[ValueTracking] Handle chain of single-pred blocks in willNotFreeBetween()

llvm/lib/Analysis/ValueTracking.cpp

@@ -93,7 +93,7 @@ static cl::opt<unsigned> DomConditionsMaxUses("dom-conditions-max-uses",
 
 /// Maximum number of instructions to check between assume and context
 /// instruction.
-static constexpr unsigned MaxInstrsToCheckForFree = 16;
+static constexpr unsigned MaxInstrsToCheckForFree = 32;
 
 /// Returns the bitwidth of the given scalar or pointer type. For vector types,
 /// returns the element type's bitwidth.
@@ -704,9 +704,10 @@ bool llvm::isValidAssumeForContext(const Instruction *Inv,
 bool llvm::willNotFreeBetween(const Instruction *Assume,
                               const Instruction *CtxI) {
   // Helper to check if there are any calls in the range that may free memory.
-  auto hasNoFreeInRange = [](auto Range) {
-    for (const auto &[Idx, I] : enumerate(Range)) {
-      if (Idx > MaxInstrsToCheckForFree)
+  unsigned NumChecked = 0;
+  auto hasNoFreeInRange = [&NumChecked](auto Range) {
+    for (const Instruction &I : Range) {
+      if (NumChecked++ > MaxInstrsToCheckForFree)
         return false;
 
       if (auto *CB = dyn_cast<CallBase>(&I)) {
@@ -718,27 +719,32 @@ bool llvm::willNotFreeBetween(const Instruction *Assume,
     return true;
   };
 
-  // Handle cross-block case: CtxI in a successor of Assume's block.
   const BasicBlock *CtxBB = CtxI->getParent();
   const BasicBlock *AssumeBB = Assume->getParent();
   BasicBlock::const_iterator CtxIter = CtxI->getIterator();
-  if (CtxBB != AssumeBB) {
-    if (CtxBB->getSinglePredecessor() != AssumeBB)
+  if (CtxBB == AssumeBB) {
+    // Same block case: check that Assume comes before CtxI.
+    if (Assume != CtxI && !Assume->comesBefore(CtxI))
       return false;
+    return hasNoFreeInRange(make_range(Assume->getIterator(), CtxIter));
+  }
+
+  // Handle chain of single-predecessor blocks.
+  const BasicBlock *CurBB = CtxBB;
+  while (true) {
+    if (CurBB == AssumeBB)
+      return hasNoFreeInRange(
+          make_range(Assume->getIterator(), AssumeBB->end()));
 
-    if (!hasNoFreeInRange(make_range(CtxBB->begin(), CtxIter)))
+    const BasicBlock *PredBB = CurBB->getSinglePredecessor();
+    if (!PredBB)
       return false;
 
-    CtxIter = AssumeBB->end();
-  } else {
-    // Same block case: check that Assume comes before CtxI.
-    if (Assume != CtxI && !Assume->comesBefore(CtxI))
+    if (!hasNoFreeInRange(make_range(CurBB->begin(),
+                                     CurBB == CtxBB ? CtxIter : CurBB->end())))
       return false;
+    CurBB = PredBB;
   }
-
-  // Check if there are any calls between Assume and CtxIter that may free
-  // memory.
-  return hasNoFreeInRange(make_range(Assume->getIterator(), CtxIter));
 }
 
 // TODO: cmpExcludesZero misses many cases where `RHS` is non-constant but

llvm/test/Transforms/LICM/hoist-deref-load.ll

@@ -1,5 +1,6 @@
 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
 ; RUN: opt -S -passes=licm -verify-memoryssa < %s | FileCheck %s
+; RUN: opt -S -passes=licm -verify-memoryssa -use-dereferenceable-at-point-semantics < %s | FileCheck %s
 ; RUN: opt -aa-pipeline=basic-aa -passes='require<opt-remark-emit>,loop-mssa(loop-simplifycfg,licm)' -verify-memoryssa -S < %s | FileCheck %s
 
 target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"

llvm/test/Transforms/SLPVectorizer/AArch64/masked-loads-side-effects-after-vec.ll

@@ -1,5 +1,6 @@
 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 5
 ; RUN: opt -S --passes=slp-vectorizer -mtriple=aarch64-unknown-linux-gnu < %s | FileCheck %s
+; RUN: opt -S --passes=slp-vectorizer -mtriple=aarch64-unknown-linux-gnu -use-dereferenceable-at-point-semantics < %s | FileCheck %s
 
 declare noalias ptr @malloc()