dnn(protobuf): backport AllowUnknownField(), SetRecursionLimit()

- limit recursion in SkipField*() calls original commit 288fa70ed9
dnn(ie): avoid conversion to legacy CNNNetworkImpl
2026-07-21 19:33:03 +04:00 · 2020-04-27 12:25:47 +00:00 · 2020-04-23 22:42:12 +00:00 · 2020-04-22 13:04:33 +00:00 · 2020-04-22 13:04:33 +00:00 · 2020-04-22 13:04:33 +00:00
3061 changed files with 165544 additions and 290311 deletions
@@ -1,15 +1,8 @@
-<!--
-If you have a question rather than reporting a bug please go to https://forum.opencv.org where you get much faster responses.
-If you need further assistance please read [How To Contribute](https://github.com/opencv/opencv/wiki/How_to_contribute).
-
-This is a template helping you to create an issue which can be processed as quickly as possible. This is the bug reporting section for the OpenCV library.
-->
-
 ##### System information (version)
 <!-- Example
- OpenCV => 3.1
+- OpenCV => 4.2
 - Operating System / Platform => Windows 64 Bit
- Compiler => Visual Studio 2015
+- Compiler => Visual Studio 2017
 -->

 - OpenCV => :grey_question:
@@ -28,3 +21,42 @@ This is a template helping you to create an issue which can be processed as quic
    ```
 or attach as .txt or .zip file
 -->
+
+##### Issue submission checklist
+
+ - [ ] I report the issue, it's not a question
+   <!--
+   OpenCV team works with answers.opencv.org, Stack Overflow and other communities
+   to discuss problems. Tickets with question without real issue statement will be
+   closed.
+   -->
+ - [ ] I checked the problem with documentation, FAQ, open issues,
+       answers.opencv.org, Stack Overflow, etc and have not found solution
+   <!--
+   Places to check:
+   * OpenCV documentation: https://docs.opencv.org
+   * FAQ page: https://github.com/opencv/opencv/wiki/FAQ
+   * OpenCV forum: https://answers.opencv.org
+   * OpenCV issue tracker: https://github.com/opencv/opencv/issues?q=is%3Aissue
+   * Stack Overflow branch: https://stackoverflow.com/questions/tagged/opencv
+   -->
+ - [ ] I updated to latest OpenCV version and the issue is still there
+   <!--
+   master branch for OpenCV 4.x and 3.4 branch for OpenCV 3.x releases.
+   OpenCV team supports only latest release for each branch.
+   The ticket is closed, if the problem is not reproduced with modern version.
+   -->
+ - [ ] There is reproducer code and related data files: videos, images, onnx, etc
+   <!--
+   The best reproducer -- test case for OpenCV that we can add to the library.
+   Recommendations for media files and binary files:
+   * Try to reproduce the issue with images and videos in opencv_extra repository
+     to reduce attachment size
+   * Use PNG for images, if you report some CV related bug, but not image reader
+     issue
+   * Attach the image as archite to the ticket, if you report some reader issue.
+     Image hosting services compress images and it breaks the repro code.
+   * Provide ONNX file for some public model or ONNX file with with random weights,
+     if you report ONNX parsing or handling issue. Architecture details diagram
+     from netron tool can be very useful too. See https://lutzroeder.github.io/netron/
+   -->
@@ -1,9 +1,11 @@
-<!-- Please use this line to close one or multiple issues when this pullrequest gets merged
-You can add another line right under the first one:
-resolves #1234
-resolves #1235
-->
+### Pull Request Readiness Checklist

-### This pullrequest changes
+See details at https://github.com/opencv/opencv/wiki/How_to_contribute#making-a-good-pull-request

-<!-- Please describe what your pullrequest is changing -->
+- [ ] I agree to contribute to the project under OpenCV (BSD) License.
+- [ ] To the best of my knowledge, the proposed patch is not based on a code under GPL or other license that is incompatible with OpenCV
+- [ ] The PR is proposed to proper branch
+- [ ] There is reference to original bug report and related work
+- [ ] There is accuracy test, performance test and test data in opencv_extra repository, if applicable
+      Patch to opencv_extra has the same branch name.
+- [ ] The feature is well documented and sample code can be built with the project CMake
@@ -1,31 +0,0 @@
-name: PR:3.4
-
-on:
-  pull_request:
-    branches:
-      - 3.4
-
-jobs:
-  Ubuntu2004-ARM64:
-    uses: opencv/ci-gha-workflow/.github/workflows/OCV-PR-3.4-ARM64.yaml@main
-
-  Ubuntu2004-x64:
-    uses: opencv/ci-gha-workflow/.github/workflows/OCV-PR-3.4-U20.yaml@main
-
-  Windows10-x64:
-    uses: opencv/ci-gha-workflow/.github/workflows/OCV-PR-3.4-W10.yaml@main
-
-  macOS-ARM64:
-    uses: opencv/ci-gha-workflow/.github/workflows/OCV-PR-3.4-macOS-ARM64.yaml@main
-
-  macOS-x64:
-    uses: opencv/ci-gha-workflow/.github/workflows/OCV-PR-3.4-macOS-x86_64.yaml@main
-
-  iOS:
-    uses: opencv/ci-gha-workflow/.github/workflows/OCV-PR-3.4-iOS.yaml@main
-
-  Android:
-    uses: opencv/ci-gha-workflow/.github/workflows/OCV-PR-3.4-Android.yaml@main
-
-  docs:
-    uses: opencv/ci-gha-workflow/.github/workflows/OCV-PR-3.4-docs.yaml@main
@@ -1,4 +1,4 @@
-cmake_minimum_required(VERSION ${MIN_VER_CMAKE} FATAL_ERROR)
+cmake_minimum_required(VERSION 2.8.11 FATAL_ERROR)

 project(Carotene)

@@ -27,11 +27,7 @@ if(CMAKE_COMPILER_IS_GNUCC)
    endif()
 endif()

-if(APPLE AND CV_CLANG AND WITH_NEON)
-    ocv_warnings_disable(CMAKE_CXX_FLAGS -Wno-unused-function)
-endif()
-
-add_library(carotene_objs OBJECT EXCLUDE_FROM_ALL
+add_library(carotene_objs OBJECT
  ${carotene_headers}
  ${carotene_sources}
 )
@@ -44,5 +40,4 @@ if(WITH_NEON)
    target_compile_definitions(carotene_objs PRIVATE "-DWITH_NEON")
 endif()

-# we add dummy file to fix XCode build
-add_library(carotene STATIC ${OPENCV_3RDPARTY_EXCLUDE_FROM_ALL} "$<TARGET_OBJECTS:carotene_objs>" "${CAROTENE_SOURCE_DIR}/dummy.cpp")
+add_library(carotene STATIC EXCLUDE_FROM_ALL "$<TARGET_OBJECTS:carotene_objs>")
@@ -1,4 +1,4 @@
-cmake_minimum_required(VERSION ${MIN_VER_CMAKE} FATAL_ERROR)
+cmake_minimum_required(VERSION 2.8.8 FATAL_ERROR)

 include(CheckCCompilerFlag)
 include(CheckCXXCompilerFlag)
@@ -12,6 +12,8 @@ elseif (CMAKE_SYSTEM_PROCESSOR MATCHES "aarch64.*|AARCH64.*")
  set(AARCH64 TRUE)
 endif()

+ocv_warnings_disable(CMAKE_CXX_FLAGS -Wunused-function)
+
 set(TEGRA_COMPILER_FLAGS "")

 if(CV_GCC OR CV_CLANG)
@@ -80,8 +82,7 @@ set_property(DIRECTORY APPEND PROPERTY COMPILE_DEFINITIONS ${carotene_defs})
 #    set_source_files_properties(impl.cpp $<TARGET_OBJECTS:carotene_objs> COMPILE_FLAGS "--param ipcp-unit-growth=100000 --param inline-unit-growth=100000 --param large-stack-frame-growth=5000")
  endif()

-# we add dummy file to fix XCode build
-add_library(tegra_hal STATIC $<TARGET_OBJECTS:carotene_objs> "dummy.cpp")
+add_library(tegra_hal STATIC $<TARGET_OBJECTS:carotene_objs>)
 set_target_properties(tegra_hal PROPERTIES ARCHIVE_OUTPUT_DIRECTORY ${3P_LIBRARY_OUTPUT_PATH})
 set(OPENCV_SRC_DIR "${CMAKE_SOURCE_DIR}")
 if(NOT BUILD_SHARED_LIBS)
@@ -1,2 +0,0 @@
-// This file is needed for compilation on some platforms e.g. with XCode generator
-// Related issue: https://gitlab.kitware.com/cmake/cmake/-/issues/17457
@@ -1296,13 +1296,13 @@ struct MorphCtx
    CAROTENE_NS::BORDER_MODE border;
    uchar borderValues[4];
 };
-inline int TEGRA_MORPHINIT(cvhalFilter2D **context, int operation, int src_type, int dst_type, int width, int height,
+inline int TEGRA_MORPHINIT(cvhalFilter2D **context, int operation, int src_type, int dst_type, int, int,
                           int kernel_type, uchar *kernel_data, size_t kernel_step, int kernel_width, int kernel_height, int anchor_x, int anchor_y,
                           int borderType, const double borderValue[4], int iterations, bool allowSubmatrix, bool allowInplace)
 {
    if(!context || !kernel_data || src_type != dst_type ||
       CV_MAT_DEPTH(src_type) != CV_8U || src_type < 0 || (src_type >> CV_CN_SHIFT) > 3 ||
-       width < kernel_width || height < kernel_height ||
+
       allowSubmatrix || allowInplace || iterations != 1 ||
       !CAROTENE_NS::isSupportedConfiguration())
        return CV_HAL_ERROR_NOT_IMPLEMENTED;
@@ -1778,30 +1778,30 @@ TegraCvtColor_Invoker(bgrx2hsvf, bgrx2hsv, src_data + static_cast<size_t>(range.
    : CV_HAL_ERROR_NOT_IMPLEMENTED \
 )

-#define TEGRA_CVT2PYUVTOBGR_EX(y_data, y_step, uv_data, uv_step, dst_data, dst_step, dst_width, dst_height, dcn, swapBlue, uIdx) \
+#define TEGRA_CVT2PYUVTOBGR(src_data, src_step, dst_data, dst_step, dst_width, dst_height, dcn, swapBlue, uIdx) \
 ( \
    CAROTENE_NS::isSupportedConfiguration() ? \
        dcn == 3 ? \
            uIdx == 0 ? \
                (swapBlue ? \
                    CAROTENE_NS::yuv420i2rgb(CAROTENE_NS::Size2D(dst_width, dst_height), \
-                                             y_data, y_step, \
-                                             uv_data, uv_step, \
+                                             src_data, src_step, \
+                                             src_data + src_step * dst_height, src_step, \
                                             dst_data, dst_step) : \
                    CAROTENE_NS::yuv420i2bgr(CAROTENE_NS::Size2D(dst_width, dst_height), \
-                                             y_data, y_step, \
-                                             uv_data, uv_step, \
+                                             src_data, src_step, \
+                                             src_data + src_step * dst_height, src_step, \
                                             dst_data, dst_step)), \
                CV_HAL_ERROR_OK : \
            uIdx == 1 ? \
                (swapBlue ? \
                    CAROTENE_NS::yuv420sp2rgb(CAROTENE_NS::Size2D(dst_width, dst_height), \
-                                              y_data, y_step, \
-                                              uv_data, uv_step, \
+                                              src_data, src_step, \
+                                              src_data + src_step * dst_height, src_step, \
                                              dst_data, dst_step) : \
                    CAROTENE_NS::yuv420sp2bgr(CAROTENE_NS::Size2D(dst_width, dst_height), \
-                                              y_data, y_step, \
-                                              uv_data, uv_step, \
+                                              src_data, src_step, \
+                                              src_data + src_step * dst_height, src_step, \
                                              dst_data, dst_step)), \
                CV_HAL_ERROR_OK : \
            CV_HAL_ERROR_NOT_IMPLEMENTED : \
@@ -1809,32 +1809,29 @@ TegraCvtColor_Invoker(bgrx2hsvf, bgrx2hsv, src_data + static_cast<size_t>(range.
            uIdx == 0 ? \
                (swapBlue ? \
                    CAROTENE_NS::yuv420i2rgbx(CAROTENE_NS::Size2D(dst_width, dst_height), \
-                                              y_data, y_step, \
-                                              uv_data, uv_step, \
+                                              src_data, src_step, \
+                                              src_data + src_step * dst_height, src_step, \
                                              dst_data, dst_step) : \
                    CAROTENE_NS::yuv420i2bgrx(CAROTENE_NS::Size2D(dst_width, dst_height), \
-                                              y_data, y_step, \
-                                              uv_data, uv_step, \
+                                              src_data, src_step, \
+                                              src_data + src_step * dst_height, src_step, \
                                              dst_data, dst_step)), \
                CV_HAL_ERROR_OK : \
            uIdx == 1 ? \
                (swapBlue ? \
                    CAROTENE_NS::yuv420sp2rgbx(CAROTENE_NS::Size2D(dst_width, dst_height), \
-                                               y_data, y_step, \
-                                               uv_data, uv_step, \
+                                               src_data, src_step, \
+                                               src_data + src_step * dst_height, src_step, \
                                               dst_data, dst_step) : \
                    CAROTENE_NS::yuv420sp2bgrx(CAROTENE_NS::Size2D(dst_width, dst_height), \
-                                               y_data, y_step, \
-                                               uv_data, uv_step, \
+                                               src_data, src_step, \
+                                               src_data + src_step * dst_height, src_step, \
                                               dst_data, dst_step)), \
                CV_HAL_ERROR_OK : \
            CV_HAL_ERROR_NOT_IMPLEMENTED : \
        CV_HAL_ERROR_NOT_IMPLEMENTED \
    : CV_HAL_ERROR_NOT_IMPLEMENTED \
 )
-#define TEGRA_CVT2PYUVTOBGR(src_data, src_step, dst_data, dst_step, dst_width, dst_height, dcn, swapBlue, uIdx) \
-    TEGRA_CVT2PYUVTOBGR_EX(src_data, src_step, src_data + src_step * dst_height, src_step, dst_data, dst_step, \
-            dst_width, dst_height, dcn, swapBlue, uIdx);

 #undef cv_hal_cvtBGRtoBGR
 #define cv_hal_cvtBGRtoBGR TEGRA_CVTBGRTOBGR
@@ -1844,18 +1841,12 @@ TegraCvtColor_Invoker(bgrx2hsvf, bgrx2hsv, src_data + static_cast<size_t>(range.
 #define cv_hal_cvtBGRtoGray TEGRA_CVTBGRTOGRAY
 #undef cv_hal_cvtGraytoBGR
 #define cv_hal_cvtGraytoBGR TEGRA_CVTGRAYTOBGR
-#if 0  // bit-exact tests are failed
 #undef cv_hal_cvtBGRtoYUV
 #define cv_hal_cvtBGRtoYUV TEGRA_CVTBGRTOYUV
-#endif
 #undef cv_hal_cvtBGRtoHSV
 #define cv_hal_cvtBGRtoHSV TEGRA_CVTBGRTOHSV
-#if 0  // bit-exact tests are failed
 #undef cv_hal_cvtTwoPlaneYUVtoBGR
 #define cv_hal_cvtTwoPlaneYUVtoBGR TEGRA_CVT2PYUVTOBGR
-#undef cv_hal_cvtTwoPlaneYUVtoBGREx
-#define cv_hal_cvtTwoPlaneYUVtoBGREx TEGRA_CVT2PYUVTOBGR_EX
-#endif

 #endif // OPENCV_IMGPROC_HAL_INTERFACE_H

@@ -109,9 +109,9 @@ template <> struct wAdd<s32>
        vgamma = vdupq_n_f32(_gamma + 0.5);
    }

-    void operator() (const VecTraits<s32>::vec128 & v_src0,
-                     const VecTraits<s32>::vec128 & v_src1,
-                     VecTraits<s32>::vec128 & v_dst) const
+    void operator() (const typename VecTraits<s32>::vec128 & v_src0,
+                     const typename VecTraits<s32>::vec128 & v_src1,
+                     typename VecTraits<s32>::vec128 & v_dst) const
    {
        float32x4_t vs1 = vcvtq_f32_s32(v_src0);
        float32x4_t vs2 = vcvtq_f32_s32(v_src1);
@@ -121,9 +121,9 @@ template <> struct wAdd<s32>
        v_dst = vcvtq_s32_f32(vs1);
    }

-    void operator() (const VecTraits<s32>::vec64 & v_src0,
-                     const VecTraits<s32>::vec64 & v_src1,
-                     VecTraits<s32>::vec64 & v_dst) const
+    void operator() (const typename VecTraits<s32>::vec64 & v_src0,
+                     const typename VecTraits<s32>::vec64 & v_src1,
+                     typename VecTraits<s32>::vec64 & v_dst) const
    {
        float32x2_t vs1 = vcvt_f32_s32(v_src0);
        float32x2_t vs2 = vcvt_f32_s32(v_src1);
@@ -153,9 +153,9 @@ template <> struct wAdd<u32>
        vgamma = vdupq_n_f32(_gamma + 0.5);
    }

-    void operator() (const VecTraits<u32>::vec128 & v_src0,
-                     const VecTraits<u32>::vec128 & v_src1,
-                     VecTraits<u32>::vec128 & v_dst) const
+    void operator() (const typename VecTraits<u32>::vec128 & v_src0,
+                     const typename VecTraits<u32>::vec128 & v_src1,
+                     typename VecTraits<u32>::vec128 & v_dst) const
    {
        float32x4_t vs1 = vcvtq_f32_u32(v_src0);
        float32x4_t vs2 = vcvtq_f32_u32(v_src1);
@@ -165,9 +165,9 @@ template <> struct wAdd<u32>
        v_dst = vcvtq_u32_f32(vs1);
    }

-    void operator() (const VecTraits<u32>::vec64 & v_src0,
-                     const VecTraits<u32>::vec64 & v_src1,
-                     VecTraits<u32>::vec64 & v_dst) const
+    void operator() (const typename VecTraits<u32>::vec64 & v_src0,
+                     const typename VecTraits<u32>::vec64 & v_src1,
+                     typename VecTraits<u32>::vec64 & v_dst) const
    {
        float32x2_t vs1 = vcvt_f32_u32(v_src0);
        float32x2_t vs2 = vcvt_f32_u32(v_src1);
@@ -197,17 +197,17 @@ template <> struct wAdd<f32>
        vgamma = vdupq_n_f32(_gamma + 0.5);
    }

-    void operator() (const VecTraits<f32>::vec128 & v_src0,
-                     const VecTraits<f32>::vec128 & v_src1,
-                     VecTraits<f32>::vec128 & v_dst) const
+    void operator() (const typename VecTraits<f32>::vec128 & v_src0,
+                     const typename VecTraits<f32>::vec128 & v_src1,
+                     typename VecTraits<f32>::vec128 & v_dst) const
    {
        float32x4_t vs1 = vmlaq_f32(vgamma, v_src0, valpha);
        v_dst = vmlaq_f32(vs1, v_src1, vbeta);
    }

-    void operator() (const VecTraits<f32>::vec64 & v_src0,
-                     const VecTraits<f32>::vec64 & v_src1,
-                     VecTraits<f32>::vec64 & v_dst) const
+    void operator() (const typename VecTraits<f32>::vec64 & v_src0,
+                     const typename VecTraits<f32>::vec64 & v_src1,
+                     typename VecTraits<f32>::vec64 & v_dst) const
    {
        float32x2_t vs1 = vmla_f32(vget_low(vgamma), v_src0, vget_low(valpha));
        v_dst = vmla_f32(vs1, v_src1, vget_low(vbeta));
@@ -49,12 +49,12 @@ namespace {

 enum
 {
-    SHIFT = 14,
+    SHIFT = 15,
    SHIFT_DELTA = 1 << (SHIFT - 1),

-    R2Y_BT601   = 4899,
-    G2Y_BT601   = 9617,
-    B2Y_BT601   = 1868,
+    R2Y_BT601   = 9798,
+    G2Y_BT601   = 19235,
+    B2Y_BT601   = 3735,

    R2Y_BT709   = 3483,
    G2Y_BT709   = 11718,
@@ -151,6 +151,10 @@ void div(const Size2D &size,
    typedef typename internal::VecTraits<T>::vec128 vec128;
    typedef typename internal::VecTraits<T>::vec64 vec64;

+#if defined(__GNUC__) && (defined(__GXX_EXPERIMENTAL_CXX0X__) || __cplusplus >= 201103L)
+    static_assert(std::numeric_limits<T>::is_integer, "template implementation is for integer types only");
+#endif
+
    if (scale == 0.0f ||
        (std::numeric_limits<T>::is_integer &&
         (scale * std::numeric_limits<T>::max()) <  1.0f &&
@@ -311,6 +315,10 @@ void recip(const Size2D &size,
    typedef typename internal::VecTraits<T>::vec128 vec128;
    typedef typename internal::VecTraits<T>::vec64 vec64;

+#if defined(__GNUC__) && (defined(__GXX_EXPERIMENTAL_CXX0X__) || __cplusplus >= 201103L)
+    static_assert(std::numeric_limits<T>::is_integer, "template implementation is for integer types only");
+#endif
+
    if (scale == 0.0f ||
        (std::numeric_limits<T>::is_integer &&
         scale <  1.0f &&
@@ -463,8 +471,6 @@ void div(const Size2D &size,
        return;
    }

-    float32x4_t v_zero = vdupq_n_f32(0.0f);
-
    size_t roiw128 = size.width >= 3 ? size.width - 3 : 0;
    size_t roiw64 = size.width >= 1 ? size.width - 1 : 0;

@@ -485,9 +491,7 @@ void div(const Size2D &size,
                float32x4_t v_src0 = vld1q_f32(src0 + j);
                float32x4_t v_src1 = vld1q_f32(src1 + j);

-                uint32x4_t v_mask = vceqq_f32(v_src1,v_zero);
-                vst1q_f32(dst + j, vreinterpretq_f32_u32(vbicq_u32(
-                                   vreinterpretq_u32_f32(vmulq_f32(v_src0, internal::vrecpq_f32(v_src1))), v_mask)));
+                vst1q_f32(dst + j, vmulq_f32(v_src0, internal::vrecpq_f32(v_src1)));
            }

            for (; j < roiw64; j += 2)
@@ -495,14 +499,12 @@ void div(const Size2D &size,
                float32x2_t v_src0 = vld1_f32(src0 + j);
                float32x2_t v_src1 = vld1_f32(src1 + j);

-                uint32x2_t v_mask = vceq_f32(v_src1,vget_low_f32(v_zero));
-                vst1_f32(dst + j, vreinterpret_f32_u32(vbic_u32(
-                                  vreinterpret_u32_f32(vmul_f32(v_src0, internal::vrecp_f32(v_src1))), v_mask)));
+                vst1_f32(dst + j, vmul_f32(v_src0, internal::vrecp_f32(v_src1)));
            }

            for (; j < size.width; j++)
            {
-                dst[j] = src1[j] ? src0[j] / src1[j] : 0.0f;
+                dst[j] = src0[j] / src1[j];
            }
        }
    }
@@ -523,10 +525,8 @@ void div(const Size2D &size,
                float32x4_t v_src0 = vld1q_f32(src0 + j);
                float32x4_t v_src1 = vld1q_f32(src1 + j);

-                uint32x4_t v_mask = vceqq_f32(v_src1,v_zero);
-                vst1q_f32(dst + j, vreinterpretq_f32_u32(vbicq_u32(
-                                   vreinterpretq_u32_f32(vmulq_f32(vmulq_n_f32(v_src0, scale),
-                                                         internal::vrecpq_f32(v_src1))), v_mask)));
+                vst1q_f32(dst + j, vmulq_f32(vmulq_n_f32(v_src0, scale),
+                                             internal::vrecpq_f32(v_src1)));
            }

            for (; j < roiw64; j += 2)
@@ -534,15 +534,13 @@ void div(const Size2D &size,
                float32x2_t v_src0 = vld1_f32(src0 + j);
                float32x2_t v_src1 = vld1_f32(src1 + j);

-                uint32x2_t v_mask = vceq_f32(v_src1,vget_low_f32(v_zero));
-                vst1_f32(dst + j, vreinterpret_f32_u32(vbic_u32(
-                                  vreinterpret_u32_f32(vmul_f32(vmul_n_f32(v_src0, scale),
-                                                                internal::vrecp_f32(v_src1))), v_mask)));
+                vst1_f32(dst + j, vmul_f32(vmul_n_f32(v_src0, scale),
+                                           internal::vrecp_f32(v_src1)));
            }

            for (; j < size.width; j++)
            {
-                dst[j] = src1[j] ? src0[j] * scale / src1[j] : 0.0f;
+                dst[j] = src0[j] * scale / src1[j];
            }
        }
    }
@@ -620,8 +618,6 @@ void reciprocal(const Size2D &size,
        return;
    }

-    float32x4_t v_zero = vdupq_n_f32(0.0f);
-
    size_t roiw128 = size.width >= 3 ? size.width - 3 : 0;
    size_t roiw64 = size.width >= 1 ? size.width - 1 : 0;

@@ -639,23 +635,19 @@ void reciprocal(const Size2D &size,

                float32x4_t v_src1 = vld1q_f32(src1 + j);

-                uint32x4_t v_mask = vceqq_f32(v_src1,v_zero);
-                vst1q_f32(dst + j, vreinterpretq_f32_u32(vbicq_u32(
-                                   vreinterpretq_u32_f32(internal::vrecpq_f32(v_src1)), v_mask)));
+                vst1q_f32(dst + j, internal::vrecpq_f32(v_src1));
            }

            for (; j < roiw64; j += 2)
            {
                float32x2_t v_src1 = vld1_f32(src1 + j);

-                uint32x2_t v_mask = vceq_f32(v_src1,vget_low_f32(v_zero));
-                vst1_f32(dst + j, vreinterpret_f32_u32(vbic_u32(
-                                  vreinterpret_u32_f32(internal::vrecp_f32(v_src1)), v_mask)));
+                vst1_f32(dst + j, internal::vrecp_f32(v_src1));
            }

            for (; j < size.width; j++)
            {
-                dst[j] = src1[j] ? 1.0f / src1[j] : 0;
+                dst[j] = 1.0f / src1[j];
            }
        }
    }
@@ -673,25 +665,19 @@ void reciprocal(const Size2D &size,

                float32x4_t v_src1 = vld1q_f32(src1 + j);

-                uint32x4_t v_mask = vceqq_f32(v_src1,v_zero);
-                vst1q_f32(dst + j, vreinterpretq_f32_u32(vbicq_u32(
-                                   vreinterpretq_u32_f32(vmulq_n_f32(internal::vrecpq_f32(v_src1),
-                                                                     scale)),v_mask)));
+                vst1q_f32(dst + j, vmulq_n_f32(internal::vrecpq_f32(v_src1), scale));
            }

            for (; j < roiw64; j += 2)
            {
                float32x2_t v_src1 = vld1_f32(src1 + j);

-                uint32x2_t v_mask = vceq_f32(v_src1,vget_low_f32(v_zero));
-                vst1_f32(dst + j, vreinterpret_f32_u32(vbic_u32(
-                                  vreinterpret_u32_f32(vmul_n_f32(internal::vrecp_f32(v_src1),
-                                                                  scale)), v_mask)));
+                vst1_f32(dst + j, vmul_n_f32(internal::vrecp_f32(v_src1), scale));
            }

            for (; j < size.width; j++)
            {
-                dst[j] = src1[j] ? scale / src1[j] : 0;
+                dst[j] = scale / src1[j];
            }
        }
    }
@@ -1,2 +0,0 @@
-// This file is needed for compilation on some platforms e.g. with XCode generator
-// Related issue: https://gitlab.kitware.com/cmake/cmake/-/issues/17457
@@ -758,7 +758,7 @@ inline void resizeAreaRounding(const Size2D &ssize, const Size2D &dsize,
    }
    else if (channels == 3)
    {
-        if ((wr == 2.0f) && (hr == 2.0f))
+        if ((wr == 2.0f) && (wr == 2.0f))
        {
 #ifndef __ANDROID__
            size_t roiw16 = dsize.width >= 15 ? (dsize.width - 15) * 3 : 0;
@@ -14,7 +14,7 @@ if(NOT DEFINED CPUFEATURES_SOURCES)
 endif()

 include_directories(${CPUFEATURES_INCLUDE_DIRS})
-add_library(${OPENCV_CPUFEATURES_TARGET_NAME} STATIC ${OPENCV_3RDPARTY_EXCLUDE_FROM_ALL} ${CPUFEATURES_SOURCES})
+add_library(${OPENCV_CPUFEATURES_TARGET_NAME} STATIC ${CPUFEATURES_SOURCES})

 set_target_properties(${OPENCV_CPUFEATURES_TARGET_NAME}
  PROPERTIES OUTPUT_NAME cpufeatures
@@ -29,7 +29,7 @@ if(ENABLE_SOLUTION_FOLDERS)
 endif()

 if(NOT BUILD_SHARED_LIBS)
-  ocv_install_target(${OPENCV_CPUFEATURES_TARGET_NAME} EXPORT OpenCVModules ARCHIVE DESTINATION ${OPENCV_3P_LIB_INSTALL_PATH} COMPONENT dev OPTIONAL)
+  ocv_install_target(${OPENCV_CPUFEATURES_TARGET_NAME} EXPORT OpenCVModules ARCHIVE DESTINATION ${OPENCV_3P_LIB_INSTALL_PATH} COMPONENT dev)
 endif()

 ocv_install_3rdparty_licenses(cpufeatures LICENSE README.md)
@@ -30,6 +30,7 @@

 #include <sys/cdefs.h>
 #include <stdint.h>
+#include <string.h>

 __BEGIN_DECLS

@@ -1,6 +1,6 @@
-$url = "https://raw.githubusercontent.com/opencv/opencv_3rdparty/@FFMPEG_BINARIES_COMMIT@/ffmpeg/opencv_ffmpeg_64.dll"
+$url = "https://raw.githubusercontent.com/opencv/opencv_3rdparty/@FFMPEG_BINARIES_COMMIT@/ffmpeg/opencv_videoio_ffmpeg_64.dll"
 $expected_md5 = "@FFMPEG_FILE_HASH_BIN64@"
-$output = "$PSScriptRoot\@OPENCV_BIN_INSTALL_PATH@\opencv_ffmpeg@OPENCV_DLLVERSION@_64.dll"
+$output = "$PSScriptRoot\@OPENCV_BIN_INSTALL_PATH@\opencv_videoio_ffmpeg@OPENCV_DLLVERSION@_64.dll"

 Write-Output ("=" * 120)
 try {
@@ -1,16 +1,16 @@
-# Binaries branch name: 3.4_20230620
-# Binaries were created for OpenCV: c97c22b7cf2ef0f82cd4203a2e9a6eda94e9f7f1
-ocv_update(FFMPEG_BINARIES_COMMIT "7c4bb90fd43a13732ae907981a88fb983a7e2197")
-ocv_update(FFMPEG_FILE_HASH_BIN32 "d7db86de29b0460294489c5ed3180b56")
-ocv_update(FFMPEG_FILE_HASH_BIN64 "9df93d8afff2eee368ad484098a12b18")
-ocv_update(FFMPEG_FILE_HASH_CMAKE "3b90f67f4b429e77d3da36698cef700c")
+# Binaries branch name: ffmpeg/master_20200311
+# Binaries were created for OpenCV: 850414a501d5e6dba111c92d73fdd05794c7061d
+ocv_update(FFMPEG_BINARIES_COMMIT "3d2e97081683265950316c65a52c2e8858ffba1b")
+ocv_update(FFMPEG_FILE_HASH_BIN32 "3b094c37d270a30f0b20a0bc8d3ecafb")
+ocv_update(FFMPEG_FILE_HASH_BIN64 "388ee23a7ca44eef2344e265fafd5940")
+ocv_update(FFMPEG_FILE_HASH_CMAKE "ad57c038ba34b868277ccbe6dd0f9602")

 function(download_win_ffmpeg script_var)
  set(${script_var} "" PARENT_SCOPE)

  set(ids BIN32 BIN64 CMAKE)
-  set(name_BIN32 "opencv_ffmpeg.dll")
-  set(name_BIN64 "opencv_ffmpeg_64.dll")
+  set(name_BIN32 "opencv_videoio_ffmpeg.dll")
+  set(name_BIN64 "opencv_videoio_ffmpeg_64.dll")
  set(name_CMAKE "ffmpeg_version.cmake")

  set(FFMPEG_DOWNLOAD_DIR "${OpenCV_BINARY_DIR}/3rdparty/ffmpeg")
@@ -13,10 +13,10 @@

 * On Windows OpenCV uses pre-built ffmpeg binaries, built with proper flags (without GPL components) and
  wrapped with simple, stable OpenCV-compatible API.
-  The binaries are opencv_ffmpeg.dll (version for 32-bit Windows) and
-  opencv_ffmpeg_64.dll (version for 64-bit Windows).
+  The binaries are opencv_videoio_ffmpeg.dll (version for 32-bit Windows) and
+  opencv_videoio_ffmpeg_64.dll (version for 64-bit Windows).

-  The pre-built opencv_ffmpeg*.dll is:
+  The pre-built opencv_videoio_ffmpeg*.dll is:
  * LGPL library, not BSD libraries.
  * Loaded at runtime by opencv_videoio module.
    If it succeeds, ffmpeg can be used to decode/encode videos;
@@ -30,7 +30,7 @@
  Or you can specify location of binary file via OPENH264_LIBRARY environment variable.

  If LGPL/GPL software can not be supplied with your OpenCV-based product, simply exclude
-  opencv_ffmpeg*.dll from your distribution; OpenCV will stay fully functional except for the ability to
+  opencv_videoio_ffmpeg*.dll from your distribution; OpenCV will stay fully functional except for the ability to
  decode/encode videos using FFMPEG (though, it may still be able to do that using other API,
  such as Video for Windows, Windows Media Foundation or our self-contained motion jpeg codec).

@@ -0,0 +1,92 @@
+//
+// File: vk_platform.h
+//
+/*
+** Copyright (c) 2014-2017 The Khronos Group Inc.
+**
+** Licensed under the Apache License, Version 2.0 (the "License");
+** you may not use this file except in compliance with the License.
+** You may obtain a copy of the License at
+**
+**     http://www.apache.org/licenses/LICENSE-2.0
+**
+** Unless required by applicable law or agreed to in writing, software
+** distributed under the License is distributed on an "AS IS" BASIS,
+** WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+** See the License for the specific language governing permissions and
+** limitations under the License.
+*/
+
+
+#ifndef VK_PLATFORM_H_
+#define VK_PLATFORM_H_
+
+#ifdef __cplusplus
+extern "C"
+{
+#endif // __cplusplus
+
+/*
+***************************************************************************************************
+*   Platform-specific directives and type declarations
+***************************************************************************************************
+*/
+
+/* Platform-specific calling convention macros.
+ *
+ * Platforms should define these so that Vulkan clients call Vulkan commands
+ * with the same calling conventions that the Vulkan implementation expects.
+ *
+ * VKAPI_ATTR - Placed before the return type in function declarations.
+ *              Useful for C++11 and GCC/Clang-style function attribute syntax.
+ * VKAPI_CALL - Placed after the return type in function declarations.
+ *              Useful for MSVC-style calling convention syntax.
+ * VKAPI_PTR  - Placed between the '(' and '*' in function pointer types.
+ *
+ * Function declaration:  VKAPI_ATTR void VKAPI_CALL vkCommand(void);
+ * Function pointer type: typedef void (VKAPI_PTR *PFN_vkCommand)(void);
+ */
+#if defined(_WIN32)
+    // On Windows, Vulkan commands use the stdcall convention
+    #define VKAPI_ATTR
+    #define VKAPI_CALL __stdcall
+    #define VKAPI_PTR  VKAPI_CALL
+#elif defined(__ANDROID__) && defined(__ARM_ARCH) && __ARM_ARCH < 7
+    #error "Vulkan isn't supported for the 'armeabi' NDK ABI"
+#elif defined(__ANDROID__) && defined(__ARM_ARCH) && __ARM_ARCH >= 7 && defined(__ARM_32BIT_STATE)
+    // On Android 32-bit ARM targets, Vulkan functions use the "hardfloat"
+    // calling convention, i.e. float parameters are passed in registers. This
+    // is true even if the rest of the application passes floats on the stack,
+    // as it does by default when compiling for the armeabi-v7a NDK ABI.
+    #define VKAPI_ATTR __attribute__((pcs("aapcs-vfp")))
+    #define VKAPI_CALL
+    #define VKAPI_PTR  VKAPI_ATTR
+#else
+    // On other platforms, use the default calling convention
+    #define VKAPI_ATTR
+    #define VKAPI_CALL
+    #define VKAPI_PTR
+#endif
+
+#include <stddef.h>
+
+#if !defined(VK_NO_STDINT_H)
+    #if defined(_MSC_VER) && (_MSC_VER < 1600)
+        typedef signed   __int8  int8_t;
+        typedef unsigned __int8  uint8_t;
+        typedef signed   __int16 int16_t;
+        typedef unsigned __int16 uint16_t;
+        typedef signed   __int32 int32_t;
+        typedef unsigned __int32 uint32_t;
+        typedef signed   __int64 int64_t;
+        typedef unsigned __int64 uint64_t;
+    #else
+        #include <stdint.h>
+    #endif
+#endif // !defined(VK_NO_STDINT_H)
+
+#ifdef __cplusplus
+} // extern "C"
+#endif // __cplusplus
+
+#endif
@@ -0,0 +1,83 @@
+#ifndef VULKAN_H_
+#define VULKAN_H_ 1
+
+/*
+** Copyright (c) 2015-2018 The Khronos Group Inc.
+**
+** Licensed under the Apache License, Version 2.0 (the "License");
+** you may not use this file except in compliance with the License.
+** You may obtain a copy of the License at
+**
+**     http://www.apache.org/licenses/LICENSE-2.0
+**
+** Unless required by applicable law or agreed to in writing, software
+** distributed under the License is distributed on an "AS IS" BASIS,
+** WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+** See the License for the specific language governing permissions and
+** limitations under the License.
+*/
+
+#include "vk_platform.h"
+#include "vulkan_core.h"
+
+#ifdef VK_USE_PLATFORM_ANDROID_KHR
+#include "vulkan_android.h"
+#endif
+
+#ifdef VK_USE_PLATFORM_FUCHSIA
+#include <zircon/types.h>
+#include "vulkan_fuchsia.h"
+#endif
+
+#ifdef VK_USE_PLATFORM_IOS_MVK
+#include "vulkan_ios.h"
+#endif
+
+
+#ifdef VK_USE_PLATFORM_MACOS_MVK
+#include "vulkan_macos.h"
+#endif
+
+
+#ifdef VK_USE_PLATFORM_MIR_KHR
+#include <mir_toolkit/client_types.h>
+#include "vulkan_mir.h"
+#endif
+
+
+#ifdef VK_USE_PLATFORM_VI_NN
+#include "vulkan_vi.h"
+#endif
+
+
+#ifdef VK_USE_PLATFORM_WAYLAND_KHR
+#include <wayland-client.h>
+#include "vulkan_wayland.h"
+#endif
+
+
+#ifdef VK_USE_PLATFORM_WIN32_KHR
+#include <windows.h>
+#include "vulkan_win32.h"
+#endif
+
+
+#ifdef VK_USE_PLATFORM_XCB_KHR
+#include <xcb/xcb.h>
+#include "vulkan_xcb.h"
+#endif
+
+
+#ifdef VK_USE_PLATFORM_XLIB_KHR
+#include <X11/Xlib.h>
+#include "vulkan_xlib.h"
+#endif
+
+
+#ifdef VK_USE_PLATFORM_XLIB_XRANDR_EXT
+#include <X11/Xlib.h>
+#include <X11/extensions/Xrandr.h>
+#include "vulkan_xlib_xrandr.h"
+#endif
+
+#endif // VULKAN_H_
@@ -0,0 +1,126 @@
+#ifndef VULKAN_ANDROID_H_
+#define VULKAN_ANDROID_H_ 1
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+/*
+** Copyright (c) 2015-2018 The Khronos Group Inc.
+**
+** Licensed under the Apache License, Version 2.0 (the "License");
+** you may not use this file except in compliance with the License.
+** You may obtain a copy of the License at
+**
+**     http://www.apache.org/licenses/LICENSE-2.0
+**
+** Unless required by applicable law or agreed to in writing, software
+** distributed under the License is distributed on an "AS IS" BASIS,
+** WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+** See the License for the specific language governing permissions and
+** limitations under the License.
+*/
+
+/*
+** This header is generated from the Khronos Vulkan XML API Registry.
+**
+*/
+
+
+#define VK_KHR_android_surface 1
+struct ANativeWindow;
+
+#define VK_KHR_ANDROID_SURFACE_SPEC_VERSION 6
+#define VK_KHR_ANDROID_SURFACE_EXTENSION_NAME "VK_KHR_android_surface"
+
+typedef VkFlags VkAndroidSurfaceCreateFlagsKHR;
+
+typedef struct VkAndroidSurfaceCreateInfoKHR {
+    VkStructureType                   sType;
+    const void*                       pNext;
+    VkAndroidSurfaceCreateFlagsKHR    flags;
+    struct ANativeWindow*             window;
+} VkAndroidSurfaceCreateInfoKHR;
+
+
+typedef VkResult (VKAPI_PTR *PFN_vkCreateAndroidSurfaceKHR)(VkInstance instance, const VkAndroidSurfaceCreateInfoKHR* pCreateInfo, const VkAllocationCallbacks* pAllocator, VkSurfaceKHR* pSurface);
+
+#ifndef VK_NO_PROTOTYPES
+VKAPI_ATTR VkResult VKAPI_CALL vkCreateAndroidSurfaceKHR(
+    VkInstance                                  instance,
+    const VkAndroidSurfaceCreateInfoKHR*        pCreateInfo,
+    const VkAllocationCallbacks*                pAllocator,
+    VkSurfaceKHR*                               pSurface);
+#endif
+
+#define VK_ANDROID_external_memory_android_hardware_buffer 1
+struct AHardwareBuffer;
+
+#define VK_ANDROID_EXTERNAL_MEMORY_ANDROID_HARDWARE_BUFFER_SPEC_VERSION 3
+#define VK_ANDROID_EXTERNAL_MEMORY_ANDROID_HARDWARE_BUFFER_EXTENSION_NAME "VK_ANDROID_external_memory_android_hardware_buffer"
+
+typedef struct VkAndroidHardwareBufferUsageANDROID {
+    VkStructureType    sType;
+    void*              pNext;
+    uint64_t           androidHardwareBufferUsage;
+} VkAndroidHardwareBufferUsageANDROID;
+
+typedef struct VkAndroidHardwareBufferPropertiesANDROID {
+    VkStructureType    sType;
+    void*              pNext;
+    VkDeviceSize       allocationSize;
+    uint32_t           memoryTypeBits;
+} VkAndroidHardwareBufferPropertiesANDROID;
+
+typedef struct VkAndroidHardwareBufferFormatPropertiesANDROID {
+    VkStructureType                  sType;
+    void*                            pNext;
+    VkFormat                         format;
+    uint64_t                         externalFormat;
+    VkFormatFeatureFlags             formatFeatures;
+    VkComponentMapping               samplerYcbcrConversionComponents;
+    VkSamplerYcbcrModelConversion    suggestedYcbcrModel;
+    VkSamplerYcbcrRange              suggestedYcbcrRange;
+    VkChromaLocation                 suggestedXChromaOffset;
+    VkChromaLocation                 suggestedYChromaOffset;
+} VkAndroidHardwareBufferFormatPropertiesANDROID;
+
+typedef struct VkImportAndroidHardwareBufferInfoANDROID {
+    VkStructureType            sType;
+    const void*                pNext;
+    struct AHardwareBuffer*    buffer;
+} VkImportAndroidHardwareBufferInfoANDROID;
+
+typedef struct VkMemoryGetAndroidHardwareBufferInfoANDROID {
+    VkStructureType    sType;
+    const void*        pNext;
+    VkDeviceMemory     memory;
+} VkMemoryGetAndroidHardwareBufferInfoANDROID;
+
+typedef struct VkExternalFormatANDROID {
+    VkStructureType    sType;
+    void*              pNext;
+    uint64_t           externalFormat;
+} VkExternalFormatANDROID;
+
+
+typedef VkResult (VKAPI_PTR *PFN_vkGetAndroidHardwareBufferPropertiesANDROID)(VkDevice device, const struct AHardwareBuffer* buffer, VkAndroidHardwareBufferPropertiesANDROID* pProperties);
+typedef VkResult (VKAPI_PTR *PFN_vkGetMemoryAndroidHardwareBufferANDROID)(VkDevice device, const VkMemoryGetAndroidHardwareBufferInfoANDROID* pInfo, struct AHardwareBuffer** pBuffer);
+
+#ifndef VK_NO_PROTOTYPES
+VKAPI_ATTR VkResult VKAPI_CALL vkGetAndroidHardwareBufferPropertiesANDROID(
+    VkDevice                                    device,
+    const struct AHardwareBuffer*               buffer,
+    VkAndroidHardwareBufferPropertiesANDROID*   pProperties);
+
+VKAPI_ATTR VkResult VKAPI_CALL vkGetMemoryAndroidHardwareBufferANDROID(
+    VkDevice                                    device,
+    const VkMemoryGetAndroidHardwareBufferInfoANDROID* pInfo,
+    struct AHardwareBuffer**                    pBuffer);
+#endif
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif
@@ -0,0 +1,58 @@
+#ifndef VULKAN_FUCHSIA_H_
+#define VULKAN_FUCHSIA_H_ 1
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+/*
+** Copyright (c) 2015-2018 The Khronos Group Inc.
+**
+** Licensed under the Apache License, Version 2.0 (the "License");
+** you may not use this file except in compliance with the License.
+** You may obtain a copy of the License at
+**
+**     http://www.apache.org/licenses/LICENSE-2.0
+**
+** Unless required by applicable law or agreed to in writing, software
+** distributed under the License is distributed on an "AS IS" BASIS,
+** WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+** See the License for the specific language governing permissions and
+** limitations under the License.
+*/
+
+/*
+** This header is generated from the Khronos Vulkan XML API Registry.
+**
+*/
+
+
+#define VK_FUCHSIA_imagepipe_surface 1
+#define VK_FUCHSIA_IMAGEPIPE_SURFACE_SPEC_VERSION 1
+#define VK_FUCHSIA_IMAGEPIPE_SURFACE_EXTENSION_NAME "VK_FUCHSIA_imagepipe_surface"
+
+typedef VkFlags VkImagePipeSurfaceCreateFlagsFUCHSIA;
+
+typedef struct VkImagePipeSurfaceCreateInfoFUCHSIA {
+    VkStructureType                         sType;
+    const void*                             pNext;
+    VkImagePipeSurfaceCreateFlagsFUCHSIA    flags;
+    zx_handle_t                             imagePipeHandle;
+} VkImagePipeSurfaceCreateInfoFUCHSIA;
+
+
+typedef VkResult (VKAPI_PTR *PFN_vkCreateImagePipeSurfaceFUCHSIA)(VkInstance instance, const VkImagePipeSurfaceCreateInfoFUCHSIA* pCreateInfo, const VkAllocationCallbacks* pAllocator, VkSurfaceKHR* pSurface);
+
+#ifndef VK_NO_PROTOTYPES
+VKAPI_ATTR VkResult VKAPI_CALL vkCreateImagePipeSurfaceFUCHSIA(
+    VkInstance                                  instance,
+    const VkImagePipeSurfaceCreateInfoFUCHSIA*  pCreateInfo,
+    const VkAllocationCallbacks*                pAllocator,
+    VkSurfaceKHR*                               pSurface);
+#endif
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif
@@ -0,0 +1,58 @@
+#ifndef VULKAN_IOS_H_
+#define VULKAN_IOS_H_ 1
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+/*
+** Copyright (c) 2015-2018 The Khronos Group Inc.
+**
+** Licensed under the Apache License, Version 2.0 (the "License");
+** you may not use this file except in compliance with the License.
+** You may obtain a copy of the License at
+**
+**     http://www.apache.org/licenses/LICENSE-2.0
+**
+** Unless required by applicable law or agreed to in writing, software
+** distributed under the License is distributed on an "AS IS" BASIS,
+** WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+** See the License for the specific language governing permissions and
+** limitations under the License.
+*/
+
+/*
+** This header is generated from the Khronos Vulkan XML API Registry.
+**
+*/
+
+
+#define VK_MVK_ios_surface 1
+#define VK_MVK_IOS_SURFACE_SPEC_VERSION   2
+#define VK_MVK_IOS_SURFACE_EXTENSION_NAME "VK_MVK_ios_surface"
+
+typedef VkFlags VkIOSSurfaceCreateFlagsMVK;
+
+typedef struct VkIOSSurfaceCreateInfoMVK {
+    VkStructureType               sType;
+    const void*                   pNext;
+    VkIOSSurfaceCreateFlagsMVK    flags;
+    const void*                   pView;
+} VkIOSSurfaceCreateInfoMVK;
+
+
+typedef VkResult (VKAPI_PTR *PFN_vkCreateIOSSurfaceMVK)(VkInstance instance, const VkIOSSurfaceCreateInfoMVK* pCreateInfo, const VkAllocationCallbacks* pAllocator, VkSurfaceKHR* pSurface);
+
+#ifndef VK_NO_PROTOTYPES
+VKAPI_ATTR VkResult VKAPI_CALL vkCreateIOSSurfaceMVK(
+    VkInstance                                  instance,
+    const VkIOSSurfaceCreateInfoMVK*            pCreateInfo,
+    const VkAllocationCallbacks*                pAllocator,
+    VkSurfaceKHR*                               pSurface);
+#endif
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif
@@ -0,0 +1,58 @@
+#ifndef VULKAN_MACOS_H_
+#define VULKAN_MACOS_H_ 1
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+/*
+** Copyright (c) 2015-2018 The Khronos Group Inc.
+**
+** Licensed under the Apache License, Version 2.0 (the "License");
+** you may not use this file except in compliance with the License.
+** You may obtain a copy of the License at
+**
+**     http://www.apache.org/licenses/LICENSE-2.0
+**
+** Unless required by applicable law or agreed to in writing, software
+** distributed under the License is distributed on an "AS IS" BASIS,
+** WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+** See the License for the specific language governing permissions and
+** limitations under the License.
+*/
+
+/*
+** This header is generated from the Khronos Vulkan XML API Registry.
+**
+*/
+
+
+#define VK_MVK_macos_surface 1
+#define VK_MVK_MACOS_SURFACE_SPEC_VERSION 2
+#define VK_MVK_MACOS_SURFACE_EXTENSION_NAME "VK_MVK_macos_surface"
+
+typedef VkFlags VkMacOSSurfaceCreateFlagsMVK;
+
+typedef struct VkMacOSSurfaceCreateInfoMVK {
+    VkStructureType                 sType;
+    const void*                     pNext;
+    VkMacOSSurfaceCreateFlagsMVK    flags;
+    const void*                     pView;
+} VkMacOSSurfaceCreateInfoMVK;
+
+
+typedef VkResult (VKAPI_PTR *PFN_vkCreateMacOSSurfaceMVK)(VkInstance instance, const VkMacOSSurfaceCreateInfoMVK* pCreateInfo, const VkAllocationCallbacks* pAllocator, VkSurfaceKHR* pSurface);
+
+#ifndef VK_NO_PROTOTYPES
+VKAPI_ATTR VkResult VKAPI_CALL vkCreateMacOSSurfaceMVK(
+    VkInstance                                  instance,
+    const VkMacOSSurfaceCreateInfoMVK*          pCreateInfo,
+    const VkAllocationCallbacks*                pAllocator,
+    VkSurfaceKHR*                               pSurface);
+#endif
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif
@@ -0,0 +1,65 @@
+#ifndef VULKAN_MIR_H_
+#define VULKAN_MIR_H_ 1
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+/*
+** Copyright (c) 2015-2018 The Khronos Group Inc.
+**
+** Licensed under the Apache License, Version 2.0 (the "License");
+** you may not use this file except in compliance with the License.
+** You may obtain a copy of the License at
+**
+**     http://www.apache.org/licenses/LICENSE-2.0
+**
+** Unless required by applicable law or agreed to in writing, software
+** distributed under the License is distributed on an "AS IS" BASIS,
+** WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+** See the License for the specific language governing permissions and
+** limitations under the License.
+*/
+
+/*
+** This header is generated from the Khronos Vulkan XML API Registry.
+**
+*/
+
+
+#define VK_KHR_mir_surface 1
+#define VK_KHR_MIR_SURFACE_SPEC_VERSION   4
+#define VK_KHR_MIR_SURFACE_EXTENSION_NAME "VK_KHR_mir_surface"
+
+typedef VkFlags VkMirSurfaceCreateFlagsKHR;
+
+typedef struct VkMirSurfaceCreateInfoKHR {
+    VkStructureType               sType;
+    const void*                   pNext;
+    VkMirSurfaceCreateFlagsKHR    flags;
+    MirConnection*                connection;
+    MirSurface*                   mirSurface;
+} VkMirSurfaceCreateInfoKHR;
+
+
+typedef VkResult (VKAPI_PTR *PFN_vkCreateMirSurfaceKHR)(VkInstance instance, const VkMirSurfaceCreateInfoKHR* pCreateInfo, const VkAllocationCallbacks* pAllocator, VkSurfaceKHR* pSurface);
+typedef VkBool32 (VKAPI_PTR *PFN_vkGetPhysicalDeviceMirPresentationSupportKHR)(VkPhysicalDevice physicalDevice, uint32_t queueFamilyIndex, MirConnection* connection);
+
+#ifndef VK_NO_PROTOTYPES
+VKAPI_ATTR VkResult VKAPI_CALL vkCreateMirSurfaceKHR(
+    VkInstance                                  instance,
+    const VkMirSurfaceCreateInfoKHR*            pCreateInfo,
+    const VkAllocationCallbacks*                pAllocator,
+    VkSurfaceKHR*                               pSurface);
+
+VKAPI_ATTR VkBool32 VKAPI_CALL vkGetPhysicalDeviceMirPresentationSupportKHR(
+    VkPhysicalDevice                            physicalDevice,
+    uint32_t                                    queueFamilyIndex,
+    MirConnection*                              connection);
+#endif
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif
@@ -0,0 +1,58 @@
+#ifndef VULKAN_VI_H_
+#define VULKAN_VI_H_ 1
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+/*
+** Copyright (c) 2015-2018 The Khronos Group Inc.
+**
+** Licensed under the Apache License, Version 2.0 (the "License");
+** you may not use this file except in compliance with the License.
+** You may obtain a copy of the License at
+**
+**     http://www.apache.org/licenses/LICENSE-2.0
+**
+** Unless required by applicable law or agreed to in writing, software
+** distributed under the License is distributed on an "AS IS" BASIS,
+** WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+** See the License for the specific language governing permissions and
+** limitations under the License.
+*/
+
+/*
+** This header is generated from the Khronos Vulkan XML API Registry.
+**
+*/
+
+
+#define VK_NN_vi_surface 1
+#define VK_NN_VI_SURFACE_SPEC_VERSION     1
+#define VK_NN_VI_SURFACE_EXTENSION_NAME   "VK_NN_vi_surface"
+
+typedef VkFlags VkViSurfaceCreateFlagsNN;
+
+typedef struct VkViSurfaceCreateInfoNN {
+    VkStructureType             sType;
+    const void*                 pNext;
+    VkViSurfaceCreateFlagsNN    flags;
+    void*                       window;
+} VkViSurfaceCreateInfoNN;
+
+
+typedef VkResult (VKAPI_PTR *PFN_vkCreateViSurfaceNN)(VkInstance instance, const VkViSurfaceCreateInfoNN* pCreateInfo, const VkAllocationCallbacks* pAllocator, VkSurfaceKHR* pSurface);
+
+#ifndef VK_NO_PROTOTYPES
+VKAPI_ATTR VkResult VKAPI_CALL vkCreateViSurfaceNN(
+    VkInstance                                  instance,
+    const VkViSurfaceCreateInfoNN*              pCreateInfo,
+    const VkAllocationCallbacks*                pAllocator,
+    VkSurfaceKHR*                               pSurface);
+#endif
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif
@@ -0,0 +1,65 @@
+#ifndef VULKAN_WAYLAND_H_
+#define VULKAN_WAYLAND_H_ 1
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+/*
+** Copyright (c) 2015-2018 The Khronos Group Inc.
+**
+** Licensed under the Apache License, Version 2.0 (the "License");
+** you may not use this file except in compliance with the License.
+** You may obtain a copy of the License at
+**
+**     http://www.apache.org/licenses/LICENSE-2.0
+**
+** Unless required by applicable law or agreed to in writing, software
+** distributed under the License is distributed on an "AS IS" BASIS,
+** WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+** See the License for the specific language governing permissions and
+** limitations under the License.
+*/
+
+/*
+** This header is generated from the Khronos Vulkan XML API Registry.
+**
+*/
+
+
+#define VK_KHR_wayland_surface 1
+#define VK_KHR_WAYLAND_SURFACE_SPEC_VERSION 6
+#define VK_KHR_WAYLAND_SURFACE_EXTENSION_NAME "VK_KHR_wayland_surface"
+
+typedef VkFlags VkWaylandSurfaceCreateFlagsKHR;
+
+typedef struct VkWaylandSurfaceCreateInfoKHR {
+    VkStructureType                   sType;
+    const void*                       pNext;
+    VkWaylandSurfaceCreateFlagsKHR    flags;
+    struct wl_display*                display;
+    struct wl_surface*                surface;
+} VkWaylandSurfaceCreateInfoKHR;
+
+
+typedef VkResult (VKAPI_PTR *PFN_vkCreateWaylandSurfaceKHR)(VkInstance instance, const VkWaylandSurfaceCreateInfoKHR* pCreateInfo, const VkAllocationCallbacks* pAllocator, VkSurfaceKHR* pSurface);
+typedef VkBool32 (VKAPI_PTR *PFN_vkGetPhysicalDeviceWaylandPresentationSupportKHR)(VkPhysicalDevice physicalDevice, uint32_t queueFamilyIndex, struct wl_display* display);
+
+#ifndef VK_NO_PROTOTYPES
+VKAPI_ATTR VkResult VKAPI_CALL vkCreateWaylandSurfaceKHR(
+    VkInstance                                  instance,
+    const VkWaylandSurfaceCreateInfoKHR*        pCreateInfo,
+    const VkAllocationCallbacks*                pAllocator,
+    VkSurfaceKHR*                               pSurface);
+
+VKAPI_ATTR VkBool32 VKAPI_CALL vkGetPhysicalDeviceWaylandPresentationSupportKHR(
+    VkPhysicalDevice                            physicalDevice,
+    uint32_t                                    queueFamilyIndex,
+    struct wl_display*                          display);
+#endif
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif
@@ -0,0 +1,276 @@
+#ifndef VULKAN_WIN32_H_
+#define VULKAN_WIN32_H_ 1
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+/*
+** Copyright (c) 2015-2018 The Khronos Group Inc.
+**
+** Licensed under the Apache License, Version 2.0 (the "License");
+** you may not use this file except in compliance with the License.
+** You may obtain a copy of the License at
+**
+**     http://www.apache.org/licenses/LICENSE-2.0
+**
+** Unless required by applicable law or agreed to in writing, software
+** distributed under the License is distributed on an "AS IS" BASIS,
+** WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+** See the License for the specific language governing permissions and
+** limitations under the License.
+*/
+
+/*
+** This header is generated from the Khronos Vulkan XML API Registry.
+**
+*/
+
+
+#define VK_KHR_win32_surface 1
+#define VK_KHR_WIN32_SURFACE_SPEC_VERSION 6
+#define VK_KHR_WIN32_SURFACE_EXTENSION_NAME "VK_KHR_win32_surface"
+
+typedef VkFlags VkWin32SurfaceCreateFlagsKHR;
+
+typedef struct VkWin32SurfaceCreateInfoKHR {
+    VkStructureType                 sType;
+    const void*                     pNext;
+    VkWin32SurfaceCreateFlagsKHR    flags;
+    HINSTANCE                       hinstance;
+    HWND                            hwnd;
+} VkWin32SurfaceCreateInfoKHR;
+
+
+typedef VkResult (VKAPI_PTR *PFN_vkCreateWin32SurfaceKHR)(VkInstance instance, const VkWin32SurfaceCreateInfoKHR* pCreateInfo, const VkAllocationCallbacks* pAllocator, VkSurfaceKHR* pSurface);
+typedef VkBool32 (VKAPI_PTR *PFN_vkGetPhysicalDeviceWin32PresentationSupportKHR)(VkPhysicalDevice physicalDevice, uint32_t queueFamilyIndex);
+
+#ifndef VK_NO_PROTOTYPES
+VKAPI_ATTR VkResult VKAPI_CALL vkCreateWin32SurfaceKHR(
+    VkInstance                                  instance,
+    const VkWin32SurfaceCreateInfoKHR*          pCreateInfo,
+    const VkAllocationCallbacks*                pAllocator,
+    VkSurfaceKHR*                               pSurface);
+
+VKAPI_ATTR VkBool32 VKAPI_CALL vkGetPhysicalDeviceWin32PresentationSupportKHR(
+    VkPhysicalDevice                            physicalDevice,
+    uint32_t                                    queueFamilyIndex);
+#endif
+
+#define VK_KHR_external_memory_win32 1
+#define VK_KHR_EXTERNAL_MEMORY_WIN32_SPEC_VERSION 1
+#define VK_KHR_EXTERNAL_MEMORY_WIN32_EXTENSION_NAME "VK_KHR_external_memory_win32"
+
+typedef struct VkImportMemoryWin32HandleInfoKHR {
+    VkStructureType                       sType;
+    const void*                           pNext;
+    VkExternalMemoryHandleTypeFlagBits    handleType;
+    HANDLE                                handle;
+    LPCWSTR                               name;
+} VkImportMemoryWin32HandleInfoKHR;
+
+typedef struct VkExportMemoryWin32HandleInfoKHR {
+    VkStructureType               sType;
+    const void*                   pNext;
+    const SECURITY_ATTRIBUTES*    pAttributes;
+    DWORD                         dwAccess;
+    LPCWSTR                       name;
+} VkExportMemoryWin32HandleInfoKHR;
+
+typedef struct VkMemoryWin32HandlePropertiesKHR {
+    VkStructureType    sType;
+    void*              pNext;
+    uint32_t           memoryTypeBits;
+} VkMemoryWin32HandlePropertiesKHR;
+
+typedef struct VkMemoryGetWin32HandleInfoKHR {
+    VkStructureType                       sType;
+    const void*                           pNext;
+    VkDeviceMemory                        memory;
+    VkExternalMemoryHandleTypeFlagBits    handleType;
+} VkMemoryGetWin32HandleInfoKHR;
+
+
+typedef VkResult (VKAPI_PTR *PFN_vkGetMemoryWin32HandleKHR)(VkDevice device, const VkMemoryGetWin32HandleInfoKHR* pGetWin32HandleInfo, HANDLE* pHandle);
+typedef VkResult (VKAPI_PTR *PFN_vkGetMemoryWin32HandlePropertiesKHR)(VkDevice device, VkExternalMemoryHandleTypeFlagBits handleType, HANDLE handle, VkMemoryWin32HandlePropertiesKHR* pMemoryWin32HandleProperties);
+
+#ifndef VK_NO_PROTOTYPES
+VKAPI_ATTR VkResult VKAPI_CALL vkGetMemoryWin32HandleKHR(
+    VkDevice                                    device,
+    const VkMemoryGetWin32HandleInfoKHR*        pGetWin32HandleInfo,
+    HANDLE*                                     pHandle);
+
+VKAPI_ATTR VkResult VKAPI_CALL vkGetMemoryWin32HandlePropertiesKHR(
+    VkDevice                                    device,
+    VkExternalMemoryHandleTypeFlagBits          handleType,
+    HANDLE                                      handle,
+    VkMemoryWin32HandlePropertiesKHR*           pMemoryWin32HandleProperties);
+#endif
+
+#define VK_KHR_win32_keyed_mutex 1
+#define VK_KHR_WIN32_KEYED_MUTEX_SPEC_VERSION 1
+#define VK_KHR_WIN32_KEYED_MUTEX_EXTENSION_NAME "VK_KHR_win32_keyed_mutex"
+
+typedef struct VkWin32KeyedMutexAcquireReleaseInfoKHR {
+    VkStructureType          sType;
+    const void*              pNext;
+    uint32_t                 acquireCount;
+    const VkDeviceMemory*    pAcquireSyncs;
+    const uint64_t*          pAcquireKeys;
+    const uint32_t*          pAcquireTimeouts;
+    uint32_t                 releaseCount;
+    const VkDeviceMemory*    pReleaseSyncs;
+    const uint64_t*          pReleaseKeys;
+} VkWin32KeyedMutexAcquireReleaseInfoKHR;
+
+
+
+#define VK_KHR_external_semaphore_win32 1
+#define VK_KHR_EXTERNAL_SEMAPHORE_WIN32_SPEC_VERSION 1
+#define VK_KHR_EXTERNAL_SEMAPHORE_WIN32_EXTENSION_NAME "VK_KHR_external_semaphore_win32"
+
+typedef struct VkImportSemaphoreWin32HandleInfoKHR {
+    VkStructureType                          sType;
+    const void*                              pNext;
+    VkSemaphore                              semaphore;
+    VkSemaphoreImportFlags                   flags;
+    VkExternalSemaphoreHandleTypeFlagBits    handleType;
+    HANDLE                                   handle;
+    LPCWSTR                                  name;
+} VkImportSemaphoreWin32HandleInfoKHR;
+
+typedef struct VkExportSemaphoreWin32HandleInfoKHR {
+    VkStructureType               sType;
+    const void*                   pNext;
+    const SECURITY_ATTRIBUTES*    pAttributes;
+    DWORD                         dwAccess;
+    LPCWSTR                       name;
+} VkExportSemaphoreWin32HandleInfoKHR;
+
+typedef struct VkD3D12FenceSubmitInfoKHR {
+    VkStructureType    sType;
+    const void*        pNext;
+    uint32_t           waitSemaphoreValuesCount;
+    const uint64_t*    pWaitSemaphoreValues;
+    uint32_t           signalSemaphoreValuesCount;
+    const uint64_t*    pSignalSemaphoreValues;
+} VkD3D12FenceSubmitInfoKHR;
+
+typedef struct VkSemaphoreGetWin32HandleInfoKHR {
+    VkStructureType                          sType;
+    const void*                              pNext;
+    VkSemaphore                              semaphore;
+    VkExternalSemaphoreHandleTypeFlagBits    handleType;
+} VkSemaphoreGetWin32HandleInfoKHR;
+
+
+typedef VkResult (VKAPI_PTR *PFN_vkImportSemaphoreWin32HandleKHR)(VkDevice device, const VkImportSemaphoreWin32HandleInfoKHR* pImportSemaphoreWin32HandleInfo);
+typedef VkResult (VKAPI_PTR *PFN_vkGetSemaphoreWin32HandleKHR)(VkDevice device, const VkSemaphoreGetWin32HandleInfoKHR* pGetWin32HandleInfo, HANDLE* pHandle);
+
+#ifndef VK_NO_PROTOTYPES
+VKAPI_ATTR VkResult VKAPI_CALL vkImportSemaphoreWin32HandleKHR(
+    VkDevice                                    device,
+    const VkImportSemaphoreWin32HandleInfoKHR*  pImportSemaphoreWin32HandleInfo);
+
+VKAPI_ATTR VkResult VKAPI_CALL vkGetSemaphoreWin32HandleKHR(
+    VkDevice                                    device,
+    const VkSemaphoreGetWin32HandleInfoKHR*     pGetWin32HandleInfo,
+    HANDLE*                                     pHandle);
+#endif
+
+#define VK_KHR_external_fence_win32 1
+#define VK_KHR_EXTERNAL_FENCE_WIN32_SPEC_VERSION 1
+#define VK_KHR_EXTERNAL_FENCE_WIN32_EXTENSION_NAME "VK_KHR_external_fence_win32"
+
+typedef struct VkImportFenceWin32HandleInfoKHR {
+    VkStructureType                      sType;
+    const void*                          pNext;
+    VkFence                              fence;
+    VkFenceImportFlags                   flags;
+    VkExternalFenceHandleTypeFlagBits    handleType;
+    HANDLE                               handle;
+    LPCWSTR                              name;
+} VkImportFenceWin32HandleInfoKHR;
+
+typedef struct VkExportFenceWin32HandleInfoKHR {
+    VkStructureType               sType;
+    const void*                   pNext;
+    const SECURITY_ATTRIBUTES*    pAttributes;
+    DWORD                         dwAccess;
+    LPCWSTR                       name;
+} VkExportFenceWin32HandleInfoKHR;
+
+typedef struct VkFenceGetWin32HandleInfoKHR {
+    VkStructureType                      sType;
+    const void*                          pNext;
+    VkFence                              fence;
+    VkExternalFenceHandleTypeFlagBits    handleType;
+} VkFenceGetWin32HandleInfoKHR;
+
+
+typedef VkResult (VKAPI_PTR *PFN_vkImportFenceWin32HandleKHR)(VkDevice device, const VkImportFenceWin32HandleInfoKHR* pImportFenceWin32HandleInfo);
+typedef VkResult (VKAPI_PTR *PFN_vkGetFenceWin32HandleKHR)(VkDevice device, const VkFenceGetWin32HandleInfoKHR* pGetWin32HandleInfo, HANDLE* pHandle);
+
+#ifndef VK_NO_PROTOTYPES
+VKAPI_ATTR VkResult VKAPI_CALL vkImportFenceWin32HandleKHR(
+    VkDevice                                    device,
+    const VkImportFenceWin32HandleInfoKHR*      pImportFenceWin32HandleInfo);
+
+VKAPI_ATTR VkResult VKAPI_CALL vkGetFenceWin32HandleKHR(
+    VkDevice                                    device,
+    const VkFenceGetWin32HandleInfoKHR*         pGetWin32HandleInfo,
+    HANDLE*                                     pHandle);
+#endif
+
+#define VK_NV_external_memory_win32 1
+#define VK_NV_EXTERNAL_MEMORY_WIN32_SPEC_VERSION 1
+#define VK_NV_EXTERNAL_MEMORY_WIN32_EXTENSION_NAME "VK_NV_external_memory_win32"
+
+typedef struct VkImportMemoryWin32HandleInfoNV {
+    VkStructureType                      sType;
+    const void*                          pNext;
+    VkExternalMemoryHandleTypeFlagsNV    handleType;
+    HANDLE                               handle;
+} VkImportMemoryWin32HandleInfoNV;
+
+typedef struct VkExportMemoryWin32HandleInfoNV {
+    VkStructureType               sType;
+    const void*                   pNext;
+    const SECURITY_ATTRIBUTES*    pAttributes;
+    DWORD                         dwAccess;
+} VkExportMemoryWin32HandleInfoNV;
+
+
+typedef VkResult (VKAPI_PTR *PFN_vkGetMemoryWin32HandleNV)(VkDevice device, VkDeviceMemory memory, VkExternalMemoryHandleTypeFlagsNV handleType, HANDLE* pHandle);
+
+#ifndef VK_NO_PROTOTYPES
+VKAPI_ATTR VkResult VKAPI_CALL vkGetMemoryWin32HandleNV(
+    VkDevice                                    device,
+    VkDeviceMemory                              memory,
+    VkExternalMemoryHandleTypeFlagsNV           handleType,
+    HANDLE*                                     pHandle);
+#endif
+
+#define VK_NV_win32_keyed_mutex 1
+#define VK_NV_WIN32_KEYED_MUTEX_SPEC_VERSION 1
+#define VK_NV_WIN32_KEYED_MUTEX_EXTENSION_NAME "VK_NV_win32_keyed_mutex"
+
+typedef struct VkWin32KeyedMutexAcquireReleaseInfoNV {
+    VkStructureType          sType;
+    const void*              pNext;
+    uint32_t                 acquireCount;
+    const VkDeviceMemory*    pAcquireSyncs;
+    const uint64_t*          pAcquireKeys;
+    const uint32_t*          pAcquireTimeoutMilliseconds;
+    uint32_t                 releaseCount;
+    const VkDeviceMemory*    pReleaseSyncs;
+    const uint64_t*          pReleaseKeys;
+} VkWin32KeyedMutexAcquireReleaseInfoNV;
+
+
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif
@@ -0,0 +1,66 @@
+#ifndef VULKAN_XCB_H_
+#define VULKAN_XCB_H_ 1
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+/*
+** Copyright (c) 2015-2018 The Khronos Group Inc.
+**
+** Licensed under the Apache License, Version 2.0 (the "License");
+** you may not use this file except in compliance with the License.
+** You may obtain a copy of the License at
+**
+**     http://www.apache.org/licenses/LICENSE-2.0
+**
+** Unless required by applicable law or agreed to in writing, software
+** distributed under the License is distributed on an "AS IS" BASIS,
+** WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+** See the License for the specific language governing permissions and
+** limitations under the License.
+*/
+
+/*
+** This header is generated from the Khronos Vulkan XML API Registry.
+**
+*/
+
+
+#define VK_KHR_xcb_surface 1
+#define VK_KHR_XCB_SURFACE_SPEC_VERSION   6
+#define VK_KHR_XCB_SURFACE_EXTENSION_NAME "VK_KHR_xcb_surface"
+
+typedef VkFlags VkXcbSurfaceCreateFlagsKHR;
+
+typedef struct VkXcbSurfaceCreateInfoKHR {
+    VkStructureType               sType;
+    const void*                   pNext;
+    VkXcbSurfaceCreateFlagsKHR    flags;
+    xcb_connection_t*             connection;
+    xcb_window_t                  window;
+} VkXcbSurfaceCreateInfoKHR;
+
+
+typedef VkResult (VKAPI_PTR *PFN_vkCreateXcbSurfaceKHR)(VkInstance instance, const VkXcbSurfaceCreateInfoKHR* pCreateInfo, const VkAllocationCallbacks* pAllocator, VkSurfaceKHR* pSurface);
+typedef VkBool32 (VKAPI_PTR *PFN_vkGetPhysicalDeviceXcbPresentationSupportKHR)(VkPhysicalDevice physicalDevice, uint32_t queueFamilyIndex, xcb_connection_t* connection, xcb_visualid_t visual_id);
+
+#ifndef VK_NO_PROTOTYPES
+VKAPI_ATTR VkResult VKAPI_CALL vkCreateXcbSurfaceKHR(
+    VkInstance                                  instance,
+    const VkXcbSurfaceCreateInfoKHR*            pCreateInfo,
+    const VkAllocationCallbacks*                pAllocator,
+    VkSurfaceKHR*                               pSurface);
+
+VKAPI_ATTR VkBool32 VKAPI_CALL vkGetPhysicalDeviceXcbPresentationSupportKHR(
+    VkPhysicalDevice                            physicalDevice,
+    uint32_t                                    queueFamilyIndex,
+    xcb_connection_t*                           connection,
+    xcb_visualid_t                              visual_id);
+#endif
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif
@@ -0,0 +1,66 @@
+#ifndef VULKAN_XLIB_H_
+#define VULKAN_XLIB_H_ 1
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+/*
+** Copyright (c) 2015-2018 The Khronos Group Inc.
+**
+** Licensed under the Apache License, Version 2.0 (the "License");
+** you may not use this file except in compliance with the License.
+** You may obtain a copy of the License at
+**
+**     http://www.apache.org/licenses/LICENSE-2.0
+**
+** Unless required by applicable law or agreed to in writing, software
+** distributed under the License is distributed on an "AS IS" BASIS,
+** WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+** See the License for the specific language governing permissions and
+** limitations under the License.
+*/
+
+/*
+** This header is generated from the Khronos Vulkan XML API Registry.
+**
+*/
+
+
+#define VK_KHR_xlib_surface 1
+#define VK_KHR_XLIB_SURFACE_SPEC_VERSION  6
+#define VK_KHR_XLIB_SURFACE_EXTENSION_NAME "VK_KHR_xlib_surface"
+
+typedef VkFlags VkXlibSurfaceCreateFlagsKHR;
+
+typedef struct VkXlibSurfaceCreateInfoKHR {
+    VkStructureType                sType;
+    const void*                    pNext;
+    VkXlibSurfaceCreateFlagsKHR    flags;
+    Display*                       dpy;
+    Window                         window;
+} VkXlibSurfaceCreateInfoKHR;
+
+
+typedef VkResult (VKAPI_PTR *PFN_vkCreateXlibSurfaceKHR)(VkInstance instance, const VkXlibSurfaceCreateInfoKHR* pCreateInfo, const VkAllocationCallbacks* pAllocator, VkSurfaceKHR* pSurface);
+typedef VkBool32 (VKAPI_PTR *PFN_vkGetPhysicalDeviceXlibPresentationSupportKHR)(VkPhysicalDevice physicalDevice, uint32_t queueFamilyIndex, Display* dpy, VisualID visualID);
+
+#ifndef VK_NO_PROTOTYPES
+VKAPI_ATTR VkResult VKAPI_CALL vkCreateXlibSurfaceKHR(
+    VkInstance                                  instance,
+    const VkXlibSurfaceCreateInfoKHR*           pCreateInfo,
+    const VkAllocationCallbacks*                pAllocator,
+    VkSurfaceKHR*                               pSurface);
+
+VKAPI_ATTR VkBool32 VKAPI_CALL vkGetPhysicalDeviceXlibPresentationSupportKHR(
+    VkPhysicalDevice                            physicalDevice,
+    uint32_t                                    queueFamilyIndex,
+    Display*                                    dpy,
+    VisualID                                    visualID);
+#endif
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif
@@ -0,0 +1,54 @@
+#ifndef VULKAN_XLIB_XRANDR_H_
+#define VULKAN_XLIB_XRANDR_H_ 1
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+/*
+** Copyright (c) 2015-2018 The Khronos Group Inc.
+**
+** Licensed under the Apache License, Version 2.0 (the "License");
+** you may not use this file except in compliance with the License.
+** You may obtain a copy of the License at
+**
+**     http://www.apache.org/licenses/LICENSE-2.0
+**
+** Unless required by applicable law or agreed to in writing, software
+** distributed under the License is distributed on an "AS IS" BASIS,
+** WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+** See the License for the specific language governing permissions and
+** limitations under the License.
+*/
+
+/*
+** This header is generated from the Khronos Vulkan XML API Registry.
+**
+*/
+
+
+#define VK_EXT_acquire_xlib_display 1
+#define VK_EXT_ACQUIRE_XLIB_DISPLAY_SPEC_VERSION 1
+#define VK_EXT_ACQUIRE_XLIB_DISPLAY_EXTENSION_NAME "VK_EXT_acquire_xlib_display"
+
+typedef VkResult (VKAPI_PTR *PFN_vkAcquireXlibDisplayEXT)(VkPhysicalDevice physicalDevice, Display* dpy, VkDisplayKHR display);
+typedef VkResult (VKAPI_PTR *PFN_vkGetRandROutputDisplayEXT)(VkPhysicalDevice physicalDevice, Display* dpy, RROutput rrOutput, VkDisplayKHR* pDisplay);
+
+#ifndef VK_NO_PROTOTYPES
+VKAPI_ATTR VkResult VKAPI_CALL vkAcquireXlibDisplayEXT(
+    VkPhysicalDevice                            physicalDevice,
+    Display*                                    dpy,
+    VkDisplayKHR                                display);
+
+VKAPI_ATTR VkResult VKAPI_CALL vkGetRandROutputDisplayEXT(
+    VkPhysicalDevice                            physicalDevice,
+    Display*                                    dpy,
+    RROutput                                    rrOutput,
+    VkDisplayKHR*                               pDisplay);
+#endif
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif
@@ -17,7 +17,7 @@ file(GLOB lib_hdrs ${IPP_IW_PATH}/include/*.h ${IPP_IW_PATH}/include/iw/*.h ${IP
 #         Define the library target:
 # ----------------------------------------------------------------------------------

-add_library(${IPP_IW_LIBRARY} STATIC ${OPENCV_3RDPARTY_EXCLUDE_FROM_ALL} ${lib_srcs} ${lib_hdrs})
+add_library(${IPP_IW_LIBRARY} STATIC ${lib_srcs} ${lib_hdrs})

 if(UNIX)
  if(CV_GCC OR CV_CLANG OR CV_ICC)
@@ -41,5 +41,5 @@ if(ENABLE_SOLUTION_FOLDERS)
 endif()

 if(NOT BUILD_SHARED_LIBS)
-  ocv_install_target(${IPP_IW_LIBRARY} EXPORT OpenCVModules ARCHIVE DESTINATION ${OPENCV_3P_LIB_INSTALL_PATH} COMPONENT dev OPTIONAL)
+  ocv_install_target(${IPP_IW_LIBRARY} EXPORT OpenCVModules ARCHIVE DESTINATION ${OPENCV_3P_LIB_INSTALL_PATH} COMPONENT dev)
 endif()
@@ -2,33 +2,32 @@ function(download_ippicv root_var)
  set(${root_var} "" PARENT_SCOPE)

  # Commit SHA in the opencv_3rdparty repo
-  set(IPPICV_COMMIT "fd27188235d85e552de31425e7ea0f53ba73ba53")
+  set(IPPICV_COMMIT "a56b6ac6f030c312b2dce17430eef13aed9af274")
  # Define actual ICV versions
  if(APPLE)
-    set(IPPICV_COMMIT "0cc4aa06bf2bef4b05d237c69a5a96b9cd0cb85a")
    set(OPENCV_ICV_PLATFORM "macosx")
    set(OPENCV_ICV_PACKAGE_SUBDIR "ippicv_mac")
-    set(OPENCV_ICV_NAME "ippicv_2021.9.1_mac_intel64_20230919_general.tgz")
-    set(OPENCV_ICV_HASH "14f01c5a4780bfae9dde9b0aaf5e56fc")
+    set(OPENCV_ICV_NAME "ippicv_2020_mac_intel64_20191018_general.tgz")
+    set(OPENCV_ICV_HASH "1c3d675c2a2395d094d523024896e01b")
  elseif((UNIX AND NOT ANDROID) OR (UNIX AND ANDROID_ABI MATCHES "x86"))
    set(OPENCV_ICV_PLATFORM "linux")
    set(OPENCV_ICV_PACKAGE_SUBDIR "ippicv_lnx")
    if(X86_64)
-      set(OPENCV_ICV_NAME "ippicv_2021.11.0_lnx_intel64_20240201_general.tgz")
-      set(OPENCV_ICV_HASH "0f2745ff705ecae31176dad437608f6f")
+      set(OPENCV_ICV_NAME "ippicv_2020_lnx_intel64_20191018_general.tgz")
+      set(OPENCV_ICV_HASH "7421de0095c7a39162ae13a6098782f9")
    else()
-      set(OPENCV_ICV_NAME "ippicv_2021.11.0_lnx_ia32_20240201_general.tgz")
-      set(OPENCV_ICV_HASH "63e381bf08076ca34fd5264203043a45")
+      set(OPENCV_ICV_NAME "ippicv_2020_lnx_ia32_20191018_general.tgz")
+      set(OPENCV_ICV_HASH "ad189a940fb60eb71f291321322fe3e8")
    endif()
  elseif(WIN32 AND NOT ARM)
    set(OPENCV_ICV_PLATFORM "windows")
    set(OPENCV_ICV_PACKAGE_SUBDIR "ippicv_win")
    if(X86_64)
-      set(OPENCV_ICV_NAME "ippicv_2021.11.0_win_intel64_20240201_general.zip")
-      set(OPENCV_ICV_HASH "59d154bf54a1e3eea20d7248f81a2a8e")
+      set(OPENCV_ICV_NAME "ippicv_2020_win_intel64_20191018_general.zip")
+      set(OPENCV_ICV_HASH "879741a7946b814455eee6c6ffde2984")
    else()
-      set(OPENCV_ICV_NAME "ippicv_2021.11.0_win_ia32_20240201_general.zip")
-      set(OPENCV_ICV_HASH "7a6d8ac5825c02fea6cbfc1201b521b5")
+      set(OPENCV_ICV_NAME "ippicv_2020_win_ia32_20191018_general.zip")
+      set(OPENCV_ICV_HASH "cd39bdf0c2e1cac9a61101dad7a2413e")
    endif()
  else()
    return()
@@ -37,7 +37,7 @@ set(ITT_SRCS
    src/ittnotify/jitprofiling.c
 )

-add_library(${ITT_LIBRARY} STATIC ${OPENCV_3RDPARTY_EXCLUDE_FROM_ALL} ${ITT_SRCS} ${ITT_PUBLIC_HDRS} ${ITT_PRIVATE_HDRS})
+add_library(${ITT_LIBRARY} STATIC ${ITT_SRCS} ${ITT_PUBLIC_HDRS} ${ITT_PRIVATE_HDRS})

 if(NOT WIN32)
  if(HAVE_DL_LIBRARY)
@@ -60,7 +60,7 @@ if(ENABLE_SOLUTION_FOLDERS)
 endif()

 if(NOT BUILD_SHARED_LIBS)
-  ocv_install_target(${ITT_LIBRARY} EXPORT OpenCVModules ARCHIVE DESTINATION ${OPENCV_3P_LIB_INSTALL_PATH} COMPONENT dev OPTIONAL)
+  ocv_install_target(${ITT_LIBRARY} EXPORT OpenCVModules ARCHIVE DESTINATION ${OPENCV_3P_LIB_INSTALL_PATH} COMPONENT dev)
 endif()

 ocv_install_3rdparty_licenses(ittnotify src/ittnotify/LICENSE.BSD src/ittnotify/LICENSE.GPL)
@@ -17,7 +17,7 @@ file(GLOB lib_ext_hdrs jasper/*.h)
 #         Define the library target:
 # ----------------------------------------------------------------------------------

-add_library(${JASPER_LIBRARY} STATIC ${OPENCV_3RDPARTY_EXCLUDE_FROM_ALL} ${lib_srcs} ${lib_hdrs} ${lib_ext_hdrs})
+add_library(${JASPER_LIBRARY} STATIC ${lib_srcs} ${lib_hdrs} ${lib_ext_hdrs})

 if(WIN32 AND NOT MINGW)
  add_definitions(-DJAS_WIN_MSVC_BUILD)
@@ -46,7 +46,7 @@ if(ENABLE_SOLUTION_FOLDERS)
 endif()

 if(NOT BUILD_SHARED_LIBS)
-  ocv_install_target(${JASPER_LIBRARY} EXPORT OpenCVModules ARCHIVE DESTINATION ${OPENCV_3P_LIB_INSTALL_PATH} COMPONENT dev OPTIONAL)
+  ocv_install_target(${JASPER_LIBRARY} EXPORT OpenCVModules ARCHIVE DESTINATION ${OPENCV_3P_LIB_INSTALL_PATH} COMPONENT dev)
 endif()

 ocv_install_3rdparty_licenses(jasper LICENSE README copyright)
@@ -3,10 +3,10 @@ project(${JPEG_LIBRARY} C)
 ocv_warnings_disable(CMAKE_C_FLAGS -Wunused-parameter -Wsign-compare -Wshorten-64-to-32 -Wimplicit-fallthrough)

 set(VERSION_MAJOR 2)
-set(VERSION_MINOR 1)
-set(VERSION_REVISION 2)
+set(VERSION_MINOR 0)
+set(VERSION_REVISION 4)
 set(VERSION ${VERSION_MAJOR}.${VERSION_MINOR}.${VERSION_REVISION})
-set(LIBJPEG_TURBO_VERSION_NUMBER 2001002)
+set(LIBJPEG_TURBO_VERSION_NUMBER 2000004)

 string(TIMESTAMP BUILD "opencv-${OPENCV_VERSION}-libjpeg-turbo")
 if(CMAKE_BUILD_TYPE STREQUAL "Debug")
@@ -46,6 +46,7 @@ if(UNIX)
  ocv_update(HAVE_UNSIGNED_SHORT 1)
  # undef INCOMPLETE_TYPES_BROKEN
  ocv_update(RIGHT_SHIFT_IS_UNSIGNED 0)
+  ocv_update(__CHAR_UNSIGNED__ 0)
 endif()


@@ -64,8 +65,6 @@ set(JPEG_LIB_VERSION 62)
 # OpenCV
 set(JPEG_LIB_VERSION "${VERSION}-${JPEG_LIB_VERSION}" PARENT_SCOPE)

-set(THREAD_LOCAL "")  # WITH_TURBOJPEG is not used
-
 if(MSVC)
  add_definitions(-W3 -wd4996 -wd4018)
 endif()
@@ -105,7 +104,7 @@ set(JPEG_SOURCES ${JPEG_SOURCES} jsimd_none.c)

 ocv_list_add_prefix(JPEG_SOURCES src/)

-add_library(${JPEG_LIBRARY} STATIC ${OPENCV_3RDPARTY_EXCLUDE_FROM_ALL} ${JPEG_SOURCES} ${SIMD_OBJS})
+add_library(${JPEG_LIBRARY} STATIC ${JPEG_SOURCES} ${SIMD_OBJS})

 set_target_properties(${JPEG_LIBRARY}
  PROPERTIES OUTPUT_NAME ${JPEG_LIBRARY}
@@ -120,7 +119,7 @@ if(ENABLE_SOLUTION_FOLDERS)
 endif()

 if(NOT BUILD_SHARED_LIBS)
-  ocv_install_target(${JPEG_LIBRARY} EXPORT OpenCVModules ARCHIVE DESTINATION ${OPENCV_3P_LIB_INSTALL_PATH} COMPONENT dev OPTIONAL)
+  ocv_install_target(${JPEG_LIBRARY} EXPORT OpenCVModules ARCHIVE DESTINATION ${OPENCV_3P_LIB_INSTALL_PATH} COMPONENT dev)
 endif()

 ocv_install_3rdparty_licenses(libjpeg-turbo README.md LICENSE.md README.ijg)
@@ -91,7 +91,7 @@ best of our understanding.
 The Modified (3-clause) BSD License
 ===================================

-Copyright (C)2009-2021 D. R. Commander.  All Rights Reserved.<br>
+Copyright (C)2009-2019 D. R. Commander.  All Rights Reserved.
 Copyright (C)2015 Viktor Szathmáry.  All Rights Reserved.

 Redistribution and use in source and binary forms, with or without
@@ -128,7 +128,7 @@ with respect to this software, its quality, accuracy, merchantability, or
 fitness for a particular purpose.  This software is provided "AS IS", and you,
 its user, assume the entire risk as to its quality and accuracy.

-This software is copyright (C) 1991-2020, Thomas G. Lane, Guido Vollbeding.
+This software is copyright (C) 1991-2016, Thomas G. Lane, Guido Vollbeding.
 All Rights Reserved except as specified below.

 Permission is hereby granted to use, copy, modify, and distribute this
@@ -159,6 +159,19 @@ commercial products, provided that all warranty or liability claims are
 assumed by the product vendor.


+The IJG distribution formerly included code to read and write GIF files.
+To avoid entanglement with the Unisys LZW patent (now expired), GIF reading
+support has been removed altogether, and the GIF writer has been simplified
+to produce "uncompressed GIFs".  This technique does not use the LZW
+algorithm; the resulting GIF files are larger than usual, but are readable
+by all standard GIF decoders.
+
+We are required to state that
+    "The Graphics Interchange Format(c) is the Copyright property of
+    CompuServe Incorporated.  GIF(sm) is a Service Mark property of
+    CompuServe Incorporated."
+
+
 REFERENCES
 ==========

@@ -210,12 +223,12 @@ https://www.iso.org/standard/54989.html and http://www.itu.int/rec/T-REC-T.871.
 A PDF file of the older JFIF 1.02 specification is available at
 http://www.w3.org/Graphics/JPEG/jfif3.pdf.

-The TIFF 6.0 file format specification can be obtained from
-http://mirrors.ctan.org/graphics/tiff/TIFF6.ps.gz.  The JPEG incorporation
-scheme found in the TIFF 6.0 spec of 3-June-92 has a number of serious
-problems.  IJG does not recommend use of the TIFF 6.0 design (TIFF Compression
-tag 6).  Instead, we recommend the JPEG design proposed by TIFF Technical Note
-#2 (Compression tag 7).  Copies of this Note can be obtained from
+The TIFF 6.0 file format specification can be obtained by FTP from
+ftp://ftp.sgi.com/graphics/tiff/TIFF6.ps.gz.  The JPEG incorporation scheme
+found in the TIFF 6.0 spec of 3-June-92 has a number of serious problems.
+IJG does not recommend use of the TIFF 6.0 design (TIFF Compression tag 6).
+Instead, we recommend the JPEG design proposed by TIFF Technical Note #2
+(Compression tag 7).  Copies of this Note can be obtained from
 http://www.ijg.org/files/.  It is expected that the next revision
 of the TIFF spec will replace the 6.0 JPEG design with the Note's design.
 Although IJG's own code does not support TIFF/JPEG, the free libtiff library
@@ -230,8 +243,14 @@ The most recent released version can always be found there in
 directory "files".

 The JPEG FAQ (Frequently Asked Questions) article is a source of some
-general information about JPEG.  It is available at
-http://www.faqs.org/faqs/jpeg-faq.
+general information about JPEG.
+It is available on the World Wide Web at http://www.faqs.org/faqs/jpeg-faq/
+and other news.answers archive sites, including the official news.answers
+archive at rtfm.mit.edu: ftp://rtfm.mit.edu/pub/usenet/news.answers/jpeg-faq/.
+If you don't have Web or FTP access, send e-mail to mail-server@rtfm.mit.edu
+with body
+        send usenet/news.answers/jpeg-faq/part1
+        send usenet/news.answers/jpeg-faq/part2


 FILE FORMAT COMPATIBILITY
@@ -2,8 +2,8 @@ Background
 ==========

 libjpeg-turbo is a JPEG image codec that uses SIMD instructions to accelerate
-baseline JPEG compression and decompression on x86, x86-64, Arm, PowerPC, and
-MIPS systems, as well as progressive JPEG compression on x86, x86-64, and Arm
+baseline JPEG compression and decompression on x86, x86-64, ARM, PowerPC, and
+MIPS systems, as well as progressive JPEG compression on x86 and x86-64
 systems.  On such systems, libjpeg-turbo is generally 2-6x as fast as libjpeg,
 all else being equal.  On other types of systems, libjpeg-turbo can still
 outperform libjpeg by a significant amount, by virtue of its highly-optimized
@@ -179,8 +179,8 @@ supported and which aren't.

 NOTE:  As of this writing, extensive research has been conducted into the
 usefulness of DCT scaling as a means of data reduction and SmartScale as a
-means of quality improvement.  Readers are invited to peruse the research at
-<http://www.libjpeg-turbo.org/About/SmartScale> and draw their own conclusions,
+means of quality improvement.  The reader is invited to peruse the research at
+<http://www.libjpeg-turbo.org/About/SmartScale> and draw his/her own conclusions,
 but it is the general belief of our project that these features have not
 demonstrated sufficient usefulness to justify inclusion in libjpeg-turbo.

@@ -287,13 +287,12 @@ following reasons:
  (and slightly faster) floating point IDCT algorithm introduced in libjpeg
  v8a as opposed to the algorithm used in libjpeg v6b.  It should be noted,
  however, that this algorithm basically brings the accuracy of the floating
-  point IDCT in line with the accuracy of the accurate integer IDCT.  The
-  floating point DCT/IDCT algorithms are mainly a legacy feature, and they do
-  not produce significantly more accuracy than the accurate integer algorithms
-  (to put numbers on this, the typical difference in PNSR between the two
-  algorithms is less than 0.10 dB, whereas changing the quality level by 1 in
-  the upper range of the quality scale is typically more like a 1.0 dB
-  difference.)
+  point IDCT in line with the accuracy of the slow integer IDCT.  The floating
+  point DCT/IDCT algorithms are mainly a legacy feature, and they do not
+  produce significantly more accuracy than the slow integer algorithms (to put
+  numbers on this, the typical difference in PNSR between the two algorithms
+  is less than 0.10 dB, whereas changing the quality level by 1 in the upper
+  range of the quality scale is typically more like a 1.0 dB difference.)

 - If the floating point algorithms in libjpeg-turbo are not implemented using
  SIMD instructions on a particular platform, then the accuracy of the
@@ -341,7 +340,7 @@ The algorithm used by the SIMD-accelerated quantization function cannot produce
 correct results whenever the fast integer forward DCT is used along with a JPEG
 quality of 98-100.  Thus, libjpeg-turbo must use the non-SIMD quantization
 function in those cases.  This causes performance to drop by as much as 40%.
-It is therefore strongly advised that you use the accurate integer forward DCT
+It is therefore strongly advised that you use the slow integer forward DCT
 whenever encoding images with a JPEG quality of 98 or higher.


@@ -61,6 +61,11 @@
   unsigned. */
 #cmakedefine RIGHT_SHIFT_IS_UNSIGNED 1

+/* Define to 1 if type `char' is unsigned and you are not using gcc.  */
+#ifndef __CHAR_UNSIGNED__
+  #cmakedefine __CHAR_UNSIGNED__ 1
+#endif
+
 /* Define to empty if `const' does not conform to ANSI C. */
 /* #undef const */

@@ -18,6 +18,7 @@
 #define HAVE_UNSIGNED_SHORT
 #undef INCOMPLETE_TYPES_BROKEN
 #undef RIGHT_SHIFT_IS_UNSIGNED
+#undef __CHAR_UNSIGNED__

 /* Define "boolean" as unsigned char, not int, per Windows custom */
 #ifndef __RPCNDR_H__            /* don't conflict if rpcndr.h already read */
@@ -15,9 +15,6 @@
 #endif
 #endif

-/* How to obtain thread-local storage */
-#define THREAD_LOCAL  @THREAD_LOCAL@
-
 /* Define to the full name of this package. */
 #define PACKAGE_NAME  "@CMAKE_PROJECT_NAME@"

@@ -40,13 +37,3 @@
 #define HAVE_BITSCANFORWARD
 #endif
 #endif
-
-#if defined(__has_attribute)
-#if __has_attribute(fallthrough)
-#define FALLTHROUGH  __attribute__((fallthrough));
-#else
-#define FALLTHROUGH
-#endif
-#else
-#define FALLTHROUGH
-#endif
@@ -48,9 +48,9 @@ rgb_ycc_convert_internal(j_compress_ptr cinfo, JSAMPARRAY input_buf,
    outptr2 = output_buf[2][output_row];
    output_row++;
    for (col = 0; col < num_cols; col++) {
-      r = inptr[RGB_RED];
-      g = inptr[RGB_GREEN];
-      b = inptr[RGB_BLUE];
+      r = GETJSAMPLE(inptr[RGB_RED]);
+      g = GETJSAMPLE(inptr[RGB_GREEN]);
+      b = GETJSAMPLE(inptr[RGB_BLUE]);
      inptr += RGB_PIXELSIZE;
      /* If the inputs are 0..MAXJSAMPLE, the outputs of these equations
       * must be too; we do not need an explicit range-limiting operation.
@@ -100,9 +100,9 @@ rgb_gray_convert_internal(j_compress_ptr cinfo, JSAMPARRAY input_buf,
    outptr = output_buf[0][output_row];
    output_row++;
    for (col = 0; col < num_cols; col++) {
-      r = inptr[RGB_RED];
-      g = inptr[RGB_GREEN];
-      b = inptr[RGB_BLUE];
+      r = GETJSAMPLE(inptr[RGB_RED]);
+      g = GETJSAMPLE(inptr[RGB_GREEN]);
+      b = GETJSAMPLE(inptr[RGB_BLUE]);
      inptr += RGB_PIXELSIZE;
      /* Y */
      outptr[col] = (JSAMPLE)((ctab[r + R_Y_OFF] + ctab[g + G_Y_OFF] +
@@ -135,9 +135,9 @@ rgb_rgb_convert_internal(j_compress_ptr cinfo, JSAMPARRAY input_buf,
    outptr2 = output_buf[2][output_row];
    output_row++;
    for (col = 0; col < num_cols; col++) {
-      outptr0[col] = inptr[RGB_RED];
-      outptr1[col] = inptr[RGB_GREEN];
-      outptr2[col] = inptr[RGB_BLUE];
+      outptr0[col] = GETJSAMPLE(inptr[RGB_RED]);
+      outptr1[col] = GETJSAMPLE(inptr[RGB_GREEN]);
+      outptr2[col] = GETJSAMPLE(inptr[RGB_BLUE]);
      inptr += RGB_PIXELSIZE;
    }
  }
@@ -392,11 +392,11 @@ cmyk_ycck_convert(j_compress_ptr cinfo, JSAMPARRAY input_buf,
    outptr3 = output_buf[3][output_row];
    output_row++;
    for (col = 0; col < num_cols; col++) {
-      r = MAXJSAMPLE - inptr[0];
-      g = MAXJSAMPLE - inptr[1];
-      b = MAXJSAMPLE - inptr[2];
+      r = MAXJSAMPLE - GETJSAMPLE(inptr[0]);
+      g = MAXJSAMPLE - GETJSAMPLE(inptr[1]);
+      b = MAXJSAMPLE - GETJSAMPLE(inptr[2]);
      /* K passes through as-is */
-      outptr3[col] = inptr[3];
+      outptr3[col] = inptr[3];  /* don't need GETJSAMPLE here */
      inptr += 4;
      /* If the inputs are 0..MAXJSAMPLE, the outputs of these equations
       * must be too; we do not need an explicit range-limiting operation.
@@ -438,7 +438,7 @@ grayscale_convert(j_compress_ptr cinfo, JSAMPARRAY input_buf,
    outptr = output_buf[0][output_row];
    output_row++;
    for (col = 0; col < num_cols; col++) {
-      outptr[col] = inptr[0];
+      outptr[col] = inptr[0];   /* don't need GETJSAMPLE() here */
      inptr += instride;
    }
  }
@@ -497,7 +497,7 @@ null_convert(j_compress_ptr cinfo, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
        inptr = *input_buf;
        outptr = output_buf[ci][output_row];
        for (col = 0; col < num_cols; col++) {
-          outptr[col] = inptr[ci];
+          outptr[col] = inptr[ci]; /* don't need GETJSAMPLE() here */
          inptr += nc;
        }
      }
@@ -381,19 +381,19 @@ convsamp(JSAMPARRAY sample_data, JDIMENSION start_col, DCTELEM *workspace)
    elemptr = sample_data[elemr] + start_col;

 #if DCTSIZE == 8                /* unroll the inner loop */
-    *workspaceptr++ = (*elemptr++) - CENTERJSAMPLE;
-    *workspaceptr++ = (*elemptr++) - CENTERJSAMPLE;
-    *workspaceptr++ = (*elemptr++) - CENTERJSAMPLE;
-    *workspaceptr++ = (*elemptr++) - CENTERJSAMPLE;
-    *workspaceptr++ = (*elemptr++) - CENTERJSAMPLE;
-    *workspaceptr++ = (*elemptr++) - CENTERJSAMPLE;
-    *workspaceptr++ = (*elemptr++) - CENTERJSAMPLE;
-    *workspaceptr++ = (*elemptr++) - CENTERJSAMPLE;
+    *workspaceptr++ = GETJSAMPLE(*elemptr++) - CENTERJSAMPLE;
+    *workspaceptr++ = GETJSAMPLE(*elemptr++) - CENTERJSAMPLE;
+    *workspaceptr++ = GETJSAMPLE(*elemptr++) - CENTERJSAMPLE;
+    *workspaceptr++ = GETJSAMPLE(*elemptr++) - CENTERJSAMPLE;
+    *workspaceptr++ = GETJSAMPLE(*elemptr++) - CENTERJSAMPLE;
+    *workspaceptr++ = GETJSAMPLE(*elemptr++) - CENTERJSAMPLE;
+    *workspaceptr++ = GETJSAMPLE(*elemptr++) - CENTERJSAMPLE;
+    *workspaceptr++ = GETJSAMPLE(*elemptr++) - CENTERJSAMPLE;
 #else
    {
      register int elemc;
      for (elemc = DCTSIZE; elemc > 0; elemc--)
-        *workspaceptr++ = (*elemptr++) - CENTERJSAMPLE;
+        *workspaceptr++ = GETJSAMPLE(*elemptr++) - CENTERJSAMPLE;
    }
 #endif
  }
@@ -533,19 +533,20 @@ convsamp_float(JSAMPARRAY sample_data, JDIMENSION start_col,
  for (elemr = 0; elemr < DCTSIZE; elemr++) {
    elemptr = sample_data[elemr] + start_col;
 #if DCTSIZE == 8                /* unroll the inner loop */
-    *workspaceptr++ = (FAST_FLOAT)((*elemptr++) - CENTERJSAMPLE);
-    *workspaceptr++ = (FAST_FLOAT)((*elemptr++) - CENTERJSAMPLE);
-    *workspaceptr++ = (FAST_FLOAT)((*elemptr++) - CENTERJSAMPLE);
-    *workspaceptr++ = (FAST_FLOAT)((*elemptr++) - CENTERJSAMPLE);
-    *workspaceptr++ = (FAST_FLOAT)((*elemptr++) - CENTERJSAMPLE);
-    *workspaceptr++ = (FAST_FLOAT)((*elemptr++) - CENTERJSAMPLE);
-    *workspaceptr++ = (FAST_FLOAT)((*elemptr++) - CENTERJSAMPLE);
-    *workspaceptr++ = (FAST_FLOAT)((*elemptr++) - CENTERJSAMPLE);
+    *workspaceptr++ = (FAST_FLOAT)(GETJSAMPLE(*elemptr++) - CENTERJSAMPLE);
+    *workspaceptr++ = (FAST_FLOAT)(GETJSAMPLE(*elemptr++) - CENTERJSAMPLE);
+    *workspaceptr++ = (FAST_FLOAT)(GETJSAMPLE(*elemptr++) - CENTERJSAMPLE);
+    *workspaceptr++ = (FAST_FLOAT)(GETJSAMPLE(*elemptr++) - CENTERJSAMPLE);
+    *workspaceptr++ = (FAST_FLOAT)(GETJSAMPLE(*elemptr++) - CENTERJSAMPLE);
+    *workspaceptr++ = (FAST_FLOAT)(GETJSAMPLE(*elemptr++) - CENTERJSAMPLE);
+    *workspaceptr++ = (FAST_FLOAT)(GETJSAMPLE(*elemptr++) - CENTERJSAMPLE);
+    *workspaceptr++ = (FAST_FLOAT)(GETJSAMPLE(*elemptr++) - CENTERJSAMPLE);
 #else
    {
      register int elemc;
      for (elemc = DCTSIZE; elemc > 0; elemc--)
-        *workspaceptr++ = (FAST_FLOAT)((*elemptr++) - CENTERJSAMPLE);
+        *workspaceptr++ = (FAST_FLOAT)
+                          (GETJSAMPLE(*elemptr++) - CENTERJSAMPLE);
    }
 #endif
  }
@@ -4,10 +4,8 @@
 * This file was part of the Independent JPEG Group's software:
 * Copyright (C) 1991-1997, Thomas G. Lane.
 * libjpeg-turbo Modifications:
- * Copyright (C) 2009-2011, 2014-2016, 2018-2021, D. R. Commander.
+ * Copyright (C) 2009-2011, 2014-2016, 2018-2019, D. R. Commander.
 * Copyright (C) 2015, Matthieu Darbois.
- * Copyright (C) 2018, Matthias Räncker.
- * Copyright (C) 2020, Arm Limited.
 * For conditions of distribution and use, see the accompanying README.ijg
 * file.
 *
@@ -36,28 +34,23 @@
 * memory footprint by 64k, which is important for some mobile applications
 * that create many isolated instances of libjpeg-turbo (web browsers, for
 * instance.)  This may improve performance on some mobile platforms as well.
- * This feature is enabled by default only on Arm processors, because some x86
+ * This feature is enabled by default only on ARM processors, because some x86
 * chips have a slow implementation of bsr, and the use of clz/bsr cannot be
 * shown to have a significant performance impact even on the x86 chips that
- * have a fast implementation of it.  When building for Armv6, you can
+ * have a fast implementation of it.  When building for ARMv6, you can
 * explicitly disable the use of clz/bsr by adding -mthumb to the compiler
 * flags (this defines __thumb__).
 */

 /* NOTE: Both GCC and Clang define __GNUC__ */
-#if (defined(__GNUC__) && (defined(__arm__) || defined(__aarch64__))) || \
-    defined(_M_ARM) || defined(_M_ARM64)
+#if defined(__GNUC__) && (defined(__arm__) || defined(__aarch64__))
 #if !defined(__thumb__) || defined(__thumb2__)
 #define USE_CLZ_INTRINSIC
 #endif
 #endif

 #ifdef USE_CLZ_INTRINSIC
-#if defined(_MSC_VER) && !defined(__clang__)
-#define JPEG_NBITS_NONZERO(x)  (32 - _CountLeadingZeros(x))
-#else
 #define JPEG_NBITS_NONZERO(x)  (32 - __builtin_clz(x))
-#endif
 #define JPEG_NBITS(x)          (x ? JPEG_NBITS_NONZERO(x) : 0)
 #else
 #include "jpeg_nbits_table.h"
@@ -72,43 +65,32 @@
 * but must not be updated permanently until we complete the MCU.
 */

-#if defined(__x86_64__) && defined(__ILP32__)
-typedef unsigned long long bit_buf_type;
-#else
-typedef size_t bit_buf_type;
-#endif
-
-/* NOTE: The more optimal Huffman encoding algorithm is only used by the
- * intrinsics implementation of the Arm Neon SIMD extensions, which is why we
- * retain the old Huffman encoder behavior when using the GAS implementation.
- */
-#if defined(WITH_SIMD) && !(defined(__arm__) || defined(__aarch64__) || \
-                            defined(_M_ARM) || defined(_M_ARM64))
-typedef unsigned long long simd_bit_buf_type;
-#else
-typedef bit_buf_type simd_bit_buf_type;
-#endif
-
-#if (defined(SIZEOF_SIZE_T) && SIZEOF_SIZE_T == 8) || defined(_WIN64) || \
-    (defined(__x86_64__) && defined(__ILP32__))
-#define BIT_BUF_SIZE  64
-#elif (defined(SIZEOF_SIZE_T) && SIZEOF_SIZE_T == 4) || defined(_WIN32)
-#define BIT_BUF_SIZE  32
-#else
-#error Cannot determine word size
-#endif
-#define SIMD_BIT_BUF_SIZE  (sizeof(simd_bit_buf_type) * 8)
-
 typedef struct {
-  union {
-    bit_buf_type c;
-    simd_bit_buf_type simd;
-  } put_buffer;                         /* current bit accumulation buffer */
-  int free_bits;                        /* # of bits available in it */
-                                        /* (Neon GAS: # of bits now in it) */
+  size_t put_buffer;                    /* current bit-accumulation buffer */
+  int put_bits;                         /* # of bits now in it */
  int last_dc_val[MAX_COMPS_IN_SCAN];   /* last DC coef for each component */
 } savable_state;

+/* This macro is to work around compilers with missing or broken
+ * structure assignment.  You'll need to fix this code if you have
+ * such a compiler and you change MAX_COMPS_IN_SCAN.
+ */
+
+#ifndef NO_STRUCT_ASSIGN
+#define ASSIGN_STATE(dest, src)  ((dest) = (src))
+#else
+#if MAX_COMPS_IN_SCAN == 4
+#define ASSIGN_STATE(dest, src) \
+  ((dest).put_buffer = (src).put_buffer, \
+   (dest).put_bits = (src).put_bits, \
+   (dest).last_dc_val[0] = (src).last_dc_val[0], \
+   (dest).last_dc_val[1] = (src).last_dc_val[1], \
+   (dest).last_dc_val[2] = (src).last_dc_val[2], \
+   (dest).last_dc_val[3] = (src).last_dc_val[3])
+#endif
+#endif
+
+
 typedef struct {
  struct jpeg_entropy_encoder pub; /* public fields */

@@ -141,7 +123,6 @@ typedef struct {
  size_t free_in_buffer;        /* # of byte spaces remaining in buffer */
  savable_state cur;            /* Current bit buffer & DC state */
  j_compress_ptr cinfo;         /* dump_buffer needs access to this */
-  int simd;
 } working_state;


@@ -220,17 +201,8 @@ start_pass_huff(j_compress_ptr cinfo, boolean gather_statistics)
  }

  /* Initialize bit buffer to empty */
-  if (entropy->simd) {
-    entropy->saved.put_buffer.simd = 0;
-#if defined(__aarch64__) && !defined(NEON_INTRINSICS)
-    entropy->saved.free_bits = 0;
-#else
-    entropy->saved.free_bits = SIMD_BIT_BUF_SIZE;
-#endif
-  } else {
-    entropy->saved.put_buffer.c = 0;
-    entropy->saved.free_bits = BIT_BUF_SIZE;
-  }
+  entropy->saved.put_buffer = 0;
+  entropy->saved.put_bits = 0;

  /* Initialize restart stuff */
  entropy->restarts_to_go = cinfo->restart_interval;
@@ -315,7 +287,6 @@ jpeg_make_c_derived_tbl(j_compress_ptr cinfo, boolean isDC, int tblno,
   * this lets us detect duplicate VAL entries here, and later
   * allows emit_bits to detect any attempt to emit such symbols.
   */
-  MEMZERO(dtbl->ehufco, sizeof(dtbl->ehufco));
  MEMZERO(dtbl->ehufsi, sizeof(dtbl->ehufsi));

  /* This is also a convenient place to check for out-of-range
@@ -363,94 +334,94 @@ dump_buffer(working_state *state)

 /* Outputting bits to the file */

-/* Output byte b and, speculatively, an additional 0 byte.  0xFF must be
- * encoded as 0xFF 0x00, so the output buffer pointer is advanced by 2 if the
- * byte is 0xFF.  Otherwise, the output buffer pointer is advanced by 1, and
- * the speculative 0 byte will be overwritten by the next byte.
+/* These macros perform the same task as the emit_bits() function in the
+ * original libjpeg code.  In addition to reducing overhead by explicitly
+ * inlining the code, additional performance is achieved by taking into
+ * account the size of the bit buffer and waiting until it is almost full
+ * before emptying it.  This mostly benefits 64-bit platforms, since 6
+ * bytes can be stored in a 64-bit bit buffer before it has to be emptied.
 */
-#define EMIT_BYTE(b) { \
-  buffer[0] = (JOCTET)(b); \
-  buffer[1] = 0; \
-  buffer -= -2 + ((JOCTET)(b) < 0xFF); \
+
+#define EMIT_BYTE() { \
+  JOCTET c; \
+  put_bits -= 8; \
+  c = (JOCTET)GETJOCTET(put_buffer >> put_bits); \
+  *buffer++ = c; \
+  if (c == 0xFF)  /* need to stuff a zero byte? */ \
+    *buffer++ = 0; \
 }

-/* Output the entire bit buffer.  If there are no 0xFF bytes in it, then write
- * directly to the output buffer.  Otherwise, use the EMIT_BYTE() macro to
- * encode 0xFF as 0xFF 0x00.
- */
-#if BIT_BUF_SIZE == 64
-
-#define FLUSH() { \
-  if (put_buffer & 0x8080808080808080 & ~(put_buffer + 0x0101010101010101)) { \
-    EMIT_BYTE(put_buffer >> 56) \
-    EMIT_BYTE(put_buffer >> 48) \
-    EMIT_BYTE(put_buffer >> 40) \
-    EMIT_BYTE(put_buffer >> 32) \
-    EMIT_BYTE(put_buffer >> 24) \
-    EMIT_BYTE(put_buffer >> 16) \
-    EMIT_BYTE(put_buffer >>  8) \
-    EMIT_BYTE(put_buffer      ) \
-  } else { \
-    buffer[0] = (JOCTET)(put_buffer >> 56); \
-    buffer[1] = (JOCTET)(put_buffer >> 48); \
-    buffer[2] = (JOCTET)(put_buffer >> 40); \
-    buffer[3] = (JOCTET)(put_buffer >> 32); \
-    buffer[4] = (JOCTET)(put_buffer >> 24); \
-    buffer[5] = (JOCTET)(put_buffer >> 16); \
-    buffer[6] = (JOCTET)(put_buffer >> 8); \
-    buffer[7] = (JOCTET)(put_buffer); \
-    buffer += 8; \
-  } \
+#define PUT_BITS(code, size) { \
+  put_bits += size; \
+  put_buffer = (put_buffer << size) | code; \
 }

-#else
+#if SIZEOF_SIZE_T != 8 && !defined(_WIN64)

-#define FLUSH() { \
-  if (put_buffer & 0x80808080 & ~(put_buffer + 0x01010101)) { \
-    EMIT_BYTE(put_buffer >> 24) \
-    EMIT_BYTE(put_buffer >> 16) \
-    EMIT_BYTE(put_buffer >>  8) \
-    EMIT_BYTE(put_buffer      ) \
-  } else { \
-    buffer[0] = (JOCTET)(put_buffer >> 24); \
-    buffer[1] = (JOCTET)(put_buffer >> 16); \
-    buffer[2] = (JOCTET)(put_buffer >> 8); \
-    buffer[3] = (JOCTET)(put_buffer); \
-    buffer += 4; \
+#define CHECKBUF15() { \
+  if (put_bits > 15) { \
+    EMIT_BYTE() \
+    EMIT_BYTE() \
  } \
 }

 #endif

-/* Fill the bit buffer to capacity with the leading bits from code, then output
- * the bit buffer and put the remaining bits from code into the bit buffer.
- */
-#define PUT_AND_FLUSH(code, size) { \
-  put_buffer = (put_buffer << (size + free_bits)) | (code >> -free_bits); \
-  FLUSH() \
-  free_bits += BIT_BUF_SIZE; \
-  put_buffer = code; \
+#define CHECKBUF31() { \
+  if (put_bits > 31) { \
+    EMIT_BYTE() \
+    EMIT_BYTE() \
+    EMIT_BYTE() \
+    EMIT_BYTE() \
+  } \
 }

-/* Insert code into the bit buffer and output the bit buffer if needed.
- * NOTE: We can't flush with free_bits == 0, since the left shift in
- * PUT_AND_FLUSH() would have undefined behavior.
- */
-#define PUT_BITS(code, size) { \
-  free_bits -= size; \
-  if (free_bits < 0) \
-    PUT_AND_FLUSH(code, size) \
-  else \
-    put_buffer = (put_buffer << size) | code; \
+#define CHECKBUF47() { \
+  if (put_bits > 47) { \
+    EMIT_BYTE() \
+    EMIT_BYTE() \
+    EMIT_BYTE() \
+    EMIT_BYTE() \
+    EMIT_BYTE() \
+    EMIT_BYTE() \
+  } \
 }

-#define PUT_CODE(code, size) { \
-  temp &= (((JLONG)1) << nbits) - 1; \
-  temp |= code << nbits; \
-  nbits += size; \
-  PUT_BITS(temp, nbits) \
+#if !defined(_WIN32) && !defined(SIZEOF_SIZE_T)
+#error Cannot determine word size
+#endif
+
+#if SIZEOF_SIZE_T == 8 || defined(_WIN64)
+
+#define EMIT_BITS(code, size) { \
+  CHECKBUF47() \
+  PUT_BITS(code, size) \
 }

+#define EMIT_CODE(code, size) { \
+  temp2 &= (((JLONG)1) << nbits) - 1; \
+  CHECKBUF31() \
+  PUT_BITS(code, size) \
+  PUT_BITS(temp2, nbits) \
+}
+
+#else
+
+#define EMIT_BITS(code, size) { \
+  PUT_BITS(code, size) \
+  CHECKBUF15() \
+}
+
+#define EMIT_CODE(code, size) { \
+  temp2 &= (((JLONG)1) << nbits) - 1; \
+  PUT_BITS(code, size) \
+  CHECKBUF15() \
+  PUT_BITS(temp2, nbits) \
+  CHECKBUF15() \
+}
+
+#endif
+

 /* Although it is exceedingly rare, it is possible for a Huffman-encoded
 * coefficient block to be larger than the 128-byte unencoded block.  For each
@@ -473,7 +444,6 @@ dump_buffer(working_state *state)

 #define STORE_BUFFER() { \
  if (localbuf) { \
-    size_t bytes, bytestocopy; \
    bytes = buffer - _buffer; \
    buffer = _buffer; \
    while (bytes > 0) { \
@@ -496,46 +466,20 @@ dump_buffer(working_state *state)
 LOCAL(boolean)
 flush_bits(working_state *state)
 {
-  JOCTET _buffer[BUFSIZE], *buffer, temp;
-  simd_bit_buf_type put_buffer;  int put_bits;
-  int localbuf = 0;
-
-  if (state->simd) {
-#if defined(__aarch64__) && !defined(NEON_INTRINSICS)
-    put_bits = state->cur.free_bits;
-#else
-    put_bits = SIMD_BIT_BUF_SIZE - state->cur.free_bits;
-#endif
-    put_buffer = state->cur.put_buffer.simd;
-  } else {
-    put_bits = BIT_BUF_SIZE - state->cur.free_bits;
-    put_buffer = state->cur.put_buffer.c;
-  }
+  JOCTET _buffer[BUFSIZE], *buffer;
+  size_t put_buffer;  int put_bits;
+  size_t bytes, bytestocopy;  int localbuf = 0;

+  put_buffer = state->cur.put_buffer;
+  put_bits = state->cur.put_bits;
  LOAD_BUFFER()

-  while (put_bits >= 8) {
-    put_bits -= 8;
-    temp = (JOCTET)(put_buffer >> put_bits);
-    EMIT_BYTE(temp)
-  }
-  if (put_bits) {
-    /* fill partial byte with ones */
-    temp = (JOCTET)((put_buffer << (8 - put_bits)) | (0xFF >> put_bits));
-    EMIT_BYTE(temp)
-  }
+  /* fill any partial byte with ones */
+  PUT_BITS(0x7F, 7)
+  while (put_bits >= 8) EMIT_BYTE()

-  if (state->simd) {                    /* and reset bit buffer to empty */
-    state->cur.put_buffer.simd = 0;
-#if defined(__aarch64__) && !defined(NEON_INTRINSICS)
-    state->cur.free_bits = 0;
-#else
-    state->cur.free_bits = SIMD_BIT_BUF_SIZE;
-#endif
-  } else {
-    state->cur.put_buffer.c = 0;
-    state->cur.free_bits = BIT_BUF_SIZE;
-  }
+  state->cur.put_buffer = 0;    /* and reset bit-buffer to empty */
+  state->cur.put_bits = 0;
  STORE_BUFFER()

  return TRUE;
@@ -549,7 +493,7 @@ encode_one_block_simd(working_state *state, JCOEFPTR block, int last_dc_val,
                      c_derived_tbl *dctbl, c_derived_tbl *actbl)
 {
  JOCTET _buffer[BUFSIZE], *buffer;
-  int localbuf = 0;
+  size_t bytes, bytestocopy;  int localbuf = 0;

  LOAD_BUFFER()

@@ -565,41 +509,53 @@ LOCAL(boolean)
 encode_one_block(working_state *state, JCOEFPTR block, int last_dc_val,
                 c_derived_tbl *dctbl, c_derived_tbl *actbl)
 {
-  int temp, nbits, free_bits;
-  bit_buf_type put_buffer;
+  int temp, temp2, temp3;
+  int nbits;
+  int r, code, size;
  JOCTET _buffer[BUFSIZE], *buffer;
-  int localbuf = 0;
+  size_t put_buffer;  int put_bits;
+  int code_0xf0 = actbl->ehufco[0xf0], size_0xf0 = actbl->ehufsi[0xf0];
+  size_t bytes, bytestocopy;  int localbuf = 0;

-  free_bits = state->cur.free_bits;
-  put_buffer = state->cur.put_buffer.c;
+  put_buffer = state->cur.put_buffer;
+  put_bits = state->cur.put_bits;
  LOAD_BUFFER()

  /* Encode the DC coefficient difference per section F.1.2.1 */

-  temp = block[0] - last_dc_val;
+  temp = temp2 = block[0] - last_dc_val;

  /* This is a well-known technique for obtaining the absolute value without a
   * branch.  It is derived from an assembly language technique presented in
   * "How to Optimize for the Pentium Processors", Copyright (c) 1996, 1997 by
-   * Agner Fog.  This code assumes we are on a two's complement machine.
+   * Agner Fog.
   */
-  nbits = temp >> (CHAR_BIT * sizeof(int) - 1);
-  temp += nbits;
-  nbits ^= temp;
+  temp3 = temp >> (CHAR_BIT * sizeof(int) - 1);
+  temp ^= temp3;
+  temp -= temp3;
+
+  /* For a negative input, want temp2 = bitwise complement of abs(input) */
+  /* This code assumes we are on a two's complement machine */
+  temp2 += temp3;

  /* Find the number of bits needed for the magnitude of the coefficient */
-  nbits = JPEG_NBITS(nbits);
+  nbits = JPEG_NBITS(temp);

-  /* Emit the Huffman-coded symbol for the number of bits.
-   * Emit that number of bits of the value, if positive,
-   * or the complement of its magnitude, if negative.
-   */
-  PUT_CODE(dctbl->ehufco[nbits], dctbl->ehufsi[nbits])
+  /* Emit the Huffman-coded symbol for the number of bits */
+  code = dctbl->ehufco[nbits];
+  size = dctbl->ehufsi[nbits];
+  EMIT_BITS(code, size)
+
+  /* Mask off any extra bits in code */
+  temp2 &= (((JLONG)1) << nbits) - 1;
+
+  /* Emit that number of bits of the value, if positive, */
+  /* or the complement of its magnitude, if negative. */
+  EMIT_BITS(temp2, nbits)

  /* Encode the AC coefficients per section F.1.2.2 */

-  {
-    int r = 0;                  /* r = run length of zeros */
+  r = 0;                        /* r = run length of zeros */

 /* Manually unroll the k loop to eliminate the counter variable.  This
 * improves performance greatly on systems with a limited number of
@@ -607,46 +563,51 @@ encode_one_block(working_state *state, JCOEFPTR block, int last_dc_val,
 */
 #define kloop(jpeg_natural_order_of_k) { \
  if ((temp = block[jpeg_natural_order_of_k]) == 0) { \
-    r += 16; \
+    r++; \
  } else { \
+    temp2 = temp; \
    /* Branch-less absolute value, bitwise complement, etc., same as above */ \
-    nbits = temp >> (CHAR_BIT * sizeof(int) - 1); \
-    temp += nbits; \
-    nbits ^= temp; \
-    nbits = JPEG_NBITS_NONZERO(nbits); \
+    temp3 = temp >> (CHAR_BIT * sizeof(int) - 1); \
+    temp ^= temp3; \
+    temp -= temp3; \
+    temp2 += temp3; \
+    nbits = JPEG_NBITS_NONZERO(temp); \
    /* if run length > 15, must emit special run-length-16 codes (0xF0) */ \
-    while (r >= 16 * 16) { \
-      r -= 16 * 16; \
-      PUT_BITS(actbl->ehufco[0xf0], actbl->ehufsi[0xf0]) \
+    while (r > 15) { \
+      EMIT_BITS(code_0xf0, size_0xf0) \
+      r -= 16; \
    } \
    /* Emit Huffman symbol for run length / number of bits */ \
-    r += nbits; \
-    PUT_CODE(actbl->ehufco[r], actbl->ehufsi[r]) \
+    temp3 = (r << 4) + nbits; \
+    code = actbl->ehufco[temp3]; \
+    size = actbl->ehufsi[temp3]; \
+    EMIT_CODE(code, size) \
    r = 0; \
  } \
 }

-    /* One iteration for each value in jpeg_natural_order[] */
-    kloop(1);   kloop(8);   kloop(16);  kloop(9);   kloop(2);   kloop(3);
-    kloop(10);  kloop(17);  kloop(24);  kloop(32);  kloop(25);  kloop(18);
-    kloop(11);  kloop(4);   kloop(5);   kloop(12);  kloop(19);  kloop(26);
-    kloop(33);  kloop(40);  kloop(48);  kloop(41);  kloop(34);  kloop(27);
-    kloop(20);  kloop(13);  kloop(6);   kloop(7);   kloop(14);  kloop(21);
-    kloop(28);  kloop(35);  kloop(42);  kloop(49);  kloop(56);  kloop(57);
-    kloop(50);  kloop(43);  kloop(36);  kloop(29);  kloop(22);  kloop(15);
-    kloop(23);  kloop(30);  kloop(37);  kloop(44);  kloop(51);  kloop(58);
-    kloop(59);  kloop(52);  kloop(45);  kloop(38);  kloop(31);  kloop(39);
-    kloop(46);  kloop(53);  kloop(60);  kloop(61);  kloop(54);  kloop(47);
-    kloop(55);  kloop(62);  kloop(63);
+  /* One iteration for each value in jpeg_natural_order[] */
+  kloop(1);   kloop(8);   kloop(16);  kloop(9);   kloop(2);   kloop(3);
+  kloop(10);  kloop(17);  kloop(24);  kloop(32);  kloop(25);  kloop(18);
+  kloop(11);  kloop(4);   kloop(5);   kloop(12);  kloop(19);  kloop(26);
+  kloop(33);  kloop(40);  kloop(48);  kloop(41);  kloop(34);  kloop(27);
+  kloop(20);  kloop(13);  kloop(6);   kloop(7);   kloop(14);  kloop(21);
+  kloop(28);  kloop(35);  kloop(42);  kloop(49);  kloop(56);  kloop(57);
+  kloop(50);  kloop(43);  kloop(36);  kloop(29);  kloop(22);  kloop(15);
+  kloop(23);  kloop(30);  kloop(37);  kloop(44);  kloop(51);  kloop(58);
+  kloop(59);  kloop(52);  kloop(45);  kloop(38);  kloop(31);  kloop(39);
+  kloop(46);  kloop(53);  kloop(60);  kloop(61);  kloop(54);  kloop(47);
+  kloop(55);  kloop(62);  kloop(63);

-    /* If the last coef(s) were zero, emit an end-of-block code */
-    if (r > 0) {
-      PUT_BITS(actbl->ehufco[0], actbl->ehufsi[0])
-    }
+  /* If the last coef(s) were zero, emit an end-of-block code */
+  if (r > 0) {
+    code = actbl->ehufco[0];
+    size = actbl->ehufsi[0];
+    EMIT_BITS(code, size)
  }

-  state->cur.put_buffer.c = put_buffer;
-  state->cur.free_bits = free_bits;
+  state->cur.put_buffer = put_buffer;
+  state->cur.put_bits = put_bits;
  STORE_BUFFER()

  return TRUE;
@@ -693,9 +654,8 @@ encode_mcu_huff(j_compress_ptr cinfo, JBLOCKROW *MCU_data)
  /* Load up working state */
  state.next_output_byte = cinfo->dest->next_output_byte;
  state.free_in_buffer = cinfo->dest->free_in_buffer;
-  state.cur = entropy->saved;
+  ASSIGN_STATE(state.cur, entropy->saved);
  state.cinfo = cinfo;
-  state.simd = entropy->simd;

  /* Emit restart marker if needed */
  if (cinfo->restart_interval) {
@@ -734,7 +694,7 @@ encode_mcu_huff(j_compress_ptr cinfo, JBLOCKROW *MCU_data)
  /* Completed MCU, so update state */
  cinfo->dest->next_output_byte = state.next_output_byte;
  cinfo->dest->free_in_buffer = state.free_in_buffer;
-  entropy->saved = state.cur;
+  ASSIGN_STATE(entropy->saved, state.cur);

  /* Update restart-interval state too */
  if (cinfo->restart_interval) {
@@ -763,9 +723,8 @@ finish_pass_huff(j_compress_ptr cinfo)
  /* Load up working state ... flush_bits needs it */
  state.next_output_byte = cinfo->dest->next_output_byte;
  state.free_in_buffer = cinfo->dest->free_in_buffer;
-  state.cur = entropy->saved;
+  ASSIGN_STATE(state.cur, entropy->saved);
  state.cinfo = cinfo;
-  state.simd = entropy->simd;

  /* Flush out the last data */
  if (!flush_bits(&state))
@@ -774,7 +733,7 @@ finish_pass_huff(j_compress_ptr cinfo)
  /* Update state */
  cinfo->dest->next_output_byte = state.next_output_byte;
  cinfo->dest->free_in_buffer = state.free_in_buffer;
-  entropy->saved = state.cur;
+  ASSIGN_STATE(entropy->saved, state.cur);
 }


@@ -1,10 +1,8 @@
 /*
 * jcinit.c
 *
- * This file was part of the Independent JPEG Group's software:
 * Copyright (C) 1991-1997, Thomas G. Lane.
- * libjpeg-turbo Modifications:
- * Copyright (C) 2020, D. R. Commander.
+ * This file is part of the Independent JPEG Group's software.
 * For conditions of distribution and use, see the accompanying README.ijg
 * file.
 *
@@ -21,7 +19,6 @@
 #define JPEG_INTERNALS
 #include "jinclude.h"
 #include "jpeglib.h"
-#include "jpegcomp.h"


 /*
@@ -493,7 +493,7 @@ prepare_for_pass(j_compress_ptr cinfo)
    master->pass_type = output_pass;
    master->pass_number++;
 #endif
-    FALLTHROUGH                 /*FALLTHROUGH*/
+    /*FALLTHROUGH*/
  case output_pass:
    /* Do a data-output pass. */
    /* We need not repeat per-scan setup if prior optimization pass did it. */
@@ -4,10 +4,8 @@
 * This file was part of the Independent JPEG Group's software:
 * Copyright (C) 1995-1997, Thomas G. Lane.
 * libjpeg-turbo Modifications:
- * Copyright (C) 2011, 2015, 2018, 2021, D. R. Commander.
+ * Copyright (C) 2011, 2015, 2018, D. R. Commander.
 * Copyright (C) 2016, 2018, Matthieu Darbois.
- * Copyright (C) 2020, Arm Limited.
- * Copyright (C) 2021, Alex Richardson.
 * For conditions of distribution and use, see the accompanying README.ijg
 * file.
 *
@@ -45,28 +43,23 @@
 * memory footprint by 64k, which is important for some mobile applications
 * that create many isolated instances of libjpeg-turbo (web browsers, for
 * instance.)  This may improve performance on some mobile platforms as well.
- * This feature is enabled by default only on Arm processors, because some x86
+ * This feature is enabled by default only on ARM processors, because some x86
 * chips have a slow implementation of bsr, and the use of clz/bsr cannot be
 * shown to have a significant performance impact even on the x86 chips that
- * have a fast implementation of it.  When building for Armv6, you can
+ * have a fast implementation of it.  When building for ARMv6, you can
 * explicitly disable the use of clz/bsr by adding -mthumb to the compiler
 * flags (this defines __thumb__).
 */

 /* NOTE: Both GCC and Clang define __GNUC__ */
-#if (defined(__GNUC__) && (defined(__arm__) || defined(__aarch64__))) || \
-    defined(_M_ARM) || defined(_M_ARM64)
+#if defined(__GNUC__) && (defined(__arm__) || defined(__aarch64__))
 #if !defined(__thumb__) || defined(__thumb2__)
 #define USE_CLZ_INTRINSIC
 #endif
 #endif

 #ifdef USE_CLZ_INTRINSIC
-#if defined(_MSC_VER) && !defined(__clang__)
-#define JPEG_NBITS_NONZERO(x)  (32 - _CountLeadingZeros(x))
-#else
 #define JPEG_NBITS_NONZERO(x)  (32 - __builtin_clz(x))
-#endif
 #define JPEG_NBITS(x)          (x ? JPEG_NBITS_NONZERO(x) : 0)
 #else
 #include "jpeg_nbits_table.h"
@@ -176,26 +169,24 @@ INLINE
 METHODDEF(int)
 count_zeroes(size_t *x)
 {
-#if defined(HAVE_BUILTIN_CTZL)
  int result;
+#if defined(HAVE_BUILTIN_CTZL)
  result = __builtin_ctzl(*x);
  *x >>= result;
 #elif defined(HAVE_BITSCANFORWARD64)
-  unsigned long result;
  _BitScanForward64(&result, *x);
  *x >>= result;
 #elif defined(HAVE_BITSCANFORWARD)
-  unsigned long result;
  _BitScanForward(&result, *x);
  *x >>= result;
 #else
-  int result = 0;
+  result = 0;
  while ((*x & 1) == 0) {
    ++result;
    *x >>= 1;
  }
 #endif
-  return (int)result;
+  return result;
 }


@@ -681,7 +672,7 @@ encode_mcu_AC_first(j_compress_ptr cinfo, JBLOCKROW *MCU_data)
      emit_restart(entropy, entropy->next_restart_num);

 #ifdef WITH_SIMD
-  cvalue = values = (JCOEF *)PAD((JUINTPTR)values_unaligned, 16);
+  cvalue = values = (JCOEF *)PAD((size_t)values_unaligned, 16);
 #else
  /* Not using SIMD, so alignment is not needed */
  cvalue = values = values_unaligned;
@@ -869,7 +860,7 @@ encode_mcu_AC_refine_prepare(const JCOEF *block,

 #define ENCODE_COEFS_AC_REFINE(label) { \
  while (zerobits) { \
-    idx = count_zeroes(&zerobits); \
+    int idx = count_zeroes(&zerobits); \
    r += idx; \
    cabsvalue += idx; \
    signbits >>= idx; \
@@ -926,7 +917,7 @@ METHODDEF(boolean)
 encode_mcu_AC_refine(j_compress_ptr cinfo, JBLOCKROW *MCU_data)
 {
  phuff_entropy_ptr entropy = (phuff_entropy_ptr)cinfo->entropy;
-  register int temp, r, idx;
+  register int temp, r;
  char *BR_buffer;
  unsigned int BR;
  int Sl = cinfo->Se - cinfo->Ss + 1;
@@ -946,7 +937,7 @@ encode_mcu_AC_refine(j_compress_ptr cinfo, JBLOCKROW *MCU_data)
      emit_restart(entropy, entropy->next_restart_num);

 #ifdef WITH_SIMD
-  cabsvalue = absvalues = (JCOEF *)PAD((JUINTPTR)absvalues_unaligned, 16);
+  cabsvalue = absvalues = (JCOEF *)PAD((size_t)absvalues_unaligned, 16);
 #else
  /* Not using SIMD, so alignment is not needed */
  cabsvalue = absvalues = absvalues_unaligned;
@@ -977,7 +968,7 @@ encode_mcu_AC_refine(j_compress_ptr cinfo, JBLOCKROW *MCU_data)

  if (zerobits) {
    int diff = ((absvalues + DCTSIZE2 / 2) - cabsvalue);
-    idx = count_zeroes(&zerobits);
+    int idx = count_zeroes(&zerobits);
    signbits >>= idx;
    idx += diff;
    r += idx;
@@ -6,7 +6,7 @@
 * libjpeg-turbo Modifications:
 * Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
 * Copyright (C) 2014, MIPS Technologies, Inc., California.
- * Copyright (C) 2015, 2019, D. R. Commander.
+ * Copyright (C) 2015, D. R. Commander.
 * For conditions of distribution and use, see the accompanying README.ijg
 * file.
 *
@@ -103,7 +103,7 @@ expand_right_edge(JSAMPARRAY image_data, int num_rows, JDIMENSION input_cols,
  if (numcols > 0) {
    for (row = 0; row < num_rows; row++) {
      ptr = image_data[row] + input_cols;
-      pixval = ptr[-1];
+      pixval = ptr[-1];         /* don't need GETJSAMPLE() here */
      for (count = numcols; count > 0; count--)
        *ptr++ = pixval;
    }
@@ -174,7 +174,7 @@ int_downsample(j_compress_ptr cinfo, jpeg_component_info *compptr,
      for (v = 0; v < v_expand; v++) {
        inptr = input_data[inrow + v] + outcol_h;
        for (h = 0; h < h_expand; h++) {
-          outvalue += (JLONG)(*inptr++);
+          outvalue += (JLONG)GETJSAMPLE(*inptr++);
        }
      }
      *outptr++ = (JSAMPLE)((outvalue + numpix2) / numpix);
@@ -237,7 +237,8 @@ h2v1_downsample(j_compress_ptr cinfo, jpeg_component_info *compptr,
    inptr = input_data[outrow];
    bias = 0;                   /* bias = 0,1,0,1,... for successive samples */
    for (outcol = 0; outcol < output_cols; outcol++) {
-      *outptr++ = (JSAMPLE)((inptr[0] + inptr[1] + bias) >> 1);
+      *outptr++ =
+        (JSAMPLE)((GETJSAMPLE(*inptr) + GETJSAMPLE(inptr[1]) + bias) >> 1);
      bias ^= 1;                /* 0=>1, 1=>0 */
      inptr += 2;
    }
@@ -276,7 +277,8 @@ h2v2_downsample(j_compress_ptr cinfo, jpeg_component_info *compptr,
    bias = 1;                   /* bias = 1,2,1,2,... for successive samples */
    for (outcol = 0; outcol < output_cols; outcol++) {
      *outptr++ =
-        (JSAMPLE)((inptr0[0] + inptr0[1] + inptr1[0] + inptr1[1] + bias) >> 2);
+        (JSAMPLE)((GETJSAMPLE(*inptr0) + GETJSAMPLE(inptr0[1]) +
+                   GETJSAMPLE(*inptr1) + GETJSAMPLE(inptr1[1]) + bias) >> 2);
      bias ^= 3;                /* 1=>2, 2=>1 */
      inptr0 += 2;  inptr1 += 2;
    }
@@ -335,25 +337,33 @@ h2v2_smooth_downsample(j_compress_ptr cinfo, jpeg_component_info *compptr,
    below_ptr = input_data[inrow + 2];

    /* Special case for first column: pretend column -1 is same as column 0 */
-    membersum = inptr0[0] + inptr0[1] + inptr1[0] + inptr1[1];
-    neighsum = above_ptr[0] + above_ptr[1] + below_ptr[0] + below_ptr[1] +
-               inptr0[0] + inptr0[2] + inptr1[0] + inptr1[2];
+    membersum = GETJSAMPLE(*inptr0) + GETJSAMPLE(inptr0[1]) +
+                GETJSAMPLE(*inptr1) + GETJSAMPLE(inptr1[1]);
+    neighsum = GETJSAMPLE(*above_ptr) + GETJSAMPLE(above_ptr[1]) +
+               GETJSAMPLE(*below_ptr) + GETJSAMPLE(below_ptr[1]) +
+               GETJSAMPLE(*inptr0) + GETJSAMPLE(inptr0[2]) +
+               GETJSAMPLE(*inptr1) + GETJSAMPLE(inptr1[2]);
    neighsum += neighsum;
-    neighsum += above_ptr[0] + above_ptr[2] + below_ptr[0] + below_ptr[2];
+    neighsum += GETJSAMPLE(*above_ptr) + GETJSAMPLE(above_ptr[2]) +
+                GETJSAMPLE(*below_ptr) + GETJSAMPLE(below_ptr[2]);
    membersum = membersum * memberscale + neighsum * neighscale;
    *outptr++ = (JSAMPLE)((membersum + 32768) >> 16);
    inptr0 += 2;  inptr1 += 2;  above_ptr += 2;  below_ptr += 2;

    for (colctr = output_cols - 2; colctr > 0; colctr--) {
      /* sum of pixels directly mapped to this output element */
-      membersum = inptr0[0] + inptr0[1] + inptr1[0] + inptr1[1];
+      membersum = GETJSAMPLE(*inptr0) + GETJSAMPLE(inptr0[1]) +
+                  GETJSAMPLE(*inptr1) + GETJSAMPLE(inptr1[1]);
      /* sum of edge-neighbor pixels */
-      neighsum = above_ptr[0] + above_ptr[1] + below_ptr[0] + below_ptr[1] +
-                 inptr0[-1] + inptr0[2] + inptr1[-1] + inptr1[2];
+      neighsum = GETJSAMPLE(*above_ptr) + GETJSAMPLE(above_ptr[1]) +
+                 GETJSAMPLE(*below_ptr) + GETJSAMPLE(below_ptr[1]) +
+                 GETJSAMPLE(inptr0[-1]) + GETJSAMPLE(inptr0[2]) +
+                 GETJSAMPLE(inptr1[-1]) + GETJSAMPLE(inptr1[2]);
      /* The edge-neighbors count twice as much as corner-neighbors */
      neighsum += neighsum;
      /* Add in the corner-neighbors */
-      neighsum += above_ptr[-1] + above_ptr[2] + below_ptr[-1] + below_ptr[2];
+      neighsum += GETJSAMPLE(above_ptr[-1]) + GETJSAMPLE(above_ptr[2]) +
+                  GETJSAMPLE(below_ptr[-1]) + GETJSAMPLE(below_ptr[2]);
      /* form final output scaled up by 2^16 */
      membersum = membersum * memberscale + neighsum * neighscale;
      /* round, descale and output it */
@@ -362,11 +372,15 @@ h2v2_smooth_downsample(j_compress_ptr cinfo, jpeg_component_info *compptr,
    }

    /* Special case for last column */
-    membersum = inptr0[0] + inptr0[1] + inptr1[0] + inptr1[1];
-    neighsum = above_ptr[0] + above_ptr[1] + below_ptr[0] + below_ptr[1] +
-               inptr0[-1] + inptr0[1] + inptr1[-1] + inptr1[1];
+    membersum = GETJSAMPLE(*inptr0) + GETJSAMPLE(inptr0[1]) +
+                GETJSAMPLE(*inptr1) + GETJSAMPLE(inptr1[1]);
+    neighsum = GETJSAMPLE(*above_ptr) + GETJSAMPLE(above_ptr[1]) +
+               GETJSAMPLE(*below_ptr) + GETJSAMPLE(below_ptr[1]) +
+               GETJSAMPLE(inptr0[-1]) + GETJSAMPLE(inptr0[1]) +
+               GETJSAMPLE(inptr1[-1]) + GETJSAMPLE(inptr1[1]);
    neighsum += neighsum;
-    neighsum += above_ptr[-1] + above_ptr[1] + below_ptr[-1] + below_ptr[1];
+    neighsum += GETJSAMPLE(above_ptr[-1]) + GETJSAMPLE(above_ptr[1]) +
+                GETJSAMPLE(below_ptr[-1]) + GETJSAMPLE(below_ptr[1]);
    membersum = membersum * memberscale + neighsum * neighscale;
    *outptr = (JSAMPLE)((membersum + 32768) >> 16);

@@ -415,18 +429,21 @@ fullsize_smooth_downsample(j_compress_ptr cinfo, jpeg_component_info *compptr,
    below_ptr = input_data[outrow + 1];

    /* Special case for first column */
-    colsum = (*above_ptr++) + (*below_ptr++) + inptr[0];
-    membersum = *inptr++;
-    nextcolsum = above_ptr[0] + below_ptr[0] + inptr[0];
+    colsum = GETJSAMPLE(*above_ptr++) + GETJSAMPLE(*below_ptr++) +
+             GETJSAMPLE(*inptr);
+    membersum = GETJSAMPLE(*inptr++);
+    nextcolsum = GETJSAMPLE(*above_ptr) + GETJSAMPLE(*below_ptr) +
+                 GETJSAMPLE(*inptr);
    neighsum = colsum + (colsum - membersum) + nextcolsum;
    membersum = membersum * memberscale + neighsum * neighscale;
    *outptr++ = (JSAMPLE)((membersum + 32768) >> 16);
    lastcolsum = colsum;  colsum = nextcolsum;

    for (colctr = output_cols - 2; colctr > 0; colctr--) {
-      membersum = *inptr++;
+      membersum = GETJSAMPLE(*inptr++);
      above_ptr++;  below_ptr++;
-      nextcolsum = above_ptr[0] + below_ptr[0] + inptr[0];
+      nextcolsum = GETJSAMPLE(*above_ptr) + GETJSAMPLE(*below_ptr) +
+                   GETJSAMPLE(*inptr);
      neighsum = lastcolsum + (colsum - membersum) + nextcolsum;
      membersum = membersum * memberscale + neighsum * neighscale;
      *outptr++ = (JSAMPLE)((membersum + 32768) >> 16);
@@ -434,7 +451,7 @@ fullsize_smooth_downsample(j_compress_ptr cinfo, jpeg_component_info *compptr,
    }

    /* Special case for last column */
-    membersum = *inptr;
+    membersum = GETJSAMPLE(*inptr);
    neighsum = lastcolsum + (colsum - membersum) + colsum;
    membersum = membersum * memberscale + neighsum * neighscale;
    *outptr = (JSAMPLE)((membersum + 32768) >> 16);
@@ -4,8 +4,8 @@
 * This file was part of the Independent JPEG Group's software:
 * Copyright (C) 1995-1998, Thomas G. Lane.
 * Modified 2000-2009 by Guido Vollbeding.
- * libjpeg-turbo Modifications:
- * Copyright (C) 2020, D. R. Commander.
+ * It was modified by The libjpeg-turbo Project to include only code relevant
+ * to libjpeg-turbo.
 * For conditions of distribution and use, see the accompanying README.ijg
 * file.
 *
@@ -17,7 +17,6 @@
 #define JPEG_INTERNALS
 #include "jinclude.h"
 #include "jpeglib.h"
-#include "jpegcomp.h"


 /* Forward declarations */
@@ -23,7 +23,6 @@
 #include "jinclude.h"
 #include "jpeglib.h"
 #include "jdmaster.h"
-#include "jconfigint.h"


 /*
@@ -309,7 +308,7 @@ jpeg_consume_input(j_decompress_ptr cinfo)
    /* Initialize application's data source module */
    (*cinfo->src->init_source) (cinfo);
    cinfo->global_state = DSTATE_INHEADER;
-    FALLTHROUGH                 /*FALLTHROUGH*/
+    /*FALLTHROUGH*/
  case DSTATE_INHEADER:
    retcode = (*cinfo->inputctl->consume_input) (cinfo);
    if (retcode == JPEG_REACHED_SOS) { /* Found SOS, prepare to decompress */
@@ -4,7 +4,7 @@
 * This file was part of the Independent JPEG Group's software:
 * Copyright (C) 1994-1996, Thomas G. Lane.
 * libjpeg-turbo Modifications:
- * Copyright (C) 2010, 2015-2020, D. R. Commander.
+ * Copyright (C) 2010, 2015-2018, D. R. Commander.
 * Copyright (C) 2015, Google, Inc.
 * For conditions of distribution and use, see the accompanying README.ijg
 * file.
@@ -21,8 +21,6 @@
 #include "jinclude.h"
 #include "jdmainct.h"
 #include "jdcoefct.h"
-#include "jdmaster.h"
-#include "jdmerge.h"
 #include "jdsample.h"
 #include "jmemsys.h"

@@ -318,10 +316,6 @@ LOCAL(void)
 read_and_discard_scanlines(j_decompress_ptr cinfo, JDIMENSION num_lines)
 {
  JDIMENSION n;
-  my_master_ptr master = (my_master_ptr)cinfo->master;
-  JSAMPLE dummy_sample[1] = { 0 };
-  JSAMPROW dummy_row = dummy_sample;
-  JSAMPARRAY scanlines = NULL;
  void (*color_convert) (j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
                         JDIMENSION input_row, JSAMPARRAY output_buf,
                         int num_rows) = NULL;
@@ -331,10 +325,6 @@ read_and_discard_scanlines(j_decompress_ptr cinfo, JDIMENSION num_lines)
  if (cinfo->cconvert && cinfo->cconvert->color_convert) {
    color_convert = cinfo->cconvert->color_convert;
    cinfo->cconvert->color_convert = noop_convert;
-    /* This just prevents UBSan from complaining about adding 0 to a NULL
-     * pointer.  The pointer isn't actually used.
-     */
-    scanlines = &dummy_row;
  }

  if (cinfo->cquantize && cinfo->cquantize->color_quantize) {
@@ -342,13 +332,8 @@ read_and_discard_scanlines(j_decompress_ptr cinfo, JDIMENSION num_lines)
    cinfo->cquantize->color_quantize = noop_quantize;
  }

-  if (master->using_merged_upsample && cinfo->max_v_samp_factor == 2) {
-    my_merged_upsample_ptr upsample = (my_merged_upsample_ptr)cinfo->upsample;
-    scanlines = &upsample->spare_row;
-  }
-
  for (n = 0; n < num_lines; n++)
-    jpeg_read_scanlines(cinfo, scanlines, 1);
+    jpeg_read_scanlines(cinfo, NULL, 1);

  if (color_convert)
    cinfo->cconvert->color_convert = color_convert;
@@ -368,12 +353,6 @@ increment_simple_rowgroup_ctr(j_decompress_ptr cinfo, JDIMENSION rows)
 {
  JDIMENSION rows_left;
  my_main_ptr main_ptr = (my_main_ptr)cinfo->main;
-  my_master_ptr master = (my_master_ptr)cinfo->master;
-
-  if (master->using_merged_upsample && cinfo->max_v_samp_factor == 2) {
-    read_and_discard_scanlines(cinfo, rows);
-    return;
-  }

  /* Increment the counter to the next row group after the skipped rows. */
  main_ptr->rowgroup_ctr += rows / cinfo->max_v_samp_factor;
@@ -403,27 +382,21 @@ jpeg_skip_scanlines(j_decompress_ptr cinfo, JDIMENSION num_lines)
 {
  my_main_ptr main_ptr = (my_main_ptr)cinfo->main;
  my_coef_ptr coef = (my_coef_ptr)cinfo->coef;
-  my_master_ptr master = (my_master_ptr)cinfo->master;
  my_upsample_ptr upsample = (my_upsample_ptr)cinfo->upsample;
  JDIMENSION i, x;
  int y;
  JDIMENSION lines_per_iMCU_row, lines_left_in_iMCU_row, lines_after_iMCU_row;
  JDIMENSION lines_to_skip, lines_to_read;

-  /* Two-pass color quantization is not supported. */
-  if (cinfo->quantize_colors && cinfo->two_pass_quantize)
-    ERREXIT(cinfo, JERR_NOTIMPL);
-
  if (cinfo->global_state != DSTATE_SCANNING)
    ERREXIT1(cinfo, JERR_BAD_STATE, cinfo->global_state);

  /* Do not skip past the bottom of the image. */
  if (cinfo->output_scanline + num_lines >= cinfo->output_height) {
-    num_lines = cinfo->output_height - cinfo->output_scanline;
    cinfo->output_scanline = cinfo->output_height;
    (*cinfo->inputctl->finish_input_pass) (cinfo);
    cinfo->inputctl->eoi_reached = TRUE;
-    return num_lines;
+    return cinfo->output_height - cinfo->output_scanline;
  }

  if (num_lines == 0)
@@ -472,10 +445,8 @@ jpeg_skip_scanlines(j_decompress_ptr cinfo, JDIMENSION num_lines)
    main_ptr->buffer_full = FALSE;
    main_ptr->rowgroup_ctr = 0;
    main_ptr->context_state = CTX_PREPARE_FOR_IMCU;
-    if (!master->using_merged_upsample) {
-      upsample->next_row_out = cinfo->max_v_samp_factor;
-      upsample->rows_to_go = cinfo->output_height - cinfo->output_scanline;
-    }
+    upsample->next_row_out = cinfo->max_v_samp_factor;
+    upsample->rows_to_go = cinfo->output_height - cinfo->output_scanline;
  }

  /* Skipping is much simpler when context rows are not required. */
@@ -487,10 +458,8 @@ jpeg_skip_scanlines(j_decompress_ptr cinfo, JDIMENSION num_lines)
      cinfo->output_scanline += lines_left_in_iMCU_row;
      main_ptr->buffer_full = FALSE;
      main_ptr->rowgroup_ctr = 0;
-      if (!master->using_merged_upsample) {
-        upsample->next_row_out = cinfo->max_v_samp_factor;
-        upsample->rows_to_go = cinfo->output_height - cinfo->output_scanline;
-      }
+      upsample->next_row_out = cinfo->max_v_samp_factor;
+      upsample->rows_to_go = cinfo->output_height - cinfo->output_scanline;
    }
  }

@@ -525,8 +494,7 @@ jpeg_skip_scanlines(j_decompress_ptr cinfo, JDIMENSION num_lines)
      cinfo->output_iMCU_row += lines_to_skip / lines_per_iMCU_row;
      increment_simple_rowgroup_ctr(cinfo, lines_to_read);
    }
-    if (!master->using_merged_upsample)
-      upsample->rows_to_go = cinfo->output_height - cinfo->output_scanline;
+    upsample->rows_to_go = cinfo->output_height - cinfo->output_scanline;
    return num_lines;
  }

@@ -538,8 +506,6 @@ jpeg_skip_scanlines(j_decompress_ptr cinfo, JDIMENSION num_lines)
         * decoded coefficients.  This is ~5% faster for large subsets, but
         * it's tough to tell a difference for smaller images.
         */
-        if (!cinfo->entropy->insufficient_data)
-          cinfo->master->last_good_iMCU_row = cinfo->input_iMCU_row;
        (*cinfo->entropy->decode_mcu) (cinfo, NULL);
      }
    }
@@ -569,8 +535,7 @@ jpeg_skip_scanlines(j_decompress_ptr cinfo, JDIMENSION num_lines)
   * bit odd, since "rows_to_go" seems to be redundantly keeping track of
   * output_scanline.
   */
-  if (!master->using_merged_upsample)
-    upsample->rows_to_go = cinfo->output_height - cinfo->output_scanline;
+  upsample->rows_to_go = cinfo->output_height - cinfo->output_scanline;

  /* Always skip the requested number of lines. */
  return num_lines;
@@ -4,7 +4,7 @@
 * This file was part of the Independent JPEG Group's software:
 * Developed 1997-2015 by Guido Vollbeding.
 * libjpeg-turbo Modifications:
- * Copyright (C) 2015-2020, D. R. Commander.
+ * Copyright (C) 2015-2018, D. R. Commander.
 * For conditions of distribution and use, see the accompanying README.ijg
 * file.
 *
@@ -80,7 +80,7 @@ get_byte(j_decompress_ptr cinfo)
    if (!(*src->fill_input_buffer) (cinfo))
      ERREXIT(cinfo, JERR_CANT_SUSPEND);
  src->bytes_in_buffer--;
-  return *src->next_input_byte++;
+  return GETJOCTET(*src->next_input_byte++);
 }


@@ -665,16 +665,8 @@ bad:
    for (ci = 0; ci < cinfo->comps_in_scan; ci++) {
      int coefi, cindex = cinfo->cur_comp_info[ci]->component_index;
      int *coef_bit_ptr = &cinfo->coef_bits[cindex][0];
-      int *prev_coef_bit_ptr =
-        &cinfo->coef_bits[cindex + cinfo->num_components][0];
      if (cinfo->Ss && coef_bit_ptr[0] < 0) /* AC without prior DC scan */
        WARNMS2(cinfo, JWRN_BOGUS_PROGRESSION, cindex, 0);
-      for (coefi = MIN(cinfo->Ss, 1); coefi <= MAX(cinfo->Se, 9); coefi++) {
-        if (cinfo->input_scan_number > 1)
-          prev_coef_bit_ptr[coefi] = coef_bit_ptr[coefi];
-        else
-          prev_coef_bit_ptr[coefi] = 0;
-      }
      for (coefi = cinfo->Ss; coefi <= cinfo->Se; coefi++) {
        int expected = (coef_bit_ptr[coefi] < 0) ? 0 : coef_bit_ptr[coefi];
        if (cinfo->Ah != expected)
@@ -735,7 +727,6 @@ bad:
  entropy->c = 0;
  entropy->a = 0;
  entropy->ct = -16;    /* force reading 2 initial bytes to fill C */
-  entropy->pub.insufficient_data = FALSE;

  /* Initialize restart counter */
  entropy->restarts_to_go = cinfo->restart_interval;
@@ -772,7 +763,7 @@ jinit_arith_decoder(j_decompress_ptr cinfo)
    int *coef_bit_ptr, ci;
    cinfo->coef_bits = (int (*)[DCTSIZE2])
      (*cinfo->mem->alloc_small) ((j_common_ptr)cinfo, JPOOL_IMAGE,
-                                  cinfo->num_components * 2 * DCTSIZE2 *
+                                  cinfo->num_components * DCTSIZE2 *
                                  sizeof(int));
    coef_bit_ptr = &cinfo->coef_bits[0][0];
    for (ci = 0; ci < cinfo->num_components; ci++)
@@ -143,7 +143,8 @@ empty_mem_output_buffer(j_compress_ptr cinfo)

  MEMCOPY(nextbuffer, dest->buffer, dest->bufsize);

-  free(dest->newbuffer);
+  if (dest->newbuffer != NULL)
+    free(dest->newbuffer);

  dest->newbuffer = nextbuffer;

@@ -5,8 +5,8 @@
 * Copyright (C) 1994-1997, Thomas G. Lane.
 * libjpeg-turbo Modifications:
 * Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
- * Copyright (C) 2010, 2015-2016, 2019-2020, D. R. Commander.
- * Copyright (C) 2015, 2020, Google, Inc.
+ * Copyright (C) 2010, 2015-2016, D. R. Commander.
+ * Copyright (C) 2015, Google, Inc.
 * For conditions of distribution and use, see the accompanying README.ijg
 * file.
 *
@@ -102,8 +102,6 @@ decompress_onepass(j_decompress_ptr cinfo, JSAMPIMAGE output_buf)
      /* Try to fetch an MCU.  Entropy decoder expects buffer to be zeroed. */
      jzero_far((void *)coef->MCU_buffer[0],
                (size_t)(cinfo->blocks_in_MCU * sizeof(JBLOCK)));
-      if (!cinfo->entropy->insufficient_data)
-        cinfo->master->last_good_iMCU_row = cinfo->input_iMCU_row;
      if (!(*cinfo->entropy->decode_mcu) (cinfo, coef->MCU_buffer)) {
        /* Suspension forced; update state counters and exit */
        coef->MCU_vert_offset = yoffset;
@@ -229,8 +227,6 @@ consume_data(j_decompress_ptr cinfo)
          }
        }
      }
-      if (!cinfo->entropy->insufficient_data)
-        cinfo->master->last_good_iMCU_row = cinfo->input_iMCU_row;
      /* Try to fetch the MCU. */
      if (!(*cinfo->entropy->decode_mcu) (cinfo, coef->MCU_buffer)) {
        /* Suspension forced; update state counters and exit */
@@ -330,22 +326,19 @@ decompress_data(j_decompress_ptr cinfo, JSAMPIMAGE output_buf)
 #ifdef BLOCK_SMOOTHING_SUPPORTED

 /*
- * This code applies interblock smoothing; the first 9 AC coefficients are
- * estimated from the DC values of a DCT block and its 24 neighboring blocks.
+ * This code applies interblock smoothing as described by section K.8
+ * of the JPEG standard: the first 5 AC coefficients are estimated from
+ * the DC values of a DCT block and its 8 neighboring blocks.
 * We apply smoothing only for progressive JPEG decoding, and only if
 * the coefficients it can estimate are not yet known to full precision.
 */

-/* Natural-order array positions of the first 9 zigzag-order coefficients */
+/* Natural-order array positions of the first 5 zigzag-order coefficients */
 #define Q01_POS  1
 #define Q10_POS  8
 #define Q20_POS  16
 #define Q11_POS  9
 #define Q02_POS  2
-#define Q03_POS  3
-#define Q12_POS  10
-#define Q21_POS  17
-#define Q30_POS  24

 /*
 * Determine whether block smoothing is applicable and safe.
@@ -363,8 +356,8 @@ smoothing_ok(j_decompress_ptr cinfo)
  int ci, coefi;
  jpeg_component_info *compptr;
  JQUANT_TBL *qtable;
-  int *coef_bits, *prev_coef_bits;
-  int *coef_bits_latch, *prev_coef_bits_latch;
+  int *coef_bits;
+  int *coef_bits_latch;

  if (!cinfo->progressive_mode || cinfo->coef_bits == NULL)
    return FALSE;
@@ -373,47 +366,34 @@ smoothing_ok(j_decompress_ptr cinfo)
  if (coef->coef_bits_latch == NULL)
    coef->coef_bits_latch = (int *)
      (*cinfo->mem->alloc_small) ((j_common_ptr)cinfo, JPOOL_IMAGE,
-                                  cinfo->num_components * 2 *
+                                  cinfo->num_components *
                                  (SAVED_COEFS * sizeof(int)));
  coef_bits_latch = coef->coef_bits_latch;
-  prev_coef_bits_latch =
-    &coef->coef_bits_latch[cinfo->num_components * SAVED_COEFS];

  for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
       ci++, compptr++) {
    /* All components' quantization values must already be latched. */
    if ((qtable = compptr->quant_table) == NULL)
      return FALSE;
-    /* Verify DC & first 9 AC quantizers are nonzero to avoid zero-divide. */
+    /* Verify DC & first 5 AC quantizers are nonzero to avoid zero-divide. */
    if (qtable->quantval[0] == 0 ||
        qtable->quantval[Q01_POS] == 0 ||
        qtable->quantval[Q10_POS] == 0 ||
        qtable->quantval[Q20_POS] == 0 ||
        qtable->quantval[Q11_POS] == 0 ||
-        qtable->quantval[Q02_POS] == 0 ||
-        qtable->quantval[Q03_POS] == 0 ||
-        qtable->quantval[Q12_POS] == 0 ||
-        qtable->quantval[Q21_POS] == 0 ||
-        qtable->quantval[Q30_POS] == 0)
+        qtable->quantval[Q02_POS] == 0)
      return FALSE;
    /* DC values must be at least partly known for all components. */
    coef_bits = cinfo->coef_bits[ci];
-    prev_coef_bits = cinfo->coef_bits[ci + cinfo->num_components];
    if (coef_bits[0] < 0)
      return FALSE;
-    coef_bits_latch[0] = coef_bits[0];
    /* Block smoothing is helpful if some AC coefficients remain inaccurate. */
-    for (coefi = 1; coefi < SAVED_COEFS; coefi++) {
-      if (cinfo->input_scan_number > 1)
-        prev_coef_bits_latch[coefi] = prev_coef_bits[coefi];
-      else
-        prev_coef_bits_latch[coefi] = -1;
+    for (coefi = 1; coefi <= 5; coefi++) {
      coef_bits_latch[coefi] = coef_bits[coefi];
      if (coef_bits[coefi] != 0)
        smoothing_useful = TRUE;
    }
    coef_bits_latch += SAVED_COEFS;
-    prev_coef_bits_latch += SAVED_COEFS;
  }

  return smoothing_useful;
@@ -432,20 +412,17 @@ decompress_smooth_data(j_decompress_ptr cinfo, JSAMPIMAGE output_buf)
  JDIMENSION block_num, last_block_column;
  int ci, block_row, block_rows, access_rows;
  JBLOCKARRAY buffer;
-  JBLOCKROW buffer_ptr, prev_prev_block_row, prev_block_row;
-  JBLOCKROW next_block_row, next_next_block_row;
+  JBLOCKROW buffer_ptr, prev_block_row, next_block_row;
  JSAMPARRAY output_ptr;
  JDIMENSION output_col;
  jpeg_component_info *compptr;
  inverse_DCT_method_ptr inverse_DCT;
-  boolean change_dc;
+  boolean first_row, last_row;
  JCOEF *workspace;
  int *coef_bits;
  JQUANT_TBL *quanttbl;
-  JLONG Q00, Q01, Q02, Q03 = 0, Q10, Q11, Q12 = 0, Q20, Q21 = 0, Q30 = 0, num;
-  int DC01, DC02, DC03, DC04, DC05, DC06, DC07, DC08, DC09, DC10, DC11, DC12,
-      DC13, DC14, DC15, DC16, DC17, DC18, DC19, DC20, DC21, DC22, DC23, DC24,
-      DC25;
+  JLONG Q00, Q01, Q02, Q10, Q11, Q20, num;
+  int DC1, DC2, DC3, DC4, DC5, DC6, DC7, DC8, DC9;
  int Al, pred;

  /* Keep a local variable to avoid looking it up more than once */
@@ -457,10 +434,10 @@ decompress_smooth_data(j_decompress_ptr cinfo, JSAMPIMAGE output_buf)
    if (cinfo->input_scan_number == cinfo->output_scan_number) {
      /* If input is working on current scan, we ordinarily want it to
       * have completed the current row.  But if input scan is DC,
-       * we want it to keep two rows ahead so that next two block rows' DC
+       * we want it to keep one row ahead so that next block row's DC
       * values are up to date.
       */
-      JDIMENSION delta = (cinfo->Ss == 0) ? 2 : 0;
+      JDIMENSION delta = (cinfo->Ss == 0) ? 1 : 0;
      if (cinfo->input_iMCU_row > cinfo->output_iMCU_row + delta)
        break;
    }
@@ -475,53 +452,34 @@ decompress_smooth_data(j_decompress_ptr cinfo, JSAMPIMAGE output_buf)
    if (!compptr->component_needed)
      continue;
    /* Count non-dummy DCT block rows in this iMCU row. */
-    if (cinfo->output_iMCU_row < last_iMCU_row - 1) {
-      block_rows = compptr->v_samp_factor;
-      access_rows = block_rows * 3; /* this and next two iMCU rows */
-    } else if (cinfo->output_iMCU_row < last_iMCU_row) {
+    if (cinfo->output_iMCU_row < last_iMCU_row) {
      block_rows = compptr->v_samp_factor;
      access_rows = block_rows * 2; /* this and next iMCU row */
+      last_row = FALSE;
    } else {
      /* NB: can't use last_row_height here; it is input-side-dependent! */
      block_rows = (int)(compptr->height_in_blocks % compptr->v_samp_factor);
      if (block_rows == 0) block_rows = compptr->v_samp_factor;
      access_rows = block_rows; /* this iMCU row only */
+      last_row = TRUE;
    }
    /* Align the virtual buffer for this component. */
-    if (cinfo->output_iMCU_row > 1) {
-      access_rows += 2 * compptr->v_samp_factor; /* prior two iMCU rows too */
-      buffer = (*cinfo->mem->access_virt_barray)
-        ((j_common_ptr)cinfo, coef->whole_image[ci],
-         (cinfo->output_iMCU_row - 2) * compptr->v_samp_factor,
-         (JDIMENSION)access_rows, FALSE);
-      buffer += 2 * compptr->v_samp_factor; /* point to current iMCU row */
-    } else if (cinfo->output_iMCU_row > 0) {
+    if (cinfo->output_iMCU_row > 0) {
+      access_rows += compptr->v_samp_factor; /* prior iMCU row too */
      buffer = (*cinfo->mem->access_virt_barray)
        ((j_common_ptr)cinfo, coef->whole_image[ci],
         (cinfo->output_iMCU_row - 1) * compptr->v_samp_factor,
         (JDIMENSION)access_rows, FALSE);
      buffer += compptr->v_samp_factor; /* point to current iMCU row */
+      first_row = FALSE;
    } else {
      buffer = (*cinfo->mem->access_virt_barray)
        ((j_common_ptr)cinfo, coef->whole_image[ci],
         (JDIMENSION)0, (JDIMENSION)access_rows, FALSE);
+      first_row = TRUE;
    }
-    /* Fetch component-dependent info.
-     * If the current scan is incomplete, then we use the component-dependent
-     * info from the previous scan.
-     */
-    if (cinfo->output_iMCU_row > cinfo->master->last_good_iMCU_row)
-      coef_bits =
-        coef->coef_bits_latch + ((ci + cinfo->num_components) * SAVED_COEFS);
-    else
-      coef_bits = coef->coef_bits_latch + (ci * SAVED_COEFS);
-
-    /* We only do DC interpolation if no AC coefficient data is available. */
-    change_dc =
-      coef_bits[1] == -1 && coef_bits[2] == -1 && coef_bits[3] == -1 &&
-      coef_bits[4] == -1 && coef_bits[5] == -1 && coef_bits[6] == -1 &&
-      coef_bits[7] == -1 && coef_bits[8] == -1 && coef_bits[9] == -1;
-
+    /* Fetch component-dependent info */
+    coef_bits = coef->coef_bits_latch + (ci * SAVED_COEFS);
    quanttbl = compptr->quant_table;
    Q00 = quanttbl->quantval[0];
    Q01 = quanttbl->quantval[Q01_POS];
@@ -529,51 +487,25 @@ decompress_smooth_data(j_decompress_ptr cinfo, JSAMPIMAGE output_buf)
    Q20 = quanttbl->quantval[Q20_POS];
    Q11 = quanttbl->quantval[Q11_POS];
    Q02 = quanttbl->quantval[Q02_POS];
-    if (change_dc) {
-      Q03 = quanttbl->quantval[Q03_POS];
-      Q12 = quanttbl->quantval[Q12_POS];
-      Q21 = quanttbl->quantval[Q21_POS];
-      Q30 = quanttbl->quantval[Q30_POS];
-    }
    inverse_DCT = cinfo->idct->inverse_DCT[ci];
    output_ptr = output_buf[ci];
    /* Loop over all DCT blocks to be processed. */
    for (block_row = 0; block_row < block_rows; block_row++) {
      buffer_ptr = buffer[block_row] + cinfo->master->first_MCU_col[ci];
-
-      if (block_row > 0 || cinfo->output_iMCU_row > 0)
-        prev_block_row =
-          buffer[block_row - 1] + cinfo->master->first_MCU_col[ci];
-      else
+      if (first_row && block_row == 0)
        prev_block_row = buffer_ptr;
-
-      if (block_row > 1 || cinfo->output_iMCU_row > 1)
-        prev_prev_block_row =
-          buffer[block_row - 2] + cinfo->master->first_MCU_col[ci];
-      else
-        prev_prev_block_row = prev_block_row;
-
-      if (block_row < block_rows - 1 || cinfo->output_iMCU_row < last_iMCU_row)
-        next_block_row =
-          buffer[block_row + 1] + cinfo->master->first_MCU_col[ci];
      else
+        prev_block_row = buffer[block_row - 1];
+      if (last_row && block_row == block_rows - 1)
        next_block_row = buffer_ptr;
-
-      if (block_row < block_rows - 2 ||
-          cinfo->output_iMCU_row < last_iMCU_row - 1)
-        next_next_block_row =
-          buffer[block_row + 2] + cinfo->master->first_MCU_col[ci];
      else
-        next_next_block_row = next_block_row;
-
+        next_block_row = buffer[block_row + 1];
      /* We fetch the surrounding DC values using a sliding-register approach.
-       * Initialize all 25 here so as to do the right thing on narrow pics.
+       * Initialize all nine here so as to do the right thing on narrow pics.
       */
-      DC01 = DC02 = DC03 = DC04 = DC05 = (int)prev_prev_block_row[0][0];
-      DC06 = DC07 = DC08 = DC09 = DC10 = (int)prev_block_row[0][0];
-      DC11 = DC12 = DC13 = DC14 = DC15 = (int)buffer_ptr[0][0];
-      DC16 = DC17 = DC18 = DC19 = DC20 = (int)next_block_row[0][0];
-      DC21 = DC22 = DC23 = DC24 = DC25 = (int)next_next_block_row[0][0];
+      DC1 = DC2 = DC3 = (int)prev_block_row[0][0];
+      DC4 = DC5 = DC6 = (int)buffer_ptr[0][0];
+      DC7 = DC8 = DC9 = (int)next_block_row[0][0];
      output_col = 0;
      last_block_column = compptr->width_in_blocks - 1;
      for (block_num = cinfo->master->first_MCU_col[ci];
@@ -581,39 +513,18 @@ decompress_smooth_data(j_decompress_ptr cinfo, JSAMPIMAGE output_buf)
        /* Fetch current DCT block into workspace so we can modify it. */
        jcopy_block_row(buffer_ptr, (JBLOCKROW)workspace, (JDIMENSION)1);
        /* Update DC values */
-        if (block_num == cinfo->master->first_MCU_col[ci] &&
-            block_num < last_block_column) {
-          DC04 = (int)prev_prev_block_row[1][0];
-          DC09 = (int)prev_block_row[1][0];
-          DC14 = (int)buffer_ptr[1][0];
-          DC19 = (int)next_block_row[1][0];
-          DC24 = (int)next_next_block_row[1][0];
+        if (block_num < last_block_column) {
+          DC3 = (int)prev_block_row[1][0];
+          DC6 = (int)buffer_ptr[1][0];
+          DC9 = (int)next_block_row[1][0];
        }
-        if (block_num + 1 < last_block_column) {
-          DC05 = (int)prev_prev_block_row[2][0];
-          DC10 = (int)prev_block_row[2][0];
-          DC15 = (int)buffer_ptr[2][0];
-          DC20 = (int)next_block_row[2][0];
-          DC25 = (int)next_next_block_row[2][0];
-        }
-        /* If DC interpolation is enabled, compute coefficient estimates using
-         * a Gaussian-like kernel, keeping the averages of the DC values.
-         *
-         * If DC interpolation is disabled, compute coefficient estimates using
-         * an algorithm similar to the one described in Section K.8 of the JPEG
-         * standard, except applied to a 5x5 window rather than a 3x3 window.
-         *
-         * An estimate is applied only if the coefficient is still zero and is
-         * not known to be fully accurate.
+        /* Compute coefficient estimates per K.8.
+         * An estimate is applied only if coefficient is still zero,
+         * and is not known to be fully accurate.
         */
        /* AC01 */
        if ((Al = coef_bits[1]) != 0 && workspace[1] == 0) {
-          num = Q00 * (change_dc ?
-                (-DC01 - DC02 + DC04 + DC05 - 3 * DC06 + 13 * DC07 -
-                 13 * DC09 + 3 * DC10 - 3 * DC11 + 38 * DC12 - 38 * DC14 +
-                 3 * DC15 - 3 * DC16 + 13 * DC17 - 13 * DC19 + 3 * DC20 -
-                 DC21 - DC22 + DC24 + DC25) :
-                (-7 * DC11 + 50 * DC12 - 50 * DC14 + 7 * DC15));
+          num = 36 * Q00 * (DC4 - DC6);
          if (num >= 0) {
            pred = (int)(((Q01 << 7) + num) / (Q01 << 8));
            if (Al > 0 && pred >= (1 << Al))
@@ -628,12 +539,7 @@ decompress_smooth_data(j_decompress_ptr cinfo, JSAMPIMAGE output_buf)
        }
        /* AC10 */
        if ((Al = coef_bits[2]) != 0 && workspace[8] == 0) {
-          num = Q00 * (change_dc ?
-                (-DC01 - 3 * DC02 - 3 * DC03 - 3 * DC04 - DC05 - DC06 +
-                 13 * DC07 + 38 * DC08 + 13 * DC09 - DC10 + DC16 -
-                 13 * DC17 - 38 * DC18 - 13 * DC19 + DC20 + DC21 +
-                 3 * DC22 + 3 * DC23 + 3 * DC24 + DC25) :
-                (-7 * DC03 + 50 * DC08 - 50 * DC18 + 7 * DC23));
+          num = 36 * Q00 * (DC2 - DC8);
          if (num >= 0) {
            pred = (int)(((Q10 << 7) + num) / (Q10 << 8));
            if (Al > 0 && pred >= (1 << Al))
@@ -648,10 +554,7 @@ decompress_smooth_data(j_decompress_ptr cinfo, JSAMPIMAGE output_buf)
        }
        /* AC20 */
        if ((Al = coef_bits[3]) != 0 && workspace[16] == 0) {
-          num = Q00 * (change_dc ?
-                (DC03 + 2 * DC07 + 7 * DC08 + 2 * DC09 - 5 * DC12 - 14 * DC13 -
-                 5 * DC14 + 2 * DC17 + 7 * DC18 + 2 * DC19 + DC23) :
-                (-DC03 + 13 * DC08 - 24 * DC13 + 13 * DC18 - DC23));
+          num = 9 * Q00 * (DC2 + DC8 - 2 * DC5);
          if (num >= 0) {
            pred = (int)(((Q20 << 7) + num) / (Q20 << 8));
            if (Al > 0 && pred >= (1 << Al))
@@ -666,11 +569,7 @@ decompress_smooth_data(j_decompress_ptr cinfo, JSAMPIMAGE output_buf)
        }
        /* AC11 */
        if ((Al = coef_bits[4]) != 0 && workspace[9] == 0) {
-          num = Q00 * (change_dc ?
-                (-DC01 + DC05 + 9 * DC07 - 9 * DC09 - 9 * DC17 +
-                 9 * DC19 + DC21 - DC25) :
-                (DC10 + DC16 - 10 * DC17 + 10 * DC19 - DC02 - DC20 + DC22 -
-                 DC24 + DC04 - DC06 + 10 * DC07 - 10 * DC09));
+          num = 5 * Q00 * (DC1 - DC3 - DC7 + DC9);
          if (num >= 0) {
            pred = (int)(((Q11 << 7) + num) / (Q11 << 8));
            if (Al > 0 && pred >= (1 << Al))
@@ -685,10 +584,7 @@ decompress_smooth_data(j_decompress_ptr cinfo, JSAMPIMAGE output_buf)
        }
        /* AC02 */
        if ((Al = coef_bits[5]) != 0 && workspace[2] == 0) {
-          num = Q00 * (change_dc ?
-                (2 * DC07 - 5 * DC08 + 2 * DC09 + DC11 + 7 * DC12 - 14 * DC13 +
-                 7 * DC14 + DC15 + 2 * DC17 - 5 * DC18 + 2 * DC19) :
-                (-DC11 + 13 * DC12 - 24 * DC13 + 13 * DC14 - DC15));
+          num = 9 * Q00 * (DC4 + DC6 - 2 * DC5);
          if (num >= 0) {
            pred = (int)(((Q02 << 7) + num) / (Q02 << 8));
            if (Al > 0 && pred >= (1 << Al))
@@ -701,96 +597,14 @@ decompress_smooth_data(j_decompress_ptr cinfo, JSAMPIMAGE output_buf)
          }
          workspace[2] = (JCOEF)pred;
        }
-        if (change_dc) {
-          /* AC03 */
-          if ((Al = coef_bits[6]) != 0 && workspace[3] == 0) {
-            num = Q00 * (DC07 - DC09 + 2 * DC12 - 2 * DC14 + DC17 - DC19);
-            if (num >= 0) {
-              pred = (int)(((Q03 << 7) + num) / (Q03 << 8));
-              if (Al > 0 && pred >= (1 << Al))
-                pred = (1 << Al) - 1;
-            } else {
-              pred = (int)(((Q03 << 7) - num) / (Q03 << 8));
-              if (Al > 0 && pred >= (1 << Al))
-                pred = (1 << Al) - 1;
-              pred = -pred;
-            }
-            workspace[3] = (JCOEF)pred;
-          }
-          /* AC12 */
-          if ((Al = coef_bits[7]) != 0 && workspace[10] == 0) {
-            num = Q00 * (DC07 - 3 * DC08 + DC09 - DC17 + 3 * DC18 - DC19);
-            if (num >= 0) {
-              pred = (int)(((Q12 << 7) + num) / (Q12 << 8));
-              if (Al > 0 && pred >= (1 << Al))
-                pred = (1 << Al) - 1;
-            } else {
-              pred = (int)(((Q12 << 7) - num) / (Q12 << 8));
-              if (Al > 0 && pred >= (1 << Al))
-                pred = (1 << Al) - 1;
-              pred = -pred;
-            }
-            workspace[10] = (JCOEF)pred;
-          }
-          /* AC21 */
-          if ((Al = coef_bits[8]) != 0 && workspace[17] == 0) {
-            num = Q00 * (DC07 - DC09 - 3 * DC12 + 3 * DC14 + DC17 - DC19);
-            if (num >= 0) {
-              pred = (int)(((Q21 << 7) + num) / (Q21 << 8));
-              if (Al > 0 && pred >= (1 << Al))
-                pred = (1 << Al) - 1;
-            } else {
-              pred = (int)(((Q21 << 7) - num) / (Q21 << 8));
-              if (Al > 0 && pred >= (1 << Al))
-                pred = (1 << Al) - 1;
-              pred = -pred;
-            }
-            workspace[17] = (JCOEF)pred;
-          }
-          /* AC30 */
-          if ((Al = coef_bits[9]) != 0 && workspace[24] == 0) {
-            num = Q00 * (DC07 + 2 * DC08 + DC09 - DC17 - 2 * DC18 - DC19);
-            if (num >= 0) {
-              pred = (int)(((Q30 << 7) + num) / (Q30 << 8));
-              if (Al > 0 && pred >= (1 << Al))
-                pred = (1 << Al) - 1;
-            } else {
-              pred = (int)(((Q30 << 7) - num) / (Q30 << 8));
-              if (Al > 0 && pred >= (1 << Al))
-                pred = (1 << Al) - 1;
-              pred = -pred;
-            }
-            workspace[24] = (JCOEF)pred;
-          }
-          /* coef_bits[0] is non-negative.  Otherwise this function would not
-           * be called.
-           */
-          num = Q00 *
-                (-2 * DC01 - 6 * DC02 - 8 * DC03 - 6 * DC04 - 2 * DC05 -
-                 6 * DC06 + 6 * DC07 + 42 * DC08 + 6 * DC09 - 6 * DC10 -
-                 8 * DC11 + 42 * DC12 + 152 * DC13 + 42 * DC14 - 8 * DC15 -
-                 6 * DC16 + 6 * DC17 + 42 * DC18 + 6 * DC19 - 6 * DC20 -
-                 2 * DC21 - 6 * DC22 - 8 * DC23 - 6 * DC24 - 2 * DC25);
-          if (num >= 0) {
-            pred = (int)(((Q00 << 7) + num) / (Q00 << 8));
-          } else {
-            pred = (int)(((Q00 << 7) - num) / (Q00 << 8));
-            pred = -pred;
-          }
-          workspace[0] = (JCOEF)pred;
-        }  /* change_dc */
-
        /* OK, do the IDCT */
        (*inverse_DCT) (cinfo, compptr, (JCOEFPTR)workspace, output_ptr,
                        output_col);
        /* Advance for next column */
-        DC01 = DC02;  DC02 = DC03;  DC03 = DC04;  DC04 = DC05;
-        DC06 = DC07;  DC07 = DC08;  DC08 = DC09;  DC09 = DC10;
-        DC11 = DC12;  DC12 = DC13;  DC13 = DC14;  DC14 = DC15;
-        DC16 = DC17;  DC17 = DC18;  DC18 = DC19;  DC19 = DC20;
-        DC21 = DC22;  DC22 = DC23;  DC23 = DC24;  DC24 = DC25;
-        buffer_ptr++, prev_block_row++, next_block_row++,
-          prev_prev_block_row++, next_next_block_row++;
+        DC1 = DC2;  DC2 = DC3;
+        DC4 = DC5;  DC5 = DC6;
+        DC7 = DC8;  DC8 = DC9;
+        buffer_ptr++, prev_block_row++, next_block_row++;
        output_col += compptr->_DCT_scaled_size;
      }
      output_ptr += compptr->_DCT_scaled_size;
@@ -839,7 +653,7 @@ jinit_d_coef_controller(j_decompress_ptr cinfo, boolean need_full_buffer)
 #ifdef BLOCK_SMOOTHING_SUPPORTED
      /* If block smoothing could be used, need a bigger window */
      if (cinfo->progressive_mode)
-        access_rows *= 5;
+        access_rows *= 3;
 #endif
      coef->whole_image[ci] = (*cinfo->mem->request_virt_barray)
        ((j_common_ptr)cinfo, JPOOL_IMAGE, TRUE,
@@ -5,7 +5,6 @@
 * Copyright (C) 1994-1997, Thomas G. Lane.
 * libjpeg-turbo Modifications:
 * Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
- * Copyright (C) 2020, Google, Inc.
 * For conditions of distribution and use, see the accompanying README.ijg
 * file.
 */
@@ -52,7 +51,7 @@ typedef struct {
 #ifdef BLOCK_SMOOTHING_SUPPORTED
  /* When doing block smoothing, we latch coefficient Al values here */
  int *coef_bits_latch;
-#define SAVED_COEFS  10         /* we save coef_bits[0..9] */
+#define SAVED_COEFS  6          /* we save coef_bits[0..5] */
 #endif
 } my_coef_controller;

@@ -45,9 +45,9 @@ ycc_rgb565_convert_internal(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
    outptr = *output_buf++;

    if (PACK_NEED_ALIGNMENT(outptr)) {
-      y  = *inptr0++;
-      cb = *inptr1++;
-      cr = *inptr2++;
+      y  = GETJSAMPLE(*inptr0++);
+      cb = GETJSAMPLE(*inptr1++);
+      cr = GETJSAMPLE(*inptr2++);
      r = range_limit[y + Crrtab[cr]];
      g = range_limit[y + ((int)RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr],
                                            SCALEBITS))];
@@ -58,18 +58,18 @@ ycc_rgb565_convert_internal(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
      num_cols--;
    }
    for (col = 0; col < (num_cols >> 1); col++) {
-      y  = *inptr0++;
-      cb = *inptr1++;
-      cr = *inptr2++;
+      y  = GETJSAMPLE(*inptr0++);
+      cb = GETJSAMPLE(*inptr1++);
+      cr = GETJSAMPLE(*inptr2++);
      r = range_limit[y + Crrtab[cr]];
      g = range_limit[y + ((int)RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr],
                                            SCALEBITS))];
      b = range_limit[y + Cbbtab[cb]];
      rgb = PACK_SHORT_565(r, g, b);

-      y  = *inptr0++;
-      cb = *inptr1++;
-      cr = *inptr2++;
+      y  = GETJSAMPLE(*inptr0++);
+      cb = GETJSAMPLE(*inptr1++);
+      cr = GETJSAMPLE(*inptr2++);
      r = range_limit[y + Crrtab[cr]];
      g = range_limit[y + ((int)RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr],
                                            SCALEBITS))];
@@ -80,9 +80,9 @@ ycc_rgb565_convert_internal(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
      outptr += 4;
    }
    if (num_cols & 1) {
-      y  = *inptr0;
-      cb = *inptr1;
-      cr = *inptr2;
+      y  = GETJSAMPLE(*inptr0);
+      cb = GETJSAMPLE(*inptr1);
+      cr = GETJSAMPLE(*inptr2);
      r = range_limit[y + Crrtab[cr]];
      g = range_limit[y + ((int)RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr],
                                            SCALEBITS))];
@@ -125,9 +125,9 @@ ycc_rgb565D_convert_internal(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
    input_row++;
    outptr = *output_buf++;
    if (PACK_NEED_ALIGNMENT(outptr)) {
-      y  = *inptr0++;
-      cb = *inptr1++;
-      cr = *inptr2++;
+      y  = GETJSAMPLE(*inptr0++);
+      cb = GETJSAMPLE(*inptr1++);
+      cr = GETJSAMPLE(*inptr2++);
      r = range_limit[DITHER_565_R(y + Crrtab[cr], d0)];
      g = range_limit[DITHER_565_G(y +
                                   ((int)RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr],
@@ -139,9 +139,9 @@ ycc_rgb565D_convert_internal(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
      num_cols--;
    }
    for (col = 0; col < (num_cols >> 1); col++) {
-      y  = *inptr0++;
-      cb = *inptr1++;
-      cr = *inptr2++;
+      y  = GETJSAMPLE(*inptr0++);
+      cb = GETJSAMPLE(*inptr1++);
+      cr = GETJSAMPLE(*inptr2++);
      r = range_limit[DITHER_565_R(y + Crrtab[cr], d0)];
      g = range_limit[DITHER_565_G(y +
                                   ((int)RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr],
@@ -150,9 +150,9 @@ ycc_rgb565D_convert_internal(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
      d0 = DITHER_ROTATE(d0);
      rgb = PACK_SHORT_565(r, g, b);

-      y  = *inptr0++;
-      cb = *inptr1++;
-      cr = *inptr2++;
+      y  = GETJSAMPLE(*inptr0++);
+      cb = GETJSAMPLE(*inptr1++);
+      cr = GETJSAMPLE(*inptr2++);
      r = range_limit[DITHER_565_R(y + Crrtab[cr], d0)];
      g = range_limit[DITHER_565_G(y +
                                   ((int)RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr],
@@ -165,9 +165,9 @@ ycc_rgb565D_convert_internal(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
      outptr += 4;
    }
    if (num_cols & 1) {
-      y  = *inptr0;
-      cb = *inptr1;
-      cr = *inptr2;
+      y  = GETJSAMPLE(*inptr0);
+      cb = GETJSAMPLE(*inptr1);
+      cr = GETJSAMPLE(*inptr2);
      r = range_limit[DITHER_565_R(y + Crrtab[cr], d0)];
      g = range_limit[DITHER_565_G(y +
                                   ((int)RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr],
@@ -202,32 +202,32 @@ rgb_rgb565_convert_internal(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
    input_row++;
    outptr = *output_buf++;
    if (PACK_NEED_ALIGNMENT(outptr)) {
-      r = *inptr0++;
-      g = *inptr1++;
-      b = *inptr2++;
+      r = GETJSAMPLE(*inptr0++);
+      g = GETJSAMPLE(*inptr1++);
+      b = GETJSAMPLE(*inptr2++);
      rgb = PACK_SHORT_565(r, g, b);
      *(INT16 *)outptr = (INT16)rgb;
      outptr += 2;
      num_cols--;
    }
    for (col = 0; col < (num_cols >> 1); col++) {
-      r = *inptr0++;
-      g = *inptr1++;
-      b = *inptr2++;
+      r = GETJSAMPLE(*inptr0++);
+      g = GETJSAMPLE(*inptr1++);
+      b = GETJSAMPLE(*inptr2++);
      rgb = PACK_SHORT_565(r, g, b);

-      r = *inptr0++;
-      g = *inptr1++;
-      b = *inptr2++;
+      r = GETJSAMPLE(*inptr0++);
+      g = GETJSAMPLE(*inptr1++);
+      b = GETJSAMPLE(*inptr2++);
      rgb = PACK_TWO_PIXELS(rgb, PACK_SHORT_565(r, g, b));

      WRITE_TWO_ALIGNED_PIXELS(outptr, rgb);
      outptr += 4;
    }
    if (num_cols & 1) {
-      r = *inptr0;
-      g = *inptr1;
-      b = *inptr2;
+      r = GETJSAMPLE(*inptr0);
+      g = GETJSAMPLE(*inptr1);
+      b = GETJSAMPLE(*inptr2);
      rgb = PACK_SHORT_565(r, g, b);
      *(INT16 *)outptr = (INT16)rgb;
    }
@@ -259,24 +259,24 @@ rgb_rgb565D_convert_internal(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
    input_row++;
    outptr = *output_buf++;
    if (PACK_NEED_ALIGNMENT(outptr)) {
-      r = range_limit[DITHER_565_R(*inptr0++, d0)];
-      g = range_limit[DITHER_565_G(*inptr1++, d0)];
-      b = range_limit[DITHER_565_B(*inptr2++, d0)];
+      r = range_limit[DITHER_565_R(GETJSAMPLE(*inptr0++), d0)];
+      g = range_limit[DITHER_565_G(GETJSAMPLE(*inptr1++), d0)];
+      b = range_limit[DITHER_565_B(GETJSAMPLE(*inptr2++), d0)];
      rgb = PACK_SHORT_565(r, g, b);
      *(INT16 *)outptr = (INT16)rgb;
      outptr += 2;
      num_cols--;
    }
    for (col = 0; col < (num_cols >> 1); col++) {
-      r = range_limit[DITHER_565_R(*inptr0++, d0)];
-      g = range_limit[DITHER_565_G(*inptr1++, d0)];
-      b = range_limit[DITHER_565_B(*inptr2++, d0)];
+      r = range_limit[DITHER_565_R(GETJSAMPLE(*inptr0++), d0)];
+      g = range_limit[DITHER_565_G(GETJSAMPLE(*inptr1++), d0)];
+      b = range_limit[DITHER_565_B(GETJSAMPLE(*inptr2++), d0)];
      d0 = DITHER_ROTATE(d0);
      rgb = PACK_SHORT_565(r, g, b);

-      r = range_limit[DITHER_565_R(*inptr0++, d0)];
-      g = range_limit[DITHER_565_G(*inptr1++, d0)];
-      b = range_limit[DITHER_565_B(*inptr2++, d0)];
+      r = range_limit[DITHER_565_R(GETJSAMPLE(*inptr0++), d0)];
+      g = range_limit[DITHER_565_G(GETJSAMPLE(*inptr1++), d0)];
+      b = range_limit[DITHER_565_B(GETJSAMPLE(*inptr2++), d0)];
      d0 = DITHER_ROTATE(d0);
      rgb = PACK_TWO_PIXELS(rgb, PACK_SHORT_565(r, g, b));

@@ -284,9 +284,9 @@ rgb_rgb565D_convert_internal(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
      outptr += 4;
    }
    if (num_cols & 1) {
-      r = range_limit[DITHER_565_R(*inptr0, d0)];
-      g = range_limit[DITHER_565_G(*inptr1, d0)];
-      b = range_limit[DITHER_565_B(*inptr2, d0)];
+      r = range_limit[DITHER_565_R(GETJSAMPLE(*inptr0), d0)];
+      g = range_limit[DITHER_565_G(GETJSAMPLE(*inptr1), d0)];
+      b = range_limit[DITHER_565_B(GETJSAMPLE(*inptr2), d0)];
      rgb = PACK_SHORT_565(r, g, b);
      *(INT16 *)outptr = (INT16)rgb;
    }
@@ -53,9 +53,9 @@ ycc_rgb_convert_internal(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
    input_row++;
    outptr = *output_buf++;
    for (col = 0; col < num_cols; col++) {
-      y  = inptr0[col];
-      cb = inptr1[col];
-      cr = inptr2[col];
+      y  = GETJSAMPLE(inptr0[col]);
+      cb = GETJSAMPLE(inptr1[col]);
+      cr = GETJSAMPLE(inptr2[col]);
      /* Range-limiting is essential due to noise introduced by DCT losses. */
      outptr[RGB_RED] =   range_limit[y + Crrtab[cr]];
      outptr[RGB_GREEN] = range_limit[y +
@@ -93,6 +93,7 @@ gray_rgb_convert_internal(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
    inptr = input_buf[0][input_row++];
    outptr = *output_buf++;
    for (col = 0; col < num_cols; col++) {
+      /* We can dispense with GETJSAMPLE() here */
      outptr[RGB_RED] = outptr[RGB_GREEN] = outptr[RGB_BLUE] = inptr[col];
      /* Set unused byte to 0xFF so it can be interpreted as an opaque */
      /* alpha channel value */
@@ -127,6 +128,7 @@ rgb_rgb_convert_internal(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
    input_row++;
    outptr = *output_buf++;
    for (col = 0; col < num_cols; col++) {
+      /* We can dispense with GETJSAMPLE() here */
      outptr[RGB_RED] = inptr0[col];
      outptr[RGB_GREEN] = inptr1[col];
      outptr[RGB_BLUE] = inptr2[col];
@@ -341,9 +341,9 @@ rgb_gray_convert(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
    input_row++;
    outptr = *output_buf++;
    for (col = 0; col < num_cols; col++) {
-      r = inptr0[col];
-      g = inptr1[col];
-      b = inptr2[col];
+      r = GETJSAMPLE(inptr0[col]);
+      g = GETJSAMPLE(inptr1[col]);
+      b = GETJSAMPLE(inptr2[col]);
      /* Y */
      outptr[col] = (JSAMPLE)((ctab[r + R_Y_OFF] + ctab[g + G_Y_OFF] +
                               ctab[b + B_Y_OFF]) >> SCALEBITS);
@@ -550,9 +550,9 @@ ycck_cmyk_convert(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
    input_row++;
    outptr = *output_buf++;
    for (col = 0; col < num_cols; col++) {
-      y  = inptr0[col];
-      cb = inptr1[col];
-      cr = inptr2[col];
+      y  = GETJSAMPLE(inptr0[col]);
+      cb = GETJSAMPLE(inptr1[col]);
+      cr = GETJSAMPLE(inptr2[col]);
      /* Range-limiting is essential due to noise introduced by DCT losses. */
      outptr[0] = range_limit[MAXJSAMPLE - (y + Crrtab[cr])];   /* red */
      outptr[1] = range_limit[MAXJSAMPLE - (y +                 /* green */
@@ -560,7 +560,7 @@ ycck_cmyk_convert(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
                                                 SCALEBITS)))];
      outptr[2] = range_limit[MAXJSAMPLE - (y + Cbbtab[cb])];   /* blue */
      /* K passes through unchanged */
-      outptr[3] = inptr3[col];
+      outptr[3] = inptr3[col];  /* don't need GETJSAMPLE here */
      outptr += 4;
    }
  }
@@ -571,10 +571,11 @@ ycck_cmyk_convert(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
 * RGB565 conversion
 */

-#define PACK_SHORT_565_LE(r, g, b) \
-  ((((r) << 8) & 0xF800) | (((g) << 3) & 0x7E0) | ((b) >> 3))
-#define PACK_SHORT_565_BE(r, g, b) \
-  (((r) & 0xF8) | ((g) >> 5) | (((g) << 11) & 0xE000) | (((b) << 5) & 0x1F00))
+#define PACK_SHORT_565_LE(r, g, b)  ((((r) << 8) & 0xF800) | \
+                                     (((g) << 3) & 0x7E0) | ((b) >> 3))
+#define PACK_SHORT_565_BE(r, g, b)  (((r) & 0xF8) | ((g) >> 5) | \
+                                     (((g) << 11) & 0xE000) | \
+                                     (((b) << 5) & 0x1F00))

 #define PACK_TWO_PIXELS_LE(l, r)    ((r << 16) | l)
 #define PACK_TWO_PIXELS_BE(l, r)    ((l << 16) | r)
@@ -5,7 +5,6 @@
 * Copyright (C) 1991-1997, Thomas G. Lane.
 * libjpeg-turbo Modifications:
 * Copyright (C) 2009-2011, 2016, 2018-2019, D. R. Commander.
- * Copyright (C) 2018, Matthias Räncker.
 * For conditions of distribution and use, see the accompanying README.ijg
 * file.
 *
@@ -40,6 +39,24 @@ typedef struct {
  int last_dc_val[MAX_COMPS_IN_SCAN]; /* last DC coef for each component */
 } savable_state;

+/* This macro is to work around compilers with missing or broken
+ * structure assignment.  You'll need to fix this code if you have
+ * such a compiler and you change MAX_COMPS_IN_SCAN.
+ */
+
+#ifndef NO_STRUCT_ASSIGN
+#define ASSIGN_STATE(dest, src)  ((dest) = (src))
+#else
+#if MAX_COMPS_IN_SCAN == 4
+#define ASSIGN_STATE(dest, src) \
+  ((dest).last_dc_val[0] = (src).last_dc_val[0], \
+   (dest).last_dc_val[1] = (src).last_dc_val[1], \
+   (dest).last_dc_val[2] = (src).last_dc_val[2], \
+   (dest).last_dc_val[3] = (src).last_dc_val[3])
+#endif
+#endif
+
+
 typedef struct {
  struct jpeg_entropy_decoder pub; /* public fields */

@@ -308,7 +325,7 @@ jpeg_fill_bit_buffer(bitread_working_state *state,
        bytes_in_buffer = cinfo->src->bytes_in_buffer;
      }
      bytes_in_buffer--;
-      c = *next_input_byte++;
+      c = GETJOCTET(*next_input_byte++);

      /* If it's 0xFF, check and discard stuffed zero byte */
      if (c == 0xFF) {
@@ -325,7 +342,7 @@ jpeg_fill_bit_buffer(bitread_working_state *state,
            bytes_in_buffer = cinfo->src->bytes_in_buffer;
          }
          bytes_in_buffer--;
-          c = *next_input_byte++;
+          c = GETJOCTET(*next_input_byte++);
        } while (c == 0xFF);

        if (c == 0) {
@@ -388,8 +405,8 @@ no_more_bytes:

 #define GET_BYTE { \
  register int c0, c1; \
-  c0 = *buffer++; \
-  c1 = *buffer; \
+  c0 = GETJOCTET(*buffer++); \
+  c1 = GETJOCTET(*buffer); \
  /* Pre-execute most common case */ \
  get_buffer = (get_buffer << 8) | c0; \
  bits_left += 8; \
@@ -406,7 +423,7 @@ no_more_bytes:
  } \
 }

-#if SIZEOF_SIZE_T == 8 || defined(_WIN64) || (defined(__x86_64__) && defined(__ILP32__))
+#if SIZEOF_SIZE_T == 8 || defined(_WIN64)

 /* Pre-fetch 48 bytes, because the holding register is 64-bit */
 #define FILL_BIT_BUFFER_FAST \
@@ -540,12 +557,6 @@ process_restart(j_decompress_ptr cinfo)
 }


-#if defined(__has_feature)
-#if __has_feature(undefined_behavior_sanitizer)
-__attribute__((no_sanitize("signed-integer-overflow"),
-               no_sanitize("unsigned-integer-overflow")))
-#endif
-#endif
 LOCAL(boolean)
 decode_mcu_slow(j_decompress_ptr cinfo, JBLOCKROW *MCU_data)
 {
@@ -557,7 +568,7 @@ decode_mcu_slow(j_decompress_ptr cinfo, JBLOCKROW *MCU_data)

  /* Load up working state */
  BITREAD_LOAD_STATE(cinfo, entropy->bitstate);
-  state = entropy->saved;
+  ASSIGN_STATE(state, entropy->saved);

  for (blkn = 0; blkn < cinfo->blocks_in_MCU; blkn++) {
    JBLOCKROW block = MCU_data ? MCU_data[blkn] : NULL;
@@ -578,15 +589,11 @@ decode_mcu_slow(j_decompress_ptr cinfo, JBLOCKROW *MCU_data)
    if (entropy->dc_needed[blkn]) {
      /* Convert DC difference to actual value, update last_dc_val */
      int ci = cinfo->MCU_membership[blkn];
-      /* Certain malformed JPEG images produce repeated DC coefficient
-       * differences of 2047 or -2047, which causes state.last_dc_val[ci] to
-       * grow until it overflows or underflows a 32-bit signed integer.  This
-       * behavior is, to the best of our understanding, innocuous, and it is
-       * unclear how to work around it without potentially affecting
-       * performance.  Thus, we (hopefully temporarily) suppress UBSan integer
-       * overflow errors for this function and decode_mcu_fast().
+      /* This is really just
+       *   s += state.last_dc_val[ci];
+       * It is written this way in order to shut up UBSan.
       */
-      s += state.last_dc_val[ci];
+      s = (int)((unsigned int)s + (unsigned int)state.last_dc_val[ci]);
      state.last_dc_val[ci] = s;
      if (block) {
        /* Output the DC coefficient (assumes jpeg_natural_order[0] = 0) */
@@ -646,17 +653,11 @@ decode_mcu_slow(j_decompress_ptr cinfo, JBLOCKROW *MCU_data)

  /* Completed MCU, so update state */
  BITREAD_SAVE_STATE(cinfo, entropy->bitstate);
-  entropy->saved = state;
+  ASSIGN_STATE(entropy->saved, state);
  return TRUE;
 }


-#if defined(__has_feature)
-#if __has_feature(undefined_behavior_sanitizer)
-__attribute__((no_sanitize("signed-integer-overflow"),
-               no_sanitize("unsigned-integer-overflow")))
-#endif
-#endif
 LOCAL(boolean)
 decode_mcu_fast(j_decompress_ptr cinfo, JBLOCKROW *MCU_data)
 {
@@ -670,7 +671,7 @@ decode_mcu_fast(j_decompress_ptr cinfo, JBLOCKROW *MCU_data)
  /* Load up working state */
  BITREAD_LOAD_STATE(cinfo, entropy->bitstate);
  buffer = (JOCTET *)br_state.next_input_byte;
-  state = entropy->saved;
+  ASSIGN_STATE(state, entropy->saved);

  for (blkn = 0; blkn < cinfo->blocks_in_MCU; blkn++) {
    JBLOCKROW block = MCU_data ? MCU_data[blkn] : NULL;
@@ -687,10 +688,7 @@ decode_mcu_fast(j_decompress_ptr cinfo, JBLOCKROW *MCU_data)

    if (entropy->dc_needed[blkn]) {
      int ci = cinfo->MCU_membership[blkn];
-      /* Refer to the comment in decode_mcu_slow() regarding the supression of
-       * a UBSan integer overflow error in this line of code.
-       */
-      s += state.last_dc_val[ci];
+      s = (int)((unsigned int)s + (unsigned int)state.last_dc_val[ci]);
      state.last_dc_val[ci] = s;
      if (block)
        (*block)[0] = (JCOEF)s;
@@ -742,7 +740,7 @@ decode_mcu_fast(j_decompress_ptr cinfo, JBLOCKROW *MCU_data)
  br_state.bytes_in_buffer -= (buffer - br_state.next_input_byte);
  br_state.next_input_byte = buffer;
  BITREAD_SAVE_STATE(cinfo, entropy->bitstate);
-  entropy->saved = state;
+  ASSIGN_STATE(entropy->saved, state);
  return TRUE;
 }

@@ -797,8 +795,7 @@ use_slow:
  }

  /* Account for restart interval (no-op if not using restarts) */
-  if (cinfo->restart_interval)
-    entropy->restarts_to_go--;
+  entropy->restarts_to_go--;

  return TRUE;
 }
@@ -4,8 +4,7 @@
 * This file was part of the Independent JPEG Group's software:
 * Copyright (C) 1991-1997, Thomas G. Lane.
 * libjpeg-turbo Modifications:
- * Copyright (C) 2010-2011, 2015-2016, 2021, D. R. Commander.
- * Copyright (C) 2018, Matthias Räncker.
+ * Copyright (C) 2010-2011, 2015-2016, D. R. Commander.
 * For conditions of distribution and use, see the accompanying README.ijg
 * file.
 *
@@ -79,11 +78,6 @@ EXTERN(void) jpeg_make_d_derived_tbl(j_decompress_ptr cinfo, boolean isDC,
 typedef size_t bit_buf_type;            /* type of bit-extraction buffer */
 #define BIT_BUF_SIZE  64                /* size of buffer in bits */

-#elif defined(__x86_64__) && defined(__ILP32__)
-
-typedef unsigned long long bit_buf_type; /* type of bit-extraction buffer */
-#define BIT_BUF_SIZE  64                 /* size of buffer in bits */
-
 #else

 typedef unsigned long bit_buf_type;     /* type of bit-extraction buffer */
@@ -234,10 +228,7 @@ slowlabel: \
      s |= GET_BITS(1); \
      nb++; \
    } \
-    if (nb > 16) \
-      s = 0; \
-    else \
-      s = htbl->pub->huffval[(int)(s + htbl->valoffset[nb]) & 0xFF]; \
+    s = htbl->pub->huffval[(int)(s + htbl->valoffset[nb]) & 0xFF]; \
  }

 /* Out-of-line case for Huffman code fetching */
@@ -38,18 +38,18 @@ marker_is_icc(jpeg_saved_marker_ptr marker)
    marker->marker == ICC_MARKER &&
    marker->data_length >= ICC_OVERHEAD_LEN &&
    /* verify the identifying string */
-    marker->data[0] == 0x49 &&
-    marker->data[1] == 0x43 &&
-    marker->data[2] == 0x43 &&
-    marker->data[3] == 0x5F &&
-    marker->data[4] == 0x50 &&
-    marker->data[5] == 0x52 &&
-    marker->data[6] == 0x4F &&
-    marker->data[7] == 0x46 &&
-    marker->data[8] == 0x49 &&
-    marker->data[9] == 0x4C &&
-    marker->data[10] == 0x45 &&
-    marker->data[11] == 0x0;
+    GETJOCTET(marker->data[0]) == 0x49 &&
+    GETJOCTET(marker->data[1]) == 0x43 &&
+    GETJOCTET(marker->data[2]) == 0x43 &&
+    GETJOCTET(marker->data[3]) == 0x5F &&
+    GETJOCTET(marker->data[4]) == 0x50 &&
+    GETJOCTET(marker->data[5]) == 0x52 &&
+    GETJOCTET(marker->data[6]) == 0x4F &&
+    GETJOCTET(marker->data[7]) == 0x46 &&
+    GETJOCTET(marker->data[8]) == 0x49 &&
+    GETJOCTET(marker->data[9]) == 0x4C &&
+    GETJOCTET(marker->data[10]) == 0x45 &&
+    GETJOCTET(marker->data[11]) == 0x0;
 }


@@ -102,12 +102,12 @@ jpeg_read_icc_profile(j_decompress_ptr cinfo, JOCTET **icc_data_ptr,
  for (marker = cinfo->marker_list; marker != NULL; marker = marker->next) {
    if (marker_is_icc(marker)) {
      if (num_markers == 0)
-        num_markers = marker->data[13];
-      else if (num_markers != marker->data[13]) {
+        num_markers = GETJOCTET(marker->data[13]);
+      else if (num_markers != GETJOCTET(marker->data[13])) {
        WARNMS(cinfo, JWRN_BOGUS_ICC);  /* inconsistent num_markers fields */
        return FALSE;
      }
-      seq_no = marker->data[12];
+      seq_no = GETJOCTET(marker->data[12]);
      if (seq_no <= 0 || seq_no > num_markers) {
        WARNMS(cinfo, JWRN_BOGUS_ICC);  /* bogus sequence number */
        return FALSE;
@@ -154,7 +154,7 @@ jpeg_read_icc_profile(j_decompress_ptr cinfo, JOCTET **icc_data_ptr,
      JOCTET FAR *src_ptr;
      JOCTET *dst_ptr;
      unsigned int length;
-      seq_no = marker->data[12];
+      seq_no = GETJOCTET(marker->data[12]);
      dst_ptr = icc_data + data_offset[seq_no];
      src_ptr = marker->data + ICC_OVERHEAD_LEN;
      length = data_length[seq_no];
@@ -18,7 +18,6 @@

 #include "jinclude.h"
 #include "jdmainct.h"
-#include "jconfigint.h"


 /*
@@ -361,7 +360,7 @@ process_data_context_main(j_decompress_ptr cinfo, JSAMPARRAY output_buf,
    main_ptr->context_state = CTX_PREPARE_FOR_IMCU;
    if (*out_row_ctr >= out_rows_avail)
      return;                   /* Postprocessor exactly filled output buf */
-    FALLTHROUGH                 /*FALLTHROUGH*/
+    /*FALLTHROUGH*/
  case CTX_PREPARE_FOR_IMCU:
    /* Prepare to process first M-1 row groups of this iMCU row */
    main_ptr->rowgroup_ctr = 0;
@@ -372,7 +371,7 @@ process_data_context_main(j_decompress_ptr cinfo, JSAMPARRAY output_buf,
    if (main_ptr->iMCU_row_ctr == cinfo->total_iMCU_rows)
      set_bottom_pointers(cinfo);
    main_ptr->context_state = CTX_PROCESS_IMCU;
-    FALLTHROUGH                 /*FALLTHROUGH*/
+    /*FALLTHROUGH*/
  case CTX_PROCESS_IMCU:
    /* Call postprocessor using previously set pointers */
    (*cinfo->post->post_process_data) (cinfo,
@@ -151,7 +151,7 @@ typedef my_marker_reader *my_marker_ptr;
 #define INPUT_BYTE(cinfo, V, action) \
  MAKESTMT( MAKE_BYTE_AVAIL(cinfo, action); \
            bytes_in_buffer--; \
-            V = *next_input_byte++; )
+            V = GETJOCTET(*next_input_byte++); )

 /* As above, but read two bytes interpreted as an unsigned 16-bit integer.
 * V should be declared unsigned int or perhaps JLONG.
@@ -159,10 +159,10 @@ typedef my_marker_reader *my_marker_ptr;
 #define INPUT_2BYTES(cinfo, V, action) \
  MAKESTMT( MAKE_BYTE_AVAIL(cinfo, action); \
            bytes_in_buffer--; \
-            V = ((unsigned int)(*next_input_byte++)) << 8; \
+            V = ((unsigned int)GETJOCTET(*next_input_byte++)) << 8; \
            MAKE_BYTE_AVAIL(cinfo, action); \
            bytes_in_buffer--; \
-            V += *next_input_byte++; )
+            V += GETJOCTET(*next_input_byte++); )


 /*
@@ -608,18 +608,18 @@ examine_app0(j_decompress_ptr cinfo, JOCTET *data, unsigned int datalen,
  JLONG totallen = (JLONG)datalen + remaining;

  if (datalen >= APP0_DATA_LEN &&
-      data[0] == 0x4A &&
-      data[1] == 0x46 &&
-      data[2] == 0x49 &&
-      data[3] == 0x46 &&
-      data[4] == 0) {
+      GETJOCTET(data[0]) == 0x4A &&
+      GETJOCTET(data[1]) == 0x46 &&
+      GETJOCTET(data[2]) == 0x49 &&
+      GETJOCTET(data[3]) == 0x46 &&
+      GETJOCTET(data[4]) == 0) {
    /* Found JFIF APP0 marker: save info */
    cinfo->saw_JFIF_marker = TRUE;
-    cinfo->JFIF_major_version = data[5];
-    cinfo->JFIF_minor_version = data[6];
-    cinfo->density_unit = data[7];
-    cinfo->X_density = (data[8] << 8) + data[9];
-    cinfo->Y_density = (data[10] << 8) + data[11];
+    cinfo->JFIF_major_version = GETJOCTET(data[5]);
+    cinfo->JFIF_minor_version = GETJOCTET(data[6]);
+    cinfo->density_unit = GETJOCTET(data[7]);
+    cinfo->X_density = (GETJOCTET(data[8]) << 8) + GETJOCTET(data[9]);
+    cinfo->Y_density = (GETJOCTET(data[10]) << 8) + GETJOCTET(data[11]);
    /* Check version.
     * Major version must be 1, anything else signals an incompatible change.
     * (We used to treat this as an error, but now it's a nonfatal warning,
@@ -634,22 +634,24 @@ examine_app0(j_decompress_ptr cinfo, JOCTET *data, unsigned int datalen,
             cinfo->JFIF_major_version, cinfo->JFIF_minor_version,
             cinfo->X_density, cinfo->Y_density, cinfo->density_unit);
    /* Validate thumbnail dimensions and issue appropriate messages */
-    if (data[12] | data[13])
-      TRACEMS2(cinfo, 1, JTRC_JFIF_THUMBNAIL, data[12], data[13]);
+    if (GETJOCTET(data[12]) | GETJOCTET(data[13]))
+      TRACEMS2(cinfo, 1, JTRC_JFIF_THUMBNAIL,
+               GETJOCTET(data[12]), GETJOCTET(data[13]));
    totallen -= APP0_DATA_LEN;
-    if (totallen != ((JLONG)data[12] * (JLONG)data[13] * (JLONG)3))
+    if (totallen !=
+        ((JLONG)GETJOCTET(data[12]) * (JLONG)GETJOCTET(data[13]) * (JLONG)3))
      TRACEMS1(cinfo, 1, JTRC_JFIF_BADTHUMBNAILSIZE, (int)totallen);
  } else if (datalen >= 6 &&
-             data[0] == 0x4A &&
-             data[1] == 0x46 &&
-             data[2] == 0x58 &&
-             data[3] == 0x58 &&
-             data[4] == 0) {
+             GETJOCTET(data[0]) == 0x4A &&
+             GETJOCTET(data[1]) == 0x46 &&
+             GETJOCTET(data[2]) == 0x58 &&
+             GETJOCTET(data[3]) == 0x58 &&
+             GETJOCTET(data[4]) == 0) {
    /* Found JFIF "JFXX" extension APP0 marker */
    /* The library doesn't actually do anything with these,
     * but we try to produce a helpful trace message.
     */
-    switch (data[5]) {
+    switch (GETJOCTET(data[5])) {
    case 0x10:
      TRACEMS1(cinfo, 1, JTRC_THUMB_JPEG, (int)totallen);
      break;
@@ -660,7 +662,8 @@ examine_app0(j_decompress_ptr cinfo, JOCTET *data, unsigned int datalen,
      TRACEMS1(cinfo, 1, JTRC_THUMB_RGB, (int)totallen);
      break;
    default:
-      TRACEMS2(cinfo, 1, JTRC_JFIF_EXTENSION, data[5], (int)totallen);
+      TRACEMS2(cinfo, 1, JTRC_JFIF_EXTENSION,
+               GETJOCTET(data[5]), (int)totallen);
      break;
    }
  } else {
@@ -681,16 +684,16 @@ examine_app14(j_decompress_ptr cinfo, JOCTET *data, unsigned int datalen,
  unsigned int version, flags0, flags1, transform;

  if (datalen >= APP14_DATA_LEN &&
-      data[0] == 0x41 &&
-      data[1] == 0x64 &&
-      data[2] == 0x6F &&
-      data[3] == 0x62 &&
-      data[4] == 0x65) {
+      GETJOCTET(data[0]) == 0x41 &&
+      GETJOCTET(data[1]) == 0x64 &&
+      GETJOCTET(data[2]) == 0x6F &&
+      GETJOCTET(data[3]) == 0x62 &&
+      GETJOCTET(data[4]) == 0x65) {
    /* Found Adobe APP14 marker */
-    version = (data[5] << 8) + data[6];
-    flags0 = (data[7] << 8) + data[8];
-    flags1 = (data[9] << 8) + data[10];
-    transform = data[11];
+    version = (GETJOCTET(data[5]) << 8) + GETJOCTET(data[6]);
+    flags0 = (GETJOCTET(data[7]) << 8) + GETJOCTET(data[8]);
+    flags1 = (GETJOCTET(data[9]) << 8) + GETJOCTET(data[10]);
+    transform = GETJOCTET(data[11]);
    TRACEMS4(cinfo, 1, JTRC_ADOBE, version, flags0, flags1, transform);
    cinfo->saw_Adobe_marker = TRUE;
    cinfo->Adobe_transform = (UINT8)transform;
@@ -5,7 +5,7 @@
 * Copyright (C) 1991-1997, Thomas G. Lane.
 * Modified 2002-2009 by Guido Vollbeding.
 * libjpeg-turbo Modifications:
- * Copyright (C) 2009-2011, 2016, 2019, D. R. Commander.
+ * Copyright (C) 2009-2011, 2016, D. R. Commander.
 * Copyright (C) 2013, Linaro Limited.
 * Copyright (C) 2015, Google, Inc.
 * For conditions of distribution and use, see the accompanying README.ijg
@@ -22,6 +22,7 @@
 #include "jpeglib.h"
 #include "jpegcomp.h"
 #include "jdmaster.h"
+#include "jsimd.h"


 /*
@@ -69,6 +70,17 @@ use_merged_upsample(j_decompress_ptr cinfo)
      cinfo->comp_info[1]._DCT_scaled_size != cinfo->_min_DCT_scaled_size ||
      cinfo->comp_info[2]._DCT_scaled_size != cinfo->_min_DCT_scaled_size)
    return FALSE;
+#ifdef WITH_SIMD
+  /* If YCbCr-to-RGB color conversion is SIMD-accelerated but merged upsampling
+     isn't, then disabling merged upsampling is likely to be faster when
+     decompressing YCbCr JPEG images. */
+  if (!jsimd_can_h2v2_merged_upsample() && !jsimd_can_h2v1_merged_upsample() &&
+      jsimd_can_ycc_rgb() && cinfo->jpeg_color_space == JCS_YCbCr &&
+      (cinfo->out_color_space == JCS_RGB ||
+       (cinfo->out_color_space >= JCS_EXT_RGB &&
+        cinfo->out_color_space <= JCS_EXT_ARGB)))
+    return FALSE;
+#endif
  /* ??? also need to test for upsample-time rescaling, when & if supported */
  return TRUE;                  /* by golly, it'll work... */
 #else
@@ -568,7 +580,6 @@ master_selection(j_decompress_ptr cinfo)
   */
  cinfo->master->first_iMCU_col = 0;
  cinfo->master->last_iMCU_col = cinfo->MCUs_per_row - 1;
-  cinfo->master->last_good_iMCU_row = 0;

 #ifdef D_MULTISCAN_FILES_SUPPORTED
  /* If jpeg_start_decompress will read the whole file, initialize
@@ -5,7 +5,7 @@
 * Copyright (C) 1994-1996, Thomas G. Lane.
 * libjpeg-turbo Modifications:
 * Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
- * Copyright (C) 2009, 2011, 2014-2015, 2020, D. R. Commander.
+ * Copyright (C) 2009, 2011, 2014-2015, D. R. Commander.
 * Copyright (C) 2013, Linaro Limited.
 * For conditions of distribution and use, see the accompanying README.ijg
 * file.
@@ -40,13 +40,41 @@
 #define JPEG_INTERNALS
 #include "jinclude.h"
 #include "jpeglib.h"
-#include "jdmerge.h"
 #include "jsimd.h"
 #include "jconfigint.h"

 #ifdef UPSAMPLE_MERGING_SUPPORTED


+/* Private subobject */
+
+typedef struct {
+  struct jpeg_upsampler pub;    /* public fields */
+
+  /* Pointer to routine to do actual upsampling/conversion of one row group */
+  void (*upmethod) (j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
+                    JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf);
+
+  /* Private state for YCC->RGB conversion */
+  int *Cr_r_tab;                /* => table for Cr to R conversion */
+  int *Cb_b_tab;                /* => table for Cb to B conversion */
+  JLONG *Cr_g_tab;              /* => table for Cr to G conversion */
+  JLONG *Cb_g_tab;              /* => table for Cb to G conversion */
+
+  /* For 2:1 vertical sampling, we produce two output rows at a time.
+   * We need a "spare" row buffer to hold the second output row if the
+   * application provides just a one-row buffer; we also use the spare
+   * to discard the dummy last row if the image height is odd.
+   */
+  JSAMPROW spare_row;
+  boolean spare_full;           /* T if spare buffer is occupied */
+
+  JDIMENSION out_row_width;     /* samples per output row */
+  JDIMENSION rows_to_go;        /* counts rows remaining in image */
+} my_upsampler;
+
+typedef my_upsampler *my_upsample_ptr;
+
 #define SCALEBITS       16      /* speediest right-shift on some machines */
 #define ONE_HALF        ((JLONG)1 << (SCALEBITS - 1))
 #define FIX(x)          ((JLONG)((x) * (1L << SCALEBITS) + 0.5))
@@ -161,7 +189,7 @@
 LOCAL(void)
 build_ycc_rgb_table(j_decompress_ptr cinfo)
 {
-  my_merged_upsample_ptr upsample = (my_merged_upsample_ptr)cinfo->upsample;
+  my_upsample_ptr upsample = (my_upsample_ptr)cinfo->upsample;
  int i;
  JLONG x;
  SHIFT_TEMPS
@@ -204,7 +232,7 @@ build_ycc_rgb_table(j_decompress_ptr cinfo)
 METHODDEF(void)
 start_pass_merged_upsample(j_decompress_ptr cinfo)
 {
-  my_merged_upsample_ptr upsample = (my_merged_upsample_ptr)cinfo->upsample;
+  my_upsample_ptr upsample = (my_upsample_ptr)cinfo->upsample;

  /* Mark the spare buffer empty */
  upsample->spare_full = FALSE;
@@ -226,7 +254,7 @@ merged_2v_upsample(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
                   JDIMENSION *out_row_ctr, JDIMENSION out_rows_avail)
 /* 2:1 vertical sampling case: may need a spare row. */
 {
-  my_merged_upsample_ptr upsample = (my_merged_upsample_ptr)cinfo->upsample;
+  my_upsample_ptr upsample = (my_upsample_ptr)cinfo->upsample;
  JSAMPROW work_ptrs[2];
  JDIMENSION num_rows;          /* number of rows returned to caller */

@@ -277,7 +305,7 @@ merged_1v_upsample(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
                   JDIMENSION *out_row_ctr, JDIMENSION out_rows_avail)
 /* 1:1 vertical sampling case: much easier, never need a spare row. */
 {
-  my_merged_upsample_ptr upsample = (my_merged_upsample_ptr)cinfo->upsample;
+  my_upsample_ptr upsample = (my_upsample_ptr)cinfo->upsample;

  /* Just do the upsampling. */
  (*upsample->upmethod) (cinfo, input_buf, *in_row_group_ctr,
@@ -392,10 +420,11 @@ h2v2_merged_upsample(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
 * RGB565 conversion
 */

-#define PACK_SHORT_565_LE(r, g, b) \
-  ((((r) << 8) & 0xF800) | (((g) << 3) & 0x7E0) | ((b) >> 3))
-#define PACK_SHORT_565_BE(r, g, b) \
-  (((r) & 0xF8) | ((g) >> 5) | (((g) << 11) & 0xE000) | (((b) << 5) & 0x1F00))
+#define PACK_SHORT_565_LE(r, g, b)  ((((r) << 8) & 0xF800) | \
+                                     (((g) << 3) & 0x7E0) | ((b) >> 3))
+#define PACK_SHORT_565_BE(r, g, b)  (((r) & 0xF8) | ((g) >> 5) | \
+                                     (((g) << 11) & 0xE000) | \
+                                     (((b) << 5) & 0x1F00))

 #define PACK_TWO_PIXELS_LE(l, r)    ((r << 16) | l)
 #define PACK_TWO_PIXELS_BE(l, r)    ((l << 16) | r)
@@ -537,11 +566,11 @@ h2v2_merged_upsample_565D(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
 GLOBAL(void)
 jinit_merged_upsampler(j_decompress_ptr cinfo)
 {
-  my_merged_upsample_ptr upsample;
+  my_upsample_ptr upsample;

-  upsample = (my_merged_upsample_ptr)
+  upsample = (my_upsample_ptr)
    (*cinfo->mem->alloc_small) ((j_common_ptr)cinfo, JPOOL_IMAGE,
-                                sizeof(my_merged_upsampler));
+                                sizeof(my_upsampler));
  cinfo->upsample = (struct jpeg_upsampler *)upsample;
  upsample->pub.start_pass = start_pass_merged_upsample;
  upsample->pub.need_context_rows = FALSE;
@@ -1,47 +0,0 @@
-/*
- * jdmerge.h
- *
- * This file was part of the Independent JPEG Group's software:
- * Copyright (C) 1994-1996, Thomas G. Lane.
- * libjpeg-turbo Modifications:
- * Copyright (C) 2020, D. R. Commander.
- * For conditions of distribution and use, see the accompanying README.ijg
- * file.
- */
-
-#define JPEG_INTERNALS
-#include "jpeglib.h"
-
-#ifdef UPSAMPLE_MERGING_SUPPORTED
-
-
-/* Private subobject */
-
-typedef struct {
-  struct jpeg_upsampler pub;    /* public fields */
-
-  /* Pointer to routine to do actual upsampling/conversion of one row group */
-  void (*upmethod) (j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
-                    JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf);
-
-  /* Private state for YCC->RGB conversion */
-  int *Cr_r_tab;                /* => table for Cr to R conversion */
-  int *Cb_b_tab;                /* => table for Cb to B conversion */
-  JLONG *Cr_g_tab;              /* => table for Cr to G conversion */
-  JLONG *Cb_g_tab;              /* => table for Cb to G conversion */
-
-  /* For 2:1 vertical sampling, we produce two output rows at a time.
-   * We need a "spare" row buffer to hold the second output row if the
-   * application provides just a one-row buffer; we also use the spare
-   * to discard the dummy last row if the image height is odd.
-   */
-  JSAMPROW spare_row;
-  boolean spare_full;           /* T if spare buffer is occupied */
-
-  JDIMENSION out_row_width;     /* samples per output row */
-  JDIMENSION rows_to_go;        /* counts rows remaining in image */
-} my_merged_upsampler;
-
-typedef my_merged_upsampler *my_merged_upsample_ptr;
-
-#endif /* UPSAMPLE_MERGING_SUPPORTED */
@@ -5,7 +5,7 @@
 * Copyright (C) 1994-1996, Thomas G. Lane.
 * libjpeg-turbo Modifications:
 * Copyright (C) 2013, Linaro Limited.
- * Copyright (C) 2014-2015, 2018, 2020, D. R. Commander.
+ * Copyright (C) 2014-2015, 2018, D. R. Commander.
 * For conditions of distribution and use, see the accompanying README.ijg
 * file.
 *
@@ -19,7 +19,7 @@ h2v1_merged_upsample_565_internal(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
                                  JDIMENSION in_row_group_ctr,
                                  JSAMPARRAY output_buf)
 {
-  my_merged_upsample_ptr upsample = (my_merged_upsample_ptr)cinfo->upsample;
+  my_upsample_ptr upsample = (my_upsample_ptr)cinfo->upsample;
  register int y, cred, cgreen, cblue;
  int cb, cr;
  register JSAMPROW outptr;
@@ -43,20 +43,20 @@ h2v1_merged_upsample_565_internal(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
  /* Loop for each pair of output pixels */
  for (col = cinfo->output_width >> 1; col > 0; col--) {
    /* Do the chroma part of the calculation */
-    cb = *inptr1++;
-    cr = *inptr2++;
+    cb = GETJSAMPLE(*inptr1++);
+    cr = GETJSAMPLE(*inptr2++);
    cred = Crrtab[cr];
    cgreen = (int)RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr], SCALEBITS);
    cblue = Cbbtab[cb];

    /* Fetch 2 Y values and emit 2 pixels */
-    y  = *inptr0++;
+    y  = GETJSAMPLE(*inptr0++);
    r = range_limit[y + cred];
    g = range_limit[y + cgreen];
    b = range_limit[y + cblue];
    rgb = PACK_SHORT_565(r, g, b);

-    y  = *inptr0++;
+    y  = GETJSAMPLE(*inptr0++);
    r = range_limit[y + cred];
    g = range_limit[y + cgreen];
    b = range_limit[y + cblue];
@@ -68,12 +68,12 @@ h2v1_merged_upsample_565_internal(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,

  /* If image width is odd, do the last output column separately */
  if (cinfo->output_width & 1) {
-    cb = *inptr1;
-    cr = *inptr2;
+    cb = GETJSAMPLE(*inptr1);
+    cr = GETJSAMPLE(*inptr2);
    cred = Crrtab[cr];
    cgreen = (int)RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr], SCALEBITS);
    cblue = Cbbtab[cb];
-    y  = *inptr0;
+    y  = GETJSAMPLE(*inptr0);
    r = range_limit[y + cred];
    g = range_limit[y + cgreen];
    b = range_limit[y + cblue];
@@ -90,7 +90,7 @@ h2v1_merged_upsample_565D_internal(j_decompress_ptr cinfo,
                                   JDIMENSION in_row_group_ctr,
                                   JSAMPARRAY output_buf)
 {
-  my_merged_upsample_ptr upsample = (my_merged_upsample_ptr)cinfo->upsample;
+  my_upsample_ptr upsample = (my_upsample_ptr)cinfo->upsample;
  register int y, cred, cgreen, cblue;
  int cb, cr;
  register JSAMPROW outptr;
@@ -115,21 +115,21 @@ h2v1_merged_upsample_565D_internal(j_decompress_ptr cinfo,
  /* Loop for each pair of output pixels */
  for (col = cinfo->output_width >> 1; col > 0; col--) {
    /* Do the chroma part of the calculation */
-    cb = *inptr1++;
-    cr = *inptr2++;
+    cb = GETJSAMPLE(*inptr1++);
+    cr = GETJSAMPLE(*inptr2++);
    cred = Crrtab[cr];
    cgreen = (int)RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr], SCALEBITS);
    cblue = Cbbtab[cb];

    /* Fetch 2 Y values and emit 2 pixels */
-    y  = *inptr0++;
+    y  = GETJSAMPLE(*inptr0++);
    r = range_limit[DITHER_565_R(y + cred, d0)];
    g = range_limit[DITHER_565_G(y + cgreen, d0)];
    b = range_limit[DITHER_565_B(y + cblue, d0)];
    d0 = DITHER_ROTATE(d0);
    rgb = PACK_SHORT_565(r, g, b);

-    y  = *inptr0++;
+    y  = GETJSAMPLE(*inptr0++);
    r = range_limit[DITHER_565_R(y + cred, d0)];
    g = range_limit[DITHER_565_G(y + cgreen, d0)];
    b = range_limit[DITHER_565_B(y + cblue, d0)];
@@ -142,12 +142,12 @@ h2v1_merged_upsample_565D_internal(j_decompress_ptr cinfo,

  /* If image width is odd, do the last output column separately */
  if (cinfo->output_width & 1) {
-    cb = *inptr1;
-    cr = *inptr2;
+    cb = GETJSAMPLE(*inptr1);
+    cr = GETJSAMPLE(*inptr2);
    cred = Crrtab[cr];
    cgreen = (int)RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr], SCALEBITS);
    cblue = Cbbtab[cb];
-    y  = *inptr0;
+    y  = GETJSAMPLE(*inptr0);
    r = range_limit[DITHER_565_R(y + cred, d0)];
    g = range_limit[DITHER_565_G(y + cgreen, d0)];
    b = range_limit[DITHER_565_B(y + cblue, d0)];
@@ -163,7 +163,7 @@ h2v2_merged_upsample_565_internal(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
                                  JDIMENSION in_row_group_ctr,
                                  JSAMPARRAY output_buf)
 {
-  my_merged_upsample_ptr upsample = (my_merged_upsample_ptr)cinfo->upsample;
+  my_upsample_ptr upsample = (my_upsample_ptr)cinfo->upsample;
  register int y, cred, cgreen, cblue;
  int cb, cr;
  register JSAMPROW outptr0, outptr1;
@@ -189,20 +189,20 @@ h2v2_merged_upsample_565_internal(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
  /* Loop for each group of output pixels */
  for (col = cinfo->output_width >> 1; col > 0; col--) {
    /* Do the chroma part of the calculation */
-    cb = *inptr1++;
-    cr = *inptr2++;
+    cb = GETJSAMPLE(*inptr1++);
+    cr = GETJSAMPLE(*inptr2++);
    cred = Crrtab[cr];
    cgreen = (int)RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr], SCALEBITS);
    cblue = Cbbtab[cb];

    /* Fetch 4 Y values and emit 4 pixels */
-    y  = *inptr00++;
+    y  = GETJSAMPLE(*inptr00++);
    r = range_limit[y + cred];
    g = range_limit[y + cgreen];
    b = range_limit[y + cblue];
    rgb = PACK_SHORT_565(r, g, b);

-    y  = *inptr00++;
+    y  = GETJSAMPLE(*inptr00++);
    r = range_limit[y + cred];
    g = range_limit[y + cgreen];
    b = range_limit[y + cblue];
@@ -211,13 +211,13 @@ h2v2_merged_upsample_565_internal(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
    WRITE_TWO_PIXELS(outptr0, rgb);
    outptr0 += 4;

-    y  = *inptr01++;
+    y  = GETJSAMPLE(*inptr01++);
    r = range_limit[y + cred];
    g = range_limit[y + cgreen];
    b = range_limit[y + cblue];
    rgb = PACK_SHORT_565(r, g, b);

-    y  = *inptr01++;
+    y  = GETJSAMPLE(*inptr01++);
    r = range_limit[y + cred];
    g = range_limit[y + cgreen];
    b = range_limit[y + cblue];
@@ -229,20 +229,20 @@ h2v2_merged_upsample_565_internal(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,

  /* If image width is odd, do the last output column separately */
  if (cinfo->output_width & 1) {
-    cb = *inptr1;
-    cr = *inptr2;
+    cb = GETJSAMPLE(*inptr1);
+    cr = GETJSAMPLE(*inptr2);
    cred = Crrtab[cr];
    cgreen = (int)RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr], SCALEBITS);
    cblue = Cbbtab[cb];

-    y  = *inptr00;
+    y  = GETJSAMPLE(*inptr00);
    r = range_limit[y + cred];
    g = range_limit[y + cgreen];
    b = range_limit[y + cblue];
    rgb = PACK_SHORT_565(r, g, b);
    *(INT16 *)outptr0 = (INT16)rgb;

-    y  = *inptr01;
+    y  = GETJSAMPLE(*inptr01);
    r = range_limit[y + cred];
    g = range_limit[y + cgreen];
    b = range_limit[y + cblue];
@@ -259,7 +259,7 @@ h2v2_merged_upsample_565D_internal(j_decompress_ptr cinfo,
                                   JDIMENSION in_row_group_ctr,
                                   JSAMPARRAY output_buf)
 {
-  my_merged_upsample_ptr upsample = (my_merged_upsample_ptr)cinfo->upsample;
+  my_upsample_ptr upsample = (my_upsample_ptr)cinfo->upsample;
  register int y, cred, cgreen, cblue;
  int cb, cr;
  register JSAMPROW outptr0, outptr1;
@@ -287,21 +287,21 @@ h2v2_merged_upsample_565D_internal(j_decompress_ptr cinfo,
  /* Loop for each group of output pixels */
  for (col = cinfo->output_width >> 1; col > 0; col--) {
    /* Do the chroma part of the calculation */
-    cb = *inptr1++;
-    cr = *inptr2++;
+    cb = GETJSAMPLE(*inptr1++);
+    cr = GETJSAMPLE(*inptr2++);
    cred = Crrtab[cr];
    cgreen = (int)RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr], SCALEBITS);
    cblue = Cbbtab[cb];

    /* Fetch 4 Y values and emit 4 pixels */
-    y  = *inptr00++;
+    y  = GETJSAMPLE(*inptr00++);
    r = range_limit[DITHER_565_R(y + cred, d0)];
    g = range_limit[DITHER_565_G(y + cgreen, d0)];
    b = range_limit[DITHER_565_B(y + cblue, d0)];
    d0 = DITHER_ROTATE(d0);
    rgb = PACK_SHORT_565(r, g, b);

-    y  = *inptr00++;
+    y  = GETJSAMPLE(*inptr00++);
    r = range_limit[DITHER_565_R(y + cred, d0)];
    g = range_limit[DITHER_565_G(y + cgreen, d0)];
    b = range_limit[DITHER_565_B(y + cblue, d0)];
@@ -311,14 +311,14 @@ h2v2_merged_upsample_565D_internal(j_decompress_ptr cinfo,
    WRITE_TWO_PIXELS(outptr0, rgb);
    outptr0 += 4;

-    y  = *inptr01++;
+    y  = GETJSAMPLE(*inptr01++);
    r = range_limit[DITHER_565_R(y + cred, d1)];
    g = range_limit[DITHER_565_G(y + cgreen, d1)];
    b = range_limit[DITHER_565_B(y + cblue, d1)];
    d1 = DITHER_ROTATE(d1);
    rgb = PACK_SHORT_565(r, g, b);

-    y  = *inptr01++;
+    y  = GETJSAMPLE(*inptr01++);
    r = range_limit[DITHER_565_R(y + cred, d1)];
    g = range_limit[DITHER_565_G(y + cgreen, d1)];
    b = range_limit[DITHER_565_B(y + cblue, d1)];
@@ -331,20 +331,20 @@ h2v2_merged_upsample_565D_internal(j_decompress_ptr cinfo,

  /* If image width is odd, do the last output column separately */
  if (cinfo->output_width & 1) {
-    cb = *inptr1;
-    cr = *inptr2;
+    cb = GETJSAMPLE(*inptr1);
+    cr = GETJSAMPLE(*inptr2);
    cred = Crrtab[cr];
    cgreen = (int)RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr], SCALEBITS);
    cblue = Cbbtab[cb];

-    y  = *inptr00;
+    y  = GETJSAMPLE(*inptr00);
    r = range_limit[DITHER_565_R(y + cred, d0)];
    g = range_limit[DITHER_565_G(y + cgreen, d0)];
    b = range_limit[DITHER_565_B(y + cblue, d0)];
    rgb = PACK_SHORT_565(r, g, b);
    *(INT16 *)outptr0 = (INT16)rgb;

-    y  = *inptr01;
+    y  = GETJSAMPLE(*inptr01);
    r = range_limit[DITHER_565_R(y + cred, d1)];
    g = range_limit[DITHER_565_G(y + cgreen, d1)];
    b = range_limit[DITHER_565_B(y + cblue, d1)];
@@ -4,7 +4,7 @@
 * This file was part of the Independent JPEG Group's software:
 * Copyright (C) 1994-1996, Thomas G. Lane.
 * libjpeg-turbo Modifications:
- * Copyright (C) 2011, 2015, 2020, D. R. Commander.
+ * Copyright (C) 2011, 2015, D. R. Commander.
 * For conditions of distribution and use, see the accompanying README.ijg
 * file.
 *
@@ -25,7 +25,7 @@ h2v1_merged_upsample_internal(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
                              JDIMENSION in_row_group_ctr,
                              JSAMPARRAY output_buf)
 {
-  my_merged_upsample_ptr upsample = (my_merged_upsample_ptr)cinfo->upsample;
+  my_upsample_ptr upsample = (my_upsample_ptr)cinfo->upsample;
  register int y, cred, cgreen, cblue;
  int cb, cr;
  register JSAMPROW outptr;
@@ -46,13 +46,13 @@ h2v1_merged_upsample_internal(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
  /* Loop for each pair of output pixels */
  for (col = cinfo->output_width >> 1; col > 0; col--) {
    /* Do the chroma part of the calculation */
-    cb = *inptr1++;
-    cr = *inptr2++;
+    cb = GETJSAMPLE(*inptr1++);
+    cr = GETJSAMPLE(*inptr2++);
    cred = Crrtab[cr];
    cgreen = (int)RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr], SCALEBITS);
    cblue = Cbbtab[cb];
    /* Fetch 2 Y values and emit 2 pixels */
-    y  = *inptr0++;
+    y  = GETJSAMPLE(*inptr0++);
    outptr[RGB_RED] =   range_limit[y + cred];
    outptr[RGB_GREEN] = range_limit[y + cgreen];
    outptr[RGB_BLUE] =  range_limit[y + cblue];
@@ -60,7 +60,7 @@ h2v1_merged_upsample_internal(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
    outptr[RGB_ALPHA] = 0xFF;
 #endif
    outptr += RGB_PIXELSIZE;
-    y  = *inptr0++;
+    y  = GETJSAMPLE(*inptr0++);
    outptr[RGB_RED] =   range_limit[y + cred];
    outptr[RGB_GREEN] = range_limit[y + cgreen];
    outptr[RGB_BLUE] =  range_limit[y + cblue];
@@ -71,12 +71,12 @@ h2v1_merged_upsample_internal(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
  }
  /* If image width is odd, do the last output column separately */
  if (cinfo->output_width & 1) {
-    cb = *inptr1;
-    cr = *inptr2;
+    cb = GETJSAMPLE(*inptr1);
+    cr = GETJSAMPLE(*inptr2);
    cred = Crrtab[cr];
    cgreen = (int)RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr], SCALEBITS);
    cblue = Cbbtab[cb];
-    y  = *inptr0;
+    y  = GETJSAMPLE(*inptr0);
    outptr[RGB_RED] =   range_limit[y + cred];
    outptr[RGB_GREEN] = range_limit[y + cgreen];
    outptr[RGB_BLUE] =  range_limit[y + cblue];
@@ -97,7 +97,7 @@ h2v2_merged_upsample_internal(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
                              JDIMENSION in_row_group_ctr,
                              JSAMPARRAY output_buf)
 {
-  my_merged_upsample_ptr upsample = (my_merged_upsample_ptr)cinfo->upsample;
+  my_upsample_ptr upsample = (my_upsample_ptr)cinfo->upsample;
  register int y, cred, cgreen, cblue;
  int cb, cr;
  register JSAMPROW outptr0, outptr1;
@@ -120,13 +120,13 @@ h2v2_merged_upsample_internal(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
  /* Loop for each group of output pixels */
  for (col = cinfo->output_width >> 1; col > 0; col--) {
    /* Do the chroma part of the calculation */
-    cb = *inptr1++;
-    cr = *inptr2++;
+    cb = GETJSAMPLE(*inptr1++);
+    cr = GETJSAMPLE(*inptr2++);
    cred = Crrtab[cr];
    cgreen = (int)RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr], SCALEBITS);
    cblue = Cbbtab[cb];
    /* Fetch 4 Y values and emit 4 pixels */
-    y  = *inptr00++;
+    y  = GETJSAMPLE(*inptr00++);
    outptr0[RGB_RED] =   range_limit[y + cred];
    outptr0[RGB_GREEN] = range_limit[y + cgreen];
    outptr0[RGB_BLUE] =  range_limit[y + cblue];
@@ -134,7 +134,7 @@ h2v2_merged_upsample_internal(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
    outptr0[RGB_ALPHA] = 0xFF;
 #endif
    outptr0 += RGB_PIXELSIZE;
-    y  = *inptr00++;
+    y  = GETJSAMPLE(*inptr00++);
    outptr0[RGB_RED] =   range_limit[y + cred];
    outptr0[RGB_GREEN] = range_limit[y + cgreen];
    outptr0[RGB_BLUE] =  range_limit[y + cblue];
@@ -142,7 +142,7 @@ h2v2_merged_upsample_internal(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
    outptr0[RGB_ALPHA] = 0xFF;
 #endif
    outptr0 += RGB_PIXELSIZE;
-    y  = *inptr01++;
+    y  = GETJSAMPLE(*inptr01++);
    outptr1[RGB_RED] =   range_limit[y + cred];
    outptr1[RGB_GREEN] = range_limit[y + cgreen];
    outptr1[RGB_BLUE] =  range_limit[y + cblue];
@@ -150,7 +150,7 @@ h2v2_merged_upsample_internal(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
    outptr1[RGB_ALPHA] = 0xFF;
 #endif
    outptr1 += RGB_PIXELSIZE;
-    y  = *inptr01++;
+    y  = GETJSAMPLE(*inptr01++);
    outptr1[RGB_RED] =   range_limit[y + cred];
    outptr1[RGB_GREEN] = range_limit[y + cgreen];
    outptr1[RGB_BLUE] =  range_limit[y + cblue];
@@ -161,19 +161,19 @@ h2v2_merged_upsample_internal(j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
  }
  /* If image width is odd, do the last output column separately */
  if (cinfo->output_width & 1) {
-    cb = *inptr1;
-    cr = *inptr2;
+    cb = GETJSAMPLE(*inptr1);
+    cr = GETJSAMPLE(*inptr2);
    cred = Crrtab[cr];
    cgreen = (int)RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr], SCALEBITS);
    cblue = Cbbtab[cb];
-    y  = *inptr00;
+    y  = GETJSAMPLE(*inptr00);
    outptr0[RGB_RED] =   range_limit[y + cred];
    outptr0[RGB_GREEN] = range_limit[y + cgreen];
    outptr0[RGB_BLUE] =  range_limit[y + cblue];
 #ifdef RGB_ALPHA
    outptr0[RGB_ALPHA] = 0xFF;
 #endif
-    y  = *inptr01;
+    y  = GETJSAMPLE(*inptr01);
    outptr1[RGB_RED] =   range_limit[y + cred];
    outptr1[RGB_GREEN] = range_limit[y + cgreen];
    outptr1[RGB_BLUE] =  range_limit[y + cblue];
@@ -4,7 +4,7 @@
 * This file was part of the Independent JPEG Group's software:
 * Copyright (C) 1995-1997, Thomas G. Lane.
 * libjpeg-turbo Modifications:
- * Copyright (C) 2015-2016, 2018-2021, D. R. Commander.
+ * Copyright (C) 2015-2016, 2018, D. R. Commander.
 * For conditions of distribution and use, see the accompanying README.ijg
 * file.
 *
@@ -41,6 +41,25 @@ typedef struct {
  int last_dc_val[MAX_COMPS_IN_SCAN];   /* last DC coef for each component */
 } savable_state;

+/* This macro is to work around compilers with missing or broken
+ * structure assignment.  You'll need to fix this code if you have
+ * such a compiler and you change MAX_COMPS_IN_SCAN.
+ */
+
+#ifndef NO_STRUCT_ASSIGN
+#define ASSIGN_STATE(dest, src)  ((dest) = (src))
+#else
+#if MAX_COMPS_IN_SCAN == 4
+#define ASSIGN_STATE(dest, src) \
+  ((dest).EOBRUN = (src).EOBRUN, \
+   (dest).last_dc_val[0] = (src).last_dc_val[0], \
+   (dest).last_dc_val[1] = (src).last_dc_val[1], \
+   (dest).last_dc_val[2] = (src).last_dc_val[2], \
+   (dest).last_dc_val[3] = (src).last_dc_val[3])
+#endif
+#endif
+
+
 typedef struct {
  struct jpeg_entropy_decoder pub; /* public fields */

@@ -83,7 +102,7 @@ start_pass_phuff_decoder(j_decompress_ptr cinfo)
  boolean is_DC_band, bad;
  int ci, coefi, tbl;
  d_derived_tbl **pdtbl;
-  int *coef_bit_ptr, *prev_coef_bit_ptr;
+  int *coef_bit_ptr;
  jpeg_component_info *compptr;

  is_DC_band = (cinfo->Ss == 0);
@@ -124,15 +143,8 @@ start_pass_phuff_decoder(j_decompress_ptr cinfo)
  for (ci = 0; ci < cinfo->comps_in_scan; ci++) {
    int cindex = cinfo->cur_comp_info[ci]->component_index;
    coef_bit_ptr = &cinfo->coef_bits[cindex][0];
-    prev_coef_bit_ptr = &cinfo->coef_bits[cindex + cinfo->num_components][0];
    if (!is_DC_band && coef_bit_ptr[0] < 0) /* AC without prior DC scan */
      WARNMS2(cinfo, JWRN_BOGUS_PROGRESSION, cindex, 0);
-    for (coefi = MIN(cinfo->Ss, 1); coefi <= MAX(cinfo->Se, 9); coefi++) {
-      if (cinfo->input_scan_number > 1)
-        prev_coef_bit_ptr[coefi] = coef_bit_ptr[coefi];
-      else
-        prev_coef_bit_ptr[coefi] = 0;
-    }
    for (coefi = cinfo->Ss; coefi <= cinfo->Se; coefi++) {
      int expected = (coef_bit_ptr[coefi] < 0) ? 0 : coef_bit_ptr[coefi];
      if (cinfo->Ah != expected)
@@ -311,7 +323,7 @@ decode_mcu_DC_first(j_decompress_ptr cinfo, JBLOCKROW *MCU_data)

    /* Load up working state */
    BITREAD_LOAD_STATE(cinfo, entropy->bitstate);
-    state = entropy->saved;
+    ASSIGN_STATE(state, entropy->saved);

    /* Outer loop handles each block in the MCU */

@@ -344,12 +356,11 @@ decode_mcu_DC_first(j_decompress_ptr cinfo, JBLOCKROW *MCU_data)

    /* Completed MCU, so update state */
    BITREAD_SAVE_STATE(cinfo, entropy->bitstate);
-    entropy->saved = state;
+    ASSIGN_STATE(entropy->saved, state);
  }

  /* Account for restart interval (no-op if not using restarts) */
-  if (cinfo->restart_interval)
-    entropy->restarts_to_go--;
+  entropy->restarts_to_go--;

  return TRUE;
 }
@@ -433,8 +444,7 @@ decode_mcu_AC_first(j_decompress_ptr cinfo, JBLOCKROW *MCU_data)
  }

  /* Account for restart interval (no-op if not using restarts) */
-  if (cinfo->restart_interval)
-    entropy->restarts_to_go--;
+  entropy->restarts_to_go--;

  return TRUE;
 }
@@ -485,8 +495,7 @@ decode_mcu_DC_refine(j_decompress_ptr cinfo, JBLOCKROW *MCU_data)
  BITREAD_SAVE_STATE(cinfo, entropy->bitstate);

  /* Account for restart interval (no-op if not using restarts) */
-  if (cinfo->restart_interval)
-    entropy->restarts_to_go--;
+  entropy->restarts_to_go--;

  return TRUE;
 }
@@ -629,8 +638,7 @@ decode_mcu_AC_refine(j_decompress_ptr cinfo, JBLOCKROW *MCU_data)
  }

  /* Account for restart interval (no-op if not using restarts) */
-  if (cinfo->restart_interval)
-    entropy->restarts_to_go--;
+  entropy->restarts_to_go--;

  return TRUE;

@@ -668,7 +676,7 @@ jinit_phuff_decoder(j_decompress_ptr cinfo)
  /* Create progression status table */
  cinfo->coef_bits = (int (*)[DCTSIZE2])
    (*cinfo->mem->alloc_small) ((j_common_ptr)cinfo, JPOOL_IMAGE,
-                                cinfo->num_components * 2 * DCTSIZE2 *
+                                cinfo->num_components * DCTSIZE2 *
                                sizeof(int));
  coef_bit_ptr = &cinfo->coef_bits[0][0];
  for (ci = 0; ci < cinfo->num_components; ci++)
@@ -8,7 +8,7 @@
 * Copyright (C) 2010, 2015-2016, D. R. Commander.
 * Copyright (C) 2014, MIPS Technologies, Inc., California.
 * Copyright (C) 2015, Google, Inc.
- * Copyright (C) 2019-2020, Arm Limited.
+ * Copyright (C) 2019, Arm Limited.
 * For conditions of distribution and use, see the accompanying README.ijg
 * file.
 *
@@ -177,7 +177,7 @@ int_upsample(j_decompress_ptr cinfo, jpeg_component_info *compptr,
    outptr = output_data[outrow];
    outend = outptr + cinfo->output_width;
    while (outptr < outend) {
-      invalue = *inptr++;
+      invalue = *inptr++;       /* don't need GETJSAMPLE() here */
      for (h = h_expand; h > 0; h--) {
        *outptr++ = invalue;
      }
@@ -213,7 +213,7 @@ h2v1_upsample(j_decompress_ptr cinfo, jpeg_component_info *compptr,
    outptr = output_data[inrow];
    outend = outptr + cinfo->output_width;
    while (outptr < outend) {
-      invalue = *inptr++;
+      invalue = *inptr++;       /* don't need GETJSAMPLE() here */
      *outptr++ = invalue;
      *outptr++ = invalue;
    }
@@ -242,7 +242,7 @@ h2v2_upsample(j_decompress_ptr cinfo, jpeg_component_info *compptr,
    outptr = output_data[outrow];
    outend = outptr + cinfo->output_width;
    while (outptr < outend) {
-      invalue = *inptr++;
+      invalue = *inptr++;       /* don't need GETJSAMPLE() here */
      *outptr++ = invalue;
      *outptr++ = invalue;
    }
@@ -283,20 +283,20 @@ h2v1_fancy_upsample(j_decompress_ptr cinfo, jpeg_component_info *compptr,
    inptr = input_data[inrow];
    outptr = output_data[inrow];
    /* Special case for first column */
-    invalue = *inptr++;
+    invalue = GETJSAMPLE(*inptr++);
    *outptr++ = (JSAMPLE)invalue;
-    *outptr++ = (JSAMPLE)((invalue * 3 + inptr[0] + 2) >> 2);
+    *outptr++ = (JSAMPLE)((invalue * 3 + GETJSAMPLE(*inptr) + 2) >> 2);

    for (colctr = compptr->downsampled_width - 2; colctr > 0; colctr--) {
      /* General case: 3/4 * nearer pixel + 1/4 * further pixel */
-      invalue = (*inptr++) * 3;
-      *outptr++ = (JSAMPLE)((invalue + inptr[-2] + 1) >> 2);
-      *outptr++ = (JSAMPLE)((invalue + inptr[0] + 2) >> 2);
+      invalue = GETJSAMPLE(*inptr++) * 3;
+      *outptr++ = (JSAMPLE)((invalue + GETJSAMPLE(inptr[-2]) + 1) >> 2);
+      *outptr++ = (JSAMPLE)((invalue + GETJSAMPLE(*inptr) + 2) >> 2);
    }

    /* Special case for last column */
-    invalue = *inptr;
-    *outptr++ = (JSAMPLE)((invalue * 3 + inptr[-1] + 1) >> 2);
+    invalue = GETJSAMPLE(*inptr);
+    *outptr++ = (JSAMPLE)((invalue * 3 + GETJSAMPLE(inptr[-1]) + 1) >> 2);
    *outptr++ = (JSAMPLE)invalue;
  }
 }
@@ -338,7 +338,7 @@ h1v2_fancy_upsample(j_decompress_ptr cinfo, jpeg_component_info *compptr,
      outptr = output_data[outrow++];

      for (colctr = 0; colctr < compptr->downsampled_width; colctr++) {
-        thiscolsum = (*inptr0++) * 3 + (*inptr1++);
+        thiscolsum = GETJSAMPLE(*inptr0++) * 3 + GETJSAMPLE(*inptr1++);
        *outptr++ = (JSAMPLE)((thiscolsum + bias) >> 2);
      }
    }
@@ -381,8 +381,8 @@ h2v2_fancy_upsample(j_decompress_ptr cinfo, jpeg_component_info *compptr,
      outptr = output_data[outrow++];

      /* Special case for first column */
-      thiscolsum = (*inptr0++) * 3 + (*inptr1++);
-      nextcolsum = (*inptr0++) * 3 + (*inptr1++);
+      thiscolsum = GETJSAMPLE(*inptr0++) * 3 + GETJSAMPLE(*inptr1++);
+      nextcolsum = GETJSAMPLE(*inptr0++) * 3 + GETJSAMPLE(*inptr1++);
      *outptr++ = (JSAMPLE)((thiscolsum * 4 + 8) >> 4);
      *outptr++ = (JSAMPLE)((thiscolsum * 3 + nextcolsum + 7) >> 4);
      lastcolsum = thiscolsum;  thiscolsum = nextcolsum;
@@ -390,7 +390,7 @@ h2v2_fancy_upsample(j_decompress_ptr cinfo, jpeg_component_info *compptr,
      for (colctr = compptr->downsampled_width - 2; colctr > 0; colctr--) {
        /* General case: 3/4 * nearer pixel + 1/4 * further pixel in each */
        /* dimension, thus 9/16, 3/16, 3/16, 1/16 overall */
-        nextcolsum = (*inptr0++) * 3 + (*inptr1++);
+        nextcolsum = GETJSAMPLE(*inptr0++) * 3 + GETJSAMPLE(*inptr1++);
        *outptr++ = (JSAMPLE)((thiscolsum * 3 + lastcolsum + 8) >> 4);
        *outptr++ = (JSAMPLE)((thiscolsum * 3 + nextcolsum + 7) >> 4);
        lastcolsum = thiscolsum;  thiscolsum = nextcolsum;
@@ -477,13 +477,7 @@ jinit_upsampler(j_decompress_ptr cinfo)
    } else if (h_in_group == h_out_group &&
               v_in_group * 2 == v_out_group && do_fancy) {
      /* Non-fancy upsampling is handled by the generic method */
-#if defined(__arm__) || defined(__aarch64__) || \
-    defined(_M_ARM) || defined(_M_ARM64)
-      if (jsimd_can_h1v2_fancy_upsample())
-        upsample->methods[ci] = jsimd_h1v2_fancy_upsample;
-      else
-#endif
-        upsample->methods[ci] = h1v2_fancy_upsample;
+      upsample->methods[ci] = h1v2_fancy_upsample;
      upsample->pub.need_context_rows = TRUE;
    } else if (h_in_group * 2 == h_out_group &&
               v_in_group * 2 == v_out_group) {
@@ -3,8 +3,8 @@
 *
 * This file was part of the Independent JPEG Group's software:
 * Copyright (C) 1995-1997, Thomas G. Lane.
- * libjpeg-turbo Modifications:
- * Copyright (C) 2020, D. R. Commander.
+ * It was modified by The libjpeg-turbo Project to include only code relevant
+ * to libjpeg-turbo.
 * For conditions of distribution and use, see the accompanying README.ijg
 * file.
 *
@@ -16,7 +16,6 @@
 #define JPEG_INTERNALS
 #include "jinclude.h"
 #include "jpeglib.h"
-#include "jpegcomp.h"


 /* Forward declarations */
@@ -207,10 +207,6 @@ JMESSAGE(JWRN_ARITH_BAD_CODE, "Corrupt JPEG data: bad arithmetic code")
 #endif
 #endif
 JMESSAGE(JWRN_BOGUS_ICC, "Corrupt JPEG data: bad ICC marker")
-#if JPEG_LIB_VERSION < 70
-JMESSAGE(JERR_BAD_DROP_SAMPLING,
-         "Component index %d: mismatching sampling ratio %d:%d, %d:%d, %c")
-#endif

 #ifdef JMAKE_ENUM_LIST

@@ -256,15 +252,6 @@ JMESSAGE(JERR_BAD_DROP_SAMPLING,
   (cinfo)->err->msg_parm.i[2] = (p3), \
   (cinfo)->err->msg_parm.i[3] = (p4), \
   (*(cinfo)->err->error_exit) ((j_common_ptr)(cinfo)))
-#define ERREXIT6(cinfo, code, p1, p2, p3, p4, p5, p6) \
-  ((cinfo)->err->msg_code = (code), \
-   (cinfo)->err->msg_parm.i[0] = (p1), \
-   (cinfo)->err->msg_parm.i[1] = (p2), \
-   (cinfo)->err->msg_parm.i[2] = (p3), \
-   (cinfo)->err->msg_parm.i[3] = (p4), \
-   (cinfo)->err->msg_parm.i[4] = (p5), \
-   (cinfo)->err->msg_parm.i[5] = (p6), \
-   (*(cinfo)->err->error_exit) ((j_common_ptr)(cinfo)))
 #define ERREXITS(cinfo, code, str) \
  ((cinfo)->err->msg_code = (code), \
   strncpy((cinfo)->err->msg_parm.s, (str), JMSG_STR_PARM_MAX), \
@@ -4,11 +4,11 @@
 * This file was part of the Independent JPEG Group's software:
 * Copyright (C) 1991-1996, Thomas G. Lane.
 * libjpeg-turbo Modifications:
- * Copyright (C) 2015, 2020, D. R. Commander.
+ * Copyright (C) 2015, D. R. Commander.
 * For conditions of distribution and use, see the accompanying README.ijg
 * file.
 *
- * This file contains a slower but more accurate integer implementation of the
+ * This file contains a slow-but-accurate integer implementation of the
 * forward DCT (Discrete Cosine Transform).
 *
 * A 2-D DCT can be done by 1-D DCT on each row followed by 1-D DCT
@@ -3,13 +3,13 @@
 *
 * This file was part of the Independent JPEG Group's software:
 * Copyright (C) 1991-1998, Thomas G. Lane.
- * Modification developed 2002-2018 by Guido Vollbeding.
+ * Modification developed 2002-2009 by Guido Vollbeding.
 * libjpeg-turbo Modifications:
- * Copyright (C) 2015, 2020, D. R. Commander.
+ * Copyright (C) 2015, D. R. Commander.
 * For conditions of distribution and use, see the accompanying README.ijg
 * file.
 *
- * This file contains a slower but more accurate integer implementation of the
+ * This file contains a slow-but-accurate integer implementation of the
 * inverse DCT (Discrete Cosine Transform).  In the IJG code, this routine
 * must also perform dequantization of the input coefficients.
 *
@@ -417,7 +417,7 @@ jpeg_idct_islow(j_decompress_ptr cinfo, jpeg_component_info *compptr,

 /*
 * Perform dequantization and inverse DCT on one block of coefficients,
- * producing a reduced-size 7x7 output block.
+ * producing a 7x7 output block.
 *
 * Optimized algorithm with 12 multiplications in the 1-D kernel.
 * cK represents sqrt(2) * cos(K*pi/14).
@@ -1258,7 +1258,7 @@ jpeg_idct_10x10(j_decompress_ptr cinfo, jpeg_component_info *compptr,

 /*
 * Perform dequantization and inverse DCT on one block of coefficients,
- * producing an 11x11 output block.
+ * producing a 11x11 output block.
 *
 * Optimized algorithm with 24 multiplications in the 1-D kernel.
 * cK represents sqrt(2) * cos(K*pi/22).
@@ -2398,7 +2398,7 @@ jpeg_idct_16x16(j_decompress_ptr cinfo, jpeg_component_info *compptr,
    tmp0 = DEQUANTIZE(inptr[DCTSIZE * 0], quantptr[DCTSIZE * 0]);
    tmp0 = LEFT_SHIFT(tmp0, CONST_BITS);
    /* Add fudge factor here for final descale. */
-    tmp0 += ONE << (CONST_BITS - PASS1_BITS - 1);
+    tmp0 += 1 << (CONST_BITS - PASS1_BITS - 1);

    z1 = DEQUANTIZE(inptr[DCTSIZE * 4], quantptr[DCTSIZE * 4]);
    tmp1 = MULTIPLY(z1, FIX(1.306562965));      /* c4[16] = c2[8] */
@@ -4,7 +4,7 @@
 * This file was part of the Independent JPEG Group's software:
 * Copyright (C) 1991-1997, Thomas G. Lane.
 * libjpeg-turbo Modifications:
- * Copyright (C) 2016, 2021, D. R. Commander.
+ * Copyright (C) 2016, D. R. Commander.
 * For conditions of distribution and use, see the accompanying README.ijg
 * file.
 *
@@ -1032,7 +1032,7 @@ free_pool(j_common_ptr cinfo, int pool_id)
    large_pool_ptr next_lhdr_ptr = lhdr_ptr->next;
    space_freed = lhdr_ptr->bytes_used +
                  lhdr_ptr->bytes_left +
-                  sizeof(large_pool_hdr) + ALIGN_SIZE - 1;
+                  sizeof(large_pool_hdr);
    jpeg_free_large(cinfo, (void *)lhdr_ptr, space_freed);
    mem->total_space_allocated -= space_freed;
    lhdr_ptr = next_lhdr_ptr;
@@ -1045,7 +1045,7 @@ free_pool(j_common_ptr cinfo, int pool_id)
  while (shdr_ptr != NULL) {
    small_pool_ptr next_shdr_ptr = shdr_ptr->next;
    space_freed = shdr_ptr->bytes_used + shdr_ptr->bytes_left +
-                  sizeof(small_pool_hdr) + ALIGN_SIZE - 1;
+                  sizeof(small_pool_hdr);
    jpeg_free_small(cinfo, (void *)shdr_ptr, space_freed);
    mem->total_space_allocated -= space_freed;
    shdr_ptr = next_shdr_ptr;
@@ -5,7 +5,7 @@
 * Copyright (C) 1991-1997, Thomas G. Lane.
 * Modified 1997-2009 by Guido Vollbeding.
 * libjpeg-turbo Modifications:
- * Copyright (C) 2009, 2011, 2014-2015, 2018, 2020, D. R. Commander.
+ * Copyright (C) 2009, 2011, 2014-2015, 2018, D. R. Commander.
 * For conditions of distribution and use, see the accompanying README.ijg
 * file.
 *
@@ -43,11 +43,25 @@

 #if BITS_IN_JSAMPLE == 8
 /* JSAMPLE should be the smallest type that will hold the values 0..255.
+ * You can use a signed char by having GETJSAMPLE mask it with 0xFF.
 */

+#ifdef HAVE_UNSIGNED_CHAR
+
 typedef unsigned char JSAMPLE;
 #define GETJSAMPLE(value)  ((int)(value))

+#else /* not HAVE_UNSIGNED_CHAR */
+
+typedef char JSAMPLE;
+#ifdef __CHAR_UNSIGNED__
+#define GETJSAMPLE(value)  ((int)(value))
+#else
+#define GETJSAMPLE(value)  ((int)(value) & 0xFF)
+#endif /* __CHAR_UNSIGNED__ */
+
+#endif /* HAVE_UNSIGNED_CHAR */
+
 #define MAXJSAMPLE      255
 #define CENTERJSAMPLE   128

@@ -83,9 +97,22 @@ typedef short JCOEF;
 * managers, this is also the data type passed to fread/fwrite.
 */

+#ifdef HAVE_UNSIGNED_CHAR
+
 typedef unsigned char JOCTET;
 #define GETJOCTET(value)  (value)

+#else /* not HAVE_UNSIGNED_CHAR */
+
+typedef char JOCTET;
+#ifdef __CHAR_UNSIGNED__
+#define GETJOCTET(value)  (value)
+#else
+#define GETJOCTET(value)  ((value) & 0xFF)
+#endif /* __CHAR_UNSIGNED__ */
+
+#endif /* HAVE_UNSIGNED_CHAR */
+

 /* These typedefs are used for various table entries and so forth.
 * They must be at least as wide as specified; but making them too big
@@ -96,7 +123,15 @@ typedef unsigned char JOCTET;

 /* UINT8 must hold at least the values 0..255. */

+#ifdef HAVE_UNSIGNED_CHAR
 typedef unsigned char UINT8;
+#else /* not HAVE_UNSIGNED_CHAR */
+#ifdef __CHAR_UNSIGNED__
+typedef char UINT8;
+#else /* not __CHAR_UNSIGNED__ */
+typedef short UINT8;
+#endif /* __CHAR_UNSIGNED__ */
+#endif /* HAVE_UNSIGNED_CHAR */

 /* UINT16 must hold at least the values 0..65535. */

@@ -238,9 +273,9 @@ typedef int boolean;

 /* Capability options common to encoder and decoder: */

-#define DCT_ISLOW_SUPPORTED     /* accurate integer method */
-#define DCT_IFAST_SUPPORTED     /* less accurate int method [legacy feature] */
-#define DCT_FLOAT_SUPPORTED     /* floating-point method [legacy feature] */
+#define DCT_ISLOW_SUPPORTED     /* slow but accurate integer algorithm */
+#define DCT_IFAST_SUPPORTED     /* faster, less accurate integer method */
+#define DCT_FLOAT_SUPPORTED     /* floating-point: accurate, fast on fast HW */

 /* Encoder capability options: */

@@ -1,7 +1,7 @@
 /*
 * jpegcomp.h
 *
- * Copyright (C) 2010, 2020, D. R. Commander.
+ * Copyright (C) 2010, D. R. Commander.
 * For conditions of distribution and use, see the accompanying README.ijg
 * file.
 *
@@ -19,7 +19,6 @@
 #define _min_DCT_v_scaled_size  min_DCT_v_scaled_size
 #define _jpeg_width  jpeg_width
 #define _jpeg_height  jpeg_height
-#define JERR_ARITH_NOTIMPL  JERR_NOT_COMPILED
 #else
 #define _DCT_scaled_size  DCT_scaled_size
 #define _DCT_h_scaled_size  DCT_scaled_size
@@ -5,9 +5,8 @@
 * Copyright (C) 1991-1997, Thomas G. Lane.
 * Modified 1997-2009 by Guido Vollbeding.
 * libjpeg-turbo Modifications:
- * Copyright (C) 2015-2016, 2019, 2021, D. R. Commander.
+ * Copyright (C) 2015-2016, D. R. Commander.
 * Copyright (C) 2015, Google, Inc.
- * Copyright (C) 2021, Alex Richardson.
 * For conditions of distribution and use, see the accompanying README.ijg
 * file.
 *
@@ -48,18 +47,6 @@ typedef enum {            /* Operating modes for buffer controllers */
 /* JLONG must hold at least signed 32-bit values. */
 typedef long JLONG;

-/* JUINTPTR must hold pointer values. */
-#ifdef __UINTPTR_TYPE__
-/*
- * __UINTPTR_TYPE__ is GNU-specific and available in GCC 4.6+ and Clang 3.0+.
- * Fortunately, that is sufficient to support the few architectures for which
- * sizeof(void *) != sizeof(size_t).  The only other options would require C99
- * or Clang-specific builtins.
- */
-typedef __UINTPTR_TYPE__ JUINTPTR;
-#else
-typedef size_t JUINTPTR;
-#endif

 /*
 * Left shift macro that handles a negative operand without causing any
@@ -171,9 +158,6 @@ struct jpeg_decomp_master {
  JDIMENSION first_MCU_col[MAX_COMPONENTS];
  JDIMENSION last_MCU_col[MAX_COMPONENTS];
  boolean jinit_upsampler_no_alloc;
-
-  /* Last iMCU row that was successfully decoded */
-  JDIMENSION last_good_iMCU_row;
 };

 /* Input control module */
@@ -5,7 +5,7 @@
 * Copyright (C) 1991-1998, Thomas G. Lane.
 * Modified 2002-2009 by Guido Vollbeding.
 * libjpeg-turbo Modifications:
- * Copyright (C) 2009-2011, 2013-2014, 2016-2017, 2020, D. R. Commander.
+ * Copyright (C) 2009-2011, 2013-2014, 2016-2017, D. R. Commander.
 * Copyright (C) 2015, Google, Inc.
 * For conditions of distribution and use, see the accompanying README.ijg
 * file.
@@ -244,9 +244,9 @@ typedef enum {
 /* DCT/IDCT algorithm options. */

 typedef enum {
-  JDCT_ISLOW,             /* accurate integer method */
-  JDCT_IFAST,             /* less accurate integer method [legacy feature] */
-  JDCT_FLOAT              /* floating-point method [legacy feature] */
+  JDCT_ISLOW,             /* slow but accurate integer algorithm */
+  JDCT_IFAST,             /* faster, less accurate integer method */
+  JDCT_FLOAT              /* floating-point: accurate, fast on fast HW */
 } J_DCT_METHOD;

 #ifndef JDCT_DEFAULT            /* may be overridden in jconfig.h */
@@ -479,7 +479,7 @@ color_quantize(j_decompress_ptr cinfo, JSAMPARRAY input_buf,
    for (col = width; col > 0; col--) {
      pixcode = 0;
      for (ci = 0; ci < nc; ci++) {
-        pixcode += colorindex[ci][*ptrin++];
+        pixcode += GETJSAMPLE(colorindex[ci][GETJSAMPLE(*ptrin++)]);
      }
      *ptrout++ = (JSAMPLE)pixcode;
    }
@@ -506,9 +506,9 @@ color_quantize3(j_decompress_ptr cinfo, JSAMPARRAY input_buf,
    ptrin = input_buf[row];
    ptrout = output_buf[row];
    for (col = width; col > 0; col--) {
-      pixcode  = colorindex0[*ptrin++];
-      pixcode += colorindex1[*ptrin++];
-      pixcode += colorindex2[*ptrin++];
+      pixcode  = GETJSAMPLE(colorindex0[GETJSAMPLE(*ptrin++)]);
+      pixcode += GETJSAMPLE(colorindex1[GETJSAMPLE(*ptrin++)]);
+      pixcode += GETJSAMPLE(colorindex2[GETJSAMPLE(*ptrin++)]);
      *ptrout++ = (JSAMPLE)pixcode;
    }
  }
@@ -552,7 +552,7 @@ quantize_ord_dither(j_decompress_ptr cinfo, JSAMPARRAY input_buf,
         * required amount of padding.
         */
        *output_ptr +=
-          colorindex_ci[*input_ptr + dither[col_index]];
+          colorindex_ci[GETJSAMPLE(*input_ptr) + dither[col_index]];
        input_ptr += nc;
        output_ptr++;
        col_index = (col_index + 1) & ODITHER_MASK;
@@ -595,9 +595,12 @@ quantize3_ord_dither(j_decompress_ptr cinfo, JSAMPARRAY input_buf,
    col_index = 0;

    for (col = width; col > 0; col--) {
-      pixcode  = colorindex0[(*input_ptr++) + dither0[col_index]];
-      pixcode += colorindex1[(*input_ptr++) + dither1[col_index]];
-      pixcode += colorindex2[(*input_ptr++) + dither2[col_index]];
+      pixcode  =
+        GETJSAMPLE(colorindex0[GETJSAMPLE(*input_ptr++) + dither0[col_index]]);
+      pixcode +=
+        GETJSAMPLE(colorindex1[GETJSAMPLE(*input_ptr++) + dither1[col_index]]);
+      pixcode +=
+        GETJSAMPLE(colorindex2[GETJSAMPLE(*input_ptr++) + dither2[col_index]]);
      *output_ptr++ = (JSAMPLE)pixcode;
      col_index = (col_index + 1) & ODITHER_MASK;
    }
@@ -674,15 +677,15 @@ quantize_fs_dither(j_decompress_ptr cinfo, JSAMPARRAY input_buf,
         * The maximum error is +- MAXJSAMPLE; this sets the required size
         * of the range_limit array.
         */
-        cur += *input_ptr;
-        cur = range_limit[cur];
+        cur += GETJSAMPLE(*input_ptr);
+        cur = GETJSAMPLE(range_limit[cur]);
        /* Select output value, accumulate into output code for this pixel */
-        pixcode = colorindex_ci[cur];
+        pixcode = GETJSAMPLE(colorindex_ci[cur]);
        *output_ptr += (JSAMPLE)pixcode;
        /* Compute actual representation error at this pixel */
        /* Note: we can do this even though we don't have the final */
        /* pixel code, because the colormap is orthogonal. */
-        cur -= colormap_ci[pixcode];
+        cur -= GETJSAMPLE(colormap_ci[pixcode]);
        /* Compute error fractions to be propagated to adjacent pixels.
         * Add these into the running sums, and simultaneously shift the
         * next-line error sums left by 1 column.
@@ -4,7 +4,7 @@
 * This file was part of the Independent JPEG Group's software:
 * Copyright (C) 1991-1996, Thomas G. Lane.
 * libjpeg-turbo Modifications:
- * Copyright (C) 2009, 2014-2015, 2020, D. R. Commander.
+ * Copyright (C) 2009, 2014-2015, D. R. Commander.
 * For conditions of distribution and use, see the accompanying README.ijg
 * file.
 *
@@ -215,9 +215,9 @@ prescan_quantize(j_decompress_ptr cinfo, JSAMPARRAY input_buf,
    ptr = input_buf[row];
    for (col = width; col > 0; col--) {
      /* get pixel value and index into the histogram */
-      histp = &histogram[ptr[0] >> C0_SHIFT]
-                        [ptr[1] >> C1_SHIFT]
-                        [ptr[2] >> C2_SHIFT];
+      histp = &histogram[GETJSAMPLE(ptr[0]) >> C0_SHIFT]
+                        [GETJSAMPLE(ptr[1]) >> C1_SHIFT]
+                        [GETJSAMPLE(ptr[2]) >> C2_SHIFT];
      /* increment, check for overflow and undo increment if so. */
      if (++(*histp) <= 0)
        (*histp)--;
@@ -665,7 +665,7 @@ find_nearby_colors(j_decompress_ptr cinfo, int minc0, int minc1, int minc2,

  for (i = 0; i < numcolors; i++) {
    /* We compute the squared-c0-distance term, then add in the other two. */
-    x = cinfo->colormap[0][i];
+    x = GETJSAMPLE(cinfo->colormap[0][i]);
    if (x < minc0) {
      tdist = (x - minc0) * C0_SCALE;
      min_dist = tdist * tdist;
@@ -688,7 +688,7 @@ find_nearby_colors(j_decompress_ptr cinfo, int minc0, int minc1, int minc2,
      }
    }

-    x = cinfo->colormap[1][i];
+    x = GETJSAMPLE(cinfo->colormap[1][i]);
    if (x < minc1) {
      tdist = (x - minc1) * C1_SCALE;
      min_dist += tdist * tdist;
@@ -710,7 +710,7 @@ find_nearby_colors(j_decompress_ptr cinfo, int minc0, int minc1, int minc2,
      }
    }

-    x = cinfo->colormap[2][i];
+    x = GETJSAMPLE(cinfo->colormap[2][i]);
    if (x < minc2) {
      tdist = (x - minc2) * C2_SCALE;
      min_dist += tdist * tdist;
@@ -788,13 +788,13 @@ find_best_colors(j_decompress_ptr cinfo, int minc0, int minc1, int minc2,
 #define STEP_C2  ((1 << C2_SHIFT) * C2_SCALE)

  for (i = 0; i < numcolors; i++) {
-    icolor = colorlist[i];
+    icolor = GETJSAMPLE(colorlist[i]);
    /* Compute (square of) distance from minc0/c1/c2 to this color */
-    inc0 = (minc0 - cinfo->colormap[0][icolor]) * C0_SCALE;
+    inc0 = (minc0 - GETJSAMPLE(cinfo->colormap[0][icolor])) * C0_SCALE;
    dist0 = inc0 * inc0;
-    inc1 = (minc1 - cinfo->colormap[1][icolor]) * C1_SCALE;
+    inc1 = (minc1 - GETJSAMPLE(cinfo->colormap[1][icolor])) * C1_SCALE;
    dist0 += inc1 * inc1;
-    inc2 = (minc2 - cinfo->colormap[2][icolor]) * C2_SCALE;
+    inc2 = (minc2 - GETJSAMPLE(cinfo->colormap[2][icolor])) * C2_SCALE;
    dist0 += inc2 * inc2;
    /* Form the initial difference increments */
    inc0 = inc0 * (2 * STEP_C0) + STEP_C0 * STEP_C0;
@@ -879,7 +879,7 @@ fill_inverse_cmap(j_decompress_ptr cinfo, int c0, int c1, int c2)
    for (ic1 = 0; ic1 < BOX_C1_ELEMS; ic1++) {
      cachep = &histogram[c0 + ic0][c1 + ic1][c2];
      for (ic2 = 0; ic2 < BOX_C2_ELEMS; ic2++) {
-        *cachep++ = (histcell)((*cptr++) + 1);
+        *cachep++ = (histcell)(GETJSAMPLE(*cptr++) + 1);
      }
    }
  }
@@ -909,9 +909,9 @@ pass2_no_dither(j_decompress_ptr cinfo, JSAMPARRAY input_buf,
    outptr = output_buf[row];
    for (col = width; col > 0; col--) {
      /* get pixel value and index into the cache */
-      c0 = (*inptr++) >> C0_SHIFT;
-      c1 = (*inptr++) >> C1_SHIFT;
-      c2 = (*inptr++) >> C2_SHIFT;
+      c0 = GETJSAMPLE(*inptr++) >> C0_SHIFT;
+      c1 = GETJSAMPLE(*inptr++) >> C1_SHIFT;
+      c2 = GETJSAMPLE(*inptr++) >> C2_SHIFT;
      cachep = &histogram[c0][c1][c2];
      /* If we have not seen this color before, find nearest colormap entry */
      /* and update the cache */
@@ -996,12 +996,12 @@ pass2_fs_dither(j_decompress_ptr cinfo, JSAMPARRAY input_buf,
       * The maximum error is +- MAXJSAMPLE (or less with error limiting);
       * this sets the required size of the range_limit array.
       */
-      cur0 += inptr[0];
-      cur1 += inptr[1];
-      cur2 += inptr[2];
-      cur0 = range_limit[cur0];
-      cur1 = range_limit[cur1];
-      cur2 = range_limit[cur2];
+      cur0 += GETJSAMPLE(inptr[0]);
+      cur1 += GETJSAMPLE(inptr[1]);
+      cur2 += GETJSAMPLE(inptr[2]);
+      cur0 = GETJSAMPLE(range_limit[cur0]);
+      cur1 = GETJSAMPLE(range_limit[cur1]);
+      cur2 = GETJSAMPLE(range_limit[cur2]);
      /* Index into the cache with adjusted pixel value */
      cachep =
        &histogram[cur0 >> C0_SHIFT][cur1 >> C1_SHIFT][cur2 >> C2_SHIFT];
@@ -1015,9 +1015,9 @@ pass2_fs_dither(j_decompress_ptr cinfo, JSAMPARRAY input_buf,
        register int pixcode = *cachep - 1;
        *outptr = (JSAMPLE)pixcode;
        /* Compute representation error for this pixel */
-        cur0 -= colormap0[pixcode];
-        cur1 -= colormap1[pixcode];
-        cur2 -= colormap2[pixcode];
+        cur0 -= GETJSAMPLE(colormap0[pixcode]);
+        cur1 -= GETJSAMPLE(colormap1[pixcode]);
+        cur2 -= GETJSAMPLE(colormap2[pixcode]);
      }
      /* Compute error fractions to be propagated to adjacent pixels.
       * Add these into the running sums, and simultaneously shift the
@@ -1145,7 +1145,7 @@ start_pass_2_quant(j_decompress_ptr cinfo, boolean is_pre_scan)
  int i;

  /* Only F-S dithering or no dithering is supported. */
-  /* If user asks for ordered dither, give them F-S. */
+  /* If user asks for ordered dither, give him F-S. */
  if (cinfo->dither_mode != JDITHER_NONE)
    cinfo->dither_mode = JDITHER_FS;

@@ -1263,7 +1263,7 @@ jinit_2pass_quantizer(j_decompress_ptr cinfo)
    cquantize->sv_colormap = NULL;

  /* Only F-S dithering or no dithering is supported. */
-  /* If user asks for ordered dither, give them F-S. */
+  /* If user asks for ordered dither, give him F-S. */
  if (cinfo->dither_mode != JDITHER_NONE)
    cinfo->dither_mode = JDITHER_FS;

@@ -4,7 +4,6 @@
 * Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
 * Copyright (C) 2011, 2014, D. R. Commander.
 * Copyright (C) 2015-2016, 2018, Matthieu Darbois.
- * Copyright (C) 2020, Arm Limited.
 *
 * Based on the x86 SIMD extension for IJG JPEG library,
 * Copyright (C) 1999-2006, MIYASAKA Masaru.
@@ -76,7 +75,6 @@ EXTERN(void) jsimd_int_upsample(j_decompress_ptr cinfo,

 EXTERN(int) jsimd_can_h2v2_fancy_upsample(void);
 EXTERN(int) jsimd_can_h2v1_fancy_upsample(void);
-EXTERN(int) jsimd_can_h1v2_fancy_upsample(void);

 EXTERN(void) jsimd_h2v2_fancy_upsample(j_decompress_ptr cinfo,
                                       jpeg_component_info *compptr,
@@ -86,10 +84,6 @@ EXTERN(void) jsimd_h2v1_fancy_upsample(j_decompress_ptr cinfo,
                                       jpeg_component_info *compptr,
                                       JSAMPARRAY input_data,
                                       JSAMPARRAY *output_data_ptr);
-EXTERN(void) jsimd_h1v2_fancy_upsample(j_decompress_ptr cinfo,
-                                       jpeg_component_info *compptr,
-                                       JSAMPARRAY input_data,
-                                       JSAMPARRAY *output_data_ptr);

 EXTERN(int) jsimd_can_h2v2_merged_upsample(void);
 EXTERN(int) jsimd_can_h2v1_merged_upsample(void);
@@ -4,7 +4,6 @@
 * Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
 * Copyright (C) 2009-2011, 2014, D. R. Commander.
 * Copyright (C) 2015-2016, 2018, Matthieu Darbois.
- * Copyright (C) 2020, Arm Limited.
 *
 * Based on the x86 SIMD extension for IJG JPEG library,
 * Copyright (C) 1999-2006, MIYASAKA Masaru.
@@ -170,12 +169,6 @@ jsimd_can_h2v1_fancy_upsample(void)
  return 0;
 }

-GLOBAL(int)
-jsimd_can_h1v2_fancy_upsample(void)
-{
-  return 0;
-}
-
 GLOBAL(void)
 jsimd_h2v2_fancy_upsample(j_decompress_ptr cinfo, jpeg_component_info *compptr,
                          JSAMPARRAY input_data, JSAMPARRAY *output_data_ptr)
@@ -188,12 +181,6 @@ jsimd_h2v1_fancy_upsample(j_decompress_ptr cinfo, jpeg_component_info *compptr,
 {
 }

-GLOBAL(void)
-jsimd_h1v2_fancy_upsample(j_decompress_ptr cinfo, jpeg_component_info *compptr,
-                          JSAMPARRAY input_data, JSAMPARRAY *output_data_ptr)
-{
-}
-
 GLOBAL(int)
 jsimd_can_h2v2_merged_upsample(void)
 {
@@ -2,9 +2,9 @@
 * jversion.h
 *
 * This file was part of the Independent JPEG Group's software:
- * Copyright (C) 1991-2020, Thomas G. Lane, Guido Vollbeding.
+ * Copyright (C) 1991-2012, Thomas G. Lane, Guido Vollbeding.
 * libjpeg-turbo Modifications:
- * Copyright (C) 2010, 2012-2021, D. R. Commander.
+ * Copyright (C) 2010, 2012-2019, D. R. Commander.
 * For conditions of distribution and use, see the accompanying README.ijg
 * file.
 *
@@ -30,25 +30,23 @@
 * NOTE: It is our convention to place the authors in the following order:
 * - libjpeg-turbo authors (2009-) in descending order of the date of their
 *   most recent contribution to the project, then in ascending order of the
- *   date of their first contribution to the project, then in alphabetical
- *   order
+ *   date of their first contribution to the project
 * - Upstream authors in descending order of the date of the first inclusion of
 *   their code
 */

 #define JCOPYRIGHT \
-  "Copyright (C) 2009-2021 D. R. Commander\n" \
-  "Copyright (C) 2015, 2020 Google, Inc.\n" \
-  "Copyright (C) 2019-2020 Arm Limited\n" \
-  "Copyright (C) 2015-2016, 2018 Matthieu Darbois\n" \
+  "Copyright (C) 2009-2019 D. R. Commander\n" \
  "Copyright (C) 2011-2016 Siarhei Siamashka\n" \
+  "Copyright (C) 2015-2016, 2018 Matthieu Darbois\n" \
  "Copyright (C) 2015 Intel Corporation\n" \
-  "Copyright (C) 2013-2014 Linaro Limited\n" \
+  "Copyright (C) 2015 Google, Inc.\n" \
  "Copyright (C) 2013-2014 MIPS Technologies, Inc.\n" \
-  "Copyright (C) 2009, 2012 Pierre Ossman for Cendio AB\n" \
+  "Copyright (C) 2013 Linaro Limited\n" \
  "Copyright (C) 2009-2011 Nokia Corporation and/or its subsidiary(-ies)\n" \
+  "Copyright (C) 2009 Pierre Ossman for Cendio AB\n" \
  "Copyright (C) 1999-2006 MIYASAKA Masaru\n" \
-  "Copyright (C) 1991-2020 Thomas G. Lane, Guido Vollbeding"
+  "Copyright (C) 1991-2016 Thomas G. Lane, Guido Vollbeding"

 #define JCOPYRIGHT_SHORT \
-  "Copyright (C) 1991-2021 The libjpeg-turbo Project and many others"
+  "Copyright (C) 1991-2019 The libjpeg-turbo Project and many others"
@@ -19,7 +19,7 @@ endif()
 #         Define the library target:
 # ----------------------------------------------------------------------------------

-add_library(${JPEG_LIBRARY} STATIC ${OPENCV_3RDPARTY_EXCLUDE_FROM_ALL} ${lib_srcs} ${lib_hdrs})
+add_library(${JPEG_LIBRARY} STATIC ${lib_srcs} ${lib_hdrs})

 if(CV_GCC OR CV_CLANG)
  set_source_files_properties(jcdctmgr.c PROPERTIES COMPILE_FLAGS "-O1")
@@ -42,7 +42,7 @@ if(ENABLE_SOLUTION_FOLDERS)
 endif()

 if(NOT BUILD_SHARED_LIBS)
-  ocv_install_target(${JPEG_LIBRARY} EXPORT OpenCVModules ARCHIVE DESTINATION ${OPENCV_3P_LIB_INSTALL_PATH} COMPONENT dev OPTIONAL)
+  ocv_install_target(${JPEG_LIBRARY} EXPORT OpenCVModules ARCHIVE DESTINATION ${OPENCV_3P_LIB_INSTALL_PATH} COMPONENT dev)
 endif()

 ocv_install_3rdparty_licenses(libjpeg README)
@@ -1,7 +1,7 @@
 The Independent JPEG Group's JPEG software
 ==========================================

-README for release 9d of 12-Jan-2020
+README for release 9c of 14-Jan-2018
 ====================================

 This distribution contains the ninth public release of the Independent JPEG
@@ -10,8 +10,8 @@ to use it for any purpose, subject to the conditions under LEGAL ISSUES, below.

 This software is the work of Tom Lane, Guido Vollbeding, Philip Gladstone,
 Bill Allombert, Jim Boucher, Lee Crocker, Bob Friesenhahn, Ben Jackson,
-John Korejwa, Julian Minguillon, Luis Ortiz, George Phillips, Davide Rossi,
-Ge' Weijers, and other members of the Independent JPEG Group.
+Julian Minguillon, Luis Ortiz, George Phillips, Davide Rossi, Ge' Weijers,
+and other members of the Independent JPEG Group.

 IJG is not affiliated with the ISO/IEC JTC1/SC29/WG1 standards committee
 (previously known as JPEG, together with ITU-T SG16).
@@ -115,7 +115,7 @@ with respect to this software, its quality, accuracy, merchantability, or
 fitness for a particular purpose.  This software is provided "AS IS", and you,
 its user, assume the entire risk as to its quality and accuracy.

-This software is copyright (C) 1991-2020, Thomas G. Lane, Guido Vollbeding.
+This software is copyright (C) 1991-2018, Thomas G. Lane, Guido Vollbeding.
 All Rights Reserved except as specified below.

 Permission is hereby granted to use, copy, modify, and distribute this
@@ -152,6 +152,13 @@ The same holds for its supporting scripts (config.guess, config.sub,
 ltmain.sh).  Another support script, install-sh, is copyright by X Consortium
 but is also freely distributable.

+The IJG distribution formerly included code to read and write GIF files.
+To avoid entanglement with the Unisys LZW patent (now expired), GIF reading
+support has been removed altogether, and the GIF writer has been simplified
+to produce "uncompressed GIFs".  This technique does not use the LZW
+algorithm; the resulting GIF files are larger than usual, but are readable
+by all standard GIF decoders.
+

 REFERENCES
 ==========
@@ -239,8 +246,8 @@ ARCHIVE LOCATIONS
 The "official" archive site for this software is www.ijg.org.
 The most recent released version can always be found there in
 directory "files".  This particular version will be archived as
-http://www.ijg.org/files/jpegsrc.v9d.tar.gz, and in Windows-compatible
-"zip" archive format as http://www.ijg.org/files/jpegsr9d.zip.
+http://www.ijg.org/files/jpegsrc.v9c.tar.gz, and in Windows-compatible
+"zip" archive format as http://www.ijg.org/files/jpegsr9c.zip.

 The JPEG FAQ (Frequently Asked Questions) article is a source of some
 general information about JPEG.
@@ -1,55 +1,6 @@
 CHANGE LOG for Independent JPEG Group's JPEG software


-Version 9d  12-Jan-2020
-----------------------
-
-Optimize the optimal Huffman code table generation to produce
-slightly smaller files.  Thank to John Korejwa for suggestion.
-Note: Requires rebuild of testimgp.jpg.
-
-Decoding Huffman: Use default tables if tables are not defined.
-Thank to Simone Azzalin for report (Motion JPEG),
-and to Martin Strunz for hint.
-
-Add sanity check in optimal Huffman code table generation.
-Thank to Adam Farley for suggestion.
-
-rdtarga.c: use read_byte(), with EOF check, instead of getc()
-in read_*_pixel().
-Thank to Chijin Zhou for cjpeg potential vulnerability report.
-
-jmemnobs.c: respect the max_memory_to_use setting in
-jpeg_mem_available() computation.  Thank to Sheng Shu and
-Dongdong She for djpeg potential vulnerability report.
-
-jdarith.c, jdhuff.c: avoid left shift of negative value
-compiler warning in decode_mcu_AC_refine().
-Thank to Indu Bhagat for suggestion.
-
-Add x64 (64-bit) platform support, avoid compiler warnings.
-Thank to Jonathan Potter, Feiyun Wang, and Sheng Shu for suggestion.
-
-Adjust libjpeg version specification for pkg-config file.
-Thank to Chen Chen for suggestion.
-
-Restore GIF read and write support from libjpeg version 6a.
-Thank to Wolfgang Werner (W.W.) Heinz for suggestion.
-
-Improve consistency in raw (downsampled) image data processing mode.
-Thank to Zhongyuan Zhou for hint.
-
-Avoid out of bounds array read (AC derived table pointers)
-in start pass in jdhuff.c.  Thank to Peng Li for report.
-
-Improve code sanity (jdhuff.c).
-Thank to Reza Mirzazade farkhani for reports.
-
-Add jpegtran -drop option; add options to the crop extension and wipe
-to fill the extra area with content from the source image region,
-instead of gray out.
-
-
 Version 9c  14-Jan-2018
 -----------------------

@@ -1,7 +1,7 @@
 /*
 * jcarith.c
 *
- * Developed 1997-2019 by Guido Vollbeding.
+ * Developed 1997-2013 by Guido Vollbeding.
 * This file is part of the Independent JPEG Group's software.
 * For conditions of distribution and use, see the accompanying README file.
 *
@@ -181,11 +181,11 @@ finish_pass (j_compress_ptr cinfo)
    if (e->zc)  /* output final pending zero bytes */
      do emit_byte(0x00, cinfo);
      while (--e->zc);
-    emit_byte((int) ((e->c >> 19) & 0xFF), cinfo);
+    emit_byte((e->c >> 19) & 0xFF, cinfo);
    if (((e->c >> 19) & 0xFF) == 0xFF)
      emit_byte(0x00, cinfo);
    if (e->c & 0x7F800L) {
-      emit_byte((int) ((e->c >> 11) & 0xFF), cinfo);
+      emit_byte((e->c >> 11) & 0xFF, cinfo);
      if (((e->c >> 11) & 0xFF) == 0xFF)
 	emit_byte(0x00, cinfo);
    }
@@ -280,8 +280,7 @@ arith_encode (j_compress_ptr cinfo, unsigned char *st, int val)
 	/* Note: The 3 spacer bits in the C register guarantee
 	 * that the new buffer byte can't be 0xFF here
 	 * (see page 160 in the P&M JPEG book). */
-	/* New output byte, might overflow later */
-	e->buffer = (int) (temp & 0xFF);
+	e->buffer = temp & 0xFF;  /* new output byte, might overflow later */
      } else if (temp == 0xFF) {
 	++e->sc;  /* stack 0xFF byte (which might overflow later) */
      } else {
@@ -303,8 +302,7 @@ arith_encode (j_compress_ptr cinfo, unsigned char *st, int val)
 	    emit_byte(0x00, cinfo);
 	  } while (--e->sc);
 	}
-	/* New output byte (can still overflow) */
-	e->buffer = (int) (temp & 0xFF);
+	e->buffer = temp & 0xFF;  /* new output byte (can still overflow) */
      }
      e->c &= 0x7FFFFL;
      e->ct += 8;
@@ -928,8 +926,9 @@ jinit_arith_encoder (j_compress_ptr cinfo)
  arith_entropy_ptr entropy;
  int i;

-  entropy = (arith_entropy_ptr) (*cinfo->mem->alloc_small)
-    ((j_common_ptr) cinfo, JPOOL_IMAGE, SIZEOF(arith_entropy_encoder));
+  entropy = (arith_entropy_ptr)
+    (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				SIZEOF(arith_entropy_encoder));
  cinfo->entropy = &entropy->pub;
  entropy->pub.start_pass = start_pass;
  entropy->pub.finish_pass = finish_pass;
@@ -2,7 +2,7 @@
 * jccolor.c
 *
 * Copyright (C) 1991-1996, Thomas G. Lane.
- * Modified 2011-2019 by Guido Vollbeding.
+ * Modified 2011-2013 by Guido Vollbeding.
 * This file is part of the Independent JPEG Group's software.
 * For conditions of distribution and use, see the accompanying README file.
 *
@@ -105,14 +105,14 @@ rgb_ycc_start (j_compress_ptr cinfo)
  /* Allocate and fill in the conversion tables. */
  cconvert->rgb_ycc_tab = rgb_ycc_tab = (INT32 *)
    (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-				TABLE_SIZE * SIZEOF(INT32));
+				(TABLE_SIZE * SIZEOF(INT32)));

  for (i = 0; i <= MAXJSAMPLE; i++) {
    rgb_ycc_tab[i+R_Y_OFF] = FIX(0.299) * i;
    rgb_ycc_tab[i+G_Y_OFF] = FIX(0.587) * i;
    rgb_ycc_tab[i+B_Y_OFF] = FIX(0.114) * i   + ONE_HALF;
-    rgb_ycc_tab[i+R_CB_OFF] = (- FIX(0.168735892)) * i;
-    rgb_ycc_tab[i+G_CB_OFF] = (- FIX(0.331264108)) * i;
+    rgb_ycc_tab[i+R_CB_OFF] = (-FIX(0.168735892)) * i;
+    rgb_ycc_tab[i+G_CB_OFF] = (-FIX(0.331264108)) * i;
    /* We use a rounding fudge-factor of 0.5-epsilon for Cb and Cr.
     * This ensures that the maximum output will round to MAXJSAMPLE
     * not MAXJSAMPLE+1, and thus that we don't have to range-limit.
@@ -121,8 +121,8 @@ rgb_ycc_start (j_compress_ptr cinfo)
 /*  B=>Cb and R=>Cr tables are the same
    rgb_ycc_tab[i+R_CR_OFF] = FIX(0.5) * i    + CBCR_OFFSET + ONE_HALF-1;
 */
-    rgb_ycc_tab[i+G_CR_OFF] = (- FIX(0.418687589)) * i;
-    rgb_ycc_tab[i+B_CR_OFF] = (- FIX(0.081312411)) * i;
+    rgb_ycc_tab[i+G_CR_OFF] = (-FIX(0.418687589)) * i;
+    rgb_ycc_tab[i+B_CR_OFF] = (-FIX(0.081312411)) * i;
  }
 }

@@ -131,12 +131,12 @@ rgb_ycc_start (j_compress_ptr cinfo)
 * Convert some rows of samples to the JPEG colorspace.
 *
 * Note that we change from the application's interleaved-pixel format
- * to our internal noninterleaved, one-plane-per-component format.  The
- * input buffer is therefore three times as wide as the output buffer.
+ * to our internal noninterleaved, one-plane-per-component format.
+ * The input buffer is therefore three times as wide as the output buffer.
 *
- * A starting row offset is provided only for the output buffer.  The
- * caller can easily adjust the passed input_buf value to accommodate
- * any row offset required on that side.
+ * A starting row offset is provided only for the output buffer.  The caller
+ * can easily adjust the passed input_buf value to accommodate any row
+ * offset required on that side.
 */

 METHODDEF(void)
@@ -145,8 +145,8 @@ rgb_ycc_convert (j_compress_ptr cinfo,
 		 JDIMENSION output_row, int num_rows)
 {
  my_cconvert_ptr cconvert = (my_cconvert_ptr) cinfo->cconvert;
-  register int r, g, b;
  register INT32 * ctab = cconvert->rgb_ycc_tab;
+  register int r, g, b;
  register JSAMPROW inptr;
  register JSAMPROW outptr0, outptr1, outptr2;
  register JDIMENSION col;
@@ -162,7 +162,6 @@ rgb_ycc_convert (j_compress_ptr cinfo,
      r = GETJSAMPLE(inptr[RGB_RED]);
      g = GETJSAMPLE(inptr[RGB_GREEN]);
      b = GETJSAMPLE(inptr[RGB_BLUE]);
-      inptr += RGB_PIXELSIZE;
      /* If the inputs are 0..MAXJSAMPLE, the outputs of these equations
       * must be too; we do not need an explicit range-limiting operation.
       * Hence the value being shifted is never negative, and we don't
@@ -180,6 +179,7 @@ rgb_ycc_convert (j_compress_ptr cinfo,
      outptr2[col] = (JSAMPLE)
 		((ctab[r+R_CR_OFF] + ctab[g+G_CR_OFF] + ctab[b+B_CR_OFF])
 		 >> SCALEBITS);
+      inptr += RGB_PIXELSIZE;
    }
  }
 }
@@ -201,8 +201,8 @@ rgb_gray_convert (j_compress_ptr cinfo,
 		  JDIMENSION output_row, int num_rows)
 {
  my_cconvert_ptr cconvert = (my_cconvert_ptr) cinfo->cconvert;
-  register int r, g, b;
  register INT32 * ctab = cconvert->rgb_ycc_tab;
+  register int r, g, b;
  register JSAMPROW inptr;
  register JSAMPROW outptr;
  register JDIMENSION col;
@@ -215,11 +215,11 @@ rgb_gray_convert (j_compress_ptr cinfo,
      r = GETJSAMPLE(inptr[RGB_RED]);
      g = GETJSAMPLE(inptr[RGB_GREEN]);
      b = GETJSAMPLE(inptr[RGB_BLUE]);
-      inptr += RGB_PIXELSIZE;
      /* Y */
      outptr[col] = (JSAMPLE)
 		((ctab[r+R_Y_OFF] + ctab[g+G_Y_OFF] + ctab[b+B_Y_OFF])
 		 >> SCALEBITS);
+      inptr += RGB_PIXELSIZE;
    }
  }
 }
@@ -228,8 +228,8 @@ rgb_gray_convert (j_compress_ptr cinfo,
 /*
 * Convert some rows of samples to the JPEG colorspace.
 * This version handles Adobe-style CMYK->YCCK conversion,
- * where we convert R=1-C, G=1-M, and B=1-Y to YCbCr using the
- * same conversion as above, while passing K (black) unchanged.
+ * where we convert R=1-C, G=1-M, and B=1-Y to YCbCr using the same
+ * conversion as above, while passing K (black) unchanged.
 * We assume rgb_ycc_start has been called.
 */

@@ -239,8 +239,8 @@ cmyk_ycck_convert (j_compress_ptr cinfo,
 		   JDIMENSION output_row, int num_rows)
 {
  my_cconvert_ptr cconvert = (my_cconvert_ptr) cinfo->cconvert;
-  register int r, g, b;
  register INT32 * ctab = cconvert->rgb_ycc_tab;
+  register int r, g, b;
  register JSAMPROW inptr;
  register JSAMPROW outptr0, outptr1, outptr2, outptr3;
  register JDIMENSION col;
@@ -259,7 +259,6 @@ cmyk_ycck_convert (j_compress_ptr cinfo,
      b = MAXJSAMPLE - GETJSAMPLE(inptr[2]);
      /* K passes through as-is */
      outptr3[col] = inptr[3];	/* don't need GETJSAMPLE here */
-      inptr += 4;
      /* If the inputs are 0..MAXJSAMPLE, the outputs of these equations
       * must be too; we do not need an explicit range-limiting operation.
       * Hence the value being shifted is never negative, and we don't
@@ -277,6 +276,7 @@ cmyk_ycck_convert (j_compress_ptr cinfo,
      outptr2[col] = (JSAMPLE)
 		((ctab[r+R_CR_OFF] + ctab[g+G_CR_OFF] + ctab[b+B_CR_OFF])
 		 >> SCALEBITS);
+      inptr += 4;
    }
  }
 }
@@ -312,13 +312,13 @@ rgb_rgb1_convert (j_compress_ptr cinfo,
      r = GETJSAMPLE(inptr[RGB_RED]);
      g = GETJSAMPLE(inptr[RGB_GREEN]);
      b = GETJSAMPLE(inptr[RGB_BLUE]);
-      inptr += RGB_PIXELSIZE;
      /* Assume that MAXJSAMPLE+1 is a power of 2, so that the MOD
       * (modulo) operator is equivalent to the bitmask operator AND.
       */
      outptr0[col] = (JSAMPLE) ((r - g + CENTERJSAMPLE) & MAXJSAMPLE);
      outptr1[col] = (JSAMPLE) g;
      outptr2[col] = (JSAMPLE) ((b - g + CENTERJSAMPLE) & MAXJSAMPLE);
+      inptr += RGB_PIXELSIZE;
    }
  }
 }
@@ -335,17 +335,17 @@ grayscale_convert (j_compress_ptr cinfo,
 		   JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
 		   JDIMENSION output_row, int num_rows)
 {
+  int instride = cinfo->input_components;
  register JSAMPROW inptr;
  register JSAMPROW outptr;
-  register JDIMENSION count;
-  register int instride = cinfo->input_components;
+  register JDIMENSION col;
  JDIMENSION num_cols = cinfo->image_width;

  while (--num_rows >= 0) {
    inptr = *input_buf++;
    outptr = output_buf[0][output_row++];
-    for (count = num_cols; count > 0; count--) {
-      *outptr++ = *inptr;	/* don't need GETJSAMPLE() here */
+    for (col = 0; col < num_cols; col++) {
+      outptr[col] = inptr[0];	/* don't need GETJSAMPLE() here */
      inptr += instride;
    }
  }
@@ -396,21 +396,21 @@ null_convert (j_compress_ptr cinfo,
 	      JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
 	      JDIMENSION output_row, int num_rows)
 {
+  int ci;
+  register int nc = cinfo->num_components;
  register JSAMPROW inptr;
  register JSAMPROW outptr;
-  register JDIMENSION count;
-  register int num_comps = cinfo->num_components;
+  register JDIMENSION col;
  JDIMENSION num_cols = cinfo->image_width;
-  int ci;

  while (--num_rows >= 0) {
    /* It seems fastest to make a separate pass for each component. */
-    for (ci = 0; ci < num_comps; ci++) {
+    for (ci = 0; ci < nc; ci++) {
      inptr = input_buf[0] + ci;
      outptr = output_buf[ci][output_row];
-      for (count = num_cols; count > 0; count--) {
+      for (col = 0; col < num_cols; col++) {
 	*outptr++ = *inptr;	/* don't need GETJSAMPLE() here */
-	inptr += num_comps;
+	inptr += nc;
      }
    }
    input_buf++;
@@ -439,8 +439,9 @@ jinit_color_converter (j_compress_ptr cinfo)
 {
  my_cconvert_ptr cconvert;

-  cconvert = (my_cconvert_ptr) (*cinfo->mem->alloc_small)
-    ((j_common_ptr) cinfo, JPOOL_IMAGE, SIZEOF(my_color_converter));
+  cconvert = (my_cconvert_ptr)
+    (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				SIZEOF(my_color_converter));
  cinfo->cconvert = &cconvert->pub;
  /* set start_pass to null method until we find out differently */
  cconvert->pub.start_pass = null_method;
@@ -454,11 +455,9 @@ jinit_color_converter (j_compress_ptr cinfo)

  case JCS_RGB:
  case JCS_BG_RGB:
-#if RGB_PIXELSIZE != 3
    if (cinfo->input_components != RGB_PIXELSIZE)
      ERREXIT(cinfo, JERR_BAD_IN_COLORSPACE);
    break;
-#endif /* else share code with YCbCr */

  case JCS_YCbCr:
  case JCS_BG_YCC:
@@ -475,6 +474,7 @@ jinit_color_converter (j_compress_ptr cinfo)
  default:			/* JCS_UNKNOWN can be anything */
    if (cinfo->input_components < 1)
      ERREXIT(cinfo, JERR_BAD_IN_COLORSPACE);
+    break;
  }

  /* Support color transform only for RGB colorspaces */
@@ -507,18 +507,19 @@ jinit_color_converter (j_compress_ptr cinfo)
  case JCS_BG_RGB:
    if (cinfo->num_components != 3)
      ERREXIT(cinfo, JERR_BAD_J_COLORSPACE);
-    if (cinfo->in_color_space != cinfo->jpeg_color_space)
+    if (cinfo->in_color_space == cinfo->jpeg_color_space) {
+      switch (cinfo->color_transform) {
+      case JCT_NONE:
+	cconvert->pub.color_convert = rgb_convert;
+	break;
+      case JCT_SUBTRACT_GREEN:
+	cconvert->pub.color_convert = rgb_rgb1_convert;
+	break;
+      default:
+	ERREXIT(cinfo, JERR_CONVERSION_NOTIMPL);
+      }
+    } else
      ERREXIT(cinfo, JERR_CONVERSION_NOTIMPL);
-    switch (cinfo->color_transform) {
-    case JCT_NONE:
-      cconvert->pub.color_convert = rgb_convert;
-      break;
-    case JCT_SUBTRACT_GREEN:
-      cconvert->pub.color_convert = rgb_rgb1_convert;
-      break;
-    default:
-      ERREXIT(cinfo, JERR_CONVERSION_NOTIMPL);
-    }
    break;

  case JCS_YCbCr:
@@ -571,9 +572,10 @@ jinit_color_converter (j_compress_ptr cinfo)
  case JCS_CMYK:
    if (cinfo->num_components != 4)
      ERREXIT(cinfo, JERR_BAD_J_COLORSPACE);
-    if (cinfo->in_color_space != JCS_CMYK)
+    if (cinfo->in_color_space == JCS_CMYK)
+      cconvert->pub.color_convert = null_convert;
+    else
      ERREXIT(cinfo, JERR_CONVERSION_NOTIMPL);
-    cconvert->pub.color_convert = null_convert;
    break;

  case JCS_YCCK:
@@ -597,5 +599,6 @@ jinit_color_converter (j_compress_ptr cinfo)
 	cinfo->num_components != cinfo->input_components)
      ERREXIT(cinfo, JERR_CONVERSION_NOTIMPL);
    cconvert->pub.color_convert = null_convert;
+    break;
  }
 }
@@ -2,7 +2,7 @@
 * jchuff.c
 *
 * Copyright (C) 1991-1997, Thomas G. Lane.
- * Modified 2006-2019 by Guido Vollbeding.
+ * Modified 2006-2013 by Guido Vollbeding.
 * This file is part of the Independent JPEG Group's software.
 * For conditions of distribution and use, see the accompanying README file.
 *
@@ -178,12 +178,13 @@ jpeg_make_c_derived_tbl (j_compress_ptr cinfo, boolean isDC, int tblno,
  htbl =
    isDC ? cinfo->dc_huff_tbl_ptrs[tblno] : cinfo->ac_huff_tbl_ptrs[tblno];
  if (htbl == NULL)
-    htbl = jpeg_std_huff_table((j_common_ptr) cinfo, isDC, tblno);
+    ERREXIT1(cinfo, JERR_NO_HUFF_TABLE, tblno);

  /* Allocate a workspace if we haven't already done so. */
  if (*pdtbl == NULL)
-    *pdtbl = (c_derived_tbl *) (*cinfo->mem->alloc_small)
-      ((j_common_ptr) cinfo, JPOOL_IMAGE, SIZEOF(c_derived_tbl));
+    *pdtbl = (c_derived_tbl *)
+      (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				  SIZEOF(c_derived_tbl));
  dtbl = *pdtbl;
  
  /* Figure C.1: make table of Huffman code length for each symbol */
@@ -1255,88 +1256,22 @@ jpeg_gen_optimal_table (j_compress_ptr cinfo, JHUFF_TBL * htbl, long freq[])
  UINT8 bits[MAX_CLEN+1];	/* bits[k] = # of symbols with code length k */
  int codesize[257];		/* codesize[k] = code length of symbol k */
  int others[257];		/* next symbol in current branch of tree */
-  int c1, c2, i, j;
-  UINT8 *p;
+  int c1, c2;
+  int p, i, j;
  long v;

-  freq[256] = 1;		/* make sure 256 has a nonzero count */
-  /* Including the pseudo-symbol 256 in the Huffman procedure guarantees
-   * that no real symbol is given code-value of all ones, because 256
-   * will be placed last in the largest codeword category.
-   * In the symbol list build procedure this element serves as sentinel
-   * for the zero run loop.
-   */
-
-#ifndef DONT_USE_FANCY_HUFF_OPT
-
-  /* Build list of symbols sorted in order of descending frequency */
-  /* This approach has several benefits (thank to John Korejwa for the idea):
-   *     1.
-   * If a codelength category is split during the length limiting procedure
-   * below, the feature that more frequent symbols are assigned shorter
-   * codewords remains valid for the adjusted code.
-   *     2.
-   * To reduce consecutive ones in a Huffman data stream (thus reducing the
-   * number of stuff bytes in JPEG) it is preferable to follow 0 branches
-   * (and avoid 1 branches) as much as possible.  This is easily done by
-   * assigning symbols to leaves of the Huffman tree in order of decreasing
-   * frequency, with no secondary sort based on codelengths.
-   *     3.
-   * The symbol list can be built independently from the assignment of code
-   * lengths by the Huffman procedure below.
-   * Note: The symbol list build procedure must be performed first, because
-   * the Huffman procedure assigning the codelengths clobbers the frequency
-   * counts!
-   */
-
-  /* Here we use the others array as a linked list of nonzero frequencies
-   * to be sorted.  Already sorted elements are removed from the list.
-   */
-
-  /* Building list */
-
-  /* This item does not correspond to a valid symbol frequency and is used
-   * as starting index.
-   */
-  j = 256;
-
-  for (i = 0;; i++) {
-    if (freq[i] == 0)		/* skip zero frequencies */
-      continue;
-    if (i > 255)
-      break;
-    others[j] = i;		/* this symbol value */
-    j = i;			/* previous symbol value */
-  }
-  others[j] = -1;		/* mark end of list */
-
-  /* Sorting list */
-
-  p = htbl->huffval;
-  while ((c1 = others[256]) >= 0) {
-    v = freq[c1];
-    i = c1;			/* first symbol value */
-    j = 256;			/* pseudo symbol value for starting index */
-    while ((c2 = others[c1]) >= 0) {
-      if (freq[c2] > v) {
-	v = freq[c2];
-	i = c2;			/* this symbol value */
-	j = c1;			/* previous symbol value */
-      }
-      c1 = c2;
-    }
-    others[j] = others[i];	/* remove this symbol i from list */
-    *p++ = (UINT8) i;
-  }
-
-#endif /* DONT_USE_FANCY_HUFF_OPT */
-
  /* This algorithm is explained in section K.2 of the JPEG standard */

  MEMZERO(bits, SIZEOF(bits));
  MEMZERO(codesize, SIZEOF(codesize));
  for (i = 0; i < 257; i++)
    others[i] = -1;		/* init links to empty */
+  
+  freq[256] = 1;		/* make sure 256 has a nonzero count */
+  /* Including the pseudo-symbol 256 in the Huffman procedure guarantees
+   * that no real symbol is given code-value of all ones, because 256
+   * will be placed last in the largest codeword category.
+   */

  /* Huffman's basic algorithm to assign optimal code lengths to symbols */

@@ -1366,7 +1301,7 @@ jpeg_gen_optimal_table (j_compress_ptr cinfo, JHUFF_TBL * htbl, long freq[])
    /* Done if we've merged everything into one frequency */
    if (c2 < 0)
      break;
-
+    
    /* Else merge the two counts/trees */
    freq[c1] += freq[c2];
    freq[c2] = 0;
@@ -1377,9 +1312,9 @@ jpeg_gen_optimal_table (j_compress_ptr cinfo, JHUFF_TBL * htbl, long freq[])
      c1 = others[c1];
      codesize[c1]++;
    }
-
+    
    others[c1] = c2;		/* chain c2 onto c1's tree branch */
-
+    
    /* Increment the codesize of everything in c2's tree branch */
    codesize[c2]++;
    while (others[c2] >= 0) {
@@ -1394,7 +1329,7 @@ jpeg_gen_optimal_table (j_compress_ptr cinfo, JHUFF_TBL * htbl, long freq[])
      /* The JPEG standard seems to think that this can't happen, */
      /* but I'm paranoid... */
      if (codesize[i] > MAX_CLEN)
-	ERREXIT(cinfo, JERR_HUFF_CLEN_OUTOFBOUNDS);
+	ERREXIT(cinfo, JERR_HUFF_CLEN_OVERFLOW);

      bits[codesize[i]]++;
    }
@@ -1410,16 +1345,13 @@ jpeg_gen_optimal_table (j_compress_ptr cinfo, JHUFF_TBL * htbl, long freq[])
   * shortest nonzero BITS entry is converted into a prefix for two code words
   * one bit longer.
   */
-
+  
  for (i = MAX_CLEN; i > 16; i--) {
    while (bits[i] > 0) {
      j = i - 2;		/* find length of new prefix to be used */
-      while (bits[j] == 0) {
-	if (j == 0)
-	  ERREXIT(cinfo, JERR_HUFF_CLEN_OUTOFBOUNDS);
+      while (bits[j] == 0)
 	j--;
-      }
-
+      
      bits[i] -= 2;		/* remove two symbols */
      bits[i-1]++;		/* one goes in this length */
      bits[j+1] += 2;		/* two new symbols in this length */
@@ -1431,27 +1363,24 @@ jpeg_gen_optimal_table (j_compress_ptr cinfo, JHUFF_TBL * htbl, long freq[])
  while (bits[i] == 0)		/* find largest codelength still in use */
    i--;
  bits[i]--;
-
+  
  /* Return final symbol counts (only for lengths 0..16) */
  MEMCOPY(htbl->bits, bits, SIZEOF(htbl->bits));
-
-#ifdef DONT_USE_FANCY_HUFF_OPT
-
+  
  /* Return a list of the symbols sorted by code length */
-  /* Note: Due to the codelength changes made above, it can happen
-   * that more frequent symbols are assigned longer codewords.
+  /* It's not real clear to me why we don't need to consider the codelength
+   * changes made above, but the JPEG spec seems to think this works.
   */
-  p = htbl->huffval;
+  p = 0;
  for (i = 1; i <= MAX_CLEN; i++) {
    for (j = 0; j <= 255; j++) {
      if (codesize[j] == i) {
-	*p++ = (UINT8) j;
+	htbl->huffval[p] = (UINT8) j;
+	p++;
      }
    }
  }

-#endif /* DONT_USE_FANCY_HUFF_OPT */
-
  /* Set sent_table FALSE so updated table will be written to JPEG file. */
  htbl->sent_table = FALSE;
 }
@@ -1471,13 +1400,13 @@ finish_pass_gather (j_compress_ptr cinfo)
  boolean did_dc[NUM_HUFF_TBLS];
  boolean did_ac[NUM_HUFF_TBLS];

+  /* It's important not to apply jpeg_gen_optimal_table more than once
+   * per table, because it clobbers the input frequency counts!
+   */
  if (cinfo->progressive_mode)
    /* Flush out buffered data (all we care about is counting the EOB symbol) */
    emit_eobrun(entropy);

-  /* It's important not to apply jpeg_gen_optimal_table more than once
-   * per table, because it clobbers the input frequency counts!
-   */
  MEMZERO(did_dc, SIZEOF(did_dc));
  MEMZERO(did_ac, SIZEOF(did_ac));

@@ -1546,8 +1475,9 @@ start_pass_huff (j_compress_ptr cinfo, boolean gather_statistics)
 	entropy->pub.encode_mcu = encode_mcu_AC_refine;
 	/* AC refinement needs a correction bit buffer */
 	if (entropy->bit_buffer == NULL)
-	  entropy->bit_buffer = (char *) (*cinfo->mem->alloc_small)
-	    ((j_common_ptr) cinfo, JPOOL_IMAGE, MAX_CORR_BITS * SIZEOF(char));
+	  entropy->bit_buffer = (char *)
+	    (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+					MAX_CORR_BITS * SIZEOF(char));
      }
    }

@@ -1575,8 +1505,9 @@ start_pass_huff (j_compress_ptr cinfo, boolean gather_statistics)
 	/* Allocate and zero the statistics tables */
 	/* Note that jpeg_gen_optimal_table expects 257 entries in each table! */
 	if (entropy->dc_count_ptrs[tbl] == NULL)
-	  entropy->dc_count_ptrs[tbl] = (long *) (*cinfo->mem->alloc_small)
-	    ((j_common_ptr) cinfo, JPOOL_IMAGE, 257 * SIZEOF(long));
+	  entropy->dc_count_ptrs[tbl] = (long *)
+	    (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+					257 * SIZEOF(long));
 	MEMZERO(entropy->dc_count_ptrs[tbl], 257 * SIZEOF(long));
      } else {
 	/* Compute derived values for Huffman tables */
@@ -1594,8 +1525,9 @@ start_pass_huff (j_compress_ptr cinfo, boolean gather_statistics)
 	if (tbl < 0 || tbl >= NUM_HUFF_TBLS)
 	  ERREXIT1(cinfo, JERR_NO_HUFF_TABLE, tbl);
 	if (entropy->ac_count_ptrs[tbl] == NULL)
-	  entropy->ac_count_ptrs[tbl] = (long *) (*cinfo->mem->alloc_small)
-	    ((j_common_ptr) cinfo, JPOOL_IMAGE, 257 * SIZEOF(long));
+	  entropy->ac_count_ptrs[tbl] = (long *)
+	    (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+					257 * SIZEOF(long));
 	MEMZERO(entropy->ac_count_ptrs[tbl], 257 * SIZEOF(long));
      } else {
 	jpeg_make_c_derived_tbl(cinfo, FALSE, tbl,
@@ -1624,8 +1556,9 @@ jinit_huff_encoder (j_compress_ptr cinfo)
  huff_entropy_ptr entropy;
  int i;

-  entropy = (huff_entropy_ptr) (*cinfo->mem->alloc_small)
-    ((j_common_ptr) cinfo, JPOOL_IMAGE, SIZEOF(huff_entropy_encoder));
+  entropy = (huff_entropy_ptr)
+    (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				SIZEOF(huff_entropy_encoder));
  cinfo->entropy = &entropy->pub;
  entropy->pub.start_pass = start_pass_huff;

@@ -2,7 +2,7 @@
 * jcmarker.c
 *
 * Copyright (C) 1991-1998, Thomas G. Lane.
- * Modified 2003-2019 by Guido Vollbeding.
+ * Modified 2003-2013 by Guido Vollbeding.
 * This file is part of the Independent JPEG Group's software.
 * For conditions of distribution and use, see the accompanying README file.
 *
@@ -471,6 +471,7 @@ emit_adobe_app14 (j_compress_ptr cinfo)
    break;
  default:
    emit_byte(cinfo, 0);	/* Color transform = 0 */
+    break;
  }
 }

@@ -701,8 +702,9 @@ jinit_marker_writer (j_compress_ptr cinfo)
  my_marker_ptr marker;

  /* Create the subobject */
-  marker = (my_marker_ptr) (*cinfo->mem->alloc_small)
-    ((j_common_ptr) cinfo, JPOOL_IMAGE, SIZEOF(my_marker_writer));
+  marker = (my_marker_ptr)
+    (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				SIZEOF(my_marker_writer));
  cinfo->marker = &marker->pub;
  /* Initialize method pointers */
  marker->pub.write_file_header = write_file_header;
@@ -2,7 +2,7 @@
 * jcmaster.c
 *
 * Copyright (C) 1991-1997, Thomas G. Lane.
- * Modified 2003-2019 by Guido Vollbeding.
+ * Modified 2003-2017 by Guido Vollbeding.
 * This file is part of the Independent JPEG Group's software.
 * For conditions of distribution and use, see the accompanying README file.
 *
@@ -62,7 +62,7 @@ initial_setup (j_compress_ptr cinfo)
  case 5: cinfo->natural_order = jpeg_natural_order5; break;
  case 6: cinfo->natural_order = jpeg_natural_order6; break;
  case 7: cinfo->natural_order = jpeg_natural_order7; break;
-  default: cinfo->natural_order = jpeg_natural_order;
+  default: cinfo->natural_order = jpeg_natural_order; break;
  }

  /* Derive lim_Se from block_size */
@@ -114,24 +114,20 @@ initial_setup (j_compress_ptr cinfo)
     */
    ssize = 1;
 #ifdef DCT_SCALING_SUPPORTED
-    if (! cinfo->raw_data_in)
-      while (cinfo->min_DCT_h_scaled_size * ssize <=
-	     (cinfo->do_fancy_downsampling ? DCTSIZE : DCTSIZE / 2) &&
-	     (cinfo->max_h_samp_factor % (compptr->h_samp_factor * ssize * 2)) ==
-	     0) {
-	ssize = ssize * 2;
-      }
+    while (cinfo->min_DCT_h_scaled_size * ssize <=
+	   (cinfo->do_fancy_downsampling ? DCTSIZE : DCTSIZE / 2) &&
+	   (cinfo->max_h_samp_factor % (compptr->h_samp_factor * ssize * 2)) == 0) {
+      ssize = ssize * 2;
+    }
 #endif
    compptr->DCT_h_scaled_size = cinfo->min_DCT_h_scaled_size * ssize;
    ssize = 1;
 #ifdef DCT_SCALING_SUPPORTED
-    if (! cinfo->raw_data_in)
-      while (cinfo->min_DCT_v_scaled_size * ssize <=
-	     (cinfo->do_fancy_downsampling ? DCTSIZE : DCTSIZE / 2) &&
-	     (cinfo->max_v_samp_factor % (compptr->v_samp_factor * ssize * 2)) ==
-	     0) {
-	ssize = ssize * 2;
-      }
+    while (cinfo->min_DCT_v_scaled_size * ssize <=
+	   (cinfo->do_fancy_downsampling ? DCTSIZE : DCTSIZE / 2) &&
+	   (cinfo->max_v_samp_factor % (compptr->v_samp_factor * ssize * 2)) == 0) {
+      ssize = ssize * 2;
+    }
 #endif
    compptr->DCT_v_scaled_size = cinfo->min_DCT_v_scaled_size * ssize;

@@ -624,8 +620,9 @@ jinit_c_master_control (j_compress_ptr cinfo, boolean transcode_only)
 {
  my_master_ptr master;

-  master = (my_master_ptr) (*cinfo->mem->alloc_small)
-    ((j_common_ptr) cinfo, JPOOL_IMAGE, SIZEOF(my_comp_master));
+  master = (my_master_ptr)
+      (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				  SIZEOF(my_comp_master));
  cinfo->master = &master->pub;
  master->pub.prepare_for_pass = prepare_for_pass;
  master->pub.pass_startup = pass_startup;
@@ -2,7 +2,6 @@
 * jcomapi.c
 *
 * Copyright (C) 1994-1997, Thomas G. Lane.
- * Modified 2019 by Guido Vollbeding.
 * This file is part of the Independent JPEG Group's software.
 * For conditions of distribution and use, see the accompanying README file.
 *
@@ -105,140 +104,3 @@ jpeg_alloc_huff_table (j_common_ptr cinfo)
  tbl->sent_table = FALSE;	/* make sure this is false in any new table */
  return tbl;
 }
-
-
-/*
- * Set up the standard Huffman tables (cf. JPEG standard section K.3).
- * IMPORTANT: these are only valid for 8-bit data precision!
- * (Would jutils.c be a more reasonable place to put this?)
- */
-
-GLOBAL(JHUFF_TBL *)
-jpeg_std_huff_table (j_common_ptr cinfo, boolean isDC, int tblno)
-{
-  JHUFF_TBL **htblptr, *htbl;
-  const UINT8 *bits, *val;
-  int nsymbols, len;
-
-  static const UINT8 bits_dc_luminance[17] =
-    { /* 0-base */ 0, 0, 1, 5, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0 };
-  static const UINT8 val_dc_luminance[] =
-    { 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11 };
-
-  static const UINT8 bits_dc_chrominance[17] =
-    { /* 0-base */ 0, 0, 3, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0 };
-  static const UINT8 val_dc_chrominance[] =
-    { 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11 };
-
-  static const UINT8 bits_ac_luminance[17] =
-    { /* 0-base */ 0, 0, 2, 1, 3, 3, 2, 4, 3, 5, 5, 4, 4, 0, 0, 1, 0x7d };
-  static const UINT8 val_ac_luminance[] =
-    { 0x01, 0x02, 0x03, 0x00, 0x04, 0x11, 0x05, 0x12,
-      0x21, 0x31, 0x41, 0x06, 0x13, 0x51, 0x61, 0x07,
-      0x22, 0x71, 0x14, 0x32, 0x81, 0x91, 0xa1, 0x08,
-      0x23, 0x42, 0xb1, 0xc1, 0x15, 0x52, 0xd1, 0xf0,
-      0x24, 0x33, 0x62, 0x72, 0x82, 0x09, 0x0a, 0x16,
-      0x17, 0x18, 0x19, 0x1a, 0x25, 0x26, 0x27, 0x28,
-      0x29, 0x2a, 0x34, 0x35, 0x36, 0x37, 0x38, 0x39,
-      0x3a, 0x43, 0x44, 0x45, 0x46, 0x47, 0x48, 0x49,
-      0x4a, 0x53, 0x54, 0x55, 0x56, 0x57, 0x58, 0x59,
-      0x5a, 0x63, 0x64, 0x65, 0x66, 0x67, 0x68, 0x69,
-      0x6a, 0x73, 0x74, 0x75, 0x76, 0x77, 0x78, 0x79,
-      0x7a, 0x83, 0x84, 0x85, 0x86, 0x87, 0x88, 0x89,
-      0x8a, 0x92, 0x93, 0x94, 0x95, 0x96, 0x97, 0x98,
-      0x99, 0x9a, 0xa2, 0xa3, 0xa4, 0xa5, 0xa6, 0xa7,
-      0xa8, 0xa9, 0xaa, 0xb2, 0xb3, 0xb4, 0xb5, 0xb6,
-      0xb7, 0xb8, 0xb9, 0xba, 0xc2, 0xc3, 0xc4, 0xc5,
-      0xc6, 0xc7, 0xc8, 0xc9, 0xca, 0xd2, 0xd3, 0xd4,
-      0xd5, 0xd6, 0xd7, 0xd8, 0xd9, 0xda, 0xe1, 0xe2,
-      0xe3, 0xe4, 0xe5, 0xe6, 0xe7, 0xe8, 0xe9, 0xea,
-      0xf1, 0xf2, 0xf3, 0xf4, 0xf5, 0xf6, 0xf7, 0xf8,
-      0xf9, 0xfa };
-
-  static const UINT8 bits_ac_chrominance[17] =
-    { /* 0-base */ 0, 0, 2, 1, 2, 4, 4, 3, 4, 7, 5, 4, 4, 0, 1, 2, 0x77 };
-  static const UINT8 val_ac_chrominance[] =
-    { 0x00, 0x01, 0x02, 0x03, 0x11, 0x04, 0x05, 0x21,
-      0x31, 0x06, 0x12, 0x41, 0x51, 0x07, 0x61, 0x71,
-      0x13, 0x22, 0x32, 0x81, 0x08, 0x14, 0x42, 0x91,
-      0xa1, 0xb1, 0xc1, 0x09, 0x23, 0x33, 0x52, 0xf0,
-      0x15, 0x62, 0x72, 0xd1, 0x0a, 0x16, 0x24, 0x34,
-      0xe1, 0x25, 0xf1, 0x17, 0x18, 0x19, 0x1a, 0x26,
-      0x27, 0x28, 0x29, 0x2a, 0x35, 0x36, 0x37, 0x38,
-      0x39, 0x3a, 0x43, 0x44, 0x45, 0x46, 0x47, 0x48,
-      0x49, 0x4a, 0x53, 0x54, 0x55, 0x56, 0x57, 0x58,
-      0x59, 0x5a, 0x63, 0x64, 0x65, 0x66, 0x67, 0x68,
-      0x69, 0x6a, 0x73, 0x74, 0x75, 0x76, 0x77, 0x78,
-      0x79, 0x7a, 0x82, 0x83, 0x84, 0x85, 0x86, 0x87,
-      0x88, 0x89, 0x8a, 0x92, 0x93, 0x94, 0x95, 0x96,
-      0x97, 0x98, 0x99, 0x9a, 0xa2, 0xa3, 0xa4, 0xa5,
-      0xa6, 0xa7, 0xa8, 0xa9, 0xaa, 0xb2, 0xb3, 0xb4,
-      0xb5, 0xb6, 0xb7, 0xb8, 0xb9, 0xba, 0xc2, 0xc3,
-      0xc4, 0xc5, 0xc6, 0xc7, 0xc8, 0xc9, 0xca, 0xd2,
-      0xd3, 0xd4, 0xd5, 0xd6, 0xd7, 0xd8, 0xd9, 0xda,
-      0xe2, 0xe3, 0xe4, 0xe5, 0xe6, 0xe7, 0xe8, 0xe9,
-      0xea, 0xf2, 0xf3, 0xf4, 0xf5, 0xf6, 0xf7, 0xf8,
-      0xf9, 0xfa };
-
-  if (cinfo->is_decompressor) {
-    if (isDC)
-      htblptr = ((j_decompress_ptr) cinfo)->dc_huff_tbl_ptrs;
-    else
-      htblptr = ((j_decompress_ptr) cinfo)->ac_huff_tbl_ptrs;
-  } else {
-    if (isDC)
-      htblptr = ((j_compress_ptr) cinfo)->dc_huff_tbl_ptrs;
-    else
-      htblptr = ((j_compress_ptr) cinfo)->ac_huff_tbl_ptrs;
-  }
-
-  switch (tblno) {
-  case 0:
-    if (isDC) {
-      bits = bits_dc_luminance;
-      val = val_dc_luminance;
-    } else {
-      bits = bits_ac_luminance;
-      val = val_ac_luminance;
-    }
-    break;
-  case 1:
-    if (isDC) {
-      bits = bits_dc_chrominance;
-      val = val_dc_chrominance;
-    } else {
-      bits = bits_ac_chrominance;
-      val = val_ac_chrominance;
-    }
-    break;
-  default:
-    ERREXIT1(cinfo, JERR_NO_HUFF_TABLE, tblno);
-    return NULL; /* avoid compiler warnings for uninitialized variables */
-  }
-
-  if (htblptr[tblno] == NULL)
-    htblptr[tblno] = jpeg_alloc_huff_table(cinfo);
-
-  htbl = htblptr[tblno];
-
-  /* Copy the number-of-symbols-of-each-code-length counts */
-  MEMCOPY(htbl->bits, bits, SIZEOF(htbl->bits));
-
-  /* Validate the counts.  We do this here mainly so we can copy the right
-   * number of symbols from the val[] array, without risking marching off
-   * the end of memory.  jxhuff.c will do a more thorough test later.
-   */
-  nsymbols = 0;
-  for (len = 1; len <= 16; len++)
-    nsymbols += bits[len];
-  if (nsymbols > 256)
-    ERREXIT(cinfo, JERR_BAD_HUFF_TABLE);
-
-  if (nsymbols > 0)
-    MEMCOPY(htbl->huffval, val, nsymbols * SIZEOF(UINT8));
-
-  /* Initialize sent_table FALSE so table will be written to JPEG file. */
-  htbl->sent_table = FALSE;
-
-  return htbl;
-}
@@ -2,7 +2,7 @@
 * jcparam.c
 *
 * Copyright (C) 1991-1998, Thomas G. Lane.
- * Modified 2003-2019 by Guido Vollbeding.
+ * Modified 2003-2013 by Guido Vollbeding.
 * This file is part of the Independent JPEG Group's software.
 * For conditions of distribution and use, see the accompanying README file.
 *
@@ -162,23 +162,112 @@ jpeg_set_quality (j_compress_ptr cinfo, int quality, boolean force_baseline)


 /*
- * Reset standard Huffman tables
+ * Huffman table setup routines
 */

 LOCAL(void)
-std_huff_tables (j_compress_ptr cinfo)
+add_huff_table (j_compress_ptr cinfo,
+		JHUFF_TBL **htblptr, const UINT8 *bits, const UINT8 *val)
+/* Define a Huffman table */
 {
-  if (cinfo->dc_huff_tbl_ptrs[0] != NULL)
-    (void) jpeg_std_huff_table((j_common_ptr) cinfo, TRUE, 0);
+  int nsymbols, len;

-  if (cinfo->ac_huff_tbl_ptrs[0] != NULL)
-    (void) jpeg_std_huff_table((j_common_ptr) cinfo, FALSE, 0);
+  if (*htblptr == NULL)
+    *htblptr = jpeg_alloc_huff_table((j_common_ptr) cinfo);

-  if (cinfo->dc_huff_tbl_ptrs[1] != NULL)
-    (void) jpeg_std_huff_table((j_common_ptr) cinfo, TRUE, 1);
+  /* Copy the number-of-symbols-of-each-code-length counts */
+  MEMCOPY((*htblptr)->bits, bits, SIZEOF((*htblptr)->bits));

-  if (cinfo->ac_huff_tbl_ptrs[1] != NULL)
-    (void) jpeg_std_huff_table((j_common_ptr) cinfo, FALSE, 1);
+  /* Validate the counts.  We do this here mainly so we can copy the right
+   * number of symbols from the val[] array, without risking marching off
+   * the end of memory.  jchuff.c will do a more thorough test later.
+   */
+  nsymbols = 0;
+  for (len = 1; len <= 16; len++)
+    nsymbols += bits[len];
+  if (nsymbols < 1 || nsymbols > 256)
+    ERREXIT(cinfo, JERR_BAD_HUFF_TABLE);
+
+  MEMCOPY((*htblptr)->huffval, val, nsymbols * SIZEOF(UINT8));
+
+  /* Initialize sent_table FALSE so table will be written to JPEG file. */
+  (*htblptr)->sent_table = FALSE;
+}
+
+
+LOCAL(void)
+std_huff_tables (j_compress_ptr cinfo)
+/* Set up the standard Huffman tables (cf. JPEG standard section K.3) */
+/* IMPORTANT: these are only valid for 8-bit data precision! */
+{
+  static const UINT8 bits_dc_luminance[17] =
+    { /* 0-base */ 0, 0, 1, 5, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0 };
+  static const UINT8 val_dc_luminance[] =
+    { 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11 };
+  
+  static const UINT8 bits_dc_chrominance[17] =
+    { /* 0-base */ 0, 0, 3, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0 };
+  static const UINT8 val_dc_chrominance[] =
+    { 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11 };
+  
+  static const UINT8 bits_ac_luminance[17] =
+    { /* 0-base */ 0, 0, 2, 1, 3, 3, 2, 4, 3, 5, 5, 4, 4, 0, 0, 1, 0x7d };
+  static const UINT8 val_ac_luminance[] =
+    { 0x01, 0x02, 0x03, 0x00, 0x04, 0x11, 0x05, 0x12,
+      0x21, 0x31, 0x41, 0x06, 0x13, 0x51, 0x61, 0x07,
+      0x22, 0x71, 0x14, 0x32, 0x81, 0x91, 0xa1, 0x08,
+      0x23, 0x42, 0xb1, 0xc1, 0x15, 0x52, 0xd1, 0xf0,
+      0x24, 0x33, 0x62, 0x72, 0x82, 0x09, 0x0a, 0x16,
+      0x17, 0x18, 0x19, 0x1a, 0x25, 0x26, 0x27, 0x28,
+      0x29, 0x2a, 0x34, 0x35, 0x36, 0x37, 0x38, 0x39,
+      0x3a, 0x43, 0x44, 0x45, 0x46, 0x47, 0x48, 0x49,
+      0x4a, 0x53, 0x54, 0x55, 0x56, 0x57, 0x58, 0x59,
+      0x5a, 0x63, 0x64, 0x65, 0x66, 0x67, 0x68, 0x69,
+      0x6a, 0x73, 0x74, 0x75, 0x76, 0x77, 0x78, 0x79,
+      0x7a, 0x83, 0x84, 0x85, 0x86, 0x87, 0x88, 0x89,
+      0x8a, 0x92, 0x93, 0x94, 0x95, 0x96, 0x97, 0x98,
+      0x99, 0x9a, 0xa2, 0xa3, 0xa4, 0xa5, 0xa6, 0xa7,
+      0xa8, 0xa9, 0xaa, 0xb2, 0xb3, 0xb4, 0xb5, 0xb6,
+      0xb7, 0xb8, 0xb9, 0xba, 0xc2, 0xc3, 0xc4, 0xc5,
+      0xc6, 0xc7, 0xc8, 0xc9, 0xca, 0xd2, 0xd3, 0xd4,
+      0xd5, 0xd6, 0xd7, 0xd8, 0xd9, 0xda, 0xe1, 0xe2,
+      0xe3, 0xe4, 0xe5, 0xe6, 0xe7, 0xe8, 0xe9, 0xea,
+      0xf1, 0xf2, 0xf3, 0xf4, 0xf5, 0xf6, 0xf7, 0xf8,
+      0xf9, 0xfa };
+  
+  static const UINT8 bits_ac_chrominance[17] =
+    { /* 0-base */ 0, 0, 2, 1, 2, 4, 4, 3, 4, 7, 5, 4, 4, 0, 1, 2, 0x77 };
+  static const UINT8 val_ac_chrominance[] =
+    { 0x00, 0x01, 0x02, 0x03, 0x11, 0x04, 0x05, 0x21,
+      0x31, 0x06, 0x12, 0x41, 0x51, 0x07, 0x61, 0x71,
+      0x13, 0x22, 0x32, 0x81, 0x08, 0x14, 0x42, 0x91,
+      0xa1, 0xb1, 0xc1, 0x09, 0x23, 0x33, 0x52, 0xf0,
+      0x15, 0x62, 0x72, 0xd1, 0x0a, 0x16, 0x24, 0x34,
+      0xe1, 0x25, 0xf1, 0x17, 0x18, 0x19, 0x1a, 0x26,
+      0x27, 0x28, 0x29, 0x2a, 0x35, 0x36, 0x37, 0x38,
+      0x39, 0x3a, 0x43, 0x44, 0x45, 0x46, 0x47, 0x48,
+      0x49, 0x4a, 0x53, 0x54, 0x55, 0x56, 0x57, 0x58,
+      0x59, 0x5a, 0x63, 0x64, 0x65, 0x66, 0x67, 0x68,
+      0x69, 0x6a, 0x73, 0x74, 0x75, 0x76, 0x77, 0x78,
+      0x79, 0x7a, 0x82, 0x83, 0x84, 0x85, 0x86, 0x87,
+      0x88, 0x89, 0x8a, 0x92, 0x93, 0x94, 0x95, 0x96,
+      0x97, 0x98, 0x99, 0x9a, 0xa2, 0xa3, 0xa4, 0xa5,
+      0xa6, 0xa7, 0xa8, 0xa9, 0xaa, 0xb2, 0xb3, 0xb4,
+      0xb5, 0xb6, 0xb7, 0xb8, 0xb9, 0xba, 0xc2, 0xc3,
+      0xc4, 0xc5, 0xc6, 0xc7, 0xc8, 0xc9, 0xca, 0xd2,
+      0xd3, 0xd4, 0xd5, 0xd6, 0xd7, 0xd8, 0xd9, 0xda,
+      0xe2, 0xe3, 0xe4, 0xe5, 0xe6, 0xe7, 0xe8, 0xe9,
+      0xea, 0xf2, 0xf3, 0xf4, 0xf5, 0xf6, 0xf7, 0xf8,
+      0xf9, 0xfa };
+  
+  add_huff_table(cinfo, &cinfo->dc_huff_tbl_ptrs[0],
+		 bits_dc_luminance, val_dc_luminance);
+  add_huff_table(cinfo, &cinfo->ac_huff_tbl_ptrs[0],
+		 bits_ac_luminance, val_ac_luminance);
+  add_huff_table(cinfo, &cinfo->dc_huff_tbl_ptrs[1],
+		 bits_dc_chrominance, val_dc_chrominance);
+  add_huff_table(cinfo, &cinfo->ac_huff_tbl_ptrs[1],
+		 bits_ac_chrominance, val_ac_chrominance);
 }


@@ -217,7 +306,7 @@ jpeg_set_defaults (j_compress_ptr cinfo)
  cinfo->data_precision = BITS_IN_JSAMPLE;
  /* Set up two quantization tables using default quality of 75 */
  jpeg_set_quality(cinfo, 75, TRUE);
-  /* Reset standard Huffman tables */
+  /* Set up two Huffman tables */
  std_huff_tables(cinfo);

  /* Initialize default arithmetic coding conditioning */
@@ -1,7 +1,7 @@
 /*
 * jdarith.c
 *
- * Developed 1997-2019 by Guido Vollbeding.
+ * Developed 1997-2015 by Guido Vollbeding.
 * This file is part of the Independent JPEG Group's software.
 * For conditions of distribution and use, see the accompanying README file.
 *
@@ -280,7 +280,7 @@ decode_mcu_DC_first (j_decompress_ptr cinfo, JBLOCKROW *MCU_data)
      if ((m = arith_decode(cinfo, st)) != 0) {
 	st = entropy->dc_stats[tbl] + 20;	/* Table F.4: X1 = 20 */
 	while (arith_decode(cinfo, st)) {
-	  if ((m <<= 1) == (int) 0x8000U) {
+	  if ((m <<= 1) == 0x8000) {
 	    WARNMS(cinfo, JWRN_ARITH_BAD_CODE);
 	    entropy->ct = -1;			/* magnitude overflow */
 	    return TRUE;
@@ -370,7 +370,7 @@ decode_mcu_AC_first (j_decompress_ptr cinfo, JBLOCKROW *MCU_data)
 	st = entropy->ac_stats[tbl] +
 	     (k <= cinfo->arith_ac_K[tbl] ? 189 : 217);
 	while (arith_decode(cinfo, st)) {
-	  if ((m <<= 1) == (int) 0x8000U) {
+	  if ((m <<= 1) == 0x8000) {
 	    WARNMS(cinfo, JWRN_ARITH_BAD_CODE);
 	    entropy->ct = -1;			/* magnitude overflow */
 	    return TRUE;
@@ -404,8 +404,7 @@ decode_mcu_DC_refine (j_decompress_ptr cinfo, JBLOCKROW *MCU_data)
 {
  arith_entropy_ptr entropy = (arith_entropy_ptr) cinfo->entropy;
  unsigned char *st;
-  JCOEF p1;
-  int blkn;
+  int p1, blkn;

  /* Process restart marker if needed */
  if (cinfo->restart_interval) {
@@ -441,7 +440,7 @@ decode_mcu_AC_refine (j_decompress_ptr cinfo, JBLOCKROW *MCU_data)
  JCOEFPTR thiscoef;
  unsigned char *st;
  int tbl, k, kex;
-  JCOEF p1, m1;
+  int p1, m1;
  const int * natural_order;

  /* Process restart marker if needed */
@@ -460,7 +459,7 @@ decode_mcu_AC_refine (j_decompress_ptr cinfo, JBLOCKROW *MCU_data)
  tbl = cinfo->cur_comp_info[0]->ac_tbl_no;

  p1 = 1 << cinfo->Al;		/* 1 in the bit position being coded */
-  m1 = -p1;			/* -1 in the bit position being coded */
+  m1 = (-1) << cinfo->Al;	/* -1 in the bit position being coded */

  /* Establish EOBx (previous stage end-of-block) index */
  kex = cinfo->Se;
@@ -556,7 +555,7 @@ decode_mcu (j_decompress_ptr cinfo, JBLOCKROW *MCU_data)
      if ((m = arith_decode(cinfo, st)) != 0) {
 	st = entropy->dc_stats[tbl] + 20;	/* Table F.4: X1 = 20 */
 	while (arith_decode(cinfo, st)) {
-	  if ((m <<= 1) == (int) 0x8000U) {
+	  if ((m <<= 1) == 0x8000) {
 	    WARNMS(cinfo, JWRN_ARITH_BAD_CODE);
 	    entropy->ct = -1;			/* magnitude overflow */
 	    return TRUE;
@@ -613,7 +612,7 @@ decode_mcu (j_decompress_ptr cinfo, JBLOCKROW *MCU_data)
 	  st = entropy->ac_stats[tbl] +
 	       (k <= cinfo->arith_ac_K[tbl] ? 189 : 217);
 	  while (arith_decode(cinfo, st)) {
-	    if ((m <<= 1) == (int) 0x8000U) {
+	    if ((m <<= 1) == 0x8000) {
 	      WARNMS(cinfo, JWRN_ARITH_BAD_CODE);
 	      entropy->ct = -1;			/* magnitude overflow */
 	      return TRUE;
@@ -767,8 +766,9 @@ jinit_arith_decoder (j_decompress_ptr cinfo)
  arith_entropy_ptr entropy;
  int i;

-  entropy = (arith_entropy_ptr) (*cinfo->mem->alloc_small)
-    ((j_common_ptr) cinfo, JPOOL_IMAGE, SIZEOF(arith_entropy_decoder));
+  entropy = (arith_entropy_ptr)
+    (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				SIZEOF(arith_entropy_decoder));
  cinfo->entropy = &entropy->pub;
  entropy->pub.start_pass = start_pass;
  entropy->pub.finish_pass = finish_pass;
@@ -785,9 +785,9 @@ jinit_arith_decoder (j_decompress_ptr cinfo)
  if (cinfo->progressive_mode) {
    /* Create progression status table */
    int *coef_bit_ptr, ci;
-    cinfo->coef_bits = (int (*)[DCTSIZE2]) (*cinfo->mem->alloc_small)
-      ((j_common_ptr) cinfo, JPOOL_IMAGE,
-       cinfo->num_components * DCTSIZE2 * SIZEOF(int));
+    cinfo->coef_bits = (int (*)[DCTSIZE2])
+      (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				  cinfo->num_components*DCTSIZE2*SIZEOF(int));
    coef_bit_ptr = & cinfo->coef_bits[0][0];
    for (ci = 0; ci < cinfo->num_components; ci++) 
      for (i = 0; i < DCTSIZE2; i++)
@@ -2,7 +2,7 @@
 * jdatadst.c
 *
 * Copyright (C) 1994-1996, Thomas G. Lane.
- * Modified 2009-2019 by Guido Vollbeding.
+ * Modified 2009-2017 by Guido Vollbeding.
 * This file is part of the Independent JPEG Group's software.
 * For conditions of distribution and use, see the accompanying README file.
 *
@@ -46,7 +46,7 @@ typedef struct {
  struct jpeg_destination_mgr pub; /* public fields */

  unsigned char ** outbuffer;	/* target buffer */
-  size_t * outsize;
+  unsigned long * outsize;
  unsigned char * newbuffer;	/* newly allocated buffer */
  JOCTET * buffer;		/* start of buffer */
  size_t bufsize;
@@ -66,8 +66,9 @@ init_destination (j_compress_ptr cinfo)
  my_dest_ptr dest = (my_dest_ptr) cinfo->dest;

  /* Allocate the output buffer --- it will be released when done with image */
-  dest->buffer = (JOCTET *) (*cinfo->mem->alloc_small)
-    ((j_common_ptr) cinfo, JPOOL_IMAGE, OUTPUT_BUF_SIZE * SIZEOF(JOCTET));
+  dest->buffer = (JOCTET *)
+      (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				  OUTPUT_BUF_SIZE * SIZEOF(JOCTET));

  dest->pub.next_output_byte = dest->buffer;
  dest->pub.free_in_buffer = OUTPUT_BUF_SIZE;
@@ -130,7 +131,7 @@ empty_mem_output_buffer (j_compress_ptr cinfo)
  nextbuffer = (JOCTET *) malloc(nextsize);

  if (nextbuffer == NULL)
-    ERREXIT1(cinfo, JERR_OUT_OF_MEMORY, 11);
+    ERREXIT1(cinfo, JERR_OUT_OF_MEMORY, 10);

  MEMCOPY(nextbuffer, dest->buffer, dest->bufsize);

@@ -203,8 +204,9 @@ jpeg_stdio_dest (j_compress_ptr cinfo, FILE * outfile)
   * sizes may be different.  Caveat programmer.
   */
  if (cinfo->dest == NULL) {	/* first time for this JPEG object? */
-    cinfo->dest = (struct jpeg_destination_mgr *) (*cinfo->mem->alloc_small)
-      ((j_common_ptr) cinfo, JPOOL_PERMANENT, SIZEOF(my_destination_mgr));
+    cinfo->dest = (struct jpeg_destination_mgr *)
+      (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_PERMANENT,
+				  SIZEOF(my_destination_mgr));
  }

  dest = (my_dest_ptr) cinfo->dest;
@@ -231,7 +233,7 @@ jpeg_stdio_dest (j_compress_ptr cinfo, FILE * outfile)

 GLOBAL(void)
 jpeg_mem_dest (j_compress_ptr cinfo,
-	       unsigned char ** outbuffer, size_t * outsize)
+	       unsigned char ** outbuffer, unsigned long * outsize)
 {
  my_mem_dest_ptr dest;

@@ -242,8 +244,9 @@ jpeg_mem_dest (j_compress_ptr cinfo,
   * can be written to the same buffer without re-executing jpeg_mem_dest.
   */
  if (cinfo->dest == NULL) {	/* first time for this JPEG object? */
-    cinfo->dest = (struct jpeg_destination_mgr *) (*cinfo->mem->alloc_small)
-      ((j_common_ptr) cinfo, JPOOL_PERMANENT, SIZEOF(my_mem_destination_mgr));
+    cinfo->dest = (struct jpeg_destination_mgr *)
+      (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_PERMANENT,
+				  SIZEOF(my_mem_destination_mgr));
  }

  dest = (my_mem_dest_ptr) cinfo->dest;
@@ -2,7 +2,7 @@
 * jdatasrc.c
 *
 * Copyright (C) 1994-1996, Thomas G. Lane.
- * Modified 2009-2019 by Guido Vollbeding.
+ * Modified 2009-2015 by Guido Vollbeding.
 * This file is part of the Independent JPEG Group's software.
 * For conditions of distribution and use, see the accompanying README file.
 *
@@ -156,23 +156,21 @@ METHODDEF(void)
 skip_input_data (j_decompress_ptr cinfo, long num_bytes)
 {
  struct jpeg_source_mgr * src = cinfo->src;
-  size_t nbytes;

  /* Just a dumb implementation for now.  Could use fseek() except
   * it doesn't work on pipes.  Not clear that being smart is worth
   * any trouble anyway --- large skips are infrequent.
   */
  if (num_bytes > 0) {
-    nbytes = (size_t) num_bytes;
-    while (nbytes > src->bytes_in_buffer) {
-      nbytes -= src->bytes_in_buffer;
+    while (num_bytes > (long) src->bytes_in_buffer) {
+      num_bytes -= (long) src->bytes_in_buffer;
      (void) (*src->fill_input_buffer) (cinfo);
      /* note we assume that fill_input_buffer will never return FALSE,
       * so suspension need not be handled.
       */
    }
-    src->next_input_byte += nbytes;
-    src->bytes_in_buffer -= nbytes;
+    src->next_input_byte += (size_t) num_bytes;
+    src->bytes_in_buffer -= (size_t) num_bytes;
  }
 }

@@ -221,11 +219,13 @@ jpeg_stdio_src (j_decompress_ptr cinfo, FILE * infile)
   * manager serially with the same JPEG object.  Caveat programmer.
   */
  if (cinfo->src == NULL) {	/* first time for this JPEG object? */
-    cinfo->src = (struct jpeg_source_mgr *) (*cinfo->mem->alloc_small)
-      ((j_common_ptr) cinfo, JPOOL_PERMANENT, SIZEOF(my_source_mgr));
+    cinfo->src = (struct jpeg_source_mgr *)
+      (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_PERMANENT,
+				  SIZEOF(my_source_mgr));
    src = (my_src_ptr) cinfo->src;
-    src->buffer = (JOCTET *) (*cinfo->mem->alloc_small)
-      ((j_common_ptr) cinfo, JPOOL_PERMANENT, INPUT_BUF_SIZE * SIZEOF(JOCTET));
+    src->buffer = (JOCTET *)
+      (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_PERMANENT,
+				  INPUT_BUF_SIZE * SIZEOF(JOCTET));
  }

  src = (my_src_ptr) cinfo->src;
@@ -247,7 +247,7 @@ jpeg_stdio_src (j_decompress_ptr cinfo, FILE * infile)

 GLOBAL(void)
 jpeg_mem_src (j_decompress_ptr cinfo,
-	      const unsigned char * inbuffer, size_t insize)
+	      const unsigned char * inbuffer, unsigned long insize)
 {
  struct jpeg_source_mgr * src;

@@ -259,8 +259,9 @@ jpeg_mem_src (j_decompress_ptr cinfo,
   * the first one.
   */
  if (cinfo->src == NULL) {	/* first time for this JPEG object? */
-    cinfo->src = (struct jpeg_source_mgr *) (*cinfo->mem->alloc_small)
-      ((j_common_ptr) cinfo, JPOOL_PERMANENT, SIZEOF(struct jpeg_source_mgr));
+    cinfo->src = (struct jpeg_source_mgr *)
+      (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_PERMANENT,
+				  SIZEOF(struct jpeg_source_mgr));
  }

  src = cinfo->src;
@@ -269,6 +270,6 @@ jpeg_mem_src (j_decompress_ptr cinfo,
  src->skip_input_data = skip_input_data;
  src->resync_to_restart = jpeg_resync_to_restart; /* use default method */
  src->term_source = term_source;
-  src->bytes_in_buffer = insize;
+  src->bytes_in_buffer = (size_t) insize;
  src->next_input_byte = (const JOCTET *) inbuffer;
 }
@@ -2,7 +2,7 @@
 * jdcolor.c
 *
 * Copyright (C) 1991-1997, Thomas G. Lane.
- * Modified 2011-2019 by Guido Vollbeding.
+ * Modified 2011-2017 by Guido Vollbeding.
 * This file is part of the Independent JPEG Group's software.
 * For conditions of distribution and use, see the accompanying README file.
 *
@@ -124,22 +124,28 @@ build_ycc_rgb_table (j_decompress_ptr cinfo)
  INT32 x;
  SHIFT_TEMPS

-  cconvert->Cr_r_tab = (int *) (*cinfo->mem->alloc_small)
-    ((j_common_ptr) cinfo, JPOOL_IMAGE, (MAXJSAMPLE+1) * SIZEOF(int));
-  cconvert->Cb_b_tab = (int *) (*cinfo->mem->alloc_small)
-    ((j_common_ptr) cinfo, JPOOL_IMAGE, (MAXJSAMPLE+1) * SIZEOF(int));
-  cconvert->Cr_g_tab = (INT32 *) (*cinfo->mem->alloc_small)
-    ((j_common_ptr) cinfo, JPOOL_IMAGE, (MAXJSAMPLE+1) * SIZEOF(INT32));
-  cconvert->Cb_g_tab = (INT32 *) (*cinfo->mem->alloc_small)
-    ((j_common_ptr) cinfo, JPOOL_IMAGE, (MAXJSAMPLE+1) * SIZEOF(INT32));
+  cconvert->Cr_r_tab = (int *)
+    (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				(MAXJSAMPLE+1) * SIZEOF(int));
+  cconvert->Cb_b_tab = (int *)
+    (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				(MAXJSAMPLE+1) * SIZEOF(int));
+  cconvert->Cr_g_tab = (INT32 *)
+    (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				(MAXJSAMPLE+1) * SIZEOF(INT32));
+  cconvert->Cb_g_tab = (INT32 *)
+    (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				(MAXJSAMPLE+1) * SIZEOF(INT32));

  for (i = 0, x = -CENTERJSAMPLE; i <= MAXJSAMPLE; i++, x++) {
    /* i is the actual input pixel value, in the range 0..MAXJSAMPLE */
    /* The Cb or Cr value we are thinking of is x = i - CENTERJSAMPLE */
    /* Cr=>R value is nearest int to 1.402 * x */
-    cconvert->Cr_r_tab[i] = (int) DESCALE(FIX(1.402) * x, SCALEBITS);
+    cconvert->Cr_r_tab[i] = (int)
+		    RIGHT_SHIFT(FIX(1.402) * x + ONE_HALF, SCALEBITS);
    /* Cb=>B value is nearest int to 1.772 * x */
-    cconvert->Cb_b_tab[i] = (int) DESCALE(FIX(1.772) * x, SCALEBITS);
+    cconvert->Cb_b_tab[i] = (int)
+		    RIGHT_SHIFT(FIX(1.772) * x + ONE_HALF, SCALEBITS);
    /* Cr=>G value is scaled-up -0.714136286 * x */
    cconvert->Cr_g_tab[i] = (- FIX(0.714136286)) * x;
    /* Cb=>G value is scaled-up -0.344136286 * x */
@@ -158,22 +164,28 @@ build_bg_ycc_rgb_table (j_decompress_ptr cinfo)
  INT32 x;
  SHIFT_TEMPS

-  cconvert->Cr_r_tab = (int *) (*cinfo->mem->alloc_small)
-    ((j_common_ptr) cinfo, JPOOL_IMAGE, (MAXJSAMPLE+1) * SIZEOF(int));
-  cconvert->Cb_b_tab = (int *) (*cinfo->mem->alloc_small)
-    ((j_common_ptr) cinfo, JPOOL_IMAGE, (MAXJSAMPLE+1) * SIZEOF(int));
-  cconvert->Cr_g_tab = (INT32 *) (*cinfo->mem->alloc_small)
-    ((j_common_ptr) cinfo, JPOOL_IMAGE, (MAXJSAMPLE+1) * SIZEOF(INT32));
-  cconvert->Cb_g_tab = (INT32 *) (*cinfo->mem->alloc_small)
-    ((j_common_ptr) cinfo, JPOOL_IMAGE, (MAXJSAMPLE+1) * SIZEOF(INT32));
+  cconvert->Cr_r_tab = (int *)
+    (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				(MAXJSAMPLE+1) * SIZEOF(int));
+  cconvert->Cb_b_tab = (int *)
+    (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				(MAXJSAMPLE+1) * SIZEOF(int));
+  cconvert->Cr_g_tab = (INT32 *)
+    (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				(MAXJSAMPLE+1) * SIZEOF(INT32));
+  cconvert->Cb_g_tab = (INT32 *)
+    (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				(MAXJSAMPLE+1) * SIZEOF(INT32));

  for (i = 0, x = -CENTERJSAMPLE; i <= MAXJSAMPLE; i++, x++) {
    /* i is the actual input pixel value, in the range 0..MAXJSAMPLE */
    /* The Cb or Cr value we are thinking of is x = i - CENTERJSAMPLE */
    /* Cr=>R value is nearest int to 2.804 * x */
-    cconvert->Cr_r_tab[i] = (int) DESCALE(FIX(2.804) * x, SCALEBITS);
+    cconvert->Cr_r_tab[i] = (int)
+		    RIGHT_SHIFT(FIX(2.804) * x + ONE_HALF, SCALEBITS);
    /* Cb=>B value is nearest int to 3.544 * x */
-    cconvert->Cb_b_tab[i] = (int) DESCALE(FIX(3.544) * x, SCALEBITS);
+    cconvert->Cb_b_tab[i] = (int)
+		    RIGHT_SHIFT(FIX(3.544) * x + ONE_HALF, SCALEBITS);
    /* Cr=>G value is scaled-up -1.428272572 * x */
    cconvert->Cr_g_tab[i] = (- FIX(1.428272572)) * x;
    /* Cb=>G value is scaled-up -0.688272572 * x */
@@ -189,7 +201,6 @@ build_bg_ycc_rgb_table (j_decompress_ptr cinfo)
 * Note that we change from noninterleaved, one-plane-per-component format
 * to interleaved-pixel format.  The output buffer is therefore three times
 * as wide as the input buffer.
- *
 * A starting row offset is provided only for the input buffer.  The caller
 * can easily adjust the passed output_buf value to accommodate any row
 * offset required on that side.
@@ -253,8 +264,9 @@ build_rgb_y_table (j_decompress_ptr cinfo)
  INT32 i;

  /* Allocate and fill in the conversion tables. */
-  cconvert->rgb_y_tab = rgb_y_tab = (INT32 *) (*cinfo->mem->alloc_small)
-    ((j_common_ptr) cinfo, JPOOL_IMAGE, TABLE_SIZE * SIZEOF(INT32));
+  cconvert->rgb_y_tab = rgb_y_tab = (INT32 *)
+    (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				(TABLE_SIZE * SIZEOF(INT32)));

  for (i = 0; i <= MAXJSAMPLE; i++) {
    rgb_y_tab[i+R_Y_OFF] = FIX(0.299) * i;
@@ -274,8 +286,8 @@ rgb_gray_convert (j_decompress_ptr cinfo,
 		  JSAMPARRAY output_buf, int num_rows)
 {
  my_cconvert_ptr cconvert = (my_cconvert_ptr) cinfo->cconvert;
-  register int r, g, b;
  register INT32 * ctab = cconvert->rgb_y_tab;
+  register int r, g, b;
  register JSAMPROW outptr;
  register JSAMPROW inptr0, inptr1, inptr2;
  register JDIMENSION col;
@@ -301,7 +313,6 @@ rgb_gray_convert (j_decompress_ptr cinfo,


 /*
- * Convert some rows of samples to the output colorspace.
 * [R-G,G,B-G] to [R,G,B] conversion with modulo calculation
 * (inverse color transform).
 * This can be seen as an adaption of the general YCbCr->RGB
@@ -353,8 +364,8 @@ rgb1_gray_convert (j_decompress_ptr cinfo,
 		   JSAMPARRAY output_buf, int num_rows)
 {
  my_cconvert_ptr cconvert = (my_cconvert_ptr) cinfo->cconvert;
-  register int r, g, b;
  register INT32 * ctab = cconvert->rgb_y_tab;
+  register int r, g, b;
  register JSAMPROW outptr;
  register JSAMPROW inptr0, inptr1, inptr2;
  register JDIMENSION col;
@@ -385,7 +396,6 @@ rgb1_gray_convert (j_decompress_ptr cinfo,


 /*
- * Convert some rows of samples to the output colorspace.
 * No colorspace change, but conversion from separate-planes
 * to interleaved representation.
 */
@@ -420,7 +430,6 @@ rgb_convert (j_decompress_ptr cinfo,
 /*
 * Color conversion for no colorspace change: just copy the data,
 * converting from separate-planes to interleaved representation.
- * We assume out_color_components == num_components.
 */

 METHODDEF(void)
@@ -428,21 +437,20 @@ null_convert (j_decompress_ptr cinfo,
 	      JSAMPIMAGE input_buf, JDIMENSION input_row,
 	      JSAMPARRAY output_buf, int num_rows)
 {
+  int ci;
+  register int nc = cinfo->num_components;
  register JSAMPROW outptr;
  register JSAMPROW inptr;
-  register JDIMENSION count;
-  register int num_comps = cinfo->num_components;
+  register JDIMENSION col;
  JDIMENSION num_cols = cinfo->output_width;
-  int ci;

  while (--num_rows >= 0) {
-    /* It seems fastest to make a separate pass for each component. */
-    for (ci = 0; ci < num_comps; ci++) {
+    for (ci = 0; ci < nc; ci++) {
      inptr = input_buf[ci][input_row];
      outptr = output_buf[0] + ci;
-      for (count = num_cols; count > 0; count--) {
-	*outptr = *inptr++;	/* don't need GETJSAMPLE() here */
-	outptr += num_comps;
+      for (col = 0; col < num_cols; col++) {
+	*outptr = *inptr++;	/* needn't bother with GETJSAMPLE() here */
+	outptr += nc;
      }
    }
    input_row++;
@@ -496,10 +504,9 @@ gray_rgb_convert (j_decompress_ptr cinfo,


 /*
- * Convert some rows of samples to the output colorspace.
- * This version handles Adobe-style YCCK->CMYK conversion,
- * where we convert YCbCr to R=1-C, G=1-M, and B=1-Y using the
- * same conversion as above, while passing K (black) unchanged.
+ * Adobe-style YCCK->CMYK conversion.
+ * We convert YCbCr to R=1-C, G=1-M, and B=1-Y using the same
+ * conversion as above, while passing K (black) unchanged.
 * We assume build_ycc_rgb_table has been called.
 */

@@ -570,8 +577,9 @@ jinit_color_deconverter (j_decompress_ptr cinfo)
  my_cconvert_ptr cconvert;
  int ci;

-  cconvert = (my_cconvert_ptr) (*cinfo->mem->alloc_small)
-    ((j_common_ptr) cinfo, JPOOL_IMAGE, SIZEOF(my_color_deconverter));
+  cconvert = (my_cconvert_ptr)
+    (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				SIZEOF(my_color_deconverter));
  cinfo->cconvert = &cconvert->pub;
  cconvert->pub.start_pass = start_pass_dcolor;

@@ -599,6 +607,7 @@ jinit_color_deconverter (j_decompress_ptr cinfo)
  default:			/* JCS_UNKNOWN can be anything */
    if (cinfo->num_components < 1)
      ERREXIT(cinfo, JERR_BAD_J_COLORSPACE);
+    break;
  }

  /* Support color transform only for RGB colorspaces */
@@ -675,18 +684,19 @@ jinit_color_deconverter (j_decompress_ptr cinfo)

  case JCS_BG_RGB:
    cinfo->out_color_components = RGB_PIXELSIZE;
-    if (cinfo->jpeg_color_space != JCS_BG_RGB)
+    if (cinfo->jpeg_color_space == JCS_BG_RGB) {
+      switch (cinfo->color_transform) {
+      case JCT_NONE:
+	cconvert->pub.color_convert = rgb_convert;
+	break;
+      case JCT_SUBTRACT_GREEN:
+	cconvert->pub.color_convert = rgb1_rgb_convert;
+	break;
+      default:
+	ERREXIT(cinfo, JERR_CONVERSION_NOTIMPL);
+      }
+    } else
      ERREXIT(cinfo, JERR_CONVERSION_NOTIMPL);
-    switch (cinfo->color_transform) {
-    case JCT_NONE:
-      cconvert->pub.color_convert = rgb_convert;
-      break;
-    case JCT_SUBTRACT_GREEN:
-      cconvert->pub.color_convert = rgb1_rgb_convert;
-      break;
-    default:
-      ERREXIT(cinfo, JERR_CONVERSION_NOTIMPL);
-    }
    break;

  case JCS_CMYK:
@@ -704,12 +714,14 @@ jinit_color_deconverter (j_decompress_ptr cinfo)
    }
    break;

-  default:		/* permit null conversion to same output space */
-    if (cinfo->out_color_space != cinfo->jpeg_color_space)
-      /* unsupported non-null conversion */
+  default:
+    /* Permit null conversion to same output space */
+    if (cinfo->out_color_space == cinfo->jpeg_color_space) {
+      cinfo->out_color_components = cinfo->num_components;
+      cconvert->pub.color_convert = null_convert;
+    } else			/* unsupported non-null conversion */
      ERREXIT(cinfo, JERR_CONVERSION_NOTIMPL);
-    cinfo->out_color_components = cinfo->num_components;
-    cconvert->pub.color_convert = null_convert;
+    break;
  }

  if (cinfo->quantize_colors)
--- a/Show More
+++ b/Show More