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 short *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 51 unsigned int vp9_variance8x8_mmx( 52 const unsigned char *src_ptr, 53 int source_stride, 54 const unsigned char *ref_ptr, 55 int recon_stride, 56 unsigned int *sse) { 57 unsigned int var; 58 int avg; 59 60 vp9_get8x8var_mmx(src_ptr, source_stride, ref_ptr, recon_stride, &var, &avg); 61 *sse = var; 62 63 return (var - (((unsigned int)avg * avg) >> 6)); 64 65 } 66 67 unsigned int vp9_mse16x16_mmx( 68 const unsigned char *src_ptr, 69 int source_stride, 70 const unsigned char *ref_ptr, 71 int recon_stride, 72 unsigned int *sse) { 73 unsigned int sse0, sse1, sse2, sse3, var; 74 int sum0, sum1, sum2, sum3; 75 76 77 vp9_get8x8var_mmx(src_ptr, source_stride, ref_ptr, recon_stride, &sse0, &sum0); 78 vp9_get8x8var_mmx(src_ptr + 8, source_stride, ref_ptr + 8, recon_stride, &sse1, &sum1); 79 vp9_get8x8var_mmx(src_ptr + 8 * source_stride, source_stride, ref_ptr + 8 * recon_stride, recon_stride, &sse2, &sum2); 80 vp9_get8x8var_mmx(src_ptr + 8 * source_stride + 8, source_stride, ref_ptr + 8 * recon_stride + 8, recon_stride, &sse3, &sum3); 81 82 var = sse0 + sse1 + sse2 + sse3; 83 *sse = var; 84 return var; 85 } 86 87 88 unsigned int vp9_variance16x16_mmx( 89 const unsigned char *src_ptr, 90 int source_stride, 91 const unsigned char *ref_ptr, 92 int recon_stride, 93 unsigned int *sse) { 94 unsigned int sse0, sse1, sse2, sse3, var; 95 int sum0, sum1, sum2, sum3, avg; 96 97 98 vp9_get8x8var_mmx(src_ptr, source_stride, ref_ptr, recon_stride, &sse0, &sum0); 99 vp9_get8x8var_mmx(src_ptr + 8, source_stride, ref_ptr + 8, recon_stride, &sse1, &sum1); 100 vp9_get8x8var_mmx(src_ptr + 8 * source_stride, source_stride, ref_ptr + 8 * recon_stride, recon_stride, &sse2, &sum2); 101 vp9_get8x8var_mmx(src_ptr + 8 * source_stride + 8, source_stride, ref_ptr + 8 * recon_stride + 8, recon_stride, &sse3, &sum3); 102 103 var = sse0 + sse1 + sse2 + sse3; 104 avg = sum0 + sum1 + sum2 + sum3; 105 *sse = var; 106 return (var - (((unsigned int)avg * avg) >> 8)); 107 } 108 109 unsigned int vp9_variance16x8_mmx( 110 const unsigned char *src_ptr, 111 int source_stride, 112 const unsigned char *ref_ptr, 113 int recon_stride, 114 unsigned int *sse) { 115 unsigned int sse0, sse1, var; 116 int sum0, sum1, avg; 117 118 vp9_get8x8var_mmx(src_ptr, source_stride, ref_ptr, recon_stride, &sse0, &sum0); 119 vp9_get8x8var_mmx(src_ptr + 8, source_stride, ref_ptr + 8, recon_stride, &sse1, &sum1); 120 121 var = sse0 + sse1; 122 avg = sum0 + sum1; 123 *sse = var; 124 return (var - (((unsigned int)avg * avg) >> 7)); 125 126 } 127 128 129 unsigned int vp9_variance8x16_mmx( 130 const unsigned char *src_ptr, 131 int source_stride, 132 const unsigned char *ref_ptr, 133 int recon_stride, 134 unsigned int *sse) { 135 unsigned int sse0, sse1, var; 136 int sum0, sum1, avg; 137 138 vp9_get8x8var_mmx(src_ptr, source_stride, ref_ptr, recon_stride, &sse0, &sum0); 139 vp9_get8x8var_mmx(src_ptr + 8 * source_stride, source_stride, ref_ptr + 8 * recon_stride, recon_stride, &sse1, &sum1); 140 141 var = sse0 + sse1; 142 avg = sum0 + sum1; 143 *sse = var; 144 145 return (var - (((unsigned int)avg * avg) >> 7)); 146 147 } 148
