2010-05-18 17:58:33 +02:00
|
|
|
/*
|
2010-09-09 14:16:39 +02:00
|
|
|
* Copyright (c) 2010 The WebM project authors. All Rights Reserved.
|
2010-05-18 17:58:33 +02:00
|
|
|
*
|
2010-06-18 18:39:21 +02:00
|
|
|
* Use of this source code is governed by a BSD-style license
|
2010-06-04 22:19:40 +02:00
|
|
|
* that can be found in the LICENSE file in the root of the source
|
|
|
|
|
* tree. An additional intellectual property rights grant can be found
|
2010-06-18 18:39:21 +02:00
|
|
|
* in the file PATENTS. All contributing project authors may
|
2010-06-04 22:19:40 +02:00
|
|
|
* be found in the AUTHORS file in the root of the source tree.
|
2010-05-18 17:58:33 +02:00
|
|
|
*/
|
|
|
|
|
|
|
|
|
|
|
WebM Experimental Codec Branch Snapshot
This is a code snapshot of experimental work currently ongoing for a
next-generation codec.
The codebase has been cut down considerably from the libvpx baseline.
For example, we are currently only supporting VBR 2-pass rate control
and have removed most of the code relating to coding speed, threading,
error resilience, partitions and various other features. This is in
part to make the codebase easier to work on and experiment with, but
also because we want to have an open discussion about how the bitstream
will be structured and partitioned and not have that conversation
constrained by past work.
Our basic working pattern has been to initially encapsulate experiments
using configure options linked to #IF CONFIG_XXX statements in the
code. Once experiments have matured and we are reasonably happy that
they give benefit and can be merged without breaking other experiments,
we remove the conditional compile statements and merge them in.
Current changes include:
* Temporal coding experiment for segments (though still only 4 max, it
will likely be increased).
* Segment feature experiment - to allow various bits of information to
be coded at the segment level. Features tested so far include mode
and reference frame information, limiting end of block offset and
transform size, alongside Q and loop filter parameters, but this set
is very fluid.
* Support for 8x8 transform - 8x8 dct with 2nd order 2x2 haar is used
in MBs using 16x16 prediction modes within inter frames.
* Compound prediction (combination of signals from existing predictors
to create a new predictor).
* 8 tap interpolation filters and 1/8th pel motion vectors.
* Loop filter modifications.
* Various entropy modifications and changes to how entropy contexts and
updates are handled.
* Extended quantizer range matched to transform precision improvements.
There are also ongoing further experiments that we hope to merge in the
near future: For example, coding of motion and other aspects of the
prediction signal to better support larger image formats, use of larger
block sizes (e.g. 32x32 and up) and lossless non-transform based coding
options (especially for key frames). It is our hope that we will be
able to make regular updates and we will warmly welcome community
contributions.
Please be warned that, at this stage, the codebase is currently slower
than VP8 stable branch as most new code has not been optimized, and
even the 'C' has been deliberately written to be simple and obvious,
not fast.
The following graphs have the initial test results, numbers in the
tables measure the compression improvement in terms of percentage. The
build has the following optional experiments configured:
--enable-experimental --enable-enhanced_interp --enable-uvintra
--enable-high_precision_mv --enable-sixteenth_subpel_uv
CIF Size clips:
http://getwebm.org/tmp/cif/
HD size clips:
http://getwebm.org/tmp/hd/
(stable_20120309 represents encoding results of WebM master branch
build as of commit#7a15907)
They were encoded using the following encode parameters:
--good --cpu-used=0 -t 0 --lag-in-frames=25 --min-q=0 --max-q=63
--end-usage=0 --auto-alt-ref=1 -p 2 --pass=2 --kf-max-dist=9999
--kf-min-dist=0 --drop-frame=0 --static-thresh=0 --bias-pct=50
--minsection-pct=0 --maxsection-pct=800 --sharpness=0
--arnr-maxframes=7 --arnr-strength=3(for HD,6 for CIF)
--arnr-type=3
Change-Id: I5c62ed09cfff5815a2bb34e7820d6a810c23183c
2012-03-10 02:32:50 +01:00
|
|
|
#include "vpx_ports/config.h"
|
2010-05-18 17:58:33 +02:00
|
|
|
#include "vpx_ports/mem.h"
|
2012-11-27 22:59:17 +01:00
|
|
|
#include "vp9/common/vp9_subpixel.h"
|
2010-05-18 17:58:33 +02:00
|
|
|
|
2012-10-31 00:25:53 +01:00
|
|
|
extern const short vp9_six_tap_mmx[16][6 * 8];
|
2012-11-01 00:09:17 +01:00
|
|
|
|
2012-10-31 01:12:12 +01:00
|
|
|
extern const short vp9_bilinear_filters_8x_mmx[16][2 * 8];
|
2010-05-18 17:58:33 +02:00
|
|
|
|
2012-11-01 00:09:17 +01:00
|
|
|
extern void vp9_filter_block1d_h6_mmx(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,
|
|
|
|
|
const short *vp9_filter);
|
|
|
|
|
|
|
|
|
|
extern void vp9_filter_block1dc_v6_mmx(unsigned short *src_ptr,
|
|
|
|
|
unsigned char *output_ptr,
|
|
|
|
|
int output_pitch,
|
|
|
|
|
unsigned int pixels_per_line,
|
|
|
|
|
unsigned int pixel_step,
|
|
|
|
|
unsigned int output_height,
|
|
|
|
|
unsigned int output_width,
|
|
|
|
|
const short *vp9_filter);
|
|
|
|
|
|
|
|
|
|
extern void vp9_filter_block1d8_h6_sse2(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,
|
|
|
|
|
const short *vp9_filter);
|
|
|
|
|
|
|
|
|
|
extern void vp9_filter_block1d16_h6_sse2(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,
|
|
|
|
|
const short *vp9_filter);
|
|
|
|
|
|
|
|
|
|
extern void vp9_filter_block1d8_v6_sse2(unsigned short *src_ptr,
|
|
|
|
|
unsigned char *output_ptr,
|
|
|
|
|
int dst_ptich,
|
|
|
|
|
unsigned int pixels_per_line,
|
|
|
|
|
unsigned int pixel_step,
|
|
|
|
|
unsigned int output_height,
|
|
|
|
|
unsigned int output_width,
|
|
|
|
|
const short *vp9_filter);
|
|
|
|
|
|
|
|
|
|
extern void vp9_filter_block1d16_v6_sse2(unsigned short *src_ptr,
|
|
|
|
|
unsigned char *output_ptr,
|
|
|
|
|
int dst_ptich,
|
|
|
|
|
unsigned int pixels_per_line,
|
|
|
|
|
unsigned int pixel_step,
|
|
|
|
|
unsigned int output_height,
|
|
|
|
|
unsigned int output_width,
|
|
|
|
|
const short *vp9_filter);
|
|
|
|
|
|
|
|
|
|
extern void vp9_unpack_block1d16_h6_sse2(unsigned char *src_ptr,
|
|
|
|
|
unsigned short *output_ptr,
|
|
|
|
|
unsigned int src_pixels_per_line,
|
|
|
|
|
unsigned int output_height,
|
|
|
|
|
unsigned int output_width);
|
|
|
|
|
|
|
|
|
|
extern void vp9_filter_block1d8_h6_only_sse2(unsigned char *src_ptr,
|
|
|
|
|
unsigned int src_pixels_per_line,
|
|
|
|
|
unsigned char *output_ptr,
|
|
|
|
|
int dst_pitch,
|
|
|
|
|
unsigned int output_height,
|
|
|
|
|
const short *vp9_filter);
|
|
|
|
|
|
|
|
|
|
extern void vp9_filter_block1d16_h6_only_sse2(unsigned char *src_ptr,
|
|
|
|
|
unsigned int src_pixels_per_lin,
|
|
|
|
|
unsigned char *output_ptr,
|
|
|
|
|
int dst_pitch,
|
|
|
|
|
unsigned int output_height,
|
|
|
|
|
const short *vp9_filter);
|
|
|
|
|
|
|
|
|
|
extern void vp9_filter_block1d8_v6_only_sse2(unsigned char *src_ptr,
|
|
|
|
|
unsigned int src_pixels_per_line,
|
|
|
|
|
unsigned char *output_ptr,
|
|
|
|
|
int dst_pitch,
|
|
|
|
|
unsigned int output_height,
|
|
|
|
|
const short *vp9_filter);
|
2010-05-18 17:58:33 +02:00
|
|
|
|
2012-11-01 00:09:17 +01:00
|
|
|
extern prototype_subpixel_predict(vp9_bilinear_predict8x8_mmx);
|
2010-05-18 17:58:33 +02:00
|
|
|
|
|
|
|
|
#if HAVE_MMX
|
2012-11-01 00:09:17 +01:00
|
|
|
void vp9_sixtap_predict4x4_mmx(unsigned char *src_ptr,
|
|
|
|
|
int src_pixels_per_line,
|
|
|
|
|
int xoffset,
|
|
|
|
|
int yoffset,
|
|
|
|
|
unsigned char *dst_ptr,
|
|
|
|
|
int dst_pitch) {
|
WebM Experimental Codec Branch Snapshot
This is a code snapshot of experimental work currently ongoing for a
next-generation codec.
The codebase has been cut down considerably from the libvpx baseline.
For example, we are currently only supporting VBR 2-pass rate control
and have removed most of the code relating to coding speed, threading,
error resilience, partitions and various other features. This is in
part to make the codebase easier to work on and experiment with, but
also because we want to have an open discussion about how the bitstream
will be structured and partitioned and not have that conversation
constrained by past work.
Our basic working pattern has been to initially encapsulate experiments
using configure options linked to #IF CONFIG_XXX statements in the
code. Once experiments have matured and we are reasonably happy that
they give benefit and can be merged without breaking other experiments,
we remove the conditional compile statements and merge them in.
Current changes include:
* Temporal coding experiment for segments (though still only 4 max, it
will likely be increased).
* Segment feature experiment - to allow various bits of information to
be coded at the segment level. Features tested so far include mode
and reference frame information, limiting end of block offset and
transform size, alongside Q and loop filter parameters, but this set
is very fluid.
* Support for 8x8 transform - 8x8 dct with 2nd order 2x2 haar is used
in MBs using 16x16 prediction modes within inter frames.
* Compound prediction (combination of signals from existing predictors
to create a new predictor).
* 8 tap interpolation filters and 1/8th pel motion vectors.
* Loop filter modifications.
* Various entropy modifications and changes to how entropy contexts and
updates are handled.
* Extended quantizer range matched to transform precision improvements.
There are also ongoing further experiments that we hope to merge in the
near future: For example, coding of motion and other aspects of the
prediction signal to better support larger image formats, use of larger
block sizes (e.g. 32x32 and up) and lossless non-transform based coding
options (especially for key frames). It is our hope that we will be
able to make regular updates and we will warmly welcome community
contributions.
Please be warned that, at this stage, the codebase is currently slower
than VP8 stable branch as most new code has not been optimized, and
even the 'C' has been deliberately written to be simple and obvious,
not fast.
The following graphs have the initial test results, numbers in the
tables measure the compression improvement in terms of percentage. The
build has the following optional experiments configured:
--enable-experimental --enable-enhanced_interp --enable-uvintra
--enable-high_precision_mv --enable-sixteenth_subpel_uv
CIF Size clips:
http://getwebm.org/tmp/cif/
HD size clips:
http://getwebm.org/tmp/hd/
(stable_20120309 represents encoding results of WebM master branch
build as of commit#7a15907)
They were encoded using the following encode parameters:
--good --cpu-used=0 -t 0 --lag-in-frames=25 --min-q=0 --max-q=63
--end-usage=0 --auto-alt-ref=1 -p 2 --pass=2 --kf-max-dist=9999
--kf-min-dist=0 --drop-frame=0 --static-thresh=0 --bias-pct=50
--minsection-pct=0 --maxsection-pct=800 --sharpness=0
--arnr-maxframes=7 --arnr-strength=3(for HD,6 for CIF)
--arnr-type=3
Change-Id: I5c62ed09cfff5815a2bb34e7820d6a810c23183c
2012-03-10 02:32:50 +01:00
|
|
|
#ifdef ANNOUNCE_FUNCTION
|
2012-10-31 00:25:53 +01:00
|
|
|
printf("vp9_sixtap_predict4x4_mmx\n");
|
WebM Experimental Codec Branch Snapshot
This is a code snapshot of experimental work currently ongoing for a
next-generation codec.
The codebase has been cut down considerably from the libvpx baseline.
For example, we are currently only supporting VBR 2-pass rate control
and have removed most of the code relating to coding speed, threading,
error resilience, partitions and various other features. This is in
part to make the codebase easier to work on and experiment with, but
also because we want to have an open discussion about how the bitstream
will be structured and partitioned and not have that conversation
constrained by past work.
Our basic working pattern has been to initially encapsulate experiments
using configure options linked to #IF CONFIG_XXX statements in the
code. Once experiments have matured and we are reasonably happy that
they give benefit and can be merged without breaking other experiments,
we remove the conditional compile statements and merge them in.
Current changes include:
* Temporal coding experiment for segments (though still only 4 max, it
will likely be increased).
* Segment feature experiment - to allow various bits of information to
be coded at the segment level. Features tested so far include mode
and reference frame information, limiting end of block offset and
transform size, alongside Q and loop filter parameters, but this set
is very fluid.
* Support for 8x8 transform - 8x8 dct with 2nd order 2x2 haar is used
in MBs using 16x16 prediction modes within inter frames.
* Compound prediction (combination of signals from existing predictors
to create a new predictor).
* 8 tap interpolation filters and 1/8th pel motion vectors.
* Loop filter modifications.
* Various entropy modifications and changes to how entropy contexts and
updates are handled.
* Extended quantizer range matched to transform precision improvements.
There are also ongoing further experiments that we hope to merge in the
near future: For example, coding of motion and other aspects of the
prediction signal to better support larger image formats, use of larger
block sizes (e.g. 32x32 and up) and lossless non-transform based coding
options (especially for key frames). It is our hope that we will be
able to make regular updates and we will warmly welcome community
contributions.
Please be warned that, at this stage, the codebase is currently slower
than VP8 stable branch as most new code has not been optimized, and
even the 'C' has been deliberately written to be simple and obvious,
not fast.
The following graphs have the initial test results, numbers in the
tables measure the compression improvement in terms of percentage. The
build has the following optional experiments configured:
--enable-experimental --enable-enhanced_interp --enable-uvintra
--enable-high_precision_mv --enable-sixteenth_subpel_uv
CIF Size clips:
http://getwebm.org/tmp/cif/
HD size clips:
http://getwebm.org/tmp/hd/
(stable_20120309 represents encoding results of WebM master branch
build as of commit#7a15907)
They were encoded using the following encode parameters:
--good --cpu-used=0 -t 0 --lag-in-frames=25 --min-q=0 --max-q=63
--end-usage=0 --auto-alt-ref=1 -p 2 --pass=2 --kf-max-dist=9999
--kf-min-dist=0 --drop-frame=0 --static-thresh=0 --bias-pct=50
--minsection-pct=0 --maxsection-pct=800 --sharpness=0
--arnr-maxframes=7 --arnr-strength=3(for HD,6 for CIF)
--arnr-type=3
Change-Id: I5c62ed09cfff5815a2bb34e7820d6a810c23183c
2012-03-10 02:32:50 +01:00
|
|
|
#endif
|
2012-11-01 00:09:17 +01:00
|
|
|
/* Temp data bufffer used in filtering */
|
|
|
|
|
DECLARE_ALIGNED_ARRAY(16, unsigned short, fdata2, 16 * 16);
|
|
|
|
|
const short *hfilter, *vfilter;
|
|
|
|
|
hfilter = vp9_six_tap_mmx[xoffset];
|
|
|
|
|
vp9_filter_block1d_h6_mmx(src_ptr - (2 * src_pixels_per_line), fdata2,
|
|
|
|
|
src_pixels_per_line, 1, 9, 8, hfilter);
|
|
|
|
|
vfilter = vp9_six_tap_mmx[yoffset];
|
|
|
|
|
vp9_filter_block1dc_v6_mmx(fdata2 + 8, dst_ptr, dst_pitch,
|
|
|
|
|
8, 4, 4, 4, vfilter);
|
2010-05-18 17:58:33 +02:00
|
|
|
}
|
|
|
|
|
|
2012-11-01 00:09:17 +01:00
|
|
|
void vp9_sixtap_predict16x16_mmx(unsigned char *src_ptr,
|
|
|
|
|
int src_pixels_per_line,
|
|
|
|
|
int xoffset,
|
|
|
|
|
int yoffset,
|
|
|
|
|
unsigned char *dst_ptr,
|
|
|
|
|
int dst_pitch) {
|
WebM Experimental Codec Branch Snapshot
This is a code snapshot of experimental work currently ongoing for a
next-generation codec.
The codebase has been cut down considerably from the libvpx baseline.
For example, we are currently only supporting VBR 2-pass rate control
and have removed most of the code relating to coding speed, threading,
error resilience, partitions and various other features. This is in
part to make the codebase easier to work on and experiment with, but
also because we want to have an open discussion about how the bitstream
will be structured and partitioned and not have that conversation
constrained by past work.
Our basic working pattern has been to initially encapsulate experiments
using configure options linked to #IF CONFIG_XXX statements in the
code. Once experiments have matured and we are reasonably happy that
they give benefit and can be merged without breaking other experiments,
we remove the conditional compile statements and merge them in.
Current changes include:
* Temporal coding experiment for segments (though still only 4 max, it
will likely be increased).
* Segment feature experiment - to allow various bits of information to
be coded at the segment level. Features tested so far include mode
and reference frame information, limiting end of block offset and
transform size, alongside Q and loop filter parameters, but this set
is very fluid.
* Support for 8x8 transform - 8x8 dct with 2nd order 2x2 haar is used
in MBs using 16x16 prediction modes within inter frames.
* Compound prediction (combination of signals from existing predictors
to create a new predictor).
* 8 tap interpolation filters and 1/8th pel motion vectors.
* Loop filter modifications.
* Various entropy modifications and changes to how entropy contexts and
updates are handled.
* Extended quantizer range matched to transform precision improvements.
There are also ongoing further experiments that we hope to merge in the
near future: For example, coding of motion and other aspects of the
prediction signal to better support larger image formats, use of larger
block sizes (e.g. 32x32 and up) and lossless non-transform based coding
options (especially for key frames). It is our hope that we will be
able to make regular updates and we will warmly welcome community
contributions.
Please be warned that, at this stage, the codebase is currently slower
than VP8 stable branch as most new code has not been optimized, and
even the 'C' has been deliberately written to be simple and obvious,
not fast.
The following graphs have the initial test results, numbers in the
tables measure the compression improvement in terms of percentage. The
build has the following optional experiments configured:
--enable-experimental --enable-enhanced_interp --enable-uvintra
--enable-high_precision_mv --enable-sixteenth_subpel_uv
CIF Size clips:
http://getwebm.org/tmp/cif/
HD size clips:
http://getwebm.org/tmp/hd/
(stable_20120309 represents encoding results of WebM master branch
build as of commit#7a15907)
They were encoded using the following encode parameters:
--good --cpu-used=0 -t 0 --lag-in-frames=25 --min-q=0 --max-q=63
--end-usage=0 --auto-alt-ref=1 -p 2 --pass=2 --kf-max-dist=9999
--kf-min-dist=0 --drop-frame=0 --static-thresh=0 --bias-pct=50
--minsection-pct=0 --maxsection-pct=800 --sharpness=0
--arnr-maxframes=7 --arnr-strength=3(for HD,6 for CIF)
--arnr-type=3
Change-Id: I5c62ed09cfff5815a2bb34e7820d6a810c23183c
2012-03-10 02:32:50 +01:00
|
|
|
#ifdef ANNOUNCE_FUNCTION
|
2012-10-31 00:25:53 +01:00
|
|
|
printf("vp9_sixtap_predict16x16_mmx\n");
|
WebM Experimental Codec Branch Snapshot
This is a code snapshot of experimental work currently ongoing for a
next-generation codec.
The codebase has been cut down considerably from the libvpx baseline.
For example, we are currently only supporting VBR 2-pass rate control
and have removed most of the code relating to coding speed, threading,
error resilience, partitions and various other features. This is in
part to make the codebase easier to work on and experiment with, but
also because we want to have an open discussion about how the bitstream
will be structured and partitioned and not have that conversation
constrained by past work.
Our basic working pattern has been to initially encapsulate experiments
using configure options linked to #IF CONFIG_XXX statements in the
code. Once experiments have matured and we are reasonably happy that
they give benefit and can be merged without breaking other experiments,
we remove the conditional compile statements and merge them in.
Current changes include:
* Temporal coding experiment for segments (though still only 4 max, it
will likely be increased).
* Segment feature experiment - to allow various bits of information to
be coded at the segment level. Features tested so far include mode
and reference frame information, limiting end of block offset and
transform size, alongside Q and loop filter parameters, but this set
is very fluid.
* Support for 8x8 transform - 8x8 dct with 2nd order 2x2 haar is used
in MBs using 16x16 prediction modes within inter frames.
* Compound prediction (combination of signals from existing predictors
to create a new predictor).
* 8 tap interpolation filters and 1/8th pel motion vectors.
* Loop filter modifications.
* Various entropy modifications and changes to how entropy contexts and
updates are handled.
* Extended quantizer range matched to transform precision improvements.
There are also ongoing further experiments that we hope to merge in the
near future: For example, coding of motion and other aspects of the
prediction signal to better support larger image formats, use of larger
block sizes (e.g. 32x32 and up) and lossless non-transform based coding
options (especially for key frames). It is our hope that we will be
able to make regular updates and we will warmly welcome community
contributions.
Please be warned that, at this stage, the codebase is currently slower
than VP8 stable branch as most new code has not been optimized, and
even the 'C' has been deliberately written to be simple and obvious,
not fast.
The following graphs have the initial test results, numbers in the
tables measure the compression improvement in terms of percentage. The
build has the following optional experiments configured:
--enable-experimental --enable-enhanced_interp --enable-uvintra
--enable-high_precision_mv --enable-sixteenth_subpel_uv
CIF Size clips:
http://getwebm.org/tmp/cif/
HD size clips:
http://getwebm.org/tmp/hd/
(stable_20120309 represents encoding results of WebM master branch
build as of commit#7a15907)
They were encoded using the following encode parameters:
--good --cpu-used=0 -t 0 --lag-in-frames=25 --min-q=0 --max-q=63
--end-usage=0 --auto-alt-ref=1 -p 2 --pass=2 --kf-max-dist=9999
--kf-min-dist=0 --drop-frame=0 --static-thresh=0 --bias-pct=50
--minsection-pct=0 --maxsection-pct=800 --sharpness=0
--arnr-maxframes=7 --arnr-strength=3(for HD,6 for CIF)
--arnr-type=3
Change-Id: I5c62ed09cfff5815a2bb34e7820d6a810c23183c
2012-03-10 02:32:50 +01:00
|
|
|
#endif
|
2012-11-01 00:09:17 +01:00
|
|
|
/* Temp data bufffer used in filtering */
|
|
|
|
|
DECLARE_ALIGNED_ARRAY(16, unsigned short, fdata2, 24 * 24);
|
|
|
|
|
const short *hfilter, *vfilter;
|
|
|
|
|
|
|
|
|
|
hfilter = vp9_six_tap_mmx[xoffset];
|
|
|
|
|
vp9_filter_block1d_h6_mmx(src_ptr - (2 * src_pixels_per_line),
|
|
|
|
|
fdata2, src_pixels_per_line, 1, 21, 32,
|
|
|
|
|
hfilter);
|
|
|
|
|
vp9_filter_block1d_h6_mmx(src_ptr - (2 * src_pixels_per_line) + 4,
|
|
|
|
|
fdata2 + 4, src_pixels_per_line, 1, 21, 32,
|
|
|
|
|
hfilter);
|
|
|
|
|
vp9_filter_block1d_h6_mmx(src_ptr - (2 * src_pixels_per_line) + 8,
|
|
|
|
|
fdata2 + 8, src_pixels_per_line, 1, 21, 32,
|
|
|
|
|
hfilter);
|
|
|
|
|
vp9_filter_block1d_h6_mmx(src_ptr - (2 * src_pixels_per_line) + 12,
|
|
|
|
|
fdata2 + 12, src_pixels_per_line, 1, 21, 32,
|
|
|
|
|
hfilter);
|
|
|
|
|
|
|
|
|
|
vfilter = vp9_six_tap_mmx[yoffset];
|
|
|
|
|
vp9_filter_block1dc_v6_mmx(fdata2 + 32, dst_ptr, dst_pitch,
|
|
|
|
|
32, 16, 16, 16, vfilter);
|
|
|
|
|
vp9_filter_block1dc_v6_mmx(fdata2 + 36, dst_ptr + 4, dst_pitch,
|
|
|
|
|
32, 16, 16, 16, vfilter);
|
|
|
|
|
vp9_filter_block1dc_v6_mmx(fdata2 + 40, dst_ptr + 8, dst_pitch,
|
|
|
|
|
32, 16, 16, 16, vfilter);
|
|
|
|
|
vp9_filter_block1dc_v6_mmx(fdata2 + 44, dst_ptr + 12, dst_pitch,
|
|
|
|
|
32, 16, 16, 16, vfilter);
|
2010-05-18 17:58:33 +02:00
|
|
|
}
|
|
|
|
|
|
2012-11-01 00:09:17 +01:00
|
|
|
void vp9_sixtap_predict8x8_mmx(unsigned char *src_ptr,
|
|
|
|
|
int src_pixels_per_line,
|
|
|
|
|
int xoffset,
|
|
|
|
|
int yoffset,
|
|
|
|
|
unsigned char *dst_ptr,
|
|
|
|
|
int dst_pitch) {
|
WebM Experimental Codec Branch Snapshot
This is a code snapshot of experimental work currently ongoing for a
next-generation codec.
The codebase has been cut down considerably from the libvpx baseline.
For example, we are currently only supporting VBR 2-pass rate control
and have removed most of the code relating to coding speed, threading,
error resilience, partitions and various other features. This is in
part to make the codebase easier to work on and experiment with, but
also because we want to have an open discussion about how the bitstream
will be structured and partitioned and not have that conversation
constrained by past work.
Our basic working pattern has been to initially encapsulate experiments
using configure options linked to #IF CONFIG_XXX statements in the
code. Once experiments have matured and we are reasonably happy that
they give benefit and can be merged without breaking other experiments,
we remove the conditional compile statements and merge them in.
Current changes include:
* Temporal coding experiment for segments (though still only 4 max, it
will likely be increased).
* Segment feature experiment - to allow various bits of information to
be coded at the segment level. Features tested so far include mode
and reference frame information, limiting end of block offset and
transform size, alongside Q and loop filter parameters, but this set
is very fluid.
* Support for 8x8 transform - 8x8 dct with 2nd order 2x2 haar is used
in MBs using 16x16 prediction modes within inter frames.
* Compound prediction (combination of signals from existing predictors
to create a new predictor).
* 8 tap interpolation filters and 1/8th pel motion vectors.
* Loop filter modifications.
* Various entropy modifications and changes to how entropy contexts and
updates are handled.
* Extended quantizer range matched to transform precision improvements.
There are also ongoing further experiments that we hope to merge in the
near future: For example, coding of motion and other aspects of the
prediction signal to better support larger image formats, use of larger
block sizes (e.g. 32x32 and up) and lossless non-transform based coding
options (especially for key frames). It is our hope that we will be
able to make regular updates and we will warmly welcome community
contributions.
Please be warned that, at this stage, the codebase is currently slower
than VP8 stable branch as most new code has not been optimized, and
even the 'C' has been deliberately written to be simple and obvious,
not fast.
The following graphs have the initial test results, numbers in the
tables measure the compression improvement in terms of percentage. The
build has the following optional experiments configured:
--enable-experimental --enable-enhanced_interp --enable-uvintra
--enable-high_precision_mv --enable-sixteenth_subpel_uv
CIF Size clips:
http://getwebm.org/tmp/cif/
HD size clips:
http://getwebm.org/tmp/hd/
(stable_20120309 represents encoding results of WebM master branch
build as of commit#7a15907)
They were encoded using the following encode parameters:
--good --cpu-used=0 -t 0 --lag-in-frames=25 --min-q=0 --max-q=63
--end-usage=0 --auto-alt-ref=1 -p 2 --pass=2 --kf-max-dist=9999
--kf-min-dist=0 --drop-frame=0 --static-thresh=0 --bias-pct=50
--minsection-pct=0 --maxsection-pct=800 --sharpness=0
--arnr-maxframes=7 --arnr-strength=3(for HD,6 for CIF)
--arnr-type=3
Change-Id: I5c62ed09cfff5815a2bb34e7820d6a810c23183c
2012-03-10 02:32:50 +01:00
|
|
|
#ifdef ANNOUNCE_FUNCTION
|
2012-10-31 00:25:53 +01:00
|
|
|
printf("vp9_sixtap_predict8x8_mmx\n");
|
WebM Experimental Codec Branch Snapshot
This is a code snapshot of experimental work currently ongoing for a
next-generation codec.
The codebase has been cut down considerably from the libvpx baseline.
For example, we are currently only supporting VBR 2-pass rate control
and have removed most of the code relating to coding speed, threading,
error resilience, partitions and various other features. This is in
part to make the codebase easier to work on and experiment with, but
also because we want to have an open discussion about how the bitstream
will be structured and partitioned and not have that conversation
constrained by past work.
Our basic working pattern has been to initially encapsulate experiments
using configure options linked to #IF CONFIG_XXX statements in the
code. Once experiments have matured and we are reasonably happy that
they give benefit and can be merged without breaking other experiments,
we remove the conditional compile statements and merge them in.
Current changes include:
* Temporal coding experiment for segments (though still only 4 max, it
will likely be increased).
* Segment feature experiment - to allow various bits of information to
be coded at the segment level. Features tested so far include mode
and reference frame information, limiting end of block offset and
transform size, alongside Q and loop filter parameters, but this set
is very fluid.
* Support for 8x8 transform - 8x8 dct with 2nd order 2x2 haar is used
in MBs using 16x16 prediction modes within inter frames.
* Compound prediction (combination of signals from existing predictors
to create a new predictor).
* 8 tap interpolation filters and 1/8th pel motion vectors.
* Loop filter modifications.
* Various entropy modifications and changes to how entropy contexts and
updates are handled.
* Extended quantizer range matched to transform precision improvements.
There are also ongoing further experiments that we hope to merge in the
near future: For example, coding of motion and other aspects of the
prediction signal to better support larger image formats, use of larger
block sizes (e.g. 32x32 and up) and lossless non-transform based coding
options (especially for key frames). It is our hope that we will be
able to make regular updates and we will warmly welcome community
contributions.
Please be warned that, at this stage, the codebase is currently slower
than VP8 stable branch as most new code has not been optimized, and
even the 'C' has been deliberately written to be simple and obvious,
not fast.
The following graphs have the initial test results, numbers in the
tables measure the compression improvement in terms of percentage. The
build has the following optional experiments configured:
--enable-experimental --enable-enhanced_interp --enable-uvintra
--enable-high_precision_mv --enable-sixteenth_subpel_uv
CIF Size clips:
http://getwebm.org/tmp/cif/
HD size clips:
http://getwebm.org/tmp/hd/
(stable_20120309 represents encoding results of WebM master branch
build as of commit#7a15907)
They were encoded using the following encode parameters:
--good --cpu-used=0 -t 0 --lag-in-frames=25 --min-q=0 --max-q=63
--end-usage=0 --auto-alt-ref=1 -p 2 --pass=2 --kf-max-dist=9999
--kf-min-dist=0 --drop-frame=0 --static-thresh=0 --bias-pct=50
--minsection-pct=0 --maxsection-pct=800 --sharpness=0
--arnr-maxframes=7 --arnr-strength=3(for HD,6 for CIF)
--arnr-type=3
Change-Id: I5c62ed09cfff5815a2bb34e7820d6a810c23183c
2012-03-10 02:32:50 +01:00
|
|
|
#endif
|
2012-11-01 00:09:17 +01:00
|
|
|
/* Temp data bufffer used in filtering */
|
|
|
|
|
DECLARE_ALIGNED_ARRAY(16, unsigned short, fdata2, 256);
|
|
|
|
|
const short *hfilter, *vfilter;
|
|
|
|
|
|
|
|
|
|
hfilter = vp9_six_tap_mmx[xoffset];
|
|
|
|
|
vp9_filter_block1d_h6_mmx(src_ptr - (2 * src_pixels_per_line),
|
|
|
|
|
fdata2, src_pixels_per_line, 1, 13, 16,
|
|
|
|
|
hfilter);
|
|
|
|
|
vp9_filter_block1d_h6_mmx(src_ptr - (2 * src_pixels_per_line) + 4,
|
|
|
|
|
fdata2 + 4, src_pixels_per_line, 1, 13, 16,
|
|
|
|
|
hfilter);
|
|
|
|
|
|
|
|
|
|
vfilter = vp9_six_tap_mmx[yoffset];
|
|
|
|
|
vp9_filter_block1dc_v6_mmx(fdata2 + 16, dst_ptr, dst_pitch,
|
|
|
|
|
16, 8, 8, 8, vfilter);
|
|
|
|
|
vp9_filter_block1dc_v6_mmx(fdata2 + 20, dst_ptr + 4, dst_pitch,
|
|
|
|
|
16, 8, 8, 8, vfilter);
|
2010-05-18 17:58:33 +02:00
|
|
|
}
|
|
|
|
|
|
2012-11-01 00:09:17 +01:00
|
|
|
void vp9_sixtap_predict8x4_mmx(unsigned char *src_ptr,
|
|
|
|
|
int src_pixels_per_line,
|
|
|
|
|
int xoffset,
|
|
|
|
|
int yoffset,
|
|
|
|
|
unsigned char *dst_ptr,
|
|
|
|
|
int dst_pitch) {
|
WebM Experimental Codec Branch Snapshot
This is a code snapshot of experimental work currently ongoing for a
next-generation codec.
The codebase has been cut down considerably from the libvpx baseline.
For example, we are currently only supporting VBR 2-pass rate control
and have removed most of the code relating to coding speed, threading,
error resilience, partitions and various other features. This is in
part to make the codebase easier to work on and experiment with, but
also because we want to have an open discussion about how the bitstream
will be structured and partitioned and not have that conversation
constrained by past work.
Our basic working pattern has been to initially encapsulate experiments
using configure options linked to #IF CONFIG_XXX statements in the
code. Once experiments have matured and we are reasonably happy that
they give benefit and can be merged without breaking other experiments,
we remove the conditional compile statements and merge them in.
Current changes include:
* Temporal coding experiment for segments (though still only 4 max, it
will likely be increased).
* Segment feature experiment - to allow various bits of information to
be coded at the segment level. Features tested so far include mode
and reference frame information, limiting end of block offset and
transform size, alongside Q and loop filter parameters, but this set
is very fluid.
* Support for 8x8 transform - 8x8 dct with 2nd order 2x2 haar is used
in MBs using 16x16 prediction modes within inter frames.
* Compound prediction (combination of signals from existing predictors
to create a new predictor).
* 8 tap interpolation filters and 1/8th pel motion vectors.
* Loop filter modifications.
* Various entropy modifications and changes to how entropy contexts and
updates are handled.
* Extended quantizer range matched to transform precision improvements.
There are also ongoing further experiments that we hope to merge in the
near future: For example, coding of motion and other aspects of the
prediction signal to better support larger image formats, use of larger
block sizes (e.g. 32x32 and up) and lossless non-transform based coding
options (especially for key frames). It is our hope that we will be
able to make regular updates and we will warmly welcome community
contributions.
Please be warned that, at this stage, the codebase is currently slower
than VP8 stable branch as most new code has not been optimized, and
even the 'C' has been deliberately written to be simple and obvious,
not fast.
The following graphs have the initial test results, numbers in the
tables measure the compression improvement in terms of percentage. The
build has the following optional experiments configured:
--enable-experimental --enable-enhanced_interp --enable-uvintra
--enable-high_precision_mv --enable-sixteenth_subpel_uv
CIF Size clips:
http://getwebm.org/tmp/cif/
HD size clips:
http://getwebm.org/tmp/hd/
(stable_20120309 represents encoding results of WebM master branch
build as of commit#7a15907)
They were encoded using the following encode parameters:
--good --cpu-used=0 -t 0 --lag-in-frames=25 --min-q=0 --max-q=63
--end-usage=0 --auto-alt-ref=1 -p 2 --pass=2 --kf-max-dist=9999
--kf-min-dist=0 --drop-frame=0 --static-thresh=0 --bias-pct=50
--minsection-pct=0 --maxsection-pct=800 --sharpness=0
--arnr-maxframes=7 --arnr-strength=3(for HD,6 for CIF)
--arnr-type=3
Change-Id: I5c62ed09cfff5815a2bb34e7820d6a810c23183c
2012-03-10 02:32:50 +01:00
|
|
|
#ifdef ANNOUNCE_FUNCTION
|
2012-10-31 00:25:53 +01:00
|
|
|
printf("vp9_sixtap_predict8x4_mmx\n");
|
WebM Experimental Codec Branch Snapshot
This is a code snapshot of experimental work currently ongoing for a
next-generation codec.
The codebase has been cut down considerably from the libvpx baseline.
For example, we are currently only supporting VBR 2-pass rate control
and have removed most of the code relating to coding speed, threading,
error resilience, partitions and various other features. This is in
part to make the codebase easier to work on and experiment with, but
also because we want to have an open discussion about how the bitstream
will be structured and partitioned and not have that conversation
constrained by past work.
Our basic working pattern has been to initially encapsulate experiments
using configure options linked to #IF CONFIG_XXX statements in the
code. Once experiments have matured and we are reasonably happy that
they give benefit and can be merged without breaking other experiments,
we remove the conditional compile statements and merge them in.
Current changes include:
* Temporal coding experiment for segments (though still only 4 max, it
will likely be increased).
* Segment feature experiment - to allow various bits of information to
be coded at the segment level. Features tested so far include mode
and reference frame information, limiting end of block offset and
transform size, alongside Q and loop filter parameters, but this set
is very fluid.
* Support for 8x8 transform - 8x8 dct with 2nd order 2x2 haar is used
in MBs using 16x16 prediction modes within inter frames.
* Compound prediction (combination of signals from existing predictors
to create a new predictor).
* 8 tap interpolation filters and 1/8th pel motion vectors.
* Loop filter modifications.
* Various entropy modifications and changes to how entropy contexts and
updates are handled.
* Extended quantizer range matched to transform precision improvements.
There are also ongoing further experiments that we hope to merge in the
near future: For example, coding of motion and other aspects of the
prediction signal to better support larger image formats, use of larger
block sizes (e.g. 32x32 and up) and lossless non-transform based coding
options (especially for key frames). It is our hope that we will be
able to make regular updates and we will warmly welcome community
contributions.
Please be warned that, at this stage, the codebase is currently slower
than VP8 stable branch as most new code has not been optimized, and
even the 'C' has been deliberately written to be simple and obvious,
not fast.
The following graphs have the initial test results, numbers in the
tables measure the compression improvement in terms of percentage. The
build has the following optional experiments configured:
--enable-experimental --enable-enhanced_interp --enable-uvintra
--enable-high_precision_mv --enable-sixteenth_subpel_uv
CIF Size clips:
http://getwebm.org/tmp/cif/
HD size clips:
http://getwebm.org/tmp/hd/
(stable_20120309 represents encoding results of WebM master branch
build as of commit#7a15907)
They were encoded using the following encode parameters:
--good --cpu-used=0 -t 0 --lag-in-frames=25 --min-q=0 --max-q=63
--end-usage=0 --auto-alt-ref=1 -p 2 --pass=2 --kf-max-dist=9999
--kf-min-dist=0 --drop-frame=0 --static-thresh=0 --bias-pct=50
--minsection-pct=0 --maxsection-pct=800 --sharpness=0
--arnr-maxframes=7 --arnr-strength=3(for HD,6 for CIF)
--arnr-type=3
Change-Id: I5c62ed09cfff5815a2bb34e7820d6a810c23183c
2012-03-10 02:32:50 +01:00
|
|
|
#endif
|
2012-11-01 00:09:17 +01:00
|
|
|
/* Temp data bufffer used in filtering */
|
|
|
|
|
DECLARE_ALIGNED_ARRAY(16, unsigned short, fdata2, 256);
|
|
|
|
|
const short *hfilter, *vfilter;
|
|
|
|
|
|
|
|
|
|
hfilter = vp9_six_tap_mmx[xoffset];
|
|
|
|
|
vp9_filter_block1d_h6_mmx(src_ptr - (2 * src_pixels_per_line),
|
|
|
|
|
fdata2, src_pixels_per_line, 1, 9, 16, hfilter);
|
|
|
|
|
vp9_filter_block1d_h6_mmx(src_ptr - (2 * src_pixels_per_line) + 4,
|
|
|
|
|
fdata2 + 4, src_pixels_per_line, 1, 9, 16, hfilter);
|
|
|
|
|
|
|
|
|
|
vfilter = vp9_six_tap_mmx[yoffset];
|
|
|
|
|
vp9_filter_block1dc_v6_mmx(fdata2 + 16, dst_ptr, dst_pitch,
|
|
|
|
|
16, 8, 4, 8, vfilter);
|
|
|
|
|
vp9_filter_block1dc_v6_mmx(fdata2 + 20, dst_ptr + 4, dst_pitch,
|
|
|
|
|
16, 8, 4, 8, vfilter);
|
2010-05-18 17:58:33 +02:00
|
|
|
}
|
|
|
|
|
|
2012-11-01 00:09:17 +01:00
|
|
|
void vp9_bilinear_predict16x16_mmx(unsigned char *src_ptr,
|
|
|
|
|
int src_pixels_per_line,
|
|
|
|
|
int xoffset,
|
|
|
|
|
int yoffset,
|
|
|
|
|
unsigned char *dst_ptr,
|
|
|
|
|
int dst_pitch) {
|
|
|
|
|
vp9_bilinear_predict8x8_mmx(src_ptr,
|
|
|
|
|
src_pixels_per_line, xoffset, yoffset,
|
|
|
|
|
dst_ptr, dst_pitch);
|
|
|
|
|
vp9_bilinear_predict8x8_mmx(src_ptr + 8,
|
|
|
|
|
src_pixels_per_line, xoffset, yoffset,
|
|
|
|
|
dst_ptr + 8, dst_pitch);
|
|
|
|
|
vp9_bilinear_predict8x8_mmx(src_ptr + 8 * src_pixels_per_line,
|
|
|
|
|
src_pixels_per_line, xoffset, yoffset,
|
|
|
|
|
dst_ptr + dst_pitch * 8, dst_pitch);
|
|
|
|
|
vp9_bilinear_predict8x8_mmx(src_ptr + 8 * src_pixels_per_line + 8,
|
|
|
|
|
src_pixels_per_line, xoffset, yoffset,
|
|
|
|
|
dst_ptr + dst_pitch * 8 + 8, dst_pitch);
|
2010-05-18 17:58:33 +02:00
|
|
|
}
|
|
|
|
|
#endif
|
|
|
|
|
|
|
|
|
|
#if HAVE_SSE2
|
2012-11-01 00:09:17 +01:00
|
|
|
void vp9_sixtap_predict16x16_sse2(unsigned char *src_ptr,
|
|
|
|
|
int src_pixels_per_line,
|
|
|
|
|
int xoffset,
|
|
|
|
|
int yoffset,
|
|
|
|
|
unsigned char *dst_ptr,
|
|
|
|
|
int dst_pitch) {
|
|
|
|
|
/* Temp data bufffer used in filtering */
|
|
|
|
|
DECLARE_ALIGNED_ARRAY(16, unsigned short, fdata2, 24 * 24);
|
|
|
|
|
const short *hfilter, *vfilter;
|
WebM Experimental Codec Branch Snapshot
This is a code snapshot of experimental work currently ongoing for a
next-generation codec.
The codebase has been cut down considerably from the libvpx baseline.
For example, we are currently only supporting VBR 2-pass rate control
and have removed most of the code relating to coding speed, threading,
error resilience, partitions and various other features. This is in
part to make the codebase easier to work on and experiment with, but
also because we want to have an open discussion about how the bitstream
will be structured and partitioned and not have that conversation
constrained by past work.
Our basic working pattern has been to initially encapsulate experiments
using configure options linked to #IF CONFIG_XXX statements in the
code. Once experiments have matured and we are reasonably happy that
they give benefit and can be merged without breaking other experiments,
we remove the conditional compile statements and merge them in.
Current changes include:
* Temporal coding experiment for segments (though still only 4 max, it
will likely be increased).
* Segment feature experiment - to allow various bits of information to
be coded at the segment level. Features tested so far include mode
and reference frame information, limiting end of block offset and
transform size, alongside Q and loop filter parameters, but this set
is very fluid.
* Support for 8x8 transform - 8x8 dct with 2nd order 2x2 haar is used
in MBs using 16x16 prediction modes within inter frames.
* Compound prediction (combination of signals from existing predictors
to create a new predictor).
* 8 tap interpolation filters and 1/8th pel motion vectors.
* Loop filter modifications.
* Various entropy modifications and changes to how entropy contexts and
updates are handled.
* Extended quantizer range matched to transform precision improvements.
There are also ongoing further experiments that we hope to merge in the
near future: For example, coding of motion and other aspects of the
prediction signal to better support larger image formats, use of larger
block sizes (e.g. 32x32 and up) and lossless non-transform based coding
options (especially for key frames). It is our hope that we will be
able to make regular updates and we will warmly welcome community
contributions.
Please be warned that, at this stage, the codebase is currently slower
than VP8 stable branch as most new code has not been optimized, and
even the 'C' has been deliberately written to be simple and obvious,
not fast.
The following graphs have the initial test results, numbers in the
tables measure the compression improvement in terms of percentage. The
build has the following optional experiments configured:
--enable-experimental --enable-enhanced_interp --enable-uvintra
--enable-high_precision_mv --enable-sixteenth_subpel_uv
CIF Size clips:
http://getwebm.org/tmp/cif/
HD size clips:
http://getwebm.org/tmp/hd/
(stable_20120309 represents encoding results of WebM master branch
build as of commit#7a15907)
They were encoded using the following encode parameters:
--good --cpu-used=0 -t 0 --lag-in-frames=25 --min-q=0 --max-q=63
--end-usage=0 --auto-alt-ref=1 -p 2 --pass=2 --kf-max-dist=9999
--kf-min-dist=0 --drop-frame=0 --static-thresh=0 --bias-pct=50
--minsection-pct=0 --maxsection-pct=800 --sharpness=0
--arnr-maxframes=7 --arnr-strength=3(for HD,6 for CIF)
--arnr-type=3
Change-Id: I5c62ed09cfff5815a2bb34e7820d6a810c23183c
2012-03-10 02:32:50 +01:00
|
|
|
#ifdef ANNOUNCE_FUNCTION
|
2012-10-31 00:25:53 +01:00
|
|
|
printf("vp9_sixtap_predict16x16_sse2\n");
|
WebM Experimental Codec Branch Snapshot
This is a code snapshot of experimental work currently ongoing for a
next-generation codec.
The codebase has been cut down considerably from the libvpx baseline.
For example, we are currently only supporting VBR 2-pass rate control
and have removed most of the code relating to coding speed, threading,
error resilience, partitions and various other features. This is in
part to make the codebase easier to work on and experiment with, but
also because we want to have an open discussion about how the bitstream
will be structured and partitioned and not have that conversation
constrained by past work.
Our basic working pattern has been to initially encapsulate experiments
using configure options linked to #IF CONFIG_XXX statements in the
code. Once experiments have matured and we are reasonably happy that
they give benefit and can be merged without breaking other experiments,
we remove the conditional compile statements and merge them in.
Current changes include:
* Temporal coding experiment for segments (though still only 4 max, it
will likely be increased).
* Segment feature experiment - to allow various bits of information to
be coded at the segment level. Features tested so far include mode
and reference frame information, limiting end of block offset and
transform size, alongside Q and loop filter parameters, but this set
is very fluid.
* Support for 8x8 transform - 8x8 dct with 2nd order 2x2 haar is used
in MBs using 16x16 prediction modes within inter frames.
* Compound prediction (combination of signals from existing predictors
to create a new predictor).
* 8 tap interpolation filters and 1/8th pel motion vectors.
* Loop filter modifications.
* Various entropy modifications and changes to how entropy contexts and
updates are handled.
* Extended quantizer range matched to transform precision improvements.
There are also ongoing further experiments that we hope to merge in the
near future: For example, coding of motion and other aspects of the
prediction signal to better support larger image formats, use of larger
block sizes (e.g. 32x32 and up) and lossless non-transform based coding
options (especially for key frames). It is our hope that we will be
able to make regular updates and we will warmly welcome community
contributions.
Please be warned that, at this stage, the codebase is currently slower
than VP8 stable branch as most new code has not been optimized, and
even the 'C' has been deliberately written to be simple and obvious,
not fast.
The following graphs have the initial test results, numbers in the
tables measure the compression improvement in terms of percentage. The
build has the following optional experiments configured:
--enable-experimental --enable-enhanced_interp --enable-uvintra
--enable-high_precision_mv --enable-sixteenth_subpel_uv
CIF Size clips:
http://getwebm.org/tmp/cif/
HD size clips:
http://getwebm.org/tmp/hd/
(stable_20120309 represents encoding results of WebM master branch
build as of commit#7a15907)
They were encoded using the following encode parameters:
--good --cpu-used=0 -t 0 --lag-in-frames=25 --min-q=0 --max-q=63
--end-usage=0 --auto-alt-ref=1 -p 2 --pass=2 --kf-max-dist=9999
--kf-min-dist=0 --drop-frame=0 --static-thresh=0 --bias-pct=50
--minsection-pct=0 --maxsection-pct=800 --sharpness=0
--arnr-maxframes=7 --arnr-strength=3(for HD,6 for CIF)
--arnr-type=3
Change-Id: I5c62ed09cfff5815a2bb34e7820d6a810c23183c
2012-03-10 02:32:50 +01:00
|
|
|
#endif
|
2010-05-18 17:58:33 +02:00
|
|
|
|
2012-07-14 00:21:29 +02:00
|
|
|
if (xoffset) {
|
|
|
|
|
if (yoffset) {
|
2012-11-01 00:09:17 +01:00
|
|
|
hfilter = vp9_six_tap_mmx[xoffset];
|
|
|
|
|
vp9_filter_block1d16_h6_sse2(src_ptr - (2 * src_pixels_per_line), fdata2,
|
|
|
|
|
src_pixels_per_line, 1, 21, 32, hfilter);
|
|
|
|
|
vfilter = vp9_six_tap_mmx[yoffset];
|
|
|
|
|
vp9_filter_block1d16_v6_sse2(fdata2 + 32, dst_ptr, dst_pitch,
|
|
|
|
|
32, 16, 16, dst_pitch, vfilter);
|
2012-07-14 00:21:29 +02:00
|
|
|
} else {
|
|
|
|
|
/* First-pass only */
|
2012-11-01 00:09:17 +01:00
|
|
|
hfilter = vp9_six_tap_mmx[xoffset];
|
|
|
|
|
vp9_filter_block1d16_h6_only_sse2(src_ptr, src_pixels_per_line,
|
|
|
|
|
dst_ptr, dst_pitch, 16, hfilter);
|
2010-05-18 17:58:33 +02:00
|
|
|
}
|
2012-07-14 00:21:29 +02:00
|
|
|
} else {
|
|
|
|
|
/* Second-pass only */
|
2012-11-01 00:09:17 +01:00
|
|
|
vfilter = vp9_six_tap_mmx[yoffset];
|
|
|
|
|
vp9_unpack_block1d16_h6_sse2(src_ptr - (2 * src_pixels_per_line), fdata2,
|
|
|
|
|
src_pixels_per_line, 21, 32);
|
|
|
|
|
vp9_filter_block1d16_v6_sse2(fdata2 + 32, dst_ptr, dst_pitch,
|
|
|
|
|
32, 16, 16, dst_pitch, vfilter);
|
2012-07-14 00:21:29 +02:00
|
|
|
}
|
2010-05-18 17:58:33 +02:00
|
|
|
}
|
|
|
|
|
|
2012-11-01 00:09:17 +01:00
|
|
|
void vp9_sixtap_predict8x8_sse2(unsigned char *src_ptr,
|
|
|
|
|
int src_pixels_per_line,
|
|
|
|
|
int xoffset,
|
|
|
|
|
int yoffset,
|
|
|
|
|
unsigned char *dst_ptr,
|
|
|
|
|
int dst_pitch) {
|
|
|
|
|
/* Temp data bufffer used in filtering */
|
|
|
|
|
DECLARE_ALIGNED_ARRAY(16, unsigned short, fdata2, 256);
|
|
|
|
|
const short *hfilter, *vfilter;
|
WebM Experimental Codec Branch Snapshot
This is a code snapshot of experimental work currently ongoing for a
next-generation codec.
The codebase has been cut down considerably from the libvpx baseline.
For example, we are currently only supporting VBR 2-pass rate control
and have removed most of the code relating to coding speed, threading,
error resilience, partitions and various other features. This is in
part to make the codebase easier to work on and experiment with, but
also because we want to have an open discussion about how the bitstream
will be structured and partitioned and not have that conversation
constrained by past work.
Our basic working pattern has been to initially encapsulate experiments
using configure options linked to #IF CONFIG_XXX statements in the
code. Once experiments have matured and we are reasonably happy that
they give benefit and can be merged without breaking other experiments,
we remove the conditional compile statements and merge them in.
Current changes include:
* Temporal coding experiment for segments (though still only 4 max, it
will likely be increased).
* Segment feature experiment - to allow various bits of information to
be coded at the segment level. Features tested so far include mode
and reference frame information, limiting end of block offset and
transform size, alongside Q and loop filter parameters, but this set
is very fluid.
* Support for 8x8 transform - 8x8 dct with 2nd order 2x2 haar is used
in MBs using 16x16 prediction modes within inter frames.
* Compound prediction (combination of signals from existing predictors
to create a new predictor).
* 8 tap interpolation filters and 1/8th pel motion vectors.
* Loop filter modifications.
* Various entropy modifications and changes to how entropy contexts and
updates are handled.
* Extended quantizer range matched to transform precision improvements.
There are also ongoing further experiments that we hope to merge in the
near future: For example, coding of motion and other aspects of the
prediction signal to better support larger image formats, use of larger
block sizes (e.g. 32x32 and up) and lossless non-transform based coding
options (especially for key frames). It is our hope that we will be
able to make regular updates and we will warmly welcome community
contributions.
Please be warned that, at this stage, the codebase is currently slower
than VP8 stable branch as most new code has not been optimized, and
even the 'C' has been deliberately written to be simple and obvious,
not fast.
The following graphs have the initial test results, numbers in the
tables measure the compression improvement in terms of percentage. The
build has the following optional experiments configured:
--enable-experimental --enable-enhanced_interp --enable-uvintra
--enable-high_precision_mv --enable-sixteenth_subpel_uv
CIF Size clips:
http://getwebm.org/tmp/cif/
HD size clips:
http://getwebm.org/tmp/hd/
(stable_20120309 represents encoding results of WebM master branch
build as of commit#7a15907)
They were encoded using the following encode parameters:
--good --cpu-used=0 -t 0 --lag-in-frames=25 --min-q=0 --max-q=63
--end-usage=0 --auto-alt-ref=1 -p 2 --pass=2 --kf-max-dist=9999
--kf-min-dist=0 --drop-frame=0 --static-thresh=0 --bias-pct=50
--minsection-pct=0 --maxsection-pct=800 --sharpness=0
--arnr-maxframes=7 --arnr-strength=3(for HD,6 for CIF)
--arnr-type=3
Change-Id: I5c62ed09cfff5815a2bb34e7820d6a810c23183c
2012-03-10 02:32:50 +01:00
|
|
|
#ifdef ANNOUNCE_FUNCTION
|
2012-10-31 00:25:53 +01:00
|
|
|
printf("vp9_sixtap_predict8x8_sse2\n");
|
WebM Experimental Codec Branch Snapshot
This is a code snapshot of experimental work currently ongoing for a
next-generation codec.
The codebase has been cut down considerably from the libvpx baseline.
For example, we are currently only supporting VBR 2-pass rate control
and have removed most of the code relating to coding speed, threading,
error resilience, partitions and various other features. This is in
part to make the codebase easier to work on and experiment with, but
also because we want to have an open discussion about how the bitstream
will be structured and partitioned and not have that conversation
constrained by past work.
Our basic working pattern has been to initially encapsulate experiments
using configure options linked to #IF CONFIG_XXX statements in the
code. Once experiments have matured and we are reasonably happy that
they give benefit and can be merged without breaking other experiments,
we remove the conditional compile statements and merge them in.
Current changes include:
* Temporal coding experiment for segments (though still only 4 max, it
will likely be increased).
* Segment feature experiment - to allow various bits of information to
be coded at the segment level. Features tested so far include mode
and reference frame information, limiting end of block offset and
transform size, alongside Q and loop filter parameters, but this set
is very fluid.
* Support for 8x8 transform - 8x8 dct with 2nd order 2x2 haar is used
in MBs using 16x16 prediction modes within inter frames.
* Compound prediction (combination of signals from existing predictors
to create a new predictor).
* 8 tap interpolation filters and 1/8th pel motion vectors.
* Loop filter modifications.
* Various entropy modifications and changes to how entropy contexts and
updates are handled.
* Extended quantizer range matched to transform precision improvements.
There are also ongoing further experiments that we hope to merge in the
near future: For example, coding of motion and other aspects of the
prediction signal to better support larger image formats, use of larger
block sizes (e.g. 32x32 and up) and lossless non-transform based coding
options (especially for key frames). It is our hope that we will be
able to make regular updates and we will warmly welcome community
contributions.
Please be warned that, at this stage, the codebase is currently slower
than VP8 stable branch as most new code has not been optimized, and
even the 'C' has been deliberately written to be simple and obvious,
not fast.
The following graphs have the initial test results, numbers in the
tables measure the compression improvement in terms of percentage. The
build has the following optional experiments configured:
--enable-experimental --enable-enhanced_interp --enable-uvintra
--enable-high_precision_mv --enable-sixteenth_subpel_uv
CIF Size clips:
http://getwebm.org/tmp/cif/
HD size clips:
http://getwebm.org/tmp/hd/
(stable_20120309 represents encoding results of WebM master branch
build as of commit#7a15907)
They were encoded using the following encode parameters:
--good --cpu-used=0 -t 0 --lag-in-frames=25 --min-q=0 --max-q=63
--end-usage=0 --auto-alt-ref=1 -p 2 --pass=2 --kf-max-dist=9999
--kf-min-dist=0 --drop-frame=0 --static-thresh=0 --bias-pct=50
--minsection-pct=0 --maxsection-pct=800 --sharpness=0
--arnr-maxframes=7 --arnr-strength=3(for HD,6 for CIF)
--arnr-type=3
Change-Id: I5c62ed09cfff5815a2bb34e7820d6a810c23183c
2012-03-10 02:32:50 +01:00
|
|
|
#endif
|
2010-05-18 17:58:33 +02:00
|
|
|
|
2012-07-14 00:21:29 +02:00
|
|
|
if (xoffset) {
|
|
|
|
|
if (yoffset) {
|
2012-11-01 00:09:17 +01:00
|
|
|
hfilter = vp9_six_tap_mmx[xoffset];
|
|
|
|
|
vp9_filter_block1d8_h6_sse2(src_ptr - (2 * src_pixels_per_line), fdata2,
|
|
|
|
|
src_pixels_per_line, 1, 13, 16, hfilter);
|
|
|
|
|
vfilter = vp9_six_tap_mmx[yoffset];
|
|
|
|
|
vp9_filter_block1d8_v6_sse2(fdata2 + 16, dst_ptr, dst_pitch,
|
|
|
|
|
16, 8, 8, dst_pitch, vfilter);
|
2012-07-14 00:21:29 +02:00
|
|
|
} else {
|
|
|
|
|
/* First-pass only */
|
2012-11-01 00:09:17 +01:00
|
|
|
hfilter = vp9_six_tap_mmx[xoffset];
|
|
|
|
|
vp9_filter_block1d8_h6_only_sse2(src_ptr, src_pixels_per_line,
|
|
|
|
|
dst_ptr, dst_pitch, 8, hfilter);
|
2010-05-18 17:58:33 +02:00
|
|
|
}
|
2012-07-14 00:21:29 +02:00
|
|
|
} else {
|
|
|
|
|
/* Second-pass only */
|
2012-11-01 00:09:17 +01:00
|
|
|
vfilter = vp9_six_tap_mmx[yoffset];
|
|
|
|
|
vp9_filter_block1d8_v6_only_sse2(src_ptr - (2 * src_pixels_per_line),
|
|
|
|
|
src_pixels_per_line,
|
|
|
|
|
dst_ptr, dst_pitch, 8, vfilter);
|
2012-07-14 00:21:29 +02:00
|
|
|
}
|
2010-05-18 17:58:33 +02:00
|
|
|
}
|
|
|
|
|
|
2012-11-01 00:09:17 +01:00
|
|
|
void vp9_sixtap_predict8x4_sse2(unsigned char *src_ptr,
|
|
|
|
|
int src_pixels_per_line,
|
|
|
|
|
int xoffset,
|
|
|
|
|
int yoffset,
|
|
|
|
|
unsigned char *dst_ptr,
|
|
|
|
|
int dst_pitch) {
|
|
|
|
|
/* Temp data bufffer used in filtering */
|
|
|
|
|
DECLARE_ALIGNED_ARRAY(16, unsigned short, fdata2, 256);
|
|
|
|
|
const short *hfilter, *vfilter;
|
WebM Experimental Codec Branch Snapshot
This is a code snapshot of experimental work currently ongoing for a
next-generation codec.
The codebase has been cut down considerably from the libvpx baseline.
For example, we are currently only supporting VBR 2-pass rate control
and have removed most of the code relating to coding speed, threading,
error resilience, partitions and various other features. This is in
part to make the codebase easier to work on and experiment with, but
also because we want to have an open discussion about how the bitstream
will be structured and partitioned and not have that conversation
constrained by past work.
Our basic working pattern has been to initially encapsulate experiments
using configure options linked to #IF CONFIG_XXX statements in the
code. Once experiments have matured and we are reasonably happy that
they give benefit and can be merged without breaking other experiments,
we remove the conditional compile statements and merge them in.
Current changes include:
* Temporal coding experiment for segments (though still only 4 max, it
will likely be increased).
* Segment feature experiment - to allow various bits of information to
be coded at the segment level. Features tested so far include mode
and reference frame information, limiting end of block offset and
transform size, alongside Q and loop filter parameters, but this set
is very fluid.
* Support for 8x8 transform - 8x8 dct with 2nd order 2x2 haar is used
in MBs using 16x16 prediction modes within inter frames.
* Compound prediction (combination of signals from existing predictors
to create a new predictor).
* 8 tap interpolation filters and 1/8th pel motion vectors.
* Loop filter modifications.
* Various entropy modifications and changes to how entropy contexts and
updates are handled.
* Extended quantizer range matched to transform precision improvements.
There are also ongoing further experiments that we hope to merge in the
near future: For example, coding of motion and other aspects of the
prediction signal to better support larger image formats, use of larger
block sizes (e.g. 32x32 and up) and lossless non-transform based coding
options (especially for key frames). It is our hope that we will be
able to make regular updates and we will warmly welcome community
contributions.
Please be warned that, at this stage, the codebase is currently slower
than VP8 stable branch as most new code has not been optimized, and
even the 'C' has been deliberately written to be simple and obvious,
not fast.
The following graphs have the initial test results, numbers in the
tables measure the compression improvement in terms of percentage. The
build has the following optional experiments configured:
--enable-experimental --enable-enhanced_interp --enable-uvintra
--enable-high_precision_mv --enable-sixteenth_subpel_uv
CIF Size clips:
http://getwebm.org/tmp/cif/
HD size clips:
http://getwebm.org/tmp/hd/
(stable_20120309 represents encoding results of WebM master branch
build as of commit#7a15907)
They were encoded using the following encode parameters:
--good --cpu-used=0 -t 0 --lag-in-frames=25 --min-q=0 --max-q=63
--end-usage=0 --auto-alt-ref=1 -p 2 --pass=2 --kf-max-dist=9999
--kf-min-dist=0 --drop-frame=0 --static-thresh=0 --bias-pct=50
--minsection-pct=0 --maxsection-pct=800 --sharpness=0
--arnr-maxframes=7 --arnr-strength=3(for HD,6 for CIF)
--arnr-type=3
Change-Id: I5c62ed09cfff5815a2bb34e7820d6a810c23183c
2012-03-10 02:32:50 +01:00
|
|
|
#ifdef ANNOUNCE_FUNCTION
|
2012-10-31 00:25:53 +01:00
|
|
|
printf("vp9_sixtap_predict8x4_sse2\n");
|
WebM Experimental Codec Branch Snapshot
This is a code snapshot of experimental work currently ongoing for a
next-generation codec.
The codebase has been cut down considerably from the libvpx baseline.
For example, we are currently only supporting VBR 2-pass rate control
and have removed most of the code relating to coding speed, threading,
error resilience, partitions and various other features. This is in
part to make the codebase easier to work on and experiment with, but
also because we want to have an open discussion about how the bitstream
will be structured and partitioned and not have that conversation
constrained by past work.
Our basic working pattern has been to initially encapsulate experiments
using configure options linked to #IF CONFIG_XXX statements in the
code. Once experiments have matured and we are reasonably happy that
they give benefit and can be merged without breaking other experiments,
we remove the conditional compile statements and merge them in.
Current changes include:
* Temporal coding experiment for segments (though still only 4 max, it
will likely be increased).
* Segment feature experiment - to allow various bits of information to
be coded at the segment level. Features tested so far include mode
and reference frame information, limiting end of block offset and
transform size, alongside Q and loop filter parameters, but this set
is very fluid.
* Support for 8x8 transform - 8x8 dct with 2nd order 2x2 haar is used
in MBs using 16x16 prediction modes within inter frames.
* Compound prediction (combination of signals from existing predictors
to create a new predictor).
* 8 tap interpolation filters and 1/8th pel motion vectors.
* Loop filter modifications.
* Various entropy modifications and changes to how entropy contexts and
updates are handled.
* Extended quantizer range matched to transform precision improvements.
There are also ongoing further experiments that we hope to merge in the
near future: For example, coding of motion and other aspects of the
prediction signal to better support larger image formats, use of larger
block sizes (e.g. 32x32 and up) and lossless non-transform based coding
options (especially for key frames). It is our hope that we will be
able to make regular updates and we will warmly welcome community
contributions.
Please be warned that, at this stage, the codebase is currently slower
than VP8 stable branch as most new code has not been optimized, and
even the 'C' has been deliberately written to be simple and obvious,
not fast.
The following graphs have the initial test results, numbers in the
tables measure the compression improvement in terms of percentage. The
build has the following optional experiments configured:
--enable-experimental --enable-enhanced_interp --enable-uvintra
--enable-high_precision_mv --enable-sixteenth_subpel_uv
CIF Size clips:
http://getwebm.org/tmp/cif/
HD size clips:
http://getwebm.org/tmp/hd/
(stable_20120309 represents encoding results of WebM master branch
build as of commit#7a15907)
They were encoded using the following encode parameters:
--good --cpu-used=0 -t 0 --lag-in-frames=25 --min-q=0 --max-q=63
--end-usage=0 --auto-alt-ref=1 -p 2 --pass=2 --kf-max-dist=9999
--kf-min-dist=0 --drop-frame=0 --static-thresh=0 --bias-pct=50
--minsection-pct=0 --maxsection-pct=800 --sharpness=0
--arnr-maxframes=7 --arnr-strength=3(for HD,6 for CIF)
--arnr-type=3
Change-Id: I5c62ed09cfff5815a2bb34e7820d6a810c23183c
2012-03-10 02:32:50 +01:00
|
|
|
#endif
|
2010-05-18 17:58:33 +02:00
|
|
|
|
2012-07-14 00:21:29 +02:00
|
|
|
if (xoffset) {
|
|
|
|
|
if (yoffset) {
|
2012-11-01 00:09:17 +01:00
|
|
|
hfilter = vp9_six_tap_mmx[xoffset];
|
|
|
|
|
vp9_filter_block1d8_h6_sse2(src_ptr - (2 * src_pixels_per_line), fdata2,
|
|
|
|
|
src_pixels_per_line, 1, 9, 16, hfilter);
|
|
|
|
|
vfilter = vp9_six_tap_mmx[yoffset];
|
|
|
|
|
vp9_filter_block1d8_v6_sse2(fdata2 + 16, dst_ptr, dst_pitch,
|
|
|
|
|
16, 8, 4, dst_pitch, vfilter);
|
2012-07-14 00:21:29 +02:00
|
|
|
} else {
|
|
|
|
|
/* First-pass only */
|
2012-11-01 00:09:17 +01:00
|
|
|
hfilter = vp9_six_tap_mmx[xoffset];
|
|
|
|
|
vp9_filter_block1d8_h6_only_sse2(src_ptr, src_pixels_per_line,
|
|
|
|
|
dst_ptr, dst_pitch, 4, hfilter);
|
2010-05-18 17:58:33 +02:00
|
|
|
}
|
2012-07-14 00:21:29 +02:00
|
|
|
} else {
|
|
|
|
|
/* Second-pass only */
|
2012-11-01 00:09:17 +01:00
|
|
|
vfilter = vp9_six_tap_mmx[yoffset];
|
|
|
|
|
vp9_filter_block1d8_v6_only_sse2(src_ptr - (2 * src_pixels_per_line),
|
|
|
|
|
src_pixels_per_line,
|
|
|
|
|
dst_ptr, dst_pitch, 4, vfilter);
|
2012-07-14 00:21:29 +02:00
|
|
|
}
|
2010-05-18 17:58:33 +02:00
|
|
|
}
|
|
|
|
|
#endif
|
2010-08-10 23:06:05 +02:00
|
|
|
|
2010-08-19 21:50:29 +02:00
|
|
|
#if HAVE_SSSE3
|
2012-11-01 00:09:17 +01:00
|
|
|
extern void vp9_filter_block1d8_h6_ssse3(unsigned char *src_ptr,
|
|
|
|
|
unsigned int src_pixels_per_line,
|
|
|
|
|
unsigned char *output_ptr,
|
|
|
|
|
unsigned int output_pitch,
|
|
|
|
|
unsigned int output_height,
|
|
|
|
|
unsigned int vp9_filter_index);
|
|
|
|
|
|
|
|
|
|
extern void vp9_filter_block1d16_h6_ssse3(unsigned char *src_ptr,
|
|
|
|
|
unsigned int src_pixels_per_line,
|
|
|
|
|
unsigned char *output_ptr,
|
|
|
|
|
unsigned int output_pitch,
|
|
|
|
|
unsigned int output_height,
|
|
|
|
|
unsigned int vp9_filter_index);
|
|
|
|
|
|
|
|
|
|
extern void vp9_filter_block1d16_v6_ssse3(unsigned char *src_ptr,
|
|
|
|
|
unsigned int src_pitch,
|
|
|
|
|
unsigned char *output_ptr,
|
|
|
|
|
unsigned int out_pitch,
|
|
|
|
|
unsigned int output_height,
|
|
|
|
|
unsigned int vp9_filter_index);
|
|
|
|
|
|
|
|
|
|
extern void vp9_filter_block1d8_v6_ssse3(unsigned char *src_ptr,
|
|
|
|
|
unsigned int src_pitch,
|
|
|
|
|
unsigned char *output_ptr,
|
|
|
|
|
unsigned int out_pitch,
|
|
|
|
|
unsigned int output_height,
|
|
|
|
|
unsigned int vp9_filter_index);
|
|
|
|
|
|
|
|
|
|
extern void vp9_filter_block1d4_h6_ssse3(unsigned char *src_ptr,
|
|
|
|
|
unsigned int src_pixels_per_line,
|
|
|
|
|
unsigned char *output_ptr,
|
|
|
|
|
unsigned int output_pitch,
|
|
|
|
|
unsigned int output_height,
|
|
|
|
|
unsigned int vp9_filter_index);
|
|
|
|
|
|
|
|
|
|
extern void vp9_filter_block1d4_v6_ssse3(unsigned char *src_ptr,
|
|
|
|
|
unsigned int src_pitch,
|
|
|
|
|
unsigned char *output_ptr,
|
|
|
|
|
unsigned int out_pitch,
|
|
|
|
|
unsigned int output_height,
|
|
|
|
|
unsigned int vp9_filter_index);
|
|
|
|
|
|
|
|
|
|
void vp9_sixtap_predict16x16_ssse3(unsigned char *src_ptr,
|
|
|
|
|
int src_pixels_per_line,
|
|
|
|
|
int xoffset,
|
|
|
|
|
int yoffset,
|
|
|
|
|
unsigned char *dst_ptr,
|
|
|
|
|
int dst_pitch) {
|
|
|
|
|
DECLARE_ALIGNED_ARRAY(16, unsigned char, fdata2, 24 * 24);
|
WebM Experimental Codec Branch Snapshot
This is a code snapshot of experimental work currently ongoing for a
next-generation codec.
The codebase has been cut down considerably from the libvpx baseline.
For example, we are currently only supporting VBR 2-pass rate control
and have removed most of the code relating to coding speed, threading,
error resilience, partitions and various other features. This is in
part to make the codebase easier to work on and experiment with, but
also because we want to have an open discussion about how the bitstream
will be structured and partitioned and not have that conversation
constrained by past work.
Our basic working pattern has been to initially encapsulate experiments
using configure options linked to #IF CONFIG_XXX statements in the
code. Once experiments have matured and we are reasonably happy that
they give benefit and can be merged without breaking other experiments,
we remove the conditional compile statements and merge them in.
Current changes include:
* Temporal coding experiment for segments (though still only 4 max, it
will likely be increased).
* Segment feature experiment - to allow various bits of information to
be coded at the segment level. Features tested so far include mode
and reference frame information, limiting end of block offset and
transform size, alongside Q and loop filter parameters, but this set
is very fluid.
* Support for 8x8 transform - 8x8 dct with 2nd order 2x2 haar is used
in MBs using 16x16 prediction modes within inter frames.
* Compound prediction (combination of signals from existing predictors
to create a new predictor).
* 8 tap interpolation filters and 1/8th pel motion vectors.
* Loop filter modifications.
* Various entropy modifications and changes to how entropy contexts and
updates are handled.
* Extended quantizer range matched to transform precision improvements.
There are also ongoing further experiments that we hope to merge in the
near future: For example, coding of motion and other aspects of the
prediction signal to better support larger image formats, use of larger
block sizes (e.g. 32x32 and up) and lossless non-transform based coding
options (especially for key frames). It is our hope that we will be
able to make regular updates and we will warmly welcome community
contributions.
Please be warned that, at this stage, the codebase is currently slower
than VP8 stable branch as most new code has not been optimized, and
even the 'C' has been deliberately written to be simple and obvious,
not fast.
The following graphs have the initial test results, numbers in the
tables measure the compression improvement in terms of percentage. The
build has the following optional experiments configured:
--enable-experimental --enable-enhanced_interp --enable-uvintra
--enable-high_precision_mv --enable-sixteenth_subpel_uv
CIF Size clips:
http://getwebm.org/tmp/cif/
HD size clips:
http://getwebm.org/tmp/hd/
(stable_20120309 represents encoding results of WebM master branch
build as of commit#7a15907)
They were encoded using the following encode parameters:
--good --cpu-used=0 -t 0 --lag-in-frames=25 --min-q=0 --max-q=63
--end-usage=0 --auto-alt-ref=1 -p 2 --pass=2 --kf-max-dist=9999
--kf-min-dist=0 --drop-frame=0 --static-thresh=0 --bias-pct=50
--minsection-pct=0 --maxsection-pct=800 --sharpness=0
--arnr-maxframes=7 --arnr-strength=3(for HD,6 for CIF)
--arnr-type=3
Change-Id: I5c62ed09cfff5815a2bb34e7820d6a810c23183c
2012-03-10 02:32:50 +01:00
|
|
|
#ifdef ANNOUNCE_FUNCTION
|
2012-10-31 00:25:53 +01:00
|
|
|
printf("vp9_sixtap_predict16x16_ssse3\n");
|
WebM Experimental Codec Branch Snapshot
This is a code snapshot of experimental work currently ongoing for a
next-generation codec.
The codebase has been cut down considerably from the libvpx baseline.
For example, we are currently only supporting VBR 2-pass rate control
and have removed most of the code relating to coding speed, threading,
error resilience, partitions and various other features. This is in
part to make the codebase easier to work on and experiment with, but
also because we want to have an open discussion about how the bitstream
will be structured and partitioned and not have that conversation
constrained by past work.
Our basic working pattern has been to initially encapsulate experiments
using configure options linked to #IF CONFIG_XXX statements in the
code. Once experiments have matured and we are reasonably happy that
they give benefit and can be merged without breaking other experiments,
we remove the conditional compile statements and merge them in.
Current changes include:
* Temporal coding experiment for segments (though still only 4 max, it
will likely be increased).
* Segment feature experiment - to allow various bits of information to
be coded at the segment level. Features tested so far include mode
and reference frame information, limiting end of block offset and
transform size, alongside Q and loop filter parameters, but this set
is very fluid.
* Support for 8x8 transform - 8x8 dct with 2nd order 2x2 haar is used
in MBs using 16x16 prediction modes within inter frames.
* Compound prediction (combination of signals from existing predictors
to create a new predictor).
* 8 tap interpolation filters and 1/8th pel motion vectors.
* Loop filter modifications.
* Various entropy modifications and changes to how entropy contexts and
updates are handled.
* Extended quantizer range matched to transform precision improvements.
There are also ongoing further experiments that we hope to merge in the
near future: For example, coding of motion and other aspects of the
prediction signal to better support larger image formats, use of larger
block sizes (e.g. 32x32 and up) and lossless non-transform based coding
options (especially for key frames). It is our hope that we will be
able to make regular updates and we will warmly welcome community
contributions.
Please be warned that, at this stage, the codebase is currently slower
than VP8 stable branch as most new code has not been optimized, and
even the 'C' has been deliberately written to be simple and obvious,
not fast.
The following graphs have the initial test results, numbers in the
tables measure the compression improvement in terms of percentage. The
build has the following optional experiments configured:
--enable-experimental --enable-enhanced_interp --enable-uvintra
--enable-high_precision_mv --enable-sixteenth_subpel_uv
CIF Size clips:
http://getwebm.org/tmp/cif/
HD size clips:
http://getwebm.org/tmp/hd/
(stable_20120309 represents encoding results of WebM master branch
build as of commit#7a15907)
They were encoded using the following encode parameters:
--good --cpu-used=0 -t 0 --lag-in-frames=25 --min-q=0 --max-q=63
--end-usage=0 --auto-alt-ref=1 -p 2 --pass=2 --kf-max-dist=9999
--kf-min-dist=0 --drop-frame=0 --static-thresh=0 --bias-pct=50
--minsection-pct=0 --maxsection-pct=800 --sharpness=0
--arnr-maxframes=7 --arnr-strength=3(for HD,6 for CIF)
--arnr-type=3
Change-Id: I5c62ed09cfff5815a2bb34e7820d6a810c23183c
2012-03-10 02:32:50 +01:00
|
|
|
#endif
|
2010-08-19 21:50:29 +02:00
|
|
|
|
2012-07-14 00:21:29 +02:00
|
|
|
if (xoffset) {
|
|
|
|
|
if (yoffset) {
|
2012-11-01 00:09:17 +01:00
|
|
|
vp9_filter_block1d16_h6_ssse3(src_ptr - (2 * src_pixels_per_line),
|
|
|
|
|
src_pixels_per_line,
|
|
|
|
|
fdata2, 16, 21, xoffset);
|
|
|
|
|
vp9_filter_block1d16_v6_ssse3(fdata2, 16, dst_ptr, dst_pitch,
|
|
|
|
|
16, yoffset);
|
2012-07-14 00:21:29 +02:00
|
|
|
} else {
|
|
|
|
|
/* First-pass only */
|
2012-11-01 00:09:17 +01:00
|
|
|
vp9_filter_block1d16_h6_ssse3(src_ptr, src_pixels_per_line,
|
|
|
|
|
dst_ptr, dst_pitch, 16, xoffset);
|
2010-08-19 21:50:29 +02:00
|
|
|
}
|
2012-07-14 00:21:29 +02:00
|
|
|
} else {
|
|
|
|
|
/* Second-pass only */
|
2012-11-01 00:09:17 +01:00
|
|
|
vp9_filter_block1d16_v6_ssse3(src_ptr - (2 * src_pixels_per_line),
|
|
|
|
|
src_pixels_per_line,
|
|
|
|
|
dst_ptr, dst_pitch, 16, yoffset);
|
2012-07-14 00:21:29 +02:00
|
|
|
}
|
2010-08-19 21:50:29 +02:00
|
|
|
}
|
|
|
|
|
|
2012-11-01 00:09:17 +01:00
|
|
|
void vp9_sixtap_predict8x8_ssse3(unsigned char *src_ptr,
|
|
|
|
|
int src_pixels_per_line,
|
|
|
|
|
int xoffset,
|
|
|
|
|
int yoffset,
|
|
|
|
|
unsigned char *dst_ptr,
|
|
|
|
|
int dst_pitch) {
|
|
|
|
|
DECLARE_ALIGNED_ARRAY(16, unsigned char, fdata2, 256);
|
WebM Experimental Codec Branch Snapshot
This is a code snapshot of experimental work currently ongoing for a
next-generation codec.
The codebase has been cut down considerably from the libvpx baseline.
For example, we are currently only supporting VBR 2-pass rate control
and have removed most of the code relating to coding speed, threading,
error resilience, partitions and various other features. This is in
part to make the codebase easier to work on and experiment with, but
also because we want to have an open discussion about how the bitstream
will be structured and partitioned and not have that conversation
constrained by past work.
Our basic working pattern has been to initially encapsulate experiments
using configure options linked to #IF CONFIG_XXX statements in the
code. Once experiments have matured and we are reasonably happy that
they give benefit and can be merged without breaking other experiments,
we remove the conditional compile statements and merge them in.
Current changes include:
* Temporal coding experiment for segments (though still only 4 max, it
will likely be increased).
* Segment feature experiment - to allow various bits of information to
be coded at the segment level. Features tested so far include mode
and reference frame information, limiting end of block offset and
transform size, alongside Q and loop filter parameters, but this set
is very fluid.
* Support for 8x8 transform - 8x8 dct with 2nd order 2x2 haar is used
in MBs using 16x16 prediction modes within inter frames.
* Compound prediction (combination of signals from existing predictors
to create a new predictor).
* 8 tap interpolation filters and 1/8th pel motion vectors.
* Loop filter modifications.
* Various entropy modifications and changes to how entropy contexts and
updates are handled.
* Extended quantizer range matched to transform precision improvements.
There are also ongoing further experiments that we hope to merge in the
near future: For example, coding of motion and other aspects of the
prediction signal to better support larger image formats, use of larger
block sizes (e.g. 32x32 and up) and lossless non-transform based coding
options (especially for key frames). It is our hope that we will be
able to make regular updates and we will warmly welcome community
contributions.
Please be warned that, at this stage, the codebase is currently slower
than VP8 stable branch as most new code has not been optimized, and
even the 'C' has been deliberately written to be simple and obvious,
not fast.
The following graphs have the initial test results, numbers in the
tables measure the compression improvement in terms of percentage. The
build has the following optional experiments configured:
--enable-experimental --enable-enhanced_interp --enable-uvintra
--enable-high_precision_mv --enable-sixteenth_subpel_uv
CIF Size clips:
http://getwebm.org/tmp/cif/
HD size clips:
http://getwebm.org/tmp/hd/
(stable_20120309 represents encoding results of WebM master branch
build as of commit#7a15907)
They were encoded using the following encode parameters:
--good --cpu-used=0 -t 0 --lag-in-frames=25 --min-q=0 --max-q=63
--end-usage=0 --auto-alt-ref=1 -p 2 --pass=2 --kf-max-dist=9999
--kf-min-dist=0 --drop-frame=0 --static-thresh=0 --bias-pct=50
--minsection-pct=0 --maxsection-pct=800 --sharpness=0
--arnr-maxframes=7 --arnr-strength=3(for HD,6 for CIF)
--arnr-type=3
Change-Id: I5c62ed09cfff5815a2bb34e7820d6a810c23183c
2012-03-10 02:32:50 +01:00
|
|
|
#ifdef ANNOUNCE_FUNCTION
|
2012-10-31 00:25:53 +01:00
|
|
|
printf("vp9_sixtap_predict8x8_ssse3\n");
|
WebM Experimental Codec Branch Snapshot
This is a code snapshot of experimental work currently ongoing for a
next-generation codec.
The codebase has been cut down considerably from the libvpx baseline.
For example, we are currently only supporting VBR 2-pass rate control
and have removed most of the code relating to coding speed, threading,
error resilience, partitions and various other features. This is in
part to make the codebase easier to work on and experiment with, but
also because we want to have an open discussion about how the bitstream
will be structured and partitioned and not have that conversation
constrained by past work.
Our basic working pattern has been to initially encapsulate experiments
using configure options linked to #IF CONFIG_XXX statements in the
code. Once experiments have matured and we are reasonably happy that
they give benefit and can be merged without breaking other experiments,
we remove the conditional compile statements and merge them in.
Current changes include:
* Temporal coding experiment for segments (though still only 4 max, it
will likely be increased).
* Segment feature experiment - to allow various bits of information to
be coded at the segment level. Features tested so far include mode
and reference frame information, limiting end of block offset and
transform size, alongside Q and loop filter parameters, but this set
is very fluid.
* Support for 8x8 transform - 8x8 dct with 2nd order 2x2 haar is used
in MBs using 16x16 prediction modes within inter frames.
* Compound prediction (combination of signals from existing predictors
to create a new predictor).
* 8 tap interpolation filters and 1/8th pel motion vectors.
* Loop filter modifications.
* Various entropy modifications and changes to how entropy contexts and
updates are handled.
* Extended quantizer range matched to transform precision improvements.
There are also ongoing further experiments that we hope to merge in the
near future: For example, coding of motion and other aspects of the
prediction signal to better support larger image formats, use of larger
block sizes (e.g. 32x32 and up) and lossless non-transform based coding
options (especially for key frames). It is our hope that we will be
able to make regular updates and we will warmly welcome community
contributions.
Please be warned that, at this stage, the codebase is currently slower
than VP8 stable branch as most new code has not been optimized, and
even the 'C' has been deliberately written to be simple and obvious,
not fast.
The following graphs have the initial test results, numbers in the
tables measure the compression improvement in terms of percentage. The
build has the following optional experiments configured:
--enable-experimental --enable-enhanced_interp --enable-uvintra
--enable-high_precision_mv --enable-sixteenth_subpel_uv
CIF Size clips:
http://getwebm.org/tmp/cif/
HD size clips:
http://getwebm.org/tmp/hd/
(stable_20120309 represents encoding results of WebM master branch
build as of commit#7a15907)
They were encoded using the following encode parameters:
--good --cpu-used=0 -t 0 --lag-in-frames=25 --min-q=0 --max-q=63
--end-usage=0 --auto-alt-ref=1 -p 2 --pass=2 --kf-max-dist=9999
--kf-min-dist=0 --drop-frame=0 --static-thresh=0 --bias-pct=50
--minsection-pct=0 --maxsection-pct=800 --sharpness=0
--arnr-maxframes=7 --arnr-strength=3(for HD,6 for CIF)
--arnr-type=3
Change-Id: I5c62ed09cfff5815a2bb34e7820d6a810c23183c
2012-03-10 02:32:50 +01:00
|
|
|
#endif
|
2010-08-19 21:50:29 +02:00
|
|
|
|
2012-07-14 00:21:29 +02:00
|
|
|
if (xoffset) {
|
|
|
|
|
if (yoffset) {
|
2012-11-01 00:09:17 +01:00
|
|
|
vp9_filter_block1d8_h6_ssse3(src_ptr - (2 * src_pixels_per_line),
|
|
|
|
|
src_pixels_per_line, fdata2, 8, 13, xoffset);
|
|
|
|
|
vp9_filter_block1d8_v6_ssse3(fdata2, 8, dst_ptr, dst_pitch, 8, yoffset);
|
2012-07-14 00:21:29 +02:00
|
|
|
} else {
|
2012-11-01 00:09:17 +01:00
|
|
|
vp9_filter_block1d8_h6_ssse3(src_ptr, src_pixels_per_line,
|
|
|
|
|
dst_ptr, dst_pitch, 8, xoffset);
|
2010-08-19 21:50:29 +02:00
|
|
|
}
|
2012-07-14 00:21:29 +02:00
|
|
|
} else {
|
|
|
|
|
/* Second-pass only */
|
2012-11-01 00:09:17 +01:00
|
|
|
vp9_filter_block1d8_v6_ssse3(src_ptr - (2 * src_pixels_per_line),
|
|
|
|
|
src_pixels_per_line,
|
|
|
|
|
dst_ptr, dst_pitch, 8, yoffset);
|
2012-07-14 00:21:29 +02:00
|
|
|
}
|
2010-08-19 21:50:29 +02:00
|
|
|
}
|
|
|
|
|
|
2012-11-01 00:09:17 +01:00
|
|
|
void vp9_sixtap_predict8x4_ssse3(unsigned char *src_ptr,
|
|
|
|
|
int src_pixels_per_line,
|
|
|
|
|
int xoffset,
|
|
|
|
|
int yoffset,
|
|
|
|
|
unsigned char *dst_ptr,
|
|
|
|
|
int dst_pitch) {
|
|
|
|
|
DECLARE_ALIGNED_ARRAY(16, unsigned char, fdata2, 256);
|
WebM Experimental Codec Branch Snapshot
This is a code snapshot of experimental work currently ongoing for a
next-generation codec.
The codebase has been cut down considerably from the libvpx baseline.
For example, we are currently only supporting VBR 2-pass rate control
and have removed most of the code relating to coding speed, threading,
error resilience, partitions and various other features. This is in
part to make the codebase easier to work on and experiment with, but
also because we want to have an open discussion about how the bitstream
will be structured and partitioned and not have that conversation
constrained by past work.
Our basic working pattern has been to initially encapsulate experiments
using configure options linked to #IF CONFIG_XXX statements in the
code. Once experiments have matured and we are reasonably happy that
they give benefit and can be merged without breaking other experiments,
we remove the conditional compile statements and merge them in.
Current changes include:
* Temporal coding experiment for segments (though still only 4 max, it
will likely be increased).
* Segment feature experiment - to allow various bits of information to
be coded at the segment level. Features tested so far include mode
and reference frame information, limiting end of block offset and
transform size, alongside Q and loop filter parameters, but this set
is very fluid.
* Support for 8x8 transform - 8x8 dct with 2nd order 2x2 haar is used
in MBs using 16x16 prediction modes within inter frames.
* Compound prediction (combination of signals from existing predictors
to create a new predictor).
* 8 tap interpolation filters and 1/8th pel motion vectors.
* Loop filter modifications.
* Various entropy modifications and changes to how entropy contexts and
updates are handled.
* Extended quantizer range matched to transform precision improvements.
There are also ongoing further experiments that we hope to merge in the
near future: For example, coding of motion and other aspects of the
prediction signal to better support larger image formats, use of larger
block sizes (e.g. 32x32 and up) and lossless non-transform based coding
options (especially for key frames). It is our hope that we will be
able to make regular updates and we will warmly welcome community
contributions.
Please be warned that, at this stage, the codebase is currently slower
than VP8 stable branch as most new code has not been optimized, and
even the 'C' has been deliberately written to be simple and obvious,
not fast.
The following graphs have the initial test results, numbers in the
tables measure the compression improvement in terms of percentage. The
build has the following optional experiments configured:
--enable-experimental --enable-enhanced_interp --enable-uvintra
--enable-high_precision_mv --enable-sixteenth_subpel_uv
CIF Size clips:
http://getwebm.org/tmp/cif/
HD size clips:
http://getwebm.org/tmp/hd/
(stable_20120309 represents encoding results of WebM master branch
build as of commit#7a15907)
They were encoded using the following encode parameters:
--good --cpu-used=0 -t 0 --lag-in-frames=25 --min-q=0 --max-q=63
--end-usage=0 --auto-alt-ref=1 -p 2 --pass=2 --kf-max-dist=9999
--kf-min-dist=0 --drop-frame=0 --static-thresh=0 --bias-pct=50
--minsection-pct=0 --maxsection-pct=800 --sharpness=0
--arnr-maxframes=7 --arnr-strength=3(for HD,6 for CIF)
--arnr-type=3
Change-Id: I5c62ed09cfff5815a2bb34e7820d6a810c23183c
2012-03-10 02:32:50 +01:00
|
|
|
#ifdef ANNOUNCE_FUNCTION
|
2012-10-31 00:25:53 +01:00
|
|
|
printf("vp9_sixtap_predict8x4_ssse3\n");
|
WebM Experimental Codec Branch Snapshot
This is a code snapshot of experimental work currently ongoing for a
next-generation codec.
The codebase has been cut down considerably from the libvpx baseline.
For example, we are currently only supporting VBR 2-pass rate control
and have removed most of the code relating to coding speed, threading,
error resilience, partitions and various other features. This is in
part to make the codebase easier to work on and experiment with, but
also because we want to have an open discussion about how the bitstream
will be structured and partitioned and not have that conversation
constrained by past work.
Our basic working pattern has been to initially encapsulate experiments
using configure options linked to #IF CONFIG_XXX statements in the
code. Once experiments have matured and we are reasonably happy that
they give benefit and can be merged without breaking other experiments,
we remove the conditional compile statements and merge them in.
Current changes include:
* Temporal coding experiment for segments (though still only 4 max, it
will likely be increased).
* Segment feature experiment - to allow various bits of information to
be coded at the segment level. Features tested so far include mode
and reference frame information, limiting end of block offset and
transform size, alongside Q and loop filter parameters, but this set
is very fluid.
* Support for 8x8 transform - 8x8 dct with 2nd order 2x2 haar is used
in MBs using 16x16 prediction modes within inter frames.
* Compound prediction (combination of signals from existing predictors
to create a new predictor).
* 8 tap interpolation filters and 1/8th pel motion vectors.
* Loop filter modifications.
* Various entropy modifications and changes to how entropy contexts and
updates are handled.
* Extended quantizer range matched to transform precision improvements.
There are also ongoing further experiments that we hope to merge in the
near future: For example, coding of motion and other aspects of the
prediction signal to better support larger image formats, use of larger
block sizes (e.g. 32x32 and up) and lossless non-transform based coding
options (especially for key frames). It is our hope that we will be
able to make regular updates and we will warmly welcome community
contributions.
Please be warned that, at this stage, the codebase is currently slower
than VP8 stable branch as most new code has not been optimized, and
even the 'C' has been deliberately written to be simple and obvious,
not fast.
The following graphs have the initial test results, numbers in the
tables measure the compression improvement in terms of percentage. The
build has the following optional experiments configured:
--enable-experimental --enable-enhanced_interp --enable-uvintra
--enable-high_precision_mv --enable-sixteenth_subpel_uv
CIF Size clips:
http://getwebm.org/tmp/cif/
HD size clips:
http://getwebm.org/tmp/hd/
(stable_20120309 represents encoding results of WebM master branch
build as of commit#7a15907)
They were encoded using the following encode parameters:
--good --cpu-used=0 -t 0 --lag-in-frames=25 --min-q=0 --max-q=63
--end-usage=0 --auto-alt-ref=1 -p 2 --pass=2 --kf-max-dist=9999
--kf-min-dist=0 --drop-frame=0 --static-thresh=0 --bias-pct=50
--minsection-pct=0 --maxsection-pct=800 --sharpness=0
--arnr-maxframes=7 --arnr-strength=3(for HD,6 for CIF)
--arnr-type=3
Change-Id: I5c62ed09cfff5815a2bb34e7820d6a810c23183c
2012-03-10 02:32:50 +01:00
|
|
|
#endif
|
2010-08-19 21:50:29 +02:00
|
|
|
|
2012-07-14 00:21:29 +02:00
|
|
|
if (xoffset) {
|
|
|
|
|
if (yoffset) {
|
2012-11-01 00:09:17 +01:00
|
|
|
vp9_filter_block1d8_h6_ssse3(src_ptr - (2 * src_pixels_per_line),
|
|
|
|
|
src_pixels_per_line, fdata2, 8, 9, xoffset);
|
|
|
|
|
vp9_filter_block1d8_v6_ssse3(fdata2, 8, dst_ptr, dst_pitch, 4, yoffset);
|
2012-07-14 00:21:29 +02:00
|
|
|
} else {
|
|
|
|
|
/* First-pass only */
|
2012-11-01 00:09:17 +01:00
|
|
|
vp9_filter_block1d8_h6_ssse3(src_ptr, src_pixels_per_line,
|
|
|
|
|
dst_ptr, dst_pitch, 4, xoffset);
|
2010-08-19 21:50:29 +02:00
|
|
|
}
|
2012-07-14 00:21:29 +02:00
|
|
|
} else {
|
|
|
|
|
/* Second-pass only */
|
2012-11-01 00:09:17 +01:00
|
|
|
vp9_filter_block1d8_v6_ssse3(src_ptr - (2 * src_pixels_per_line),
|
|
|
|
|
src_pixels_per_line,
|
|
|
|
|
dst_ptr, dst_pitch, 4, yoffset);
|
2012-07-14 00:21:29 +02:00
|
|
|
}
|
2010-08-19 21:50:29 +02:00
|
|
|
}
|
|
|
|
|
|
2012-11-01 00:09:17 +01:00
|
|
|
void vp9_sixtap_predict4x4_ssse3(unsigned char *src_ptr,
|
|
|
|
|
int src_pixels_per_line,
|
|
|
|
|
int xoffset,
|
|
|
|
|
int yoffset,
|
|
|
|
|
unsigned char *dst_ptr,
|
|
|
|
|
int dst_pitch) {
|
|
|
|
|
DECLARE_ALIGNED_ARRAY(16, unsigned char, fdata2, 4 * 9);
|
WebM Experimental Codec Branch Snapshot
This is a code snapshot of experimental work currently ongoing for a
next-generation codec.
The codebase has been cut down considerably from the libvpx baseline.
For example, we are currently only supporting VBR 2-pass rate control
and have removed most of the code relating to coding speed, threading,
error resilience, partitions and various other features. This is in
part to make the codebase easier to work on and experiment with, but
also because we want to have an open discussion about how the bitstream
will be structured and partitioned and not have that conversation
constrained by past work.
Our basic working pattern has been to initially encapsulate experiments
using configure options linked to #IF CONFIG_XXX statements in the
code. Once experiments have matured and we are reasonably happy that
they give benefit and can be merged without breaking other experiments,
we remove the conditional compile statements and merge them in.
Current changes include:
* Temporal coding experiment for segments (though still only 4 max, it
will likely be increased).
* Segment feature experiment - to allow various bits of information to
be coded at the segment level. Features tested so far include mode
and reference frame information, limiting end of block offset and
transform size, alongside Q and loop filter parameters, but this set
is very fluid.
* Support for 8x8 transform - 8x8 dct with 2nd order 2x2 haar is used
in MBs using 16x16 prediction modes within inter frames.
* Compound prediction (combination of signals from existing predictors
to create a new predictor).
* 8 tap interpolation filters and 1/8th pel motion vectors.
* Loop filter modifications.
* Various entropy modifications and changes to how entropy contexts and
updates are handled.
* Extended quantizer range matched to transform precision improvements.
There are also ongoing further experiments that we hope to merge in the
near future: For example, coding of motion and other aspects of the
prediction signal to better support larger image formats, use of larger
block sizes (e.g. 32x32 and up) and lossless non-transform based coding
options (especially for key frames). It is our hope that we will be
able to make regular updates and we will warmly welcome community
contributions.
Please be warned that, at this stage, the codebase is currently slower
than VP8 stable branch as most new code has not been optimized, and
even the 'C' has been deliberately written to be simple and obvious,
not fast.
The following graphs have the initial test results, numbers in the
tables measure the compression improvement in terms of percentage. The
build has the following optional experiments configured:
--enable-experimental --enable-enhanced_interp --enable-uvintra
--enable-high_precision_mv --enable-sixteenth_subpel_uv
CIF Size clips:
http://getwebm.org/tmp/cif/
HD size clips:
http://getwebm.org/tmp/hd/
(stable_20120309 represents encoding results of WebM master branch
build as of commit#7a15907)
They were encoded using the following encode parameters:
--good --cpu-used=0 -t 0 --lag-in-frames=25 --min-q=0 --max-q=63
--end-usage=0 --auto-alt-ref=1 -p 2 --pass=2 --kf-max-dist=9999
--kf-min-dist=0 --drop-frame=0 --static-thresh=0 --bias-pct=50
--minsection-pct=0 --maxsection-pct=800 --sharpness=0
--arnr-maxframes=7 --arnr-strength=3(for HD,6 for CIF)
--arnr-type=3
Change-Id: I5c62ed09cfff5815a2bb34e7820d6a810c23183c
2012-03-10 02:32:50 +01:00
|
|
|
#ifdef ANNOUNCE_FUNCTION
|
2012-10-31 00:25:53 +01:00
|
|
|
printf("vp9_sixtap_predict4x4_ssse3\n");
|
WebM Experimental Codec Branch Snapshot
This is a code snapshot of experimental work currently ongoing for a
next-generation codec.
The codebase has been cut down considerably from the libvpx baseline.
For example, we are currently only supporting VBR 2-pass rate control
and have removed most of the code relating to coding speed, threading,
error resilience, partitions and various other features. This is in
part to make the codebase easier to work on and experiment with, but
also because we want to have an open discussion about how the bitstream
will be structured and partitioned and not have that conversation
constrained by past work.
Our basic working pattern has been to initially encapsulate experiments
using configure options linked to #IF CONFIG_XXX statements in the
code. Once experiments have matured and we are reasonably happy that
they give benefit and can be merged without breaking other experiments,
we remove the conditional compile statements and merge them in.
Current changes include:
* Temporal coding experiment for segments (though still only 4 max, it
will likely be increased).
* Segment feature experiment - to allow various bits of information to
be coded at the segment level. Features tested so far include mode
and reference frame information, limiting end of block offset and
transform size, alongside Q and loop filter parameters, but this set
is very fluid.
* Support for 8x8 transform - 8x8 dct with 2nd order 2x2 haar is used
in MBs using 16x16 prediction modes within inter frames.
* Compound prediction (combination of signals from existing predictors
to create a new predictor).
* 8 tap interpolation filters and 1/8th pel motion vectors.
* Loop filter modifications.
* Various entropy modifications and changes to how entropy contexts and
updates are handled.
* Extended quantizer range matched to transform precision improvements.
There are also ongoing further experiments that we hope to merge in the
near future: For example, coding of motion and other aspects of the
prediction signal to better support larger image formats, use of larger
block sizes (e.g. 32x32 and up) and lossless non-transform based coding
options (especially for key frames). It is our hope that we will be
able to make regular updates and we will warmly welcome community
contributions.
Please be warned that, at this stage, the codebase is currently slower
than VP8 stable branch as most new code has not been optimized, and
even the 'C' has been deliberately written to be simple and obvious,
not fast.
The following graphs have the initial test results, numbers in the
tables measure the compression improvement in terms of percentage. The
build has the following optional experiments configured:
--enable-experimental --enable-enhanced_interp --enable-uvintra
--enable-high_precision_mv --enable-sixteenth_subpel_uv
CIF Size clips:
http://getwebm.org/tmp/cif/
HD size clips:
http://getwebm.org/tmp/hd/
(stable_20120309 represents encoding results of WebM master branch
build as of commit#7a15907)
They were encoded using the following encode parameters:
--good --cpu-used=0 -t 0 --lag-in-frames=25 --min-q=0 --max-q=63
--end-usage=0 --auto-alt-ref=1 -p 2 --pass=2 --kf-max-dist=9999
--kf-min-dist=0 --drop-frame=0 --static-thresh=0 --bias-pct=50
--minsection-pct=0 --maxsection-pct=800 --sharpness=0
--arnr-maxframes=7 --arnr-strength=3(for HD,6 for CIF)
--arnr-type=3
Change-Id: I5c62ed09cfff5815a2bb34e7820d6a810c23183c
2012-03-10 02:32:50 +01:00
|
|
|
#endif
|
2010-08-19 21:50:29 +02:00
|
|
|
|
2012-07-14 00:21:29 +02:00
|
|
|
if (xoffset) {
|
|
|
|
|
if (yoffset) {
|
2012-11-01 00:09:17 +01:00
|
|
|
vp9_filter_block1d4_h6_ssse3(src_ptr - (2 * src_pixels_per_line),
|
|
|
|
|
src_pixels_per_line, fdata2, 4, 9, xoffset);
|
|
|
|
|
vp9_filter_block1d4_v6_ssse3(fdata2, 4, dst_ptr, dst_pitch, 4, yoffset);
|
2012-07-14 00:21:29 +02:00
|
|
|
} else {
|
2012-11-01 00:09:17 +01:00
|
|
|
vp9_filter_block1d4_h6_ssse3(src_ptr, src_pixels_per_line,
|
|
|
|
|
dst_ptr, dst_pitch, 4, xoffset);
|
2012-07-14 00:21:29 +02:00
|
|
|
}
|
|
|
|
|
} else {
|
2012-11-01 00:09:17 +01:00
|
|
|
vp9_filter_block1d4_v6_ssse3(src_ptr - (2 * src_pixels_per_line),
|
|
|
|
|
src_pixels_per_line,
|
|
|
|
|
dst_ptr, dst_pitch, 4, yoffset);
|
2010-08-19 21:50:29 +02:00
|
|
|
}
|
2012-10-26 02:24:50 +02:00
|
|
|
}
|
2010-08-19 21:50:29 +02:00
|
|
|
|
2012-10-31 00:25:53 +01:00
|
|
|
void vp9_filter_block1d16_v8_ssse3(const unsigned char *src_ptr,
|
2012-10-26 02:24:50 +02:00
|
|
|
const unsigned int src_pitch,
|
|
|
|
|
unsigned char *output_ptr,
|
|
|
|
|
unsigned int out_pitch,
|
|
|
|
|
unsigned int output_height,
|
|
|
|
|
const short *filter);
|
2012-11-01 00:09:17 +01:00
|
|
|
|
2012-10-31 00:25:53 +01:00
|
|
|
void vp9_filter_block1d16_h8_ssse3(const unsigned char *src_ptr,
|
2012-10-26 02:24:50 +02:00
|
|
|
const unsigned int src_pitch,
|
|
|
|
|
unsigned char *output_ptr,
|
|
|
|
|
unsigned int out_pitch,
|
|
|
|
|
unsigned int output_height,
|
|
|
|
|
const short *filter);
|
|
|
|
|
|
2012-11-01 00:09:17 +01:00
|
|
|
void vp9_filter_block2d_16x16_8_ssse3(const unsigned char *src_ptr,
|
|
|
|
|
const unsigned int src_stride,
|
|
|
|
|
const short *hfilter_aligned16,
|
|
|
|
|
const short *vfilter_aligned16,
|
|
|
|
|
unsigned char *dst_ptr,
|
|
|
|
|
unsigned int dst_stride) {
|
|
|
|
|
if (hfilter_aligned16[3] != 128 && vfilter_aligned16[3] != 128) {
|
|
|
|
|
DECLARE_ALIGNED_ARRAY(16, unsigned char, fdata2, 23 * 16);
|
|
|
|
|
|
2012-10-31 00:25:53 +01:00
|
|
|
vp9_filter_block1d16_h8_ssse3(src_ptr - (3 * src_stride), src_stride,
|
2012-11-01 00:09:17 +01:00
|
|
|
fdata2, 16, 23, hfilter_aligned16);
|
|
|
|
|
vp9_filter_block1d16_v8_ssse3(fdata2, 16, dst_ptr, dst_stride, 16,
|
|
|
|
|
vfilter_aligned16);
|
2012-10-26 02:24:50 +02:00
|
|
|
} else {
|
2012-11-01 00:09:17 +01:00
|
|
|
if (hfilter_aligned16[3] != 128) {
|
2012-10-31 00:25:53 +01:00
|
|
|
vp9_filter_block1d16_h8_ssse3(src_ptr, src_stride, dst_ptr, dst_stride,
|
2012-11-01 00:09:17 +01:00
|
|
|
16, hfilter_aligned16);
|
2012-10-26 02:24:50 +02:00
|
|
|
} else {
|
2012-10-31 00:25:53 +01:00
|
|
|
vp9_filter_block1d16_v8_ssse3(src_ptr - (3 * src_stride), src_stride,
|
2012-11-01 00:09:17 +01:00
|
|
|
dst_ptr, dst_stride, 16, vfilter_aligned16);
|
2012-10-26 02:24:50 +02:00
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
2012-10-31 00:25:53 +01:00
|
|
|
void vp9_filter_block1d8_v8_ssse3(const unsigned char *src_ptr,
|
2012-10-26 02:24:50 +02:00
|
|
|
const unsigned int src_pitch,
|
|
|
|
|
unsigned char *output_ptr,
|
|
|
|
|
unsigned int out_pitch,
|
|
|
|
|
unsigned int output_height,
|
|
|
|
|
const short *filter);
|
2012-11-01 00:09:17 +01:00
|
|
|
|
2012-10-31 00:25:53 +01:00
|
|
|
void vp9_filter_block1d8_h8_ssse3(const unsigned char *src_ptr,
|
2012-10-26 02:24:50 +02:00
|
|
|
const unsigned int src_pitch,
|
|
|
|
|
unsigned char *output_ptr,
|
|
|
|
|
unsigned int out_pitch,
|
|
|
|
|
unsigned int output_height,
|
|
|
|
|
const short *filter);
|
2012-11-01 00:09:17 +01:00
|
|
|
|
|
|
|
|
void vp9_filter_block2d_8x8_8_ssse3(const unsigned char *src_ptr,
|
|
|
|
|
const unsigned int src_stride,
|
|
|
|
|
const short *hfilter_aligned16,
|
|
|
|
|
const short *vfilter_aligned16,
|
|
|
|
|
unsigned char *dst_ptr,
|
|
|
|
|
unsigned int dst_stride) {
|
|
|
|
|
if (hfilter_aligned16[3] != 128 && vfilter_aligned16[3] != 128) {
|
|
|
|
|
DECLARE_ALIGNED_ARRAY(16, unsigned char, fdata2, 23 * 16);
|
|
|
|
|
|
2012-10-31 00:25:53 +01:00
|
|
|
vp9_filter_block1d8_h8_ssse3(src_ptr - (3 * src_stride), src_stride,
|
2012-11-01 00:09:17 +01:00
|
|
|
fdata2, 16, 15, hfilter_aligned16);
|
|
|
|
|
vp9_filter_block1d8_v8_ssse3(fdata2, 16, dst_ptr, dst_stride, 8,
|
|
|
|
|
vfilter_aligned16);
|
2012-10-26 02:24:50 +02:00
|
|
|
} else {
|
2012-11-01 00:09:17 +01:00
|
|
|
if (hfilter_aligned16[3] != 128) {
|
2012-10-31 00:25:53 +01:00
|
|
|
vp9_filter_block1d8_h8_ssse3(src_ptr, src_stride, dst_ptr, dst_stride, 8,
|
2012-11-01 00:09:17 +01:00
|
|
|
hfilter_aligned16);
|
2012-10-26 02:24:50 +02:00
|
|
|
} else {
|
2012-10-31 00:25:53 +01:00
|
|
|
vp9_filter_block1d8_v8_ssse3(src_ptr - (3 * src_stride), src_stride,
|
2012-11-01 00:09:17 +01:00
|
|
|
dst_ptr, dst_stride, 8, vfilter_aligned16);
|
2012-10-26 02:24:50 +02:00
|
|
|
}
|
|
|
|
|
}
|
2010-08-19 21:50:29 +02:00
|
|
|
}
|
|
|
|
|
|
2012-11-01 00:09:17 +01:00
|
|
|
void vp9_filter_block2d_8x4_8_ssse3(const unsigned char *src_ptr,
|
|
|
|
|
const unsigned int src_stride,
|
|
|
|
|
const short *hfilter_aligned16,
|
|
|
|
|
const short *vfilter_aligned16,
|
|
|
|
|
unsigned char *dst_ptr,
|
|
|
|
|
unsigned int dst_stride) {
|
|
|
|
|
if (hfilter_aligned16[3] !=128 && vfilter_aligned16[3] != 128) {
|
|
|
|
|
DECLARE_ALIGNED_ARRAY(16, unsigned char, fdata2, 23 * 16);
|
|
|
|
|
|
2012-10-31 00:25:53 +01:00
|
|
|
vp9_filter_block1d8_h8_ssse3(src_ptr - (3 * src_stride), src_stride,
|
2012-11-01 00:09:17 +01:00
|
|
|
fdata2, 16, 11, hfilter_aligned16);
|
|
|
|
|
vp9_filter_block1d8_v8_ssse3(fdata2, 16, dst_ptr, dst_stride, 4,
|
|
|
|
|
vfilter_aligned16);
|
2012-10-26 02:24:50 +02:00
|
|
|
} else {
|
2012-11-01 00:09:17 +01:00
|
|
|
if (hfilter_aligned16[3] != 128) {
|
2012-10-31 00:25:53 +01:00
|
|
|
vp9_filter_block1d8_h8_ssse3(src_ptr, src_stride, dst_ptr, dst_stride, 4,
|
2012-11-01 00:09:17 +01:00
|
|
|
hfilter_aligned16);
|
2012-10-26 02:24:50 +02:00
|
|
|
} else {
|
2012-10-31 00:25:53 +01:00
|
|
|
vp9_filter_block1d8_v8_ssse3(src_ptr - (3 * src_stride), src_stride,
|
2012-11-01 00:09:17 +01:00
|
|
|
dst_ptr, dst_stride, 4, vfilter_aligned16);
|
2012-10-26 02:24:50 +02:00
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
}
|
2010-08-19 21:50:29 +02:00
|
|
|
#endif
|