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
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
|
///*****************************************************************************
//*
//* Copyright (C) 2012 Ittiam Systems Pvt Ltd, Bangalore
//*
//* Licensed under the Apache License, Version 2.0 (the "License");
//* you may not use this file except in compliance with the License.
//* You may obtain a copy of the License at:
//*
//* http://www.apache.org/licenses/LICENSE-2.0
//*
//* Unless required by applicable law or agreed to in writing, software
//* distributed under the License is distributed on an "AS IS" BASIS,
//* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
//* See the License for the specific language governing permissions and
//* limitations under the License.
//*
//*****************************************************************************/
///**
//*******************************************************************************
//* //file
//* ihevc_inter_pred_chroma_vert_neon_w16inp_neon.s
//*
//* //brief
//* contains function definitions for inter prediction interpolation.
//* functions are coded using neon intrinsics and can be compiled using
//* rvct
//*
//* //author
//* yogeswaran rs / parthiban
//*
//* //par list of functions:
//*
//*
//* //remarks
//* none
//*
//*******************************************************************************
//*/
///**
///**
//*******************************************************************************
//*
//* //brief
//* chroma interprediction filter for 16bit vertical input.
//*
//* //par description:
//* applies a vertical filter with coefficients pointed to by 'pi1_coeff' to
//* the elements pointed by 'pu1_src' and writes to the location pointed by
//* 'pu1_dst' input is 16 bits the filter output is downshifted by 12 and
//* clipped to lie between 0 and 255 assumptions : the function is
//* optimized considering the fact width and height are multiple of 2.
//*
//* //param[in] pi2_src
//* word16 pointer to the source
//*
//* //param[out] pu1_dst
//* uword8 pointer to the destination
//*
//* //param[in] src_strd
//* integer source stride
//*
//* //param[in] dst_strd
//* integer destination stride
//*
//* //param[in] pi1_coeff
//* word8 pointer to the filter coefficients
//*
//* //param[in] ht
//* integer height of the array
//*
//* //param[in] wd
//* integer width of the array
//*
//* //returns
//*
//* //remarks
//* none
//*
//*******************************************************************************
//*/
//void ihevc_inter_pred_chroma_vert_w16inp(word16 *pi2_src,
// uword8 *pu1_dst,
// word32 src_strd,
// word32 dst_strd,
// word8 *pi1_coeff,
// word32 ht,
// word32 wd)
//**************variables vs registers*****************************************
//x0 => *pu1_src
//x1 => *pi2_dst
//x2 => src_strd
//x3 => dst_strd
.text
.align 4
.include "ihevc_neon_macros.s"
.globl ihevc_inter_pred_chroma_vert_w16inp_av8
.type ihevc_inter_pred_chroma_vert_w16inp_av8, %function
ihevc_inter_pred_chroma_vert_w16inp_av8:
// stmfd sp!, {x4-x12, x14} //stack stores the values of the arguments
stp x19, x20,[sp,#-16]!
mov x15,x4 // pi1_coeff
mov x16,x5 // ht
mov x17,x6 // wd
mov x4, x15 //loads pi1_coeff
mov x6, x17 //wd
lsl x2,x2,#1 //src_strd = 2* src_strd
mov x5,x16 //loads ht
ld1 {v0.8b},[x4] //loads pi1_coeff
sub x4,x0,x2 //pu1_src - src_strd
sxtl v0.8h, v0.8b //long the value
tst x6,#3 //checks wd == 2
dup v16.4h, v0.4h[0] //coeff_0
dup v17.4h, v0.4h[1] //coeff_1
dup v18.4h, v0.4h[2] //coeff_2
dup v19.4h, v0.4h[3] //coeff_3
bgt core_loop_ht_2 //jumps to loop handles wd 2
tst x5,#3 //checks ht == mul of 4
beq core_loop_ht_4 //jumps to loop handles ht mul of 4
core_loop_ht_2:
lsl x7,x2,#1 //2*src_strd
lsl x12,x3,#1 //2*dst_strd
lsl x9,x6,#2 //4*wd
sub x6,x12,x6,lsl #1 //2*dst_strd - 2*wd
sub x8,x7,x9 //2*src_strd - 4*wd
mov x12,x9 //4wd
inner_loop_ht_2:
add x0,x4,x2 //increments pi2_src
ld1 {v0.4h},[x4],#8 //loads pu1_src
smull v0.4s, v0.4h, v16.4h //vmull_s16(src_tmp1, coeff_0)
subs x12,x12,#8 //2wd + 8
ld1 {v2.4h},[x0],x2 //loads pi2_src
smull v7.4s, v2.4h, v16.4h //vmull_s16(src_tmp2, coeff_0)
ld1 {v3.4h},[x0],x2 //loads pi2_src
smlal v0.4s, v2.4h, v17.4h
ld1 {v6.4h},[x0],x2
smlal v7.4s, v3.4h, v17.4h
ld1 {v2.4h},[x0]
add x7,x1,x3 //pu1_dst + dst_strd
smlal v0.4s, v3.4h, v18.4h
smlal v7.4s, v6.4h, v18.4h
smlal v0.4s, v6.4h, v19.4h
smlal v7.4s, v2.4h, v19.4h
sqshrn v0.4h, v0.4s,#6 //right shift
sqshrn v30.4h, v7.4s,#6 //right shift
sqrshrun v0.8b, v0.8h,#6 //rounding shift
sqrshrun v30.8b, v30.8h,#6 //rounding shift
st1 {v0.s}[0],[x1],#4 //stores the loaded value
st1 {v30.s}[0],[x7] //stores the loaded value
bgt inner_loop_ht_2 //inner loop -again
//inner loop ends
subs x5,x5,#2 //increments ht
add x1,x1,x6 //pu1_dst += 2*dst_strd - 2*wd
mov x12,x9 //4wd
add x4,x4,x8 //pi1_src_tmp1 += 2*src_strd - 4*wd
bgt inner_loop_ht_2 //loop again
b end_loops //jumps to end
core_loop_ht_4:
lsl x7,x2,#2 //2*src_strd
lsl x12,x3,#2 //2*dst_strd
lsr x11, x6, #1 //divide by 2
sub x14,x12,x6,lsl #1 //2*dst_strd - 2*wd
sub x8,x7,x6,lsl #2 //2*src_strd - 4*wd
mul x12, x5 , x11 //multiply height by width
sub x12, x12,#4 //subtract by one for epilog
lsl x11, x6, #1 //2*wd
prolog:
add x0,x4,x2 //increments pi2_src
ld1 {v0.4h},[x4],#8 //loads pu1_src
ld1 {v1.4h},[x0],x2 //loads pi2_src
subs x11,x11,#4
ld1 {v2.4h},[x0],x2 //loads pi2_src
smull v30.4s, v0.4h, v16.4h //vmull_s16(src_tmp1, coeff_0)
ld1 {v3.4h},[x0],x2
smlal v30.4s, v1.4h, v17.4h
smlal v30.4s, v2.4h, v18.4h
add x9,x1,x3 //pu1_dst + dst_strd
smlal v30.4s, v3.4h, v19.4h
ld1 {v4.4h},[x0],x2
smull v28.4s, v1.4h, v16.4h //vmull_s16(src_tmp2, coeff_0)
add x20,x4,x8
csel x4, x20, x4,le
smlal v28.4s, v2.4h, v17.4h
ld1 {v5.4h},[x0],x2
smlal v28.4s, v3.4h, v18.4h
ld1 {v6.4h},[x0],x2
smlal v28.4s, v4.4h, v19.4h
lsl x20,x6,#1
csel x11, x20, x11,le
sqshrn v30.4h, v30.4s,#6 //right shift
smull v26.4s, v2.4h, v16.4h //vmull_s16(src_tmp2, coeff_0)
add x0,x4,x2
smlal v26.4s, v3.4h, v17.4h
smlal v26.4s, v4.4h, v18.4h
ld1 {v0.4h},[x4],#8 //loads pu1_src
smlal v26.4s, v5.4h, v19.4h
sqrshrun v30.8b, v30.8h,#6 //rounding shift
sqshrn v28.4h, v28.4s,#6 //right shift
ld1 {v1.4h},[x0],x2 //loads pi2_src
smull v24.4s, v3.4h, v16.4h //vmull_s16(src_tmp2, coeff_0)
st1 {v30.s}[0],[x1],#4 //stores the loaded value
smlal v24.4s, v4.4h, v17.4h
ld1 {v2.4h},[x0],x2 //loads pi2_src
smlal v24.4s, v5.4h, v18.4h
ld1 {v3.4h},[x0],x2
smlal v24.4s, v6.4h, v19.4h
add x20,x1,x14
csel x1, x20, x1,le
sqshrn v26.4h, v26.4s,#6 //right shift
subs x12,x12,#4
sqrshrun v28.8b, v28.8h,#6 //rounding shift
beq epilog //jumps to epilog
kernel_4:
smull v30.4s, v0.4h, v16.4h //vmull_s16(src_tmp1, coeff_0)
subs x11,x11,#4
smlal v30.4s, v1.4h, v17.4h
st1 {v28.s}[0],[x9],x3 //stores the loaded value
smlal v30.4s, v2.4h, v18.4h
smlal v30.4s, v3.4h, v19.4h
sqshrn v24.4h, v24.4s,#6 //right shift
sqrshrun v26.8b, v26.8h,#6 //rounding shift
ld1 {v4.4h},[x0],x2
smull v28.4s, v1.4h, v16.4h //vmull_s16(src_tmp2, coeff_0)
smlal v28.4s, v2.4h, v17.4h
smlal v28.4s, v3.4h, v18.4h
smlal v28.4s, v4.4h, v19.4h
st1 {v26.s}[0],[x9],x3 //stores the loaded value
add x20,x4,x8
csel x4, x20, x4,le
lsl x20,x6,#1
csel x11, x20, x11,le
sqshrn v30.4h, v30.4s,#6 //right shift
sqrshrun v24.8b, v24.8h,#6 //rounding shift
ld1 {v5.4h},[x0],x2
smull v26.4s, v2.4h, v16.4h //vmull_s16(src_tmp2, coeff_0)
ld1 {v6.4h},[x0],x2
smlal v26.4s, v3.4h, v17.4h
st1 {v24.s}[0],[x9] //stores the loaded value
add x0,x4,x2
smlal v26.4s, v4.4h, v18.4h
ld1 {v0.4h},[x4],#8 //loads pu1_src
smlal v26.4s, v5.4h, v19.4h
sqshrn v28.4h, v28.4s,#6 //right shift
sqrshrun v30.8b, v30.8h,#6 //rounding shift
ld1 {v1.4h},[x0],x2 //loads pi2_src
smull v24.4s, v3.4h, v16.4h //vmull_s16(src_tmp2, coeff_0)
add x9,x1,x3 //pu1_dst + dst_strd
ld1 {v2.4h},[x0],x2 //loads pi2_src
smlal v24.4s, v4.4h, v17.4h
ld1 {v3.4h},[x0],x2
smlal v24.4s, v5.4h, v18.4h
st1 {v30.s}[0],[x1],#4 //stores the loaded value
smlal v24.4s, v6.4h, v19.4h
sqshrn v26.4h, v26.4s,#6 //right shift
sqrshrun v28.8b, v28.8h,#6 //rounding shift
add x20,x1,x14
csel x1, x20, x1,le
subs x12,x12,#4
bgt kernel_4 //jumps to kernel_4
epilog:
smull v30.4s, v0.4h, v16.4h //vmull_s16(src_tmp1, coeff_0)
st1 {v28.s}[0],[x9],x3 //stores the loaded value
smlal v30.4s, v1.4h, v17.4h
smlal v30.4s, v2.4h, v18.4h
smlal v30.4s, v3.4h, v19.4h
sqshrn v24.4h, v24.4s,#6 //right shift
sqrshrun v26.8b, v26.8h,#6 //rounding shift
smull v28.4s, v1.4h, v16.4h //vmull_s16(src_tmp2, coeff_0)
ld1 {v4.4h},[x0],x2
smlal v28.4s, v2.4h, v17.4h
st1 {v26.s}[0],[x9],x3 //stores the loaded value
smlal v28.4s, v3.4h, v18.4h
smlal v28.4s, v4.4h, v19.4h
sqshrn v30.4h, v30.4s,#6 //right shift
sqrshrun v24.8b, v24.8h,#6 //rounding shift
smull v26.4s, v2.4h, v16.4h //vmull_s16(src_tmp2, coeff_0)
ld1 {v5.4h},[x0],x2
smlal v26.4s, v3.4h, v17.4h
smlal v26.4s, v4.4h, v18.4h
smlal v26.4s, v5.4h, v19.4h
sqshrn v28.4h, v28.4s,#6 //right shift
sqrshrun v30.8b, v30.8h,#6 //rounding shift
st1 {v24.s}[0],[x9] //stores the loaded value
smull v24.4s, v3.4h, v16.4h //vmull_s16(src_tmp2, coeff_0)
smlal v24.4s, v4.4h, v17.4h
add x9,x1,x3 //pu1_dst + dst_strd
ld1 {v6.4h},[x0],x2
smlal v24.4s, v5.4h, v18.4h
smlal v24.4s, v6.4h, v19.4h
st1 {v30.s}[0],[x1],#4 //stores the loaded value
sqrshrun v28.8b, v28.8h,#6 //rounding shift
sqshrn v26.4h, v26.4s,#6 //right shift
st1 {v28.s}[0],[x9],x3 //stores the loaded value
sqrshrun v26.8b, v26.8h,#6 //rounding shift
sqshrn v24.4h, v24.4s,#6 //right shift
st1 {v26.s}[0],[x9],x3 //stores the loaded value
sqrshrun v24.8b, v24.8h,#6 //rounding shift
st1 {v24.s}[0],[x9] //stores the loaded value
end_loops:
// ldmfd sp!,{x4-x12,x15} //reload the registers from sp
ldp x19, x20,[sp],#16
ret
|