[LoopVectorizer] Fix in getScalarizationOverhead()

[jonpa]

Summary

If target returns false in TTI.prefersVectorizedAddressing(), it means the address registers will not need to be extracted. Therefore, there should be no operands scalarization overhead for a load instruction.

Test for SystemZ.

[hfinkel]

LGTM

[jonpa]

Thanks for review.

Updated just the test to make it more clear that the address computation is in fact scalar (the multiply).

[uweigand]

LGTM as well.

[jonpa]

Thanks for review. I am waiting with committing this until https://reviews.llvm.org/D52417 is approved, which is a simple fix I found necessary after applying this patch.

[jonpa]

Committed in r345603.

lib/Transforms/Vectorize/LoopVectorize.cpp

static unsigned getScalarizationOverhead(Instruction *I, unsigned VF,
 
   unsigned Cost = 0;
   Type *RetTy = ToVectorTy(I->getType(), VF);
   if (!RetTy->isVoidTy() &&
       (!isa<LoadInst>(I) ||
        !TTI.supportsEfficientVectorElementLoadStore()))
     Cost += TTI.getScalarizationOverhead(RetTy, true, false);
 
+  // Some targets keep addresses scalar.
+  if (isa<LoadInst>(I) && !TTI.prefersVectorizedAddressing())
+    return Cost;
+
   if (CallInst *CI = dyn_cast<CallInst>(I)) {
     SmallVector<const Value *, 4> Operands(CI->arg_operands());
     Cost += TTI.getOperandsScalarizationOverhead(Operands, VF);
   }
   else if (!isa<StoreInst>(I) ||
            !TTI.supportsEfficientVectorElementLoadStore()) {
     SmallVector<const Value *, 4> Operands(I->operand_values());
     Cost += TTI.getOperandsScalarizationOverhead(Operands, VF);

test/Transforms/LoopVectorize/SystemZ/load-scalarization-cost-0.ll

+; RUN: opt -mtriple=s390x-unknown-linux -mcpu=z13 -loop-vectorize \
+; RUN:   -force-vector-width=2 -debug-only=loop-vectorize \
+; RUN:   -disable-output < %s 2>&1 | FileCheck %s
+; REQUIRES: asserts
+;
+; Check that a scalarized load does not get operands scalarization costs added.
+
+define void @fun(i64* %data, i64 %n, i64 %s, double* %Src) {
+entry:
+  br label %for.body
+
+for.body:
+  %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]
+  %mul = mul nsw i64 %iv, %s
+  %gep = getelementptr inbounds double, double* %Src, i64 %mul
+  %bct = bitcast double* %gep to i64*
+  %ld = load i64, i64* %bct
+  %iv.next = add nuw nsw i64 %iv, 1
+  %cmp110.us = icmp slt i64 %iv.next, %n
+  br i1 %cmp110.us, label %for.body, label %for.end
+
+for.end:
+  ret void
+
+; CHECK: LV: Found an estimated cost of 2 for VF 2 For instruction:   %mul = mul nsw i64 %iv, %s
+; CHECK: LV: Found an estimated cost of 2 for VF 2 For instruction:   %ld = load i64, i64* %bct
+}