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
|
/*
* Optimization of some functions from mpegvideo.c for armv5te
* Copyright (c) 2007 Siarhei Siamashka <ssvb@users.sourceforge.net>
*
* 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 "config.h"
#include "asm.S"
/*
* Special optimized version of dct_unquantize_h263_helper_c, it
* requires the block to be at least 8 bytes aligned, and may process
* more elements than requested. But it is guaranteed to never
* process more than 64 elements provided that count argument is <= 64,
* so it is safe. This function is optimized for a common distribution
* of values for nCoeffs (they are mostly multiple of 8 plus one or
* two extra elements). So this function processes data as 8 elements
* per loop iteration and contains optional 2 elements processing in
* the end.
*
* Inner loop should take 6 cycles per element on arm926ej-s (Nokia 770)
*/
function ff_dct_unquantize_h263_armv5te, export=1
push {r4-r9,lr}
mov ip, #0
subs r3, r3, #2
ble 2f
ldrd r4, [r0, #0]
1:
ldrd r6, [r0, #8]
rsbs r9, ip, r4, asr #16
addgt r9, r2, #0
rsblt r9, r2, #0
smlatbne r9, r4, r1, r9
rsbs lr, ip, r5, asr #16
addgt lr, r2, #0
rsblt lr, r2, #0
smlatbne lr, r5, r1, lr
rsbs r8, ip, r4, asl #16
addgt r8, r2, #0
rsblt r8, r2, #0
smlabbne r4, r4, r1, r8
rsbs r8, ip, r5, asl #16
addgt r8, r2, #0
rsblt r8, r2, #0
smlabbne r5, r5, r1, r8
strh r4, [r0], #2
strh r9, [r0], #2
strh r5, [r0], #2
strh lr, [r0], #2
rsbs r9, ip, r6, asr #16
addgt r9, r2, #0
rsblt r9, r2, #0
smlatbne r9, r6, r1, r9
rsbs lr, ip, r7, asr #16
addgt lr, r2, #0
rsblt lr, r2, #0
smlatbne lr, r7, r1, lr
rsbs r8, ip, r6, asl #16
addgt r8, r2, #0
rsblt r8, r2, #0
smlabbne r6, r6, r1, r8
rsbs r8, ip, r7, asl #16
addgt r8, r2, #0
rsblt r8, r2, #0
smlabbne r7, r7, r1, r8
strh r6, [r0], #2
strh r9, [r0], #2
strh r7, [r0], #2
strh lr, [r0], #2
subs r3, r3, #8
ldrgtd r4, [r0, #0] /* load data early to avoid load/use pipeline stall */
bgt 1b
adds r3, r3, #2
pople {r4-r9,pc}
2:
ldrsh r9, [r0, #0]
ldrsh lr, [r0, #2]
mov r8, r2
cmp r9, #0
rsblt r8, r2, #0
smlabbne r9, r9, r1, r8
mov r8, r2
cmp lr, #0
rsblt r8, r2, #0
smlabbne lr, lr, r1, r8
strh r9, [r0], #2
strh lr, [r0], #2
pop {r4-r9,pc}
endfunc
|
