release: OpenCV 3.4.17

Merge pull request #21341 from alalek:issue_20930_3.4
python: use '((x,y), (w,h), angle)' in std::vector<RotatedRect>
2026-07-21 19:33:03 +04:00 · 2021-12-24 16:39:15 +00:00 · 2021-12-24 16:33:19 +00:00 · 2021-12-24 15:01:45 +00:00 · 2021-12-24 11:56:12 +00:00 · 2021-12-24 11:55:55 +00:00
562 changed files with 28067 additions and 9293 deletions
@@ -1778,30 +1778,30 @@ TegraCvtColor_Invoker(bgrx2hsvf, bgrx2hsv, src_data + static_cast<size_t>(range.
    : CV_HAL_ERROR_NOT_IMPLEMENTED \
 )

-#define TEGRA_CVT2PYUVTOBGR(src_data, src_step, dst_data, dst_step, dst_width, dst_height, dcn, swapBlue, uIdx) \
+#define TEGRA_CVT2PYUVTOBGR_EX(y_data, y_step, uv_data, uv_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), \
-                                             src_data, src_step, \
-                                             src_data + src_step * dst_height, src_step, \
+                                             y_data, y_step, \
+                                             uv_data, uv_step, \
                                             dst_data, dst_step) : \
                    CAROTENE_NS::yuv420i2bgr(CAROTENE_NS::Size2D(dst_width, dst_height), \
-                                             src_data, src_step, \
-                                             src_data + src_step * dst_height, src_step, \
+                                             y_data, y_step, \
+                                             uv_data, uv_step, \
                                             dst_data, dst_step)), \
                CV_HAL_ERROR_OK : \
            uIdx == 1 ? \
                (swapBlue ? \
                    CAROTENE_NS::yuv420sp2rgb(CAROTENE_NS::Size2D(dst_width, dst_height), \
-                                              src_data, src_step, \
-                                              src_data + src_step * dst_height, src_step, \
+                                              y_data, y_step, \
+                                              uv_data, uv_step, \
                                              dst_data, dst_step) : \
                    CAROTENE_NS::yuv420sp2bgr(CAROTENE_NS::Size2D(dst_width, dst_height), \
-                                              src_data, src_step, \
-                                              src_data + src_step * dst_height, src_step, \
+                                              y_data, y_step, \
+                                              uv_data, uv_step, \
                                              dst_data, dst_step)), \
                CV_HAL_ERROR_OK : \
            CV_HAL_ERROR_NOT_IMPLEMENTED : \
@@ -1809,29 +1809,32 @@ TegraCvtColor_Invoker(bgrx2hsvf, bgrx2hsv, src_data + static_cast<size_t>(range.
            uIdx == 0 ? \
                (swapBlue ? \
                    CAROTENE_NS::yuv420i2rgbx(CAROTENE_NS::Size2D(dst_width, dst_height), \
-                                              src_data, src_step, \
-                                              src_data + src_step * dst_height, src_step, \
+                                              y_data, y_step, \
+                                              uv_data, uv_step, \
                                              dst_data, dst_step) : \
                    CAROTENE_NS::yuv420i2bgrx(CAROTENE_NS::Size2D(dst_width, dst_height), \
-                                              src_data, src_step, \
-                                              src_data + src_step * dst_height, src_step, \
+                                              y_data, y_step, \
+                                              uv_data, uv_step, \
                                              dst_data, dst_step)), \
                CV_HAL_ERROR_OK : \
            uIdx == 1 ? \
                (swapBlue ? \
                    CAROTENE_NS::yuv420sp2rgbx(CAROTENE_NS::Size2D(dst_width, dst_height), \
-                                               src_data, src_step, \
-                                               src_data + src_step * dst_height, src_step, \
+                                               y_data, y_step, \
+                                               uv_data, uv_step, \
                                               dst_data, dst_step) : \
                    CAROTENE_NS::yuv420sp2bgrx(CAROTENE_NS::Size2D(dst_width, dst_height), \
-                                               src_data, src_step, \
-                                               src_data + src_step * dst_height, src_step, \
+                                               y_data, y_step, \
+                                               uv_data, uv_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
@@ -1841,12 +1844,18 @@ 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

@@ -1,8 +1,8 @@
-# Binaries branch name: ffmpeg/3.4_20200907
-# Binaries were created for OpenCV: 03bee14372f5537daa56c62e771ec16181ca1f98
-ocv_update(FFMPEG_BINARIES_COMMIT "2a96257b743695a47f8012aab1ffb995a1dee8b4")
-ocv_update(FFMPEG_FILE_HASH_BIN32 "5e68a3ff82f43ac6524e50e448a34c9c")
-ocv_update(FFMPEG_FILE_HASH_BIN64 "205db629d893e7d4865fd1459807ff47")
+# Binaries branch name: ffmpeg/3.4_20211220
+# Binaries were created for OpenCV: a22dd28e0272ec0f1cfee8811d3f5f0392827c65
+ocv_update(FFMPEG_BINARIES_COMMIT "5a7644ec3940c6eed41c6ebb5a0602a5615fdb3f")
+ocv_update(FFMPEG_FILE_HASH_BIN32 "8ad9de6f1f2ca77786748d1f3a4e83ea")
+ocv_update(FFMPEG_FILE_HASH_BIN64 "2c670f068252e7cd28d3883993dc1d6e")
 ocv_update(FFMPEG_FILE_HASH_CMAKE "3b90f67f4b429e77d3da36698cef700c")

 function(download_win_ffmpeg script_var)
@@ -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 0)
-set(VERSION_REVISION 6)
+set(VERSION_MINOR 1)
+set(VERSION_REVISION 2)
 set(VERSION ${VERSION_MAJOR}.${VERSION_MINOR}.${VERSION_REVISION})
-set(LIBJPEG_TURBO_VERSION_NUMBER 2000006)
+set(LIBJPEG_TURBO_VERSION_NUMBER 2001002)

 string(TIMESTAMP BUILD "opencv-${OPENCV_VERSION}-libjpeg-turbo")
 if(CMAKE_BUILD_TYPE STREQUAL "Debug")
@@ -46,7 +46,6 @@ 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()


@@ -91,7 +91,7 @@ best of our understanding.
 The Modified (3-clause) BSD License
 ===================================

-Copyright (C)2009-2020 D. R. Commander.  All Rights Reserved.
+Copyright (C)2009-2021 D. R. Commander.  All Rights Reserved.<br>
 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) 1991-2016, Thomas G. Lane, Guido Vollbeding.
+This software is copyright (C) 1991-2020, Thomas G. Lane, Guido Vollbeding.
 All Rights Reserved except as specified below.

 Permission is hereby granted to use, copy, modify, and distribute this
@@ -159,19 +159,6 @@ 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
 ==========

@@ -3,7 +3,7 @@ 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 and x86-64
+MIPS systems, as well as progressive JPEG compression on x86, x86-64, and Arm
 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
@@ -61,11 +61,6 @@
   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 */

@@ -18,7 +18,6 @@
 #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 */
@@ -40,3 +40,13 @@
 #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 = GETJSAMPLE(inptr[RGB_RED]);
-      g = GETJSAMPLE(inptr[RGB_GREEN]);
-      b = GETJSAMPLE(inptr[RGB_BLUE]);
+      r = inptr[RGB_RED];
+      g = inptr[RGB_GREEN];
+      b = 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 = GETJSAMPLE(inptr[RGB_RED]);
-      g = GETJSAMPLE(inptr[RGB_GREEN]);
-      b = GETJSAMPLE(inptr[RGB_BLUE]);
+      r = inptr[RGB_RED];
+      g = inptr[RGB_GREEN];
+      b = 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] = GETJSAMPLE(inptr[RGB_RED]);
-      outptr1[col] = GETJSAMPLE(inptr[RGB_GREEN]);
-      outptr2[col] = GETJSAMPLE(inptr[RGB_BLUE]);
+      outptr0[col] = inptr[RGB_RED];
+      outptr1[col] = inptr[RGB_GREEN];
+      outptr2[col] = 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 - GETJSAMPLE(inptr[0]);
-      g = MAXJSAMPLE - GETJSAMPLE(inptr[1]);
-      b = MAXJSAMPLE - GETJSAMPLE(inptr[2]);
+      r = MAXJSAMPLE - inptr[0];
+      g = MAXJSAMPLE - inptr[1];
+      b = MAXJSAMPLE - inptr[2];
      /* K passes through as-is */
-      outptr3[col] = inptr[3];  /* don't need GETJSAMPLE here */
+      outptr3[col] = inptr[3];
      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];   /* don't need GETJSAMPLE() here */
+      outptr[col] = inptr[0];
      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]; /* don't need GETJSAMPLE() here */
+          outptr[col] = inptr[ci];
          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++ = 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;
+    *workspaceptr++ = (*elemptr++) - CENTERJSAMPLE;
+    *workspaceptr++ = (*elemptr++) - CENTERJSAMPLE;
+    *workspaceptr++ = (*elemptr++) - CENTERJSAMPLE;
+    *workspaceptr++ = (*elemptr++) - CENTERJSAMPLE;
+    *workspaceptr++ = (*elemptr++) - CENTERJSAMPLE;
+    *workspaceptr++ = (*elemptr++) - CENTERJSAMPLE;
+    *workspaceptr++ = (*elemptr++) - CENTERJSAMPLE;
+    *workspaceptr++ = (*elemptr++) - CENTERJSAMPLE;
 #else
    {
      register int elemc;
      for (elemc = DCTSIZE; elemc > 0; elemc--)
-        *workspaceptr++ = GETJSAMPLE(*elemptr++) - CENTERJSAMPLE;
+        *workspaceptr++ = (*elemptr++) - CENTERJSAMPLE;
    }
 #endif
  }
@@ -533,20 +533,19 @@ 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)(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);
+    *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);
 #else
    {
      register int elemc;
      for (elemc = DCTSIZE; elemc > 0; elemc--)
-        *workspaceptr++ = (FAST_FLOAT)
-                          (GETJSAMPLE(*elemptr++) - CENTERJSAMPLE);
+        *workspaceptr++ = (FAST_FLOAT)((*elemptr++) - CENTERJSAMPLE);
    }
 #endif
  }
@@ -4,8 +4,10 @@
 * 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, 2018-2019, D. R. Commander.
+ * Copyright (C) 2009-2011, 2014-2016, 2018-2021, 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.
 *
@@ -43,14 +45,19 @@
 */

 /* NOTE: Both GCC and Clang define __GNUC__ */
-#if defined(__GNUC__) && (defined(__arm__) || defined(__aarch64__))
+#if (defined(__GNUC__) && (defined(__arm__) || defined(__aarch64__))) || \
+    defined(_M_ARM) || defined(_M_ARM64)
 #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"
@@ -65,32 +72,43 @@
 * 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 {
-  size_t put_buffer;                    /* current bit-accumulation buffer */
-  int put_bits;                         /* # of bits now in it */
+  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) */
  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 */

@@ -123,6 +141,7 @@ 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;


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

  /* Initialize bit buffer to empty */
-  entropy->saved.put_buffer = 0;
-  entropy->saved.put_bits = 0;
+  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;
+  }

  /* Initialize restart stuff */
  entropy->restarts_to_go = cinfo->restart_interval;
@@ -287,6 +315,7 @@ 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
@@ -334,94 +363,94 @@ dump_buffer(working_state *state)

 /* Outputting bits to the file */

-/* 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.
+/* 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.
 */
-
-#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; \
+#define EMIT_BYTE(b) { \
+  buffer[0] = (JOCTET)(b); \
+  buffer[1] = 0; \
+  buffer -= -2 + ((JOCTET)(b) < 0xFF); \
 }

-#define PUT_BITS(code, size) { \
-  put_bits += size; \
-  put_buffer = (put_buffer << size) | code; \
-}
+/* 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

-#if SIZEOF_SIZE_T != 8 && !defined(_WIN64)
-
-#define CHECKBUF15() { \
-  if (put_bits > 15) { \
-    EMIT_BYTE() \
-    EMIT_BYTE() \
+#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; \
  } \
 }

-#endif
-
-#define CHECKBUF31() { \
-  if (put_bits > 31) { \
-    EMIT_BYTE() \
-    EMIT_BYTE() \
-    EMIT_BYTE() \
-    EMIT_BYTE() \
-  } \
-}
-
-#define CHECKBUF47() { \
-  if (put_bits > 47) { \
-    EMIT_BYTE() \
-    EMIT_BYTE() \
-    EMIT_BYTE() \
-    EMIT_BYTE() \
-    EMIT_BYTE() \
-    EMIT_BYTE() \
-  } \
-}
-
-#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() \
+#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; \
+  } \
 }

 #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; \
+}
+
+/* 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 PUT_CODE(code, size) { \
+  temp &= (((JLONG)1) << nbits) - 1; \
+  temp |= code << nbits; \
+  nbits += size; \
+  PUT_BITS(temp, nbits) \
+}
+

 /* 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
@@ -444,6 +473,7 @@ dump_buffer(working_state *state)

 #define STORE_BUFFER() { \
  if (localbuf) { \
+    size_t bytes, bytestocopy; \
    bytes = buffer - _buffer; \
    buffer = _buffer; \
    while (bytes > 0) { \
@@ -466,20 +496,46 @@ dump_buffer(working_state *state)
 LOCAL(boolean)
 flush_bits(working_state *state)
 {
-  JOCTET _buffer[BUFSIZE], *buffer;
-  size_t put_buffer;  int put_bits;
-  size_t bytes, bytestocopy;  int localbuf = 0;
+  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;
+  }

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

-  /* fill any partial byte with ones */
-  PUT_BITS(0x7F, 7)
-  while (put_bits >= 8) EMIT_BYTE()
+  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)
+  }

-  state->cur.put_buffer = 0;    /* and reset bit-buffer to empty */
-  state->cur.put_bits = 0;
+  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;
+  }
  STORE_BUFFER()

  return TRUE;
@@ -493,7 +549,7 @@ 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;
-  size_t bytes, bytestocopy;  int localbuf = 0;
+  int localbuf = 0;

  LOAD_BUFFER()

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

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

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

-  temp = temp2 = block[0] - last_dc_val;
+  temp = 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.
+   * Agner Fog.  This code assumes we are on a two's complement machine.
   */
-  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;
+  nbits = temp >> (CHAR_BIT * sizeof(int) - 1);
+  temp += nbits;
+  nbits ^= temp;

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

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

-  r = 0;                        /* r = run length of zeros */
+  {
+    int 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
@@ -563,51 +607,46 @@ 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++; \
+    r += 16; \
  } else { \
-    temp2 = temp; \
    /* Branch-less absolute value, bitwise complement, etc., same as above */ \
-    temp3 = temp >> (CHAR_BIT * sizeof(int) - 1); \
-    temp ^= temp3; \
-    temp -= temp3; \
-    temp2 += temp3; \
-    nbits = JPEG_NBITS_NONZERO(temp); \
+    nbits = temp >> (CHAR_BIT * sizeof(int) - 1); \
+    temp += nbits; \
+    nbits ^= temp; \
+    nbits = JPEG_NBITS_NONZERO(nbits); \
    /* if run length > 15, must emit special run-length-16 codes (0xF0) */ \
-    while (r > 15) { \
-      EMIT_BITS(code_0xf0, size_0xf0) \
-      r -= 16; \
+    while (r >= 16 * 16) { \
+      r -= 16 * 16; \
+      PUT_BITS(actbl->ehufco[0xf0], actbl->ehufsi[0xf0]) \
    } \
    /* Emit Huffman symbol for run length / number of bits */ \
-    temp3 = (r << 4) + nbits; \
-    code = actbl->ehufco[temp3]; \
-    size = actbl->ehufsi[temp3]; \
-    EMIT_CODE(code, size) \
+    r += nbits; \
+    PUT_CODE(actbl->ehufco[r], actbl->ehufsi[r]) \
    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) {
-    code = actbl->ehufco[0];
-    size = actbl->ehufsi[0];
-    EMIT_BITS(code, size)
+    /* If the last coef(s) were zero, emit an end-of-block code */
+    if (r > 0) {
+      PUT_BITS(actbl->ehufco[0], actbl->ehufsi[0])
+    }
  }

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

  return TRUE;
@@ -654,8 +693,9 @@ 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;
-  ASSIGN_STATE(state.cur, entropy->saved);
+  state.cur = entropy->saved;
  state.cinfo = cinfo;
+  state.simd = entropy->simd;

  /* Emit restart marker if needed */
  if (cinfo->restart_interval) {
@@ -694,7 +734,7 @@ 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;
-  ASSIGN_STATE(entropy->saved, state.cur);
+  entropy->saved = state.cur;

  /* Update restart-interval state too */
  if (cinfo->restart_interval) {
@@ -723,8 +763,9 @@ 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;
-  ASSIGN_STATE(state.cur, entropy->saved);
+  state.cur = entropy->saved;
  state.cinfo = cinfo;
+  state.simd = entropy->simd;

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


@@ -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. */
@@ -4,8 +4,10 @@
 * 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, D. R. Commander.
+ * Copyright (C) 2011, 2015, 2018, 2021, 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.
 *
@@ -52,14 +54,19 @@
 */

 /* NOTE: Both GCC and Clang define __GNUC__ */
-#if defined(__GNUC__) && (defined(__arm__) || defined(__aarch64__))
+#if (defined(__GNUC__) && (defined(__arm__) || defined(__aarch64__))) || \
+    defined(_M_ARM) || defined(_M_ARM64)
 #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"
@@ -169,24 +176,26 @@ INLINE
 METHODDEF(int)
 count_zeroes(size_t *x)
 {
-  int result;
 #if defined(HAVE_BUILTIN_CTZL)
+  int result;
  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
-  result = 0;
+  int result = 0;
  while ((*x & 1) == 0) {
    ++result;
    *x >>= 1;
  }
 #endif
-  return result;
+  return (int)result;
 }


@@ -672,7 +681,7 @@ 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((size_t)values_unaligned, 16);
+  cvalue = values = (JCOEF *)PAD((JUINTPTR)values_unaligned, 16);
 #else
  /* Not using SIMD, so alignment is not needed */
  cvalue = values = values_unaligned;
@@ -860,7 +869,7 @@ encode_mcu_AC_refine_prepare(const JCOEF *block,

 #define ENCODE_COEFS_AC_REFINE(label) { \
  while (zerobits) { \
-    int idx = count_zeroes(&zerobits); \
+    idx = count_zeroes(&zerobits); \
    r += idx; \
    cabsvalue += idx; \
    signbits >>= idx; \
@@ -917,7 +926,7 @@ 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;
+  register int temp, r, idx;
  char *BR_buffer;
  unsigned int BR;
  int Sl = cinfo->Se - cinfo->Ss + 1;
@@ -937,7 +946,7 @@ 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((size_t)absvalues_unaligned, 16);
+  cabsvalue = absvalues = (JCOEF *)PAD((JUINTPTR)absvalues_unaligned, 16);
 #else
  /* Not using SIMD, so alignment is not needed */
  cabsvalue = absvalues = absvalues_unaligned;
@@ -968,7 +977,7 @@ encode_mcu_AC_refine(j_compress_ptr cinfo, JBLOCKROW *MCU_data)

  if (zerobits) {
    int diff = ((absvalues + DCTSIZE2 / 2) - cabsvalue);
-    int idx = count_zeroes(&zerobits);
+    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, D. R. Commander.
+ * Copyright (C) 2015, 2019, 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];         /* don't need GETJSAMPLE() here */
+      pixval = ptr[-1];
      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)GETJSAMPLE(*inptr++);
+          outvalue += (JLONG)(*inptr++);
        }
      }
      *outptr++ = (JSAMPLE)((outvalue + numpix2) / numpix);
@@ -237,8 +237,7 @@ 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)((GETJSAMPLE(*inptr) + GETJSAMPLE(inptr[1]) + bias) >> 1);
+      *outptr++ = (JSAMPLE)((inptr[0] + inptr[1] + bias) >> 1);
      bias ^= 1;                /* 0=>1, 1=>0 */
      inptr += 2;
    }
@@ -277,8 +276,7 @@ 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)((GETJSAMPLE(*inptr0) + GETJSAMPLE(inptr0[1]) +
-                   GETJSAMPLE(*inptr1) + GETJSAMPLE(inptr1[1]) + bias) >> 2);
+        (JSAMPLE)((inptr0[0] + inptr0[1] + inptr1[0] + inptr1[1] + bias) >> 2);
      bias ^= 3;                /* 1=>2, 2=>1 */
      inptr0 += 2;  inptr1 += 2;
    }
@@ -337,33 +335,25 @@ 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 = 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]);
+    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];
    neighsum += neighsum;
-    neighsum += GETJSAMPLE(*above_ptr) + GETJSAMPLE(above_ptr[2]) +
-                GETJSAMPLE(*below_ptr) + GETJSAMPLE(below_ptr[2]);
+    neighsum += above_ptr[0] + above_ptr[2] + below_ptr[0] + 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 = GETJSAMPLE(*inptr0) + GETJSAMPLE(inptr0[1]) +
-                  GETJSAMPLE(*inptr1) + GETJSAMPLE(inptr1[1]);
+      membersum = inptr0[0] + inptr0[1] + inptr1[0] + inptr1[1];
      /* sum of edge-neighbor pixels */
-      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]);
+      neighsum = above_ptr[0] + above_ptr[1] + below_ptr[0] + below_ptr[1] +
+                 inptr0[-1] + inptr0[2] + inptr1[-1] + inptr1[2];
      /* The edge-neighbors count twice as much as corner-neighbors */
      neighsum += neighsum;
      /* Add in the corner-neighbors */
-      neighsum += GETJSAMPLE(above_ptr[-1]) + GETJSAMPLE(above_ptr[2]) +
-                  GETJSAMPLE(below_ptr[-1]) + GETJSAMPLE(below_ptr[2]);
+      neighsum += above_ptr[-1] + above_ptr[2] + below_ptr[-1] + below_ptr[2];
      /* form final output scaled up by 2^16 */
      membersum = membersum * memberscale + neighsum * neighscale;
      /* round, descale and output it */
@@ -372,15 +362,11 @@ h2v2_smooth_downsample(j_compress_ptr cinfo, jpeg_component_info *compptr,
    }

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

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

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

    /* Special case for last column */
-    membersum = GETJSAMPLE(*inptr);
+    membersum = *inptr;
    neighsum = lastcolsum + (colsum - membersum) + colsum;
    membersum = membersum * memberscale + neighsum * neighscale;
    *outptr = (JSAMPLE)((membersum + 32768) >> 16);
@@ -23,6 +23,7 @@
 #include "jinclude.h"
 #include "jpeglib.h"
 #include "jdmaster.h"
+#include "jconfigint.h"


 /*
@@ -308,7 +309,7 @@ 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, 2015-2018, 2020, D. R. Commander.
+ * Copyright (C) 2010, 2015-2020, D. R. Commander.
 * Copyright (C) 2015, Google, Inc.
 * For conditions of distribution and use, see the accompanying README.ijg
 * file.
@@ -319,6 +319,8 @@ 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,
@@ -329,6 +331,10 @@ 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) {
@@ -532,6 +538,8 @@ 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);
      }
    }
@@ -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) 2015-2018, D. R. Commander.
+ * Copyright (C) 2015-2020, 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 GETJOCTET(*src->next_input_byte++);
+  return *src->next_input_byte++;
 }


@@ -665,8 +665,16 @@ 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)
@@ -727,6 +735,7 @@ 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;
@@ -763,7 +772,7 @@ 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 * DCTSIZE2 *
+                                  cinfo->num_components * 2 * DCTSIZE2 *
                                  sizeof(int));
    coef_bit_ptr = &cinfo->coef_bits[0][0];
    for (ci = 0; ci < cinfo->num_components; ci++)
@@ -5,7 +5,7 @@
 * 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, D. R. Commander.
+ * Copyright (C) 2010, 2015-2016, 2019-2020, D. R. Commander.
 * Copyright (C) 2015, 2020, Google, Inc.
 * For conditions of distribution and use, see the accompanying README.ijg
 * file.
@@ -102,6 +102,8 @@ 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;
@@ -227,6 +229,8 @@ 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 */
@@ -326,19 +330,22 @@ decompress_data(j_decompress_ptr cinfo, JSAMPIMAGE output_buf)
 #ifdef BLOCK_SMOOTHING_SUPPORTED

 /*
- * 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.
+ * 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.
 * 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 5 zigzag-order coefficients */
+/* Natural-order array positions of the first 9 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.
@@ -356,8 +363,8 @@ smoothing_ok(j_decompress_ptr cinfo)
  int ci, coefi;
  jpeg_component_info *compptr;
  JQUANT_TBL *qtable;
-  int *coef_bits;
-  int *coef_bits_latch;
+  int *coef_bits, *prev_coef_bits;
+  int *coef_bits_latch, *prev_coef_bits_latch;

  if (!cinfo->progressive_mode || cinfo->coef_bits == NULL)
    return FALSE;
@@ -366,34 +373,47 @@ 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 *
+                                  cinfo->num_components * 2 *
                                  (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 5 AC quantizers are nonzero to avoid zero-divide. */
+    /* Verify DC & first 9 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[Q02_POS] == 0 ||
+        qtable->quantval[Q03_POS] == 0 ||
+        qtable->quantval[Q12_POS] == 0 ||
+        qtable->quantval[Q21_POS] == 0 ||
+        qtable->quantval[Q30_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 <= 5; coefi++) {
+    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;
      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;
@@ -412,17 +432,20 @@ 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_block_row, next_block_row;
+  JBLOCKROW buffer_ptr, prev_prev_block_row, prev_block_row;
+  JBLOCKROW next_block_row, next_next_block_row;
  JSAMPARRAY output_ptr;
  JDIMENSION output_col;
  jpeg_component_info *compptr;
  inverse_DCT_method_ptr inverse_DCT;
-  boolean first_row, last_row;
+  boolean change_dc;
  JCOEF *workspace;
  int *coef_bits;
  JQUANT_TBL *quanttbl;
-  JLONG Q00, Q01, Q02, Q10, Q11, Q20, num;
-  int DC1, DC2, DC3, DC4, DC5, DC6, DC7, DC8, DC9;
+  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;
  int Al, pred;

  /* Keep a local variable to avoid looking it up more than once */
@@ -434,10 +457,10 @@ 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 one row ahead so that next block row's DC
+       * we want it to keep two rows ahead so that next two block rows' DC
       * values are up to date.
       */
-      JDIMENSION delta = (cinfo->Ss == 0) ? 1 : 0;
+      JDIMENSION delta = (cinfo->Ss == 0) ? 2 : 0;
      if (cinfo->input_iMCU_row > cinfo->output_iMCU_row + delta)
        break;
    }
@@ -452,34 +475,53 @@ 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) {
+    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) {
      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 > 0) {
-      access_rows += compptr->v_samp_factor; /* prior iMCU row too */
+    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) {
      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 */
-    coef_bits = coef->coef_bits_latch + (ci * SAVED_COEFS);
+    /* 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;
+
    quanttbl = compptr->quant_table;
    Q00 = quanttbl->quantval[0];
    Q01 = quanttbl->quantval[Q01_POS];
@@ -487,27 +529,51 @@ 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 (first_row && block_row == 0)
+
+      if (block_row > 0 || cinfo->output_iMCU_row > 0)
+        prev_block_row =
+          buffer[block_row - 1] + cinfo->master->first_MCU_col[ci];
+      else
        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] +
-                         cinfo->master->first_MCU_col[ci];
-      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_block_row = buffer[block_row + 1] +
-                         cinfo->master->first_MCU_col[ci];
+        next_next_block_row = next_block_row;
+
      /* We fetch the surrounding DC values using a sliding-register approach.
-       * Initialize all nine here so as to do the right thing on narrow pics.
+       * Initialize all 25 here so as to do the right thing on narrow pics.
       */
-      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];
+      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];
      output_col = 0;
      last_block_column = compptr->width_in_blocks - 1;
      for (block_num = cinfo->master->first_MCU_col[ci];
@@ -515,18 +581,39 @@ 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 < 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 == 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];
        }
-        /* Compute coefficient estimates per K.8.
-         * An estimate is applied only if coefficient is still zero,
-         * and is not known to be fully accurate.
+        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.
         */
        /* AC01 */
        if ((Al = coef_bits[1]) != 0 && workspace[1] == 0) {
-          num = 36 * Q00 * (DC4 - DC6);
+          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));
          if (num >= 0) {
            pred = (int)(((Q01 << 7) + num) / (Q01 << 8));
            if (Al > 0 && pred >= (1 << Al))
@@ -541,7 +628,12 @@ decompress_smooth_data(j_decompress_ptr cinfo, JSAMPIMAGE output_buf)
        }
        /* AC10 */
        if ((Al = coef_bits[2]) != 0 && workspace[8] == 0) {
-          num = 36 * Q00 * (DC2 - DC8);
+          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));
          if (num >= 0) {
            pred = (int)(((Q10 << 7) + num) / (Q10 << 8));
            if (Al > 0 && pred >= (1 << Al))
@@ -556,7 +648,10 @@ decompress_smooth_data(j_decompress_ptr cinfo, JSAMPIMAGE output_buf)
        }
        /* AC20 */
        if ((Al = coef_bits[3]) != 0 && workspace[16] == 0) {
-          num = 9 * Q00 * (DC2 + DC8 - 2 * DC5);
+          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));
          if (num >= 0) {
            pred = (int)(((Q20 << 7) + num) / (Q20 << 8));
            if (Al > 0 && pred >= (1 << Al))
@@ -571,7 +666,11 @@ decompress_smooth_data(j_decompress_ptr cinfo, JSAMPIMAGE output_buf)
        }
        /* AC11 */
        if ((Al = coef_bits[4]) != 0 && workspace[9] == 0) {
-          num = 5 * Q00 * (DC1 - DC3 - DC7 + DC9);
+          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));
          if (num >= 0) {
            pred = (int)(((Q11 << 7) + num) / (Q11 << 8));
            if (Al > 0 && pred >= (1 << Al))
@@ -586,7 +685,10 @@ decompress_smooth_data(j_decompress_ptr cinfo, JSAMPIMAGE output_buf)
        }
        /* AC02 */
        if ((Al = coef_bits[5]) != 0 && workspace[2] == 0) {
-          num = 9 * Q00 * (DC4 + DC6 - 2 * DC5);
+          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));
          if (num >= 0) {
            pred = (int)(((Q02 << 7) + num) / (Q02 << 8));
            if (Al > 0 && pred >= (1 << Al))
@@ -599,14 +701,96 @@ 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 */
-        DC1 = DC2;  DC2 = DC3;
-        DC4 = DC5;  DC5 = DC6;
-        DC7 = DC8;  DC8 = DC9;
-        buffer_ptr++, prev_block_row++, next_block_row++;
+        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++;
        output_col += compptr->_DCT_scaled_size;
      }
      output_ptr += compptr->_DCT_scaled_size;
@@ -655,7 +839,7 @@ 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 *= 3;
+        access_rows *= 5;
 #endif
      coef->whole_image[ci] = (*cinfo->mem->request_virt_barray)
        ((j_common_ptr)cinfo, JPOOL_IMAGE, TRUE,
@@ -5,6 +5,7 @@
 * 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.
 */
@@ -51,7 +52,7 @@ typedef struct {
 #ifdef BLOCK_SMOOTHING_SUPPORTED
  /* When doing block smoothing, we latch coefficient Al values here */
  int *coef_bits_latch;
-#define SAVED_COEFS  6          /* we save coef_bits[0..5] */
+#define SAVED_COEFS  10         /* we save coef_bits[0..9] */
 #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  = GETJSAMPLE(*inptr0++);
-      cb = GETJSAMPLE(*inptr1++);
-      cr = GETJSAMPLE(*inptr2++);
+      y  = *inptr0++;
+      cb = *inptr1++;
+      cr = *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  = GETJSAMPLE(*inptr0++);
-      cb = GETJSAMPLE(*inptr1++);
-      cr = GETJSAMPLE(*inptr2++);
+      y  = *inptr0++;
+      cb = *inptr1++;
+      cr = *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  = GETJSAMPLE(*inptr0++);
-      cb = GETJSAMPLE(*inptr1++);
-      cr = GETJSAMPLE(*inptr2++);
+      y  = *inptr0++;
+      cb = *inptr1++;
+      cr = *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  = GETJSAMPLE(*inptr0);
-      cb = GETJSAMPLE(*inptr1);
-      cr = GETJSAMPLE(*inptr2);
+      y  = *inptr0;
+      cb = *inptr1;
+      cr = *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  = GETJSAMPLE(*inptr0++);
-      cb = GETJSAMPLE(*inptr1++);
-      cr = GETJSAMPLE(*inptr2++);
+      y  = *inptr0++;
+      cb = *inptr1++;
+      cr = *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  = GETJSAMPLE(*inptr0++);
-      cb = GETJSAMPLE(*inptr1++);
-      cr = GETJSAMPLE(*inptr2++);
+      y  = *inptr0++;
+      cb = *inptr1++;
+      cr = *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  = GETJSAMPLE(*inptr0++);
-      cb = GETJSAMPLE(*inptr1++);
-      cr = GETJSAMPLE(*inptr2++);
+      y  = *inptr0++;
+      cb = *inptr1++;
+      cr = *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  = GETJSAMPLE(*inptr0);
-      cb = GETJSAMPLE(*inptr1);
-      cr = GETJSAMPLE(*inptr2);
+      y  = *inptr0;
+      cb = *inptr1;
+      cr = *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 = GETJSAMPLE(*inptr0++);
-      g = GETJSAMPLE(*inptr1++);
-      b = GETJSAMPLE(*inptr2++);
+      r = *inptr0++;
+      g = *inptr1++;
+      b = *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 = GETJSAMPLE(*inptr0++);
-      g = GETJSAMPLE(*inptr1++);
-      b = GETJSAMPLE(*inptr2++);
+      r = *inptr0++;
+      g = *inptr1++;
+      b = *inptr2++;
      rgb = PACK_SHORT_565(r, g, b);

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

-      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)];
+      r = range_limit[DITHER_565_R(*inptr0++, d0)];
+      g = range_limit[DITHER_565_G(*inptr1++, d0)];
+      b = range_limit[DITHER_565_B(*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(GETJSAMPLE(*inptr0), d0)];
-      g = range_limit[DITHER_565_G(GETJSAMPLE(*inptr1), d0)];
-      b = range_limit[DITHER_565_B(GETJSAMPLE(*inptr2), d0)];
+      r = range_limit[DITHER_565_R(*inptr0, d0)];
+      g = range_limit[DITHER_565_G(*inptr1, d0)];
+      b = range_limit[DITHER_565_B(*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  = GETJSAMPLE(inptr0[col]);
-      cb = GETJSAMPLE(inptr1[col]);
-      cr = GETJSAMPLE(inptr2[col]);
+      y  = inptr0[col];
+      cb = inptr1[col];
+      cr = 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 +
@@ -93,7 +93,6 @@ 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 */
@@ -128,7 +127,6 @@ 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 = GETJSAMPLE(inptr0[col]);
-      g = GETJSAMPLE(inptr1[col]);
-      b = GETJSAMPLE(inptr2[col]);
+      r = inptr0[col];
+      g = inptr1[col];
+      b = 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  = GETJSAMPLE(inptr0[col]);
-      cb = GETJSAMPLE(inptr1[col]);
-      cr = GETJSAMPLE(inptr2[col]);
+      y  = inptr0[col];
+      cb = inptr1[col];
+      cr = 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];  /* don't need GETJSAMPLE here */
+      outptr[3] = inptr3[col];
      outptr += 4;
    }
  }
@@ -5,6 +5,7 @@
 * 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.
 *
@@ -39,24 +40,6 @@ 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 */

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

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

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

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

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

 /* Pre-fetch 48 bytes, because the holding register is 64-bit */
 #define FILL_BIT_BUFFER_FAST \
@@ -557,6 +540,12 @@ 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)
 {
@@ -568,7 +557,7 @@ decode_mcu_slow(j_decompress_ptr cinfo, JBLOCKROW *MCU_data)

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

  for (blkn = 0; blkn < cinfo->blocks_in_MCU; blkn++) {
    JBLOCKROW block = MCU_data ? MCU_data[blkn] : NULL;
@@ -589,11 +578,15 @@ 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];
-      /* This is really just
-       *   s += state.last_dc_val[ci];
-       * It is written this way in order to shut up UBSan.
+      /* 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().
       */
-      s = (int)((unsigned int)s + (unsigned int)state.last_dc_val[ci]);
+      s += state.last_dc_val[ci];
      state.last_dc_val[ci] = s;
      if (block) {
        /* Output the DC coefficient (assumes jpeg_natural_order[0] = 0) */
@@ -653,11 +646,17 @@ decode_mcu_slow(j_decompress_ptr cinfo, JBLOCKROW *MCU_data)

  /* Completed MCU, so update state */
  BITREAD_SAVE_STATE(cinfo, entropy->bitstate);
-  ASSIGN_STATE(entropy->saved, 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)
 {
@@ -671,7 +670,7 @@ 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;
-  ASSIGN_STATE(state, entropy->saved);
+  state = entropy->saved;

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

    if (entropy->dc_needed[blkn]) {
      int ci = cinfo->MCU_membership[blkn];
-      s = (int)((unsigned int)s + (unsigned int)state.last_dc_val[ci]);
+      /* 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];
      state.last_dc_val[ci] = s;
      if (block)
        (*block)[0] = (JCOEF)s;
@@ -740,7 +742,7 @@ 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);
-  ASSIGN_STATE(entropy->saved, state);
+  entropy->saved = state;
  return TRUE;
 }

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

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

  return TRUE;
 }
@@ -4,7 +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) 2010-2011, 2015-2016, D. R. Commander.
+ * Copyright (C) 2010-2011, 2015-2016, 2021, D. R. Commander.
+ * Copyright (C) 2018, Matthias Räncker.
 * For conditions of distribution and use, see the accompanying README.ijg
 * file.
 *
@@ -78,6 +79,11 @@ 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 */
@@ -228,7 +234,10 @@ slowlabel: \
      s |= GET_BITS(1); \
      nb++; \
    } \
-    s = htbl->pub->huffval[(int)(s + htbl->valoffset[nb]) & 0xFF]; \
+    if (nb > 16) \
+      s = 0; \
+    else \
+      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 */
-    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;
+    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;
 }


@@ -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 = GETJOCTET(marker->data[13]);
-      else if (num_markers != GETJOCTET(marker->data[13])) {
+        num_markers = marker->data[13];
+      else if (num_markers != marker->data[13]) {
        WARNMS(cinfo, JWRN_BOGUS_ICC);  /* inconsistent num_markers fields */
        return FALSE;
      }
-      seq_no = GETJOCTET(marker->data[12]);
+      seq_no = 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 = GETJOCTET(marker->data[12]);
+      seq_no = marker->data[12];
      dst_ptr = icc_data + data_offset[seq_no];
      src_ptr = marker->data + ICC_OVERHEAD_LEN;
      length = data_length[seq_no];
@@ -18,6 +18,7 @@

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


 /*
@@ -360,7 +361,7 @@ 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;
@@ -371,7 +372,7 @@ 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 = GETJOCTET(*next_input_byte++); )
+            V = *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)GETJOCTET(*next_input_byte++)) << 8; \
+            V = ((unsigned int)(*next_input_byte++)) << 8; \
            MAKE_BYTE_AVAIL(cinfo, action); \
            bytes_in_buffer--; \
-            V += GETJOCTET(*next_input_byte++); )
+            V += *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 &&
-      GETJOCTET(data[0]) == 0x4A &&
-      GETJOCTET(data[1]) == 0x46 &&
-      GETJOCTET(data[2]) == 0x49 &&
-      GETJOCTET(data[3]) == 0x46 &&
-      GETJOCTET(data[4]) == 0) {
+      data[0] == 0x4A &&
+      data[1] == 0x46 &&
+      data[2] == 0x49 &&
+      data[3] == 0x46 &&
+      data[4] == 0) {
    /* Found JFIF APP0 marker: save info */
    cinfo->saw_JFIF_marker = TRUE;
-    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]);
+    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];
    /* 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,
@@ -634,24 +634,22 @@ 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 (GETJOCTET(data[12]) | GETJOCTET(data[13]))
-      TRACEMS2(cinfo, 1, JTRC_JFIF_THUMBNAIL,
-               GETJOCTET(data[12]), GETJOCTET(data[13]));
+    if (data[12] | data[13])
+      TRACEMS2(cinfo, 1, JTRC_JFIF_THUMBNAIL, data[12], data[13]);
    totallen -= APP0_DATA_LEN;
-    if (totallen !=
-        ((JLONG)GETJOCTET(data[12]) * (JLONG)GETJOCTET(data[13]) * (JLONG)3))
+    if (totallen != ((JLONG)data[12] * (JLONG)data[13] * (JLONG)3))
      TRACEMS1(cinfo, 1, JTRC_JFIF_BADTHUMBNAILSIZE, (int)totallen);
  } else if (datalen >= 6 &&
-             GETJOCTET(data[0]) == 0x4A &&
-             GETJOCTET(data[1]) == 0x46 &&
-             GETJOCTET(data[2]) == 0x58 &&
-             GETJOCTET(data[3]) == 0x58 &&
-             GETJOCTET(data[4]) == 0) {
+             data[0] == 0x4A &&
+             data[1] == 0x46 &&
+             data[2] == 0x58 &&
+             data[3] == 0x58 &&
+             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 (GETJOCTET(data[5])) {
+    switch (data[5]) {
    case 0x10:
      TRACEMS1(cinfo, 1, JTRC_THUMB_JPEG, (int)totallen);
      break;
@@ -662,8 +660,7 @@ 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,
-               GETJOCTET(data[5]), (int)totallen);
+      TRACEMS2(cinfo, 1, JTRC_JFIF_EXTENSION, data[5], (int)totallen);
      break;
    }
  } else {
@@ -684,16 +681,16 @@ examine_app14(j_decompress_ptr cinfo, JOCTET *data, unsigned int datalen,
  unsigned int version, flags0, flags1, transform;

  if (datalen >= APP14_DATA_LEN &&
-      GETJOCTET(data[0]) == 0x41 &&
-      GETJOCTET(data[1]) == 0x64 &&
-      GETJOCTET(data[2]) == 0x6F &&
-      GETJOCTET(data[3]) == 0x62 &&
-      GETJOCTET(data[4]) == 0x65) {
+      data[0] == 0x41 &&
+      data[1] == 0x64 &&
+      data[2] == 0x6F &&
+      data[3] == 0x62 &&
+      data[4] == 0x65) {
    /* Found Adobe APP14 marker */
-    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]);
+    version = (data[5] << 8) + data[6];
+    flags0 = (data[7] << 8) + data[8];
+    flags1 = (data[9] << 8) + data[10];
+    transform = 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, D. R. Commander.
+ * Copyright (C) 2009-2011, 2016, 2019, D. R. Commander.
 * Copyright (C) 2013, Linaro Limited.
 * Copyright (C) 2015, Google, Inc.
 * For conditions of distribution and use, see the accompanying README.ijg
@@ -22,7 +22,6 @@
 #include "jpeglib.h"
 #include "jpegcomp.h"
 #include "jdmaster.h"
-#include "jsimd.h"


 /*
@@ -70,17 +69,6 @@ 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
@@ -580,6 +568,7 @@ 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
@@ -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 = GETJSAMPLE(*inptr1++);
-    cr = GETJSAMPLE(*inptr2++);
+    cb = *inptr1++;
+    cr = *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  = GETJSAMPLE(*inptr0++);
+    y  = *inptr0++;
    r = range_limit[y + cred];
    g = range_limit[y + cgreen];
    b = range_limit[y + cblue];
    rgb = PACK_SHORT_565(r, g, b);

-    y  = GETJSAMPLE(*inptr0++);
+    y  = *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 = GETJSAMPLE(*inptr1);
-    cr = GETJSAMPLE(*inptr2);
+    cb = *inptr1;
+    cr = *inptr2;
    cred = Crrtab[cr];
    cgreen = (int)RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr], SCALEBITS);
    cblue = Cbbtab[cb];
-    y  = GETJSAMPLE(*inptr0);
+    y  = *inptr0;
    r = range_limit[y + cred];
    g = range_limit[y + cgreen];
    b = range_limit[y + cblue];
@@ -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 = GETJSAMPLE(*inptr1++);
-    cr = GETJSAMPLE(*inptr2++);
+    cb = *inptr1++;
+    cr = *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  = GETJSAMPLE(*inptr0++);
+    y  = *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  = GETJSAMPLE(*inptr0++);
+    y  = *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 = GETJSAMPLE(*inptr1);
-    cr = GETJSAMPLE(*inptr2);
+    cb = *inptr1;
+    cr = *inptr2;
    cred = Crrtab[cr];
    cgreen = (int)RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr], SCALEBITS);
    cblue = Cbbtab[cb];
-    y  = GETJSAMPLE(*inptr0);
+    y  = *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)];
@@ -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 = GETJSAMPLE(*inptr1++);
-    cr = GETJSAMPLE(*inptr2++);
+    cb = *inptr1++;
+    cr = *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  = GETJSAMPLE(*inptr00++);
+    y  = *inptr00++;
    r = range_limit[y + cred];
    g = range_limit[y + cgreen];
    b = range_limit[y + cblue];
    rgb = PACK_SHORT_565(r, g, b);

-    y  = GETJSAMPLE(*inptr00++);
+    y  = *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  = GETJSAMPLE(*inptr01++);
+    y  = *inptr01++;
    r = range_limit[y + cred];
    g = range_limit[y + cgreen];
    b = range_limit[y + cblue];
    rgb = PACK_SHORT_565(r, g, b);

-    y  = GETJSAMPLE(*inptr01++);
+    y  = *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 = GETJSAMPLE(*inptr1);
-    cr = GETJSAMPLE(*inptr2);
+    cb = *inptr1;
+    cr = *inptr2;
    cred = Crrtab[cr];
    cgreen = (int)RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr], SCALEBITS);
    cblue = Cbbtab[cb];

-    y  = GETJSAMPLE(*inptr00);
+    y  = *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  = GETJSAMPLE(*inptr01);
+    y  = *inptr01;
    r = range_limit[y + cred];
    g = range_limit[y + cgreen];
    b = range_limit[y + cblue];
@@ -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 = GETJSAMPLE(*inptr1++);
-    cr = GETJSAMPLE(*inptr2++);
+    cb = *inptr1++;
+    cr = *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  = GETJSAMPLE(*inptr00++);
+    y  = *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  = GETJSAMPLE(*inptr00++);
+    y  = *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  = GETJSAMPLE(*inptr01++);
+    y  = *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  = GETJSAMPLE(*inptr01++);
+    y  = *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 = GETJSAMPLE(*inptr1);
-    cr = GETJSAMPLE(*inptr2);
+    cb = *inptr1;
+    cr = *inptr2;
    cred = Crrtab[cr];
    cgreen = (int)RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr], SCALEBITS);
    cblue = Cbbtab[cb];

-    y  = GETJSAMPLE(*inptr00);
+    y  = *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  = GETJSAMPLE(*inptr01);
+    y  = *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)];
@@ -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 = GETJSAMPLE(*inptr1++);
-    cr = GETJSAMPLE(*inptr2++);
+    cb = *inptr1++;
+    cr = *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  = GETJSAMPLE(*inptr0++);
+    y  = *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  = GETJSAMPLE(*inptr0++);
+    y  = *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 = GETJSAMPLE(*inptr1);
-    cr = GETJSAMPLE(*inptr2);
+    cb = *inptr1;
+    cr = *inptr2;
    cred = Crrtab[cr];
    cgreen = (int)RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr], SCALEBITS);
    cblue = Cbbtab[cb];
-    y  = GETJSAMPLE(*inptr0);
+    y  = *inptr0;
    outptr[RGB_RED] =   range_limit[y + cred];
    outptr[RGB_GREEN] = range_limit[y + cgreen];
    outptr[RGB_BLUE] =  range_limit[y + cblue];
@@ -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 = GETJSAMPLE(*inptr1++);
-    cr = GETJSAMPLE(*inptr2++);
+    cb = *inptr1++;
+    cr = *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  = GETJSAMPLE(*inptr00++);
+    y  = *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  = GETJSAMPLE(*inptr00++);
+    y  = *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  = GETJSAMPLE(*inptr01++);
+    y  = *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  = GETJSAMPLE(*inptr01++);
+    y  = *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 = GETJSAMPLE(*inptr1);
-    cr = GETJSAMPLE(*inptr2);
+    cb = *inptr1;
+    cr = *inptr2;
    cred = Crrtab[cr];
    cgreen = (int)RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr], SCALEBITS);
    cblue = Cbbtab[cb];
-    y  = GETJSAMPLE(*inptr00);
+    y  = *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  = GETJSAMPLE(*inptr01);
+    y  = *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, D. R. Commander.
+ * Copyright (C) 2015-2016, 2018-2021, D. R. Commander.
 * For conditions of distribution and use, see the accompanying README.ijg
 * file.
 *
@@ -41,25 +41,6 @@ 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 */

@@ -102,7 +83,7 @@ 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;
+  int *coef_bit_ptr, *prev_coef_bit_ptr;
  jpeg_component_info *compptr;

  is_DC_band = (cinfo->Ss == 0);
@@ -143,8 +124,15 @@ 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)
@@ -323,7 +311,7 @@ decode_mcu_DC_first(j_decompress_ptr cinfo, JBLOCKROW *MCU_data)

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

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

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

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

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

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

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

  return TRUE;
 }
@@ -495,7 +485,8 @@ 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) */
-  entropy->restarts_to_go--;
+  if (cinfo->restart_interval)
+    entropy->restarts_to_go--;

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

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

  return TRUE;

@@ -676,7 +668,7 @@ 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 * DCTSIZE2 *
+                                cinfo->num_components * 2 * 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, Arm Limited.
+ * Copyright (C) 2019-2020, 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++;       /* don't need GETJSAMPLE() here */
+      invalue = *inptr++;
      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++;       /* don't need GETJSAMPLE() here */
+      invalue = *inptr++;
      *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++;       /* don't need GETJSAMPLE() here */
+      invalue = *inptr++;
      *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 = GETJSAMPLE(*inptr++);
+    invalue = *inptr++;
    *outptr++ = (JSAMPLE)invalue;
-    *outptr++ = (JSAMPLE)((invalue * 3 + GETJSAMPLE(*inptr) + 2) >> 2);
+    *outptr++ = (JSAMPLE)((invalue * 3 + inptr[0] + 2) >> 2);

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

    /* Special case for last column */
-    invalue = GETJSAMPLE(*inptr);
-    *outptr++ = (JSAMPLE)((invalue * 3 + GETJSAMPLE(inptr[-1]) + 1) >> 2);
+    invalue = *inptr;
+    *outptr++ = (JSAMPLE)((invalue * 3 + 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 = GETJSAMPLE(*inptr0++) * 3 + GETJSAMPLE(*inptr1++);
+        thiscolsum = (*inptr0++) * 3 + (*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 = GETJSAMPLE(*inptr0++) * 3 + GETJSAMPLE(*inptr1++);
-      nextcolsum = GETJSAMPLE(*inptr0++) * 3 + GETJSAMPLE(*inptr1++);
+      thiscolsum = (*inptr0++) * 3 + (*inptr1++);
+      nextcolsum = (*inptr0++) * 3 + (*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 = GETJSAMPLE(*inptr0++) * 3 + GETJSAMPLE(*inptr1++);
+        nextcolsum = (*inptr0++) * 3 + (*inptr1++);
        *outptr++ = (JSAMPLE)((thiscolsum * 3 + lastcolsum + 8) >> 4);
        *outptr++ = (JSAMPLE)((thiscolsum * 3 + nextcolsum + 7) >> 4);
        lastcolsum = thiscolsum;  thiscolsum = nextcolsum;
@@ -477,7 +477,13 @@ 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 */
-      upsample->methods[ci] = h1v2_fancy_upsample;
+#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->pub.need_context_rows = TRUE;
    } else if (h_in_group * 2 == h_out_group &&
               v_in_group * 2 == v_out_group) {
@@ -207,6 +207,10 @@ 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

@@ -252,6 +256,15 @@ JMESSAGE(JWRN_BOGUS_ICC, "Corrupt JPEG data: bad ICC marker")
   (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), \
@@ -3,7 +3,7 @@
 *
 * This file was part of the Independent JPEG Group's software:
 * Copyright (C) 1991-1998, Thomas G. Lane.
- * Modification developed 2002-2009 by Guido Vollbeding.
+ * Modification developed 2002-2018 by Guido Vollbeding.
 * libjpeg-turbo Modifications:
 * Copyright (C) 2015, 2020, D. R. Commander.
 * For conditions of distribution and use, see the accompanying README.ijg
@@ -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 7x7 output block.
+ * producing a reduced-size 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 a 11x11 output block.
+ * producing an 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 += 1 << (CONST_BITS - PASS1_BITS - 1);
+    tmp0 += ONE << (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, D. R. Commander.
+ * Copyright (C) 2016, 2021, 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);
+                  sizeof(large_pool_hdr) + ALIGN_SIZE - 1;
    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);
+                  sizeof(small_pool_hdr) + ALIGN_SIZE - 1;
    jpeg_free_small(cinfo, (void *)shdr_ptr, space_freed);
    mem->total_space_allocated -= space_freed;
    shdr_ptr = next_shdr_ptr;
@@ -43,25 +43,11 @@

 #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

@@ -97,22 +83,9 @@ 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
@@ -123,15 +96,7 @@ typedef 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. */

@@ -5,8 +5,9 @@
 * Copyright (C) 1991-1997, Thomas G. Lane.
 * Modified 1997-2009 by Guido Vollbeding.
 * libjpeg-turbo Modifications:
- * Copyright (C) 2015-2016, D. R. Commander.
+ * Copyright (C) 2015-2016, 2019, 2021, 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.
 *
@@ -47,6 +48,18 @@ 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
@@ -158,6 +171,9 @@ 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 */
@@ -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 += GETJSAMPLE(colorindex[ci][GETJSAMPLE(*ptrin++)]);
+        pixcode += colorindex[ci][*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  = GETJSAMPLE(colorindex0[GETJSAMPLE(*ptrin++)]);
-      pixcode += GETJSAMPLE(colorindex1[GETJSAMPLE(*ptrin++)]);
-      pixcode += GETJSAMPLE(colorindex2[GETJSAMPLE(*ptrin++)]);
+      pixcode  = colorindex0[*ptrin++];
+      pixcode += colorindex1[*ptrin++];
+      pixcode += colorindex2[*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[GETJSAMPLE(*input_ptr) + dither[col_index]];
+          colorindex_ci[*input_ptr + dither[col_index]];
        input_ptr += nc;
        output_ptr++;
        col_index = (col_index + 1) & ODITHER_MASK;
@@ -595,12 +595,9 @@ quantize3_ord_dither(j_decompress_ptr cinfo, JSAMPARRAY input_buf,
    col_index = 0;

    for (col = width; col > 0; col--) {
-      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]]);
+      pixcode  = colorindex0[(*input_ptr++) + dither0[col_index]];
+      pixcode += colorindex1[(*input_ptr++) + dither1[col_index]];
+      pixcode += colorindex2[(*input_ptr++) + dither2[col_index]];
      *output_ptr++ = (JSAMPLE)pixcode;
      col_index = (col_index + 1) & ODITHER_MASK;
    }
@@ -677,15 +674,15 @@ 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 += GETJSAMPLE(*input_ptr);
-        cur = GETJSAMPLE(range_limit[cur]);
+        cur += *input_ptr;
+        cur = range_limit[cur];
        /* Select output value, accumulate into output code for this pixel */
-        pixcode = GETJSAMPLE(colorindex_ci[cur]);
+        pixcode = 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 -= GETJSAMPLE(colormap_ci[pixcode]);
+        cur -= 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.
@@ -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[GETJSAMPLE(ptr[0]) >> C0_SHIFT]
-                        [GETJSAMPLE(ptr[1]) >> C1_SHIFT]
-                        [GETJSAMPLE(ptr[2]) >> C2_SHIFT];
+      histp = &histogram[ptr[0] >> C0_SHIFT]
+                        [ptr[1] >> C1_SHIFT]
+                        [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 = GETJSAMPLE(cinfo->colormap[0][i]);
+    x = 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 = GETJSAMPLE(cinfo->colormap[1][i]);
+    x = 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 = GETJSAMPLE(cinfo->colormap[2][i]);
+    x = 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 = GETJSAMPLE(colorlist[i]);
+    icolor = colorlist[i];
    /* Compute (square of) distance from minc0/c1/c2 to this color */
-    inc0 = (minc0 - GETJSAMPLE(cinfo->colormap[0][icolor])) * C0_SCALE;
+    inc0 = (minc0 - cinfo->colormap[0][icolor]) * C0_SCALE;
    dist0 = inc0 * inc0;
-    inc1 = (minc1 - GETJSAMPLE(cinfo->colormap[1][icolor])) * C1_SCALE;
+    inc1 = (minc1 - cinfo->colormap[1][icolor]) * C1_SCALE;
    dist0 += inc1 * inc1;
-    inc2 = (minc2 - GETJSAMPLE(cinfo->colormap[2][icolor])) * C2_SCALE;
+    inc2 = (minc2 - 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)(GETJSAMPLE(*cptr++) + 1);
+        *cachep++ = (histcell)((*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 = GETJSAMPLE(*inptr++) >> C0_SHIFT;
-      c1 = GETJSAMPLE(*inptr++) >> C1_SHIFT;
-      c2 = GETJSAMPLE(*inptr++) >> C2_SHIFT;
+      c0 = (*inptr++) >> C0_SHIFT;
+      c1 = (*inptr++) >> C1_SHIFT;
+      c2 = (*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 += 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]);
+      cur0 += inptr[0];
+      cur1 += inptr[1];
+      cur2 += inptr[2];
+      cur0 = range_limit[cur0];
+      cur1 = range_limit[cur1];
+      cur2 = 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 -= GETJSAMPLE(colormap0[pixcode]);
-        cur1 -= GETJSAMPLE(colormap1[pixcode]);
-        cur2 -= GETJSAMPLE(colormap2[pixcode]);
+        cur0 -= colormap0[pixcode];
+        cur1 -= colormap1[pixcode];
+        cur2 -= colormap2[pixcode];
      }
      /* Compute error fractions to be propagated to adjacent pixels.
       * Add these into the running sums, and simultaneously shift the
@@ -4,6 +4,7 @@
 * 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.
@@ -75,6 +76,7 @@ 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,
@@ -84,6 +86,10 @@ 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);
@@ -4,6 +4,7 @@
 * 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.
@@ -169,6 +170,12 @@ 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)
@@ -181,6 +188,12 @@ 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) 1991-2012, Thomas G. Lane, Guido Vollbeding.
+ * Copyright (C) 1991-2020, Thomas G. Lane, Guido Vollbeding.
 * libjpeg-turbo Modifications:
- * Copyright (C) 2010, 2012-2020, D. R. Commander.
+ * Copyright (C) 2010, 2012-2021, D. R. Commander.
 * For conditions of distribution and use, see the accompanying README.ijg
 * file.
 *
@@ -37,9 +37,9 @@
 */

 #define JCOPYRIGHT \
-  "Copyright (C) 2009-2020 D. R. Commander\n" \
+  "Copyright (C) 2009-2021 D. R. Commander\n" \
  "Copyright (C) 2015, 2020 Google, Inc.\n" \
-  "Copyright (C) 2019 Arm Limited\n" \
+  "Copyright (C) 2019-2020 Arm Limited\n" \
  "Copyright (C) 2015-2016, 2018 Matthieu Darbois\n" \
  "Copyright (C) 2011-2016 Siarhei Siamashka\n" \
  "Copyright (C) 2015 Intel Corporation\n" \
@@ -48,7 +48,7 @@
  "Copyright (C) 2009, 2012 Pierre Ossman for Cendio AB\n" \
  "Copyright (C) 2009-2011 Nokia Corporation and/or its subsidiary(-ies)\n" \
  "Copyright (C) 1999-2006 MIYASAKA Masaru\n" \
-  "Copyright (C) 1991-2017 Thomas G. Lane, Guido Vollbeding"
+  "Copyright (C) 1991-2020 Thomas G. Lane, Guido Vollbeding"

 #define JCOPYRIGHT_SHORT \
-  "Copyright (C) 1991-2020 The libjpeg-turbo Project and many others"
+  "Copyright (C) 1991-2021 The libjpeg-turbo Project and many others"
@@ -923,6 +923,11 @@ int ovx_hal_cvtGraytoBGR(const uchar * a, size_t astep, uchar * b, size_t bstep,
 }

 int ovx_hal_cvtTwoPlaneYUVtoBGR(const uchar * a, size_t astep, uchar * b, size_t bstep, int w, int h, int bcn, bool swapBlue, int uIdx)
+{
+    return ovx_hal_cvtTwoPlaneYUVtoBGREx(a, astep, a + h * astep, astep, b, bstep, w, h, bcn, swapBlue, uIdx);
+}
+
+int ovx_hal_cvtTwoPlaneYUVtoBGREx(const uchar * a, size_t astep, const uchar * b, size_t bstep, uchar * c, size_t cstep, int w, int h, int bcn, bool swapBlue, int uIdx)
 {
    if (skipSmallImages<VX_KERNEL_COLOR_CONVERT>(w, h))
        return CV_HAL_ERROR_NOT_IMPLEMENTED;
@@ -933,8 +938,7 @@ int ovx_hal_cvtTwoPlaneYUVtoBGR(const uchar * a, size_t astep, uchar * b, size_t

    if (w & 1 || h & 1) // It's not described in spec but sample implementation unable to convert odd sized images
        return CV_HAL_ERROR_NOT_IMPLEMENTED;
-    refineStep(w, h, uIdx ? VX_DF_IMAGE_NV21 : VX_DF_IMAGE_NV12, astep);
-    refineStep(w, h, bcn == 3 ? VX_DF_IMAGE_RGB : VX_DF_IMAGE_RGBX, bstep);
+
    try
    {
        ivx::Context ctx = getOpenVXHALContext();
@@ -943,8 +947,8 @@ int ovx_hal_cvtTwoPlaneYUVtoBGR(const uchar * a, size_t astep, uchar * b, size_t
        std::vector<void *> ptrs;
            addr.push_back(ivx::Image::createAddressing(w, h, 1, (vx_int32)astep));
            ptrs.push_back((void*)a);
-            addr.push_back(ivx::Image::createAddressing(w / 2, h / 2, 2, (vx_int32)astep));
-            ptrs.push_back((void*)(a + h * astep));
+            addr.push_back(ivx::Image::createAddressing(w / 2, h / 2, 2, (vx_int32)bstep));
+            ptrs.push_back((void*)b);

        vxImage
            ia = ivx::Image::createFromHandle(ctx, uIdx ? VX_DF_IMAGE_NV21 : VX_DF_IMAGE_NV12, addr, ptrs);
@@ -952,7 +956,7 @@ int ovx_hal_cvtTwoPlaneYUVtoBGR(const uchar * a, size_t astep, uchar * b, size_t
            return CV_HAL_ERROR_NOT_IMPLEMENTED; // OpenCV store NV12/NV21 as RANGE_RESTRICTED while OpenVX expect RANGE_FULL
        vxImage
            ib = ivx::Image::createFromHandle(ctx, bcn == 3 ? VX_DF_IMAGE_RGB : VX_DF_IMAGE_RGBX,
-                ivx::Image::createAddressing(w, h, bcn, (vx_int32)bstep), b);
+                ivx::Image::createAddressing(w, h, bcn, (vx_int32)cstep), c);
        ivx::IVX_CHECK_STATUS(vxuColorConvert(ctx, ia, ib));
    }
    catch (ivx::RuntimeError & e)
@@ -49,6 +49,7 @@ int ovx_hal_morph(cvhalFilter2D *filter_context, uchar *a, size_t astep, uchar *
 int ovx_hal_cvtBGRtoBGR(const uchar * a, size_t astep, uchar * b, size_t bstep, int w, int h, int depth, int acn, int bcn, bool swapBlue);
 int ovx_hal_cvtGraytoBGR(const uchar * a, size_t astep, uchar * b, size_t bstep, int w, int h, int depth, int bcn);
 int ovx_hal_cvtTwoPlaneYUVtoBGR(const uchar * a, size_t astep, uchar * b, size_t bstep, int w, int h, int bcn, bool swapBlue, int uIdx);
+int ovx_hal_cvtTwoPlaneYUVtoBGREx(const uchar * a, size_t astep, const uchar * b, size_t bstep, uchar * c, size_t cstep, int w, int h, int bcn, bool swapBlue, int uIdx);
 int ovx_hal_cvtThreePlaneYUVtoBGR(const uchar * a, size_t astep, uchar * b, size_t bstep, int w, int h, int bcn, bool swapBlue, int uIdx);
 int ovx_hal_cvtBGRtoThreePlaneYUV(const uchar * a, size_t astep, uchar * b, size_t bstep, int w, int h, int acn, bool swapBlue, int uIdx);
 int ovx_hal_cvtOnePlaneYUVtoBGR(const uchar * a, size_t astep, uchar * b, size_t bstep, int w, int h, int bcn, bool swapBlue, int uIdx, int ycn);
@@ -130,6 +131,8 @@ int ovx_hal_integral(int depth, int sdepth, int, const uchar * a, size_t astep,
 #define cv_hal_cvtGraytoBGR ovx_hal_cvtGraytoBGR
 #undef cv_hal_cvtTwoPlaneYUVtoBGR
 #define cv_hal_cvtTwoPlaneYUVtoBGR ovx_hal_cvtTwoPlaneYUVtoBGR
+#undef cv_hal_cvtTwoPlaneYUVtoBGREx
+#define cv_hal_cvtTwoPlaneYUVtoBGREx ovx_hal_cvtTwoPlaneYUVtoBGREx
 #undef cv_hal_cvtThreePlaneYUVtoBGR
 #define cv_hal_cvtThreePlaneYUVtoBGR ovx_hal_cvtThreePlaneYUVtoBGR
 #undef cv_hal_cvtBGRtoThreePlaneYUV
@@ -31,7 +31,7 @@ libpng                Portable Network Graphics library.
 libtiff               Tag Image File Format (TIFF) Software
                      Copyright (c) 1988-1997 Sam Leffler
                      Copyright (c) 1991-1997 Silicon Graphics, Inc.
-                      See libtiff home page http://www.remotesensing.org/libtiff/
+                      See libtiff home page http://www.libtiff.org/
                      for details and links to the source code

                      WITH_TIFF CMake option must be ON to add libtiff & zlib support to imgcodecs.
@@ -51,7 +51,9 @@ jasper                JasPer is a collection of software
                      Copyright (c) 1999-2000 The University of British Columbia
                      Copyright (c) 2001-2003 Michael David Adams

-                      The JasPer license can be found in libjasper.
+                      See JasPer official GitHub repository
+                      https://github.com/jasper-software/jasper.git
+                      for details and links to source code
 ------------------------------------------------------------------------------------
 openexr               OpenEXR is a high dynamic-range (HDR) image file format developed
                      by Industrial Light & Magic for use in computer imaging applications.
@@ -98,6 +98,12 @@ ocv_cmake_hook(CMAKE_INIT)

 # must go before the project()/enable_language() commands
 ocv_update(CMAKE_CONFIGURATION_TYPES "Debug;Release" CACHE STRING "Configs" FORCE)
+if(NOT DEFINED CMAKE_BUILD_TYPE
+    AND NOT OPENCV_SKIP_DEFAULT_BUILD_TYPE
+)
+  message(STATUS "'Release' build type is used by default. Use CMAKE_BUILD_TYPE to specify build type (Release or Debug)")
+  set(CMAKE_BUILD_TYPE "Release" CACHE STRING "Choose the type of build")
+endif()
 if(DEFINED CMAKE_BUILD_TYPE)
  set_property(CACHE CMAKE_BUILD_TYPE PROPERTY STRINGS "${CMAKE_CONFIGURATION_TYPES}")
 endif()
@@ -107,7 +113,7 @@ set(CMAKE_POSITION_INDEPENDENT_CODE ${ENABLE_PIC})

 ocv_cmake_hook(PRE_CMAKE_BOOTSTRAP)

-# Bootstap CMake system: setup CMAKE_SYSTEM_NAME and other vars
+# Bootstrap CMake system: setup CMAKE_SYSTEM_NAME and other vars
 enable_language(CXX C)

 ocv_cmake_hook(POST_CMAKE_BOOTSTRAP)
@@ -218,7 +224,7 @@ OCV_OPTION(BUILD_TIFF               "Build libtiff from source"          (WIN32
 OCV_OPTION(BUILD_JASPER             "Build libjasper from source"        (WIN32 OR ANDROID OR APPLE OR OPENCV_FORCE_3RDPARTY_BUILD) )
 OCV_OPTION(BUILD_JPEG               "Build libjpeg from source"          (WIN32 OR ANDROID OR APPLE OR OPENCV_FORCE_3RDPARTY_BUILD) )
 OCV_OPTION(BUILD_PNG                "Build libpng from source"           (WIN32 OR ANDROID OR APPLE OR OPENCV_FORCE_3RDPARTY_BUILD) )
-OCV_OPTION(BUILD_OPENEXR            "Build openexr from source"          (((WIN32 OR ANDROID OR APPLE) AND NOT WINRT) OR OPENCV_FORCE_3RDPARTY_BUILD) )
+OCV_OPTION(BUILD_OPENEXR            "Build openexr from source"          (OPENCV_FORCE_3RDPARTY_BUILD) )
 OCV_OPTION(BUILD_WEBP               "Build WebP from source"             (((WIN32 OR ANDROID OR APPLE) AND NOT WINRT) OR OPENCV_FORCE_3RDPARTY_BUILD) )
 OCV_OPTION(BUILD_TBB                "Download and build TBB from source" (ANDROID OR OPENCV_FORCE_3RDPARTY_BUILD) )
 OCV_OPTION(BUILD_IPP_IW             "Build IPP IW from source"           (NOT MINGW OR OPENCV_FORCE_3RDPARTY_BUILD) IF (X86_64 OR X86) AND NOT WINRT )
@@ -300,7 +306,7 @@ OCV_OPTION(WITH_JPEG "Include JPEG support" ON
 OCV_OPTION(WITH_WEBP "Include WebP support" ON
  VISIBLE_IF NOT WINRT
  VERIFY HAVE_WEBP)
-OCV_OPTION(WITH_OPENEXR "Include ILM support via OpenEXR" BUILD_OPENEXR OR NOT CMAKE_CROSSCOMPILING
+OCV_OPTION(WITH_OPENEXR "Include ILM support via OpenEXR" ((WIN32 OR ANDROID OR APPLE) OR BUILD_OPENEXR) OR NOT CMAKE_CROSSCOMPILING
  VISIBLE_IF NOT APPLE_FRAMEWORK AND NOT WINRT
  VERIFY HAVE_OPENEXR)
 OCV_OPTION(WITH_OPENGL "Include OpenGL support" OFF
@@ -599,10 +605,6 @@ endif()
 # ----------------------------------------------------------------------------
 # OpenCV compiler and linker options
 # ----------------------------------------------------------------------------
-# In case of Makefiles if the user does not setup CMAKE_BUILD_TYPE, assume it's Release:
-if(CMAKE_GENERATOR MATCHES "Makefiles|Ninja" AND "${CMAKE_BUILD_TYPE}" STREQUAL "")
-  set(CMAKE_BUILD_TYPE Release)
-endif()

 ocv_cmake_hook(POST_CMAKE_BUILD_OPTIONS)

@@ -648,6 +650,8 @@ if(UNIX)
      set(OPENCV_LINKER_LIBS ${OPENCV_LINKER_LIBS} m pthread)
    elseif(EMSCRIPTEN)
      # no need to link to system libs with emscripten
+    elseif(QNXNTO)
+      set(OPENCV_LINKER_LIBS ${OPENCV_LINKER_LIBS} m)
    else()
      set(OPENCV_LINKER_LIBS ${OPENCV_LINKER_LIBS} dl m pthread rt)
    endif()
@@ -1000,6 +1004,12 @@ if(COMMAND ocv_pylint_finalize)
  ocv_pylint_add_directory_recurse(${CMAKE_CURRENT_LIST_DIR}/samples/python/tutorial_code)
  ocv_pylint_finalize()
 endif()
+if(TARGET check_pylint)
+  message(STATUS "Registered 'check_pylint' target: using ${PYLINT_EXECUTABLE} (ver: ${PYLINT_VERSION}), checks: ${PYLINT_TOTAL_TARGETS}")
+endif()
+if(TARGET check_flake8)
+  message(STATUS "Registered 'check_flake8' target: using ${FLAKE8_EXECUTABLE} (ver: ${FLAKE8_VERSION})")
+endif()

 if(OPENCV_GENERATE_SETUPVARS)
  include(cmake/OpenCVGenSetupVars.cmake)
@@ -1224,12 +1234,17 @@ status("")
 status("  GUI: ")

 if(WITH_QT OR HAVE_QT)
-  if(HAVE_QT5)
-    status("    QT:" "YES (ver ${Qt5Core_VERSION_STRING})")
-    status("      QT OpenGL support:" HAVE_QT_OPENGL THEN "YES (${Qt5OpenGL_LIBRARIES} ${Qt5OpenGL_VERSION_STRING})" ELSE NO)
-  elseif(HAVE_QT)
+  if(HAVE_QT)
    status("    QT:" "YES (ver ${QT_VERSION_MAJOR}.${QT_VERSION_MINOR}.${QT_VERSION_PATCH} ${QT_EDITION})")
-    status("      QT OpenGL support:" HAVE_QT_OPENGL THEN "YES (${QT_QTOPENGL_LIBRARY})" ELSE NO)
+    if(HAVE_QT_OPENGL)
+      if(Qt${QT_VERSION_MAJOR}OpenGL_LIBRARIES)
+        status("      QT OpenGL support:" HAVE_QT_OPENGL THEN "YES (${Qt${QT_VERSION_MAJOR}OpenGL_LIBRARIES} ${Qt${QT_VERSION_MAJOR}OpenGL_VERSION_STRING})" ELSE NO)
+      else()
+        status("      QT OpenGL support:" HAVE_QT_OPENGL THEN "YES (${QT_QTOPENGL_LIBRARY})" ELSE NO)
+      endif()
+    else()
+      status("      QT OpenGL support:" "NO")
+    endif()
  else()
    status("    QT:" "NO")
  endif()
@@ -1633,12 +1648,6 @@ endif()

 status("")
 status("  Python (for build):"  PYTHON_DEFAULT_AVAILABLE THEN "${PYTHON_DEFAULT_EXECUTABLE}" ELSE NO)
-if(PYLINT_FOUND AND PYLINT_EXECUTABLE)
-  status("    Pylint:"  PYLINT_FOUND THEN "${PYLINT_EXECUTABLE} (ver: ${PYLINT_VERSION}, checks: ${PYLINT_TOTAL_TARGETS})" ELSE NO)
-endif()
-if(FLAKE8_FOUND AND FLAKE8_EXECUTABLE)
-  status("    Flake8:"  FLAKE8_FOUND THEN "${FLAKE8_EXECUTABLE} (ver: ${FLAKE8_VERSION})" ELSE NO)
-endif()

 # ========================== java ==========================
 if(BUILD_JAVA)
@@ -16,7 +16,7 @@ if(PYLINT_EXECUTABLE AND NOT DEFINED PYLINT_VERSION)
  execute_process(COMMAND ${PYLINT_EXECUTABLE} --version RESULT_VARIABLE _result OUTPUT_VARIABLE PYLINT_VERSION_RAW)
  if(NOT _result EQUAL 0)
    ocv_clear_vars(PYLINT_EXECUTABLE PYLINT_VERSION)
-  elseif(PYLINT_VERSION_RAW MATCHES "pylint([^,]*) ([0-9\\.]+[0-9])")
+  elseif(PYLINT_VERSION_RAW MATCHES "pylint([^,\n]*) ([0-9\\.]+[0-9])")
    set(PYLINT_VERSION "${CMAKE_MATCH_2}")
  else()
    set(PYLINT_VERSION "unknown")
@@ -87,11 +87,3 @@ endif()
 set( CMAKE_SHARED_LINKER_FLAGS "${CMAKE_SHARED_LINKER_FLAGS} ${OPENCV_LINKER_DEFENSES_FLAGS_COMMON}" )
 set( CMAKE_MODULE_LINKER_FLAGS "${CMAKE_MODULE_LINKER_FLAGS} ${OPENCV_LINKER_DEFENSES_FLAGS_COMMON}" )
 set( CMAKE_EXE_LINKER_FLAGS "${CMAKE_EXE_LINKER_FLAGS} ${OPENCV_LINKER_DEFENSES_FLAGS_COMMON}" )
-
-if(CV_GCC OR CV_CLANG)
-  foreach(flags
-          CMAKE_CXX_FLAGS CMAKE_CXX_FLAGS_RELEASE CMAKE_CXX_FLAGS_DEBUG
-          CMAKE_C_FLAGS CMAKE_C_FLAGS_RELEASE CMAKE_C_FLAGS_DEBUG)
-    string(REPLACE "-O3" "-O2" ${flags} "${${flags}}")
-  endforeach()
-endif()
@@ -238,7 +238,7 @@ if(X86 OR X86_64)
    ocv_intel_compiler_optimization_option(FP16 "-mavx" "/arch:AVX")
    ocv_intel_compiler_optimization_option(AVX "-mavx" "/arch:AVX")
    ocv_intel_compiler_optimization_option(FMA3 "" "")
-    ocv_intel_compiler_optimization_option(POPCNT "" "")
+    ocv_intel_compiler_optimization_option(POPCNT "-mpopcnt" "")  # -mpopcnt is available since ICC 19.0.0
    ocv_intel_compiler_optimization_option(SSE4_2 "-msse4.2" "/arch:SSE4.2")
    ocv_intel_compiler_optimization_option(SSE4_1 "-msse4.1" "/arch:SSE4.1")
    ocv_intel_compiler_optimization_option(SSE3 "-msse3" "/arch:SSE3")
@@ -177,7 +177,13 @@ if(CV_GCC OR CV_CLANG)
  endif()

  # We need pthread's
-  if(UNIX AND NOT ANDROID AND NOT (APPLE AND CV_CLANG)) # TODO
+  if((UNIX
+          AND NOT ANDROID
+          AND NOT (APPLE AND CV_CLANG)
+          AND NOT EMSCRIPTEN
+      )
+      OR (EMSCRIPTEN AND WITH_PTHREADS_PF)  # https://github.com/opencv/opencv/issues/20285
+  )
    add_extra_compiler_option(-pthread)
  endif()

@@ -225,9 +231,11 @@ if(CV_GCC OR CV_CLANG)
        if(APPLE)
          set(OPENCV_EXTRA_EXE_LINKER_FLAGS "${OPENCV_EXTRA_EXE_LINKER_FLAGS} -Wl,-dead_strip")
          set(OPENCV_EXTRA_SHARED_LINKER_FLAGS "${OPENCV_EXTRA_SHARED_LINKER_FLAGS} -Wl,-dead_strip")
+          set(OPENCV_EXTRA_MODULE_LINKER_FLAGS "${OPENCV_EXTRA_MODULE_LINKER_FLAGS} -Wl,-dead_strip")
        else()
          set(OPENCV_EXTRA_EXE_LINKER_FLAGS "${OPENCV_EXTRA_EXE_LINKER_FLAGS} -Wl,--gc-sections")
          set(OPENCV_EXTRA_SHARED_LINKER_FLAGS "${OPENCV_EXTRA_SHARED_LINKER_FLAGS} -Wl,--gc-sections")
+          set(OPENCV_EXTRA_MODULE_LINKER_FLAGS "${OPENCV_EXTRA_MODULE_LINKER_FLAGS} -Wl,--gc-sections")
        endif()
      endif()
    endif()
@@ -279,6 +287,7 @@ if(MSVC)
    set(OPENCV_EXTRA_FLAGS_RELEASE "${OPENCV_EXTRA_FLAGS_RELEASE} /Zi")
    set(OPENCV_EXTRA_EXE_LINKER_FLAGS_RELEASE "${OPENCV_EXTRA_EXE_LINKER_FLAGS_RELEASE} /debug")
    set(OPENCV_EXTRA_SHARED_LINKER_FLAGS_RELEASE "${OPENCV_EXTRA_SHARED_LINKER_FLAGS_RELEASE} /debug")
+    set(OPENCV_EXTRA_MODULE_LINKER_FLAGS_RELEASE "${OPENCV_EXTRA_MODULE_LINKER_FLAGS_RELEASE} /debug")
  endif()

  # Remove unreferenced functions: function level linking
@@ -346,6 +355,7 @@ if(NOT OPENCV_SKIP_LINK_AS_NEEDED)
    if(HAVE_LINK_AS_NEEDED)
      set(OPENCV_EXTRA_EXE_LINKER_FLAGS "${OPENCV_EXTRA_EXE_LINKER_FLAGS} ${_option}")
      set(OPENCV_EXTRA_SHARED_LINKER_FLAGS "${OPENCV_EXTRA_SHARED_LINKER_FLAGS} ${_option}")
+      set(OPENCV_EXTRA_MODULE_LINKER_FLAGS "${OPENCV_EXTRA_MODULE_LINKER_FLAGS} ${_option}")
    endif()
  endif()
 endif()
@@ -364,6 +374,9 @@ if(NOT OPENCV_SKIP_EXTRA_COMPILER_FLAGS)
  set(CMAKE_SHARED_LINKER_FLAGS         "${CMAKE_SHARED_LINKER_FLAGS} ${OPENCV_EXTRA_SHARED_LINKER_FLAGS}")
  set(CMAKE_SHARED_LINKER_FLAGS_RELEASE "${CMAKE_SHARED_LINKER_FLAGS_RELEASE} ${OPENCV_EXTRA_SHARED_LINKER_FLAGS_RELEASE}")
  set(CMAKE_SHARED_LINKER_FLAGS_DEBUG   "${CMAKE_SHARED_LINKER_FLAGS_DEBUG} ${OPENCV_EXTRA_SHARED_LINKER_FLAGS_DEBUG}")
+  set(CMAKE_MODULE_LINKER_FLAGS         "${CMAKE_MODULE_LINKER_FLAGS} ${OPENCV_EXTRA_MODULE_LINKER_FLAGS}")
+  set(CMAKE_MODULE_LINKER_FLAGS_RELEASE "${CMAKE_MODULE_LINKER_FLAGS_RELEASE} ${OPENCV_EXTRA_MODULE_LINKER_FLAGS_RELEASE}")
+  set(CMAKE_MODULE_LINKER_FLAGS_DEBUG   "${CMAKE_MODULE_LINKER_FLAGS_DEBUG} ${OPENCV_EXTRA_MODULE_LINKER_FLAGS_DEBUG}")
 endif()

 if(MSVC)
@@ -387,6 +400,9 @@ if(MSVC)
    endif()
  endif()

+  # Enable [[attribute]] syntax checking to prevent silent failure: "attribute is ignored in this syntactic position"
+  add_extra_compiler_option("/w15240")
+
  if(NOT ENABLE_NOISY_WARNINGS)
    ocv_warnings_disable(CMAKE_CXX_FLAGS /wd4127) # conditional expression is constant
    ocv_warnings_disable(CMAKE_CXX_FLAGS /wd4251) # class 'std::XXX' needs to have dll-interface to be used by clients of YYY
@@ -102,7 +102,7 @@ if(CUDA_FOUND)
  if(CUDA_GENERATION)
    if(NOT ";${_generations};" MATCHES ";${CUDA_GENERATION};")
      string(REPLACE ";" ", " _generations "${_generations}")
-      message(FATAL_ERROR "ERROR: ${_generations} Generations are suppered.")
+      message(FATAL_ERROR "ERROR: ${_generations} Generations are supported.")
    endif()
    unset(CUDA_ARCH_BIN CACHE)
    unset(CUDA_ARCH_PTX CACHE)
@@ -169,6 +169,8 @@ elseif(MSVC)
    set(OpenCV_RUNTIME vc15)
  elseif(MSVC_VERSION MATCHES "^192[0-9]$")
    set(OpenCV_RUNTIME vc16)
+  elseif(MSVC_VERSION MATCHES "^193[0-9]$")
+    set(OpenCV_RUNTIME vc17)
  else()
    message(WARNING "OpenCV does not recognize MSVC_VERSION \"${MSVC_VERSION}\". Cannot set OpenCV_RUNTIME")
  endif()
@@ -9,9 +9,14 @@ set(HALIDE_ROOT_DIR "${HALIDE_ROOT_DIR}" CACHE PATH "Halide root directory")
 if(NOT HAVE_HALIDE)
  find_package(Halide QUIET) # Try CMake-based config files
  if(Halide_FOUND)
-    set(HALIDE_INCLUDE_DIRS "${Halide_INCLUDE_DIRS}" CACHE PATH "Halide include directories" FORCE)
-    set(HALIDE_LIBRARIES "${Halide_LIBRARIES}" CACHE PATH "Halide libraries" FORCE)
-    set(HAVE_HALIDE TRUE)
+    if(TARGET Halide::Halide)  # modern Halide scripts defines imported target
+      set(HALIDE_INCLUDE_DIRS "")
+      set(HALIDE_LIBRARIES "Halide::Halide")
+      set(HAVE_HALIDE TRUE)
+    else()
+      # using HALIDE_INCLUDE_DIRS / Halide_LIBRARIES
+      set(HAVE_HALIDE TRUE)
+    endif()
  endif()
 endif()

@@ -28,18 +33,15 @@ if(NOT HAVE_HALIDE AND HALIDE_ROOT_DIR)
  )
  if(HALIDE_LIBRARY AND HALIDE_INCLUDE_DIR)
    # TODO try_compile
-    set(HALIDE_INCLUDE_DIRS "${HALIDE_INCLUDE_DIR}" CACHE PATH "Halide include directories" FORCE)
-    set(HALIDE_LIBRARIES "${HALIDE_LIBRARY}" CACHE PATH "Halide libraries" FORCE)
+    set(HALIDE_INCLUDE_DIRS "${HALIDE_INCLUDE_DIR}")
+    set(HALIDE_LIBRARIES "${HALIDE_LIBRARY}")
    set(HAVE_HALIDE TRUE)
  endif()
-  if(NOT HAVE_HALIDE)
-    ocv_clear_vars(HALIDE_LIBRARIES HALIDE_INCLUDE_DIRS CACHE)
-  endif()
 endif()

 if(HAVE_HALIDE)
-  include_directories(${HALIDE_INCLUDE_DIRS})
+  if(HALIDE_INCLUDE_DIRS)
+    include_directories(${HALIDE_INCLUDE_DIRS})
+  endif()
  list(APPEND OPENCV_LINKER_LIBS ${HALIDE_LIBRARIES})
-else()
-  ocv_clear_vars(HALIDE_INCLUDE_DIRS HALIDE_LIBRARIES)
 endif()
@@ -105,6 +105,20 @@ if(InferenceEngine_FOUND)
  message(STATUS "Detected InferenceEngine: cmake package (${InferenceEngine_VERSION})")
 endif()

+if(DEFINED InferenceEngine_VERSION)
+  message(STATUS "InferenceEngine: ${InferenceEngine_VERSION}")
+  if(NOT INF_ENGINE_RELEASE AND NOT (InferenceEngine_VERSION VERSION_LESS "2021.4"))
+    math(EXPR INF_ENGINE_RELEASE_INIT "${InferenceEngine_VERSION_MAJOR} * 1000000 + ${InferenceEngine_VERSION_MINOR} * 10000 + ${InferenceEngine_VERSION_PATCH} * 100")
+  endif()
+endif()
+if(NOT INF_ENGINE_RELEASE AND NOT INF_ENGINE_RELEASE_INIT)
+  message(STATUS "WARNING: InferenceEngine version has not been set, 2021.4.2 will be used by default. Set INF_ENGINE_RELEASE variable if you experience build errors.")
+  set(INF_ENGINE_RELEASE_INIT "2021040200")
+elseif(DEFINED INF_ENGINE_RELEASE)
+  set(INF_ENGINE_RELEASE_INIT "${INF_ENGINE_RELEASE}")
+endif()
+set(INF_ENGINE_RELEASE "${INF_ENGINE_RELEASE_INIT}" CACHE STRING "Force IE version, should be in form YYYYAABBCC (e.g. 2020.1.0.2 -> 2020010002)")
+
 if(NOT INF_ENGINE_TARGET AND INF_ENGINE_LIB_DIRS AND INF_ENGINE_INCLUDE_DIRS)
  find_path(ie_custom_inc "inference_engine.hpp" PATHS "${INF_ENGINE_INCLUDE_DIRS}" NO_DEFAULT_PATH)
  if(CMAKE_BUILD_TYPE STREQUAL "Debug")
@@ -140,12 +154,8 @@ endif()
 # Add more features to the target

 if(INF_ENGINE_TARGET)
-  if(NOT INF_ENGINE_RELEASE)
-    message(WARNING "InferenceEngine version has not been set, 2021.2 will be used by default. Set INF_ENGINE_RELEASE variable if you experience build errors.")
-  endif()
-  set(INF_ENGINE_RELEASE "2021020000" CACHE STRING "Force IE version, should be in form YYYYAABBCC (e.g. 2020.1.0.2 -> 2020010002)")
  set_target_properties(${INF_ENGINE_TARGET} PROPERTIES
-    INTERFACE_COMPILE_DEFINITIONS "HAVE_INF_ENGINE=1;INF_ENGINE_RELEASE=${INF_ENGINE_RELEASE}"
+      INTERFACE_COMPILE_DEFINITIONS "HAVE_INF_ENGINE=1;INF_ENGINE_RELEASE=${INF_ENGINE_RELEASE}"
  )
 endif()

@@ -177,7 +177,7 @@ if(NOT ${found})

    if(NOT ANDROID AND NOT IOS)
      if(CMAKE_HOST_UNIX)
-        execute_process(COMMAND ${_executable} -c "from distutils.sysconfig import *; print(get_python_lib())"
+        execute_process(COMMAND ${_executable} -c "from sysconfig import *; print(get_path('purelib'))"
                        RESULT_VARIABLE _cvpy_process
                        OUTPUT_VARIABLE _std_packages_path
                        OUTPUT_STRIP_TRAILING_WHITESPACE)
@@ -26,31 +26,14 @@ if(VTK_VERSION VERSION_LESS "5.8.0")
 endif()

 # Different Qt versions can't be linked together
-if(HAVE_QT5 AND VTK_VERSION VERSION_LESS "6.0.0")
-  if(VTK_USE_QT)
-    message(STATUS "VTK support is disabled. Incompatible combination: OpenCV + Qt5 and VTK ver.${VTK_VERSION} + Qt4")
-  endif()
-endif()
-
-# Different Qt versions can't be linked together. VTK 6.0.0 doesn't provide a way to get Qt version it was linked with
-if(HAVE_QT5 AND VTK_VERSION VERSION_EQUAL "6.0.0" AND NOT DEFINED FORCE_VTK)
-  message(STATUS "VTK support is disabled. Possible incompatible combination: OpenCV+Qt5, and VTK ver.${VTK_VERSION} with Qt4")
-  message(STATUS "If it is known that VTK was compiled without Qt4, please define '-DFORCE_VTK=TRUE' flag in CMake")
+if((HAVE_QT AND VTK_USE_QT)
+    AND NOT DEFINED FORCE_VTK  # deprecated
+    AND NOT DEFINED OPENCV_FORCE_VTK
+)
+  message(STATUS "VTK support is disabled. Possible incompatible combination: OpenCV+Qt, and VTK ver.${VTK_VERSION} with Qt")
+  message(STATUS "If it is known that VTK was compiled without Qt4, please define '-DOPENCV_FORCE_VTK=TRUE' flag in CMake")
  return()
 endif()

-# Different Qt versions can't be linked together
-if(HAVE_QT AND VTK_VERSION VERSION_GREATER "6.0.0" AND NOT ${VTK_QT_VERSION} STREQUAL "")
-  if(HAVE_QT5 AND ${VTK_QT_VERSION} EQUAL "4")
-    message(STATUS "VTK support is disabled. Incompatible combination: OpenCV + Qt5 and VTK ver.${VTK_VERSION} + Qt4")
-    return()
-  endif()
-
-  if(NOT HAVE_QT5 AND ${VTK_QT_VERSION} EQUAL "5")
-    message(STATUS "VTK support is disabled. Incompatible combination: OpenCV + Qt4 and VTK ver.${VTK_VERSION} + Qt5")
-    return()
-  endif()
-endif()
-
 set(HAVE_VTK ON)
 message(STATUS "Found VTK ${VTK_VERSION} (${VTK_USE_FILE})")
@@ -23,7 +23,7 @@ set(OPENCV_DOWNLOAD_LOG "${OpenCV_BINARY_DIR}/CMakeDownloadLog.txt")
 set(OPENCV_DOWNLOAD_WITH_CURL "${OpenCV_BINARY_DIR}/download_with_curl.sh")
 set(OPENCV_DOWNLOAD_WITH_WGET "${OpenCV_BINARY_DIR}/download_with_wget.sh")
 set(OPENCV_DOWNLOAD_TRIES_LIST 1 CACHE STRING "List of download tries") # a list
-set(OPENCV_DOWNLOAD_PARAMS INACTIVITY_TIMEOUT 60 TIMEOUT 600 CACHE STRING "Download parameters to be passed to file(DOWNLAOD ...)")
+set(OPENCV_DOWNLOAD_PARAMS INACTIVITY_TIMEOUT 60 TIMEOUT 600 CACHE STRING "Download parameters to be passed to file(DOWNLOAD ...)")
 mark_as_advanced(OPENCV_DOWNLOAD_TRIES_LIST OPENCV_DOWNLOAD_PARAMS)

 # Init download cache directory and log file and helper scripts
@@ -143,10 +143,25 @@ macro(ipp_detect_version)
        list(APPEND IPP_LIBRARIES ${IPP_LIBRARY_DIR}/${IPP_LIB_PREFIX}${IPP_PREFIX}${name}${IPP_SUFFIX}${IPP_LIB_SUFFIX})
      else ()
        add_library(ipp${name} STATIC IMPORTED)
+        set(_filename "${IPP_LIB_PREFIX}${IPP_PREFIX}${name}${IPP_SUFFIX}${IPP_LIB_SUFFIX}")
        set_target_properties(ipp${name} PROPERTIES
          IMPORTED_LINK_INTERFACE_LIBRARIES ""
-          IMPORTED_LOCATION ${IPP_LIBRARY_DIR}/${IPP_LIB_PREFIX}${IPP_PREFIX}${name}${IPP_SUFFIX}${IPP_LIB_SUFFIX}
+          IMPORTED_LOCATION ${IPP_LIBRARY_DIR}/${_filename}
        )
+        if("${name}" STREQUAL "core")  # https://github.com/opencv/opencv/pull/19681
+          if(OPENCV_FORCE_IPP_EXCLUDE_LIBS OR OPENCV_FORCE_IPP_EXCLUDE_LIBS_CORE
+              OR (UNIX AND NOT ANDROID AND NOT APPLE
+                  AND CMAKE_CXX_COMPILER_ID MATCHES "GNU|Clang|Intel"
+              )
+              AND NOT OPENCV_SKIP_IPP_EXCLUDE_LIBS_CORE
+          )
+            if(CMAKE_VERSION VERSION_LESS "3.13.0")
+              set(CMAKE_SHARED_LINKER_FLAGS "-Wl,--exclude-libs,${_filename} ${CMAKE_SHARED_LINKER_FLAGS}")
+            else()
+              target_link_options(ipp${name} INTERFACE "LINKER:--exclude-libs,${_filename}")
+            endif()
+          endif()
+        endif()
        list(APPEND IPP_LIBRARIES ipp${name})
        if (NOT BUILD_SHARED_LIBS AND (HAVE_IPP_ICV OR ";${OPENCV_INSTALL_EXTERNAL_DEPENDENCIES};" MATCHES ";ipp;"))
          # CMake doesn't support "install(TARGETS ${IPP_PREFIX}${name} " command with imported targets
@@ -51,6 +51,23 @@ macro(ocv_lapack_check)
    if(NOT "${OPENCV_CBLAS_H_PATH_${_lapack_impl}}" STREQUAL "${OPENCV_LAPACKE_H_PATH_${_lapack_impl}}")
      list(APPEND _lapack_content "#include \"${OPENCV_LAPACKE_H_PATH_${_lapack_impl}}\"")
    endif()
+    list(APPEND _lapack_content "
+#if defined(LAPACK_GLOBAL) || defined(LAPACK_NAME)
+/*
+ * Using netlib's reference LAPACK implementation version >= 3.4.0 (first with C interface).
+ * Use LAPACK_xxxx to transparently (via predefined lapack macros) deal with pre and post 3.9.1 versions.
+ * LAPACK 3.9.1 introduces LAPACK_FORTRAN_STRLEN_END and modifies (through preprocessing) the declarations of the following functions used in opencv
+ *        sposv_, dposv_, spotrf_, dpotrf_, sgesdd_, dgesdd_, sgels_, dgels_
+ * which end up with an extra parameter.
+ * So we also need to preprocess the function calls in opencv coding by prefixing them with LAPACK_.
+ * The good news is the preprocessing works fine whatever netlib's LAPACK version.
+ */
+#define OCV_LAPACK_FUNC(f) LAPACK_##f
+#else
+/* Using other LAPACK implementations so fall back to opencv's assumption until now */
+#define OCV_LAPACK_FUNC(f) f##_
+#endif
+")
    if(${_lapack_add_extern_c})
      list(APPEND _lapack_content "}")
    endif()
@@ -11,25 +11,50 @@ if(WITH_WIN32UI)
    CMAKE_FLAGS "-DLINK_LIBRARIES:STRING=user32;gdi32")
 endif()

-# --- QT4 ---
-ocv_clear_vars(HAVE_QT HAVE_QT5)
-if(WITH_QT)
-  if(NOT WITH_QT EQUAL 4)
-    find_package(Qt5 COMPONENTS Core Gui Widgets Test Concurrent REQUIRED NO_MODULE)
-    if(Qt5_FOUND)
-      set(HAVE_QT5 ON)
-      set(HAVE_QT  ON)
-      find_package(Qt5 COMPONENTS OpenGL QUIET)
-      if(Qt5OpenGL_FOUND)
-        set(QT_QTOPENGL_FOUND ON)
-      endif()
-    endif()
+macro(ocv_find_package_Qt4)
+  find_package(Qt4 COMPONENTS QtCore QtGui QtTest ${ARGN})
+  if(QT4_FOUND)
+    set(QT_FOUND 1)
+    ocv_assert(QT_VERSION_MAJOR EQUAL 4)
  endif()
+endmacro()

-  if(NOT HAVE_QT)
-    find_package(Qt4 REQUIRED QtCore QtGui QtTest)
-    if(QT4_FOUND)
-      set(HAVE_QT TRUE)
+macro(ocv_find_package_Qt OCV_QT_VER)
+  find_package(Qt${OCV_QT_VER} COMPONENTS Core Gui Widgets Test Concurrent ${ARGN} NO_MODULE)
+  if(Qt${OCV_QT_VER}_FOUND)
+    set(QT_FOUND 1)
+    set(QT_VERSION "${Qt${OCV_QT_VER}_VERSION}")
+    set(QT_VERSION_MAJOR "${Qt${OCV_QT_VER}_VERSION_MAJOR}")
+    set(QT_VERSION_MINOR "${Qt${OCV_QT_VER}_VERSION_MINOR}")
+    set(QT_VERSION_PATCH "${Qt${OCV_QT_VER}_VERSION_PATCH}")
+    set(QT_VERSION_TWEAK "${Qt${OCV_QT_VER}_VERSION_TWEAK}")
+    set(QT_VERSION_COUNT "${Qt${OCV_QT_VER}_VERSION_COUNT}")
+  endif()
+endmacro()
+
+# --- QT4 ---
+if(WITH_QT)
+  if(NOT WITH_QT GREATER 0)
+    # BUG: Qt5Config.cmake script can't handle components properly: find_package(QT NAMES Qt6 Qt5 REQUIRED NO_MODULE COMPONENTS Core Gui Widgets Test Concurrent)
+    ocv_find_package_Qt(6 QUIET)
+    if(NOT QT_FOUND)
+      ocv_find_package_Qt(5 QUIET)
+    endif()
+    if(NOT QT_FOUND)
+      ocv_find_package_Qt4(QUIET)
+    endif()
+  elseif(WITH_QT EQUAL 4)
+    ocv_find_package_Qt4(REQUIRED)
+  else()  # WITH_QT=<major version>
+    ocv_find_package_Qt("${WITH_QT}" REQUIRED)
+  endif()
+  if(QT_FOUND)
+    set(HAVE_QT ON)
+    if(QT_VERSION_MAJOR GREATER 4)
+      find_package(Qt${QT_VERSION_MAJOR} COMPONENTS OpenGL QUIET)
+      if(Qt${QT_VERSION_MAJOR}OpenGL_FOUND)
+        set(QT_QTOPENGL_FOUND ON)  # HAVE_QT_OPENGL is defined below
+      endif()
    endif()
  endif()
 endif()
@@ -76,7 +101,6 @@ if(WITH_OPENGL)
    find_package (OpenGL QUIET)
    if(OPENGL_FOUND)
      set(HAVE_OPENGL TRUE)
-      list(APPEND OPENCV_LINKER_LIBS ${OPENGL_LIBRARIES})
      if(QT_QTOPENGL_FOUND)
        set(HAVE_QT_OPENGL TRUE)
      else()
@@ -240,7 +240,9 @@ if(WITH_OPENEXR)
    set(OPENEXR_LIBRARIES IlmImf)
    add_subdirectory("${OpenCV_SOURCE_DIR}/3rdparty/openexr")
    if(OPENEXR_VERSION)  # check via TARGET doesn't work
+      set(BUILD_OPENEXR ON)
      set(HAVE_OPENEXR YES)
+      set(BUILD_OPENEXR ON)
    endif()
  endif()
 endif()
@@ -29,7 +29,7 @@ if(WITH_IPP)
    if(OPENCV_FORCE_IPP_EXCLUDE_LIBS
        OR (HAVE_IPP_ICV
            AND UNIX AND NOT ANDROID AND NOT APPLE
-            AND (CMAKE_CXX_COMPILER_ID MATCHES "GNU" OR CMAKE_CXX_COMPILER_ID MATCHES "Clang")
+            AND CMAKE_CXX_COMPILER_ID MATCHES "GNU|Clang|Intel"
        )
        AND NOT OPENCV_SKIP_IPP_EXCLUDE_LIBS
    )
@@ -9,6 +9,27 @@
 # OPENEXR_LIBRARIES = libraries that are needed to use OpenEXR.
 #

+if(NOT HAVE_CXX11)
+  message(STATUS "OpenEXR: enable C++11 to use external OpenEXR")
+  return()
+endif()
+
+if(NOT OPENCV_SKIP_OPENEXR_FIND_PACKAGE)
+  find_package(OpenEXR 3 QUIET)
+  #ocv_cmake_dump_vars(EXR)
+  if(OpenEXR_FOUND)
+    if(TARGET OpenEXR::OpenEXR)  # OpenEXR 3+
+      set(OPENEXR_LIBRARIES OpenEXR::OpenEXR)
+      set(OPENEXR_INCLUDE_PATHS "")
+      set(OPENEXR_VERSION "${OpenEXR_VERSION}")
+      set(OPENEXR_FOUND 1)
+      return()
+    else()
+      message(STATUS "Unsupported find_package(OpenEXR) - missing OpenEXR::OpenEXR target (version ${OpenEXR_VERSION})")
+    endif()
+  endif()
+endif()
+
 SET(OPENEXR_LIBRARIES "")
 SET(OPENEXR_LIBSEARCH_SUFFIXES "")
 file(TO_CMAKE_PATH "$ENV{ProgramFiles}" ProgramFiles_ENV_PATH)
@@ -876,7 +876,11 @@ endmacro()
 macro(_ocv_create_module)

  ocv_compiler_optimization_process_sources(OPENCV_MODULE_${the_module}_SOURCES OPENCV_MODULE_${the_module}_DEPS_EXT ${the_module})
-  set(OPENCV_MODULE_${the_module}_HEADERS ${OPENCV_MODULE_${the_module}_HEADERS} CACHE INTERNAL "List of header files for ${the_module}")
+  set(__module_headers ${OPENCV_MODULE_${the_module}_HEADERS})
+  if(__module_headers)
+    list(SORT __module_headers)  # fix headers order, useful for bindings
+  endif()
+  set(OPENCV_MODULE_${the_module}_HEADERS ${__module_headers} CACHE INTERNAL "List of header files for ${the_module}")
  set(OPENCV_MODULE_${the_module}_SOURCES ${OPENCV_MODULE_${the_module}_SOURCES} CACHE INTERNAL "List of source files for ${the_module}")

  # The condition we ought to be testing here is whether ocv_add_precompiled_headers will
@@ -122,7 +122,6 @@ function(ocv_pylint_finalize)

  list(LENGTH PYLINT_TARGET_ID __total)
  set(PYLINT_TOTAL_TARGETS "${__total}" CACHE INTERNAL "")
-  message(STATUS "Pylint: registered ${__total} targets. Build 'check_pylint' target to run checks (\"cmake --build . --target check_pylint\" or \"make check_pylint\")")
  configure_file("${OpenCV_SOURCE_DIR}/cmake/templates/pylint.cmake.in" "${CMAKE_BINARY_DIR}/pylint.cmake" @ONLY)

  add_custom_target(check_pylint
@@ -1488,8 +1488,8 @@ function(ocv_target_link_libraries target)
      if(NOT LINK_PENDING STREQUAL "")
        __ocv_push_target_link_libraries(${LINK_MODE} ${LINK_PENDING})
        set(LINK_PENDING "")
-        set(LINK_MODE "${dep}")
      endif()
+      set(LINK_MODE "${dep}")
    else()
      if(BUILD_opencv_world)
        if(OPENCV_MODULE_${dep}_IS_PART_OF_WORLD)
@@ -137,6 +137,20 @@ elseif(MSVC)
        set(OpenCV_RUNTIME vc14) # selecting previous compatible runtime version
      endif()
    endif()
+  elseif(MSVC_VERSION MATCHES "^193[0-9]$")
+    set(OpenCV_RUNTIME vc17)
+    check_one_config(has_VS2022)
+    if(NOT has_VS2022)
+     set(OpenCV_RUNTIME vc16)
+     check_one_config(has_VS2019)
+     if(NOT has_VS2019)
+       set(OpenCV_RUNTIME vc15) # selecting previous compatible runtime version
+       check_one_config(has_VS2017)
+       if(NOT has_VS2017)
+         set(OpenCV_RUNTIME vc14) # selecting previous compatible runtime version
+       endif()
+     endif()
+    endif()
  endif()
 elseif(MINGW)
  set(OpenCV_RUNTIME mingw)
@@ -31,7 +31,7 @@ MULTILINE_CPP_IS_BRIEF = NO
 INHERIT_DOCS           = YES
 SEPARATE_MEMBER_PAGES  = NO
 TAB_SIZE               = 4
-ALIASES               += add_toggle{1}="@htmlonly[block] <div class='newInnerHTML'>\1</div><div> <script type="text/javascript"> addToggle(); </script>@endhtmlonly"
+ALIASES               += add_toggle{1}="@htmlonly[block] <div class='newInnerHTML'>\1</div><div> <script type='text/javascript'> addToggle(); </script>@endhtmlonly"
 ALIASES               += add_toggle_cpp="@htmlonly[block] <div class='newInnerHTML' title='cpp' style='display: none;'>C++</div><div class='toggleable_div label_cpp' style='display: none;'>@endhtmlonly"
 ALIASES               += add_toggle_java="@htmlonly[block] <div class='newInnerHTML' title='java' style='display: none;'>Java</div><div class='toggleable_div label_java' style='display: none;'>@endhtmlonly"
 ALIASES               += add_toggle_python="@htmlonly[block] <div class='newInnerHTML' title='python' style='display: none;'>Python</div><div class='toggleable_div label_python' style='display: none;'>@endhtmlonly"
@@ -249,6 +249,12 @@ PREDEFINED             = __cplusplus=1 \
                         CV_DEFAULT(x)=" = x" \
                         CV_NEON=1 \
                         CV_SSE2=1 \
+                         CV_SIMD128=1 \
+                         CV_SIMD256=1 \
+                         CV_SIMD512=1 \
+                         CV_SIMD128_64F=1 \
+                         CV_SIMD256_64F=1 \
+                         CV_SIMD512_64F=1 \
                         CV__DEBUG_NS_BEGIN= \
                         CV__DEBUG_NS_END= \
                         CV_DEPRECATED_EXTERNAL= \
@@ -77,7 +77,7 @@ cv.normalize(roiHist, roiHist, 0, 255, cv.NORM_MINMAX);
 // delete useless mats.
 roi.delete(); hsvRoi.delete(); mask.delete(); low.delete(); high.delete(); hsvRoiVec.delete();

-// Setup the termination criteria, either 10 iteration or move by atleast 1 pt
+// Setup the termination criteria, either 10 iteration or move by at least 1 pt
 let termCrit = new cv.TermCriteria(cv.TERM_CRITERIA_EPS | cv.TERM_CRITERIA_COUNT, 10, 1);

 let hsv = new cv.Mat(video.height, video.width, cv.CV_8UC3);
@@ -116,7 +116,7 @@ swapRB = false;
 needSoftmax = false;

 // url for label file, can from local or Internet
-labelsUrl = "https://raw.githubusercontent.com/opencv/opencv/master/samples/data/dnn/classification_classes_ILSVRC2012.txt";
+labelsUrl = "https://raw.githubusercontent.com/opencv/opencv/3.4/samples/data/dnn/classification_classes_ILSVRC2012.txt";
 </script>

 <script id="codeSnippet1" type="text/code-snippet">
@@ -6,7 +6,7 @@
            "std": "1",
            "swapRB": "false",
            "needSoftmax": "false",
-            "labelsUrl": "https://raw.githubusercontent.com/opencv/opencv/master/samples/data/dnn/classification_classes_ILSVRC2012.txt",
+            "labelsUrl": "https://raw.githubusercontent.com/opencv/opencv/3.4/samples/data/dnn/classification_classes_ILSVRC2012.txt",
            "modelUrl": "http://dl.caffe.berkeleyvision.org/bvlc_alexnet.caffemodel",
            "configUrl": "https://raw.githubusercontent.com/BVLC/caffe/master/models/bvlc_alexnet/deploy.prototxt"
        },
@@ -16,7 +16,7 @@
            "std": "0.007843",
            "swapRB": "false",
            "needSoftmax": "true",
-            "labelsUrl": "https://raw.githubusercontent.com/opencv/opencv/master/samples/data/dnn/classification_classes_ILSVRC2012.txt",
+            "labelsUrl": "https://raw.githubusercontent.com/opencv/opencv/3.4/samples/data/dnn/classification_classes_ILSVRC2012.txt",
            "modelUrl": "https://drive.google.com/open?id=0B7ubpZO7HnlCcHlfNmJkU2VPelE",
            "configUrl": "https://raw.githubusercontent.com/shicai/DenseNet-Caffe/master/DenseNet_121.prototxt"
        },
@@ -26,7 +26,7 @@
            "std": "1",
            "swapRB": "false",
            "needSoftmax": "false",
-            "labelsUrl": "https://raw.githubusercontent.com/opencv/opencv/master/samples/data/dnn/classification_classes_ILSVRC2012.txt",
+            "labelsUrl": "https://raw.githubusercontent.com/opencv/opencv/3.4/samples/data/dnn/classification_classes_ILSVRC2012.txt",
            "modelUrl": "http://dl.caffe.berkeleyvision.org/bvlc_googlenet.caffemodel",
            "configUrl": "https://raw.githubusercontent.com/BVLC/caffe/master/models/bvlc_googlenet/deploy.prototxt"
        },
@@ -36,7 +36,7 @@
            "std": "1",
            "swapRB": "false",
            "needSoftmax": "false",
-            "labelsUrl": "https://raw.githubusercontent.com/opencv/opencv/master/samples/data/dnn/classification_classes_ILSVRC2012.txt",
+            "labelsUrl": "https://raw.githubusercontent.com/opencv/opencv/3.4/samples/data/dnn/classification_classes_ILSVRC2012.txt",
            "modelUrl": "https://raw.githubusercontent.com/forresti/SqueezeNet/master/SqueezeNet_v1.0/squeezenet_v1.0.caffemodel",
            "configUrl": "https://raw.githubusercontent.com/forresti/SqueezeNet/master/SqueezeNet_v1.0/deploy.prototxt"
        },
@@ -46,7 +46,7 @@
            "std": "1",
            "swapRB": "false",
            "needSoftmax": "false",
-            "labelsUrl": "https://raw.githubusercontent.com/opencv/opencv/master/samples/data/dnn/classification_classes_ILSVRC2012.txt",
+            "labelsUrl": "https://raw.githubusercontent.com/opencv/opencv/3.4/samples/data/dnn/classification_classes_ILSVRC2012.txt",
            "modelUrl": "http://www.robots.ox.ac.uk/~vgg/software/very_deep/caffe/VGG_ILSVRC_19_layers.caffemodel",
            "configUrl": "https://gist.githubusercontent.com/ksimonyan/3785162f95cd2d5fee77/raw/f02f8769e64494bcd3d7e97d5d747ac275825721/VGG_ILSVRC_19_layers_deploy.prototxt"
        }
@@ -116,7 +116,7 @@ swapRB = false;
 needSoftmax = false;

 // url for label file, can from local or Internet
-labelsUrl = "https://raw.githubusercontent.com/opencv/opencv/master/samples/data/dnn/classification_classes_ILSVRC2012.txt";
+labelsUrl = "https://raw.githubusercontent.com/opencv/opencv/3.4/samples/data/dnn/classification_classes_ILSVRC2012.txt";
 </script>

 <script id="codeSnippet1" type="text/code-snippet">
@@ -77,7 +77,7 @@ cv.normalize(roiHist, roiHist, 0, 255, cv.NORM_MINMAX);
 // delete useless mats.
 roi.delete(); hsvRoi.delete(); mask.delete(); low.delete(); high.delete(); hsvRoiVec.delete();

-// Setup the termination criteria, either 10 iteration or move by atleast 1 pt
+// Setup the termination criteria, either 10 iteration or move by at least 1 pt
 let termCrit = new cv.TermCriteria(cv.TERM_CRITERIA_EPS | cv.TERM_CRITERIA_COUNT, 10, 1);

 let hsv = new cv.Mat(video.height, video.width, cv.CV_8UC3);
@@ -94,7 +94,7 @@ nmsThreshold = 0.4;
 outType = "SSD";

 // url for label file, can from local or Internet
-labelsUrl = "https://raw.githubusercontent.com/opencv/opencv/master/samples/data/dnn/object_detection_classes_pascal_voc.txt";
+labelsUrl = "https://raw.githubusercontent.com/opencv/opencv/3.4/samples/data/dnn/object_detection_classes_pascal_voc.txt";
 </script>

 <script id="codeSnippet1" type="text/code-snippet">
@@ -7,7 +7,7 @@
            "std": "0.007843",
            "swapRB": "false",
            "outType": "SSD",
-            "labelsUrl": "https://raw.githubusercontent.com/opencv/opencv/master/samples/data/dnn/object_detection_classes_pascal_voc.txt",
+            "labelsUrl": "https://raw.githubusercontent.com/opencv/opencv/3.4/samples/data/dnn/object_detection_classes_pascal_voc.txt",
            "modelUrl": "https://raw.githubusercontent.com/chuanqi305/MobileNet-SSD/master/mobilenet_iter_73000.caffemodel",
            "configUrl": "https://raw.githubusercontent.com/chuanqi305/MobileNet-SSD/master/deploy.prototxt"
        },
@@ -18,7 +18,7 @@
            "std": "1",
            "swapRB": "false",
            "outType": "SSD",
-            "labelsUrl": "https://raw.githubusercontent.com/opencv/opencv/master/samples/data/dnn/object_detection_classes_pascal_voc.txt",
+            "labelsUrl": "https://raw.githubusercontent.com/opencv/opencv/3.4/samples/data/dnn/object_detection_classes_pascal_voc.txt",
            "modelUrl": "https://drive.google.com/uc?id=0BzKzrI_SkD1_WVVTSmQxU0dVRzA&export=download",
            "configUrl": "https://drive.google.com/uc?id=0BzKzrI_SkD1_WVVTSmQxU0dVRzA&export=download"
        }
@@ -31,7 +31,7 @@
            "std": "0.00392",
            "swapRB": "false",
            "outType": "YOLO",
-            "labelsUrl": "https://raw.githubusercontent.com/opencv/opencv/master/samples/data/dnn/object_detection_classes_yolov3.txt",
+            "labelsUrl": "https://raw.githubusercontent.com/opencv/opencv/3.4/samples/data/dnn/object_detection_classes_yolov3.txt",
            "modelUrl": "https://pjreddie.com/media/files/yolov2-tiny.weights",
            "configUrl": "https://raw.githubusercontent.com/pjreddie/darknet/master/cfg/yolov2-tiny.cfg"
        }
@@ -94,7 +94,7 @@ nmsThreshold = 0.4;
 outType = "SSD";

 // url for label file, can from local or Internet
-labelsUrl = "https://raw.githubusercontent.com/opencv/opencv/master/samples/data/dnn/object_detection_classes_pascal_voc.txt";
+labelsUrl = "https://raw.githubusercontent.com/opencv/opencv/3.4/samples/data/dnn/object_detection_classes_pascal_voc.txt";
 </script>

 <script id="codeSnippet1" type="text/code-snippet">
@@ -1,6 +1,9 @@
 Contour Features {#tutorial_js_contour_features}
 ================

+@prev_tutorial{tutorial_js_contours_begin}
+@next_tutorial{tutorial_js_contour_properties}
+
 Goal
 ----

@@ -1,6 +1,9 @@
 Contour Properties {#tutorial_js_contour_properties}
 ==================

+@prev_tutorial{tutorial_js_contour_features}
+@next_tutorial{tutorial_js_contours_more_functions}
+
 Goal
 ----

@@ -1,6 +1,8 @@
 Contours : Getting Started {#tutorial_js_contours_begin}
 ==========================

+@next_tutorial{tutorial_js_contour_features}
+
 Goal
 ----

@@ -1,6 +1,8 @@
 Contours Hierarchy {#tutorial_js_contours_hierarchy}
 ==================

+@prev_tutorial{tutorial_js_contours_more_functions}
+
 Goal
 ----

@@ -1,6 +1,9 @@
 Contours : More Functions {#tutorial_js_contours_more_functions}
 =========================

+@prev_tutorial{tutorial_js_contour_properties}
+@next_tutorial{tutorial_js_contours_hierarchy}
+
 Goal
 ----

@@ -333,7 +333,7 @@ function installDOM(){
 ### Execute it ###

 -   Save the file as `exampleNodeCanvasData.js`.
-   Make sure the files `aarcascade_frontalface_default.xml` and `haarcascade_eye.xml` are present in project's directory. They can be obtained from [OpenCV sources](https://github.com/opencv/opencv/tree/master/data/haarcascades).
+-   Make sure the files `aarcascade_frontalface_default.xml` and `haarcascade_eye.xml` are present in project's directory. They can be obtained from [OpenCV sources](https://github.com/opencv/opencv/tree/3.4/data/haarcascades).
 -   Make sure a sample image file `lena.jpg` exists in project's directory. It should display people's faces for this example to make sense. The following image is known to work:

 ![image](lena.jpg)
@@ -138,6 +138,7 @@ Building OpenCV.js from Source
    python ./platforms/js/build_js.py build_js --cmake_option="-DOPENCV_EXTRA_MODULES_PATH=opencv_contrib/modules"
    @endcode

+
 Running OpenCV.js Tests
 ---------------------------------------

@@ -303,6 +304,12 @@ The example uses latest version of emscripten. If the build fails you should try
 docker run --rm -v $(pwd):/src -u $(id -u):$(id -g) emscripten/emsdk:2.0.10 emcmake python3 ./platforms/js/build_js.py build_js
@endcode

+In Windows use the following PowerShell command:
+
+@code{.bash}
+docker run --rm --workdir /src -v "$(get-location):/src" "emscripten/emsdk:2.0.10" emcmake python3 ./platforms/js/build_js.py build_js
+@endcode
+
 ### Building the documentation with Docker

 To build the documentation `doxygen` needs to be installed. Create a file named `Dockerfile` with the following content:
@@ -4,7 +4,9 @@ Using OpenCV.js {#tutorial_js_usage}
 Steps
 -----

-In this tutorial, you will learn how to include and start to use `opencv.js` inside a web page. You can get a copy of `opencv.js` from `opencv-{VERSION_NUMBER}-docs.zip` in each [release](https://github.com/opencv/opencv/releases), or simply download the prebuilt script from the online documentations at "https://docs.opencv.org/{VERSION_NUMBER}/opencv.js" (For example, [https://docs.opencv.org/3.4.0/opencv.js](https://docs.opencv.org/3.4.0/opencv.js). Use `master` if you want the latest build). You can also build your own copy by following the tutorial on Build Opencv.js.
+In this tutorial, you will learn how to include and start to use `opencv.js` inside a web page.
+You can get a copy of `opencv.js` from `opencv-{VERSION_NUMBER}-docs.zip` in each [release](https://github.com/opencv/opencv/releases), or simply download the prebuilt script from the online documentations at "https://docs.opencv.org/{VERSION_NUMBER}/opencv.js" (For example, [https://docs.opencv.org/3.4.0/opencv.js](https://docs.opencv.org/3.4.0/opencv.js). Use `3.4` if you want the latest build).
+You can also build your own copy by following the tutorial @ref tutorial_js_setup.

 ### Create a web page

@@ -133,9 +133,9 @@ Dense Optical Flow in OpenCV.js

 Lucas-Kanade method computes optical flow for a sparse feature set (in our example, corners detected
 using Shi-Tomasi algorithm). OpenCV.js provides another algorithm to find the dense optical flow. It
-computes the optical flow for all the points in the frame. It is based on Gunner Farneback's
+computes the optical flow for all the points in the frame. It is based on Gunnar Farneback's
 algorithm which is explained in "Two-Frame Motion Estimation Based on Polynomial Expansion" by
-Gunner Farneback in 2003.
+Gunnar Farneback in 2003.

 We use the function: **cv.calcOpticalFlowFarneback (prev, next, flow, pyrScale, levels, winsize,
 iterations, polyN, polySigma, flags)**
@@ -240,7 +240,7 @@
  hal_id = {inria-00350283},
  hal_version = {v1},
 }
-@article{Collins14
+@article{Collins14,
  year = {2014},
  issn = {0920-5691},
  journal = {International Journal of Computer Vision},
@@ -835,12 +835,12 @@
  journal = {IEEE Transactions on Robotics and Automation},
  title = {Robot sensor calibration: solving AX=XB on the Euclidean group},
  year = {1994},
+  month = oct,
  volume = {10},
  number = {5},
  pages = {717-721},
  doi = {10.1109/70.326576},
-  ISSN = {1042-296X},
-  month = {Oct}
+  issn = {1042-296X}
 }
@inproceedings{PM03,
  author = {P{\'e}rez, Patrick and Gangnet, Michel and Blake, Andrew},
@@ -1028,12 +1028,12 @@
  journal = {IEEE Transactions on Robotics and Automation},
  title = {A new technique for fully autonomous and efficient 3D robotics hand/eye calibration},
  year = {1989},
+  month = jun,
  volume = {5},
  number = {3},
  pages = {345-358},
  doi = {10.1109/70.34770},
-  ISSN = {1042-296X},
-  month = {June}
+  issn = {1042-296X}
 }
@inproceedings{UES01,
  author = {Uyttendaele, Matthew and Eden, Ashley and Skeliski, R},
@@ -1271,6 +1271,12 @@
  number={2},
  pages={117-135},
 }
+@inproceedings{Zuliani2014RANSACFD,
+  title={RANSAC for Dummies With examples using the RANSAC toolbox for Matlab \& Octave and more...},
+  author={Marco Zuliani},
+  year={2014},
+  url = {https://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.475.1243&rep=rep1&type=pdf}
+}
@inproceedings{forstner1987fast,
  title={A fast operator for detection and precise location of distincs points, corners and center of circular features},
  author={FORSTNER, W},
@@ -1278,3 +1284,11 @@
  pages={281--305},
  year={1987}
 }
+@article{Bolelli2021,
+  title={One DAG to Rule Them All},
+  author={Bolelli, Federico and Allegretti, Stefano and Grana, Costantino},
+  journal={IEEE Transactions on Pattern Analysis and Machine Intelligence},
+  year={2021},
+  publisher={IEEE},
+  doi = {10.1109/TPAMI.2021.3055337}
+}
@@ -98,7 +98,7 @@ import numpy as np
 import cv2 as cv
 from matplotlib import pyplot as plt

-img = cv.imread('simple.jpg',0)
+img = cv.imread('blox.jpg',0) # `<opencv_root>/samples/data/blox.jpg`

 # Initiate FAST object with default values
 fast = cv.FastFeatureDetector_create()
@@ -113,17 +113,17 @@ print( "nonmaxSuppression:{}".format(fast.getNonmaxSuppression()) )
 print( "neighborhood: {}".format(fast.getType()) )
 print( "Total Keypoints with nonmaxSuppression: {}".format(len(kp)) )

-cv.imwrite('fast_true.png',img2)
+cv.imwrite('fast_true.png', img2)

 # Disable nonmaxSuppression
 fast.setNonmaxSuppression(0)
-kp = fast.detect(img,None)
+kp = fast.detect(img, None)

 print( "Total Keypoints without nonmaxSuppression: {}".format(len(kp)) )

 img3 = cv.drawKeypoints(img, kp, None, color=(255,0,0))

-cv.imwrite('fast_false.png',img3)
+cv.imwrite('fast_false.png', img3)
@endcode
 See the results. First image shows FAST with nonmaxSuppression and second one without
 nonmaxSuppression:
@@ -78,7 +78,7 @@ if len(good)>MIN_MATCH_COUNT:
    M, mask = cv.findHomography(src_pts, dst_pts, cv.RANSAC,5.0)
    matchesMask = mask.ravel().tolist()

-    h,w,d = img1.shape
+    h,w = img1.shape
    pts = np.float32([ [0,0],[0,h-1],[w-1,h-1],[w-1,0] ]).reshape(-1,1,2)
    dst = cv.perspectiveTransform(pts,M)

@@ -40,12 +40,12 @@ using **cv.Sobel()**).
 Then comes the main part. After this, they created a score, basically an equation, which
 determines if a window can contain a corner or not.

-\f[R = det(M) - k(trace(M))^2\f]
+\f[R = \det(M) - k(\operatorname{trace}(M))^2\f]

 where
-    -   \f$det(M) = \lambda_1 \lambda_2\f$
-    -   \f$trace(M) = \lambda_1 + \lambda_2\f$
-    -   \f$\lambda_1\f$ and \f$\lambda_2\f$ are the eigenvalues of M
+    -   \f$\det(M) = \lambda_1 \lambda_2\f$
+    -   \f$\operatorname{trace}(M) = \lambda_1 + \lambda_2\f$
+    -   \f$\lambda_1\f$ and \f$\lambda_2\f$ are the eigenvalues of \f$M\f$

 So the magnitudes of these eigenvalues decide whether a region is a corner, an edge, or flat.

@@ -20,7 +20,7 @@ Harris Corner Detector. The scoring function in Harris Corner Detector was given

 Instead of this, Shi-Tomasi proposed:

-\f[R = min(\lambda_1, \lambda_2)\f]
+\f[R = \min(\lambda_1, \lambda_2)\f]

 If it is a greater than a threshold value, it is considered as a corner. If we plot it in
 \f$\lambda_1 - \lambda_2\f$ space as we did in Harris Corner Detector, we get an image as below:
@@ -28,7 +28,7 @@ If it is a greater than a threshold value, it is considered as a corner. If we p
 ![image](images/shitomasi_space.png)

 From the figure, you can see that only when \f$\lambda_1\f$ and \f$\lambda_2\f$ are above a minimum value,
-\f$\lambda_{min}\f$, it is considered as a corner(green region).
+\f$\lambda_{\min}\f$, it is considered as a corner(green region).

 Code
 ----
@@ -156,7 +156,7 @@ sift = cv.SIFT_create()
 kp, des = sift.detectAndCompute(gray,None)
@endcode
 Here kp will be a list of keypoints and des is a numpy array of shape
-\f$Number\_of\_Keypoints \times 128\f$.
+\f$\text{(Number of Keypoints)} \times 128\f$.

 So we got keypoints, descriptors etc. Now we want to see how to match keypoints in different images.
 That we will learn in coming chapters.
@@ -74,7 +74,7 @@ Canny Edge Detection in OpenCV

 OpenCV puts all the above in single function, **cv.Canny()**. We will see how to use it. First
 argument is our input image. Second and third arguments are our minVal and maxVal respectively.
-Third argument is aperture_size. It is the size of Sobel kernel used for find image gradients. By
+Fourth argument is aperture_size. It is the size of Sobel kernel used for find image gradients. By
 default it is 3. Last argument is L2gradient which specifies the equation for finding gradient
 magnitude. If it is True, it uses the equation mentioned above which is more accurate, otherwise it
 uses this function: \f$Edge\_Gradient \; (G) = |G_x| + |G_y|\f$. By default, it is False.
@@ -1,6 +1,9 @@
 Contour Features {#tutorial_py_contour_features}
 ================

+@prev_tutorial{tutorial_py_contours_begin}
+@next_tutorial{tutorial_py_contour_properties}
+
 Goal
 ----

@@ -91,7 +94,7 @@ convexity defects, which are the local maximum deviations of hull from contours.

 There is a little bit things to discuss about it its syntax:
@code{.py}
-hull = cv.convexHull(points[, hull[, clockwise[, returnPoints]]
+hull = cv.convexHull(points[, hull[, clockwise[, returnPoints]]])
@endcode
 Arguments details:

@@ -1,6 +1,9 @@
 Contour Properties {#tutorial_py_contour_properties}
 ==================

+@prev_tutorial{tutorial_py_contour_features}
+@next_tutorial{tutorial_py_contours_more_functions}
+
 Here we will learn to extract some frequently used properties of objects like Solidity, Equivalent
 Diameter, Mask image, Mean Intensity etc. More features can be found at [Matlab regionprops
 documentation](http://www.mathworks.in/help/images/ref/regionprops.html).
@@ -1,6 +1,8 @@
 Contours : Getting Started {#tutorial_py_contours_begin}
 ==========================

+@next_tutorial{tutorial_py_contour_features}
+
 Goal
 ----

@@ -1,6 +1,8 @@
 Contours Hierarchy {#tutorial_py_contours_hierarchy}
 ==================

+@prev_tutorial{tutorial_py_contours_more_functions}
+
 Goal
 ----

@@ -1,6 +1,10 @@
 Contours : More Functions {#tutorial_py_contours_more_functions}
 =========================

+@prev_tutorial{tutorial_py_contour_properties}
+@next_tutorial{tutorial_py_contours_hierarchy}
+
+
 Goal
 ----

@@ -117,7 +117,7 @@ for i in range(5,0,-1):
 LS = []
 for la,lb in zip(lpA,lpB):
    rows,cols,dpt = la.shape
-    ls = np.hstack((la[:,0:cols/2], lb[:,cols/2:]))
+    ls = np.hstack((la[:,0:cols//2], lb[:,cols//2:]))
    LS.append(ls)

 # now reconstruct
@@ -127,7 +127,7 @@ for i in range(1,6):
    ls_ = cv.add(ls_, LS[i])

 # image with direct connecting each half
-real = np.hstack((A[:,:cols/2],B[:,cols/2:]))
+real = np.hstack((A[:,:cols//2],B[:,cols//2:]))

 cv.imwrite('Pyramid_blending2.jpg',ls_)
 cv.imwrite('Direct_blending.jpg',real)
--- a/Show More
+++ b/Show More