4 * Copyright (c) 2003 Romain Dolbeau <romain@dolbeau.org>
5 * Based on code Copyright (c) 2002 Fabrice Bellard.
7 * This file is part of FFmpeg.
9 * FFmpeg is free software; you can redistribute it and/or
10 * modify it under the terms of the GNU Lesser General Public
11 * License as published by the Free Software Foundation; either
12 * version 2.1 of the License, or (at your option) any later version.
14 * FFmpeg 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 GNU
17 * Lesser General Public License for more details.
19 * You should have received a copy of the GNU Lesser General Public
20 * License along with FFmpeg; if not, write to the Free Software
21 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
23 #include "libavcodec/dsputil.h"
25 #include "gcc_fixes.h"
27 #include "dsputil_ppc.h"
28 #include "util_altivec.h"
30 * Do a complex FFT with the parameters defined in ff_fft_init(). The
31 * input data must be permuted before with s->revtab table. No
32 * 1.0/sqrt(n) normalization is done.
34 * This code assumes that the 'z' pointer is 16 bytes-aligned
35 * It also assumes all FFTComplex are 8 bytes-aligned pair of float
36 * The code is exactly the same as the SSE version, except
37 * that successive MUL + ADD/SUB have been merged into
38 * fused multiply-add ('vec_madd' in altivec)
40 void ff_fft_calc_altivec(FFTContext *s, FFTComplex *z)
42 POWERPC_PERF_DECLARE(altivec_fft_num, s->nbits >= 6);
43 register const vector float vczero = (const vector float)vec_splat_u32(0.);
48 register FFTComplex *p, *q;
49 FFTComplex *cptr, *cptr1;
52 POWERPC_PERF_START_COUNT(altivec_fft_num, s->nbits >= 6);
57 vector float *r, a, b, a1, c1, c2;
59 r = (vector float *)&z[0];
72 a1 = vec_ld(sizeof(vector float), r);
74 b = vec_perm(a,a,vcprmle(1,0,3,2));
76 /* do the pass 0 butterfly */
78 b = vec_perm(a1,a1,vcprmle(1,0,3,2));
79 b = vec_madd(a1,c1,b);
80 /* do the pass 0 butterfly */
82 /* multiply third by -i */
83 b = vec_perm(b,b,vcprmle(2,3,1,0));
85 /* do the pass 1 butterfly */
86 vec_st(vec_madd(b,c2,a), 0, r);
87 vec_st(vec_nmsub(b,c2,a), sizeof(vector float), r);
107 vector float a,b,c,t1;
109 a = vec_ld(0, (float*)p);
110 b = vec_ld(0, (float*)q);
113 c = vec_ld(0, (float*)cptr);
115 t1 = vec_madd(c, vec_perm(b,b,vcprmle(2,2,0,0)),vczero);
116 c = vec_ld(sizeof(vector float), (float*)cptr);
118 b = vec_madd(c, vec_perm(b,b,vcprmle(3,3,1,1)),t1);
121 vec_st(vec_add(a,b), 0, (float*)p);
122 vec_st(vec_sub(a,b), 0, (float*)q);
133 nblocks = nblocks >> 1;
134 nloops = nloops << 1;
135 } while (nblocks != 0);
137 POWERPC_PERF_STOP_COUNT(altivec_fft_num, s->nbits >= 6);