1 /*****************************************************************************
2 * ac3_srfft_3dn.c: accelerated 3D Now! ac3 fft functions
3 *****************************************************************************
4 * Copyright (C) 1999, 2000, 2001 VideoLAN
5 * $Id: ac3_srfft_3dn.c,v 1.1 2001/05/16 14:51:29 reno 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 #define MODULE_NAME imdct3dn
25 #include "modules_inner.h"
27 /*****************************************************************************
29 * *****************************************************************************/
42 #include "ac3_imdct.h"
43 #include "ac3_srfft.h"
45 void hsqrt2_3dn (void);
47 static void fft_4_3dn (complex_t *x);
48 static void fft_8_3dn (complex_t *x);
49 static void fft_asmb_3dn (int k, complex_t *x, complex_t *wTB,
50 const complex_t *d, const complex_t *d_3);
52 void _M( fft_64p ) ( complex_t *a )
54 fft_8_3dn(&a[0]); fft_4_3dn(&a[8]); fft_4_3dn(&a[12]);
55 fft_asmb_3dn(2, &a[0], &a[8], &delta16[0], &delta16_3[0]);
57 fft_8_3dn(&a[16]), fft_8_3dn(&a[24]);
58 fft_asmb_3dn(4, &a[0], &a[16],&delta32[0], &delta32_3[0]);
60 fft_8_3dn(&a[32]); fft_4_3dn(&a[40]); fft_4_3dn(&a[44]);
61 fft_asmb_3dn(2, &a[32], &a[40], &delta16[0], &delta16_3[0]);
63 fft_8_3dn(&a[48]); fft_4_3dn(&a[56]); fft_4_3dn(&a[60]);
64 fft_asmb_3dn(2, &a[48], &a[56], &delta16[0], &delta16_3[0]);
66 fft_asmb_3dn(8, &a[0], &a[32],&delta64[0], &delta64_3[0]);
69 void _M( fft_128p ) ( complex_t *a )
71 fft_8_3dn(&a[0]); fft_4_3dn(&a[8]); fft_4_3dn(&a[12]);
72 fft_asmb_3dn(2, &a[0], &a[8], &delta16[0], &delta16_3[0]);
74 fft_8_3dn(&a[16]), fft_8_3dn(&a[24]);
75 fft_asmb_3dn(4, &a[0], &a[16],&delta32[0], &delta32_3[0]);
77 fft_8_3dn(&a[32]); fft_4_3dn(&a[40]); fft_4_3dn(&a[44]);
78 fft_asmb_3dn(2, &a[32], &a[40], &delta16[0], &delta16_3[0]);
80 fft_8_3dn(&a[48]); fft_4_3dn(&a[56]); fft_4_3dn(&a[60]);
81 fft_asmb_3dn(2, &a[48], &a[56], &delta16[0], &delta16_3[0]);
83 fft_asmb_3dn(8, &a[0], &a[32],&delta64[0], &delta64_3[0]);
85 fft_8_3dn(&a[64]); fft_4_3dn(&a[72]); fft_4_3dn(&a[76]);
87 fft_asmb_3dn(2, &a[64], &a[72], &delta16[0], &delta16_3[0]);
89 fft_8_3dn(&a[80]); fft_8_3dn(&a[88]);
92 fft_asmb_3dn(4, &a[64], &a[80],&delta32[0], &delta32_3[0]);
94 fft_8_3dn(&a[96]); fft_4_3dn(&a[104]), fft_4_3dn(&a[108]);
96 fft_asmb_3dn(2, &a[96], &a[104], &delta16[0], &delta16_3[0]);
98 fft_8_3dn(&a[112]), fft_8_3dn(&a[120]);
100 fft_asmb_3dn(4, &a[96], &a[112], &delta32[0], &delta32_3[0]);
102 /* fft_128(&a[0]); */
103 fft_asmb_3dn(16, &a[0], &a[64], &delta128[0], &delta128_3[0]);
106 void hsqrt2_3dn (void)
109 ".float 0f0.707106781188\n"
110 ".float 0f0.707106781188\n"
111 ".float 0f-0.707106781188\n"
112 ".float 0f-0.707106781188\n"
126 static void fft_4_3dn (complex_t *x)
128 __asm__ __volatile__ (
129 "movq (%%eax), %%mm0\n" /* x[0] */
130 "movq 8(%%eax), %%mm1\n" /* x[1] */
131 "movq 16(%%eax), %%mm2\n" /* x[2] */
132 "movq 24(%%eax), %%mm3\n" /* x[3] */
133 "movq %%mm0, %%mm4\n" /* x[1] */
134 "movq %%mm1, %%mm5\n" /* x[1] */
135 "movq %%mm0, %%mm6\n" /* x[0] */
136 "pfadd %%mm2, %%mm0\n" /* x[0] + x[2] */
137 "pfadd %%mm3, %%mm1\n" /* x[1] + x[3] */
138 "pfsub %%mm2, %%mm4\n" /* x[0] - x[2] */
139 "pfsub %%mm3, %%mm5\n" /* x[1] - x[3] */
141 "pfadd %%mm1, %%mm0\n" /* x[0] + x[2] + x[1] + x[3] */
142 "pfsub %%mm1, %%mm6\n" /* x[0] + x[2] - x[1] - x[3] */
144 "movq %%mm0, (%%eax)\n"
145 "movq %%mm6, 16(%%eax)\n"
147 "pxor %%mm6, %%mm6\n"
148 "movq %%mm5, %%mm2\n" /* x[1] - x[3] */
149 "movq %%mm4, %%mm3\n" /* x[0] - x[2] */
150 "pfsub %%mm5, %%mm6\n" /* x[3] - x[1] */
152 "punpckhdq %%mm2,%%mm2\n" /* x[1] - x[3].im */
153 "punpckldq %%mm6,%%mm6\n" /* x[3] - x[1].re */
154 "punpckhdq %%mm6,%%mm2\n" /* x[3] - x[1].re, x[1] - x[3].im */
156 "pfsub %%mm2, %%mm4\n" /* x0i-x2i-x3r+x1.r,x0r-x2r-x1i+x3i */
157 "pfadd %%mm3, %%mm2\n" /* x0i-x2i+x3r-x1.r, x0r-x2r+x1i-x3.i */
159 "movq %%mm2, 8(%%eax)\n" /* mm4_2 + mm6_1, mm4_1 + mm5_2 */
160 "movq %%mm4, 24(%%eax)\n" /* mm4_2 - mm6_1, mm4_1 - mm5_2 */
166 static void fft_8_3dn (complex_t *x)
168 register float wT1_r, wT1_i, wB1_r, wB1_i, wT2_r, wT2_i, wB2_r, wB2_i;
179 register float yt_r, yt_i, yb_r, yb_i, u_r, u_i, vi_r, vi_i;
182 yb_r = yt_r - x[2].real;
186 vi_i = x[3].real - u_r;
190 vi_r = u_i - x[3].imag;
205 yb_i = yt_i - x[2].imag;
227 wT2_i = x[0].real - wT2_i;
237 wT2_i = x[0].imag - wT2_i;
248 wT2_i = x[2].real - wT2_i;
258 wT2_i = x[2].imag - wT2_i;
279 wT2_r = x[1].real - wT2_r;
287 wT2_i = x[1].imag - wT2_i;
295 wB1_r = wB2_i - x[7].imag;
296 wB1_i = wB2_r + x[7].real;
299 wB1_r = wT1_r + wT1_i;
304 wB2_i = wB2_r + wT1_i;
309 wB2_r = wB2_i + wB1_r;
316 static void fft_asmb_3dn (int k, complex_t *x, complex_t *wTB,
317 const complex_t *d, const complex_t *d_3)
319 register complex_t *x2k, *x3k, *x4k, *wB;
320 register float a_r, a_i, a1_r, a1_i, u_r, u_i, v_r, v_i;
327 TRANSZERO(x[0],x2k[0],x3k[0],x4k[0]);
328 TRANS(x[1],x2k[1],x3k[1],x4k[1],wTB[1],wB[1],d[1],d_3[1]);
332 TRANS(x[2],x2k[2],x3k[2],x4k[2],wTB[2],wB[2],d[2],d_3[2]);
333 TRANS(x[3],x2k[3],x3k[3],x4k[3],wTB[3],wB[3],d[3],d_3[3]);