Index: llvm/lib/Transforms/InstCombine/InstCombineVectorOps.cpp
===================================================================
--- llvm/lib/Transforms/InstCombine/InstCombineVectorOps.cpp
+++ llvm/lib/Transforms/InstCombine/InstCombineVectorOps.cpp
@@ -1889,9 +1889,9 @@
 Instruction *InstCombiner::visitShuffleVectorInst(ShuffleVectorInst &SVI) {
   Value *LHS = SVI.getOperand(0);
   Value *RHS = SVI.getOperand(1);
-  if (auto *V =
-          SimplifyShuffleVectorInst(LHS, RHS, SVI.getShuffleMask(),
-                                    SVI.getType(), SQ.getWithInstruction(&SVI)))
+  SimplifyQuery ShufQuery = SQ.getWithInstruction(&SVI);
+  if (auto *V = SimplifyShuffleVectorInst(LHS, RHS, SVI.getShuffleMask(),
+                                          SVI.getType(), ShufQuery))
     return replaceInstUsesWith(SVI, V);
 
   // shuffle x, x, mask --> shuffle x, undef, mask'
@@ -1899,6 +1899,32 @@
   unsigned LHSWidth = LHS->getType()->getVectorNumElements();
   ArrayRef<int> Mask = SVI.getShuffleMask();
   Type *Int32Ty = Type::getInt32Ty(SVI.getContext());
+
+  // Peek through a bitcasted shuffle operand by scaling the mask. If the
+  // simulated shuffle can simplify, then this shuffle is unnecessary:
+  // shuf (bitcast X), undef, Mask --> bitcast X'
+  // TODO: This could be extended to allow length-changing shuffles and/or casts
+  //       to narrower elements. The transform might also be obsoleted if we
+  //       allowed canonicalization of bitcasted shuffles.
+  Value *X;
+  if (match(LHS, m_BitCast(m_Value(X))) && match(RHS, m_Undef()) &&
+      X->getType()->isVectorTy() && VWidth == LHSWidth &&
+      X->getType()->getVectorNumElements() >= VWidth) {
+    // Create the scaled mask constant.
+    Type *XType = X->getType();
+    unsigned XNumElts = XType->getVectorNumElements();
+    assert(XNumElts % VWidth == 0 && "Unexpected vector bitcast");
+    unsigned ScaleFactor = XNumElts / VWidth;
+    SmallVector<int, 16> ScaledMask;
+    scaleShuffleMask(ScaleFactor, Mask, ScaledMask);
+
+    // If the shuffled source vector simplifies, cast that value to this
+    // shuffle's type.
+    if (auto *V = SimplifyShuffleVectorInst(X, UndefValue::get(XType),
+                                            ScaledMask, XType, ShufQuery))
+      return BitCastInst::Create(Instruction::BitCast, V, SVI.getType());
+  }
+
   if (LHS == RHS) {
     assert(!isa<UndefValue>(RHS) && "Shuffle with 2 undef ops not simplified?");
     // Remap any references to RHS to use LHS.
Index: llvm/test/Transforms/InstCombine/shufflevec-bitcast.ll
===================================================================
--- llvm/test/Transforms/InstCombine/shufflevec-bitcast.ll
+++ llvm/test/Transforms/InstCombine/shufflevec-bitcast.ll
@@ -21,11 +21,12 @@
   ret void
 }
 
+; Shuffle-of-bitcast-splat --> splat-bitcast
+
 define <4 x i16> @splat_bitcast_operand(<8 x i8> %x) {
 ; CHECK-LABEL: @splat_bitcast_operand(
-; CHECK-NEXT:    [[S1:%.*]] = shufflevector <8 x i8> [[X:%.*]], <8 x i8> undef, <8 x i32> <i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 undef, i32 undef>
-; CHECK-NEXT:    [[BC:%.*]] = bitcast <8 x i8> [[S1]] to <4 x i16>
-; CHECK-NEXT:    [[S2:%.*]] = shufflevector <4 x i16> [[BC]], <4 x i16> undef, <4 x i32> <i32 0, i32 2, i32 1, i32 0>
+; CHECK-NEXT:    [[S1:%.*]] = shufflevector <8 x i8> [[X:%.*]], <8 x i8> undef, <8 x i32> <i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1>
+; CHECK-NEXT:    [[S2:%.*]] = bitcast <8 x i8> [[S1]] to <4 x i16>
 ; CHECK-NEXT:    ret <4 x i16> [[S2]]
 ;
   %s1 = shufflevector <8 x i8> %x, <8 x i8> undef, <8 x i32> <i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1>
@@ -34,12 +35,14 @@
   ret <4 x i16> %s2
 }
 
+; Shuffle-of-bitcast-splat --> splat-bitcast
+
 define <4 x i16> @splat_bitcast_operand_uses(<8 x i8> %x) {
 ; CHECK-LABEL: @splat_bitcast_operand_uses(
 ; CHECK-NEXT:    [[S1:%.*]] = shufflevector <8 x i8> [[X:%.*]], <8 x i8> undef, <8 x i32> <i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1>
 ; CHECK-NEXT:    [[BC:%.*]] = bitcast <8 x i8> [[S1]] to <4 x i16>
 ; CHECK-NEXT:    call void @use(<4 x i16> [[BC]])
-; CHECK-NEXT:    [[S2:%.*]] = shufflevector <4 x i16> [[BC]], <4 x i16> undef, <4 x i32> <i32 0, i32 2, i32 1, i32 0>
+; CHECK-NEXT:    [[S2:%.*]] = bitcast <8 x i8> [[S1]] to <4 x i16>
 ; CHECK-NEXT:    ret <4 x i16> [[S2]]
 ;
   %s1 = shufflevector <8 x i8> %x, <8 x i8> undef, <8 x i32> <i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1>
@@ -49,11 +52,12 @@
   ret <4 x i16> %s2
 }
 
+; Shuffle-of-bitcast-splat --> splat-bitcast
+
 define <4 x i32> @splat_bitcast_operand_same_size_src_elt(<4 x float> %x) {
 ; CHECK-LABEL: @splat_bitcast_operand_same_size_src_elt(
-; CHECK-NEXT:    [[S1:%.*]] = shufflevector <4 x float> [[X:%.*]], <4 x float> undef, <4 x i32> <i32 2, i32 2, i32 2, i32 undef>
-; CHECK-NEXT:    [[BC:%.*]] = bitcast <4 x float> [[S1]] to <4 x i32>
-; CHECK-NEXT:    [[S2:%.*]] = shufflevector <4 x i32> [[BC]], <4 x i32> undef, <4 x i32> <i32 0, i32 2, i32 1, i32 0>
+; CHECK-NEXT:    [[S1:%.*]] = shufflevector <4 x float> [[X:%.*]], <4 x float> undef, <4 x i32> <i32 2, i32 2, i32 2, i32 2>
+; CHECK-NEXT:    [[S2:%.*]] = bitcast <4 x float> [[S1]] to <4 x i32>
 ; CHECK-NEXT:    ret <4 x i32> [[S2]]
 ;
   %s1 = shufflevector <4 x float> %x, <4 x float> undef, <4 x i32> <i32 2, i32 2, i32 2, i32 2>
@@ -66,9 +70,7 @@
 
 define <4 x i32> @shuf_bitcast_operand(<16 x i8> %x) {
 ; CHECK-LABEL: @shuf_bitcast_operand(
-; CHECK-NEXT:    [[S1:%.*]] = shufflevector <16 x i8> [[X:%.*]], <16 x i8> undef, <16 x i32> <i32 12, i32 13, i32 14, i32 15, i32 8, i32 9, i32 10, i32 11, i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>
-; CHECK-NEXT:    [[BC:%.*]] = bitcast <16 x i8> [[S1]] to <4 x i32>
-; CHECK-NEXT:    [[S2:%.*]] = shufflevector <4 x i32> [[BC]], <4 x i32> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; CHECK-NEXT:    [[S2:%.*]] = bitcast <16 x i8> [[X:%.*]] to <4 x i32>
 ; CHECK-NEXT:    ret <4 x i32> [[S2]]
 ;
   %s1 = shufflevector <16 x i8> %x, <16 x i8> undef, <16 x i32> <i32 12, i32 13, i32 14, i32 15, i32 8, i32 9, i32 10, i32 11, i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>
@@ -77,6 +79,8 @@
   ret <4 x i32> %s2
 }
 
+; TODO: Could allow fold for length-changing shuffles.
+
 define <5 x i16> @splat_bitcast_operand_change_type(<8 x i8> %x) {
 ; CHECK-LABEL: @splat_bitcast_operand_change_type(
 ; CHECK-NEXT:    [[S1:%.*]] = shufflevector <8 x i8> [[X:%.*]], <8 x i8> undef, <8 x i32> <i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1>
@@ -90,6 +94,8 @@
   ret <5 x i16> %s2
 }
 
+; TODO: Could allow fold for cast to narrow element.
+
 define <4 x i16> @splat_bitcast_operand_wider_src_elt(<2 x i32> %x) {
 ; CHECK-LABEL: @splat_bitcast_operand_wider_src_elt(
 ; CHECK-NEXT:    [[S1:%.*]] = shufflevector <2 x i32> [[X:%.*]], <2 x i32> undef, <2 x i32> <i32 1, i32 undef>