1 /*****************************************************************************
2 * ac3_srfft_3dn.c: accelerated 3D Now! ac3 fft functions
3 *****************************************************************************
4 * Copyright (C) 1999-2001 VideoLAN
5 * $Id: ac3_srfft_3dn.c,v 1.7 2002/07/31 20:56:51 sam Exp $
7 * Authors: Renaud Dartus <reno@videolan.org>
9 * This program is free software; you can redistribute it and/or modify
10 * it under the terms of the GNU General Public License as published by
11 * the Free Software Foundation; either version 2 of the License, or
12 * (at your option) any later version.
14 * This program is distributed in the hope that it will be useful,
15 * but WITHOUT ANY WARRANTY; without even the implied warranty of
16 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
17 * GNU General Public License for more details.
19 * You should have received a copy of the GNU General Public License
20 * along with this program; if not, write to the Free Software
21 * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111, USA.
22 *****************************************************************************/
24 /*****************************************************************************
26 *****************************************************************************/
32 #include "ac3_imdct.h"
33 #include "ac3_srfft.h"
35 void hsqrt2_3dn (void);
37 static void fft_4_3dn (complex_t *x);
38 static void fft_8_3dn (complex_t *x);
39 static void fft_asmb_3dn (int k, complex_t *x, complex_t *wTB,
40 const complex_t *d, const complex_t *d_3);
42 void E_( fft_64p ) ( complex_t *a )
44 fft_8_3dn(&a[0]); fft_4_3dn(&a[8]); fft_4_3dn(&a[12]);
45 fft_asmb_3dn(2, &a[0], &a[8], &delta16[0], &delta16_3[0]);
47 fft_8_3dn(&a[16]), fft_8_3dn(&a[24]);
48 fft_asmb_3dn(4, &a[0], &a[16],&delta32[0], &delta32_3[0]);
50 fft_8_3dn(&a[32]); fft_4_3dn(&a[40]); fft_4_3dn(&a[44]);
51 fft_asmb_3dn(2, &a[32], &a[40], &delta16[0], &delta16_3[0]);
53 fft_8_3dn(&a[48]); fft_4_3dn(&a[56]); fft_4_3dn(&a[60]);
54 fft_asmb_3dn(2, &a[48], &a[56], &delta16[0], &delta16_3[0]);
56 fft_asmb_3dn(8, &a[0], &a[32],&delta64[0], &delta64_3[0]);
59 void E_( fft_128p ) ( complex_t *a )
61 fft_8_3dn(&a[0]); fft_4_3dn(&a[8]); fft_4_3dn(&a[12]);
62 fft_asmb_3dn(2, &a[0], &a[8], &delta16[0], &delta16_3[0]);
64 fft_8_3dn(&a[16]), fft_8_3dn(&a[24]);
65 fft_asmb_3dn(4, &a[0], &a[16],&delta32[0], &delta32_3[0]);
67 fft_8_3dn(&a[32]); fft_4_3dn(&a[40]); fft_4_3dn(&a[44]);
68 fft_asmb_3dn(2, &a[32], &a[40], &delta16[0], &delta16_3[0]);
70 fft_8_3dn(&a[48]); fft_4_3dn(&a[56]); fft_4_3dn(&a[60]);
71 fft_asmb_3dn(2, &a[48], &a[56], &delta16[0], &delta16_3[0]);
73 fft_asmb_3dn(8, &a[0], &a[32],&delta64[0], &delta64_3[0]);
75 fft_8_3dn(&a[64]); fft_4_3dn(&a[72]); fft_4_3dn(&a[76]);
77 fft_asmb_3dn(2, &a[64], &a[72], &delta16[0], &delta16_3[0]);
79 fft_8_3dn(&a[80]); fft_8_3dn(&a[88]);
82 fft_asmb_3dn(4, &a[64], &a[80],&delta32[0], &delta32_3[0]);
84 fft_8_3dn(&a[96]); fft_4_3dn(&a[104]), fft_4_3dn(&a[108]);
86 fft_asmb_3dn(2, &a[96], &a[104], &delta16[0], &delta16_3[0]);
88 fft_8_3dn(&a[112]), fft_8_3dn(&a[120]);
90 fft_asmb_3dn(4, &a[96], &a[112], &delta32[0], &delta32_3[0]);
93 fft_asmb_3dn(16, &a[0], &a[64], &delta128[0], &delta128_3[0]);
96 void hsqrt2_3dn (void)
99 ".float 0f0.707106781188\n"
100 ".float 0f0.707106781188\n"
101 ".float 0f-0.707106781188\n"
102 ".float 0f-0.707106781188\n"
116 static void fft_4_3dn (complex_t *x)
118 __asm__ __volatile__ (
120 "movq (%%eax), %%mm0\n" /* x[0] */
121 "movq 8(%%eax), %%mm1\n" /* x[1] */
122 "movq 16(%%eax), %%mm2\n" /* x[2] */
123 "movq 24(%%eax), %%mm3\n" /* x[3] */
124 "movq %%mm0, %%mm4\n" /* x[1] */
125 "movq %%mm1, %%mm5\n" /* x[1] */
126 "movq %%mm0, %%mm6\n" /* x[0] */
127 "pfadd %%mm2, %%mm0\n" /* x[0] + x[2] */
128 "pfadd %%mm3, %%mm1\n" /* x[1] + x[3] */
129 "pfsub %%mm2, %%mm4\n" /* x[0] - x[2] */
130 "pfsub %%mm3, %%mm5\n" /* x[1] - x[3] */
132 "pfadd %%mm1, %%mm0\n" /* x[0] + x[2] + x[1] + x[3] */
133 "pfsub %%mm1, %%mm6\n" /* x[0] + x[2] - x[1] - x[3] */
135 "movq %%mm0, (%%eax)\n"
136 "movq %%mm6, 16(%%eax)\n"
138 "pxor %%mm6, %%mm6\n"
139 "movq %%mm5, %%mm2\n" /* x[1] - x[3] */
140 "movq %%mm4, %%mm3\n" /* x[0] - x[2] */
141 "pfsub %%mm5, %%mm6\n" /* x[3] - x[1] */
143 "punpckhdq %%mm2,%%mm2\n" /* x[1] - x[3].im */
144 "punpckldq %%mm6,%%mm6\n" /* x[3] - x[1].re */
145 "punpckhdq %%mm6,%%mm2\n" /* x[3] - x[1].re, x[1] - x[3].im */
147 "pfsub %%mm2, %%mm4\n" /* x0i-x2i-x3r+x1.r,x0r-x2r-x1i+x3i */
148 "pfadd %%mm3, %%mm2\n" /* x0i-x2i+x3r-x1.r, x0r-x2r+x1i-x3.i */
150 "movq %%mm2, 8(%%eax)\n" /* mm4_2 + mm6_1, mm4_1 + mm5_2 */
151 "movq %%mm4, 24(%%eax)\n" /* mm4_2 - mm6_1, mm4_1 - mm5_2 */
157 static void fft_8_3dn (complex_t *x)
159 register float wT1_r, wT1_i, wB1_r, wB1_i, wT2_r, wT2_i, wB2_r, wB2_i;
170 register float yt_r, yt_i, yb_r, yb_i, u_r, u_i, vi_r, vi_i;
173 yb_r = yt_r - x[2].real;
177 vi_i = x[3].real - u_r;
181 vi_r = u_i - x[3].imag;
196 yb_i = yt_i - x[2].imag;
218 wT2_i = x[0].real - wT2_i;
228 wT2_i = x[0].imag - wT2_i;
239 wT2_i = x[2].real - wT2_i;
249 wT2_i = x[2].imag - wT2_i;
270 wT2_r = x[1].real - wT2_r;
278 wT2_i = x[1].imag - wT2_i;
286 wB1_r = wB2_i - x[7].imag;
287 wB1_i = wB2_r + x[7].real;
290 wB1_r = wT1_r + wT1_i;
295 wB2_i = wB2_r + wT1_i;
300 wB2_r = wB2_i + wB1_r;
307 static void fft_asmb_3dn (int k, complex_t *x, complex_t *wTB,
308 const complex_t *d, const complex_t *d_3)
310 register complex_t *x2k, *x3k, *x4k, *wB;
311 register float a_r, a_i, a1_r, a1_i, u_r, u_i, v_r, v_i;
318 TRANSZERO(x[0],x2k[0],x3k[0],x4k[0]);
319 TRANS(x[1],x2k[1],x3k[1],x4k[1],wTB[1],wB[1],d[1],d_3[1]);
323 TRANS(x[2],x2k[2],x3k[2],x4k[2],wTB[2],wB[2],d[2],d_3[2]);
324 TRANS(x[3],x2k[3],x3k[3],x4k[3],wTB[3],wB[3],d[3],d_3[3]);