SimplifyCFG: Register cloned assume intrinsics with assumption cache when creating critical edge.

[pcc]

• N may be null

[hfinkel]

LGTM

llvm/trunk/lib/Transforms/Utils/SimplifyCFG.cpp

 #include "llvm/ADT/Optional.h"
 #include "llvm/ADT/SetOperations.h"
 #include "llvm/ADT/SetVector.h"
 #include "llvm/ADT/SmallPtrSet.h"
 #include "llvm/ADT/SmallSet.h"
 #include "llvm/ADT/SmallVector.h"
 #include "llvm/ADT/Statistic.h"
 #include "llvm/ADT/STLExtras.h"
+#include "llvm/Analysis/AssumptionCache.h"
 #include "llvm/Analysis/ConstantFolding.h"
 #include "llvm/Analysis/EHPersonalities.h"
 #include "llvm/Analysis/InstructionSimplify.h"
 #include "llvm/Analysis/TargetTransformInfo.h"
 #include "llvm/Analysis/ValueTracking.h"
 #include "llvm/IR/BasicBlock.h"
 #include "llvm/IR/CallSite.h"
 #include "llvm/IR/CFG.h"
static bool BlockIsSimpleEnoughToThreadThrough(BasicBlock *BB) {
   }
 
   return true;
 }
 
 /// If we have a conditional branch on a PHI node value that is defined in the
 /// same block as the branch and if any PHI entries are constants, thread edges
 /// corresponding to that entry to be branches to their ultimate destination.
-static bool FoldCondBranchOnPHI(BranchInst *BI, const DataLayout &DL) {
+static bool FoldCondBranchOnPHI(BranchInst *BI, const DataLayout &DL,
+                                AssumptionCache *AC) {
   BasicBlock *BB = BI->getParent();
   PHINode *PN = dyn_cast<PHINode>(BI->getCondition());
   // NOTE: we currently cannot transform this case if the PHI node is used
   // outside of the block.
   if (!PN || PN->getParent() != BB || !PN->hasOneUse())
     return false;
 
   // Degenerate case of a single entry PHI.
for (BasicBlock::iterator BBI = BB->begin(); &*BBI != BI; ++BBI) {
         }
       } else {
         if (!BBI->use_empty())
           TranslateMap[&*BBI] = N;
       }
       // Insert the new instruction into its new home.
       if (N)
         EdgeBB->getInstList().insert(InsertPt, N);
+
+      // Register the new instruction with the assumption cache if necessary.
+      if (auto *II = dyn_cast_or_null<IntrinsicInst>(N))
+        if (II->getIntrinsicID() == Intrinsic::assume)
+          AC->registerAssumption(II);
     }
 
     // Loop over all of the edges from PredBB to BB, changing them to branch
     // to EdgeBB instead.
     TerminatorInst *PredBBTI = PredBB->getTerminator();
     for (unsigned i = 0, e = PredBBTI->getNumSuccessors(); i != e; ++i)
       if (PredBBTI->getSuccessor(i) == BB) {
         BB->removePredecessor(PredBB);
         PredBBTI->setSuccessor(i, EdgeBB);
       }
 
     // Recurse, simplifying any other constants.
-    return FoldCondBranchOnPHI(BI, DL) | true;
+    return FoldCondBranchOnPHI(BI, DL, AC) | true;
   }
 
   return false;
 }
 
 /// Given a BB that starts with the specified two-entry PHI node,
 /// see if we can eliminate it.
 static bool FoldTwoEntryPHINode(PHINode *PN, const TargetTransformInfo &TTI,
if (Succ1TI->getNumSuccessors() == 1 &&
       if (SpeculativelyExecuteBB(BI, BI->getSuccessor(1), TTI))
         return SimplifyCFG(BB, TTI, BonusInstThreshold, AC) | true;
   }
 
   // If this is a branch on a phi node in the current block, thread control
   // through this block if any PHI node entries are constants.
   if (PHINode *PN = dyn_cast<PHINode>(BI->getCondition()))
     if (PN->getParent() == BI->getParent())
-      if (FoldCondBranchOnPHI(BI, DL))
+      if (FoldCondBranchOnPHI(BI, DL, AC))
         return SimplifyCFG(BB, TTI, BonusInstThreshold, AC) | true;
 
   // Scan predecessor blocks for conditional branches.
   for (pred_iterator PI = pred_begin(BB), E = pred_end(BB); PI != E; ++PI)
     if (BranchInst *PBI = dyn_cast<BranchInst>((*PI)->getTerminator()))
       if (PBI != BI && PBI->isConditional())
         if (SimplifyCondBranchToCondBranch(PBI, BI, DL))
           return SimplifyCFG(BB, TTI, BonusInstThreshold, AC) | true;

llvm/trunk/test/Transforms/SimplifyCFG/critedge-assume.ll

+; RUN: opt -o %t %s -instcombine -simplifycfg -thinlto-bc
+; RUN: llvm-dis -o - %t | FileCheck %s
+
+; Test that the simplifycfg pass correctly updates the assumption cache
+; when it clones the llvm.assume call as part of creating a critical
+; edge. To do that, we set up a pass pipeline such that (1) an assumption
+; cache is created for foo before simplifycfg updates it, and (2) foo's
+; assumption cache is verified after simplifycfg has run. To satisfy 1, we
+; run the instcombine pass first in our pipeline. To satisfy 2, we use the
+; ThinLTOBitcodeWriter pass to write bitcode (that pass uses the assumption
+; cache). That ensures that the pass manager does not call releaseMemory()
+; on the AssumptionCacheTracker before the end of the pipeline, which would
+; wipe out the bad assumption cache before it is verified.
+
+target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
+target triple = "x86_64-unknown-linux-gnu"
+
+%class.F = type { i8 }
+%class.B = type { i8 }
+%class.A = type { %class.C }
+%class.C = type { i32 (...)** }
+
+define void @foo(%class.F* %this, %class.B* %out) {
+entry:
+  %call = tail call i32 @_ZNK1F5beginEv(%class.F* %this)
+  %call2 = tail call i32 @_ZNK1F3endEv(%class.F* %this)
+  %cmp.i22 = icmp eq i32 %call, %call2
+  br i1 %cmp.i22, label %while.end, label %while.body.preheader
+
+while.body.preheader:
+  br label %while.body
+
+while.body:
+  %frame_node.sroa.0.023 = phi i32 [ %inc.i, %_ZN10unique_ptrD2Ev.exit ], [ %call, %while.body.preheader ]
+  %call8 = tail call i8* @_Znwm(i64 8)
+  %inc.i = add nsw i32 %frame_node.sroa.0.023, 1
+  %cmp = icmp eq i32 %inc.i, %call2
+  br i1 %cmp, label %_ZN10unique_ptrD2Ev.exit, label %if.then
+
+if.then:
+  tail call void @_ZN1B6appendEv(%class.B* %out)
+  br label %_ZN10unique_ptrD2Ev.exit
+
+_ZN10unique_ptrD2Ev.exit:
+  %x1 = bitcast i8* %call8 to void (%class.A*)***
+  %vtable.i.i = load void (%class.A*)**, void (%class.A*)*** %x1, align 8
+  %x2 = bitcast void (%class.A*)** %vtable.i.i to i8*
+  %x3 = tail call i1 @llvm.type.test(i8* %x2, metadata !"foo")
+  ; CHECK: call void @llvm.assume
+  ; CHECK: call void @llvm.assume
+  tail call void @llvm.assume(i1 %x3) #5
+  br i1 %cmp, label %while.end.loopexit, label %while.body
+
+while.end.loopexit:
+  br label %while.end
+
+while.end:
+  ret void
+}
+
+declare void @llvm.lifetime.start(i64, i8* nocapture)
+
+declare i32 @_ZNK1F5beginEv(%class.F*)
+
+declare i32 @_ZNK1F3endEv(%class.F*)
+
+declare void @llvm.memcpy.p0i8.p0i8.i64(i8* nocapture writeonly, i8* nocapture readonly, i64, i32, i1)
+
+declare noalias nonnull i8* @_Znwm(i64)
+
+declare void @_ZN1B6appendEv(%class.B*)
+
+declare void @llvm.lifetime.end(i64, i8* nocapture)
+
+declare i1 @llvm.type.test(i8*, metadata)
+
+declare void @llvm.assume(i1)
+
+!llvm.module.flags = !{!0}
+!llvm.ident = !{!1}
+
+!0 = !{i32 1, !"PIC Level", i32 2}
+!1 = !{!"clang version 5.0.0 "}