; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc -mtriple=thumbv8.1m.main-none-none-eabi -mattr=+mve.fp -enable-arm-maskedgatscat %s -o - | FileCheck %s

; VLDRH.16 Qd, [base, offs, uxtw #1]
define arm_aapcs_vfpcc void @scaled_v8i16_i16(i16* %base, <8 x i16>* %offptr, <8 x i16> %input) {
; CHECK-LABEL: scaled_v8i16_i16:
; CHECK:       @ %bb.0: @ %entry
; CHECK-NEXT:    vldrh.u16 q1, [r1]
; CHECK-NEXT:    vstrh.16 q0, [r0, q1, uxtw #1]
; CHECK-NEXT:    bx lr
entry:
  %offs = load <8 x i16>, <8 x i16>* %offptr, align 2
  %offs.zext = zext <8 x i16> %offs to <8 x i32>
  %ptrs = getelementptr inbounds i16, i16* %base, <8 x i32> %offs.zext
  call void @llvm.masked.scatter.v8i16.v8p0i16(<8 x i16> %input, <8 x i16*> %ptrs, i32 2, <8 x i1> <i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true>)
  ret void
}

; VLDRH.16 Qd, [base, offs, uxtw #1]
define arm_aapcs_vfpcc void @scaled_v8f16_i16(i16* %base, <8 x i16>* %offptr, <8 x half> %input) {
; CHECK-LABEL: scaled_v8f16_i16:
; CHECK:       @ %bb.0: @ %entry
; CHECK-NEXT:    vldrh.u16 q1, [r1]
; CHECK-NEXT:    vstrh.16 q0, [r0, q1, uxtw #1]
; CHECK-NEXT:    bx lr
entry:
  %offs = load <8 x i16>, <8 x i16>* %offptr, align 2
  %offs.zext = zext <8 x i16> %offs to <8 x i32>
  %i16_ptrs = getelementptr inbounds i16, i16* %base, <8 x i32> %offs.zext
  %ptrs = bitcast <8 x i16*> %i16_ptrs to <8 x half*>
  call void @llvm.masked.scatter.v8f16.v8p0f16(<8 x half> %input, <8 x half*> %ptrs, i32 2, <8 x i1> <i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true>)
  ret void
}

; VLDRH.16 Qd, [base, offs, uxtw #1]
define arm_aapcs_vfpcc void @scaled_v8f16_half(half* %base, <8 x i16>* %offptr, <8 x half> %input) {
; CHECK-LABEL: scaled_v8f16_half:
; CHECK:       @ %bb.0: @ %entry
; CHECK-NEXT:    vldrh.u16 q1, [r1]
; CHECK-NEXT:    vstrh.16 q0, [r0, q1, uxtw #1]
; CHECK-NEXT:    bx lr
entry:
  %offs = load <8 x i16>, <8 x i16>* %offptr, align 2
  %offs.zext = zext <8 x i16> %offs to <8 x i32>
  %ptrs = getelementptr inbounds half, half* %base, <8 x i32> %offs.zext
  call void @llvm.masked.scatter.v8f16.v8p0f16(<8 x half> %input, <8 x half*> %ptrs, i32 2, <8 x i1> <i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true>)
  ret void
}

; Expand
define arm_aapcs_vfpcc void @scaled_v8i16_sext(i16* %base, <8 x i16>* %offptr, <8 x i16> %input) {
; CHECK-LABEL: scaled_v8i16_sext:
; CHECK:       @ %bb.0: @ %entry
; CHECK-NEXT:    vldrh.s32 q2, [r1]
; CHECK-NEXT:    vldrh.s32 q1, [r1, #8]
; CHECK-NEXT:    vmov.u16 r1, q0[0]
; CHECK-NEXT:    vshl.i32 q2, q2, #1
; CHECK-NEXT:    vshl.i32 q1, q1, #1
; CHECK-NEXT:    vadd.i32 q2, q2, r0
; CHECK-NEXT:    vadd.i32 q1, q1, r0
; CHECK-NEXT:    vmov r0, s8
; CHECK-NEXT:    strh r1, [r0]
; CHECK-NEXT:    vmov r0, s9
; CHECK-NEXT:    vmov.u16 r1, q0[1]
; CHECK-NEXT:    strh r1, [r0]
; CHECK-NEXT:    vmov r0, s10
; CHECK-NEXT:    vmov.u16 r1, q0[2]
; CHECK-NEXT:    strh r1, [r0]
; CHECK-NEXT:    vmov r0, s11
; CHECK-NEXT:    vmov.u16 r1, q0[3]
; CHECK-NEXT:    strh r1, [r0]
; CHECK-NEXT:    vmov r0, s4
; CHECK-NEXT:    vmov.u16 r1, q0[4]
; CHECK-NEXT:    strh r1, [r0]
; CHECK-NEXT:    vmov r0, s5
; CHECK-NEXT:    vmov.u16 r1, q0[5]
; CHECK-NEXT:    strh r1, [r0]
; CHECK-NEXT:    vmov r0, s6
; CHECK-NEXT:    vmov.u16 r1, q0[6]
; CHECK-NEXT:    strh r1, [r0]
; CHECK-NEXT:    vmov r0, s7
; CHECK-NEXT:    vmov.u16 r1, q0[7]
; CHECK-NEXT:    strh r1, [r0]
; CHECK-NEXT:    bx lr
entry:
  %offs = load <8 x i16>, <8 x i16>* %offptr, align 2
  %offs.sext = sext <8 x i16> %offs to <8 x i32>
  %ptrs = getelementptr inbounds i16, i16* %base, <8 x i32> %offs.sext
  call void @llvm.masked.scatter.v8i16.v8p0i16(<8 x i16> %input, <8 x i16*> %ptrs, i32 2, <8 x i1> <i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true>)
  ret void
}

; Expand
define arm_aapcs_vfpcc void @scaled_v8f16_sext(i16* %base, <8 x i16>* %offptr, <8 x half> %input) {
; CHECK-LABEL: scaled_v8f16_sext:
; CHECK:       @ %bb.0: @ %entry
; CHECK-NEXT:    vldrh.s32 q1, [r1]
; CHECK-NEXT:    vmovx.f16 s12, s0
; CHECK-NEXT:    vshl.i32 q2, q1, #1
; CHECK-NEXT:    vldrh.s32 q1, [r1, #8]
; CHECK-NEXT:    vadd.i32 q2, q2, r0
; CHECK-NEXT:    vmov r1, s8
; CHECK-NEXT:    vshl.i32 q1, q1, #1
; CHECK-NEXT:    vstr.16 s0, [r1]
; CHECK-NEXT:    vmov r1, s9
; CHECK-NEXT:    vadd.i32 q1, q1, r0
; CHECK-NEXT:    vstr.16 s12, [r1]
; CHECK-NEXT:    vmov r1, s10
; CHECK-NEXT:    vmovx.f16 s0, s3
; CHECK-NEXT:    vstr.16 s1, [r1]
; CHECK-NEXT:    vmov r1, s11
; CHECK-NEXT:    vmovx.f16 s8, s1
; CHECK-NEXT:    vmov r0, s4
; CHECK-NEXT:    vstr.16 s8, [r1]
; CHECK-NEXT:    vstr.16 s2, [r0]
; CHECK-NEXT:    vmov r0, s5
; CHECK-NEXT:    vmovx.f16 s8, s2
; CHECK-NEXT:    vstr.16 s8, [r0]
; CHECK-NEXT:    vmov r0, s6
; CHECK-NEXT:    vstr.16 s3, [r0]
; CHECK-NEXT:    vmov r0, s7
; CHECK-NEXT:    vstr.16 s0, [r0]
; CHECK-NEXT:    bx lr
entry:
  %offs = load <8 x i16>, <8 x i16>* %offptr, align 2
  %offs.sext = sext <8 x i16> %offs to <8 x i32>
  %i16_ptrs = getelementptr inbounds i16, i16* %base, <8 x i32> %offs.sext
  %ptrs = bitcast <8 x i16*> %i16_ptrs to <8 x half*>
  call void @llvm.masked.scatter.v8f16.v8p0f16(<8 x half> %input, <8 x half*> %ptrs, i32 2, <8 x i1> <i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true>)
  ret void
}

; VLDRH.16 Qd, [base, zext(offs), uxtw #1]
define arm_aapcs_vfpcc void @unsigned_scaled_v8i16_i8(i16* %base, <8 x i8>* %offptr, <8 x i16> %input) {
; CHECK-LABEL: unsigned_scaled_v8i16_i8:
; CHECK:       @ %bb.0: @ %entry
; CHECK-NEXT:    vldrb.u16 q1, [r1]
; CHECK-NEXT:    vstrh.16 q0, [r0, q1, uxtw #1]
; CHECK-NEXT:    bx lr
entry:
  %offs = load <8 x i8>, <8 x i8>* %offptr, align 1
  %offs.zext = zext <8 x i8> %offs to <8 x i32>
  %ptrs = getelementptr inbounds i16, i16* %base, <8 x i32> %offs.zext
  call void @llvm.masked.scatter.v8i16.v8p0i16(<8 x i16> %input, <8 x i16*> %ptrs, i32 2, <8 x i1> <i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true>)
  ret void
}

; VLDRH.16 Qd, [base, zext(offs), uxtw #1]
define arm_aapcs_vfpcc void @unsigned_scaled_v8f16_i8(i16* %base, <8 x i8>* %offptr, <8 x half> %input) {
; CHECK-LABEL: unsigned_scaled_v8f16_i8:
; CHECK:       @ %bb.0: @ %entry
; CHECK-NEXT:    vldrb.u16 q1, [r1]
; CHECK-NEXT:    vstrh.16 q0, [r0, q1, uxtw #1]
; CHECK-NEXT:    bx lr
entry:
  %offs = load <8 x i8>, <8 x i8>* %offptr, align 1
  %offs.zext = zext <8 x i8> %offs to <8 x i32>
  %i16_ptrs = getelementptr inbounds i16, i16* %base, <8 x i32> %offs.zext
  %ptrs = bitcast <8 x i16*> %i16_ptrs to <8 x half*>
  call void @llvm.masked.scatter.v8f16.v8p0f16(<8 x half> %input, <8 x half*> %ptrs, i32 2, <8 x i1> <i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true>)
  ret void
}

define arm_aapcs_vfpcc void @scaled_v8i16_i16_passthru_icmp0(i16* %base, <8 x i16>* %offptr, <8 x i16> %input) {
; CHECK-LABEL: scaled_v8i16_i16_passthru_icmp0:
; CHECK:       @ %bb.0: @ %entry
; CHECK-NEXT:    vldrh.u16 q1, [r1]
; CHECK-NEXT:    vpt.s16 gt, q1, zr
; CHECK-NEXT:    vstrht.16 q0, [r0, q1, uxtw #1]
; CHECK-NEXT:    bx lr
entry:
  %offs = load <8 x i16>, <8 x i16>* %offptr, align 2
  %offs.zext = zext <8 x i16> %offs to <8 x i32>
  %ptrs = getelementptr inbounds i16, i16* %base, <8 x i32> %offs.zext
  %mask = icmp sgt <8 x i16> %offs, zeroinitializer
  call void @llvm.masked.scatter.v8i16.v8p0i16(<8 x i16> %input, <8 x i16*> %ptrs, i32 2, <8 x i1> %mask)
  ret void
}

declare void @llvm.masked.scatter.v8i16.v8p0i16(<8 x i16>, <8 x i16*>, i32, <8 x i1>)
declare void @llvm.masked.scatter.v8f16.v8p0f16(<8 x half>, <8 x half*>, i32, <8 x i1>)