[InstCombine] recognize bswap disguised as shufflevector

bitcast <N x i8> (shuf X, undef, <N, N-1,...0>) to i{N*8} --> bswap (bitcast X to i{N*8})

In PR43146:
https://bugs.llvm.org/show_bug.cgi?id=43146
...we have a more complicated case where SLP is making a mess of bswap. This patch won't
do anything for that currently, but we need to improve bswap recognition in instcombine,
SLP, and/or a standalone pass to avoid that problem.

This is limited using the data-layout so we don't try to do this transform with actual
vector types. The backend does not appear to have folds to convert in either direction,
so we don't want to mess up something that is actually better lowered as a shuffle.

On x86, we're trading something like this:

  vmovd	%edi, %xmm0
  vpshufb	LCPI0_0(%rip), %xmm0, %xmm0 ## xmm0 = xmm0[3,2,1,0,u,u,u,u,u,u,u,u,u,u,u,u]
  vmovd	%xmm0, %eax

For:

  movl	%edi, %eax
  bswapl	%eax

Differential Revision: https://reviews.llvm.org/D66965

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@370659 91177308-0d34-0410-b5e6-96231b3b80d8

diff --git a/lib/Transforms/InstCombine/InstCombineCasts.cpp b/lib/Transforms/InstCombine/InstCombineCasts.cpp
index 4be44dac4ca6760c4cb40797b764e8194cc20b25..1a569d2638f3a5882d4ad0f64990b8f30404fa3f 100644 (file)
--- a/lib/Transforms/InstCombine/InstCombineCasts.cpp
+++ b/lib/Transforms/InstCombine/InstCombineCasts.cpp
@@ -2416,6 +2416,22 @@ Instruction *InstCombiner::visitBitCast(BitCastInst &CI) {
         return new ShuffleVectorInst(LHS, RHS, Shuf->getOperand(2));
       }
     }
+
+    // A bitcasted-to-scalar and byte-reversing shuffle is better recognized as
+    // a byte-swap:
+    // bitcast <N x i8> (shuf X, undef, <N, N-1,...0>) --> bswap (bitcast X)
+    // TODO: We should match the related pattern for bitreverse.
+    if (DestTy->isIntegerTy() &&
+        DL.isLegalInteger(DestTy->getScalarSizeInBits()) &&
+        SrcTy->getScalarSizeInBits() == 8 && NumShufElts % 2 == 0 &&
+        Shuf->hasOneUse() && Shuf->isReverse()) {
+      assert(ShufOp0->getType() == SrcTy && "Unexpected shuffle mask");
+      assert(isa<UndefValue>(ShufOp1) && "Unexpected shuffle op");
+      Function *Bswap =
+          Intrinsic::getDeclaration(CI.getModule(), Intrinsic::bswap, DestTy);
+      Value *ScalarX = Builder.CreateBitCast(ShufOp0, DestTy);
+      return IntrinsicInst::Create(Bswap, { ScalarX });
+    }
   }
 
   // Handle the A->B->A cast, and there is an intervening PHI node.

diff --git a/test/Transforms/InstCombine/bswap.ll b/test/Transforms/InstCombine/bswap.ll
index 9d5d499e6439c6c89550285f3ef44ed2a1762f79..2d8a69b86a59c6df0e6ed085495c37bc5d122bc9 100644 (file)
--- a/test/Transforms/InstCombine/bswap.ll
+++ b/test/Transforms/InstCombine/bswap.ll
@@ -233,8 +233,8 @@ define i16 @test10(i32 %a) {
 
 define i32 @shuf_4bytes(<4 x i8> %x) {
 ; CHECK-LABEL: @shuf_4bytes(
-; CHECK-NEXT:    [[BSWAP:%.*]] = shufflevector <4 x i8> [[X:%.*]], <4 x i8> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
-; CHECK-NEXT:    [[CAST:%.*]] = bitcast <4 x i8> [[BSWAP]] to i32
+; CHECK-NEXT:    [[TMP1:%.*]] = bitcast <4 x i8> [[X:%.*]] to i32
+; CHECK-NEXT:    [[CAST:%.*]] = call i32 @llvm.bswap.i32(i32 [[TMP1]])
 ; CHECK-NEXT:    ret i32 [[CAST]]
 ;
   %bswap = shufflevector <4 x i8> %x, <4 x i8> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
@@ -244,9 +244,9 @@ define i32 @shuf_4bytes(<4 x i8> %x) {
 
 define i32 @shuf_load_4bytes(<4 x i8>* %p) {
 ; CHECK-LABEL: @shuf_load_4bytes(
-; CHECK-NEXT:    [[X:%.*]] = load <4 x i8>, <4 x i8>* [[P:%.*]], align 4
-; CHECK-NEXT:    [[BSWAP:%.*]] = shufflevector <4 x i8> [[X]], <4 x i8> undef, <4 x i32> <i32 3, i32 2, i32 undef, i32 0>
-; CHECK-NEXT:    [[CAST:%.*]] = bitcast <4 x i8> [[BSWAP]] to i32
+; CHECK-NEXT:    [[TMP1:%.*]] = bitcast <4 x i8>* [[P:%.*]] to i32*
+; CHECK-NEXT:    [[X1:%.*]] = load i32, i32* [[TMP1]], align 4
+; CHECK-NEXT:    [[CAST:%.*]] = call i32 @llvm.bswap.i32(i32 [[X1]])
 ; CHECK-NEXT:    ret i32 [[CAST]]
 ;
   %x = load <4 x i8>, <4 x i8>* %p
@@ -257,9 +257,7 @@ define i32 @shuf_load_4bytes(<4 x i8>* %p) {
 
 define i32 @shuf_bitcast_twice_4bytes(i32 %x) {
 ; CHECK-LABEL: @shuf_bitcast_twice_4bytes(
-; CHECK-NEXT:    [[CAST1:%.*]] = bitcast i32 [[X:%.*]] to <4 x i8>
-; CHECK-NEXT:    [[BSWAP:%.*]] = shufflevector <4 x i8> [[CAST1]], <4 x i8> undef, <4 x i32> <i32 undef, i32 2, i32 1, i32 0>
-; CHECK-NEXT:    [[CAST2:%.*]] = bitcast <4 x i8> [[BSWAP]] to i32
+; CHECK-NEXT:    [[CAST2:%.*]] = call i32 @llvm.bswap.i32(i32 [[X:%.*]])
 ; CHECK-NEXT:    ret i32 [[CAST2]]
 ;
   %cast1 = bitcast i32 %x to <4 x i8>
@@ -268,6 +266,7 @@ define i32 @shuf_bitcast_twice_4bytes(i32 %x) {
   ret i32 %cast2
 }
 
+; Negative test - extra use
 declare void @use(<4 x i8>)
 
 define i32 @shuf_4bytes_extra_use(<4 x i8> %x) {
@@ -283,6 +282,8 @@ define i32 @shuf_4bytes_extra_use(<4 x i8> %x) {
   ret i32 %cast
 }
 
+; Negative test - scalar type is not in the data layout
+
 define i128 @shuf_16bytes(<16 x i8> %x) {
 ; CHECK-LABEL: @shuf_16bytes(
 ; CHECK-NEXT:    [[BSWAP:%.*]] = shufflevector <16 x i8> [[X:%.*]], <16 x i8> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
@@ -294,6 +295,8 @@ define i128 @shuf_16bytes(<16 x i8> %x) {
   ret i128 %cast
 }
 
+; Negative test - don't touch widening shuffles (for now)
+
 define i32 @shuf_2bytes_widening(<2 x i8> %x) {
 ; CHECK-LABEL: @shuf_2bytes_widening(
 ; CHECK-NEXT:    [[BSWAP:%.*]] = shufflevector <2 x i8> [[X:%.*]], <2 x i8> undef, <4 x i32> <i32 1, i32 0, i32 undef, i32 undef>