[VectorCombine] transform bitcasted shuffle to wider elements

bitcast (shuf V, MaskC) --> shuf (bitcast V), MaskC'

This is the widen shuffle elements enhancement to D76727.
It builds on the analysis and simplifications in
D77881 and rG6a7e958a423e.

The phase ordering tests show that we can simplify inverse
shuffles across a binop in both directions (widen/narrow or
narrow/widen) now.

There's another potential transform visible in some of the
remaining TODOs - move a bitcasted operand of a shuffle
after the shuffle.

Differential Revision: https://reviews.llvm.org/D78371

Author: rotateright

llvm/lib/Transforms/Vectorize/VectorCombine.cpp

@@ -259,42 +259,49 @@ static bool foldExtractExtract(Instruction &I, const TargetTransformInfo &TTI) {
   return true;
 }
 
-/// If this is a bitcast to narrow elements from a shuffle of wider elements,
-/// try to bitcast the source vector to the narrow type followed by shuffle.
-/// This can enable further transforms by moving bitcasts or shuffles together.
+/// If this is a bitcast of a shuffle, try to bitcast the source vector to the
+/// destination type followed by shuffle. This can enable further transforms by
+/// moving bitcasts or shuffles together.
 static bool foldBitcastShuf(Instruction &I, const TargetTransformInfo &TTI) {
   Value *V;
   ArrayRef<int> Mask;
   if (!match(&I, m_BitCast(m_OneUse(m_ShuffleVector(m_Value(V), m_Undef(),
                                                     m_Mask(Mask))))))
     return false;
 
+  // Disallow non-vector casts and length-changing shuffles.
+  // TODO: We could allow any shuffle.
   auto *DestTy = dyn_cast<VectorType>(I.getType());
   auto *SrcTy = cast<VectorType>(V->getType());
   if (!DestTy || I.getOperand(0)->getType() != SrcTy)
     return false;
 
-  // TODO: Handle bitcast from narrow element type to wide element type.
-  unsigned DestNumElts = DestTy->getNumElements();
-  unsigned SrcNumElts = SrcTy->getNumElements();
-  if (SrcNumElts > DestNumElts)
-    return false;
-
   // The new shuffle must not cost more than the old shuffle. The bitcast is
   // moved ahead of the shuffle, so assume that it has the same cost as before.
   if (TTI.getShuffleCost(TargetTransformInfo::SK_PermuteSingleSrc, DestTy) >
       TTI.getShuffleCost(TargetTransformInfo::SK_PermuteSingleSrc, SrcTy))
     return false;
 
-  // Bitcast the source vector and expand the shuffle mask to the equivalent for
-  // narrow elements.
+  unsigned DestNumElts = DestTy->getNumElements();
+  unsigned SrcNumElts = SrcTy->getNumElements();
+  SmallVector<int, 16> NewMask;
+  if (SrcNumElts <= DestNumElts) {
+    // The bitcast is from wide to narrow/equal elements. The shuffle mask can
+    // always be expanded to the equivalent form choosing narrower elements.
+    assert(DestNumElts % SrcNumElts == 0 && "Unexpected shuffle mask");
+    unsigned ScaleFactor = DestNumElts / SrcNumElts;
+    narrowShuffleMaskElts(ScaleFactor, Mask, NewMask);
+  } else {
+    // The bitcast is from narrow elements to wide elements. The shuffle mask
+    // must choose consecutive elements to allow casting first.
+    assert(SrcNumElts % DestNumElts == 0 && "Unexpected shuffle mask");
+    unsigned ScaleFactor = SrcNumElts / DestNumElts;
+    if (!widenShuffleMaskElts(ScaleFactor, Mask, NewMask))
+      return false;
+  }
   // bitcast (shuf V, MaskC) --> shuf (bitcast V), MaskC'
   IRBuilder<> Builder(&I);
   Value *CastV = Builder.CreateBitCast(V, DestTy);
-  SmallVector<int, 16> NewMask;
-  assert(DestNumElts % SrcNumElts == 0 && "Unexpected shuffle mask");
-  unsigned ScaleFactor = DestNumElts / SrcNumElts;
-  narrowShuffleMaskElts(ScaleFactor, Mask, NewMask);
   Value *Shuf = Builder.CreateShuffleVector(CastV, UndefValue::get(DestTy),
                                             NewMask);
   I.replaceAllUsesWith(Shuf);

llvm/test/Transforms/PhaseOrdering/X86/shuffle.ll

@@ -47,17 +47,13 @@ define <2 x i64> @shuffle_32_add_8_shuffle_32_masks_are_eq(<2 x i64> %v) {
   ret <2 x i64> %bc5
 }
 
-; TODO: Eliminate redundant shuffles
+; Eliminate redundant shuffles
 
 define <2 x i64> @shuffle_8_add_32_shuffle_8_masks_are_eq(<2 x i64> %v) {
 ; CHECK-LABEL: @shuffle_8_add_32_shuffle_8_masks_are_eq(
-; CHECK-NEXT:    [[BC0:%.*]] = bitcast <2 x i64> [[V:%.*]] to <16 x i8>
-; CHECK-NEXT:    [[SHUFFLE:%.*]] = shufflevector <16 x i8> [[BC0]], <16 x i8> undef, <16 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
-; CHECK-NEXT:    [[BC2:%.*]] = bitcast <16 x i8> [[SHUFFLE]] to <4 x i32>
-; CHECK-NEXT:    [[ADD_I:%.*]] = shl <4 x i32> [[BC2]], <i32 1, i32 1, i32 1, i32 1>
-; CHECK-NEXT:    [[BC4:%.*]] = bitcast <4 x i32> [[ADD_I]] to <16 x i8>
-; CHECK-NEXT:    [[SHUFFLE4:%.*]] = shufflevector <16 x i8> [[BC4]], <16 x i8> undef, <16 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
-; CHECK-NEXT:    [[BC5:%.*]] = bitcast <16 x i8> [[SHUFFLE4]] to <2 x i64>
+; CHECK-NEXT:    [[TMP1:%.*]] = bitcast <2 x i64> [[V:%.*]] to <4 x i32>
+; CHECK-NEXT:    [[TMP2:%.*]] = shl <4 x i32> [[TMP1]], <i32 1, i32 1, i32 1, i32 1>
+; CHECK-NEXT:    [[BC5:%.*]] = bitcast <4 x i32> [[TMP2]] to <2 x i64>
 ; CHECK-NEXT:    ret <2 x i64> [[BC5]]
 ;
   %bc0 = bitcast <2 x i64> %v to <16 x i8>
@@ -126,15 +122,14 @@ define <16 x i8> @shuffle_16_add_8_masks_are_eq(<8 x i16> %v1, <8 x i16> %v2) {
   ret <16 x i8> %add
 }
 
-; TODO: Sink single shuffle.
+; Sink single shuffle.
 
 define <4 x i32> @shuffle_16_add_32_masks_are_eq_and_can_be_converted_up(<8 x i16> %v1, <8 x i16> %v2) {
 ; CHECK-LABEL: @shuffle_16_add_32_masks_are_eq_and_can_be_converted_up(
-; CHECK-NEXT:    [[SHUFFLE1:%.*]] = shufflevector <8 x i16> [[V1:%.*]], <8 x i16> undef, <8 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>
-; CHECK-NEXT:    [[SHUFFLE2:%.*]] = shufflevector <8 x i16> [[V2:%.*]], <8 x i16> undef, <8 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>
-; CHECK-NEXT:    [[BC1:%.*]] = bitcast <8 x i16> [[SHUFFLE1]] to <4 x i32>
-; CHECK-NEXT:    [[BC2:%.*]] = bitcast <8 x i16> [[SHUFFLE2]] to <4 x i32>
-; CHECK-NEXT:    [[ADD:%.*]] = add <4 x i32> [[BC2]], [[BC1]]
+; CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i16> [[V1:%.*]] to <4 x i32>
+; CHECK-NEXT:    [[TMP2:%.*]] = bitcast <8 x i16> [[V2:%.*]] to <4 x i32>
+; CHECK-NEXT:    [[TMP3:%.*]] = add <4 x i32> [[TMP2]], [[TMP1]]
+; CHECK-NEXT:    [[ADD:%.*]] = shufflevector <4 x i32> [[TMP3]], <4 x i32> undef, <4 x i32> <i32 2, i32 3, i32 0, i32 1>
 ; CHECK-NEXT:    ret <4 x i32> [[ADD]]
 ;
   %shuffle1 = shufflevector <8 x i16> %v1, <8 x i16> undef, <8 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>
@@ -145,15 +140,14 @@ define <4 x i32> @shuffle_16_add_32_masks_are_eq_and_can_be_converted_up(<8 x i1
   ret <4 x i32> %add
 }
 
-; TODO: Sink single shuffle.
+; Sink single shuffle.
 
 define <4 x i32> @shuffle_8_add_32_masks_are_eq_and_can_be_converted_up(<16 x i8> %v1, <16 x i8> %v2) {
 ; CHECK-LABEL: @shuffle_8_add_32_masks_are_eq_and_can_be_converted_up(
-; CHECK-NEXT:    [[SHUFFLE1:%.*]] = shufflevector <16 x i8> [[V1:%.*]], <16 x i8> undef, <16 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
-; CHECK-NEXT:    [[SHUFFLE2:%.*]] = shufflevector <16 x i8> [[V2:%.*]], <16 x i8> undef, <16 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
-; CHECK-NEXT:    [[BC1:%.*]] = bitcast <16 x i8> [[SHUFFLE1]] to <4 x i32>
-; CHECK-NEXT:    [[BC2:%.*]] = bitcast <16 x i8> [[SHUFFLE2]] to <4 x i32>
-; CHECK-NEXT:    [[ADD:%.*]] = add <4 x i32> [[BC2]], [[BC1]]
+; CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[V1:%.*]] to <4 x i32>
+; CHECK-NEXT:    [[TMP2:%.*]] = bitcast <16 x i8> [[V2:%.*]] to <4 x i32>
+; CHECK-NEXT:    [[TMP3:%.*]] = add <4 x i32> [[TMP2]], [[TMP1]]
+; CHECK-NEXT:    [[ADD:%.*]] = shufflevector <4 x i32> [[TMP3]], <4 x i32> undef, <4 x i32> <i32 1, i32 0, i32 2, i32 3>
 ; CHECK-NEXT:    ret <4 x i32> [[ADD]]
 ;
   %shuffle1 = shufflevector <16 x i8> %v1, <16 x i8> undef, <16 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
@@ -229,13 +223,13 @@ define <16 x i8> @shuffle_32_bitcast_8_shuffle_8_can_not_be_converted_up(<4 x i3
 }
 
 ; shuffle<4 x i32>( bitcast<4 x i32>( shuffle<16 x i8>(v)))
-; TODO: Narrow, squash shuffles, and widen type?
+; TODO: squash shuffles?
 
 define <4 x i32> @shuffle_8_bitcast_32_shuffle_32_can_be_converted_up(<16 x i8> %v1) {
 ; CHECK-LABEL: @shuffle_8_bitcast_32_shuffle_32_can_be_converted_up(
-; CHECK-NEXT:    [[SHUFFLE1:%.*]] = shufflevector <16 x i8> [[V1:%.*]], <16 x i8> undef, <16 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>
-; CHECK-NEXT:    [[BC1:%.*]] = bitcast <16 x i8> [[SHUFFLE1]] to <4 x i32>
-; CHECK-NEXT:    [[SHUFFLE2:%.*]] = shufflevector <4 x i32> [[BC1]], <4 x i32> undef, <4 x i32> <i32 2, i32 3, i32 1, i32 0>
+; CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[V1:%.*]] to <4 x i32>
+; CHECK-NEXT:    [[TMP2:%.*]] = shufflevector <4 x i32> [[TMP1]], <4 x i32> undef, <4 x i32> <i32 2, i32 3, i32 1, i32 0>
+; CHECK-NEXT:    [[SHUFFLE2:%.*]] = shufflevector <4 x i32> [[TMP2]], <4 x i32> undef, <4 x i32> <i32 2, i32 3, i32 1, i32 0>
 ; CHECK-NEXT:    ret <4 x i32> [[SHUFFLE2]]
 ;
   %shuffle1 = shufflevector <16 x i8> %v1, <16 x i8> undef, <16 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>
@@ -245,13 +239,13 @@ define <4 x i32> @shuffle_8_bitcast_32_shuffle_32_can_be_converted_up(<16 x i8>
 }
 
 ; shuffle<4 x i32>( bitcast<4 x i32>( shuffle<8 x i16>(v)))
-; TODO: Narrow, squash shuffles, and widen type?
+; TODO: squash shuffles?
 
 define <4 x i32> @shuffle_16_bitcast_32_shuffle_32_can_be_converted_up(<8 x i16> %v1) {
 ; CHECK-LABEL: @shuffle_16_bitcast_32_shuffle_32_can_be_converted_up(
-; CHECK-NEXT:    [[SHUFFLE1:%.*]] = shufflevector <8 x i16> [[V1:%.*]], <8 x i16> undef, <8 x i32> <i32 4, i32 5, i32 6, i32 7, i32 2, i32 3, i32 0, i32 1>
-; CHECK-NEXT:    [[BC1:%.*]] = bitcast <8 x i16> [[SHUFFLE1]] to <4 x i32>
-; CHECK-NEXT:    [[SHUFFLE2:%.*]] = shufflevector <4 x i32> [[BC1]], <4 x i32> undef, <4 x i32> <i32 2, i32 3, i32 1, i32 0>
+; CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i16> [[V1:%.*]] to <4 x i32>
+; CHECK-NEXT:    [[TMP2:%.*]] = shufflevector <4 x i32> [[TMP1]], <4 x i32> undef, <4 x i32> <i32 2, i32 3, i32 1, i32 0>
+; CHECK-NEXT:    [[SHUFFLE2:%.*]] = shufflevector <4 x i32> [[TMP2]], <4 x i32> undef, <4 x i32> <i32 2, i32 3, i32 1, i32 0>
 ; CHECK-NEXT:    ret <4 x i32> [[SHUFFLE2]]
 ;
   %shuffle1 = shufflevector <8 x i16> %v1, <8 x i16> undef, <8 x i32> <i32 4, i32 5, i32 6, i32 7, i32 2, i32 3, i32 0, i32 1>
@@ -293,13 +287,13 @@ define <4 x i32> @shuffle_16_bitcast_32_shuffle_32_can_not_be_converted_up(<8 x
 }
 
 ; shuffle<8 x i16>( bitcast<8 x i16>( shuffle<16 x i8>(v)))
-; TODO: Narrow, squash shuffles, and widen type?
+; TODO: squash shuffles and widen type?
 
 define <8 x i16> @shuffle_8_bitcast_16_shuffle_16_can__be_converted_up(<16 x i8> %v1) {
 ; CHECK-LABEL: @shuffle_8_bitcast_16_shuffle_16_can__be_converted_up(
-; CHECK-NEXT:    [[SHUFFLE1:%.*]] = shufflevector <16 x i8> [[V1:%.*]], <16 x i8> undef, <16 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>
-; CHECK-NEXT:    [[BC1:%.*]] = bitcast <16 x i8> [[SHUFFLE1]] to <8 x i16>
-; CHECK-NEXT:    [[SHUFFLE2:%.*]] = shufflevector <8 x i16> [[BC1]], <8 x i16> undef, <8 x i32> <i32 4, i32 5, i32 6, i32 7, i32 2, i32 3, i32 0, i32 1>
+; CHECK-NEXT:    [[TMP1:%.*]] = bitcast <16 x i8> [[V1:%.*]] to <8 x i16>
+; CHECK-NEXT:    [[TMP2:%.*]] = shufflevector <8 x i16> [[TMP1]], <8 x i16> undef, <8 x i32> <i32 4, i32 5, i32 6, i32 7, i32 2, i32 3, i32 0, i32 1>
+; CHECK-NEXT:    [[SHUFFLE2:%.*]] = shufflevector <8 x i16> [[TMP2]], <8 x i16> undef, <8 x i32> <i32 4, i32 5, i32 6, i32 7, i32 2, i32 3, i32 0, i32 1>
 ; CHECK-NEXT:    ret <8 x i16> [[SHUFFLE2]]
 ;
   %shuffle1 = shufflevector <16 x i8> %v1, <16 x i8> undef, <16 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>

llvm/test/Transforms/VectorCombine/X86/shuffle.ll

@@ -59,13 +59,13 @@ define i128 @bitcast_shuf_narrow_element_wrong_type(<4 x i32> %v) {
   ret i128 %r
 }
 
-; Negative test - but might want to try this
+; Widen shuffle elements
 
 define <4 x i32> @bitcast_shuf_wide_element(<8 x i16> %v) {
 ; CHECK-LABEL: @bitcast_shuf_wide_element(
-; CHECK-NEXT:    [[SHUF:%.*]] = shufflevector <8 x i16> [[V:%.*]], <8 x i16> undef, <8 x i32> <i32 0, i32 1, i32 0, i32 1, i32 2, i32 3, i32 2, i32 3>
-; CHECK-NEXT:    [[R:%.*]] = bitcast <8 x i16> [[SHUF]] to <4 x i32>
-; CHECK-NEXT:    ret <4 x i32> [[R]]
+; CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i16> [[V:%.*]] to <4 x i32>
+; CHECK-NEXT:    [[TMP2:%.*]] = shufflevector <4 x i32> [[TMP1]], <4 x i32> undef, <4 x i32> <i32 0, i32 0, i32 1, i32 1>
+; CHECK-NEXT:    ret <4 x i32> [[TMP2]]
 ;
   %shuf = shufflevector <8 x i16> %v, <8 x i16> undef, <8 x i32> <i32 0, i32 1, i32 0, i32 1, i32 2, i32 3, i32 2, i32 3>
   %r = bitcast <8 x i16> %shuf to <4 x i32>