Hugging Face's logo

Androidonnxfork
/
test

Model card Files Community
6
test
File size: 7,879 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
 
// 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 NR % 4 == 0
$ABC = "0123456789ABCDEFGHIJKLMNOPQRSTUVWXYZ"

#include <assert.h>

#include <arm_neon.h>

#include <xnnpack/gemm.h>


void xnn_bf16_gemm_minmax_ukernel_${MR}x${NR}c2__neonbf16_bfdot_lane_ld128(
    size_t mr,
    size_t nc,
    size_t kc,
    const void* restrict a,
    size_t a_stride,
    const void* restrict w_ptr,
    void* restrict c,
    size_t cm_stride,
    size_t cn_stride,
    const union xnn_bf16_minmax_params params[restrict XNN_MIN_ELEMENTS(1)])
{
  assert(mr != 0);
  assert(mr <= ${MR});
  assert(nc != 0);
  assert(kc != 0);
  assert(kc % sizeof(bfloat16_t) == 0);
  assert(a != NULL);
  assert(w_ptr != NULL);
  assert(c != NULL);

  const bfloat16_t* a0 = (const bfloat16_t*) a;
  bfloat16_t* c0 = (bfloat16_t*) c;
  $for M in range(1, MR):
    const bfloat16_t* a${M} = (const bfloat16_t*) ((uintptr_t) a${M-1} + a_stride);
    bfloat16_t* c${M} = (bfloat16_t*) ((uintptr_t) c${M-1} + cm_stride);
    $if M % 2 == 0:
      if XNN_UNPREDICTABLE(mr <= ${M}) {
        a${M} = a${M-1};
        c${M} = c${M-1};
      }
    $elif M + 1 == MR:
      if XNN_UNPREDICTABLE(mr != ${M+1}) {
        a${M} = a${M-1};
        c${M} = c${M-1};
      }
    $else:
      if XNN_UNPREDICTABLE(mr < ${M+1}) {
        a${M} = a${M-1};
        c${M} = c${M-1};
      }

  const bfloat16_t* w = (const bfloat16_t*) w_ptr;
  do {
    $for N in range(0, NR, 4):
      float32x4_t vacc0x${ABC[N:N+4]} = vcvt_f32_bf16(vld1_bf16(w)); w += 4;
    $for M in range(1, MR):
      $for N in range(0, NR, 4):
        float32x4_t vacc${M}x${ABC[N:N+4]} = vacc0x${ABC[N:N+4]};

    size_t k = kc;
    for (; k >= 8 * sizeof(bfloat16_t); k -= 8 * sizeof(bfloat16_t)) {
      $for M in range(MR):
        const bfloat16x8_t va${M} = vld1q_bf16(a${M}); a${M} += 8;

      $for K in range(4):
        $for N in range(0, NR, 4):
          const bfloat16x8_t vb${ABC[N:N+4]}c${ABC[2*K:2*K+2]} = vld1q_bf16(w); w += 8;

        $for N in range(0, NR, 4):
          $for M in range(MR):
            vacc${M}x${ABC[N:N+4]} = vbfdotq_laneq_f32(vacc${M}x${ABC[N:N+4]}, vb${ABC[N:N+4]}c${ABC[2*K:2*K+2]}, va${M}, ${K});
    }
    if XNN_UNLIKELY(k != 0) {
      $for M in range(MR):
        const bfloat16x8_t va${M} = vld1q_bf16(a${M}); a${M} = (const bfloat16_t*) ((uintptr_t) a${M} + k);

      $for N in range(0, NR, 4):
        const bfloat16x8_t vb${ABC[N:N+4]}c${ABC[0:2]} = vld1q_bf16(w); w += 8;

      $for M in range(MR):
        const uint32x4_t va${M}c${ABC[0:2]} = vdupq_lane_u32(vreinterpret_u32_bf16(vget_low_bf16(va${M})), 0);

      $for N in range(0, NR, 4):
        const uint32x4_t vm${ABC[N:N+4]}c${ABC[0:2]} = vreinterpretq_u32_u16(vceqq_u16(vreinterpretq_u16_bf16(vb${ABC[N:N+4]}c${ABC[0:2]}), vmovq_n_u16(0)));

      $for N in range(0, NR, 4):
        $for M in range(MR):
          const uint32x4_t va${M}x${ABC[N:N+4]}c${ABC[0:2]} = vbicq_u32(va${M}c${ABC[0:2]}, vm${ABC[N:N+4]}c${ABC[0:2]});
          vacc${M}x${ABC[N:N+4]} = vbfdotq_f32(vacc${M}x${ABC[N:N+4]}, vb${ABC[N:N+4]}c${ABC[0:2]}, vreinterpretq_bf16_u32(va${M}x${ABC[N:N+4]}c${ABC[0:2]}));

      if (k > 2 * sizeof(bfloat16_t)) {
        $for N in range(0, NR, 4):
          const bfloat16x8_t vb${ABC[N:N+4]}c${ABC[2:4]} = vld1q_bf16(w); w += 8;

        $for M in range(MR):
          const uint32x4_t va${M}c${ABC[2:4]} = vdupq_lane_u32(vreinterpret_u32_bf16(vget_low_bf16(va${M})), 1);

        $for N in range(0, NR, 4):
          const uint32x4_t vm${ABC[N:N+4]}c${ABC[2:4]} = vreinterpretq_u32_u16(vceqq_u16(vreinterpretq_u16_bf16(vb${ABC[N:N+4]}c${ABC[2:4]}), vmovq_n_u16(0)));

        $for N in range(0, NR, 4):
          $for M in range(MR):
            const uint32x4_t va${M}x${ABC[N:N+4]}c${ABC[2:4]} = vbicq_u32(va${M}c${ABC[2:4]}, vm${ABC[N:N+4]}c${ABC[2:4]});
            vacc${M}x${ABC[N:N+4]} = vbfdotq_f32(vacc${M}x${ABC[N:N+4]}, vb${ABC[N:N+4]}c${ABC[2:4]}, vreinterpretq_bf16_u32(va${M}x${ABC[N:N+4]}c${ABC[2:4]}));

        if (k > 4 * sizeof(bfloat16_t)) {
          $for N in range(0, NR, 4):
            const bfloat16x8_t vb${ABC[N:N+4]}c${ABC[4:6]} = vld1q_bf16(w); w += 8;

          $for M in range(MR):
            const uint32x4_t va${M}c${ABC[4:6]} = vdupq_lane_u32(vreinterpret_u32_bf16(vget_high_bf16(va${M})), 0);

          $for N in range(0, NR, 4):
            const uint32x4_t vm${ABC[N:N+4]}c${ABC[4:6]} = vreinterpretq_u32_u16(vceqq_u16(vreinterpretq_u16_bf16(vb${ABC[N:N+4]}c${ABC[4:6]}), vmovq_n_u16(0)));

          $for N in range(0, NR, 4):
            $for M in range(MR):
              const uint32x4_t va${M}x${ABC[N:N+4]}c${ABC[4:6]} = vbicq_u32(va${M}c${ABC[4:6]}, vm${ABC[N:N+4]}c${ABC[4:6]});
              vacc${M}x${ABC[N:N+4]} = vbfdotq_f32(vacc${M}x${ABC[N:N+4]}, vb${ABC[N:N+4]}c${ABC[4:6]}, vreinterpretq_bf16_u32(va${M}x${ABC[N:N+4]}c${ABC[4:6]}));

          if (k > 6 * sizeof(bfloat16_t)) {
            $for N in range(0, NR, 4):
              const bfloat16x8_t vb${ABC[N:N+4]}c${ABC[6:8]} = vld1q_bf16(w); w += 8;

            $for M in range(MR):
              const uint32x4_t va${M}c${ABC[6:8]} = vdupq_lane_u32(vreinterpret_u32_bf16(vget_high_bf16(va${M})), 1);

            $for N in range(0, NR, 4):
              const uint32x4_t vm${ABC[N:N+4]}c${ABC[6:8]} = vreinterpretq_u32_u16(vceqq_u16(vreinterpretq_u16_bf16(vb${ABC[N:N+4]}c${ABC[6:8]}), vmovq_n_u16(0)));

            $for N in range(0, NR, 4):
              $for M in range(MR):
                const uint32x4_t va${M}x${ABC[N:N+4]}c${ABC[6:8]} = vbicq_u32(va${M}c${ABC[6:8]}, vm${ABC[N:N+4]}c${ABC[6:8]});
                vacc${M}x${ABC[N:N+4]} = vbfdotq_f32(vacc${M}x${ABC[N:N+4]}, vb${ABC[N:N+4]}c${ABC[6:8]}, vreinterpretq_bf16_u32(va${M}x${ABC[N:N+4]}c${ABC[6:8]}));
          }
        }
      }
    }

    const float32x4_t vmax = vld1q_dup_f32(&params->scalar.max);
    $for N in range(0, NR, 4):
      $for M in range(MR):
        vacc${M}x${ABC[N:N+4]} = vminq_f32(vacc${M}x${ABC[N:N+4]}, vmax);

    const float32x4_t vmin = vld1q_dup_f32(&params->scalar.min);
    $for N in range(0, NR, 4):
      $for M in range(MR):
        vacc${M}x${ABC[N:N+4]} = vmaxq_f32(vacc${M}x${ABC[N:N+4]}, vmin);

    $for N in range(0, NR, 4):
      $for M in range(MR):
        bfloat16x4_t vout${M}x${ABC[N:N+4]} = vcvt_bf16_f32(vacc${M}x${ABC[N:N+4]});

    if XNN_LIKELY(nc >= ${NR}) {
      $for M in range(MR):
        vst1_bf16(c${M}, vout${M}x${ABC[0:4]});
        $for N in range(4, NR, 4):
          vst1_bf16(c${M} + ${N}, vout${M}x${ABC[N:N+4]});
        c${M} = (bfloat16_t*) ((uintptr_t) c${M} + cn_stride);

      $for M in range(MR):
        a${M} = (const bfloat16_t*) ((uintptr_t) a${M} - kc);

      nc -= ${NR};
    } else {
      $for LOG2N in reversed(range(NR.bit_length())):
        $if NR != 1 << LOG2N:
          if (nc & ${1 << LOG2N}) {
            $if LOG2N >= 2:
              $for N in range(0, 1 << LOG2N, 4):
                $for M in range(MR):
                  vst1_bf16(c${M}, vout${M}x${ABC[N:N+4]}); c${M} += 4;

              $for M in range(MR):
                $for N in range(0, NR - (1 << LOG2N), 4):
                  vout${M}x${ABC[N:N+4]} = vout${M}x${ABC[N + (1 << LOG2N):N + (1 << LOG2N)+4]};
            $elif LOG2N == 1:
              $for M in range(MR):
                vst1_lane_u32((void*) c${M}, vreinterpret_u32_bf16(vout${M}x${ABC[0:4]}), 0); c${M} += 2;

              $for M in range(MR):
                vout${M}x${ABC[0:4]} = vreinterpret_bf16_u16(vext_u16(vreinterpret_u16_bf16(vout${M}x${ABC[0:4]}), vreinterpret_u16_bf16(vout${M}x${ABC[0:4]}), 2));
            $elif LOG2N == 0:
              $for M in range(MR):
                vst1_lane_bf16(c${M}, vout${M}x${ABC[0:4]}, 0);
          }

      nc = 0;
    }
  } while (nc != 0);
}