Index: llvm/lib/Transforms/Scalar/GVN.cpp
===================================================================
--- llvm/lib/Transforms/Scalar/GVN.cpp
+++ llvm/lib/Transforms/Scalar/GVN.cpp
@@ -1398,6 +1398,10 @@
       NewLoad->setMetadata(LLVMContext::MD_invariant_group, InvGroupMD);
     if (auto *RangeMD = LI->getMetadata(LLVMContext::MD_range))
       NewLoad->setMetadata(LLVMContext::MD_range, RangeMD);
+    if (auto *AccessMD = LI->getMetadata(LLVMContext::MD_access_group))
+      if (this->LI && this->LI->getLoopFor(LI->getParent()) ==
+                          this->LI->getLoopFor(UnavailablePred))
+        NewLoad->setMetadata(LLVMContext::MD_access_group, AccessMD);
 
     // We do not propagate the old load's debug location, because the new
     // load now lives in a different BB, and we want to avoid a jumpy line
Index: llvm/test/Transforms/GVN/PRE/load-pre-metadata-accsess-group.ll
===================================================================
--- /dev/null
+++ llvm/test/Transforms/GVN/PRE/load-pre-metadata-accsess-group.ll
@@ -0,0 +1,56 @@
+; RUN: opt < %s -basic-aa -gvn -enable-load-pre -S | FileCheck %s
+
+define dso_local void @test1(i32* nocapture readonly %aa, i32* nocapture readonly %bb, i32* nocapture readonly %cc, i32* nocapture readonly %dd, i32* nocapture %ee, i32* nocapture %ff) local_unnamed_addr {
+; CHECK-LABEL: @test1(
+entry:
+  %0 = load i32, i32* %aa, align 4
+  %idxprom = sext i32 %0 to i64
+  %arrayidx = getelementptr inbounds i32, i32* %ee, i64 %idxprom
+  %1 = load i32, i32* %arrayidx, align 4
+  %arrayidx2 = getelementptr inbounds i32, i32* %ff, i64 %idxprom
+  %2 = load i32, i32* %arrayidx2, align 4
+  %arrayidx4 = getelementptr inbounds i32, i32* %dd, i64 %idxprom
+  %3 = load i32, i32* %arrayidx4, align 4
+  %div = sdiv i32 %2, %3
+  %sub = sub nsw i32 %1, %div
+  store i32 %sub, i32* %arrayidx, align 4
+  %arrayidx8 = getelementptr inbounds i32, i32* %bb, i64 %idxprom
+  %4 = load i32, i32* %arrayidx8, align 4
+  %cmp37 = icmp sgt i32 %4, 0
+  br i1 %cmp37, label %for.body.preheader, label %for.end
+
+for.body.preheader:
+  %wide.trip.count = zext i32 %4 to i64
+  br label %for.body
+
+for.body:
+  %indvars.iv = phi i64 [ 0, %for.body.preheader ], [ %indvars.iv.next, %for.body ]
+  %arrayidx10 = getelementptr inbounds i32, i32* %cc, i64 %indvars.iv
+  %5 = load i32, i32* %arrayidx10, align 4, !llvm.access.group !1
+  %idxprom11 = sext i32 %5 to i64
+  %arrayidx12 = getelementptr inbounds i32, i32* %ff, i64 %idxprom11
+  %6 = load i32, i32* %arrayidx12, align 4, !llvm.access.group !1
+  %arrayidx14 = getelementptr inbounds i32, i32* %dd, i64 %indvars.iv
+  %7 = load i32, i32* %arrayidx14, align 4, !llvm.access.group !1
+  %8 = load i32, i32* %arrayidx, align 4, !llvm.access.group !1
+  %mul = mul nsw i32 %8, %7
+  %add = add nsw i32 %mul, %6
+  store i32 %add, i32* %arrayidx12, align 4, !llvm.access.group !1
+  %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1
+  %exitcond = icmp ne i64 %indvars.iv.next, %wide.trip.count
+  br i1 %exitcond, label %for.body, label %for.end.loopexit, !llvm.loop !2
+
+;CHECK-LABEL: for.body.for.body_crit_edge:
+;CHECK-NEXT: %.pre = load i32, i32* %arrayidx, align 4, !llvm.access.group !0
+
+for.end.loopexit:
+  br label %for.end
+
+for.end:
+  ret void
+}
+
+!1 = distinct !{}
+!2 = distinct !{!2, !3, !4}
+!3 = !{!"llvm.loop.parallel_accesses", !1}
+!4 = !{!"llvm.loop.vectorize.enable", i1 true}