1.1 --- /dev/null Thu Jan 01 00:00:00 1970 +0000
1.2 +++ b/media/libvpx/vp8/common/x86/variance_mmx.c Wed Dec 31 06:09:35 2014 +0100
1.3 @@ -0,0 +1,398 @@
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 "vp8/common/variance.h"
1.16 +#include "vp8/common/pragmas.h"
1.17 +#include "vpx_ports/mem.h"
1.18 +#include "vp8/common/x86/filter_x86.h"
1.19 +
1.20 +extern void filter_block1d_h6_mmx
1.21 +(
1.22 + const unsigned char *src_ptr,
1.23 + unsigned short *output_ptr,
1.24 + unsigned int src_pixels_per_line,
1.25 + unsigned int pixel_step,
1.26 + unsigned int output_height,
1.27 + unsigned int output_width,
1.28 + short *filter
1.29 +);
1.30 +extern void filter_block1d_v6_mmx
1.31 +(
1.32 + const short *src_ptr,
1.33 + unsigned char *output_ptr,
1.34 + unsigned int pixels_per_line,
1.35 + unsigned int pixel_step,
1.36 + unsigned int output_height,
1.37 + unsigned int output_width,
1.38 + short *filter
1.39 +);
1.40 +
1.41 +extern unsigned int vp8_get_mb_ss_mmx(const short *src_ptr);
1.42 +extern unsigned int vp8_get8x8var_mmx
1.43 +(
1.44 + const unsigned char *src_ptr,
1.45 + int source_stride,
1.46 + const unsigned char *ref_ptr,
1.47 + int recon_stride,
1.48 + unsigned int *SSE,
1.49 + int *Sum
1.50 +);
1.51 +extern unsigned int vp8_get4x4var_mmx
1.52 +(
1.53 + const unsigned char *src_ptr,
1.54 + int source_stride,
1.55 + const unsigned char *ref_ptr,
1.56 + int recon_stride,
1.57 + unsigned int *SSE,
1.58 + int *Sum
1.59 +);
1.60 +extern void vp8_filter_block2d_bil4x4_var_mmx
1.61 +(
1.62 + const unsigned char *ref_ptr,
1.63 + int ref_pixels_per_line,
1.64 + const unsigned char *src_ptr,
1.65 + int src_pixels_per_line,
1.66 + const short *HFilter,
1.67 + const short *VFilter,
1.68 + int *sum,
1.69 + unsigned int *sumsquared
1.70 +);
1.71 +extern void vp8_filter_block2d_bil_var_mmx
1.72 +(
1.73 + const unsigned char *ref_ptr,
1.74 + int ref_pixels_per_line,
1.75 + const unsigned char *src_ptr,
1.76 + int src_pixels_per_line,
1.77 + unsigned int Height,
1.78 + const short *HFilter,
1.79 + const short *VFilter,
1.80 + int *sum,
1.81 + unsigned int *sumsquared
1.82 +);
1.83 +
1.84 +
1.85 +unsigned int vp8_variance4x4_mmx(
1.86 + const unsigned char *src_ptr,
1.87 + int source_stride,
1.88 + const unsigned char *ref_ptr,
1.89 + int recon_stride,
1.90 + unsigned int *sse)
1.91 +{
1.92 + unsigned int var;
1.93 + int avg;
1.94 +
1.95 + vp8_get4x4var_mmx(src_ptr, source_stride, ref_ptr, recon_stride, &var, &avg) ;
1.96 + *sse = var;
1.97 + return (var - (((unsigned int)avg * avg) >> 4));
1.98 +
1.99 +}
1.100 +
1.101 +unsigned int vp8_variance8x8_mmx(
1.102 + const unsigned char *src_ptr,
1.103 + int source_stride,
1.104 + const unsigned char *ref_ptr,
1.105 + int recon_stride,
1.106 + unsigned int *sse)
1.107 +{
1.108 + unsigned int var;
1.109 + int avg;
1.110 +
1.111 + vp8_get8x8var_mmx(src_ptr, source_stride, ref_ptr, recon_stride, &var, &avg) ;
1.112 + *sse = var;
1.113 +
1.114 + return (var - (((unsigned int)avg * avg) >> 6));
1.115 +
1.116 +}
1.117 +
1.118 +unsigned int vp8_mse16x16_mmx(
1.119 + const unsigned char *src_ptr,
1.120 + int source_stride,
1.121 + const unsigned char *ref_ptr,
1.122 + int recon_stride,
1.123 + unsigned int *sse)
1.124 +{
1.125 + unsigned int sse0, sse1, sse2, sse3, var;
1.126 + int sum0, sum1, sum2, sum3;
1.127 +
1.128 +
1.129 + vp8_get8x8var_mmx(src_ptr, source_stride, ref_ptr, recon_stride, &sse0, &sum0) ;
1.130 + vp8_get8x8var_mmx(src_ptr + 8, source_stride, ref_ptr + 8, recon_stride, &sse1, &sum1);
1.131 + vp8_get8x8var_mmx(src_ptr + 8 * source_stride, source_stride, ref_ptr + 8 * recon_stride, recon_stride, &sse2, &sum2) ;
1.132 + vp8_get8x8var_mmx(src_ptr + 8 * source_stride + 8, source_stride, ref_ptr + 8 * recon_stride + 8, recon_stride, &sse3, &sum3);
1.133 +
1.134 + var = sse0 + sse1 + sse2 + sse3;
1.135 + *sse = var;
1.136 + return var;
1.137 +}
1.138 +
1.139 +
1.140 +unsigned int vp8_variance16x16_mmx(
1.141 + const unsigned char *src_ptr,
1.142 + int source_stride,
1.143 + const unsigned char *ref_ptr,
1.144 + int recon_stride,
1.145 + unsigned int *sse)
1.146 +{
1.147 + unsigned int sse0, sse1, sse2, sse3, var;
1.148 + int sum0, sum1, sum2, sum3, avg;
1.149 +
1.150 +
1.151 + vp8_get8x8var_mmx(src_ptr, source_stride, ref_ptr, recon_stride, &sse0, &sum0) ;
1.152 + vp8_get8x8var_mmx(src_ptr + 8, source_stride, ref_ptr + 8, recon_stride, &sse1, &sum1);
1.153 + vp8_get8x8var_mmx(src_ptr + 8 * source_stride, source_stride, ref_ptr + 8 * recon_stride, recon_stride, &sse2, &sum2) ;
1.154 + vp8_get8x8var_mmx(src_ptr + 8 * source_stride + 8, source_stride, ref_ptr + 8 * recon_stride + 8, recon_stride, &sse3, &sum3);
1.155 +
1.156 + var = sse0 + sse1 + sse2 + sse3;
1.157 + avg = sum0 + sum1 + sum2 + sum3;
1.158 + *sse = var;
1.159 + return (var - (((unsigned int)avg * avg) >> 8));
1.160 +}
1.161 +
1.162 +unsigned int vp8_variance16x8_mmx(
1.163 + const unsigned char *src_ptr,
1.164 + int source_stride,
1.165 + const unsigned char *ref_ptr,
1.166 + int recon_stride,
1.167 + unsigned int *sse)
1.168 +{
1.169 + unsigned int sse0, sse1, var;
1.170 + int sum0, sum1, avg;
1.171 +
1.172 + vp8_get8x8var_mmx(src_ptr, source_stride, ref_ptr, recon_stride, &sse0, &sum0) ;
1.173 + vp8_get8x8var_mmx(src_ptr + 8, source_stride, ref_ptr + 8, recon_stride, &sse1, &sum1);
1.174 +
1.175 + var = sse0 + sse1;
1.176 + avg = sum0 + sum1;
1.177 + *sse = var;
1.178 + return (var - (((unsigned int)avg * avg) >> 7));
1.179 +
1.180 +}
1.181 +
1.182 +
1.183 +unsigned int vp8_variance8x16_mmx(
1.184 + const unsigned char *src_ptr,
1.185 + int source_stride,
1.186 + const unsigned char *ref_ptr,
1.187 + int recon_stride,
1.188 + unsigned int *sse)
1.189 +{
1.190 + unsigned int sse0, sse1, var;
1.191 + int sum0, sum1, avg;
1.192 +
1.193 + vp8_get8x8var_mmx(src_ptr, source_stride, ref_ptr, recon_stride, &sse0, &sum0) ;
1.194 + vp8_get8x8var_mmx(src_ptr + 8 * source_stride, source_stride, ref_ptr + 8 * recon_stride, recon_stride, &sse1, &sum1) ;
1.195 +
1.196 + var = sse0 + sse1;
1.197 + avg = sum0 + sum1;
1.198 + *sse = var;
1.199 +
1.200 + return (var - (((unsigned int)avg * avg) >> 7));
1.201 +
1.202 +}
1.203 +
1.204 +
1.205 +unsigned int vp8_sub_pixel_variance4x4_mmx
1.206 +(
1.207 + const unsigned char *src_ptr,
1.208 + int src_pixels_per_line,
1.209 + int xoffset,
1.210 + int yoffset,
1.211 + const unsigned char *dst_ptr,
1.212 + int dst_pixels_per_line,
1.213 + unsigned int *sse)
1.214 +
1.215 +{
1.216 + int xsum;
1.217 + unsigned int xxsum;
1.218 + vp8_filter_block2d_bil4x4_var_mmx(
1.219 + src_ptr, src_pixels_per_line,
1.220 + dst_ptr, dst_pixels_per_line,
1.221 + vp8_bilinear_filters_x86_4[xoffset], vp8_bilinear_filters_x86_4[yoffset],
1.222 + &xsum, &xxsum
1.223 + );
1.224 + *sse = xxsum;
1.225 + return (xxsum - (((unsigned int)xsum * xsum) >> 4));
1.226 +}
1.227 +
1.228 +
1.229 +unsigned int vp8_sub_pixel_variance8x8_mmx
1.230 +(
1.231 + const unsigned char *src_ptr,
1.232 + int src_pixels_per_line,
1.233 + int xoffset,
1.234 + int yoffset,
1.235 + const unsigned char *dst_ptr,
1.236 + int dst_pixels_per_line,
1.237 + unsigned int *sse
1.238 +)
1.239 +{
1.240 +
1.241 + int xsum;
1.242 + unsigned int xxsum;
1.243 + vp8_filter_block2d_bil_var_mmx(
1.244 + src_ptr, src_pixels_per_line,
1.245 + dst_ptr, dst_pixels_per_line, 8,
1.246 + vp8_bilinear_filters_x86_4[xoffset], vp8_bilinear_filters_x86_4[yoffset],
1.247 + &xsum, &xxsum
1.248 + );
1.249 + *sse = xxsum;
1.250 + return (xxsum - (((unsigned int)xsum * xsum) >> 6));
1.251 +}
1.252 +
1.253 +unsigned int vp8_sub_pixel_variance16x16_mmx
1.254 +(
1.255 + const unsigned char *src_ptr,
1.256 + int src_pixels_per_line,
1.257 + int xoffset,
1.258 + int yoffset,
1.259 + const unsigned char *dst_ptr,
1.260 + int dst_pixels_per_line,
1.261 + unsigned int *sse
1.262 +)
1.263 +{
1.264 +
1.265 + int xsum0, xsum1;
1.266 + unsigned int xxsum0, xxsum1;
1.267 +
1.268 +
1.269 + vp8_filter_block2d_bil_var_mmx(
1.270 + src_ptr, src_pixels_per_line,
1.271 + dst_ptr, dst_pixels_per_line, 16,
1.272 + vp8_bilinear_filters_x86_4[xoffset], vp8_bilinear_filters_x86_4[yoffset],
1.273 + &xsum0, &xxsum0
1.274 + );
1.275 +
1.276 +
1.277 + vp8_filter_block2d_bil_var_mmx(
1.278 + src_ptr + 8, src_pixels_per_line,
1.279 + dst_ptr + 8, dst_pixels_per_line, 16,
1.280 + vp8_bilinear_filters_x86_4[xoffset], vp8_bilinear_filters_x86_4[yoffset],
1.281 + &xsum1, &xxsum1
1.282 + );
1.283 +
1.284 + xsum0 += xsum1;
1.285 + xxsum0 += xxsum1;
1.286 +
1.287 + *sse = xxsum0;
1.288 + return (xxsum0 - (((unsigned int)xsum0 * xsum0) >> 8));
1.289 +
1.290 +
1.291 +}
1.292 +
1.293 +unsigned int vp8_sub_pixel_mse16x16_mmx(
1.294 + const unsigned char *src_ptr,
1.295 + int src_pixels_per_line,
1.296 + int xoffset,
1.297 + int yoffset,
1.298 + const unsigned char *dst_ptr,
1.299 + int dst_pixels_per_line,
1.300 + unsigned int *sse
1.301 +)
1.302 +{
1.303 + vp8_sub_pixel_variance16x16_mmx(src_ptr, src_pixels_per_line, xoffset, yoffset, dst_ptr, dst_pixels_per_line, sse);
1.304 + return *sse;
1.305 +}
1.306 +
1.307 +unsigned int vp8_sub_pixel_variance16x8_mmx
1.308 +(
1.309 + const unsigned char *src_ptr,
1.310 + int src_pixels_per_line,
1.311 + int xoffset,
1.312 + int yoffset,
1.313 + const unsigned char *dst_ptr,
1.314 + int dst_pixels_per_line,
1.315 + unsigned int *sse
1.316 +)
1.317 +{
1.318 + int xsum0, xsum1;
1.319 + unsigned int xxsum0, xxsum1;
1.320 +
1.321 +
1.322 + vp8_filter_block2d_bil_var_mmx(
1.323 + src_ptr, src_pixels_per_line,
1.324 + dst_ptr, dst_pixels_per_line, 8,
1.325 + vp8_bilinear_filters_x86_4[xoffset], vp8_bilinear_filters_x86_4[yoffset],
1.326 + &xsum0, &xxsum0
1.327 + );
1.328 +
1.329 +
1.330 + vp8_filter_block2d_bil_var_mmx(
1.331 + src_ptr + 8, src_pixels_per_line,
1.332 + dst_ptr + 8, dst_pixels_per_line, 8,
1.333 + vp8_bilinear_filters_x86_4[xoffset], vp8_bilinear_filters_x86_4[yoffset],
1.334 + &xsum1, &xxsum1
1.335 + );
1.336 +
1.337 + xsum0 += xsum1;
1.338 + xxsum0 += xxsum1;
1.339 +
1.340 + *sse = xxsum0;
1.341 + return (xxsum0 - (((unsigned int)xsum0 * xsum0) >> 7));
1.342 +}
1.343 +
1.344 +unsigned int vp8_sub_pixel_variance8x16_mmx
1.345 +(
1.346 + const unsigned char *src_ptr,
1.347 + int src_pixels_per_line,
1.348 + int xoffset,
1.349 + int yoffset,
1.350 + const unsigned char *dst_ptr,
1.351 + int dst_pixels_per_line,
1.352 + unsigned int *sse
1.353 +)
1.354 +{
1.355 + int xsum;
1.356 + unsigned int xxsum;
1.357 + vp8_filter_block2d_bil_var_mmx(
1.358 + src_ptr, src_pixels_per_line,
1.359 + dst_ptr, dst_pixels_per_line, 16,
1.360 + vp8_bilinear_filters_x86_4[xoffset], vp8_bilinear_filters_x86_4[yoffset],
1.361 + &xsum, &xxsum
1.362 + );
1.363 + *sse = xxsum;
1.364 + return (xxsum - (((unsigned int)xsum * xsum) >> 7));
1.365 +}
1.366 +
1.367 +
1.368 +unsigned int vp8_variance_halfpixvar16x16_h_mmx(
1.369 + const unsigned char *src_ptr,
1.370 + int source_stride,
1.371 + const unsigned char *ref_ptr,
1.372 + int recon_stride,
1.373 + unsigned int *sse)
1.374 +{
1.375 + return vp8_sub_pixel_variance16x16_mmx(src_ptr, source_stride, 4, 0,
1.376 + ref_ptr, recon_stride, sse);
1.377 +}
1.378 +
1.379 +
1.380 +unsigned int vp8_variance_halfpixvar16x16_v_mmx(
1.381 + const unsigned char *src_ptr,
1.382 + int source_stride,
1.383 + const unsigned char *ref_ptr,
1.384 + int recon_stride,
1.385 + unsigned int *sse)
1.386 +{
1.387 + return vp8_sub_pixel_variance16x16_mmx(src_ptr, source_stride, 0, 4,
1.388 + ref_ptr, recon_stride, sse);
1.389 +}
1.390 +
1.391 +
1.392 +unsigned int vp8_variance_halfpixvar16x16_hv_mmx(
1.393 + const unsigned char *src_ptr,
1.394 + int source_stride,
1.395 + const unsigned char *ref_ptr,
1.396 + int recon_stride,
1.397 + unsigned int *sse)
1.398 +{
1.399 + return vp8_sub_pixel_variance16x16_mmx(src_ptr, source_stride, 4, 4,
1.400 + ref_ptr, recon_stride, sse);
1.401 +}
