aboutsummaryrefslogtreecommitdiffstats
path: root/libavcodec/x86/lossless_videodsp.asm
blob: eb1b80506ee93293ca7def37a5c8657522917a01 (plain) (blame)
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
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
;******************************************************************************
;* SIMD lossless video DSP utils
;* Copyright (c) 2008 Loren Merritt
;* Copyright (c) 2014 Michael Niedermayer
;* Copyright (c) 2017 Jokyo Images
;*
;* This file is part of FFmpeg.
;*
;* FFmpeg is free software; you can redistribute it and/or
;* modify it under the terms of the GNU Lesser General Public
;* License as published by the Free Software Foundation; either
;* version 2.1 of the License, or (at your option) any later version.
;*
;* FFmpeg is distributed in the hope that it will be useful,
;* but WITHOUT ANY WARRANTY; without even the implied warranty of
;* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
;* Lesser General Public License for more details.
;*
;* You should have received a copy of the GNU Lesser General Public
;* License along with FFmpeg; if not, write to the Free Software
;* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
;******************************************************************************

%include "libavutil/x86/x86util.asm"

SECTION_RODATA

cextern pb_15
pb_zzzzzzzz77777777: times 8 db -1
pb_7: times 8 db 7
pb_ef: times 8 db 14,15
pb_67: times 8 db  6, 7
pb_zzzz3333zzzzbbbb: db -1,-1,-1,-1,3,3,3,3,-1,-1,-1,-1,11,11,11,11
pb_zz11zz55zz99zzdd: db -1,-1,1,1,-1,-1,5,5,-1,-1,9,9,-1,-1,13,13
pb_zzzz2323zzzzabab: db -1,-1,-1,-1, 2, 3, 2, 3,-1,-1,-1,-1,10,11,10,11
pb_zzzzzzzz67676767: db -1,-1,-1,-1,-1,-1,-1,-1, 6, 7, 6, 7, 6, 7, 6, 7

SECTION .text

;------------------------------------------------------------------------------
; void ff_add_median_pred(uint8_t *dst, const uint8_t *top,
;                         const uint8_t *diff, int w,
;                         int *left, int *left_top)
;------------------------------------------------------------------------------
INIT_XMM sse2
cglobal add_median_pred, 6,6,8, dst, top, diff, w, left, left_top
    movu    m0, [topq]
    mova    m2, m0
    movd    m4, [left_topq]
    LSHIFT  m2, 1
    mova    m1, m0
    por     m4, m2
    movd    m3, [leftq]
    psubb   m0, m4 ; t-tl
    add    dstq, wq
    add    topq, wq
    add   diffq, wq
    neg      wq
    jmp .skip
.loop:
    movu    m4, [topq+wq]
    mova    m0, m4
    LSHIFT  m4, 1
    por     m4, m1
    mova    m1, m0 ; t
    psubb   m0, m4 ; t-tl
.skip:
    movu    m2, [diffq+wq]
%assign i 0
%rep mmsize
    mova    m4, m0
    paddb   m4, m3 ; t-tl+l
    mova    m5, m3
    pmaxub  m3, m1
    pminub  m5, m1
    pminub  m3, m4
    pmaxub  m3, m5 ; median
    paddb   m3, m2 ; +residual
%if i==0
    mova    m7, m3
    LSHIFT  m7, mmsize-1
%else
    mova    m6, m3
    RSHIFT  m7, 1
    LSHIFT  m6, mmsize-1
    por     m7, m6
%endif
%if i<mmsize-1
    RSHIFT  m0, 1
    RSHIFT  m1, 1
    RSHIFT  m2, 1
%endif
%assign i i+1
%endrep
    movu [dstq+wq], m7
    add      wq, mmsize
    jl .loop
    movzx   r2d, byte [dstq-1]
    mov [leftq], r2d
    movzx   r2d, byte [topq-1]
    mov [left_topq], r2d
    RET


%macro ADD_LEFT_LOOP 2 ; %1 = dst_is_aligned, %2 = src_is_aligned
    add     srcq, wq
    add     dstq, wq
    neg     wq
%%.loop:
    pshufb  xm0, xm5
%if %2
    mova    m1, [srcq+wq]
%else
    movu    m1, [srcq+wq]
%endif
    psllw   m2, m1, 8
    paddb   m1, m2
    pshufb  m2, m1, m3
    paddb   m1, m2
    pshufb  m2, m1, m4
    paddb   m1, m2
%if mmsize >= 16
    pshufb  m2, m1, m6
    paddb   m1, m2
%endif
    paddb   xm0, xm1
%if %1
    mova    [dstq+wq], xm0
%else
    movq    [dstq+wq], xm0
    movhps  [dstq+wq+8], xm0
%endif

%if mmsize == 32
    vextracti128    xm2, m1, 1 ; get second lane of the ymm
    pshufb          xm0, xm5   ; set alls val to last val of the first lane
    paddb           xm0, xm2
;store val
%if %1
    mova    [dstq+wq+16], xm0
%else;
    movq    [dstq+wq+16], xm0
    movhps  [dstq+wq+16+8], xm0
%endif
%endif
    add     wq, mmsize
    jl %%.loop
%if mmsize == 32
    movzx   eax, byte [dstq - 1]
%else;
    mov     eax, mmsize-1
    sub     eax, wd
    movd    m1, eax
    pshufb  m0, m1
    movd    eax, m0
%endif
    RET
%endmacro

;------------------------------------------------------------------------------
; int ff_add_left_pred(uint8_t *dst, const uint8_t *src, int w, int left)
;------------------------------------------------------------------------------
INIT_MMX ssse3
cglobal add_left_pred, 3,3,7, dst, src, w, left
.skip_prologue:
    mova    m5, [pb_7]
    mova    m4, [pb_zzzz3333zzzzbbbb]
    mova    m3, [pb_zz11zz55zz99zzdd]
    movd    m0, leftm
    psllq   m0, 56
    ADD_LEFT_LOOP 1, 1

%macro ADD_LEFT_PRED_UNALIGNED 0
cglobal add_left_pred_unaligned, 3,3,7, dst, src, w, left
    mova    xm5, [pb_15]
    VBROADCASTI128    m6, [pb_zzzzzzzz77777777]
    VBROADCASTI128    m4, [pb_zzzz3333zzzzbbbb]
    VBROADCASTI128    m3, [pb_zz11zz55zz99zzdd]
    movd    xm0, leftm
    pslldq  xm0, 15
    test    srcq, mmsize - 1
    jnz .src_unaligned
    test    dstq, mmsize - 1
    jnz .dst_unaligned
    ADD_LEFT_LOOP 1, 1
.dst_unaligned:
    ADD_LEFT_LOOP 0, 1
.src_unaligned:
    ADD_LEFT_LOOP 0, 0
%endmacro

INIT_XMM ssse3
ADD_LEFT_PRED_UNALIGNED

%if HAVE_AVX2_EXTERNAL
INIT_YMM avx2
ADD_LEFT_PRED_UNALIGNED
%endif

;------------------------------------------------------------------------------
; void ff_add_bytes(uint8_t *dst, uint8_t *src, ptrdiff_t w);
;------------------------------------------------------------------------------
%macro ADD_BYTES 0
cglobal add_bytes, 3,4,2, dst, src, w, size
    mov  sizeq, wq
    and  sizeq, -2*mmsize
    jz  .2
    add   dstq, sizeq
    add   srcq, sizeq
    neg  sizeq
.1:
    mova    m0, [srcq + sizeq]
    mova    m1, [srcq + sizeq + mmsize]
    paddb   m0, [dstq + sizeq]
    paddb   m1, [dstq + sizeq + mmsize]
    mova   [dstq + sizeq], m0
    mova   [dstq + sizeq + mmsize], m1
    add  sizeq, 2*mmsize
    jl .1
.2:
    and     wq, 2*mmsize-1
    jz    .end
    add   dstq, wq
    add   srcq, wq
    neg     wq
.3:
    mov  sizeb, [srcq + wq]
    add [dstq + wq], sizeb
    inc     wq
    jl .3
.end:
    REP_RET
%endmacro

INIT_XMM sse2
ADD_BYTES

%if HAVE_AVX2_EXTERNAL
INIT_YMM avx2
ADD_BYTES
%endif

%macro ADD_HFYU_LEFT_LOOP_INT16 2 ; %1 = dst alignment (a/u), %2 = src alignment (a/u)
    add     wd, wd
    add     srcq, wq
    add     dstq, wq
    neg     wq
%%.loop:
    mov%2   m1, [srcq+wq]
    mova    m2, m1
    pslld   m1, 16
    paddw   m1, m2
    mova    m2, m1

    pshufb  m1, m3
    paddw   m1, m2
    pshufb  m0, m5
%if mmsize == 16
    mova    m2, m1
    pshufb  m1, m4
    paddw   m1, m2
%endif
    paddw   m0, m1
    pand    m0, m7
%ifidn %1, a
    mova    [dstq+wq], m0
%else
    movq    [dstq+wq], m0
    movhps  [dstq+wq+8], m0
%endif
    add     wq, mmsize
    jl %%.loop
    mov     eax, mmsize-1
    sub     eax, wd
    mov     wd, eax
    shl     wd, 8
    lea     eax, [wd+eax-1]
    movd    m1, eax
    pshufb  m0, m1
    movd    eax, m0
    RET
%endmacro

;---------------------------------------------------------------------------------------------
; int add_left_pred_int16(uint16_t *dst, const uint16_t *src, unsigned mask, int w, int left)
;---------------------------------------------------------------------------------------------
INIT_MMX ssse3
cglobal add_left_pred_int16, 4,4,8, dst, src, mask, w, left
.skip_prologue:
    mova    m5, [pb_67]
    mova    m3, [pb_zzzz2323zzzzabab]
    movd    m0, leftm
    psllq   m0, 48
    movd    m7, maskm
    SPLATW  m7 ,m7
    ADD_HFYU_LEFT_LOOP_INT16 a, a

INIT_XMM ssse3
cglobal add_left_pred_int16_unaligned, 4,4,8, dst, src, mask, w, left
    mova    m5, [pb_ef]
    mova    m4, [pb_zzzzzzzz67676767]
    mova    m3, [pb_zzzz2323zzzzabab]
    movd    m0, leftm
    pslldq  m0, 14
    movd    m7, maskm
    SPLATW  m7 ,m7
    test    srcq, 15
    jnz .src_unaligned
    test    dstq, 15
    jnz .dst_unaligned
    ADD_HFYU_LEFT_LOOP_INT16 a, a
.dst_unaligned:
    ADD_HFYU_LEFT_LOOP_INT16 u, a
.src_unaligned:
    ADD_HFYU_LEFT_LOOP_INT16 u, u


;---------------------------------------------------------------------------------------------
; void add_gradient_pred(uint8_t *src, const ptrdiff_t stride, const ptrdiff_t width)
;---------------------------------------------------------------------------------------------
%macro ADD_GRADIENT_PRED 0
cglobal add_gradient_pred, 3,4,5, src, stride, width, tmp
    mova         xm0, [pb_15]

;load src - 1 in xm1
    movd         xm1, [srcq-1]
%if cpuflag(avx2)
    vpbroadcastb xm1, xm1
%else
    pxor         xm2, xm2
    pshufb       xm1, xm2
%endif

    add    srcq, widthq
    neg  widthq
    neg strideq

.loop:
    lea    tmpq, [srcq + strideq]
    mova     m2, [tmpq + widthq] ; A = src[x-stride]
    movu     m3, [tmpq + widthq - 1] ; B = src[x - (stride + 1)]
    mova     m4, [srcq + widthq] ; current val (src[x])

    psubb    m2, m3; A - B

; prefix sum A-B
    pslldq   m3, m2, 1
    paddb    m2, m3
    pslldq   m3, m2, 2
    paddb    m2, m3
    pslldq   m3, m2, 4
    paddb    m2, m3
    pslldq   m3, m2, 8
    paddb    m2, m3

; prefix sum current val
    pslldq   m3, m4, 1
    paddb    m4, m3
    pslldq   m3, m4, 2
    paddb    m4, m3
    pslldq   m3, m4, 4
    paddb    m4, m3
    pslldq   m3, m4, 8
    paddb    m4, m3

; last sum
    paddb                    m2, m4 ; current + (A - B)

    paddb                   xm1, xm2 ; += C
    mova        [srcq + widthq], xm1 ; store

    pshufb                  xm1, xm0 ; put last val in all val of xm1

%if mmsize == 32
    vextracti128            xm2, m2, 1 ; get second lane of the ymm
    paddb                   xm1, xm2; += C

    mova   [srcq + widthq + 16], xm1 ; store
    pshufb                  xm1, xm0 ; put last val in all val of m1
%endif

    add         widthq, mmsize
    jl .loop
    RET

%endmacro

INIT_XMM ssse3
ADD_GRADIENT_PRED

%if HAVE_AVX2_EXTERNAL
INIT_YMM avx2
ADD_GRADIENT_PRED
%endif