PR for llvm/llvm-project#79861

llvm/include/llvm/Analysis/ScalarEvolution.h

@@ -1314,6 +1314,13 @@ class ScalarEvolution {
   void getPoisonGeneratingValues(SmallPtrSetImpl<const Value *> &Result,
                                  const SCEV *S);
 
+  /// Check whether it is poison-safe to represent the expression S using the
+  /// instruction I. If such a replacement is performed, the poison flags of
+  /// instructions in DropPoisonGeneratingInsts must be dropped.
+  bool canReuseInstruction(
+      const SCEV *S, Instruction *I,
+      SmallVectorImpl<Instruction *> &DropPoisonGeneratingInsts);
+
   class FoldID {
     const SCEV *Op = nullptr;
     const Type *Ty = nullptr;

llvm/lib/Analysis/ScalarEvolution.cpp

@@ -4184,6 +4184,68 @@ void ScalarEvolution::getPoisonGeneratingValues(
     Result.insert(SU->getValue());
 }
 
+bool ScalarEvolution::canReuseInstruction(
+    const SCEV *S, Instruction *I,
+    SmallVectorImpl<Instruction *> &DropPoisonGeneratingInsts) {
+  // If the instruction cannot be poison, it's always safe to reuse.
+  if (programUndefinedIfPoison(I))
+    return true;
+
+  // Otherwise, it is possible that I is more poisonous that S. Collect the
+  // poison-contributors of S, and then check whether I has any additional
+  // poison-contributors. Poison that is contributed through poison-generating
+  // flags is handled by dropping those flags instead.
+  SmallPtrSet<const Value *, 8> PoisonVals;
+  getPoisonGeneratingValues(PoisonVals, S);
+
+  SmallVector<Value *> Worklist;
+  SmallPtrSet<Value *, 8> Visited;
+  Worklist.push_back(I);
+  while (!Worklist.empty()) {
+    Value *V = Worklist.pop_back_val();
+    if (!Visited.insert(V).second)
+      continue;
+
+    // Avoid walking large instruction graphs.
+    if (Visited.size() > 16)
+      return false;
+
+    // Either the value can't be poison, or the S would also be poison if it
+    // is.
+    if (PoisonVals.contains(V) || isGuaranteedNotToBePoison(V))
+      continue;
+
+    auto *I = dyn_cast<Instruction>(V);
+    if (!I)
+      return false;
+
+    // Disjoint or instructions are interpreted as adds by SCEV. However, we
+    // can't replace an arbitrary add with disjoint or, even if we drop the
+    // flag. We would need to convert the or into an add.
+    if (auto *PDI = dyn_cast<PossiblyDisjointInst>(I))
+      if (PDI->isDisjoint())
+        return false;
+
+    // FIXME: Ignore vscale, even though it technically could be poison. Do this
+    // because SCEV currently assumes it can't be poison. Remove this special
+    // case once we proper model when vscale can be poison.
+    if (auto *II = dyn_cast<IntrinsicInst>(I);
+        II && II->getIntrinsicID() == Intrinsic::vscale)
+      continue;
+
+    if (canCreatePoison(cast<Operator>(I), /*ConsiderFlagsAndMetadata*/ false))
+      return false;
+
+    // If the instruction can't create poison, we can recurse to its operands.
+    if (I->hasPoisonGeneratingFlagsOrMetadata())
+      DropPoisonGeneratingInsts.push_back(I);
+
+    for (Value *Op : I->operands())
+      Worklist.push_back(Op);
+  }
+  return true;
+}
+
 const SCEV *
 ScalarEvolution::getSequentialMinMaxExpr(SCEVTypes Kind,
                                          SmallVectorImpl<const SCEV *> &Ops) {

llvm/lib/Transforms/Utils/ScalarEvolutionExpander.cpp

@@ -1366,61 +1366,6 @@ Value *SCEVExpander::expandCodeFor(const SCEV *SH, Type *Ty) {
   return V;
 }
 
-static bool
-canReuseInstruction(ScalarEvolution &SE, const SCEV *S, Instruction *I,
-                    SmallVectorImpl<Instruction *> &DropPoisonGeneratingInsts) {
-  // If the instruction cannot be poison, it's always safe to reuse.
-  if (programUndefinedIfPoison(I))
-    return true;
-
-  // Otherwise, it is possible that I is more poisonous that S. Collect the
-  // poison-contributors of S, and then check whether I has any additional
-  // poison-contributors. Poison that is contributed through poison-generating
-  // flags is handled by dropping those flags instead.
-  SmallPtrSet<const Value *, 8> PoisonVals;
-  SE.getPoisonGeneratingValues(PoisonVals, S);
-
-  SmallVector<Value *> Worklist;
-  SmallPtrSet<Value *, 8> Visited;
-  Worklist.push_back(I);
-  while (!Worklist.empty()) {
-    Value *V = Worklist.pop_back_val();
-    if (!Visited.insert(V).second)
-      continue;
-
-    // Avoid walking large instruction graphs.
-    if (Visited.size() > 16)
-      return false;
-
-    // Either the value can't be poison, or the S would also be poison if it
-    // is.
-    if (PoisonVals.contains(V) || isGuaranteedNotToBePoison(V))
-      continue;
-
-    auto *I = dyn_cast<Instruction>(V);
-    if (!I)
-      return false;
-
-    // FIXME: Ignore vscale, even though it technically could be poison. Do this
-    // because SCEV currently assumes it can't be poison. Remove this special
-    // case once we proper model when vscale can be poison.
-    if (auto *II = dyn_cast<IntrinsicInst>(I);
-        II && II->getIntrinsicID() == Intrinsic::vscale)
-      continue;
-
-    if (canCreatePoison(cast<Operator>(I), /*ConsiderFlagsAndMetadata*/ false))
-      return false;
-
-    // If the instruction can't create poison, we can recurse to its operands.
-    if (I->hasPoisonGeneratingFlagsOrMetadata())
-      DropPoisonGeneratingInsts.push_back(I);
-
-    for (Value *Op : I->operands())
-      Worklist.push_back(Op);
-  }
-  return true;
-}
-
 Value *SCEVExpander::FindValueInExprValueMap(
     const SCEV *S, const Instruction *InsertPt,
     SmallVectorImpl<Instruction *> &DropPoisonGeneratingInsts) {
@@ -1448,7 +1393,7 @@ Value *SCEVExpander::FindValueInExprValueMap(
       continue;
 
     // Make sure reusing the instruction is poison-safe.
-    if (canReuseInstruction(SE, S, EntInst, DropPoisonGeneratingInsts))
+    if (SE.canReuseInstruction(S, EntInst, DropPoisonGeneratingInsts))
       return V;
     DropPoisonGeneratingInsts.clear();
   }

llvm/lib/Transforms/Utils/SimplifyIndVar.cpp

@@ -16,6 +16,7 @@
 #include "llvm/ADT/SmallVector.h"
 #include "llvm/ADT/Statistic.h"
 #include "llvm/Analysis/LoopInfo.h"
+#include "llvm/Analysis/ValueTracking.h"
 #include "llvm/IR/Dominators.h"
 #include "llvm/IR/IRBuilder.h"
 #include "llvm/IR/Instructions.h"
@@ -713,8 +714,11 @@ bool SimplifyIndvar::replaceFloatIVWithIntegerIV(Instruction *UseInst) {
 bool SimplifyIndvar::eliminateIdentitySCEV(Instruction *UseInst,
                                            Instruction *IVOperand) {
   if (!SE->isSCEVable(UseInst->getType()) ||
-      (UseInst->getType() != IVOperand->getType()) ||
-      (SE->getSCEV(UseInst) != SE->getSCEV(IVOperand)))
+      UseInst->getType() != IVOperand->getType())
+    return false;
+
+  const SCEV *UseSCEV = SE->getSCEV(UseInst);
+  if (UseSCEV != SE->getSCEV(IVOperand))
     return false;
 
   // getSCEV(X) == getSCEV(Y) does not guarantee that X and Y are related in the
@@ -742,6 +746,16 @@ bool SimplifyIndvar::eliminateIdentitySCEV(Instruction *UseInst,
   if (!LI->replacementPreservesLCSSAForm(UseInst, IVOperand))
     return false;
 
+  // Make sure the operand is not more poisonous than the instruction.
+  if (!impliesPoison(IVOperand, UseInst)) {
+    SmallVector<Instruction *> DropPoisonGeneratingInsts;
+    if (!SE->canReuseInstruction(UseSCEV, IVOperand, DropPoisonGeneratingInsts))
+      return false;
+
+    for (Instruction *I : DropPoisonGeneratingInsts)
+      I->dropPoisonGeneratingFlagsAndMetadata();
+  }
+
   LLVM_DEBUG(dbgs() << "INDVARS: Eliminated identity: " << *UseInst << '\n');
 
   SE->forgetValue(UseInst);

llvm/test/Transforms/IndVarSimplify/pr55925.ll

@@ -18,9 +18,9 @@ define void @test(ptr %p) personality ptr undef {
 ; CHECK-NEXT:    [[RES:%.*]] = invoke i32 @foo(i32 returned [[TMP0]])
 ; CHECK-NEXT:            to label [[LOOP_LATCH]] unwind label [[EXIT:%.*]]
 ; CHECK:       loop.latch:
-; CHECK-NEXT:    [[INDVARS_IV_NEXT]] = add i64 [[INDVARS_IV]], 1
 ; CHECK-NEXT:    [[TMP1:%.*]] = trunc i64 [[INDVARS_IV]] to i32
 ; CHECK-NEXT:    [[TMP2:%.*]] = call i32 @foo(i32 [[TMP1]])
+; CHECK-NEXT:    [[INDVARS_IV_NEXT]] = add i64 [[INDVARS_IV]], 1
 ; CHECK-NEXT:    br label [[LOOP]]
 ; CHECK:       exit:
 ; CHECK-NEXT:    [[LP:%.*]] = landingpad { ptr, i32 }
@@ -64,8 +64,8 @@ define void @test_critedge(i1 %c, ptr %p) personality ptr undef {
 ; CHECK-NEXT:    br label [[LOOP_LATCH]]
 ; CHECK:       loop.latch:
 ; CHECK-NEXT:    [[PHI:%.*]] = phi i32 [ [[TMP1]], [[LOOP_INVOKE]] ], [ 0, [[LOOP_OTHER]] ]
-; CHECK-NEXT:    [[INDVARS_IV_NEXT]] = add i64 [[INDVARS_IV]], 1
 ; CHECK-NEXT:    [[TMP2:%.*]] = call i32 @foo(i32 [[PHI]])
+; CHECK-NEXT:    [[INDVARS_IV_NEXT]] = add i64 [[INDVARS_IV]], 1
 ; CHECK-NEXT:    br label [[LOOP]]
 ; CHECK:       exit:
 ; CHECK-NEXT:    [[LP:%.*]] = landingpad { ptr, i32 }

llvm/test/Transforms/IndVarSimplify/pr79861.ll

@@ -0,0 +1,107 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 4
+; RUN: opt -S -passes=indvars < %s | FileCheck %s
+
+target datalayout = "n64"
+
+declare void @use(i64)
+
+define void @or_disjoint() {
+; CHECK-LABEL: define void @or_disjoint() {
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    br label [[LOOP:%.*]]
+; CHECK:       loop:
+; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ 2, [[ENTRY:%.*]] ], [ [[IV_DEC:%.*]], [[LOOP]] ]
+; CHECK-NEXT:    [[OR:%.*]] = or disjoint i64 [[IV]], 1
+; CHECK-NEXT:    [[ADD:%.*]] = add nuw nsw i64 [[IV]], 1
+; CHECK-NEXT:    [[SEL:%.*]] = select i1 false, i64 [[OR]], i64 [[ADD]]
+; CHECK-NEXT:    call void @use(i64 [[SEL]])
+; CHECK-NEXT:    [[IV_DEC]] = add nsw i64 [[IV]], -1
+; CHECK-NEXT:    [[EXIT_COND:%.*]] = icmp eq i64 [[IV_DEC]], 0
+; CHECK-NEXT:    br i1 [[EXIT_COND]], label [[EXIT:%.*]], label [[LOOP]]
+; CHECK:       exit:
+; CHECK-NEXT:    ret void
+;
+entry:
+  br label %loop
+
+loop:
+  %iv = phi i64 [ 2, %entry ], [ %iv.dec, %loop ]
+  %or = or disjoint i64 %iv, 1
+  %add = add nsw i64 %iv, 1
+  %sel = select i1 false, i64 %or, i64 %add
+  call void @use(i64 %sel)
+
+  %iv.dec = add nsw i64 %iv, -1
+  %exit.cond = icmp eq i64 %iv.dec, 0
+  br i1 %exit.cond, label %exit, label %loop
+
+exit:
+  ret void
+}
+
+define void @add_nowrap_flags(i64 %n) {
+; CHECK-LABEL: define void @add_nowrap_flags(
+; CHECK-SAME: i64 [[N:%.*]]) {
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    br label [[LOOP:%.*]]
+; CHECK:       loop:
+; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ], [ [[IV_INC:%.*]], [[LOOP]] ]
+; CHECK-NEXT:    [[ADD1:%.*]] = add i64 [[IV]], 123
+; CHECK-NEXT:    call void @use(i64 [[ADD1]])
+; CHECK-NEXT:    [[IV_INC]] = add i64 [[IV]], 1
+; CHECK-NEXT:    [[EXIT_COND:%.*]] = icmp eq i64 [[IV_INC]], [[N]]
+; CHECK-NEXT:    br i1 [[EXIT_COND]], label [[EXIT:%.*]], label [[LOOP]]
+; CHECK:       exit:
+; CHECK-NEXT:    ret void
+;
+entry:
+  br label %loop
+
+loop:
+  %iv = phi i64 [ 0, %entry ], [ %iv.inc, %loop ]
+  %add1 = add nuw nsw i64 %iv, 123
+  %add2 = add i64 %iv, 123
+  %sel = select i1 false, i64 %add1, i64 %add2
+  call void @use(i64 %sel)
+
+  %iv.inc = add i64 %iv, 1
+  %exit.cond = icmp eq i64 %iv.inc, %n
+  br i1 %exit.cond, label %exit, label %loop
+
+exit:
+  ret void
+}
+
+
+define void @expander_or_disjoint(i64 %n) {
+; CHECK-LABEL: define void @expander_or_disjoint(
+; CHECK-SAME: i64 [[N:%.*]]) {
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[OR:%.*]] = or disjoint i64 [[N]], 1
+; CHECK-NEXT:    [[TMP0:%.*]] = add i64 [[N]], 1
+; CHECK-NEXT:    br label [[LOOP:%.*]]
+; CHECK:       loop:
+; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ], [ [[IV_INC:%.*]], [[LOOP]] ]
+; CHECK-NEXT:    [[IV_INC]] = add i64 [[IV]], 1
+; CHECK-NEXT:    [[ADD:%.*]] = add i64 [[IV]], [[OR]]
+; CHECK-NEXT:    call void @use(i64 [[ADD]])
+; CHECK-NEXT:    [[EXITCOND:%.*]] = icmp ne i64 [[IV_INC]], [[TMP0]]
+; CHECK-NEXT:    br i1 [[EXITCOND]], label [[LOOP]], label [[EXIT:%.*]]
+; CHECK:       exit:
+; CHECK-NEXT:    ret void
+;
+entry:
+  %or = or disjoint i64 %n, 1
+  br label %loop
+
+loop:
+  %iv = phi i64 [ 0, %entry ], [ %iv.inc, %loop ]
+  %iv.inc = add i64 %iv, 1
+  %add = add i64 %iv, %or
+  call void @use(i64 %add)
+  %cmp = icmp ult i64 %iv, %n
+  br i1 %cmp, label %loop, label %exit
+
+exit:
+  ret void
+}