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
|
;************************************************************************
;* SIMD-optimized lossless video encoding functions
;* Copyright (c) 2000, 2001 Fabrice Bellard
;* Copyright (c) 2002-2004 Michael Niedermayer <michaelni@gmx.at>
;*
;* MMX optimization by Nick Kurshev <nickols_k@mail.ru>
;* Conversion to NASM format by Tiancheng "Timothy" Gu <timothygu99@gmail.com>
;*
;* 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 "libavutil/x86/x86util.asm"
SECTION .text
; void ff_diff_bytes(uint8_t *dst, const uint8_t *src1, const uint8_t *src2,
; intptr_t w);
%macro DIFF_BYTES_PROLOGUE 0
%if ARCH_X86_32
cglobal diff_bytes, 3,5,2, dst, src1, src2
%define wq r4q
DECLARE_REG_TMP 3
mov wq, r3mp
%else
cglobal diff_bytes, 4,5,2, dst, src1, src2, w
DECLARE_REG_TMP 4
%endif ; ARCH_X86_32
%define i t0q
%endmacro
; labels to jump to if w < regsize and w < 0
%macro DIFF_BYTES_LOOP_PREP 2
mov i, wq
and i, -2 * regsize
js %2
jz %1
add dstq, i
add src1q, i
add src2q, i
neg i
%endmacro
; mov type used for src1q, dstq, first reg, second reg
%macro DIFF_BYTES_LOOP_CORE 4
%if mmsize != 16
mov%1 %3, [src1q + i]
mov%1 %4, [src1q + i + regsize]
psubb %3, [src2q + i]
psubb %4, [src2q + i + regsize]
mov%2 [dstq + i], %3
mov%2 [regsize + dstq + i], %4
%else
; SSE enforces alignment of psubb operand
mov%1 %3, [src1q + i]
movu %4, [src2q + i]
psubb %3, %4
mov%2 [dstq + i], %3
mov%1 %3, [src1q + i + regsize]
movu %4, [src2q + i + regsize]
psubb %3, %4
mov%2 [regsize + dstq + i], %3
%endif
%endmacro
%macro DIFF_BYTES_BODY 2 ; mov type used for src1q, for dstq
%define regsize mmsize
.loop_%1%2:
DIFF_BYTES_LOOP_CORE %1, %2, m0, m1
add i, 2 * regsize
jl .loop_%1%2
.skip_main_%1%2:
and wq, 2 * regsize - 1
jz .end_%1%2
%if mmsize > 16
; fall back to narrower xmm
%define regsize (mmsize / 2)
DIFF_BYTES_LOOP_PREP .setup_loop_gpr_aa, .end_aa
.loop2_%1%2:
DIFF_BYTES_LOOP_CORE %1, %2, xm0, xm1
add i, 2 * regsize
jl .loop2_%1%2
.setup_loop_gpr_%1%2:
and wq, 2 * regsize - 1
jz .end_%1%2
%endif
add dstq, wq
add src1q, wq
add src2q, wq
neg wq
.loop_gpr_%1%2:
mov t0b, [src1q + wq]
sub t0b, [src2q + wq]
mov [dstq + wq], t0b
inc wq
jl .loop_gpr_%1%2
.end_%1%2:
REP_RET
%endmacro
%if ARCH_X86_32
INIT_MMX mmx
DIFF_BYTES_PROLOGUE
%define regsize mmsize
DIFF_BYTES_LOOP_PREP .skip_main_aa, .end_aa
DIFF_BYTES_BODY a, a
%undef i
%endif
INIT_XMM sse2
DIFF_BYTES_PROLOGUE
%define regsize mmsize
DIFF_BYTES_LOOP_PREP .skip_main_aa, .end_aa
test dstq, regsize - 1
jnz .loop_uu
test src1q, regsize - 1
jnz .loop_ua
DIFF_BYTES_BODY a, a
DIFF_BYTES_BODY u, a
DIFF_BYTES_BODY u, u
%undef i
%if HAVE_AVX2_EXTERNAL
INIT_YMM avx2
DIFF_BYTES_PROLOGUE
%define regsize mmsize
; Directly using unaligned SSE2 version is marginally faster than
; branching based on arguments.
DIFF_BYTES_LOOP_PREP .skip_main_uu, .end_uu
test dstq, regsize - 1
jnz .loop_uu
test src1q, regsize - 1
jnz .loop_ua
DIFF_BYTES_BODY a, a
DIFF_BYTES_BODY u, a
DIFF_BYTES_BODY u, u
%undef i
%endif
|
