1/* 2 * FFT/IFFT transforms 3 * AltiVec-enabled 4 * Copyright (c) 2009 Loren Merritt 5 * 6 * This file is part of FFmpeg. 7 * 8 * FFmpeg is free software; you can redistribute it and/or 9 * modify it under the terms of the GNU Lesser General Public 10 * License as published by the Free Software Foundation; either 11 * version 2.1 of the License, or (at your option) any later version. 12 * 13 * FFmpeg is distributed in the hope that it will be useful, 14 * but WITHOUT ANY WARRANTY; without even the implied warranty of 15 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU 16 * Lesser General Public License for more details. 17 * 18 * You should have received a copy of the GNU Lesser General Public 19 * License along with FFmpeg; if not, write to the Free Software 20 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA 21 */ 22 23#include "config.h" 24#include "libavutil/cpu.h" 25#include "libavutil/ppc/cpu.h" 26#include "libavutil/ppc/types_altivec.h" 27#include "libavutil/ppc/util_altivec.h" 28#include "libavcodec/fft.h" 29 30/** 31 * Do a complex FFT with the parameters defined in ff_fft_init(). 32 * The input data must be permuted before with s->revtab table. 33 * No 1.0 / sqrt(n) normalization is done. 34 * AltiVec-enabled: 35 * This code assumes that the 'z' pointer is 16 bytes-aligned. 36 * It also assumes all FFTComplex are 8 bytes-aligned pairs of floats. 37 */ 38 39#if HAVE_VSX 40#include "fft_vsx.h" 41#else 42void ff_fft_calc_altivec(FFTContext *s, FFTComplex *z); 43void ff_fft_calc_interleave_altivec(FFTContext *s, FFTComplex *z); 44#endif 45 46#if HAVE_GNU_AS && HAVE_ALTIVEC 47static void imdct_half_altivec(FFTContext *s, FFTSample *output, const FFTSample *input) 48{ 49 int j, k; 50 int n = 1 << s->mdct_bits; 51 int n4 = n >> 2; 52 int n8 = n >> 3; 53 int n32 = n >> 5; 54 const uint16_t *revtabj = s->revtab; 55 const uint16_t *revtabk = s->revtab+n4; 56 const vec_f *tcos = (const vec_f*)(s->tcos+n8); 57 const vec_f *tsin = (const vec_f*)(s->tsin+n8); 58 const vec_f *pin = (const vec_f*)(input+n4); 59 vec_f *pout = (vec_f*)(output+n4); 60 61 /* pre rotation */ 62 k = n32-1; 63 do { 64 vec_f cos,sin,cos0,sin0,cos1,sin1,re,im,r0,i0,r1,i1,a,b,c,d; 65#define CMULA(p,o0,o1,o2,o3)\ 66 a = pin[ k*2+p]; /* { z[k].re, z[k].im, z[k+1].re, z[k+1].im } */\ 67 b = pin[-k*2-p-1]; /* { z[-k-2].re, z[-k-2].im, z[-k-1].re, z[-k-1].im } */\ 68 re = vec_perm(a, b, vcprm(0,2,s0,s2)); /* { z[k].re, z[k+1].re, z[-k-2].re, z[-k-1].re } */\ 69 im = vec_perm(a, b, vcprm(s3,s1,3,1)); /* { z[-k-1].im, z[-k-2].im, z[k+1].im, z[k].im } */\ 70 cos = vec_perm(cos0, cos1, vcprm(o0,o1,s##o2,s##o3)); /* { cos[k], cos[k+1], cos[-k-2], cos[-k-1] } */\ 71 sin = vec_perm(sin0, sin1, vcprm(o0,o1,s##o2,s##o3));\ 72 r##p = im*cos - re*sin;\ 73 i##p = re*cos + im*sin; 74#define STORE2(v,dst)\ 75 j = dst;\ 76 vec_ste(v, 0, output+j*2);\ 77 vec_ste(v, 4, output+j*2); 78#define STORE8(p)\ 79 a = vec_perm(r##p, i##p, vcprm(0,s0,0,s0));\ 80 b = vec_perm(r##p, i##p, vcprm(1,s1,1,s1));\ 81 c = vec_perm(r##p, i##p, vcprm(2,s2,2,s2));\ 82 d = vec_perm(r##p, i##p, vcprm(3,s3,3,s3));\ 83 STORE2(a, revtabk[ p*2-4]);\ 84 STORE2(b, revtabk[ p*2-3]);\ 85 STORE2(c, revtabj[-p*2+2]);\ 86 STORE2(d, revtabj[-p*2+3]); 87 88 cos0 = tcos[k]; 89 sin0 = tsin[k]; 90 cos1 = tcos[-k-1]; 91 sin1 = tsin[-k-1]; 92 CMULA(0, 0,1,2,3); 93 CMULA(1, 2,3,0,1); 94 STORE8(0); 95 STORE8(1); 96 revtabj += 4; 97 revtabk -= 4; 98 k--; 99 } while(k >= 0); 100 101#if HAVE_VSX 102 ff_fft_calc_vsx(s, (FFTComplex*)output); 103#else 104 ff_fft_calc_altivec(s, (FFTComplex*)output); 105#endif 106 107 /* post rotation + reordering */ 108 j = -n32; 109 k = n32-1; 110 do { 111 vec_f cos,sin,re,im,a,b,c,d; 112#define CMULB(d0,d1,o)\ 113 re = pout[o*2];\ 114 im = pout[o*2+1];\ 115 cos = tcos[o];\ 116 sin = tsin[o];\ 117 d0 = im*sin - re*cos;\ 118 d1 = re*sin + im*cos; 119 120 CMULB(a,b,j); 121 CMULB(c,d,k); 122 pout[2*j] = vec_perm(a, d, vcprm(0,s3,1,s2)); 123 pout[2*j+1] = vec_perm(a, d, vcprm(2,s1,3,s0)); 124 pout[2*k] = vec_perm(c, b, vcprm(0,s3,1,s2)); 125 pout[2*k+1] = vec_perm(c, b, vcprm(2,s1,3,s0)); 126 j++; 127 k--; 128 } while(k >= 0); 129} 130 131static void imdct_calc_altivec(FFTContext *s, FFTSample *output, const FFTSample *input) 132{ 133 int k; 134 int n = 1 << s->mdct_bits; 135 int n4 = n >> 2; 136 int n16 = n >> 4; 137 vec_u32 sign = {1U<<31,1U<<31,1U<<31,1U<<31}; 138 vec_u32 *p0 = (vec_u32*)(output+n4); 139 vec_u32 *p1 = (vec_u32*)(output+n4*3); 140 141 imdct_half_altivec(s, output + n4, input); 142 143 for (k = 0; k < n16; k++) { 144 vec_u32 a = p0[k] ^ sign; 145 vec_u32 b = p1[-k-1]; 146 p0[-k-1] = vec_perm(a, a, vcprm(3,2,1,0)); 147 p1[k] = vec_perm(b, b, vcprm(3,2,1,0)); 148 } 149} 150#endif /* HAVE_GNU_AS && HAVE_ALTIVEC */ 151 152av_cold void ff_fft_init_ppc(FFTContext *s) 153{ 154#if HAVE_GNU_AS && HAVE_ALTIVEC 155 if (!PPC_ALTIVEC(av_get_cpu_flags())) 156 return; 157 158#if HAVE_VSX 159 s->fft_calc = ff_fft_calc_interleave_vsx; 160#else 161 s->fft_calc = ff_fft_calc_interleave_altivec; 162#endif 163 if (s->mdct_bits >= 5) { 164 s->imdct_calc = imdct_calc_altivec; 165 s->imdct_half = imdct_half_altivec; 166 } 167#endif /* HAVE_GNU_AS && HAVE_ALTIVEC */ 168} 169