Hugging Face's logo

Androidonnxfork
/
test

Model card Files Community
5
test
File size: 5,424 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
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
 
// 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
$SIMD_TILE = BATCH_TILE // 8
#include <assert.h>

#include <immintrin.h>

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


void xnn_f16_velu_ukernel__avx2_rr1_p3_x${BATCH_TILE}(
    size_t batch,
    const void* input,
    void* output,
    const union xnn_f16_elu_params params[restrict XNN_MIN_ELEMENTS(1)])
{
  assert(batch != 0);
  assert(batch % sizeof(uint16_t) == 0);
  assert(input != NULL);
  assert(output != NULL);

  const __m256 vprescale = _mm256_load_ps(params->avx2_rr1_p3.prescale);
  const __m256 vsat_cutoff = _mm256_load_ps(params->avx2_rr1_p3.sat_cutoff);
  const __m256 vmagic_bias = _mm256_load_ps(params->avx2_rr1_p3.magic_bias);
  const __m256 vlog2e = _mm256_load_ps(params->avx2_rr1_p3.log2e);
  const __m256 vminus_ln2 = _mm256_load_ps(params->avx2_rr1_p3.minus_ln2);
  const __m256 vc3 = _mm256_load_ps(params->avx2_rr1_p3.c3);
  const __m256 vc2 = _mm256_load_ps(params->avx2_rr1_p3.c2);
  const __m256 vc1 = _mm256_load_ps(params->avx2_rr1_p3.c1);
  const __m256 valpha = _mm256_load_ps(params->avx2_rr1_p3.alpha);
  const __m256 vbeta = _mm256_load_ps(params->avx2_rr1_p3.beta);

  const uint16_t* i = (const uint16_t*) input;
  uint16_t* o = (uint16_t*) output;
  $if BATCH_TILE > 8:
    for (; batch >= ${BATCH_TILE} * sizeof(uint16_t); batch -= ${BATCH_TILE} * sizeof(uint16_t)) {
      __m256 vx0 = _mm256_cvtph_ps(_mm_loadu_si128((const __m128i*) i));
      $for N in range(1, SIMD_TILE):
        __m256 vx${N} = _mm256_cvtph_ps(_mm_loadu_si128((const __m128i*) (i + ${N * 8})));
      i += ${BATCH_TILE};

      $for N in range(SIMD_TILE):
        const __m256 vz${N} = _mm256_max_ps(vsat_cutoff, _mm256_mul_ps(vx${N}, vprescale));

      $for N in range(SIMD_TILE):
        __m256 vn${N} = _mm256_fmadd_ps(vz${N}, vlog2e, vmagic_bias);

      $for N in range(SIMD_TILE):
        __m256 vs${N} = _mm256_castsi256_ps(_mm256_slli_epi32(_mm256_castps_si256(vn${N}), 23));
        vn${N} = _mm256_sub_ps(vn${N}, vmagic_bias);

      $for N in range(SIMD_TILE):
        __m256 vt${N} = _mm256_fmadd_ps(vn${N}, vminus_ln2, vz${N});

      $for N in range(SIMD_TILE):
        __m256 vp${N} = _mm256_fmadd_ps(vc3, vt${N}, vc2);

      $for N in range(SIMD_TILE):
        vp${N} = _mm256_fmadd_ps(vp${N}, vt${N}, vc1);
        vt${N} = _mm256_mul_ps(vt${N}, valpha);

      $for N in range(SIMD_TILE):
        vt${N} = _mm256_mul_ps(vt${N}, vs${N});
        vs${N} = _mm256_fmsub_ps(vs${N}, valpha, valpha);

      $for N in range(SIMD_TILE):
        const __m256 ve${N} = _mm256_fmadd_ps(vp${N}, vt${N}, vs${N});
        vx${N} = _mm256_mul_ps(vx${N}, vbeta);

      $for N in range(SIMD_TILE):
        const __m256 vy${N} = _mm256_blendv_ps(vx${N}, ve${N}, vx${N});

      _mm_storeu_si128((__m128i*) o, _mm256_cvtps_ph(vy0, _MM_FROUND_TO_NEAREST_INT));
      $for N in range(1, SIMD_TILE):
        _mm_storeu_si128((__m128i*) (o + ${N * 8}), _mm256_cvtps_ph(vy${N}, _MM_FROUND_TO_NEAREST_INT));
      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;

    const __m256 vz = _mm256_max_ps(vsat_cutoff, _mm256_mul_ps(vx, vprescale));

    __m256 vn = _mm256_fmadd_ps(vz, vlog2e, vmagic_bias);
    __m256 vs = _mm256_castsi256_ps(_mm256_slli_epi32(_mm256_castps_si256(vn), 23));
    vn = _mm256_sub_ps(vn, vmagic_bias);
    __m256 vt = _mm256_fmadd_ps(vn, vminus_ln2, vz);

    __m256 vp = _mm256_fmadd_ps(vc3, vt, vc2);
    vp = _mm256_fmadd_ps(vp, vt, vc1);
    vt = _mm256_mul_ps(vt, valpha);
    vt = _mm256_mul_ps(vt, vs);
    vs = _mm256_fmsub_ps(vs, valpha, valpha);
    const __m256 ve = _mm256_fmadd_ps(vp, vt, vs);
    vx = _mm256_mul_ps(vx, vbeta);
    const __m256 vy = _mm256_blendv_ps(vx, ve, vx);

    _mm_storeu_si128((__m128i*) o, _mm256_cvtps_ph(vy, _MM_FROUND_TO_NEAREST_INT));
    o += 8;
  }
  if XNN_UNLIKELY(batch != 0) {
    assert(batch >= 1 * sizeof(uint16_t));
    assert(batch <= 7 * sizeof(uint16_t));
    __m256 vx = _mm256_cvtph_ps(_mm_loadu_si128((const __m128i*) i));

    const __m256 vz = _mm256_max_ps(vsat_cutoff, _mm256_mul_ps(vx, vprescale));

    __m256 vn = _mm256_fmadd_ps(vz, vlog2e, vmagic_bias);
    __m256 vs = _mm256_castsi256_ps(_mm256_slli_epi32(_mm256_castps_si256(vn), 23));
    vn = _mm256_sub_ps(vn, vmagic_bias);
    __m256 vt = _mm256_fmadd_ps(vn, vminus_ln2, vz);

    __m256 vp = _mm256_fmadd_ps(vc3, vt, vc2);
    vp = _mm256_fmadd_ps(vp, vt, vc1);
    vt = _mm256_mul_ps(vt, valpha);
    vt = _mm256_mul_ps(vt, vs);
    vs = _mm256_fmsub_ps(vs, valpha, valpha);
    const __m256 ve = _mm256_fmadd_ps(vp, vt, vs);
    vx = _mm256_mul_ps(vx, vbeta);
    const __m256 vy = _mm256_blendv_ps(vx, ve, vx);

    __m128i vh = _mm256_cvtps_ph(vy, _MM_FROUND_TO_NEAREST_INT);
    if (batch & (4 * sizeof(uint16_t))) {
      _mm_storel_epi64((__m128i*) o, vh);
      vh = _mm_unpackhi_epi64(vh, vh);
      o += 4;
    }
    if (batch & (2 * sizeof(uint16_t))) {
      _mm_storeu_si32(o, vh);
      vh = _mm_srli_epi64(vh, 32);
      o += 2;
    }
    if (batch & (1 * sizeof(uint16_t))) {
      *o = (uint16_t) _mm_extract_epi16(vh, 0);
    }
  }
}