Add README_RS

Change-Id: I046b97ce4badb0730283fd44206c9e920cbea290

.clang-format

@@ -1,91 +0,0 @@
----
-Language:        Cpp
-# BasedOnStyle:  Google
-# Generated with clang-format 3.8.1
-AccessModifierOffset: -1
-AlignAfterOpenBracket: Align
-AlignConsecutiveAssignments: false
-AlignConsecutiveDeclarations: false
-AlignEscapedNewlinesLeft: true
-AlignOperands:   true
-AlignTrailingComments: true
-AllowAllParametersOfDeclarationOnNextLine: true
-AllowShortBlocksOnASingleLine: false
-AllowShortCaseLabelsOnASingleLine: true
-AllowShortFunctionsOnASingleLine: All
-AllowShortIfStatementsOnASingleLine: true
-AllowShortLoopsOnASingleLine: true
-AlwaysBreakAfterDefinitionReturnType: None
-AlwaysBreakAfterReturnType: None
-AlwaysBreakBeforeMultilineStrings: true
-AlwaysBreakTemplateDeclarations: true
-BinPackArguments: true
-BinPackParameters: true
-BraceWrapping:
-  AfterClass:      false
-  AfterControlStatement: false
-  AfterEnum:       false
-  AfterFunction:   false
-  AfterNamespace:  false
-  AfterObjCDeclaration: false
-  AfterStruct:     false
-  AfterUnion:      false
-  BeforeCatch:     false
-  BeforeElse:      false
-  IndentBraces:    false
-BreakBeforeBinaryOperators: None
-BreakBeforeBraces: Attach
-BreakBeforeTernaryOperators: true
-BreakConstructorInitializersBeforeComma: false
-ColumnLimit:     80
-CommentPragmas:  '^ IWYU pragma:'
-ConstructorInitializerAllOnOneLineOrOnePerLine: false
-ConstructorInitializerIndentWidth: 4
-ContinuationIndentWidth: 4
-Cpp11BracedListStyle: false
-DerivePointerAlignment: false
-DisableFormat:   false
-ExperimentalAutoDetectBinPacking: false
-ForEachMacros:   [ foreach, Q_FOREACH, BOOST_FOREACH ]
-IncludeCategories:
-  - Regex:           '^<.*\.h>'
-    Priority:        1
-  - Regex:           '^<.*'
-    Priority:        2
-  - Regex:           '.*'
-    Priority:        3
-IndentCaseLabels: true
-IndentWidth:     2
-IndentWrappedFunctionNames: false
-KeepEmptyLinesAtTheStartOfBlocks: false
-MacroBlockBegin: ''
-MacroBlockEnd:   ''
-MaxEmptyLinesToKeep: 1
-NamespaceIndentation: None
-ObjCBlockIndentWidth: 2
-ObjCSpaceAfterProperty: false
-ObjCSpaceBeforeProtocolList: false
-PenaltyBreakBeforeFirstCallParameter: 1
-PenaltyBreakComment: 300
-PenaltyBreakFirstLessLess: 120
-PenaltyBreakString: 1000
-PenaltyExcessCharacter: 1000000
-PenaltyReturnTypeOnItsOwnLine: 200
-PointerAlignment: Right
-ReflowComments:  true
-SortIncludes:    false
-SpaceAfterCStyleCast: false
-SpaceBeforeAssignmentOperators: true
-SpaceBeforeParens: ControlStatements
-SpaceInEmptyParentheses: false
-SpacesBeforeTrailingComments: 2
-SpacesInAngles:  false
-SpacesInContainerLiterals: true
-SpacesInCStyleCastParentheses: false
-SpacesInParentheses: false
-SpacesInSquareBrackets: false
-Standard:        Auto
-TabWidth:        8
-UseTab:          Never
-...
-

.gitignore

@@ -1,19 +1,15 @@
-*.S
 *.a
 *.asm.s
 *.d
-*.gcda
 *.gcno
+*.gcda
 *.o
 *~
-.cproject
-.project
-.settings
+/*.ivf
+/*.ivf.md5
 /*-*.mk
 /*.asm
 /*.doxy
-/*.ivf
-/*.ivf.md5
 /.bins
 /.deps
 /.docs
@@ -22,43 +18,58 @@
 /Makefile
 /config.log
 /config.mk
+/decode_to_md5
+/decode_to_md5.c
+/decode_to_md5.dox
+/decode_with_drops
+/decode_with_drops.c
+/decode_with_drops.dox
 /docs/
 /doxyfile
-/examples/*.dox
-/examples/decode_to_md5
-/examples/decode_with_drops
-/examples/decode_with_partial_drops
-/examples/example_xma
-/examples/lossless_encoder
-/examples/postproc
-/examples/resize_util
-/examples/set_maps
-/examples/simple_decoder
-/examples/simple_encoder
-/examples/twopass_encoder
-/examples/aom_cx_set_ref
-/examples/av1_spatial_scalable_encoder
-/examples/aom_temporal_scalable_patterns
-/examples/aom_temporal_svc_encoder
+/error_resilient
+/error_resilient.c
+/error_resilient.dox
+/force_keyframe
+/force_keyframe.c
+/force_keyframe.dox
 /ivfdec
 /ivfdec.dox
 /ivfenc
 /ivfenc.dox
-/libaom.so*
-/libaom.ver
+/libvpx.so*
+/libvpx.ver
+/obj_int_extract
+/postproc
+/postproc.c
+/postproc.dox
 /samples.dox
-/test_intra_pred_speed
-/test_libaom
-/aom_api1_migration.dox
-/av1_rtcd.h
-/aom.pc
-/aom_config.c
-/aom_config.h
-/aom_dsp_rtcd.h
-/aom_scale_rtcd.h
-/aom_version.h
-/aomdec
-/aomdec.dox
-/aomenc
-/aomenc.dox
+/simple_decoder
+/simple_decoder.c
+/simple_decoder.dox
+/simple_encoder
+/simple_encoder.c
+/simple_encoder.dox
+/test_libvpx
+/twopass_encoder
+/twopass_encoder.c
+/twopass_encoder.dox
+/vp8_api1_migration.dox
+/vp8_scalable_patterns
+/vp8_scalable_patterns.dox
+/vp8_set_maps
+/vp8_set_maps.c
+/vp8_set_maps.dox
+/vp8cx_set_ref
+/vp8cx_set_ref.c
+/vp8cx_set_ref.dox
+/vpx.pc
+/vpx_config.c
+/vpx_config.h
+/vpx_rtcd.h
+/vpx_version.h
+/vpxdec
+/vpxenc
 TAGS
+.cproject
+.project
+.settings

.mailmap

@@ -1,32 +1,18 @@
 Adrian Grange <agrange@google.com>
-Aℓex Converse <aconverse@google.com>
-Aℓex Converse <aconverse@google.com> <alex.converse@gmail.com>
 Alexis Ballier <aballier@gentoo.org> <alexis.ballier@gmail.com>
-Alpha Lam <hclam@google.com> <hclam@chromium.org>
-Deb Mukherjee <debargha@google.com>
-Erik Niemeyer <erik.a.niemeyer@intel.com> <erik.a.niemeyer@gmail.com>
-Guillaume Martres <gmartres@google.com> <smarter3@gmail.com>
 Hangyu Kuang <hkuang@google.com>
-Hui Su <huisu@google.com>
-Jacky Chen <jackychen@google.com>
 Jim Bankoski <jimbankoski@google.com>
+John Koleszar <jkoleszar@google.com>
 Johann Koenig <johannkoenig@google.com>
 Johann Koenig <johannkoenig@google.com> <johann.koenig@duck.com>
-Johann Koenig <johannkoenig@google.com> <johann.koenig@gmail.com>
-John Koleszar <jkoleszar@google.com>
-Joshua Litt <joshualitt@google.com> <joshualitt@chromium.org>
-Marco Paniconi <marpan@google.com>
-Marco Paniconi <marpan@google.com> <marpan@chromium.org>
+Johann Koenig <johannkoenig@google.com> <johannkoenig@dhcp-172-19-7-52.mtv.corp.google.com>
 Pascal Massimino <pascal.massimino@gmail.com>
-Paul Wilkins <paulwilkins@google.com>
-Ralph Giles <giles@xiph.org> <giles@entropywave.com>
-Ralph Giles <giles@xiph.org> <giles@mozilla.com>
-Ronald S. Bultje <rsbultje@gmail.com> <rbultje@google.com>
 Sami Pietilä <samipietila@google.com>
-Tamar Levy <tamar.levy@intel.com>
-Tamar Levy <tamar.levy@intel.com> <levytamar82@gmail.com>
 Tero Rintaluoma <teror@google.com> <tero.rintaluoma@on2.com>
 Timothy B. Terriberry <tterribe@xiph.org> Tim Terriberry <tterriberry@mozilla.com>
 Tom Finegan <tomfinegan@google.com>
-Tom Finegan <tomfinegan@google.com> <tomfinegan@chromium.org>
+Ralph Giles <giles@xiph.org> <giles@entropywave.com>
+Ralph Giles <giles@xiph.org> <giles@mozilla.com>
+Alpha Lam <hclam@google.com> <hclam@chromium.org>
+Deb Mukherjee <debargha@google.com>
 Yaowu Xu <yaowu@google.com> <yaowu@xuyaowu.com>

AUTHORS

@@ -3,11 +3,10 @@
 
 Aaron Watry <awatry@gmail.com>
 Abo Talib Mahfoodh <ab.mahfoodh@gmail.com>
-Adam Xu <adam@xuyaowu.com>
 Adrian Grange <agrange@google.com>
-Aℓex Converse <aconverse@google.com>
 Ahmad Sharif <asharif@google.com>
 Alexander Voronov <avoronov@graphics.cs.msu.ru>
+Alex Converse <alex.converse@gmail.com>
 Alexis Ballier <aballier@gentoo.org>
 Alok Ahuja <waveletcoeff@gmail.com>
 Alpha Lam <hclam@google.com>
@@ -15,68 +14,44 @@ A.Mahfoodh <ab.mahfoodh@gmail.com>
 Ami Fischman <fischman@chromium.org>
 Andoni Morales Alastruey <ylatuya@gmail.com>
 Andres Mejia <mcitadel@gmail.com>
-Andrew Russell <anrussell@google.com>
-Angie Chiang <angiebird@google.com>
 Aron Rosenberg <arosenberg@logitech.com>
 Attila Nagy <attilanagy@google.com>
-Brion Vibber <bvibber@wikimedia.org>
 changjun.yang <changjun.yang@intel.com>
-Charles 'Buck' Krasic <ckrasic@google.com>
 chm <chm@rock-chips.com>
 Christian Duvivier <cduvivier@google.com>
 Daniel Kang <ddkang@google.com>
 Deb Mukherjee <debargha@google.com>
-Dim Temp <dimtemp0@gmail.com>
 Dmitry Kovalev <dkovalev@google.com>
 Dragan Mrdjan <dmrdjan@mips.com>
-Ed Baker <edward.baker@intel.com>
-Ehsan Akhgari <ehsan.akhgari@gmail.com>
-Erik Niemeyer <erik.a.niemeyer@intel.com>
+Erik Niemeyer <erik.a.niemeyer@gmail.com>
 Fabio Pedretti <fabio.ped@libero.it>
 Frank Galligan <fgalligan@google.com>
 Fredrik Söderquist <fs@opera.com>
 Fritz Koenig <frkoenig@google.com>
 Gaute Strokkenes <gaute.strokkenes@broadcom.com>
-Geza Lore <gezalore@gmail.com>
-Ghislain MARY <ghislainmary2@gmail.com>
 Giuseppe Scrivano <gscrivano@gnu.org>
-Gordana Cmiljanovic <gordana.cmiljanovic@imgtec.com>
 Guillaume Martres <gmartres@google.com>
 Guillermo Ballester Valor <gbvalor@gmail.com>
 Hangyu Kuang <hkuang@google.com>
-Hanno Böck <hanno@hboeck.de>
 Henrik Lundin <hlundin@google.com>
 Hui Su <huisu@google.com>
 Ivan Maltz <ivanmaltz@google.com>
-Jacek Caban <cjacek@gmail.com>
-Jacky Chen <jackychen@google.com>
 James Berry <jamesberry@google.com>
-James Yu <james.yu@linaro.org>
 James Zern <jzern@google.com>
-Jan Gerber <j@mailb.org>
 Jan Kratochvil <jan.kratochvil@redhat.com>
 Janne Salonen <jsalonen@google.com>
-Jean-Marc Valin <jmvalin@jmvalin.ca>
 Jeff Faust <jfaust@google.com>
 Jeff Muizelaar <jmuizelaar@mozilla.com>
 Jeff Petkau <jpet@chromium.org>
-Jia Jia <jia.jia@linaro.org>
-Jian Zhou <zhoujian@google.com>
 Jim Bankoski <jimbankoski@google.com>
 Jingning Han <jingning@google.com>
-Joey Parrish <joeyparrish@google.com>
-Johann Koenig <johannkoenig@chromium.org>
 Johann Koenig <johannkoenig@google.com>
 John Koleszar <jkoleszar@google.com>
-Johnny Klonaris <google@jawknee.com>
-John Stark <jhnstrk@gmail.com>
 Joshua Bleecher Snyder <josh@treelinelabs.com>
 Joshua Litt <joshualitt@google.com>
-Julia Robson <juliamrobson@gmail.com>
 Justin Clift <justin@salasaga.org>
 Justin Lebar <justin.lebar@gmail.com>
 KO Myung-Hun <komh@chollian.net>
-Lawrence Velázquez <larryv@macports.org>
 Lou Quillio <louquillio@google.com>
 Luca Barbato <lu_zero@gentoo.org>
 Makoto Kato <makoto.kt@gmail.com>
@@ -90,55 +65,36 @@ Michael Kohler <michaelkohler@live.com>
 Mike Frysinger <vapier@chromium.org>
 Mike Hommey <mhommey@mozilla.com>
 Mikhal Shemer <mikhal@google.com>
-Minghai Shang <minghai@google.com>
 Morton Jonuschat <yabawock@gmail.com>
-Nathan E. Egge <negge@dgql.org>
-Nico Weber <thakis@chromium.org>
 Parag Salasakar <img.mips1@gmail.com>
 Pascal Massimino <pascal.massimino@gmail.com>
 Patrik Westin <patrik.westin@gmail.com>
 Paul Wilkins <paulwilkins@google.com>
 Pavol Rusnak <stick@gk2.sk>
 Paweł Hajdan <phajdan@google.com>
-Pengchong Jin <pengchong@google.com>
-Peter de Rivaz <peter.derivaz@argondesign.com>
-Peter de Rivaz <peter.derivaz@gmail.com>
 Philip Jägenstedt <philipj@opera.com>
 Priit Laes <plaes@plaes.org>
 Rafael Ávila de Espíndola <rafael.espindola@gmail.com>
 Rafaël Carré <funman@videolan.org>
 Ralph Giles <giles@xiph.org>
 Rob Bradford <rob@linux.intel.com>
-Ronald S. Bultje <rsbultje@gmail.com>
-Rui Ueyama <ruiu@google.com>
+Ronald S. Bultje <rbultje@google.com>
 Sami Pietilä <samipietila@google.com>
-Sasi Inguva <isasi@google.com>
 Scott Graham <scottmg@chromium.org>
 Scott LaVarnway <slavarnway@google.com>
-Sean McGovern <gseanmcg@gmail.com>
-Sergey Kolomenkin <kolomenkin@gmail.com>
-Sergey Ulanov <sergeyu@chromium.org>
 Shimon Doodkin <helpmepro1@gmail.com>
-Shunyao Li <shunyaoli@google.com>
 Stefan Holmer <holmer@google.com>
-Steinar Midtskogen <stemidts@cisco.com>
 Suman Sunkara <sunkaras@google.com>
 Taekhyun Kim <takim@nvidia.com>
 Takanori MATSUURA <t.matsuu@gmail.com>
 Tamar Levy <tamar.levy@intel.com>
-Tao Bai <michaelbai@chromium.org>
 Tero Rintaluoma <teror@google.com>
 Thijs Vermeir <thijsvermeir@gmail.com>
-Thomas Daede <tdaede@mozilla.com>
-Thomas Davies <thdavies@cisco.com>
-Thomas <thdavies@cisco.com>
-Tim Kopp <tkopp@google.com>
 Timothy B. Terriberry <tterribe@xiph.org>
 Tom Finegan <tomfinegan@google.com>
-Tristan Matthews <le.businessman@gmail.com>
-Tristan Matthews <tmatth@videolan.org>
 Vignesh Venkatasubramanian <vigneshv@google.com>
 Yaowu Xu <yaowu@google.com>
-Yongzhe Wang <yongzhe@google.com>
 Yunqing Wang <yunqingwang@google.com>
-Zoe Liu <zoeliu@google.com>
+Google Inc.
+The Mozilla Foundation
+The Xiph.Org Foundation

CHANGELOG

@@ -1,49 +1,3 @@
-Next Release
-  - Incompatible changes:
-    The AV1 encoder's default keyframe interval changed to 128 from 9999.
-
-2016-04-07 v0.1.0 "AOMedia Codec 1"
-  This release is the first Alliance for Open Media codec.
-2015-11-09 v1.5.0 "Javan Whistling Duck"
-  This release improves upon the VP9 encoder and speeds up the encoding and
-  decoding processes.
-
-  - Upgrading:
-    This release is ABI incompatible with 1.4.0. It drops deprecated VP8
-    controls and adds a variety of VP9 controls for testing.
-
-    The vpxenc utility now prefers VP9 by default.
-
-  - Enhancements:
-    Faster VP9 encoding and decoding
-    Smaller library size by combining functions used by VP8 and VP9
-
-  - Bug Fixes:
-    A variety of fuzzing issues
-
-2015-04-03 v1.4.0 "Indian Runner Duck"
-  This release includes significant improvements to the VP9 codec.
-
-  - Upgrading:
-    This release is ABI incompatible with 1.3.0. It drops the compatibility
-    layer, requiring VPX_IMG_FMT_* instead of IMG_FMT_*, and adds several codec
-    controls for VP9.
-
-  - Enhancements:
-    Faster VP9 encoding and decoding
-    Multithreaded VP9 decoding (tile and frame-based)
-    Multithreaded VP9 encoding - on by default
-    YUV 4:2:2 and 4:4:4 support in VP9
-    10 and 12bit support in VP9
-    64bit ARM support by replacing ARM assembly with intrinsics
-
-  - Bug Fixes:
-    Fixes a VP9 bitstream issue in Profile 1. This only affected non-YUV 4:2:0
-    files.
-
-  - Known Issues:
-    Frame Parallel decoding fails for segmented and non-420 files.
-
 2013-11-15 v1.3.0 "Forest"
   This release introduces the VP9 codec in a backward-compatible way.
   All existing users of VP8 can continue to use the library without

CMakeLists.txt

@@ -1,270 +0,0 @@
-##
-## Copyright (c) 2016, Alliance for Open Media. All rights reserved
-##
-## This source code is subject to the terms of the BSD 2 Clause License and
-## the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
-## was not distributed with this source code in the LICENSE file, you can
-## obtain it at www.aomedia.org/license/software. If the Alliance for Open
-## Media Patent License 1.0 was not distributed with this source code in the
-## PATENTS file, you can obtain it at www.aomedia.org/license/patent.
-##
-cmake_minimum_required(VERSION 3.2)
-project(AOM C CXX)
-
-set(AOM_ROOT "${CMAKE_CURRENT_SOURCE_DIR}")
-set(AOM_CONFIG_DIR "${CMAKE_CURRENT_BINARY_DIR}")
-include("${AOM_ROOT}/build/cmake/aom_configure.cmake")
-
-set(AOM_SRCS
-    "${AOM_CONFIG_DIR}/aom_config.c"
-    "${AOM_CONFIG_DIR}/aom_config.h"
-    "${AOM_ROOT}/aom/aom.h"
-    "${AOM_ROOT}/aom/aom_codec.h"
-    "${AOM_ROOT}/aom/aom_decoder.h"
-    "${AOM_ROOT}/aom/aom_encoder.h"
-    "${AOM_ROOT}/aom/aom_frame_buffer.h"
-    "${AOM_ROOT}/aom/aom_image.h"
-    "${AOM_ROOT}/aom/aom_integer.h"
-    "${AOM_ROOT}/aom/aomcx.h"
-    "${AOM_ROOT}/aom/aomdx.h"
-    "${AOM_ROOT}/aom/internal/aom_codec_internal.h"
-    "${AOM_ROOT}/aom/src/aom_codec.c"
-    "${AOM_ROOT}/aom/src/aom_decoder.c"
-    "${AOM_ROOT}/aom/src/aom_encoder.c"
-    "${AOM_ROOT}/aom/src/aom_image.c")
-
-set(AOM_DSP_SRCS
-    "${AOM_ROOT}/aom_dsp/aom_convolve.c"
-    "${AOM_ROOT}/aom_dsp/aom_convolve.h"
-    "${AOM_ROOT}/aom_dsp/aom_dsp_common.h"
-    "${AOM_ROOT}/aom_dsp/aom_dsp_rtcd.c"
-    "${AOM_ROOT}/aom_dsp/aom_filter.h"
-    "${AOM_ROOT}/aom_dsp/aom_simd.c"
-    "${AOM_ROOT}/aom_dsp/aom_simd.h"
-    "${AOM_ROOT}/aom_dsp/aom_simd_inline.h"
-    "${AOM_ROOT}/aom_dsp/avg.c"
-    "${AOM_ROOT}/aom_dsp/bitreader.h"
-    "${AOM_ROOT}/aom_dsp/bitreader_buffer.c"
-    "${AOM_ROOT}/aom_dsp/bitreader_buffer.h"
-    "${AOM_ROOT}/aom_dsp/bitwriter.h"
-    "${AOM_ROOT}/aom_dsp/bitwriter_buffer.c"
-    "${AOM_ROOT}/aom_dsp/bitwriter_buffer.h"
-    "${AOM_ROOT}/aom_dsp/blend.h"
-    "${AOM_ROOT}/aom_dsp/blend_a64_hmask.c"
-    "${AOM_ROOT}/aom_dsp/blend_a64_mask.c"
-    "${AOM_ROOT}/aom_dsp/blend_a64_vmask.c"
-    "${AOM_ROOT}/aom_dsp/dkboolreader.c"
-    "${AOM_ROOT}/aom_dsp/dkboolreader.h"
-    "${AOM_ROOT}/aom_dsp/dkboolwriter.c"
-    "${AOM_ROOT}/aom_dsp/dkboolwriter.h"
-    "${AOM_ROOT}/aom_dsp/fwd_txfm.c"
-    "${AOM_ROOT}/aom_dsp/fwd_txfm.h"
-    "${AOM_ROOT}/aom_dsp/intrapred.c"
-    "${AOM_ROOT}/aom_dsp/inv_txfm.c"
-    "${AOM_ROOT}/aom_dsp/inv_txfm.h"
-    "${AOM_ROOT}/aom_dsp/loopfilter.c"
-    "${AOM_ROOT}/aom_dsp/prob.c"
-    "${AOM_ROOT}/aom_dsp/prob.h"
-    "${AOM_ROOT}/aom_dsp/psnr.c"
-    "${AOM_ROOT}/aom_dsp/psnr.h"
-    "${AOM_ROOT}/aom_dsp/quantize.c"
-    "${AOM_ROOT}/aom_dsp/quantize.h"
-    "${AOM_ROOT}/aom_dsp/sad.c"
-    "${AOM_ROOT}/aom_dsp/simd/v128_intrinsics.h"
-    "${AOM_ROOT}/aom_dsp/simd/v128_intrinsics_c.h"
-    "${AOM_ROOT}/aom_dsp/simd/v256_intrinsics.h"
-    "${AOM_ROOT}/aom_dsp/simd/v256_intrinsics_c.h"
-    "${AOM_ROOT}/aom_dsp/simd/v64_intrinsics.h"
-    "${AOM_ROOT}/aom_dsp/simd/v64_intrinsics_c.h"
-    "${AOM_ROOT}/aom_dsp/subtract.c"
-    "${AOM_ROOT}/aom_dsp/txfm_common.h"
-    "${AOM_ROOT}/aom_dsp/variance.c"
-    "${AOM_ROOT}/aom_dsp/variance.h")
-
-set(AOM_MEM_SRCS
-    "${AOM_ROOT}/aom_mem/aom_mem.c"
-    "${AOM_ROOT}/aom_mem/aom_mem.h"
-    "${AOM_ROOT}/aom_mem/include/aom_mem_intrnl.h")
-
-set(AOM_SCALE_SRCS
-    "${AOM_ROOT}/aom_scale/aom_scale.h"
-    "${AOM_ROOT}/aom_scale/aom_scale_rtcd.c"
-    "${AOM_ROOT}/aom_scale/generic/aom_scale.c"
-    "${AOM_ROOT}/aom_scale/generic/gen_scalers.c"
-    "${AOM_ROOT}/aom_scale/generic/yv12config.c"
-    "${AOM_ROOT}/aom_scale/generic/yv12extend.c"
-    "${AOM_ROOT}/aom_scale/yv12config.h")
-
-# TODO(tomfinegan): Extract aom_ports from aom_util if possible.
-set(AOM_UTIL_SRCS
-    "${AOM_ROOT}/aom_ports/aom_once.h"
-    "${AOM_ROOT}/aom_ports/aom_timer.h"
-    "${AOM_ROOT}/aom_ports/bitops.h"
-    "${AOM_ROOT}/aom_ports/emmintrin_compat.h"
-    "${AOM_ROOT}/aom_ports/mem.h"
-    "${AOM_ROOT}/aom_ports/mem_ops.h"
-    "${AOM_ROOT}/aom_ports/mem_ops_aligned.h"
-    "${AOM_ROOT}/aom_ports/msvc.h"
-    "${AOM_ROOT}/aom_ports/system_state.h"
-    "${AOM_ROOT}/aom_util/aom_thread.c"
-    "${AOM_ROOT}/aom_util/aom_thread.h"
-    "${AOM_ROOT}/aom_util/endian_inl.h")
-
-set(AOM_AV1_COMMON_SRCS
-    "${AOM_ROOT}/av1/av1_iface_common.h"
-    "${AOM_ROOT}/av1/common/alloccommon.c"
-    "${AOM_ROOT}/av1/common/alloccommon.h"
-    "${AOM_ROOT}/av1/common/av1_fwd_txfm.c"
-    "${AOM_ROOT}/av1/common/av1_fwd_txfm.h"
-    "${AOM_ROOT}/av1/common/av1_inv_txfm.c"
-    "${AOM_ROOT}/av1/common/av1_inv_txfm.h"
-    "${AOM_ROOT}/av1/common/av1_rtcd.c"
-    "${AOM_ROOT}/av1/common/blockd.c"
-    "${AOM_ROOT}/av1/common/blockd.h"
-    "${AOM_ROOT}/av1/common/common.h"
-    "${AOM_ROOT}/av1/common/common_data.h"
-    "${AOM_ROOT}/av1/common/convolve.c"
-    "${AOM_ROOT}/av1/common/convolve.h"
-    "${AOM_ROOT}/av1/common/debugmodes.c"
-    "${AOM_ROOT}/av1/common/entropy.c"
-    "${AOM_ROOT}/av1/common/entropy.h"
-    "${AOM_ROOT}/av1/common/entropymode.c"
-    "${AOM_ROOT}/av1/common/entropymode.h"
-    "${AOM_ROOT}/av1/common/entropymv.c"
-    "${AOM_ROOT}/av1/common/entropymv.h"
-    "${AOM_ROOT}/av1/common/enums.h"
-    "${AOM_ROOT}/av1/common/filter.c"
-    "${AOM_ROOT}/av1/common/filter.h"
-    "${AOM_ROOT}/av1/common/frame_buffers.c"
-    "${AOM_ROOT}/av1/common/frame_buffers.h"
-    "${AOM_ROOT}/av1/common/idct.c"
-    "${AOM_ROOT}/av1/common/idct.h"
-    "${AOM_ROOT}/av1/common/loopfilter.c"
-    "${AOM_ROOT}/av1/common/loopfilter.h"
-    "${AOM_ROOT}/av1/common/mv.h"
-    "${AOM_ROOT}/av1/common/mvref_common.c"
-    "${AOM_ROOT}/av1/common/mvref_common.h"
-    "${AOM_ROOT}/av1/common/odintrin.c"
-    "${AOM_ROOT}/av1/common/odintrin.h"
-    "${AOM_ROOT}/av1/common/onyxc_int.h"
-    "${AOM_ROOT}/av1/common/pred_common.c"
-    "${AOM_ROOT}/av1/common/pred_common.h"
-    "${AOM_ROOT}/av1/common/quant_common.c"
-    "${AOM_ROOT}/av1/common/quant_common.h"
-    "${AOM_ROOT}/av1/common/reconinter.c"
-    "${AOM_ROOT}/av1/common/reconinter.h"
-    "${AOM_ROOT}/av1/common/reconintra.c"
-    "${AOM_ROOT}/av1/common/reconintra.h"
-    "${AOM_ROOT}/av1/common/scale.c"
-    "${AOM_ROOT}/av1/common/scale.h"
-    "${AOM_ROOT}/av1/common/scan.c"
-    "${AOM_ROOT}/av1/common/scan.h"
-    "${AOM_ROOT}/av1/common/seg_common.c"
-    "${AOM_ROOT}/av1/common/seg_common.h"
-    "${AOM_ROOT}/av1/common/thread_common.c"
-    "${AOM_ROOT}/av1/common/thread_common.h"
-    "${AOM_ROOT}/av1/common/tile_common.c"
-    "${AOM_ROOT}/av1/common/tile_common.h")
-
-set(AOM_AV1_DECODER_SRCS
-    "${AOM_ROOT}/av1/av1_dx_iface.c"
-    "${AOM_ROOT}/av1/decoder/decodeframe.c"
-    "${AOM_ROOT}/av1/decoder/decodeframe.h"
-    "${AOM_ROOT}/av1/decoder/decodemv.c"
-    "${AOM_ROOT}/av1/decoder/decodemv.h"
-    "${AOM_ROOT}/av1/decoder/decoder.c"
-    "${AOM_ROOT}/av1/decoder/decoder.h"
-    "${AOM_ROOT}/av1/decoder/detokenize.c"
-    "${AOM_ROOT}/av1/decoder/detokenize.h"
-    "${AOM_ROOT}/av1/decoder/dsubexp.c"
-    "${AOM_ROOT}/av1/decoder/dsubexp.h"
-    "${AOM_ROOT}/av1/decoder/dthread.c"
-    "${AOM_ROOT}/av1/decoder/dthread.h")
-
-set(AOM_AV1_ENCODER_SRCS
-    "${AOM_ROOT}/av1/av1_cx_iface.c"
-    "${AOM_ROOT}/av1/encoder/aq_complexity.c"
-    "${AOM_ROOT}/av1/encoder/aq_complexity.h"
-    "${AOM_ROOT}/av1/encoder/aq_cyclicrefresh.c"
-    "${AOM_ROOT}/av1/encoder/aq_cyclicrefresh.h"
-    "${AOM_ROOT}/av1/encoder/aq_variance.c"
-    "${AOM_ROOT}/av1/encoder/aq_variance.h"
-    "${AOM_ROOT}/av1/encoder/bitstream.c"
-    "${AOM_ROOT}/av1/encoder/bitstream.h"
-    "${AOM_ROOT}/av1/encoder/block.h"
-    "${AOM_ROOT}/av1/encoder/context_tree.c"
-    "${AOM_ROOT}/av1/encoder/context_tree.h"
-    "${AOM_ROOT}/av1/encoder/cost.c"
-    "${AOM_ROOT}/av1/encoder/cost.h"
-    "${AOM_ROOT}/av1/encoder/dct.c"
-    "${AOM_ROOT}/av1/encoder/encodeframe.c"
-    "${AOM_ROOT}/av1/encoder/encodeframe.h"
-    "${AOM_ROOT}/av1/encoder/encodemb.c"
-    "${AOM_ROOT}/av1/encoder/encodemb.h"
-    "${AOM_ROOT}/av1/encoder/encodemv.c"
-    "${AOM_ROOT}/av1/encoder/encodemv.h"
-    "${AOM_ROOT}/av1/encoder/encoder.c"
-    "${AOM_ROOT}/av1/encoder/encoder.h"
-    "${AOM_ROOT}/av1/encoder/ethread.c"
-    "${AOM_ROOT}/av1/encoder/ethread.h"
-    "${AOM_ROOT}/av1/encoder/extend.c"
-    "${AOM_ROOT}/av1/encoder/extend.h"
-    "${AOM_ROOT}/av1/encoder/firstpass.c"
-    "${AOM_ROOT}/av1/encoder/firstpass.h"
-    "${AOM_ROOT}/av1/encoder/hybrid_fwd_txfm.c"
-    "${AOM_ROOT}/av1/encoder/hybrid_fwd_txfm.h"
-    "${AOM_ROOT}/av1/encoder/lookahead.c"
-    "${AOM_ROOT}/av1/encoder/lookahead.h"
-    "${AOM_ROOT}/av1/encoder/mbgraph.c"
-    "${AOM_ROOT}/av1/encoder/mbgraph.h"
-    "${AOM_ROOT}/av1/encoder/mcomp.c"
-    "${AOM_ROOT}/av1/encoder/mcomp.h"
-    "${AOM_ROOT}/av1/encoder/picklpf.c"
-    "${AOM_ROOT}/av1/encoder/picklpf.h"
-    "${AOM_ROOT}/av1/encoder/quantize.c"
-    "${AOM_ROOT}/av1/encoder/quantize.h"
-    "${AOM_ROOT}/av1/encoder/ratectrl.c"
-    "${AOM_ROOT}/av1/encoder/ratectrl.h"
-    "${AOM_ROOT}/av1/encoder/rd.c"
-    "${AOM_ROOT}/av1/encoder/rd.h"
-    "${AOM_ROOT}/av1/encoder/rdopt.c"
-    "${AOM_ROOT}/av1/encoder/rdopt.h"
-    "${AOM_ROOT}/av1/encoder/resize.c"
-    "${AOM_ROOT}/av1/encoder/resize.h"
-    "${AOM_ROOT}/av1/encoder/segmentation.c"
-    "${AOM_ROOT}/av1/encoder/segmentation.h"
-    "${AOM_ROOT}/av1/encoder/speed_features.c"
-    "${AOM_ROOT}/av1/encoder/speed_features.h"
-    "${AOM_ROOT}/av1/encoder/subexp.c"
-    "${AOM_ROOT}/av1/encoder/subexp.h"
-    "${AOM_ROOT}/av1/encoder/temporal_filter.c"
-    "${AOM_ROOT}/av1/encoder/temporal_filter.h"
-    "${AOM_ROOT}/av1/encoder/tokenize.c"
-    "${AOM_ROOT}/av1/encoder/tokenize.h"
-    "${AOM_ROOT}/av1/encoder/treewriter.c"
-    "${AOM_ROOT}/av1/encoder/treewriter.h")
-
-# Targets
-add_library(aom_dsp ${AOM_DSP_SRCS})
-include_directories(${AOM_ROOT} ${AOM_CONFIG_DIR})
-add_library(aom_mem ${AOM_MEM_SRCS})
-add_library(aom_scale ${AOM_SCALE_SRCS})
-include_directories(${AOM_ROOT} ${AOM_CONFIG_DIR})
-add_library(aom_util ${AOM_UTIL_SRCS})
-add_library(aom_av1_decoder ${AOM_AV1_DECODER_SRCS})
-add_library(aom_av1_encoder ${AOM_AV1_ENCODER_SRCS})
-add_library(aom ${AOM_SRCS})
-target_link_libraries(aom LINK_PUBLIC
-                      aom_dsp
-                      aom_mem
-                      aom_scale
-                      aom_util
-                      aom_av1_decoder
-                      aom_av1_encoder)
-add_executable(simple_decoder examples/simple_decoder.c)
-include_directories(${AOM_ROOT})
-target_link_libraries(simple_decoder LINK_PUBLIC aom)
-add_executable(simple_encoder examples/simple_encoder.c)
-include_directories(${AOM_ROOT})
-target_link_libraries(simple_encoder LINK_PUBLIC aom)
-

LICENSE

@@ -1,27 +1,31 @@
-Copyright (c) 2016, Alliance for Open Media. All rights reserved.
+Copyright (c) 2010, The WebM Project authors. All rights reserved.
 
 Redistribution and use in source and binary forms, with or without
-modification, are permitted provided that the following conditions
-are met:
+modification, are permitted provided that the following conditions are
+met:
 
-1. Redistributions of source code must retain the above copyright
-   notice, this list of conditions and the following disclaimer.
+  * Redistributions of source code must retain the above copyright
+    notice, this list of conditions and the following disclaimer.
 
-2. Redistributions in binary form must reproduce the above copyright
-   notice, this list of conditions and the following disclaimer in
-   the documentation and/or other materials provided with the
-   distribution.
+  * Redistributions in binary form must reproduce the above copyright
+    notice, this list of conditions and the following disclaimer in
+    the documentation and/or other materials provided with the
+    distribution.
+
+  * Neither the name of Google, nor the WebM Project, nor the names
+    of its contributors may be used to endorse or promote products
+    derived from this software without specific prior written
+    permission.
 
 THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
 "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
-LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
-FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
-COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
-INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
-BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
-LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
-CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
-LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN
-ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
-POSSIBILITY OF SUCH DAMAGE.
+LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+HOLDER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+(INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 

PATENTS

@@ -1,108 +1,22 @@
-Alliance for Open Media Patent License 1.0
+Additional IP Rights Grant (Patents)
 
-1. License Terms.
-
-1.1. Patent License. Subject to the terms and conditions of this License, each
-     Licensor, on behalf of itself and successors in interest and assigns,
-     grants Licensee a non-sublicensable, perpetual, worldwide, non-exclusive,
-     no-charge, royalty-free, irrevocable (except as expressly stated in this
-     License) patent license to its Necessary Claims to make, use, sell, offer
-     for sale, import or distribute any Implementation.
-
-1.2. Conditions.
-
-1.2.1. Availability. As a condition to the grant of rights to Licensee to make,
-       sell, offer for sale, import or distribute an Implementation under
-       Section 1.1, Licensee must make its Necessary Claims available under
-       this License, and must reproduce this License with any Implementation
-       as follows:
-
-       a. For distribution in source code, by including this License in the
-          root directory of the source code with its Implementation.
-
-       b. For distribution in any other form (including binary, object form,
-          and/or hardware description code (e.g., HDL, RTL, Gate Level Netlist,
-          GDSII, etc.)), by including this License in the documentation, legal
-          notices, and/or other written materials provided with the
-          Implementation.
-
-1.2.2. Additional Conditions. This license is directly from Licensor to
-       Licensee.  Licensee acknowledges as a condition of benefiting from it
-       that no rights from Licensor are received from suppliers, distributors,
-       or otherwise in connection with this License.
-
-1.3. Defensive Termination. If any Licensee, its Affiliates, or its agents
-     initiates patent litigation or files, maintains, or voluntarily
-     participates in a lawsuit against another entity or any person asserting
-     that any Implementation infringes Necessary Claims, any patent licenses
-     granted under this License directly to the Licensee are immediately
-     terminated as of the date of the initiation of action unless 1) that suit
-     was in response to a corresponding suit regarding an Implementation first
-     brought against an initiating entity, or 2) that suit was brought to
-     enforce the terms of this License (including intervention in a third-party
-     action by a Licensee).
-
-1.4. Disclaimers. The Reference Implementation and Specification are provided
-     "AS IS" and without warranty. The entire risk as to implementing or
-     otherwise using the Reference Implementation or Specification is assumed
-     by the implementer and user. Licensor expressly disclaims any warranties
-     (express, implied, or otherwise), including implied warranties of
-     merchantability, non-infringement, fitness for a particular purpose, or
-     title, related to the material. IN NO EVENT WILL LICENSOR BE LIABLE TO
-     ANY OTHER PARTY FOR LOST PROFITS OR ANY FORM OF INDIRECT, SPECIAL,
-     INCIDENTAL, OR CONSEQUENTIAL DAMAGES OF ANY CHARACTER FROM ANY CAUSES OF
-     ACTION OF ANY KIND WITH RESPECT TO THIS LICENSE, WHETHER BASED ON BREACH
-     OF CONTRACT, TORT (INCLUDING NEGLIGENCE), OR OTHERWISE, AND WHETHER OR
-     NOT THE OTHER PARTRY HAS BEEN ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
-
-2. Definitions.
-
-2.1. Affiliate.  <20>Affiliate<74> means an entity that directly or indirectly
-     Controls, is Controlled by, or is under common Control of that party.
-
-2.2. Control. <20>Control<6F> means direct or indirect control of more than 50% of
-     the voting power to elect directors of that corporation, or for any other
-     entity, the power to direct management of such entity.
-
-2.3. Decoder.  "Decoder" means any decoder that conforms fully with all
-     non-optional portions of the Specification.
-
-2.4. Encoder.  "Encoder" means any encoder that produces a bitstream that can
-     be decoded by a Decoder only to the extent it produces such a bitstream.
-
-2.5. Final Deliverable.  <20>Final Deliverable<6C> means the final version of a
-     deliverable approved by the Alliance for Open Media as a Final
-     Deliverable.
-
-2.6. Implementation.  "Implementation" means any implementation, including the
-     Reference Implementation, that is an Encoder and/or a Decoder. An
-     Implementation also includes components of an Implementation only to the
-     extent they are used as part of an Implementation.
-
-2.7. License. <20>License<73> means this license.
-
-2.8. Licensee. <20>Licensee<65> means any person or entity who exercises patent
-     rights granted under this License.
-
-2.9. Licensor.  "Licensor" means (i) any Licensee that makes, sells, offers
-     for sale, imports or distributes any Implementation, or (ii) a person
-     or entity that has a licensing obligation to the Implementation as a
-     result of its membership and/or participation in the Alliance for Open
-     Media working group that developed the Specification.
-
-2.10. Necessary Claims.  "Necessary Claims" means all claims of patents or
-      patent applications, (a) that currently or at any time in the future,
-      are owned or controlled by the Licensor, and (b) (i) would be an
-      Essential Claim as defined by the W3C Policy as of February 5, 2004
-      (https://www.w3.org/Consortium/Patent-Policy-20040205/#def-essential)
-      as if the Specification was a W3C Recommendation; or (ii) are infringed
-      by the Reference Implementation.
-
-2.11. Reference Implementation. <20>Reference Implementation<6F> means an Encoder
-      and/or Decoder released by the Alliance for Open Media as a Final
-      Deliverable.
-
-2.12. Specification. <20>Specification<6F> means the specification designated by
-      the Alliance for Open Media as a Final Deliverable for which this
-      License was issued.
+"This implementation" means the copyrightable works distributed by
+Google as part of the WebM Project.
 
+Google hereby grants to you a perpetual, worldwide, non-exclusive,
+no-charge, royalty-free, irrevocable (except as stated in this section)
+patent license to make, have made, use, offer to sell, sell, import,
+transfer, and otherwise run, modify and propagate the contents of this
+implementation of VP8, where such license applies only to those patent
+claims, both currently owned by Google and acquired in the future,
+licensable by Google that are necessarily infringed by this
+implementation of VP8. This grant does not include claims that would be
+infringed only as a consequence of further modification of this
+implementation. If you or your agent or exclusive licensee institute or
+order or agree to the institution of patent litigation against any
+entity (including a cross-claim or counterclaim in a lawsuit) alleging
+that this implementation of VP8 or any code incorporated within this
+implementation of VP8 constitutes direct or contributory patent
+infringement, or inducement of patent infringement, then any patent
+rights granted to you under this License for this implementation of VP8
+shall terminate as of the date such litigation is filed.

README

@@ -1,6 +1,7 @@
-README - 23 March 2015
+vpx Multi-Format Codec SDK
+README - 1 August 2013
 
-Welcome to the WebM VP8/AV1 Codec SDK!
+Welcome to the WebM VP8/VP9 Codec SDK!
 
 COMPILING THE APPLICATIONS/LIBRARIES:
   The build system used is similar to autotools. Building generally consists of
@@ -11,20 +12,22 @@ COMPILING THE APPLICATIONS/LIBRARIES:
 
     * All x86 targets require the Yasm[1] assembler be installed.
     * All Windows builds require that Cygwin[2] be installed.
-    * Building the documentation requires Doxygen[3]. If you do not
-      have this package, the install-docs option will be disabled.
-    * Downloading the data for the unit tests requires curl[4] and sha1sum.
+    * Building the documentation requires PHP[3] and Doxygen[4]. If you do not
+      have these packages, you must pass --disable-install-docs to the
+      configure script.
+    * Downloading the data for the unit tests requires curl[5] and sha1sum.
       sha1sum is provided via the GNU coreutils, installed by default on
       many *nix platforms, as well as MinGW and Cygwin. If coreutils is not
       available, a compatible version of sha1sum can be built from
-      source[5]. These requirements are optional if not running the unit
+      source[6]. These requirements are optional if not running the unit
       tests.
 
     [1]: http://www.tortall.net/projects/yasm
     [2]: http://www.cygwin.com
-    [3]: http://www.doxygen.org
-    [4]: http://curl.haxx.se
-    [5]: http://www.microbrew.org/tools/md5sha1sum/
+    [3]: http://php.net
+    [4]: http://www.doxygen.org
+    [5]: http://curl.haxx.se
+    [6]: http://www.microbrew.org/tools/md5sha1sum/
 
   2. Out-of-tree builds
   Out of tree builds are a supported method of building the application. For
@@ -33,13 +36,13 @@ COMPILING THE APPLICATIONS/LIBRARIES:
 
     $ mkdir build
     $ cd build
-    $ ../libaom/configure <options>
+    $ ../libvpx/configure <options>
     $ make
 
   3. Configuration options
   The 'configure' script supports a number of options. The --help option can be
   used to get a list of supported options:
-    $ ../libaom/configure --help
+    $ ../libvpx/configure --help
 
   4. Cross development
   For cross development, the most notable option is the --target option. The
@@ -47,10 +50,14 @@ COMPILING THE APPLICATIONS/LIBRARIES:
   --help output of the configure script. As of this writing, the list of
   available targets is:
 
+    armv5te-android-gcc
+    armv5te-linux-rvct
+    armv5te-linux-gcc
+    armv5te-none-rvct
+    armv6-darwin-gcc
     armv6-linux-rvct
     armv6-linux-gcc
     armv6-none-rvct
-    arm64-darwin-gcc
     armv7-android-gcc
     armv7-darwin-gcc
     armv7-linux-rvct
@@ -58,10 +65,13 @@ COMPILING THE APPLICATIONS/LIBRARIES:
     armv7-none-rvct
     armv7-win32-vs11
     armv7-win32-vs12
-    armv7-win32-vs14
-    armv7s-darwin-gcc
     mips32-linux-gcc
-    mips64-linux-gcc
+    ppc32-darwin8-gcc
+    ppc32-darwin9-gcc
+    ppc32-linux-gcc
+    ppc64-darwin8-gcc
+    ppc64-darwin9-gcc
+    ppc64-linux-gcc
     sparc-solaris-gcc
     x86-android-gcc
     x86-darwin8-gcc
@@ -72,33 +82,37 @@ COMPILING THE APPLICATIONS/LIBRARIES:
     x86-darwin11-gcc
     x86-darwin12-gcc
     x86-darwin13-gcc
-    x86-darwin14-gcc
-    x86-iphonesimulator-gcc
     x86-linux-gcc
     x86-linux-icc
     x86-os2-gcc
     x86-solaris-gcc
     x86-win32-gcc
+    x86-win32-vs7
+    x86-win32-vs8
+    x86-win32-vs9
     x86-win32-vs10
     x86-win32-vs11
     x86-win32-vs12
-    x86-win32-vs14
-    x86_64-android-gcc
     x86_64-darwin9-gcc
     x86_64-darwin10-gcc
     x86_64-darwin11-gcc
     x86_64-darwin12-gcc
     x86_64-darwin13-gcc
-    x86_64-darwin14-gcc
-    x86_64-iphonesimulator-gcc
     x86_64-linux-gcc
     x86_64-linux-icc
     x86_64-solaris-gcc
     x86_64-win64-gcc
+    x86_64-win64-vs8
+    x86_64-win64-vs9
     x86_64-win64-vs10
     x86_64-win64-vs11
     x86_64-win64-vs12
-    x86_64-win64-vs14
+    universal-darwin8-gcc
+    universal-darwin9-gcc
+    universal-darwin10-gcc
+    universal-darwin11-gcc
+    universal-darwin12-gcc
+    universal-darwin13-gcc
     generic-gnu
 
   The generic-gnu target, in conjunction with the CROSS environment variable,
@@ -108,7 +122,7 @@ COMPILING THE APPLICATIONS/LIBRARIES:
   toolchain, the following command could be used (note, POSIX SH syntax, adapt
   to your shell as necessary):
 
-    $ CROSS=mipsel-linux-uclibc- ../libaom/configure
+    $ CROSS=mipsel-linux-uclibc- ../libvpx/configure
 
   In addition, the executables to be invoked can be overridden by specifying the
   environment variables: CC, AR, LD, AS, STRIP, NM. Additional flags can be
@@ -119,29 +133,6 @@ COMPILING THE APPLICATIONS/LIBRARIES:
   This defaults to config.log. This should give a good indication of what went
   wrong. If not, contact us for support.
 
-VP8/AV1 TEST VECTORS:
-  The test vectors can be downloaded and verified using the build system after
-  running configure. To specify an alternate directory the
-  LIBAOM_TEST_DATA_PATH environment variable can be used.
-
-  $ ./configure --enable-unit-tests
-  $ LIBAOM_TEST_DATA_PATH=../-test-data make testdata
-
-CODE STYLE:
-  The coding style used by this project is enforced with clang-format using the
-  configuration contained in the .clang-format file in the root of the
-  repository.
-
-  Before pushing changes for review you can format your code with:
-  # Apply clang-format to modified .c, .h and .cc files
-  $ clang-format -i --style=file \
-    $(git diff --name-only --diff-filter=ACMR '*.[hc]' '*.cc')
-
-  Check the .clang-format file for the version used to generate it if there is
-  any difference between your local formatting and the review system.
-
-  See also: http://clang.llvm.org/docs/ClangFormat.html
-
 SUPPORT
   This library is an open source project supported by its community. Please
   please email webm-discuss@webmproject.org for help.

README_RS

@@ -0,0 +1,37 @@
+Welcome to read this page!
+
+How to run renderscript in VP9 decode
+
+  1. Compile RenderScript Kernel(We use zero copy this version,
+                                 if ndk rs module don't support this,
+                                 it can't pass compile)
+  Because RenderScript need g++, but VP9 use gcc.
+  So we need to compile rs alone as a so library. Then VP9 project load/use it.
+  RS code is in vp9/common/kernel/rs/, there is a compile.sh in that folder,
+  We can use sh to compile rs kernel.(need NDK-r9b or higher, we need to config ndk in path).
+  If compile success, it will gen:
+  vp9_rs_packing.so(vp9/common/kernel/rs/)
+  librs.inter_rs.so(vp9/common/kernel/rs/obj/)
+  libstlport_shared.so(vp9/common/kernel/rs/obj/)
+
+  2. Compile VP9 project
+  Just Compile it normally.
+
+  3. Run vpxdec with RenderScript(We use zero copy this version,
+                                  if system RS module don't support this,
+                                  it won't work)
+  [1]: push vp9_rs_packing.so to /data/local/tmp/(or /vendor/lib, /system/lib,
+       need adb remount)
+  [2]: push librs.inter_rs.so to /data/local/tmp/(or /vendor/lib, /system/lib,
+       need adb remount)
+  [3]: push libstlport_shared.so to /data/local/tmp/(or /vendor/lib, /system/lib,
+       need adb remount)
+  [4]: push vpxdec to /data/local/tmp
+  [5]: adb shell          (as root)
+  [6]: cd /data/local/tmp (on device)
+  [7]: export RSENABLE=1  (on device, if not, it won't use RS)
+  [8]: export QCTENABLE=1 (on device, enable qct kernel)
+  [9]: export OPENABLE=1  (on device, enable optimized kernel, dependence on system and driver)
+  [10]:export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/data/local/tmp (on device,
+       if we push .so file to /data/local/tmp, we need set this)
+  [11]:run vpxdec normally(on device)

aom/aom.h

@@ -1,160 +0,0 @@
-/*
- * Copyright (c) 2016, Alliance for Open Media. All rights reserved
- *
- * This source code is subject to the terms of the BSD 2 Clause License and
- * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
- * was not distributed with this source code in the LICENSE file, you can
- * obtain it at www.aomedia.org/license/software. If the Alliance for Open
- * Media Patent License 1.0 was not distributed with this source code in the
- * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
- */
-
-/*!\defgroup aom AOM
- * \ingroup codecs
- * AOM is aom's newest video compression algorithm that uses motion
- * compensated prediction, Discrete Cosine Transform (DCT) coding of the
- * prediction error signal and context dependent entropy coding techniques
- * based on arithmetic principles. It features:
- *  - YUV 4:2:0 image format
- *  - Macro-block based coding (16x16 luma plus two 8x8 chroma)
- *  - 1/4 (1/8) pixel accuracy motion compensated prediction
- *  - 4x4 DCT transform
- *  - 128 level linear quantizer
- *  - In loop deblocking filter
- *  - Context-based entropy coding
- *
- * @{
- */
-/*!\file
- * \brief Provides controls common to both the AOM encoder and decoder.
- */
-#ifndef AOM_AOM_H_
-#define AOM_AOM_H_
-
-#include "./aom_codec.h"
-#include "./aom_image.h"
-
-#ifdef __cplusplus
-extern "C" {
-#endif
-
-/*!\brief Control functions
- *
- * The set of macros define the control functions of AOM interface
- */
-enum aom_com_control_id {
-  /*!\brief pass in an external frame into decoder to be used as reference frame
-   */
-  AOM_SET_REFERENCE = 1,
-  AOM_COPY_REFERENCE = 2, /**< get a copy of reference frame from the decoder */
-  AOM_SET_POSTPROC = 3,   /**< set the decoder's post processing settings  */
-  AOM_SET_DBG_COLOR_REF_FRAME =
-      4, /**< set the reference frames to color for each macroblock */
-  AOM_SET_DBG_COLOR_MB_MODES = 5, /**< set which macro block modes to color */
-  AOM_SET_DBG_COLOR_B_MODES = 6,  /**< set which blocks modes to color */
-  AOM_SET_DBG_DISPLAY_MV = 7,     /**< set which motion vector modes to draw */
-
-  /* TODO(jkoleszar): The encoder incorrectly reuses some of these values (5+)
-   * for its control ids. These should be migrated to something like the
-   * AOM_DECODER_CTRL_ID_START range next time we're ready to break the ABI.
-   */
-  AV1_GET_REFERENCE = 128, /**< get a pointer to a reference frame */
-  AOM_COMMON_CTRL_ID_MAX,
-
-  AV1_GET_NEW_FRAME_IMAGE = 192, /**< get a pointer to the new frame */
-
-  AOM_DECODER_CTRL_ID_START = 256
-};
-
-/*!\brief post process flags
- *
- * The set of macros define AOM decoder post processing flags
- */
-enum aom_postproc_level {
-  AOM_NOFILTERING = 0,
-  AOM_DEBLOCK = 1 << 0,
-  AOM_DEMACROBLOCK = 1 << 1,
-  AOM_ADDNOISE = 1 << 2,
-  AOM_DEBUG_TXT_FRAME_INFO = 1 << 3, /**< print frame information */
-  AOM_DEBUG_TXT_MBLK_MODES =
-      1 << 4, /**< print macro block modes over each macro block */
-  AOM_DEBUG_TXT_DC_DIFF = 1 << 5,   /**< print dc diff for each macro block */
-  AOM_DEBUG_TXT_RATE_INFO = 1 << 6, /**< print video rate info (encoder only) */
-  AOM_MFQE = 1 << 10
-};
-
-/*!\brief post process flags
- *
- * This define a structure that describe the post processing settings. For
- * the best objective measure (using the PSNR metric) set post_proc_flag
- * to AOM_DEBLOCK and deblocking_level to 1.
- */
-
-typedef struct aom_postproc_cfg {
-  /*!\brief the types of post processing to be done, should be combination of
-   * "aom_postproc_level" */
-  int post_proc_flag;
-  int deblocking_level; /**< the strength of deblocking, valid range [0, 16] */
-  int noise_level; /**< the strength of additive noise, valid range [0, 16] */
-} aom_postproc_cfg_t;
-
-/*!\brief reference frame type
- *
- * The set of macros define the type of AOM reference frames
- */
-typedef enum aom_ref_frame_type {
-  AOM_LAST_FRAME = 1,
-  AOM_GOLD_FRAME = 2,
-  AOM_ALTR_FRAME = 4
-} aom_ref_frame_type_t;
-
-/*!\brief reference frame data struct
- *
- * Define the data struct to access aom reference frames.
- */
-typedef struct aom_ref_frame {
-  aom_ref_frame_type_t frame_type; /**< which reference frame */
-  aom_image_t img;                 /**< reference frame data in image format */
-} aom_ref_frame_t;
-
-/*!\brief AV1 specific reference frame data struct
- *
- * Define the data struct to access av1 reference frames.
- */
-typedef struct av1_ref_frame {
-  int idx;         /**< frame index to get (input) */
-  aom_image_t img; /**< img structure to populate (output) */
-} av1_ref_frame_t;
-
-/*!\cond */
-/*!\brief aom decoder control function parameter type
- *
- * defines the data type for each of AOM decoder control function requires
- */
-AOM_CTRL_USE_TYPE(AOM_SET_REFERENCE, aom_ref_frame_t *)
-#define AOM_CTRL_AOM_SET_REFERENCE
-AOM_CTRL_USE_TYPE(AOM_COPY_REFERENCE, aom_ref_frame_t *)
-#define AOM_CTRL_AOM_COPY_REFERENCE
-AOM_CTRL_USE_TYPE(AOM_SET_POSTPROC, aom_postproc_cfg_t *)
-#define AOM_CTRL_AOM_SET_POSTPROC
-AOM_CTRL_USE_TYPE(AOM_SET_DBG_COLOR_REF_FRAME, int)
-#define AOM_CTRL_AOM_SET_DBG_COLOR_REF_FRAME
-AOM_CTRL_USE_TYPE(AOM_SET_DBG_COLOR_MB_MODES, int)
-#define AOM_CTRL_AOM_SET_DBG_COLOR_MB_MODES
-AOM_CTRL_USE_TYPE(AOM_SET_DBG_COLOR_B_MODES, int)
-#define AOM_CTRL_AOM_SET_DBG_COLOR_B_MODES
-AOM_CTRL_USE_TYPE(AOM_SET_DBG_DISPLAY_MV, int)
-#define AOM_CTRL_AOM_SET_DBG_DISPLAY_MV
-AOM_CTRL_USE_TYPE(AV1_GET_REFERENCE, av1_ref_frame_t *)
-#define AOM_CTRL_AV1_GET_REFERENCE
-AOM_CTRL_USE_TYPE(AV1_GET_NEW_FRAME_IMAGE, aom_image_t *)
-#define AOM_CTRL_AV1_GET_NEW_FRAME_IMAGE
-
-/*!\endcond */
-/*! @} - end defgroup aom */
-
-#ifdef __cplusplus
-}  // extern "C"
-#endif
-
-#endif  // AOM_AOM_H_

aom/aom_codec.h

@@ -1,487 +0,0 @@
-/*
- * Copyright (c) 2016, Alliance for Open Media. All rights reserved
- *
- * This source code is subject to the terms of the BSD 2 Clause License and
- * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
- * was not distributed with this source code in the LICENSE file, you can
- * obtain it at www.aomedia.org/license/software. If the Alliance for Open
- * Media Patent License 1.0 was not distributed with this source code in the
- * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
- */
-
-/*!\defgroup codec Common Algorithm Interface
- * This abstraction allows applications to easily support multiple video
- * formats with minimal code duplication. This section describes the interface
- * common to all codecs (both encoders and decoders).
- * @{
- */
-
-/*!\file
- * \brief Describes the codec algorithm interface to applications.
- *
- * This file describes the interface between an application and a
- * video codec algorithm.
- *
- * An application instantiates a specific codec instance by using
- * aom_codec_init() and a pointer to the algorithm's interface structure:
- *     <pre>
- *     my_app.c:
- *       extern aom_codec_iface_t my_codec;
- *       {
- *           aom_codec_ctx_t algo;
- *           res = aom_codec_init(&algo, &my_codec);
- *       }
- *     </pre>
- *
- * Once initialized, the instance is manged using other functions from
- * the aom_codec_* family.
- */
-#ifndef AOM_AOM_CODEC_H_
-#define AOM_AOM_CODEC_H_
-
-#ifdef __cplusplus
-extern "C" {
-#endif
-
-#include "./aom_integer.h"
-#include "./aom_image.h"
-
-/*!\brief Decorator indicating a function is deprecated */
-#ifndef DEPRECATED
-#if defined(__GNUC__) && __GNUC__
-#define DEPRECATED __attribute__((deprecated))
-#elif defined(_MSC_VER)
-#define DEPRECATED
-#else
-#define DEPRECATED
-#endif
-#endif /* DEPRECATED */
-
-#ifndef DECLSPEC_DEPRECATED
-#if defined(__GNUC__) && __GNUC__
-#define DECLSPEC_DEPRECATED /**< \copydoc #DEPRECATED */
-#elif defined(_MSC_VER)
-/*!\brief \copydoc #DEPRECATED */
-#define DECLSPEC_DEPRECATED __declspec(deprecated)
-#else
-#define DECLSPEC_DEPRECATED /**< \copydoc #DEPRECATED */
-#endif
-#endif /* DECLSPEC_DEPRECATED */
-
-/*!\brief Decorator indicating a function is potentially unused */
-#ifdef UNUSED
-#elif defined(__GNUC__) || defined(__clang__)
-#define UNUSED __attribute__((unused))
-#else
-#define UNUSED
-#endif
-
-/*!\brief Decorator indicating that given struct/union/enum is packed */
-#ifndef ATTRIBUTE_PACKED
-#if defined(__GNUC__) && __GNUC__
-#define ATTRIBUTE_PACKED __attribute__((packed))
-#elif defined(_MSC_VER)
-#define ATTRIBUTE_PACKED
-#else
-#define ATTRIBUTE_PACKED
-#endif
-#endif /* ATTRIBUTE_PACKED */
-
-/*!\brief Current ABI version number
- *
- * \internal
- * If this file is altered in any way that changes the ABI, this value
- * must be bumped.  Examples include, but are not limited to, changing
- * types, removing or reassigning enums, adding/removing/rearranging
- * fields to structures
- */
-#define AOM_CODEC_ABI_VERSION (3 + AOM_IMAGE_ABI_VERSION) /**<\hideinitializer*/
-
-/*!\brief Algorithm return codes */
-typedef enum {
-  /*!\brief Operation completed without error */
-  AOM_CODEC_OK,
-
-  /*!\brief Unspecified error */
-  AOM_CODEC_ERROR,
-
-  /*!\brief Memory operation failed */
-  AOM_CODEC_MEM_ERROR,
-
-  /*!\brief ABI version mismatch */
-  AOM_CODEC_ABI_MISMATCH,
-
-  /*!\brief Algorithm does not have required capability */
-  AOM_CODEC_INCAPABLE,
-
-  /*!\brief The given bitstream is not supported.
-   *
-   * The bitstream was unable to be parsed at the highest level. The decoder
-   * is unable to proceed. This error \ref SHOULD be treated as fatal to the
-   * stream. */
-  AOM_CODEC_UNSUP_BITSTREAM,
-
-  /*!\brief Encoded bitstream uses an unsupported feature
-   *
-   * The decoder does not implement a feature required by the encoder. This
-   * return code should only be used for features that prevent future
-   * pictures from being properly decoded. This error \ref MAY be treated as
-   * fatal to the stream or \ref MAY be treated as fatal to the current GOP.
-   */
-  AOM_CODEC_UNSUP_FEATURE,
-
-  /*!\brief The coded data for this stream is corrupt or incomplete
-   *
-   * There was a problem decoding the current frame.  This return code
-   * should only be used for failures that prevent future pictures from
-   * being properly decoded. This error \ref MAY be treated as fatal to the
-   * stream or \ref MAY be treated as fatal to the current GOP. If decoding
-   * is continued for the current GOP, artifacts may be present.
-   */
-  AOM_CODEC_CORRUPT_FRAME,
-
-  /*!\brief An application-supplied parameter is not valid.
-   *
-   */
-  AOM_CODEC_INVALID_PARAM,
-
-  /*!\brief An iterator reached the end of list.
-   *
-   */
-  AOM_CODEC_LIST_END
-
-} aom_codec_err_t;
-
-/*! \brief Codec capabilities bitfield
- *
- *  Each codec advertises the capabilities it supports as part of its
- *  ::aom_codec_iface_t interface structure. Capabilities are extra interfaces
- *  or functionality, and are not required to be supported.
- *
- *  The available flags are specified by AOM_CODEC_CAP_* defines.
- */
-typedef long aom_codec_caps_t;
-#define AOM_CODEC_CAP_DECODER 0x1 /**< Is a decoder */
-#define AOM_CODEC_CAP_ENCODER 0x2 /**< Is an encoder */
-
-/*! \brief Initialization-time Feature Enabling
- *
- *  Certain codec features must be known at initialization time, to allow for
- *  proper memory allocation.
- *
- *  The available flags are specified by AOM_CODEC_USE_* defines.
- */
-typedef long aom_codec_flags_t;
-
-/*!\brief Codec interface structure.
- *
- * Contains function pointers and other data private to the codec
- * implementation. This structure is opaque to the application.
- */
-typedef const struct aom_codec_iface aom_codec_iface_t;
-
-/*!\brief Codec private data structure.
- *
- * Contains data private to the codec implementation. This structure is opaque
- * to the application.
- */
-typedef struct aom_codec_priv aom_codec_priv_t;
-
-/*!\brief Iterator
- *
- * Opaque storage used for iterating over lists.
- */
-typedef const void *aom_codec_iter_t;
-
-/*!\brief Codec context structure
- *
- * All codecs \ref MUST support this context structure fully. In general,
- * this data should be considered private to the codec algorithm, and
- * not be manipulated or examined by the calling application. Applications
- * may reference the 'name' member to get a printable description of the
- * algorithm.
- */
-typedef struct aom_codec_ctx {
-  const char *name;             /**< Printable interface name */
-  aom_codec_iface_t *iface;     /**< Interface pointers */
-  aom_codec_err_t err;          /**< Last returned error */
-  const char *err_detail;       /**< Detailed info, if available */
-  aom_codec_flags_t init_flags; /**< Flags passed at init time */
-  union {
-    /**< Decoder Configuration Pointer */
-    const struct aom_codec_dec_cfg *dec;
-    /**< Encoder Configuration Pointer */
-    const struct aom_codec_enc_cfg *enc;
-    const void *raw;
-  } config;               /**< Configuration pointer aliasing union */
-  aom_codec_priv_t *priv; /**< Algorithm private storage */
-} aom_codec_ctx_t;
-
-/*!\brief Bit depth for codec
- * *
- * This enumeration determines the bit depth of the codec.
- */
-typedef enum aom_bit_depth {
-  AOM_BITS_8 = 8,   /**<  8 bits */
-  AOM_BITS_10 = 10, /**< 10 bits */
-  AOM_BITS_12 = 12, /**< 12 bits */
-} aom_bit_depth_t;
-
-/*!\brief Superblock size selection.
- *
- * Defines the superblock size used for encoding. The superblock size can
- * either be fixed at 64x64 or 128x128 pixels, or it can be dynamically
- * selected by the encoder for each frame.
- */
-typedef enum aom_superblock_size {
-  AOM_SUPERBLOCK_SIZE_64X64,   /**< Always use 64x64 superblocks. */
-  AOM_SUPERBLOCK_SIZE_128X128, /**< Always use 128x128 superblocks. */
-  AOM_SUPERBLOCK_SIZE_DYNAMIC  /**< Select superblock size dynamically. */
-} aom_superblock_size_t;
-
-/*
- * Library Version Number Interface
- *
- * For example, see the following sample return values:
- *     aom_codec_version()           (1<<16 | 2<<8 | 3)
- *     aom_codec_version_str()       "v1.2.3-rc1-16-gec6a1ba"
- *     aom_codec_version_extra_str() "rc1-16-gec6a1ba"
- */
-
-/*!\brief Return the version information (as an integer)
- *
- * Returns a packed encoding of the library version number. This will only
- * include
- * the major.minor.patch component of the version number. Note that this encoded
- * value should be accessed through the macros provided, as the encoding may
- * change
- * in the future.
- *
- */
-int aom_codec_version(void);
-#define AOM_VERSION_MAJOR(v) \
-  ((v >> 16) & 0xff) /**< extract major from packed version */
-#define AOM_VERSION_MINOR(v) \
-  ((v >> 8) & 0xff) /**< extract minor from packed version */
-#define AOM_VERSION_PATCH(v) \
-  ((v >> 0) & 0xff) /**< extract patch from packed version */
-
-/*!\brief Return the version major number */
-#define aom_codec_version_major() ((aom_codec_version() >> 16) & 0xff)
-
-/*!\brief Return the version minor number */
-#define aom_codec_version_minor() ((aom_codec_version() >> 8) & 0xff)
-
-/*!\brief Return the version patch number */
-#define aom_codec_version_patch() ((aom_codec_version() >> 0) & 0xff)
-
-/*!\brief Return the version information (as a string)
- *
- * Returns a printable string containing the full library version number. This
- * may
- * contain additional text following the three digit version number, as to
- * indicate
- * release candidates, prerelease versions, etc.
- *
- */
-const char *aom_codec_version_str(void);
-
-/*!\brief Return the version information (as a string)
- *
- * Returns a printable "extra string". This is the component of the string
- * returned
- * by aom_codec_version_str() following the three digit version number.
- *
- */
-const char *aom_codec_version_extra_str(void);
-
-/*!\brief Return the build configuration
- *
- * Returns a printable string containing an encoded version of the build
- * configuration. This may be useful to aom support.
- *
- */
-const char *aom_codec_build_config(void);
-
-/*!\brief Return the name for a given interface
- *
- * Returns a human readable string for name of the given codec interface.
- *
- * \param[in]    iface     Interface pointer
- *
- */
-const char *aom_codec_iface_name(aom_codec_iface_t *iface);
-
-/*!\brief Convert error number to printable string
- *
- * Returns a human readable string for the last error returned by the
- * algorithm. The returned error will be one line and will not contain
- * any newline characters.
- *
- *
- * \param[in]    err     Error number.
- *
- */
-const char *aom_codec_err_to_string(aom_codec_err_t err);
-
-/*!\brief Retrieve error synopsis for codec context
- *
- * Returns a human readable string for the last error returned by the
- * algorithm. The returned error will be one line and will not contain
- * any newline characters.
- *
- *
- * \param[in]    ctx     Pointer to this instance's context.
- *
- */
-const char *aom_codec_error(aom_codec_ctx_t *ctx);
-
-/*!\brief Retrieve detailed error information for codec context
- *
- * Returns a human readable string providing detailed information about
- * the last error.
- *
- * \param[in]    ctx     Pointer to this instance's context.
- *
- * \retval NULL
- *     No detailed information is available.
- */
-const char *aom_codec_error_detail(aom_codec_ctx_t *ctx);
-
-/* REQUIRED FUNCTIONS
- *
- * The following functions are required to be implemented for all codecs.
- * They represent the base case functionality expected of all codecs.
- */
-
-/*!\brief Destroy a codec instance
- *
- * Destroys a codec context, freeing any associated memory buffers.
- *
- * \param[in] ctx   Pointer to this instance's context
- *
- * \retval #AOM_CODEC_OK
- *     The codec algorithm initialized.
- * \retval #AOM_CODEC_MEM_ERROR
- *     Memory allocation failed.
- */
-aom_codec_err_t aom_codec_destroy(aom_codec_ctx_t *ctx);
-
-/*!\brief Get the capabilities of an algorithm.
- *
- * Retrieves the capabilities bitfield from the algorithm's interface.
- *
- * \param[in] iface   Pointer to the algorithm interface
- *
- */
-aom_codec_caps_t aom_codec_get_caps(aom_codec_iface_t *iface);
-
-/*!\brief Control algorithm
- *
- * This function is used to exchange algorithm specific data with the codec
- * instance. This can be used to implement features specific to a particular
- * algorithm.
- *
- * This wrapper function dispatches the request to the helper function
- * associated with the given ctrl_id. It tries to call this function
- * transparently, but will return #AOM_CODEC_ERROR if the request could not
- * be dispatched.
- *
- * Note that this function should not be used directly. Call the
- * #aom_codec_control wrapper macro instead.
- *
- * \param[in]     ctx              Pointer to this instance's context
- * \param[in]     ctrl_id          Algorithm specific control identifier
- *
- * \retval #AOM_CODEC_OK
- *     The control request was processed.
- * \retval #AOM_CODEC_ERROR
- *     The control request was not processed.
- * \retval #AOM_CODEC_INVALID_PARAM
- *     The data was not valid.
- */
-aom_codec_err_t aom_codec_control_(aom_codec_ctx_t *ctx, int ctrl_id, ...);
-#if defined(AOM_DISABLE_CTRL_TYPECHECKS) && AOM_DISABLE_CTRL_TYPECHECKS
-#define aom_codec_control(ctx, id, data) aom_codec_control_(ctx, id, data)
-#define AOM_CTRL_USE_TYPE(id, typ)
-#define AOM_CTRL_USE_TYPE_DEPRECATED(id, typ)
-#define AOM_CTRL_VOID(id, typ)
-
-#else
-/*!\brief aom_codec_control wrapper macro
- *
- * This macro allows for type safe conversions across the variadic parameter
- * to aom_codec_control_().
- *
- * \internal
- * It works by dispatching the call to the control function through a wrapper
- * function named with the id parameter.
- */
-#define aom_codec_control(ctx, id, data) \
-  aom_codec_control_##id(ctx, id, data) /**<\hideinitializer*/
-
-/*!\brief aom_codec_control type definition macro
- *
- * This macro allows for type safe conversions across the variadic parameter
- * to aom_codec_control_(). It defines the type of the argument for a given
- * control identifier.
- *
- * \internal
- * It defines a static function with
- * the correctly typed arguments as a wrapper to the type-unsafe internal
- * function.
- */
-#define AOM_CTRL_USE_TYPE(id, typ)                                           \
-  static aom_codec_err_t aom_codec_control_##id(aom_codec_ctx_t *, int, typ) \
-      UNUSED;                                                                \
-                                                                             \
-  static aom_codec_err_t aom_codec_control_##id(aom_codec_ctx_t *ctx,        \
-                                                int ctrl_id, typ data) {     \
-    return aom_codec_control_(ctx, ctrl_id, data);                           \
-  } /**<\hideinitializer*/
-
-/*!\brief aom_codec_control deprecated type definition macro
- *
- * Like #AOM_CTRL_USE_TYPE, but indicates that the specified control is
- * deprecated and should not be used. Consult the documentation for your
- * codec for more information.
- *
- * \internal
- * It defines a static function with the correctly typed arguments as a
- * wrapper to the type-unsafe internal function.
- */
-#define AOM_CTRL_USE_TYPE_DEPRECATED(id, typ)                        \
-  DECLSPEC_DEPRECATED static aom_codec_err_t aom_codec_control_##id( \
-      aom_codec_ctx_t *, int, typ) DEPRECATED UNUSED;                \
-                                                                     \
-  DECLSPEC_DEPRECATED static aom_codec_err_t aom_codec_control_##id( \
-      aom_codec_ctx_t *ctx, int ctrl_id, typ data) {                 \
-    return aom_codec_control_(ctx, ctrl_id, data);                   \
-  } /**<\hideinitializer*/
-
-/*!\brief aom_codec_control void type definition macro
- *
- * This macro allows for type safe conversions across the variadic parameter
- * to aom_codec_control_(). It indicates that a given control identifier takes
- * no argument.
- *
- * \internal
- * It defines a static function without a data argument as a wrapper to the
- * type-unsafe internal function.
- */
-#define AOM_CTRL_VOID(id)                                               \
-  static aom_codec_err_t aom_codec_control_##id(aom_codec_ctx_t *, int) \
-      UNUSED;                                                           \
-                                                                        \
-  static aom_codec_err_t aom_codec_control_##id(aom_codec_ctx_t *ctx,   \
-                                                int ctrl_id) {          \
-    return aom_codec_control_(ctx, ctrl_id);                            \
-  } /**<\hideinitializer*/
-
-#endif
-
-/*!@} - end defgroup codec*/
-#ifdef __cplusplus
-}
-#endif
-#endif  // AOM_AOM_CODEC_H_

aom/aom_codec.mk

@@ -1,42 +0,0 @@
-##
-## Copyright (c) 2016, Alliance for Open Media. All rights reserved
-##
-## This source code is subject to the terms of the BSD 2 Clause License and
-## the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
-## was not distributed with this source code in the LICENSE file, you can
-## obtain it at www.aomedia.org/license/software. If the Alliance for Open
-## Media Patent License 1.0 was not distributed with this source code in the
-## PATENTS file, you can obtain it at www.aomedia.org/license/patent.
-##
-
-
-API_EXPORTS += exports
-
-API_SRCS-$(CONFIG_AV1_ENCODER) += aom.h
-API_SRCS-$(CONFIG_AV1_ENCODER) += aomcx.h
-API_DOC_SRCS-$(CONFIG_AV1_ENCODER) += aom.h
-API_DOC_SRCS-$(CONFIG_AV1_ENCODER) += aomcx.h
-
-API_SRCS-$(CONFIG_AV1_DECODER) += aom.h
-API_SRCS-$(CONFIG_AV1_DECODER) += aomdx.h
-API_DOC_SRCS-$(CONFIG_AV1_DECODER) += aom.h
-API_DOC_SRCS-$(CONFIG_AV1_DECODER) += aomdx.h
-
-API_DOC_SRCS-yes += aom_codec.h
-API_DOC_SRCS-yes += aom_decoder.h
-API_DOC_SRCS-yes += aom_encoder.h
-API_DOC_SRCS-yes += aom_frame_buffer.h
-API_DOC_SRCS-yes += aom_image.h
-
-API_SRCS-yes += src/aom_decoder.c
-API_SRCS-yes += aom_decoder.h
-API_SRCS-yes += src/aom_encoder.c
-API_SRCS-yes += aom_encoder.h
-API_SRCS-yes += internal/aom_codec_internal.h
-API_SRCS-yes += src/aom_codec.c
-API_SRCS-yes += src/aom_image.c
-API_SRCS-yes += aom_codec.h
-API_SRCS-yes += aom_codec.mk
-API_SRCS-yes += aom_frame_buffer.h
-API_SRCS-yes += aom_image.h
-API_SRCS-yes += aom_integer.h

aom/aom_decoder.h

@@ -1,366 +0,0 @@
-/*
- * Copyright (c) 2016, Alliance for Open Media. All rights reserved
- *
- * This source code is subject to the terms of the BSD 2 Clause License and
- * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
- * was not distributed with this source code in the LICENSE file, you can
- * obtain it at www.aomedia.org/license/software. If the Alliance for Open
- * Media Patent License 1.0 was not distributed with this source code in the
- * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
- */
-#ifndef AOM_AOM_DECODER_H_
-#define AOM_AOM_DECODER_H_
-
-/*!\defgroup decoder Decoder Algorithm Interface
- * \ingroup codec
- * This abstraction allows applications using this decoder to easily support
- * multiple video formats with minimal code duplication. This section describes
- * the interface common to all decoders.
- * @{
- */
-
-/*!\file
- * \brief Describes the decoder algorithm interface to applications.
- *
- * This file describes the interface between an application and a
- * video decoder algorithm.
- *
- */
-#ifdef __cplusplus
-extern "C" {
-#endif
-
-#include "./aom_codec.h"
-#include "./aom_frame_buffer.h"
-
-/*!\brief Current ABI version number
- *
- * \internal
- * If this file is altered in any way that changes the ABI, this value
- * must be bumped.  Examples include, but are not limited to, changing
- * types, removing or reassigning enums, adding/removing/rearranging
- * fields to structures
- */
-#define AOM_DECODER_ABI_VERSION \
-  (3 + AOM_CODEC_ABI_VERSION) /**<\hideinitializer*/
-
-/*! \brief Decoder capabilities bitfield
- *
- *  Each decoder advertises the capabilities it supports as part of its
- *  ::aom_codec_iface_t interface structure. Capabilities are extra interfaces
- *  or functionality, and are not required to be supported by a decoder.
- *
- *  The available flags are specified by AOM_CODEC_CAP_* defines.
- */
-#define AOM_CODEC_CAP_PUT_SLICE 0x10000 /**< Will issue put_slice callbacks */
-#define AOM_CODEC_CAP_PUT_FRAME 0x20000 /**< Will issue put_frame callbacks */
-#define AOM_CODEC_CAP_POSTPROC 0x40000  /**< Can postprocess decoded frame */
-/*!\brief Can conceal errors due to packet loss */
-#define AOM_CODEC_CAP_ERROR_CONCEALMENT 0x80000
-/*!\brief Can receive encoded frames one fragment at a time */
-#define AOM_CODEC_CAP_INPUT_FRAGMENTS 0x100000
-
-/*! \brief Initialization-time Feature Enabling
- *
- *  Certain codec features must be known at initialization time, to allow for
- *  proper memory allocation.
- *
- *  The available flags are specified by AOM_CODEC_USE_* defines.
- */
-/*!\brief Can support frame-based multi-threading */
-#define AOM_CODEC_CAP_FRAME_THREADING 0x200000
-/*!brief Can support external frame buffers */
-#define AOM_CODEC_CAP_EXTERNAL_FRAME_BUFFER 0x400000
-
-#define AOM_CODEC_USE_POSTPROC 0x10000 /**< Postprocess decoded frame */
-/*!\brief Conceal errors in decoded frames */
-#define AOM_CODEC_USE_ERROR_CONCEALMENT 0x20000
-/*!\brief The input frame should be passed to the decoder one fragment at a
- * time */
-#define AOM_CODEC_USE_INPUT_FRAGMENTS 0x40000
-/*!\brief Enable frame-based multi-threading */
-#define AOM_CODEC_USE_FRAME_THREADING 0x80000
-
-/*!\brief Stream properties
- *
- * This structure is used to query or set properties of the decoded
- * stream. Algorithms may extend this structure with data specific
- * to their bitstream by setting the sz member appropriately.
- */
-typedef struct aom_codec_stream_info {
-  unsigned int sz;    /**< Size of this structure */
-  unsigned int w;     /**< Width (or 0 for unknown/default) */
-  unsigned int h;     /**< Height (or 0 for unknown/default) */
-  unsigned int is_kf; /**< Current frame is a keyframe */
-} aom_codec_stream_info_t;
-
-/* REQUIRED FUNCTIONS
- *
- * The following functions are required to be implemented for all decoders.
- * They represent the base case functionality expected of all decoders.
- */
-
-/*!\brief Initialization Configurations
- *
- * This structure is used to pass init time configuration options to the
- * decoder.
- */
-typedef struct aom_codec_dec_cfg {
-  unsigned int threads; /**< Maximum number of threads to use, default 1 */
-  unsigned int w;       /**< Width */
-  unsigned int h;       /**< Height */
-} aom_codec_dec_cfg_t;  /**< alias for struct aom_codec_dec_cfg */
-
-/*!\brief Initialize a decoder instance
- *
- * Initializes a decoder context using the given interface. Applications
- * should call the aom_codec_dec_init convenience macro instead of this
- * function directly, to ensure that the ABI version number parameter
- * is properly initialized.
- *
- * If the library was configured with --disable-multithread, this call
- * is not thread safe and should be guarded with a lock if being used
- * in a multithreaded context.
- *
- * \param[in]    ctx     Pointer to this instance's context.
- * \param[in]    iface   Pointer to the algorithm interface to use.
- * \param[in]    cfg     Configuration to use, if known. May be NULL.
- * \param[in]    flags   Bitfield of AOM_CODEC_USE_* flags
- * \param[in]    ver     ABI version number. Must be set to
- *                       AOM_DECODER_ABI_VERSION
- * \retval #AOM_CODEC_OK
- *     The decoder algorithm initialized.
- * \retval #AOM_CODEC_MEM_ERROR
- *     Memory allocation failed.
- */
-aom_codec_err_t aom_codec_dec_init_ver(aom_codec_ctx_t *ctx,
-                                       aom_codec_iface_t *iface,
-                                       const aom_codec_dec_cfg_t *cfg,
-                                       aom_codec_flags_t flags, int ver);
-
-/*!\brief Convenience macro for aom_codec_dec_init_ver()
- *
- * Ensures the ABI version parameter is properly set.
- */
-#define aom_codec_dec_init(ctx, iface, cfg, flags) \
-  aom_codec_dec_init_ver(ctx, iface, cfg, flags, AOM_DECODER_ABI_VERSION)
-
-/*!\brief Parse stream info from a buffer
- *
- * Performs high level parsing of the bitstream. Construction of a decoder
- * context is not necessary. Can be used to determine if the bitstream is
- * of the proper format, and to extract information from the stream.
- *
- * \param[in]      iface   Pointer to the algorithm interface
- * \param[in]      data    Pointer to a block of data to parse
- * \param[in]      data_sz Size of the data buffer
- * \param[in,out]  si      Pointer to stream info to update. The size member
- *                         \ref MUST be properly initialized, but \ref MAY be
- *                         clobbered by the algorithm. This parameter \ref MAY
- *                         be NULL.
- *
- * \retval #AOM_CODEC_OK
- *     Bitstream is parsable and stream information updated
- */
-aom_codec_err_t aom_codec_peek_stream_info(aom_codec_iface_t *iface,
-                                           const uint8_t *data,
-                                           unsigned int data_sz,
-                                           aom_codec_stream_info_t *si);
-
-/*!\brief Return information about the current stream.
- *
- * Returns information about the stream that has been parsed during decoding.
- *
- * \param[in]      ctx     Pointer to this instance's context
- * \param[in,out]  si      Pointer to stream info to update. The size member
- *                         \ref MUST be properly initialized, but \ref MAY be
- *                         clobbered by the algorithm. This parameter \ref MAY
- *                         be NULL.
- *
- * \retval #AOM_CODEC_OK
- *     Bitstream is parsable and stream information updated
- */
-aom_codec_err_t aom_codec_get_stream_info(aom_codec_ctx_t *ctx,
-                                          aom_codec_stream_info_t *si);
-
-/*!\brief Decode data
- *
- * Processes a buffer of coded data. If the processing results in a new
- * decoded frame becoming available, PUT_SLICE and PUT_FRAME events may be
- * generated, as appropriate. Encoded data \ref MUST be passed in DTS (decode
- * time stamp) order. Frames produced will always be in PTS (presentation
- * time stamp) order.
- * If the decoder is configured with AOM_CODEC_USE_INPUT_FRAGMENTS enabled,
- * data and data_sz can contain a fragment of the encoded frame. Fragment
- * \#n must contain at least partition \#n, but can also contain subsequent
- * partitions (\#n+1 - \#n+i), and if so, fragments \#n+1, .., \#n+i must
- * be empty. When no more data is available, this function should be called
- * with NULL as data and 0 as data_sz. The memory passed to this function
- * must be available until the frame has been decoded.
- *
- * \param[in] ctx          Pointer to this instance's context
- * \param[in] data         Pointer to this block of new coded data. If
- *                         NULL, a AOM_CODEC_CB_PUT_FRAME event is posted
- *                         for the previously decoded frame.
- * \param[in] data_sz      Size of the coded data, in bytes.
- * \param[in] user_priv    Application specific data to associate with
- *                         this frame.
- * \param[in] deadline     Soft deadline the decoder should attempt to meet,
- *                         in us. Set to zero for unlimited.
- *
- * \return Returns #AOM_CODEC_OK if the coded data was processed completely
- *         and future pictures can be decoded without error. Otherwise,
- *         see the descriptions of the other error codes in ::aom_codec_err_t
- *         for recoverability capabilities.
- */
-aom_codec_err_t aom_codec_decode(aom_codec_ctx_t *ctx, const uint8_t *data,
-                                 unsigned int data_sz, void *user_priv,
-                                 long deadline);
-
-/*!\brief Decoded frames iterator
- *
- * Iterates over a list of the frames available for display. The iterator
- * storage should be initialized to NULL to start the iteration. Iteration is
- * complete when this function returns NULL.
- *
- * The list of available frames becomes valid upon completion of the
- * aom_codec_decode call, and remains valid until the next call to
- * aom_codec_decode.
- *
- * \param[in]     ctx      Pointer to this instance's context
- * \param[in,out] iter     Iterator storage, initialized to NULL
- *
- * \return Returns a pointer to an image, if one is ready for display. Frames
- *         produced will always be in PTS (presentation time stamp) order.
- */
-aom_image_t *aom_codec_get_frame(aom_codec_ctx_t *ctx, aom_codec_iter_t *iter);
-
-/*!\defgroup cap_put_frame Frame-Based Decoding Functions
- *
- * The following functions are required to be implemented for all decoders
- * that advertise the AOM_CODEC_CAP_PUT_FRAME capability. Calling these
- * functions
- * for codecs that don't advertise this capability will result in an error
- * code being returned, usually AOM_CODEC_ERROR
- * @{
- */
-
-/*!\brief put frame callback prototype
- *
- * This callback is invoked by the decoder to notify the application of
- * the availability of decoded image data.
- */
-typedef void (*aom_codec_put_frame_cb_fn_t)(void *user_priv,
-                                            const aom_image_t *img);
-
-/*!\brief Register for notification of frame completion.
- *
- * Registers a given function to be called when a decoded frame is
- * available.
- *
- * \param[in] ctx          Pointer to this instance's context
- * \param[in] cb           Pointer to the callback function
- * \param[in] user_priv    User's private data
- *
- * \retval #AOM_CODEC_OK
- *     Callback successfully registered.
- * \retval #AOM_CODEC_ERROR
- *     Decoder context not initialized, or algorithm not capable of
- *     posting slice completion.
- */
-aom_codec_err_t aom_codec_register_put_frame_cb(aom_codec_ctx_t *ctx,
-                                                aom_codec_put_frame_cb_fn_t cb,
-                                                void *user_priv);
-
-/*!@} - end defgroup cap_put_frame */
-
-/*!\defgroup cap_put_slice Slice-Based Decoding Functions
- *
- * The following functions are required to be implemented for all decoders
- * that advertise the AOM_CODEC_CAP_PUT_SLICE capability. Calling these
- * functions
- * for codecs that don't advertise this capability will result in an error
- * code being returned, usually AOM_CODEC_ERROR
- * @{
- */
-
-/*!\brief put slice callback prototype
- *
- * This callback is invoked by the decoder to notify the application of
- * the availability of partially decoded image data. The
- */
-typedef void (*aom_codec_put_slice_cb_fn_t)(void *user_priv,
-                                            const aom_image_t *img,
-                                            const aom_image_rect_t *valid,
-                                            const aom_image_rect_t *update);
-
-/*!\brief Register for notification of slice completion.
- *
- * Registers a given function to be called when a decoded slice is
- * available.
- *
- * \param[in] ctx          Pointer to this instance's context
- * \param[in] cb           Pointer to the callback function
- * \param[in] user_priv    User's private data
- *
- * \retval #AOM_CODEC_OK
- *     Callback successfully registered.
- * \retval #AOM_CODEC_ERROR
- *     Decoder context not initialized, or algorithm not capable of
- *     posting slice completion.
- */
-aom_codec_err_t aom_codec_register_put_slice_cb(aom_codec_ctx_t *ctx,
-                                                aom_codec_put_slice_cb_fn_t cb,
-                                                void *user_priv);
-
-/*!@} - end defgroup cap_put_slice*/
-
-/*!\defgroup cap_external_frame_buffer External Frame Buffer Functions
- *
- * The following section is required to be implemented for all decoders
- * that advertise the AOM_CODEC_CAP_EXTERNAL_FRAME_BUFFER capability.
- * Calling this function for codecs that don't advertise this capability
- * will result in an error code being returned, usually AOM_CODEC_ERROR.
- *
- * \note
- * Currently this only works with AV1.
- * @{
- */
-
-/*!\brief Pass in external frame buffers for the decoder to use.
- *
- * Registers functions to be called when libaom needs a frame buffer
- * to decode the current frame and a function to be called when libaom does
- * not internally reference the frame buffer. This set function must
- * be called before the first call to decode or libaom will assume the
- * default behavior of allocating frame buffers internally.
- *
- * \param[in] ctx          Pointer to this instance's context
- * \param[in] cb_get       Pointer to the get callback function
- * \param[in] cb_release   Pointer to the release callback function
- * \param[in] cb_priv      Callback's private data
- *
- * \retval #AOM_CODEC_OK
- *     External frame buffers will be used by libaom.
- * \retval #AOM_CODEC_INVALID_PARAM
- *     One or more of the callbacks were NULL.
- * \retval #AOM_CODEC_ERROR
- *     Decoder context not initialized, or algorithm not capable of
- *     using external frame buffers.
- *
- * \note
- * When decoding AV1, the application may be required to pass in at least
- * #AOM_MAXIMUM_WORK_BUFFERS external frame
- * buffers.
- */
-aom_codec_err_t aom_codec_set_frame_buffer_functions(
-    aom_codec_ctx_t *ctx, aom_get_frame_buffer_cb_fn_t cb_get,
-    aom_release_frame_buffer_cb_fn_t cb_release, void *cb_priv);
-
-/*!@} - end defgroup cap_external_frame_buffer */
-
-/*!@} - end defgroup decoder*/
-#ifdef __cplusplus
-}
-#endif
-#endif  // AOM_AOM_DECODER_H_

aom/aom_encoder.h

@@ -1,837 +0,0 @@
-/*
- * Copyright (c) 2016, Alliance for Open Media. All rights reserved
- *
- * This source code is subject to the terms of the BSD 2 Clause License and
- * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
- * was not distributed with this source code in the LICENSE file, you can
- * obtain it at www.aomedia.org/license/software. If the Alliance for Open
- * Media Patent License 1.0 was not distributed with this source code in the
- * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
- */
-#ifndef AOM_AOM_ENCODER_H_
-#define AOM_AOM_ENCODER_H_
-
-/*!\defgroup encoder Encoder Algorithm Interface
- * \ingroup codec
- * This abstraction allows applications using this encoder to easily support
- * multiple video formats with minimal code duplication. This section describes
- * the interface common to all encoders.
- * @{
- */
-
-/*!\file
- * \brief Describes the encoder algorithm interface to applications.
- *
- * This file describes the interface between an application and a
- * video encoder algorithm.
- *
- */
-#ifdef __cplusplus
-extern "C" {
-#endif
-
-#include "./aom_codec.h"
-
-/*!\brief Current ABI version number
- *
- * \internal
- * If this file is altered in any way that changes the ABI, this value
- * must be bumped.  Examples include, but are not limited to, changing
- * types, removing or reassigning enums, adding/removing/rearranging
- * fields to structures
- */
-#define AOM_ENCODER_ABI_VERSION \
-  (5 + AOM_CODEC_ABI_VERSION) /**<\hideinitializer*/
-
-/*! \brief Encoder capabilities bitfield
- *
- *  Each encoder advertises the capabilities it supports as part of its
- *  ::aom_codec_iface_t interface structure. Capabilities are extra
- *  interfaces or functionality, and are not required to be supported
- *  by an encoder.
- *
- *  The available flags are specified by AOM_CODEC_CAP_* defines.
- */
-#define AOM_CODEC_CAP_PSNR 0x10000 /**< Can issue PSNR packets */
-
-/*! Can output one partition at a time. Each partition is returned in its
- *  own AOM_CODEC_CX_FRAME_PKT, with the FRAME_IS_FRAGMENT flag set for
- *  every partition but the last. In this mode all frames are always
- *  returned partition by partition.
- */
-#define AOM_CODEC_CAP_OUTPUT_PARTITION 0x20000
-
-/*! Can support input images at greater than 8 bitdepth.
- */
-#define AOM_CODEC_CAP_HIGHBITDEPTH 0x40000
-
-/*! \brief Initialization-time Feature Enabling
- *
- *  Certain codec features must be known at initialization time, to allow
- *  for proper memory allocation.
- *
- *  The available flags are specified by AOM_CODEC_USE_* defines.
- */
-#define AOM_CODEC_USE_PSNR 0x10000 /**< Calculate PSNR on each frame */
-/*!\brief Make the encoder output one  partition at a time. */
-#define AOM_CODEC_USE_OUTPUT_PARTITION 0x20000
-#define AOM_CODEC_USE_HIGHBITDEPTH 0x40000 /**< Use high bitdepth */
-
-/*!\brief Generic fixed size buffer structure
- *
- * This structure is able to hold a reference to any fixed size buffer.
- */
-typedef struct aom_fixed_buf {
-  void *buf;       /**< Pointer to the data */
-  size_t sz;       /**< Length of the buffer, in chars */
-} aom_fixed_buf_t; /**< alias for struct aom_fixed_buf */
-
-/*!\brief Time Stamp Type
- *
- * An integer, which when multiplied by the stream's time base, provides
- * the absolute time of a sample.
- */
-typedef int64_t aom_codec_pts_t;
-
-/*!\brief Compressed Frame Flags
- *
- * This type represents a bitfield containing information about a compressed
- * frame that may be useful to an application. The most significant 16 bits
- * can be used by an algorithm to provide additional detail, for example to
- * support frame types that are codec specific (MPEG-1 D-frames for example)
- */
-typedef uint32_t aom_codec_frame_flags_t;
-#define AOM_FRAME_IS_KEY 0x1 /**< frame is the start of a GOP */
-/*!\brief frame can be dropped without affecting the stream (no future frame
- * depends on this one) */
-#define AOM_FRAME_IS_DROPPABLE 0x2
-/*!\brief frame should be decoded but will not be shown */
-#define AOM_FRAME_IS_INVISIBLE 0x4
-/*!\brief this is a fragment of the encoded frame */
-#define AOM_FRAME_IS_FRAGMENT 0x8
-
-/*!\brief Error Resilient flags
- *
- * These flags define which error resilient features to enable in the
- * encoder. The flags are specified through the
- * aom_codec_enc_cfg::g_error_resilient variable.
- */
-typedef uint32_t aom_codec_er_flags_t;
-/*!\brief Improve resiliency against losses of whole frames */
-#define AOM_ERROR_RESILIENT_DEFAULT 0x1
-/*!\brief The frame partitions are independently decodable by the bool decoder,
- * meaning that partitions can be decoded even though earlier partitions have
- * been lost. Note that intra prediction is still done over the partition
- * boundary. */
-#define AOM_ERROR_RESILIENT_PARTITIONS 0x2
-
-/*!\brief Encoder output packet variants
- *
- * This enumeration lists the different kinds of data packets that can be
- * returned by calls to aom_codec_get_cx_data(). Algorithms \ref MAY
- * extend this list to provide additional functionality.
- */
-enum aom_codec_cx_pkt_kind {
-  AOM_CODEC_CX_FRAME_PKT,    /**< Compressed video frame */
-  AOM_CODEC_STATS_PKT,       /**< Two-pass statistics for this frame */
-  AOM_CODEC_FPMB_STATS_PKT,  /**< first pass mb statistics for this frame */
-  AOM_CODEC_PSNR_PKT,        /**< PSNR statistics for this frame */
-  AOM_CODEC_CUSTOM_PKT = 256 /**< Algorithm extensions  */
-};
-
-/*!\brief Encoder output packet
- *
- * This structure contains the different kinds of output data the encoder
- * may produce while compressing a frame.
- */
-typedef struct aom_codec_cx_pkt {
-  enum aom_codec_cx_pkt_kind kind; /**< packet variant */
-  union {
-    struct {
-      void *buf; /**< compressed data buffer */
-      size_t sz; /**< length of compressed data */
-      /*!\brief time stamp to show frame (in timebase units) */
-      aom_codec_pts_t pts;
-      /*!\brief duration to show frame (in timebase units) */
-      unsigned long duration;
-      aom_codec_frame_flags_t flags; /**< flags for this frame */
-      /*!\brief the partition id defines the decoding order of the partitions.
-       * Only applicable when "output partition" mode is enabled. First
-       * partition has id 0.*/
-      int partition_id;
-    } frame;                            /**< data for compressed frame packet */
-    aom_fixed_buf_t twopass_stats;      /**< data for two-pass packet */
-    aom_fixed_buf_t firstpass_mb_stats; /**< first pass mb packet */
-    struct aom_psnr_pkt {
-      unsigned int samples[4]; /**< Number of samples, total/y/u/v */
-      uint64_t sse[4];         /**< sum squared error, total/y/u/v */
-      double psnr[4];          /**< PSNR, total/y/u/v */
-    } psnr;                    /**< data for PSNR packet */
-    aom_fixed_buf_t raw;       /**< data for arbitrary packets */
-
-    /* This packet size is fixed to allow codecs to extend this
-     * interface without having to manage storage for raw packets,
-     * i.e., if it's smaller than 128 bytes, you can store in the
-     * packet list directly.
-     */
-    char pad[128 - sizeof(enum aom_codec_cx_pkt_kind)]; /**< fixed sz */
-  } data;                                               /**< packet data */
-} aom_codec_cx_pkt_t; /**< alias for struct aom_codec_cx_pkt */
-
-/*!\brief Rational Number
- *
- * This structure holds a fractional value.
- */
-typedef struct aom_rational {
-  int num;        /**< fraction numerator */
-  int den;        /**< fraction denominator */
-} aom_rational_t; /**< alias for struct aom_rational */
-
-/*!\brief Multi-pass Encoding Pass */
-enum aom_enc_pass {
-  AOM_RC_ONE_PASS,   /**< Single pass mode */
-  AOM_RC_FIRST_PASS, /**< First pass of multi-pass mode */
-  AOM_RC_LAST_PASS   /**< Final pass of multi-pass mode */
-};
-
-/*!\brief Rate control mode */
-enum aom_rc_mode {
-  AOM_VBR, /**< Variable Bit Rate (VBR) mode */
-  AOM_CBR, /**< Constant Bit Rate (CBR) mode */
-  AOM_CQ,  /**< Constrained Quality (CQ)  mode */
-  AOM_Q,   /**< Constant Quality (Q) mode */
-};
-
-/*!\brief Keyframe placement mode.
- *
- * This enumeration determines whether keyframes are placed automatically by
- * the encoder or whether this behavior is disabled. Older releases of this
- * SDK were implemented such that AOM_KF_FIXED meant keyframes were disabled.
- * This name is confusing for this behavior, so the new symbols to be used
- * are AOM_KF_AUTO and AOM_KF_DISABLED.
- */
-enum aom_kf_mode {
-  AOM_KF_FIXED,       /**< deprecated, implies AOM_KF_DISABLED */
-  AOM_KF_AUTO,        /**< Encoder determines optimal placement automatically */
-  AOM_KF_DISABLED = 0 /**< Encoder does not place keyframes. */
-};
-
-/*!\brief Encoded Frame Flags
- *
- * This type indicates a bitfield to be passed to aom_codec_encode(), defining
- * per-frame boolean values. By convention, bits common to all codecs will be
- * named AOM_EFLAG_*, and bits specific to an algorithm will be named
- * /algo/_eflag_*. The lower order 16 bits are reserved for common use.
- */
-typedef long aom_enc_frame_flags_t;
-#define AOM_EFLAG_FORCE_KF (1 << 0) /**< Force this frame to be a keyframe */
-
-/*!\brief Encoder configuration structure
- *
- * This structure contains the encoder settings that have common representations
- * across all codecs. This doesn't imply that all codecs support all features,
- * however.
- */
-typedef struct aom_codec_enc_cfg {
-  /*
-   * generic settings (g)
-   */
-
-  /*!\brief Algorithm specific "usage" value
-   *
-   * Algorithms may define multiple values for usage, which may convey the
-   * intent of how the application intends to use the stream. If this value
-   * is non-zero, consult the documentation for the codec to determine its
-   * meaning.
-   */
-  unsigned int g_usage;
-
-  /*!\brief Maximum number of threads to use
-   *
-   * For multi-threaded implementations, use no more than this number of
-   * threads. The codec may use fewer threads than allowed. The value
-   * 0 is equivalent to the value 1.
-   */
-  unsigned int g_threads;
-
-  /*!\brief Bitstream profile to use
-   *
-   * Some codecs support a notion of multiple bitstream profiles. Typically
-   * this maps to a set of features that are turned on or off. Often the
-   * profile to use is determined by the features of the intended decoder.
-   * Consult the documentation for the codec to determine the valid values
-   * for this parameter, or set to zero for a sane default.
-   */
-  unsigned int g_profile; /**< profile of bitstream to use */
-
-  /*!\brief Width of the frame
-   *
-   * This value identifies the presentation resolution of the frame,
-   * in pixels. Note that the frames passed as input to the encoder must
-   * have this resolution. Frames will be presented by the decoder in this
-   * resolution, independent of any spatial resampling the encoder may do.
-   */
-  unsigned int g_w;
-
-  /*!\brief Height of the frame
-   *
-   * This value identifies the presentation resolution of the frame,
-   * in pixels. Note that the frames passed as input to the encoder must
-   * have this resolution. Frames will be presented by the decoder in this
-   * resolution, independent of any spatial resampling the encoder may do.
-   */
-  unsigned int g_h;
-
-  /*!\brief Bit-depth of the codec
-   *
-   * This value identifies the bit_depth of the codec,
-   * Only certain bit-depths are supported as identified in the
-   * aom_bit_depth_t enum.
-   */
-  aom_bit_depth_t g_bit_depth;
-
-  /*!\brief Bit-depth of the input frames
-   *
-   * This value identifies the bit_depth of the input frames in bits.
-   * Note that the frames passed as input to the encoder must have
-   * this bit-depth.
-   */
-  unsigned int g_input_bit_depth;
-
-  /*!\brief Stream timebase units
-   *
-   * Indicates the smallest interval of time, in seconds, used by the stream.
-   * For fixed frame rate material, or variable frame rate material where
-   * frames are timed at a multiple of a given clock (ex: video capture),
-   * the \ref RECOMMENDED method is to set the timebase to the reciprocal
-   * of the frame rate (ex: 1001/30000 for 29.970 Hz NTSC). This allows the
-   * pts to correspond to the frame number, which can be handy. For
-   * re-encoding video from containers with absolute time timestamps, the
-   * \ref RECOMMENDED method is to set the timebase to that of the parent
-   * container or multimedia framework (ex: 1/1000 for ms, as in FLV).
-   */
-  struct aom_rational g_timebase;
-
-  /*!\brief Enable error resilient modes.
-   *
-   * The error resilient bitfield indicates to the encoder which features
-   * it should enable to take measures for streaming over lossy or noisy
-   * links.
-   */
-  aom_codec_er_flags_t g_error_resilient;
-
-  /*!\brief Multi-pass Encoding Mode
-   *
-   * This value should be set to the current phase for multi-pass encoding.
-   * For single pass, set to #AOM_RC_ONE_PASS.
-   */
-  enum aom_enc_pass g_pass;
-
-  /*!\brief Allow lagged encoding
-   *
-   * If set, this value allows the encoder to consume a number of input
-   * frames before producing output frames. This allows the encoder to
-   * base decisions for the current frame on future frames. This does
-   * increase the latency of the encoding pipeline, so it is not appropriate
-   * in all situations (ex: realtime encoding).
-   *
-   * Note that this is a maximum value -- the encoder may produce frames
-   * sooner than the given limit. Set this value to 0 to disable this
-   * feature.
-   */
-  unsigned int g_lag_in_frames;
-
-  /*
-   * rate control settings (rc)
-   */
-
-  /*!\brief Temporal resampling configuration, if supported by the codec.
-   *
-   * Temporal resampling allows the codec to "drop" frames as a strategy to
-   * meet its target data rate. This can cause temporal discontinuities in
-   * the encoded video, which may appear as stuttering during playback. This
-   * trade-off is often acceptable, but for many applications is not. It can
-   * be disabled in these cases.
-   *
-   * Note that not all codecs support this feature. All aom AVx codecs do.
-   * For other codecs, consult the documentation for that algorithm.
-   *
-   * This threshold is described as a percentage of the target data buffer.
-   * When the data buffer falls below this percentage of fullness, a
-   * dropped frame is indicated. Set the threshold to zero (0) to disable
-   * this feature.
-   */
-  unsigned int rc_dropframe_thresh;
-
-  /*!\brief Enable/disable spatial resampling, if supported by the codec.
-   *
-   * Spatial resampling allows the codec to compress a lower resolution
-   * version of the frame, which is then upscaled by the encoder to the
-   * correct presentation resolution. This increases visual quality at
-   * low data rates, at the expense of CPU time on the encoder/decoder.
-   */
-  unsigned int rc_resize_allowed;
-
-  /*!\brief Internal coded frame width.
-   *
-   * If spatial resampling is enabled this specifies the width of the
-   * encoded frame.
-   */
-  unsigned int rc_scaled_width;
-
-  /*!\brief Internal coded frame height.
-   *
-   * If spatial resampling is enabled this specifies the height of the
-   * encoded frame.
-   */
-  unsigned int rc_scaled_height;
-
-  /*!\brief Spatial resampling up watermark.
-   *
-   * This threshold is described as a percentage of the target data buffer.
-   * When the data buffer rises above this percentage of fullness, the
-   * encoder will step up to a higher resolution version of the frame.
-   */
-  unsigned int rc_resize_up_thresh;
-
-  /*!\brief Spatial resampling down watermark.
-   *
-   * This threshold is described as a percentage of the target data buffer.
-   * When the data buffer falls below this percentage of fullness, the
-   * encoder will step down to a lower resolution version of the frame.
-   */
-  unsigned int rc_resize_down_thresh;
-
-  /*!\brief Rate control algorithm to use.
-   *
-   * Indicates whether the end usage of this stream is to be streamed over
-   * a bandwidth constrained link, indicating that Constant Bit Rate (CBR)
-   * mode should be used, or whether it will be played back on a high
-   * bandwidth link, as from a local disk, where higher variations in
-   * bitrate are acceptable.
-   */
-  enum aom_rc_mode rc_end_usage;
-
-  /*!\brief Two-pass stats buffer.
-   *
-   * A buffer containing all of the stats packets produced in the first
-   * pass, concatenated.
-   */
-  aom_fixed_buf_t rc_twopass_stats_in;
-
-  /*!\brief first pass mb stats buffer.
-   *
-   * A buffer containing all of the first pass mb stats packets produced
-   * in the first pass, concatenated.
-   */
-  aom_fixed_buf_t rc_firstpass_mb_stats_in;
-
-  /*!\brief Target data rate
-   *
-   * Target bandwidth to use for this stream, in kilobits per second.
-   */
-  unsigned int rc_target_bitrate;
-
-  /*
-   * quantizer settings
-   */
-
-  /*!\brief Minimum (Best Quality) Quantizer
-   *
-   * The quantizer is the most direct control over the quality of the
-   * encoded image. The range of valid values for the quantizer is codec
-   * specific. Consult the documentation for the codec to determine the
-   * values to use. To determine the range programmatically, call
-   * aom_codec_enc_config_default() with a usage value of 0.
-   */
-  unsigned int rc_min_quantizer;
-
-  /*!\brief Maximum (Worst Quality) Quantizer
-   *
-   * The quantizer is the most direct control over the quality of the
-   * encoded image. The range of valid values for the quantizer is codec
-   * specific. Consult the documentation for the codec to determine the
-   * values to use. To determine the range programmatically, call
-   * aom_codec_enc_config_default() with a usage value of 0.
-   */
-  unsigned int rc_max_quantizer;
-
-  /*
-   * bitrate tolerance
-   */
-
-  /*!\brief Rate control adaptation undershoot control
-   *
-   * This value, expressed as a percentage of the target bitrate,
-   * controls the maximum allowed adaptation speed of the codec.
-   * This factor controls the maximum amount of bits that can
-   * be subtracted from the target bitrate in order to compensate
-   * for prior overshoot.
-   *
-   * Valid values in the range 0-1000.
-   */
-  unsigned int rc_undershoot_pct;
-
-  /*!\brief Rate control adaptation overshoot control
-   *
-   * This value, expressed as a percentage of the target bitrate,
-   * controls the maximum allowed adaptation speed of the codec.
-   * This factor controls the maximum amount of bits that can
-   * be added to the target bitrate in order to compensate for
-   * prior undershoot.
-   *
-   * Valid values in the range 0-1000.
-   */
-  unsigned int rc_overshoot_pct;
-
-  /*
-   * decoder buffer model parameters
-   */
-
-  /*!\brief Decoder Buffer Size
-   *
-   * This value indicates the amount of data that may be buffered by the
-   * decoding application. Note that this value is expressed in units of
-   * time (milliseconds). For example, a value of 5000 indicates that the
-   * client will buffer (at least) 5000ms worth of encoded data. Use the
-   * target bitrate (#rc_target_bitrate) to convert to bits/bytes, if
-   * necessary.
-   */
-  unsigned int rc_buf_sz;
-
-  /*!\brief Decoder Buffer Initial Size
-   *
-   * This value indicates the amount of data that will be buffered by the
-   * decoding application prior to beginning playback. This value is
-   * expressed in units of time (milliseconds). Use the target bitrate
-   * (#rc_target_bitrate) to convert to bits/bytes, if necessary.
-   */
-  unsigned int rc_buf_initial_sz;
-
-  /*!\brief Decoder Buffer Optimal Size
-   *
-   * This value indicates the amount of data that the encoder should try
-   * to maintain in the decoder's buffer. This value is expressed in units
-   * of time (milliseconds). Use the target bitrate (#rc_target_bitrate)
-   * to convert to bits/bytes, if necessary.
-   */
-  unsigned int rc_buf_optimal_sz;
-
-  /*
-   * 2 pass rate control parameters
-   */
-
-  /*!\brief Two-pass mode CBR/VBR bias
-   *
-   * Bias, expressed on a scale of 0 to 100, for determining target size
-   * for the current frame. The value 0 indicates the optimal CBR mode
-   * value should be used. The value 100 indicates the optimal VBR mode
-   * value should be used. Values in between indicate which way the
-   * encoder should "lean."
-   */
-  unsigned int rc_2pass_vbr_bias_pct;
-
-  /*!\brief Two-pass mode per-GOP minimum bitrate
-   *
-   * This value, expressed as a percentage of the target bitrate, indicates
-   * the minimum bitrate to be used for a single GOP (aka "section")
-   */
-  unsigned int rc_2pass_vbr_minsection_pct;
-
-  /*!\brief Two-pass mode per-GOP maximum bitrate
-   *
-   * This value, expressed as a percentage of the target bitrate, indicates
-   * the maximum bitrate to be used for a single GOP (aka "section")
-   */
-  unsigned int rc_2pass_vbr_maxsection_pct;
-
-  /*
-   * keyframing settings (kf)
-   */
-
-  /*!\brief Keyframe placement mode
-   *
-   * This value indicates whether the encoder should place keyframes at a
-   * fixed interval, or determine the optimal placement automatically
-   * (as governed by the #kf_min_dist and #kf_max_dist parameters)
-   */
-  enum aom_kf_mode kf_mode;
-
-  /*!\brief Keyframe minimum interval
-   *
-   * This value, expressed as a number of frames, prevents the encoder from
-   * placing a keyframe nearer than kf_min_dist to the previous keyframe. At
-   * least kf_min_dist frames non-keyframes will be coded before the next
-   * keyframe. Set kf_min_dist equal to kf_max_dist for a fixed interval.
-   */
-  unsigned int kf_min_dist;
-
-  /*!\brief Keyframe maximum interval
-   *
-   * This value, expressed as a number of frames, forces the encoder to code
-   * a keyframe if one has not been coded in the last kf_max_dist frames.
-   * A value of 0 implies all frames will be keyframes. Set kf_min_dist
-   * equal to kf_max_dist for a fixed interval.
-   */
-  unsigned int kf_max_dist;
-} aom_codec_enc_cfg_t; /**< alias for struct aom_codec_enc_cfg */
-
-/*!\brief Initialize an encoder instance
- *
- * Initializes a encoder context using the given interface. Applications
- * should call the aom_codec_enc_init convenience macro instead of this
- * function directly, to ensure that the ABI version number parameter
- * is properly initialized.
- *
- * If the library was configured with --disable-multithread, this call
- * is not thread safe and should be guarded with a lock if being used
- * in a multithreaded context.
- *
- * \param[in]    ctx     Pointer to this instance's context.
- * \param[in]    iface   Pointer to the algorithm interface to use.
- * \param[in]    cfg     Configuration to use, if known. May be NULL.
- * \param[in]    flags   Bitfield of AOM_CODEC_USE_* flags
- * \param[in]    ver     ABI version number. Must be set to
- *                       AOM_ENCODER_ABI_VERSION
- * \retval #AOM_CODEC_OK
- *     The decoder algorithm initialized.
- * \retval #AOM_CODEC_MEM_ERROR
- *     Memory allocation failed.
- */
-aom_codec_err_t aom_codec_enc_init_ver(aom_codec_ctx_t *ctx,
-                                       aom_codec_iface_t *iface,
-                                       const aom_codec_enc_cfg_t *cfg,
-                                       aom_codec_flags_t flags, int ver);
-
-/*!\brief Convenience macro for aom_codec_enc_init_ver()
- *
- * Ensures the ABI version parameter is properly set.
- */
-#define aom_codec_enc_init(ctx, iface, cfg, flags) \
-  aom_codec_enc_init_ver(ctx, iface, cfg, flags, AOM_ENCODER_ABI_VERSION)
-
-/*!\brief Initialize multi-encoder instance
- *
- * Initializes multi-encoder context using the given interface.
- * Applications should call the aom_codec_enc_init_multi convenience macro
- * instead of this function directly, to ensure that the ABI version number
- * parameter is properly initialized.
- *
- * \param[in]    ctx     Pointer to this instance's context.
- * \param[in]    iface   Pointer to the algorithm interface to use.
- * \param[in]    cfg     Configuration to use, if known. May be NULL.
- * \param[in]    num_enc Total number of encoders.
- * \param[in]    flags   Bitfield of AOM_CODEC_USE_* flags
- * \param[in]    dsf     Pointer to down-sampling factors.
- * \param[in]    ver     ABI version number. Must be set to
- *                       AOM_ENCODER_ABI_VERSION
- * \retval #AOM_CODEC_OK
- *     The decoder algorithm initialized.
- * \retval #AOM_CODEC_MEM_ERROR
- *     Memory allocation failed.
- */
-aom_codec_err_t aom_codec_enc_init_multi_ver(
-    aom_codec_ctx_t *ctx, aom_codec_iface_t *iface, aom_codec_enc_cfg_t *cfg,
-    int num_enc, aom_codec_flags_t flags, aom_rational_t *dsf, int ver);
-
-/*!\brief Convenience macro for aom_codec_enc_init_multi_ver()
- *
- * Ensures the ABI version parameter is properly set.
- */
-#define aom_codec_enc_init_multi(ctx, iface, cfg, num_enc, flags, dsf) \
-  aom_codec_enc_init_multi_ver(ctx, iface, cfg, num_enc, flags, dsf,   \
-                               AOM_ENCODER_ABI_VERSION)
-
-/*!\brief Get a default configuration
- *
- * Initializes a encoder configuration structure with default values. Supports
- * the notion of "usages" so that an algorithm may offer different default
- * settings depending on the user's intended goal. This function \ref SHOULD
- * be called by all applications to initialize the configuration structure
- * before specializing the configuration with application specific values.
- *
- * \param[in]    iface     Pointer to the algorithm interface to use.
- * \param[out]   cfg       Configuration buffer to populate.
- * \param[in]    reserved  Must set to 0 for VP8 and AV1.
- *
- * \retval #AOM_CODEC_OK
- *     The configuration was populated.
- * \retval #AOM_CODEC_INCAPABLE
- *     Interface is not an encoder interface.
- * \retval #AOM_CODEC_INVALID_PARAM
- *     A parameter was NULL, or the usage value was not recognized.
- */
-aom_codec_err_t aom_codec_enc_config_default(aom_codec_iface_t *iface,
-                                             aom_codec_enc_cfg_t *cfg,
-                                             unsigned int reserved);
-
-/*!\brief Set or change configuration
- *
- * Reconfigures an encoder instance according to the given configuration.
- *
- * \param[in]    ctx     Pointer to this instance's context
- * \param[in]    cfg     Configuration buffer to use
- *
- * \retval #AOM_CODEC_OK
- *     The configuration was populated.
- * \retval #AOM_CODEC_INCAPABLE
- *     Interface is not an encoder interface.
- * \retval #AOM_CODEC_INVALID_PARAM
- *     A parameter was NULL, or the usage value was not recognized.
- */
-aom_codec_err_t aom_codec_enc_config_set(aom_codec_ctx_t *ctx,
-                                         const aom_codec_enc_cfg_t *cfg);
-
-/*!\brief Get global stream headers
- *
- * Retrieves a stream level global header packet, if supported by the codec.
- *
- * \param[in]    ctx     Pointer to this instance's context
- *
- * \retval NULL
- *     Encoder does not support global header
- * \retval Non-NULL
- *     Pointer to buffer containing global header packet
- */
-aom_fixed_buf_t *aom_codec_get_global_headers(aom_codec_ctx_t *ctx);
-
-/*!\brief deadline parameter analogous to AVx REALTIME mode. */
-#define AOM_DL_REALTIME (1)
-/*!\brief deadline parameter analogous to  AVx GOOD QUALITY mode. */
-#define AOM_DL_GOOD_QUALITY (1000000)
-/*!\brief deadline parameter analogous to AVx BEST QUALITY mode. */
-#define AOM_DL_BEST_QUALITY (0)
-/*!\brief Encode a frame
- *
- * Encodes a video frame at the given "presentation time." The presentation
- * time stamp (PTS) \ref MUST be strictly increasing.
- *
- * The encoder supports the notion of a soft real-time deadline. Given a
- * non-zero value to the deadline parameter, the encoder will make a "best
- * effort" guarantee to  return before the given time slice expires. It is
- * implicit that limiting the available time to encode will degrade the
- * output quality. The encoder can be given an unlimited time to produce the
- * best possible frame by specifying a deadline of '0'. This deadline
- * supercedes the AVx notion of "best quality, good quality, realtime".
- * Applications that wish to map these former settings to the new deadline
- * based system can use the symbols #AOM_DL_REALTIME, #AOM_DL_GOOD_QUALITY,
- * and #AOM_DL_BEST_QUALITY.
- *
- * When the last frame has been passed to the encoder, this function should
- * continue to be called, with the img parameter set to NULL. This will
- * signal the end-of-stream condition to the encoder and allow it to encode
- * any held buffers. Encoding is complete when aom_codec_encode() is called
- * and aom_codec_get_cx_data() returns no data.
- *
- * \param[in]    ctx       Pointer to this instance's context
- * \param[in]    img       Image data to encode, NULL to flush.
- * \param[in]    pts       Presentation time stamp, in timebase units.
- * \param[in]    duration  Duration to show frame, in timebase units.
- * \param[in]    flags     Flags to use for encoding this frame.
- * \param[in]    deadline  Time to spend encoding, in microseconds. (0=infinite)
- *
- * \retval #AOM_CODEC_OK
- *     The configuration was populated.
- * \retval #AOM_CODEC_INCAPABLE
- *     Interface is not an encoder interface.
- * \retval #AOM_CODEC_INVALID_PARAM
- *     A parameter was NULL, the image format is unsupported, etc.
- */
-aom_codec_err_t aom_codec_encode(aom_codec_ctx_t *ctx, const aom_image_t *img,
-                                 aom_codec_pts_t pts, unsigned long duration,
-                                 aom_enc_frame_flags_t flags,
-                                 unsigned long deadline);
-
-/*!\brief Set compressed data output buffer
- *
- * Sets the buffer that the codec should output the compressed data
- * into. This call effectively sets the buffer pointer returned in the
- * next AOM_CODEC_CX_FRAME_PKT packet. Subsequent packets will be
- * appended into this buffer. The buffer is preserved across frames,
- * so applications must periodically call this function after flushing
- * the accumulated compressed data to disk or to the network to reset
- * the pointer to the buffer's head.
- *
- * `pad_before` bytes will be skipped before writing the compressed
- * data, and `pad_after` bytes will be appended to the packet. The size
- * of the packet will be the sum of the size of the actual compressed
- * data, pad_before, and pad_after. The padding bytes will be preserved
- * (not overwritten).
- *
- * Note that calling this function does not guarantee that the returned
- * compressed data will be placed into the specified buffer. In the
- * event that the encoded data will not fit into the buffer provided,
- * the returned packet \ref MAY point to an internal buffer, as it would
- * if this call were never used. In this event, the output packet will
- * NOT have any padding, and the application must free space and copy it
- * to the proper place. This is of particular note in configurations
- * that may output multiple packets for a single encoded frame (e.g., lagged
- * encoding) or if the application does not reset the buffer periodically.
- *
- * Applications may restore the default behavior of the codec providing
- * the compressed data buffer by calling this function with a NULL
- * buffer.
- *
- * Applications \ref MUSTNOT call this function during iteration of
- * aom_codec_get_cx_data().
- *
- * \param[in]    ctx         Pointer to this instance's context
- * \param[in]    buf         Buffer to store compressed data into
- * \param[in]    pad_before  Bytes to skip before writing compressed data
- * \param[in]    pad_after   Bytes to skip after writing compressed data
- *
- * \retval #AOM_CODEC_OK
- *     The buffer was set successfully.
- * \retval #AOM_CODEC_INVALID_PARAM
- *     A parameter was NULL, the image format is unsupported, etc.
- */
-aom_codec_err_t aom_codec_set_cx_data_buf(aom_codec_ctx_t *ctx,
-                                          const aom_fixed_buf_t *buf,
-                                          unsigned int pad_before,
-                                          unsigned int pad_after);
-
-/*!\brief Encoded data iterator
- *
- * Iterates over a list of data packets to be passed from the encoder to the
- * application. The different kinds of packets available are enumerated in
- * #aom_codec_cx_pkt_kind.
- *
- * #AOM_CODEC_CX_FRAME_PKT packets should be passed to the application's
- * muxer. Multiple compressed frames may be in the list.
- * #AOM_CODEC_STATS_PKT packets should be appended to a global buffer.
- *
- * The application \ref MUST silently ignore any packet kinds that it does
- * not recognize or support.
- *
- * The data buffers returned from this function are only guaranteed to be
- * valid until the application makes another call to any aom_codec_* function.
- *
- * \param[in]     ctx      Pointer to this instance's context
- * \param[in,out] iter     Iterator storage, initialized to NULL
- *
- * \return Returns a pointer to an output data packet (compressed frame data,
- *         two-pass statistics, etc.) or NULL to signal end-of-list.
- *
- */
-const aom_codec_cx_pkt_t *aom_codec_get_cx_data(aom_codec_ctx_t *ctx,
-                                                aom_codec_iter_t *iter);
-
-/*!\brief Get Preview Frame
- *
- * Returns an image that can be used as a preview. Shows the image as it would
- * exist at the decompressor. The application \ref MUST NOT write into this
- * image buffer.
- *
- * \param[in]     ctx      Pointer to this instance's context
- *
- * \return Returns a pointer to a preview image, or NULL if no image is
- *         available.
- *
- */
-const aom_image_t *aom_codec_get_preview_frame(aom_codec_ctx_t *ctx);
-
-/*!@} - end defgroup encoder*/
-#ifdef __cplusplus
-}
-#endif
-#endif  // AOM_AOM_ENCODER_H_

aom/aom_frame_buffer.h

@@ -1,84 +0,0 @@
-/*
- * Copyright (c) 2016, Alliance for Open Media. All rights reserved
- *
- * This source code is subject to the terms of the BSD 2 Clause License and
- * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
- * was not distributed with this source code in the LICENSE file, you can
- * obtain it at www.aomedia.org/license/software. If the Alliance for Open
- * Media Patent License 1.0 was not distributed with this source code in the
- * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
- */
-
-#ifndef AOM_AOM_FRAME_BUFFER_H_
-#define AOM_AOM_FRAME_BUFFER_H_
-
-/*!\file
- * \brief Describes the decoder external frame buffer interface.
- */
-
-#ifdef __cplusplus
-extern "C" {
-#endif
-
-#include "./aom_integer.h"
-
-/*!\brief The maximum number of work buffers used by libaom.
- *  Support maximum 4 threads to decode video in parallel.
- *  Each thread will use one work buffer.
- * TODO(hkuang): Add support to set number of worker threads dynamically.
- */
-#define AOM_MAXIMUM_WORK_BUFFERS 8
-
-/*!\brief The maximum number of reference buffers that a AV1 encoder may use.
- */
-#define AOM_MAXIMUM_REF_BUFFERS 8
-
-/*!\brief External frame buffer
- *
- * This structure holds allocated frame buffers used by the decoder.
- */
-typedef struct aom_codec_frame_buffer {
-  uint8_t *data; /**< Pointer to the data buffer */
-  size_t size;   /**< Size of data in bytes */
-  void *priv;    /**< Frame's private data */
-} aom_codec_frame_buffer_t;
-
-/*!\brief get frame buffer callback prototype
- *
- * This callback is invoked by the decoder to retrieve data for the frame
- * buffer in order for the decode call to complete. The callback must
- * allocate at least min_size in bytes and assign it to fb->data. The callback
- * must zero out all the data allocated. Then the callback must set fb->size
- * to the allocated size. The application does not need to align the allocated
- * data. The callback is triggered when the decoder needs a frame buffer to
- * decode a compressed image into. This function may be called more than once
- * for every call to aom_codec_decode. The application may set fb->priv to
- * some data which will be passed back in the ximage and the release function
- * call. |fb| is guaranteed to not be NULL. On success the callback must
- * return 0. Any failure the callback must return a value less than 0.
- *
- * \param[in] priv         Callback's private data
- * \param[in] new_size     Size in bytes needed by the buffer
- * \param[in,out] fb       Pointer to aom_codec_frame_buffer_t
- */
-typedef int (*aom_get_frame_buffer_cb_fn_t)(void *priv, size_t min_size,
-                                            aom_codec_frame_buffer_t *fb);
-
-/*!\brief release frame buffer callback prototype
- *
- * This callback is invoked by the decoder when the frame buffer is not
- * referenced by any other buffers. |fb| is guaranteed to not be NULL. On
- * success the callback must return 0. Any failure the callback must return
- * a value less than 0.
- *
- * \param[in] priv         Callback's private data
- * \param[in] fb           Pointer to aom_codec_frame_buffer_t
- */
-typedef int (*aom_release_frame_buffer_cb_fn_t)(void *priv,
-                                                aom_codec_frame_buffer_t *fb);
-
-#ifdef __cplusplus
-}  // extern "C"
-#endif
-
-#endif  // AOM_AOM_FRAME_BUFFER_H_

aom/aom_image.h

@@ -1,225 +0,0 @@
-/*
- * Copyright (c) 2016, Alliance for Open Media. All rights reserved
- *
- * This source code is subject to the terms of the BSD 2 Clause License and
- * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
- * was not distributed with this source code in the LICENSE file, you can
- * obtain it at www.aomedia.org/license/software. If the Alliance for Open
- * Media Patent License 1.0 was not distributed with this source code in the
- * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
- */
-
-/*!\file
- * \brief Describes the aom image descriptor and associated operations
- *
- */
-#ifndef AOM_AOM_IMAGE_H_
-#define AOM_AOM_IMAGE_H_
-
-#ifdef __cplusplus
-extern "C" {
-#endif
-
-/*!\brief Current ABI version number
- *
- * \internal
- * If this file is altered in any way that changes the ABI, this value
- * must be bumped.  Examples include, but are not limited to, changing
- * types, removing or reassigning enums, adding/removing/rearranging
- * fields to structures
- */
-#define AOM_IMAGE_ABI_VERSION (4) /**<\hideinitializer*/
-
-#define AOM_IMG_FMT_PLANAR 0x100       /**< Image is a planar format. */
-#define AOM_IMG_FMT_UV_FLIP 0x200      /**< V plane precedes U in memory. */
-#define AOM_IMG_FMT_HAS_ALPHA 0x400    /**< Image has an alpha channel. */
-#define AOM_IMG_FMT_HIGHBITDEPTH 0x800 /**< Image uses 16bit framebuffer. */
-
-/*!\brief List of supported image formats */
-typedef enum aom_img_fmt {
-  AOM_IMG_FMT_NONE,
-  AOM_IMG_FMT_RGB24,     /**< 24 bit per pixel packed RGB */
-  AOM_IMG_FMT_RGB32,     /**< 32 bit per pixel packed 0RGB */
-  AOM_IMG_FMT_RGB565,    /**< 16 bit per pixel, 565 */
-  AOM_IMG_FMT_RGB555,    /**< 16 bit per pixel, 555 */
-  AOM_IMG_FMT_UYVY,      /**< UYVY packed YUV */
-  AOM_IMG_FMT_YUY2,      /**< YUYV packed YUV */
-  AOM_IMG_FMT_YVYU,      /**< YVYU packed YUV */
-  AOM_IMG_FMT_BGR24,     /**< 24 bit per pixel packed BGR */
-  AOM_IMG_FMT_RGB32_LE,  /**< 32 bit packed BGR0 */
-  AOM_IMG_FMT_ARGB,      /**< 32 bit packed ARGB, alpha=255 */
-  AOM_IMG_FMT_ARGB_LE,   /**< 32 bit packed BGRA, alpha=255 */
-  AOM_IMG_FMT_RGB565_LE, /**< 16 bit per pixel, gggbbbbb rrrrrggg */
-  AOM_IMG_FMT_RGB555_LE, /**< 16 bit per pixel, gggbbbbb 0rrrrrgg */
-  AOM_IMG_FMT_YV12 =
-      AOM_IMG_FMT_PLANAR | AOM_IMG_FMT_UV_FLIP | 1, /**< planar YVU */
-  AOM_IMG_FMT_I420 = AOM_IMG_FMT_PLANAR | 2,
-  AOM_IMG_FMT_AOMYV12 = AOM_IMG_FMT_PLANAR | AOM_IMG_FMT_UV_FLIP |
-                        3, /** < planar 4:2:0 format with aom color space */
-  AOM_IMG_FMT_AOMI420 = AOM_IMG_FMT_PLANAR | 4,
-  AOM_IMG_FMT_I422 = AOM_IMG_FMT_PLANAR | 5,
-  AOM_IMG_FMT_I444 = AOM_IMG_FMT_PLANAR | 6,
-  AOM_IMG_FMT_I440 = AOM_IMG_FMT_PLANAR | 7,
-  AOM_IMG_FMT_444A = AOM_IMG_FMT_PLANAR | AOM_IMG_FMT_HAS_ALPHA | 6,
-  AOM_IMG_FMT_I42016 = AOM_IMG_FMT_I420 | AOM_IMG_FMT_HIGHBITDEPTH,
-  AOM_IMG_FMT_I42216 = AOM_IMG_FMT_I422 | AOM_IMG_FMT_HIGHBITDEPTH,
-  AOM_IMG_FMT_I44416 = AOM_IMG_FMT_I444 | AOM_IMG_FMT_HIGHBITDEPTH,
-  AOM_IMG_FMT_I44016 = AOM_IMG_FMT_I440 | AOM_IMG_FMT_HIGHBITDEPTH
-} aom_img_fmt_t; /**< alias for enum aom_img_fmt */
-
-/*!\brief List of supported color spaces */
-typedef enum aom_color_space {
-  AOM_CS_UNKNOWN = 0,   /**< Unknown */
-  AOM_CS_BT_601 = 1,    /**< BT.601 */
-  AOM_CS_BT_709 = 2,    /**< BT.709 */
-  AOM_CS_SMPTE_170 = 3, /**< SMPTE.170 */
-  AOM_CS_SMPTE_240 = 4, /**< SMPTE.240 */
-  AOM_CS_BT_2020 = 5,   /**< BT.2020 */
-  AOM_CS_RESERVED = 6,  /**< Reserved */
-  AOM_CS_SRGB = 7       /**< sRGB */
-} aom_color_space_t;    /**< alias for enum aom_color_space */
-
-/*!\brief List of supported color range */
-typedef enum aom_color_range {
-  AOM_CR_STUDIO_RANGE = 0, /**< Y [16..235], UV [16..240] */
-  AOM_CR_FULL_RANGE = 1    /**< YUV/RGB [0..255] */
-} aom_color_range_t;       /**< alias for enum aom_color_range */
-
-/**\brief Image Descriptor */
-typedef struct aom_image {
-  aom_img_fmt_t fmt;       /**< Image Format */
-  aom_color_space_t cs;    /**< Color Space */
-  aom_color_range_t range; /**< Color Range */
-
-  /* Image storage dimensions */
-  unsigned int w;         /**< Stored image width */
-  unsigned int h;         /**< Stored image height */
-  unsigned int bit_depth; /**< Stored image bit-depth */
-
-  /* Image display dimensions */
-  unsigned int d_w; /**< Displayed image width */
-  unsigned int d_h; /**< Displayed image height */
-
-  /* Image intended rendering dimensions */
-  unsigned int r_w; /**< Intended rendering image width */
-  unsigned int r_h; /**< Intended rendering image height */
-
-  /* Chroma subsampling info */
-  unsigned int x_chroma_shift; /**< subsampling order, X */
-  unsigned int y_chroma_shift; /**< subsampling order, Y */
-
-/* Image data pointers. */
-#define AOM_PLANE_PACKED 0  /**< To be used for all packed formats */
-#define AOM_PLANE_Y 0       /**< Y (Luminance) plane */
-#define AOM_PLANE_U 1       /**< U (Chroma) plane */
-#define AOM_PLANE_V 2       /**< V (Chroma) plane */
-#define AOM_PLANE_ALPHA 3   /**< A (Transparency) plane */
-  unsigned char *planes[4]; /**< pointer to the top left pixel for each plane */
-  int stride[4];            /**< stride between rows for each plane */
-
-  int bps; /**< bits per sample (for packed formats) */
-
-  /*!\brief The following member may be set by the application to associate
-   * data with this image.
-   */
-  void *user_priv;
-
-  /* The following members should be treated as private. */
-  unsigned char *img_data; /**< private */
-  int img_data_owner;      /**< private */
-  int self_allocd;         /**< private */
-
-  void *fb_priv; /**< Frame buffer data associated with the image. */
-} aom_image_t;   /**< alias for struct aom_image */
-
-/**\brief Representation of a rectangle on a surface */
-typedef struct aom_image_rect {
-  unsigned int x;   /**< leftmost column */
-  unsigned int y;   /**< topmost row */
-  unsigned int w;   /**< width */
-  unsigned int h;   /**< height */
-} aom_image_rect_t; /**< alias for struct aom_image_rect */
-
-/*!\brief Open a descriptor, allocating storage for the underlying image
- *
- * Returns a descriptor for storing an image of the given format. The
- * storage for the descriptor is allocated on the heap.
- *
- * \param[in]    img       Pointer to storage for descriptor. If this parameter
- *                         is NULL, the storage for the descriptor will be
- *                         allocated on the heap.
- * \param[in]    fmt       Format for the image
- * \param[in]    d_w       Width of the image
- * \param[in]    d_h       Height of the image
- * \param[in]    align     Alignment, in bytes, of the image buffer and
- *                         each row in the image(stride).
- *
- * \return Returns a pointer to the initialized image descriptor. If the img
- *         parameter is non-null, the value of the img parameter will be
- *         returned.
- */
-aom_image_t *aom_img_alloc(aom_image_t *img, aom_img_fmt_t fmt,
-                           unsigned int d_w, unsigned int d_h,
-                           unsigned int align);
-
-/*!\brief Open a descriptor, using existing storage for the underlying image
- *
- * Returns a descriptor for storing an image of the given format. The
- * storage for descriptor has been allocated elsewhere, and a descriptor is
- * desired to "wrap" that storage.
- *
- * \param[in]    img       Pointer to storage for descriptor. If this parameter
- *                         is NULL, the storage for the descriptor will be
- *                         allocated on the heap.
- * \param[in]    fmt       Format for the image
- * \param[in]    d_w       Width of the image
- * \param[in]    d_h       Height of the image
- * \param[in]    align     Alignment, in bytes, of each row in the image.
- * \param[in]    img_data  Storage to use for the image
- *
- * \return Returns a pointer to the initialized image descriptor. If the img
- *         parameter is non-null, the value of the img parameter will be
- *         returned.
- */
-aom_image_t *aom_img_wrap(aom_image_t *img, aom_img_fmt_t fmt, unsigned int d_w,
-                          unsigned int d_h, unsigned int align,
-                          unsigned char *img_data);
-
-/*!\brief Set the rectangle identifying the displayed portion of the image
- *
- * Updates the displayed rectangle (aka viewport) on the image surface to
- * match the specified coordinates and size.
- *
- * \param[in]    img       Image descriptor
- * \param[in]    x         leftmost column
- * \param[in]    y         topmost row
- * \param[in]    w         width
- * \param[in]    h         height
- *
- * \return 0 if the requested rectangle is valid, nonzero otherwise.
- */
-int aom_img_set_rect(aom_image_t *img, unsigned int x, unsigned int y,
-                     unsigned int w, unsigned int h);
-
-/*!\brief Flip the image vertically (top for bottom)
- *
- * Adjusts the image descriptor's pointers and strides to make the image
- * be referenced upside-down.
- *
- * \param[in]    img       Image descriptor
- */
-void aom_img_flip(aom_image_t *img);
-
-/*!\brief Close an image descriptor
- *
- * Frees all allocated storage associated with an image descriptor.
- *
- * \param[in]    img       Image descriptor
- */
-void aom_img_free(aom_image_t *img);
-
-#ifdef __cplusplus
-}  // extern "C"
-#endif
-
-#endif  // AOM_AOM_IMAGE_H_

aom/aom_integer.h

@@ -1,64 +0,0 @@
-/*
- * Copyright (c) 2016, Alliance for Open Media. All rights reserved
- *
- * This source code is subject to the terms of the BSD 2 Clause License and
- * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
- * was not distributed with this source code in the LICENSE file, you can
- * obtain it at www.aomedia.org/license/software. If the Alliance for Open
- * Media Patent License 1.0 was not distributed with this source code in the
- * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
- */
-
-#ifndef AOM_AOM_INTEGER_H_
-#define AOM_AOM_INTEGER_H_
-
-/* get ptrdiff_t, size_t, wchar_t, NULL */
-#include <stddef.h>
-
-#if defined(_MSC_VER)
-#define AOM_FORCE_INLINE __forceinline
-#define AOM_INLINE __inline
-#else
-#define AOM_FORCE_INLINE __inline__ __attribute__((always_inline))
-// TODO(jbb): Allow a way to force inline off for older compilers.
-#define AOM_INLINE inline
-#endif
-
-#if defined(AOM_EMULATE_INTTYPES)
-typedef signed char int8_t;
-typedef signed short int16_t;
-typedef signed int int32_t;
-
-typedef unsigned char uint8_t;
-typedef unsigned short uint16_t;
-typedef unsigned int uint32_t;
-
-#ifndef _UINTPTR_T_DEFINED
-typedef size_t uintptr_t;
-#endif
-
-#else
-
-/* Most platforms have the C99 standard integer types. */
-
-#if defined(__cplusplus)
-#if !defined(__STDC_FORMAT_MACROS)
-#define __STDC_FORMAT_MACROS
-#endif
-#if !defined(__STDC_LIMIT_MACROS)
-#define __STDC_LIMIT_MACROS
-#endif
-#endif  // __cplusplus
-
-#include <stdint.h>
-
-#endif
-
-/* VS2010 defines stdint.h, but not inttypes.h */
-#if defined(_MSC_VER) && _MSC_VER < 1800
-#define PRId64 "I64d"
-#else
-#include <inttypes.h>
-#endif
-
-#endif  // AOM_AOM_INTEGER_H_

aom/aomcx.h

@@ -1,759 +0,0 @@
-/*
- * Copyright (c) 2016, Alliance for Open Media. All rights reserved
- *
- * This source code is subject to the terms of the BSD 2 Clause License and
- * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
- * was not distributed with this source code in the LICENSE file, you can
- * obtain it at www.aomedia.org/license/software. If the Alliance for Open
- * Media Patent License 1.0 was not distributed with this source code in the
- * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
- */
-#ifndef AOM_AOMCX_H_
-#define AOM_AOMCX_H_
-
-/*!\defgroup aom_encoder AOMedia AOM/AV1 Encoder
- * \ingroup aom
- *
- * @{
- */
-#include "./aom.h"
-#include "./aom_encoder.h"
-
-/*!\file
- * \brief Provides definitions for using AOM or AV1 encoder algorithm within the
- *        aom Codec Interface.
- */
-
-#ifdef __cplusplus
-extern "C" {
-#endif
-
-/*!\name Algorithm interface for AV1
- *
- * This interface provides the capability to encode raw AV1 streams.
- * @{
- */
-extern aom_codec_iface_t aom_codec_av1_cx_algo;
-extern aom_codec_iface_t *aom_codec_av1_cx(void);
-/*!@} - end algorithm interface member group*/
-
-/*
- * Algorithm Flags
- */
-
-/*!\brief Don't reference the last frame
- *
- * When this flag is set, the encoder will not use the last frame as a
- * predictor. When not set, the encoder will choose whether to use the
- * last frame or not automatically.
- */
-#define AOM_EFLAG_NO_REF_LAST (1 << 16)
-
-/*!\brief Don't reference the golden frame
- *
- * When this flag is set, the encoder will not use the golden frame as a
- * predictor. When not set, the encoder will choose whether to use the
- * golden frame or not automatically.
- */
-#define AOM_EFLAG_NO_REF_GF (1 << 17)
-
-/*!\brief Don't reference the alternate reference frame
- *
- * When this flag is set, the encoder will not use the alt ref frame as a
- * predictor. When not set, the encoder will choose whether to use the
- * alt ref frame or not automatically.
- */
-#define AOM_EFLAG_NO_REF_ARF (1 << 21)
-
-/*!\brief Don't update the last frame
- *
- * When this flag is set, the encoder will not update the last frame with
- * the contents of the current frame.
- */
-#define AOM_EFLAG_NO_UPD_LAST (1 << 18)
-
-/*!\brief Don't update the golden frame
- *
- * When this flag is set, the encoder will not update the golden frame with
- * the contents of the current frame.
- */
-#define AOM_EFLAG_NO_UPD_GF (1 << 22)
-
-/*!\brief Don't update the alternate reference frame
- *
- * When this flag is set, the encoder will not update the alt ref frame with
- * the contents of the current frame.
- */
-#define AOM_EFLAG_NO_UPD_ARF (1 << 23)
-
-/*!\brief Force golden frame update
- *
- * When this flag is set, the encoder copy the contents of the current frame
- * to the golden frame buffer.
- */
-#define AOM_EFLAG_FORCE_GF (1 << 19)
-
-/*!\brief Force alternate reference frame update
- *
- * When this flag is set, the encoder copy the contents of the current frame
- * to the alternate reference frame buffer.
- */
-#define AOM_EFLAG_FORCE_ARF (1 << 24)
-
-/*!\brief Disable entropy update
- *
- * When this flag is set, the encoder will not update its internal entropy
- * model based on the entropy of this frame.
- */
-#define AOM_EFLAG_NO_UPD_ENTROPY (1 << 20)
-
-/*!\brief AVx encoder control functions
- *
- * This set of macros define the control functions available for AVx
- * encoder interface.
- *
- * \sa #aom_codec_control
- */
-enum aome_enc_control_id {
-  /*!\brief Codec control function to set which reference frame encoder can use.
-   *
-   * Supported in codecs: VP8, AV1
-   */
-  AOME_USE_REFERENCE = 7,
-
-  /*!\brief Codec control function to pass an ROI map to encoder.
-   *
-   * Supported in codecs: VP8, AV1
-   */
-  AOME_SET_ROI_MAP = 8,
-
-  /*!\brief Codec control function to pass an Active map to encoder.
-   *
-   * Supported in codecs: VP8, AV1
-   */
-  AOME_SET_ACTIVEMAP,
-
-  /*!\brief Codec control function to set encoder scaling mode.
-   *
-   * Supported in codecs: VP8, AV1
-   */
-  AOME_SET_SCALEMODE = 11,
-
-  /*!\brief Codec control function to set encoder internal speed settings.
-   *
-   * Changes in this value influences, among others, the encoder's selection
-   * of motion estimation methods. Values greater than 0 will increase encoder
-   * speed at the expense of quality.
-   *
-   * \note Valid range for VP8: -16..16
-   * \note Valid range for AV1: -8..8
-   *
-   * Supported in codecs: VP8, AV1
-   */
-  AOME_SET_CPUUSED = 13,
-
-  /*!\brief Codec control function to enable automatic set and use alf frames.
-   *
-   * Supported in codecs: VP8, AV1
-   */
-  AOME_SET_ENABLEAUTOALTREF,
-
-#if CONFIG_EXT_REFS
-  /*!\brief Codec control function to enable automatic set and use
-   * bwd-pred frames.
-   *
-   * Supported in codecs: AV1
-   */
-  AOME_SET_ENABLEAUTOBWDREF,
-#endif  // CONFIG_EXT_REFS
-
-  /*!\brief control function to set noise sensitivity
-   *
-   * 0: off, 1: OnYOnly, 2: OnYUV,
-   * 3: OnYUVAggressive, 4: Adaptive
-   *
-   * Supported in codecs: VP8
-   */
-  AOME_SET_NOISE_SENSITIVITY,
-
-  /*!\brief Codec control function to set sharpness.
-   *
-   * Supported in codecs: VP8, AV1
-   */
-  AOME_SET_SHARPNESS,
-
-  /*!\brief Codec control function to set the threshold for MBs treated static.
-   *
-   * Supported in codecs: VP8, AV1
-   */
-  AOME_SET_STATIC_THRESHOLD,
-
-  /*!\brief Codec control function to set the number of token partitions.
-   *
-   * Supported in codecs: VP8
-   */
-  AOME_SET_TOKEN_PARTITIONS,
-
-  /*!\brief Codec control function to get last quantizer chosen by the encoder.
-   *
-   * Return value uses internal quantizer scale defined by the codec.
-   *
-   * Supported in codecs: VP8, AV1
-   */
-  AOME_GET_LAST_QUANTIZER,
-
-  /*!\brief Codec control function to get last quantizer chosen by the encoder.
-   *
-   * Return value uses the 0..63 scale as used by the rc_*_quantizer config
-   * parameters.
-   *
-   * Supported in codecs: VP8, AV1
-   */
-  AOME_GET_LAST_QUANTIZER_64,
-
-  /*!\brief Codec control function to set the max no of frames to create arf.
-   *
-   * Supported in codecs: VP8, AV1
-   */
-  AOME_SET_ARNR_MAXFRAMES,
-
-  /*!\brief Codec control function to set the filter strength for the arf.
-   *
-   * Supported in codecs: VP8, AV1
-   */
-  AOME_SET_ARNR_STRENGTH,
-
-  /*!\deprecated control function to set the filter type to use for the arf. */
-  AOME_SET_ARNR_TYPE,
-
-  /*!\brief Codec control function to set visual tuning.
-   *
-   * Supported in codecs: VP8, AV1
-   */
-  AOME_SET_TUNING,
-
-  /*!\brief Codec control function to set constrained quality level.
-   *
-   * \attention For this value to be used aom_codec_enc_cfg_t::g_usage must be
-   *            set to #AOM_CQ.
-   * \note Valid range: 0..63
-   *
-   * Supported in codecs: VP8, AV1
-   */
-  AOME_SET_CQ_LEVEL,
-
-  /*!\brief Codec control function to set Max data rate for Intra frames.
-   *
-   * This value controls additional clamping on the maximum size of a
-   * keyframe. It is expressed as a percentage of the average
-   * per-frame bitrate, with the special (and default) value 0 meaning
-   * unlimited, or no additional clamping beyond the codec's built-in
-   * algorithm.
-   *
-   * For example, to allocate no more than 4.5 frames worth of bitrate
-   * to a keyframe, set this to 450.
-   *
-   * Supported in codecs: VP8, AV1
-   */
-  AOME_SET_MAX_INTRA_BITRATE_PCT,
-
-  /*!\brief Codec control function to set reference and update frame flags.
-   *
-   *  Supported in codecs: VP8
-   */
-  AOME_SET_FRAME_FLAGS,
-
-  /*!\brief Codec control function to set max data rate for Inter frames.
-   *
-   * This value controls additional clamping on the maximum size of an
-   * inter frame. It is expressed as a percentage of the average
-   * per-frame bitrate, with the special (and default) value 0 meaning
-   * unlimited, or no additional clamping beyond the codec's built-in
-   * algorithm.
-   *
-   * For example, to allow no more than 4.5 frames worth of bitrate
-   * to an inter frame, set this to 450.
-   *
-   * Supported in codecs: AV1
-   */
-  AV1E_SET_MAX_INTER_BITRATE_PCT,
-
-  /*!\brief Boost percentage for Golden Frame in CBR mode.
-   *
-   * This value controls the amount of boost given to Golden Frame in
-   * CBR mode. It is expressed as a percentage of the average
-   * per-frame bitrate, with the special (and default) value 0 meaning
-   * the feature is off, i.e., no golden frame boost in CBR mode and
-   * average bitrate target is used.
-   *
-   * For example, to allow 100% more bits, i.e, 2X, in a golden frame
-   * than average frame, set this to 100.
-   *
-   * Supported in codecs: AV1
-   */
-  AV1E_SET_GF_CBR_BOOST_PCT,
-
-  /*!\brief Codec control function to set encoder screen content mode.
-   *
-   * 0: off, 1: On, 2: On with more aggressive rate control.
-   *
-   * Supported in codecs: VP8
-   */
-  AOME_SET_SCREEN_CONTENT_MODE,
-
-  /*!\brief Codec control function to set lossless encoding mode.
-   *
-   * AV1 can operate in lossless encoding mode, in which the bitstream
-   * produced will be able to decode and reconstruct a perfect copy of
-   * input source. This control function provides a mean to switch encoder
-   * into lossless coding mode(1) or normal coding mode(0) that may be lossy.
-   *                          0 = lossy coding mode
-   *                          1 = lossless coding mode
-   *
-   *  By default, encoder operates in normal coding mode (maybe lossy).
-   *
-   * Supported in codecs: AV1
-   */
-  AV1E_SET_LOSSLESS,
-#if CONFIG_AOM_QM
-  /*!\brief Codec control function to encode with quantisation matrices.
-   *
-   * AOM can operate with default quantisation matrices dependent on
-   * quantisation level and block type.
-   *                          0 = do not use quantisation matrices
-   *                          1 = use quantisation matrices
-   *
-   *  By default, the encoder operates without quantisation matrices.
-   *
-   * Supported in codecs: AOM
-   */
-
-  AV1E_SET_ENABLE_QM,
-
-  /*!\brief Codec control function to set the min quant matrix flatness.
-   *
-   * AOM can operate with different ranges of quantisation matrices.
-   * As quantisation levels increase, the matrices get flatter. This
-   * control sets the minimum level of flatness from which the matrices
-   * are determined.
-   *
-   *  By default, the encoder sets this minimum at half the available
-   *  range.
-   *
-   * Supported in codecs: AOM
-   */
-  AV1E_SET_QM_MIN,
-
-  /*!\brief Codec control function to set the max quant matrix flatness.
-   *
-   * AOM can operate with different ranges of quantisation matrices.
-   * As quantisation levels increase, the matrices get flatter. This
-   * control sets the maximum level of flatness possible.
-   *
-   * By default, the encoder sets this maximum at the top of the
-   * available range.
-   *
-   * Supported in codecs: AOM
-   */
-  AV1E_SET_QM_MAX,
-#endif
-
-  /*!\brief Codec control function to set number of tile columns.
-   *
-   * In encoding and decoding, AV1 allows an input image frame be partitioned
-   * into separated vertical tile columns, which can be encoded or decoded
-   * independently. This enables easy implementation of parallel encoding and
-   * decoding. This control requests the encoder to use column tiles in
-   * encoding an input frame, with number of tile columns (in Log2 unit) as
-   * the parameter:
-   *             0 = 1 tile column
-   *             1 = 2 tile columns
-   *             2 = 4 tile columns
-   *             .....
-   *             n = 2**n tile columns
-   * The requested tile columns will be capped by encoder based on image size
-   * limitation (The minimum width of a tile column is 256 pixel, the maximum
-   * is 4096).
-   *
-   * By default, the value is 0, i.e. one single column tile for entire image.
-   *
-   * Supported in codecs: AV1
-   */
-  AV1E_SET_TILE_COLUMNS,
-
-  /*!\brief Codec control function to set number of tile rows.
-   *
-   * In encoding and decoding, AV1 allows an input image frame be partitioned
-   * into separated horizontal tile rows. Tile rows are encoded or decoded
-   * sequentially. Even though encoding/decoding of later tile rows depends on
-   * earlier ones, this allows the encoder to output data packets for tile rows
-   * prior to completely processing all tile rows in a frame, thereby reducing
-   * the latency in processing between input and output. The parameter
-   * for this control describes the number of tile rows, which has a valid
-   * range [0, 2]:
-   *            0 = 1 tile row
-   *            1 = 2 tile rows
-   *            2 = 4 tile rows
-   *
-   * By default, the value is 0, i.e. one single row tile for entire image.
-   *
-   * Supported in codecs: AV1
-   */
-  AV1E_SET_TILE_ROWS,
-
-  /*!\brief Codec control function to enable frame parallel decoding feature.
-   *
-   * AV1 has a bitstream feature to reduce decoding dependency between frames
-   * by turning off backward update of probability context used in encoding
-   * and decoding. This allows staged parallel processing of more than one
-   * video frames in the decoder. This control function provides a mean to
-   * turn this feature on or off for bitstreams produced by encoder.
-   *
-   * By default, this feature is off.
-   *
-   * Supported in codecs: AV1
-   */
-  AV1E_SET_FRAME_PARALLEL_DECODING,
-
-  /*!\brief Codec control function to set adaptive quantization mode.
-   *
-   * AV1 has a segment based feature that allows encoder to adaptively change
-   * quantization parameter for each segment within a frame to improve the
-   * subjective quality. This control makes encoder operate in one of the
-   * several AQ_modes supported.
-   *
-   * By default, encoder operates with AQ_Mode 0(adaptive quantization off).
-   *
-   * Supported in codecs: AV1
-   */
-  AV1E_SET_AQ_MODE,
-
-  /*!\brief Codec control function to enable/disable periodic Q boost.
-   *
-   * One AV1 encoder speed feature is to enable quality boost by lowering
-   * frame level Q periodically. This control function provides a mean to
-   * turn on/off this feature.
-   *               0 = off
-   *               1 = on
-   *
-   * By default, the encoder is allowed to use this feature for appropriate
-   * encoding modes.
-   *
-   * Supported in codecs: AV1
-   */
-  AV1E_SET_FRAME_PERIODIC_BOOST,
-
-  /*!\brief Codec control function to set noise sensitivity.
-   *
-   *  0: off, 1: On(YOnly)
-   *
-   * Supported in codecs: AV1
-   */
-  AV1E_SET_NOISE_SENSITIVITY,
-
-  /*!\brief Codec control function to set content type.
-   * \note Valid parameter range:
-   *              AOM_CONTENT_DEFAULT = Regular video content (Default)
-   *              AOM_CONTENT_SCREEN  = Screen capture content
-   *
-   * Supported in codecs: AV1
-   */
-  AV1E_SET_TUNE_CONTENT,
-
-  /*!\brief Codec control function to set color space info.
-   * \note Valid ranges: 0..7, default is "UNKNOWN".
-   *                     0 = UNKNOWN,
-   *                     1 = BT_601
-   *                     2 = BT_709
-   *                     3 = SMPTE_170
-   *                     4 = SMPTE_240
-   *                     5 = BT_2020
-   *                     6 = RESERVED
-   *                     7 = SRGB
-   *
-   * Supported in codecs: AV1
-   */
-  AV1E_SET_COLOR_SPACE,
-
-  /*!\brief Codec control function to set minimum interval between GF/ARF frames
-   *
-   * By default the value is set as 4.
-   *
-   * Supported in codecs: AV1
-   */
-  AV1E_SET_MIN_GF_INTERVAL,
-
-  /*!\brief Codec control function to set minimum interval between GF/ARF frames
-   *
-   * By default the value is set as 16.
-   *
-   * Supported in codecs: AV1
-   */
-  AV1E_SET_MAX_GF_INTERVAL,
-
-  /*!\brief Codec control function to get an Active map back from the encoder.
-   *
-   * Supported in codecs: AV1
-   */
-  AV1E_GET_ACTIVEMAP,
-
-  /*!\brief Codec control function to set color range bit.
-   * \note Valid ranges: 0..1, default is 0
-   *                     0 = Limited range (16..235 or HBD equivalent)
-   *                     1 = Full range (0..255 or HBD equivalent)
-   *
-   * Supported in codecs: AV1
-   */
-  AV1E_SET_COLOR_RANGE,
-
-  /*!\brief Codec control function to set intended rendering image size.
-   *
-   * By default, this is identical to the image size in pixels.
-   *
-   * Supported in codecs: AV1
-   */
-  AV1E_SET_RENDER_SIZE,
-
-  /*!\brief Codec control function to set target level.
-   *
-   * 255: off (default); 0: only keep level stats; 10: target for level 1.0;
-   * 11: target for level 1.1; ... 62: target for level 6.2
-   *
-   * Supported in codecs: AV1
-   */
-  AV1E_SET_TARGET_LEVEL,
-
-  /*!\brief Codec control function to get bitstream level.
-   *
-   * Supported in codecs: AV1
-   */
-  AV1E_GET_LEVEL,
-
-  /*!\brief Codec control function to set intended superblock size.
-   *
-   * By default, the superblock size is determined separately for each
-   * frame by the encoder.
-   *
-   * Supported in codecs: AV1
-   */
-  AV1E_SET_SUPERBLOCK_SIZE,
-};
-
-/*!\brief aom 1-D scaling mode
- *
- * This set of constants define 1-D aom scaling modes
- */
-typedef enum aom_scaling_mode_1d {
-  AOME_NORMAL = 0,
-  AOME_FOURFIVE = 1,
-  AOME_THREEFIVE = 2,
-  AOME_ONETWO = 3
-} AOM_SCALING_MODE;
-
-/*!\brief  aom region of interest map
- *
- * These defines the data structures for the region of interest map
- *
- */
-
-typedef struct aom_roi_map {
-  /*! An id between 0 and 3 for each 16x16 region within a frame. */
-  unsigned char *roi_map;
-  unsigned int rows; /**< Number of rows. */
-  unsigned int cols; /**< Number of columns. */
-  // TODO(paulwilkins): broken for AV1 which has 8 segments
-  // q and loop filter deltas for each segment
-  // (see MAX_MB_SEGMENTS)
-  int delta_q[4];  /**< Quantizer deltas. */
-  int delta_lf[4]; /**< Loop filter deltas. */
-  /*! Static breakout threshold for each segment. */
-  unsigned int static_threshold[4];
-} aom_roi_map_t;
-
-/*!\brief  aom active region map
- *
- * These defines the data structures for active region map
- *
- */
-
-typedef struct aom_active_map {
-  /*!\brief specify an on (1) or off (0) each 16x16 region within a frame */
-  unsigned char *active_map;
-  unsigned int rows; /**< number of rows */
-  unsigned int cols; /**< number of cols */
-} aom_active_map_t;
-
-/*!\brief  aom image scaling mode
- *
- * This defines the data structure for image scaling mode
- *
- */
-typedef struct aom_scaling_mode {
-  AOM_SCALING_MODE h_scaling_mode; /**< horizontal scaling mode */
-  AOM_SCALING_MODE v_scaling_mode; /**< vertical scaling mode   */
-} aom_scaling_mode_t;
-
-/*!\brief VP8 token partition mode
- *
- * This defines VP8 partitioning mode for compressed data, i.e., the number of
- * sub-streams in the bitstream. Used for parallelized decoding.
- *
- */
-
-typedef enum {
-  AOM_ONE_TOKENPARTITION = 0,
-  AOM_TWO_TOKENPARTITION = 1,
-  AOM_FOUR_TOKENPARTITION = 2,
-  AOM_EIGHT_TOKENPARTITION = 3
-} aome_token_partitions;
-
-/*!brief AV1 encoder content type */
-typedef enum {
-  AOM_CONTENT_DEFAULT,
-  AOM_CONTENT_SCREEN,
-  AOM_CONTENT_INVALID
-} aom_tune_content;
-
-/*!\brief VP8 model tuning parameters
- *
- * Changes the encoder to tune for certain types of input material.
- *
- */
-typedef enum { AOM_TUNE_PSNR, AOM_TUNE_SSIM } aom_tune_metric;
-
-/*!\cond */
-/*!\brief VP8 encoder control function parameter type
- *
- * Defines the data types that VP8E control functions take. Note that
- * additional common controls are defined in aom.h
- *
- */
-
-AOM_CTRL_USE_TYPE_DEPRECATED(AOME_USE_REFERENCE, int)
-#define AOM_CTRL_AOME_USE_REFERENCE
-AOM_CTRL_USE_TYPE(AOME_SET_FRAME_FLAGS, int)
-#define AOM_CTRL_AOME_SET_FRAME_FLAGS
-AOM_CTRL_USE_TYPE(AOME_SET_ROI_MAP, aom_roi_map_t *)
-#define AOM_CTRL_AOME_SET_ROI_MAP
-AOM_CTRL_USE_TYPE(AOME_SET_ACTIVEMAP, aom_active_map_t *)
-#define AOM_CTRL_AOME_SET_ACTIVEMAP
-AOM_CTRL_USE_TYPE(AOME_SET_SCALEMODE, aom_scaling_mode_t *)
-#define AOM_CTRL_AOME_SET_SCALEMODE
-
-AOM_CTRL_USE_TYPE(AOME_SET_CPUUSED, int)
-#define AOM_CTRL_AOME_SET_CPUUSED
-AOM_CTRL_USE_TYPE(AOME_SET_ENABLEAUTOALTREF, unsigned int)
-#define AOM_CTRL_AOME_SET_ENABLEAUTOALTREF
-
-#if CONFIG_EXT_REFS
-AOM_CTRL_USE_TYPE(AOME_SET_ENABLEAUTOBWDREF, unsigned int)
-#define AOM_CTRL_AOME_SET_ENABLEAUTOBWDREF
-#endif  // CONFIG_EXT_REFS
-
-AOM_CTRL_USE_TYPE(AOME_SET_NOISE_SENSITIVITY, unsigned int)
-#define AOM_CTRL_AOME_SET_NOISE_SENSITIVITY
-AOM_CTRL_USE_TYPE(AOME_SET_SHARPNESS, unsigned int)
-#define AOM_CTRL_AOME_SET_SHARPNESS
-AOM_CTRL_USE_TYPE(AOME_SET_STATIC_THRESHOLD, unsigned int)
-#define AOM_CTRL_AOME_SET_STATIC_THRESHOLD
-AOM_CTRL_USE_TYPE(AOME_SET_TOKEN_PARTITIONS, int) /* aome_token_partitions */
-#define AOM_CTRL_AOME_SET_TOKEN_PARTITIONS
-
-AOM_CTRL_USE_TYPE(AOME_SET_ARNR_MAXFRAMES, unsigned int)
-#define AOM_CTRL_AOME_SET_ARNR_MAXFRAMES
-AOM_CTRL_USE_TYPE(AOME_SET_ARNR_STRENGTH, unsigned int)
-#define AOM_CTRL_AOME_SET_ARNR_STRENGTH
-AOM_CTRL_USE_TYPE_DEPRECATED(AOME_SET_ARNR_TYPE, unsigned int)
-#define AOM_CTRL_AOME_SET_ARNR_TYPE
-AOM_CTRL_USE_TYPE(AOME_SET_TUNING, int) /* aom_tune_metric */
-#define AOM_CTRL_AOME_SET_TUNING
-AOM_CTRL_USE_TYPE(AOME_SET_CQ_LEVEL, unsigned int)
-#define AOM_CTRL_AOME_SET_CQ_LEVEL
-
-AOM_CTRL_USE_TYPE(AV1E_SET_TILE_COLUMNS, int)
-#define AOM_CTRL_AV1E_SET_TILE_COLUMNS
-AOM_CTRL_USE_TYPE(AV1E_SET_TILE_ROWS, int)
-#define AOM_CTRL_AV1E_SET_TILE_ROWS
-
-AOM_CTRL_USE_TYPE(AOME_GET_LAST_QUANTIZER, int *)
-#define AOM_CTRL_AOME_GET_LAST_QUANTIZER
-AOM_CTRL_USE_TYPE(AOME_GET_LAST_QUANTIZER_64, int *)
-#define AOM_CTRL_AOME_GET_LAST_QUANTIZER_64
-
-AOM_CTRL_USE_TYPE(AOME_SET_MAX_INTRA_BITRATE_PCT, unsigned int)
-#define AOM_CTRL_AOME_SET_MAX_INTRA_BITRATE_PCT
-AOM_CTRL_USE_TYPE(AOME_SET_MAX_INTER_BITRATE_PCT, unsigned int)
-#define AOM_CTRL_AOME_SET_MAX_INTER_BITRATE_PCT
-
-AOM_CTRL_USE_TYPE(AOME_SET_SCREEN_CONTENT_MODE, unsigned int)
-#define AOM_CTRL_AOME_SET_SCREEN_CONTENT_MODE
-
-AOM_CTRL_USE_TYPE(AV1E_SET_GF_CBR_BOOST_PCT, unsigned int)
-#define AOM_CTRL_AV1E_SET_GF_CBR_BOOST_PCT
-
-AOM_CTRL_USE_TYPE(AV1E_SET_LOSSLESS, unsigned int)
-#define AOM_CTRL_AV1E_SET_LOSSLESS
-
-#if CONFIG_AOM_QM
-AOM_CTRL_USE_TYPE(AV1E_SET_ENABLE_QM, unsigned int)
-#define AOM_CTRL_AV1E_SET_ENABLE_QM
-
-AOM_CTRL_USE_TYPE(AV1E_SET_QM_MIN, unsigned int)
-#define AOM_CTRL_AV1E_SET_QM_MIN
-
-AOM_CTRL_USE_TYPE(AV1E_SET_QM_MAX, unsigned int)
-#define AOM_CTRL_AV1E_SET_QM_MAX
-#endif
-
-AOM_CTRL_USE_TYPE(AV1E_SET_FRAME_PARALLEL_DECODING, unsigned int)
-#define AOM_CTRL_AV1E_SET_FRAME_PARALLEL_DECODING
-
-AOM_CTRL_USE_TYPE(AV1E_SET_AQ_MODE, unsigned int)
-#define AOM_CTRL_AV1E_SET_AQ_MODE
-
-AOM_CTRL_USE_TYPE(AV1E_SET_FRAME_PERIODIC_BOOST, unsigned int)
-#define AOM_CTRL_AV1E_SET_FRAME_PERIODIC_BOOST
-
-AOM_CTRL_USE_TYPE(AV1E_SET_NOISE_SENSITIVITY, unsigned int)
-#define AOM_CTRL_AV1E_SET_NOISE_SENSITIVITY
-
-AOM_CTRL_USE_TYPE(AV1E_SET_TUNE_CONTENT, int) /* aom_tune_content */
-#define AOM_CTRL_AV1E_SET_TUNE_CONTENT
-
-AOM_CTRL_USE_TYPE(AV1E_SET_COLOR_SPACE, int)
-#define AOM_CTRL_AV1E_SET_COLOR_SPACE
-
-AOM_CTRL_USE_TYPE(AV1E_SET_MIN_GF_INTERVAL, unsigned int)
-#define AOM_CTRL_AV1E_SET_MIN_GF_INTERVAL
-
-AOM_CTRL_USE_TYPE(AV1E_SET_MAX_GF_INTERVAL, unsigned int)
-#define AOM_CTRL_AV1E_SET_MAX_GF_INTERVAL
-
-AOM_CTRL_USE_TYPE(AV1E_GET_ACTIVEMAP, aom_active_map_t *)
-#define AOM_CTRL_AV1E_GET_ACTIVEMAP
-
-AOM_CTRL_USE_TYPE(AV1E_SET_COLOR_RANGE, int)
-#define AOM_CTRL_AV1E_SET_COLOR_RANGE
-
-/*!\brief
- *
- * TODO(rbultje) : add support of the control in ffmpeg
- */
-#define AOM_CTRL_AV1E_SET_RENDER_SIZE
-AOM_CTRL_USE_TYPE(AV1E_SET_RENDER_SIZE, int *)
-
-AOM_CTRL_USE_TYPE(AV1E_SET_SUPERBLOCK_SIZE, unsigned int)
-#define AOM_CTRL_AV1E_SET_SUPERBLOCK_SIZE
-
-AOM_CTRL_USE_TYPE(AV1E_SET_TARGET_LEVEL, unsigned int)
-#define AOM_CTRL_AV1E_SET_TARGET_LEVEL
-
-AOM_CTRL_USE_TYPE(AV1E_GET_LEVEL, int *)
-#define AOM_CTRL_AV1E_GET_LEVEL
-/*!\endcond */
-/*! @} - end defgroup vp8_encoder */
-#ifdef __cplusplus
-}  // extern "C"
-#endif
-
-#endif  // AOM_AOMCX_H_

aom/aomdx.h

@@ -1,191 +0,0 @@
-/*
- * Copyright (c) 2016, Alliance for Open Media. All rights reserved
- *
- * This source code is subject to the terms of the BSD 2 Clause License and
- * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
- * was not distributed with this source code in the LICENSE file, you can
- * obtain it at www.aomedia.org/license/software. If the Alliance for Open
- * Media Patent License 1.0 was not distributed with this source code in the
- * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
- */
-
-/*!\defgroup aom_decoder AOMedia AOM/AV1 Decoder
- * \ingroup aom
- *
- * @{
- */
-/*!\file
- * \brief Provides definitions for using AOM or AV1 within the aom Decoder
- *        interface.
- */
-#ifndef AOM_AOMDX_H_
-#define AOM_AOMDX_H_
-
-#ifdef __cplusplus
-extern "C" {
-#endif
-
-/* Include controls common to both the encoder and decoder */
-#include "./aom.h"
-
-/*!\name Algorithm interface for AV1
- *
- * This interface provides the capability to decode AV1 streams.
- * @{
- */
-extern aom_codec_iface_t aom_codec_av1_dx_algo;
-extern aom_codec_iface_t *aom_codec_av1_dx(void);
-/*!@} - end algorithm interface member group*/
-
-/** Data structure that stores bit accounting for debug
- */
-typedef struct Accounting Accounting;
-
-/*!\enum aom_dec_control_id
- * \brief AOM decoder control functions
- *
- * This set of macros define the control functions available for the AOM
- * decoder interface.
- *
- * \sa #aom_codec_control
- */
-enum aom_dec_control_id {
-  /** control function to get info on which reference frames were updated
-   *  by the last decode
-   */
-  AOMD_GET_LAST_REF_UPDATES = AOM_DECODER_CTRL_ID_START,
-
-  /** check if the indicated frame is corrupted */
-  AOMD_GET_FRAME_CORRUPTED,
-
-  /** control function to get info on which reference frames were used
-   *  by the last decode
-   */
-  AOMD_GET_LAST_REF_USED,
-
-  /** decryption function to decrypt encoded buffer data immediately
-   * before decoding. Takes a aom_decrypt_init, which contains
-   * a callback function and opaque context pointer.
-   */
-  AOMD_SET_DECRYPTOR,
-  // AOMD_SET_DECRYPTOR = AOMD_SET_DECRYPTOR,
-
-  /** control function to get the dimensions that the current frame is decoded
-   * at. This may be different to the intended display size for the frame as
-   * specified in the wrapper or frame header (see AV1D_GET_DISPLAY_SIZE). */
-  AV1D_GET_FRAME_SIZE,
-
-  /** control function to get the current frame's intended display dimensions
-   * (as specified in the wrapper or frame header). This may be different to
-   * the decoded dimensions of this frame (see AV1D_GET_FRAME_SIZE). */
-  AV1D_GET_DISPLAY_SIZE,
-
-  /** control function to get the bit depth of the stream. */
-  AV1D_GET_BIT_DEPTH,
-
-  /** control function to set the byte alignment of the planes in the reference
-   * buffers. Valid values are power of 2, from 32 to 1024. A value of 0 sets
-   * legacy alignment. I.e. Y plane is aligned to 32 bytes, U plane directly
-   * follows Y plane, and V plane directly follows U plane. Default value is 0.
-   */
-  AV1_SET_BYTE_ALIGNMENT,
-
-  /** control function to invert the decoding order to from right to left. The
-   * function is used in a test to confirm the decoding independence of tile
-   * columns. The function may be used in application where this order
-   * of decoding is desired.
-   *
-   * TODO(yaowu): Rework the unit test that uses this control, and in a future
-   *              release, this test-only control shall be removed.
-   */
-  AV1_INVERT_TILE_DECODE_ORDER,
-
-  /** control function to set the skip loop filter flag. Valid values are
-   * integers. The decoder will skip the loop filter when its value is set to
-   * nonzero. If the loop filter is skipped the decoder may accumulate decode
-   * artifacts. The default value is 0.
-   */
-  AV1_SET_SKIP_LOOP_FILTER,
-
-  /** control function to retrieve a pointer to the Accounting struct.  When
-   * compiled without --enable-accounting, this returns AOM_CODEC_INCAPABLE.
-   * If called before a frame has been decoded, this returns AOM_CODEC_ERROR.
-   * The caller should ensure that AOM_CODEC_OK is returned before attempting
-   * to dereference the Accounting pointer.
-   */
-  AV1_GET_ACCOUNTING,
-
-  AOM_DECODER_CTRL_ID_MAX,
-
-  /** control function to set the range of tile decoding. A value that is
-   * greater and equal to zero indicates only the specific row/column is
-   * decoded. A value that is -1 indicates the whole row/column is decoded.
-   * A special case is both values are -1 that means the whole frame is
-   * decoded.
-   */
-  AV1_SET_DECODE_TILE_ROW,
-  AV1_SET_DECODE_TILE_COL
-};
-
-/** Decrypt n bytes of data from input -> output, using the decrypt_state
- *  passed in AOMD_SET_DECRYPTOR.
- */
-typedef void (*aom_decrypt_cb)(void *decrypt_state, const unsigned char *input,
-                               unsigned char *output, int count);
-
-/*!\brief Structure to hold decryption state
- *
- * Defines a structure to hold the decryption state and access function.
- */
-typedef struct aom_decrypt_init {
-  /*! Decrypt callback. */
-  aom_decrypt_cb decrypt_cb;
-
-  /*! Decryption state. */
-  void *decrypt_state;
-} aom_decrypt_init;
-
-/*!\brief A deprecated alias for aom_decrypt_init.
- */
-typedef aom_decrypt_init aom_decrypt_init;
-
-/*!\cond */
-/*!\brief AOM decoder control function parameter type
- *
- * Defines the data types that AOMD control functions take. Note that
- * additional common controls are defined in aom.h
- *
- */
-
-AOM_CTRL_USE_TYPE(AOMD_GET_LAST_REF_UPDATES, int *)
-#define AOM_CTRL_AOMD_GET_LAST_REF_UPDATES
-AOM_CTRL_USE_TYPE(AOMD_GET_FRAME_CORRUPTED, int *)
-#define AOM_CTRL_AOMD_GET_FRAME_CORRUPTED
-AOM_CTRL_USE_TYPE(AOMD_GET_LAST_REF_USED, int *)
-#define AOM_CTRL_AOMD_GET_LAST_REF_USED
-AOM_CTRL_USE_TYPE(AOMD_SET_DECRYPTOR, aom_decrypt_init *)
-#define AOM_CTRL_AOMD_SET_DECRYPTOR
-// AOM_CTRL_USE_TYPE(AOMD_SET_DECRYPTOR, aom_decrypt_init *)
-//#define AOM_CTRL_AOMD_SET_DECRYPTOR
-AOM_CTRL_USE_TYPE(AV1D_GET_DISPLAY_SIZE, int *)
-#define AOM_CTRL_AV1D_GET_DISPLAY_SIZE
-AOM_CTRL_USE_TYPE(AV1D_GET_BIT_DEPTH, unsigned int *)
-#define AOM_CTRL_AV1D_GET_BIT_DEPTH
-AOM_CTRL_USE_TYPE(AV1D_GET_FRAME_SIZE, int *)
-#define AOM_CTRL_AV1D_GET_FRAME_SIZE
-AOM_CTRL_USE_TYPE(AV1_INVERT_TILE_DECODE_ORDER, int)
-#define AOM_CTRL_AV1_INVERT_TILE_DECODE_ORDER
-AOM_CTRL_USE_TYPE(AV1_GET_ACCOUNTING, Accounting **)
-#define AOM_CTRL_AV1_GET_ACCOUNTING
-AOM_CTRL_USE_TYPE(AV1_SET_DECODE_TILE_ROW, int)
-#define AOM_CTRL_AV1_SET_DECODE_TILE_ROW
-AOM_CTRL_USE_TYPE(AV1_SET_DECODE_TILE_COL, int)
-#define AOM_CTRL_AV1_SET_DECODE_TILE_COL
-/*!\endcond */
-/*! @} - end defgroup aom_decoder */
-
-#ifdef __cplusplus
-}  // extern "C"
-#endif
-
-#endif  // AOM_AOMDX_H_

aom/exports_com

@@ -1,16 +0,0 @@
-text aom_codec_build_config
-text aom_codec_control_
-text aom_codec_destroy
-text aom_codec_err_to_string
-text aom_codec_error
-text aom_codec_error_detail
-text aom_codec_get_caps
-text aom_codec_iface_name
-text aom_codec_version
-text aom_codec_version_extra_str
-text aom_codec_version_str
-text aom_img_alloc
-text aom_img_flip
-text aom_img_free
-text aom_img_set_rect
-text aom_img_wrap

aom/exports_dec

@@ -1,8 +0,0 @@
-text aom_codec_dec_init_ver
-text aom_codec_decode
-text aom_codec_get_frame
-text aom_codec_get_stream_info
-text aom_codec_peek_stream_info
-text aom_codec_register_put_frame_cb
-text aom_codec_register_put_slice_cb
-text aom_codec_set_frame_buffer_functions

aom/exports_enc

@@ -1,9 +0,0 @@
-text aom_codec_enc_config_default
-text aom_codec_enc_config_set
-text aom_codec_enc_init_multi_ver
-text aom_codec_enc_init_ver
-text aom_codec_encode
-text aom_codec_get_cx_data
-text aom_codec_get_global_headers
-text aom_codec_get_preview_frame
-text aom_codec_set_cx_data_buf

aom/internal/aom_codec_internal.h

@@ -1,465 +0,0 @@
-/*
- * Copyright (c) 2016, Alliance for Open Media. All rights reserved
- *
- * This source code is subject to the terms of the BSD 2 Clause License and
- * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
- * was not distributed with this source code in the LICENSE file, you can
- * obtain it at www.aomedia.org/license/software. If the Alliance for Open
- * Media Patent License 1.0 was not distributed with this source code in the
- * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
- */
-
-/*!\file
- * \brief Describes the decoder algorithm interface for algorithm
- *        implementations.
- *
- * This file defines the private structures and data types that are only
- * relevant to implementing an algorithm, as opposed to using it.
- *
- * To create a decoder algorithm class, an interface structure is put
- * into the global namespace:
- *     <pre>
- *     my_codec.c:
- *       aom_codec_iface_t my_codec = {
- *           "My Codec v1.0",
- *           AOM_CODEC_ALG_ABI_VERSION,
- *           ...
- *       };
- *     </pre>
- *
- * An application instantiates a specific decoder instance by using
- * aom_codec_init() and a pointer to the algorithm's interface structure:
- *     <pre>
- *     my_app.c:
- *       extern aom_codec_iface_t my_codec;
- *       {
- *           aom_codec_ctx_t algo;
- *           res = aom_codec_init(&algo, &my_codec);
- *       }
- *     </pre>
- *
- * Once initialized, the instance is manged using other functions from
- * the aom_codec_* family.
- */
-#ifndef AOM_INTERNAL_AOM_CODEC_INTERNAL_H_
-#define AOM_INTERNAL_AOM_CODEC_INTERNAL_H_
-#include "./aom_config.h"
-#include "../aom_decoder.h"
-#include "../aom_encoder.h"
-#include <stdarg.h>
-
-#ifdef __cplusplus
-extern "C" {
-#endif
-
-/*!\brief Current ABI version number
- *
- * \internal
- * If this file is altered in any way that changes the ABI, this value
- * must be bumped.  Examples include, but are not limited to, changing
- * types, removing or reassigning enums, adding/removing/rearranging
- * fields to structures
- */
-#define AOM_CODEC_INTERNAL_ABI_VERSION (5) /**<\hideinitializer*/
-
-typedef struct aom_codec_alg_priv aom_codec_alg_priv_t;
-typedef struct aom_codec_priv_enc_mr_cfg aom_codec_priv_enc_mr_cfg_t;
-
-/*!\brief init function pointer prototype
- *
- * Performs algorithm-specific initialization of the decoder context. This
- * function is called by the generic aom_codec_init() wrapper function, so
- * plugins implementing this interface may trust the input parameters to be
- * properly initialized.
- *
- * \param[in] ctx   Pointer to this instance's context
- * \retval #AOM_CODEC_OK
- *     The input stream was recognized and decoder initialized.
- * \retval #AOM_CODEC_MEM_ERROR
- *     Memory operation failed.
- */
-typedef aom_codec_err_t (*aom_codec_init_fn_t)(
-    aom_codec_ctx_t *ctx, aom_codec_priv_enc_mr_cfg_t *data);
-
-/*!\brief destroy function pointer prototype
- *
- * Performs algorithm-specific destruction of the decoder context. This
- * function is called by the generic aom_codec_destroy() wrapper function,
- * so plugins implementing this interface may trust the input parameters
- * to be properly initialized.
- *
- * \param[in] ctx   Pointer to this instance's context
- * \retval #AOM_CODEC_OK
- *     The input stream was recognized and decoder initialized.
- * \retval #AOM_CODEC_MEM_ERROR
- *     Memory operation failed.
- */
-typedef aom_codec_err_t (*aom_codec_destroy_fn_t)(aom_codec_alg_priv_t *ctx);
-
-/*!\brief parse stream info function pointer prototype
- *
- * Performs high level parsing of the bitstream. This function is called by the
- * generic aom_codec_peek_stream_info() wrapper function, so plugins
- * implementing this interface may trust the input parameters to be properly
- * initialized.
- *
- * \param[in]      data    Pointer to a block of data to parse
- * \param[in]      data_sz Size of the data buffer
- * \param[in,out]  si      Pointer to stream info to update. The size member
- *                         \ref MUST be properly initialized, but \ref MAY be
- *                         clobbered by the algorithm. This parameter \ref MAY
- *                         be NULL.
- *
- * \retval #AOM_CODEC_OK
- *     Bitstream is parsable and stream information updated
- */
-typedef aom_codec_err_t (*aom_codec_peek_si_fn_t)(const uint8_t *data,
-                                                  unsigned int data_sz,
-                                                  aom_codec_stream_info_t *si);
-
-/*!\brief Return information about the current stream.
- *
- * Returns information about the stream that has been parsed during decoding.
- *
- * \param[in]      ctx     Pointer to this instance's context
- * \param[in,out]  si      Pointer to stream info to update. The size member
- *                         \ref MUST be properly initialized, but \ref MAY be
- *                         clobbered by the algorithm. This parameter \ref MAY
- *                         be NULL.
- *
- * \retval #AOM_CODEC_OK
- *     Bitstream is parsable and stream information updated
- */
-typedef aom_codec_err_t (*aom_codec_get_si_fn_t)(aom_codec_alg_priv_t *ctx,
-                                                 aom_codec_stream_info_t *si);
-
-/*!\brief control function pointer prototype
- *
- * This function is used to exchange algorithm specific data with the decoder
- * instance. This can be used to implement features specific to a particular
- * algorithm.
- *
- * This function is called by the generic aom_codec_control() wrapper
- * function, so plugins implementing this interface may trust the input
- * parameters to be properly initialized. However,  this interface does not
- * provide type safety for the exchanged data or assign meanings to the
- * control codes. Those details should be specified in the algorithm's
- * header file. In particular, the ctrl_id parameter is guaranteed to exist
- * in the algorithm's control mapping table, and the data parameter may be NULL.
- *
- *
- * \param[in]     ctx              Pointer to this instance's context
- * \param[in]     ctrl_id          Algorithm specific control identifier
- * \param[in,out] data             Data to exchange with algorithm instance.
- *
- * \retval #AOM_CODEC_OK
- *     The internal state data was deserialized.
- */
-typedef aom_codec_err_t (*aom_codec_control_fn_t)(aom_codec_alg_priv_t *ctx,
-                                                  va_list ap);
-
-/*!\brief control function pointer mapping
- *
- * This structure stores the mapping between control identifiers and
- * implementing functions. Each algorithm provides a list of these
- * mappings. This list is searched by the aom_codec_control() wrapper
- * function to determine which function to invoke. The special
- * value {0, NULL} is used to indicate end-of-list, and must be
- * present. The special value {0, <non-null>} can be used as a catch-all
- * mapping. This implies that ctrl_id values chosen by the algorithm
- * \ref MUST be non-zero.
- */
-typedef const struct aom_codec_ctrl_fn_map {
-  int ctrl_id;
-  aom_codec_control_fn_t fn;
-} aom_codec_ctrl_fn_map_t;
-
-/*!\brief decode data function pointer prototype
- *
- * Processes a buffer of coded data. If the processing results in a new
- * decoded frame becoming available, #AOM_CODEC_CB_PUT_SLICE and
- * #AOM_CODEC_CB_PUT_FRAME events are generated as appropriate. This
- * function is called by the generic aom_codec_decode() wrapper function,
- * so plugins implementing this interface may trust the input parameters
- * to be properly initialized.
- *
- * \param[in] ctx          Pointer to this instance's context
- * \param[in] data         Pointer to this block of new coded data. If
- *                         NULL, a #AOM_CODEC_CB_PUT_FRAME event is posted
- *                         for the previously decoded frame.
- * \param[in] data_sz      Size of the coded data, in bytes.
- *
- * \return Returns #AOM_CODEC_OK if the coded data was processed completely
- *         and future pictures can be decoded without error. Otherwise,
- *         see the descriptions of the other error codes in ::aom_codec_err_t
- *         for recoverability capabilities.
- */
-typedef aom_codec_err_t (*aom_codec_decode_fn_t)(aom_codec_alg_priv_t *ctx,
-                                                 const uint8_t *data,
-                                                 unsigned int data_sz,
-                                                 void *user_priv,
-                                                 long deadline);
-
-/*!\brief Decoded frames iterator
- *
- * Iterates over a list of the frames available for display. The iterator
- * storage should be initialized to NULL to start the iteration. Iteration is
- * complete when this function returns NULL.
- *
- * The list of available frames becomes valid upon completion of the
- * aom_codec_decode call, and remains valid until the next call to
- * aom_codec_decode.
- *
- * \param[in]     ctx      Pointer to this instance's context
- * \param[in out] iter     Iterator storage, initialized to NULL
- *
- * \return Returns a pointer to an image, if one is ready for display. Frames
- *         produced will always be in PTS (presentation time stamp) order.
- */
-typedef aom_image_t *(*aom_codec_get_frame_fn_t)(aom_codec_alg_priv_t *ctx,
-                                                 aom_codec_iter_t *iter);
-
-/*!\brief Pass in external frame buffers for the decoder to use.
- *
- * Registers functions to be called when libaom needs a frame buffer
- * to decode the current frame and a function to be called when libaom does
- * not internally reference the frame buffer. This set function must
- * be called before the first call to decode or libaom will assume the
- * default behavior of allocating frame buffers internally.
- *
- * \param[in] ctx          Pointer to this instance's context
- * \param[in] cb_get       Pointer to the get callback function
- * \param[in] cb_release   Pointer to the release callback function
- * \param[in] cb_priv      Callback's private data
- *
- * \retval #AOM_CODEC_OK
- *     External frame buffers will be used by libaom.
- * \retval #AOM_CODEC_INVALID_PARAM
- *     One or more of the callbacks were NULL.
- * \retval #AOM_CODEC_ERROR
- *     Decoder context not initialized, or algorithm not capable of
- *     using external frame buffers.
- *
- * \note
- * When decoding AV1, the application may be required to pass in at least
- * #AOM_MAXIMUM_WORK_BUFFERS external frame
- * buffers.
- */
-typedef aom_codec_err_t (*aom_codec_set_fb_fn_t)(
-    aom_codec_alg_priv_t *ctx, aom_get_frame_buffer_cb_fn_t cb_get,
-    aom_release_frame_buffer_cb_fn_t cb_release, void *cb_priv);
-
-typedef aom_codec_err_t (*aom_codec_encode_fn_t)(aom_codec_alg_priv_t *ctx,
-                                                 const aom_image_t *img,
-                                                 aom_codec_pts_t pts,
-                                                 unsigned long duration,
-                                                 aom_enc_frame_flags_t flags,
-                                                 unsigned long deadline);
-typedef const aom_codec_cx_pkt_t *(*aom_codec_get_cx_data_fn_t)(
-    aom_codec_alg_priv_t *ctx, aom_codec_iter_t *iter);
-
-typedef aom_codec_err_t (*aom_codec_enc_config_set_fn_t)(
-    aom_codec_alg_priv_t *ctx, const aom_codec_enc_cfg_t *cfg);
-typedef aom_fixed_buf_t *(*aom_codec_get_global_headers_fn_t)(
-    aom_codec_alg_priv_t *ctx);
-
-typedef aom_image_t *(*aom_codec_get_preview_frame_fn_t)(
-    aom_codec_alg_priv_t *ctx);
-
-typedef aom_codec_err_t (*aom_codec_enc_mr_get_mem_loc_fn_t)(
-    const aom_codec_enc_cfg_t *cfg, void **mem_loc);
-
-/*!\brief usage configuration mapping
- *
- * This structure stores the mapping between usage identifiers and
- * configuration structures. Each algorithm provides a list of these
- * mappings. This list is searched by the aom_codec_enc_config_default()
- * wrapper function to determine which config to return. The special value
- * {-1, {0}} is used to indicate end-of-list, and must be present. At least
- * one mapping must be present, in addition to the end-of-list.
- *
- */
-typedef const struct aom_codec_enc_cfg_map {
-  int usage;
-  aom_codec_enc_cfg_t cfg;
-} aom_codec_enc_cfg_map_t;
-
-/*!\brief Decoder algorithm interface interface
- *
- * All decoders \ref MUST expose a variable of this type.
- */
-struct aom_codec_iface {
-  const char *name;                   /**< Identification String  */
-  int abi_version;                    /**< Implemented ABI version */
-  aom_codec_caps_t caps;              /**< Decoder capabilities */
-  aom_codec_init_fn_t init;           /**< \copydoc ::aom_codec_init_fn_t */
-  aom_codec_destroy_fn_t destroy;     /**< \copydoc ::aom_codec_destroy_fn_t */
-  aom_codec_ctrl_fn_map_t *ctrl_maps; /**< \copydoc ::aom_codec_ctrl_fn_map_t */
-  struct aom_codec_dec_iface {
-    aom_codec_peek_si_fn_t peek_si; /**< \copydoc ::aom_codec_peek_si_fn_t */
-    aom_codec_get_si_fn_t get_si;   /**< \copydoc ::aom_codec_get_si_fn_t */
-    aom_codec_decode_fn_t decode;   /**< \copydoc ::aom_codec_decode_fn_t */
-    aom_codec_get_frame_fn_t
-        get_frame;                   /**< \copydoc ::aom_codec_get_frame_fn_t */
-    aom_codec_set_fb_fn_t set_fb_fn; /**< \copydoc ::aom_codec_set_fb_fn_t */
-  } dec;
-  struct aom_codec_enc_iface {
-    int cfg_map_count;
-    aom_codec_enc_cfg_map_t
-        *cfg_maps;                /**< \copydoc ::aom_codec_enc_cfg_map_t */
-    aom_codec_encode_fn_t encode; /**< \copydoc ::aom_codec_encode_fn_t */
-    aom_codec_get_cx_data_fn_t
-        get_cx_data; /**< \copydoc ::aom_codec_get_cx_data_fn_t */
-    aom_codec_enc_config_set_fn_t
-        cfg_set; /**< \copydoc ::aom_codec_enc_config_set_fn_t */
-    aom_codec_get_global_headers_fn_t
-        get_glob_hdrs; /**< \copydoc ::aom_codec_get_global_headers_fn_t */
-    aom_codec_get_preview_frame_fn_t
-        get_preview; /**< \copydoc ::aom_codec_get_preview_frame_fn_t */
-    aom_codec_enc_mr_get_mem_loc_fn_t
-        mr_get_mem_loc; /**< \copydoc ::aom_codec_enc_mr_get_mem_loc_fn_t */
-  } enc;
-};
-
-/*!\brief Callback function pointer / user data pair storage */
-typedef struct aom_codec_priv_cb_pair {
-  union {
-    aom_codec_put_frame_cb_fn_t put_frame;
-    aom_codec_put_slice_cb_fn_t put_slice;
-  } u;
-  void *user_priv;
-} aom_codec_priv_cb_pair_t;
-
-/*!\brief Instance private storage
- *
- * This structure is allocated by the algorithm's init function. It can be
- * extended in one of two ways. First, a second, algorithm specific structure
- * can be allocated and the priv member pointed to it. Alternatively, this
- * structure can be made the first member of the algorithm specific structure,
- * and the pointer cast to the proper type.
- */
-struct aom_codec_priv {
-  const char *err_detail;
-  aom_codec_flags_t init_flags;
-  struct {
-    aom_codec_priv_cb_pair_t put_frame_cb;
-    aom_codec_priv_cb_pair_t put_slice_cb;
-  } dec;
-  struct {
-    aom_fixed_buf_t cx_data_dst_buf;
-    unsigned int cx_data_pad_before;
-    unsigned int cx_data_pad_after;
-    aom_codec_cx_pkt_t cx_data_pkt;
-    unsigned int total_encoders;
-  } enc;
-};
-
-/*
- * Multi-resolution encoding internal configuration
- */
-struct aom_codec_priv_enc_mr_cfg {
-  unsigned int mr_total_resolutions;
-  unsigned int mr_encoder_id;
-  struct aom_rational mr_down_sampling_factor;
-  void *mr_low_res_mode_info;
-};
-
-#undef AOM_CTRL_USE_TYPE
-#define AOM_CTRL_USE_TYPE(id, typ) \
-  static AOM_INLINE typ id##__value(va_list args) { return va_arg(args, typ); }
-
-#undef AOM_CTRL_USE_TYPE_DEPRECATED
-#define AOM_CTRL_USE_TYPE_DEPRECATED(id, typ) \
-  static AOM_INLINE typ id##__value(va_list args) { return va_arg(args, typ); }
-
-#define CAST(id, arg) id##__value(arg)
-
-/* CODEC_INTERFACE convenience macro
- *
- * By convention, each codec interface is a struct with extern linkage, where
- * the symbol is suffixed with _algo. A getter function is also defined to
- * return a pointer to the struct, since in some cases it's easier to work
- * with text symbols than data symbols (see issue #169). This function has
- * the same name as the struct, less the _algo suffix. The CODEC_INTERFACE
- * macro is provided to define this getter function automatically.
- */
-#define CODEC_INTERFACE(id)                          \
-  aom_codec_iface_t *id(void) { return &id##_algo; } \
-  aom_codec_iface_t id##_algo
-
-/* Internal Utility Functions
- *
- * The following functions are intended to be used inside algorithms as
- * utilities for manipulating aom_codec_* data structures.
- */
-struct aom_codec_pkt_list {
-  unsigned int cnt;
-  unsigned int max;
-  struct aom_codec_cx_pkt pkts[1];
-};
-
-#define aom_codec_pkt_list_decl(n)     \
-  union {                              \
-    struct aom_codec_pkt_list head;    \
-    struct {                           \
-      struct aom_codec_pkt_list head;  \
-      struct aom_codec_cx_pkt pkts[n]; \
-    } alloc;                           \
-  }
-
-#define aom_codec_pkt_list_init(m) \
-  (m)->alloc.head.cnt = 0,         \
-  (m)->alloc.head.max = sizeof((m)->alloc.pkts) / sizeof((m)->alloc.pkts[0])
-
-int aom_codec_pkt_list_add(struct aom_codec_pkt_list *,
-                           const struct aom_codec_cx_pkt *);
-
-const aom_codec_cx_pkt_t *aom_codec_pkt_list_get(
-    struct aom_codec_pkt_list *list, aom_codec_iter_t *iter);
-
-#include <stdio.h>
-#include <setjmp.h>
-
-struct aom_internal_error_info {
-  aom_codec_err_t error_code;
-  int has_detail;
-  char detail[80];
-  int setjmp;
-  jmp_buf jmp;
-};
-
-#define CLANG_ANALYZER_NORETURN
-#if defined(__has_feature)
-#if __has_feature(attribute_analyzer_noreturn)
-#undef CLANG_ANALYZER_NORETURN
-#define CLANG_ANALYZER_NORETURN __attribute__((analyzer_noreturn))
-#endif
-#endif
-
-void aom_internal_error(struct aom_internal_error_info *info,
-                        aom_codec_err_t error, const char *fmt,
-                        ...) CLANG_ANALYZER_NORETURN;
-
-#if CONFIG_DEBUG
-#define AOM_CHECK_MEM_ERROR(error_info, lval, expr)                         \
-  do {                                                                      \
-    lval = (expr);                                                          \
-    if (!lval)                                                              \
-      aom_internal_error(error_info, AOM_CODEC_MEM_ERROR,                   \
-                         "Failed to allocate " #lval " at %s:%d", __FILE__, \
-                         __LINE__);                                         \
-  } while (0)
-#else
-#define AOM_CHECK_MEM_ERROR(error_info, lval, expr)       \
-  do {                                                    \
-    lval = (expr);                                        \
-    if (!lval)                                            \
-      aom_internal_error(error_info, AOM_CODEC_MEM_ERROR, \
-                         "Failed to allocate " #lval);    \
-  } while (0)
-#endif
-#ifdef __cplusplus
-}  // extern "C"
-#endif
-
-#endif  // AOM_INTERNAL_AOM_CODEC_INTERNAL_H_

aom/src/aom_codec.c

@@ -1,134 +0,0 @@
-/*
- * Copyright (c) 2016, Alliance for Open Media. All rights reserved
- *
- * This source code is subject to the terms of the BSD 2 Clause License and
- * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
- * was not distributed with this source code in the LICENSE file, you can
- * obtain it at www.aomedia.org/license/software. If the Alliance for Open
- * Media Patent License 1.0 was not distributed with this source code in the
- * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
- */
-
-/*!\file
- * \brief Provides the high level interface to wrap decoder algorithms.
- *
- */
-#include <stdarg.h>
-#include <stdlib.h>
-#include "aom/aom_integer.h"
-#include "aom/internal/aom_codec_internal.h"
-#include "aom_version.h"
-
-#define SAVE_STATUS(ctx, var) (ctx ? (ctx->err = var) : var)
-
-int aom_codec_version(void) { return VERSION_PACKED; }
-
-const char *aom_codec_version_str(void) { return VERSION_STRING_NOSP; }
-
-const char *aom_codec_version_extra_str(void) { return VERSION_EXTRA; }
-
-const char *aom_codec_iface_name(aom_codec_iface_t *iface) {
-  return iface ? iface->name : "<invalid interface>";
-}
-
-const char *aom_codec_err_to_string(aom_codec_err_t err) {
-  switch (err) {
-    case AOM_CODEC_OK: return "Success";
-    case AOM_CODEC_ERROR: return "Unspecified internal error";
-    case AOM_CODEC_MEM_ERROR: return "Memory allocation error";
-    case AOM_CODEC_ABI_MISMATCH: return "ABI version mismatch";
-    case AOM_CODEC_INCAPABLE:
-      return "Codec does not implement requested capability";
-    case AOM_CODEC_UNSUP_BITSTREAM:
-      return "Bitstream not supported by this decoder";
-    case AOM_CODEC_UNSUP_FEATURE:
-      return "Bitstream required feature not supported by this decoder";
-    case AOM_CODEC_CORRUPT_FRAME: return "Corrupt frame detected";
-    case AOM_CODEC_INVALID_PARAM: return "Invalid parameter";
-    case AOM_CODEC_LIST_END: return "End of iterated list";
-  }
-
-  return "Unrecognized error code";
-}
-
-const char *aom_codec_error(aom_codec_ctx_t *ctx) {
-  return (ctx) ? aom_codec_err_to_string(ctx->err)
-               : aom_codec_err_to_string(AOM_CODEC_INVALID_PARAM);
-}
-
-const char *aom_codec_error_detail(aom_codec_ctx_t *ctx) {
-  if (ctx && ctx->err)
-    return ctx->priv ? ctx->priv->err_detail : ctx->err_detail;
-
-  return NULL;
-}
-
-aom_codec_err_t aom_codec_destroy(aom_codec_ctx_t *ctx) {
-  aom_codec_err_t res;
-
-  if (!ctx)
-    res = AOM_CODEC_INVALID_PARAM;
-  else if (!ctx->iface || !ctx->priv)
-    res = AOM_CODEC_ERROR;
-  else {
-    ctx->iface->destroy((aom_codec_alg_priv_t *)ctx->priv);
-
-    ctx->iface = NULL;
-    ctx->name = NULL;
-    ctx->priv = NULL;
-    res = AOM_CODEC_OK;
-  }
-
-  return SAVE_STATUS(ctx, res);
-}
-
-aom_codec_caps_t aom_codec_get_caps(aom_codec_iface_t *iface) {
-  return (iface) ? iface->caps : 0;
-}
-
-aom_codec_err_t aom_codec_control_(aom_codec_ctx_t *ctx, int ctrl_id, ...) {
-  aom_codec_err_t res;
-
-  if (!ctx || !ctrl_id)
-    res = AOM_CODEC_INVALID_PARAM;
-  else if (!ctx->iface || !ctx->priv || !ctx->iface->ctrl_maps)
-    res = AOM_CODEC_ERROR;
-  else {
-    aom_codec_ctrl_fn_map_t *entry;
-
-    res = AOM_CODEC_ERROR;
-
-    for (entry = ctx->iface->ctrl_maps; entry && entry->fn; entry++) {
-      if (!entry->ctrl_id || entry->ctrl_id == ctrl_id) {
-        va_list ap;
-
-        va_start(ap, ctrl_id);
-        res = entry->fn((aom_codec_alg_priv_t *)ctx->priv, ap);
-        va_end(ap);
-        break;
-      }
-    }
-  }
-
-  return SAVE_STATUS(ctx, res);
-}
-
-void aom_internal_error(struct aom_internal_error_info *info,
-                        aom_codec_err_t error, const char *fmt, ...) {
-  va_list ap;
-
-  info->error_code = error;
-  info->has_detail = 0;
-
-  if (fmt) {
-    size_t sz = sizeof(info->detail);
-
-    info->has_detail = 1;
-    va_start(ap, fmt);
-    vsnprintf(info->detail, sz - 1, fmt, ap);
-    va_end(ap);
-    info->detail[sz - 1] = '\0';
-  }
-
-  if (info->setjmp) longjmp(info->jmp, info->error_code);
-}

aom/src/aom_decoder.c

@@ -1,189 +0,0 @@
-/*
- * Copyright (c) 2016, Alliance for Open Media. All rights reserved
- *
- * This source code is subject to the terms of the BSD 2 Clause License and
- * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
- * was not distributed with this source code in the LICENSE file, you can
- * obtain it at www.aomedia.org/license/software. If the Alliance for Open
- * Media Patent License 1.0 was not distributed with this source code in the
- * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
- */
-
-/*!\file
- * \brief Provides the high level interface to wrap decoder algorithms.
- *
- */
-#include <string.h>
-#include "aom/internal/aom_codec_internal.h"
-
-#define SAVE_STATUS(ctx, var) (ctx ? (ctx->err = var) : var)
-
-static aom_codec_alg_priv_t *get_alg_priv(aom_codec_ctx_t *ctx) {
-  return (aom_codec_alg_priv_t *)ctx->priv;
-}
-
-aom_codec_err_t aom_codec_dec_init_ver(aom_codec_ctx_t *ctx,
-                                       aom_codec_iface_t *iface,
-                                       const aom_codec_dec_cfg_t *cfg,
-                                       aom_codec_flags_t flags, int ver) {
-  aom_codec_err_t res;
-
-  if (ver != AOM_DECODER_ABI_VERSION)
-    res = AOM_CODEC_ABI_MISMATCH;
-  else if (!ctx || !iface)
-    res = AOM_CODEC_INVALID_PARAM;
-  else if (iface->abi_version != AOM_CODEC_INTERNAL_ABI_VERSION)
-    res = AOM_CODEC_ABI_MISMATCH;
-  else if ((flags & AOM_CODEC_USE_POSTPROC) &&
-           !(iface->caps & AOM_CODEC_CAP_POSTPROC))
-    res = AOM_CODEC_INCAPABLE;
-  else if ((flags & AOM_CODEC_USE_ERROR_CONCEALMENT) &&
-           !(iface->caps & AOM_CODEC_CAP_ERROR_CONCEALMENT))
-    res = AOM_CODEC_INCAPABLE;
-  else if ((flags & AOM_CODEC_USE_INPUT_FRAGMENTS) &&
-           !(iface->caps & AOM_CODEC_CAP_INPUT_FRAGMENTS))
-    res = AOM_CODEC_INCAPABLE;
-  else if (!(iface->caps & AOM_CODEC_CAP_DECODER))
-    res = AOM_CODEC_INCAPABLE;
-  else {
-    memset(ctx, 0, sizeof(*ctx));
-    ctx->iface = iface;
-    ctx->name = iface->name;
-    ctx->priv = NULL;
-    ctx->init_flags = flags;
-    ctx->config.dec = cfg;
-
-    res = ctx->iface->init(ctx, NULL);
-    if (res) {
-      ctx->err_detail = ctx->priv ? ctx->priv->err_detail : NULL;
-      aom_codec_destroy(ctx);
-    }
-  }
-
-  return SAVE_STATUS(ctx, res);
-}
-
-aom_codec_err_t aom_codec_peek_stream_info(aom_codec_iface_t *iface,
-                                           const uint8_t *data,
-                                           unsigned int data_sz,
-                                           aom_codec_stream_info_t *si) {
-  aom_codec_err_t res;
-
-  if (!iface || !data || !data_sz || !si ||
-      si->sz < sizeof(aom_codec_stream_info_t))
-    res = AOM_CODEC_INVALID_PARAM;
-  else {
-    /* Set default/unknown values */
-    si->w = 0;
-    si->h = 0;
-
-    res = iface->dec.peek_si(data, data_sz, si);
-  }
-
-  return res;
-}
-
-aom_codec_err_t aom_codec_get_stream_info(aom_codec_ctx_t *ctx,
-                                          aom_codec_stream_info_t *si) {
-  aom_codec_err_t res;
-
-  if (!ctx || !si || si->sz < sizeof(aom_codec_stream_info_t))
-    res = AOM_CODEC_INVALID_PARAM;
-  else if (!ctx->iface || !ctx->priv)
-    res = AOM_CODEC_ERROR;
-  else {
-    /* Set default/unknown values */
-    si->w = 0;
-    si->h = 0;
-
-    res = ctx->iface->dec.get_si(get_alg_priv(ctx), si);
-  }
-
-  return SAVE_STATUS(ctx, res);
-}
-
-aom_codec_err_t aom_codec_decode(aom_codec_ctx_t *ctx, const uint8_t *data,
-                                 unsigned int data_sz, void *user_priv,
-                                 long deadline) {
-  aom_codec_err_t res;
-
-  /* Sanity checks */
-  /* NULL data ptr allowed if data_sz is 0 too */
-  if (!ctx || (!data && data_sz) || (data && !data_sz))
-    res = AOM_CODEC_INVALID_PARAM;
-  else if (!ctx->iface || !ctx->priv)
-    res = AOM_CODEC_ERROR;
-  else {
-    res = ctx->iface->dec.decode(get_alg_priv(ctx), data, data_sz, user_priv,
-                                 deadline);
-  }
-
-  return SAVE_STATUS(ctx, res);
-}
-
-aom_image_t *aom_codec_get_frame(aom_codec_ctx_t *ctx, aom_codec_iter_t *iter) {
-  aom_image_t *img;
-
-  if (!ctx || !iter || !ctx->iface || !ctx->priv)
-    img = NULL;
-  else
-    img = ctx->iface->dec.get_frame(get_alg_priv(ctx), iter);
-
-  return img;
-}
-
-aom_codec_err_t aom_codec_register_put_frame_cb(aom_codec_ctx_t *ctx,
-                                                aom_codec_put_frame_cb_fn_t cb,
-                                                void *user_priv) {
-  aom_codec_err_t res;
-
-  if (!ctx || !cb)
-    res = AOM_CODEC_INVALID_PARAM;
-  else if (!ctx->iface || !ctx->priv ||
-           !(ctx->iface->caps & AOM_CODEC_CAP_PUT_FRAME))
-    res = AOM_CODEC_ERROR;
-  else {
-    ctx->priv->dec.put_frame_cb.u.put_frame = cb;
-    ctx->priv->dec.put_frame_cb.user_priv = user_priv;
-    res = AOM_CODEC_OK;
-  }
-
-  return SAVE_STATUS(ctx, res);
-}
-
-aom_codec_err_t aom_codec_register_put_slice_cb(aom_codec_ctx_t *ctx,
-                                                aom_codec_put_slice_cb_fn_t cb,
-                                                void *user_priv) {
-  aom_codec_err_t res;
-
-  if (!ctx || !cb)
-    res = AOM_CODEC_INVALID_PARAM;
-  else if (!ctx->iface || !ctx->priv ||
-           !(ctx->iface->caps & AOM_CODEC_CAP_PUT_SLICE))
-    res = AOM_CODEC_ERROR;
-  else {
-    ctx->priv->dec.put_slice_cb.u.put_slice = cb;
-    ctx->priv->dec.put_slice_cb.user_priv = user_priv;
-    res = AOM_CODEC_OK;
-  }
-
-  return SAVE_STATUS(ctx, res);
-}
-
-aom_codec_err_t aom_codec_set_frame_buffer_functions(
-    aom_codec_ctx_t *ctx, aom_get_frame_buffer_cb_fn_t cb_get,
-    aom_release_frame_buffer_cb_fn_t cb_release, void *cb_priv) {
-  aom_codec_err_t res;
-
-  if (!ctx || !cb_get || !cb_release) {
-    res = AOM_CODEC_INVALID_PARAM;
-  } else if (!ctx->iface || !ctx->priv ||
-             !(ctx->iface->caps & AOM_CODEC_CAP_EXTERNAL_FRAME_BUFFER)) {
-    res = AOM_CODEC_ERROR;
-  } else {
-    res = ctx->iface->dec.set_fb_fn(get_alg_priv(ctx), cb_get, cb_release,
-                                    cb_priv);
-  }
-
-  return SAVE_STATUS(ctx, res);
-}

aom/src/aom_encoder.c

@@ -1,380 +0,0 @@
-/*
- * Copyright (c) 2016, Alliance for Open Media. All rights reserved
- *
- * This source code is subject to the terms of the BSD 2 Clause License and
- * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
- * was not distributed with this source code in the LICENSE file, you can
- * obtain it at www.aomedia.org/license/software. If the Alliance for Open
- * Media Patent License 1.0 was not distributed with this source code in the
- * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
- */
-
-/*!\file
- * \brief Provides the high level interface to wrap encoder algorithms.
- *
- */
-#include <limits.h>
-#include <string.h>
-#include "aom_config.h"
-#include "aom/internal/aom_codec_internal.h"
-
-#define SAVE_STATUS(ctx, var) (ctx ? (ctx->err = var) : var)
-
-static aom_codec_alg_priv_t *get_alg_priv(aom_codec_ctx_t *ctx) {
-  return (aom_codec_alg_priv_t *)ctx->priv;
-}
-
-aom_codec_err_t aom_codec_enc_init_ver(aom_codec_ctx_t *ctx,
-                                       aom_codec_iface_t *iface,
-                                       const aom_codec_enc_cfg_t *cfg,
-                                       aom_codec_flags_t flags, int ver) {
-  aom_codec_err_t res;
-
-  if (ver != AOM_ENCODER_ABI_VERSION)
-    res = AOM_CODEC_ABI_MISMATCH;
-  else if (!ctx || !iface || !cfg)
-    res = AOM_CODEC_INVALID_PARAM;
-  else if (iface->abi_version != AOM_CODEC_INTERNAL_ABI_VERSION)
-    res = AOM_CODEC_ABI_MISMATCH;
-  else if (!(iface->caps & AOM_CODEC_CAP_ENCODER))
-    res = AOM_CODEC_INCAPABLE;
-  else if ((flags & AOM_CODEC_USE_PSNR) && !(iface->caps & AOM_CODEC_CAP_PSNR))
-    res = AOM_CODEC_INCAPABLE;
-  else if ((flags & AOM_CODEC_USE_OUTPUT_PARTITION) &&
-           !(iface->caps & AOM_CODEC_CAP_OUTPUT_PARTITION))
-    res = AOM_CODEC_INCAPABLE;
-  else {
-    ctx->iface = iface;
-    ctx->name = iface->name;
-    ctx->priv = NULL;
-    ctx->init_flags = flags;
-    ctx->config.enc = cfg;
-    res = ctx->iface->init(ctx, NULL);
-
-    if (res) {
-      ctx->err_detail = ctx->priv ? ctx->priv->err_detail : NULL;
-      aom_codec_destroy(ctx);
-    }
-  }
-
-  return SAVE_STATUS(ctx, res);
-}
-
-aom_codec_err_t aom_codec_enc_init_multi_ver(
-    aom_codec_ctx_t *ctx, aom_codec_iface_t *iface, aom_codec_enc_cfg_t *cfg,
-    int num_enc, aom_codec_flags_t flags, aom_rational_t *dsf, int ver) {
-  aom_codec_err_t res = AOM_CODEC_OK;
-
-  if (ver != AOM_ENCODER_ABI_VERSION)
-    res = AOM_CODEC_ABI_MISMATCH;
-  else if (!ctx || !iface || !cfg || (num_enc > 16 || num_enc < 1))
-    res = AOM_CODEC_INVALID_PARAM;
-  else if (iface->abi_version != AOM_CODEC_INTERNAL_ABI_VERSION)
-    res = AOM_CODEC_ABI_MISMATCH;
-  else if (!(iface->caps & AOM_CODEC_CAP_ENCODER))
-    res = AOM_CODEC_INCAPABLE;
-  else if ((flags & AOM_CODEC_USE_PSNR) && !(iface->caps & AOM_CODEC_CAP_PSNR))
-    res = AOM_CODEC_INCAPABLE;
-  else if ((flags & AOM_CODEC_USE_OUTPUT_PARTITION) &&
-           !(iface->caps & AOM_CODEC_CAP_OUTPUT_PARTITION))
-    res = AOM_CODEC_INCAPABLE;
-  else {
-    int i;
-    void *mem_loc = NULL;
-
-    if (!(res = iface->enc.mr_get_mem_loc(cfg, &mem_loc))) {
-      for (i = 0; i < num_enc; i++) {
-        aom_codec_priv_enc_mr_cfg_t mr_cfg;
-
-        /* Validate down-sampling factor. */
-        if (dsf->num < 1 || dsf->num > 4096 || dsf->den < 1 ||
-            dsf->den > dsf->num) {
-          res = AOM_CODEC_INVALID_PARAM;
-          break;
-        }
-
-        mr_cfg.mr_low_res_mode_info = mem_loc;
-        mr_cfg.mr_total_resolutions = num_enc;
-        mr_cfg.mr_encoder_id = num_enc - 1 - i;
-        mr_cfg.mr_down_sampling_factor.num = dsf->num;
-        mr_cfg.mr_down_sampling_factor.den = dsf->den;
-
-        /* Force Key-frame synchronization. Namely, encoder at higher
-         * resolution always use the same frame_type chosen by the
-         * lowest-resolution encoder.
-         */
-        if (mr_cfg.mr_encoder_id) cfg->kf_mode = AOM_KF_DISABLED;
-
-        ctx->iface = iface;
-        ctx->name = iface->name;
-        ctx->priv = NULL;
-        ctx->init_flags = flags;
-        ctx->config.enc = cfg;
-        res = ctx->iface->init(ctx, &mr_cfg);
-
-        if (res) {
-          const char *error_detail = ctx->priv ? ctx->priv->err_detail : NULL;
-          /* Destroy current ctx */
-          ctx->err_detail = error_detail;
-          aom_codec_destroy(ctx);
-
-          /* Destroy already allocated high-level ctx */
-          while (i) {
-            ctx--;
-            ctx->err_detail = error_detail;
-            aom_codec_destroy(ctx);
-            i--;
-          }
-        }
-
-        if (res) break;
-
-        ctx++;
-        cfg++;
-        dsf++;
-      }
-      ctx--;
-    }
-  }
-
-  return SAVE_STATUS(ctx, res);
-}
-
-aom_codec_err_t aom_codec_enc_config_default(aom_codec_iface_t *iface,
-                                             aom_codec_enc_cfg_t *cfg,
-                                             unsigned int usage) {
-  aom_codec_err_t res;
-  aom_codec_enc_cfg_map_t *map;
-  int i;
-
-  if (!iface || !cfg || usage > INT_MAX)
-    res = AOM_CODEC_INVALID_PARAM;
-  else if (!(iface->caps & AOM_CODEC_CAP_ENCODER))
-    res = AOM_CODEC_INCAPABLE;
-  else {
-    res = AOM_CODEC_INVALID_PARAM;
-
-    for (i = 0; i < iface->enc.cfg_map_count; ++i) {
-      map = iface->enc.cfg_maps + i;
-      if (map->usage == (int)usage) {
-        *cfg = map->cfg;
-        cfg->g_usage = usage;
-        res = AOM_CODEC_OK;
-        break;
-      }
-    }
-  }
-
-  return res;
-}
-
-#if ARCH_X86 || ARCH_X86_64
-/* On X86, disable the x87 unit's internal 80 bit precision for better
- * consistency with the SSE unit's 64 bit precision.
- */
-#include "aom_ports/x86.h"
-#define FLOATING_POINT_INIT() \
-  do {                        \
-    unsigned short x87_orig_mode = x87_set_double_precision();
-#define FLOATING_POINT_RESTORE()       \
-  x87_set_control_word(x87_orig_mode); \
-  }                                    \
-  while (0)
-
-#else
-static void FLOATING_POINT_INIT() {}
-static void FLOATING_POINT_RESTORE() {}
-#endif
-
-aom_codec_err_t aom_codec_encode(aom_codec_ctx_t *ctx, const aom_image_t *img,
-                                 aom_codec_pts_t pts, unsigned long duration,
-                                 aom_enc_frame_flags_t flags,
-                                 unsigned long deadline) {
-  aom_codec_err_t res = AOM_CODEC_OK;
-
-  if (!ctx || (img && !duration))
-    res = AOM_CODEC_INVALID_PARAM;
-  else if (!ctx->iface || !ctx->priv)
-    res = AOM_CODEC_ERROR;
-  else if (!(ctx->iface->caps & AOM_CODEC_CAP_ENCODER))
-    res = AOM_CODEC_INCAPABLE;
-  else {
-    unsigned int num_enc = ctx->priv->enc.total_encoders;
-
-    /* Execute in a normalized floating point environment, if the platform
-     * requires it.
-     */
-    FLOATING_POINT_INIT();
-
-    if (num_enc == 1)
-      res = ctx->iface->enc.encode(get_alg_priv(ctx), img, pts, duration, flags,
-                                   deadline);
-    else {
-      /* Multi-resolution encoding:
-       * Encode multi-levels in reverse order. For example,
-       * if mr_total_resolutions = 3, first encode level 2,
-       * then encode level 1, and finally encode level 0.
-       */
-      int i;
-
-      ctx += num_enc - 1;
-      if (img) img += num_enc - 1;
-
-      for (i = num_enc - 1; i >= 0; i--) {
-        if ((res = ctx->iface->enc.encode(get_alg_priv(ctx), img, pts, duration,
-                                          flags, deadline)))
-          break;
-
-        ctx--;
-        if (img) img--;
-      }
-      ctx++;
-    }
-
-    FLOATING_POINT_RESTORE();
-  }
-
-  return SAVE_STATUS(ctx, res);
-}
-
-const aom_codec_cx_pkt_t *aom_codec_get_cx_data(aom_codec_ctx_t *ctx,
-                                                aom_codec_iter_t *iter) {
-  const aom_codec_cx_pkt_t *pkt = NULL;
-
-  if (ctx) {
-    if (!iter)
-      ctx->err = AOM_CODEC_INVALID_PARAM;
-    else if (!ctx->iface || !ctx->priv)
-      ctx->err = AOM_CODEC_ERROR;
-    else if (!(ctx->iface->caps & AOM_CODEC_CAP_ENCODER))
-      ctx->err = AOM_CODEC_INCAPABLE;
-    else
-      pkt = ctx->iface->enc.get_cx_data(get_alg_priv(ctx), iter);
-  }
-
-  if (pkt && pkt->kind == AOM_CODEC_CX_FRAME_PKT) {
-    // If the application has specified a destination area for the
-    // compressed data, and the codec has not placed the data there,
-    // and it fits, copy it.
-    aom_codec_priv_t *const priv = ctx->priv;
-    char *const dst_buf = (char *)priv->enc.cx_data_dst_buf.buf;
-
-    if (dst_buf && pkt->data.raw.buf != dst_buf &&
-        pkt->data.raw.sz + priv->enc.cx_data_pad_before +
-                priv->enc.cx_data_pad_after <=
-            priv->enc.cx_data_dst_buf.sz) {
-      aom_codec_cx_pkt_t *modified_pkt = &priv->enc.cx_data_pkt;
-
-      memcpy(dst_buf + priv->enc.cx_data_pad_before, pkt->data.raw.buf,
-             pkt->data.raw.sz);
-      *modified_pkt = *pkt;
-      modified_pkt->data.raw.buf = dst_buf;
-      modified_pkt->data.raw.sz +=
-          priv->enc.cx_data_pad_before + priv->enc.cx_data_pad_after;
-      pkt = modified_pkt;
-    }
-
-    if (dst_buf == pkt->data.raw.buf) {
-      priv->enc.cx_data_dst_buf.buf = dst_buf + pkt->data.raw.sz;
-      priv->enc.cx_data_dst_buf.sz -= pkt->data.raw.sz;
-    }
-  }
-
-  return pkt;
-}
-
-aom_codec_err_t aom_codec_set_cx_data_buf(aom_codec_ctx_t *ctx,
-                                          const aom_fixed_buf_t *buf,
-                                          unsigned int pad_before,
-                                          unsigned int pad_after) {
-  if (!ctx || !ctx->priv) return AOM_CODEC_INVALID_PARAM;
-
-  if (buf) {
-    ctx->priv->enc.cx_data_dst_buf = *buf;
-    ctx->priv->enc.cx_data_pad_before = pad_before;
-    ctx->priv->enc.cx_data_pad_after = pad_after;
-  } else {
-    ctx->priv->enc.cx_data_dst_buf.buf = NULL;
-    ctx->priv->enc.cx_data_dst_buf.sz = 0;
-    ctx->priv->enc.cx_data_pad_before = 0;
-    ctx->priv->enc.cx_data_pad_after = 0;
-  }
-
-  return AOM_CODEC_OK;
-}
-
-const aom_image_t *aom_codec_get_preview_frame(aom_codec_ctx_t *ctx) {
-  aom_image_t *img = NULL;
-
-  if (ctx) {
-    if (!ctx->iface || !ctx->priv)
-      ctx->err = AOM_CODEC_ERROR;
-    else if (!(ctx->iface->caps & AOM_CODEC_CAP_ENCODER))
-      ctx->err = AOM_CODEC_INCAPABLE;
-    else if (!ctx->iface->enc.get_preview)
-      ctx->err = AOM_CODEC_INCAPABLE;
-    else
-      img = ctx->iface->enc.get_preview(get_alg_priv(ctx));
-  }
-
-  return img;
-}
-
-aom_fixed_buf_t *aom_codec_get_global_headers(aom_codec_ctx_t *ctx) {
-  aom_fixed_buf_t *buf = NULL;
-
-  if (ctx) {
-    if (!ctx->iface || !ctx->priv)
-      ctx->err = AOM_CODEC_ERROR;
-    else if (!(ctx->iface->caps & AOM_CODEC_CAP_ENCODER))
-      ctx->err = AOM_CODEC_INCAPABLE;
-    else if (!ctx->iface->enc.get_glob_hdrs)
-      ctx->err = AOM_CODEC_INCAPABLE;
-    else
-      buf = ctx->iface->enc.get_glob_hdrs(get_alg_priv(ctx));
-  }
-
-  return buf;
-}
-
-aom_codec_err_t aom_codec_enc_config_set(aom_codec_ctx_t *ctx,
-                                         const aom_codec_enc_cfg_t *cfg) {
-  aom_codec_err_t res;
-
-  if (!ctx || !ctx->iface || !ctx->priv || !cfg)
-    res = AOM_CODEC_INVALID_PARAM;
-  else if (!(ctx->iface->caps & AOM_CODEC_CAP_ENCODER))
-    res = AOM_CODEC_INCAPABLE;
-  else
-    res = ctx->iface->enc.cfg_set(get_alg_priv(ctx), cfg);
-
-  return SAVE_STATUS(ctx, res);
-}
-
-int aom_codec_pkt_list_add(struct aom_codec_pkt_list *list,
-                           const struct aom_codec_cx_pkt *pkt) {
-  if (list->cnt < list->max) {
-    list->pkts[list->cnt++] = *pkt;
-    return 0;
-  }
-
-  return 1;
-}
-
-const aom_codec_cx_pkt_t *aom_codec_pkt_list_get(
-    struct aom_codec_pkt_list *list, aom_codec_iter_t *iter) {
-  const aom_codec_cx_pkt_t *pkt;
-
-  if (!(*iter)) {
-    *iter = list->pkts;
-  }
-
-  pkt = (const aom_codec_cx_pkt_t *)*iter;
-
-  if ((size_t)(pkt - list->pkts) < list->cnt)
-    *iter = pkt + 1;
-  else
-    pkt = NULL;
-
-  return pkt;
-}

aom/src/aom_image.c

@@ -1,240 +0,0 @@
-/*
- * Copyright (c) 2016, Alliance for Open Media. All rights reserved
- *
- * This source code is subject to the terms of the BSD 2 Clause License and
- * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
- * was not distributed with this source code in the LICENSE file, you can
- * obtain it at www.aomedia.org/license/software. If the Alliance for Open
- * Media Patent License 1.0 was not distributed with this source code in the
- * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
- */
-
-#include <stdlib.h>
-#include <string.h>
-
-#include "aom/aom_image.h"
-#include "aom/aom_integer.h"
-#include "aom_mem/aom_mem.h"
-
-static aom_image_t *img_alloc_helper(aom_image_t *img, aom_img_fmt_t fmt,
-                                     unsigned int d_w, unsigned int d_h,
-                                     unsigned int buf_align,
-                                     unsigned int stride_align,
-                                     unsigned char *img_data) {
-  unsigned int h, w, s, xcs, ycs, bps;
-  unsigned int stride_in_bytes;
-  int align;
-
-  /* Treat align==0 like align==1 */
-  if (!buf_align) buf_align = 1;
-
-  /* Validate alignment (must be power of 2) */
-  if (buf_align & (buf_align - 1)) goto fail;
-
-  /* Treat align==0 like align==1 */
-  if (!stride_align) stride_align = 1;
-
-  /* Validate alignment (must be power of 2) */
-  if (stride_align & (stride_align - 1)) goto fail;
-
-  /* Get sample size for this format */
-  switch (fmt) {
-    case AOM_IMG_FMT_RGB32:
-    case AOM_IMG_FMT_RGB32_LE:
-    case AOM_IMG_FMT_ARGB:
-    case AOM_IMG_FMT_ARGB_LE: bps = 32; break;
-    case AOM_IMG_FMT_RGB24:
-    case AOM_IMG_FMT_BGR24: bps = 24; break;
-    case AOM_IMG_FMT_RGB565:
-    case AOM_IMG_FMT_RGB565_LE:
-    case AOM_IMG_FMT_RGB555:
-    case AOM_IMG_FMT_RGB555_LE:
-    case AOM_IMG_FMT_UYVY:
-    case AOM_IMG_FMT_YUY2:
-    case AOM_IMG_FMT_YVYU: bps = 16; break;
-    case AOM_IMG_FMT_I420:
-    case AOM_IMG_FMT_YV12:
-    case AOM_IMG_FMT_AOMI420:
-    case AOM_IMG_FMT_AOMYV12: bps = 12; break;
-    case AOM_IMG_FMT_I422:
-    case AOM_IMG_FMT_I440: bps = 16; break;
-    case AOM_IMG_FMT_I444: bps = 24; break;
-    case AOM_IMG_FMT_I42016: bps = 24; break;
-    case AOM_IMG_FMT_I42216:
-    case AOM_IMG_FMT_I44016: bps = 32; break;
-    case AOM_IMG_FMT_I44416: bps = 48; break;
-    default: bps = 16; break;
-  }
-
-  /* Get chroma shift values for this format */
-  switch (fmt) {
-    case AOM_IMG_FMT_I420:
-    case AOM_IMG_FMT_YV12:
-    case AOM_IMG_FMT_AOMI420:
-    case AOM_IMG_FMT_AOMYV12:
-    case AOM_IMG_FMT_I422:
-    case AOM_IMG_FMT_I42016:
-    case AOM_IMG_FMT_I42216: xcs = 1; break;
-    default: xcs = 0; break;
-  }
-
-  switch (fmt) {
-    case AOM_IMG_FMT_I420:
-    case AOM_IMG_FMT_I440:
-    case AOM_IMG_FMT_YV12:
-    case AOM_IMG_FMT_AOMI420:
-    case AOM_IMG_FMT_AOMYV12:
-    case AOM_IMG_FMT_I42016:
-    case AOM_IMG_FMT_I44016: ycs = 1; break;
-    default: ycs = 0; break;
-  }
-
-  /* Calculate storage sizes given the chroma subsampling */
-  align = (1 << xcs) - 1;
-  w = (d_w + align) & ~align;
-  align = (1 << ycs) - 1;
-  h = (d_h + align) & ~align;
-  s = (fmt & AOM_IMG_FMT_PLANAR) ? w : bps * w / 8;
-  s = (s + stride_align - 1) & ~(stride_align - 1);
-  stride_in_bytes = (fmt & AOM_IMG_FMT_HIGHBITDEPTH) ? s * 2 : s;
-
-  /* Allocate the new image */
-  if (!img) {
-    img = (aom_image_t *)calloc(1, sizeof(aom_image_t));
-
-    if (!img) goto fail;
-
-    img->self_allocd = 1;
-  } else {
-    memset(img, 0, sizeof(aom_image_t));
-  }
-
-  img->img_data = img_data;
-
-  if (!img_data) {
-    const uint64_t alloc_size = (fmt & AOM_IMG_FMT_PLANAR)
-                                    ? (uint64_t)h * s * bps / 8
-                                    : (uint64_t)h * s;
-
-    if (alloc_size != (size_t)alloc_size) goto fail;
-
-    img->img_data = (uint8_t *)aom_memalign(buf_align, (size_t)alloc_size);
-    img->img_data_owner = 1;
-  }
-
-  if (!img->img_data) goto fail;
-
-  img->fmt = fmt;
-  img->bit_depth = (fmt & AOM_IMG_FMT_HIGHBITDEPTH) ? 16 : 8;
-  img->w = w;
-  img->h = h;
-  img->x_chroma_shift = xcs;
-  img->y_chroma_shift = ycs;
-  img->bps = bps;
-
-  /* Calculate strides */
-  img->stride[AOM_PLANE_Y] = img->stride[AOM_PLANE_ALPHA] = stride_in_bytes;
-  img->stride[AOM_PLANE_U] = img->stride[AOM_PLANE_V] = stride_in_bytes >> xcs;
-
-  /* Default viewport to entire image */
-  if (!aom_img_set_rect(img, 0, 0, d_w, d_h)) return img;
-
-fail:
-  aom_img_free(img);
-  return NULL;
-}
-
-aom_image_t *aom_img_alloc(aom_image_t *img, aom_img_fmt_t fmt,
-                           unsigned int d_w, unsigned int d_h,
-                           unsigned int align) {
-  return img_alloc_helper(img, fmt, d_w, d_h, align, align, NULL);
-}
-
-aom_image_t *aom_img_wrap(aom_image_t *img, aom_img_fmt_t fmt, unsigned int d_w,
-                          unsigned int d_h, unsigned int stride_align,
-                          unsigned char *img_data) {
-  /* By setting buf_align = 1, we don't change buffer alignment in this
-   * function. */
-  return img_alloc_helper(img, fmt, d_w, d_h, 1, stride_align, img_data);
-}
-
-int aom_img_set_rect(aom_image_t *img, unsigned int x, unsigned int y,
-                     unsigned int w, unsigned int h) {
-  unsigned char *data;
-
-  if (x + w <= img->w && y + h <= img->h) {
-    img->d_w = w;
-    img->d_h = h;
-
-    /* Calculate plane pointers */
-    if (!(img->fmt & AOM_IMG_FMT_PLANAR)) {
-      img->planes[AOM_PLANE_PACKED] =
-          img->img_data + x * img->bps / 8 + y * img->stride[AOM_PLANE_PACKED];
-    } else {
-      const int bytes_per_sample =
-          (img->fmt & AOM_IMG_FMT_HIGHBITDEPTH) ? 2 : 1;
-      data = img->img_data;
-
-      if (img->fmt & AOM_IMG_FMT_HAS_ALPHA) {
-        img->planes[AOM_PLANE_ALPHA] =
-            data + x * bytes_per_sample + y * img->stride[AOM_PLANE_ALPHA];
-        data += img->h * img->stride[AOM_PLANE_ALPHA];
-      }
-
-      img->planes[AOM_PLANE_Y] =
-          data + x * bytes_per_sample + y * img->stride[AOM_PLANE_Y];
-      data += img->h * img->stride[AOM_PLANE_Y];
-
-      if (!(img->fmt & AOM_IMG_FMT_UV_FLIP)) {
-        img->planes[AOM_PLANE_U] =
-            data + (x >> img->x_chroma_shift) * bytes_per_sample +
-            (y >> img->y_chroma_shift) * img->stride[AOM_PLANE_U];
-        data += (img->h >> img->y_chroma_shift) * img->stride[AOM_PLANE_U];
-        img->planes[AOM_PLANE_V] =
-            data + (x >> img->x_chroma_shift) * bytes_per_sample +
-            (y >> img->y_chroma_shift) * img->stride[AOM_PLANE_V];
-      } else {
-        img->planes[AOM_PLANE_V] =
-            data + (x >> img->x_chroma_shift) * bytes_per_sample +
-            (y >> img->y_chroma_shift) * img->stride[AOM_PLANE_V];
-        data += (img->h >> img->y_chroma_shift) * img->stride[AOM_PLANE_V];
-        img->planes[AOM_PLANE_U] =
-            data + (x >> img->x_chroma_shift) * bytes_per_sample +
-            (y >> img->y_chroma_shift) * img->stride[AOM_PLANE_U];
-      }
-    }
-    return 0;
-  }
-  return -1;
-}
-
-void aom_img_flip(aom_image_t *img) {
-  /* Note: In the calculation pointer adjustment calculation, we want the
-   * rhs to be promoted to a signed type. Section 6.3.1.8 of the ISO C99
-   * standard indicates that if the adjustment parameter is unsigned, the
-   * stride parameter will be promoted to unsigned, causing errors when
-   * the lhs is a larger type than the rhs.
-   */
-  img->planes[AOM_PLANE_Y] += (signed)(img->d_h - 1) * img->stride[AOM_PLANE_Y];
-  img->stride[AOM_PLANE_Y] = -img->stride[AOM_PLANE_Y];
-
-  img->planes[AOM_PLANE_U] += (signed)((img->d_h >> img->y_chroma_shift) - 1) *
-                              img->stride[AOM_PLANE_U];
-  img->stride[AOM_PLANE_U] = -img->stride[AOM_PLANE_U];
-
-  img->planes[AOM_PLANE_V] += (signed)((img->d_h >> img->y_chroma_shift) - 1) *
-                              img->stride[AOM_PLANE_V];
-  img->stride[AOM_PLANE_V] = -img->stride[AOM_PLANE_V];
-
-  img->planes[AOM_PLANE_ALPHA] +=
-      (signed)(img->d_h - 1) * img->stride[AOM_PLANE_ALPHA];
-  img->stride[AOM_PLANE_ALPHA] = -img->stride[AOM_PLANE_ALPHA];
-}
-
-void aom_img_free(aom_image_t *img) {
-  if (img) {
-    if (img->img_data && img->img_data_owner) aom_free(img->img_data);
-
-    if (img->self_allocd) free(img);
-  }
-}

aom_dsp/add_noise.c

@@ -1,72 +0,0 @@
-/*
- *  Copyright (c) 2015 The WebM project authors. All Rights Reserved.
- *
- *  Use of this source code is governed by a BSD-style license
- *  that can be found in the LICENSE file in the root of the source
- *  tree. An additional intellectual property rights grant can be found
- *  in the file PATENTS.  All contributing project authors may
- *  be found in the AUTHORS file in the root of the source tree.
- */
-
-#include <math.h>
-#include <stdlib.h>
-
-#include "./aom_config.h"
-#include "./aom_dsp_rtcd.h"
-
-#include "aom/aom_integer.h"
-#include "aom_ports/mem.h"
-
-void aom_plane_add_noise_c(uint8_t *start, char *noise, char blackclamp[16],
-                           char whiteclamp[16], char bothclamp[16],
-                           unsigned int width, unsigned int height, int pitch) {
-  unsigned int i, j;
-
-  for (i = 0; i < height; ++i) {
-    uint8_t *pos = start + i * pitch;
-    char *ref = (char *)(noise + (rand() & 0xff));  // NOLINT
-
-    for (j = 0; j < width; ++j) {
-      int v = pos[j];
-
-      v = clamp(v - blackclamp[0], 0, 255);
-      v = clamp(v + bothclamp[0], 0, 255);
-      v = clamp(v - whiteclamp[0], 0, 255);
-
-      pos[j] = v + ref[j];
-    }
-  }
-}
-
-static double gaussian(double sigma, double mu, double x) {
-  return 1 / (sigma * sqrt(2.0 * 3.14159265)) *
-         (exp(-(x - mu) * (x - mu) / (2 * sigma * sigma)));
-}
-
-int aom_setup_noise(double sigma, int size, char *noise) {
-  char char_dist[256];
-  int next = 0, i, j;
-
-  // set up a 256 entry lookup that matches gaussian distribution
-  for (i = -32; i < 32; ++i) {
-    const int a_i = (int)(0.5 + 256 * gaussian(sigma, 0, i));
-    if (a_i) {
-      for (j = 0; j < a_i; ++j) {
-        char_dist[next + j] = (char)i;
-      }
-      next = next + j;
-    }
-  }
-
-  // Rounding error - might mean we have less than 256.
-  for (; next < 256; ++next) {
-    char_dist[next] = 0;
-  }
-
-  for (i = 0; i < size; ++i) {
-    noise[i] = char_dist[rand() & 0xff];  // NOLINT
-  }
-
-  // Returns the highest non 0 value used in distribution.
-  return -char_dist[0];
-}

aom_dsp/ans.c

@@ -1,64 +0,0 @@
-/*
- * Copyright (c) 2016, Alliance for Open Media. All rights reserved
- *
- * This source code is subject to the terms of the BSD 2 Clause License and
- * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
- * was not distributed with this source code in the LICENSE file, you can
- * obtain it at www.aomedia.org/license/software. If the Alliance for Open
- * Media Patent License 1.0 was not distributed with this source code in the
- * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
- */
-
-#include <assert.h>
-#include "./aom_config.h"
-#include "aom/aom_integer.h"
-#include "aom_dsp/ans.h"
-#include "aom_dsp/prob.h"
-
-static int find_largest(const aom_cdf_prob *const pdf_tab, int num_syms) {
-  int largest_idx = -1;
-  int largest_p = -1;
-  int i;
-  for (i = 0; i < num_syms; ++i) {
-    int p = pdf_tab[i];
-    if (p > largest_p) {
-      largest_p = p;
-      largest_idx = i;
-    }
-  }
-  return largest_idx;
-}
-
-void aom_rans_merge_prob8_pdf(aom_cdf_prob *const out_pdf,
-                              const AnsP8 node_prob,
-                              const aom_cdf_prob *const src_pdf, int in_syms) {
-  int i;
-  int adjustment = RANS_PRECISION;
-  const int round_fact = ANS_P8_PRECISION >> 1;
-  const AnsP8 p1 = ANS_P8_PRECISION - node_prob;
-  const int out_syms = in_syms + 1;
-  assert(src_pdf != out_pdf);
-
-  out_pdf[0] = node_prob << (RANS_PROB_BITS - ANS_P8_SHIFT);
-  adjustment -= out_pdf[0];
-  for (i = 0; i < in_syms; ++i) {
-    int p = (p1 * src_pdf[i] + round_fact) >> ANS_P8_SHIFT;
-    p = AOMMIN(p, (int)RANS_PRECISION - in_syms);
-    p = AOMMAX(p, 1);
-    out_pdf[i + 1] = p;
-    adjustment -= p;
-  }
-
-  // Adjust probabilities so they sum to the total probability
-  if (adjustment > 0) {
-    i = find_largest(out_pdf, out_syms);
-    out_pdf[i] += adjustment;
-  } else {
-    while (adjustment < 0) {
-      i = find_largest(out_pdf, out_syms);
-      --out_pdf[i];
-      assert(out_pdf[i] > 0);
-      adjustment++;
-    }
-  }
-}

aom_dsp/ans.h

@@ -1,44 +0,0 @@
-/*
- * Copyright (c) 2016, Alliance for Open Media. All rights reserved
- *
- * This source code is subject to the terms of the BSD 2 Clause License and
- * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
- * was not distributed with this source code in the LICENSE file, you can
- * obtain it at www.aomedia.org/license/software. If the Alliance for Open
- * Media Patent License 1.0 was not distributed with this source code in the
- * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
- */
-
-#ifndef AOM_DSP_ANS_H_
-#define AOM_DSP_ANS_H_
-// Constants, types and utilities for Asymmetric Numeral Systems
-// http://arxiv.org/abs/1311.2540v2
-
-#include <assert.h>
-#include "./aom_config.h"
-#include "aom/aom_integer.h"
-#include "aom_dsp/prob.h"
-
-#ifdef __cplusplus
-extern "C" {
-#endif  // __cplusplus
-
-typedef uint8_t AnsP8;
-#define ANS_P8_PRECISION 256u
-#define ANS_P8_SHIFT 8
-#define RANS_PROB_BITS 15
-#define RANS_PRECISION (1u << RANS_PROB_BITS)
-
-// L_BASE % PRECISION must be 0. Increasing L_BASE beyond 2**15 will cause uabs
-// to overflow.
-#define L_BASE (RANS_PRECISION)
-#define IO_BASE 256
-// Range I = { L_BASE, L_BASE + 1, ..., L_BASE * IO_BASE - 1 }
-
-void aom_rans_merge_prob8_pdf(aom_cdf_prob *const out_pdf,
-                              const AnsP8 node_prob,
-                              const aom_cdf_prob *const src_pdf, int in_syms);
-#ifdef __cplusplus
-}  // extern "C"
-#endif  // __cplusplus
-#endif  // AOM_DSP_ANS_H_

aom_dsp/ansreader.h

@@ -1,146 +0,0 @@
-/*
- * Copyright (c) 2016, Alliance for Open Media. All rights reserved
- *
- * This source code is subject to the terms of the BSD 2 Clause License and
- * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
- * was not distributed with this source code in the LICENSE file, you can
- * obtain it at www.aomedia.org/license/software. If the Alliance for Open
- * Media Patent License 1.0 was not distributed with this source code in the
- * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
- */
-
-#ifndef AOM_DSP_ANSREADER_H_
-#define AOM_DSP_ANSREADER_H_
-// A uABS and rANS decoder implementation of Asymmetric Numeral Systems
-// http://arxiv.org/abs/1311.2540v2
-
-#include <assert.h>
-#include "./aom_config.h"
-#include "aom/aom_integer.h"
-#include "aom_dsp/prob.h"
-#include "aom_dsp/ans.h"
-#include "aom_ports/mem_ops.h"
-#if CONFIG_ACCOUNTING
-#include "av1/common/accounting.h"
-#endif
-
-#ifdef __cplusplus
-extern "C" {
-#endif  // __cplusplus
-
-struct AnsDecoder {
-  const uint8_t *buf;
-  int buf_offset;
-  uint32_t state;
-#if CONFIG_ACCOUNTING
-  Accounting *accounting;
-#endif
-};
-
-static INLINE int uabs_read(struct AnsDecoder *ans, AnsP8 p0) {
-  AnsP8 p = ANS_P8_PRECISION - p0;
-  int s;
-  unsigned xp, sp;
-  unsigned state = ans->state;
-  while (state < L_BASE && ans->buf_offset > 0) {
-    state = state * IO_BASE + ans->buf[--ans->buf_offset];
-  }
-  sp = state * p;
-  xp = sp / ANS_P8_PRECISION;
-  s = (sp & 0xFF) >= p0;
-  if (s)
-    ans->state = xp;
-  else
-    ans->state = state - xp;
-  return s;
-}
-
-static INLINE int uabs_read_bit(struct AnsDecoder *ans) {
-  int s;
-  unsigned state = ans->state;
-  while (state < L_BASE && ans->buf_offset > 0) {
-    state = state * IO_BASE + ans->buf[--ans->buf_offset];
-  }
-  s = (int)(state & 1);
-  ans->state = state >> 1;
-  return s;
-}
-
-struct rans_dec_sym {
-  uint8_t val;
-  aom_cdf_prob prob;
-  aom_cdf_prob cum_prob;  // not-inclusive
-};
-
-static INLINE void fetch_sym(struct rans_dec_sym *out, const aom_cdf_prob *cdf,
-                             aom_cdf_prob rem) {
-  int i;
-  aom_cdf_prob cum_prob = 0, top_prob;
-  // TODO(skal): if critical, could be a binary search.
-  // Or, better, an O(1) alias-table.
-  for (i = 0; rem >= (top_prob = cdf[i]); ++i) {
-    cum_prob = top_prob;
-  }
-  out->val = i;
-  out->prob = top_prob - cum_prob;
-  out->cum_prob = cum_prob;
-}
-
-static INLINE int rans_read(struct AnsDecoder *ans, const aom_cdf_prob *tab) {
-  unsigned rem;
-  unsigned quo;
-  struct rans_dec_sym sym;
-  while (ans->state < L_BASE && ans->buf_offset > 0) {
-    ans->state = ans->state * IO_BASE + ans->buf[--ans->buf_offset];
-  }
-  quo = ans->state / RANS_PRECISION;
-  rem = ans->state % RANS_PRECISION;
-  fetch_sym(&sym, tab, rem);
-  ans->state = quo * sym.prob + rem - sym.cum_prob;
-  return sym.val;
-}
-
-static INLINE int ans_read_init(struct AnsDecoder *const ans,
-                                const uint8_t *const buf, int offset) {
-  unsigned x;
-  if (offset < 1) return 1;
-  ans->buf = buf;
-  x = buf[offset - 1] >> 6;
-  if (x == 0) {
-    ans->buf_offset = offset - 1;
-    ans->state = buf[offset - 1] & 0x3F;
-  } else if (x == 1) {
-    if (offset < 2) return 1;
-    ans->buf_offset = offset - 2;
-    ans->state = mem_get_le16(buf + offset - 2) & 0x3FFF;
-  } else if (x == 2) {
-    if (offset < 3) return 1;
-    ans->buf_offset = offset - 3;
-    ans->state = mem_get_le24(buf + offset - 3) & 0x3FFFFF;
-  } else if ((buf[offset - 1] & 0xE0) == 0xE0) {
-    if (offset < 4) return 1;
-    ans->buf_offset = offset - 4;
-    ans->state = mem_get_le32(buf + offset - 4) & 0x1FFFFFFF;
-  } else {
-    // 110xxxxx implies this byte is a superframe marker
-    return 1;
-  }
-#if CONFIG_ACCOUNTING
-  ans->accounting = NULL;
-#endif
-  ans->state += L_BASE;
-  if (ans->state >= L_BASE * IO_BASE) return 1;
-  return 0;
-}
-
-static INLINE int ans_read_end(struct AnsDecoder *const ans) {
-  return ans->state == L_BASE;
-}
-
-static INLINE int ans_reader_has_error(const struct AnsDecoder *const ans) {
-  return ans->state < L_BASE && ans->buf_offset == 0;
-}
-#ifdef __cplusplus
-}  // extern "C"
-#endif  // __cplusplus
-#endif  // AOM_DSP_ANSREADER_H_

aom_dsp/answriter.h

@@ -1,120 +0,0 @@
-/*
- * Copyright (c) 2016, Alliance for Open Media. All rights reserved
- *
- * This source code is subject to the terms of the BSD 2 Clause License and
- * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
- * was not distributed with this source code in the LICENSE file, you can
- * obtain it at www.aomedia.org/license/software. If the Alliance for Open
- * Media Patent License 1.0 was not distributed with this source code in the
- * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
- */
-
-#ifndef AOM_DSP_ANSWRITER_H_
-#define AOM_DSP_ANSWRITER_H_
-// A uABS and rANS encoder implementation of Asymmetric Numeral Systems
-// http://arxiv.org/abs/1311.2540v2
-
-#include <assert.h>
-#include "./aom_config.h"
-#include "aom/aom_integer.h"
-#include "aom_dsp/ans.h"
-#include "aom_dsp/prob.h"
-#include "aom_ports/mem_ops.h"
-#include "av1/common/odintrin.h"
-
-#if RANS_PRECISION <= OD_DIVU_DMAX
-#define ANS_DIVREM(quotient, remainder, dividend, divisor) \
-  do {                                                     \
-    quotient = OD_DIVU_SMALL((dividend), (divisor));       \
-    remainder = (dividend) - (quotient) * (divisor);       \
-  } while (0)
-#else
-#define ANS_DIVREM(quotient, remainder, dividend, divisor) \
-  do {                                                     \
-    quotient = (dividend) / (divisor);                     \
-    remainder = (dividend) % (divisor);                    \
-  } while (0)
-#endif
-
-#define ANS_DIV8(dividend, divisor) OD_DIVU_SMALL((dividend), (divisor))
-
-#ifdef __cplusplus
-extern "C" {
-#endif  // __cplusplus
-
-struct AnsCoder {
-  uint8_t *buf;
-  int buf_offset;
-  uint32_t state;
-};
-
-static INLINE void ans_write_init(struct AnsCoder *const ans,
-                                  uint8_t *const buf) {
-  ans->buf = buf;
-  ans->buf_offset = 0;
-  ans->state = L_BASE;
-}
-
-static INLINE int ans_write_end(struct AnsCoder *const ans) {
-  uint32_t state;
-  assert(ans->state >= L_BASE);
-  assert(ans->state < L_BASE * IO_BASE);
-  state = ans->state - L_BASE;
-  if (state < (1 << 6)) {
-    ans->buf[ans->buf_offset] = (0x00 << 6) + state;
-    return ans->buf_offset + 1;
-  } else if (state < (1 << 14)) {
-    mem_put_le16(ans->buf + ans->buf_offset, (0x01 << 14) + state);
-    return ans->buf_offset + 2;
-  } else if (state < (1 << 22)) {
-    mem_put_le24(ans->buf + ans->buf_offset, (0x02 << 22) + state);
-    return ans->buf_offset + 3;
-  } else if (state < (1 << 29)) {
-    mem_put_le32(ans->buf + ans->buf_offset, (0x07 << 29) + state);
-    return ans->buf_offset + 4;
-  } else {
-    assert(0 && "State is too large to be serialized");
-    return ans->buf_offset;
-  }
-}
-
-// uABS with normalization
-static INLINE void uabs_write(struct AnsCoder *ans, int val, AnsP8 p0) {
-  AnsP8 p = ANS_P8_PRECISION - p0;
-  const unsigned l_s = val ? p : p0;
-  while (ans->state >= L_BASE / ANS_P8_PRECISION * IO_BASE * l_s) {
-    ans->buf[ans->buf_offset++] = ans->state % IO_BASE;
-    ans->state /= IO_BASE;
-  }
-  if (!val)
-    ans->state = ANS_DIV8(ans->state * ANS_P8_PRECISION, p0);
-  else
-    ans->state = ANS_DIV8((ans->state + 1) * ANS_P8_PRECISION + p - 1, p) - 1;
-}
-
-struct rans_sym {
-  aom_cdf_prob prob;
-  aom_cdf_prob cum_prob;  // not-inclusive
-};
-
-// rANS with normalization
-// sym->prob takes the place of l_s from the paper
-// ANS_P10_PRECISION is m
-static INLINE void rans_write(struct AnsCoder *ans,
-                              const struct rans_sym *const sym) {
-  const aom_cdf_prob p = sym->prob;
-  unsigned quot, rem;
-  while (ans->state >= L_BASE / RANS_PRECISION * IO_BASE * p) {
-    ans->buf[ans->buf_offset++] = ans->state % IO_BASE;
-    ans->state /= IO_BASE;
-  }
-  ANS_DIVREM(quot, rem, ans->state, p);
-  ans->state = quot * RANS_PRECISION + rem + sym->cum_prob;
-}
-
-#undef ANS_DIV8
-#undef ANS_DIVREM
-#ifdef __cplusplus
-}  // extern "C"
-#endif  // __cplusplus
-#endif  // AOM_DSP_ANSWRITER_H_

aom_dsp/aom_convolve.c

@@ -1,600 +0,0 @@
-/*
- * Copyright (c) 2016, Alliance for Open Media. All rights reserved
- *
- * This source code is subject to the terms of the BSD 2 Clause License and
- * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
- * was not distributed with this source code in the LICENSE file, you can
- * obtain it at www.aomedia.org/license/software. If the Alliance for Open
- * Media Patent License 1.0 was not distributed with this source code in the
- * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
- */
-
-#include <assert.h>
-#include <string.h>
-
-#include "./aom_config.h"
-#include "./aom_dsp_rtcd.h"
-#include "aom/aom_integer.h"
-#include "aom_dsp/aom_convolve.h"
-#include "aom_dsp/aom_dsp_common.h"
-#include "aom_dsp/aom_filter.h"
-#include "aom_ports/mem.h"
-
-static void convolve_horiz(const uint8_t *src, ptrdiff_t src_stride,
-                           uint8_t *dst, ptrdiff_t dst_stride,
-                           const InterpKernel *x_filters, int x0_q4,
-                           int x_step_q4, int w, int h) {
-  int x, y;
-  src -= SUBPEL_TAPS / 2 - 1;
-  for (y = 0; y < h; ++y) {
-    int x_q4 = x0_q4;
-    for (x = 0; x < w; ++x) {
-      const uint8_t *const src_x = &src[x_q4 >> SUBPEL_BITS];
-      const int16_t *const x_filter = x_filters[x_q4 & SUBPEL_MASK];
-      int k, sum = 0;
-      for (k = 0; k < SUBPEL_TAPS; ++k) sum += src_x[k] * x_filter[k];
-      dst[x] = clip_pixel(ROUND_POWER_OF_TWO(sum, FILTER_BITS));
-      x_q4 += x_step_q4;
-    }
-    src += src_stride;
-    dst += dst_stride;
-  }
-}
-
-static void convolve_avg_horiz(const uint8_t *src, ptrdiff_t src_stride,
-                               uint8_t *dst, ptrdiff_t dst_stride,
-                               const InterpKernel *x_filters, int x0_q4,
-                               int x_step_q4, int w, int h) {
-  int x, y;
-  src -= SUBPEL_TAPS / 2 - 1;
-  for (y = 0; y < h; ++y) {
-    int x_q4 = x0_q4;
-    for (x = 0; x < w; ++x) {
-      const uint8_t *const src_x = &src[x_q4 >> SUBPEL_BITS];
-      const int16_t *const x_filter = x_filters[x_q4 & SUBPEL_MASK];
-      int k, sum = 0;
-      for (k = 0; k < SUBPEL_TAPS; ++k) sum += src_x[k] * x_filter[k];
-      dst[x] = ROUND_POWER_OF_TWO(
-          dst[x] + clip_pixel(ROUND_POWER_OF_TWO(sum, FILTER_BITS)), 1);
-      x_q4 += x_step_q4;
-    }
-    src += src_stride;
-    dst += dst_stride;
-  }
-}
-
-static void convolve_vert(const uint8_t *src, ptrdiff_t src_stride,
-                          uint8_t *dst, ptrdiff_t dst_stride,
-                          const InterpKernel *y_filters, int y0_q4,
-                          int y_step_q4, int w, int h) {
-  int x, y;
-  src -= src_stride * (SUBPEL_TAPS / 2 - 1);
-
-  for (x = 0; x < w; ++x) {
-    int y_q4 = y0_q4;
-    for (y = 0; y < h; ++y) {
-      const unsigned char *src_y = &src[(y_q4 >> SUBPEL_BITS) * src_stride];
-      const int16_t *const y_filter = y_filters[y_q4 & SUBPEL_MASK];
-      int k, sum = 0;
-      for (k = 0; k < SUBPEL_TAPS; ++k)
-        sum += src_y[k * src_stride] * y_filter[k];
-      dst[y * dst_stride] = clip_pixel(ROUND_POWER_OF_TWO(sum, FILTER_BITS));
-      y_q4 += y_step_q4;
-    }
-    ++src;
-    ++dst;
-  }
-}
-
-static void convolve_avg_vert(const uint8_t *src, ptrdiff_t src_stride,
-                              uint8_t *dst, ptrdiff_t dst_stride,
-                              const InterpKernel *y_filters, int y0_q4,
-                              int y_step_q4, int w, int h) {
-  int x, y;
-  src -= src_stride * (SUBPEL_TAPS / 2 - 1);
-
-  for (x = 0; x < w; ++x) {
-    int y_q4 = y0_q4;
-    for (y = 0; y < h; ++y) {
-      const unsigned char *src_y = &src[(y_q4 >> SUBPEL_BITS) * src_stride];
-      const int16_t *const y_filter = y_filters[y_q4 & SUBPEL_MASK];
-      int k, sum = 0;
-      for (k = 0; k < SUBPEL_TAPS; ++k)
-        sum += src_y[k * src_stride] * y_filter[k];
-      dst[y * dst_stride] = ROUND_POWER_OF_TWO(
-          dst[y * dst_stride] +
-              clip_pixel(ROUND_POWER_OF_TWO(sum, FILTER_BITS)),
-          1);
-      y_q4 += y_step_q4;
-    }
-    ++src;
-    ++dst;
-  }
-}
-
-static void convolve(const uint8_t *src, ptrdiff_t src_stride, uint8_t *dst,
-                     ptrdiff_t dst_stride, const InterpKernel *const x_filters,
-                     int x0_q4, int x_step_q4,
-                     const InterpKernel *const y_filters, int y0_q4,
-                     int y_step_q4, int w, int h) {
-  // Note: Fixed size intermediate buffer, temp, places limits on parameters.
-  // 2d filtering proceeds in 2 steps:
-  //   (1) Interpolate horizontally into an intermediate buffer, temp.
-  //   (2) Interpolate temp vertically to derive the sub-pixel result.
-  // Deriving the maximum number of rows in the temp buffer (135):
-  // --Smallest scaling factor is x1/2 ==> y_step_q4 = 32 (Normative).
-  // --Largest block size is 64x64 pixels.
-  // --64 rows in the downscaled frame span a distance of (64 - 1) * 32 in the
-  //   original frame (in 1/16th pixel units).
-  // --Must round-up because block may be located at sub-pixel position.
-  // --Require an additional SUBPEL_TAPS rows for the 8-tap filter tails.
-  // --((64 - 1) * 32 + 15) >> 4 + 8 = 135.
-  uint8_t temp[MAX_EXT_SIZE * MAX_SB_SIZE];
-  int intermediate_height =
-      (((h - 1) * y_step_q4 + y0_q4) >> SUBPEL_BITS) + SUBPEL_TAPS;
-
-  assert(w <= MAX_SB_SIZE);
-  assert(h <= MAX_SB_SIZE);
-
-  assert(y_step_q4 <= 32);
-  assert(x_step_q4 <= 32);
-
-  convolve_horiz(src - src_stride * (SUBPEL_TAPS / 2 - 1), src_stride, temp,
-                 MAX_SB_SIZE, x_filters, x0_q4, x_step_q4, w,
-                 intermediate_height);
-  convolve_vert(temp + MAX_SB_SIZE * (SUBPEL_TAPS / 2 - 1), MAX_SB_SIZE, dst,
-                dst_stride, y_filters, y0_q4, y_step_q4, w, h);
-}
-
-static const InterpKernel *get_filter_base(const int16_t *filter) {
-  // NOTE: This assumes that the filter table is 256-byte aligned.
-  // TODO(agrange) Modify to make independent of table alignment.
-  return (const InterpKernel *)(((intptr_t)filter) & ~((intptr_t)0xFF));
-}
-
-static int get_filter_offset(const int16_t *f, const InterpKernel *base) {
-  return (int)((const InterpKernel *)(intptr_t)f - base);
-}
-
-void aom_convolve8_horiz_c(const uint8_t *src, ptrdiff_t src_stride,
-                           uint8_t *dst, ptrdiff_t dst_stride,
-                           const int16_t *filter_x, int x_step_q4,
-                           const int16_t *filter_y, int y_step_q4, int w,
-                           int h) {
-  const InterpKernel *const filters_x = get_filter_base(filter_x);
-  const int x0_q4 = get_filter_offset(filter_x, filters_x);
-
-  (void)filter_y;
-  (void)y_step_q4;
-
-  convolve_horiz(src, src_stride, dst, dst_stride, filters_x, x0_q4, x_step_q4,
-                 w, h);
-}
-
-void aom_convolve8_avg_horiz_c(const uint8_t *src, ptrdiff_t src_stride,
-                               uint8_t *dst, ptrdiff_t dst_stride,
-                               const int16_t *filter_x, int x_step_q4,
-                               const int16_t *filter_y, int y_step_q4, int w,
-                               int h) {
-  const InterpKernel *const filters_x = get_filter_base(filter_x);
-  const int x0_q4 = get_filter_offset(filter_x, filters_x);
-
-  (void)filter_y;
-  (void)y_step_q4;
-
-  convolve_avg_horiz(src, src_stride, dst, dst_stride, filters_x, x0_q4,
-                     x_step_q4, w, h);
-}
-
-void aom_convolve8_vert_c(const uint8_t *src, ptrdiff_t src_stride,
-                          uint8_t *dst, ptrdiff_t dst_stride,
-                          const int16_t *filter_x, int x_step_q4,
-                          const int16_t *filter_y, int y_step_q4, int w,
-                          int h) {
-  const InterpKernel *const filters_y = get_filter_base(filter_y);
-  const int y0_q4 = get_filter_offset(filter_y, filters_y);
-
-  (void)filter_x;
-  (void)x_step_q4;
-
-  convolve_vert(src, src_stride, dst, dst_stride, filters_y, y0_q4, y_step_q4,
-                w, h);
-}
-
-void aom_convolve8_avg_vert_c(const uint8_t *src, ptrdiff_t src_stride,
-                              uint8_t *dst, ptrdiff_t dst_stride,
-                              const int16_t *filter_x, int x_step_q4,
-                              const int16_t *filter_y, int y_step_q4, int w,
-                              int h) {
-  const InterpKernel *const filters_y = get_filter_base(filter_y);
-  const int y0_q4 = get_filter_offset(filter_y, filters_y);
-
-  (void)filter_x;
-  (void)x_step_q4;
-
-  convolve_avg_vert(src, src_stride, dst, dst_stride, filters_y, y0_q4,
-                    y_step_q4, w, h);
-}
-
-void aom_convolve8_c(const uint8_t *src, ptrdiff_t src_stride, uint8_t *dst,
-                     ptrdiff_t dst_stride, const int16_t *filter_x,
-                     int x_step_q4, const int16_t *filter_y, int y_step_q4,
-                     int w, int h) {
-  const InterpKernel *const filters_x = get_filter_base(filter_x);
-  const int x0_q4 = get_filter_offset(filter_x, filters_x);
-
-  const InterpKernel *const filters_y = get_filter_base(filter_y);
-  const int y0_q4 = get_filter_offset(filter_y, filters_y);
-
-  convolve(src, src_stride, dst, dst_stride, filters_x, x0_q4, x_step_q4,
-           filters_y, y0_q4, y_step_q4, w, h);
-}
-
-void aom_convolve8_avg_c(const uint8_t *src, ptrdiff_t src_stride, uint8_t *dst,
-                         ptrdiff_t dst_stride, const int16_t *filter_x,
-                         int x_step_q4, const int16_t *filter_y, int y_step_q4,
-                         int w, int h) {
-  /* Fixed size intermediate buffer places limits on parameters. */
-  DECLARE_ALIGNED(16, uint8_t, temp[MAX_SB_SIZE * MAX_SB_SIZE]);
-  assert(w <= MAX_SB_SIZE);
-  assert(h <= MAX_SB_SIZE);
-
-  aom_convolve8_c(src, src_stride, temp, MAX_SB_SIZE, filter_x, x_step_q4,
-                  filter_y, y_step_q4, w, h);
-  aom_convolve_avg_c(temp, MAX_SB_SIZE, dst, dst_stride, NULL, 0, NULL, 0, w,
-                     h);
-}
-
-void aom_convolve_copy_c(const uint8_t *src, ptrdiff_t src_stride, uint8_t *dst,
-                         ptrdiff_t dst_stride, const int16_t *filter_x,
-                         int filter_x_stride, const int16_t *filter_y,
-                         int filter_y_stride, int w, int h) {
-  int r;
-
-  (void)filter_x;
-  (void)filter_x_stride;
-  (void)filter_y;
-  (void)filter_y_stride;
-
-  for (r = h; r > 0; --r) {
-    memcpy(dst, src, w);
-    src += src_stride;
-    dst += dst_stride;
-  }
-}
-
-void aom_convolve_avg_c(const uint8_t *src, ptrdiff_t src_stride, uint8_t *dst,
-                        ptrdiff_t dst_stride, const int16_t *filter_x,
-                        int filter_x_stride, const int16_t *filter_y,
-                        int filter_y_stride, int w, int h) {
-  int x, y;
-
-  (void)filter_x;
-  (void)filter_x_stride;
-  (void)filter_y;
-  (void)filter_y_stride;
-
-  for (y = 0; y < h; ++y) {
-    for (x = 0; x < w; ++x) dst[x] = ROUND_POWER_OF_TWO(dst[x] + src[x], 1);
-
-    src += src_stride;
-    dst += dst_stride;
-  }
-}
-
-void aom_scaled_horiz_c(const uint8_t *src, ptrdiff_t src_stride, uint8_t *dst,
-                        ptrdiff_t dst_stride, const int16_t *filter_x,
-                        int x_step_q4, const int16_t *filter_y, int y_step_q4,
-                        int w, int h) {
-  aom_convolve8_horiz_c(src, src_stride, dst, dst_stride, filter_x, x_step_q4,
-                        filter_y, y_step_q4, w, h);
-}
-
-void aom_scaled_vert_c(const uint8_t *src, ptrdiff_t src_stride, uint8_t *dst,
-                       ptrdiff_t dst_stride, const int16_t *filter_x,
-                       int x_step_q4, const int16_t *filter_y, int y_step_q4,
-                       int w, int h) {
-  aom_convolve8_vert_c(src, src_stride, dst, dst_stride, filter_x, x_step_q4,
-                       filter_y, y_step_q4, w, h);
-}
-
-void aom_scaled_2d_c(const uint8_t *src, ptrdiff_t src_stride, uint8_t *dst,
-                     ptrdiff_t dst_stride, const int16_t *filter_x,
-                     int x_step_q4, const int16_t *filter_y, int y_step_q4,
-                     int w, int h) {
-  aom_convolve8_c(src, src_stride, dst, dst_stride, filter_x, x_step_q4,
-                  filter_y, y_step_q4, w, h);
-}
-
-void aom_scaled_avg_horiz_c(const uint8_t *src, ptrdiff_t src_stride,
-                            uint8_t *dst, ptrdiff_t dst_stride,
-                            const int16_t *filter_x, int x_step_q4,
-                            const int16_t *filter_y, int y_step_q4, int w,
-                            int h) {
-  aom_convolve8_avg_horiz_c(src, src_stride, dst, dst_stride, filter_x,
-                            x_step_q4, filter_y, y_step_q4, w, h);
-}
-
-void aom_scaled_avg_vert_c(const uint8_t *src, ptrdiff_t src_stride,
-                           uint8_t *dst, ptrdiff_t dst_stride,
-                           const int16_t *filter_x, int x_step_q4,
-                           const int16_t *filter_y, int y_step_q4, int w,
-                           int h) {
-  aom_convolve8_avg_vert_c(src, src_stride, dst, dst_stride, filter_x,
-                           x_step_q4, filter_y, y_step_q4, w, h);
-}
-
-void aom_scaled_avg_2d_c(const uint8_t *src, ptrdiff_t src_stride, uint8_t *dst,
-                         ptrdiff_t dst_stride, const int16_t *filter_x,
-                         int x_step_q4, const int16_t *filter_y, int y_step_q4,
-                         int w, int h) {
-  aom_convolve8_avg_c(src, src_stride, dst, dst_stride, filter_x, x_step_q4,
-                      filter_y, y_step_q4, w, h);
-}
-
-#if CONFIG_AOM_HIGHBITDEPTH
-static void highbd_convolve_horiz(const uint8_t *src8, ptrdiff_t src_stride,
-                                  uint8_t *dst8, ptrdiff_t dst_stride,
-                                  const InterpKernel *x_filters, int x0_q4,
-                                  int x_step_q4, int w, int h, int bd) {
-  int x, y;
-  uint16_t *src = CONVERT_TO_SHORTPTR(src8);
-  uint16_t *dst = CONVERT_TO_SHORTPTR(dst8);
-  src -= SUBPEL_TAPS / 2 - 1;
-  for (y = 0; y < h; ++y) {
-    int x_q4 = x0_q4;
-    for (x = 0; x < w; ++x) {
-      const uint16_t *const src_x = &src[x_q4 >> SUBPEL_BITS];
-      const int16_t *const x_filter = x_filters[x_q4 & SUBPEL_MASK];
-      int k, sum = 0;
-      for (k = 0; k < SUBPEL_TAPS; ++k) sum += src_x[k] * x_filter[k];
-      dst[x] = clip_pixel_highbd(ROUND_POWER_OF_TWO(sum, FILTER_BITS), bd);
-      x_q4 += x_step_q4;
-    }
-    src += src_stride;
-    dst += dst_stride;
-  }
-}
-
-static void highbd_convolve_avg_horiz(const uint8_t *src8, ptrdiff_t src_stride,
-                                      uint8_t *dst8, ptrdiff_t dst_stride,
-                                      const InterpKernel *x_filters, int x0_q4,
-                                      int x_step_q4, int w, int h, int bd) {
-  int x, y;
-  uint16_t *src = CONVERT_TO_SHORTPTR(src8);
-  uint16_t *dst = CONVERT_TO_SHORTPTR(dst8);
-  src -= SUBPEL_TAPS / 2 - 1;
-  for (y = 0; y < h; ++y) {
-    int x_q4 = x0_q4;
-    for (x = 0; x < w; ++x) {
-      const uint16_t *const src_x = &src[x_q4 >> SUBPEL_BITS];
-      const int16_t *const x_filter = x_filters[x_q4 & SUBPEL_MASK];
-      int k, sum = 0;
-      for (k = 0; k < SUBPEL_TAPS; ++k) sum += src_x[k] * x_filter[k];
-      dst[x] = ROUND_POWER_OF_TWO(
-          dst[x] + clip_pixel_highbd(ROUND_POWER_OF_TWO(sum, FILTER_BITS), bd),
-          1);
-      x_q4 += x_step_q4;
-    }
-    src += src_stride;
-    dst += dst_stride;
-  }
-}
-
-static void highbd_convolve_vert(const uint8_t *src8, ptrdiff_t src_stride,
-                                 uint8_t *dst8, ptrdiff_t dst_stride,
-                                 const InterpKernel *y_filters, int y0_q4,
-                                 int y_step_q4, int w, int h, int bd) {
-  int x, y;
-  uint16_t *src = CONVERT_TO_SHORTPTR(src8);
-  uint16_t *dst = CONVERT_TO_SHORTPTR(dst8);
-  src -= src_stride * (SUBPEL_TAPS / 2 - 1);
-  for (x = 0; x < w; ++x) {
-    int y_q4 = y0_q4;
-    for (y = 0; y < h; ++y) {
-      const uint16_t *src_y = &src[(y_q4 >> SUBPEL_BITS) * src_stride];
-      const int16_t *const y_filter = y_filters[y_q4 & SUBPEL_MASK];
-      int k, sum = 0;
-      for (k = 0; k < SUBPEL_TAPS; ++k)
-        sum += src_y[k * src_stride] * y_filter[k];
-      dst[y * dst_stride] =
-          clip_pixel_highbd(ROUND_POWER_OF_TWO(sum, FILTER_BITS), bd);
-      y_q4 += y_step_q4;
-    }
-    ++src;
-    ++dst;
-  }
-}
-
-static void highbd_convolve_avg_vert(const uint8_t *src8, ptrdiff_t src_stride,
-                                     uint8_t *dst8, ptrdiff_t dst_stride,
-                                     const InterpKernel *y_filters, int y0_q4,
-                                     int y_step_q4, int w, int h, int bd) {
-  int x, y;
-  uint16_t *src = CONVERT_TO_SHORTPTR(src8);
-  uint16_t *dst = CONVERT_TO_SHORTPTR(dst8);
-  src -= src_stride * (SUBPEL_TAPS / 2 - 1);
-  for (x = 0; x < w; ++x) {
-    int y_q4 = y0_q4;
-    for (y = 0; y < h; ++y) {
-      const uint16_t *src_y = &src[(y_q4 >> SUBPEL_BITS) * src_stride];
-      const int16_t *const y_filter = y_filters[y_q4 & SUBPEL_MASK];
-      int k, sum = 0;
-      for (k = 0; k < SUBPEL_TAPS; ++k)
-        sum += src_y[k * src_stride] * y_filter[k];
-      dst[y * dst_stride] = ROUND_POWER_OF_TWO(
-          dst[y * dst_stride] +
-              clip_pixel_highbd(ROUND_POWER_OF_TWO(sum, FILTER_BITS), bd),
-          1);
-      y_q4 += y_step_q4;
-    }
-    ++src;
-    ++dst;
-  }
-}
-
-static void highbd_convolve(const uint8_t *src, ptrdiff_t src_stride,
-                            uint8_t *dst, ptrdiff_t dst_stride,
-                            const InterpKernel *const x_filters, int x0_q4,
-                            int x_step_q4, const InterpKernel *const y_filters,
-                            int y0_q4, int y_step_q4, int w, int h, int bd) {
-  // Note: Fixed size intermediate buffer, temp, places limits on parameters.
-  // 2d filtering proceeds in 2 steps:
-  //   (1) Interpolate horizontally into an intermediate buffer, temp.
-  //   (2) Interpolate temp vertically to derive the sub-pixel result.
-  // Deriving the maximum number of rows in the temp buffer (135):
-  // --Smallest scaling factor is x1/2 ==> y_step_q4 = 32 (Normative).
-  // --Largest block size is 64x64 pixels.
-  // --64 rows in the downscaled frame span a distance of (64 - 1) * 32 in the
-  //   original frame (in 1/16th pixel units).
-  // --Must round-up because block may be located at sub-pixel position.
-  // --Require an additional SUBPEL_TAPS rows for the 8-tap filter tails.
-  // --((64 - 1) * 32 + 15) >> 4 + 8 = 135.
-  uint16_t temp[MAX_EXT_SIZE * MAX_SB_SIZE];
-  int intermediate_height =
-      (((h - 1) * y_step_q4 + y0_q4) >> SUBPEL_BITS) + SUBPEL_TAPS;
-
-  assert(w <= MAX_SB_SIZE);
-  assert(h <= MAX_SB_SIZE);
-  assert(y_step_q4 <= 32);
-  assert(x_step_q4 <= 32);
-
-  highbd_convolve_horiz(src - src_stride * (SUBPEL_TAPS / 2 - 1), src_stride,
-                        CONVERT_TO_BYTEPTR(temp), MAX_SB_SIZE, x_filters, x0_q4,
-                        x_step_q4, w, intermediate_height, bd);
-  highbd_convolve_vert(
-      CONVERT_TO_BYTEPTR(temp) + MAX_SB_SIZE * (SUBPEL_TAPS / 2 - 1),
-      MAX_SB_SIZE, dst, dst_stride, y_filters, y0_q4, y_step_q4, w, h, bd);
-}
-
-void aom_highbd_convolve8_horiz_c(const uint8_t *src, ptrdiff_t src_stride,
-                                  uint8_t *dst, ptrdiff_t dst_stride,
-                                  const int16_t *filter_x, int x_step_q4,
-                                  const int16_t *filter_y, int y_step_q4, int w,
-                                  int h, int bd) {
-  const InterpKernel *const filters_x = get_filter_base(filter_x);
-  const int x0_q4 = get_filter_offset(filter_x, filters_x);
-  (void)filter_y;
-  (void)y_step_q4;
-
-  highbd_convolve_horiz(src, src_stride, dst, dst_stride, filters_x, x0_q4,
-                        x_step_q4, w, h, bd);
-}
-
-void aom_highbd_convolve8_avg_horiz_c(const uint8_t *src, ptrdiff_t src_stride,
-                                      uint8_t *dst, ptrdiff_t dst_stride,
-                                      const int16_t *filter_x, int x_step_q4,
-                                      const int16_t *filter_y, int y_step_q4,
-                                      int w, int h, int bd) {
-  const InterpKernel *const filters_x = get_filter_base(filter_x);
-  const int x0_q4 = get_filter_offset(filter_x, filters_x);
-  (void)filter_y;
-  (void)y_step_q4;
-
-  highbd_convolve_avg_horiz(src, src_stride, dst, dst_stride, filters_x, x0_q4,
-                            x_step_q4, w, h, bd);
-}
-
-void aom_highbd_convolve8_vert_c(const uint8_t *src, ptrdiff_t src_stride,
-                                 uint8_t *dst, ptrdiff_t dst_stride,
-                                 const int16_t *filter_x, int x_step_q4,
-                                 const int16_t *filter_y, int y_step_q4, int w,
-                                 int h, int bd) {
-  const InterpKernel *const filters_y = get_filter_base(filter_y);
-  const int y0_q4 = get_filter_offset(filter_y, filters_y);
-  (void)filter_x;
-  (void)x_step_q4;
-
-  highbd_convolve_vert(src, src_stride, dst, dst_stride, filters_y, y0_q4,
-                       y_step_q4, w, h, bd);
-}
-
-void aom_highbd_convolve8_avg_vert_c(const uint8_t *src, ptrdiff_t src_stride,
-                                     uint8_t *dst, ptrdiff_t dst_stride,
-                                     const int16_t *filter_x, int x_step_q4,
-                                     const int16_t *filter_y, int y_step_q4,
-                                     int w, int h, int bd) {
-  const InterpKernel *const filters_y = get_filter_base(filter_y);
-  const int y0_q4 = get_filter_offset(filter_y, filters_y);
-  (void)filter_x;
-  (void)x_step_q4;
-
-  highbd_convolve_avg_vert(src, src_stride, dst, dst_stride, filters_y, y0_q4,
-                           y_step_q4, w, h, bd);
-}
-
-void aom_highbd_convolve8_c(const uint8_t *src, ptrdiff_t src_stride,
-                            uint8_t *dst, ptrdiff_t dst_stride,
-                            const int16_t *filter_x, int x_step_q4,
-                            const int16_t *filter_y, int y_step_q4, int w,
-                            int h, int bd) {
-  const InterpKernel *const filters_x = get_filter_base(filter_x);
-  const int x0_q4 = get_filter_offset(filter_x, filters_x);
-
-  const InterpKernel *const filters_y = get_filter_base(filter_y);
-  const int y0_q4 = get_filter_offset(filter_y, filters_y);
-
-  highbd_convolve(src, src_stride, dst, dst_stride, filters_x, x0_q4, x_step_q4,
-                  filters_y, y0_q4, y_step_q4, w, h, bd);
-}
-
-void aom_highbd_convolve8_avg_c(const uint8_t *src, ptrdiff_t src_stride,
-                                uint8_t *dst, ptrdiff_t dst_stride,
-                                const int16_t *filter_x, int x_step_q4,
-                                const int16_t *filter_y, int y_step_q4, int w,
-                                int h, int bd) {
-  // Fixed size intermediate buffer places limits on parameters.
-  DECLARE_ALIGNED(16, uint16_t, temp[MAX_SB_SIZE * MAX_SB_SIZE]);
-  assert(w <= MAX_SB_SIZE);
-  assert(h <= MAX_SB_SIZE);
-
-  aom_highbd_convolve8_c(src, src_stride, CONVERT_TO_BYTEPTR(temp), MAX_SB_SIZE,
-                         filter_x, x_step_q4, filter_y, y_step_q4, w, h, bd);
-  aom_highbd_convolve_avg_c(CONVERT_TO_BYTEPTR(temp), MAX_SB_SIZE, dst,
-                            dst_stride, NULL, 0, NULL, 0, w, h, bd);
-}
-
-void aom_highbd_convolve_copy_c(const uint8_t *src8, ptrdiff_t src_stride,
-                                uint8_t *dst8, ptrdiff_t dst_stride,
-                                const int16_t *filter_x, int filter_x_stride,
-                                const int16_t *filter_y, int filter_y_stride,
-                                int w, int h, int bd) {
-  int r;
-  uint16_t *src = CONVERT_TO_SHORTPTR(src8);
-  uint16_t *dst = CONVERT_TO_SHORTPTR(dst8);
-  (void)filter_x;
-  (void)filter_y;
-  (void)filter_x_stride;
-  (void)filter_y_stride;
-  (void)bd;
-
-  for (r = h; r > 0; --r) {
-    memcpy(dst, src, w * sizeof(uint16_t));
-    src += src_stride;
-    dst += dst_stride;
-  }
-}
-
-void aom_highbd_convolve_avg_c(const uint8_t *src8, ptrdiff_t src_stride,
-                               uint8_t *dst8, ptrdiff_t dst_stride,
-                               const int16_t *filter_x, int filter_x_stride,
-                               const int16_t *filter_y, int filter_y_stride,
-                               int w, int h, int bd) {
-  int x, y;
-  uint16_t *src = CONVERT_TO_SHORTPTR(src8);
-  uint16_t *dst = CONVERT_TO_SHORTPTR(dst8);
-  (void)filter_x;
-  (void)filter_y;
-  (void)filter_x_stride;
-  (void)filter_y_stride;
-  (void)bd;
-
-  for (y = 0; y < h; ++y) {
-    for (x = 0; x < w; ++x) {
-      dst[x] = ROUND_POWER_OF_TWO(dst[x] + src[x], 1);
-    }
-    src += src_stride;
-    dst += dst_stride;
-  }
-}
-#endif

aom_dsp/aom_convolve.h

@@ -1,57 +0,0 @@
-/*
- * Copyright (c) 2016, Alliance for Open Media. All rights reserved
- *
- * This source code is subject to the terms of the BSD 2 Clause License and
- * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
- * was not distributed with this source code in the LICENSE file, you can
- * obtain it at www.aomedia.org/license/software. If the Alliance for Open
- * Media Patent License 1.0 was not distributed with this source code in the
- * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
- */
-#ifndef AOM_DSP_AOM_CONVOLVE_H_
-#define AOM_DSP_AOM_CONVOLVE_H_
-
-#include "./aom_config.h"
-#include "aom/aom_integer.h"
-
-#ifdef __cplusplus
-extern "C" {
-#endif
-
-// Note: Fixed size intermediate buffers, place limits on parameters
-// of some functions. 2d filtering proceeds in 2 steps:
-//   (1) Interpolate horizontally into an intermediate buffer, temp.
-//   (2) Interpolate temp vertically to derive the sub-pixel result.
-// Deriving the maximum number of rows in the temp buffer (135):
-// --Smallest scaling factor is x1/2 ==> y_step_q4 = 32 (Normative).
-// --Largest block size is 64x64 pixels.
-// --64 rows in the downscaled frame span a distance of (64 - 1) * 32 in the
-//   original frame (in 1/16th pixel units).
-// --Must round-up because block may be located at sub-pixel position.
-// --Require an additional SUBPEL_TAPS rows for the 8-tap filter tails.
-// --((64 - 1) * 32 + 15) >> 4 + 8 = 135.
-#if CONFIG_AV1 && CONFIG_EXT_PARTITION
-#define MAX_EXT_SIZE 263
-#else
-#define MAX_EXT_SIZE 135
-#endif  // CONFIG_AV1 && CONFIG_EXT_PARTITION
-
-typedef void (*convolve_fn_t)(const uint8_t *src, ptrdiff_t src_stride,
-                              uint8_t *dst, ptrdiff_t dst_stride,
-                              const int16_t *filter_x, int x_step_q4,
-                              const int16_t *filter_y, int y_step_q4, int w,
-                              int h);
-
-#if CONFIG_AOM_HIGHBITDEPTH
-typedef void (*highbd_convolve_fn_t)(const uint8_t *src, ptrdiff_t src_stride,
-                                     uint8_t *dst, ptrdiff_t dst_stride,
-                                     const int16_t *filter_x, int x_step_q4,
-                                     const int16_t *filter_y, int y_step_q4,
-                                     int w, int h, int bd);
-#endif
-
-#ifdef __cplusplus
-}  // extern "C"
-#endif
-
-#endif  // AOM_DSP_AOM_CONVOLVE_H_

aom_dsp/aom_dsp.mk

@@ -1,426 +0,0 @@
-##
-## Copyright (c) 2016, Alliance for Open Media. All rights reserved
-##
-## This source code is subject to the terms of the BSD 2 Clause License and
-## the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
-## was not distributed with this source code in the LICENSE file, you can
-## obtain it at www.aomedia.org/license/software. If the Alliance for Open
-## Media Patent License 1.0 was not distributed with this source code in the
-## PATENTS file, you can obtain it at www.aomedia.org/license/patent.
-##
-
-
-DSP_SRCS-yes += aom_dsp.mk
-DSP_SRCS-yes += aom_dsp_common.h
-
-DSP_SRCS-$(HAVE_MSA)    += mips/macros_msa.h
-
-DSP_SRCS-$(ARCH_X86)$(ARCH_X86_64)   += x86/synonyms.h
-
-# bit reader
-DSP_SRCS-yes += prob.h
-DSP_SRCS-yes += prob.c
-DSP_SRCS-$(CONFIG_ANS) += ans.h
-DSP_SRCS-$(CONFIG_ANS) += ans.c
-
-ifeq ($(CONFIG_ENCODERS),yes)
-DSP_SRCS-$(CONFIG_ANS) += answriter.h
-DSP_SRCS-yes += bitwriter.h
-DSP_SRCS-yes += dkboolwriter.h
-DSP_SRCS-yes += dkboolwriter.c
-DSP_SRCS-yes += bitwriter_buffer.c
-DSP_SRCS-yes += bitwriter_buffer.h
-DSP_SRCS-yes += psnr.c
-DSP_SRCS-yes += psnr.h
-DSP_SRCS-$(CONFIG_ANS) += buf_ans.h
-DSP_SRCS-$(CONFIG_ANS) += buf_ans.c
-DSP_SRCS-$(CONFIG_INTERNAL_STATS) += ssim.c
-DSP_SRCS-$(CONFIG_INTERNAL_STATS) += ssim.h
-DSP_SRCS-$(CONFIG_INTERNAL_STATS) += psnrhvs.c
-DSP_SRCS-$(CONFIG_INTERNAL_STATS) += fastssim.c
-endif
-
-ifeq ($(CONFIG_DECODERS),yes)
-DSP_SRCS-$(CONFIG_ANS) += ansreader.h
-DSP_SRCS-yes += bitreader.h
-DSP_SRCS-yes += dkboolreader.h
-DSP_SRCS-yes += dkboolreader.c
-DSP_SRCS-yes += bitreader_buffer.c
-DSP_SRCS-yes += bitreader_buffer.h
-endif
-
-# intra predictions
-DSP_SRCS-yes += intrapred.c
-
-ifeq ($(CONFIG_DAALA_EC),yes)
-DSP_SRCS-yes += entenc.c
-DSP_SRCS-yes += entenc.h
-DSP_SRCS-yes += entdec.c
-DSP_SRCS-yes += entdec.h
-DSP_SRCS-yes += entcode.c
-DSP_SRCS-yes += entcode.h
-DSP_SRCS-yes += daalaboolreader.c
-DSP_SRCS-yes += daalaboolreader.h
-DSP_SRCS-yes += daalaboolwriter.c
-DSP_SRCS-yes += daalaboolwriter.h
-endif
-
-DSP_SRCS-$(HAVE_SSE) += x86/intrapred_sse2.asm
-DSP_SRCS-$(HAVE_SSE2) += x86/intrapred_sse2.asm
-DSP_SRCS-$(HAVE_SSSE3) += x86/intrapred_ssse3.asm
-DSP_SRCS-$(HAVE_SSSE3) += x86/aom_subpixel_8t_ssse3.asm
-
-ifeq ($(CONFIG_AOM_HIGHBITDEPTH),yes)
-DSP_SRCS-$(HAVE_SSE)  += x86/highbd_intrapred_sse2.asm
-DSP_SRCS-$(HAVE_SSE2) += x86/highbd_intrapred_sse2.asm
-endif  # CONFIG_AOM_HIGHBITDEPTH
-
-DSP_SRCS-$(HAVE_NEON_ASM) += arm/intrapred_neon_asm$(ASM)
-DSP_SRCS-$(HAVE_NEON) += arm/intrapred_neon.c
-DSP_SRCS-$(HAVE_MSA) += mips/intrapred_msa.c
-DSP_SRCS-$(HAVE_DSPR2)  += mips/intrapred4_dspr2.c
-DSP_SRCS-$(HAVE_DSPR2)  += mips/intrapred8_dspr2.c
-DSP_SRCS-$(HAVE_DSPR2)  += mips/intrapred16_dspr2.c
-
-DSP_SRCS-$(HAVE_DSPR2)  += mips/common_dspr2.h
-DSP_SRCS-$(HAVE_DSPR2)  += mips/common_dspr2.c
-
-# inter predictions
-DSP_SRCS-yes            += blend.h
-DSP_SRCS-yes            += blend_a64_mask.c
-DSP_SRCS-yes            += blend_a64_hmask.c
-DSP_SRCS-yes            += blend_a64_vmask.c
-DSP_SRCS-$(HAVE_SSE4_1) += x86/blend_sse4.h
-DSP_SRCS-$(HAVE_SSE4_1) += x86/blend_a64_mask_sse4.c
-DSP_SRCS-$(HAVE_SSE4_1) += x86/blend_a64_hmask_sse4.c
-DSP_SRCS-$(HAVE_SSE4_1) += x86/blend_a64_vmask_sse4.c
-
-# interpolation filters
-DSP_SRCS-yes += aom_convolve.c
-DSP_SRCS-yes += aom_convolve.h
-DSP_SRCS-yes += aom_filter.h
-
-DSP_SRCS-$(ARCH_X86)$(ARCH_X86_64) += x86/convolve.h
-DSP_SRCS-$(ARCH_X86)$(ARCH_X86_64) += x86/aom_asm_stubs.c
-DSP_SRCS-$(HAVE_SSE2)  += x86/aom_subpixel_8t_sse2.asm
-DSP_SRCS-$(HAVE_SSE2)  += x86/aom_subpixel_bilinear_sse2.asm
-DSP_SRCS-$(HAVE_SSSE3) += x86/aom_subpixel_8t_ssse3.asm
-DSP_SRCS-$(HAVE_SSSE3) += x86/aom_subpixel_bilinear_ssse3.asm
-DSP_SRCS-$(HAVE_AVX2)  += x86/aom_subpixel_8t_intrin_avx2.c
-DSP_SRCS-$(HAVE_SSSE3) += x86/aom_subpixel_8t_intrin_ssse3.c
-ifeq ($(CONFIG_AOM_HIGHBITDEPTH),yes)
-DSP_SRCS-$(HAVE_SSE2)  += x86/aom_high_subpixel_8t_sse2.asm
-DSP_SRCS-$(HAVE_SSE2)  += x86/aom_high_subpixel_bilinear_sse2.asm
-endif
-DSP_SRCS-$(HAVE_SSE2)  += x86/aom_convolve_copy_sse2.asm
-
-ifeq ($(HAVE_NEON_ASM),yes)
-DSP_SRCS-yes += arm/aom_convolve_copy_neon_asm$(ASM)
-DSP_SRCS-yes += arm/aom_convolve8_avg_neon_asm$(ASM)
-DSP_SRCS-yes += arm/aom_convolve8_neon_asm$(ASM)
-DSP_SRCS-yes += arm/aom_convolve_avg_neon_asm$(ASM)
-DSP_SRCS-yes += arm/aom_convolve_neon.c
-else
-ifeq ($(HAVE_NEON),yes)
-DSP_SRCS-yes += arm/aom_convolve_copy_neon.c
-DSP_SRCS-yes += arm/aom_convolve8_avg_neon.c
-DSP_SRCS-yes += arm/aom_convolve8_neon.c
-DSP_SRCS-yes += arm/aom_convolve_avg_neon.c
-DSP_SRCS-yes += arm/aom_convolve_neon.c
-endif  # HAVE_NEON
-endif  # HAVE_NEON_ASM
-
-# common (msa)
-DSP_SRCS-$(HAVE_MSA) += mips/aom_convolve8_avg_horiz_msa.c
-DSP_SRCS-$(HAVE_MSA) += mips/aom_convolve8_avg_msa.c
-DSP_SRCS-$(HAVE_MSA) += mips/aom_convolve8_avg_vert_msa.c
-DSP_SRCS-$(HAVE_MSA) += mips/aom_convolve8_horiz_msa.c
-DSP_SRCS-$(HAVE_MSA) += mips/aom_convolve8_msa.c
-DSP_SRCS-$(HAVE_MSA) += mips/aom_convolve8_vert_msa.c
-DSP_SRCS-$(HAVE_MSA) += mips/aom_convolve_avg_msa.c
-DSP_SRCS-$(HAVE_MSA) += mips/aom_convolve_copy_msa.c
-DSP_SRCS-$(HAVE_MSA) += mips/aom_convolve_msa.h
-
-# common (dspr2)
-DSP_SRCS-$(HAVE_DSPR2)  += mips/convolve_common_dspr2.h
-DSP_SRCS-$(HAVE_DSPR2)  += mips/convolve2_avg_dspr2.c
-DSP_SRCS-$(HAVE_DSPR2)  += mips/convolve2_avg_horiz_dspr2.c
-DSP_SRCS-$(HAVE_DSPR2)  += mips/convolve2_dspr2.c
-DSP_SRCS-$(HAVE_DSPR2)  += mips/convolve2_horiz_dspr2.c
-DSP_SRCS-$(HAVE_DSPR2)  += mips/convolve2_vert_dspr2.c
-DSP_SRCS-$(HAVE_DSPR2)  += mips/convolve8_avg_dspr2.c
-DSP_SRCS-$(HAVE_DSPR2)  += mips/convolve8_avg_horiz_dspr2.c
-DSP_SRCS-$(HAVE_DSPR2)  += mips/convolve8_dspr2.c
-DSP_SRCS-$(HAVE_DSPR2)  += mips/convolve8_horiz_dspr2.c
-DSP_SRCS-$(HAVE_DSPR2)  += mips/convolve8_vert_dspr2.c
-
-# loop filters
-DSP_SRCS-yes += loopfilter.c
-
-DSP_SRCS-$(ARCH_X86)$(ARCH_X86_64)   += x86/loopfilter_sse2.c
-DSP_SRCS-$(HAVE_AVX2)                += x86/loopfilter_avx2.c
-
-DSP_SRCS-$(HAVE_NEON)   += arm/loopfilter_neon.c
-ifeq ($(HAVE_NEON_ASM),yes)
-DSP_SRCS-yes  += arm/loopfilter_mb_neon$(ASM)
-DSP_SRCS-yes  += arm/loopfilter_16_neon$(ASM)
-DSP_SRCS-yes  += arm/loopfilter_8_neon$(ASM)
-DSP_SRCS-yes  += arm/loopfilter_4_neon$(ASM)
-else
-ifeq ($(HAVE_NEON),yes)
-DSP_SRCS-yes   += arm/loopfilter_16_neon.c
-DSP_SRCS-yes   += arm/loopfilter_8_neon.c
-DSP_SRCS-yes   += arm/loopfilter_4_neon.c
-endif  # HAVE_NEON
-endif  # HAVE_NEON_ASM
-
-DSP_SRCS-$(HAVE_MSA)    += mips/loopfilter_msa.h
-DSP_SRCS-$(HAVE_MSA)    += mips/loopfilter_16_msa.c
-DSP_SRCS-$(HAVE_MSA)    += mips/loopfilter_8_msa.c
-DSP_SRCS-$(HAVE_MSA)    += mips/loopfilter_4_msa.c
-DSP_SRCS-$(HAVE_DSPR2)  += mips/loopfilter_filters_dspr2.h
-DSP_SRCS-$(HAVE_DSPR2)  += mips/loopfilter_filters_dspr2.c
-DSP_SRCS-$(HAVE_DSPR2)  += mips/loopfilter_macros_dspr2.h
-DSP_SRCS-$(HAVE_DSPR2)  += mips/loopfilter_masks_dspr2.h
-DSP_SRCS-$(HAVE_DSPR2)  += mips/loopfilter_mb_dspr2.c
-DSP_SRCS-$(HAVE_DSPR2)  += mips/loopfilter_mb_horiz_dspr2.c
-DSP_SRCS-$(HAVE_DSPR2)  += mips/loopfilter_mb_vert_dspr2.c
-
-ifeq ($(CONFIG_AOM_HIGHBITDEPTH),yes)
-DSP_SRCS-$(HAVE_SSE2)   += x86/highbd_loopfilter_sse2.c
-endif  # CONFIG_AOM_HIGHBITDEPTH
-
-DSP_SRCS-yes            += txfm_common.h
-DSP_SRCS-yes            += x86/txfm_common_intrin.h
-DSP_SRCS-$(HAVE_SSE2)   += x86/txfm_common_sse2.h
-DSP_SRCS-$(HAVE_MSA)    += mips/txfm_macros_msa.h
-# forward transform
-ifeq ($(CONFIG_AV1),yes)
-DSP_SRCS-yes            += fwd_txfm.c
-DSP_SRCS-yes            += fwd_txfm.h
-DSP_SRCS-$(HAVE_SSE2)   += x86/fwd_txfm_sse2.h
-DSP_SRCS-$(HAVE_SSE2)   += x86/fwd_txfm_sse2.c
-DSP_SRCS-$(HAVE_SSE2)   += x86/fwd_dct32_8cols_sse2.c
-DSP_SRCS-$(HAVE_SSE2)   += x86/fwd_txfm_impl_sse2.h
-DSP_SRCS-$(HAVE_SSE2)   += x86/fwd_dct32x32_impl_sse2.h
-ifeq ($(ARCH_X86_64),yes)
-DSP_SRCS-$(HAVE_SSSE3)  += x86/fwd_txfm_ssse3_x86_64.asm
-endif
-DSP_SRCS-$(HAVE_AVX2)   += x86/fwd_txfm_avx2.h
-DSP_SRCS-$(HAVE_AVX2)   += x86/fwd_txfm_avx2.c
-DSP_SRCS-$(HAVE_AVX2)   += x86/txfm_common_avx2.h
-DSP_SRCS-$(HAVE_AVX2)   += x86/fwd_dct32x32_impl_avx2.h
-DSP_SRCS-$(HAVE_NEON)   += arm/fwd_txfm_neon.c
-DSP_SRCS-$(HAVE_MSA)    += mips/fwd_txfm_msa.h
-DSP_SRCS-$(HAVE_MSA)    += mips/fwd_txfm_msa.c
-DSP_SRCS-$(HAVE_MSA)    += mips/fwd_dct32x32_msa.c
-endif  # CONFIG_AV1_ENCODER
-
-ifeq ($(CONFIG_PVQ),yes)
-DSP_SRCS-yes            += fwd_txfm.c
-DSP_SRCS-yes            += fwd_txfm.h
-DSP_SRCS-$(HAVE_SSE2)   += x86/fwd_txfm_sse2.h
-DSP_SRCS-$(HAVE_SSE2)   += x86/fwd_txfm_sse2.c
-DSP_SRCS-$(HAVE_SSE2)   += x86/fwd_txfm_impl_sse2.h
-DSP_SRCS-$(HAVE_SSE2)   += x86/fwd_dct32x32_impl_sse2.h
-ifeq ($(ARCH_X86_64),yes)
-DSP_SRCS-$(HAVE_SSSE3)  += x86/fwd_txfm_ssse3_x86_64.asm
-endif
-DSP_SRCS-$(HAVE_AVX2)   += x86/fwd_txfm_avx2.c
-DSP_SRCS-$(HAVE_AVX2)   += x86/fwd_dct32x32_impl_avx2.h
-DSP_SRCS-$(HAVE_NEON)   += arm/fwd_txfm_neon.c
-DSP_SRCS-$(HAVE_MSA)    += mips/fwd_txfm_msa.h
-DSP_SRCS-$(HAVE_MSA)    += mips/fwd_txfm_msa.c
-DSP_SRCS-$(HAVE_MSA)    += mips/fwd_dct32x32_msa.c
-endif  # CONFIG_PVQ
-
-# inverse transform
-ifeq ($(CONFIG_AV1), yes)
-DSP_SRCS-yes            += inv_txfm.h
-DSP_SRCS-yes            += inv_txfm.c
-DSP_SRCS-$(HAVE_SSE2)   += x86/inv_txfm_sse2.h
-DSP_SRCS-$(HAVE_SSE2)   += x86/inv_txfm_sse2.c
-DSP_SRCS-$(HAVE_SSE2)   += x86/inv_wht_sse2.asm
-ifeq ($(ARCH_X86_64),yes)
-DSP_SRCS-$(HAVE_SSSE3)  += x86/inv_txfm_ssse3_x86_64.asm
-endif  # ARCH_X86_64
-
-ifeq ($(HAVE_NEON_ASM),yes)
-DSP_SRCS-yes  += arm/save_reg_neon$(ASM)
-DSP_SRCS-yes  += arm/idct4x4_1_add_neon$(ASM)
-DSP_SRCS-yes  += arm/idct4x4_add_neon$(ASM)
-DSP_SRCS-yes  += arm/idct8x8_1_add_neon$(ASM)
-DSP_SRCS-yes  += arm/idct8x8_add_neon$(ASM)
-DSP_SRCS-yes  += arm/idct16x16_1_add_neon$(ASM)
-DSP_SRCS-yes  += arm/idct16x16_add_neon$(ASM)
-DSP_SRCS-yes  += arm/idct32x32_1_add_neon$(ASM)
-DSP_SRCS-yes  += arm/idct32x32_add_neon$(ASM)
-else
-ifeq ($(HAVE_NEON),yes)
-DSP_SRCS-yes  += arm/idct4x4_1_add_neon.c
-DSP_SRCS-yes  += arm/idct4x4_add_neon.c
-DSP_SRCS-yes  += arm/idct8x8_1_add_neon.c
-DSP_SRCS-yes  += arm/idct8x8_add_neon.c
-DSP_SRCS-yes  += arm/idct16x16_1_add_neon.c
-DSP_SRCS-yes  += arm/idct16x16_add_neon.c
-DSP_SRCS-yes  += arm/idct32x32_1_add_neon.c
-DSP_SRCS-yes  += arm/idct32x32_add_neon.c
-endif  # HAVE_NEON
-endif  # HAVE_NEON_ASM
-DSP_SRCS-$(HAVE_NEON)  += arm/idct16x16_neon.c
-
-DSP_SRCS-$(HAVE_MSA)   += mips/inv_txfm_msa.h
-DSP_SRCS-$(HAVE_MSA)   += mips/idct4x4_msa.c
-DSP_SRCS-$(HAVE_MSA)   += mips/idct8x8_msa.c
-DSP_SRCS-$(HAVE_MSA)   += mips/idct16x16_msa.c
-DSP_SRCS-$(HAVE_MSA)   += mips/idct32x32_msa.c
-
-ifneq ($(CONFIG_AOM_HIGHBITDEPTH),yes)
-DSP_SRCS-$(HAVE_DSPR2) += mips/inv_txfm_dspr2.h
-DSP_SRCS-$(HAVE_DSPR2) += mips/itrans4_dspr2.c
-DSP_SRCS-$(HAVE_DSPR2) += mips/itrans8_dspr2.c
-DSP_SRCS-$(HAVE_DSPR2) += mips/itrans16_dspr2.c
-DSP_SRCS-$(HAVE_DSPR2) += mips/itrans32_dspr2.c
-DSP_SRCS-$(HAVE_DSPR2) += mips/itrans32_cols_dspr2.c
-endif  # CONFIG_AOM_HIGHBITDEPTH
-endif  # CONFIG_AV1
-
-# quantization
-ifneq ($(filter yes,$(CONFIG_AV1_ENCODER)),)
-DSP_SRCS-yes            += quantize.c
-DSP_SRCS-yes            += quantize.h
-
-DSP_SRCS-$(HAVE_SSE2)   += x86/quantize_sse2.c
-ifeq ($(CONFIG_AOM_HIGHBITDEPTH),yes)
-DSP_SRCS-$(HAVE_SSE2)   += x86/highbd_quantize_intrin_sse2.c
-endif
-ifeq ($(ARCH_X86_64),yes)
-DSP_SRCS-$(HAVE_SSSE3)  += x86/quantize_ssse3_x86_64.asm
-DSP_SRCS-$(HAVE_AVX)    += x86/quantize_avx_x86_64.asm
-endif
-
-# avg
-DSP_SRCS-yes           += avg.c
-DSP_SRCS-$(HAVE_SSE2)  += x86/avg_intrin_sse2.c
-DSP_SRCS-$(HAVE_NEON)  += arm/avg_neon.c
-DSP_SRCS-$(HAVE_MSA)   += mips/avg_msa.c
-DSP_SRCS-$(HAVE_NEON)  += arm/hadamard_neon.c
-ifeq ($(ARCH_X86_64),yes)
-DSP_SRCS-$(HAVE_SSSE3) += x86/avg_ssse3_x86_64.asm
-endif
-
-# high bit depth subtract
-ifeq ($(CONFIG_AOM_HIGHBITDEPTH),yes)
-DSP_SRCS-$(HAVE_SSE2)  += x86/highbd_subtract_sse2.c
-endif
-
-endif  # CONFIG_AV1_ENCODER
-
-ifeq ($(CONFIG_AV1_ENCODER),yes)
-DSP_SRCS-yes            += sum_squares.c
-
-DSP_SRCS-$(HAVE_SSE2)   += x86/sum_squares_sse2.c
-endif # CONFIG_AV1_ENCODER
-
-ifeq ($(CONFIG_ENCODERS),yes)
-DSP_SRCS-yes            += sad.c
-DSP_SRCS-yes            += subtract.c
-
-DSP_SRCS-$(HAVE_MEDIA)  += arm/sad_media$(ASM)
-DSP_SRCS-$(HAVE_NEON)   += arm/sad4d_neon.c
-DSP_SRCS-$(HAVE_NEON)   += arm/sad_neon.c
-DSP_SRCS-$(HAVE_NEON)   += arm/subtract_neon.c
-
-DSP_SRCS-$(HAVE_MSA)    += mips/sad_msa.c
-DSP_SRCS-$(HAVE_MSA)    += mips/subtract_msa.c
-
-DSP_SRCS-$(HAVE_SSE3)   += x86/sad_sse3.asm
-DSP_SRCS-$(HAVE_SSSE3)  += x86/sad_ssse3.asm
-DSP_SRCS-$(HAVE_SSE4_1) += x86/sad_sse4.asm
-DSP_SRCS-$(HAVE_AVX2)   += x86/sad4d_avx2.c
-DSP_SRCS-$(HAVE_AVX2)   += x86/sad_avx2.c
-
-ifeq ($(CONFIG_AV1_ENCODER),yes)
-ifeq ($(CONFIG_EXT_INTER),yes)
-DSP_SRCS-$(HAVE_SSSE3)  += x86/masked_sad_intrin_ssse3.c
-DSP_SRCS-$(HAVE_SSSE3)  += x86/masked_variance_intrin_ssse3.c
-endif  #CONFIG_EXT_INTER
-ifeq ($(CONFIG_MOTION_VAR),yes)
-DSP_SRCS-$(HAVE_SSE4_1) += x86/obmc_sad_sse4.c
-DSP_SRCS-$(HAVE_SSE4_1) += x86/obmc_variance_sse4.c
-endif  #CONFIG_MOTION_VAR
-endif  #CONFIG_AV1_ENCODER
-
-DSP_SRCS-$(HAVE_SSE)    += x86/sad4d_sse2.asm
-DSP_SRCS-$(HAVE_SSE)    += x86/sad_sse2.asm
-DSP_SRCS-$(HAVE_SSE2)   += x86/sad4d_sse2.asm
-DSP_SRCS-$(HAVE_SSE2)   += x86/sad_sse2.asm
-DSP_SRCS-$(HAVE_SSE2)   += x86/subtract_sse2.asm
-
-ifeq ($(CONFIG_AOM_HIGHBITDEPTH),yes)
-DSP_SRCS-$(HAVE_SSE2) += x86/highbd_sad4d_sse2.asm
-DSP_SRCS-$(HAVE_SSE2) += x86/highbd_sad_sse2.asm
-endif  # CONFIG_AOM_HIGHBITDEPTH
-
-endif  # CONFIG_ENCODERS
-
-ifneq ($(filter yes,$(CONFIG_ENCODERS)),)
-DSP_SRCS-yes            += variance.c
-DSP_SRCS-yes            += variance.h
-
-DSP_SRCS-$(HAVE_MEDIA)  += arm/bilinear_filter_media$(ASM)
-DSP_SRCS-$(HAVE_MEDIA)  += arm/subpel_variance_media.c
-DSP_SRCS-$(HAVE_MEDIA)  += arm/variance_halfpixvar16x16_h_media$(ASM)
-DSP_SRCS-$(HAVE_MEDIA)  += arm/variance_halfpixvar16x16_hv_media$(ASM)
-DSP_SRCS-$(HAVE_MEDIA)  += arm/variance_halfpixvar16x16_v_media$(ASM)
-DSP_SRCS-$(HAVE_MEDIA)  += arm/variance_media$(ASM)
-DSP_SRCS-$(HAVE_NEON)   += arm/subpel_variance_neon.c
-DSP_SRCS-$(HAVE_NEON)   += arm/variance_neon.c
-
-DSP_SRCS-$(HAVE_MSA)    += mips/variance_msa.c
-DSP_SRCS-$(HAVE_MSA)    += mips/sub_pixel_variance_msa.c
-
-DSP_SRCS-$(HAVE_SSE)    += x86/variance_sse2.c
-DSP_SRCS-$(HAVE_SSE2)   += x86/variance_sse2.c  # Contains SSE2 and SSSE3
-DSP_SRCS-$(HAVE_SSE2)   += x86/halfpix_variance_sse2.c
-DSP_SRCS-$(HAVE_SSE2)   += x86/halfpix_variance_impl_sse2.asm
-DSP_SRCS-$(HAVE_AVX2)   += x86/variance_avx2.c
-DSP_SRCS-$(HAVE_AVX2)   += x86/variance_impl_avx2.c
-
-ifeq ($(ARCH_X86_64),yes)
-DSP_SRCS-$(HAVE_SSE2)   += x86/ssim_opt_x86_64.asm
-endif  # ARCH_X86_64
-
-DSP_SRCS-$(HAVE_SSE)    += x86/subpel_variance_sse2.asm
-DSP_SRCS-$(HAVE_SSE2)   += x86/subpel_variance_sse2.asm  # Contains SSE2 and SSSE3
-
-ifeq ($(CONFIG_AOM_HIGHBITDEPTH),yes)
-DSP_SRCS-$(HAVE_SSE2)   += x86/highbd_variance_sse2.c
-DSP_SRCS-$(HAVE_SSE4_1) += x86/highbd_variance_sse4.c
-DSP_SRCS-$(HAVE_SSE2)   += x86/highbd_variance_impl_sse2.asm
-DSP_SRCS-$(HAVE_SSE2)   += x86/highbd_subpel_variance_impl_sse2.asm
-endif  # CONFIG_AOM_HIGHBITDEPTH
-endif  # CONFIG_ENCODERS
-
-DSP_SRCS-no += $(DSP_SRCS_REMOVE-yes)
-
-DSP_SRCS-yes += aom_dsp_rtcd.c
-DSP_SRCS-yes += aom_dsp_rtcd_defs.pl
-
-DSP_SRCS-yes += aom_simd.c
-DSP_SRCS-yes += aom_simd.h
-DSP_SRCS-yes += aom_simd_inline.h
-DSP_SRCS-yes += simd/v64_intrinsics.h
-DSP_SRCS-yes += simd/v64_intrinsics_c.h
-DSP_SRCS-yes += simd/v128_intrinsics.h
-DSP_SRCS-yes += simd/v128_intrinsics_c.h
-DSP_SRCS-yes += simd/v256_intrinsics.h
-DSP_SRCS-yes += simd/v256_intrinsics_c.h
-DSP_SRCS-$(HAVE_SSE2) += simd/v64_intrinsics_x86.h
-DSP_SRCS-$(HAVE_SSE2) += simd/v128_intrinsics_x86.h
-DSP_SRCS-$(HAVE_SSE2) += simd/v256_intrinsics_x86.h
-DSP_SRCS-$(HAVE_NEON) += simd/v64_intrinsics_arm.h
-DSP_SRCS-$(HAVE_NEON) += simd/v128_intrinsics_arm.h
-DSP_SRCS-$(HAVE_NEON) += simd/v256_intrinsics_arm.h
-
-$(eval $(call rtcd_h_template,aom_dsp_rtcd,aom_dsp/aom_dsp_rtcd_defs.pl))

aom_dsp/aom_dsp_common.h

@@ -1,102 +0,0 @@
-/*
- * Copyright (c) 2016, Alliance for Open Media. All rights reserved
- *
- * This source code is subject to the terms of the BSD 2 Clause License and
- * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
- * was not distributed with this source code in the LICENSE file, you can
- * obtain it at www.aomedia.org/license/software. If the Alliance for Open
- * Media Patent License 1.0 was not distributed with this source code in the
- * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
- */
-
-#ifndef AOM_DSP_AOM_DSP_COMMON_H_
-#define AOM_DSP_AOM_DSP_COMMON_H_
-
-#include "./aom_config.h"
-#include "aom/aom_integer.h"
-#include "aom_ports/mem.h"
-
-#ifdef __cplusplus
-extern "C" {
-#endif
-
-#ifndef MAX_SB_SIZE
-#if CONFIG_AV1 && CONFIG_EXT_PARTITION
-#define MAX_SB_SIZE 128
-#else
-#define MAX_SB_SIZE 64
-#endif  // CONFIG_AV1 && CONFIG_EXT_PARTITION
-#endif  // ndef MAX_SB_SIZE
-
-#define AOMMIN(x, y) (((x) < (y)) ? (x) : (y))
-#define AOMMAX(x, y) (((x) > (y)) ? (x) : (y))
-
-#define IMPLIES(a, b) (!(a) || (b))  //  Logical 'a implies b' (or 'a -> b')
-
-#define IS_POWER_OF_TWO(x) (((x) & ((x)-1)) == 0)
-
-// These can be used to give a hint about branch outcomes.
-// This can have an effect, even if your target processor has a
-// good branch predictor, as these hints can affect basic block
-// ordering by the compiler.
-#ifdef __GNUC__
-#define LIKELY(v) __builtin_expect(v, 1)
-#define UNLIKELY(v) __builtin_expect(v, 0)
-#else
-#define LIKELY(v) (v)
-#define UNLIKELY(v) (v)
-#endif
-
-#define AOM_SWAP(type, a, b) \
-  do {                       \
-    type c = (b);            \
-    b = a;                   \
-    a = c;                   \
-  } while (0)
-
-#if CONFIG_AOM_QM
-typedef uint16_t qm_val_t;
-#define AOM_QM_BITS 6
-#endif
-#if CONFIG_AOM_HIGHBITDEPTH
-// Note:
-// tran_low_t  is the datatype used for final transform coefficients.
-// tran_high_t is the datatype used for intermediate transform stages.
-typedef int64_t tran_high_t;
-typedef int32_t tran_low_t;
-#else
-// Note:
-// tran_low_t  is the datatype used for final transform coefficients.
-// tran_high_t is the datatype used for intermediate transform stages.
-typedef int32_t tran_high_t;
-typedef int16_t tran_low_t;
-#endif  // CONFIG_AOM_HIGHBITDEPTH
-
-static INLINE uint8_t clip_pixel(int val) {
-  return (val > 255) ? 255 : (val < 0) ? 0 : val;
-}
-
-static INLINE int clamp(int value, int low, int high) {
-  return value < low ? low : (value > high ? high : value);
-}
-
-static INLINE double fclamp(double value, double low, double high) {
-  return value < low ? low : (value > high ? high : value);
-}
-
-#if CONFIG_AOM_HIGHBITDEPTH
-static INLINE uint16_t clip_pixel_highbd(int val, int bd) {
-  switch (bd) {
-    case 8:
-    default: return (uint16_t)clamp(val, 0, 255);
-    case 10: return (uint16_t)clamp(val, 0, 1023);
-    case 12: return (uint16_t)clamp(val, 0, 4095);
-  }
-}
-#endif  // CONFIG_AOM_HIGHBITDEPTH
-
-#ifdef __cplusplus
-}  // extern "C"
-#endif
-
-#endif  // AOM_DSP_AOM_DSP_COMMON_H_

aom_dsp/aom_dsp_rtcd.c

@@ -1,16 +0,0 @@
-/*
- * Copyright (c) 2016, Alliance for Open Media. All rights reserved
- *
- * This source code is subject to the terms of the BSD 2 Clause License and
- * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
- * was not distributed with this source code in the LICENSE file, you can
- * obtain it at www.aomedia.org/license/software. If the Alliance for Open
- * Media Patent License 1.0 was not distributed with this source code in the
- * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
- */
-#include "./aom_config.h"
-#define RTCD_C
-#include "./aom_dsp_rtcd.h"
-#include "aom_ports/aom_once.h"
-
-void aom_dsp_rtcd() { once(setup_rtcd_internal); }

aom_dsp/aom_filter.h

@@ -1,43 +0,0 @@
-/*
- * Copyright (c) 2016, Alliance for Open Media. All rights reserved
- *
- * This source code is subject to the terms of the BSD 2 Clause License and
- * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
- * was not distributed with this source code in the LICENSE file, you can
- * obtain it at www.aomedia.org/license/software. If the Alliance for Open
- * Media Patent License 1.0 was not distributed with this source code in the
- * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
- */
-
-#ifndef AOM_DSP_AOM_FILTER_H_
-#define AOM_DSP_AOM_FILTER_H_
-
-#include "aom/aom_integer.h"
-
-#ifdef __cplusplus
-extern "C" {
-#endif
-
-#define FILTER_BITS 7
-
-#define SUBPEL_BITS 4
-#define SUBPEL_MASK ((1 << SUBPEL_BITS) - 1)
-#define SUBPEL_SHIFTS (1 << SUBPEL_BITS)
-#define SUBPEL_TAPS 8
-
-typedef int16_t InterpKernel[SUBPEL_TAPS];
-
-#define BIL_SUBPEL_BITS 3
-#define BIL_SUBPEL_SHIFTS (1 << BIL_SUBPEL_BITS)
-
-// 2 tap bilinear filters
-static const uint8_t bilinear_filters_2t[BIL_SUBPEL_SHIFTS][2] = {
-  { 128, 0 }, { 112, 16 }, { 96, 32 }, { 80, 48 },
-  { 64, 64 }, { 48, 80 },  { 32, 96 }, { 16, 112 },
-};
-
-#ifdef __cplusplus
-}  // extern "C"
-#endif
-
-#endif  // AOM_DSP_AOM_FILTER_H_

aom_dsp/aom_simd.c

@@ -1,13 +0,0 @@
-/*
- * Copyright (c) 2016, Alliance for Open Media. All rights reserved
- *
- * This source code is subject to the terms of the BSD 2 Clause License and
- * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
- * was not distributed with this source code in the LICENSE file, you can
- * obtain it at www.aomedia.org/license/software. If the Alliance for Open
- * Media Patent License 1.0 was not distributed with this source code in the
- * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
- */
-
-// Set to 1 to add some sanity checks in the fallback C code
-const int simd_check = 1;

aom_dsp/aom_simd.h

@@ -1,32 +0,0 @@
-/*
- * Copyright (c) 2016, Alliance for Open Media. All rights reserved
- *
- * This source code is subject to the terms of the BSD 2 Clause License and
- * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
- * was not distributed with this source code in the LICENSE file, you can
- * obtain it at www.aomedia.org/license/software. If the Alliance for Open
- * Media Patent License 1.0 was not distributed with this source code in the
- * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
- */
-
-#ifndef AOM_DSP_AOM_AOM_SIMD_H_
-#define AOM_DSP_AOM_AOM_SIMD_H_
-
-#include <stdint.h>
-
-#if defined(_WIN32)
-#include <intrin.h>
-#endif
-
-#include "./aom_config.h"
-#include "./aom_simd_inline.h"
-
-#if HAVE_NEON
-#include "simd/v256_intrinsics_arm.h"
-#elif HAVE_SSE2
-#include "simd/v256_intrinsics_x86.h"
-#else
-#include "simd/v256_intrinsics.h"
-#endif
-
-#endif  // AOM_DSP_AOM_AOM_SIMD_H_

aom_dsp/aom_simd_inline.h

@@ -1,21 +0,0 @@
-/*
- * Copyright (c) 2016, Alliance for Open Media. All rights reserved
- *
- * This source code is subject to the terms of the BSD 2 Clause License and
- * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
- * was not distributed with this source code in the LICENSE file, you can
- * obtain it at www.aomedia.org/license/software. If the Alliance for Open
- * Media Patent License 1.0 was not distributed with this source code in the
- * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
- */
-
-#ifndef AOM_DSP_AOM_SIMD_INLINE_H_
-#define AOM_DSP_AOM_SIMD_INLINE_H_
-
-#include "aom/aom_integer.h"
-
-#ifndef SIMD_INLINE
-#define SIMD_INLINE static AOM_FORCE_INLINE
-#endif
-
-#endif  // AOM_DSP_AOM_SIMD_INLINE_H_

aom_dsp/arm/aom_convolve8_avg_neon.c

@@ -1,364 +0,0 @@
-/*
- * Copyright (c) 2016, Alliance for Open Media. All rights reserved
- *
- * This source code is subject to the terms of the BSD 2 Clause License and
- * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
- * was not distributed with this source code in the LICENSE file, you can
- * obtain it at www.aomedia.org/license/software. If the Alliance for Open
- * Media Patent License 1.0 was not distributed with this source code in the
- * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
- */
-
-#include <arm_neon.h>
-#include <assert.h>
-
-#include "./aom_config.h"
-#include "./aom_dsp_rtcd.h"
-#include "aom/aom_integer.h"
-#include "aom_ports/mem.h"
-
-static INLINE int32x4_t MULTIPLY_BY_Q0(int16x4_t dsrc0, int16x4_t dsrc1,
-                                       int16x4_t dsrc2, int16x4_t dsrc3,
-                                       int16x4_t dsrc4, int16x4_t dsrc5,
-                                       int16x4_t dsrc6, int16x4_t dsrc7,
-                                       int16x8_t q0s16) {
-  int32x4_t qdst;
-  int16x4_t d0s16, d1s16;
-
-  d0s16 = vget_low_s16(q0s16);
-  d1s16 = vget_high_s16(q0s16);
-
-  qdst = vmull_lane_s16(dsrc0, d0s16, 0);
-  qdst = vmlal_lane_s16(qdst, dsrc1, d0s16, 1);
-  qdst = vmlal_lane_s16(qdst, dsrc2, d0s16, 2);
-  qdst = vmlal_lane_s16(qdst, dsrc3, d0s16, 3);
-  qdst = vmlal_lane_s16(qdst, dsrc4, d1s16, 0);
-  qdst = vmlal_lane_s16(qdst, dsrc5, d1s16, 1);
-  qdst = vmlal_lane_s16(qdst, dsrc6, d1s16, 2);
-  qdst = vmlal_lane_s16(qdst, dsrc7, d1s16, 3);
-  return qdst;
-}
-
-void aom_convolve8_avg_horiz_neon(const uint8_t *src, ptrdiff_t src_stride,
-                                  uint8_t *dst, ptrdiff_t dst_stride,
-                                  const int16_t *filter_x, int x_step_q4,
-                                  const int16_t *filter_y,  // unused
-                                  int y_step_q4,            // unused
-                                  int w, int h) {
-  int width;
-  const uint8_t *s;
-  uint8_t *d;
-  uint8x8_t d2u8, d3u8, d24u8, d25u8, d26u8, d27u8, d28u8, d29u8;
-  uint32x2_t d2u32, d3u32, d6u32, d7u32, d28u32, d29u32, d30u32, d31u32;
-  uint8x16_t q1u8, q3u8, q12u8, q13u8, q14u8, q15u8;
-  int16x4_t d16s16, d17s16, d18s16, d19s16, d20s16, d22s16, d23s16;
-  int16x4_t d24s16, d25s16, d26s16, d27s16;
-  uint16x4_t d2u16, d3u16, d4u16, d5u16, d16u16, d17u16, d18u16, d19u16;
-  int16x8_t q0s16;
-  uint16x8_t q1u16, q2u16, q8u16, q9u16, q10u16, q11u16, q12u16, q13u16;
-  int32x4_t q1s32, q2s32, q14s32, q15s32;
-  uint16x8x2_t q0x2u16;
-  uint8x8x2_t d0x2u8, d1x2u8;
-  uint32x2x2_t d0x2u32;
-  uint16x4x2_t d0x2u16, d1x2u16;
-  uint32x4x2_t q0x2u32;
-
-  assert(x_step_q4 == 16);
-
-  (void)x_step_q4;
-  (void)y_step_q4;
-  (void)filter_y;
-
-  q0s16 = vld1q_s16(filter_x);
-
-  src -= 3;                // adjust for taps
-  for (; h > 0; h -= 4) {  // loop_horiz_v
-    s = src;
-    d24u8 = vld1_u8(s);
-    s += src_stride;
-    d25u8 = vld1_u8(s);
-    s += src_stride;
-    d26u8 = vld1_u8(s);
-    s += src_stride;
-    d27u8 = vld1_u8(s);
-
-    q12u8 = vcombine_u8(d24u8, d25u8);
-    q13u8 = vcombine_u8(d26u8, d27u8);
-
-    q0x2u16 =
-        vtrnq_u16(vreinterpretq_u16_u8(q12u8), vreinterpretq_u16_u8(q13u8));
-    d24u8 = vreinterpret_u8_u16(vget_low_u16(q0x2u16.val[0]));
-    d25u8 = vreinterpret_u8_u16(vget_high_u16(q0x2u16.val[0]));
-    d26u8 = vreinterpret_u8_u16(vget_low_u16(q0x2u16.val[1]));
-    d27u8 = vreinterpret_u8_u16(vget_high_u16(q0x2u16.val[1]));
-    d0x2u8 = vtrn_u8(d24u8, d25u8);
-    d1x2u8 = vtrn_u8(d26u8, d27u8);
-
-    __builtin_prefetch(src + src_stride * 4);
-    __builtin_prefetch(src + src_stride * 5);
-
-    q8u16 = vmovl_u8(d0x2u8.val[0]);
-    q9u16 = vmovl_u8(d0x2u8.val[1]);
-    q10u16 = vmovl_u8(d1x2u8.val[0]);
-    q11u16 = vmovl_u8(d1x2u8.val[1]);
-
-    src += 7;
-    d16u16 = vget_low_u16(q8u16);
-    d17u16 = vget_high_u16(q8u16);
-    d18u16 = vget_low_u16(q9u16);
-    d19u16 = vget_high_u16(q9u16);
-    q8u16 = vcombine_u16(d16u16, d18u16);  // vswp 17 18
-    q9u16 = vcombine_u16(d17u16, d19u16);
-
-    d20s16 = vreinterpret_s16_u16(vget_low_u16(q10u16));
-    d23s16 = vreinterpret_s16_u16(vget_high_u16(q10u16));         // vmov 23 21
-    for (width = w; width > 0; width -= 4, src += 4, dst += 4) {  // loop_horiz
-      s = src;
-      d28u32 = vld1_dup_u32((const uint32_t *)s);
-      s += src_stride;
-      d29u32 = vld1_dup_u32((const uint32_t *)s);
-      s += src_stride;
-      d31u32 = vld1_dup_u32((const uint32_t *)s);
-      s += src_stride;
-      d30u32 = vld1_dup_u32((const uint32_t *)s);
-
-      __builtin_prefetch(src + 64);
-
-      d0x2u16 =
-          vtrn_u16(vreinterpret_u16_u32(d28u32), vreinterpret_u16_u32(d31u32));
-      d1x2u16 =
-          vtrn_u16(vreinterpret_u16_u32(d29u32), vreinterpret_u16_u32(d30u32));
-      d0x2u8 = vtrn_u8(vreinterpret_u8_u16(d0x2u16.val[0]),   // d28
-                       vreinterpret_u8_u16(d1x2u16.val[0]));  // d29
-      d1x2u8 = vtrn_u8(vreinterpret_u8_u16(d0x2u16.val[1]),   // d31
-                       vreinterpret_u8_u16(d1x2u16.val[1]));  // d30
-
-      __builtin_prefetch(src + 64 + src_stride);
-
-      q14u8 = vcombine_u8(d0x2u8.val[0], d0x2u8.val[1]);
-      q15u8 = vcombine_u8(d1x2u8.val[1], d1x2u8.val[0]);
-      q0x2u32 =
-          vtrnq_u32(vreinterpretq_u32_u8(q14u8), vreinterpretq_u32_u8(q15u8));
-
-      d28u8 = vreinterpret_u8_u32(vget_low_u32(q0x2u32.val[0]));
-      d29u8 = vreinterpret_u8_u32(vget_high_u32(q0x2u32.val[0]));
-      q12u16 = vmovl_u8(d28u8);
-      q13u16 = vmovl_u8(d29u8);
-
-      __builtin_prefetch(src + 64 + src_stride * 2);
-
-      d = dst;
-      d6u32 = vld1_lane_u32((const uint32_t *)d, d6u32, 0);
-      d += dst_stride;
-      d7u32 = vld1_lane_u32((const uint32_t *)d, d7u32, 0);
-      d += dst_stride;
-      d6u32 = vld1_lane_u32((const uint32_t *)d, d6u32, 1);
-      d += dst_stride;
-      d7u32 = vld1_lane_u32((const uint32_t *)d, d7u32, 1);
-
-      d16s16 = vreinterpret_s16_u16(vget_low_u16(q8u16));
-      d17s16 = vreinterpret_s16_u16(vget_high_u16(q8u16));
-      d18s16 = vreinterpret_s16_u16(vget_low_u16(q9u16));
-      d19s16 = vreinterpret_s16_u16(vget_high_u16(q9u16));
-      d22s16 = vreinterpret_s16_u16(vget_low_u16(q11u16));
-      d24s16 = vreinterpret_s16_u16(vget_low_u16(q12u16));
-      d25s16 = vreinterpret_s16_u16(vget_high_u16(q12u16));
-      d26s16 = vreinterpret_s16_u16(vget_low_u16(q13u16));
-      d27s16 = vreinterpret_s16_u16(vget_high_u16(q13u16));
-
-      q1s32 = MULTIPLY_BY_Q0(d16s16, d17s16, d20s16, d22s16, d18s16, d19s16,
-                             d23s16, d24s16, q0s16);
-      q2s32 = MULTIPLY_BY_Q0(d17s16, d20s16, d22s16, d18s16, d19s16, d23s16,
-                             d24s16, d26s16, q0s16);
-      q14s32 = MULTIPLY_BY_Q0(d20s16, d22s16, d18s16, d19s16, d23s16, d24s16,
-                              d26s16, d27s16, q0s16);
-      q15s32 = MULTIPLY_BY_Q0(d22s16, d18s16, d19s16, d23s16, d24s16, d26s16,
-                              d27s16, d25s16, q0s16);
-
-      __builtin_prefetch(src + 64 + src_stride * 3);
-
-      d2u16 = vqrshrun_n_s32(q1s32, 7);
-      d3u16 = vqrshrun_n_s32(q2s32, 7);
-      d4u16 = vqrshrun_n_s32(q14s32, 7);
-      d5u16 = vqrshrun_n_s32(q15s32, 7);
-
-      q1u16 = vcombine_u16(d2u16, d3u16);
-      q2u16 = vcombine_u16(d4u16, d5u16);
-
-      d2u8 = vqmovn_u16(q1u16);
-      d3u8 = vqmovn_u16(q2u16);
-
-      d0x2u16 = vtrn_u16(vreinterpret_u16_u8(d2u8), vreinterpret_u16_u8(d3u8));
-      d0x2u32 = vtrn_u32(vreinterpret_u32_u16(d0x2u16.val[0]),
-                         vreinterpret_u32_u16(d0x2u16.val[1]));
-      d0x2u8 = vtrn_u8(vreinterpret_u8_u32(d0x2u32.val[0]),
-                       vreinterpret_u8_u32(d0x2u32.val[1]));
-
-      q1u8 = vcombine_u8(d0x2u8.val[0], d0x2u8.val[1]);
-      q3u8 = vreinterpretq_u8_u32(vcombine_u32(d6u32, d7u32));
-
-      q1u8 = vrhaddq_u8(q1u8, q3u8);
-
-      d2u32 = vreinterpret_u32_u8(vget_low_u8(q1u8));
-      d3u32 = vreinterpret_u32_u8(vget_high_u8(q1u8));
-
-      d = dst;
-      vst1_lane_u32((uint32_t *)d, d2u32, 0);
-      d += dst_stride;
-      vst1_lane_u32((uint32_t *)d, d3u32, 0);
-      d += dst_stride;
-      vst1_lane_u32((uint32_t *)d, d2u32, 1);
-      d += dst_stride;
-      vst1_lane_u32((uint32_t *)d, d3u32, 1);
-
-      q8u16 = q9u16;
-      d20s16 = d23s16;
-      q11u16 = q12u16;
-      q9u16 = q13u16;
-      d23s16 = vreinterpret_s16_u16(vget_high_u16(q11u16));
-    }
-    src += src_stride * 4 - w - 7;
-    dst += dst_stride * 4 - w;
-  }
-  return;
-}
-
-void aom_convolve8_avg_vert_neon(const uint8_t *src, ptrdiff_t src_stride,
-                                 uint8_t *dst, ptrdiff_t dst_stride,
-                                 const int16_t *filter_x,  // unused
-                                 int x_step_q4,            // unused
-                                 const int16_t *filter_y, int y_step_q4, int w,
-                                 int h) {
-  int height;
-  const uint8_t *s;
-  uint8_t *d;
-  uint8x8_t d2u8, d3u8;
-  uint32x2_t d2u32, d3u32, d6u32, d7u32;
-  uint32x2_t d16u32, d18u32, d20u32, d22u32, d24u32, d26u32;
-  uint8x16_t q1u8, q3u8;
-  int16x4_t d16s16, d17s16, d18s16, d19s16, d20s16, d21s16, d22s16;
-  int16x4_t d24s16, d25s16, d26s16, d27s16;
-  uint16x4_t d2u16, d3u16, d4u16, d5u16;
-  int16x8_t q0s16;
-  uint16x8_t q1u16, q2u16, q8u16, q9u16, q10u16, q11u16, q12u16, q13u16;
-  int32x4_t q1s32, q2s32, q14s32, q15s32;
-
-  assert(y_step_q4 == 16);
-
-  (void)x_step_q4;
-  (void)y_step_q4;
-  (void)filter_x;
-
-  src -= src_stride * 3;
-  q0s16 = vld1q_s16(filter_y);
-  for (; w > 0; w -= 4, src += 4, dst += 4) {  // loop_vert_h
-    s = src;
-    d16u32 = vld1_lane_u32((const uint32_t *)s, d16u32, 0);
-    s += src_stride;
-    d16u32 = vld1_lane_u32((const uint32_t *)s, d16u32, 1);
-    s += src_stride;
-    d18u32 = vld1_lane_u32((const uint32_t *)s, d18u32, 0);
-    s += src_stride;
-    d18u32 = vld1_lane_u32((const uint32_t *)s, d18u32, 1);
-    s += src_stride;
-    d20u32 = vld1_lane_u32((const uint32_t *)s, d20u32, 0);
-    s += src_stride;
-    d20u32 = vld1_lane_u32((const uint32_t *)s, d20u32, 1);
-    s += src_stride;
-    d22u32 = vld1_lane_u32((const uint32_t *)s, d22u32, 0);
-    s += src_stride;
-
-    q8u16 = vmovl_u8(vreinterpret_u8_u32(d16u32));
-    q9u16 = vmovl_u8(vreinterpret_u8_u32(d18u32));
-    q10u16 = vmovl_u8(vreinterpret_u8_u32(d20u32));
-    q11u16 = vmovl_u8(vreinterpret_u8_u32(d22u32));
-
-    d18s16 = vreinterpret_s16_u16(vget_low_u16(q9u16));
-    d19s16 = vreinterpret_s16_u16(vget_high_u16(q9u16));
-    d22s16 = vreinterpret_s16_u16(vget_low_u16(q11u16));
-    d = dst;
-    for (height = h; height > 0; height -= 4) {  // loop_vert
-      d24u32 = vld1_lane_u32((const uint32_t *)s, d24u32, 0);
-      s += src_stride;
-      d26u32 = vld1_lane_u32((const uint32_t *)s, d26u32, 0);
-      s += src_stride;
-      d26u32 = vld1_lane_u32((const uint32_t *)s, d26u32, 1);
-      s += src_stride;
-      d24u32 = vld1_lane_u32((const uint32_t *)s, d24u32, 1);
-      s += src_stride;
-
-      q12u16 = vmovl_u8(vreinterpret_u8_u32(d24u32));
-      q13u16 = vmovl_u8(vreinterpret_u8_u32(d26u32));
-
-      d6u32 = vld1_lane_u32((const uint32_t *)d, d6u32, 0);
-      d += dst_stride;
-      d6u32 = vld1_lane_u32((const uint32_t *)d, d6u32, 1);
-      d += dst_stride;
-      d7u32 = vld1_lane_u32((const uint32_t *)d, d7u32, 0);
-      d += dst_stride;
-      d7u32 = vld1_lane_u32((const uint32_t *)d, d7u32, 1);
-      d -= dst_stride * 3;
-
-      d16s16 = vreinterpret_s16_u16(vget_low_u16(q8u16));
-      d17s16 = vreinterpret_s16_u16(vget_high_u16(q8u16));
-      d20s16 = vreinterpret_s16_u16(vget_low_u16(q10u16));
-      d21s16 = vreinterpret_s16_u16(vget_high_u16(q10u16));
-      d24s16 = vreinterpret_s16_u16(vget_low_u16(q12u16));
-      d25s16 = vreinterpret_s16_u16(vget_high_u16(q12u16));
-      d26s16 = vreinterpret_s16_u16(vget_low_u16(q13u16));
-      d27s16 = vreinterpret_s16_u16(vget_high_u16(q13u16));
-
-      __builtin_prefetch(s);
-      __builtin_prefetch(s + src_stride);
-      q1s32 = MULTIPLY_BY_Q0(d16s16, d17s16, d18s16, d19s16, d20s16, d21s16,
-                             d22s16, d24s16, q0s16);
-      __builtin_prefetch(s + src_stride * 2);
-      __builtin_prefetch(s + src_stride * 3);
-      q2s32 = MULTIPLY_BY_Q0(d17s16, d18s16, d19s16, d20s16, d21s16, d22s16,
-                             d24s16, d26s16, q0s16);
-      __builtin_prefetch(d);
-      __builtin_prefetch(d + dst_stride);
-      q14s32 = MULTIPLY_BY_Q0(d18s16, d19s16, d20s16, d21s16, d22s16, d24s16,
-                              d26s16, d27s16, q0s16);
-      __builtin_prefetch(d + dst_stride * 2);
-      __builtin_prefetch(d + dst_stride * 3);
-      q15s32 = MULTIPLY_BY_Q0(d19s16, d20s16, d21s16, d22s16, d24s16, d26s16,
-                              d27s16, d25s16, q0s16);
-
-      d2u16 = vqrshrun_n_s32(q1s32, 7);
-      d3u16 = vqrshrun_n_s32(q2s32, 7);
-      d4u16 = vqrshrun_n_s32(q14s32, 7);
-      d5u16 = vqrshrun_n_s32(q15s32, 7);
-
-      q1u16 = vcombine_u16(d2u16, d3u16);
-      q2u16 = vcombine_u16(d4u16, d5u16);
-
-      d2u8 = vqmovn_u16(q1u16);
-      d3u8 = vqmovn_u16(q2u16);
-
-      q1u8 = vcombine_u8(d2u8, d3u8);
-      q3u8 = vreinterpretq_u8_u32(vcombine_u32(d6u32, d7u32));
-
-      q1u8 = vrhaddq_u8(q1u8, q3u8);
-
-      d2u32 = vreinterpret_u32_u8(vget_low_u8(q1u8));
-      d3u32 = vreinterpret_u32_u8(vget_high_u8(q1u8));
-
-      vst1_lane_u32((uint32_t *)d, d2u32, 0);
-      d += dst_stride;
-      vst1_lane_u32((uint32_t *)d, d2u32, 1);
-      d += dst_stride;
-      vst1_lane_u32((uint32_t *)d, d3u32, 0);
-      d += dst_stride;
-      vst1_lane_u32((uint32_t *)d, d3u32, 1);
-      d += dst_stride;
-
-      q8u16 = q10u16;
-      d18s16 = d22s16;
-      d19s16 = d24s16;
-      q10u16 = q13u16;
-      d22s16 = d25s16;
-    }
-  }
-  return;
-}

aom_dsp/arm/aom_convolve8_neon.c

@@ -1,331 +0,0 @@
-/*
- * Copyright (c) 2016, Alliance for Open Media. All rights reserved
- *
- * This source code is subject to the terms of the BSD 2 Clause License and
- * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
- * was not distributed with this source code in the LICENSE file, you can
- * obtain it at www.aomedia.org/license/software. If the Alliance for Open
- * Media Patent License 1.0 was not distributed with this source code in the
- * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
- */
-
-#include <arm_neon.h>
-#include <assert.h>
-
-#include "./aom_config.h"
-#include "./aom_dsp_rtcd.h"
-#include "aom/aom_integer.h"
-#include "aom_ports/mem.h"
-
-static INLINE int32x4_t MULTIPLY_BY_Q0(int16x4_t dsrc0, int16x4_t dsrc1,
-                                       int16x4_t dsrc2, int16x4_t dsrc3,
-                                       int16x4_t dsrc4, int16x4_t dsrc5,
-                                       int16x4_t dsrc6, int16x4_t dsrc7,
-                                       int16x8_t q0s16) {
-  int32x4_t qdst;
-  int16x4_t d0s16, d1s16;
-
-  d0s16 = vget_low_s16(q0s16);
-  d1s16 = vget_high_s16(q0s16);
-
-  qdst = vmull_lane_s16(dsrc0, d0s16, 0);
-  qdst = vmlal_lane_s16(qdst, dsrc1, d0s16, 1);
-  qdst = vmlal_lane_s16(qdst, dsrc2, d0s16, 2);
-  qdst = vmlal_lane_s16(qdst, dsrc3, d0s16, 3);
-  qdst = vmlal_lane_s16(qdst, dsrc4, d1s16, 0);
-  qdst = vmlal_lane_s16(qdst, dsrc5, d1s16, 1);
-  qdst = vmlal_lane_s16(qdst, dsrc6, d1s16, 2);
-  qdst = vmlal_lane_s16(qdst, dsrc7, d1s16, 3);
-  return qdst;
-}
-
-void aom_convolve8_horiz_neon(const uint8_t *src, ptrdiff_t src_stride,
-                              uint8_t *dst, ptrdiff_t dst_stride,
-                              const int16_t *filter_x, int x_step_q4,
-                              const int16_t *filter_y,  // unused
-                              int y_step_q4,            // unused
-                              int w, int h) {
-  int width;
-  const uint8_t *s, *psrc;
-  uint8_t *d, *pdst;
-  uint8x8_t d2u8, d3u8, d24u8, d25u8, d26u8, d27u8, d28u8, d29u8;
-  uint32x2_t d2u32, d3u32, d28u32, d29u32, d30u32, d31u32;
-  uint8x16_t q12u8, q13u8, q14u8, q15u8;
-  int16x4_t d16s16, d17s16, d18s16, d19s16, d20s16, d22s16, d23s16;
-  int16x4_t d24s16, d25s16, d26s16, d27s16;
-  uint16x4_t d2u16, d3u16, d4u16, d5u16, d16u16, d17u16, d18u16, d19u16;
-  int16x8_t q0s16;
-  uint16x8_t q1u16, q2u16, q8u16, q9u16, q10u16, q11u16, q12u16, q13u16;
-  int32x4_t q1s32, q2s32, q14s32, q15s32;
-  uint16x8x2_t q0x2u16;
-  uint8x8x2_t d0x2u8, d1x2u8;
-  uint32x2x2_t d0x2u32;
-  uint16x4x2_t d0x2u16, d1x2u16;
-  uint32x4x2_t q0x2u32;
-
-  assert(x_step_q4 == 16);
-
-  (void)x_step_q4;
-  (void)y_step_q4;
-  (void)filter_y;
-
-  q0s16 = vld1q_s16(filter_x);
-
-  src -= 3;  // adjust for taps
-  for (; h > 0; h -= 4, src += src_stride * 4,
-                dst += dst_stride * 4) {  // loop_horiz_v
-    s = src;
-    d24u8 = vld1_u8(s);
-    s += src_stride;
-    d25u8 = vld1_u8(s);
-    s += src_stride;
-    d26u8 = vld1_u8(s);
-    s += src_stride;
-    d27u8 = vld1_u8(s);
-
-    q12u8 = vcombine_u8(d24u8, d25u8);
-    q13u8 = vcombine_u8(d26u8, d27u8);
-
-    q0x2u16 =
-        vtrnq_u16(vreinterpretq_u16_u8(q12u8), vreinterpretq_u16_u8(q13u8));
-    d24u8 = vreinterpret_u8_u16(vget_low_u16(q0x2u16.val[0]));
-    d25u8 = vreinterpret_u8_u16(vget_high_u16(q0x2u16.val[0]));
-    d26u8 = vreinterpret_u8_u16(vget_low_u16(q0x2u16.val[1]));
-    d27u8 = vreinterpret_u8_u16(vget_high_u16(q0x2u16.val[1]));
-    d0x2u8 = vtrn_u8(d24u8, d25u8);
-    d1x2u8 = vtrn_u8(d26u8, d27u8);
-
-    __builtin_prefetch(src + src_stride * 4);
-    __builtin_prefetch(src + src_stride * 5);
-    __builtin_prefetch(src + src_stride * 6);
-
-    q8u16 = vmovl_u8(d0x2u8.val[0]);
-    q9u16 = vmovl_u8(d0x2u8.val[1]);
-    q10u16 = vmovl_u8(d1x2u8.val[0]);
-    q11u16 = vmovl_u8(d1x2u8.val[1]);
-
-    d16u16 = vget_low_u16(q8u16);
-    d17u16 = vget_high_u16(q8u16);
-    d18u16 = vget_low_u16(q9u16);
-    d19u16 = vget_high_u16(q9u16);
-    q8u16 = vcombine_u16(d16u16, d18u16);  // vswp 17 18
-    q9u16 = vcombine_u16(d17u16, d19u16);
-
-    d20s16 = vreinterpret_s16_u16(vget_low_u16(q10u16));
-    d23s16 = vreinterpret_s16_u16(vget_high_u16(q10u16));  // vmov 23 21
-    for (width = w, psrc = src + 7, pdst = dst; width > 0;
-         width -= 4, psrc += 4, pdst += 4) {  // loop_horiz
-      s = psrc;
-      d28u32 = vld1_dup_u32((const uint32_t *)s);
-      s += src_stride;
-      d29u32 = vld1_dup_u32((const uint32_t *)s);
-      s += src_stride;
-      d31u32 = vld1_dup_u32((const uint32_t *)s);
-      s += src_stride;
-      d30u32 = vld1_dup_u32((const uint32_t *)s);
-
-      __builtin_prefetch(psrc + 64);
-
-      d0x2u16 =
-          vtrn_u16(vreinterpret_u16_u32(d28u32), vreinterpret_u16_u32(d31u32));
-      d1x2u16 =
-          vtrn_u16(vreinterpret_u16_u32(d29u32), vreinterpret_u16_u32(d30u32));
-      d0x2u8 = vtrn_u8(vreinterpret_u8_u16(d0x2u16.val[0]),   // d28
-                       vreinterpret_u8_u16(d1x2u16.val[0]));  // d29
-      d1x2u8 = vtrn_u8(vreinterpret_u8_u16(d0x2u16.val[1]),   // d31
-                       vreinterpret_u8_u16(d1x2u16.val[1]));  // d30
-
-      __builtin_prefetch(psrc + 64 + src_stride);
-
-      q14u8 = vcombine_u8(d0x2u8.val[0], d0x2u8.val[1]);
-      q15u8 = vcombine_u8(d1x2u8.val[1], d1x2u8.val[0]);
-      q0x2u32 =
-          vtrnq_u32(vreinterpretq_u32_u8(q14u8), vreinterpretq_u32_u8(q15u8));
-
-      d28u8 = vreinterpret_u8_u32(vget_low_u32(q0x2u32.val[0]));
-      d29u8 = vreinterpret_u8_u32(vget_high_u32(q0x2u32.val[0]));
-      q12u16 = vmovl_u8(d28u8);
-      q13u16 = vmovl_u8(d29u8);
-
-      __builtin_prefetch(psrc + 64 + src_stride * 2);
-
-      d16s16 = vreinterpret_s16_u16(vget_low_u16(q8u16));
-      d17s16 = vreinterpret_s16_u16(vget_high_u16(q8u16));
-      d18s16 = vreinterpret_s16_u16(vget_low_u16(q9u16));
-      d19s16 = vreinterpret_s16_u16(vget_high_u16(q9u16));
-      d22s16 = vreinterpret_s16_u16(vget_low_u16(q11u16));
-      d24s16 = vreinterpret_s16_u16(vget_low_u16(q12u16));
-      d25s16 = vreinterpret_s16_u16(vget_high_u16(q12u16));
-      d26s16 = vreinterpret_s16_u16(vget_low_u16(q13u16));
-      d27s16 = vreinterpret_s16_u16(vget_high_u16(q13u16));
-
-      q1s32 = MULTIPLY_BY_Q0(d16s16, d17s16, d20s16, d22s16, d18s16, d19s16,
-                             d23s16, d24s16, q0s16);
-      q2s32 = MULTIPLY_BY_Q0(d17s16, d20s16, d22s16, d18s16, d19s16, d23s16,
-                             d24s16, d26s16, q0s16);
-      q14s32 = MULTIPLY_BY_Q0(d20s16, d22s16, d18s16, d19s16, d23s16, d24s16,
-                              d26s16, d27s16, q0s16);
-      q15s32 = MULTIPLY_BY_Q0(d22s16, d18s16, d19s16, d23s16, d24s16, d26s16,
-                              d27s16, d25s16, q0s16);
-
-      __builtin_prefetch(psrc + 60 + src_stride * 3);
-
-      d2u16 = vqrshrun_n_s32(q1s32, 7);
-      d3u16 = vqrshrun_n_s32(q2s32, 7);
-      d4u16 = vqrshrun_n_s32(q14s32, 7);
-      d5u16 = vqrshrun_n_s32(q15s32, 7);
-
-      q1u16 = vcombine_u16(d2u16, d3u16);
-      q2u16 = vcombine_u16(d4u16, d5u16);
-
-      d2u8 = vqmovn_u16(q1u16);
-      d3u8 = vqmovn_u16(q2u16);
-
-      d0x2u16 = vtrn_u16(vreinterpret_u16_u8(d2u8), vreinterpret_u16_u8(d3u8));
-      d0x2u32 = vtrn_u32(vreinterpret_u32_u16(d0x2u16.val[0]),
-                         vreinterpret_u32_u16(d0x2u16.val[1]));
-      d0x2u8 = vtrn_u8(vreinterpret_u8_u32(d0x2u32.val[0]),
-                       vreinterpret_u8_u32(d0x2u32.val[1]));
-
-      d2u32 = vreinterpret_u32_u8(d0x2u8.val[0]);
-      d3u32 = vreinterpret_u32_u8(d0x2u8.val[1]);
-
-      d = pdst;
-      vst1_lane_u32((uint32_t *)d, d2u32, 0);
-      d += dst_stride;
-      vst1_lane_u32((uint32_t *)d, d3u32, 0);
-      d += dst_stride;
-      vst1_lane_u32((uint32_t *)d, d2u32, 1);
-      d += dst_stride;
-      vst1_lane_u32((uint32_t *)d, d3u32, 1);
-
-      q8u16 = q9u16;
-      d20s16 = d23s16;
-      q11u16 = q12u16;
-      q9u16 = q13u16;
-      d23s16 = vreinterpret_s16_u16(vget_high_u16(q11u16));
-    }
-  }
-  return;
-}
-
-void aom_convolve8_vert_neon(const uint8_t *src, ptrdiff_t src_stride,
-                             uint8_t *dst, ptrdiff_t dst_stride,
-                             const int16_t *filter_x,  // unused
-                             int x_step_q4,            // unused
-                             const int16_t *filter_y, int y_step_q4, int w,
-                             int h) {
-  int height;
-  const uint8_t *s;
-  uint8_t *d;
-  uint32x2_t d2u32, d3u32;
-  uint32x2_t d16u32, d18u32, d20u32, d22u32, d24u32, d26u32;
-  int16x4_t d16s16, d17s16, d18s16, d19s16, d20s16, d21s16, d22s16;
-  int16x4_t d24s16, d25s16, d26s16, d27s16;
-  uint16x4_t d2u16, d3u16, d4u16, d5u16;
-  int16x8_t q0s16;
-  uint16x8_t q1u16, q2u16, q8u16, q9u16, q10u16, q11u16, q12u16, q13u16;
-  int32x4_t q1s32, q2s32, q14s32, q15s32;
-
-  assert(y_step_q4 == 16);
-
-  (void)x_step_q4;
-  (void)y_step_q4;
-  (void)filter_x;
-
-  src -= src_stride * 3;
-  q0s16 = vld1q_s16(filter_y);
-  for (; w > 0; w -= 4, src += 4, dst += 4) {  // loop_vert_h
-    s = src;
-    d16u32 = vld1_lane_u32((const uint32_t *)s, d16u32, 0);
-    s += src_stride;
-    d16u32 = vld1_lane_u32((const uint32_t *)s, d16u32, 1);
-    s += src_stride;
-    d18u32 = vld1_lane_u32((const uint32_t *)s, d18u32, 0);
-    s += src_stride;
-    d18u32 = vld1_lane_u32((const uint32_t *)s, d18u32, 1);
-    s += src_stride;
-    d20u32 = vld1_lane_u32((const uint32_t *)s, d20u32, 0);
-    s += src_stride;
-    d20u32 = vld1_lane_u32((const uint32_t *)s, d20u32, 1);
-    s += src_stride;
-    d22u32 = vld1_lane_u32((const uint32_t *)s, d22u32, 0);
-    s += src_stride;
-
-    q8u16 = vmovl_u8(vreinterpret_u8_u32(d16u32));
-    q9u16 = vmovl_u8(vreinterpret_u8_u32(d18u32));
-    q10u16 = vmovl_u8(vreinterpret_u8_u32(d20u32));
-    q11u16 = vmovl_u8(vreinterpret_u8_u32(d22u32));
-
-    d18s16 = vreinterpret_s16_u16(vget_low_u16(q9u16));
-    d19s16 = vreinterpret_s16_u16(vget_high_u16(q9u16));
-    d22s16 = vreinterpret_s16_u16(vget_low_u16(q11u16));
-    d = dst;
-    for (height = h; height > 0; height -= 4) {  // loop_vert
-      d24u32 = vld1_lane_u32((const uint32_t *)s, d24u32, 0);
-      s += src_stride;
-      d26u32 = vld1_lane_u32((const uint32_t *)s, d26u32, 0);
-      s += src_stride;
-      d26u32 = vld1_lane_u32((const uint32_t *)s, d26u32, 1);
-      s += src_stride;
-      d24u32 = vld1_lane_u32((const uint32_t *)s, d24u32, 1);
-      s += src_stride;
-
-      q12u16 = vmovl_u8(vreinterpret_u8_u32(d24u32));
-      q13u16 = vmovl_u8(vreinterpret_u8_u32(d26u32));
-
-      d16s16 = vreinterpret_s16_u16(vget_low_u16(q8u16));
-      d17s16 = vreinterpret_s16_u16(vget_high_u16(q8u16));
-      d20s16 = vreinterpret_s16_u16(vget_low_u16(q10u16));
-      d21s16 = vreinterpret_s16_u16(vget_high_u16(q10u16));
-      d24s16 = vreinterpret_s16_u16(vget_low_u16(q12u16));
-      d25s16 = vreinterpret_s16_u16(vget_high_u16(q12u16));
-      d26s16 = vreinterpret_s16_u16(vget_low_u16(q13u16));
-      d27s16 = vreinterpret_s16_u16(vget_high_u16(q13u16));
-
-      __builtin_prefetch(d);
-      __builtin_prefetch(d + dst_stride);
-      q1s32 = MULTIPLY_BY_Q0(d16s16, d17s16, d18s16, d19s16, d20s16, d21s16,
-                             d22s16, d24s16, q0s16);
-      __builtin_prefetch(d + dst_stride * 2);
-      __builtin_prefetch(d + dst_stride * 3);
-      q2s32 = MULTIPLY_BY_Q0(d17s16, d18s16, d19s16, d20s16, d21s16, d22s16,
-                             d24s16, d26s16, q0s16);
-      __builtin_prefetch(s);
-      __builtin_prefetch(s + src_stride);
-      q14s32 = MULTIPLY_BY_Q0(d18s16, d19s16, d20s16, d21s16, d22s16, d24s16,
-                              d26s16, d27s16, q0s16);
-      __builtin_prefetch(s + src_stride * 2);
-      __builtin_prefetch(s + src_stride * 3);
-      q15s32 = MULTIPLY_BY_Q0(d19s16, d20s16, d21s16, d22s16, d24s16, d26s16,
-                              d27s16, d25s16, q0s16);
-
-      d2u16 = vqrshrun_n_s32(q1s32, 7);
-      d3u16 = vqrshrun_n_s32(q2s32, 7);
-      d4u16 = vqrshrun_n_s32(q14s32, 7);
-      d5u16 = vqrshrun_n_s32(q15s32, 7);
-
-      q1u16 = vcombine_u16(d2u16, d3u16);
-      q2u16 = vcombine_u16(d4u16, d5u16);
-
-      d2u32 = vreinterpret_u32_u8(vqmovn_u16(q1u16));
-      d3u32 = vreinterpret_u32_u8(vqmovn_u16(q2u16));
-
-      vst1_lane_u32((uint32_t *)d, d2u32, 0);
-      d += dst_stride;
-      vst1_lane_u32((uint32_t *)d, d2u32, 1);
-      d += dst_stride;
-      vst1_lane_u32((uint32_t *)d, d3u32, 0);
-      d += dst_stride;
-      vst1_lane_u32((uint32_t *)d, d3u32, 1);
-      d += dst_stride;
-
-      q8u16 = q10u16;
-      d18s16 = d22s16;
-      d19s16 = d24s16;
-      q10u16 = q13u16;
-      d22s16 = d25s16;
-    }
-  }
-  return;
-}

aom_dsp/arm/aom_convolve_avg_neon.c

@@ -1,145 +0,0 @@
-/*
- * Copyright (c) 2016, Alliance for Open Media. All rights reserved
- *
- * This source code is subject to the terms of the BSD 2 Clause License and
- * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
- * was not distributed with this source code in the LICENSE file, you can
- * obtain it at www.aomedia.org/license/software. If the Alliance for Open
- * Media Patent License 1.0 was not distributed with this source code in the
- * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
- */
-
-#include <arm_neon.h>
-
-#include "./aom_dsp_rtcd.h"
-#include "aom/aom_integer.h"
-
-void aom_convolve_avg_neon(const uint8_t *src,    // r0
-                           ptrdiff_t src_stride,  // r1
-                           uint8_t *dst,          // r2
-                           ptrdiff_t dst_stride,  // r3
-                           const int16_t *filter_x, int filter_x_stride,
-                           const int16_t *filter_y, int filter_y_stride, int w,
-                           int h) {
-  uint8_t *d;
-  uint8x8_t d0u8, d1u8, d2u8, d3u8;
-  uint32x2_t d0u32, d2u32;
-  uint8x16_t q0u8, q1u8, q2u8, q3u8, q8u8, q9u8, q10u8, q11u8;
-  (void)filter_x;
-  (void)filter_x_stride;
-  (void)filter_y;
-  (void)filter_y_stride;
-
-  d = dst;
-  if (w > 32) {  // avg64
-    for (; h > 0; h -= 1) {
-      q0u8 = vld1q_u8(src);
-      q1u8 = vld1q_u8(src + 16);
-      q2u8 = vld1q_u8(src + 32);
-      q3u8 = vld1q_u8(src + 48);
-      src += src_stride;
-      q8u8 = vld1q_u8(d);
-      q9u8 = vld1q_u8(d + 16);
-      q10u8 = vld1q_u8(d + 32);
-      q11u8 = vld1q_u8(d + 48);
-      d += dst_stride;
-
-      q0u8 = vrhaddq_u8(q0u8, q8u8);
-      q1u8 = vrhaddq_u8(q1u8, q9u8);
-      q2u8 = vrhaddq_u8(q2u8, q10u8);
-      q3u8 = vrhaddq_u8(q3u8, q11u8);
-
-      vst1q_u8(dst, q0u8);
-      vst1q_u8(dst + 16, q1u8);
-      vst1q_u8(dst + 32, q2u8);
-      vst1q_u8(dst + 48, q3u8);
-      dst += dst_stride;
-    }
-  } else if (w == 32) {  // avg32
-    for (; h > 0; h -= 2) {
-      q0u8 = vld1q_u8(src);
-      q1u8 = vld1q_u8(src + 16);
-      src += src_stride;
-      q2u8 = vld1q_u8(src);
-      q3u8 = vld1q_u8(src + 16);
-      src += src_stride;
-      q8u8 = vld1q_u8(d);
-      q9u8 = vld1q_u8(d + 16);
-      d += dst_stride;
-      q10u8 = vld1q_u8(d);
-      q11u8 = vld1q_u8(d + 16);
-      d += dst_stride;
-
-      q0u8 = vrhaddq_u8(q0u8, q8u8);
-      q1u8 = vrhaddq_u8(q1u8, q9u8);
-      q2u8 = vrhaddq_u8(q2u8, q10u8);
-      q3u8 = vrhaddq_u8(q3u8, q11u8);
-
-      vst1q_u8(dst, q0u8);
-      vst1q_u8(dst + 16, q1u8);
-      dst += dst_stride;
-      vst1q_u8(dst, q2u8);
-      vst1q_u8(dst + 16, q3u8);
-      dst += dst_stride;
-    }
-  } else if (w > 8) {  // avg16
-    for (; h > 0; h -= 2) {
-      q0u8 = vld1q_u8(src);
-      src += src_stride;
-      q1u8 = vld1q_u8(src);
-      src += src_stride;
-      q2u8 = vld1q_u8(d);
-      d += dst_stride;
-      q3u8 = vld1q_u8(d);
-      d += dst_stride;
-
-      q0u8 = vrhaddq_u8(q0u8, q2u8);
-      q1u8 = vrhaddq_u8(q1u8, q3u8);
-
-      vst1q_u8(dst, q0u8);
-      dst += dst_stride;
-      vst1q_u8(dst, q1u8);
-      dst += dst_stride;
-    }
-  } else if (w == 8) {  // avg8
-    for (; h > 0; h -= 2) {
-      d0u8 = vld1_u8(src);
-      src += src_stride;
-      d1u8 = vld1_u8(src);
-      src += src_stride;
-      d2u8 = vld1_u8(d);
-      d += dst_stride;
-      d3u8 = vld1_u8(d);
-      d += dst_stride;
-
-      q0u8 = vcombine_u8(d0u8, d1u8);
-      q1u8 = vcombine_u8(d2u8, d3u8);
-      q0u8 = vrhaddq_u8(q0u8, q1u8);
-
-      vst1_u8(dst, vget_low_u8(q0u8));
-      dst += dst_stride;
-      vst1_u8(dst, vget_high_u8(q0u8));
-      dst += dst_stride;
-    }
-  } else {  // avg4
-    for (; h > 0; h -= 2) {
-      d0u32 = vld1_lane_u32((const uint32_t *)src, d0u32, 0);
-      src += src_stride;
-      d0u32 = vld1_lane_u32((const uint32_t *)src, d0u32, 1);
-      src += src_stride;
-      d2u32 = vld1_lane_u32((const uint32_t *)d, d2u32, 0);
-      d += dst_stride;
-      d2u32 = vld1_lane_u32((const uint32_t *)d, d2u32, 1);
-      d += dst_stride;
-
-      d0u8 = vrhadd_u8(vreinterpret_u8_u32(d0u32), vreinterpret_u8_u32(d2u32));
-
-      d0u32 = vreinterpret_u32_u8(d0u8);
-      vst1_lane_u32((uint32_t *)dst, d0u32, 0);
-      dst += dst_stride;
-      vst1_lane_u32((uint32_t *)dst, d0u32, 1);
-      dst += dst_stride;
-    }
-  }
-  return;
-}

aom_dsp/arm/aom_convolve_copy_neon.c

@@ -1,93 +0,0 @@
-/*
- * Copyright (c) 2016, Alliance for Open Media. All rights reserved
- *
- * This source code is subject to the terms of the BSD 2 Clause License and
- * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
- * was not distributed with this source code in the LICENSE file, you can
- * obtain it at www.aomedia.org/license/software. If the Alliance for Open
- * Media Patent License 1.0 was not distributed with this source code in the
- * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
- */
-
-#include <arm_neon.h>
-
-#include "./aom_dsp_rtcd.h"
-#include "aom/aom_integer.h"
-
-void aom_convolve_copy_neon(const uint8_t *src,    // r0
-                            ptrdiff_t src_stride,  // r1
-                            uint8_t *dst,          // r2
-                            ptrdiff_t dst_stride,  // r3
-                            const int16_t *filter_x, int filter_x_stride,
-                            const int16_t *filter_y, int filter_y_stride, int w,
-                            int h) {
-  uint8x8_t d0u8, d2u8;
-  uint8x16_t q0u8, q1u8, q2u8, q3u8;
-  (void)filter_x;
-  (void)filter_x_stride;
-  (void)filter_y;
-  (void)filter_y_stride;
-
-  if (w > 32) {  // copy64
-    for (; h > 0; h--) {
-      q0u8 = vld1q_u8(src);
-      q1u8 = vld1q_u8(src + 16);
-      q2u8 = vld1q_u8(src + 32);
-      q3u8 = vld1q_u8(src + 48);
-      src += src_stride;
-
-      vst1q_u8(dst, q0u8);
-      vst1q_u8(dst + 16, q1u8);
-      vst1q_u8(dst + 32, q2u8);
-      vst1q_u8(dst + 48, q3u8);
-      dst += dst_stride;
-    }
-  } else if (w == 32) {  // copy32
-    for (; h > 0; h -= 2) {
-      q0u8 = vld1q_u8(src);
-      q1u8 = vld1q_u8(src + 16);
-      src += src_stride;
-      q2u8 = vld1q_u8(src);
-      q3u8 = vld1q_u8(src + 16);
-      src += src_stride;
-
-      vst1q_u8(dst, q0u8);
-      vst1q_u8(dst + 16, q1u8);
-      dst += dst_stride;
-      vst1q_u8(dst, q2u8);
-      vst1q_u8(dst + 16, q3u8);
-      dst += dst_stride;
-    }
-  } else if (w > 8) {  // copy16
-    for (; h > 0; h -= 2) {
-      q0u8 = vld1q_u8(src);
-      src += src_stride;
-      q1u8 = vld1q_u8(src);
-      src += src_stride;
-
-      vst1q_u8(dst, q0u8);
-      dst += dst_stride;
-      vst1q_u8(dst, q1u8);
-      dst += dst_stride;
-    }
-  } else if (w == 8) {  // copy8
-    for (; h > 0; h -= 2) {
-      d0u8 = vld1_u8(src);
-      src += src_stride;
-      d2u8 = vld1_u8(src);
-      src += src_stride;
-
-      vst1_u8(dst, d0u8);
-      dst += dst_stride;
-      vst1_u8(dst, d2u8);
-      dst += dst_stride;
-    }
-  } else {  // copy4
-    for (; h > 0; h--) {
-      *(uint32_t *)dst = *(const uint32_t *)src;
-      src += src_stride;
-      dst += dst_stride;
-    }
-  }
-  return;
-}

aom_dsp/arm/aom_convolve_neon.c

@@ -1,66 +0,0 @@
-/*
- * Copyright (c) 2016, Alliance for Open Media. All rights reserved
- *
- * This source code is subject to the terms of the BSD 2 Clause License and
- * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
- * was not distributed with this source code in the LICENSE file, you can
- * obtain it at www.aomedia.org/license/software. If the Alliance for Open
- * Media Patent License 1.0 was not distributed with this source code in the
- * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
- */
-
-#include <assert.h>
-
-#include "./aom_dsp_rtcd.h"
-#include "aom_dsp/aom_dsp_common.h"
-#include "aom_ports/mem.h"
-
-void aom_convolve8_neon(const uint8_t *src, ptrdiff_t src_stride, uint8_t *dst,
-                        ptrdiff_t dst_stride, const int16_t *filter_x,
-                        int x_step_q4, const int16_t *filter_y, int y_step_q4,
-                        int w, int h) {
-  /* Given our constraints: w <= 64, h <= 64, taps == 8 we can reduce the
-   * maximum buffer size to 64 * 64 + 7 (+ 1 to make it divisible by 4).
-   */
-  DECLARE_ALIGNED(8, uint8_t, temp[64 * 72]);
-
-  // Account for the vertical phase needing 3 lines prior and 4 lines post
-  int intermediate_height = h + 7;
-
-  assert(y_step_q4 == 16);
-  assert(x_step_q4 == 16);
-
-  /* Filter starting 3 lines back. The neon implementation will ignore the
-   * given height and filter a multiple of 4 lines. Since this goes in to
-   * the temp buffer which has lots of extra room and is subsequently discarded
-   * this is safe if somewhat less than ideal.
-   */
-  aom_convolve8_horiz_neon(src - src_stride * 3, src_stride, temp, 64, filter_x,
-                           x_step_q4, filter_y, y_step_q4, w,
-                           intermediate_height);
-
-  /* Step into the temp buffer 3 lines to get the actual frame data */
-  aom_convolve8_vert_neon(temp + 64 * 3, 64, dst, dst_stride, filter_x,
-                          x_step_q4, filter_y, y_step_q4, w, h);
-}
-
-void aom_convolve8_avg_neon(const uint8_t *src, ptrdiff_t src_stride,
-                            uint8_t *dst, ptrdiff_t dst_stride,
-                            const int16_t *filter_x, int x_step_q4,
-                            const int16_t *filter_y, int y_step_q4, int w,
-                            int h) {
-  DECLARE_ALIGNED(8, uint8_t, temp[64 * 72]);
-  int intermediate_height = h + 7;
-
-  assert(y_step_q4 == 16);
-  assert(x_step_q4 == 16);
-
-  /* This implementation has the same issues as above. In addition, we only want
-   * to average the values after both passes.
-   */
-  aom_convolve8_horiz_neon(src - src_stride * 3, src_stride, temp, 64, filter_x,
-                           x_step_q4, filter_y, y_step_q4, w,
-                           intermediate_height);
-  aom_convolve8_avg_vert_neon(temp + 64 * 3, 64, dst, dst_stride, filter_x,
-                              x_step_q4, filter_y, y_step_q4, w, h);
-}

aom_dsp/arm/avg_neon.c

@@ -1,254 +0,0 @@
-/*
- * Copyright (c) 2016, Alliance for Open Media. All rights reserved
- *
- * This source code is subject to the terms of the BSD 2 Clause License and
- * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
- * was not distributed with this source code in the LICENSE file, you can
- * obtain it at www.aomedia.org/license/software. If the Alliance for Open
- * Media Patent License 1.0 was not distributed with this source code in the
- * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
- */
-
-#include <arm_neon.h>
-#include <assert.h>
-
-#include "./aom_dsp_rtcd.h"
-#include "./aom_config.h"
-
-#include "aom/aom_integer.h"
-
-static INLINE unsigned int horizontal_add_u16x8(const uint16x8_t v_16x8) {
-  const uint32x4_t a = vpaddlq_u16(v_16x8);
-  const uint64x2_t b = vpaddlq_u32(a);
-  const uint32x2_t c = vadd_u32(vreinterpret_u32_u64(vget_low_u64(b)),
-                                vreinterpret_u32_u64(vget_high_u64(b)));
-  return vget_lane_u32(c, 0);
-}
-
-unsigned int aom_avg_4x4_neon(const uint8_t *s, int p) {
-  uint16x8_t v_sum;
-  uint32x2_t v_s0 = vdup_n_u32(0);
-  uint32x2_t v_s1 = vdup_n_u32(0);
-  v_s0 = vld1_lane_u32((const uint32_t *)s, v_s0, 0);
-  v_s0 = vld1_lane_u32((const uint32_t *)(s + p), v_s0, 1);
-  v_s1 = vld1_lane_u32((const uint32_t *)(s + 2 * p), v_s1, 0);
-  v_s1 = vld1_lane_u32((const uint32_t *)(s + 3 * p), v_s1, 1);
-  v_sum = vaddl_u8(vreinterpret_u8_u32(v_s0), vreinterpret_u8_u32(v_s1));
-  return (horizontal_add_u16x8(v_sum) + 8) >> 4;
-}
-
-unsigned int aom_avg_8x8_neon(const uint8_t *s, int p) {
-  uint8x8_t v_s0 = vld1_u8(s);
-  const uint8x8_t v_s1 = vld1_u8(s + p);
-  uint16x8_t v_sum = vaddl_u8(v_s0, v_s1);
-
-  v_s0 = vld1_u8(s + 2 * p);
-  v_sum = vaddw_u8(v_sum, v_s0);
-
-  v_s0 = vld1_u8(s + 3 * p);
-  v_sum = vaddw_u8(v_sum, v_s0);
-
-  v_s0 = vld1_u8(s + 4 * p);
-  v_sum = vaddw_u8(v_sum, v_s0);
-
-  v_s0 = vld1_u8(s + 5 * p);
-  v_sum = vaddw_u8(v_sum, v_s0);
-
-  v_s0 = vld1_u8(s + 6 * p);
-  v_sum = vaddw_u8(v_sum, v_s0);
-
-  v_s0 = vld1_u8(s + 7 * p);
-  v_sum = vaddw_u8(v_sum, v_s0);
-
-  return (horizontal_add_u16x8(v_sum) + 32) >> 6;
-}
-
-// coeff: 16 bits, dynamic range [-32640, 32640].
-// length: value range {16, 64, 256, 1024}.
-int aom_satd_neon(const int16_t *coeff, int length) {
-  const int16x4_t zero = vdup_n_s16(0);
-  int32x4_t accum = vdupq_n_s32(0);
-
-  do {
-    const int16x8_t src0 = vld1q_s16(coeff);
-    const int16x8_t src8 = vld1q_s16(coeff + 8);
-    accum = vabal_s16(accum, vget_low_s16(src0), zero);
-    accum = vabal_s16(accum, vget_high_s16(src0), zero);
-    accum = vabal_s16(accum, vget_low_s16(src8), zero);
-    accum = vabal_s16(accum, vget_high_s16(src8), zero);
-    length -= 16;
-    coeff += 16;
-  } while (length != 0);
-
-  {
-    // satd: 26 bits, dynamic range [-32640 * 1024, 32640 * 1024]
-    const int64x2_t s0 = vpaddlq_s32(accum);  // cascading summation of 'accum'.
-    const int32x2_t s1 = vadd_s32(vreinterpret_s32_s64(vget_low_s64(s0)),
-                                  vreinterpret_s32_s64(vget_high_s64(s0)));
-    const int satd = vget_lane_s32(s1, 0);
-    return satd;
-  }
-}
-
-void aom_int_pro_row_neon(int16_t hbuf[16], uint8_t const *ref,
-                          const int ref_stride, const int height) {
-  int i;
-  uint16x8_t vec_sum_lo = vdupq_n_u16(0);
-  uint16x8_t vec_sum_hi = vdupq_n_u16(0);
-  const int shift_factor = ((height >> 5) + 3) * -1;
-  const int16x8_t vec_shift = vdupq_n_s16(shift_factor);
-
-  for (i = 0; i < height; i += 8) {
-    const uint8x16_t vec_row1 = vld1q_u8(ref);
-    const uint8x16_t vec_row2 = vld1q_u8(ref + ref_stride);
-    const uint8x16_t vec_row3 = vld1q_u8(ref + ref_stride * 2);
-    const uint8x16_t vec_row4 = vld1q_u8(ref + ref_stride * 3);
-    const uint8x16_t vec_row5 = vld1q_u8(ref + ref_stride * 4);
-    const uint8x16_t vec_row6 = vld1q_u8(ref + ref_stride * 5);
-    const uint8x16_t vec_row7 = vld1q_u8(ref + ref_stride * 6);
-    const uint8x16_t vec_row8 = vld1q_u8(ref + ref_stride * 7);
-
-    vec_sum_lo = vaddw_u8(vec_sum_lo, vget_low_u8(vec_row1));
-    vec_sum_hi = vaddw_u8(vec_sum_hi, vget_high_u8(vec_row1));
-
-    vec_sum_lo = vaddw_u8(vec_sum_lo, vget_low_u8(vec_row2));
-    vec_sum_hi = vaddw_u8(vec_sum_hi, vget_high_u8(vec_row2));
-
-    vec_sum_lo = vaddw_u8(vec_sum_lo, vget_low_u8(vec_row3));
-    vec_sum_hi = vaddw_u8(vec_sum_hi, vget_high_u8(vec_row3));
-
-    vec_sum_lo = vaddw_u8(vec_sum_lo, vget_low_u8(vec_row4));
-    vec_sum_hi = vaddw_u8(vec_sum_hi, vget_high_u8(vec_row4));
-
-    vec_sum_lo = vaddw_u8(vec_sum_lo, vget_low_u8(vec_row5));
-    vec_sum_hi = vaddw_u8(vec_sum_hi, vget_high_u8(vec_row5));
-
-    vec_sum_lo = vaddw_u8(vec_sum_lo, vget_low_u8(vec_row6));
-    vec_sum_hi = vaddw_u8(vec_sum_hi, vget_high_u8(vec_row6));
-
-    vec_sum_lo = vaddw_u8(vec_sum_lo, vget_low_u8(vec_row7));
-    vec_sum_hi = vaddw_u8(vec_sum_hi, vget_high_u8(vec_row7));
-
-    vec_sum_lo = vaddw_u8(vec_sum_lo, vget_low_u8(vec_row8));
-    vec_sum_hi = vaddw_u8(vec_sum_hi, vget_high_u8(vec_row8));
-
-    ref += ref_stride * 8;
-  }
-
-  vec_sum_lo = vshlq_u16(vec_sum_lo, vec_shift);
-  vec_sum_hi = vshlq_u16(vec_sum_hi, vec_shift);
-
-  vst1q_s16(hbuf, vreinterpretq_s16_u16(vec_sum_lo));
-  hbuf += 8;
-  vst1q_s16(hbuf, vreinterpretq_s16_u16(vec_sum_hi));
-}
-
-int16_t aom_int_pro_col_neon(uint8_t const *ref, const int width) {
-  int i;
-  uint16x8_t vec_sum = vdupq_n_u16(0);
-
-  for (i = 0; i < width; i += 16) {
-    const uint8x16_t vec_row = vld1q_u8(ref);
-    vec_sum = vaddw_u8(vec_sum, vget_low_u8(vec_row));
-    vec_sum = vaddw_u8(vec_sum, vget_high_u8(vec_row));
-    ref += 16;
-  }
-
-  return horizontal_add_u16x8(vec_sum);
-}
-
-// ref, src = [0, 510] - max diff = 16-bits
-// bwl = {2, 3, 4}, width = {16, 32, 64}
-int aom_vector_var_neon(int16_t const *ref, int16_t const *src, const int bwl) {
-  int width = 4 << bwl;
-  int32x4_t sse = vdupq_n_s32(0);
-  int16x8_t total = vdupq_n_s16(0);
-
-  assert(width >= 8);
-  assert((width % 8) == 0);
-
-  do {
-    const int16x8_t r = vld1q_s16(ref);
-    const int16x8_t s = vld1q_s16(src);
-    const int16x8_t diff = vsubq_s16(r, s);  // [-510, 510], 10 bits.
-    const int16x4_t diff_lo = vget_low_s16(diff);
-    const int16x4_t diff_hi = vget_high_s16(diff);
-    sse = vmlal_s16(sse, diff_lo, diff_lo);  // dynamic range 26 bits.
-    sse = vmlal_s16(sse, diff_hi, diff_hi);
-    total = vaddq_s16(total, diff);  // dynamic range 16 bits.
-
-    ref += 8;
-    src += 8;
-    width -= 8;
-  } while (width != 0);
-
-  {
-    // Note: 'total''s pairwise addition could be implemented similarly to
-    // horizontal_add_u16x8(), but one less vpaddl with 'total' when paired
-    // with the summation of 'sse' performed better on a Cortex-A15.
-    const int32x4_t t0 = vpaddlq_s16(total);  // cascading summation of 'total'
-    const int32x2_t t1 = vadd_s32(vget_low_s32(t0), vget_high_s32(t0));
-    const int32x2_t t2 = vpadd_s32(t1, t1);
-    const int t = vget_lane_s32(t2, 0);
-    const int64x2_t s0 = vpaddlq_s32(sse);  // cascading summation of 'sse'.
-    const int32x2_t s1 = vadd_s32(vreinterpret_s32_s64(vget_low_s64(s0)),
-                                  vreinterpret_s32_s64(vget_high_s64(s0)));
-    const int s = vget_lane_s32(s1, 0);
-    const int shift_factor = bwl + 2;
-    return s - ((t * t) >> shift_factor);
-  }
-}
-
-void aom_minmax_8x8_neon(const uint8_t *a, int a_stride, const uint8_t *b,
-                         int b_stride, int *min, int *max) {
-  // Load and concatenate.
-  const uint8x16_t a01 = vcombine_u8(vld1_u8(a), vld1_u8(a + a_stride));
-  const uint8x16_t a23 =
-      vcombine_u8(vld1_u8(a + 2 * a_stride), vld1_u8(a + 3 * a_stride));
-  const uint8x16_t a45 =
-      vcombine_u8(vld1_u8(a + 4 * a_stride), vld1_u8(a + 5 * a_stride));
-  const uint8x16_t a67 =
-      vcombine_u8(vld1_u8(a + 6 * a_stride), vld1_u8(a + 7 * a_stride));
-
-  const uint8x16_t b01 = vcombine_u8(vld1_u8(b), vld1_u8(b + b_stride));
-  const uint8x16_t b23 =
-      vcombine_u8(vld1_u8(b + 2 * b_stride), vld1_u8(b + 3 * b_stride));
-  const uint8x16_t b45 =
-      vcombine_u8(vld1_u8(b + 4 * b_stride), vld1_u8(b + 5 * b_stride));
-  const uint8x16_t b67 =
-      vcombine_u8(vld1_u8(b + 6 * b_stride), vld1_u8(b + 7 * b_stride));
-
-  // Absolute difference.
-  const uint8x16_t ab01_diff = vabdq_u8(a01, b01);
-  const uint8x16_t ab23_diff = vabdq_u8(a23, b23);
-  const uint8x16_t ab45_diff = vabdq_u8(a45, b45);
-  const uint8x16_t ab67_diff = vabdq_u8(a67, b67);
-
-  // Max values between the Q vectors.
-  const uint8x16_t ab0123_max = vmaxq_u8(ab01_diff, ab23_diff);
-  const uint8x16_t ab4567_max = vmaxq_u8(ab45_diff, ab67_diff);
-  const uint8x16_t ab0123_min = vminq_u8(ab01_diff, ab23_diff);
-  const uint8x16_t ab4567_min = vminq_u8(ab45_diff, ab67_diff);
-
-  const uint8x16_t ab07_max = vmaxq_u8(ab0123_max, ab4567_max);
-  const uint8x16_t ab07_min = vminq_u8(ab0123_min, ab4567_min);
-
-  // Split to D and start doing pairwise.
-  uint8x8_t ab_max = vmax_u8(vget_high_u8(ab07_max), vget_low_u8(ab07_max));
-  uint8x8_t ab_min = vmin_u8(vget_high_u8(ab07_min), vget_low_u8(ab07_min));
-
-  // Enough runs of vpmax/min propogate the max/min values to every position.
-  ab_max = vpmax_u8(ab_max, ab_max);
-  ab_min = vpmin_u8(ab_min, ab_min);
-
-  ab_max = vpmax_u8(ab_max, ab_max);
-  ab_min = vpmin_u8(ab_min, ab_min);
-
-  ab_max = vpmax_u8(ab_max, ab_max);
-  ab_min = vpmin_u8(ab_min, ab_min);
-
-  *min = *max = 0;  // Clear high bits
-  // Store directly to avoid costly neon->gpr transfer.
-  vst1_lane_u8((uint8_t *)max, ab_max, 0);
-  vst1_lane_u8((uint8_t *)min, ab_min, 0);
-}

aom_dsp/arm/fwd_txfm_neon.c

@@ -1,221 +0,0 @@
-/*
- * Copyright (c) 2016, Alliance for Open Media. All rights reserved
- *
- * This source code is subject to the terms of the BSD 2 Clause License and
- * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
- * was not distributed with this source code in the LICENSE file, you can
- * obtain it at www.aomedia.org/license/software. If the Alliance for Open
- * Media Patent License 1.0 was not distributed with this source code in the
- * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
- */
-
-#include <arm_neon.h>
-
-#include "./aom_config.h"
-#include "aom_dsp/txfm_common.h"
-
-void aom_fdct8x8_neon(const int16_t *input, int16_t *final_output, int stride) {
-  int i;
-  // stage 1
-  int16x8_t input_0 = vshlq_n_s16(vld1q_s16(&input[0 * stride]), 2);
-  int16x8_t input_1 = vshlq_n_s16(vld1q_s16(&input[1 * stride]), 2);
-  int16x8_t input_2 = vshlq_n_s16(vld1q_s16(&input[2 * stride]), 2);
-  int16x8_t input_3 = vshlq_n_s16(vld1q_s16(&input[3 * stride]), 2);
-  int16x8_t input_4 = vshlq_n_s16(vld1q_s16(&input[4 * stride]), 2);
-  int16x8_t input_5 = vshlq_n_s16(vld1q_s16(&input[5 * stride]), 2);
-  int16x8_t input_6 = vshlq_n_s16(vld1q_s16(&input[6 * stride]), 2);
-  int16x8_t input_7 = vshlq_n_s16(vld1q_s16(&input[7 * stride]), 2);
-  for (i = 0; i < 2; ++i) {
-    int16x8_t out_0, out_1, out_2, out_3, out_4, out_5, out_6, out_7;
-    const int16x8_t v_s0 = vaddq_s16(input_0, input_7);
-    const int16x8_t v_s1 = vaddq_s16(input_1, input_6);
-    const int16x8_t v_s2 = vaddq_s16(input_2, input_5);
-    const int16x8_t v_s3 = vaddq_s16(input_3, input_4);
-    const int16x8_t v_s4 = vsubq_s16(input_3, input_4);
-    const int16x8_t v_s5 = vsubq_s16(input_2, input_5);
-    const int16x8_t v_s6 = vsubq_s16(input_1, input_6);
-    const int16x8_t v_s7 = vsubq_s16(input_0, input_7);
-    // fdct4(step, step);
-    int16x8_t v_x0 = vaddq_s16(v_s0, v_s3);
-    int16x8_t v_x1 = vaddq_s16(v_s1, v_s2);
-    int16x8_t v_x2 = vsubq_s16(v_s1, v_s2);
-    int16x8_t v_x3 = vsubq_s16(v_s0, v_s3);
-    // fdct4(step, step);
-    int32x4_t v_t0_lo = vaddl_s16(vget_low_s16(v_x0), vget_low_s16(v_x1));
-    int32x4_t v_t0_hi = vaddl_s16(vget_high_s16(v_x0), vget_high_s16(v_x1));
-    int32x4_t v_t1_lo = vsubl_s16(vget_low_s16(v_x0), vget_low_s16(v_x1));
-    int32x4_t v_t1_hi = vsubl_s16(vget_high_s16(v_x0), vget_high_s16(v_x1));
-    int32x4_t v_t2_lo = vmull_n_s16(vget_low_s16(v_x2), (int16_t)cospi_24_64);
-    int32x4_t v_t2_hi = vmull_n_s16(vget_high_s16(v_x2), (int16_t)cospi_24_64);
-    int32x4_t v_t3_lo = vmull_n_s16(vget_low_s16(v_x3), (int16_t)cospi_24_64);
-    int32x4_t v_t3_hi = vmull_n_s16(vget_high_s16(v_x3), (int16_t)cospi_24_64);
-    v_t2_lo = vmlal_n_s16(v_t2_lo, vget_low_s16(v_x3), (int16_t)cospi_8_64);
-    v_t2_hi = vmlal_n_s16(v_t2_hi, vget_high_s16(v_x3), (int16_t)cospi_8_64);
-    v_t3_lo = vmlsl_n_s16(v_t3_lo, vget_low_s16(v_x2), (int16_t)cospi_8_64);
-    v_t3_hi = vmlsl_n_s16(v_t3_hi, vget_high_s16(v_x2), (int16_t)cospi_8_64);
-    v_t0_lo = vmulq_n_s32(v_t0_lo, (int32_t)cospi_16_64);
-    v_t0_hi = vmulq_n_s32(v_t0_hi, (int32_t)cospi_16_64);
-    v_t1_lo = vmulq_n_s32(v_t1_lo, (int32_t)cospi_16_64);
-    v_t1_hi = vmulq_n_s32(v_t1_hi, (int32_t)cospi_16_64);
-    {
-      const int16x4_t a = vrshrn_n_s32(v_t0_lo, DCT_CONST_BITS);
-      const int16x4_t b = vrshrn_n_s32(v_t0_hi, DCT_CONST_BITS);
-      const int16x4_t c = vrshrn_n_s32(v_t1_lo, DCT_CONST_BITS);
-      const int16x4_t d = vrshrn_n_s32(v_t1_hi, DCT_CONST_BITS);
-      const int16x4_t e = vrshrn_n_s32(v_t2_lo, DCT_CONST_BITS);
-      const int16x4_t f = vrshrn_n_s32(v_t2_hi, DCT_CONST_BITS);
-      const int16x4_t g = vrshrn_n_s32(v_t3_lo, DCT_CONST_BITS);
-      const int16x4_t h = vrshrn_n_s32(v_t3_hi, DCT_CONST_BITS);
-      out_0 = vcombine_s16(a, c);  // 00 01 02 03 40 41 42 43
-      out_2 = vcombine_s16(e, g);  // 20 21 22 23 60 61 62 63
-      out_4 = vcombine_s16(b, d);  // 04 05 06 07 44 45 46 47
-      out_6 = vcombine_s16(f, h);  // 24 25 26 27 64 65 66 67
-    }
-    // Stage 2
-    v_x0 = vsubq_s16(v_s6, v_s5);
-    v_x1 = vaddq_s16(v_s6, v_s5);
-    v_t0_lo = vmull_n_s16(vget_low_s16(v_x0), (int16_t)cospi_16_64);
-    v_t0_hi = vmull_n_s16(vget_high_s16(v_x0), (int16_t)cospi_16_64);
-    v_t1_lo = vmull_n_s16(vget_low_s16(v_x1), (int16_t)cospi_16_64);
-    v_t1_hi = vmull_n_s16(vget_high_s16(v_x1), (int16_t)cospi_16_64);
-    {
-      const int16x4_t a = vrshrn_n_s32(v_t0_lo, DCT_CONST_BITS);
-      const int16x4_t b = vrshrn_n_s32(v_t0_hi, DCT_CONST_BITS);
-      const int16x4_t c = vrshrn_n_s32(v_t1_lo, DCT_CONST_BITS);
-      const int16x4_t d = vrshrn_n_s32(v_t1_hi, DCT_CONST_BITS);
-      const int16x8_t ab = vcombine_s16(a, b);
-      const int16x8_t cd = vcombine_s16(c, d);
-      // Stage 3
-      v_x0 = vaddq_s16(v_s4, ab);
-      v_x1 = vsubq_s16(v_s4, ab);
-      v_x2 = vsubq_s16(v_s7, cd);
-      v_x3 = vaddq_s16(v_s7, cd);
-    }
-    // Stage 4
-    v_t0_lo = vmull_n_s16(vget_low_s16(v_x3), (int16_t)cospi_4_64);
-    v_t0_hi = vmull_n_s16(vget_high_s16(v_x3), (int16_t)cospi_4_64);
-    v_t0_lo = vmlal_n_s16(v_t0_lo, vget_low_s16(v_x0), (int16_t)cospi_28_64);
-    v_t0_hi = vmlal_n_s16(v_t0_hi, vget_high_s16(v_x0), (int16_t)cospi_28_64);
-    v_t1_lo = vmull_n_s16(vget_low_s16(v_x1), (int16_t)cospi_12_64);
-    v_t1_hi = vmull_n_s16(vget_high_s16(v_x1), (int16_t)cospi_12_64);
-    v_t1_lo = vmlal_n_s16(v_t1_lo, vget_low_s16(v_x2), (int16_t)cospi_20_64);
-    v_t1_hi = vmlal_n_s16(v_t1_hi, vget_high_s16(v_x2), (int16_t)cospi_20_64);
-    v_t2_lo = vmull_n_s16(vget_low_s16(v_x2), (int16_t)cospi_12_64);
-    v_t2_hi = vmull_n_s16(vget_high_s16(v_x2), (int16_t)cospi_12_64);
-    v_t2_lo = vmlsl_n_s16(v_t2_lo, vget_low_s16(v_x1), (int16_t)cospi_20_64);
-    v_t2_hi = vmlsl_n_s16(v_t2_hi, vget_high_s16(v_x1), (int16_t)cospi_20_64);
-    v_t3_lo = vmull_n_s16(vget_low_s16(v_x3), (int16_t)cospi_28_64);
-    v_t3_hi = vmull_n_s16(vget_high_s16(v_x3), (int16_t)cospi_28_64);
-    v_t3_lo = vmlsl_n_s16(v_t3_lo, vget_low_s16(v_x0), (int16_t)cospi_4_64);
-    v_t3_hi = vmlsl_n_s16(v_t3_hi, vget_high_s16(v_x0), (int16_t)cospi_4_64);
-    {
-      const int16x4_t a = vrshrn_n_s32(v_t0_lo, DCT_CONST_BITS);
-      const int16x4_t b = vrshrn_n_s32(v_t0_hi, DCT_CONST_BITS);
-      const int16x4_t c = vrshrn_n_s32(v_t1_lo, DCT_CONST_BITS);
-      const int16x4_t d = vrshrn_n_s32(v_t1_hi, DCT_CONST_BITS);
-      const int16x4_t e = vrshrn_n_s32(v_t2_lo, DCT_CONST_BITS);
-      const int16x4_t f = vrshrn_n_s32(v_t2_hi, DCT_CONST_BITS);
-      const int16x4_t g = vrshrn_n_s32(v_t3_lo, DCT_CONST_BITS);
-      const int16x4_t h = vrshrn_n_s32(v_t3_hi, DCT_CONST_BITS);
-      out_1 = vcombine_s16(a, c);  // 10 11 12 13 50 51 52 53
-      out_3 = vcombine_s16(e, g);  // 30 31 32 33 70 71 72 73
-      out_5 = vcombine_s16(b, d);  // 14 15 16 17 54 55 56 57
-      out_7 = vcombine_s16(f, h);  // 34 35 36 37 74 75 76 77
-    }
-    // transpose 8x8
-    {
-      // 00 01 02 03 40 41 42 43
-      // 10 11 12 13 50 51 52 53
-      // 20 21 22 23 60 61 62 63
-      // 30 31 32 33 70 71 72 73
-      // 04 05 06 07 44 45 46 47
-      // 14 15 16 17 54 55 56 57
-      // 24 25 26 27 64 65 66 67
-      // 34 35 36 37 74 75 76 77
-      const int32x4x2_t r02_s32 =
-          vtrnq_s32(vreinterpretq_s32_s16(out_0), vreinterpretq_s32_s16(out_2));
-      const int32x4x2_t r13_s32 =
-          vtrnq_s32(vreinterpretq_s32_s16(out_1), vreinterpretq_s32_s16(out_3));
-      const int32x4x2_t r46_s32 =
-          vtrnq_s32(vreinterpretq_s32_s16(out_4), vreinterpretq_s32_s16(out_6));
-      const int32x4x2_t r57_s32 =
-          vtrnq_s32(vreinterpretq_s32_s16(out_5), vreinterpretq_s32_s16(out_7));
-      const int16x8x2_t r01_s16 =
-          vtrnq_s16(vreinterpretq_s16_s32(r02_s32.val[0]),
-                    vreinterpretq_s16_s32(r13_s32.val[0]));
-      const int16x8x2_t r23_s16 =
-          vtrnq_s16(vreinterpretq_s16_s32(r02_s32.val[1]),
-                    vreinterpretq_s16_s32(r13_s32.val[1]));
-      const int16x8x2_t r45_s16 =
-          vtrnq_s16(vreinterpretq_s16_s32(r46_s32.val[0]),
-                    vreinterpretq_s16_s32(r57_s32.val[0]));
-      const int16x8x2_t r67_s16 =
-          vtrnq_s16(vreinterpretq_s16_s32(r46_s32.val[1]),
-                    vreinterpretq_s16_s32(r57_s32.val[1]));
-      input_0 = r01_s16.val[0];
-      input_1 = r01_s16.val[1];
-      input_2 = r23_s16.val[0];
-      input_3 = r23_s16.val[1];
-      input_4 = r45_s16.val[0];
-      input_5 = r45_s16.val[1];
-      input_6 = r67_s16.val[0];
-      input_7 = r67_s16.val[1];
-      // 00 10 20 30 40 50 60 70
-      // 01 11 21 31 41 51 61 71
-      // 02 12 22 32 42 52 62 72
-      // 03 13 23 33 43 53 63 73
-      // 04 14 24 34 44 54 64 74
-      // 05 15 25 35 45 55 65 75
-      // 06 16 26 36 46 56 66 76
-      // 07 17 27 37 47 57 67 77
-    }
-  }  // for
-  {
-    // from aom_dct_sse2.c
-    // Post-condition (division by two)
-    //    division of two 16 bits signed numbers using shifts
-    //    n / 2 = (n - (n >> 15)) >> 1
-    const int16x8_t sign_in0 = vshrq_n_s16(input_0, 15);
-    const int16x8_t sign_in1 = vshrq_n_s16(input_1, 15);
-    const int16x8_t sign_in2 = vshrq_n_s16(input_2, 15);
-    const int16x8_t sign_in3 = vshrq_n_s16(input_3, 15);
-    const int16x8_t sign_in4 = vshrq_n_s16(input_4, 15);
-    const int16x8_t sign_in5 = vshrq_n_s16(input_5, 15);
-    const int16x8_t sign_in6 = vshrq_n_s16(input_6, 15);
-    const int16x8_t sign_in7 = vshrq_n_s16(input_7, 15);
-    input_0 = vhsubq_s16(input_0, sign_in0);
-    input_1 = vhsubq_s16(input_1, sign_in1);
-    input_2 = vhsubq_s16(input_2, sign_in2);
-    input_3 = vhsubq_s16(input_3, sign_in3);
-    input_4 = vhsubq_s16(input_4, sign_in4);
-    input_5 = vhsubq_s16(input_5, sign_in5);
-    input_6 = vhsubq_s16(input_6, sign_in6);
-    input_7 = vhsubq_s16(input_7, sign_in7);
-    // store results
-    vst1q_s16(&final_output[0 * 8], input_0);
-    vst1q_s16(&final_output[1 * 8], input_1);
-    vst1q_s16(&final_output[2 * 8], input_2);
-    vst1q_s16(&final_output[3 * 8], input_3);
-    vst1q_s16(&final_output[4 * 8], input_4);
-    vst1q_s16(&final_output[5 * 8], input_5);
-    vst1q_s16(&final_output[6 * 8], input_6);
-    vst1q_s16(&final_output[7 * 8], input_7);
-  }
-}
-
-void aom_fdct8x8_1_neon(const int16_t *input, int16_t *output, int stride) {
-  int r;
-  int16x8_t sum = vld1q_s16(&input[0]);
-  for (r = 1; r < 8; ++r) {
-    const int16x8_t input_00 = vld1q_s16(&input[r * stride]);
-    sum = vaddq_s16(sum, input_00);
-  }
-  {
-    const int32x4_t a = vpaddlq_s16(sum);
-    const int64x2_t b = vpaddlq_s32(a);
-    const int32x2_t c = vadd_s32(vreinterpret_s32_s64(vget_low_s64(b)),
-                                 vreinterpret_s32_s64(vget_high_s64(b)));
-    output[0] = vget_lane_s16(vreinterpret_s16_s32(c), 0);
-    output[1] = 0;
-  }
-}

aom_dsp/arm/hadamard_neon.c

@@ -1,199 +0,0 @@
-/*
- *  Copyright (c) 2016 The WebM project authors. All Rights Reserved.
- *
- *  Use of this source code is governed by a BSD-style license
- *  that can be found in the LICENSE file in the root of the source
- *  tree. An additional intellectual property rights grant can be found
- *  in the file PATENTS.  All contributing project authors may
- *  be found in the AUTHORS file in the root of the source tree.
- */
-
-#include <arm_neon.h>
-
-#include "./aom_dsp_rtcd.h"
-
-static void hadamard8x8_one_pass(int16x8_t *a0, int16x8_t *a1, int16x8_t *a2,
-                                 int16x8_t *a3, int16x8_t *a4, int16x8_t *a5,
-                                 int16x8_t *a6, int16x8_t *a7) {
-  const int16x8_t b0 = vaddq_s16(*a0, *a1);
-  const int16x8_t b1 = vsubq_s16(*a0, *a1);
-  const int16x8_t b2 = vaddq_s16(*a2, *a3);
-  const int16x8_t b3 = vsubq_s16(*a2, *a3);
-  const int16x8_t b4 = vaddq_s16(*a4, *a5);
-  const int16x8_t b5 = vsubq_s16(*a4, *a5);
-  const int16x8_t b6 = vaddq_s16(*a6, *a7);
-  const int16x8_t b7 = vsubq_s16(*a6, *a7);
-
-  const int16x8_t c0 = vaddq_s16(b0, b2);
-  const int16x8_t c1 = vaddq_s16(b1, b3);
-  const int16x8_t c2 = vsubq_s16(b0, b2);
-  const int16x8_t c3 = vsubq_s16(b1, b3);
-  const int16x8_t c4 = vaddq_s16(b4, b6);
-  const int16x8_t c5 = vaddq_s16(b5, b7);
-  const int16x8_t c6 = vsubq_s16(b4, b6);
-  const int16x8_t c7 = vsubq_s16(b5, b7);
-
-  *a0 = vaddq_s16(c0, c4);
-  *a1 = vsubq_s16(c2, c6);
-  *a2 = vsubq_s16(c0, c4);
-  *a3 = vaddq_s16(c2, c6);
-  *a4 = vaddq_s16(c3, c7);
-  *a5 = vsubq_s16(c3, c7);
-  *a6 = vsubq_s16(c1, c5);
-  *a7 = vaddq_s16(c1, c5);
-}
-
-// TODO(johannkoenig): Make a transpose library and dedup with idct. Consider
-// reversing transpose order which may make it easier for the compiler to
-// reconcile the vtrn.64 moves.
-static void transpose8x8(int16x8_t *a0, int16x8_t *a1, int16x8_t *a2,
-                         int16x8_t *a3, int16x8_t *a4, int16x8_t *a5,
-                         int16x8_t *a6, int16x8_t *a7) {
-  // Swap 64 bit elements. Goes from:
-  // a0: 00 01 02 03 04 05 06 07
-  // a1: 08 09 10 11 12 13 14 15
-  // a2: 16 17 18 19 20 21 22 23
-  // a3: 24 25 26 27 28 29 30 31
-  // a4: 32 33 34 35 36 37 38 39
-  // a5: 40 41 42 43 44 45 46 47
-  // a6: 48 49 50 51 52 53 54 55
-  // a7: 56 57 58 59 60 61 62 63
-  // to:
-  // a04_lo: 00 01 02 03 32 33 34 35
-  // a15_lo: 08 09 10 11 40 41 42 43
-  // a26_lo: 16 17 18 19 48 49 50 51
-  // a37_lo: 24 25 26 27 56 57 58 59
-  // a04_hi: 04 05 06 07 36 37 38 39
-  // a15_hi: 12 13 14 15 44 45 46 47
-  // a26_hi: 20 21 22 23 52 53 54 55
-  // a37_hi: 28 29 30 31 60 61 62 63
-  const int16x8_t a04_lo = vcombine_s16(vget_low_s16(*a0), vget_low_s16(*a4));
-  const int16x8_t a15_lo = vcombine_s16(vget_low_s16(*a1), vget_low_s16(*a5));
-  const int16x8_t a26_lo = vcombine_s16(vget_low_s16(*a2), vget_low_s16(*a6));
-  const int16x8_t a37_lo = vcombine_s16(vget_low_s16(*a3), vget_low_s16(*a7));
-  const int16x8_t a04_hi = vcombine_s16(vget_high_s16(*a0), vget_high_s16(*a4));
-  const int16x8_t a15_hi = vcombine_s16(vget_high_s16(*a1), vget_high_s16(*a5));
-  const int16x8_t a26_hi = vcombine_s16(vget_high_s16(*a2), vget_high_s16(*a6));
-  const int16x8_t a37_hi = vcombine_s16(vget_high_s16(*a3), vget_high_s16(*a7));
-
-  // Swap 32 bit elements resulting in:
-  // a0246_lo:
-  // 00 01 16 17 32 33 48 49
-  // 02 03 18 19 34 35 50 51
-  // a1357_lo:
-  // 08 09 24 25 40 41 56 57
-  // 10 11 26 27 42 43 58 59
-  // a0246_hi:
-  // 04 05 20 21 36 37 52 53
-  // 06 07 22 23 38 39 54 55
-  // a1657_hi:
-  // 12 13 28 29 44 45 60 61
-  // 14 15 30 31 46 47 62 63
-  const int32x4x2_t a0246_lo =
-      vtrnq_s32(vreinterpretq_s32_s16(a04_lo), vreinterpretq_s32_s16(a26_lo));
-  const int32x4x2_t a1357_lo =
-      vtrnq_s32(vreinterpretq_s32_s16(a15_lo), vreinterpretq_s32_s16(a37_lo));
-  const int32x4x2_t a0246_hi =
-      vtrnq_s32(vreinterpretq_s32_s16(a04_hi), vreinterpretq_s32_s16(a26_hi));
-  const int32x4x2_t a1357_hi =
-      vtrnq_s32(vreinterpretq_s32_s16(a15_hi), vreinterpretq_s32_s16(a37_hi));
-
-  // Swap 16 bit elements resulting in:
-  // b0:
-  // 00 08 16 24 32 40 48 56
-  // 01 09 17 25 33 41 49 57
-  // b1:
-  // 02 10 18 26 34 42 50 58
-  // 03 11 19 27 35 43 51 59
-  // b2:
-  // 04 12 20 28 36 44 52 60
-  // 05 13 21 29 37 45 53 61
-  // b3:
-  // 06 14 22 30 38 46 54 62
-  // 07 15 23 31 39 47 55 63
-  const int16x8x2_t b0 = vtrnq_s16(vreinterpretq_s16_s32(a0246_lo.val[0]),
-                                   vreinterpretq_s16_s32(a1357_lo.val[0]));
-  const int16x8x2_t b1 = vtrnq_s16(vreinterpretq_s16_s32(a0246_lo.val[1]),
-                                   vreinterpretq_s16_s32(a1357_lo.val[1]));
-  const int16x8x2_t b2 = vtrnq_s16(vreinterpretq_s16_s32(a0246_hi.val[0]),
-                                   vreinterpretq_s16_s32(a1357_hi.val[0]));
-  const int16x8x2_t b3 = vtrnq_s16(vreinterpretq_s16_s32(a0246_hi.val[1]),
-                                   vreinterpretq_s16_s32(a1357_hi.val[1]));
-
-  *a0 = b0.val[0];
-  *a1 = b0.val[1];
-  *a2 = b1.val[0];
-  *a3 = b1.val[1];
-  *a4 = b2.val[0];
-  *a5 = b2.val[1];
-  *a6 = b3.val[0];
-  *a7 = b3.val[1];
-}
-
-void aom_hadamard_8x8_neon(const int16_t *src_diff, int src_stride,
-                           int16_t *coeff) {
-  int16x8_t a0 = vld1q_s16(src_diff);
-  int16x8_t a1 = vld1q_s16(src_diff + src_stride);
-  int16x8_t a2 = vld1q_s16(src_diff + 2 * src_stride);
-  int16x8_t a3 = vld1q_s16(src_diff + 3 * src_stride);
-  int16x8_t a4 = vld1q_s16(src_diff + 4 * src_stride);
-  int16x8_t a5 = vld1q_s16(src_diff + 5 * src_stride);
-  int16x8_t a6 = vld1q_s16(src_diff + 6 * src_stride);
-  int16x8_t a7 = vld1q_s16(src_diff + 7 * src_stride);
-
-  hadamard8x8_one_pass(&a0, &a1, &a2, &a3, &a4, &a5, &a6, &a7);
-
-  transpose8x8(&a0, &a1, &a2, &a3, &a4, &a5, &a6, &a7);
-
-  hadamard8x8_one_pass(&a0, &a1, &a2, &a3, &a4, &a5, &a6, &a7);
-
-  // Skip the second transpose because it is not required.
-
-  vst1q_s16(coeff + 0, a0);
-  vst1q_s16(coeff + 8, a1);
-  vst1q_s16(coeff + 16, a2);
-  vst1q_s16(coeff + 24, a3);
-  vst1q_s16(coeff + 32, a4);
-  vst1q_s16(coeff + 40, a5);
-  vst1q_s16(coeff + 48, a6);
-  vst1q_s16(coeff + 56, a7);
-}
-
-void aom_hadamard_16x16_neon(const int16_t *src_diff, int src_stride,
-                             int16_t *coeff) {
-  int i;
-
-  /* Rearrange 16x16 to 8x32 and remove stride.
-   * Top left first. */
-  aom_hadamard_8x8_neon(src_diff + 0 + 0 * src_stride, src_stride, coeff + 0);
-  /* Top right. */
-  aom_hadamard_8x8_neon(src_diff + 8 + 0 * src_stride, src_stride, coeff + 64);
-  /* Bottom left. */
-  aom_hadamard_8x8_neon(src_diff + 0 + 8 * src_stride, src_stride, coeff + 128);
-  /* Bottom right. */
-  aom_hadamard_8x8_neon(src_diff + 8 + 8 * src_stride, src_stride, coeff + 192);
-
-  for (i = 0; i < 64; i += 8) {
-    const int16x8_t a0 = vld1q_s16(coeff + 0);
-    const int16x8_t a1 = vld1q_s16(coeff + 64);
-    const int16x8_t a2 = vld1q_s16(coeff + 128);
-    const int16x8_t a3 = vld1q_s16(coeff + 192);
-
-    const int16x8_t b0 = vhaddq_s16(a0, a1);
-    const int16x8_t b1 = vhsubq_s16(a0, a1);
-    const int16x8_t b2 = vhaddq_s16(a2, a3);
-    const int16x8_t b3 = vhsubq_s16(a2, a3);
-
-    const int16x8_t c0 = vaddq_s16(b0, b2);
-    const int16x8_t c1 = vaddq_s16(b1, b3);
-    const int16x8_t c2 = vsubq_s16(b0, b2);
-    const int16x8_t c3 = vsubq_s16(b1, b3);
-
-    vst1q_s16(coeff + 0, c0);
-    vst1q_s16(coeff + 64, c1);
-    vst1q_s16(coeff + 128, c2);
-    vst1q_s16(coeff + 192, c3);
-
-    coeff += 8;
-  }
-}

aom_dsp/arm/idct16x16_1_add_neon.c

@@ -1,59 +0,0 @@
-/*
- * Copyright (c) 2016, Alliance for Open Media. All rights reserved
- *
- * This source code is subject to the terms of the BSD 2 Clause License and
- * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
- * was not distributed with this source code in the LICENSE file, you can
- * obtain it at www.aomedia.org/license/software. If the Alliance for Open
- * Media Patent License 1.0 was not distributed with this source code in the
- * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
- */
-
-#include <arm_neon.h>
-
-#include "aom_dsp/inv_txfm.h"
-#include "aom_ports/mem.h"
-
-void aom_idct16x16_1_add_neon(int16_t *input, uint8_t *dest, int dest_stride) {
-  uint8x8_t d2u8, d3u8, d30u8, d31u8;
-  uint64x1_t d2u64, d3u64, d4u64, d5u64;
-  uint16x8_t q0u16, q9u16, q10u16, q11u16, q12u16;
-  int16x8_t q0s16;
-  uint8_t *d1, *d2;
-  int16_t i, j, a1;
-  int16_t out = dct_const_round_shift(input[0] * cospi_16_64);
-  out = dct_const_round_shift(out * cospi_16_64);
-  a1 = ROUND_POWER_OF_TWO(out, 6);
-
-  q0s16 = vdupq_n_s16(a1);
-  q0u16 = vreinterpretq_u16_s16(q0s16);
-
-  for (d1 = d2 = dest, i = 0; i < 4; i++) {
-    for (j = 0; j < 2; j++) {
-      d2u64 = vld1_u64((const uint64_t *)d1);
-      d3u64 = vld1_u64((const uint64_t *)(d1 + 8));
-      d1 += dest_stride;
-      d4u64 = vld1_u64((const uint64_t *)d1);
-      d5u64 = vld1_u64((const uint64_t *)(d1 + 8));
-      d1 += dest_stride;
-
-      q9u16 = vaddw_u8(q0u16, vreinterpret_u8_u64(d2u64));
-      q10u16 = vaddw_u8(q0u16, vreinterpret_u8_u64(d3u64));
-      q11u16 = vaddw_u8(q0u16, vreinterpret_u8_u64(d4u64));
-      q12u16 = vaddw_u8(q0u16, vreinterpret_u8_u64(d5u64));
-
-      d2u8 = vqmovun_s16(vreinterpretq_s16_u16(q9u16));
-      d3u8 = vqmovun_s16(vreinterpretq_s16_u16(q10u16));
-      d30u8 = vqmovun_s16(vreinterpretq_s16_u16(q11u16));
-      d31u8 = vqmovun_s16(vreinterpretq_s16_u16(q12u16));
-
-      vst1_u64((uint64_t *)d2, vreinterpret_u64_u8(d2u8));
-      vst1_u64((uint64_t *)(d2 + 8), vreinterpret_u64_u8(d3u8));
-      d2 += dest_stride;
-      vst1_u64((uint64_t *)d2, vreinterpret_u64_u8(d30u8));
-      vst1_u64((uint64_t *)(d2 + 8), vreinterpret_u64_u8(d31u8));
-      d2 += dest_stride;
-    }
-  }
-  return;
-}

aom_dsp/arm/idct16x16_neon.c

@@ -1,152 +0,0 @@
-/*
- * Copyright (c) 2016, Alliance for Open Media. All rights reserved
- *
- * This source code is subject to the terms of the BSD 2 Clause License and
- * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
- * was not distributed with this source code in the LICENSE file, you can
- * obtain it at www.aomedia.org/license/software. If the Alliance for Open
- * Media Patent License 1.0 was not distributed with this source code in the
- * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
- */
-
-#include "aom_dsp/aom_dsp_common.h"
-
-void aom_idct16x16_256_add_neon_pass1(const int16_t *input, int16_t *output,
-                                      int output_stride);
-void aom_idct16x16_256_add_neon_pass2(const int16_t *src, int16_t *output,
-                                      int16_t *pass1Output, int16_t skip_adding,
-                                      uint8_t *dest, int dest_stride);
-void aom_idct16x16_10_add_neon_pass1(const int16_t *input, int16_t *output,
-                                     int output_stride);
-void aom_idct16x16_10_add_neon_pass2(const int16_t *src, int16_t *output,
-                                     int16_t *pass1Output, int16_t skip_adding,
-                                     uint8_t *dest, int dest_stride);
-
-#if HAVE_NEON_ASM
-/* For ARM NEON, d8-d15 are callee-saved registers, and need to be saved. */
-extern void aom_push_neon(int64_t *store);
-extern void aom_pop_neon(int64_t *store);
-#endif  // HAVE_NEON_ASM
-
-void aom_idct16x16_256_add_neon(const int16_t *input, uint8_t *dest,
-                                int dest_stride) {
-#if HAVE_NEON_ASM
-  int64_t store_reg[8];
-#endif
-  int16_t pass1_output[16 * 16] = { 0 };
-  int16_t row_idct_output[16 * 16] = { 0 };
-
-#if HAVE_NEON_ASM
-  // save d8-d15 register values.
-  aom_push_neon(store_reg);
-#endif
-
-  /* Parallel idct on the upper 8 rows */
-  // First pass processes even elements 0, 2, 4, 6, 8, 10, 12, 14 and save the
-  // stage 6 result in pass1_output.
-  aom_idct16x16_256_add_neon_pass1(input, pass1_output, 8);
-
-  // Second pass processes odd elements 1, 3, 5, 7, 9, 11, 13, 15 and combines
-  // with result in pass1(pass1_output) to calculate final result in stage 7
-  // which will be saved into row_idct_output.
-  aom_idct16x16_256_add_neon_pass2(input + 1, row_idct_output, pass1_output, 0,
-                                   dest, dest_stride);
-
-  /* Parallel idct on the lower 8 rows */
-  // First pass processes even elements 0, 2, 4, 6, 8, 10, 12, 14 and save the
-  // stage 6 result in pass1_output.
-  aom_idct16x16_256_add_neon_pass1(input + 8 * 16, pass1_output, 8);
-
-  // Second pass processes odd elements 1, 3, 5, 7, 9, 11, 13, 15 and combines
-  // with result in pass1(pass1_output) to calculate final result in stage 7
-  // which will be saved into row_idct_output.
-  aom_idct16x16_256_add_neon_pass2(input + 8 * 16 + 1, row_idct_output + 8,
-                                   pass1_output, 0, dest, dest_stride);
-
-  /* Parallel idct on the left 8 columns */
-  // First pass processes even elements 0, 2, 4, 6, 8, 10, 12, 14 and save the
-  // stage 6 result in pass1_output.
-  aom_idct16x16_256_add_neon_pass1(row_idct_output, pass1_output, 8);
-
-  // Second pass processes odd elements 1, 3, 5, 7, 9, 11, 13, 15 and combines
-  // with result in pass1(pass1_output) to calculate final result in stage 7.
-  // Then add the result to the destination data.
-  aom_idct16x16_256_add_neon_pass2(row_idct_output + 1, row_idct_output,
-                                   pass1_output, 1, dest, dest_stride);
-
-  /* Parallel idct on the right 8 columns */
-  // First pass processes even elements 0, 2, 4, 6, 8, 10, 12, 14 and save the
-  // stage 6 result in pass1_output.
-  aom_idct16x16_256_add_neon_pass1(row_idct_output + 8 * 16, pass1_output, 8);
-
-  // Second pass processes odd elements 1, 3, 5, 7, 9, 11, 13, 15 and combines
-  // with result in pass1(pass1_output) to calculate final result in stage 7.
-  // Then add the result to the destination data.
-  aom_idct16x16_256_add_neon_pass2(row_idct_output + 8 * 16 + 1,
-                                   row_idct_output + 8, pass1_output, 1,
-                                   dest + 8, dest_stride);
-
-#if HAVE_NEON_ASM
-  // restore d8-d15 register values.
-  aom_pop_neon(store_reg);
-#endif
-
-  return;
-}
-
-void aom_idct16x16_10_add_neon(const int16_t *input, uint8_t *dest,
-                               int dest_stride) {
-#if HAVE_NEON_ASM
-  int64_t store_reg[8];
-#endif
-  int16_t pass1_output[16 * 16] = { 0 };
-  int16_t row_idct_output[16 * 16] = { 0 };
-
-#if HAVE_NEON_ASM
-  // save d8-d15 register values.
-  aom_push_neon(store_reg);
-#endif
-
-  /* Parallel idct on the upper 8 rows */
-  // First pass processes even elements 0, 2, 4, 6, 8, 10, 12, 14 and save the
-  // stage 6 result in pass1_output.
-  aom_idct16x16_10_add_neon_pass1(input, pass1_output, 8);
-
-  // Second pass processes odd elements 1, 3, 5, 7, 9, 11, 13, 15 and combines
-  // with result in pass1(pass1_output) to calculate final result in stage 7
-  // which will be saved into row_idct_output.
-  aom_idct16x16_10_add_neon_pass2(input + 1, row_idct_output, pass1_output, 0,
-                                  dest, dest_stride);
-
-  /* Skip Parallel idct on the lower 8 rows as they are all 0s */
-
-  /* Parallel idct on the left 8 columns */
-  // First pass processes even elements 0, 2, 4, 6, 8, 10, 12, 14 and save the
-  // stage 6 result in pass1_output.
-  aom_idct16x16_256_add_neon_pass1(row_idct_output, pass1_output, 8);
-
-  // Second pass processes odd elements 1, 3, 5, 7, 9, 11, 13, 15 and combines
-  // with result in pass1(pass1_output) to calculate final result in stage 7.
-  // Then add the result to the destination data.
-  aom_idct16x16_256_add_neon_pass2(row_idct_output + 1, row_idct_output,
-                                   pass1_output, 1, dest, dest_stride);
-
-  /* Parallel idct on the right 8 columns */
-  // First pass processes even elements 0, 2, 4, 6, 8, 10, 12, 14 and save the
-  // stage 6 result in pass1_output.
-  aom_idct16x16_256_add_neon_pass1(row_idct_output + 8 * 16, pass1_output, 8);
-
-  // Second pass processes odd elements 1, 3, 5, 7, 9, 11, 13, 15 and combines
-  // with result in pass1(pass1_output) to calculate final result in stage 7.
-  // Then add the result to the destination data.
-  aom_idct16x16_256_add_neon_pass2(row_idct_output + 8 * 16 + 1,
-                                   row_idct_output + 8, pass1_output, 1,
-                                   dest + 8, dest_stride);
-
-#if HAVE_NEON_ASM
-  // restore d8-d15 register values.
-  aom_pop_neon(store_reg);
-#endif
-
-  return;
-}

aom_dsp/arm/idct32x32_1_add_neon.c

@@ -1,141 +0,0 @@
-/*
- * Copyright (c) 2016, Alliance for Open Media. All rights reserved
- *
- * This source code is subject to the terms of the BSD 2 Clause License and
- * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
- * was not distributed with this source code in the LICENSE file, you can
- * obtain it at www.aomedia.org/license/software. If the Alliance for Open
- * Media Patent License 1.0 was not distributed with this source code in the
- * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
- */
-
-#include <arm_neon.h>
-
-#include "./aom_config.h"
-
-#include "aom_dsp/inv_txfm.h"
-#include "aom_ports/mem.h"
-
-static INLINE void LD_16x8(uint8_t *d, int d_stride, uint8x16_t *q8u8,
-                           uint8x16_t *q9u8, uint8x16_t *q10u8,
-                           uint8x16_t *q11u8, uint8x16_t *q12u8,
-                           uint8x16_t *q13u8, uint8x16_t *q14u8,
-                           uint8x16_t *q15u8) {
-  *q8u8 = vld1q_u8(d);
-  d += d_stride;
-  *q9u8 = vld1q_u8(d);
-  d += d_stride;
-  *q10u8 = vld1q_u8(d);
-  d += d_stride;
-  *q11u8 = vld1q_u8(d);
-  d += d_stride;
-  *q12u8 = vld1q_u8(d);
-  d += d_stride;
-  *q13u8 = vld1q_u8(d);
-  d += d_stride;
-  *q14u8 = vld1q_u8(d);
-  d += d_stride;
-  *q15u8 = vld1q_u8(d);
-  return;
-}
-
-static INLINE void ADD_DIFF_16x8(uint8x16_t qdiffu8, uint8x16_t *q8u8,
-                                 uint8x16_t *q9u8, uint8x16_t *q10u8,
-                                 uint8x16_t *q11u8, uint8x16_t *q12u8,
-                                 uint8x16_t *q13u8, uint8x16_t *q14u8,
-                                 uint8x16_t *q15u8) {
-  *q8u8 = vqaddq_u8(*q8u8, qdiffu8);
-  *q9u8 = vqaddq_u8(*q9u8, qdiffu8);
-  *q10u8 = vqaddq_u8(*q10u8, qdiffu8);
-  *q11u8 = vqaddq_u8(*q11u8, qdiffu8);
-  *q12u8 = vqaddq_u8(*q12u8, qdiffu8);
-  *q13u8 = vqaddq_u8(*q13u8, qdiffu8);
-  *q14u8 = vqaddq_u8(*q14u8, qdiffu8);
-  *q15u8 = vqaddq_u8(*q15u8, qdiffu8);
-  return;
-}
-
-static INLINE void SUB_DIFF_16x8(uint8x16_t qdiffu8, uint8x16_t *q8u8,
-                                 uint8x16_t *q9u8, uint8x16_t *q10u8,
-                                 uint8x16_t *q11u8, uint8x16_t *q12u8,
-                                 uint8x16_t *q13u8, uint8x16_t *q14u8,
-                                 uint8x16_t *q15u8) {
-  *q8u8 = vqsubq_u8(*q8u8, qdiffu8);
-  *q9u8 = vqsubq_u8(*q9u8, qdiffu8);
-  *q10u8 = vqsubq_u8(*q10u8, qdiffu8);
-  *q11u8 = vqsubq_u8(*q11u8, qdiffu8);
-  *q12u8 = vqsubq_u8(*q12u8, qdiffu8);
-  *q13u8 = vqsubq_u8(*q13u8, qdiffu8);
-  *q14u8 = vqsubq_u8(*q14u8, qdiffu8);
-  *q15u8 = vqsubq_u8(*q15u8, qdiffu8);
-  return;
-}
-
-static INLINE void ST_16x8(uint8_t *d, int d_stride, uint8x16_t *q8u8,
-                           uint8x16_t *q9u8, uint8x16_t *q10u8,
-                           uint8x16_t *q11u8, uint8x16_t *q12u8,
-                           uint8x16_t *q13u8, uint8x16_t *q14u8,
-                           uint8x16_t *q15u8) {
-  vst1q_u8(d, *q8u8);
-  d += d_stride;
-  vst1q_u8(d, *q9u8);
-  d += d_stride;
-  vst1q_u8(d, *q10u8);
-  d += d_stride;
-  vst1q_u8(d, *q11u8);
-  d += d_stride;
-  vst1q_u8(d, *q12u8);
-  d += d_stride;
-  vst1q_u8(d, *q13u8);
-  d += d_stride;
-  vst1q_u8(d, *q14u8);
-  d += d_stride;
-  vst1q_u8(d, *q15u8);
-  return;
-}
-
-void aom_idct32x32_1_add_neon(int16_t *input, uint8_t *dest, int dest_stride) {
-  uint8x16_t q0u8, q8u8, q9u8, q10u8, q11u8, q12u8, q13u8, q14u8, q15u8;
-  int i, j, dest_stride8;
-  uint8_t *d;
-  int16_t a1;
-  int16_t out = dct_const_round_shift(input[0] * cospi_16_64);
-
-  out = dct_const_round_shift(out * cospi_16_64);
-  a1 = ROUND_POWER_OF_TWO(out, 6);
-
-  dest_stride8 = dest_stride * 8;
-  if (a1 >= 0) {  // diff_positive_32_32
-    a1 = a1 < 0 ? 0 : a1 > 255 ? 255 : a1;
-    q0u8 = vdupq_n_u8(a1);
-    for (i = 0; i < 2; i++, dest += 16) {  // diff_positive_32_32_loop
-      d = dest;
-      for (j = 0; j < 4; j++) {
-        LD_16x8(d, dest_stride, &q8u8, &q9u8, &q10u8, &q11u8, &q12u8, &q13u8,
-                &q14u8, &q15u8);
-        ADD_DIFF_16x8(q0u8, &q8u8, &q9u8, &q10u8, &q11u8, &q12u8, &q13u8,
-                      &q14u8, &q15u8);
-        ST_16x8(d, dest_stride, &q8u8, &q9u8, &q10u8, &q11u8, &q12u8, &q13u8,
-                &q14u8, &q15u8);
-        d += dest_stride8;
-      }
-    }
-  } else {  // diff_negative_32_32
-    a1 = -a1;
-    a1 = a1 < 0 ? 0 : a1 > 255 ? 255 : a1;
-    q0u8 = vdupq_n_u8(a1);
-    for (i = 0; i < 2; i++, dest += 16) {  // diff_negative_32_32_loop
-      d = dest;
-      for (j = 0; j < 4; j++) {
-        LD_16x8(d, dest_stride, &q8u8, &q9u8, &q10u8, &q11u8, &q12u8, &q13u8,
-                &q14u8, &q15u8);
-        SUB_DIFF_16x8(q0u8, &q8u8, &q9u8, &q10u8, &q11u8, &q12u8, &q13u8,
-                      &q14u8, &q15u8);
-        ST_16x8(d, dest_stride, &q8u8, &q9u8, &q10u8, &q11u8, &q12u8, &q13u8,
-                &q14u8, &q15u8);
-        d += dest_stride8;
-      }
-    }
-  }
-  return;
-}

aom_dsp/arm/idct32x32_add_neon.c

@@ -1,686 +0,0 @@
-/*
- * Copyright (c) 2016, Alliance for Open Media. All rights reserved
- *
- * This source code is subject to the terms of the BSD 2 Clause License and
- * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
- * was not distributed with this source code in the LICENSE file, you can
- * obtain it at www.aomedia.org/license/software. If the Alliance for Open
- * Media Patent License 1.0 was not distributed with this source code in the
- * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
- */
-
-#include <arm_neon.h>
-
-#include "./aom_config.h"
-#include "aom_dsp/txfm_common.h"
-
-#define LOAD_FROM_TRANSPOSED(prev, first, second) \
-  q14s16 = vld1q_s16(trans_buf + first * 8);      \
-  q13s16 = vld1q_s16(trans_buf + second * 8);
-
-#define LOAD_FROM_OUTPUT(prev, first, second, qA, qB) \
-  qA = vld1q_s16(out + first * 32);                   \
-  qB = vld1q_s16(out + second * 32);
-
-#define STORE_IN_OUTPUT(prev, first, second, qA, qB) \
-  vst1q_s16(out + first * 32, qA);                   \
-  vst1q_s16(out + second * 32, qB);
-
-#define STORE_COMBINE_CENTER_RESULTS(r10, r9) \
-  __STORE_COMBINE_CENTER_RESULTS(r10, r9, stride, q6s16, q7s16, q8s16, q9s16);
-static INLINE void __STORE_COMBINE_CENTER_RESULTS(uint8_t *p1, uint8_t *p2,
-                                                  int stride, int16x8_t q6s16,
-                                                  int16x8_t q7s16,
-                                                  int16x8_t q8s16,
-                                                  int16x8_t q9s16) {
-  int16x4_t d8s16, d9s16, d10s16, d11s16;
-
-  d8s16 = vld1_s16((int16_t *)p1);
-  p1 += stride;
-  d11s16 = vld1_s16((int16_t *)p2);
-  p2 -= stride;
-  d9s16 = vld1_s16((int16_t *)p1);
-  d10s16 = vld1_s16((int16_t *)p2);
-
-  q7s16 = vrshrq_n_s16(q7s16, 6);
-  q8s16 = vrshrq_n_s16(q8s16, 6);
-  q9s16 = vrshrq_n_s16(q9s16, 6);
-  q6s16 = vrshrq_n_s16(q6s16, 6);
-
-  q7s16 = vreinterpretq_s16_u16(
-      vaddw_u8(vreinterpretq_u16_s16(q7s16), vreinterpret_u8_s16(d9s16)));
-  q8s16 = vreinterpretq_s16_u16(
-      vaddw_u8(vreinterpretq_u16_s16(q8s16), vreinterpret_u8_s16(d10s16)));
-  q9s16 = vreinterpretq_s16_u16(
-      vaddw_u8(vreinterpretq_u16_s16(q9s16), vreinterpret_u8_s16(d11s16)));
-  q6s16 = vreinterpretq_s16_u16(
-      vaddw_u8(vreinterpretq_u16_s16(q6s16), vreinterpret_u8_s16(d8s16)));
-
-  d9s16 = vreinterpret_s16_u8(vqmovun_s16(q7s16));
-  d10s16 = vreinterpret_s16_u8(vqmovun_s16(q8s16));
-  d11s16 = vreinterpret_s16_u8(vqmovun_s16(q9s16));
-  d8s16 = vreinterpret_s16_u8(vqmovun_s16(q6s16));
-
-  vst1_s16((int16_t *)p1, d9s16);
-  p1 -= stride;
-  vst1_s16((int16_t *)p2, d10s16);
-  p2 += stride;
-  vst1_s16((int16_t *)p1, d8s16);
-  vst1_s16((int16_t *)p2, d11s16);
-  return;
-}
-
-#define STORE_COMBINE_EXTREME_RESULTS(r7, r6) \
-  ;                                           \
-  __STORE_COMBINE_EXTREME_RESULTS(r7, r6, stride, q4s16, q5s16, q6s16, q7s16);
-static INLINE void __STORE_COMBINE_EXTREME_RESULTS(uint8_t *p1, uint8_t *p2,
-                                                   int stride, int16x8_t q4s16,
-                                                   int16x8_t q5s16,
-                                                   int16x8_t q6s16,
-                                                   int16x8_t q7s16) {
-  int16x4_t d4s16, d5s16, d6s16, d7s16;
-
-  d4s16 = vld1_s16((int16_t *)p1);
-  p1 += stride;
-  d7s16 = vld1_s16((int16_t *)p2);
-  p2 -= stride;
-  d5s16 = vld1_s16((int16_t *)p1);
-  d6s16 = vld1_s16((int16_t *)p2);
-
-  q5s16 = vrshrq_n_s16(q5s16, 6);
-  q6s16 = vrshrq_n_s16(q6s16, 6);
-  q7s16 = vrshrq_n_s16(q7s16, 6);
-  q4s16 = vrshrq_n_s16(q4s16, 6);
-
-  q5s16 = vreinterpretq_s16_u16(
-      vaddw_u8(vreinterpretq_u16_s16(q5s16), vreinterpret_u8_s16(d5s16)));
-  q6s16 = vreinterpretq_s16_u16(
-      vaddw_u8(vreinterpretq_u16_s16(q6s16), vreinterpret_u8_s16(d6s16)));
-  q7s16 = vreinterpretq_s16_u16(
-      vaddw_u8(vreinterpretq_u16_s16(q7s16), vreinterpret_u8_s16(d7s16)));
-  q4s16 = vreinterpretq_s16_u16(
-      vaddw_u8(vreinterpretq_u16_s16(q4s16), vreinterpret_u8_s16(d4s16)));
-
-  d5s16 = vreinterpret_s16_u8(vqmovun_s16(q5s16));
-  d6s16 = vreinterpret_s16_u8(vqmovun_s16(q6s16));
-  d7s16 = vreinterpret_s16_u8(vqmovun_s16(q7s16));
-  d4s16 = vreinterpret_s16_u8(vqmovun_s16(q4s16));
-
-  vst1_s16((int16_t *)p1, d5s16);
-  p1 -= stride;
-  vst1_s16((int16_t *)p2, d6s16);
-  p2 += stride;
-  vst1_s16((int16_t *)p2, d7s16);
-  vst1_s16((int16_t *)p1, d4s16);
-  return;
-}
-
-#define DO_BUTTERFLY_STD(const_1, const_2, qA, qB) \
-  DO_BUTTERFLY(q14s16, q13s16, const_1, const_2, qA, qB);
-static INLINE void DO_BUTTERFLY(int16x8_t q14s16, int16x8_t q13s16,
-                                int16_t first_const, int16_t second_const,
-                                int16x8_t *qAs16, int16x8_t *qBs16) {
-  int16x4_t d30s16, d31s16;
-  int32x4_t q8s32, q9s32, q10s32, q11s32, q12s32, q15s32;
-  int16x4_t dCs16, dDs16, dAs16, dBs16;
-
-  dCs16 = vget_low_s16(q14s16);
-  dDs16 = vget_high_s16(q14s16);
-  dAs16 = vget_low_s16(q13s16);
-  dBs16 = vget_high_s16(q13s16);
-
-  d30s16 = vdup_n_s16(first_const);
-  d31s16 = vdup_n_s16(second_const);
-
-  q8s32 = vmull_s16(dCs16, d30s16);
-  q10s32 = vmull_s16(dAs16, d31s16);
-  q9s32 = vmull_s16(dDs16, d30s16);
-  q11s32 = vmull_s16(dBs16, d31s16);
-  q12s32 = vmull_s16(dCs16, d31s16);
-
-  q8s32 = vsubq_s32(q8s32, q10s32);
-  q9s32 = vsubq_s32(q9s32, q11s32);
-
-  q10s32 = vmull_s16(dDs16, d31s16);
-  q11s32 = vmull_s16(dAs16, d30s16);
-  q15s32 = vmull_s16(dBs16, d30s16);
-
-  q11s32 = vaddq_s32(q12s32, q11s32);
-  q10s32 = vaddq_s32(q10s32, q15s32);
-
-  *qAs16 = vcombine_s16(vqrshrn_n_s32(q8s32, 14), vqrshrn_n_s32(q9s32, 14));
-  *qBs16 = vcombine_s16(vqrshrn_n_s32(q11s32, 14), vqrshrn_n_s32(q10s32, 14));
-  return;
-}
-
-static INLINE void idct32_transpose_pair(int16_t *input, int16_t *t_buf) {
-  int16_t *in;
-  int i;
-  const int stride = 32;
-  int16x4_t d16s16, d17s16, d18s16, d19s16, d20s16, d21s16, d22s16, d23s16;
-  int16x4_t d24s16, d25s16, d26s16, d27s16, d28s16, d29s16, d30s16, d31s16;
-  int16x8_t q8s16, q9s16, q10s16, q11s16, q12s16, q13s16, q14s16, q15s16;
-  int32x4x2_t q0x2s32, q1x2s32, q2x2s32, q3x2s32;
-  int16x8x2_t q0x2s16, q1x2s16, q2x2s16, q3x2s16;
-
-  for (i = 0; i < 4; i++, input += 8) {
-    in = input;
-    q8s16 = vld1q_s16(in);
-    in += stride;
-    q9s16 = vld1q_s16(in);
-    in += stride;
-    q10s16 = vld1q_s16(in);
-    in += stride;
-    q11s16 = vld1q_s16(in);
-    in += stride;
-    q12s16 = vld1q_s16(in);
-    in += stride;
-    q13s16 = vld1q_s16(in);
-    in += stride;
-    q14s16 = vld1q_s16(in);
-    in += stride;
-    q15s16 = vld1q_s16(in);
-
-    d16s16 = vget_low_s16(q8s16);
-    d17s16 = vget_high_s16(q8s16);
-    d18s16 = vget_low_s16(q9s16);
-    d19s16 = vget_high_s16(q9s16);
-    d20s16 = vget_low_s16(q10s16);
-    d21s16 = vget_high_s16(q10s16);
-    d22s16 = vget_low_s16(q11s16);
-    d23s16 = vget_high_s16(q11s16);
-    d24s16 = vget_low_s16(q12s16);
-    d25s16 = vget_high_s16(q12s16);
-    d26s16 = vget_low_s16(q13s16);
-    d27s16 = vget_high_s16(q13s16);
-    d28s16 = vget_low_s16(q14s16);
-    d29s16 = vget_high_s16(q14s16);
-    d30s16 = vget_low_s16(q15s16);
-    d31s16 = vget_high_s16(q15s16);
-
-    q8s16 = vcombine_s16(d16s16, d24s16);   // vswp d17, d24
-    q9s16 = vcombine_s16(d18s16, d26s16);   // vswp d19, d26
-    q10s16 = vcombine_s16(d20s16, d28s16);  // vswp d21, d28
-    q11s16 = vcombine_s16(d22s16, d30s16);  // vswp d23, d30
-    q12s16 = vcombine_s16(d17s16, d25s16);
-    q13s16 = vcombine_s16(d19s16, d27s16);
-    q14s16 = vcombine_s16(d21s16, d29s16);
-    q15s16 = vcombine_s16(d23s16, d31s16);
-
-    q0x2s32 =
-        vtrnq_s32(vreinterpretq_s32_s16(q8s16), vreinterpretq_s32_s16(q10s16));
-    q1x2s32 =
-        vtrnq_s32(vreinterpretq_s32_s16(q9s16), vreinterpretq_s32_s16(q11s16));
-    q2x2s32 =
-        vtrnq_s32(vreinterpretq_s32_s16(q12s16), vreinterpretq_s32_s16(q14s16));
-    q3x2s32 =
-        vtrnq_s32(vreinterpretq_s32_s16(q13s16), vreinterpretq_s32_s16(q15s16));
-
-    q0x2s16 = vtrnq_s16(vreinterpretq_s16_s32(q0x2s32.val[0]),   // q8
-                        vreinterpretq_s16_s32(q1x2s32.val[0]));  // q9
-    q1x2s16 = vtrnq_s16(vreinterpretq_s16_s32(q0x2s32.val[1]),   // q10
-                        vreinterpretq_s16_s32(q1x2s32.val[1]));  // q11
-    q2x2s16 = vtrnq_s16(vreinterpretq_s16_s32(q2x2s32.val[0]),   // q12
-                        vreinterpretq_s16_s32(q3x2s32.val[0]));  // q13
-    q3x2s16 = vtrnq_s16(vreinterpretq_s16_s32(q2x2s32.val[1]),   // q14
-                        vreinterpretq_s16_s32(q3x2s32.val[1]));  // q15
-
-    vst1q_s16(t_buf, q0x2s16.val[0]);
-    t_buf += 8;
-    vst1q_s16(t_buf, q0x2s16.val[1]);
-    t_buf += 8;
-    vst1q_s16(t_buf, q1x2s16.val[0]);
-    t_buf += 8;
-    vst1q_s16(t_buf, q1x2s16.val[1]);
-    t_buf += 8;
-    vst1q_s16(t_buf, q2x2s16.val[0]);
-    t_buf += 8;
-    vst1q_s16(t_buf, q2x2s16.val[1]);
-    t_buf += 8;
-    vst1q_s16(t_buf, q3x2s16.val[0]);
-    t_buf += 8;
-    vst1q_s16(t_buf, q3x2s16.val[1]);
-    t_buf += 8;
-  }
-  return;
-}
-
-static INLINE void idct32_bands_end_1st_pass(int16_t *out, int16x8_t q2s16,
-                                             int16x8_t q3s16, int16x8_t q6s16,
-                                             int16x8_t q7s16, int16x8_t q8s16,
-                                             int16x8_t q9s16, int16x8_t q10s16,
-                                             int16x8_t q11s16, int16x8_t q12s16,
-                                             int16x8_t q13s16, int16x8_t q14s16,
-                                             int16x8_t q15s16) {
-  int16x8_t q0s16, q1s16, q4s16, q5s16;
-
-  STORE_IN_OUTPUT(17, 16, 17, q6s16, q7s16);
-  STORE_IN_OUTPUT(17, 14, 15, q8s16, q9s16);
-
-  LOAD_FROM_OUTPUT(15, 30, 31, q0s16, q1s16);
-  q4s16 = vaddq_s16(q2s16, q1s16);
-  q5s16 = vaddq_s16(q3s16, q0s16);
-  q6s16 = vsubq_s16(q3s16, q0s16);
-  q7s16 = vsubq_s16(q2s16, q1s16);
-  STORE_IN_OUTPUT(31, 30, 31, q6s16, q7s16);
-  STORE_IN_OUTPUT(31, 0, 1, q4s16, q5s16);
-
-  LOAD_FROM_OUTPUT(1, 12, 13, q0s16, q1s16);
-  q2s16 = vaddq_s16(q10s16, q1s16);
-  q3s16 = vaddq_s16(q11s16, q0s16);
-  q4s16 = vsubq_s16(q11s16, q0s16);
-  q5s16 = vsubq_s16(q10s16, q1s16);
-
-  LOAD_FROM_OUTPUT(13, 18, 19, q0s16, q1s16);
-  q8s16 = vaddq_s16(q4s16, q1s16);
-  q9s16 = vaddq_s16(q5s16, q0s16);
-  q6s16 = vsubq_s16(q5s16, q0s16);
-  q7s16 = vsubq_s16(q4s16, q1s16);
-  STORE_IN_OUTPUT(19, 18, 19, q6s16, q7s16);
-  STORE_IN_OUTPUT(19, 12, 13, q8s16, q9s16);
-
-  LOAD_FROM_OUTPUT(13, 28, 29, q0s16, q1s16);
-  q4s16 = vaddq_s16(q2s16, q1s16);
-  q5s16 = vaddq_s16(q3s16, q0s16);
-  q6s16 = vsubq_s16(q3s16, q0s16);
-  q7s16 = vsubq_s16(q2s16, q1s16);
-  STORE_IN_OUTPUT(29, 28, 29, q6s16, q7s16);
-  STORE_IN_OUTPUT(29, 2, 3, q4s16, q5s16);
-
-  LOAD_FROM_OUTPUT(3, 10, 11, q0s16, q1s16);
-  q2s16 = vaddq_s16(q12s16, q1s16);
-  q3s16 = vaddq_s16(q13s16, q0s16);
-  q4s16 = vsubq_s16(q13s16, q0s16);
-  q5s16 = vsubq_s16(q12s16, q1s16);
-
-  LOAD_FROM_OUTPUT(11, 20, 21, q0s16, q1s16);
-  q8s16 = vaddq_s16(q4s16, q1s16);
-  q9s16 = vaddq_s16(q5s16, q0s16);
-  q6s16 = vsubq_s16(q5s16, q0s16);
-  q7s16 = vsubq_s16(q4s16, q1s16);
-  STORE_IN_OUTPUT(21, 20, 21, q6s16, q7s16);
-  STORE_IN_OUTPUT(21, 10, 11, q8s16, q9s16);
-
-  LOAD_FROM_OUTPUT(11, 26, 27, q0s16, q1s16);
-  q4s16 = vaddq_s16(q2s16, q1s16);
-  q5s16 = vaddq_s16(q3s16, q0s16);
-  q6s16 = vsubq_s16(q3s16, q0s16);
-  q7s16 = vsubq_s16(q2s16, q1s16);
-  STORE_IN_OUTPUT(27, 26, 27, q6s16, q7s16);
-  STORE_IN_OUTPUT(27, 4, 5, q4s16, q5s16);
-
-  LOAD_FROM_OUTPUT(5, 8, 9, q0s16, q1s16);
-  q2s16 = vaddq_s16(q14s16, q1s16);
-  q3s16 = vaddq_s16(q15s16, q0s16);
-  q4s16 = vsubq_s16(q15s16, q0s16);
-  q5s16 = vsubq_s16(q14s16, q1s16);
-
-  LOAD_FROM_OUTPUT(9, 22, 23, q0s16, q1s16);
-  q8s16 = vaddq_s16(q4s16, q1s16);
-  q9s16 = vaddq_s16(q5s16, q0s16);
-  q6s16 = vsubq_s16(q5s16, q0s16);
-  q7s16 = vsubq_s16(q4s16, q1s16);
-  STORE_IN_OUTPUT(23, 22, 23, q6s16, q7s16);
-  STORE_IN_OUTPUT(23, 8, 9, q8s16, q9s16);
-
-  LOAD_FROM_OUTPUT(9, 24, 25, q0s16, q1s16);
-  q4s16 = vaddq_s16(q2s16, q1s16);
-  q5s16 = vaddq_s16(q3s16, q0s16);
-  q6s16 = vsubq_s16(q3s16, q0s16);
-  q7s16 = vsubq_s16(q2s16, q1s16);
-  STORE_IN_OUTPUT(25, 24, 25, q6s16, q7s16);
-  STORE_IN_OUTPUT(25, 6, 7, q4s16, q5s16);
-  return;
-}
-
-static INLINE void idct32_bands_end_2nd_pass(
-    int16_t *out, uint8_t *dest, int stride, int16x8_t q2s16, int16x8_t q3s16,
-    int16x8_t q6s16, int16x8_t q7s16, int16x8_t q8s16, int16x8_t q9s16,
-    int16x8_t q10s16, int16x8_t q11s16, int16x8_t q12s16, int16x8_t q13s16,
-    int16x8_t q14s16, int16x8_t q15s16) {
-  uint8_t *r6 = dest + 31 * stride;
-  uint8_t *r7 = dest /* +  0 * stride*/;
-  uint8_t *r9 = dest + 15 * stride;
-  uint8_t *r10 = dest + 16 * stride;
-  int str2 = stride << 1;
-  int16x8_t q0s16, q1s16, q4s16, q5s16;
-
-  STORE_COMBINE_CENTER_RESULTS(r10, r9);
-  r10 += str2;
-  r9 -= str2;
-
-  LOAD_FROM_OUTPUT(17, 30, 31, q0s16, q1s16)
-  q4s16 = vaddq_s16(q2s16, q1s16);
-  q5s16 = vaddq_s16(q3s16, q0s16);
-  q6s16 = vsubq_s16(q3s16, q0s16);
-  q7s16 = vsubq_s16(q2s16, q1s16);
-  STORE_COMBINE_EXTREME_RESULTS(r7, r6);
-  r7 += str2;
-  r6 -= str2;
-
-  LOAD_FROM_OUTPUT(31, 12, 13, q0s16, q1s16)
-  q2s16 = vaddq_s16(q10s16, q1s16);
-  q3s16 = vaddq_s16(q11s16, q0s16);
-  q4s16 = vsubq_s16(q11s16, q0s16);
-  q5s16 = vsubq_s16(q10s16, q1s16);
-
-  LOAD_FROM_OUTPUT(13, 18, 19, q0s16, q1s16)
-  q8s16 = vaddq_s16(q4s16, q1s16);
-  q9s16 = vaddq_s16(q5s16, q0s16);
-  q6s16 = vsubq_s16(q5s16, q0s16);
-  q7s16 = vsubq_s16(q4s16, q1s16);
-  STORE_COMBINE_CENTER_RESULTS(r10, r9);
-  r10 += str2;
-  r9 -= str2;
-
-  LOAD_FROM_OUTPUT(19, 28, 29, q0s16, q1s16)
-  q4s16 = vaddq_s16(q2s16, q1s16);
-  q5s16 = vaddq_s16(q3s16, q0s16);
-  q6s16 = vsubq_s16(q3s16, q0s16);
-  q7s16 = vsubq_s16(q2s16, q1s16);
-  STORE_COMBINE_EXTREME_RESULTS(r7, r6);
-  r7 += str2;
-  r6 -= str2;
-
-  LOAD_FROM_OUTPUT(29, 10, 11, q0s16, q1s16)
-  q2s16 = vaddq_s16(q12s16, q1s16);
-  q3s16 = vaddq_s16(q13s16, q0s16);
-  q4s16 = vsubq_s16(q13s16, q0s16);
-  q5s16 = vsubq_s16(q12s16, q1s16);
-
-  LOAD_FROM_OUTPUT(11, 20, 21, q0s16, q1s16)
-  q8s16 = vaddq_s16(q4s16, q1s16);
-  q9s16 = vaddq_s16(q5s16, q0s16);
-  q6s16 = vsubq_s16(q5s16, q0s16);
-  q7s16 = vsubq_s16(q4s16, q1s16);
-  STORE_COMBINE_CENTER_RESULTS(r10, r9);
-  r10 += str2;
-  r9 -= str2;
-
-  LOAD_FROM_OUTPUT(21, 26, 27, q0s16, q1s16)
-  q4s16 = vaddq_s16(q2s16, q1s16);
-  q5s16 = vaddq_s16(q3s16, q0s16);
-  q6s16 = vsubq_s16(q3s16, q0s16);
-  q7s16 = vsubq_s16(q2s16, q1s16);
-  STORE_COMBINE_EXTREME_RESULTS(r7, r6);
-  r7 += str2;
-  r6 -= str2;
-
-  LOAD_FROM_OUTPUT(27, 8, 9, q0s16, q1s16)
-  q2s16 = vaddq_s16(q14s16, q1s16);
-  q3s16 = vaddq_s16(q15s16, q0s16);
-  q4s16 = vsubq_s16(q15s16, q0s16);
-  q5s16 = vsubq_s16(q14s16, q1s16);
-
-  LOAD_FROM_OUTPUT(9, 22, 23, q0s16, q1s16)
-  q8s16 = vaddq_s16(q4s16, q1s16);
-  q9s16 = vaddq_s16(q5s16, q0s16);
-  q6s16 = vsubq_s16(q5s16, q0s16);
-  q7s16 = vsubq_s16(q4s16, q1s16);
-  STORE_COMBINE_CENTER_RESULTS(r10, r9);
-
-  LOAD_FROM_OUTPUT(23, 24, 25, q0s16, q1s16)
-  q4s16 = vaddq_s16(q2s16, q1s16);
-  q5s16 = vaddq_s16(q3s16, q0s16);
-  q6s16 = vsubq_s16(q3s16, q0s16);
-  q7s16 = vsubq_s16(q2s16, q1s16);
-  STORE_COMBINE_EXTREME_RESULTS(r7, r6);
-  return;
-}
-
-void aom_idct32x32_1024_add_neon(int16_t *input, uint8_t *dest, int stride) {
-  int i, idct32_pass_loop;
-  int16_t trans_buf[32 * 8];
-  int16_t pass1[32 * 32];
-  int16_t pass2[32 * 32];
-  int16_t *out;
-  int16x8_t q0s16, q1s16, q2s16, q3s16, q4s16, q5s16, q6s16, q7s16;
-  int16x8_t q8s16, q9s16, q10s16, q11s16, q12s16, q13s16, q14s16, q15s16;
-
-  for (idct32_pass_loop = 0, out = pass1; idct32_pass_loop < 2;
-       idct32_pass_loop++,
-      input = pass1,  // the input of pass2 is the result of pass1
-       out = pass2) {
-    for (i = 0; i < 4; i++, input += 32 * 8, out += 8) {  // idct32_bands_loop
-      idct32_transpose_pair(input, trans_buf);
-
-      // -----------------------------------------
-      // BLOCK A: 16-19,28-31
-      // -----------------------------------------
-      // generate 16,17,30,31
-      // part of stage 1
-      LOAD_FROM_TRANSPOSED(0, 1, 31)
-      DO_BUTTERFLY_STD(cospi_31_64, cospi_1_64, &q0s16, &q2s16)
-      LOAD_FROM_TRANSPOSED(31, 17, 15)
-      DO_BUTTERFLY_STD(cospi_15_64, cospi_17_64, &q1s16, &q3s16)
-      // part of stage 2
-      q4s16 = vaddq_s16(q0s16, q1s16);
-      q13s16 = vsubq_s16(q0s16, q1s16);
-      q6s16 = vaddq_s16(q2s16, q3s16);
-      q14s16 = vsubq_s16(q2s16, q3s16);
-      // part of stage 3
-      DO_BUTTERFLY_STD(cospi_28_64, cospi_4_64, &q5s16, &q7s16)
-
-      // generate 18,19,28,29
-      // part of stage 1
-      LOAD_FROM_TRANSPOSED(15, 9, 23)
-      DO_BUTTERFLY_STD(cospi_23_64, cospi_9_64, &q0s16, &q2s16)
-      LOAD_FROM_TRANSPOSED(23, 25, 7)
-      DO_BUTTERFLY_STD(cospi_7_64, cospi_25_64, &q1s16, &q3s16)
-      // part of stage 2
-      q13s16 = vsubq_s16(q3s16, q2s16);
-      q3s16 = vaddq_s16(q3s16, q2s16);
-      q14s16 = vsubq_s16(q1s16, q0s16);
-      q2s16 = vaddq_s16(q1s16, q0s16);
-      // part of stage 3
-      DO_BUTTERFLY_STD(-cospi_4_64, -cospi_28_64, &q1s16, &q0s16)
-      // part of stage 4
-      q8s16 = vaddq_s16(q4s16, q2s16);
-      q9s16 = vaddq_s16(q5s16, q0s16);
-      q10s16 = vaddq_s16(q7s16, q1s16);
-      q15s16 = vaddq_s16(q6s16, q3s16);
-      q13s16 = vsubq_s16(q5s16, q0s16);
-      q14s16 = vsubq_s16(q7s16, q1s16);
-      STORE_IN_OUTPUT(0, 16, 31, q8s16, q15s16)
-      STORE_IN_OUTPUT(31, 17, 30, q9s16, q10s16)
-      // part of stage 5
-      DO_BUTTERFLY_STD(cospi_24_64, cospi_8_64, &q0s16, &q1s16)
-      STORE_IN_OUTPUT(30, 29, 18, q1s16, q0s16)
-      // part of stage 4
-      q13s16 = vsubq_s16(q4s16, q2s16);
-      q14s16 = vsubq_s16(q6s16, q3s16);
-      // part of stage 5
-      DO_BUTTERFLY_STD(cospi_24_64, cospi_8_64, &q4s16, &q6s16)
-      STORE_IN_OUTPUT(18, 19, 28, q4s16, q6s16)
-
-      // -----------------------------------------
-      // BLOCK B: 20-23,24-27
-      // -----------------------------------------
-      // generate 20,21,26,27
-      // part of stage 1
-      LOAD_FROM_TRANSPOSED(7, 5, 27)
-      DO_BUTTERFLY_STD(cospi_27_64, cospi_5_64, &q0s16, &q2s16)
-      LOAD_FROM_TRANSPOSED(27, 21, 11)
-      DO_BUTTERFLY_STD(cospi_11_64, cospi_21_64, &q1s16, &q3s16)
-      // part of stage 2
-      q13s16 = vsubq_s16(q0s16, q1s16);
-      q0s16 = vaddq_s16(q0s16, q1s16);
-      q14s16 = vsubq_s16(q2s16, q3s16);
-      q2s16 = vaddq_s16(q2s16, q3s16);
-      // part of stage 3
-      DO_BUTTERFLY_STD(cospi_12_64, cospi_20_64, &q1s16, &q3s16)
-
-      // generate 22,23,24,25
-      // part of stage 1
-      LOAD_FROM_TRANSPOSED(11, 13, 19)
-      DO_BUTTERFLY_STD(cospi_19_64, cospi_13_64, &q5s16, &q7s16)
-      LOAD_FROM_TRANSPOSED(19, 29, 3)
-      DO_BUTTERFLY_STD(cospi_3_64, cospi_29_64, &q4s16, &q6s16)
-      // part of stage 2
-      q14s16 = vsubq_s16(q4s16, q5s16);
-      q5s16 = vaddq_s16(q4s16, q5s16);
-      q13s16 = vsubq_s16(q6s16, q7s16);
-      q6s16 = vaddq_s16(q6s16, q7s16);
-      // part of stage 3
-      DO_BUTTERFLY_STD(-cospi_20_64, -cospi_12_64, &q4s16, &q7s16)
-      // part of stage 4
-      q10s16 = vaddq_s16(q7s16, q1s16);
-      q11s16 = vaddq_s16(q5s16, q0s16);
-      q12s16 = vaddq_s16(q6s16, q2s16);
-      q15s16 = vaddq_s16(q4s16, q3s16);
-      // part of stage 6
-      LOAD_FROM_OUTPUT(28, 16, 17, q14s16, q13s16)
-      q8s16 = vaddq_s16(q14s16, q11s16);
-      q9s16 = vaddq_s16(q13s16, q10s16);
-      q13s16 = vsubq_s16(q13s16, q10s16);
-      q11s16 = vsubq_s16(q14s16, q11s16);
-      STORE_IN_OUTPUT(17, 17, 16, q9s16, q8s16)
-      LOAD_FROM_OUTPUT(16, 30, 31, q14s16, q9s16)
-      q8s16 = vsubq_s16(q9s16, q12s16);
-      q10s16 = vaddq_s16(q14s16, q15s16);
-      q14s16 = vsubq_s16(q14s16, q15s16);
-      q12s16 = vaddq_s16(q9s16, q12s16);
-      STORE_IN_OUTPUT(31, 30, 31, q10s16, q12s16)
-      // part of stage 7
-      DO_BUTTERFLY_STD(cospi_16_64, cospi_16_64, &q13s16, &q14s16)
-      STORE_IN_OUTPUT(31, 25, 22, q14s16, q13s16)
-      q13s16 = q11s16;
-      q14s16 = q8s16;
-      DO_BUTTERFLY_STD(cospi_16_64, cospi_16_64, &q13s16, &q14s16)
-      STORE_IN_OUTPUT(22, 24, 23, q14s16, q13s16)
-      // part of stage 4
-      q14s16 = vsubq_s16(q5s16, q0s16);
-      q13s16 = vsubq_s16(q6s16, q2s16);
-      DO_BUTTERFLY_STD(-cospi_8_64, -cospi_24_64, &q5s16, &q6s16);
-      q14s16 = vsubq_s16(q7s16, q1s16);
-      q13s16 = vsubq_s16(q4s16, q3s16);
-      DO_BUTTERFLY_STD(-cospi_8_64, -cospi_24_64, &q0s16, &q1s16);
-      // part of stage 6
-      LOAD_FROM_OUTPUT(23, 18, 19, q14s16, q13s16)
-      q8s16 = vaddq_s16(q14s16, q1s16);
-      q9s16 = vaddq_s16(q13s16, q6s16);
-      q13s16 = vsubq_s16(q13s16, q6s16);
-      q1s16 = vsubq_s16(q14s16, q1s16);
-      STORE_IN_OUTPUT(19, 18, 19, q8s16, q9s16)
-      LOAD_FROM_OUTPUT(19, 28, 29, q8s16, q9s16)
-      q14s16 = vsubq_s16(q8s16, q5s16);
-      q10s16 = vaddq_s16(q8s16, q5s16);
-      q11s16 = vaddq_s16(q9s16, q0s16);
-      q0s16 = vsubq_s16(q9s16, q0s16);
-      STORE_IN_OUTPUT(29, 28, 29, q10s16, q11s16)
-      // part of stage 7
-      DO_BUTTERFLY_STD(cospi_16_64, cospi_16_64, &q13s16, &q14s16)
-      STORE_IN_OUTPUT(29, 20, 27, q13s16, q14s16)
-      DO_BUTTERFLY(q0s16, q1s16, cospi_16_64, cospi_16_64, &q1s16, &q0s16);
-      STORE_IN_OUTPUT(27, 21, 26, q1s16, q0s16)
-
-      // -----------------------------------------
-      // BLOCK C: 8-10,11-15
-      // -----------------------------------------
-      // generate 8,9,14,15
-      // part of stage 2
-      LOAD_FROM_TRANSPOSED(3, 2, 30)
-      DO_BUTTERFLY_STD(cospi_30_64, cospi_2_64, &q0s16, &q2s16)
-      LOAD_FROM_TRANSPOSED(30, 18, 14)
-      DO_BUTTERFLY_STD(cospi_14_64, cospi_18_64, &q1s16, &q3s16)
-      // part of stage 3
-      q13s16 = vsubq_s16(q0s16, q1s16);
-      q0s16 = vaddq_s16(q0s16, q1s16);
-      q14s16 = vsubq_s16(q2s16, q3s16);
-      q2s16 = vaddq_s16(q2s16, q3s16);
-      // part of stage 4
-      DO_BUTTERFLY_STD(cospi_24_64, cospi_8_64, &q1s16, &q3s16)
-
-      // generate 10,11,12,13
-      // part of stage 2
-      LOAD_FROM_TRANSPOSED(14, 10, 22)
-      DO_BUTTERFLY_STD(cospi_22_64, cospi_10_64, &q5s16, &q7s16)
-      LOAD_FROM_TRANSPOSED(22, 26, 6)
-      DO_BUTTERFLY_STD(cospi_6_64, cospi_26_64, &q4s16, &q6s16)
-      // part of stage 3
-      q14s16 = vsubq_s16(q4s16, q5s16);
-      q5s16 = vaddq_s16(q4s16, q5s16);
-      q13s16 = vsubq_s16(q6s16, q7s16);
-      q6s16 = vaddq_s16(q6s16, q7s16);
-      // part of stage 4
-      DO_BUTTERFLY_STD(-cospi_8_64, -cospi_24_64, &q4s16, &q7s16)
-      // part of stage 5
-      q8s16 = vaddq_s16(q0s16, q5s16);
-      q9s16 = vaddq_s16(q1s16, q7s16);
-      q13s16 = vsubq_s16(q1s16, q7s16);
-      q14s16 = vsubq_s16(q3s16, q4s16);
-      q10s16 = vaddq_s16(q3s16, q4s16);
-      q15s16 = vaddq_s16(q2s16, q6s16);
-      STORE_IN_OUTPUT(26, 8, 15, q8s16, q15s16)
-      STORE_IN_OUTPUT(15, 9, 14, q9s16, q10s16)
-      // part of stage 6
-      DO_BUTTERFLY_STD(cospi_16_64, cospi_16_64, &q1s16, &q3s16)
-      STORE_IN_OUTPUT(14, 13, 10, q3s16, q1s16)
-      q13s16 = vsubq_s16(q0s16, q5s16);
-      q14s16 = vsubq_s16(q2s16, q6s16);
-      DO_BUTTERFLY_STD(cospi_16_64, cospi_16_64, &q1s16, &q3s16)
-      STORE_IN_OUTPUT(10, 11, 12, q1s16, q3s16)
-
-      // -----------------------------------------
-      // BLOCK D: 0-3,4-7
-      // -----------------------------------------
-      // generate 4,5,6,7
-      // part of stage 3
-      LOAD_FROM_TRANSPOSED(6, 4, 28)
-      DO_BUTTERFLY_STD(cospi_28_64, cospi_4_64, &q0s16, &q2s16)
-      LOAD_FROM_TRANSPOSED(28, 20, 12)
-      DO_BUTTERFLY_STD(cospi_12_64, cospi_20_64, &q1s16, &q3s16)
-      // part of stage 4
-      q13s16 = vsubq_s16(q0s16, q1s16);
-      q0s16 = vaddq_s16(q0s16, q1s16);
-      q14s16 = vsubq_s16(q2s16, q3s16);
-      q2s16 = vaddq_s16(q2s16, q3s16);
-      // part of stage 5
-      DO_BUTTERFLY_STD(cospi_16_64, cospi_16_64, &q1s16, &q3s16)
-
-      // generate 0,1,2,3
-      // part of stage 4
-      LOAD_FROM_TRANSPOSED(12, 0, 16)
-      DO_BUTTERFLY_STD(cospi_16_64, cospi_16_64, &q5s16, &q7s16)
-      LOAD_FROM_TRANSPOSED(16, 8, 24)
-      DO_BUTTERFLY_STD(cospi_24_64, cospi_8_64, &q14s16, &q6s16)
-      // part of stage 5
-      q4s16 = vaddq_s16(q7s16, q6s16);
-      q7s16 = vsubq_s16(q7s16, q6s16);
-      q6s16 = vsubq_s16(q5s16, q14s16);
-      q5s16 = vaddq_s16(q5s16, q14s16);
-      // part of stage 6
-      q8s16 = vaddq_s16(q4s16, q2s16);
-      q9s16 = vaddq_s16(q5s16, q3s16);
-      q10s16 = vaddq_s16(q6s16, q1s16);
-      q11s16 = vaddq_s16(q7s16, q0s16);
-      q12s16 = vsubq_s16(q7s16, q0s16);
-      q13s16 = vsubq_s16(q6s16, q1s16);
-      q14s16 = vsubq_s16(q5s16, q3s16);
-      q15s16 = vsubq_s16(q4s16, q2s16);
-      // part of stage 7
-      LOAD_FROM_OUTPUT(12, 14, 15, q0s16, q1s16)
-      q2s16 = vaddq_s16(q8s16, q1s16);
-      q3s16 = vaddq_s16(q9s16, q0s16);
-      q4s16 = vsubq_s16(q9s16, q0s16);
-      q5s16 = vsubq_s16(q8s16, q1s16);
-      LOAD_FROM_OUTPUT(15, 16, 17, q0s16, q1s16)
-      q8s16 = vaddq_s16(q4s16, q1s16);
-      q9s16 = vaddq_s16(q5s16, q0s16);
-      q6s16 = vsubq_s16(q5s16, q0s16);
-      q7s16 = vsubq_s16(q4s16, q1s16);
-
-      if (idct32_pass_loop == 0) {
-        idct32_bands_end_1st_pass(out, q2s16, q3s16, q6s16, q7s16, q8s16, q9s16,
-                                  q10s16, q11s16, q12s16, q13s16, q14s16,
-                                  q15s16);
-      } else {
-        idct32_bands_end_2nd_pass(out, dest, stride, q2s16, q3s16, q6s16, q7s16,
-                                  q8s16, q9s16, q10s16, q11s16, q12s16, q13s16,
-                                  q14s16, q15s16);
-        dest += 8;
-      }
-    }
-  }
-  return;
-}

aom_dsp/arm/idct4x4_1_add_neon.c

@@ -1,47 +0,0 @@
-/*
- * Copyright (c) 2016, Alliance for Open Media. All rights reserved
- *
- * This source code is subject to the terms of the BSD 2 Clause License and
- * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
- * was not distributed with this source code in the LICENSE file, you can
- * obtain it at www.aomedia.org/license/software. If the Alliance for Open
- * Media Patent License 1.0 was not distributed with this source code in the
- * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
- */
-
-#include <arm_neon.h>
-
-#include "aom_dsp/inv_txfm.h"
-#include "aom_ports/mem.h"
-
-void aom_idct4x4_1_add_neon(int16_t *input, uint8_t *dest, int dest_stride) {
-  uint8x8_t d6u8;
-  uint32x2_t d2u32 = vdup_n_u32(0);
-  uint16x8_t q8u16;
-  int16x8_t q0s16;
-  uint8_t *d1, *d2;
-  int16_t i, a1;
-  int16_t out = dct_const_round_shift(input[0] * cospi_16_64);
-  out = dct_const_round_shift(out * cospi_16_64);
-  a1 = ROUND_POWER_OF_TWO(out, 4);
-
-  q0s16 = vdupq_n_s16(a1);
-
-  // dc_only_idct_add
-  d1 = d2 = dest;
-  for (i = 0; i < 2; i++) {
-    d2u32 = vld1_lane_u32((const uint32_t *)d1, d2u32, 0);
-    d1 += dest_stride;
-    d2u32 = vld1_lane_u32((const uint32_t *)d1, d2u32, 1);
-    d1 += dest_stride;
-
-    q8u16 = vaddw_u8(vreinterpretq_u16_s16(q0s16), vreinterpret_u8_u32(d2u32));
-    d6u8 = vqmovun_s16(vreinterpretq_s16_u16(q8u16));
-
-    vst1_lane_u32((uint32_t *)d2, vreinterpret_u32_u8(d6u8), 0);
-    d2 += dest_stride;
-    vst1_lane_u32((uint32_t *)d2, vreinterpret_u32_u8(d6u8), 1);
-    d2 += dest_stride;
-  }
-  return;
-}

aom_dsp/arm/idct4x4_add_neon.c

@@ -1,146 +0,0 @@
-/*
- * Copyright (c) 2016, Alliance for Open Media. All rights reserved
- *
- * This source code is subject to the terms of the BSD 2 Clause License and
- * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
- * was not distributed with this source code in the LICENSE file, you can
- * obtain it at www.aomedia.org/license/software. If the Alliance for Open
- * Media Patent License 1.0 was not distributed with this source code in the
- * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
- */
-
-#include <arm_neon.h>
-
-#include "aom_dsp/txfm_common.h"
-
-void aom_idct4x4_16_add_neon(int16_t *input, uint8_t *dest, int dest_stride) {
-  uint8x8_t d26u8, d27u8;
-  uint32x2_t d26u32, d27u32;
-  uint16x8_t q8u16, q9u16;
-  int16x4_t d16s16, d17s16, d18s16, d19s16, d20s16, d21s16;
-  int16x4_t d22s16, d23s16, d24s16, d26s16, d27s16, d28s16, d29s16;
-  int16x8_t q8s16, q9s16, q13s16, q14s16;
-  int32x4_t q1s32, q13s32, q14s32, q15s32;
-  int16x4x2_t d0x2s16, d1x2s16;
-  int32x4x2_t q0x2s32;
-  uint8_t *d;
-
-  d26u32 = d27u32 = vdup_n_u32(0);
-
-  q8s16 = vld1q_s16(input);
-  q9s16 = vld1q_s16(input + 8);
-
-  d16s16 = vget_low_s16(q8s16);
-  d17s16 = vget_high_s16(q8s16);
-  d18s16 = vget_low_s16(q9s16);
-  d19s16 = vget_high_s16(q9s16);
-
-  d0x2s16 = vtrn_s16(d16s16, d17s16);
-  d1x2s16 = vtrn_s16(d18s16, d19s16);
-  q8s16 = vcombine_s16(d0x2s16.val[0], d0x2s16.val[1]);
-  q9s16 = vcombine_s16(d1x2s16.val[0], d1x2s16.val[1]);
-
-  d20s16 = vdup_n_s16((int16_t)cospi_8_64);
-  d21s16 = vdup_n_s16((int16_t)cospi_16_64);
-
-  q0x2s32 =
-      vtrnq_s32(vreinterpretq_s32_s16(q8s16), vreinterpretq_s32_s16(q9s16));
-  d16s16 = vget_low_s16(vreinterpretq_s16_s32(q0x2s32.val[0]));
-  d17s16 = vget_high_s16(vreinterpretq_s16_s32(q0x2s32.val[0]));
-  d18s16 = vget_low_s16(vreinterpretq_s16_s32(q0x2s32.val[1]));
-  d19s16 = vget_high_s16(vreinterpretq_s16_s32(q0x2s32.val[1]));
-
-  d22s16 = vdup_n_s16((int16_t)cospi_24_64);
-
-  // stage 1
-  d23s16 = vadd_s16(d16s16, d18s16);
-  d24s16 = vsub_s16(d16s16, d18s16);
-
-  q15s32 = vmull_s16(d17s16, d22s16);
-  q1s32 = vmull_s16(d17s16, d20s16);
-  q13s32 = vmull_s16(d23s16, d21s16);
-  q14s32 = vmull_s16(d24s16, d21s16);
-
-  q15s32 = vmlsl_s16(q15s32, d19s16, d20s16);
-  q1s32 = vmlal_s16(q1s32, d19s16, d22s16);
-
-  d26s16 = vqrshrn_n_s32(q13s32, 14);
-  d27s16 = vqrshrn_n_s32(q14s32, 14);
-  d29s16 = vqrshrn_n_s32(q15s32, 14);
-  d28s16 = vqrshrn_n_s32(q1s32, 14);
-  q13s16 = vcombine_s16(d26s16, d27s16);
-  q14s16 = vcombine_s16(d28s16, d29s16);
-
-  // stage 2
-  q8s16 = vaddq_s16(q13s16, q14s16);
-  q9s16 = vsubq_s16(q13s16, q14s16);
-
-  d16s16 = vget_low_s16(q8s16);
-  d17s16 = vget_high_s16(q8s16);
-  d18s16 = vget_high_s16(q9s16);  // vswp d18 d19
-  d19s16 = vget_low_s16(q9s16);
-
-  d0x2s16 = vtrn_s16(d16s16, d17s16);
-  d1x2s16 = vtrn_s16(d18s16, d19s16);
-  q8s16 = vcombine_s16(d0x2s16.val[0], d0x2s16.val[1]);
-  q9s16 = vcombine_s16(d1x2s16.val[0], d1x2s16.val[1]);
-
-  q0x2s32 =
-      vtrnq_s32(vreinterpretq_s32_s16(q8s16), vreinterpretq_s32_s16(q9s16));
-  d16s16 = vget_low_s16(vreinterpretq_s16_s32(q0x2s32.val[0]));
-  d17s16 = vget_high_s16(vreinterpretq_s16_s32(q0x2s32.val[0]));
-  d18s16 = vget_low_s16(vreinterpretq_s16_s32(q0x2s32.val[1]));
-  d19s16 = vget_high_s16(vreinterpretq_s16_s32(q0x2s32.val[1]));
-
-  // do the transform on columns
-  // stage 1
-  d23s16 = vadd_s16(d16s16, d18s16);
-  d24s16 = vsub_s16(d16s16, d18s16);
-
-  q15s32 = vmull_s16(d17s16, d22s16);
-  q1s32 = vmull_s16(d17s16, d20s16);
-  q13s32 = vmull_s16(d23s16, d21s16);
-  q14s32 = vmull_s16(d24s16, d21s16);
-
-  q15s32 = vmlsl_s16(q15s32, d19s16, d20s16);
-  q1s32 = vmlal_s16(q1s32, d19s16, d22s16);
-
-  d26s16 = vqrshrn_n_s32(q13s32, 14);
-  d27s16 = vqrshrn_n_s32(q14s32, 14);
-  d29s16 = vqrshrn_n_s32(q15s32, 14);
-  d28s16 = vqrshrn_n_s32(q1s32, 14);
-  q13s16 = vcombine_s16(d26s16, d27s16);
-  q14s16 = vcombine_s16(d28s16, d29s16);
-
-  // stage 2
-  q8s16 = vaddq_s16(q13s16, q14s16);
-  q9s16 = vsubq_s16(q13s16, q14s16);
-
-  q8s16 = vrshrq_n_s16(q8s16, 4);
-  q9s16 = vrshrq_n_s16(q9s16, 4);
-
-  d = dest;
-  d26u32 = vld1_lane_u32((const uint32_t *)d, d26u32, 0);
-  d += dest_stride;
-  d26u32 = vld1_lane_u32((const uint32_t *)d, d26u32, 1);
-  d += dest_stride;
-  d27u32 = vld1_lane_u32((const uint32_t *)d, d27u32, 1);
-  d += dest_stride;
-  d27u32 = vld1_lane_u32((const uint32_t *)d, d27u32, 0);
-
-  q8u16 = vaddw_u8(vreinterpretq_u16_s16(q8s16), vreinterpret_u8_u32(d26u32));
-  q9u16 = vaddw_u8(vreinterpretq_u16_s16(q9s16), vreinterpret_u8_u32(d27u32));
-
-  d26u8 = vqmovun_s16(vreinterpretq_s16_u16(q8u16));
-  d27u8 = vqmovun_s16(vreinterpretq_s16_u16(q9u16));
-
-  d = dest;
-  vst1_lane_u32((uint32_t *)d, vreinterpret_u32_u8(d26u8), 0);
-  d += dest_stride;
-  vst1_lane_u32((uint32_t *)d, vreinterpret_u32_u8(d26u8), 1);
-  d += dest_stride;
-  vst1_lane_u32((uint32_t *)d, vreinterpret_u32_u8(d27u8), 1);
-  d += dest_stride;
-  vst1_lane_u32((uint32_t *)d, vreinterpret_u32_u8(d27u8), 0);
-  return;
-}

aom_dsp/arm/idct8x8_1_add_neon.c

@@ -1,62 +0,0 @@
-/*
- * Copyright (c) 2016, Alliance for Open Media. All rights reserved
- *
- * This source code is subject to the terms of the BSD 2 Clause License and
- * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
- * was not distributed with this source code in the LICENSE file, you can
- * obtain it at www.aomedia.org/license/software. If the Alliance for Open
- * Media Patent License 1.0 was not distributed with this source code in the
- * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
- */
-
-#include <arm_neon.h>
-
-#include "aom_dsp/inv_txfm.h"
-#include "aom_ports/mem.h"
-
-void aom_idct8x8_1_add_neon(int16_t *input, uint8_t *dest, int dest_stride) {
-  uint8x8_t d2u8, d3u8, d30u8, d31u8;
-  uint64x1_t d2u64, d3u64, d4u64, d5u64;
-  uint16x8_t q0u16, q9u16, q10u16, q11u16, q12u16;
-  int16x8_t q0s16;
-  uint8_t *d1, *d2;
-  int16_t i, a1;
-  int16_t out = dct_const_round_shift(input[0] * cospi_16_64);
-  out = dct_const_round_shift(out * cospi_16_64);
-  a1 = ROUND_POWER_OF_TWO(out, 5);
-
-  q0s16 = vdupq_n_s16(a1);
-  q0u16 = vreinterpretq_u16_s16(q0s16);
-
-  d1 = d2 = dest;
-  for (i = 0; i < 2; i++) {
-    d2u64 = vld1_u64((const uint64_t *)d1);
-    d1 += dest_stride;
-    d3u64 = vld1_u64((const uint64_t *)d1);
-    d1 += dest_stride;
-    d4u64 = vld1_u64((const uint64_t *)d1);
-    d1 += dest_stride;
-    d5u64 = vld1_u64((const uint64_t *)d1);
-    d1 += dest_stride;
-
-    q9u16 = vaddw_u8(q0u16, vreinterpret_u8_u64(d2u64));
-    q10u16 = vaddw_u8(q0u16, vreinterpret_u8_u64(d3u64));
-    q11u16 = vaddw_u8(q0u16, vreinterpret_u8_u64(d4u64));
-    q12u16 = vaddw_u8(q0u16, vreinterpret_u8_u64(d5u64));
-
-    d2u8 = vqmovun_s16(vreinterpretq_s16_u16(q9u16));
-    d3u8 = vqmovun_s16(vreinterpretq_s16_u16(q10u16));
-    d30u8 = vqmovun_s16(vreinterpretq_s16_u16(q11u16));
-    d31u8 = vqmovun_s16(vreinterpretq_s16_u16(q12u16));
-
-    vst1_u64((uint64_t *)d2, vreinterpret_u64_u8(d2u8));
-    d2 += dest_stride;
-    vst1_u64((uint64_t *)d2, vreinterpret_u64_u8(d3u8));
-    d2 += dest_stride;
-    vst1_u64((uint64_t *)d2, vreinterpret_u64_u8(d30u8));
-    d2 += dest_stride;
-    vst1_u64((uint64_t *)d2, vreinterpret_u64_u8(d31u8));
-    d2 += dest_stride;
-  }
-  return;
-}

aom_dsp/arm/idct8x8_add_neon.c

@@ -1,509 +0,0 @@
-/*
- * Copyright (c) 2016, Alliance for Open Media. All rights reserved
- *
- * This source code is subject to the terms of the BSD 2 Clause License and
- * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
- * was not distributed with this source code in the LICENSE file, you can
- * obtain it at www.aomedia.org/license/software. If the Alliance for Open
- * Media Patent License 1.0 was not distributed with this source code in the
- * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
- */
-
-#include <arm_neon.h>
-
-#include "./aom_config.h"
-#include "aom_dsp/txfm_common.h"
-
-static INLINE void TRANSPOSE8X8(int16x8_t *q8s16, int16x8_t *q9s16,
-                                int16x8_t *q10s16, int16x8_t *q11s16,
-                                int16x8_t *q12s16, int16x8_t *q13s16,
-                                int16x8_t *q14s16, int16x8_t *q15s16) {
-  int16x4_t d16s16, d17s16, d18s16, d19s16, d20s16, d21s16, d22s16, d23s16;
-  int16x4_t d24s16, d25s16, d26s16, d27s16, d28s16, d29s16, d30s16, d31s16;
-  int32x4x2_t q0x2s32, q1x2s32, q2x2s32, q3x2s32;
-  int16x8x2_t q0x2s16, q1x2s16, q2x2s16, q3x2s16;
-
-  d16s16 = vget_low_s16(*q8s16);
-  d17s16 = vget_high_s16(*q8s16);
-  d18s16 = vget_low_s16(*q9s16);
-  d19s16 = vget_high_s16(*q9s16);
-  d20s16 = vget_low_s16(*q10s16);
-  d21s16 = vget_high_s16(*q10s16);
-  d22s16 = vget_low_s16(*q11s16);
-  d23s16 = vget_high_s16(*q11s16);
-  d24s16 = vget_low_s16(*q12s16);
-  d25s16 = vget_high_s16(*q12s16);
-  d26s16 = vget_low_s16(*q13s16);
-  d27s16 = vget_high_s16(*q13s16);
-  d28s16 = vget_low_s16(*q14s16);
-  d29s16 = vget_high_s16(*q14s16);
-  d30s16 = vget_low_s16(*q15s16);
-  d31s16 = vget_high_s16(*q15s16);
-
-  *q8s16 = vcombine_s16(d16s16, d24s16);   // vswp d17, d24
-  *q9s16 = vcombine_s16(d18s16, d26s16);   // vswp d19, d26
-  *q10s16 = vcombine_s16(d20s16, d28s16);  // vswp d21, d28
-  *q11s16 = vcombine_s16(d22s16, d30s16);  // vswp d23, d30
-  *q12s16 = vcombine_s16(d17s16, d25s16);
-  *q13s16 = vcombine_s16(d19s16, d27s16);
-  *q14s16 = vcombine_s16(d21s16, d29s16);
-  *q15s16 = vcombine_s16(d23s16, d31s16);
-
-  q0x2s32 =
-      vtrnq_s32(vreinterpretq_s32_s16(*q8s16), vreinterpretq_s32_s16(*q10s16));
-  q1x2s32 =
-      vtrnq_s32(vreinterpretq_s32_s16(*q9s16), vreinterpretq_s32_s16(*q11s16));
-  q2x2s32 =
-      vtrnq_s32(vreinterpretq_s32_s16(*q12s16), vreinterpretq_s32_s16(*q14s16));
-  q3x2s32 =
-      vtrnq_s32(vreinterpretq_s32_s16(*q13s16), vreinterpretq_s32_s16(*q15s16));
-
-  q0x2s16 = vtrnq_s16(vreinterpretq_s16_s32(q0x2s32.val[0]),   // q8
-                      vreinterpretq_s16_s32(q1x2s32.val[0]));  // q9
-  q1x2s16 = vtrnq_s16(vreinterpretq_s16_s32(q0x2s32.val[1]),   // q10
-                      vreinterpretq_s16_s32(q1x2s32.val[1]));  // q11
-  q2x2s16 = vtrnq_s16(vreinterpretq_s16_s32(q2x2s32.val[0]),   // q12
-                      vreinterpretq_s16_s32(q3x2s32.val[0]));  // q13
-  q3x2s16 = vtrnq_s16(vreinterpretq_s16_s32(q2x2s32.val[1]),   // q14
-                      vreinterpretq_s16_s32(q3x2s32.val[1]));  // q15
-
-  *q8s16 = q0x2s16.val[0];
-  *q9s16 = q0x2s16.val[1];
-  *q10s16 = q1x2s16.val[0];
-  *q11s16 = q1x2s16.val[1];
-  *q12s16 = q2x2s16.val[0];
-  *q13s16 = q2x2s16.val[1];
-  *q14s16 = q3x2s16.val[0];
-  *q15s16 = q3x2s16.val[1];
-  return;
-}
-
-static INLINE void IDCT8x8_1D(int16x8_t *q8s16, int16x8_t *q9s16,
-                              int16x8_t *q10s16, int16x8_t *q11s16,
-                              int16x8_t *q12s16, int16x8_t *q13s16,
-                              int16x8_t *q14s16, int16x8_t *q15s16) {
-  int16x4_t d0s16, d1s16, d2s16, d3s16;
-  int16x4_t d8s16, d9s16, d10s16, d11s16, d12s16, d13s16, d14s16, d15s16;
-  int16x4_t d16s16, d17s16, d18s16, d19s16, d20s16, d21s16, d22s16, d23s16;
-  int16x4_t d24s16, d25s16, d26s16, d27s16, d28s16, d29s16, d30s16, d31s16;
-  int16x8_t q0s16, q1s16, q2s16, q3s16, q4s16, q5s16, q6s16, q7s16;
-  int32x4_t q2s32, q3s32, q5s32, q6s32, q8s32, q9s32;
-  int32x4_t q10s32, q11s32, q12s32, q13s32, q15s32;
-
-  d0s16 = vdup_n_s16((int16_t)cospi_28_64);
-  d1s16 = vdup_n_s16((int16_t)cospi_4_64);
-  d2s16 = vdup_n_s16((int16_t)cospi_12_64);
-  d3s16 = vdup_n_s16((int16_t)cospi_20_64);
-
-  d16s16 = vget_low_s16(*q8s16);
-  d17s16 = vget_high_s16(*q8s16);
-  d18s16 = vget_low_s16(*q9s16);
-  d19s16 = vget_high_s16(*q9s16);
-  d20s16 = vget_low_s16(*q10s16);
-  d21s16 = vget_high_s16(*q10s16);
-  d22s16 = vget_low_s16(*q11s16);
-  d23s16 = vget_high_s16(*q11s16);
-  d24s16 = vget_low_s16(*q12s16);
-  d25s16 = vget_high_s16(*q12s16);
-  d26s16 = vget_low_s16(*q13s16);
-  d27s16 = vget_high_s16(*q13s16);
-  d28s16 = vget_low_s16(*q14s16);
-  d29s16 = vget_high_s16(*q14s16);
-  d30s16 = vget_low_s16(*q15s16);
-  d31s16 = vget_high_s16(*q15s16);
-
-  q2s32 = vmull_s16(d18s16, d0s16);
-  q3s32 = vmull_s16(d19s16, d0s16);
-  q5s32 = vmull_s16(d26s16, d2s16);
-  q6s32 = vmull_s16(d27s16, d2s16);
-
-  q2s32 = vmlsl_s16(q2s32, d30s16, d1s16);
-  q3s32 = vmlsl_s16(q3s32, d31s16, d1s16);
-  q5s32 = vmlsl_s16(q5s32, d22s16, d3s16);
-  q6s32 = vmlsl_s16(q6s32, d23s16, d3s16);
-
-  d8s16 = vqrshrn_n_s32(q2s32, 14);
-  d9s16 = vqrshrn_n_s32(q3s32, 14);
-  d10s16 = vqrshrn_n_s32(q5s32, 14);
-  d11s16 = vqrshrn_n_s32(q6s32, 14);
-  q4s16 = vcombine_s16(d8s16, d9s16);
-  q5s16 = vcombine_s16(d10s16, d11s16);
-
-  q2s32 = vmull_s16(d18s16, d1s16);
-  q3s32 = vmull_s16(d19s16, d1s16);
-  q9s32 = vmull_s16(d26s16, d3s16);
-  q13s32 = vmull_s16(d27s16, d3s16);
-
-  q2s32 = vmlal_s16(q2s32, d30s16, d0s16);
-  q3s32 = vmlal_s16(q3s32, d31s16, d0s16);
-  q9s32 = vmlal_s16(q9s32, d22s16, d2s16);
-  q13s32 = vmlal_s16(q13s32, d23s16, d2s16);
-
-  d14s16 = vqrshrn_n_s32(q2s32, 14);
-  d15s16 = vqrshrn_n_s32(q3s32, 14);
-  d12s16 = vqrshrn_n_s32(q9s32, 14);
-  d13s16 = vqrshrn_n_s32(q13s32, 14);
-  q6s16 = vcombine_s16(d12s16, d13s16);
-  q7s16 = vcombine_s16(d14s16, d15s16);
-
-  d0s16 = vdup_n_s16((int16_t)cospi_16_64);
-
-  q2s32 = vmull_s16(d16s16, d0s16);
-  q3s32 = vmull_s16(d17s16, d0s16);
-  q13s32 = vmull_s16(d16s16, d0s16);
-  q15s32 = vmull_s16(d17s16, d0s16);
-
-  q2s32 = vmlal_s16(q2s32, d24s16, d0s16);
-  q3s32 = vmlal_s16(q3s32, d25s16, d0s16);
-  q13s32 = vmlsl_s16(q13s32, d24s16, d0s16);
-  q15s32 = vmlsl_s16(q15s32, d25s16, d0s16);
-
-  d0s16 = vdup_n_s16((int16_t)cospi_24_64);
-  d1s16 = vdup_n_s16((int16_t)cospi_8_64);
-
-  d18s16 = vqrshrn_n_s32(q2s32, 14);
-  d19s16 = vqrshrn_n_s32(q3s32, 14);
-  d22s16 = vqrshrn_n_s32(q13s32, 14);
-  d23s16 = vqrshrn_n_s32(q15s32, 14);
-  *q9s16 = vcombine_s16(d18s16, d19s16);
-  *q11s16 = vcombine_s16(d22s16, d23s16);
-
-  q2s32 = vmull_s16(d20s16, d0s16);
-  q3s32 = vmull_s16(d21s16, d0s16);
-  q8s32 = vmull_s16(d20s16, d1s16);
-  q12s32 = vmull_s16(d21s16, d1s16);
-
-  q2s32 = vmlsl_s16(q2s32, d28s16, d1s16);
-  q3s32 = vmlsl_s16(q3s32, d29s16, d1s16);
-  q8s32 = vmlal_s16(q8s32, d28s16, d0s16);
-  q12s32 = vmlal_s16(q12s32, d29s16, d0s16);
-
-  d26s16 = vqrshrn_n_s32(q2s32, 14);
-  d27s16 = vqrshrn_n_s32(q3s32, 14);
-  d30s16 = vqrshrn_n_s32(q8s32, 14);
-  d31s16 = vqrshrn_n_s32(q12s32, 14);
-  *q13s16 = vcombine_s16(d26s16, d27s16);
-  *q15s16 = vcombine_s16(d30s16, d31s16);
-
-  q0s16 = vaddq_s16(*q9s16, *q15s16);
-  q1s16 = vaddq_s16(*q11s16, *q13s16);
-  q2s16 = vsubq_s16(*q11s16, *q13s16);
-  q3s16 = vsubq_s16(*q9s16, *q15s16);
-
-  *q13s16 = vsubq_s16(q4s16, q5s16);
-  q4s16 = vaddq_s16(q4s16, q5s16);
-  *q14s16 = vsubq_s16(q7s16, q6s16);
-  q7s16 = vaddq_s16(q7s16, q6s16);
-  d26s16 = vget_low_s16(*q13s16);
-  d27s16 = vget_high_s16(*q13s16);
-  d28s16 = vget_low_s16(*q14s16);
-  d29s16 = vget_high_s16(*q14s16);
-
-  d16s16 = vdup_n_s16((int16_t)cospi_16_64);
-
-  q9s32 = vmull_s16(d28s16, d16s16);
-  q10s32 = vmull_s16(d29s16, d16s16);
-  q11s32 = vmull_s16(d28s16, d16s16);
-  q12s32 = vmull_s16(d29s16, d16s16);
-
-  q9s32 = vmlsl_s16(q9s32, d26s16, d16s16);
-  q10s32 = vmlsl_s16(q10s32, d27s16, d16s16);
-  q11s32 = vmlal_s16(q11s32, d26s16, d16s16);
-  q12s32 = vmlal_s16(q12s32, d27s16, d16s16);
-
-  d10s16 = vqrshrn_n_s32(q9s32, 14);
-  d11s16 = vqrshrn_n_s32(q10s32, 14);
-  d12s16 = vqrshrn_n_s32(q11s32, 14);
-  d13s16 = vqrshrn_n_s32(q12s32, 14);
-  q5s16 = vcombine_s16(d10s16, d11s16);
-  q6s16 = vcombine_s16(d12s16, d13s16);
-
-  *q8s16 = vaddq_s16(q0s16, q7s16);
-  *q9s16 = vaddq_s16(q1s16, q6s16);
-  *q10s16 = vaddq_s16(q2s16, q5s16);
-  *q11s16 = vaddq_s16(q3s16, q4s16);
-  *q12s16 = vsubq_s16(q3s16, q4s16);
-  *q13s16 = vsubq_s16(q2s16, q5s16);
-  *q14s16 = vsubq_s16(q1s16, q6s16);
-  *q15s16 = vsubq_s16(q0s16, q7s16);
-  return;
-}
-
-void aom_idct8x8_64_add_neon(int16_t *input, uint8_t *dest, int dest_stride) {
-  uint8_t *d1, *d2;
-  uint8x8_t d0u8, d1u8, d2u8, d3u8;
-  uint64x1_t d0u64, d1u64, d2u64, d3u64;
-  int16x8_t q8s16, q9s16, q10s16, q11s16, q12s16, q13s16, q14s16, q15s16;
-  uint16x8_t q8u16, q9u16, q10u16, q11u16;
-
-  q8s16 = vld1q_s16(input);
-  q9s16 = vld1q_s16(input + 8);
-  q10s16 = vld1q_s16(input + 16);
-  q11s16 = vld1q_s16(input + 24);
-  q12s16 = vld1q_s16(input + 32);
-  q13s16 = vld1q_s16(input + 40);
-  q14s16 = vld1q_s16(input + 48);
-  q15s16 = vld1q_s16(input + 56);
-
-  TRANSPOSE8X8(&q8s16, &q9s16, &q10s16, &q11s16, &q12s16, &q13s16, &q14s16,
-               &q15s16);
-
-  IDCT8x8_1D(&q8s16, &q9s16, &q10s16, &q11s16, &q12s16, &q13s16, &q14s16,
-             &q15s16);
-
-  TRANSPOSE8X8(&q8s16, &q9s16, &q10s16, &q11s16, &q12s16, &q13s16, &q14s16,
-               &q15s16);
-
-  IDCT8x8_1D(&q8s16, &q9s16, &q10s16, &q11s16, &q12s16, &q13s16, &q14s16,
-             &q15s16);
-
-  q8s16 = vrshrq_n_s16(q8s16, 5);
-  q9s16 = vrshrq_n_s16(q9s16, 5);
-  q10s16 = vrshrq_n_s16(q10s16, 5);
-  q11s16 = vrshrq_n_s16(q11s16, 5);
-  q12s16 = vrshrq_n_s16(q12s16, 5);
-  q13s16 = vrshrq_n_s16(q13s16, 5);
-  q14s16 = vrshrq_n_s16(q14s16, 5);
-  q15s16 = vrshrq_n_s16(q15s16, 5);
-
-  d1 = d2 = dest;
-
-  d0u64 = vld1_u64((uint64_t *)d1);
-  d1 += dest_stride;
-  d1u64 = vld1_u64((uint64_t *)d1);
-  d1 += dest_stride;
-  d2u64 = vld1_u64((uint64_t *)d1);
-  d1 += dest_stride;
-  d3u64 = vld1_u64((uint64_t *)d1);
-  d1 += dest_stride;
-
-  q8u16 = vaddw_u8(vreinterpretq_u16_s16(q8s16), vreinterpret_u8_u64(d0u64));
-  q9u16 = vaddw_u8(vreinterpretq_u16_s16(q9s16), vreinterpret_u8_u64(d1u64));
-  q10u16 = vaddw_u8(vreinterpretq_u16_s16(q10s16), vreinterpret_u8_u64(d2u64));
-  q11u16 = vaddw_u8(vreinterpretq_u16_s16(q11s16), vreinterpret_u8_u64(d3u64));
-
-  d0u8 = vqmovun_s16(vreinterpretq_s16_u16(q8u16));
-  d1u8 = vqmovun_s16(vreinterpretq_s16_u16(q9u16));
-  d2u8 = vqmovun_s16(vreinterpretq_s16_u16(q10u16));
-  d3u8 = vqmovun_s16(vreinterpretq_s16_u16(q11u16));
-
-  vst1_u64((uint64_t *)d2, vreinterpret_u64_u8(d0u8));
-  d2 += dest_stride;
-  vst1_u64((uint64_t *)d2, vreinterpret_u64_u8(d1u8));
-  d2 += dest_stride;
-  vst1_u64((uint64_t *)d2, vreinterpret_u64_u8(d2u8));
-  d2 += dest_stride;
-  vst1_u64((uint64_t *)d2, vreinterpret_u64_u8(d3u8));
-  d2 += dest_stride;
-
-  q8s16 = q12s16;
-  q9s16 = q13s16;
-  q10s16 = q14s16;
-  q11s16 = q15s16;
-
-  d0u64 = vld1_u64((uint64_t *)d1);
-  d1 += dest_stride;
-  d1u64 = vld1_u64((uint64_t *)d1);
-  d1 += dest_stride;
-  d2u64 = vld1_u64((uint64_t *)d1);
-  d1 += dest_stride;
-  d3u64 = vld1_u64((uint64_t *)d1);
-  d1 += dest_stride;
-
-  q8u16 = vaddw_u8(vreinterpretq_u16_s16(q8s16), vreinterpret_u8_u64(d0u64));
-  q9u16 = vaddw_u8(vreinterpretq_u16_s16(q9s16), vreinterpret_u8_u64(d1u64));
-  q10u16 = vaddw_u8(vreinterpretq_u16_s16(q10s16), vreinterpret_u8_u64(d2u64));
-  q11u16 = vaddw_u8(vreinterpretq_u16_s16(q11s16), vreinterpret_u8_u64(d3u64));
-
-  d0u8 = vqmovun_s16(vreinterpretq_s16_u16(q8u16));
-  d1u8 = vqmovun_s16(vreinterpretq_s16_u16(q9u16));
-  d2u8 = vqmovun_s16(vreinterpretq_s16_u16(q10u16));
-  d3u8 = vqmovun_s16(vreinterpretq_s16_u16(q11u16));
-
-  vst1_u64((uint64_t *)d2, vreinterpret_u64_u8(d0u8));
-  d2 += dest_stride;
-  vst1_u64((uint64_t *)d2, vreinterpret_u64_u8(d1u8));
-  d2 += dest_stride;
-  vst1_u64((uint64_t *)d2, vreinterpret_u64_u8(d2u8));
-  d2 += dest_stride;
-  vst1_u64((uint64_t *)d2, vreinterpret_u64_u8(d3u8));
-  d2 += dest_stride;
-  return;
-}
-
-void aom_idct8x8_12_add_neon(int16_t *input, uint8_t *dest, int dest_stride) {
-  uint8_t *d1, *d2;
-  uint8x8_t d0u8, d1u8, d2u8, d3u8;
-  int16x4_t d10s16, d11s16, d12s16, d13s16, d16s16;
-  int16x4_t d26s16, d27s16, d28s16, d29s16;
-  uint64x1_t d0u64, d1u64, d2u64, d3u64;
-  int16x8_t q0s16, q1s16, q2s16, q3s16, q4s16, q5s16, q6s16, q7s16;
-  int16x8_t q8s16, q9s16, q10s16, q11s16, q12s16, q13s16, q14s16, q15s16;
-  uint16x8_t q8u16, q9u16, q10u16, q11u16;
-  int32x4_t q9s32, q10s32, q11s32, q12s32;
-
-  q8s16 = vld1q_s16(input);
-  q9s16 = vld1q_s16(input + 8);
-  q10s16 = vld1q_s16(input + 16);
-  q11s16 = vld1q_s16(input + 24);
-  q12s16 = vld1q_s16(input + 32);
-  q13s16 = vld1q_s16(input + 40);
-  q14s16 = vld1q_s16(input + 48);
-  q15s16 = vld1q_s16(input + 56);
-
-  TRANSPOSE8X8(&q8s16, &q9s16, &q10s16, &q11s16, &q12s16, &q13s16, &q14s16,
-               &q15s16);
-
-  // First transform rows
-  // stage 1
-  q0s16 = vdupq_n_s16((int16_t)cospi_28_64 * 2);
-  q1s16 = vdupq_n_s16((int16_t)cospi_4_64 * 2);
-
-  q4s16 = vqrdmulhq_s16(q9s16, q0s16);
-
-  q0s16 = vdupq_n_s16(-(int16_t)cospi_20_64 * 2);
-
-  q7s16 = vqrdmulhq_s16(q9s16, q1s16);
-
-  q1s16 = vdupq_n_s16((int16_t)cospi_12_64 * 2);
-
-  q5s16 = vqrdmulhq_s16(q11s16, q0s16);
-
-  q0s16 = vdupq_n_s16((int16_t)cospi_16_64 * 2);
-
-  q6s16 = vqrdmulhq_s16(q11s16, q1s16);
-
-  // stage 2 & stage 3 - even half
-  q1s16 = vdupq_n_s16((int16_t)cospi_24_64 * 2);
-
-  q9s16 = vqrdmulhq_s16(q8s16, q0s16);
-
-  q0s16 = vdupq_n_s16((int16_t)cospi_8_64 * 2);
-
-  q13s16 = vqrdmulhq_s16(q10s16, q1s16);
-
-  q15s16 = vqrdmulhq_s16(q10s16, q0s16);
-
-  // stage 3 -odd half
-  q0s16 = vaddq_s16(q9s16, q15s16);
-  q1s16 = vaddq_s16(q9s16, q13s16);
-  q2s16 = vsubq_s16(q9s16, q13s16);
-  q3s16 = vsubq_s16(q9s16, q15s16);
-
-  // stage 2 - odd half
-  q13s16 = vsubq_s16(q4s16, q5s16);
-  q4s16 = vaddq_s16(q4s16, q5s16);
-  q14s16 = vsubq_s16(q7s16, q6s16);
-  q7s16 = vaddq_s16(q7s16, q6s16);
-  d26s16 = vget_low_s16(q13s16);
-  d27s16 = vget_high_s16(q13s16);
-  d28s16 = vget_low_s16(q14s16);
-  d29s16 = vget_high_s16(q14s16);
-
-  d16s16 = vdup_n_s16((int16_t)cospi_16_64);
-  q9s32 = vmull_s16(d28s16, d16s16);
-  q10s32 = vmull_s16(d29s16, d16s16);
-  q11s32 = vmull_s16(d28s16, d16s16);
-  q12s32 = vmull_s16(d29s16, d16s16);
-
-  q9s32 = vmlsl_s16(q9s32, d26s16, d16s16);
-  q10s32 = vmlsl_s16(q10s32, d27s16, d16s16);
-  q11s32 = vmlal_s16(q11s32, d26s16, d16s16);
-  q12s32 = vmlal_s16(q12s32, d27s16, d16s16);
-
-  d10s16 = vqrshrn_n_s32(q9s32, 14);
-  d11s16 = vqrshrn_n_s32(q10s32, 14);
-  d12s16 = vqrshrn_n_s32(q11s32, 14);
-  d13s16 = vqrshrn_n_s32(q12s32, 14);
-  q5s16 = vcombine_s16(d10s16, d11s16);
-  q6s16 = vcombine_s16(d12s16, d13s16);
-
-  // stage 4
-  q8s16 = vaddq_s16(q0s16, q7s16);
-  q9s16 = vaddq_s16(q1s16, q6s16);
-  q10s16 = vaddq_s16(q2s16, q5s16);
-  q11s16 = vaddq_s16(q3s16, q4s16);
-  q12s16 = vsubq_s16(q3s16, q4s16);
-  q13s16 = vsubq_s16(q2s16, q5s16);
-  q14s16 = vsubq_s16(q1s16, q6s16);
-  q15s16 = vsubq_s16(q0s16, q7s16);
-
-  TRANSPOSE8X8(&q8s16, &q9s16, &q10s16, &q11s16, &q12s16, &q13s16, &q14s16,
-               &q15s16);
-
-  IDCT8x8_1D(&q8s16, &q9s16, &q10s16, &q11s16, &q12s16, &q13s16, &q14s16,
-             &q15s16);
-
-  q8s16 = vrshrq_n_s16(q8s16, 5);
-  q9s16 = vrshrq_n_s16(q9s16, 5);
-  q10s16 = vrshrq_n_s16(q10s16, 5);
-  q11s16 = vrshrq_n_s16(q11s16, 5);
-  q12s16 = vrshrq_n_s16(q12s16, 5);
-  q13s16 = vrshrq_n_s16(q13s16, 5);
-  q14s16 = vrshrq_n_s16(q14s16, 5);
-  q15s16 = vrshrq_n_s16(q15s16, 5);
-
-  d1 = d2 = dest;
-
-  d0u64 = vld1_u64((uint64_t *)d1);
-  d1 += dest_stride;
-  d1u64 = vld1_u64((uint64_t *)d1);
-  d1 += dest_stride;
-  d2u64 = vld1_u64((uint64_t *)d1);
-  d1 += dest_stride;
-  d3u64 = vld1_u64((uint64_t *)d1);
-  d1 += dest_stride;
-
-  q8u16 = vaddw_u8(vreinterpretq_u16_s16(q8s16), vreinterpret_u8_u64(d0u64));
-  q9u16 = vaddw_u8(vreinterpretq_u16_s16(q9s16), vreinterpret_u8_u64(d1u64));
-  q10u16 = vaddw_u8(vreinterpretq_u16_s16(q10s16), vreinterpret_u8_u64(d2u64));
-  q11u16 = vaddw_u8(vreinterpretq_u16_s16(q11s16), vreinterpret_u8_u64(d3u64));
-
-  d0u8 = vqmovun_s16(vreinterpretq_s16_u16(q8u16));
-  d1u8 = vqmovun_s16(vreinterpretq_s16_u16(q9u16));
-  d2u8 = vqmovun_s16(vreinterpretq_s16_u16(q10u16));
-  d3u8 = vqmovun_s16(vreinterpretq_s16_u16(q11u16));
-
-  vst1_u64((uint64_t *)d2, vreinterpret_u64_u8(d0u8));
-  d2 += dest_stride;
-  vst1_u64((uint64_t *)d2, vreinterpret_u64_u8(d1u8));
-  d2 += dest_stride;
-  vst1_u64((uint64_t *)d2, vreinterpret_u64_u8(d2u8));
-  d2 += dest_stride;
-  vst1_u64((uint64_t *)d2, vreinterpret_u64_u8(d3u8));
-  d2 += dest_stride;
-
-  q8s16 = q12s16;
-  q9s16 = q13s16;
-  q10s16 = q14s16;
-  q11s16 = q15s16;
-
-  d0u64 = vld1_u64((uint64_t *)d1);
-  d1 += dest_stride;
-  d1u64 = vld1_u64((uint64_t *)d1);
-  d1 += dest_stride;
-  d2u64 = vld1_u64((uint64_t *)d1);
-  d1 += dest_stride;
-  d3u64 = vld1_u64((uint64_t *)d1);
-  d1 += dest_stride;
-
-  q8u16 = vaddw_u8(vreinterpretq_u16_s16(q8s16), vreinterpret_u8_u64(d0u64));
-  q9u16 = vaddw_u8(vreinterpretq_u16_s16(q9s16), vreinterpret_u8_u64(d1u64));
-  q10u16 = vaddw_u8(vreinterpretq_u16_s16(q10s16), vreinterpret_u8_u64(d2u64));
-  q11u16 = vaddw_u8(vreinterpretq_u16_s16(q11s16), vreinterpret_u8_u64(d3u64));
-
-  d0u8 = vqmovun_s16(vreinterpretq_s16_u16(q8u16));
-  d1u8 = vqmovun_s16(vreinterpretq_s16_u16(q9u16));
-  d2u8 = vqmovun_s16(vreinterpretq_s16_u16(q10u16));
-  d3u8 = vqmovun_s16(vreinterpretq_s16_u16(q11u16));
-
-  vst1_u64((uint64_t *)d2, vreinterpret_u64_u8(d0u8));
-  d2 += dest_stride;
-  vst1_u64((uint64_t *)d2, vreinterpret_u64_u8(d1u8));
-  d2 += dest_stride;
-  vst1_u64((uint64_t *)d2, vreinterpret_u64_u8(d2u8));
-  d2 += dest_stride;
-  vst1_u64((uint64_t *)d2, vreinterpret_u64_u8(d3u8));
-  d2 += dest_stride;
-  return;
-}

aom_dsp/arm/intrapred_neon.c

@@ -1,819 +0,0 @@
-/*
- * Copyright (c) 2016, Alliance for Open Media. All rights reserved
- *
- * This source code is subject to the terms of the BSD 2 Clause License and
- * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
- * was not distributed with this source code in the LICENSE file, you can
- * obtain it at www.aomedia.org/license/software. If the Alliance for Open
- * Media Patent License 1.0 was not distributed with this source code in the
- * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
- */
-
-#include <arm_neon.h>
-
-#include "./aom_config.h"
-#include "./aom_dsp_rtcd.h"
-#include "aom/aom_integer.h"
-
-//------------------------------------------------------------------------------
-// DC 4x4
-
-// 'do_above' and 'do_left' facilitate branch removal when inlined.
-static INLINE void dc_4x4(uint8_t *dst, ptrdiff_t stride, const uint8_t *above,
-                          const uint8_t *left, int do_above, int do_left) {
-  uint16x8_t sum_top;
-  uint16x8_t sum_left;
-  uint8x8_t dc0;
-
-  if (do_above) {
-    const uint8x8_t A = vld1_u8(above);  // top row
-    const uint16x4_t p0 = vpaddl_u8(A);  // cascading summation of the top
-    const uint16x4_t p1 = vpadd_u16(p0, p0);
-    sum_top = vcombine_u16(p1, p1);
-  }
-
-  if (do_left) {
-    const uint8x8_t L = vld1_u8(left);   // left border
-    const uint16x4_t p0 = vpaddl_u8(L);  // cascading summation of the left
-    const uint16x4_t p1 = vpadd_u16(p0, p0);
-    sum_left = vcombine_u16(p1, p1);
-  }
-
-  if (do_above && do_left) {
-    const uint16x8_t sum = vaddq_u16(sum_left, sum_top);
-    dc0 = vrshrn_n_u16(sum, 3);
-  } else if (do_above) {
-    dc0 = vrshrn_n_u16(sum_top, 2);
-  } else if (do_left) {
-    dc0 = vrshrn_n_u16(sum_left, 2);
-  } else {
-    dc0 = vdup_n_u8(0x80);
-  }
-
-  {
-    const uint8x8_t dc = vdup_lane_u8(dc0, 0);
-    int i;
-    for (i = 0; i < 4; ++i) {
-      vst1_lane_u32((uint32_t *)(dst + i * stride), vreinterpret_u32_u8(dc), 0);
-    }
-  }
-}
-
-void aom_dc_predictor_4x4_neon(uint8_t *dst, ptrdiff_t stride,
-                               const uint8_t *above, const uint8_t *left) {
-  dc_4x4(dst, stride, above, left, 1, 1);
-}
-
-void aom_dc_left_predictor_4x4_neon(uint8_t *dst, ptrdiff_t stride,
-                                    const uint8_t *above, const uint8_t *left) {
-  (void)above;
-  dc_4x4(dst, stride, NULL, left, 0, 1);
-}
-
-void aom_dc_top_predictor_4x4_neon(uint8_t *dst, ptrdiff_t stride,
-                                   const uint8_t *above, const uint8_t *left) {
-  (void)left;
-  dc_4x4(dst, stride, above, NULL, 1, 0);
-}
-
-void aom_dc_128_predictor_4x4_neon(uint8_t *dst, ptrdiff_t stride,
-                                   const uint8_t *above, const uint8_t *left) {
-  (void)above;
-  (void)left;
-  dc_4x4(dst, stride, NULL, NULL, 0, 0);
-}
-
-//------------------------------------------------------------------------------
-// DC 8x8
-
-// 'do_above' and 'do_left' facilitate branch removal when inlined.
-static INLINE void dc_8x8(uint8_t *dst, ptrdiff_t stride, const uint8_t *above,
-                          const uint8_t *left, int do_above, int do_left) {
-  uint16x8_t sum_top;
-  uint16x8_t sum_left;
-  uint8x8_t dc0;
-
-  if (do_above) {
-    const uint8x8_t A = vld1_u8(above);  // top row
-    const uint16x4_t p0 = vpaddl_u8(A);  // cascading summation of the top
-    const uint16x4_t p1 = vpadd_u16(p0, p0);
-    const uint16x4_t p2 = vpadd_u16(p1, p1);
-    sum_top = vcombine_u16(p2, p2);
-  }
-
-  if (do_left) {
-    const uint8x8_t L = vld1_u8(left);   // left border
-    const uint16x4_t p0 = vpaddl_u8(L);  // cascading summation of the left
-    const uint16x4_t p1 = vpadd_u16(p0, p0);
-    const uint16x4_t p2 = vpadd_u16(p1, p1);
-    sum_left = vcombine_u16(p2, p2);
-  }
-
-  if (do_above && do_left) {
-    const uint16x8_t sum = vaddq_u16(sum_left, sum_top);
-    dc0 = vrshrn_n_u16(sum, 4);
-  } else if (do_above) {
-    dc0 = vrshrn_n_u16(sum_top, 3);
-  } else if (do_left) {
-    dc0 = vrshrn_n_u16(sum_left, 3);
-  } else {
-    dc0 = vdup_n_u8(0x80);
-  }
-
-  {
-    const uint8x8_t dc = vdup_lane_u8(dc0, 0);
-    int i;
-    for (i = 0; i < 8; ++i) {
-      vst1_u32((uint32_t *)(dst + i * stride), vreinterpret_u32_u8(dc));
-    }
-  }
-}
-
-void aom_dc_predictor_8x8_neon(uint8_t *dst, ptrdiff_t stride,
-                               const uint8_t *above, const uint8_t *left) {
-  dc_8x8(dst, stride, above, left, 1, 1);
-}
-
-void aom_dc_left_predictor_8x8_neon(uint8_t *dst, ptrdiff_t stride,
-                                    const uint8_t *above, const uint8_t *left) {
-  (void)above;
-  dc_8x8(dst, stride, NULL, left, 0, 1);
-}
-
-void aom_dc_top_predictor_8x8_neon(uint8_t *dst, ptrdiff_t stride,
-                                   const uint8_t *above, const uint8_t *left) {
-  (void)left;
-  dc_8x8(dst, stride, above, NULL, 1, 0);
-}
-
-void aom_dc_128_predictor_8x8_neon(uint8_t *dst, ptrdiff_t stride,
-                                   const uint8_t *above, const uint8_t *left) {
-  (void)above;
-  (void)left;
-  dc_8x8(dst, stride, NULL, NULL, 0, 0);
-}
-
-//------------------------------------------------------------------------------
-// DC 16x16
-
-// 'do_above' and 'do_left' facilitate branch removal when inlined.
-static INLINE void dc_16x16(uint8_t *dst, ptrdiff_t stride,
-                            const uint8_t *above, const uint8_t *left,
-                            int do_above, int do_left) {
-  uint16x8_t sum_top;
-  uint16x8_t sum_left;
-  uint8x8_t dc0;
-
-  if (do_above) {
-    const uint8x16_t A = vld1q_u8(above);  // top row
-    const uint16x8_t p0 = vpaddlq_u8(A);   // cascading summation of the top
-    const uint16x4_t p1 = vadd_u16(vget_low_u16(p0), vget_high_u16(p0));
-    const uint16x4_t p2 = vpadd_u16(p1, p1);
-    const uint16x4_t p3 = vpadd_u16(p2, p2);
-    sum_top = vcombine_u16(p3, p3);
-  }
-
-  if (do_left) {
-    const uint8x16_t L = vld1q_u8(left);  // left row
-    const uint16x8_t p0 = vpaddlq_u8(L);  // cascading summation of the left
-    const uint16x4_t p1 = vadd_u16(vget_low_u16(p0), vget_high_u16(p0));
-    const uint16x4_t p2 = vpadd_u16(p1, p1);
-    const uint16x4_t p3 = vpadd_u16(p2, p2);
-    sum_left = vcombine_u16(p3, p3);
-  }
-
-  if (do_above && do_left) {
-    const uint16x8_t sum = vaddq_u16(sum_left, sum_top);
-    dc0 = vrshrn_n_u16(sum, 5);
-  } else if (do_above) {
-    dc0 = vrshrn_n_u16(sum_top, 4);
-  } else if (do_left) {
-    dc0 = vrshrn_n_u16(sum_left, 4);
-  } else {
-    dc0 = vdup_n_u8(0x80);
-  }
-
-  {
-    const uint8x16_t dc = vdupq_lane_u8(dc0, 0);
-    int i;
-    for (i = 0; i < 16; ++i) {
-      vst1q_u8(dst + i * stride, dc);
-    }
-  }
-}
-
-void aom_dc_predictor_16x16_neon(uint8_t *dst, ptrdiff_t stride,
-                                 const uint8_t *above, const uint8_t *left) {
-  dc_16x16(dst, stride, above, left, 1, 1);
-}
-
-void aom_dc_left_predictor_16x16_neon(uint8_t *dst, ptrdiff_t stride,
-                                      const uint8_t *above,
-                                      const uint8_t *left) {
-  (void)above;
-  dc_16x16(dst, stride, NULL, left, 0, 1);
-}
-
-void aom_dc_top_predictor_16x16_neon(uint8_t *dst, ptrdiff_t stride,
-                                     const uint8_t *above,
-                                     const uint8_t *left) {
-  (void)left;
-  dc_16x16(dst, stride, above, NULL, 1, 0);
-}
-
-void aom_dc_128_predictor_16x16_neon(uint8_t *dst, ptrdiff_t stride,
-                                     const uint8_t *above,
-                                     const uint8_t *left) {
-  (void)above;
-  (void)left;
-  dc_16x16(dst, stride, NULL, NULL, 0, 0);
-}
-
-//------------------------------------------------------------------------------
-// DC 32x32
-
-// 'do_above' and 'do_left' facilitate branch removal when inlined.
-static INLINE void dc_32x32(uint8_t *dst, ptrdiff_t stride,
-                            const uint8_t *above, const uint8_t *left,
-                            int do_above, int do_left) {
-  uint16x8_t sum_top;
-  uint16x8_t sum_left;
-  uint8x8_t dc0;
-
-  if (do_above) {
-    const uint8x16_t A0 = vld1q_u8(above);  // top row
-    const uint8x16_t A1 = vld1q_u8(above + 16);
-    const uint16x8_t p0 = vpaddlq_u8(A0);  // cascading summation of the top
-    const uint16x8_t p1 = vpaddlq_u8(A1);
-    const uint16x8_t p2 = vaddq_u16(p0, p1);
-    const uint16x4_t p3 = vadd_u16(vget_low_u16(p2), vget_high_u16(p2));
-    const uint16x4_t p4 = vpadd_u16(p3, p3);
-    const uint16x4_t p5 = vpadd_u16(p4, p4);
-    sum_top = vcombine_u16(p5, p5);
-  }
-
-  if (do_left) {
-    const uint8x16_t L0 = vld1q_u8(left);  // left row
-    const uint8x16_t L1 = vld1q_u8(left + 16);
-    const uint16x8_t p0 = vpaddlq_u8(L0);  // cascading summation of the left
-    const uint16x8_t p1 = vpaddlq_u8(L1);
-    const uint16x8_t p2 = vaddq_u16(p0, p1);
-    const uint16x4_t p3 = vadd_u16(vget_low_u16(p2), vget_high_u16(p2));
-    const uint16x4_t p4 = vpadd_u16(p3, p3);
-    const uint16x4_t p5 = vpadd_u16(p4, p4);
-    sum_left = vcombine_u16(p5, p5);
-  }
-
-  if (do_above && do_left) {
-    const uint16x8_t sum = vaddq_u16(sum_left, sum_top);
-    dc0 = vrshrn_n_u16(sum, 6);
-  } else if (do_above) {
-    dc0 = vrshrn_n_u16(sum_top, 5);
-  } else if (do_left) {
-    dc0 = vrshrn_n_u16(sum_left, 5);
-  } else {
-    dc0 = vdup_n_u8(0x80);
-  }
-
-  {
-    const uint8x16_t dc = vdupq_lane_u8(dc0, 0);
-    int i;
-    for (i = 0; i < 32; ++i) {
-      vst1q_u8(dst + i * stride, dc);
-      vst1q_u8(dst + i * stride + 16, dc);
-    }
-  }
-}
-
-void aom_dc_predictor_32x32_neon(uint8_t *dst, ptrdiff_t stride,
-                                 const uint8_t *above, const uint8_t *left) {
-  dc_32x32(dst, stride, above, left, 1, 1);
-}
-
-void aom_dc_left_predictor_32x32_neon(uint8_t *dst, ptrdiff_t stride,
-                                      const uint8_t *above,
-                                      const uint8_t *left) {
-  (void)above;
-  dc_32x32(dst, stride, NULL, left, 0, 1);
-}
-
-void aom_dc_top_predictor_32x32_neon(uint8_t *dst, ptrdiff_t stride,
-                                     const uint8_t *above,
-                                     const uint8_t *left) {
-  (void)left;
-  dc_32x32(dst, stride, above, NULL, 1, 0);
-}
-
-void aom_dc_128_predictor_32x32_neon(uint8_t *dst, ptrdiff_t stride,
-                                     const uint8_t *above,
-                                     const uint8_t *left) {
-  (void)above;
-  (void)left;
-  dc_32x32(dst, stride, NULL, NULL, 0, 0);
-}
-
-// -----------------------------------------------------------------------------
-
-void aom_d45_predictor_4x4_neon(uint8_t *dst, ptrdiff_t stride,
-                                const uint8_t *above, const uint8_t *left) {
-  const uint64x1_t A0 = vreinterpret_u64_u8(vld1_u8(above));  // top row
-  const uint64x1_t A1 = vshr_n_u64(A0, 8);
-  const uint64x1_t A2 = vshr_n_u64(A0, 16);
-  const uint8x8_t ABCDEFGH = vreinterpret_u8_u64(A0);
-  const uint8x8_t BCDEFGH0 = vreinterpret_u8_u64(A1);
-  const uint8x8_t CDEFGH00 = vreinterpret_u8_u64(A2);
-  const uint8x8_t avg1 = vhadd_u8(ABCDEFGH, CDEFGH00);
-  const uint8x8_t avg2 = vrhadd_u8(avg1, BCDEFGH0);
-  const uint64x1_t avg2_u64 = vreinterpret_u64_u8(avg2);
-  const uint32x2_t r0 = vreinterpret_u32_u8(avg2);
-  const uint32x2_t r1 = vreinterpret_u32_u64(vshr_n_u64(avg2_u64, 8));
-  const uint32x2_t r2 = vreinterpret_u32_u64(vshr_n_u64(avg2_u64, 16));
-  const uint32x2_t r3 = vreinterpret_u32_u64(vshr_n_u64(avg2_u64, 24));
-  (void)left;
-  vst1_lane_u32((uint32_t *)(dst + 0 * stride), r0, 0);
-  vst1_lane_u32((uint32_t *)(dst + 1 * stride), r1, 0);
-  vst1_lane_u32((uint32_t *)(dst + 2 * stride), r2, 0);
-  vst1_lane_u32((uint32_t *)(dst + 3 * stride), r3, 0);
-  dst[3 * stride + 3] = above[7];
-}
-
-void aom_d45_predictor_8x8_neon(uint8_t *dst, ptrdiff_t stride,
-                                const uint8_t *above, const uint8_t *left) {
-  static const uint8_t shuffle1[8] = { 1, 2, 3, 4, 5, 6, 7, 7 };
-  static const uint8_t shuffle2[8] = { 2, 3, 4, 5, 6, 7, 7, 7 };
-  const uint8x8_t sh_12345677 = vld1_u8(shuffle1);
-  const uint8x8_t sh_23456777 = vld1_u8(shuffle2);
-  const uint8x8_t A0 = vld1_u8(above);  // top row
-  const uint8x8_t A1 = vtbl1_u8(A0, sh_12345677);
-  const uint8x8_t A2 = vtbl1_u8(A0, sh_23456777);
-  const uint8x8_t avg1 = vhadd_u8(A0, A2);
-  uint8x8_t row = vrhadd_u8(avg1, A1);
-  int i;
-  (void)left;
-  for (i = 0; i < 7; ++i) {
-    vst1_u8(dst + i * stride, row);
-    row = vtbl1_u8(row, sh_12345677);
-  }
-  vst1_u8(dst + i * stride, row);
-}
-
-void aom_d45_predictor_16x16_neon(uint8_t *dst, ptrdiff_t stride,
-                                  const uint8_t *above, const uint8_t *left) {
-  const uint8x16_t A0 = vld1q_u8(above);  // top row
-  const uint8x16_t above_right = vld1q_dup_u8(above + 15);
-  const uint8x16_t A1 = vextq_u8(A0, above_right, 1);
-  const uint8x16_t A2 = vextq_u8(A0, above_right, 2);
-  const uint8x16_t avg1 = vhaddq_u8(A0, A2);
-  uint8x16_t row = vrhaddq_u8(avg1, A1);
-  int i;
-  (void)left;
-  for (i = 0; i < 15; ++i) {
-    vst1q_u8(dst + i * stride, row);
-    row = vextq_u8(row, above_right, 1);
-  }
-  vst1q_u8(dst + i * stride, row);
-}
-
-// -----------------------------------------------------------------------------
-
-void aom_d135_predictor_4x4_neon(uint8_t *dst, ptrdiff_t stride,
-                                 const uint8_t *above, const uint8_t *left) {
-  const uint8x8_t XABCD_u8 = vld1_u8(above - 1);
-  const uint64x1_t XABCD = vreinterpret_u64_u8(XABCD_u8);
-  const uint64x1_t ____XABC = vshl_n_u64(XABCD, 32);
-  const uint32x2_t zero = vdup_n_u32(0);
-  const uint32x2_t IJKL = vld1_lane_u32((const uint32_t *)left, zero, 0);
-  const uint8x8_t IJKL_u8 = vreinterpret_u8_u32(IJKL);
-  const uint64x1_t LKJI____ = vreinterpret_u64_u8(vrev32_u8(IJKL_u8));
-  const uint64x1_t LKJIXABC = vorr_u64(LKJI____, ____XABC);
-  const uint8x8_t KJIXABC_ = vreinterpret_u8_u64(vshr_n_u64(LKJIXABC, 8));
-  const uint8x8_t JIXABC__ = vreinterpret_u8_u64(vshr_n_u64(LKJIXABC, 16));
-  const uint8_t D = vget_lane_u8(XABCD_u8, 4);
-  const uint8x8_t JIXABCD_ = vset_lane_u8(D, JIXABC__, 6);
-  const uint8x8_t LKJIXABC_u8 = vreinterpret_u8_u64(LKJIXABC);
-  const uint8x8_t avg1 = vhadd_u8(JIXABCD_, LKJIXABC_u8);
-  const uint8x8_t avg2 = vrhadd_u8(avg1, KJIXABC_);
-  const uint64x1_t avg2_u64 = vreinterpret_u64_u8(avg2);
-  const uint32x2_t r3 = vreinterpret_u32_u8(avg2);
-  const uint32x2_t r2 = vreinterpret_u32_u64(vshr_n_u64(avg2_u64, 8));
-  const uint32x2_t r1 = vreinterpret_u32_u64(vshr_n_u64(avg2_u64, 16));
-  const uint32x2_t r0 = vreinterpret_u32_u64(vshr_n_u64(avg2_u64, 24));
-  vst1_lane_u32((uint32_t *)(dst + 0 * stride), r0, 0);
-  vst1_lane_u32((uint32_t *)(dst + 1 * stride), r1, 0);
-  vst1_lane_u32((uint32_t *)(dst + 2 * stride), r2, 0);
-  vst1_lane_u32((uint32_t *)(dst + 3 * stride), r3, 0);
-}
-
-#if !HAVE_NEON_ASM
-
-void aom_v_predictor_4x4_neon(uint8_t *dst, ptrdiff_t stride,
-                              const uint8_t *above, const uint8_t *left) {
-  int i;
-  uint32x2_t d0u32 = vdup_n_u32(0);
-  (void)left;
-
-  d0u32 = vld1_lane_u32((const uint32_t *)above, d0u32, 0);
-  for (i = 0; i < 4; i++, dst += stride)
-    vst1_lane_u32((uint32_t *)dst, d0u32, 0);
-}
-
-void aom_v_predictor_8x8_neon(uint8_t *dst, ptrdiff_t stride,
-                              const uint8_t *above, const uint8_t *left) {
-  int i;
-  uint8x8_t d0u8 = vdup_n_u8(0);
-  (void)left;
-
-  d0u8 = vld1_u8(above);
-  for (i = 0; i < 8; i++, dst += stride) vst1_u8(dst, d0u8);
-}
-
-void aom_v_predictor_16x16_neon(uint8_t *dst, ptrdiff_t stride,
-                                const uint8_t *above, const uint8_t *left) {
-  int i;
-  uint8x16_t q0u8 = vdupq_n_u8(0);
-  (void)left;
-
-  q0u8 = vld1q_u8(above);
-  for (i = 0; i < 16; i++, dst += stride) vst1q_u8(dst, q0u8);
-}
-
-void aom_v_predictor_32x32_neon(uint8_t *dst, ptrdiff_t stride,
-                                const uint8_t *above, const uint8_t *left) {
-  int i;
-  uint8x16_t q0u8 = vdupq_n_u8(0);
-  uint8x16_t q1u8 = vdupq_n_u8(0);
-  (void)left;
-
-  q0u8 = vld1q_u8(above);
-  q1u8 = vld1q_u8(above + 16);
-  for (i = 0; i < 32; i++, dst += stride) {
-    vst1q_u8(dst, q0u8);
-    vst1q_u8(dst + 16, q1u8);
-  }
-}
-
-void aom_h_predictor_4x4_neon(uint8_t *dst, ptrdiff_t stride,
-                              const uint8_t *above, const uint8_t *left) {
-  uint8x8_t d0u8 = vdup_n_u8(0);
-  uint32x2_t d1u32 = vdup_n_u32(0);
-  (void)above;
-
-  d1u32 = vld1_lane_u32((const uint32_t *)left, d1u32, 0);
-
-  d0u8 = vdup_lane_u8(vreinterpret_u8_u32(d1u32), 0);
-  vst1_lane_u32((uint32_t *)dst, vreinterpret_u32_u8(d0u8), 0);
-  dst += stride;
-  d0u8 = vdup_lane_u8(vreinterpret_u8_u32(d1u32), 1);
-  vst1_lane_u32((uint32_t *)dst, vreinterpret_u32_u8(d0u8), 0);
-  dst += stride;
-  d0u8 = vdup_lane_u8(vreinterpret_u8_u32(d1u32), 2);
-  vst1_lane_u32((uint32_t *)dst, vreinterpret_u32_u8(d0u8), 0);
-  dst += stride;
-  d0u8 = vdup_lane_u8(vreinterpret_u8_u32(d1u32), 3);
-  vst1_lane_u32((uint32_t *)dst, vreinterpret_u32_u8(d0u8), 0);
-}
-
-void aom_h_predictor_8x8_neon(uint8_t *dst, ptrdiff_t stride,
-                              const uint8_t *above, const uint8_t *left) {
-  uint8x8_t d0u8 = vdup_n_u8(0);
-  uint64x1_t d1u64 = vdup_n_u64(0);
-  (void)above;
-
-  d1u64 = vld1_u64((const uint64_t *)left);
-
-  d0u8 = vdup_lane_u8(vreinterpret_u8_u64(d1u64), 0);
-  vst1_u8(dst, d0u8);
-  dst += stride;
-  d0u8 = vdup_lane_u8(vreinterpret_u8_u64(d1u64), 1);
-  vst1_u8(dst, d0u8);
-  dst += stride;
-  d0u8 = vdup_lane_u8(vreinterpret_u8_u64(d1u64), 2);
-  vst1_u8(dst, d0u8);
-  dst += stride;
-  d0u8 = vdup_lane_u8(vreinterpret_u8_u64(d1u64), 3);
-  vst1_u8(dst, d0u8);
-  dst += stride;
-  d0u8 = vdup_lane_u8(vreinterpret_u8_u64(d1u64), 4);
-  vst1_u8(dst, d0u8);
-  dst += stride;
-  d0u8 = vdup_lane_u8(vreinterpret_u8_u64(d1u64), 5);
-  vst1_u8(dst, d0u8);
-  dst += stride;
-  d0u8 = vdup_lane_u8(vreinterpret_u8_u64(d1u64), 6);
-  vst1_u8(dst, d0u8);
-  dst += stride;
-  d0u8 = vdup_lane_u8(vreinterpret_u8_u64(d1u64), 7);
-  vst1_u8(dst, d0u8);
-}
-
-void aom_h_predictor_16x16_neon(uint8_t *dst, ptrdiff_t stride,
-                                const uint8_t *above, const uint8_t *left) {
-  int j;
-  uint8x8_t d2u8 = vdup_n_u8(0);
-  uint8x16_t q0u8 = vdupq_n_u8(0);
-  uint8x16_t q1u8 = vdupq_n_u8(0);
-  (void)above;
-
-  q1u8 = vld1q_u8(left);
-  d2u8 = vget_low_u8(q1u8);
-  for (j = 0; j < 2; j++, d2u8 = vget_high_u8(q1u8)) {
-    q0u8 = vdupq_lane_u8(d2u8, 0);
-    vst1q_u8(dst, q0u8);
-    dst += stride;
-    q0u8 = vdupq_lane_u8(d2u8, 1);
-    vst1q_u8(dst, q0u8);
-    dst += stride;
-    q0u8 = vdupq_lane_u8(d2u8, 2);
-    vst1q_u8(dst, q0u8);
-    dst += stride;
-    q0u8 = vdupq_lane_u8(d2u8, 3);
-    vst1q_u8(dst, q0u8);
-    dst += stride;
-    q0u8 = vdupq_lane_u8(d2u8, 4);
-    vst1q_u8(dst, q0u8);
-    dst += stride;
-    q0u8 = vdupq_lane_u8(d2u8, 5);
-    vst1q_u8(dst, q0u8);
-    dst += stride;
-    q0u8 = vdupq_lane_u8(d2u8, 6);
-    vst1q_u8(dst, q0u8);
-    dst += stride;
-    q0u8 = vdupq_lane_u8(d2u8, 7);
-    vst1q_u8(dst, q0u8);
-    dst += stride;
-  }
-}
-
-void aom_h_predictor_32x32_neon(uint8_t *dst, ptrdiff_t stride,
-                                const uint8_t *above, const uint8_t *left) {
-  int j, k;
-  uint8x8_t d2u8 = vdup_n_u8(0);
-  uint8x16_t q0u8 = vdupq_n_u8(0);
-  uint8x16_t q1u8 = vdupq_n_u8(0);
-  (void)above;
-
-  for (k = 0; k < 2; k++, left += 16) {
-    q1u8 = vld1q_u8(left);
-    d2u8 = vget_low_u8(q1u8);
-    for (j = 0; j < 2; j++, d2u8 = vget_high_u8(q1u8)) {
-      q0u8 = vdupq_lane_u8(d2u8, 0);
-      vst1q_u8(dst, q0u8);
-      vst1q_u8(dst + 16, q0u8);
-      dst += stride;
-      q0u8 = vdupq_lane_u8(d2u8, 1);
-      vst1q_u8(dst, q0u8);
-      vst1q_u8(dst + 16, q0u8);
-      dst += stride;
-      q0u8 = vdupq_lane_u8(d2u8, 2);
-      vst1q_u8(dst, q0u8);
-      vst1q_u8(dst + 16, q0u8);
-      dst += stride;
-      q0u8 = vdupq_lane_u8(d2u8, 3);
-      vst1q_u8(dst, q0u8);
-      vst1q_u8(dst + 16, q0u8);
-      dst += stride;
-      q0u8 = vdupq_lane_u8(d2u8, 4);
-      vst1q_u8(dst, q0u8);
-      vst1q_u8(dst + 16, q0u8);
-      dst += stride;
-      q0u8 = vdupq_lane_u8(d2u8, 5);
-      vst1q_u8(dst, q0u8);
-      vst1q_u8(dst + 16, q0u8);
-      dst += stride;
-      q0u8 = vdupq_lane_u8(d2u8, 6);
-      vst1q_u8(dst, q0u8);
-      vst1q_u8(dst + 16, q0u8);
-      dst += stride;
-      q0u8 = vdupq_lane_u8(d2u8, 7);
-      vst1q_u8(dst, q0u8);
-      vst1q_u8(dst + 16, q0u8);
-      dst += stride;
-    }
-  }
-}
-
-void aom_tm_predictor_4x4_neon(uint8_t *dst, ptrdiff_t stride,
-                               const uint8_t *above, const uint8_t *left) {
-  int i;
-  uint16x8_t q1u16, q3u16;
-  int16x8_t q1s16;
-  uint8x8_t d0u8 = vdup_n_u8(0);
-  uint32x2_t d2u32 = vdup_n_u32(0);
-
-  d0u8 = vld1_dup_u8(above - 1);
-  d2u32 = vld1_lane_u32((const uint32_t *)above, d2u32, 0);
-  q3u16 = vsubl_u8(vreinterpret_u8_u32(d2u32), d0u8);
-  for (i = 0; i < 4; i++, dst += stride) {
-    q1u16 = vdupq_n_u16((uint16_t)left[i]);
-    q1s16 =
-        vaddq_s16(vreinterpretq_s16_u16(q1u16), vreinterpretq_s16_u16(q3u16));
-    d0u8 = vqmovun_s16(q1s16);
-    vst1_lane_u32((uint32_t *)dst, vreinterpret_u32_u8(d0u8), 0);
-  }
-}
-
-void aom_tm_predictor_8x8_neon(uint8_t *dst, ptrdiff_t stride,
-                               const uint8_t *above, const uint8_t *left) {
-  int j;
-  uint16x8_t q0u16, q3u16, q10u16;
-  int16x8_t q0s16;
-  uint16x4_t d20u16;
-  uint8x8_t d0u8, d2u8, d30u8;
-
-  d0u8 = vld1_dup_u8(above - 1);
-  d30u8 = vld1_u8(left);
-  d2u8 = vld1_u8(above);
-  q10u16 = vmovl_u8(d30u8);
-  q3u16 = vsubl_u8(d2u8, d0u8);
-  d20u16 = vget_low_u16(q10u16);
-  for (j = 0; j < 2; j++, d20u16 = vget_high_u16(q10u16)) {
-    q0u16 = vdupq_lane_u16(d20u16, 0);
-    q0s16 =
-        vaddq_s16(vreinterpretq_s16_u16(q3u16), vreinterpretq_s16_u16(q0u16));
-    d0u8 = vqmovun_s16(q0s16);
-    vst1_u64((uint64_t *)dst, vreinterpret_u64_u8(d0u8));
-    dst += stride;
-    q0u16 = vdupq_lane_u16(d20u16, 1);
-    q0s16 =
-        vaddq_s16(vreinterpretq_s16_u16(q3u16), vreinterpretq_s16_u16(q0u16));
-    d0u8 = vqmovun_s16(q0s16);
-    vst1_u64((uint64_t *)dst, vreinterpret_u64_u8(d0u8));
-    dst += stride;
-    q0u16 = vdupq_lane_u16(d20u16, 2);
-    q0s16 =
-        vaddq_s16(vreinterpretq_s16_u16(q3u16), vreinterpretq_s16_u16(q0u16));
-    d0u8 = vqmovun_s16(q0s16);
-    vst1_u64((uint64_t *)dst, vreinterpret_u64_u8(d0u8));
-    dst += stride;
-    q0u16 = vdupq_lane_u16(d20u16, 3);
-    q0s16 =
-        vaddq_s16(vreinterpretq_s16_u16(q3u16), vreinterpretq_s16_u16(q0u16));
-    d0u8 = vqmovun_s16(q0s16);
-    vst1_u64((uint64_t *)dst, vreinterpret_u64_u8(d0u8));
-    dst += stride;
-  }
-}
-
-void aom_tm_predictor_16x16_neon(uint8_t *dst, ptrdiff_t stride,
-                                 const uint8_t *above, const uint8_t *left) {
-  int j, k;
-  uint16x8_t q0u16, q2u16, q3u16, q8u16, q10u16;
-  uint8x16_t q0u8, q1u8;
-  int16x8_t q0s16, q1s16, q8s16, q11s16;
-  uint16x4_t d20u16;
-  uint8x8_t d2u8, d3u8, d18u8, d22u8, d23u8;
-
-  q0u8 = vld1q_dup_u8(above - 1);
-  q1u8 = vld1q_u8(above);
-  q2u16 = vsubl_u8(vget_low_u8(q1u8), vget_low_u8(q0u8));
-  q3u16 = vsubl_u8(vget_high_u8(q1u8), vget_high_u8(q0u8));
-  for (k = 0; k < 2; k++, left += 8) {
-    d18u8 = vld1_u8(left);
-    q10u16 = vmovl_u8(d18u8);
-    d20u16 = vget_low_u16(q10u16);
-    for (j = 0; j < 2; j++, d20u16 = vget_high_u16(q10u16)) {
-      q0u16 = vdupq_lane_u16(d20u16, 0);
-      q8u16 = vdupq_lane_u16(d20u16, 1);
-      q1s16 =
-          vaddq_s16(vreinterpretq_s16_u16(q0u16), vreinterpretq_s16_u16(q2u16));
-      q0s16 =
-          vaddq_s16(vreinterpretq_s16_u16(q0u16), vreinterpretq_s16_u16(q3u16));
-      q11s16 =
-          vaddq_s16(vreinterpretq_s16_u16(q8u16), vreinterpretq_s16_u16(q2u16));
-      q8s16 =
-          vaddq_s16(vreinterpretq_s16_u16(q8u16), vreinterpretq_s16_u16(q3u16));
-      d2u8 = vqmovun_s16(q1s16);
-      d3u8 = vqmovun_s16(q0s16);
-      d22u8 = vqmovun_s16(q11s16);
-      d23u8 = vqmovun_s16(q8s16);
-      vst1_u64((uint64_t *)dst, vreinterpret_u64_u8(d2u8));
-      vst1_u64((uint64_t *)(dst + 8), vreinterpret_u64_u8(d3u8));
-      dst += stride;
-      vst1_u64((uint64_t *)dst, vreinterpret_u64_u8(d22u8));
-      vst1_u64((uint64_t *)(dst + 8), vreinterpret_u64_u8(d23u8));
-      dst += stride;
-
-      q0u16 = vdupq_lane_u16(d20u16, 2);
-      q8u16 = vdupq_lane_u16(d20u16, 3);
-      q1s16 =
-          vaddq_s16(vreinterpretq_s16_u16(q0u16), vreinterpretq_s16_u16(q2u16));
-      q0s16 =
-          vaddq_s16(vreinterpretq_s16_u16(q0u16), vreinterpretq_s16_u16(q3u16));
-      q11s16 =
-          vaddq_s16(vreinterpretq_s16_u16(q8u16), vreinterpretq_s16_u16(q2u16));
-      q8s16 =
-          vaddq_s16(vreinterpretq_s16_u16(q8u16), vreinterpretq_s16_u16(q3u16));
-      d2u8 = vqmovun_s16(q1s16);
-      d3u8 = vqmovun_s16(q0s16);
-      d22u8 = vqmovun_s16(q11s16);
-      d23u8 = vqmovun_s16(q8s16);
-      vst1_u64((uint64_t *)dst, vreinterpret_u64_u8(d2u8));
-      vst1_u64((uint64_t *)(dst + 8), vreinterpret_u64_u8(d3u8));
-      dst += stride;
-      vst1_u64((uint64_t *)dst, vreinterpret_u64_u8(d22u8));
-      vst1_u64((uint64_t *)(dst + 8), vreinterpret_u64_u8(d23u8));
-      dst += stride;
-    }
-  }
-}
-
-void aom_tm_predictor_32x32_neon(uint8_t *dst, ptrdiff_t stride,
-                                 const uint8_t *above, const uint8_t *left) {
-  int j, k;
-  uint16x8_t q0u16, q3u16, q8u16, q9u16, q10u16, q11u16;
-  uint8x16_t q0u8, q1u8, q2u8;
-  int16x8_t q12s16, q13s16, q14s16, q15s16;
-  uint16x4_t d6u16;
-  uint8x8_t d0u8, d1u8, d2u8, d3u8, d26u8;
-
-  q0u8 = vld1q_dup_u8(above - 1);
-  q1u8 = vld1q_u8(above);
-  q2u8 = vld1q_u8(above + 16);
-  q8u16 = vsubl_u8(vget_low_u8(q1u8), vget_low_u8(q0u8));
-  q9u16 = vsubl_u8(vget_high_u8(q1u8), vget_high_u8(q0u8));
-  q10u16 = vsubl_u8(vget_low_u8(q2u8), vget_low_u8(q0u8));
-  q11u16 = vsubl_u8(vget_high_u8(q2u8), vget_high_u8(q0u8));
-  for (k = 0; k < 4; k++, left += 8) {
-    d26u8 = vld1_u8(left);
-    q3u16 = vmovl_u8(d26u8);
-    d6u16 = vget_low_u16(q3u16);
-    for (j = 0; j < 2; j++, d6u16 = vget_high_u16(q3u16)) {
-      q0u16 = vdupq_lane_u16(d6u16, 0);
-      q12s16 =
-          vaddq_s16(vreinterpretq_s16_u16(q0u16), vreinterpretq_s16_u16(q8u16));
-      q13s16 =
-          vaddq_s16(vreinterpretq_s16_u16(q0u16), vreinterpretq_s16_u16(q9u16));
-      q14s16 = vaddq_s16(vreinterpretq_s16_u16(q0u16),
-                         vreinterpretq_s16_u16(q10u16));
-      q15s16 = vaddq_s16(vreinterpretq_s16_u16(q0u16),
-                         vreinterpretq_s16_u16(q11u16));
-      d0u8 = vqmovun_s16(q12s16);
-      d1u8 = vqmovun_s16(q13s16);
-      d2u8 = vqmovun_s16(q14s16);
-      d3u8 = vqmovun_s16(q15s16);
-      q0u8 = vcombine_u8(d0u8, d1u8);
-      q1u8 = vcombine_u8(d2u8, d3u8);
-      vst1q_u64((uint64_t *)dst, vreinterpretq_u64_u8(q0u8));
-      vst1q_u64((uint64_t *)(dst + 16), vreinterpretq_u64_u8(q1u8));
-      dst += stride;
-
-      q0u16 = vdupq_lane_u16(d6u16, 1);
-      q12s16 =
-          vaddq_s16(vreinterpretq_s16_u16(q0u16), vreinterpretq_s16_u16(q8u16));
-      q13s16 =
-          vaddq_s16(vreinterpretq_s16_u16(q0u16), vreinterpretq_s16_u16(q9u16));
-      q14s16 = vaddq_s16(vreinterpretq_s16_u16(q0u16),
-                         vreinterpretq_s16_u16(q10u16));
-      q15s16 = vaddq_s16(vreinterpretq_s16_u16(q0u16),
-                         vreinterpretq_s16_u16(q11u16));
-      d0u8 = vqmovun_s16(q12s16);
-      d1u8 = vqmovun_s16(q13s16);
-      d2u8 = vqmovun_s16(q14s16);
-      d3u8 = vqmovun_s16(q15s16);
-      q0u8 = vcombine_u8(d0u8, d1u8);
-      q1u8 = vcombine_u8(d2u8, d3u8);
-      vst1q_u64((uint64_t *)dst, vreinterpretq_u64_u8(q0u8));
-      vst1q_u64((uint64_t *)(dst + 16), vreinterpretq_u64_u8(q1u8));
-      dst += stride;
-
-      q0u16 = vdupq_lane_u16(d6u16, 2);
-      q12s16 =
-          vaddq_s16(vreinterpretq_s16_u16(q0u16), vreinterpretq_s16_u16(q8u16));
-      q13s16 =
-          vaddq_s16(vreinterpretq_s16_u16(q0u16), vreinterpretq_s16_u16(q9u16));
-      q14s16 = vaddq_s16(vreinterpretq_s16_u16(q0u16),
-                         vreinterpretq_s16_u16(q10u16));
-      q15s16 = vaddq_s16(vreinterpretq_s16_u16(q0u16),
-                         vreinterpretq_s16_u16(q11u16));
-      d0u8 = vqmovun_s16(q12s16);
-      d1u8 = vqmovun_s16(q13s16);
-      d2u8 = vqmovun_s16(q14s16);
-      d3u8 = vqmovun_s16(q15s16);
-      q0u8 = vcombine_u8(d0u8, d1u8);
-      q1u8 = vcombine_u8(d2u8, d3u8);
-      vst1q_u64((uint64_t *)dst, vreinterpretq_u64_u8(q0u8));
-      vst1q_u64((uint64_t *)(dst + 16), vreinterpretq_u64_u8(q1u8));
-      dst += stride;
-
-      q0u16 = vdupq_lane_u16(d6u16, 3);
-      q12s16 =
-          vaddq_s16(vreinterpretq_s16_u16(q0u16), vreinterpretq_s16_u16(q8u16));
-      q13s16 =
-          vaddq_s16(vreinterpretq_s16_u16(q0u16), vreinterpretq_s16_u16(q9u16));
-      q14s16 = vaddq_s16(vreinterpretq_s16_u16(q0u16),
-                         vreinterpretq_s16_u16(q10u16));
-      q15s16 = vaddq_s16(vreinterpretq_s16_u16(q0u16),
-                         vreinterpretq_s16_u16(q11u16));
-      d0u8 = vqmovun_s16(q12s16);
-      d1u8 = vqmovun_s16(q13s16);
-      d2u8 = vqmovun_s16(q14s16);
-      d3u8 = vqmovun_s16(q15s16);
-      q0u8 = vcombine_u8(d0u8, d1u8);
-      q1u8 = vcombine_u8(d2u8, d3u8);
-      vst1q_u64((uint64_t *)dst, vreinterpretq_u64_u8(q0u8));
-      vst1q_u64((uint64_t *)(dst + 16), vreinterpretq_u64_u8(q1u8));
-      dst += stride;
-    }
-  }
-}
-#endif  // !HAVE_NEON_ASM

aom_dsp/arm/intrapred_neon_asm.asm

@@ -1,633 +0,0 @@
-;
-; Copyright (c) 2016, Alliance for Open Media. All rights reserved
-;
-; This source code is subject to the terms of the BSD 2 Clause License and
-; the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
-; was not distributed with this source code in the LICENSE file, you can
-; obtain it at www.aomedia.org/license/software. If the Alliance for Open
-; Media Patent License 1.0 was not distributed with this source code in the
-; PATENTS file, you can obtain it at www.aomedia.org/license/patent.
-;
-
-;
-
-    EXPORT  |aom_v_predictor_4x4_neon|
-    EXPORT  |aom_v_predictor_8x8_neon|
-    EXPORT  |aom_v_predictor_16x16_neon|
-    EXPORT  |aom_v_predictor_32x32_neon|
-    EXPORT  |aom_h_predictor_4x4_neon|
-    EXPORT  |aom_h_predictor_8x8_neon|
-    EXPORT  |aom_h_predictor_16x16_neon|
-    EXPORT  |aom_h_predictor_32x32_neon|
-    EXPORT  |aom_tm_predictor_4x4_neon|
-    EXPORT  |aom_tm_predictor_8x8_neon|
-    EXPORT  |aom_tm_predictor_16x16_neon|
-    EXPORT  |aom_tm_predictor_32x32_neon|
-    ARM
-    REQUIRE8
-    PRESERVE8
-
-    AREA ||.text||, CODE, READONLY, ALIGN=2
-
-;void aom_v_predictor_4x4_neon(uint8_t *dst, ptrdiff_t y_stride,
-;                              const uint8_t *above,
-;                              const uint8_t *left)
-; r0  uint8_t *dst
-; r1  ptrdiff_t y_stride
-; r2  const uint8_t *above
-; r3  const uint8_t *left
-
-|aom_v_predictor_4x4_neon| PROC
-    vld1.32             {d0[0]}, [r2]
-    vst1.32             {d0[0]}, [r0], r1
-    vst1.32             {d0[0]}, [r0], r1
-    vst1.32             {d0[0]}, [r0], r1
-    vst1.32             {d0[0]}, [r0], r1
-    bx                  lr
-    ENDP                ; |aom_v_predictor_4x4_neon|
-
-;void aom_v_predictor_8x8_neon(uint8_t *dst, ptrdiff_t y_stride,
-;                              const uint8_t *above,
-;                              const uint8_t *left)
-; r0  uint8_t *dst
-; r1  ptrdiff_t y_stride
-; r2  const uint8_t *above
-; r3  const uint8_t *left
-
-|aom_v_predictor_8x8_neon| PROC
-    vld1.8              {d0}, [r2]
-    vst1.8              {d0}, [r0], r1
-    vst1.8              {d0}, [r0], r1
-    vst1.8              {d0}, [r0], r1
-    vst1.8              {d0}, [r0], r1
-    vst1.8              {d0}, [r0], r1
-    vst1.8              {d0}, [r0], r1
-    vst1.8              {d0}, [r0], r1
-    vst1.8              {d0}, [r0], r1
-    bx                  lr
-    ENDP                ; |aom_v_predictor_8x8_neon|
-
-;void aom_v_predictor_16x16_neon(uint8_t *dst, ptrdiff_t y_stride,
-;                                const uint8_t *above,
-;                                const uint8_t *left)
-; r0  uint8_t *dst
-; r1  ptrdiff_t y_stride
-; r2  const uint8_t *above
-; r3  const uint8_t *left
-
-|aom_v_predictor_16x16_neon| PROC
-    vld1.8              {q0}, [r2]
-    vst1.8              {q0}, [r0], r1
-    vst1.8              {q0}, [r0], r1
-    vst1.8              {q0}, [r0], r1
-    vst1.8              {q0}, [r0], r1
-    vst1.8              {q0}, [r0], r1
-    vst1.8              {q0}, [r0], r1
-    vst1.8              {q0}, [r0], r1
-    vst1.8              {q0}, [r0], r1
-    vst1.8              {q0}, [r0], r1
-    vst1.8              {q0}, [r0], r1
-    vst1.8              {q0}, [r0], r1
-    vst1.8              {q0}, [r0], r1
-    vst1.8              {q0}, [r0], r1
-    vst1.8              {q0}, [r0], r1
-    vst1.8              {q0}, [r0], r1
-    vst1.8              {q0}, [r0], r1
-    bx                  lr
-    ENDP                ; |aom_v_predictor_16x16_neon|
-
-;void aom_v_predictor_32x32_neon(uint8_t *dst, ptrdiff_t y_stride,
-;                                const uint8_t *above,
-;                                const uint8_t *left)
-; r0  uint8_t *dst
-; r1  ptrdiff_t y_stride
-; r2  const uint8_t *above
-; r3  const uint8_t *left
-
-|aom_v_predictor_32x32_neon| PROC
-    vld1.8              {q0, q1}, [r2]
-    mov                 r2, #2
-loop_v
-    vst1.8              {q0, q1}, [r0], r1
-    vst1.8              {q0, q1}, [r0], r1
-    vst1.8              {q0, q1}, [r0], r1
-    vst1.8              {q0, q1}, [r0], r1
-    vst1.8              {q0, q1}, [r0], r1
-    vst1.8              {q0, q1}, [r0], r1
-    vst1.8              {q0, q1}, [r0], r1
-    vst1.8              {q0, q1}, [r0], r1
-    vst1.8              {q0, q1}, [r0], r1
-    vst1.8              {q0, q1}, [r0], r1
-    vst1.8              {q0, q1}, [r0], r1
-    vst1.8              {q0, q1}, [r0], r1
-    vst1.8              {q0, q1}, [r0], r1
-    vst1.8              {q0, q1}, [r0], r1
-    vst1.8              {q0, q1}, [r0], r1
-    vst1.8              {q0, q1}, [r0], r1
-    subs                r2, r2, #1
-    bgt                 loop_v
-    bx                  lr
-    ENDP                ; |aom_v_predictor_32x32_neon|
-
-;void aom_h_predictor_4x4_neon(uint8_t *dst, ptrdiff_t y_stride,
-;                              const uint8_t *above,
-;                              const uint8_t *left)
-; r0  uint8_t *dst
-; r1  ptrdiff_t y_stride
-; r2  const uint8_t *above
-; r3  const uint8_t *left
-
-|aom_h_predictor_4x4_neon| PROC
-    vld1.32             {d1[0]}, [r3]
-    vdup.8              d0, d1[0]
-    vst1.32             {d0[0]}, [r0], r1
-    vdup.8              d0, d1[1]
-    vst1.32             {d0[0]}, [r0], r1
-    vdup.8              d0, d1[2]
-    vst1.32             {d0[0]}, [r0], r1
-    vdup.8              d0, d1[3]
-    vst1.32             {d0[0]}, [r0], r1
-    bx                  lr
-    ENDP                ; |aom_h_predictor_4x4_neon|
-
-;void aom_h_predictor_8x8_neon(uint8_t *dst, ptrdiff_t y_stride,
-;                              const uint8_t *above,
-;                              const uint8_t *left)
-; r0  uint8_t *dst
-; r1  ptrdiff_t y_stride
-; r2  const uint8_t *above
-; r3  const uint8_t *left
-
-|aom_h_predictor_8x8_neon| PROC
-    vld1.64             {d1}, [r3]
-    vdup.8              d0, d1[0]
-    vst1.64             {d0}, [r0], r1
-    vdup.8              d0, d1[1]
-    vst1.64             {d0}, [r0], r1
-    vdup.8              d0, d1[2]
-    vst1.64             {d0}, [r0], r1
-    vdup.8              d0, d1[3]
-    vst1.64             {d0}, [r0], r1
-    vdup.8              d0, d1[4]
-    vst1.64             {d0}, [r0], r1
-    vdup.8              d0, d1[5]
-    vst1.64             {d0}, [r0], r1
-    vdup.8              d0, d1[6]
-    vst1.64             {d0}, [r0], r1
-    vdup.8              d0, d1[7]
-    vst1.64             {d0}, [r0], r1
-    bx                  lr
-    ENDP                ; |aom_h_predictor_8x8_neon|
-
-;void aom_h_predictor_16x16_neon(uint8_t *dst, ptrdiff_t y_stride,
-;                                const uint8_t *above,
-;                                const uint8_t *left)
-; r0  uint8_t *dst
-; r1  ptrdiff_t y_stride
-; r2  const uint8_t *above
-; r3  const uint8_t *left
-
-|aom_h_predictor_16x16_neon| PROC
-    vld1.8              {q1}, [r3]
-    vdup.8              q0, d2[0]
-    vst1.8              {q0}, [r0], r1
-    vdup.8              q0, d2[1]
-    vst1.8              {q0}, [r0], r1
-    vdup.8              q0, d2[2]
-    vst1.8              {q0}, [r0], r1
-    vdup.8              q0, d2[3]
-    vst1.8              {q0}, [r0], r1
-    vdup.8              q0, d2[4]
-    vst1.8              {q0}, [r0], r1
-    vdup.8              q0, d2[5]
-    vst1.8              {q0}, [r0], r1
-    vdup.8              q0, d2[6]
-    vst1.8              {q0}, [r0], r1
-    vdup.8              q0, d2[7]
-    vst1.8              {q0}, [r0], r1
-    vdup.8              q0, d3[0]
-    vst1.8              {q0}, [r0], r1
-    vdup.8              q0, d3[1]
-    vst1.8              {q0}, [r0], r1
-    vdup.8              q0, d3[2]
-    vst1.8              {q0}, [r0], r1
-    vdup.8              q0, d3[3]
-    vst1.8              {q0}, [r0], r1
-    vdup.8              q0, d3[4]
-    vst1.8              {q0}, [r0], r1
-    vdup.8              q0, d3[5]
-    vst1.8              {q0}, [r0], r1
-    vdup.8              q0, d3[6]
-    vst1.8              {q0}, [r0], r1
-    vdup.8              q0, d3[7]
-    vst1.8              {q0}, [r0], r1
-    bx                  lr
-    ENDP                ; |aom_h_predictor_16x16_neon|
-
-;void aom_h_predictor_32x32_neon(uint8_t *dst, ptrdiff_t y_stride,
-;                                const uint8_t *above,
-;                                const uint8_t *left)
-; r0  uint8_t *dst
-; r1  ptrdiff_t y_stride
-; r2  const uint8_t *above
-; r3  const uint8_t *left
-
-|aom_h_predictor_32x32_neon| PROC
-    sub                 r1, r1, #16
-    mov                 r2, #2
-loop_h
-    vld1.8              {q1}, [r3]!
-    vdup.8              q0, d2[0]
-    vst1.8              {q0}, [r0]!
-    vst1.8              {q0}, [r0], r1
-    vdup.8              q0, d2[1]
-    vst1.8              {q0}, [r0]!
-    vst1.8              {q0}, [r0], r1
-    vdup.8              q0, d2[2]
-    vst1.8              {q0}, [r0]!
-    vst1.8              {q0}, [r0], r1
-    vdup.8              q0, d2[3]
-    vst1.8              {q0}, [r0]!
-    vst1.8              {q0}, [r0], r1
-    vdup.8              q0, d2[4]
-    vst1.8              {q0}, [r0]!
-    vst1.8              {q0}, [r0], r1
-    vdup.8              q0, d2[5]
-    vst1.8              {q0}, [r0]!
-    vst1.8              {q0}, [r0], r1
-    vdup.8              q0, d2[6]
-    vst1.8              {q0}, [r0]!
-    vst1.8              {q0}, [r0], r1
-    vdup.8              q0, d2[7]
-    vst1.8              {q0}, [r0]!
-    vst1.8              {q0}, [r0], r1
-    vdup.8              q0, d3[0]
-    vst1.8              {q0}, [r0]!
-    vst1.8              {q0}, [r0], r1
-    vdup.8              q0, d3[1]
-    vst1.8              {q0}, [r0]!
-    vst1.8              {q0}, [r0], r1
-    vdup.8              q0, d3[2]
-    vst1.8              {q0}, [r0]!
-    vst1.8              {q0}, [r0], r1
-    vdup.8              q0, d3[3]
-    vst1.8              {q0}, [r0]!
-    vst1.8              {q0}, [r0], r1
-    vdup.8              q0, d3[4]
-    vst1.8              {q0}, [r0]!
-    vst1.8              {q0}, [r0], r1
-    vdup.8              q0, d3[5]
-    vst1.8              {q0}, [r0]!
-    vst1.8              {q0}, [r0], r1
-    vdup.8              q0, d3[6]
-    vst1.8              {q0}, [r0]!
-    vst1.8              {q0}, [r0], r1
-    vdup.8              q0, d3[7]
-    vst1.8              {q0}, [r0]!
-    vst1.8              {q0}, [r0], r1
-    subs                r2, r2, #1
-    bgt                 loop_h
-    bx                  lr
-    ENDP                ; |aom_h_predictor_32x32_neon|
-
-;void aom_tm_predictor_4x4_neon (uint8_t *dst, ptrdiff_t y_stride,
-;                                const uint8_t *above,
-;                                const uint8_t *left)
-; r0  uint8_t *dst
-; r1  ptrdiff_t y_stride
-; r2  const uint8_t *above
-; r3  const uint8_t *left
-
-|aom_tm_predictor_4x4_neon| PROC
-    ; Load ytop_left = above[-1];
-    sub                 r12, r2, #1
-    vld1.u8             {d0[]}, [r12]
-
-    ; Load above 4 pixels
-    vld1.32             {d2[0]}, [r2]
-
-    ; Compute above - ytop_left
-    vsubl.u8            q3, d2, d0
-
-    ; Load left row by row and compute left + (above - ytop_left)
-    ; 1st row and 2nd row
-    vld1.u8             {d2[]}, [r3]!
-    vld1.u8             {d4[]}, [r3]!
-    vmovl.u8            q1, d2
-    vmovl.u8            q2, d4
-    vadd.s16            q1, q1, q3
-    vadd.s16            q2, q2, q3
-    vqmovun.s16         d0, q1
-    vqmovun.s16         d1, q2
-    vst1.32             {d0[0]}, [r0], r1
-    vst1.32             {d1[0]}, [r0], r1
-
-    ; 3rd row and 4th row
-    vld1.u8             {d2[]}, [r3]!
-    vld1.u8             {d4[]}, [r3]
-    vmovl.u8            q1, d2
-    vmovl.u8            q2, d4
-    vadd.s16            q1, q1, q3
-    vadd.s16            q2, q2, q3
-    vqmovun.s16         d0, q1
-    vqmovun.s16         d1, q2
-    vst1.32             {d0[0]}, [r0], r1
-    vst1.32             {d1[0]}, [r0], r1
-    bx                  lr
-    ENDP                ; |aom_tm_predictor_4x4_neon|
-
-;void aom_tm_predictor_8x8_neon (uint8_t *dst, ptrdiff_t y_stride,
-;                                const uint8_t *above,
-;                                const uint8_t *left)
-; r0  uint8_t *dst
-; r1  ptrdiff_t y_stride
-; r2  const uint8_t *above
-; r3  const uint8_t *left
-
-|aom_tm_predictor_8x8_neon| PROC
-    ; Load ytop_left = above[-1];
-    sub                 r12, r2, #1
-    vld1.8              {d0[]}, [r12]
-
-    ; preload 8 left
-    vld1.8              {d30}, [r3]
-
-    ; Load above 8 pixels
-    vld1.64             {d2}, [r2]
-
-    vmovl.u8            q10, d30
-
-    ; Compute above - ytop_left
-    vsubl.u8            q3, d2, d0
-
-    ; Load left row by row and compute left + (above - ytop_left)
-    ; 1st row and 2nd row
-    vdup.16             q0, d20[0]
-    vdup.16             q1, d20[1]
-    vadd.s16            q0, q3, q0
-    vadd.s16            q1, q3, q1
-
-    ; 3rd row and 4th row
-    vdup.16             q8, d20[2]
-    vdup.16             q9, d20[3]
-    vadd.s16            q8, q3, q8
-    vadd.s16            q9, q3, q9
-
-    vqmovun.s16         d0, q0
-    vqmovun.s16         d1, q1
-    vqmovun.s16         d2, q8
-    vqmovun.s16         d3, q9
-
-    vst1.64             {d0}, [r0], r1
-    vst1.64             {d1}, [r0], r1
-    vst1.64             {d2}, [r0], r1
-    vst1.64             {d3}, [r0], r1
-
-    ; 5th row and 6th row
-    vdup.16             q0, d21[0]
-    vdup.16             q1, d21[1]
-    vadd.s16            q0, q3, q0
-    vadd.s16            q1, q3, q1
-
-    ; 7th row and 8th row
-    vdup.16             q8, d21[2]
-    vdup.16             q9, d21[3]
-    vadd.s16            q8, q3, q8
-    vadd.s16            q9, q3, q9
-
-    vqmovun.s16         d0, q0
-    vqmovun.s16         d1, q1
-    vqmovun.s16         d2, q8
-    vqmovun.s16         d3, q9
-
-    vst1.64             {d0}, [r0], r1
-    vst1.64             {d1}, [r0], r1
-    vst1.64             {d2}, [r0], r1
-    vst1.64             {d3}, [r0], r1
-
-    bx                  lr
-    ENDP                ; |aom_tm_predictor_8x8_neon|
-
-;void aom_tm_predictor_16x16_neon (uint8_t *dst, ptrdiff_t y_stride,
-;                                const uint8_t *above,
-;                                const uint8_t *left)
-; r0  uint8_t *dst
-; r1  ptrdiff_t y_stride
-; r2  const uint8_t *above
-; r3  const uint8_t *left
-
-|aom_tm_predictor_16x16_neon| PROC
-    ; Load ytop_left = above[-1];
-    sub                 r12, r2, #1
-    vld1.8              {d0[]}, [r12]
-
-    ; Load above 8 pixels
-    vld1.8              {q1}, [r2]
-
-    ; preload 8 left into r12
-    vld1.8              {d18}, [r3]!
-
-    ; Compute above - ytop_left
-    vsubl.u8            q2, d2, d0
-    vsubl.u8            q3, d3, d0
-
-    vmovl.u8            q10, d18
-
-    ; Load left row by row and compute left + (above - ytop_left)
-    ; Process 8 rows in each single loop and loop 2 times to process 16 rows.
-    mov                 r2, #2
-
-loop_16x16_neon
-    ; Process two rows.
-    vdup.16             q0, d20[0]
-    vdup.16             q8, d20[1]
-    vadd.s16            q1, q0, q2
-    vadd.s16            q0, q0, q3
-    vadd.s16            q11, q8, q2
-    vadd.s16            q8, q8, q3
-    vqmovun.s16         d2, q1
-    vqmovun.s16         d3, q0
-    vqmovun.s16         d22, q11
-    vqmovun.s16         d23, q8
-    vdup.16             q0, d20[2]                  ; proload next 2 rows data
-    vdup.16             q8, d20[3]
-    vst1.64             {d2,d3}, [r0], r1
-    vst1.64             {d22,d23}, [r0], r1
-
-    ; Process two rows.
-    vadd.s16            q1, q0, q2
-    vadd.s16            q0, q0, q3
-    vadd.s16            q11, q8, q2
-    vadd.s16            q8, q8, q3
-    vqmovun.s16         d2, q1
-    vqmovun.s16         d3, q0
-    vqmovun.s16         d22, q11
-    vqmovun.s16         d23, q8
-    vdup.16             q0, d21[0]                  ; proload next 2 rows data
-    vdup.16             q8, d21[1]
-    vst1.64             {d2,d3}, [r0], r1
-    vst1.64             {d22,d23}, [r0], r1
-
-    vadd.s16            q1, q0, q2
-    vadd.s16            q0, q0, q3
-    vadd.s16            q11, q8, q2
-    vadd.s16            q8, q8, q3
-    vqmovun.s16         d2, q1
-    vqmovun.s16         d3, q0
-    vqmovun.s16         d22, q11
-    vqmovun.s16         d23, q8
-    vdup.16             q0, d21[2]                  ; proload next 2 rows data
-    vdup.16             q8, d21[3]
-    vst1.64             {d2,d3}, [r0], r1
-    vst1.64             {d22,d23}, [r0], r1
-
-
-    vadd.s16            q1, q0, q2
-    vadd.s16            q0, q0, q3
-    vadd.s16            q11, q8, q2
-    vadd.s16            q8, q8, q3
-    vqmovun.s16         d2, q1
-    vqmovun.s16         d3, q0
-    vqmovun.s16         d22, q11
-    vqmovun.s16         d23, q8
-    vld1.8              {d18}, [r3]!                  ; preload 8 left into r12
-    vmovl.u8            q10, d18
-    vst1.64             {d2,d3}, [r0], r1
-    vst1.64             {d22,d23}, [r0], r1
-
-    subs                r2, r2, #1
-    bgt                 loop_16x16_neon
-
-    bx                  lr
-    ENDP                ; |aom_tm_predictor_16x16_neon|
-
-;void aom_tm_predictor_32x32_neon (uint8_t *dst, ptrdiff_t y_stride,
-;                                  const uint8_t *above,
-;                                  const uint8_t *left)
-; r0  uint8_t *dst
-; r1  ptrdiff_t y_stride
-; r2  const uint8_t *above
-; r3  const uint8_t *left
-
-|aom_tm_predictor_32x32_neon| PROC
-    ; Load ytop_left = above[-1];
-    sub                 r12, r2, #1
-    vld1.8              {d0[]}, [r12]
-
-    ; Load above 32 pixels
-    vld1.8              {q1}, [r2]!
-    vld1.8              {q2}, [r2]
-
-    ; preload 8 left pixels
-    vld1.8              {d26}, [r3]!
-
-    ; Compute above - ytop_left
-    vsubl.u8            q8, d2, d0
-    vsubl.u8            q9, d3, d0
-    vsubl.u8            q10, d4, d0
-    vsubl.u8            q11, d5, d0
-
-    vmovl.u8            q3, d26
-
-    ; Load left row by row and compute left + (above - ytop_left)
-    ; Process 8 rows in each single loop and loop 4 times to process 32 rows.
-    mov                 r2, #4
-
-loop_32x32_neon
-    ; Process two rows.
-    vdup.16             q0, d6[0]
-    vdup.16             q2, d6[1]
-    vadd.s16            q12, q0, q8
-    vadd.s16            q13, q0, q9
-    vadd.s16            q14, q0, q10
-    vadd.s16            q15, q0, q11
-    vqmovun.s16         d0, q12
-    vqmovun.s16         d1, q13
-    vadd.s16            q12, q2, q8
-    vadd.s16            q13, q2, q9
-    vqmovun.s16         d2, q14
-    vqmovun.s16         d3, q15
-    vadd.s16            q14, q2, q10
-    vadd.s16            q15, q2, q11
-    vst1.64             {d0-d3}, [r0], r1
-    vqmovun.s16         d24, q12
-    vqmovun.s16         d25, q13
-    vqmovun.s16         d26, q14
-    vqmovun.s16         d27, q15
-    vdup.16             q1, d6[2]
-    vdup.16             q2, d6[3]
-    vst1.64             {d24-d27}, [r0], r1
-
-    ; Process two rows.
-    vadd.s16            q12, q1, q8
-    vadd.s16            q13, q1, q9
-    vadd.s16            q14, q1, q10
-    vadd.s16            q15, q1, q11
-    vqmovun.s16         d0, q12
-    vqmovun.s16         d1, q13
-    vadd.s16            q12, q2, q8
-    vadd.s16            q13, q2, q9
-    vqmovun.s16         d2, q14
-    vqmovun.s16         d3, q15
-    vadd.s16            q14, q2, q10
-    vadd.s16            q15, q2, q11
-    vst1.64             {d0-d3}, [r0], r1
-    vqmovun.s16         d24, q12
-    vqmovun.s16         d25, q13
-    vqmovun.s16         d26, q14
-    vqmovun.s16         d27, q15
-    vdup.16             q0, d7[0]
-    vdup.16             q2, d7[1]
-    vst1.64             {d24-d27}, [r0], r1
-
-    ; Process two rows.
-    vadd.s16            q12, q0, q8
-    vadd.s16            q13, q0, q9
-    vadd.s16            q14, q0, q10
-    vadd.s16            q15, q0, q11
-    vqmovun.s16         d0, q12
-    vqmovun.s16         d1, q13
-    vadd.s16            q12, q2, q8
-    vadd.s16            q13, q2, q9
-    vqmovun.s16         d2, q14
-    vqmovun.s16         d3, q15
-    vadd.s16            q14, q2, q10
-    vadd.s16            q15, q2, q11
-    vst1.64             {d0-d3}, [r0], r1
-    vqmovun.s16         d24, q12
-    vqmovun.s16         d25, q13
-    vqmovun.s16         d26, q14
-    vqmovun.s16         d27, q15
-    vdup.16             q0, d7[2]
-    vdup.16             q2, d7[3]
-    vst1.64             {d24-d27}, [r0], r1
-
-    ; Process two rows.
-    vadd.s16            q12, q0, q8
-    vadd.s16            q13, q0, q9
-    vadd.s16            q14, q0, q10
-    vadd.s16            q15, q0, q11
-    vqmovun.s16         d0, q12
-    vqmovun.s16         d1, q13
-    vadd.s16            q12, q2, q8
-    vadd.s16            q13, q2, q9
-    vqmovun.s16         d2, q14
-    vqmovun.s16         d3, q15
-    vadd.s16            q14, q2, q10
-    vadd.s16            q15, q2, q11
-    vst1.64             {d0-d3}, [r0], r1
-    vqmovun.s16         d24, q12
-    vqmovun.s16         d25, q13
-    vld1.8              {d0}, [r3]!                   ; preload 8 left pixels
-    vqmovun.s16         d26, q14
-    vqmovun.s16         d27, q15
-    vmovl.u8            q3, d0
-    vst1.64             {d24-d27}, [r0], r1
-
-    subs                r2, r2, #1
-    bgt                 loop_32x32_neon
-
-    bx                  lr
-    ENDP                ; |aom_tm_predictor_32x32_neon|
-
-    END

aom_dsp/arm/loopfilter_16_neon.c

@@ -1,174 +0,0 @@
-/*
- * Copyright (c) 2016, Alliance for Open Media. All rights reserved
- *
- * This source code is subject to the terms of the BSD 2 Clause License and
- * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
- * was not distributed with this source code in the LICENSE file, you can
- * obtain it at www.aomedia.org/license/software. If the Alliance for Open
- * Media Patent License 1.0 was not distributed with this source code in the
- * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
- */
-
-#include <arm_neon.h>
-
-#include "./aom_dsp_rtcd.h"
-#include "./aom_config.h"
-#include "aom/aom_integer.h"
-
-static INLINE void loop_filter_neon_16(uint8x16_t qblimit,  // blimit
-                                       uint8x16_t qlimit,   // limit
-                                       uint8x16_t qthresh,  // thresh
-                                       uint8x16_t q3,       // p3
-                                       uint8x16_t q4,       // p2
-                                       uint8x16_t q5,       // p1
-                                       uint8x16_t q6,       // p0
-                                       uint8x16_t q7,       // q0
-                                       uint8x16_t q8,       // q1
-                                       uint8x16_t q9,       // q2
-                                       uint8x16_t q10,      // q3
-                                       uint8x16_t *q5r,     // p1
-                                       uint8x16_t *q6r,     // p0
-                                       uint8x16_t *q7r,     // q0
-                                       uint8x16_t *q8r) {   // q1
-  uint8x16_t q1u8, q2u8, q11u8, q12u8, q13u8, q14u8, q15u8;
-  int16x8_t q2s16, q11s16;
-  uint16x8_t q4u16;
-  int8x16_t q0s8, q1s8, q2s8, q11s8, q12s8, q13s8;
-  int8x8_t d2s8, d3s8;
-
-  q11u8 = vabdq_u8(q3, q4);
-  q12u8 = vabdq_u8(q4, q5);
-  q13u8 = vabdq_u8(q5, q6);
-  q14u8 = vabdq_u8(q8, q7);
-  q3 = vabdq_u8(q9, q8);
-  q4 = vabdq_u8(q10, q9);
-
-  q11u8 = vmaxq_u8(q11u8, q12u8);
-  q12u8 = vmaxq_u8(q13u8, q14u8);
-  q3 = vmaxq_u8(q3, q4);
-  q15u8 = vmaxq_u8(q11u8, q12u8);
-
-  q9 = vabdq_u8(q6, q7);
-
-  // aom_hevmask
-  q13u8 = vcgtq_u8(q13u8, qthresh);
-  q14u8 = vcgtq_u8(q14u8, qthresh);
-  q15u8 = vmaxq_u8(q15u8, q3);
-
-  q2u8 = vabdq_u8(q5, q8);
-  q9 = vqaddq_u8(q9, q9);
-
-  q15u8 = vcgeq_u8(qlimit, q15u8);
-
-  // aom_filter() function
-  // convert to signed
-  q10 = vdupq_n_u8(0x80);
-  q8 = veorq_u8(q8, q10);
-  q7 = veorq_u8(q7, q10);
-  q6 = veorq_u8(q6, q10);
-  q5 = veorq_u8(q5, q10);
-
-  q2u8 = vshrq_n_u8(q2u8, 1);
-  q9 = vqaddq_u8(q9, q2u8);
-
-  q2s16 = vsubl_s8(vget_low_s8(vreinterpretq_s8_u8(q7)),
-                   vget_low_s8(vreinterpretq_s8_u8(q6)));
-  q11s16 = vsubl_s8(vget_high_s8(vreinterpretq_s8_u8(q7)),
-                    vget_high_s8(vreinterpretq_s8_u8(q6)));
-
-  q9 = vcgeq_u8(qblimit, q9);
-
-  q1s8 = vqsubq_s8(vreinterpretq_s8_u8(q5), vreinterpretq_s8_u8(q8));
-
-  q14u8 = vorrq_u8(q13u8, q14u8);
-
-  q4u16 = vdupq_n_u16(3);
-  q2s16 = vmulq_s16(q2s16, vreinterpretq_s16_u16(q4u16));
-  q11s16 = vmulq_s16(q11s16, vreinterpretq_s16_u16(q4u16));
-
-  q1u8 = vandq_u8(vreinterpretq_u8_s8(q1s8), q14u8);
-  q15u8 = vandq_u8(q15u8, q9);
-
-  q1s8 = vreinterpretq_s8_u8(q1u8);
-  q2s16 = vaddw_s8(q2s16, vget_low_s8(q1s8));
-  q11s16 = vaddw_s8(q11s16, vget_high_s8(q1s8));
-
-  q4 = vdupq_n_u8(3);
-  q9 = vdupq_n_u8(4);
-  // aom_filter = clamp(aom_filter + 3 * ( qs0 - ps0))
-  d2s8 = vqmovn_s16(q2s16);
-  d3s8 = vqmovn_s16(q11s16);
-  q1s8 = vcombine_s8(d2s8, d3s8);
-  q1u8 = vandq_u8(vreinterpretq_u8_s8(q1s8), q15u8);
-  q1s8 = vreinterpretq_s8_u8(q1u8);
-
-  q2s8 = vqaddq_s8(q1s8, vreinterpretq_s8_u8(q4));
-  q1s8 = vqaddq_s8(q1s8, vreinterpretq_s8_u8(q9));
-  q2s8 = vshrq_n_s8(q2s8, 3);
-  q1s8 = vshrq_n_s8(q1s8, 3);
-
-  q11s8 = vqaddq_s8(vreinterpretq_s8_u8(q6), q2s8);
-  q0s8 = vqsubq_s8(vreinterpretq_s8_u8(q7), q1s8);
-
-  q1s8 = vrshrq_n_s8(q1s8, 1);
-  q1s8 = vbicq_s8(q1s8, vreinterpretq_s8_u8(q14u8));
-
-  q13s8 = vqaddq_s8(vreinterpretq_s8_u8(q5), q1s8);
-  q12s8 = vqsubq_s8(vreinterpretq_s8_u8(q8), q1s8);
-
-  *q8r = veorq_u8(vreinterpretq_u8_s8(q12s8), q10);
-  *q7r = veorq_u8(vreinterpretq_u8_s8(q0s8), q10);
-  *q6r = veorq_u8(vreinterpretq_u8_s8(q11s8), q10);
-  *q5r = veorq_u8(vreinterpretq_u8_s8(q13s8), q10);
-  return;
-}
-
-void aom_lpf_horizontal_4_dual_neon(
-    uint8_t *s, int p /* pitch */, const uint8_t *blimit0,
-    const uint8_t *limit0, const uint8_t *thresh0, const uint8_t *blimit1,
-    const uint8_t *limit1, const uint8_t *thresh1) {
-  uint8x8_t dblimit0, dlimit0, dthresh0, dblimit1, dlimit1, dthresh1;
-  uint8x16_t qblimit, qlimit, qthresh;
-  uint8x16_t q3u8, q4u8, q5u8, q6u8, q7u8, q8u8, q9u8, q10u8;
-
-  dblimit0 = vld1_u8(blimit0);
-  dlimit0 = vld1_u8(limit0);
-  dthresh0 = vld1_u8(thresh0);
-  dblimit1 = vld1_u8(blimit1);
-  dlimit1 = vld1_u8(limit1);
-  dthresh1 = vld1_u8(thresh1);
-  qblimit = vcombine_u8(dblimit0, dblimit1);
-  qlimit = vcombine_u8(dlimit0, dlimit1);
-  qthresh = vcombine_u8(dthresh0, dthresh1);
-
-  s -= (p << 2);
-
-  q3u8 = vld1q_u8(s);
-  s += p;
-  q4u8 = vld1q_u8(s);
-  s += p;
-  q5u8 = vld1q_u8(s);
-  s += p;
-  q6u8 = vld1q_u8(s);
-  s += p;
-  q7u8 = vld1q_u8(s);
-  s += p;
-  q8u8 = vld1q_u8(s);
-  s += p;
-  q9u8 = vld1q_u8(s);
-  s += p;
-  q10u8 = vld1q_u8(s);
-
-  loop_filter_neon_16(qblimit, qlimit, qthresh, q3u8, q4u8, q5u8, q6u8, q7u8,
-                      q8u8, q9u8, q10u8, &q5u8, &q6u8, &q7u8, &q8u8);
-
-  s -= (p * 5);
-  vst1q_u8(s, q5u8);
-  s += p;
-  vst1q_u8(s, q6u8);
-  s += p;
-  vst1q_u8(s, q7u8);
-  s += p;
-  vst1q_u8(s, q8u8);
-  return;
-}

aom_dsp/arm/loopfilter_4_neon.asm

@@ -1,252 +0,0 @@
-;
-; Copyright (c) 2016, Alliance for Open Media. All rights reserved
-;
-; This source code is subject to the terms of the BSD 2 Clause License and
-; the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
-; was not distributed with this source code in the LICENSE file, you can
-; obtain it at www.aomedia.org/license/software. If the Alliance for Open
-; Media Patent License 1.0 was not distributed with this source code in the
-; PATENTS file, you can obtain it at www.aomedia.org/license/patent.
-;
-
-;
-
-    EXPORT  |aom_lpf_horizontal_4_neon|
-    EXPORT  |aom_lpf_vertical_4_neon|
-    ARM
-
-    AREA ||.text||, CODE, READONLY, ALIGN=2
-
-; Currently aom only works on iterations 8 at a time. The aom loop filter
-; works on 16 iterations at a time.
-;
-; void aom_lpf_horizontal_4_neon(uint8_t *s,
-;                                int p /* pitch */,
-;                                const uint8_t *blimit,
-;                                const uint8_t *limit,
-;                                const uint8_t *thresh)
-;
-; r0    uint8_t *s,
-; r1    int p, /* pitch */
-; r2    const uint8_t *blimit,
-; r3    const uint8_t *limit,
-; sp    const uint8_t *thresh,
-|aom_lpf_horizontal_4_neon| PROC
-    push        {lr}
-
-    vld1.8      {d0[]}, [r2]               ; duplicate *blimit
-    ldr         r2, [sp, #4]               ; load thresh
-    add         r1, r1, r1                 ; double pitch
-
-    vld1.8      {d1[]}, [r3]               ; duplicate *limit
-    vld1.8      {d2[]}, [r2]               ; duplicate *thresh
-
-    sub         r2, r0, r1, lsl #1         ; move src pointer down by 4 lines
-    add         r3, r2, r1, lsr #1         ; set to 3 lines down
-
-    vld1.u8     {d3}, [r2@64], r1          ; p3
-    vld1.u8     {d4}, [r3@64], r1          ; p2
-    vld1.u8     {d5}, [r2@64], r1          ; p1
-    vld1.u8     {d6}, [r3@64], r1          ; p0
-    vld1.u8     {d7}, [r2@64], r1          ; q0
-    vld1.u8     {d16}, [r3@64], r1         ; q1
-    vld1.u8     {d17}, [r2@64]             ; q2
-    vld1.u8     {d18}, [r3@64]             ; q3
-
-    sub         r2, r2, r1, lsl #1
-    sub         r3, r3, r1, lsl #1
-
-    bl          aom_loop_filter_neon
-
-    vst1.u8     {d4}, [r2@64], r1          ; store op1
-    vst1.u8     {d5}, [r3@64], r1          ; store op0
-    vst1.u8     {d6}, [r2@64], r1          ; store oq0
-    vst1.u8     {d7}, [r3@64], r1          ; store oq1
-
-    pop         {pc}
-    ENDP        ; |aom_lpf_horizontal_4_neon|
-
-; Currently aom only works on iterations 8 at a time. The aom loop filter
-; works on 16 iterations at a time.
-;
-; void aom_lpf_vertical_4_neon(uint8_t *s,
-;                              int p /* pitch */,
-;                              const uint8_t *blimit,
-;                              const uint8_t *limit,
-;                              const uint8_t *thresh)
-;
-; r0    uint8_t *s,
-; r1    int p, /* pitch */
-; r2    const uint8_t *blimit,
-; r3    const uint8_t *limit,
-; sp    const uint8_t *thresh,
-|aom_lpf_vertical_4_neon| PROC
-    push        {lr}
-
-    vld1.8      {d0[]}, [r2]              ; duplicate *blimit
-    vld1.8      {d1[]}, [r3]              ; duplicate *limit
-
-    ldr         r3, [sp, #4]              ; load thresh
-    sub         r2, r0, #4                ; move s pointer down by 4 columns
-
-    vld1.8      {d2[]}, [r3]              ; duplicate *thresh
-
-    vld1.u8     {d3}, [r2], r1             ; load s data
-    vld1.u8     {d4}, [r2], r1
-    vld1.u8     {d5}, [r2], r1
-    vld1.u8     {d6}, [r2], r1
-    vld1.u8     {d7}, [r2], r1
-    vld1.u8     {d16}, [r2], r1
-    vld1.u8     {d17}, [r2], r1
-    vld1.u8     {d18}, [r2]
-
-    ;transpose to 8x16 matrix
-    vtrn.32     d3, d7
-    vtrn.32     d4, d16
-    vtrn.32     d5, d17
-    vtrn.32     d6, d18
-
-    vtrn.16     d3, d5
-    vtrn.16     d4, d6
-    vtrn.16     d7, d17
-    vtrn.16     d16, d18
-
-    vtrn.8      d3, d4
-    vtrn.8      d5, d6
-    vtrn.8      d7, d16
-    vtrn.8      d17, d18
-
-    bl          aom_loop_filter_neon
-
-    sub         r0, r0, #2
-
-    ;store op1, op0, oq0, oq1
-    vst4.8      {d4[0], d5[0], d6[0], d7[0]}, [r0], r1
-    vst4.8      {d4[1], d5[1], d6[1], d7[1]}, [r0], r1
-    vst4.8      {d4[2], d5[2], d6[2], d7[2]}, [r0], r1
-    vst4.8      {d4[3], d5[3], d6[3], d7[3]}, [r0], r1
-    vst4.8      {d4[4], d5[4], d6[4], d7[4]}, [r0], r1
-    vst4.8      {d4[5], d5[5], d6[5], d7[5]}, [r0], r1
-    vst4.8      {d4[6], d5[6], d6[6], d7[6]}, [r0], r1
-    vst4.8      {d4[7], d5[7], d6[7], d7[7]}, [r0]
-
-    pop         {pc}
-    ENDP        ; |aom_lpf_vertical_4_neon|
-
-; void aom_loop_filter_neon();
-; This is a helper function for the loopfilters. The invidual functions do the
-; necessary load, transpose (if necessary) and store. The function does not use
-; registers d8-d15.
-;
-; Inputs:
-; r0-r3, r12 PRESERVE
-; d0    blimit
-; d1    limit
-; d2    thresh
-; d3    p3
-; d4    p2
-; d5    p1
-; d6    p0
-; d7    q0
-; d16   q1
-; d17   q2
-; d18   q3
-;
-; Outputs:
-; d4    op1
-; d5    op0
-; d6    oq0
-; d7    oq1
-|aom_loop_filter_neon| PROC
-    ; filter_mask
-    vabd.u8     d19, d3, d4                 ; m1 = abs(p3 - p2)
-    vabd.u8     d20, d4, d5                 ; m2 = abs(p2 - p1)
-    vabd.u8     d21, d5, d6                 ; m3 = abs(p1 - p0)
-    vabd.u8     d22, d16, d7                ; m4 = abs(q1 - q0)
-    vabd.u8     d3, d17, d16                ; m5 = abs(q2 - q1)
-    vabd.u8     d4, d18, d17                ; m6 = abs(q3 - q2)
-
-    ; only compare the largest value to limit
-    vmax.u8     d19, d19, d20               ; m1 = max(m1, m2)
-    vmax.u8     d20, d21, d22               ; m2 = max(m3, m4)
-
-    vabd.u8     d17, d6, d7                 ; abs(p0 - q0)
-
-    vmax.u8     d3, d3, d4                  ; m3 = max(m5, m6)
-
-    vmov.u8     d18, #0x80
-
-    vmax.u8     d23, d19, d20               ; m1 = max(m1, m2)
-
-    ; hevmask
-    vcgt.u8     d21, d21, d2                ; (abs(p1 - p0) > thresh)*-1
-    vcgt.u8     d22, d22, d2                ; (abs(q1 - q0) > thresh)*-1
-    vmax.u8     d23, d23, d3                ; m1 = max(m1, m3)
-
-    vabd.u8     d28, d5, d16                ; a = abs(p1 - q1)
-    vqadd.u8    d17, d17, d17               ; b = abs(p0 - q0) * 2
-
-    veor        d7, d7, d18                 ; qs0
-
-    vcge.u8     d23, d1, d23                ; abs(m1) > limit
-
-    ; filter() function
-    ; convert to signed
-
-    vshr.u8     d28, d28, #1                ; a = a / 2
-    veor        d6, d6, d18                 ; ps0
-
-    veor        d5, d5, d18                 ; ps1
-    vqadd.u8    d17, d17, d28               ; a = b + a
-
-    veor        d16, d16, d18               ; qs1
-
-    vmov.u8     d19, #3
-
-    vsub.s8     d28, d7, d6                 ; ( qs0 - ps0)
-
-    vcge.u8     d17, d0, d17                ; a > blimit
-
-    vqsub.s8    d27, d5, d16                ; filter = clamp(ps1-qs1)
-    vorr        d22, d21, d22               ; hevmask
-
-    vmull.s8    q12, d28, d19               ; 3 * ( qs0 - ps0)
-
-    vand        d27, d27, d22               ; filter &= hev
-    vand        d23, d23, d17               ; filter_mask
-
-    vaddw.s8    q12, q12, d27               ; filter + 3 * (qs0 - ps0)
-
-    vmov.u8     d17, #4
-
-    ; filter = clamp(filter + 3 * ( qs0 - ps0))
-    vqmovn.s16  d27, q12
-
-    vand        d27, d27, d23               ; filter &= mask
-
-    vqadd.s8    d28, d27, d19               ; filter2 = clamp(filter+3)
-    vqadd.s8    d27, d27, d17               ; filter1 = clamp(filter+4)
-    vshr.s8     d28, d28, #3                ; filter2 >>= 3
-    vshr.s8     d27, d27, #3                ; filter1 >>= 3
-
-    vqadd.s8    d19, d6, d28                ; u = clamp(ps0 + filter2)
-    vqsub.s8    d26, d7, d27                ; u = clamp(qs0 - filter1)
-
-    ; outer tap adjustments
-    vrshr.s8    d27, d27, #1                ; filter = ++filter1 >> 1
-
-    veor        d6, d26, d18                ; *oq0 = u^0x80
-
-    vbic        d27, d27, d22               ; filter &= ~hev
-
-    vqadd.s8    d21, d5, d27                ; u = clamp(ps1 + filter)
-    vqsub.s8    d20, d16, d27               ; u = clamp(qs1 - filter)
-
-    veor        d5, d19, d18                ; *op0 = u^0x80
-    veor        d4, d21, d18                ; *op1 = u^0x80
-    veor        d7, d20, d18                ; *oq1 = u^0x80
-
-    bx          lr
-    ENDP        ; |aom_loop_filter_neon|
-
-    END

aom_dsp/arm/loopfilter_4_neon.c

@@ -1,250 +0,0 @@
-/*
- * Copyright (c) 2016, Alliance for Open Media. All rights reserved
- *
- * This source code is subject to the terms of the BSD 2 Clause License and
- * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
- * was not distributed with this source code in the LICENSE file, you can
- * obtain it at www.aomedia.org/license/software. If the Alliance for Open
- * Media Patent License 1.0 was not distributed with this source code in the
- * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
- */
-
-#include <arm_neon.h>
-
-#include "./aom_dsp_rtcd.h"
-
-static INLINE void loop_filter_neon(uint8x8_t dblimit,   // flimit
-                                    uint8x8_t dlimit,    // limit
-                                    uint8x8_t dthresh,   // thresh
-                                    uint8x8_t d3u8,      // p3
-                                    uint8x8_t d4u8,      // p2
-                                    uint8x8_t d5u8,      // p1
-                                    uint8x8_t d6u8,      // p0
-                                    uint8x8_t d7u8,      // q0
-                                    uint8x8_t d16u8,     // q1
-                                    uint8x8_t d17u8,     // q2
-                                    uint8x8_t d18u8,     // q3
-                                    uint8x8_t *d4ru8,    // p1
-                                    uint8x8_t *d5ru8,    // p0
-                                    uint8x8_t *d6ru8,    // q0
-                                    uint8x8_t *d7ru8) {  // q1
-  uint8x8_t d19u8, d20u8, d21u8, d22u8, d23u8, d27u8, d28u8;
-  int16x8_t q12s16;
-  int8x8_t d19s8, d20s8, d21s8, d26s8, d27s8, d28s8;
-
-  d19u8 = vabd_u8(d3u8, d4u8);
-  d20u8 = vabd_u8(d4u8, d5u8);
-  d21u8 = vabd_u8(d5u8, d6u8);
-  d22u8 = vabd_u8(d16u8, d7u8);
-  d3u8 = vabd_u8(d17u8, d16u8);
-  d4u8 = vabd_u8(d18u8, d17u8);
-
-  d19u8 = vmax_u8(d19u8, d20u8);
-  d20u8 = vmax_u8(d21u8, d22u8);
-  d3u8 = vmax_u8(d3u8, d4u8);
-  d23u8 = vmax_u8(d19u8, d20u8);
-
-  d17u8 = vabd_u8(d6u8, d7u8);
-
-  d21u8 = vcgt_u8(d21u8, dthresh);
-  d22u8 = vcgt_u8(d22u8, dthresh);
-  d23u8 = vmax_u8(d23u8, d3u8);
-
-  d28u8 = vabd_u8(d5u8, d16u8);
-  d17u8 = vqadd_u8(d17u8, d17u8);
-
-  d23u8 = vcge_u8(dlimit, d23u8);
-
-  d18u8 = vdup_n_u8(0x80);
-  d5u8 = veor_u8(d5u8, d18u8);
-  d6u8 = veor_u8(d6u8, d18u8);
-  d7u8 = veor_u8(d7u8, d18u8);
-  d16u8 = veor_u8(d16u8, d18u8);
-
-  d28u8 = vshr_n_u8(d28u8, 1);
-  d17u8 = vqadd_u8(d17u8, d28u8);
-
-  d19u8 = vdup_n_u8(3);
-
-  d28s8 = vsub_s8(vreinterpret_s8_u8(d7u8), vreinterpret_s8_u8(d6u8));
-
-  d17u8 = vcge_u8(dblimit, d17u8);
-
-  d27s8 = vqsub_s8(vreinterpret_s8_u8(d5u8), vreinterpret_s8_u8(d16u8));
-
-  d22u8 = vorr_u8(d21u8, d22u8);
-
-  q12s16 = vmull_s8(d28s8, vreinterpret_s8_u8(d19u8));
-
-  d27u8 = vand_u8(vreinterpret_u8_s8(d27s8), d22u8);
-  d23u8 = vand_u8(d23u8, d17u8);
-
-  q12s16 = vaddw_s8(q12s16, vreinterpret_s8_u8(d27u8));
-
-  d17u8 = vdup_n_u8(4);
-
-  d27s8 = vqmovn_s16(q12s16);
-  d27u8 = vand_u8(vreinterpret_u8_s8(d27s8), d23u8);
-  d27s8 = vreinterpret_s8_u8(d27u8);
-
-  d28s8 = vqadd_s8(d27s8, vreinterpret_s8_u8(d19u8));
-  d27s8 = vqadd_s8(d27s8, vreinterpret_s8_u8(d17u8));
-  d28s8 = vshr_n_s8(d28s8, 3);
-  d27s8 = vshr_n_s8(d27s8, 3);
-
-  d19s8 = vqadd_s8(vreinterpret_s8_u8(d6u8), d28s8);
-  d26s8 = vqsub_s8(vreinterpret_s8_u8(d7u8), d27s8);
-
-  d27s8 = vrshr_n_s8(d27s8, 1);
-  d27s8 = vbic_s8(d27s8, vreinterpret_s8_u8(d22u8));
-
-  d21s8 = vqadd_s8(vreinterpret_s8_u8(d5u8), d27s8);
-  d20s8 = vqsub_s8(vreinterpret_s8_u8(d16u8), d27s8);
-
-  *d4ru8 = veor_u8(vreinterpret_u8_s8(d21s8), d18u8);
-  *d5ru8 = veor_u8(vreinterpret_u8_s8(d19s8), d18u8);
-  *d6ru8 = veor_u8(vreinterpret_u8_s8(d26s8), d18u8);
-  *d7ru8 = veor_u8(vreinterpret_u8_s8(d20s8), d18u8);
-  return;
-}
-
-void aom_lpf_horizontal_4_neon(uint8_t *src, int pitch, const uint8_t *blimit,
-                               const uint8_t *limit, const uint8_t *thresh) {
-  int i;
-  uint8_t *s, *psrc;
-  uint8x8_t dblimit, dlimit, dthresh;
-  uint8x8_t d3u8, d4u8, d5u8, d6u8, d7u8, d16u8, d17u8, d18u8;
-
-  dblimit = vld1_u8(blimit);
-  dlimit = vld1_u8(limit);
-  dthresh = vld1_u8(thresh);
-
-  psrc = src - (pitch << 2);
-  for (i = 0; i < 1; i++) {
-    s = psrc + i * 8;
-
-    d3u8 = vld1_u8(s);
-    s += pitch;
-    d4u8 = vld1_u8(s);
-    s += pitch;
-    d5u8 = vld1_u8(s);
-    s += pitch;
-    d6u8 = vld1_u8(s);
-    s += pitch;
-    d7u8 = vld1_u8(s);
-    s += pitch;
-    d16u8 = vld1_u8(s);
-    s += pitch;
-    d17u8 = vld1_u8(s);
-    s += pitch;
-    d18u8 = vld1_u8(s);
-
-    loop_filter_neon(dblimit, dlimit, dthresh, d3u8, d4u8, d5u8, d6u8, d7u8,
-                     d16u8, d17u8, d18u8, &d4u8, &d5u8, &d6u8, &d7u8);
-
-    s -= (pitch * 5);
-    vst1_u8(s, d4u8);
-    s += pitch;
-    vst1_u8(s, d5u8);
-    s += pitch;
-    vst1_u8(s, d6u8);
-    s += pitch;
-    vst1_u8(s, d7u8);
-  }
-  return;
-}
-
-void aom_lpf_vertical_4_neon(uint8_t *src, int pitch, const uint8_t *blimit,
-                             const uint8_t *limit, const uint8_t *thresh) {
-  int i, pitch8;
-  uint8_t *s;
-  uint8x8_t dblimit, dlimit, dthresh;
-  uint8x8_t d3u8, d4u8, d5u8, d6u8, d7u8, d16u8, d17u8, d18u8;
-  uint32x2x2_t d2tmp0, d2tmp1, d2tmp2, d2tmp3;
-  uint16x4x2_t d2tmp4, d2tmp5, d2tmp6, d2tmp7;
-  uint8x8x2_t d2tmp8, d2tmp9, d2tmp10, d2tmp11;
-  uint8x8x4_t d4Result;
-
-  dblimit = vld1_u8(blimit);
-  dlimit = vld1_u8(limit);
-  dthresh = vld1_u8(thresh);
-
-  pitch8 = pitch * 8;
-  for (i = 0; i < 1; i++, src += pitch8) {
-    s = src - (i + 1) * 4;
-
-    d3u8 = vld1_u8(s);
-    s += pitch;
-    d4u8 = vld1_u8(s);
-    s += pitch;
-    d5u8 = vld1_u8(s);
-    s += pitch;
-    d6u8 = vld1_u8(s);
-    s += pitch;
-    d7u8 = vld1_u8(s);
-    s += pitch;
-    d16u8 = vld1_u8(s);
-    s += pitch;
-    d17u8 = vld1_u8(s);
-    s += pitch;
-    d18u8 = vld1_u8(s);
-
-    d2tmp0 = vtrn_u32(vreinterpret_u32_u8(d3u8), vreinterpret_u32_u8(d7u8));
-    d2tmp1 = vtrn_u32(vreinterpret_u32_u8(d4u8), vreinterpret_u32_u8(d16u8));
-    d2tmp2 = vtrn_u32(vreinterpret_u32_u8(d5u8), vreinterpret_u32_u8(d17u8));
-    d2tmp3 = vtrn_u32(vreinterpret_u32_u8(d6u8), vreinterpret_u32_u8(d18u8));
-
-    d2tmp4 = vtrn_u16(vreinterpret_u16_u32(d2tmp0.val[0]),
-                      vreinterpret_u16_u32(d2tmp2.val[0]));
-    d2tmp5 = vtrn_u16(vreinterpret_u16_u32(d2tmp1.val[0]),
-                      vreinterpret_u16_u32(d2tmp3.val[0]));
-    d2tmp6 = vtrn_u16(vreinterpret_u16_u32(d2tmp0.val[1]),
-                      vreinterpret_u16_u32(d2tmp2.val[1]));
-    d2tmp7 = vtrn_u16(vreinterpret_u16_u32(d2tmp1.val[1]),
-                      vreinterpret_u16_u32(d2tmp3.val[1]));
-
-    d2tmp8 = vtrn_u8(vreinterpret_u8_u16(d2tmp4.val[0]),
-                     vreinterpret_u8_u16(d2tmp5.val[0]));
-    d2tmp9 = vtrn_u8(vreinterpret_u8_u16(d2tmp4.val[1]),
-                     vreinterpret_u8_u16(d2tmp5.val[1]));
-    d2tmp10 = vtrn_u8(vreinterpret_u8_u16(d2tmp6.val[0]),
-                      vreinterpret_u8_u16(d2tmp7.val[0]));
-    d2tmp11 = vtrn_u8(vreinterpret_u8_u16(d2tmp6.val[1]),
-                      vreinterpret_u8_u16(d2tmp7.val[1]));
-
-    d3u8 = d2tmp8.val[0];
-    d4u8 = d2tmp8.val[1];
-    d5u8 = d2tmp9.val[0];
-    d6u8 = d2tmp9.val[1];
-    d7u8 = d2tmp10.val[0];
-    d16u8 = d2tmp10.val[1];
-    d17u8 = d2tmp11.val[0];
-    d18u8 = d2tmp11.val[1];
-
-    loop_filter_neon(dblimit, dlimit, dthresh, d3u8, d4u8, d5u8, d6u8, d7u8,
-                     d16u8, d17u8, d18u8, &d4u8, &d5u8, &d6u8, &d7u8);
-
-    d4Result.val[0] = d4u8;
-    d4Result.val[1] = d5u8;
-    d4Result.val[2] = d6u8;
-    d4Result.val[3] = d7u8;
-
-    src -= 2;
-    vst4_lane_u8(src, d4Result, 0);
-    src += pitch;
-    vst4_lane_u8(src, d4Result, 1);
-    src += pitch;
-    vst4_lane_u8(src, d4Result, 2);
-    src += pitch;
-    vst4_lane_u8(src, d4Result, 3);
-    src += pitch;
-    vst4_lane_u8(src, d4Result, 4);
-    src += pitch;
-    vst4_lane_u8(src, d4Result, 5);
-    src += pitch;
-    vst4_lane_u8(src, d4Result, 6);
-    src += pitch;
-    vst4_lane_u8(src, d4Result, 7);
-  }
-  return;
-}

aom_dsp/arm/loopfilter_8_neon.c

@@ -1,430 +0,0 @@
-/*
- * Copyright (c) 2016, Alliance for Open Media. All rights reserved
- *
- * This source code is subject to the terms of the BSD 2 Clause License and
- * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
- * was not distributed with this source code in the LICENSE file, you can
- * obtain it at www.aomedia.org/license/software. If the Alliance for Open
- * Media Patent License 1.0 was not distributed with this source code in the
- * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
- */
-
-#include <arm_neon.h>
-
-#include "./aom_dsp_rtcd.h"
-
-static INLINE void mbloop_filter_neon(uint8x8_t dblimit,   // mblimit
-                                      uint8x8_t dlimit,    // limit
-                                      uint8x8_t dthresh,   // thresh
-                                      uint8x8_t d3u8,      // p2
-                                      uint8x8_t d4u8,      // p2
-                                      uint8x8_t d5u8,      // p1
-                                      uint8x8_t d6u8,      // p0
-                                      uint8x8_t d7u8,      // q0
-                                      uint8x8_t d16u8,     // q1
-                                      uint8x8_t d17u8,     // q2
-                                      uint8x8_t d18u8,     // q3
-                                      uint8x8_t *d0ru8,    // p1
-                                      uint8x8_t *d1ru8,    // p1
-                                      uint8x8_t *d2ru8,    // p0
-                                      uint8x8_t *d3ru8,    // q0
-                                      uint8x8_t *d4ru8,    // q1
-                                      uint8x8_t *d5ru8) {  // q1
-  uint32_t flat;
-  uint8x8_t d0u8, d1u8, d2u8, d19u8, d20u8, d21u8, d22u8, d23u8, d24u8;
-  uint8x8_t d25u8, d26u8, d27u8, d28u8, d29u8, d30u8, d31u8;
-  int16x8_t q15s16;
-  uint16x8_t q10u16, q14u16;
-  int8x8_t d21s8, d24s8, d25s8, d26s8, d28s8, d29s8, d30s8;
-
-  d19u8 = vabd_u8(d3u8, d4u8);
-  d20u8 = vabd_u8(d4u8, d5u8);
-  d21u8 = vabd_u8(d5u8, d6u8);
-  d22u8 = vabd_u8(d16u8, d7u8);
-  d23u8 = vabd_u8(d17u8, d16u8);
-  d24u8 = vabd_u8(d18u8, d17u8);
-
-  d19u8 = vmax_u8(d19u8, d20u8);
-  d20u8 = vmax_u8(d21u8, d22u8);
-
-  d25u8 = vabd_u8(d6u8, d4u8);
-
-  d23u8 = vmax_u8(d23u8, d24u8);
-
-  d26u8 = vabd_u8(d7u8, d17u8);
-
-  d19u8 = vmax_u8(d19u8, d20u8);
-
-  d24u8 = vabd_u8(d6u8, d7u8);
-  d27u8 = vabd_u8(d3u8, d6u8);
-  d28u8 = vabd_u8(d18u8, d7u8);
-
-  d19u8 = vmax_u8(d19u8, d23u8);
-
-  d23u8 = vabd_u8(d5u8, d16u8);
-  d24u8 = vqadd_u8(d24u8, d24u8);
-
-  d19u8 = vcge_u8(dlimit, d19u8);
-
-  d25u8 = vmax_u8(d25u8, d26u8);
-  d26u8 = vmax_u8(d27u8, d28u8);
-
-  d23u8 = vshr_n_u8(d23u8, 1);
-
-  d25u8 = vmax_u8(d25u8, d26u8);
-
-  d24u8 = vqadd_u8(d24u8, d23u8);
-
-  d20u8 = vmax_u8(d20u8, d25u8);
-
-  d23u8 = vdup_n_u8(1);
-  d24u8 = vcge_u8(dblimit, d24u8);
-
-  d21u8 = vcgt_u8(d21u8, dthresh);
-
-  d20u8 = vcge_u8(d23u8, d20u8);
-
-  d19u8 = vand_u8(d19u8, d24u8);
-
-  d23u8 = vcgt_u8(d22u8, dthresh);
-
-  d20u8 = vand_u8(d20u8, d19u8);
-
-  d22u8 = vdup_n_u8(0x80);
-
-  d23u8 = vorr_u8(d21u8, d23u8);
-
-  q10u16 = vcombine_u16(vreinterpret_u16_u8(d20u8), vreinterpret_u16_u8(d21u8));
-
-  d30u8 = vshrn_n_u16(q10u16, 4);
-  flat = vget_lane_u32(vreinterpret_u32_u8(d30u8), 0);
-
-  if (flat == 0xffffffff) {  // Check for all 1's, power_branch_only
-    d27u8 = vdup_n_u8(3);
-    d21u8 = vdup_n_u8(2);
-    q14u16 = vaddl_u8(d6u8, d7u8);
-    q14u16 = vmlal_u8(q14u16, d3u8, d27u8);
-    q14u16 = vmlal_u8(q14u16, d4u8, d21u8);
-    q14u16 = vaddw_u8(q14u16, d5u8);
-    *d0ru8 = vqrshrn_n_u16(q14u16, 3);
-
-    q14u16 = vsubw_u8(q14u16, d3u8);
-    q14u16 = vsubw_u8(q14u16, d4u8);
-    q14u16 = vaddw_u8(q14u16, d5u8);
-    q14u16 = vaddw_u8(q14u16, d16u8);
-    *d1ru8 = vqrshrn_n_u16(q14u16, 3);
-
-    q14u16 = vsubw_u8(q14u16, d3u8);
-    q14u16 = vsubw_u8(q14u16, d5u8);
-    q14u16 = vaddw_u8(q14u16, d6u8);
-    q14u16 = vaddw_u8(q14u16, d17u8);
-    *d2ru8 = vqrshrn_n_u16(q14u16, 3);
-
-    q14u16 = vsubw_u8(q14u16, d3u8);
-    q14u16 = vsubw_u8(q14u16, d6u8);
-    q14u16 = vaddw_u8(q14u16, d7u8);
-    q14u16 = vaddw_u8(q14u16, d18u8);
-    *d3ru8 = vqrshrn_n_u16(q14u16, 3);
-
-    q14u16 = vsubw_u8(q14u16, d4u8);
-    q14u16 = vsubw_u8(q14u16, d7u8);
-    q14u16 = vaddw_u8(q14u16, d16u8);
-    q14u16 = vaddw_u8(q14u16, d18u8);
-    *d4ru8 = vqrshrn_n_u16(q14u16, 3);
-
-    q14u16 = vsubw_u8(q14u16, d5u8);
-    q14u16 = vsubw_u8(q14u16, d16u8);
-    q14u16 = vaddw_u8(q14u16, d17u8);
-    q14u16 = vaddw_u8(q14u16, d18u8);
-    *d5ru8 = vqrshrn_n_u16(q14u16, 3);
-  } else {
-    d21u8 = veor_u8(d7u8, d22u8);
-    d24u8 = veor_u8(d6u8, d22u8);
-    d25u8 = veor_u8(d5u8, d22u8);
-    d26u8 = veor_u8(d16u8, d22u8);
-
-    d27u8 = vdup_n_u8(3);
-
-    d28s8 = vsub_s8(vreinterpret_s8_u8(d21u8), vreinterpret_s8_u8(d24u8));
-    d29s8 = vqsub_s8(vreinterpret_s8_u8(d25u8), vreinterpret_s8_u8(d26u8));
-
-    q15s16 = vmull_s8(d28s8, vreinterpret_s8_u8(d27u8));
-
-    d29s8 = vand_s8(d29s8, vreinterpret_s8_u8(d23u8));
-
-    q15s16 = vaddw_s8(q15s16, d29s8);
-
-    d29u8 = vdup_n_u8(4);
-
-    d28s8 = vqmovn_s16(q15s16);
-
-    d28s8 = vand_s8(d28s8, vreinterpret_s8_u8(d19u8));
-
-    d30s8 = vqadd_s8(d28s8, vreinterpret_s8_u8(d27u8));
-    d29s8 = vqadd_s8(d28s8, vreinterpret_s8_u8(d29u8));
-    d30s8 = vshr_n_s8(d30s8, 3);
-    d29s8 = vshr_n_s8(d29s8, 3);
-
-    d24s8 = vqadd_s8(vreinterpret_s8_u8(d24u8), d30s8);
-    d21s8 = vqsub_s8(vreinterpret_s8_u8(d21u8), d29s8);
-
-    d29s8 = vrshr_n_s8(d29s8, 1);
-    d29s8 = vbic_s8(d29s8, vreinterpret_s8_u8(d23u8));
-
-    d25s8 = vqadd_s8(vreinterpret_s8_u8(d25u8), d29s8);
-    d26s8 = vqsub_s8(vreinterpret_s8_u8(d26u8), d29s8);
-
-    if (flat == 0) {  // filter_branch_only
-      *d0ru8 = d4u8;
-      *d1ru8 = veor_u8(vreinterpret_u8_s8(d25s8), d22u8);
-      *d2ru8 = veor_u8(vreinterpret_u8_s8(d24s8), d22u8);
-      *d3ru8 = veor_u8(vreinterpret_u8_s8(d21s8), d22u8);
-      *d4ru8 = veor_u8(vreinterpret_u8_s8(d26s8), d22u8);
-      *d5ru8 = d17u8;
-      return;
-    }
-
-    d21u8 = veor_u8(vreinterpret_u8_s8(d21s8), d22u8);
-    d24u8 = veor_u8(vreinterpret_u8_s8(d24s8), d22u8);
-    d25u8 = veor_u8(vreinterpret_u8_s8(d25s8), d22u8);
-    d26u8 = veor_u8(vreinterpret_u8_s8(d26s8), d22u8);
-
-    d23u8 = vdup_n_u8(2);
-    q14u16 = vaddl_u8(d6u8, d7u8);
-    q14u16 = vmlal_u8(q14u16, d3u8, d27u8);
-    q14u16 = vmlal_u8(q14u16, d4u8, d23u8);
-
-    d0u8 = vbsl_u8(d20u8, dblimit, d4u8);
-
-    q14u16 = vaddw_u8(q14u16, d5u8);
-
-    d1u8 = vbsl_u8(d20u8, dlimit, d25u8);
-
-    d30u8 = vqrshrn_n_u16(q14u16, 3);
-
-    q14u16 = vsubw_u8(q14u16, d3u8);
-    q14u16 = vsubw_u8(q14u16, d4u8);
-    q14u16 = vaddw_u8(q14u16, d5u8);
-    q14u16 = vaddw_u8(q14u16, d16u8);
-
-    d2u8 = vbsl_u8(d20u8, dthresh, d24u8);
-
-    d31u8 = vqrshrn_n_u16(q14u16, 3);
-
-    q14u16 = vsubw_u8(q14u16, d3u8);
-    q14u16 = vsubw_u8(q14u16, d5u8);
-    q14u16 = vaddw_u8(q14u16, d6u8);
-    q14u16 = vaddw_u8(q14u16, d17u8);
-
-    *d0ru8 = vbsl_u8(d20u8, d30u8, d0u8);
-
-    d23u8 = vqrshrn_n_u16(q14u16, 3);
-
-    q14u16 = vsubw_u8(q14u16, d3u8);
-    q14u16 = vsubw_u8(q14u16, d6u8);
-    q14u16 = vaddw_u8(q14u16, d7u8);
-
-    *d1ru8 = vbsl_u8(d20u8, d31u8, d1u8);
-
-    q14u16 = vaddw_u8(q14u16, d18u8);
-
-    *d2ru8 = vbsl_u8(d20u8, d23u8, d2u8);
-
-    d22u8 = vqrshrn_n_u16(q14u16, 3);
-
-    q14u16 = vsubw_u8(q14u16, d4u8);
-    q14u16 = vsubw_u8(q14u16, d7u8);
-    q14u16 = vaddw_u8(q14u16, d16u8);
-
-    d3u8 = vbsl_u8(d20u8, d3u8, d21u8);
-
-    q14u16 = vaddw_u8(q14u16, d18u8);
-
-    d4u8 = vbsl_u8(d20u8, d4u8, d26u8);
-
-    d6u8 = vqrshrn_n_u16(q14u16, 3);
-
-    q14u16 = vsubw_u8(q14u16, d5u8);
-    q14u16 = vsubw_u8(q14u16, d16u8);
-    q14u16 = vaddw_u8(q14u16, d17u8);
-    q14u16 = vaddw_u8(q14u16, d18u8);
-
-    d5u8 = vbsl_u8(d20u8, d5u8, d17u8);
-
-    d7u8 = vqrshrn_n_u16(q14u16, 3);
-
-    *d3ru8 = vbsl_u8(d20u8, d22u8, d3u8);
-    *d4ru8 = vbsl_u8(d20u8, d6u8, d4u8);
-    *d5ru8 = vbsl_u8(d20u8, d7u8, d5u8);
-  }
-  return;
-}
-
-void aom_lpf_horizontal_8_neon(uint8_t *src, int pitch, const uint8_t *blimit,
-                               const uint8_t *limit, const uint8_t *thresh) {
-  int i;
-  uint8_t *s, *psrc;
-  uint8x8_t dblimit, dlimit, dthresh;
-  uint8x8_t d0u8, d1u8, d2u8, d3u8, d4u8, d5u8, d6u8, d7u8;
-  uint8x8_t d16u8, d17u8, d18u8;
-
-  dblimit = vld1_u8(blimit);
-  dlimit = vld1_u8(limit);
-  dthresh = vld1_u8(thresh);
-
-  psrc = src - (pitch << 2);
-  for (i = 0; i < 1; i++) {
-    s = psrc + i * 8;
-
-    d3u8 = vld1_u8(s);
-    s += pitch;
-    d4u8 = vld1_u8(s);
-    s += pitch;
-    d5u8 = vld1_u8(s);
-    s += pitch;
-    d6u8 = vld1_u8(s);
-    s += pitch;
-    d7u8 = vld1_u8(s);
-    s += pitch;
-    d16u8 = vld1_u8(s);
-    s += pitch;
-    d17u8 = vld1_u8(s);
-    s += pitch;
-    d18u8 = vld1_u8(s);
-
-    mbloop_filter_neon(dblimit, dlimit, dthresh, d3u8, d4u8, d5u8, d6u8, d7u8,
-                       d16u8, d17u8, d18u8, &d0u8, &d1u8, &d2u8, &d3u8, &d4u8,
-                       &d5u8);
-
-    s -= (pitch * 6);
-    vst1_u8(s, d0u8);
-    s += pitch;
-    vst1_u8(s, d1u8);
-    s += pitch;
-    vst1_u8(s, d2u8);
-    s += pitch;
-    vst1_u8(s, d3u8);
-    s += pitch;
-    vst1_u8(s, d4u8);
-    s += pitch;
-    vst1_u8(s, d5u8);
-  }
-  return;
-}
-
-void aom_lpf_vertical_8_neon(uint8_t *src, int pitch, const uint8_t *blimit,
-                             const uint8_t *limit, const uint8_t *thresh) {
-  int i;
-  uint8_t *s;
-  uint8x8_t dblimit, dlimit, dthresh;
-  uint8x8_t d0u8, d1u8, d2u8, d3u8, d4u8, d5u8, d6u8, d7u8;
-  uint8x8_t d16u8, d17u8, d18u8;
-  uint32x2x2_t d2tmp0, d2tmp1, d2tmp2, d2tmp3;
-  uint16x4x2_t d2tmp4, d2tmp5, d2tmp6, d2tmp7;
-  uint8x8x2_t d2tmp8, d2tmp9, d2tmp10, d2tmp11;
-  uint8x8x4_t d4Result;
-  uint8x8x2_t d2Result;
-
-  dblimit = vld1_u8(blimit);
-  dlimit = vld1_u8(limit);
-  dthresh = vld1_u8(thresh);
-
-  for (i = 0; i < 1; i++) {
-    s = src + (i * (pitch << 3)) - 4;
-
-    d3u8 = vld1_u8(s);
-    s += pitch;
-    d4u8 = vld1_u8(s);
-    s += pitch;
-    d5u8 = vld1_u8(s);
-    s += pitch;
-    d6u8 = vld1_u8(s);
-    s += pitch;
-    d7u8 = vld1_u8(s);
-    s += pitch;
-    d16u8 = vld1_u8(s);
-    s += pitch;
-    d17u8 = vld1_u8(s);
-    s += pitch;
-    d18u8 = vld1_u8(s);
-
-    d2tmp0 = vtrn_u32(vreinterpret_u32_u8(d3u8), vreinterpret_u32_u8(d7u8));
-    d2tmp1 = vtrn_u32(vreinterpret_u32_u8(d4u8), vreinterpret_u32_u8(d16u8));
-    d2tmp2 = vtrn_u32(vreinterpret_u32_u8(d5u8), vreinterpret_u32_u8(d17u8));
-    d2tmp3 = vtrn_u32(vreinterpret_u32_u8(d6u8), vreinterpret_u32_u8(d18u8));
-
-    d2tmp4 = vtrn_u16(vreinterpret_u16_u32(d2tmp0.val[0]),
-                      vreinterpret_u16_u32(d2tmp2.val[0]));
-    d2tmp5 = vtrn_u16(vreinterpret_u16_u32(d2tmp1.val[0]),
-                      vreinterpret_u16_u32(d2tmp3.val[0]));
-    d2tmp6 = vtrn_u16(vreinterpret_u16_u32(d2tmp0.val[1]),
-                      vreinterpret_u16_u32(d2tmp2.val[1]));
-    d2tmp7 = vtrn_u16(vreinterpret_u16_u32(d2tmp1.val[1]),
-                      vreinterpret_u16_u32(d2tmp3.val[1]));
-
-    d2tmp8 = vtrn_u8(vreinterpret_u8_u16(d2tmp4.val[0]),
-                     vreinterpret_u8_u16(d2tmp5.val[0]));
-    d2tmp9 = vtrn_u8(vreinterpret_u8_u16(d2tmp4.val[1]),
-                     vreinterpret_u8_u16(d2tmp5.val[1]));
-    d2tmp10 = vtrn_u8(vreinterpret_u8_u16(d2tmp6.val[0]),
-                      vreinterpret_u8_u16(d2tmp7.val[0]));
-    d2tmp11 = vtrn_u8(vreinterpret_u8_u16(d2tmp6.val[1]),
-                      vreinterpret_u8_u16(d2tmp7.val[1]));
-
-    d3u8 = d2tmp8.val[0];
-    d4u8 = d2tmp8.val[1];
-    d5u8 = d2tmp9.val[0];
-    d6u8 = d2tmp9.val[1];
-    d7u8 = d2tmp10.val[0];
-    d16u8 = d2tmp10.val[1];
-    d17u8 = d2tmp11.val[0];
-    d18u8 = d2tmp11.val[1];
-
-    mbloop_filter_neon(dblimit, dlimit, dthresh, d3u8, d4u8, d5u8, d6u8, d7u8,
-                       d16u8, d17u8, d18u8, &d0u8, &d1u8, &d2u8, &d3u8, &d4u8,
-                       &d5u8);
-
-    d4Result.val[0] = d0u8;
-    d4Result.val[1] = d1u8;
-    d4Result.val[2] = d2u8;
-    d4Result.val[3] = d3u8;
-
-    d2Result.val[0] = d4u8;
-    d2Result.val[1] = d5u8;
-
-    s = src - 3;
-    vst4_lane_u8(s, d4Result, 0);
-    s += pitch;
-    vst4_lane_u8(s, d4Result, 1);
-    s += pitch;
-    vst4_lane_u8(s, d4Result, 2);
-    s += pitch;
-    vst4_lane_u8(s, d4Result, 3);
-    s += pitch;
-    vst4_lane_u8(s, d4Result, 4);
-    s += pitch;
-    vst4_lane_u8(s, d4Result, 5);
-    s += pitch;
-    vst4_lane_u8(s, d4Result, 6);
-    s += pitch;
-    vst4_lane_u8(s, d4Result, 7);
-
-    s = src + 1;
-    vst2_lane_u8(s, d2Result, 0);
-    s += pitch;
-    vst2_lane_u8(s, d2Result, 1);
-    s += pitch;
-    vst2_lane_u8(s, d2Result, 2);
-    s += pitch;
-    vst2_lane_u8(s, d2Result, 3);
-    s += pitch;
-    vst2_lane_u8(s, d2Result, 4);
-    s += pitch;
-    vst2_lane_u8(s, d2Result, 5);
-    s += pitch;
-    vst2_lane_u8(s, d2Result, 6);
-    s += pitch;
-    vst2_lane_u8(s, d2Result, 7);
-  }
-  return;
-}

aom_dsp/arm/loopfilter_neon.c

@@ -1,49 +0,0 @@
-/*
- * Copyright (c) 2016, Alliance for Open Media. All rights reserved
- *
- * This source code is subject to the terms of the BSD 2 Clause License and
- * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
- * was not distributed with this source code in the LICENSE file, you can
- * obtain it at www.aomedia.org/license/software. If the Alliance for Open
- * Media Patent License 1.0 was not distributed with this source code in the
- * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
- */
-
-#include <arm_neon.h>
-
-#include "./aom_dsp_rtcd.h"
-#include "./aom_config.h"
-#include "aom/aom_integer.h"
-
-void aom_lpf_vertical_4_dual_neon(uint8_t *s, int p, const uint8_t *blimit0,
-                                  const uint8_t *limit0, const uint8_t *thresh0,
-                                  const uint8_t *blimit1, const uint8_t *limit1,
-                                  const uint8_t *thresh1) {
-  aom_lpf_vertical_4_neon(s, p, blimit0, limit0, thresh0);
-  aom_lpf_vertical_4_neon(s + 8 * p, p, blimit1, limit1, thresh1);
-}
-
-#if HAVE_NEON_ASM
-void aom_lpf_horizontal_8_dual_neon(
-    uint8_t *s, int p /* pitch */, const uint8_t *blimit0,
-    const uint8_t *limit0, const uint8_t *thresh0, const uint8_t *blimit1,
-    const uint8_t *limit1, const uint8_t *thresh1) {
-  aom_lpf_horizontal_8_neon(s, p, blimit0, limit0, thresh0);
-  aom_lpf_horizontal_8_neon(s + 8, p, blimit1, limit1, thresh1);
-}
-
-void aom_lpf_vertical_8_dual_neon(uint8_t *s, int p, const uint8_t *blimit0,
-                                  const uint8_t *limit0, const uint8_t *thresh0,
-                                  const uint8_t *blimit1, const uint8_t *limit1,
-                                  const uint8_t *thresh1) {
-  aom_lpf_vertical_8_neon(s, p, blimit0, limit0, thresh0);
-  aom_lpf_vertical_8_neon(s + 8 * p, p, blimit1, limit1, thresh1);
-}
-
-void aom_lpf_vertical_16_dual_neon(uint8_t *s, int p, const uint8_t *blimit,
-                                   const uint8_t *limit,
-                                   const uint8_t *thresh) {
-  aom_lpf_vertical_16_neon(s, p, blimit, limit, thresh);
-  aom_lpf_vertical_16_neon(s + 8 * p, p, blimit, limit, thresh);
-}
-#endif  // HAVE_NEON_ASM

aom_dsp/arm/sad4d_neon.c

@@ -1,225 +0,0 @@
-/*
- * Copyright (c) 2016, Alliance for Open Media. All rights reserved
- *
- * This source code is subject to the terms of the BSD 2 Clause License and
- * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
- * was not distributed with this source code in the LICENSE file, you can
- * obtain it at www.aomedia.org/license/software. If the Alliance for Open
- * Media Patent License 1.0 was not distributed with this source code in the
- * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
- */
-
-#include <arm_neon.h>
-
-#include "./aom_config.h"
-#include "./aom_dsp_rtcd.h"
-#include "aom/aom_integer.h"
-
-static INLINE unsigned int horizontal_long_add_16x8(const uint16x8_t vec_lo,
-                                                    const uint16x8_t vec_hi) {
-  const uint32x4_t vec_l_lo =
-      vaddl_u16(vget_low_u16(vec_lo), vget_high_u16(vec_lo));
-  const uint32x4_t vec_l_hi =
-      vaddl_u16(vget_low_u16(vec_hi), vget_high_u16(vec_hi));
-  const uint32x4_t a = vaddq_u32(vec_l_lo, vec_l_hi);
-  const uint64x2_t b = vpaddlq_u32(a);
-  const uint32x2_t c = vadd_u32(vreinterpret_u32_u64(vget_low_u64(b)),
-                                vreinterpret_u32_u64(vget_high_u64(b)));
-  return vget_lane_u32(c, 0);
-}
-
-// Calculate the absolute difference of 64 bytes from vec_src_00, vec_src_16,
-// vec_src_32, vec_src_48 and ref. Accumulate partial sums in vec_sum_ref_lo
-// and vec_sum_ref_hi.
-static void sad_neon_64(const uint8x16_t vec_src_00,
-                        const uint8x16_t vec_src_16,
-                        const uint8x16_t vec_src_32,
-                        const uint8x16_t vec_src_48, const uint8_t *ref,
-                        uint16x8_t *vec_sum_ref_lo,
-                        uint16x8_t *vec_sum_ref_hi) {
-  const uint8x16_t vec_ref_00 = vld1q_u8(ref);
-  const uint8x16_t vec_ref_16 = vld1q_u8(ref + 16);
-  const uint8x16_t vec_ref_32 = vld1q_u8(ref + 32);
-  const uint8x16_t vec_ref_48 = vld1q_u8(ref + 48);
-
-  *vec_sum_ref_lo = vabal_u8(*vec_sum_ref_lo, vget_low_u8(vec_src_00),
-                             vget_low_u8(vec_ref_00));
-  *vec_sum_ref_hi = vabal_u8(*vec_sum_ref_hi, vget_high_u8(vec_src_00),
-                             vget_high_u8(vec_ref_00));
-  *vec_sum_ref_lo = vabal_u8(*vec_sum_ref_lo, vget_low_u8(vec_src_16),
-                             vget_low_u8(vec_ref_16));
-  *vec_sum_ref_hi = vabal_u8(*vec_sum_ref_hi, vget_high_u8(vec_src_16),
-                             vget_high_u8(vec_ref_16));
-  *vec_sum_ref_lo = vabal_u8(*vec_sum_ref_lo, vget_low_u8(vec_src_32),
-                             vget_low_u8(vec_ref_32));
-  *vec_sum_ref_hi = vabal_u8(*vec_sum_ref_hi, vget_high_u8(vec_src_32),
-                             vget_high_u8(vec_ref_32));
-  *vec_sum_ref_lo = vabal_u8(*vec_sum_ref_lo, vget_low_u8(vec_src_48),
-                             vget_low_u8(vec_ref_48));
-  *vec_sum_ref_hi = vabal_u8(*vec_sum_ref_hi, vget_high_u8(vec_src_48),
-                             vget_high_u8(vec_ref_48));
-}
-
-// Calculate the absolute difference of 32 bytes from vec_src_00, vec_src_16,
-// and ref. Accumulate partial sums in vec_sum_ref_lo and vec_sum_ref_hi.
-static void sad_neon_32(const uint8x16_t vec_src_00,
-                        const uint8x16_t vec_src_16, const uint8_t *ref,
-                        uint16x8_t *vec_sum_ref_lo,
-                        uint16x8_t *vec_sum_ref_hi) {
-  const uint8x16_t vec_ref_00 = vld1q_u8(ref);
-  const uint8x16_t vec_ref_16 = vld1q_u8(ref + 16);
-
-  *vec_sum_ref_lo = vabal_u8(*vec_sum_ref_lo, vget_low_u8(vec_src_00),
-                             vget_low_u8(vec_ref_00));
-  *vec_sum_ref_hi = vabal_u8(*vec_sum_ref_hi, vget_high_u8(vec_src_00),
-                             vget_high_u8(vec_ref_00));
-  *vec_sum_ref_lo = vabal_u8(*vec_sum_ref_lo, vget_low_u8(vec_src_16),
-                             vget_low_u8(vec_ref_16));
-  *vec_sum_ref_hi = vabal_u8(*vec_sum_ref_hi, vget_high_u8(vec_src_16),
-                             vget_high_u8(vec_ref_16));
-}
-
-void aom_sad64x64x4d_neon(const uint8_t *src, int src_stride,
-                          const uint8_t *const ref[4], int ref_stride,
-                          uint32_t *res) {
-  int i;
-  uint16x8_t vec_sum_ref0_lo = vdupq_n_u16(0);
-  uint16x8_t vec_sum_ref0_hi = vdupq_n_u16(0);
-  uint16x8_t vec_sum_ref1_lo = vdupq_n_u16(0);
-  uint16x8_t vec_sum_ref1_hi = vdupq_n_u16(0);
-  uint16x8_t vec_sum_ref2_lo = vdupq_n_u16(0);
-  uint16x8_t vec_sum_ref2_hi = vdupq_n_u16(0);
-  uint16x8_t vec_sum_ref3_lo = vdupq_n_u16(0);
-  uint16x8_t vec_sum_ref3_hi = vdupq_n_u16(0);
-  const uint8_t *ref0, *ref1, *ref2, *ref3;
-  ref0 = ref[0];
-  ref1 = ref[1];
-  ref2 = ref[2];
-  ref3 = ref[3];
-
-  for (i = 0; i < 64; ++i) {
-    const uint8x16_t vec_src_00 = vld1q_u8(src);
-    const uint8x16_t vec_src_16 = vld1q_u8(src + 16);
-    const uint8x16_t vec_src_32 = vld1q_u8(src + 32);
-    const uint8x16_t vec_src_48 = vld1q_u8(src + 48);
-
-    sad_neon_64(vec_src_00, vec_src_16, vec_src_32, vec_src_48, ref0,
-                &vec_sum_ref0_lo, &vec_sum_ref0_hi);
-    sad_neon_64(vec_src_00, vec_src_16, vec_src_32, vec_src_48, ref1,
-                &vec_sum_ref1_lo, &vec_sum_ref1_hi);
-    sad_neon_64(vec_src_00, vec_src_16, vec_src_32, vec_src_48, ref2,
-                &vec_sum_ref2_lo, &vec_sum_ref2_hi);
-    sad_neon_64(vec_src_00, vec_src_16, vec_src_32, vec_src_48, ref3,
-                &vec_sum_ref3_lo, &vec_sum_ref3_hi);
-
-    src += src_stride;
-    ref0 += ref_stride;
-    ref1 += ref_stride;
-    ref2 += ref_stride;
-    ref3 += ref_stride;
-  }
-
-  res[0] = horizontal_long_add_16x8(vec_sum_ref0_lo, vec_sum_ref0_hi);
-  res[1] = horizontal_long_add_16x8(vec_sum_ref1_lo, vec_sum_ref1_hi);
-  res[2] = horizontal_long_add_16x8(vec_sum_ref2_lo, vec_sum_ref2_hi);
-  res[3] = horizontal_long_add_16x8(vec_sum_ref3_lo, vec_sum_ref3_hi);
-}
-
-void aom_sad32x32x4d_neon(const uint8_t *src, int src_stride,
-                          const uint8_t *const ref[4], int ref_stride,
-                          uint32_t *res) {
-  int i;
-  uint16x8_t vec_sum_ref0_lo = vdupq_n_u16(0);
-  uint16x8_t vec_sum_ref0_hi = vdupq_n_u16(0);
-  uint16x8_t vec_sum_ref1_lo = vdupq_n_u16(0);
-  uint16x8_t vec_sum_ref1_hi = vdupq_n_u16(0);
-  uint16x8_t vec_sum_ref2_lo = vdupq_n_u16(0);
-  uint16x8_t vec_sum_ref2_hi = vdupq_n_u16(0);
-  uint16x8_t vec_sum_ref3_lo = vdupq_n_u16(0);
-  uint16x8_t vec_sum_ref3_hi = vdupq_n_u16(0);
-  const uint8_t *ref0, *ref1, *ref2, *ref3;
-  ref0 = ref[0];
-  ref1 = ref[1];
-  ref2 = ref[2];
-  ref3 = ref[3];
-
-  for (i = 0; i < 32; ++i) {
-    const uint8x16_t vec_src_00 = vld1q_u8(src);
-    const uint8x16_t vec_src_16 = vld1q_u8(src + 16);
-
-    sad_neon_32(vec_src_00, vec_src_16, ref0, &vec_sum_ref0_lo,
-                &vec_sum_ref0_hi);
-    sad_neon_32(vec_src_00, vec_src_16, ref1, &vec_sum_ref1_lo,
-                &vec_sum_ref1_hi);
-    sad_neon_32(vec_src_00, vec_src_16, ref2, &vec_sum_ref2_lo,
-                &vec_sum_ref2_hi);
-    sad_neon_32(vec_src_00, vec_src_16, ref3, &vec_sum_ref3_lo,
-                &vec_sum_ref3_hi);
-
-    src += src_stride;
-    ref0 += ref_stride;
-    ref1 += ref_stride;
-    ref2 += ref_stride;
-    ref3 += ref_stride;
-  }
-
-  res[0] = horizontal_long_add_16x8(vec_sum_ref0_lo, vec_sum_ref0_hi);
-  res[1] = horizontal_long_add_16x8(vec_sum_ref1_lo, vec_sum_ref1_hi);
-  res[2] = horizontal_long_add_16x8(vec_sum_ref2_lo, vec_sum_ref2_hi);
-  res[3] = horizontal_long_add_16x8(vec_sum_ref3_lo, vec_sum_ref3_hi);
-}
-
-void aom_sad16x16x4d_neon(const uint8_t *src, int src_stride,
-                          const uint8_t *const ref[4], int ref_stride,
-                          uint32_t *res) {
-  int i;
-  uint16x8_t vec_sum_ref0_lo = vdupq_n_u16(0);
-  uint16x8_t vec_sum_ref0_hi = vdupq_n_u16(0);
-  uint16x8_t vec_sum_ref1_lo = vdupq_n_u16(0);
-  uint16x8_t vec_sum_ref1_hi = vdupq_n_u16(0);
-  uint16x8_t vec_sum_ref2_lo = vdupq_n_u16(0);
-  uint16x8_t vec_sum_ref2_hi = vdupq_n_u16(0);
-  uint16x8_t vec_sum_ref3_lo = vdupq_n_u16(0);
-  uint16x8_t vec_sum_ref3_hi = vdupq_n_u16(0);
-  const uint8_t *ref0, *ref1, *ref2, *ref3;
-  ref0 = ref[0];
-  ref1 = ref[1];
-  ref2 = ref[2];
-  ref3 = ref[3];
-
-  for (i = 0; i < 16; ++i) {
-    const uint8x16_t vec_src = vld1q_u8(src);
-    const uint8x16_t vec_ref0 = vld1q_u8(ref0);
-    const uint8x16_t vec_ref1 = vld1q_u8(ref1);
-    const uint8x16_t vec_ref2 = vld1q_u8(ref2);
-    const uint8x16_t vec_ref3 = vld1q_u8(ref3);
-
-    vec_sum_ref0_lo =
-        vabal_u8(vec_sum_ref0_lo, vget_low_u8(vec_src), vget_low_u8(vec_ref0));
-    vec_sum_ref0_hi = vabal_u8(vec_sum_ref0_hi, vget_high_u8(vec_src),
-                               vget_high_u8(vec_ref0));
-    vec_sum_ref1_lo =
-        vabal_u8(vec_sum_ref1_lo, vget_low_u8(vec_src), vget_low_u8(vec_ref1));
-    vec_sum_ref1_hi = vabal_u8(vec_sum_ref1_hi, vget_high_u8(vec_src),
-                               vget_high_u8(vec_ref1));
-    vec_sum_ref2_lo =
-        vabal_u8(vec_sum_ref2_lo, vget_low_u8(vec_src), vget_low_u8(vec_ref2));
-    vec_sum_ref2_hi = vabal_u8(vec_sum_ref2_hi, vget_high_u8(vec_src),
-                               vget_high_u8(vec_ref2));
-    vec_sum_ref3_lo =
-        vabal_u8(vec_sum_ref3_lo, vget_low_u8(vec_src), vget_low_u8(vec_ref3));
-    vec_sum_ref3_hi = vabal_u8(vec_sum_ref3_hi, vget_high_u8(vec_src),
-                               vget_high_u8(vec_ref3));
-
-    src += src_stride;
-    ref0 += ref_stride;
-    ref1 += ref_stride;
-    ref2 += ref_stride;
-    ref3 += ref_stride;
-  }
-
-  res[0] = horizontal_long_add_16x8(vec_sum_ref0_lo, vec_sum_ref0_hi);
-  res[1] = horizontal_long_add_16x8(vec_sum_ref1_lo, vec_sum_ref1_hi);
-  res[2] = horizontal_long_add_16x8(vec_sum_ref2_lo, vec_sum_ref2_hi);
-  res[3] = horizontal_long_add_16x8(vec_sum_ref3_lo, vec_sum_ref3_hi);
-}

aom_dsp/arm/sad_neon.c

@@ -1,224 +0,0 @@
-/*
- * Copyright (c) 2016, Alliance for Open Media. All rights reserved
- *
- * This source code is subject to the terms of the BSD 2 Clause License and
- * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
- * was not distributed with this source code in the LICENSE file, you can
- * obtain it at www.aomedia.org/license/software. If the Alliance for Open
- * Media Patent License 1.0 was not distributed with this source code in the
- * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
- */
-
-#include <arm_neon.h>
-
-#include "./aom_config.h"
-
-#include "aom/aom_integer.h"
-
-unsigned int aom_sad8x16_neon(unsigned char *src_ptr, int src_stride,
-                              unsigned char *ref_ptr, int ref_stride) {
-  uint8x8_t d0, d8;
-  uint16x8_t q12;
-  uint32x4_t q1;
-  uint64x2_t q3;
-  uint32x2_t d5;
-  int i;
-
-  d0 = vld1_u8(src_ptr);
-  src_ptr += src_stride;
-  d8 = vld1_u8(ref_ptr);
-  ref_ptr += ref_stride;
-  q12 = vabdl_u8(d0, d8);
-
-  for (i = 0; i < 15; i++) {
-    d0 = vld1_u8(src_ptr);
-    src_ptr += src_stride;
-    d8 = vld1_u8(ref_ptr);
-    ref_ptr += ref_stride;
-    q12 = vabal_u8(q12, d0, d8);
-  }
-
-  q1 = vpaddlq_u16(q12);
-  q3 = vpaddlq_u32(q1);
-  d5 = vadd_u32(vreinterpret_u32_u64(vget_low_u64(q3)),
-                vreinterpret_u32_u64(vget_high_u64(q3)));
-
-  return vget_lane_u32(d5, 0);
-}
-
-unsigned int aom_sad4x4_neon(unsigned char *src_ptr, int src_stride,
-                             unsigned char *ref_ptr, int ref_stride) {
-  uint8x8_t d0, d8;
-  uint16x8_t q12;
-  uint32x2_t d1;
-  uint64x1_t d3;
-  int i;
-
-  d0 = vld1_u8(src_ptr);
-  src_ptr += src_stride;
-  d8 = vld1_u8(ref_ptr);
-  ref_ptr += ref_stride;
-  q12 = vabdl_u8(d0, d8);
-
-  for (i = 0; i < 3; i++) {
-    d0 = vld1_u8(src_ptr);
-    src_ptr += src_stride;
-    d8 = vld1_u8(ref_ptr);
-    ref_ptr += ref_stride;
-    q12 = vabal_u8(q12, d0, d8);
-  }
-
-  d1 = vpaddl_u16(vget_low_u16(q12));
-  d3 = vpaddl_u32(d1);
-
-  return vget_lane_u32(vreinterpret_u32_u64(d3), 0);
-}
-
-unsigned int aom_sad16x8_neon(unsigned char *src_ptr, int src_stride,
-                              unsigned char *ref_ptr, int ref_stride) {
-  uint8x16_t q0, q4;
-  uint16x8_t q12, q13;
-  uint32x4_t q1;
-  uint64x2_t q3;
-  uint32x2_t d5;
-  int i;
-
-  q0 = vld1q_u8(src_ptr);
-  src_ptr += src_stride;
-  q4 = vld1q_u8(ref_ptr);
-  ref_ptr += ref_stride;
-  q12 = vabdl_u8(vget_low_u8(q0), vget_low_u8(q4));
-  q13 = vabdl_u8(vget_high_u8(q0), vget_high_u8(q4));
-
-  for (i = 0; i < 7; i++) {
-    q0 = vld1q_u8(src_ptr);
-    src_ptr += src_stride;
-    q4 = vld1q_u8(ref_ptr);
-    ref_ptr += ref_stride;
-    q12 = vabal_u8(q12, vget_low_u8(q0), vget_low_u8(q4));
-    q13 = vabal_u8(q13, vget_high_u8(q0), vget_high_u8(q4));
-  }
-
-  q12 = vaddq_u16(q12, q13);
-  q1 = vpaddlq_u16(q12);
-  q3 = vpaddlq_u32(q1);
-  d5 = vadd_u32(vreinterpret_u32_u64(vget_low_u64(q3)),
-                vreinterpret_u32_u64(vget_high_u64(q3)));
-
-  return vget_lane_u32(d5, 0);
-}
-
-static INLINE unsigned int horizontal_long_add_16x8(const uint16x8_t vec_lo,
-                                                    const uint16x8_t vec_hi) {
-  const uint32x4_t vec_l_lo =
-      vaddl_u16(vget_low_u16(vec_lo), vget_high_u16(vec_lo));
-  const uint32x4_t vec_l_hi =
-      vaddl_u16(vget_low_u16(vec_hi), vget_high_u16(vec_hi));
-  const uint32x4_t a = vaddq_u32(vec_l_lo, vec_l_hi);
-  const uint64x2_t b = vpaddlq_u32(a);
-  const uint32x2_t c = vadd_u32(vreinterpret_u32_u64(vget_low_u64(b)),
-                                vreinterpret_u32_u64(vget_high_u64(b)));
-  return vget_lane_u32(c, 0);
-}
-static INLINE unsigned int horizontal_add_16x8(const uint16x8_t vec_16x8) {
-  const uint32x4_t a = vpaddlq_u16(vec_16x8);
-  const uint64x2_t b = vpaddlq_u32(a);
-  const uint32x2_t c = vadd_u32(vreinterpret_u32_u64(vget_low_u64(b)),
-                                vreinterpret_u32_u64(vget_high_u64(b)));
-  return vget_lane_u32(c, 0);
-}
-
-unsigned int aom_sad64x64_neon(const uint8_t *src, int src_stride,
-                               const uint8_t *ref, int ref_stride) {
-  int i;
-  uint16x8_t vec_accum_lo = vdupq_n_u16(0);
-  uint16x8_t vec_accum_hi = vdupq_n_u16(0);
-  for (i = 0; i < 64; ++i) {
-    const uint8x16_t vec_src_00 = vld1q_u8(src);
-    const uint8x16_t vec_src_16 = vld1q_u8(src + 16);
-    const uint8x16_t vec_src_32 = vld1q_u8(src + 32);
-    const uint8x16_t vec_src_48 = vld1q_u8(src + 48);
-    const uint8x16_t vec_ref_00 = vld1q_u8(ref);
-    const uint8x16_t vec_ref_16 = vld1q_u8(ref + 16);
-    const uint8x16_t vec_ref_32 = vld1q_u8(ref + 32);
-    const uint8x16_t vec_ref_48 = vld1q_u8(ref + 48);
-    src += src_stride;
-    ref += ref_stride;
-    vec_accum_lo = vabal_u8(vec_accum_lo, vget_low_u8(vec_src_00),
-                            vget_low_u8(vec_ref_00));
-    vec_accum_hi = vabal_u8(vec_accum_hi, vget_high_u8(vec_src_00),
-                            vget_high_u8(vec_ref_00));
-    vec_accum_lo = vabal_u8(vec_accum_lo, vget_low_u8(vec_src_16),
-                            vget_low_u8(vec_ref_16));
-    vec_accum_hi = vabal_u8(vec_accum_hi, vget_high_u8(vec_src_16),
-                            vget_high_u8(vec_ref_16));
-    vec_accum_lo = vabal_u8(vec_accum_lo, vget_low_u8(vec_src_32),
-                            vget_low_u8(vec_ref_32));
-    vec_accum_hi = vabal_u8(vec_accum_hi, vget_high_u8(vec_src_32),
-                            vget_high_u8(vec_ref_32));
-    vec_accum_lo = vabal_u8(vec_accum_lo, vget_low_u8(vec_src_48),
-                            vget_low_u8(vec_ref_48));
-    vec_accum_hi = vabal_u8(vec_accum_hi, vget_high_u8(vec_src_48),
-                            vget_high_u8(vec_ref_48));
-  }
-  return horizontal_long_add_16x8(vec_accum_lo, vec_accum_hi);
-}
-
-unsigned int aom_sad32x32_neon(const uint8_t *src, int src_stride,
-                               const uint8_t *ref, int ref_stride) {
-  int i;
-  uint16x8_t vec_accum_lo = vdupq_n_u16(0);
-  uint16x8_t vec_accum_hi = vdupq_n_u16(0);
-
-  for (i = 0; i < 32; ++i) {
-    const uint8x16_t vec_src_00 = vld1q_u8(src);
-    const uint8x16_t vec_src_16 = vld1q_u8(src + 16);
-    const uint8x16_t vec_ref_00 = vld1q_u8(ref);
-    const uint8x16_t vec_ref_16 = vld1q_u8(ref + 16);
-    src += src_stride;
-    ref += ref_stride;
-    vec_accum_lo = vabal_u8(vec_accum_lo, vget_low_u8(vec_src_00),
-                            vget_low_u8(vec_ref_00));
-    vec_accum_hi = vabal_u8(vec_accum_hi, vget_high_u8(vec_src_00),
-                            vget_high_u8(vec_ref_00));
-    vec_accum_lo = vabal_u8(vec_accum_lo, vget_low_u8(vec_src_16),
-                            vget_low_u8(vec_ref_16));
-    vec_accum_hi = vabal_u8(vec_accum_hi, vget_high_u8(vec_src_16),
-                            vget_high_u8(vec_ref_16));
-  }
-  return horizontal_add_16x8(vaddq_u16(vec_accum_lo, vec_accum_hi));
-}
-
-unsigned int aom_sad16x16_neon(const uint8_t *src, int src_stride,
-                               const uint8_t *ref, int ref_stride) {
-  int i;
-  uint16x8_t vec_accum_lo = vdupq_n_u16(0);
-  uint16x8_t vec_accum_hi = vdupq_n_u16(0);
-
-  for (i = 0; i < 16; ++i) {
-    const uint8x16_t vec_src = vld1q_u8(src);
-    const uint8x16_t vec_ref = vld1q_u8(ref);
-    src += src_stride;
-    ref += ref_stride;
-    vec_accum_lo =
-        vabal_u8(vec_accum_lo, vget_low_u8(vec_src), vget_low_u8(vec_ref));
-    vec_accum_hi =
-        vabal_u8(vec_accum_hi, vget_high_u8(vec_src), vget_high_u8(vec_ref));
-  }
-  return horizontal_add_16x8(vaddq_u16(vec_accum_lo, vec_accum_hi));
-}
-
-unsigned int aom_sad8x8_neon(const uint8_t *src, int src_stride,
-                             const uint8_t *ref, int ref_stride) {
-  int i;
-  uint16x8_t vec_accum = vdupq_n_u16(0);
-
-  for (i = 0; i < 8; ++i) {
-    const uint8x8_t vec_src = vld1_u8(src);
-    const uint8x8_t vec_ref = vld1_u8(ref);
-    src += src_stride;
-    ref += ref_stride;
-    vec_accum = vabal_u8(vec_accum, vec_src, vec_ref);
-  }
-  return horizontal_add_16x8(vec_accum);
-}

aom_dsp/arm/save_reg_neon.asm

@@ -1,39 +0,0 @@
-;
-; Copyright (c) 2016, Alliance for Open Media. All rights reserved
-;
-; This source code is subject to the terms of the BSD 2 Clause License and
-; the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
-; was not distributed with this source code in the LICENSE file, you can
-; obtain it at www.aomedia.org/license/software. If the Alliance for Open
-; Media Patent License 1.0 was not distributed with this source code in the
-; PATENTS file, you can obtain it at www.aomedia.org/license/patent.
-;
-
-;
-
-
-    EXPORT  |aom_push_neon|
-    EXPORT  |aom_pop_neon|
-
-    ARM
-    REQUIRE8
-    PRESERVE8
-
-    AREA ||.text||, CODE, READONLY, ALIGN=2
-
-|aom_push_neon| PROC
-    vst1.i64            {d8, d9, d10, d11}, [r0]!
-    vst1.i64            {d12, d13, d14, d15}, [r0]!
-    bx              lr
-
-    ENDP
-
-|aom_pop_neon| PROC
-    vld1.i64            {d8, d9, d10, d11}, [r0]!
-    vld1.i64            {d12, d13, d14, d15}, [r0]!
-    bx              lr
-
-    ENDP
-
-    END
-

aom_dsp/arm/subpel_variance_media.c

@@ -1,81 +0,0 @@
-/*
- * Copyright (c) 2016, Alliance for Open Media. All rights reserved
- *
- * This source code is subject to the terms of the BSD 2 Clause License and
- * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
- * was not distributed with this source code in the LICENSE file, you can
- * obtain it at www.aomedia.org/license/software. If the Alliance for Open
- * Media Patent License 1.0 was not distributed with this source code in the
- * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
- */
-
-#include "./aom_config.h"
-#include "./aom_dsp_rtcd.h"
-#include "aom/aom_integer.h"
-#include "aom_ports/mem.h"
-
-#if HAVE_MEDIA
-static const int16_t bilinear_filters_media[8][2] = { { 128, 0 }, { 112, 16 },
-                                                      { 96, 32 }, { 80, 48 },
-                                                      { 64, 64 }, { 48, 80 },
-                                                      { 32, 96 }, { 16, 112 } };
-
-extern void aom_filter_block2d_bil_first_pass_media(
-    const uint8_t *src_ptr, uint16_t *dst_ptr, uint32_t src_pitch,
-    uint32_t height, uint32_t width, const int16_t *filter);
-
-extern void aom_filter_block2d_bil_second_pass_media(
-    const uint16_t *src_ptr, uint8_t *dst_ptr, int32_t src_pitch,
-    uint32_t height, uint32_t width, const int16_t *filter);
-
-unsigned int aom_sub_pixel_variance8x8_media(
-    const uint8_t *src_ptr, int src_pixels_per_line, int xoffset, int yoffset,
-    const uint8_t *dst_ptr, int dst_pixels_per_line, unsigned int *sse) {
-  uint16_t first_pass[10 * 8];
-  uint8_t second_pass[8 * 8];
-  const int16_t *HFilter, *VFilter;
-
-  HFilter = bilinear_filters_media[xoffset];
-  VFilter = bilinear_filters_media[yoffset];
-
-  aom_filter_block2d_bil_first_pass_media(src_ptr, first_pass,
-                                          src_pixels_per_line, 9, 8, HFilter);
-  aom_filter_block2d_bil_second_pass_media(first_pass, second_pass, 8, 8, 8,
-                                           VFilter);
-
-  return aom_variance8x8_media(second_pass, 8, dst_ptr, dst_pixels_per_line,
-                               sse);
-}
-
-unsigned int aom_sub_pixel_variance16x16_media(
-    const uint8_t *src_ptr, int src_pixels_per_line, int xoffset, int yoffset,
-    const uint8_t *dst_ptr, int dst_pixels_per_line, unsigned int *sse) {
-  uint16_t first_pass[36 * 16];
-  uint8_t second_pass[20 * 16];
-  const int16_t *HFilter, *VFilter;
-  unsigned int var;
-
-  if (xoffset == 4 && yoffset == 0) {
-    var = aom_variance_halfpixvar16x16_h_media(
-        src_ptr, src_pixels_per_line, dst_ptr, dst_pixels_per_line, sse);
-  } else if (xoffset == 0 && yoffset == 4) {
-    var = aom_variance_halfpixvar16x16_v_media(
-        src_ptr, src_pixels_per_line, dst_ptr, dst_pixels_per_line, sse);
-  } else if (xoffset == 4 && yoffset == 4) {
-    var = aom_variance_halfpixvar16x16_hv_media(
-        src_ptr, src_pixels_per_line, dst_ptr, dst_pixels_per_line, sse);
-  } else {
-    HFilter = bilinear_filters_media[xoffset];
-    VFilter = bilinear_filters_media[yoffset];
-
-    aom_filter_block2d_bil_first_pass_media(
-        src_ptr, first_pass, src_pixels_per_line, 17, 16, HFilter);
-    aom_filter_block2d_bil_second_pass_media(first_pass, second_pass, 16, 16,
-                                             16, VFilter);
-
-    var = aom_variance16x16_media(second_pass, 16, dst_ptr, dst_pixels_per_line,
-                                  sse);
-  }
-  return var;
-}
-#endif  // HAVE_MEDIA

aom_dsp/arm/subpel_variance_neon.c

@@ -1,134 +0,0 @@
-/*
- * Copyright (c) 2016, Alliance for Open Media. All rights reserved
- *
- * This source code is subject to the terms of the BSD 2 Clause License and
- * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
- * was not distributed with this source code in the LICENSE file, you can
- * obtain it at www.aomedia.org/license/software. If the Alliance for Open
- * Media Patent License 1.0 was not distributed with this source code in the
- * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
- */
-
-#include <arm_neon.h>
-#include "./aom_dsp_rtcd.h"
-#include "./aom_config.h"
-
-#include "aom_ports/mem.h"
-#include "aom/aom_integer.h"
-
-#include "aom_dsp/variance.h"
-
-static const uint8_t bilinear_filters[8][2] = {
-  { 128, 0 }, { 112, 16 }, { 96, 32 }, { 80, 48 },
-  { 64, 64 }, { 48, 80 },  { 32, 96 }, { 16, 112 },
-};
-
-static void var_filter_block2d_bil_w8(const uint8_t *src_ptr,
-                                      uint8_t *output_ptr,
-                                      unsigned int src_pixels_per_line,
-                                      int pixel_step,
-                                      unsigned int output_height,
-                                      unsigned int output_width,
-                                      const uint8_t *filter) {
-  const uint8x8_t f0 = vmov_n_u8(filter[0]);
-  const uint8x8_t f1 = vmov_n_u8(filter[1]);
-  unsigned int i;
-  for (i = 0; i < output_height; ++i) {
-    const uint8x8_t src_0 = vld1_u8(&src_ptr[0]);
-    const uint8x8_t src_1 = vld1_u8(&src_ptr[pixel_step]);
-    const uint16x8_t a = vmull_u8(src_0, f0);
-    const uint16x8_t b = vmlal_u8(a, src_1, f1);
-    const uint8x8_t out = vrshrn_n_u16(b, FILTER_BITS);
-    vst1_u8(&output_ptr[0], out);
-    // Next row...
-    src_ptr += src_pixels_per_line;
-    output_ptr += output_width;
-  }
-}
-
-static void var_filter_block2d_bil_w16(const uint8_t *src_ptr,
-                                       uint8_t *output_ptr,
-                                       unsigned int src_pixels_per_line,
-                                       int pixel_step,
-                                       unsigned int output_height,
-                                       unsigned int output_width,
-                                       const uint8_t *filter) {
-  const uint8x8_t f0 = vmov_n_u8(filter[0]);
-  const uint8x8_t f1 = vmov_n_u8(filter[1]);
-  unsigned int i, j;
-  for (i = 0; i < output_height; ++i) {
-    for (j = 0; j < output_width; j += 16) {
-      const uint8x16_t src_0 = vld1q_u8(&src_ptr[j]);
-      const uint8x16_t src_1 = vld1q_u8(&src_ptr[j + pixel_step]);
-      const uint16x8_t a = vmull_u8(vget_low_u8(src_0), f0);
-      const uint16x8_t b = vmlal_u8(a, vget_low_u8(src_1), f1);
-      const uint8x8_t out_lo = vrshrn_n_u16(b, FILTER_BITS);
-      const uint16x8_t c = vmull_u8(vget_high_u8(src_0), f0);
-      const uint16x8_t d = vmlal_u8(c, vget_high_u8(src_1), f1);
-      const uint8x8_t out_hi = vrshrn_n_u16(d, FILTER_BITS);
-      vst1q_u8(&output_ptr[j], vcombine_u8(out_lo, out_hi));
-    }
-    // Next row...
-    src_ptr += src_pixels_per_line;
-    output_ptr += output_width;
-  }
-}
-
-unsigned int aom_sub_pixel_variance8x8_neon(const uint8_t *src, int src_stride,
-                                            int xoffset, int yoffset,
-                                            const uint8_t *dst, int dst_stride,
-                                            unsigned int *sse) {
-  DECLARE_ALIGNED(16, uint8_t, temp2[8 * 8]);
-  DECLARE_ALIGNED(16, uint8_t, fdata3[9 * 8]);
-
-  var_filter_block2d_bil_w8(src, fdata3, src_stride, 1, 9, 8,
-                            bilinear_filters[xoffset]);
-  var_filter_block2d_bil_w8(fdata3, temp2, 8, 8, 8, 8,
-                            bilinear_filters[yoffset]);
-  return aom_variance8x8_neon(temp2, 8, dst, dst_stride, sse);
-}
-
-unsigned int aom_sub_pixel_variance16x16_neon(const uint8_t *src,
-                                              int src_stride, int xoffset,
-                                              int yoffset, const uint8_t *dst,
-                                              int dst_stride,
-                                              unsigned int *sse) {
-  DECLARE_ALIGNED(16, uint8_t, temp2[16 * 16]);
-  DECLARE_ALIGNED(16, uint8_t, fdata3[17 * 16]);
-
-  var_filter_block2d_bil_w16(src, fdata3, src_stride, 1, 17, 16,
-                             bilinear_filters[xoffset]);
-  var_filter_block2d_bil_w16(fdata3, temp2, 16, 16, 16, 16,
-                             bilinear_filters[yoffset]);
-  return aom_variance16x16_neon(temp2, 16, dst, dst_stride, sse);
-}
-
-unsigned int aom_sub_pixel_variance32x32_neon(const uint8_t *src,
-                                              int src_stride, int xoffset,
-                                              int yoffset, const uint8_t *dst,
-                                              int dst_stride,
-                                              unsigned int *sse) {
-  DECLARE_ALIGNED(16, uint8_t, temp2[32 * 32]);
-  DECLARE_ALIGNED(16, uint8_t, fdata3[33 * 32]);
-
-  var_filter_block2d_bil_w16(src, fdata3, src_stride, 1, 33, 32,
-                             bilinear_filters[xoffset]);
-  var_filter_block2d_bil_w16(fdata3, temp2, 32, 32, 32, 32,
-                             bilinear_filters[yoffset]);
-  return aom_variance32x32_neon(temp2, 32, dst, dst_stride, sse);
-}
-
-unsigned int aom_sub_pixel_variance64x64_neon(const uint8_t *src,
-                                              int src_stride, int xoffset,
-                                              int yoffset, const uint8_t *dst,
-                                              int dst_stride,
-                                              unsigned int *sse) {
-  DECLARE_ALIGNED(16, uint8_t, temp2[64 * 64]);
-  DECLARE_ALIGNED(16, uint8_t, fdata3[65 * 64]);
-
-  var_filter_block2d_bil_w16(src, fdata3, src_stride, 1, 65, 64,
-                             bilinear_filters[xoffset]);
-  var_filter_block2d_bil_w16(fdata3, temp2, 64, 64, 64, 64,
-                             bilinear_filters[yoffset]);
-  return aom_variance64x64_neon(temp2, 64, dst, dst_stride, sse);
-}

aom_dsp/arm/subtract_neon.c

@@ -1,80 +0,0 @@
-/*
- * Copyright (c) 2016, Alliance for Open Media. All rights reserved
- *
- * This source code is subject to the terms of the BSD 2 Clause License and
- * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
- * was not distributed with this source code in the LICENSE file, you can
- * obtain it at www.aomedia.org/license/software. If the Alliance for Open
- * Media Patent License 1.0 was not distributed with this source code in the
- * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
- */
-
-#include <arm_neon.h>
-
-#include "./aom_config.h"
-#include "aom/aom_integer.h"
-
-void aom_subtract_block_neon(int rows, int cols, int16_t *diff,
-                             ptrdiff_t diff_stride, const uint8_t *src,
-                             ptrdiff_t src_stride, const uint8_t *pred,
-                             ptrdiff_t pred_stride) {
-  int r, c;
-
-  if (cols > 16) {
-    for (r = 0; r < rows; ++r) {
-      for (c = 0; c < cols; c += 32) {
-        const uint8x16_t v_src_00 = vld1q_u8(&src[c + 0]);
-        const uint8x16_t v_src_16 = vld1q_u8(&src[c + 16]);
-        const uint8x16_t v_pred_00 = vld1q_u8(&pred[c + 0]);
-        const uint8x16_t v_pred_16 = vld1q_u8(&pred[c + 16]);
-        const uint16x8_t v_diff_lo_00 =
-            vsubl_u8(vget_low_u8(v_src_00), vget_low_u8(v_pred_00));
-        const uint16x8_t v_diff_hi_00 =
-            vsubl_u8(vget_high_u8(v_src_00), vget_high_u8(v_pred_00));
-        const uint16x8_t v_diff_lo_16 =
-            vsubl_u8(vget_low_u8(v_src_16), vget_low_u8(v_pred_16));
-        const uint16x8_t v_diff_hi_16 =
-            vsubl_u8(vget_high_u8(v_src_16), vget_high_u8(v_pred_16));
-        vst1q_s16(&diff[c + 0], vreinterpretq_s16_u16(v_diff_lo_00));
-        vst1q_s16(&diff[c + 8], vreinterpretq_s16_u16(v_diff_hi_00));
-        vst1q_s16(&diff[c + 16], vreinterpretq_s16_u16(v_diff_lo_16));
-        vst1q_s16(&diff[c + 24], vreinterpretq_s16_u16(v_diff_hi_16));
-      }
-      diff += diff_stride;
-      pred += pred_stride;
-      src += src_stride;
-    }
-  } else if (cols > 8) {
-    for (r = 0; r < rows; ++r) {
-      const uint8x16_t v_src = vld1q_u8(&src[0]);
-      const uint8x16_t v_pred = vld1q_u8(&pred[0]);
-      const uint16x8_t v_diff_lo =
-          vsubl_u8(vget_low_u8(v_src), vget_low_u8(v_pred));
-      const uint16x8_t v_diff_hi =
-          vsubl_u8(vget_high_u8(v_src), vget_high_u8(v_pred));
-      vst1q_s16(&diff[0], vreinterpretq_s16_u16(v_diff_lo));
-      vst1q_s16(&diff[8], vreinterpretq_s16_u16(v_diff_hi));
-      diff += diff_stride;
-      pred += pred_stride;
-      src += src_stride;
-    }
-  } else if (cols > 4) {
-    for (r = 0; r < rows; ++r) {
-      const uint8x8_t v_src = vld1_u8(&src[0]);
-      const uint8x8_t v_pred = vld1_u8(&pred[0]);
-      const uint16x8_t v_diff = vsubl_u8(v_src, v_pred);
-      vst1q_s16(&diff[0], vreinterpretq_s16_u16(v_diff));
-      diff += diff_stride;
-      pred += pred_stride;
-      src += src_stride;
-    }
-  } else {
-    for (r = 0; r < rows; ++r) {
-      for (c = 0; c < cols; ++c) diff[c] = src[c] - pred[c];
-
-      diff += diff_stride;
-      pred += pred_stride;
-      src += src_stride;
-    }
-  }
-}

aom_dsp/arm/variance_media.asm

@@ -1,361 +0,0 @@
-;
-; Copyright (c) 2016, Alliance for Open Media. All rights reserved
-;
-; This source code is subject to the terms of the BSD 2 Clause License and
-; the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
-; was not distributed with this source code in the LICENSE file, you can
-; obtain it at www.aomedia.org/license/software. If the Alliance for Open
-; Media Patent License 1.0 was not distributed with this source code in the
-; PATENTS file, you can obtain it at www.aomedia.org/license/patent.
-;
-
-;
-
-
-    EXPORT  |aom_variance16x16_media|
-    EXPORT  |aom_variance8x8_media|
-    EXPORT  |aom_mse16x16_media|
-
-    ARM
-    REQUIRE8
-    PRESERVE8
-
-    AREA ||.text||, CODE, READONLY, ALIGN=2
-
-; r0    unsigned char *src_ptr
-; r1    int source_stride
-; r2    unsigned char *ref_ptr
-; r3    int  recon_stride
-; stack unsigned int *sse
-|aom_variance16x16_media| PROC
-
-    stmfd   sp!, {r4-r12, lr}
-
-    pld     [r0, r1, lsl #0]
-    pld     [r2, r3, lsl #0]
-
-    mov     r8, #0              ; initialize sum = 0
-    mov     r11, #0             ; initialize sse = 0
-    mov     r12, #16            ; set loop counter to 16 (=block height)
-
-loop16x16
-    ; 1st 4 pixels
-    ldr     r4, [r0, #0]        ; load 4 src pixels
-    ldr     r5, [r2, #0]        ; load 4 ref pixels
-
-    mov     lr, #0              ; constant zero
-
-    usub8   r6, r4, r5          ; calculate difference
-    pld     [r0, r1, lsl #1]
-    sel     r7, r6, lr          ; select bytes with positive difference
-    usub8   r9, r5, r4          ; calculate difference with reversed operands
-    pld     [r2, r3, lsl #1]
-    sel     r6, r9, lr          ; select bytes with negative difference
-
-    ; calculate partial sums
-    usad8   r4, r7, lr          ; calculate sum of positive differences
-    usad8   r5, r6, lr          ; calculate sum of negative differences
-    orr     r6, r6, r7          ; differences of all 4 pixels
-    ; calculate total sum
-    adds    r8, r8, r4          ; add positive differences to sum
-    subs    r8, r8, r5          ; subtract negative differences from sum
-
-    ; calculate sse
-    uxtb16  r5, r6              ; byte (two pixels) to halfwords
-    uxtb16  r10, r6, ror #8     ; another two pixels to halfwords
-    smlad   r11, r5, r5, r11    ; dual signed multiply, add and accumulate (1)
-
-    ; 2nd 4 pixels
-    ldr     r4, [r0, #4]        ; load 4 src pixels
-    ldr     r5, [r2, #4]        ; load 4 ref pixels
-    smlad   r11, r10, r10, r11  ; dual signed multiply, add and accumulate (2)
-
-    usub8   r6, r4, r5          ; calculate difference
-    sel     r7, r6, lr          ; select bytes with positive difference
-    usub8   r9, r5, r4          ; calculate difference with reversed operands
-    sel     r6, r9, lr          ; select bytes with negative difference
-
-    ; calculate partial sums
-    usad8   r4, r7, lr          ; calculate sum of positive differences
-    usad8   r5, r6, lr          ; calculate sum of negative differences
-    orr     r6, r6, r7          ; differences of all 4 pixels
-
-    ; calculate total sum
-    add     r8, r8, r4          ; add positive differences to sum
-    sub     r8, r8, r5          ; subtract negative differences from sum
-
-    ; calculate sse
-    uxtb16  r5, r6              ; byte (two pixels) to halfwords
-    uxtb16  r10, r6, ror #8     ; another two pixels to halfwords
-    smlad   r11, r5, r5, r11    ; dual signed multiply, add and accumulate (1)
-
-    ; 3rd 4 pixels
-    ldr     r4, [r0, #8]        ; load 4 src pixels
-    ldr     r5, [r2, #8]        ; load 4 ref pixels
-    smlad   r11, r10, r10, r11  ; dual signed multiply, add and accumulate (2)
-
-    usub8   r6, r4, r5          ; calculate difference
-    sel     r7, r6, lr          ; select bytes with positive difference
-    usub8   r9, r5, r4          ; calculate difference with reversed operands
-    sel     r6, r9, lr          ; select bytes with negative difference
-
-    ; calculate partial sums
-    usad8   r4, r7, lr          ; calculate sum of positive differences
-    usad8   r5, r6, lr          ; calculate sum of negative differences
-    orr     r6, r6, r7          ; differences of all 4 pixels
-
-    ; calculate total sum
-    add     r8, r8, r4          ; add positive differences to sum
-    sub     r8, r8, r5          ; subtract negative differences from sum
-
-    ; calculate sse
-    uxtb16  r5, r6              ; byte (two pixels) to halfwords
-    uxtb16  r10, r6, ror #8     ; another two pixels to halfwords
-    smlad   r11, r5, r5, r11    ; dual signed multiply, add and accumulate (1)
-
-    ; 4th 4 pixels
-    ldr     r4, [r0, #12]       ; load 4 src pixels
-    ldr     r5, [r2, #12]       ; load 4 ref pixels
-    smlad   r11, r10, r10, r11  ; dual signed multiply, add and accumulate (2)
-
-    usub8   r6, r4, r5          ; calculate difference
-    add     r0, r0, r1          ; set src_ptr to next row
-    sel     r7, r6, lr          ; select bytes with positive difference
-    usub8   r9, r5, r4          ; calculate difference with reversed operands
-    add     r2, r2, r3          ; set dst_ptr to next row
-    sel     r6, r9, lr          ; select bytes with negative difference
-
-    ; calculate partial sums
-    usad8   r4, r7, lr          ; calculate sum of positive differences
-    usad8   r5, r6, lr          ; calculate sum of negative differences
-    orr     r6, r6, r7          ; differences of all 4 pixels
-
-    ; calculate total sum
-    add     r8, r8, r4          ; add positive differences to sum
-    sub     r8, r8, r5          ; subtract negative differences from sum
-
-    ; calculate sse
-    uxtb16  r5, r6              ; byte (two pixels) to halfwords
-    uxtb16  r10, r6, ror #8     ; another two pixels to halfwords
-    smlad   r11, r5, r5, r11    ; dual signed multiply, add and accumulate (1)
-    smlad   r11, r10, r10, r11  ; dual signed multiply, add and accumulate (2)
-
-
-    subs    r12, r12, #1
-
-    bne     loop16x16
-
-    ; return stuff
-    ldr     r6, [sp, #40]       ; get address of sse
-    mul     r0, r8, r8          ; sum * sum
-    str     r11, [r6]           ; store sse
-    sub     r0, r11, r0, lsr #8 ; return (sse - ((sum * sum) >> 8))
-
-    ldmfd   sp!, {r4-r12, pc}
-
-    ENDP
-
-; r0    unsigned char *src_ptr
-; r1    int source_stride
-; r2    unsigned char *ref_ptr
-; r3    int  recon_stride
-; stack unsigned int *sse
-|aom_variance8x8_media| PROC
-
-    push    {r4-r10, lr}
-
-    pld     [r0, r1, lsl #0]
-    pld     [r2, r3, lsl #0]
-
-    mov     r12, #8             ; set loop counter to 8 (=block height)
-    mov     r4, #0              ; initialize sum = 0
-    mov     r5, #0              ; initialize sse = 0
-
-loop8x8
-    ; 1st 4 pixels
-    ldr     r6, [r0, #0x0]      ; load 4 src pixels
-    ldr     r7, [r2, #0x0]      ; load 4 ref pixels
-
-    mov     lr, #0              ; constant zero
-
-    usub8   r8, r6, r7          ; calculate difference
-    pld     [r0, r1, lsl #1]
-    sel     r10, r8, lr         ; select bytes with positive difference
-    usub8   r9, r7, r6          ; calculate difference with reversed operands
-    pld     [r2, r3, lsl #1]
-    sel     r8, r9, lr          ; select bytes with negative difference
-
-    ; calculate partial sums
-    usad8   r6, r10, lr         ; calculate sum of positive differences
-    usad8   r7, r8, lr          ; calculate sum of negative differences
-    orr     r8, r8, r10         ; differences of all 4 pixels
-    ; calculate total sum
-    add    r4, r4, r6           ; add positive differences to sum
-    sub    r4, r4, r7           ; subtract negative differences from sum
-
-    ; calculate sse
-    uxtb16  r7, r8              ; byte (two pixels) to halfwords
-    uxtb16  r10, r8, ror #8     ; another two pixels to halfwords
-    smlad   r5, r7, r7, r5      ; dual signed multiply, add and accumulate (1)
-
-    ; 2nd 4 pixels
-    ldr     r6, [r0, #0x4]      ; load 4 src pixels
-    ldr     r7, [r2, #0x4]      ; load 4 ref pixels
-    smlad   r5, r10, r10, r5    ; dual signed multiply, add and accumulate (2)
-
-    usub8   r8, r6, r7          ; calculate difference
-    add     r0, r0, r1          ; set src_ptr to next row
-    sel     r10, r8, lr         ; select bytes with positive difference
-    usub8   r9, r7, r6          ; calculate difference with reversed operands
-    add     r2, r2, r3          ; set dst_ptr to next row
-    sel     r8, r9, lr          ; select bytes with negative difference
-
-    ; calculate partial sums
-    usad8   r6, r10, lr         ; calculate sum of positive differences
-    usad8   r7, r8, lr          ; calculate sum of negative differences
-    orr     r8, r8, r10         ; differences of all 4 pixels
-
-    ; calculate total sum
-    add     r4, r4, r6          ; add positive differences to sum
-    sub     r4, r4, r7          ; subtract negative differences from sum
-
-    ; calculate sse
-    uxtb16  r7, r8              ; byte (two pixels) to halfwords
-    uxtb16  r10, r8, ror #8     ; another two pixels to halfwords
-    smlad   r5, r7, r7, r5      ; dual signed multiply, add and accumulate (1)
-    subs    r12, r12, #1        ; next row
-    smlad   r5, r10, r10, r5    ; dual signed multiply, add and accumulate (2)
-
-    bne     loop8x8
-
-    ; return stuff
-    ldr     r8, [sp, #32]       ; get address of sse
-    mul     r1, r4, r4          ; sum * sum
-    str     r5, [r8]            ; store sse
-    sub     r0, r5, r1, ASR #6  ; return (sse - ((sum * sum) >> 6))
-
-    pop     {r4-r10, pc}
-
-    ENDP
-
-; r0    unsigned char *src_ptr
-; r1    int source_stride
-; r2    unsigned char *ref_ptr
-; r3    int  recon_stride
-; stack unsigned int *sse
-;
-;note: Based on aom_variance16x16_media. In this function, sum is never used.
-;      So, we can remove this part of calculation.
-
-|aom_mse16x16_media| PROC
-
-    push    {r4-r9, lr}
-
-    pld     [r0, r1, lsl #0]
-    pld     [r2, r3, lsl #0]
-
-    mov     r12, #16            ; set loop counter to 16 (=block height)
-    mov     r4, #0              ; initialize sse = 0
-
-loopmse
-    ; 1st 4 pixels
-    ldr     r5, [r0, #0x0]      ; load 4 src pixels
-    ldr     r6, [r2, #0x0]      ; load 4 ref pixels
-
-    mov     lr, #0              ; constant zero
-
-    usub8   r8, r5, r6          ; calculate difference
-    pld     [r0, r1, lsl #1]
-    sel     r7, r8, lr          ; select bytes with positive difference
-    usub8   r9, r6, r5          ; calculate difference with reversed operands
-    pld     [r2, r3, lsl #1]
-    sel     r8, r9, lr          ; select bytes with negative difference
-
-    ; calculate partial sums
-    usad8   r5, r7, lr          ; calculate sum of positive differences
-    usad8   r6, r8, lr          ; calculate sum of negative differences
-    orr     r8, r8, r7          ; differences of all 4 pixels
-
-    ldr     r5, [r0, #0x4]      ; load 4 src pixels
-
-    ; calculate sse
-    uxtb16  r6, r8              ; byte (two pixels) to halfwords
-    uxtb16  r7, r8, ror #8      ; another two pixels to halfwords
-    smlad   r4, r6, r6, r4      ; dual signed multiply, add and accumulate (1)
-
-    ; 2nd 4 pixels
-    ldr     r6, [r2, #0x4]      ; load 4 ref pixels
-    smlad   r4, r7, r7, r4      ; dual signed multiply, add and accumulate (2)
-
-    usub8   r8, r5, r6          ; calculate difference
-    sel     r7, r8, lr          ; select bytes with positive difference
-    usub8   r9, r6, r5          ; calculate difference with reversed operands
-    sel     r8, r9, lr          ; select bytes with negative difference
-
-    ; calculate partial sums
-    usad8   r5, r7, lr          ; calculate sum of positive differences
-    usad8   r6, r8, lr          ; calculate sum of negative differences
-    orr     r8, r8, r7          ; differences of all 4 pixels
-    ldr     r5, [r0, #0x8]      ; load 4 src pixels
-    ; calculate sse
-    uxtb16  r6, r8              ; byte (two pixels) to halfwords
-    uxtb16  r7, r8, ror #8      ; another two pixels to halfwords
-    smlad   r4, r6, r6, r4      ; dual signed multiply, add and accumulate (1)
-
-    ; 3rd 4 pixels
-    ldr     r6, [r2, #0x8]      ; load 4 ref pixels
-    smlad   r4, r7, r7, r4      ; dual signed multiply, add and accumulate (2)
-
-    usub8   r8, r5, r6          ; calculate difference
-    sel     r7, r8, lr          ; select bytes with positive difference
-    usub8   r9, r6, r5          ; calculate difference with reversed operands
-    sel     r8, r9, lr          ; select bytes with negative difference
-
-    ; calculate partial sums
-    usad8   r5, r7, lr          ; calculate sum of positive differences
-    usad8   r6, r8, lr          ; calculate sum of negative differences
-    orr     r8, r8, r7          ; differences of all 4 pixels
-
-    ldr     r5, [r0, #0xc]      ; load 4 src pixels
-
-    ; calculate sse
-    uxtb16  r6, r8              ; byte (two pixels) to halfwords
-    uxtb16  r7, r8, ror #8      ; another two pixels to halfwords
-    smlad   r4, r6, r6, r4      ; dual signed multiply, add and accumulate (1)
-
-    ; 4th 4 pixels
-    ldr     r6, [r2, #0xc]      ; load 4 ref pixels
-    smlad   r4, r7, r7, r4      ; dual signed multiply, add and accumulate (2)
-
-    usub8   r8, r5, r6          ; calculate difference
-    add     r0, r0, r1          ; set src_ptr to next row
-    sel     r7, r8, lr          ; select bytes with positive difference
-    usub8   r9, r6, r5          ; calculate difference with reversed operands
-    add     r2, r2, r3          ; set dst_ptr to next row
-    sel     r8, r9, lr          ; select bytes with negative difference
-
-    ; calculate partial sums
-    usad8   r5, r7, lr          ; calculate sum of positive differences
-    usad8   r6, r8, lr          ; calculate sum of negative differences
-    orr     r8, r8, r7          ; differences of all 4 pixels
-
-    subs    r12, r12, #1        ; next row
-
-    ; calculate sse
-    uxtb16  r6, r8              ; byte (two pixels) to halfwords
-    uxtb16  r7, r8, ror #8      ; another two pixels to halfwords
-    smlad   r4, r6, r6, r4      ; dual signed multiply, add and accumulate (1)
-    smlad   r4, r7, r7, r4      ; dual signed multiply, add and accumulate (2)
-
-    bne     loopmse
-
-    ; return stuff
-    ldr     r1, [sp, #28]       ; get address of sse
-    mov     r0, r4              ; return sse
-    str     r4, [r1]            ; store sse
-
-    pop     {r4-r9, pc}
-
-    ENDP
-
-    END

aom_dsp/arm/variance_neon.c

@@ -1,400 +0,0 @@
-/*
- * Copyright (c) 2016, Alliance for Open Media. All rights reserved
- *
- * This source code is subject to the terms of the BSD 2 Clause License and
- * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
- * was not distributed with this source code in the LICENSE file, you can
- * obtain it at www.aomedia.org/license/software. If the Alliance for Open
- * Media Patent License 1.0 was not distributed with this source code in the
- * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
- */
-
-#include <arm_neon.h>
-
-#include "./aom_dsp_rtcd.h"
-#include "./aom_config.h"
-
-#include "aom/aom_integer.h"
-#include "aom_ports/mem.h"
-
-static INLINE int horizontal_add_s16x8(const int16x8_t v_16x8) {
-  const int32x4_t a = vpaddlq_s16(v_16x8);
-  const int64x2_t b = vpaddlq_s32(a);
-  const int32x2_t c = vadd_s32(vreinterpret_s32_s64(vget_low_s64(b)),
-                               vreinterpret_s32_s64(vget_high_s64(b)));
-  return vget_lane_s32(c, 0);
-}
-
-static INLINE int horizontal_add_s32x4(const int32x4_t v_32x4) {
-  const int64x2_t b = vpaddlq_s32(v_32x4);
-  const int32x2_t c = vadd_s32(vreinterpret_s32_s64(vget_low_s64(b)),
-                               vreinterpret_s32_s64(vget_high_s64(b)));
-  return vget_lane_s32(c, 0);
-}
-
-// w * h must be less than 2048 or local variable v_sum may overflow.
-static void variance_neon_w8(const uint8_t *a, int a_stride, const uint8_t *b,
-                             int b_stride, int w, int h, uint32_t *sse,
-                             int *sum) {
-  int i, j;
-  int16x8_t v_sum = vdupq_n_s16(0);
-  int32x4_t v_sse_lo = vdupq_n_s32(0);
-  int32x4_t v_sse_hi = vdupq_n_s32(0);
-
-  for (i = 0; i < h; ++i) {
-    for (j = 0; j < w; j += 8) {
-      const uint8x8_t v_a = vld1_u8(&a[j]);
-      const uint8x8_t v_b = vld1_u8(&b[j]);
-      const uint16x8_t v_diff = vsubl_u8(v_a, v_b);
-      const int16x8_t sv_diff = vreinterpretq_s16_u16(v_diff);
-      v_sum = vaddq_s16(v_sum, sv_diff);
-      v_sse_lo =
-          vmlal_s16(v_sse_lo, vget_low_s16(sv_diff), vget_low_s16(sv_diff));
-      v_sse_hi =
-          vmlal_s16(v_sse_hi, vget_high_s16(sv_diff), vget_high_s16(sv_diff));
-    }
-    a += a_stride;
-    b += b_stride;
-  }
-
-  *sum = horizontal_add_s16x8(v_sum);
-  *sse = (unsigned int)horizontal_add_s32x4(vaddq_s32(v_sse_lo, v_sse_hi));
-}
-
-void aom_get8x8var_neon(const uint8_t *a, int a_stride, const uint8_t *b,
-                        int b_stride, unsigned int *sse, int *sum) {
-  variance_neon_w8(a, a_stride, b, b_stride, 8, 8, sse, sum);
-}
-
-void aom_get16x16var_neon(const uint8_t *a, int a_stride, const uint8_t *b,
-                          int b_stride, unsigned int *sse, int *sum) {
-  variance_neon_w8(a, a_stride, b, b_stride, 16, 16, sse, sum);
-}
-
-unsigned int aom_variance8x8_neon(const uint8_t *a, int a_stride,
-                                  const uint8_t *b, int b_stride,
-                                  unsigned int *sse) {
-  int sum;
-  variance_neon_w8(a, a_stride, b, b_stride, 8, 8, sse, &sum);
-  return *sse - (((int64_t)sum * sum) >> 6);  //  >> 6 = / 8 * 8
-}
-
-unsigned int aom_variance16x16_neon(const uint8_t *a, int a_stride,
-                                    const uint8_t *b, int b_stride,
-                                    unsigned int *sse) {
-  int sum;
-  variance_neon_w8(a, a_stride, b, b_stride, 16, 16, sse, &sum);
-  return *sse - (((int64_t)sum * sum) >> 8);  //  >> 8 = / 16 * 16
-}
-
-unsigned int aom_variance32x32_neon(const uint8_t *a, int a_stride,
-                                    const uint8_t *b, int b_stride,
-                                    unsigned int *sse) {
-  int sum;
-  variance_neon_w8(a, a_stride, b, b_stride, 32, 32, sse, &sum);
-  return *sse - (((int64_t)sum * sum) >> 10);  // >> 10 = / 32 * 32
-}
-
-unsigned int aom_variance32x64_neon(const uint8_t *a, int a_stride,
-                                    const uint8_t *b, int b_stride,
-                                    unsigned int *sse) {
-  int sum1, sum2;
-  uint32_t sse1, sse2;
-  variance_neon_w8(a, a_stride, b, b_stride, 32, 32, &sse1, &sum1);
-  variance_neon_w8(a + (32 * a_stride), a_stride, b + (32 * b_stride), b_stride,
-                   32, 32, &sse2, &sum2);
-  *sse = sse1 + sse2;
-  sum1 += sum2;
-  return *sse - (((int64_t)sum1 * sum1) >> 11);  // >> 11 = / 32 * 64
-}
-
-unsigned int aom_variance64x32_neon(const uint8_t *a, int a_stride,
-                                    const uint8_t *b, int b_stride,
-                                    unsigned int *sse) {
-  int sum1, sum2;
-  uint32_t sse1, sse2;
-  variance_neon_w8(a, a_stride, b, b_stride, 64, 16, &sse1, &sum1);
-  variance_neon_w8(a + (16 * a_stride), a_stride, b + (16 * b_stride), b_stride,
-                   64, 16, &sse2, &sum2);
-  *sse = sse1 + sse2;
-  sum1 += sum2;
-  return *sse - (((int64_t)sum1 * sum1) >> 11);  // >> 11 = / 32 * 64
-}
-
-unsigned int aom_variance64x64_neon(const uint8_t *a, int a_stride,
-                                    const uint8_t *b, int b_stride,
-                                    unsigned int *sse) {
-  int sum1, sum2;
-  uint32_t sse1, sse2;
-
-  variance_neon_w8(a, a_stride, b, b_stride, 64, 16, &sse1, &sum1);
-  variance_neon_w8(a + (16 * a_stride), a_stride, b + (16 * b_stride), b_stride,
-                   64, 16, &sse2, &sum2);
-  sse1 += sse2;
-  sum1 += sum2;
-
-  variance_neon_w8(a + (16 * 2 * a_stride), a_stride, b + (16 * 2 * b_stride),
-                   b_stride, 64, 16, &sse2, &sum2);
-  sse1 += sse2;
-  sum1 += sum2;
-
-  variance_neon_w8(a + (16 * 3 * a_stride), a_stride, b + (16 * 3 * b_stride),
-                   b_stride, 64, 16, &sse2, &sum2);
-  *sse = sse1 + sse2;
-  sum1 += sum2;
-  return *sse - (((int64_t)sum1 * sum1) >> 12);  // >> 12 = / 64 * 64
-}
-
-unsigned int aom_variance16x8_neon(const unsigned char *src_ptr,
-                                   int source_stride,
-                                   const unsigned char *ref_ptr,
-                                   int recon_stride, unsigned int *sse) {
-  int i;
-  int16x4_t d22s16, d23s16, d24s16, d25s16, d26s16, d27s16, d28s16, d29s16;
-  uint32x2_t d0u32, d10u32;
-  int64x1_t d0s64, d1s64;
-  uint8x16_t q0u8, q1u8, q2u8, q3u8;
-  uint16x8_t q11u16, q12u16, q13u16, q14u16;
-  int32x4_t q8s32, q9s32, q10s32;
-  int64x2_t q0s64, q1s64, q5s64;
-
-  q8s32 = vdupq_n_s32(0);
-  q9s32 = vdupq_n_s32(0);
-  q10s32 = vdupq_n_s32(0);
-
-  for (i = 0; i < 4; i++) {
-    q0u8 = vld1q_u8(src_ptr);
-    src_ptr += source_stride;
-    q1u8 = vld1q_u8(src_ptr);
-    src_ptr += source_stride;
-    __builtin_prefetch(src_ptr);
-
-    q2u8 = vld1q_u8(ref_ptr);
-    ref_ptr += recon_stride;
-    q3u8 = vld1q_u8(ref_ptr);
-    ref_ptr += recon_stride;
-    __builtin_prefetch(ref_ptr);
-
-    q11u16 = vsubl_u8(vget_low_u8(q0u8), vget_low_u8(q2u8));
-    q12u16 = vsubl_u8(vget_high_u8(q0u8), vget_high_u8(q2u8));
-    q13u16 = vsubl_u8(vget_low_u8(q1u8), vget_low_u8(q3u8));
-    q14u16 = vsubl_u8(vget_high_u8(q1u8), vget_high_u8(q3u8));
-
-    d22s16 = vreinterpret_s16_u16(vget_low_u16(q11u16));
-    d23s16 = vreinterpret_s16_u16(vget_high_u16(q11u16));
-    q8s32 = vpadalq_s16(q8s32, vreinterpretq_s16_u16(q11u16));
-    q9s32 = vmlal_s16(q9s32, d22s16, d22s16);
-    q10s32 = vmlal_s16(q10s32, d23s16, d23s16);
-
-    d24s16 = vreinterpret_s16_u16(vget_low_u16(q12u16));
-    d25s16 = vreinterpret_s16_u16(vget_high_u16(q12u16));
-    q8s32 = vpadalq_s16(q8s32, vreinterpretq_s16_u16(q12u16));
-    q9s32 = vmlal_s16(q9s32, d24s16, d24s16);
-    q10s32 = vmlal_s16(q10s32, d25s16, d25s16);
-
-    d26s16 = vreinterpret_s16_u16(vget_low_u16(q13u16));
-    d27s16 = vreinterpret_s16_u16(vget_high_u16(q13u16));
-    q8s32 = vpadalq_s16(q8s32, vreinterpretq_s16_u16(q13u16));
-    q9s32 = vmlal_s16(q9s32, d26s16, d26s16);
-    q10s32 = vmlal_s16(q10s32, d27s16, d27s16);
-
-    d28s16 = vreinterpret_s16_u16(vget_low_u16(q14u16));
-    d29s16 = vreinterpret_s16_u16(vget_high_u16(q14u16));
-    q8s32 = vpadalq_s16(q8s32, vreinterpretq_s16_u16(q14u16));
-    q9s32 = vmlal_s16(q9s32, d28s16, d28s16);
-    q10s32 = vmlal_s16(q10s32, d29s16, d29s16);
-  }
-
-  q10s32 = vaddq_s32(q10s32, q9s32);
-  q0s64 = vpaddlq_s32(q8s32);
-  q1s64 = vpaddlq_s32(q10s32);
-
-  d0s64 = vadd_s64(vget_low_s64(q0s64), vget_high_s64(q0s64));
-  d1s64 = vadd_s64(vget_low_s64(q1s64), vget_high_s64(q1s64));
-
-  q5s64 = vmull_s32(vreinterpret_s32_s64(d0s64), vreinterpret_s32_s64(d0s64));
-  vst1_lane_u32((uint32_t *)sse, vreinterpret_u32_s64(d1s64), 0);
-
-  d10u32 = vshr_n_u32(vreinterpret_u32_s64(vget_low_s64(q5s64)), 7);
-  d0u32 = vsub_u32(vreinterpret_u32_s64(d1s64), d10u32);
-
-  return vget_lane_u32(d0u32, 0);
-}
-
-unsigned int aom_variance8x16_neon(const unsigned char *src_ptr,
-                                   int source_stride,
-                                   const unsigned char *ref_ptr,
-                                   int recon_stride, unsigned int *sse) {
-  int i;
-  uint8x8_t d0u8, d2u8, d4u8, d6u8;
-  int16x4_t d22s16, d23s16, d24s16, d25s16;
-  uint32x2_t d0u32, d10u32;
-  int64x1_t d0s64, d1s64;
-  uint16x8_t q11u16, q12u16;
-  int32x4_t q8s32, q9s32, q10s32;
-  int64x2_t q0s64, q1s64, q5s64;
-
-  q8s32 = vdupq_n_s32(0);
-  q9s32 = vdupq_n_s32(0);
-  q10s32 = vdupq_n_s32(0);
-
-  for (i = 0; i < 8; i++) {
-    d0u8 = vld1_u8(src_ptr);
-    src_ptr += source_stride;
-    d2u8 = vld1_u8(src_ptr);
-    src_ptr += source_stride;
-    __builtin_prefetch(src_ptr);
-
-    d4u8 = vld1_u8(ref_ptr);
-    ref_ptr += recon_stride;
-    d6u8 = vld1_u8(ref_ptr);
-    ref_ptr += recon_stride;
-    __builtin_prefetch(ref_ptr);
-
-    q11u16 = vsubl_u8(d0u8, d4u8);
-    q12u16 = vsubl_u8(d2u8, d6u8);
-
-    d22s16 = vreinterpret_s16_u16(vget_low_u16(q11u16));
-    d23s16 = vreinterpret_s16_u16(vget_high_u16(q11u16));
-    q8s32 = vpadalq_s16(q8s32, vreinterpretq_s16_u16(q11u16));
-    q9s32 = vmlal_s16(q9s32, d22s16, d22s16);
-    q10s32 = vmlal_s16(q10s32, d23s16, d23s16);
-
-    d24s16 = vreinterpret_s16_u16(vget_low_u16(q12u16));
-    d25s16 = vreinterpret_s16_u16(vget_high_u16(q12u16));
-    q8s32 = vpadalq_s16(q8s32, vreinterpretq_s16_u16(q12u16));
-    q9s32 = vmlal_s16(q9s32, d24s16, d24s16);
-    q10s32 = vmlal_s16(q10s32, d25s16, d25s16);
-  }
-
-  q10s32 = vaddq_s32(q10s32, q9s32);
-  q0s64 = vpaddlq_s32(q8s32);
-  q1s64 = vpaddlq_s32(q10s32);
-
-  d0s64 = vadd_s64(vget_low_s64(q0s64), vget_high_s64(q0s64));
-  d1s64 = vadd_s64(vget_low_s64(q1s64), vget_high_s64(q1s64));
-
-  q5s64 = vmull_s32(vreinterpret_s32_s64(d0s64), vreinterpret_s32_s64(d0s64));
-  vst1_lane_u32((uint32_t *)sse, vreinterpret_u32_s64(d1s64), 0);
-
-  d10u32 = vshr_n_u32(vreinterpret_u32_s64(vget_low_s64(q5s64)), 7);
-  d0u32 = vsub_u32(vreinterpret_u32_s64(d1s64), d10u32);
-
-  return vget_lane_u32(d0u32, 0);
-}
-
-unsigned int aom_mse16x16_neon(const unsigned char *src_ptr, int source_stride,
-                               const unsigned char *ref_ptr, int recon_stride,
-                               unsigned int *sse) {
-  int i;
-  int16x4_t d22s16, d23s16, d24s16, d25s16, d26s16, d27s16, d28s16, d29s16;
-  int64x1_t d0s64;
-  uint8x16_t q0u8, q1u8, q2u8, q3u8;
-  int32x4_t q7s32, q8s32, q9s32, q10s32;
-  uint16x8_t q11u16, q12u16, q13u16, q14u16;
-  int64x2_t q1s64;
-
-  q7s32 = vdupq_n_s32(0);
-  q8s32 = vdupq_n_s32(0);
-  q9s32 = vdupq_n_s32(0);
-  q10s32 = vdupq_n_s32(0);
-
-  for (i = 0; i < 8; i++) {  // mse16x16_neon_loop
-    q0u8 = vld1q_u8(src_ptr);
-    src_ptr += source_stride;
-    q1u8 = vld1q_u8(src_ptr);
-    src_ptr += source_stride;
-    q2u8 = vld1q_u8(ref_ptr);
-    ref_ptr += recon_stride;
-    q3u8 = vld1q_u8(ref_ptr);
-    ref_ptr += recon_stride;
-
-    q11u16 = vsubl_u8(vget_low_u8(q0u8), vget_low_u8(q2u8));
-    q12u16 = vsubl_u8(vget_high_u8(q0u8), vget_high_u8(q2u8));
-    q13u16 = vsubl_u8(vget_low_u8(q1u8), vget_low_u8(q3u8));
-    q14u16 = vsubl_u8(vget_high_u8(q1u8), vget_high_u8(q3u8));
-
-    d22s16 = vreinterpret_s16_u16(vget_low_u16(q11u16));
-    d23s16 = vreinterpret_s16_u16(vget_high_u16(q11u16));
-    q7s32 = vmlal_s16(q7s32, d22s16, d22s16);
-    q8s32 = vmlal_s16(q8s32, d23s16, d23s16);
-
-    d24s16 = vreinterpret_s16_u16(vget_low_u16(q12u16));
-    d25s16 = vreinterpret_s16_u16(vget_high_u16(q12u16));
-    q9s32 = vmlal_s16(q9s32, d24s16, d24s16);
-    q10s32 = vmlal_s16(q10s32, d25s16, d25s16);
-
-    d26s16 = vreinterpret_s16_u16(vget_low_u16(q13u16));
-    d27s16 = vreinterpret_s16_u16(vget_high_u16(q13u16));
-    q7s32 = vmlal_s16(q7s32, d26s16, d26s16);
-    q8s32 = vmlal_s16(q8s32, d27s16, d27s16);
-
-    d28s16 = vreinterpret_s16_u16(vget_low_u16(q14u16));
-    d29s16 = vreinterpret_s16_u16(vget_high_u16(q14u16));
-    q9s32 = vmlal_s16(q9s32, d28s16, d28s16);
-    q10s32 = vmlal_s16(q10s32, d29s16, d29s16);
-  }
-
-  q7s32 = vaddq_s32(q7s32, q8s32);
-  q9s32 = vaddq_s32(q9s32, q10s32);
-  q10s32 = vaddq_s32(q7s32, q9s32);
-
-  q1s64 = vpaddlq_s32(q10s32);
-  d0s64 = vadd_s64(vget_low_s64(q1s64), vget_high_s64(q1s64));
-
-  vst1_lane_u32((uint32_t *)sse, vreinterpret_u32_s64(d0s64), 0);
-  return vget_lane_u32(vreinterpret_u32_s64(d0s64), 0);
-}
-
-unsigned int aom_get4x4sse_cs_neon(const unsigned char *src_ptr,
-                                   int source_stride,
-                                   const unsigned char *ref_ptr,
-                                   int recon_stride) {
-  int16x4_t d22s16, d24s16, d26s16, d28s16;
-  int64x1_t d0s64;
-  uint8x8_t d0u8, d1u8, d2u8, d3u8, d4u8, d5u8, d6u8, d7u8;
-  int32x4_t q7s32, q8s32, q9s32, q10s32;
-  uint16x8_t q11u16, q12u16, q13u16, q14u16;
-  int64x2_t q1s64;
-
-  d0u8 = vld1_u8(src_ptr);
-  src_ptr += source_stride;
-  d4u8 = vld1_u8(ref_ptr);
-  ref_ptr += recon_stride;
-  d1u8 = vld1_u8(src_ptr);
-  src_ptr += source_stride;
-  d5u8 = vld1_u8(ref_ptr);
-  ref_ptr += recon_stride;
-  d2u8 = vld1_u8(src_ptr);
-  src_ptr += source_stride;
-  d6u8 = vld1_u8(ref_ptr);
-  ref_ptr += recon_stride;
-  d3u8 = vld1_u8(src_ptr);
-  src_ptr += source_stride;
-  d7u8 = vld1_u8(ref_ptr);
-  ref_ptr += recon_stride;
-
-  q11u16 = vsubl_u8(d0u8, d4u8);
-  q12u16 = vsubl_u8(d1u8, d5u8);
-  q13u16 = vsubl_u8(d2u8, d6u8);
-  q14u16 = vsubl_u8(d3u8, d7u8);
-
-  d22s16 = vget_low_s16(vreinterpretq_s16_u16(q11u16));
-  d24s16 = vget_low_s16(vreinterpretq_s16_u16(q12u16));
-  d26s16 = vget_low_s16(vreinterpretq_s16_u16(q13u16));
-  d28s16 = vget_low_s16(vreinterpretq_s16_u16(q14u16));
-
-  q7s32 = vmull_s16(d22s16, d22s16);
-  q8s32 = vmull_s16(d24s16, d24s16);
-  q9s32 = vmull_s16(d26s16, d26s16);
-  q10s32 = vmull_s16(d28s16, d28s16);
-
-  q7s32 = vaddq_s32(q7s32, q8s32);
-  q9s32 = vaddq_s32(q9s32, q10s32);
-  q9s32 = vaddq_s32(q7s32, q9s32);
-
-  q1s64 = vpaddlq_s32(q9s32);
-  d0s64 = vadd_s64(vget_low_s64(q1s64), vget_high_s64(q1s64));
-
-  return vget_lane_u32(vreinterpret_u32_s64(d0s64), 0);
-}

aom_dsp/avg.c

@@ -1,232 +0,0 @@
-/*
- * Copyright (c) 2016, Alliance for Open Media. All rights reserved
- *
- * This source code is subject to the terms of the BSD 2 Clause License and
- * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
- * was not distributed with this source code in the LICENSE file, you can
- * obtain it at www.aomedia.org/license/software. If the Alliance for Open
- * Media Patent License 1.0 was not distributed with this source code in the
- * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
- */
-#include <stdlib.h>
-
-#include "./aom_dsp_rtcd.h"
-#include "aom_ports/mem.h"
-
-unsigned int aom_avg_8x8_c(const uint8_t *src, int stride) {
-  int i, j;
-  int sum = 0;
-  for (i = 0; i < 8; ++i, src += stride)
-    for (j = 0; j < 8; sum += src[j], ++j) {
-    }
-
-  return ROUND_POWER_OF_TWO(sum, 6);
-}
-
-unsigned int aom_avg_4x4_c(const uint8_t *src, int stride) {
-  int i, j;
-  int sum = 0;
-  for (i = 0; i < 4; ++i, src += stride)
-    for (j = 0; j < 4; sum += src[j], ++j) {
-    }
-
-  return ROUND_POWER_OF_TWO(sum, 4);
-}
-
-// src_diff: first pass, 9 bit, dynamic range [-255, 255]
-//           second pass, 12 bit, dynamic range [-2040, 2040]
-static void hadamard_col8(const int16_t *src_diff, int src_stride,
-                          int16_t *coeff) {
-  int16_t b0 = src_diff[0 * src_stride] + src_diff[1 * src_stride];
-  int16_t b1 = src_diff[0 * src_stride] - src_diff[1 * src_stride];
-  int16_t b2 = src_diff[2 * src_stride] + src_diff[3 * src_stride];
-  int16_t b3 = src_diff[2 * src_stride] - src_diff[3 * src_stride];
-  int16_t b4 = src_diff[4 * src_stride] + src_diff[5 * src_stride];
-  int16_t b5 = src_diff[4 * src_stride] - src_diff[5 * src_stride];
-  int16_t b6 = src_diff[6 * src_stride] + src_diff[7 * src_stride];
-  int16_t b7 = src_diff[6 * src_stride] - src_diff[7 * src_stride];
-
-  int16_t c0 = b0 + b2;
-  int16_t c1 = b1 + b3;
-  int16_t c2 = b0 - b2;
-  int16_t c3 = b1 - b3;
-  int16_t c4 = b4 + b6;
-  int16_t c5 = b5 + b7;
-  int16_t c6 = b4 - b6;
-  int16_t c7 = b5 - b7;
-
-  coeff[0] = c0 + c4;
-  coeff[7] = c1 + c5;
-  coeff[3] = c2 + c6;
-  coeff[4] = c3 + c7;
-  coeff[2] = c0 - c4;
-  coeff[6] = c1 - c5;
-  coeff[1] = c2 - c6;
-  coeff[5] = c3 - c7;
-}
-
-// The order of the output coeff of the hadamard is not important. For
-// optimization purposes the final transpose may be skipped.
-void aom_hadamard_8x8_c(const int16_t *src_diff, int src_stride,
-                        int16_t *coeff) {
-  int idx;
-  int16_t buffer[64];
-  int16_t *tmp_buf = &buffer[0];
-  for (idx = 0; idx < 8; ++idx) {
-    hadamard_col8(src_diff, src_stride, tmp_buf);  // src_diff: 9 bit
-                                                   // dynamic range [-255, 255]
-    tmp_buf += 8;
-    ++src_diff;
-  }
-
-  tmp_buf = &buffer[0];
-  for (idx = 0; idx < 8; ++idx) {
-    hadamard_col8(tmp_buf, 8, coeff);  // tmp_buf: 12 bit
-                                       // dynamic range [-2040, 2040]
-    coeff += 8;                        // coeff: 15 bit
-                                       // dynamic range [-16320, 16320]
-    ++tmp_buf;
-  }
-}
-
-// In place 16x16 2D Hadamard transform
-void aom_hadamard_16x16_c(const int16_t *src_diff, int src_stride,
-                          int16_t *coeff) {
-  int idx;
-  for (idx = 0; idx < 4; ++idx) {
-    // src_diff: 9 bit, dynamic range [-255, 255]
-    const int16_t *src_ptr =
-        src_diff + (idx >> 1) * 8 * src_stride + (idx & 0x01) * 8;
-    aom_hadamard_8x8_c(src_ptr, src_stride, coeff + idx * 64);
-  }
-
-  // coeff: 15 bit, dynamic range [-16320, 16320]
-  for (idx = 0; idx < 64; ++idx) {
-    int16_t a0 = coeff[0];
-    int16_t a1 = coeff[64];
-    int16_t a2 = coeff[128];
-    int16_t a3 = coeff[192];
-
-    int16_t b0 = (a0 + a1) >> 1;  // (a0 + a1): 16 bit, [-32640, 32640]
-    int16_t b1 = (a0 - a1) >> 1;  // b0-b3: 15 bit, dynamic range
-    int16_t b2 = (a2 + a3) >> 1;  // [-16320, 16320]
-    int16_t b3 = (a2 - a3) >> 1;
-
-    coeff[0] = b0 + b2;  // 16 bit, [-32640, 32640]
-    coeff[64] = b1 + b3;
-    coeff[128] = b0 - b2;
-    coeff[192] = b1 - b3;
-
-    ++coeff;
-  }
-}
-
-// coeff: 16 bits, dynamic range [-32640, 32640].
-// length: value range {16, 64, 256, 1024}.
-int aom_satd_c(const int16_t *coeff, int length) {
-  int i;
-  int satd = 0;
-  for (i = 0; i < length; ++i) satd += abs(coeff[i]);
-
-  // satd: 26 bits, dynamic range [-32640 * 1024, 32640 * 1024]
-  return satd;
-}
-
-// Integer projection onto row vectors.
-// height: value range {16, 32, 64}.
-void aom_int_pro_row_c(int16_t hbuf[16], const uint8_t *ref,
-                       const int ref_stride, const int height) {
-  int idx;
-  const int norm_factor = height >> 1;
-  for (idx = 0; idx < 16; ++idx) {
-    int i;
-    hbuf[idx] = 0;
-    // hbuf[idx]: 14 bit, dynamic range [0, 16320].
-    for (i = 0; i < height; ++i) hbuf[idx] += ref[i * ref_stride];
-    // hbuf[idx]: 9 bit, dynamic range [0, 510].
-    hbuf[idx] /= norm_factor;
-    ++ref;
-  }
-}
-
-// width: value range {16, 32, 64}.
-int16_t aom_int_pro_col_c(const uint8_t *ref, const int width) {
-  int idx;
-  int16_t sum = 0;
-  // sum: 14 bit, dynamic range [0, 16320]
-  for (idx = 0; idx < width; ++idx) sum += ref[idx];
-  return sum;
-}
-
-// ref: [0 - 510]
-// src: [0 - 510]
-// bwl: {2, 3, 4}
-int aom_vector_var_c(const int16_t *ref, const int16_t *src, const int bwl) {
-  int i;
-  int width = 4 << bwl;
-  int sse = 0, mean = 0, var;
-
-  for (i = 0; i < width; ++i) {
-    int diff = ref[i] - src[i];  // diff: dynamic range [-510, 510], 10 bits.
-    mean += diff;                // mean: dynamic range 16 bits.
-    sse += diff * diff;          // sse:  dynamic range 26 bits.
-  }
-
-  // (mean * mean): dynamic range 31 bits.
-  var = sse - ((mean * mean) >> (bwl + 2));
-  return var;
-}
-
-void aom_minmax_8x8_c(const uint8_t *src, int src_stride, const uint8_t *ref,
-                      int ref_stride, int *min, int *max) {
-  int i, j;
-  *min = 255;
-  *max = 0;
-  for (i = 0; i < 8; ++i, src += src_stride, ref += ref_stride) {
-    for (j = 0; j < 8; ++j) {
-      int diff = abs(src[j] - ref[j]);
-      *min = diff < *min ? diff : *min;
-      *max = diff > *max ? diff : *max;
-    }
-  }
-}
-
-#if CONFIG_AOM_HIGHBITDEPTH
-unsigned int aom_highbd_avg_8x8_c(const uint8_t *src, int stride) {
-  int i, j;
-  int sum = 0;
-  const uint16_t *s = CONVERT_TO_SHORTPTR(src);
-  for (i = 0; i < 8; ++i, s += stride)
-    for (j = 0; j < 8; sum += s[j], ++j) {
-    }
-
-  return ROUND_POWER_OF_TWO(sum, 6);
-}
-
-unsigned int aom_highbd_avg_4x4_c(const uint8_t *src, int stride) {
-  int i, j;
-  int sum = 0;
-  const uint16_t *s = CONVERT_TO_SHORTPTR(src);
-  for (i = 0; i < 4; ++i, s += stride)
-    for (j = 0; j < 4; sum += s[j], ++j) {
-    }
-
-  return ROUND_POWER_OF_TWO(sum, 4);
-}
-
-void aom_highbd_minmax_8x8_c(const uint8_t *s8, int p, const uint8_t *d8,
-                             int dp, int *min, int *max) {
-  int i, j;
-  const uint16_t *s = CONVERT_TO_SHORTPTR(s8);
-  const uint16_t *d = CONVERT_TO_SHORTPTR(d8);
-  *min = 255;
-  *max = 0;
-  for (i = 0; i < 8; ++i, s += p, d += dp) {
-    for (j = 0; j < 8; ++j) {
-      int diff = abs(s[j] - d[j]);
-      *min = diff < *min ? diff : *min;
-      *max = diff > *max ? diff : *max;
-    }
-  }
-}
-#endif  // CONFIG_AOM_HIGHBITDEPTH

aom_dsp/bitreader.h

@@ -1,240 +0,0 @@
-/*
- * Copyright (c) 2016, Alliance for Open Media. All rights reserved
- *
- * This source code is subject to the terms of the BSD 2 Clause License and
- * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
- * was not distributed with this source code in the LICENSE file, you can
- * obtain it at www.aomedia.org/license/software. If the Alliance for Open
- * Media Patent License 1.0 was not distributed with this source code in the
- * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
- */
-
-#ifndef AOM_DSP_BITREADER_H_
-#define AOM_DSP_BITREADER_H_
-
-#include <assert.h>
-#include <limits.h>
-
-#include "./aom_config.h"
-#if CONFIG_EC_ADAPT && !CONFIG_EC_MULTISYMBOL
-#error "CONFIG_EC_ADAPT is enabled without enabling CONFIG_EC_MULTISYMBOL."
-#endif
-
-#include "aom/aomdx.h"
-#include "aom/aom_integer.h"
-#if CONFIG_ANS
-#include "aom_dsp/ansreader.h"
-#elif CONFIG_DAALA_EC
-#include "aom_dsp/daalaboolreader.h"
-#else
-#include "aom_dsp/dkboolreader.h"
-#endif
-#include "aom_dsp/prob.h"
-#include "av1/common/odintrin.h"
-
-#if CONFIG_ACCOUNTING
-#include "av1/common/accounting.h"
-#define ACCT_STR_NAME acct_str
-#define ACCT_STR_PARAM , const char *ACCT_STR_NAME
-#define ACCT_STR_ARG(s) , s
-#else
-#define ACCT_STR_PARAM
-#define ACCT_STR_ARG(s)
-#endif
-
-#define aom_read(r, prob, ACCT_STR_NAME) \
-  aom_read_(r, prob ACCT_STR_ARG(ACCT_STR_NAME))
-#define aom_read_bit(r, ACCT_STR_NAME) \
-  aom_read_bit_(r ACCT_STR_ARG(ACCT_STR_NAME))
-#define aom_read_tree(r, tree, probs, ACCT_STR_NAME) \
-  aom_read_tree_(r, tree, probs ACCT_STR_ARG(ACCT_STR_NAME))
-#define aom_read_literal(r, bits, ACCT_STR_NAME) \
-  aom_read_literal_(r, bits ACCT_STR_ARG(ACCT_STR_NAME))
-#define aom_read_tree_bits(r, tree, probs, ACCT_STR_NAME) \
-  aom_read_tree_bits_(r, tree, probs ACCT_STR_ARG(ACCT_STR_NAME))
-#define aom_read_symbol(r, cdf, nsymbs, ACCT_STR_NAME) \
-  aom_read_symbol_(r, cdf, nsymbs ACCT_STR_ARG(ACCT_STR_NAME))
-
-#ifdef __cplusplus
-extern "C" {
-#endif
-
-#if CONFIG_ANS
-typedef struct AnsDecoder aom_reader;
-#elif CONFIG_DAALA_EC
-typedef struct daala_reader aom_reader;
-#else
-typedef struct aom_dk_reader aom_reader;
-#endif
-
-static INLINE int aom_reader_init(aom_reader *r, const uint8_t *buffer,
-                                  size_t size, aom_decrypt_cb decrypt_cb,
-                                  void *decrypt_state) {
-#if CONFIG_ANS
-  (void)decrypt_cb;
-  (void)decrypt_state;
-  assert(size <= INT_MAX);
-  return ans_read_init(r, buffer, size);
-#elif CONFIG_DAALA_EC
-  (void)decrypt_cb;
-  (void)decrypt_state;
-  return aom_daala_reader_init(r, buffer, size);
-#else
-  return aom_dk_reader_init(r, buffer, size, decrypt_cb, decrypt_state);
-#endif
-}
-
-static INLINE const uint8_t *aom_reader_find_end(aom_reader *r) {
-#if CONFIG_ANS
-  (void)r;
-  assert(0 && "Use the raw buffer size with ANS");
-  return NULL;
-#elif CONFIG_DAALA_EC
-  return aom_daala_reader_find_end(r);
-#else
-  return aom_dk_reader_find_end(r);
-#endif
-}
-
-static INLINE int aom_reader_has_error(aom_reader *r) {
-#if CONFIG_ANS
-  return ans_reader_has_error(r);
-#elif CONFIG_DAALA_EC
-  return aom_daala_reader_has_error(r);
-#else
-  return aom_dk_reader_has_error(r);
-#endif
-}
-
-// Returns the position in the bit reader in bits.
-static INLINE uint32_t aom_reader_tell(const aom_reader *r) {
-#if CONFIG_ANS
-  (void)r;
-  assert(0 && "aom_reader_tell() is unimplemented for ANS");
-  return 0;
-#elif CONFIG_DAALA_EC
-  return aom_daala_reader_tell(r);
-#else
-  return aom_dk_reader_tell(r);
-#endif
-}
-
-// Returns the position in the bit reader in 1/8th bits.
-static INLINE uint32_t aom_reader_tell_frac(const aom_reader *r) {
-#if CONFIG_ANS
-  (void)r;
-  assert(0 && "aom_reader_tell_frac() is unimplemented for ANS");
-  return 0;
-#elif CONFIG_DAALA_EC
-  return aom_daala_reader_tell_frac(r);
-#else
-  return aom_dk_reader_tell_frac(r);
-#endif
-}
-
-#if CONFIG_ACCOUNTING
-static INLINE void aom_process_accounting(const aom_reader *r ACCT_STR_PARAM) {
-  if (r->accounting != NULL) {
-    uint32_t tell_frac;
-    tell_frac = aom_reader_tell_frac(r);
-    aom_accounting_record(r->accounting, ACCT_STR_NAME,
-                          tell_frac - r->accounting->last_tell_frac);
-    r->accounting->last_tell_frac = tell_frac;
-  }
-}
-#endif
-
-static INLINE int aom_read_(aom_reader *r, int prob ACCT_STR_PARAM) {
-  int ret;
-#if CONFIG_ANS
-  ret = uabs_read(r, prob);
-#elif CONFIG_DAALA_EC
-  ret = aom_daala_read(r, prob);
-#else
-  ret = aom_dk_read(r, prob);
-#endif
-#if CONFIG_ACCOUNTING
-  if (ACCT_STR_NAME) aom_process_accounting(r, ACCT_STR_NAME);
-#endif
-  return ret;
-}
-
-static INLINE int aom_read_bit_(aom_reader *r ACCT_STR_PARAM) {
-  int ret;
-#if CONFIG_ANS
-  ret = uabs_read_bit(r);  // Non trivial optimization at half probability
-#else
-  ret = aom_read(r, 128, NULL);  // aom_prob_half
-#endif
-#if CONFIG_ACCOUNTING
-  if (ACCT_STR_NAME) aom_process_accounting(r, ACCT_STR_NAME);
-#endif
-  return ret;
-}
-
-static INLINE int aom_read_literal_(aom_reader *r, int bits ACCT_STR_PARAM) {
-  int literal = 0, bit;
-
-  for (bit = bits - 1; bit >= 0; bit--) literal |= aom_read_bit(r, NULL) << bit;
-#if CONFIG_ACCOUNTING
-  if (ACCT_STR_NAME) aom_process_accounting(r, ACCT_STR_NAME);
-#endif
-  return literal;
-}
-
-static INLINE int aom_read_tree_bits_(aom_reader *r, const aom_tree_index *tree,
-                                      const aom_prob *probs ACCT_STR_PARAM) {
-  aom_tree_index i = 0;
-
-  while ((i = tree[i + aom_read(r, probs[i >> 1], NULL)]) > 0) continue;
-#if CONFIG_ACCOUNTING
-  if (ACCT_STR_NAME) aom_process_accounting(r, ACCT_STR_NAME);
-#endif
-  return -i;
-}
-
-static INLINE int aom_read_tree_(aom_reader *r, const aom_tree_index *tree,
-                                 const aom_prob *probs ACCT_STR_PARAM) {
-  int ret;
-#if CONFIG_DAALA_EC
-  ret = daala_read_tree_bits(r, tree, probs);
-#else
-  ret = aom_read_tree_bits(r, tree, probs, NULL);
-#endif
-#if CONFIG_ACCOUNTING
-  if (ACCT_STR_NAME) aom_process_accounting(r, ACCT_STR_NAME);
-#endif
-  return ret;
-}
-
-#if CONFIG_EC_MULTISYMBOL
-static INLINE int aom_read_symbol_(aom_reader *r, aom_cdf_prob *cdf,
-                                   int nsymbs ACCT_STR_PARAM) {
-  int ret;
-#if CONFIG_RANS
-  (void)nsymbs;
-  ret = rans_read(r, cdf);
-#elif CONFIG_DAALA_EC
-  ret = daala_read_symbol(r, cdf, nsymbs);
-#else
-#error \
-    "CONFIG_EC_MULTISYMBOL is selected without a valid backing entropy " \
-  "coder. Enable daala_ec or ans for a valid configuration."
-#endif
-
-#if CONFIG_EC_ADAPT
-  update_cdf(cdf, ret, nsymbs);
-#endif
-
-#if CONFIG_ACCOUNTING
-  if (ACCT_STR_NAME) aom_process_accounting(r, ACCT_STR_NAME);
-#endif
-  return ret;
-}
-#endif  // CONFIG_EC_MULTISYMBOL
-
-#ifdef __cplusplus
-}  // extern "C"
-#endif
-
-#endif  // AOM_DSP_BITREADER_H_

aom_dsp/bitreader_buffer.c

@@ -1,47 +0,0 @@
-/*
- * Copyright (c) 2016, Alliance for Open Media. All rights reserved
- *
- * This source code is subject to the terms of the BSD 2 Clause License and
- * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
- * was not distributed with this source code in the LICENSE file, you can
- * obtain it at www.aomedia.org/license/software. If the Alliance for Open
- * Media Patent License 1.0 was not distributed with this source code in the
- * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
- */
-#include "./aom_config.h"
-#include "./bitreader_buffer.h"
-
-size_t aom_rb_bytes_read(struct aom_read_bit_buffer *rb) {
-  return (rb->bit_offset + 7) >> 3;
-}
-
-int aom_rb_read_bit(struct aom_read_bit_buffer *rb) {
-  const size_t off = rb->bit_offset;
-  const size_t p = off >> 3;
-  const int q = 7 - (int)(off & 0x7);
-  if (rb->bit_buffer + p < rb->bit_buffer_end) {
-    const int bit = (rb->bit_buffer[p] >> q) & 1;
-    rb->bit_offset = off + 1;
-    return bit;
-  } else {
-    rb->error_handler(rb->error_handler_data);
-    return 0;
-  }
-}
-
-int aom_rb_read_literal(struct aom_read_bit_buffer *rb, int bits) {
-  int value = 0, bit;
-  for (bit = bits - 1; bit >= 0; bit--) value |= aom_rb_read_bit(rb) << bit;
-  return value;
-}
-
-int aom_rb_read_signed_literal(struct aom_read_bit_buffer *rb, int bits) {
-  const int value = aom_rb_read_literal(rb, bits);
-  return aom_rb_read_bit(rb) ? -value : value;
-}
-
-int aom_rb_read_inv_signed_literal(struct aom_read_bit_buffer *rb, int bits) {
-  const int nbits = sizeof(unsigned) * 8 - bits - 1;
-  const unsigned value = (unsigned)aom_rb_read_literal(rb, bits + 1) << nbits;
-  return ((int)value) >> nbits;
-}

aom_dsp/bitreader_buffer.h

@@ -1,48 +0,0 @@
-/*
- * Copyright (c) 2016, Alliance for Open Media. All rights reserved
- *
- * This source code is subject to the terms of the BSD 2 Clause License and
- * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
- * was not distributed with this source code in the LICENSE file, you can
- * obtain it at www.aomedia.org/license/software. If the Alliance for Open
- * Media Patent License 1.0 was not distributed with this source code in the
- * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
- */
-
-#ifndef AOM_DSP_BITREADER_BUFFER_H_
-#define AOM_DSP_BITREADER_BUFFER_H_
-
-#include <limits.h>
-
-#include "aom/aom_integer.h"
-
-#ifdef __cplusplus
-extern "C" {
-#endif
-
-typedef void (*aom_rb_error_handler)(void *data);
-
-struct aom_read_bit_buffer {
-  const uint8_t *bit_buffer;
-  const uint8_t *bit_buffer_end;
-  size_t bit_offset;
-
-  void *error_handler_data;
-  aom_rb_error_handler error_handler;
-};
-
-size_t aom_rb_bytes_read(struct aom_read_bit_buffer *rb);
-
-int aom_rb_read_bit(struct aom_read_bit_buffer *rb);
-
-int aom_rb_read_literal(struct aom_read_bit_buffer *rb, int bits);
-
-int aom_rb_read_signed_literal(struct aom_read_bit_buffer *rb, int bits);
-
-int aom_rb_read_inv_signed_literal(struct aom_read_bit_buffer *rb, int bits);
-
-#ifdef __cplusplus
-}  // extern "C"
-#endif
-
-#endif  // AOM_DSP_BITREADER_BUFFER_H_

aom_dsp/bitwriter.h

@@ -1,179 +0,0 @@
-/*
- * Copyright (c) 2016, Alliance for Open Media. All rights reserved
- *
- * This source code is subject to the terms of the BSD 2 Clause License and
- * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
- * was not distributed with this source code in the LICENSE file, you can
- * obtain it at www.aomedia.org/license/software. If the Alliance for Open
- * Media Patent License 1.0 was not distributed with this source code in the
- * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
- */
-
-#ifndef AOM_DSP_BITWRITER_H_
-#define AOM_DSP_BITWRITER_H_
-
-#include <assert.h>
-#include "./aom_config.h"
-#if CONFIG_EC_ADAPT && !CONFIG_EC_MULTISYMBOL
-#error "CONFIG_EC_ADAPT is enabled without enabling CONFIG_EC_MULTISYMBOL"
-#endif
-
-#if CONFIG_ANS
-#include "aom_dsp/buf_ans.h"
-#elif CONFIG_DAALA_EC
-#include "aom_dsp/daalaboolwriter.h"
-#else
-#include "aom_dsp/dkboolwriter.h"
-#endif
-#include "aom_dsp/prob.h"
-
-#if CONFIG_RD_DEBUG
-#include "av1/encoder/cost.h"
-#endif
-
-#ifdef __cplusplus
-extern "C" {
-#endif
-
-#if CONFIG_ANS
-typedef struct BufAnsCoder aom_writer;
-#elif CONFIG_DAALA_EC
-typedef struct daala_writer aom_writer;
-#else
-typedef struct aom_dk_writer aom_writer;
-#endif
-
-typedef struct TOKEN_STATS { int64_t cost; } TOKEN_STATS;
-
-static INLINE void aom_start_encode(aom_writer *bc, uint8_t *buffer) {
-#if CONFIG_ANS
-  (void)bc;
-  (void)buffer;
-  assert(0 && "buf_ans requires a more complicated startup procedure");
-#elif CONFIG_DAALA_EC
-  aom_daala_start_encode(bc, buffer);
-#else
-  aom_dk_start_encode(bc, buffer);
-#endif
-}
-
-static INLINE void aom_stop_encode(aom_writer *bc) {
-#if CONFIG_ANS
-  (void)bc;
-  assert(0 && "buf_ans requires a more complicated shutdown procedure");
-#elif CONFIG_DAALA_EC
-  aom_daala_stop_encode(bc);
-#else
-  aom_dk_stop_encode(bc);
-#endif
-}
-
-static INLINE void aom_write(aom_writer *br, int bit, int probability) {
-#if CONFIG_ANS
-  buf_uabs_write(br, bit, probability);
-#elif CONFIG_DAALA_EC
-  aom_daala_write(br, bit, probability);
-#else
-  aom_dk_write(br, bit, probability);
-#endif
-}
-
-static INLINE void aom_write_record(aom_writer *br, int bit, int probability,
-                                    TOKEN_STATS *token_stats) {
-  aom_write(br, bit, probability);
-#if CONFIG_RD_DEBUG
-  token_stats->cost += av1_cost_bit(probability, bit);
-#else
-  (void)token_stats;
-#endif
-}
-
-static INLINE void aom_write_bit(aom_writer *w, int bit) {
-  aom_write(w, bit, 128);  // aom_prob_half
-}
-
-static INLINE void aom_write_bit_record(aom_writer *w, int bit,
-                                        TOKEN_STATS *token_stats) {
-  aom_write_record(w, bit, 128, token_stats);  // aom_prob_half
-}
-
-static INLINE void aom_write_literal(aom_writer *w, int data, int bits) {
-  int bit;
-
-  for (bit = bits - 1; bit >= 0; bit--) aom_write_bit(w, 1 & (data >> bit));
-}
-
-static INLINE void aom_write_tree_bits(aom_writer *w, const aom_tree_index *tr,
-                                       const aom_prob *probs, int bits, int len,
-                                       aom_tree_index i) {
-  do {
-    const int bit = (bits >> --len) & 1;
-    aom_write(w, bit, probs[i >> 1]);
-    i = tr[i + bit];
-  } while (len);
-}
-
-static INLINE void aom_write_tree_bits_record(aom_writer *w,
-                                              const aom_tree_index *tr,
-                                              const aom_prob *probs, int bits,
-                                              int len, aom_tree_index i,
-                                              TOKEN_STATS *token_stats) {
-  do {
-    const int bit = (bits >> --len) & 1;
-    aom_write_record(w, bit, probs[i >> 1], token_stats);
-    i = tr[i + bit];
-  } while (len);
-}
-
-static INLINE void aom_write_tree(aom_writer *w, const aom_tree_index *tree,
-                                  const aom_prob *probs, int bits, int len,
-                                  aom_tree_index i) {
-#if CONFIG_DAALA_EC
-  daala_write_tree_bits(w, tree, probs, bits, len, i);
-#else
-  aom_write_tree_bits(w, tree, probs, bits, len, i);
-#endif
-}
-
-static INLINE void aom_write_tree_record(aom_writer *w,
-                                         const aom_tree_index *tree,
-                                         const aom_prob *probs, int bits,
-                                         int len, aom_tree_index i,
-                                         TOKEN_STATS *token_stats) {
-#if CONFIG_DAALA_EC
-  (void)token_stats;
-  daala_write_tree_bits(w, tree, probs, bits, len, i);
-#else
-  aom_write_tree_bits_record(w, tree, probs, bits, len, i, token_stats);
-#endif
-}
-
-#if CONFIG_EC_MULTISYMBOL
-static INLINE void aom_write_symbol(aom_writer *w, int symb, aom_cdf_prob *cdf,
-                                    int nsymbs) {
-#if CONFIG_RANS
-  struct rans_sym s;
-  (void)nsymbs;
-  assert(cdf);
-  s.cum_prob = symb > 0 ? cdf[symb - 1] : 0;
-  s.prob = cdf[symb] - s.cum_prob;
-  buf_rans_write(w, &s);
-#elif CONFIG_DAALA_EC
-  daala_write_symbol(w, symb, cdf, nsymbs);
-#else
-#error \
-    "CONFIG_EC_MULTISYMBOL is selected without a valid backing entropy " \
-  "coder. Enable daala_ec or ans for a valid configuration."
-#endif
-
-#if CONFIG_EC_ADAPT
-  update_cdf(cdf, symb, nsymbs);
-#endif
-}
-#endif  // CONFIG_EC_MULTISYMBOL
-
-#ifdef __cplusplus
-}  // extern "C"
-#endif
-
-#endif  // AOM_DSP_BITWRITER_H_

aom_dsp/bitwriter_buffer.c

@@ -1,43 +0,0 @@
-/*
- * Copyright (c) 2016, Alliance for Open Media. All rights reserved
- *
- * This source code is subject to the terms of the BSD 2 Clause License and
- * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
- * was not distributed with this source code in the LICENSE file, you can
- * obtain it at www.aomedia.org/license/software. If the Alliance for Open
- * Media Patent License 1.0 was not distributed with this source code in the
- * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
- */
-
-#include <limits.h>
-#include <stdlib.h>
-
-#include "./aom_config.h"
-#include "./bitwriter_buffer.h"
-
-size_t aom_wb_bytes_written(const struct aom_write_bit_buffer *wb) {
-  return wb->bit_offset / CHAR_BIT + (wb->bit_offset % CHAR_BIT > 0);
-}
-
-void aom_wb_write_bit(struct aom_write_bit_buffer *wb, int bit) {
-  const int off = (int)wb->bit_offset;
-  const int p = off / CHAR_BIT;
-  const int q = CHAR_BIT - 1 - off % CHAR_BIT;
-  if (q == CHAR_BIT - 1) {
-    wb->bit_buffer[p] = bit << q;
-  } else {
-    wb->bit_buffer[p] &= ~(1 << q);
-    wb->bit_buffer[p] |= bit << q;
-  }
-  wb->bit_offset = off + 1;
-}
-
-void aom_wb_write_literal(struct aom_write_bit_buffer *wb, int data, int bits) {
-  int bit;
-  for (bit = bits - 1; bit >= 0; bit--) aom_wb_write_bit(wb, (data >> bit) & 1);
-}
-
-void aom_wb_write_inv_signed_literal(struct aom_write_bit_buffer *wb, int data,
-                                     int bits) {
-  aom_wb_write_literal(wb, data, bits + 1);
-}

aom_dsp/bitwriter_buffer.h

@@ -1,39 +0,0 @@
-/*
- * Copyright (c) 2016, Alliance for Open Media. All rights reserved
- *
- * This source code is subject to the terms of the BSD 2 Clause License and
- * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
- * was not distributed with this source code in the LICENSE file, you can
- * obtain it at www.aomedia.org/license/software. If the Alliance for Open
- * Media Patent License 1.0 was not distributed with this source code in the
- * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
- */
-
-#ifndef AOM_DSP_BITWRITER_BUFFER_H_
-#define AOM_DSP_BITWRITER_BUFFER_H_
-
-#include "aom/aom_integer.h"
-
-#ifdef __cplusplus
-extern "C" {
-#endif
-
-struct aom_write_bit_buffer {
-  uint8_t *bit_buffer;
-  size_t bit_offset;
-};
-
-size_t aom_wb_bytes_written(const struct aom_write_bit_buffer *wb);
-
-void aom_wb_write_bit(struct aom_write_bit_buffer *wb, int bit);
-
-void aom_wb_write_literal(struct aom_write_bit_buffer *wb, int data, int bits);
-
-void aom_wb_write_inv_signed_literal(struct aom_write_bit_buffer *wb, int data,
-                                     int bits);
-
-#ifdef __cplusplus
-}  // extern "C"
-#endif
-
-#endif  // AOM_DSP_BITWRITER_BUFFER_H_

aom_dsp/blend.h

@@ -1,42 +0,0 @@
-/*
- * Copyright (c) 2016, Alliance for Open Media. All rights reserved
- *
- * This source code is subject to the terms of the BSD 2 Clause License and
- * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
- * was not distributed with this source code in the LICENSE file, you can
- * obtain it at www.aomedia.org/license/software. If the Alliance for Open
- * Media Patent License 1.0 was not distributed with this source code in the
- * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
- */
-
-#ifndef AOM_DSP_BLEND_H_
-#define AOM_DSP_BLEND_H_
-
-#include "aom_ports/mem.h"
-
-// Various blending functions and macros.
-// See also the aom_blend_* functions in aom_dsp_rtcd.h
-
-// Alpha blending with alpha values from the range [0, 64], where 64
-// means use the first input and 0 means use the second input.
-
-#define AOM_BLEND_A64_ROUND_BITS 6
-#define AOM_BLEND_A64_MAX_ALPHA (1 << AOM_BLEND_A64_ROUND_BITS)  // 64
-
-#define AOM_BLEND_A64(a, v0, v1)                                          \
-  ROUND_POWER_OF_TWO((a) * (v0) + (AOM_BLEND_A64_MAX_ALPHA - (a)) * (v1), \
-                     AOM_BLEND_A64_ROUND_BITS)
-
-// Alpha blending with alpha values from the range [0, 256], where 256
-// means use the first input and 0 means use the second input.
-#define AOM_BLEND_A256_ROUND_BITS 8
-#define AOM_BLEND_A256_MAX_ALPHA (1 << AOM_BLEND_A256_ROUND_BITS)  // 256
-
-#define AOM_BLEND_A256(a, v0, v1)                                          \
-  ROUND_POWER_OF_TWO((a) * (v0) + (AOM_BLEND_A256_MAX_ALPHA - (a)) * (v1), \
-                     AOM_BLEND_A256_ROUND_BITS)
-
-// Blending by averaging.
-#define AOM_BLEND_AVG(v0, v1) ROUND_POWER_OF_TWO((v0) + (v1), 1)
-
-#endif  // AOM_DSP_BLEND_H_

aom_dsp/blend_a64_hmask.c

@@ -1,71 +0,0 @@
-/*
- * Copyright (c) 2016, Alliance for Open Media. All rights reserved
- *
- * This source code is subject to the terms of the BSD 2 Clause License and
- * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
- * was not distributed with this source code in the LICENSE file, you can
- * obtain it at www.aomedia.org/license/software. If the Alliance for Open
- * Media Patent License 1.0 was not distributed with this source code in the
- * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
- */
-
-#include <assert.h>
-
-#include "aom/aom_integer.h"
-#include "aom_ports/mem.h"
-#include "aom_dsp/aom_dsp_common.h"
-#include "aom_dsp/blend.h"
-
-#include "./aom_dsp_rtcd.h"
-
-void aom_blend_a64_hmask_c(uint8_t *dst, uint32_t dst_stride,
-                           const uint8_t *src0, uint32_t src0_stride,
-                           const uint8_t *src1, uint32_t src1_stride,
-                           const uint8_t *mask, int h, int w) {
-  int i, j;
-
-  assert(IMPLIES(src0 == dst, src0_stride == dst_stride));
-  assert(IMPLIES(src1 == dst, src1_stride == dst_stride));
-
-  assert(h >= 1);
-  assert(w >= 1);
-  assert(IS_POWER_OF_TWO(h));
-  assert(IS_POWER_OF_TWO(w));
-
-  for (i = 0; i < h; ++i) {
-    for (j = 0; j < w; ++j) {
-      dst[i * dst_stride + j] = AOM_BLEND_A64(
-          mask[j], src0[i * src0_stride + j], src1[i * src1_stride + j]);
-    }
-  }
-}
-
-#if CONFIG_AOM_HIGHBITDEPTH
-void aom_highbd_blend_a64_hmask_c(uint8_t *dst_8, uint32_t dst_stride,
-                                  const uint8_t *src0_8, uint32_t src0_stride,
-                                  const uint8_t *src1_8, uint32_t src1_stride,
-                                  const uint8_t *mask, int h, int w, int bd) {
-  int i, j;
-  uint16_t *dst = CONVERT_TO_SHORTPTR(dst_8);
-  const uint16_t *src0 = CONVERT_TO_SHORTPTR(src0_8);
-  const uint16_t *src1 = CONVERT_TO_SHORTPTR(src1_8);
-  (void)bd;
-
-  assert(IMPLIES(src0 == dst, src0_stride == dst_stride));
-  assert(IMPLIES(src1 == dst, src1_stride == dst_stride));
-
-  assert(h >= 1);
-  assert(w >= 1);
-  assert(IS_POWER_OF_TWO(h));
-  assert(IS_POWER_OF_TWO(w));
-
-  assert(bd == 8 || bd == 10 || bd == 12);
-
-  for (i = 0; i < h; ++i) {
-    for (j = 0; j < w; ++j) {
-      dst[i * dst_stride + j] = AOM_BLEND_A64(
-          mask[j], src0[i * src0_stride + j], src1[i * src1_stride + j]);
-    }
-  }
-}
-#endif  // CONFIG_AOM_HIGHBITDEPTH

aom_dsp/blend_a64_mask.c

@@ -1,145 +0,0 @@
-/*
- * Copyright (c) 2016, Alliance for Open Media. All rights reserved
- *
- * This source code is subject to the terms of the BSD 2 Clause License and
- * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
- * was not distributed with this source code in the LICENSE file, you can
- * obtain it at www.aomedia.org/license/software. If the Alliance for Open
- * Media Patent License 1.0 was not distributed with this source code in the
- * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
- */
-
-#include <assert.h>
-
-#include "aom/aom_integer.h"
-#include "aom_ports/mem.h"
-#include "aom_dsp/blend.h"
-#include "aom_dsp/aom_dsp_common.h"
-
-#include "./aom_dsp_rtcd.h"
-
-// Blending with alpha mask. Mask values come from the range [0, 64],
-// as described for AOM_BLEND_A64 in aom_dsp/blend.h. src0 or src1 can
-// be the same as dst, or dst can be different from both sources.
-
-void aom_blend_a64_mask_c(uint8_t *dst, uint32_t dst_stride,
-                          const uint8_t *src0, uint32_t src0_stride,
-                          const uint8_t *src1, uint32_t src1_stride,
-                          const uint8_t *mask, uint32_t mask_stride, int h,
-                          int w, int subh, int subw) {
-  int i, j;
-
-  assert(IMPLIES(src0 == dst, src0_stride == dst_stride));
-  assert(IMPLIES(src1 == dst, src1_stride == dst_stride));
-
-  assert(h >= 1);
-  assert(w >= 1);
-  assert(IS_POWER_OF_TWO(h));
-  assert(IS_POWER_OF_TWO(w));
-
-  if (subw == 0 && subh == 0) {
-    for (i = 0; i < h; ++i) {
-      for (j = 0; j < w; ++j) {
-        const int m = mask[i * mask_stride + j];
-        dst[i * dst_stride + j] = AOM_BLEND_A64(m, src0[i * src0_stride + j],
-                                                src1[i * src1_stride + j]);
-      }
-    }
-  } else if (subw == 1 && subh == 1) {
-    for (i = 0; i < h; ++i) {
-      for (j = 0; j < w; ++j) {
-        const int m = ROUND_POWER_OF_TWO(
-            mask[(2 * i) * mask_stride + (2 * j)] +
-                mask[(2 * i + 1) * mask_stride + (2 * j)] +
-                mask[(2 * i) * mask_stride + (2 * j + 1)] +
-                mask[(2 * i + 1) * mask_stride + (2 * j + 1)],
-            2);
-        dst[i * dst_stride + j] = AOM_BLEND_A64(m, src0[i * src0_stride + j],
-                                                src1[i * src1_stride + j]);
-      }
-    }
-  } else if (subw == 1 && subh == 0) {
-    for (i = 0; i < h; ++i) {
-      for (j = 0; j < w; ++j) {
-        const int m = AOM_BLEND_AVG(mask[i * mask_stride + (2 * j)],
-                                    mask[i * mask_stride + (2 * j + 1)]);
-        dst[i * dst_stride + j] = AOM_BLEND_A64(m, src0[i * src0_stride + j],
-                                                src1[i * src1_stride + j]);
-      }
-    }
-  } else {
-    for (i = 0; i < h; ++i) {
-      for (j = 0; j < w; ++j) {
-        const int m = AOM_BLEND_AVG(mask[(2 * i) * mask_stride + j],
-                                    mask[(2 * i + 1) * mask_stride + j]);
-        dst[i * dst_stride + j] = AOM_BLEND_A64(m, src0[i * src0_stride + j],
-                                                src1[i * src1_stride + j]);
-      }
-    }
-  }
-}
-
-#if CONFIG_AOM_HIGHBITDEPTH
-void aom_highbd_blend_a64_mask_c(uint8_t *dst_8, uint32_t dst_stride,
-                                 const uint8_t *src0_8, uint32_t src0_stride,
-                                 const uint8_t *src1_8, uint32_t src1_stride,
-                                 const uint8_t *mask, uint32_t mask_stride,
-                                 int h, int w, int subh, int subw, int bd) {
-  int i, j;
-  uint16_t *dst = CONVERT_TO_SHORTPTR(dst_8);
-  const uint16_t *src0 = CONVERT_TO_SHORTPTR(src0_8);
-  const uint16_t *src1 = CONVERT_TO_SHORTPTR(src1_8);
-  (void)bd;
-
-  assert(IMPLIES(src0 == dst, src0_stride == dst_stride));
-  assert(IMPLIES(src1 == dst, src1_stride == dst_stride));
-
-  assert(h >= 1);
-  assert(w >= 1);
-  assert(IS_POWER_OF_TWO(h));
-  assert(IS_POWER_OF_TWO(w));
-
-  assert(bd == 8 || bd == 10 || bd == 12);
-
-  if (subw == 0 && subh == 0) {
-    for (i = 0; i < h; ++i) {
-      for (j = 0; j < w; ++j) {
-        const int m = mask[i * mask_stride + j];
-        dst[i * dst_stride + j] = AOM_BLEND_A64(m, src0[i * src0_stride + j],
-                                                src1[i * src1_stride + j]);
-      }
-    }
-  } else if (subw == 1 && subh == 1) {
-    for (i = 0; i < h; ++i) {
-      for (j = 0; j < w; ++j) {
-        const int m = ROUND_POWER_OF_TWO(
-            mask[(2 * i) * mask_stride + (2 * j)] +
-                mask[(2 * i + 1) * mask_stride + (2 * j)] +
-                mask[(2 * i) * mask_stride + (2 * j + 1)] +
-                mask[(2 * i + 1) * mask_stride + (2 * j + 1)],
-            2);
-        dst[i * dst_stride + j] = AOM_BLEND_A64(m, src0[i * src0_stride + j],
-                                                src1[i * src1_stride + j]);
-      }
-    }
-  } else if (subw == 1 && subh == 0) {
-    for (i = 0; i < h; ++i) {
-      for (j = 0; j < w; ++j) {
-        const int m = AOM_BLEND_AVG(mask[i * mask_stride + (2 * j)],
-                                    mask[i * mask_stride + (2 * j + 1)]);
-        dst[i * dst_stride + j] = AOM_BLEND_A64(m, src0[i * src0_stride + j],
-                                                src1[i * src1_stride + j]);
-      }
-    }
-  } else {
-    for (i = 0; i < h; ++i) {
-      for (j = 0; j < w; ++j) {
-        const int m = AOM_BLEND_AVG(mask[(2 * i) * mask_stride + j],
-                                    mask[(2 * i + 1) * mask_stride + j]);
-        dst[i * dst_stride + j] = AOM_BLEND_A64(m, src0[i * src0_stride + j],
-                                                src1[i * src1_stride + j]);
-      }
-    }
-  }
-}
-#endif  // CONFIG_AOM_HIGHBITDEPTH

aom_dsp/blend_a64_vmask.c

@@ -1,73 +0,0 @@
-/*
- * Copyright (c) 2016, Alliance for Open Media. All rights reserved
- *
- * This source code is subject to the terms of the BSD 2 Clause License and
- * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
- * was not distributed with this source code in the LICENSE file, you can
- * obtain it at www.aomedia.org/license/software. If the Alliance for Open
- * Media Patent License 1.0 was not distributed with this source code in the
- * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
- */
-
-#include <assert.h>
-
-#include "aom/aom_integer.h"
-#include "aom_ports/mem.h"
-#include "aom_dsp/aom_dsp_common.h"
-#include "aom_dsp/blend.h"
-
-#include "./aom_dsp_rtcd.h"
-
-void aom_blend_a64_vmask_c(uint8_t *dst, uint32_t dst_stride,
-                           const uint8_t *src0, uint32_t src0_stride,
-                           const uint8_t *src1, uint32_t src1_stride,
-                           const uint8_t *mask, int h, int w) {
-  int i, j;
-
-  assert(IMPLIES(src0 == dst, src0_stride == dst_stride));
-  assert(IMPLIES(src1 == dst, src1_stride == dst_stride));
-
-  assert(h >= 1);
-  assert(w >= 1);
-  assert(IS_POWER_OF_TWO(h));
-  assert(IS_POWER_OF_TWO(w));
-
-  for (i = 0; i < h; ++i) {
-    const int m = mask[i];
-    for (j = 0; j < w; ++j) {
-      dst[i * dst_stride + j] = AOM_BLEND_A64(m, src0[i * src0_stride + j],
-                                              src1[i * src1_stride + j]);
-    }
-  }
-}
-
-#if CONFIG_AOM_HIGHBITDEPTH
-void aom_highbd_blend_a64_vmask_c(uint8_t *dst_8, uint32_t dst_stride,
-                                  const uint8_t *src0_8, uint32_t src0_stride,
-                                  const uint8_t *src1_8, uint32_t src1_stride,
-                                  const uint8_t *mask, int h, int w, int bd) {
-  int i, j;
-  uint16_t *dst = CONVERT_TO_SHORTPTR(dst_8);
-  const uint16_t *src0 = CONVERT_TO_SHORTPTR(src0_8);
-  const uint16_t *src1 = CONVERT_TO_SHORTPTR(src1_8);
-  (void)bd;
-
-  assert(IMPLIES(src0 == dst, src0_stride == dst_stride));
-  assert(IMPLIES(src1 == dst, src1_stride == dst_stride));
-
-  assert(h >= 1);
-  assert(w >= 1);
-  assert(IS_POWER_OF_TWO(h));
-  assert(IS_POWER_OF_TWO(w));
-
-  assert(bd == 8 || bd == 10 || bd == 12);
-
-  for (i = 0; i < h; ++i) {
-    const int m = mask[i];
-    for (j = 0; j < w; ++j) {
-      dst[i * dst_stride + j] = AOM_BLEND_A64(m, src0[i * src0_stride + j],
-                                              src1[i * src1_stride + j]);
-    }
-  }
-}
-#endif  // CONFIG_AOM_HIGHBITDEPTH

aom_dsp/buf_ans.c

@@ -1,42 +0,0 @@
-/*
- * Copyright (c) 2016, Alliance for Open Media. All rights reserved
- *
- * This source code is subject to the terms of the BSD 2 Clause License and
- * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
- * was not distributed with this source code in the LICENSE file, you can
- * obtain it at www.aomedia.org/license/software. If the Alliance for Open
- * Media Patent License 1.0 was not distributed with this source code in the
- * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
- */
-
-#include <string.h>
-
-#include "aom_dsp/buf_ans.h"
-#include "aom_mem/aom_mem.h"
-#include "aom/internal/aom_codec_internal.h"
-
-void aom_buf_ans_alloc(struct BufAnsCoder *c,
-                       struct aom_internal_error_info *error, int size_hint) {
-  c->error = error;
-  c->size = size_hint;
-  AOM_CHECK_MEM_ERROR(error, c->buf, aom_malloc(c->size * sizeof(*c->buf)));
-  // Initialize to overfull to trigger the assert in write.
-  c->offset = c->size + 1;
-}
-
-void aom_buf_ans_free(struct BufAnsCoder *c) {
-  aom_free(c->buf);
-  c->buf = NULL;
-  c->size = 0;
-}
-
-void aom_buf_ans_grow(struct BufAnsCoder *c) {
-  struct buffered_ans_symbol *new_buf = NULL;
-  int new_size = c->size * 2;
-  AOM_CHECK_MEM_ERROR(c->error, new_buf,
-                      aom_malloc(new_size * sizeof(*new_buf)));
-  memcpy(new_buf, c->buf, c->size * sizeof(*c->buf));
-  aom_free(c->buf);
-  c->buf = new_buf;
-  c->size = new_size;
-}

aom_dsp/buf_ans.h

@@ -1,112 +0,0 @@
-/*
- * Copyright (c) 2016, Alliance for Open Media. All rights reserved
- *
- * This source code is subject to the terms of the BSD 2 Clause License and
- * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
- * was not distributed with this source code in the LICENSE file, you can
- * obtain it at www.aomedia.org/license/software. If the Alliance for Open
- * Media Patent License 1.0 was not distributed with this source code in the
- * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
- */
-
-#ifndef AOM_DSP_BUF_ANS_H_
-#define AOM_DSP_BUF_ANS_H_
-// Buffered forward ANS writer.
-// Symbols are written to the writer in forward (decode) order and serialized
-// backwards due to ANS's stack like behavior.
-
-#include <assert.h>
-#include "./aom_config.h"
-#include "aom/aom_integer.h"
-#include "aom_dsp/ans.h"
-#include "aom_dsp/answriter.h"
-
-#ifdef __cplusplus
-extern "C" {
-#endif  // __cplusplus
-
-#define ANS_METHOD_UABS 0
-#define ANS_METHOD_RANS 1
-
-struct buffered_ans_symbol {
-  unsigned int method : 1;  // one of ANS_METHOD_UABS or ANS_METHOD_RANS
-  // TODO(aconverse): Should be possible to write this in terms of start for ABS
-  unsigned int val_start : RANS_PROB_BITS;  // Boolean value for ABS
-                                            // start in symbol cycle for Rans
-  unsigned int prob : RANS_PROB_BITS;       // Probability of this symbol
-};
-
-struct BufAnsCoder {
-  struct aom_internal_error_info *error;
-  struct buffered_ans_symbol *buf;
-  int size;
-  int offset;
-};
-
-void aom_buf_ans_alloc(struct BufAnsCoder *c,
-                       struct aom_internal_error_info *error, int size_hint);
-
-void aom_buf_ans_free(struct BufAnsCoder *c);
-
-void aom_buf_ans_grow(struct BufAnsCoder *c);
-
-static INLINE void buf_ans_write_reset(struct BufAnsCoder *const c) {
-  c->offset = 0;
-}
-
-static INLINE void buf_uabs_write(struct BufAnsCoder *const c, uint8_t val,
-                                  AnsP8 prob) {
-  assert(c->offset <= c->size);
-  if (c->offset == c->size) {
-    aom_buf_ans_grow(c);
-  }
-  c->buf[c->offset].method = ANS_METHOD_UABS;
-  c->buf[c->offset].val_start = val;
-  c->buf[c->offset].prob = prob;
-  ++c->offset;
-}
-
-static INLINE void buf_rans_write(struct BufAnsCoder *const c,
-                                  const struct rans_sym *const sym) {
-  assert(c->offset <= c->size);
-  if (c->offset == c->size) {
-    aom_buf_ans_grow(c);
-  }
-  c->buf[c->offset].method = ANS_METHOD_RANS;
-  c->buf[c->offset].val_start = sym->cum_prob;
-  c->buf[c->offset].prob = sym->prob;
-  ++c->offset;
-}
-
-static INLINE void buf_ans_flush(const struct BufAnsCoder *const c,
-                                 struct AnsCoder *ans) {
-  int offset;
-  for (offset = c->offset - 1; offset >= 0; --offset) {
-    if (c->buf[offset].method == ANS_METHOD_RANS) {
-      struct rans_sym sym;
-      sym.prob = c->buf[offset].prob;
-      sym.cum_prob = c->buf[offset].val_start;
-      rans_write(ans, &sym);
-    } else {
-      uabs_write(ans, (uint8_t)c->buf[offset].val_start,
-                 (AnsP8)c->buf[offset].prob);
-    }
-  }
-}
-
-static INLINE void buf_uabs_write_bit(struct BufAnsCoder *c, int bit) {
-  buf_uabs_write(c, bit, 128);
-}
-
-static INLINE void buf_uabs_write_literal(struct BufAnsCoder *c, int literal,
-                                          int bits) {
-  int bit;
-
-  assert(bits < 31);
-  for (bit = bits - 1; bit >= 0; bit--)
-    buf_uabs_write_bit(c, 1 & (literal >> bit));
-}
-#ifdef __cplusplus
-}  // extern "C"
-#endif  // __cplusplus
-#endif  // AOM_DSP_BUF_ANS_H_

aom_dsp/daalaboolreader.c

@@ -1,37 +0,0 @@
-/*
- * Copyright (c) 2016, Alliance for Open Media. All rights reserved
- *
- * This source code is subject to the terms of the BSD 2 Clause License and
- * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
- * was not distributed with this source code in the LICENSE file, you can
- * obtain it at www.aomedia.org/license/software. If the Alliance for Open
- * Media Patent License 1.0 was not distributed with this source code in the
- * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
- */
-
-#include "aom_dsp/daalaboolreader.h"
-
-int aom_daala_reader_init(daala_reader *r, const uint8_t *buffer, int size) {
-  if (size && !buffer) {
-    return 1;
-  }
-  r->buffer_end = buffer + size;
-  r->buffer = buffer;
-  od_ec_dec_init(&r->ec, buffer, size - 1);
-#if CONFIG_ACCOUNTING
-  r->accounting = NULL;
-#endif
-  return 0;
-}
-
-const uint8_t *aom_daala_reader_find_end(daala_reader *r) {
-  return r->buffer_end;
-}
-
-uint32_t aom_daala_reader_tell(const daala_reader *r) {
-  return od_ec_dec_tell(&r->ec);
-}
-
-uint32_t aom_daala_reader_tell_frac(const daala_reader *r) {
-  return od_ec_dec_tell_frac(&r->ec);
-}

aom_dsp/daalaboolreader.h

@@ -1,87 +0,0 @@
-/*
- * Copyright (c) 2016, Alliance for Open Media. All rights reserved
- *
- * This source code is subject to the terms of the BSD 2 Clause License and
- * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
- * was not distributed with this source code in the LICENSE file, you can
- * obtain it at www.aomedia.org/license/software. If the Alliance for Open
- * Media Patent License 1.0 was not distributed with this source code in the
- * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
- */
-
-#ifndef AOM_DSP_DAALABOOLREADER_H_
-#define AOM_DSP_DAALABOOLREADER_H_
-
-#include "aom/aom_integer.h"
-#include "aom_dsp/entdec.h"
-#include "aom_dsp/prob.h"
-#if CONFIG_ACCOUNTING
-#include "av1/common/accounting.h"
-#endif
-
-#ifdef __cplusplus
-extern "C" {
-#endif
-
-struct daala_reader {
-  const uint8_t *buffer;
-  const uint8_t *buffer_end;
-  od_ec_dec ec;
-#if CONFIG_ACCOUNTING
-  Accounting *accounting;
-#endif
-};
-
-typedef struct daala_reader daala_reader;
-
-int aom_daala_reader_init(daala_reader *r, const uint8_t *buffer, int size);
-const uint8_t *aom_daala_reader_find_end(daala_reader *r);
-uint32_t aom_daala_reader_tell(const daala_reader *r);
-uint32_t aom_daala_reader_tell_frac(const daala_reader *r);
-
-static INLINE int aom_daala_read(daala_reader *r, int prob) {
-  if (prob == 128) {
-    return od_ec_dec_bits(&r->ec, 1, "aom_bits");
-  } else {
-    int p = ((prob << 15) + (256 - prob)) >> 8;
-    return od_ec_decode_bool_q15(&r->ec, p);
-  }
-}
-
-static INLINE int aom_daala_read_bit(daala_reader *r) {
-  return aom_daala_read(r, 128);
-}
-
-static INLINE int aom_daala_reader_has_error(daala_reader *r) {
-  return r->ec.error;
-}
-
-static INLINE int daala_read_tree_bits(daala_reader *r,
-                                       const aom_tree_index *tree,
-                                       const aom_prob *probs) {
-  aom_tree_index i = 0;
-  do {
-    aom_cdf_prob cdf[16];
-    aom_tree_index index[16];
-    int path[16];
-    int dist[16];
-    int nsymbs;
-    int symb;
-    nsymbs = tree_to_cdf(tree, probs, i, cdf, index, path, dist);
-    symb = od_ec_decode_cdf_q15(&r->ec, cdf, nsymbs);
-    OD_ASSERT(symb >= 0 && symb < nsymbs);
-    i = index[symb];
-  } while (i > 0);
-  return -i;
-}
-
-static INLINE int daala_read_symbol(daala_reader *r, const aom_cdf_prob *cdf,
-                                    int nsymbs) {
-  return od_ec_decode_cdf_q15(&r->ec, cdf, nsymbs);
-}
-
-#ifdef __cplusplus
-}  // extern "C"
-#endif
-
-#endif

aom_dsp/daalaboolwriter.c

@@ -1,32 +0,0 @@
-/*
- * Copyright (c) 2016, Alliance for Open Media. All rights reserved
- *
- * This source code is subject to the terms of the BSD 2 Clause License and
- * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
- * was not distributed with this source code in the LICENSE file, you can
- * obtain it at www.aomedia.org/license/software. If the Alliance for Open
- * Media Patent License 1.0 was not distributed with this source code in the
- * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
- */
-
-#include <string.h>
-#include "aom_dsp/daalaboolwriter.h"
-
-void aom_daala_start_encode(daala_writer *br, uint8_t *source) {
-  br->buffer = source;
-  br->pos = 0;
-  od_ec_enc_init(&br->ec, 62025);
-}
-
-void aom_daala_stop_encode(daala_writer *br) {
-  uint32_t daala_bytes;
-  unsigned char *daala_data;
-  daala_data = od_ec_enc_done(&br->ec, &daala_bytes);
-  memcpy(br->buffer, daala_data, daala_bytes);
-  br->pos = daala_bytes;
-  /* Prevent ec bitstream from being detected as a superframe marker.
-     Must always be added, so that rawbits knows the exact length of the
-      bitstream. */
-  br->buffer[br->pos++] = 0;
-  od_ec_enc_clear(&br->ec);
-}

aom_dsp/daalaboolwriter.h

@@ -1,90 +0,0 @@
-/*
- * Copyright (c) 2016, Alliance for Open Media. All rights reserved
- *
- * This source code is subject to the terms of the BSD 2 Clause License and
- * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
- * was not distributed with this source code in the LICENSE file, you can
- * obtain it at www.aomedia.org/license/software. If the Alliance for Open
- * Media Patent License 1.0 was not distributed with this source code in the
- * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
- */
-
-#ifndef AOM_DSP_DAALABOOLWRITER_H_
-#define AOM_DSP_DAALABOOLWRITER_H_
-
-#include "aom_dsp/entenc.h"
-#include "aom_dsp/prob.h"
-
-#ifdef __cplusplus
-extern "C" {
-#endif
-
-struct daala_writer {
-  unsigned int pos;
-  uint8_t *buffer;
-  od_ec_enc ec;
-};
-
-typedef struct daala_writer daala_writer;
-
-void aom_daala_start_encode(daala_writer *w, uint8_t *buffer);
-void aom_daala_stop_encode(daala_writer *w);
-
-static INLINE void aom_daala_write(daala_writer *w, int bit, int prob) {
-  if (prob == 128) {
-    od_ec_enc_bits(&w->ec, bit, 1);
-  } else {
-    int p = ((prob << 15) + (256 - prob)) >> 8;
-    od_ec_encode_bool_q15(&w->ec, bit, p);
-  }
-}
-
-static INLINE void daala_write_tree_bits(daala_writer *w,
-                                         const aom_tree_index *tree,
-                                         const aom_prob *probs, int bits,
-                                         int len, aom_tree_index i) {
-  aom_tree_index root;
-  root = i;
-  do {
-    aom_cdf_prob cdf[16];
-    aom_tree_index index[16];
-    int path[16];
-    int dist[16];
-    int nsymbs;
-    int symb;
-    int j;
-    /* Compute the CDF of the binary tree using the given probabilities. */
-    nsymbs = tree_to_cdf(tree, probs, root, cdf, index, path, dist);
-    /* Find the symbol to code. */
-    symb = -1;
-    for (j = 0; j < nsymbs; j++) {
-      /* If this symbol codes a leaf node,  */
-      if (index[j] <= 0) {
-        if (len == dist[j] && path[j] == bits) {
-          symb = j;
-          break;
-        }
-      } else {
-        if (len > dist[j] && path[j] == bits >> (len - dist[j])) {
-          symb = j;
-          break;
-        }
-      }
-    }
-    OD_ASSERT(symb != -1);
-    od_ec_encode_cdf_q15(&w->ec, symb, cdf, nsymbs);
-    bits &= (1 << (len - dist[symb])) - 1;
-    len -= dist[symb];
-  } while (len);
-}
-
-static INLINE void daala_write_symbol(daala_writer *w, int symb,
-                                      const aom_cdf_prob *cdf, int nsymbs) {
-  od_ec_encode_cdf_q15(&w->ec, symb, cdf, nsymbs);
-}
-
-#ifdef __cplusplus
-}  // extern "C"
-#endif
-
-#endif

aom_dsp/deblock.c

@@ -1,195 +0,0 @@
-/*
- * Copyright (c) 2016, Alliance for Open Media. All rights reserved
- *
- * This source code is subject to the terms of the BSD 2 Clause License and
- * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
- * was not distributed with this source code in the LICENSE file, you can
- * obtain it at www.aomedia.org/license/software. If the Alliance for Open
- * Media Patent License 1.0 was not distributed with this source code in the
- * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
- *
- */
-
-#include <stdlib.h>
-#include "aom/aom_integer.h"
-
-const int16_t aom_rv[] = {
-  8,  5,  2,  2,  8,  12, 4,  9,  8,  3,  0,  3,  9,  0,  0,  0,  8,  3,  14,
-  4,  10, 1,  11, 14, 1,  14, 9,  6,  12, 11, 8,  6,  10, 0,  0,  8,  9,  0,
-  3,  14, 8,  11, 13, 4,  2,  9,  0,  3,  9,  6,  1,  2,  3,  14, 13, 1,  8,
-  2,  9,  7,  3,  3,  1,  13, 13, 6,  6,  5,  2,  7,  11, 9,  11, 8,  7,  3,
-  2,  0,  13, 13, 14, 4,  12, 5,  12, 10, 8,  10, 13, 10, 4,  14, 4,  10, 0,
-  8,  11, 1,  13, 7,  7,  14, 6,  14, 13, 2,  13, 5,  4,  4,  0,  10, 0,  5,
-  13, 2,  12, 7,  11, 13, 8,  0,  4,  10, 7,  2,  7,  2,  2,  5,  3,  4,  7,
-  3,  3,  14, 14, 5,  9,  13, 3,  14, 3,  6,  3,  0,  11, 8,  13, 1,  13, 1,
-  12, 0,  10, 9,  7,  6,  2,  8,  5,  2,  13, 7,  1,  13, 14, 7,  6,  7,  9,
-  6,  10, 11, 7,  8,  7,  5,  14, 8,  4,  4,  0,  8,  7,  10, 0,  8,  14, 11,
-  3,  12, 5,  7,  14, 3,  14, 5,  2,  6,  11, 12, 12, 8,  0,  11, 13, 1,  2,
-  0,  5,  10, 14, 7,  8,  0,  4,  11, 0,  8,  0,  3,  10, 5,  8,  0,  11, 6,
-  7,  8,  10, 7,  13, 9,  2,  5,  1,  5,  10, 2,  4,  3,  5,  6,  10, 8,  9,
-  4,  11, 14, 0,  10, 0,  5,  13, 2,  12, 7,  11, 13, 8,  0,  4,  10, 7,  2,
-  7,  2,  2,  5,  3,  4,  7,  3,  3,  14, 14, 5,  9,  13, 3,  14, 3,  6,  3,
-  0,  11, 8,  13, 1,  13, 1,  12, 0,  10, 9,  7,  6,  2,  8,  5,  2,  13, 7,
-  1,  13, 14, 7,  6,  7,  9,  6,  10, 11, 7,  8,  7,  5,  14, 8,  4,  4,  0,
-  8,  7,  10, 0,  8,  14, 11, 3,  12, 5,  7,  14, 3,  14, 5,  2,  6,  11, 12,
-  12, 8,  0,  11, 13, 1,  2,  0,  5,  10, 14, 7,  8,  0,  4,  11, 0,  8,  0,
-  3,  10, 5,  8,  0,  11, 6,  7,  8,  10, 7,  13, 9,  2,  5,  1,  5,  10, 2,
-  4,  3,  5,  6,  10, 8,  9,  4,  11, 14, 3,  8,  3,  7,  8,  5,  11, 4,  12,
-  3,  11, 9,  14, 8,  14, 13, 4,  3,  1,  2,  14, 6,  5,  4,  4,  11, 4,  6,
-  2,  1,  5,  8,  8,  12, 13, 5,  14, 10, 12, 13, 0,  9,  5,  5,  11, 10, 13,
-  9,  10, 13,
-};
-
-void aom_post_proc_down_and_across_mb_row_c(unsigned char *src_ptr,
-                                            unsigned char *dst_ptr,
-                                            int src_pixels_per_line,
-                                            int dst_pixels_per_line, int cols,
-                                            unsigned char *f, int size) {
-  unsigned char *p_src, *p_dst;
-  int row;
-  int col;
-  unsigned char v;
-  unsigned char d[4];
-
-  for (row = 0; row < size; row++) {
-    /* post_proc_down for one row */
-    p_src = src_ptr;
-    p_dst = dst_ptr;
-
-    for (col = 0; col < cols; col++) {
-      unsigned char p_above2 = p_src[col - 2 * src_pixels_per_line];
-      unsigned char p_above1 = p_src[col - src_pixels_per_line];
-      unsigned char p_below1 = p_src[col + src_pixels_per_line];
-      unsigned char p_below2 = p_src[col + 2 * src_pixels_per_line];
-
-      v = p_src[col];
-
-      if ((abs(v - p_above2) < f[col]) && (abs(v - p_above1) < f[col]) &&
-          (abs(v - p_below1) < f[col]) && (abs(v - p_below2) < f[col])) {
-        unsigned char k1, k2, k3;
-        k1 = (p_above2 + p_above1 + 1) >> 1;
-        k2 = (p_below2 + p_below1 + 1) >> 1;
-        k3 = (k1 + k2 + 1) >> 1;
-        v = (k3 + v + 1) >> 1;
-      }
-
-      p_dst[col] = v;
-    }
-
-    /* now post_proc_across */
-    p_src = dst_ptr;
-    p_dst = dst_ptr;
-
-    p_src[-2] = p_src[-1] = p_src[0];
-    p_src[cols] = p_src[cols + 1] = p_src[cols - 1];
-
-    for (col = 0; col < cols; col++) {
-      v = p_src[col];
-
-      if ((abs(v - p_src[col - 2]) < f[col]) &&
-          (abs(v - p_src[col - 1]) < f[col]) &&
-          (abs(v - p_src[col + 1]) < f[col]) &&
-          (abs(v - p_src[col + 2]) < f[col])) {
-        unsigned char k1, k2, k3;
-        k1 = (p_src[col - 2] + p_src[col - 1] + 1) >> 1;
-        k2 = (p_src[col + 2] + p_src[col + 1] + 1) >> 1;
-        k3 = (k1 + k2 + 1) >> 1;
-        v = (k3 + v + 1) >> 1;
-      }
-
-      d[col & 3] = v;
-
-      if (col >= 2) p_dst[col - 2] = d[(col - 2) & 3];
-    }
-
-    /* handle the last two pixels */
-    p_dst[col - 2] = d[(col - 2) & 3];
-    p_dst[col - 1] = d[(col - 1) & 3];
-
-    /* next row */
-    src_ptr += src_pixels_per_line;
-    dst_ptr += dst_pixels_per_line;
-  }
-}
-
-void aom_mbpost_proc_across_ip_c(unsigned char *src, int pitch, int rows,
-                                 int cols, int flimit) {
-  int r, c, i;
-
-  unsigned char *s = src;
-  unsigned char d[16];
-
-  for (r = 0; r < rows; r++) {
-    int sumsq = 0;
-    int sum = 0;
-
-    for (i = -8; i < 0; i++) s[i] = s[0];
-
-    /* 17 avoids valgrind warning - we buffer values in c in d
-     * and only write them when we've read 8 ahead...
-     */
-    for (i = 0; i < 17; i++) s[i + cols] = s[cols - 1];
-
-    for (i = -8; i <= 6; i++) {
-      sumsq += s[i] * s[i];
-      sum += s[i];
-      d[i + 8] = 0;
-    }
-
-    for (c = 0; c < cols + 8; c++) {
-      int x = s[c + 7] - s[c - 8];
-      int y = s[c + 7] + s[c - 8];
-
-      sum += x;
-      sumsq += x * y;
-
-      d[c & 15] = s[c];
-
-      if (sumsq * 15 - sum * sum < flimit) {
-        d[c & 15] = (8 + sum + s[c]) >> 4;
-      }
-
-      s[c - 8] = d[(c - 8) & 15];
-    }
-
-    s += pitch;
-  }
-}
-
-void aom_mbpost_proc_down_c(unsigned char *dst, int pitch, int rows, int cols,
-                            int flimit) {
-  int r, c, i;
-  const int16_t *rv3 = &aom_rv[63 & rand()];
-
-  for (c = 0; c < cols; c++) {
-    unsigned char *s = &dst[c];
-    int sumsq = 0;
-    int sum = 0;
-    unsigned char d[16];
-    const int16_t *rv2 = rv3 + ((c * 17) & 127);
-
-    for (i = -8; i < 0; i++) s[i * pitch] = s[0];
-
-    /* 17 avoids valgrind warning - we buffer values in c in d
-     * and only write them when we've read 8 ahead...
-     */
-    for (i = 0; i < 17; i++) s[(i + rows) * pitch] = s[(rows - 1) * pitch];
-
-    for (i = -8; i <= 6; i++) {
-      sumsq += s[i * pitch] * s[i * pitch];
-      sum += s[i * pitch];
-    }
-
-    for (r = 0; r < rows + 8; r++) {
-      sumsq += s[7 * pitch] * s[7 * pitch] - s[-8 * pitch] * s[-8 * pitch];
-      sum += s[7 * pitch] - s[-8 * pitch];
-      d[r & 15] = s[0];
-
-      if (sumsq * 15 - sum * sum < flimit) {
-        d[r & 15] = (rv2[r & 127] + sum + s[0]) >> 4;
-      }
-      if (r >= 8) s[-8 * pitch] = d[(r - 8) & 15];
-      s += pitch;
-    }
-  }
-}

aom_dsp/dkboolreader.c

@@ -1,108 +0,0 @@
-/*
- *  Copyright (c) 2010 The WebM project authors. All Rights Reserved.
- *
- *  Use of this source code is governed by a BSD-style license
- *  that can be found in the LICENSE file in the root of the source
- *  tree. An additional intellectual property rights grant can be found
- *  in the file PATENTS.  All contributing project authors may
- *  be found in the AUTHORS file in the root of the source tree.
- */
-#include <stdlib.h>
-
-#include "./aom_config.h"
-
-#include "aom_dsp/dkboolreader.h"
-#include "aom_dsp/prob.h"
-#include "aom_dsp/aom_dsp_common.h"
-#include "aom_ports/mem.h"
-#include "aom_mem/aom_mem.h"
-#include "aom_util/endian_inl.h"
-
-static INLINE int aom_dk_read_bit(struct aom_dk_reader *r) {
-  return aom_dk_read(r, 128);  // aom_prob_half
-}
-
-int aom_dk_reader_init(struct aom_dk_reader *r, const uint8_t *buffer,
-                       size_t size, aom_decrypt_cb decrypt_cb,
-                       void *decrypt_state) {
-  if (size && !buffer) {
-    return 1;
-  } else {
-    r->buffer_end = buffer + size;
-    r->buffer_start = r->buffer = buffer;
-    r->value = 0;
-    r->count = -8;
-    r->range = 255;
-    r->decrypt_cb = decrypt_cb;
-    r->decrypt_state = decrypt_state;
-    aom_dk_reader_fill(r);
-#if CONFIG_ACCOUNTING
-    r->accounting = NULL;
-#endif
-    return aom_dk_read_bit(r) != 0;  // marker bit
-  }
-}
-
-void aom_dk_reader_fill(struct aom_dk_reader *r) {
-  const uint8_t *const buffer_end = r->buffer_end;
-  const uint8_t *buffer = r->buffer;
-  const uint8_t *buffer_start = buffer;
-  BD_VALUE value = r->value;
-  int count = r->count;
-  const size_t bytes_left = buffer_end - buffer;
-  const size_t bits_left = bytes_left * CHAR_BIT;
-  int shift = BD_VALUE_SIZE - CHAR_BIT - (count + CHAR_BIT);
-
-  if (r->decrypt_cb) {
-    size_t n = AOMMIN(sizeof(r->clear_buffer), bytes_left);
-    r->decrypt_cb(r->decrypt_state, buffer, r->clear_buffer, (int)n);
-    buffer = r->clear_buffer;
-    buffer_start = r->clear_buffer;
-  }
-  if (bits_left > BD_VALUE_SIZE) {
-    const int bits = (shift & 0xfffffff8) + CHAR_BIT;
-    BD_VALUE nv;
-    BD_VALUE big_endian_values;
-    memcpy(&big_endian_values, buffer, sizeof(BD_VALUE));
-#if SIZE_MAX == 0xffffffffffffffffULL
-    big_endian_values = HToBE64(big_endian_values);
-#else
-    big_endian_values = HToBE32(big_endian_values);
-#endif
-    nv = big_endian_values >> (BD_VALUE_SIZE - bits);
-    count += bits;
-    buffer += (bits >> 3);
-    value = r->value | (nv << (shift & 0x7));
-  } else {
-    const int bits_over = (int)(shift + CHAR_BIT - (int)bits_left);
-    int loop_end = 0;
-    if (bits_over >= 0) {
-      count += LOTS_OF_BITS;
-      loop_end = bits_over;
-    }
-
-    if (bits_over < 0 || bits_left) {
-      while (shift >= loop_end) {
-        count += CHAR_BIT;
-        value |= (BD_VALUE)*buffer++ << shift;
-        shift -= CHAR_BIT;
-      }
-    }
-  }
-
-  // NOTE: Variable 'buffer' may not relate to 'r->buffer' after decryption,
-  // so we increase 'r->buffer' by the amount that 'buffer' moved, rather than
-  // assign 'buffer' to 'r->buffer'.
-  r->buffer += buffer - buffer_start;
-  r->value = value;
-  r->count = count;
-}
-
-const uint8_t *aom_dk_reader_find_end(struct aom_dk_reader *r) {
-  // Find the end of the coded buffer
-  while (r->count > CHAR_BIT && r->count < BD_VALUE_SIZE) {
-    r->count -= CHAR_BIT;
-    r->buffer--;
-  }
-  return r->buffer;
-}

aom_dsp/dkboolreader.h

@@ -1,180 +0,0 @@
-/*
- * Copyright (c) 2016, Alliance for Open Media. All rights reserved
- *
- * This source code is subject to the terms of the BSD 2 Clause License and
- * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
- * was not distributed with this source code in the LICENSE file, you can
- * obtain it at www.aomedia.org/license/software. If the Alliance for Open
- * Media Patent License 1.0 was not distributed with this source code in the
- * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
- */
-
-#ifndef AOM_DSP_DKBOOLREADER_H_
-#define AOM_DSP_DKBOOLREADER_H_
-
-#include <assert.h>
-#include <stddef.h>
-#include <limits.h>
-
-#include "./aom_config.h"
-#if CONFIG_BITSTREAM_DEBUG
-#include <assert.h>
-#include <stdio.h>
-#include "aom_util/debug_util.h"
-#endif  // CONFIG_BITSTREAM_DEBUG
-
-#include "aom_ports/mem.h"
-#include "aom/aomdx.h"
-#include "aom/aom_integer.h"
-#include "aom_dsp/prob.h"
-#if CONFIG_ACCOUNTING
-#include "av1/common/accounting.h"
-#endif
-
-#ifdef __cplusplus
-extern "C" {
-#endif
-
-typedef size_t BD_VALUE;
-
-#define BD_VALUE_SIZE ((int)sizeof(BD_VALUE) * CHAR_BIT)
-
-// This is meant to be a large, positive constant that can still be efficiently
-// loaded as an immediate (on platforms like ARM, for example).
-// Even relatively modest values like 100 would work fine.
-#define LOTS_OF_BITS 0x40000000
-
-struct aom_dk_reader {
-  // Be careful when reordering this struct, it may impact the cache negatively.
-  BD_VALUE value;
-  unsigned int range;
-  int count;
-  const uint8_t *buffer_start;
-  const uint8_t *buffer_end;
-  const uint8_t *buffer;
-  aom_decrypt_cb decrypt_cb;
-  void *decrypt_state;
-  uint8_t clear_buffer[sizeof(BD_VALUE) + 1];
-#if CONFIG_ACCOUNTING
-  Accounting *accounting;
-#endif
-};
-
-int aom_dk_reader_init(struct aom_dk_reader *r, const uint8_t *buffer,
-                       size_t size, aom_decrypt_cb decrypt_cb,
-                       void *decrypt_state);
-
-void aom_dk_reader_fill(struct aom_dk_reader *r);
-
-const uint8_t *aom_dk_reader_find_end(struct aom_dk_reader *r);
-
-static INLINE uint32_t aom_dk_reader_tell(const struct aom_dk_reader *r) {
-  const uint32_t bits_read = (r->buffer - r->buffer_start) * CHAR_BIT;
-  const int count =
-      (r->count < LOTS_OF_BITS) ? r->count : r->count - LOTS_OF_BITS;
-  assert(r->buffer >= r->buffer_start);
-  return bits_read - (count + CHAR_BIT);
-}
-
-/*The resolution of fractional-precision bit usage measurements, i.e.,
-   3 => 1/8th bits.*/
-#define DK_BITRES (3)
-
-static INLINE uint32_t aom_dk_reader_tell_frac(const struct aom_dk_reader *r) {
-  uint32_t num_bits;
-  uint32_t range;
-  int l;
-  int i;
-  num_bits = aom_dk_reader_tell(r) << DK_BITRES;
-  range = r->range;
-  l = 0;
-  for (i = DK_BITRES; i-- > 0;) {
-    int b;
-    range = range * range >> 7;
-    b = (int)(range >> 8);
-    l = l << 1 | b;
-    range >>= b;
-  }
-  return num_bits - l;
-}
-
-static INLINE int aom_dk_reader_has_error(struct aom_dk_reader *r) {
-  // Check if we have reached the end of the buffer.
-  //
-  // Variable 'count' stores the number of bits in the 'value' buffer, minus
-  // 8. The top byte is part of the algorithm, and the remainder is buffered
-  // to be shifted into it. So if count == 8, the top 16 bits of 'value' are
-  // occupied, 8 for the algorithm and 8 in the buffer.
-  //
-  // When reading a byte from the user's buffer, count is filled with 8 and
-  // one byte is filled into the value buffer. When we reach the end of the
-  // data, count is additionally filled with LOTS_OF_BITS. So when
-  // count == LOTS_OF_BITS - 1, the user's data has been exhausted.
-  //
-  // 1 if we have tried to decode bits after the end of stream was encountered.
-  // 0 No error.
-  return r->count > BD_VALUE_SIZE && r->count < LOTS_OF_BITS;
-}
-
-static INLINE int aom_dk_read(struct aom_dk_reader *r, int prob) {
-  unsigned int bit = 0;
-  BD_VALUE value;
-  BD_VALUE bigsplit;
-  int count;
-  unsigned int range;
-  unsigned int split = (r->range * prob + (256 - prob)) >> CHAR_BIT;
-
-  if (r->count < 0) aom_dk_reader_fill(r);
-
-  value = r->value;
-  count = r->count;
-
-  bigsplit = (BD_VALUE)split << (BD_VALUE_SIZE - CHAR_BIT);
-
-  range = split;
-
-  if (value >= bigsplit) {
-    range = r->range - split;
-    value = value - bigsplit;
-    bit = 1;
-  }
-
-  {
-    register int shift = aom_norm[range];
-    range <<= shift;
-    value <<= shift;
-    count -= shift;
-  }
-  r->value = value;
-  r->count = count;
-  r->range = range;
-
-#if CONFIG_BITSTREAM_DEBUG
-  {
-    int ref_bit, ref_prob;
-    const int queue_r = bitstream_queue_get_read();
-    const int frame_idx = bitstream_queue_get_frame_read();
-    bitstream_queue_pop(&ref_bit, &ref_prob);
-    if (prob != ref_prob) {
-      fprintf(
-          stderr,
-          "\n *** prob error, frame_idx_r %d prob %d ref_prob %d queue_r %d\n",
-          frame_idx, prob, ref_prob, queue_r);
-      assert(0);
-    }
-    if ((int)bit != ref_bit) {
-      fprintf(stderr, "\n *** bit error, frame_idx_r %d bit %d ref_bit %d\n",
-              frame_idx, bit, ref_bit);
-      assert(0);
-    }
-  }
-#endif  // CONFIG_BITSTREAM_DEBUG
-
-  return bit;
-}
-
-#ifdef __cplusplus
-}  // extern "C"
-#endif
-
-#endif  // AOM_DSP_DKBOOLREADER_H_

aom_dsp/dkboolwriter.c

@@ -1,44 +0,0 @@
-/*
- * Copyright (c) 2016, Alliance for Open Media. All rights reserved
- *
- * This source code is subject to the terms of the BSD 2 Clause License and
- * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
- * was not distributed with this source code in the LICENSE file, you can
- * obtain it at www.aomedia.org/license/software. If the Alliance for Open
- * Media Patent License 1.0 was not distributed with this source code in the
- * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
- */
-
-#include <assert.h>
-
-#include "./dkboolwriter.h"
-
-static INLINE void aom_dk_write_bit(aom_dk_writer *w, int bit) {
-  aom_dk_write(w, bit, 128);  // aom_prob_half
-}
-
-void aom_dk_start_encode(aom_dk_writer *br, uint8_t *source) {
-  br->lowvalue = 0;
-  br->range = 255;
-  br->count = -24;
-  br->buffer = source;
-  br->pos = 0;
-  aom_dk_write_bit(br, 0);
-}
-
-void aom_dk_stop_encode(aom_dk_writer *br) {
-  int i;
-
-#if CONFIG_BITSTREAM_DEBUG
-  bitstream_queue_set_skip_write(1);
-#endif  // CONFIG_BITSTREAM_DEBUG
-
-  for (i = 0; i < 32; i++) aom_dk_write_bit(br, 0);
-
-#if CONFIG_BITSTREAM_DEBUG
-  bitstream_queue_set_skip_write(0);
-#endif  // CONFIG_BITSTREAM_DEBUG
-
-  // Ensure there's no ambigous collision with any index marker bytes
-  if ((br->buffer[br->pos - 1] & 0xe0) == 0xc0) br->buffer[br->pos++] = 0;
-}

aom_dsp/dkboolwriter.h

@@ -1,104 +0,0 @@
-/*
- * Copyright (c) 2016, Alliance for Open Media. All rights reserved
- *
- * This source code is subject to the terms of the BSD 2 Clause License and
- * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
- * was not distributed with this source code in the LICENSE file, you can
- * obtain it at www.aomedia.org/license/software. If the Alliance for Open
- * Media Patent License 1.0 was not distributed with this source code in the
- * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
- */
-
-#ifndef AOM_DSP_DKBOOLWRITER_H_
-#define AOM_DSP_DKBOOLWRITER_H_
-
-#include "./aom_config.h"
-
-#if CONFIG_BITSTREAM_DEBUG
-#include <stdio.h>
-#include "aom_util/debug_util.h"
-#endif  // CONFIG_BITSTREAM_DEBUG
-
-#include "aom_dsp/prob.h"
-#include "aom_ports/mem.h"
-
-#ifdef __cplusplus
-extern "C" {
-#endif
-
-typedef struct aom_dk_writer {
-  unsigned int lowvalue;
-  unsigned int range;
-  int count;
-  unsigned int pos;
-  uint8_t *buffer;
-} aom_dk_writer;
-
-void aom_dk_start_encode(aom_dk_writer *bc, uint8_t *buffer);
-void aom_dk_stop_encode(aom_dk_writer *bc);
-
-static INLINE void aom_dk_write(aom_dk_writer *br, int bit, int probability) {
-  unsigned int split;
-  int count = br->count;
-  unsigned int range = br->range;
-  unsigned int lowvalue = br->lowvalue;
-  register int shift;
-
-#if CONFIG_BITSTREAM_DEBUG
-  // int queue_r = 0;
-  // int frame_idx_r = 0;
-  // int queue_w = bitstream_queue_get_write();
-  // int frame_idx_w = bitstream_queue_get_frame_write();
-  // if (frame_idx_w == frame_idx_r && queue_w == queue_r) {
-  //   fprintf(stderr, "\n *** bitstream queue at frame_idx_w %d queue_w %d\n",
-  //   frame_idx_w, queue_w);
-  // }
-  bitstream_queue_push(bit, probability);
-#endif  // CONFIG_BITSTREAM_DEBUG
-
-  split = 1 + (((range - 1) * probability) >> 8);
-
-  range = split;
-
-  if (bit) {
-    lowvalue += split;
-    range = br->range - split;
-  }
-
-  shift = aom_norm[range];
-
-  range <<= shift;
-  count += shift;
-
-  if (count >= 0) {
-    int offset = shift - count;
-
-    if ((lowvalue << (offset - 1)) & 0x80000000) {
-      int x = br->pos - 1;
-
-      while (x >= 0 && br->buffer[x] == 0xff) {
-        br->buffer[x] = 0;
-        x--;
-      }
-
-      br->buffer[x] += 1;
-    }
-
-    br->buffer[br->pos++] = (lowvalue >> (24 - offset));
-    lowvalue <<= offset;
-    shift = count;
-    lowvalue &= 0xffffff;
-    count -= 8;
-  }
-
-  lowvalue <<= shift;
-  br->count = count;
-  br->lowvalue = lowvalue;
-  br->range = range;
-}
-
-#ifdef __cplusplus
-}  // extern "C"
-#endif
-
-#endif  // AOM_DSP_DKBOOLWRITER_H_

aom_dsp/entcode.c

@@ -1,80 +0,0 @@
-/*
- * Copyright (c) 2001-2016, Alliance for Open Media. All rights reserved
- *
- * This source code is subject to the terms of the BSD 2 Clause License and
- * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
- * was not distributed with this source code in the LICENSE file, you can
- * obtain it at www.aomedia.org/license/software. If the Alliance for Open
- * Media Patent License 1.0 was not distributed with this source code in the
- * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
- */
-
-#ifdef HAVE_CONFIG_H
-#include "./config.h"
-#endif
-
-#include "aom_dsp/entcode.h"
-
-/*CDFs for uniform probability distributions of small sizes (2 through 16,
-   inclusive).*/
-// clang-format off
-const uint16_t OD_UNIFORM_CDFS_Q15[135] = {
-  16384, 32768,
-  10923, 21845, 32768,
-  8192,  16384, 24576, 32768,
-  6554,  13107, 19661, 26214, 32768,
-  5461,  10923, 16384, 21845, 27307, 32768,
-  4681,   9362, 14043, 18725, 23406, 28087, 32768,
-  4096,   8192, 12288, 16384, 20480, 24576, 28672, 32768,
-  3641,   7282, 10923, 14564, 18204, 21845, 25486, 29127, 32768,
-  3277,   6554,  9830, 13107, 16384, 19661, 22938, 26214, 29491, 32768,
-  2979,   5958,  8937, 11916, 14895, 17873, 20852, 23831, 26810, 29789, 32768,
-  2731,   5461,  8192, 10923, 13653, 16384, 19115, 21845, 24576, 27307, 30037,
-  32768,
-  2521,   5041,  7562, 10082, 12603, 15124, 17644, 20165, 22686, 25206, 27727,
-  30247, 32768,
-  2341,   4681,  7022,  9362, 11703, 14043, 16384, 18725, 21065, 23406, 25746,
-  28087, 30427, 32768,
-  2185,   4369,  6554,  8738, 10923, 13107, 15292, 17476, 19661, 21845, 24030,
-  26214, 28399, 30583, 32768,
-  2048,   4096,  6144,  8192, 10240, 12288, 14336, 16384, 18432, 20480, 22528,
-  24576, 26624, 28672, 30720, 32768
-};
-// clang-format on
-
-/*Given the current total integer number of bits used and the current value of
-   rng, computes the fraction number of bits used to OD_BITRES precision.
-  This is used by od_ec_enc_tell_frac() and od_ec_dec_tell_frac().
-  nbits_total: The number of whole bits currently used, i.e., the value
-                returned by od_ec_enc_tell() or od_ec_dec_tell().
-  rng: The current value of rng from either the encoder or decoder state.
-  Return: The number of bits scaled by 2**OD_BITRES.
-          This will always be slightly larger than the exact value (e.g., all
-           rounding error is in the positive direction).*/
-uint32_t od_ec_tell_frac(uint32_t nbits_total, uint32_t rng) {
-  uint32_t nbits;
-  int l;
-  int i;
-  /*To handle the non-integral number of bits still left in the encoder/decoder
-     state, we compute the worst-case number of bits of val that must be
-     encoded to ensure that the value is inside the range for any possible
-     subsequent bits.
-    The computation here is independent of val itself (the decoder does not
-     even track that value), even though the real number of bits used after
-     od_ec_enc_done() may be 1 smaller if rng is a power of two and the
-     corresponding trailing bits of val are all zeros.
-    If we did try to track that special case, then coding a value with a
-     probability of 1/(1 << n) might sometimes appear to use more than n bits.
-    This may help explain the surprising result that a newly initialized
-     encoder or decoder claims to have used 1 bit.*/
-  nbits = nbits_total << OD_BITRES;
-  l = 0;
-  for (i = OD_BITRES; i-- > 0;) {
-    int b;
-    rng = rng * rng >> 15;
-    b = (int)(rng >> 16);
-    l = l << 1 | b;
-    rng >>= b;
-  }
-  return nbits - l;
-}

aom_dsp/entcode.h

@@ -1,105 +0,0 @@
-/*
- * Copyright (c) 2001-2016, Alliance for Open Media. All rights reserved
- *
- * This source code is subject to the terms of the BSD 2 Clause License and
- * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
- * was not distributed with this source code in the LICENSE file, you can
- * obtain it at www.aomedia.org/license/software. If the Alliance for Open
- * Media Patent License 1.0 was not distributed with this source code in the
- * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
- */
-
-#if !defined(_entcode_H)
-#define _entcode_H (1)
-#include <limits.h>
-#include <stddef.h>
-#include "av1/common/odintrin.h"
-
-/*Set this flag 1 to enable a "reduced overhead" version of the entropy coder.
-  This uses a partition function that more accurately follows the input
-   probability estimates at the expense of some additional CPU cost (though
-   still an order of magnitude less than a full division).
-
-  In classic arithmetic coding, the partition function maps a value x in the
-   range [0, ft] to a value in y in [0, r] with 0 < ft <= r via
-    y = x*r/ft.
-  Any deviation from this value increases coding inefficiency.
-
-  To avoid divisions, we require ft <= r < 2*ft (enforcing it by shifting up
-   ft if necessary), and replace that function with
-    y = x + OD_MINI(x, r - ft).
-  This counts values of x smaller than r - ft double compared to values larger
-   than r - ft, which over-estimates the probability of symbols at the start of
-   the alphabet, and under-estimates the probability of symbols at the end of
-   the alphabet.
-  The overall coding inefficiency assuming accurate probability models and
-   independent symbols is in the 1% range, which is similar to that of CABAC.
-
-  To reduce overhead even further, we split this into two cases:
-  1) r - ft > ft - (r - ft).
-     That is, we have more values of x that are double-counted than
-      single-counted.
-     In this case, we still double-count the first 2*r - 3*ft values of x, but
-      after that we alternate between single-counting and double-counting for
-      the rest.
-  2) r - ft < ft - (r - ft).
-     That is, we have more values of x that are single-counted than
-      double-counted.
-     In this case, we alternate between single-counting and double-counting for
-      the first 2*(r - ft) values of x, and single-count the rest.
-  For two equiprobable symbols in different places in the alphabet, this
-   reduces the maximum ratio of over-estimation to under-estimation from 2:1
-   for the previous partition function to either 4:3 or 3:2 (for each of the
-   two cases above, respectively), assuming symbol probabilities significantly
-   greater than 1/32768.
-  That reduces the worst-case per-symbol overhead from 1 bit to 0.58 bits.
-
-  The resulting function is
-    e = OD_MAXI(2*r - 3*ft, 0);
-    y = x + OD_MINI(x, e) + OD_MINI(OD_MAXI(x - e, 0) >> 1, r - ft).
-  Here, e is a value that is greater than 0 in case 1, and 0 in case 2.
-  This function is about 3 times as expensive to evaluate as the high-overhead
-   version, but still an order of magnitude cheaper than a division, since it
-   is composed only of very simple operations.
-  Because we want to fit in 16-bit registers and must use unsigned values to do
-   so, we use saturating subtraction to enforce the maximums with 0.
-
-  Enabling this reduces the measured overhead in ectest from 0.805% to 0.621%
-   (vs. 0.022% for the division-based partition function with r much greater
-   than ft).
-  It improves performance on ntt-short-1 by about 0.3%.*/
-#define OD_EC_REDUCED_OVERHEAD (1)
-
-/*OPT: od_ec_window must be at least 32 bits, but if you have fast arithmetic
-   on a larger type, you can speed up the decoder by using it here.*/
-typedef uint32_t od_ec_window;
-
-#define OD_EC_WINDOW_SIZE ((int)sizeof(od_ec_window) * CHAR_BIT)
-
-/*Unsigned subtraction with unsigned saturation.
-  This implementation of the macro is intentionally chosen to increase the
-   number of common subexpressions in the reduced-overhead partition function.
-  This matters for C code, but it would not for hardware with a saturating
-   subtraction instruction.*/
-#define OD_SUBSATU(a, b) ((a)-OD_MINI(a, b))
-
-/*The number of bits to use for the range-coded part of unsigned integers.*/
-#define OD_EC_UINT_BITS (4)
-
-/*The resolution of fractional-precision bit usage measurements, i.e.,
-   3 => 1/8th bits.*/
-#define OD_BITRES (3)
-
-extern const uint16_t OD_UNIFORM_CDFS_Q15[135];
-
-/*Returns a Q15 CDF for a uniform probability distribution of the given size.
-  n: The size of the distribution.
-     This must be at least 2, and no more than 16.*/
-#define OD_UNIFORM_CDF_Q15(n) (OD_UNIFORM_CDFS_Q15 + ((n) * ((n)-1) >> 1) - 1)
-
-/*See entcode.c for further documentation.*/
-
-OD_WARN_UNUSED_RESULT uint32_t od_ec_tell_frac(uint32_t nbits_total,
-                                               uint32_t rng);
-
-#endif

aom_dsp/entdec.c

@@ -1,494 +0,0 @@
-/*
- * Copyright (c) 2001-2016, Alliance for Open Media. All rights reserved
- *
- * This source code is subject to the terms of the BSD 2 Clause License and
- * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
- * was not distributed with this source code in the LICENSE file, you can
- * obtain it at www.aomedia.org/license/software. If the Alliance for Open
- * Media Patent License 1.0 was not distributed with this source code in the
- * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
- */
-
-#ifdef HAVE_CONFIG_H
-#include "./config.h"
-#endif
-
-#include "aom_dsp/entdec.h"
-
-/*A range decoder.
-  This is an entropy decoder based upon \cite{Mar79}, which is itself a
-   rediscovery of the FIFO arithmetic code introduced by \cite{Pas76}.
-  It is very similar to arithmetic encoding, except that encoding is done with
-   digits in any base, instead of with bits, and so it is faster when using
-   larger bases (i.e.: a byte).
-  The author claims an average waste of $\frac{1}{2}\log_b(2b)$ bits, where $b$
-   is the base, longer than the theoretical optimum, but to my knowledge there
-   is no published justification for this claim.
-  This only seems true when using near-infinite precision arithmetic so that
-   the process is carried out with no rounding errors.
-
-  An excellent description of implementation details is available at
-   http://www.arturocampos.com/ac_range.html
-  A recent work \cite{MNW98} which proposes several changes to arithmetic
-   encoding for efficiency actually re-discovers many of the principles
-   behind range encoding, and presents a good theoretical analysis of them.
-
-  End of stream is handled by writing out the smallest number of bits that
-   ensures that the stream will be correctly decoded regardless of the value of
-   any subsequent bits.
-  od_ec_dec_tell() can be used to determine how many bits were needed to decode
-   all the symbols thus far; other data can be packed in the remaining bits of
-   the input buffer.
-  @PHDTHESIS{Pas76,
-    author="Richard Clark Pasco",
-    title="Source coding algorithms for fast data compression",
-    school="Dept. of Electrical Engineering, Stanford University",
-    address="Stanford, CA",
-    month=May,
-    year=1976,
-    URL="http://www.richpasco.org/scaffdc.pdf"
-  }
-  @INPROCEEDINGS{Mar79,
-   author="Martin, G.N.N.",
-   title="Range encoding: an algorithm for removing redundancy from a digitised
-    message",
-   booktitle="Video & Data Recording Conference",
-   year=1979,
-   address="Southampton",
-   month=Jul,
-   URL="http://www.compressconsult.com/rangecoder/rngcod.pdf.gz"
-  }
-  @ARTICLE{MNW98,
-   author="Alistair Moffat and Radford Neal and Ian H. Witten",
-   title="Arithmetic Coding Revisited",
-   journal="{ACM} Transactions on Information Systems",
-   year=1998,
-   volume=16,
-   number=3,
-   pages="256--294",
-   month=Jul,
-   URL="http://researchcommons.waikato.ac.nz/bitstream/handle/10289/78/content.pdf"
-  }*/
-
-/*This is meant to be a large, positive constant that can still be efficiently
-   loaded as an immediate (on platforms like ARM, for example).
-  Even relatively modest values like 100 would work fine.*/
-#define OD_EC_LOTS_OF_BITS (0x4000)
-
-static void od_ec_dec_refill(od_ec_dec *dec) {
-  int s;
-  od_ec_window dif;
-  int16_t cnt;
-  const unsigned char *bptr;
-  const unsigned char *end;
-  dif = dec->dif;
-  cnt = dec->cnt;
-  bptr = dec->bptr;
-  end = dec->end;
-  s = OD_EC_WINDOW_SIZE - 9 - (cnt + 15);
-  for (; s >= 0 && bptr < end; s -= 8, bptr++) {
-    OD_ASSERT(s <= OD_EC_WINDOW_SIZE - 8);
-    dif |= (od_ec_window)bptr[0] << s;
-    cnt += 8;
-  }
-  if (bptr >= end) {
-    dec->tell_offs += OD_EC_LOTS_OF_BITS - cnt;
-    cnt = OD_EC_LOTS_OF_BITS;
-  }
-  dec->dif = dif;
-  dec->cnt = cnt;
-  dec->bptr = bptr;
-}
-
-/*Takes updated dif and range values, renormalizes them so that
-   32768 <= rng < 65536 (reading more bytes from the stream into dif if
-   necessary), and stores them back in the decoder context.
-  dif: The new value of dif.
-  rng: The new value of the range.
-  ret: The value to return.
-  Return: ret.
-          This allows the compiler to jump to this function via a tail-call.*/
-static int od_ec_dec_normalize(od_ec_dec *dec, od_ec_window dif, unsigned rng,
-                               int ret) {
-  int d;
-  OD_ASSERT(rng <= 65535U);
-  d = 16 - OD_ILOG_NZ(rng);
-  dec->cnt -= d;
-  dec->dif = dif << d;
-  dec->rng = rng << d;
-  if (dec->cnt < 0) od_ec_dec_refill(dec);
-  return ret;
-}
-
-/*Initializes the decoder.
-  buf: The input buffer to use.
-  Return: 0 on success, or a negative value on error.*/
-void od_ec_dec_init(od_ec_dec *dec, const unsigned char *buf,
-                    uint32_t storage) {
-  dec->buf = buf;
-  dec->eptr = buf + storage;
-  dec->end_window = 0;
-  dec->nend_bits = 0;
-  dec->tell_offs = 10 - (OD_EC_WINDOW_SIZE - 8);
-  dec->end = buf + storage;
-  dec->bptr = buf;
-  dec->dif = 0;
-  dec->rng = 0x8000;
-  dec->cnt = -15;
-  dec->error = 0;
-  od_ec_dec_refill(dec);
-}
-
-/*Decode a bit that has an fz/ft probability of being a zero.
-  fz: The probability that the bit is zero, scaled by _ft.
-  ft: The total probability.
-      This must be at least 16384 and no more than 32768.
-  Return: The value decoded (0 or 1).*/
-int od_ec_decode_bool(od_ec_dec *dec, unsigned fz, unsigned ft) {
-  od_ec_window dif;
-  od_ec_window vw;
-  unsigned r;
-  int s;
-  unsigned v;
-  int ret;
-  OD_ASSERT(0 < fz);
-  OD_ASSERT(fz < ft);
-  OD_ASSERT(16384 <= ft);
-  OD_ASSERT(ft <= 32768U);
-  dif = dec->dif;
-  r = dec->rng;
-  OD_ASSERT(dif >> (OD_EC_WINDOW_SIZE - 16) < r);
-  OD_ASSERT(ft <= r);
-  s = r - ft >= ft;
-  ft <<= s;
-  fz <<= s;
-  OD_ASSERT(r - ft < ft);
-#if OD_EC_REDUCED_OVERHEAD
-  {
-    unsigned d;
-    unsigned e;
-    d = r - ft;
-    e = OD_SUBSATU(2 * d, ft);
-    v = fz + OD_MINI(fz, e) + OD_MINI(OD_SUBSATU(fz, e) >> 1, d);
-  }
-#else
-  v = fz + OD_MINI(fz, r - ft);
-#endif
-  vw = (od_ec_window)v << (OD_EC_WINDOW_SIZE - 16);
-  ret = dif >= vw;
-  if (ret) dif -= vw;
-  r = ret ? r - v : v;
-  return od_ec_dec_normalize(dec, dif, r, ret);
-}
-
-/*Decode a bit that has an fz probability of being a zero in Q15.
-  This is a simpler, lower overhead version of od_ec_decode_bool() for use when
-   ft == 32768.
-  To be decoded properly by this function, symbols cannot have been encoded by
-   od_ec_encode(), but must have been encoded with one of the equivalent _q15()
-   or _dyadic() functions instead.
-  fz: The probability that the bit is zero, scaled by 32768.
-  Return: The value decoded (0 or 1).*/
-int od_ec_decode_bool_q15(od_ec_dec *dec, unsigned fz) {
-  od_ec_window dif;
-  od_ec_window vw;
-  unsigned r;
-  unsigned r_new;
-  unsigned v;
-  int ret;
-  OD_ASSERT(0 < fz);
-  OD_ASSERT(fz < 32768U);
-  dif = dec->dif;
-  r = dec->rng;
-  OD_ASSERT(dif >> (OD_EC_WINDOW_SIZE - 16) < r);
-  OD_ASSERT(32768U <= r);
-  v = fz * (uint32_t)r >> 15;
-  vw = (od_ec_window)v << (OD_EC_WINDOW_SIZE - 16);
-  ret = 0;
-  r_new = v;
-  if (dif >= vw) {
-    r_new = r - v;
-    dif -= vw;
-    ret = 1;
-  }
-  return od_ec_dec_normalize(dec, dif, r_new, ret);
-}
-
-/*Decodes a symbol given a cumulative distribution function (CDF) table.
-  cdf: The CDF, such that symbol s falls in the range
-        [s > 0 ? cdf[s - 1] : 0, cdf[s]).
-       The values must be monotonically non-increasing, and cdf[nsyms - 1]
-        must be at least 16384, and no more than 32768.
-  nsyms: The number of symbols in the alphabet.
-         This should be at most 16.
-  Return: The decoded symbol s.*/
-int od_ec_decode_cdf(od_ec_dec *dec, const uint16_t *cdf, int nsyms) {
-  od_ec_window dif;
-  unsigned r;
-  unsigned c;
-  unsigned d;
-#if OD_EC_REDUCED_OVERHEAD
-  unsigned e;
-#endif
-  int s;
-  unsigned u;
-  unsigned v;
-  unsigned q;
-  unsigned fl;
-  unsigned fh;
-  unsigned ft;
-  int ret;
-  dif = dec->dif;
-  r = dec->rng;
-  OD_ASSERT(dif >> (OD_EC_WINDOW_SIZE - 16) < r);
-  OD_ASSERT(nsyms > 0);
-  ft = cdf[nsyms - 1];
-  OD_ASSERT(16384 <= ft);
-  OD_ASSERT(ft <= 32768U);
-  OD_ASSERT(ft <= r);
-  s = r - ft >= ft;
-  ft <<= s;
-  d = r - ft;
-  OD_ASSERT(d < ft);
-  c = (unsigned)(dif >> (OD_EC_WINDOW_SIZE - 16));
-  q = OD_MAXI((int)(c >> 1), (int)(c - d));
-#if OD_EC_REDUCED_OVERHEAD
-  e = OD_SUBSATU(2 * d, ft);
-  /*The correctness of this inverse partition function is not obvious, but it
-     was checked exhaustively for all possible values of r, ft, and c.
-    TODO: It should be possible to optimize this better than the compiler,
-     given that we do not care about the accuracy of negative results (as we
-     will not use them).
-    It would also be nice to get rid of the 32-bit dividend, as it requires a
-     32x32->64 bit multiply to invert.*/
-  q = OD_MAXI((int)q, (int)((2 * (int32_t)c + 1 - (int32_t)e) / 3));
-#endif
-  q >>= s;
-  OD_ASSERT(q<ft>> s);
-  fl = 0;
-  ret = 0;
-  for (fh = cdf[ret]; fh <= q; fh = cdf[++ret]) fl = fh;
-  OD_ASSERT(fh <= ft >> s);
-  fl <<= s;
-  fh <<= s;
-#if OD_EC_REDUCED_OVERHEAD
-  u = fl + OD_MINI(fl, e) + OD_MINI(OD_SUBSATU(fl, e) >> 1, d);
-  v = fh + OD_MINI(fh, e) + OD_MINI(OD_SUBSATU(fh, e) >> 1, d);
-#else
-  u = fl + OD_MINI(fl, d);
-  v = fh + OD_MINI(fh, d);
-#endif
-  r = v - u;
-  dif -= (od_ec_window)u << (OD_EC_WINDOW_SIZE - 16);
-  return od_ec_dec_normalize(dec, dif, r, ret);
-}
-
-/*Decodes a symbol given a cumulative distribution function (CDF) table.
-  cdf: The CDF, such that symbol s falls in the range
-        [s > 0 ? cdf[s - 1] : 0, cdf[s]).
-       The values must be monotonically non-increasing, and cdf[nsyms - 1]
-       must be at least 2, and no more than 32768.
-  nsyms: The number of symbols in the alphabet.
-         This should be at most 16.
-  Return: The decoded symbol s.*/
-int od_ec_decode_cdf_unscaled(od_ec_dec *dec, const uint16_t *cdf, int nsyms) {
-  od_ec_window dif;
-  unsigned r;
-  unsigned c;
-  unsigned d;
-#if OD_EC_REDUCED_OVERHEAD
-  unsigned e;
-#endif
-  int s;
-  unsigned u;
-  unsigned v;
-  unsigned q;
-  unsigned fl;
-  unsigned fh;
-  unsigned ft;
-  int ret;
-  dif = dec->dif;
-  r = dec->rng;
-  OD_ASSERT(dif >> (OD_EC_WINDOW_SIZE - 16) < r);
-  OD_ASSERT(nsyms > 0);
-  ft = cdf[nsyms - 1];
-  OD_ASSERT(2 <= ft);
-  OD_ASSERT(ft <= 32768U);
-  s = 15 - OD_ILOG_NZ(ft - 1);
-  ft <<= s;
-  OD_ASSERT(ft <= r);
-  if (r - ft >= ft) {
-    ft <<= 1;
-    s++;
-  }
-  d = r - ft;
-  OD_ASSERT(d < ft);
-  c = (unsigned)(dif >> (OD_EC_WINDOW_SIZE - 16));
-  q = OD_MAXI((int)(c >> 1), (int)(c - d));
-#if OD_EC_REDUCED_OVERHEAD
-  e = OD_SUBSATU(2 * d, ft);
-  /*TODO: See TODO above.*/
-  q = OD_MAXI((int)q, (int)((2 * (int32_t)c + 1 - (int32_t)e) / 3));
-#endif
-  q >>= s;
-  OD_ASSERT(q<ft>> s);
-  fl = 0;
-  ret = 0;
-  for (fh = cdf[ret]; fh <= q; fh = cdf[++ret]) fl = fh;
-  OD_ASSERT(fh <= ft >> s);
-  fl <<= s;
-  fh <<= s;
-#if OD_EC_REDUCED_OVERHEAD
-  u = fl + OD_MINI(fl, e) + OD_MINI(OD_SUBSATU(fl, e) >> 1, d);
-  v = fh + OD_MINI(fh, e) + OD_MINI(OD_SUBSATU(fh, e) >> 1, d);
-#else
-  u = fl + OD_MINI(fl, d);
-  v = fh + OD_MINI(fh, d);
-#endif
-  r = v - u;
-  dif -= (od_ec_window)u << (OD_EC_WINDOW_SIZE - 16);
-  return od_ec_dec_normalize(dec, dif, r, ret);
-}
-
-/*Decodes a symbol given a cumulative distribution function (CDF) table that
-   sums to a power of two.
-  This is a simpler, lower overhead version of od_ec_decode_cdf() for use when
-   cdf[nsyms - 1] is a power of two.
-  To be decoded properly by this function, symbols cannot have been encoded by
-   od_ec_encode(), but must have been encoded with one of the equivalent _q15()
-   functions instead.
-  cdf: The CDF, such that symbol s falls in the range
-        [s > 0 ? cdf[s - 1] : 0, cdf[s]).
-       The values must be monotonically non-increasing, and cdf[nsyms - 1]
-       must be exactly 1 << ftb.
-  nsyms: The number of symbols in the alphabet.
-         This should be at most 16.
-  ftb: The number of bits of precision in the cumulative distribution.
-       This must be no more than 15.
-  Return: The decoded symbol s.*/
-int od_ec_decode_cdf_unscaled_dyadic(od_ec_dec *dec, const uint16_t *cdf,
-                                     int nsyms, unsigned ftb) {
-  od_ec_window dif;
-  unsigned r;
-  unsigned c;
-  unsigned u;
-  unsigned v;
-  int ret;
-  (void)nsyms;
-  dif = dec->dif;
-  r = dec->rng;
-  OD_ASSERT(dif >> (OD_EC_WINDOW_SIZE - 16) < r);
-  OD_ASSERT(ftb <= 15);
-  OD_ASSERT(cdf[nsyms - 1] == 1U << ftb);
-  OD_ASSERT(32768U <= r);
-  c = (unsigned)(dif >> (OD_EC_WINDOW_SIZE - 16));
-  v = 0;
-  ret = -1;
-  do {
-    u = v;
-    v = cdf[++ret] * (uint32_t)r >> ftb;
-  } while (v <= c);
-  OD_ASSERT(v <= r);
-  r = v - u;
-  dif -= (od_ec_window)u << (OD_EC_WINDOW_SIZE - 16);
-  return od_ec_dec_normalize(dec, dif, r, ret);
-}
-
-/*Decodes a symbol given a cumulative distribution function (CDF) table in Q15.
-  This is a simpler, lower overhead version of od_ec_decode_cdf() for use when
-   cdf[nsyms - 1] == 32768.
-  To be decoded properly by this function, symbols cannot have been encoded by
-   od_ec_encode(), but must have been encoded with one of the equivalent _q15()
-   or dyadic() functions instead.
-  cdf: The CDF, such that symbol s falls in the range
-        [s > 0 ? cdf[s - 1] : 0, cdf[s]).
-       The values must be monotonically non-increasing, and cdf[nsyms - 1]
-        must be 32768.
-  nsyms: The number of symbols in the alphabet.
-         This should be at most 16.
-  Return: The decoded symbol s.*/
-int od_ec_decode_cdf_q15(od_ec_dec *dec, const uint16_t *cdf, int nsyms) {
-  return od_ec_decode_cdf_unscaled_dyadic(dec, cdf, nsyms, 15);
-}
-
-/*Extracts a raw unsigned integer with a non-power-of-2 range from the stream.
-  The integer must have been encoded with od_ec_enc_uint().
-  ft: The number of integers that can be decoded (one more than the max).
-      This must be at least 2, and no more than 2**29.
-  Return: The decoded bits.*/
-uint32_t od_ec_dec_uint(od_ec_dec *dec, uint32_t ft) {
-  OD_ASSERT(ft >= 2);
-  OD_ASSERT(ft <= (uint32_t)1 << (25 + OD_EC_UINT_BITS));
-  if (ft > 1U << OD_EC_UINT_BITS) {
-    uint32_t t;
-    int ft1;
-    int ftb;
-    ft--;
-    ftb = OD_ILOG_NZ(ft) - OD_EC_UINT_BITS;
-    ft1 = (int)(ft >> ftb) + 1;
-    t = od_ec_decode_cdf_q15(dec, OD_UNIFORM_CDF_Q15(ft1), ft1);
-    t = t << ftb | od_ec_dec_bits(dec, ftb, "");
-    if (t <= ft) return t;
-    dec->error = 1;
-    return ft;
-  }
-  return od_ec_decode_cdf_q15(dec, OD_UNIFORM_CDF_Q15(ft), (int)ft);
-}
-
-/*Extracts a sequence of raw bits from the stream.
-  The bits must have been encoded with od_ec_enc_bits().
-  ftb: The number of bits to extract.
-       This must be between 0 and 25, inclusive.
-  Return: The decoded bits.*/
-uint32_t od_ec_dec_bits_(od_ec_dec *dec, unsigned ftb) {
-  od_ec_window window;
-  int available;
-  uint32_t ret;
-  OD_ASSERT(ftb <= 25);
-  window = dec->end_window;
-  available = dec->nend_bits;
-  if ((unsigned)available < ftb) {
-    const unsigned char *buf;
-    const unsigned char *eptr;
-    buf = dec->buf;
-    eptr = dec->eptr;
-    OD_ASSERT(available <= OD_EC_WINDOW_SIZE - 8);
-    do {
-      if (eptr <= buf) {
-        dec->tell_offs += OD_EC_LOTS_OF_BITS - available;
-        available = OD_EC_LOTS_OF_BITS;
-        break;
-      }
-      window |= (od_ec_window) * --eptr << available;
-      available += 8;
-    } while (available <= OD_EC_WINDOW_SIZE - 8);
-    dec->eptr = eptr;
-  }
-  ret = (uint32_t)window & (((uint32_t)1 << ftb) - 1);
-  window >>= ftb;
-  available -= ftb;
-  dec->end_window = window;
-  dec->nend_bits = available;
-  return ret;
-}
-
-/*Returns the number of bits "used" by the decoded symbols so far.
-  This same number can be computed in either the encoder or the decoder, and is
-   suitable for making coding decisions.
-  Return: The number of bits.
-          This will always be slightly larger than the exact value (e.g., all
-           rounding error is in the positive direction).*/
-int od_ec_dec_tell(const od_ec_dec *dec) {
-  return ((dec->end - dec->eptr) + (dec->bptr - dec->buf)) * 8 - dec->cnt -
-         dec->nend_bits + dec->tell_offs;
-}
-
-/*Returns the number of bits "used" by the decoded symbols so far.
-  This same number can be computed in either the encoder or the decoder, and is
-   suitable for making coding decisions.
-  Return: The number of bits scaled by 2**OD_BITRES.
-          This will always be slightly larger than the exact value (e.g., all
-           rounding error is in the positive direction).*/
-uint32_t od_ec_dec_tell_frac(const od_ec_dec *dec) {
-  return od_ec_tell_frac(od_ec_dec_tell(dec), dec->rng);
-}