Hugging Face's logo

Androidonnxfork
/
test

Model card Files Community
6
test
File size: 8,364 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
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
 
// Copyright 2021 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 <wasm_simd128.h>

#include <xnnpack/common.h>
#include <xnnpack/vcvt.h>


$WASM_V16X8_LANESELECT = "wasm_i16x8_relaxed_laneselect" if RELAXED else "wasm_v128_bitselect"
$ISA = "wasmrelaxedsimd" if RELAXED else "wasmsimd"
void xnn_f32_f16_vcvt_ukernel__${ISA}_x${BATCH_TILE}(
    size_t batch,
    const float* input,
    void* output,
    const union xnn_f32_f16_cvt_params params[restrict XNN_MIN_ELEMENTS(1)]) XNN_OOB_READS
{
  assert(batch != 0);
  assert(batch % sizeof(float) == 0);
  assert(input != NULL);
  assert(output != NULL);

  const v128_t vexp_bias = wasm_v128_load64_splat(params->wasmsimd.exp_bias);
  const v128_t vscale_to_inf = wasm_v128_load64_splat(params->wasmsimd.scale_to_inf);
  const v128_t vexpw_max = wasm_v128_load64_splat(params->wasmsimd.expw_max);
  const v128_t vscale_to_zero = wasm_v128_load64_splat(params->wasmsimd.scale_to_zero);
  const v128_t vbias_min = wasm_v128_load64_splat(params->wasmsimd.bias_min);
  const v128_t vmanth_mask = wasm_v128_load64_splat(params->wasmsimd.manth_mask);
  const v128_t vexph_mask = wasm_v128_load64_splat(params->wasmsimd.exph_mask);
  const v128_t vnanh = wasm_v128_load64_splat(params->wasmsimd.nanh);

  uint16_t* o = (uint16_t*) output;
  $if BATCH_TILE > 8:
    for (; batch >= ${BATCH_TILE} * sizeof(float); batch -= ${BATCH_TILE} * sizeof(float)) {
      const v128_t vx0 = wasm_v128_load(input);
      $for N in range(1, 2*SIMD_TILE):
        const v128_t vx${N} = wasm_v128_load(input + ${N * 4});
      input += ${BATCH_TILE};

      $for N in range(2*SIMD_TILE):
        const v128_t vabsx${N} = wasm_f32x4_abs(vx${N});

      $for N in range(2*SIMD_TILE):
        const v128_t vsignx${N} = wasm_v128_xor(vx${N}, vabsx${N});

      $for N in range(2*SIMD_TILE):
        v128_t vbias${N} = wasm_i32x4_add(vabsx${N}, vexp_bias);

      $for N in range(2*SIMD_TILE):
        v128_t vf${N} = wasm_f32x4_mul(vabsx${N}, vscale_to_inf);

      $for N in range(2*SIMD_TILE):
        const v128_t vnanmaskw${N} = wasm_i32x4_gt(vabsx${N}, vexpw_max);

      $for N in range(2*SIMD_TILE):
        vbias${N} = wasm_v128_and(vbias${N}, vexpw_max);

      $for N in range(2*SIMD_TILE):
        vf${N} = wasm_f32x4_mul(vf${N}, vscale_to_zero);

      $for N in range(SIMD_TILE):
        const v128_t vnanmaskh${N} = wasm_i16x8_narrow_i32x4(vnanmaskw${2*N}, vnanmaskw${2*N+1});

      $for N in range(SIMD_TILE):
        const v128_t vsignh${N} = wasm_i16x8_narrow_i32x4(vsignx${2*N}, vsignx${2*N+1});

      $for N in range(2*SIMD_TILE):
        vbias${N} = wasm_i16x8_max(vbias${N}, vbias_min);

      $for N in range(2*SIMD_TILE):
        vf${N} = wasm_f32x4_add(vf${N}, vbias${N});

      $for N in range(2*SIMD_TILE):
        v128_t vexpw${N} = wasm_i32x4_shr(vf${N}, 13);

      $for N in range(2*SIMD_TILE):
        const v128_t vmantw${N} = wasm_v128_and(vf${N}, vmanth_mask);

      $for N in range(2*SIMD_TILE):
        vexpw${N} = wasm_v128_and(vexpw${N}, vexph_mask);

      $for N in range(2*SIMD_TILE):
        const v128_t vnonsignw${N} = wasm_i32x4_add(vmantw${N}, vexpw${N});

      $for N in range(SIMD_TILE):
        const v128_t vnonsignh${N} = wasm_i16x8_narrow_i32x4(vnonsignw${2*N}, vnonsignw${2*N+1});

      $for N in range(SIMD_TILE):
        const v128_t vabsh${N} = ${WASM_V16X8_LANESELECT}(vnanh, vnonsignh${N}, vnanmaskh${N});

      $for N in range(SIMD_TILE):
        const v128_t vh${N} = wasm_v128_or(vabsh${N}, vsignh${N});

      wasm_v128_store(o, vh0);
      $for N in range(1, SIMD_TILE):
        wasm_v128_store(o + ${N * 8}, vh${N});
      o += ${BATCH_TILE};
    }
  for (; batch >= 8 * sizeof(float); batch -= 8 * sizeof(float)) {
    const v128_t vx_lo = wasm_v128_load(input);
    const v128_t vx_hi = wasm_v128_load(input + 4);
    input += 8;

    const v128_t vabsx_lo = wasm_f32x4_abs(vx_lo);
    const v128_t vabsx_hi = wasm_f32x4_abs(vx_hi);

    const v128_t vsignx_lo = wasm_v128_xor(vx_lo, vabsx_lo);
    const v128_t vsignx_hi = wasm_v128_xor(vx_hi, vabsx_hi);
    v128_t vbias_lo = wasm_i32x4_add(vabsx_lo, vexp_bias);
    v128_t vbias_hi = wasm_i32x4_add(vabsx_hi, vexp_bias);
    v128_t vf_lo = wasm_f32x4_mul(vabsx_lo, vscale_to_inf);
    v128_t vf_hi = wasm_f32x4_mul(vabsx_hi, vscale_to_inf);
    const v128_t vnanmaskw_lo = wasm_i32x4_gt(vabsx_lo, vexpw_max);
    const v128_t vnanmaskw_hi = wasm_i32x4_gt(vabsx_hi, vexpw_max);

    vbias_lo = wasm_v128_and(vbias_lo, vexpw_max);
    vbias_hi = wasm_v128_and(vbias_hi, vexpw_max);
    vf_lo = wasm_f32x4_mul(vf_lo, vscale_to_zero);
    vf_hi = wasm_f32x4_mul(vf_hi, vscale_to_zero);
    const v128_t vnanmaskh = wasm_i16x8_narrow_i32x4(vnanmaskw_lo, vnanmaskw_hi);
    const v128_t vsignh = wasm_i16x8_narrow_i32x4(vsignx_lo, vsignx_hi);

    vbias_lo = wasm_i16x8_max(vbias_lo, vbias_min);
    vbias_hi = wasm_i16x8_max(vbias_hi, vbias_min);

    vf_lo = wasm_f32x4_add(vf_lo, vbias_lo);
    vf_hi = wasm_f32x4_add(vf_hi, vbias_hi);

    v128_t vexpw_lo = wasm_i32x4_shr(vf_lo, 13);
    v128_t vexpw_hi = wasm_i32x4_shr(vf_hi, 13);
    const v128_t vmantw_lo = wasm_v128_and(vf_lo, vmanth_mask);
    const v128_t vmantw_hi = wasm_v128_and(vf_hi, vmanth_mask);

    vexpw_lo = wasm_v128_and(vexpw_lo, vexph_mask);
    vexpw_hi = wasm_v128_and(vexpw_hi, vexph_mask);

    const v128_t vnonsignw_lo = wasm_i32x4_add(vmantw_lo, vexpw_lo);
    const v128_t vnonsignw_hi = wasm_i32x4_add(vmantw_hi, vexpw_hi);

    const v128_t vnonsignh = wasm_i16x8_narrow_i32x4(vnonsignw_lo, vnonsignw_hi);

    const v128_t vabsh = ${WASM_V16X8_LANESELECT}(vnanh, vnonsignh, vnanmaskh);

    const v128_t vh = wasm_v128_or(vabsh, vsignh);

    wasm_v128_store(o, vh);
    o += 8;
  }
  if XNN_UNPREDICTABLE(batch != 0) {
    const v128_t vx_lo = wasm_v128_load(input);
    const float* input_hi = (const float*) ((uintptr_t) input + (batch & (4 * sizeof(float))));
    const v128_t vx_hi = wasm_v128_load(input_hi);

    const v128_t vabsx_lo = wasm_f32x4_abs(vx_lo);
    const v128_t vabsx_hi = wasm_f32x4_abs(vx_hi);

    const v128_t vsignx_lo = wasm_v128_xor(vx_lo, vabsx_lo);
    const v128_t vsignx_hi = wasm_v128_xor(vx_hi, vabsx_hi);
    v128_t vbias_lo = wasm_i32x4_add(vabsx_lo, vexp_bias);
    v128_t vbias_hi = wasm_i32x4_add(vabsx_hi, vexp_bias);
    v128_t vf_lo = wasm_f32x4_mul(vabsx_lo, vscale_to_inf);
    v128_t vf_hi = wasm_f32x4_mul(vabsx_hi, vscale_to_inf);
    const v128_t vnanmaskw_lo = wasm_i32x4_gt(vabsx_lo, vexpw_max);
    const v128_t vnanmaskw_hi = wasm_i32x4_gt(vabsx_hi, vexpw_max);

    vbias_lo = wasm_v128_and(vbias_lo, vexpw_max);
    vbias_hi = wasm_v128_and(vbias_hi, vexpw_max);
    vf_lo = wasm_f32x4_mul(vf_lo, vscale_to_zero);
    vf_hi = wasm_f32x4_mul(vf_hi, vscale_to_zero);
    const v128_t vnanmaskh = wasm_i16x8_narrow_i32x4(vnanmaskw_lo, vnanmaskw_hi);
    const v128_t vsignh = wasm_i16x8_narrow_i32x4(vsignx_lo, vsignx_hi);

    vbias_lo = wasm_i16x8_max(vbias_lo, vbias_min);
    vbias_hi = wasm_i16x8_max(vbias_hi, vbias_min);

    vf_lo = wasm_f32x4_add(vf_lo, vbias_lo);
    vf_hi = wasm_f32x4_add(vf_hi, vbias_hi);

    v128_t vexpw_lo = wasm_i32x4_shr(vf_lo, 13);
    v128_t vexpw_hi = wasm_i32x4_shr(vf_hi, 13);
    const v128_t vmantw_lo = wasm_v128_and(vf_lo, vmanth_mask);
    const v128_t vmantw_hi = wasm_v128_and(vf_hi, vmanth_mask);

    vexpw_lo = wasm_v128_and(vexpw_lo, vexph_mask);
    vexpw_hi = wasm_v128_and(vexpw_hi, vexph_mask);

    const v128_t vnonsignw_lo = wasm_i32x4_add(vmantw_lo, vexpw_lo);
    const v128_t vnonsignw_hi = wasm_i32x4_add(vmantw_hi, vexpw_hi);

    const v128_t vnonsignh = wasm_i16x8_narrow_i32x4(vnonsignw_lo, vnonsignw_hi);

    const v128_t vabsh = ${WASM_V16X8_LANESELECT}(vnanh, vnonsignh, vnanmaskh);

    v128_t vh = wasm_v128_or(vabsh, vsignh);

    if (batch & (4 * sizeof(float))) {
      wasm_v128_store64_lane(o, vh, 0);
      vh = wasm_v64x2_shuffle(vh, vh, 1, 1);
      o += 4;
    }
    if (batch & (2 * sizeof(float))) {
      wasm_v128_store32_lane(o, vh, 0);
      vh = wasm_i64x2_shr(vh, 32);
      o += 2;
    }
    if (batch & (1 * sizeof(float))) {
      wasm_v128_store16_lane(o, vh, 0);
    }
  }
}