1 ;******************************************************************************
2 ;* AAC Spectral Band Replication decoding functions
3 ;* Copyright (C) 2012 Christophe Gisquet <christophe.gisquet@gmail.com>
5 ;* This file is part of Libav.
7 ;* Libav 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 ;* Libav 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 Libav; 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"
25 ; mask equivalent for multiply by -1.0 1.0
26 ps_mask times 2 dd 1<<31, 0
27 ps_mask2 times 2 dd 0, 1<<31
28 ps_neg times 4 dd 1<<31
33 cglobal sbr_sum_square, 2, 3, 6
59 ; len is a multiple of 2, thus there are at least 4 elements to process
81 cglobal sbr_hf_g_filt, 5, 6, 5
82 lea r1, [r1 + 8*r4] ; offset by ixh elements into X_high
90 movlps m0, [r2 + 4*r3 + 0]
91 movlps m1, [r2 + 4*r3 + 8]
92 movlps m2, [r1 + 0*STEP]
93 movlps m3, [r1 + 2*STEP]
94 movhps m2, [r1 + 1*STEP]
95 movhps m3, [r1 + 3*STEP]
100 movu [r0 + 8*r3 + 0], m0
101 movu [r0 + 8*r3 + 16], m1
105 and r5, 3 ; number of single element loops
107 .loop1: ; element 0 and 1 can be computed at the same time
121 ; void ff_sbr_hf_gen_sse(float (*X_high)[2], const float (*X_low)[2],
122 ; const float alpha0[2], const float alpha1[2],
123 ; float bw, int start, int end)
125 cglobal sbr_hf_gen, 4,4,8, X_high, X_low, alpha0, alpha1, BW, S, E
128 %if ARCH_X86_64 == 0 || WIN64
134 mulps m2, bw ; (a1[0] a1[1])*bw
135 mulps m1, bw ; (a0[0] a0[1])*bw = (a2 a3)
136 mulps m2, bw ; (a1[0] a1[1])*bw*bw = (a0 a1)
142 %if ARCH_X86_64 == 0 || WIN64
143 ; start and end 6th and 7th args on stack
149 ; BW does not actually occupy a register, so shift by 1
153 sub start, end ; neg num of loops
154 lea X_highq, [X_highq + end*2*4]
155 lea X_lowq, [X_lowq + end*2*4 - 2*2*4]
156 shl start, 3 ; offset from num loops
158 mova m0, [X_lowq + start]
159 movlhps m1, m1 ; (a2 a3 a2 a3)
160 movlhps m2, m2 ; (a0 a1 a0 a1)
161 shufps m3, m3, q0101 ; (a3 a2 a3 a2)
162 shufps m4, m4, q0101 ; (a1 a0 a1 a0)
163 xorps m3, m7 ; (-a3 a2 -a3 a2)
164 xorps m4, m7 ; (-a1 a0 -a1 a0)
168 shufps m0, m0, q2200 ; {Xl[-2][0],",Xl[-1][0],"}
169 shufps m5, m5, q3311 ; {Xl[-2][1],",Xl[-1][1],"}
174 mova m0, [X_lowq + start + 2*2*4]
175 shufps m6, m0, q0022 ; {Xl[-1][0],",Xl[0][0],"}
176 shufps m7, m0, q1133 ; {Xl[-1][1],",Xl[1][1],"}
182 mova [X_highq + start], m5
187 cglobal sbr_sum64x5, 1,2,4,z
210 cglobal sbr_qmf_post_shuffle, 2,3,4,W,z
211 lea r2q, [zq + (64-4)*4]
216 shufps m0, m0, m0, q0123
229 cglobal sbr_neg_odd_64, 1,2,4,z
250 ; void ff_sbr_qmf_deint_bfly_sse2(float *v, const float *src0, const float *src1)
251 cglobal sbr_qmf_deint_bfly, 3,5,8, v,src0,src1,vrev,c
252 mov cq, 64*4-2*mmsize
253 lea vrevq, [vq + 64*4]
257 mova m2, [src0q+cq+mmsize]
258 mova m3, [src1q+mmsize]
268 mova [vrevq+mmsize], m3
270 mova [vq+cq+mmsize], m2
278 cglobal sbr_qmf_pre_shuffle, 1,4,6,z
279 %define OFFSET (32*4-2*mmsize)
281 lea r1q, [zq + (32+1)*4]
286 movu m2, [r1q + mmsize]
287 movu m1, [zq + r3q + 4 + mmsize]
288 movu m3, [zq + r3q + 4]
294 SBUTTERFLY dq, 2, 3, 4
295 SBUTTERFLY dq, 0, 1, 4
296 mova [r2q + 2*r3q + 0*mmsize], m2
297 mova [r2q + 2*r3q + 1*mmsize], m3
298 mova [r2q + 2*r3q + 2*mmsize], m0
299 mova [r2q + 2*r3q + 3*mmsize], m1