; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
; RUN: opt < %s -passes=instcombine -S | FileCheck %s

target datalayout = "e-p:32:32:32-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:32:64-f32:32:32-f64:32:64-v64:64:64-v128:128:128-a0:0:64-f80:128:128-n8:16:32"

; A copy of bswap.ll, with undef at insertelement/shufflevector replaced with
; poison

define i32 @shuf_4bytes(<4 x i8> %x) {
; CHECK-LABEL: @shuf_4bytes(
; 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> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
  %cast = bitcast <4 x i8> %bswap to i32
  ret i32 %cast
}

define i32 @shuf_load_4bytes(<4 x i8>* %p) {
; CHECK-LABEL: @shuf_load_4bytes(
; 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
  %bswap = shufflevector <4 x i8> %x, <4 x i8> poison, <4 x i32> <i32 3, i32 2, i32 undef, i32 0>
  %cast = bitcast <4 x i8> %bswap to i32
  ret i32 %cast
}

define i32 @shuf_bitcast_twice_4bytes(i32 %x) {
; CHECK-LABEL: @shuf_bitcast_twice_4bytes(
; CHECK-NEXT:    [[CAST2:%.*]] = call i32 @llvm.bswap.i32(i32 [[X:%.*]])
; CHECK-NEXT:    ret i32 [[CAST2]]
;
  %cast1 = bitcast i32 %x to <4 x i8>
  %bswap = shufflevector <4 x i8> %cast1, <4 x i8> poison, <4 x i32> <i32 undef, i32 2, i32 1, i32 0>
  %cast2 = bitcast <4 x i8> %bswap to i32
  ret i32 %cast2
}

; Negative test - extra use
declare void @use(<4 x i8>)

define i32 @shuf_4bytes_extra_use(<4 x i8> %x) {
; CHECK-LABEL: @shuf_4bytes_extra_use(
; CHECK-NEXT:    [[BSWAP:%.*]] = shufflevector <4 x i8> [[X:%.*]], <4 x i8> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
; CHECK-NEXT:    call void @use(<4 x i8> [[BSWAP]])
; CHECK-NEXT:    [[CAST:%.*]] = bitcast <4 x i8> [[BSWAP]] to i32
; CHECK-NEXT:    ret i32 [[CAST]]
;
  %bswap = shufflevector <4 x i8> %x, <4 x i8> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
  call void @use(<4 x i8> %bswap)
  %cast = bitcast <4 x i8> %bswap to i32
  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> poison, <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>
; CHECK-NEXT:    [[CAST:%.*]] = bitcast <16 x i8> [[BSWAP]] to i128
; CHECK-NEXT:    ret i128 [[CAST]]
;
  %bswap = shufflevector <16 x i8> %x, <16 x i8> poison, <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>
  %cast = bitcast <16 x i8> %bswap to i128
  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> poison, <4 x i32> <i32 1, i32 0, i32 undef, i32 undef>
; CHECK-NEXT:    [[CAST:%.*]] = bitcast <4 x i8> [[BSWAP]] to i32
; CHECK-NEXT:    ret i32 [[CAST]]
;
  %bswap = shufflevector <2 x i8> %x, <2 x i8> poison, <4 x i32> <i32 1, i32 0, i32 undef, i32 undef>
  %cast = bitcast <4 x i8> %bswap to i32
  ret i32 %cast
}

declare i32 @llvm.fshl.i32(i32, i32, i32)
declare i32 @llvm.fshr.i32(i32, i32, i32)