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
|
;******************************************************************************
;* MMX/SSE2-optimized functions for the RV40 decoder
;* Copyright (C) 2012 Christophe Gisquet <christophe.gisquet@gmail.com>
;*
;* This file is part of Libav.
;*
;* Libav 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.
;*
;* Libav 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 Libav; if not, write to the Free Software
;* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
;******************************************************************************
%include "x86inc.asm"
%include "x86util.asm"
SECTION_RODATA
align 16
shift_round: times 8 dw 1 << (16 - 6)
cextern pw_16
SECTION .text
; %1=5bits weights?, %2=dst %3=src1 %4=src3 %5=stride if sse2
%macro RV40_WCORE 4-5
movh m4, [%3 + 0]
movh m5, [%4 + 0]
%if %0 == 4
%define OFFSET mmsize / 2
%else
; 8x8 block and sse2, stride was provided
%define OFFSET %5
%endif
movh m6, [%3 + OFFSET]
movh m7, [%4 + OFFSET]
%if %1 == 0
; 14bits weights
punpcklbw m4, m0
punpcklbw m5, m0
punpcklbw m6, m0
punpcklbw m7, m0
psllw m4, 7
psllw m5, 7
psllw m6, 7
psllw m7, 7
pmulhw m4, m3
pmulhw m5, m2
pmulhw m6, m3
pmulhw m7, m2
paddw m4, m5
paddw m6, m7
%else
; 5bits weights
%if cpuflag(ssse3)
punpcklbw m4, m5
punpcklbw m6, m7
pmaddubsw m4, m3
pmaddubsw m6, m3
%else
punpcklbw m4, m0
punpcklbw m5, m0
punpcklbw m6, m0
punpcklbw m7, m0
pmullw m4, m3
pmullw m5, m2
pmullw m6, m3
pmullw m7, m2
paddw m4, m5
paddw m6, m7
%endif
%endif
; bias and shift down
%if cpuflag(ssse3)
pmulhrsw m4, m1
pmulhrsw m6, m1
%else
paddw m4, m1
paddw m6, m1
psrlw m4, 5
psrlw m6, 5
%endif
packuswb m4, m6
%if %0 == 5
; Only called for 8x8 blocks and sse2
movh [%2 + 0], m4
movhps [%2 + %5], m4
%else
mova [%2], m4
%endif
%endmacro
%macro MAIN_LOOP 2
%if mmsize == 8
RV40_WCORE %2, r0, r1, r2
%if %1 == 16
RV40_WCORE %2, r0 + 8, r1 + 8, r2 + 8
%endif
; Prepare for next loop
add r0, r5
add r1, r5
add r2, r5
%else
%ifidn %1, 8
RV40_WCORE %2, r0, r1, r2, r5
; Prepare 2 next lines
lea r0, [r0 + 2 * r5]
lea r1, [r1 + 2 * r5]
lea r2, [r2 + 2 * r5]
%else
RV40_WCORE %2, r0, r1, r2
; Prepare single next line
add r0, r5
add r1, r5
add r2, r5
%endif
%endif
dec r6
%endmacro
; rv40_weight_func_%1(uint8_t *dst, uint8_t *src1, uint8_t *src2, int w1, int w2, int stride)
; %1=size %2=num of xmm regs
; The weights are FP0.14 notation of fractions depending on pts.
; For timebases without rounding error (i.e. PAL), the fractions
; can be simplified, and several operations can be avoided.
; Therefore, we check here whether they are multiples of 2^9 for
; those simplifications to occur.
%macro RV40_WEIGHT 3
cglobal rv40_weight_func_%1_%2, 6, 7, %3
%if cpuflag(ssse3)
mova m1, [shift_round]
%else
mova m1, [pw_16]
%endif
pxor m0, m0
; Set loop counter and increments
%if mmsize == 8
mov r6, %2
%else
mov r6, (%2 * %2) / mmsize
%endif
movd m2, r3
movd m3, r4
%ifidn %1,rnd
%define RND 0
SPLATW m2, m2
%else
%define RND 1
%if cpuflag(ssse3)
punpcklbw m3, m2
%else
SPLATW m2, m2
%endif
%endif
SPLATW m3, m3
.loop:
MAIN_LOOP %2, RND
jnz .loop
REP_RET
%endmacro
INIT_MMX mmx
RV40_WEIGHT rnd, 8, 3
RV40_WEIGHT rnd, 16, 4
RV40_WEIGHT nornd, 8, 3
RV40_WEIGHT nornd, 16, 4
INIT_XMM sse2
RV40_WEIGHT rnd, 8, 3
RV40_WEIGHT rnd, 16, 4
RV40_WEIGHT nornd, 8, 3
RV40_WEIGHT nornd, 16, 4
INIT_XMM ssse3
RV40_WEIGHT rnd, 8, 3
RV40_WEIGHT rnd, 16, 4
RV40_WEIGHT nornd, 8, 3
RV40_WEIGHT nornd, 16, 4
|
