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/dsputil.h" 00022 00023 #include "dsputil_altivec.h" 00024 #include "util_altivec.h" 00025 00026 static void vector_fmul_altivec(float *dst, const float *src0, const float *src1, int len) 00027 { 00028 int i; 00029 vector float d0, d1, s, zero = (vector float)vec_splat_u32(0); 00030 for(i=0; i<len-7; i+=8) { 00031 d0 = vec_ld(0, src0+i); 00032 s = vec_ld(0, src1+i); 00033 d1 = vec_ld(16, src0+i); 00034 d0 = vec_madd(d0, s, zero); 00035 d1 = vec_madd(d1, vec_ld(16,src1+i), zero); 00036 vec_st(d0, 0, dst+i); 00037 vec_st(d1, 16, dst+i); 00038 } 00039 } 00040 00041 static void vector_fmul_reverse_altivec(float *dst, const float *src0, 00042 const float *src1, int len) 00043 { 00044 int i; 00045 vector float d, s0, s1, h0, l0, 00046 s2, s3, zero = (vector float)vec_splat_u32(0); 00047 src1 += len-4; 00048 for(i=0; i<len-7; i+=8) { 00049 s1 = vec_ld(0, src1-i); // [a,b,c,d] 00050 s0 = vec_ld(0, src0+i); 00051 l0 = vec_mergel(s1, s1); // [c,c,d,d] 00052 s3 = vec_ld(-16, src1-i); 00053 h0 = vec_mergeh(s1, s1); // [a,a,b,b] 00054 s2 = vec_ld(16, src0+i); 00055 s1 = vec_mergeh(vec_mergel(l0,h0), // [d,b,d,b] 00056 vec_mergeh(l0,h0)); // [c,a,c,a] 00057 // [d,c,b,a] 00058 l0 = vec_mergel(s3, s3); 00059 d = vec_madd(s0, s1, zero); 00060 h0 = vec_mergeh(s3, s3); 00061 vec_st(d, 0, dst+i); 00062 s3 = vec_mergeh(vec_mergel(l0,h0), 00063 vec_mergeh(l0,h0)); 00064 d = vec_madd(s2, s3, zero); 00065 vec_st(d, 16, dst+i); 00066 } 00067 } 00068 00069 static void vector_fmul_add_altivec(float *dst, const float *src0, 00070 const float *src1, const float *src2, 00071 int len) 00072 { 00073 int i; 00074 vector float d, s0, s1, s2, t0, t1, edges; 00075 vector unsigned char align = vec_lvsr(0,dst), 00076 mask = vec_lvsl(0, dst); 00077 00078 for (i=0; i<len-3; i+=4) { 00079 t0 = vec_ld(0, dst+i); 00080 t1 = vec_ld(15, dst+i); 00081 s0 = vec_ld(0, src0+i); 00082 s1 = vec_ld(0, src1+i); 00083 s2 = vec_ld(0, src2+i); 00084 edges = vec_perm(t1 ,t0, mask); 00085 d = vec_madd(s0,s1,s2); 00086 t1 = vec_perm(d, edges, align); 00087 t0 = vec_perm(edges, d, align); 00088 vec_st(t1, 15, dst+i); 00089 vec_st(t0, 0, dst+i); 00090 } 00091 } 00092 00093 static void vector_fmul_window_altivec(float *dst, const float *src0, const float *src1, const float *win, int len) 00094 { 00095 vector float zero, t0, t1, s0, s1, wi, wj; 00096 const vector unsigned char reverse = vcprm(3,2,1,0); 00097 int i,j; 00098 00099 dst += len; 00100 win += len; 00101 src0+= len; 00102 00103 zero = (vector float)vec_splat_u32(0); 00104 00105 for(i=-len*4, j=len*4-16; i<0; i+=16, j-=16) { 00106 s0 = vec_ld(i, src0); 00107 s1 = vec_ld(j, src1); 00108 wi = vec_ld(i, win); 00109 wj = vec_ld(j, win); 00110 00111 s1 = vec_perm(s1, s1, reverse); 00112 wj = vec_perm(wj, wj, reverse); 00113 00114 t0 = vec_madd(s0, wj, zero); 00115 t0 = vec_nmsub(s1, wi, t0); 00116 t1 = vec_madd(s0, wi, zero); 00117 t1 = vec_madd(s1, wj, t1); 00118 t1 = vec_perm(t1, t1, reverse); 00119 00120 vec_st(t0, i, dst); 00121 vec_st(t1, j, dst); 00122 } 00123 } 00124 00125 void float_init_altivec(DSPContext* c, AVCodecContext *avctx) 00126 { 00127 c->vector_fmul = vector_fmul_altivec; 00128 c->vector_fmul_reverse = vector_fmul_reverse_altivec; 00129 c->vector_fmul_add = vector_fmul_add_altivec; 00130 if(!(avctx->flags & CODEC_FLAG_BITEXACT)) { 00131 c->vector_fmul_window = vector_fmul_window_altivec; 00132 } 00133 }