An error occurred while loading the file. Please try again.
-
John Koleszar authored
This patch fixes the system dependent entries for the half-pixel variance functions in both the RTCD and non-RTCD cases: - The generic C versions of these functions are now correct. Before all three cases called the hv code. - Wire up the ARM functions in RTCD mode - Created stubs for x86 to call the optimized subpixel functions with the correct parameters, rather than falling back to C code. Change-Id: I1d937d074d929e0eb93aacb1232cc5e0ad1c6184a0ae3682
variance_c.c 15.63 KiB
/* * 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 "variance.h"const int vp8_six_tap[8][6] ={ { 0, 0, 128, 0, 0, 0 }, // note that 1/8 pel positions are just as per alpha -0.5 bicubic { 0, -6, 123, 12, -1, 0 }, { 2, -11, 108, 36, -8, 1 }, // New 1/4 pel 6 tap filter { 0, -9, 93, 50, -6, 0 }, { 3, -16, 77, 77, -16, 3 }, // New 1/2 pel 6 tap filter { 0, -6, 50, 93, -9, 0 }, { 1, -8, 36, 108, -11, 2 }, // New 1/4 pel 6 tap filter { 0, -1, 12, 123, -6, 0 }};const int VP8_FILTER_WEIGHT = 128;const int VP8_FILTER_SHIFT = 7;const int vp8_bilinear_taps[8][2] ={ { 128, 0 }, { 112, 16 }, { 96, 32 }, { 80, 48 }, { 64, 64 }, { 48, 80 }, { 32, 96 }, { 16, 112 }};unsigned int vp8_get_mb_ss_c( const short *src_ptr){ unsigned int i = 0, sum = 0; do { sum += (src_ptr[i] * src_ptr[i]); i++; } while (i < 256); return sum;}void vp8_variance( const unsigned char *src_ptr, int source_stride, const unsigned char *ref_ptr, int recon_stride, int w, int h, unsigned int *sse, int *sum){ int i, j; int diff;7172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140
*sum = 0;
*sse = 0;
for (i = 0; i < h; i++)
{
for (j = 0; j < w; j++)
{
diff = src_ptr[j] - ref_ptr[j];
*sum += diff;
*sse += diff * diff;
}
src_ptr += source_stride;
ref_ptr += recon_stride;
}
}
unsigned int
vp8_get8x8var_c
(
const unsigned char *src_ptr,
int source_stride,
const unsigned char *ref_ptr,
int recon_stride,
unsigned int *SSE,
int *Sum
)
{
vp8_variance(src_ptr, source_stride, ref_ptr, recon_stride, 8, 8, SSE, Sum);
return (*SSE - (((*Sum) * (*Sum)) >> 6));
}
unsigned int
vp8_get16x16var_c
(
const unsigned char *src_ptr,
int source_stride,
const unsigned char *ref_ptr,
int recon_stride,
unsigned int *SSE,
int *Sum
)
{
vp8_variance(src_ptr, source_stride, ref_ptr, recon_stride, 16, 16, SSE, Sum);
return (*SSE - (((*Sum) * (*Sum)) >> 8));
}
unsigned int vp8_variance16x16_c(
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_variance(src_ptr, source_stride, ref_ptr, recon_stride, 16, 16, &var, &avg);
*sse = var;
return (var - ((avg * avg) >> 8));
}
unsigned int vp8_variance8x16_c(
141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210
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_variance(src_ptr, source_stride, ref_ptr, recon_stride, 8, 16, &var, &avg);
*sse = var;
return (var - ((avg * avg) >> 7));
}
unsigned int vp8_variance16x8_c(
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_variance(src_ptr, source_stride, ref_ptr, recon_stride, 16, 8, &var, &avg);
*sse = var;
return (var - ((avg * avg) >> 7));
}
unsigned int vp8_variance8x8_c(
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_variance(src_ptr, source_stride, ref_ptr, recon_stride, 8, 8, &var, &avg);
*sse = var;
return (var - ((avg * avg) >> 6));
}
unsigned int vp8_variance4x4_c(
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_variance(src_ptr, source_stride, ref_ptr, recon_stride, 4, 4, &var, &avg);
*sse = var;
return (var - ((avg * avg) >> 4));
}
unsigned int vp8_mse16x16_c(
const unsigned char *src_ptr,
int source_stride,
const unsigned char *ref_ptr,
int recon_stride,
211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280
unsigned int *sse)
{
unsigned int var;
int avg;
vp8_variance(src_ptr, source_stride, ref_ptr, recon_stride, 16, 16, &var, &avg);
*sse = var;
return var;
}
/****************************************************************************
*
* ROUTINE : filter_block2d_bil_first_pass
*
* INPUTS : UINT8 *src_ptr : Pointer to source block.
* UINT32 src_pixels_per_line : Stride of input block.
* UINT32 pixel_step : Offset between filter input samples (see notes).
* UINT32 output_height : Input block height.
* UINT32 output_width : Input block width.
* INT32 *vp8_filter : Array of 2 bi-linear filter taps.
*
* OUTPUTS : INT32 *output_ptr : Pointer to filtered block.
*
* RETURNS : void
*
* FUNCTION : Applies a 1-D 2-tap bi-linear filter to the source block in
* either horizontal or vertical direction to produce the
* filtered output block. Used to implement first-pass
* of 2-D separable filter.
*
* SPECIAL NOTES : Produces INT32 output to retain precision for next pass.
* Two filter taps should sum to VP8_FILTER_WEIGHT.
* pixel_step defines whether the filter is applied
* horizontally (pixel_step=1) or vertically (pixel_step=stride).
* It defines the offset required to move from one input
* to the next.
*
****************************************************************************/
void vp8e_filter_block2d_bil_first_pass
(
const unsigned char *src_ptr,
unsigned short *output_ptr,
unsigned int src_pixels_per_line,
int pixel_step,
unsigned int output_height,
unsigned int output_width,
const int *vp8_filter
)
{
unsigned int i, j;
for (i = 0; i < output_height; i++)
{
for (j = 0; j < output_width; j++)
{
// Apply bilinear filter
output_ptr[j] = (((int)src_ptr[0] * vp8_filter[0]) +
((int)src_ptr[pixel_step] * vp8_filter[1]) +
(VP8_FILTER_WEIGHT / 2)) >> VP8_FILTER_SHIFT;
src_ptr++;
}
// Next row...
src_ptr += src_pixels_per_line - output_width;
output_ptr += output_width;
}
}
/****************************************************************************
281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350
*
* ROUTINE : filter_block2d_bil_second_pass
*
* INPUTS : INT32 *src_ptr : Pointer to source block.
* UINT32 src_pixels_per_line : Stride of input block.
* UINT32 pixel_step : Offset between filter input samples (see notes).
* UINT32 output_height : Input block height.
* UINT32 output_width : Input block width.
* INT32 *vp8_filter : Array of 2 bi-linear filter taps.
*
* OUTPUTS : UINT16 *output_ptr : Pointer to filtered block.
*
* RETURNS : void
*
* FUNCTION : Applies a 1-D 2-tap bi-linear filter to the source block in
* either horizontal or vertical direction to produce the
* filtered output block. Used to implement second-pass
* of 2-D separable filter.
*
* SPECIAL NOTES : Requires 32-bit input as produced by filter_block2d_bil_first_pass.
* Two filter taps should sum to VP8_FILTER_WEIGHT.
* pixel_step defines whether the filter is applied
* horizontally (pixel_step=1) or vertically (pixel_step=stride).
* It defines the offset required to move from one input
* to the next.
*
****************************************************************************/
void vp8e_filter_block2d_bil_second_pass
(
const unsigned short *src_ptr,
unsigned char *output_ptr,
unsigned int src_pixels_per_line,
unsigned int pixel_step,
unsigned int output_height,
unsigned int output_width,
const int *vp8_filter
)
{
unsigned int i, j;
int Temp;
for (i = 0; i < output_height; i++)
{
for (j = 0; j < output_width; j++)
{
// Apply filter
Temp = ((int)src_ptr[0] * vp8_filter[0]) +
((int)src_ptr[pixel_step] * vp8_filter[1]) +
(VP8_FILTER_WEIGHT / 2);
output_ptr[j] = (unsigned int)(Temp >> VP8_FILTER_SHIFT);
src_ptr++;
}
// Next row...
src_ptr += src_pixels_per_line - output_width;
output_ptr += output_width;
}
}
/****************************************************************************
*
* ROUTINE : filter_block2d_bil
*
* INPUTS : UINT8 *src_ptr : Pointer to source block.
* UINT32 src_pixels_per_line : Stride of input block.
* INT32 *HFilter : Array of 2 horizontal filter taps.
* INT32 *VFilter : Array of 2 vertical filter taps.
*
* OUTPUTS : UINT16 *output_ptr : Pointer to filtered block.
351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420
*
* RETURNS : void
*
* FUNCTION : 2-D filters an 8x8 input block by applying a 2-tap
* bi-linear filter horizontally followed by a 2-tap
* bi-linear filter vertically on the result.
*
* SPECIAL NOTES : The intermediate horizontally filtered block must produce
* 1 more point than the input block in each column. This
* is to ensure that the 2-tap filter has one extra data-point
* at the top of each column so filter taps do not extend
* beyond data. Thus the output of the first stage filter
* is an 8x9 (hx_v) block.
*
****************************************************************************/
void vp8e_filter_block2d_bil
(
const unsigned char *src_ptr,
unsigned char *output_ptr,
unsigned int src_pixels_per_line,
int *HFilter,
int *VFilter
)
{
unsigned short FData[20*16]; // Temp data bufffer used in filtering
// First filter 1-D horizontally...
vp8e_filter_block2d_bil_first_pass(src_ptr, FData, src_pixels_per_line, 1, 9, 8, HFilter);
// then 1-D vertically...
vp8e_filter_block2d_bil_second_pass(FData, output_ptr, 8, 8, 8, 8, VFilter);
}
unsigned int vp8_sub_pixel_variance4x4_c
(
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
)
{
unsigned char temp2[20*16];
const int *HFilter, *VFilter;
unsigned short FData3[5*4]; // Temp data bufffer used in filtering
HFilter = vp8_bilinear_taps[xoffset];
VFilter = vp8_bilinear_taps[yoffset];
// First filter 1d Horizontal
vp8e_filter_block2d_bil_first_pass(src_ptr, FData3, src_pixels_per_line, 1, 5, 4, HFilter);
// Now filter Verticaly
vp8e_filter_block2d_bil_second_pass(FData3, temp2, 4, 4, 4, 4, VFilter);
return vp8_variance4x4_c(temp2, 4, dst_ptr, dst_pixels_per_line, sse);
}
unsigned int vp8_sub_pixel_variance8x8_c
(
const unsigned char *src_ptr,
int src_pixels_per_line,
int xoffset,
int yoffset,
421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490
const unsigned char *dst_ptr,
int dst_pixels_per_line,
unsigned int *sse
)
{
unsigned short FData3[9*8]; // Temp data bufffer used in filtering
unsigned char temp2[20*16];
const int *HFilter, *VFilter;
HFilter = vp8_bilinear_taps[xoffset];
VFilter = vp8_bilinear_taps[yoffset];
vp8e_filter_block2d_bil_first_pass(src_ptr, FData3, src_pixels_per_line, 1, 9, 8, HFilter);
vp8e_filter_block2d_bil_second_pass(FData3, temp2, 8, 8, 8, 8, VFilter);
return vp8_variance8x8_c(temp2, 8, dst_ptr, dst_pixels_per_line, sse);
}
unsigned int vp8_sub_pixel_variance16x16_c
(
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
)
{
unsigned short FData3[17*16]; // Temp data bufffer used in filtering
unsigned char temp2[20*16];
const int *HFilter, *VFilter;
HFilter = vp8_bilinear_taps[xoffset];
VFilter = vp8_bilinear_taps[yoffset];
vp8e_filter_block2d_bil_first_pass(src_ptr, FData3, src_pixels_per_line, 1, 17, 16, HFilter);
vp8e_filter_block2d_bil_second_pass(FData3, temp2, 16, 16, 16, 16, VFilter);
return vp8_variance16x16_c(temp2, 16, dst_ptr, dst_pixels_per_line, sse);
}
unsigned int vp8_variance_halfpixvar16x16_h_c(
const unsigned char *src_ptr,
int source_stride,
const unsigned char *ref_ptr,
int recon_stride,
unsigned int *sse)
{
return vp8_sub_pixel_variance16x16_c(src_ptr, source_stride, 4, 0,
ref_ptr, recon_stride, sse);
}
unsigned int vp8_variance_halfpixvar16x16_v_c(
const unsigned char *src_ptr,
int source_stride,
const unsigned char *ref_ptr,
int recon_stride,
unsigned int *sse)
{
return vp8_sub_pixel_variance16x16_c(src_ptr, source_stride, 0, 4,
ref_ptr, recon_stride, sse);
}
unsigned int vp8_variance_halfpixvar16x16_hv_c(
const unsigned char *src_ptr,
int source_stride,
491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521522523524525526527528529530531532533534535536537538539540541542543544545546547548549550551552553554555556557558559560
const unsigned char *ref_ptr,
int recon_stride,
unsigned int *sse)
{
return vp8_sub_pixel_variance16x16_c(src_ptr, source_stride, 4, 4,
ref_ptr, recon_stride, sse);
}
unsigned int vp8_sub_pixel_mse16x16_c
(
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_c(src_ptr, src_pixels_per_line, xoffset, yoffset, dst_ptr, dst_pixels_per_line, sse);
return *sse;
}
unsigned int vp8_sub_pixel_variance16x8_c
(
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
)
{
unsigned short FData3[16*9]; // Temp data bufffer used in filtering
unsigned char temp2[20*16];
const int *HFilter, *VFilter;
HFilter = vp8_bilinear_taps[xoffset];
VFilter = vp8_bilinear_taps[yoffset];
vp8e_filter_block2d_bil_first_pass(src_ptr, FData3, src_pixels_per_line, 1, 9, 16, HFilter);
vp8e_filter_block2d_bil_second_pass(FData3, temp2, 16, 16, 8, 16, VFilter);
return vp8_variance16x8_c(temp2, 16, dst_ptr, dst_pixels_per_line, sse);
}
unsigned int vp8_sub_pixel_variance8x16_c
(
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
)
{
unsigned short FData3[9*16]; // Temp data bufffer used in filtering
unsigned char temp2[20*16];
const int *HFilter, *VFilter;
HFilter = vp8_bilinear_taps[xoffset];
VFilter = vp8_bilinear_taps[yoffset];
vp8e_filter_block2d_bil_first_pass(src_ptr, FData3, src_pixels_per_line, 1, 17, 8, HFilter);
vp8e_filter_block2d_bil_second_pass(FData3, temp2, 8, 8, 16, 8, VFilter);
561562563564
return vp8_variance8x16_c(temp2, 8, dst_ptr, dst_pixels_per_line, sse);
}