vpx/vp8/encoder/x86/variance_sse2.c
Deb Mukherjee 18e90d744e Supporting high precision 1/8-pel motion vectors
This is the initial patch for supporting 1/8th pel
motion. Currently if we configure with enable-high-precision-mv,
all motion vectors would default to 1/8 pel. Encode and
decode syncs fine with the current code. In the next phase
the code will be refactored so that we can choose the 1/8
pel mode adaptively at a frame/segment/mb level.

Derf results:
http://www.corp.google.com/~debargha/vp8_results/enhinterp_hpmv.html
(about 0.83% better than 8-tap interpoaltion)

Patch 3: Rebased. Also adding 1/16th pel interpolation for U and V

Patch 4: HD results.
http://www.corp.google.com/~debargha/vp8_results/enhinterp_hd_hpmv.html
Seems impressive (unless I am doing something wrong).

Patch 5: Added mmx/sse for bilateral filtering, as well as enforced
use of c-versions of subpel filters with 8-taps and 1/16th pel;
Also redesigned the 8-tap filters to reduce the cut-off in order to
introduce a denoising effect. There is a new configure option
sixteenth-subpel-uv which will use 1/16 th pel interpolation for
uv, if the motion vectors have 1/8 pel accuracy.

With the fixes the results are promising on the derf set. The enhanced
interpolation option with 8-taps alone gives 3% improvement over thei
derf set:
http://www.corp.google.com/~debargha/vp8_results/enhinterpn.html

Results on high precision mv and on the hd set are to follow.

Patch 6: Adding a missing condition for CONFIG_SIXTEENTH_SUBPEL_UV in
vp8/common/x86/x86_systemdependent.c

Patch 7: Cleaning up various debug messages.

Patch 8: Merge conflict

Change-Id: I5b1d844457aefd7414a9e4e0e06c6ed38fd8cc04
2012-02-23 09:25:21 -08:00

569 lines
14 KiB
C

/*
* Copyright (c) 2010 The WebM project authors. All Rights Reserved.
*
* Use of this source code is governed by a BSD-style license
* that can be found in the LICENSE file in the root of the source
* tree. An additional intellectual property rights grant can be found
* in the file PATENTS. All contributing project authors may
* be found in the AUTHORS file in the root of the source tree.
*/
#include "vpx_config.h"
#include "vp8/encoder/variance.h"
#include "vp8/common/pragmas.h"
#include "vpx_ports/mem.h"
#if CONFIG_SIXTEENTH_SUBPEL_UV
#define HALFNDX 8
#else
#define HALFNDX 4
#endif
extern void filter_block1d_h6_mmx(const unsigned char *src_ptr, unsigned short *output_ptr, unsigned int src_pixels_per_line, unsigned int pixel_step, unsigned int output_height, unsigned int output_width, short *vp7_filter);
extern void filter_block1d_v6_mmx(const short *src_ptr, unsigned char *output_ptr, unsigned int pixels_per_line, unsigned int pixel_step, unsigned int output_height, unsigned int output_width, short *vp7_filter);
extern void filter_block1d8_h6_sse2(const unsigned char *src_ptr, unsigned short *output_ptr, unsigned int src_pixels_per_line, unsigned int pixel_step, unsigned int output_height, unsigned int output_width, short *vp7_filter);
extern void filter_block1d8_v6_sse2(const short *src_ptr, unsigned char *output_ptr, unsigned int pixels_per_line, unsigned int pixel_step, unsigned int output_height, unsigned int output_width, short *vp7_filter);
extern void vp8_filter_block2d_bil4x4_var_mmx
(
const unsigned char *ref_ptr,
int ref_pixels_per_line,
const unsigned char *src_ptr,
int src_pixels_per_line,
const short *HFilter,
const short *VFilter,
int *sum,
unsigned int *sumsquared
);
extern unsigned int vp8_get4x4var_mmx
(
const unsigned char *src_ptr,
int source_stride,
const unsigned char *ref_ptr,
int recon_stride,
unsigned int *SSE,
int *Sum
);
unsigned int vp8_get_mb_ss_sse2
(
const short *src_ptr
);
unsigned int vp8_get16x16var_sse2
(
const unsigned char *src_ptr,
int source_stride,
const unsigned char *ref_ptr,
int recon_stride,
unsigned int *SSE,
int *Sum
);
unsigned int vp8_get8x8var_sse2
(
const unsigned char *src_ptr,
int source_stride,
const unsigned char *ref_ptr,
int recon_stride,
unsigned int *SSE,
int *Sum
);
void vp8_filter_block2d_bil_var_sse2
(
const unsigned char *ref_ptr,
int ref_pixels_per_line,
const unsigned char *src_ptr,
int src_pixels_per_line,
unsigned int Height,
int xoffset,
int yoffset,
int *sum,
unsigned int *sumsquared
);
void vp8_half_horiz_vert_variance8x_h_sse2
(
const unsigned char *ref_ptr,
int ref_pixels_per_line,
const unsigned char *src_ptr,
int src_pixels_per_line,
unsigned int Height,
int *sum,
unsigned int *sumsquared
);
void vp8_half_horiz_vert_variance16x_h_sse2
(
const unsigned char *ref_ptr,
int ref_pixels_per_line,
const unsigned char *src_ptr,
int src_pixels_per_line,
unsigned int Height,
int *sum,
unsigned int *sumsquared
);
void vp8_half_horiz_variance8x_h_sse2
(
const unsigned char *ref_ptr,
int ref_pixels_per_line,
const unsigned char *src_ptr,
int src_pixels_per_line,
unsigned int Height,
int *sum,
unsigned int *sumsquared
);
void vp8_half_horiz_variance16x_h_sse2
(
const unsigned char *ref_ptr,
int ref_pixels_per_line,
const unsigned char *src_ptr,
int src_pixels_per_line,
unsigned int Height,
int *sum,
unsigned int *sumsquared
);
void vp8_half_vert_variance8x_h_sse2
(
const unsigned char *ref_ptr,
int ref_pixels_per_line,
const unsigned char *src_ptr,
int src_pixels_per_line,
unsigned int Height,
int *sum,
unsigned int *sumsquared
);
void vp8_half_vert_variance16x_h_sse2
(
const unsigned char *ref_ptr,
int ref_pixels_per_line,
const unsigned char *src_ptr,
int src_pixels_per_line,
unsigned int Height,
int *sum,
unsigned int *sumsquared
);
#if CONFIG_SIXTEENTH_SUBPEL_UV
DECLARE_ALIGNED(16, extern short, vp8_vp7_bilinear_filters_mmx[16][8]);
#else
DECLARE_ALIGNED(16, extern short, vp8_vp7_bilinear_filters_mmx[8][8]);
#endif
unsigned int vp8_variance4x4_wmt(
const unsigned char *src_ptr,
int source_stride,
const unsigned char *ref_ptr,
int recon_stride,
unsigned int *sse)
{
unsigned int var;
int avg;
vp8_get4x4var_mmx(src_ptr, source_stride, ref_ptr, recon_stride, &var, &avg) ;
*sse = var;
return (var - ((avg * avg) >> 4));
}
unsigned int vp8_variance8x8_wmt
(
const unsigned char *src_ptr,
int source_stride,
const unsigned char *ref_ptr,
int recon_stride,
unsigned int *sse)
{
unsigned int var;
int avg;
vp8_get8x8var_sse2(src_ptr, source_stride, ref_ptr, recon_stride, &var, &avg) ;
*sse = var;
return (var - ((avg * avg) >> 6));
}
unsigned int vp8_variance16x16_wmt
(
const unsigned char *src_ptr,
int source_stride,
const unsigned char *ref_ptr,
int recon_stride,
unsigned int *sse)
{
unsigned int sse0;
int sum0;
vp8_get16x16var_sse2(src_ptr, source_stride, ref_ptr, recon_stride, &sse0, &sum0) ;
*sse = sse0;
return (sse0 - ((sum0 * sum0) >> 8));
}
unsigned int vp8_mse16x16_wmt(
const unsigned char *src_ptr,
int source_stride,
const unsigned char *ref_ptr,
int recon_stride,
unsigned int *sse)
{
unsigned int sse0;
int sum0;
vp8_get16x16var_sse2(src_ptr, source_stride, ref_ptr, recon_stride, &sse0, &sum0) ;
*sse = sse0;
return sse0;
}
unsigned int vp8_variance16x8_wmt
(
const unsigned char *src_ptr,
int source_stride,
const unsigned char *ref_ptr,
int recon_stride,
unsigned int *sse)
{
unsigned int sse0, sse1, var;
int sum0, sum1, avg;
vp8_get8x8var_sse2(src_ptr, source_stride, ref_ptr, recon_stride, &sse0, &sum0) ;
vp8_get8x8var_sse2(src_ptr + 8, source_stride, ref_ptr + 8, recon_stride, &sse1, &sum1);
var = sse0 + sse1;
avg = sum0 + sum1;
*sse = var;
return (var - ((avg * avg) >> 7));
}
unsigned int vp8_variance8x16_wmt
(
const unsigned char *src_ptr,
int source_stride,
const unsigned char *ref_ptr,
int recon_stride,
unsigned int *sse)
{
unsigned int sse0, sse1, var;
int sum0, sum1, avg;
vp8_get8x8var_sse2(src_ptr, source_stride, ref_ptr, recon_stride, &sse0, &sum0) ;
vp8_get8x8var_sse2(src_ptr + 8 * source_stride, source_stride, ref_ptr + 8 * recon_stride, recon_stride, &sse1, &sum1) ;
var = sse0 + sse1;
avg = sum0 + sum1;
*sse = var;
return (var - ((avg * avg) >> 7));
}
unsigned int vp8_sub_pixel_variance4x4_wmt
(
const unsigned char *src_ptr,
int src_pixels_per_line,
int xoffset,
int yoffset,
const unsigned char *dst_ptr,
int dst_pixels_per_line,
unsigned int *sse
)
{
int xsum;
unsigned int xxsum;
vp8_filter_block2d_bil4x4_var_mmx(
src_ptr, src_pixels_per_line,
dst_ptr, dst_pixels_per_line,
vp8_vp7_bilinear_filters_mmx[xoffset], vp8_vp7_bilinear_filters_mmx[yoffset],
&xsum, &xxsum
);
*sse = xxsum;
return (xxsum - ((xsum * xsum) >> 4));
}
unsigned int vp8_sub_pixel_variance8x8_wmt
(
const unsigned char *src_ptr,
int src_pixels_per_line,
int xoffset,
int yoffset,
const unsigned char *dst_ptr,
int dst_pixels_per_line,
unsigned int *sse
)
{
int xsum;
unsigned int xxsum;
if (xoffset == HALFNDX && yoffset == 0)
{
vp8_half_horiz_variance8x_h_sse2(
src_ptr, src_pixels_per_line,
dst_ptr, dst_pixels_per_line, 8,
&xsum, &xxsum);
}
else if (xoffset == 0 && yoffset == HALFNDX)
{
vp8_half_vert_variance8x_h_sse2(
src_ptr, src_pixels_per_line,
dst_ptr, dst_pixels_per_line, 8,
&xsum, &xxsum);
}
else if (xoffset == HALFNDX && yoffset == HALFNDX)
{
vp8_half_horiz_vert_variance8x_h_sse2(
src_ptr, src_pixels_per_line,
dst_ptr, dst_pixels_per_line, 8,
&xsum, &xxsum);
}
else
{
vp8_filter_block2d_bil_var_sse2(
src_ptr, src_pixels_per_line,
dst_ptr, dst_pixels_per_line, 8,
xoffset, yoffset,
&xsum, &xxsum);
}
*sse = xxsum;
return (xxsum - ((xsum * xsum) >> 6));
}
unsigned int vp8_sub_pixel_variance16x16_wmt
(
const unsigned char *src_ptr,
int src_pixels_per_line,
int xoffset,
int yoffset,
const unsigned char *dst_ptr,
int dst_pixels_per_line,
unsigned int *sse
)
{
int xsum0, xsum1;
unsigned int xxsum0, xxsum1;
// note we could avoid these if statements if the calling function
// just called the appropriate functions inside.
if (xoffset == HALFNDX && yoffset == 0)
{
vp8_half_horiz_variance16x_h_sse2(
src_ptr, src_pixels_per_line,
dst_ptr, dst_pixels_per_line, 16,
&xsum0, &xxsum0);
}
else if (xoffset == 0 && yoffset == HALFNDX)
{
vp8_half_vert_variance16x_h_sse2(
src_ptr, src_pixels_per_line,
dst_ptr, dst_pixels_per_line, 16,
&xsum0, &xxsum0);
}
else if (xoffset == HALFNDX && yoffset == HALFNDX)
{
vp8_half_horiz_vert_variance16x_h_sse2(
src_ptr, src_pixels_per_line,
dst_ptr, dst_pixels_per_line, 16,
&xsum0, &xxsum0);
}
else
{
vp8_filter_block2d_bil_var_sse2(
src_ptr, src_pixels_per_line,
dst_ptr, dst_pixels_per_line, 16,
xoffset, yoffset,
&xsum0, &xxsum0
);
vp8_filter_block2d_bil_var_sse2(
src_ptr + 8, src_pixels_per_line,
dst_ptr + 8, dst_pixels_per_line, 16,
xoffset, yoffset,
&xsum1, &xxsum1
);
xsum0 += xsum1;
xxsum0 += xxsum1;
}
*sse = xxsum0;
return (xxsum0 - ((xsum0 * xsum0) >> 8));
}
unsigned int vp8_sub_pixel_mse16x16_wmt(
const unsigned char *src_ptr,
int src_pixels_per_line,
int xoffset,
int yoffset,
const unsigned char *dst_ptr,
int dst_pixels_per_line,
unsigned int *sse
)
{
vp8_sub_pixel_variance16x16_wmt(src_ptr, src_pixels_per_line, xoffset, yoffset, dst_ptr, dst_pixels_per_line, sse);
return *sse;
}
unsigned int vp8_sub_pixel_variance16x8_wmt
(
const unsigned char *src_ptr,
int src_pixels_per_line,
int xoffset,
int yoffset,
const unsigned char *dst_ptr,
int dst_pixels_per_line,
unsigned int *sse
)
{
int xsum0, xsum1;
unsigned int xxsum0, xxsum1;
if (xoffset == HALFNDX && yoffset == 0)
{
vp8_half_horiz_variance16x_h_sse2(
src_ptr, src_pixels_per_line,
dst_ptr, dst_pixels_per_line, 8,
&xsum0, &xxsum0);
}
else if (xoffset == 0 && yoffset == HALFNDX)
{
vp8_half_vert_variance16x_h_sse2(
src_ptr, src_pixels_per_line,
dst_ptr, dst_pixels_per_line, 8,
&xsum0, &xxsum0);
}
else if (xoffset == HALFNDX && yoffset == HALFNDX)
{
vp8_half_horiz_vert_variance16x_h_sse2(
src_ptr, src_pixels_per_line,
dst_ptr, dst_pixels_per_line, 8,
&xsum0, &xxsum0);
}
else
{
vp8_filter_block2d_bil_var_sse2(
src_ptr, src_pixels_per_line,
dst_ptr, dst_pixels_per_line, 8,
xoffset, yoffset,
&xsum0, &xxsum0);
vp8_filter_block2d_bil_var_sse2(
src_ptr + 8, src_pixels_per_line,
dst_ptr + 8, dst_pixels_per_line, 8,
xoffset, yoffset,
&xsum1, &xxsum1);
xsum0 += xsum1;
xxsum0 += xxsum1;
}
*sse = xxsum0;
return (xxsum0 - ((xsum0 * xsum0) >> 7));
}
unsigned int vp8_sub_pixel_variance8x16_wmt
(
const unsigned char *src_ptr,
int src_pixels_per_line,
int xoffset,
int yoffset,
const unsigned char *dst_ptr,
int dst_pixels_per_line,
unsigned int *sse
)
{
int xsum;
unsigned int xxsum;
if (xoffset == HALFNDX && yoffset == 0)
{
vp8_half_horiz_variance8x_h_sse2(
src_ptr, src_pixels_per_line,
dst_ptr, dst_pixels_per_line, 16,
&xsum, &xxsum);
}
else if (xoffset == 0 && yoffset == HALFNDX)
{
vp8_half_vert_variance8x_h_sse2(
src_ptr, src_pixels_per_line,
dst_ptr, dst_pixels_per_line, 16,
&xsum, &xxsum);
}
else if (xoffset == HALFNDX && yoffset == HALFNDX)
{
vp8_half_horiz_vert_variance8x_h_sse2(
src_ptr, src_pixels_per_line,
dst_ptr, dst_pixels_per_line, 16,
&xsum, &xxsum);
}
else
{
vp8_filter_block2d_bil_var_sse2(
src_ptr, src_pixels_per_line,
dst_ptr, dst_pixels_per_line, 16,
xoffset, yoffset,
&xsum, &xxsum);
}
*sse = xxsum;
return (xxsum - ((xsum * xsum) >> 7));
}
unsigned int vp8_variance_halfpixvar16x16_h_wmt(
const unsigned char *src_ptr,
int src_pixels_per_line,
const unsigned char *dst_ptr,
int dst_pixels_per_line,
unsigned int *sse)
{
int xsum0;
unsigned int xxsum0;
vp8_half_horiz_variance16x_h_sse2(
src_ptr, src_pixels_per_line,
dst_ptr, dst_pixels_per_line, 16,
&xsum0, &xxsum0);
*sse = xxsum0;
return (xxsum0 - ((xsum0 * xsum0) >> 8));
}
unsigned int vp8_variance_halfpixvar16x16_v_wmt(
const unsigned char *src_ptr,
int src_pixels_per_line,
const unsigned char *dst_ptr,
int dst_pixels_per_line,
unsigned int *sse)
{
int xsum0;
unsigned int xxsum0;
vp8_half_vert_variance16x_h_sse2(
src_ptr, src_pixels_per_line,
dst_ptr, dst_pixels_per_line, 16,
&xsum0, &xxsum0);
*sse = xxsum0;
return (xxsum0 - ((xsum0 * xsum0) >> 8));
}
unsigned int vp8_variance_halfpixvar16x16_hv_wmt(
const unsigned char *src_ptr,
int src_pixels_per_line,
const unsigned char *dst_ptr,
int dst_pixels_per_line,
unsigned int *sse)
{
int xsum0;
unsigned int xxsum0;
vp8_half_horiz_vert_variance16x_h_sse2(
src_ptr, src_pixels_per_line,
dst_ptr, dst_pixels_per_line, 16,
&xsum0, &xxsum0);
*sse = xxsum0;
return (xxsum0 - ((xsum0 * xsum0) >> 8));
}