Libav 0.7.1
|
00001 /* 00002 * Copyright (c) 2006 Luca Barbato <lu_zero@gentoo.org> 00003 * 00004 * This file is part of Libav. 00005 * 00006 * Libav is free software; you can redistribute it and/or 00007 * modify it under the terms of the GNU Lesser General Public 00008 * License as published by the Free Software Foundation; either 00009 * version 2.1 of the License, or (at your option) any later version. 00010 * 00011 * Libav is distributed in the hope that it will be useful, 00012 * but WITHOUT ANY WARRANTY; without even the implied warranty of 00013 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU 00014 * Lesser General Public License for more details. 00015 * 00016 * You should have received a copy of the GNU Lesser General Public 00017 * License along with Libav; if not, write to the Free Software 00018 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA 00019 */ 00020 00021 #include "libavcodec/fmtconvert.h" 00022 00023 #include "dsputil_altivec.h" 00024 #include "util_altivec.h" 00025 00026 static void int32_to_float_fmul_scalar_altivec(float *dst, const int *src, float mul, int len) 00027 { 00028 union { 00029 vector float v; 00030 float s[4]; 00031 } mul_u; 00032 int i; 00033 vector float src1, src2, dst1, dst2, mul_v, zero; 00034 00035 zero = (vector float)vec_splat_u32(0); 00036 mul_u.s[0] = mul; 00037 mul_v = vec_splat(mul_u.v, 0); 00038 00039 for(i=0; i<len; i+=8) { 00040 src1 = vec_ctf(vec_ld(0, src+i), 0); 00041 src2 = vec_ctf(vec_ld(16, src+i), 0); 00042 dst1 = vec_madd(src1, mul_v, zero); 00043 dst2 = vec_madd(src2, mul_v, zero); 00044 vec_st(dst1, 0, dst+i); 00045 vec_st(dst2, 16, dst+i); 00046 } 00047 } 00048 00049 00050 static vector signed short 00051 float_to_int16_one_altivec(const float *src) 00052 { 00053 vector float s0 = vec_ld(0, src); 00054 vector float s1 = vec_ld(16, src); 00055 vector signed int t0 = vec_cts(s0, 0); 00056 vector signed int t1 = vec_cts(s1, 0); 00057 return vec_packs(t0,t1); 00058 } 00059 00060 static void float_to_int16_altivec(int16_t *dst, const float *src, long len) 00061 { 00062 int i; 00063 vector signed short d0, d1, d; 00064 vector unsigned char align; 00065 if(((long)dst)&15) //FIXME 00066 for(i=0; i<len-7; i+=8) { 00067 d0 = vec_ld(0, dst+i); 00068 d = float_to_int16_one_altivec(src+i); 00069 d1 = vec_ld(15, dst+i); 00070 d1 = vec_perm(d1, d0, vec_lvsl(0,dst+i)); 00071 align = vec_lvsr(0, dst+i); 00072 d0 = vec_perm(d1, d, align); 00073 d1 = vec_perm(d, d1, align); 00074 vec_st(d0, 0, dst+i); 00075 vec_st(d1,15, dst+i); 00076 } 00077 else 00078 for(i=0; i<len-7; i+=8) { 00079 d = float_to_int16_one_altivec(src+i); 00080 vec_st(d, 0, dst+i); 00081 } 00082 } 00083 00084 static void 00085 float_to_int16_interleave_altivec(int16_t *dst, const float **src, 00086 long len, int channels) 00087 { 00088 int i; 00089 vector signed short d0, d1, d2, c0, c1, t0, t1; 00090 vector unsigned char align; 00091 if(channels == 1) 00092 float_to_int16_altivec(dst, src[0], len); 00093 else 00094 if (channels == 2) { 00095 if(((long)dst)&15) 00096 for(i=0; i<len-7; i+=8) { 00097 d0 = vec_ld(0, dst + i); 00098 t0 = float_to_int16_one_altivec(src[0] + i); 00099 d1 = vec_ld(31, dst + i); 00100 t1 = float_to_int16_one_altivec(src[1] + i); 00101 c0 = vec_mergeh(t0, t1); 00102 c1 = vec_mergel(t0, t1); 00103 d2 = vec_perm(d1, d0, vec_lvsl(0, dst + i)); 00104 align = vec_lvsr(0, dst + i); 00105 d0 = vec_perm(d2, c0, align); 00106 d1 = vec_perm(c0, c1, align); 00107 vec_st(d0, 0, dst + i); 00108 d0 = vec_perm(c1, d2, align); 00109 vec_st(d1, 15, dst + i); 00110 vec_st(d0, 31, dst + i); 00111 dst+=8; 00112 } 00113 else 00114 for(i=0; i<len-7; i+=8) { 00115 t0 = float_to_int16_one_altivec(src[0] + i); 00116 t1 = float_to_int16_one_altivec(src[1] + i); 00117 d0 = vec_mergeh(t0, t1); 00118 d1 = vec_mergel(t0, t1); 00119 vec_st(d0, 0, dst + i); 00120 vec_st(d1, 16, dst + i); 00121 dst+=8; 00122 } 00123 } else { 00124 DECLARE_ALIGNED(16, int16_t, tmp)[len]; 00125 int c, j; 00126 for (c = 0; c < channels; c++) { 00127 float_to_int16_altivec(tmp, src[c], len); 00128 for (i = 0, j = c; i < len; i++, j+=channels) { 00129 dst[j] = tmp[i]; 00130 } 00131 } 00132 } 00133 } 00134 00135 void ff_fmt_convert_init_altivec(FmtConvertContext *c, AVCodecContext *avctx) 00136 { 00137 c->int32_to_float_fmul_scalar = int32_to_float_fmul_scalar_altivec; 00138 if(!(avctx->flags & CODEC_FLAG_BITEXACT)) { 00139 c->float_to_int16 = float_to_int16_altivec; 00140 c->float_to_int16_interleave = float_to_int16_interleave_altivec; 00141 } 00142 }