[LV] Clamp the Unroll Factor to at least 1

[simoll]

Summary

LV requires the unroll factor (UF) to be >= 1. This patch adds an explicit assertion for this and fixes a case in the cost model where the UF was set to zero. This was exposed by an out-of-tree TTI implementation that reported zero scores for candidate vectorization factors.

[fhahn]

Is the problem TTI.getMaxInterleaveFactor() returning 0? If that's the case, I think it would be better to make sure TTI.getMaxInterleaveFactor() returns at least 1.

[fhahn]

Marking as changes requested to clear up review queue, as this has a pending comment.

llvm/lib/Transforms/Vectorize/LoopVectorize.cpp

unsigned LoopVectorizationCostModel::selectInterleaveCount(unsigned VF,
   } else {
     if (ForceTargetMaxVectorInterleaveFactor.getNumOccurrences() > 0)
       MaxInterleaveCount = ForceTargetMaxVectorInterleaveFactor;
   }
 
   // If trip count is known or estimated compile time constant, limit the
   // interleave count to be less than the trip count divided by VF.
   if (BestKnownTC) {
-    MaxInterleaveCount = std::min(*BestKnownTC / VF, MaxInterleaveCount);
+    MaxInterleaveCount =
+        std::max(1u, std::min(*BestKnownTC / VF, MaxInterleaveCount));
   }
 
   // If we did not calculate the cost for VF (because the user selected the VF)
   // then we calculate the cost of VF here.
   if (LoopCost == 0)
     LoopCost = expectedCost(VF).first;
 
   assert(LoopCost && "Non-zero loop cost expected");
 
   // Clamp the calculated IC to be between the 1 and the max interleave count
   // that the target and trip count allows.
   if (IC > MaxInterleaveCount)
     IC = MaxInterleaveCount;
   else if (IC < 1)
     IC = 1;
 
   // Interleave if we vectorized this loop and there is a reduction that could
   // benefit from interleaving.
   if (VF > 1 && !Legal->getReductionVars().empty()) {
     LLVM_DEBUG(dbgs() << "LV: Interleaving because of reductions.\n");
+    assert(IC > 0);
     return IC;
   }
 
   // Note that if we've already vectorized the loop we will have done the
   // runtime check and so interleaving won't require further checks.
   bool InterleavingRequiresRuntimePointerCheck =
       (VF == 1 && Legal->getRuntimePointerChecking()->Need);