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
|
;******************************************************************************
;* MMX optimized discrete wavelet trasnform
;* Copyright (c) 2010 David Conrad
;*
;* 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
;* 51, Inc., Foundation Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
;******************************************************************************
%include "x86inc.asm"
cextern horizontal_compose_dd97i_end_c
cextern horizontal_compose_haar0i_end_c
cextern horizontal_compose_haar1i_end_c
SECTION_RODATA
pw_1: times 8 dw 1
pw_2: times 8 dw 2
pw_8: times 8 dw 8
pw_16: times 8 dw 16
pw_1991: times 4 dw 9,-1
section .text
; %1 -= (%2 + %3 + 2)>>2 %4 is pw_2
%macro COMPOSE_53iL0 4
paddw %2, %3
paddw %2, %4
psraw %2, 2
psubw %1, %2
%endm
; m1 = %1 + (-m0 + 9*m1 + 9*%2 -%3 + 8)>>4
; if %4 is supplied, %1 is loaded unaligned from there
; m2: clobbered m3: pw_8 m4: pw_1991
%macro COMPOSE_DD97iH0 3-4
paddw m0, %3
paddw m1, %2
psubw m0, m3
mova m2, m1
punpcklwd m1, m0
punpckhwd m2, m0
pmaddwd m1, m4
pmaddwd m2, m4
%if %0 > 3
movu %1, %4
%endif
psrad m1, 4
psrad m2, 4
packssdw m1, m2
paddw m1, %1
%endm
%macro COMPOSE_VERTICAL 1
; void vertical_compose53iL0(IDWTELEM *b0, IDWTELEM *b1, IDWTELEM *b2,
; int width)
cglobal vertical_compose53iL0_%1, 4,4,1, b0, b1, b2, width
mova m2, [pw_2]
.loop:
sub widthd, mmsize/2
mova m1, [b0q+2*widthq]
mova m0, [b1q+2*widthq]
COMPOSE_53iL0 m0, m1, [b2q+2*widthq], m2
mova [b1q+2*widthq], m0
jg .loop
REP_RET
; void vertical_compose_dirac53iH0(IDWTELEM *b0, IDWTELEM *b1, IDWTELEM *b2,
; int width)
cglobal vertical_compose_dirac53iH0_%1, 4,4,1, b0, b1, b2, width
mova m1, [pw_1]
.loop:
sub widthd, mmsize/2
mova m0, [b0q+2*widthq]
paddw m0, [b2q+2*widthq]
paddw m0, m1
psraw m0, 1
paddw m0, [b1q+2*widthq]
mova [b1q+2*widthq], m0
jg .loop
REP_RET
; void vertical_compose_dd97iH0(IDWTELEM *b0, IDWTELEM *b1, IDWTELEM *b2,
; IDWTELEM *b3, IDWTELEM *b4, int width)
cglobal vertical_compose_dd97iH0_%1, 6,6,5, b0, b1, b2, b3, b4, width
mova m3, [pw_8]
mova m4, [pw_1991]
.loop:
sub widthd, mmsize/2
mova m0, [b0q+2*widthq]
mova m1, [b1q+2*widthq]
COMPOSE_DD97iH0 [b2q+2*widthq], [b3q+2*widthq], [b4q+2*widthq]
mova [b2q+2*widthq], m1
jg .loop
REP_RET
; void vertical_compose_dd137iL0(IDWTELEM *b0, IDWTELEM *b1, IDWTELEM *b2,
; IDWTELEM *b3, IDWTELEM *b4, int width)
cglobal vertical_compose_dd137iL0_%1, 6,6,6, b0, b1, b2, b3, b4, width
mova m3, [pw_16]
mova m4, [pw_1991]
.loop:
sub widthd, mmsize/2
mova m0, [b0q+2*widthq]
mova m1, [b1q+2*widthq]
mova m5, [b2q+2*widthq]
paddw m0, [b4q+2*widthq]
paddw m1, [b3q+2*widthq]
psubw m0, m3
mova m2, m1
punpcklwd m1, m0
punpckhwd m2, m0
pmaddwd m1, m4
pmaddwd m2, m4
psrad m1, 5
psrad m2, 5
packssdw m1, m2
psubw m5, m1
mova [b2q+2*widthq], m5
jg .loop
REP_RET
; void vertical_compose_haar(IDWTELEM *b0, IDWTELEM *b1, int width)
cglobal vertical_compose_haar_%1, 3,4,3, b0, b1, width
mova m3, [pw_1]
.loop:
sub widthd, mmsize/2
mova m1, [b1q+2*widthq]
mova m0, [b0q+2*widthq]
mova m2, m1
paddw m1, m3
psraw m1, 1
psubw m0, m1
mova [b0q+2*widthq], m0
paddw m2, m0
mova [b1q+2*widthq], m2
jg .loop
REP_RET
%endmacro
; extend the left and right edges of the tmp array by %1 and %2 respectively
%macro EDGE_EXTENSION 3
mov %3, [tmpq]
%assign %%i 1
%rep %1
mov [tmpq-2*%%i], %3
%assign %%i %%i+1
%endrep
mov %3, [tmpq+2*w2q-2]
%assign %%i 0
%rep %2
mov [tmpq+2*w2q+2*%%i], %3
%assign %%i %%i+1
%endrep
%endmacro
; On x86-64 this does a tail call to the C function to do the final bit
; x86-32 doesn't because isn't enough stack space for the additional argument x
%macro END_HORIZONTAL 1
shr wd, 1
%ifdef ARCH_X86_64
RET ;This RET was a CLEANUP call
jmp %1
%else
push xd
push wd
push tmpd
push bd
call %1
add esp, 16
RET
%endif
%endmacro
%macro HAAR_HORIZONTAL 2
; void horizontal_compose_haari(IDWTELEM *b, IDWTELEM *tmp, int width)
cglobal horizontal_compose_haar%2i_%1, 3,6,4, b, tmp, w, x, w2, b_w2
mov w2d, wd
xor xd, xd
shr w2d, 1
lea b_w2q, [bq+wq]
mova m3, [pw_1]
.lowpass_loop:
movu m1, [b_w2q + 2*xq]
mova m0, [bq + 2*xq]
paddw m1, m3
psraw m1, 1
psubw m0, m1
mova [tmpq + 2*xq], m0
add xd, mmsize/2
cmp xd, w2d
jl .lowpass_loop
xor xd, xd
and w2d, ~(mmsize/2 - 1)
cmp w2d, mmsize/2
jl .end
.highpass_loop:
mova m1, [b_w2q + 2*xq]
mova m0, [tmpq + 2*xq]
paddw m1, m0
; shift and interleave
%if %2 == 1
paddw m0, m3
paddw m1, m3
psraw m0, 1
psraw m1, 1
%endif
mova m2, m0
punpcklwd m0, m1
punpckhwd m2, m1
mova [bq+4*xq], m0
mova [bq+4*xq+mmsize], m2
add xd, mmsize/2
cmp xd, w2d
jl .highpass_loop
.end:
END_HORIZONTAL horizontal_compose_haar%2i_end_c
%endmacro
INIT_XMM
; void horizontal_compose_dd97i(IDWTELEM *b, IDWTELEM *tmp, int width)
cglobal horizontal_compose_dd97i_ssse3, 3,6,8, b, tmp, w, x, w2, b_w2
mov w2d, wd
xor xd, xd
shr w2d, 1
lea b_w2q, [bq+wq]
movu m4, [bq+wq]
mova m7, [pw_2]
pslldq m4, 14
.lowpass_loop:
movu m1, [b_w2q + 2*xq]
mova m0, [bq + 2*xq]
mova m2, m1
palignr m1, m4, 14
mova m4, m2
COMPOSE_53iL0 m0, m1, m2, m7
mova [tmpq + 2*xq], m0
add xd, mmsize/2
cmp xd, w2d
jl .lowpass_loop
EDGE_EXTENSION 1, 2, xw
; leave the last up to 7 (sse) or 3 (mmx) values for C
xor xd, xd
and w2d, ~(mmsize/2 - 1)
cmp w2d, mmsize/2
jl .end
mova m7, [tmpq-mmsize]
mova m0, [tmpq]
mova m5, [pw_1]
mova m3, [pw_8]
mova m4, [pw_1991]
.highpass_loop:
mova m6, m0
palignr m0, m7, 14
mova m7, [tmpq + 2*xq + 16]
mova m1, m7
mova m2, m7
palignr m1, m6, 2
palignr m2, m6, 4
COMPOSE_DD97iH0 m0, m6, m2, [b_w2q + 2*xq]
mova m0, m7
mova m7, m6
; shift and interleave
paddw m6, m5
paddw m1, m5
psraw m6, 1
psraw m1, 1
mova m2, m6
punpcklwd m6, m1
punpckhwd m2, m1
mova [bq+4*xq], m6
mova [bq+4*xq+mmsize], m2
add xd, mmsize/2
cmp xd, w2d
jl .highpass_loop
.end:
END_HORIZONTAL horizontal_compose_dd97i_end_c
%ifndef ARCH_X86_64
INIT_MMX
COMPOSE_VERTICAL mmx
HAAR_HORIZONTAL mmx, 0
HAAR_HORIZONTAL mmx, 1
%endif
;;INIT_XMM
INIT_XMM
COMPOSE_VERTICAL sse2
HAAR_HORIZONTAL sse2, 0
HAAR_HORIZONTAL sse2, 1
|
