Home | History | Annotate | Download | only in x86 
      1 /*
      2  *  Copyright (c) 2010 The WebM project authors. All Rights Reserved.
      3  *
      4  *  Use of this source code is governed by a BSD-style license
      5  *  that can be found in the LICENSE file in the root of the source
      6  *  tree. An additional intellectual property rights grant can be found
      7  *  in the file PATENTS.  All contributing project authors may
      8  *  be found in the AUTHORS file in the root of the source tree.
      9  */
     10 
     11 #include "./vpx_config.h"
     12 #include "vp9/encoder/vp9_variance.h"
     13 #include "vp9/common/vp9_pragmas.h"
     14 #include "vpx_ports/mem.h"
     15 
     16 extern unsigned int vp9_get_mb_ss_mmx(const int16_t *src_ptr);
     17 extern unsigned int vp9_get8x8var_mmx
     18 (
     19   const unsigned char *src_ptr,
     20   int  source_stride,
     21   const unsigned char *ref_ptr,
     22   int  recon_stride,
     23   unsigned int *SSE,
     24   int *Sum
     25 );
     26 extern unsigned int vp9_get4x4var_mmx
     27 (
     28   const unsigned char *src_ptr,
     29   int  source_stride,
     30   const unsigned char *ref_ptr,
     31   int  recon_stride,
     32   unsigned int *SSE,
     33   int *Sum
     34 );
     35 
     36 unsigned int vp9_variance4x4_mmx(
     37   const unsigned char *src_ptr,
     38   int  source_stride,
     39   const unsigned char *ref_ptr,
     40   int  recon_stride,
     41   unsigned int *sse) {
     42   unsigned int var;
     43   int avg;
     44 
     45   vp9_get4x4var_mmx(src_ptr, source_stride, ref_ptr, recon_stride, &var, &avg);
     46   *sse = var;
     47   return (var - (((unsigned int)avg * avg) >> 4));
     48 }
     49 
     50 unsigned int vp9_variance8x8_mmx(
     51   const unsigned char *src_ptr,
     52   int  source_stride,
     53   const unsigned char *ref_ptr,
     54   int  recon_stride,
     55   unsigned int *sse) {
     56   unsigned int var;
     57   int avg;
     58 
     59   vp9_get8x8var_mmx(src_ptr, source_stride, ref_ptr, recon_stride, &var, &avg);
     60   *sse = var;
     61 
     62   return (var - (((unsigned int)avg * avg) >> 6));
     63 }
     64 
     65 unsigned int vp9_mse16x16_mmx(
     66   const unsigned char *src_ptr,
     67   int  source_stride,
     68   const unsigned char *ref_ptr,
     69   int  recon_stride,
     70   unsigned int *sse) {
     71   unsigned int sse0, sse1, sse2, sse3, var;
     72   int sum0, sum1, sum2, sum3;
     73 
     74 
     75   vp9_get8x8var_mmx(src_ptr, source_stride, ref_ptr, recon_stride, &sse0,
     76                     &sum0);
     77   vp9_get8x8var_mmx(src_ptr + 8, source_stride, ref_ptr + 8, recon_stride,
     78                     &sse1, &sum1);
     79   vp9_get8x8var_mmx(src_ptr + 8 * source_stride, source_stride,
     80                     ref_ptr + 8 * recon_stride, recon_stride, &sse2, &sum2);
     81   vp9_get8x8var_mmx(src_ptr + 8 * source_stride + 8, source_stride,
     82                     ref_ptr + 8 * recon_stride + 8, recon_stride, &sse3, &sum3);
     83 
     84   var = sse0 + sse1 + sse2 + sse3;
     85   *sse = var;
     86   return var;
     87 }
     88 
     89 
     90 unsigned int vp9_variance16x16_mmx(
     91   const unsigned char *src_ptr,
     92   int  source_stride,
     93   const unsigned char *ref_ptr,
     94   int  recon_stride,
     95   unsigned int *sse) {
     96   unsigned int sse0, sse1, sse2, sse3, var;
     97   int sum0, sum1, sum2, sum3, avg;
     98 
     99   vp9_get8x8var_mmx(src_ptr, source_stride, ref_ptr, recon_stride, &sse0,
    100                     &sum0);
    101   vp9_get8x8var_mmx(src_ptr + 8, source_stride, ref_ptr + 8, recon_stride,
    102                     &sse1, &sum1);
    103   vp9_get8x8var_mmx(src_ptr + 8 * source_stride, source_stride,
    104                     ref_ptr + 8 * recon_stride, recon_stride, &sse2, &sum2);
    105   vp9_get8x8var_mmx(src_ptr + 8 * source_stride + 8, source_stride,
    106                     ref_ptr + 8 * recon_stride + 8, recon_stride, &sse3, &sum3);
    107 
    108   var = sse0 + sse1 + sse2 + sse3;
    109   avg = sum0 + sum1 + sum2 + sum3;
    110   *sse = var;
    111   return (var - (((unsigned int)avg * avg) >> 8));
    112 }
    113 
    114 unsigned int vp9_variance16x8_mmx(
    115   const unsigned char *src_ptr,
    116   int  source_stride,
    117   const unsigned char *ref_ptr,
    118   int  recon_stride,
    119   unsigned int *sse) {
    120   unsigned int sse0, sse1, var;
    121   int sum0, sum1, avg;
    122 
    123   vp9_get8x8var_mmx(src_ptr, source_stride, ref_ptr, recon_stride, &sse0,
    124                     &sum0);
    125   vp9_get8x8var_mmx(src_ptr + 8, source_stride, ref_ptr + 8, recon_stride,
    126                     &sse1, &sum1);
    127 
    128   var = sse0 + sse1;
    129   avg = sum0 + sum1;
    130   *sse = var;
    131   return (var - (((unsigned int)avg * avg) >> 7));
    132 }
    133 
    134 
    135 unsigned int vp9_variance8x16_mmx(
    136   const unsigned char *src_ptr,
    137   int  source_stride,
    138   const unsigned char *ref_ptr,
    139   int  recon_stride,
    140   unsigned int *sse) {
    141   unsigned int sse0, sse1, var;
    142   int sum0, sum1, avg;
    143 
    144   vp9_get8x8var_mmx(src_ptr, source_stride, ref_ptr, recon_stride, &sse0,
    145                     &sum0);
    146   vp9_get8x8var_mmx(src_ptr + 8 * source_stride, source_stride,
    147                     ref_ptr + 8 * recon_stride, recon_stride, &sse1, &sum1);
    148 
    149   var = sse0 + sse1;
    150   avg = sum0 + sum1;
    151   *sse = var;
    152 
    153   return (var - (((unsigned int)avg * avg) >> 7));
    154 }
    155