File size: 3,745 Bytes
8b7c501
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
// Copyright 2022 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 <immintrin.h>

#include <xnnpack/common.h>
#include <xnnpack/intrinsics-polyfill.h>
#include <xnnpack/vunary.h>


void xnn_f16_vhswish_ukernel__f16c_x${BATCH_TILE}(
    size_t batch,
    const void* restrict input,
    void* restrict output,
    const union xnn_f16_hswish_params params[restrict XNN_MIN_ELEMENTS(1)]) XNN_OOB_READS
{
  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;

  const __m256 vsixth = _mm256_load_ps(params->avx.sixth);
  const __m256 vthree = _mm256_load_ps(params->avx.three);
  const __m128i vsix = _mm_load_si128((const __m128i*) params->avx.six);
  const __m128i vzero = _mm_setzero_si128();

  $if BATCH_TILE > 8:
    for (; batch >= ${BATCH_TILE} * sizeof(uint16_t); batch -= ${BATCH_TILE} * sizeof(uint16_t)) {
      __m256 vx${ABC[0:8]} = _mm256_cvtph_ps(_mm_loadu_si128((const __m128i*) i));
      $for N in range(8, BATCH_TILE, 8):
        __m256 vx${ABC[N:N+8]} = _mm256_cvtph_ps(_mm_loadu_si128((const __m128i*) (i + ${N})));
      i += ${BATCH_TILE};

      $for N in range(0, BATCH_TILE, 8):
        __m128i vacc${ABC[N:N+8]} = _mm256_cvtps_ph(_mm256_add_ps(vx${ABC[N:N+8]}, vthree), _MM_FROUND_TO_NEAREST_INT);
        vx${ABC[N:N+8]} = _mm256_cvtph_ps(_mm256_cvtps_ph(_mm256_mul_ps(vx${ABC[N:N+8]}, vsixth), _MM_FROUND_TO_NEAREST_INT));

      $for N in range(0, BATCH_TILE, 8):
        vacc${ABC[N:N+8]} = _mm_max_epi16(vacc${ABC[N:N+8]}, vzero);

      $for N in range(0, BATCH_TILE, 8):
        vacc${ABC[N:N+8]} = _mm_min_epi16(vacc${ABC[N:N+8]}, vsix);

      $for N in range(0, BATCH_TILE, 8):
        vacc${ABC[N:N+8]} = _mm256_cvtps_ph(_mm256_mul_ps(_mm256_cvtph_ps(vacc${ABC[N:N+8]}), vx${ABC[N:N+8]}), _MM_FROUND_TO_NEAREST_INT);

      _mm_storeu_si128((__m128i*) o, vacc${ABC[0:8]});
      $for N in range(8, BATCH_TILE, 8):
        _mm_storeu_si128((__m128i*) (o + ${N}), vacc${ABC[N:N+8]});
      o += ${BATCH_TILE};
    }
  for (; batch >= 8 * sizeof(uint16_t); batch -= 8 * sizeof(uint16_t)) {
    __m256 vx = _mm256_cvtph_ps(_mm_loadu_si128((const __m128i*) i));
    i += 8;
    __m128i vacc = _mm256_cvtps_ph(_mm256_add_ps(vx, vthree), _MM_FROUND_TO_NEAREST_INT);
    vx = _mm256_cvtph_ps(_mm256_cvtps_ph(_mm256_mul_ps(vx, vsixth), _MM_FROUND_TO_NEAREST_INT));
    vacc = _mm_max_epi16(vacc, vzero);
    vacc = _mm_min_epi16(vacc, vsix);
    vacc = _mm256_cvtps_ph(_mm256_mul_ps(_mm256_cvtph_ps(vacc), vx), _MM_FROUND_TO_NEAREST_INT);
    _mm_storeu_si128((__m128i*) o, vacc);
    o += 8;
  }
  if XNN_UNLIKELY(batch != 0) {
    __m256 vx = _mm256_cvtph_ps(_mm_loadu_si128((const __m128i*) i));
    __m128i vacc = _mm256_cvtps_ph(_mm256_add_ps(vx, vthree), _MM_FROUND_TO_NEAREST_INT);
    vx = _mm256_cvtph_ps(_mm256_cvtps_ph(_mm256_mul_ps(vx, vsixth), _MM_FROUND_TO_NEAREST_INT));
    vacc = _mm_max_epi16(vacc, vzero);
    vacc = _mm_min_epi16(vacc, vsix);
    vacc = _mm256_cvtps_ph(_mm256_mul_ps(_mm256_cvtph_ps(vacc), vx), _MM_FROUND_TO_NEAREST_INT);

    if (batch & (4 * sizeof(uint16_t))) {
      _mm_storel_epi64((__m128i*) o, vacc);
      vacc = _mm_unpackhi_epi64(vacc, vacc);
      o += 4;
    }
    if (batch & (2 * sizeof(uint16_t))) {
      _mm_storeu_si32(o, vacc);
      vacc = _mm_srli_epi64(vacc, 32);
      o += 2;
    }
    if (batch & (1 * sizeof(uint16_t))) {
      *o = (uint16_t) _mm_extract_epi16(vacc, 0);
    }
  }
}