00001 /* 00002 * Copyright (c) 2007 Luca Barbato <lu_zero@gentoo.org> 00003 * 00004 * This file is part of FFmpeg. 00005 * 00006 * FFmpeg 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 * FFmpeg 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 FFmpeg; if not, write to the Free Software 00018 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA 00019 */ 00020 00026 #include "libavcodec/dsputil.h" 00027 00028 #include "gcc_fixes.h" 00029 00030 #include "dsputil_altivec.h" 00031 00032 #include "types_altivec.h" 00033 00034 static int ssd_int8_vs_int16_altivec(const int8_t *pix1, const int16_t *pix2, 00035 int size) { 00036 int i, size16; 00037 vector signed char vpix1; 00038 vector signed short vpix2, vdiff, vpix1l,vpix1h; 00039 union { vector signed int vscore; 00040 int32_t score[4]; 00041 } u; 00042 u.vscore = vec_splat_s32(0); 00043 // 00044 //XXX lazy way, fix it later 00045 00046 #define vec_unaligned_load(b) \ 00047 vec_perm(vec_ld(0,b),vec_ld(15,b),vec_lvsl(0, b)); 00048 00049 size16 = size >> 4; 00050 while(size16) { 00051 // score += (pix1[i]-pix2[i])*(pix1[i]-pix2[i]); 00052 //load pix1 and the first batch of pix2 00053 00054 vpix1 = vec_unaligned_load(pix1); 00055 vpix2 = vec_unaligned_load(pix2); 00056 pix2 += 8; 00057 //unpack 00058 vpix1h = vec_unpackh(vpix1); 00059 vdiff = vec_sub(vpix1h, vpix2); 00060 vpix1l = vec_unpackl(vpix1); 00061 // load another batch from pix2 00062 vpix2 = vec_unaligned_load(pix2); 00063 u.vscore = vec_msum(vdiff, vdiff, u.vscore); 00064 vdiff = vec_sub(vpix1l, vpix2); 00065 u.vscore = vec_msum(vdiff, vdiff, u.vscore); 00066 pix1 += 16; 00067 pix2 += 8; 00068 size16--; 00069 } 00070 u.vscore = vec_sums(u.vscore, vec_splat_s32(0)); 00071 00072 size %= 16; 00073 for (i = 0; i < size; i++) { 00074 u.score[3] += (pix1[i]-pix2[i])*(pix1[i]-pix2[i]); 00075 } 00076 return u.score[3]; 00077 } 00078 00079 static void add_int16_altivec(int16_t * v1, int16_t * v2, int order) 00080 { 00081 int i; 00082 register vec_s16 vec, *pv; 00083 00084 for(i = 0; i < order; i += 8){ 00085 pv = (vec_s16*)v2; 00086 vec = vec_perm(pv[0], pv[1], vec_lvsl(0, v2)); 00087 vec_st(vec_add(vec_ld(0, v1), vec), 0, v1); 00088 v1 += 8; 00089 v2 += 8; 00090 } 00091 } 00092 00093 static void sub_int16_altivec(int16_t * v1, int16_t * v2, int order) 00094 { 00095 int i; 00096 register vec_s16 vec, *pv; 00097 00098 for(i = 0; i < order; i += 8){ 00099 pv = (vec_s16*)v2; 00100 vec = vec_perm(pv[0], pv[1], vec_lvsl(0, v2)); 00101 vec_st(vec_sub(vec_ld(0, v1), vec), 0, v1); 00102 v1 += 8; 00103 v2 += 8; 00104 } 00105 } 00106 00107 static int32_t scalarproduct_int16_altivec(int16_t * v1, int16_t * v2, int order, const int shift) 00108 { 00109 int i; 00110 LOAD_ZERO; 00111 register vec_s16 vec1, *pv; 00112 register vec_s32 res = vec_splat_s32(0), t; 00113 register vec_u32 shifts; 00114 DECLARE_ALIGNED_16(int32_t, ires); 00115 00116 shifts = zero_u32v; 00117 if(shift & 0x10) shifts = vec_add(shifts, vec_sl(vec_splat_u32(0x08), vec_splat_u32(0x1))); 00118 if(shift & 0x08) shifts = vec_add(shifts, vec_splat_u32(0x08)); 00119 if(shift & 0x04) shifts = vec_add(shifts, vec_splat_u32(0x04)); 00120 if(shift & 0x02) shifts = vec_add(shifts, vec_splat_u32(0x02)); 00121 if(shift & 0x01) shifts = vec_add(shifts, vec_splat_u32(0x01)); 00122 00123 for(i = 0; i < order; i += 8){ 00124 pv = (vec_s16*)v1; 00125 vec1 = vec_perm(pv[0], pv[1], vec_lvsl(0, v1)); 00126 t = vec_msum(vec1, vec_ld(0, v2), zero_s32v); 00127 t = vec_sr(t, shifts); 00128 res = vec_sums(t, res); 00129 v1 += 8; 00130 v2 += 8; 00131 } 00132 res = vec_splat(res, 3); 00133 vec_ste(res, 0, &ires); 00134 return ires; 00135 } 00136 00137 void int_init_altivec(DSPContext* c, AVCodecContext *avctx) 00138 { 00139 c->ssd_int8_vs_int16 = ssd_int8_vs_int16_altivec; 00140 c->add_int16 = add_int16_altivec; 00141 c->sub_int16 = sub_int16_altivec; 00142 c->scalarproduct_int16 = scalarproduct_int16_altivec; 00143 }