diff --git a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
--- a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
+++ b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
@@ -3094,6 +3094,16 @@
     case Instruction::Store: {
       // Check if the stores are consecutive or if we need to swizzle them.
       llvm::Type *ScalarTy = cast<StoreInst>(VL0)->getValueOperand()->getType();
+      // Avoid types that are padded when being allocated as scalars, while
+      // being packed together in a vector (such as i1).
+      if (DL->getTypeSizeInBits(ScalarTy) !=
+          DL->getTypeAllocSizeInBits(ScalarTy)) {
+        BS.cancelScheduling(VL, VL0);
+        newTreeEntry(VL, None /*not vectorized*/, S, UserTreeIdx,
+                     ReuseShuffleIndicies);
+        LLVM_DEBUG(dbgs() << "SLP: Gathering stores of non-packed type.\n");
+        return;
+      }
       // Make sure all stores in the bundle are simple - we can't vectorize
       // atomic or volatile stores.
       SmallVector<Value *, 4> PointerOps(VL.size());
diff --git a/llvm/test/Transforms/SLPVectorizer/X86/bad_types.ll b/llvm/test/Transforms/SLPVectorizer/X86/bad_types.ll
--- a/llvm/test/Transforms/SLPVectorizer/X86/bad_types.ll
+++ b/llvm/test/Transforms/SLPVectorizer/X86/bad_types.ll
@@ -113,7 +113,6 @@
 
 define void @test4(i32 %a, i28* %ptr) {
 ; Check that we do not vectorize types that are padded to a bigger ones.
-; FIXME: This is not correct! See D94446.
 ;
 ; CHECK-LABEL: @test4(
 ; CHECK-NEXT:  entry:
@@ -121,12 +120,10 @@
 ; CHECK-NEXT:    [[GEP1:%.*]] = getelementptr i28, i28* [[PTR:%.*]], i32 1
 ; CHECK-NEXT:    [[GEP2:%.*]] = getelementptr i28, i28* [[PTR]], i32 2
 ; CHECK-NEXT:    [[GEP3:%.*]] = getelementptr i28, i28* [[PTR]], i32 3
-; CHECK-NEXT:    [[TMP0:%.*]] = insertelement <4 x i28> poison, i28 [[TRUNC]], i32 0
-; CHECK-NEXT:    [[TMP1:%.*]] = insertelement <4 x i28> [[TMP0]], i28 [[TRUNC]], i32 1
-; CHECK-NEXT:    [[TMP2:%.*]] = insertelement <4 x i28> [[TMP1]], i28 [[TRUNC]], i32 2
-; CHECK-NEXT:    [[TMP3:%.*]] = insertelement <4 x i28> [[TMP2]], i28 [[TRUNC]], i32 3
-; CHECK-NEXT:    [[TMP4:%.*]] = bitcast i28* [[PTR]] to <4 x i28>*
-; CHECK-NEXT:    store <4 x i28> [[TMP3]], <4 x i28>* [[TMP4]], align 4
+; CHECK-NEXT:    store i28 [[TRUNC]], i28* [[PTR]], align 4
+; CHECK-NEXT:    store i28 [[TRUNC]], i28* [[GEP1]], align 4
+; CHECK-NEXT:    store i28 [[TRUNC]], i28* [[GEP2]], align 4
+; CHECK-NEXT:    store i28 [[TRUNC]], i28* [[GEP3]], align 4
 ; CHECK-NEXT:    ret void
 ;
 entry: