test/CodeGen/AArch64/build-vector-reconstructshuffle.ll

; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc < %s | FileCheck %s

; Make sure that AArch64TargetLowering::ReconstructShuffle does not try to construct
; a NEON tbl if the inputs are wider than a NEON-sized vector.

target triple = "aarch64"

; Function Attrs: vscale_range(2,2)
define <8 x i16> @deinterleave_4() #0 {
; CHECK-LABEL: deinterleave_4:
; CHECK:       // %bb.0: // %iter.check
; CHECK-NEXT:    mov x8, xzr
; CHECK-NEXT:    mov w9, #32 // =0x20
; CHECK-NEXT:    ldr z0, [x8]
; CHECK-NEXT:    mov z1.s, z0.s[4]
; CHECK-NEXT:    fmov w8, s1
; CHECK-NEXT:    ldr z1, [x9]
; CHECK-NEXT:    mov v0.h[1], w8
; CHECK-NEXT:    fmov w8, s1
; CHECK-NEXT:    mov z1.s, z1.s[4]
; CHECK-NEXT:    fmov w9, s1
; CHECK-NEXT:    mov v0.h[2], w8
; CHECK-NEXT:    mov w8, #64 // =0x40
; CHECK-NEXT:    ldr z1, [x8]
; CHECK-NEXT:    mov v0.h[3], w9
; CHECK-NEXT:    fmov w8, s1
; CHECK-NEXT:    mov z1.s, z1.s[4]
; CHECK-NEXT:    fmov w9, s1
; CHECK-NEXT:    mov v0.h[4], w8
; CHECK-NEXT:    mov w8, #96 // =0x60
; CHECK-NEXT:    ldr z1, [x8]
; CHECK-NEXT:    mov v0.h[5], w9
; CHECK-NEXT:    fmov w8, s1
; CHECK-NEXT:    mov z1.s, z1.s[4]
; CHECK-NEXT:    mov v0.h[6], w8
; CHECK-NEXT:    fmov w8, s1
; CHECK-NEXT:    mov v0.h[7], w8
; CHECK-NEXT:    // kill: def $q0 killed $q0 killed $z0
; CHECK-NEXT:    ret
iter.check:
  %0 = load <32 x i32>, ptr null, align 8
  %1 = shufflevector <32 x i32> %0, <32 x i32> zeroinitializer, <8 x i32> <i32 0, i32 4, i32 8, i32 12, i32 16, i32 20, i32 24, i32 28>
  %2 = trunc <8 x i32> %1 to <8 x i16>
  ret <8 x i16> %2
}

; Function Attrs: vscale_range(2,2)
define <8 x i16> @extract_3x_v8i32_loads(ptr %pa, ptr %pb, ptr %pc) #0 {
; CHECK-LABEL: extract_3x_v8i32_loads:
; CHECK:       // %bb.0:
; CHECK-NEXT:    ldr z0, [x0]
; CHECK-NEXT:    mov z1.s, z0.s[4]
; CHECK-NEXT:    fmov w9, s0
; CHECK-NEXT:    fmov w8, s1
; CHECK-NEXT:    ldr z1, [x1]
; CHECK-NEXT:    mov v0.h[1], w8
; CHECK-NEXT:    fmov w10, s1
; CHECK-NEXT:    mov z1.s, z1.s[4]
; CHECK-NEXT:    mov v0.h[2], w10
; CHECK-NEXT:    fmov w10, s1
; CHECK-NEXT:    ldr z1, [x2]
; CHECK-NEXT:    mov v0.h[3], w10
; CHECK-NEXT:    fmov w10, s1
; CHECK-NEXT:    mov z1.s, z1.s[4]
; CHECK-NEXT:    mov v0.h[4], w10
; CHECK-NEXT:    fmov w10, s1
; CHECK-NEXT:    mov v0.h[5], w10
; CHECK-NEXT:    mov v0.h[6], w9
; CHECK-NEXT:    mov v0.h[7], w8
; CHECK-NEXT:    // kill: def $q0 killed $q0 killed $z0
; CHECK-NEXT:    ret
  %a = load <8 x i32>, ptr %pa, align 8
  %b = load <8 x i32>, ptr %pb, align 8
  %c = load <8 x i32>, ptr %pc, align 8
  %a0 = extractelement <8 x i32> %a, i64 0
  %a4 = extractelement <8 x i32> %a, i64 4
  %b0 = extractelement <8 x i32> %b, i64 0
  %b4 = extractelement <8 x i32> %b, i64 4
  %c0 = extractelement <8 x i32> %c, i64 0
  %c4 = extractelement <8 x i32> %c, i64 4
  %a0t = trunc i32 %a0 to i16
  %a4t = trunc i32 %a4 to i16
  %b0t = trunc i32 %b0 to i16
  %b4t = trunc i32 %b4 to i16
  %c0t = trunc i32 %c0 to i16
  %c4t = trunc i32 %c4 to i16
  %r0 = insertelement <8 x i16> poison, i16 %a0t, i64 0
  %r1 = insertelement <8 x i16> %r0, i16 %a4t, i64 1
  %r2 = insertelement <8 x i16> %r1, i16 %b0t, i64 2
  %r3 = insertelement <8 x i16> %r2, i16 %b4t, i64 3
  %r4 = insertelement <8 x i16> %r3, i16 %c0t, i64 4
  %r5 = insertelement <8 x i16> %r4, i16 %c4t, i64 5
  %r6 = insertelement <8 x i16> %r5, i16 %a0t, i64 6
  %r7 = insertelement <8 x i16> %r6, i16 %a4t, i64 7
  ret <8 x i16> %r7
}

attributes #0 = { vscale_range(2,2) "target-features"="+neon,+sve" }