[LoopCacheAnalysis]: handle mismatch type for Numerator and CacheLineSize

[shchenz]

Summary

fix an assertion due to mismatch type for Numerator and CacheLineSize in loop cache analysis pass.

[bmahjour]

LGTM

[jsji]

Commit it on behalf of @shchenz for the timing issue.

llvm/lib/Analysis/LoopCacheAnalysis.cpp

CacheCostTy IndexedReference::computeRefCost(const Loop &L,
   // If the indexed reference is 'consecutive' the cost is
   // (TripCount*Stride)/CLS, otherwise the cost is TripCount.
   const SCEV *RefCost = TripCount;
 
   if (isConsecutive(L, CLS)) {
     const SCEV *Coeff = getLastCoefficient();
     const SCEV *ElemSize = Sizes.back();
     const SCEV *Stride = SE.getMulExpr(Coeff, ElemSize);
-    const SCEV *CacheLineSize = SE.getConstant(Stride->getType(), CLS);
     Type *WiderType = SE.getWiderType(Stride->getType(), TripCount->getType());
+    const SCEV *CacheLineSize = SE.getConstant(WiderType, CLS);
     if (SE.isKnownNegative(Stride))
       Stride = SE.getNegativeSCEV(Stride);
     Stride = SE.getNoopOrAnyExtend(Stride, WiderType);
     TripCount = SE.getNoopOrAnyExtend(TripCount, WiderType);
     const SCEV *Numerator = SE.getMulExpr(Stride, TripCount);
     RefCost = SE.getUDivExpr(Numerator, CacheLineSize);
 
     LLVM_DEBUG(dbgs().indent(4)

llvm/test/Analysis/LoopCacheAnalysis/PowerPC/compute-cost-m32.ll

+; RUN: opt < %s -passes='print<loop-cache-cost>' -disable-output 2>&1 | FileCheck %s
+
+target datalayout = "E-m:a-i64:64-p:32:32-n32-S128-v256:256:256-v512:512:512"
+target triple = "powerpc-ibm-aix7.2.0.0"
+
+; Check IndexedReference::computeRefCost can handle type differences between
+; CacheLineSize and Numerator
+
+; CHECK: Loop '_loop_1_do_' has cost = 2
+
+%_elem_type_of_v = type <{ i32 }>
+
+define signext i32 @foo(%_elem_type_of_v* %v) {
+_entry:
+  br label %_loop_1_do_
+
+_loop_1_do_:                                      ; preds = %_entry, %_loop_1_do_
+  %i.011 = phi i64 [ 1, %_entry ], [ %_loop_1_update_loop_ix, %_loop_1_do_ ]
+  %_conv = trunc i64 %i.011 to i32
+  %_ind_cast = getelementptr %_elem_type_of_v, %_elem_type_of_v* %v, i32 %_conv, i32 0
+  store i32 %_conv, i32* %_ind_cast, align 4
+  %_loop_1_update_loop_ix = add nuw nsw i64 %i.011, 1
+  %_leq_tmp = icmp ult i64 %_loop_1_update_loop_ix, 33
+  br i1 %_leq_tmp, label %_loop_1_do_, label %_loop_1_endl_
+
+_loop_1_endl_:                                    ; preds = %_loop_1_do_
+  ret i32 0
+}