diff --git a/llvm/lib/Transforms/IPO/GlobalOpt.cpp b/llvm/lib/Transforms/IPO/GlobalOpt.cpp
--- a/llvm/lib/Transforms/IPO/GlobalOpt.cpp
+++ b/llvm/lib/Transforms/IPO/GlobalOpt.cpp
@@ -1598,6 +1598,33 @@
     if (optimizeOnceStoredGlobal(GV, StoredOnceValue, GS.Ordering, DL, GetTLI))
       return true;
 
+    // If the store dominates any loads, those loads will always load the
+    // stored value (as opposed to the initializer), even in the presence of
+    // recursion. Restrict to the case where all loads/stores of the global are
+    // in the same function.
+    if (!GS.HasMultipleAccessingFunctions && GS.AccessingFunction) {
+      SmallVector<LoadInst *> Loads;
+      for (User *U : GV->users()) {
+        if (auto *LI = dyn_cast<LoadInst>(U))
+          Loads.push_back(LI);
+      }
+      // Only compute DT if we have any loads to examine.
+      bool MadeChange = false;
+      if (!Loads.empty()) {
+        auto &DT = LookupDomTree(*const_cast<Function *>(GS.AccessingFunction));
+        for (auto *LI : Loads) {
+          if (LI->getType() == StoredOnceValue->getType() &&
+              DT.dominates(GS.StoredOnceStore, LI)) {
+            LI->replaceAllUsesWith(StoredOnceValue);
+            LI->eraseFromParent();
+            MadeChange = true;
+          }
+        }
+      }
+      if (MadeChange)
+        return true;
+    }
+
     // Otherwise, if the global was not a boolean, we can shrink it to be a
     // boolean. Skip this optimization for AS that doesn't allow an initializer.
     if (SOVConstant && GS.Ordering == AtomicOrdering::NotAtomic &&
diff --git a/llvm/test/Transforms/GlobalOpt/stored-once-forward-value.ll b/llvm/test/Transforms/GlobalOpt/stored-once-forward-value.ll
--- a/llvm/test/Transforms/GlobalOpt/stored-once-forward-value.ll
+++ b/llvm/test/Transforms/GlobalOpt/stored-once-forward-value.ll
@@ -10,10 +10,8 @@
 
 define i1 @dom_const() {
 ; CHECK-LABEL: @dom_const(
-; CHECK-NEXT:    store i1 true, ptr @g1, align 1
 ; CHECK-NEXT:    call void @b()
-; CHECK-NEXT:    [[R:%.*]] = load i1, ptr @g1, align 1
-; CHECK-NEXT:    ret i1 [[R]]
+; CHECK-NEXT:    ret i1 true
 ;
   store i1 true, ptr @g1
   call void @b()
@@ -23,10 +21,8 @@
 
 define i32 @dom_arg(i32 %a) {
 ; CHECK-LABEL: @dom_arg(
-; CHECK-NEXT:    store i32 [[A:%.*]], ptr @g2, align 4
 ; CHECK-NEXT:    call void @b()
-; CHECK-NEXT:    [[R:%.*]] = load i32, ptr @g2, align 4
-; CHECK-NEXT:    ret i32 [[R]]
+; CHECK-NEXT:    ret i32 [[A:%.*]]
 ;
   store i32 %a, ptr @g2
   call void @b()