1.1 --- /dev/null Thu Jan 01 00:00:00 1970 +0000
1.2 +++ b/media/libvpx/vp9/encoder/x86/vp9_variance_mmx.c Wed Dec 31 06:09:35 2014 +0100
1.3 @@ -0,0 +1,154 @@
1.4 +/*
1.5 + * Copyright (c) 2010 The WebM project authors. All Rights Reserved.
1.6 + *
1.7 + * Use of this source code is governed by a BSD-style license
1.8 + * that can be found in the LICENSE file in the root of the source
1.9 + * tree. An additional intellectual property rights grant can be found
1.10 + * in the file PATENTS. All contributing project authors may
1.11 + * be found in the AUTHORS file in the root of the source tree.
1.12 + */
1.13 +
1.14 +#include "./vpx_config.h"
1.15 +#include "vp9/encoder/vp9_variance.h"
1.16 +#include "vp9/common/vp9_pragmas.h"
1.17 +#include "vpx_ports/mem.h"
1.18 +
1.19 +extern unsigned int vp9_get_mb_ss_mmx(const int16_t *src_ptr);
1.20 +extern unsigned int vp9_get8x8var_mmx
1.21 +(
1.22 + const unsigned char *src_ptr,
1.23 + int source_stride,
1.24 + const unsigned char *ref_ptr,
1.25 + int recon_stride,
1.26 + unsigned int *SSE,
1.27 + int *Sum
1.28 +);
1.29 +extern unsigned int vp9_get4x4var_mmx
1.30 +(
1.31 + const unsigned char *src_ptr,
1.32 + int source_stride,
1.33 + const unsigned char *ref_ptr,
1.34 + int recon_stride,
1.35 + unsigned int *SSE,
1.36 + int *Sum
1.37 +);
1.38 +
1.39 +unsigned int vp9_variance4x4_mmx(
1.40 + const unsigned char *src_ptr,
1.41 + int source_stride,
1.42 + const unsigned char *ref_ptr,
1.43 + int recon_stride,
1.44 + unsigned int *sse) {
1.45 + unsigned int var;
1.46 + int avg;
1.47 +
1.48 + vp9_get4x4var_mmx(src_ptr, source_stride, ref_ptr, recon_stride, &var, &avg);
1.49 + *sse = var;
1.50 + return (var - (((unsigned int)avg * avg) >> 4));
1.51 +}
1.52 +
1.53 +unsigned int vp9_variance8x8_mmx(
1.54 + const unsigned char *src_ptr,
1.55 + int source_stride,
1.56 + const unsigned char *ref_ptr,
1.57 + int recon_stride,
1.58 + unsigned int *sse) {
1.59 + unsigned int var;
1.60 + int avg;
1.61 +
1.62 + vp9_get8x8var_mmx(src_ptr, source_stride, ref_ptr, recon_stride, &var, &avg);
1.63 + *sse = var;
1.64 +
1.65 + return (var - (((unsigned int)avg * avg) >> 6));
1.66 +}
1.67 +
1.68 +unsigned int vp9_mse16x16_mmx(
1.69 + const unsigned char *src_ptr,
1.70 + int source_stride,
1.71 + const unsigned char *ref_ptr,
1.72 + int recon_stride,
1.73 + unsigned int *sse) {
1.74 + unsigned int sse0, sse1, sse2, sse3, var;
1.75 + int sum0, sum1, sum2, sum3;
1.76 +
1.77 +
1.78 + vp9_get8x8var_mmx(src_ptr, source_stride, ref_ptr, recon_stride, &sse0,
1.79 + &sum0);
1.80 + vp9_get8x8var_mmx(src_ptr + 8, source_stride, ref_ptr + 8, recon_stride,
1.81 + &sse1, &sum1);
1.82 + vp9_get8x8var_mmx(src_ptr + 8 * source_stride, source_stride,
1.83 + ref_ptr + 8 * recon_stride, recon_stride, &sse2, &sum2);
1.84 + vp9_get8x8var_mmx(src_ptr + 8 * source_stride + 8, source_stride,
1.85 + ref_ptr + 8 * recon_stride + 8, recon_stride, &sse3, &sum3);
1.86 +
1.87 + var = sse0 + sse1 + sse2 + sse3;
1.88 + *sse = var;
1.89 + return var;
1.90 +}
1.91 +
1.92 +
1.93 +unsigned int vp9_variance16x16_mmx(
1.94 + const unsigned char *src_ptr,
1.95 + int source_stride,
1.96 + const unsigned char *ref_ptr,
1.97 + int recon_stride,
1.98 + unsigned int *sse) {
1.99 + unsigned int sse0, sse1, sse2, sse3, var;
1.100 + int sum0, sum1, sum2, sum3, avg;
1.101 +
1.102 + vp9_get8x8var_mmx(src_ptr, source_stride, ref_ptr, recon_stride, &sse0,
1.103 + &sum0);
1.104 + vp9_get8x8var_mmx(src_ptr + 8, source_stride, ref_ptr + 8, recon_stride,
1.105 + &sse1, &sum1);
1.106 + vp9_get8x8var_mmx(src_ptr + 8 * source_stride, source_stride,
1.107 + ref_ptr + 8 * recon_stride, recon_stride, &sse2, &sum2);
1.108 + vp9_get8x8var_mmx(src_ptr + 8 * source_stride + 8, source_stride,
1.109 + ref_ptr + 8 * recon_stride + 8, recon_stride, &sse3, &sum3);
1.110 +
1.111 + var = sse0 + sse1 + sse2 + sse3;
1.112 + avg = sum0 + sum1 + sum2 + sum3;
1.113 + *sse = var;
1.114 + return (var - (((unsigned int)avg * avg) >> 8));
1.115 +}
1.116 +
1.117 +unsigned int vp9_variance16x8_mmx(
1.118 + const unsigned char *src_ptr,
1.119 + int source_stride,
1.120 + const unsigned char *ref_ptr,
1.121 + int recon_stride,
1.122 + unsigned int *sse) {
1.123 + unsigned int sse0, sse1, var;
1.124 + int sum0, sum1, avg;
1.125 +
1.126 + vp9_get8x8var_mmx(src_ptr, source_stride, ref_ptr, recon_stride, &sse0,
1.127 + &sum0);
1.128 + vp9_get8x8var_mmx(src_ptr + 8, source_stride, ref_ptr + 8, recon_stride,
1.129 + &sse1, &sum1);
1.130 +
1.131 + var = sse0 + sse1;
1.132 + avg = sum0 + sum1;
1.133 + *sse = var;
1.134 + return (var - (((unsigned int)avg * avg) >> 7));
1.135 +}
1.136 +
1.137 +
1.138 +unsigned int vp9_variance8x16_mmx(
1.139 + const unsigned char *src_ptr,
1.140 + int source_stride,
1.141 + const unsigned char *ref_ptr,
1.142 + int recon_stride,
1.143 + unsigned int *sse) {
1.144 + unsigned int sse0, sse1, var;
1.145 + int sum0, sum1, avg;
1.146 +
1.147 + vp9_get8x8var_mmx(src_ptr, source_stride, ref_ptr, recon_stride, &sse0,
1.148 + &sum0);
1.149 + vp9_get8x8var_mmx(src_ptr + 8 * source_stride, source_stride,
1.150 + ref_ptr + 8 * recon_stride, recon_stride, &sse1, &sum1);
1.151 +
1.152 + var = sse0 + sse1;
1.153 + avg = sum0 + sum1;
1.154 + *sse = var;
1.155 +
1.156 + return (var - (((unsigned int)avg * avg) >> 7));
1.157 +}
