1 ;*****************************************************************************
2 ;* x86-optimized AC-3 DSP utils
3 ;* Copyright (c) 2011 Justin Ruggles
5 ;* This file is part of FFmpeg.
7 ;* FFmpeg is free software; you can redistribute it and/or
8 ;* modify it under the terms of the GNU Lesser General Public
9 ;* License as published by the Free Software Foundation; either
10 ;* version 2.1 of the License, or (at your option) any later version.
12 ;* FFmpeg is distributed in the hope that it will be useful,
13 ;* but WITHOUT ANY WARRANTY; without even the implied warranty of
14 ;* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
15 ;* Lesser General Public License for more details.
17 ;* You should have received a copy of the GNU Lesser General Public
18 ;* License along with FFmpeg; if not, write to the Free Software
19 ;* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
20 ;******************************************************************************
23 %include "x86util.asm"
27 ; 16777216.0f - used in ff_float_to_fixed24()
28 pf_1_24: times 4 dd 0x4B800000
32 ;-----------------------------------------------------------------------------
33 ; void ff_ac3_exponent_min(uint8_t *exp, int num_reuse_blocks, int nb_coefs)
34 ;-----------------------------------------------------------------------------
36 %macro AC3_EXPONENT_MIN 1
37 cglobal ac3_exponent_min_%1, 3,4,2, exp, reuse_blks, expn, offset
42 mov offsetq, reuse_blksq
43 mova m0, [expq+offsetq]
47 PMINUB m0, [expq+offsetq], m1
58 %define PMINUB PMINUB_MMX
63 %define PMINUB PMINUB_MMXEXT
64 %define LOOP_ALIGN ALIGN 16
65 AC3_EXPONENT_MIN mmxext
74 ;-----------------------------------------------------------------------------
75 ; int ff_ac3_max_msb_abs_int16(const int16_t *src, int len)
77 ; This function uses 2 different methods to calculate a valid result.
78 ; 1) logical 'or' of abs of each element
79 ; This is used for ssse3 because of the pabsw instruction.
80 ; It is also used for mmx because of the lack of min/max instructions.
81 ; 2) calculate min/max for the array, then or(abs(min),abs(max))
82 ; This is used for mmxext and sse2 because they have pminsw/pmaxsw.
83 ;-----------------------------------------------------------------------------
85 %macro AC3_MAX_MSB_ABS_INT16 2
86 cglobal ac3_max_msb_abs_int16_%1, 2,2,5, src, len
92 mova m1, [srcq+mmsize]
100 mova m1, [srcq+mmsize]
103 ; using memory args is faster for ssse3
105 pabsw m1, [srcq+mmsize]
131 %define ABS2 ABS2_MMX
132 %define PSHUFLW pshufw
133 AC3_MAX_MSB_ABS_INT16 mmx, or_abs
134 %define ABS2 ABS2_MMX2
135 AC3_MAX_MSB_ABS_INT16 mmxext, min_max
137 %define PSHUFLW pshuflw
138 AC3_MAX_MSB_ABS_INT16 sse2, min_max
139 %define ABS2 ABS2_SSSE3
140 AC3_MAX_MSB_ABS_INT16 ssse3, or_abs
142 ;-----------------------------------------------------------------------------
143 ; macro used for ff_ac3_lshift_int16() and ff_ac3_rshift_int32()
144 ;-----------------------------------------------------------------------------
146 %macro AC3_SHIFT 4 ; l/r, 16/32, shift instruction, instruction set
147 cglobal ac3_%1shift_int%2_%4, 3,3,5, src, len, shift
151 mova m2, [srcq+mmsize ]
152 mova m3, [srcq+mmsize*2]
153 mova m4, [srcq+mmsize*3]
159 mova [srcq+mmsize ], m2
160 mova [srcq+mmsize*2], m3
161 mova [srcq+mmsize*3], m4
163 sub lend, mmsize*32/%2
169 ;-----------------------------------------------------------------------------
170 ; void ff_ac3_lshift_int16(int16_t *src, unsigned int len, unsigned int shift)
171 ;-----------------------------------------------------------------------------
174 AC3_SHIFT l, 16, psllw, mmx
176 AC3_SHIFT l, 16, psllw, sse2
178 ;-----------------------------------------------------------------------------
179 ; void ff_ac3_rshift_int32(int32_t *src, unsigned int len, unsigned int shift)
180 ;-----------------------------------------------------------------------------
183 AC3_SHIFT r, 32, psrad, mmx
185 AC3_SHIFT r, 32, psrad, sse2
187 ;-----------------------------------------------------------------------------
188 ; void ff_float_to_fixed24(int32_t *dst, const float *src, unsigned int len)
189 ;-----------------------------------------------------------------------------
191 ; The 3DNow! version is not bit-identical because pf2id uses truncation rather
192 ; than round-to-nearest.
194 cglobal float_to_fixed24_3dnow, 3,3,0, dst, src, len
220 cglobal float_to_fixed24_sse, 3,3,3, dst, src, len
244 cglobal float_to_fixed24_sse2, 3,3,9, dst, src, len
248 movaps m2, [srcq+16 ]
249 movaps m3, [srcq+32 ]
250 movaps m4, [srcq+48 ]
252 movaps m5, [srcq+64 ]
253 movaps m6, [srcq+80 ]
254 movaps m7, [srcq+96 ]
255 movaps m8, [srcq+112]
278 movdqa [dstq+16 ], m2
279 movdqa [dstq+32 ], m3
280 movdqa [dstq+48 ], m4
282 movdqa [dstq+64 ], m5
283 movdqa [dstq+80 ], m6
284 movdqa [dstq+96 ], m7
285 movdqa [dstq+112], m8