Fix LoopLoadElimination to keep original alignment on the inital hoisted store

[mehdi_amini]

Summary

This is fixing a bug where Loop Vectorization is widening a load but
with a lower alignment. Hoisting the load without propagating the alignment
will allow inst-combine to later deduce a higher alignment that what the pointer
actually is.

[anemet]

LGTM. Mehdi, thanks very much for fixing this.

llvm/lib/Transforms/Scalar/LoopLoadElimination.cpp

propagateStoredValueToLoadUsers(const StoreToLoadForwardingCandidate &Cand,
     //      store %y, %gep_i_plus_1
 
     Value *Ptr = Cand.Load->getPointerOperand();
     auto *PtrSCEV = cast<SCEVAddRecExpr>(PSE.getSCEV(Ptr));
     auto *PH = L->getLoopPreheader();
     Value *InitialPtr = SEE.expandCodeFor(PtrSCEV->getStart(), Ptr->getType(),
                                           PH->getTerminator());
     Value *Initial =
-        new LoadInst(InitialPtr, "load_initial", PH->getTerminator());
+        new LoadInst(InitialPtr, "load_initial", /* isVolatile */ false,
+                     Cand.Load->getAlignment(), PH->getTerminator());
+
     PHINode *PHI = PHINode::Create(Initial->getType(), 2, "store_forwarded",
                                    &L->getHeader()->front());
     PHI->addIncoming(Initial, PH);
     PHI->addIncoming(Cand.Store->getOperand(0), L->getLoopLatch());
 
     Cand.Load->replaceAllUsesWith(PHI);
   }
 

llvm/test/Transforms/LoopLoadElim/forward.ll

 
 define void @f(i32* %A, i32* %B, i32* %C, i64 %N) {
 
 ; CHECK:   for.body.lver.check:
 ; CHECK:     %found.conflict{{.*}} =
 ; CHECK-NOT: %found.conflict{{.*}} =
 
 entry:
-; for.body.ph:
-; CHECK: %load_initial = load i32, i32* %A
+; Make sure the hoisted load keeps the alignment
+; CHECK: %load_initial = load i32, i32* %A, align 1
   br label %for.body
 
 for.body:                                         ; preds = %for.body, %entry
 ; CHECK: %store_forwarded = phi i32 [ %load_initial, %for.body.ph ], [ %a_p1, %for.body ]
   %indvars.iv = phi i64 [ 0, %entry ], [ %indvars.iv.next, %for.body ]
   %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1
 
   %Aidx_next = getelementptr inbounds i32, i32* %A, i64 %indvars.iv.next
   %Bidx = getelementptr inbounds i32, i32* %B, i64 %indvars.iv
   %Cidx = getelementptr inbounds i32, i32* %C, i64 %indvars.iv
   %Aidx = getelementptr inbounds i32, i32* %A, i64 %indvars.iv
 
   %b = load i32, i32* %Bidx, align 4
   %a_p1 = add i32 %b, 2
   store i32 %a_p1, i32* %Aidx_next, align 4
 
-  %a = load i32, i32* %Aidx, align 4
+  %a = load i32, i32* %Aidx, align 1
 ; CHECK: %c = mul i32 %store_forwarded, 2
   %c = mul i32 %a, 2
   store i32 %c, i32* %Cidx, align 4
 
   %exitcond = icmp eq i64 %indvars.iv.next, %N
   br i1 %exitcond, label %for.end, label %for.body
 
 for.end:                                          ; preds = %for.body
   ret void
 }