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_rtcd.h"
michael@0: #include "vp8/common/variance.h"
michael@0: #include "vp8/common/filter.h"
michael@0: 
michael@0: #if HAVE_MEDIA
michael@0: #include "vp8/common/arm/bilinearfilter_arm.h"
michael@0: 
michael@0: unsigned int vp8_sub_pixel_variance8x8_armv6
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:     unsigned short first_pass[10*8];
michael@0:     unsigned char  second_pass[8*8];
michael@0:     const short *HFilter, *VFilter;
michael@0: 
michael@0:     HFilter = vp8_bilinear_filters[xoffset];
michael@0:     VFilter = vp8_bilinear_filters[yoffset];
michael@0: 
michael@0:     vp8_filter_block2d_bil_first_pass_armv6(src_ptr, first_pass,
michael@0:                                             src_pixels_per_line,
michael@0:                                             9, 8, HFilter);
michael@0:     vp8_filter_block2d_bil_second_pass_armv6(first_pass, second_pass,
michael@0:                                              8, 8, 8, VFilter);
michael@0: 
michael@0:     return vp8_variance8x8_armv6(second_pass, 8, dst_ptr,
michael@0:                                    dst_pixels_per_line, sse);
michael@0: }
michael@0: 
michael@0: unsigned int vp8_sub_pixel_variance16x16_armv6
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:     unsigned short first_pass[36*16];
michael@0:     unsigned char  second_pass[20*16];
michael@0:     const short *HFilter, *VFilter;
michael@0:     unsigned int var;
michael@0: 
michael@0:     if (xoffset == 4 && yoffset == 0)
michael@0:     {
michael@0:         var = vp8_variance_halfpixvar16x16_h_armv6(src_ptr, src_pixels_per_line,
michael@0:                                                    dst_ptr, dst_pixels_per_line, sse);
michael@0:     }
michael@0:     else if (xoffset == 0 && yoffset == 4)
michael@0:     {
michael@0:         var = vp8_variance_halfpixvar16x16_v_armv6(src_ptr, src_pixels_per_line,
michael@0:                                                    dst_ptr, dst_pixels_per_line, sse);
michael@0:     }
michael@0:     else if (xoffset == 4 && yoffset == 4)
michael@0:     {
michael@0:         var = vp8_variance_halfpixvar16x16_hv_armv6(src_ptr, src_pixels_per_line,
michael@0:                                                    dst_ptr, dst_pixels_per_line, sse);
michael@0:     }
michael@0:     else
michael@0:     {
michael@0:         HFilter = vp8_bilinear_filters[xoffset];
michael@0:         VFilter = vp8_bilinear_filters[yoffset];
michael@0: 
michael@0:         vp8_filter_block2d_bil_first_pass_armv6(src_ptr, first_pass,
michael@0:                                                 src_pixels_per_line,
michael@0:                                                 17, 16, HFilter);
michael@0:         vp8_filter_block2d_bil_second_pass_armv6(first_pass, second_pass,
michael@0:                                                  16, 16, 16, VFilter);
michael@0: 
michael@0:         var = vp8_variance16x16_armv6(second_pass, 16, dst_ptr,
michael@0:                                        dst_pixels_per_line, sse);
michael@0:     }
michael@0:     return var;
michael@0: }
michael@0: 
michael@0: #endif /* HAVE_MEDIA */
michael@0: 
michael@0: 
michael@0: #if HAVE_NEON
michael@0: 
michael@0: extern unsigned int vp8_sub_pixel_variance16x16_neon_func
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: unsigned int vp8_sub_pixel_variance16x16_neon
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:   if (xoffset == 4 && yoffset == 0)
michael@0:     return vp8_variance_halfpixvar16x16_h_neon(src_ptr, src_pixels_per_line, dst_ptr, dst_pixels_per_line, sse);
michael@0:   else if (xoffset == 0 && yoffset == 4)
michael@0:     return vp8_variance_halfpixvar16x16_v_neon(src_ptr, src_pixels_per_line, dst_ptr, dst_pixels_per_line, sse);
michael@0:   else if (xoffset == 4 && yoffset == 4)
michael@0:     return vp8_variance_halfpixvar16x16_hv_neon(src_ptr, src_pixels_per_line, dst_ptr, dst_pixels_per_line, sse);
michael@0:   else
michael@0:     return vp8_sub_pixel_variance16x16_neon_func(src_ptr, src_pixels_per_line, xoffset, yoffset, dst_ptr, dst_pixels_per_line, sse);
michael@0: }
michael@0: 
michael@0: #endif