Request 6540 (accepted)

Overview

Request 6540 (accepted)

Update to version 1.1.1:

Created by olh about 2 hours ago
In state accepted

Submit libde265

Submit package Staging / libde265 to package Essentials / libde265

libde265.changes Changed

@@ -1,4 +1,48 @@
 -------------------------------------------------------------------
+Tue Jun 16 07:40:59 UTC 2026 - Bjørn Lie <zaitor@opensuse.org>
+
+- Update to version 1.1.1:
+  + The decoding speed has been improved by about 8% on x86 CPUs
+    thanks to more SIMD acceleration and optimized CABAC code. Also
+    the startup time has been improved, which gives a 3% speed
+    improvement when decoding HEIC files with similar-sized tiles.
+  + Security:
+    - CVE-2026-54240 (GHSA-ccfw-29x7-rrx3) - Pixel accessor signed
+      integer overflow causes heap OOB read/write
+    - CVE-2026-54241 (GHSA-j2qq-x2xq-g9wr) - SAO sequential filter
+      heap buffer overflow via signed integer overflow
+- Changes from version 1.1.0:
+  + Added de265_security_limits parameters to limit the maximum
+    image size and memory that libde265 will use during decoding.
+  + Security fixes:
+    - CVE-2026-49295 (GHSA-g2rg-wj66-w594) - Out-of-bounds write in
+      process_reference_picture_set via predicted short-term RPS
+    - CVE-2026-49337 (GHSA-g5hj-rf9f-7vxm) - Unbounded memory
+      accumulation via orphaned slice headers in read_slice_NAL
+    - CVE-2026-49346 (GHSA-vv8h-932h-7r86) - Heap buffer overflow
+      in de265_image_get_buffer via SPS dimension integer overflow
+    - (GHSA-x27c-jp65-g395) - Quadratic CPU consumption in NAL
+      parser (remove_stuffing_bytes, resize)
+- Changes from version 1.0.19:
+  + This release contains a number of security fixes, correctness
+    fixes for edge cases, and build/packaging improvements.
+    The public API is binary-compatible with v1.0.18.
+  + Security fixes:
+    - CVE-2026-45382 (GHSA-hwhx-x2mq-ccr9) : Heap-buffer-overflow
+      READ in decode_slice_unit_tiles via unvalidated PPS tile
+      geometry.
+    - CVE-2026-45383 (GHSA-wg9q-ppqw-6q38) : Heap buffer overflow
+      (OOB read) in decode_slice_unit_WPP() via out-of-bounds
+      CtbAddrRStoTS access
+- Changes from version 1.0.18:
+  + libde265ConfigVersion.cmake renamed to
+    libde265-config-version.cmake
+  + fix pkg-config when installing to absolute paths
+  + fix compilation with MSVC in Debug mode
+  + removed the (defunct) encoder code and the internal development
+    tools from the tarball.
+
+-------------------------------------------------------------------
 Wed Mar 18 13:35:29 UTC 2026 - Bjørn Lie <zaitor@opensuse.org>
 
 - Update to version 1.0.17:

libde265.spec Changed

libde265-1.0.17.tar.gz/enc265 Deleted

libde265-1.0.17.tar.gz/enc265/CMakeLists.txt Deleted

libde265-1.0.17.tar.gz/enc265/COPYING Deleted

@@ -1,20 +0,0 @@
-
-                           MIT License
-
-  Permission is hereby granted, free of charge, to any person obtaining a copy
-  of this software and associated documentation files (the "Software"), to deal
-  in the Software without restriction, including without limitation the rights
-  to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
-  copies of the Software, and to permit persons to whom the Software is
-  furnished to do so, subject to the following conditions:
-
-  The above copyright notice and this permission notice shall be included in all
-  copies or substantial portions of the Software.
-
-  THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
-  IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
-  FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
-  AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
-  LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
-  OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
-  SOFTWARE.

libde265-1.0.17.tar.gz/enc265/enc265.cc Deleted

@@ -1,361 +0,0 @@
-/*
-  libde265 example application "enc265".
-
-  MIT License
-
-  Copyright (c) 2013-2014 struktur AG, Dirk Farin <farin@struktur.de>
-
-  Permission is hereby granted, free of charge, to any person obtaining a copy
-  of this software and associated documentation files (the "Software"), to deal
-  in the Software without restriction, including without limitation the rights
-  to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
-  copies of the Software, and to permit persons to whom the Software is
-  furnished to do so, subject to the following conditions:
-
-  The above copyright notice and this permission notice shall be included in all
-  copies or substantial portions of the Software.
-
-  THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
-  IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
-  FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
-  AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
-  LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
-  OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
-  SOFTWARE.
-*/
-
-#include "libde265/en265.h"
-#ifdef HAVE_CONFIG_H
-#include "config.h"
-#endif
-
-#include "libde265/encoder/configparam.h"
-#include "libde265/image-io.h"
-#include "libde265/encoder/encoder-core.h"
-#include "libde265/util.h"
-#include "image-io-png.h"
-
-#include <getopt.h>
-
-
-
-#if HAVE_VIDEOGFX
-#include <libvideogfx.hh>
-using namespace videogfx;
-
-void debug_show_image_libvideogfx(const de265_image* input, int slot)
-{
-    static X11Win debugwin;
-    static bool opened=false;
-    int w = input->get_width();
-    int h = input->get_height();
-    if (!opened) {
-      opened=true;
-      debugwin.Create(w,h, "debug");
-    }
-
-    Image<Pixel> img;
-    img.Create(w,h,Colorspace_YUV, Chroma_420);
-
-    for (int y=0;y<h;y++)
-      memcpy(img.AskFrameY()y, input->get_image_plane_at_pos(0,0,y), w);
-
-    for (int y=0;y<h/2;y++) {
-      memcpy(img.AskFrameU()y, input->get_image_plane_at_pos(1,0,y), w/2);
-      memcpy(img.AskFrameV()y, input->get_image_plane_at_pos(2,0,y), w/2);
-    }
-
-    debugwin.Display(img);
-    //debugwin.WaitForKeypress();
-}
-#endif
-
-
-
-int show_help=false;
-int verbosity=0;
-
-static struct option long_options = {
-  {"help",       no_argument,       &show_help, 1 },
-  {"verbose",    no_argument,       0, 'v' },
-  {0,            0,                 0,  0 }
-};
-
-
-struct inout_params
-{
-  inout_params();
-
-  // input
-
-  option_int first_frame;
-  option_int max_number_of_frames;
-
-  option_string input_yuv;
-  option_int input_width;
-  option_int input_height;
-
-  option_bool input_is_rgb;
-
-  // output
-
-  option_string output_filename;
-
-  // debug
-
-  option_string reconstruction_yuv;
-
-
-  void register_params(config_parameters& config);
-};
-
-
-inout_params::inout_params()
-{
-  input_yuv.set_ID("input"); input_yuv.set_short_option('i');
-  input_yuv.set_default("paris_cif.yuv");
-
-  output_filename.set_ID("output"); output_filename.set_short_option('o');
-  output_filename.set_default("out.bin");
-
-  reconstruction_yuv.set_ID("input");
-  reconstruction_yuv.set_default("recon.yuv");
-
-  first_frame.set_ID("first-frame");
-  first_frame.set_default(0);
-  first_frame.set_minimum(0);
-
-  max_number_of_frames.set_ID("frames");
-  max_number_of_frames.set_short_option('f');
-  max_number_of_frames.set_minimum(1);
-  //max_number_of_frames.set_default(INT_MAX);
-
-  input_width.set_ID("width"); input_width.set_short_option('w');
-  input_width.set_minimum(1);  input_width.set_default(352);
-
-  input_height.set_ID("height"); input_height.set_short_option('h');
-  input_height.set_minimum(1); input_height.set_default(288);
-
-  input_is_rgb.set_ID("rgb");
-  input_is_rgb.set_default(false);
-  input_is_rgb.set_description("input is sequence of RGB PNG images");
-}
-
-
-void inout_params::register_params(config_parameters& config)
-{
-  config.add_option(&input_yuv);
-  config.add_option(&output_filename);
-  config.add_option(&first_frame);
-  config.add_option(&max_number_of_frames);
-  config.add_option(&input_width);
-  config.add_option(&input_height);
-#if HAVE_VIDEOGFX
-  if (videogfx::PNG_Supported()) {
-    config.add_option(&input_is_rgb);
-  }
-#endif
-}
-
-
-void test_parameters_API(en265_encoder_context* ectx)
-{
-  const char** param = en265_list_parameters(ectx);
-  if (param) {
-    for (int i=0; parami; i++) {
-      printf("|%s| ",parami);
-
-      enum en265_parameter_type type = en265_get_parameter_type(ectx, parami);
-      const char* type_name="unknown";
-      switch (type) {
-      case en265_parameter_int: type_name="int"; break;
-      case en265_parameter_bool: type_name="bool"; break;
-      case en265_parameter_string: type_name="string"; break;
-      case en265_parameter_choice: type_name="choice"; break;
-      }
-
-      printf("(%s)",type_name);
-
-      if (type==en265_parameter_choice) {
-        const char** choices = en265_list_parameter_choices(ectx, parami);
-        if (choices) {
-          for (int k=0; choicesk; k++) {
-            printf(" %s",choicesk);
-          }
-        }
-      }
-
-      printf("\n");
-    }
-  }
-
-  // en265_set_parameter_int(ectx, "min-tb-size", 8);
-}
-
-
-extern int skipTBSplit, noskipTBSplit;
-extern int zeroBlockCorrelation625;
-
-/*LIBDE265_API*/ ImageSink_YUV reconstruction_sink;
-
-
-int main(int argc, char** argv)
-{
-  de265_init();
-
-  en265_encoder_context* ectx = en265_new_encoder();
-
-
-  bool cmdline_errors = false;
-
-  // --- in/out parameters ---
-
-  struct inout_params inout_params;
-  config_parameters inout_param_config;
-  inout_params.register_params(inout_param_config);
-
-  int first_idx=1;
-  if (!inout_param_config.parse_command_line_params(&argc,argv, &first_idx, true)) {
-    cmdline_errors = true;
-  }
-
-
-  // --- read encoder parameters ---
-
-  if (en265_parse_command_line_parameters(ectx, &argc, argv) != DE265_OK) {
-    cmdline_errors = true;
-  }
-
-
-
-  while (1) {
-    int option_index = 0;
-
-    int c = getopt_long(argc, argv, "v"
-                        , long_options, &option_index);
-    if (c == -1)
-      break;
-
-    switch (c) {
-    case 'v': verbosity++; break;
-    }
-  }
-
-
-  // --- show usage information ---
-
-  if (optind != argc || cmdline_errors || show_help) {
-    fprintf(stderr," enc265  v%s\n", de265_get_version());
-    fprintf(stderr,"--------------\n");
-    fprintf(stderr,"usage: enc265 options\n");
-    fprintf(stderr,"The video file must be a raw YUV file or a PNG sequence for RGB input\n");
-    fprintf(stderr,"\n");
-    fprintf(stderr,"options:\n");
-    fprintf(stderr,"      --help         show help\n");
-    fprintf(stderr,"  -v, --verbose      increase verbosity level (up to 3 times)\n");
-
-    inout_param_config.print_params();
-    fprintf(stderr,"\n");
-    en265_show_parameters(ectx);
-
-    exit(show_help ? 0 : 5);
-  }
-
-
-
-  de265_set_verbosity(verbosity);
-
-#if HAVE_VIDEOGFX
-  //debug_set_image_output(debug_show_image_libvideogfx);
-#endif
-
-  //test_parameters_API(ectx);
-
-
-  if (strlen(inout_params.reconstruction_yuv.get().c_str()) != 0) {
-    reconstruction_sink.set_filename(inout_params.reconstruction_yuv.get().c_str());
-    //ectx.reconstruction_sink = &reconstruction_sink;
-  }
-
-  ImageSource* image_source;
-  ImageSource_YUV image_source_yuv;
-#if HAVE_VIDEOGFX
-  ImageSource_PNG image_source_png;
-#endif
-
-
-  if (inout_params.input_is_rgb) {
-#if HAVE_VIDEOGFX
-    image_source_png.set_input_file(inout_params.input_yuv.get().c_str());
-    image_source = &image_source_png;
-#endif
-  }
-  else {
-    image_source_yuv.set_input_file(inout_params.input_yuv.get().c_str(),
-                                    inout_params.input_width,
-                                    inout_params.input_height);
-    image_source = &image_source_yuv;
-  }
-
-
-  PacketSink_File packet_sink;
-  packet_sink.set_filename(inout_params.output_filename.get().c_str());
-
-
-  // --- run encoder ---
-
-  image_source->skip_frames( inout_params.first_frame );
-
-  en265_start_encoder(ectx, 0);
-
-  int maxPoc = INT_MAX;
-  if (inout_params.max_number_of_frames.is_defined()) {
-    maxPoc = inout_params.max_number_of_frames;
-  }
-
-  bool eof = false;
-  for (int poc=0; poc<maxPoc && !eof ;poc++)
-    {
-      // push one image into the encoder
-
-      de265_image* input_image = image_source->get_image();
-      if (input_image==NULL) {
-        en265_push_eof(ectx);
-        eof=true;
-      }
-      else {
-        en265_push_image(ectx, input_image);
-      }
-
-
-
-      // encode images while more are available
-
-      en265_encode(ectx);
-
-
-      // write all pending packets
-
-      for (;;) {
-        en265_packet* pck = en265_get_packet(ectx,0);
-        if (pck==NULL)
-          break;
-
-        packet_sink.send_packet(pck->data, pck->length);
-
-        en265_free_packet(ectx,pck);
-      }
-    }
-
-
-  // --- print statistics ---
-
-  en265_print_logging((encoder_context*)ectx, "tb-split", NULL);
-
-
-  en265_free_encoder(ectx);
-
-  de265_free();
-
-  return 0;
-}

libde265-1.0.17.tar.gz/enc265/image-io-png.cc Deleted

@@ -1,114 +0,0 @@
-/*
-  libde265 example application "enc265".
-
-  MIT License
-
-  Copyright (c) 2013-2014 struktur AG, Dirk Farin <farin@struktur.de>
-
-  Permission is hereby granted, free of charge, to any person obtaining a copy
-  of this software and associated documentation files (the "Software"), to deal
-  in the Software without restriction, including without limitation the rights
-  to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
-  copies of the Software, and to permit persons to whom the Software is
-  furnished to do so, subject to the following conditions:
-
-  The above copyright notice and this permission notice shall be included in all
-  copies or substantial portions of the Software.
-
-  THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
-  IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
-  FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
-  AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
-  LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
-  OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
-  SOFTWARE.
-*/
-
-#include "image-io-png.h"
-#include <assert.h>
-
-#if HAVE_VIDEOGFX
-#include <libvideogfx.hh>
-using namespace videogfx;
-
-
-ImageSource_PNG::ImageSource_PNG()
-{
-  mFilenameTemplate = NULL;
-  mNextImageNumber = 1;
-
-  mReachedEndOfStream = false;
-
-  mWidth=mHeight=0;
-}
-
-ImageSource_PNG::~ImageSource_PNG()
-{
-}
-
-bool ImageSource_PNG::set_input_file(const char* filename)
-{
-  mFilenameTemplate = filename;
-  return true;
-}
-
-de265_image* ImageSource_PNG::get_image(bool block)
-{
-  if (mReachedEndOfStream) return NULL;
-
-
-  // --- construct image filename ---
-
-  char filename1000;
-  sprintf(filename,mFilenameTemplate,mNextImageNumber);
-  mNextImageNumber++;
-
-
-  // --- load image ---
-
-  Image<Pixel> input;
-  bool success = videogfx::ReadImage_PNG(input, filename);
-  if (!success) {
-    mReachedEndOfStream = true;
-    return NULL;
-  }
-
-
-  mWidth = input.AskWidth();
-  mHeight= input.AskHeight();
-
-  de265_image* img = new de265_image;
-  img->alloc_image(mWidth,mHeight,de265_chroma_444, NULL, false,
-                   NULL, /*NULL,*/ 0, NULL, false);
-  assert(img); // TODO: error handling
-
-
-  uint8_t* p;
-  int stride;
-
-  for (int c=0;c<3;c++) {
-    int h265channel;
-    switch (c) {
-    case 0: h265channel=2; break; // R
-    case 1: h265channel=0; break; // G
-    case 2: h265channel=1; break; // B
-    }
-
-    p = img->get_image_plane(h265channel);
-    stride = img->get_image_stride(h265channel);
-
-    for (int y=0;y<mHeight;y++) {
-      memcpy(p, input.AskFrame((BitmapChannel(c)))y, mWidth);
-      p += stride;
-    }
-  }
-
-  return img;
-}
-
-void ImageSource_PNG::skip_frames(int n)
-{
-  mNextImageNumber += n;
-}
-
-#endif

libde265-1.0.17.tar.gz/enc265/image-io-png.h Deleted

@@ -1,60 +0,0 @@
-/*
-  libde265 example application "enc265".
-
-  MIT License
-
-  Copyright (c) 2013-2014 struktur AG, Dirk Farin <farin@struktur.de>
-
-  Permission is hereby granted, free of charge, to any person obtaining a copy
-  of this software and associated documentation files (the "Software"), to deal
-  in the Software without restriction, including without limitation the rights
-  to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
-  copies of the Software, and to permit persons to whom the Software is
-  furnished to do so, subject to the following conditions:
-
-  The above copyright notice and this permission notice shall be included in all
-  copies or substantial portions of the Software.
-
-  THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
-  IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
-  FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
-  AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
-  LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
-  OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
-  SOFTWARE.
- */
-
-#ifndef IMAGE_IO_PNG_H
-#define IMAGE_IO_PNG_H
-
-#include "libde265/image-io.h"
-#include <deque>
-
-
-#if HAVE_VIDEOGFX
-class ImageSource_PNG : public ImageSource
-{
- public:
-  LIBDE265_API ImageSource_PNG();
-  virtual LIBDE265_API ~ImageSource_PNG();
-
-  bool LIBDE265_API set_input_file(const char* filename);
-
-  //virtual ImageStatus  get_status();
-  virtual LIBDE265_API de265_image* get_image(bool block=true);
-  virtual LIBDE265_API void skip_frames(int n);
-
-  virtual LIBDE265_API int get_width() const { return mWidth; }
-  virtual LIBDE265_API int get_height() const { return mHeight; }
-
- private:
-  const char* mFilenameTemplate;
-  int mNextImageNumber;
-
-  bool mReachedEndOfStream;
-
-  int mWidth,mHeight;
-};
-#endif
-
-#endif

libde265-1.0.17.tar.gz/libde265/arm Deleted

libde265-1.0.17.tar.gz/libde265/arm/CMakeLists.txt Deleted

libde265-1.0.17.tar.gz/libde265/en265.cc Deleted

@@ -1,321 +0,0 @@
-/*
- * H.265 video codec.
- * Copyright (c) 2013-2014 struktur AG, Dirk Farin <farin@struktur.de>
- *
- * Authors: Dirk Farin <farin@struktur.de>
- *
- * This file is part of libde265.
- *
- * libde265 is free software: you can redistribute it and/or modify
- * it under the terms of the GNU Lesser General Public License as
- * published by the Free Software Foundation, either version 3 of
- * the License, or (at your option) any later version.
- *
- * libde265 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 Lesser General Public License for more details.
- *
- * You should have received a copy of the GNU Lesser General Public License
- * along with libde265.  If not, see <http://www.gnu.org/licenses/>.
- */
-
-#include "libde265/en265.h"
-#include "libde265/encoder/encoder-context.h"
-
-
-LIBDE265_API en265_encoder_context* en265_new_encoder(void)
-{
-  de265_error init_err = de265_init();
-  if (init_err != DE265_OK) {
-    return NULL;
-  }
-
-  encoder_context* ectx = new encoder_context();
-  return reinterpret_cast<en265_encoder_context*>(ectx);
-}
-
-
-LIBDE265_API de265_error en265_free_encoder(en265_encoder_context* e)
-{
-  assert(e);
-  encoder_context* ectx = reinterpret_cast<encoder_context*>(e);
-  delete ectx;
-
-  return de265_free();
-}
-
-
-/*
-LIBDE265_API void en265_set_image_release_function(en265_encoder_context* e,
-                                                   void (*release_func)(en265_encoder_context*,
-                                                                        de265_image*,
-                                                                        void* userdata),
-                                                   void* alloc_userdata)
-{
-  assert(e);
-  encoder_context* ectx = reinterpret_cast<encoder_context*>(e);
-
-  ectx->param_image_allocation_userdata = alloc_userdata;
-  ectx->release_func = release_func;
-}
-*/
-
-
-// ========== encoder parameters ==========
-
-LIBDE265_API de265_error en265_parse_command_line_parameters(en265_encoder_context* e,
-                                                             int* argc, char** argv)
-{
-  assert(e);
-  encoder_context* ectx = reinterpret_cast<encoder_context*>(e);
-
-  //if (!ectx->params_config.parse_command_line_params(argc,argv, &ectx->params, true)) {
-  int first_idx=1;
-  if (!ectx->params_config.parse_command_line_params(argc,argv, &first_idx, true)) {
-    return DE265_ERROR_PARAMETER_PARSING;
-  }
-  else {
-    return DE265_OK;
-  }
-}
-
-LIBDE265_API void en265_show_parameters(en265_encoder_context* e)
-{
-  assert(e);
-  encoder_context* ectx = reinterpret_cast<encoder_context*>(e);
-
-  //ectx->params_config.show_params(&ectx->params);
-
-  ectx->params_config.print_params();
-}
-
-
-LIBDE265_API const char** en265_list_parameters(en265_encoder_context* e)
-{
-  assert(e);
-  encoder_context* ectx = reinterpret_cast<encoder_context*>(e);
-
-  return ectx->params_config.get_parameter_string_table();
-}
-
-
-LIBDE265_API enum en265_parameter_type en265_get_parameter_type(en265_encoder_context* e,
-                                                                const char* parametername)
-{
-  assert(e);
-  encoder_context* ectx = reinterpret_cast<encoder_context*>(e);
-
-  return ectx->params_config.get_parameter_type(parametername);
-}
-
-
-LIBDE265_API de265_error en265_set_parameter_bool(en265_encoder_context* e,
-                                                  const char* param,int value)
-{
-  assert(e);
-  encoder_context* ectx = reinterpret_cast<encoder_context*>(e);
-
-  return ectx->params_config.set_bool(param,value) ? DE265_OK : DE265_ERROR_PARAMETER_PARSING;
-}
-
-
-LIBDE265_API de265_error en265_set_parameter_int(en265_encoder_context* e,
-                                                 const char* param,int value)
-{
-  assert(e);
-  encoder_context* ectx = reinterpret_cast<encoder_context*>(e);
-
-  return ectx->params_config.set_int(param,value) ? DE265_OK : DE265_ERROR_PARAMETER_PARSING;
-}
-
-LIBDE265_API de265_error en265_set_parameter_string(en265_encoder_context* e,
-                                                    const char* param,const char* value)
-{
-  assert(e);
-  encoder_context* ectx = reinterpret_cast<encoder_context*>(e);
-
-  return ectx->params_config.set_string(param,value) ? DE265_OK : DE265_ERROR_PARAMETER_PARSING;
-}
-
-LIBDE265_API de265_error en265_set_parameter_choice(en265_encoder_context* e,
-                                                    const char* param,const char* value)
-{
-  assert(e);
-  encoder_context* ectx = reinterpret_cast<encoder_context*>(e);
-
-  return ectx->params_config.set_choice(param,value) ? DE265_OK : DE265_ERROR_PARAMETER_PARSING;
-}
-
-
-LIBDE265_API const char** en265_list_parameter_choices(en265_encoder_context* e,
-                                                       const char* parametername)
-{
-  assert(e);
-  encoder_context* ectx = reinterpret_cast<encoder_context*>(e);
-
-  return ectx->params_config.get_parameter_choices_table(parametername);
-}
-
-
-
-// ========== encoding loop ==========
-
-
-LIBDE265_API de265_error en265_start_encoder(en265_encoder_context* e, int number_of_threads)
-{
-  assert(e);
-  encoder_context* ectx = reinterpret_cast<encoder_context*>(e);
-
-  ectx->start_encoder();
-
-  return DE265_OK;
-}
-
-
-LIBDE265_API struct de265_image* en265_allocate_image(en265_encoder_context* e,
-                                                      int width, int height, de265_chroma chroma,
-                                                      de265_PTS pts, void* image_userdata)
-{
-  assert(e);
-  encoder_context* ectx = reinterpret_cast<encoder_context*>(e);
-
-  de265_image* img = new de265_image;
-  if (img->alloc_image(width,height,de265_chroma_420, NULL, false,
-                       NULL, /*ectx,*/ pts, image_userdata, true) != DE265_OK) {
-    delete img;
-    return NULL;
-  }
-
-  return img;
-}
-
-// Request a specification of the image memory layout for an image of the specified dimensions.
-LIBDE265_API void en265_get_image_spec(en265_encoder_context* e,
-                                       int width, int height, de265_chroma chroma,
-                                       struct de265_image_spec* out_spec)
-{
-  out_spec->format = de265_image_format_YUV420P8;
-  out_spec->width = width;
-  out_spec->height= height;
-  out_spec->alignment = 1;
-
-  out_spec->crop_left  =0;
-  out_spec->crop_right =0;
-  out_spec->crop_top   =0;
-  out_spec->crop_bottom=0;
-
-  out_spec->visible_width  = out_spec->width  - out_spec->crop_left - out_spec->crop_right;
-  out_spec->visible_height = out_spec->height - out_spec->crop_top  - out_spec->crop_bottom;
-}
-
-// Image memory layout specification for an image returned by en265_allocate_image().
-//LIBDE265_API void de265_get_image_spec_from_image(de265_image* img, struct de265_image_spec* spec);
-
-
-
-LIBDE265_API de265_error en265_push_image(en265_encoder_context* e,
-                                          struct de265_image* img)
-{
-  assert(e);
-  encoder_context* ectx = reinterpret_cast<encoder_context*>(e);
-
-  ectx->sop->insert_new_input_image(img);
-  return DE265_OK;
-}
-
-
-LIBDE265_API de265_error en265_push_eof(en265_encoder_context* e)
-{
-  assert(e);
-  encoder_context* ectx = reinterpret_cast<encoder_context*>(e);
-
-  ectx->sop->insert_end_of_stream();
-  return DE265_OK;
-}
-
-
-LIBDE265_API de265_error en265_block_on_input_queue_length(en265_encoder_context*,
-                                                           int max_pending_images,
-                                                           int timeout_ms)
-{
-  // TODO
-  return DE265_OK;
-}
-
-LIBDE265_API de265_error en265_trim_input_queue(en265_encoder_context*, int max_pending_images)
-{
-  // TODO
-  return DE265_OK;
-}
-
-LIBDE265_API int  en265_current_input_queue_length(en265_encoder_context*)
-{
-  // TODO
-  return DE265_OK;
-}
-
-LIBDE265_API de265_error en265_encode(en265_encoder_context* e)
-{
-  assert(e);
-  encoder_context* ectx = reinterpret_cast<encoder_context*>(e);
-
-  while (ectx->picbuf.have_more_frames_to_encode())
-    {
-      de265_error result = ectx->encode_picture_from_input_buffer();
-      if (result != DE265_OK) return result;
-    }
-
-  return DE265_OK;
-}
-
-LIBDE265_API enum en265_encoder_state en265_get_encoder_state(en265_encoder_context* e)
-{
-  // TODO
-  return EN265_STATE_IDLE;
-}
-
-LIBDE265_API struct en265_packet* en265_get_packet(en265_encoder_context* e, int timeout_ms)
-{
-  assert(e);
-  encoder_context* ectx = reinterpret_cast<encoder_context*>(e);
-
-  assert(timeout_ms==0); // TODO: blocking not implemented yet
-
-  if (ectx->output_packets.size()>0) {
-    en265_packet* pck = ectx->output_packets.front();
-    ectx->output_packets.pop_front();
-
-    return pck;
-  }
-  else {
-    return NULL;
-  }
-}
-
-LIBDE265_API void en265_free_packet(en265_encoder_context* e, struct en265_packet* pck)
-{
-  assert(e);
-  encoder_context* ectx = reinterpret_cast<encoder_context*>(e);
-
-  // Do not delete images here. They are owned by the EncPicBuf.
-  //delete   pck->input_image;
-  //delete   pck->reconstruction;
-
-  if (pck->frame_number >= 0) {
-    ectx->mark_image_is_outputted(pck->frame_number);
-
-    ectx->release_input_image(pck->frame_number);
-  }
-
-  delete pck->data;
-  delete   pck;
-}
-
-LIBDE265_API int en265_number_of_queued_packets(en265_encoder_context* e)
-{
-  assert(e);
-  encoder_context* ectx = reinterpret_cast<encoder_context*>(e);
-
-  return ectx->output_packets.size();
-}

libde265-1.0.17.tar.gz/libde265/en265.h Deleted

@@ -1,218 +0,0 @@
-/*
- * H.265 video codec.
- * Copyright (c) 2013-2014 struktur AG, Dirk Farin <farin@struktur.de>
- *
- * Authors: Dirk Farin <farin@struktur.de>
- *
- * This file is part of libde265.
- *
- * libde265 is free software: you can redistribute it and/or modify
- * it under the terms of the GNU Lesser General Public License as
- * published by the Free Software Foundation, either version 3 of
- * the License, or (at your option) any later version.
- *
- * libde265 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 Lesser General Public License for more details.
- *
- * You should have received a copy of the GNU Lesser General Public License
- * along with libde265.  If not, see <http://www.gnu.org/licenses/>.
- */
-
-#ifndef EN265_H
-#define EN265_H
-
-#ifdef __cplusplus
-extern "C" {
-#endif
-
-#include <libde265/de265.h>
-
-
-// ========== encoder context ==========
-
-struct en265_encoder_context; // private structure
-
-/* Get a new encoder context. Must be freed with en265_free_encoder(). */
-LIBDE265_API en265_encoder_context* en265_new_encoder(void);
-
-/* Free encoder context. May only be called once on a context. */
-LIBDE265_API de265_error en265_free_encoder(en265_encoder_context*);
-
-/* The alloc_userdata pointer will be given to the release_func(). */
-/*
-LIBDE265_API void en265_set_image_release_function(en265_encoder_context*,
-                                                   void (*release_func)(en265_encoder_context*,
-                                                                        struct de265_image*,
-                                                                        void* userdata),
-                                                   void* alloc_userdata);
-*/
-
-// ========== encoder parameters ==========
-
-LIBDE265_API de265_error en265_set_parameter_bool(en265_encoder_context*,
-                                                  const char* parametername,int value);
-LIBDE265_API de265_error en265_set_parameter_int(en265_encoder_context*,
-                                                 const char* parametername,int value);
-LIBDE265_API de265_error en265_set_parameter_string(en265_encoder_context*,
-                                                    const char* parametername,const char* value);
-LIBDE265_API de265_error en265_set_parameter_choice(en265_encoder_context*,
-                                                    const char* parametername,const char* value);
-
-
-LIBDE265_API const char** en265_list_parameters(en265_encoder_context*);
-
-enum en265_parameter_type {
-  en265_parameter_bool,
-  en265_parameter_int,
-  en265_parameter_string,
-  en265_parameter_choice
-};
-
-LIBDE265_API enum en265_parameter_type en265_get_parameter_type(en265_encoder_context*,
-                                                                const char* parametername);
-
-LIBDE265_API const char** en265_list_parameter_choices(en265_encoder_context*,
-                                                       const char* parametername);
-
-
-// --- convenience functions for command-line parameters ---
-
-LIBDE265_API de265_error en265_parse_command_line_parameters(en265_encoder_context*,
-                                                             int* argc, char** argv);
-LIBDE265_API void en265_show_parameters(en265_encoder_context*);
-
-
-
-// ========== encoding loop ==========
-
-LIBDE265_API de265_error en265_start_encoder(en265_encoder_context*, int number_of_threads);
-
-// If we have provided our own memory release function, no image memory will be allocated.
-LIBDE265_API struct de265_image* en265_allocate_image(en265_encoder_context*,
-                                                      int width, int height,
-                                                      enum de265_chroma chroma,
-                                                      de265_PTS pts, void* image_userdata);
-
-LIBDE265_API void* de265_alloc_image_plane(struct de265_image* img, int cIdx,
-                                           void* inputdata, int inputstride, void *userdata);
-LIBDE265_API void de265_free_image_plane(struct de265_image* img, int cIdx);
-
-
-// Request a specification of the image memory layout for an image of the specified dimensions.
-LIBDE265_API void en265_get_image_spec(en265_encoder_context*,
-                                       int width, int height, enum de265_chroma chroma,
-                                       struct de265_image_spec* out_spec);
-
-// Image memory layout specification for an image returned by en265_allocate_image().
-/* TODO: do we need this?
-LIBDE265_API void de265_get_image_spec_from_image(de265_image* img, struct de265_image_spec* spec);
-*/
-
-
-LIBDE265_API de265_error en265_push_image(en265_encoder_context*,
-                                          struct de265_image*); // non-blocking
-
-LIBDE265_API de265_error en265_push_eof(en265_encoder_context*);
-
-// block when there are more than max_input_images in the input queue
-LIBDE265_API de265_error en265_block_on_input_queue_length(en265_encoder_context*,
-                                                           int max_pending_images,
-                                                           int timeout_ms);
-
-LIBDE265_API de265_error en265_trim_input_queue(en265_encoder_context*, int max_pending_images);
-
-LIBDE265_API int  en265_current_input_queue_length(en265_encoder_context*);
-
-// Run encoder in main thread. Only use this when not using background threads.
-LIBDE265_API de265_error en265_encode(en265_encoder_context*);
-
-enum en265_encoder_state
-{
-  EN265_STATE_IDLE,
-  EN265_STATE_WAITING_FOR_INPUT,
-  EN265_STATE_WORKING,
-  EN265_STATE_OUTPUT_QUEUE_FULL,
-  EN265_STATE_EOS
-};
-
-
-LIBDE265_API enum en265_encoder_state en265_get_encoder_state(en265_encoder_context*);
-
-
-enum en265_packet_content_type {
-  EN265_PACKET_VPS,
-  EN265_PACKET_SPS,
-  EN265_PACKET_PPS,
-  EN265_PACKET_SEI,
-  EN265_PACKET_SLICE,
-  EN265_PACKET_SKIPPED_IMAGE
-};
-
-
-enum en265_nal_unit_type {
-  EN265_NUT_TRAIL_N = 0,
-  EN265_NUT_TRAIL_R = 1,
-  EN265_NUT_TSA_N   = 2,
-  EN265_NUT_TSA_R   = 3,
-  EN265_NUT_STSA_N  = 4,
-  EN265_NUT_STSA_R  = 5,
-  EN265_NUT_RADL_N  = 6,
-  EN265_NUT_RADL_R  = 7,
-  EN265_NUT_RASL_N  = 8,
-  EN265_NUT_RASL_R  = 9,
-  EN265_NUT_BLA_W_LP  = 16,
-  EN265_NUT_BLA_W_RADL= 17,
-  EN265_NUT_BLA_N_LP  = 18,
-  EN265_NUT_IDR_W_RADL= 19,
-  EN265_NUT_IDR_N_LP  = 20,
-  EN265_NUT_CRA       = 21,
-  EN265_NUT_VPS   =    32,
-  EN265_NUT_SPS   =    33,
-  EN265_NUT_PPS   =    34,
-  EN265_NUT_AUD   =    35,
-  EN265_NUT_EOS   =    36,
-  EN265_NUT_EOB   =    37,
-  EN265_NUT_FD    =    38,
-  EN265_NUT_PREFIX_SEI = 39,
-  EN265_NUT_SUFFIX_SEI = 40
-};
-
-
-struct en265_packet
-{
-  int version; // currently: 1
-
-  const uint8_t* data;
-  int   length;
-
-  int  frame_number;
-
-  enum en265_packet_content_type content_type;
-  char complete_picture : 1;
-  char final_slice      : 1;
-  char dependent_slice  : 1;
-
-  enum en265_nal_unit_type nal_unit_type;
-  unsigned char nuh_layer_id;
-  unsigned char nuh_temporal_id;
-
-  en265_encoder_context* encoder_context;
-
-  const struct de265_image* input_image;
-  const struct de265_image* reconstruction;
-};
-
-// timeout_ms - timeout in milliseconds. 0 - no timeout, -1 - block forever
-LIBDE265_API struct en265_packet* en265_get_packet(en265_encoder_context*, int timeout_ms);
-LIBDE265_API void en265_free_packet(en265_encoder_context*, struct en265_packet*);
-
-LIBDE265_API int en265_number_of_queued_packets(en265_encoder_context*);
-
-#ifdef __cplusplus
-}
-#endif
-
-
-#endif

libde265-1.0.17.tar.gz/libde265/encoder Deleted

libde265-1.0.17.tar.gz/libde265/encoder/CMakeLists.txt Deleted

libde265-1.0.17.tar.gz/libde265/encoder/algo Deleted

libde265-1.0.17.tar.gz/libde265/encoder/algo/CMakeLists.txt Deleted

libde265-1.0.17.tar.gz/libde265/encoder/algo/algo.cc Deleted

@@ -1,95 +0,0 @@
-/*
- * H.265 video codec.
- * Copyright (c) 2013-2014 struktur AG, Dirk Farin <farin@struktur.de>
- *
- * Authors: Dirk Farin <farin@struktur.de>
- *
- * This file is part of libde265.
- *
- * libde265 is free software: you can redistribute it and/or modify
- * it under the terms of the GNU Lesser General Public License as
- * published by the Free Software Foundation, either version 3 of
- * the License, or (at your option) any later version.
- *
- * libde265 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 Lesser General Public License for more details.
- *
- * You should have received a copy of the GNU Lesser General Public License
- * along with libde265.  If not, see <http://www.gnu.org/licenses/>.
- */
-
-#include "libde265/encoder/algo/algo.h"
-#include "libde265/encoder/encoder-context.h"
-
-#include <stdarg.h>
-
-
-
-#ifdef DE265_LOG_DEBUG
-static int descendLevel = 0;
-
-void Algo::enter()
-{
-  if (logdebug_enabled(LogEncoder)) {
-    printf("%d",descendLevel+1);
-    for (int i=0;i<descendLevel+1;i++) { printf(" "); }
-    printf(":%s\n",name());
-  }
-}
-
-void Algo::descend(const enc_node* node, const char* option, ...)
-{
-  if (logdebug_enabled(LogEncoder)) {
-    descendLevel++;
-    printf("%d ",descendLevel);
-    for (int i=0;i<descendLevel;i++) { printf(" "); }
-
-    va_list va;
-    va_start(va, option);
-    va_end(va);
-
-    fprintf(stdout, ">%s(", name());
-    vfprintf(stdout, option, va);
-    fprintf(stdout, ") %d;%d %dx%d %p\n",node->x,node->y,1<<node->log2Size,1<<node->log2Size,static_cast<void*>(node));
-  }
-}
-
-void Algo::ascend(const enc_node* resultNode, const char* fmt, ...)
-{
-  if (logdebug_enabled(LogEncoder)) {
-    if (fmt != NULL) {
-      printf("%d ",descendLevel);
-      for (int i=0;i<descendLevel;i++) { printf(" "); }
-
-      va_list va;
-      va_start(va, fmt);
-      va_end(va);
-
-      fprintf(stdout, "<%s(", name());
-      vfprintf(stdout, fmt, va);
-      fprintf(stdout, ") <- %p\n",static_cast<void*>(resultNode));
-    }
-
-    descendLevel--;
-  }
-}
-
-void Algo::leaf(const enc_node* node, const char* option, ...)
-{
-  if (logdebug_enabled(LogEncoder)) {
-    printf("%d ",descendLevel+1);
-    for (int i=0;i<descendLevel+1;i++) { printf(" "); }
-
-    va_list va;
-    va_start(va, option);
-    va_end(va);
-
-    fprintf(stdout, "%s(", name());
-    vfprintf(stdout, option, va);
-    fprintf(stdout, ") %d;%d %dx%d\n",node->x,node->y,1<<node->log2Size,1<<node->log2Size);
-  }
-}
-
-#endif

libde265-1.0.17.tar.gz/libde265/encoder/algo/algo.h Deleted

@@ -1,95 +0,0 @@
-/*
- * H.265 video codec.
- * Copyright (c) 2013-2014 struktur AG, Dirk Farin <farin@struktur.de>
- *
- * Authors: Dirk Farin <farin@struktur.de>
- *
- * This file is part of libde265.
- *
- * libde265 is free software: you can redistribute it and/or modify
- * it under the terms of the GNU Lesser General Public License as
- * published by the Free Software Foundation, either version 3 of
- * the License, or (at your option) any later version.
- *
- * libde265 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 Lesser General Public License for more details.
- *
- * You should have received a copy of the GNU Lesser General Public License
- * along with libde265.  If not, see <http://www.gnu.org/licenses/>.
- */
-
-#ifndef ALGO_H
-#define ALGO_H
-
-#include "libde265/encoder/encoder-types.h"
-
-
-/* When entering the next recursion level, it is assumed that
-   a valid CB structure is passed down. Ownership is transferred to
-   the new algorithm. That algorithm passes back a (possibly different)
-   CB structure that the first algorithm should use. The receiving
-   algorithm will be the owner of the passed back algorithm.
-   The original CB structure might have been deleted in the called algorithm.
-
-   When using CodingOptions, it is important to set the passed back
-   enc_node in the CodingOption (set_node()), so that the CodingOption
-   can correctly handle ownership and delete nodes as needed.
-
-   The context_model_table passed down is at the current state.
-   When the algorithm returns, the state should represent the state
-   after running this algorithm.
- */
-
-class Algo
-{
- public:
-  virtual ~Algo() { }
-
-  virtual const char* name() const { return "noname"; }
-
-#ifdef DE265_LOG_DEBUG
-  void enter();
-  void descend(const enc_node* node,const char* option_description, ...);
-  void ascend(const enc_node* resultNode=NULL, const char* fmt=NULL, ...);
-  void leaf(const enc_node* node,const char* option_description, ...);
-#else
-  inline void enter() { }
-  inline void descend(const enc_node*,const char*, ...) { }
-  inline void ascend(const enc_node* resultNode=NULL,const char* fmt=NULL, ...) { }
-  inline void leaf(const enc_node*,const char*, ...) { }
-#endif
-};
-
-
-class Algo_CB : public Algo
-{
- public:
-  virtual ~Algo_CB() { }
-
-  /* The context_model_table that is provided can be modified and
-     even released in the function. On exit, it should be filled with
-     a (optionally new) context_model_table that represents the state
-     after encoding the syntax element. However, to speed up computation,
-     it is also allowed to not modify the context_model_table at all.
-   */
-  virtual enc_cb* analyze(encoder_context*,
-                          context_model_table&,
-                          enc_cb* cb) = 0;
-};
-
-
-class Algo_PB : public Algo
-{
- public:
-  virtual ~Algo_PB() { }
-
-  virtual enc_cb* analyze(encoder_context*,
-                          context_model_table&,
-                          enc_cb* cb,
-                          int PBidx, int x,int y,int w,int h) = 0;
-};
-
-
-#endif

libde265-1.0.17.tar.gz/libde265/encoder/algo/cb-interpartmode.cc Deleted

@@ -1,113 +0,0 @@
-/*
- * H.265 video codec.
- * Copyright (c) 2013-2014 struktur AG, Dirk Farin <farin@struktur.de>
- *
- * Authors: struktur AG, Dirk Farin <farin@struktur.de>
- *
- * This file is part of libde265.
- *
- * libde265 is free software: you can redistribute it and/or modify
- * it under the terms of the GNU Lesser General Public License as
- * published by the Free Software Foundation, either version 3 of
- * the License, or (at your option) any later version.
- *
- * libde265 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 Lesser General Public License for more details.
- *
- * You should have received a copy of the GNU Lesser General Public License
- * along with libde265.  If not, see <http://www.gnu.org/licenses/>.
- */
-
-
-#include "libde265/encoder/algo/cb-interpartmode.h"
-#include "libde265/encoder/algo/coding-options.h"
-#include "libde265/encoder/encoder-context.h"
-#include <assert.h>
-#include <limits>
-#include <math.h>
-
-
-
-enc_cb* Algo_CB_InterPartMode::codeAllPBs(encoder_context* ectx,
-                                          context_model_table& ctxModel,
-                                          enc_cb* cb)
-{
-  int x = cb->x;
-  int y = cb->y;
-  int log2Size = cb->log2Size;
-  int w = 1<<log2Size;
-  int s; // splitSize;
-
-  int nPB;
-  switch (cb->PartMode) {
-  case PART_2Nx2N:
-    cb = mChildAlgo->analyze(ectx, ctxModel, cb, 0, x,y,1<<log2Size,1<<log2Size);
-    break;
-
-  case PART_NxN:
-    s = 1<<(log2Size-1);
-    descend(cb,"NxN(1/4)"); cb = mChildAlgo->analyze(ectx, ctxModel, cb, 0, x  ,y  ,s,s); ascend();
-    descend(cb,"NxN(2/4)"); cb = mChildAlgo->analyze(ectx, ctxModel, cb, 1, x+s,y  ,s,s); ascend();
-    descend(cb,"NxN(3/4)"); cb = mChildAlgo->analyze(ectx, ctxModel, cb, 2, x  ,y+s,s,s); ascend();
-    descend(cb,"NxN(4/4)"); cb = mChildAlgo->analyze(ectx, ctxModel, cb, 3, x+s,y+s,s,s); ascend();
-    break;
-
-  case PART_2NxN:
-    s = 1<<(log2Size-1);
-    descend(cb,"2NxN(1/2)"); cb = mChildAlgo->analyze(ectx, ctxModel, cb, 0, x,y  ,w,s); ascend();
-    descend(cb,"2NxN(2/2)"); cb = mChildAlgo->analyze(ectx, ctxModel, cb, 1, x,y+s,w,s); ascend();
-    break;
-
-  case PART_Nx2N:
-    s = 1<<(log2Size-1);
-    descend(cb,"Nx2N(1/2)"); cb = mChildAlgo->analyze(ectx, ctxModel, cb, 0, x  ,y,s,w); ascend();
-    descend(cb,"Nx2N(2/2)"); cb = mChildAlgo->analyze(ectx, ctxModel, cb, 1, x+s,y,s,w); ascend();
-    break;
-
-  case PART_2NxnU:
-    s = 1<<(log2Size-2);
-    descend(cb,"2NxnU(1/2)"); cb = mChildAlgo->analyze(ectx, ctxModel, cb, 0, x,y  ,w,s);   ascend();
-    descend(cb,"2NxnU(2/2)"); cb = mChildAlgo->analyze(ectx, ctxModel, cb, 1, x,y+s,w,w-s); ascend();
-    break;
-
-  case PART_2NxnD:
-    s = 1<<(log2Size-2);
-    descend(cb,"2NxnD(1/2)"); cb = mChildAlgo->analyze(ectx, ctxModel, cb, 0, x,y    ,w,w-s); ascend();
-    descend(cb,"2NxnD(2/2)"); cb = mChildAlgo->analyze(ectx, ctxModel, cb, 1, x,y+w-s,w,s);   ascend();
-    break;
-
-  case PART_nLx2N:
-    s = 1<<(log2Size-2);
-    descend(cb,"nLx2N(1/2)"); cb = mChildAlgo->analyze(ectx, ctxModel, cb, 0, x  ,y,s  ,w); ascend();
-    descend(cb,"nLx2N(2/2)"); cb = mChildAlgo->analyze(ectx, ctxModel, cb, 1, x+s,y,w-s,w); ascend();
-    break;
-
-  case PART_nRx2N:
-    s = 1<<(log2Size-2);
-    descend(cb,"nRx2N(1/2)"); cb = mChildAlgo->analyze(ectx, ctxModel, cb, 0, x    ,y,w-s,w); ascend();
-    descend(cb,"nRx2N(2/2)"); cb = mChildAlgo->analyze(ectx, ctxModel, cb, 1, x+w-s,y,s  ,w); ascend();
-    break;
-  }
-
-  return cb;
-}
-
-
-enc_cb* Algo_CB_InterPartMode_Fixed::analyze(encoder_context* ectx,
-                                             context_model_table& ctxModel,
-                                             enc_cb* cb)
-{
-  const int x = cb->x;
-  const int y = cb->y;
-
-  enum PartMode partMode = mParams.partMode();
-
-  cb->PartMode = partMode;
-  ectx->img->set_PartMode(x,y, partMode);
-
-  cb = codeAllPBs(ectx,ctxModel,cb);
-
-  return cb;
-}

libde265-1.0.17.tar.gz/libde265/encoder/algo/cb-interpartmode.h Deleted

@@ -1,108 +0,0 @@
-/*
- * H.265 video codec.
- * Copyright (c) 2013-2014 struktur AG, Dirk Farin <farin@struktur.de>
- *
- * Authors: Dirk Farin <farin@struktur.de>
- *
- * This file is part of libde265.
- *
- * libde265 is free software: you can redistribute it and/or modify
- * it under the terms of the GNU Lesser General Public License as
- * published by the Free Software Foundation, either version 3 of
- * the License, or (at your option) any later version.
- *
- * libde265 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 Lesser General Public License for more details.
- *
- * You should have received a copy of the GNU Lesser General Public License
- * along with libde265.  If not, see <http://www.gnu.org/licenses/>.
- */
-
-#ifndef CB_INTERPARTMODE_H
-#define CB_INTERPARTMODE_H
-
-#include "libde265/nal-parser.h"
-#include "libde265/decctx.h"
-#include "libde265/slice.h"
-#include "libde265/scan.h"
-#include "libde265/intrapred.h"
-#include "libde265/transform.h"
-#include "libde265/fallback-dct.h"
-#include "libde265/quality.h"
-#include "libde265/fallback.h"
-#include "libde265/encoder/configparam.h"
-
-#include "libde265/encoder/algo/algo.h"
-#include "libde265/encoder/algo/tb-intrapredmode.h"
-#include "libde265/encoder/algo/tb-split.h"
-#include "libde265/encoder/algo/cb-intrapartmode.h"
-
-
-// ========== CB Intra/Inter decision ==========
-
-class Algo_CB_InterPartMode : public Algo_CB
-{
- public:
-  virtual ~Algo_CB_InterPartMode() { }
-
-  void setChildAlgo(Algo_PB* algo) { mChildAlgo = algo; }
-
-  virtual const char* name() const { return "cb-interpartmode"; }
-
- protected:
-  Algo_PB* mChildAlgo;
-
-  enc_cb* codeAllPBs(encoder_context*,
-                     context_model_table&,
-                     enc_cb* cb);
-};
-
-
-
-
-class option_InterPartMode : public choice_option<enum PartMode> // choice_option
-{
- public:
-  option_InterPartMode() {
-    add_choice("2Nx2N", PART_2Nx2N, true);
-    add_choice("NxN",   PART_NxN);
-    add_choice("Nx2N",  PART_Nx2N);
-    add_choice("2NxN",  PART_2NxN);
-    add_choice("2NxnU", PART_2NxnU);
-    add_choice("2NxnD", PART_2NxnD);
-    add_choice("nLx2N", PART_nLx2N);
-    add_choice("nRx2N", PART_nRx2N);
-  }
-};
-
-class Algo_CB_InterPartMode_Fixed : public Algo_CB_InterPartMode
-{
- public:
-  struct params
-  {
-    params() {
-      partMode.set_ID("CB-InterPartMode-Fixed-partMode");
-    }
-
-    option_InterPartMode partMode;
-  };
-
-  void registerParams(config_parameters& config) {
-    config.add_option(&mParams.partMode);
-  }
-
-  void setParams(const params& p) { mParams=p; }
-
-  virtual enc_cb* analyze(encoder_context*,
-                          context_model_table&,
-                          enc_cb* cb);
-
-  virtual const char* name() const { return "cb-interpartmode-fixed"; }
-
- private:
-  params mParams;
-};
-
-#endif

libde265-1.0.17.tar.gz/libde265/encoder/algo/cb-intra-inter.cc Deleted

@@ -1,132 +0,0 @@
-/*
- * H.265 video codec.
- * Copyright (c) 2013-2014 struktur AG, Dirk Farin <farin@struktur.de>
- *
- * Authors: struktur AG, Dirk Farin <farin@struktur.de>
- *
- * This file is part of libde265.
- *
- * libde265 is free software: you can redistribute it and/or modify
- * it under the terms of the GNU Lesser General Public License as
- * published by the Free Software Foundation, either version 3 of
- * the License, or (at your option) any later version.
- *
- * libde265 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 Lesser General Public License for more details.
- *
- * You should have received a copy of the GNU Lesser General Public License
- * along with libde265.  If not, see <http://www.gnu.org/licenses/>.
- */
-
-
-#include "libde265/encoder/algo/cb-intra-inter.h"
-#include "libde265/encoder/algo/coding-options.h"
-#include "libde265/encoder/encoder-context.h"
-#include <assert.h>
-#include <limits>
-#include <math.h>
-
-
-
-enc_cb* Algo_CB_IntraInter_BruteForce::analyze(encoder_context* ectx,
-                                               context_model_table& ctxModel,
-                                               enc_cb* cb)
-{
-  assert(cb->pcm_flag==0);
-
-  bool try_intra = true;
-  bool try_inter = (ectx->shdr->slice_type != SLICE_TYPE_I);
-
-  bool debug_halt = try_inter;
-  try_inter = false;
-  //try_intra = !try_inter; // TODO HACK: no intra in inter frames
-
-  if (ectx->imgdata->frame_number > 0) {
-    //printf("%d\n",ectx->imgdata->frame_number);
-  }
-
-  // 0: intra
-  // 1: inter
-
-  CodingOptions<enc_cb> options(ectx,cb,ctxModel);
-
-  CodingOption<enc_cb> option_intra = options.new_option(try_intra);
-  CodingOption<enc_cb> option_inter = options.new_option(try_inter);
-
-  options.start();
-
-  enc_cb* cb_inter = NULL;
-  enc_cb* cb_intra = NULL;
-
-  const int log2CbSize = cb->log2Size;
-  const int x = cb->x;
-  const int y = cb->y;
-
-
-  // try encoding with inter
-
-  if (option_inter) {
-    option_inter.begin();
-    cb_inter = option_inter.get_node();
-
-    cb_inter->PredMode = MODE_INTER;
-    ectx->img->set_pred_mode(x,y, log2CbSize, MODE_INTER);
-
-    enc_cb* cb_result;
-    descend(cb,"inter");
-    cb_result=mInterAlgo->analyze(ectx, option_inter.get_context(), cb_inter);
-    ascend();
-
-    if (cb_result->PredMode != MODE_SKIP) {
-      CABAC_encoder_estim* cabac = option_inter.get_cabac();
-      cabac->reset();
-
-      cabac->write_CABAC_bit(CONTEXT_MODEL_PRED_MODE_FLAG, 0); // 0 - inter
-      float rate_pred_mode_flag = cabac->getRDBits();
-      //printf("inter bits: %f\n", rate_pred_mode_flag);
-
-      cb_result->rate += rate_pred_mode_flag;
-    }
-
-    option_inter.set_node(cb_result);
-
-    option_inter.end();
-  }
-
-
-  // try intra
-
-  if (option_intra) {
-    option_intra.begin();
-    cb_intra = option_intra.get_node();
-
-    cb_intra->PredMode = MODE_INTRA;
-    ectx->img->set_pred_mode(x,y, log2CbSize, MODE_INTRA);
-
-    enc_cb* cb_result;
-    descend(cb,"intra");
-    cb_result=mIntraAlgo->analyze(ectx, option_intra.get_context(), cb_intra);
-    ascend();
-
-    if (ectx->shdr->slice_type != SLICE_TYPE_I) {
-      CABAC_encoder_estim* cabac = option_intra.get_cabac();
-      cabac->reset();
-
-      cabac->write_CABAC_bit(CONTEXT_MODEL_PRED_MODE_FLAG, 1); // 1 - intra
-      float rate_pred_mode_flag = cabac->getRDBits();
-      //printf("intra bits: %f\n", rate_pred_mode_flag);
-
-      cb_result->rate += rate_pred_mode_flag;
-    }
-
-    option_intra.set_node(cb_result);
-
-    option_intra.end();
-  }
-
-
-  options.compute_rdo_costs();
-  return options.return_best_rdo_node();
-}

libde265-1.0.17.tar.gz/libde265/encoder/algo/cb-intra-inter.h Deleted

@@ -1,68 +0,0 @@
-/*
- * H.265 video codec.
- * Copyright (c) 2013-2014 struktur AG, Dirk Farin <farin@struktur.de>
- *
- * Authors: Dirk Farin <farin@struktur.de>
- *
- * This file is part of libde265.
- *
- * libde265 is free software: you can redistribute it and/or modify
- * it under the terms of the GNU Lesser General Public License as
- * published by the Free Software Foundation, either version 3 of
- * the License, or (at your option) any later version.
- *
- * libde265 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 Lesser General Public License for more details.
- *
- * You should have received a copy of the GNU Lesser General Public License
- * along with libde265.  If not, see <http://www.gnu.org/licenses/>.
- */
-
-#ifndef CB_INTRA_INTER_H
-#define CB_INTRA_INTER_H
-
-#include "libde265/nal-parser.h"
-#include "libde265/decctx.h"
-#include "libde265/slice.h"
-#include "libde265/scan.h"
-#include "libde265/intrapred.h"
-#include "libde265/transform.h"
-#include "libde265/fallback-dct.h"
-#include "libde265/quality.h"
-#include "libde265/fallback.h"
-#include "libde265/encoder/configparam.h"
-
-#include "libde265/encoder/algo/algo.h"
-#include "libde265/encoder/algo/tb-intrapredmode.h"
-#include "libde265/encoder/algo/tb-split.h"
-#include "libde265/encoder/algo/cb-intrapartmode.h"
-
-
-// ========== CB Intra/Inter decision ==========
-
-class Algo_CB_IntraInter : public Algo_CB
-{
- public:
-  virtual ~Algo_CB_IntraInter() { }
-
-  void setIntraChildAlgo(Algo_CB* algo) { mIntraAlgo = algo; }
-  void setInterChildAlgo(Algo_CB* algo) { mInterAlgo = algo; }
-
-  virtual const char* name() const { return "cb-intra-inter"; }
-
- protected:
-  Algo_CB* mIntraAlgo;
-  Algo_CB* mInterAlgo;
-};
-
-class Algo_CB_IntraInter_BruteForce : public Algo_CB_IntraInter
-{
- public:
-  virtual enc_cb* analyze(encoder_context*,
-                          context_model_table&,
-                          enc_cb* cb);
-};
-
-#endif

libde265-1.0.17.tar.gz/libde265/encoder/algo/cb-intrapartmode.cc Deleted

@@ -1,185 +0,0 @@
-/*
- * H.265 video codec.
- * Copyright (c) 2013-2014 struktur AG, Dirk Farin <farin@struktur.de>
- *
- * Authors: struktur AG, Dirk Farin <farin@struktur.de>
- *
- * This file is part of libde265.
- *
- * libde265 is free software: you can redistribute it and/or modify
- * it under the terms of the GNU Lesser General Public License as
- * published by the Free Software Foundation, either version 3 of
- * the License, or (at your option) any later version.
- *
- * libde265 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 Lesser General Public License for more details.
- *
- * You should have received a copy of the GNU Lesser General Public License
- * along with libde265.  If not, see <http://www.gnu.org/licenses/>.
- */
-
-
-#include "libde265/encoder/algo/cb-intrapartmode.h"
-#include "libde265/encoder/algo/coding-options.h"
-#include "libde265/encoder/encoder-context.h"
-#include <assert.h>
-#include <limits>
-#include <math.h>
-#include <iostream>
-
-
-#define ENCODER_DEVELOPMENT 1
-
-
-
-enc_cb* Algo_CB_IntraPartMode_BruteForce::analyze(encoder_context* ectx,
-                                                  context_model_table& ctxModel,
-                                                  enc_cb* cb_in)
-{
-  const int log2CbSize = cb_in->log2Size;
-  const int x = cb_in->x;
-  const int y = cb_in->y;
-
-  const bool can_use_NxN = ((log2CbSize == ectx->get_sps().Log2MinCbSizeY) &&
-                            (log2CbSize >  ectx->get_sps().Log2MinTrafoSize));
-
-
-  // test all modes
-
-  assert(cb_in->pcm_flag==0);
-
-
-  // 0: 2Nx2N  (always checked)
-  // 1:  NxN   (only checked at MinCbSize)
-
-  CodingOptions<enc_cb> options(ectx,cb_in,ctxModel);
-  CodingOption<enc_cb> option2;
-  option0 = options.new_option(true);
-  option1 = options.new_option(can_use_NxN);
-
-  options.start();
-
-  for (int p=0;p<2;p++)
-    if (optionp) {
-      optionp.begin();
-
-      enc_cb* cb = optionp.get_node();
-      *(cb_in->downPtr) = cb;
-
-      // --- set intra prediction mode ---
-
-      cb->PartMode = (p==0 ? PART_2Nx2N : PART_NxN);
-
-      ectx->img->set_pred_mode(x,y, log2CbSize, cb->PredMode);  // TODO: probably unnecessary
-      ectx->img->set_PartMode (x,y, cb->PartMode);
-
-
-      // encode transform tree
-
-      int IntraSplitFlag= (cb->PredMode == MODE_INTRA && cb->PartMode == PART_NxN);
-      int MaxTrafoDepth = ectx->get_sps().max_transform_hierarchy_depth_intra + IntraSplitFlag;
-
-      descend(cb,p==0 ? "2Nx2N" : "NxN");
-
-      enc_tb* tb = new enc_tb(x,y,log2CbSize,cb);
-      tb->downPtr = &cb->transform_tree;
-
-      cb->transform_tree = mTBIntraPredModeAlgo->analyze(ectx, optionp.get_context(),
-                                                         ectx->imgdata->input, tb,
-                                                         0, MaxTrafoDepth, IntraSplitFlag);
-
-      ascend();
-
-      cb->distortion = cb->transform_tree->distortion;
-      cb->rate       = cb->transform_tree->rate;
-
-
-      // rate for cu syntax
-
-      logtrace(LogSymbols,"$1 part_mode=%d\n",cb->PartMode);
-      if (log2CbSize == ectx->get_sps().Log2MinCbSizeY) {
-        int bin = (cb->PartMode==PART_2Nx2N);
-        optionp.get_cabac()->reset();
-        optionp.get_cabac()->write_CABAC_bit(CONTEXT_MODEL_PART_MODE+0, bin);
-        cb->rate += optionp.get_cabac()->getRDBits();
-      }
-
-      optionp.end();
-    }
-
-  options.compute_rdo_costs();
-  enc_cb* bestCB = options.return_best_rdo_node();
-
-  return bestCB;
-}
-
-
-enc_cb* Algo_CB_IntraPartMode_Fixed::analyze(encoder_context* ectx,
-                                             context_model_table& ctxModel,
-                                             enc_cb* cb)
-{
-  enum PartMode PartMode = mParams.partMode();
-
-
-  const int log2CbSize = cb->log2Size;
-  const int x = cb->x;
-  const int y = cb->y;
-
-
-  // NxN can only be applied at minimum CB size.
-  // If we are not at the minimum size, we have to use 2Nx2N.
-
-  if (PartMode==PART_NxN && log2CbSize != ectx->get_sps().Log2MinCbSizeY) {
-    PartMode = PART_2Nx2N;
-  }
-
-
-  // --- set intra prediction mode ---
-
-  cb->PartMode = PartMode;
-  ectx->img->set_PartMode(x,y, PartMode);
-
-
-  // encode transform tree
-
-  int IntraSplitFlag= (cb->PredMode == MODE_INTRA && cb->PartMode == PART_NxN);
-  int MaxTrafoDepth = ectx->get_sps().max_transform_hierarchy_depth_intra + IntraSplitFlag;
-
-  enc_tb* tb = new enc_tb(x,y,log2CbSize,cb);
-  tb->blkIdx = 0;
-  tb->downPtr = &cb->transform_tree;
-
-  descend(cb,"fixed:%s", (PartMode==PART_2Nx2N ? "2Nx2N":"NxN"));
-  cb->transform_tree = mTBIntraPredModeAlgo->analyze(ectx, ctxModel,
-                                                     ectx->imgdata->input, tb,
-                                                     0, MaxTrafoDepth, IntraSplitFlag);
-  ascend();
-
-
-  // rate and distortion for this CB
-
-  cb->distortion = cb->transform_tree->distortion;
-  cb->rate       = cb->transform_tree->rate;
-
-
-  // rate for cu syntax
-
-  CABAC_encoder_estim estim;
-  estim.set_context_models(&ctxModel);
-
-  //encode_coding_unit(ectx,&estim,cb,x,y,log2CbSize, false);
-
-  //encode_part_mode(ectx,&estim, MODE_INTRA, PartMode, 0);
-
-  logtrace(LogSymbols,"$1 part_mode=%d\n",PartMode);
-  if (log2CbSize == ectx->get_sps().Log2MinCbSizeY) {
-    int bin = (PartMode==PART_2Nx2N);
-    estim.write_CABAC_bit(CONTEXT_MODEL_PART_MODE+0, bin);
-  }
-
-  cb->rate += estim.getRDBits();
-
-  return cb;
-}

libde265-1.0.17.tar.gz/libde265/encoder/algo/cb-intrapartmode.h Deleted

@@ -1,149 +0,0 @@
-/*
- * H.265 video codec.
- * Copyright (c) 2013-2014 struktur AG, Dirk Farin <farin@struktur.de>
- *
- * Authors: Dirk Farin <farin@struktur.de>
- *
- * This file is part of libde265.
- *
- * libde265 is free software: you can redistribute it and/or modify
- * it under the terms of the GNU Lesser General Public License as
- * published by the Free Software Foundation, either version 3 of
- * the License, or (at your option) any later version.
- *
- * libde265 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 Lesser General Public License for more details.
- *
- * You should have received a copy of the GNU Lesser General Public License
- * along with libde265.  If not, see <http://www.gnu.org/licenses/>.
- */
-
-#ifndef CB_INTRAPARTMODE_H
-#define CB_INTRAPARTMODE_H
-
-#include "libde265/nal-parser.h"
-#include "libde265/decctx.h"
-#include "libde265/encoder/encoder-types.h"
-#include "libde265/slice.h"
-#include "libde265/scan.h"
-#include "libde265/intrapred.h"
-#include "libde265/transform.h"
-#include "libde265/fallback-dct.h"
-#include "libde265/quality.h"
-#include "libde265/fallback.h"
-#include "libde265/encoder/configparam.h"
-
-#include "libde265/encoder/algo/algo.h"
-#include "libde265/encoder/algo/tb-intrapredmode.h"
-#include "libde265/encoder/algo/tb-split.h"
-
-
-/*  Encoder search tree, bottom up:
-
-    - Algo_TB_Split - whether TB is split or not
-
-    - Algo_TB_IntraPredMode - choose the intra prediction mode (or NOP, if at the wrong tree level)
-
-    - Algo_CB_IntraPartMode - choose between NxN and 2Nx2N intra parts
-
-    - Algo_CB_Split - whether CB is split or not
-
-    - Algo_CTB_QScale - select QScale on CTB granularity
- */
-
-
-// ========== CB intra NxN vs. 2Nx2N decision ==========
-
-enum ALGO_CB_IntraPartMode {
-  ALGO_CB_IntraPartMode_BruteForce,
-  ALGO_CB_IntraPartMode_Fixed
-};
-
-class option_ALGO_CB_IntraPartMode : public choice_option<enum ALGO_CB_IntraPartMode>
-{
- public:
-  option_ALGO_CB_IntraPartMode() {
-    add_choice("fixed",      ALGO_CB_IntraPartMode_Fixed);
-    add_choice("brute-force",ALGO_CB_IntraPartMode_BruteForce, true);
-  }
-};
-
-
-class Algo_CB_IntraPartMode : public Algo_CB
-{
- public:
-  Algo_CB_IntraPartMode() : mTBIntraPredModeAlgo(NULL) { }
-  virtual ~Algo_CB_IntraPartMode() { }
-
-  virtual enc_cb* analyze(encoder_context*,
-                          context_model_table&,
-                          enc_cb* cb) = 0;
-
-  void setChildAlgo(Algo_TB_IntraPredMode* algo) { mTBIntraPredModeAlgo = algo; }
-
-  virtual const char* name() const { return "cb-intrapartmode"; }
-
- protected:
-  Algo_TB_IntraPredMode* mTBIntraPredModeAlgo;
-};
-
-/* Try both NxN, 2Nx2N and choose better one.
- */
-class Algo_CB_IntraPartMode_BruteForce : public Algo_CB_IntraPartMode
-{
- public:
-  virtual enc_cb* analyze(encoder_context*,
-                          context_model_table&,
-                          enc_cb* cb);
-
-  virtual const char* name() const { return "cb-intrapartmode-bruteforce"; }
-};
-
-
-class option_PartMode : public choice_option<enum PartMode> // choice_option
-{
- public:
-  option_PartMode() {
-    add_choice("NxN",   PART_NxN);
-    add_choice("2Nx2N", PART_2Nx2N, true);
-  }
-};
-
-
-/* Always use choose selected part mode.
-   If NxN is chosen but cannot be applied (CB tree not at maximum depth), 2Nx2N is used instead.
- */
-class Algo_CB_IntraPartMode_Fixed : public Algo_CB_IntraPartMode
-{
- public:
- Algo_CB_IntraPartMode_Fixed() { }
-
-  struct params
-  {
-    params() {
-      partMode.set_ID("CB-IntraPartMode-Fixed-partMode");
-    }
-
-    option_PartMode partMode;
-  };
-
-  void registerParams(config_parameters& config) {
-    config.add_option(&mParams.partMode);
-  }
-
-  void setParams(const params& p) { mParams=p; }
-
-  virtual enc_cb* analyze(encoder_context* ectx,
-                          context_model_table& ctxModel,
-                          enc_cb* cb);
-
-  virtual const char* name() const { return "cb-intrapartmode-fixed"; }
-
- private:
-  params mParams;
-};
-
-
-#endif

libde265-1.0.17.tar.gz/libde265/encoder/algo/cb-mergeindex.cc Deleted

@@ -1,176 +0,0 @@
-/*
- * H.265 video codec.
- * Copyright (c) 2013-2014 struktur AG, Dirk Farin <farin@struktur.de>
- *
- * Authors: struktur AG, Dirk Farin <farin@struktur.de>
- *
- * This file is part of libde265.
- *
- * libde265 is free software: you can redistribute it and/or modify
- * it under the terms of the GNU Lesser General Public License as
- * published by the Free Software Foundation, either version 3 of
- * the License, or (at your option) any later version.
- *
- * libde265 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 Lesser General Public License for more details.
- *
- * You should have received a copy of the GNU Lesser General Public License
- * along with libde265.  If not, see <http://www.gnu.org/licenses/>.
- */
-
-
-#include "libde265/encoder/algo/cb-mergeindex.h"
-#include "libde265/encoder/encoder-context.h"
-#include "libde265/encoder/encoder-syntax.h"
-#include "libde265/encoder/encoder-motion.h"
-#include <assert.h>
-#include <limits>
-#include <math.h>
-
-
-
-
-enc_cb* Algo_CB_MergeIndex_Fixed::analyze(encoder_context* ectx,
-                                          context_model_table& ctxModel,
-                                          enc_cb* cb)
-{
-  assert(cb->split_cu_flag==false);
-  assert(cb->PredMode==MODE_SKIP); // TODO: || (cb->PredMode==MODE_INTER && cb->inter.skip_flag));
-
-
-  PBMotion mergeCandList5;
-
-  int partIdx = 0;
-
-  int cbSize = 1 << cb->log2Size;
-
-  get_merge_candidate_list_from_tree(ectx, ectx->shdr,
-                                     cb->x, cb->y, // xC/yC
-                                     cb->x, cb->y, // xP/yP
-                                     cbSize, // nCS
-                                     cbSize,cbSize, // nPbW/nPbH
-                                     partIdx, // partIdx
-                                     mergeCandList);
-
-  PBMotionCoding&   spec = cb->inter.pbpartIdx.spec;
-
-  spec.merge_flag = 1; // we use merge mode
-  spec.merge_idx  = 0;
-
-
-  // build prediction
-
-  // previous frame (TODO)
-  const de265_image* refimg = ectx->get_image(ectx->imgdata->frame_number -1);
-
-  //printf("prev frame: %p %d\n",refimg,ectx->imgdata->frame_number);
-
-  /*
-    printf("#l0: %d\n",ectx->imgdata->shdr.num_ref_idx_l0_active);
-    printf("#l1: %d\n",ectx->imgdata->shdr.num_ref_idx_l1_active);
-
-    for (int i=0;i<ectx->imgdata->shdr.num_ref_idx_l0_active;i++)
-    printf("RefPixList0%d = %d\n", i, ectx->imgdata->shdr.RefPicList0i);
-  */
-
-  // TODO: fake motion data
-
-  const PBMotion& vec = mergeCandListspec.merge_idx;
-  cb->inter.pbpartIdx.motion = vec;
-
-  //ectx->img->set_mv_info(cb->x,cb->y, 1<<cb->log2Size,1<<cb->log2Size, vec);
-
-  /*
-  generate_inter_prediction_samples(ectx, ectx->shdr, ectx->prediction,
-                                    cb->x,cb->y, // int xC,int yC,
-                                    0,0,         // int xB,int yB,
-                                    1<<cb->log2Size, // int nCS,
-                                    1<<cb->log2Size,
-                                    1<<cb->log2Size, // int nPbW,int nPbH,
-                                    &vec);
-  */
-
-  generate_inter_prediction_samples(ectx, ectx->shdr, ectx->img,
-                                    cb->x,cb->y, // int xC,int yC,
-                                    0,0,         // int xB,int yB,
-                                    1<<cb->log2Size, // int nCS,
-                                    1<<cb->log2Size,
-                                    1<<cb->log2Size, // int nPbW,int nPbH,
-                                    &vec);
-
-  /*
-  printBlk("merge prediction:",
-           ectx->img->get_image_plane_at_pos(0, cb->x,cb->y), 1<<cb->log2Size,
-           ectx->img->get_image_stride(0),
-           "merge ");
-  */
-
-  // estimate rate for sending merge index
-
-  //CABAC_encoder_estim cabac;
-  //cabac.write_bits();
-
-  int IntraSplitFlag = 0;
-  int MaxTrafoDepth = ectx->get_sps().max_transform_hierarchy_depth_inter;
-
-  if (mCodeResidual) {
-    assert(false);
-    descend(cb,"with residual");
-    assert(false);
-    /* TODO
-    cb->transform_tree = mTBSplit->analyze(ectx,ctxModel, ectx->imgdata->input, NULL, cb,
-                                           cb->x,cb->y,cb->x,cb->y, cb->log2Size,0,
-                                           0, MaxTrafoDepth, IntraSplitFlag);
-    */
-    ascend();
-
-    cb->inter.rqt_root_cbf = ! cb->transform_tree->isZeroBlock();
-
-    cb->distortion = cb->transform_tree->distortion;
-    cb->rate       = cb->transform_tree->rate;
-  }
-  else {
-    const de265_image* input = ectx->imgdata->input;
-    //de265_image* img   = ectx->prediction;
-    int x0 = cb->x;
-    int y0 = cb->y;
-    int tbSize = 1<<cb->log2Size;
-
-    CABAC_encoder_estim cabac;
-    cabac.set_context_models(&ctxModel);
-    encode_merge_idx(ectx, &cabac, spec.merge_idx);
-
-    leaf(cb,"no residual");
-
-    cb->rate = cabac.getRDBits();
-
-    cb->inter.rqt_root_cbf = 0;
-
-
-    enc_tb* tb = new enc_tb(x0,y0,cb->log2Size,cb);
-    tb->downPtr = &cb->transform_tree;
-    cb->transform_tree = tb;
-
-    tb->reconstruct(ectx, ectx->img); // reconstruct luma
-
-    /*
-    printBlk("distortion input:",
-             input->get_image_plane_at_pos(0,x0,y0), 1<<cb->log2Size,
-             input->get_image_stride(0),
-             "input ");
-
-    printBlk("distortion prediction:",
-             ectx->img->get_image_plane_at_pos(0,x0,y0), 1<<cb->log2Size,
-             ectx->img->get_image_stride(0),
-             "pred ");
-    */
-
-    cb->distortion = compute_distortion_ssd(input, ectx->img, x0,y0, cb->log2Size, 0);
-  }
-
-  //printf("%d;%d rqt_root_cbf=%d\n",cb->x,cb->y,cb->inter.rqt_root_cbf);
-
-  return cb;
-}

libde265-1.0.17.tar.gz/libde265/encoder/algo/cb-mergeindex.h Deleted

@@ -1,70 +0,0 @@
-/*
- * H.265 video codec.
- * Copyright (c) 2013-2014 struktur AG, Dirk Farin <farin@struktur.de>
- *
- * Authors: Dirk Farin <farin@struktur.de>
- *
- * This file is part of libde265.
- *
- * libde265 is free software: you can redistribute it and/or modify
- * it under the terms of the GNU Lesser General Public License as
- * published by the Free Software Foundation, either version 3 of
- * the License, or (at your option) any later version.
- *
- * libde265 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 Lesser General Public License for more details.
- *
- * You should have received a copy of the GNU Lesser General Public License
- * along with libde265.  If not, see <http://www.gnu.org/licenses/>.
- */
-
-#ifndef CB_MERGEINDEX_H
-#define CB_MERGEINDEX_H
-
-#include "libde265/nal-parser.h"
-#include "libde265/decctx.h"
-#include "libde265/slice.h"
-#include "libde265/scan.h"
-#include "libde265/intrapred.h"
-#include "libde265/transform.h"
-#include "libde265/fallback-dct.h"
-#include "libde265/quality.h"
-#include "libde265/fallback.h"
-#include "libde265/encoder/configparam.h"
-
-#include "libde265/encoder/algo/algo.h"
-#include "libde265/encoder/algo/tb-split.h"
-
-
-// ========== CB Skip/Inter decision ==========
-
-class Algo_CB_MergeIndex : public Algo_CB
-{
- public:
-  Algo_CB_MergeIndex() : mCodeResidual(false) { }
-  virtual ~Algo_CB_MergeIndex() { }
-
-  void set_code_residual(bool flag=true) { mCodeResidual=flag; }
-
-  void setChildAlgo(Algo_TB_Split* algo) { mTBSplit = algo; }
-  // TODO void setInterChildAlgo(Algo_CB_IntraPartMode* algo) { mInterPartModeAlgo = algo; }
-
-  virtual const char* name() const { return "cb-mergeindex"; }
-
- protected:
-  Algo_TB_Split* mTBSplit;
-
-  bool mCodeResidual;
-};
-
-class Algo_CB_MergeIndex_Fixed : public Algo_CB_MergeIndex
-{
- public:
-  virtual enc_cb* analyze(encoder_context*,
-                          context_model_table&,
-                          enc_cb* cb);
-};
-
-#endif

libde265-1.0.17.tar.gz/libde265/encoder/algo/cb-skip.cc Deleted

@@ -1,114 +0,0 @@
-/*
- * H.265 video codec.
- * Copyright (c) 2013-2014 struktur AG, Dirk Farin <farin@struktur.de>
- *
- * Authors: struktur AG, Dirk Farin <farin@struktur.de>
- *
- * This file is part of libde265.
- *
- * libde265 is free software: you can redistribute it and/or modify
- * it under the terms of the GNU Lesser General Public License as
- * published by the Free Software Foundation, either version 3 of
- * the License, or (at your option) any later version.
- *
- * libde265 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 Lesser General Public License for more details.
- *
- * You should have received a copy of the GNU Lesser General Public License
- * along with libde265.  If not, see <http://www.gnu.org/licenses/>.
- */
-
-
-#include "libde265/encoder/algo/cb-skip.h"
-#include "libde265/encoder/algo/coding-options.h"
-#include "libde265/encoder/encoder-syntax.h"
-#include "libde265/encoder/encoder-context.h"
-#include <assert.h>
-#include <limits>
-#include <math.h>
-
-
-
-
-enc_cb* Algo_CB_Skip_BruteForce::analyze(encoder_context* ectx,
-                                         context_model_table& ctxModel,
-                                         enc_cb* cb)
-{
-  bool try_skip  = (ectx->shdr->slice_type != SLICE_TYPE_I);
-  bool try_nonskip = true;
-
-  //try_nonskip = !try_skip;
-
-  CodingOptions<enc_cb> options(ectx,cb,ctxModel);
-  CodingOption<enc_cb> option_skip    = options.new_option(try_skip);
-  CodingOption<enc_cb> option_nonskip = options.new_option(try_nonskip);
-  options.start();
-
-  for (int i=0;i<CONTEXT_MODEL_TABLE_LENGTH;i++) {
-    //printf("%i: %d/%d\n",i, ctxModeli.state, ctxModeli.MPSbit);
-  }
-
-  if (option_skip) {
-    CodingOption<enc_cb>& opt = option_skip;
-    opt.begin();
-
-    enc_cb* cb = opt.get_node();
-
-    // calc rate for skip flag (=true)
-
-    CABAC_encoder_estim* cabac = opt.get_cabac();
-    encode_cu_skip_flag(ectx, cabac, cb, true);
-    float rate_pred_mode = cabac->getRDBits();
-    cabac->reset();
-
-    // set skip flag
-
-    cb->PredMode = MODE_SKIP;
-    ectx->img->set_pred_mode(cb->x,cb->y, cb->log2Size, cb->PredMode);
-
-    // encode CB
-
-    descend(cb,"yes");
-    cb = mSkipAlgo->analyze(ectx, opt.get_context(), cb);
-    ascend();
-
-    // add rate for PredMode
-
-    cb->rate += rate_pred_mode;
-    opt.set_node(cb);
-    opt.end();
-  }
-
-  if (option_nonskip) {
-    CodingOption<enc_cb>& opt = option_nonskip;
-    enc_cb* cb = opt.get_node();
-
-    opt.begin();
-
-    // calc rate for skip flag (=true)
-
-    float rate_pred_mode = 0;
-
-    if (try_skip) {
-      CABAC_encoder_estim* cabac = opt.get_cabac();
-      encode_cu_skip_flag(ectx, cabac, cb, false);
-      rate_pred_mode = cabac->getRDBits();
-      cabac->reset();
-    }
-
-    descend(cb,"no");
-    cb = mNonSkipAlgo->analyze(ectx, opt.get_context(), cb);
-    ascend();
-
-    // add rate for PredMode
-
-    cb->rate += rate_pred_mode;
-    opt.set_node(cb);
-    opt.end();
-  }
-
-  options.compute_rdo_costs();
-  return options.return_best_rdo_node();
-}

libde265-1.0.17.tar.gz/libde265/encoder/algo/cb-skip.h Deleted

@@ -1,72 +0,0 @@
-/*
- * H.265 video codec.
- * Copyright (c) 2013-2014 struktur AG, Dirk Farin <farin@struktur.de>
- *
- * Authors: Dirk Farin <farin@struktur.de>
- *
- * This file is part of libde265.
- *
- * libde265 is free software: you can redistribute it and/or modify
- * it under the terms of the GNU Lesser General Public License as
- * published by the Free Software Foundation, either version 3 of
- * the License, or (at your option) any later version.
- *
- * libde265 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 Lesser General Public License for more details.
- *
- * You should have received a copy of the GNU Lesser General Public License
- * along with libde265.  If not, see <http://www.gnu.org/licenses/>.
- */
-
-#ifndef CB_SKIP_H
-#define CB_SKIP_H
-
-#include "libde265/nal-parser.h"
-#include "libde265/decctx.h"
-#include "libde265/slice.h"
-#include "libde265/scan.h"
-#include "libde265/intrapred.h"
-#include "libde265/transform.h"
-#include "libde265/fallback-dct.h"
-#include "libde265/quality.h"
-#include "libde265/fallback.h"
-#include "libde265/encoder/configparam.h"
-
-#include "libde265/encoder/algo/algo.h"
-#include "libde265/encoder/algo/cb-mergeindex.h"
-
-
-// ========== CB Skip/Inter decision ==========
-
-class Algo_CB_Skip : public Algo_CB
-{
- public:
-  virtual ~Algo_CB_Skip() { }
-
-  void setSkipAlgo(Algo_CB_MergeIndex* algo) {
-    mSkipAlgo = algo;
-    mSkipAlgo->set_code_residual(false);
-  }
-
-  void setNonSkipAlgo(Algo_CB* algo) { mNonSkipAlgo = algo; }
-
-  const char* name() const { return "cb-skip"; }
-
- protected:
-  Algo_CB_MergeIndex* mSkipAlgo;
-  Algo_CB*            mNonSkipAlgo;
-};
-
-class Algo_CB_Skip_BruteForce : public Algo_CB_Skip
-{
- public:
-  virtual enc_cb* analyze(encoder_context*,
-                          context_model_table&,
-                          enc_cb* cb);
-
-  const char* name() const { return "cb-skip-bruteforce"; }
-};
-
-#endif

libde265-1.0.17.tar.gz/libde265/encoder/algo/cb-split.cc Deleted

@@ -1,178 +0,0 @@
-/*
- * H.265 video codec.
- * Copyright (c) 2013-2014 struktur AG, Dirk Farin <farin@struktur.de>
- *
- * Authors: struktur AG, Dirk Farin <farin@struktur.de>
- *
- * This file is part of libde265.
- *
- * libde265 is free software: you can redistribute it and/or modify
- * it under the terms of the GNU Lesser General Public License as
- * published by the Free Software Foundation, either version 3 of
- * the License, or (at your option) any later version.
- *
- * libde265 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 Lesser General Public License for more details.
- *
- * You should have received a copy of the GNU Lesser General Public License
- * along with libde265.  If not, see <http://www.gnu.org/licenses/>.
- */
-
-
-#include "libde265/encoder/algo/cb-split.h"
-#include "libde265/encoder/algo/coding-options.h"
-#include "libde265/encoder/encoder-context.h"
-#include "libde265/encoder/encoder-syntax.h"
-#include <assert.h>
-#include <limits>
-#include <math.h>
-#include <iostream>
-
-
-// Utility function to encode all four children in a split CB.
-// Children are coded with the specified algo_cb_split.
-enc_cb* Algo_CB_Split::encode_cb_split(encoder_context* ectx,
-                                       context_model_table& ctxModel,
-                                       enc_cb* cb)
-{
-  int w = ectx->imgdata->input->get_width();
-  int h = ectx->imgdata->input->get_height();
-
-
-  cb->split_cu_flag = true;
-
-
-  // encode all 4 children and sum their distortions and rates
-
-  for (int i=0;i<4;i++) {
-    cb->childreni = NULL;
-  }
-
-  for (int i=0;i<4;i++) {
-    int child_x = cb->x + ((i&1)  << (cb->log2Size-1));
-    int child_y = cb->y + ((i>>1) << (cb->log2Size-1));
-
-    if (child_x>=w || child_y>=h) {
-      // NOP
-    }
-    else {
-      enc_cb* childCB = new enc_cb;
-      childCB->log2Size = cb->log2Size-1;
-      childCB->ctDepth  = cb->ctDepth+1;
-
-      childCB->x = child_x;
-      childCB->y = child_y;
-      childCB->parent  = cb;
-      childCB->downPtr = &cb->childreni;
-
-      descend(cb,"yes %d/4",i+1);
-      cb->childreni = analyze(ectx, ctxModel, childCB);
-      ascend();
-
-      cb->distortion += cb->childreni->distortion;
-      cb->rate       += cb->childreni->rate;
-    }
-  }
-
-  return cb;
-}
-
-
-
-
-enc_cb* Algo_CB_Split_BruteForce::analyze(encoder_context* ectx,
-                                          context_model_table& ctxModel,
-                                          enc_cb* cb_input)
-{
-  assert(cb_input->pcm_flag==0);
-
-  // --- prepare coding options ---
-
-  const SplitType split_type = get_split_type(&ectx->get_sps(),
-                                              cb_input->x, cb_input->y,
-                                              cb_input->log2Size);
-
-
-  bool can_split_CB   = (split_type != ForcedNonSplit);
-  bool can_nosplit_CB = (split_type != ForcedSplit);
-
-  //if (can_split_CB) { can_nosplit_CB=false; } // TODO TMP
-  //if (can_nosplit_CB) { can_split_CB=false; } // TODO TMP
-
-  CodingOptions<enc_cb> options(ectx, cb_input, ctxModel);
-
-  CodingOption<enc_cb> option_no_split = options.new_option(can_nosplit_CB);
-  CodingOption<enc_cb> option_split    = options.new_option(can_split_CB);
-
-  options.start();
-
-  /*
-  cb_input->writeSurroundingMetadata(ectx, ectx->img,
-                                     enc_node::METADATA_CT_DEPTH, // for estimation cb-split bits
-                                     cb_input->get_rectangle());
-  */
-
-  // --- encode without splitting ---
-
-  if (option_no_split) {
-    CodingOption<enc_cb>& opt = option_no_split; // abbrev.
-
-    opt.begin();
-
-    enc_cb* cb = opt.get_node();
-    *cb_input->downPtr = cb;
-
-    // set CB size in image data-structure
-    //ectx->img->set_ctDepth(cb->x,cb->y,cb->log2Size, cb->ctDepth);
-    //ectx->img->set_log2CbSize(cb->x,cb->y,cb->log2Size, true);
-
-    /* We set QP here, because this is required at in non-split CBs only.
-     */
-    cb->qp = ectx->active_qp;
-
-    // analyze subtree
-    assert(mChildAlgo);
-
-    descend(cb,"no");
-    cb = mChildAlgo->analyze(ectx, opt.get_context(), cb);
-    ascend();
-
-    // add rate for split flag
-    if (split_type == OptionalSplit) {
-      encode_split_cu_flag(ectx,opt.get_cabac(), cb->x,cb->y, cb->ctDepth, 0);
-      cb->rate += opt.get_cabac_rate();
-    }
-
-    opt.set_node(cb);
-    opt.end();
-  }
-
-  // --- encode with splitting ---
-
-  if (option_split) {
-    option_split.begin();
-
-    enc_cb* cb = option_split.get_node();
-    *cb_input->downPtr = cb;
-
-    cb = encode_cb_split(ectx, option_split.get_context(), cb);
-
-    // add rate for split flag
-    if (split_type == OptionalSplit) {
-      encode_split_cu_flag(ectx,option_split.get_cabac(), cb->x,cb->y, cb->ctDepth, 1);
-      cb->rate += option_split.get_cabac_rate();
-    }
-
-    option_split.set_node(cb);
-    option_split.end();
-  }
-
-  options.compute_rdo_costs();
-  enc_cb* bestCB = options.return_best_rdo_node();
-
-  //bestCB->debug_assertTreeConsistency(ectx->img);
-
-  return bestCB;
-}

libde265-1.0.17.tar.gz/libde265/encoder/algo/cb-split.h Deleted

@@ -1,88 +0,0 @@
-/*
- * H.265 video codec.
- * Copyright (c) 2013-2014 struktur AG, Dirk Farin <farin@struktur.de>
- *
- * Authors: Dirk Farin <farin@struktur.de>
- *
- * This file is part of libde265.
- *
- * libde265 is free software: you can redistribute it and/or modify
- * it under the terms of the GNU Lesser General Public License as
- * published by the Free Software Foundation, either version 3 of
- * the License, or (at your option) any later version.
- *
- * libde265 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 Lesser General Public License for more details.
- *
- * You should have received a copy of the GNU Lesser General Public License
- * along with libde265.  If not, see <http://www.gnu.org/licenses/>.
- */
-
-#ifndef CB_SPLIT_H
-#define CB_SPLIT_H
-
-#include "libde265/nal-parser.h"
-#include "libde265/decctx.h"
-#include "libde265/slice.h"
-#include "libde265/scan.h"
-#include "libde265/intrapred.h"
-#include "libde265/transform.h"
-#include "libde265/fallback-dct.h"
-#include "libde265/quality.h"
-#include "libde265/fallback.h"
-#include "libde265/encoder/configparam.h"
-
-#include "libde265/encoder/algo/algo.h"
-#include "libde265/encoder/algo/tb-intrapredmode.h"
-#include "libde265/encoder/algo/tb-split.h"
-
-
-/*  Encoder search tree, bottom up:
-
-    - Algo_TB_Split - whether TB is split or not
-
-    - Algo_TB_IntraPredMode - choose the intra prediction mode (or NOP, if at the wrong tree level)
-
-    - Algo_CB_IntraPartMode - choose between NxN and 2Nx2N intra parts
-
-    - Algo_CB_Split - whether CB is split or not
-
-    - Algo_CTB_QScale - select QScale on CTB granularity
- */
-
-
-// ========== CB split decision ==========
-
-class Algo_CB_Split : public Algo_CB
-{
- public:
-  virtual ~Algo_CB_Split() { }
-
-  // TODO: probably, this will later be a intra/inter decision which again
-  // has two child algorithms, depending on the coding mode.
-  void setChildAlgo(Algo_CB* algo) { mChildAlgo = algo; }
-
-  const char* name() const { return "cb-split"; }
-
- protected:
-  Algo_CB* mChildAlgo;
-
-  enc_cb* encode_cb_split(encoder_context* ectx,
-                          context_model_table& ctxModel,
-                          enc_cb* cb);
-};
-
-
-class Algo_CB_Split_BruteForce : public Algo_CB_Split
-{
- public:
-  virtual enc_cb* analyze(encoder_context*,
-                          context_model_table&,
-                          enc_cb* cb);
-
-  const char* name() const { return "cb-split-bruteforce"; }
-};
-
-#endif

libde265-1.0.17.tar.gz/libde265/encoder/algo/coding-options.cc Deleted

@@ -1,202 +0,0 @@
-/*
- * H.265 video codec.
- * Copyright (c) 2013-2014 struktur AG, Dirk Farin <farin@struktur.de>
- *
- * Authors: Dirk Farin <farin@struktur.de>
- *
- * This file is part of libde265.
- *
- * libde265 is free software: you can redistribute it and/or modify
- * it under the terms of the GNU Lesser General Public License as
- * published by the Free Software Foundation, either version 3 of
- * the License, or (at your option) any later version.
- *
- * libde265 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 Lesser General Public License for more details.
- *
- * You should have received a copy of the GNU Lesser General Public License
- * along with libde265.  If not, see <http://www.gnu.org/licenses/>.
- */
-
-#include "libde265/encoder/algo/coding-options.h"
-#include "libde265/encoder/encoder-context.h"
-
-
-template <class node>
-CodingOptions<node>::CodingOptions(encoder_context* ectx, node* _node, context_model_table& tab)
-{
-  //mCBMode = true;
-  mInputNode = _node;
-  mContextModelInput = &tab;
-
-  mBestRDO=-1;
-
-  mECtx = ectx;
-}
-
-template <class node>
-CodingOptions<node>::~CodingOptions()
-{
-}
-
-template <class node>
-CodingOption<node> CodingOptions<node>::new_option(bool active)
-{
-  if (!active) {
-    return CodingOption<node>();
-  }
-
-
-  CodingOptionData opt;
-
-  bool firstOption = mOptions.empty();
-  if (firstOption) {
-    opt.mNode = mInputNode;
-  }
-  else {
-    opt.mNode = new node(*mInputNode);
-  }
-
-  opt.context  = *mContextModelInput;
-  opt.computed = false;
-
-  CodingOption<node> option(this, mOptions.size());
-
-  mOptions.push_back( std::move(opt) );
-
-  return option;
-}
-
-
-template <class node>
-void CodingOptions<node>::start(enum RateEstimationMethod rateMethod)
-{
-  /* We don't need the input context model anymore.
-     Releasing it now may save a copy during a later decouple().
-  */
-  mContextModelInput->release();
-
-  bool adaptiveContext;
-  switch (rateMethod) {
-  case Rate_Default:
-    adaptiveContext = mECtx->use_adaptive_context;
-    break;
-  case Rate_FixedContext:
-    adaptiveContext = false;
-    break;
-  case Rate_AdaptiveContext:
-    adaptiveContext = true;
-    break;
-  }
-
-  if (adaptiveContext) {
-    /* If we modify the context models in this algorithm,
-       we need separate models for each option.
-    */
-    for (auto& option : mOptions) {
-      option.context.decouple();
-    }
-
-    cabac = &cabac_adaptive;
-  }
-  else {
-    cabac = &cabac_constant;
-  }
-}
-
-
-template <class node>
-void CodingOptions<node>::compute_rdo_costs()
-{
-  for (size_t i=0;i<mOptions.size();i++) {
-    if (mOptionsi.computed) {
-      //printf("compute_rdo_costs %d: %f\n",i, mOptionsi.mNode->rate);
-      mOptionsi.rdoCost = mOptionsi.mNode->distortion + mECtx->lambda * mOptionsi.mNode->rate;
-    }
-  }
-}
-
-
-template <class node>
-int CodingOptions<node>::find_best_rdo_index()
-{
-  assert(mOptions.size()>0);
-
-
-  float bestRDOCost = 0;
-  bool  first=true;
-  int   bestRDO=-1;
-
-  for (size_t i=0;i<mOptions.size();i++) {
-    if (mOptionsi.computed) {
-      float cost = mOptionsi.rdoCost;
-
-      //printf("option %d cost: %f\n",i,cost);
-
-      if (first || cost < bestRDOCost) {
-        bestRDOCost = cost;
-        first = false;
-        bestRDO = i;
-      }
-    }
-  }
-
-  return bestRDO;
-}
-
-
-template <class node>
-node* CodingOptions<node>::return_best_rdo_node()
-{
-  int bestRDO = find_best_rdo_index();
-
-  assert(bestRDO>=0);
-
-  *mContextModelInput = mOptionsbestRDO.context;
-
-
-  // delete all CBs except the best one
-
-  for (size_t i=0;i<mOptions.size();i++) {
-    if (i != bestRDO)
-      {
-        delete mOptionsi.mNode;
-        mOptionsi.mNode = NULL;
-      }
-  }
-
-  return mOptionsbestRDO.mNode;
-}
-
-
-template <class node>
-void CodingOption<node>::begin()
-{
-  assert(mParent);
-  assert(mParent->cabac); // did you call CodingOptions.start() ?
-
-  mParent->cabac->reset();
-  mParent->cabac->set_context_models( &get_context() );
-
-  mParent->mOptionsmOptionIdx.computed = true;
-
-  // link this node into the coding tree
-
-  node* n = get_node();
-  *(n->downPtr) = n;
-}
-
-
-template <class node>
-void CodingOption<node>::end()
-{
-}
-
-
-template class CodingOptions<enc_tb>;
-template class CodingOptions<enc_cb>;
-
-template class CodingOption<enc_tb>;
-template class CodingOption<enc_cb>;

libde265-1.0.17.tar.gz/libde265/encoder/algo/coding-options.h Deleted

@@ -1,151 +0,0 @@
-/*
- * H.265 video codec.
- * Copyright (c) 2013-2014 struktur AG, Dirk Farin <farin@struktur.de>
- *
- * Authors: Dirk Farin <farin@struktur.de>
- *
- * This file is part of libde265.
- *
- * libde265 is free software: you can redistribute it and/or modify
- * it under the terms of the GNU Lesser General Public License as
- * published by the Free Software Foundation, either version 3 of
- * the License, or (at your option) any later version.
- *
- * libde265 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 Lesser General Public License for more details.
- *
- * You should have received a copy of the GNU Lesser General Public License
- * along with libde265.  If not, see <http://www.gnu.org/licenses/>.
- */
-
-#ifndef CODING_OPTIONS_H
-#define CODING_OPTIONS_H
-
-#include "libde265/encoder/encoder-types.h"
-
-
-template <class node> class CodingOption;
-
-
-template <class node>
-class CodingOptions
-{
- public:
-  CodingOptions(encoder_context*, node*, context_model_table& tab);
-  ~CodingOptions();
-
-  typedef CodingOption<node> Option;
-
-  // --- init --- call before object use
-
-  CodingOption<node> new_option(bool active=true);
-
-  enum RateEstimationMethod
-  {
-    Rate_Default,  // take default value from encoder_context
-    Rate_AdaptiveContext,
-    Rate_FixedContext
-  };
-
-  void start(enum RateEstimationMethod = Rate_Default);
-
-
-  // --- processing ---
-
-  // compute RDO cost (D + lambda*R) for all options
-  void compute_rdo_costs();
-
-
-  // --- end processing --- do not call any function after this one
-
-  /* Return the CB with the lowest RDO cost. All other CBs are destroyed.
-     If the current metadata stored in the image are not from the returned block,
-     its metadata flags are set to zero.
-   */
-  node* return_best_rdo_node();
-
- private:
-  struct CodingOptionData
-  {
-    node* mNode;
-
-    context_model_table context;
-    bool  mOptionActive;
-    bool  computed;
-    float rdoCost;
-  };
-
-
-  encoder_context* mECtx;
-
-  bool mCBMode;
-  node* mInputNode;
-
-  context_model_table* mContextModelInput;
-
-  int mBestRDO;
-
-  std::vector<CodingOptionData> mOptions;
-
-  CABAC_encoder_estim           cabac_adaptive;
-  CABAC_encoder_estim_constant  cabac_constant;
-  CABAC_encoder_estim*          cabac;
-
-  friend class CodingOption<node>;
-
-  int find_best_rdo_index();
-};
-
-
-template <class node>
-class CodingOption
-{
- public:
-  CodingOption() {
-    mParent = NULL;
-    mOptionIdx = 0;
-  }
-
-  node* get_node() { return mParent->mOptionsmOptionIdx.mNode; }
-  void set_node(node* _node) {
-    if (_node != mParent->mOptionsmOptionIdx.mNode) {
-      //printf("delete TB %p\n", mParent->mOptionsmOptionIdx.tb);
-      //delete mParent->mOptionsmOptionIdx.mNode;
-    }
-    mParent->mOptionsmOptionIdx.mNode = _node;
-  }
-
-  context_model_table& get_context() { return mParent->mOptionsmOptionIdx.context; }
-
-  /** @return True if the option is active.
-   */
-  operator bool() const { return mParent; }
-
-  /* When modifying the metadata stored in the image, you have to
-     encapsulate the modification between these two functions to ensure
-     that the correct reconstruction will be active after return_best_rdo().
-   */
-  void begin();
-  void end();
-
-  // Manually set RDO costs instead of computing them with compute_rdo_costs.
-  // Only required when using custom costs.
-  void set_rdo_cost(float rdo) { mParent->mOptionsmOptionIdx.rdoCost=rdo; }
-
-  CABAC_encoder_estim* get_cabac() { return mParent->cabac; }
-  float get_cabac_rate() const { return mParent->cabac->getRDBits(); }
-
-private:
-  CodingOption(class CodingOptions<node>* parent, int idx)
-    : mParent(parent), mOptionIdx(idx) { }
-
-  class CodingOptions<node>* mParent;
-  int   mOptionIdx;
-
-  friend class CodingOptions<node>;
-};
-
-
-#endif

libde265-1.0.17.tar.gz/libde265/encoder/algo/ctb-qscale.cc Deleted

@@ -1,61 +0,0 @@
-/*
- * H.265 video codec.
- * Copyright (c) 2013-2014 struktur AG, Dirk Farin <farin@struktur.de>
- *
- * Authors: struktur AG, Dirk Farin <farin@struktur.de>
- *
- * This file is part of libde265.
- *
- * libde265 is free software: you can redistribute it and/or modify
- * it under the terms of the GNU Lesser General Public License as
- * published by the Free Software Foundation, either version 3 of
- * the License, or (at your option) any later version.
- *
- * libde265 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 Lesser General Public License for more details.
- *
- * You should have received a copy of the GNU Lesser General Public License
- * along with libde265.  If not, see <http://www.gnu.org/licenses/>.
- */
-
-
-#include "libde265/encoder/algo/ctb-qscale.h"
-#include "libde265/encoder/encoder-context.h"
-#include <assert.h>
-#include <limits>
-#include <math.h>
-
-
-#define ENCODER_DEVELOPMENT 1
-
-
-enc_cb* Algo_CTB_QScale_Constant::analyze(encoder_context* ectx,
-                                          context_model_table& ctxModel,
-                                          int x,int y)
-{
-  enc_cb* cb = new enc_cb();
-
-  cb->log2Size = ectx->get_sps().Log2CtbSizeY;
-  cb->ctDepth = 0;
-  cb->x = x;
-  cb->y = y;
-  cb->downPtr = ectx->ctbs.getCTBRootPointer(x,y);
-  *cb->downPtr = cb;
-
-  cb->qp = ectx->active_qp;
-
-  // write currently unused coding options
-  cb->cu_transquant_bypass_flag = false;
-  cb->pcm_flag = false;
-
-  assert(mChildAlgo);
-  descend(cb, "Q=%d",ectx->active_qp);
-  enc_cb* result_cb = mChildAlgo->analyze(ectx,ctxModel,cb);
-  ascend();
-
-  *cb->downPtr = result_cb;
-
-  return result_cb;
-}

libde265-1.0.17.tar.gz/libde265/encoder/algo/ctb-qscale.h Deleted

@@ -1,109 +0,0 @@
-/*
- * H.265 video codec.
- * Copyright (c) 2013-2014 struktur AG, Dirk Farin <farin@struktur.de>
- *
- * Authors: Dirk Farin <farin@struktur.de>
- *
- * This file is part of libde265.
- *
- * libde265 is free software: you can redistribute it and/or modify
- * it under the terms of the GNU Lesser General Public License as
- * published by the Free Software Foundation, either version 3 of
- * the License, or (at your option) any later version.
- *
- * libde265 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 Lesser General Public License for more details.
- *
- * You should have received a copy of the GNU Lesser General Public License
- * along with libde265.  If not, see <http://www.gnu.org/licenses/>.
- */
-
-#ifndef CTB_QSCALE_H
-#define CTB_QSCALE_H
-
-#include "libde265/nal-parser.h"
-#include "libde265/decctx.h"
-#include "libde265/encoder/algo/algo.h"
-#include "libde265/slice.h"
-#include "libde265/scan.h"
-#include "libde265/intrapred.h"
-#include "libde265/transform.h"
-#include "libde265/fallback-dct.h"
-#include "libde265/quality.h"
-#include "libde265/fallback.h"
-#include "libde265/encoder/configparam.h"
-
-#include "libde265/encoder/algo/cb-split.h"
-
-
-/*  Encoder search tree, bottom up:
-
-    - Algo_TB_Split - whether TB is split or not
-
-    - Algo_TB_IntraPredMode - choose the intra prediction mode (or NOP, if at the wrong tree level)
-
-    - Algo_CB_IntraPartMode - choose between NxN and 2Nx2N intra parts
-
-    - Algo_CB_Split - whether CB is split or not
-
-    - Algo_CTB_QScale - select QScale on CTB granularity
- */
-
-
-// ========== choose a qscale at CTB level ==========
-
-class Algo_CTB_QScale : public Algo
-{
- public:
- Algo_CTB_QScale() : mChildAlgo(NULL) { }
-  virtual ~Algo_CTB_QScale() { }
-
-  virtual enc_cb* analyze(encoder_context*,
-                          context_model_table&,
-                          int ctb_x,int ctb_y) = 0;
-
-  void setChildAlgo(Algo_CB_Split* algo) { mChildAlgo = algo; }
-
- protected:
-  Algo_CB_Split* mChildAlgo;
-};
-
-
-
-class Algo_CTB_QScale_Constant : public Algo_CTB_QScale
-{
- public:
-  struct params
-  {
-    params() {
-      mQP.set_range(1,51);
-      mQP.set_default(27);
-      mQP.set_ID("CTB-QScale-Constant");
-      mQP.set_cmd_line_options("qp",'q');
-    }
-
-    option_int mQP;
-  };
-
-  void setParams(const params& p) { mParams=p; }
-
-  void registerParams(config_parameters& config) {
-    config.add_option(&mParams.mQP);
-  }
-
-  virtual enc_cb* analyze(encoder_context*,
-                          context_model_table&,
-                          int ctb_x,int ctb_y);
-
-  int getQP() const { return mParams.mQP; }
-
-  const char* name() const { return "ctb-qscale-constant"; }
-
- private:
-  params mParams;
-};
-
-
-#endif

libde265-1.0.17.tar.gz/libde265/encoder/algo/pb-mv.cc Deleted

@@ -1,318 +0,0 @@
-/*
- * H.265 video codec.
- * Copyright (c) 2013-2014 struktur AG, Dirk Farin <farin@struktur.de>
- *
- * Authors: struktur AG, Dirk Farin <farin@struktur.de>
- *
- * This file is part of libde265.
- *
- * libde265 is free software: you can redistribute it and/or modify
- * it under the terms of the GNU Lesser General Public License as
- * published by the Free Software Foundation, either version 3 of
- * the License, or (at your option) any later version.
- *
- * libde265 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 Lesser General Public License for more details.
- *
- * You should have received a copy of the GNU Lesser General Public License
- * along with libde265.  If not, see <http://www.gnu.org/licenses/>.
- */
-
-
-#include "libde265/encoder/algo/pb-mv.h"
-#include "libde265/encoder/algo/coding-options.h"
-#include "libde265/encoder/encoder-context.h"
-#include <assert.h>
-#include <limits>
-#include <math.h>
-
-
-
-enc_cb* Algo_PB_MV_Test::analyze(encoder_context* ectx,
-                                 context_model_table& ctxModel,
-                                 enc_cb* cb,
-                                 int PBidx, int x,int y,int w,int h)
-{
-  enum MVTestMode testMode = mParams.testMode();
-
-
-  MotionVector mvp2;
-
-  fill_luma_motion_vector_predictors(ectx, ectx->shdr, ectx->img,
-                                     cb->x,cb->y,1<<cb->log2Size, x,y,w,h,
-                                     0, // l
-                                     0, 0, // int refIdx, int partIdx,
-                                     mvp);
-
-  //printf("%d/%d: %d;%d %d;%d\n",cb->x,cb->y, mvp0.x,mvp0.y, mvp1.x,mvp1.y);
-
-
-  PBMotionCoding& spec = cb->inter.pbPBidx.spec;
-  PBMotion&        vec = cb->inter.pbPBidx.motion;
-
-  spec.merge_flag = 0;
-  spec.merge_idx  = 0;
-
-  spec.inter_pred_idc = PRED_L0;
-  spec.refIdx0 = vec.refIdx0 = 0;
-  spec.mvp_l0_flag = 0;
-
-  int value = mParams.range();
-
-  switch (testMode) {
-  case MVTestMode_Zero:
-    spec.mvd00=0;
-    spec.mvd01=0;
-    break;
-
-  case MVTestMode_Random:
-    spec.mvd00 = (rand() % (2*value+1)) - value;
-    spec.mvd01 = (rand() % (2*value+1)) - value;
-    break;
-
-  case MVTestMode_Horizontal:
-    spec.mvd00=value;
-    spec.mvd01=0;
-    break;
-
-  case MVTestMode_Vertical:
-    spec.mvd00=0;
-    spec.mvd01=value;
-    break;
-  }
-
-  spec.mvd00 -= mvp0.x;
-  spec.mvd01 -= mvp0.y;
-
-  vec.mv0.x = mvp0.x + spec.mvd00;
-  vec.mv0.y = mvp0.y + spec.mvd01;
-  vec.predFlag0 = 1;
-  vec.predFlag1 = 0;
-
-  ectx->img->set_mv_info(x,y,w,h, vec);
-
-  /* TMP REMOVE: ectx->prediction does not exist anymore
-  generate_inter_prediction_samples(ectx, ectx->shdr, ectx->prediction,
-                                    cb->x,cb->y, // int xC,int yC,
-                                    0,0,         // int xB,int yB,
-                                    1<<cb->log2Size, // int nCS,
-                                    1<<cb->log2Size,
-                                    1<<cb->log2Size, // int nPbW,int nPbH,
-                                    &vec);
-  */
-
-  // TODO estimate rate for sending MV
-
-  int IntraSplitFlag = 0;
-  int MaxTrafoDepth = ectx->get_sps().max_transform_hierarchy_depth_inter;
-
-  mCodeResidual=true;
-  if (mCodeResidual) {
-    assert(mTBSplitAlgo);
-    assert(false);
-    /*
-    cb->transform_tree = mTBSplitAlgo->analyze(ectx,ctxModel, ectx->imgdata->input, NULL, cb,
-                                               cb->x,cb->y,cb->x,cb->y, cb->log2Size,0,
-                                               0, MaxTrafoDepth, IntraSplitFlag);
-    */
-
-    cb->inter.rqt_root_cbf = ! cb->transform_tree->isZeroBlock();
-
-    cb->distortion = cb->transform_tree->distortion;
-    cb->rate       = cb->transform_tree->rate;
-  }
-  else {
-    const de265_image* input = ectx->imgdata->input;
-    /* TODO TMP REMOVE: prediction does not exist anymore
-    de265_image* img   = ectx->prediction;
-    int x0 = cb->x;
-    int y0 = cb->y;
-    int tbSize = 1<<cb->log2Size;
-
-    cb->distortion = compute_distortion_ssd(input, img, x0,y0, cb->log2Size, 0);
-    cb->rate = 5; // fake (MV)
-
-    cb->inter.rqt_root_cbf = 0;
-    */
-  }
-
-  return cb;
-}
-
-
-
-
-int sad(const uint8_t* p1,int stride1,
-        const uint8_t* p2,int stride2,
-        int w,int h)
-{
-  int cost=0;
-
-  for (int y=0;y<h;y++) {
-    for (int x=0;x<w;x++) {
-      cost += abs_value(*p1 - *p2);
-      p1++;
-      p2++;
-    }
-
-    p1 += stride1-w;
-    p2 += stride2-w;
-  }
-
-  return cost;
-}
-
-
-enc_cb* Algo_PB_MV_Search::analyze(encoder_context* ectx,
-                                   context_model_table& ctxModel,
-                                   enc_cb* cb,
-                                   int PBidx, int x,int y,int pbW,int pbH)
-{
-  enum MVSearchAlgo searchAlgo = mParams.mvSearchAlgo();
-
-
-  MotionVector mvp2;
-
-  fill_luma_motion_vector_predictors(ectx, ectx->shdr, ectx->img,
-                                     cb->x,cb->y,1<<cb->log2Size, x,y,pbW,pbH,
-                                     0, // l
-                                     0, 0, // int refIdx, int partIdx,
-                                     mvp);
-
-  PBMotionCoding& spec = cb->inter.pbPBidx.spec;
-  PBMotion&        vec = cb->inter.pbPBidx.motion;
-
-  spec.merge_flag = 0;
-  spec.merge_idx  = 0;
-
-  spec.inter_pred_idc = PRED_L0;
-  spec.refIdx0 = vec.refIdx0 = 0;
-  spec.mvp_l0_flag = 0;
-
-  int hrange = mParams.hrange();
-  int vrange = mParams.vrange();
-
-  // previous frame (TODO)
-  const de265_image* refimg   = ectx->get_image(ectx->imgdata->frame_number -1);
-  const de265_image* inputimg = ectx->imgdata->input;
-
-  int w = refimg->get_width();
-  int h = refimg->get_height();
-
-  int mincost = 0x7fffffff;
-
-  double lambda = 10.0;
-
-  double *bits_h = new double2*hrange+1;
-  double *bits_v = new double2*vrange+1;
-
-  for (int i=-hrange;i<=hrange;i++) {
-    int diff = (i - mvp0.x);
-    int b;
-
-    if (diff==0) { b=0; }
-    else if (diff==1 || diff==-1) { b=2; }
-    else { b=abs_value(b+2); }
-
-    bits_hi+hrange=b;
-  }
-
-  for (int i=-vrange;i<=vrange;i++) {
-    int diff = (i - mvp0.y);
-    int b;
-
-    if (diff==0) { b=0; }
-    else if (diff==1 || diff==-1) { b=2; }
-    else { b=abs_value(b+2); }
-
-    bits_vi+vrange=b;
-  }
-
-  for (int my = y-vrange; my<=y+vrange; my++)
-    for (int mx = x-hrange; mx<=x+hrange; mx++)
-      {
-        if (mx<0 || mx+pbW>w || my<0 || my+pbH>h) continue;
-
-        int cost = sad(refimg->get_image_plane_at_pos(0,mx,my),
-                       refimg->get_image_stride(0),
-                       inputimg->get_image_plane_at_pos(0,x,y),
-                       inputimg->get_image_stride(0),
-                       pbW,pbH);
-
-        int bits = bits_hmx-x+hrange + bits_vmy-y+vrange;
-
-        cost += lambda * bits;
-
-        //printf("%d %d : %d\n",mx,my,cost);
-
-        if (cost<mincost) {
-          mincost=cost;
-
-          spec.mvd00=(mx-x)<<2;
-          spec.mvd01=(my-y)<<2;
-        }
-      }
-
-  spec.mvd00 -= mvp0.x;
-  spec.mvd01 -= mvp0.y;
-
-  vec.mv0.x = mvp0.x + spec.mvd00;
-  vec.mv0.y = mvp0.y + spec.mvd01;
-  vec.predFlag0 = 1;
-  vec.predFlag1 = 0;
-
-  ectx->img->set_mv_info(x,y,pbW,pbH, vec);
-
-  /* TMP REMOVE: ectx->prediction does not exist anymore
-  generate_inter_prediction_samples(ectx, ectx->shdr, ectx->prediction,
-                                    cb->x,cb->y, // int xC,int yC,
-                                    0,0,         // int xB,int yB,
-                                    1<<cb->log2Size, // int nCS,
-                                    1<<cb->log2Size,
-                                    1<<cb->log2Size, // int nPbW,int nPbH,
-                                    &vec);
-  */
-
-  // --- create residual ---
-
-
-
-  // TODO estimate rate for sending MV
-
-  int IntraSplitFlag = 0;
-  int MaxTrafoDepth = ectx->get_sps().max_transform_hierarchy_depth_inter;
-
-  mCodeResidual=true;
-  if (mCodeResidual) {
-    assert(false);
-    /*
-    cb->transform_tree = mTBSplitAlgo->analyze(ectx,ctxModel, ectx->imgdata->input, NULL, cb,
-                                               cb->x,cb->y,cb->x,cb->y, cb->log2Size,0,
-                                               0, MaxTrafoDepth, IntraSplitFlag);
-    */
-
-    cb->inter.rqt_root_cbf = ! cb->transform_tree->isZeroBlock();
-
-    cb->distortion = cb->transform_tree->distortion;
-    cb->rate       = cb->transform_tree->rate;
-  }
-  else {
-    const de265_image* input = ectx->imgdata->input;
-    de265_image* img   = ectx->img;
-    int x0 = cb->x;
-    int y0 = cb->y;
-    int tbSize = 1<<cb->log2Size;
-
-    cb->distortion = compute_distortion_ssd(input, img, x0,y0, cb->log2Size, 0);
-    cb->rate = 5; // fake (MV)
-
-    cb->inter.rqt_root_cbf = 0;
-  }
-
-  delete bits_h;
-  delete bits_v;
-
-  return cb;
-}

libde265-1.0.17.tar.gz/libde265/encoder/algo/pb-mv.h Deleted

@@ -1,177 +0,0 @@
-/*
- * H.265 video codec.
- * Copyright (c) 2013-2014 struktur AG, Dirk Farin <farin@struktur.de>
- *
- * Authors: Dirk Farin <farin@struktur.de>
- *
- * This file is part of libde265.
- *
- * libde265 is free software: you can redistribute it and/or modify
- * it under the terms of the GNU Lesser General Public License as
- * published by the Free Software Foundation, either version 3 of
- * the License, or (at your option) any later version.
- *
- * libde265 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 Lesser General Public License for more details.
- *
- * You should have received a copy of the GNU Lesser General Public License
- * along with libde265.  If not, see <http://www.gnu.org/licenses/>.
- */
-
-#ifndef PB_MV_H
-#define PB_MV_H
-
-#include "libde265/nal-parser.h"
-#include "libde265/decctx.h"
-#include "libde265/slice.h"
-#include "libde265/scan.h"
-#include "libde265/intrapred.h"
-#include "libde265/transform.h"
-#include "libde265/fallback-dct.h"
-#include "libde265/quality.h"
-#include "libde265/fallback.h"
-#include "libde265/encoder/configparam.h"
-
-#include "libde265/encoder/algo/algo.h"
-
-
-// ========== CB Intra/Inter decision ==========
-
-class Algo_TB_Split;
-
-
-class Algo_PB_MV : public Algo_PB
-{
- public:
- Algo_PB_MV() : mTBSplitAlgo(NULL) { }
-  virtual ~Algo_PB_MV() { }
-
-  void setChildAlgo(Algo_TB_Split* algo) { mTBSplitAlgo = algo; }
-
- protected:
-  Algo_TB_Split* mTBSplitAlgo;
-};
-
-
-
-
-enum MVTestMode
-  {
-    MVTestMode_Zero,
-    MVTestMode_Random,
-    MVTestMode_Horizontal,
-    MVTestMode_Vertical
-  };
-
-class option_MVTestMode : public choice_option<enum MVTestMode>
-{
- public:
-  option_MVTestMode() {
-    add_choice("zero",   MVTestMode_Zero);
-    add_choice("random", MVTestMode_Random);
-    add_choice("horiz",  MVTestMode_Horizontal, true);
-    add_choice("verti",  MVTestMode_Vertical);
-  }
-};
-
-
-class Algo_PB_MV_Test : public Algo_PB_MV
-{
- public:
- Algo_PB_MV_Test() : mCodeResidual(false) { }
-
-  struct params
-  {
-    params() {
-      testMode.set_ID("PB-MV-TestMode");
-      range.set_ID   ("PB-MV-Range");
-      range.set_default(4);
-    }
-
-    option_MVTestMode testMode;
-    option_int        range;
-  };
-
-  void registerParams(config_parameters& config) {
-    config.add_option(&mParams.testMode);
-    config.add_option(&mParams.range);
-  }
-
-  void setParams(const params& p) { mParams=p; }
-
-  virtual enc_cb* analyze(encoder_context*,
-                          context_model_table&,
-                          enc_cb* cb,
-                          int PBidx, int x,int y,int w,int h);
-
- private:
-  params mParams;
-
-  bool mCodeResidual;
-};
-
-
-
-
-enum MVSearchAlgo
-  {
-    MVSearchAlgo_Zero,
-    MVSearchAlgo_Full,
-    MVSearchAlgo_Diamond,
-    MVSearchAlgo_PMVFast
-  };
-
-class option_MVSearchAlgo : public choice_option<enum MVSearchAlgo>
-{
- public:
-  option_MVSearchAlgo() {
-    add_choice("zero",   MVSearchAlgo_Zero);
-    add_choice("full",   MVSearchAlgo_Full, true);
-    add_choice("diamond",MVSearchAlgo_Diamond);
-    add_choice("pmvfast",MVSearchAlgo_PMVFast);
-  }
-};
-
-
-class Algo_PB_MV_Search : public Algo_PB_MV
-{
- public:
- Algo_PB_MV_Search() : mCodeResidual(false) { }
-
-  struct params
-  {
-    params() {
-      mvSearchAlgo.set_ID("PB-MV-Search-Algo");
-      hrange.set_ID      ("PB-MV-Search-HRange");
-      vrange.set_ID      ("PB-MV-Search-VRange");
-      hrange.set_default(8);
-      vrange.set_default(8);
-    }
-
-    option_MVSearchAlgo mvSearchAlgo;
-    option_int        hrange;
-    option_int        vrange;
-  };
-
-  void registerParams(config_parameters& config) {
-    config.add_option(&mParams.mvSearchAlgo);
-    config.add_option(&mParams.hrange);
-    config.add_option(&mParams.vrange);
-  }
-
-  void setParams(const params& p) { mParams=p; }
-
-  virtual enc_cb* analyze(encoder_context*,
-                          context_model_table&,
-                          enc_cb* cb,
-                          int PBidx, int x,int y,int w,int h);
-
- private:
-  params mParams;
-
-  bool mCodeResidual;
-};
-
-#endif

libde265-1.0.17.tar.gz/libde265/encoder/algo/tb-intrapredmode.cc Deleted

@@ -1,532 +0,0 @@
-/*
- * H.265 video codec.
- * Copyright (c) 2013-2014 struktur AG, Dirk Farin <farin@struktur.de>
- *
- * Authors: struktur AG, Dirk Farin <farin@struktur.de>
- *
- * This file is part of libde265.
- *
- * libde265 is free software: you can redistribute it and/or modify
- * it under the terms of the GNU Lesser General Public License as
- * published by the Free Software Foundation, either version 3 of
- * the License, or (at your option) any later version.
- *
- * libde265 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 Lesser General Public License for more details.
- *
- * You should have received a copy of the GNU Lesser General Public License
- * along with libde265.  If not, see <http://www.gnu.org/licenses/>.
- */
-
-
-#include "libde265/encoder/encoder-context.h"
-#include "libde265/encoder/algo/tb-split.h"
-#include "libde265/encoder/algo/coding-options.h"
-#include "libde265/encoder/encoder-intrapred.h"
-#include <assert.h>
-#include <limits>
-#include <math.h>
-#include <algorithm>
-#include <iostream>
-
-
-float get_intra_pred_mode_bits(const enum IntraPredMode candidates3,
-                               enum IntraPredMode intraMode,
-                               enum IntraPredMode intraModeC,
-                               context_model_table& context_models,
-                               bool includeChroma)
-{
-  float rate;
-  int enc_bin;
-
-  /**/ if (candidates0==intraMode) { rate = 1; enc_bin=1; }
-  else if (candidates1==intraMode) { rate = 2; enc_bin=1; }
-  else if (candidates2==intraMode) { rate = 2; enc_bin=1; }
-  else { rate = 5; enc_bin=0; }
-
-  CABAC_encoder_estim estim;
-  estim.set_context_models(&context_models);
-  logtrace(LogSymbols,"$1 prev_intra_luma_pred_flag=%d\n",enc_bin);
-  estim.write_CABAC_bit(CONTEXT_MODEL_PREV_INTRA_LUMA_PRED_FLAG, enc_bin);
-
-  // TODO: currently we make the chroma-pred-mode decision for each part even
-  // in NxN part mode. Since we always set this to the same value, it does not
-  // matter. However, we should only add the rate for it once (for blkIdx=0).
-
-  if (includeChroma) {
-    assert(intraMode == intraModeC);
-
-    logtrace(LogSymbols,"$1 intra_chroma_pred_mode=%d\n",0);
-    estim.write_CABAC_bit(CONTEXT_MODEL_INTRA_CHROMA_PRED_MODE,0);
-  }
-  rate += estim.getRDBits();
-
-  return rate;
-}
-
-
-
-
-float estim_TB_bitrate(const encoder_context* ectx,
-                       const de265_image* input,
-                       const enc_tb* tb,
-                       enum TBBitrateEstimMethod method)
-{
-  int x0 = tb->x;
-  int y0 = tb->y;
-  int blkSize = 1 << tb->log2Size;
-
-  float distortion;
-
-  switch (method)
-    {
-    case TBBitrateEstim_SSD:
-      return SSD(input->get_image_plane_at_pos(0, x0,y0),
-                 input->get_image_stride(0),
-                 tb->intra_prediction0->get_buffer_u8(),
-                 tb->intra_prediction0->getStride(),
-                 blkSize, blkSize);
-      break;
-
-    case TBBitrateEstim_SAD:
-      return SAD(input->get_image_plane_at_pos(0, x0,y0),
-                 input->get_image_stride(0),
-                 tb->intra_prediction0->get_buffer_u8(),
-                 tb->intra_prediction0->getStride(),
-                 blkSize, blkSize);
-      break;
-
-    case TBBitrateEstim_SATD_DCT:
-    case TBBitrateEstim_SATD_Hadamard:
-      {
-        int16_t coeffs64*64;
-        int16_t diff64*64;
-
-        // Usually, TBs are max. 32x32 big. However, it may be that this is still called
-        // for 64x64 blocks, because we are sometimes computing an intra pred mode for a
-        // whole CTB at once.
-        assert(blkSize <= 64);
-
-        diff_blk(diff,blkSize,
-                 input->get_image_plane_at_pos(0, x0,y0), input->get_image_stride(0),
-                 tb->intra_prediction0->get_buffer_u8(),
-                 tb->intra_prediction0->getStride(),
-                 blkSize);
-
-        void (*transform)(int16_t *coeffs, const int16_t *src, ptrdiff_t stride);
-
-
-        if (tb->log2Size == 6) {
-          // hack for 64x64 blocks: compute 4 times 32x32 blocks
-
-          if (method == TBBitrateEstim_SATD_Hadamard) {
-            transform = ectx->acceleration.hadamard_transform_86-1-2;
-
-            transform(coeffs,         &diff0       , 64);
-            transform(coeffs+1*32*32, &diff32      , 64);
-            transform(coeffs+2*32*32, &diff32*64   , 64);
-            transform(coeffs+3*32*32, &diff32*64+32, 64);
-          }
-          else {
-            transform = ectx->acceleration.fwd_transform_86-1-2;
-
-            transform(coeffs,         &diff0       , 64);
-            transform(coeffs+1*32*32, &diff32      , 64);
-            transform(coeffs+2*32*32, &diff32*64   , 64);
-            transform(coeffs+3*32*32, &diff32*64+32, 64);
-          }
-        }
-        else {
-          assert(tb->log2Size-2 <= 3);
-
-          if (method == TBBitrateEstim_SATD_Hadamard) {
-            ectx->acceleration.hadamard_transform_8tb->log2Size-2(coeffs, diff, &diffblkSize - &diff0);
-          }
-          else {
-            ectx->acceleration.fwd_transform_8tb->log2Size-2(coeffs, diff, &diffblkSize - &diff0);
-          }
-        }
-
-        float distortion=0;
-        for (int i=0;i<blkSize*blkSize;i++) {
-          distortion += abs_value((int)coeffsi);
-        }
-
-        return distortion;
-      }
-      break;
-
-      /*
-    case TBBitrateEstim_AccurateBits:
-      assert(false);
-      return 0;
-      */
-    }
-
-  assert(false);
-  return 0;
-}
-
-
-
-enc_tb*
-Algo_TB_IntraPredMode_BruteForce::analyze(encoder_context* ectx,
-                                          context_model_table& ctxModel,
-                                          const de265_image* input,
-                                          enc_tb* tb,
-                                          int TrafoDepth, int MaxTrafoDepth,
-                                          int IntraSplitFlag)
-{
-  enter();
-
-  enc_cb* cb = tb->cb;
-
-  bool selectIntraPredMode = false;
-  selectIntraPredMode |= (cb->PredMode==MODE_INTRA && cb->PartMode==PART_2Nx2N && TrafoDepth==0);
-  selectIntraPredMode |= (cb->PredMode==MODE_INTRA && cb->PartMode==PART_NxN   && TrafoDepth==1);
-
-  if (selectIntraPredMode) {
-
-    CodingOptions<enc_tb> options(ectx, tb, ctxModel);
-    CodingOption<enc_tb>  option35;
-
-    for (int i=0;i<35;i++) {
-      bool computeIntraMode = isPredModeEnabled((enum IntraPredMode)i);
-      optioni = options.new_option(computeIntraMode);
-    }
-
-    options.start();
-
-
-    const seq_parameter_set* sps = &ectx->get_sps();
-    enum IntraPredMode candidates3;
-    fillIntraPredModeCandidates(candidates, tb->x,tb->y,
-                                tb->x > 0, tb->y > 0, ectx->ctbs, &ectx->get_sps());
-
-
-    for (int i = 0; i<35; i++) {
-      if (!optioni) {
-        continue;
-      }
-
-
-      enum IntraPredMode intraMode = (IntraPredMode)i;
-
-
-      optioni.begin();
-
-      enc_tb* tb_option = optioni.get_node();
-
-      *tb_option->downPtr = tb_option;
-
-      tb_option->intra_mode        = intraMode;
-
-      // set chroma mode to same mode is its luma mode
-      enum IntraPredMode intraModeC;
-
-      if (cb->PartMode==PART_2Nx2N || ectx->get_sps().ChromaArrayType==CHROMA_444) {
-        intraModeC = intraMode;
-      }
-      else {
-        intraModeC = tb_option->parent->children0->intra_mode;
-      }
-
-      tb_option->intra_mode_chroma = intraModeC;
-
-
-      descend(tb_option,"%d",intraMode);
-      tb_option = mTBSplitAlgo->analyze(ectx,optioni.get_context(),input,tb_option,
-                                        TrafoDepth, MaxTrafoDepth, IntraSplitFlag);
-      optioni.set_node(tb_option);
-      ascend();
-
-      float intraPredModeBits = get_intra_pred_mode_bits(candidates,
-                                                         intraMode,
-                                                         intraModeC,
-                                                         optioni.get_context(),
-                                                         tb_option->blkIdx == 0);
-
-      tb_option->rate_withoutCbfChroma += intraPredModeBits;
-      tb_option->rate += intraPredModeBits;
-
-      optioni.end();
-    }
-
-
-    options.compute_rdo_costs();
-
-    enc_tb* bestTB = options.return_best_rdo_node();
-
-    return bestTB;
-  }
-  else {
-    descend(tb,"NOP"); // TODO: not parent
-    enc_tb* new_tb = mTBSplitAlgo->analyze(ectx, ctxModel, input, tb,
-                                           TrafoDepth, MaxTrafoDepth, IntraSplitFlag);
-    ascend();
-
-    return new_tb;
-  }
-
-  assert(false);
-  return NULL;
-}
-
-
-
-enc_tb*
-Algo_TB_IntraPredMode_MinResidual::analyze(encoder_context* ectx,
-                                           context_model_table& ctxModel,
-                                           const de265_image* input,
-                                           enc_tb* tb,
-                                           int TrafoDepth, int MaxTrafoDepth, int IntraSplitFlag)
-{
-  enter();
-
-  enc_cb* cb = tb->cb;
-
-  int x0 = tb->x;
-  int y0 = tb->y;
-  int xBase = cb->x;
-  int yBase = cb->y;
-  int log2TbSize = tb->log2Size;
-
-  bool selectIntraPredMode = false;
-  selectIntraPredMode |= (cb->PredMode==MODE_INTRA && cb->PartMode==PART_2Nx2N && TrafoDepth==0);
-  selectIntraPredMode |= (cb->PredMode==MODE_INTRA && cb->PartMode==PART_NxN   && TrafoDepth==1);
-
-  //printf("tb-intrapredmode: %d %d %d\n", cb->PredMode, cb->PartMode, TrafoDepth);
-
-  if (selectIntraPredMode) {
-
-    *tb->downPtr = tb;
-
-    enum IntraPredMode intraMode;
-    float minDistortion = std::numeric_limits<float>::max();
-
-    assert(nPredModesEnabled()>=1);
-
-    if (nPredModesEnabled()==1) {
-      intraMode = getPredMode(0);
-    }
-    else {
-      tb->intra_prediction0 = std::make_shared<small_image_buffer>(log2TbSize, sizeof(uint8_t));
-
-      for (int idx=0;idx<nPredModesEnabled();idx++) {
-        enum IntraPredMode mode = getPredMode(idx);
-
-        tb->intra_mode = mode;
-        decode_intra_prediction_from_tree(ectx->img, tb, ectx->ctbs, ectx->get_sps(), 0);
-
-        float distortion;
-        distortion = estim_TB_bitrate(ectx, input, tb,
-                                      mParams.bitrateEstimMethod());
-
-        if (distortion<minDistortion) {
-          minDistortion = distortion;
-          intraMode = mode;
-        }
-      }
-    }
-
-    //cb->intra.pred_modeblkIdx = intraMode;
-    //if (blkIdx==0) { cb->intra.chroma_mode = intraMode; }
-
-    //intraMode=(IntraPredMode)17; // HACK
-
-    //printf("INTRA MODE (%d;%d) = %d\n",x0,y0,intraMode);
-
-    tb->intra_mode        = intraMode;
-
-    // set chroma mode to same mode is its luma mode
-    enum IntraPredMode intraModeC;
-
-    if (cb->PartMode==PART_2Nx2N || ectx->get_sps().ChromaArrayType==CHROMA_444) {
-      intraModeC = intraMode;
-    }
-    else {
-      intraModeC = tb->parent->children0->intra_mode;
-    }
-
-    tb->intra_mode_chroma = intraModeC;
-
-
-    // Note: cannot prepare intra prediction pixels here, because this has to
-    // be done at the lowest TB split level.
-
-
-    descend(tb,"%d",intraMode);
-    tb = mTBSplitAlgo->analyze(ectx,ctxModel,input,tb,
-                               TrafoDepth, MaxTrafoDepth, IntraSplitFlag);
-    ascend();
-
-    debug_show_image(ectx->img, 0);
-
-
-    enum IntraPredMode candidates3;
-    fillIntraPredModeCandidates(candidates, x0,y0,
-                                x0>0, y0>0, ectx->ctbs, &ectx->get_sps());
-
-    float intraPredModeBits = get_intra_pred_mode_bits(candidates,
-                                                       intraMode,
-                                                       intraModeC,
-                                                       ctxModel,
-                                                       tb->blkIdx == 0);
-
-    tb->rate_withoutCbfChroma += intraPredModeBits;
-    tb->rate += intraPredModeBits;
-
-    return tb;
-  }
-  else {
-    descend(tb,"NOP");
-    enc_tb* nop_tb = mTBSplitAlgo->analyze(ectx, ctxModel, input, tb,
-                                           TrafoDepth, MaxTrafoDepth,
-                                           IntraSplitFlag);
-    ascend();
-    return nop_tb;
-  }
-
-  assert(false);
-  return NULL;
-}
-
-static bool sortDistortions(std::pair<enum IntraPredMode,float> i,
-                            std::pair<enum IntraPredMode,float> j)
-{
-  return i.second < j.second;
-}
-
-
-enc_tb*
-Algo_TB_IntraPredMode_FastBrute::analyze(encoder_context* ectx,
-                                         context_model_table& ctxModel,
-                                         const de265_image* input,
-                                         enc_tb* tb,
-                                         int TrafoDepth, int MaxTrafoDepth, int IntraSplitFlag)
-{
-  enc_cb* cb = tb->cb;
-
-  bool selectIntraPredMode = false;
-  selectIntraPredMode |= (cb->PredMode==MODE_INTRA && cb->PartMode==PART_2Nx2N && TrafoDepth==0);
-  selectIntraPredMode |= (cb->PredMode==MODE_INTRA && cb->PartMode==PART_NxN   && TrafoDepth==1);
-
-  if (selectIntraPredMode) {
-    float minCost = std::numeric_limits<float>::max();
-    int   minCostIdx=0;
-    float minCandCost;
-
-    const seq_parameter_set* sps = &ectx->get_sps();
-    enum IntraPredMode candidates3;
-    fillIntraPredModeCandidates(candidates, tb->x,tb->y,
-                                tb->x>0, tb->y>0, ectx->ctbs, &ectx->get_sps());
-
-
-
-    std::vector< std::pair<enum IntraPredMode,float> > distortions;
-
-    int log2TbSize = tb->log2Size;
-    tb->intra_prediction0 = std::make_shared<small_image_buffer>(log2TbSize, sizeof(uint8_t));
-
-    for (int idx=0;idx<35;idx++)
-      if (idx!=candidates0 && idx!=candidates1 && idx!=candidates2 &&
-          isPredModeEnabled((enum IntraPredMode)idx))
-        {
-          enum IntraPredMode mode = (enum IntraPredMode)idx;
-
-          tb->intra_mode = mode;
-          decode_intra_prediction_from_tree(ectx->img, tb, ectx->ctbs, ectx->get_sps(), 0);
-
-          float distortion;
-          distortion = estim_TB_bitrate(ectx, input, tb,
-                                        mParams.bitrateEstimMethod());
-
-          distortions.push_back( std::make_pair((enum IntraPredMode)idx, distortion) );
-        }
-
-    std::sort( distortions.begin(), distortions.end(), sortDistortions );
-
-
-    for (int i=0;i<distortions.size();i++)
-      {
-        //printf("%d -> %f\n",i,distortionsi.second);
-      }
-
-    int keepNBest=std::min((int)mParams.keepNBest, (int)distortions.size());
-    distortions.resize(keepNBest);
-    distortions.push_back(std::make_pair((enum IntraPredMode)candidates0,0));
-    distortions.push_back(std::make_pair((enum IntraPredMode)candidates1,0));
-    distortions.push_back(std::make_pair((enum IntraPredMode)candidates2,0));
-
-
-    CodingOptions<enc_tb> options(ectx, tb, ctxModel);
-    std::vector<CodingOption<enc_tb> >  option;
-
-    for (size_t i=0;i<distortions.size();i++) {
-      enum IntraPredMode intraMode  = (IntraPredMode)distortionsi.first;
-      if (!isPredModeEnabled(intraMode)) { continue; }
-
-      CodingOption<enc_tb> opt = options.new_option(isPredModeEnabled(intraMode));
-      opt.get_node()->intra_mode = intraMode;
-      option.push_back(opt);
-    }
-
-    options.start();
-
-
-    for (int i=0;i<option.size();i++) {
-
-      enc_tb* opt_tb = optioni.get_node();
-
-      *opt_tb->downPtr = opt_tb;
-
-      // set chroma mode to same mode is its luma mode
-      enum IntraPredMode intraModeC;
-      if (cb->PartMode==PART_2Nx2N || ectx->get_sps().ChromaArrayType==CHROMA_444) {
-        intraModeC = opt_tb->intra_mode;
-      }
-      else {
-        intraModeC = opt_tb->parent->children0->intra_mode;
-      }
-
-      opt_tb->intra_mode_chroma = intraModeC;
-
-      optioni.begin();
-
-      descend(opt_tb,"%d",opt_tb->intra_mode);
-      opt_tb = mTBSplitAlgo->analyze(ectx,optioni.get_context(),input,opt_tb,
-                                     TrafoDepth, MaxTrafoDepth, IntraSplitFlag);
-      optioni.set_node(opt_tb);
-      ascend();
-
-
-      float intraPredModeBits = get_intra_pred_mode_bits(candidates,
-                                                         opt_tb->intra_mode,
-                                                         intraModeC,
-                                                         optioni.get_context(),
-                                                         tb->blkIdx == 0);
-
-      opt_tb->rate_withoutCbfChroma += intraPredModeBits;
-      opt_tb->rate += intraPredModeBits;
-
-      optioni.end();
-    }
-
-
-    options.compute_rdo_costs();
-
-    return options.return_best_rdo_node();
-  }
-  else {
-    descend(tb,"NOP");
-    enc_tb* new_tb = mTBSplitAlgo->analyze(ectx, ctxModel, input, tb,
-                                           TrafoDepth, MaxTrafoDepth, IntraSplitFlag);
-    ascend();
-    return new_tb;
-  }
-
-  assert(false);
-  return NULL;
-}

libde265-1.0.17.tar.gz/libde265/encoder/algo/tb-intrapredmode.h Deleted

@@ -1,297 +0,0 @@
-/*
- * H.265 video codec.
- * Copyright (c) 2013-2014 struktur AG, Dirk Farin <farin@struktur.de>
- *
- * Authors: Dirk Farin <farin@struktur.de>
- *
- * This file is part of libde265.
- *
- * libde265 is free software: you can redistribute it and/or modify
- * it under the terms of the GNU Lesser General Public License as
- * published by the Free Software Foundation, either version 3 of
- * the License, or (at your option) any later version.
- *
- * libde265 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 Lesser General Public License for more details.
- *
- * You should have received a copy of the GNU Lesser General Public License
- * along with libde265.  If not, see <http://www.gnu.org/licenses/>.
- */
-
-#ifndef TB_INTRAPREDMODE_H
-#define TB_INTRAPREDMODE_H
-
-#include "libde265/nal-parser.h"
-#include "libde265/decctx.h"
-#include "libde265/encoder/encoder-types.h"
-#include "libde265/encoder/algo/algo.h"
-#include "libde265/slice.h"
-#include "libde265/scan.h"
-#include "libde265/intrapred.h"
-#include "libde265/transform.h"
-#include "libde265/fallback-dct.h"
-#include "libde265/quality.h"
-#include "libde265/fallback.h"
-#include "libde265/encoder/configparam.h"
-
-
-// ========== TB intra prediction mode ==========
-
-enum ALGO_TB_IntraPredMode {
-  ALGO_TB_IntraPredMode_BruteForce,
-  ALGO_TB_IntraPredMode_FastBrute,
-  ALGO_TB_IntraPredMode_MinResidual
-};
-
-class option_ALGO_TB_IntraPredMode : public choice_option<enum ALGO_TB_IntraPredMode>
-{
- public:
-  option_ALGO_TB_IntraPredMode() {
-    add_choice("min-residual",ALGO_TB_IntraPredMode_MinResidual);
-    add_choice("brute-force" ,ALGO_TB_IntraPredMode_BruteForce);
-    add_choice("fast-brute"  ,ALGO_TB_IntraPredMode_FastBrute, true);
-  }
-};
-
-
-enum TBBitrateEstimMethod {
-  //TBBitrateEstim_AccurateBits,
-  TBBitrateEstim_SSD,
-  TBBitrateEstim_SAD,
-  TBBitrateEstim_SATD_DCT,
-  TBBitrateEstim_SATD_Hadamard
-};
-
-class option_TBBitrateEstimMethod : public choice_option<enum TBBitrateEstimMethod>
-{
- public:
-  option_TBBitrateEstimMethod() {
-    add_choice("ssd",TBBitrateEstim_SSD);
-    add_choice("sad",TBBitrateEstim_SAD);
-    add_choice("satd-dct",TBBitrateEstim_SATD_DCT);
-    add_choice("satd",TBBitrateEstim_SATD_Hadamard, true);
-  }
-};
-
-class Algo_TB_Split;
-
-
-/** Base class for intra prediction-mode algorithms.
-    Selects one of the 35 prediction modes.
- */
-class Algo_TB_IntraPredMode : public Algo
-{
- public:
-  Algo_TB_IntraPredMode() : mTBSplitAlgo(NULL) { }
-  virtual ~Algo_TB_IntraPredMode() { }
-
-  virtual enc_tb* analyze(encoder_context*,
-                          context_model_table&,
-                          const de265_image* input,
-                          enc_tb* tb,
-                          int TrafoDepth, int MaxTrafoDepth, int IntraSplitFlag) = 0;
-
-  void setChildAlgo(Algo_TB_Split* algo) { mTBSplitAlgo = algo; }
-
-  const char* name() const { return "tb-intrapredmode"; }
-
- protected:
-  Algo_TB_Split* mTBSplitAlgo;
-};
-
-
-enum ALGO_TB_IntraPredMode_Subset {
-  ALGO_TB_IntraPredMode_Subset_All,
-  ALGO_TB_IntraPredMode_Subset_HVPlus,
-  ALGO_TB_IntraPredMode_Subset_DC,
-  ALGO_TB_IntraPredMode_Subset_Planar
-};
-
-class option_ALGO_TB_IntraPredMode_Subset : public choice_option<enum ALGO_TB_IntraPredMode_Subset>
-{
- public:
-  option_ALGO_TB_IntraPredMode_Subset() {
-    add_choice("all"   ,ALGO_TB_IntraPredMode_Subset_All, true);
-    add_choice("HV+"   ,ALGO_TB_IntraPredMode_Subset_HVPlus);
-    add_choice("DC"    ,ALGO_TB_IntraPredMode_Subset_DC);
-    add_choice("planar",ALGO_TB_IntraPredMode_Subset_Planar);
-  }
-};
-
-
-/** Utility class for intra prediction-mode algorithm that uses a subset of modes.
- */
-class Algo_TB_IntraPredMode_ModeSubset : public Algo_TB_IntraPredMode
-{
- public:
-  Algo_TB_IntraPredMode_ModeSubset() {
-    enableAllIntraPredModes();
-  }
-
-  void enableAllIntraPredModes() {
-    for (int i=0;i<35;i++) {
-      mPredMode_enabledi = true;
-      mPredModei = (enum IntraPredMode)i;
-    }
-
-    mNumPredModesEnabled = 35;
-  }
-
-  void disableAllIntraPredModes() {
-    for (int i=0;i<35;i++) {
-      mPredMode_enabledi = false;
-    }
-
-    mNumPredModesEnabled = 0;
-  }
-
-  void enableIntraPredMode(enum IntraPredMode mode) {
-    if (!mPredMode_enabledmode) {
-      mPredModemNumPredModesEnabled = mode;
-      mPredMode_enabledmode = true;
-      mNumPredModesEnabled++;
-    }
-  }
-
-  // TODO: method to disable modes
-
-  void enableIntraPredModeSubset(enum ALGO_TB_IntraPredMode_Subset subset) {
-    switch (subset)
-      {
-      case ALGO_TB_IntraPredMode_Subset_All: // activate all is the default
-        for (int i=0;i<35;i++) { enableIntraPredMode((enum IntraPredMode)i); }
-        break;
-      case ALGO_TB_IntraPredMode_Subset_DC:
-        disableAllIntraPredModes();
-        enableIntraPredMode(INTRA_DC);
-        break;
-      case ALGO_TB_IntraPredMode_Subset_Planar:
-        disableAllIntraPredModes();
-        enableIntraPredMode(INTRA_PLANAR);
-        break;
-      case ALGO_TB_IntraPredMode_Subset_HVPlus:
-        disableAllIntraPredModes();
-        enableIntraPredMode(INTRA_DC);
-        enableIntraPredMode(INTRA_PLANAR);
-        enableIntraPredMode(INTRA_ANGULAR_10);
-        enableIntraPredMode(INTRA_ANGULAR_26);
-        break;
-      }
-  }
-
-
-  enum IntraPredMode getPredMode(int idx) const {
-    assert(idx<mNumPredModesEnabled);
-    return mPredModeidx;
-  }
-
-  int nPredModesEnabled() const {
-    return mNumPredModesEnabled;
-  }
-
-  bool isPredModeEnabled(enum IntraPredMode mode) {
-    return mPredMode_enabledmode;
-  }
-
- private:
-  IntraPredMode mPredMode35;
-  bool mPredMode_enabled35;
-  int  mNumPredModesEnabled;
-};
-
-
-/** Algorithm that brute-forces through all intra prediction mode.
- */
-class Algo_TB_IntraPredMode_BruteForce : public Algo_TB_IntraPredMode_ModeSubset
-{
- public:
-
-  virtual enc_tb* analyze(encoder_context*,
-                          context_model_table&,
-                          const de265_image* input,
-                          enc_tb* tb,
-                          int TrafoDepth, int MaxTrafoDepth, int IntraSplitFlag);
-
-
-  const char* name() const { return "tb-intrapredmode_BruteForce"; }
-};
-
-
-/** Algorithm that makes a quick pre-selection of modes and then brute-forces through them.
- */
-class Algo_TB_IntraPredMode_FastBrute : public Algo_TB_IntraPredMode_ModeSubset
-{
- public:
-
-  struct params
-  {
-    params() {
-      keepNBest.set_ID("IntraPredMode-FastBrute-keepNBest");
-      keepNBest.set_range(0,32);
-      keepNBest.set_default(5);
-
-      bitrateEstimMethod.set_ID("IntraPredMode-FastBrute-estimator");
-    }
-
-    option_TBBitrateEstimMethod bitrateEstimMethod;
-    option_int keepNBest;
-  };
-
-  void registerParams(config_parameters& config) {
-    config.add_option(&mParams.keepNBest);
-    config.add_option(&mParams.bitrateEstimMethod);
-  }
-
-  void setParams(const params& p) { mParams=p; }
-
-
-  virtual enc_tb* analyze(encoder_context*,
-                          context_model_table&,
-                          const de265_image* input,
-                          enc_tb* tb,
-                          int TrafoDepth, int MaxTrafoDepth, int IntraSplitFlag);
-
-
-  const char* name() const { return "tb-intrapredmode_FastBrute"; }
-
- private:
-  params mParams;
-};
-
-
-/** Algorithm that selects the intra prediction mode on minimum residual only.
- */
-class Algo_TB_IntraPredMode_MinResidual : public Algo_TB_IntraPredMode_ModeSubset
-{
- public:
-
-  struct params
-  {
-    params() {
-      bitrateEstimMethod.set_ID("IntraPredMode-MinResidual-estimator");
-    }
-
-    option_TBBitrateEstimMethod bitrateEstimMethod;
-  };
-
-  void setParams(const params& p) { mParams=p; }
-
-  void registerParams(config_parameters& config) {
-    config.add_option(&mParams.bitrateEstimMethod);
-  }
-
-  enc_tb* analyze(encoder_context*,
-                  context_model_table&,
-                  const de265_image* input,
-                  enc_tb* tb,
-                  int TrafoDepth, int MaxTrafoDepth, int IntraSplitFlag);
-
-  const char* name() const { return "tb-intrapredmode_MinResidual"; }
-
- private:
-  params mParams;
-};
-
-#endif

libde265-1.0.17.tar.gz/libde265/encoder/algo/tb-rateestim.cc Deleted

@@ -1,46 +0,0 @@
-/*
- * H.265 video codec.
- * Copyright (c) 2013-2014 struktur AG, Dirk Farin <farin@struktur.de>
- *
- * Authors: struktur AG, Dirk Farin <farin@struktur.de>
- *
- * This file is part of libde265.
- *
- * libde265 is free software: you can redistribute it and/or modify
- * it under the terms of the GNU Lesser General Public License as
- * published by the Free Software Foundation, either version 3 of
- * the License, or (at your option) any later version.
- *
- * libde265 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 Lesser General Public License for more details.
- *
- * You should have received a copy of the GNU Lesser General Public License
- * along with libde265.  If not, see <http://www.gnu.org/licenses/>.
- */
-
-
-#include "libde265/encoder/algo/tb-rateestim.h"
-#include "libde265/encoder/encoder-syntax.h"
-#include <assert.h>
-#include <iostream>
-
-
-float Algo_TB_RateEstimation_Exact::encode_transform_unit(encoder_context* ectx,
-                                                          context_model_table& ctxModel,
-                                                          const enc_tb* tb, const enc_cb* cb,
-                                                          int x0,int y0, int xBase,int yBase,
-                                                          int log2TrafoSize, int trafoDepth,
-                                                          int blkIdx)
-{
-  CABAC_encoder_estim estim;
-  estim.set_context_models(&ctxModel);
-
-  leaf(cb, NULL);
-
-  ::encode_transform_unit(ectx, &estim, tb,cb, x0,y0, xBase,yBase,
-                          log2TrafoSize, trafoDepth, blkIdx);
-
-  return estim.getRDBits();
-}

libde265-1.0.17.tar.gz/libde265/encoder/algo/tb-rateestim.h Deleted

@@ -1,101 +0,0 @@
-/*
- * H.265 video codec.
- * Copyright (c) 2013-2014 struktur AG, Dirk Farin <farin@struktur.de>
- *
- * Authors: Dirk Farin <farin@struktur.de>
- *
- * This file is part of libde265.
- *
- * libde265 is free software: you can redistribute it and/or modify
- * it under the terms of the GNU Lesser General Public License as
- * published by the Free Software Foundation, either version 3 of
- * the License, or (at your option) any later version.
- *
- * libde265 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 Lesser General Public License for more details.
- *
- * You should have received a copy of the GNU Lesser General Public License
- * along with libde265.  If not, see <http://www.gnu.org/licenses/>.
- */
-
-#ifndef TB_RATEESTIM_H
-#define TB_RATEESTIM_H
-
-#include "libde265/nal-parser.h"
-#include "libde265/decctx.h"
-#include "libde265/encoder/encoder-types.h"
-#include "libde265/encoder/algo/algo.h"
-#include "libde265/slice.h"
-#include "libde265/scan.h"
-#include "libde265/intrapred.h"
-#include "libde265/transform.h"
-#include "libde265/fallback-dct.h"
-#include "libde265/quality.h"
-#include "libde265/fallback.h"
-#include "libde265/encoder/configparam.h"
-
-
-enum ALGO_TB_RateEstimation {
-  ALGO_TB_RateEstimation_None,
-  ALGO_TB_RateEstimation_Exact
-};
-
-class option_ALGO_TB_RateEstimation : public choice_option<enum ALGO_TB_RateEstimation>
-{
- public:
-  option_ALGO_TB_RateEstimation() {
-    add_choice("none" ,ALGO_TB_RateEstimation_None);
-    add_choice("exact",ALGO_TB_RateEstimation_Exact, true);
-  }
-};
-
-
-
-class Algo_TB_RateEstimation : public Algo
-{
- public:
-  virtual ~Algo_TB_RateEstimation() { }
-
-  virtual float encode_transform_unit(encoder_context* ectx,
-                                      context_model_table& ctxModel,
-                                      const enc_tb* tb, const enc_cb* cb,
-                                      int x0,int y0, int xBase,int yBase,
-                                      int log2TrafoSize, int trafoDepth, int blkIdx) = 0;
-
-  virtual const char* name() const { return "tb-rateestimation"; }
-};
-
-
-class Algo_TB_RateEstimation_None : public Algo_TB_RateEstimation
-{
- public:
-  virtual float encode_transform_unit(encoder_context* ectx,
-                                      context_model_table& ctxModel,
-                                      const enc_tb* tb, const enc_cb* cb,
-                                      int x0,int y0, int xBase,int yBase,
-                                      int log2TrafoSize, int trafoDepth, int blkIdx)
-  {
-    leaf(cb, NULL);
-    return 0.0f;
-  }
-
-  virtual const char* name() const { return "tb-rateestimation-none"; }
-};
-
-
-class Algo_TB_RateEstimation_Exact : public Algo_TB_RateEstimation
-{
- public:
-  virtual float encode_transform_unit(encoder_context* ectx,
-                                      context_model_table& ctxModel,
-                                      const enc_tb* tb, const enc_cb* cb,
-                                      int x0,int y0, int xBase,int yBase,
-                                      int log2TrafoSize, int trafoDepth, int blkIdx);
-
-  virtual const char* name() const { return "tb-rateestimation-exact"; }
-};
-
-
-#endif

libde265-1.0.17.tar.gz/libde265/encoder/algo/tb-split.cc Deleted

@@ -1,378 +0,0 @@
-/*
- * H.265 video codec.
- * Copyright (c) 2013-2014 struktur AG, Dirk Farin <farin@struktur.de>
- *
- * Authors: struktur AG, Dirk Farin <farin@struktur.de>
- *
- * This file is part of libde265.
- *
- * libde265 is free software: you can redistribute it and/or modify
- * it under the terms of the GNU Lesser General Public License as
- * published by the Free Software Foundation, either version 3 of
- * the License, or (at your option) any later version.
- *
- * libde265 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 Lesser General Public License for more details.
- *
- * You should have received a copy of the GNU Lesser General Public License
- * along with libde265.  If not, see <http://www.gnu.org/licenses/>.
- */
-
-
-#include "libde265/encoder/encoder-core.h"
-#include "libde265/encoder/encoder-context.h"
-#include "libde265/encoder/encoder-syntax.h"
-#include "libde265/encoder/encoder-intrapred.h"
-#include "libde265/encoder/algo/coding-options.h"
-#include <assert.h>
-#include <limits>
-#include <math.h>
-#include <iostream>
-
-
-struct Logging_TB_Split : public Logging
-{
-  int skipTBSplit, noskipTBSplit;
-  int zeroBlockCorrelation625;
-
-  const char* name() const { return "tb-split"; }
-
-  void print(const encoder_context* ectx, const char* filename)
-  {
-    for (int tb=3;tb<=5;tb++) {
-      for (int z=0;z<=1;z++) {
-        float total = 0;
-
-        for (int c=0;c<5;c++)
-          total += zeroBlockCorrelationtbzc;
-
-        for (int c=0;c<5;c++) {
-          printf("%d %d %d : %d %5.2f\n", tb,z,c,
-                 zeroBlockCorrelationtbzc,
-                 total==0 ? 0 : zeroBlockCorrelationtbzc/total*100);
-        }
-      }
-    }
-
-
-    for (int z=0;z<2;z++) {
-      printf("\n");
-      for (int tb=3;tb<=5;tb++) {
-        float total = 0;
-
-        for (int c=0;c<5;c++)
-          total += zeroBlockCorrelationtbzc;
-
-        printf("%dx%d ",1<<tb,1<<tb);
-
-        for (int c=0;c<5;c++) {
-          printf("%5.2f ", total==0 ? 0 : zeroBlockCorrelationtbzc/total*100);
-        }
-        printf("\n");
-      }
-    }
-  }
-} logging_tb_split;
-
-
-template <class pixel_t>
-void diff_blk(int16_t* out,int out_stride,
-              const pixel_t* a_ptr, int a_stride,
-              const pixel_t* b_ptr, int b_stride,
-              int blkSize)
-{
-  for (int by=0;by<blkSize;by++)
-    for (int bx=0;bx<blkSize;bx++)
-      {
-        outby*out_stride+bx = a_ptrby*a_stride+bx - b_ptrby*b_stride+bx;
-      }
-}
-
-
-template <class pixel_t>
-void compute_residual_channel(encoder_context* ectx, enc_tb* tb, const de265_image* input,
-                              int cIdx, int x,int y,int log2Size)
-{
-  int blkSize = (1<<log2Size);
-
-  enum IntraPredMode mode;
-
-  if (cIdx==0) {
-    mode = tb->intra_mode;
-  }
-  else {
-    mode = tb->intra_mode_chroma;
-  }
-
-  // decode intra prediction
-
-  tb->intra_predictioncIdx = std::make_shared<small_image_buffer>(log2Size, sizeof(pixel_t));
-
-  decode_intra_prediction_from_tree(ectx->img, tb, ectx->ctbs, ectx->get_sps(), cIdx);
-
-  // create residual buffer and compute differences
-
-  tb->residualcIdx = std::make_shared<small_image_buffer>(log2Size, sizeof(int16_t));
-
-  diff_blk<pixel_t>(tb->residualcIdx->get_buffer_s16(), blkSize,
-                    input->get_image_plane_at_pos(cIdx,x,y),
-                    input->get_image_stride(cIdx),
-                    tb->intra_predictioncIdx->get_buffer<pixel_t>(), blkSize,
-                    blkSize);
-}
-
-
-template <class pixel_t>
-void compute_residual(encoder_context* ectx, enc_tb* tb, const de265_image* input, int blkIdx)
-{
-  int tbSize = 1<<tb->log2Size;
-
-  /*
-  tb->writeSurroundingMetadata(ectx, ectx->img,
-                               enc_node::METADATA_RECONSTRUCTION_BORDERS,
-                               tb->get_rectangle_with_width(1<<(tb->log2Size+1)));
-  */
-
-  compute_residual_channel<pixel_t>(ectx,tb,input, 0,tb->x,tb->y,tb->log2Size);
-
-  if (ectx->get_sps().chroma_format_idc == CHROMA_444) {
-    compute_residual_channel<pixel_t>(ectx,tb,input, 1,tb->x,tb->y,tb->log2Size);
-    compute_residual_channel<pixel_t>(ectx,tb,input, 2,tb->x,tb->y,tb->log2Size);
-  }
-  else if (tb->log2Size > 2) {
-    int x = tb->x / input->SubWidthC;
-    int y = tb->y / input->SubHeightC;
-    int log2BlkSize = tb->log2Size -1;  // TODO chroma 422/444
-
-    compute_residual_channel<pixel_t>(ectx,tb,input, 1,x,y,log2BlkSize);
-    compute_residual_channel<pixel_t>(ectx,tb,input, 2,x,y,log2BlkSize);
-  }
-  else if (blkIdx==3) {
-    int x = tb->parent->x / input->SubWidthC;
-    int y = tb->parent->y / input->SubHeightC;
-    int log2BlkSize = tb->log2Size;
-
-    compute_residual_channel<pixel_t>(ectx,tb,input, 1,x,y,log2BlkSize);
-    compute_residual_channel<pixel_t>(ectx,tb,input, 2,x,y,log2BlkSize);
-  }
-}
-
-
-enc_tb*
-Algo_TB_Split_BruteForce::analyze(encoder_context* ectx,
-                                  context_model_table& ctxModel,
-                                  const de265_image* input,
-                                  enc_tb* tb,
-                                  int TrafoDepth, int MaxTrafoDepth, int IntraSplitFlag)
-{
-  enter();
-
-  enc_cb* cb = tb->cb;
-
-  int log2TbSize = tb->log2Size;
-
-  bool test_split = (log2TbSize > 2 &&
-                     TrafoDepth < MaxTrafoDepth &&
-                     log2TbSize > ectx->get_sps().Log2MinTrafoSize);
-
-  bool test_no_split = true;
-  if (IntraSplitFlag && TrafoDepth==0) test_no_split=false; // we have to split
-  if (log2TbSize > ectx->get_sps().Log2MaxTrafoSize) test_no_split=false;
-
-  assert(test_no_split || test_split);
-
-  CodingOptions<enc_tb> options(ectx, tb, ctxModel);
-
-  CodingOption<enc_tb> option_no_split = options.new_option(test_no_split);
-  CodingOption<enc_tb> option_split    = options.new_option(test_split);
-
-  //if (test_no_split) test_split = false;
-  //if (test_split) test_no_split = false;   // HACK for debugging
-
-  options.start();
-
-
-  enc_tb* tb_no_split = NULL;
-  enc_tb* tb_split    = NULL;
-
-  if (test_no_split) {
-    descend(tb,"no split");
-    option_no_split.begin();
-    tb_no_split = option_no_split.get_node();
-    //tb_no_split = new enc_tb(*tb);
-    *tb->downPtr = tb_no_split;
-
-    if (cb->PredMode == MODE_INTRA) {
-      compute_residual<uint8_t>(ectx, tb_no_split, input, tb->blkIdx);
-    }
-
-    tb_no_split = mAlgo_TB_Residual->analyze(ectx, option_no_split.get_context(),
-                                             input, tb_no_split, TrafoDepth,MaxTrafoDepth,IntraSplitFlag);
-    ascend(tb_no_split,"bits:%f/%f",tb_no_split->rate,tb_no_split->rate_withoutCbfChroma);
-
-
-    option_no_split.set_node(tb_no_split);
-    option_no_split.end();
-
-
-    // --- some statistics ---
-
-    if (log2TbSize <= mParams.zeroBlockPrune()) {
-      bool zeroBlock = tb_no_split->isZeroBlock();
-
-      if (zeroBlock) {
-        test_split = false;
-        logging_tb_split.skipTBSplit++;
-      }
-      else
-        logging_tb_split.noskipTBSplit++;
-    }
-  }
-
-
-  if (test_split) {
-    option_split.begin();
-
-    tb_split = option_split.get_node();
-    *tb->downPtr = tb_split;
-
-    //descend(tb,"split");
-    tb_split = encode_transform_tree_split(ectx, option_split.get_context(), input, tb_split, cb,
-                                           TrafoDepth, MaxTrafoDepth, IntraSplitFlag);
-    option_split.set_node(tb_split);
-    //ascend("bits:%f/%f",tb_split->rate,tb_split->rate_withoutCbfChroma);
-
-    option_split.end();
-  }
-
-
-  // --- do some statistics that will help us develop a fast algorithm ---
-
-  if (test_split && test_no_split) {
-    bool zero_block = tb_no_split->isZeroBlock();
-
-    int nChildZero = 0;
-    for (int i=0;i<4;i++) {
-      if (tb_split->childreni->isZeroBlock()) nChildZero++;
-    }
-
-    logging_tb_split.zeroBlockCorrelationlog2TbSizezero_block ? 0 : 1nChildZero++;
-  }
-
-
-  //bool split = (rd_cost_split < rd_cost_no_split);
-
-  //if (test_split) split=true;  /// DEBUGGING HACK
-
-  options.compute_rdo_costs();
-
-  enc_tb* bestTB = options.return_best_rdo_node();
-  return bestTB;
-}
-
-
-
-enc_tb* Algo_TB_Split::encode_transform_tree_split(encoder_context* ectx,
-                                                   context_model_table& ctxModel,
-                                                   const de265_image* input,
-                                                   enc_tb* tb,
-                                                   enc_cb* cb,
-                                                   int TrafoDepth, int MaxTrafoDepth,
-                                                   int IntraSplitFlag)
-{
-  const de265_image* img = ectx->img;
-
-  int log2TbSize = tb->log2Size;
-  int x0 = tb->x;
-  int y0 = tb->y;
-
-  context_model ctxModelCbfChroma4;
-  for (int i=0;i<4;i++) {
-    ctxModelCbfChromai = ctxModelCONTEXT_MODEL_CBF_CHROMA+i;
-  }
-
-  tb->split_transform_flag = true;
-  tb->rate_withoutCbfChroma = 0;
-  tb->distortion = 0;
-
-  // --- encode all child nodes ---
-
-  for (int i=0;i<4;i++) {
-    tb->childreni = NULL;
-  }
-
-  for (int i=0;i<4;i++) {
-
-    // generate child node and propagate values down
-
-    int dx = (i&1)  << (log2TbSize-1);
-    int dy = (i>>1) << (log2TbSize-1);
-
-    enc_tb* child_tb = new enc_tb(x0+dx,y0+dy, log2TbSize-1,cb);
-
-    child_tb->intra_mode        = tb->intra_mode;
-    child_tb->intra_mode_chroma = tb->intra_mode_chroma;
-    child_tb->TrafoDepth = tb->TrafoDepth + 1;
-    child_tb->parent = tb;
-    child_tb->blkIdx = i;
-    child_tb->downPtr = &tb->childreni;
-
-    descend(tb,"split %d/4",i+1);
-
-    if (cb->PredMode == MODE_INTRA) {
-      //descend(tb,"intra");
-      tb->childreni = mAlgo_TB_IntraPredMode->analyze(ectx, ctxModel, input,
-                                                        child_tb,
-                                                        TrafoDepth+1, MaxTrafoDepth,
-                                                        IntraSplitFlag);
-      //ascend("bits:%f",tb->rate);
-    }
-    else {
-      //descend(tb,"inter");
-      tb->childreni = this->analyze(ectx, ctxModel, input,
-                                      child_tb, TrafoDepth+1, MaxTrafoDepth, IntraSplitFlag);
-      //ascend();
-    }
-
-    ascend();
-
-    tb->distortion            += tb->childreni->distortion;
-    tb->rate_withoutCbfChroma += tb->childreni->rate_withoutCbfChroma;
-  }
-
-  tb->set_cbf_flags_from_children();
-
-
-  // --- add rate for this TB level ---
-
-  CABAC_encoder_estim estim;
-  estim.set_context_models(&ctxModel);
-
-
-
-
-  const seq_parameter_set* sps = &ectx->img->get_sps();
-
-  if (log2TbSize <= sps->Log2MaxTrafoSize &&
-      log2TbSize >  sps->Log2MinTrafoSize &&
-      TrafoDepth < MaxTrafoDepth &&
-      !(IntraSplitFlag && TrafoDepth==0))
-    {
-      encode_split_transform_flag(ectx, &estim, log2TbSize, 1);
-      tb->rate_withoutCbfChroma += estim.getRDBits();
-      estim.reset();
-    }
-
-  // restore chroma CBF context models
-
-  for (int i=0;i<4;i++) {
-    ctxModelCONTEXT_MODEL_CBF_CHROMA+i = ctxModelCbfChromai;
-  }
-
-  tb->rate = (tb->rate_withoutCbfChroma +
-              recursive_cbfChroma_rate(&estim,tb, log2TbSize, TrafoDepth));
-
-  return tb;
-}

libde265-1.0.17.tar.gz/libde265/encoder/algo/tb-split.h Deleted

@@ -1,126 +0,0 @@
-/*
- * H.265 video codec.
- * Copyright (c) 2013-2014 struktur AG, Dirk Farin <farin@struktur.de>
- *
- * Authors: Dirk Farin <farin@struktur.de>
- *
- * This file is part of libde265.
- *
- * libde265 is free software: you can redistribute it and/or modify
- * it under the terms of the GNU Lesser General Public License as
- * published by the Free Software Foundation, either version 3 of
- * the License, or (at your option) any later version.
- *
- * libde265 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 Lesser General Public License for more details.
- *
- * You should have received a copy of the GNU Lesser General Public License
- * along with libde265.  If not, see <http://www.gnu.org/licenses/>.
- */
-
-#ifndef TB_SPLIT_H
-#define TB_SPLIT_H
-
-#include "libde265/nal-parser.h"
-#include "libde265/decctx.h"
-#include "libde265/encoder/encoder-types.h"
-#include "libde265/encoder/algo/algo.h"
-#include "libde265/slice.h"
-#include "libde265/scan.h"
-#include "libde265/intrapred.h"
-#include "libde265/transform.h"
-#include "libde265/fallback-dct.h"
-#include "libde265/quality.h"
-#include "libde265/fallback.h"
-#include "libde265/encoder/configparam.h"
-
-#include "libde265/encoder/algo/tb-intrapredmode.h"
-#include "libde265/encoder/algo/tb-rateestim.h"
-#include "libde265/encoder/algo/tb-transform.h"
-
-
-// ========== TB split decision ==========
-
-class Algo_TB_Split : public Algo
-{
- public:
- Algo_TB_Split() : mAlgo_TB_IntraPredMode(NULL) { }
-  virtual ~Algo_TB_Split() { }
-
-  virtual enc_tb* analyze(encoder_context*,
-                          context_model_table&,
-                          const de265_image* input,
-                          enc_tb* tb,
-                          int TrafoDepth, int MaxTrafoDepth, int IntraSplitFlag) = 0;
-
-  void setAlgo_TB_IntraPredMode(Algo_TB_IntraPredMode* algo) { mAlgo_TB_IntraPredMode=algo; }
-  void setAlgo_TB_Residual(Algo_TB_Residual* algo) { mAlgo_TB_Residual=algo; }
-
- protected:
-  enc_tb* encode_transform_tree_split(encoder_context* ectx,
-                                      context_model_table& ctxModel,
-                                      const de265_image* input,
-                                      enc_tb* tb,
-                                      enc_cb* cb,
-                                      int TrafoDepth, int MaxTrafoDepth, int IntraSplitFlag);
-
-  Algo_TB_IntraPredMode* mAlgo_TB_IntraPredMode;
-  Algo_TB_Residual*      mAlgo_TB_Residual;
-};
-
-
-
-enum ALGO_TB_Split_BruteForce_ZeroBlockPrune {
-  // numeric value specifies the maximum size for log2Tb for which the pruning is applied
-  ALGO_TB_BruteForce_ZeroBlockPrune_off = 0,
-  ALGO_TB_BruteForce_ZeroBlockPrune_8x8 = 3,
-  ALGO_TB_BruteForce_ZeroBlockPrune_8x8_16x16 = 4,
-  ALGO_TB_BruteForce_ZeroBlockPrune_all = 5
-};
-
-class option_ALGO_TB_Split_BruteForce_ZeroBlockPrune
-: public choice_option<enum ALGO_TB_Split_BruteForce_ZeroBlockPrune>
-{
- public:
-  option_ALGO_TB_Split_BruteForce_ZeroBlockPrune() {
-    add_choice("off"     ,ALGO_TB_BruteForce_ZeroBlockPrune_off);
-    add_choice("8x8"     ,ALGO_TB_BruteForce_ZeroBlockPrune_8x8);
-    add_choice("8-16"    ,ALGO_TB_BruteForce_ZeroBlockPrune_8x8_16x16);
-    add_choice("all"     ,ALGO_TB_BruteForce_ZeroBlockPrune_all, true);
-  }
-};
-
-
-class Algo_TB_Split_BruteForce : public Algo_TB_Split
-{
- public:
-  struct params
-  {
-    params() {
-      zeroBlockPrune.set_ID("TB-Split-BruteForce-ZeroBlockPrune");
-    }
-
-    option_ALGO_TB_Split_BruteForce_ZeroBlockPrune zeroBlockPrune;
-  };
-
-  void setParams(const params& p) { mParams=p; }
-
-  void registerParams(config_parameters& config) {
-    config.add_option(&mParams.zeroBlockPrune);
-  }
-
-  virtual enc_tb* analyze(encoder_context*,
-                          context_model_table&,
-                          const de265_image* input,
-                          enc_tb* tb,
-                          int TrafoDepth, int MaxTrafoDepth, int IntraSplitFlag);
-
-  const char* name() const { return "tb-split-bruteforce"; }
-
- private:
-  params mParams;
-};
-
-#endif

libde265-1.0.17.tar.gz/libde265/encoder/algo/tb-transform.cc Deleted

@@ -1,254 +0,0 @@
-/*
- * H.265 video codec.
- * Copyright (c) 2013-2014 struktur AG, Dirk Farin <farin@struktur.de>
- *
- * Authors: struktur AG, Dirk Farin <farin@struktur.de>
- *
- * This file is part of libde265.
- *
- * libde265 is free software: you can redistribute it and/or modify
- * it under the terms of the GNU Lesser General Public License as
- * published by the Free Software Foundation, either version 3 of
- * the License, or (at your option) any later version.
- *
- * libde265 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 Lesser General Public License for more details.
- *
- * You should have received a copy of the GNU Lesser General Public License
- * along with libde265.  If not, see <http://www.gnu.org/licenses/>.
- */
-
-
-#include "libde265/encoder/algo/tb-transform.h"
-#include "libde265/encoder/encoder-core.h"
-#include "libde265/encoder/encoder-context.h"
-#include "libde265/encoder/encoder-syntax.h"
-#include <assert.h>
-#include <limits>
-#include <math.h>
-#include <iostream>
-
-
-// DEPRECATED IN THIS FILE
-void diff_blk(int16_t* out,int out_stride,
-              const uint8_t* a_ptr, int a_stride,
-              const uint8_t* b_ptr, int b_stride,
-              int blkSize)
-{
-  for (int by=0;by<blkSize;by++)
-    for (int bx=0;bx<blkSize;bx++)
-      {
-        outby*out_stride+bx = a_ptrby*a_stride+bx - b_ptrby*b_stride+bx;
-      }
-}
-
-
-static bool has_nonzero_value(const int16_t* data, int n)
-{
-  for (int i=0;i<n;i++)
-    if (datai) return true;
-
-  return false;
-}
-
-
-void compute_transform_coeffs(encoder_context* ectx,
-                              enc_tb* tb,
-                              const de265_image* input, // TODO: probably pass pixels/stride directly
-                              //int16_t* residual, int stride,
-                              int x0,int y0, // luma position
-                              int log2TbSize, // chroma adapted
-                              const enc_cb* cb,
-                              int cIdx)
-{
-  int xC = x0;
-  int yC = y0;
-  int tbSize = 1<<log2TbSize;
-  if (cIdx>0) { xC>>=1; yC>>=1; }
-
-  enum PredMode predMode = cb->PredMode;
-
-  int16_t blk32*32; // residual
-  int16_t* residual;
-
-
-  //printf("transform-coeffs %d;%d size:%d cIdx:%d\n", tb->x,tb->y,1<<tb->log2Size,cIdx);
-
-  // --- do intra prediction ---
-
-  if (predMode==MODE_INTRA) {
-    residual = tb->residualcIdx->get_buffer_s16();
-
-    //printf("intra residual: %p stride: %d\n",residual, tb->residualcIdx->get_stride());
-  }
-  else {
-    // --- subtract prediction from input image ---
-
-    /* TMP REMOVE: ectx->prediction does not exist anymore
-
-    uint8_t* pred = ectx->prediction->get_image_plane(cIdx);
-    int stride = ectx->prediction->get_image_stride(cIdx);
-
-    //printBlk("input",input->get_image_plane_at_pos(cIdx,xC,yC), tbSize, input->get_image_stride(cIdx));
-    //printBlk("prediction", pred,tbSize, stride);
-
-    diff_blk(blk,tbSize,
-             input->get_image_plane_at_pos(cIdx,xC,yC), input->get_image_stride(cIdx),
-             &predyC*stride+xC,stride, tbSize);
-
-    residual=blk;
-    */
-
-    //printBlk("residual", blk,tbSize,tbSize);
-  }
-
-
-
-
-
-  // --- forward transform ---
-
-  tb->alloc_coeff_memory(cIdx, tbSize);
-
-
-  // transformation mode (DST or DCT)
-
-  int trType;
-  if (cIdx==0 && log2TbSize==2 && predMode==MODE_INTRA) trType=1; // TODO: inter mode
-  else trType=0;
-
-
-  // do forward transform
-
-  fwd_transform(&ectx->acceleration, tb->coeffcIdx, tbSize, log2TbSize, trType,  residual, tbSize);
-
-
-  // --- quantization ---
-
-  quant_coefficients(tb->coeffcIdx, tb->coeffcIdx, log2TbSize,  cb->qp, true);
-
-
-  // set CBF to 0 if there are no non-zero coefficients
-
-  tb->cbfcIdx = has_nonzero_value(tb->coeffcIdx, 1<<(log2TbSize<<1));
-}
-
-
-enc_tb* Algo_TB_Transform::analyze(encoder_context* ectx,
-                                   context_model_table& ctxModel,
-                                   const de265_image* input,
-                                   enc_tb* tb,
-                                   int trafoDepth, int MaxTrafoDepth,
-                                   int IntraSplitFlag)
-{
-  enter();
-
-  const enc_cb* cb = tb->cb;
-  *tb->downPtr = tb; // TODO: should be obsolet
-
-  de265_image* img = ectx->img;
-
-  int stride = ectx->img->get_image_stride(0);
-
-  uint8_t* luma_plane = ectx->img->get_image_plane(0);
-  uint8_t* cb_plane = ectx->img->get_image_plane(1);
-  uint8_t* cr_plane = ectx->img->get_image_plane(2);
-
-  // --- compute transform coefficients ---
-
-  int x0 = tb->x;
-  int y0 = tb->y;
-  int xBase = cb->x;
-  int yBase = cb->y;
-  int log2TbSize = tb->log2Size;
-
-  // luma block
-
-  compute_transform_coeffs(ectx, tb, input, x0,y0, log2TbSize, cb, 0 /* Y */);
-
-
-  // chroma blocks
-
-  if (ectx->get_sps().chroma_format_idc == CHROMA_444) {
-    compute_transform_coeffs(ectx, tb, input, x0,y0, log2TbSize, cb, 1 /* Cb */);
-    compute_transform_coeffs(ectx, tb, input, x0,y0, log2TbSize, cb, 2 /* Cr */);
-  }
-  else if (log2TbSize > 2) {
-    // if TB is > 4x4, do chroma transform of half size
-    compute_transform_coeffs(ectx, tb, input, x0,y0, log2TbSize-1, cb, 1 /* Cb */);
-    compute_transform_coeffs(ectx, tb, input, x0,y0, log2TbSize-1, cb, 2 /* Cr */);
-  }
-  else if (tb->blkIdx==3) {
-    // if TB size is 4x4, do chroma transform for last sub-block
-    compute_transform_coeffs(ectx, tb, input, xBase,yBase, log2TbSize, cb, 1 /* Cb */);
-    compute_transform_coeffs(ectx, tb, input, xBase,yBase, log2TbSize, cb, 2 /* Cr */);
-  }
-
-
-  // --- reconstruction ---
-
-  /* We could compute the reconstruction lazy on first access. However, we currently
-     use it right away for computing the distortion.
-  */
-  tb->reconstruct(ectx, ectx->img);
-
-
-  // measure rate
-
-  CABAC_encoder_estim estim;
-  estim.set_context_models(&ctxModel);
-
-
-  tb->rate_withoutCbfChroma = 0;
-
-  const seq_parameter_set* sps = &ectx->img->get_sps();
-
-
-  if (log2TbSize <= sps->Log2MaxTrafoSize &&
-      log2TbSize >  sps->Log2MinTrafoSize &&
-      trafoDepth < MaxTrafoDepth &&
-      !(IntraSplitFlag && trafoDepth==0))
-    {
-      encode_split_transform_flag(ectx, &estim, log2TbSize, 0);
-      tb->rate_withoutCbfChroma += estim.getRDBits();
-      estim.reset();
-    }
-
-  // --- CBF CB/CR ---
-
-  float luma_cbf_bits = 0;
-  if (cb->PredMode == MODE_INTRA || trafoDepth != 0 ||
-      tb->cbf1 || tb->cbf2) {
-    encode_cbf_luma(&estim, trafoDepth==0, tb->cbf0);
-    luma_cbf_bits = estim.getRDBits();
-  }
-
-  descend(tb,"DCT");
-  float bits = mAlgo_TB_RateEstimation->encode_transform_unit(ectx,ctxModel,
-                                                              tb,cb, x0,y0, xBase,yBase,
-                                                              log2TbSize, trafoDepth, tb->blkIdx);
-  ascend();
-
-  tb->rate_withoutCbfChroma += bits + luma_cbf_bits;
-
-  estim.reset(); // TODO: not needed ?
-
-  tb->rate = (tb->rate_withoutCbfChroma +
-              recursive_cbfChroma_rate(&estim,tb,log2TbSize,trafoDepth));
-
-  //float rate_cbfChroma = estim.getRDBits();
-  //tb->rate = tb->rate_withoutCbfChroma + rate_cbfChroma;
-
-
-  // measure distortion
-
-  int tbSize = 1<<log2TbSize;
-  tb->distortion = SSD(input->get_image_plane_at_pos(0, x0,y0), input->get_image_stride(0),
-                       tb->reconstruction0->get_buffer_u8(),
-                       tb->reconstruction0->getStride(),
-                       tbSize, tbSize);
-
-  return tb;
-}

libde265-1.0.17.tar.gz/libde265/encoder/algo/tb-transform.h Deleted

@@ -1,86 +0,0 @@
-/*
- * H.265 video codec.
- * Copyright (c) 2013-2014 struktur AG, Dirk Farin <farin@struktur.de>
- *
- * Authors: Dirk Farin <farin@struktur.de>
- *
- * This file is part of libde265.
- *
- * libde265 is free software: you can redistribute it and/or modify
- * it under the terms of the GNU Lesser General Public License as
- * published by the Free Software Foundation, either version 3 of
- * the License, or (at your option) any later version.
- *
- * libde265 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 Lesser General Public License for more details.
- *
- * You should have received a copy of the GNU Lesser General Public License
- * along with libde265.  If not, see <http://www.gnu.org/licenses/>.
- */
-
-#ifndef TB_TRANSFORM_H
-#define TB_TRANSFORM_H
-
-#include "libde265/nal-parser.h"
-#include "libde265/decctx.h"
-#include "libde265/encoder/encoder-types.h"
-#include "libde265/encoder/algo/algo.h"
-#include "libde265/slice.h"
-#include "libde265/scan.h"
-#include "libde265/intrapred.h"
-#include "libde265/transform.h"
-#include "libde265/fallback-dct.h"
-#include "libde265/quality.h"
-#include "libde265/fallback.h"
-#include "libde265/encoder/configparam.h"
-
-#include "libde265/encoder/algo/tb-intrapredmode.h"
-#include "libde265/encoder/algo/tb-rateestim.h"
-
-
-void diff_blk(int16_t* out,int out_stride,
-              const uint8_t* a_ptr, int a_stride,
-              const uint8_t* b_ptr, int b_stride,
-              int blkSize);
-
-
-// ========== TB split decision ==========
-
-class Algo_TB_Residual : public Algo
-{
-public:
-  Algo_TB_Residual() { }
-
-  virtual enc_tb* analyze(encoder_context*,
-                          context_model_table&,
-                          const de265_image* input,
-                          enc_tb* tb,
-                          int TrafoDepth, int MaxTrafoDepth, int IntraSplitFlag) = 0;
-
-  const char* name() const { return "residual-unknown"; }
-};
-
-
-class Algo_TB_Transform : public Algo_TB_Residual
-{
-public:
-  Algo_TB_Transform() : mAlgo_TB_RateEstimation(NULL) { }
-
-  virtual enc_tb* analyze(encoder_context*,
-                          context_model_table&,
-                          const de265_image* input,
-                          enc_tb* parent,
-                          int TrafoDepth, int MaxTrafoDepth, int IntraSplitFlag);
-
-  void setAlgo_TB_RateEstimation(Algo_TB_RateEstimation* algo) { mAlgo_TB_RateEstimation=algo; }
-
-  const char* name() const { return "residual-FDCT"; }
-
- protected:
-  Algo_TB_RateEstimation* mAlgo_TB_RateEstimation;
-};
-
-
-#endif

libde265-1.0.17.tar.gz/libde265/encoder/configparam.cc Deleted

@@ -1,471 +0,0 @@
-/*
- * H.265 video codec.
- * Copyright (c) 2013-2014 struktur AG, Dirk Farin <farin@struktur.de>
- *
- * Authors: struktur AG, Dirk Farin <farin@struktur.de>
- *
- * This file is part of libde265.
- *
- * libde265 is free software: you can redistribute it and/or modify
- * it under the terms of the GNU Lesser General Public License as
- * published by the Free Software Foundation, either version 3 of
- * the License, or (at your option) any later version.
- *
- * libde265 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 Lesser General Public License for more details.
- *
- * You should have received a copy of the GNU Lesser General Public License
- * along with libde265.  If not, see <http://www.gnu.org/licenses/>.
- */
-
-#include "configparam.h"
-
-#include <string.h>
-#include <ctype.h>
-#include <sstream>
-#include <iomanip>
-#include <iostream>
-#include <algorithm>
-#include <typeinfo>
-
-#ifndef RTTI_ENABLED
-#error "Need to compile with RTTI enabled."
-#endif
-
-static void remove_option(int* argc,char** argv,int idx, int n=1)
-{
-  for (int i=idx+n;i<*argc;i++) {
-    argvi-n = argvi;
-  }
-
-  *argc-=n;
-}
-
-
-bool option_string::processCmdLineArguments(char** argv, int* argc, int idx)
-{
-  if (argv==NULL)   { return false; }
-  if (idx >= *argc) { return false; }
-
-  value = argvidx;
-  value_set = true;
-
-  remove_option(argc,argv,idx,1);
-
-  return true;
-}
-
-
-void option_int::set_range(int mini,int maxi)
-{
-  have_low_limit =true;
-  have_high_limit=true;
-  low_limit =mini;
-  high_limit=maxi;
-}
-
-std::string option_int::getTypeDescr() const
-{
-  std::stringstream sstr;
-  sstr << "(int)";
-
-  if (have_low_limit || have_high_limit) { sstr << " "; }
-  if (have_low_limit) { sstr << low_limit << " <= "; }
-  if (have_low_limit || have_high_limit) { sstr << "x"; }
-  if (have_high_limit) { sstr << " <= " << high_limit; }
-
-  if (!valid_values_set.empty()) {
-    sstr << " {";
-    bool first=true;
-    for (int v : valid_values_set) {
-      if (!first) sstr << ","; else first=false;
-      sstr << v;
-    }
-    sstr << "}";
-  }
-
-  return sstr.str();
-}
-
-bool option_int::processCmdLineArguments(char** argv, int* argc, int idx)
-{
-  if (argv==NULL)   { return false; }
-  if (idx >= *argc) { return false; }
-
-  int v = atoi(argvidx);
-  if (!is_valid(v)) { return false; }
-
-  value = v;
-  value_set = true;
-
-  remove_option(argc,argv,idx,1);
-
-  return true;
-}
-
-bool option_int::is_valid(int v) const
-{
-  if (have_low_limit  && v<low_limit)  { return false; }
-  if (have_high_limit && v>high_limit) { return false; }
-
-  if (!valid_values_set.empty()) {
-    auto iter = std::find(valid_values_set.begin(), valid_values_set.end(), v);
-    if (iter==valid_values_set.end()) { return false; }
-  }
-
-  return true;
-}
-
-std::string option_int::get_default_string() const
-{
-  std::stringstream sstr;
-  sstr << default_value;
-  return sstr.str();
-}
-
-
-std::string choice_option_base::getTypeDescr() const
-{
-  std::vector<std::string> choices = get_choice_names();
-
-  std::stringstream sstr;
-  sstr << "{";
-
-  bool first=true;
-  for (const auto& c : choices) {
-    if (first) { first=false; }
-    else { sstr << ","; }
-
-    sstr << c;
-  }
-
-  sstr << "}";
-  return sstr.str();
-}
-
-
-bool choice_option_base::processCmdLineArguments(char** argv, int* argc, int idx)
-{
-  if (argv==NULL)   { return false; }
-  if (idx >= *argc) { return false; }
-
-  std::string value = argvidx;
-
-  std::cout << "set " << value << "\n";
-  bool success = set_value(value);
-  std::cout << "success " << success << "\n";
-
-  remove_option(argc,argv,idx,1);
-
-  return success;
-}
-
-
-static char* fill_strings_into_memory(const std::vector<std::string>& strings_list)
-{
-  // calculate memory requirement
-
-  int totalStringLengths = 0;
-  for (const auto& str : strings_list) {
-    totalStringLengths += str.length() +1; // +1 for null termination
-  }
-
-  int numStrings = strings_list.size();
-
-  int pointersSize = (numStrings+1) * sizeof(const char*);
-
-  char* memory = new charpointersSize + totalStringLengths;
-
-
-  // copy strings to memory area
-
-  char* stringPtr = memory + (numStrings+1) * sizeof(const char*);
-  const char** tablePtr = (const char**)memory;
-
-  for (const auto& str : strings_list) {
-    *tablePtr++ = stringPtr;
-
-    strcpy(stringPtr, str.c_str());
-    stringPtr += str.length()+1;
-  }
-
-  *tablePtr = NULL;
-
-  return memory;
-}
-
-
-const char** choice_option_base::get_choices_string_table() const
-{
-  if (choice_string_table==NULL) {
-    choice_string_table = fill_strings_into_memory(get_choice_names());
-  }
-
-  return (const char**)choice_string_table;
-}
-
-
-
-bool config_parameters::parse_command_line_params(int* argc, char** argv, int* first_idx_ptr,
-                                                  bool ignore_unknown_options)
-{
-  int first_idx=1;
-  if (first_idx_ptr) { first_idx = *first_idx_ptr; }
-
-  for (int i=first_idx;i < *argc;i++) {
-
-    if (argvi0=='-') {
-      // option
-
-      if (argvi1=='-') {
-        // long option
-
-        bool option_found=false;
-
-        for (size_t o=0;o<mOptions.size();o++) {
-          if (mOptionso->hasLongOption() && strcmp(mOptionso->getLongOption().c_str(),
-                                                     argvi+2)==0) {
-            option_found=true;
-
-            printf("FOUND %s\n",argvi);
-
-            bool success = mOptionso->processCmdLineArguments(argv,argc, i+1);
-            if (!success) {
-              if (first_idx_ptr) { *first_idx_ptr = i; }
-              return false;
-            }
-
-            remove_option(argc,argv,i);
-            i--;
-
-            break;
-          }
-        }
-
-        if (option_found == false && !ignore_unknown_options) {
-          return false;
-        }
-      }
-      else {
-        // short option
-
-        bool is_single_option = (argvi1 != 0 && argvi2==0);
-        bool do_remove_option = true;
-
-        for (int n=1; argvin; n++) {
-          char option = argvin;
-
-          bool option_found=false;
-
-          for (size_t o=0;o<mOptions.size();o++) {
-            if (mOptionso->getShortOption() == option) {
-              option_found=true;
-
-              bool success;
-              if (is_single_option) {
-                success = mOptionso->processCmdLineArguments(argv,argc, i+1);
-              }
-              else {
-                success = mOptionso->processCmdLineArguments(NULL,NULL, 0);
-              }
-
-              if (!success) {
-                if (first_idx_ptr) { *first_idx_ptr = i; }
-                return false;
-              }
-
-              break;
-            }
-          }
-
-          if (!option_found) {
-            if (!ignore_unknown_options) {
-              fprintf(stderr, "unknown option -%c\n",option);
-              return false;
-            }
-            else {
-              do_remove_option=false;
-            }
-          }
-
-        } // all short options
-
-        if (do_remove_option) {
-          remove_option(argc,argv,i);
-          i--;
-        }
-      } // is short option
-    } // is option
-  } // all command line arguments
-
-  return true;
-}
-
-
-void config_parameters::print_params() const
-{
-  for (size_t i=0;i<mOptions.size();i++) {
-    const option_base* o = mOptionsi;
-
-    std::stringstream sstr;
-    sstr << "  ";
-    if (o->hasShortOption()) {
-      sstr << '-' << o->getShortOption();
-    } else {
-      sstr << "  ";
-    }
-
-    if (o->hasShortOption() && o->hasLongOption()) {
-      sstr << ", ";
-    } else {
-      sstr << "  ";
-    }
-
-    if (o->hasLongOption()) {
-      sstr << "--" << std::setw(12) << std::left << o->getLongOption();
-    } else {
-      sstr << "              ";
-    }
-
-    sstr << " ";
-    sstr << o->getTypeDescr();
-
-    if (o->has_default()) {
-      sstr << ", default=" << o->get_default_string();
-    }
-
-    if (o->has_description()) {
-      sstr << " : " << o->get_description();
-    }
-
-    sstr << "\n";
-
-    std::cerr << sstr.str();
-  }
-}
-
-
-void config_parameters::add_option(option_base* o)
-{
-  mOptions.push_back(o);
-  delete param_string_table; // delete old table, since we got a new parameter
-  param_string_table = NULL;
-}
-
-
-std::vector<std::string> config_parameters::get_parameter_IDs() const
-{
-  std::vector<std::string> ids;
-
-  for (auto option : mOptions) {
-    ids.push_back(option->get_name());
-  }
-
-  return ids;
-}
-
-
-enum en265_parameter_type config_parameters::get_parameter_type(const char* param) const
-{
-  option_base* option = find_option(param);
-  assert(option);
-
-  if (dynamic_cast<option_int*>   (option)) { return en265_parameter_int; }
-  if (dynamic_cast<option_bool*>  (option)) { return en265_parameter_bool; }
-  if (dynamic_cast<option_string*>(option)) { return en265_parameter_string; }
-  if (dynamic_cast<choice_option_base*>(option)) { return en265_parameter_choice; }
-
-  assert(false);
-  return en265_parameter_bool;
-}
-
-
-std::vector<std::string> config_parameters::get_parameter_choices(const char* param) const
-{
-  option_base* option = find_option(param);
-  assert(option);
-
-  choice_option_base* o = dynamic_cast<choice_option_base*>(option);
-  assert(o);
-
-  return o->get_choice_names();
-}
-
-
-option_base* config_parameters::find_option(const char* param) const
-{
-  for (auto o : mOptions) {
-    if (strcmp(o->get_name().c_str(), param)==0) { return o; }
-  }
-
-  return NULL;
-}
-
-
-bool config_parameters::set_bool(const char* param, bool value)
-{
-  option_base* option = find_option(param);
-  assert(option);
-
-  option_bool* o = dynamic_cast<option_bool*>(option);
-  assert(o);
-
-  return o->set(value);
-}
-
-bool config_parameters::set_int(const char* param, int value)
-{
-  option_base* option = find_option(param);
-  assert(option);
-
-  option_int* o = dynamic_cast<option_int*>(option);
-  assert(o);
-
-  return o->set(value);
-}
-
-bool config_parameters::set_string(const char* param, const char* value)
-{
-  option_base* option = find_option(param);
-  assert(option);
-
-  option_string* o = dynamic_cast<option_string*>(option);
-  assert(o);
-
-  return o->set(value);
-}
-
-bool config_parameters::set_choice(const char* param, const char* value)
-{
-  option_base* option = find_option(param);
-  assert(option);
-
-  choice_option_base* o = dynamic_cast<choice_option_base*>(option);
-  assert(o);
-
-  return o->set(value);
-}
-
-
-
-const char** config_parameters::get_parameter_choices_table(const char* param) const
-{
-  option_base* option = find_option(param);
-  assert(option);
-
-  choice_option_base* o = dynamic_cast<choice_option_base*>(option);
-  assert(o);
-
-  return o->get_choices_string_table();
-}
-
-const char** config_parameters::get_parameter_string_table() const
-{
-  if (param_string_table==NULL) {
-    param_string_table = fill_strings_into_memory(get_parameter_IDs());
-  }
-
-  return (const char**)param_string_table;
-}

libde265-1.0.17.tar.gz/libde265/encoder/configparam.h Deleted

@@ -1,386 +0,0 @@
-/*
- * H.265 video codec.
- * Copyright (c) 2013-2014 struktur AG, Dirk Farin <farin@struktur.de>
- *
- * Authors: struktur AG, Dirk Farin <farin@struktur.de>
- *
- * This file is part of libde265.
- *
- * libde265 is free software: you can redistribute it and/or modify
- * it under the terms of the GNU Lesser General Public License as
- * published by the Free Software Foundation, either version 3 of
- * the License, or (at your option) any later version.
- *
- * libde265 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 Lesser General Public License for more details.
- *
- * You should have received a copy of the GNU Lesser General Public License
- * along with libde265.  If not, see <http://www.gnu.org/licenses/>.
- */
-
-#ifndef CONFIG_PARAM_H
-#define CONFIG_PARAM_H
-
-#include "en265.h"
-#include "util.h"
-
-#include <climits>
-#include <vector>
-#include <string>
-#include <stddef.h>
-#include <assert.h>
-
-
-/* Notes: probably best to keep cmd-line-options here. So it will be:
-   - automatically consistent even when having different combinations of algorithms
-   - no other place to edit
-   - if needed, one can still override it at another place
- */
-
-// TODO: set a stack of default prefixes in config_parameters, such that all options added
-// will receive this namespace prefix.
-
-// TODO: add the possibility to remove long options again, i.e., not use the default id name
-class option_base
-{
- public:
- option_base() : mShortOption(0), mLongOption(NULL) { }
- option_base(const char* name) : mIDName(name), mShortOption(0), mLongOption(NULL) { }
-  virtual ~option_base() { }
-
-
-  // --- option identifier ---
-
-  void set_ID(const char* name) { mIDName=name; }
-  void add_namespace_prefix(std::string prefix) { mPrefix = prefix + ":" + mPrefix; }
-
-  std::string get_name() const { return mPrefix + mIDName; }
-
-
-  // --- description ---
-
-  void set_description(std::string descr) { mDescription = descr; }
-  std::string get_description() const { return mDescription; }
-  bool has_description() const { return !mDescription.empty(); }
-
-
-  // --- value ---
-
-  virtual bool is_defined() const = 0;
-  bool is_undefined() const { return !is_defined(); }
-
-  virtual bool has_default() const  = 0;
-
-
-  // --- command line options ----
-
-  void set_cmd_line_options(const char* long_option, char short_option = 0)
-  {
-    mShortOption = short_option;
-    mLongOption  = long_option;
-  }
-
-  void set_short_option(char short_option) { mShortOption=short_option; }
-
-  void unsetCmdLineOption()
-  {
-    mShortOption = 0;
-    mLongOption  = NULL;
-  }
-
-  bool hasShortOption() const { return mShortOption!=0; }
-  char getShortOption() const { return mShortOption; }
-  bool hasLongOption() const { return true; } //mLongOption!=NULL; }
-  std::string getLongOption() const { return mLongOption ? std::string(mLongOption) : get_name(); }
-
-  virtual LIBDE265_API bool processCmdLineArguments(char** argv, int* argc, int idx) { return false; }
-
-
-
-  virtual std::string getTypeDescr() const = 0;
-
-  virtual std::string get_default_string() const { return "N/A"; }
-
- private:
-  std::string mPrefix;
-  std::string mIDName;
-
-  std::string mDescription;
-
-  char mShortOption;
-  const char* mLongOption;
-};
-
-
-
-class option_bool : public option_base
-{
-public:
-  option_bool() : value_set(false), default_set(false) { }
-
-  operator bool() const {
-    assert(value_set || default_set);
-    return value_set ? value : default_value;
-  }
-
-  virtual bool is_defined() const { return value_set || default_set; }
-  virtual bool has_default() const { return default_set; }
-
-  void set_default(bool v) { default_value=v; default_set=true; }
-  virtual std::string get_default_string() const { return default_value ? "true":"false"; }
-
-  virtual std::string getTypeDescr() const { return "(boolean)"; }
-  virtual LIBDE265_API bool processCmdLineArguments(char** argv, int* argc, int idx) { set(true); return true; }
-
-  bool set(bool v) { value_set=true; value=v; return true; }
-
- private:
-  bool value_set;
-  bool value;
-
-  bool default_set;
-  bool default_value;
-};
-
-
-class option_string : public option_base
-{
-public:
-  option_string() : value_set(false), default_set(false) { }
-
-  const option_string& operator=(std::string v) { value=v; value_set=true; return *this; }
-
-  operator std::string() const { return get(); }
-  std::string get() const {
-    assert(value_set || default_set);
-    return value_set ? value : default_value;
-  }
-
-  virtual bool is_defined() const { return value_set || default_set; }
-  virtual bool has_default() const { return default_set; }
-
-  void set_default(std::string v) { default_value=v; default_set=true; }
-  virtual LIBDE265_API std::string get_default_string() const { return default_value; }
-
-  virtual LIBDE265_API std::string getTypeDescr() const { return "(string)"; }
-  virtual LIBDE265_API bool processCmdLineArguments(char** argv, int* argc, int idx);
-
-  bool set(std::string v) { value_set=true; value=v; return true; }
-
- private:
-  bool value_set;
-  std::string value;
-
-  bool default_set;
-  std::string default_value;
-};
-
-
-class option_int : public option_base
-{
-public:
-  option_int() : value_set(false), default_set(false),
-    have_low_limit(false), have_high_limit(false) { }
-
-  void set_minimum(int mini) { have_low_limit =true; low_limit =mini; }
-  void set_maximum(int maxi) { have_high_limit=true; high_limit=maxi; }
-  void set_range(int mini,int maxi);
-  void set_valid_values(const std::vector<int>& v) { valid_values_set = v; }
-
-  const option_int& operator=(int v) { value=v; value_set=true; return *this; }
-
-  int operator() () const {
-    assert(value_set || default_set);
-    return value_set ? value : default_value;
-  }
-  operator int() const { return operator()(); }
-
-  virtual bool is_defined() const { return value_set || default_set; }
-  virtual bool has_default() const { return default_set; }
-
-  void set_default(int v) { default_value=v; default_set=true; }
-  virtual LIBDE265_API std::string get_default_string() const;
-
-  virtual LIBDE265_API std::string getTypeDescr() const;
-  virtual LIBDE265_API bool processCmdLineArguments(char** argv, int* argc, int idx);
-
-  bool set(int v) {
-    if (is_valid(v)) { value_set=true; value=v; return true; }
-    else { return false; }
-  }
-
- private:
-  bool value_set;
-  int value;
-
-  bool default_set;
-  int default_value;
-
-  bool have_low_limit, have_high_limit;
-  int  low_limit, high_limit;
-
-  std::vector<int> valid_values_set;
-
-  bool is_valid(int v) const;
-};
-
-
-
-class choice_option_base : public option_base
-{
-public:
-  choice_option_base() : choice_string_table(NULL) { }
-  ~choice_option_base() { delete choice_string_table; }
-
-  bool set(std::string v) { return set_value(v); }
-  virtual bool set_value(const std::string& val) = 0;
-  virtual std::vector<std::string> get_choice_names() const = 0;
-
-  virtual std::string getTypeDescr() const;
-  virtual LIBDE265_API bool processCmdLineArguments(char** argv, int* argc, int idx);
-
-  const char** get_choices_string_table() const;
-
- protected:
-  void invalidate_choices_string_table() {
-    delete choice_string_table;
-    choice_string_table = NULL;
-  }
-
- private:
-  mutable char* choice_string_table;
-};
-
-
-template <class T> class choice_option : public choice_option_base
-{
- public:
- choice_option() : default_set(false), value_set(false) { }
-
-  // --- initialization ---
-
-  void add_choice(const std::string& s, T id, bool default_value=false) {
-    choices.push_back( std::make_pair(s,id) );
-    if (default_value) {
-      defaultID = id;
-      defaultValue = s;
-      default_set = true;
-    }
-
-    invalidate_choices_string_table();
-  }
-
-  void set_default(T val) {
-    for (const auto& c : choices) {
-      if (c.second == val) {
-        defaultID = val;
-        defaultValue = c.first;
-        default_set = true;
-        return;
-      }
-    }
-
-    assert(false); // value does not exist
-  }
-
-
-  // --- usage ---
-
-  bool set_value(const std::string& val) // returns false if it is not a valid option
-  {
-    value_set = true;
-    selectedValue=val;
-
-    validValue = false;
-
-    for (const auto& c : choices) {
-      if (val == c.first) {
-        selectedID = c.second;
-        validValue = true;
-      }
-    }
-
-    return validValue;
-  }
-
-  bool isValidValue() const { return validValue; }
-
-  const std::string& getValue() const {
-    assert(value_set || default_set);
-    return value_set ? selectedValue : defaultValue;
-  }
-  void setID(T id) { selectedID=id; validValue=true; }
-  const T getID() const { return value_set ? selectedID : defaultID; }
-
-  virtual bool is_defined() const { return value_set || default_set; }
-  virtual bool has_default() const { return default_set; }
-
-  std::vector<std::string> get_choice_names() const
-  {
-    std::vector<std::string> names;
-    for (const auto& p : choices) {
-      names.push_back(p.first);
-    }
-    return names;
-  }
-
-  std::string get_default_string() const { return defaultValue; }
-
-  T operator() () const { return (T)getID(); }
-
- private:
-  std::vector< std::pair<std::string,T> > choices;
-
-  bool default_set;
-  std::string defaultValue;
-  T defaultID;
-
-  bool value_set;
-  std::string selectedValue;
-  T selectedID;
-
-  bool validValue;
-};
-
-
-
-
-class config_parameters
-{
- public:
- config_parameters() : param_string_table(NULL) { }
-  ~config_parameters() { delete param_string_table; }
-
-  void LIBDE265_API add_option(option_base* o);
-
-  void LIBDE265_API print_params() const;
-  bool LIBDE265_API parse_command_line_params(int* argc, char** argv, int* first_idx=NULL,
-                                 bool ignore_unknown_options=false);
-
-
-  // --- connection to C API ---
-
-  std::vector<std::string> get_parameter_IDs() const;
-  enum en265_parameter_type get_parameter_type(const char* param) const;
-
-  std::vector<std::string> get_parameter_choices(const char* param) const;
-
-  bool set_bool(const char* param, bool value);
-  bool set_int(const char* param, int value);
-  bool set_string(const char* param, const char* value);
-  bool set_choice(const char* param, const char* value);
-
-  const char** get_parameter_string_table() const;
-  const char** get_parameter_choices_table(const char* param) const;
-
- private:
-  std::vector<option_base*> mOptions;
-
-  option_base* find_option(const char* param) const;
-
-  mutable char* param_string_table;
-};
-
-#endif

libde265-1.0.17.tar.gz/libde265/encoder/encoder-context.cc Deleted

@@ -1,313 +0,0 @@
-/*
- * H.265 video codec.
- * Copyright (c) 2013-2014 struktur AG, Dirk Farin <farin@struktur.de>
- *
- * Authors: Dirk Farin <farin@struktur.de>
- *
- * This file is part of libde265.
- *
- * libde265 is free software: you can redistribute it and/or modify
- * it under the terms of the GNU Lesser General Public License as
- * published by the Free Software Foundation, either version 3 of
- * the License, or (at your option) any later version.
- *
- * libde265 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 Lesser General Public License for more details.
- *
- * You should have received a copy of the GNU Lesser General Public License
- * along with libde265.  If not, see <http://www.gnu.org/licenses/>.
- */
-
-#include "encoder/encoder-context.h"
-#include "libde265/util.h"
-
-#include <math.h>
-
-
-encoder_context::encoder_context()
-{
-  encoder_started=false;
-
-  vps = std::make_shared<video_parameter_set>();
-  sps = std::make_shared<seq_parameter_set>();
-  pps = std::make_shared<pic_parameter_set>();
-
-  //img_source = NULL;
-  //reconstruction_sink = NULL;
-  //packet_sink = NULL;
-
-  image_spec_is_defined = false;
-  parameters_have_been_set = false;
-  headers_have_been_sent = false;
-
-  param_image_allocation_userdata = NULL;
-  //release_func = NULL;
-
-  use_adaptive_context = true; //false;
-
-  //enc_coeff_pool.set_blk_size(64*64*20); // TODO: this a guess
-
-  //switch_CABAC_to_bitstream();
-
-
-  params.registerParams(params_config);
-  algo.registerParams(params_config);
-}
-
-
-encoder_context::~encoder_context()
-{
-  while (!output_packets.empty()) {
-    en265_free_packet((en265_encoder_context*)this, output_packets.front());
-    output_packets.pop_front();
-  }
-}
-
-
-void encoder_context::start_encoder()
-{
-  if (encoder_started) {
-    return;
-  }
-
-
-  if (params.sop_structure() == SOP_Intra) {
-    sop = std::shared_ptr<sop_creator_intra_only>(new sop_creator_intra_only());
-  }
-  else {
-    auto s = std::shared_ptr<sop_creator_trivial_low_delay>(new sop_creator_trivial_low_delay());
-    s->setParams(params.mSOP_LowDelay);
-    sop = s;
-  }
-
-  sop->set_encoder_context(this);
-  sop->set_encoder_picture_buffer(&picbuf);
-
-
-  encoder_started=true;
-}
-
-
-en265_packet* encoder_context::create_packet(en265_packet_content_type t)
-{
-  en265_packet* pck = new en265_packet;
-
-  uint8_t* data = new uint8_tcabac_encoder.size();
-  memcpy(data, cabac_encoder.data(), cabac_encoder.size());
-
-  pck->version = 1;
-
-  pck->data = data;
-  pck->length = cabac_encoder.size();
-
-  pck->frame_number = -1;
-  pck->content_type = t;
-  pck->complete_picture = 0;
-  pck->final_slice = 0;
-  pck->dependent_slice = 0;
-  //pck->pts = 0;
-  //pck->user_data = NULL;
-  pck->nuh_layer_id = 0;
-  pck->nuh_temporal_id = 0;
-
-  pck->encoder_context = (en265_encoder_context*)this;
-
-  pck->input_image = NULL;
-  pck->reconstruction = NULL;
-
-  cabac_encoder.reset();
-
-  return pck;
-}
-
-
-de265_error encoder_context::encode_headers()
-{
-  nal_header nal;
-
-  // VPS
-
-  vps->set_defaults(Profile_Main, 6,2);
-
-
-  // SPS
-
-  sps->set_defaults();
-  sps->set_CB_log2size_range( Log2(params.min_cb_size), Log2(params.max_cb_size));
-  sps->set_TB_log2size_range( Log2(params.min_tb_size), Log2(params.max_tb_size));
-  sps->max_transform_hierarchy_depth_intra = params.max_transform_hierarchy_depth_intra;
-  sps->max_transform_hierarchy_depth_inter = params.max_transform_hierarchy_depth_inter;
-
-  if (imgdata->input->get_chroma_format() == de265_chroma_444) {
-    sps->chroma_format_idc = CHROMA_444;
-  }
-
-  sps->set_resolution(image_width, image_height);
-  sop->set_SPS_header_values();
-  de265_error err = sps->compute_derived_values(true);
-  if (err != DE265_OK) {
-    fprintf(stderr,"invalid SPS parameters\n");
-    exit(10);
-  }
-
-
-  // PPS
-
-  pps->set_defaults();
-  pps->sps = sps; //sps.get();
-  pps->pic_init_qp = algo.getPPS_QP();
-
-  // turn off deblocking filter
-  pps->deblocking_filter_control_present_flag = true;
-  pps->deblocking_filter_override_enabled_flag = false;
-  pps->pic_disable_deblocking_filter_flag = true;
-  pps->pps_loop_filter_across_slices_enabled_flag = false;
-
-  pps->set_derived_values(sps.get());
-
-
-
-  // write headers
-
-  en265_packet* pck;
-
-  nal.set(NAL_UNIT_VPS_NUT);
-  nal.write(cabac_encoder);
-  vps->write(this, cabac_encoder);
-  cabac_encoder.add_trailing_bits();
-  cabac_encoder.flush_VLC();
-  pck = create_packet(EN265_PACKET_VPS);
-  pck->nal_unit_type = EN265_NUT_VPS;
-  output_packets.push_back(pck);
-
-  nal.set(NAL_UNIT_SPS_NUT);
-  nal.write(cabac_encoder);
-  sps->write(this, cabac_encoder);
-  cabac_encoder.add_trailing_bits();
-  cabac_encoder.flush_VLC();
-  pck = create_packet(EN265_PACKET_SPS);
-  pck->nal_unit_type = EN265_NUT_SPS;
-  output_packets.push_back(pck);
-
-  nal.set(NAL_UNIT_PPS_NUT);
-  nal.write(cabac_encoder);
-  pps->write(this, cabac_encoder, sps.get());
-  cabac_encoder.add_trailing_bits();
-  cabac_encoder.flush_VLC();
-  pck = create_packet(EN265_PACKET_PPS);
-  pck->nal_unit_type = EN265_NUT_PPS;
-  output_packets.push_back(pck);
-
-
-
-  headers_have_been_sent = true;
-
-  return DE265_OK;
-}
-
-
-de265_error encoder_context::encode_picture_from_input_buffer()
-{
-  if (!picbuf.have_more_frames_to_encode()) {
-    return DE265_OK;
-  }
-
-
-  if (!image_spec_is_defined) {
-    const image_data* id = picbuf.peek_next_picture_to_encode();
-    image_width  = id->input->get_width();
-    image_height = id->input->get_height();
-    image_spec_is_defined = true;
-
-    ctbs.alloc(image_width, image_height, Log2(params.max_cb_size));
-  }
-
-
-  if (!parameters_have_been_set) {
-    algo.setParams(params);
-
-
-    // TODO: must be <30, because Y->C mapping (tab8_22) is not implemented yet
-    int qp = algo.getPPS_QP();
-
-    //lambda = ectx->params.lambda;
-    lambda = 0.0242 * pow(1.27245, qp);
-
-    parameters_have_been_set = true;
-  }
-
-
-
-
-
-  image_data* imgdata;
-  imgdata = picbuf.get_next_picture_to_encode();
-  assert(imgdata);
-  picbuf.mark_encoding_started(imgdata->frame_number);
-
-  this->imgdata = imgdata;
-  this->shdr    = &imgdata->shdr;
-  loginfo(LogEncoder,"encoding frame %d\n",imgdata->frame_number);
-
-
-  // write headers if not written yet
-
-  if (!headers_have_been_sent) {
-    encode_headers();
-  }
-
-
-  // write slice header
-
-  // slice
-
-  imgdata->shdr.slice_deblocking_filter_disabled_flag = true;
-  imgdata->shdr.slice_loop_filter_across_slices_enabled_flag = false;
-  imgdata->shdr.compute_derived_values(pps.get());
-
-  imgdata->shdr.pps = pps;
-
-  //shdr.slice_pic_order_cnt_lsb = poc & 0xFF;
-
-  imgdata->nal.write(cabac_encoder);
-  imgdata->shdr.write(this, cabac_encoder, sps.get(), pps.get(), imgdata->nal.nal_unit_type);
-  cabac_encoder.add_trailing_bits();
-  cabac_encoder.flush_VLC();
-
-
-  // encode image
-
-  cabac_encoder.init_CABAC();
-  double psnr = encode_image(this,imgdata->input, algo);
-  loginfo(LogEncoder,"  PSNR-Y: %f\n", psnr);
-  cabac_encoder.flush_CABAC();
-  cabac_encoder.add_trailing_bits();
-  cabac_encoder.flush_VLC();
-
-
-  // set reconstruction image
-
-  picbuf.set_reconstruction_image(imgdata->frame_number, img);
-  //picbuf.set_prediction_image(imgdata->frame_number, prediction);
-  img=NULL;
-  this->imgdata = NULL;
-  this->shdr = NULL;
-
-  // build output packet
-
-  en265_packet* pck = create_packet(EN265_PACKET_SLICE);
-  pck->input_image    = imgdata->input;
-  pck->reconstruction = imgdata->reconstruction;
-  pck->frame_number   = imgdata->frame_number;
-  pck->nal_unit_type  = (enum en265_nal_unit_type)imgdata->nal.nal_unit_type;
-  pck->nuh_layer_id   = imgdata->nal.nuh_layer_id;
-  pck->nuh_temporal_id= imgdata->nal.nuh_temporal_id;
-  output_packets.push_back(pck);
-
-
-  picbuf.mark_encoding_finished(imgdata->frame_number);
-
-  return DE265_OK;
-}

libde265-1.0.17.tar.gz/libde265/encoder/encoder-context.h Deleted

@@ -1,173 +0,0 @@
-/*
- * H.265 video codec.
- * Copyright (c) 2013-2014 struktur AG, Dirk Farin <farin@struktur.de>
- *
- * Authors: Dirk Farin <farin@struktur.de>
- *
- * This file is part of libde265.
- *
- * libde265 is free software: you can redistribute it and/or modify
- * it under the terms of the GNU Lesser General Public License as
- * published by the Free Software Foundation, either version 3 of
- * the License, or (at your option) any later version.
- *
- * libde265 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 Lesser General Public License for more details.
- *
- * You should have received a copy of the GNU Lesser General Public License
- * along with libde265.  If not, see <http://www.gnu.org/licenses/>.
- */
-
-#ifndef ENCODER_CONTEXT_H
-#define ENCODER_CONTEXT_H
-
-#include "libde265/image.h"
-#include "libde265/decctx.h"
-#include "libde265/image-io.h"
-#include "libde265/encoder/encoder-params.h"
-#include "libde265/encoder/encpicbuf.h"
-#include "libde265/encoder/sop.h"
-#include "libde265/en265.h"
-#include "libde265/util.h"
-
-#include <memory>
-
-
-class encoder_context : public base_context
-{
- public:
-  encoder_context();
-  ~encoder_context();
-
-  virtual const de265_image* get_image(uint16_t frame_id) const {
-    return picbuf.get_picture(frame_id)->reconstruction;
-  }
-
-  virtual bool has_image(uint16_t frame_id) const {
-    return picbuf.has_picture(frame_id);
-  }
-
-  bool encoder_started;
-
-  encoder_params params;
-  config_parameters params_config;
-
-  EncoderCore_Custom algo;
-
-  int image_width, image_height;
-  bool image_spec_is_defined;  // whether we know the input image size
-
-  void* param_image_allocation_userdata;
-  /*
-  void (*release_func)(en265_encoder_context*,
-                       de265_image*,
-                       void* userdata);
-  */
-
-  //error_queue errqueue;
-  //acceleration_functions accel;
-
-  // quick links
-  de265_image* img; // reconstruction
-  //de265_image* prediction;
-  image_data* imgdata; // input image
-  slice_segment_header* shdr;
-
-  CTBTreeMatrix ctbs;
-
-  // temporary memory for motion compensated pixels (when CB-algo passes this down to TB-algo)
-  //uint8_t prediction364*64; // stride: 1<<(cb->log2Size)
-  //int prediction_x0,prediction_y0;
-
-
-  int active_qp; // currently active QP
-  /*int target_qp;*/ /* QP we want to code at.
-                     (Not actually the real QP. Check image.get_QPY() for that.) */
-
-  const seq_parameter_set& get_sps() const { return *sps; }
-  const pic_parameter_set& get_pps() const { return *pps; }
-
-  seq_parameter_set& get_sps() { return *sps; }
-  pic_parameter_set& get_pps() { return *pps; }
-
-  std::shared_ptr<video_parameter_set>& get_shared_vps() { return vps; }
-  std::shared_ptr<seq_parameter_set>& get_shared_sps() { return sps; }
-  std::shared_ptr<pic_parameter_set>& get_shared_pps() { return pps; }
-
- private:
-  std::shared_ptr<video_parameter_set>  vps;
-  std::shared_ptr<seq_parameter_set>    sps;
-  std::shared_ptr<pic_parameter_set>    pps;
-  //slice_segment_header shdr;
-
- public:
-  bool parameters_have_been_set;
-  bool headers_have_been_sent;
-
-  encoder_picture_buffer picbuf;
-  std::shared_ptr<sop_creator> sop;
-
-  std::deque<en265_packet*> output_packets;
-
-
-  // --- rate-control ---
-
-  float lambda;
-
-
-  // --- CABAC output and rate estimation ---
-
-  //CABAC_encoder*  cabac;      // currently active CABAC output (estim or bitstream)
-  //context_model_table2* ctx_model;  // currently active ctx models (estim or bitstream)
-
-  // CABAC bitstream writer
-  CABAC_encoder_bitstream cabac_encoder;
-  context_model_table     cabac_ctx_models;
-
-  //std::shared_ptr<CABAC_encoder> cabac_estim;
-
-  bool use_adaptive_context;
-
-
-  /*** TODO: CABAC_encoder direkt an encode-Funktion übergeben, anstatt hier
-       aussenrum zwischenzuspeichern (mit undefinierter Lifetime).
-       Das Context-Model kann dann gleich mit in den Encoder rein cabac_encoder(ctxtable).
-       write_bits() wird dann mit dem context-index aufgerufen, nicht mit dem model direkt.
-  ***/
-
-
-  /*
-  void switch_CABAC(context_model_table2* model) {
-    cabac      = cabac_estim.get();
-    ctx_model  = model;
-  }
-
-  void switch_CABAC_to_bitstream() {
-    cabac     = &cabac_bitstream;
-    ctx_model = &ctx_model_bitstream;
-  }
-  */
-
-  en265_packet* create_packet(en265_packet_content_type t);
-
-
-  // --- encoding control ---
-
-  void start_encoder();
-  de265_error encode_headers();
-  de265_error encode_picture_from_input_buffer();
-
-
-  // Input images can be released after encoding and when the output packet is released.
-  // This is important to do as soon as possible, as the image might actually wrap
-  // scarce resources like camera picture buffers.
-  // This function does release (only) the raw input data.
-  void release_input_image(int frame_number) { picbuf.release_input_image(frame_number); }
-
-  void mark_image_is_outputted(int frame_number) { picbuf.mark_image_is_outputted(frame_number); }
-};
-
-
-#endif

libde265-1.0.17.tar.gz/libde265/encoder/encoder-core.cc Deleted

@@ -1,428 +0,0 @@
-/*
- * H.265 video codec.
- * Copyright (c) 2013-2014 struktur AG, Dirk Farin <farin@struktur.de>
- *
- * Authors: struktur AG, Dirk Farin <farin@struktur.de>
- *
- * This file is part of libde265.
- *
- * libde265 is free software: you can redistribute it and/or modify
- * it under the terms of the GNU Lesser General Public License as
- * published by the Free Software Foundation, either version 3 of
- * the License, or (at your option) any later version.
- *
- * libde265 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 Lesser General Public License for more details.
- *
- * You should have received a copy of the GNU Lesser General Public License
- * along with libde265.  If not, see <http://www.gnu.org/licenses/>.
- */
-
-
-#include "libde265/encoder/encoder-core.h"
-#include "libde265/encoder/encoder-context.h"
-#include "libde265/encoder/encoder-syntax.h"
-#include <assert.h>
-#include <limits>
-#include <math.h>
-#include <iostream>
-#include <fstream>
-
-
-#define ENCODER_DEVELOPMENT 0
-#define COMPARE_ESTIMATED_RATE_TO_REAL_RATE 0
-
-
-static int IntraPredModeCnt735;
-static int MPM_used735;
-
-static int IntraPredModeCnt_total35;
-static int MPM_used_total35;
-
-/*
-void statistics_IntraPredMode(const encoder_context* ectx, int x,int y, const enc_cb* cb)
-{
-  if (cb->split_cu_flag) {
-    for (int i=0;i<4;i++)
-      if (cb->childreni) {
-        statistics_IntraPredMode(ectx, childX(x,i,cb->log2Size), childY(y,i,cb->log2Size), cb->childreni);
-      }
-  }
-  else {
-    int cnt;
-    int size = cb->log2Size;
-
-    if (cb->PartMode == PART_NxN) { cnt=4; size--; } else cnt=1;
-
-    for (int i=0;i<cnt;i++) {
-      IntraPredModeCntsize cb->intra.pred_modei ++;
-      IntraPredModeCnt_total cb->intra.pred_modei ++;
-
-      int xi = childX(x,i,cb->log2Size);
-      int yi = childY(y,i,cb->log2Size);
-
-      enum IntraPredMode candModeList3;
-      fillIntraPredModeCandidates(candModeList,xi,yi, xi>0, yi>0, ectx->img);
-
-      int predmode = cb->intra.pred_modei;
-      if (candModeList0==predmode ||
-          candModeList1==predmode ||
-          candModeList2==predmode) {
-        MPM_usedsizepredmode++;
-        MPM_used_totalpredmode++;
-      }
-    }
-  }
-}
-*/
-
-void statistics_print()
-{
-  for (int i=0;i<35;i++) {
-    printf("%d",i);
-    printf("  %d %d",IntraPredModeCnt_totali, MPM_used_totali);
-
-    for (int k=2;k<=6;k++) {
-      printf("  %d %d",IntraPredModeCntki, MPM_usedki);
-    }
-
-    printf("\n");
-  }
-}
-
-
-void print_tb_tree_rates(const enc_tb* tb, int level)
-{
-  for (int i=0;i<level;i++)
-    std::cout << "  ";
-
-  std::cout << "TB rate=" << tb->rate << " (" << tb->rate_withoutCbfChroma << ")\n";
-  if (tb->split_transform_flag) {
-    for (int i=0;i<4;i++)
-      print_tb_tree_rates(tb->childreni, level+1);
-  }
-}
-
-
-void print_cb_tree_rates(const enc_cb* cb, int level)
-{
-  for (int i=0;i<level;i++)
-    std::cout << "  ";
-
-  std::cout << "CB rate=" << cb->rate << "\n";
-  if (cb->split_cu_flag) {
-    for (int i=0;i<4;i++)
-      print_cb_tree_rates(cb->childreni, level+1);
-  }
-  else {
-    print_tb_tree_rates(cb->transform_tree, level+1);
-  }
-}
-
-
-// /*LIBDE265_API*/ ImageSink_YUV reconstruction_sink;
-
-double encode_image(encoder_context* ectx,
-                    const de265_image* input,
-                    EncoderCore& algo)
-{
-  int stride=input->get_image_stride(0);
-
-  int w = ectx->get_sps().pic_width_in_luma_samples;
-  int h = ectx->get_sps().pic_height_in_luma_samples;
-
-  // --- create reconstruction image ---
-  ectx->img = new de265_image;
-  ectx->img->set_headers(ectx->get_shared_vps(), ectx->get_shared_sps(), ectx->get_shared_pps());
-  ectx->img->PicOrderCntVal = input->PicOrderCntVal;
-
-  ectx->img->alloc_image(w,h, input->get_chroma_format(), ectx->get_shared_sps(), true,
-                         NULL /* no decctx */, /*ectx,*/ 0,NULL,false);
-  //ectx->img->alloc_encoder_data(&ectx->sps);
-  ectx->img->clear_metadata();
-
-#if 0
-  if (1) {
-    ectx->prediction = new de265_image;
-    ectx->prediction->alloc_image(w,h, input->get_chroma_format(), &ectx->sps, false /* no metadata */,
-                                  NULL /* no decctx */, NULL /* no encctx */, 0,NULL,false);
-    ectx->prediction->vps = ectx->vps;
-    ectx->prediction->sps = ectx->sps;
-    ectx->prediction->pps = ectx->pps;
-  }
-#endif
-
-  ectx->active_qp = ectx->get_pps().pic_init_qp; // TODO take current qp from slice
-
-
-  ectx->cabac_ctx_models.init(ectx->shdr->initType, ectx->shdr->SliceQPY);
-  ectx->cabac_encoder.set_context_models(&ectx->cabac_ctx_models);
-
-
-  context_model_table modelEstim;
-  CABAC_encoder_estim cabacEstim;
-
-  modelEstim.init(ectx->shdr->initType, ectx->shdr->SliceQPY);
-  cabacEstim.set_context_models(&modelEstim);
-
-
-  int Log2CtbSize = ectx->get_sps().Log2CtbSizeY;
-
-  uint8_t* luma_plane = ectx->img->get_image_plane(0);
-  uint8_t* cb_plane   = ectx->img->get_image_plane(1);
-  uint8_t* cr_plane   = ectx->img->get_image_plane(2);
-
-  double mse=0;
-
-
-  // encode CTB by CTB
-
-  ectx->ctbs.clear();
-
-  for (int y=0;y<ectx->get_sps().PicHeightInCtbsY;y++)
-    for (int x=0;x<ectx->get_sps().PicWidthInCtbsY;x++)
-      {
-        ectx->img->set_SliceAddrRS(x, y, ectx->shdr->SliceAddrRS);
-
-        int x0 = x<<Log2CtbSize;
-        int y0 = y<<Log2CtbSize;
-
-        logtrace(LogSlice,"encode CTB at %d %d\n",x0,y0);
-
-        // make a copy of the context model that we can modify for testing alternatives
-
-        context_model_table ctxModel;
-        //copy_context_model_table(ctxModel, ectx->ctx_model_bitstream);
-        ctxModel = ectx->cabac_ctx_models.copy();
-        ctxModel = modelEstim.copy(); // TODO TMP
-
-        disable_logging(LogSymbols);
-        enable_logging(LogSymbols);  // TODO TMP
-
-        //printf("================================================== ANALYZE\n");
-
-#if 1
-        /*
-          enc_cb* cb = encode_cb_may_split(ectx, ctxModel,
-          input, x0,y0, Log2CtbSize, 0, qp);
-        */
-
-        enc_cb* cb = algo.getAlgoCTBQScale()->analyze(ectx,ctxModel, x0,y0);
-#else
-        float minCost = std::numeric_limits<float>::max();
-        int bestQ = 0;
-        int qp = ectx->params.constant_QP;
-
-        enc_cb* cb;
-        for (int q=1;q<51;q++) {
-          copy_context_model_table(ctxModel, ectx->ctx_model_bitstream);
-
-          enc_cb* cbq = encode_cb_may_split(ectx, ctxModel,
-                                            input, x0,y0, Log2CtbSize, 0, q);
-
-          float cost = cbq->distortion + ectx->lambda * cbq->rate;
-          if (cost<minCost) { minCost=cost; bestQ=q; }
-
-          if (q==qp) { cb=cbq; }
-        }
-
-        printf("Q %d\n",bestQ);
-        fflush(stdout);
-#endif
-
-        //print_cb_tree_rates(cb,0);
-
-        //statistics_IntraPredMode(ectx, x0,y0, cb);
-
-
-        // --- write bitstream ---
-
-        //ectx->switch_CABAC_to_bitstream();
-
-        enable_logging(LogSymbols);
-
-        logdebug(LogEncoder,"write CTB %d;%d\n",x,y);
-
-        if (logdebug_enabled(LogEncoder)) {
-          cb->debug_dumpTree(enc_tb::DUMPTREE_ALL);
-        }
-
-        /*
-        cb->debug_assertTreeConsistency(ectx->img);
-
-        //cb->invalidateMetadataInSubTree(ectx->img);
-        cb->writeMetadata(ectx, ectx->img,
-                          enc_node::METADATA_INTRA_MODES |
-                          enc_node::METADATA_RECONSTRUCTION |
-                          enc_node::METADATA_CT_DEPTH);
-
-        cb->debug_assertTreeConsistency(ectx->img);
-        */
-
-        encode_ctb(ectx, &ectx->cabac_encoder, cb, x,y);
-
-        //printf("================================================== WRITE\n");
-
-
-        if (COMPARE_ESTIMATED_RATE_TO_REAL_RATE) {
-          float realPre = cabacEstim.getRDBits();
-          encode_ctb(ectx, &cabacEstim, cb, x,y);
-          float realPost = cabacEstim.getRDBits();
-
-          printf("estim: %f  real: %f  diff: %f\n",
-                 cb->rate,
-                 realPost-realPre,
-                 cb->rate - (realPost-realPre));
-        }
-
-
-        int last = (y==ectx->get_sps().PicHeightInCtbsY-1 &&
-                    x==ectx->get_sps().PicWidthInCtbsY-1);
-        ectx->cabac_encoder.write_CABAC_term_bit(last);
-
-        //delete cb;
-
-        //ectx->free_all_pools();
-
-        mse += cb->distortion;
-      }
-
-  mse /= ectx->img->get_width() * ectx->img->get_height();
-
-
-  //reconstruction_sink.send_image(ectx->img);
-
-
-  //statistics_print();
-
-
-  //delete ectx->prediction;
-
-
-  // frame PSNR
-
-  ectx->ctbs.writeReconstructionToImage(ectx->img, &ectx->get_sps());
-
-#if 0
-  std::ofstream ostr("out.pgm");
-  ostr << "P5\n" << ectx->img->get_width() << " " << ectx->img->get_height() << "\n255\n";
-  for (int y=0;y<ectx->img->get_height();y++) {
-    ostr.write( (char*)ectx->img->get_image_plane_at_pos(0,0,y), ectx->img->get_width() );
-  }
-#endif
-
-  double psnr = 10*log10(255.0*255.0 / mse);
-
-#if 0
-  double psnr2 = PSNR(MSE(input->get_image_plane(0), input->get_image_stride(0),
-                          luma_plane, ectx->img->get_image_stride(0),
-                          input->get_width(), input->get_height()));
-
-  printf("rate-estim PSNR: %f vs %f\n",psnr,psnr2);
-#endif
-
-  return psnr;
-}
-
-
-
-void EncoderCore_Custom::setParams(encoder_params& params)
-{
-  // build algorithm tree
-
-  mAlgo_CTB_QScale_Constant.setChildAlgo(&mAlgo_CB_Split_BruteForce);
-  mAlgo_CB_Split_BruteForce.setChildAlgo(&mAlgo_CB_Skip_BruteForce);
-
-  mAlgo_CB_Skip_BruteForce.setSkipAlgo(&mAlgo_CB_MergeIndex_Fixed);
-  mAlgo_CB_Skip_BruteForce.setNonSkipAlgo(&mAlgo_CB_IntraInter_BruteForce);
-  //&mAlgo_CB_InterPartMode_Fixed);
-
-  Algo_CB_IntraPartMode* algo_CB_IntraPartMode = NULL;
-  switch (params.mAlgo_CB_IntraPartMode()) {
-  case ALGO_CB_IntraPartMode_BruteForce:
-    algo_CB_IntraPartMode = &mAlgo_CB_IntraPartMode_BruteForce;
-    break;
-  case ALGO_CB_IntraPartMode_Fixed:
-    algo_CB_IntraPartMode = &mAlgo_CB_IntraPartMode_Fixed;
-    break;
-  }
-
-  mAlgo_CB_IntraInter_BruteForce.setIntraChildAlgo(algo_CB_IntraPartMode);
-  mAlgo_CB_IntraInter_BruteForce.setInterChildAlgo(&mAlgo_CB_InterPartMode_Fixed);
-
-  mAlgo_CB_MergeIndex_Fixed.setChildAlgo(&mAlgo_TB_Split_BruteForce);
-
-  Algo_PB_MV* pbAlgo = NULL;
-  switch (params.mAlgo_MEMode()) {
-  case MEMode_Test:
-    pbAlgo = &mAlgo_PB_MV_Test;
-    break;
-  case MEMode_Search:
-    pbAlgo = &mAlgo_PB_MV_Search;
-    break;
-  }
-
-  mAlgo_CB_InterPartMode_Fixed.setChildAlgo(pbAlgo);
-  pbAlgo->setChildAlgo(&mAlgo_TB_Split_BruteForce);
-
-
-  Algo_TB_IntraPredMode_ModeSubset* algo_TB_IntraPredMode = NULL;
-  switch (params.mAlgo_TB_IntraPredMode()) {
-  case ALGO_TB_IntraPredMode_BruteForce:
-    algo_TB_IntraPredMode = &mAlgo_TB_IntraPredMode_BruteForce;
-    break;
-  case ALGO_TB_IntraPredMode_FastBrute:
-    algo_TB_IntraPredMode = &mAlgo_TB_IntraPredMode_FastBrute;
-    break;
-  case ALGO_TB_IntraPredMode_MinResidual:
-    algo_TB_IntraPredMode = &mAlgo_TB_IntraPredMode_MinResidual;
-    break;
-  }
-
-  algo_CB_IntraPartMode->setChildAlgo(algo_TB_IntraPredMode);
-
-  mAlgo_TB_Split_BruteForce.setAlgo_TB_IntraPredMode(algo_TB_IntraPredMode);
-  mAlgo_TB_Split_BruteForce.setAlgo_TB_Residual(&mAlgo_TB_Transform);
-
-  Algo_TB_RateEstimation* algo_TB_RateEstimation = NULL;
-  switch (params.mAlgo_TB_RateEstimation()) {
-  case ALGO_TB_RateEstimation_None:  algo_TB_RateEstimation = &mAlgo_TB_RateEstimation_None;  break;
-  case ALGO_TB_RateEstimation_Exact: algo_TB_RateEstimation = &mAlgo_TB_RateEstimation_Exact; break;
-  }
-  mAlgo_TB_Transform.setAlgo_TB_RateEstimation(algo_TB_RateEstimation);
-  //mAlgo_TB_Split_BruteForce.setParams(params.TB_Split_BruteForce);
-
-  algo_TB_IntraPredMode->setChildAlgo(&mAlgo_TB_Split_BruteForce);
-
-
-  // ===== set algorithm parameters ======
-
-  //mAlgo_CB_IntraPartMode_Fixed.setParams(params.CB_IntraPartMode_Fixed);
-
-  //mAlgo_TB_IntraPredMode_FastBrute.setParams(params.TB_IntraPredMode_FastBrute);
-  //mAlgo_TB_IntraPredMode_MinResidual.setParams(params.TB_IntraPredMode_MinResidual);
-
-
-  //mAlgo_CTB_QScale_Constant.setParams(params.CTB_QScale_Constant);
-
-
-  algo_TB_IntraPredMode->enableIntraPredModeSubset( params.mAlgo_TB_IntraPredMode_Subset() );
-}
-
-
-void Logging::print_logging(const encoder_context* ectx, const char* id, const char* filename)
-{
-#if 000
-  if (strcmp(id,logging_tb_split.name())==0) {
-    logging_tb_split.print(ectx,filename);
-  }
-#endif
-}
-
-
-void en265_print_logging(const encoder_context* ectx, const char* id, const char* filename)
-{
-  Logging::print_logging(ectx,id,filename);
-}

libde265-1.0.17.tar.gz/libde265/encoder/encoder-core.h Deleted

@@ -1,151 +0,0 @@
-/*
- * H.265 video codec.
- * Copyright (c) 2013-2014 struktur AG, Dirk Farin <farin@struktur.de>
- *
- * Authors: Dirk Farin <farin@struktur.de>
- *
- * This file is part of libde265.
- *
- * libde265 is free software: you can redistribute it and/or modify
- * it under the terms of the GNU Lesser General Public License as
- * published by the Free Software Foundation, either version 3 of
- * the License, or (at your option) any later version.
- *
- * libde265 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 Lesser General Public License for more details.
- *
- * You should have received a copy of the GNU Lesser General Public License
- * along with libde265.  If not, see <http://www.gnu.org/licenses/>.
- */
-
-#ifndef ANALYZE_H
-#define ANALYZE_H
-
-#include "libde265/nal-parser.h"
-#include "libde265/decctx.h"
-#include "libde265/encoder/encoder-types.h"
-#include "libde265/slice.h"
-#include "libde265/scan.h"
-#include "libde265/intrapred.h"
-#include "libde265/transform.h"
-#include "libde265/fallback-dct.h"
-#include "libde265/quality.h"
-#include "libde265/fallback.h"
-#include "libde265/encoder/configparam.h"
-
-#include "libde265/encoder/algo/tb-intrapredmode.h"
-#include "libde265/encoder/algo/tb-transform.h"
-#include "libde265/encoder/algo/tb-split.h"
-#include "libde265/encoder/algo/cb-intrapartmode.h"
-#include "libde265/encoder/algo/cb-interpartmode.h"
-#include "libde265/encoder/algo/cb-split.h"
-#include "libde265/encoder/algo/ctb-qscale.h"
-#include "libde265/encoder/algo/cb-mergeindex.h"
-//#include "libde265/encoder/algo/cb-skip-or-inter.h"
-#include "libde265/encoder/algo/pb-mv.h"
-#include "libde265/encoder/algo/cb-skip.h"
-#include "libde265/encoder/algo/cb-intra-inter.h"
-
-
-/*  Encoder search tree, bottom up:
-
-    - Algo_TB_Split - whether TB is split or not
-
-    - Algo_TB_IntraPredMode - choose the intra prediction mode (or NOP, if at the wrong tree level)
-
-    - Algo_CB_IntraPartMode - choose between NxN and 2Nx2N intra parts
-
-    - Algo_CB_PredMode - intra / inter
-
-    - Algo_CB_Split - whether CB is split or not
-
-    - Algo_CTB_QScale - select QScale on CTB granularity
- */
-
-
-// ========== an encoding algorithm combines a set of algorithm modules ==========
-
-class EncoderCore
-{
- public:
-  virtual ~EncoderCore() { }
-
-  virtual Algo_CTB_QScale* getAlgoCTBQScale() = 0;
-
-  virtual int getPPS_QP() const = 0;
-  virtual int getSlice_QPDelta() const { return 0; }
-};
-
-
-class EncoderCore_Custom : public EncoderCore
-{
- public:
-
-  void setParams(struct encoder_params& params);
-
-  void registerParams(config_parameters& config) {
-    mAlgo_CTB_QScale_Constant.registerParams(config);
-    mAlgo_CB_IntraPartMode_Fixed.registerParams(config);
-    mAlgo_CB_InterPartMode_Fixed.registerParams(config);
-    mAlgo_PB_MV_Test.registerParams(config);
-    mAlgo_PB_MV_Search.registerParams(config);
-    mAlgo_TB_IntraPredMode_FastBrute.registerParams(config);
-    mAlgo_TB_IntraPredMode_MinResidual.registerParams(config);
-    mAlgo_TB_Split_BruteForce.registerParams(config);
-  }
-
-  virtual Algo_CTB_QScale* getAlgoCTBQScale() { return &mAlgo_CTB_QScale_Constant; }
-
-  virtual int getPPS_QP() const { return mAlgo_CTB_QScale_Constant.getQP(); }
-
- private:
-  Algo_CTB_QScale_Constant         mAlgo_CTB_QScale_Constant;
-
-  Algo_CB_Split_BruteForce         mAlgo_CB_Split_BruteForce;
-  Algo_CB_Skip_BruteForce          mAlgo_CB_Skip_BruteForce;
-  Algo_CB_IntraInter_BruteForce    mAlgo_CB_IntraInter_BruteForce;
-
-  Algo_CB_IntraPartMode_BruteForce mAlgo_CB_IntraPartMode_BruteForce;
-  Algo_CB_IntraPartMode_Fixed      mAlgo_CB_IntraPartMode_Fixed;
-
-  Algo_CB_InterPartMode_Fixed      mAlgo_CB_InterPartMode_Fixed;
-  Algo_CB_MergeIndex_Fixed         mAlgo_CB_MergeIndex_Fixed;
-
-  Algo_PB_MV_Test                  mAlgo_PB_MV_Test;
-  Algo_PB_MV_Search                mAlgo_PB_MV_Search;
-
-  Algo_TB_Split_BruteForce          mAlgo_TB_Split_BruteForce;
-
-  Algo_TB_IntraPredMode_BruteForce  mAlgo_TB_IntraPredMode_BruteForce;
-  Algo_TB_IntraPredMode_FastBrute   mAlgo_TB_IntraPredMode_FastBrute;
-  Algo_TB_IntraPredMode_MinResidual mAlgo_TB_IntraPredMode_MinResidual;
-
-  Algo_TB_Transform                 mAlgo_TB_Transform;
-  Algo_TB_RateEstimation_None       mAlgo_TB_RateEstimation_None;
-  Algo_TB_RateEstimation_Exact      mAlgo_TB_RateEstimation_Exact;
-};
-
-
-
-double encode_image(encoder_context*, const de265_image* input, EncoderCore&);
-
-void encode_sequence(encoder_context*);
-
-
-class Logging
-{
-public:
-  virtual ~Logging() { }
-
-  static void print_logging(const encoder_context* ectx, const char* id, const char* filename);
-
-  virtual const char* name() const = 0;
-  virtual void print(const encoder_context* ectx, const char* filename) = 0;
-};
-
-
-LIBDE265_API void en265_print_logging(const encoder_context* ectx, const char* id, const char* filename);
-
-#endif

libde265-1.0.17.tar.gz/libde265/encoder/encoder-intrapred.cc Deleted

@@ -1,340 +0,0 @@
-/*
- * H.265 video codec.
- * Copyright (c) 2013-2014 struktur AG, Dirk Farin <farin@struktur.de>
- *
- * This file is part of libde265.
- *
- * libde265 is free software: you can redistribute it and/or modify
- * it under the terms of the GNU Lesser General Public License as
- * published by the Free Software Foundation, either version 3 of
- * the License, or (at your option) any later version.
- *
- * libde265 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 Lesser General Public License for more details.
- *
- * You should have received a copy of the GNU Lesser General Public License
- * along with libde265.  If not, see <http://www.gnu.org/licenses/>.
- */
-
-#include "intrapred.h"
-#include "encoder/encoder-intrapred.h"
-#include "encoder/encoder-types.h"
-#include "transform.h"
-#include "util.h"
-#include <assert.h>
-
-
-#include <sys/types.h>
-#include <string.h>
-
-
-void fillIntraPredModeCandidates(enum IntraPredMode candModeList3,
-                                 int x,int y,
-                                 bool availableA, // left
-                                 bool availableB, // top
-                                 const CTBTreeMatrix& ctbs,
-                                 const seq_parameter_set* sps)
-{
-
-  // block on left side
-
-  enum IntraPredMode candIntraPredModeA, candIntraPredModeB;
-
-  if (availableA==false) {
-    candIntraPredModeA=INTRA_DC;
-  }
-  else {
-    const enc_cb* cbL = ctbs.getCB(x-1,y);
-    assert(cbL != NULL);
-
-    if (cbL->PredMode != MODE_INTRA ||
-        cbL->pcm_flag) {
-      candIntraPredModeA=INTRA_DC;
-    }
-    else {
-      const enc_tb* tbL = cbL->getTB(x-1,y);
-      assert(tbL);
-      candIntraPredModeA = tbL->intra_mode;
-    }
-  }
-
-  // block above
-
-  if (availableB==false) {
-    candIntraPredModeB=INTRA_DC;
-  }
-  else {
-    const enc_cb* cbA = ctbs.getCB(x,y-1);
-    assert(cbA != NULL);
-
-    if (cbA->PredMode != MODE_INTRA ||
-        cbA->pcm_flag) {
-      candIntraPredModeB=INTRA_DC;
-    }
-    else if (y-1 < ((y >> sps->Log2CtbSizeY) << sps->Log2CtbSizeY)) {
-      candIntraPredModeB=INTRA_DC;
-    }
-    else {
-      const enc_tb* tbA = cbA->getTB(x,y-1);
-      assert(tbA);
-
-      candIntraPredModeB = tbA->intra_mode;
-    }
-  }
-
-
-  logtrace(LogSlice,"%d;%d candA:%d / candB:%d\n", x,y,
-           availableA ? candIntraPredModeA : -999,
-           availableB ? candIntraPredModeB : -999);
-
-
-  fillIntraPredModeCandidates(candModeList,
-                              candIntraPredModeA,
-                              candIntraPredModeB);
-}
-
-
-template <class pixel_t>
-class intra_border_computer_ctbtree : public intra_border_computer<pixel_t>
-{
-public:
-  void fill_from_ctbtree(const enc_tb* tb,
-                         const CTBTreeMatrix& ctbs);
-};
-
-
-template <class pixel_t>
-void intra_border_computer_ctbtree<pixel_t>::fill_from_ctbtree(const enc_tb* blkTb,
-                                                               const CTBTreeMatrix& ctbs)
-{
-  int xBLuma = this->xB * this->SubWidth;
-  int yBLuma = this->yB * this->SubHeight;
-
-  int currBlockAddr = this->pps->MinTbAddrZS (xBLuma >> this->sps->Log2MinTrafoSize) +
-                                              (yBLuma >> this->sps->Log2MinTrafoSize) * this->sps->PicWidthInTbsY ;
-
-
-  // copy pixels at left column
-
-  for (int y=this->nBottom-1 ; y>=0 ; y-=4)
-    if (this->availableLeft)
-      {
-        int NBlockAddr = this->pps->MinTbAddrZS (((this->xB-1)*this->SubWidth )>>this->sps->Log2MinTrafoSize) +
-                                                 (((this->yB+y)*this->SubHeight)>>this->sps->Log2MinTrafoSize)
-                                                 * this->sps->PicWidthInTbsY ;
-
-        bool availableN = NBlockAddr <= currBlockAddr;
-
-        int xN = this->xB-1;
-        int yN = this->yB+y;
-
-        const enc_cb* cb = ctbs.getCB(xN*this->SubWidth, yN*this->SubHeight);
-
-        if (this->pps->constrained_intra_pred_flag) {
-          if (cb->PredMode != MODE_INTRA)
-            availableN = false;
-        }
-
-        if (availableN) {
-          PixelAccessor pa = cb->transform_tree->getPixels(xN,yN, this->cIdx, *this->sps);
-
-          if (!this->nAvail) this->firstValue = pathis->yB+ythis->xB-1;
-
-          for (int i=0;i<4;i++) {
-            this->available-y+i-1 = availableN;
-            this->out_border-y+i-1 = pathis->yB+y-ithis->xB-1;
-          }
-
-          this->nAvail+=4;
-        }
-      }
-
-  // copy pixel at top-left position
-
-  if (this->availableTopLeft)
-    {
-      int NBlockAddr = this->pps->MinTbAddrZS (((this->xB-1)*this->SubWidth )>>this->sps->Log2MinTrafoSize) +
-                                               (((this->yB-1)*this->SubHeight)>>this->sps->Log2MinTrafoSize)
-                                               * this->sps->PicWidthInTbsY ;
-
-      bool availableN = NBlockAddr <= currBlockAddr;
-
-      int xN = this->xB-1;
-      int yN = this->yB-1;
-
-      const enc_cb* cb = ctbs.getCB(xN*this->SubWidth, yN*this->SubHeight);
-
-      if (this->pps->constrained_intra_pred_flag) {
-        if (cb->PredMode!=MODE_INTRA) {
-          availableN = false;
-        }
-      }
-
-      if (availableN) {
-        PixelAccessor pa = cb->transform_tree->getPixels(xN,yN, this->cIdx, *this->sps);
-
-        this->out_border0 = pathis->yB-1this->xB-1;
-        this->available0 = availableN;
-
-        if (!this->nAvail) this->firstValue = this->out_border0;
-        this->nAvail++;
-      }
-    }
-
-
-  // copy pixels at top row
-
-  for (int x=0 ; x<this->nRight ; x+=4) {
-    bool borderAvailable;
-    if (x<this->nT) borderAvailable = this->availableTop;
-    else            borderAvailable = this->availableTopRight;
-
-    if (borderAvailable)
-      {
-        int NBlockAddr = this->pps->MinTbAddrZS (((this->xB+x)*this->SubWidth )>>this->sps->Log2MinTrafoSize) +
-                                                 (((this->yB-1)*this->SubHeight)>>this->sps->Log2MinTrafoSize)
-                                                 * this->sps->PicWidthInTbsY ;
-
-        bool availableN = NBlockAddr <= currBlockAddr;
-
-        int xN = this->xB+x;
-        int yN = this->yB-1;
-
-        const enc_cb* cb = ctbs.getCB(xN*this->SubWidth, yN*this->SubHeight);
-
-        if (this->pps->constrained_intra_pred_flag) {
-          if (cb->PredMode!=MODE_INTRA) {
-            availableN = false;
-          }
-        }
-
-
-        if (availableN) {
-          PixelAccessor pa = cb->transform_tree->getPixels(xN,yN, this->cIdx, *this->sps);
-
-          if (!this->nAvail) this->firstValue = pathis->yB-1this->xB+x;
-
-          for (int i=0;i<4;i++) {
-            this->out_borderx+i+1 = pathis->yB-1this->xB+x+i;
-            this->availablex+i+1 = availableN;
-          }
-
-          this->nAvail+=4;
-        }
-      }
-  }
-}
-
-
-// (8.4.4.2.2)
-template <class pixel_t>
-void fill_border_samples_from_tree(const de265_image* img,
-                                   const enc_tb* tb,
-                                   const CTBTreeMatrix& ctbs,
-                                   int cIdx,
-                                   pixel_t* out_border)
-{
-  intra_border_computer_ctbtree<pixel_t> c;
-
-  // xB,yB in component specific resolution
-  int xB,yB;
-  int nT = 1<<tb->log2Size;
-
-  xB = tb->x;
-  yB = tb->y;
-
-  if (img->get_sps().chroma_format_idc == CHROMA_444) {
-  }
-  else if (cIdx > 0) {
-    // TODO: proper chroma handling
-    xB >>= 1;
-    yB >>= 1;
-    nT >>= 1;
-
-    if (tb->log2Size==2) {
-      xB = tb->parent->x >> 1;
-      yB = tb->parent->y >> 1;
-      nT = 4;
-    }
-  }
-
-  c.init(out_border, img, nT, cIdx, xB, yB);
-  c.preproc();
-  c.fill_from_ctbtree(tb, ctbs);
-  c.reference_sample_substitution();
-}
-
-
-
-template <class pixel_t>
-void decode_intra_prediction_from_tree_internal(const de265_image* img,
-                                                const enc_tb* tb,
-                                                const CTBTreeMatrix& ctbs,
-                                                const seq_parameter_set& sps,
-                                                int cIdx)
-{
-  enum IntraPredMode intraPredMode;
-  if (cIdx==0) intraPredMode = tb->intra_mode;
-  else         intraPredMode = tb->intra_mode_chroma;
-
-  pixel_t* dst = tb->intra_predictioncIdx->get_buffer<pixel_t>();
-  int dstStride = tb->intra_predictioncIdx->getStride();
-
-  pixel_t  border_pixels_mem4*MAX_INTRA_PRED_BLOCK_SIZE+1;
-  pixel_t* border_pixels = &border_pixels_mem2*MAX_INTRA_PRED_BLOCK_SIZE;
-
-  fill_border_samples_from_tree(img, tb, ctbs, cIdx, border_pixels);
-
-  if (cIdx==0) {
-    // memcpy(tb->debug_intra_border, border_pixels_mem, 2*64+1);
-  }
-
-  int nT = 1<<tb->log2Size;
-  if (cIdx>0 && tb->log2Size>2 && sps.chroma_format_idc == CHROMA_420) {
-    nT >>= 1; // TODO: 4:2:2
-  }
-
-
-  if (sps.range_extension.intra_smoothing_disabled_flag == 0 &&
-      (cIdx==0 || sps.ChromaArrayType==CHROMA_444))
-    {
-      intra_prediction_sample_filtering(sps, border_pixels, nT, cIdx, intraPredMode);
-    }
-
-
-  switch (intraPredMode) {
-  case INTRA_PLANAR:
-    intra_prediction_planar(dst,dstStride, nT, cIdx, border_pixels);
-    break;
-  case INTRA_DC:
-    intra_prediction_DC(dst,dstStride, nT, cIdx, border_pixels);
-    break;
-  default:
-    {
-      //int bit_depth = img->get_bit_depth(cIdx);
-      int bit_depth = 8; // TODO
-
-      bool disableIntraBoundaryFilter =
-        (sps.range_extension.implicit_rdpcm_enabled_flag &&
-         tb->cb->cu_transquant_bypass_flag);
-
-      intra_prediction_angular(dst,dstStride, bit_depth,disableIntraBoundaryFilter,
-                               tb->x,tb->y,intraPredMode,nT,cIdx, border_pixels);
-    }
-    break;
-  }
-}
-
-
-void decode_intra_prediction_from_tree(const de265_image* img,
-                                       const enc_tb* tb,
-                                       const CTBTreeMatrix& ctbs,
-                                       const seq_parameter_set& sps,
-                                       int cIdx)
-{
-  // TODO: high bit depths
-
-  decode_intra_prediction_from_tree_internal<uint8_t>(img ,tb, ctbs, sps, cIdx);
-}

libde265-1.0.17.tar.gz/libde265/encoder/encoder-intrapred.h Deleted

@@ -1,40 +0,0 @@
-/*
- * H.265 video codec.
- * Copyright (c) 2013-2014 struktur AG, Dirk Farin <farin@struktur.de>
- *
- * This file is part of libde265.
- *
- * libde265 is free software: you can redistribute it and/or modify
- * it under the terms of the GNU Lesser General Public License as
- * published by the Free Software Foundation, either version 3 of
- * the License, or (at your option) any later version.
- *
- * libde265 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 Lesser General Public License for more details.
- *
- * You should have received a copy of the GNU Lesser General Public License
- * along with libde265.  If not, see <http://www.gnu.org/licenses/>.
- */
-
-#ifndef DE265_ENCODER_INTRAPRED_H
-#define DE265_ENCODER_INTRAPRED_H
-
-#include "libde265/decctx.h"
-
-
-void fillIntraPredModeCandidates(enum IntraPredMode candModeList3,
-                                 int x,int y,
-                                 bool availableA, // left
-                                 bool availableB, // top
-                                 const class CTBTreeMatrix& ctbs,
-                                 const seq_parameter_set* sps);
-
-void decode_intra_prediction_from_tree(const de265_image* img,
-                                       const class enc_tb* tb,
-                                       const class CTBTreeMatrix& ctbs,
-                                       const class seq_parameter_set& sps,
-                                       int cIdx);
-
-#endif

libde265-1.0.17.tar.gz/libde265/encoder/encoder-motion.cc Deleted

@@ -1,80 +0,0 @@
-/*
- * H.265 video codec.
- * Copyright (c) 2013-2014 struktur AG, Dirk Farin <farin@struktur.de>
- *
- * This file is part of libde265.
- *
- * libde265 is free software: you can redistribute it and/or modify
- * it under the terms of the GNU Lesser General Public License as
- * published by the Free Software Foundation, either version 3 of
- * the License, or (at your option) any later version.
- *
- * libde265 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 Lesser General Public License for more details.
- *
- * You should have received a copy of the GNU Lesser General Public License
- * along with libde265.  If not, see <http://www.gnu.org/licenses/>.
- */
-
-#include "encoder/encoder-motion.h"
-#include "encoder/encoder-context.h"
-#include "decctx.h"
-#include "util.h"
-#include "dpb.h"
-#include "motion.h"
-
-#include <assert.h>
-
-
-#include <sys/types.h>
-#include <signal.h>
-#include <string.h>
-
-#if defined(_MSC_VER) || defined(__MINGW32__)
-# include <malloc.h>
-#elif defined(HAVE_ALLOCA_H)
-# include <alloca.h>
-#endif
-
-
-class MotionVectorAccess_encoder_context : public MotionVectorAccess
-{
-public:
-  MotionVectorAccess_encoder_context(const encoder_context* e) : ectx(e) { }
-
-  enum PartMode get_PartMode(int x,int y) const override { return ectx->ctbs.getCB(x,y)->PartMode; }
-  const PBMotion& get_mv_info(int x,int y) const override { return ectx->ctbs.getPB(x,y)->motion; }
-
-private:
-  const encoder_context* ectx;
-};
-
-
-
-void get_merge_candidate_list_from_tree(encoder_context* ectx,
-                                        const slice_segment_header* shdr,
-                                        int xC,int yC, int xP,int yP,
-                                        int nCS, int nPbW,int nPbH, int partIdx,
-                                        PBMotion* mergeCandList)
-{
-  int max_merge_idx = 5-shdr->five_minus_max_num_merge_cand -1;
-
-  get_merge_candidate_list_without_step_9(ectx, shdr,
-                                          MotionVectorAccess_encoder_context(ectx), ectx->img,
-                                          xC,yC,xP,yP,nCS,nPbW,nPbH, partIdx,
-                                          max_merge_idx, mergeCandList);
-
-  // 9. for encoder: modify all merge candidates
-
-  for (int i=0;i<=max_merge_idx;i++) {
-    if (mergeCandListi.predFlag0 &&
-        mergeCandListi.predFlag1 &&
-        nPbW+nPbH==12)
-      {
-        mergeCandListi.refIdx1   = 0;
-        mergeCandListi.predFlag1 = 0;
-      }
-  }
-}

libde265-1.0.17.tar.gz/libde265/encoder/encoder-motion.h Deleted

@@ -1,32 +0,0 @@
-/*
- * H.265 video codec.
- * Copyright (c) 2013-2014 struktur AG, Dirk Farin <farin@struktur.de>
- *
- * This file is part of libde265.
- *
- * libde265 is free software: you can redistribute it and/or modify
- * it under the terms of the GNU Lesser General Public License as
- * published by the Free Software Foundation, either version 3 of
- * the License, or (at your option) any later version.
- *
- * libde265 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 Lesser General Public License for more details.
- *
- * You should have received a copy of the GNU Lesser General Public License
- * along with libde265.  If not, see <http://www.gnu.org/licenses/>.
- */
-
-#ifndef DE265_ENCODER_MOTION_H
-#define DE265_ENCODER_MOTION_H
-
-#include "libde265/motion.h"
-
-void get_merge_candidate_list_from_tree(class encoder_context* ectx,
-                                        const slice_segment_header* shdr,
-                                        int xC,int yC, int xP,int yP,
-                                        int nCS, int nPbW,int nPbH, int partIdx,
-                                        PBMotion* mergeCandList);
-
-#endif

libde265-1.0.17.tar.gz/libde265/encoder/encoder-params.cc Deleted

@@ -1,83 +0,0 @@
-/*
- * H.265 video codec.
- * Copyright (c) 2013-2014 struktur AG, Dirk Farin <farin@struktur.de>
- *
- * Authors: struktur AG, Dirk Farin <farin@struktur.de>
- *
- * This file is part of libde265.
- *
- * libde265 is free software: you can redistribute it and/or modify
- * it under the terms of the GNU Lesser General Public License as
- * published by the Free Software Foundation, either version 3 of
- * the License, or (at your option) any later version.
- *
- * libde265 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 Lesser General Public License for more details.
- *
- * You should have received a copy of the GNU Lesser General Public License
- * along with libde265.  If not, see <http://www.gnu.org/licenses/>.
- */
-
-#include "encoder-params.h"
-
-
-
-static std::vector<int> power2range(int low,int high)
-{
-  std::vector<int> vals;
-  for (int i=low; i<=high; i*=2)
-    vals.push_back(i);
-  return vals;
-}
-
-encoder_params::encoder_params()
-{
-  //rateControlMethod = RateControlMethod_ConstantQP;
-
-  min_cb_size.set_ID("min-cb-size"); min_cb_size.set_valid_values(power2range(8,64)); min_cb_size.set_default(8);
-  max_cb_size.set_ID("max-cb-size"); max_cb_size.set_valid_values(power2range(8,64)); max_cb_size.set_default(32);
-  min_tb_size.set_ID("min-tb-size"); min_tb_size.set_valid_values(power2range(4,32)); min_tb_size.set_default(4);
-  max_tb_size.set_ID("max-tb-size"); max_tb_size.set_valid_values(power2range(8,32)); max_tb_size.set_default(32);
-
-  max_transform_hierarchy_depth_intra.set_ID("max-transform-hierarchy-depth-intra");
-  max_transform_hierarchy_depth_intra.set_range(0,4);
-  max_transform_hierarchy_depth_intra.set_default(3);
-
-  max_transform_hierarchy_depth_inter.set_ID("max-transform-hierarchy-depth-inter");
-  max_transform_hierarchy_depth_inter.set_range(0,4);
-  max_transform_hierarchy_depth_inter.set_default(3);
-
-  sop_structure.set_ID("sop-structure");
-
-  mAlgo_TB_IntraPredMode.set_ID("TB-IntraPredMode");
-  mAlgo_TB_IntraPredMode_Subset.set_ID("TB-IntraPredMode-subset");
-  mAlgo_CB_IntraPartMode.set_ID("CB-IntraPartMode");
-
-  mAlgo_TB_RateEstimation.set_ID("TB-RateEstimation");
-
-  mAlgo_MEMode.set_ID("MEMode");
-}
-
-
-void encoder_params::registerParams(config_parameters& config)
-{
-  config.add_option(&min_cb_size);
-  config.add_option(&max_cb_size);
-  config.add_option(&min_tb_size);
-  config.add_option(&max_tb_size);
-  config.add_option(&max_transform_hierarchy_depth_intra);
-  config.add_option(&max_transform_hierarchy_depth_inter);
-
-  config.add_option(&sop_structure);
-
-  config.add_option(&mAlgo_TB_IntraPredMode);
-  config.add_option(&mAlgo_TB_IntraPredMode_Subset);
-  config.add_option(&mAlgo_CB_IntraPartMode);
-
-  config.add_option(&mAlgo_MEMode);
-  config.add_option(&mAlgo_TB_RateEstimation);
-
-  mSOP_LowDelay.registerParams(config);
-}

libde265-1.0.17.tar.gz/libde265/encoder/encoder-params.h Deleted

@@ -1,143 +0,0 @@
-/*
- * H.265 video codec.
- * Copyright (c) 2013-2014 struktur AG, Dirk Farin <farin@struktur.de>
- *
- * Authors: Dirk Farin <farin@struktur.de>
- *
- * This file is part of libde265.
- *
- * libde265 is free software: you can redistribute it and/or modify
- * it under the terms of the GNU Lesser General Public License as
- * published by the Free Software Foundation, either version 3 of
- * the License, or (at your option) any later version.
- *
- * libde265 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 Lesser General Public License for more details.
- *
- * You should have received a copy of the GNU Lesser General Public License
- * along with libde265.  If not, see <http://www.gnu.org/licenses/>.
- */
-
-#ifndef ENCODER_PARAMS_H
-#define ENCODER_PARAMS_H
-
-#include "libde265/encoder/encoder-types.h"
-#include "libde265/encoder/encoder-core.h"
-#include "libde265/encoder/sop.h"
-
-
-enum RateControlMethod
-  {
-    RateControlMethod_ConstantQP,
-    RateControlMethod_ConstantLambda
-  };
-
-enum IntraPredSearch
-  {
-    IntraPredSearch_Complete
-  };
-
-
-enum SOP_Structure
-  {
-    SOP_Intra,
-    SOP_LowDelay
-  };
-
-class option_SOP_Structure : public choice_option<enum SOP_Structure>
-{
- public:
-  option_SOP_Structure() {
-    add_choice("intra",     SOP_Intra);
-    add_choice("low-delay", SOP_LowDelay, true);
-  }
-};
-
-
-enum MEMode
-  {
-    MEMode_Test,
-    MEMode_Search
-  };
-
-class option_MEMode : public choice_option<enum MEMode>
-{
- public:
-  option_MEMode() {
-    add_choice("test",   MEMode_Test, true);
-    add_choice("search", MEMode_Search);
-  }
-};
-
-
-struct encoder_params
-{
-  encoder_params();
-
-  void registerParams(config_parameters& config);
-
-
-  // CB quad-tree
-
-  option_int min_cb_size;
-  option_int max_cb_size;
-
-  option_int min_tb_size;
-  option_int max_tb_size;
-
-  option_int max_transform_hierarchy_depth_intra;
-  option_int max_transform_hierarchy_depth_inter;
-
-
-  option_SOP_Structure sop_structure;
-
-  sop_creator_trivial_low_delay::params mSOP_LowDelay;
-
-
-  // --- Algo_TB_IntraPredMode
-
-  option_ALGO_TB_IntraPredMode        mAlgo_TB_IntraPredMode;
-  option_ALGO_TB_IntraPredMode_Subset mAlgo_TB_IntraPredMode_Subset;
-
-  //Algo_TB_IntraPredMode_FastBrute::params TB_IntraPredMode_FastBrute;
-  //Algo_TB_IntraPredMode_MinResidual::params TB_IntraPredMode_MinResidual;
-
-
-  // --- Algo_TB_Split_BruteForce
-
-  //Algo_TB_Split_BruteForce::params  TB_Split_BruteForce;
-
-
-  // --- Algo_CB_IntraPartMode
-
-  option_ALGO_CB_IntraPartMode mAlgo_CB_IntraPartMode;
-
-  //Algo_CB_IntraPartMode_Fixed::params CB_IntraPartMode_Fixed;
-
-  // --- Algo_CB_Split
-
-  // --- Algo_CTB_QScale
-
-  //Algo_CTB_QScale_Constant::params    CTB_QScale_Constant;
-
-  option_MEMode mAlgo_MEMode;
-
-
-  // intra-prediction
-
-  enum IntraPredSearch intraPredSearch;
-
-
-  // rate-control
-
-  enum RateControlMethod rateControlMethod;
-  option_ALGO_TB_RateEstimation mAlgo_TB_RateEstimation;
-
-  //int constant_QP;
-  //int lambda;
-};
-
-
-#endif

libde265-1.0.17.tar.gz/libde265/encoder/encoder-syntax.cc Deleted

@@ -1,1730 +0,0 @@
-/*
- * H.265 video codec.
- * Copyright (c) 2013-2014 struktur AG, Dirk Farin <farin@struktur.de>
- *
- * Authors: struktur AG, Dirk Farin <farin@struktur.de>
- *
- * This file is part of libde265.
- *
- * libde265 is free software: you can redistribute it and/or modify
- * it under the terms of the GNU Lesser General Public License as
- * published by the Free Software Foundation, either version 3 of
- * the License, or (at your option) any later version.
- *
- * libde265 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 Lesser General Public License for more details.
- *
- * You should have received a copy of the GNU Lesser General Public License
- * along with libde265.  If not, see <http://www.gnu.org/licenses/>.
- */
-
-#include "encoder-syntax.h"
-#include "encoder-context.h"
-#include "encoder-intrapred.h"
-#include "slice.h"
-#include "scan.h"
-#include "intrapred.h"
-#include "libde265/transform.h"
-#include "libde265/fallback-dct.h"
-#include <iostream>
-
-
-#ifdef DE265_LOG_DEBUG
-#define ESTIM_BITS_BEGIN \
-  CABAC_encoder_estim* log_estim; \
-  float log_bits_pre = 0; \
-  if (logdebug_enabled(LogEncoder)) { \
-  log_estim = dynamic_cast<CABAC_encoder_estim*>(cabac); \
-  if (log_estim) { \
-    log_bits_pre = log_estim->getRDBits(); \
-  } \
-  }
-
-#define ESTIM_BITS_END(name) \
-  if (logdebug_enabled(LogEncoder)) { \
-  if (log_estim) { \
-    float bits_post = log_estim->getRDBits(); \
-    printf("%s=%f\n",name,bits_post - log_bits_pre);  \
-  } \
-  }
-#else
-#define ESTIM_BITS_BEGIN
-#define ESTIM_BITS_END(name)
-#endif
-
-
-
-static void internal_recursive_cbfChroma_rate(CABAC_encoder_estim* cabac,
-                                              enc_tb* tb, int log2TrafoSize, int trafoDepth)
-{
-  // --- CBF CB/CR ---
-
-  // For 4x4 luma, there is no signaling of chroma CBF, because only the
-  // chroma CBF for 8x8 is relevant.
-  if (log2TrafoSize>2) {
-    if (trafoDepth==0 || tb->parent->cbf1) {
-      encode_cbf_chroma(cabac, trafoDepth, tb->cbf1);
-    }
-    if (trafoDepth==0 || tb->parent->cbf2) {
-      encode_cbf_chroma(cabac, trafoDepth, tb->cbf2);
-    }
-  }
-
-  if (tb->split_transform_flag) {
-    for (int i=0;i<4;i++) {
-      internal_recursive_cbfChroma_rate(cabac, tb->childreni, log2TrafoSize-1, trafoDepth+1);
-    }
-  }
-}
-
-
-float recursive_cbfChroma_rate(CABAC_encoder_estim* cabac,
-                               enc_tb* tb, int log2TrafoSize, int trafoDepth)
-{
-  float bits_pre = cabac->getRDBits();
-
-  internal_recursive_cbfChroma_rate(cabac, tb, log2TrafoSize, trafoDepth);
-
-  float bits_post = cabac->getRDBits();
-
-  return bits_post - bits_pre;
-}
-
-
-
-void encode_split_cu_flag(encoder_context* ectx,
-                          CABAC_encoder* cabac,
-                          int x0, int y0, int ctDepth, int split_flag)
-{
-  logtrace(LogSymbols,"$1 split_cu_flag=%d\n",split_flag);
-
-  // check if neighbors are available
-
-  int availableL = check_CTB_available(ectx->img, x0,y0, x0-1,y0);
-  int availableA = check_CTB_available(ectx->img, x0,y0, x0,y0-1);
-
-  int condL = 0;
-  int condA = 0;
-
-  if (availableL && ectx->ctbs.getCB(x0-1,y0)->ctDepth > ctDepth) condL=1;
-  if (availableA && ectx->ctbs.getCB(x0,y0-1)->ctDepth > ctDepth) condA=1;
-
-  int contextOffset = condL + condA;
-  int context = contextOffset;
-
-  // decode bit
-
-  logtrace(LogSlice,"> split_cu_flag = %d (context=%d)\n",split_flag,context);
-
-  cabac->write_CABAC_bit(CONTEXT_MODEL_SPLIT_CU_FLAG + context, split_flag);
-}
-
-
-void encode_part_mode(encoder_context* ectx,
-                      CABAC_encoder* cabac,
-                      enum PredMode PredMode, enum PartMode PartMode, int cLog2CbSize)
-{
-  logtrace(LogSymbols,"$1 part_mode=%d\n",PartMode);
-  logtrace(LogSlice,"> part_mode = %d\n",PartMode);
-
-  if (PredMode == MODE_INTRA) {
-    int bin = (PartMode==PART_2Nx2N);
-    cabac->write_CABAC_bit(CONTEXT_MODEL_PART_MODE+0, bin);
-  }
-  else {
-    if (PartMode==PART_2Nx2N) {
-      cabac->write_CABAC_bit(CONTEXT_MODEL_PART_MODE+0, 1);
-      return;
-    }
-    else {
-      cabac->write_CABAC_bit(CONTEXT_MODEL_PART_MODE+0, 0);
-    }
-
-    if (cLog2CbSize > ectx->get_sps().Log2MinCbSizeY) {
-      if (ectx->get_sps().amp_enabled_flag) {
-        switch (PartMode) {
-        case PART_2NxN:
-          cabac->write_CABAC_bit(CONTEXT_MODEL_PART_MODE+1, 1);
-          cabac->write_CABAC_bit(CONTEXT_MODEL_PART_MODE+3, 1);
-          break;
-        case PART_Nx2N:
-          cabac->write_CABAC_bit(CONTEXT_MODEL_PART_MODE+1, 0);
-          cabac->write_CABAC_bit(CONTEXT_MODEL_PART_MODE+3, 1);
-          break;
-        case PART_2NxnU:
-          cabac->write_CABAC_bit(CONTEXT_MODEL_PART_MODE+1, 1);
-          cabac->write_CABAC_bit(CONTEXT_MODEL_PART_MODE+3, 0);
-          cabac->write_CABAC_bypass(0);
-          break;
-        case PART_2NxnD:
-          cabac->write_CABAC_bit(CONTEXT_MODEL_PART_MODE+1, 1);
-          cabac->write_CABAC_bit(CONTEXT_MODEL_PART_MODE+3, 0);
-          cabac->write_CABAC_bypass(1);
-          break;
-        case PART_nLx2N:
-          cabac->write_CABAC_bit(CONTEXT_MODEL_PART_MODE+1, 0);
-          cabac->write_CABAC_bit(CONTEXT_MODEL_PART_MODE+3, 0);
-          cabac->write_CABAC_bypass(0);
-          break;
-        case PART_nRx2N:
-          cabac->write_CABAC_bit(CONTEXT_MODEL_PART_MODE+1, 0);
-          cabac->write_CABAC_bit(CONTEXT_MODEL_PART_MODE+3, 0);
-          cabac->write_CABAC_bypass(1);
-          break;
-        case PART_NxN:
-        case PART_2Nx2N:
-          assert(false);
-          break;
-        }
-      }
-      else {
-        if (PartMode==PART_2NxN) {
-          cabac->write_CABAC_bit(CONTEXT_MODEL_PART_MODE+1, 1);
-        }
-        else {
-          assert(PartMode==PART_Nx2N);
-          cabac->write_CABAC_bit(CONTEXT_MODEL_PART_MODE+1, 0);
-        }
-      }
-    }
-    else {
-      if (PartMode==PART_2NxN) {
-        cabac->write_CABAC_bit(CONTEXT_MODEL_PART_MODE+1, 1);
-      }
-      else {
-        cabac->write_CABAC_bit(CONTEXT_MODEL_PART_MODE+1, 0);
-
-        if (cLog2CbSize==3) {
-          assert(PartMode==PART_Nx2N);
-        }
-        else {
-          if (PartMode==PART_Nx2N) {
-            cabac->write_CABAC_bit(CONTEXT_MODEL_PART_MODE+3, 1);
-          }
-          else {
-            assert(PartMode==PART_NxN);
-            cabac->write_CABAC_bit(CONTEXT_MODEL_PART_MODE+3, 0);
-          }
-        }
-      }
-    }
-  }
-}
-
-
-static void encode_pred_mode_flag(encoder_context* ectx,
-                                  CABAC_encoder* cabac,
-                                  enum PredMode PredMode)
-{
-  logtrace(LogSlice,"> pred_mode = %d\n",PredMode);
-
-  int flag = (PredMode == MODE_INTRA) ? 1 : 0;
-
-  logtrace(LogSymbols,"$1 pred_mode=%d\n",flag);
-
-  cabac->write_CABAC_bit(CONTEXT_MODEL_PRED_MODE_FLAG, flag);
-}
-
-
-static void encode_prev_intra_luma_pred_flag(encoder_context* ectx,
-                                             CABAC_encoder* cabac,
-                                             int intraPred)
-{
-  logtrace(LogSymbols,"$1 prev_intra_luma_pred_flag=%d\n",intraPred>=0);
-  int bin = (intraPred>=0);
-
-  logtrace(LogSlice,"> prev_intra_luma_pred_flag = %d\n",bin);
-
-  cabac->write_CABAC_bit(CONTEXT_MODEL_PREV_INTRA_LUMA_PRED_FLAG, bin);
-}
-
-static void encode_intra_mpm_or_rem(encoder_context* ectx,
-                                    CABAC_encoder* cabac,
-                                    int intraPred)
-{
-  if (intraPred>=0) {
-    logtrace(LogSymbols,"$1 mpm_idx=%d\n",intraPred);
-    logtrace(LogSlice,"> mpm_idx = %d\n",intraPred);
-    assert(intraPred<=2);
-    cabac->write_CABAC_TU_bypass(intraPred, 2);
-  }
-  else {
-    logtrace(LogSymbols,"$1 rem_intra_luma_pred_mode=%d\n",-intraPred-1);
-    logtrace(LogSlice,"> rem_intra_luma_pred_mode = %d\n",-intraPred-1);
-    cabac->write_CABAC_FL_bypass(-intraPred-1, 5);
-  }
-}
-
-
-static void encode_intra_chroma_pred_mode(encoder_context* ectx,
-                                          CABAC_encoder* cabac,
-                                          int mode)
-{
-  logtrace(LogSymbols,"$1 intra_chroma_pred_mode=%d\n",mode);
-  logtrace(LogSlice,"> intra_chroma_pred_mode = %d\n",mode);
-
-  if (mode==4) {
-    cabac->write_CABAC_bit(CONTEXT_MODEL_INTRA_CHROMA_PRED_MODE,0);
-  }
-  else {
-    assert(mode<4);
-
-    cabac->write_CABAC_bit(CONTEXT_MODEL_INTRA_CHROMA_PRED_MODE,1);
-    cabac->write_CABAC_FL_bypass(mode, 2);
-  }
-}
-
-
-/* Optimized variant that tests most likely branch first.
- */
-enum IntraChromaPredMode find_chroma_pred_mode(enum IntraPredMode chroma_mode,
-                                               enum IntraPredMode luma_mode)
-{
-  // most likely mode: chroma mode = luma mode
-
-  if (luma_mode==chroma_mode) {
-    return INTRA_CHROMA_LIKE_LUMA;
-  }
-
-
-  // check remaining candidates
-
-  IntraPredMode mode = chroma_mode;
-
-  // angular-34 is coded by setting the coded mode equal to the luma_mode
-  if (chroma_mode == INTRA_ANGULAR_34) {
-    mode = luma_mode;
-  }
-
-  switch (mode) {
-  case INTRA_PLANAR:     return INTRA_CHROMA_PLANAR_OR_34;
-  case INTRA_ANGULAR_26: return INTRA_CHROMA_ANGULAR_26_OR_34;
-  case INTRA_ANGULAR_10: return INTRA_CHROMA_ANGULAR_10_OR_34;
-  case INTRA_DC:         return INTRA_CHROMA_DC_OR_34;
-  default:
-    assert(false);
-    return INTRA_CHROMA_DC_OR_34;
-  }
-}
-
-
-
-void encode_split_transform_flag(encoder_context* ectx,
-                                 CABAC_encoder* cabac,
-                                 int log2TrafoSize, int split_flag)
-{
-  logtrace(LogSymbols,"$1 split_transform_flag=%d\n",split_flag);
-  logtrace(LogSlice,"> split_transform_flag = %d\n",split_flag);
-
-  int context = 5-log2TrafoSize;
-  assert(context >= 0 && context <= 2);
-
-  cabac->write_CABAC_bit(CONTEXT_MODEL_SPLIT_TRANSFORM_FLAG + context, split_flag);
-}
-
-
-void encode_cbf_luma(CABAC_encoder* cabac,
-                     bool zeroTrafoDepth, int cbf_luma)
-{
-  logtrace(LogSymbols,"$1 cbf_luma=%d\n",cbf_luma);
-  logtrace(LogSlice,"> cbf_luma = %d\n",cbf_luma);
-
-  int context = (zeroTrafoDepth ? 1 : 0);
-
-  cabac->write_CABAC_bit(CONTEXT_MODEL_CBF_LUMA + context, cbf_luma);
-}
-
-
-void encode_cbf_chroma(CABAC_encoder* cabac,
-                       int trafoDepth, int cbf_chroma)
-{
-  logtrace(LogSymbols,"$1 cbf_chroma=%d\n",cbf_chroma);
-  logtrace(LogSlice,"> cbf_chroma = %d\n",cbf_chroma);
-
-  int context = trafoDepth;
-  assert(context >= 0 && context <= 3);
-
-  cabac->write_CABAC_bit(CONTEXT_MODEL_CBF_CHROMA + context, cbf_chroma);
-}
-
-static inline void encode_coded_sub_block_flag(encoder_context* ectx,
-                                               CABAC_encoder* cabac,
-                                               int cIdx,
-                                               uint8_t coded_sub_block_neighbors,
-                                               int flag)
-{
-  logtrace(LogSymbols,"$1 coded_sub_block_flag=%d\n",flag);
-  logtrace(LogSlice,"# coded_sub_block_flag = %d\n",flag);
-
-  // tricky computation of csbfCtx
-  int csbfCtx = ((coded_sub_block_neighbors &  1) |  // right neighbor set  or
-                 (coded_sub_block_neighbors >> 1));  // bottom neighbor set   -> csbfCtx=1
-
-  int ctxIdxInc = csbfCtx;
-  if (cIdx!=0) {
-    ctxIdxInc += 2;
-  }
-
-  cabac->write_CABAC_bit(CONTEXT_MODEL_CODED_SUB_BLOCK_FLAG + ctxIdxInc, flag);
-}
-
-static inline void encode_significant_coeff_flag_lookup(encoder_context* ectx,
-                                                        CABAC_encoder* cabac,
-                                                        uint8_t ctxIdxInc,
-                                                        int significantFlag)
-{
-  logtrace(LogSymbols,"$1 significant_coeff_flag=%d\n",significantFlag);
-  logtrace(LogSlice,"# significant_coeff_flag = significantFlag\n");
-  logtrace(LogSlice,"context: %d\n",ctxIdxInc);
-
-  cabac->write_CABAC_bit(CONTEXT_MODEL_SIGNIFICANT_COEFF_FLAG + ctxIdxInc, significantFlag);
-}
-
-static inline void encode_coeff_abs_level_greater1(encoder_context* ectx,
-                                                   CABAC_encoder* cabac,
-                                                   int cIdx, int i,
-                                                   bool firstCoeffInSubblock,
-                                                   bool firstSubblock,
-                                                   int  lastSubblock_greater1Ctx,
-                                                   int* lastInvocation_greater1Ctx,
-                                                   int* lastInvocation_coeff_abs_level_greater1_flag,
-                                                   int* lastInvocation_ctxSet, int c1,
-                                                   int value)
-{
-  logtrace(LogSymbols,"$1 coeff_abs_level_greater1=%d\n",value);
-  logtrace(LogSlice,"# coeff_abs_level_greater1 = %d\n",value);
-
-  logtrace(LogSlice,"  cIdx:%d i:%d firstCoeffInSB:%d firstSB:%d lastSB>1:%d last>1Ctx:%d lastLev>1:%d lastCtxSet:%d\n", cIdx,i,firstCoeffInSubblock,firstSubblock,lastSubblock_greater1Ctx,
-	   *lastInvocation_greater1Ctx,
-	   *lastInvocation_coeff_abs_level_greater1_flag,
-	   *lastInvocation_ctxSet);
-
-  int lastGreater1Ctx;
-  int greater1Ctx;
-  int ctxSet;
-
-  logtrace(LogSlice,"c1: %d\n",c1);
-
-  if (firstCoeffInSubblock) {
-    // block with real DC -> ctx 0
-    if (i==0 || cIdx>0) { ctxSet=0; }
-    else { ctxSet=2; }
-
-    if (firstSubblock) { lastGreater1Ctx=1; }
-    else { lastGreater1Ctx = lastSubblock_greater1Ctx; }
-
-    if (lastGreater1Ctx==0) { ctxSet++; }
-
-    logtrace(LogSlice,"ctxSet: %d\n",ctxSet);
-
-    greater1Ctx=1;
-  }
-  else { // !firstCoeffInSubblock
-    ctxSet = *lastInvocation_ctxSet;
-    logtrace(LogSlice,"ctxSet (old): %d\n",ctxSet);
-
-    greater1Ctx = *lastInvocation_greater1Ctx;
-    if (greater1Ctx>0) {
-      int lastGreater1Flag=*lastInvocation_coeff_abs_level_greater1_flag;
-      if (lastGreater1Flag==1) greater1Ctx=0;
-      else { /*if (greater1Ctx>0)*/ greater1Ctx++; }
-    }
-  }
-
-  ctxSet = c1; // use HM algo
-
-  int ctxIdxInc = (ctxSet*4) + (greater1Ctx>=3 ? 3 : greater1Ctx);
-
-  if (cIdx>0) { ctxIdxInc+=16; }
-
-  cabac->write_CABAC_bit(CONTEXT_MODEL_COEFF_ABS_LEVEL_GREATER1_FLAG + ctxIdxInc,  value);
-
-  *lastInvocation_greater1Ctx = greater1Ctx;
-  *lastInvocation_coeff_abs_level_greater1_flag = value;
-  *lastInvocation_ctxSet = ctxSet;
-}
-
-static void encode_coeff_abs_level_greater2(encoder_context* ectx,
-                                            CABAC_encoder* cabac,
-                                            int cIdx, // int i,int n,
-                                            int ctxSet,
-                                            int value)
-{
-  logtrace(LogSymbols,"$1 coeff_abs_level_greater2=%d\n",value);
-  logtrace(LogSlice,"# coeff_abs_level_greater2 = %d\n",value);
-
-  int ctxIdxInc = ctxSet;
-
-  if (cIdx>0) ctxIdxInc+=4;
-
-  cabac->write_CABAC_bit(CONTEXT_MODEL_COEFF_ABS_LEVEL_GREATER2_FLAG + ctxIdxInc,  value);
-}
-
-
-bool TU(int val, int maxi)
-{
-  for (int i=0;i<val;i++) {
-    printf("1");
-  }
-  if (val<maxi) { printf("0"); return false; }
-  else return true;
-}
-
-void bin(int val, int bits)
-{
-  for (int i=0;i<bits;i++) {
-    int bit = (1<<(bits-1-i));
-    if (val&bit) printf("1"); else printf("0");
-  }
-}
-
-void ExpG(int level, int riceParam)
-{
-  int prefix = level >> riceParam;
-  int suffix = level - (prefix<<riceParam);
-
-  //printf("%d %d ",prefix,suffix);
-
-  int base=0;
-  int range=1;
-  int nBits=0;
-  while (prefix >= base+range) {
-    printf("1");
-    base+=range;
-    range*=2;
-    nBits++;
-  }
-
-  printf("0.");
-  bin(prefix-base, nBits);
-  printf(":");
-  bin(suffix,riceParam);
-}
-
-int blamain()
-{
-  int riceParam=2;
-  int TRMax = 4<<riceParam;
-
-  for (int level=0;level<128;level++)
-    {
-      printf("%d: ",level);
-
-      int prefixPart = std::min(TRMax, level);
-
-      // code TR prefix
-
-      bool isMaxi = TU(prefixPart>>riceParam, TRMax>>riceParam);
-      printf(":");
-      if (TRMax>prefixPart) {
-        int remain = prefixPart & ((1<<riceParam)-1);
-        bin(remain, riceParam);
-      }
-      printf("|");
-
-      if (isMaxi) {
-        ExpG(level-TRMax, riceParam+1);
-      }
-
-      printf("\n");
-    }
-
-  return 0;
-}
-
-
-static void encode_coeff_abs_level_remaining(encoder_context* ectx,
-                                             CABAC_encoder* cabac,
-                                             int cRiceParam,
-                                             int level)
-{
-  logtrace(LogSymbols,"$1 coeff_abs_level_remaining=%d\n",level);
-  logtrace(LogSlice,"# encode_coeff_abs_level_remaining = %d\n",level);
-
-  int cTRMax = 4<<cRiceParam;
-  int prefixPart = std::min(level, cTRMax);
-
-  // --- code prefix with TR ---
-
-  // TU part, length 4 (cTRMax>>riceParam)
-
-  int nOnes = (prefixPart>>cRiceParam);
-  cabac->write_CABAC_TU_bypass(nOnes, 4);
-
-  // TR suffix
-
-  if (cTRMax > prefixPart) {
-    int remain = prefixPart & ((1<<cRiceParam)-1);
-    cabac->write_CABAC_FL_bypass(remain, cRiceParam);
-  }
-
-
-  // --- remainder suffix ---
-
-  if (nOnes==4) {
-    int remain = level-cTRMax;
-    int ExpGRiceParam = cRiceParam+1;
-
-    int prefix = remain >> ExpGRiceParam;
-    int suffix = remain - (prefix<<ExpGRiceParam);
-
-    int base=0;
-    int range=1;
-    int nBits=0;
-    while (prefix >= base+range) {
-      cabac->write_CABAC_bypass(1);
-      base+=range;
-      range*=2;
-      nBits++;
-    }
-
-    cabac->write_CABAC_bypass(0);
-    cabac->write_CABAC_FL_bypass(prefix-base, nBits);
-    cabac->write_CABAC_FL_bypass(suffix, ExpGRiceParam);
-  }
-}
-
-// ---------------------------------------------------------------------------
-
-void findLastSignificantCoeff(const position* sbScan, const position* cScan,
-                              const int16_t* coeff, int log2TrafoSize,
-                              int* lastSignificantX, int* lastSignificantY,
-                              int* lastSb, int* lastPos)
-{
-  int nSb = 1<<((log2TrafoSize-2)<<1); // number of sub-blocks
-
-  // find last significant coefficient
-
-  for (int i=nSb ; i-->0 ;) {
-    int x0 = sbScani.x << 2;
-    int y0 = sbScani.y << 2;
-    for (int c=16 ; c-->0 ;) {
-      int x = x0 + cScanc.x;
-      int y = y0 + cScanc.y;
-
-      if (coeffx+(y<<log2TrafoSize)) {
-        *lastSignificantX = x;
-        *lastSignificantY = y;
-        *lastSb = i;
-        *lastPos= c;
-
-        logtrace(LogSlice,"last significant coeff at: %d;%d, Sb:%d Pos:%d\n", x,y,i,c);
-
-        return;
-      }
-    }
-  }
-
-  // all coefficients == 0 ? cannot be since cbf should be false in this case
-  assert(false);
-}
-
-
-bool subblock_has_nonzero_coefficient(const int16_t* coeff, int coeffStride,
-                                      const position& sbPos)
-{
-  int x0 = sbPos.x << 2;
-  int y0 = sbPos.y << 2;
-
-  coeff += x0 + y0*coeffStride;
-
-  for (int y=0;y<4;y++) {
-    if (coeff0 || coeff1 || coeff2 || coeff3) { return true; }
-    coeff += coeffStride;
-  }
-
-  return false;
-}
-
-/*
-  Example 16x16:  prefix in 0;7
-
-  prefix       | last pos
-  =============|=============
-  0            |   0
-  1            |   1
-  2            |   2
-  3            |   3
-  -------------+-------------
-     lsb nBits |
-  4   0    1   |   4, 5
-  5   1    1   |   6, 7
-  6   0    2   |   8, 9,10,11
-  7   1    2   |  12,13,14,15
-*/
-void encode_last_signficiant_coeff_prefix(encoder_context* ectx,
-                                          CABAC_encoder* cabac,
-                                          int log2TrafoSize,
-                                          int cIdx, int lastSignificant,
-                                          int context_model_index)
-{
-  logtrace(LogSlice,"> last_significant_coeff_prefix=%d log2TrafoSize:%d cIdx:%d\n",
-           lastSignificant,log2TrafoSize,cIdx);
-
-  int cMax = (log2TrafoSize<<1)-1;
-
-  int ctxOffset, ctxShift;
-  if (cIdx==0) {
-    ctxOffset = 3*(log2TrafoSize-2) + ((log2TrafoSize-1)>>2);
-    ctxShift  = (log2TrafoSize+1)>>2;
-  }
-  else {
-    ctxOffset = 15;
-    ctxShift  = log2TrafoSize-2;
-  }
-
-  for (int binIdx=0;binIdx<lastSignificant;binIdx++)
-    {
-      int ctxIdxInc = (binIdx >> ctxShift);
-      cabac->write_CABAC_bit(context_model_index + ctxOffset + ctxIdxInc, 1);
-    }
-
-  if (lastSignificant != cMax) {
-    int binIdx = lastSignificant;
-    int ctxIdxInc = (binIdx >> ctxShift);
-    cabac->write_CABAC_bit(context_model_index + ctxOffset + ctxIdxInc, 0);
-  }
-}
-
-
-void split_last_significant_position(int pos, int* prefix, int* suffix, int* nSuffixBits)
-{
-  logtrace(LogSlice,"split position %d : ",pos);
-
-  // most frequent case
-
-  if (pos<=3) {
-    *prefix=pos;
-    *suffix=-1; // just to have some defined value
-    *nSuffixBits=0;
-    logtrace(LogSlice,"prefix=%d suffix=%d (%d bits)\n",*prefix,*suffix,*nSuffixBits);
-    return;
-  }
-
-  pos -= 4;
-  int nBits=1;
-  int range=4;
-  while (pos>=range) {
-    nBits++;
-    pos-=range;
-    range<<=1;
-  }
-
-  *prefix = (1+nBits)<<1;
-  if (pos >= (range>>1)) {
-    *prefix |= 1;
-    pos -= (range>>1);
-  }
-  *suffix = pos;
-  *nSuffixBits = nBits;
-
-  logtrace(LogSlice,"prefix=%d suffix=%d (%d bits)\n",*prefix,*suffix,*nSuffixBits);
-}
-
-
-extern uint8_t* ctxIdxLookup4 /* 4-log2-32 */2 /* !!cIdx */2 /* !!scanIdx */4 /* prevCsbf */;
-
-/* These values are read from the image metadata:
-   - intra prediction mode (x0;y0)
- */
-void encode_residual(encoder_context* ectx,
-                     CABAC_encoder* cabac,
-                     const enc_tb* tb, const enc_cb* cb,
-                     int x0,int y0,int log2TrafoSize,int cIdx)
-{
-  logdebug(LogEncoder,"encode_residual %s\n",typeid(*cabac).name());
-
-  const de265_image* img = ectx->img;
-  const seq_parameter_set& sps = img->get_sps();
-  const pic_parameter_set& pps = img->get_pps();
-
-  int16_t* coeff = tb->coeffcIdx;
-
-  if (pps.transform_skip_enabled_flag && true /* TODO */) {
-  }
-
-
-#if 1
-  logdebug(LogEncoder,"write coefficients\n");
-  for (int y=0;y<(1<<log2TrafoSize);y++)
-    {
-      for (int x=0;x<(1<<log2TrafoSize);x++)
-        {
-          logdebug(LogEncoder,"*%4d ",coeffx+y*(1<<log2TrafoSize));
-        }
-      logdebug(LogEncoder,"*\n");
-    }
-#endif
-
-
-  // --- get scan orders ---
-
-  enum PredMode PredMode = cb->PredMode;
-  int scanIdx;
-
-  if (PredMode == MODE_INTRA) {
-    if (cIdx==0) {
-      scanIdx = get_intra_scan_idx(log2TrafoSize, tb->intra_mode,  cIdx, &sps);
-      //printf("luma scan idx=%d <- intra mode=%d\n",scanIdx, tb->intra_mode);
-    }
-    else {
-      scanIdx = get_intra_scan_idx(log2TrafoSize, tb->intra_mode_chroma,  cIdx, &sps);
-      //printf("chroma scan idx=%d <- intra mode=%d chroma:%d trsize:%d\n",scanIdx,
-      //       tb->intra_mode_chroma, sps.chroma_format_idc, 1<<log2TrafoSize);
-    }
-  }
-  else {
-    scanIdx=0;
-  }
-
-
-  const position* ScanOrderSub = get_scan_order(log2TrafoSize-2, scanIdx);
-  const position* ScanOrderPos = get_scan_order(2, scanIdx);
-
-  int lastSignificantX, lastSignificantY;
-  int lastScanPos;
-  int lastSubBlock;
-  findLastSignificantCoeff(ScanOrderSub, ScanOrderPos,
-                           coeff, log2TrafoSize,
-                           &lastSignificantX, &lastSignificantY,
-                           &lastSubBlock, &lastScanPos);
-
-  int codedSignificantX = lastSignificantX;
-  int codedSignificantY = lastSignificantY;
-
-  if (scanIdx==2) {
-    std::swap(codedSignificantX, codedSignificantY);
-  }
-
-
-
-  int prefixX, suffixX, suffixBitsX;
-  int prefixY, suffixY, suffixBitsY;
-
-  split_last_significant_position(codedSignificantX, &prefixX,&suffixX,&suffixBitsX);
-  split_last_significant_position(codedSignificantY, &prefixY,&suffixY,&suffixBitsY);
-
-  encode_last_signficiant_coeff_prefix(ectx, cabac, log2TrafoSize, cIdx, prefixX,
-                                       CONTEXT_MODEL_LAST_SIGNIFICANT_COEFFICIENT_X_PREFIX);
-
-  encode_last_signficiant_coeff_prefix(ectx, cabac, log2TrafoSize, cIdx, prefixY,
-                                       CONTEXT_MODEL_LAST_SIGNIFICANT_COEFFICIENT_Y_PREFIX);
-
-
-  if (codedSignificantX > 3) {
-    cabac->write_CABAC_FL_bypass(suffixX, suffixBitsX);
-  }
-  if (codedSignificantY > 3) {
-    cabac->write_CABAC_FL_bypass(suffixY, suffixBitsY);
-  }
-
-
-
-  int sbWidth = 1<<(log2TrafoSize-2);
-  int CoeffStride = 1<<log2TrafoSize;
-
-  uint8_t coded_sub_block_neighbors32/4*32/4;  // 64*2 flags
-  memset(coded_sub_block_neighbors,0,sbWidth*sbWidth);
-
-  int  c1 = 1;
-  bool firstSubblock = true;           // for coeff_abs_level_greater1_flag context model
-  int  lastSubblock_greater1Ctx=false; /* for coeff_abs_level_greater1_flag context model
-                                          (initialization not strictly needed)
-                                       */
-
-  int  lastInvocation_greater1Ctx=0;
-  int  lastInvocation_coeff_abs_level_greater1_flag=0;
-  int  lastInvocation_ctxSet=0;
-
-
-
-  // ----- encode coefficients -----
-
-  //tctx->nCoeffcIdx = 0;
-
-
-  // i - subblock index
-  // n - coefficient index in subblock
-
-  for (int i=lastSubBlock;i>=0;i--) {
-    position S = ScanOrderSubi;
-    int inferSbDcSigCoeffFlag=0;
-
-    logtrace(LogSlice,"sub block scan idx: %d\n",i);
-
-
-    // --- check whether this sub-block has to be coded ---
-
-    int sub_block_is_coded = 0;
-
-    if ((i<lastSubBlock) && (i>0)) {
-      sub_block_is_coded = subblock_has_nonzero_coefficient(coeff, CoeffStride, S);
-      encode_coded_sub_block_flag(ectx, cabac, cIdx,
-                                  coded_sub_block_neighborsS.x+S.y*sbWidth,
-                                  sub_block_is_coded);
-      inferSbDcSigCoeffFlag=1;
-    }
-    else if (i==0 || i==lastSubBlock) {
-      // first (DC) and last sub-block are always coded
-      // - the first will most probably contain coefficients
-      // - the last obviously contains the last coded coefficient
-
-      sub_block_is_coded = 1;
-    }
-
-    if (sub_block_is_coded) {
-      if (S.x > 0) coded_sub_block_neighborsS.x-1 + S.y  *sbWidth |= 1;
-      if (S.y > 0) coded_sub_block_neighborsS.x + (S.y-1)*sbWidth |= 2;
-    }
-
-    logtrace(LogSlice,"subblock is coded: %s\n", sub_block_is_coded ? "yes":"no");
-
-
-    // --- write significant coefficient flags ---
-
-    int16_t  coeff_value16;
-    int16_t  coeff_baseLevel16;
-    int8_t   coeff_scan_pos16;
-    int8_t   coeff_sign16;
-    int8_t   coeff_has_max_base_level16;
-    int nCoefficients=0;
-
-
-    if (sub_block_is_coded) {
-      int x0 = S.x<<2;
-      int y0 = S.y<<2;
-
-      int log2w = log2TrafoSize-2;
-      int prevCsbf = coded_sub_block_neighborsS.x+S.y*sbWidth;
-      uint8_t* ctxIdxMap = ctxIdxLookuplog2w!!cIdx!!scanIdxprevCsbf;
-
-      logdebug(LogSlice,"log2w:%d cIdx:%d scanIdx:%d prevCsbf:%d\n",
-               log2w,cIdx,scanIdx,prevCsbf);
-
-
-      // set the last coded coefficient in the last subblock
-
-      if (i==lastSubBlock) {
-        coeff_valuenCoefficients = coefflastSignificantX+(lastSignificantY<<log2TrafoSize);
-        coeff_has_max_base_levelnCoefficients = 1;  // TODO
-        coeff_scan_posnCoefficients = lastScanPos;
-        nCoefficients++;
-      }
-
-
-      // --- encode all coefficients' significant_coeff flags except for the DC coefficient ---
-
-      int last_coeff =  (i==lastSubBlock) ? lastScanPos-1 : 15;
-
-      for (int n= last_coeff ; n>0 ; n--) {
-        int subX = ScanOrderPosn.x;
-        int subY = ScanOrderPosn.y;
-        int xC = x0 + subX;
-        int yC = y0 + subY;
-
-
-        // for all AC coefficients in sub-block, a significant_coeff flag is coded
-
-        int isSignificant = !!tb->coeffcIdxxC + (yC<<log2TrafoSize);
-
-        logtrace(LogSlice,"coeff %d is significant: %d\n", n, isSignificant);
-
-        logtrace(LogSlice,"trafoSize: %d\n",1<<log2TrafoSize);
-        logtrace(LogSlice,"context idx: %d;%d\n",xC,yC);
-
-        encode_significant_coeff_flag_lookup(ectx, cabac,
-                                             ctxIdxMapxC+(yC<<log2TrafoSize),
-                                             isSignificant);
-        //ctxIdxMap(i<<4)+n);
-
-        if (isSignificant) {
-          coeff_valuenCoefficients = coeffxC+(yC<<log2TrafoSize);
-          coeff_has_max_base_levelnCoefficients = 1;
-          coeff_scan_posnCoefficients = n;
-          nCoefficients++;
-
-          // since we have a coefficient in the sub-block,
-          // we cannot infer the DC coefficient anymore
-          inferSbDcSigCoeffFlag = 0;
-        }
-      }
-
-
-      // --- decode DC coefficient significance ---
-
-      if (last_coeff>=0) // last coded coefficient (always set to 1) is not the DC coefficient
-        {
-          if (inferSbDcSigCoeffFlag==0) {
-            // if we cannot infert the DC coefficient, it is coded
-            int isSignificant = !!tb->coeffcIdxx0 + (y0<<log2TrafoSize);
-
-            logtrace(LogSlice,"DC coeff is significant: %d\n", isSignificant);
-
-            encode_significant_coeff_flag_lookup(ectx, cabac,
-                                                 ctxIdxMapx0+(y0<<log2TrafoSize),
-                                                 isSignificant);
-
-            if (isSignificant) {
-              coeff_valuenCoefficients = coeffx0+(y0<<log2TrafoSize);
-              coeff_has_max_base_levelnCoefficients = 1;
-              coeff_scan_posnCoefficients = 0;
-              nCoefficients++;
-            }
-          }
-          else {
-            // we can infer that the DC coefficient must be present
-            coeff_valuenCoefficients = coeffx0+(y0<<log2TrafoSize);
-            coeff_has_max_base_levelnCoefficients = 1;
-            coeff_scan_posnCoefficients = 0;
-            nCoefficients++;
-          }
-        }
-    }
-
-
-
-    // --- encode coefficient values ---
-
-    if (nCoefficients) {
-
-      // separate absolute coefficient value and sign
-
-      logtrace(LogSlice,"coefficients to code: ");
-
-      for (int l=0;l<nCoefficients;l++) {
-        logtrace(LogSlice,"%d ",coeff_valuel);
-
-        if (coeff_valuel<0) {
-          coeff_valuel = -coeff_valuel;
-          coeff_signl = 1;
-        }
-        else {
-          coeff_signl = 0;
-        }
-
-        coeff_baseLevell = 1;
-
-        logtrace(LogSlice,"(%d) ",coeff_scan_posl);
-      }
-
-      logtrace(LogSlice,"\n");
-
-
-      int ctxSet;
-      if (i==0 || cIdx>0) { ctxSet=0; }
-      else { ctxSet=2; }
-
-      if (c1==0) { ctxSet++; }
-      c1=1;
-
-
-      // --- encode greater-1 flags ---
-
-      int newLastGreater1ScanPos=-1;
-
-      int lastGreater1Coefficient = libde265_min(8,nCoefficients);
-      for (int c=0;c<lastGreater1Coefficient;c++) {
-        int greater1_flag = (coeff_valuec>1);
-
-        encode_coeff_abs_level_greater1(ectx, cabac, cIdx,i,
-                                        c==0,
-                                        firstSubblock,
-                                        lastSubblock_greater1Ctx,
-                                        &lastInvocation_greater1Ctx,
-                                        &lastInvocation_coeff_abs_level_greater1_flag,
-                                        &lastInvocation_ctxSet, ctxSet,
-                                        greater1_flag);
-
-        if (greater1_flag) {
-          coeff_baseLevelc++;
-
-          c1=0;
-
-          if (newLastGreater1ScanPos == -1) {
-            newLastGreater1ScanPos=c;
-          }
-        }
-        else {
-          coeff_has_max_base_levelc = 0;
-
-          if (c1<3 && c1>0) {
-            c1++;
-          }
-        }
-      }
-
-      firstSubblock = false;
-      lastSubblock_greater1Ctx = lastInvocation_greater1Ctx;
-
-
-      // --- decode greater-2 flag ---
-
-      if (newLastGreater1ScanPos != -1) {
-        int greater2_flag = (coeff_valuenewLastGreater1ScanPos>2);
-        encode_coeff_abs_level_greater2(ectx,cabac, cIdx, lastInvocation_ctxSet, greater2_flag);
-        coeff_baseLevelnewLastGreater1ScanPos += greater2_flag;
-        coeff_has_max_base_levelnewLastGreater1ScanPos = greater2_flag;
-      }
-
-
-      // --- encode coefficient signs ---
-
-      int signHidden = (coeff_scan_pos0-coeff_scan_posnCoefficients-1 > 3 &&
-                        !cb->cu_transquant_bypass_flag);
-
-      for (int n=0;n<nCoefficients-1;n++) {
-        cabac->write_CABAC_bypass(coeff_signn);
-        //logtrace(LogSlice,"a) sign%d = %d\n", n, coeff_signn);
-      }
-
-      // n==nCoefficients-1
-      if (!pps.sign_data_hiding_flag || !signHidden) {
-        cabac->write_CABAC_bypass(coeff_signnCoefficients-1);
-        //logtrace(LogSlice,"b) sign%d = %d\n", nCoefficients-1, coeff_signnCoefficients-1);
-      }
-      else {
-        assert(coeff_signnCoefficients-1 == 0);
-      }
-
-      // --- decode coefficient value ---
-
-      int sumAbsLevel=0;
-      int uiGoRiceParam=0;
-
-      for (int n=0;n<nCoefficients;n++) {
-        int baseLevel = coeff_baseLeveln;
-
-        int coeff_abs_level_remaining;
-
-        if (coeff_has_max_base_leveln) {
-          logtrace(LogSlice,"value%d=%d, base level: %d\n",n,coeff_valuen,coeff_baseLeveln);
-
-          coeff_abs_level_remaining = coeff_valuen - coeff_baseLeveln;
-
-          encode_coeff_abs_level_remaining(ectx, cabac, uiGoRiceParam,
-                                           coeff_abs_level_remaining);
-
-          // (9-462)
-          if (baseLevel + coeff_abs_level_remaining > 3*(1<<uiGoRiceParam)) {
-            uiGoRiceParam++;
-            if (uiGoRiceParam>4) uiGoRiceParam=4;
-          }
-        }
-        else {
-          coeff_abs_level_remaining = 0;
-        }
-
-
-        // --- DEBUG: check coefficient ---
-
-#if 0
-        int16_t currCoeff = baseLevel + coeff_abs_level_remaining;
-        if (coeff_signn) {
-          currCoeff = -currCoeff;
-        }
-
-        if (pps.sign_data_hiding_flag && signHidden) {
-          sumAbsLevel += baseLevel + coeff_abs_level_remaining;
-
-          if (n==nCoefficients-1 && (sumAbsLevel & 1)) {
-            currCoeff = -currCoeff;
-          }
-        }
-
-        assert(currCoeff == coeff_valuen);
-#endif
-      }  // iterate through coefficients in sub-block
-    }  // if nonZero
-
-  }
-}
-
-
-void encode_transform_unit(encoder_context* ectx,
-                           CABAC_encoder* cabac,
-                           const enc_tb* tb, const enc_cb* cb,
-                           int x0,int y0, int xBase,int yBase,
-                           int log2TrafoSize, int trafoDepth, int blkIdx)
-{
-  ESTIM_BITS_BEGIN;
-
-  if (tb->cbf0 || tb->cbf1 || tb->cbf2) {
-    if (ectx->img->get_pps().cu_qp_delta_enabled_flag &&
-        true /*!ectx->IsCuQpDeltaCoded*/) {
-      assert(0);
-    }
-
-    if (tb->cbf0) {
-      encode_residual(ectx,cabac, tb,cb,x0,y0,log2TrafoSize,0);
-    }
-
-    if (ectx->get_sps().chroma_format_idc == CHROMA_444) {
-      if (tb->cbf1) {
-        encode_residual(ectx,cabac, tb,cb,x0,y0,log2TrafoSize,1);
-      }
-      if (tb->cbf2) {
-        encode_residual(ectx,cabac, tb,cb,x0,y0,log2TrafoSize,2);
-      }
-    }
-    else if (log2TrafoSize>2) {
-      // larger than 4x4
-
-      if (tb->cbf1) {
-        encode_residual(ectx,cabac,tb,cb,x0,y0,log2TrafoSize-1,1);
-      }
-      if (tb->cbf2) {
-        encode_residual(ectx,cabac,tb,cb,x0,y0,log2TrafoSize-1,2);
-      }
-    }
-    else if (blkIdx==3) {
-      // cannot check for tb->parent->cbf, because this may not yet be set
-      if (tb->cbf1) {
-        encode_residual(ectx,cabac,tb,cb,xBase,yBase,log2TrafoSize,1);
-      }
-      if (tb->cbf2) {
-        encode_residual(ectx,cabac,tb,cb,xBase,yBase,log2TrafoSize,2);
-      }
-    }
-  }
-
-  ESTIM_BITS_END("encode_transform_unit");
-}
-
-
-void encode_transform_tree(encoder_context* ectx,
-                           CABAC_encoder* cabac,
-                           const enc_tb* tb, const enc_cb* cb,
-                           int x0,int y0, int xBase,int yBase,
-                           int log2TrafoSize, int trafoDepth, int blkIdx,
-                           int MaxTrafoDepth, int IntraSplitFlag, bool recurse)
-{
-  ESTIM_BITS_BEGIN;
-
-  //de265_image* img = ectx->img;
-  const seq_parameter_set& sps = ectx->img->get_sps();
-
-  if (log2TrafoSize <= sps.Log2MaxTrafoSize &&
-      log2TrafoSize >  sps.Log2MinTrafoSize &&
-      trafoDepth < MaxTrafoDepth &&
-      !(IntraSplitFlag && trafoDepth==0))
-    {
-      int split_transform_flag = tb->split_transform_flag;
-      encode_split_transform_flag(ectx, cabac, log2TrafoSize, split_transform_flag);
-    }
-  else
-    {
-      int interSplitFlag=0; // TODO
-
-      bool split_transform_flag = (log2TrafoSize > sps.Log2MaxTrafoSize ||
-                                   (IntraSplitFlag==1 && trafoDepth==0) ||
-                                   interSplitFlag==1) ? 1:0;
-
-      /*
-      printf("split_transform_flag log2TrafoSize:%d Log2MaxTrafoSize:%d "
-             "IntraSplitFlag:%d trafoDepth:%d -> %d\n",
-             log2TrafoSize,sps->Log2MaxTrafoSize,
-             IntraSplitFlag, trafoDepth,
-             split_transform_flag);
-      */
-
-      assert(tb->split_transform_flag == split_transform_flag);
-    }
-
-  // --- CBF CB/CR ---
-
-  // For 4x4 luma, there is no signaling of chroma CBF, because only the
-  // chroma CBF for 8x8 is relevant.
-  if (log2TrafoSize>2 || sps.ChromaArrayType == CHROMA_444) {
-    if (trafoDepth==0 || tb->parent->cbf1) {
-      encode_cbf_chroma(cabac, trafoDepth, tb->cbf1);
-    }
-    if (trafoDepth==0 || tb->parent->cbf2) {
-      encode_cbf_chroma(cabac, trafoDepth, tb->cbf2);
-    }
-  }
-
-  if (tb->split_transform_flag) {
-    if (recurse) {
-      int x1 = x0 + (1<<(log2TrafoSize-1));
-      int y1 = y0 + (1<<(log2TrafoSize-1));
-
-      encode_transform_tree(ectx, cabac, tb->children0, cb, x0,y0,x0,y0,log2TrafoSize-1,
-                            trafoDepth+1, 0, MaxTrafoDepth, IntraSplitFlag, true);
-      encode_transform_tree(ectx, cabac, tb->children1, cb, x1,y0,x0,y0,log2TrafoSize-1,
-                            trafoDepth+1, 1, MaxTrafoDepth, IntraSplitFlag, true);
-      encode_transform_tree(ectx, cabac, tb->children2, cb, x0,y1,x0,y0,log2TrafoSize-1,
-                            trafoDepth+1, 2, MaxTrafoDepth, IntraSplitFlag, true);
-      encode_transform_tree(ectx, cabac, tb->children3, cb, x1,y1,x0,y0,log2TrafoSize-1,
-                            trafoDepth+1, 3, MaxTrafoDepth, IntraSplitFlag, true);
-    }
-  }
-  else {
-    if (cb->PredMode == MODE_INTRA || trafoDepth != 0 ||
-        tb->cbf1 || tb->cbf2) {
-      encode_cbf_luma(cabac, trafoDepth==0, tb->cbf0);
-    }
-    else {
-    /* Note: usually, cbf0 should be TRUE, but while estimating the bitrate, this
-       function can also be called with all CBFs FALSE. Usually, this is handled by
-       the rqt_root_cbf flag, but during analysis, this is set after the bitrate is estimated.
-     */
-      // assert(tb->cbf0==true);
-    }
-
-    encode_transform_unit(ectx,cabac, tb,cb, x0,y0, xBase,yBase, log2TrafoSize, trafoDepth, blkIdx);
-  }
-
-  ESTIM_BITS_END("encode_transform_tree");
-}
-
-
-void encode_cu_skip_flag(encoder_context* ectx,
-                         CABAC_encoder* cabac,
-                         const enc_cb* cb,
-                         bool skip)
-{
-  logtrace(LogSymbols,"$1 cu_skip_flag=%d\n",skip);
-
-  const de265_image* img = ectx->img;
-
-  int x0 = cb->x;
-  int y0 = cb->y;
-
-  // check if neighbors are available
-
-  int availableL = check_CTB_available(img, x0,y0, x0-1,y0);
-  int availableA = check_CTB_available(img, x0,y0, x0,y0-1);
-
-  int condL = 0;
-  int condA = 0;
-
-  if (availableL && ectx->ctbs.getCB(x0-1,y0)->PredMode == MODE_SKIP) condL=1;
-  if (availableA && ectx->ctbs.getCB(x0,y0-1)->PredMode == MODE_SKIP) condA=1;
-
-  int contextOffset = condL + condA;
-  int context = contextOffset;
-
-  // decode bit
-
-  int bit = skip;
-
-  logtrace(LogSlice,"> cu_skip_flag ctx=%d, bit=%d\n", context,bit);
-
-  cabac->write_CABAC_bit(CONTEXT_MODEL_CU_SKIP_FLAG + context, bit);
-}
-
-
-void encode_merge_idx(encoder_context* ectx,
-                      CABAC_encoder* cabac,
-                      int mergeIdx)
-{
-  logtrace(LogSymbols,"$1 merge_idx=%d\n",mergeIdx);
-  logtrace(LogSlice,"# merge_idx %d\n", mergeIdx);
-
-  if (ectx->shdr->MaxNumMergeCand <= 1) {
-    return; // code nothing, we use only a single merge candidate
-  }
-
-  // TU coding, first bin is CABAC, remaining are bypass.
-  // cMax = MaxNumMergeCand-1
-
-  cabac->write_CABAC_bit(CONTEXT_MODEL_MERGE_IDX, mergeIdx ? 1 : 0);
-
-  if (mergeIdx>0) {
-    int idx=1;
-
-    while (idx<ectx->shdr->MaxNumMergeCand-1) {
-      int increase = (idx < mergeIdx);
-
-      cabac->write_CABAC_bypass(increase);
-      if (increase) {
-        idx++;
-      }
-      else {
-        break;
-      }
-    }
-  }
-}
-
-
-static inline void encode_rqt_root_cbf(encoder_context* ectx,
-                                       CABAC_encoder* cabac,
-                                       int rqt_root_cbf)
-{
-  logtrace(LogSymbols,"$1 rqt_root_cbf=%d\n",rqt_root_cbf);
-  cabac->write_CABAC_bit(CONTEXT_MODEL_RQT_ROOT_CBF, rqt_root_cbf);
-}
-
-
-void encode_mvd(encoder_context* ectx,
-                CABAC_encoder* cabac,
-                const int16_t mvd2)
-{
-  int mvd0abs = abs_value(mvd0);
-  int mvd1abs = abs_value(mvd1);
-
-  int mvd0_greater_0 = !!(mvd0abs);
-  int mvd1_greater_0 = !!(mvd1abs);
-
-  cabac->write_CABAC_bit(CONTEXT_MODEL_ABS_MVD_GREATER01_FLAG+0, mvd0_greater_0);
-  cabac->write_CABAC_bit(CONTEXT_MODEL_ABS_MVD_GREATER01_FLAG+0, mvd1_greater_0);
-
-  if (mvd0_greater_0) {
-    cabac->write_CABAC_bit(CONTEXT_MODEL_ABS_MVD_GREATER01_FLAG+1, mvd0abs>1);
-  }
-  if (mvd1_greater_0) {
-    cabac->write_CABAC_bit(CONTEXT_MODEL_ABS_MVD_GREATER01_FLAG+1, mvd1abs>1);
-  }
-
-  if (mvd0abs) {
-    if (mvd0abs>1) {
-      cabac->write_CABAC_EGk(mvd0abs-2,1);
-    }
-    cabac->write_CABAC_bypass(mvd0<0);
-  }
-
-  if (mvd1abs) {
-    if (mvd1abs>1) {
-      cabac->write_CABAC_EGk(mvd1abs-2,1);
-    }
-    cabac->write_CABAC_bypass(mvd1<0);
-  }
-}
-
-
-void encode_prediction_unit(encoder_context* ectx,
-                            CABAC_encoder* cabac,
-                            const enc_cb* cb, int pbIdx,
-                            int x0,int y0, int w, int h)
-{
-  const enc_pb_inter& pb = cb->inter.pbpbIdx;
-
-  logtrace(LogSymbols,"$1 merge_flag=%d\n",pb.spec.merge_flag);
-  cabac->write_CABAC_bit(CONTEXT_MODEL_MERGE_FLAG, pb.spec.merge_flag);
-
-  if (pb.spec.merge_flag) {
-    assert(false); // TODO
-  }
-  else {
-    if (ectx->shdr->slice_type == SLICE_TYPE_B) {
-      assert(false); // TODO
-    }
-
-    if (pb.spec.inter_pred_idc != PRED_L1) {
-      if (ectx->shdr->num_ref_idx_l0_active > 1) {
-        assert(false); // TODO
-      //cabac->write_CABAC_bit(CONTEXT_MODEL_REF_IDX_LX, pb.spec.mvp_l0_flag);
-      }
-
-      encode_mvd(ectx,cabac, pb.spec.mvd0);
-
-      logtrace(LogSymbols,"$1 mvp_lx_flag=%d\n",pb.spec.mvp_l0_flag);
-      cabac->write_CABAC_bit(CONTEXT_MODEL_MVP_LX_FLAG, pb.spec.mvp_l0_flag);
-    }
-
-    if (pb.spec.inter_pred_idc != PRED_L0) {
-      assert(false); // TODO
-    }
-
-    /*
-enum InterPredIdc
-    InterPredIdc::PRED_L0=0,
-    InterPredIdc::PRED_L1=1,
-    InterPredIdc::PRED_BI=2
-    */
-  }
-}
-
-
-void encode_coding_unit(encoder_context* ectx,
-                        CABAC_encoder* cabac,
-                        const enc_cb* cb, int x0,int y0, int log2CbSize, bool recurse)
-{
-  logtrace(LogSlice,"--- encode CU (%d;%d) ---\n",x0,y0);
-
-  de265_image* img = ectx->img;
-  const slice_segment_header* shdr = &ectx->imgdata->shdr;
-  const seq_parameter_set& sps = ectx->img->get_sps();
-
-
-  int nCbS = 1<<log2CbSize;
-
-
-  // write skip_flag
-
-  if (shdr->slice_type != SLICE_TYPE_I) {
-    encode_cu_skip_flag(ectx,cabac, cb, cb->PredMode==MODE_SKIP);
-  }
-
-  if (cb->PredMode==MODE_SKIP) {
-    assert(cb->inter.pb0.spec.merge_flag);
-    encode_merge_idx(ectx,cabac, cb->inter.pb0.spec.merge_idx);
-  }
-  else {
-
-    enum PredMode PredMode = cb->PredMode;
-    enum PartMode PartMode = PART_2Nx2N;
-    int IntraSplitFlag=0;
-
-    if (shdr->slice_type != SLICE_TYPE_I) {
-      encode_pred_mode_flag(ectx,cabac, PredMode);
-    }
-
-    if (PredMode != MODE_INTRA ||
-        log2CbSize == sps.Log2MinCbSizeY) {
-      PartMode = cb->PartMode;
-      encode_part_mode(ectx,cabac, PredMode, PartMode, log2CbSize);
-    }
-
-    if (PredMode == MODE_INTRA) {
-
-      assert(cb->split_cu_flag == 0);
-
-      int availableA0 = check_CTB_available(img, x0,y0, x0-1,y0);
-      int availableB0 = check_CTB_available(img, x0,y0, x0,y0-1);
-
-      if (PartMode==PART_2Nx2N) {
-        logtrace(LogSlice,"x0,y0: %d,%d\n",x0,y0);
-        int PUidx = (x0>>sps.Log2MinPUSize) + (y0>>sps.Log2MinPUSize)*sps.PicWidthInMinPUs;
-
-        enum IntraPredMode candModeList3;
-        fillIntraPredModeCandidates(candModeList,x0,y0,
-                                    availableA0,availableB0, ectx->ctbs, &sps);
-
-        for (int i=0;i<3;i++)
-          logtrace(LogSlice,"candModeList%d = %d\n", i, candModeListi);
-
-        enum IntraPredMode mode = cb->transform_tree->intra_mode;
-
-        int intraPred = find_intra_pred_mode(mode, candModeList);
-        encode_prev_intra_luma_pred_flag(ectx,cabac, intraPred);
-        encode_intra_mpm_or_rem(ectx,cabac, intraPred);
-
-        logtrace(LogSlice,"IntraPredMode: %d (candidates: %d %d %d)\n", mode,
-                 candModeList0, candModeList1, candModeList2);
-        logtrace(LogSlice,"  MPM/REM = %d\n",intraPred);
-
-
-        IntraChromaPredMode chromaPredMode;
-        chromaPredMode = find_chroma_pred_mode(cb->transform_tree->intra_mode_chroma,
-                                               cb->transform_tree->intra_mode);
-        encode_intra_chroma_pred_mode(ectx,cabac, chromaPredMode);
-      }
-      else {
-        IntraSplitFlag=1;
-
-        int pbOffset = nCbS/2;
-        int PUidx;
-
-        int intraPred4;
-        int childIdx=0;
-
-        for (int j=0;j<nCbS;j+=pbOffset)
-          for (int i=0;i<nCbS;i+=pbOffset, childIdx++)
-            {
-              int x=x0+i, y=y0+j;
-
-              int availableA = availableA0 || (i>0); // left candidate always available for right blk
-              int availableB = availableB0 || (j>0); // top candidate always available for bottom blk
-
-              PUidx = (x>>sps.Log2MinPUSize) + (y>>sps.Log2MinPUSize)*sps.PicWidthInMinPUs;
-
-              enum IntraPredMode candModeList3;
-              fillIntraPredModeCandidates(candModeList,x,y,
-                                          availableA,availableB, ectx->ctbs, &sps);
-
-              for (int i=0;i<3;i++)
-                logtrace(LogSlice,"candModeList%d = %d\n", i, candModeListi);
-
-              enum IntraPredMode mode = cb->transform_tree->childrenchildIdx->intra_mode;
-
-              intraPredchildIdx = find_intra_pred_mode(mode, candModeList);
-
-              logtrace(LogSlice,"IntraPredMode: %d (candidates: %d %d %d)\n", mode,
-                       candModeList0, candModeList1, candModeList2);
-              logtrace(LogSlice,"  MPM/REM = %d\n",intraPredchildIdx);
-            }
-
-        for (int i=0;i<4;i++)
-          encode_prev_intra_luma_pred_flag(ectx,cabac, intraPredi);
-
-        for (int i=0;i<4;i++) {
-          encode_intra_mpm_or_rem(ectx,cabac, intraPredi);
-        }
-
-
-        // send chroma mode
-
-        if (sps.ChromaArrayType == CHROMA_444) {
-          for (int i=0;i<4;i++) {
-            IntraChromaPredMode chromaPredMode;
-            chromaPredMode = find_chroma_pred_mode(cb->transform_tree->childreni->intra_mode_chroma,
-                                                   cb->transform_tree->childreni->intra_mode);
-            encode_intra_chroma_pred_mode(ectx,cabac, chromaPredMode);
-          }
-        }
-        else {
-          IntraChromaPredMode chromaPredMode;
-          chromaPredMode = find_chroma_pred_mode(cb->transform_tree->children0->intra_mode_chroma,
-                                                 cb->transform_tree->children0->intra_mode);
-          encode_intra_chroma_pred_mode(ectx,cabac, chromaPredMode);
-        }
-      }
-
-      /*
-      printf("write intra modes. Luma=%d Chroma=%d\n",
-             cb->intra.pred_mode0,
-             cb->intra.chroma_mode);
-      */
-    }
-    else {
-      switch (cb->PartMode) {
-      case PART_2Nx2N:
-        encode_prediction_unit(ectx,cabac,cb, 0, cb->x,cb->y,1<<cb->log2Size,1<<cb->log2Size);
-        break;
-      case PART_2NxN:
-      case PART_Nx2N:
-      case PART_NxN:
-      case PART_2NxnU:
-      case PART_2NxnD:
-      case PART_nLx2N:
-      case PART_nRx2N:
-        assert(false); // TODO
-      }
-    }
-
-
-    if (true) { // !pcm
-
-      if (cb->PredMode != MODE_INTRA &&
-          !(cb->PartMode == PART_2Nx2N && cb->inter.pb0.spec.merge_flag)) {
-
-        //printf("%d %d %d\n",cb->PredMode,cb->PartMode,cb->inter.pb0.merge_flag);
-
-        encode_rqt_root_cbf(ectx,cabac, cb->inter.rqt_root_cbf);
-      }
-
-      //printf("%d;%d encode rqt_root_cbf=%d\n",x0,y0,cb->inter.rqt_root_cbf);
-
-      if (cb->PredMode == MODE_INTRA || cb->inter.rqt_root_cbf) {
-        int MaxTrafoDepth;
-        if (PredMode == MODE_INTRA)
-          { MaxTrafoDepth = sps.max_transform_hierarchy_depth_intra + IntraSplitFlag; }
-        else
-          { MaxTrafoDepth = sps.max_transform_hierarchy_depth_inter; }
-
-
-        if (recurse) {
-          //printf("%d;%d store transform tree\n",x0,y0);
-
-          encode_transform_tree(ectx,cabac, cb->transform_tree, cb,
-                                x0,y0, x0,y0, log2CbSize, 0, 0, MaxTrafoDepth, IntraSplitFlag, true);
-        }
-      }
-    }
-  }
-}
-
-
-SplitType get_split_type(const seq_parameter_set* sps,
-                         int x0,int y0, int log2CbSize)
-{
-  /*
-    CU split flag:
-
-         | overlaps | minimum ||
-    case | border   | size    ||  split
-    -----+----------+---------++----------
-      A  |    0     |     0   || optional
-      B  |    0     |     1   ||    0
-      C  |    1     |     0   ||    1
-      D  |    1     |     1   ||    0
-  */
-  if (x0+(1<<log2CbSize) <= sps->pic_width_in_luma_samples &&
-      y0+(1<<log2CbSize) <= sps->pic_height_in_luma_samples &&
-      log2CbSize > sps->Log2MinCbSizeY) {
-
-    // case A
-
-    return OptionalSplit;
-  } else {
-    // case B/C/D
-
-    if (log2CbSize > sps->Log2MinCbSizeY) { return ForcedSplit;    }
-    else                                  { return ForcedNonSplit; }
-  }
-}
-
-
-void encode_quadtree(encoder_context* ectx,
-                     CABAC_encoder* cabac,
-                     const enc_cb* cb, int x0,int y0, int log2CbSize, int ctDepth,
-                     bool recurse)
-{
-  //de265_image* img = ectx->img;
-  const seq_parameter_set& sps = ectx->img->get_sps();
-
-  int split_flag = get_split_type(&sps,x0,y0,log2CbSize);
-
-  // if it is an optional split, take the decision from the CU flag
-  if (split_flag == OptionalSplit) {
-    split_flag = cb->split_cu_flag;
-
-    encode_split_cu_flag(ectx,cabac, x0,y0, ctDepth, split_flag);
-  }
-
-
-  if (split_flag) {
-    if (recurse) {
-      int x1 = x0 + (1<<(log2CbSize-1));
-      int y1 = y0 + (1<<(log2CbSize-1));
-
-      encode_quadtree(ectx,cabac, cb->children0, x0,y0, log2CbSize-1, ctDepth+1, true);
-
-      if (x1<sps.pic_width_in_luma_samples)
-        encode_quadtree(ectx,cabac, cb->children1, x1,y0, log2CbSize-1, ctDepth+1, true);
-
-      if (y1<sps.pic_height_in_luma_samples)
-        encode_quadtree(ectx,cabac, cb->children2, x0,y1, log2CbSize-1, ctDepth+1, true);
-
-      if (x1<sps.pic_width_in_luma_samples &&
-          y1<sps.pic_height_in_luma_samples)
-        encode_quadtree(ectx,cabac, cb->children3, x1,y1, log2CbSize-1, ctDepth+1, true);
-    }
-  }
-  else {
-    encode_coding_unit(ectx,cabac, cb,x0,y0, log2CbSize, true);
-  }
-}
-
-
-void encode_ctb(encoder_context* ectx,
-                CABAC_encoder* cabac,
-                enc_cb* cb, int ctbX,int ctbY)
-{
-  logtrace(LogSlice,"----- encode CTB (%d;%d) -----\n",ctbX,ctbY);
-
-#if 0
-  printf("MODEL:\n");
-  for (int i=0;i<CONTEXT_MODEL_TABLE_LENGTH;i++)
-    {
-      printf("%d;%d ",
-             ectx->ctx_modeli.state,
-             ectx->ctx_modeli.MPSbit);
-
-      if ((i%16)==15) printf("\n");
-    }
-  printf("\n");
-#endif
-
-  de265_image* img = ectx->img;
-  int log2ctbSize = img->get_sps().Log2CtbSizeY;
-
-  encode_quadtree(ectx,cabac, cb, ctbX<<log2ctbSize, ctbY<<log2ctbSize, log2ctbSize, 0, true);
-}
-
-
-// ---------------------------------------------------------------------------

libde265-1.0.17.tar.gz/libde265/encoder/encoder-syntax.h Deleted

@@ -1,102 +0,0 @@
-/*
- * H.265 video codec.
- * Copyright (c) 2013-2014 struktur AG, Dirk Farin <farin@struktur.de>
- *
- * Authors: Dirk Farin <farin@struktur.de>
- *
- * This file is part of libde265.
- *
- * libde265 is free software: you can redistribute it and/or modify
- * it under the terms of the GNU Lesser General Public License as
- * published by the Free Software Foundation, either version 3 of
- * the License, or (at your option) any later version.
- *
- * libde265 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 Lesser General Public License for more details.
- *
- * You should have received a copy of the GNU Lesser General Public License
- * along with libde265.  If not, see <http://www.gnu.org/licenses/>.
- */
-
-#ifndef ENCODER_SYNTAX_H
-#define ENCODER_SYNTAX_H
-
-#include "libde265/image.h"
-#include "libde265/encoder/encoder-types.h"
-
-
-void encode_split_cu_flag(encoder_context* ectx,
-                          CABAC_encoder* cabac,
-                          int x0, int y0, int ctDepth, int split_flag);
-
-void encode_transform_tree(encoder_context* ectx,
-                           CABAC_encoder* cabac,
-                           const enc_tb* tb, const enc_cb* cb,
-                           int x0,int y0, int xBase,int yBase,
-                           int log2TrafoSize, int trafoDepth, int blkIdx,
-                           int MaxTrafoDepth, int IntraSplitFlag, bool recurse);
-
-void encode_coding_unit(encoder_context* ectx,
-                        CABAC_encoder* cabac,
-                        const enc_cb* cb, int x0,int y0, int log2CbSize, bool recurse);
-
-/* returns
-   1  - forced split
-   0  - forced non-split
-   -1 - optional split
-*/
-enum SplitType {
-  ForcedNonSplit = 0,
-  ForcedSplit    = 1,
-  OptionalSplit  = 2
-};
-
-SplitType get_split_type(const seq_parameter_set* sps,
-                         int x0,int y0, int log2CbSize);
-
-
-/* Compute how much rate is required for sending the chroma CBF flags
-   in the whole TB tree.
- */
-float recursive_cbfChroma_rate(CABAC_encoder_estim* cabac,
-                               enc_tb* tb, int log2TrafoSize, int trafoDepth);
-
-
-void encode_split_transform_flag(encoder_context* ectx,
-                                 CABAC_encoder* cabac,
-                                 int log2TrafoSize, int split_flag);
-
-void encode_merge_idx(encoder_context* ectx,
-                      CABAC_encoder* cabac,
-                      int mergeIdx);
-
-void encode_cu_skip_flag(encoder_context* ectx,
-                         CABAC_encoder* cabac,
-                         const enc_cb* cb,
-                         bool skip);
-
-void encode_cbf_luma(CABAC_encoder* cabac,
-                     bool zeroTrafoDepth, int cbf_luma);
-
-void encode_cbf_chroma(CABAC_encoder* cabac,
-                       int trafoDepth, int cbf_chroma);
-
-void encode_transform_unit(encoder_context* ectx,
-                           CABAC_encoder* cabac,
-                           const enc_tb* tb, const enc_cb* cb,
-                           int x0,int y0, int xBase,int yBase,
-                           int log2TrafoSize, int trafoDepth, int blkIdx);
-
-
-void encode_quadtree(encoder_context* ectx,
-                     CABAC_encoder* cabac,
-                     const enc_cb* cb, int x0,int y0, int log2CbSize, int ctDepth,
-                     bool recurse);
-
-void encode_ctb(encoder_context* ectx,
-                CABAC_encoder* cabac,
-                enc_cb* cb, int ctbX,int ctbY);
-
-#endif

libde265-1.0.17.tar.gz/libde265/encoder/encoder-types.cc Deleted

@@ -1,766 +0,0 @@
-/*
- * H.265 video codec.
- * Copyright (c) 2013-2014 struktur AG, Dirk Farin <farin@struktur.de>
- *
- * Authors: struktur AG, Dirk Farin <farin@struktur.de>
- *
- * This file is part of libde265.
- *
- * libde265 is free software: you can redistribute it and/or modify
- * it under the terms of the GNU Lesser General Public License as
- * published by the Free Software Foundation, either version 3 of
- * the License, or (at your option) any later version.
- *
- * libde265 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 Lesser General Public License for more details.
- *
- * You should have received a copy of the GNU Lesser General Public License
- * along with libde265.  If not, see <http://www.gnu.org/licenses/>.
- */
-
-#include "encoder-types.h"
-#include "encoder-context.h"
-#include "slice.h"
-#include "scan.h"
-#include "intrapred.h"
-#include "libde265/transform.h"
-#include "libde265/fallback-dct.h"
-#include <iostream>
-
-
-int allocTB = 0;
-int allocCB = 0;
-
-#define DEBUG_ALLOCS 0
-
-
-small_image_buffer::small_image_buffer(int log2Size,int bytes_per_pixel)
-{
-  mWidth  = 1<<log2Size;
-  mHeight = 1<<log2Size;
-  mStride = 1<<log2Size;
-  mBytesPerRow = bytes_per_pixel * (1<<log2Size);
-
-  int nBytes = mWidth*mHeight*bytes_per_pixel;
-  mBuf = new uint8_tnBytes;
-}
-
-
-small_image_buffer::~small_image_buffer()
-{
-  delete mBuf;
-}
-
-
-void enc_cb::set_rqt_root_bf_from_children_cbf()
-{
-  assert(transform_tree);
-  inter.rqt_root_cbf = (transform_tree->cbf0 |
-                        transform_tree->cbf1 |
-                        transform_tree->cbf2);
-}
-
-
-
-
-alloc_pool enc_tb::mMemPool(sizeof(enc_tb));
-
-enc_tb::enc_tb(int x,int y,int log2TbSize, enc_cb* _cb)
-  : enc_node(x,y,log2TbSize)
-{
-  parent = NULL;
-  cb = _cb;
-  downPtr = NULL;
-  blkIdx = 0;
-
-  split_transform_flag = false;
-  coeff0=coeff1=coeff2=NULL;
-
-  TrafoDepth = 0;
-  cbf0 = cbf1 = cbf2 = 0;
-
-  distortion = 0;
-  rate = 0;
-  rate_withoutCbfChroma = 0;
-
-  intra_mode = INTRA_PLANAR;
-  intra_mode_chroma = INTRA_PLANAR;
-
-  if (DEBUG_ALLOCS) { allocTB++; printf("TB  : %d\n",allocTB); }
-}
-
-
-enc_tb::~enc_tb()
-{
-  if (split_transform_flag) {
-    for (int i=0;i<4;i++) {
-      delete childreni;
-    }
-  }
-  else {
-    for (int i=0;i<3;i++) {
-      delete coeffi;
-    }
-  }
-
-  if (DEBUG_ALLOCS) { allocTB--; printf("TB ~: %d\n",allocTB); }
-}
-
-
-void enc_tb::alloc_coeff_memory(int cIdx, int tbSize)
-{
-  assert(coeffcIdx==NULL);
-  coeffcIdx = new int16_ttbSize*tbSize;
-}
-
-
-void enc_tb::reconstruct_tb(encoder_context* ectx,
-                            de265_image* img,
-                            int x0,int y0,  // luma
-                            int log2TbSize, // chroma adapted
-                            int cIdx) const
-{
-  // chroma adapted position
-  int xC=x0;
-  int yC=y0;
-
-  if (cIdx>0 && ectx->get_sps().chroma_format_idc == CHROMA_420) {
-    xC>>=1;
-    yC>>=1;
-  }
-
-
-  if (!reconstructioncIdx) {
-
-    reconstructioncIdx = std::make_shared<small_image_buffer>(log2TbSize, sizeof(uint8_t));
-
-    if (cb->PredMode == MODE_SKIP) {
-      PixelAccessor dstPixels(*reconstructioncIdx, xC,yC);
-      dstPixels.copyFromImage(img, cIdx);
-    }
-    else { // not SKIP mode
-      if (cb->PredMode == MODE_INTRA) {
-
-        //enum IntraPredMode intraPredMode  = img->get_IntraPredMode(x0,y0);
-        enum IntraPredMode intraPredMode  = intra_mode;
-
-        if (cIdx>0) {
-          intraPredMode = intra_mode_chroma;
-        }
-
-        //printf("reconstruct TB (%d;%d): intra mode (cIdx=%d) = %d\n",xC,yC,cIdx,intraPredMode);
-
-        //decode_intra_prediction(img, xC,yC,  intraPredMode, 1<< log2TbSize   , cIdx);
-
-        //printf("access intra-prediction of TB %p\n",this);
-
-        intra_predictioncIdx->copy_to(*reconstructioncIdx);
-        /*
-          copy_subimage(img->get_image_plane_at_pos(cIdx,xC,yC),
-          img->get_image_stride(cIdx),
-          intra_predictioncIdx->get_buffer<uint8_t>(), 1<<log2TbSize,
-          1<<log2TbSize, 1<<log2TbSize);
-        */
-      }
-      else {
-        assert(0); // -> TODO: now only store in tb_enc
-
-        int size = 1<<log2TbSize;
-
-        uint8_t* dst_ptr  = img->get_image_plane_at_pos(cIdx, xC,  yC  );
-        int dst_stride  = img->get_image_stride(cIdx);
-
-        /* TODO TMP HACK: prediction does not exist anymore
-        uint8_t* src_ptr  = ectx->prediction->get_image_plane_at_pos(cIdx, xC,  yC  );
-        int src_stride  = ectx->prediction->get_image_stride(cIdx);
-
-        for (int y=0;y<size;y++) {
-          for (int x=0;x<size;x++) {
-            dst_ptry*dst_stride+x = src_ptry*src_stride+x;
-          }
-        }
-        */
-      }
-
-      ALIGNED_16(int16_t) dequant_coeff32*32;
-
-      if (cbfcIdx) dequant_coefficients(dequant_coeff, coeffcIdx, log2TbSize, cb->qp);
-
-      if (0 && cbfcIdx) {
-        printf("--- quantized coeffs ---\n");
-        printBlk("qcoeffs",coeff0,1<<log2TbSize,1<<log2TbSize);
-
-        printf("--- dequantized coeffs ---\n");
-        printBlk("dequant",dequant_coeff,1<<log2TbSize,1<<log2TbSize);
-      }
-
-#if 0
-      uint8_t* ptr  = img->get_image_plane_at_pos(cIdx, xC,  yC  );
-      int stride  = img->get_image_stride(cIdx);
-#endif
-
-      int trType = (cIdx==0 && log2TbSize==2); // TODO: inter
-
-      //printf("--- prediction %d %d / %d ---\n",x0,y0,cIdx);
-      //printBlk("prediction",ptr,1<<log2TbSize,stride);
-
-      if (cbfcIdx) inv_transform(&ectx->acceleration,
-                                   reconstructioncIdx->get_buffer<uint8_t>(), 1<<log2TbSize,
-                                   dequant_coeff, log2TbSize,   trType);
-
-      //printBlk("RECO",reconstructioncIdx->get_buffer_u8(),1<<log2TbSize,
-      //         reconstructioncIdx->getStride());
-    }
-  }
-
-
-  // copy reconstruction into image
-
-#if 0
-  copy_subimage(img->get_image_plane_at_pos(cIdx,xC,yC),
-                img->get_image_stride(cIdx),
-                reconstructioncIdx->get_buffer<uint8_t>(), 1<<log2TbSize,
-                1<<log2TbSize, 1<<log2TbSize);
-#endif
-
-  //printf("--- RECO intra prediction %d %d ---\n",x0,y0);
-  //printBlk("RECO",ptr,1<<log2TbSize,stride);
-}
-
-
-void enc_tb::debug_writeBlack(encoder_context* ectx, de265_image* img) const
-{
-  if (split_transform_flag) {
-    for (int i=0;i<4;i++) {
-      childreni->debug_writeBlack(ectx,img);
-    }
-  }
-  else {
-    //reconstruct_tb(ectx, img, x,y, log2Size, 0);
-
-    int size = 1<<(log2Size<<1);
-    std::vector<uint8_t> buf(size);
-    memset(&buf0,0x12,size);
-
-    int cIdx=0;
-    int xC=x,yC=y;
-
-    copy_subimage(img->get_image_plane_at_pos(cIdx,xC,yC),
-                  img->get_image_stride(cIdx),
-                  &buf0, 1<<log2Size,
-                  1<<log2Size, 1<<log2Size);
-  }
-}
-
-
-void enc_tb::reconstruct(encoder_context* ectx, de265_image* img) const
-{
-  if (split_transform_flag) {
-    for (int i=0;i<4;i++) {
-      childreni->reconstruct(ectx,img);
-    }
-  }
-  else {
-    reconstruct_tb(ectx, img, x,y, log2Size, 0);
-
-    if (ectx->get_sps().chroma_format_idc == CHROMA_444) {
-      reconstruct_tb(ectx, img, x,y, log2Size, 1);
-      reconstruct_tb(ectx, img, x,y, log2Size, 2);
-    }
-    else if (log2Size>2) {
-      reconstruct_tb(ectx, img, x,y, log2Size-1, 1);
-      reconstruct_tb(ectx, img, x,y, log2Size-1, 2);
-    }
-    else if (blkIdx==3) {
-      int xBase = x - (1<<log2Size);
-      int yBase = y - (1<<log2Size);
-
-      reconstruct_tb(ectx, img, xBase,yBase, log2Size, 1);
-      reconstruct_tb(ectx, img, xBase,yBase, log2Size, 2);
-    }
-  }
-}
-
-
-void enc_tb::set_cbf_flags_from_children()
-{
-  assert(split_transform_flag);
-
-  cbf0 = 0;
-  cbf1 = 0;
-  cbf2 = 0;
-
-  for (int i=0;i<4;i++) {
-    cbf0 |= childreni->cbf0;
-    cbf1 |= childreni->cbf1;
-    cbf2 |= childreni->cbf2;
-  }
-}
-
-
-
-alloc_pool enc_cb::mMemPool(sizeof(enc_cb), 200);
-
-
-enc_cb::enc_cb()
-  : split_cu_flag(false),
-    cu_transquant_bypass_flag(false),
-    pcm_flag(false),
-    transform_tree(NULL),
-    distortion(0),
-    rate(0)
-{
-  parent = NULL;
-  downPtr = NULL;
-
-  if (DEBUG_ALLOCS) { allocCB++; printf("CB  : %d\n",allocCB); }
-}
-
-enc_cb::~enc_cb()
-{
-  if (split_cu_flag) {
-    for (int i=0;i<4;i++) {
-      delete childreni;
-    }
-  }
-  else {
-    delete transform_tree;
-  }
-
-  if (DEBUG_ALLOCS) { allocCB--; printf("CB ~: %d\n",allocCB); }
-}
-
-
-/*
-void enc_cb::write_to_image(de265_image* img) const
-{
-  //printf("write_to_image %d %d size:%d\n",x,y,1<<log2Size);
-
-
-  if (!split_cu_flag) {
-    img->set_log2CbSize(x,y,log2Size, true);
-    img->set_ctDepth(x,y,log2Size, ctDepth);
-    assert(pcm_flag==0);
-    img->set_pcm_flag(x,y,log2Size, pcm_flag);
-    img->set_cu_transquant_bypass(x,y,log2Size, cu_transquant_bypass_flag);
-    img->set_QPY(x,y,log2Size, qp);
-    img->set_pred_mode(x,y, log2Size, PredMode);
-    img->set_PartMode(x,y, PartMode);
-
-    if (PredMode == MODE_INTRA) {
-
-      if (PartMode == PART_NxN) {
-        int h = 1<<(log2Size-1);
-        img->set_IntraPredMode(x  ,y  ,log2Size-1, transform_tree->children0->intra_mode);
-        img->set_IntraPredMode(x+h,y  ,log2Size-1, transform_tree->children1->intra_mode);
-        img->set_IntraPredMode(x  ,y+h,log2Size-1, transform_tree->children2->intra_mode);
-        img->set_IntraPredMode(x+h,y+h,log2Size-1, transform_tree->children3->intra_mode);
-      }
-      else {
-        img->set_IntraPredMode(x,y,log2Size, transform_tree->intra_mode);
-      }
-    }
-    else {
-      int nC = 1<<log2Size;
-      int nC2 = nC>>1;
-      int nC4 = nC>>2;
-      int nC3 = nC-nC4;
-      switch (PartMode) {
-      case PART_2Nx2N:
-        img->set_mv_info(x,y,nC,nC, inter.pb0.motion);
-        break;
-      case PART_NxN:
-        img->set_mv_info(x    ,y    ,nC2,nC2, inter.pb0.motion);
-        img->set_mv_info(x+nC2,y    ,nC2,nC2, inter.pb1.motion);
-        img->set_mv_info(x    ,y+nC2,nC2,nC2, inter.pb2.motion);
-        img->set_mv_info(x+nC2,y+nC2,nC2,nC2, inter.pb3.motion);
-        break;
-      case PART_2NxN:
-        img->set_mv_info(x,y    ,nC,nC2, inter.pb0.motion);
-        img->set_mv_info(x,y+nC2,nC,nC2, inter.pb1.motion);
-        break;
-      case PART_Nx2N:
-        img->set_mv_info(x    ,y,nC2,nC, inter.pb0.motion);
-        img->set_mv_info(x+nC2,y,nC2,nC, inter.pb1.motion);
-        break;
-      case PART_2NxnU:
-        img->set_mv_info(x,y    ,nC,nC4, inter.pb0.motion);
-        img->set_mv_info(x,y+nC4,nC,nC3, inter.pb1.motion);
-        break;
-      case PART_2NxnD:
-        img->set_mv_info(x,y    ,nC,nC3, inter.pb0.motion);
-        img->set_mv_info(x,y+nC3,nC,nC4, inter.pb1.motion);
-        break;
-      case PART_nLx2N:
-        img->set_mv_info(x    ,y,nC4,nC, inter.pb0.motion);
-        img->set_mv_info(x+nC4,y,nC3,nC, inter.pb1.motion);
-        break;
-      case PART_nRx2N:
-        img->set_mv_info(x    ,y,nC3,nC, inter.pb0.motion);
-        img->set_mv_info(x+nC3,y,nC4,nC, inter.pb1.motion);
-        break;
-      }
-    }
-  }
-  else {
-    for (int i=0;i<4;i++) {
-      if (childreni) {
-        childreni->write_to_image(img);
-      }
-    }
-  }
-}
-*/
-
-void enc_cb::reconstruct(encoder_context* ectx, de265_image* img) const
-{
-  assert(0);
-  if (split_cu_flag) {
-    for (int i=0;i<4;i++) {
-      childreni->reconstruct(ectx, img);
-    }
-  }
-  else {
-    //write_to_image(img);
-    transform_tree->reconstruct(ectx,img);
-  }
-}
-
-
-void enc_cb::debug_dumpTree(int flags, int indent) const
-{
-  std::string indentStr;
-  indentStr.insert(0,indent,' ');
-
-  std::cout << indentStr << "CB " << x << ";" << y << " "
-            << (1<<log2Size) << "x" << (1<<log2Size) << " " << this << "\n";
-
-  std::cout << indentStr << "| split_cu_flag: " << int(split_cu_flag) << "\n";
-  std::cout << indentStr << "| ctDepth:       " << int(ctDepth) << "\n";
-
-  if (split_cu_flag) {
-    for (int i=0;i<4;i++)
-      if (childreni) {
-        std::cout << indentStr << "| child CB " << i << ":\n";
-        childreni->debug_dumpTree(flags, indent+2);
-      }
-  }
-  else {
-    std::cout << indentStr << "| qp: " << int(qp) << "\n";
-    std::cout << indentStr << "| PredMode: " << PredMode << "\n";
-    std::cout << indentStr << "| PartMode: " << part_mode_name(PartMode) << "\n";
-    std::cout << indentStr << "| transform_tree:\n";
-
-    transform_tree->debug_dumpTree(flags, indent+2);
-  }
-}
-
-
-void enc_tb::debug_dumpTree(int flags, int indent) const
-{
-  std::string indentStr;
-  indentStr.insert(0,indent,' ');
-
-  std::cout << indentStr << "TB " << x << ";" << y << " "
-            << (1<<log2Size) << "x" << (1<<log2Size) << " " << this << "\n";
-
-  std::cout << indentStr << "| split_transform_flag: " << int(split_transform_flag) << "\n";
-  std::cout << indentStr << "| TrafoDepth:           " << int(TrafoDepth) << "\n";
-  std::cout << indentStr << "| blkIdx:               " << int(blkIdx) << "\n";
-
-  std::cout << indentStr << "| intra_mode:           " << int(intra_mode) << "\n";
-  std::cout << indentStr << "| intra_mode_chroma:    " << int(intra_mode_chroma) << "\n";
-
-  std::cout << indentStr << "| cbf:                  "
-            << int(cbf0) << ":"
-            << int(cbf1) << ":"
-            << int(cbf2) << "\n";
-
-
-  if (flags & DUMPTREE_RECONSTRUCTION) {
-    for (int i=0;i<3;i++)
-      if (reconstructioni) {
-        std::cout << indentStr << "| Reconstruction, channel " << i << ":\n";
-        printBlk(NULL,
-                 reconstructioni->get_buffer_u8(),
-                 reconstructioni->getWidth(),
-                 reconstructioni->getStride(),
-                 indentStr + "| ");
-      }
-  }
-
-  if (flags & DUMPTREE_INTRA_PREDICTION) {
-    for (int i=0;i<3;i++)
-      if (intra_predictioni) {
-        //if (i==0) print_border(debug_intra_border+64, NULL, 1<<log2Size);
-
-        std::cout << indentStr << "| Intra prediction, channel " << i << ":\n";
-        printBlk(NULL,
-                 intra_predictioni->get_buffer_u8(),
-                 intra_predictioni->getWidth(),
-                 intra_predictioni->getStride(),
-                 indentStr + "| ");
-      }
-  }
-
-  if (split_transform_flag) {
-    for (int i=0;i<4;i++)
-      if (childreni) {
-        std::cout << indentStr << "| child TB " << i << ":\n";
-        childreni->debug_dumpTree(flags, indent+2);
-      }
-  }
-}
-
-
-
-const enc_tb* enc_cb::getTB(int x,int y) const
-{
-  assert(!split_cu_flag);
-  assert(transform_tree);
-
-  return transform_tree->getTB(x,y);
-}
-
-
-const enc_tb* enc_tb::getTB(int px,int py) const
-{
-  if (split_transform_flag) {
-    int xHalf = x + (1<<(log2Size-1));
-    int yHalf = y + (1<<(log2Size-1));
-
-    enc_tb* child;
-
-    if (px<xHalf) {
-      if (py<yHalf) {
-        child = children0;
-      }
-      else {
-        child = children2;
-      }
-    }
-    else {
-      if (py<yHalf) {
-        child = children1;
-      }
-      else {
-        child = children3;
-      }
-    }
-
-    if (!child) { return NULL; }
-
-    return child->getTB(px,py);
-  }
-
-  return this;
-}
-
-
-
-void CTBTreeMatrix::alloc(int w,int h, int log2CtbSize)
-{
-  free();
-
-  int ctbSize = 1<<log2CtbSize;
-
-  mWidthCtbs  = (w+ctbSize-1) >> log2CtbSize;
-  mHeightCtbs = (h+ctbSize-1) >> log2CtbSize;
-  mLog2CtbSize = log2CtbSize;
-
-  mCTBs.resize(mWidthCtbs * mHeightCtbs, NULL);
-}
-
-
-const enc_cb* CTBTreeMatrix::getCB(int px,int py) const
-{
-  int xCTB = px>>mLog2CtbSize;
-  int yCTB = py>>mLog2CtbSize;
-
-  int idx = xCTB + yCTB*mWidthCtbs;
-  assert(idx < mCTBs.size());
-
-  enc_cb* cb = mCTBsidx;
-  if (!cb) { return NULL; }
-
-  while (cb->split_cu_flag) {
-    int xHalf = cb->x + (1<<(cb->log2Size-1));
-    int yHalf = cb->y + (1<<(cb->log2Size-1));
-
-    if (px<xHalf) {
-      if (py<yHalf) {
-        cb = cb->children0;
-      }
-      else {
-        cb = cb->children2;
-      }
-    }
-    else {
-      if (py<yHalf) {
-        cb = cb->children1;
-      }
-      else {
-        cb = cb->children3;
-      }
-    }
-
-    if (!cb) { return NULL; }
-  }
-
-  return cb;
-}
-
-
-const enc_tb* CTBTreeMatrix::getTB(int x,int y) const
-{
-  const enc_cb* cb = getCB(x,y);
-  if (!cb) { return NULL; }
-
-  const enc_tb* tb = cb->transform_tree;
-  if (!tb) { return NULL; }
-
-  return tb->getTB(x,y);
-}
-
-
-const enc_pb_inter* CTBTreeMatrix::getPB(int x,int y) const
-{
-  const enc_cb* cb = getCB(x,y);
-
-  // TODO: get PB block based on partitioning
-
-  return &cb->inter.pb0;
-}
-
-
-void CTBTreeMatrix::writeReconstructionToImage(de265_image* img,
-                                               const seq_parameter_set* sps) const
-{
-  for (size_t i=0;i<mCTBs.size();i++) {
-    const enc_cb* cb = mCTBsi;
-    cb->writeReconstructionToImage(img, sps);
-  }
-}
-
-void enc_cb::writeReconstructionToImage(de265_image* img,
-                                        const seq_parameter_set* sps) const
-{
-  if (split_cu_flag) {
-    for (int i=0;i<4;i++) {
-      if (childreni) {
-        childreni->writeReconstructionToImage(img,sps);
-      }
-    }
-  }
-  else {
-    transform_tree->writeReconstructionToImage(img,sps);
-  }
-}
-
-void enc_tb::writeReconstructionToImage(de265_image* img,
-                                        const seq_parameter_set* sps) const
-{
-  if (split_transform_flag) {
-    for (int i=0;i<4;i++) {
-      if (childreni) {
-        childreni->writeReconstructionToImage(img,sps);
-      }
-    }
-  }
-  else {
-    // luma pixels
-
-    PixelAccessor lumaPixels(*reconstruction0, x,y);
-    lumaPixels.copyToImage(img, 0);
-
-    // chroma pixels
-
-    if (sps->chroma_format_idc == CHROMA_444) {
-      PixelAccessor chroma1Pixels(*reconstruction1, x,y);
-      chroma1Pixels.copyToImage(img, 1);
-      PixelAccessor chroma2Pixels(*reconstruction2, x,y);
-      chroma2Pixels.copyToImage(img, 2);
-    }
-    else if (log2Size>2) {
-      PixelAccessor chroma1Pixels(*reconstruction1, x>>1,y>>1);
-      chroma1Pixels.copyToImage(img, 1);
-      PixelAccessor chroma2Pixels(*reconstruction2, x>>1,y>>1);
-      chroma2Pixels.copyToImage(img, 2);
-
-      //reconstruct_tb(ectx, img, x,y, log2Size-1, 1);
-      //reconstruct_tb(ectx, img, x,y, log2Size-1, 2);
-    }
-    else if (blkIdx==3) {
-      int xBase = x - (1<<log2Size);
-      int yBase = y - (1<<log2Size);
-
-      PixelAccessor chroma1Pixels(*reconstruction1, xBase>>1,yBase>>1);
-      chroma1Pixels.copyToImage(img, 1);
-      PixelAccessor chroma2Pixels(*reconstruction2, xBase>>1,yBase>>1);
-      chroma2Pixels.copyToImage(img, 2);
-
-      //reconstruct_tb(ectx, img, xBase,yBase, log2Size, 1);
-      //reconstruct_tb(ectx, img, xBase,yBase, log2Size, 2);
-    }
-  }
-}
-
-
-PixelAccessor enc_tb::getPixels(int x,int y, int cIdx, const seq_parameter_set& sps)
-{
-  int xL = x <<  sps.get_chroma_shift_W(cIdx);
-  int yL = y <<  sps.get_chroma_shift_H(cIdx);
-
-  const enc_tb* tb = getTB(xL,yL);
-
-  if (cIdx==0 || sps.chroma_format_idc == CHROMA_444) {
-    return PixelAccessor(*tb->reconstructioncIdx, tb->x, tb->y);
-  }
-  else if (sps.chroma_format_idc == CHROMA_420) {
-    if (tb->log2Size > 2) {
-      return PixelAccessor(*tb->reconstructioncIdx,
-                           tb->x >> 1,
-                           tb->y >> 1);
-    }
-    else {
-      enc_tb* parent = tb->parent;
-      tb = parent->children3;
-
-      return PixelAccessor(*tb->reconstructioncIdx,
-                           parent->x >> 1,
-                           parent->y >> 1);
-    }
-  }
-  else {
-    assert(sps.chroma_format_idc == CHROMA_422);
-
-    assert(false); // not supported yet
-
-    return PixelAccessor::invalid();
-  }
-}
-
-
-void PixelAccessor::copyToImage(de265_image* img, int cIdx) const
-{
-  uint8_t* p = img->get_image_plane_at_pos(cIdx, mXMin, mYMin);
-  int stride = img->get_image_stride(cIdx);
-
-  for (int y=0;y<mHeight;y++) {
-    memcpy(p, mBase+mXMin+(y+mYMin)*mStride, mWidth);
-    p += stride;
-  }
-}
-
-void PixelAccessor::copyFromImage(const de265_image* img, int cIdx)
-{
-  const uint8_t* p = img->get_image_plane_at_pos(cIdx, mXMin, mYMin);
-  int stride = img->get_image_stride(cIdx);
-
-  for (int y=0;y<mHeight;y++) {
-    memcpy(mBase+mXMin+(y+mYMin)*mStride, p, mWidth);
-    p += stride;
-  }
-}

libde265-1.0.17.tar.gz/libde265/encoder/encoder-types.h Deleted

@@ -1,409 +0,0 @@
-/*
- * H.265 video codec.
- * Copyright (c) 2013-2014 struktur AG, Dirk Farin <farin@struktur.de>
- *
- * Authors: Dirk Farin <farin@struktur.de>
- *
- * This file is part of libde265.
- *
- * libde265 is free software: you can redistribute it and/or modify
- * it under the terms of the GNU Lesser General Public License as
- * published by the Free Software Foundation, either version 3 of
- * the License, or (at your option) any later version.
- *
- * libde265 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 Lesser General Public License for more details.
- *
- * You should have received a copy of the GNU Lesser General Public License
- * along with libde265.  If not, see <http://www.gnu.org/licenses/>.
- */
-
-#ifndef ENCODE_H
-#define ENCODE_H
-
-#include "libde265/image.h"
-#include "libde265/decctx.h"
-#include "libde265/image-io.h"
-#include "libde265/alloc_pool.h"
-
-#include <memory>
-
-class encoder_context;
-class enc_cb;
-
-
-class small_image_buffer
-{
- public:
-  explicit small_image_buffer(int log2Size,int bytes_per_pixel=1);
-  ~small_image_buffer();
-
-  uint8_t*  get_buffer_u8() const { return mBuf; }
-  int16_t*  get_buffer_s16() const { return (int16_t*)mBuf; }
-  uint16_t* get_buffer_u16() const { return (uint16_t*)mBuf; }
-  template <class pixel_t> pixel_t* get_buffer() const { return (pixel_t*)mBuf; }
-
-  void copy_to(small_image_buffer& b) const {
-    assert(b.mHeight==mHeight);
-    assert(b.mBytesPerRow==mBytesPerRow);
-    memcpy(b.mBuf, mBuf, mBytesPerRow*mHeight);
-  }
-
-  int getWidth() const { return mWidth; }
-  int getHeight() const { return mHeight; }
-
-  int getStride() const { return mStride; } // pixels per row
-
- private:
-  uint8_t*  mBuf;
-  uint16_t  mStride;
-  uint16_t  mBytesPerRow;
-
-  uint8_t   mWidth, mHeight;
-
-  // small_image_buffer cannot be copied
-
-  small_image_buffer(const small_image_buffer&) { assert(false); } // = delete;
-  small_image_buffer& operator=(const small_image_buffer&) { assert(false); return *this; } // = delete;
-};
-
-
-class enc_node
-{
-public:
-  enc_node() { }
-  enc_node(int _x,int _y, int _log2Size) : x(_x), y(_y), log2Size(_log2Size) { }
-  virtual ~enc_node() { }
-
-  uint16_t x,y;
-  uint8_t  log2Size : 3;
-
-
-  static const int DUMPTREE_INTRA_PREDICTION = (1<<0);
-  static const int DUMPTREE_RESIDUAL         = (1<<1);
-  static const int DUMPTREE_RECONSTRUCTION   = (1<<2);
-  static const int DUMPTREE_ALL              = 0xFFFF;
-
-  virtual void debug_dumpTree(int flags, int indent=0) const = 0;
-};
-
-
-class PixelAccessor
-{
- public:
-  PixelAccessor(small_image_buffer& buf, int x0,int y0) {
-    mBase = buf.get_buffer_u8();
-    mStride = buf.getStride();
-    mXMin = x0;
-    mYMin = y0;
-    mWidth = buf.getWidth();
-    mHeight= buf.getHeight();
-
-    mBase -= x0 + y0*mStride;
-  }
-
-  const uint8_t* operator(int y) const { return mBase+y*mStride; }
-
-  int getLeft() const { return mXMin; }
-  int getWidth() const { return mWidth; }
-  int getTop() const { return mYMin; }
-  int getHeight() const { return mHeight; }
-
-  void copyToImage(de265_image* img, int cIdx) const;
-  void copyFromImage(const de265_image* img, int cIdx);
-
-  static PixelAccessor invalid() {
-    return PixelAccessor();
-  }
-
- private:
-  uint8_t* mBase;
-  short   mStride;
-  short   mXMin,  mYMin;
-  uint8_t mWidth, mHeight;
-
-  PixelAccessor() {
-    mBase = NULL;
-    mStride = mXMin = mYMin = mWidth = mHeight = 0;
-  }
-};
-
-
-class enc_tb : public enc_node
-{
- public:
-  enc_tb(int x,int y,int log2TbSize, enc_cb* _cb);
-  ~enc_tb();
-
-  enc_tb* parent;
-  enc_cb* cb;
-  enc_tb** downPtr;
-
-  uint8_t split_transform_flag : 1;
-  uint8_t TrafoDepth : 2;  // 2 bits enough ? (TODO)
-  uint8_t blkIdx : 2;
-
-  enum IntraPredMode intra_mode;
-
-  // Note: in NxN partition mode, the chroma mode is always derived from
-  // the top-left child's intra mode (for chroma 4:2:0).
-  enum IntraPredMode intra_mode_chroma;
-
-  uint8_t cbf3;
-
-
-  /* intra_prediction and residual is filled in tb-split, because this is where we decide
-     on the final block-size the TB is coded with.
-   */
-  //mutable uint8_t debug_intra_border2*64+1;
-  std::shared_ptr<small_image_buffer> intra_prediction3;
-  std::shared_ptr<small_image_buffer> residual3;
-
-  /* Reconstruction is computed on-demand in writeMetadata().
-   */
-  mutable std::shared_ptr<small_image_buffer> reconstruction3;
-
-  union {
-    // split
-    struct {
-      enc_tb* children4;
-    };
-
-    // leaf node
-    struct {
-      int16_t* coeff3;
-
-      bool    skip_transform32;
-      uint8_t explicit_rdpcm32;
-    };
-  };
-
-  float distortion;  // total distortion for this level of the TB tree (including all children)
-  float rate;        // total rate for coding this TB level and all children
-  float rate_withoutCbfChroma;
-
-  void set_cbf_flags_from_children();
-
-  void reconstruct(encoder_context* ectx, de265_image* img) const;
-  void debug_writeBlack(encoder_context* ectx, de265_image* img) const;
-
-  bool isZeroBlock() const { return cbf0==false && cbf1==false && cbf2==false; }
-
-  void alloc_coeff_memory(int cIdx, int tbSize);
-
-  const enc_tb* getTB(int x,int y) const;
-
-  PixelAccessor getPixels(int x,int y, int cIdx, const seq_parameter_set& sps);
-
-  void writeReconstructionToImage(de265_image* img,
-                                  const seq_parameter_set* sps) const;
-
-  /*
-  static void* operator new(const size_t size) { return mMemPool.new_obj(size); }
-  static void operator delete(void* obj) { mMemPool.delete_obj(obj); }
-  */
-
-
-  virtual void debug_dumpTree(int flags, int indent=0) const;
-
-private:
-  static alloc_pool mMemPool;
-
-  void reconstruct_tb(encoder_context* ectx,
-                      de265_image* img, int x0,int y0, int log2TbSize,
-                      int cIdx) const;
-};
-
-
-struct enc_pb_inter
-{
-  /* absolute motion information (for MV-prediction candidates)
-   */
-  PBMotion         motion;
-
-  /* specification how to code the motion vector in the bitstream
-   */
-  PBMotionCoding   spec;
-
-
-  // NOT TRUE: refIdx in 'spec' is not used. It is taken from 'motion'
-  // Currently, information is duplicated. Same as with inter_pred_idc/predFlag.
-
-  /* SPEC:
-  int8_t  refIdx2; // not used
-  int16_t mvd22;
-
-  uint8_t inter_pred_idc : 2; // enum InterPredIdc
-  uint8_t mvp_l0_flag : 1;
-  uint8_t mvp_l1_flag : 1;
-  uint8_t merge_flag : 1;
-  uint8_t merge_idx  : 3;
-  */
-};
-
-
-class enc_cb : public enc_node
-{
-public:
-  enc_cb();
-  ~enc_cb();
-
-  enc_cb* parent;
-  enc_cb** downPtr;
-
-  uint8_t split_cu_flag : 1;
-  uint8_t ctDepth : 2;
-
-
-  union {
-    // split
-    struct {
-      enc_cb* children4;   // undefined when split_cu_flag==false
-    };
-
-    // non-split
-    struct {
-      uint8_t qp : 6;
-      uint8_t cu_transquant_bypass_flag : 1; // currently unused
-      uint8_t pcm_flag : 1;
-
-      enum PredMode PredMode; // : 6;
-      enum PartMode PartMode; // : 3;
-
-      union {
-        struct {
-          //enum IntraPredMode pred_mode4;
-          //enum IntraPredMode chroma_mode;
-        } intra;
-
-        struct {
-          enc_pb_inter pb4;
-
-          uint8_t rqt_root_cbf : 1;
-        } inter;
-      };
-
-      enc_tb* transform_tree;
-    };
-  };
-
-
-  float distortion;
-  float rate;
-
-
-  void set_rqt_root_bf_from_children_cbf();
-
-  /* Save CB reconstruction in the node and restore it again to the image.
-     Pixel data and metadata.
-   */
-  //virtual void save(const de265_image*);
-  //virtual void restore(de265_image*);
-
-
-  /* Decode this CB: pixel data and write metadata to image.
-   */
-  void reconstruct(encoder_context* ectx,de265_image* img) const;
-
-  // can only be called on the lowest-level CB (with TB-tree as its direct child)
-  const enc_tb* getTB(int x,int y) const;
-
-  void writeReconstructionToImage(de265_image* img,
-                                  const seq_parameter_set* sps) const;
-
-
-  virtual void debug_dumpTree(int flags, int indent=0) const;
-
-
-  // memory management
-
-  static void* operator new(const size_t size) {
-    void* p = mMemPool.new_obj(size);
-    //printf("ALLOC %p\n",p);
-    return p;
-  }
-  static void operator delete(void* obj) {
-    //printf("DELETE %p\n",obj);
-    mMemPool.delete_obj(obj);
-  }
-
- private:
-  //void write_to_image(de265_image*) const;
-
-  static alloc_pool mMemPool;
-};
-
-
-
-class CTBTreeMatrix
-{
- public:
- CTBTreeMatrix() : mWidthCtbs(0), mHeightCtbs(0), mLog2CtbSize(0) { }
-  ~CTBTreeMatrix() { free(); }
-
-  void alloc(int w,int h, int log2CtbSize);
-  void clear() { free(); }
-
-  void setCTB(int xCTB, int yCTB, enc_cb* ctb) {
-    int idx = xCTB + yCTB*mWidthCtbs;
-    assert(idx < mCTBs.size());
-    if (mCTBsidx) { delete mCTBsidx; }
-    mCTBsidx = ctb;
-  }
-
-  const enc_cb* getCTB(int xCTB, int yCTB) const {
-    int idx = xCTB + yCTB*mWidthCtbs;
-    assert(idx < mCTBs.size());
-    return mCTBsidx;
-  }
-
-  enc_cb** getCTBRootPointer(int x, int y) {
-    x >>= mLog2CtbSize;
-    y >>= mLog2CtbSize;
-
-    int idx = x + y*mWidthCtbs;
-    assert(idx < mCTBs.size());
-    return &mCTBsidx;
-  }
-
-  const enc_cb* getCB(int x,int y) const;
-  const enc_tb* getTB(int x,int y) const;
-  const enc_pb_inter* getPB(int x,int y) const;
-
-  void writeReconstructionToImage(de265_image* img,
-                                  const seq_parameter_set*) const;
-
- private:
-  std::vector<enc_cb*> mCTBs;
-  int mWidthCtbs;
-  int mHeightCtbs;
-  int mLog2CtbSize;
-
-  void free() {
-    for (int i=0 ; i<mWidthCtbs*mHeightCtbs ; i++) {
-      if (mCTBsi) {
-        delete mCTBsi;
-        mCTBsi=NULL;
-      }
-    }
-  }
-};
-
-
-
-
-inline int childX(int x0, int idx, int log2CbSize)
-{
-  return x0 + ((idx&1) << (log2CbSize-1));
-}
-
-inline int childY(int y0, int idx, int log2CbSize)
-{
-  return y0 + ((idx>>1) << (log2CbSize-1));
-}
-
-
-
-#endif

libde265-1.0.17.tar.gz/libde265/encoder/encpicbuf.cc Deleted

@@ -1,313 +0,0 @@
-/*
- * H.265 video codec.
- * Copyright (c) 2013-2014 struktur AG, Dirk Farin <farin@struktur.de>
- *
- * This file is part of libde265.
- *
- * libde265 is free software: you can redistribute it and/or modify
- * it under the terms of the GNU Lesser General Public License as
- * published by the Free Software Foundation, either version 3 of
- * the License, or (at your option) any later version.
- *
- * libde265 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 Lesser General Public License for more details.
- *
- * You should have received a copy of the GNU Lesser General Public License
- * along with libde265.  If not, see <http://www.gnu.org/licenses/>.
- */
-
-#include "libde265/encoder/encpicbuf.h"
-#include "libde265/util.h"
-
-
-encoder_picture_buffer::encoder_picture_buffer()
-{
-}
-
-encoder_picture_buffer::~encoder_picture_buffer()
-{
-  flush_images();
-}
-
-
-image_data::image_data()
-{
-  //printf("new %p\n",this);
-
-  frame_number = 0;
-
-  input = NULL;
-  prediction = NULL;
-  reconstruction = NULL;
-
-  // SOP metadata
-
-  sps_index = -1;
-  skip_priority = 0;
-  is_intra = true;
-
-  state = state_unprocessed;
-
-  is_in_output_queue = true;
-}
-
-image_data::~image_data()
-{
-  //printf("delete %p\n",this);
-
-  delete input;
-  // TODO: this could still be referenced in the packet output queue, so the
-  //       images should really be refcounted. release for now to prevent leaks
-  delete reconstruction;
-  delete prediction;
-}
-
-
-// --- input pushed by the input process ---
-
-void encoder_picture_buffer::reset()
-{
-  flush_images();
-
-  mEndOfStream = false;
-}
-
-
-void encoder_picture_buffer::flush_images()
-{
-  while (!mImages.empty()) {
-    delete mImages.front();
-    mImages.pop_front();
-  }
-}
-
-
-image_data* encoder_picture_buffer::insert_next_image_in_encoding_order(const de265_image* img,
-                                                                        int frame_number)
-{
-  image_data* data = new image_data();
-  data->frame_number = frame_number;
-  data->input = img;
-  data->shdr.set_defaults();
-
-  mImages.push_back(data);
-
-  return data;
-}
-
-void encoder_picture_buffer::insert_end_of_stream()
-{
-  mEndOfStream = true;
-}
-
-
-// --- SOP structure ---
-
-void image_data::set_intra()
-{
-  is_intra = true;
-}
-
-void image_data::set_NAL_type(uint8_t nalType)
-{
-  nal.nal_unit_type = nalType;
-}
-
-void image_data::set_references(int sps_index, // -1 -> custom
-                                const std::vector<int>& l0,
-                                const std::vector<int>& l1,
-                                const std::vector<int>& lt,
-                                const std::vector<int>& keepMoreReferences)
-{
-  this->sps_index = sps_index;
-  ref0 = l0;
-  ref1 = l1;
-  longterm = lt;
-  keep = keepMoreReferences;
-
-
-  // TODO: pps.num_ref_idx_l0_default_active
-
-  shdr.num_ref_idx_l0_active = l0.size();
-  //shdr.num_ref_idx_l1_active = l1.size();
-
-  assert(l0.size() < MAX_NUM_REF_PICS);
-  for (size_t i=0;i<l0.size();i++) {
-    shdr.RefPicList0i = l0i;
-  }
-
-  /*
-  assert(l1.size() < MAX_NUM_REF_PICS);
-  for (int i=0;i<l1.size();i++) {
-    shdr.RefPicList1i = l1i;
-  }
-  */
-}
-
-void image_data::set_NAL_temporal_id(int temporal_id)
-{
-  this->nal.nuh_temporal_id = temporal_id;
-}
-
-void image_data::set_skip_priority(int skip_priority)
-{
-  this->skip_priority = skip_priority;
-}
-
-void encoder_picture_buffer::sop_metadata_commit(int frame_number)
-{
-  image_data* data = mImages.back();
-  assert(data->frame_number == frame_number);
-
-  data->state = image_data::state_sop_metadata_available;
-}
-
-
-
-// --- infos pushed by encoder ---
-
-void encoder_picture_buffer::mark_encoding_started(int frame_number)
-{
-  image_data* data = get_picture(frame_number);
-
-  data->state = image_data::state_encoding;
-}
-
-void encoder_picture_buffer::set_prediction_image(int frame_number, de265_image* pred)
-{
-  image_data* data = get_picture(frame_number);
-
-  data->prediction = pred;
-}
-
-void encoder_picture_buffer::set_reconstruction_image(int frame_number, de265_image* reco)
-{
-  image_data* data = get_picture(frame_number);
-
-  data->reconstruction = reco;
-}
-
-void encoder_picture_buffer::mark_encoding_finished(int frame_number)
-{
-  image_data* data = get_picture(frame_number);
-
-  data->state = image_data::state_keep_for_reference;
-
-
-  // --- delete images that are not required anymore ---
-
-  // first, mark all images unused
-
-  for (auto imgdata : mImages) {
-    imgdata->mark_used = false;
-  }
-
-  // mark all images that will be used later
-
-  for (int f : data->ref0)     { get_picture(f)->mark_used=true; }
-  for (int f : data->ref1)     { get_picture(f)->mark_used=true; }
-  for (int f : data->longterm) { get_picture(f)->mark_used=true; }
-  for (int f : data->keep)     { get_picture(f)->mark_used=true; }
-  data->mark_used=true;
-
-  // copy over all images that we still keep
-
-  std::deque<image_data*> newImageSet;
-  for (auto imgdata : mImages) {
-    if (imgdata->mark_used || imgdata->is_in_output_queue) {
-      imgdata->reconstruction->PicState = UsedForShortTermReference; // TODO: this is only a hack
-
-      newImageSet.push_back(imgdata);
-    }
-    else {
-      // image is not needed anymore for reference, remove it from EncPicBuf
-
-      delete imgdata;
-    }
-  }
-
-  mImages = newImageSet;
-}
-
-
-
-// --- data access ---
-
-bool encoder_picture_buffer::have_more_frames_to_encode() const
-{
-  for (size_t i=0;i<mImages.size();i++) {
-    if (mImagesi->state < image_data::state_encoding) {
-      return true;
-    }
-  }
-
-  return false;
-}
-
-
-image_data* encoder_picture_buffer::get_next_picture_to_encode()
-{
-  for (size_t i=0;i<mImages.size();i++) {
-    if (mImagesi->state < image_data::state_encoding) {
-      return mImagesi;
-    }
-  }
-
-  return NULL;
-}
-
-
-const image_data* encoder_picture_buffer::get_picture(int frame_number) const
-{
-  for (size_t i=0;i<mImages.size();i++) {
-    if (mImagesi->frame_number == frame_number)
-      return mImagesi;
-  }
-
-  assert(false);
-  return NULL;
-}
-
-
-image_data* encoder_picture_buffer::get_picture(int frame_number)
-{
-  for (size_t i=0;i<mImages.size();i++) {
-    if (mImagesi->frame_number == frame_number)
-      return mImagesi;
-  }
-
-  assert(false);
-  return NULL;
-}
-
-
-bool encoder_picture_buffer::has_picture(int frame_number) const
-{
-  for (size_t i=0;i<mImages.size();i++) {
-    if (mImagesi->frame_number == frame_number)
-      return true;
-  }
-
-  return false;
-}
-
-
-void encoder_picture_buffer::mark_image_is_outputted(int frame_number)
-{
-  image_data* idata = get_picture(frame_number);
-  assert(idata);
-
-  idata->is_in_output_queue = false;
-}
-
-
-void encoder_picture_buffer::release_input_image(int frame_number)
-{
-  image_data* idata = get_picture(frame_number);
-  assert(idata);
-
-  delete idata->input;
-  idata->input = NULL;
-}

libde265-1.0.17.tar.gz/libde265/encoder/encpicbuf.h Deleted

@@ -1,144 +0,0 @@
-/*
- * H.265 video codec.
- * Copyright (c) 2013-2014 struktur AG, Dirk Farin <farin@struktur.de>
- *
- * This file is part of libde265.
- *
- * libde265 is free software: you can redistribute it and/or modify
- * it under the terms of the GNU Lesser General Public License as
- * published by the Free Software Foundation, either version 3 of
- * the License, or (at your option) any later version.
- *
- * libde265 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 Lesser General Public License for more details.
- *
- * You should have received a copy of the GNU Lesser General Public License
- * along with libde265.  If not, see <http://www.gnu.org/licenses/>.
- */
-
-#ifndef DE265_ENCPICBUF_H
-#define DE265_ENCPICBUF_H
-
-#include "libde265/image.h"
-#include "libde265/sps.h"
-
-#include <deque>
-#include <vector>
-
-
-/* TODO: we need a way to quickly access pictures with a stable ID, like in the DPB.
- */
-
-struct image_data
-{
-  image_data();
-  ~image_data();
-
-  int frame_number;
-
-  const de265_image* input; // owner
-  de265_image* prediction;  // owner
-  de265_image* reconstruction; // owner
-
-  // SOP metadata
-
-  nal_header nal; // TODO: image split into several NALs (always same NAL header?)
-
-  slice_segment_header shdr; // TODO: multi-slice pictures
-
-  std::vector<int> ref0;
-  std::vector<int> ref1;
-  std::vector<int> longterm;
-  std::vector<int> keep;
-  int sps_index;
-  int skip_priority;
-  bool is_intra;  // TODO: remove, use shdr.slice_type instead
-
-  /* unprocessed              only input image has been inserted, no metadata
-     sop_metadata_available   sop-creator has filled in references and skipping metadata
-     a) encoding              encoding started for this frame, reconstruction image was created
-     .  keep_for_reference    encoding finished, picture is kept in the buffer for reference
-     b) skipped               image was skipped, no encoding was done, no reconstruction image
-  */
-  enum state {
-    state_unprocessed,
-    state_sop_metadata_available,
-    state_encoding,
-    state_keep_for_reference,
-    state_skipped
-  } state;
-
-  bool is_in_output_queue;
-
-  bool mark_used;
-
-
-  // --- SOP structure ---
-
-  void set_intra();
-  void set_NAL_type(uint8_t nalType);
-  void set_NAL_temporal_id(int temporal_id);
-  void set_references(int sps_index, // -1 -> custom
-                      const std::vector<int>& l0, const std::vector<int>& l1,
-                      const std::vector<int>& lt,
-                      const std::vector<int>& keepMoreReferences);
-  void set_skip_priority(int skip_priority);
-};
-
-
-class encoder_picture_buffer
-{
- public:
-  encoder_picture_buffer();
-  ~encoder_picture_buffer();
-
-
-  // --- input pushed by the input process ---
-
-  void reset();
-
-  image_data* insert_next_image_in_encoding_order(const de265_image*, int frame_number);
-  void insert_end_of_stream();
-
-
-  // --- SOP structure ---
-
-  void sop_metadata_commit(int frame_number); // note: frame_number is only for consistency checking
-
-
-  // --- infos pushed by encoder ---
-
-  void mark_encoding_started(int frame_number);
-  void set_prediction_image(int frame_number, de265_image*); // store it just for debugging fun
-  void set_reconstruction_image(int frame_number, de265_image*);
-  void mark_encoding_finished(int frame_number);
-
-
-
-  // --- data access ---
-
-  bool have_more_frames_to_encode() const;
-  image_data* get_next_picture_to_encode(); // or return NULL if no picture is available
-  const image_data* get_picture(int frame_number) const;
-  bool has_picture(int frame_number) const;
-
-  const image_data* peek_next_picture_to_encode() const {
-    assert(!mImages.empty());
-    return mImages.front();
-  }
-
-  void mark_image_is_outputted(int frame_number);
-  void release_input_image(int frame_number);
-
- private:
-  bool mEndOfStream;
-  std::deque<image_data*> mImages;
-
-  void flush_images();
-  image_data* get_picture(int frame_number);
-};
-
-
-#endif

libde265-1.0.17.tar.gz/libde265/encoder/sop.cc Deleted

@@ -1,106 +0,0 @@
-/*
- * H.265 video codec.
- * Copyright (c) 2013-2014 struktur AG, Dirk Farin <farin@struktur.de>
- *
- * This file is part of libde265.
- *
- * libde265 is free software: you can redistribute it and/or modify
- * it under the terms of the GNU Lesser General Public License as
- * published by the Free Software Foundation, either version 3 of
- * the License, or (at your option) any later version.
- *
- * libde265 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 Lesser General Public License for more details.
- *
- * You should have received a copy of the GNU Lesser General Public License
- * along with libde265.  If not, see <http://www.gnu.org/licenses/>.
- */
-
-#include "libde265/encoder/sop.h"
-#include "libde265/encoder/encoder-context.h"
-
-
-sop_creator_intra_only::sop_creator_intra_only()
-{
-}
-
-
-void sop_creator_intra_only::set_SPS_header_values()
-{
-  mEncCtx->get_sps().log2_max_pic_order_cnt_lsb = get_num_poc_lsb_bits();
-}
-
-
-void sop_creator_intra_only::insert_new_input_image(de265_image* img)
-{
-  img->PicOrderCntVal = get_pic_order_count();
-
-  reset_poc();
-  int poc = get_pic_order_count();
-
-  assert(mEncPicBuf);
-  image_data* imgdata = mEncPicBuf->insert_next_image_in_encoding_order(img, get_frame_number());
-
-  imgdata->set_intra();
-  imgdata->set_NAL_type(NAL_UNIT_IDR_N_LP);
-  imgdata->shdr.slice_type = SLICE_TYPE_I;
-  imgdata->shdr.slice_pic_order_cnt_lsb = get_pic_order_count_lsb();
-
-  mEncPicBuf->sop_metadata_commit(get_frame_number());
-
-  advance_frame();
-}
-
-
-// ---------------------------------------------------------------------------
-
-
-sop_creator_trivial_low_delay::sop_creator_trivial_low_delay()
-{
-}
-
-
-void sop_creator_trivial_low_delay::set_SPS_header_values()
-{
-  ref_pic_set rps;
-  rps.DeltaPocS00 = -1;
-  rps.UsedByCurrPicS00 = true;
-  rps.NumNegativePics = 1;
-  rps.NumPositivePics = 0;
-  rps.compute_derived_values();
-  mEncCtx->get_sps().ref_pic_sets.push_back(rps);
-  mEncCtx->get_sps().log2_max_pic_order_cnt_lsb = get_num_poc_lsb_bits();
-}
-
-
-void sop_creator_trivial_low_delay::insert_new_input_image(de265_image* img)
-{
-  img->PicOrderCntVal = get_pic_order_count();
-
-  int frame = get_frame_number();
-
-  std::vector<int> l0, l1, empty;
-  if (!isIntra(frame)) {
-    l0.push_back(frame-1);
-  }
-
-  assert(mEncPicBuf);
-  image_data* imgdata = mEncPicBuf->insert_next_image_in_encoding_order(img, get_frame_number());
-
-  if (isIntra(frame)) {
-    reset_poc();
-    imgdata->set_intra();
-    imgdata->set_NAL_type(NAL_UNIT_IDR_N_LP);
-    imgdata->shdr.slice_type = SLICE_TYPE_I;
-  } else {
-    imgdata->set_references(0, l0,l1, empty,empty);
-    imgdata->set_NAL_type(NAL_UNIT_TRAIL_R);
-    imgdata->shdr.slice_type = SLICE_TYPE_P;
-  }
-  imgdata->shdr.slice_pic_order_cnt_lsb = get_pic_order_count_lsb();
-  mEncPicBuf->sop_metadata_commit(get_frame_number());
-
-  advance_frame();
-}

libde265-1.0.17.tar.gz/libde265/encoder/sop.h Deleted

@@ -1,147 +0,0 @@
-/*
- * H.265 video codec.
- * Copyright (c) 2013-2014 struktur AG, Dirk Farin <farin@struktur.de>
- *
- * This file is part of libde265.
- *
- * libde265 is free software: you can redistribute it and/or modify
- * it under the terms of the GNU Lesser General Public License as
- * published by the Free Software Foundation, either version 3 of
- * the License, or (at your option) any later version.
- *
- * libde265 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 Lesser General Public License for more details.
- *
- * You should have received a copy of the GNU Lesser General Public License
- * along with libde265.  If not, see <http://www.gnu.org/licenses/>.
- */
-
-#ifndef DE265_SOP_H
-#define DE265_SOP_H
-
-#include "libde265/image.h"
-#include "libde265/sps.h"
-#include "libde265/encoder/configparam.h"
-//#include "libde265/encoder/encoder-context.h"
-#include "libde265/encoder/encpicbuf.h"
-
-#include <deque>
-#include <vector>
-
-/*
-struct refpic_set
-{
-  std::vector<int> l0;
-  std::vector<int> l1;
-};
-*/
-
-class encoder_context;
-
-
-class pic_order_counter
-{
- public:
-  pic_order_counter() { mFrameNumber=0; mPOC=0; mNumLsbBits=6; }
-
-  void reset_poc() { mPOC=0; }
-
-  int get_frame_number() const { return mFrameNumber; }
-
-  int get_pic_order_count() const { return mPOC; }
-  int get_pic_order_count_lsb() const {
-    return mPOC & ((1<<mNumLsbBits)-1);
-  }
-
-  void advance_frame(int n=1) { mFrameNumber+=n; mPOC+=n; }
-
-  void set_num_poc_lsb_bits(int n) { mNumLsbBits=n; }
-  int  get_num_poc_lsb_bits() const { return mNumLsbBits; }
-
- private:
-  int mFrameNumber;
-  int mPOC;
-  int mNumLsbBits;
-};
-
-
-class sop_creator : public pic_order_counter
-{
- public:
-  sop_creator() { mEncCtx=NULL; mEncPicBuf=NULL; }
-  virtual ~sop_creator() { }
-
-  void set_encoder_context(encoder_context* encctx) { mEncCtx=encctx; }
-  void set_encoder_picture_buffer(encoder_picture_buffer* encbuf) { mEncPicBuf=encbuf; }
-
-  /* Fills in the following fields:
-     - SPS.ref_pic_sets
-     - SPS.log2_max_pic_order_cnt_lsb
-   */
-  virtual void set_SPS_header_values() = 0;
-
-  /* Fills in the following fields:
-     - NAL.nal_type
-     - SHDR.slice_type
-     - SHDR.slice_pic_order_cnt_lsb
-     - IMGDATA.references
-   */
-  virtual void insert_new_input_image(de265_image*) = 0;
-  virtual void insert_end_of_stream() { mEncPicBuf->insert_end_of_stream(); }
-
-  virtual int  get_number_of_temporal_layers() const { return 1; }
-
-  //virtual std::vector<refpic_set> get_sps_refpic_sets() const = 0;
-
- protected:
-  encoder_context*        mEncCtx;
-  encoder_picture_buffer* mEncPicBuf;
-};
-
-
-
-class sop_creator_intra_only : public sop_creator
-{
- public:
-  sop_creator_intra_only();
-
-  virtual void set_SPS_header_values();
-  virtual void insert_new_input_image(de265_image* img);
-};
-
-
-
-class sop_creator_trivial_low_delay : public sop_creator
-{
- public:
-  struct params {
-    params() {
-      intraPeriod.set_ID("sop-lowDelay-intraPeriod");
-      intraPeriod.set_minimum(1);
-      intraPeriod.set_default(250);
-    }
-
-    void registerParams(config_parameters& config) {
-      config.add_option(&intraPeriod);
-    }
-
-    option_int intraPeriod;
-  };
-
-  sop_creator_trivial_low_delay();
-
-  void setParams(const params& p) { mParams=p; }
-
-  virtual void set_SPS_header_values();
-  virtual void insert_new_input_image(de265_image* img);
-
- private:
-  params mParams;
-
-  bool isIntra(int frame) const { return (frame % mParams.intraPeriod)==0; }
-};
-
-
-#endif

libde265-1.0.17.tar.gz/tools Deleted

libde265-1.0.17.tar.gz/tools/CMakeLists.txt Deleted

libde265-1.0.17.tar.gz/tools/bjoentegaard.cc Deleted

@@ -1,344 +0,0 @@
-/*
- * H.265 video codec.
- * Copyright (c) 2013-2014 struktur AG, Dirk Farin <farin@struktur.de>
- *
- * This file is part of libde265.
- *
- * libde265 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 3 of the License, or
- * (at your option) any later version.
- *
- * libde265 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 libde265.  If not, see <http://www.gnu.org/licenses/>.
- */
-
-#include <stdio.h>
-#include <stdlib.h>
-#include <fstream>
-#include <sstream>
-#include <string>
-#include <vector>
-#include <math.h>
-#include <unistd.h>
-
-
-const bool D = false;
-
-
-/* There are numerical stability problems in the matrix inverse.
-   Switching to long double seems to provide enough accuracy.
-   TODO: in the long term, use a better regression algorithm.
- */
-typedef long double FP;
-
-
-struct datapoint
-{
-  double rate;
-  double distortion;
-};
-
-struct BjoentegaardParams
-{
-  // a*log^3 R + b*log^2 R + c*log R + d
-  double a,b,c,d;
-
-  double minRate, maxRate;
-};
-
-std::vector<datapoint> curveA,curveB;
-BjoentegaardParams paramsA,paramsB;
-
-#define RATE_NORMALIZATION_FACTOR 1 //(1/1000.0)
-
-
-
-FP invf(int i,int j,const FP* m)
-{
-  int o = 2+(j-i);
-
-  i += 4+o;
-  j += 4-o;
-
-#define e(a,b) m ((j+b)%4)*4 + ((i+a)%4) 
-
-    FP inv =
-      + e(+1,-1)*e(+0,+0)*e(-1,+1)
-      + e(+1,+1)*e(+0,-1)*e(-1,+0)
-      + e(-1,-1)*e(+1,+0)*e(+0,+1)
-      - e(-1,-1)*e(+0,+0)*e(+1,+1)
-      - e(-1,+1)*e(+0,-1)*e(+1,+0)
-      - e(+1,-1)*e(-1,+0)*e(+0,+1);
-
-    return (o%2)?inv : -inv;
-
-    #undef e
-
-}
-
-bool inverseMatrix4x4(const FP *m, FP *out)
-{
-  FP inv16;
-
-  for(int i=0;i<4;i++)
-    for(int j=0;j<4;j++)
-      invj*4+i = invf(i,j,m);
-
-  FP D = 0;
-
-  for(int k=0;k<4;k++) D += mk * invk*4;
-
-  if (D == 0) return false;
-
-  D = 1.0 / D;
-
-  for (int i = 0; i < 16; i++)
-    outi = invi * D;
-
-  return true;
-
-}
-
-
-
-BjoentegaardParams fitParams(const std::vector<datapoint>& curve)
-{
-  // build regression matrix
-
-  int n = curve.size();
-
-  FP X4*n;  // regression matrix
-  FP XTn*4; // transpose of X
-
-  for (int i=0;i<n;i++) {
-    FP x = log(curvei.rate) * RATE_NORMALIZATION_FACTOR;
-
-    X4*i + 0 = 1;
-    X4*i + 1 = x;
-    X4*i + 2 = x*x;
-    X4*i + 3 = x*x*x;
-
-    if (D) printf("%f %f %f %f ;\n",1.0,(double)x,(double)(x*x),(double)(x*x*x));
-
-    XTi+0*n = 1;
-    XTi+1*n = x;
-    XTi+2*n = x*x;
-    XTi+3*n = x*x*x;
-  }
-
-  if (D) {
-    printf("rate: ");
-    for (int i=0;i<n;i++) {
-      printf("%f ; ",curvei.rate);
-    }
-    printf("\n");
-
-    printf("distortion: ");
-    for (int i=0;i<n;i++) {
-      printf("%f ; ",curvei.distortion);
-    }
-    printf("\n");
-  }
-
-  // calc X^T * X
-
-  FP XTX4*4;
-  for (int y=0;y<4;y++)
-    for (int x=0;x<4;x++) {
-      FP sum=0;
-
-      for (int i=0;i<n;i++)
-        {
-          sum += XTy*n + i * Xx + i*4;
-        }
-
-      XTXy*4+x = sum;
-    }
-
-  FP XTXinv4*4;
-
-  inverseMatrix4x4(XTX, XTXinv);
-
-  if (D) {
-    for (int y=0;y<4;y++) {
-      for (int x=0;x<4;x++) {
-        printf("%f ",(double)XTXinvy*4+x);
-      }
-      printf("\n");
-    }
-  }
-
-  // calculate pseudo-inverse XP = (X^T * X)^-1 * X^T
-
-  FP XPn*4;
-
-  for (int y=0;y<4;y++) {
-    for (int x=0;x<n;x++) {
-      FP sum=0;
-
-      for (int i=0;i<4;i++)
-        {
-          sum += XTXinvy*4 + i * XTx + i*n;
-        }
-
-      XPy*n+x = sum;
-    }
-  }
-
-  // calculate regression parameters
-
-  FP p4;
-
-  for (int k=0;k<4;k++)
-    {
-      FP sum=0;
-
-      for (int i=0;i<n;i++) {
-        sum += XPk*n + i * curvei.distortion;
-      }
-
-      pk=sum;
-    }
-
-
-  BjoentegaardParams param;
-  param.d = p0;
-  param.c = p1;
-  param.b = p2;
-  param.a = p3;
-
-
-  // find min and max rate
-
-  param.minRate = curve0.rate;
-  param.maxRate = curve0.rate;
-
-  for (int i=1;i<n;i++) {
-    param.minRate = std::min(param.minRate, curvei.rate);
-    param.maxRate = std::max(param.maxRate, curvei.rate);
-  }
-
-  return param;
-}
-
-FP evalIntegralAt(const BjoentegaardParams& p, double x)
-{
-  FP sum = 0;
-
-  // integral of: d
-
-  sum += p.d * x;
-
-  // integral of: c*log(x)
-
-  sum += p.c * x* (log(x)-1);
-
-  // integral of: b*log(x)^2
-
-  sum += p.b * x * ((log(x)-2)*log(x)+2);
-
-  // integral of: a*log(x)^3
-
-  sum += p.a * x * (log(x)*((log(x)-3)*log(x)+6)-6);
-
-  return sum;
-}
-
-
-double calcBjoentegaard(const BjoentegaardParams& paramsA,
-                        const BjoentegaardParams& paramsB,
-                        double min_rate, double max_rate)
-{
-  double mini = std::max(paramsA.minRate, paramsB.minRate);
-  double maxi = std::min(paramsA.maxRate, paramsB.maxRate);
-
-  if (min_rate >= 0) mini = std::max(mini, min_rate);
-  if (max_rate >= 0) maxi = std::min(maxi, max_rate);
-
-  if (D) printf("range: %f %f\n",mini,maxi);
-
-  FP intA = evalIntegralAt(paramsA, maxi) - evalIntegralAt(paramsA, mini);
-  FP intB = evalIntegralAt(paramsB, maxi) - evalIntegralAt(paramsB, mini);
-
-  if (D) printf("int1:%f int2:%f\n",(double)intA,(double)intB);
-
-  return (intA-intB)/(maxi-mini);
-}
-
-
-std::vector<datapoint> readRDFile(const char* filename, float min_rate, float max_rate)
-{
-  std::vector<datapoint> curve;
-  std::ifstream istr(filename);
-
-  for (;;)
-    {
-      std::string line;
-      getline(istr, line);
-      if (istr.eof())
-        break;
-
-      if (line0=='#') continue;
-
-      std::stringstream sstr(line);
-      datapoint p;
-      sstr >> p.rate >> p.distortion;
-
-      if (min_rate>=0 && p.rate < min_rate) continue;
-      if (max_rate>=0 && p.rate > max_rate) continue;
-
-      curve.push_back(p);
-    }
-
-  return curve;
-}
-
-
-int main(int argc, char** argv)
-{
-  float min_rate = -1;
-  float max_rate = -1;
-
-  int c;
-  while ((c=getopt(argc,argv, "l:h:")) != -1) {
-    switch (c) {
-    case 'l': min_rate = atof(optarg); break;
-    case 'h': max_rate = atof(optarg); break;
-    }
-  }
-
-  curveA = readRDFile(argvoptind, min_rate, max_rate);
-  paramsA = fitParams(curveA);
-
-  printf("params A: %f %f %f %f\n",paramsA.a,paramsA.b,paramsA.c,paramsA.d);
-
-  printf("gnuplot: %f*log(x)**3+%f*log(x)**2+%f*log(x)+%f\n",paramsA.a,paramsA.b,paramsA.c,paramsA.d);
-
-  if (optind+1<argc) {
-    curveB = readRDFile(argvoptind+1, min_rate, max_rate);
-    paramsB = fitParams(curveB);
-
-    printf("params B: %f %f %f %f\n",paramsB.a,paramsB.b,paramsB.c,paramsB.d);
-
-    printf("gnuplot: %f*log(x)**3+%f*log(x)**2+%f*log(x)+%f\n",paramsB.a,paramsB.b,paramsB.c,paramsB.d);
-
-    double delta = calcBjoentegaard(paramsA,paramsB, min_rate,max_rate);
-
-    printf("Bjoentegaard delta: %f dB   (A-B -> >0 -> first (A) is better)\n",delta);
-
-    if (delta>=0) {
-      printf("-> first is better by %f dB\n",delta);
-    }
-    else {
-      printf("-> second is better by %f dB\n",-delta);
-    }
-  }
-
-  return 0;
-}

libde265-1.0.17.tar.gz/tools/block-rate-estim.cc Deleted

@@ -1,127 +0,0 @@
-/*
- * H.265 video codec.
- * Copyright (c) 2013-2014 struktur AG, Dirk Farin <farin@struktur.de>
- *
- * This file is part of libde265.
- *
- * libde265 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 3 of the License, or
- * (at your option) any later version.
- *
- * libde265 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 libde265.  If not, see <http://www.gnu.org/licenses/>.
- */
-
-#include <vector>
-#include <string>
-#include <fstream>
-#include <math.h>
-
-
-struct datapoint {
-  int   log2blksize;
-  float rate;
-  float estim;
-};
-
-std::vector<datapoint> pts;
-
-#define NBINS 100
-
-/*
- #define ESTIMDIV 100
- #define MAXESTIM 80000
- std::vector<float> bitestim2MAXESTIM/ESTIMDIV;
-*/
-
-
-
-void print_bitestim_results(int log2blksize)
-{
-  float max_estim=0;
-
-  for (int i=0;i<pts.size();i++) {
-    if (log2blksize==0 || ptsi.log2blksize==log2blksize) {
-      max_estim = std::max(max_estim, ptsi.estim);
-    }
-  }
-
-
-
-  float epsilon = 0.0001;
-  float interval = (max_estim+epsilon) / NBINS;
-
-  for (int b=0;b<NBINS;b++) {
-
-    int cnt=0;
-    double sum=0;
-    float mini=999999;
-    float maxi=0;
-
-    for (int i=0;i<pts.size();i++)
-      if (log2blksize==0 || ptsi.log2blksize==log2blksize) {
-        int bin = ptsi.estim/interval;
-        if (bin==b) {
-          sum += ptsi.rate;
-
-          mini = std::min(mini,ptsi.rate);
-          maxi = std::max(maxi,ptsi.rate);
-          cnt++;
-        }
-      }
-
-    if (cnt>0) {
-      double mean = sum/cnt;
-
-      double var = 0;
-
-      for (int i=0;i<pts.size();i++)
-        if (log2blksize==0 || ptsi.log2blksize==log2blksize) {
-          int bin = ptsi.estim/interval;
-          if (bin==b) {
-            var += (ptsi.rate-mean)*(ptsi.rate-mean);
-          }
-        }
-
-      var /= cnt;
-      double stddev = sqrt(var);
-
-      printf("%f  %f %f  %f %f  %f %f %d\n",
-             (b+0.5)*interval,mean,var,
-             mean-stddev,mean+stddev, mini,maxi,
-             cnt);
-    }
-  }
-}
-
-
-int main(int argc,char** argv)
-{
-  std::string tag = argv1;
-
-  std::ifstream istr(argv2);
-  for (;;)
-    {
-      std::string t;
-      int log2blksize;
-      datapoint pt;
-
-      istr >> t >> pt.log2blksize >> pt.rate >> pt.estim;
-
-      if (istr.eof()) break;
-
-      if (t == tag) {
-        pts.push_back(pt);
-      }
-    }
-
-  print_bitestim_results(0);
-
-  return 0;
-}

libde265-1.0.17.tar.gz/tools/gen-entropy-table.cc Deleted

@@ -1,477 +0,0 @@
-/*
- * H.265 video codec.
- * Copyright (c) 2013-2014 struktur AG, Dirk Farin <farin@struktur.de>
- *
- * This file is part of libde265.
- *
- * libde265 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 3 of the License, or
- * (at your option) any later version.
- *
- * libde265 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 libde265.  If not, see <http://www.gnu.org/licenses/>.
- */
-
-
-#include "libde265/cabac.h"
-#include <assert.h>
-#include <time.h>
-#include <stdlib.h>
-#include <stdio.h>
-
-
-void simple_getline(char** lineptr,size_t* linelen,FILE* fh)
-{
-  const int LINESIZE=1000;
-
-  if (*lineptr==NULL) {
-    *linelen = LINESIZE;
-    *lineptr = (char*)malloc(LINESIZE);
-  }
-
-  char* p = *lineptr;
-
-  for (;;) {
-    assert(p - *lineptr < LINESIZE);
-
-    int c = fgetc(fh);
-    if (c == EOF || c == '\n') {
-      *p = 0;
-      break;
-    }
-
-    *p++ = c;
-  }
-}
-
-
-void generate_entropy_table()
-{
-#if 000
-  const int nSymbols=1000*1000*10;
-  const int oversample = 10;
-
-  double tab642;
-
-  for (int i=0;i<64;i++)
-    for (int k=0;k<2;k++)
-      tabik=0;
-
-  srand(time(0));
-  //srand(123);
-
-  int cnt=1;
-  for (;;cnt++) {
-    printf("-------------------- %d --------------------\n",cnt);
-
-    for (int s=0;s<63;s++) {
-      CABAC_encoder_bitstream cabac_mix0;
-      CABAC_encoder_bitstream cabac_mix1;
-      CABAC_encoder_bitstream cabac_ref;
-
-      for (int i=0;i<nSymbols*oversample;i++) {
-        int r = rand();
-        int n = (r>>2) % 63;
-        int m = (r>>1) & 1;
-        int b = r & 1;
-
-        context_model model;
-        model.MPSbit = 1;
-        model.state  = n;
-        cabac_ref.write_CABAC_bit(&model, b);
-
-        model.MPSbit = 1;
-        model.state  = n;
-        cabac_mix0.write_CABAC_bit(&model, b);
-
-        model.MPSbit = 1;
-        model.state  = n;
-        cabac_mix1.write_CABAC_bit(&model, b);
-
-        if (i%oversample == oversample/2) {
-          model.MPSbit = 1;
-          model.state  = s;
-          cabac_mix0.write_CABAC_bit(&model, 0);
-
-          model.MPSbit = 1;
-          model.state  = s;
-          cabac_mix1.write_CABAC_bit(&model, 1);
-
-          //b = rand() & 1;
-          //cabac_mix.write_CABAC_bypass(1);
-        }
-
-      }
-
-      cabac_ref.flush_CABAC();
-      cabac_mix0.flush_CABAC();
-      cabac_mix1.flush_CABAC();
-
-      int bits_ref  = cabac_ref.size()*8;
-      int bits_mix0 = cabac_mix0.size()*8;
-      int bits_mix1 = cabac_mix1.size()*8;
-
-      //printf("bits: %d %d\n",bits_ref,bits_mix);
-      int bits_diff0 = bits_mix0-bits_ref;
-      int bits_diff1 = bits_mix1-bits_ref;
-      //printf("bits diff: %d\n",bits_diff);
-
-      double bits_per_symbol0 = bits_diff0 / double(nSymbols);
-      double bits_per_symbol1 = bits_diff1 / double(nSymbols);
-
-      tabs0 += bits_per_symbol0;
-      tabs1 += bits_per_symbol1;
-
-      double bps0 = tabs0/cnt;
-      double bps1 = tabs1/cnt;
-
-      printf("/* state=%2d */  0x%05x /* %f */,  0x%05x /* %f */,\n", s,
-             (int)(bps1*0x8000), bps1,
-             (int)(bps0*0x8000), bps0);
-    }
-
-    printf("                0x0010c, 0x3bfbb /* dummy, should never be used */\n");
-  }
-#endif
-}
-
-int probTab128+2 = {
-  1537234,1602970,
-  1608644,1815493,
-  1822246,2245961,
-  916773,1329391,
-  1337504,1930659,
-  1063692,1707588,
-  868294,1532108,
-  842934,1555538,
-  689043,1396941,
-  860184,1789964,
-  534165,1258482,
-  672508,1598821,
-  578782,1476240,
-  602247,1613140,
-  409393,1206638,
-  459294,1356779,
-  430124,1359893,
-  308326,1050647,
-  313100,1099956,
-  293887,1088978,
-  220901,869582,
-  214967,881695,
-  197226,856990,
-  166131,767761,
-  152514,737406,
-  128332,663998,
-  117638,632653,
-  106178,595539,
-  90898,539506,
-  83437,509231,
-  76511,492801,
-  64915,443096,
-  57847,409809,
-  52730,385395,
-  45707,354059,
-  42018,333028,
-  37086,308073,
-  33256,284497,
-  36130,299172,
-  28831,270716,
-  25365,244840,
-  22850,221896,
-  19732,201462,
-  17268,183729,
-  15252,168106,
-  13787,153979,
-  12187,141455,
-  10821,130337,
-  9896,120165,
-  8626,112273,
-  8162,103886,
-  7201,96441,
-  6413,89805,
-  5886,83733,
-  5447,78084,
-  4568,73356,
-  4388,68831,
-  3959,64688,
-  3750,60804,
-  3407,57271,
-  3109,54024,
-  2820,51099,
-  48569,1451987,
-  0,    0,
-  0*22686225,    0
-};
-
-
-void generate_entropy_table_probTableWeighted()
-{
-#if 000
-  int64_t probTabSum=0;
-  for (int i=0;i<130;i++)
-    probTabSum += probTabi;
-
-
-  const int nSymbols=1000*1000*10;
-  const int oversample = 10;
-
-  double tab642;
-
-  for (int i=0;i<64;i++)
-    for (int k=0;k<2;k++)
-      tabik=0;
-
-  srand(time(0));
-  //srand(123);
-
-  int cnt=1;
-  for (;;cnt++) {
-    printf("-------------------- %d --------------------\n",cnt);
-
-    for (int s=0;s<63;s++) {
-      CABAC_encoder_bitstream cabac_mix0;
-      CABAC_encoder_bitstream cabac_mix1;
-      CABAC_encoder_bitstream cabac_ref;
-
-      for (int i=0;i<nSymbols*oversample;i++) {
-        int r = rand();
-
-        r %= probTabSum;
-        int idx=0;
-        while (r>probTabidx) {
-          r-=probTabidx;
-          idx++;
-        }
-
-        assert(idx<=128);
-
-        int n = idx/2;
-        int b = idx&1;
-        bool bypass = (idx==128);
-
-        printf("%d %d %d\n",n,b,bypass);
-
-        context_model model;
-        model.MPSbit = 1;
-        model.state  = n;
-        if (bypass) cabac_ref.write_CABAC_bypass(1);
-        else        cabac_ref.write_CABAC_bit(&model, b);
-
-        model.MPSbit = 1;
-        model.state  = n;
-        if (bypass) cabac_mix0.write_CABAC_bypass(1);
-        else        cabac_mix0.write_CABAC_bit(&model, b);
-
-        model.MPSbit = 1;
-        model.state  = n;
-        if (bypass) cabac_mix1.write_CABAC_bypass(1);
-        else        cabac_mix1.write_CABAC_bit(&model, b);
-
-        if (i%oversample == oversample/2) {
-          model.MPSbit = 1;
-          model.state  = s;
-          cabac_mix0.write_CABAC_bit(&model, 0);
-
-          model.MPSbit = 1;
-          model.state  = s;
-          cabac_mix1.write_CABAC_bit(&model, 1);
-
-          //b = rand() & 1;
-          //cabac_mix.write_CABAC_bypass(1);
-        }
-
-      }
-
-      cabac_ref.flush_CABAC();
-      cabac_mix0.flush_CABAC();
-      cabac_mix1.flush_CABAC();
-
-      int bits_ref  = cabac_ref.size()*8;
-      int bits_mix0 = cabac_mix0.size()*8;
-      int bits_mix1 = cabac_mix1.size()*8;
-
-      //printf("bits: %d %d\n",bits_ref,bits_mix);
-      int bits_diff0 = bits_mix0-bits_ref;
-      int bits_diff1 = bits_mix1-bits_ref;
-      //printf("bits diff: %d\n",bits_diff);
-
-      double bits_per_symbol0 = bits_diff0 / double(nSymbols);
-      double bits_per_symbol1 = bits_diff1 / double(nSymbols);
-
-      tabs0 += bits_per_symbol0;
-      tabs1 += bits_per_symbol1;
-
-      double bps0 = tabs0/cnt;
-      double bps1 = tabs1/cnt;
-
-      printf("/* state=%2d */  0x%05x /* %f */,  0x%05x /* %f */,\n", s,
-             (int)(bps1*0x8000), bps1,
-             (int)(bps0*0x8000), bps0);
-    }
-
-    printf("                0x0010c, 0x3bfbb /* dummy, should never be used */\n");
-  }
-#endif
-}
-
-
-void generate_entropy_table_replay()
-{
-#if 000
-  const int oversample = 10;
-
-  char* lineptr = NULL;
-  size_t linelen = 0;
-
-  for (int s=0;s<63;s++) {
-    CABAC_encoder_bitstream cabac_mix0;
-    CABAC_encoder_bitstream cabac_mix1;
-    CABAC_encoder_bitstream cabac_ref;
-
-    int nSymbols = 0;
-
-    FILE* fh = fopen("streamdump-paris-intra","r");
-
-    for (int i=0;i<80000000;i++) {
-      simple_getline(&lineptr,&linelen,fh);
-      if (feof(fh))
-        break;
-
-      int n,b;
-      sscanf(lineptr,"%d %d",&n,&b);
-
-      bool bypass = (n==64);
-
-      if ((i%10000)==0)
-        { printf("%d  %d %d    \r",i,n,b);
-        }
-
-      //printf("%d %d %d\n",n,b,bypass);
-
-      context_model model;
-      model.MPSbit = 1;
-      model.state  = n;
-      if (bypass) cabac_ref.write_CABAC_bypass(1);
-      else        cabac_ref.write_CABAC_bit(&model, b);
-
-      model.MPSbit = 1;
-      model.state  = n;
-      if (bypass) cabac_mix0.write_CABAC_bypass(1);
-      else        cabac_mix0.write_CABAC_bit(&model, b);
-
-      model.MPSbit = 1;
-      model.state  = n;
-      if (bypass) cabac_mix1.write_CABAC_bypass(1);
-      else        cabac_mix1.write_CABAC_bit(&model, b);
-
-      if (i%oversample == oversample/2) {
-        model.MPSbit = 1;
-        model.state  = s;
-        cabac_mix0.write_CABAC_bit(&model, 0);
-
-        model.MPSbit = 1;
-        model.state  = s;
-        cabac_mix1.write_CABAC_bit(&model, 1);
-
-        nSymbols++;
-
-        //b = rand() & 1;
-        //cabac_mix.write_CABAC_bypass(1);
-      }
-    }
-
-    fclose(fh);
-
-    cabac_ref.flush_CABAC();
-    cabac_mix0.flush_CABAC();
-    cabac_mix1.flush_CABAC();
-
-    int bits_ref  = cabac_ref.size()*8;
-    int bits_mix0 = cabac_mix0.size()*8;
-    int bits_mix1 = cabac_mix1.size()*8;
-
-    //printf("bits: %d %d\n",bits_ref,bits_mix);
-    int bits_diff0 = bits_mix0-bits_ref;
-    int bits_diff1 = bits_mix1-bits_ref;
-    //printf("bits diff: %d\n",bits_diff);
-
-    double bits_per_symbol0 = bits_diff0 / double(nSymbols);
-    double bits_per_symbol1 = bits_diff1 / double(nSymbols);
-
-    double bps0 = bits_per_symbol0;
-    double bps1 = bits_per_symbol1;
-
-    printf("/* state=%2d */  0x%05x /* %f */,  0x%05x /* %f */,\n", s,
-           (int)(bps1*0x8000), bps1,
-           (int)(bps0*0x8000), bps0);
-  }
-
-  printf("                0x0010c, 0x3bfbb /* dummy, should never be used */\n");
-#endif
-}
-
-
-void test_entropy_table_replay()
-{
-#if 000
-  char* lineptr = NULL;
-  size_t linelen = 0;
-
-
-  CABAC_encoder_bitstream cabac_bs;
-  CABAC_encoder_estim     cabac_estim;
-
-  //FILE* fh = fopen("y","r");
-  //FILE* fh = fopen("own-dump","r");
-  //FILE* fh = fopen("rawstream-dump","r");
-  //FILE* fh = fopen("johnny-stream-dump","r");
-  FILE* fh = fopen("streamdump-paris-intra","r");
-
-  for (int i=0;i<80000000;i++) {
-    simple_getline(&lineptr,&linelen,fh);
-    if (feof(fh))
-      break;
-
-    int n,b;
-    sscanf(lineptr,"%d %d",&n,&b);
-    b=!b;
-    bool bypass = (n==64);
-
-    if ((i%10000)==0)
-      { printf("%d  %d %d    \n",i,n,b);
-      }
-
-    context_model model;
-    model.MPSbit = 1;
-    model.state  = n;
-    if (bypass) cabac_bs.write_CABAC_bypass(1);
-    else        cabac_bs.write_CABAC_bit(&model, b);
-
-    model.MPSbit = 1;
-    model.state  = n;
-    if (bypass) cabac_estim.write_CABAC_bypass(1);
-    else        cabac_estim.write_CABAC_bit(&model, b);
-  }
-
-  fclose(fh);
-
-  printf("bs:%d estim:%d\n",cabac_bs.size(),cabac_estim.size());
-#endif
-}
-
-
-int main(int argc, char** argv)
-{
-  //generate_entropy_table();
-  //generate_entropy_table_replay();
-
-  test_entropy_table_replay();
-
-  return 0;
-}

libde265-1.0.17.tar.gz/tools/rd-curves.cc Deleted

@@ -1,1075 +0,0 @@
-/*
- * H.265 video codec.
- * Copyright (c) 2013-2014 struktur AG, Dirk Farin <farin@struktur.de>
- *
- * This file is part of libde265.
- *
- * libde265 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 3 of the License, or
- * (at your option) any later version.
- *
- * libde265 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 libde265.  If not, see <http://www.gnu.org/licenses/>.
- */
-
-#ifdef HAVE_CONFIG_H
-#include "config.h"
-#endif
-
-#include <stdio.h>
-#include <stdlib.h>
-#include <sys/types.h>
-#include <sys/stat.h>
-#include <unistd.h>
-#include <assert.h>
-#include <cstdint>
-#include <string.h>
-#include <getopt.h>
-
-#include <sys/time.h>
-
-#if !defined(_WIN32) && !defined(WIN32)
-#include <sys/times.h>
-#endif
-
-#include <string>
-#include <sstream>
-#include <iostream>
-#include <iomanip>
-#include <fstream>
-#include <vector>
-
-
-
-static struct {
-  const char* name;
-  const char* value;
-} variables = {
-  { "$HOME"    , "/home/domain/farindk" },
-  { "$ROOT"    , "/home/domain/farindk/prog/h265" },
-  { "$ENC265"  , "$ROOT/libde265/enc265/enc265" },
-  { "$DEC265"  , "$ROOT/libde265/dec265/dec265" },
-  { "$YUVDIST" , "$ROOT/libde265/tools/yuv-distortion" },
-  { "$YUVTMP"  , "/mnt/temp/dirk/yuv/ftp.tnt.uni-hannover.de/testsequences" },
-  { "$YUV"     , "/storage/users/farindk/yuv" },
-  { "$HMENC"   , "HM13enc" },
-  { "$HM13CFG" , "$ROOT/HM/HM-13.0-dev/cfg" },
-  { "$HMSCCENC", "HM-SCC-enc" },
-  { "$HMSCCCFG", "$ROOT/HM/HM-SCC-extensions/cfg" },
-  { "$X265ENC" , "$ROOT/x265/build/linux/x265" },
-  { "$X264"    , "x264" },
-  { "$FFMPEG"  , "ffmpeg" },
-  { "$F265"    , "$ROOT/f265/build/f265cli" },
-  { 0,0 }
-};
-
-
-bool keepStreams = false;
-int  maxFrames = 0;
-std::string encoderParameters;
-
-
-std::string replace_variables(std::string str)
-{
-  bool replaced = false;
-  for (int i=0;variablesi.name;i++) {
-    size_t pos = str.find(variablesi.name);
-    if (pos != std::string::npos) {
-      replaced = true;
-      str = str.replace(pos, strlen(variablesi.name), variablesi.value);
-      break;
-    }
-  }
-
-  if (!replaced) return str;
-  else return replace_variables(str);
-}
-
-
-// ---------------------------------------------------------------------------
-
-struct Preset
-{
-  const int ID;
-  const char* name;
-  const char* descr;
-
-  const char* options_de265;
-  const char* options_hm;
-  const char* options_hm_scc;
-  const char* options_x265;
-  const char* options_f265;
-  const char* options_x264;
-  const char* options_x264_ffmpeg;
-  const char* options_ffmpeg_mpeg2;
-
-  //int nFrames;
-};
-
-
-Preset preset = {
-  { 1, "pre01-intra-noLF", "intra, no LF, no SBH, CTB-size 32, min CB=8",
-    /* de265  */ "--sop-structure intra",
-    /* HM     */ "-c $HM13CFG/encoder_intra_main.cfg -SBH 0 --SAO=0 --LoopFilterDisable --DeblockingFilterControlPresent --MaxCUSize=32 --MaxPartitionDepth=2",
-    /* HM SCC */ "-c $HMSCCCFG/encoder_intra_main_scc.cfg -SBH 0 --SAO=0 --LoopFilterDisable --DeblockingFilterControlPresent --MaxCUSize=32 --MaxPartitionDepth=2",
-    /* x265   */ "--no-lft -I 1 --no-signhide",
-    /* f265   */ "key-frame-spacing=1",
-    /* x264   */ "-I 1",
-    /* ffmpeg */ "-g 1",
-    /* mpeg-2 */ "-g 1"
-    // 0 // all frames
-  },
-
-  { 2, "pre02-fastIntra", "intra, no LF, no SBH, CTB-size 32, min CB=8",
-    /* de265  */ "--sop-structure intra --TB-IntraPredMode minSSD",
-    /* HM     */ "-c $HM13CFG/encoder_intra_main.cfg -SBH 0 --SAO=0 --LoopFilterDisable --DeblockingFilterControlPresent --MaxCUSize=32 --MaxPartitionDepth=2",
-    /* HM SCC */ "-c $HMSCCCFG/encoder_intra_main_scc.cfg -SBH 0 --SAO=0 --LoopFilterDisable --DeblockingFilterControlPresent --MaxCUSize=32 --MaxPartitionDepth=2",
-    /* x265   */ "--no-lft -I 1 --no-signhide",
-    /* f265   */ "key-frame-spacing=1",
-    /* x264   */ "-I 1",
-    /* ffmpeg */ "-g 1",
-    /* mpeg-2 */ "-g 1"
-    // 0 // all frames
-  },
-
-  { 3, "pre03-fastIntra", "pre02, but fast-brute",
-    /* de265  */ "--sop-structure intra --TB-IntraPredMode fast-brute",
-    /* HM     */ "-c $HM13CFG/encoder_intra_main.cfg -SBH 0 --SAO=0 --LoopFilterDisable --DeblockingFilterControlPresent --MaxCUSize=32 --MaxPartitionDepth=2",
-    /* HM SCC */ "-c $HMSCCCFG/encoder_intra_main_scc.cfg -SBH 0 --SAO=0 --LoopFilterDisable --DeblockingFilterControlPresent --MaxCUSize=32 --MaxPartitionDepth=2",
-    /* x265   */ "--no-lft -I 1 --no-signhide",
-    /* f265   */ "key-frame-spacing=1",
-    /* x264   */ "-I 1",
-    /* ffmpeg */ "-g 1",
-    /* mpeg-2 */ "-g 1"
-    // 0 // all frames
-  },
-
-  { 50, "cb-auto16", "(development test)",
-    /* de265  */ "--max-cb-size 16 --min-cb-size 8",
-    /* HM     */ "-c $HM13CFG/encoder_intra_main.cfg -SBH 0 --SAO=0 --LoopFilterDisable --DeblockingFilterControlPresent --MaxCUSize=32 --MaxPartitionDepth=2",
-    /* HM SCC */ "-c $HMSCCCFG/encoder_intra_main_scc.cfg -SBH 0 --SAO=0 --LoopFilterDisable --DeblockingFilterControlPresent --MaxCUSize=32 --MaxPartitionDepth=2",
-    /* x265   */ "--no-lft -I 1 --no-signhide",
-    /* f265   */ "key-frame-spacing=1",
-    /* x264   */ "-I 1",
-    /* ffmpeg */ "-g 1",
-    /* mpeg-2 */ "-g 1"
-    // 0 // all frames
-  },
-
-  { 80, "lowdelay", "default (low-default) encoder parameters",
-    "--MEMode search --max-cb-size 32 --min-cb-size 8 --min-tb-size 4 --CB-IntraPartMode-Fixed-partMode 2Nx2N --CB-IntraPartMode fixed --TB-IntraPredMode min-residual --PB-MV-TestMode zero",
-    /* de265  */ //"--sop-structure low-delay --MEMode search --max-cb-size 32 --min-cb-size 8 --min-tb-size 4 --CB-IntraPartMode fixed --TB-IntraPredMode min-residual",
-    /* HM     */ "-c $HM13CFG/encoder_lowdelay_main.cfg -ip 248",
-    /* HM SCC */ "-c $HMSCCCFG/encoder_lowdelay_main_scc.cfg -ip 248",
-    /* x265   */ "-I 248 --no-wpp --bframes 0", // GOP size: 248
-    /* f265   */ 0, //"key-frame-spacing=248",
-    /* x264   */ "",
-    /* ffmpeg */ "-g 248 -bf 0",
-    /* mpeg-2 */ "" // GOP size 248 does not make sense here
-    // 0 // all frames
-  },
-
-  { 98, "best", "default (random-access) encoder parameters",
-    /* de265  */ "--max-cb-size 16 --min-cb-size 8",
-    /* HM     */ "-c $HM13CFG/encoder_randomaccess_main.cfg",
-    /* HM SCC */ "-c $HMSCCCFG/encoder_randomaccess_main_scc.cfg",
-    /* x265   */ "",
-    /* f265   */ "",
-    /* x264   */ "",
-    /* ffmpeg */ "",
-    /* mpeg-2 */ ""
-    // 0 // all frames
-  },
-
-  { 99, "besteq", "default (random-access) encoder parameters, I-frame distance = 248",
-    /* de265  */ "",
-    /* HM     */ "-c $HM13CFG/encoder_randomaccess_main.cfg -ip 248",
-    /* HM SCC */ "-c $HMSCCCFG/encoder_randomaccess_main_scc.cfg -ip 248",
-    /* x265   */ "-I 248 --no-wpp", // GOP size: 248
-    /* f265   */ "key-frame-spacing=248",
-    /* x264   */ "",
-    /* ffmpeg */ "-g 248",
-    /* mpeg-2 */ "" // GOP size 248 does not make sense here
-    // 0 // all frames
-  },
-
-  { 0, NULL }
-};
-
-// ---------------------------------------------------------------------------
-
-class Input
-{
-public:
-  Input() {
-    width=height=0;
-    maxFrames=0;
-  }
-
-  void setInput(const char* yuvfilename,int w,int h, float fps) {
-    mInputFilename = yuvfilename;
-    width = w;
-    height = h;
-    mFPS = fps;
-  }
-
-  void setMaxFrames(int n) { maxFrames=n; }
-
-  std::string options_de265() const {
-    std::stringstream sstr;
-    sstr << " -i " << mInputFilename << " --width " << width << " --height " << height;
-    if (maxFrames) sstr << " --frames " << maxFrames;
-
-    return sstr.str();
-  }
-
-  std::string options_HM() const {
-    std::stringstream sstr;
-    sstr << "-i " << mInputFilename << " -wdt " << width << " -hgt " << height
-         << " -fr " << mFPS;
-    if (maxFrames) sstr << " -f " << maxFrames;
-
-    return sstr.str();
-  }
-
-  std::string options_x265() const {
-    std::stringstream sstr;
-    sstr << mInputFilename << " --input-res " << width << "x" << height
-         << " --fps " << mFPS;
-    if (maxFrames) sstr << " -f " << maxFrames;
-
-    return sstr.str();
-  }
-
-  std::string options_x264() const {
-    std::stringstream sstr;
-    sstr << mInputFilename << " --input-res " << width << "x" << height;
-    sstr << " --fps 25"; // TODO: check why crf/qp rate-control freaks out when fps is != 25
-    if (maxFrames) sstr << " --frames " << maxFrames;
-
-    return sstr.str();
-  }
-
-  std::string options_ffmpeg() const {
-    std::stringstream sstr;
-    sstr << "-f rawvideo -vcodec rawvideo -s " << width << "x" << height; // << " -r " << mFPS
-    sstr << " -pix_fmt yuv420p -i " << mInputFilename;
-    if (maxFrames) sstr << " -vframes " << maxFrames;
-
-    return sstr.str();
-  }
-
-  std::string options_f265() const {
-    std::stringstream sstr;
-    sstr << mInputFilename << " -w " << width << ":" << height;
-    if (maxFrames) sstr << " -c " << maxFrames;
-
-    return sstr.str();
-  }
-
-  std::string getFilename() const { return mInputFilename; }
-  float getFPS() const { return mFPS; }
-  int   getNFrames() const { return maxFrames; }
-  int   getWidth() const { return width; }
-  int   getHeight() const { return height; }
-
-private:
-  std::string mInputFilename;
-  int width, height;
-  int maxFrames;
-  float mFPS;
-};
-
-Input input;
-
-struct InputSpec
-{
-  const char* name;
-  const char* filename;
-  int width,height, nFrames;
-  float fps;
-} inputSpec = {
-  { "paris",       "$YUV/paris_cif.yuv",352,288,1065, 30.0 },
-  { "paris10",     "$YUV/paris_cif.yuv",352,288,  10, 30.0 },
-  { "paris100",    "$YUV/paris_cif.yuv",352,288, 100, 30.0 },
-  { "johnny",      "$YUV/Johnny_1280x720_60.yuv",1280,720,600,60.0 },
-  { "johnny10",    "$YUV/Johnny_1280x720_60.yuv",1280,720, 10,60.0 },
-  { "johnny100",   "$YUV/Johnny_1280x720_60.yuv",1280,720,100,60.0 },
-  { "cactus",      "$YUV/Cactus_1920x1080_50.yuv",1920,1080,500,50.0 },
-  { "cactus10",    "$YUV/Cactus_1920x1080_50.yuv",1920,1080, 10,50.0 },
-  { "4people",     "$YUVTMP/FourPeople_1280x720_60.yuv",1280,720,600,60.0 },
-  { "4people100",  "$YUVTMP/FourPeople_1280x720_60.yuv",1280,720,100,60.0 },
-  { "slideedit",   "$YUVTMP/SlideEditing_1280x720_30.yuv",1280,720,300,30.0 },
-  { "slideedit100","$YUVTMP/SlideEditing_1280x720_30.yuv",1280,720,100,30.0 },
-  { "slideshow",   "$YUVTMP/SlideShow_1280x720_20.yuv",1280,720,500,20.0 },
-  { "slideshow100","$YUVTMP/SlideShow_1280x720_20.yuv",1280,720,100,20.0 },
-  { "screensharing","$HOME/test-screensharing-encoding/Screensharing.yuv",1360,768,4715,60.0 },
-  { NULL }
-};
-
-
-void setInput(const char* input_preset)
-{
-  bool presetFound=false;
-
-  for (int i=0;inputSpeci.name;i++) {
-    if (strcmp(input_preset, inputSpeci.name)==0) {
-      input.setInput(inputSpeci.filename,
-                     inputSpeci.width,
-                     inputSpeci.height,
-                     inputSpeci.fps);
-      input.setMaxFrames(inputSpeci.nFrames);
-      presetFound=true;
-      break;
-    }
-  }
-
-  if (!presetFound) {
-    fprintf(stderr,"no input preset '%s'\n",input_preset);
-    exit(5);
-  }
-}
-
-
-float bitrate(const char* filename)
-{
-  struct stat s;
-  stat(filename,&s);
-
-  long size = s.st_size;
-
-  int frames = input.getNFrames();
-  assert(frames!=0);
-
-  float bitrate = size*8/(frames/input.getFPS());
-  return bitrate;
-}
-
-
-// ---------------------------------------------------------------------------
-
-class Quality
-{
-public:
-  virtual ~Quality() { }
-
-  virtual void measure(const char* h265filename);
-  virtual void measure_yuv(const char* yuvfilename);
-
-  float psnr, ssim;
-};
-
-
-void Quality::measure(const char* h265filename)
-{
-  std::stringstream sstr;
-  sstr << "$DEC265 " << h265filename << " -q -t6 -m " << input.getFilename() << " | grep total "
-    //"| awk '{print $2}' "
-    ">/tmp/xtmp";
-
-  //std::cout << sstr.str() << "\n";
-  int retval = system(replace_variables(sstr.str()).c_str());
-
-  std::ifstream istr;
-  istr.open("/tmp/xtmp");
-  std::string dummy;
-  istr >> dummy >> psnr >> dummy >> dummy >> ssim;
-
-  unlink("/tmp/xtmp");
-}
-
-
-void Quality::measure_yuv(const char* yuvfilename)
-{
-  std::stringstream sstr;
-
-  sstr << "$YUVDIST " << input.getFilename() << " " << yuvfilename
-       << " " << input.getWidth() << " " << input.getHeight()
-       << "|grep total "
-    //"|awk '{print $2}' "
-    ">/tmp/ytmp";
-
-  //std::cout << sstr.str() << "\n";
-  int retval = system(replace_variables(sstr.str()).c_str());
-
-  std::ifstream istr;
-  istr.open("/tmp/ytmp");
-  std::string dummy;
-  istr >> dummy >> psnr >> ssim;
-
-  unlink("/tmp/ytmp");
-}
-
-Quality quality;
-
-// ---------------------------------------------------------------------------
-
-long ticks_per_second;
-
-void init_clock()
-{
-#if !defined(_WIN32) && !defined(WIN32)
-  ticks_per_second = sysconf(_SC_CLK_TCK);
-#endif
-}
-
-double get_cpu_time()
-{
-#if !defined(_WIN32) && !defined(WIN32)
-  struct tms t;
-  times(&t);
-  return double(t.tms_cutime)/ticks_per_second;
-#else
-  return 0; // not supported on windows (TODO)
-#endif
-}
-
-double get_wall_time()
-{
-  struct timeval tv;
-  gettimeofday(&tv, NULL);
-  double t  = tv.tv_sec;
-  double ut = tv.tv_usec/1000000.0f;
-  t += ut;
-  return t;
-}
-
-
-struct RDPoint
-{
-  float rate;
-  float psnr;
-  float ssim;
-  double cpu_time; // computation time in seconds
-  double wall_time;
-
-
-  RDPoint() { }
-
-  void compute_from_h265(std::string stream_name) {
-    rate = bitrate(stream_name.c_str());
-    quality.measure(stream_name.c_str());
-    psnr = quality.psnr;
-    ssim = quality.ssim;
-  }
-
-  void compute_from_yuv(std::string stream_name, std::string yuv_name) {
-    rate = bitrate(stream_name.c_str());
-    quality.measure_yuv(yuv_name.c_str());
-    psnr = quality.psnr;
-    ssim = quality.ssim;
-  }
-
-  void start_timer() {
-    cpu_time = get_cpu_time();
-    wall_time= get_wall_time();
-  }
-
-  void end_timer() {
-    cpu_time = get_cpu_time() - cpu_time;
-    wall_time= get_wall_time()- wall_time;
-  }
-};
-
-
-FILE* output_fh;
-
-void write_rd_line(RDPoint p)
-{
-  fprintf(output_fh,"%9.2f %6.4f %5.3f %5.4f %5.4f\n",
-          p.rate/1024, p.psnr, p.ssim,
-          p.cpu_time/60, p.wall_time/60);
-  fflush(output_fh);
-}
-
-
-
-
-class Encoder
-{
-public:
-  virtual ~Encoder() { }
-
-  virtual std::vector<RDPoint> encode_curve(const Preset& preset) const = 0;
-
-private:
-};
-
-
-class Encoder_de265 : public Encoder
-{
-public:
-  Encoder_de265();
-  void setQPRange(int low,int high,int step) { mQPLow=low; mQPHigh=high; mQPStep=step; }
-
-  virtual std::vector<RDPoint> encode_curve(const Preset& preset) const;
-
-private:
-  RDPoint encode(const Preset& preset,int qp) const;
-
-  int mQPLow,mQPHigh,mQPStep;
-};
-
-
-Encoder_de265::Encoder_de265()
-{
-  mQPLow = 14;
-  mQPHigh= 40;
-  mQPStep=  2;
-}
-
-
-std::vector<RDPoint> Encoder_de265::encode_curve(const Preset& preset) const
-{
-  std::vector<RDPoint> curve;
-
-  for (int qp=mQPHigh ; qp>=mQPLow ; qp-=mQPStep) {
-    curve.push_back(encode(preset, qp));
-  }
-
-  return curve;
-}
-
-
-RDPoint Encoder_de265::encode(const Preset& preset,int qp) const
-{
-  std::stringstream streamname;
-  streamname << "de265-" << preset.name << "-" << qp << ".265";
-
-  std::stringstream cmd1;
-  cmd1 << "$ENC265 " << input.options_de265()
-       << " " << preset.options_de265
-       << " -q " << qp << " -o " << streamname.str()
-       << " " << encoderParameters;
-
-  std::string cmd2 = replace_variables(cmd1.str());
-
-  printf("cmdline: %s\n",cmd2.c_str());
-
-  RDPoint rd;
-  rd.start_timer();
-  int retval = system(cmd2.c_str());
-  rd.end_timer();
-
-  rd.compute_from_h265(streamname.str());
-
-  if (!keepStreams) { unlink(streamname.str().c_str()); }
-
-  write_rd_line(rd);
-
-  return rd;
-}
-
-
-
-
-class Encoder_HM : public Encoder
-{
-public:
-  Encoder_HM();
-
-  void enableSCC(bool flag=true) { useSCC = flag; }
-  void setQPRange(int low,int high,int step) { mQPLow=low; mQPHigh=high; mQPStep=step; }
-
-  virtual std::vector<RDPoint> encode_curve(const Preset& preset) const;
-
-private:
-  RDPoint encode(const Preset& preset,int qp) const;
-
-  bool useSCC;
-  int mQPLow,mQPHigh,mQPStep;
-};
-
-
-Encoder_HM::Encoder_HM()
-{
-  mQPLow = 14;
-  mQPHigh= 40;
-  mQPStep=  2;
-
-  useSCC = false;
-}
-
-
-std::vector<RDPoint> Encoder_HM::encode_curve(const Preset& preset) const
-{
-  std::vector<RDPoint> curve;
-
-  for (int qp=mQPHigh ; qp>=mQPLow ; qp-=mQPStep) {
-    curve.push_back(encode(preset, qp));
-  }
-
-  return curve;
-}
-
-
-RDPoint Encoder_HM::encode(const Preset& preset,int qp) const
-{
-  std::stringstream streamname;
-  streamname << (useSCC ? "hmscc-" : "hm-") << preset.name << "-" << qp << ".265";
-
-  char recoyuv_prefix = "/tmp/reco-XXXXXX";
-  char *tempfile = mktemp(recoyuv_prefix);
-  assert(tempfile != NULL && tempfile0 != 0);
-  std::string recoyuv = std::string(recoyuv_prefix) + ".yuv";
-
-  std::stringstream cmd1;
-  cmd1 << (useSCC ? "$HMSCCENC " : "$HMENC ")
-       << input.options_HM()
-       << " " << (useSCC ? preset.options_hm_scc : preset.options_hm)
-       << " -q " << qp << " -o " << recoyuv << " -b " << streamname.str()
-       << " " << encoderParameters << " >&2";
-
-  std::string cmd2 = replace_variables(cmd1.str());
-
-  std::cout << "CMD: '" << cmd2 << "'\n";
-  RDPoint rd;
-  rd.start_timer();
-  int retval = system(cmd2.c_str());
-  rd.end_timer();
-
-  rd.compute_from_yuv(streamname.str(), recoyuv);
-  if (!keepStreams) { unlink(streamname.str().c_str()); }
-  unlink(recoyuv.c_str());
-
-  write_rd_line(rd);
-
-  return rd;
-}
-
-
-
-class Encoder_x265 : public Encoder
-{
-public:
-  Encoder_x265();
-  void setQPRange(int low,int high,int step) { mQPLow=low; mQPHigh=high; mQPStep=step; }
-
-  virtual std::vector<RDPoint> encode_curve(const Preset& preset) const;
-
-private:
-  RDPoint encode(const Preset& preset,int qp) const;
-
-  int mQPLow,mQPHigh,mQPStep;
-};
-
-
-Encoder_x265::Encoder_x265()
-{
-  /* CRF
-  mQPLow =  4;
-  mQPHigh= 34;
-  mQPStep=  2;
-  */
-
-  mQPLow = 14;
-  mQPHigh= 40;
-  mQPStep=  2;
-}
-
-
-std::vector<RDPoint> Encoder_x265::encode_curve(const Preset& preset) const
-{
-  std::vector<RDPoint> curve;
-
-  for (int qp=mQPHigh ; qp>=mQPLow ; qp-=mQPStep) {
-    curve.push_back(encode(preset, qp));
-  }
-
-  return curve;
-}
-
-
-RDPoint Encoder_x265::encode(const Preset& preset,int qp) const
-{
-  std::stringstream streamname;
-  streamname << "x265-" << preset.name << "-" << qp << ".265";
-
-  std::stringstream cmd1;
-  cmd1 << "$X265ENC " << input.options_x265()
-       << " " << preset.options_x265
-       << " --qp " << qp << " " << streamname.str()
-       << " " << encoderParameters
-       << " >&2";
-
-  std::string cmd2 = replace_variables(cmd1.str());
-
-  //std::cout << "CMD: '" << cmd2 << "'\n";
-  RDPoint rd;
-  rd.start_timer();
-  int retval = system(cmd2.c_str());
-  rd.end_timer();
-
-  rd.compute_from_h265(streamname.str());
-  if (!keepStreams) { unlink(streamname.str().c_str()); }
-
-  write_rd_line(rd);
-
-  return rd;
-}
-
-
-
-
-class Encoder_f265 : public Encoder
-{
-public:
-  Encoder_f265();
-  void setQPRange(int low,int high,int step) { mQPLow=low; mQPHigh=high; mQPStep=step; }
-
-  virtual std::vector<RDPoint> encode_curve(const Preset& preset) const;
-
-private:
-  RDPoint encode(const Preset& preset,int qp) const;
-
-  int mQPLow,mQPHigh,mQPStep;
-};
-
-
-Encoder_f265::Encoder_f265()
-{
-  mQPLow = 14;
-  mQPHigh= 40;
-  mQPStep=  2;
-}
-
-
-std::vector<RDPoint> Encoder_f265::encode_curve(const Preset& preset) const
-{
-  std::vector<RDPoint> curve;
-
-  for (int qp=mQPHigh ; qp>=mQPLow ; qp-=mQPStep) {
-    curve.push_back(encode(preset, qp));
-  }
-
-  return curve;
-}
-
-
-RDPoint Encoder_f265::encode(const Preset& preset,int qp) const
-{
-  std::stringstream cmd1;
-  cmd1 << "$F265 " << input.options_f265()
-       << " f265.out -v -p\"" << preset.options_f265 << " qp=" << qp
-       << " " << encoderParameters
-       << "\" >&2";
-
-  std::string cmd2 = replace_variables(cmd1.str());
-
-  std::cout << "CMD: '" << cmd2 << "'\n";
-  RDPoint rd;
-  rd.start_timer();
-  int retval = system(cmd2.c_str());
-  rd.end_timer();
-
-  rd.compute_from_h265("f265.out");
-  if (!keepStreams) { unlink("f265.out"); }
-
-  write_rd_line(rd);
-
-  return rd;
-}
-
-
-
-class Encoder_x264 : public Encoder
-{
-public:
-  Encoder_x264();
-  //void setCRFRange(int low,int high,int step) { mCRFLow=low; mCRFHigh=high; mCRFStep=step; }
-
-  virtual std::vector<RDPoint> encode_curve(const Preset& preset) const;
-
-private:
-  RDPoint encode(const Preset& preset,int crf) const;
-
-  int mCRFLow,mCRFMid,mCRFHigh;
-  int mCRFStepHigh, mCRFStepLow;
-};
-
-
-Encoder_x264::Encoder_x264()
-{
-  // in the upper bit-rate range mid;high, use larger CRF step-size 'StepHigh'
-  // in the lower bit-rate range low;mid, use smaller CRF step-size 'StepLow'
-
-  mCRFLow = 10;
-  mCRFMid = 20;
-  mCRFHigh= 36;
-  mCRFStepHigh= 2;
-  mCRFStepLow = 1;
-}
-
-
-std::vector<RDPoint> Encoder_x264::encode_curve(const Preset& preset) const
-{
-  std::vector<RDPoint> curve;
-
-  for (int crf=mCRFLow ; crf<mCRFMid ; crf+=mCRFStepHigh) {
-    curve.push_back(encode(preset, crf));
-  }
-
-  for (int crf=mCRFMid ; crf<=mCRFHigh ; crf+=mCRFStepLow) {
-    curve.push_back(encode(preset, crf));
-  }
-
-  return curve;
-}
-
-
-RDPoint Encoder_x264::encode(const Preset& preset,int qp_crf) const
-{
-  std::stringstream streamname;
-  streamname << "x264-" << preset.name << "-" << qp_crf << ".264";
-
-  std::stringstream cmd1;
-#if 0
-  cmd1 << "$X264 " << input.options_x264()
-       << " " << preset.options_x264
-       << " --crf " << qp_crf
-       << " -o " << streamname.str();
-#else
-  cmd1 << "$FFMPEG " << input.options_ffmpeg()
-       << " " << preset.options_x264_ffmpeg
-       << " -crf " << qp_crf
-       << " -threads 6"
-       << " -f h264 " << streamname.str()
-       << " " << encoderParameters;
-#endif
-
-  std::string cmd2 = replace_variables(cmd1.str());
-
-  std::cerr << "-----------------------------\n";
-
-  std::cerr << "CMD: '" << cmd2 << "'\n";
-
-  RDPoint rd;
-  rd.start_timer();
-  int retval = system(cmd2.c_str());
-  rd.end_timer();
-
-  char tmpyuv_prefix = "/tmp/rdout-XXXXXX";
-  char *tempfile = mktemp(tmpyuv_prefix);
-  assert(tempfile != NULL && tempfile0 != 0);
-  std::string tmpyuv = std::string(tmpyuv_prefix) + ".yuv";
-
-  std::string cmd3 = "ffmpeg -i " + streamname.str() + " -threads 6 " + tmpyuv;
-
-  retval = system(cmd3.c_str());
-
-  rd.compute_from_yuv(streamname.str(), tmpyuv);
-
-  unlink(tmpyuv.c_str());
-  if (!keepStreams) { unlink(streamname.str().c_str()); }
-
-  write_rd_line(rd);
-
-  return rd;
-}
-
-
-class Encoder_mpeg2 : public Encoder
-{
-public:
-  Encoder_mpeg2();
-
-  virtual std::vector<RDPoint> encode_curve(const Preset& preset) const;
-
-private:
-  RDPoint encode(const Preset& preset,int bitrate) const;
-};
-
-
-Encoder_mpeg2::Encoder_mpeg2()
-{
-}
-
-
-std::vector<RDPoint> Encoder_mpeg2::encode_curve(const Preset& preset) const
-{
-  std::vector<RDPoint> curve;
-
-  int bitrates = { 250,500,750,1000,1250,1500,1750,2000,2500,3000,3500,4000,4500,5000,
-                     6000,7000,8000,9000,10000,12000,14000,16000,18000,20000,25000,30000,
-                     -1 };
-
-  for (int i=0; bitratesi>0; i++) {
-    curve.push_back(encode(preset, bitratesi));
-  }
-
-  return curve;
-}
-
-
-RDPoint Encoder_mpeg2::encode(const Preset& preset,int br) const
-{
-  std::stringstream streamname;
-  streamname << "mpeg2-" << preset.name << "-"
-             << std::setfill('0') << std::setw(5) << br << ".mp2";
-
-  std::stringstream cmd1;
-  cmd1 << "$FFMPEG " << input.options_ffmpeg()
-       << " " << preset.options_x264_ffmpeg
-       << " -b " << br << "k "
-       << " -threads 6"
-       << " -f mpeg2video " << streamname.str()
-       << " " << encoderParameters;
-
-  std::string cmd2 = replace_variables(cmd1.str());
-
-  std::cerr << "-----------------------------\n";
-
-  std::cerr << "CMD: '" << cmd2 << "'\n";
-
-  RDPoint rd;
-  rd.start_timer();
-  int retval = system(cmd2.c_str());
-  rd.end_timer();
-
-  char tmpyuv_prefix = "/tmp/rdout-XXXXXX";
-  char *tempfile = mktemp(tmpyuv_prefix);
-  assert(tempfile != NULL && tempfile0 != 0);
-  std::string tmpyuv = std::string(tmpyuv_prefix) + ".yuv";
-
-  std::string cmd3 = "ffmpeg -i " + streamname.str() + " -threads 6 " + tmpyuv;
-
-  retval = system(cmd3.c_str());
-
-  rd.compute_from_yuv(streamname.str(), tmpyuv);
-
-  unlink(tmpyuv.c_str());
-  if (!keepStreams) { unlink(streamname.str().c_str()); }
-
-  write_rd_line(rd);
-
-  return rd;
-}
-
-
-Encoder_de265 enc_de265;
-Encoder_HM enc_hm;
-Encoder_x265 enc_x265;
-Encoder_f265 enc_f265;
-Encoder_x264 enc_x264;
-Encoder_mpeg2 enc_mpeg2;
-
-// ---------------------------------------------------------------------------
-
-static struct option long_options = {
-  {"keep-streams",      no_argument,       0, 'k' },
-  //{"write-bytestream", required_argument,0, 'B' },
-  {0,         0,                 0,  0 }
-};
-
-
-void show_usage()
-{
-  fprintf(stderr,
-          "usage: rd-curves 'preset_id' 'input_preset' 'encoder'\n"
-          "supported encoders: de265 / hm / hmscc / x265 / f265 / x264 / mpeg2\n");
-  fprintf(stderr,
-          "presets:\n");
-
-  for (int i=0;preseti.name!=NULL;i++) {
-    fprintf(stderr,
-            " %2d %-20s %s\n",preseti.ID,preseti.name,preseti.descr);
-  }
-
-  fprintf(stderr,
-          "\ninput presets:\n");
-  for (int i=0;inputSpeci.name;i++) {
-    fprintf(stderr,
-            " %-12s %-30s %4dx%4d, %4d frames, %5.2f fps\n",
-            inputSpeci.name,
-            inputSpeci.filename,
-            inputSpeci.width,
-            inputSpeci.height,
-            inputSpeci.nFrames,
-            inputSpeci.fps);
-  }
-}
-
-int main(int argc, char** argv)
-{
-  init_clock();
-
-  while (1) {
-    int option_index = 0;
-
-    int c = getopt_long(argc, argv, "kf:p:",
-                        long_options, &option_index);
-    if (c == -1)
-      break;
-
-    switch (c) {
-    case 'k': keepStreams=true; break;
-    case 'f': maxFrames=atoi(optarg); break;
-    case 'p': encoderParameters=optarg; break;
-    }
-  }
-
-  if (optind != argc-3) {
-    show_usage();
-    exit(5);
-  }
-
-  int presetID = atoi( argvoptind );
-  const char* inputName = argvoptind+1;
-  const char* encoderName = argvoptind+2;
-
-  int presetIdx = -1;
-
-  for (int i=0;preseti.name != NULL;i++) {
-    if (preseti.ID == presetID) {
-      presetIdx = i;
-      break;
-    }
-  }
-
-  if (presetIdx == -1) {
-    fprintf(stderr,"preset ID %d does not exist\n",presetID);
-    exit(5);
-  }
-
-  setInput(inputName);
-  if (maxFrames) input.setMaxFrames(maxFrames);
-
-
-  Encoder* enc = NULL;
-  /**/ if (strcmp(encoderName,"de265")==0) { enc = &enc_de265; }
-  else if (strcmp(encoderName,"hm"   )==0) { enc = &enc_hm;   }
-  else if (strcmp(encoderName,"hmscc")==0) { enc = &enc_hm;   enc_hm.enableSCC(); }
-  else if (strcmp(encoderName,"x265" )==0) { enc = &enc_x265; }
-  else if (strcmp(encoderName,"f265" )==0) { enc = &enc_f265; }
-  else if (strcmp(encoderName,"x264" )==0) { enc = &enc_x264; }
-  else if (strcmp(encoderName,"mpeg2")==0) { enc = &enc_mpeg2; }
-
-  if (enc==NULL) {
-    fprintf(stderr, "unknown encoder");
-    exit(5);
-  }
-
-
-  std::stringstream data_filename;
-  data_filename << encoderName << "-" << inputName << "-" << presetpresetIdx.name << ".rd";
-  output_fh = fopen(data_filename.str().c_str(), "wb");
-
-  fprintf(output_fh,"# %s\n", presetpresetIdx.descr);
-  fprintf(output_fh,"# 1:rate 2:psnr 3:ssim 4:cputime(min) 5:walltime(min)\n");
-
-  std::vector<RDPoint> curve = enc->encode_curve(presetpresetIdx);
-
-  for (int i=0;i<curve.size();i++) {
-    //fprintf(out_fh,"%7.2f %6.4f\n", curvei.rate/1024, curvei.psnr);
-  }
-
-  fclose(output_fh);
-
-  return 0;
-}

libde265-1.0.17.tar.gz/tools/tests.cc Deleted

@@ -1,100 +0,0 @@
-/*
- * H.265 video codec.
- * Copyright (c) 2013-2014 struktur AG, Dirk Farin <farin@struktur.de>
- *
- * This file is part of libde265.
- *
- * libde265 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 3 of the License, or
- * (at your option) any later version.
- *
- * libde265 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 libde265.  If not, see <http://www.gnu.org/licenses/>.
- */
-
-#include <stdio.h>
-#include <iostream>
-#include <string.h>
-
-
-class Test
-{
-public:
-  Test() { next=s_firstTest; s_firstTest=this; }
-  virtual ~Test() { }
-
-  virtual const char* getName() const { return "noname"; }
-  virtual const char* getDescription() const { return "no description"; }
-  virtual bool work(bool quiet=false) = 0;
-
-  static void runTest(const char* name) {
-    Test* t = s_firstTest;
-    while (t) {
-      if (strcmp(t->getName(), name)==0) {
-        t->work();
-        break;
-      }
-      t=t->next;
-    }
-  }
-
-  static void runAllTests() {
-    Test* t = s_firstTest;
-    while (t) {
-      printf("%s ... ",t->getName());
-      fflush(stdout);
-      if (t->work(true) == false) {
-        printf("*** FAILED ***\n");
-      }
-      else {
-        printf("passed\n");
-      }
-
-      t=t->next;
-    }
-  }
-
-public:
-  Test* next;
-  static Test* s_firstTest;
-};
-
-Test* Test::s_firstTest = NULL;
-
-
-class ListTests : public Test
-{
-public:
-  const char* getName() const { return "list"; }
-  const char* getDescription() const { return "list all available tests"; }
-  bool work(bool quiet) {
-    if (!quiet) {
-      Test* t = s_firstTest;
-      while (t) {
-        printf("- %s: %s\n",t->getName(), t->getDescription());
-        t=t->next;
-      }
-    }
-    return true;
-  }
-} listtest;
-
-
-
-int main(int argc,char** argv)
-{
-  if (argc>=2) {
-    Test::runTest(argv1);
-  }
-  else {
-    Test::runAllTests();
-  }
-
-  return 0;
-}

libde265-1.0.17.tar.gz/tools/yuv-distortion.cc Deleted

@@ -1,113 +0,0 @@
-/*
- * H.265 video codec.
- * Copyright (c) 2013-2014 struktur AG, Dirk Farin <farin@struktur.de>
- *
- * This file is part of libde265.
- *
- * libde265 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 3 of the License, or
- * (at your option) any later version.
- *
- * libde265 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 libde265.  If not, see <http://www.gnu.org/licenses/>.
- */
-
-#ifdef HAVE_CONFIG_H
-#include "config.h"
-#endif
-
-#include <stdio.h>
-#include <stdlib.h>
-
-#include <libde265/quality.h>
-
-#if HAVE_VIDEOGFX
-#include <libvideogfx.hh>
-using namespace videogfx;
-#endif
-
-
-float ssim(const uint8_t* img1,
-           const uint8_t* img2,
-           int width, int height)
-{
-#if HAVE_VIDEOGFX
-  Bitmap<Pixel> ref, coded;
-  ref  .Create(width, height); // reference image
-  coded.Create(width, height); // coded image
-
-  for (int y=0;y<height;y++) {
-    memcpy(codedy, img1 + y*width, width);
-    memcpy(refy,   img2 + y*width, width);
-  }
-
-  SSIM ssimAlgo;
-  return ssimAlgo.calcMSSIM(ref,coded);
-#else
-  return 0;
-#endif
-}
-
-
-int main(int argc, char** argv)
-{
-  if (argc != 5) {
-    fprintf(stderr,"need two YUV files and image size as input: FILE1 FILE2 WIDTH HEIGHT\n");
-    exit(5);
-  }
-
-
-  FILE* fh_ref = fopen(argv1,"rb");
-  FILE* fh_cmp = fopen(argv2,"rb");
-
-  int width  = atoi(argv3);
-  int height = atoi(argv4);
-
-  uint8_t* yp_ref = (uint8_t*)malloc(width*height);
-  uint8_t* yp_cmp = (uint8_t*)malloc(width*height);
-
-  double mse_y=0.0, ssim_y=0.0;
-  int nFrames=0;
-
-  for (;;)
-    {
-      if (fread(yp_ref,1,width*height,fh_ref) != width*height) {
-        break;
-      }
-      if (fread(yp_cmp,1,width*height,fh_cmp) != width*height) {
-        break;
-      }
-
-      if (feof(fh_ref)) break;
-      if (feof(fh_cmp)) break;
-
-      fprintf(stderr,"yuv-distortion processing frame %d\r",nFrames+1);
-
-      fseek(fh_ref,width*height/2,SEEK_CUR);
-      fseek(fh_cmp,width*height/2,SEEK_CUR);
-
-      double curr_mse_y = MSE(yp_ref, width,  yp_cmp, width,  width, height);
-      mse_y += curr_mse_y;
-
-      double curr_ssim_y = ssim(yp_ref, yp_cmp, width, height);
-      ssim_y += curr_ssim_y;
-
-      printf("%4d %f %f\n",nFrames,PSNR(curr_mse_y),curr_ssim_y);
-
-      nFrames++;
-    }
-
-  printf("total: %f %f\n",PSNR(mse_y/nFrames),ssim_y/nFrames);
-  fprintf(stderr,"\n");
-
-  fclose(fh_ref);
-  fclose(fh_cmp);
-
-  return 0;
-}

libde265-1.0.17.tar.gz/CMakeLists.txt -> libde265-1.1.1.tar.gz/CMakeLists.txt Changed

@@ -2,7 +2,7 @@
 
 project (libde265
     LANGUAGES C CXX
-    VERSION 1.0.17
+    VERSION 1.1.1
 )
 
 # Auto-compute BCD-encoded numeric version from project version.
@@ -19,10 +19,10 @@
 # This controls the shared library filename:
 #   libde265.so -> libde265.so.0 -> libde265.so.0.1.9
 #                  ^^^^^^^^^^^^^^    ^^^^^^^^^^^^^^^^^^
-#                  SOVERSION         VERSION
+#                  DE265_SOVERSION   DE265_LIBRARY_VERSION
 #
-# VERSION is a three-part number: MAJOR.MINOR.PATCH
-#   - MAJOR = the ABI major version (same as SOVERSION).
+# DE265_LIBRARY_VERSION is a three-part number: MAJOR.MINOR.PATCH
+#   - MAJOR = the ABI major version (same as DE265_SOVERSION).
 #             Bump when the ABI breaks (exported functions removed or signatures changed).
 #             Reset MINOR and PATCH to 0.
 #   - MINOR = backward-compatible ABI additions.
@@ -31,20 +31,21 @@
 #   - PATCH = implementation-only changes (bug fixes, performance improvements).
 #             Bump when the ABI is unchanged.
 #
-# SOVERSION must always equal the MAJOR part of VERSION.
+# DE265_SOVERSION must always equal the MAJOR part of DE265_LIBRARY_VERSION.
 # Programs linked against libde265.so.0 will work with any libde265.so.0.x.y.
 #
 set(DE265_SOVERSION 0)
-set(DE265_LIBRARY_VERSION "0.1.10")
+set(DE265_LIBRARY_VERSION "0.2.1")
 
 set(CMAKE_CXX_STANDARD 17)
 set(CMAKE_CXX_STANDARD_REQUIRED ON)
 set(CMAKE_CXX_EXTENSIONS OFF)
 set(CMAKE_POSITION_INDEPENDENT_CODE ON)
 
-if (CMAKE_BUILD_TYPE STREQUAL "Debug")
-  add_compile_options(-Wall -Wextra -Wpedantic -Wno-unused-parameter)
-endif()
+add_compile_options(
+  "$<$<AND:$<CONFIG:Debug>,$<NOT:$<CXX_COMPILER_ID:MSVC>>>:-Wall;-Wextra;-Wpedantic;-Wno-unused-parameter>"
+  "$<$<AND:$<CONFIG:Debug>,$<CXX_COMPILER_ID:MSVC>>:/W4>"
+)
 
 option(USE_IWYU "Run include-what-you-use analysis during build" OFF)
 if (USE_IWYU)
@@ -72,42 +73,66 @@
 CHECK_INCLUDE_FILE(malloc.h HAVE_MALLOC_H)
 CHECK_FUNCTION_EXISTS(posix_memalign HAVE_POSIX_MEMALIGN)
 
+option(ENABLE_SIMD "Enable SIMD acceleration (SSE on x86, NEON on ARM)" ON)
+option(ENABLE_AVX2 "Enable AVX2 kernels on x86 (in addition to SSE)" ON)
+option(ENABLE_AVX512 "Enable AVX-512 kernels on x86 (in addition to AVX2)" ON)
+
 include(CheckCSourceCompiles)
-check_c_source_compiles(
-  "#if !defined(__x86_64) && !defined(__i386__) \
-  && !defined(_M_IX86) && !defined(_M_AMD64) \
-  || defined(_M_ARM64EC) || defined(_ARM64EC_)
-  #error not x86
-  #endif
-  int main(){return 0;}"
-  HAVE_X86)
-
-if(HAVE_X86)
-  if (MSVC)
-    set(SUPPORTS_SSE2 1)
-    set(SUPPORTS_SSSE3 1)
-    set(SUPPORTS_SSE4_1 1)
-  else()
-    check_c_compiler_flag(-msse2 SUPPORTS_SSE2)
-    check_c_compiler_flag(-mssse3 SUPPORTS_SSSE3)
-    check_c_compiler_flag(-msse4.1 SUPPORTS_SSE4_1)
-  endif()
-  if(SUPPORTS_SSE4_1)
-    set(HAVE_SSE4_1 TRUE)
+if(ENABLE_SIMD)
+  check_c_source_compiles(
+    "#if !defined(__x86_64) && !defined(__i386__) \
+    && !defined(_M_IX86) && !defined(_M_AMD64) \
+    || defined(_M_ARM64EC) || defined(_ARM64EC_)
+    #error not x86
+    #endif
+    int main(){return 0;}"
+    HAVE_X86)
+
+  if(HAVE_X86)
+    if (MSVC)
+      set(SUPPORTS_SSE2 1)
+      set(SUPPORTS_SSSE3 1)
+      set(SUPPORTS_SSE4_1 1)
+    else()
+      check_c_compiler_flag(-msse2 SUPPORTS_SSE2)
+      check_c_compiler_flag(-mssse3 SUPPORTS_SSSE3)
+      check_c_compiler_flag(-msse4.1 SUPPORTS_SSE4_1)
+      check_c_compiler_flag(-mavx2 SUPPORTS_AVX2)
+      check_c_compiler_flag("-mavx512f -mavx512bw" SUPPORTS_AVX512)
+    endif()
+    if(SUPPORTS_SSE4_1)
+      set(HAVE_SSE4_1 TRUE)
+    endif()
+    # AVX2 kernels back the SSE4.1 ones; only built on 64-bit targets.
+    if(ENABLE_AVX2 AND SUPPORTS_AVX2 AND HAVE_SSE4_1 AND CMAKE_SIZEOF_VOID_P EQUAL 8)
+      set(HAVE_AVX2 TRUE)
+    endif()
+    # AVX-512 (F+BW) kernels back the AVX2 ones.
+    if(ENABLE_AVX512 AND SUPPORTS_AVX512 AND HAVE_AVX2)
+      set(HAVE_AVX512 TRUE)
+    endif()
   endif()
-endif()
 
-check_c_source_compiles(
-  "#if !defined(__arm__) && !defined(__aarch64__) \
-  && !defined(_M_ARM) && !defined(_M_ARM64)
-  #error not ARM
-  #endif
-  int main(){return 0;}"
-  HAVE_ARM)
+  check_c_source_compiles(
+    "#if !defined(__arm__) && !defined(_M_ARM)
+    #error not 32-bit ARM
+    #endif
+    int main(){return 0;}"
+    HAVE_ARM32)
+
+  check_c_source_compiles(
+    "#if !defined(__aarch64__) && !defined(_M_ARM64)
+    #error not 64-bit ARM
+    #endif
+    int main(){return 0;}"
+    HAVE_ARM64)
 
-if(HAVE_ARM)
-  enable_language(ASM)
-  check_c_compiler_flag(-mfpu=neon HAVE_NEON)
+  if(HAVE_ARM32)
+    enable_language(ASM)
+    check_c_compiler_flag(-mfpu=neon HAVE_NEON)
+  endif()
+else()
+  message(STATUS "SIMD acceleration disabled (ENABLE_SIMD=OFF) — using scalar fallback only")
 endif()
 
 configure_file (libde265/de265-version.h.in libde265/de265-version.h)
@@ -118,16 +143,31 @@
   add_definitions(-Wall -Werror=return-type -Werror=unused-result -Werror=reorder)
 endif()
 
+set(THREADS_PREFER_PTHREAD_FLAG TRUE)
+find_package(Threads REQUIRED)
+
 include(CheckCXXSymbolExists)
 check_cxx_symbol_exists(_LIBCPP_VERSION cstdlib HAVE_LIBCPP)
-if(HAVE_LIBCPP)
-  set(LIBS_PRIVATE "-lc++")
-else()
-  set(LIBS_PRIVATE "-lstdc++")
+
+# Build Libs.private contents for libde265.pc.
+set(_libs_private "")
+if(NOT MSVC)
+  if(HAVE_LIBCPP)
+    list(APPEND _libs_private "-lc++")
+  else()
+    list(APPEND _libs_private "-lstdc++")
+  endif()
 endif()
+# Emit '-lpthread' rather than '-pthread' so consumers using
+# `pkg-config --libs --static` get an unambiguous linker flag (see #453).
+if(CMAKE_USE_PTHREADS_INIT)
+  list(APPEND _libs_private "-lpthread")
+endif()
+if(UNIX)
+  list(APPEND _libs_private "-lm")
+endif()
+string(JOIN " " LIBS_PRIVATE ${_libs_private})
 
-configure_file (libde265.pc.in libde265.pc @ONLY)
-install(FILES ${CMAKE_CURRENT_BINARY_DIR}/libde265.pc DESTINATION ${CMAKE_INSTALL_LIBDIR}/pkgconfig)
 
 option(BUILD_SHARED_LIBS "Build shared library" ON)
 if(NOT BUILD_SHARED_LIBS)
@@ -145,12 +185,10 @@
   include_directories ("${PROJECT_SOURCE_DIR}/extra")
 endif()
 
-find_package(Threads)
-
 option(ENABLE_DECODER "Enable Decoder" ON)
 option(ENABLE_ENCODER "Enable Encoder" OFF)
 option(ENABLE_SHERLOCK265 "Build sherlock265 visual inspection tool" OFF)
-option(ENABLE_TOOLS "Build tools" OFF)
+option(ENABLE_INTERNAL_DEVELOPMENT_TOOLS "Build internal development tools (not for end users)" OFF)
 option(WITH_FUZZERS "Build the fuzzers" OFF)
 set(FUZZING_SANITIZER_OPTIONS "-fsanitize=address,shift,integer" "-fno-sanitize=unsigned-shift-base" "-fno-sanitize-recover=shift,integer" CACHE STRING "Sanitizer flags for fuzzing builds")
 
@@ -159,6 +197,47 @@
   add_link_options(${FUZZING_SANITIZER_OPTIONS})
 endif()
 
+# --- Symbol visibility -------------------------------------------------------
+#
+# Hiding the library's non-API symbols turns cross-translation-unit calls inside
+# libde265 from indirect PLT/GOT dispatch into direct calls — a measurable win
+# on the hot CABAC decode path — and shrinks the dynamic symbol table. Only the
+# symbols tagged LIBDE265_API (the public de265_*/en265_* C API) stay exported.
+#
+# By default we reduce visibility for optimized (non-Debug) builds and keep
+# every symbol exported for Debug builds, so a debug build stays fully available
+# for introspection. The encoder CLI, sherlock265 and the dev tools all link
+# against *internal* (non-API) symbols of the library, so hiding would break
+# their link; visibility is therefore never reduced when any of those is built.
+#
+# Set FORCE_FULL_VISIBILITY=ON to override the default and export all symbols
+# even from an optimized build.
+option(FORCE_FULL_VISIBILITY
+  "Export all symbols, even from optimized builds (default: hide non-API symbols in optimized builds without the encoder/tools)"
+  OFF)
+
+set(_de265_internal_consumers OFF)
+if(ENABLE_ENCODER OR ENABLE_SHERLOCK265 OR ENABLE_INTERNAL_DEVELOPMENT_TOOLS)
+  set(_de265_internal_consumers ON)
+endif()
+
+set(_de265_optimized_build OFF)
+if(CMAKE_BUILD_TYPE)
+  string(TOUPPER "${CMAKE_BUILD_TYPE}" _de265_build_type_upper)
+  if(NOT _de265_build_type_upper STREQUAL "DEBUG")
+    set(_de265_optimized_build ON)
+  endif()
+endif()
+
+set(DE265_REDUCED_VISIBILITY OFF)
+if(_de265_optimized_build AND NOT _de265_internal_consumers AND NOT FORCE_FULL_VISIBILITY)
+  set(DE265_REDUCED_VISIBILITY ON)
+  set(CMAKE_C_VISIBILITY_PRESET hidden)
+  set(CMAKE_CXX_VISIBILITY_PRESET hidden)
+  set(CMAKE_VISIBILITY_INLINES_HIDDEN ON)
+  message(STATUS "Reduced symbol visibility enabled (only LIBDE265_API symbols exported)")
+endif()
+
 add_subdirectory (libde265)
 if (ENABLE_DECODER)
   add_subdirectory (dec265)
@@ -169,8 +248,8 @@
 if (ENABLE_SHERLOCK265)
   add_subdirectory (sherlock265)
 endif()
-if (ENABLE_TOOLS)
-  add_subdirectory (tools)
+if (ENABLE_INTERNAL_DEVELOPMENT_TOOLS)
+  add_subdirectory (dev-tools)
 endif()
 if (WITH_FUZZERS)
   add_subdirectory (fuzzing)

libde265-1.0.17.tar.gz/README.md -> libde265-1.1.1.tar.gz/README.md Changed

@@ -54,7 +54,7 @@
 
 !Build Status(https://github.com/strukturag/libde265/workflows/build/badge.svg)(https://github.com/strukturag/libde265/actions) !Build Status(https://ci.appveyor.com/api/projects/status/github/strukturag/libde265?svg=true)(https://ci.appveyor.com/project/strukturag/libde265)
 
-libde265 uses the CMake build system. Please do not use to deprecated autotools scripts.
+libde265 uses the CMake build system.
 To compile libde265, run
 ````
 mkdir build
@@ -81,20 +81,6 @@
   http://github.com/farindk/libvideogfx
 
 
-You can disable building of the example programs by running `./configure` with
-<pre>
-  --disable-dec265        Do not build the dec265 decoder program.
-  --disable-sherlock265   Do not build the sherlock265 visual inspection program.
-</pre>
-
-Additional logging information can be turned on and off using these `./configure` flags:
-<pre>
-  --enable-log-error      turn on logging at error level (default=yes)
-  --enable-log-info       turn on logging at info level (default=no)
-  --enable-log-trace      turn on logging at trace level (default=no)
-</pre>
-
-
 Build using cmake
 =================
 
@@ -108,8 +94,16 @@
 make
 ```
 
-See the cmake documentation(http://www.cmake.org) for further information on
-using cmake on other platforms.
+You can disable building of the example programs by running `cmake` with
+<pre>
+  -DENABLE_DECODER=off     Do not build the dec265 decoder program.
+  -DENABLE_SHERLOCK265=off Do not build the sherlock265 visual inspection program.
+</pre>
+
+Additional logging information can be turned on and off using these `./configure` flags:
+<pre>
+  -DDE265_LOG_LEVEL={error;info;debug;trace}
+</pre>
 
 
 Building using vcpkg

libde265-1.1.1.tar.gz/SECURITY.md Added

@@ -0,0 +1,58 @@
+# Security Policy
+
+## Reporting a Vulnerability
+
+If you believe you have found a security vulnerability in libde265, **please do
+not open a public GitHub issue**. Instead, report it privately via one of the
+following channels:
+
+- **Preferred:** Use GitHub's private vulnerability reporting by clicking
+  *Report a vulnerability* on the
+  Security tab(https://github.com/strukturag/libde265/security/advisories/new)
+  of this repository. This creates a private draft advisory that only the
+  maintainers can see.
+- **Alternative:** If you cannot use GitHub, send an email to
+  <dirk.farin@gmail.com> describing the issue.
+
+When reporting, please include as much of the following as you can:
+
+- A description of the vulnerability and its potential impact.
+- Steps to reproduce, ideally with a minimal H.265 bitstream that triggers the
+  issue.
+- The libde265 version (release tag or commit hash) and build configuration
+  (compiler, sanitizers, etc.) where the issue was observed.
+- Any suggested fix or mitigation, if known.
+
+We will acknowledge your report, work with you on a fix, and credit you in the
+release notes and advisory unless you prefer to remain anonymous.
+
+## Supported Versions
+
+Security fixes are applied to the latest release on the `master` branch. Older
+releases are not maintained; please update to the current release before
+reporting.
+
+## Scope
+
+In scope:
+
+- Memory safety issues in the decoder (out-of-bounds reads/writes, use-after-
+  free, double-free, uninitialized memory use).
+- Crashes, hangs, or excessive resource consumption triggered by crafted H.265
+  bitstreams.
+- Issues in the public C API (`libde265/de265.h`) that lead to the above when
+  used as documented.
+
+Out of scope:
+
+- Bugs in the experimental encoder (`libde265/encoder/`, built with
+  `-DENABLE_ENCODER=ON`).
+- Bugs in the sample applications (`dec265/`, `enc265/`, `sherlock265/`) that
+  do not also affect the library.
+- Issues that require a non-default, unsafe build configuration or modified
+  source.
+
+## Disclosure
+
+We aim to coordinate disclosure with the reporter. Once a fix is available, we
+will publish a GitHub Security Advisory and request a CVE where appropriate.

libde265-1.0.17.tar.gz/cmake/config.h.in -> libde265-1.1.1.tar.gz/cmake/config.h.in Changed

libde265-1.1.1.tar.gz/cmake/toolchains/arm-linux-gnueabihf-clang.cmake Added

libde265-1.0.17.tar.gz/dec265/dec265.cc -> libde265-1.1.1.tar.gz/dec265/dec265.cc Changed

@@ -271,18 +271,20 @@
 
   de265_chroma chroma = de265_get_chroma_format(img);
 
+  enum SDL_YUV_Display::SDL_Chroma sdlChroma;
+  switch (chroma) {
+  case de265_chroma_420:  sdlChroma = SDL_YUV_Display::SDL_CHROMA_420;  break;
+  case de265_chroma_422:  sdlChroma = SDL_YUV_Display::SDL_CHROMA_422;  break;
+  case de265_chroma_444:  sdlChroma = SDL_YUV_Display::SDL_CHROMA_444;  break;
+  case de265_chroma_mono: sdlChroma = SDL_YUV_Display::SDL_CHROMA_MONO; break;
+  default: assert(false); sdlChroma = SDL_YUV_Display::SDL_CHROMA_MONO;
+  }
+
   if (!sdl_active) {
     sdl_active=true;
-    enum SDL_YUV_Display::SDL_Chroma sdlChroma;
-    switch (chroma) {
-    case de265_chroma_420:  sdlChroma = SDL_YUV_Display::SDL_CHROMA_420;  break;
-    case de265_chroma_422:  sdlChroma = SDL_YUV_Display::SDL_CHROMA_422;  break;
-    case de265_chroma_444:  sdlChroma = SDL_YUV_Display::SDL_CHROMA_444;  break;
-    case de265_chroma_mono: sdlChroma = SDL_YUV_Display::SDL_CHROMA_MONO; break;
-      default: assert(false); sdlChroma = SDL_YUV_Display::SDL_CHROMA_MONO;
-    }
-
-    sdlWin.init(width,height, sdlChroma);
+    if (!sdlWin.init(width,height, sdlChroma)) return true;
+  } else {
+    if (!sdlWin.resize(width,height, sdlChroma)) return true;
   }
 
   int stride,chroma_stride;

libde265-1.0.17.tar.gz/dec265/sdl-display.cc -> libde265-1.1.1.tar.gz/dec265/sdl-display.cc Changed

@@ -25,6 +25,7 @@
 */
 
 #include "sdl-display.h"
+#include <cstdio>
 #include <cstring>
 
 
@@ -94,6 +95,43 @@
   return true;
 }
 
+bool SDL_YUV_Display::resize(int frame_width, int frame_height, enum SDL_Chroma chroma)
+{
+  if (!mWindowOpen) {
+    return init(frame_width, frame_height, chroma);
+  }
+
+  // SDL_PIXELFORMAT_YV12 requires even dimensions; init() rounds down to a
+  // multiple of 8, so we do the same here for consistency.
+  frame_width  &= ~7;
+  frame_height &= ~7;
+
+  if (frame_width == rect.w && frame_height == rect.h && mChroma == chroma) {
+    return true;
+  }
+
+  // All chroma formats currently map to SDL_PIXELFORMAT_YV12 (we down-convert
+  // 4:2:2 and 4:4:4 ourselves), so the texture pixel format never changes.
+  // Only the texture dimensions and the window size need updating.
+  SDL_DestroyTexture(mTexture);
+  mTexture = SDL_CreateTexture(mRenderer, SDL_PIXELFORMAT_YV12,
+                               SDL_TEXTUREACCESS_STREAMING,
+                               frame_width, frame_height);
+  if (!mTexture) {
+    printf("SDL: Couldn't recreate SDL texture: %s\n", SDL_GetError());
+    mWindowOpen = false;
+    return false;
+  }
+
+  SDL_SetWindowSize(mWindow, frame_width, frame_height);
+
+  mChroma = chroma;
+  rect.w  = frame_width;
+  rect.h  = frame_height;
+
+  return true;
+}
+
 void SDL_YUV_Display::display(const unsigned char *Y,
                               const unsigned char *U,
                               const unsigned char *V,
@@ -247,7 +285,7 @@
     }
 
   uint8_t *startV = mPixels + (rect.h*mStride);
-  uint8_t *startU = startV + (rect.h*mStride/2);
+  uint8_t *startU = startV + (rect.h*mStride/4);
   for (int y=0;y<rect.h;y+=2)
     {
       unsigned char* u = startU + y/2*mStride/2;

libde265-1.0.17.tar.gz/dec265/sdl-display.h -> libde265-1.1.1.tar.gz/dec265/sdl-display.h Changed

libde265-1.0.17.tar.gz/libde265.pc.in -> libde265-1.1.1.tar.gz/libde265.pc.in Changed

libde265-1.0.17.tar.gz/libde265/CMakeLists.txt -> libde265-1.1.1.tar.gz/libde265/CMakeLists.txt Changed

@@ -10,7 +10,9 @@
   decctx.cc
   dpb.cc
   fallback-dct.cc
-  fallback-motion.cc 
+  fallback-deblk.cc
+  fallback-intrapred.cc
+  fallback-motion.cc
   fallback.cc
   image-io.cc
   image.cc
@@ -47,6 +49,7 @@
   decctx.h
   dpb.h
   fallback-dct.h
+  fallback-intrapred.h
   fallback-motion.h
   fallback.h
   image-io.h
@@ -87,6 +90,13 @@
 
 add_definitions(-DLIBDE265_EXPORTS)
 
+# With hidden default visibility, LIBDE265_API must expand to
+# visibility("default") so the public C API stays exported. de265.h gates that
+# on HAVE_VISIBILITY (see the symbol-visibility block in the top CMakeLists).
+if(DE265_REDUCED_VISIBILITY)
+  add_definitions(-DHAVE_VISIBILITY)
+endif()
+
 if (ENABLE_ENCODER)
   add_subdirectory (encoder)
   list(APPEND libde265_sources en265.cc)
@@ -100,14 +110,18 @@
   endif()
 endif()
 
-if(HAVE_ARM)
-  add_subdirectory (arm)
+if(HAVE_ARM32)
+  add_subdirectory (arm32)
 endif()
 
-add_library(de265 ${libde265_sources} ${libde265_public_headers} ${ENCODER_OBJECTS} ${X86_OBJECTS} ${ARM_OBJECTS})
+add_library(de265 ${libde265_sources} ${libde265_public_headers} ${ENCODER_OBJECTS} ${X86_OBJECTS} ${ARM32_OBJECTS})
 target_include_directories(de265 PRIVATE ${CMAKE_BINARY_DIR} ${CMAKE_CURRENT_BINARY_DIR})
+target_link_libraries(de265 PRIVATE Threads::Threads)
+if(UNIX)
+  target_link_libraries(de265 PRIVATE m)
+endif()
 
-write_basic_package_version_file(libde265ConfigVersion.cmake COMPATIBILITY ExactVersion)
+write_basic_package_version_file(libde265-config-version.cmake COMPATIBILITY ExactVersion)
 
 # --- debug output
 
@@ -155,7 +169,7 @@
     MACOSX_RPATH TRUE)
 endif()
 
-install(TARGETS de265 EXPORT libde265Config
+install(TARGETS de265 EXPORT libde265-config
     RUNTIME DESTINATION ${CMAKE_INSTALL_BINDIR}
     LIBRARY DESTINATION ${CMAKE_INSTALL_LIBDIR}
     ARCHIVE DESTINATION ${CMAKE_INSTALL_LIBDIR}
@@ -165,9 +179,9 @@
 )
 
 install(FILES ${libde265_public_headers} DESTINATION ${CMAKE_INSTALL_INCLUDEDIR}/libde265)
-install(EXPORT libde265Config DESTINATION "${CMAKE_INSTALL_LIBDIR}/cmake/libde265")
+install(EXPORT libde265-config DESTINATION "${CMAKE_INSTALL_LIBDIR}/cmake/libde265")
 
-install(FILES ${CMAKE_CURRENT_BINARY_DIR}/libde265ConfigVersion.cmake DESTINATION
+install(FILES ${CMAKE_CURRENT_BINARY_DIR}/libde265-config-version.cmake DESTINATION
     "${CMAKE_INSTALL_LIBDIR}/cmake/libde265")
 
 
@@ -186,6 +200,5 @@
     set(includedir "\${prefix}/${CMAKE_INSTALL_INCLUDEDIR}")
 endif()
 
-set(VERSION ${PROJECT_VERSION}) # so that the replacement in libde265.pc will work with both autotools and CMake
 configure_file(../libde265.pc.in ${CMAKE_CURRENT_BINARY_DIR}/libde265.pc @ONLY)
 install(FILES ${CMAKE_CURRENT_BINARY_DIR}/libde265.pc DESTINATION ${CMAKE_INSTALL_LIBDIR}/pkgconfig)

libde265-1.0.17.tar.gz/libde265/acceleration.h -> libde265-1.1.1.tar.gz/libde265/acceleration.h Changed

@@ -172,6 +172,20 @@
   template <class pixel_t>
   void add_residual(pixel_t *dst, ptrdiff_t stride, const int32_t* r, int nT, int bit_depth) const;
 
+  // Inverse quantization (no scaling list): for each of the nCoeff entries,
+  // coeffBufcoeffPosi = Clip16( (coeffListi*fact + offset) >> bdShift ).
+  // Contract: fact small enough that coeffListi*fact+offset fits in int32
+  // (caller checks this; the rare int64 case stays scalar). bdShift >= 1.
+  void (*dequant_coeff_block)(int16_t* coeffBuf, const int16_t* coeffList,
+                              const int16_t* coeffPos, int nCoeff,
+                              int32_t fact, int32_t offset, int32_t bdShift);
+
+  // --- deblocking (8 bit; one 4-line edge segment) ---
+  void (*deblock_luma_8)(uint8_t* ptr, ptrdiff_t stride, int vertical,
+                         int dE, int dEp, int dEq, int tc, int filterP, int filterQ);
+  void (*deblock_chroma_8)(uint8_t* ptr, ptrdiff_t stride, int vertical,
+                           int tc, int filterP, int filterQ);
+
   void (*rdpcm_v)(int32_t* residual, const int16_t* coeffs, int nT,int tsShift,int bdShift);
   void (*rdpcm_h)(int32_t* residual, const int16_t* coeffs, int nT,int tsShift,int bdShift);
 
@@ -186,6 +200,22 @@
   template <class pixel_t> void transform_add(int sizeIdx, pixel_t *dst, const int16_t *coeffs, ptrdiff_t stride, int bit_depth) const;
 
 
+  // --- intra prediction ---
+
+  void (*intra_pred_dc_8 )(uint8_t*  dst, ptrdiff_t stride, int nT, int cIdx, const uint8_t*  border);
+  void (*intra_pred_dc_16)(uint16_t* dst, ptrdiff_t stride, int nT, int cIdx, const uint16_t* border);
+  void (*intra_pred_planar_8 )(uint8_t*  dst, ptrdiff_t stride, int nT, int cIdx, const uint8_t*  border);
+  void (*intra_pred_planar_16)(uint16_t* dst, ptrdiff_t stride, int nT, int cIdx, const uint16_t* border);
+  void (*intra_pred_angular_8 )(uint8_t*  dst, ptrdiff_t stride, int bit_depth, int disableBoundaryFilter,
+                                int xB0, int yB0, int mode, int nT, int cIdx, const uint8_t*  border);
+  void (*intra_pred_angular_16)(uint16_t* dst, ptrdiff_t stride, int bit_depth, int disableBoundaryFilter,
+                                int xB0, int yB0, int mode, int nT, int cIdx, const uint16_t* border);
+
+  template <class pixel_t> void intra_pred_dc(pixel_t* dst, ptrdiff_t stride, int nT, int cIdx, const pixel_t* border) const;
+  template <class pixel_t> void intra_pred_planar(pixel_t* dst, ptrdiff_t stride, int nT, int cIdx, const pixel_t* border) const;
+  template <class pixel_t> void intra_pred_angular(pixel_t* dst, ptrdiff_t stride, int bit_depth, int disableBoundaryFilter,
+                                                   int xB0, int yB0, int mode, int nT, int cIdx, const pixel_t* border) const;
+
 
   // --- forward transforms ---
 
@@ -356,4 +386,13 @@
 template <> inline void acceleration_functions::add_residual(uint8_t *dst,  ptrdiff_t stride, const int32_t* r, int nT, int bit_depth) const { add_residual_8(dst,stride,r,nT,bit_depth); }
 template <> inline void acceleration_functions::add_residual(uint16_t *dst, ptrdiff_t stride, const int32_t* r, int nT, int bit_depth) const { add_residual_16(dst,stride,r,nT,bit_depth); }
 
+template <> inline void acceleration_functions::intra_pred_dc<uint8_t> (uint8_t*  dst, ptrdiff_t stride, int nT, int cIdx, const uint8_t*  border) const { intra_pred_dc_8 (dst,stride,nT,cIdx,border); }
+template <> inline void acceleration_functions::intra_pred_dc<uint16_t>(uint16_t* dst, ptrdiff_t stride, int nT, int cIdx, const uint16_t* border) const { intra_pred_dc_16(dst,stride,nT,cIdx,border); }
+
+template <> inline void acceleration_functions::intra_pred_planar<uint8_t> (uint8_t*  dst, ptrdiff_t stride, int nT, int cIdx, const uint8_t*  border) const { intra_pred_planar_8 (dst,stride,nT,cIdx,border); }
+template <> inline void acceleration_functions::intra_pred_planar<uint16_t>(uint16_t* dst, ptrdiff_t stride, int nT, int cIdx, const uint16_t* border) const { intra_pred_planar_16(dst,stride,nT,cIdx,border); }
+
+template <> inline void acceleration_functions::intra_pred_angular<uint8_t> (uint8_t*  dst, ptrdiff_t stride, int bit_depth, int disableBoundaryFilter, int xB0, int yB0, int mode, int nT, int cIdx, const uint8_t*  border) const { intra_pred_angular_8 (dst,stride,bit_depth,disableBoundaryFilter,xB0,yB0,mode,nT,cIdx,border); }
+template <> inline void acceleration_functions::intra_pred_angular<uint16_t>(uint16_t* dst, ptrdiff_t stride, int bit_depth, int disableBoundaryFilter, int xB0, int yB0, int mode, int nT, int cIdx, const uint16_t* border) const { intra_pred_angular_16(dst,stride,bit_depth,disableBoundaryFilter,xB0,yB0,mode,nT,cIdx,border); }
+
 #endif

libde265-1.1.1.tar.gz/libde265/arm32 Added

libde265-1.1.1.tar.gz/libde265/arm32/CMakeLists.txt Added

libde265-1.1.1.tar.gz/libde265/arm32/arm.cc Changed

libde265-1.1.1.tar.gz/libde265/arm32/arm.h Changed

libde265-1.1.1.tar.gz/libde265/arm32/asm.S Changed

libde265-1.1.1.tar.gz/libde265/arm32/cpudetect.S Changed

libde265-1.1.1.tar.gz/libde265/arm32/hevcdsp_qpel_neon.S Changed

libde265-1.1.1.tar.gz/libde265/arm32/neon.S Changed

libde265-1.0.17.tar.gz/libde265/cabac.cc -> libde265-1.1.1.tar.gz/libde265/cabac.cc Changed

@@ -158,8 +158,101 @@
 }
 
 
+#if defined(__x86_64__) && defined(__GNUC__) && !defined(DE265_LOG_TRACE)
+#define DE265_CABAC_ASM_X86_64 1
+// Combined state-transition table (folds the MPS-flip at state 0), indexed by
+// the packed context byte (state*2+MPSbit) for MPS and (packed ^ 0xFF) for LPS.
+static const uint8_t cabac_transition256 =
+  {
+      2,  3,  4,  5,  6,  7,  8,  9, 10, 11, 12, 13, 14, 15, 16, 17,
+     18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33,
+     34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49,
+     50, 51, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, 64, 65,
+     66, 67, 68, 69, 70, 71, 72, 73, 74, 75, 76, 77, 78, 79, 80, 81,
+     82, 83, 84, 85, 86, 87, 88, 89, 90, 91, 92, 93, 94, 95, 96, 97,
+     98, 99,100,101,102,103,104,105,106,107,108,109,110,111,112,113,
+    114,115,116,117,118,119,120,121,122,123,124,125,124,125,126,127,
+    127,126, 77, 76, 77, 76, 75, 74, 75, 74, 75, 74, 73, 72, 73, 72,
+     73, 72, 71, 70, 71, 70, 71, 70, 69, 68, 69, 68, 67, 66, 67, 66,
+     67, 66, 65, 64, 65, 64, 63, 62, 61, 60, 61, 60, 61, 60, 59, 58,
+     59, 58, 57, 56, 55, 54, 55, 54, 53, 52, 53, 52, 51, 50, 49, 48,
+     49, 48, 47, 46, 45, 44, 45, 44, 43, 42, 43, 42, 39, 38, 39, 38,
+     37, 36, 37, 36, 33, 32, 33, 32, 31, 30, 31, 30, 27, 26, 27, 26,
+     25, 24, 23, 22, 23, 22, 19, 18, 19, 18, 17, 16, 15, 14, 13, 12,
+     11, 10,  9,  8,  9,  8,  5,  4,  5,  4,  3,  2,  1,  0,  0,  1,
+  };
+#endif
+
 int  CABAC_decoder::decode_bit(context_model* model)
 {
+#ifdef DE265_CABAC_ASM_X86_64
+  // x86-64 branchless arithmetic decoder. Bit-identical to the C fallback below.
+  uint32_t range_l = range, value_l = value;
+  int      bn_l    = bits_needed;
+  uint8_t* curr_l = bitstream_curr;
+  uint8_t* sp = reinterpret_cast<uint8_t*>(model);
+  int bit_l;
+
+  __asm__ (
+    "movzbl (%sp), %%eax            \n\t" // eax = b (state*2 + MPS)
+    "mov    %range, %%ecx           \n\t"
+    "shr    $6, %%ecx                 \n\t"
+    "sub    $4, %%ecx                 \n\t" // ecx = (range>>6)-4
+    "mov    %%eax, %%edx              \n\t"
+    "shr    $1, %%edx                 \n\t" // edx = state
+    "lea    (%%rcx,%%rdx,4), %%rcx    \n\t" // rcx = state*4 + ridx
+    "lea    %lps, %%rdx             \n\t" // rdx = &LPS_table00
+    "movzbl (%%rdx,%%rcx,1), %%edx    \n\t" // edx = RangeLPS
+    "sub    %%edx, %range           \n\t" // range = range - RangeLPS (MPS range)
+    "mov    %range, %%r8d           \n\t"
+    "shl    $7, %%r8d                 \n\t" // r8d = scaled_range
+    "mov    %value, %%r9d           \n\t"
+    "sub    %%r8d, %%r9d              \n\t"
+    "sar    $31, %%r9d                \n\t" // r9d = mps_mask (-1 if MPS)
+    "not    %%r9d                     \n\t" // r9d = lps_mask (-1 if LPS)
+    "mov    %%r8d, %%ecx              \n\t"
+    "and    %%r9d, %%ecx              \n\t"
+    "sub    %%ecx, %value           \n\t" // value -= scaled & lps_mask
+    "mov    %%edx, %%ecx              \n\t" // RangeLPS
+    "sub    %range, %%ecx           \n\t"
+    "and    %%r9d, %%ecx              \n\t"
+    "add    %%ecx, %range           \n\t" // range = RangeLPS if LPS, else unchanged
+    "and    $0xFF, %%r9d              \n\t"
+    "xor    %%r9d, %%eax              \n\t" // eax = idx
+    "mov    %%eax, %bit             \n\t"
+    "and    $1, %bit                \n\t" // bit = idx & 1
+    "lea    %trans, %%rcx           \n\t" // rcx = &cabac_transition0
+    "movzbl (%%rcx,%%rax,1), %%edx    \n\t"
+    "mov    %%dl, (%sp)             \n\t" // *sp = cabac_transitionidx
+    "bsr    %range, %%ecx           \n\t" // ecx = MSB index
+    "mov    $8, %%edx                 \n\t"
+    "sub    %%ecx, %%edx              \n\t" // edx = renorm shift = 8 - bsr
+    "mov    %%edx, %%ecx              \n\t"
+    "shl    %%cl, %value            \n\t"
+    "shl    %%cl, %range            \n\t"
+    "add    %%edx, %bn              \n\t" // bits_needed += shift
+    "test   %bn, %bn              \n\t"
+    "js     1f                        \n\t" // bits_needed < 0: no refill
+    "cmp    %end, %curr           \n\t"
+    "jae    2f                        \n\t" // curr >= end: no read
+    "movzbl (%curr), %%eax          \n\t"
+    "mov    %bn, %%ecx              \n\t"
+    "shl    %%cl, %%eax              \n\t"
+    "or     %%eax, %value           \n\t" // value |= (*curr) << bits_needed
+    "inc    %curr                   \n\t"
+    "2:                               \n\t"
+    "sub    $8, %bn                 \n\t"
+    "1:                               \n\t"
+    : range"+r"(range_l), value"+r"(value_l), bn"+r"(bn_l),
+      curr"+r"(curr_l), bit"=&r"(bit_l)
+    : sp"r"(sp), end"r"(bitstream_end),
+      lps"m"(LPS_table00), trans"m"(cabac_transition0)
+    : "rax","rcx","rdx","r8","r9","cc","memory"
+  );
+
+  range = range_l; value = value_l; bits_needed = (int16_t)bn_l; bitstream_curr = curr_l;
+  return bit_l;
+#else
   logtrace(LogCABAC,"%3d decodeBin r:%x v:%x state:%d\n",logcnt,range, value, model->state);
 
   int decoded_bit;
@@ -237,6 +330,7 @@
 #endif
 
   return decoded_bit;
+#endif
 }
 
 int  CABAC_decoder::decode_term_bit()
@@ -279,6 +373,43 @@
 // we will eventually only return zeros.
 int  CABAC_decoder::decode_bypass()
 {
+#ifdef DE265_CABAC_ASM_X86_64
+  uint32_t value_l = value;
+  int      bn_l    = bits_needed;
+  uint8_t* curr_l  = bitstream_curr;
+  int bit_l;
+
+  __asm__ (
+    "add    %value, %value        \n\t" // value <<= 1
+    "addl   $1, %bn                 \n\t" // bits_needed++
+    "js     1f                        \n\t" // bits_needed < 0: no refill
+    "cmp    %end, %curr           \n\t"
+    "jae    2f                        \n\t" // curr >= end: no read
+    "movzbl (%curr), %%eax          \n\t"
+    "or     %%eax, %value           \n\t" // value |= *curr (bits_needed==0)
+    "inc    %curr                   \n\t"
+    "2:                               \n\t"
+    "movl   $-8, %bn                \n\t" // bits_needed = -8
+    "1:                               \n\t"
+    "mov    %range, %%ecx           \n\t"
+    "shl    $7, %%ecx                 \n\t" // ecx = scaled_range
+    "mov    %value, %%edx           \n\t"
+    "sub    %%ecx, %%edx              \n\t" // edx = value - scaled_range
+    "mov    %%edx, %%eax              \n\t"
+    "sar    $31, %%eax               \n\t"  // eax = mask (-1 if value<scaled -> bit 0)
+    "mov    %%eax, %bit             \n\t"
+    "add    $1, %bit                \n\t" // bit = mask + 1
+    "not    %%eax                     \n\t" // eax = ~mask
+    "and    %%eax, %%ecx              \n\t" // ecx = scaled & ~mask
+    "sub    %%ecx, %value           \n\t" // value -= scaled when bit==1
+    : value"+r"(value_l), bn"+r"(bn_l), curr"+r"(curr_l), bit"=&r"(bit_l)
+    : range"r"(range), end"r"(bitstream_end)
+    : "rax","rcx","rdx","cc","memory"
+  );
+
+  value = value_l; bits_needed = (int16_t)bn_l; bitstream_curr = curr_l;
+  return bit_l;
+#else
   logtrace(LogCABAC,"%3d bypass r:%x v:%x\n",logcnt,range, value);
 
   value <<= 1;
@@ -314,6 +445,7 @@
 #endif
 
   return bit;
+#endif
 }

libde265-1.0.17.tar.gz/libde265/de265.cc -> libde265-1.1.1.tar.gz/libde265/de265.cc Changed

@@ -82,6 +82,8 @@
   case DE265_ERROR_OUT_OF_MEMORY: return "out of memory";
   case DE265_ERROR_CODED_PARAMETER_OUT_OF_RANGE: return "coded parameter out of range";
   case DE265_ERROR_IMAGE_BUFFER_FULL: return "DPB/output queue full";
+  case DE265_ERROR_IMAGE_SIZE_EXCEEDS_SECURITY_LIMIT: return "image size exceeds security limit";
+  case DE265_ERROR_NAL_SIZE_EXCEEDS_SECURITY_LIMIT: return "NAL unit size exceeds security limit";
   case DE265_ERROR_CANNOT_START_THREADPOOL: return "cannot start decoding threads";
   case DE265_ERROR_LIBRARY_INITIALIZATION_FAILED: return "global library initialization failed";
   case DE265_ERROR_LIBRARY_NOT_INITIALIZED: return "cannot free library data (not initialized";
@@ -178,6 +180,10 @@
     return "Access with invalid slice header index";
   case DE265_WARNING_INVALID_TU_BLOCK_SPLIT:
     return "Transform block split below minimum transform size";
+  case DE265_WARNING_RICE_PARAMETER_OUT_OF_RANGE:
+    return "Rice parameter or StatCoeff out of range, clamped";
+  case DE265_WARNING_MAX_NUMBER_OF_SEI_MESSAGES_EXCEEDED:
+    return "number of SEI messages exceeds security limit, dropped";
 
   default: return "unknown error";
   }
@@ -215,8 +221,10 @@
   // do initializations
 
   init_scan_orders();
+  pps_scan_cache_init();
 
   if (!alloc_and_init_significant_coeff_ctxIdx_lookupTable()) {
+    pps_scan_cache_free();
     de265_init_count--;
     return DE265_ERROR_LIBRARY_INITIALIZATION_FAILED;
   }
@@ -236,6 +244,7 @@
 
   if (de265_init_count==0) {
     free_significant_coeff_ctxIdx_lookupTable();
+    pps_scan_cache_free();
   }
 
   return DE265_OK;
@@ -610,6 +619,53 @@
 }
 
 
+static const de265_security_limits disabled_security_limits = {
+  1,    // version
+  0,    // max_image_size_pixels
+  0,    // max_NAL_size_bytes
+  0     // max_SEI_messages
+};
+
+
+LIBDE265_API de265_security_limits* de265_get_security_limits(de265_decoder_context* de265ctx)
+{
+  decoder_context* ctx = reinterpret_cast<decoder_context*>(de265ctx);
+  return &ctx->param_security_limits;
+}
+
+
+// Copy security limits field by field, but only those fields that exist in both
+// structs, i.e. up to min(dst->version, src->version). This keeps copying safe
+// when 'src' was compiled against a different (older or newer) header version
+// than 'dst'. Fields not covered by the common version keep their value in 'dst'.
+static void copy_security_limits(de265_security_limits* dst, const de265_security_limits* src)
+{
+  uint8_t version = (dst->version < src->version) ? dst->version : src->version;
+
+  if (version >= 1) {
+    dst->max_image_size_pixels = src->max_image_size_pixels;
+    dst->max_NAL_size_bytes    = src->max_NAL_size_bytes;
+    dst->max_SEI_messages      = src->max_SEI_messages;
+  }
+}
+
+
+LIBDE265_API void de265_set_security_limits(de265_decoder_context* de265ctx,
+                                            const de265_security_limits* limits)
+{
+  decoder_context* ctx = reinterpret_cast<decoder_context*>(de265ctx);
+  if (limits) {
+    copy_security_limits(&ctx->param_security_limits, limits);
+  }
+}
+
+
+LIBDE265_API const de265_security_limits* de265_get_disabled_security_limits()
+{
+  return &disabled_security_limits;
+}
+
+
 LIBDE265_API int de265_get_number_of_input_bytes_pending(de265_decoder_context* de265ctx)
 {
   decoder_context* ctx = reinterpret_cast<decoder_context*>(de265ctx);
@@ -676,7 +732,7 @@
 
   uint8_t* data = img->pixels_confwinchannel;
 
-  if (stride) *stride = img->get_image_stride(channel) * ((de265_get_bits_per_pixel(img, channel)+7) / 8);
+  if (stride) *stride = static_cast<int>(img->get_image_stride(channel) * ((de265_get_bits_per_pixel(img, channel)+7) / 8));
 
   return data;
 }

libde265-1.0.17.tar.gz/libde265/de265.h -> libde265-1.1.1.tar.gz/libde265/de265.h Changed

@@ -28,8 +28,6 @@
 
 #include <libde265/de265-version.h>
 
-//#define inline static __inline
-
 
 #ifndef __STDC_LIMIT_MACROS
 #define __STDC_LIMIT_MACROS 1
@@ -60,12 +58,6 @@
 #define LIBDE265_DEPRECATED
 #endif
 
-#if defined(_MSC_VER)
-#define LIBDE265_INLINE __inline
-#else
-#define LIBDE265_INLINE inline
-#endif
-
 /* === version numbers === */
 
 // version of linked libde265 library
@@ -103,6 +95,8 @@
   DE265_ERROR_NO_INITIAL_SLICE_HEADER=16,
   DE265_ERROR_PREMATURE_END_OF_SLICE=17,
   DE265_ERROR_UNSPECIFIED_DECODING_ERROR=18,
+  DE265_ERROR_IMAGE_SIZE_EXCEEDS_SECURITY_LIMIT=19,
+  DE265_ERROR_NAL_SIZE_EXCEEDS_SECURITY_LIMIT=20,
 
   // --- errors that should become obsolete in later libde265 versions ---
 
@@ -147,7 +141,9 @@
   DE265_WARNING_BIT_DEPTH_OF_CURRENT_IMAGE_DOES_NOT_MATCH_SPS=1031,
   DE265_WARNING_REFERENCE_IMAGE_CHROMA_FORMAT_DOES_NOT_MATCH=1032,
   DE265_WARNING_INVALID_SLICE_HEADER_INDEX_ACCESS=1033,
-  DE265_WARNING_INVALID_TU_BLOCK_SPLIT=1034
+  DE265_WARNING_INVALID_TU_BLOCK_SPLIT=1034,
+  DE265_WARNING_RICE_PARAMETER_OUT_OF_RANGE=1035,
+  DE265_WARNING_MAX_NUMBER_OF_SEI_MESSAGES_EXCEEDED=1036
 } de265_error;
 
 LIBDE265_API const char* de265_get_error_text(de265_error err);
@@ -209,6 +205,17 @@
 typedef void de265_decoder_context; // private structure
 
 
+/* Thread-safety:
+   A de265_decoder_context must not be accessed concurrently from multiple
+   threads. All API calls that take a de265_decoder_context (push data, decode,
+   query state, retrieve images, free) must be serialized by the caller.
+   To decode multiple streams in parallel, create one context per thread.
+
+   This is independent from de265_start_worker_threads(), which only enables
+   internal worker threads inside a single context to parallelize WPP/tile
+   decoding. Those internal threads are managed by libde265 and do not relax
+   the single-owner-thread requirement above.
+*/
 
 /* Get a new decoder context. Must be freed with de265_free_decoder(). */
 LIBDE265_API de265_decoder_context* de265_new_decoder(void);
@@ -428,6 +435,25 @@
 LIBDE265_API int  de265_get_parameter_bool(de265_decoder_context*, de265_param param);
 
 
+/* --- security limits --- */
+
+typedef struct de265_security_limits {
+  uint8_t version;
+
+  // --- version 1 ---
+
+  uint32_t max_image_size_pixels;
+  uint32_t max_NAL_size_bytes;
+  uint32_t max_SEI_messages;   // max number of SEI messages per access unit (0 = unlimited)
+
+} de265_security_limits;
+
+LIBDE265_API de265_security_limits* de265_get_security_limits(de265_decoder_context*);
+
+LIBDE265_API void de265_set_security_limits(de265_decoder_context*, const de265_security_limits* limits);
+
+LIBDE265_API const de265_security_limits* de265_get_disabled_security_limits();
+
 
 /* --- optional library initialization --- */

libde265-1.0.17.tar.gz/libde265/deblock.cc -> libde265-1.1.1.tar.gz/libde265/deblock.cc Changed

@@ -22,6 +22,8 @@
 #include "util.h"
 #include "transform.h"
 #include "de265.h"
+#include "decctx.h"
+#include "fallback-deblk.h"
 
 #include <assert.h>
 
@@ -187,8 +189,8 @@
               filterLeftCbEdge = 0;
             }
           else if (pps.loop_filter_across_tiles_enabled_flag == 0 &&
-                   pps.TileIdRS  x0ctb           +y0ctb*picWidthInCtbs !=
-                   pps.TileIdRS((x0-1)>>ctbshift)+y0ctb*picWidthInCtbs) {
+                   pps.scan->TileIdRS  x0ctb           +y0ctb*picWidthInCtbs !=
+                   pps.scan->TileIdRS((x0-1)>>ctbshift)+y0ctb*picWidthInCtbs) {
             filterLeftCbEdge = 0;
           }
         }
@@ -201,8 +203,8 @@
               filterTopCbEdge = 0;
             }
           else if (pps.loop_filter_across_tiles_enabled_flag == 0 &&
-                   pps.TileIdRSx0ctb+  y0ctb           *picWidthInCtbs !=
-                   pps.TileIdRSx0ctb+((y0-1)>>ctbshift)*picWidthInCtbs) {
+                   pps.scan->TileIdRSx0ctb+  y0ctb           *picWidthInCtbs !=
+                   pps.scan->TileIdRSx0ctb+((y0-1)>>ctbshift)*picWidthInCtbs) {
             filterTopCbEdge = 0;
           }
         }
@@ -250,8 +252,8 @@
     (DEBLOCK_FLAG_HORIZ | DEBLOCK_PB_EDGE_HORIZ);
   int transformEdgeMask = vertical ? DEBLOCK_FLAG_VERTI : DEBLOCK_FLAG_HORIZ;
 
-  xEnd = libde265_min(xEnd,img->get_deblk_width());
-  yEnd = libde265_min(yEnd,img->get_deblk_height());
+  xEnd = std::min(xEnd,img->get_deblk_width());
+  yEnd = std::min(yEnd,img->get_deblk_height());
 
   //int TUShift = img->get_sps().Log2MinTrafoSize;
   //int TUStride= img->get_sps().PicWidthInTbsY;
@@ -326,18 +328,18 @@
 		if (refPicP0 != refPicP1) {
 
 		  if (refPicP0 == refPicQ0) {
-		    if (abs_value(mvP0.x-mvQ0.x) >= 4 ||
-			abs_value(mvP0.y-mvQ0.y) >= 4 ||
-			abs_value(mvP1.x-mvQ1.x) >= 4 ||
-			abs_value(mvP1.y-mvQ1.y) >= 4) {
+		    if (std::abs(mvP0.x-mvQ0.x) >= 4 ||
+			std::abs(mvP0.y-mvQ0.y) >= 4 ||
+			std::abs(mvP1.x-mvQ1.x) >= 4 ||
+			std::abs(mvP1.y-mvQ1.y) >= 4) {
 		      bS = 1;
 		    }
 		  }
 		  else {
-		    if (abs_value(mvP0.x-mvQ1.x) >= 4 ||
-			abs_value(mvP0.y-mvQ1.y) >= 4 ||
-			abs_value(mvP1.x-mvQ0.x) >= 4 ||
-			abs_value(mvP1.y-mvQ0.y) >= 4) {
+		    if (std::abs(mvP0.x-mvQ1.x) >= 4 ||
+			std::abs(mvP0.y-mvQ1.y) >= 4 ||
+			std::abs(mvP1.x-mvQ0.x) >= 4 ||
+			std::abs(mvP1.y-mvQ0.y) >= 4) {
 		      bS = 1;
 		    }
 		  }
@@ -345,15 +347,15 @@
 		else {
 		  assert(refPicQ0==refPicQ1);
 
-		  if ((abs_value(mvP0.x-mvQ0.x) >= 4 ||
-		       abs_value(mvP0.y-mvQ0.y) >= 4 ||
-		       abs_value(mvP1.x-mvQ1.x) >= 4 ||
-		       abs_value(mvP1.y-mvQ1.y) >= 4)
+		  if ((std::abs(mvP0.x-mvQ0.x) >= 4 ||
+		       std::abs(mvP0.y-mvQ0.y) >= 4 ||
+		       std::abs(mvP1.x-mvQ1.x) >= 4 ||
+		       std::abs(mvP1.y-mvQ1.y) >= 4)
 		      &&
-		      (abs_value(mvP0.x-mvQ1.x) >= 4 ||
-		       abs_value(mvP0.y-mvQ1.y) >= 4 ||
-		       abs_value(mvP1.x-mvQ0.x) >= 4 ||
-		       abs_value(mvP1.y-mvQ0.y) >= 4)) {
+		      (std::abs(mvP0.x-mvQ1.x) >= 4 ||
+		       std::abs(mvP0.y-mvQ1.y) >= 4 ||
+		       std::abs(mvP1.x-mvQ0.x) >= 4 ||
+		       std::abs(mvP1.y-mvQ0.y) >= 4)) {
 		    bS = 1;
 		  }
 		}
@@ -422,8 +424,8 @@
 
   int bitDepth_Y = sps.BitDepth_Y;
 
-  xEnd = libde265_min(xEnd,img->get_deblk_width());
-  yEnd = libde265_min(yEnd,img->get_deblk_height());
+  xEnd = std::min(xEnd,img->get_deblk_width());
+  yEnd = std::min(yEnd,img->get_deblk_height());
 
   for (int y=yStart;y<yEnd;y+=yIncr)
     for (int x=xStart;x<xEnd;x+=xIncr) {
@@ -532,10 +534,10 @@
 
         int dE=0, dEp=0, dEq=0;
 
-        int dp0 = abs_value(p02 - 2*p01 + p00);
-        int dp3 = abs_value(p32 - 2*p31 + p30);
-        int dq0 = abs_value(q02 - 2*q01 + q00);
-        int dq3 = abs_value(q32 - 2*q31 + q30);
+        int dp0 = std::abs(p02 - 2*p01 + p00);
+        int dp3 = std::abs(p32 - 2*p31 + p30);
+        int dq0 = std::abs(q02 - 2*q01 + q00);
+        int dq3 = std::abs(q32 - 2*q31 + q30);
 
         int dpq0 = dp0 + dq0;
         int dpq3 = dp3 + dq3;
@@ -547,12 +549,12 @@
         if (d < beta) {
           //int dpq = 2*dpq0;
           bool dSam0 = (2 * dpq0 < (beta >> 2) &&
-                        abs_value(p03-p00) + abs_value(q00-q03) < (beta >> 3) &&
-                        abs_value(p00-q00) < ((5 * tc + 1) >> 1));
+                        std::abs(p03-p00) + std::abs(q00-q03) < (beta >> 3) &&
+                        std::abs(p00-q00) < ((5 * tc + 1) >> 1));
 
           bool dSam3 = (2 * dpq3 < (beta >> 2) &&
-                        abs_value(p33-p30) + abs_value(q30-q33) < (beta >> 3) &&
-                        abs_value(p30-q30) < ((5 * tc + 1) >> 1));
+                        std::abs(p33-p30) + std::abs(q30-q33) < (beta >> 3) &&
+                        std::abs(p30-q30) < ((5 * tc + 1) >> 1));
 
           if (dSam0 && dSam3) {
             dE = 2;
@@ -589,109 +591,13 @@
             if (img->get_cu_transquant_bypass(xDi,yDi)) filterQ=false;
           }
 
-          for (int k=0;k<4;k++) {
-            //int nDp,nDq;
-
-            logtrace(LogDeblock,"line:%d\n",k);
-
-            const pixel_t p0 = pk0;
-            const pixel_t p1 = pk1;
-            const pixel_t p2 = pk2;
-            const pixel_t p3 = pk3;
-            const pixel_t q0 = qk0;
-            const pixel_t q1 = qk1;
-            const pixel_t q2 = qk2;
-            const pixel_t q3 = qk3;
-
-            if (dE==2) {
-              // strong filtering
-
-              //nDp=nDq=3;
-
-              pixel_t pnew3,qnew3;
-              pnew0 = Clip3(p0-2*tc,p0+2*tc, (p2 + 2*p1 + 2*p0 + 2*q0 + q1 +4)>>3);
-              pnew1 = Clip3(p1-2*tc,p1+2*tc, (p2 + p1 + p0 + q0+2)>>2);
-              pnew2 = Clip3(p2-2*tc,p2+2*tc, (2*p3 + 3*p2 + p1 + p0 + q0 + 4)>>3);
-              qnew0 = Clip3(q0-2*tc,q0+2*tc, (p1+2*p0+2*q0+2*q1+q2+4)>>3);
-              qnew1 = Clip3(q1-2*tc,q1+2*tc, (p0+q0+q1+q2+2)>>2);
-              qnew2 = Clip3(q2-2*tc,q2+2*tc, (p0+q0+q1+3*q2+2*q3+4)>>3);
-
-              logtrace(LogDeblock,"strong filtering\n");
-
-              if (vertical) {
-                for (int i=0;i<3;i++) {
-                  if (filterP) { ptr-i-1+k*stride = pnewi; }
-                  if (filterQ) { ptr i + k*stride = qnewi; }
-                }
-
-                // ptr-1+k*stride = ptr 0+k*stride = 200;
-              }
-              else {
-                for (int i=0;i<3;i++) {
-                  if (filterP) { ptr k -(i+1)*stride = pnewi; }
-                  if (filterQ) { ptr k + i   *stride = qnewi; }
-                }
-              }
-            }
-            else {
-              // weak filtering
-
-              //nDp=nDq=0;
-
-              int delta = (9*(q0-p0) - 3*(q1-p1) + 8)>>4;
-              logtrace(LogDeblock,"delta=%d, tc=%d\n",delta,tc);
-
-              if (abs_value(delta) < tc*10) {
-
-                delta = Clip3(-tc,tc,delta);
-                logtrace(LogDeblock," deblk + %d;%d %02x->%02x  - %d;%d %02x->%02x delta:%d\n",
-                         vertical ? xDi-1 : xDi+k,
-                         vertical ? yDi+k : yDi-1, p0,Clip_BitDepth(p0+delta, bitDepth_Y),
-                         vertical ? xDi   : xDi+k,
-                         vertical ? yDi+k : yDi,   q0,Clip_BitDepth(q0-delta, bitDepth_Y),
-                         delta);
-
-                if (vertical) {
-                  if (filterP) { ptr-0-1+k*stride = Clip_BitDepth(p0+delta, bitDepth_Y); }
-                  if (filterQ) { ptr 0  +k*stride = Clip_BitDepth(q0-delta, bitDepth_Y); }
-                }
-                else {
-                  if (filterP) { ptr k -1*stride = Clip_BitDepth(p0+delta, bitDepth_Y); }
-                  if (filterQ) { ptr k +0*stride = Clip_BitDepth(q0-delta, bitDepth_Y); }
-                }
-
-                //ptr 0+k*stride = 200;
-
-                if (dEp==1 && filterP) {
-                  int delta_p = Clip3(-(tc>>1), tc>>1, (((p2+p0+1)>>1)-p1+delta)>>1);
-
-                  logtrace(LogDeblock," deblk dEp %d;%d delta:%d\n",
-                           vertical ? xDi-2 : xDi+k,
-                           vertical ? yDi+k : yDi-2,
-                           delta_p);
-
-                  if (vertical) { ptr-1-1+k*stride = Clip_BitDepth(p1+delta_p, bitDepth_Y); }
-                  else          { ptr k  -2*stride = Clip_BitDepth(p1+delta_p, bitDepth_Y); }
-                }
-
-                if (dEq==1 && filterQ) {
-                  int delta_q = Clip3(-(tc>>1), tc>>1, (((q2+q0+1)>>1)-q1-delta)>>1);
-
-                  logtrace(LogDeblock," delkb dEq %d;%d delta:%d\n",
-                           vertical ? xDi+1 : xDi+k,
-                           vertical ? yDi+k : yDi+1,
-                           delta_q);
-
-                  if (vertical) { ptr 1  +k*stride = Clip_BitDepth(q1+delta_q, bitDepth_Y); }
-                  else          { ptr k  +1*stride = Clip_BitDepth(q1+delta_q, bitDepth_Y); }
-                }
-
-                //nDp = dEp+1;
-                //nDq = dEq+1;
-
-                //logtrace(LogDeblock,"weak filtering (%d:%d)\n",nDp,nDq);
-              }
-            }
+          if constexpr (sizeof(pixel_t)==1) {
+            img->decctx->acceleration.deblock_luma_8((uint8_t*)ptr, stride, vertical,
+                                                     dE,dEp,dEq,tc, filterP,filterQ);
+          }
+          else {
+            deblock_luma_kernel<pixel_t>(ptr, stride, vertical,
+                                         dE,dEp,dEq,tc, filterP,filterQ, bitDepth_Y);
           }
         }
       }
@@ -746,8 +652,8 @@
 
   const int stride = img->get_image_stride(1);
 
-  xEnd = libde265_min(xEnd,img->get_deblk_width());
-  yEnd = libde265_min(yEnd,img->get_deblk_height());
+  xEnd = std::min(xEnd,img->get_deblk_width());
+  yEnd = std::min(yEnd,img->get_deblk_height());
 
   int bitDepth_C = sps.BitDepth_C;
 
@@ -770,23 +676,6 @@
 
           pixel_t* ptr = img->get_image_plane_at_pos_NEW<pixel_t>(cplane+1, xDi,yDi);
 
-          pixel_t p24;
-          pixel_t q24;
-
-          logtrace(LogDeblock,"-%s- %d %d\n",cplane==0 ? "Cb" : "Cr",xDi,yDi);
-
-          for (int i=0;i<2;i++)
-            for (int k=0;k<4;k++)
-              {
-                if (vertical) {
-                  qik = ptr i  +k*stride;
-                  pik = ptr-i-1+k*stride;
-                }
-                else {
-                  qik = ptrk + i   *stride;
-                  pik = ptrk -(i+1)*stride;
-                }
-              }
 
 #if 0
           for (int k=0;k<4;k++)
@@ -815,7 +704,7 @@
           if (sps.ChromaArrayType == CHROMA_420) {
             QP_C = table8_22(qP_i);
           } else {
-            QP_C = libde265_min(qP_i, 51);
+            QP_C = std::min(qP_i, 51);
           }
 
 
@@ -843,11 +732,11 @@
             if (img->get_cu_transquant_bypass(SubWidthC*xDi,SubHeightC*yDi)) filterQ=false;
 
 
-            for (int k=0;k<4;k++) {
-              int delta = Clip3(-tc,tc, ((((q0k-p0k)*4)+p1k-q1k+4)>>3)); // standard says <<2 in eq. (8-356), but the value can also be negative
-              logtrace(LogDeblock,"delta=%d\n",delta);
-              if (filterP) { ptr-1+k*stride = Clip_BitDepth(p0k+delta, bitDepth_C); }
-              if (filterQ) { ptr 0+k*stride = Clip_BitDepth(q0k-delta, bitDepth_C); }
+            if constexpr (sizeof(pixel_t)==1) {
+              img->decctx->acceleration.deblock_chroma_8((uint8_t*)ptr, stride, 1, tc, filterP,filterQ);
+            }
+            else {
+              deblock_chroma_kernel<pixel_t>(ptr, stride, true, tc, filterP,filterQ, bitDepth_C);
             }
           }
           else {
@@ -859,10 +748,11 @@
             if (sps.pcm_loop_filter_disable_flag && img->get_pcm_flag(SubWidthC*xDi,SubHeightC*yDi)) filterQ=false;
             if (img->get_cu_transquant_bypass(SubWidthC*xDi,SubHeightC*yDi)) filterQ=false;
 
-            for (int k=0;k<4;k++) {
-              int delta = Clip3(-tc,tc, ((((q0k-p0k)*4)+p1k-q1k+4)>>3)); // standard says <<2, but the value can also be negative
-              if (filterP) { ptr k-1*stride = Clip_BitDepth(p0k+delta, bitDepth_C); }
-              if (filterQ) { ptr k+0*stride = Clip_BitDepth(q0k-delta, bitDepth_C); }
+            if constexpr (sizeof(pixel_t)==1) {
+              img->decctx->acceleration.deblock_chroma_8((uint8_t*)ptr, stride, 0, tc, filterP,filterQ);
+            }
+            else {
+              deblock_chroma_kernel<pixel_t>(ptr, stride, false, tc, filterP,filterQ, bitDepth_C);
             }
           }
         }

libde265-1.0.17.tar.gz/libde265/decctx.cc -> libde265-1.1.1.tar.gz/libde265/decctx.cc Changed

@@ -41,8 +41,8 @@
 #include "x86/sse.h"
 #endif
 
-#ifdef HAVE_ARM
-#include "arm/arm.h"
+#ifdef HAVE_ARM32
+#include "arm32/arm.h"
 #endif
 
 #define SAVE_INTERMEDIATE_IMAGES 0
@@ -133,6 +133,7 @@
 decoder_context::decoder_context()
 {
   param_image_allocation_functions = de265_image::default_image_allocation;
+  nal_parser.set_security_limits(&param_security_limits);
   compute_framedrop_table();
 }
 
@@ -249,7 +250,19 @@
     init_acceleration_functions_sse(&acceleration);
   }
 #endif
-#ifdef HAVE_ARM
+#if HAVE_AVX2
+  // layered on top of SSE: overrides a few transform kernels (runtime-checked)
+  if (l>=de265_acceleration_AVX2) {
+    init_acceleration_functions_avx2(&acceleration);
+  }
+#endif
+#if HAVE_AVX512
+  // layered on top of AVX2: overrides the 32x32 transform (runtime-checked)
+  if (l>=de265_acceleration_AVX2) {
+    init_acceleration_functions_avx512(&acceleration);
+  }
+#endif
+#ifdef HAVE_ARM32
   if (l>=de265_acceleration_ARM) {
     init_acceleration_functions_arm(&acceleration);
   }
@@ -276,7 +289,7 @@
 
 
   if (tctx->shdr->slice_segment_address > 0) {
-    int prevCtb = pps.CtbAddrTStoRS pps.CtbAddrRStoTStctx->shdr->slice_segment_address -1 ;
+    int prevCtb = pps.scan->CtbAddrTStoRS pps.scan->CtbAddrRStoTStctx->shdr->slice_segment_address -1 ;
 
     int ctbX = prevCtb % sps.PicWidthInCtbsY;
     int ctbY = prevCtb / sps.PicWidthInCtbsY;
@@ -301,7 +314,7 @@
 
 void decoder_context::add_task_decode_CTB_row(thread_context* tctx,
                                               bool firstSliceSubstream,
-                                              int ctbRow)
+                                              uint16_t ctbRow)
 {
   thread_task_ctb_row* task = new thread_task_ctb_row;
   task->firstSliceSubstream = firstSliceSubstream;
@@ -316,7 +329,7 @@
 
 
 void decoder_context::add_task_decode_slice_segment(thread_context* tctx, bool firstSliceSubstream,
-                                                    int ctbx,int ctby)
+                                                    uint16_t ctbx, uint16_t ctby)
 {
   thread_task_slice_segment* task = new thread_task_slice_segment;
   task->firstSliceSubstream = firstSliceSubstream;
@@ -413,6 +426,14 @@
     dump_sei(&sei, current_sps.get());
 
     if (image_units.empty()==false && suffix) {
+      uint32_t max_SEI_messages = param_security_limits.max_SEI_messages;
+      if (max_SEI_messages != 0 &&
+          image_units.back()->suffix_SEIs.size() >= max_SEI_messages) {
+        // too many SEI messages for this access unit -> drop to bound memory usage
+        add_warning(DE265_WARNING_MAX_NUMBER_OF_SEI_MESSAGES_EXCEEDED, false);
+        return DE265_WARNING_MAX_NUMBER_OF_SEI_MESSAGES_EXCEEDED;
+      }
+
       image_units.back()->suffix_SEIs.push_back(sei);
     }
   }
@@ -466,7 +487,7 @@
   // modify entry_point_offsets
 
   uint32_t headerLength = reader.data - nal->data();
-  for (int i=0;i<shdr->num_entry_point_offsets;i++) {
+  for (uint32_t i=0;i<shdr->num_entry_point_offsets;i++) {
     uint32_t skipped = nal->num_skipped_bytes_before(shdr->entry_point_offseti,
                                                      headerLength);
     if (skipped > shdr->entry_point_offseti) {
@@ -478,16 +499,18 @@
     shdr->entry_point_offseti -= skipped;
   }
 
-  this->img->add_slice_segment_header(shdr);
-
-
-
   // --- start a new image if this is the first slice ---
 
   if (shdr->first_slice_segment_in_pic_flag) {
     image_unit* imgunit = new image_unit;
     imgunit->img = this->img;
     image_units.push_back(imgunit);
+
+    // A new picture starts here. Drop the reference to the previous picture's
+    // slice header, whose storage may be released independently of this decoder
+    // state. Dependent slices only ever reference a preceding slice header
+    // within the same picture, which is set below as slices are retained.
+    previous_slice_header = nullptr;
   }
 
 
@@ -495,6 +518,18 @@
 
   if ( ! image_units.empty() ) {
 
+    // Hand the slice header to the picture (which takes ownership and frees it
+    // on release). Only do this when there is an active image unit to decode
+    // the slice; otherwise the header would be retained on img->slices forever,
+    // which a crafted stream of non-first slice NALs can exploit to grow memory
+    // without bound.
+    this->img->add_slice_segment_header(shdr);
+
+    // The header is now owned by the image and stays alive at least until the
+    // image is released, so it is safe for a following dependent slice to copy
+    // from it. Only retained headers may become 'previous_slice_header'.
+    previous_slice_header = shdr;
+
     slice_unit* sliceunit = new slice_unit(this);
     sliceunit->nal = nal;
     sliceunit->shdr = shdr;
@@ -507,6 +542,7 @@
   }
   else {
     nal_parser.free_NAL_unit(nal);
+    delete shdr;
   }
 
   bool did_work;
@@ -631,7 +667,7 @@
 
   remove_images_from_dpb(sliceunit->shdr->RemoveReferencesList);
 
-  if (sliceunit->shdr->slice_segment_address >= imgunit->img->get_pps().CtbAddrRStoTS.size()) {
+  if (sliceunit->shdr->slice_segment_address >= imgunit->img->get_pps().scan->CtbAddrRStoTS.size()) {
     return DE265_ERROR_CTB_OUTSIDE_IMAGE_AREA;
   }
 
@@ -643,7 +679,7 @@
   tctx.decctx = this;
   tctx.imgunit = imgunit;
   tctx.sliceunit= sliceunit;
-  tctx.CtbAddrInTS = imgunit->img->get_pps().CtbAddrRStoTStctx.shdr->slice_segment_address;
+  tctx.CtbAddrInTS = imgunit->img->get_pps().scan->CtbAddrRStoTStctx.shdr->slice_segment_address;
   tctx.task = nullptr;
 
   init_thread_context(&tctx);
@@ -660,6 +696,7 @@
   if (imgunit->img->get_pps().entropy_coding_sync_enabled_flag &&
       sliceunit->shdr->first_slice_segment_in_pic_flag) {
     imgunit->ctx_models.resize( (img->get_sps().PicHeightInCtbsY-1) ); //* CONTEXT_MODEL_TABLE_LENGTH );
+    imgunit->StatCoeff_models.assign( (img->get_sps().PicHeightInCtbsY-1), {{0,0,0,0}} );
   }
 
   sliceunit->nThreads=1;
@@ -809,8 +846,8 @@
   slice_segment_header* shdr = sliceunit->shdr;
   const pic_parameter_set& pps = img->get_pps();
 
-  int nRows = shdr->num_entry_point_offsets +1;
-  int ctbsWidth = img->get_sps().PicWidthInCtbsY;
+  uint16_t nRows = shdr->num_entry_point_offsets +1;
+  uint16_t ctbsWidth = img->get_sps().PicWidthInCtbsY;
 
 
   assert(img->num_threads_active() == 0);
@@ -821,6 +858,7 @@
   if (shdr->first_slice_segment_in_pic_flag) {
     // reserve space for nRows-1 because we don't need to save the CABAC model in the last CTB row
     imgunit->ctx_models.resize( (img->get_sps().PicHeightInCtbsY-1) ); //* CONTEXT_MODEL_TABLE_LENGTH );
+    imgunit->StatCoeff_models.assign( (img->get_sps().PicHeightInCtbsY-1), {{0,0,0,0}} );
   }
 
 
@@ -828,10 +866,14 @@
 
 
   // first CTB in this slice
-  int ctbAddrRS = shdr->slice_segment_address;
-  int ctbRow    = ctbAddrRS / ctbsWidth;
+  uint32_t ctbAddrRS = shdr->slice_segment_address;
+  uint16_t ctbRow    = ctbAddrRS / ctbsWidth;
 
-  for (int entryPt=0;entryPt<nRows;entryPt++) {
+  if (ctbRow + nRows > img->get_sps().PicHeightInCtbsY) {
+    return DE265_WARNING_SLICEHEADER_INVALID;
+  }
+
+  for (uint16_t entryPt=0;entryPt<nRows;entryPt++) {
     // entry points other than the first start at CTB rows
     if (entryPt>0) {
       ctbRow++;
@@ -857,7 +899,12 @@
     tctx->img     = img;
     tctx->imgunit = imgunit;
     tctx->sliceunit= sliceunit;
-    tctx->CtbAddrInTS = pps.CtbAddrRStoTSctbAddrRS;
+
+    if (ctbAddrRS >= pps.scan->CtbAddrRStoTS.size()) {
+      err = DE265_WARNING_SLICEHEADER_INVALID;
+      break;
+    }
+    tctx->CtbAddrInTS = pps.scan->CtbAddrRStoTSctbAddrRS;
 
     init_thread_context(tctx);
 
@@ -922,8 +969,8 @@
   slice_segment_header* shdr = sliceunit->shdr;
   const pic_parameter_set& pps = img->get_pps();
 
-  int nTiles = shdr->num_entry_point_offsets +1;
-  int ctbsWidth = img->get_sps().PicWidthInCtbsY;
+  uint16_t nTiles = shdr->num_entry_point_offsets +1;
+  uint16_t ctbsWidth = img->get_sps().PicWidthInCtbsY;
 
 
   assert(img->num_threads_active() == 0);
@@ -932,10 +979,16 @@
 
 
   // first CTB in this slice
-  int ctbAddrRS = shdr->slice_segment_address;
-  int tileID = pps.TileIdRSctbAddrRS;
+  uint32_t ctbAddrRS = shdr->slice_segment_address;
 
-  for (int entryPt=0;entryPt<nTiles;entryPt++) {
+  // pps.scan->TileIdRS and pps.scan->CtbAddrRStoTS are both sized to PicSizeInCtbsY in
+  // set_derived_values(), so one bound covers both accesses below.
+  if (ctbAddrRS >= pps.scan->CtbAddrRStoTS.size()) {
+    return DE265_WARNING_SLICEHEADER_INVALID;
+  }
+  int tileID = pps.scan->TileIdRSctbAddrRS;
+
+  for (uint16_t entryPt=0;entryPt<nTiles;entryPt++) {
     // entry points other than the first start at tile beginnings
     if (entryPt>0) {
       tileID++;
@@ -945,9 +998,14 @@
         break;
       }
 
-      int ctbX = pps.colBdtileID % pps.num_tile_columns;
-      int ctbY = pps.rowBdtileID / pps.num_tile_columns;
+      uint16_t ctbX = pps.colBdtileID % pps.num_tile_columns;
+      uint16_t ctbY = pps.rowBdtileID / pps.num_tile_columns;
       ctbAddrRS = ctbY * ctbsWidth + ctbX;
+
+      if (ctbAddrRS >= pps.scan->CtbAddrRStoTS.size()) {
+        err = DE265_WARNING_SLICEHEADER_INVALID;
+        break;
+      }
     }
 
     // set thread context
@@ -959,7 +1017,7 @@
     tctx->img    = img;
     tctx->imgunit = imgunit;
     tctx->sliceunit= sliceunit;
-    tctx->CtbAddrInTS = pps.CtbAddrRStoTSctbAddrRS;
+    tctx->CtbAddrInTS = pps.scan->CtbAddrRStoTSctbAddrRS;
 
     init_thread_context(tctx);
 
@@ -989,8 +1047,8 @@
     img->thread_start(1);
     sliceunit->nThreads++;
     add_task_decode_slice_segment(tctx, entryPt==0,
-                                  ctbAddrRS % ctbsWidth,
-                                  ctbAddrRS / ctbsWidth);
+                                  static_cast<uint16_t>(ctbAddrRS % ctbsWidth),
+                                  static_cast<uint16_t>(ctbAddrRS / ctbsWidth));
   }
 
   img->wait_for_completion();
@@ -1571,7 +1629,7 @@
 bool decoder_context::construct_reference_picture_lists(slice_segment_header* hdr)
 {
   int NumPocTotalCurr = hdr->NumPocTotalCurr;
-  int NumRpsCurrTempList0 = libde265_max(hdr->num_ref_idx_l0_active, NumPocTotalCurr);
+  int NumRpsCurrTempList0 = std::max((int)hdr->num_ref_idx_l0_active, NumPocTotalCurr);
 
   // TODO: fold code for both lists together
 
@@ -1643,7 +1701,7 @@
   */
 
   if (hdr->slice_type == SLICE_TYPE_B) {
-    int NumRpsCurrTempList1 = libde265_max(hdr->num_ref_idx_l1_active, NumPocTotalCurr);
+    int NumRpsCurrTempList1 = std::max((int)hdr->num_ref_idx_l1_active, NumPocTotalCurr);
 
     int rIdx=0;
     while (rIdx < NumRpsCurrTempList1) {
@@ -1966,7 +2024,9 @@
     hdr->SliceAddrRS = previous_slice_header->SliceAddrRS;
   }
 
-  previous_slice_header = hdr;
+  // Note: previous_slice_header is updated by the caller (read_slice_NAL) only
+  // once the slice header is actually retained by the image. Setting it here
+  // would leave a dangling pointer when the caller discards/deletes 'hdr'.
 
 
   loginfo(LogHeaders,"SliceAddrRS = %d\n",hdr->SliceAddrRS);

libde265-1.0.17.tar.gz/libde265/decctx.h -> libde265-1.1.1.tar.gz/libde265/decctx.h Changed

@@ -35,6 +35,7 @@
 #include "libde265/acceleration.h"
 #include "libde265/nal-parser.h"
 
+#include <array>
 #include <memory>
 
 constexpr int DE265_MAX_VPS_SETS = 16;   // this is the maximum as defined in the standard
@@ -262,6 +263,11 @@
      There is one saved model for the initialization of each CTB row.
      The array is unused for non-WPP streams. */
   std::vector<context_model_table> ctx_models;  // TODO: move this into image ?
+
+  /* Saved StatCoeff (persistent_rice_adaptation state) parallel to ctx_models.
+     Per HEVC RExt, this state must be carried across WPP CTB rows together
+     with the CABAC context. */
+  std::vector<std::array<uint8_t, 4>> StatCoeff_models;
 };
 
 
@@ -350,6 +356,13 @@
   //bool param_disable_mc_residual_idct;  // not implemented yet
   //bool param_disable_intra_residual_idct;  // not implemented yet
 
+  de265_security_limits param_security_limits = {
+    1,                // version
+    8192 * 8192,      // max_image_size_pixels
+    16 * 1024 * 1024, // max_NAL_size_bytes
+    256               // max_SEI_messages
+  };
+
   void set_image_allocation_functions(de265_image_allocation* allocfunc, void* userdata);
 
   de265_image_allocation param_image_allocation_functions; // initialized in constructor
@@ -499,9 +512,9 @@
 
  private:
   void init_thread_context(thread_context* tctx);
-  void add_task_decode_CTB_row(thread_context* tctx, bool firstSliceSubstream, int ctbRow);
+  void add_task_decode_CTB_row(thread_context* tctx, bool firstSliceSubstream, uint16_t ctbRow);
   void add_task_decode_slice_segment(thread_context* tctx, bool firstSliceSubstream,
-                                     int ctbX,int ctbY);
+                                     uint16_t ctbX, uint16_t ctbY);
 
   void mark_whole_slice_as_processed(image_unit* imgunit,
                                      slice_unit* sliceunit,

libde265-1.0.17.tar.gz/libde265/dpb.cc -> libde265-1.1.1.tar.gz/libde265/dpb.cc Changed

libde265-1.0.17.tar.gz/libde265/fallback-dct.cc -> libde265-1.1.1.tar.gz/libde265/fallback-dct.cc Changed

libde265-1.0.17.tar.gz/libde265/fallback-dct.h -> libde265-1.1.1.tar.gz/libde265/fallback-dct.h Changed

libde265-1.1.1.tar.gz/libde265/fallback-deblk.cc Added

@@ -0,0 +1,34 @@
+/*
+ * H.265 video codec.
+ * Copyright (c) 2026 Dirk Farin <dirk.farin@gmail.com>
+ *
+ * This file is part of libde265.
+ *
+ * libde265 is free software: you can redistribute it and/or modify
+ * it under the terms of the GNU Lesser General Public License as
+ * published by the Free Software Foundation, either version 3 of
+ * the License, or (at your option) any later version.
+ *
+ * libde265 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 Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public License
+ * along with libde265.  If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#include "fallback-deblk.h"
+
+void deblock_luma_8_fallback(uint8_t* ptr, ptrdiff_t stride, int vertical,
+                             int dE, int dEp, int dEq, int tc, int filterP, int filterQ)
+{
+  deblock_luma_kernel<uint8_t>(ptr, stride, vertical!=0, dE, dEp, dEq, tc,
+                               filterP!=0, filterQ!=0, 8);
+}
+
+void deblock_chroma_8_fallback(uint8_t* ptr, ptrdiff_t stride, int vertical,
+                               int tc, int filterP, int filterQ)
+{
+  deblock_chroma_kernel<uint8_t>(ptr, stride, vertical!=0, tc, filterP!=0, filterQ!=0, 8);
+}

libde265-1.1.1.tar.gz/libde265/fallback-deblk.h Added

@@ -0,0 +1,133 @@
+/*
+ * H.265 video codec.
+ * Copyright (c) 2026 Dirk Farin <dirk.farin@gmail.com>
+ *
+ * This file is part of libde265.
+ *
+ * libde265 is free software: you can redistribute it and/or modify
+ * it under the terms of the GNU Lesser General Public License as
+ * published by the Free Software Foundation, either version 3 of
+ * the License, or (at your option) any later version.
+ *
+ * libde265 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 Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public License
+ * along with libde265.  If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#ifndef DE265_FALLBACK_DEBLK_H
+#define DE265_FALLBACK_DEBLK_H
+
+#include <stddef.h>
+#include <stdint.h>
+#include "util.h"
+
+// One luma edge-filter segment (4 lines along the edge), spec 8.7.2.4.4.
+// 'ptr' points at the q0 sample of line 0. dE in {1,2} (weak/strong); the
+// caller guarantees dE != 0. The filterP/filterQ flags disable the p- resp.
+// q-side (PCM / transquant-bypass).
+template <class pixel_t>
+void deblock_luma_kernel(pixel_t* ptr, ptrdiff_t stride, bool vertical,
+                         int dE, int dEp, int dEq, int tc,
+                         bool filterP, bool filterQ, int bitDepth)
+{
+  for (int k=0;k<4;k++) {
+    pixel_t p0,p1,p2,p3,q0,q1,q2,q3;
+    if (vertical) {
+      p0=ptr-1+k*stride; p1=ptr-2+k*stride; p2=ptr-3+k*stride; p3=ptr-4+k*stride;
+      q0=ptr 0+k*stride; q1=ptr 1+k*stride; q2=ptr 2+k*stride; q3=ptr 3+k*stride;
+    } else {
+      p0=ptrk-1*stride; p1=ptrk-2*stride; p2=ptrk-3*stride; p3=ptrk-4*stride;
+      q0=ptrk+0*stride; q1=ptrk+1*stride; q2=ptrk+2*stride; q3=ptrk+3*stride;
+    }
+
+    if (dE==2) {
+      // strong filtering
+      pixel_t pnew3,qnew3;
+      pnew0 = Clip3(p0-2*tc,p0+2*tc, (p2 + 2*p1 + 2*p0 + 2*q0 + q1 +4)>>3);
+      pnew1 = Clip3(p1-2*tc,p1+2*tc, (p2 + p1 + p0 + q0+2)>>2);
+      pnew2 = Clip3(p2-2*tc,p2+2*tc, (2*p3 + 3*p2 + p1 + p0 + q0 + 4)>>3);
+      qnew0 = Clip3(q0-2*tc,q0+2*tc, (p1+2*p0+2*q0+2*q1+q2+4)>>3);
+      qnew1 = Clip3(q1-2*tc,q1+2*tc, (p0+q0+q1+q2+2)>>2);
+      qnew2 = Clip3(q2-2*tc,q2+2*tc, (p0+q0+q1+3*q2+2*q3+4)>>3);
+
+      if (vertical) {
+        for (int i=0;i<3;i++) {
+          if (filterP) { ptr-i-1+k*stride = pnewi; }
+          if (filterQ) { ptr i + k*stride = qnewi; }
+        }
+      } else {
+        for (int i=0;i<3;i++) {
+          if (filterP) { ptr k -(i+1)*stride = pnewi; }
+          if (filterQ) { ptr k + i   *stride = qnewi; }
+        }
+      }
+    }
+    else {
+      // weak filtering
+      int delta = (9*(q0-p0) - 3*(q1-p1) + 8)>>4;
+
+      if (std::abs(delta) < tc*10) {
+        delta = Clip3(-tc,tc,delta);
+
+        if (vertical) {
+          if (filterP) { ptr-0-1+k*stride = Clip_BitDepth(p0+delta, bitDepth); }
+          if (filterQ) { ptr 0  +k*stride = Clip_BitDepth(q0-delta, bitDepth); }
+        } else {
+          if (filterP) { ptr k -1*stride = Clip_BitDepth(p0+delta, bitDepth); }
+          if (filterQ) { ptr k +0*stride = Clip_BitDepth(q0-delta, bitDepth); }
+        }
+
+        if (dEp==1 && filterP) {
+          int delta_p = Clip3(-(tc>>1), tc>>1, (((p2+p0+1)>>1)-p1+delta)>>1);
+          if (vertical) { ptr-1-1+k*stride = Clip_BitDepth(p1+delta_p, bitDepth); }
+          else          { ptr k  -2*stride = Clip_BitDepth(p1+delta_p, bitDepth); }
+        }
+
+        if (dEq==1 && filterQ) {
+          int delta_q = Clip3(-(tc>>1), tc>>1, (((q2+q0+1)>>1)-q1-delta)>>1);
+          if (vertical) { ptr 1  +k*stride = Clip_BitDepth(q1+delta_q, bitDepth); }
+          else          { ptr k  +1*stride = Clip_BitDepth(q1+delta_q, bitDepth); }
+        }
+      }
+    }
+  }
+}
+
+
+// One chroma edge-filter segment (4 lines), spec 8.7.2.4.5.
+template <class pixel_t>
+void deblock_chroma_kernel(pixel_t* ptr, ptrdiff_t stride, bool vertical,
+                           int tc, bool filterP, bool filterQ, int bitDepth)
+{
+  for (int k=0;k<4;k++) {
+    pixel_t p0,p1,q0,q1;
+    if (vertical) {
+      q0=ptr 0+k*stride; q1=ptr 1+k*stride; p0=ptr-1+k*stride; p1=ptr-2+k*stride;
+    } else {
+      q0=ptrk+0*stride; q1=ptrk+1*stride; p0=ptrk-1*stride; p1=ptrk-2*stride;
+    }
+
+    int delta = Clip3(-tc,tc, ((((q0-p0)*4)+p1-q1+4)>>3));
+
+    if (vertical) {
+      if (filterP) { ptr-1+k*stride = Clip_BitDepth(p0+delta, bitDepth); }
+      if (filterQ) { ptr 0+k*stride = Clip_BitDepth(q0-delta, bitDepth); }
+    } else {
+      if (filterP) { ptr k-1*stride = Clip_BitDepth(p0+delta, bitDepth); }
+      if (filterQ) { ptr k+0*stride = Clip_BitDepth(q0-delta, bitDepth); }
+    }
+  }
+}
+
+
+// 8-bit fallback wrappers stored in the acceleration table.
+void deblock_luma_8_fallback(uint8_t* ptr, ptrdiff_t stride, int vertical,
+                             int dE, int dEp, int dEq, int tc, int filterP, int filterQ);
+void deblock_chroma_8_fallback(uint8_t* ptr, ptrdiff_t stride, int vertical,
+                               int tc, int filterP, int filterQ);
+
+#endif

libde265-1.1.1.tar.gz/libde265/fallback-intrapred.cc Added

@@ -0,0 +1,54 @@
+/*
+ * H.265 video codec.
+ * Copyright (c) 2026 Dirk Farin <dirk.farin@gmail.com>
+ *
+ * This file is part of libde265.
+ *
+ * libde265 is free software: you can redistribute it and/or modify
+ * it under the terms of the GNU Lesser General Public License as
+ * published by the Free Software Foundation, either version 3 of
+ * the License, or (at your option) any later version.
+ *
+ * libde265 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 Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public License
+ * along with libde265.  If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#include "fallback-intrapred.h"
+#include "intrapred.h"
+
+
+template <class pixel_t>
+void intra_pred_dc_fallback(pixel_t* dst, ptrdiff_t stride, int nT, int cIdx, const pixel_t* border)
+{
+  intra_prediction_DC<pixel_t>(dst, (int)stride, nT, cIdx, const_cast<pixel_t*>(border));
+}
+
+template <class pixel_t>
+void intra_pred_planar_fallback(pixel_t* dst, ptrdiff_t stride, int nT, int cIdx, const pixel_t* border)
+{
+  intra_prediction_planar<pixel_t>(dst, (int)stride, nT, cIdx, const_cast<pixel_t*>(border));
+}
+
+template <class pixel_t>
+void intra_pred_angular_fallback(pixel_t* dst, ptrdiff_t stride, int bit_depth, int disableBoundaryFilter,
+                                 int xB0, int yB0, int mode, int nT, int cIdx, const pixel_t* border)
+{
+  intra_prediction_angular<pixel_t>(dst, (int)stride, bit_depth, (bool)disableBoundaryFilter,
+                                    xB0, yB0, (enum IntraPredMode)mode, nT, cIdx,
+                                    const_cast<pixel_t*>(border));
+}
+
+
+// explicit instantiations so the symbols can be installed into the acceleration table
+
+template void intra_pred_dc_fallback<uint8_t> (uint8_t*,  ptrdiff_t, int, int, const uint8_t*);
+template void intra_pred_dc_fallback<uint16_t>(uint16_t*, ptrdiff_t, int, int, const uint16_t*);
+template void intra_pred_planar_fallback<uint8_t> (uint8_t*,  ptrdiff_t, int, int, const uint8_t*);
+template void intra_pred_planar_fallback<uint16_t>(uint16_t*, ptrdiff_t, int, int, const uint16_t*);
+template void intra_pred_angular_fallback<uint8_t> (uint8_t*,  ptrdiff_t, int, int, int, int, int, int, int, const uint8_t*);
+template void intra_pred_angular_fallback<uint16_t>(uint16_t*, ptrdiff_t, int, int, int, int, int, int, int, const uint16_t*);

libde265-1.1.1.tar.gz/libde265/fallback-intrapred.h Added

@@ -0,0 +1,41 @@
+/*
+ * H.265 video codec.
+ * Copyright (c) 2026 Dirk Farin <dirk.farin@gmail.com>
+ *
+ * This file is part of libde265.
+ *
+ * libde265 is free software: you can redistribute it and/or modify
+ * it under the terms of the GNU Lesser General Public License as
+ * published by the Free Software Foundation, either version 3 of
+ * the License, or (at your option) any later version.
+ *
+ * libde265 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 Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public License
+ * along with libde265.  If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#ifndef DE265_FALLBACK_INTRAPRED_H
+#define DE265_FALLBACK_INTRAPRED_H
+
+#include <stddef.h>
+#include <stdint.h>
+
+// Scalar fallback wrappers around the intra-prediction kernels in intrapred.h.
+// They have plain (non-templated mode/flag) signatures so they can be stored in
+// the acceleration_functions function-pointer table.
+
+template <class pixel_t>
+void intra_pred_dc_fallback(pixel_t* dst, ptrdiff_t stride, int nT, int cIdx, const pixel_t* border);
+
+template <class pixel_t>
+void intra_pred_planar_fallback(pixel_t* dst, ptrdiff_t stride, int nT, int cIdx, const pixel_t* border);
+
+template <class pixel_t>
+void intra_pred_angular_fallback(pixel_t* dst, ptrdiff_t stride, int bit_depth, int disableBoundaryFilter,
+                                 int xB0, int yB0, int mode, int nT, int cIdx, const pixel_t* border);
+
+#endif

libde265-1.0.17.tar.gz/libde265/fallback-motion.cc -> libde265-1.1.1.tar.gz/libde265/fallback-motion.cc Changed

@@ -165,9 +165,11 @@
                                      const int16_t *src, ptrdiff_t srcstride,
                                      int width, int height, int bit_depth)
 {
-  int shift1 = 14-bit_depth;
-  int offset1 = 0;
-  if (shift1>0) { offset1 = 1<<(shift1-1); }
+  // shift1 per HEVC v2 (10/2014) spec 8.5.3.3.4.2: Max(2, 14 - BitDepth).
+  // The Max() was added with the Range Extensions in v2 to handle BitDepth up to 16;
+  // the v1 (04/2013) formula was just (14 - BitDepth), valid only for BitDepth <= 14.
+  int shift1 = std::max(2, 14-bit_depth);
+  int offset1 = 1<<(shift1-1);
 
   assert((width&1)==0);
 
@@ -232,7 +234,10 @@
                                        ptrdiff_t srcstride, int width,
                                        int height, int bit_depth)
 {
-  int shift2 = 15-bit_depth;
+  // shift2 per HEVC v2 (10/2014) spec 8.5.3.3.4.2: Max(3, 15 - BitDepth).
+  // The Max() was added with the Range Extensions in v2 to handle BitDepth up to 16;
+  // the v1 (04/2013) formula was just (15 - BitDepth), valid only for BitDepth <= 14.
+  int shift2 = std::max(3, 15-bit_depth);
   int offset2 = 1<<(shift2-1);
 
   assert((width&1)==0);
@@ -279,7 +284,10 @@
                           int width, int height,
                           int mx, int my, int16_t* mcbuffer, int bit_depth)
 {
-  int shift3 = 14 - bit_depth;
+  // shift3 per HEVC v2 (10/2014) spec 8.5.3.3.3.3 (chroma): Max(2, 14 - BitDepth).
+  // The Max() was added with the Range Extensions in v2 to handle BitDepth up to 16;
+  // the v1 (04/2013) formula was just (14 - BitDepth), valid only for BitDepth <= 14.
+  int shift3 = std::max(2, 14 - bit_depth);
 
   for (int y=0;y<height;y++) {
     int16_t* o = &outy*out_stride;
@@ -459,7 +467,10 @@
 {
   //const int shift1 = bit_depth-8;
   //const int shift2 = 6;
-  const int shift3 = 14-bit_depth;
+  // shift3 per HEVC v2 (10/2014) spec 8.5.3.3.3.2 (luma): Max(2, 14 - BitDepth).
+  // The Max() was added with the Range Extensions in v2 to handle BitDepth up to 16;
+  // the v1 (04/2013) formula was just (14 - BitDepth), valid only for BitDepth <= 14.
+  const int shift3 = std::max(2, 14-bit_depth);
 
   // straight copy

libde265-1.0.17.tar.gz/libde265/fallback.cc -> libde265-1.1.1.tar.gz/libde265/fallback.cc Changed

@@ -21,6 +21,8 @@
 #include "fallback.h"
 #include "fallback-motion.h"
 #include "fallback-dct.h"
+#include "fallback-intrapred.h"
+#include "fallback-deblk.h"
 
 
 void init_acceleration_functions_fallback(struct acceleration_functions* accel)
@@ -104,6 +106,7 @@
   accel->rotate_coefficients = rotate_coefficients_fallback;
   accel->add_residual_8  = add_residual_fallback<uint8_t>;
   accel->add_residual_16 = add_residual_fallback<uint16_t>;
+  accel->dequant_coeff_block = dequant_coeff_block_fallback;
   accel->rdpcm_h = rdpcm_h_fallback;
   accel->rdpcm_v = rdpcm_v_fallback;
   accel->transform_skip_residual = transform_skip_residual_fallback;
@@ -124,4 +127,14 @@
   accel->hadamard_transform_81 = hadamard_8x8_8_fallback;
   accel->hadamard_transform_82 = hadamard_16x16_8_fallback;
   accel->hadamard_transform_83 = hadamard_32x32_8_fallback;
+
+  accel->intra_pred_dc_8  = intra_pred_dc_fallback<uint8_t>;
+  accel->intra_pred_dc_16 = intra_pred_dc_fallback<uint16_t>;
+  accel->intra_pred_planar_8  = intra_pred_planar_fallback<uint8_t>;
+  accel->intra_pred_planar_16 = intra_pred_planar_fallback<uint16_t>;
+  accel->intra_pred_angular_8  = intra_pred_angular_fallback<uint8_t>;
+  accel->intra_pred_angular_16 = intra_pred_angular_fallback<uint16_t>;
+
+  accel->deblock_luma_8   = deblock_luma_8_fallback;
+  accel->deblock_chroma_8 = deblock_chroma_8_fallback;
 }

libde265-1.0.17.tar.gz/libde265/image-io.cc -> libde265-1.1.1.tar.gz/libde265/image-io.cc Changed

libde265-1.0.17.tar.gz/libde265/image.cc -> libde265-1.1.1.tar.gz/libde265/image.cc Changed

@@ -20,7 +20,6 @@
 
 #include "image.h"
 #include "decctx.h"
-#include "en265.h"
 
 #include <atomic>
 
@@ -45,10 +44,10 @@
 
 #define STANDARD_ALIGNMENT 16
 
-#ifdef HAVE___MINGW_ALIGNED_MALLOC
+#if defined(__MINGW32__)
 #define ALLOC_ALIGNED(alignment, size)         __mingw_aligned_malloc((size), (alignment))
 #define FREE_ALIGNED(mem)                      __mingw_aligned_free((mem))
-#elif _WIN32
+#elif defined(_MSC_VER)
 #define ALLOC_ALIGNED(alignment, size)         _aligned_malloc((size), (alignment))
 #define FREE_ALIGNED(mem)                      _aligned_free((mem))
 #elif defined(HAVE_POSIX_MEMALIGN)
@@ -71,10 +70,11 @@
                                            void* inputdata, int inputstride, void *userdata)
 {
   int alignment = STANDARD_ALIGNMENT;
-  int stride = (img->get_width(cIdx) + alignment-1) / alignment * alignment;
-  int height = img->get_height(cIdx);
+  uint32_t stride = (img->get_width(cIdx) + alignment-1) / alignment * alignment;
+  uint32_t height = img->get_height(cIdx);
 
-  uint8_t* p = static_cast<uint8_t*>(ALLOC_ALIGNED_16(stride * height + MEMORY_PADDING));
+  // size computed in size_t: stride*height can exceed UINT32_MAX for large planes
+  uint8_t* p = static_cast<uint8_t*>(ALLOC_ALIGNED_16(static_cast<size_t>(stride) * height + MEMORY_PADDING));
 
   if (p==nullptr) { return nullptr; }
 
@@ -83,12 +83,14 @@
   // copy input data if provided
 
   if (inputdata != nullptr) {
-    if (inputstride == stride) {
-      memcpy(p, inputdata, stride*height);
+    if (inputstride == static_cast<int>(stride)) {
+      memcpy(p, inputdata, static_cast<size_t>(stride) * height);
     }
     else {
-      for (int y=0;y<height;y++) {
-        memcpy(p+y*stride, static_cast<char*>(inputdata) + inputstride*y, inputstride);
+      for (uint32_t y=0;y<height;y++) {
+        memcpy(p + static_cast<size_t>(y) * stride,
+               static_cast<char*>(inputdata) + static_cast<size_t>(inputstride) * y,
+               inputstride);
       }
     }
   }
@@ -108,30 +110,35 @@
 static int  de265_image_get_buffer(de265_decoder_context* ctx,
                                    de265_image_spec* spec, de265_image* img, void* userdata)
 {
-  const int rawChromaWidth  = spec->width  / img->SubWidthC;
-  const int rawChromaHeight = spec->height / img->SubHeightC;
+  const uint32_t rawChromaWidth  = spec->width  / img->SubWidthC;
+  const uint32_t rawChromaHeight = spec->height / img->SubHeightC;
 
-  int luma_stride   = (spec->width    + spec->alignment-1) / spec->alignment * spec->alignment;
-  int chroma_stride = (rawChromaWidth + spec->alignment-1) / spec->alignment * spec->alignment;
+  uint32_t luma_stride   = (spec->width    + spec->alignment-1) / spec->alignment * spec->alignment;
+  uint32_t chroma_stride = (rawChromaWidth + spec->alignment-1) / spec->alignment * spec->alignment;
 
   assert(img->BitDepth_Y >= 8 && img->BitDepth_Y <= 16);
   assert(img->BitDepth_C >= 8 && img->BitDepth_C <= 16);
 
-  int luma_bpl   = luma_stride   * ((img->BitDepth_Y+7)/8);
-  int chroma_bpl = chroma_stride * ((img->BitDepth_C+7)/8);
+  uint32_t luma_bpl   = luma_stride   * ((img->BitDepth_Y+7)/8);
+  uint32_t chroma_bpl = chroma_stride * ((img->BitDepth_C+7)/8);
 
-  int luma_height   = spec->height;
-  int chroma_height = rawChromaHeight;
+  uint32_t luma_height   = spec->height;
+  uint32_t chroma_height = rawChromaHeight;
 
   bool alloc_failed = false;
 
-  uint8_t* p3 = { 0,0,0 };
-  p0 = static_cast<uint8_t*>(ALLOC_ALIGNED_16(luma_height   * luma_bpl   + MEMORY_PADDING));
+  // Compute the plane sizes in size_t. Each operand fits in uint32_t, but the
+  // height * bytes-per-line product can exceed UINT32_MAX for large frames, so
+  // the multiplication must be done in 64 bits. Computing it in 32 bits wraps
+  // the allocation size to a small value while fill_image() later writes the
+  // real (size_t) size -> heap buffer overflow (GHSA-vv8h-932h-7r86).
+  uint8_t* p3 = { nullptr,nullptr,nullptr };
+  p0 = static_cast<uint8_t*>(ALLOC_ALIGNED_16(static_cast<size_t>(luma_height) * luma_bpl + MEMORY_PADDING));
   if (p0==nullptr) { alloc_failed=true; }
 
   if (img->get_chroma_format() != de265_chroma_mono) {
-    p1 = static_cast<uint8_t*>(ALLOC_ALIGNED_16(chroma_height * chroma_bpl + MEMORY_PADDING));
-    p2 = static_cast<uint8_t*>(ALLOC_ALIGNED_16(chroma_height * chroma_bpl + MEMORY_PADDING));
+    p1 = static_cast<uint8_t*>(ALLOC_ALIGNED_16(static_cast<size_t>(chroma_height) * chroma_bpl + MEMORY_PADDING));
+    p2 = static_cast<uint8_t*>(ALLOC_ALIGNED_16(static_cast<size_t>(chroma_height) * chroma_bpl + MEMORY_PADDING));
 
     if (p1==nullptr || p2==nullptr) { alloc_failed=true; }
   }
@@ -177,7 +184,7 @@
 };
 
 
-void de265_image::set_image_plane(int cIdx, uint8_t* mem, int stride, void *userdata)
+void de265_image::set_image_plane(int cIdx, uint8_t* mem, ptrdiff_t stride, void *userdata)
 {
   pixelscIdx = mem;
   plane_user_datacIdx = userdata;
@@ -490,24 +497,21 @@
   int bytes_per_pixel = get_bytes_per_pixel(channel);
   assert(value >= 0); // needed for the shift operation in the check below
 
+  // Each plane is allocated with MEMORY_PADDING trailing bytes for safe SSE overread; the
+  // memsets below cover that padding too so it never contains uninitialized heap data.
+  const size_t plane_bytes =
+      (channel == 0 ? static_cast<size_t>(stride) * height
+                    : static_cast<size_t>(chroma_stride) * chroma_height)
+      * bytes_per_pixel;
+
   if (bytes_per_pixel == 1) {
-    if (channel==0) {
-      memset(pixelschannel, value, stride * height);
-    }
-    else {
-      memset(pixelschannel, value, chroma_stride * chroma_height);
-    }
+    memset(pixelschannel, value, plane_bytes + MEMORY_PADDING);
   }
   else if ((value >> 8) == (value & 0xFF)) {
     assert(bytes_per_pixel == 2);
 
     // if we fill the same byte value to all bytes, we can still use memset()
-    if (channel==0) {
-      memset(pixelschannel, 0, stride * height * bytes_per_pixel);
-    }
-    else {
-      memset(pixelschannel, 0, chroma_stride * chroma_height * bytes_per_pixel);
-    }
+    memset(pixelschannel, 0, plane_bytes + MEMORY_PADDING);
   }
   else {
     assert(bytes_per_pixel == 2);
@@ -535,6 +539,10 @@
         memcpy(pixelschannel + y * chroma_stride * 2, pixelschannel, chroma_width * 2);
       }
     }
+
+#if MEMORY_PADDING > 0
+    memset(pixelschannel + plane_bytes, 0, MEMORY_PADDING);
+#endif
   }
 }
 
@@ -782,9 +790,9 @@
   if (xN>=sps->pic_width_in_luma_samples ||
       yN>=sps->pic_height_in_luma_samples) return false;
 
-  int minBlockAddrN = pps->MinTbAddrZS (xN>>sps->Log2MinTrafoSize) +
+  int minBlockAddrN = pps->scan->MinTbAddrZS (xN>>sps->Log2MinTrafoSize) +
                                         (yN>>sps->Log2MinTrafoSize) * sps->PicWidthInTbsY ;
-  int minBlockAddrCurr = pps->MinTbAddrZS (xCurr>>sps->Log2MinTrafoSize) +
+  int minBlockAddrCurr = pps->scan->MinTbAddrZS (xCurr>>sps->Log2MinTrafoSize) +
                                            (yCurr>>sps->Log2MinTrafoSize) * sps->PicWidthInTbsY ;
 
   if (minBlockAddrN > minBlockAddrCurr) return false;
@@ -799,8 +807,8 @@
     return false;
   }
 
-  if (pps->TileIdRSxCurrCtb + yCurrCtb*sps->PicWidthInCtbsY !=
-      pps->TileIdRSxNCtb    + yNCtb   *sps->PicWidthInCtbsY) {
+  if (pps->scan->TileIdRSxCurrCtb + yCurrCtb*sps->PicWidthInCtbsY !=
+      pps->scan->TileIdRSxNCtb    + yNCtb   *sps->PicWidthInCtbsY) {
     return false;
   }

libde265-1.0.17.tar.gz/libde265/image.h -> libde265-1.1.1.tar.gz/libde265/image.h Changed

@@ -26,6 +26,7 @@
 #endif
 
 #include <assert.h>
+#include <stddef.h>
 #include <stdint.h>
 #include <stdlib.h>
 #include <string.h>
@@ -235,11 +236,11 @@
   /* */ uint8_t* get_image_plane(int cIdx)       { return pixelscIdx; }
   const uint8_t* get_image_plane(int cIdx) const { return pixelscIdx; }
 
-  void set_image_plane(int cIdx, uint8_t* mem, int stride, void *userdata);
+  void set_image_plane(int cIdx, uint8_t* mem, ptrdiff_t stride, void *userdata);
 
   uint8_t* get_image_plane_at_pos(int cIdx, int xpos,int ypos)
   {
-    int stride = get_image_stride(cIdx);
+    ptrdiff_t stride = get_image_stride(cIdx);
     return pixelscIdx + xpos + ypos*stride;
   }
 
@@ -248,38 +249,38 @@
   template <class pixel_t>
   pixel_t* get_image_plane_at_pos_NEW(int cIdx, int xpos,int ypos)
   {
-    int stride = get_image_stride(cIdx);
+    ptrdiff_t stride = get_image_stride(cIdx);
     return (pixel_t*)(pixelscIdx + (xpos + ypos*stride)*sizeof(pixel_t));
   }
 
   const uint8_t* get_image_plane_at_pos(int cIdx, int xpos,int ypos) const
   {
-    int stride = get_image_stride(cIdx);
+    ptrdiff_t stride = get_image_stride(cIdx);
     return pixelscIdx + xpos + ypos*stride;
   }
 
   void* get_image_plane_at_pos_any_depth(int cIdx, int xpos,int ypos)
   {
-    int stride = get_image_stride(cIdx);
+    ptrdiff_t stride = get_image_stride(cIdx);
     return pixelscIdx + ((xpos + ypos*stride) << bpp_shiftcIdx);
   }
 
   const void* get_image_plane_at_pos_any_depth(int cIdx, int xpos,int ypos) const
   {
-    int stride = get_image_stride(cIdx);
+    ptrdiff_t stride = get_image_stride(cIdx);
     return pixelscIdx + ((xpos + ypos*stride) << bpp_shiftcIdx);
   }
 
   /* Number of pixels in one row (not number of bytes).
    */
-  int get_image_stride(int cIdx) const
+  ptrdiff_t get_image_stride(int cIdx) const
   {
     if (cIdx==0) return stride;
     else         return chroma_stride;
   }
 
-  int get_luma_stride() const { return stride; }
-  int get_chroma_stride() const { return chroma_stride; }
+  ptrdiff_t get_luma_stride() const { return stride; }
+  ptrdiff_t get_chroma_stride() const { return chroma_stride; }
 
   int get_width (int cIdx=0) const { return cIdx==0 ? width  : chroma_width;  }
   int get_height(int cIdx=0) const { return cIdx==0 ? height : chroma_height; }
@@ -333,7 +334,7 @@
   int width = 0, height = 0;  // size in luma pixels
 
   int chroma_width = 0, chroma_height = 0;
-  int stride = 0, chroma_stride = 0;
+  ptrdiff_t stride = 0, chroma_stride = 0;
 
 public:
   uint8_t BitDepth_Y = 0, BitDepth_C = 0;

libde265-1.0.17.tar.gz/libde265/intrapred.cc -> libde265-1.1.1.tar.gz/libde265/intrapred.cc Changed

@@ -21,6 +21,7 @@
 #include "intrapred.h"
 #include "transform.h"
 #include "util.h"
+#include "decctx.h"
 #include <assert.h>
 
 
@@ -292,12 +293,14 @@
     }
 
 
+  const acceleration_functions& acceleration = img->decctx->acceleration;
+
   switch (intraPredMode) {
   case INTRA_PLANAR:
-    intra_prediction_planar(dst,dstStride, nT,cIdx, border_pixels);
+    acceleration.intra_pred_planar<pixel_t>(dst,dstStride, nT,cIdx, border_pixels);
     break;
   case INTRA_DC:
-    intra_prediction_DC(dst,dstStride, nT,cIdx, border_pixels);
+    acceleration.intra_pred_dc<pixel_t>(dst,dstStride, nT,cIdx, border_pixels);
     break;
   default:
     {
@@ -306,8 +309,8 @@
         (img->get_sps().range_extension.implicit_rdpcm_enabled_flag &&
          img->get_cu_transquant_bypass(xB0,yB0));
 
-      intra_prediction_angular(dst,dstStride, bit_depth,disableIntraBoundaryFilter,
-                               xB0,yB0,intraPredMode,nT,cIdx, border_pixels);
+      acceleration.intra_pred_angular<pixel_t>(dst,dstStride, bit_depth,disableIntraBoundaryFilter,
+                                               xB0,yB0,intraPredMode,nT,cIdx, border_pixels);
     }
     break;
   }

libde265-1.0.17.tar.gz/libde265/intrapred.h -> libde265-1.1.1.tar.gz/libde265/intrapred.h Changed

@@ -195,9 +195,8 @@
   if (intraPredMode==INTRA_DC || nT==4) {
     filterFlag = 0;
   } else {
-    // int-cast below prevents a typing problem that leads to wrong results when abs_value is a macro
-    int minDistVerHor = libde265_min( abs_value((int)intraPredMode-26),
-                                      abs_value((int)intraPredMode-10) );
+    int minDistVerHor = std::min( std::abs((int)intraPredMode-26),
+                                  std::abs((int)intraPredMode-10) );
 
     //printf("mindist: %d\n",minDistVerHor);
 
@@ -217,8 +216,8 @@
     int biIntFlag = (sps.strong_intra_smoothing_enable_flag &&
                      cIdx==0 &&
                      nT==32 &&
-                     abs_value(p0+p 64-2*p 32) < (1<<(sps.bit_depth_luma-5)) &&
-                     abs_value(p0+p-64-2*p-32) < (1<<(sps.bit_depth_luma-5)))
+                     std::abs(p0+p 64-2*p 32) < (1<<(sps.bit_depth_luma-5)) &&
+                     std::abs(p0+p-64-2*p-32) < (1<<(sps.bit_depth_luma-5)))
       ? 1 : 0;
 
     pixel_t  pF_mem4*32+1;
@@ -491,17 +490,17 @@
   int topleftCTBSlice  = availableTopLeft  ? img->get_SliceAddrRS(xLeftCtb, yTopCtb) : -1;
 
   /*
-  printf("size: %d\n",pps->TileIdRS.size());
+  printf("size: %d\n",pps->scan->TileIdRS.size());
   printf("curr: %d left: %d top: %d\n",
          xCurrCtb+yCurrCtb*picWidthInCtbs,
          availableLeft ? xLeftCtb+yCurrCtb*picWidthInCtbs : 9999,
          availableTop  ? xCurrCtb+yTopCtb*picWidthInCtbs  : 9999);
   */
-  uint32_t currCTBTileID = pps->TileIdRSxCurrCtb+yCurrCtb*picWidthInCtbs;
-  uint32_t leftCTBTileID = availableLeft ? pps->TileIdRSxLeftCtb+yCurrCtb*picWidthInCtbs : UINT32_MAX;
-  uint32_t topCTBTileID  = availableTop ? pps->TileIdRSxCurrCtb+yTopCtb*picWidthInCtbs : UINT32_MAX;
-  uint32_t topleftCTBTileID = availableTopLeft ? pps->TileIdRSxLeftCtb+yTopCtb*picWidthInCtbs : UINT32_MAX;
-  uint32_t toprightCTBTileID= availableTopRight? pps->TileIdRSxRightCtb+yTopCtb*picWidthInCtbs : UINT32_MAX;
+  uint32_t currCTBTileID = pps->scan->TileIdRSxCurrCtb+yCurrCtb*picWidthInCtbs;
+  uint32_t leftCTBTileID = availableLeft ? pps->scan->TileIdRSxLeftCtb+yCurrCtb*picWidthInCtbs : UINT32_MAX;
+  uint32_t topCTBTileID  = availableTop ? pps->scan->TileIdRSxCurrCtb+yTopCtb*picWidthInCtbs : UINT32_MAX;
+  uint32_t topleftCTBTileID = availableTopLeft ? pps->scan->TileIdRSxLeftCtb+yTopCtb*picWidthInCtbs : UINT32_MAX;
+  uint32_t toprightCTBTileID= availableTopRight? pps->scan->TileIdRSxRightCtb+yTopCtb*picWidthInCtbs : UINT32_MAX;
 
   if (leftCTBSlice != currCTBSlice  || leftCTBTileID != currCTBTileID ) availableLeft    = false;
   if (topCTBSlice  != currCTBSlice  || topCTBTileID  != currCTBTileID ) availableTop     = false;
@@ -533,14 +532,14 @@
   assert(nT<=32);
 
   pixel_t* image;
-  int stride;
+  ptrdiff_t stride;
   image  = (pixel_t*)img->get_image_plane(cIdx);
   stride = img->get_image_stride(cIdx);
 
   int xBLuma = xB * SubWidth;
   int yBLuma = yB * SubHeight;
 
-  int currBlockAddr = pps->MinTbAddrZS (xBLuma>>sps->Log2MinTrafoSize) +
+  int currBlockAddr = pps->scan->MinTbAddrZS (xBLuma>>sps->Log2MinTrafoSize) +
                                         (yBLuma>>sps->Log2MinTrafoSize) * sps->PicWidthInTbsY ;
 
 
@@ -549,7 +548,7 @@
   for (int y=nBottom-1 ; y>=0 ; y-=4)
     if (availableLeft)
       {
-        int NBlockAddr = pps->MinTbAddrZS (((xB-1)*SubWidth )>>sps->Log2MinTrafoSize) +
+        int NBlockAddr = pps->scan->MinTbAddrZS (((xB-1)*SubWidth )>>sps->Log2MinTrafoSize) +
                                            (((yB+y)*SubHeight)>>sps->Log2MinTrafoSize)
                                            * sps->PicWidthInTbsY ;
 
@@ -576,7 +575,7 @@
 
   if (availableTopLeft)
     {
-      int NBlockAddr = pps->MinTbAddrZS (((xB-1)*SubWidth )>>sps->Log2MinTrafoSize) +
+      int NBlockAddr = pps->scan->MinTbAddrZS (((xB-1)*SubWidth )>>sps->Log2MinTrafoSize) +
                                          (((yB-1)*SubHeight)>>sps->Log2MinTrafoSize)
                                          * sps->PicWidthInTbsY ;
 
@@ -606,7 +605,7 @@
 
     if (borderAvailable)
       {
-        int NBlockAddr = pps->MinTbAddrZS (((xB+x)*SubWidth )>>sps->Log2MinTrafoSize) +
+        int NBlockAddr = pps->scan->MinTbAddrZS (((xB+x)*SubWidth )>>sps->Log2MinTrafoSize) +
                                            (((yB-1)*SubHeight)>>sps->Log2MinTrafoSize)
                                            * sps->PicWidthInTbsY ;

libde265-1.0.17.tar.gz/libde265/motion.cc -> libde265-1.1.1.tar.gz/libde265/motion.cc Changed

@@ -50,7 +50,7 @@
              const seq_parameter_set* sps, int mv_x, int mv_y,
              int xP,int yP,
              int16_t* out, int out_stride,
-             const pixel_t* ref, int ref_stride,
+             const pixel_t* ref, ptrdiff_t ref_stride,
              int nPbW, int nPbH, int bitDepth_L)
 {
   int xFracL = mv_x & 3;
@@ -63,7 +63,7 @@
 
   //const int shift1 = sps->BitDepth_Y-8;
   //const int shift2 = 6;
-  const int shift3 = 14 - sps->BitDepth_Y;
+  const int shift3 = std::max(2, 14 - sps->BitDepth_Y);
 
   int w = sps->pic_width_in_luma_samples;
   int h = sps->pic_height_in_luma_samples;
@@ -129,7 +129,7 @@
     pixel_t padbuf(MAX_CU_SIZE+16)*(MAX_CU_SIZE+7);
 
     const pixel_t* src_ptr;
-    int src_stride;
+    ptrdiff_t src_stride;
 
     if (-extra_left + xIntOffsL >= 0 &&
         -extra_top  + yIntOffsL >= 0 &&
@@ -181,14 +181,14 @@
                int mv_x, int mv_y,
                int xP,int yP,
                int16_t* out, int out_stride,
-               const pixel_t* ref, int ref_stride,
+               const pixel_t* ref, ptrdiff_t ref_stride,
                int nPbWC, int nPbHC, int bit_depth_C)
 {
   // chroma sample interpolation process (8.5.3.2.2.2)
 
   //const int shift1 = sps->BitDepth_C-8;
   //const int shift2 = 6;
-  const int shift3 = 14 - sps->BitDepth_C;
+  const int shift3 = std::max(2, 14 - sps->BitDepth_C);
 
   int wC = sps->pic_width_in_luma_samples /sps->SubWidthC;
   int hC = sps->pic_height_in_luma_samples/sps->SubHeightC;
@@ -227,7 +227,7 @@
     pixel_t padbuf(MAX_CU_SIZE+16)*(MAX_CU_SIZE+3);
 
     const pixel_t* src_ptr;
-    int src_stride;
+    ptrdiff_t src_stride;
 
     int extra_top  = 1;
     int extra_left = 1;
@@ -453,9 +453,9 @@
 
   // weighted sample prediction  (8.5.3.2.3)
 
-  const int shift1_L = libde265_max(2,14-sps->BitDepth_Y);
+  const int shift1_L = std::max(2,14-sps->BitDepth_Y);
   const int offset_shift1_L = img->get_sps().WpOffsetBdShiftY;
-  const int shift1_C = libde265_max(2,14-sps->BitDepth_C);
+  const int shift1_C = std::max(2,14-sps->BitDepth_C);
   const int offset_shift1_C = img->get_sps().WpOffsetBdShiftC;
 
   /*
@@ -1074,7 +1074,7 @@
     numRefIdx = shdr->num_ref_idx_l0_active;
   }
   else {
-    numRefIdx = libde265_min(shdr->num_ref_idx_l0_active,
+    numRefIdx = std::min(shdr->num_ref_idx_l0_active,
                              shdr->num_ref_idx_l1_active);
   }
 
@@ -1128,12 +1128,12 @@
     return false;
   }
   else {
-    int tx = (16384 + (abs_value(td)>>1)) / td;
+    int tx = (16384 + (std::abs(td)>>1)) / td;
     int distScaleFactor = Clip3(-4096,4095, (tb*tx+32)>>6);
     out_mv->x = Clip3(-32768,32767,
-                      Sign(distScaleFactor*mv.x)*((abs_value(distScaleFactor*mv.x)+127)>>8));
+                      Sign(distScaleFactor*mv.x)*((std::abs(distScaleFactor*mv.x)+127)>>8));
     out_mv->y = Clip3(-32768,32767,
-                      Sign(distScaleFactor*mv.y)*((abs_value(distScaleFactor*mv.y)+127)>>8));
+                      Sign(distScaleFactor*mv.y)*((std::abs(distScaleFactor*mv.y)+127)>>8));
     return true;
   }
 }

libde265-1.0.17.tar.gz/libde265/nal-parser.cc -> libde265-1.1.1.tar.gz/libde265/nal-parser.cc Changed

@@ -24,6 +24,8 @@
 #include <assert.h>
 #include <stdlib.h>
 #include <stdio.h>
+#include <stdint.h>
+#include <limits.h>
 
 #ifdef HAVE_CONFIG_H
 #include "config.h"
@@ -55,7 +57,21 @@
 LIBDE265_CHECK_RESULT bool NAL_unit::resize(int new_size)
 {
   if (capacity < new_size) {
-    unsigned char* newbuffer = static_cast<unsigned char*>(malloc(new_size));
+    // Grow the buffer geometrically (1.5x) rather than to the exact requested
+    // size. NAL_Parser::push_data() appends to the pending NAL one input chunk
+    // at a time, increasing the request by a roughly constant amount each call.
+    // With exact-size allocation every chunk would reallocate and copy the
+    // whole accumulated buffer (O(n^2) for a single oversized NAL); spare
+    // capacity amortizes the total copying to O(n). Here new_size > capacity >= 0,
+    // so the 1.5x term is computed in 64 bits and only used when it both exceeds
+    // the request and still fits in 'int'.
+    int alloc_size = new_size;
+    int64_t grow = static_cast<int64_t>(capacity) + capacity / 2;
+    if (grow > new_size && grow <= INT_MAX) {
+      alloc_size = static_cast<int>(grow);
+    }
+
+    unsigned char* newbuffer = static_cast<unsigned char*>(malloc(alloc_size));
     if (newbuffer == nullptr) {
       return false;
     }
@@ -66,7 +82,7 @@
     }
 
     nal_data = newbuffer;
-    capacity = new_size;
+    capacity = alloc_size;
   }
   return true;
 }
@@ -113,38 +129,36 @@
 
 void NAL_unit::remove_stuffing_bytes()
 {
-  uint8_t* p = data();
-
-  for (int i=0;i<size()-2;i++)
-    {
-#if 0
-        for (int k=i;k<i+64;k++) 
-          if (i*0+k<size()) {
-            printf("%c%02x", (k==i) ? '':' ', data()k);
-          }
-        printf("\n");
-#endif
-
-      if (p2!=3 && p2!=0) {
-        // fast forward 3 bytes (2+1)
-        p+=2;
-        i+=2;
-      }
-      else {
-        if (p0==0 && p1==0 && p2==3) {
-          //printf("SKIP NAL @ %d\n",i+2+num_skipped_bytes);
-          insert_skipped_byte(i+2 + num_skipped_bytes());
-
-          memmove(p+2, p+3, size()-i-3);
-          set_size(size()-1);
+  // Remove emulation-prevention bytes: every 0x03 that immediately follows two
+  // 0x00 bytes is dropped (and the zero-run reset, so 00 00 03 03 keeps the
+  // trailing 03). This is done in a single in-place forward-compaction pass in
+  // O(n) time. A previous implementation called memmove() on the remaining tail
+  // for each removed byte, which is O(n^2) and can be abused by a payload that
+  // is densely packed with 00 00 03 triplets.
+
+  uint8_t* d = data();
+  const int n = size();
+
+  int w = 0;       // write position == length of the compacted output so far
+  int zeros = 0;   // number of consecutive 0x00 bytes already written to output
+
+  for (int r=0; r<n; r++) {
+    uint8_t b = dr;
+
+    if (zeros >= 2 && b == 3) {
+      // 'r' is the position of this byte in the original (uncompacted) NAL,
+      // which equals (compacted position) + num_skipped_bytes() — the value the
+      // previous memmove-based code recorded here.
+      insert_skipped_byte(r);
+      zeros = 0;
+      continue;
+    }
 
-          p++;
-          i++;
-        }
-      }
+    dw++ = b;
+    zeros = (b == 0) ? zeros + 1 : 0;
+  }
 
-      p++;
-    }
+  set_size(w);
 }
 
 
@@ -321,6 +335,14 @@
         }
 #endif
 
+        // enforce the maximum NAL size: drop an oversized NAL and resync
+        if (!nal_size_within_limit(out - nal->data())) {
+          free_NAL_unit(pending_input_NAL);
+          pending_input_NAL = nullptr;
+          input_push_state = 0;
+          return DE265_ERROR_NAL_SIZE_EXCEEDS_SECURITY_LIMIT;
+        }
+
         nal->set_size(out - nal->data());;
 
         // push this NAL decoder queue
@@ -355,6 +377,18 @@
   }
 
   nal->set_size(out - nal->data());
+
+  // Enforce the maximum NAL size on the still-incomplete pending NAL. This bounds
+  // memory when a single NAL grows across many push_data() calls without ever
+  // reaching a start code. The oversized pending NAL is dropped and the parser
+  // resyncs at the next start code.
+  if (!nal_size_within_limit(nal->size())) {
+    free_NAL_unit(pending_input_NAL);
+    pending_input_NAL = nullptr;
+    input_push_state = 0;
+    return DE265_ERROR_NAL_SIZE_EXCEEDS_SECURITY_LIMIT;
+  }
+
   return DE265_OK;
 }
 
@@ -368,6 +402,11 @@
 
   end_of_frame = false;
 
+  // enforce the maximum NAL size to bound memory usage
+  if (!nal_size_within_limit(len)) {
+    return DE265_ERROR_NAL_SIZE_EXCEEDS_SECURITY_LIMIT;
+  }
+
   NAL_unit* nal = alloc_NAL_unit(len);
   if (nal == nullptr || !nal->set_data(data, len)) {
     free_NAL_unit(nal);

libde265-1.0.17.tar.gz/libde265/nal-parser.h -> libde265-1.1.1.tar.gz/libde265/nal-parser.h Changed

@@ -25,6 +25,7 @@
 #include "libde265/pps.h"
 #include "libde265/nal.h"
 #include "libde265/util.h"
+#include "libde265/de265.h"
 
 #include <vector>
 #include <queue>
@@ -90,6 +91,10 @@
   NAL_Parser();
   ~NAL_Parser();
 
+  // Point the parser at the live security limits struct so that runtime
+  // changes (via de265_get_security_limits()) take effect immediately.
+  void set_security_limits(const de265_security_limits* limits) { m_security_limits = limits; }
+
   de265_error push_data(const unsigned char* data, int len,
                         de265_PTS pts, void* user_data = nullptr);
 
@@ -134,6 +139,8 @@
 
   NAL_unit* pending_input_NAL = nullptr;
 
+  const de265_security_limits* m_security_limits = nullptr;
+
 
   // NAL level
 
@@ -142,6 +149,14 @@
 
   void push_to_NAL_queue(NAL_unit*);
 
+  // Returns true if a NAL unit of the given size is within the configured
+  // security limit (or if no limit is set).
+  bool nal_size_within_limit(int64_t nal_size) const {
+    return m_security_limits == nullptr ||
+           m_security_limits->max_NAL_size_bytes == 0 ||
+           nal_size <= m_security_limits->max_NAL_size_bytes;
+  }
+
 
   // pool of unused NAL memory

libde265-1.0.17.tar.gz/libde265/pps.cc -> libde265-1.1.1.tar.gz/libde265/pps.cc Changed

@@ -25,6 +25,9 @@
 #include <assert.h>
 #include <stdlib.h>
 #include <string.h>
+#include <mutex>
+#include <memory>
+#include <atomic>
 #if defined(_MSC_VER) || defined(__MINGW32__)
 # include <malloc.h>
 #elif defined(HAVE_ALLOCA_H)
@@ -62,15 +65,19 @@
   }
 
   cross_component_prediction_enabled_flag = br->get_bits(1);
+  // shall be 0 when ChromaArrayType is not 3 (Sec. 7.4.3.3.2)
   if (sps->ChromaArrayType != CHROMA_444 &&
       cross_component_prediction_enabled_flag) {
       ctx->add_warning(DE265_WARNING_PPS_HEADER_INVALID, false);
+      return false;
   }
 
   chroma_qp_offset_list_enabled_flag = br->get_bits(1);
+  // shall be 0 when ChromaArrayType is 0 (mono) (Sec. 7.4.3.3.2)
   if (sps->ChromaArrayType == CHROMA_MONO &&
       chroma_qp_offset_list_enabled_flag) {
       ctx->add_warning(DE265_WARNING_PPS_HEADER_INVALID, false);
+      return false;
   }
 
   if (chroma_qp_offset_list_enabled_flag) {
@@ -118,7 +125,7 @@
 
   uvlc = br->get_uvlc();
   if (uvlc == UVLC_ERROR ||
-      uvlc > static_cast<uint32_t>(libde265_max(0, sps->BitDepth_Y-10))) {
+      uvlc > static_cast<uint32_t>(std::max(0, sps->BitDepth_Y-10))) {
     ctx->add_warning(DE265_WARNING_PPS_HEADER_INVALID, false);
     return false;
   }
@@ -127,7 +134,7 @@
 
   uvlc = br->get_uvlc();
   if (uvlc == UVLC_ERROR ||
-      uvlc > static_cast<uint32_t>(libde265_max(0, sps->BitDepth_C-10))) {
+      uvlc > static_cast<uint32_t>(std::max(0, sps->BitDepth_C-10))) {
     ctx->add_warning(DE265_WARNING_PPS_HEADER_INVALID, false);
     return false;
   }
@@ -231,11 +238,7 @@
   for (int i=0;i<=DE265_MAX_TILE_COLUMNS;i++) { colBdi=0; }
   for (int i=0;i<=DE265_MAX_TILE_ROWS;i++)    { rowBdi=0; }
 
-  CtbAddrRStoTS.clear();
-  CtbAddrTStoRS.clear();
-  TileId.clear();
-  TileIdRS.clear();
-  MinTbAddrZS.clear();
+  scan.reset();
 
 
   Log2MinCuQpDeltaSize = 0;
@@ -260,6 +263,8 @@
   pps_range_extension_flag = 0;
   pps_multilayer_extension_flag = 0;
   pps_extension_6bits = 0;
+
+  range_extension.reset();
 }
 
 
@@ -312,7 +317,9 @@
 
   {
     int32_t svlc;
-    if ((svlc = br->get_svlc()) == SVLC_ERROR) {
+    // init_qp_minus26 shall be in -(26 + QpBdOffset_Y), +25 (Sec. 7.4.3.3.1)
+    if ((svlc = br->get_svlc()) == SVLC_ERROR ||
+        svlc < -(26 + sps->QpBdOffset_Y) || svlc > 25) {
       ctx->add_warning(DE265_WARNING_PPS_HEADER_INVALID, false);
       return false;
     }
@@ -365,18 +372,20 @@
 
   if (tiles_enabled_flag) {
     if ((uvlc = br->get_uvlc()) == UVLC_ERROR ||
-	uvlc+1 > DE265_MAX_TILE_COLUMNS) {
+        uvlc + 1 > DE265_MAX_TILE_COLUMNS ||
+        uvlc + 1 > sps->PicWidthInCtbsY) {
       ctx->add_warning(DE265_WARNING_PPS_HEADER_INVALID, false);
       return false;
     }
-    num_tile_columns = uvlc+1;
+    num_tile_columns = uvlc + 1;
 
     if ((uvlc = br->get_uvlc()) == UVLC_ERROR ||
-	uvlc+1 > DE265_MAX_TILE_ROWS) {
+        uvlc + 1 > DE265_MAX_TILE_ROWS ||
+        uvlc + 1 > sps->PicHeightInCtbsY) {
       ctx->add_warning(DE265_WARNING_PPS_HEADER_INVALID, false);
       return false;
     }
-    num_tile_rows = uvlc+1;
+    num_tile_rows = uvlc + 1;
 
     uniform_spacing_flag = br->get_bits(1);
 
@@ -384,30 +393,29 @@
       uint16_t lastColumnWidth = sps->PicWidthInCtbsY;
       uint16_t lastRowHeight   = sps->PicHeightInCtbsY;
 
-      for (int i=0; i<num_tile_columns-1; i++)
-        {
-          if ((uvlc = br->get_uvlc()) == UVLC_ERROR ||
-              uvlc >= lastColumnWidth) {
-	    ctx->add_warning(DE265_WARNING_PPS_HEADER_INVALID, false);
-	    return false;
-	  }
-          colWidthi = uvlc+1;
-
-          lastColumnWidth -= colWidthi;
+      for (int i = 0; i < num_tile_columns - 1; i++) {
+        if ((uvlc = br->get_uvlc()) == UVLC_ERROR ||
+            uvlc + 1 >= lastColumnWidth) {
+          ctx->add_warning(DE265_WARNING_PPS_HEADER_INVALID, false);
+          return false;
         }
 
-      colWidthnum_tile_columns-1 = lastColumnWidth;
+        colWidthi = uvlc + 1;
 
-      for (int i=0; i<num_tile_rows-1; i++)
-        {
-          if ((uvlc = br->get_uvlc()) == UVLC_ERROR ||
-              uvlc >= lastRowHeight) {
-	    ctx->add_warning(DE265_WARNING_PPS_HEADER_INVALID, false);
-	    return false;
-	  }
-          rowHeighti = uvlc+1;
-          lastRowHeight -= rowHeighti;
+        lastColumnWidth -= colWidthi;
+      }
+
+      colWidthnum_tile_columns - 1 = lastColumnWidth;
+
+      for (int i = 0; i < num_tile_rows - 1; i++) {
+        if ((uvlc = br->get_uvlc()) == UVLC_ERROR ||
+            uvlc + 1 >= lastRowHeight) {
+          ctx->add_warning(DE265_WARNING_PPS_HEADER_INVALID, false);
+          return false;
         }
+        rowHeighti = uvlc + 1;
+        lastRowHeight -= rowHeighti;
+      }
 
 
       rowHeightnum_tile_rows-1 = lastRowHeight;
@@ -481,7 +489,7 @@
     }
   }
   else {
-    memcpy(&scaling_list, &sps->scaling_list, sizeof(scaling_list_data));
+    scaling_list = sps->scaling_list;
   }
 
 
@@ -514,16 +522,12 @@
       }
     }
 
-    //assert(false);
-    /*
-      while( more_rbsp_data() )
-
-      pps_extension_data_flag
-      u(1)
-      rbsp_trailing_bits()
-
-      }
-    */
+    // Multilayer extension and the 6 reserved extension bits would carry
+    // additional payload that we do not parse. Reject the stream.
+    if (pps_multilayer_extension_flag || pps_extension_6bits) {
+      ctx->add_warning(DE265_ERROR_NOT_IMPLEMENTED_YET, false);
+      return false;
+    }
   }
 
 
@@ -535,6 +539,153 @@
 }
 
 
+//----------------------------------------------------------------------------
+// Library-scope cache for the geometry-derived scan tables (HEVC Sec. 6.5).
+//
+// The tables depend only on the picture/tile geometry. Many independent decoder
+// contexts (e.g. libheif tile grids) decode images of the same geometry, so we
+// compute the tables once and share them read-only via shared_ptr. A small LRU
+// cache (a few distinct geometries) protected by a mutex serves concurrent
+// decoders. The compute is done while holding the lock on purpose: a burst of
+// contexts with the same new geometry then computes the tables exactly once
+// (the others block briefly and pick up the cached result).
+//----------------------------------------------------------------------------
+
+namespace {
+
+struct pps_scan_key {
+  uint8_t  log2CtbSize;
+  uint8_t  log2MinTrafo;
+  uint16_t picWidthInCtbs, picHeightInCtbs;
+  uint16_t picWidthInTbs,  picHeightInTbs;
+  uint32_t picSizeInCtbs,  picSizeInTbs;
+  uint16_t numTileCols,    numTileRows;
+  uint16_t colBdDE265_MAX_TILE_COLUMNS+1;
+  uint16_t rowBdDE265_MAX_TILE_ROWS+1;
+
+  bool operator==(const pps_scan_key& o) const {
+    if (log2CtbSize    != o.log2CtbSize    || log2MinTrafo   != o.log2MinTrafo   ||
+        picWidthInCtbs != o.picWidthInCtbs || picHeightInCtbs!= o.picHeightInCtbs||
+        picWidthInTbs  != o.picWidthInTbs  || picHeightInTbs != o.picHeightInTbs ||
+        picSizeInCtbs  != o.picSizeInCtbs  || picSizeInTbs   != o.picSizeInTbs   ||
+        numTileCols    != o.numTileCols    || numTileRows    != o.numTileRows) return false;
+    for (int i=0;i<=numTileCols;i++) if (colBdi!=o.colBdi) return false;
+    for (int i=0;i<=numTileRows;i++) if (rowBdi!=o.rowBdi) return false;
+    return true;
+  }
+};
+
+// Build the five scan tables from the geometry key (HEVC 6.5.1 + 6.5.2).
+std::shared_ptr<const pps_scan_tables> compute_scan_tables(const pps_scan_key& k)
+{
+  std::shared_ptr<pps_scan_tables> t = std::make_shared<pps_scan_tables>();
+  t->CtbAddrRStoTS.resize(k.picSizeInCtbs);
+  t->CtbAddrTStoRS.resize(k.picSizeInCtbs);
+  t->TileId       .resize(k.picSizeInCtbs);
+  t->TileIdRS     .resize(k.picSizeInCtbs);
+  t->MinTbAddrZS  .resize(k.picSizeInTbs);
+
+  // 6.5.1 raster (RS) <-> tile scan (TS) conversion + tile-ID assignment.
+  uint32_t ctbAddrTS = 0;
+  uint32_t tIdx = 0;
+  for (int tileY=0; tileY<k.numTileRows; tileY++) {
+    for (int tileX=0; tileX<k.numTileCols; tileX++) {
+      for (int y=k.rowBdtileY; y<k.rowBdtileY+1; y++) {
+        for (int x=k.colBdtileX; x<k.colBdtileX+1; x++) {
+          uint32_t ctbAddrRS = y * k.picWidthInCtbs + x;
+          t->CtbAddrRStoTSctbAddrRS = ctbAddrTS;
+          t->CtbAddrTStoRSctbAddrTS = ctbAddrRS;
+          t->TileId  ctbAddrTS = tIdx;
+          t->TileIdRSctbAddrRS = tIdx;
+          ctbAddrTS++;
+        }
+      }
+      tIdx++;
+    }
+  }
+  assert(ctbAddrTS == k.picSizeInCtbs);
+
+  // 6.5.2 Z-scan order array initialization process.
+  const int shift = k.log2CtbSize - k.log2MinTrafo;
+  for (int y=0; y<k.picHeightInTbs; y++)
+    for (int x=0; x<k.picWidthInTbs; x++) {
+      int tbX = (x<<k.log2MinTrafo)>>k.log2CtbSize;
+      int tbY = (y<<k.log2MinTrafo)>>k.log2CtbSize;
+      int ctbAddrRS = k.picWidthInCtbs*tbY + tbX;
+
+      uint32_t v = t->CtbAddrRStoTSctbAddrRS << (shift*2);
+      int p=0;
+      for (int i=0;i<shift;i++) {
+        int m=1<<i;
+        p += (m & x ? m*m : 0) + (m & y ? 2*m*m : 0);
+      }
+      t->MinTbAddrZSx + y*k.picWidthInTbs = v + p;
+    }
+
+  return t;
+}
+
+class pps_scan_cache {
+public:
+  std::shared_ptr<const pps_scan_tables> get(const pps_scan_key& key) {
+    std::lock_guard<std::mutex> lock(mMutex);
+
+    for (size_t i=0; i<mEntries.size(); i++) {
+      if (mEntriesi.key == key) {
+        std::shared_ptr<const pps_scan_tables> tables = mEntriesi.tables;
+        if (i != 0) {  // move-to-front (LRU)
+          Entry e = mEntriesi;
+          mEntries.erase(mEntries.begin()+i);
+          mEntries.insert(mEntries.begin(), e);
+        }
+        return tables;
+      }
+    }
+
+    // Miss: compute while holding the lock so that a burst of concurrent decoders
+    // with the same new geometry computes the tables exactly once.
+    std::shared_ptr<const pps_scan_tables> tables = compute_scan_tables(key);
+    mEntries.insert(mEntries.begin(), Entry{key, tables});
+    if (mEntries.size() > kMaxEntries) mEntries.pop_back();  // evict LRU
+    return tables;
+  }
+
+private:
+  static const size_t kMaxEntries = 3;
+  struct Entry { pps_scan_key key; std::shared_ptr<const pps_scan_tables> tables; };
+  std::mutex mMutex;
+  std::vector<Entry> mEntries;
+};
+
+// Owned by the de265_init()/de265_free() lifecycle (see de265.cc). It is created
+// and destroyed (under de265's init mutex) while no decoder is running, so it is
+// read locklessly during decoding; the cache's own mutex guards concurrent get()
+// calls. Atomic so the publish/read of the pointer is well-defined.
+std::atomic<pps_scan_cache*> g_pps_scan_cache{nullptr};
+
+std::shared_ptr<const pps_scan_tables> get_pps_scan_tables(const pps_scan_key& key)
+{
+  pps_scan_cache* cache = g_pps_scan_cache.load(std::memory_order_acquire);
+  if (cache) return cache->get(key);
+  return compute_scan_tables(key);  // library not initialized: compute without caching
+}
+
+} // namespace
+
+
+void pps_scan_cache_init()
+{
+  if (!g_pps_scan_cache.load(std::memory_order_relaxed)) {
+    g_pps_scan_cache.store(new pps_scan_cache(), std::memory_order_release);
+  }
+}
+
+void pps_scan_cache_free()
+{
+  delete g_pps_scan_cache.exchange(nullptr, std::memory_order_acq_rel);
+}
+
+
 void pic_parameter_set::set_derived_values(const seq_parameter_set* sps)
 {
   Log2MinCuQpDeltaSize = sps->Log2CtbSizeY - diff_cu_qp_delta_depth;
@@ -582,143 +733,28 @@
 
 
 
-  // alloc raster scan arrays
-
-  CtbAddrRStoTS.resize(sps->PicSizeInCtbsY);
-  CtbAddrTStoRS.resize(sps->PicSizeInCtbsY);
-  TileId       .resize(sps->PicSizeInCtbsY);
-  TileIdRS     .resize(sps->PicSizeInCtbsY);
-  MinTbAddrZS  .resize(sps->PicSizeInTbsY );
-
-
-  // raster scan (RS) <-> tile scan (TS) conversion
-
-  for (uint32_t ctbAddrRS=0 ; ctbAddrRS < sps->PicSizeInCtbsY ; ctbAddrRS++)
-    {
-      int tbX = ctbAddrRS % sps->PicWidthInCtbsY;
-      int tbY = ctbAddrRS / sps->PicWidthInCtbsY;
-      int tileX=-1,tileY=-1;
-
-      for (int i=0;i<num_tile_columns;i++)
-        if (tbX >= colBdi)
-          tileX=i;
-
-      for (int j=0;j<num_tile_rows;j++)
-        if (tbY >= rowBdj)
-          tileY=j;
-
-      CtbAddrRStoTSctbAddrRS = 0;
-      for (int i=0;i<tileX;i++)
-        CtbAddrRStoTSctbAddrRS += rowHeighttileY*colWidthi;
-
-      for (int j=0;j<tileY;j++)
-        {
-          //pps->CtbAddrRStoTSctbAddrRS += (tbY - pps->rowBdtileY)*pps->colWidthtileX;
-          //pps->CtbAddrRStoTSctbAddrRS += tbX - pps->colBdtileX;
-
-          CtbAddrRStoTSctbAddrRS += sps->PicWidthInCtbsY * rowHeightj;
-        }
-
-      assert(tileX>=0 && tileY>=0);
-
-      CtbAddrRStoTSctbAddrRS += (tbY-rowBdtileY)*colWidthtileX;
-      CtbAddrRStoTSctbAddrRS +=  tbX - colBdtileX;
-
-
-      // inverse mapping
-
-      CtbAddrTStoRS CtbAddrRStoTSctbAddrRS  = ctbAddrRS;
-    }
-
-
-#if 0
-  logtrace(LogHeaders,"6.5.1 CtbAddrRSToTS\n");
-  for (int y=0;y<sps->PicHeightInCtbsY;y++)
-    {
-      for (int x=0;x<sps->PicWidthInCtbsY;x++)
-        {
-          logtrace(LogHeaders,"%3d ", CtbAddrRStoTSx + y*sps->PicWidthInCtbsY);
-        }
-
-      logtrace(LogHeaders,"\n");
-    }
-#endif
-
-  // tile id
-
-  for (int j=0, tIdx=0 ; j<num_tile_rows ; j++)
-    for (int i=0 ; i<num_tile_columns;i++)
-      {
-        for (int y=rowBdj ; y<rowBdj+1 ; y++)
-          for (int x=colBdi ; x<colBdi+1 ; x++) {
-            TileId   CtbAddrRStoTSy*sps->PicWidthInCtbsY + x  = tIdx;
-            TileIdRS y*sps->PicWidthInCtbsY + x  = tIdx;
-
-            //logtrace(LogHeaders,"tileID%d,%d = %d\n",x,y,pps->TileIdRS y*sps->PicWidthInCtbsY + x );
-          }
-
-        tIdx++;
-      }
-
-#if 0
-  logtrace(LogHeaders,"Tile IDs RS:\n");
-  for (int y=0;y<sps->PicHeightInCtbsY;y++) {
-    for (int x=0;x<sps->PicWidthInCtbsY;x++) {
-      logtrace(LogHeaders,"%2d ",TileIdRSy*sps->PicWidthInCtbsY+x);
-    }
-    logtrace(LogHeaders,"\n");
-  }
-#endif
-
-  // 6.5.2 Z-scan order array initialization process
-
-  for (int y=0;y<sps->PicHeightInTbsY;y++)
-    for (int x=0;x<sps->PicWidthInTbsY;x++)
-      {
-        int tbX = (x<<sps->Log2MinTrafoSize)>>sps->Log2CtbSizeY;
-        int tbY = (y<<sps->Log2MinTrafoSize)>>sps->Log2CtbSizeY;
-        int ctbAddrRS = sps->PicWidthInCtbsY*tbY + tbX;
-
-        MinTbAddrZSx + y*sps->PicWidthInTbsY = CtbAddrRStoTSctbAddrRS
-          << ((sps->Log2CtbSizeY-sps->Log2MinTrafoSize)*2);
-
-        int p=0;
-        for (int i=0 ; i<(sps->Log2CtbSizeY - sps->Log2MinTrafoSize) ; i++) {
-          int m=1<<i;
-          p += (m & x ? m*m : 0) + (m & y ? 2*m*m : 0);
-        }
-
-        MinTbAddrZSx + y*sps->PicWidthInTbsY += p;
-      }
-
-
-  // --- debug logging ---
-
-  /*
-    logtrace(LogHeaders,"6.5.2 Z-scan order array\n");
-    for (int y=0;y<sps->PicHeightInTbsY;y++)
-    {
-    for (int x=0;x<sps->PicWidthInTbsY;x++)
-    {
-    logtrace(LogHeaders,"%4d ", pps->MinTbAddrZSx + y*sps->PicWidthInTbsY);
-    }
-
-    logtrace(LogHeaders,"\n");
-    }
-
-    for (int i=0;i<sps->PicSizeInTbsY;i++)
-    {
-    for (int y=0;y<sps->PicHeightInTbsY;y++)
-    {
-    for (int x=0;x<sps->PicWidthInTbsY;x++)
-    {
-    if (pps->MinTbAddrZSx + y*sps->PicWidthInTbsY == i) {
-    logtrace(LogHeaders,"%d %d\n",x,y);
-    }
-    }
-    }
-    }
-  */
+  // The derived scan tables (Sec. 6.5.1 + 6.5.2) depend only on the picture/tile
+  // geometry computed above. Build the geometry key and fetch the shared tables
+  // from the library-scope cache (computing+caching them on a miss). This avoids
+  // recomputing the (potentially large) MinTbAddrZS table for every decoder
+  // context when many contexts decode images of the same geometry.
+
+  pps_scan_key key;
+  memset(&key, 0, sizeof(key));   // zero padding/unused tile entries for clean compares
+  key.log2CtbSize     = sps->Log2CtbSizeY;
+  key.log2MinTrafo    = sps->Log2MinTrafoSize;
+  key.picWidthInCtbs  = sps->PicWidthInCtbsY;
+  key.picHeightInCtbs = sps->PicHeightInCtbsY;
+  key.picWidthInTbs   = sps->PicWidthInTbsY;
+  key.picHeightInTbs  = sps->PicHeightInTbsY;
+  key.picSizeInCtbs   = sps->PicSizeInCtbsY;
+  key.picSizeInTbs    = sps->PicSizeInTbsY;
+  key.numTileCols     = num_tile_columns;
+  key.numTileRows     = num_tile_rows;
+  for (int i=0;i<=num_tile_columns;i++) key.colBdi = colBdi;
+  for (int i=0;i<=num_tile_rows;   i++) key.rowBdi = rowBdi;
+
+  scan = get_pps_scan_tables(key);
 }
 
 
@@ -731,7 +767,7 @@
   }
   out.write_uvlc(pic_parameter_set_id);
 
-  if (seq_parameter_set_id >= DE265_MAX_PPS_SETS) {
+  if (seq_parameter_set_id >= DE265_MAX_SPS_SETS) {
     errqueue->add_warning(DE265_WARNING_NONEXISTING_SPS_REFERENCED, false);
     return false;
   }

libde265-1.0.17.tar.gz/libde265/pps.h -> libde265-1.1.1.tar.gz/libde265/pps.h Changed

@@ -56,6 +56,26 @@
 };
 
 
+// Picture-geometry-derived scan-order tables (HEVC Sec. 6.5). They depend only
+// on the picture/tile geometry, are immutable after construction, and are shared
+// read-only across pic_parameter_set instances (and across independent decoder
+// contexts) through a small library-scope cache -- see pps.cc. Sharing avoids
+// recomputing the (potentially large) MinTbAddrZS table once per decoder when
+// many contexts decode images of the same geometry (e.g. libheif tile grids).
+struct pps_scan_tables {
+  std::vector<uint32_t> CtbAddrRStoTS; // #CTBs
+  std::vector<uint32_t> CtbAddrTStoRS; // #CTBs
+  std::vector<uint32_t> TileId;        // #CTBs  // index in tile-scan order
+  std::vector<uint32_t> TileIdRS;      // #CTBs  // index in raster-scan order
+  std::vector<uint32_t> MinTbAddrZS;   // #TBs   x + y*PicWidthInTbsY
+};
+
+// Library-scope scan-table cache lifecycle, tied to de265_init() / de265_free()
+// so the cache is released when the library is de-initialized (no leak at exit).
+void pps_scan_cache_init();
+void pps_scan_cache_free();
+
+
 class pic_parameter_set {
 public:
   pic_parameter_set();
@@ -126,7 +146,7 @@
   int8_t tc_offset;    // -12;12
 
   bool pic_scaling_list_data_present_flag;
-  struct scaling_list_data scaling_list; // contains valid data if sps->scaling_list_enabled_flag set
+  scaling_list_data scaling_list; // contains valid data if sps->scaling_list_enabled_flag set
 
   bool lists_modification_present_flag;
   uint8_t log2_parallel_merge_level; // 2 ; log2(max CB size)
@@ -146,16 +166,14 @@
   int Log2MinCuChromaQpOffsetSize;
   int Log2MaxTransformSkipSize;
 
-  int colWidth  DE265_MAX_TILE_COLUMNS ;
-  int rowHeight DE265_MAX_TILE_ROWS ;
-  int colBd     DE265_MAX_TILE_COLUMNS+1 ;
-  int rowBd     DE265_MAX_TILE_ROWS+1 ;
+  uint16_t colWidth  DE265_MAX_TILE_COLUMNS ;
+  uint16_t rowHeight DE265_MAX_TILE_ROWS ;
+  uint16_t colBd     DE265_MAX_TILE_COLUMNS+1 ;
+  uint16_t rowBd     DE265_MAX_TILE_ROWS+1 ;
 
-  std::vector<uint32_t> CtbAddrRStoTS; // #CTBs
-  std::vector<uint32_t> CtbAddrTStoRS; // #CTBs
-  std::vector<uint32_t> TileId;        // #CTBs  // index in tile-scan order
-  std::vector<uint32_t> TileIdRS;      // #CTBs  // index in raster-scan order
-  std::vector<uint32_t> MinTbAddrZS;   // #TBs   x + y*PicWidthInTbsY
+  // Derived scan-order tables (Sec. 6.5), shared read-only via the library-scope
+  // cache. Access as scan->CtbAddrRStoTS..., scan->MinTbAddrZS..., etc.
+  std::shared_ptr<const pps_scan_tables> scan;
 
   void set_derived_values(const seq_parameter_set* sps);
 };

libde265-1.0.17.tar.gz/libde265/quality.cc -> libde265-1.1.1.tar.gz/libde265/quality.cc Changed

libde265-1.0.17.tar.gz/libde265/refpic.cc -> libde265-1.1.1.tar.gz/libde265/refpic.cc Changed

@@ -86,7 +86,7 @@
                                  const seq_parameter_set* sps,
                                  bitreader* br,
                                  ref_pic_set* out_set, // where to store the read set
-                                 int idxRps,  // index of the set to be read
+                                 uint32_t idxRps,  // index of the set to be read
                                  const std::vector<ref_pic_set>& sets, // previously read sets
                                  bool sliceRefPicSet) // is this in the slice header?
 {
@@ -109,16 +109,13 @@
     /* Only for the last ref_pic_set (that's the one coded in the slice header),
        we can specify relative to which reference set we code the set. */
 
-    int delta_idx;
+    uint32_t delta_idx;
     if (sliceRefPicSet) { // idxRps == num_short_term_ref_pic_sets) {
-      delta_idx = vlc = br->get_uvlc();
-      if (vlc==UVLC_ERROR) {
-        return false;
-      }
-
-      if (delta_idx>=idxRps) {
+      vlc = br->get_uvlc();
+      if (vlc==UVLC_ERROR || vlc >= idxRps) {
         return false;
       }
+      delta_idx = vlc;
 
       delta_idx++;
     } else {
@@ -325,6 +322,22 @@
 
   out_set->compute_derived_values();
 
+  // The unused short-term references are all collected into a single PocStFoll array
+  // of MAX_NUM_REF_PICS entries (see decoder_context::process_reference_picture_set).
+  // While each individual list is bounded above, the predicted-RPS construction can
+  // append the current-picture delta to an already-full source set, pushing the
+  // combined count past MAX_NUM_REF_PICS. Reject such sets to avoid an out-of-bounds
+  // write when filling PocStFoll.
+  if (out_set->NumDeltaPocs > MAX_NUM_REF_PICS) {
+    out_set->NumNegativePics = 0;
+    out_set->NumPositivePics = 0;
+    out_set->NumDeltaPocs = 0;
+    out_set->NumPocTotalCurr_shortterm_only = 0;
+
+    errqueue->add_warning(DE265_WARNING_MAX_NUM_REF_PICS_EXCEEDED, false);
+    return false;
+  }
+
   return true;
 }
 
@@ -333,7 +346,7 @@
                                          const seq_parameter_set* sps,
                                          CABAC_encoder& out,
                                          const ref_pic_set* in_set, // which set to write
-                                         int idxRps,  // index of the set to be written
+                                         uint32_t idxRps,  // index of the set to be written
                                          const std::vector<ref_pic_set>& sets, // previously read sets
                                          bool sliceRefPicSet) // is this in the slice header?
 {
@@ -384,7 +397,7 @@
                                   const seq_parameter_set* sps,
                                   CABAC_encoder& out,
                                   const ref_pic_set* in_set, // which set to write
-                                  int idxRps,  // index of the set to be read
+                                  uint32_t idxRps,  // index of the set to be read
                                   const std::vector<ref_pic_set>& sets, // previously read sets
                                   bool sliceRefPicSet) // is this in the slice header?
 {

libde265-1.0.17.tar.gz/libde265/sao.cc -> libde265-1.1.1.tar.gz/libde265/sao.cc Changed

@@ -28,8 +28,8 @@
 template <class pixel_t>
 void apply_sao_internal(de265_image* img, int xCtb,int yCtb,
                         const slice_segment_header* shdr, int cIdx, int nSW,int nSH,
-                        const pixel_t* in_img,  int in_stride,
-                        /* */ pixel_t* out_img, int out_stride)
+                        const pixel_t* in_img,  ptrdiff_t in_stride,
+                        /* */ pixel_t* out_img, ptrdiff_t out_stride)
 {
   const sao_info* saoinfo = img->get_sao_info(xCtb,yCtb);
 
@@ -77,7 +77,7 @@
 
   if (SaoTypeIdx==2) {
     int hPos2, vPos2;
-    int vPosStride2; // vPos multiplied by image stride
+    ptrdiff_t vPosStride2; // vPos multiplied by image stride
     int SaoEoClass = (saoinfo->SaoEoClass >> (2*cIdx)) & 0x3;
 
     switch (SaoEoClass) {
@@ -156,8 +156,8 @@
 
 
             if (pps->loop_filter_across_tiles_enabled_flag==0 &&
-                pps->TileIdRS(xS>>ctbshiftW) + (yS>>ctbshiftH)*picWidthInCtbs !=
-                pps->TileIdRS(xC>>ctbshiftW) + (yC>>ctbshiftH)*picWidthInCtbs) {
+                pps->scan->TileIdRS(xS>>ctbshiftW) + (yS>>ctbshiftH)*picWidthInCtbs !=
+                pps->scan->TileIdRS(xC>>ctbshiftW) + (yC>>ctbshiftH)*picWidthInCtbs) {
               edgeIdx=0;
               break;
             }
@@ -266,8 +266,8 @@
 template <class pixel_t>
 void apply_sao(de265_image* img, int xCtb,int yCtb,
                const slice_segment_header* shdr, int cIdx, int nSW,int nSH,
-               const pixel_t* in_img,  int in_stride,
-               /* */ pixel_t* out_img, int out_stride)
+               const pixel_t* in_img,  ptrdiff_t in_stride,
+               /* */ pixel_t* out_img, ptrdiff_t out_stride)
 {
   if (img->high_bit_depth(cIdx)) {
     apply_sao_internal<uint16_t>(img,xCtb,yCtb, shdr,cIdx,nSW,nSH,
@@ -332,10 +332,10 @@
     return;
   }
 
-  int lumaImageSize   = img->get_image_stride(0) * img->get_height(0) * img->get_bytes_per_pixel(0);
-  int chromaImageSize = img->get_image_stride(1) * img->get_height(1) * img->get_bytes_per_pixel(1);
+  size_t lumaImageSize   = static_cast<size_t>(img->get_image_stride(0)) * img->get_height(0) * img->get_bytes_per_pixel(0);
+  size_t chromaImageSize = static_cast<size_t>(img->get_image_stride(1)) * img->get_height(1) * img->get_bytes_per_pixel(1);
 
-  uint8_t* inputCopy = new uint8_t libde265_max(lumaImageSize, chromaImageSize) ;
+  uint8_t* inputCopy = new uint8_t std::max(lumaImageSize, chromaImageSize) ;
   if (inputCopy == nullptr) {
     img->decctx->add_warning(DE265_WARNING_CANNOT_APPLY_SAO_OUT_OF_MEMORY,false);
     return;
@@ -347,10 +347,10 @@
 
   for (int cIdx=0;cIdx<nChannels;cIdx++) {
 
-    int stride = img->get_image_stride(cIdx);
+    ptrdiff_t stride = img->get_image_stride(cIdx);
     int height = img->get_height(cIdx);
 
-    memcpy(inputCopy, img->get_image_plane(cIdx), stride * height * img->get_bytes_per_pixel(cIdx));
+    memcpy(inputCopy, img->get_image_plane(cIdx), static_cast<size_t>(stride) * height * img->get_bytes_per_pixel(cIdx));
 
     for (int yCtb=0; yCtb<sps.PicHeightInCtbsY; yCtb++)
       for (int xCtb=0; xCtb<sps.PicWidthInCtbsY; xCtb++)

libde265-1.0.17.tar.gz/libde265/scan.cc -> libde265-1.1.1.tar.gz/libde265/scan.cc Changed

@@ -105,30 +105,25 @@
   return scanposscanIdxlog2BlkSize y*(1<<log2BlkSize) + x ;
 }
 
-static void fill_scan_pos(scan_position* pos, int x,int y,int scanIdx, int log2TrafoSize)
+static void fill_scan_pos_table(scan_position* pos, int scanIdx, int log2TrafoSize)
 {
-  int lastScanPos = 16;
-  int lastSubBlock = (1<<(log2TrafoSize-2)) * (1<<(log2TrafoSize-2)) -1;
+  int numSubBlocks = (1<<(log2TrafoSize-2)) * (1<<(log2TrafoSize-2));
+  int blkSize = 1 << log2TrafoSize;
 
   const position* ScanOrderSub = get_scan_order(log2TrafoSize-2, scanIdx);
   const position* ScanOrderPos = get_scan_order(2, scanIdx);
 
-  int xC,yC;
-  do {
-    if (lastScanPos==0) {
-      lastScanPos=16;
-      lastSubBlock--;
+  for (int sb = 0; sb < numSubBlocks; sb++)
+    {
+      position S = ScanOrderSubsb;
+      for (int sp = 0; sp < 16; sp++)
+        {
+          int xC = (S.x<<2) + ScanOrderPossp.x;
+          int yC = (S.y<<2) + ScanOrderPossp.y;
+          posyC * blkSize + xC.subBlock = sb;
+          posyC * blkSize + xC.scanPos  = sp;
+        }
     }
-    lastScanPos--;
-
-    position S = ScanOrderSublastSubBlock;
-    xC = (S.x<<2) + ScanOrderPoslastScanPos.x;
-    yC = (S.y<<2) + ScanOrderPoslastScanPos.y;
-
-  } while ( (xC != x) || (yC != y));
-
-  pos->subBlock = lastSubBlock;
-  pos->scanPos  = lastScanPos;
 }
 
 
@@ -141,12 +136,7 @@
       init_scan_d(scan_dlog2size, 1<<log2size);
     }
 
-
   for (int log2size=2;log2size<=5;log2size++)
     for (int scanIdx=0;scanIdx<3;scanIdx++)
-      for (int y=0;y<(1<<log2size);y++)
-        for (int x=0;x<(1<<log2size);x++)
-          {
-            fill_scan_pos(&scanposscanIdxlog2size y*(1<<log2size) + x ,x,y,scanIdx,log2size);
-          }
+      fill_scan_pos_table(scanposscanIdxlog2size, scanIdx, log2size);
 }

libde265-1.0.17.tar.gz/libde265/sei.cc -> libde265-1.1.1.tar.gz/libde265/sei.cc Changed

@@ -99,7 +99,7 @@
 class raw_hash_data
 {
 public:
-  raw_hash_data(int w, int stride);
+  raw_hash_data(int w, ptrdiff_t stride);
   ~raw_hash_data();
 
   struct data_chunk {
@@ -111,13 +111,14 @@
   data_chunk prepare_16bit(const uint8_t* data,int y);
 
 private:
-  int mWidth, mStride;
+  int mWidth;
+  ptrdiff_t mStride;
 
   uint8_t* mMem;
 };
 
 
-raw_hash_data::raw_hash_data(int w, int stride)
+raw_hash_data::raw_hash_data(int w, ptrdiff_t stride)
 {
   mWidth=w;
   mStride=stride;
@@ -157,7 +158,7 @@
 }
 
 
-static uint32_t compute_checksum(uint8_t* data,int w,int h,int stride, int bit_depth)
+static uint32_t compute_checksum(uint8_t* data,int w,int h,ptrdiff_t stride, int bit_depth)
 {
   uint32_t sum = 0;
 
@@ -170,7 +171,7 @@
   }
   else {
     auto* data16 = reinterpret_cast<uint16_t*>(data);
-    int stride16 = stride / 2;
+    ptrdiff_t stride16 = stride / 2;
     for (int y=0; y<h; y++)
       for(int x=0; x<w; x++) {
         uint8_t xorMask = ( x & 0xFF ) ^ ( y & 0xFF ) ^ ( x  >>  8 ) ^ ( y  >>  8 );
@@ -224,7 +225,7 @@
 	   (t << 12)) & 0xFFFF;
 }
 
-static uint32_t compute_CRC_8bit_fast(const uint8_t* data,int w,int h,int stride, int bit_depth)
+static uint32_t compute_CRC_8bit_fast(const uint8_t* data,int w,int h,ptrdiff_t stride, int bit_depth)
 {
   raw_hash_data raw_data(w,stride);
 
@@ -250,7 +251,7 @@
 }
 
 
-static void compute_MD5(uint8_t* data,int w,int h,int stride, uint8_t* result, int bit_depth)
+static void compute_MD5(uint8_t* data,int w,int h,ptrdiff_t stride, uint8_t* result, int bit_depth)
 {
   MD5_CTX md5;
   MD5_Init(&md5);
@@ -289,7 +290,8 @@
   int nHashes = img->get_sps().chroma_format_idc==0 ? 1 : 3;
   for (int i=0;i<nHashes;i++) {
     uint8_t* data;
-    int w,h,stride;
+    int w,h;
+    ptrdiff_t stride;
 
     w = img->get_width(i);
     h = img->get_height(i);

libde265-1.0.17.tar.gz/libde265/slice.cc -> libde265-1.1.1.tar.gz/libde265/slice.cc Changed

@@ -42,7 +42,7 @@
                                         const seq_parameter_set* sps,
                                         bitreader* br,
                                         ref_pic_set* out_set,
-                                        int idxRps, // index of the set to be read
+                                        uint32_t idxRps, // index of the set to be read
                                         const std::vector<ref_pic_set>& sets,
                                         bool sliceRefPicSet);
 
@@ -360,6 +360,9 @@
 
   RemoveReferencesList.clear();
 
+  for (int i = 0; i < 4; i++) {
+    ctx_model_storage_StatCoeffi = 0;
+  }
   ctx_model_storage_defined = false;
 }
 
@@ -851,7 +854,7 @@
       }
       offset_len = uvlc + 1;
 
-      for (int i = 0; i < num_entry_point_offsets; i++) {
+      for (uint32_t i = 0; i < num_entry_point_offsets; i++) {
         {
           uint32_t offset_minus1 = br->get_bits(offset_len);
           if (offset_minus1 == UINT32_MAX) {
@@ -1235,7 +1238,7 @@
     if (num_entry_point_offsets > 0) {
       out.write_uvlc(offset_len - 1);
 
-      for (int i = 0; i < num_entry_point_offsets; i++) {
+      for (uint32_t i = 0; i < num_entry_point_offsets; i++) {
         {
           uint32_t prev = 0;
           if (i > 0) prev = entry_point_offseti - 1;
@@ -1505,7 +1508,7 @@
     if (num_entry_point_offsets > 0) {
       LOG1("offset_len                 : %d\n", offset_len);
 
-      for (int i = 0; i < num_entry_point_offsets; i++) {
+      for (uint32_t i = 0; i < num_entry_point_offsets; i++) {
         LOG2("entry point %i : %d\n", i, entry_point_offseti);
       }
     }
@@ -1599,7 +1602,7 @@
 static uint8_t decode_sao_offset_abs(thread_context* tctx, int bitDepth)
 {
   logtrace(LogSlice, "# sao_offset_abs\n");
-  int cMax = (1 << (libde265_min(bitDepth, 10) - 5)) - 1;
+  int cMax = (1 << (std::min(bitDepth, 10) - 5)) - 1;
   assert(cMax >= 7 && cMax<=31);
   uint8_t value = static_cast<uint8_t>(tctx->cabac_decoder.decode_TU_bypass( cMax));
   logtrace(LogSymbols, "$1 sao_offset_abs=%d\n", value);
@@ -2474,12 +2477,21 @@
 
 #define MAX_PREFIX (15+3)
 
+// Defensive bounds against non-conforming bitstreams. The spec (eq. 9-25 / 9-23) does
+// not impose an explicit upper bound on cRiceParam or StatCoeff in the persistent-rice
+// path, but a malformed stream can push them arbitrarily high. We clamp so that the
+// signed-int shift expressions in residual_coding stay well-defined:
+//   - 3 * (1 << uiGoRiceParam) requires uiGoRiceParam <= 29 (else int32 overflow)
+//   - 3 << (StatCoeff/4)       requires StatCoeff/4   <= 29 (same)
+#define MAX_RICE_PARAM 29
+#define MAX_STAT_COEFF (4 * MAX_RICE_PARAM + 3)  // 119: largest value with /4 <= 29
+
 static int32_t decode_coeff_abs_level_remaining(thread_context* tctx,
                                                 int cRiceParam)
 {
   logtrace(LogSlice, "# decode_coeff_abs_level_remaining\n");
 
-  uint16_t prefix = 0;
+  uint32_t prefix = 0;
   while (tctx->cabac_decoder.decode_bypass()) {
     prefix++;
     if (prefix > MAX_PREFIX) {
@@ -2502,7 +2514,7 @@
     // included in the 'prefix' counter above.
 
     int codeword = tctx->cabac_decoder.decode_FL_bypass( prefix - 3 + cRiceParam);
-    value = (((UINT16_C(1) << (prefix - 3)) + 3 - 1) << cRiceParam) + codeword;
+    value = (((UINT32_C(1) << (prefix - 3)) + 3 - 1) << cRiceParam) + codeword;
   }
 
   logtrace(LogSymbols, "$1 coeff_abs_level_remaining=%d\n", value);
@@ -2696,7 +2708,7 @@
   const seq_parameter_set& sps = tctx->img->get_sps();
 
   if (tctx->CtbAddrInTS < sps.PicSizeInCtbsY) {
-    tctx->CtbAddrInRS = tctx->img->get_pps().CtbAddrTStoRStctx->CtbAddrInTS;
+    tctx->CtbAddrInRS = tctx->img->get_pps().scan->CtbAddrTStoRStctx->CtbAddrInTS;
 
     tctx->CtbX = tctx->CtbAddrInRS % sps.PicWidthInCtbsY;
     tctx->CtbY = tctx->CtbAddrInRS / sps.PicWidthInCtbsY;
@@ -2741,8 +2753,8 @@
   if (xCtb > 0) {
     //char leftCtbInSliceSeg = (CtbAddrInSliceSeg>0);
     char leftCtbInSliceSeg = (tctx->CtbAddrInRS > shdr->SliceAddrRS);
-    char leftCtbInTile = (pps.TileIdRSxCtb + yCtb * sps.PicWidthInCtbsY ==
-                          pps.TileIdRSxCtb - 1 + yCtb * sps.PicWidthInCtbsY);
+    char leftCtbInTile = (pps.scan->TileIdRSxCtb + yCtb * sps.PicWidthInCtbsY ==
+                          pps.scan->TileIdRSxCtb - 1 + yCtb * sps.PicWidthInCtbsY);
 
     if (leftCtbInSliceSeg && leftCtbInTile) {
       sao_merge_left_flag = decode_sao_merge_flag(tctx);
@@ -2756,8 +2768,8 @@
              sps.PicWidthInCtbsY,
              shdr->slice_segment_address);
     bool upCtbInSliceSeg = (tctx->CtbAddrInRS - sps.PicWidthInCtbsY) >= shdr->SliceAddrRS;
-    bool upCtbInTile = (pps.TileIdRSxCtb + yCtb * sps.PicWidthInCtbsY ==
-                        pps.TileIdRSxCtb + (yCtb - 1) * sps.PicWidthInCtbsY);
+    bool upCtbInTile = (pps.scan->TileIdRSxCtb + yCtb * sps.PicWidthInCtbsY ==
+                        pps.scan->TileIdRSxCtb + (yCtb - 1) * sps.PicWidthInCtbsY);
 
     if (upCtbInSliceSeg && upCtbInTile) {
       sao_merge_up_flag = decode_sao_merge_flag(tctx);
@@ -2888,7 +2900,7 @@
 }
 
 
-LIBDE265_INLINE static int luma_pos_to_ctbAddrRS(const seq_parameter_set* sps, int x, int y)
+inline static int luma_pos_to_ctbAddrRS(const seq_parameter_set* sps, int x, int y)
 {
   int ctbX = x >> sps->Log2CtbSizeY;
   int ctbY = y >> sps->Log2CtbSizeY;
@@ -2919,8 +2931,8 @@
 
   // check if both CTBs are in the same tile.
 
-  if (img->get_pps().TileIdRScurrent_ctbAddrRS !=
-      img->get_pps().TileIdRSneighbor_ctbAddrRS) {
+  if (img->get_pps().scan->TileIdRScurrent_ctbAddrRS !=
+      img->get_pps().scan->TileIdRSneighbor_ctbAddrRS) {
     return 0;
   }
 
@@ -3258,7 +3270,7 @@
 
       int newLastGreater1ScanPos = -1;
 
-      int lastGreater1Coefficient = libde265_min(8, nCoefficients);
+      int lastGreater1Coefficient = std::min(8, nCoefficients);
       for (int c = 0; c < lastGreater1Coefficient; c++) {
         int greater1_flag =
             decode_coeff_abs_level_greater1(tctx, cIdx, i,
@@ -3372,15 +3384,24 @@
             }
           }
           else {
-            if (baseLevel + coeff_abs_level_remaining > 3 * (1 << uiGoRiceParam))
+            if (baseLevel + coeff_abs_level_remaining > 3 * (1 << uiGoRiceParam)) {
               uiGoRiceParam++;
+              if (uiGoRiceParam > MAX_RICE_PARAM) {
+                uiGoRiceParam = MAX_RICE_PARAM;
+                tctx->decctx->add_warning(DE265_WARNING_RICE_PARAMETER_OUT_OF_RANGE, true);
+              }
+            }
           }
 
           // persistent_rice_adaptation_enabled_flag
           if (sps.range_extension.persistent_rice_adaptation_enabled_flag &&
               firstCoeffWithAbsLevelRemaining) {
             if (coeff_abs_level_remaining >= (3 << (tctx->StatCoeffsbType / 4))) {
-              tctx->StatCoeffsbType++;
+              if (tctx->StatCoeffsbType < MAX_STAT_COEFF) {
+                tctx->StatCoeffsbType++;
+              } else {
+                tctx->decctx->add_warning(DE265_WARNING_RICE_PARAMETER_OUT_OF_RANGE, true);
+              }
             }
             else if (2 * coeff_abs_level_remaining < (1 << (tctx->StatCoeffsbType / 4)) &&
                      tctx->StatCoeffsbType > 0) {
@@ -3581,7 +3602,7 @@
   const int ChromaArrayType = sps.ChromaArrayType;
 
   int log2TrafoSizeC = (ChromaArrayType == CHROMA_444 ? log2TrafoSize : log2TrafoSize - 1);
-  log2TrafoSizeC = libde265_max(2, log2TrafoSizeC);
+  log2TrafoSizeC = std::max(2, log2TrafoSizeC);
 
   const int cbfLuma = cbf_luma;
   const int cbfChroma = cbf_cb | cbf_cr;
@@ -4739,6 +4760,11 @@
       // copy CABAC model from previous CTB row
       tctx->ctx_model = tctx->imgunit->ctx_models(tctx->CtbY - 1);
       tctx->imgunit->ctx_models(tctx->CtbY - 1).release(); // not used anymore
+
+      // also restore the StatCoeff state for persistent_rice_adaptation
+      for (int i = 0; i < 4; i++) {
+        tctx->StatCoeffi = tctx->imgunit->StatCoeff_models(tctx->CtbY - 1)i;
+      }
     }
     else {
       tctx->img->wait_for_progress(tctx->task, 0, tctx->CtbY - 1,CTB_PROGRESS_PREFILTER);
@@ -4751,7 +4777,7 @@
     const uint32_t ctbx = tctx->CtbX;
     const uint32_t ctby = tctx->CtbY;
 
-    if (ctbx + ctby * ctbW >= pps.CtbAddrRStoTS.size()) {
+    if (ctbx + ctby * ctbW >= pps.scan->CtbAddrRStoTS.size()) {
       return Decode_Error;
     }
 
@@ -4792,6 +4818,11 @@
 
       tctx->imgunit->ctx_modelsctby = tctx->ctx_model;
       tctx->imgunit->ctx_modelsctby.decouple(); // store an independent copy
+
+      // also save the StatCoeff state for persistent_rice_adaptation
+      for (int i = 0; i < 4; i++) {
+        tctx->imgunit->StatCoeff_modelsctbyi = tctx->StatCoeffi;
+      }
     }
 
 
@@ -4808,6 +4839,11 @@
         tctx->shdr->ctx_model_storage = tctx->ctx_model;
         tctx->shdr->ctx_model_storage.decouple(); // store an independent copy
 
+        // also save the StatCoeff state for persistent_rice_adaptation
+        for (int i = 0; i < 4; i++) {
+          tctx->shdr->ctx_model_storage_StatCoeffi = tctx->StatCoeffi;
+        }
+
         tctx->shdr->ctx_model_storage_defined = true;
       }
     }
@@ -4851,7 +4887,7 @@
     if (!end_of_slice_segment_flag) {
       bool end_of_sub_stream = false;
       end_of_sub_stream |= (pps.tiles_enabled_flag &&
-                            pps.TileIdtctx->CtbAddrInTS != pps.TileIdtctx->CtbAddrInTS - 1);
+                            pps.scan->TileIdtctx->CtbAddrInTS != pps.scan->TileIdtctx->CtbAddrInTS - 1);
       end_of_sub_stream |= (pps.entropy_coding_sync_enabled_flag &&
                             lastCtbY != tctx->CtbY);
 
@@ -4879,7 +4915,7 @@
   slice_segment_header* shdr = tctx->shdr;
 
   if (shdr->dependent_slice_segment_flag) {
-    int prevCtb = pps.CtbAddrTStoRSpps.CtbAddrRStoTSshdr->slice_segment_address - 1;
+    int prevCtb = pps.scan->CtbAddrTStoRSpps.scan->CtbAddrRStoTSshdr->slice_segment_address - 1;
 
     uint16_t sliceIdx = img->get_SliceHeaderIndex_atIndex(prevCtb);
     if (sliceIdx >= img->slices.size()) {
@@ -4920,6 +4956,11 @@
 
       tctx->ctx_model = prevCtbHdr->ctx_model_storage;
       prevCtbHdr->ctx_model_storage.release();
+
+      // also restore the StatCoeff state for persistent_rice_adaptation
+      for (int i = 0; i < 4; i++) {
+        tctx->StatCoeffi = prevCtbHdr->ctx_model_storage_StatCoeffi;
+      }
     }
   }
   else {

libde265-1.0.17.tar.gz/libde265/slice.h -> libde265-1.1.1.tar.gz/libde265/slice.h Changed

@@ -218,8 +218,8 @@
 
   bool slice_loop_filter_across_slices_enabled_flag;
 
-  int  num_entry_point_offsets;
-  int  offset_len;
+  uint32_t num_entry_point_offsets;
+  int      offset_len;
   std::vector<uint32_t> entry_point_offset;
 
   int  slice_segment_header_extension_length;
@@ -255,6 +255,8 @@
 
   // context storage for dependent slices (stores CABAC model at end of slice segment)
   context_model_table ctx_model_storage;
+  // StatCoeff (persistent_rice_adaptation state) saved alongside ctx_model_storage
+  uint8_t ctx_model_storage_StatCoeff4;
   bool ctx_model_storage_defined; // whether there is valid data in ctx_model_storage
 
   std::vector<int> RemoveReferencesList; // images that can be removed from the DPB before decoding this slice
@@ -286,7 +288,7 @@
 {
 public:
   bool   firstSliceSubstream;
-  int    debug_startCtbRow;
+  uint16_t debug_startCtbRow;
   thread_context* tctx;
 
   void work() override;
@@ -297,7 +299,7 @@
 {
 public:
   bool   firstSliceSubstream;
-  int    debug_startCtbX, debug_startCtbY;
+  uint16_t debug_startCtbX, debug_startCtbY;
   thread_context* tctx;
 
   void work() override;

libde265-1.0.17.tar.gz/libde265/sps.cc -> libde265-1.1.1.tar.gz/libde265/sps.cc Changed

@@ -50,7 +50,7 @@
                                         const seq_parameter_set* sps,
                                         bitreader* br,
                                         ref_pic_set* out_set,
-                                        int idxRps,  // index of the set to be read
+                                        uint32_t idxRps,  // index of the set to be read
                                         const std::vector<ref_pic_set>& sets,
                                         bool sliceRefPicSet);
 
@@ -58,7 +58,7 @@
                                          const seq_parameter_set* sps,
                                          CABAC_encoder& out,
                                          const ref_pic_set* in_set, // which set to write
-                                         int idxRps,  // index of the set to be read
+                                         uint32_t idxRps,  // index of the set to be read
                                          const std::vector<ref_pic_set>& sets, // previously read sets
                                          bool sliceRefPicSet); // is this in the slice header?
 
@@ -570,7 +570,7 @@
   PicHeightInCtbsY   = ceil_div(pic_height_in_luma_samples,CtbSizeY);
   PicSizeInMinCbsY   = PicWidthInMinCbsY * PicHeightInMinCbsY;
   PicSizeInCtbsY = PicWidthInCtbsY * PicHeightInCtbsY;
-  PicSizeInSamplesY = pic_width_in_luma_samples * pic_height_in_luma_samples;
+  PicSizeInSamplesY = static_cast<uint32_t>(pic_width_in_luma_samples) * pic_height_in_luma_samples;
 
   if (chroma_format_idc==0 || separate_colour_plane_flag) {
     CtbWidthC  = 0;
@@ -583,6 +583,7 @@
 
   Log2MinTrafoSize = log2_min_transform_block_size;
   Log2MaxTrafoSize = log2_min_transform_block_size + log2_diff_max_min_transform_block_size;
+  assert(Log2MaxTrafoSize >= 2);  // log2_min_transform_block_size >= 2 by spec; relied on by pps.cc
 
   if (max_transform_hierarchy_depth_inter > Log2CtbSizeY - Log2MinTrafoSize) {
     if (sanitize_values) {
@@ -656,7 +657,7 @@
     return DE265_ERROR_CODED_PARAMETER_OUT_OF_RANGE;
   }
 
-  if (Log2MaxTrafoSize > libde265_min(Log2CtbSizeY,5)) {
+  if (Log2MaxTrafoSize > std::min((int)Log2CtbSizeY,5)) {
     if (D) fprintf(stderr,"SPS error: TB_max > 32 or CTB\n");
     return DE265_ERROR_CODED_PARAMETER_OUT_OF_RANGE;
   }
@@ -1324,7 +1325,7 @@
   PicHeightInCtbsY   = ceil_div(pic_height_in_luma_samples,CtbSizeY);
   PicSizeInMinCbsY   = PicWidthInMinCbsY * PicHeightInMinCbsY;
   PicSizeInCtbsY = PicWidthInCtbsY * PicHeightInCtbsY;
-  PicSizeInSamplesY = pic_width_in_luma_samples * pic_height_in_luma_samples;
+  PicSizeInSamplesY = static_cast<uint32_t>(pic_width_in_luma_samples) * pic_height_in_luma_samples;
   if (chroma_format_idc==0 || separate_colour_plane_flag) {
     CtbWidthC  = 0;
     CtbHeightC = 0;
@@ -1335,6 +1336,7 @@
   }
   Log2MinTrafoSize = log2_min_transform_block_size;
   Log2MaxTrafoSize = log2_min_transform_block_size + log2_diff_max_min_transform_block_size;
+  assert(Log2MaxTrafoSize >= 2);  // log2_min_transform_block_size >= 2 by spec; relied on by pps.cc
   Log2MinPUSize = Log2MinCbSizeY-1;
   PicWidthInMinPUs  = PicWidthInCtbsY  << (Log2CtbSizeY - Log2MinPUSize);
   PicHeightInMinPUs = PicHeightInCtbsY << (Log2CtbSizeY - Log2MinPUSize);

libde265-1.0.17.tar.gz/libde265/sps.h -> libde265-1.1.1.tar.gz/libde265/sps.h Changed

@@ -39,6 +39,13 @@
 constexpr int MAX_PICTURE_WIDTH  = 65535;
 constexpr int MAX_PICTURE_HEIGHT = 65535;
 
+// pic_width/height_in_luma_samples are stored as uint16_t and PicSizeInSamplesY as uint32_t,
+// so these limits must keep width/height in 16 bits and their product in 32 bits.
+static_assert(MAX_PICTURE_WIDTH  <= 0xFFFF, "picture width must fit in uint16_t");
+static_assert(MAX_PICTURE_HEIGHT <= 0xFFFF, "picture height must fit in uint16_t");
+static_assert((uint64_t)MAX_PICTURE_WIDTH * MAX_PICTURE_HEIGHT <= 0xFFFFFFFFu,
+              "total luma sample count must fit in uint32_t");
+
 enum {
   CHROMA_MONO = 0,
   CHROMA_420 = 1,
@@ -111,8 +118,8 @@
   uint8_t chroma_format_idc;          // 0;3
 
   bool separate_colour_plane_flag;
-  int  pic_width_in_luma_samples;
-  int  pic_height_in_luma_samples;
+  uint16_t pic_width_in_luma_samples;   // <= MAX_PICTURE_WIDTH  (validated on parse)
+  uint16_t pic_height_in_luma_samples;  // <= MAX_PICTURE_HEIGHT (validated on parse)
   bool conformance_window_flag;
 
   int conf_win_left_offset;

libde265-1.0.17.tar.gz/libde265/transform.cc -> libde265-1.1.1.tar.gz/libde265/transform.cc Changed

@@ -110,7 +110,7 @@
   if (tctx->img->available_zscan(xQG,yQG, xQG-1,yQG)) {
     int xTmp = (xQG-1) >> sps.Log2MinTrafoSize;
     int yTmp = (yQG  ) >> sps.Log2MinTrafoSize;
-    int minTbAddrA = pps.MinTbAddrZSxTmp + yTmp*sps.PicWidthInTbsY;
+    int minTbAddrA = pps.scan->MinTbAddrZSxTmp + yTmp*sps.PicWidthInTbsY;
     uint32_t ctbAddrA = minTbAddrA >> (2 * (sps.Log2CtbSizeY-sps.Log2MinTrafoSize));
     if (ctbAddrA == tctx->CtbAddrInTS) {
       qPYA = tctx->img->get_QPY(xQG-1,yQG);
@@ -126,7 +126,7 @@
   if (tctx->img->available_zscan(xQG,yQG, xQG,yQG-1)) {
     int xTmp = (xQG  ) >> sps.Log2MinTrafoSize;
     int yTmp = (yQG-1) >> sps.Log2MinTrafoSize;
-    uint32_t minTbAddrB = pps.MinTbAddrZSxTmp + yTmp*sps.PicWidthInTbsY;
+    uint32_t minTbAddrB = pps.scan->MinTbAddrZSxTmp + yTmp*sps.PicWidthInTbsY;
     uint32_t ctbAddrB = minTbAddrB >> (2 * (sps.Log2CtbSizeY-sps.Log2MinTrafoSize));
     if (ctbAddrB == tctx->CtbAddrInTS) {
       qPYB = tctx->img->get_QPY(xQG,yQG-1);
@@ -467,19 +467,23 @@
       const int offset = (1<<(bdShift-1));
       const int fact = m_x_y * levelScaleqP%6 << (qP/6);
 
-      for (int i=0;i<tctx->nCoeffcIdx;i++) {
-
-        int64_t currCoeff  = tctx->coeffListcIdxi;
-
-        //logtrace(LogTransform,"coefficient%d = %d\n",tctx->coeffPoscIdxi,
-        //tctx->coeffListcIdxi);
-
-        currCoeff = Clip3(-32768,32767,
-                          ( (currCoeff * fact + offset ) >> bdShift));
-
-        //logtrace(LogTransform," -> %d\n",currCoeff);
-
-        tctx->coeffBuf tctx->coeffPoscIdxi  = currCoeff;
+      // Fast path: when coeffListi*fact (|coeffList| <= 32767) fits in int32,
+      // the dequant can run in int32 SIMD. Otherwise (very high QP, high bit
+      // depth) fall back to the scalar int64 loop.
+      if (fact <= 32767) {
+        tctx->decctx->acceleration.dequant_coeff_block(tctx->coeffBuf,
+                                                       tctx->coeffListcIdx,
+                                                       tctx->coeffPoscIdx,
+                                                       tctx->nCoeffcIdx,
+                                                       fact, offset, bdShift);
+      }
+      else {
+        for (int i=0;i<tctx->nCoeffcIdx;i++) {
+          int64_t currCoeff  = tctx->coeffListcIdxi;
+          currCoeff = Clip3(-32768,32767,
+                            ( (currCoeff * fact + offset ) >> bdShift));
+          tctx->coeffBuf tctx->coeffPoscIdxi  = currCoeff;
+        }
       }
     }
     else {
@@ -545,8 +549,8 @@
 
       int extended_precision_processing_flag = 0;
       int Log2nTbS = Log2(nT);
-      int bdShift = libde265_max( 20 - bit_depth, extended_precision_processing_flag ? 11 : 0 );
-      int tsShift = (extended_precision_processing_flag ? libde265_min( 5, bdShift - 2 ) : 5 )
+      int bdShift = std::max( 20 - bit_depth, extended_precision_processing_flag ? 11 : 0 );
+      int tsShift = (extended_precision_processing_flag ? std::min( 5, bdShift - 2 ) : 5 )
         + Log2nTbS;
 
       if (rotateCoeffs) {
@@ -699,7 +703,7 @@
       //logtrace(LogTransform,"(%d,%d) %d -> ", x,y,level);
       sign   = (level < 0 ? -1: 1);
 
-      level = (abs_value(level) * uiQ + rnd ) >> qBits;
+      level = (std::abs(level) * uiQ + rnd ) >> qBits;
       level *= sign;
       out_coeffblockPos = Clip3(-32768, 32767, level);
       //logtrace(LogTransform,"%d\n", out_coeffblockPos);

libde265-1.0.17.tar.gz/libde265/transform.h -> libde265-1.1.1.tar.gz/libde265/transform.h Changed

libde265-1.0.17.tar.gz/libde265/util.h -> libde265-1.1.1.tar.gz/libde265/util.h Changed

@@ -31,6 +31,7 @@
 
 #include <stdio.h>
 #include <string>
+#include <cstdlib>
 
 #include "libde265/de265.h"
 
@@ -74,22 +75,35 @@
   #endif
 #endif
 
-//inline uint8_t Clip1_8bit(int16_t value) { if (value<=0) return 0; else if (value>=255) return 255; else return value; }
-#define Clip1_8bit(value) ((value)<0 ? 0 : (value)>255 ? 255 : (value))
-#define Clip_BitDepth(value, bit_depth) ((value)<0 ? 0 : (value)>((1<<bit_depth)-1) ? ((1<<bit_depth)-1) : (value))
-#define Clip3(low,high,value) ((value)<(low) ? (low) : (value)>(high) ? (high) : (value))
-#define Sign(value) (((value)<0) ? -1 : ((value)>0) ? 1 : 0)
-#define abs_value(a) (((a)<0) ? -(a) : (a))
-#define libde265_min(a,b) (((a)<(b)) ? (a) : (b))
-#define libde265_max(a,b) (((a)>(b)) ? (a) : (b))
+inline static int Clip1_8bit(int value)
+{
+  return value<0 ? 0 : value>255 ? 255 : value;
+}
+
+inline static int Clip_BitDepth(int value, int bit_depth)
+{
+  const int maxval = (1<<bit_depth)-1;
+  return value<0 ? 0 : value>maxval ? maxval : value;
+}
+
+inline static int Clip3(int low, int high, int value)
+{
+  return value<low ? low : value>high ? high : value;
+}
+
+// three-valued sign: returns -1, 0, or +1
+template <typename T> inline int Sign(T value)
+{
+  return (T(0) < value) - (value < T(0));
+}
 
-LIBDE265_INLINE static int ceil_div(int num,int denom)
+inline static int ceil_div(int num,int denom)
 {
   num += denom-1;
   return num/denom;
 }
 
-LIBDE265_INLINE static int ceil_log2(int val)
+inline static int ceil_log2(int val)
 {
   int n=0;
   while (val > (1<<n)) {
@@ -99,7 +113,7 @@
   return n;
 }
 
-LIBDE265_INLINE static int Log2(int v)
+inline static int Log2(int v)
 {
   int n=0;
   while (v>1) {
@@ -110,7 +124,7 @@
   return n;
 }
 
-LIBDE265_INLINE static int Log2SizeToArea(int v)
+inline static int Log2SizeToArea(int v)
 {
   return (1<<(v<<1));
 }

libde265-1.0.17.tar.gz/libde265/visualize.cc -> libde265-1.1.1.tar.gz/libde265/visualize.cc Changed

libde265-1.0.17.tar.gz/libde265/x86/CMakeLists.txt -> libde265-1.1.1.tar.gz/libde265/x86/CMakeLists.txt Changed

@@ -2,8 +2,17 @@
   sse.cc sse.h
 )
 
-set (x86_sse_sources 
-  sse-motion.cc sse-motion.h sse-dct.h sse-dct.cc
+set (x86_sse_sources
+  sse-motion.cc sse-motion.h sse-dct.h sse-dct.cc sse-intrapred.h sse-intrapred.cc
+  sse-deblk.h sse-deblk.cc
+)
+
+set (x86_avx2_sources
+  transform-avx2.cc transform-avx2.h transform-dct-tables.h
+)
+
+set (x86_avx512_sources
+  transform-avx512.cc transform-avx512.h
 )
 
 add_library(x86 OBJECT ${x86_sources})
@@ -20,6 +29,31 @@
   endif(CMAKE_SIZEOF_VOID_P EQUAL 8)
 endif()
 
-set(X86_OBJECTS $<TARGET_OBJECTS:x86> $<TARGET_OBJECTS:x86_sse> PARENT_SCOPE)
-
 SET_TARGET_PROPERTIES(x86_sse PROPERTIES COMPILE_FLAGS "${sse_flags}")
+
+set(X86_OBJECTS $<TARGET_OBJECTS:x86> $<TARGET_OBJECTS:x86_sse>)
+
+# AVX2 kernels (only compiled where supported). The dispatch/detection code lives
+# in the plain x86 object (sse.cc) so it never runs AVX2 instructions before the
+# runtime CPU check.
+if(HAVE_AVX2)
+  add_library(x86_avx2 OBJECT ${x86_avx2_sources})
+  if(MSVC)
+    SET_TARGET_PROPERTIES(x86_avx2 PROPERTIES COMPILE_FLAGS "/arch:AVX2")
+  else()
+    SET_TARGET_PROPERTIES(x86_avx2 PROPERTIES COMPILE_FLAGS "-mavx2")
+  endif()
+  set(X86_OBJECTS ${X86_OBJECTS} $<TARGET_OBJECTS:x86_avx2>)
+endif()
+
+if(HAVE_AVX512)
+  add_library(x86_avx512 OBJECT ${x86_avx512_sources})
+  if(MSVC)
+    SET_TARGET_PROPERTIES(x86_avx512 PROPERTIES COMPILE_FLAGS "/arch:AVX512")
+  else()
+    SET_TARGET_PROPERTIES(x86_avx512 PROPERTIES COMPILE_FLAGS "-mavx512f -mavx512bw")
+  endif()
+  set(X86_OBJECTS ${X86_OBJECTS} $<TARGET_OBJECTS:x86_avx512>)
+endif()
+
+set(X86_OBJECTS ${X86_OBJECTS} PARENT_SCOPE)

libde265-1.0.17.tar.gz/libde265/x86/sse-dct.cc -> libde265-1.1.1.tar.gz/libde265/x86/sse-dct.cc Changed

@@ -2909,6 +2909,13 @@
 
 
 #if HAVE_SSE4_1
+// All m128iS0..m128iS31 are unconditionally loaded at function entry before any
+// use, but GCC's path analysis gives up inside this very large inlined function
+// and emits spurious -Wmaybe-uninitialized warnings for them. Suppress them here.
+#if defined(__GNUC__) && !defined(__clang__)
+#pragma GCC diagnostic push
+#pragma GCC diagnostic ignored "-Wmaybe-uninitialized"
+#endif
 void ff_hevc_transform_32x32_add_8_sse4(uint8_t *_dst, const int16_t *coeffs,
         ptrdiff_t _stride) {
     uint8_t shift_2nd = 12; // 20 - Bit depth
@@ -5197,6 +5204,9 @@
         }
     }
 }
+#if defined(__GNUC__) && !defined(__clang__)
+#pragma GCC diagnostic pop
+#endif
 #endif
 
 
@@ -7092,3 +7102,135 @@
 }
 #endif
 
+
+#if HAVE_SSE4_1
+// Add the int32 residual block 'r' (nT x nT, row-major, nT values per row) to
+// the prediction samples in 'dst' and clip into the valid pixel range.
+// Equivalent to add_residual_fallback<uint8_t> (bit_depth is always 8 here).
+void add_residual_8_sse4(uint8_t *dst, ptrdiff_t stride,
+                         const int32_t* r, int nT, int bit_depth)
+{
+  if (nT==4) {
+    for (int y=0;y<4;y++) {
+      uint8_t* drow = dst + y*stride;
+
+      __m128i res = _mm_loadu_si128((const __m128i*)(r + y*4));     // 4 x int32 residual
+      __m128i pix = _mm_cvtsi32_si128(*(const int32_t*)drow);       // 4 x uint8
+      pix = _mm_cvtepu8_epi32(pix);                                 // -> 4 x int32
+
+      __m128i sum = _mm_add_epi32(res, pix);
+      sum = _mm_packs_epi32(sum, sum);                              // -> int16 (saturate)
+      sum = _mm_packus_epi16(sum, sum);                             // -> uint8  (clip 0..255)
+
+      *(int32_t*)drow = _mm_cvtsi128_si32(sum);
+    }
+  }
+  else {
+    // nT is 8, 16 or 32 -> always a multiple of 8
+    for (int y=0;y<nT;y++) {
+      const int32_t* rrow = r + y*nT;
+      uint8_t*       drow = dst + y*stride;
+
+      for (int x=0;x<nT;x+=8) {
+        __m128i r0  = _mm_loadu_si128((const __m128i*)(rrow + x));    // 4 x int32
+        __m128i r1  = _mm_loadu_si128((const __m128i*)(rrow + x+4));  // 4 x int32
+        __m128i pix = _mm_loadl_epi64((const __m128i*)(drow + x));    // 8 x uint8
+
+        __m128i p0 = _mm_cvtepu8_epi32(pix);                         // 4 x int32
+        __m128i p1 = _mm_cvtepu8_epi32(_mm_srli_si128(pix,4));       // 4 x int32
+
+        __m128i s0 = _mm_add_epi32(r0, p0);
+        __m128i s1 = _mm_add_epi32(r1, p1);
+
+        __m128i p16 = _mm_packs_epi32(s0, s1);                       // 8 x int16 (saturate)
+        __m128i p8  = _mm_packus_epi16(p16, p16);                    // 8 x uint8 (clip 0..255)
+
+        _mm_storel_epi64((__m128i*)(drow + x), p8);
+      }
+    }
+  }
+}
+
+
+// 16-bit (high bit-depth) variant. Equivalent to add_residual_fallback<uint16_t>.
+void add_residual_16_sse4(uint16_t *dst, ptrdiff_t stride,
+                          const int32_t* r, int nT, int bit_depth)
+{
+  const int32_t maxval = (1<<bit_depth)-1;
+  const __m128i vmax  = _mm_set1_epi32(maxval);
+  const __m128i vzero = _mm_setzero_si128();
+
+  if (nT==4) {
+    for (int y=0;y<4;y++) {
+      uint16_t* drow = dst + y*stride;
+
+      __m128i res = _mm_loadu_si128((const __m128i*)(r + y*4));     // 4 x int32 residual
+      __m128i pix = _mm_loadl_epi64((const __m128i*)drow);          // 4 x uint16
+      pix = _mm_cvtepu16_epi32(pix);                                // -> 4 x int32
+
+      __m128i sum = _mm_add_epi32(res, pix);
+      sum = _mm_min_epi32(_mm_max_epi32(sum, vzero), vmax);         // clip 0..maxval
+      sum = _mm_packus_epi32(sum, sum);                             // -> uint16
+
+      _mm_storel_epi64((__m128i*)drow, sum);
+    }
+  }
+  else {
+    // nT is 8, 16 or 32 -> always a multiple of 8
+    for (int y=0;y<nT;y++) {
+      const int32_t* rrow = r + y*nT;
+      uint16_t*      drow = dst + y*stride;
+
+      for (int x=0;x<nT;x+=8) {
+        __m128i r0  = _mm_loadu_si128((const __m128i*)(rrow + x));    // 4 x int32
+        __m128i r1  = _mm_loadu_si128((const __m128i*)(rrow + x+4));  // 4 x int32
+        __m128i pix = _mm_loadu_si128((const __m128i*)(drow + x));    // 8 x uint16
+
+        __m128i p0 = _mm_cvtepu16_epi32(pix);                        // 4 x int32
+        __m128i p1 = _mm_cvtepu16_epi32(_mm_srli_si128(pix,8));      // 4 x int32
+
+        __m128i s0 = _mm_add_epi32(r0, p0);
+        __m128i s1 = _mm_add_epi32(r1, p1);
+
+        s0 = _mm_min_epi32(_mm_max_epi32(s0, vzero), vmax);          // clip 0..maxval
+        s1 = _mm_min_epi32(_mm_max_epi32(s1, vzero), vmax);
+
+        __m128i out = _mm_packus_epi32(s0, s1);                      // 8 x uint16
+        _mm_storeu_si128((__m128i*)(drow + x), out);
+      }
+    }
+  }
+}
+
+
+// Inverse quantization without scaling list, int32 fast path (see acceleration.h).
+// Vectorizes the multiply/round/clip 8 coefficients at a time; the scatter into
+// coeffBufcoeffPosi stays scalar (no 16-bit SIMD scatter exists).
+void dequant_coeff_block_sse4(int16_t* coeffBuf, const int16_t* coeffList,
+                              const int16_t* coeffPos, int nCoeff,
+                              int32_t fact, int32_t offset, int32_t bdShift)
+{
+  const __m128i vfact = _mm_set1_epi32(fact);
+  const __m128i voff  = _mm_set1_epi32(offset);
+  const __m128i vsh   = _mm_cvtsi32_si128(bdShift);
+
+  alignas(16) int16_t tmp8;
+  int i = 0;
+  for (; i+8 <= nCoeff; i += 8) {
+    __m128i c  = _mm_loadu_si128((const __m128i*)(coeffList + i));  // 8 int16
+    __m128i lo = _mm_cvtepi16_epi32(c);                            // ci..i+3
+    __m128i hi = _mm_cvtepi16_epi32(_mm_srli_si128(c, 8));         // ci+4..i+7
+    lo = _mm_sra_epi32(_mm_add_epi32(_mm_mullo_epi32(lo, vfact), voff), vsh);
+    hi = _mm_sra_epi32(_mm_add_epi32(_mm_mullo_epi32(hi, vfact), voff), vsh);
+    __m128i r = _mm_packs_epi32(lo, hi);                           // signed sat == Clip3
+    _mm_store_si128((__m128i*)tmp, r);
+    for (int k=0;k<8;k++) coeffBuf coeffPosi+k  = tmpk;      // scatter
+  }
+  for (; i < nCoeff; i++) {
+    int32_t v = (coeffListi*fact + offset) >> bdShift;
+    v = v < -32768 ? -32768 : (v > 32767 ? 32767 : v);
+    coeffBuf coeffPosi  = (int16_t)v;
+  }
+}
+#endif
+

libde265-1.0.17.tar.gz/libde265/x86/sse-dct.h -> libde265-1.1.1.tar.gz/libde265/x86/sse-dct.h Changed

libde265-1.1.1.tar.gz/libde265/x86/sse-deblk.cc Added

@@ -0,0 +1,185 @@
+/*
+ * H.265 video codec.
+ * Copyright (c) 2026 Dirk Farin <dirk.farin@gmail.com>
+ *
+ * This file is part of libde265.
+ *
+ * libde265 is free software: you can redistribute it and/or modify
+ * it under the terms of the GNU Lesser General Public License as
+ * published by the Free Software Foundation, either version 3 of
+ * the License, or (at your option) any later version.
+ *
+ * libde265 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 Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public License
+ * along with libde265.  If not, see <http://www.gnu.org/licenses/>.
+ */
+
+// SSE4.1 8-bit deblocking. One edge segment = 4 lines along the edge. The four
+// lines all use the same per-edge parameters and (for luma) the same strong/
+// weak choice, so they are processed in parallel as the 4 int32 lanes of an
+// xmm register. Each sample position (p3..q3) becomes one vector-of-4-lines.
+// For horizontal edges those vectors are 4 contiguous bytes (one per line);
+// for vertical edges the 4 lines are strided, so load/store transpose a small
+// block. The arithmetic is identical to the scalar kernels -> bit-exact.
+
+#include "x86/sse-deblk.h"
+
+#ifdef HAVE_CONFIG_H
+#include "config.h"
+#endif
+
+#if HAVE_SSE4_1
+
+#include <string.h>
+#include <smmintrin.h> // SSE4.1
+
+namespace {
+
+inline __m128i load4(const uint8_t* p) {           // 4 bytes -> 4 int32
+  int32_t t; memcpy(&t, p, 4);
+  return _mm_cvtepu8_epi32(_mm_cvtsi32_si128(t));
+}
+inline void store4(uint8_t* p, __m128i v) {        // 4 int32 (0..255) -> 4 bytes
+  __m128i b = _mm_packus_epi16(_mm_packus_epi32(v, v), _mm_setzero_si128());
+  int32_t t = _mm_cvtsi128_si32(b);
+  memcpy(p, &t, 4);
+}
+
+inline __m128i clip3(__m128i lo, __m128i hi, __m128i v) {
+  return _mm_min_epi32(_mm_max_epi32(v, lo), hi);
+}
+inline __m128i clip_u8(__m128i v) {
+  return _mm_min_epi32(_mm_max_epi32(v, _mm_setzero_si128()), _mm_set1_epi32(255));
+}
+inline __m128i x2(__m128i a){ return _mm_add_epi32(a,a); }
+inline __m128i add3(__m128i a,__m128i b,__m128i c){ return _mm_add_epi32(_mm_add_epi32(a,b),c); }
+
+// --- vertical: load/store transpose a 4(lines) x 8(samples) block ----------
+
+inline void load_vert(const uint8_t* ptr, ptrdiff_t stride, __m128i s8) {
+  __m128i r0 = _mm_loadl_epi64((const __m128i*)(ptr-4));
+  __m128i r1 = _mm_loadl_epi64((const __m128i*)(ptr-4+stride));
+  __m128i r2 = _mm_loadl_epi64((const __m128i*)(ptr-4+2*stride));
+  __m128i r3 = _mm_loadl_epi64((const __m128i*)(ptr-4+3*stride));
+  __m128i e  = _mm_unpacklo_epi8(r0, r1);
+  __m128i f  = _mm_unpacklo_epi8(r2, r3);
+  __m128i lo = _mm_unpacklo_epi16(e, f);   // samples p3 p2 p1 p0 (4 bytes each, 4 lines)
+  __m128i hi = _mm_unpackhi_epi16(e, f);   // samples q0 q1 q2 q3
+  s0=_mm_cvtepu8_epi32(lo);
+  s1=_mm_cvtepu8_epi32(_mm_srli_si128(lo,4));
+  s2=_mm_cvtepu8_epi32(_mm_srli_si128(lo,8));
+  s3=_mm_cvtepu8_epi32(_mm_srli_si128(lo,12));
+  s4=_mm_cvtepu8_epi32(hi);
+  s5=_mm_cvtepu8_epi32(_mm_srli_si128(hi,4));
+  s6=_mm_cvtepu8_epi32(_mm_srli_si128(hi,8));
+  s7=_mm_cvtepu8_epi32(_mm_srli_si128(hi,12));
+}
+
+inline void store_vert(uint8_t* ptr, ptrdiff_t stride, const __m128i s8) {
+  __m128i lo = _mm_packus_epi16(_mm_packus_epi32(s0,s1), _mm_packus_epi32(s2,s3));
+  __m128i hi = _mm_packus_epi16(_mm_packus_epi32(s4,s5), _mm_packus_epi32(s6,s7));
+  const __m128i base = _mm_setr_epi8(0,4,8,12, (char)0x80,(char)0x80,(char)0x80,(char)0x80,
+                                     (char)0x80,(char)0x80,(char)0x80,(char)0x80,
+                                     (char)0x80,(char)0x80,(char)0x80,(char)0x80);
+  for (int k=0;k<4;k++) {
+    __m128i mask = _mm_add_epi8(base, _mm_set1_epi8((char)k));  // {k,4+k,8+k,12+k, 0x80..}
+    __m128i a = _mm_shuffle_epi8(lo, mask);   // low4 = s0..s3 of line k
+    __m128i b = _mm_shuffle_epi8(hi, mask);   // low4 = s4..s7 of line k
+    __m128i row = _mm_unpacklo_epi32(a, b);   // low8 = the 8 samples of line k
+    _mm_storel_epi64((__m128i*)(ptr-4+k*stride), row);
+  }
+}
+
+inline void load_horiz(const uint8_t* ptr, ptrdiff_t stride, __m128i s8) {
+  s0=load4(ptr-4*stride); s1=load4(ptr-3*stride); s2=load4(ptr-2*stride); s3=load4(ptr-1*stride);
+  s4=load4(ptr+0*stride); s5=load4(ptr+1*stride); s6=load4(ptr+2*stride); s7=load4(ptr+3*stride);
+}
+inline void store_horiz(uint8_t* ptr, ptrdiff_t stride, const __m128i s8) {
+  store4(ptr-4*stride,s0); store4(ptr-3*stride,s1); store4(ptr-2*stride,s2); store4(ptr-1*stride,s3);
+  store4(ptr+0*stride,s4); store4(ptr+1*stride,s5); store4(ptr+2*stride,s6); store4(ptr+3*stride,s7);
+}
+
+} // namespace
+
+
+void deblock_luma_8_sse4(uint8_t* ptr, ptrdiff_t stride, int vertical,
+                         int dE, int dEp, int dEq, int tc, int filterP, int filterQ)
+{
+  __m128i s8;
+  if (vertical) load_vert(ptr, stride, s); else load_horiz(ptr, stride, s);
+
+  const __m128i p3=s0, p2=s1, p1=s2, p0=s3;
+  const __m128i q0=s4, q1=s5, q2=s6, q3=s7;
+
+  if (dE==2) {
+    // strong filtering
+    const __m128i v2tc = _mm_set1_epi32(2*tc);
+    const __m128i c4   = _mm_set1_epi32(4);
+    const __m128i c2   = _mm_set1_epi32(2);
+
+    __m128i pn0 = _mm_srai_epi32(_mm_add_epi32(add3(p2, x2(p1), x2(p0)), add3(x2(q0), q1, c4)), 3);
+    pn0 = clip3(_mm_sub_epi32(p0,v2tc), _mm_add_epi32(p0,v2tc), pn0);
+    __m128i pn1 = _mm_srai_epi32(_mm_add_epi32(add3(p2,p1,p0), _mm_add_epi32(q0,c2)), 2);
+    pn1 = clip3(_mm_sub_epi32(p1,v2tc), _mm_add_epi32(p1,v2tc), pn1);
+    __m128i pn2 = _mm_srai_epi32(_mm_add_epi32(add3(x2(p3), _mm_add_epi32(x2(p2),p2), p1), add3(p0,q0,c4)), 3);
+    pn2 = clip3(_mm_sub_epi32(p2,v2tc), _mm_add_epi32(p2,v2tc), pn2);
+
+    __m128i qn0 = _mm_srai_epi32(_mm_add_epi32(add3(p1, x2(p0), x2(q0)), add3(x2(q1), q2, c4)), 3);
+    qn0 = clip3(_mm_sub_epi32(q0,v2tc), _mm_add_epi32(q0,v2tc), qn0);
+    __m128i qn1 = _mm_srai_epi32(_mm_add_epi32(add3(p0,q0,q1), _mm_add_epi32(q2,c2)), 2);
+    qn1 = clip3(_mm_sub_epi32(q1,v2tc), _mm_add_epi32(q1,v2tc), qn1);
+    __m128i qn2 = _mm_srai_epi32(_mm_add_epi32(add3(p0,q0,q1), add3(_mm_add_epi32(x2(q2),q2), x2(q3), c4)), 3);
+    qn2 = clip3(_mm_sub_epi32(q2,v2tc), _mm_add_epi32(q2,v2tc), qn2);
+
+    if (filterP) { s3=pn0; s2=pn1; s1=pn2; }
+    if (filterQ) { s4=qn0; s5=qn1; s6=qn2; }
+  }
+  else {
+    // weak filtering
+    const __m128i vtc  = _mm_set1_epi32(tc);
+    const __m128i delta0 = _mm_srai_epi32(
+        _mm_add_epi32(_mm_sub_epi32(_mm_mullo_epi32(_mm_set1_epi32(9), _mm_sub_epi32(q0,p0)),
+                                    _mm_mullo_epi32(_mm_set1_epi32(3), _mm_sub_epi32(q1,p1))),
+                      _mm_set1_epi32(8)), 4);
+    // per-line mask: abs(delta) < tc*10
+    __m128i mask = _mm_cmpgt_epi32(_mm_set1_epi32(tc*10), _mm_abs_epi32(delta0));
+    __m128i delta = clip3(_mm_set1_epi32(-tc), vtc, delta0);
+
+    if (filterP) {
+      __m128i p0n = clip_u8(_mm_add_epi32(p0, delta));
+      s3 = _mm_blendv_epi8(p0, p0n, mask);
+    }
+    if (filterQ) {
+      __m128i q0n = clip_u8(_mm_sub_epi32(q0, delta));
+      s4 = _mm_blendv_epi8(q0, q0n, mask);
+    }
+    if (dEp && filterP) {
+      const __m128i htc = _mm_set1_epi32(tc>>1);
+      __m128i dp = _mm_srai_epi32(_mm_add_epi32(_mm_sub_epi32(_mm_srai_epi32(_mm_add_epi32(_mm_add_epi32(p2,p0),_mm_set1_epi32(1)),1), p1), delta), 1);
+      dp = clip3(_mm_sub_epi32(_mm_setzero_si128(),htc), htc, dp);
+      __m128i p1n = clip_u8(_mm_add_epi32(p1, dp));
+      s2 = _mm_blendv_epi8(p1, p1n, mask);
+    }
+    if (dEq && filterQ) {
+      const __m128i htc = _mm_set1_epi32(tc>>1);
+      __m128i dq = _mm_srai_epi32(_mm_sub_epi32(_mm_sub_epi32(_mm_srai_epi32(_mm_add_epi32(_mm_add_epi32(q2,q0),_mm_set1_epi32(1)),1), q1), delta), 1);
+      dq = clip3(_mm_sub_epi32(_mm_setzero_si128(),htc), htc, dq);
+      __m128i q1n = clip_u8(_mm_add_epi32(q1, dq));
+      s5 = _mm_blendv_epi8(q1, q1n, mask);
+    }
+  }
+
+  if (vertical) store_vert(ptr, stride, s); else store_horiz(ptr, stride, s);
+}
+
+// Note: an SSE chroma deblock filter was implemented and benchmarked too, but
+// the chroma filter is a single delta per line -- so trivial that it is fully
+// load/store-bound (the vertical case needs a strided 2-column scatter), and
+// SSE measured slower than scalar (~0.5-0.9x). Chroma deblock therefore stays
+// on the scalar fallback; only luma is accelerated here.
+
+#endif // HAVE_SSE4_1

libde265-1.1.1.tar.gz/libde265/x86/sse-deblk.h Added

@@ -0,0 +1,33 @@
+/*
+ * H.265 video codec.
+ * Copyright (c) 2026 Dirk Farin <dirk.farin@gmail.com>
+ *
+ * This file is part of libde265.
+ *
+ * libde265 is free software: you can redistribute it and/or modify
+ * it under the terms of the GNU Lesser General Public License as
+ * published by the Free Software Foundation, either version 3 of
+ * the License, or (at your option) any later version.
+ *
+ * libde265 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 Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public License
+ * along with libde265.  If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#ifndef SSE_DEBLK_H
+#define SSE_DEBLK_H
+
+#include <stddef.h>
+#include <stdint.h>
+
+// SSE4.1 8-bit deblocking of one 4-line edge segment. Bit-identical to the
+// scalar kernels in fallback-deblk.h (verified by dev-tools/test-deblk).
+void deblock_luma_8_sse4(uint8_t* ptr, ptrdiff_t stride, int vertical,
+                         int dE, int dEp, int dEq, int tc, int filterP, int filterQ);
+// (chroma deblock stays scalar — SSE measured slower; see sse-deblk.cc)
+
+#endif

libde265-1.1.1.tar.gz/libde265/x86/sse-intrapred.cc Added

@@ -0,0 +1,224 @@
+/*
+ * H.265 video codec.
+ * Copyright (c) 2026 Dirk Farin <dirk.farin@gmail.com>
+ *
+ * This file is part of libde265.
+ *
+ * libde265 is free software: you can redistribute it and/or modify
+ * it under the terms of the GNU Lesser General Public License as
+ * published by the Free Software Foundation, either version 3 of
+ * the License, or (at your option) any later version.
+ *
+ * libde265 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 Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public License
+ * along with libde265.  If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#include "x86/sse-intrapred.h"
+#include "libde265/util.h"
+
+#ifdef HAVE_CONFIG_H
+#include "config.h"
+#endif
+
+#include <string.h>
+#include <emmintrin.h> // SSE2
+
+#if HAVE_SSE4_1
+#include <smmintrin.h> // SSE4.1
+
+// angle / inverse-angle lookup tables, defined in intrapred.cc
+extern const int intraPredAngle_table1+34;
+extern const int invAngle_table25-10;
+
+namespace {
+
+const int kMaxBlk = 64; // == MAX_INTRA_PRED_BLOCK_SIZE
+
+// load 4 / 8 consecutive uint8_t and zero-extend to 8x int16
+inline __m128i load4_epi16(const uint8_t* p) {
+  int32_t t; memcpy(&t, p, 4);
+  return _mm_cvtepu8_epi16(_mm_cvtsi32_si128(t));
+}
+inline __m128i load8_epi16(const uint8_t* p) {
+  return _mm_cvtepu8_epi16(_mm_loadl_epi64((const __m128i*)p));
+}
+
+// store nT bytes from the low lanes of an u8-packed register (nT in {4,8,16,32})
+inline void store_row(uint8_t* d, __m128i v, int nT) {
+  switch (nT) {
+    case 4:  { int32_t t = _mm_cvtsi128_si32(v); memcpy(d, &t, 4); } break;
+    case 8:  _mm_storel_epi64((__m128i*)d, v); break;
+    case 16: _mm_storeu_si128((__m128i*)d, v); break;
+    default: _mm_storeu_si128((__m128i*)d, v);
+             _mm_storeu_si128((__m128i*)(d+16), v); break;
+  }
+}
+
+// copy exactly nT bytes (nT in {4,8,16,32})
+inline void copy_row(uint8_t* d, const uint8_t* s, int nT) {
+  switch (nT) {
+    case 4:  memcpy(d, s, 4); break;
+    case 8:  memcpy(d, s, 8); break;
+    case 16: _mm_storeu_si128((__m128i*)d, _mm_loadu_si128((const __m128i*)s)); break;
+    default: _mm_storeu_si128((__m128i*)d,      _mm_loadu_si128((const __m128i*)s));
+             _mm_storeu_si128((__m128i*)(d+16), _mm_loadu_si128((const __m128i*)(s+16))); break;
+  }
+}
+
+inline int shift_for(int nT) { return (nT==4)?3 : (nT==8)?4 : (nT==16)?5 : 6; } // Log2(nT)+1
+
+} // namespace
+
+
+void intra_pred_dc_8_sse4(uint8_t* dst, ptrdiff_t stride, int nT, int cIdx, const uint8_t* border)
+{
+  const int shift = shift_for(nT);
+
+  int dcVal = 0;
+  for (int i=0;i<nT;i++) { dcVal += borderi+1; dcVal += border-i-1; }
+  dcVal += nT;
+  dcVal >>= shift;
+
+  const __m128i v = _mm_set1_epi8((char)dcVal);
+  for (int y=0;y<nT;y++) store_row(dst + y*stride, v, nT);
+
+  // luma edge smoothing overwrites first row and first column (disjoint cells)
+  if (cIdx==0 && nT<32) {
+    dst0 = (uint8_t)((border-1 + 2*dcVal + border1 + 2) >> 2);
+    for (int x=1;x<nT;x++) dstx          = (uint8_t)((border x+1 + 3*dcVal + 2) >> 2);
+    for (int y=1;y<nT;y++) dsty*stride   = (uint8_t)((border-y-1 + 3*dcVal + 2) >> 2);
+  }
+}
+
+
+void intra_pred_planar_8_sse4(uint8_t* dst, ptrdiff_t stride, int nT, int cIdx, const uint8_t* border)
+{
+  const int shift = shift_for(nT);
+  const int TR = border 1+nT;   // top-right corner sample
+  const int BL = border-1-nT;   // bottom-left corner sample
+
+  const __m128i base  = _mm_setr_epi16(0,1,2,3,4,5,6,7);
+  const __m128i vTR   = _mm_set1_epi16((short)TR);
+  const __m128i vNTm1 = _mm_set1_epi16((short)(nT-1));
+  const __m128i one   = _mm_set1_epi16(1);
+
+  for (int y=0;y<nT;y++) {
+    const int left_y = border-1-y;
+    const int Cy = (y+1)*BL + nT;                 // constant term for this row
+    const __m128i vL    = _mm_set1_epi16((short)left_y);
+    const __m128i vNT1Y = _mm_set1_epi16((short)(nT-1-y));
+    const __m128i vC    = _mm_set1_epi16((short)Cy);
+
+    for (int x=0;x<nT;x+=8) {
+      const __m128i xidx = _mm_add_epi16(_mm_set1_epi16((short)x), base);
+      const __m128i vA   = _mm_sub_epi16(vNTm1, xidx);     // (nT-1-x)
+      const __m128i vB   = _mm_add_epi16(xidx, one);       // (x+1)
+      const __m128i top  = (nT==4) ? load4_epi16(border+1+x) : load8_epi16(border+1+x);
+
+      __m128i acc = _mm_mullo_epi16(vA,  vL);              // (nT-1-x)*border-1-y
+      acc = _mm_add_epi16(acc, _mm_mullo_epi16(vB,  vTR)); // (x+1)*border1+nT
+      acc = _mm_add_epi16(acc, _mm_mullo_epi16(top, vNT1Y));// (nT-1-y)*border1+x
+      acc = _mm_add_epi16(acc, vC);                        // (y+1)*border-1-nT + nT
+      acc = _mm_srli_epi16(acc, shift);
+
+      const __m128i p = _mm_packus_epi16(acc, acc);
+      store_row(dst + y*stride + x, p, (nT<8)?nT:8);
+    }
+  }
+}
+
+
+void intra_pred_angular_8_sse4(uint8_t* dst, ptrdiff_t stride, int bit_depth, int disableBoundaryFilter,
+                               int xB0, int yB0, int mode, int nT, int cIdx, const uint8_t* border)
+{
+  const int intraPredAngle = intraPredAngle_tablemode;
+
+  uint8_t  ref_mem4*kMaxBlk+1;
+  uint8_t* ref = &ref_mem2*kMaxBlk;
+
+  if (mode >= 18) {
+    for (int x=0;x<=nT;x++) refx = borderx;
+
+    if (intraPredAngle<0) {
+      const int invAngle = invAngle_tablemode-11;
+      if (((nT*intraPredAngle)>>5) < -1) {
+        for (int x=(nT*intraPredAngle)>>5; x<=-1; x++)
+          refx = border0-((x*invAngle+128)>>8);
+      }
+    } else {
+      for (int x=nT+1; x<=2*nT; x++) refx = borderx;
+    }
+
+    for (int y=0;y<nT;y++) {
+      const int iIdx  = ((y+1)*intraPredAngle)>>5;
+      const int iFact = ((y+1)*intraPredAngle)&31;
+      const uint8_t* src = ref + iIdx + 1;
+      uint8_t* d = dst + y*stride;
+
+      if (iFact==0) {
+        copy_row(d, src, nT);                     // dstx = refx+iIdx+1
+      } else {
+        const __m128i w0  = _mm_set1_epi16((short)(32-iFact));
+        const __m128i w1  = _mm_set1_epi16((short)iFact);
+        const __m128i r16 = _mm_set1_epi16(16);
+        if (nT==4) {
+          __m128i a = load4_epi16(src), b = load4_epi16(src+1);
+          __m128i acc = _mm_add_epi16(_mm_add_epi16(_mm_mullo_epi16(a,w0), _mm_mullo_epi16(b,w1)), r16);
+          acc = _mm_srli_epi16(acc, 5);
+          store_row(d, _mm_packus_epi16(acc,acc), 4);
+        } else {
+          for (int x=0;x<nT;x+=8) {
+            __m128i a = load8_epi16(src+x), b = load8_epi16(src+x+1);
+            __m128i acc = _mm_add_epi16(_mm_add_epi16(_mm_mullo_epi16(a,w0), _mm_mullo_epi16(b,w1)), r16);
+            acc = _mm_srli_epi16(acc, 5);
+            store_row(d+x, _mm_packus_epi16(acc,acc), 8);
+          }
+        }
+      }
+    }
+
+    if (mode==26 && cIdx==0 && nT<32 && !disableBoundaryFilter) {
+      for (int y=0;y<nT;y++)
+        dsty*stride = (uint8_t)Clip_BitDepth(border1 + ((border-1-y - border0)>>1), bit_depth);
+    }
+  }
+  else {
+    // Modes 2..17: the reference projection is transposed (per-column iIdx/iFact and
+    // a row-indexed reference fetch), which does not map onto contiguous SIMD loads or
+    // stores. Use the scalar reference path here -- kept bit-identical to
+    // intra_prediction_angular() in intrapred.h.
+    for (int x=0;x<=nT;x++) refx = border-x;
+
+    if (intraPredAngle<0) {
+      const int invAngle = invAngle_tablemode-11;
+      if (((nT*intraPredAngle)>>5) < -1) {
+        for (int x=(nT*intraPredAngle)>>5; x<=-1; x++)
+          refx = border(x*invAngle+128)>>8;
+      }
+    } else {
+      for (int x=nT+1; x<=2*nT; x++) refx = border-x;
+    }
+
+    for (int y=0;y<nT;y++)
+      for (int x=0;x<nT;x++) {
+        const int iIdx  = ((x+1)*intraPredAngle)>>5;
+        const int iFact = ((x+1)*intraPredAngle)&31;
+        if (iFact != 0)
+          dstx+y*stride = (uint8_t)(((32-iFact)*refy+iIdx+1 + iFact*refy+iIdx+2 + 16)>>5);
+        else
+          dstx+y*stride = refy+iIdx+1;
+      }
+
+    if (mode==10 && cIdx==0 && nT<32 && !disableBoundaryFilter) {
+      for (int x=0;x<nT;x++)
+        dstx = (uint8_t)Clip_BitDepth(border-1 + ((border1+x - border0)>>1), bit_depth);
+    }
+  }
+}
+
+#endif // HAVE_SSE4_1

libde265-1.1.1.tar.gz/libde265/x86/sse-intrapred.h Added

@@ -0,0 +1,35 @@
+/*
+ * H.265 video codec.
+ * Copyright (c) 2026 Dirk Farin <dirk.farin@gmail.com>
+ *
+ * This file is part of libde265.
+ *
+ * libde265 is free software: you can redistribute it and/or modify
+ * it under the terms of the GNU Lesser General Public License as
+ * published by the Free Software Foundation, either version 3 of
+ * the License, or (at your option) any later version.
+ *
+ * libde265 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 Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public License
+ * along with libde265.  If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#ifndef SSE_INTRAPRED_H
+#define SSE_INTRAPRED_H
+
+#include <stddef.h>
+#include <stdint.h>
+
+// SSE4.1 accelerated 8-bit intra prediction. All three are bit-identical to the
+// scalar kernels in intrapred.h (verified by dev-tools/test-intrapred.cc).
+
+void intra_pred_dc_8_sse4    (uint8_t* dst, ptrdiff_t stride, int nT, int cIdx, const uint8_t* border);
+void intra_pred_planar_8_sse4(uint8_t* dst, ptrdiff_t stride, int nT, int cIdx, const uint8_t* border);
+void intra_pred_angular_8_sse4(uint8_t* dst, ptrdiff_t stride, int bit_depth, int disableBoundaryFilter,
+                               int xB0, int yB0, int mode, int nT, int cIdx, const uint8_t* border);
+
+#endif

libde265-1.0.17.tar.gz/libde265/x86/sse.cc -> libde265-1.1.1.tar.gz/libde265/x86/sse.cc Changed

@@ -25,11 +25,20 @@
 #include "x86/sse.h"
 #include "x86/sse-motion.h"
 #include "x86/sse-dct.h"
+#include "x86/sse-intrapred.h"
+#include "x86/sse-deblk.h"
 
 #ifdef HAVE_CONFIG_H
 #include "config.h"
 #endif
 
+#if HAVE_AVX2
+#include "x86/transform-avx2.h"
+#endif
+#if HAVE_AVX512
+#include "x86/transform-avx512.h"
+#endif
+
 #if defined(__GNUC__) && !defined(__EMSCRIPTEN__)
 #include <cpuid.h>
 #endif
@@ -109,7 +118,55 @@
     accel->transform_add_81 = ff_hevc_transform_8x8_add_8_sse4;
     accel->transform_add_82 = ff_hevc_transform_16x16_add_8_sse4;
     accel->transform_add_83 = ff_hevc_transform_32x32_add_8_sse4;
+
+    accel->add_residual_8  = add_residual_8_sse4;
+    accel->add_residual_16 = add_residual_16_sse4;
+    accel->dequant_coeff_block = dequant_coeff_block_sse4;
+
+    accel->intra_pred_dc_8      = intra_pred_dc_8_sse4;
+    accel->intra_pred_planar_8  = intra_pred_planar_8_sse4;
+    accel->intra_pred_angular_8 = intra_pred_angular_8_sse4;
+
+    accel->deblock_luma_8   = deblock_luma_8_sse4;
+    // chroma deblock stays on the scalar fallback: the filter is too cheap to
+    // amortize the SIMD load/transpose/scatter overhead (SSE measured slower).
+  }
+#endif
+}
+
+
+void init_acceleration_functions_avx2(struct acceleration_functions* accel)
+{
+#if HAVE_AVX2
+  // __builtin_cpu_supports("avx2") handles the OSXSAVE / XGETBV (YMM-enabled)
+  // checks internally, so this is safe to call on any CPU. This TU is *not*
+  // compiled with -mavx2, so reaching here never executes an AVX2 instruction.
+#if defined(__GNUC__) && !defined(__EMSCRIPTEN__)
+  __builtin_cpu_init();
+  if (__builtin_cpu_supports("avx2")) {
+    accel->transform_add_82 = transform_16x16_add_8_avx2;
+    accel->transform_add_83 = transform_32x32_add_8_avx2;
+    // NB: dequant intentionally stays on the SSE version. An AVX2 variant was
+    // implemented and benchmarked, but inverse quantization is scatter-bound, so
+    // the wider arithmetic gave no benefit and AVX2 measured actually slightly
+    // slower than SSE. (AVX-512 would be no better, for the same reason.)
   }
 #endif
+#endif
+}
+
+
+void init_acceleration_functions_avx512(struct acceleration_functions* accel)
+{
+#if HAVE_AVX512
+#if defined(__GNUC__) && !defined(__EMSCRIPTEN__)
+  __builtin_cpu_init();
+  // need AVX-512F + AVX-512BW (16-bit ops). __builtin_cpu_supports handles the
+  // OS (XCR0/ZMM-enabled) check. This TU is not compiled with -mavx512*.
+  if (__builtin_cpu_supports("avx512f") && __builtin_cpu_supports("avx512bw")) {
+    accel->transform_add_83 = transform_32x32_add_8_avx512;
+  }
+#endif
+#endif
 }

libde265-1.0.17.tar.gz/libde265/x86/sse.h -> libde265-1.1.1.tar.gz/libde265/x86/sse.h Changed

libde265-1.1.1.tar.gz/libde265/x86/transform-avx2.cc Added

@@ -0,0 +1,318 @@
+/*
+ * H.265 video codec.
+ * Copyright (c) 2026 Dirk Farin <dirk.farin@gmail.com>
+ *
+ * This file is part of libde265.
+ *
+ * libde265 is free software: you can redistribute it and/or modify
+ * it under the terms of the GNU Lesser General Public License as
+ * published by the Free Software Foundation, either version 3 of
+ * the License, or (at your option) any later version.
+ *
+ * libde265 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 Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public License
+ * along with libde265.  If not, see <http://www.gnu.org/licenses/>.
+ */
+
+// AVX2 inverse transforms (16x16, 32x32) for 8-bit.
+//
+// These produce bit-identical results to the SSE4.1 / scalar kernels but
+// process all columns of a block in 256-bit registers (16 int16 per register)
+// instead of 8-column slabs. The per-column butterfly arithmetic matches the
+// SSE version: every 256-bit op acts independently on its two 128-bit halves,
+// so each half does exactly what the SSE code does for 8 columns. The only
+// genuinely new piece is a full 16x16 / 32x32 in-register transpose between the
+// vertical and horizontal passes (the SSE code only transposed 8x8 sub-blocks).
+//
+// The 32-point inverse transform's even part is itself a 16-point inverse
+// transform of the even-indexed input rows, so idct16_core() is reused for it.
+
+#include "x86/transform-avx2.h"
+
+#ifdef HAVE_CONFIG_H
+#include "config.h"
+#endif
+
+#if HAVE_AVX2
+
+#include <immintrin.h>
+#include "x86/transform-dct-tables.h"   // idct_T16_{1,2,3}, idct_T32
+
+namespace {
+
+#define shift_1st 7
+#define add_1st   (1 << (shift_1st - 1))   // 64
+// 8-bit: shift_2nd = 20 - 8 = 12, add_2nd = 1<<11
+#define shift_2nd 12
+#define add_2nd   (1 << (shift_2nd - 1))   // 2048
+
+static inline __m256i bc(const int16_t* p) {
+  return _mm256_broadcastsi128_si256(_mm_load_si128((const __m128i*)p));
+}
+
+// --- transposes ----------------------------------------------------------
+
+// 8x8 int16 transpose performed independently in each 128-bit lane.
+static inline void transpose8x8_lanes(__m256i a8) {
+  __m256i b0=_mm256_unpacklo_epi16(a0,a1), b1=_mm256_unpackhi_epi16(a0,a1);
+  __m256i b2=_mm256_unpacklo_epi16(a2,a3), b3=_mm256_unpackhi_epi16(a2,a3);
+  __m256i b4=_mm256_unpacklo_epi16(a4,a5), b5=_mm256_unpackhi_epi16(a4,a5);
+  __m256i b6=_mm256_unpacklo_epi16(a6,a7), b7=_mm256_unpackhi_epi16(a6,a7);
+  __m256i c0=_mm256_unpacklo_epi32(b0,b2), c1=_mm256_unpackhi_epi32(b0,b2);
+  __m256i c2=_mm256_unpacklo_epi32(b1,b3), c3=_mm256_unpackhi_epi32(b1,b3);
+  __m256i c4=_mm256_unpacklo_epi32(b4,b6), c5=_mm256_unpackhi_epi32(b4,b6);
+  __m256i c6=_mm256_unpacklo_epi32(b5,b7), c7=_mm256_unpackhi_epi32(b5,b7);
+  a0=_mm256_unpacklo_epi64(c0,c4); a1=_mm256_unpackhi_epi64(c0,c4);
+  a2=_mm256_unpacklo_epi64(c1,c5); a3=_mm256_unpackhi_epi64(c1,c5);
+  a4=_mm256_unpacklo_epi64(c2,c6); a5=_mm256_unpackhi_epi64(c2,c6);
+  a6=_mm256_unpacklo_epi64(c3,c7); a7=_mm256_unpackhi_epi64(c3,c7);
+}
+
+// Full 16x16 int16 transpose of S0..15 (each ymm = one row: lane0 cols0-7,
+// lane1 cols8-15): four 8x8 quadrant transposes + a lane swap.
+static inline void transpose16x16(__m256i S16) {
+  __m256i top8, bot8;
+  for (int r=0;r<8;r++){ topr=Sr; botr=Sr+8; }
+  transpose8x8_lanes(top);
+  transpose8x8_lanes(bot);
+  for (int r=0;r<8;r++){
+    Sr   = _mm256_permute2x128_si256(topr, botr, 0x20);
+    Sr+8 = _mm256_permute2x128_si256(topr, botr, 0x31);
+  }
+}
+
+// Full 32x32 int16 transpose. Lor=cols0-15 of row r, Hir=cols16-31.
+// Built from four 16x16 block transposes: out = A^T,C^T,B^T,D^T.
+static inline void transpose32x32(__m256i Lo32, __m256i Hi32) {
+  __m256i A16,B16,C16,D16;
+  for (int r=0;r<16;r++){ Ar=Lor; Br=Hir; Cr=Lor+16; Dr=Hir+16; }
+  transpose16x16(A); transpose16x16(B); transpose16x16(C); transpose16x16(D);
+  for (int r=0;r<16;r++){ Lor=Ar; Hir=Cr; Lor+16=Br; Hir+16=Dr; }
+}
+
+// --- 16-point core -------------------------------------------------------
+
+// Computes the 16 outputs of a 1-D 16-point inverse transform applied down the
+// 16 rows S0..15, for 16 columns at once, as int32 (no rounding/shift). The
+// result lanes split lo = cols{0-3,8-11}, hi = cols{4-7,12-15} (recombined to
+// natural order by packs_epi32 in the caller).
+static inline void idct16_core(const __m256i S16, __m256i resL16, __m256i resH16) {
+  const __m256i T00=bc(idct_T16_100),T01=bc(idct_T16_101),T02=bc(idct_T16_102),T03=bc(idct_T16_103);
+  const __m256i T04=bc(idct_T16_104),T05=bc(idct_T16_105),T06=bc(idct_T16_106),T07=bc(idct_T16_107);
+  const __m256i T10=bc(idct_T16_110),T11=bc(idct_T16_111),T12=bc(idct_T16_112),T13=bc(idct_T16_113);
+  const __m256i T14=bc(idct_T16_114),T15=bc(idct_T16_115),T16=bc(idct_T16_116),T17=bc(idct_T16_117);
+  const __m256i T20=bc(idct_T16_120),T21=bc(idct_T16_121),T22=bc(idct_T16_122),T23=bc(idct_T16_123);
+  const __m256i T24=bc(idct_T16_124),T25=bc(idct_T16_125),T26=bc(idct_T16_126),T27=bc(idct_T16_127);
+  const __m256i T30=bc(idct_T16_130),T31=bc(idct_T16_131),T32_=bc(idct_T16_132),T33=bc(idct_T16_133);
+  const __m256i T34=bc(idct_T16_134),T35=bc(idct_T16_135),T36=bc(idct_T16_136),T37=bc(idct_T16_137);
+  const __m256i U00=bc(idct_T16_200),U01=bc(idct_T16_201),U02=bc(idct_T16_202),U03=bc(idct_T16_203);
+  const __m256i U10=bc(idct_T16_210),U11=bc(idct_T16_211),U12=bc(idct_T16_212),U13=bc(idct_T16_213);
+  const __m256i V00=bc(idct_T16_300),V01=bc(idct_T16_301),V10=bc(idct_T16_310),V11=bc(idct_T16_311);
+
+  __m256i m0,m1,m2,m3,m4,m5,m6,m7;
+  __m256i E0l,E1l,E2l,E3l,E0h,E1h,E2h,E3h;
+  __m256i O0l,O1l,O2l,O3l,O4l,O5l,O6l,O7l,O0h,O1h,O2h,O3h,O4h,O5h,O6h,O7h;
+  __m256i E00l,E01l,E00h,E01h,EE0l,EE1l,EE2l,EE3l,EE0h,EE1h,EE2h,EE3h;
+  __m256i E4l,E5l,E6l,E7l,E4h,E5h,E6h,E7h;
+
+  m0=_mm256_unpacklo_epi16(S1,S3);  E0l=_mm256_madd_epi16(m0,T00);
+  m1=_mm256_unpackhi_epi16(S1,S3);  E0h=_mm256_madd_epi16(m1,T00);
+  m2=_mm256_unpacklo_epi16(S5,S7);  E1l=_mm256_madd_epi16(m2,T10);
+  m3=_mm256_unpackhi_epi16(S5,S7);  E1h=_mm256_madd_epi16(m3,T10);
+  m4=_mm256_unpacklo_epi16(S9,S11); E2l=_mm256_madd_epi16(m4,T20);
+  m5=_mm256_unpackhi_epi16(S9,S11); E2h=_mm256_madd_epi16(m5,T20);
+  m6=_mm256_unpacklo_epi16(S13,S15);E3l=_mm256_madd_epi16(m6,T30);
+  m7=_mm256_unpackhi_epi16(S13,S15);E3h=_mm256_madd_epi16(m7,T30);
+  O0l=_mm256_add_epi32(_mm256_add_epi32(E0l,E1l),_mm256_add_epi32(E2l,E3l));
+  O0h=_mm256_add_epi32(_mm256_add_epi32(E0h,E1h),_mm256_add_epi32(E2h,E3h));
+
+  E0l=_mm256_madd_epi16(m0,T01);E0h=_mm256_madd_epi16(m1,T01);E1l=_mm256_madd_epi16(m2,T11);E1h=_mm256_madd_epi16(m3,T11);
+  E2l=_mm256_madd_epi16(m4,T21);E2h=_mm256_madd_epi16(m5,T21);E3l=_mm256_madd_epi16(m6,T31);E3h=_mm256_madd_epi16(m7,T31);
+  O1l=_mm256_add_epi32(_mm256_add_epi32(E0l,E1l),_mm256_add_epi32(E2l,E3l));
+  O1h=_mm256_add_epi32(_mm256_add_epi32(E0h,E1h),_mm256_add_epi32(E2h,E3h));
+
+  E0l=_mm256_madd_epi16(m0,T02);E0h=_mm256_madd_epi16(m1,T02);E1l=_mm256_madd_epi16(m2,T12);E1h=_mm256_madd_epi16(m3,T12);
+  E2l=_mm256_madd_epi16(m4,T22);E2h=_mm256_madd_epi16(m5,T22);E3l=_mm256_madd_epi16(m6,T32_);E3h=_mm256_madd_epi16(m7,T32_);
+  O2l=_mm256_add_epi32(_mm256_add_epi32(E0l,E1l),_mm256_add_epi32(E2l,E3l));
+  O2h=_mm256_add_epi32(_mm256_add_epi32(E0h,E1h),_mm256_add_epi32(E2h,E3h));
+
+  E0l=_mm256_madd_epi16(m0,T03);E0h=_mm256_madd_epi16(m1,T03);E1l=_mm256_madd_epi16(m2,T13);E1h=_mm256_madd_epi16(m3,T13);
+  E2l=_mm256_madd_epi16(m4,T23);E2h=_mm256_madd_epi16(m5,T23);E3l=_mm256_madd_epi16(m6,T33);E3h=_mm256_madd_epi16(m7,T33);
+  O3l=_mm256_add_epi32(_mm256_add_epi32(E0l,E1l),_mm256_add_epi32(E2l,E3l));
+  O3h=_mm256_add_epi32(_mm256_add_epi32(E0h,E1h),_mm256_add_epi32(E2h,E3h));
+
+  E0l=_mm256_madd_epi16(m0,T04);E0h=_mm256_madd_epi16(m1,T04);E1l=_mm256_madd_epi16(m2,T14);E1h=_mm256_madd_epi16(m3,T14);
+  E2l=_mm256_madd_epi16(m4,T24);E2h=_mm256_madd_epi16(m5,T24);E3l=_mm256_madd_epi16(m6,T34);E3h=_mm256_madd_epi16(m7,T34);
+  O4l=_mm256_add_epi32(_mm256_add_epi32(E0l,E1l),_mm256_add_epi32(E2l,E3l));
+  O4h=_mm256_add_epi32(_mm256_add_epi32(E0h,E1h),_mm256_add_epi32(E2h,E3h));
+
+  E0l=_mm256_madd_epi16(m0,T05);E0h=_mm256_madd_epi16(m1,T05);E1l=_mm256_madd_epi16(m2,T15);E1h=_mm256_madd_epi16(m3,T15);
+  E2l=_mm256_madd_epi16(m4,T25);E2h=_mm256_madd_epi16(m5,T25);E3l=_mm256_madd_epi16(m6,T35);E3h=_mm256_madd_epi16(m7,T35);
+  O5l=_mm256_add_epi32(_mm256_add_epi32(E0l,E1l),_mm256_add_epi32(E2l,E3l));
+  O5h=_mm256_add_epi32(_mm256_add_epi32(E0h,E1h),_mm256_add_epi32(E2h,E3h));
+
+  E0l=_mm256_madd_epi16(m0,T06);E0h=_mm256_madd_epi16(m1,T06);E1l=_mm256_madd_epi16(m2,T16);E1h=_mm256_madd_epi16(m3,T16);
+  E2l=_mm256_madd_epi16(m4,T26);E2h=_mm256_madd_epi16(m5,T26);E3l=_mm256_madd_epi16(m6,T36);E3h=_mm256_madd_epi16(m7,T36);
+  O6l=_mm256_add_epi32(_mm256_add_epi32(E0l,E1l),_mm256_add_epi32(E2l,E3l));
+  O6h=_mm256_add_epi32(_mm256_add_epi32(E0h,E1h),_mm256_add_epi32(E2h,E3h));
+
+  E0l=_mm256_madd_epi16(m0,T07);E0h=_mm256_madd_epi16(m1,T07);E1l=_mm256_madd_epi16(m2,T17);E1h=_mm256_madd_epi16(m3,T17);
+  E2l=_mm256_madd_epi16(m4,T27);E2h=_mm256_madd_epi16(m5,T27);E3l=_mm256_madd_epi16(m6,T37);E3h=_mm256_madd_epi16(m7,T37);
+  O7l=_mm256_add_epi32(_mm256_add_epi32(E0l,E1l),_mm256_add_epi32(E2l,E3l));
+  O7h=_mm256_add_epi32(_mm256_add_epi32(E0h,E1h),_mm256_add_epi32(E2h,E3h));
+
+  // even part
+  m0=_mm256_unpacklo_epi16(S2,S6);  E0l=_mm256_madd_epi16(m0,U00);
+  m1=_mm256_unpackhi_epi16(S2,S6);  E0h=_mm256_madd_epi16(m1,U00);
+  m2=_mm256_unpacklo_epi16(S10,S14);E0l=_mm256_add_epi32(E0l,_mm256_madd_epi16(m2,U10));
+  m3=_mm256_unpackhi_epi16(S10,S14);E0h=_mm256_add_epi32(E0h,_mm256_madd_epi16(m3,U10));
+  E1l=_mm256_madd_epi16(m0,U01);E1h=_mm256_madd_epi16(m1,U01);
+  E1l=_mm256_add_epi32(E1l,_mm256_madd_epi16(m2,U11));E1h=_mm256_add_epi32(E1h,_mm256_madd_epi16(m3,U11));
+  E2l=_mm256_madd_epi16(m0,U02);E2h=_mm256_madd_epi16(m1,U02);
+  E2l=_mm256_add_epi32(E2l,_mm256_madd_epi16(m2,U12));E2h=_mm256_add_epi32(E2h,_mm256_madd_epi16(m3,U12));
+  E3l=_mm256_madd_epi16(m0,U03);E3h=_mm256_madd_epi16(m1,U03);
+  E3l=_mm256_add_epi32(E3l,_mm256_madd_epi16(m2,U13));E3h=_mm256_add_epi32(E3h,_mm256_madd_epi16(m3,U13));
+
+  m0=_mm256_unpacklo_epi16(S4,S12); E00l=_mm256_madd_epi16(m0,V00);
+  m1=_mm256_unpackhi_epi16(S4,S12); E00h=_mm256_madd_epi16(m1,V00);
+  m2=_mm256_unpacklo_epi16(S0,S8);  EE0l=_mm256_madd_epi16(m2,V10);
+  m3=_mm256_unpackhi_epi16(S0,S8);  EE0h=_mm256_madd_epi16(m3,V10);
+  E01l=_mm256_madd_epi16(m0,V01);E01h=_mm256_madd_epi16(m1,V01);
+  EE1l=_mm256_madd_epi16(m2,V11);EE1h=_mm256_madd_epi16(m3,V11);
+
+  EE2l=_mm256_sub_epi32(EE1l,E01l);EE3l=_mm256_sub_epi32(EE0l,E00l);
+  EE2h=_mm256_sub_epi32(EE1h,E01h);EE3h=_mm256_sub_epi32(EE0h,E00h);
+  EE0l=_mm256_add_epi32(EE0l,E00l);EE1l=_mm256_add_epi32(EE1l,E01l);
+  EE0h=_mm256_add_epi32(EE0h,E00h);EE1h=_mm256_add_epi32(EE1h,E01h);
+
+  E4l=_mm256_sub_epi32(EE3l,E3l); E5l=_mm256_sub_epi32(EE2l,E2l);
+  E6l=_mm256_sub_epi32(EE1l,E1l); E7l=_mm256_sub_epi32(EE0l,E0l);
+  E4h=_mm256_sub_epi32(EE3h,E3h); E5h=_mm256_sub_epi32(EE2h,E2h);
+  E6h=_mm256_sub_epi32(EE1h,E1h); E7h=_mm256_sub_epi32(EE0h,E0h);
+  E0l=_mm256_add_epi32(EE0l,E0l); E1l=_mm256_add_epi32(EE1l,E1l);
+  E2l=_mm256_add_epi32(EE2l,E2l); E3l=_mm256_add_epi32(EE3l,E3l);
+  E0h=_mm256_add_epi32(EE0h,E0h); E1h=_mm256_add_epi32(EE1h,E1h);
+  E2h=_mm256_add_epi32(EE2h,E2h); E3h=_mm256_add_epi32(EE3h,E3h);
+
+  resL0=_mm256_add_epi32(E0l,O0l); resH0=_mm256_add_epi32(E0h,O0h);
+  resL1=_mm256_add_epi32(E1l,O1l); resH1=_mm256_add_epi32(E1h,O1h);
+  resL2=_mm256_add_epi32(E2l,O2l); resH2=_mm256_add_epi32(E2h,O2h);
+  resL3=_mm256_add_epi32(E3l,O3l); resH3=_mm256_add_epi32(E3h,O3h);
+  resL4=_mm256_add_epi32(E4l,O4l); resH4=_mm256_add_epi32(E4h,O4h);
+  resL5=_mm256_add_epi32(E5l,O5l); resH5=_mm256_add_epi32(E5h,O5h);
+  resL6=_mm256_add_epi32(E6l,O6l); resH6=_mm256_add_epi32(E6h,O6h);
+  resL7=_mm256_add_epi32(E7l,O7l); resH7=_mm256_add_epi32(E7h,O7h);
+  resL15=_mm256_sub_epi32(E0l,O0l); resH15=_mm256_sub_epi32(E0h,O0h);
+  resL14=_mm256_sub_epi32(E1l,O1l); resH14=_mm256_sub_epi32(E1h,O1h);
+  resL13=_mm256_sub_epi32(E2l,O2l); resH13=_mm256_sub_epi32(E2h,O2h);
+  resL12=_mm256_sub_epi32(E3l,O3l); resH12=_mm256_sub_epi32(E3h,O3h);
+  resL11=_mm256_sub_epi32(E4l,O4l); resH11=_mm256_sub_epi32(E4h,O4h);
+  resL10=_mm256_sub_epi32(E5l,O5l); resH10=_mm256_sub_epi32(E5h,O5h);
+  resL9 =_mm256_sub_epi32(E6l,O6l); resH9 =_mm256_sub_epi32(E6h,O6h);
+  resL8 =_mm256_sub_epi32(E7l,O7l); resH8 =_mm256_sub_epi32(E7h,O7h);
+}
+
+static inline __m256i round_shift_pack(__m256i lo, __m256i hi, __m256i vr, int shift) {
+  return _mm256_packs_epi32(
+      _mm256_srai_epi32(_mm256_add_epi32(lo,vr),shift),
+      _mm256_srai_epi32(_mm256_add_epi32(hi,vr),shift));
+}
+
+// 1-D 16-point inverse transform in place on S0..15 (16 columns).
+static inline void idct16_vpass(__m256i S16, int add, int shift) {
+  __m256i resL16, resH16;
+  idct16_core(S, resL, resH);
+  const __m256i vr = _mm256_set1_epi32(add);
+  for (int k=0;k<16;k++) Sk = round_shift_pack(resLk, resHk, vr, shift);
+}
+
+// 1-D 32-point inverse transform in place on S0..31 (16 columns). Even part
+// reuses idct16_core on the even rows; odd part uses the T32 coefficients.
+static inline void idct32_vpass(__m256i S32, int add, int shift) {
+  __m256i ev16;
+  for (int i=0;i<16;i++) evi=S2*i;
+  __m256i EL16, EH16;
+  idct16_core(ev, EL, EH);
+
+  __m256i pl8, ph8;
+  for (int g=0; g<8; g++) {
+    plg=_mm256_unpacklo_epi16(S4*g+1, S4*g+3);
+    phg=_mm256_unpackhi_epi16(S4*g+1, S4*g+3);
+  }
+
+  const __m256i vr = _mm256_set1_epi32(add);
+  for (int k=0;k<16;k++) {
+    __m256i OL=_mm256_madd_epi16(pl0, bc(idct_T320k));
+    __m256i OH=_mm256_madd_epi16(ph0, bc(idct_T320k));
+    for (int g=1; g<8; g++) {
+      OL=_mm256_add_epi32(OL,_mm256_madd_epi16(plg, bc(idct_T32gk)));
+      OH=_mm256_add_epi32(OH,_mm256_madd_epi16(phg, bc(idct_T32gk)));
+    }
+    Sk    = round_shift_pack(_mm256_add_epi32(ELk,OL), _mm256_add_epi32(EHk,OH), vr, shift);
+    S31-k = round_shift_pack(_mm256_sub_epi32(ELk,OL), _mm256_sub_epi32(EHk,OH), vr, shift);
+  }
+}
+
+} // namespace
+
+
+void transform_16x16_add_8_avx2(uint8_t *dst, const int16_t *coeffs, ptrdiff_t stride)
+{
+  __m256i S16;
+  for (int i=0;i<16;i++) Si=_mm256_loadu_si256((const __m256i*)(coeffs + i*16));
+
+  idct16_vpass(S, add_1st, shift_1st);
+  transpose16x16(S);
+  idct16_vpass(S, add_2nd, shift_2nd);
+  transpose16x16(S);
+
+  for (int r=0;r<16;r++) {
+    uint8_t* d = dst + r*stride;
+    __m256i pred = _mm256_cvtepu8_epi16(_mm_loadu_si128((const __m128i*)d));
+    __m256i sum  = _mm256_adds_epi16(Sr, pred);
+    __m256i pk   = _mm256_packus_epi16(sum, sum);          // c0-7|c0-7 ; c8-15|c8-15
+    pk = _mm256_permute4x64_epi64(pk, 0x08);               // low 128 = c0-7,c8-15
+    _mm_storeu_si128((__m128i*)d, _mm256_castsi256_si128(pk));
+  }
+}
+
+
+void transform_32x32_add_8_avx2(uint8_t *dst, const int16_t *coeffs, ptrdiff_t stride)
+{
+  __m256i Lo32, Hi32;
+  for (int r=0;r<32;r++) {
+    Lor=_mm256_loadu_si256((const __m256i*)(coeffs + r*32));
+    Hir=_mm256_loadu_si256((const __m256i*)(coeffs + r*32 + 16));
+  }
+
+  idct32_vpass(Lo, add_1st, shift_1st);
+  idct32_vpass(Hi, add_1st, shift_1st);
+  transpose32x32(Lo, Hi);
+  idct32_vpass(Lo, add_2nd, shift_2nd);
+  idct32_vpass(Hi, add_2nd, shift_2nd);
+  transpose32x32(Lo, Hi);
+
+  for (int r=0;r<32;r++) {
+    uint8_t* d = dst + r*stride;
+    __m256i predLo = _mm256_cvtepu8_epi16(_mm_loadu_si128((const __m128i*)d));
+    __m256i predHi = _mm256_cvtepu8_epi16(_mm_loadu_si128((const __m128i*)(d+16)));
+    __m256i sumLo  = _mm256_adds_epi16(Lor, predLo);
+    __m256i sumHi  = _mm256_adds_epi16(Hir, predHi);
+    // pack to bytes: qwords come out as c0-7, c16-23, c8-15, c24-31
+    __m256i pk = _mm256_packus_epi16(sumLo, sumHi);
+    pk = _mm256_permute4x64_epi64(pk, _MM_SHUFFLE(3,1,2,0));  // -> c0-31
+    _mm256_storeu_si256((__m256i*)d, pk);
+  }
+}
+
+// Note: an AVX2 inverse-quantization kernel was implemented and benchmarked too,
+// but inverse quantization is scatter-bound (the result is scattered to
+// coeffBufcoeffPosi, and there is no 16-bit SIMD scatter), so the wider
+// arithmetic gave no benefit -- AVX2 measured ~equal-to-slightly-slower than the
+// SSE version. The SSE version (dequant_coeff_block_sse4) is used instead.
+
+#endif // HAVE_AVX2

libde265-1.1.1.tar.gz/libde265/x86/transform-avx2.h Added

@@ -0,0 +1,33 @@
+/*
+ * H.265 video codec.
+ * Copyright (c) 2026 Dirk Farin <dirk.farin@gmail.com>
+ *
+ * This file is part of libde265.
+ *
+ * libde265 is free software: you can redistribute it and/or modify
+ * it under the terms of the GNU Lesser General Public License as
+ * published by the Free Software Foundation, either version 3 of
+ * the License, or (at your option) any later version.
+ *
+ * libde265 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 Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public License
+ * along with libde265.  If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#ifndef SSE_TRANSFORM_AVX2_H
+#define SSE_TRANSFORM_AVX2_H
+
+#include <stddef.h>
+#include <stdint.h>
+
+// AVX2 inverse-DCT + add. Bit-identical to the SSE4.1 / scalar versions
+// (verified by dev-tools/test-transform). Same signatures as transform_add_8.
+
+void transform_16x16_add_8_avx2(uint8_t *dst, const int16_t *coeffs, ptrdiff_t stride);
+void transform_32x32_add_8_avx2(uint8_t *dst, const int16_t *coeffs, ptrdiff_t stride);
+
+#endif

libde265-1.1.1.tar.gz/libde265/x86/transform-avx512.cc Added

@@ -0,0 +1,275 @@
+/*
+ * H.265 video codec.
+ * Copyright (c) 2026 Dirk Farin <dirk.farin@gmail.com>
+ *
+ * This file is part of libde265.
+ *
+ * libde265 is free software: you can redistribute it and/or modify
+ * it under the terms of the GNU Lesser General Public License as
+ * published by the Free Software Foundation, either version 3 of
+ * the License, or (at your option) any later version.
+ *
+ * libde265 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 Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public License
+ * along with libde265.  If not, see <http://www.gnu.org/licenses/>.
+ */
+
+// AVX-512 (F + BW) inverse transform for 32x32, 8-bit.
+//
+// A full 32-column row fits in one zmm (32 int16), so the vertical pass
+// processes all 32 columns at once (twice the AVX2 width). The per-column
+// butterfly is the same arithmetic as the AVX2/SSE versions: every 512-bit op
+// acts independently per 128-bit lane, so each of the 4 lanes does what the SSE
+// does for 8 columns -> 32 columns, bit-identical. The even part reuses a
+// 16-point core (a 32-pt inverse DCT's even half is a 16-pt inverse DCT of the
+// even rows). The 32x32 transpose between passes is done at 256-bit width using
+// the proven AVX2 path (split each zmm into its two ymm halves, transpose,
+// recombine) since the transpose is shuffle-bound, not the hot path.
+
+#include "x86/transform-avx512.h"
+
+#ifdef HAVE_CONFIG_H
+#include "config.h"
+#endif
+
+#if HAVE_AVX512
+
+#include <immintrin.h>
+#include "x86/transform-dct-tables.h"   // idct_T16_{1,2,3}, idct_T32
+
+namespace {
+
+#define shift_1st 7
+#define add_1st   (1 << (shift_1st - 1))   // 64
+#define shift_2nd 12
+#define add_2nd   (1 << (shift_2nd - 1))   // 2048
+
+// broadcast a 128-bit (8 int16) madd pattern to all four 128-bit lanes
+static inline __m512i bc(const int16_t* p) {
+  return _mm512_broadcast_i32x4(_mm_load_si128((const __m128i*)p));
+}
+
+// --- 256-bit transpose helpers (identical to transform-avx2.cc) ----------
+
+static inline void transpose8x8_lanes(__m256i a8) {
+  __m256i b0=_mm256_unpacklo_epi16(a0,a1), b1=_mm256_unpackhi_epi16(a0,a1);
+  __m256i b2=_mm256_unpacklo_epi16(a2,a3), b3=_mm256_unpackhi_epi16(a2,a3);
+  __m256i b4=_mm256_unpacklo_epi16(a4,a5), b5=_mm256_unpackhi_epi16(a4,a5);
+  __m256i b6=_mm256_unpacklo_epi16(a6,a7), b7=_mm256_unpackhi_epi16(a6,a7);
+  __m256i c0=_mm256_unpacklo_epi32(b0,b2), c1=_mm256_unpackhi_epi32(b0,b2);
+  __m256i c2=_mm256_unpacklo_epi32(b1,b3), c3=_mm256_unpackhi_epi32(b1,b3);
+  __m256i c4=_mm256_unpacklo_epi32(b4,b6), c5=_mm256_unpackhi_epi32(b4,b6);
+  __m256i c6=_mm256_unpacklo_epi32(b5,b7), c7=_mm256_unpackhi_epi32(b5,b7);
+  a0=_mm256_unpacklo_epi64(c0,c4); a1=_mm256_unpackhi_epi64(c0,c4);
+  a2=_mm256_unpacklo_epi64(c1,c5); a3=_mm256_unpackhi_epi64(c1,c5);
+  a4=_mm256_unpacklo_epi64(c2,c6); a5=_mm256_unpackhi_epi64(c2,c6);
+  a6=_mm256_unpacklo_epi64(c3,c7); a7=_mm256_unpackhi_epi64(c3,c7);
+}
+static inline void transpose16x16(__m256i S16) {
+  __m256i top8, bot8;
+  for (int r=0;r<8;r++){ topr=Sr; botr=Sr+8; }
+  transpose8x8_lanes(top); transpose8x8_lanes(bot);
+  for (int r=0;r<8;r++){
+    Sr   = _mm256_permute2x128_si256(topr, botr, 0x20);
+    Sr+8 = _mm256_permute2x128_si256(topr, botr, 0x31);
+  }
+}
+static inline void transpose32x32_ymm(__m256i Lo32, __m256i Hi32) {
+  __m256i A16,B16,C16,D16;
+  for (int r=0;r<16;r++){ Ar=Lor; Br=Hir; Cr=Lor+16; Dr=Hir+16; }
+  transpose16x16(A); transpose16x16(B); transpose16x16(C); transpose16x16(D);
+  for (int r=0;r<16;r++){ Lor=Ar; Hir=Cr; Lor+16=Br; Hir+16=Dr; }
+}
+
+// 32x32 transpose of zmm rows, via the 256-bit path.
+static inline void transpose32x32_z(__m512i S32) {
+  __m256i Lo32, Hi32;
+  for (int r=0;r<32;r++){
+    Lor=_mm512_castsi512_si256(Sr);
+    Hir=_mm512_extracti64x4_epi64(Sr, 1);
+  }
+  transpose32x32_ymm(Lo, Hi);
+  for (int r=0;r<32;r++)
+    Sr=_mm512_inserti64x4(_mm512_castsi256_si512(Lor), Hir, 1);
+}
+
+// --- 16-point core (32 columns) ------------------------------------------
+
+static inline void idct16_core_512(const __m512i S16, __m512i resL16, __m512i resH16) {
+  const __m512i T00=bc(idct_T16_100),T01=bc(idct_T16_101),T02=bc(idct_T16_102),T03=bc(idct_T16_103);
+  const __m512i T04=bc(idct_T16_104),T05=bc(idct_T16_105),T06=bc(idct_T16_106),T07=bc(idct_T16_107);
+  const __m512i T10=bc(idct_T16_110),T11=bc(idct_T16_111),T12=bc(idct_T16_112),T13=bc(idct_T16_113);
+  const __m512i T14=bc(idct_T16_114),T15=bc(idct_T16_115),T16=bc(idct_T16_116),T17=bc(idct_T16_117);
+  const __m512i T20=bc(idct_T16_120),T21=bc(idct_T16_121),T22=bc(idct_T16_122),T23=bc(idct_T16_123);
+  const __m512i T24=bc(idct_T16_124),T25=bc(idct_T16_125),T26=bc(idct_T16_126),T27=bc(idct_T16_127);
+  const __m512i T30=bc(idct_T16_130),T31=bc(idct_T16_131),T32_=bc(idct_T16_132),T33=bc(idct_T16_133);
+  const __m512i T34=bc(idct_T16_134),T35=bc(idct_T16_135),T36=bc(idct_T16_136),T37=bc(idct_T16_137);
+  const __m512i U00=bc(idct_T16_200),U01=bc(idct_T16_201),U02=bc(idct_T16_202),U03=bc(idct_T16_203);
+  const __m512i U10=bc(idct_T16_210),U11=bc(idct_T16_211),U12=bc(idct_T16_212),U13=bc(idct_T16_213);
+  const __m512i V00=bc(idct_T16_300),V01=bc(idct_T16_301),V10=bc(idct_T16_310),V11=bc(idct_T16_311);
+
+  __m512i m0,m1,m2,m3,m4,m5,m6,m7;
+  __m512i E0l,E1l,E2l,E3l,E0h,E1h,E2h,E3h;
+  __m512i O0l,O1l,O2l,O3l,O4l,O5l,O6l,O7l,O0h,O1h,O2h,O3h,O4h,O5h,O6h,O7h;
+  __m512i E00l,E01l,E00h,E01h,EE0l,EE1l,EE2l,EE3l,EE0h,EE1h,EE2h,EE3h;
+  __m512i E4l,E5l,E6l,E7l,E4h,E5h,E6h,E7h;
+
+  m0=_mm512_unpacklo_epi16(S1,S3);  E0l=_mm512_madd_epi16(m0,T00);
+  m1=_mm512_unpackhi_epi16(S1,S3);  E0h=_mm512_madd_epi16(m1,T00);
+  m2=_mm512_unpacklo_epi16(S5,S7);  E1l=_mm512_madd_epi16(m2,T10);
+  m3=_mm512_unpackhi_epi16(S5,S7);  E1h=_mm512_madd_epi16(m3,T10);
+  m4=_mm512_unpacklo_epi16(S9,S11); E2l=_mm512_madd_epi16(m4,T20);
+  m5=_mm512_unpackhi_epi16(S9,S11); E2h=_mm512_madd_epi16(m5,T20);
+  m6=_mm512_unpacklo_epi16(S13,S15);E3l=_mm512_madd_epi16(m6,T30);
+  m7=_mm512_unpackhi_epi16(S13,S15);E3h=_mm512_madd_epi16(m7,T30);
+  O0l=_mm512_add_epi32(_mm512_add_epi32(E0l,E1l),_mm512_add_epi32(E2l,E3l));
+  O0h=_mm512_add_epi32(_mm512_add_epi32(E0h,E1h),_mm512_add_epi32(E2h,E3h));
+
+  E0l=_mm512_madd_epi16(m0,T01);E0h=_mm512_madd_epi16(m1,T01);E1l=_mm512_madd_epi16(m2,T11);E1h=_mm512_madd_epi16(m3,T11);
+  E2l=_mm512_madd_epi16(m4,T21);E2h=_mm512_madd_epi16(m5,T21);E3l=_mm512_madd_epi16(m6,T31);E3h=_mm512_madd_epi16(m7,T31);
+  O1l=_mm512_add_epi32(_mm512_add_epi32(E0l,E1l),_mm512_add_epi32(E2l,E3l));
+  O1h=_mm512_add_epi32(_mm512_add_epi32(E0h,E1h),_mm512_add_epi32(E2h,E3h));
+
+  E0l=_mm512_madd_epi16(m0,T02);E0h=_mm512_madd_epi16(m1,T02);E1l=_mm512_madd_epi16(m2,T12);E1h=_mm512_madd_epi16(m3,T12);
+  E2l=_mm512_madd_epi16(m4,T22);E2h=_mm512_madd_epi16(m5,T22);E3l=_mm512_madd_epi16(m6,T32_);E3h=_mm512_madd_epi16(m7,T32_);
+  O2l=_mm512_add_epi32(_mm512_add_epi32(E0l,E1l),_mm512_add_epi32(E2l,E3l));
+  O2h=_mm512_add_epi32(_mm512_add_epi32(E0h,E1h),_mm512_add_epi32(E2h,E3h));
+
+  E0l=_mm512_madd_epi16(m0,T03);E0h=_mm512_madd_epi16(m1,T03);E1l=_mm512_madd_epi16(m2,T13);E1h=_mm512_madd_epi16(m3,T13);
+  E2l=_mm512_madd_epi16(m4,T23);E2h=_mm512_madd_epi16(m5,T23);E3l=_mm512_madd_epi16(m6,T33);E3h=_mm512_madd_epi16(m7,T33);
+  O3l=_mm512_add_epi32(_mm512_add_epi32(E0l,E1l),_mm512_add_epi32(E2l,E3l));
+  O3h=_mm512_add_epi32(_mm512_add_epi32(E0h,E1h),_mm512_add_epi32(E2h,E3h));
+
+  E0l=_mm512_madd_epi16(m0,T04);E0h=_mm512_madd_epi16(m1,T04);E1l=_mm512_madd_epi16(m2,T14);E1h=_mm512_madd_epi16(m3,T14);
+  E2l=_mm512_madd_epi16(m4,T24);E2h=_mm512_madd_epi16(m5,T24);E3l=_mm512_madd_epi16(m6,T34);E3h=_mm512_madd_epi16(m7,T34);
+  O4l=_mm512_add_epi32(_mm512_add_epi32(E0l,E1l),_mm512_add_epi32(E2l,E3l));
+  O4h=_mm512_add_epi32(_mm512_add_epi32(E0h,E1h),_mm512_add_epi32(E2h,E3h));
+
+  E0l=_mm512_madd_epi16(m0,T05);E0h=_mm512_madd_epi16(m1,T05);E1l=_mm512_madd_epi16(m2,T15);E1h=_mm512_madd_epi16(m3,T15);
+  E2l=_mm512_madd_epi16(m4,T25);E2h=_mm512_madd_epi16(m5,T25);E3l=_mm512_madd_epi16(m6,T35);E3h=_mm512_madd_epi16(m7,T35);
+  O5l=_mm512_add_epi32(_mm512_add_epi32(E0l,E1l),_mm512_add_epi32(E2l,E3l));
+  O5h=_mm512_add_epi32(_mm512_add_epi32(E0h,E1h),_mm512_add_epi32(E2h,E3h));
+
+  E0l=_mm512_madd_epi16(m0,T06);E0h=_mm512_madd_epi16(m1,T06);E1l=_mm512_madd_epi16(m2,T16);E1h=_mm512_madd_epi16(m3,T16);
+  E2l=_mm512_madd_epi16(m4,T26);E2h=_mm512_madd_epi16(m5,T26);E3l=_mm512_madd_epi16(m6,T36);E3h=_mm512_madd_epi16(m7,T36);
+  O6l=_mm512_add_epi32(_mm512_add_epi32(E0l,E1l),_mm512_add_epi32(E2l,E3l));
+  O6h=_mm512_add_epi32(_mm512_add_epi32(E0h,E1h),_mm512_add_epi32(E2h,E3h));
+
+  E0l=_mm512_madd_epi16(m0,T07);E0h=_mm512_madd_epi16(m1,T07);E1l=_mm512_madd_epi16(m2,T17);E1h=_mm512_madd_epi16(m3,T17);
+  E2l=_mm512_madd_epi16(m4,T27);E2h=_mm512_madd_epi16(m5,T27);E3l=_mm512_madd_epi16(m6,T37);E3h=_mm512_madd_epi16(m7,T37);
+  O7l=_mm512_add_epi32(_mm512_add_epi32(E0l,E1l),_mm512_add_epi32(E2l,E3l));
+  O7h=_mm512_add_epi32(_mm512_add_epi32(E0h,E1h),_mm512_add_epi32(E2h,E3h));
+
+  // even part
+  m0=_mm512_unpacklo_epi16(S2,S6);  E0l=_mm512_madd_epi16(m0,U00);
+  m1=_mm512_unpackhi_epi16(S2,S6);  E0h=_mm512_madd_epi16(m1,U00);
+  m2=_mm512_unpacklo_epi16(S10,S14);E0l=_mm512_add_epi32(E0l,_mm512_madd_epi16(m2,U10));
+  m3=_mm512_unpackhi_epi16(S10,S14);E0h=_mm512_add_epi32(E0h,_mm512_madd_epi16(m3,U10));
+  E1l=_mm512_madd_epi16(m0,U01);E1h=_mm512_madd_epi16(m1,U01);
+  E1l=_mm512_add_epi32(E1l,_mm512_madd_epi16(m2,U11));E1h=_mm512_add_epi32(E1h,_mm512_madd_epi16(m3,U11));
+  E2l=_mm512_madd_epi16(m0,U02);E2h=_mm512_madd_epi16(m1,U02);
+  E2l=_mm512_add_epi32(E2l,_mm512_madd_epi16(m2,U12));E2h=_mm512_add_epi32(E2h,_mm512_madd_epi16(m3,U12));
+  E3l=_mm512_madd_epi16(m0,U03);E3h=_mm512_madd_epi16(m1,U03);
+  E3l=_mm512_add_epi32(E3l,_mm512_madd_epi16(m2,U13));E3h=_mm512_add_epi32(E3h,_mm512_madd_epi16(m3,U13));
+
+  m0=_mm512_unpacklo_epi16(S4,S12); E00l=_mm512_madd_epi16(m0,V00);
+  m1=_mm512_unpackhi_epi16(S4,S12); E00h=_mm512_madd_epi16(m1,V00);
+  m2=_mm512_unpacklo_epi16(S0,S8);  EE0l=_mm512_madd_epi16(m2,V10);
+  m3=_mm512_unpackhi_epi16(S0,S8);  EE0h=_mm512_madd_epi16(m3,V10);
+  E01l=_mm512_madd_epi16(m0,V01);E01h=_mm512_madd_epi16(m1,V01);
+  EE1l=_mm512_madd_epi16(m2,V11);EE1h=_mm512_madd_epi16(m3,V11);
+
+  EE2l=_mm512_sub_epi32(EE1l,E01l);EE3l=_mm512_sub_epi32(EE0l,E00l);
+  EE2h=_mm512_sub_epi32(EE1h,E01h);EE3h=_mm512_sub_epi32(EE0h,E00h);
+  EE0l=_mm512_add_epi32(EE0l,E00l);EE1l=_mm512_add_epi32(EE1l,E01l);
+  EE0h=_mm512_add_epi32(EE0h,E00h);EE1h=_mm512_add_epi32(EE1h,E01h);
+
+  E4l=_mm512_sub_epi32(EE3l,E3l); E5l=_mm512_sub_epi32(EE2l,E2l);
+  E6l=_mm512_sub_epi32(EE1l,E1l); E7l=_mm512_sub_epi32(EE0l,E0l);
+  E4h=_mm512_sub_epi32(EE3h,E3h); E5h=_mm512_sub_epi32(EE2h,E2h);
+  E6h=_mm512_sub_epi32(EE1h,E1h); E7h=_mm512_sub_epi32(EE0h,E0h);
+  E0l=_mm512_add_epi32(EE0l,E0l); E1l=_mm512_add_epi32(EE1l,E1l);
+  E2l=_mm512_add_epi32(EE2l,E2l); E3l=_mm512_add_epi32(EE3l,E3l);
+  E0h=_mm512_add_epi32(EE0h,E0h); E1h=_mm512_add_epi32(EE1h,E1h);
+  E2h=_mm512_add_epi32(EE2h,E2h); E3h=_mm512_add_epi32(EE3h,E3h);
+
+  resL0=_mm512_add_epi32(E0l,O0l); resH0=_mm512_add_epi32(E0h,O0h);
+  resL1=_mm512_add_epi32(E1l,O1l); resH1=_mm512_add_epi32(E1h,O1h);
+  resL2=_mm512_add_epi32(E2l,O2l); resH2=_mm512_add_epi32(E2h,O2h);
+  resL3=_mm512_add_epi32(E3l,O3l); resH3=_mm512_add_epi32(E3h,O3h);
+  resL4=_mm512_add_epi32(E4l,O4l); resH4=_mm512_add_epi32(E4h,O4h);
+  resL5=_mm512_add_epi32(E5l,O5l); resH5=_mm512_add_epi32(E5h,O5h);
+  resL6=_mm512_add_epi32(E6l,O6l); resH6=_mm512_add_epi32(E6h,O6h);
+  resL7=_mm512_add_epi32(E7l,O7l); resH7=_mm512_add_epi32(E7h,O7h);
+  resL15=_mm512_sub_epi32(E0l,O0l); resH15=_mm512_sub_epi32(E0h,O0h);
+  resL14=_mm512_sub_epi32(E1l,O1l); resH14=_mm512_sub_epi32(E1h,O1h);
+  resL13=_mm512_sub_epi32(E2l,O2l); resH13=_mm512_sub_epi32(E2h,O2h);
+  resL12=_mm512_sub_epi32(E3l,O3l); resH12=_mm512_sub_epi32(E3h,O3h);
+  resL11=_mm512_sub_epi32(E4l,O4l); resH11=_mm512_sub_epi32(E4h,O4h);
+  resL10=_mm512_sub_epi32(E5l,O5l); resH10=_mm512_sub_epi32(E5h,O5h);
+  resL9 =_mm512_sub_epi32(E6l,O6l); resH9 =_mm512_sub_epi32(E6h,O6h);
+  resL8 =_mm512_sub_epi32(E7l,O7l); resH8 =_mm512_sub_epi32(E7h,O7h);
+}
+
+static inline __m512i round_shift_pack(__m512i lo, __m512i hi, __m512i vr, int shift) {
+  return _mm512_packs_epi32(
+      _mm512_srai_epi32(_mm512_add_epi32(lo,vr),shift),
+      _mm512_srai_epi32(_mm512_add_epi32(hi,vr),shift));
+}
+
+// 1-D 32-point inverse transform in place on S0..31 (32 columns).
+static inline void idct32_vpass_512(__m512i S32, int add, int shift) {
+  __m512i ev16;
+  for (int i=0;i<16;i++) evi=S2*i;
+  __m512i EL16, EH16;
+  idct16_core_512(ev, EL, EH);
+
+  __m512i pl8, ph8;
+  for (int g=0; g<8; g++) {
+    plg=_mm512_unpacklo_epi16(S4*g+1, S4*g+3);
+    phg=_mm512_unpackhi_epi16(S4*g+1, S4*g+3);
+  }
+
+  const __m512i vr = _mm512_set1_epi32(add);
+  for (int k=0;k<16;k++) {
+    __m512i OL=_mm512_madd_epi16(pl0, bc(idct_T320k));
+    __m512i OH=_mm512_madd_epi16(ph0, bc(idct_T320k));
+    for (int g=1; g<8; g++) {
+      OL=_mm512_add_epi32(OL,_mm512_madd_epi16(plg, bc(idct_T32gk)));
+      OH=_mm512_add_epi32(OH,_mm512_madd_epi16(phg, bc(idct_T32gk)));
+    }
+    Sk    = round_shift_pack(_mm512_add_epi32(ELk,OL), _mm512_add_epi32(EHk,OH), vr, shift);
+    S31-k = round_shift_pack(_mm512_sub_epi32(ELk,OL), _mm512_sub_epi32(EHk,OH), vr, shift);
+  }
+}
+
+} // namespace
+
+
+void transform_32x32_add_8_avx512(uint8_t *dst, const int16_t *coeffs, ptrdiff_t stride)
+{
+  __m512i S32;
+  for (int r=0;r<32;r++) Sr=_mm512_loadu_si512((const void*)(coeffs + r*32));
+
+  idct32_vpass_512(S, add_1st, shift_1st);
+  transpose32x32_z(S);
+  idct32_vpass_512(S, add_2nd, shift_2nd);
+  transpose32x32_z(S);
+
+  // index to gather the low qword of each 128-bit lane into a contiguous 256
+  const __m512i gather = _mm512_setr_epi64(0,2,4,6, 1,3,5,7);
+  for (int r=0;r<32;r++) {
+    uint8_t* d = dst + r*stride;
+    __m512i pred = _mm512_cvtepu8_epi16(_mm256_loadu_si256((const __m256i*)d)); // 32 int16
+    __m512i sum  = _mm512_adds_epi16(Sr, pred);
+    __m512i pk   = _mm512_packus_epi16(sum, sum);          // per lane: c|c
+    pk = _mm512_permutexvar_epi64(gather, pk);             // low 256 = c0..c31
+    _mm256_storeu_si256((__m256i*)d, _mm512_castsi512_si256(pk));
+  }
+}
+
+#endif // HAVE_AVX512

libde265-1.1.1.tar.gz/libde265/x86/transform-avx512.h Added

@@ -0,0 +1,32 @@
+/*
+ * H.265 video codec.
+ * Copyright (c) 2026 Dirk Farin <dirk.farin@gmail.com>
+ *
+ * This file is part of libde265.
+ *
+ * libde265 is free software: you can redistribute it and/or modify
+ * it under the terms of the GNU Lesser General Public License as
+ * published by the Free Software Foundation, either version 3 of
+ * the License, or (at your option) any later version.
+ *
+ * libde265 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 Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public License
+ * along with libde265.  If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#ifndef SSE_TRANSFORM_AVX512_H
+#define SSE_TRANSFORM_AVX512_H
+
+#include <stddef.h>
+#include <stdint.h>
+
+// AVX-512 (F+BW) inverse DCT + add. Bit-identical to the SSE4.1 / AVX2 / scalar
+// versions (verified by dev-tools/test-transform). Same signature as
+// transform_add_8.
+void transform_32x32_add_8_avx512(uint8_t *dst, const int16_t *coeffs, ptrdiff_t stride);
+
+#endif

libde265-1.1.1.tar.gz/libde265/x86/transform-dct-tables.h Added

@@ -0,0 +1,88 @@
+/*
+ * H.265 video codec.
+ * Copyright (c) 2026 Dirk Farin <dirk.farin@gmail.com>
+ *
+ * This file is part of libde265.
+ *
+ * libde265 is free software: you can redistribute it and/or modify
+ * it under the terms of the GNU Lesser General Public License as
+ * published by the Free Software Foundation, either version 3 of
+ * the License, or (at your option) any later version.
+ *
+ * libde265 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 Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public License
+ * along with libde265.  If not, see <http://www.gnu.org/licenses/>.
+ */
+
+// Coefficient tables for the AVX2 / AVX-512 inverse transforms. Each inner row
+// holds one madd coefficient pattern (a,b,a,b,...) and is broadcast to all
+// 128-bit lanes of the SIMD register. Values match transform16x16_{1,2,3} /
+// transform32x32 in sse-dct.cc. Included by transform-avx2.cc and
+// transform-avx512.cc (static -> a private copy per translation unit).
+
+#ifndef X86_TRANSFORM_DCT_TABLES_H
+#define X86_TRANSFORM_DCT_TABLES_H
+
+#include <stdint.h>
+
+#ifndef ALIGNED_64
+#define ALIGNED_64(decl) decl __attribute__((aligned(64)))
+#endif
+
+ALIGNED_64(static const int16_t) idct_T16_1488 = {
+  {{ 90, 87, 90, 87, 90, 87, 90, 87},{ 87, 57, 87, 57, 87, 57, 87, 57},
+   { 80,  9, 80,  9, 80,  9, 80,  9},{ 70,-43, 70,-43, 70,-43, 70,-43},
+   { 57,-80, 57,-80, 57,-80, 57,-80},{ 43,-90, 43,-90, 43,-90, 43,-90},
+   { 25,-70, 25,-70, 25,-70, 25,-70},{  9,-25,  9,-25,  9,-25,  9,-25}},
+  {{ 80, 70, 80, 70, 80, 70, 80, 70},{  9,-43,  9,-43,  9,-43,  9,-43},
+   {-70,-87,-70,-87,-70,-87,-70,-87},{-87,  9,-87,  9,-87,  9,-87,  9},
+   {-25, 90,-25, 90,-25, 90,-25, 90},{ 57, 25, 57, 25, 57, 25, 57, 25},
+   { 90,-80, 90,-80, 90,-80, 90,-80},{ 43,-57, 43,-57, 43,-57, 43,-57}},
+  {{ 57, 43, 57, 43, 57, 43, 57, 43},{-80,-90,-80,-90,-80,-90,-80,-90},
+   {-25, 57,-25, 57,-25, 57,-25, 57},{ 90, 25, 90, 25, 90, 25, 90, 25},
+   { -9,-87, -9,-87, -9,-87, -9,-87},{-87, 70,-87, 70,-87, 70,-87, 70},
+   { 43,  9, 43,  9, 43,  9, 43,  9},{ 70,-80, 70,-80, 70,-80, 70,-80}},
+  {{ 25,  9, 25,  9, 25,  9, 25,  9},{-70,-25,-70,-25,-70,-25,-70,-25},
+   { 90, 43, 90, 43, 90, 43, 90, 43},{-80,-57,-80,-57,-80,-57,-80,-57},
+   { 43, 70, 43, 70, 43, 70, 43, 70},{  9,-80,  9,-80,  9,-80,  9,-80},
+   {-57, 87,-57, 87,-57, 87,-57, 87},{ 87,-90, 87,-90, 87,-90, 87,-90}}
+};
+
+ALIGNED_64(static const int16_t) idct_T16_2248 = {
+  {{ 89, 75, 89, 75, 89, 75, 89, 75},{ 75,-18, 75,-18, 75,-18, 75,-18},
+   { 50,-89, 50,-89, 50,-89, 50,-89},{ 18,-50, 18,-50, 18,-50, 18,-50}},
+  {{ 50, 18, 50, 18, 50, 18, 50, 18},{-89,-50,-89,-50,-89,-50,-89,-50},
+   { 18, 75, 18, 75, 18, 75, 18, 75},{ 75,-89, 75,-89, 75,-89, 75,-89}}
+};
+
+ALIGNED_64(static const int16_t) idct_T16_3228 = {
+  {{ 83, 36, 83, 36, 83, 36, 83, 36},{ 36,-83, 36,-83, 36,-83, 36,-83}},
+  {{ 64, 64, 64, 64, 64, 64, 64, 64},{ 64,-64, 64,-64, 64,-64, 64,-64}}
+};
+
+#define IDCT_R8(a,b) { a,b,a,b,a,b,a,b }
+ALIGNED_64(static const int16_t) idct_T328168 = {
+  { IDCT_R8(90,90),IDCT_R8(90,82),IDCT_R8(88,67),IDCT_R8(85,46),IDCT_R8(82,22),IDCT_R8(78,-4),IDCT_R8(73,-31),IDCT_R8(67,-54),
+    IDCT_R8(61,-73),IDCT_R8(54,-85),IDCT_R8(46,-90),IDCT_R8(38,-88),IDCT_R8(31,-78),IDCT_R8(22,-61),IDCT_R8(13,-38),IDCT_R8(4,-13) },
+  { IDCT_R8(88,85),IDCT_R8(67,46),IDCT_R8(31,-13),IDCT_R8(-13,-67),IDCT_R8(-54,-90),IDCT_R8(-82,-73),IDCT_R8(-90,-22),IDCT_R8(-78,38),
+    IDCT_R8(-46,82),IDCT_R8(-4,88),IDCT_R8(38,54),IDCT_R8(73,-4),IDCT_R8(90,-61),IDCT_R8(85,-90),IDCT_R8(61,-78),IDCT_R8(22,-31) },
+  { IDCT_R8(82,78),IDCT_R8(22,-4),IDCT_R8(-54,-82),IDCT_R8(-90,-73),IDCT_R8(-61,13),IDCT_R8(13,85),IDCT_R8(78,67),IDCT_R8(85,-22),
+    IDCT_R8(31,-88),IDCT_R8(-46,-61),IDCT_R8(-90,31),IDCT_R8(-67,90),IDCT_R8(4,54),IDCT_R8(73,-38),IDCT_R8(88,-90),IDCT_R8(38,-46) },
+  { IDCT_R8(73,67),IDCT_R8(-31,-54),IDCT_R8(-90,-78),IDCT_R8(-22,38),IDCT_R8(78,85),IDCT_R8(67,-22),IDCT_R8(-38,-90),IDCT_R8(-90,4),
+    IDCT_R8(-13,90),IDCT_R8(82,13),IDCT_R8(61,-88),IDCT_R8(-46,-31),IDCT_R8(-88,82),IDCT_R8(-4,46),IDCT_R8(85,-73),IDCT_R8(54,-61) },
+  { IDCT_R8(61,54),IDCT_R8(-73,-85),IDCT_R8(-46,-4),IDCT_R8(82,88),IDCT_R8(31,-46),IDCT_R8(-88,-61),IDCT_R8(-13,82),IDCT_R8(90,13),
+    IDCT_R8(-4,-90),IDCT_R8(-90,38),IDCT_R8(22,67),IDCT_R8(85,-78),IDCT_R8(-38,-22),IDCT_R8(-78,90),IDCT_R8(54,-31),IDCT_R8(67,-73) },
+  { IDCT_R8(46,38),IDCT_R8(-90,-88),IDCT_R8(38,73),IDCT_R8(54,-4),IDCT_R8(-90,-67),IDCT_R8(31,90),IDCT_R8(61,-46),IDCT_R8(-88,-31),
+    IDCT_R8(22,85),IDCT_R8(67,-78),IDCT_R8(-85,13),IDCT_R8(13,61),IDCT_R8(73,-90),IDCT_R8(-82,54),IDCT_R8(4,22),IDCT_R8(78,-82) },
+  { IDCT_R8(31,22),IDCT_R8(-78,-61),IDCT_R8(90,85),IDCT_R8(-61,-90),IDCT_R8(4,73),IDCT_R8(54,-38),IDCT_R8(-88,-4),IDCT_R8(82,46),
+    IDCT_R8(-38,-78),IDCT_R8(-22,90),IDCT_R8(73,-82),IDCT_R8(-90,54),IDCT_R8(67,-13),IDCT_R8(-13,-31),IDCT_R8(-46,67),IDCT_R8(85,-88) },
+  { IDCT_R8(13,4),IDCT_R8(-38,-13),IDCT_R8(61,22),IDCT_R8(-78,-31),IDCT_R8(88,38),IDCT_R8(-90,-46),IDCT_R8(85,54),IDCT_R8(-73,-61),
+    IDCT_R8(54,67),IDCT_R8(-31,-73),IDCT_R8(4,78),IDCT_R8(22,-82),IDCT_R8(-46,85),IDCT_R8(67,-88),IDCT_R8(-82,90),IDCT_R8(90,-90) }
+};
+#undef IDCT_R8
+
+#endif

libde265-1.0.17.tar.gz/sherlock265/VideoDecoder.cc -> libde265-1.1.1.tar.gz/sherlock265/VideoDecoder.cc Changed

@@ -25,6 +25,7 @@
 */
 
 #include "VideoDecoder.h"
+#include <chrono>
 #ifdef HAVE_VIDEOGFX
 #include <libvideogfx.hh>
 #endif
@@ -44,6 +45,7 @@
     img(NULL),
     mNextBuffer(0),
     mFrameCount(0),
+    mFramerate(30),
     mPlayingVideo(false),
     mVideoEnded(false),
     mSingleStep(false),
@@ -114,9 +116,15 @@
 
 void VideoDecoder::decoder_loop()
 {
+  using std::chrono::steady_clock;
+  using std::chrono::microseconds;
+  using std::chrono::duration_cast;
+
   for (;;)
     {
       if (mPlayingVideo) {
+        auto frame_start_time = steady_clock::now();
+
         mutex.lock();
 
         if (img) {
@@ -140,8 +148,13 @@
             else if (more && err == DE265_ERROR_WAITING_FOR_INPUT_DATA) {
               uint8_t buf4096;
               int buf_size = fread(buf,1,sizeof(buf),mFH);
-              int err = de265_push_data(ctx,buf,buf_size ,0,0);
-	      (void)err;
+              if (buf_size > 0) {
+                int err = de265_push_data(ctx,buf,buf_size ,0,0);
+                (void)err;
+              }
+              if (feof(mFH)) {
+                de265_flush_data(ctx); // signal end-of-stream so trailing frames are emitted
+              }
             }
             else if (!more)
               {
@@ -168,6 +181,16 @@
         // process events
 
         QCoreApplication::processEvents();
+
+        // Throttle to target frame rate (skip when we are about to go idle anyway)
+        int framerate = mFramerate.load();
+        if (mPlayingVideo && framerate > 0) {
+          auto target_interval = microseconds(1000000 / framerate);
+          auto elapsed = duration_cast<microseconds>(steady_clock::now() - frame_start_time);
+          if (elapsed < target_interval) {
+            QThread::usleep((target_interval - elapsed).count());
+          }
+        }
       }
       else {
         exec();
@@ -430,6 +453,10 @@
 }
 
 
+void VideoDecoder::setFramerate(int framerate)
+{
+  mFramerate = framerate;
+}
 
 
 void VideoDecoder::init_decoder(const char* filename)

libde265-1.0.17.tar.gz/sherlock265/VideoDecoder.h -> libde265-1.1.1.tar.gz/sherlock265/VideoDecoder.h Changed

libde265-1.0.17.tar.gz/sherlock265/VideoPlayer.cc -> libde265-1.1.1.tar.gz/sherlock265/VideoPlayer.cc Changed

Build Results
RPM Lint

Refresh

Mentioned Issues (7)

CVE-2026-45382
CVE-2026-45383
CVE-2026-49295
CVE-2026-49337
CVE-2026-49346
CVE-2026-54240
CVE-2026-54241

Comments for request 6540 0

Request History

olh created request about 2 hours ago

Update to version 1.1.1:

olh accepted request about 2 hours ago