test/CodeGen/AArch64/arm64-rev.ll

; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc < %s -mtriple=aarch64-eabi -aarch64-neon-syntax=apple | FileCheck %s
; RUN: llc < %s -global-isel -global-isel-abort=2 -pass-remarks-missed=gisel* -mtriple=aarch64-eabi -aarch64-neon-syntax=apple | FileCheck %s --check-prefixes=FALLBACK,GISEL

; FALLBACK-NOT: remark{{.*}}test_rev_w
define i32 @test_rev_w(i32 %a) nounwind {
; CHECK-LABEL: test_rev_w:
; CHECK:       // %bb.0: // %entry
; CHECK-NEXT:    rev w0, w0
; CHECK-NEXT:    ret
;
; FALLBACK-LABEL: test_rev_w:
; FALLBACK:       // %bb.0: // %entry
; FALLBACK-NEXT:    rev w0, w0
; FALLBACK-NEXT:    ret
entry:
  %0 = tail call i32 @llvm.bswap.i32(i32 %a)
  ret i32 %0
}

; FALLBACK-NOT: remark{{.*}}test_rev_x
define i64 @test_rev_x(i64 %a) nounwind {
; CHECK-LABEL: test_rev_x:
; CHECK:       // %bb.0: // %entry
; CHECK-NEXT:    rev x0, x0
; CHECK-NEXT:    ret
;
; FALLBACK-LABEL: test_rev_x:
; FALLBACK:       // %bb.0: // %entry
; FALLBACK-NEXT:    rev x0, x0
; FALLBACK-NEXT:    ret
entry:
  %0 = tail call i64 @llvm.bswap.i64(i64 %a)
  ret i64 %0
}

; Canonicalize (srl (bswap x), 16) to (rotr (bswap x), 16) if the high 16-bits
; of %a are zero. This optimizes rev + lsr 16 to rev16.
define i32 @test_rev_w_srl16(i16 %a) {
; CHECK-LABEL: test_rev_w_srl16:
; CHECK:       // %bb.0: // %entry
; CHECK-NEXT:    and w8, w0, #0xffff
; CHECK-NEXT:    rev16 w0, w8
; CHECK-NEXT:    ret
;
; FALLBACK-LABEL: test_rev_w_srl16:
; FALLBACK:       // %bb.0: // %entry
; FALLBACK-NEXT:    and w8, w0, #0xffff
; FALLBACK-NEXT:    rev w8, w8
; FALLBACK-NEXT:    lsr w0, w8, #16
; FALLBACK-NEXT:    ret
entry:
  %0 = zext i16 %a to i32
  %1 = tail call i32 @llvm.bswap.i32(i32 %0)
  %2 = lshr i32 %1, 16
  ret i32 %2
}

define i32 @test_rev_w_srl16_load(i16 *%a) {
; CHECK-LABEL: test_rev_w_srl16_load:
; CHECK:       // %bb.0: // %entry
; CHECK-NEXT:    ldrh w8, [x0]
; CHECK-NEXT:    rev16 w0, w8
; CHECK-NEXT:    ret
;
; FALLBACK-LABEL: test_rev_w_srl16_load:
; FALLBACK:       // %bb.0: // %entry
; FALLBACK-NEXT:    ldrh w8, [x0]
; FALLBACK-NEXT:    rev w8, w8
; FALLBACK-NEXT:    lsr w0, w8, #16
; FALLBACK-NEXT:    ret
entry:
  %0 = load i16, i16 *%a
  %1 = zext i16 %0 to i32
  %2 = tail call i32 @llvm.bswap.i32(i32 %1)
  %3 = lshr i32 %2, 16
  ret i32 %3
}

define i32 @test_rev_w_srl16_add(i8 %a, i8 %b) {
; CHECK-LABEL: test_rev_w_srl16_add:
; CHECK:       // %bb.0: // %entry
; CHECK-NEXT:    and w8, w0, #0xff
; CHECK-NEXT:    add w8, w8, w1, uxtb
; CHECK-NEXT:    rev16 w0, w8
; CHECK-NEXT:    ret
;
; FALLBACK-LABEL: test_rev_w_srl16_add:
; FALLBACK:       // %bb.0: // %entry
; FALLBACK-NEXT:    and w8, w1, #0xff
; FALLBACK-NEXT:    add w8, w8, w0, uxtb
; FALLBACK-NEXT:    rev w8, w8
; FALLBACK-NEXT:    lsr w0, w8, #16
; FALLBACK-NEXT:    ret
entry:
  %0 = zext i8 %a to i32
  %1 = zext i8 %b to i32
  %2 = add i32 %0, %1
  %3 = tail call i32 @llvm.bswap.i32(i32 %2)
  %4 = lshr i32 %3, 16
  ret i32 %4
}

; Canonicalize (srl (bswap x), 32) to (rotr (bswap x), 32) if the high 32-bits
; of %a are zero. This optimizes rev + lsr 32 to rev32.
define i64 @test_rev_x_srl32(i32 %a) {
; CHECK-LABEL: test_rev_x_srl32:
; CHECK:       // %bb.0: // %entry
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    rev32 x0, x8
; CHECK-NEXT:    ret
;
; FALLBACK-LABEL: test_rev_x_srl32:
; FALLBACK:       // %bb.0: // %entry
; FALLBACK-NEXT:    // kill: def $w0 killed $w0 def $x0
; FALLBACK-NEXT:    ubfx x8, x0, #0, #32
; FALLBACK-NEXT:    rev x8, x8
; FALLBACK-NEXT:    lsr x0, x8, #32
; FALLBACK-NEXT:    ret
entry:
  %0 = zext i32 %a to i64
  %1 = tail call i64 @llvm.bswap.i64(i64 %0)
  %2 = lshr i64 %1, 32
  ret i64 %2
}

define i64 @test_rev_x_srl32_load(i32 *%a) {
; CHECK-LABEL: test_rev_x_srl32_load:
; CHECK:       // %bb.0: // %entry
; CHECK-NEXT:    ldr w8, [x0]
; CHECK-NEXT:    rev32 x0, x8
; CHECK-NEXT:    ret
;
; FALLBACK-LABEL: test_rev_x_srl32_load:
; FALLBACK:       // %bb.0: // %entry
; FALLBACK-NEXT:    ldr w8, [x0]
; FALLBACK-NEXT:    rev x8, x8
; FALLBACK-NEXT:    lsr x0, x8, #32
; FALLBACK-NEXT:    ret
entry:
  %0 = load i32, i32 *%a
  %1 = zext i32 %0 to i64
  %2 = tail call i64 @llvm.bswap.i64(i64 %1)
  %3 = lshr i64 %2, 32
  ret i64 %3
}

define i64 @test_rev_x_srl32_shift(i64 %a) {
; CHECK-LABEL: test_rev_x_srl32_shift:
; CHECK:       // %bb.0: // %entry
; CHECK-NEXT:    ubfx x8, x0, #2, #29
; CHECK-NEXT:    rev32 x0, x8
; CHECK-NEXT:    ret
;
; FALLBACK-LABEL: test_rev_x_srl32_shift:
; FALLBACK:       // %bb.0: // %entry
; FALLBACK-NEXT:    lsl x8, x0, #33
; FALLBACK-NEXT:    lsr x8, x8, #35
; FALLBACK-NEXT:    rev x8, x8
; FALLBACK-NEXT:    lsr x0, x8, #32
; FALLBACK-NEXT:    ret
entry:
  %0 = shl i64 %a, 33
  %1 = lshr i64 %0, 35
  %2 = tail call i64 @llvm.bswap.i64(i64 %1)
  %3 = lshr i64 %2, 32
  ret i64 %3
}

declare i32 @llvm.bswap.i32(i32) nounwind readnone
declare i64 @llvm.bswap.i64(i64) nounwind readnone

define i32 @test_rev16_w(i32 %X) nounwind {
; CHECK-LABEL: test_rev16_w:
; CHECK:       // %bb.0: // %entry
; CHECK-NEXT:    rev16 w0, w0
; CHECK-NEXT:    ret
;
; FALLBACK-LABEL: test_rev16_w:
; FALLBACK:       // %bb.0: // %entry
; FALLBACK-NEXT:    lsr w8, w0, #8
; FALLBACK-NEXT:    lsl w9, w0, #8
; FALLBACK-NEXT:    and w10, w8, #0xff0000
; FALLBACK-NEXT:    and w11, w9, #0xff000000
; FALLBACK-NEXT:    and w9, w9, #0xff00
; FALLBACK-NEXT:    orr w10, w11, w10
; FALLBACK-NEXT:    and w8, w8, #0xff
; FALLBACK-NEXT:    orr w9, w10, w9
; FALLBACK-NEXT:    orr w0, w9, w8
; FALLBACK-NEXT:    ret
entry:
  %tmp1 = lshr i32 %X, 8
  %X15 = bitcast i32 %X to i32
  %tmp4 = shl i32 %X15, 8
  %tmp2 = and i32 %tmp1, 16711680
  %tmp5 = and i32 %tmp4, -16777216
  %tmp9 = and i32 %tmp1, 255
  %tmp13 = and i32 %tmp4, 65280
  %tmp6 = or i32 %tmp5, %tmp2
  %tmp10 = or i32 %tmp6, %tmp13
  %tmp14 = or i32 %tmp10, %tmp9
  ret i32 %tmp14
}

; 64-bit REV16 is *not* a swap then a 16-bit rotation:
;   01234567 ->(bswap) 76543210 ->(rotr) 10765432
;   01234567 ->(rev16) 10325476
define i64 @test_rev16_x(i64 %a) nounwind {
; CHECK-LABEL: test_rev16_x:
; CHECK:       // %bb.0: // %entry
; CHECK-NEXT:    rev x8, x0
; CHECK-NEXT:    ror x0, x8, #16
; CHECK-NEXT:    ret
;
; FALLBACK-LABEL: test_rev16_x:
; FALLBACK:       // %bb.0: // %entry
; FALLBACK-NEXT:    rev x8, x0
; FALLBACK-NEXT:    lsl x9, x8, #48
; FALLBACK-NEXT:    orr x0, x9, x8, lsr #16
; FALLBACK-NEXT:    ret
entry:
  %0 = tail call i64 @llvm.bswap.i64(i64 %a)
  %1 = lshr i64 %0, 16
  %2 = shl i64 %0, 48
  %3 = or i64 %1, %2
  ret i64 %3
}

define i64 @test_rev32_x(i64 %a) nounwind {
; CHECK-LABEL: test_rev32_x:
; CHECK:       // %bb.0: // %entry
; CHECK-NEXT:    rev32 x0, x0
; CHECK-NEXT:    ret
;
; FALLBACK-LABEL: test_rev32_x:
; FALLBACK:       // %bb.0: // %entry
; FALLBACK-NEXT:    rev x8, x0
; FALLBACK-NEXT:    lsl x9, x8, #32
; FALLBACK-NEXT:    orr x0, x9, x8, lsr #32
; FALLBACK-NEXT:    ret
entry:
  %0 = tail call i64 @llvm.bswap.i64(i64 %a)
  %1 = lshr i64 %0, 32
  %2 = shl i64 %0, 32
  %3 = or i64 %1, %2
  ret i64 %3
}

define <8 x i8> @test_vrev64D8(<8 x i8>* %A) nounwind {
; CHECK-LABEL: test_vrev64D8:
; CHECK:       // %bb.0:
; CHECK-NEXT:    ldr d0, [x0]
; CHECK-NEXT:    rev64.8b v0, v0
; CHECK-NEXT:    ret
;
; FALLBACK-LABEL: test_vrev64D8:
; FALLBACK:       // %bb.0:
; FALLBACK-NEXT:    ldr d0, [x0]
; FALLBACK-NEXT:    rev64.8b v0, v0
; FALLBACK-NEXT:    ret
	%tmp1 = load <8 x i8>, <8 x i8>* %A
	%tmp2 = shufflevector <8 x i8> %tmp1, <8 x i8> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
	ret <8 x i8> %tmp2
}

define <4 x i16> @test_vrev64D16(<4 x i16>* %A) nounwind {
; CHECK-LABEL: test_vrev64D16:
; CHECK:       // %bb.0:
; CHECK-NEXT:    ldr d0, [x0]
; CHECK-NEXT:    rev64.4h v0, v0
; CHECK-NEXT:    ret
;
; FALLBACK-LABEL: test_vrev64D16:
; FALLBACK:       // %bb.0:
; FALLBACK-NEXT:    ldr d0, [x0]
; FALLBACK-NEXT:    rev64.4h v0, v0
; FALLBACK-NEXT:    ret
	%tmp1 = load <4 x i16>, <4 x i16>* %A
	%tmp2 = shufflevector <4 x i16> %tmp1, <4 x i16> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
	ret <4 x i16> %tmp2
}

define <2 x i32> @test_vrev64D32(<2 x i32>* %A) nounwind {
; CHECK-LABEL: test_vrev64D32:
; CHECK:       // %bb.0:
; CHECK-NEXT:    ldr d0, [x0]
; CHECK-NEXT:    rev64.2s v0, v0
; CHECK-NEXT:    ret
;
; FALLBACK-LABEL: test_vrev64D32:
; FALLBACK:       // %bb.0:
; FALLBACK-NEXT:    ldr d0, [x0]
; FALLBACK-NEXT:    adrp x8, .LCPI13_0
; FALLBACK-NEXT:    ldr d1, [x8, :lo12:.LCPI13_0]
; FALLBACK-NEXT:    mov.s v2[1], w8
; FALLBACK-NEXT:    mov.d v0[1], v2[0]
; FALLBACK-NEXT:    tbl.16b v0, { v0 }, v1
; FALLBACK-NEXT:    // kill: def $d0 killed $d0 killed $q0
; FALLBACK-NEXT:    ret
	%tmp1 = load <2 x i32>, <2 x i32>* %A
	%tmp2 = shufflevector <2 x i32> %tmp1, <2 x i32> undef, <2 x i32> <i32 1, i32 0>
	ret <2 x i32> %tmp2
}

define <2 x float> @test_vrev64Df(<2 x float>* %A) nounwind {
; CHECK-LABEL: test_vrev64Df:
; CHECK:       // %bb.0:
; CHECK-NEXT:    ldr d0, [x0]
; CHECK-NEXT:    rev64.2s v0, v0
; CHECK-NEXT:    ret
;
; FALLBACK-LABEL: test_vrev64Df:
; FALLBACK:       // %bb.0:
; FALLBACK-NEXT:    ldr d0, [x0]
; FALLBACK-NEXT:    adrp x8, .LCPI14_0
; FALLBACK-NEXT:    ldr d1, [x8, :lo12:.LCPI14_0]
; FALLBACK-NEXT:    mov.s v2[1], w8
; FALLBACK-NEXT:    mov.d v0[1], v2[0]
; FALLBACK-NEXT:    tbl.16b v0, { v0 }, v1
; FALLBACK-NEXT:    // kill: def $d0 killed $d0 killed $q0
; FALLBACK-NEXT:    ret
	%tmp1 = load <2 x float>, <2 x float>* %A
	%tmp2 = shufflevector <2 x float> %tmp1, <2 x float> undef, <2 x i32> <i32 1, i32 0>
	ret <2 x float> %tmp2
}

define <16 x i8> @test_vrev64Q8(<16 x i8>* %A) nounwind {
; CHECK-LABEL: test_vrev64Q8:
; CHECK:       // %bb.0:
; CHECK-NEXT:    ldr q0, [x0]
; CHECK-NEXT:    rev64.16b v0, v0
; CHECK-NEXT:    ret
;
; FALLBACK-LABEL: test_vrev64Q8:
; FALLBACK:       // %bb.0:
; FALLBACK-NEXT:    ldr q0, [x0]
; FALLBACK-NEXT:    rev64.16b v0, v0
; FALLBACK-NEXT:    ret
	%tmp1 = load <16 x i8>, <16 x i8>* %A
	%tmp2 = shufflevector <16 x i8> %tmp1, <16 x i8> undef, <16 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8>
	ret <16 x i8> %tmp2
}

define <8 x i16> @test_vrev64Q16(<8 x i16>* %A) nounwind {
; CHECK-LABEL: test_vrev64Q16:
; CHECK:       // %bb.0:
; CHECK-NEXT:    ldr q0, [x0]
; CHECK-NEXT:    rev64.8h v0, v0
; CHECK-NEXT:    ret
;
; FALLBACK-LABEL: test_vrev64Q16:
; FALLBACK:       // %bb.0:
; FALLBACK-NEXT:    ldr q0, [x0]
; FALLBACK-NEXT:    rev64.8h v0, v0
; FALLBACK-NEXT:    ret
	%tmp1 = load <8 x i16>, <8 x i16>* %A
	%tmp2 = shufflevector <8 x i16> %tmp1, <8 x i16> undef, <8 x i32> <i32 3, i32 2, i32 1, i32 0, i32 7, i32 6, i32 5, i32 4>
	ret <8 x i16> %tmp2
}

define <4 x i32> @test_vrev64Q32(<4 x i32>* %A) nounwind {
; CHECK-LABEL: test_vrev64Q32:
; CHECK:       // %bb.0:
; CHECK-NEXT:    ldr q0, [x0]
; CHECK-NEXT:    rev64.4s v0, v0
; CHECK-NEXT:    ret
;
; FALLBACK-LABEL: test_vrev64Q32:
; FALLBACK:       // %bb.0:
; FALLBACK-NEXT:    adrp x8, .LCPI17_0
; FALLBACK-NEXT:    ldr q0, [x0]
; FALLBACK-NEXT:    ldr q2, [x8, :lo12:.LCPI17_0]
; FALLBACK-NEXT:    tbl.16b v0, { v0, v1 }, v2
; FALLBACK-NEXT:    ret
	%tmp1 = load <4 x i32>, <4 x i32>* %A
	%tmp2 = shufflevector <4 x i32> %tmp1, <4 x i32> undef, <4 x i32> <i32 1, i32 0, i32 3, i32 2>
	ret <4 x i32> %tmp2
}

define <4 x float> @test_vrev64Qf(<4 x float>* %A) nounwind {
; CHECK-LABEL: test_vrev64Qf:
; CHECK:       // %bb.0:
; CHECK-NEXT:    ldr q0, [x0]
; CHECK-NEXT:    rev64.4s v0, v0
; CHECK-NEXT:    ret
;
; FALLBACK-LABEL: test_vrev64Qf:
; FALLBACK:       // %bb.0:
; FALLBACK-NEXT:    adrp x8, .LCPI18_0
; FALLBACK-NEXT:    ldr q0, [x0]
; FALLBACK-NEXT:    ldr q2, [x8, :lo12:.LCPI18_0]
; FALLBACK-NEXT:    tbl.16b v0, { v0, v1 }, v2
; FALLBACK-NEXT:    ret
	%tmp1 = load <4 x float>, <4 x float>* %A
	%tmp2 = shufflevector <4 x float> %tmp1, <4 x float> undef, <4 x i32> <i32 1, i32 0, i32 3, i32 2>
	ret <4 x float> %tmp2
}

define <8 x i8> @test_vrev32D8(<8 x i8>* %A) nounwind {
; CHECK-LABEL: test_vrev32D8:
; CHECK:       // %bb.0:
; CHECK-NEXT:    ldr d0, [x0]
; CHECK-NEXT:    rev32.8b v0, v0
; CHECK-NEXT:    ret
;
; FALLBACK-LABEL: test_vrev32D8:
; FALLBACK:       // %bb.0:
; FALLBACK-NEXT:    ldr d0, [x0]
; FALLBACK-NEXT:    rev32.8b v0, v0
; FALLBACK-NEXT:    ret
	%tmp1 = load <8 x i8>, <8 x i8>* %A
	%tmp2 = shufflevector <8 x i8> %tmp1, <8 x i8> undef, <8 x i32> <i32 3, i32 2, i32 1, i32 0, i32 7, i32 6, i32 5, i32 4>
	ret <8 x i8> %tmp2
}

define <4 x i16> @test_vrev32D16(<4 x i16>* %A) nounwind {
; CHECK-LABEL: test_vrev32D16:
; CHECK:       // %bb.0:
; CHECK-NEXT:    ldr d0, [x0]
; CHECK-NEXT:    rev32.4h v0, v0
; CHECK-NEXT:    ret
;
; FALLBACK-LABEL: test_vrev32D16:
; FALLBACK:       // %bb.0:
; FALLBACK-NEXT:    ldr d0, [x0]
; FALLBACK-NEXT:    rev32.4h v0, v0
; FALLBACK-NEXT:    ret
	%tmp1 = load <4 x i16>, <4 x i16>* %A
	%tmp2 = shufflevector <4 x i16> %tmp1, <4 x i16> undef, <4 x i32> <i32 1, i32 0, i32 3, i32 2>
	ret <4 x i16> %tmp2
}

define <16 x i8> @test_vrev32Q8(<16 x i8>* %A) nounwind {
; CHECK-LABEL: test_vrev32Q8:
; CHECK:       // %bb.0:
; CHECK-NEXT:    ldr q0, [x0]
; CHECK-NEXT:    rev32.16b v0, v0
; CHECK-NEXT:    ret
;
; FALLBACK-LABEL: test_vrev32Q8:
; FALLBACK:       // %bb.0:
; FALLBACK-NEXT:    ldr q0, [x0]
; FALLBACK-NEXT:    rev32.16b v0, v0
; FALLBACK-NEXT:    ret
	%tmp1 = load <16 x i8>, <16 x i8>* %A
	%tmp2 = shufflevector <16 x i8> %tmp1, <16 x i8> undef, <16 x i32> <i32 3, i32 2, i32 1, i32 0, i32 7, i32 6, i32 5, i32 4, i32 11, i32 10, i32 9, i32 8, i32 15, i32 14, i32 13, i32 12>
	ret <16 x i8> %tmp2
}

define <8 x i16> @test_vrev32Q16(<8 x i16>* %A) nounwind {
; CHECK-LABEL: test_vrev32Q16:
; CHECK:       // %bb.0:
; CHECK-NEXT:    ldr q0, [x0]
; CHECK-NEXT:    rev32.8h v0, v0
; CHECK-NEXT:    ret
;
; FALLBACK-LABEL: test_vrev32Q16:
; FALLBACK:       // %bb.0:
; FALLBACK-NEXT:    ldr q0, [x0]
; FALLBACK-NEXT:    rev32.8h v0, v0
; FALLBACK-NEXT:    ret
	%tmp1 = load <8 x i16>, <8 x i16>* %A
	%tmp2 = shufflevector <8 x i16> %tmp1, <8 x i16> undef, <8 x i32> <i32 1, i32 0, i32 3, i32 2, i32 5, i32 4, i32 7, i32 6>
	ret <8 x i16> %tmp2
}

define <8 x i8> @test_vrev16D8(<8 x i8>* %A) nounwind {
; CHECK-LABEL: test_vrev16D8:
; CHECK:       // %bb.0:
; CHECK-NEXT:    ldr d0, [x0]
; CHECK-NEXT:    rev16.8b v0, v0
; CHECK-NEXT:    ret
;
; FALLBACK-LABEL: test_vrev16D8:
; FALLBACK:       // %bb.0:
; FALLBACK-NEXT:    ldr d0, [x0]
; FALLBACK-NEXT:    rev16.8b v0, v0
; FALLBACK-NEXT:    ret
	%tmp1 = load <8 x i8>, <8 x i8>* %A
	%tmp2 = shufflevector <8 x i8> %tmp1, <8 x i8> undef, <8 x i32> <i32 1, i32 0, i32 3, i32 2, i32 5, i32 4, i32 7, i32 6>
	ret <8 x i8> %tmp2
}

define <16 x i8> @test_vrev16Q8(<16 x i8>* %A) nounwind {
; CHECK-LABEL: test_vrev16Q8:
; CHECK:       // %bb.0:
; CHECK-NEXT:    ldr q0, [x0]
; CHECK-NEXT:    rev16.16b v0, v0
; CHECK-NEXT:    ret
;
; FALLBACK-LABEL: test_vrev16Q8:
; FALLBACK:       // %bb.0:
; FALLBACK-NEXT:    ldr q0, [x0]
; FALLBACK-NEXT:    rev16.16b v0, v0
; FALLBACK-NEXT:    ret
	%tmp1 = load <16 x i8>, <16 x i8>* %A
	%tmp2 = shufflevector <16 x i8> %tmp1, <16 x i8> undef, <16 x i32> <i32 1, i32 0, i32 3, i32 2, i32 5, i32 4, i32 7, i32 6, i32 9, i32 8, i32 11, i32 10, i32 13, i32 12, i32 15, i32 14>
	ret <16 x i8> %tmp2
}

; Undef shuffle indices should not prevent matching to VREV:

define <8 x i8> @test_vrev64D8_undef(<8 x i8>* %A) nounwind {
; CHECK-LABEL: test_vrev64D8_undef:
; CHECK:       // %bb.0:
; CHECK-NEXT:    ldr d0, [x0]
; CHECK-NEXT:    rev64.8b v0, v0
; CHECK-NEXT:    ret
;
; FALLBACK-LABEL: test_vrev64D8_undef:
; FALLBACK:       // %bb.0:
; FALLBACK-NEXT:    ldr d0, [x0]
; FALLBACK-NEXT:    rev64.8b v0, v0
; FALLBACK-NEXT:    ret
	%tmp1 = load <8 x i8>, <8 x i8>* %A
	%tmp2 = shufflevector <8 x i8> %tmp1, <8 x i8> undef, <8 x i32> <i32 7, i32 undef, i32 undef, i32 4, i32 3, i32 2, i32 1, i32 0>
	ret <8 x i8> %tmp2
}

define <8 x i16> @test_vrev32Q16_undef(<8 x i16>* %A) nounwind {
; CHECK-LABEL: test_vrev32Q16_undef:
; CHECK:       // %bb.0:
; CHECK-NEXT:    ldr q0, [x0]
; CHECK-NEXT:    rev32.8h v0, v0
; CHECK-NEXT:    ret
;
; FALLBACK-LABEL: test_vrev32Q16_undef:
; FALLBACK:       // %bb.0:
; FALLBACK-NEXT:    ldr q0, [x0]
; FALLBACK-NEXT:    rev32.8h v0, v0
; FALLBACK-NEXT:    ret
	%tmp1 = load <8 x i16>, <8 x i16>* %A
	%tmp2 = shufflevector <8 x i16> %tmp1, <8 x i16> undef, <8 x i32> <i32 undef, i32 0, i32 undef, i32 2, i32 5, i32 4, i32 7, i32 undef>
	ret <8 x i16> %tmp2
}

; vrev <4 x i16> should use REV32 and not REV64
define void @test_vrev64(<4 x i16>* nocapture %source, <2 x i16>* nocapture %dst) nounwind ssp {
; CHECK-LABEL: test_vrev64:
; CHECK:       // %bb.0: // %entry
; CHECK-NEXT:    ldr q0, [x0]
; CHECK-NEXT:    add x8, x1, #2 // =2
; CHECK-NEXT:    st1.h { v0 }[5], [x8]
; CHECK-NEXT:    st1.h { v0 }[6], [x1]
; CHECK-NEXT:    ret
;
; FALLBACK-LABEL: test_vrev64:
; FALLBACK:       // %bb.0: // %entry
; FALLBACK-NEXT:    ldr q0, [x0]
; FALLBACK-NEXT:    add x8, x1, #2 // =2
; FALLBACK-NEXT:    st1.h { v0 }[5], [x8]
; FALLBACK-NEXT:    st1.h { v0 }[6], [x1]
; FALLBACK-NEXT:    ret
entry:
  %0 = bitcast <4 x i16>* %source to <8 x i16>*
  %tmp2 = load <8 x i16>, <8 x i16>* %0, align 4
  %tmp3 = extractelement <8 x i16> %tmp2, i32 6
  %tmp5 = insertelement <2 x i16> undef, i16 %tmp3, i32 0
  %tmp9 = extractelement <8 x i16> %tmp2, i32 5
  %tmp11 = insertelement <2 x i16> %tmp5, i16 %tmp9, i32 1
  store <2 x i16> %tmp11, <2 x i16>* %dst, align 4
  ret void
}

; Test vrev of float4
define void @float_vrev64(float* nocapture %source, <4 x float>* nocapture %dest) nounwind noinline ssp {
; CHECK-LABEL: float_vrev64:
; CHECK:       // %bb.0: // %entry
; CHECK-NEXT:    ldr q0, [x0]
; CHECK-NEXT:    movi.2d v1, #0000000000000000
; CHECK-NEXT:    dup.4s v1, v1[0]
; CHECK-NEXT:    ext.16b v0, v0, v1, #12
; CHECK-NEXT:    rev64.4s v0, v0
; CHECK-NEXT:    str q0, [x1, #176]
; CHECK-NEXT:    ret
;
; FALLBACK-LABEL: float_vrev64:
; FALLBACK:       // %bb.0: // %entry
; FALLBACK-NEXT:    fmov s0, wzr
; FALLBACK-NEXT:    mov.s v0[1], v0[0]
; FALLBACK-NEXT:    mov.s v0[2], v0[0]
; FALLBACK-NEXT:    adrp x8, .LCPI28_0
; FALLBACK-NEXT:    mov.s v0[3], v0[0]
; FALLBACK-NEXT:    ldr q1, [x0]
; FALLBACK-NEXT:    ldr q2, [x8, :lo12:.LCPI28_0]
; FALLBACK-NEXT:    tbl.16b v0, { v0, v1 }, v2
; FALLBACK-NEXT:    str q0, [x1, #176]
; FALLBACK-NEXT:    ret
entry:
  %0 = bitcast float* %source to <4 x float>*
  %tmp2 = load <4 x float>, <4 x float>* %0, align 4
  %tmp5 = shufflevector <4 x float> <float 0.000000e+00, float undef, float undef, float undef>, <4 x float> %tmp2, <4 x i32> <i32 0, i32 7, i32 0, i32 0>
  %arrayidx8 = getelementptr inbounds <4 x float>, <4 x float>* %dest, i32 11
  store <4 x float> %tmp5, <4 x float>* %arrayidx8, align 4
  ret void
}


define <4 x i32> @test_vrev32_bswap(<4 x i32> %source) nounwind {
; CHECK-LABEL: test_vrev32_bswap:
; CHECK:       // %bb.0:
; CHECK-NEXT:    rev32.16b v0, v0
; CHECK-NEXT:    ret
;
; FALLBACK-LABEL: test_vrev32_bswap:
; FALLBACK:       // %bb.0:
; FALLBACK-NEXT:    rev32.16b v0, v0
; FALLBACK-NEXT:    ret
  %bswap = call <4 x i32> @llvm.bswap.v4i32(<4 x i32> %source)
  ret <4 x i32> %bswap
}

declare <4 x i32> @llvm.bswap.v4i32(<4 x i32>) nounwind readnone