// Copyright 2020 Google LLC
//
// This source code is licensed under the BSD-style license found in the
// LICENSE file in the root directory of this source tree.

$assert BATCH_TILE % 8 == 0
$assert BATCH_TILE >= 8
$ABC = "0123456789ABCDEFGHIJKLMNOPQRSTUVWXYZ"
#include <assert.h>

#include <arm_neon.h>

#include "xnnpack/common.h"
#include "xnnpack/vunary.h"


void xnn_f16_vclamp_ukernel__neonfp16arith_u${BATCH_TILE}(
    size_t batch,
    const xnn_float16* restrict input,
    xnn_float16* restrict output,
    const union xnn_f16_minmax_params params[restrict XNN_MIN_ELEMENTS(1)])
{
  assert(batch != 0);
  assert(batch % sizeof(uint16_t) == 0);
  assert(input != NULL);
  assert(output != NULL);

  const uint16_t* i = (const uint16_t*) input;
  uint16_t* o = (uint16_t*) output;

  #if XNN_ARCH_ARM64
    const uint16x8x2_t vminmax = vld2q_dup_u16((const uint16_t*) &params->scalar.min);
    const float16x8_t vmin = vreinterpretq_f16_u16(vminmax.val[0]);
    const float16x8_t vmax = vreinterpretq_f16_u16(vminmax.val[1]);
  #else
    // vld2_dup is to work around aarch32 clang bug with vld1q_dup
    const uint16x4x2_t vminmax = vld2_dup_u16((const uint16_t*) &params->scalar.min);
    const float16x8_t vmin = vreinterpretq_f16_u16(vcombine_u16(vminmax.val[0], vminmax.val[0]));
    const float16x8_t vmax = vreinterpretq_f16_u16(vcombine_u16(vminmax.val[1], vminmax.val[1]));
  #endif

  $if BATCH_TILE > 8:
    for (; batch >= ${BATCH_TILE} * sizeof(uint16_t); batch -= ${BATCH_TILE} * sizeof(uint16_t)) {
      $for N in range(0, BATCH_TILE, 8):
        float16x8_t vacc${ABC[N:N+8]} = vreinterpretq_f16_u16(vld1q_u16(i)); i += 8;

      $for N in range(0, BATCH_TILE, 8):
        vacc${ABC[N:N+8]} = vmaxq_f16(vacc${ABC[N:N+8]}, vmin);

      $for N in range(0, BATCH_TILE, 8):
        vacc${ABC[N:N+8]} = vminq_f16(vacc${ABC[N:N+8]}, vmax);

      $for N in range(0, BATCH_TILE, 8):
        vst1q_u16(o, vreinterpretq_u16_f16(vacc${ABC[N:N+8]})); o += 8;
    }
  for (; batch >= 8 * sizeof(uint16_t); batch -= 8 * sizeof(uint16_t)) {
    float16x8_t vacc = vreinterpretq_f16_u16(vld1q_u16(i)); i += 8;
    vacc = vmaxq_f16(vacc, vmin);
    vacc = vminq_f16(vacc, vmax);
    vst1q_u16(o, vreinterpretq_u16_f16(vacc)); o += 8;
  }
  if XNN_UNLIKELY(batch != 0) {
    if (batch & (4 * sizeof(uint16_t))) {
      float16x4_t vacc = vreinterpret_f16_u16(vld1_u16(i)); i += 4;
      vacc = vmax_f16(vacc, vget_low_f16(vmin));
      vacc = vmin_f16(vacc, vget_low_f16(vmax));
      vst1_u16(o, vreinterpret_u16_f16(vacc)); o += 4;
    }
    if (batch & (2 * sizeof(uint16_t))) {
      float16x4_t vacc = vreinterpret_f16_u32(vld1_dup_u32((const void*) i)); i += 2;
      vacc = vmax_f16(vacc, vget_low_f16(vmin));
      vacc = vmin_f16(vacc, vget_low_f16(vmax));
      vst1_lane_u32((void*) o, vreinterpret_u32_f16(vacc), 0); o += 2;
    }
    if (batch & (1 * sizeof(uint16_t))) {
      float16x4_t vacc = vreinterpret_f16_u16(vld1_dup_u16(i));
      vacc = vmax_f16(vacc, vget_low_f16(vmin));
      vacc = vmin_f16(vacc, vget_low_f16(vmax));
      vst1_lane_u16(o, vreinterpret_u16_f16(vacc), 0);
    }
  }
}