1 ;*****************************************************************************
2 ;* x86-optimized AC-3 DSP functions
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 ;******************************************************************************
22 %include "libavutil/x86/x86util.asm"
26 ; 16777216.0f - used in ff_float_to_fixed24()
27 pf_1_24: times 4 dd 0x4B800000
29 ; used in ff_ac3_compute_mantissa_size()
31 pw_bap_mul1: dw 21846, 21846, 0, 32768, 21846, 21846, 0, 32768
32 pw_bap_mul2: dw 5, 7, 0, 7, 5, 7, 0, 7
34 ; used in ff_ac3_extract_exponents()
36 pd_151: times 4 dd 151
40 ;-----------------------------------------------------------------------------
41 ; void ff_ac3_exponent_min(uint8_t *exp, int num_reuse_blocks, int nb_coefs)
42 ;-----------------------------------------------------------------------------
44 %macro AC3_EXPONENT_MIN 0
45 cglobal ac3_exponent_min, 3, 4, 2, exp, reuse_blks, expn, offset
50 mov offsetq, reuse_blksq
51 mova m0, [expq+offsetq]
55 PMINUB m0, [expq+offsetq], m1
69 %if HAVE_MMXEXT_EXTERNAL
70 %define LOOP_ALIGN ALIGN 16
74 %if HAVE_SSE2_EXTERNAL
80 ;-----------------------------------------------------------------------------
81 ; void ff_float_to_fixed24(int32_t *dst, const float *src, unsigned int len)
82 ;-----------------------------------------------------------------------------
84 ; The 3DNow! version is not bit-identical because pf2id uses truncation rather
85 ; than round-to-nearest.
87 cglobal float_to_fixed24, 3, 3, 0, dst, src, len
114 cglobal float_to_fixed24, 3, 3, 3, dst, src, len
139 cglobal float_to_fixed24, 3, 3, 9, dst, src, len
143 movaps m2, [srcq+16 ]
144 movaps m3, [srcq+32 ]
145 movaps m4, [srcq+48 ]
147 movaps m5, [srcq+64 ]
148 movaps m6, [srcq+80 ]
149 movaps m7, [srcq+96 ]
150 movaps m8, [srcq+112]
173 movdqa [dstq+16 ], m2
174 movdqa [dstq+32 ], m3
175 movdqa [dstq+48 ], m4
177 movdqa [dstq+64 ], m5
178 movdqa [dstq+80 ], m6
179 movdqa [dstq+96 ], m7
180 movdqa [dstq+112], m8
192 ;------------------------------------------------------------------------------
193 ; int ff_ac3_compute_mantissa_size(uint16_t mant_cnt[6][16])
194 ;------------------------------------------------------------------------------
196 %macro PHADDD4 2 ; xmm src, xmm tmp
204 cglobal ac3_compute_mantissa_size, 1, 2, 4, mant_cnt, sum
205 movdqa m0, [mant_cntq ]
206 movdqa m1, [mant_cntq+ 1*16]
207 paddw m0, [mant_cntq+ 2*16]
208 paddw m1, [mant_cntq+ 3*16]
209 paddw m0, [mant_cntq+ 4*16]
210 paddw m1, [mant_cntq+ 5*16]
211 paddw m0, [mant_cntq+ 6*16]
212 paddw m1, [mant_cntq+ 7*16]
213 paddw m0, [mant_cntq+ 8*16]
214 paddw m1, [mant_cntq+ 9*16]
215 paddw m0, [mant_cntq+10*16]
216 paddw m1, [mant_cntq+11*16]
217 pmaddwd m0, [ac3_bap_bits ]
218 pmaddwd m1, [ac3_bap_bits+16]
222 movdqa m3, [pw_bap_mul1]
223 movhpd m0, [mant_cntq +2]
224 movlpd m0, [mant_cntq+1*32+2]
225 movhpd m1, [mant_cntq+2*32+2]
226 movlpd m1, [mant_cntq+3*32+2]
227 movhpd m2, [mant_cntq+4*32+2]
228 movlpd m2, [mant_cntq+5*32+2]
234 pmaddwd m0, [pw_bap_mul2]
240 ;------------------------------------------------------------------------------
241 ; void ff_ac3_extract_exponents(uint8_t *exp, int32_t *coef, int nb_coefs)
242 ;------------------------------------------------------------------------------
244 %macro PABSD 1-2 ; src/dst, unused
255 %macro AC3_EXTRACT_EXPONENTS 0
256 cglobal ac3_extract_exponents, 3, 3, 4, exp, coef, len
258 lea coefq, [coefq+4*lenq]
263 ; move 4 32-bit coefs to xmm0
264 mova m0, [coefq+4*lenq]
267 ; convert to float and extract exponents
274 ; move the lowest byte in each of 4 dwords to the low dword
275 ; NOTE: We cannot just extract the low bytes with pshufb because the dword
276 ; result for 16777215 is -1 due to float inaccuracy. Using packuswb
277 ; clips this to 0, which is the correct exponent.
287 %if HAVE_SSE2_EXTERNAL
289 AC3_EXTRACT_EXPONENTS
291 %if HAVE_SSSE3_EXTERNAL
293 AC3_EXTRACT_EXPONENTS