michael@0: /*
michael@0:  *  Copyright (c) 2010 The WebM project authors. All Rights Reserved.
michael@0:  *
michael@0:  *  Use of this source code is governed by a BSD-style license
michael@0:  *  that can be found in the LICENSE file in the root of the source
michael@0:  *  tree. An additional intellectual property rights grant can be found
michael@0:  *  in the file PATENTS.  All contributing project authors may
michael@0:  *  be found in the AUTHORS file in the root of the source tree.
michael@0:  */
michael@0: 
michael@0: #include "vpx_config.h"
michael@0: #include "vp8/common/variance.h"
michael@0: #include "vp8/common/pragmas.h"
michael@0: #include "vpx_ports/mem.h"
michael@0: 
michael@0: extern unsigned int vp8_get16x16var_sse2
michael@0: (
michael@0:     const unsigned char *src_ptr,
michael@0:     int source_stride,
michael@0:     const unsigned char *ref_ptr,
michael@0:     int recon_stride,
michael@0:     unsigned int *SSE,
michael@0:     int *Sum
michael@0: );
michael@0: extern void vp8_half_horiz_vert_variance16x_h_sse2
michael@0: (
michael@0:     const unsigned char *ref_ptr,
michael@0:     int ref_pixels_per_line,
michael@0:     const unsigned char *src_ptr,
michael@0:     int src_pixels_per_line,
michael@0:     unsigned int Height,
michael@0:     int *sum,
michael@0:     unsigned int *sumsquared
michael@0: );
michael@0: extern void vp8_half_horiz_variance16x_h_sse2
michael@0: (
michael@0:     const unsigned char *ref_ptr,
michael@0:     int ref_pixels_per_line,
michael@0:     const unsigned char *src_ptr,
michael@0:     int src_pixels_per_line,
michael@0:     unsigned int Height,
michael@0:     int *sum,
michael@0:     unsigned int *sumsquared
michael@0: );
michael@0: extern void vp8_half_vert_variance16x_h_sse2
michael@0: (
michael@0:     const unsigned char *ref_ptr,
michael@0:     int ref_pixels_per_line,
michael@0:     const unsigned char *src_ptr,
michael@0:     int src_pixels_per_line,
michael@0:     unsigned int Height,
michael@0:     int *sum,
michael@0:     unsigned int *sumsquared
michael@0: );
michael@0: extern void vp8_filter_block2d_bil_var_ssse3
michael@0: (
michael@0:     const unsigned char *ref_ptr,
michael@0:     int ref_pixels_per_line,
michael@0:     const unsigned char *src_ptr,
michael@0:     int src_pixels_per_line,
michael@0:     unsigned int Height,
michael@0:     int  xoffset,
michael@0:     int  yoffset,
michael@0:     int *sum,
michael@0:     unsigned int *sumsquared
michael@0: );
michael@0: 
michael@0: unsigned int vp8_sub_pixel_variance16x16_ssse3
michael@0: (
michael@0:     const unsigned char  *src_ptr,
michael@0:     int  src_pixels_per_line,
michael@0:     int  xoffset,
michael@0:     int  yoffset,
michael@0:     const unsigned char *dst_ptr,
michael@0:     int dst_pixels_per_line,
michael@0:     unsigned int *sse
michael@0: )
michael@0: {
michael@0:     int xsum0;
michael@0:     unsigned int xxsum0;
michael@0: 
michael@0:     /* note we could avoid these if statements if the calling function
michael@0:      * just called the appropriate functions inside.
michael@0:      */
michael@0:     if (xoffset == 4 && yoffset == 0)
michael@0:     {
michael@0:         vp8_half_horiz_variance16x_h_sse2(
michael@0:             src_ptr, src_pixels_per_line,
michael@0:             dst_ptr, dst_pixels_per_line, 16,
michael@0:             &xsum0, &xxsum0);
michael@0:     }
michael@0:     else if (xoffset == 0 && yoffset == 4)
michael@0:     {
michael@0:         vp8_half_vert_variance16x_h_sse2(
michael@0:             src_ptr, src_pixels_per_line,
michael@0:             dst_ptr, dst_pixels_per_line, 16,
michael@0:             &xsum0, &xxsum0);
michael@0:     }
michael@0:     else if (xoffset == 4 && yoffset == 4)
michael@0:     {
michael@0:         vp8_half_horiz_vert_variance16x_h_sse2(
michael@0:             src_ptr, src_pixels_per_line,
michael@0:             dst_ptr, dst_pixels_per_line, 16,
michael@0:             &xsum0, &xxsum0);
michael@0:     }
michael@0:     else
michael@0:     {
michael@0:         vp8_filter_block2d_bil_var_ssse3(
michael@0:             src_ptr, src_pixels_per_line,
michael@0:             dst_ptr, dst_pixels_per_line, 16,
michael@0:             xoffset, yoffset,
michael@0:             &xsum0, &xxsum0);
michael@0:     }
michael@0: 
michael@0:     *sse = xxsum0;
michael@0:     return (xxsum0 - (((unsigned int)xsum0 * xsum0) >> 8));
michael@0: }
michael@0: 
michael@0: unsigned int vp8_sub_pixel_variance16x8_ssse3
michael@0: (
michael@0:     const unsigned char  *src_ptr,
michael@0:     int  src_pixels_per_line,
michael@0:     int  xoffset,
michael@0:     int  yoffset,
michael@0:     const unsigned char *dst_ptr,
michael@0:     int dst_pixels_per_line,
michael@0:     unsigned int *sse
michael@0: 
michael@0: )
michael@0: {
michael@0:     int xsum0;
michael@0:     unsigned int xxsum0;
michael@0: 
michael@0:     if (xoffset == 4 && yoffset == 0)
michael@0:     {
michael@0:         vp8_half_horiz_variance16x_h_sse2(
michael@0:             src_ptr, src_pixels_per_line,
michael@0:             dst_ptr, dst_pixels_per_line, 8,
michael@0:             &xsum0, &xxsum0);
michael@0:     }
michael@0:     else if (xoffset == 0 && yoffset == 4)
michael@0:     {
michael@0:         vp8_half_vert_variance16x_h_sse2(
michael@0:             src_ptr, src_pixels_per_line,
michael@0:             dst_ptr, dst_pixels_per_line, 8,
michael@0:             &xsum0, &xxsum0);
michael@0:     }
michael@0:     else if (xoffset == 4 && yoffset == 4)
michael@0:     {
michael@0:         vp8_half_horiz_vert_variance16x_h_sse2(
michael@0:             src_ptr, src_pixels_per_line,
michael@0:             dst_ptr, dst_pixels_per_line, 8,
michael@0:             &xsum0, &xxsum0);
michael@0:     }
michael@0:     else
michael@0:     {
michael@0:         vp8_filter_block2d_bil_var_ssse3(
michael@0:             src_ptr, src_pixels_per_line,
michael@0:             dst_ptr, dst_pixels_per_line, 8,
michael@0:             xoffset, yoffset,
michael@0:             &xsum0, &xxsum0);
michael@0:     }
michael@0: 
michael@0:     *sse = xxsum0;
michael@0:     return (xxsum0 - (((unsigned int)xsum0 * xsum0) >> 7));
michael@0: }