• Main Page
  • Related Pages
  • Modules
  • Data Structures
  • Files
  • File List
  • Globals

libavcodec/ppc/int_altivec.c

Go to the documentation of this file.
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 }

Generated on Sat Feb 16 2013 09:23:13 for ffmpeg by  doxygen 1.7.1