// Auto-generated file. Do not edit!
//   Template: src/qs8-vadd/avx2-mul32-ld64.c.in
//   Generator: tools/xngen
//
// 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.

#include <assert.h>

#include <immintrin.h>

#include "xnnpack/common.h"
#include "xnnpack/intrinsics-polyfill.h"
#include "xnnpack/vbinary.h"


void xnn_qs8_vadd_minmax_ukernel__avx2_mul32_ld64_u24(
    size_t batch,
    const int8_t* input_a,
    const int8_t* input_b,
    int8_t* output,
    const struct xnn_qs8_add_minmax_params params[restrict XNN_MIN_ELEMENTS(1)]) XNN_OOB_READS
{
  assert(batch != 0);
  assert(batch % sizeof(int8_t) == 0);
  assert(input_a != NULL);
  assert(input_b != NULL);
  assert(output != NULL);

  const __m256i vbias = _mm256_set1_epi32(params->scalar.bias);
  const __m256i va_multiplier = _mm256_set1_epi32(params->scalar.a_multiplier);
  const __m256i vb_multiplier = _mm256_set1_epi32(params->scalar.b_multiplier);
  const __m128i vshift = _mm_set1_epi64x(params->scalar.shift);
  const __m256i voutput_zero_point = _mm256_set1_epi16(params->scalar.output_zero_point);
  const __m128i voutput_min = _mm_set1_epi8(params->scalar.output_min);
  const __m128i voutput_max = _mm_set1_epi8(params->scalar.output_max);

  XNN_FORCE_REALIZATION(vbias);
  XNN_FORCE_REALIZATION(va_multiplier);
  XNN_FORCE_REALIZATION(vb_multiplier);
  XNN_FORCE_REALIZATION(vshift);
  XNN_FORCE_REALIZATION(voutput_zero_point);
  XNN_FORCE_REALIZATION(voutput_min);
  XNN_FORCE_REALIZATION(voutput_max);

  for (; batch >= 24 * sizeof(int8_t); batch -= 24 * sizeof(int8_t)) {
    const __m256i va01234567 = _mm256_cvtepi8_epi32(_mm_loadl_epi64((const __m128i*) input_a));
    const __m256i vb01234567 = _mm256_cvtepi8_epi32(_mm_loadl_epi64((const __m128i*) input_b));
    const __m256i va89ABCDEF = _mm256_cvtepi8_epi32(_mm_loadl_epi64((const __m128i*) (input_a + 8)));
    const __m256i vb89ABCDEF = _mm256_cvtepi8_epi32(_mm_loadl_epi64((const __m128i*) (input_b + 8)));
    const __m256i vaGHIJKLMN = _mm256_cvtepi8_epi32(_mm_loadl_epi64((const __m128i*) (input_a + 16)));
    const __m256i vbGHIJKLMN = _mm256_cvtepi8_epi32(_mm_loadl_epi64((const __m128i*) (input_b + 16)));
    input_a += 24;
    input_b += 24;

    __m256i vacc01234567 = _mm256_add_epi32(vbias, _mm256_mullo_epi32(va01234567, va_multiplier));
    __m256i vacc89ABCDEF = _mm256_add_epi32(vbias, _mm256_mullo_epi32(va89ABCDEF, va_multiplier));
    __m256i vaccGHIJKLMN = _mm256_add_epi32(vbias, _mm256_mullo_epi32(vaGHIJKLMN, va_multiplier));

    vacc01234567 = _mm256_add_epi32(vacc01234567, _mm256_mullo_epi32(vb01234567, vb_multiplier));
    vacc89ABCDEF = _mm256_add_epi32(vacc89ABCDEF, _mm256_mullo_epi32(vb89ABCDEF, vb_multiplier));
    vaccGHIJKLMN = _mm256_add_epi32(vaccGHIJKLMN, _mm256_mullo_epi32(vbGHIJKLMN, vb_multiplier));

    vacc01234567 = _mm256_sra_epi32(vacc01234567, vshift);
    vacc89ABCDEF = _mm256_sra_epi32(vacc89ABCDEF, vshift);
    vaccGHIJKLMN = _mm256_sra_epi32(vaccGHIJKLMN, vshift);

    __m256i vout012389AB4567CDEF = _mm256_adds_epi16(_mm256_packs_epi32(vacc01234567, vacc89ABCDEF), voutput_zero_point);
    __m128i voutGHIJKLMN = _mm_adds_epi16(_mm_packs_epi32(_mm256_castsi256_si128(vaccGHIJKLMN), _mm256_extracti128_si256(vaccGHIJKLMN, 1)), _mm256_castsi256_si128(voutput_zero_point));

    __m128i vout0123456789ABCDEF = _mm_shuffle_epi32(_mm_packs_epi16(_mm256_castsi256_si128(vout012389AB4567CDEF), _mm256_extracti128_si256(vout012389AB4567CDEF, 1)), _MM_SHUFFLE(3, 1, 2, 0));
    __m128i voutGHIJKLMNGHIJKLMN = _mm_packs_epi16(voutGHIJKLMN, voutGHIJKLMN);

    vout0123456789ABCDEF = _mm_max_epi8(vout0123456789ABCDEF, voutput_min);
    voutGHIJKLMNGHIJKLMN = _mm_max_epi8(voutGHIJKLMNGHIJKLMN, voutput_min);

    vout0123456789ABCDEF = _mm_min_epi8(vout0123456789ABCDEF, voutput_max);
    voutGHIJKLMNGHIJKLMN = _mm_min_epi8(voutGHIJKLMNGHIJKLMN, voutput_max);

    _mm_storeu_si128((__m128i*) output, vout0123456789ABCDEF);
    _mm_storel_epi64((__m128i*) (output + 16), voutGHIJKLMNGHIJKLMN);
    output += 24;
  }
  if XNN_UNLIKELY(batch != 0) {
    do {
      const __m256i va01234567 = _mm256_cvtepi8_epi32(_mm_loadl_epi64((const __m128i*) input_a));
      const __m256i vb01234567 = _mm256_cvtepi8_epi32(_mm_loadl_epi64((const __m128i*) input_b));
      input_a += 8;
      input_b += 8;

      __m256i vacc01234567 = _mm256_add_epi32(vbias, _mm256_mullo_epi32(va01234567, va_multiplier));

      vacc01234567 = _mm256_add_epi32(vacc01234567, _mm256_mullo_epi32(vb01234567, vb_multiplier));

      vacc01234567 = _mm256_sra_epi32(vacc01234567, vshift);

      __m128i vout01234567 = _mm_adds_epi16(_mm_packs_epi32(_mm256_castsi256_si128(vacc01234567), _mm256_extracti128_si256(vacc01234567, 1)), _mm256_castsi256_si128(voutput_zero_point));
      __m128i vout0123456701234567 = _mm_packs_epi16(vout01234567, vout01234567);
      vout0123456701234567 = _mm_max_epi8(vout0123456701234567, voutput_min);
      vout0123456701234567 = _mm_min_epi8(vout0123456701234567, voutput_max);

      if XNN_LIKELY(batch >= (8 * sizeof(int8_t))) {
        _mm_storel_epi64((__m128i*) output, vout0123456701234567);
        output += 8;
        batch -= 8 * sizeof(int8_t);
      } else {
        if (batch & (4 * sizeof(int8_t))) {
          _mm_storeu_si32(output, vout0123456701234567);
          vout0123456701234567 = _mm_srli_epi64(vout0123456701234567, 32);
          output += 4;
        }
        if (batch & (2 * sizeof(int8_t))) {
          _mm_storeu_si16(output, vout0123456701234567);
          vout0123456701234567 = _mm_srli_epi32(vout0123456701234567, 16);
          output += 2;
        }
        if (batch & (1 * sizeof(int8_t))) {
          *output = (int8_t) _mm_extract_epi8(vout0123456701234567, 0);
        }
        batch = 0;
      }
    } while (batch != 0);
  }
}