@@ -1,4 +1,40 @@

 -------------------------------------------------------------------

+Mon Jun  1 17:51:22 UTC 2020 - Luigi Baldoni <aloisio@gmx.com>

+

+- Update to version 3.4

+  New features:

+  * Edge-aware quadtree partitioning to terminate CU depth

+    recursion based on edge information. --rskip level 2 enables

+    the feature and --rskip-edge-threshold denotes the minimum

+    expected edge-density percentage within the CU, below which

+    the recursion is skipped. Experimental feature.

+  * Application-level feature --abr-ladder for automating

+    efficient ABR ladder generation. Shows ~65% savings in the

+    over-all turn-around time required for the generation of a

+    typical Apple HLS ladder in Intel(R) Xeon(R) Platinum 8280

+    CPU @ 2.70GHz over a sequential ABR-ladder generation

+    approach that leverages save-load architecture.

+  Enhancements to existing features:

+  * Improved efficiency in 2-pass rate-control algorithm. The

+    savings in the bitrate is ~1.72% with visual improvement in

+    quality in the initial 1-2 secs.

+  Encoder enhancements:

+  * Faster ARM64 encodes enabled by ASM contributions from

+    Huawei. The speed-up over no-asm version for 1080p encodes @

+    medium preset is ~15% in a 16 core H/W.

+  * Strict VBV conformance in zone encoding.

+  Bug fixes:

+  * Multi-pass encode failures with --frame-dup.

+  * Corrupted bitstreams with --hist-scenecut when input depth

+    and internal bit-depth differ.

+  * Incorrect analysis propagation in multi-level save-load

+    architecture.

+  * Failure in detecting NUMA packages installed in non-standard

+    directories.

+

+- Refreshed arm.patch

+

+-------------------------------------------------------------------

 Sat Mar 28 14:28:56 UTC 2020 - Luigi Baldoni <aloisio@gmx.com>

 - Update to version 3.3

​

@@ -17,11 +17,11 @@

 #

-%define sover  188

+%define sover  192

 %define libname lib%{name}

 %define libsoname %{libname}-%{sover}

 Name:           x265

-Version:        3.3

+Version:        3.4

 Release:        0

 Summary:        A free h265/HEVC encoder - encoder binary

 License:        GPL-2.0-or-later

@@ -67,7 +67,6 @@

 %patch0 -p1

 %patch1 -p1

 %patch2 -p1

-

 sed -i -e "s/0.0/%{sover}.0/g" source/cmake/version.cmake

​

@@ -1,8 +1,8 @@

-Index: x265_2.2/source/CMakeLists.txt

+Index: x265_3.4/source/CMakeLists.txt

 ===================================================================

---- x265_2.2.orig/source/CMakeLists.txt

-+++ x265_2.2/source/CMakeLists.txt

-@@ -65,15 +65,22 @@ elseif(POWERMATCH GREATER "-1")

+--- x265_3.4.orig/source/CMakeLists.txt

++++ x265_3.4/source/CMakeLists.txt

+@@ -64,26 +64,26 @@ elseif(POWERMATCH GREATER "-1")

          add_definitions(-DPPC64=1)

          message(STATUS "Detected POWER PPC64 target processor")

      endif()

@@ -12,41 +12,62 @@

 -    else()

 -        set(CROSS_COMPILE_ARM 0)

 -    endif()

--    message(STATUS "Detected ARM target processor")

 -    set(ARM 1)

--    add_definitions(-DX265_ARCH_ARM=1 -DHAVE_ARMV6=1)

+-    if("${CMAKE_SIZEOF_VOID_P}" MATCHES 8)

+-        message(STATUS "Detected ARM64 target processor")

+-        set(ARM64 1)

+-        add_definitions(-DX265_ARCH_ARM=1 -DX265_ARCH_ARM64=1 -DHAVE_ARMV6=0)

+-    else()

+-        message(STATUS "Detected ARM target processor")

+-        add_definitions(-DX265_ARCH_ARM=1 -DX265_ARCH_ARM64=0 -DHAVE_ARMV6=1)

+-    endif()

 +elseif(${SYSPROC} MATCHES "armv5.*")

 +    message(STATUS "Detected ARMV5 system processor")

 +    set(ARMV5 1)

-+    add_definitions(-DX265_ARCH_ARM=1 -DHAVE_ARMV6=0 -DHAVE_NEON=0)

++    add_definitions(-DX265_ARCH_ARM=1 -DX265_ARCH_ARM64=0 -DHAVE_ARMV6=0 -DHAVE_NEON=0)

 +elseif(${SYSPROC} STREQUAL "armv6l")

 +    message(STATUS "Detected ARMV6 system processor")

 +    set(ARMV6 1)

-+    add_definitions(-DX265_ARCH_ARM=1 -DHAVE_ARMV6=1 -DHAVE_NEON=0)

++    add_definitions(-DX265_ARCH_ARM=1 -DX265_ARCH_ARM64=0 -DHAVE_ARMV6=1 -DHAVE_NEON=0)

 +elseif(${SYSPROC} STREQUAL "armv7l")

 +    message(STATUS "Detected ARMV7 system processor")

 +    set(ARMV7 1)

-+    add_definitions(-DX265_ARCH_ARM=1 -DHAVE_ARMV6=1 -DHAVE_NEON=0)

++    add_definitions(-DX265_ARCH_ARM=1 -DX265_ARCH_ARM64=0 -DHAVE_ARMV6=1 -DHAVE_NEON=0)

 +elseif(${SYSPROC} STREQUAL "aarch64")

 +    message(STATUS "Detected AArch64 system processor")

 +    set(ARMV7 1)

-+    add_definitions(-DX265_ARCH_ARM=1 -DHAVE_ARMV6=1 -DHAVE_NEON=0)

++    add_definitions(-DX265_ARCH_ARM=1 -DX265_ARCH_ARM64=1 -DHAVE_ARMV6=0 -DHAVE_NEON=0)

  else()

      message(STATUS "CMAKE_SYSTEM_PROCESSOR value `${CMAKE_SYSTEM_PROCESSOR}` is unknown")

      message(STATUS "Please add this value near ${CMAKE_CURRENT_LIST_FILE}:${CMAKE_CURRENT_LIST_LINE}")

-@@ -208,18 +215,9 @@ if(GCC)

+ endif()

+-

+ if(UNIX)

+     list(APPEND PLATFORM_LIBS pthread)

+     find_library(LIBRT rt)

+@@ -238,28 +238,9 @@ if(GCC)

              endif()

          endif()

      endif()

 -    if(ARM AND CROSS_COMPILE_ARM)

--        set(ARM_ARGS -march=armv6 -mfloat-abi=soft -mfpu=vfp -marm -fPIC)

+-        if(ARM64)

+-            set(ARM_ARGS -fPIC)

+-        else()

+-            set(ARM_ARGS -march=armv6 -mfloat-abi=soft -mfpu=vfp -marm -fPIC)

+-        endif()

+-        message(STATUS "cross compile arm")

 -    elseif(ARM)

--        find_package(Neon)

--        if(CPU_HAS_NEON)

--            set(ARM_ARGS -mcpu=native -mfloat-abi=hard -mfpu=neon -marm -fPIC)

+-        if(ARM64)

+-            set(ARM_ARGS -fPIC)

 -            add_definitions(-DHAVE_NEON)

 -        else()

--            set(ARM_ARGS -mcpu=native -mfloat-abi=hard -mfpu=vfp -marm)

+-            find_package(Neon)

+-            if(CPU_HAS_NEON)

+-                set(ARM_ARGS -mcpu=native -mfloat-abi=hard -mfpu=neon -marm -fPIC)

+-                add_definitions(-DHAVE_NEON)

+-            else()

+-                set(ARM_ARGS -mcpu=native -mfloat-abi=hard -mfpu=vfp -marm)

+-            endif()

 -        endif()

 +    if(ARMV7)

 +        add_definitions(-fPIC)

@@ -55,11 +76,11 @@

      if(FPROFILE_GENERATE)

          if(INTEL_CXX)

              add_definitions(-prof-gen -prof-dir="${CMAKE_CURRENT_BINARY_DIR}")

-Index: x265_2.2/source/common/cpu.cpp

+Index: x265_3.4/source/common/cpu.cpp

 ===================================================================

---- x265_2.2.orig/source/common/cpu.cpp

-+++ x265_2.2/source/common/cpu.cpp

-@@ -37,7 +37,7 @@

+--- x265_3.4.orig/source/common/cpu.cpp

++++ x265_3.4/source/common/cpu.cpp

+@@ -39,7 +39,7 @@

  #include <machine/cpu.h>

  #endif

@@ -68,7 +89,7 @@

  #include <signal.h>

  #include <setjmp.h>

  static sigjmp_buf jmpbuf;

-@@ -344,7 +344,6 @@ uint32_t cpu_detect(void)

+@@ -350,7 +350,6 @@ uint32_t cpu_detect(bool benableavx512)

      }

      canjump = 1;

@@ -76,7 +97,7 @@

      canjump = 0;

      signal(SIGILL, oldsig);

  #endif // if !HAVE_NEON

-@@ -360,7 +359,7 @@ uint32_t cpu_detect(void)

+@@ -366,7 +365,7 @@ uint32_t cpu_detect(bool benableavx512)

      // which may result in incorrect detection and the counters stuck enabled.

      // right now Apple does not seem to support performance counters for this test

  #ifndef __MACH__

@@ -84,4 +105,4 @@

 +    //flags |= PFX(cpu_fast_neon_mrc_test)() ? X265_CPU_FAST_NEON_MRC : 0;

  #endif

      // TODO: write dual issue test? currently it's A8 (dual issue) vs. A9 (fast mrc)

- #endif // if HAVE_ARMV6

+ #elif X265_ARCH_ARM64

​

@@ -1,1 +1,1 @@

-libx265-179

+libx265-192

​

@@ -1,5 +1,4 @@

 repo: 09fe40627f03a0f9c3e6ac78b22ac93da23f9fdf

-node: f94b0d32737d40b2b9a9d74df57fee45e6be5cb0

-branch: Release_3.3

-latesttag: 3.3

-latesttagdistance: 1

+node: 2a65b720985096bcb1664f7cb05c3d04aeb576f5

+branch: Release_3.4

+tag: 3.4

​

@@ -40,3 +40,4 @@

 5ee3593ebd82b4d8957909bbc1b68b99b59ba773 3.3_RC1

 96a10df63c0b778b480330bdf3be8da7db8a5fb1 3.3_RC2

 057215961bc4b51b6260a584ff3d506e6d65cfd6 3.3

+ee92f36782800f145970131e01c79955a3ed5c10 3.4_RC1

​

@@ -0,0 +1,15 @@

+# CMake toolchain file for cross compiling x265 for aarch64

+# This feature is only supported as experimental. Use with caution.

+# Please report bugs on bitbucket

+# Run cmake with: cmake -DCMAKE_TOOLCHAIN_FILE=crosscompile.cmake -G "Unix Makefiles" ../../source && ccmake ../../source

+

+set(CROSS_COMPILE_ARM 1)

+set(CMAKE_SYSTEM_NAME Linux)

+set(CMAKE_SYSTEM_PROCESSOR aarch64)

+

+# specify the cross compiler

+set(CMAKE_C_COMPILER aarch64-linux-gnu-gcc)

+set(CMAKE_CXX_COMPILER aarch64-linux-gnu-g++)

+

+# specify the target environment

+SET(CMAKE_FIND_ROOT_PATH  /usr/aarch64-linux-gnu)

​

@@ -0,0 +1,4 @@

+#!/bin/bash

+# Run this from within a bash shell

+

+cmake -DCMAKE_TOOLCHAIN_FILE="crosscompile.cmake" -G "Unix Makefiles" ../../source && ccmake ../../source

​

@@ -107,6 +107,9 @@

    **BufferFillFinal** Buffer bits available after removing the frame out of CPB.

+   **UnclippedBufferFillFinal** Unclipped buffer bits available after removing the frame 

+   out of CPB only used for csv logging purpose.

+   

    **Latency** Latency in terms of number of frames between when the frame 

    was given in and when the frame is given out.

@@ -842,15 +845,31 @@

    Measure 2Nx2N merge candidates first; if no residual is found, 

    additional modes at that depth are not analysed. Default disabled

-.. option:: --rskip, --no-rskip

+.. option:: --rskip <0|1|2>

+

+   This option determines early exit from CU depth recursion in modes 1 and 2. When a skip CU is

+   found, additional heuristics (depending on the RD level and rskip mode) are used to decide whether

+   to terminate recursion. The following table summarizes the behavior.

+   

+   +----------+------------+----------------------------------------------------------------+

+   | RD Level | Rskip Mode |   Skip Recursion Heuristic                                     |

+   +==========+============+================================================================+

+   |   0 - 4  |      1     |   Neighbour costs and CU homogenity.                           |

+   +----------+------------+----------------------------------------------------------------+

+   |   5 - 6  |      1     |   Comparison with inter2Nx2N.                                  |

+   +----------+------------+----------------------------------------------------------------+

+   |   0 - 6  |      2     |   CU edge density.                                             |

+   +----------+------------+----------------------------------------------------------------+

+

+   Provides minimal quality degradation at good performance gains for non-zero modes.

+   :option:`--rskip mode 0` means disabled. Default: 1, disabled when :option:`--tune grain` is used.

+   This is a integer value representing the edge-density percentage within the CU. Internally normalized to a number between 0.0 to 1.0 in x265. 

+   Recommended low thresholds for slow encodes and high for fast encodes.

-   This option determines early exit from CU depth recursion. When a skip CU is

-   found, additional heuristics (depending on rd-level) are used to decide whether

-   to terminate recursion. In rdlevels 5 and 6, comparison with inter2Nx2N is used, 

-   while at rdlevels 4 and neighbour costs are used to skip recursion.

-   Provides minimal quality degradation at good performance gains when enabled. 

+.. option:: --rskip-edge-threshold <0..100>

-   Default: enabled, disabled for :option:`--tune grain`

+   Denotes the minimum expected edge-density percentage within the CU, below which the recursion is skipped.

+   Default: 5, requires :option:`--rskip mode 2` to be enabled.

 .. option:: --splitrd-skip, --no-splitrd-skip

@@ -2501,6 +2520,28 @@

    --recon-y4m-exec "ffplay -i pipe:0 -autoexit"

    **CLI ONLY**

+   

+ABR-ladder Options

+==================

+

+.. option:: --abr-ladder <filename>

+

+   File containing the encoder configurations to generate ABR ladder.

+   The format of each line is:

+

+   **<encID:reuse-level:refID> <CLI>**

+   

+   where, encID indicates the unique name given to the encode, refID indicates

+   the name of the encode from which analysis info has to be re-used ( set to 'nil'

+   if analysis reuse isn't preferred ), and reuse-level indicates the level ( :option:`--analysis-load-reuse-level`)

+   at which analysis info has to be reused.

+   

+   A sample config file is available in `the downloads page <https://bitbucket.org/multicoreware/x265/downloads/Sample_ABR_ladder_config>`_

+   

+   Default: Disabled ( Conventional single encode generation ). Experimental feature.

+

+   **CLI ONLY**

+

 SVT-HEVC Encoder Options

 ========================

​

@@ -2,6 +2,32 @@

 Release Notes

 *************

+Version 3.4

+===========

+

+Release date - 29th May, 2020.

+

+New features

+------------

+1. **Edge-aware quadtree partitioning** to terminate CU depth recursion based on edge information. :option:`--rskip` level 2 enables the feature and  :option:`--rskip-edge-threshold` denotes the minimum expected edge-density percentage within the CU, below which the recursion is skipped. Experimental feature.

+2. Application-level feature :option:`--abr-ladder` for automating efficient ABR ladder generation. Shows ~65% savings in the over-all turn-around time required for the generation of a typical Apple HLS ladder in Intel(R) Xeon(R) Platinum 8280 CPU @ 2.70GHz over a sequential ABR-ladder generation approach that leverages save-load architecture.

+

+Enhancements to existing features

+---------------------------------

+1. Improved efficiency in 2-pass rate-control algorithm. The savings in the bitrate is ~1.72% with visual improvement in quality in the initial 1-2 secs.

+

+Encoder enhancements

+--------------------

+1. Faster ARM64 encodes enabled by ASM contributions from Huawei. The speed-up over no-asm version for 1080p encodes @ medium preset is ~15% in a 16 core H/W.

+2. Strict VBV conformance in zone encoding.

+

+Bug fixes

+---------

+1. Multi-pass encode failures with :option:`--frame-dup`.

+2. Corrupted bitstreams with :option:`--hist-scenecut` when input depth and internal bit-depth differ.

+3. Incorrect analysis propagation in multi-level save-load architecture.

+4. Failure in detecting NUMA packages installed in non-standard directories.

+

 Version 3.3

 ===========

​

@@ -29,7 +29,7 @@

 option(STATIC_LINK_CRT "Statically link C runtime for release builds" OFF)

 mark_as_advanced(FPROFILE_USE FPROFILE_GENERATE NATIVE_BUILD)

 # X265_BUILD must be incremented each time the public API is changed

-set(X265_BUILD 188)

+set(X265_BUILD 192)

 configure_file("${PROJECT_SOURCE_DIR}/x265.def.in"

                "${PROJECT_BINARY_DIR}/x265.def")

 configure_file("${PROJECT_SOURCE_DIR}/x265_config.h.in"

@@ -40,7 +40,7 @@

 # System architecture detection

 string(TOLOWER "${CMAKE_SYSTEM_PROCESSOR}" SYSPROC)

 set(X86_ALIASES x86 i386 i686 x86_64 amd64)

-set(ARM_ALIASES armv6l armv7l)

+set(ARM_ALIASES armv6l armv7l aarch64)

 list(FIND X86_ALIASES "${SYSPROC}" X86MATCH)

 list(FIND ARM_ALIASES "${SYSPROC}" ARMMATCH)

 set(POWER_ALIASES ppc64 ppc64le)

@@ -70,9 +70,15 @@

     else()

         set(CROSS_COMPILE_ARM 0)

     endif()

-    message(STATUS "Detected ARM target processor")

     set(ARM 1)

-    add_definitions(-DX265_ARCH_ARM=1 -DHAVE_ARMV6=1)

+    if("${CMAKE_SIZEOF_VOID_P}" MATCHES 8)

+        message(STATUS "Detected ARM64 target processor")

+        set(ARM64 1)

+        add_definitions(-DX265_ARCH_ARM=1 -DX265_ARCH_ARM64=1 -DHAVE_ARMV6=0)

+    else()

+        message(STATUS "Detected ARM target processor")

+        add_definitions(-DX265_ARCH_ARM=1 -DX265_ARCH_ARM64=0 -DHAVE_ARMV6=1)

+    endif()

 else()

     message(STATUS "CMAKE_SYSTEM_PROCESSOR value `${CMAKE_SYSTEM_PROCESSOR}` is unknown")

     message(STATUS "Please add this value near ${CMAKE_CURRENT_LIST_FILE}:${CMAKE_CURRENT_LIST_LINE}")

@@ -95,6 +101,8 @@

         if(NUMA_FOUND)

             link_directories(${NUMA_LIBRARY_DIR})

             list(APPEND CMAKE_REQUIRED_LIBRARIES numa)

+            list(APPEND CMAKE_REQUIRED_INCLUDES ${NUMA_INCLUDE_DIR})

+            list(APPEND CMAKE_REQUIRED_LINK_OPTIONS "-L${NUMA_LIBRARY_DIR}")

             check_symbol_exists(numa_node_of_cpu numa.h NUMA_V2)

             if(NUMA_V2)

                 add_definitions(-DHAVE_LIBNUMA)

@@ -231,14 +239,24 @@

         endif()

     endif()

     if(ARM AND CROSS_COMPILE_ARM)

-        set(ARM_ARGS -march=armv6 -mfloat-abi=soft -mfpu=vfp -marm -fPIC)

+        if(ARM64)

+            set(ARM_ARGS -fPIC)

+        else()

+            set(ARM_ARGS -march=armv6 -mfloat-abi=soft -mfpu=vfp -marm -fPIC)

+        endif()

+        message(STATUS "cross compile arm")

     elseif(ARM)

-        find_package(Neon)

-        if(CPU_HAS_NEON)

-            set(ARM_ARGS -mcpu=native -mfloat-abi=hard -mfpu=neon -marm -fPIC)

+        if(ARM64)

+            set(ARM_ARGS -fPIC)

             add_definitions(-DHAVE_NEON)

         else()

-            set(ARM_ARGS -mcpu=native -mfloat-abi=hard -mfpu=vfp -marm)

+            find_package(Neon)

+            if(CPU_HAS_NEON)

+                set(ARM_ARGS -mcpu=native -mfloat-abi=hard -mfpu=neon -marm -fPIC)

+                add_definitions(-DHAVE_NEON)

+            else()

+                set(ARM_ARGS -mcpu=native -mfloat-abi=hard -mfpu=vfp -marm)

+            endif()

         endif()

     endif()

     add_definitions(${ARM_ARGS})

@@ -518,7 +536,11 @@

     # compile ARM arch asm files here

         enable_language(ASM)

         foreach(ASM ${ARM_ASMS})

-            set(ASM_SRC ${CMAKE_CURRENT_SOURCE_DIR}/common/arm/${ASM})

+            if(ARM64)

+                set(ASM_SRC ${CMAKE_CURRENT_SOURCE_DIR}/common/aarch64/${ASM})

+            else()

+                set(ASM_SRC ${CMAKE_CURRENT_SOURCE_DIR}/common/arm/${ASM})

+            endif()

             list(APPEND ASM_SRCS ${ASM_SRC})

             list(APPEND ASM_OBJS ${ASM}.${SUFFIX})

             add_custom_command(

@@ -725,16 +747,16 @@

         # Xcode seems unable to link the CLI with libs, so link as one targget

         if(ENABLE_HDR10_PLUS)

         add_executable(cli ../COPYING ${InputFiles} ${OutputFiles} ${GETOPT}

-                        x265.cpp x265.h x265cli.h

+                        x265.cpp x265.h x265cli.cpp x265cli.h abrEncApp.cpp abrEncApp.h

                         $<TARGET_OBJECTS:encoder> $<TARGET_OBJECTS:common> $<TARGET_OBJECTS:dynamicHDR10> ${ASM_OBJS})

         else()

             add_executable(cli ../COPYING ${InputFiles} ${OutputFiles} ${GETOPT}

-                        x265.cpp x265.h x265cli.h

+                        x265.cpp x265.h x265cli.cpp x265cli.h abrEncApp.cpp abrEncApp.h

                         $<TARGET_OBJECTS:encoder> $<TARGET_OBJECTS:common> ${ASM_OBJS})

         endif()

     else()

         add_executable(cli ../COPYING ${InputFiles} ${OutputFiles} ${GETOPT} ${X265_RC_FILE}

-                       ${ExportDefs} x265.cpp x265.h x265cli.h)

+                       ${ExportDefs} x265.cpp x265.h x265cli.cpp x265cli.h abrEncApp.cpp abrEncApp.h)

         if(WIN32 OR NOT ENABLE_SHARED OR INTEL_CXX)

             # The CLI cannot link to the shared library on Windows, it

             # requires internal APIs not exported from the DLL

​

@@ -0,0 +1,1108 @@

+/*****************************************************************************

+* Copyright (C) 2013-2020 MulticoreWare, Inc

+*

+* Authors: Pooja Venkatesan <pooja@multicorewareinc.com>

+*          Aruna Matheswaran <aruna@multicorewareinc.com>

+*

+* This program is free software; you can redistribute it and/or modify

+* it under the terms of the GNU General Public License as published by

+* the Free Software Foundation; either version 2 of the License, or

+* (at your option) any later version.

+*

+* This program is distributed in the hope that it will be useful,

+* but WITHOUT ANY WARRANTY; without even the implied warranty of

+* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the

+* GNU General Public License for more details.

+*

+* You should have received a copy of the GNU General Public License

+* along with this program; if not, write to the Free Software

+* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02111, USA.

+*

+* This program is also available under a commercial proprietary license.

+* For more information, contact us at license @ x265.com.

+*****************************************************************************/

+

+#include "abrEncApp.h"

+#include "mv.h"

+#include "slice.h"

+#include "param.h"

+

+#include <signal.h>

+#include <errno.h>

+

+#include <queue>

+

+using namespace X265_NS;

+

+/* Ctrl-C handler */

+static volatile sig_atomic_t b_ctrl_c /* = 0 */;

+static void sigint_handler(int)

+{

+    b_ctrl_c = 1;

+}

+

+namespace X265_NS {

+    // private namespace

+#define X265_INPUT_QUEUE_SIZE 250

+

+    AbrEncoder::AbrEncoder(CLIOptions cliopt[], uint8_t numEncodes, int &ret)

+    {

+        m_numEncodes = numEncodes;

+        m_numActiveEncodes.set(numEncodes);

+        m_queueSize = (numEncodes > 1) ? X265_INPUT_QUEUE_SIZE : 1;

+        m_passEnc = X265_MALLOC(PassEncoder*, m_numEncodes);

+

+        for (uint8_t i = 0; i < m_numEncodes; i++)

+        {

+            m_passEnc[i] = new PassEncoder(i, cliopt[i], this);

+            if (!m_passEnc[i])

+            {

+                x265_log(NULL, X265_LOG_ERROR, "Unable to allocate memory for passEncoder\n");

+                ret = 4;

+            }

+            m_passEnc[i]->init(ret);

+        }

+

+        if (!allocBuffers())

+        {

+            x265_log(NULL, X265_LOG_ERROR, "Unable to allocate memory for buffers\n");

+            ret = 4;

+        }

+

+        /* start passEncoder worker threads */

+        for (uint8_t pass = 0; pass < m_numEncodes; pass++)

+            m_passEnc[pass]->startThreads();

+    }

+

+    bool AbrEncoder::allocBuffers()

+    {

+        m_inputPicBuffer = X265_MALLOC(x265_picture**, m_numEncodes);

+        m_analysisBuffer = X265_MALLOC(x265_analysis_data*, m_numEncodes);

+

+        m_picWriteCnt = new ThreadSafeInteger[m_numEncodes];

+        m_picReadCnt = new ThreadSafeInteger[m_numEncodes];

+        m_analysisWriteCnt = new ThreadSafeInteger[m_numEncodes];

+        m_analysisReadCnt = new ThreadSafeInteger[m_numEncodes];

+

+        m_picIdxReadCnt = X265_MALLOC(ThreadSafeInteger*, m_numEncodes);

+        m_analysisWrite = X265_MALLOC(ThreadSafeInteger*, m_numEncodes);

+        m_analysisRead = X265_MALLOC(ThreadSafeInteger*, m_numEncodes);

+        m_readFlag = X265_MALLOC(int*, m_numEncodes);

+

+        for (uint8_t pass = 0; pass < m_numEncodes; pass++)

+        {

+            m_inputPicBuffer[pass] = X265_MALLOC(x265_picture*, m_queueSize);

+            for (uint32_t idx = 0; idx < m_queueSize; idx++)

+            {

+                m_inputPicBuffer[pass][idx] = x265_picture_alloc();

+                x265_picture_init(m_passEnc[pass]->m_param, m_inputPicBuffer[pass][idx]);

+            }

+

+            m_analysisBuffer[pass] = X265_MALLOC(x265_analysis_data, m_queueSize);

+            m_picIdxReadCnt[pass] = new ThreadSafeInteger[m_queueSize];

+            m_analysisWrite[pass] = new ThreadSafeInteger[m_queueSize];

+            m_analysisRead[pass] = new ThreadSafeInteger[m_queueSize];

+            m_readFlag[pass] = X265_MALLOC(int, m_queueSize);

+        }

+        return true;

+    }

+

+    void AbrEncoder::destroy()

+    {

+        x265_cleanup(); /* Free library singletons */

+        for (uint8_t pass = 0; pass < m_numEncodes; pass++)

+        {

+            for (uint32_t index = 0; index < m_queueSize; index++)

+            {

+                X265_FREE(m_inputPicBuffer[pass][index]->planes[0]);

+                x265_picture_free(m_inputPicBuffer[pass][index]);

+            }

+

+            X265_FREE(m_inputPicBuffer[pass]);

+            X265_FREE(m_analysisBuffer[pass]);

+            X265_FREE(m_readFlag[pass]);

+            delete[] m_picIdxReadCnt[pass];

+            delete[] m_analysisWrite[pass];

+            delete[] m_analysisRead[pass];

+            m_passEnc[pass]->destroy();

+            delete m_passEnc[pass];

+        }

+        X265_FREE(m_inputPicBuffer);

+        X265_FREE(m_analysisBuffer);

+        X265_FREE(m_readFlag);

+

+        delete[] m_picWriteCnt;

+        delete[] m_picReadCnt;

+        delete[] m_analysisWriteCnt;

+        delete[] m_analysisReadCnt;

+

+        X265_FREE(m_picIdxReadCnt);

+        X265_FREE(m_analysisWrite);

+        X265_FREE(m_analysisRead);

+

+        X265_FREE(m_passEnc);

+    }

+

+    PassEncoder::PassEncoder(uint32_t id, CLIOptions cliopt, AbrEncoder *parent)

+    {

+        m_id = id;

+        m_cliopt = cliopt;

+        m_parent = parent;

+        if(!(m_cliopt.enableScaler && m_id))

+            m_input = m_cliopt.input;

+        m_param = cliopt.param;

+        m_inputOver = false;

+        m_lastIdx = -1;

+        m_encoder = NULL;

+        m_scaler = NULL;

+        m_reader = NULL;

+        m_ret = 0;

+    }

+

+    int PassEncoder::init(int &result)

+    {

+        if (m_parent->m_numEncodes > 1)

+            setReuseLevel();

+                

+        if (!(m_cliopt.enableScaler && m_id))

+            m_reader = new Reader(m_id, this);

+        else

+        {

+            VideoDesc *src = NULL, *dst = NULL;

+            dst = new VideoDesc(m_param->sourceWidth, m_param->sourceHeight, m_param->internalCsp, m_param->internalBitDepth);

+            int dstW = m_parent->m_passEnc[m_id - 1]->m_param->sourceWidth;

+            int dstH = m_parent->m_passEnc[m_id - 1]->m_param->sourceHeight;

+            src = new VideoDesc(dstW, dstH, m_param->internalCsp, m_param->internalBitDepth);

+            if (src != NULL && dst != NULL)

+            {

+                m_scaler = new Scaler(0, 1, m_id, src, dst, this);

+                if (!m_scaler)

+                {

+                    x265_log(m_param, X265_LOG_ERROR, "\n MALLOC failure in Scaler");

+                    result = 4;

+                }

+            }

+        }

+

+        /* note: we could try to acquire a different libx265 API here based on

+        * the profile found during option parsing, but it must be done before

+        * opening an encoder */

+

+        if (m_param)

+            m_encoder = m_cliopt.api->encoder_open(m_param);

+        if (!m_encoder)

+        {

+            x265_log(NULL, X265_LOG_ERROR, "x265_encoder_open() failed for Enc, \n");

+            m_ret = 2;

+            return -1;

+        }

+

​

@@ -0,0 +1,153 @@

+/*****************************************************************************

+* Copyright (C) 2013-2020 MulticoreWare, Inc

+*

+* Authors: Pooja Venkatesan <pooja@multicorewareinc.com>

+*          Aruna Matheswaran <aruna@multicorewareinc.com>

+*           

+*

+* This program is free software; you can redistribute it and/or modify

+* it under the terms of the GNU General Public License as published by

+* the Free Software Foundation; either version 2 of the License, or

+* (at your option) any later version.

+*

+* This program is distributed in the hope that it will be useful,

+* but WITHOUT ANY WARRANTY; without even the implied warranty of

+* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the

+* GNU General Public License for more details.

+*

+* You should have received a copy of the GNU General Public License

+* along with this program; if not, write to the Free Software

+* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02111, USA.

+*

+* This program is also available under a commercial proprietary license.

+* For more information, contact us at license @ x265.com.

+*****************************************************************************/

+

+#ifndef ABR_ENCODE_H

+#define ABR_ENCODE_H

+

+#include "x265.h"

+#include "scaler.h"

+#include "threading.h"

+#include "x265cli.h"

+

+namespace X265_NS {

+    // private namespace

+

+    class PassEncoder;

+    class Scaler;

+    class Reader;

+

+    class AbrEncoder

+    {

+    public:

+        uint8_t           m_numEncodes;

+        PassEncoder        **m_passEnc;

+        uint32_t           m_queueSize;

+        ThreadSafeInteger  m_numActiveEncodes;

+

+        x265_picture       ***m_inputPicBuffer; //[numEncodes][queueSize]

+        x265_analysis_data **m_analysisBuffer; //[numEncodes][queueSize]

+        int                **m_readFlag;

+

+        ThreadSafeInteger  *m_picWriteCnt;

+        ThreadSafeInteger  *m_picReadCnt;

+        ThreadSafeInteger  **m_picIdxReadCnt;

+        ThreadSafeInteger  *m_analysisWriteCnt; //[numEncodes][queueSize]

+        ThreadSafeInteger  *m_analysisReadCnt; //[numEncodes][queueSize]

+        ThreadSafeInteger  **m_analysisWrite; //[numEncodes][queueSize]

+        ThreadSafeInteger  **m_analysisRead; //[numEncodes][queueSize]

+

+        AbrEncoder(CLIOptions cliopt[], uint8_t numEncodes, int& ret);

+        bool allocBuffers();

+        void destroy();

+

+    };

+

+    class PassEncoder : public Thread

+    {

+    public:

+

+        uint32_t m_id;

+        x265_param *m_param;

+        AbrEncoder *m_parent;

+        x265_encoder *m_encoder;

+        Reader *m_reader;

+        Scaler *m_scaler;

+        bool m_inputOver;

+

+        int m_threadActive;

+        int m_lastIdx;

+        uint32_t m_outputNalsCount;

+

+        x265_picture **m_inputPicBuffer;

+        x265_analysis_data **m_analysisBuffer;

+        x265_nal **m_outputNals;

+        x265_picture **m_outputRecon;

+

+        CLIOptions m_cliopt;

+        InputFile* m_input;

+        const char* m_reconPlayCmd;

+        FILE*    m_qpfile;

+        FILE*    m_zoneFile;

+        FILE*    m_dolbyVisionRpu;/* File containing Dolby Vision BL RPU metadata */

+

+        int m_ret;

+

+        PassEncoder(uint32_t id, CLIOptions cliopt, AbrEncoder *parent);

+        int init(int &result);

+        void setReuseLevel();

+

+        void startThreads();

+        void copyInfo(x265_analysis_data *src);

+

+        bool readPicture(x265_picture*);

+        void destroy();

+

+    private:

+        void threadMain();

+    };

+

+    class Scaler : public Thread

+    {

+    public:

+        PassEncoder *m_parentEnc;

+        int m_id;

+        int m_scalePlanes[3];

+        int m_scaleFrameSize;

+        uint32_t m_threadId;

+        uint32_t m_threadTotal;

+        ThreadSafeInteger m_scaledWriteCnt;

+        VideoDesc* m_srcFormat;

+        VideoDesc* m_dstFormat;

+        int m_threadActive;

+        ScalerFilterManager* m_filterManager;

+

+        Scaler(int threadId, int threadNum, int id, VideoDesc *src, VideoDesc * dst, PassEncoder *parentEnc);

+        bool scalePic(x265_picture *destination, x265_picture *source);

+        void threadMain();

+        void destroy()

+        {

+            if (m_filterManager)

+            {

+                delete m_filterManager;

+                m_filterManager = NULL;

+            }

+        }

+    };

+

+    class Reader : public Thread

+    {

+    public:

+        PassEncoder *m_parentEnc;

+        int m_id;

+        InputFile* m_input;

+        int m_threadActive;

+

+        Reader(int id, PassEncoder *parentEnc);

+        void threadMain();

+    };

+}

+

+#endif // ifndef ABR_ENCODE_H

+#pragma once

​

@@ -14,7 +14,7 @@

 endif(EXTRA_LIB)

 if(ENABLE_ASSEMBLY)

-    set_source_files_properties(threading.cpp primitives.cpp PROPERTIES COMPILE_FLAGS -DENABLE_ASSEMBLY=1)

+    set_source_files_properties(threading.cpp primitives.cpp pixel.cpp PROPERTIES COMPILE_FLAGS -DENABLE_ASSEMBLY=1)

     list(APPEND VFLAGS "-DENABLE_ASSEMBLY=1")

 endif(ENABLE_ASSEMBLY)

@@ -84,16 +84,33 @@

 endif(ENABLE_ASSEMBLY AND X86)

 if(ENABLE_ASSEMBLY AND (ARM OR CROSS_COMPILE_ARM))

-    set(C_SRCS asm-primitives.cpp pixel.h mc.h ipfilter8.h blockcopy8.h dct8.h loopfilter.h)

+    if(ARM64)

+        if(GCC AND (CMAKE_CXX_FLAGS_RELEASE MATCHES "-O3"))

+            message(STATUS "Detected CXX compiler using -O3 optimization level")

+            add_definitions(-DAUTO_VECTORIZE=1)

+        endif()

+        set(C_SRCS asm-primitives.cpp pixel.h ipfilter8.h)

-    # add ARM assembly/intrinsic files here

-    set(A_SRCS asm.S cpu-a.S mc-a.S sad-a.S pixel-util.S ssd-a.S blockcopy8.S ipfilter8.S dct-a.S)

-    set(VEC_PRIMITIVES)

+        # add ARM assembly/intrinsic files here

+        set(A_SRCS asm.S mc-a.S sad-a.S pixel-util.S ipfilter8.S)

+        set(VEC_PRIMITIVES)

-    set(ARM_ASMS "${A_SRCS}" CACHE INTERNAL "ARM Assembly Sources")

-    foreach(SRC ${C_SRCS})

-        set(ASM_PRIMITIVES ${ASM_PRIMITIVES} arm/${SRC})

-    endforeach()

+        set(ARM_ASMS "${A_SRCS}" CACHE INTERNAL "ARM Assembly Sources")

+        foreach(SRC ${C_SRCS})

+            set(ASM_PRIMITIVES ${ASM_PRIMITIVES} aarch64/${SRC})

+        endforeach()

+    else()

+        set(C_SRCS asm-primitives.cpp pixel.h mc.h ipfilter8.h blockcopy8.h dct8.h loopfilter.h)

+

+        # add ARM assembly/intrinsic files here

+        set(A_SRCS asm.S cpu-a.S mc-a.S sad-a.S pixel-util.S ssd-a.S blockcopy8.S ipfilter8.S dct-a.S)

+        set(VEC_PRIMITIVES)

+

+        set(ARM_ASMS "${A_SRCS}" CACHE INTERNAL "ARM Assembly Sources")

+        foreach(SRC ${C_SRCS})

+            set(ASM_PRIMITIVES ${ASM_PRIMITIVES} arm/${SRC})

+        endforeach()

+    endif()

     source_group(Assembly FILES ${ASM_PRIMITIVES})

 endif(ENABLE_ASSEMBLY AND (ARM OR CROSS_COMPILE_ARM))

@@ -151,4 +168,5 @@

     predict.cpp  predict.h

     scalinglist.cpp scalinglist.h

     quant.cpp quant.h contexts.h

-    deblock.cpp deblock.h)

+    deblock.cpp deblock.h

+    scaler.cpp scaler.h)

​

@@ -0,0 +1,219 @@

+/*****************************************************************************

+ * Copyright (C) 2020 MulticoreWare, Inc

+ *

+ * Authors: Hongbin Liu <liuhongbin1@huawei.com>

+ *          Yimeng Su <yimeng.su@huawei.com>

+ *

+ * This program is free software; you can redistribute it and/or modify

+ * it under the terms of the GNU General Public License as published by

+ * the Free Software Foundation; either version 2 of the License, or

+ * (at your option) any later version.

+ *

+ * This program is distributed in the hope that it will be useful,

+ * but WITHOUT ANY WARRANTY; without even the implied warranty of

+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the

+ * GNU General Public License for more details.

+ *

+ * You should have received a copy of the GNU General Public License

+ * along with this program; if not, write to the Free Software

+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02111, USA.

+ *

+ * This program is also available under a commercial proprietary license.

+ * For more information, contact us at license @ x265.com.

+ *****************************************************************************/

+

+#include "common.h"

+#include "primitives.h"

+#include "x265.h"

+#include "cpu.h"

+

+

+#if defined(__GNUC__)

+#define GCC_VERSION (__GNUC__ * 10000 + __GNUC_MINOR__ * 100 + __GNUC_PATCHLEVEL__)

+#endif

+

+#define GCC_4_9_0 40900

+#define GCC_5_1_0 50100

+

+extern "C" {

+#include "pixel.h"

+#include "pixel-util.h"

+#include "ipfilter8.h"

+}

+

+namespace X265_NS {

+// private x265 namespace

+

+

+template<int size>

+void interp_8tap_hv_pp_cpu(const pixel* src, intptr_t srcStride, pixel* dst, intptr_t dstStride, int idxX, int idxY)

+{

+    ALIGN_VAR_32(int16_t, immed[MAX_CU_SIZE * (MAX_CU_SIZE + NTAPS_LUMA - 1)]);

+    const int halfFilterSize = NTAPS_LUMA >> 1;

+    const int immedStride = MAX_CU_SIZE;

+

+    primitives.pu[size].luma_hps(src, srcStride, immed, immedStride, idxX, 1);

+    primitives.pu[size].luma_vsp(immed + (halfFilterSize - 1) * immedStride, immedStride, dst, dstStride, idxY);

+}

+

+

+/* Temporary workaround because luma_vsp assembly primitive has not been completed

+ * but interp_8tap_hv_pp_cpu uses mixed C primitive and assembly primitive.

+ * Otherwise, segment fault occurs. */

+void setupAliasCPrimitives(EncoderPrimitives &cp, EncoderPrimitives &asmp, int cpuMask)

+{

+    if (cpuMask & X265_CPU_NEON)

+    {

+        asmp.pu[LUMA_8x4].luma_vsp   = cp.pu[LUMA_8x4].luma_vsp;

+        asmp.pu[LUMA_8x8].luma_vsp   = cp.pu[LUMA_8x8].luma_vsp;

+        asmp.pu[LUMA_8x16].luma_vsp  = cp.pu[LUMA_8x16].luma_vsp;

+        asmp.pu[LUMA_8x32].luma_vsp  = cp.pu[LUMA_8x32].luma_vsp;

+        asmp.pu[LUMA_12x16].luma_vsp = cp.pu[LUMA_12x16].luma_vsp;

+#if !AUTO_VECTORIZE || GCC_VERSION < GCC_5_1_0 /* gcc_version < gcc-5.1.0 */

+        asmp.pu[LUMA_16x4].luma_vsp  = cp.pu[LUMA_16x4].luma_vsp;

+        asmp.pu[LUMA_16x8].luma_vsp  = cp.pu[LUMA_16x8].luma_vsp;

+        asmp.pu[LUMA_16x12].luma_vsp = cp.pu[LUMA_16x12].luma_vsp;

+        asmp.pu[LUMA_16x16].luma_vsp = cp.pu[LUMA_16x16].luma_vsp;

+        asmp.pu[LUMA_16x32].luma_vsp = cp.pu[LUMA_16x32].luma_vsp;

+        asmp.pu[LUMA_16x64].luma_vsp = cp.pu[LUMA_16x64].luma_vsp;

+        asmp.pu[LUMA_32x16].luma_vsp = cp.pu[LUMA_32x16].luma_vsp;

+        asmp.pu[LUMA_32x24].luma_vsp = cp.pu[LUMA_32x24].luma_vsp;

+        asmp.pu[LUMA_32x32].luma_vsp = cp.pu[LUMA_32x32].luma_vsp;

+        asmp.pu[LUMA_32x64].luma_vsp = cp.pu[LUMA_32x64].luma_vsp;

+        asmp.pu[LUMA_48x64].luma_vsp = cp.pu[LUMA_48x64].luma_vsp;

+        asmp.pu[LUMA_64x16].luma_vsp = cp.pu[LUMA_64x16].luma_vsp;

+        asmp.pu[LUMA_64x32].luma_vsp = cp.pu[LUMA_64x32].luma_vsp;

+        asmp.pu[LUMA_64x48].luma_vsp = cp.pu[LUMA_64x48].luma_vsp;

+        asmp.pu[LUMA_64x64].luma_vsp = cp.pu[LUMA_64x64].luma_vsp;    

+#if !AUTO_VECTORIZE || GCC_VERSION < GCC_4_9_0 /* gcc_version < gcc-4.9.0 */

+        asmp.pu[LUMA_4x4].luma_vsp   = cp.pu[LUMA_4x4].luma_vsp;

+        asmp.pu[LUMA_4x8].luma_vsp   = cp.pu[LUMA_4x8].luma_vsp;

+        asmp.pu[LUMA_4x16].luma_vsp  = cp.pu[LUMA_4x16].luma_vsp;

+        asmp.pu[LUMA_24x32].luma_vsp = cp.pu[LUMA_24x32].luma_vsp;

+        asmp.pu[LUMA_32x8].luma_vsp  = cp.pu[LUMA_32x8].luma_vsp;

+#endif

+#endif

+    }

+}

+

+

+void setupAssemblyPrimitives(EncoderPrimitives &p, int cpuMask) 

+{

+    if (cpuMask & X265_CPU_NEON)

+    {

+        p.pu[LUMA_4x4].satd   = PFX(pixel_satd_4x4_neon);

+        p.pu[LUMA_4x8].satd   = PFX(pixel_satd_4x8_neon);

+        p.pu[LUMA_4x16].satd  = PFX(pixel_satd_4x16_neon);

+        p.pu[LUMA_8x4].satd   = PFX(pixel_satd_8x4_neon);

+        p.pu[LUMA_8x8].satd   = PFX(pixel_satd_8x8_neon);

+        p.pu[LUMA_12x16].satd = PFX(pixel_satd_12x16_neon);

+        

+        p.chroma[X265_CSP_I420].pu[CHROMA_420_4x4].satd    = PFX(pixel_satd_4x4_neon);

+        p.chroma[X265_CSP_I420].pu[CHROMA_420_4x8].satd    = PFX(pixel_satd_4x8_neon);

+        p.chroma[X265_CSP_I420].pu[CHROMA_420_4x16].satd   = PFX(pixel_satd_4x16_neon);

+        p.chroma[X265_CSP_I420].pu[CHROMA_420_8x4].satd    = PFX(pixel_satd_8x4_neon);

+        p.chroma[X265_CSP_I420].pu[CHROMA_420_8x8].satd    = PFX(pixel_satd_8x8_neon);

+        p.chroma[X265_CSP_I420].pu[CHROMA_420_12x16].satd  = PFX(pixel_satd_12x16_neon);

+        

+        p.chroma[X265_CSP_I422].pu[CHROMA_422_4x4].satd    = PFX(pixel_satd_4x4_neon);

+        p.chroma[X265_CSP_I422].pu[CHROMA_422_4x8].satd    = PFX(pixel_satd_4x8_neon);

+        p.chroma[X265_CSP_I422].pu[CHROMA_422_4x16].satd   = PFX(pixel_satd_4x16_neon);

+        p.chroma[X265_CSP_I422].pu[CHROMA_422_4x32].satd   = PFX(pixel_satd_4x32_neon);

+        p.chroma[X265_CSP_I422].pu[CHROMA_422_8x4].satd    = PFX(pixel_satd_8x4_neon);

+        p.chroma[X265_CSP_I422].pu[CHROMA_422_8x8].satd    = PFX(pixel_satd_8x8_neon);

+        p.chroma[X265_CSP_I422].pu[CHROMA_422_12x32].satd  = PFX(pixel_satd_12x32_neon);

+

+        p.pu[LUMA_4x4].pixelavg_pp[NONALIGNED]   = PFX(pixel_avg_pp_4x4_neon);

+        p.pu[LUMA_4x8].pixelavg_pp[NONALIGNED]   = PFX(pixel_avg_pp_4x8_neon);

+        p.pu[LUMA_4x16].pixelavg_pp[NONALIGNED]  = PFX(pixel_avg_pp_4x16_neon);

+        p.pu[LUMA_8x4].pixelavg_pp[NONALIGNED]   = PFX(pixel_avg_pp_8x4_neon);

+        p.pu[LUMA_8x8].pixelavg_pp[NONALIGNED]   = PFX(pixel_avg_pp_8x8_neon);

+        p.pu[LUMA_8x16].pixelavg_pp[NONALIGNED]  = PFX(pixel_avg_pp_8x16_neon);

+        p.pu[LUMA_8x32].pixelavg_pp[NONALIGNED]  = PFX(pixel_avg_pp_8x32_neon);

+

+        p.pu[LUMA_4x4].pixelavg_pp[ALIGNED]   = PFX(pixel_avg_pp_4x4_neon);

+        p.pu[LUMA_4x8].pixelavg_pp[ALIGNED]   = PFX(pixel_avg_pp_4x8_neon);

+        p.pu[LUMA_4x16].pixelavg_pp[ALIGNED]  = PFX(pixel_avg_pp_4x16_neon);

+        p.pu[LUMA_8x4].pixelavg_pp[ALIGNED]   = PFX(pixel_avg_pp_8x4_neon);

+        p.pu[LUMA_8x8].pixelavg_pp[ALIGNED]   = PFX(pixel_avg_pp_8x8_neon);

+        p.pu[LUMA_8x16].pixelavg_pp[ALIGNED]  = PFX(pixel_avg_pp_8x16_neon);

+        p.pu[LUMA_8x32].pixelavg_pp[ALIGNED]  = PFX(pixel_avg_pp_8x32_neon);

+

+        p.pu[LUMA_8x4].sad_x3   = PFX(sad_x3_8x4_neon);

+        p.pu[LUMA_8x8].sad_x3   = PFX(sad_x3_8x8_neon);

+        p.pu[LUMA_8x16].sad_x3  = PFX(sad_x3_8x16_neon);

+        p.pu[LUMA_8x32].sad_x3  = PFX(sad_x3_8x32_neon);

+

+        p.pu[LUMA_8x4].sad_x4   = PFX(sad_x4_8x4_neon);

+        p.pu[LUMA_8x8].sad_x4   = PFX(sad_x4_8x8_neon);

+        p.pu[LUMA_8x16].sad_x4  = PFX(sad_x4_8x16_neon);

+        p.pu[LUMA_8x32].sad_x4  = PFX(sad_x4_8x32_neon);

+

+        // quant

+        p.quant = PFX(quant_neon);

+        // luma_hps

+        p.pu[LUMA_4x4].luma_hps   = PFX(interp_8tap_horiz_ps_4x4_neon);

+        p.pu[LUMA_4x8].luma_hps   = PFX(interp_8tap_horiz_ps_4x8_neon);

+        p.pu[LUMA_4x16].luma_hps  = PFX(interp_8tap_horiz_ps_4x16_neon);

+        p.pu[LUMA_8x4].luma_hps   = PFX(interp_8tap_horiz_ps_8x4_neon);

+        p.pu[LUMA_8x8].luma_hps   = PFX(interp_8tap_horiz_ps_8x8_neon);

+        p.pu[LUMA_8x16].luma_hps  = PFX(interp_8tap_horiz_ps_8x16_neon);

+        p.pu[LUMA_8x32].luma_hps  = PFX(interp_8tap_horiz_ps_8x32_neon);

+        p.pu[LUMA_12x16].luma_hps = PFX(interp_8tap_horiz_ps_12x16_neon);

+        p.pu[LUMA_24x32].luma_hps = PFX(interp_8tap_horiz_ps_24x32_neon);

+#if !AUTO_VECTORIZE || GCC_VERSION < GCC_5_1_0 /* gcc_version < gcc-5.1.0 */

+        p.pu[LUMA_16x4].luma_hps  = PFX(interp_8tap_horiz_ps_16x4_neon);

+        p.pu[LUMA_16x8].luma_hps  = PFX(interp_8tap_horiz_ps_16x8_neon);

+        p.pu[LUMA_16x12].luma_hps = PFX(interp_8tap_horiz_ps_16x12_neon);

+        p.pu[LUMA_16x16].luma_hps = PFX(interp_8tap_horiz_ps_16x16_neon);

+        p.pu[LUMA_16x32].luma_hps = PFX(interp_8tap_horiz_ps_16x32_neon);

+        p.pu[LUMA_16x64].luma_hps = PFX(interp_8tap_horiz_ps_16x64_neon);

+        p.pu[LUMA_32x8].luma_hps  = PFX(interp_8tap_horiz_ps_32x8_neon);

+        p.pu[LUMA_32x16].luma_hps = PFX(interp_8tap_horiz_ps_32x16_neon);

+        p.pu[LUMA_32x24].luma_hps = PFX(interp_8tap_horiz_ps_32x24_neon);

+        p.pu[LUMA_32x32].luma_hps = PFX(interp_8tap_horiz_ps_32x32_neon);

+        p.pu[LUMA_32x64].luma_hps = PFX(interp_8tap_horiz_ps_32x64_neon);

+        p.pu[LUMA_48x64].luma_hps = PFX(interp_8tap_horiz_ps_48x64_neon);

+        p.pu[LUMA_64x16].luma_hps = PFX(interp_8tap_horiz_ps_64x16_neon);

+        p.pu[LUMA_64x32].luma_hps = PFX(interp_8tap_horiz_ps_64x32_neon);

+        p.pu[LUMA_64x48].luma_hps = PFX(interp_8tap_horiz_ps_64x48_neon);

+        p.pu[LUMA_64x64].luma_hps = PFX(interp_8tap_horiz_ps_64x64_neon);

+#endif

+

+        p.pu[LUMA_8x4].luma_hvpp   =  interp_8tap_hv_pp_cpu<LUMA_8x4>;

+        p.pu[LUMA_8x8].luma_hvpp   =  interp_8tap_hv_pp_cpu<LUMA_8x8>;

+        p.pu[LUMA_8x16].luma_hvpp  =  interp_8tap_hv_pp_cpu<LUMA_8x16>;

+        p.pu[LUMA_8x32].luma_hvpp  =  interp_8tap_hv_pp_cpu<LUMA_8x32>;

+        p.pu[LUMA_12x16].luma_hvpp =  interp_8tap_hv_pp_cpu<LUMA_12x16>;

+#if !AUTO_VECTORIZE || GCC_VERSION < GCC_5_1_0 /* gcc_version < gcc-5.1.0 */

+        p.pu[LUMA_16x4].luma_hvpp  =  interp_8tap_hv_pp_cpu<LUMA_16x4>;

+        p.pu[LUMA_16x8].luma_hvpp  =  interp_8tap_hv_pp_cpu<LUMA_16x8>;

+        p.pu[LUMA_16x12].luma_hvpp =  interp_8tap_hv_pp_cpu<LUMA_16x12>;

+        p.pu[LUMA_16x16].luma_hvpp =  interp_8tap_hv_pp_cpu<LUMA_16x16>;

+        p.pu[LUMA_16x32].luma_hvpp =  interp_8tap_hv_pp_cpu<LUMA_16x32>;

+        p.pu[LUMA_16x64].luma_hvpp =  interp_8tap_hv_pp_cpu<LUMA_16x64>;

+        p.pu[LUMA_32x16].luma_hvpp =  interp_8tap_hv_pp_cpu<LUMA_32x16>;

+        p.pu[LUMA_32x24].luma_hvpp =  interp_8tap_hv_pp_cpu<LUMA_32x24>;

+        p.pu[LUMA_32x32].luma_hvpp =  interp_8tap_hv_pp_cpu<LUMA_32x32>;

+        p.pu[LUMA_32x64].luma_hvpp =  interp_8tap_hv_pp_cpu<LUMA_32x64>;

+        p.pu[LUMA_48x64].luma_hvpp =  interp_8tap_hv_pp_cpu<LUMA_48x64>;

​

@@ -0,0 +1,69 @@

+/*****************************************************************************

+ * Copyright (C) 2020 MulticoreWare, Inc

+ *

+ * Authors: Hongbin Liu <liuhongbin1@huawei.com>

+ *

+ * This program is free software; you can redistribute it and/or modify

+ * it under the terms of the GNU General Public License as published by

+ * the Free Software Foundation; either version 2 of the License, or

+ * (at your option) any later version.

+ *

+ * This program is distributed in the hope that it will be useful,

+ * but WITHOUT ANY WARRANTY; without even the implied warranty of

+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the

+ * GNU General Public License for more details.

+ *

+ * You should have received a copy of the GNU General Public License

+ * along with this program; if not, write to the Free Software

+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02111, USA.

+ *

+ * This program is also available under a commercial proprietary license.

+ * For more information, contact us at license @ x265.com.

+ *****************************************************************************/

+

+.arch           armv8-a

+

+#ifdef PREFIX

+#define EXTERN_ASM _

+#else

+#define EXTERN_ASM

+#endif

+

+#ifdef __ELF__

+#define ELF

+#else

+#define ELF @

+#endif

+

+#define HAVE_AS_FUNC 1

+

+#if HAVE_AS_FUNC

+#define FUNC

+#else

+#define FUNC @

+#endif

+

+.macro function name, export=1

+    .macro endfunc

+ELF     .size   \name, . - \name

+FUNC    .endfunc

+        .purgem endfunc

+    .endm

+        .align  2

+.if \export == 1

+        .global EXTERN_ASM\name

+ELF     .hidden EXTERN_ASM\name

+ELF     .type   EXTERN_ASM\name, %function

+FUNC    .func   EXTERN_ASM\name

+EXTERN_ASM\name:

+.else

+ELF     .hidden \name

+ELF     .type   \name, %function

+FUNC    .func   \name

+\name:

+.endif

+.endm

+

+

+#define FENC_STRIDE 64

+#define FDEC_STRIDE 32

​

@@ -0,0 +1,414 @@

+/*****************************************************************************

+ * Copyright (C) 2020 MulticoreWare, Inc

+ *

+ * Authors: Yimeng Su <yimeng.su@huawei.com>

+ *

+ * This program is free software; you can redistribute it and/or modify

+ * it under the terms of the GNU General Public License as published by

+ * the Free Software Foundation; either version 2 of the License, or

+ * (at your option) any later version.

+ *

+ * This program is distributed in the hope that it will be useful,

+ * but WITHOUT ANY WARRANTY; without even the implied warranty of

+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the

+ * GNU General Public License for more details.

+ *

+ * You should have received a copy of the GNU General Public License

+ * along with this program; if not, write to the Free Software

+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02111, USA.

+ *

+ * This program is also available under a commercial proprietary license.

+ * For more information, contact us at license @ x265.com.

+ *****************************************************************************/

+

+#include "asm.S"

+

+.section .rodata

+

+.align 4

+

+.text

+

+

+

+.macro qpel_filter_0_32b

+    movi            v24.8h, #64

+    uxtl            v19.8h, v5.8b

+    smull           v17.4s, v19.4h, v24.4h

+    smull2          v18.4s, v19.8h, v24.8h

+.endm

+

+.macro qpel_filter_1_32b

+    movi            v16.8h, #58

+    uxtl            v19.8h, v5.8b

+    smull           v17.4s, v19.4h, v16.4h

+    smull2          v18.4s, v19.8h, v16.8h

+

+    movi            v24.8h, #10

+    uxtl            v21.8h, v1.8b

+    smull           v19.4s, v21.4h, v24.4h

+    smull2          v20.4s, v21.8h, v24.8h

+

+    movi            v16.8h, #17

+    uxtl            v23.8h, v2.8b

+    smull           v21.4s, v23.4h, v16.4h

+    smull2          v22.4s, v23.8h, v16.8h

+

+    movi            v24.8h, #5

+    uxtl            v1.8h, v6.8b

+    smull           v23.4s, v1.4h, v24.4h

+    smull2          v16.4s, v1.8h, v24.8h

+

+    sub             v17.4s, v17.4s, v19.4s

+    sub             v18.4s, v18.4s, v20.4s

+

+    uxtl            v1.8h, v4.8b

+    sshll           v19.4s, v1.4h, #2

+    sshll2          v20.4s, v1.8h, #2

+

+    add             v17.4s, v17.4s, v21.4s

+    add             v18.4s, v18.4s, v22.4s