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
66 %define LOOP_ALIGN ALIGN 16
67 %if HAVE_SSE2_EXTERNAL
73 ;-----------------------------------------------------------------------------
74 ; void ff_float_to_fixed24(int32_t *dst, const float *src, unsigned int len)
75 ;-----------------------------------------------------------------------------
78 cglobal float_to_fixed24, 3, 3, 9, dst, src, len
112 movdqa [dstq+16 ], m2
113 movdqa [dstq+32 ], m3
114 movdqa [dstq+48 ], m4
116 movdqa [dstq+64 ], m5
117 movdqa [dstq+80 ], m6
118 movdqa [dstq+96 ], m7
119 movdqa [dstq+112], m8
131 ;------------------------------------------------------------------------------
132 ; int ff_ac3_compute_mantissa_size(uint16_t mant_cnt[6][16])
133 ;------------------------------------------------------------------------------
135 %macro PHADDD4 2 ; xmm src, xmm tmp
143 cglobal ac3_compute_mantissa_size, 1, 2, 4, mant_cnt, sum
144 movdqa m0, [mant_cntq ]
145 movdqa m1, [mant_cntq+ 1*16]
146 paddw m0, [mant_cntq+ 2*16]
147 paddw m1, [mant_cntq+ 3*16]
148 paddw m0, [mant_cntq+ 4*16]
149 paddw m1, [mant_cntq+ 5*16]
150 paddw m0, [mant_cntq+ 6*16]
151 paddw m1, [mant_cntq+ 7*16]
152 paddw m0, [mant_cntq+ 8*16]
153 paddw m1, [mant_cntq+ 9*16]
154 paddw m0, [mant_cntq+10*16]
155 paddw m1, [mant_cntq+11*16]
156 pmaddwd m0, [ac3_bap_bits ]
157 pmaddwd m1, [ac3_bap_bits+16]
161 movdqa m3, [pw_bap_mul1]
162 movhpd m0, [mant_cntq +2]
163 movlpd m0, [mant_cntq+1*32+2]
164 movhpd m1, [mant_cntq+2*32+2]
165 movlpd m1, [mant_cntq+3*32+2]
166 movhpd m2, [mant_cntq+4*32+2]
167 movlpd m2, [mant_cntq+5*32+2]
173 pmaddwd m0, [pw_bap_mul2]
179 ;------------------------------------------------------------------------------
180 ; void ff_ac3_extract_exponents(uint8_t *exp, int32_t *coef, int nb_coefs)
181 ;------------------------------------------------------------------------------
183 %macro PABSD 1-2 ; src/dst, unused
194 %macro AC3_EXTRACT_EXPONENTS 0
195 cglobal ac3_extract_exponents, 3, 3, 4, exp, coef, len
197 lea coefq, [coefq+4*lenq]
202 ; move 4 32-bit coefs to xmm0
203 mova m0, [coefq+4*lenq]
206 ; convert to float and extract exponents
213 ; move the lowest byte in each of 4 dwords to the low dword
214 ; NOTE: We cannot just extract the low bytes with pshufb because the dword
215 ; result for 16777215 is -1 due to float inaccuracy. Using packuswb
216 ; clips this to 0, which is the correct exponent.
226 %if HAVE_SSE2_EXTERNAL
228 AC3_EXTRACT_EXPONENTS
230 %if HAVE_SSSE3_EXTERNAL
232 AC3_EXTRACT_EXPONENTS