[llvm] [LoopVectorize] Add support for reverse loops in isDereferenceableAndAlignedInLoop (PR #96752)

Florian Hahn via llvm-commits llvm-commits at lists.llvm.org
Wed Jul 24 14:45:10 PDT 2024


================
@@ -293,26 +293,65 @@ bool llvm::isDereferenceableAndAlignedInLoop(LoadInst *LI, Loop *L,
 
   // TODO: Handle overlapping accesses.
   // We should be computing AccessSize as (TC - 1) * Step + EltSize.
-  if (EltSize.sgt(Step->getAPInt()))
+  bool StepIsNegative = Step->getAPInt().isNegative();
+  APInt AbsStep = Step->getAPInt().abs();
+  if (EltSize.ugt(AbsStep))
+    return false;
+
+  // For the moment, restrict ourselves to the case where the access size is a
+  // multiple of the requested alignment and the base is aligned.
+  // TODO: generalize if a case found which warrants
+  if (EltSize.urem(Alignment.value()) != 0)
     return false;
 
   // Compute the total access size for access patterns with unit stride and
   // patterns with gaps. For patterns with unit stride, Step and EltSize are the
   // same.
   // For patterns with gaps (i.e. non unit stride), we are
   // accessing EltSize bytes at every Step.
-  APInt AccessSize = TC * Step->getAPInt();
+  APInt AccessSize = TC * AbsStep;
 
   assert(SE.isLoopInvariant(AddRec->getStart(), L) &&
          "implied by addrec definition");
   Value *Base = nullptr;
   if (auto *StartS = dyn_cast<SCEVUnknown>(AddRec->getStart())) {
+    if (StepIsNegative)
+      return false;
     Base = StartS->getValue();
   } else if (auto *StartS = dyn_cast<SCEVAddExpr>(AddRec->getStart())) {
-    // Handle (NewBase + offset) as start value.
-    const auto *Offset = dyn_cast<SCEVConstant>(StartS->getOperand(0));
-    const auto *NewBase = dyn_cast<SCEVUnknown>(StartS->getOperand(1));
-    if (StartS->getNumOperands() == 2 && Offset && NewBase) {
+    const SCEV *End = AddRec->evaluateAtIteration(
+        SE.getConstant(StartS->getType(), TC - 1), SE);
+
+    // The step recurrence could be negative so it's necessary to find the min
+    // and max accessed addresses in the loop.
+    const SCEV *Min = SE.getUMinExpr(StartS, End);
+    const SCEV *Max = SE.getUMaxExpr(StartS, End);
+    if (isa<SCEVCouldNotCompute>(Min) || isa<SCEVCouldNotCompute>(Max))
+      return false;
+
+    // Now calculate the total access size, which is (max - min) + element_size.
+    const SCEV *Diff = SE.getMinusSCEV(Max, Min);
+    if (isa<SCEVCouldNotCompute>(Diff))
+      return false;
+
+    const SCEV *AS = SE.getAddExpr(
+        Diff, SE.getConstant(Diff->getType(), EltSize.getZExtValue()));
+    auto *ASC = dyn_cast<SCEVConstant>(AS);
+    if (!ASC)
+      return false;
+
+    if (const SCEVUnknown *NewBase = dyn_cast<SCEVUnknown>(Min)) {
----------------
fhahn wrote:

Retrieving the new base and access size here seems to unfortunately add a bit of complexity.

What could make this much simpler would be if we would get the base pointer (underlying object) and the # of known dereferenceable bits. Then construct SCEVs for the start and end of the object and check if the Start/End of the accessed pointer is >= and <= the bounds of the object.

https://github.com/llvm/llvm-project/blob/main/llvm/lib/Analysis/Loads.cpp#L97 would highlight a way to get the number of dereferenceable bytes

https://github.com/llvm/llvm-project/pull/96752


More information about the llvm-commits mailing list