  entry:
+// CHECK4-IRBUILDER-NEXT:    [[A_ADDR:%.*]] = alloca float*, align 8
+// CHECK4-IRBUILDER-NEXT:    [[B_ADDR:%.*]] = alloca float*, align 8
+// CHECK4-IRBUILDER-NEXT:    [[C_ADDR:%.*]] = alloca float*, align 8
+// CHECK4-IRBUILDER-NEXT:    [[D_ADDR:%.*]] = alloca float*, align 8
+// CHECK4-IRBUILDER-NEXT:    [[DOTOMP_IV:%.*]] = alloca i64, align 8
+// CHECK4-IRBUILDER-NEXT:    [[TMP:%.*]] = alloca i64, align 8
+// CHECK4-IRBUILDER-NEXT:    [[DOTOMP_LB:%.*]] = alloca i64, align 8
+// CHECK4-IRBUILDER-NEXT:    [[DOTOMP_UB:%.*]] = alloca i64, align 8
+// CHECK4-IRBUILDER-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i64, align 8
+// CHECK4-IRBUILDER-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
+// CHECK4-IRBUILDER-NEXT:    [[I:%.*]] = alloca i64, align 8
+// CHECK4-IRBUILDER-NEXT:    store float* [[A]], float** [[A_ADDR]], align 8
+// CHECK4-IRBUILDER-NEXT:    store float* [[B]], float** [[B_ADDR]], align 8
+// CHECK4-IRBUILDER-NEXT:    store float* [[C]], float** [[C_ADDR]], align 8
+// CHECK4-IRBUILDER-NEXT:    store float* [[D]], float** [[D_ADDR]], align 8
+// CHECK4-IRBUILDER-NEXT:    store i64 0, i64* [[DOTOMP_LB]], align 8
+// CHECK4-IRBUILDER-NEXT:    store i64 16908287, i64* [[DOTOMP_UB]], align 8
+// CHECK4-IRBUILDER-NEXT:    store i64 1, i64* [[DOTOMP_STRIDE]], align 8
+// CHECK4-IRBUILDER-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
+// CHECK4-IRBUILDER-NEXT:    [[OMP_GLOBAL_THREAD_NUM:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB6:[0-9]+]])
+// CHECK4-IRBUILDER-NEXT:    call void @__kmpc_dispatch_init_8u(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM]], i32 67, i64 0, i64 16908287, i64 1, i64 1)
+// CHECK4-IRBUILDER-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
+// CHECK4-IRBUILDER:       omp.dispatch.cond:
+// CHECK4-IRBUILDER-NEXT:    [[OMP_GLOBAL_THREAD_NUM1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB6]])
+// CHECK4-IRBUILDER-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_dispatch_next_8u(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM1]], i32* [[DOTOMP_IS_LAST]], i64* [[DOTOMP_LB]], i64* [[DOTOMP_UB]], i64* [[DOTOMP_STRIDE]])
+// CHECK4-IRBUILDER-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP0]], 0
+// CHECK4-IRBUILDER-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
+// CHECK4-IRBUILDER:       omp.dispatch.body:
+// CHECK4-IRBUILDER-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTOMP_LB]], align 8
+// CHECK4-IRBUILDER-NEXT:    store i64 [[TMP1]], i64* [[DOTOMP_IV]], align 8
+// CHECK4-IRBUILDER-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
+// CHECK4-IRBUILDER:       omp.inner.for.cond:
+// CHECK4-IRBUILDER-NEXT:    [[TMP2:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
+// CHECK4-IRBUILDER-NEXT:    [[TMP3:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
+// CHECK4-IRBUILDER-NEXT:    [[ADD:%.*]] = add i64 [[TMP3]], 1
+// CHECK4-IRBUILDER-NEXT:    [[CMP:%.*]] = icmp ult i64 [[TMP2]], [[ADD]]
+// CHECK4-IRBUILDER-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// CHECK4-IRBUILDER:       omp.inner.for.body:
+// CHECK4-IRBUILDER-NEXT:    [[TMP4:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
+// CHECK4-IRBUILDER-NEXT:    [[MUL:%.*]] = mul i64 [[TMP4]], 127
+// CHECK4-IRBUILDER-NEXT:    [[ADD2:%.*]] = add i64 131071, [[MUL]]
+// CHECK4-IRBUILDER-NEXT:    store i64 [[ADD2]], i64* [[I]], align 8
+// CHECK4-IRBUILDER-NEXT:    [[OMP_GLOBAL_THREAD_NUM3:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
+// CHECK4-IRBUILDER-NEXT:    call void @__kmpc_ordered(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM3]])
+// CHECK4-IRBUILDER-NEXT:    [[TMP5:%.*]] = load float*, float** [[B_ADDR]], align 8
+// CHECK4-IRBUILDER-NEXT:    [[TMP6:%.*]] = load i64, i64* [[I]], align 8
+// CHECK4-IRBUILDER-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, float* [[TMP5]], i64 [[TMP6]]
+// CHECK4-IRBUILDER-NEXT:    [[TMP7:%.*]] = load float, float* [[ARRAYIDX]], align 4
+// CHECK4-IRBUILDER-NEXT:    [[TMP8:%.*]] = load float*, float** [[C_ADDR]], align 8
+// CHECK4-IRBUILDER-NEXT:    [[TMP9:%.*]] = load i64, i64* [[I]], align 8
+// CHECK4-IRBUILDER-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds float, float* [[TMP8]], i64 [[TMP9]]
+// CHECK4-IRBUILDER-NEXT:    [[TMP10:%.*]] = load float, float* [[ARRAYIDX4]], align 4
+// CHECK4-IRBUILDER-NEXT:    [[MUL5:%.*]] = fmul float [[TMP7]], [[TMP10]]
+// CHECK4-IRBUILDER-NEXT:    [[TMP11:%.*]] = load float*, float** [[D_ADDR]], align 8
+// CHECK4-IRBUILDER-NEXT:    [[TMP12:%.*]] = load i64, i64* [[I]], align 8
+// CHECK4-IRBUILDER-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds float, float* [[TMP11]], i64 [[TMP12]]
+// CHECK4-IRBUILDER-NEXT:    [[TMP13:%.*]] = load float, float* [[ARRAYIDX6]], align 4
+// CHECK4-IRBUILDER-NEXT:    [[MUL7:%.*]] = fmul float [[MUL5]], [[TMP13]]
+// CHECK4-IRBUILDER-NEXT:    [[TMP14:%.*]] = load float*, float** [[A_ADDR]], align 8
+// CHECK4-IRBUILDER-NEXT:    [[TMP15:%.*]] = load i64, i64* [[I]], align 8
+// CHECK4-IRBUILDER-NEXT:    [[ARRAYIDX8:%.*]] = getelementptr inbounds float, float* [[TMP14]], i64 [[TMP15]]
+// CHECK4-IRBUILDER-NEXT:    store float [[MUL7]], float* [[ARRAYIDX8]], align 4
+// CHECK4-IRBUILDER-NEXT:    br label [[OMP_INNER_FOR_BODY_ORDERED_AFTER:%.*]]
+// CHECK4-IRBUILDER:       omp.inner.for.body.ordered.after:
+// CHECK4-IRBUILDER-NEXT:    call void @__kmpc_end_ordered(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM3]])
+// CHECK4-IRBUILDER-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
+// CHECK4-IRBUILDER:       omp.body.continue:
+// CHECK4-IRBUILDER-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
+// CHECK4-IRBUILDER:       omp.inner.for.inc:
+// CHECK4-IRBUILDER-NEXT:    [[TMP16:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
+// CHECK4-IRBUILDER-NEXT:    [[ADD9:%.*]] = add i64 [[TMP16]], 1
+// CHECK4-IRBUILDER-NEXT:    store i64 [[ADD9]], i64* [[DOTOMP_IV]], align 8
+// CHECK4-IRBUILDER-NEXT:    [[OMP_GLOBAL_THREAD_NUM10:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB6]])
+// CHECK4-IRBUILDER-NEXT:    call void @__kmpc_dispatch_fini_8u(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM10]])
+// CHECK4-IRBUILDER-NEXT:    br label [[OMP_INNER_FOR_COND]]
+// CHECK4-IRBUILDER:       omp.inner.for.end:
+// CHECK4-IRBUILDER-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
+// CHECK4-IRBUILDER:       omp.dispatch.inc:
+// CHECK4-IRBUILDER-NEXT:    br label [[OMP_DISPATCH_COND]]
+// CHECK4-IRBUILDER:       omp.dispatch.end:
+// CHECK4-IRBUILDER-NEXT:    [[OMP_GLOBAL_THREAD_NUM11:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
+// CHECK4-IRBUILDER-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB4]], i32 [[OMP_GLOBAL_THREAD_NUM11]])
+// CHECK4-IRBUILDER-NEXT:    ret void
+//
+//
+// CHECK4-IRBUILDER-LABEL: define {{[^@]+}}@_Z9test_autoPfS_S_S_
+// CHECK4-IRBUILDER-SAME: (float* noundef [[A:%.*]], float* noundef [[B:%.*]], float* noundef [[C:%.*]], float* noundef [[D:%.*]]) #[[ATTR0]] {
+// CHECK4-IRBUILDER-NEXT:  entry:
+// CHECK4-IRBUILDER-NEXT:    [[A_ADDR:%.*]] = alloca float*, align 8
+// CHECK4-IRBUILDER-NEXT:    [[B_ADDR:%.*]] = alloca float*, align 8
+// CHECK4-IRBUILDER-NEXT:    [[C_ADDR:%.*]] = alloca float*, align 8
+// CHECK4-IRBUILDER-NEXT:    [[D_ADDR:%.*]] = alloca float*, align 8
+// CHECK4-IRBUILDER-NEXT:    [[X:%.*]] = alloca i32, align 4
+// CHECK4-IRBUILDER-NEXT:    [[Y:%.*]] = alloca i32, align 4
+// CHECK4-IRBUILDER-NEXT:    [[DOTOMP_IV:%.*]] = alloca i64, align 8
+// CHECK4-IRBUILDER-NEXT:    [[TMP:%.*]] = alloca i8, align 1
+// CHECK4-IRBUILDER-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
+// CHECK4-IRBUILDER-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i8, align 1
+// CHECK4-IRBUILDER-NEXT:    [[DOTCAPTURE_EXPR_2:%.*]] = alloca i64, align 8
+// CHECK4-IRBUILDER-NEXT:    [[I:%.*]] = alloca i8, align 1
+// CHECK4-IRBUILDER-NEXT:    [[X6:%.*]] = alloca i32, align 4
+// CHECK4-IRBUILDER-NEXT:    [[DOTOMP_LB:%.*]] = alloca i64, align 8
+// CHECK4-IRBUILDER-NEXT:    [[DOTOMP_UB:%.*]] = alloca i64, align 8
+// CHECK4-IRBUILDER-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i64, align 8
+// CHECK4-IRBUILDER-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
+// CHECK4-IRBUILDER-NEXT:    [[I8:%.*]] = alloca i8, align 1
+// CHECK4-IRBUILDER-NEXT:    [[X9:%.*]] = alloca i32, align 4
+// CHECK4-IRBUILDER-NEXT:    store float* [[A]], float** [[A_ADDR]], align 8
+// CHECK4-IRBUILDER-NEXT:    store float* [[B]], float** [[B_ADDR]], align 8
+// CHECK4-IRBUILDER-NEXT:    store float* [[C]], float** [[C_ADDR]], align 8
+// CHECK4-IRBUILDER-NEXT:    store float* [[D]], float** [[D_ADDR]], align 8
+// CHECK4-IRBUILDER-NEXT:    store i32 0, i32* [[X]], align 4
+// CHECK4-IRBUILDER-NEXT:    store i32 0, i32* [[Y]], align 4
+// CHECK4-IRBUILDER-NEXT:    [[TMP0:%.*]] = load i32, i32* [[Y]], align 4
+// CHECK4-IRBUILDER-NEXT:    [[CONV:%.*]] = trunc i32 [[TMP0]] to i8
+// CHECK4-IRBUILDER-NEXT:    store i8 [[CONV]], i8* [[DOTCAPTURE_EXPR_]], align 1
+// CHECK4-IRBUILDER-NEXT:    [[TMP1:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1
+// CHECK4-IRBUILDER-NEXT:    [[CONV3:%.*]] = sext i8 [[TMP1]] to i32
+// CHECK4-IRBUILDER-NEXT:    [[SUB:%.*]] = sub i32 57, [[CONV3]]
+// CHECK4-IRBUILDER-NEXT:    [[ADD:%.*]] = add i32 [[SUB]], 1
+// CHECK4-IRBUILDER-NEXT:    [[DIV:%.*]] = udiv i32 [[ADD]], 1
+// CHECK4-IRBUILDER-NEXT:    [[CONV4:%.*]] = zext i32 [[DIV]] to i64
+// CHECK4-IRBUILDER-NEXT:    [[MUL:%.*]] = mul nsw i64 [[CONV4]], 11
+// CHECK4-IRBUILDER-NEXT:    [[SUB5:%.*]] = sub nsw i64 [[MUL]], 1
+// CHECK4-IRBUILDER-NEXT:    store i64 [[SUB5]], i64* [[DOTCAPTURE_EXPR_2]], align 8
+// CHECK4-IRBUILDER-NEXT:    [[TMP2:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1
+// CHECK4-IRBUILDER-NEXT:    store i8 [[TMP2]], i8* [[I]], align 1
+// CHECK4-IRBUILDER-NEXT:    store i32 11, i32* [[X6]], align 4
+// CHECK4-IRBUILDER-NEXT:    [[TMP3:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1
+// CHECK4-IRBUILDER-NEXT:    [[CONV7:%.*]] = sext i8 [[TMP3]] to i32
+// CHECK4-IRBUILDER-NEXT:    [[CMP:%.*]] = icmp sle i32 [[CONV7]], 57
+// CHECK4-IRBUILDER-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
+// CHECK4-IRBUILDER:       omp.precond.then:
+// CHECK4-IRBUILDER-NEXT:    store i64 0, i64* [[DOTOMP_LB]], align 8
+// CHECK4-IRBUILDER-NEXT:    [[TMP4:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR_2]], align 8
+// CHECK4-IRBUILDER-NEXT:    store i64 [[TMP4]], i64* [[DOTOMP_UB]], align 8
+// CHECK4-IRBUILDER-NEXT:    store i64 1, i64* [[DOTOMP_STRIDE]], align 8
+// CHECK4-IRBUILDER-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
+// CHECK4-IRBUILDER-NEXT:    [[TMP5:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR_2]], align 8
+// CHECK4-IRBUILDER-NEXT:    [[OMP_GLOBAL_THREAD_NUM:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB8:[0-9]+]])
+// CHECK4-IRBUILDER-NEXT:    call void @__kmpc_dispatch_init_8(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM]], i32 70, i64 0, i64 [[TMP5]], i64 1, i64 1)
+// CHECK4-IRBUILDER-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
+// CHECK4-IRBUILDER:       omp.dispatch.cond:
+// CHECK4-IRBUILDER-NEXT:    [[OMP_GLOBAL_THREAD_NUM10:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB8]])
+// CHECK4-IRBUILDER-NEXT:    [[TMP6:%.*]] = call i32 @__kmpc_dispatch_next_8(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM10]], i32* [[DOTOMP_IS_LAST]], i64* [[DOTOMP_LB]], i64* [[DOTOMP_UB]], i64* [[DOTOMP_STRIDE]])
+// CHECK4-IRBUILDER-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP6]], 0
+// CHECK4-IRBUILDER-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
+// CHECK4-IRBUILDER:       omp.dispatch.body:
+// CHECK4-IRBUILDER-NEXT:    [[TMP7:%.*]] = load i64, i64* [[DOTOMP_LB]], align 8
+// CHECK4-IRBUILDER-NEXT:    store i64 [[TMP7]], i64* [[DOTOMP_IV]], align 8
+// CHECK4-IRBUILDER-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
+// CHECK4-IRBUILDER:       omp.inner.for.cond:
+// CHECK4-IRBUILDER-NEXT:    [[TMP8:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
+// CHECK4-IRBUILDER-NEXT:    [[TMP9:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
+// CHECK4-IRBUILDER-NEXT:    [[CMP11:%.*]] = icmp sle i64 [[TMP8]], [[TMP9]]
+// CHECK4-IRBUILDER-NEXT:    br i1 [[CMP11]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// CHECK4-IRBUILDER:       omp.inner.for.body:
+// CHECK4-IRBUILDER-NEXT:    [[TMP10:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1
+// CHECK4-IRBUILDER-NEXT:    [[CONV12:%.*]] = sext i8 [[TMP10]] to i64
+// CHECK4-IRBUILDER-NEXT:    [[TMP11:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
+// CHECK4-IRBUILDER-NEXT:    [[DIV13:%.*]] = sdiv i64 [[TMP11]], 11
+// CHECK4-IRBUILDER-NEXT:    [[MUL14:%.*]] = mul nsw i64 [[DIV13]], 1
+// CHECK4-IRBUILDER-NEXT:    [[ADD15:%.*]] = add nsw i64 [[CONV12]], [[MUL14]]
+// CHECK4-IRBUILDER-NEXT:    [[CONV16:%.*]] = trunc i64 [[ADD15]] to i8
+// CHECK4-IRBUILDER-NEXT:    store i8 [[CONV16]], i8* [[I8]], align 1
+// CHECK4-IRBUILDER-NEXT:    [[TMP12:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
+// CHECK4-IRBUILDER-NEXT:    [[TMP13:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
+// CHECK4-IRBUILDER-NEXT:    [[DIV17:%.*]] = sdiv i64 [[TMP13]], 11
+// CHECK4-IRBUILDER-NEXT:    [[MUL18:%.*]] = mul nsw i64 [[DIV17]], 11
+// CHECK4-IRBUILDER-NEXT:    [[SUB19:%.*]] = sub nsw i64 [[TMP12]], [[MUL18]]
+// CHECK4-IRBUILDER-NEXT:    [[MUL20:%.*]] = mul nsw i64 [[SUB19]], 1
+// CHECK4-IRBUILDER-NEXT:    [[SUB21:%.*]] = sub nsw i64 11, [[MUL20]]
+// CHECK4-IRBUILDER-NEXT:    [[CONV22:%.*]] = trunc i64 [[SUB21]] to i32
+// CHECK4-IRBUILDER-NEXT:    store i32 [[CONV22]], i32* [[X9]], align 4
+// CHECK4-IRBUILDER-NEXT:    [[OMP_GLOBAL_THREAD_NUM23:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
+// CHECK4-IRBUILDER-NEXT:    call void @__kmpc_ordered(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM23]])
+// CHECK4-IRBUILDER-NEXT:    [[TMP14:%.*]] = load float*, float** [[B_ADDR]], align 8
+// CHECK4-IRBUILDER-NEXT:    [[TMP15:%.*]] = load i8, i8* [[I8]], align 1
+// CHECK4-IRBUILDER-NEXT:    [[IDXPROM:%.*]] = sext i8 [[TMP15]] to i64
+// CHECK4-IRBUILDER-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, float* [[TMP14]], i64 [[IDXPROM]]
+// CHECK4-IRBUILDER-NEXT:    [[TMP16:%.*]] = load float, float* [[ARRAYIDX]], align 4
+// CHECK4-IRBUILDER-NEXT:    [[TMP17:%.*]] = load float*, float** [[C_ADDR]], align 8
+// CHECK4-IRBUILDER-NEXT:    [[TMP18:%.*]] = load i8, i8* [[I8]], align 1
+// CHECK4-IRBUILDER-NEXT:    [[IDXPROM24:%.*]] = sext i8 [[TMP18]] to i64
+// CHECK4-IRBUILDER-NEXT:    [[ARRAYIDX25:%.*]] = getelementptr inbounds float, float* [[TMP17]], i64 [[IDXPROM24]]
+// CHECK4-IRBUILDER-NEXT:    [[TMP19:%.*]] = load float, float* [[ARRAYIDX25]], align 4
+// CHECK4-IRBUILDER-NEXT:    [[MUL26:%.*]] = fmul float [[TMP16]], [[TMP19]]
+// CHECK4-IRBUILDER-NEXT:    [[TMP20:%.*]] = load float*, float** [[D_ADDR]], align 8
+// CHECK4-IRBUILDER-NEXT:    [[TMP21:%.*]] = load i8, i8* [[I8]], align 1
+// CHECK4-IRBUILDER-NEXT:    [[IDXPROM27:%.*]] = sext i8 [[TMP21]] to i64
+// CHECK4-IRBUILDER-NEXT:    [[ARRAYIDX28:%.*]] = getelementptr inbounds float, float* [[TMP20]], i64 [[IDXPROM27]]
+// CHECK4-IRBUILDER-NEXT:    [[TMP22:%.*]] = load float, float* [[ARRAYIDX28]], align 4
+// CHECK4-IRBUILDER-NEXT:    [[MUL29:%.*]] = fmul float [[MUL26]], [[TMP22]]
+// CHECK4-IRBUILDER-NEXT:    [[TMP23:%.*]] = load float*, float** [[A_ADDR]], align 8
+// CHECK4-IRBUILDER-NEXT:    [[TMP24:%.*]] = load i8, i8* [[I8]], align 1
+// CHECK4-IRBUILDER-NEXT:    [[IDXPROM30:%.*]] = sext i8 [[TMP24]] to i64
+// CHECK4-IRBUILDER-NEXT:    [[ARRAYIDX31:%.*]] = getelementptr inbounds float, float* [[TMP23]], i64 [[IDXPROM30]]
+// CHECK4-IRBUILDER-NEXT:    store float [[MUL29]], float* [[ARRAYIDX31]], align 4
+// CHECK4-IRBUILDER-NEXT:    br label [[OMP_INNER_FOR_BODY_ORDERED_AFTER:%.*]]
+// CHECK4-IRBUILDER:       omp.inner.for.body.ordered.after:
+// CHECK4-IRBUILDER-NEXT:    call void @__kmpc_end_ordered(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM23]])
+// CHECK4-IRBUILDER-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
+// CHECK4-IRBUILDER:       omp.body.continue:
+// CHECK4-IRBUILDER-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
+// CHECK4-IRBUILDER:       omp.inner.for.inc:
+// CHECK4-IRBUILDER-NEXT:    [[TMP25:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
+// CHECK4-IRBUILDER-NEXT:    [[ADD32:%.*]] = add nsw i64 [[TMP25]], 1
+// CHECK4-IRBUILDER-NEXT:    store i64 [[ADD32]], i64* [[DOTOMP_IV]], align 8
+// CHECK4-IRBUILDER-NEXT:    [[OMP_GLOBAL_THREAD_NUM33:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB8]])
+// CHECK4-IRBUILDER-NEXT:    call void @__kmpc_dispatch_fini_8(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM33]])
+// CHECK4-IRBUILDER-NEXT:    br label [[OMP_INNER_FOR_COND]]
+// CHECK4-IRBUILDER:       omp.inner.for.end:
+// CHECK4-IRBUILDER-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
+// CHECK4-IRBUILDER:       omp.dispatch.inc:
+// CHECK4-IRBUILDER-NEXT:    br label [[OMP_DISPATCH_COND]]
+// CHECK4-IRBUILDER:       omp.dispatch.end:
+// CHECK4-IRBUILDER-NEXT:    br label [[OMP_PRECOND_END]]
+// CHECK4-IRBUILDER:       omp.precond.end:
+// CHECK4-IRBUILDER-NEXT:    [[OMP_GLOBAL_THREAD_NUM34:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
+// CHECK4-IRBUILDER-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB4]], i32 [[OMP_GLOBAL_THREAD_NUM34]])
+// CHECK4-IRBUILDER-NEXT:    ret void
+//
+//
+// CHECK4-IRBUILDER-LABEL: define {{[^@]+}}@_Z7runtimePfS_S_S_
+// CHECK4-IRBUILDER-SAME: (float* noundef [[A:%.*]], float* noundef [[B:%.*]], float* noundef [[C:%.*]], float* noundef [[D:%.*]]) #[[ATTR0]] {
+// CHECK4-IRBUILDER-NEXT:  entry:
+// CHECK4-IRBUILDER-NEXT:    [[A_ADDR:%.*]] = alloca float*, align 8
+// CHECK4-IRBUILDER-NEXT:    [[B_ADDR:%.*]] = alloca float*, align 8
+// CHECK4-IRBUILDER-NEXT:    [[C_ADDR:%.*]] = alloca float*, align 8
+// CHECK4-IRBUILDER-NEXT:    [[D_ADDR:%.*]] = alloca float*, align 8
+// CHECK4-IRBUILDER-NEXT:    [[X:%.*]] = alloca i32, align 4
+// CHECK4-IRBUILDER-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
+// CHECK4-IRBUILDER-NEXT:    [[TMP:%.*]] = alloca i8, align 1
+// CHECK4-IRBUILDER-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
+// CHECK4-IRBUILDER-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
+// CHECK4-IRBUILDER-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
+// CHECK4-IRBUILDER-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
+// CHECK4-IRBUILDER-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
+// CHECK4-IRBUILDER-NEXT:    [[I:%.*]] = alloca i8, align 1
+// CHECK4-IRBUILDER-NEXT:    [[X2:%.*]] = alloca i32, align 4
+// CHECK4-IRBUILDER-NEXT:    store float* [[A]], float** [[A_ADDR]], align 8
+// CHECK4-IRBUILDER-NEXT:    store float* [[B]], float** [[B_ADDR]], align 8
+// CHECK4-IRBUILDER-NEXT:    store float* [[C]], float** [[C_ADDR]], align 8
+// CHECK4-IRBUILDER-NEXT:    store float* [[D]], float** [[D_ADDR]], align 8
+// CHECK4-IRBUILDER-NEXT:    store i32 0, i32* [[X]], align 4
+// CHECK4-IRBUILDER-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
+// CHECK4-IRBUILDER-NEXT:    store i32 199, i32* [[DOTOMP_UB]], align 4
+// CHECK4-IRBUILDER-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
+// CHECK4-IRBUILDER-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
+// CHECK4-IRBUILDER-NEXT:    [[OMP_GLOBAL_THREAD_NUM:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB10:[0-9]+]])
+// CHECK4-IRBUILDER-NEXT:    call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM]], i32 69, i32 0, i32 199, i32 1, i32 1)
+// CHECK4-IRBUILDER-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
+// CHECK4-IRBUILDER:       omp.dispatch.cond:
+// CHECK4-IRBUILDER-NEXT:    [[OMP_GLOBAL_THREAD_NUM3:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB10]])
+// CHECK4-IRBUILDER-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM3]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
+// CHECK4-IRBUILDER-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP0]], 0
+// CHECK4-IRBUILDER-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
+// CHECK4-IRBUILDER:       omp.dispatch.body:
+// CHECK4-IRBUILDER-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
+// CHECK4-IRBUILDER-NEXT:    store i32 [[TMP1]], i32* [[DOTOMP_IV]], align 4
+// CHECK4-IRBUILDER-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
+// CHECK4-IRBUILDER:       omp.inner.for.cond:
+// CHECK4-IRBUILDER-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
+// CHECK4-IRBUILDER-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
+// CHECK4-IRBUILDER-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP2]], [[TMP3]]
+// CHECK4-IRBUILDER-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// CHECK4-IRBUILDER:       omp.inner.for.body:
+// CHECK4-IRBUILDER-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
+// CHECK4-IRBUILDER-NEXT:    [[DIV:%.*]] = sdiv i32 [[TMP4]], 20
+// CHECK4-IRBUILDER-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV]], 1
+// CHECK4-IRBUILDER-NEXT:    [[ADD:%.*]] = add nsw i32 48, [[MUL]]
+// CHECK4-IRBUILDER-NEXT:    [[CONV:%.*]] = trunc i32 [[ADD]] to i8
+// CHECK4-IRBUILDER-NEXT:    store i8 [[CONV]], i8* [[I]], align 1
+// CHECK4-IRBUILDER-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
+// CHECK4-IRBUILDER-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
+// CHECK4-IRBUILDER-NEXT:    [[DIV4:%.*]] = sdiv i32 [[TMP6]], 20
+// CHECK4-IRBUILDER-NEXT:    [[MUL5:%.*]] = mul nsw i32 [[DIV4]], 20
+// CHECK4-IRBUILDER-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP5]], [[MUL5]]
+// CHECK4-IRBUILDER-NEXT:    [[MUL6:%.*]] = mul nsw i32 [[SUB]], 1
+// CHECK4-IRBUILDER-NEXT:    [[ADD7:%.*]] = add nsw i32 -10, [[MUL6]]
+// CHECK4-IRBUILDER-NEXT:    store i32 [[ADD7]], i32* [[X2]], align 4
+// CHECK4-IRBUILDER-NEXT:    [[OMP_GLOBAL_THREAD_NUM8:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
+// CHECK4-IRBUILDER-NEXT:    call void @__kmpc_ordered(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM8]])
+// CHECK4-IRBUILDER-NEXT:    [[TMP7:%.*]] = load float*, float** [[B_ADDR]], align 8
+// CHECK4-IRBUILDER-NEXT:    [[TMP8:%.*]] = load i8, i8* [[I]], align 1
+// CHECK4-IRBUILDER-NEXT:    [[IDXPROM:%.*]] = zext i8 [[TMP8]] to i64
+// CHECK4-IRBUILDER-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, float* [[TMP7]], i64 [[IDXPROM]]
+// CHECK4-IRBUILDER-NEXT:    [[TMP9:%.*]] = load float, float* [[ARRAYIDX]], align 4
+// CHECK4-IRBUILDER-NEXT:    [[TMP10:%.*]] = load float*, float** [[C_ADDR]], align 8
+// CHECK4-IRBUILDER-NEXT:    [[TMP11:%.*]] = load i8, i8* [[I]], align 1
+// CHECK4-IRBUILDER-NEXT:    [[IDXPROM9:%.*]] = zext i8 [[TMP11]] to i64
+// CHECK4-IRBUILDER-NEXT:    [[ARRAYIDX10:%.*]] = getelementptr inbounds float, float* [[TMP10]], i64 [[IDXPROM9]]
+// CHECK4-IRBUILDER-NEXT:    [[TMP12:%.*]] = load float, float* [[ARRAYIDX10]], align 4
+// CHECK4-IRBUILDER-NEXT:    [[MUL11:%.*]] = fmul float [[TMP9]], [[TMP12]]
+// CHECK4-IRBUILDER-NEXT:    [[TMP13:%.*]] = load float*, float** [[D_ADDR]], align 8
+// CHECK4-IRBUILDER-NEXT:    [[TMP14:%.*]] = load i8, i8* [[I]], align 1
+// CHECK4-IRBUILDER-NEXT:    [[IDXPROM12:%.*]] = zext i8 [[TMP14]] to i64
+// CHECK4-IRBUILDER-NEXT:    [[ARRAYIDX13:%.*]] = getelementptr inbounds float, float* [[TMP13]], i64 [[IDXPROM12]]
+// CHECK4-IRBUILDER-NEXT:    [[TMP15:%.*]] = load float, float* [[ARRAYIDX13]], align 4
+// CHECK4-IRBUILDER-NEXT:    [[MUL14:%.*]] = fmul float [[MUL11]], [[TMP15]]
+// CHECK4-IRBUILDER-NEXT:    [[TMP16:%.*]] = load float*, float** [[A_ADDR]], align 8
+// CHECK4-IRBUILDER-NEXT:    [[TMP17:%.*]] = load i8, i8* [[I]], align 1
+// CHECK4-IRBUILDER-NEXT:    [[IDXPROM15:%.*]] = zext i8 [[TMP17]] to i64
+// CHECK4-IRBUILDER-NEXT:    [[ARRAYIDX16:%.*]] = getelementptr inbounds float, float* [[TMP16]], i64 [[IDXPROM15]]
+// CHECK4-IRBUILDER-NEXT:    store float [[MUL14]], float* [[ARRAYIDX16]], align 4
+// CHECK4-IRBUILDER-NEXT:    br label [[OMP_INNER_FOR_BODY_ORDERED_AFTER:%.*]]
+// CHECK4-IRBUILDER:       omp.inner.for.body.ordered.after:
+// CHECK4-IRBUILDER-NEXT:    call void @__kmpc_end_ordered(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM8]])
+// CHECK4-IRBUILDER-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
+// CHECK4-IRBUILDER:       omp.body.continue:
+// CHECK4-IRBUILDER-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
+// CHECK4-IRBUILDER:       omp.inner.for.inc:
+// CHECK4-IRBUILDER-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
+// CHECK4-IRBUILDER-NEXT:    [[ADD17:%.*]] = add nsw i32 [[TMP18]], 1
+// CHECK4-IRBUILDER-NEXT:    store i32 [[ADD17]], i32* [[DOTOMP_IV]], align 4
+// CHECK4-IRBUILDER-NEXT:    [[OMP_GLOBAL_THREAD_NUM18:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB10]])
+// CHECK4-IRBUILDER-NEXT:    call void @__kmpc_dispatch_fini_4(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM18]])
+// CHECK4-IRBUILDER-NEXT:    br label [[OMP_INNER_FOR_COND]]
+// CHECK4-IRBUILDER:       omp.inner.for.end:
+// CHECK4-IRBUILDER-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
+// CHECK4-IRBUILDER:       omp.dispatch.inc:
+// CHECK4-IRBUILDER-NEXT:    br label [[OMP_DISPATCH_COND]]
+// CHECK4-IRBUILDER:       omp.dispatch.end:
+// CHECK4-IRBUILDER-NEXT:    [[OMP_GLOBAL_THREAD_NUM19:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
+// CHECK4-IRBUILDER-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB4]], i32 [[OMP_GLOBAL_THREAD_NUM19]])
+// CHECK4-IRBUILDER-NEXT:    ret void
+//
+//
+// CHECK4-IRBUILDER-LABEL: define {{[^@]+}}@_Z8foo_simdii
+// CHECK4-IRBUILDER-SAME: (i32 noundef [[LOW:%.*]], i32 noundef [[UP:%.*]]) #[[ATTR0]] {
+// CHECK4-IRBUILDER-NEXT:  entry:
+// CHECK4-IRBUILDER-NEXT:    [[LOW_ADDR:%.*]] = alloca i32, align 4
+// CHECK4-IRBUILDER-NEXT:    [[UP_ADDR:%.*]] = alloca i32, align 4
+// CHECK4-IRBUILDER-NEXT:    [[TMP:%.*]] = alloca i32, align 4
+// CHECK4-IRBUILDER-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
+// CHECK4-IRBUILDER-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
+// CHECK4-IRBUILDER-NEXT:    [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
+// CHECK4-IRBUILDER-NEXT:    [[I:%.*]] = alloca i32, align 4
+// CHECK4-IRBUILDER-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
+// CHECK4-IRBUILDER-NEXT:    [[I5:%.*]] = alloca i32, align 4
+// CHECK4-IRBUILDER-NEXT:    [[DOTOMP_IV16:%.*]] = alloca i32, align 4
+// CHECK4-IRBUILDER-NEXT:    [[_TMP17:%.*]] = alloca i32, align 4
+// CHECK4-IRBUILDER-NEXT:    [[DOTCAPTURE_EXPR_18:%.*]] = alloca i32, align 4
+// CHECK4-IRBUILDER-NEXT:    [[DOTCAPTURE_EXPR_19:%.*]] = alloca i32, align 4
+// CHECK4-IRBUILDER-NEXT:    [[DOTCAPTURE_EXPR_20:%.*]] = alloca i32, align 4
+// CHECK4-IRBUILDER-NEXT:    [[I26:%.*]] = alloca i32, align 4
+// CHECK4-IRBUILDER-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
+// CHECK4-IRBUILDER-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
+// CHECK4-IRBUILDER-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
+// CHECK4-IRBUILDER-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
+// CHECK4-IRBUILDER-NEXT:    [[I28:%.*]] = alloca i32, align 4
+// CHECK4-IRBUILDER-NEXT:    store i32 [[LOW]], i32* [[LOW_ADDR]], align 4
+// CHECK4-IRBUILDER-NEXT:    store i32 [[UP]], i32* [[UP_ADDR]], align 4
+// CHECK4-IRBUILDER-NEXT:    [[TMP0:%.*]] = load i32, i32* [[LOW_ADDR]], align 4
+// CHECK4-IRBUILDER-NEXT:    store i32 [[TMP0]], i32* [[DOTCAPTURE_EXPR_]], align 4
+// CHECK4-IRBUILDER-NEXT:    [[TMP1:%.*]] = load i32, i32* [[UP_ADDR]], align 4
+// CHECK4-IRBUILDER-NEXT:    store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_1]], align 4
+// CHECK4-IRBUILDER-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
+// CHECK4-IRBUILDER-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
+// CHECK4-IRBUILDER-NEXT:    [[SUB:%.*]] = sub i32 [[TMP2]], [[TMP3]]
+// CHECK4-IRBUILDER-NEXT:    [[SUB3:%.*]] = sub i32 [[SUB]], 1
+// CHECK4-IRBUILDER-NEXT:    [[ADD:%.*]] = add i32 [[SUB3]], 1
+// CHECK4-IRBUILDER-NEXT:    [[DIV:%.*]] = udiv i32 [[ADD]], 1
+// CHECK4-IRBUILDER-NEXT:    [[SUB4:%.*]] = sub i32 [[DIV]], 1
+// CHECK4-IRBUILDER-NEXT:    store i32 [[SUB4]], i32* [[DOTCAPTURE_EXPR_2]], align 4
+// CHECK4-IRBUILDER-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
+// CHECK4-IRBUILDER-NEXT:    store i32 [[TMP4]], i32* [[I]], align 4
+// CHECK4-IRBUILDER-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
+// CHECK4-IRBUILDER-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
+// CHECK4-IRBUILDER-NEXT:    [[CMP:%.*]] = icmp slt i32 [[TMP5]], [[TMP6]]
+// CHECK4-IRBUILDER-NEXT:    br i1 [[CMP]], label [[SIMD_IF_THEN:%.*]], label [[SIMD_IF_END:%.*]]
+// CHECK4-IRBUILDER:       simd.if.then:
+// CHECK4-IRBUILDER-NEXT:    store i32 0, i32* [[DOTOMP_IV]], align 4
+// CHECK4-IRBUILDER-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
+// CHECK4-IRBUILDER:       omp.inner.for.cond:
+// CHECK4-IRBUILDER-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !3
+// CHECK4-IRBUILDER-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4, !llvm.access.group !3
+// CHECK4-IRBUILDER-NEXT:    [[ADD6:%.*]] = add i32 [[TMP8]], 1
+// CHECK4-IRBUILDER-NEXT:    [[CMP7:%.*]] = icmp ult i32 [[TMP7]], [[ADD6]]
+// CHECK4-IRBUILDER-NEXT:    br i1 [[CMP7]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// CHECK4-IRBUILDER:       omp.inner.for.body:
+// CHECK4-IRBUILDER-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4, !llvm.access.group !3
+// CHECK4-IRBUILDER-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !3
+// CHECK4-IRBUILDER-NEXT:    [[MUL:%.*]] = mul i32 [[TMP10]], 1
+// CHECK4-IRBUILDER-NEXT:    [[ADD8:%.*]] = add i32 [[TMP9]], [[MUL]]
+// CHECK4-IRBUILDER-NEXT:    store i32 [[ADD8]], i32* [[I5]], align 4, !llvm.access.group !3
+// CHECK4-IRBUILDER-NEXT:    [[TMP11:%.*]] = load i32, i32* [[I5]], align 4, !llvm.access.group !3
+// CHECK4-IRBUILDER-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64
+// CHECK4-IRBUILDER-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x float], [10 x float]* @f, i64 0, i64 [[IDXPROM]]
+// CHECK4-IRBUILDER-NEXT:    store float 0.000000e+00, float* [[ARRAYIDX]], align 4, !llvm.access.group !3
+// CHECK4-IRBUILDER-NEXT:    call void @__captured_stmt(i32* [[I5]]), !llvm.access.group !3
+// CHECK4-IRBUILDER-NEXT:    br label [[OMP_INNER_FOR_BODY_ORDERED_AFTER:%.*]]
+// CHECK4-IRBUILDER:       omp.inner.for.body.ordered.after:
+// CHECK4-IRBUILDER-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
+// CHECK4-IRBUILDER:       omp.body.continue:
+// CHECK4-IRBUILDER-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
+// CHECK4-IRBUILDER:       omp.inner.for.inc:
+// CHECK4-IRBUILDER-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !3
+// CHECK4-IRBUILDER-NEXT:    [[ADD9:%.*]] = add i32 [[TMP12]], 1
+// CHECK4-IRBUILDER-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !3
+// CHECK4-IRBUILDER-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP4:![0-9]+]]
+// CHECK4-IRBUILDER:       omp.inner.for.end:
+// CHECK4-IRBUILDER-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
+// CHECK4-IRBUILDER-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
+// CHECK4-IRBUILDER-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
+// CHECK4-IRBUILDER-NEXT:    [[SUB10:%.*]] = sub i32 [[TMP14]], [[TMP15]]
+// CHECK4-IRBUILDER-NEXT:    [[SUB11:%.*]] = sub i32 [[SUB10]], 1
+// CHECK4-IRBUILDER-NEXT:    [[ADD12:%.*]] = add i32 [[SUB11]], 1
+// CHECK4-IRBUILDER-NEXT:    [[DIV13:%.*]] = udiv i32 [[ADD12]], 1
+// CHECK4-IRBUILDER-NEXT:    [[MUL14:%.*]] = mul i32 [[DIV13]], 1
+// CHECK4-IRBUILDER-NEXT:    [[ADD15:%.*]] = add i32 [[TMP13]], [[MUL14]]
+// CHECK4-IRBUILDER-NEXT:    store i32 [[ADD15]], i32* [[I5]], align 4
+// CHECK4-IRBUILDER-NEXT:    br label [[SIMD_IF_END]]
+// CHECK4-IRBUILDER:       simd.if.end:
+// CHECK4-IRBUILDER-NEXT:    [[TMP16:%.*]] = load i32, i32* [[LOW_ADDR]], align 4
+// CHECK4-IRBUILDER-NEXT:    store i32 [[TMP16]], i32* [[DOTCAPTURE_EXPR_18]], align 4
+// CHECK4-IRBUILDER-NEXT:    [[TMP17:%.*]] = load i32, i32* [[UP_ADDR]], align 4
+// CHECK4-IRBUILDER-NEXT:    store i32 [[TMP17]], i32* [[DOTCAPTURE_EXPR_19]], align 4
+// CHECK4-IRBUILDER-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_19]], align 4
+// CHECK4-IRBUILDER-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_18]], align 4
+// CHECK4-IRBUILDER-NEXT:    [[SUB21:%.*]] = sub i32 [[TMP18]], [[TMP19]]
+// CHECK4-IRBUILDER-NEXT:    [[SUB22:%.*]] = sub i32 [[SUB21]], 1
+// CHECK4-IRBUILDER-NEXT:    [[ADD23:%.*]] = add i32 [[SUB22]], 1
+// CHECK4-IRBUILDER-NEXT:    [[DIV24:%.*]] = udiv i32 [[ADD23]], 1
+// CHECK4-IRBUILDER-NEXT:    [[SUB25:%.*]] = sub i32 [[DIV24]], 1
+// CHECK4-IRBUILDER-NEXT:    store i32 [[SUB25]], i32* [[DOTCAPTURE_EXPR_20]], align 4
+// CHECK4-IRBUILDER-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_18]], align 4
+// CHECK4-IRBUILDER-NEXT:    store i32 [[TMP20]], i32* [[I26]], align 4
+// CHECK4-IRBUILDER-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_18]], align 4
+// CHECK4-IRBUILDER-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_19]], align 4
+// CHECK4-IRBUILDER-NEXT:    [[CMP27:%.*]] = icmp slt i32 [[TMP21]], [[TMP22]]
+// CHECK4-IRBUILDER-NEXT:    br i1 [[CMP27]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
+// CHECK4-IRBUILDER:       omp.precond.then:
+// CHECK4-IRBUILDER-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
+// CHECK4-IRBUILDER-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_20]], align 4
+// CHECK4-IRBUILDER-NEXT:    store i32 [[TMP23]], i32* [[DOTOMP_UB]], align 4
+// CHECK4-IRBUILDER-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
+// CHECK4-IRBUILDER-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
+// CHECK4-IRBUILDER-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_20]], align 4
+// CHECK4-IRBUILDER-NEXT:    [[OMP_GLOBAL_THREAD_NUM:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB12:[0-9]+]])
+// CHECK4-IRBUILDER-NEXT:    call void @__kmpc_dispatch_init_4u(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM]], i32 66, i32 0, i32 [[TMP24]], i32 1, i32 1)
+// CHECK4-IRBUILDER-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
+// CHECK4-IRBUILDER:       omp.dispatch.cond:
+// CHECK4-IRBUILDER-NEXT:    [[OMP_GLOBAL_THREAD_NUM29:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB12]])
+// CHECK4-IRBUILDER-NEXT:    [[TMP25:%.*]] = call i32 @__kmpc_dispatch_next_4u(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM29]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
+// CHECK4-IRBUILDER-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP25]], 0
+// CHECK4-IRBUILDER-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
+// CHECK4-IRBUILDER:       omp.dispatch.body:
+// CHECK4-IRBUILDER-NEXT:    [[TMP26:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
+// CHECK4-IRBUILDER-NEXT:    store i32 [[TMP26]], i32* [[DOTOMP_IV16]], align 4
+// CHECK4-IRBUILDER-NEXT:    br label [[OMP_INNER_FOR_COND30:%.*]]
+// CHECK4-IRBUILDER:       omp.inner.for.cond30:
+// CHECK4-IRBUILDER-NEXT:    [[TMP27:%.*]] = load i32, i32* [[DOTOMP_IV16]], align 4, !llvm.access.group !7
+// CHECK4-IRBUILDER-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !7
+// CHECK4-IRBUILDER-NEXT:    [[ADD31:%.*]] = add i32 [[TMP28]], 1
+// CHECK4-IRBUILDER-NEXT:    [[CMP32:%.*]] = icmp ult i32 [[TMP27]], [[ADD31]]
+// CHECK4-IRBUILDER-NEXT:    br i1 [[CMP32]], label [[OMP_INNER_FOR_BODY33:%.*]], label [[OMP_INNER_FOR_END42:%.*]]
+// CHECK4-IRBUILDER:       omp.inner.for.body33:
+// CHECK4-IRBUILDER-NEXT:    [[TMP29:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_18]], align 4, !llvm.access.group !7
+// CHECK4-IRBUILDER-NEXT:    [[TMP30:%.*]] = load i32, i32* [[DOTOMP_IV16]], align 4, !llvm.access.group !7
+// CHECK4-IRBUILDER-NEXT:    [[MUL34:%.*]] = mul i32 [[TMP30]], 1
+// CHECK4-IRBUILDER-NEXT:    [[ADD35:%.*]] = add i32 [[TMP29]], [[MUL34]]
+// CHECK4-IRBUILDER-NEXT:    store i32 [[ADD35]], i32* [[I28]], align 4, !llvm.access.group !7
+// CHECK4-IRBUILDER-NEXT:    [[TMP31:%.*]] = load i32, i32* [[I28]], align 4, !llvm.access.group !7
+// CHECK4-IRBUILDER-NEXT:    [[IDXPROM36:%.*]] = sext i32 [[TMP31]] to i64
+// CHECK4-IRBUILDER-NEXT:    [[ARRAYIDX37:%.*]] = getelementptr inbounds [10 x float], [10 x float]* @f, i64 0, i64 [[IDXPROM36]]
+// CHECK4-IRBUILDER-NEXT:    store float 0.000000e+00, float* [[ARRAYIDX37]], align 4, !llvm.access.group !7
+// CHECK4-IRBUILDER-NEXT:    call void @__captured_stmt.1(i32* [[I28]]), !llvm.access.group !7
+// CHECK4-IRBUILDER-NEXT:    br label [[OMP_INNER_FOR_BODY33_ORDERED_AFTER:%.*]]
+// CHECK4-IRBUILDER:       omp.inner.for.body33.ordered.after:
+// CHECK4-IRBUILDER-NEXT:    br label [[OMP_BODY_CONTINUE38:%.*]]
+// CHECK4-IRBUILDER:       omp.body.continue38:
+// CHECK4-IRBUILDER-NEXT:    br label [[OMP_INNER_FOR_INC39:%.*]]
+// CHECK4-IRBUILDER:       omp.inner.for.inc39:
+// CHECK4-IRBUILDER-NEXT:    [[TMP32:%.*]] = load i32, i32* [[DOTOMP_IV16]], align 4, !llvm.access.group !7
+// CHECK4-IRBUILDER-NEXT:    [[ADD40:%.*]] = add i32 [[TMP32]], 1
+// CHECK4-IRBUILDER-NEXT:    store i32 [[ADD40]], i32* [[DOTOMP_IV16]], align 4, !llvm.access.group !7
+// CHECK4-IRBUILDER-NEXT:    [[OMP_GLOBAL_THREAD_NUM41:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB12]])
+// CHECK4-IRBUILDER-NEXT:    call void @__kmpc_dispatch_fini_4u(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM41]]), !llvm.access.group !7
+// CHECK4-IRBUILDER-NEXT:    br label [[OMP_INNER_FOR_COND30]], !llvm.loop [[LOOP8:![0-9]+]]
+// CHECK4-IRBUILDER:       omp.inner.for.end42:
+// CHECK4-IRBUILDER-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
+// CHECK4-IRBUILDER:       omp.dispatch.inc:
+// CHECK4-IRBUILDER-NEXT:    br label [[OMP_DISPATCH_COND]]
+// CHECK4-IRBUILDER:       omp.dispatch.end:
+// CHECK4-IRBUILDER-NEXT:    [[TMP33:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
+// CHECK4-IRBUILDER-NEXT:    [[TMP34:%.*]] = icmp ne i32 [[TMP33]], 0
+// CHECK4-IRBUILDER-NEXT:    br i1 [[TMP34]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
+// CHECK4-IRBUILDER:       .omp.final.then:
+// CHECK4-IRBUILDER-NEXT:    [[TMP35:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_18]], align 4
+// CHECK4-IRBUILDER-NEXT:    [[TMP36:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_19]], align 4
+// CHECK4-IRBUILDER-NEXT:    [[TMP37:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_18]], align 4
+// CHECK4-IRBUILDER-NEXT:    [[SUB43:%.*]] = sub i32 [[TMP36]], [[TMP37]]
+// CHECK4-IRBUILDER-NEXT:    [[SUB44:%.*]] = sub i32 [[SUB43]], 1
+// CHECK4-IRBUILDER-NEXT:    [[ADD45:%.*]] = add i32 [[SUB44]], 1
+// CHECK4-IRBUILDER-NEXT:    [[DIV46:%.*]] = udiv i32 [[ADD45]], 1
+// CHECK4-IRBUILDER-NEXT:    [[MUL47:%.*]] = mul i32 [[DIV46]], 1
+// CHECK4-IRBUILDER-NEXT:    [[ADD48:%.*]] = add i32 [[TMP35]], [[MUL47]]
+// CHECK4-IRBUILDER-NEXT:    store i32 [[ADD48]], i32* [[I28]], align 4
+// CHECK4-IRBUILDER-NEXT:    br label [[DOTOMP_FINAL_DONE]]
+// CHECK4-IRBUILDER:       .omp.final.done:
+// CHECK4-IRBUILDER-NEXT:    br label [[OMP_PRECOND_END]]
+// CHECK4-IRBUILDER:       omp.precond.end:
+// CHECK4-IRBUILDER-NEXT:    [[OMP_GLOBAL_THREAD_NUM49:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
+// CHECK4-IRBUILDER-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB4]], i32 [[OMP_GLOBAL_THREAD_NUM49]])
+// CHECK4-IRBUILDER-NEXT:    ret void
+//
+//
+// CHECK4-IRBUILDER-LABEL: define {{[^@]+}}@__captured_stmt
+// CHECK4-IRBUILDER-SAME: (i32* noundef nonnull align 4 dereferenceable(4) [[I:%.*]]) #[[ATTR3:[0-9]+]] {
+// CHECK4-IRBUILDER-NEXT:  entry:
+// CHECK4-IRBUILDER-NEXT:    [[I_ADDR:%.*]] = alloca i32*, align 8
+// CHECK4-IRBUILDER-NEXT:    store i32* [[I]], i32** [[I_ADDR]], align 8
+// CHECK4-IRBUILDER-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[I_ADDR]], align 8
+// CHECK4-IRBUILDER-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
+// CHECK4-IRBUILDER-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP1]] to i64
+// CHECK4-IRBUILDER-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x float], [10 x float]* @f, i64 0, i64 [[IDXPROM]]
+// CHECK4-IRBUILDER-NEXT:    store float 1.000000e+00, float* [[ARRAYIDX]], align 4
+// CHECK4-IRBUILDER-NEXT:    ret void
+//
+//
+// CHECK4-IRBUILDER-LABEL: define {{[^@]+}}@__captured_stmt.1
+// CHECK4-IRBUILDER-SAME: (i32* noundef nonnull align 4 dereferenceable(4) [[I:%.*]]) #[[ATTR3]] {
+// CHECK4-IRBUILDER-NEXT:  entry:
+// CHECK4-IRBUILDER-NEXT:    [[I_ADDR:%.*]] = alloca i32*, align 8
+// CHECK4-IRBUILDER-NEXT:    store i32* [[I]], i32** [[I_ADDR]], align 8
+// CHECK4-IRBUILDER-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[I_ADDR]], align 8
+// CHECK4-IRBUILDER-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
+// CHECK4-IRBUILDER-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP1]] to i64
+// CHECK4-IRBUILDER-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x float], [10 x float]* @f, i64 0, i64 [[IDXPROM]]
+// CHECK4-IRBUILDER-NEXT:    store float 1.000000e+00, float* [[ARRAYIDX]], align 4
+// CHECK4-IRBUILDER-NEXT:    ret void
+//
+//
 // CHECK5-LABEL: define {{[^@]+}}@_Z18static_not_chunkedPfS_S_S_
 // CHECK5-SAME: (float* noundef [[A:%.*]], float* noundef [[B:%.*]], float* noundef [[C:%.*]], float* noundef [[D:%.*]]) #[[ATTR0:[0-9]+]] {
 // CHECK5-NEXT:  entry:
 // CHECK5-NEXT:    [[A_ADDR:%.*]] = alloca float*, align 8
 // CHECK5-NEXT:    [[B_ADDR:%.*]] = alloca float*, align 8
 // CHECK5-NEXT:    [[C_ADDR:%.*]] = alloca float*, align 8
 // CHECK5-NEXT:    [[D_ADDR:%.*]] = alloca float*, align 8
 // CHECK5-NEXT:    [[I:%.*]] = alloca i32, align 4

llvm/include/llvm/Frontend/OpenMP/OMPIRBuilder.h

 #include "llvm/IR/DebugLoc.h"
 #include "llvm/IR/IRBuilder.h"
 #include "llvm/Support/Allocator.h"
 #include <forward_list>
 
 namespace llvm {
 class CanonicalLoopInfo;
 
+/// Move the instruction after an InsertPoint to the beginning of another
+/// BasicBlock.
+///
+/// The instructions after \p IP are moved to the beginning of \p New which must
+/// not have any PHINodes. If \p CreateBranch is true, a branch instruction to
+/// \p New will be added such that there is no semantic change. Otherwise, the
+/// \p IP insert block remains degenerate and it is up to the caller to insert a
+/// terminator.
+void spliceBB(IRBuilderBase::InsertPoint IP, BasicBlock *New,
+              bool CreateBranch);
+
+/// Splice a BasicBlock at an IRBuilder's current insertion point. Its new
+/// insert location will stick to after the instruction before the insertion
+/// point (instead of moving with the instruction the InsertPoint stores
+/// internally).
+void spliceBB(IRBuilder<> &Builder, BasicBlock *New, bool CreateBranch);
+
+/// Split a BasicBlock at an InsertPoint, even if the block is degenerate
+/// (missing the terminator).
+///
+/// llvm::SplitBasicBlock and BasicBlock::splitBasicBlock require a well-formed
+/// BasicBlock. \p Name is used for the new successor block. If \p CreateBranch
+/// is true, a branch to the new successor will new created such that
+/// semantically there is no change; otherwise the block of the insertion point
+/// remains degenerate and it is the caller's responsibility to insert a
+/// terminator. Returns the new successor block.
+BasicBlock *splitBB(IRBuilderBase::InsertPoint IP, bool CreateBranch,
+                    llvm::Twine Name = {});
+
+/// Split a BasicBlock at \p Builder's insertion point, even if the block is
+/// degenerate (missing the terminator).  Its new insert location will stick to
+/// after the instruction before the insertion point (instead of moving with the
+/// instruction the InsertPoint stores internally).
+BasicBlock *splitBB(IRBuilderBase &Builder, bool CreateBranch,
+                    llvm::Twine Name = {});
+
+/// Split a BasicBlock at \p Builder's insertion point, even if the block is
+/// degenerate (missing the terminator).  Its new insert location will stick to
+/// after the instruction before the insertion point (instead of moving with the
+/// instruction the InsertPoint stores internally).
+BasicBlock *splitBB(IRBuilder<> &Builder, bool CreateBranch, llvm::Twine Name);
+
+/// Like splitBB, but reuses the current block's name for the new name.
+BasicBlock *splitBBWithSuffix(IRBuilderBase &Builder, bool CreateBranch,
+                              llvm::Twine Suffix = ".split");
+
 /// An interface to create LLVM-IR for OpenMP directives.
 ///
 /// Each OpenMP directive has a corresponding public generator method.
 class OpenMPIRBuilder {
 public:
   /// Create a new OpenMPIRBuilder operating on the given module \p M. This will
   /// not have an effect on \p M (see initialize).
   OpenMPIRBuilder(Module &M) : M(M), Builder(M.getContext()) {}
public:
   /// Pop the last finalization callback from the finalization stack.
   ///
   /// NOTE: Temporary solution until Clang CG is gone.
   void popFinalizationCB() { FinalizationStack.pop_back(); }
 
   /// Callback type for body (=inner region) code generation
   ///
   /// The callback takes code locations as arguments, each describing a
-  /// location at which code might need to be generated or a location that is
-  /// the target of control transfer.
+  /// location where additional instructions can be inserted.
+  ///
+  /// The CodeGenIP may be in the middle of a basic block or point to the end of
+  /// it. The basic block may have a terminator or be degenerate. The callback
+  /// function may just insert instructions at that position, but also split the
+  /// block (without the Before argument of BasicBlock::splitBasicBlock such
+  /// that the identify of the split predecessor block is preserved) and insert
+  /// additional control flow, including branches that do not lead back to what
+  /// follows the CodeGenIP. Note that since the callback is allowed to split
+  /// the block, callers must assume that InsertPoints to positions in the
+  /// BasicBlock after CodeGenIP including CodeGenIP itself are invalidated. If
+  /// such InsertPoints need to be preserved, it can split the block itself
+  /// before calling the callback.
+  ///
+  /// AllocaIP and CodeGenIP must not point to the same position.
   ///
   /// \param AllocaIP is the insertion point at which new alloca instructions
-  ///                 should be placed.
+  ///                 should be placed. The BasicBlock it is pointing to must
+  ///                 not be split.
   /// \param CodeGenIP is the insertion point at which the body code should be
   ///                  placed.
-  /// \param ContinuationBB is the basic block target to leave the body.
-  ///
-  /// Note that all blocks pointed to by the arguments have terminators.
   using BodyGenCallbackTy =
-      function_ref<void(InsertPointTy AllocaIP, InsertPointTy CodeGenIP,
-                        BasicBlock &ContinuationBB)>;
+      function_ref<void(InsertPointTy AllocaIP, InsertPointTy CodeGenIP)>;
 
   // This is created primarily for sections construct as llvm::function_ref
   // (BodyGenCallbackTy) is not storable (as described in the comments of
   // function_ref class - function_ref contains non-ownable reference
   // to the callable.
   using StorableBodyGenCallbackTy =
-      std::function<void(InsertPointTy AllocaIP, InsertPointTy CodeGenIP,
-                         BasicBlock &ContinuationBB)>;
+      std::function<void(InsertPointTy AllocaIP, InsertPointTy CodeGenIP)>;
 
   /// Callback type for loop body code generation.
   ///
   /// \param CodeGenIP is the insertion point where the loop's body code must be
   ///                  placed. This will be a dedicated BasicBlock with a
   ///                  conditional branch from the loop condition check and
   ///                  terminated with an unconditional branch to the loop
   ///                  latch.

llvm/lib/Frontend/OpenMP/OMPIRBuilder.cpp

if (Instruction *Term = Source->getTerminator()) {
     Br->setSuccessor(0, Target);
     return;
   }
 
   auto *NewBr = BranchInst::Create(Target, Source);
   NewBr->setDebugLoc(DL);
 }
 
-/// Move the instruction after an InsertPoint to the beginning of another
-/// BasicBlock.
-///
-/// The instructions after \p IP are moved to the beginning of \p New which must
-/// not have any PHINodes. If \p CreateBranch is true, a branch instruction to
-/// \p New will be added such that there is no semantic change. Otherwise, the
-/// \p IP insert block remains degenerate and it is up to the caller to insert a
-/// terminator.
-static void spliceBB(OpenMPIRBuilder::InsertPointTy IP, BasicBlock *New,
-                     bool CreateBranch) {
+void llvm::spliceBB(IRBuilderBase::InsertPoint IP, BasicBlock *New,
   assert(New->getFirstInsertionPt() == New->begin() &&
          "Target BB must not have PHI nodes");
 
   // Move instructions to new block.
   BasicBlock *Old = IP.getBlock();
   New->getInstList().splice(New->begin(), Old->getInstList(), IP.getPoint(),
                             Old->end());
 
   if (CreateBranch)
     BranchInst::Create(New, Old);
 }
 
-/// Splice a BasicBlock at an IRBuilder's current insertion point. Its new
-/// insert location will stick to after the instruction before the insertion
-/// point (instead of moving with the instruction the InsertPoint stores
-/// internally).
-static void spliceBB(IRBuilder<> &Builder, BasicBlock *New, bool CreateBranch) {
+void llvm::spliceBB(IRBuilder<> &Builder, BasicBlock *New, bool CreateBranch) {
   DebugLoc DebugLoc = Builder.getCurrentDebugLocation();
   BasicBlock *Old = Builder.GetInsertBlock();
 
   spliceBB(Builder.saveIP(), New, CreateBranch);
   if (CreateBranch)
     Builder.SetInsertPoint(Old->getTerminator());
   else
     Builder.SetInsertPoint(Old);
 
   // SetInsertPoint also updates the Builder's debug location, but we want to
   // keep the one the Builder was configured to use.
   Builder.SetCurrentDebugLocation(DebugLoc);
 }
 
-/// Split a BasicBlock at an InsertPoint, even if the block is degenerate
-/// (missing the terminator).
-///
-/// llvm::SplitBasicBlock and BasicBlock::splitBasicBlock require a well-formed
-/// BasicBlock. \p Name is used for the new successor block. If \p CreateBranch
-/// is true, a branch to the new successor will new created such that
-/// semantically there is no change; otherwise the block of the insertion point
-/// remains degenerate and it is the caller's responsibility to insert a
-/// terminator. Returns the new successor block.
-static BasicBlock *splitBB(OpenMPIRBuilder::InsertPointTy IP, bool CreateBranch,
-                           llvm::Twine Name = {}) {
+BasicBlock *llvm::splitBB(IRBuilderBase::InsertPoint IP, bool CreateBranch,
+                          llvm::Twine Name) {
   BasicBlock *Old = IP.getBlock();
   BasicBlock *New = BasicBlock::Create(
       Old->getContext(), Name.isTriviallyEmpty() ? Old->getName() : Name,
       Old->getParent(), Old->getNextNode());
   spliceBB(IP, New, CreateBranch);
   New->replaceSuccessorsPhiUsesWith(Old, New);
   return New;
 }
 
-/// Split a BasicBlock at \p Builder's insertion point, even if the block is
-/// degenerate (missing the terminator).  Its new insert location will stick to
-/// after the instruction before the insertion point (instead of moving with the
-/// instruction the InsertPoint stores internally).
-static BasicBlock *splitBB(IRBuilder<> &Builder, bool CreateBranch,
-                           llvm::Twine Name = {}) {
+BasicBlock *llvm::splitBB(IRBuilderBase &Builder, bool CreateBranch,
+                          llvm::Twine Name) {
   DebugLoc DebugLoc = Builder.getCurrentDebugLocation();
   BasicBlock *New = splitBB(Builder.saveIP(), CreateBranch, Name);
   if (CreateBranch)
     Builder.SetInsertPoint(Builder.GetInsertBlock()->getTerminator());
   else
     Builder.SetInsertPoint(Builder.GetInsertBlock());
   // SetInsertPoint also updates the Builder's debug location, but we want to
   // keep the one the Builder was configured to use.
   Builder.SetCurrentDebugLocation(DebugLoc);
   return New;
 }
 
+BasicBlock *llvm::splitBB(IRBuilder<> &Builder, bool CreateBranch,
+                          llvm::Twine Name) {
+  DebugLoc DebugLoc = Builder.getCurrentDebugLocation();
+  BasicBlock *New = splitBB(Builder.saveIP(), CreateBranch, Name);
+  if (CreateBranch)
+    Builder.SetInsertPoint(Builder.GetInsertBlock()->getTerminator());
+  else
+    Builder.SetInsertPoint(Builder.GetInsertBlock());
+  // SetInsertPoint also updates the Builder's debug location, but we want to
+  // keep the one the Builder was configured to use.
+  Builder.SetCurrentDebugLocation(DebugLoc);
+  return New;
+}
+
+BasicBlock *llvm::splitBBWithSuffix(IRBuilderBase &Builder, bool CreateBranch,
+                                    llvm::Twine Suffix) {
+  BasicBlock *Old = Builder.GetInsertBlock();
+  return splitBB(Builder, CreateBranch, Old->getName() + Suffix);
+}
+
 void OpenMPIRBuilder::addAttributes(omp::RuntimeFunction FnID, Function &Fn) {
   LLVMContext &Ctx = Fn.getContext();
 
   // Get the function's current attributes.
   auto Attrs = Fn.getAttributes();
   auto FnAttrs = Attrs.getFnAttrs();
   auto RetAttrs = Attrs.getRetAttrs();
   SmallVector<AttributeSet, 4> ArgAttrs;
IRBuilder<>::InsertPoint OpenMPIRBuilder::createParallel(
   // PRegionExitBB          <- A common exit to simplify block collection.
   //
 
   LLVM_DEBUG(dbgs() << "Before body codegen: " << *OuterFn << "\n");
 
   // Let the caller create the body.
   assert(BodyGenCB && "Expected body generation callback!");
   InsertPointTy CodeGenIP(PRegBodyBB, PRegBodyBB->begin());
-  BodyGenCB(InnerAllocaIP, CodeGenIP, *PRegPreFiniBB);
+  BodyGenCB(InnerAllocaIP, CodeGenIP);
 
   LLVM_DEBUG(dbgs() << "After  body codegen: " << *OuterFn << "\n");
 
   FunctionCallee RTLFn = getOrCreateRuntimeFunctionPtr(OMPRTL___kmpc_fork_call);
   if (auto *F = dyn_cast<llvm::Function>(RTLFn.getCallee())) {
     if (!F->hasMetadata(llvm::LLVMContext::MD_callback)) {
       llvm::LLVMContext &Ctx = F->getContext();
       MDBuilder MDB(Ctx);
OpenMPIRBuilder::InsertPointTy OpenMPIRBuilder::createSections(
   //   case <NumSection> - 1:
   //     <SectionStmt[<NumSection> - 1]>;
   //     break;
   // }
   // ...
   // section_loop.after:
   // <FiniCB>;
   auto LoopBodyGenCB = [&](InsertPointTy CodeGenIP, Value *IndVar) {
-    auto *CurFn = CodeGenIP.getBlock()->getParent();
-    auto *ForIncBB = CodeGenIP.getBlock()->getSingleSuccessor();
-    auto *ForExitBB = CodeGenIP.getBlock()
-                          ->getSinglePredecessor()
-                          ->getTerminator()
-                          ->getSuccessor(1);
-    SwitchInst *SwitchStmt = Builder.CreateSwitch(IndVar, ForIncBB);
     Builder.restoreIP(CodeGenIP);
+    BasicBlock *Continue =
+        splitBBWithSuffix(Builder, /*CreateBranch=*/false, ".sections.after");
+    Function *CurFn = Continue->getParent();
+    SwitchInst *SwitchStmt = Builder.CreateSwitch(IndVar, Continue);
+
     unsigned CaseNumber = 0;
     for (auto SectionCB : SectionCBs) {
-      auto *CaseBB = BasicBlock::Create(M.getContext(),
-                                        "omp_section_loop.body.case", CurFn);
+      BasicBlock *CaseBB = BasicBlock::Create(
+          M.getContext(), "omp_section_loop.body.case", CurFn, Continue);
       SwitchStmt->addCase(Builder.getInt32(CaseNumber), CaseBB);
       Builder.SetInsertPoint(CaseBB);
-      SectionCB(InsertPointTy(), Builder.saveIP(), *ForExitBB);
+      BranchInst *CaseEndBr = Builder.CreateBr(Continue);
+      SectionCB(InsertPointTy(),
+                {CaseEndBr->getParent(), CaseEndBr->getIterator()});
       CaseNumber++;
     }
     // remove the existing terminator from body BB since there can be no
     // terminators after switch/case
-    CodeGenIP.getBlock()->getTerminator()->eraseFromParent();
   };
   // Loop body ends here
   // LowerBound, UpperBound, and STride for createCanonicalLoop
   Type *I32Ty = Type::getInt32Ty(M.getContext());
   Value *LB = ConstantInt::get(I32Ty, 0);
   Value *UB = ConstantInt::get(I32Ty, SectionCBs.size());
   Value *ST = ConstantInt::get(I32Ty, 1);
   llvm::CanonicalLoopInfo *LoopInfo = createCanonicalLoop(
       Loc, LoopBodyGenCB, LB, UB, ST, true, false, AllocaIP, "section_loop");
-  Builder.SetInsertPoint(AllocaIP.getBlock()->getTerminator());
-  AllocaIP = Builder.saveIP();
   InsertPointTy AfterIP =
       applyStaticWorkshareLoop(Loc.DL, LoopInfo, AllocaIP, !IsNowait);
-  BasicBlock *LoopAfterBB = AfterIP.getBlock();
-  Instruction *SplitPos = LoopAfterBB->getTerminator();
-  if (!isa_and_nonnull<BranchInst>(SplitPos))
-    SplitPos = new UnreachableInst(Builder.getContext(), LoopAfterBB);
-  // ExitBB after LoopAfterBB because LoopAfterBB is used for FinalizationCB,
-  // which requires a BB with branch
-  BasicBlock *ExitBB =
-      LoopAfterBB->splitBasicBlock(SplitPos, "omp_sections.end");
-  SplitPos->eraseFromParent();
 
   // Apply the finalization callback in LoopAfterBB
   auto FiniInfo = FinalizationStack.pop_back_val();
   assert(FiniInfo.DK == OMPD_sections &&
          "Unexpected finalization stack state!");
-  Builder.SetInsertPoint(LoopAfterBB->getTerminator());
-  FiniInfo.FiniCB(Builder.saveIP());
-  Builder.SetInsertPoint(ExitBB);
+  if (FinalizeCallbackTy &CB = FiniInfo.FiniCB) {
+    Builder.restoreIP(AfterIP);
+    BasicBlock *FiniBB =
+        splitBBWithSuffix(Builder, /*CreateBranch=*/true, "sections.fini");
+    CB(Builder.saveIP());
+    AfterIP = {FiniBB, FiniBB->begin()};
+  }
 
-  return Builder.saveIP();
+  return AfterIP;
 }
 
 OpenMPIRBuilder::InsertPointTy
 OpenMPIRBuilder::createSection(const LocationDescription &Loc,
                                BodyGenCallbackTy BodyGenCB,
                                FinalizeCallbackTy FiniCB) {
   if (!updateToLocation(Loc))
     return Loc.IP;
OpenMPIRBuilder::InsertPointTy OpenMPIRBuilder::EmitOMPInlinedRegion(
   BasicBlock *FiniBB =
       EntryBB->splitBasicBlock(EntryBB->getTerminator(), "omp_region.finalize");
 
   Builder.SetInsertPoint(EntryBB->getTerminator());
   emitCommonDirectiveEntry(OMPD, EntryCall, ExitBB, Conditional);
 
   // generate body
   BodyGenCB(/* AllocaIP */ InsertPointTy(),
-            /* CodeGenIP */ Builder.saveIP(), *FiniBB);
+            /* CodeGenIP */ Builder.saveIP());
 
-  // If we didn't emit a branch to FiniBB during body generation, it means
-  // FiniBB is unreachable (e.g. while(1);). stop generating all the
-  // unreachable blocks, and remove anything we are not going to use.
-  auto SkipEmittingRegion = FiniBB->hasNPredecessors(0);
-  if (SkipEmittingRegion) {
-    FiniBB->eraseFromParent();
-    ExitCall->eraseFromParent();
-    // Discard finalization if we have it.
-    if (HasFinalize) {
-      assert(!FinalizationStack.empty() &&
-             "Unexpected finalization stack state!");
-      FinalizationStack.pop_back();
-    }
-  } else {
-    // emit exit call and do any needed finalization.
-    auto FinIP = InsertPointTy(FiniBB, FiniBB->getFirstInsertionPt());
-    assert(FiniBB->getTerminator()->getNumSuccessors() == 1 &&
-           FiniBB->getTerminator()->getSuccessor(0) == ExitBB &&
-           "Unexpected control flow graph state!!");
-    emitCommonDirectiveExit(OMPD, FinIP, ExitCall, HasFinalize);
-    assert(FiniBB->getUniquePredecessor()->getUniqueSuccessor() == FiniBB &&
-           "Unexpected Control Flow State!");
-    MergeBlockIntoPredecessor(FiniBB);
-  }
 
   // If we are skipping the region of a non conditional, remove the exit
   // block, and clear the builder's insertion point.
   assert(SplitPos->getParent() == ExitBB &&
          "Unexpected Insertion point location!");
-  if (!Conditional && SkipEmittingRegion) {
-    ExitBB->eraseFromParent();
-    Builder.ClearInsertionPoint();
-  } else {
-    auto merged = MergeBlockIntoPredecessor(ExitBB);
-    BasicBlock *ExitPredBB = SplitPos->getParent();
-    auto InsertBB = merged ? ExitPredBB : ExitBB;
-    if (!isa_and_nonnull<BranchInst>(SplitPos))
-      SplitPos->eraseFromParent();
-    Builder.SetInsertPoint(InsertBB);
-  }
 
   return Builder.saveIP();
 }
 
 OpenMPIRBuilder::InsertPointTy OpenMPIRBuilder::emitCommonDirectiveEntry(
     Directive OMPD, Value *EntryCall, BasicBlock *ExitBB, bool Conditional) {
   // if nothing to do, Return current insertion point.
   if (!Conditional || !EntryCall)

llvm/lib/Transforms/IPO/OpenMPOpt.cpp

bool mergeParallelRegions() {
 
     bool Changed = false;
     LoopInfo *LI = nullptr;
     DominatorTree *DT = nullptr;
 
     SmallDenseMap<BasicBlock *, SmallPtrSet<Instruction *, 4>> BB2PRMap;
 
     BasicBlock *StartBB = nullptr, *EndBB = nullptr;
-    auto BodyGenCB = [&](InsertPointTy AllocaIP, InsertPointTy CodeGenIP,
-                         BasicBlock &ContinuationIP) {
+    auto BodyGenCB = [&](InsertPointTy AllocaIP, InsertPointTy CodeGenIP) {
       BasicBlock *CGStartBB = CodeGenIP.getBlock();
       BasicBlock *CGEndBB =
           SplitBlock(CGStartBB, &*CodeGenIP.getPoint(), DT, LI);
       assert(StartBB != nullptr && "StartBB should not be null");
       CGStartBB->getTerminator()->setSuccessor(0, StartBB);
       assert(EndBB != nullptr && "EndBB should not be null");
       EndBB->getTerminator()->setSuccessor(0, CGEndBB);
     };
auto CreateSequentialRegion = [&](Function *OuterFn,
       BasicBlock *SeqStartBB =
           SplitBlock(ParentBB, SeqStartI, DT, LI, nullptr, "seq.par.merged");
 
       assert(ParentBB->getUniqueSuccessor() == SeqStartBB &&
              "Expected a different CFG");
       const DebugLoc DL = ParentBB->getTerminator()->getDebugLoc();
       ParentBB->getTerminator()->eraseFromParent();
 
-      auto BodyGenCB = [&](InsertPointTy AllocaIP, InsertPointTy CodeGenIP,
-                           BasicBlock &ContinuationIP) {
+      auto BodyGenCB = [&](InsertPointTy AllocaIP, InsertPointTy CodeGenIP) {
         BasicBlock *CGStartBB = CodeGenIP.getBlock();
         BasicBlock *CGEndBB =
             SplitBlock(CGStartBB, &*CodeGenIP.getPoint(), DT, LI);
         assert(SeqStartBB != nullptr && "SeqStartBB should not be null");
         CGStartBB->getTerminator()->setSuccessor(0, SeqStartBB);
         assert(SeqEndBB != nullptr && "SeqEndBB should not be null");
         SeqEndBB->getTerminator()->setSuccessor(0, CGEndBB);
       };

llvm/unittests/Frontend/OpenMPIRBuilderTest.cpp

TEST_F(OpenMPIRBuilderTest, ParallelSimple) {
   OpenMPIRBuilder::LocationDescription Loc({Builder.saveIP(), DL});
 
   AllocaInst *PrivAI = nullptr;
 
   unsigned NumBodiesGenerated = 0;
   unsigned NumPrivatizedVars = 0;
   unsigned NumFinalizationPoints = 0;
 
-  auto BodyGenCB = [&](InsertPointTy AllocaIP, InsertPointTy CodeGenIP,
-                       BasicBlock &ContinuationIP) {
+  auto BodyGenCB = [&](InsertPointTy AllocaIP, InsertPointTy CodeGenIP) {
     ++NumBodiesGenerated;
 
     Builder.restoreIP(AllocaIP);
     PrivAI = Builder.CreateAlloca(F->arg_begin()->getType());
     Builder.CreateStore(F->arg_begin(), PrivAI);
 
     Builder.restoreIP(CodeGenIP);
     Value *PrivLoad =
         Builder.CreateLoad(PrivAI->getAllocatedType(), PrivAI, "local.use");
     Value *Cmp = Builder.CreateICmpNE(F->arg_begin(), PrivLoad);
     Instruction *ThenTerm, *ElseTerm;
     SplitBlockAndInsertIfThenElse(Cmp, CodeGenIP.getBlock()->getTerminator(),
                                   &ThenTerm, &ElseTerm);
-
-    Builder.SetInsertPoint(ThenTerm);
-    Builder.CreateBr(&ContinuationIP);
-    ThenTerm->eraseFromParent();
   };
 
   auto PrivCB = [&](InsertPointTy AllocaIP, InsertPointTy CodeGenIP,
                     Value &Orig, Value &Inner,
                     Value *&ReplacementValue) -> InsertPointTy {
     ++NumPrivatizedVars;
 
     if (!isa<AllocaInst>(Orig)) {
TEST_F(OpenMPIRBuilderTest, ParallelNested) {
   Builder.CreateBr(EnterBB);
   Builder.SetInsertPoint(EnterBB);
   OpenMPIRBuilder::LocationDescription Loc({Builder.saveIP(), DL});
 
   unsigned NumInnerBodiesGenerated = 0;
   unsigned NumOuterBodiesGenerated = 0;
   unsigned NumFinalizationPoints = 0;
 
-  auto InnerBodyGenCB = [&](InsertPointTy AllocaIP, InsertPointTy CodeGenIP,
-                            BasicBlock &ContinuationIP) {
+  auto InnerBodyGenCB = [&](InsertPointTy AllocaIP, InsertPointTy CodeGenIP) {
     ++NumInnerBodiesGenerated;
   };
 
   auto PrivCB = [&](InsertPointTy AllocaIP, InsertPointTy CodeGenIP,
                     Value &Orig, Value &Inner,
                     Value *&ReplacementValue) -> InsertPointTy {
     // Trivial copy (=firstprivate).
     Builder.restoreIP(AllocaIP);
     Type *VTy = ReplacementValue->getType();
     Value *V = Builder.CreateLoad(VTy, &Inner, Orig.getName() + ".reload");
     ReplacementValue = Builder.CreateAlloca(VTy, 0, Orig.getName() + ".copy");
     Builder.restoreIP(CodeGenIP);
     Builder.CreateStore(V, ReplacementValue);
     return CodeGenIP;
   };
 
   auto FiniCB = [&](InsertPointTy CodeGenIP) { ++NumFinalizationPoints; };
 
-  auto OuterBodyGenCB = [&](InsertPointTy AllocaIP, InsertPointTy CodeGenIP,
-                            BasicBlock &ContinuationIP) {
+  auto OuterBodyGenCB = [&](InsertPointTy AllocaIP, InsertPointTy CodeGenIP) {
     ++NumOuterBodiesGenerated;
     Builder.restoreIP(CodeGenIP);
     BasicBlock *CGBB = CodeGenIP.getBlock();
     BasicBlock *NewBB = SplitBlock(CGBB, &*CodeGenIP.getPoint());
     CGBB->getTerminator()->eraseFromParent();
     ;
 
     IRBuilder<>::InsertPoint AfterIP = OMPBuilder.createParallel(
TEST_F(OpenMPIRBuilderTest, ParallelNested2Inner) {
   Builder.CreateBr(EnterBB);
   Builder.SetInsertPoint(EnterBB);
   OpenMPIRBuilder::LocationDescription Loc({Builder.saveIP(), DL});
 
   unsigned NumInnerBodiesGenerated = 0;
   unsigned NumOuterBodiesGenerated = 0;
   unsigned NumFinalizationPoints = 0;
 
-  auto InnerBodyGenCB = [&](InsertPointTy AllocaIP, InsertPointTy CodeGenIP,
-                            BasicBlock &ContinuationIP) {
+  auto InnerBodyGenCB = [&](InsertPointTy AllocaIP, InsertPointTy CodeGenIP) {
     ++NumInnerBodiesGenerated;
   };
 
   auto PrivCB = [&](InsertPointTy AllocaIP, InsertPointTy CodeGenIP,
                     Value &Orig, Value &Inner,
                     Value *&ReplacementValue) -> InsertPointTy {
     // Trivial copy (=firstprivate).
     Builder.restoreIP(AllocaIP);
     Type *VTy = ReplacementValue->getType();
     Value *V = Builder.CreateLoad(VTy, &Inner, Orig.getName() + ".reload");
     ReplacementValue = Builder.CreateAlloca(VTy, 0, Orig.getName() + ".copy");
     Builder.restoreIP(CodeGenIP);
     Builder.CreateStore(V, ReplacementValue);
     return CodeGenIP;
   };
 
   auto FiniCB = [&](InsertPointTy CodeGenIP) { ++NumFinalizationPoints; };
 
-  auto OuterBodyGenCB = [&](InsertPointTy AllocaIP, InsertPointTy CodeGenIP,
-                            BasicBlock &ContinuationIP) {
+  auto OuterBodyGenCB = [&](InsertPointTy AllocaIP, InsertPointTy CodeGenIP) {
     ++NumOuterBodiesGenerated;
     Builder.restoreIP(CodeGenIP);
     BasicBlock *CGBB = CodeGenIP.getBlock();
     BasicBlock *NewBB1 = SplitBlock(CGBB, &*CodeGenIP.getPoint());
     BasicBlock *NewBB2 = SplitBlock(NewBB1, &*NewBB1->getFirstInsertionPt());
     CGBB->getTerminator()->eraseFromParent();
     ;
     NewBB1->getTerminator()->eraseFromParent();
TEST_F(OpenMPIRBuilderTest, ParallelIfCond) {
   OpenMPIRBuilder::LocationDescription Loc({Builder.saveIP(), DL});
 
   AllocaInst *PrivAI = nullptr;
 
   unsigned NumBodiesGenerated = 0;
   unsigned NumPrivatizedVars = 0;
   unsigned NumFinalizationPoints = 0;
 
-  auto BodyGenCB = [&](InsertPointTy AllocaIP, InsertPointTy CodeGenIP,
-                       BasicBlock &ContinuationIP) {
+  auto BodyGenCB = [&](InsertPointTy AllocaIP, InsertPointTy CodeGenIP) {
     ++NumBodiesGenerated;
 
     Builder.restoreIP(AllocaIP);
     PrivAI = Builder.CreateAlloca(F->arg_begin()->getType());
     Builder.CreateStore(F->arg_begin(), PrivAI);
 
     Builder.restoreIP(CodeGenIP);
     Value *PrivLoad =
         Builder.CreateLoad(PrivAI->getAllocatedType(), PrivAI, "local.use");
     Value *Cmp = Builder.CreateICmpNE(F->arg_begin(), PrivLoad);
     Instruction *ThenTerm, *ElseTerm;
-    SplitBlockAndInsertIfThenElse(Cmp, CodeGenIP.getBlock()->getTerminator(),
-                                  &ThenTerm, &ElseTerm);
-
-    Builder.SetInsertPoint(ThenTerm);
-    Builder.CreateBr(&ContinuationIP);
-    ThenTerm->eraseFromParent();
+    SplitBlockAndInsertIfThenElse(Cmp, &*Builder.GetInsertPoint(), &ThenTerm,
+                                  &ElseTerm);
   };
 
   auto PrivCB = [&](InsertPointTy AllocaIP, InsertPointTy CodeGenIP,
                     Value &Orig, Value &Inner,
                     Value *&ReplacementValue) -> InsertPointTy {
     ++NumPrivatizedVars;
 
     if (!isa<AllocaInst>(Orig)) {
TEST_F(OpenMPIRBuilderTest, ParallelCancelBarrier) {
   Builder.SetInsertPoint(EnterBB);
   OpenMPIRBuilder::LocationDescription Loc({Builder.saveIP(), DL});
 
   unsigned NumBodiesGenerated = 0;
   unsigned NumPrivatizedVars = 0;
   unsigned NumFinalizationPoints = 0;
 
   CallInst *CheckedBarrier = nullptr;
-  auto BodyGenCB = [&](InsertPointTy AllocaIP, InsertPointTy CodeGenIP,
-                       BasicBlock &ContinuationIP) {
+  auto BodyGenCB = [&](InsertPointTy AllocaIP, InsertPointTy CodeGenIP) {
     ++NumBodiesGenerated;
 
     Builder.restoreIP(CodeGenIP);
 
     // Create three barriers, two cancel barriers but only one checked.
     Function *CBFn, *BFn;
 
     Builder.restoreIP(
TEST_F(OpenMPIRBuilderTest, ParallelForwardAsPointers) {
   FunctionCallee TakeStructPtrFunc =
       M->getOrInsertFunction("take_structPtr", VoidTy, StructPtrTy);
   Value *I32Val = Builder.CreateCall(RetI32Func);
   Value *I32PtrVal = Builder.CreateCall(RetI32PtrFunc);
   Value *StructVal = Builder.CreateCall(RetStructFunc);
   Value *StructPtrVal = Builder.CreateCall(RetStructPtrFunc);
 
   Instruction *Internal;
-  auto BodyGenCB = [&](InsertPointTy AllocaIP, InsertPointTy CodeGenIP,
-                       BasicBlock &ContinuationBB) {
+  auto BodyGenCB = [&](InsertPointTy AllocaIP, InsertPointTy CodeGenIP) {
     IRBuilder<>::InsertPointGuard Guard(Builder);
     Builder.restoreIP(CodeGenIP);
     Internal = Builder.CreateCall(TakeI32Func, I32Val);
     Builder.CreateCall(TakeI32PtrFunc, I32PtrVal);
     Builder.CreateCall(TakeStructFunc, StructVal);
     Builder.CreateCall(TakeStructPtrFunc, StructPtrVal);
   };
   auto PrivCB = [&](InsertPointTy AllocaIP, InsertPointTy CodeGenIP, Value &,
TEST_F(OpenMPIRBuilderTest, MasterDirective) {
   F->setName("func");
   IRBuilder<> Builder(BB);
 
   OpenMPIRBuilder::LocationDescription Loc({Builder.saveIP(), DL});
 
   AllocaInst *PrivAI = nullptr;
 
   BasicBlock *EntryBB = nullptr;
-  BasicBlock *ExitBB = nullptr;
   BasicBlock *ThenBB = nullptr;
 
-  auto BodyGenCB = [&](InsertPointTy AllocaIP, InsertPointTy CodeGenIP,
-                       BasicBlock &FiniBB) {
+  auto BodyGenCB = [&](InsertPointTy AllocaIP, InsertPointTy CodeGenIP) {
     if (AllocaIP.isSet())
       Builder.restoreIP(AllocaIP);
     else
       Builder.SetInsertPoint(&*(F->getEntryBlock().getFirstInsertionPt()));
     PrivAI = Builder.CreateAlloca(F->arg_begin()->getType());
     Builder.CreateStore(F->arg_begin(), PrivAI);
 
     llvm::BasicBlock *CodeGenIPBB = CodeGenIP.getBlock();
     llvm::Instruction *CodeGenIPInst = &*CodeGenIP.getPoint();
     EXPECT_EQ(CodeGenIPBB->getTerminator(), CodeGenIPInst);
 
     Builder.restoreIP(CodeGenIP);
 
     // collect some info for checks later
-    ExitBB = FiniBB.getUniqueSuccessor();
     ThenBB = Builder.GetInsertBlock();
     EntryBB = ThenBB->getUniquePredecessor();
 
     // simple instructions for body
     Value *PrivLoad =
         Builder.CreateLoad(PrivAI->getAllocatedType(), PrivAI, "local.use");
     Builder.CreateICmpNE(F->arg_begin(), PrivLoad);
   };
 
   auto FiniCB = [&](InsertPointTy IP) {
     BasicBlock *IPBB = IP.getBlock();
     EXPECT_NE(IPBB->end(), IP.getPoint());
   };
 
   Builder.restoreIP(OMPBuilder.createMaster(Builder, BodyGenCB, FiniCB));
   Value *EntryBBTI = EntryBB->getTerminator();
   EXPECT_NE(EntryBBTI, nullptr);
   EXPECT_TRUE(isa<BranchInst>(EntryBBTI));
   BranchInst *EntryBr = cast<BranchInst>(EntryBB->getTerminator());
   EXPECT_TRUE(EntryBr->isConditional());
   EXPECT_EQ(EntryBr->getSuccessor(0), ThenBB);
-  EXPECT_EQ(ThenBB->getUniqueSuccessor(), ExitBB);
+  BasicBlock *ExitBB = ThenBB->getUniqueSuccessor();
   EXPECT_EQ(EntryBr->getSuccessor(1), ExitBB);
 
   CmpInst *CondInst = cast<CmpInst>(EntryBr->getCondition());
   EXPECT_TRUE(isa<CallInst>(CondInst->getOperand(0)));
 
   CallInst *MasterEntryCI = cast<CallInst>(CondInst->getOperand(0));
   EXPECT_EQ(MasterEntryCI->arg_size(), 2U);
   EXPECT_EQ(MasterEntryCI->getCalledFunction()->getName(), "__kmpc_master");
TEST_F(OpenMPIRBuilderTest, MaskedDirective) {
   F->setName("func");
   IRBuilder<> Builder(BB);
 
   OpenMPIRBuilder::LocationDescription Loc({Builder.saveIP(), DL});
 
   AllocaInst *PrivAI = nullptr;
 
   BasicBlock *EntryBB = nullptr;
-  BasicBlock *ExitBB = nullptr;
   BasicBlock *ThenBB = nullptr;
 
-  auto BodyGenCB = [&](InsertPointTy AllocaIP, InsertPointTy CodeGenIP,
-                       BasicBlock &FiniBB) {
+  auto BodyGenCB = [&](InsertPointTy AllocaIP, InsertPointTy CodeGenIP) {
     if (AllocaIP.isSet())
       Builder.restoreIP(AllocaIP);
     else
       Builder.SetInsertPoint(&*(F->getEntryBlock().getFirstInsertionPt()));
     PrivAI = Builder.CreateAlloca(F->arg_begin()->getType());
     Builder.CreateStore(F->arg_begin(), PrivAI);
 
     llvm::BasicBlock *CodeGenIPBB = CodeGenIP.getBlock();
     llvm::Instruction *CodeGenIPInst = &*CodeGenIP.getPoint();
     EXPECT_EQ(CodeGenIPBB->getTerminator(), CodeGenIPInst);
 
     Builder.restoreIP(CodeGenIP);
 
     // collect some info for checks later
-    ExitBB = FiniBB.getUniqueSuccessor();
     ThenBB = Builder.GetInsertBlock();
     EntryBB = ThenBB->getUniquePredecessor();
 
     // simple instructions for body
     Value *PrivLoad =
         Builder.CreateLoad(PrivAI->getAllocatedType(), PrivAI, "local.use");
     Builder.CreateICmpNE(F->arg_begin(), PrivLoad);
   };
 
   auto FiniCB = [&](InsertPointTy IP) {
     BasicBlock *IPBB = IP.getBlock();
     EXPECT_NE(IPBB->end(), IP.getPoint());
   };
 
   Constant *Filter = ConstantInt::get(Type::getInt32Ty(M->getContext()), 0);
   Builder.restoreIP(
       OMPBuilder.createMasked(Builder, BodyGenCB, FiniCB, Filter));
   Value *EntryBBTI = EntryBB->getTerminator();
   EXPECT_NE(EntryBBTI, nullptr);
   EXPECT_TRUE(isa<BranchInst>(EntryBBTI));
   BranchInst *EntryBr = cast<BranchInst>(EntryBB->getTerminator());
   EXPECT_TRUE(EntryBr->isConditional());
   EXPECT_EQ(EntryBr->getSuccessor(0), ThenBB);
-  EXPECT_EQ(ThenBB->getUniqueSuccessor(), ExitBB);
+  BasicBlock *ExitBB = ThenBB->getUniqueSuccessor();
   EXPECT_EQ(EntryBr->getSuccessor(1), ExitBB);
 
   CmpInst *CondInst = cast<CmpInst>(EntryBr->getCondition());
   EXPECT_TRUE(isa<CallInst>(CondInst->getOperand(0)));
 
   CallInst *MaskedEntryCI = cast<CallInst>(CondInst->getOperand(0));
   EXPECT_EQ(MaskedEntryCI->arg_size(), 3U);
   EXPECT_EQ(MaskedEntryCI->getCalledFunction()->getName(), "__kmpc_masked");
TEST_F(OpenMPIRBuilderTest, CriticalDirective) {
   OMPBuilder.initialize();
   F->setName("func");
   IRBuilder<> Builder(BB);
 
   OpenMPIRBuilder::LocationDescription Loc({Builder.saveIP(), DL});
 
   AllocaInst *PrivAI = Builder.CreateAlloca(F->arg_begin()->getType());
 
-  BasicBlock *EntryBB = nullptr;
-
-  auto BodyGenCB = [&](InsertPointTy AllocaIP, InsertPointTy CodeGenIP,
-                       BasicBlock &FiniBB) {
-    // collect some info for checks later
-    EntryBB = FiniBB.getUniquePredecessor();
-
+  auto BodyGenCB = [&](InsertPointTy AllocaIP, InsertPointTy CodeGenIP) {
     // actual start for bodyCB
     llvm::BasicBlock *CodeGenIPBB = CodeGenIP.getBlock();
     llvm::Instruction *CodeGenIPInst = &*CodeGenIP.getPoint();
     EXPECT_EQ(CodeGenIPBB->getTerminator(), CodeGenIPInst);
-    EXPECT_EQ(EntryBB, CodeGenIPBB);
 
     // body begin
     Builder.restoreIP(CodeGenIP);
     Builder.CreateStore(F->arg_begin(), PrivAI);
     Value *PrivLoad =
         Builder.CreateLoad(PrivAI->getAllocatedType(), PrivAI, "local.use");
     Builder.CreateICmpNE(F->arg_begin(), PrivLoad);
   };
 
   auto FiniCB = [&](InsertPointTy IP) {
     BasicBlock *IPBB = IP.getBlock();
     EXPECT_NE(IPBB->end(), IP.getPoint());
   };
+  BasicBlock *EntryBB = Builder.GetInsertBlock();
 
   Builder.restoreIP(OMPBuilder.createCritical(Builder, BodyGenCB, FiniCB,
                                               "testCRT", nullptr));
 
-  Value *EntryBBTI = EntryBB->getTerminator();
-  EXPECT_EQ(EntryBBTI, nullptr);
-
   CallInst *CriticalEntryCI = nullptr;
   for (auto &EI : *EntryBB) {
     Instruction *cur = &EI;
     if (isa<CallInst>(cur)) {
       CriticalEntryCI = cast<CallInst>(cur);
       if (CriticalEntryCI->getCalledFunction()->getName() == "__kmpc_critical")
         break;
       CriticalEntryCI = nullptr;
TEST_F(OpenMPIRBuilderTest, OrderedDirectiveThreads) {
   F->setName("func");
   IRBuilder<> Builder(BB);
 
   OpenMPIRBuilder::LocationDescription Loc({Builder.saveIP(), DL});
 
   AllocaInst *PrivAI =
       Builder.CreateAlloca(F->arg_begin()->getType(), nullptr, "priv.inst");
 
-  BasicBlock *EntryBB = nullptr;
-
-  auto BodyGenCB = [&](InsertPointTy AllocaIP, InsertPointTy CodeGenIP,
-                       BasicBlock &FiniBB) {
-    EntryBB = FiniBB.getUniquePredecessor();
-
+  auto BodyGenCB = [&](InsertPointTy AllocaIP, InsertPointTy CodeGenIP) {
     llvm::BasicBlock *CodeGenIPBB = CodeGenIP.getBlock();
     llvm::Instruction *CodeGenIPInst = &*CodeGenIP.getPoint();
     EXPECT_EQ(CodeGenIPBB->getTerminator(), CodeGenIPInst);
-    EXPECT_EQ(EntryBB, CodeGenIPBB);
 
     Builder.restoreIP(CodeGenIP);
     Builder.CreateStore(F->arg_begin(), PrivAI);
     Value *PrivLoad =
         Builder.CreateLoad(PrivAI->getAllocatedType(), PrivAI, "local.use");
     Builder.CreateICmpNE(F->arg_begin(), PrivLoad);
   };
 
   auto FiniCB = [&](InsertPointTy IP) {
     BasicBlock *IPBB = IP.getBlock();
     EXPECT_NE(IPBB->end(), IP.getPoint());
   };
 
   // Test for "#omp ordered [threads]"
+  BasicBlock *EntryBB = Builder.GetInsertBlock();
   Builder.restoreIP(
       OMPBuilder.createOrderedThreadsSimd(Builder, BodyGenCB, FiniCB, true));
 
   Builder.CreateRetVoid();
   OMPBuilder.finalize();
   EXPECT_FALSE(verifyModule(*M, &errs()));
 
   EXPECT_NE(EntryBB->getTerminator(), nullptr);
TEST_F(OpenMPIRBuilderTest, OrderedDirectiveSimd) {
   F->setName("func");
   IRBuilder<> Builder(BB);
 
   OpenMPIRBuilder::LocationDescription Loc({Builder.saveIP(), DL});
 
   AllocaInst *PrivAI =
       Builder.CreateAlloca(F->arg_begin()->getType(), nullptr, "priv.inst");
 
-  BasicBlock *EntryBB = nullptr;
-
-  auto BodyGenCB = [&](InsertPointTy AllocaIP, InsertPointTy CodeGenIP,
-                       BasicBlock &FiniBB) {
-    EntryBB = FiniBB.getUniquePredecessor();
-
+  auto BodyGenCB = [&](InsertPointTy AllocaIP, InsertPointTy CodeGenIP) {
     llvm::BasicBlock *CodeGenIPBB = CodeGenIP.getBlock();
     llvm::Instruction *CodeGenIPInst = &*CodeGenIP.getPoint();
     EXPECT_EQ(CodeGenIPBB->getTerminator(), CodeGenIPInst);
-    EXPECT_EQ(EntryBB, CodeGenIPBB);
 
     Builder.restoreIP(CodeGenIP);
     Builder.CreateStore(F->arg_begin(), PrivAI);
     Value *PrivLoad =
         Builder.CreateLoad(PrivAI->getAllocatedType(), PrivAI, "local.use");
     Builder.CreateICmpNE(F->arg_begin(), PrivLoad);
   };
 
   auto FiniCB = [&](InsertPointTy IP) {
     BasicBlock *IPBB = IP.getBlock();
     EXPECT_NE(IPBB->end(), IP.getPoint());
   };
 
   // Test for "#omp ordered simd"
+  BasicBlock *EntryBB = Builder.GetInsertBlock();
   Builder.restoreIP(
       OMPBuilder.createOrderedThreadsSimd(Builder, BodyGenCB, FiniCB, false));
 
   Builder.CreateRetVoid();
   OMPBuilder.finalize();
   EXPECT_FALSE(verifyModule(*M, &errs()));
 
   EXPECT_NE(EntryBB->getTerminator(), nullptr);
TEST_F(OpenMPIRBuilderTest, SingleDirective) {
   F->setName("func");
   IRBuilder<> Builder(BB);
 
   OpenMPIRBuilder::LocationDescription Loc({Builder.saveIP(), DL});
 
   AllocaInst *PrivAI = nullptr;
 
   BasicBlock *EntryBB = nullptr;
-  BasicBlock *ExitBB = nullptr;
   BasicBlock *ThenBB = nullptr;
 
-  auto BodyGenCB = [&](InsertPointTy AllocaIP, InsertPointTy CodeGenIP,
-                       BasicBlock &FiniBB) {
+  auto BodyGenCB = [&](InsertPointTy AllocaIP, InsertPointTy CodeGenIP) {
     if (AllocaIP.isSet())
       Builder.restoreIP(AllocaIP);
     else
       Builder.SetInsertPoint(&*(F->getEntryBlock().getFirstInsertionPt()));
     PrivAI = Builder.CreateAlloca(F->arg_begin()->getType());
     Builder.CreateStore(F->arg_begin(), PrivAI);
 
     llvm::BasicBlock *CodeGenIPBB = CodeGenIP.getBlock();
     llvm::Instruction *CodeGenIPInst = &*CodeGenIP.getPoint();
     EXPECT_EQ(CodeGenIPBB->getTerminator(), CodeGenIPInst);
 
     Builder.restoreIP(CodeGenIP);
 
     // collect some info for checks later
-    ExitBB = FiniBB.getUniqueSuccessor();
     ThenBB = Builder.GetInsertBlock();
     EntryBB = ThenBB->getUniquePredecessor();
 
     // simple instructions for body
     Value *PrivLoad =
         Builder.CreateLoad(PrivAI->getAllocatedType(), PrivAI, "local.use");
     Builder.CreateICmpNE(F->arg_begin(), PrivLoad);
   };
 
   auto FiniCB = [&](InsertPointTy IP) {
     BasicBlock *IPBB = IP.getBlock();
     EXPECT_NE(IPBB->end(), IP.getPoint());
   };
 
   Builder.restoreIP(OMPBuilder.createSingle(
       Builder, BodyGenCB, FiniCB, /*IsNowait*/ false, /*DidIt*/ nullptr));
   Value *EntryBBTI = EntryBB->getTerminator();
   EXPECT_NE(EntryBBTI, nullptr);
   EXPECT_TRUE(isa<BranchInst>(EntryBBTI));
   BranchInst *EntryBr = cast<BranchInst>(EntryBB->getTerminator());
   EXPECT_TRUE(EntryBr->isConditional());
   EXPECT_EQ(EntryBr->getSuccessor(0), ThenBB);
-  EXPECT_EQ(ThenBB->getUniqueSuccessor(), ExitBB);
+  BasicBlock *ExitBB = ThenBB->getUniqueSuccessor();
   EXPECT_EQ(EntryBr->getSuccessor(1), ExitBB);
 
   CmpInst *CondInst = cast<CmpInst>(EntryBr->getCondition());
   EXPECT_TRUE(isa<CallInst>(CondInst->getOperand(0)));
 
   CallInst *SingleEntryCI = cast<CallInst>(CondInst->getOperand(0));
   EXPECT_EQ(SingleEntryCI->arg_size(), 2U);
   EXPECT_EQ(SingleEntryCI->getCalledFunction()->getName(), "__kmpc_single");
TEST_F(OpenMPIRBuilderTest, SingleDirectiveNowait) {
   F->setName("func");
   IRBuilder<> Builder(BB);
 
   OpenMPIRBuilder::LocationDescription Loc({Builder.saveIP(), DL});
 
   AllocaInst *PrivAI = nullptr;
 
   BasicBlock *EntryBB = nullptr;
-  BasicBlock *ExitBB = nullptr;
   BasicBlock *ThenBB = nullptr;
 
-  auto BodyGenCB = [&](InsertPointTy AllocaIP, InsertPointTy CodeGenIP,
-                       BasicBlock &FiniBB) {
+  auto BodyGenCB = [&](InsertPointTy AllocaIP, InsertPointTy CodeGenIP) {
     if (AllocaIP.isSet())
       Builder.restoreIP(AllocaIP);
     else
       Builder.SetInsertPoint(&*(F->getEntryBlock().getFirstInsertionPt()));
     PrivAI = Builder.CreateAlloca(F->arg_begin()->getType());
     Builder.CreateStore(F->arg_begin(), PrivAI);
 
     llvm::BasicBlock *CodeGenIPBB = CodeGenIP.getBlock();
     llvm::Instruction *CodeGenIPInst = &*CodeGenIP.getPoint();
     EXPECT_EQ(CodeGenIPBB->getTerminator(), CodeGenIPInst);
 
     Builder.restoreIP(CodeGenIP);
 
     // collect some info for checks later
-    ExitBB = FiniBB.getUniqueSuccessor();
     ThenBB = Builder.GetInsertBlock();
     EntryBB = ThenBB->getUniquePredecessor();
 
     // simple instructions for body
     Value *PrivLoad =
         Builder.CreateLoad(PrivAI->getAllocatedType(), PrivAI, "local.use");
     Builder.CreateICmpNE(F->arg_begin(), PrivLoad);
   };
 
   auto FiniCB = [&](InsertPointTy IP) {
     BasicBlock *IPBB = IP.getBlock();
     EXPECT_NE(IPBB->end(), IP.getPoint());
   };
 
   Builder.restoreIP(OMPBuilder.createSingle(
       Builder, BodyGenCB, FiniCB, /*IsNowait*/ true, /*DidIt*/ nullptr));
   Value *EntryBBTI = EntryBB->getTerminator();
   EXPECT_NE(EntryBBTI, nullptr);
   EXPECT_TRUE(isa<BranchInst>(EntryBBTI));
   BranchInst *EntryBr = cast<BranchInst>(EntryBB->getTerminator());
   EXPECT_TRUE(EntryBr->isConditional());
   EXPECT_EQ(EntryBr->getSuccessor(0), ThenBB);
-  EXPECT_EQ(ThenBB->getUniqueSuccessor(), ExitBB);
+  BasicBlock *ExitBB = ThenBB->getUniqueSuccessor();
   EXPECT_EQ(EntryBr->getSuccessor(1), ExitBB);
 
   CmpInst *CondInst = cast<CmpInst>(EntryBr->getCondition());
   EXPECT_TRUE(isa<CallInst>(CondInst->getOperand(0)));
 
   CallInst *SingleEntryCI = cast<CallInst>(CondInst->getOperand(0));
   EXPECT_EQ(SingleEntryCI->arg_size(), 2U);
   EXPECT_EQ(SingleEntryCI->getCalledFunction()->getName(), "__kmpc_single");
TEST_F(OpenMPIRBuilderTest, CreateReductions) {
   Builder.CreateStore(ConstantFP::get(Builder.getFloatTy(), 0.0), SumReduced);
   Builder.CreateStore(Builder.getInt32(1), XorReduced);
 
   // The loop body computes two reductions:
   //   sum of (float) thread-id;
   //   xor of thread-id;
   // and store the result in global variables.
   InsertPointTy BodyIP, BodyAllocaIP;
-  auto BodyGenCB = [&](InsertPointTy InnerAllocaIP, InsertPointTy CodeGenIP,
-                       BasicBlock &ContinuationBB) {
+  auto BodyGenCB = [&](InsertPointTy InnerAllocaIP, InsertPointTy CodeGenIP) {
     IRBuilderBase::InsertPointGuard Guard(Builder);
     Builder.restoreIP(CodeGenIP);
 
     uint32_t StrSize;
     Constant *SrcLocStr = OMPBuilder.getOrCreateSrcLocStr(Loc, StrSize);
     Value *Ident = OMPBuilder.getOrCreateIdent(SrcLocStr, StrSize);
     Value *TID = OMPBuilder.getOrCreateThreadID(Ident);
     Value *SumLocal =
TEST_F(OpenMPIRBuilderTest, CreateTwoReductions) {
   }
 
   // Store initial values of reductions into global variables.
   Builder.CreateStore(ConstantFP::get(Builder.getFloatTy(), 0.0), SumReduced);
   Builder.CreateStore(Builder.getInt32(1), XorReduced);
 
   InsertPointTy FirstBodyIP, FirstBodyAllocaIP;
   auto FirstBodyGenCB = [&](InsertPointTy InnerAllocaIP,
-                            InsertPointTy CodeGenIP,
-                            BasicBlock &ContinuationBB) {
+                            InsertPointTy CodeGenIP) {
     IRBuilderBase::InsertPointGuard Guard(Builder);
     Builder.restoreIP(CodeGenIP);
 
     uint32_t StrSize;
     Constant *SrcLocStr = OMPBuilder.getOrCreateSrcLocStr(Loc, StrSize);
     Value *Ident = OMPBuilder.getOrCreateIdent(SrcLocStr, StrSize);
     Value *TID = OMPBuilder.getOrCreateThreadID(Ident);
     Value *SumLocal =
         Builder.CreateUIToFP(TID, Builder.getFloatTy(), "sum.local");
     Value *SumPartial = Builder.CreateLoad(SumType, SumReduced, "sum.partial");
     Value *Sum = Builder.CreateFAdd(SumPartial, SumLocal, "sum");
     Builder.CreateStore(Sum, SumReduced);
 
     FirstBodyIP = Builder.saveIP();
     FirstBodyAllocaIP = InnerAllocaIP;
   };
 
   InsertPointTy SecondBodyIP, SecondBodyAllocaIP;
   auto SecondBodyGenCB = [&](InsertPointTy InnerAllocaIP,
-                             InsertPointTy CodeGenIP,
-                             BasicBlock &ContinuationBB) {
+                             InsertPointTy CodeGenIP) {
     IRBuilderBase::InsertPointGuard Guard(Builder);
     Builder.restoreIP(CodeGenIP);
 
     uint32_t StrSize;
     Constant *SrcLocStr = OMPBuilder.getOrCreateSrcLocStr(Loc, StrSize);
     Value *Ident = OMPBuilder.getOrCreateIdent(SrcLocStr, StrSize);
     Value *TID = OMPBuilder.getOrCreateThreadID(Ident);
     Value *XorPartial = Builder.CreateLoad(XorType, XorReduced, "xor.partial");
TEST_F(OpenMPIRBuilderTest, CreateSectionsSimple) {
   Builder.CreateBr(EnterBB);
   Builder.SetInsertPoint(EnterBB);
   OpenMPIRBuilder::LocationDescription Loc({Builder.saveIP(), DL});
 
   llvm::SmallVector<BodyGenCallbackTy, 4> SectionCBVector;
   llvm::SmallVector<BasicBlock *, 4> CaseBBs;
 
   auto FiniCB = [&](InsertPointTy IP) {};
-  auto SectionCB = [&](InsertPointTy AllocaIP, InsertPointTy CodeGenIP,
-                       BasicBlock &FiniBB) {
-    Builder.restoreIP(CodeGenIP);
-    Builder.CreateBr(&FiniBB);
-  };
+  auto SectionCB = [&](InsertPointTy AllocaIP, InsertPointTy CodeGenIP) {};
   SectionCBVector.push_back(SectionCB);
 
   auto PrivCB = [](InsertPointTy AllocaIP, InsertPointTy CodeGenIP,
                    llvm::Value &, llvm::Value &Val,
                    llvm::Value *&ReplVal) { return CodeGenIP; };
   IRBuilder<>::InsertPoint AllocaIP(&F->getEntryBlock(),
                                     F->getEntryBlock().getFirstInsertionPt());
   Builder.restoreIP(OMPBuilder.createSections(Loc, AllocaIP, SectionCBVector,
TEST_F(OpenMPIRBuilderTest, CreateSections) {
   F->setName("func");
   IRBuilder<> Builder(BB);
 
   OpenMPIRBuilder::LocationDescription Loc({Builder.saveIP(), DL});
   llvm::SmallVector<BodyGenCallbackTy, 4> SectionCBVector;
   llvm::SmallVector<BasicBlock *, 4> CaseBBs;
 
   BasicBlock *SwitchBB = nullptr;
-  BasicBlock *ForExitBB = nullptr;
-  BasicBlock *ForIncBB = nullptr;
   AllocaInst *PrivAI = nullptr;
   SwitchInst *Switch = nullptr;
 
   unsigned NumBodiesGenerated = 0;
   unsigned NumFiniCBCalls = 0;
   PrivAI = Builder.CreateAlloca(F->arg_begin()->getType());
 
   auto FiniCB = [&](InsertPointTy IP) {
     ++NumFiniCBCalls;
     BasicBlock *IPBB = IP.getBlock();
     EXPECT_NE(IPBB->end(), IP.getPoint());
   };
 
-  auto SectionCB = [&](InsertPointTy AllocaIP, InsertPointTy CodeGenIP,
-                       BasicBlock &FiniBB) {
+  auto SectionCB = [&](InsertPointTy AllocaIP, InsertPointTy CodeGenIP) {
     ++NumBodiesGenerated;
     CaseBBs.push_back(CodeGenIP.getBlock());
     SwitchBB = CodeGenIP.getBlock()->getSinglePredecessor();
     Builder.restoreIP(CodeGenIP);
     Builder.CreateStore(F->arg_begin(), PrivAI);
     Value *PrivLoad =
         Builder.CreateLoad(F->arg_begin()->getType(), PrivAI, "local.alloca");
     Builder.CreateICmpNE(F->arg_begin(), PrivLoad);
-    Builder.CreateBr(&FiniBB);
-    ForIncBB =
-        CodeGenIP.getBlock()->getSinglePredecessor()->getSingleSuccessor();
   };
   auto PrivCB = [](InsertPointTy AllocaIP, InsertPointTy CodeGenIP,
                    llvm::Value &, llvm::Value &Val, llvm::Value *&ReplVal) {
     // TODO: Privatization not implemented yet
     return CodeGenIP;
   };
 
   SectionCBVector.push_back(SectionCB);
   SectionCBVector.push_back(SectionCB);
 
   IRBuilder<>::InsertPoint AllocaIP(&F->getEntryBlock(),
                                     F->getEntryBlock().getFirstInsertionPt());
   Builder.restoreIP(OMPBuilder.createSections(Loc, AllocaIP, SectionCBVector,
                                               PrivCB, FiniCB, false, false));
   Builder.CreateRetVoid(); // Required at the end of the function
 
   // Switch BB's predecessor is loop condition BB, whose successor at index 1 is
   // loop's exit BB
-  ForExitBB =
+  BasicBlock *ForExitBB =
       SwitchBB->getSinglePredecessor()->getTerminator()->getSuccessor(1);
   EXPECT_NE(ForExitBB, nullptr);
 
   EXPECT_NE(PrivAI, nullptr);
   Function *OutlinedFn = PrivAI->getFunction();
   EXPECT_EQ(F, OutlinedFn);
   EXPECT_FALSE(verifyModule(*M, &errs()));
   EXPECT_EQ(OutlinedFn->arg_size(), 1U);
-  EXPECT_EQ(OutlinedFn->getBasicBlockList().size(), size_t(11));
 
   BasicBlock *LoopPreheaderBB =
       OutlinedFn->getEntryBlock().getSingleSuccessor();
   // loop variables are 5 - lower bound, upper bound, stride, islastiter, and
   // iterator/counter
   bool FoundForInit = false;
   for (Instruction &Inst : *LoopPreheaderBB) {
     if (isa<CallInst>(Inst)) {
TEST_F(OpenMPIRBuilderTest, CreateSections) {
   EXPECT_EQ(FoundForExit, true);
   EXPECT_EQ(FoundBarrier, true);
 
   EXPECT_NE(SwitchBB, nullptr);
   EXPECT_NE(SwitchBB->getTerminator(), nullptr);
   EXPECT_EQ(isa<SwitchInst>(SwitchBB->getTerminator()), true);
   Switch = cast<SwitchInst>(SwitchBB->getTerminator());
   EXPECT_EQ(Switch->getNumCases(), 2U);
-  EXPECT_NE(ForIncBB, nullptr);
-  EXPECT_EQ(Switch->getSuccessor(0), ForIncBB);
 
   EXPECT_EQ(CaseBBs.size(), 2U);
   for (auto *&CaseBB : CaseBBs) {
     EXPECT_EQ(CaseBB->getParent(), OutlinedFn);
-    EXPECT_EQ(CaseBB->getSingleSuccessor(), ForExitBB);
   }
 
   ASSERT_EQ(NumBodiesGenerated, 2U);
   ASSERT_EQ(NumFiniCBCalls, 1U);
   EXPECT_FALSE(verifyModule(*M, &errs()));
 }
 
 TEST_F(OpenMPIRBuilderTest, CreateSectionsNoWait) {

mlir/lib/Target/LLVMIR/Dialect/OpenMP/OpenMPToLLVMIRTranslation.cpp

findAllocaInsertPoint(llvm::IRBuilderBase &builder,
   if (walkResult.wasInterrupted())
     return allocaInsertPoint;
 
   // Otherwise, insert to the entry block of the surrounding function.
   // If the current IRBuilder InsertPoint is the function's entry, it cannot
   // also be used for alloca insertion which would result in insertion order
   // confusion. Create a new BasicBlock for the Builder and use the entry block
   // for the allocs.
+  // TODO: Create a dedicated alloca BasicBlock at function creation such that
+  // we do not need to move the current InertPoint here.
   if (builder.GetInsertBlock() ==
       &builder.GetInsertBlock()->getParent()->getEntryBlock()) {
     assert(builder.GetInsertPoint() == builder.GetInsertBlock()->end() &&
            "Assuming end of basic block");
     llvm::BasicBlock *entryBB = llvm::BasicBlock::Create(
         builder.getContext(), "entry", builder.GetInsertBlock()->getParent(),
         builder.GetInsertBlock()->getNextNode());
     builder.CreateBr(entryBB);
     builder.SetInsertPoint(entryBB);
   }
 
   llvm::BasicBlock &funcEntryBlock =
       builder.GetInsertBlock()->getParent()->getEntryBlock();
   return llvm::OpenMPIRBuilder::InsertPointTy(
       &funcEntryBlock, funcEntryBlock.getFirstInsertionPt());
 }
 
 /// Converts the given region that appears within an OpenMP dialect operation to
 /// LLVM IR, creating a branch from the `sourceBlock` to the entry block of the
 /// region, and a branch from any block with an successor-less OpenMP terminator
 /// to `continuationBlock`. Populates `continuationBlockPHIs` with the PHI nodes
 /// of the continuation block if provided.
-static void convertOmpOpRegions(
-    Region &region, StringRef blockName, llvm::BasicBlock &sourceBlock,
-    llvm::BasicBlock &continuationBlock, llvm::IRBuilderBase &builder,
+static llvm::BasicBlock *convertOmpOpRegions(
+    Region &region, StringRef blockName, llvm::IRBuilderBase &builder,
     LLVM::ModuleTranslation &moduleTranslation, LogicalResult &bodyGenStatus,
     SmallVectorImpl<llvm::PHINode *> *continuationBlockPHIs = nullptr) {
+  llvm::BasicBlock *continuationBlock =
+      splitBB(builder, true, "omp.region.cont");
+  llvm::BasicBlock *sourceBlock = builder.GetInsertBlock();
+
   llvm::LLVMContext &llvmContext = builder.getContext();
   for (Block &bb : region) {
     llvm::BasicBlock *llvmBB = llvm::BasicBlock::Create(
         llvmContext, blockName, builder.GetInsertBlock()->getParent(),
         builder.GetInsertBlock()->getNextNode());
     moduleTranslation.mapBlock(&bb, llvmBB);
   }
 
-  llvm::Instruction *sourceTerminator = sourceBlock.getTerminator();
+  llvm::Instruction *sourceTerminator = sourceBlock->getTerminator();
 
   // Terminators (namely YieldOp) may be forwarding values to the region that
   // need to be available in the continuation block. Collect the types of these
   // operands in preparation of creating PHI nodes.
   SmallVector<llvm::Type *> continuationBlockPHITypes;
   bool operandsProcessed = false;
   unsigned numYields = 0;
   for (Block &bb : region.getBlocks()) {
static llvm::BasicBlock *convertOmpOpRegions(
   // terminators in this region.
   if (!continuationBlockPHITypes.empty())
     assert(
         continuationBlockPHIs &&
         "expected continuation block PHIs if converted regions yield values");
   if (continuationBlockPHIs) {
     llvm::IRBuilderBase::InsertPointGuard guard(builder);
     continuationBlockPHIs->reserve(continuationBlockPHITypes.size());
-    builder.SetInsertPoint(&continuationBlock, continuationBlock.begin());
+    builder.SetInsertPoint(continuationBlock, continuationBlock->begin());
     for (llvm::Type *ty : continuationBlockPHITypes)
       continuationBlockPHIs->push_back(builder.CreatePHI(ty, numYields));
   }
 
   // Convert blocks one by one in topological order to ensure
   // defs are converted before uses.
   SetVector<Block *> blocks =
       LLVM::detail::getTopologicallySortedBlocks(region);
   for (Block *bb : blocks) {
     llvm::BasicBlock *llvmBB = moduleTranslation.lookupBlock(bb);
     // Retarget the branch of the entry block to the entry block of the
     // converted region (regions are single-entry).
     if (bb->isEntryBlock()) {
       assert(sourceTerminator->getNumSuccessors() == 1 &&
              "provided entry block has multiple successors");
-      assert(sourceTerminator->getSuccessor(0) == &continuationBlock &&
+      assert(sourceTerminator->getSuccessor(0) == continuationBlock &&
              "ContinuationBlock is not the successor of the entry block");
       sourceTerminator->setSuccessor(0, llvmBB);
     }
 
     llvm::IRBuilderBase::InsertPointGuard guard(builder);
     if (failed(
             moduleTranslation.convertBlock(*bb, bb->isEntryBlock(), builder))) {
       bodyGenStatus = failure();
-      return;
+      return continuationBlock;
     }
 
     // Special handling for `omp.yield` and `omp.terminator` (we may have more
     // than one): they return the control to the parent OpenMP dialect operation
     // so replace them with the branch to the continuation block. We handle this
     // here to avoid relying inter-function communication through the
     // ModuleTranslation class to set up the correct insertion point. This is
     // also consistent with MLIR's idiom of handling special region terminators
     // in the same code that handles the region-owning operation.
     Operation *terminator = bb->getTerminator();
     if (isa<omp::TerminatorOp, omp::YieldOp>(terminator)) {
-      builder.CreateBr(&continuationBlock);
+      builder.CreateBr(continuationBlock);
 
       for (unsigned i = 0, e = terminator->getNumOperands(); i < e; ++i)
         (*continuationBlockPHIs)[i]->addIncoming(
             moduleTranslation.lookupValue(terminator->getOperand(i)), llvmBB);
     }
   }
   // After all blocks have been traversed and values mapped, connect the PHI
   // nodes to the results of preceding blocks.
   LLVM::detail::connectPHINodes(region, moduleTranslation);
 
   // Remove the blocks and values defined in this region from the mapping since
   // they are not visible outside of this region. This allows the same region to
   // be converted several times, that is cloned, without clashes, and slightly
   // speeds up the lookups.
   moduleTranslation.forgetMapping(region);
+
+  return continuationBlock;
 }
 
 /// Convert ProcBindKind from MLIR-generated enum to LLVM enum.
 static llvm::omp::ProcBindKind getProcBindKind(omp::ClauseProcBindKind kind) {
   switch (kind) {
   case omp::ClauseProcBindKind::Close:
     return llvm::omp::ProcBindKind::OMP_PROC_BIND_close;
   case omp::ClauseProcBindKind::Master:

 static LogicalResult
 convertOmpParallel(omp::ParallelOp opInst, llvm::IRBuilderBase &builder,
                    LLVM::ModuleTranslation &moduleTranslation) {
   using InsertPointTy = llvm::OpenMPIRBuilder::InsertPointTy;
   // TODO: support error propagation in OpenMPIRBuilder and use it instead of
   // relying on captured variables.
   LogicalResult bodyGenStatus = success();
 
-  auto bodyGenCB = [&](InsertPointTy allocaIP, InsertPointTy codeGenIP,
-                       llvm::BasicBlock &continuationBlock) {
+  auto bodyGenCB = [&](InsertPointTy allocaIP, InsertPointTy codeGenIP) {
     // Save the alloca insertion point on ModuleTranslation stack for use in
     // nested regions.
     LLVM::ModuleTranslation::SaveStack<OpenMPAllocaStackFrame> frame(
         moduleTranslation, allocaIP);
 
     // ParallelOp has only one region associated with it.
-    convertOmpOpRegions(opInst.getRegion(), "omp.par.region",
-                        *codeGenIP.getBlock(), continuationBlock, builder,
+    builder.restoreIP(codeGenIP);
+    convertOmpOpRegions(opInst.getRegion(), "omp.par.region", builder,
                         moduleTranslation, bodyGenStatus);
   };
 
   // TODO: Perform appropriate actions according to the data-sharing
   // attribute (shared, private, firstprivate, ...) of variables.
   // Currently defaults to shared.
   auto privCB = [&](InsertPointTy allocaIP, InsertPointTy codeGenIP,
                     llvm::Value &, llvm::Value &vPtr,

 static LogicalResult
 convertOmpMaster(Operation &opInst, llvm::IRBuilderBase &builder,
                  LLVM::ModuleTranslation &moduleTranslation) {
   using InsertPointTy = llvm::OpenMPIRBuilder::InsertPointTy;
   // TODO: support error propagation in OpenMPIRBuilder and use it instead of
   // relying on captured variables.
   LogicalResult bodyGenStatus = success();
 
-  auto bodyGenCB = [&](InsertPointTy allocaIP, InsertPointTy codeGenIP,
-                       llvm::BasicBlock &continuationBlock) {
+  auto bodyGenCB = [&](InsertPointTy allocaIP, InsertPointTy codeGenIP) {
     // MasterOp has only one region associated with it.
     auto &region = cast<omp::MasterOp>(opInst).getRegion();
-    convertOmpOpRegions(region, "omp.master.region", *codeGenIP.getBlock(),
-                        continuationBlock, builder, moduleTranslation,
+    builder.restoreIP(codeGenIP);
+    convertOmpOpRegions(region, "omp.master.region", builder, moduleTranslation,
                         bodyGenStatus);
   };
 
   // TODO: Perform finalization actions for variables. This has to be
   // called for variables which have destructors/finalizers.
   auto finiCB = [&](InsertPointTy codeGenIP) {};
 
   llvm::OpenMPIRBuilder::LocationDescription ompLoc(builder);
   builder.restoreIP(moduleTranslation.getOpenMPBuilder()->createMaster(
       ompLoc, bodyGenCB, finiCB));
   return success();
 }
 
 /// Converts an OpenMP 'critical' operation into LLVM IR using OpenMPIRBuilder.
 static LogicalResult
 convertOmpCritical(Operation &opInst, llvm::IRBuilderBase &builder,
                    LLVM::ModuleTranslation &moduleTranslation) {
   using InsertPointTy = llvm::OpenMPIRBuilder::InsertPointTy;
   auto criticalOp = cast<omp::CriticalOp>(opInst);
   // TODO: support error propagation in OpenMPIRBuilder and use it instead of
   // relying on captured variables.
   LogicalResult bodyGenStatus = success();
 
-  auto bodyGenCB = [&](InsertPointTy allocaIP, InsertPointTy codeGenIP,
-                       llvm::BasicBlock &continuationBlock) {
+  auto bodyGenCB = [&](InsertPointTy allocaIP, InsertPointTy codeGenIP) {
     // CriticalOp has only one region associated with it.
     auto &region = cast<omp::CriticalOp>(opInst).getRegion();
-    convertOmpOpRegions(region, "omp.critical.region", *codeGenIP.getBlock(),
-                        continuationBlock, builder, moduleTranslation,
-                        bodyGenStatus);
+    builder.restoreIP(codeGenIP);
+    convertOmpOpRegions(region, "omp.critical.region", builder,
+                        moduleTranslation, bodyGenStatus);
   };
 
   // TODO: Perform finalization actions for variables. This has to be
   // called for variables which have destructors/finalizers.
   auto finiCB = [&](InsertPointTy codeGenIP) {};
 
   llvm::OpenMPIRBuilder::LocationDescription ompLoc(builder);
   llvm::LLVMContext &llvmContext = moduleTranslation.getLLVMContext();
if (continuationBlockArgs)
           moduleTranslation.lookupValues(region.front().back().getOperands()));
 
     // Drop the mapping that is no longer necessary so that the same region can
     // be processed multiple times.
     moduleTranslation.forgetMapping(region);
     return success();
   }
 
-  // Create the continuation block manually instead of calling splitBlock
-  // because the current insertion block may not have a terminator.
-  llvm::BasicBlock *continuationBlock =
-      llvm::BasicBlock::Create(builder.getContext(), blockName + ".cont",
-                               builder.GetInsertBlock()->getParent(),
-                               builder.GetInsertBlock()->getNextNode());
-  builder.CreateBr(continuationBlock);
-
   LogicalResult bodyGenStatus = success();
   SmallVector<llvm::PHINode *> phis;
-  convertOmpOpRegions(region, blockName, *builder.GetInsertBlock(),
-                      *continuationBlock, builder, moduleTranslation,
-                      bodyGenStatus, &phis);
+  llvm::BasicBlock *continuationBlock = convertOmpOpRegions(
+      region, blockName, builder, moduleTranslation, bodyGenStatus, &phis);
   if (failed(bodyGenStatus))
     return failure();
   if (continuationBlockArgs)
     llvm::append_range(*continuationBlockArgs, phis);
   builder.SetInsertPoint(continuationBlock,
                          continuationBlock->getFirstInsertionPt());
   return success();
 }
convertOmpOrderedRegion(Operation &opInst, llvm::IRBuilderBase &builder,
   // TODO: The code generation for ordered simd directive is not supported yet.
   if (orderedRegionOp.simd())
     return failure();
 
   // TODO: support error propagation in OpenMPIRBuilder and use it instead of
   // relying on captured variables.
   LogicalResult bodyGenStatus = success();
 
-  auto bodyGenCB = [&](InsertPointTy allocaIP, InsertPointTy codeGenIP,
-                       llvm::BasicBlock &continuationBlock) {
+  auto bodyGenCB = [&](InsertPointTy allocaIP, InsertPointTy codeGenIP) {
     // OrderedOp has only one region associated with it.
     auto &region = cast<omp::OrderedRegionOp>(opInst).getRegion();
-    convertOmpOpRegions(region, "omp.ordered.region", *codeGenIP.getBlock(),
-                        continuationBlock, builder, moduleTranslation,
-                        bodyGenStatus);
+    builder.restoreIP(codeGenIP);
+    convertOmpOpRegions(region, "omp.ordered.region", builder,
+                        moduleTranslation, bodyGenStatus);
   };
 
   // TODO: Perform finalization actions for variables. This has to be
   // called for variables which have destructors/finalizers.
   auto finiCB = [&](InsertPointTy codeGenIP) {};
 
   llvm::OpenMPIRBuilder::LocationDescription ompLoc(builder);
   builder.restoreIP(
convertOmpSections(Operation &opInst, llvm::IRBuilderBase &builder,
 
   for (Operation &op : *sectionsOp.region().begin()) {
     auto sectionOp = dyn_cast<omp::SectionOp>(op);
     if (!sectionOp) // omp.terminator
       continue;
 
     Region &region = sectionOp.region();
     auto sectionCB = [&region, &builder, &moduleTranslation, &bodyGenStatus](
-                         InsertPointTy allocaIP, InsertPointTy codeGenIP,
-                         llvm::BasicBlock &finiBB) {
+                         InsertPointTy allocaIP, InsertPointTy codeGenIP) {
       builder.restoreIP(codeGenIP);
-      builder.CreateBr(&finiBB);
-      convertOmpOpRegions(region, "omp.section.region", *codeGenIP.getBlock(),
-                          finiBB, builder, moduleTranslation, bodyGenStatus);
+      convertOmpOpRegions(region, "omp.section.region", builder,
+                          moduleTranslation, bodyGenStatus);
     };
     sectionCBs.push_back(sectionCB);
   }
 
   // No sections within omp.sections operation - skip generation. This situation
   // is only possible if there is only a terminator operation inside the
   // sections operation
   if (sectionCBs.empty())

 
 /// Converts an OpenMP single construct into LLVM IR using OpenMPIRBuilder.
 static LogicalResult
 convertOmpSingle(omp::SingleOp &singleOp, llvm::IRBuilderBase &builder,
                  LLVM::ModuleTranslation &moduleTranslation) {
   using InsertPointTy = llvm::OpenMPIRBuilder::InsertPointTy;
   llvm::OpenMPIRBuilder::LocationDescription ompLoc(builder);
   LogicalResult bodyGenStatus = success();
-  auto bodyCB = [&](InsertPointTy allocaIP, InsertPointTy codegenIP,
-                    llvm::BasicBlock &continuationBB) {
-    convertOmpOpRegions(singleOp.region(), "omp.single.region",
-                        *codegenIP.getBlock(), continuationBB, builder,
+  auto bodyCB = [&](InsertPointTy allocaIP, InsertPointTy codegenIP) {
+    builder.restoreIP(codegenIP);
+    convertOmpOpRegions(singleOp.region(), "omp.single.region", builder,
                         moduleTranslation, bodyGenStatus);
   };
   auto finiCB = [&](InsertPointTy codeGenIP) {};
   builder.restoreIP(moduleTranslation.getOpenMPBuilder()->createSingle(
       ompLoc, bodyCB, finiCB, singleOp.nowait(), /*DidIt=*/nullptr));
   return bodyGenStatus;
 }
 
auto bodyGen = [&](llvm::OpenMPIRBuilder::InsertPointTy ip, llvm::Value *iv) {
     // CanonicalLoopInfo always points to the beginning of the entry block of
     // the body.
     bodyInsertPoints.push_back(ip);
 
     if (loopInfos.size() != loop.getNumLoops() - 1)
       return;
 
     // Convert the body of the loop.
-    llvm::BasicBlock *entryBlock = ip.getBlock();
-    llvm::BasicBlock *exitBlock =
-        entryBlock->splitBasicBlock(ip.getPoint(), "omp.wsloop.exit");
-    convertOmpOpRegions(loop.region(), "omp.wsloop.region", *entryBlock,
-                        *exitBlock, builder, moduleTranslation, bodyGenStatus);
+    builder.restoreIP(ip);
+    convertOmpOpRegions(loop.region(), "omp.wsloop.region", builder,
+                        moduleTranslation, bodyGenStatus);
   };
 
   // Delegate actual loop construction to the OpenMP IRBuilder.
   // TODO: this currently assumes WsLoop is semantically similar to SCF loop,
   // i.e. it has a positive step, uses signed integer semantics. Reconsider
   // this code when WsLoop clearly supports more cases.
   llvm::OpenMPIRBuilder *ompBuilder = moduleTranslation.getOpenMPBuilder();
   for (unsigned i = 0, e = loop.getNumLoops(); i < e; ++i) {
auto bodyGen = [&](llvm::OpenMPIRBuilder::InsertPointTy ip, llvm::Value *iv) {
     // CanonicalLoopInfo always points to the beginning of the entry block of
     // the body.
     bodyInsertPoints.push_back(ip);
 
     if (loopInfos.size() != loop.getNumLoops() - 1)
       return;
 
     // Convert the body of the loop.
-    llvm::BasicBlock *entryBlock = ip.getBlock();
-    llvm::BasicBlock *exitBlock =
-        entryBlock->splitBasicBlock(ip.getPoint(), "omp.simdloop.exit");
-    convertOmpOpRegions(loop.region(), "omp.simdloop.region", *entryBlock,
-                        *exitBlock, builder, moduleTranslation, bodyGenStatus);
+    builder.restoreIP(ip);
+    convertOmpOpRegions(loop.region(), "omp.simdloop.region", builder,
+                        moduleTranslation, bodyGenStatus);
   };
 
   // Delegate actual loop construction to the OpenMP IRBuilder.
   // TODO: this currently assumes SimdLoop is semantically similar to SCF loop,
   // i.e. it has a positive step, uses signed integer semantics. Reconsider
   // this code when SimdLoop clearly supports more cases.
   llvm::OpenMPIRBuilder *ompBuilder = moduleTranslation.getOpenMPBuilder();
   for (unsigned i = 0, e = loop.getNumLoops(); i < e; ++i) {

mlir/test/Target/LLVMIR/openmp-llvm.mlir

llvm.func @omp_ordered(%arg0 : i32, %arg1 : i32, %arg2 : i32, %arg3 : i64,
   // CHECK-NEXT:  call void @__kmpc_ordered(%struct.ident_t* @[[GLOB1]], i32 [[OMP_THREAD]])
   omp.ordered_region {
     omp.terminator
   // CHECK: call void @__kmpc_end_ordered(%struct.ident_t* @[[GLOB1]], i32 [[OMP_THREAD]])
   }
 
   omp.wsloop ordered(0)
   for (%arg7) : i32 = (%arg0) to (%arg1) step (%arg2) {
-    // CHECK: [[OMP_THREAD:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1:[0-9]+]])
-    // CHECK-NEXT:  call void @__kmpc_ordered(%struct.ident_t* @[[GLOB1]], i32 [[OMP_THREAD]])
+    // CHECK:  call void @__kmpc_ordered(%struct.ident_t* @[[GLOB3:[0-9]+]], i32 [[OMP_THREAD2:%.*]])
     omp.ordered_region  {
       omp.terminator
-    // CHECK: call void @__kmpc_end_ordered(%struct.ident_t* @[[GLOB1]], i32 [[OMP_THREAD]])
+    // CHECK: call void @__kmpc_end_ordered(%struct.ident_t* @[[GLOB3]], i32 [[OMP_THREAD2]])
     }
     omp.yield
   }
 
   omp.wsloop ordered(1)
   for (%arg7) : i32 = (%arg0) to (%arg1) step (%arg2) {
     // CHECK: [[TMP:%.*]] = getelementptr inbounds [1 x i64], [1 x i64]* [[ADDR]], i64 0, i64 0
     // CHECK: store i64 [[ARG0:%.*]], i64* [[TMP]], align 8
llvm.func @omp_sections_trivial() -> () {
   // CHECK: [[COND]]:
   // CHECK:   br i1 %{{.*}}, label %[[BODY:.*]], label %[[EXIT:.*]]
   // CHECK: [[BODY]]:
   // CHECK:   switch i32 %{{.*}}, label %[[INC:.*]] [
   // CHECK-NEXT:     i32 0, label %[[SECTION1:.*]]
   // CHECK-NEXT:     i32 1, label %[[SECTION2:.*]]
   // CHECK-NEXT: ]
 
-  // CHECK: [[INC]]:
-  // CHECK:   %{{.*}} = add {{.*}}, 1
-  // CHECK:   br label %[[HEADER]]
-
-  // CHECK: [[EXIT]]:
-  // CHECK:   call void @__kmpc_for_static_fini({{.*}})
-  // CHECK:   call void @__kmpc_barrier({{.*}})
-  // CHECK:   br label %[[AFTER:.*]]
-
-  // CHECK: [[AFTER]]:
-  // CHECK:   br label %[[END:.*]]
-
-  // CHECK: [[END]]:
-  // CHECK:   ret void
   omp.sections {
     omp.section {
       // CHECK: [[SECTION1]]:
-      // CHECK-NEXT: br label %[[REGION1:[^ ,]*]]
-      // CHECK: [[REGION1]]:
-      // CHECK-NEXT: br label %[[EXIT]]
+      // CHECK-NEXT: br label %[[SECTION1_REGION1:[^ ,]*]]
+      // CHECK-EMPTY:
+      // CHECK-NEXT: [[SECTION1_REGION1]]:
+      // CHECK-NEXT: br label %[[SECTION1_REGION2:[^ ,]*]]
+      // CHECK-EMPTY:
+      // CHECK-NEXT: [[SECTION1_REGION2]]:
+      // CHECK-NEXT: br label %[[INC]]
       omp.terminator
     }
     omp.section {
       // CHECK: [[SECTION2]]:
-      // CHECK-NEXT: br label %[[REGION2:[^ ,]*]]
-      // CHECK: [[REGION2]]:
-      // CHECK-NEXT: br label %[[EXIT]]
+      // CHECK: br label %[[INC]]
       omp.terminator
     }
     omp.terminator
   }
+
+  // CHECK: [[INC]]:
+  // CHECK:   %{{.*}} = add {{.*}}, 1
+  // CHECK:   br label %[[HEADER]]
+
+  // CHECK: [[EXIT]]:
+  // CHECK:   call void @__kmpc_for_static_fini({{.*}})
+  // CHECK:   call void @__kmpc_barrier({{.*}})
+  // CHECK:   br label %[[AFTER:.*]]
+
+  // CHECK: [[AFTER]]:
+  // CHECK:   ret void
   llvm.return
 }
 
 // -----
 
 // CHECK: declare void @foo()
 llvm.func @foo()