File size: 15,558 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
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
// 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.

$assert ROW_TILE >= 1
$assert ACCUMULATORS >= 1
#include <assert.h>

#include <arm_neon.h>

#include <xnnpack/dwconv.h>
#include <xnnpack/intrinsics-polyfill.h>
#include <xnnpack/math.h>


void xnn_f16_dwconv2d_chw_ukernel_3x3s2p1__neonfp16arith_${ROW_TILE}x8${"_acc%d" % ACCUMULATORS if ACCUMULATORS > 1 else ""}(
    size_t input_height,
    size_t input_width,
    const void* input,
    const void* weights,
    const void* zero,
    void* output,
    uint32_t padding_top,
    const union xnn_f16_chw_params params[restrict XNN_MIN_ELEMENTS(1)]) XNN_OOB_READS
{
  assert(input_height != 0);
  assert(input_width != 0);
  assert(input_width % sizeof(uint16_t) == 0);
  assert(padding_top <= 1);

  #if XNN_ARCH_ARM64
    const uint16x8x2_t vminmax = vld2q_dup_u16(&params->neonfp16arith_stride2.min);
    const float16x8_t vmin = vreinterpretq_f16_u16(vminmax.val[0]);
    const float16x8_t vmax = vreinterpretq_f16_u16(vminmax.val[1]);
  #else
    // vld2_dup is to work around aarch32 clang bug with vld1q_dup
    const uint16x4x2_t vminmax = vld2_dup_u16(&params->neonfp16arith_stride2.min);
    const float16x8_t vmin = vreinterpretq_f16_u16(vcombine_u16(vminmax.val[0], vminmax.val[0]));
    const float16x8_t vmax = vreinterpretq_f16_u16(vcombine_u16(vminmax.val[1], vminmax.val[1]));
  #endif
  const uint16x8_t vmask_even = vld1q_u16(params->neonfp16arith_stride2.mask_even);
  const uint16x8_t vmask_odd  = vld1q_u16(params->neonfp16arith_stride2.mask_odd);

  const uint16_t* w = (const uint16_t*) weights;
  const float16x8_t vw01234567 = vreinterpretq_f16_u16(vld1q_u16(w));
  const float16x4_t vw89 = vreinterpret_f16_u32(vld1_dup_u32((const void*) (w + 8)));

  const size_t input_decrement = round_down_po2(input_width, 8 /* SIMD output width */ * 2 /* subsampling */ * sizeof(uint16_t));
  $if ROW_TILE > 1:
    const size_t output_width = round_down_po2((input_width + (2 /* padding */ - 3 /* kernel size */ + 2 /* subsampling */) * sizeof(uint16_t)) / 2, sizeof(uint16_t));

  const uint16_t* i0 = (const uint16_t*) ((uintptr_t) input - ((-padding_top) & input_width));
  const uint16_t* i1 = (const uint16_t*) ((uintptr_t) i0 + input_width);
  if XNN_UNPREDICTABLE(padding_top != 0) {
    i0 = zero;
  }
  $for M in range(2, 1 + 2 * ROW_TILE):
    const uint16_t* i${M} = (const uint16_t*) ((uintptr_t) i${M-1} + input_width);

  uint16_t* o0 = output;
  $for M in range(1, ROW_TILE):
    uint16_t* o${M} = (uint16_t*) ((uintptr_t) o${M-1} + output_width);

  size_t padded_input_height = input_height + padding_top + 1 /* padding bottom */;
  size_t output_height = (padded_input_height - 3 /* kernel size */ + 2 /* subsampling */) / 2;
  do {
    $for M in range(2, 1 + 2 * ROW_TILE):
      if XNN_UNPREDICTABLE(padded_input_height < ${2 + M}) {
        i${M} = zero;
        $if M % 2 == 1:
          o${(M - 1) // 2} = o${(M - 1) // 2 - 1};
      }

    $for M in range(1 + 2 * ROW_TILE):
      float16x8_t vi${M}x13579BDF = vreinterpretq_f16_u16(vmovq_n_u16(0));

    size_t w = input_width;
    for (; w >= 16 * sizeof(uint16_t); w -= 16 * sizeof(uint16_t)) {
      $for M in range(ROW_TILE):
        float16x8_t vo${M}p0 = vdupq_lane_f16(vget_low_f16(vw01234567), 0);

      $for M in range(1 + 2 * ROW_TILE):
        const uint16x8x2_t vi${M}xGIKMOQSUHJLNPRTV = vld2q_u16(i${M}); i${M} += 16;

      // Center column
      $for M in range(ROW_TILE):
        $if ACCUMULATORS > 1:
          float16x8_t vo${M}p1 = vmulq_lane_f16(vreinterpretq_f16_u16(vi${2*M}xGIKMOQSUHJLNPRTV.val[0]), vget_low_f16(vw01234567), 2);
        $else:
          #if XNN_ARCH_ARM64
            vo${M}p0 = vfmaq_laneq_f16(vo${M}p0, vreinterpretq_f16_u16(vi${2*M}xGIKMOQSUHJLNPRTV.val[0]), vw01234567, 2);
          #else
            vo${M}p0 = vmlaq_lane_f16(vo${M}p0, vreinterpretq_f16_u16(vi${2*M}xGIKMOQSUHJLNPRTV.val[0]), vget_low_f16(vw01234567), 2);
          #endif
      $for M in range(ROW_TILE):
        $if ACCUMULATORS > 2:
          float16x8_t vo${M}p2 = vmulq_lane_f16(vreinterpretq_f16_u16(vi${2*M+1}xGIKMOQSUHJLNPRTV.val[0]), vget_high_f16(vw01234567), 1);
        $else:
          #if XNN_ARCH_ARM64
            vo${M}p0 = vfmaq_laneq_f16(vo${M}p0, vreinterpretq_f16_u16(vi${2*M+1}xGIKMOQSUHJLNPRTV.val[0]), vw01234567, 5);
          #else
            vo${M}p0 = vmlaq_lane_f16(vo${M}p0, vreinterpretq_f16_u16(vi${2*M+1}xGIKMOQSUHJLNPRTV.val[0]), vget_high_f16(vw01234567), 1);
          #endif
      $for M in range(ROW_TILE):
        $if ACCUMULATORS > 3:
          float16x8_t vo${M}p3 = vmulq_lane_f16(vreinterpretq_f16_u16(vi${2*M+2}xGIKMOQSUHJLNPRTV.val[0]), vw89, 0);
        $else:
          #if XNN_ARCH_ARM64
            vo${M}p${4 % ACCUMULATORS} = vfmaq_lane_f16(vo${M}p${4 % ACCUMULATORS}, vreinterpretq_f16_u16(vi${2*M+2}xGIKMOQSUHJLNPRTV.val[0]), vw89, 0);
          #else
            vo${M}p${4 % ACCUMULATORS} = vmlaq_lane_f16(vo${M}p${4 % ACCUMULATORS}, vreinterpretq_f16_u16(vi${2*M+2}xGIKMOQSUHJLNPRTV.val[0]), vw89, 0);
          #endif
      // Left column
      $for M in range(1 + 2 * ROW_TILE):
        const float16x8_t vi${M}xFHJLNPRT = vextq_f16(vi${M}x13579BDF, vreinterpretq_f16_u16(vi${M}xGIKMOQSUHJLNPRTV.val[1]), 7);
        vi${M}x13579BDF = vreinterpretq_f16_u16(vi${M}xGIKMOQSUHJLNPRTV.val[1]);

      $for M in range(ROW_TILE):
        $if ACCUMULATORS > 4:
          float16x8_t vo${M}p4 = vmulq_lane_f16(vi${2*M}xFHJLNPRT, vget_low_f16(vw01234567), 1);
        $else:
          #if XNN_ARCH_ARM64
            vo${M}p${5 % ACCUMULATORS} = vfmaq_laneq_f16(vo${M}p${5 % ACCUMULATORS}, vi${2*M}xFHJLNPRT, vw01234567, 1);
          #else
            vo${M}p${5 % ACCUMULATORS} = vmlaq_lane_f16(vo${M}p${5 % ACCUMULATORS}, vi${2*M}xFHJLNPRT, vget_low_f16(vw01234567), 1);
          #endif
      $for M in range(ROW_TILE):
        $if ACCUMULATORS > 5:
          float16x8_t vo${M}p5 = vmulq_lane_f16(vi${2*M+1}xFHJLNPRT, vget_high_f16(vw01234567), 0);
        $else:
          #if XNN_ARCH_ARM64
            vo${M}p${6 % ACCUMULATORS} = vfmaq_laneq_f16(vo${M}p${6 % ACCUMULATORS}, vi${2*M+1}xFHJLNPRT, vw01234567, 4);
          #else
            vo${M}p${6 % ACCUMULATORS} = vmlaq_lane_f16(vo${M}p${6 % ACCUMULATORS}, vi${2*M+1}xFHJLNPRT, vget_high_f16(vw01234567), 0);
          #endif
      $for M in range(ROW_TILE):
        $if ACCUMULATORS > 6:
          float16x8_t vo${M}p6 = vmulq_lane_f16(vi${2*M+2}xFHJLNPRT, vget_high_f16(vw01234567), 1);
        $else:
          #if XNN_ARCH_ARM64
            vo${M}p${7 % ACCUMULATORS} = vfmaq_laneq_f16(vo${M}p${7 % ACCUMULATORS}, vi${2*M+2}xFHJLNPRT, vw01234567, 7);
          #else
            vo${M}p${7 % ACCUMULATORS} = vmlaq_lane_f16(vo${M}p${7 % ACCUMULATORS}, vi${2*M+2}xFHJLNPRT, vget_high_f16(vw01234567), 3);
          #endif
      // Right column
      $for M in range(ROW_TILE):
        #if XNN_ARCH_ARM64
          vo${M}p${8 % ACCUMULATORS} = vfmaq_laneq_f16(vo${M}p${8 % ACCUMULATORS}, vreinterpretq_f16_u16(vi${2*M}xGIKMOQSUHJLNPRTV.val[1]), vw01234567, 3);
        #else
          vo${M}p${8 % ACCUMULATORS} = vmlaq_lane_f16(vo${M}p${8 % ACCUMULATORS}, vreinterpretq_f16_u16(vi${2*M}xGIKMOQSUHJLNPRTV.val[1]), vget_low_f16(vw01234567), 3);
        #endif
      $for M in range(ROW_TILE):
        #if XNN_ARCH_ARM64
          vo${M}p${9 % ACCUMULATORS} = vfmaq_laneq_f16(vo${M}p${9 % ACCUMULATORS}, vreinterpretq_f16_u16(vi${2*M+1}xGIKMOQSUHJLNPRTV.val[1]), vw01234567, 6);
        #else
          vo${M}p${9 % ACCUMULATORS} = vmlaq_lane_f16(vo${M}p${9 % ACCUMULATORS}, vreinterpretq_f16_u16(vi${2*M+1}xGIKMOQSUHJLNPRTV.val[1]), vget_high_f16(vw01234567), 2);
        #endif
      $for M in range(ROW_TILE):
        #if XNN_ARCH_ARM64
          vo${M}p${10 % ACCUMULATORS} = vfmaq_lane_f16(vo${M}p${10 % ACCUMULATORS}, vreinterpretq_f16_u16(vi${2*M+2}xGIKMOQSUHJLNPRTV.val[1]), vw89, 1);
        #else
          vo${M}p${10 % ACCUMULATORS} = vmlaq_lane_f16(vo${M}p${10 % ACCUMULATORS}, vreinterpretq_f16_u16(vi${2*M+2}xGIKMOQSUHJLNPRTV.val[1]), vw89, 1);
        #endif
      $if ACCUMULATORS > 1:
        $ACC_SLICE = 1
        $while ACC_SLICE < ACCUMULATORS:
          $for A in range(0, ACCUMULATORS, ACC_SLICE * 2):
            $if A + ACC_SLICE < ACCUMULATORS:
              $for M in range(ROW_TILE):
                vo${M}p${A} = vaddq_f16(vo${M}p${A}, vo${M}p${A + ACC_SLICE});
          $ACC_SLICE *= 2

      $for M in range(ROW_TILE):
        float16x8_t vo${M} = vmaxq_f16(vo${M}p0, vmin);

      $for M in range(ROW_TILE):
        vo${M} = vminq_f16(vo${M}, vmax);

      $for M in reversed(range(ROW_TILE)):
        vst1q_u16(o${M}, vreinterpretq_u16_f16(vo${M})); o${M} += 8;
    }

    // Last block has 0-15 pixels to process.
    assert(w < 16 * sizeof(uint16_t));
    if XNN_LIKELY(w != 0) {
      $for M in range(ROW_TILE):
        float16x8_t vo${M}p0 = vdupq_lane_f16(vget_low_f16(vw01234567), 0);

      $for M in range(1 + 2 * ROW_TILE):
        const uint16x8x2_t vi${M}xGIKMOQSUHJLNPRTV = vld2q_u16(i${M});

      $for M in range(1 + 2 * ROW_TILE):
        const float16x8_t vi${M}xGIKMOQSU = vreinterpretq_f16_u16(vandq_u16(vmask_even, vi${M}xGIKMOQSUHJLNPRTV.val[0]));
        const float16x8_t vi${M}xHJLNPRTV = vreinterpretq_f16_u16(vandq_u16(vmask_odd,  vi${M}xGIKMOQSUHJLNPRTV.val[1]));

      // Center column
      $for M in range(ROW_TILE):
        $if ACCUMULATORS > 1:
          float16x8_t vo${M}p1 = vmulq_lane_f16(vi${2*M}xGIKMOQSU, vget_low_f16(vw01234567), 2);
        $else:
          #if XNN_ARCH_ARM64
            vo${M}p0 = vfmaq_laneq_f16(vo${M}p0, vi${2*M}xGIKMOQSU, vw01234567, 2);
          #else
            vo${M}p0 = vmlaq_lane_f16(vo${M}p0, vi${2*M}xGIKMOQSU, vget_low_f16(vw01234567), 2);
          #endif
      $for M in range(ROW_TILE):
        $if ACCUMULATORS > 2:
          float16x8_t vo${M}p2 = vmulq_lane_f16(vi${2*M+1}xGIKMOQSU, vget_high_f16(vw01234567), 1);
        $else:
          #if XNN_ARCH_ARM64
            vo${M}p0 = vfmaq_laneq_f16(vo${M}p0, vi${2*M+1}xGIKMOQSU, vw01234567, 5);
          #else
            vo${M}p0 = vmlaq_lane_f16(vo${M}p0, vi${2*M+1}xGIKMOQSU, vget_high_f16(vw01234567), 1);
          #endif
      $for M in range(ROW_TILE):
        $if ACCUMULATORS > 3:
          float16x8_t vo${M}p3 = vmulq_lane_f16(vi${2*M+2}xGIKMOQSU, vw89, 0);
        $else:
          #if XNN_ARCH_ARM64
            vo${M}p${4 % ACCUMULATORS} = vfmaq_lane_f16(vo${M}p${4 % ACCUMULATORS}, vi${2*M+2}xGIKMOQSU, vw89, 0);
          #else
            vo${M}p${4 % ACCUMULATORS} = vmlaq_lane_f16(vo${M}p${4 % ACCUMULATORS}, vi${2*M+2}xGIKMOQSU, vw89, 0);
          #endif
      // Left column
      $for M in range(1 + 2 * ROW_TILE):
        const float16x8_t vi${M}xFHJLNPRT = vextq_f16(vi${M}x13579BDF, vi${M}xHJLNPRTV, 7);

      $for M in range(ROW_TILE):
        $if ACCUMULATORS > 4:
          float16x8_t vo${M}p4 = vmulq_lane_f16(vi${2*M}xFHJLNPRT, vget_low_f16(vw01234567), 1);
        $else:
          #if XNN_ARCH_ARM64
            vo${M}p${5 % ACCUMULATORS} = vfmaq_laneq_f16(vo${M}p${5 % ACCUMULATORS}, vi${2*M}xFHJLNPRT, vw01234567, 1);
          #else
            vo${M}p${5 % ACCUMULATORS} = vmlaq_lane_f16(vo${M}p${5 % ACCUMULATORS}, vi${2*M}xFHJLNPRT, vget_low_f16(vw01234567), 1);
          #endif
      $for M in range(ROW_TILE):
        $if ACCUMULATORS > 5:
          float16x8_t vo${M}p5 = vmulq_lane_f16(vi${2*M+1}xFHJLNPRT, vget_high_f16(vw01234567), 0);
        $else:
          #if XNN_ARCH_ARM64
            vo${M}p${6 % ACCUMULATORS} = vfmaq_laneq_f16(vo${M}p${6 % ACCUMULATORS}, vi${2*M+1}xFHJLNPRT, vw01234567, 4);
          #else
            vo${M}p${6 % ACCUMULATORS} = vmlaq_lane_f16(vo${M}p${6 % ACCUMULATORS}, vi${2*M+1}xFHJLNPRT, vget_high_f16(vw01234567), 0);
          #endif
      $for M in range(ROW_TILE):
        $if ACCUMULATORS > 6:
          float16x8_t vo${M}p6 = vmulq_lane_f16(vi${2*M+2}xFHJLNPRT, vget_high_f16(vw01234567), 1);
        $else:
          #if XNN_ARCH_ARM64
            vo${M}p${7 % ACCUMULATORS} = vfmaq_laneq_f16(vo${M}p${7 % ACCUMULATORS}, vi${2*M+2}xFHJLNPRT, vw01234567, 7);
          #else
            vo${M}p${7 % ACCUMULATORS} = vmlaq_lane_f16(vo${M}p${7 % ACCUMULATORS}, vi${2*M+2}xFHJLNPRT, vget_high_f16(vw01234567), 3);
          #endif
      // Right column
      $for M in range(ROW_TILE):
        #if XNN_ARCH_ARM64
          vo${M}p${8 % ACCUMULATORS} = vfmaq_laneq_f16(vo${M}p${8 % ACCUMULATORS}, vi${2*M}xHJLNPRTV, vw01234567, 3);
        #else
          vo${M}p${8 % ACCUMULATORS} = vmlaq_lane_f16(vo${M}p${8 % ACCUMULATORS}, vi${2*M}xHJLNPRTV, vget_low_f16(vw01234567), 3);
        #endif
      $for M in range(ROW_TILE):
        #if XNN_ARCH_ARM64
          vo${M}p${9 % ACCUMULATORS} = vfmaq_laneq_f16(vo${M}p${9 % ACCUMULATORS}, vi${2*M+1}xHJLNPRTV, vw01234567, 6);
        #else
          vo${M}p${9 % ACCUMULATORS} = vmlaq_lane_f16(vo${M}p${9 % ACCUMULATORS}, vi${2*M+1}xHJLNPRTV, vget_high_f16(vw01234567), 2);
        #endif
      $for M in range(ROW_TILE):
        #if XNN_ARCH_ARM64
          vo${M}p${10 % ACCUMULATORS} = vfmaq_lane_f16(vo${M}p${10 % ACCUMULATORS}, vi${2*M+2}xHJLNPRTV, vw89, 1);
        #else
          vo${M}p${10 % ACCUMULATORS} = vmlaq_lane_f16(vo${M}p${10 % ACCUMULATORS}, vi${2*M+2}xHJLNPRTV, vw89, 1);
        #endif
      $if ACCUMULATORS > 1:
        $ACC_SLICE = 1
        $while ACC_SLICE < ACCUMULATORS:
          $for A in range(0, ACCUMULATORS, ACC_SLICE * 2):
            $if A + ACC_SLICE < ACCUMULATORS:
              $for M in range(ROW_TILE):
                vo${M}p${A} = vaddq_f16(vo${M}p${A}, vo${M}p${A + ACC_SLICE});
          $ACC_SLICE *= 2

      $for M in range(ROW_TILE):
        float16x8_t vo${M} = vmaxq_f16(vo${M}p0, vmin);

      $for M in range(ROW_TILE):
        vo${M} = vminq_f16(vo${M}, vmax);

      w += 1 * sizeof(uint16_t);

      if XNN_LIKELY(w == 16 * sizeof(uint16_t)) {
        $for M in reversed(range(ROW_TILE)):
          vst1q_u16(o${M}, vreinterpretq_u16_f16(vo${M})); o${M} += 8;
      } else {
        $for M in reversed(range(ROW_TILE)):
          float16x4_t vo${M}_lo = vget_low_f16(vo${M});

        if (w & (8 * sizeof(uint16_t))) {
         $for M in reversed(range(ROW_TILE)):
            vst1_u16(o${M}, vreinterpret_u16_f16(vo${M}_lo)); o${M} += 4;

          $for M in reversed(range(ROW_TILE)):
            vo${M}_lo = vget_high_f16(vo${M});
        }
        if (w & (4 * sizeof(uint16_t))) {
          $for M in reversed(range(ROW_TILE)):
            vst1_lane_u32((void*) o${M}, vreinterpret_u32_f16(vo${M}_lo), 0); o${M} += 2;

          $for M in range(ROW_TILE):
            vo${M}_lo = vext_f16(vo${M}_lo, vo${M}_lo, 2);
        }
        if (w & (2 * sizeof(uint16_t))) {
          $for M in reversed(range(ROW_TILE)):
            vst1_lane_u16(o${M}, vreinterpret_u16_f16(vo${M}_lo), 0); o${M} += 1;
        }
      }
    }

    i0 = (const uint16_t*) ((uintptr_t) i${2 * ROW_TILE} - input_decrement);
    $for M in range(1, 1 + 2 * ROW_TILE):
      i${M} = (const uint16_t*) ((uintptr_t) i${M-1} + input_width);

    $if ROW_TILE > 1:
      o0 = o${ROW_TILE - 1};
      $for M in range(1, ROW_TILE):
        o${M} = (uint16_t*) ((uintptr_t) o${M-1} + output_width);

    $if ROW_TILE > 1:
      output_height = doz(output_height, ${ROW_TILE});
      padded_input_height = doz(padded_input_height, ${ROW_TILE * 2});
    $else:
      output_height -= 1;
      padded_input_height -= 2;
  } while (output_height != 0);
}