aboutsummaryrefslogtreecommitdiffstats
path: root/libavcodec/x86/ac3dsp.asm
blob: 0ba980aa7ba6e011d552083b35385c2a3f73a089 (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
;*****************************************************************************
;* x86-optimized AC-3 DSP functions
;* Copyright (c) 2011 Justin Ruggles
;*
;* 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

; 16777216.0f - used in ff_float_to_fixed24()
pf_1_24: times 4 dd 0x4B800000

; used in ff_ac3_compute_mantissa_size()
cextern ac3_bap_bits
pw_bap_mul1: dw 21846, 21846, 0, 32768, 21846, 21846, 0, 32768
pw_bap_mul2: dw 5, 7, 0, 7, 5, 7, 0, 7

; used in ff_ac3_extract_exponents()
cextern pd_1
pd_151: times 4 dd 151

SECTION .text

;-----------------------------------------------------------------------------
; void ff_ac3_exponent_min(uint8_t *exp, int num_reuse_blocks, int nb_coefs)
;-----------------------------------------------------------------------------

%macro AC3_EXPONENT_MIN 0
cglobal ac3_exponent_min, 3, 4, 2, exp, reuse_blks, expn, offset
    shl  reuse_blksq, 8
    jz .end
    LOOP_ALIGN
.nextexp:
    mov      offsetq, reuse_blksq
    mova          m0, [expq+offsetq]
    sub      offsetq, 256
    LOOP_ALIGN
.nextblk:
    PMINUB        m0, [expq+offsetq], m1
    sub      offsetq, 256
    jae .nextblk
    mova      [expq], m0
    add         expq, mmsize
    sub        expnq, mmsize
    jg .nextexp
.end:
    RET
%endmacro

%define LOOP_ALIGN ALIGN 16
%if HAVE_SSE2_EXTERNAL
INIT_XMM sse2
AC3_EXPONENT_MIN
%endif
%undef LOOP_ALIGN

;-----------------------------------------------------------------------------
; void ff_float_to_fixed24(int32_t *dst, const float *src, unsigned int len)
;-----------------------------------------------------------------------------

INIT_XMM sse2
cglobal float_to_fixed24, 3, 3, 9, dst, src, len
    movaps     m0, [pf_1_24]
    shl      lenq, 2
    add      srcq, lenq
    add      dstq, lenq
    neg      lenq
.loop:
    movaps     m1, [srcq+lenq    ]
    movaps     m2, [srcq+lenq+16 ]
    movaps     m3, [srcq+lenq+32 ]
    movaps     m4, [srcq+lenq+48 ]
%ifdef m8
    movaps     m5, [srcq+lenq+64 ]
    movaps     m6, [srcq+lenq+80 ]
    movaps     m7, [srcq+lenq+96 ]
    movaps     m8, [srcq+lenq+112]
%endif
    mulps      m1, m0
    mulps      m2, m0
    mulps      m3, m0
    mulps      m4, m0
%ifdef m8
    mulps      m5, m0
    mulps      m6, m0
    mulps      m7, m0
    mulps      m8, m0
%endif
    cvtps2dq   m1, m1
    cvtps2dq   m2, m2
    cvtps2dq   m3, m3
    cvtps2dq   m4, m4
%ifdef m8
    cvtps2dq   m5, m5
    cvtps2dq   m6, m6
    cvtps2dq   m7, m7
    cvtps2dq   m8, m8
%endif
    movdqa  [dstq+lenq    ], m1
    movdqa  [dstq+lenq+16 ], m2
    movdqa  [dstq+lenq+32 ], m3
    movdqa  [dstq+lenq+48 ], m4
%ifdef m8
    movdqa  [dstq+lenq+64 ], m5
    movdqa  [dstq+lenq+80 ], m6
    movdqa  [dstq+lenq+96 ], m7
    movdqa  [dstq+lenq+112], m8
    add      lenq, 128
%else
    add      lenq, 64
%endif
    jl .loop
    RET

INIT_YMM avx
cglobal float_to_fixed24, 3, 3, 5, dst, src, len
    vbroadcastf128 m0, [pf_1_24]
    shl      lenq, 2
    add      srcq, lenq
    add      dstq, lenq
    neg      lenq
.loop:
    mulps      m1, m0, [srcq+lenq+mmsize*0]
    mulps      m2, m0, [srcq+lenq+mmsize*1]
    mulps      m3, m0, [srcq+lenq+mmsize*2]
    mulps      m4, m0, [srcq+lenq+mmsize*3]
    cvtps2dq   m1, m1
    cvtps2dq   m2, m2
    cvtps2dq   m3, m3
    cvtps2dq   m4, m4
    mova  [dstq+lenq+mmsize*0], m1
    mova  [dstq+lenq+mmsize*1], m2
    mova  [dstq+lenq+mmsize*2], m3
    mova  [dstq+lenq+mmsize*3], m4
    add      lenq, mmsize*4
    jl .loop
    RET

;------------------------------------------------------------------------------
; int ff_ac3_compute_mantissa_size(uint16_t mant_cnt[6][16])
;------------------------------------------------------------------------------

%macro PHADDD4 2 ; xmm src, xmm tmp
    movhlps  %2, %1
    paddd    %1, %2
    pshufd   %2, %1, 0x1
    paddd    %1, %2
%endmacro

INIT_XMM sse2
cglobal ac3_compute_mantissa_size, 1, 2, 4, mant_cnt, sum
    movdqa      m0, [mant_cntq      ]
    movdqa      m1, [mant_cntq+ 1*16]
    paddw       m0, [mant_cntq+ 2*16]
    paddw       m1, [mant_cntq+ 3*16]
    paddw       m0, [mant_cntq+ 4*16]
    paddw       m1, [mant_cntq+ 5*16]
    paddw       m0, [mant_cntq+ 6*16]
    paddw       m1, [mant_cntq+ 7*16]
    paddw       m0, [mant_cntq+ 8*16]
    paddw       m1, [mant_cntq+ 9*16]
    paddw       m0, [mant_cntq+10*16]
    paddw       m1, [mant_cntq+11*16]
    pmaddwd     m0, [ac3_bap_bits   ]
    pmaddwd     m1, [ac3_bap_bits+16]
    paddd       m0, m1
    PHADDD4     m0, m1
    movd      sumd, m0
    movdqa      m3, [pw_bap_mul1]
    movhpd      m0, [mant_cntq     +2]
    movlpd      m0, [mant_cntq+1*32+2]
    movhpd      m1, [mant_cntq+2*32+2]
    movlpd      m1, [mant_cntq+3*32+2]
    movhpd      m2, [mant_cntq+4*32+2]
    movlpd      m2, [mant_cntq+5*32+2]
    pmulhuw     m0, m3
    pmulhuw     m1, m3
    pmulhuw     m2, m3
    paddusw     m0, m1
    paddusw     m0, m2
    pmaddwd     m0, [pw_bap_mul2]
    PHADDD4     m0, m1
    movd       eax, m0
    add        eax, sumd
    RET

;------------------------------------------------------------------------------
; void ff_ac3_extract_exponents(uint8_t *exp, int32_t *coef, int nb_coefs)
;------------------------------------------------------------------------------

%macro PABSD 1-2 ; src/dst, unused
%if cpuflag(ssse3)
    pabsd    %1, %1
%else ; src/dst, tmp
    pxor     %2, %2
    pcmpgtd  %2, %1
    pxor     %1, %2
    psubd    %1, %2
%endif
%endmacro

%macro AC3_EXTRACT_EXPONENTS 0
cglobal ac3_extract_exponents, 3, 3, 4, exp, coef, len
    add     expq, lenq
    lea    coefq, [coefq+4*lenq]
    neg     lenq
    mova      m2, [pd_1]
    mova      m3, [pd_151]
.loop:
    ; move 4 32-bit coefs to xmm0
    mova      m0, [coefq+4*lenq]
    ; absolute value
    PABSD     m0, m1
    ; convert to float and extract exponents
    pslld     m0, 1
    por       m0, m2
    cvtdq2ps  m1, m0
    psrld     m1, 23
    mova      m0, m3
    psubd     m0, m1
    ; move the lowest byte in each of 4 dwords to the low dword
    ; NOTE: We cannot just extract the low bytes with pshufb because the dword
    ;       result for 16777215 is -1 due to float inaccuracy. Using packuswb
    ;       clips this to 0, which is the correct exponent.
    packssdw  m0, m0
    packuswb  m0, m0
    movd  [expq+lenq], m0

    add     lenq, 4
    jl .loop
    RET
%endmacro

%if HAVE_SSE2_EXTERNAL
INIT_XMM sse2
AC3_EXTRACT_EXPONENTS
%endif
%if HAVE_SSSE3_EXTERNAL
INIT_XMM ssse3
AC3_EXTRACT_EXPONENTS
%endif