00001
00002
00003
00004
00005
00006
00007
00008
00009
00010
00011
00012
00013
00014
00015
00016
00017
00018
00019
00020
00021
00022 #include "dsputil_altivec.h"
00023 #include "util_altivec.h"
00024
00025 #define CONFIG_FLOAT 1
00026 #include "libavcodec/mpegaudio.h"
00027
00028 #define MACS(rt, ra, rb) rt+=(ra)*(rb)
00029 #define MLSS(rt, ra, rb) rt-=(ra)*(rb)
00030
00031 #define SUM8(op, sum, w, p) \
00032 { \
00033 op(sum, (w)[0 * 64], (p)[0 * 64]); \
00034 op(sum, (w)[1 * 64], (p)[1 * 64]); \
00035 op(sum, (w)[2 * 64], (p)[2 * 64]); \
00036 op(sum, (w)[3 * 64], (p)[3 * 64]); \
00037 op(sum, (w)[4 * 64], (p)[4 * 64]); \
00038 op(sum, (w)[5 * 64], (p)[5 * 64]); \
00039 op(sum, (w)[6 * 64], (p)[6 * 64]); \
00040 op(sum, (w)[7 * 64], (p)[7 * 64]); \
00041 }
00042
00043 static void apply_window(const float *buf, const float *win1,
00044 const float *win2, float *sum1, float *sum2, int len)
00045 {
00046 const vector float *win1a = (const vector float *) win1;
00047 const vector float *win2a = (const vector float *) win2;
00048 const vector float *bufa = (const vector float *) buf;
00049 vector float *sum1a = (vector float *) sum1;
00050 vector float *sum2a = (vector float *) sum2;
00051 vector float av_uninit(v0), av_uninit(v4);
00052 vector float v1, v2, v3;
00053
00054 len = len >> 2;
00055
00056 #define MULT(a, b) \
00057 { \
00058 v1 = vec_ld(a, win1a); \
00059 v2 = vec_ld(b, win2a); \
00060 v3 = vec_ld(a, bufa); \
00061 v0 = vec_madd(v3, v1, v0); \
00062 v4 = vec_madd(v2, v3, v4); \
00063 }
00064
00065 while (len--) {
00066 v0 = vec_xor(v0, v0);
00067 v4 = vec_xor(v4, v4);
00068
00069 MULT( 0, 0);
00070 MULT( 256, 64);
00071 MULT( 512, 128);
00072 MULT( 768, 192);
00073 MULT(1024, 256);
00074 MULT(1280, 320);
00075 MULT(1536, 384);
00076 MULT(1792, 448);
00077
00078 vec_st(v0, 0, sum1a);
00079 vec_st(v4, 0, sum2a);
00080 sum1a++;
00081 sum2a++;
00082 win1a++;
00083 win2a++;
00084 bufa++;
00085 }
00086 }
00087
00088 static void apply_window_mp3(float *in, float *win, int *unused, float *out,
00089 int incr)
00090 {
00091 LOCAL_ALIGNED_16(float, suma, [17]);
00092 LOCAL_ALIGNED_16(float, sumb, [17]);
00093 LOCAL_ALIGNED_16(float, sumc, [17]);
00094 LOCAL_ALIGNED_16(float, sumd, [17]);
00095
00096 float sum;
00097 int j;
00098 float *out2 = out + 32 * incr;
00099
00100
00101 memcpy(in + 512, in, 32 * sizeof(*in));
00102
00103 apply_window(in + 16, win , win + 512, suma, sumc, 16);
00104 apply_window(in + 32, win + 48, win + 640, sumb, sumd, 16);
00105
00106 SUM8(MLSS, suma[0], win + 32, in + 48);
00107
00108 sumc[ 0] = 0;
00109 sumb[16] = 0;
00110 sumd[16] = 0;
00111
00112 out[0 ] = suma[ 0];
00113 out += incr;
00114 out2 -= incr;
00115 for(j=1;j<16;j++) {
00116 *out = suma[ j] - sumd[16-j];
00117 *out2 = -sumb[16-j] - sumc[ j];
00118 out += incr;
00119 out2 -= incr;
00120 }
00121
00122 sum = 0;
00123 SUM8(MLSS, sum, win + 16 + 32, in + 32);
00124 *out = sum;
00125 }
00126
00127 void ff_mpegaudiodec_init_altivec(MPADecodeContext *s)
00128 {
00129 s->apply_window_mp3 = apply_window_mp3;
00130 }