[M114-LTS] Fix OOB write in BuildHuffmanTable.

M114 merge issues:
  dec/vp8l_dec.c:
    - Conflicting checks before ReadHuffmanCodeLengths() return statement;
    In 114, an assignment follows the check instead of a return.
    - ReadHuffmanCodes(): Conflict after the changed huffman_tables check,
    there's an assignment in 114 instead of the setter call.

First, BuildHuffmanTable is called to check if the data is valid.
If it is and the table is not big enough, more memory is allocated.

This will make sure that valid (but unoptimized because of unbalanced
codes) streams are still decodable.

Bug: chromium:1479274
Change-Id: I31c36dbf3aa78d35ecf38706b50464fd3d375741
(cherry picked from commit 902bc91903)

.gitignore

@@ -3,7 +3,9 @@
 *.pc
 .DS_Store
 .deps
+.idea
 .libs
+.vscode
 /aclocal.m4
 /ar-lib
 /autom4te.cache

.mailmap

@@ -8,8 +8,11 @@ Vikas Arora <vikasa@google.com>
 <vikasa@google.com> <vikaas.arora@gmail.com>
 <slobodan.prijic@imgtec.com> <Slobodan.Prijic@imgtec.com>
 <vrabaud@google.com> <vincent.rabaud@gmail.com>
+Vincent Rabaud <vrabaud@google.com>
 Tamar Levy <tamar.levy@intel.com>
 <qrczak@google.com> <qrczak>
 Hui Su <huisu@google.com>
 James Zern <jzern@google.com>
 Roberto Alanis <alanisbaez@google.com>
+Brian Ledger <brianpl@google.com>
+Maryla Ustarroz-Calonge <maryla@google.com>

AUTHORS

@@ -1,18 +1,23 @@
 Contributors:
 - Aidan O'Loan (aidanol at gmail dot com)
 - Alan Browning (browning at google dot com)
+- Alexandru Ardelean (ardeleanalex at gmail dot com)
+- Brian Ledger (brianpl at google dot com)
 - Charles Munger (clm at google dot com)
 - Cheng Yi (cyi at google dot com)
 - Christian Duvivier (cduvivier at google dot com)
 - Christopher Degawa (ccom at randomderp dot com)
 - Clement Courbet (courbet at google dot com)
 - Djordje Pesut (djordje dot pesut at imgtec dot com)
+- Frank Barchard (fbarchard at google dot com)
 - Hui Su (huisu at google dot com)
+- H. Vetinari (h dot vetinari at gmx dot com)
 - Ilya Kurdyukov (jpegqs at gmail dot com)
 - Ingvar Stepanyan (rreverser at google dot com)
 - James Zern (jzern at google dot com)
 - Jan Engelhardt (jengelh at medozas dot de)
 - Jehan (jehan at girinstud dot io)
+- Jeremy Maitin-Shepard (jbms at google dot com)
 - Johann Koenig (johann dot koenig at duck dot com)
 - Jovan Zelincevic (jovan dot zelincevic at imgtec dot com)
 - Jyrki Alakuijala (jyrki at google dot com)
@@ -22,6 +27,7 @@ Contributors:
 - Mans Rullgard (mans at mansr dot com)
 - Marcin Kowalczyk (qrczak at google dot com)
 - Martin Olsson (mnemo at minimum dot se)
+- Maryla Ustarroz-Calonge (maryla at google dot com)
 - Mikołaj Zalewski (mikolajz at google dot com)
 - Mislav Bradac (mislavm at google dot com)
 - Nico Weber (thakis at chromium dot org)

Android.mk

@@ -1,3 +1,5 @@
+# Ignore this file during non-NDK builds.
+ifdef NDK_ROOT
 LOCAL_PATH := $(call my-dir)
 
 WEBP_CFLAGS := -Wall -DANDROID -DHAVE_MALLOC_H -DHAVE_PTHREAD -DWEBP_USE_THREAD
@@ -33,6 +35,15 @@ else
   NEON := c
 endif
 
+sharpyuv_srcs := \
+    sharpyuv/sharpyuv.c \
+    sharpyuv/sharpyuv_cpu.c \
+    sharpyuv/sharpyuv_csp.c \
+    sharpyuv/sharpyuv_dsp.c \
+    sharpyuv/sharpyuv_gamma.c \
+    sharpyuv/sharpyuv_neon.$(NEON) \
+    sharpyuv/sharpyuv_sse2.c \
+
 dec_srcs := \
     src/dec/alpha_dec.c \
     src/dec/buffer_dec.c \
@@ -204,12 +215,13 @@ endif  # ENABLE_SHARED=1
 include $(CLEAR_VARS)
 
 LOCAL_SRC_FILES := \
+    $(sharpyuv_srcs) \
     $(dsp_enc_srcs) \
     $(enc_srcs) \
     $(utils_enc_srcs) \
 
 LOCAL_CFLAGS := $(WEBP_CFLAGS)
-LOCAL_EXPORT_C_INCLUDES += $(LOCAL_PATH)/src
+LOCAL_EXPORT_C_INCLUDES += $(LOCAL_PATH)/src $(LOCAL_PATH)
 
 # prefer arm over thumb mode for performance gains
 LOCAL_ARM_MODE := arm
@@ -279,3 +291,4 @@ include $(WEBP_SRC_PATH)/examples/Android.mk
 ifeq ($(USE_CPUFEATURES),yes)
   $(call import-module,android/cpufeatures)
 endif
+endif  # NDK_ROOT

CMakeLists.txt

@@ -6,7 +6,11 @@
 #  in the file PATENTS.  All contributing project authors may
 #  be found in the AUTHORS file in the root of the source tree.
 
-cmake_minimum_required(VERSION 3.7)
+if(APPLE)
+  cmake_minimum_required(VERSION 3.17)
+else()
+  cmake_minimum_required(VERSION 3.7)
+endif()
 
 if(POLICY CMP0072)
   cmake_policy(SET CMP0072 NEW)
@@ -15,6 +19,14 @@ endif()
 project(WebP C)
 
 # Options for coder / decoder executables.
+if(BUILD_SHARED_LIBS)
+  set(WEBP_LINK_STATIC_DEFAULT OFF)
+else()
+  set(WEBP_LINK_STATIC_DEFAULT ON)
+endif()
+option(WEBP_LINK_STATIC
+       "Link using static libraries. If OFF, use dynamic libraries."
+       ${WEBP_LINK_STATIC_DEFAULT})
 if(NOT EMSCRIPTEN)
   # Disable SIMD on Emscripten by default, as it's a new unstable Wasm feature.
   # Users can still explicitly opt-in to make a SIMD-enabled build.
@@ -40,6 +52,18 @@ option(WEBP_ENABLE_SWAP_16BIT_CSP "Enable byte swap for 16 bit colorspaces."
 set(WEBP_BITTRACE "0" CACHE STRING "Bit trace mode (0=none, 1=bit, 2=bytes)")
 set_property(CACHE WEBP_BITTRACE PROPERTY STRINGS 0 1 2)
 
+if(WEBP_LINK_STATIC)
+  if(WIN32)
+    set(CMAKE_FIND_LIBRARY_SUFFIXES .lib .a ${CMAKE_FIND_LIBRARY_SUFFIXES})
+  else()
+    set(CMAKE_FIND_LIBRARY_SUFFIXES .a ${CMAKE_FIND_LIBRARY_SUFFIXES})
+  endif()
+  set(CMAKE_POSITION_INDEPENDENT_CODE ON)
+  # vwebp does not compile on Ubuntu with static libraries so disabling it for
+  # now.
+  set(WEBP_BUILD_VWEBP OFF)
+endif()
+
 # Option needed for handling Unicode file names on Windows.
 if(WIN32)
   option(WEBP_UNICODE "Build Unicode executables." ON)
@@ -62,19 +86,27 @@ if(WEBP_BUILD_WEBP_JS)
   endif()
 endif()
 
+set(SHARPYUV_DEP_LIBRARIES)
+set(SHARPYUV_DEP_INCLUDE_DIRS)
 set(WEBP_DEP_LIBRARIES)
 set(WEBP_DEP_INCLUDE_DIRS)
 
 if(NOT CMAKE_BUILD_TYPE)
   set(CMAKE_BUILD_TYPE "Release"
       CACHE STRING "Build type: Release, Debug, MinSizeRel or RelWithDebInfo"
-      FORCE)
+            FORCE)
 endif()
 
 # Include dependencies.
 include(cmake/deps.cmake)
 include(GNUInstallDirs)
 
+if(BUILD_SHARED_LIBS AND NOT CMAKE_INSTALL_RPATH)
+  # Set the rpath to match autoconf/libtool behavior. Note this must be set
+  # before target creation.
+  set(CMAKE_INSTALL_RPATH "${CMAKE_INSTALL_PREFIX}/${CMAKE_INSTALL_LIBDIR}")
+endif()
+
 # ##############################################################################
 # Options.
 if(WEBP_ENABLE_SWAP_16BIT_CSP)
@@ -90,24 +122,50 @@ if(WEBP_UNICODE)
   add_definitions(-DUNICODE -D_UNICODE)
 endif()
 
+if(MSVC AND BUILD_SHARED_LIBS)
+  add_definitions(-DWEBP_DLL)
+endif()
+
+# pkg-config variables used by *.pc.in.
 set(prefix ${CMAKE_INSTALL_PREFIX})
-set(exec_prefix "\$\{prefix\}")
+set(exec_prefix "\${prefix}")
-set(libdir "\$\{prefix\}/lib")
+if(IS_ABSOLUTE "${CMAKE_INSTALL_LIBDIR}")
-set(includedir "\$\{prefix\}/include")
+  set(libdir "${CMAKE_INSTALL_LIBDIR}")
+else()
+  set(libdir "\${exec_prefix}/${CMAKE_INSTALL_LIBDIR}")
+endif()
+if(IS_ABSOLUTE "${CMAKE_INSTALL_INCLUDEDIR}")
+  set(includedir "${CMAKE_INSTALL_INCLUDEDIR}")
+else()
+  set(includedir "\${prefix}/${CMAKE_INSTALL_INCLUDEDIR}")
+endif()
 set(PTHREAD_LIBS ${CMAKE_THREAD_LIBS_INIT})
 set(INSTALLED_LIBRARIES)
 
+if(MSVC)
+  # match the naming convention used by nmake
+  set(webp_libname_prefix "lib")
+  set(CMAKE_SHARED_LIBRARY_PREFIX "${webp_libname_prefix}")
+  set(CMAKE_IMPORT_LIBRARY_PREFIX "${webp_libname_prefix}")
+  set(CMAKE_STATIC_LIBRARY_PREFIX "${webp_libname_prefix}")
+endif()
+
+set(CMAKE_C_VISIBILITY_PRESET hidden)
+
 # ##############################################################################
 # Android only.
 if(ANDROID)
   include_directories(${ANDROID_NDK}/sources/android/cpufeatures)
-  add_library(cpufeatures STATIC
+  add_library(cpufeatures-webp STATIC
               ${ANDROID_NDK}/sources/android/cpufeatures/cpu-features.c)
-  list(APPEND INSTALLED_LIBRARIES cpufeatures)
+  list(APPEND INSTALLED_LIBRARIES cpufeatures-webp)
-  target_link_libraries(cpufeatures dl)
+  target_link_libraries(cpufeatures-webp dl)
-  set(WEBP_DEP_LIBRARIES ${WEBP_DEP_LIBRARIES} cpufeatures)
+  set(SHARPYUV_DEP_LIBRARIES ${SHARPYUV_DEP_LIBRARIES} cpufeatures-webp)
-  set(WEBP_DEP_INCLUDE_DIRS ${WEBP_DEP_INCLUDE_DIRS}
+  set(WEBP_DEP_LIBRARIES ${WEBP_DEP_LIBRARIES} cpufeatures-webp)
-      ${ANDROID_NDK}/sources/android/cpufeatures)
+  set(cpufeatures_include_dir ${ANDROID_NDK}/sources/android/cpufeatures)
+  set(SHARPYUV_DEP_INCLUDE_DIRS ${SHARPYUV_DEP_INCLUDE_DIRS}
+                                ${cpufeatures_include_dir})
+  set(WEBP_DEP_INCLUDE_DIRS ${WEBP_DEP_INCLUDE_DIRS} ${cpufeatures_include_dir})
   add_definitions(-DHAVE_CPU_FEATURES_H=1)
   set(HAVE_CPU_FEATURES_H 1)
 else()
@@ -116,7 +174,7 @@ endif()
 
 function(configure_pkg_config FILE)
   configure_file("${CMAKE_CURRENT_SOURCE_DIR}/${FILE}.in"
-                 "${CMAKE_CURRENT_BINARY_DIR}/${FILE}")
+                 "${CMAKE_CURRENT_BINARY_DIR}/${FILE}" @ONLY)
 
   if(HAVE_MATH_LIBRARY)
     # MSVC doesn't have libm
@@ -125,10 +183,8 @@ function(configure_pkg_config FILE)
     file(WRITE ${CMAKE_CURRENT_BINARY_DIR}/${FILE} ${data})
   endif()
 
-  install(
+  install(FILES "${CMAKE_CURRENT_BINARY_DIR}/${FILE}"
-    FILES "${CMAKE_CURRENT_BINARY_DIR}/${FILE}"
+          DESTINATION ${CMAKE_INSTALL_LIBDIR}/pkgconfig)
-    DESTINATION ${CMAKE_INSTALL_LIBDIR}/pkgconfig
-  )
 endfunction()
 
 # ##############################################################################
@@ -138,23 +194,15 @@ endfunction()
 # E.g.: libimagedec_la_SOURCES  = image_dec.c image_dec.h
 function(parse_Makefile_am FOLDER VAR SRC_REGEX)
   file(READ ${FOLDER}/Makefile.am MAKEFILE_AM)
-  string(REGEX MATCHALL
+  string(REGEX MATCHALL "${SRC_REGEX}_SOURCES[ ]*\\+?=[ ]+[0-9a-z\\._ ]*"
-               "${SRC_REGEX}_SOURCES[ ]*\\+?=[ ]+[0-9a-z\\._ ]*"
+               FILES_PER_LINE ${MAKEFILE_AM})
-               FILES_PER_LINE
-               ${MAKEFILE_AM})
   set(SRCS ${${VAR}})
   foreach(FILES ${FILES_PER_LINE})
     string(FIND ${FILES} "=" OFFSET)
     math(EXPR OFFSET "${OFFSET} + 2")
-    string(SUBSTRING ${FILES}
+    string(SUBSTRING ${FILES} ${OFFSET} -1 FILES)
-                     ${OFFSET}
-                     -1
-                     FILES)
     if(FILES)
-      string(REGEX MATCHALL
+      string(REGEX MATCHALL "[0-9a-z\\._]+" FILES ${FILES})
-                   "[0-9a-z\\._]+"
-                   FILES
-                   ${FILES})
       foreach(FILE ${FILES})
         list(APPEND SRCS ${FOLDER}/${FILE})
       endforeach()
@@ -183,161 +231,17 @@ endforeach()
 
 # Generate the config.h file.
 configure_file(${CMAKE_CURRENT_LIST_DIR}/cmake/config.h.in
-               ${CMAKE_CURRENT_BINARY_DIR}/src/webp/config.h)
+               ${CMAKE_CURRENT_BINARY_DIR}/src/webp/config.h @ONLY)
 add_definitions(-DHAVE_CONFIG_H)
 
-# ##############################################################################
-# Build the webpdecoder library.
-
-# Creates a source file with an unused stub function in $CMAKE_BINARY_DIR and
-# adds it to the specified target. Currently used only with Xcode.
-#
-# See also:
-# https://cmake.org/cmake/help/v3.18/command/add_library.html#object-libraries
-# "Some native build systems (such as Xcode) may not like targets that have
-# only object files, so consider adding at least one real source file to any
-# target that references $<TARGET_OBJECTS:objlib>."
-function(libwebp_add_stub_file TARGET)
-  set(stub_source_dir "${CMAKE_BINARY_DIR}")
-  set(stub_source_file
-      "${stub_source_dir}/libwebp_${TARGET}_stub.c")
-  set(stub_source_code
-      "// Generated file. DO NOT EDIT!\n"
-      "// C source file created for target ${TARGET}.\n"
-      "void libwebp_${TARGET}_stub_function(void)\;\n"
-      "void libwebp_${TARGET}_stub_function(void) {}\n")
-  file(WRITE "${stub_source_file}" ${stub_source_code})
-
-  target_sources(${TARGET} PRIVATE ${stub_source_file})
-endfunction()
-
-if(MSVC)
-  # avoid security warnings for e.g., fopen() used in the examples.
-  add_definitions(-D_CRT_SECURE_NO_WARNINGS)
-else()
-  add_definitions(-Wall)
-endif()
-include_directories(${WEBP_DEP_INCLUDE_DIRS})
-add_library(webpdecode OBJECT ${WEBP_DEC_SRCS})
-target_include_directories(webpdecode
-                           PRIVATE ${CMAKE_CURRENT_BINARY_DIR}
-                                   ${CMAKE_CURRENT_SOURCE_DIR})
-add_library(webpdspdecode OBJECT ${WEBP_DSP_COMMON_SRCS} ${WEBP_DSP_DEC_SRCS})
-target_include_directories(webpdspdecode
-                           PRIVATE ${CMAKE_CURRENT_BINARY_DIR}
-                                   ${CMAKE_CURRENT_SOURCE_DIR})
-add_library(webputilsdecode
-            OBJECT
-            ${WEBP_UTILS_COMMON_SRCS}
-            ${WEBP_UTILS_DEC_SRCS})
-target_include_directories(webputilsdecode
-                           PRIVATE ${CMAKE_CURRENT_BINARY_DIR}
-                                   ${CMAKE_CURRENT_SOURCE_DIR})
-add_library(webpdecoder
-            $<TARGET_OBJECTS:webpdecode>
-            $<TARGET_OBJECTS:webpdspdecode>
-            $<TARGET_OBJECTS:webputilsdecode>)
-if(XCODE)
-  libwebp_add_stub_file(webpdecoder)
-endif()
-target_link_libraries(webpdecoder ${WEBP_DEP_LIBRARIES})
-target_include_directories(
-  webpdecoder
-  PRIVATE ${CMAKE_CURRENT_BINARY_DIR} ${CMAKE_CURRENT_SOURCE_DIR}
-  INTERFACE $<BUILD_INTERFACE:${CMAKE_CURRENT_SOURCE_DIR}>
-            $<INSTALL_INTERFACE:${CMAKE_INSTALL_INCLUDEDIR}>)
-set_target_properties(
-  webpdecoder
-  PROPERTIES PUBLIC_HEADER "${CMAKE_CURRENT_SOURCE_DIR}/src/webp/decode.h;\
-${CMAKE_CURRENT_SOURCE_DIR}/src/webp/types.h")
-
-configure_pkg_config("src/libwebpdecoder.pc")
-
-# Build the webp library.
-add_library(webpencode OBJECT ${WEBP_ENC_SRCS})
-target_include_directories(webpencode
-                           PRIVATE ${CMAKE_CURRENT_BINARY_DIR}
-                                   ${CMAKE_CURRENT_SOURCE_DIR})
-add_library(webpdsp
-            OBJECT
-            ${WEBP_DSP_COMMON_SRCS}
-            ${WEBP_DSP_DEC_SRCS}
-            ${WEBP_DSP_ENC_SRCS})
-target_include_directories(webpdsp
-                           PRIVATE ${CMAKE_CURRENT_BINARY_DIR}
-                                   ${CMAKE_CURRENT_SOURCE_DIR})
-add_library(webputils
-            OBJECT
-            ${WEBP_UTILS_COMMON_SRCS}
-            ${WEBP_UTILS_DEC_SRCS}
-            ${WEBP_UTILS_ENC_SRCS})
-target_include_directories(webputils
-                           PRIVATE ${CMAKE_CURRENT_BINARY_DIR}
-                                   ${CMAKE_CURRENT_SOURCE_DIR})
-add_library(webp
-            $<TARGET_OBJECTS:webpdecode>
-            $<TARGET_OBJECTS:webpdsp>
-            $<TARGET_OBJECTS:webpencode>
-            $<TARGET_OBJECTS:webputils>)
-if(XCODE)
-  libwebp_add_stub_file(webp)
-endif()
-target_link_libraries(webp ${WEBP_DEP_LIBRARIES})
-target_include_directories(
-  webp
-  PRIVATE ${CMAKE_CURRENT_SOURCE_DIR}
-          ${CMAKE_CURRENT_BINARY_DIR}
-  PUBLIC $<BUILD_INTERFACE:${CMAKE_CURRENT_SOURCE_DIR}/src>
-         $<INSTALL_INTERFACE:include>)
-set_target_properties(
-  webp
-  PROPERTIES PUBLIC_HEADER "${CMAKE_CURRENT_SOURCE_DIR}/src/webp/decode.h;\
-${CMAKE_CURRENT_SOURCE_DIR}/src/webp/encode.h;\
-${CMAKE_CURRENT_SOURCE_DIR}/src/webp/types.h")
-
-# Make sure the OBJECT libraries are built with position independent code (it is
-# not ON by default).
-set_target_properties(webpdecode
-                      webpdspdecode
-                      webputilsdecode
-                      webpencode
-                      webpdsp
-                      webputils
-                      PROPERTIES POSITION_INDEPENDENT_CODE ON)
-configure_pkg_config("src/libwebp.pc")
-
-# Build the webp demux library.
-add_library(webpdemux ${WEBP_DEMUX_SRCS})
-target_link_libraries(webpdemux webp)
-target_include_directories(webpdemux
-                           PRIVATE ${CMAKE_CURRENT_SOURCE_DIR}
-                                   ${CMAKE_CURRENT_BINARY_DIR}
-                           PUBLIC $<INSTALL_INTERFACE:include>)
-set_target_properties(
-  webpdemux
-  PROPERTIES PUBLIC_HEADER "${CMAKE_CURRENT_SOURCE_DIR}/src/webp/decode.h;\
-${CMAKE_CURRENT_SOURCE_DIR}/src/webp/demux.h;\
-${CMAKE_CURRENT_SOURCE_DIR}/src/webp/mux_types.h;\
-${CMAKE_CURRENT_SOURCE_DIR}/src/webp/types.h")
-
-configure_pkg_config("src/demux/libwebpdemux.pc")
-
 # Set the version numbers.
 macro(set_version FILE TARGET_NAME NAME_IN_MAKEFILE)
-  file(READ ${CMAKE_CURRENT_SOURCE_DIR}/src/${FILE} SOURCE_FILE)
+  file(READ ${CMAKE_CURRENT_SOURCE_DIR}/${FILE} SOURCE_FILE)
   string(REGEX MATCH
-               "${NAME_IN_MAKEFILE}_la_LDFLAGS[^\n]* -version-info [0-9:]+"
+               "${NAME_IN_MAKEFILE}_la_LDFLAGS[^\n]* -version-info [0-9:]+" TMP
-               TMP
                ${SOURCE_FILE})
-  string(REGEX MATCH
+  string(REGEX MATCH "[0-9:]+" TMP ${TMP})
-               "[0-9:]+"
+  string(REGEX REPLACE ":" " " LT_VERSION ${TMP})
-               TMP
-               ${TMP})
-  string(REGEX
-         REPLACE ":"
-                 " "
-                 LT_VERSION
-                 ${TMP})
 
   # See the libtool docs for more information:
   # https://www.gnu.org/software/libtool/manual/libtool.html#Updating-version-info
@@ -356,23 +260,164 @@ macro(set_version FILE TARGET_NAME NAME_IN_MAKEFILE)
 
   set_target_properties(
     ${TARGET_NAME}
-    PROPERTIES VERSION
+    PROPERTIES VERSION ${LT_CURRENT_MINUS_AGE}.${LT_AGE}.${LT_REVISION}
-               ${LT_CURRENT_MINUS_AGE}.${LT_AGE}.${LT_REVISION}
+               SOVERSION ${LT_CURRENT_MINUS_AGE})
-               SOVERSION
+  if(APPLE)
-               ${LT_CURRENT_MINUS_AGE})
+    # For compatibility, set MACHO_COMPATIBILITY_VERSION and
+    # MACHO_CURRENT_VERSION to match libtool. These properties were introduced
+    # in 3.17:
+    # https://cmake.org/cmake/help/latest/prop_tgt/MACHO_COMPATIBILITY_VERSION.html
+    math(EXPR LIBWEBP_MACHO_COMPATIBILITY_VERSION "${LT_CURRENT} + 1")
+    set_target_properties(
+      ${TARGET_NAME}
+      PROPERTIES MACHO_COMPATIBILITY_VERSION
+                 ${LIBWEBP_MACHO_COMPATIBILITY_VERSION}
+                 MACHO_CURRENT_VERSION
+                 ${LIBWEBP_MACHO_COMPATIBILITY_VERSION}.${LT_REVISION})
+  endif()
 endmacro()
-set_version(Makefile.am webp webp)
+
-set_version(Makefile.am webpdecoder webpdecoder)
+# ##############################################################################
-set_version(demux/Makefile.am webpdemux webpdemux)
+# Build the webpdecoder library.
+
+# Creates a source file with an unused stub function in $CMAKE_BINARY_DIR and
+# adds it to the specified target. Currently used only with Xcode.
+#
+# See also:
+# https://cmake.org/cmake/help/v3.18/command/add_library.html#object-libraries
+# "Some native build systems (such as Xcode) may not like targets that have only
+# object files, so consider adding at least one real source file to any target
+# that references $<TARGET_OBJECTS:objlib>."
+function(libwebp_add_stub_file TARGET)
+  set(stub_source_dir "${CMAKE_BINARY_DIR}")
+  set(stub_source_file "${stub_source_dir}/libwebp_${TARGET}_stub.c")
+  set(stub_source_code
+      "// Generated file. DO NOT EDIT!\n"
+      "// C source file created for target ${TARGET}.\n"
+      "void libwebp_${TARGET}_stub_function(void)\;\n"
+      "void libwebp_${TARGET}_stub_function(void) {}\n")
+  file(WRITE "${stub_source_file}" ${stub_source_code})
+
+  target_sources(${TARGET} PRIVATE ${stub_source_file})
+endfunction()
+
+parse_makefile_am(${CMAKE_CURRENT_SOURCE_DIR}/sharpyuv "WEBP_SHARPYUV_SRCS" "")
+add_library(sharpyuv ${WEBP_SHARPYUV_SRCS})
+target_link_libraries(sharpyuv ${SHARPYUV_DEP_LIBRARIES})
+set_version(sharpyuv/Makefile.am sharpyuv sharpyuv)
+target_include_directories(
+  sharpyuv PRIVATE ${CMAKE_CURRENT_BINARY_DIR} ${CMAKE_CURRENT_SOURCE_DIR}
+                   ${CMAKE_CURRENT_SOURCE_DIR}/src)
+set_target_properties(
+  sharpyuv
+  PROPERTIES PUBLIC_HEADER "${CMAKE_CURRENT_SOURCE_DIR}/sharpyuv/sharpyuv.h;\
+${CMAKE_CURRENT_SOURCE_DIR}/sharpyuv/sharpyuv_csp.h")
+configure_pkg_config("sharpyuv/libsharpyuv.pc")
+install(
+  TARGETS sharpyuv
+  EXPORT ${PROJECT_NAME}Targets
+  PUBLIC_HEADER DESTINATION ${CMAKE_INSTALL_INCLUDEDIR}/webp/sharpyuv
+  INCLUDES
+  DESTINATION ${CMAKE_INSTALL_INCLUDEDIR}
+  ${CMAKE_INSTALL_INCLUDEDIR}/webp
+  ARCHIVE DESTINATION ${CMAKE_INSTALL_LIBDIR}
+  LIBRARY DESTINATION ${CMAKE_INSTALL_LIBDIR}
+  RUNTIME DESTINATION ${CMAKE_INSTALL_BINDIR})
+
+if(MSVC)
+  # avoid security warnings for e.g., fopen() used in the examples.
+  add_definitions(-D_CRT_SECURE_NO_WARNINGS)
+else()
+  add_compile_options(-Wall)
+endif()
+include_directories(${WEBP_DEP_INCLUDE_DIRS})
+add_library(webpdecode OBJECT ${WEBP_DEC_SRCS})
+target_include_directories(webpdecode PRIVATE ${CMAKE_CURRENT_BINARY_DIR}
+                                              ${CMAKE_CURRENT_SOURCE_DIR})
+add_library(webpdspdecode OBJECT ${WEBP_DSP_COMMON_SRCS} ${WEBP_DSP_DEC_SRCS})
+target_include_directories(webpdspdecode PRIVATE ${CMAKE_CURRENT_BINARY_DIR}
+                                                 ${CMAKE_CURRENT_SOURCE_DIR})
+add_library(webputilsdecode OBJECT ${WEBP_UTILS_COMMON_SRCS}
+                                   ${WEBP_UTILS_DEC_SRCS})
+target_include_directories(webputilsdecode PRIVATE ${CMAKE_CURRENT_BINARY_DIR}
+                                                   ${CMAKE_CURRENT_SOURCE_DIR})
+add_library(
+  webpdecoder $<TARGET_OBJECTS:webpdecode> $<TARGET_OBJECTS:webpdspdecode>
+              $<TARGET_OBJECTS:webputilsdecode>)
+if(XCODE)
+  libwebp_add_stub_file(webpdecoder)
+endif()
+target_link_libraries(webpdecoder ${WEBP_DEP_LIBRARIES})
+target_include_directories(
+  webpdecoder PRIVATE ${CMAKE_CURRENT_BINARY_DIR} ${CMAKE_CURRENT_SOURCE_DIR}
+  INTERFACE $<BUILD_INTERFACE:${CMAKE_CURRENT_SOURCE_DIR}>
+            $<INSTALL_INTERFACE:${CMAKE_INSTALL_INCLUDEDIR}>)
+set_target_properties(
+  webpdecoder
+  PROPERTIES PUBLIC_HEADER "${CMAKE_CURRENT_SOURCE_DIR}/src/webp/decode.h;\
+${CMAKE_CURRENT_SOURCE_DIR}/src/webp/types.h")
+
+configure_pkg_config("src/libwebpdecoder.pc")
+
+# Build the webp library.
+add_library(webpencode OBJECT ${WEBP_ENC_SRCS})
+target_include_directories(
+  webpencode PRIVATE ${CMAKE_CURRENT_BINARY_DIR} ${CMAKE_CURRENT_SOURCE_DIR}
+                     ${CMAKE_CURRENT_SOURCE_DIR}/src)
+add_library(webpdsp OBJECT ${WEBP_DSP_COMMON_SRCS} ${WEBP_DSP_DEC_SRCS}
+                           ${WEBP_DSP_ENC_SRCS})
+target_include_directories(webpdsp PRIVATE ${CMAKE_CURRENT_BINARY_DIR}
+                                           ${CMAKE_CURRENT_SOURCE_DIR})
+add_library(webputils OBJECT ${WEBP_UTILS_COMMON_SRCS} ${WEBP_UTILS_DEC_SRCS}
+                             ${WEBP_UTILS_ENC_SRCS})
+target_include_directories(webputils PRIVATE ${CMAKE_CURRENT_BINARY_DIR}
+                                             ${CMAKE_CURRENT_SOURCE_DIR})
+add_library(webp $<TARGET_OBJECTS:webpdecode> $<TARGET_OBJECTS:webpdsp>
+                 $<TARGET_OBJECTS:webpencode> $<TARGET_OBJECTS:webputils>)
+target_link_libraries(webp sharpyuv)
+if(XCODE)
+  libwebp_add_stub_file(webp)
+endif()
+target_link_libraries(webp ${WEBP_DEP_LIBRARIES})
+target_include_directories(
+  webp PRIVATE ${CMAKE_CURRENT_SOURCE_DIR} ${CMAKE_CURRENT_BINARY_DIR}
+  PUBLIC $<BUILD_INTERFACE:${CMAKE_CURRENT_SOURCE_DIR}/src>
+         $<INSTALL_INTERFACE:include>)
+set_target_properties(
+  webp
+  PROPERTIES PUBLIC_HEADER "${CMAKE_CURRENT_SOURCE_DIR}/src/webp/decode.h;\
+${CMAKE_CURRENT_SOURCE_DIR}/src/webp/encode.h;\
+${CMAKE_CURRENT_SOURCE_DIR}/src/webp/types.h")
+
+# Make sure the OBJECT libraries are built with position independent code (it is
+# not ON by default).
+set_target_properties(webpdecode webpdspdecode webputilsdecode webpencode
+                      webpdsp webputils PROPERTIES POSITION_INDEPENDENT_CODE ON)
+configure_pkg_config("src/libwebp.pc")
+
+# Build the webp demux library.
+add_library(webpdemux ${WEBP_DEMUX_SRCS})
+target_link_libraries(webpdemux webp)
+target_include_directories(
+  webpdemux PRIVATE ${CMAKE_CURRENT_SOURCE_DIR} ${CMAKE_CURRENT_BINARY_DIR}
+  PUBLIC $<INSTALL_INTERFACE:include>)
+set_target_properties(
+  webpdemux
+  PROPERTIES
+    PUBLIC_HEADER
+    "${CMAKE_CURRENT_SOURCE_DIR}/src/webp/decode.h;\
+${CMAKE_CURRENT_SOURCE_DIR}/src/webp/demux.h;\
+${CMAKE_CURRENT_SOURCE_DIR}/src/webp/mux_types.h;\
+${CMAKE_CURRENT_SOURCE_DIR}/src/webp/types.h")
+
+configure_pkg_config("src/demux/libwebpdemux.pc")
+
+set_version(src/Makefile.am webp webp)
+set_version(src/Makefile.am webpdecoder webpdecoder)
+set_version(src/demux/Makefile.am webpdemux webpdemux)
 file(READ ${CMAKE_CURRENT_SOURCE_DIR}/configure.ac CONFIGURE_FILE)
-string(REGEX MATCH
+string(REGEX MATCH "AC_INIT\\([^\n]*\\[[0-9\\.]+\\]" TMP ${CONFIGURE_FILE})
-             "AC_INIT\\([^\n]*\\[[0-9\\.]+\\]"
+string(REGEX MATCH "[0-9\\.]+" PROJECT_VERSION ${TMP})
-             TMP
-             ${CONFIGURE_FILE})
-string(REGEX MATCH
-             "[0-9\\.]+"
-             PROJECT_VERSION
-             ${TMP})
 
 # Define the libraries to install.
 list(APPEND INSTALLED_LIBRARIES webpdecoder webp webpdemux)
@@ -385,10 +430,8 @@ math(EXPR WEBP_SIMD_FILES_TO_INCLUDE_RANGE
 foreach(I_FILE RANGE ${WEBP_SIMD_FILES_TO_INCLUDE_RANGE})
   list(GET WEBP_SIMD_FILES_TO_INCLUDE ${I_FILE} FILE)
   list(GET WEBP_SIMD_FLAGS_TO_INCLUDE ${I_FILE} SIMD_COMPILE_FLAG)
-  set_source_files_properties(${FILE}
+  set_source_files_properties(${FILE} PROPERTIES COMPILE_FLAGS
-                              PROPERTIES
+                                                 ${SIMD_COMPILE_FLAG})
-                              COMPILE_FLAGS
-                              ${SIMD_COMPILE_FLAG})
 endforeach()
 
 if(NOT WEBP_BUILD_LIBWEBPMUX)
@@ -411,30 +454,28 @@ if(WEBP_BUILD_ANIM_UTILS
    OR WEBP_BUILD_DWEBP
    OR WEBP_BUILD_GIF2WEBP
    OR WEBP_BUILD_IMG2WEBP
-   OR WEBP_BUILD_VWEBP)
+   OR WEBP_BUILD_VWEBP
+   OR WEBP_BUILD_WEBPMUX
+   OR WEBP_BUILD_WEBPINFO)
   # Example utility library.
   parse_makefile_am(${CMAKE_CURRENT_SOURCE_DIR}/examples "EXAMPLEUTIL_SRCS"
                     "example_util_[^ ]*")
   list(APPEND EXAMPLEUTIL_SRCS ${CMAKE_CURRENT_SOURCE_DIR}/examples/stopwatch.h)
   add_library(exampleutil STATIC ${EXAMPLEUTIL_SRCS})
-  target_link_libraries(exampleutil imageioutil)
   target_include_directories(
-    exampleutil
+    exampleutil PUBLIC $<BUILD_INTERFACE:${CMAKE_CURRENT_SOURCE_DIR}/src>)
-    PUBLIC $<BUILD_INTERFACE:${CMAKE_CURRENT_SOURCE_DIR}/src>)
 
   parse_makefile_am(${CMAKE_CURRENT_SOURCE_DIR}/imageio "IMAGEIOUTILS_SRCS"
                     "imageio_util_[^ ]*")
   add_library(imageioutil STATIC ${IMAGEIOUTILS_SRCS})
   target_link_libraries(imageioutil webp)
+  target_link_libraries(exampleutil imageioutil)
 
   # Image-decoding utility library.
   parse_makefile_am(${CMAKE_CURRENT_SOURCE_DIR}/imageio "IMAGEDEC_SRCS"
                     "imagedec_[^ ]*")
   add_library(imagedec STATIC ${IMAGEDEC_SRCS})
-  target_link_libraries(imagedec
+  target_link_libraries(imagedec imageioutil webpdemux webp
-                        imageioutil
-                        webpdemux
-                        webp
                         ${WEBP_DEP_IMG_LIBRARIES})
 
   # Image-encoding utility library.
@@ -443,12 +484,10 @@ if(WEBP_BUILD_ANIM_UTILS
   add_library(imageenc STATIC ${IMAGEENC_SRCS})
   target_link_libraries(imageenc imageioutil webp)
 
-  set_property(TARGET exampleutil
+  set_property(
-                      imageioutil
+    TARGET exampleutil imageioutil imagedec imageenc
-                      imagedec
+    PROPERTY INCLUDE_DIRECTORIES ${CMAKE_CURRENT_SOURCE_DIR}/src
-                      imageenc
+             ${CMAKE_CURRENT_BINARY_DIR}/src)
-               PROPERTY INCLUDE_DIRECTORIES ${CMAKE_CURRENT_SOURCE_DIR}/src
-                        ${CMAKE_CURRENT_BINARY_DIR}/src)
 endif()
 
 if(WEBP_BUILD_DWEBP)
@@ -465,7 +504,8 @@ if(WEBP_BUILD_CWEBP)
   parse_makefile_am(${CMAKE_CURRENT_SOURCE_DIR}/examples "CWEBP_SRCS" "cwebp")
   add_executable(cwebp ${CWEBP_SRCS})
   target_link_libraries(cwebp exampleutil imagedec webp)
-  target_include_directories(cwebp PRIVATE ${CMAKE_CURRENT_BINARY_DIR}/src)
+  target_include_directories(cwebp PRIVATE ${CMAKE_CURRENT_BINARY_DIR}/src
+                                           ${CMAKE_CURRENT_SOURCE_DIR})
   install(TARGETS cwebp RUNTIME DESTINATION ${CMAKE_INSTALL_BINDIR})
 endif()
 
@@ -473,13 +513,12 @@ if(WEBP_BUILD_LIBWEBPMUX)
   parse_makefile_am(${CMAKE_CURRENT_SOURCE_DIR}/src/mux "WEBP_MUX_SRCS" "")
   add_library(libwebpmux ${WEBP_MUX_SRCS})
   target_link_libraries(libwebpmux webp)
-  target_include_directories(libwebpmux
+  target_include_directories(libwebpmux PRIVATE ${CMAKE_CURRENT_BINARY_DIR}
-                             PRIVATE ${CMAKE_CURRENT_BINARY_DIR}
+                                                ${CMAKE_CURRENT_SOURCE_DIR})
-                                     ${CMAKE_CURRENT_SOURCE_DIR})
+  set_version(src/mux/Makefile.am libwebpmux webpmux)
-  set_version(mux/Makefile.am libwebpmux webpmux)
+  set_target_properties(
-  set_target_properties(libwebpmux
+    libwebpmux
-                        PROPERTIES PUBLIC_HEADER
+    PROPERTIES PUBLIC_HEADER "${CMAKE_CURRENT_SOURCE_DIR}/src/webp/mux.h;\
-                                   "${CMAKE_CURRENT_SOURCE_DIR}/src/webp/mux.h;\
 ${CMAKE_CURRENT_SOURCE_DIR}/src/webp/mux_types.h;\
 ${CMAKE_CURRENT_SOURCE_DIR}/src/webp/types.h;")
   set_target_properties(libwebpmux PROPERTIES OUTPUT_NAME webpmux)
@@ -493,11 +532,7 @@ if(WEBP_BUILD_GIF2WEBP)
   parse_makefile_am(${CMAKE_CURRENT_SOURCE_DIR}/examples "GIF2WEBP_SRCS"
                     "gif2webp")
   add_executable(gif2webp ${GIF2WEBP_SRCS})
-  target_link_libraries(gif2webp
+  target_link_libraries(gif2webp exampleutil imageioutil webp libwebpmux
-                        exampleutil
-                        imageioutil
-                        webp
-                        libwebpmux
                         ${WEBP_DEP_GIF_LIBRARIES})
   target_include_directories(gif2webp PRIVATE ${CMAKE_CURRENT_BINARY_DIR}/src)
   install(TARGETS gif2webp RUNTIME DESTINATION ${CMAKE_INSTALL_BINDIR})
@@ -509,11 +544,7 @@ if(WEBP_BUILD_IMG2WEBP)
   parse_makefile_am(${CMAKE_CURRENT_SOURCE_DIR}/examples "IMG2WEBP_SRCS"
                     "img2webp")
   add_executable(img2webp ${IMG2WEBP_SRCS})
-  target_link_libraries(img2webp
+  target_link_libraries(img2webp exampleutil imagedec imageioutil webp
-                        exampleutil
-                        imagedec
-                        imageioutil
-                        webp
                         libwebpmux)
   target_include_directories(img2webp PRIVATE ${CMAKE_CURRENT_BINARY_DIR}/src)
   install(TARGETS img2webp RUNTIME DESTINATION ${CMAKE_INSTALL_BINDIR})
@@ -526,17 +557,17 @@ if(WEBP_BUILD_VWEBP)
     include_directories(${WEBP_DEP_IMG_INCLUDE_DIRS})
     parse_makefile_am(${CMAKE_CURRENT_SOURCE_DIR}/examples "VWEBP_SRCS" "vwebp")
     add_executable(vwebp ${VWEBP_SRCS})
-    target_link_libraries(vwebp
+    target_link_libraries(
-                          ${OPENGL_LIBRARIES}
+      vwebp
-                          exampleutil
+      ${OPENGL_LIBRARIES}
-                          ${GLUT_glut_LIBRARY}
+      exampleutil
-                          imageioutil
+      GLUT::GLUT
-                          webp
+      imageioutil
-                          webpdemux)
+      webp
-    target_include_directories(vwebp
+      webpdemux)
-                               PRIVATE ${GLUT_INCLUDE_DIR}
+    target_include_directories(
-                                       ${CMAKE_CURRENT_BINARY_DIR}/src
+      vwebp PRIVATE ${GLUT_INCLUDE_DIR} ${CMAKE_CURRENT_BINARY_DIR}/src
-                                       ${OPENGL_INCLUDE_DIR})
+                    ${OPENGL_INCLUDE_DIR})
     install(TARGETS vwebp RUNTIME DESTINATION ${CMAKE_INSTALL_BINDIR})
     if(${CMAKE_SYSTEM_NAME} STREQUAL "Darwin")
       check_c_compiler_flag("-Wno-deprecated-declarations" HAS_NO_DEPRECATED)
@@ -554,7 +585,8 @@ if(WEBP_BUILD_WEBPINFO)
                     "webpinfo")
   add_executable(webpinfo ${WEBPINFO_SRCS})
   target_link_libraries(webpinfo exampleutil imageioutil)
-  target_include_directories(webpinfo PRIVATE ${CMAKE_CURRENT_BINARY_DIR}/src)
+  target_include_directories(webpinfo PRIVATE ${CMAKE_CURRENT_BINARY_DIR}/src
+                                              ${CMAKE_CURRENT_SOURCE_DIR}/src)
   install(TARGETS webpinfo RUNTIME DESTINATION ${CMAKE_INSTALL_BINDIR})
 endif()
 
@@ -575,30 +607,32 @@ if(WEBP_BUILD_EXTRAS)
   parse_makefile_am(${EXTRAS_MAKEFILE} "WEBP_QUALITY_SRCS" "webp_quality")
   parse_makefile_am(${EXTRAS_MAKEFILE} "VWEBP_SDL_SRCS" "vwebp_sdl")
 
+  # libextras
+  add_library(extras STATIC ${WEBP_EXTRAS_SRCS})
+  target_include_directories(
+    extras PRIVATE ${CMAKE_CURRENT_BINARY_DIR} ${CMAKE_CURRENT_SOURCE_DIR}
+                   ${CMAKE_CURRENT_SOURCE_DIR}/src)
+
   # get_disto
   add_executable(get_disto ${GET_DISTO_SRCS})
   target_link_libraries(get_disto imagedec)
-  target_include_directories(get_disto
+  target_include_directories(get_disto PRIVATE ${CMAKE_CURRENT_SOURCE_DIR}
-                             PRIVATE ${CMAKE_CURRENT_SOURCE_DIR}
+                                               ${CMAKE_CURRENT_BINARY_DIR}/src)
-                                     ${CMAKE_CURRENT_BINARY_DIR}/src)
 
   # webp_quality
-  add_executable(webp_quality ${WEBP_QUALITY_SRCS} ${WEBP_EXTRAS_SRCS})
+  add_executable(webp_quality ${WEBP_QUALITY_SRCS})
-  target_link_libraries(webp_quality exampleutil imagedec)
+  target_link_libraries(webp_quality exampleutil imagedec extras)
-  target_include_directories(webp_quality
+  target_include_directories(webp_quality PRIVATE ${CMAKE_CURRENT_SOURCE_DIR}
-                             PRIVATE ${CMAKE_CURRENT_SOURCE_DIR}
+                                                  ${CMAKE_CURRENT_BINARY_DIR})
-                                     ${CMAKE_CURRENT_BINARY_DIR})
 
   # vwebp_sdl
   find_package(SDL)
-  if(SDL_FOUND)
+  if(WEBP_BUILD_VWEBP AND SDL_FOUND)
     add_executable(vwebp_sdl ${VWEBP_SDL_SRCS})
     target_link_libraries(vwebp_sdl ${SDL_LIBRARY} imageioutil webp)
-    target_include_directories(vwebp_sdl
+    target_include_directories(
-                               PRIVATE ${CMAKE_CURRENT_SOURCE_DIR}
+      vwebp_sdl PRIVATE ${CMAKE_CURRENT_SOURCE_DIR} ${CMAKE_CURRENT_BINARY_DIR}
-                                       ${CMAKE_CURRENT_BINARY_DIR}
+                        ${CMAKE_CURRENT_BINARY_DIR}/src ${SDL_INCLUDE_DIR})
-                                       ${CMAKE_CURRENT_BINARY_DIR}/src
-                                       ${SDL_INCLUDE_DIR})
     set(WEBP_HAVE_SDL 1)
     target_compile_definitions(vwebp_sdl PUBLIC WEBP_HAVE_SDL)
   endif()
@@ -614,9 +648,9 @@ if(WEBP_BUILD_WEBP_JS)
     set(WEBP_HAVE_SDL 1)
     set_target_properties(
       webp_js
-      PROPERTIES LINK_FLAGS "-s WASM=0 \
+      PROPERTIES LINK_FLAGS "-sWASM=0 \
-         -s EXPORTED_FUNCTIONS='[\"_WebpToSDL\"]' -s INVOKE_RUN=0 \
+         -sEXPORTED_FUNCTIONS=_WebPToSDL -sINVOKE_RUN=0 \
-         -s EXPORTED_RUNTIME_METHODS='[\"cwrap\"]'")
+         -sEXPORTED_RUNTIME_METHODS=cwrap")
     set_target_properties(webp_js PROPERTIES OUTPUT_NAME webp)
     target_compile_definitions(webp_js PUBLIC EMSCRIPTEN WEBP_HAVE_SDL)
   endif()
@@ -627,9 +661,9 @@ if(WEBP_BUILD_WEBP_JS)
   target_include_directories(webp_wasm PRIVATE ${CMAKE_CURRENT_BINARY_DIR})
   set_target_properties(
     webp_wasm
-    PROPERTIES LINK_FLAGS "-s WASM=1 \
+    PROPERTIES LINK_FLAGS "-sWASM=1 \
-       -s EXPORTED_FUNCTIONS='[\"_WebpToSDL\"]' -s INVOKE_RUN=0 \
+       -sEXPORTED_FUNCTIONS=_WebPToSDL -sINVOKE_RUN=0 \
-       -s EXPORTED_RUNTIME_METHODS='[\"cwrap\"]'")
+       -sEXPORTED_RUNTIME_METHODS=cwrap")
   target_compile_definitions(webp_wasm PUBLIC EMSCRIPTEN WEBP_HAVE_SDL)
 
   target_compile_definitions(webpdspdecode PUBLIC EMSCRIPTEN)
@@ -641,14 +675,15 @@ if(WEBP_BUILD_ANIM_UTILS)
   parse_makefile_am(${CMAKE_CURRENT_SOURCE_DIR}/examples "ANIM_DIFF_SRCS"
                     "anim_diff")
   add_executable(anim_diff ${ANIM_DIFF_SRCS})
-  target_link_libraries(anim_diff
+  target_link_libraries(
-                        exampleutil
+    anim_diff
-                        imagedec
+    exampleutil
-                        imageenc
+    imagedec
-                        imageioutil
+    imageenc
-                        webp
+    imageioutil
-                        webpdemux
+    webp
-                        ${WEBP_DEP_GIF_LIBRARIES})
+    webpdemux
+    ${WEBP_DEP_GIF_LIBRARIES})
   target_include_directories(anim_diff PRIVATE ${CMAKE_CURRENT_BINARY_DIR}/src)
 
   # anim_dump
@@ -656,45 +691,58 @@ if(WEBP_BUILD_ANIM_UTILS)
   parse_makefile_am(${CMAKE_CURRENT_SOURCE_DIR}/examples "ANIM_DUMP_SRCS"
                     "anim_dump")
   add_executable(anim_dump ${ANIM_DUMP_SRCS})
-  target_link_libraries(anim_dump
+  target_link_libraries(
-                        exampleutil
+    anim_dump
-                        imagedec
+    exampleutil
-                        imageenc
+    imagedec
-                        imageioutil
+    imageenc
-                        webp
+    imageioutil
-                        webpdemux
+    webp
-                        ${WEBP_DEP_GIF_LIBRARIES})
+    webpdemux
+    ${WEBP_DEP_GIF_LIBRARIES})
   target_include_directories(anim_dump PRIVATE ${CMAKE_CURRENT_BINARY_DIR}/src)
 endif()
 
 # Install the different headers and libraries.
-install(TARGETS ${INSTALLED_LIBRARIES}
+install(
-        EXPORT ${PROJECT_NAME}Targets
+  TARGETS ${INSTALLED_LIBRARIES}
-        PUBLIC_HEADER DESTINATION ${CMAKE_INSTALL_INCLUDEDIR}/webp
+  EXPORT ${PROJECT_NAME}Targets
-        INCLUDES
+  PUBLIC_HEADER DESTINATION ${CMAKE_INSTALL_INCLUDEDIR}/webp
-        DESTINATION ${CMAKE_INSTALL_INCLUDEDIR}
+  INCLUDES
-        ARCHIVE DESTINATION ${CMAKE_INSTALL_LIBDIR}
+  DESTINATION ${CMAKE_INSTALL_INCLUDEDIR}
-        LIBRARY DESTINATION ${CMAKE_INSTALL_LIBDIR}
+  ARCHIVE DESTINATION ${CMAKE_INSTALL_LIBDIR}
-        RUNTIME DESTINATION ${CMAKE_INSTALL_BINDIR})
+  LIBRARY DESTINATION ${CMAKE_INSTALL_LIBDIR}
+  RUNTIME DESTINATION ${CMAKE_INSTALL_BINDIR})
 set(ConfigPackageLocation ${CMAKE_INSTALL_DATADIR}/${PROJECT_NAME}/cmake/)
-install(EXPORT ${PROJECT_NAME}Targets
+install(EXPORT ${PROJECT_NAME}Targets NAMESPACE ${PROJECT_NAME}::
-        NAMESPACE ${PROJECT_NAME}::
         DESTINATION ${ConfigPackageLocation})
 
 # Create the CMake version file.
 include(CMakePackageConfigHelpers)
 write_basic_package_version_file(
   "${CMAKE_CURRENT_BINARY_DIR}/WebPConfigVersion.cmake"
-  VERSION ${PACKAGE_VERSION}
+  VERSION ${PACKAGE_VERSION} COMPATIBILITY AnyNewerVersion)
-  COMPATIBILITY AnyNewerVersion)
 
 # Create the Config file.
 include(CMakePackageConfigHelpers)
+# Fix libwebpmux reference. The target name libwebpmux is used for compatibility
+# purposes, but the library mentioned in WebPConfig.cmake should be the
+# unprefixed version. Note string(...) can be replaced with list(TRANSFORM ...)
+# if cmake_minimum_required is >= 3.12.
+string(REGEX REPLACE "libwebpmux" "webpmux" INSTALLED_LIBRARIES
+                     "${INSTALLED_LIBRARIES}")
+
+if(MSVC)
+  # For compatibility with nmake, MSVC builds use a custom prefix (lib) that
+  # needs to be included in the library name.
+  string(REGEX REPLACE "[A-Za-z0-9_]+" "${CMAKE_STATIC_LIBRARY_PREFIX}\\0"
+                       INSTALLED_LIBRARIES "${INSTALLED_LIBRARIES}")
+endif()
+
 configure_package_config_file(
   ${CMAKE_CURRENT_SOURCE_DIR}/cmake/WebPConfig.cmake.in
   ${CMAKE_CURRENT_BINARY_DIR}/WebPConfig.cmake
-  INSTALL_DESTINATION
+  INSTALL_DESTINATION ${ConfigPackageLocation})
-  ${ConfigPackageLocation})
 
 # Install the generated CMake files.
 install(FILES "${CMAKE_CURRENT_BINARY_DIR}/WebPConfigVersion.cmake"
@@ -726,7 +774,6 @@ foreach(I_MAN RANGE ${MAN_PAGES_RANGE})
   if(WEBP_BUILD_${EXEC_BUILD})
     list(GET MAN_PAGES ${I_MAN} MAN_PAGE)
     install(FILES ${CMAKE_CURRENT_SOURCE_DIR}/man/${MAN_PAGE}
-            DESTINATION ${CMAKE_INSTALL_MANDIR}/man1
+            DESTINATION ${CMAKE_INSTALL_MANDIR}/man1 COMPONENT doc)
-            COMPONENT doc)
   endif()
 endforeach()

ChangeLog

@@ -1,3 +1,278 @@
+0ba77244 update NEWS
+e763eb1e bump version to 1.3.0
+2a8686fc update AUTHORS
+106a57c1 Merge "*/Android.mk: add a check for NDK_ROOT" into main
+c5e841c4 Merge "extras: WebpToSDL -> WebPToSDL" into main
+dbc30715 Merge "xcframeworkbuild.sh: bump MACOSX_CATALYST_MIN_VERSION" into main
+6fc1a9f9 */Android.mk: add a check for NDK_ROOT
+d3e151fc doc/api.md,webp_js/README.md: Webp -> WebP
+ed92a626 extras: WebpToSDL -> WebPToSDL
+6eb0189b xcframeworkbuild.sh: bump MACOSX_CATALYST_MIN_VERSION
+1d58575b CMake: align .pc variables with autoconf
+e5fe2cfc webp-lossless-bitstream-spec,cosmetics: reflow paragraphs
+0ceeeab9 webp-lossless-bitstream-spec: add amendment note
+607611cd Merge "webp-container-spec: normalize section title case" into main
+f853685e lossless: SUBTRACT_GREEN -> SUBTRACT_GREEN_TRANSFORM
+786497e4 webp-lossless-bitstream-spec: fix inv color txfm description
+c6ac672d webp-lossless-bitstream-spec: fix num_code_lengths check
+b5700efb webp-lossless-bitstream-spec,cosmetics: grammar/capitalization
+d8ed8c11 webp-container-spec: normalize section title case
+52ec0b8f Merge changes Ie975dbb5,Ifc8c93af,I6ca7c5d6,I2e8d66f5,I152477b8 into main
+5097ef62 webp-container-spec,cosmetics: grammar/capitalization
+e3ba2b1f webp-lossless-bitstream-spec,cosmetics: reflow abstract
+1e8e3ded webp-lossless-bitstream-spec: reword abstract re alpha
+017cb6fa webp-container-spec,cosmetics: normalize range syntax
+f6a4684b webp-lossless-bitstream-spec,cosmetics: normalize range syntax
+54ebd5a3 webp-lossless-bitstream-spec: limit dist map lut to 69 cols
+44741f9c webp-lossless-bitstream-spec: fix dist mapping example
+fad0ece7 pnmdec.c: use snprintf instead of sprintf
+3f73e8f7 sharpyuv: add SharpYuvGetVersion()
+ce2f2d66 SharpYuvConvert: fix a race on SharpYuvGetCPUInfo
+a458e308 sharpyuv_dsp.h: restore sharpyuv_cpu.h include
+9ba800a7 Merge changes Id72fbf3b,Ic59d23a2 into main
+979c0ebb sharpyuv: add SharpYuvGetCPUInfo
+8bab09a4 Merge "*.pc.in: rename lib_prefix to webp_libname_prefix" into main
+769387c5 cpu.c,cosmetics: fix a typo
+a02978c2 sharpyuv/Makefile.am+cmake: add missing -lm
+28aedcb9 *.pc.in: rename lib_prefix to webp_libname_prefix
+c42e6d5a configure.ac: export an empty lib_prefix variable
+dfc843aa Merge "*.pc.in: add lib prefix to lib names w/MSVC" into main
+2498209b *.pc.in: add lib prefix to lib names w/MSVC
+ac252b61 Merge "analysis_enc.c: fix a dead store warning" into main
+56944762 analysis_enc.c: fix a dead store warning
+d34f9b99 Merge "webp-lossless-bitstream-spec: convert BNF to ABNF" into main
+dc05b4db Merge changes I96bc063c,I45880467,If9e18e5a,I6ee938e4,I0a410b28, ... into main
+83270c7f webp-container-spec: add prose for rendering process
+73b19b64 webp-container-spec: note reserved fields MUST be ignored
+57101d3f webp-lossless-bitstream-spec: improve 'small' color table stmt
+dfd32e45 webp-container-spec: remove redundant sentence
+8a6185dd doc/webp-*: fix some punctuation, grammar
+72776530 webp-lossless-bitstream-spec: convert BNF to ABNF
+d992bb08 cmake: rename cpufeatures target to cpufeatures-webp
+3ed2b275 webp-container-spec: clarify background color note
+951c292d webp-container-spec: come too late -> out of order
+902dd787 webp-container-spec: prefer hex literals
+a8f6b5ee webp-container-spec: change SHOULD to MUST w/ANIM chunk
+1dc59435 webp-container-spec: add unknown fields MUST be ignored
+280a810f webp-container-spec: make padding byte=0 a MUST
+41f0bf68 webp-container-spec: update note on trailing data
+6bdd36db webp-container-spec: clarify Chunk Size is in bytes
+87e36c48 Merge "webp_js/README.md,cosmetics: reflow some lines" into main
+5b01f321 Merge "Update Windows makefile to build libsharpyuv library." into main
+19b1a71c webp_js/README.md,cosmetics: reflow some lines
+780db756 Update Windows makefile to build libsharpyuv library.
+e407d4b3 CMakeLists.txt: replace GLUT_glut_LIBRARY w/GLUT::GLUT
+abf73d62 Merge "WebPConfig.cmake.in: add find_dependency(Threads)" into main
+25807fb4 Merge "cmake: restore compatibility with cmake < 3.12" into main
+5dbc4bfa WebPConfig.cmake.in: add find_dependency(Threads)
+b2a175dd Merge "Update wasm instructions." into main
+cb90f76b Update wasm instructions.
+02d15258 cmake: restore compatibility with cmake < 3.12
+5ba046e2 CMake: add_definitions -> add_compile_options
+e68765af dsp,neon: use vaddv in a few more places
+e8f83de2 Set libsharpyuv include dir to 'webp' subdirectory.
+15a91ab1 cmake,cosmetics: apply cmake-format
+0dd49d1a CMakeLists.txt: set @ONLY in configure_file() calls
+62b1bfe8 Merge changes I2877e7bb,I777cad70,I15af7d1a,I686e6740,If10538a9, ... into main
+95c8fe5f Merge changes Iecea3603,I9dc228ab into main
+e7c805cf picture_csp_enc.c: remove SafeInitSharpYuv
+6af8845a sharpyuv: prefer webp/types.h
+639619ce cmake: fix dll exports
+782ed48c sharpyuv,SharpYuvInit: add mutex protection when available
+cad0d5ad sharyuv_{neon,sse2}.c: merge WEBP_USE_* sections
+ef70ee06 add a few missing <stddef.h> includes for NULL
+f0f9eda4 sharpyuv.h: remove <inttypes.h>
+9b902cba Merge "picture_csp_enc.c,CheckNonOpaque: rm unneeded local" into main
+9c1d457c cmake/cpu.cmake: remove unused variable
+9ac25bcb CMakeLists.txt,win32: match naming convention used by nmake
+76c353ba picture_csp_enc.c,CheckNonOpaque: rm unneeded local
+5000de54 Merge "cwebp: fix WebPPictureHasTransparency call" into main
+e1729309 Merge "WebPPictureHasTransparency: add missing pointer check" into main
+00ff988a vp8l_enc,AddSingleSubGreen: clear int sanitizer warnings
+e2fecc22 dsp/lossless_enc.c: clear int sanitizer warnings
+129cf9e9 dsp/lossless.c: clear int sanitizer warnings
+ad7d1753 dsp/lossless_enc.c: clear int sanitizer warnings
+5037220e VP8LSubtractGreenFromBlueAndRed_C: clear int sanitizer warnings
+2ee786c7 upsampling_sse2.c: clear int sanitizer warnings
+4cc157d4 ParseOptionalChunks: clear int sanitizer warning
+892cf033 BuildHuffmanTable: clear int sanitizer warning
+3a9a4d45 VP8GetSigned: clear int sanitizer warnings
+704a3d0a dsp/lossless.c: quiet int sanitizer warnings
+1a6c109c WebPPictureHasTransparency: add missing pointer check
+c626e7d5 cwebp: fix WebPPictureHasTransparency call
+866e349c Merge tag 'v1.2.4'
+c170df38 Merge "Create libsharpyuv.a in makefile.unix." into main
+9d7ff74a Create libsharpyuv.a in makefile.unix.
+0d1f1254 update ChangeLog (tag: v1.2.4, origin/1.2.4)
+fcbc2d78 Merge "doc/*.txt: restrict code to 69 columns" into main
+4ad0e189 Merge "webp-container-spec.txt: normalize fourcc spelling" into main
+980d2488 update NEWS
+9fde8127 bump version to 1.2.4
+7a0a9935 doc/*.txt: restrict code to 69 columns
+c040a615 webp-container-spec.txt: normalize fourcc spelling
+aff1c546 dsp,x86: normalize types w/_mm_cvtsi128_si32 calls
+ab540ae0 dsp,x86: normalize types w/_mm_cvtsi32_si128 calls
+8980362e dsp,x86: normalize types w/_mm_set* calls (2)
+e626925c lossless: fix crunch mode w/WEBP_REDUCE_SIZE
+83539239 dsp,x86: normalize types w/_mm_set* calls
+8a4576ce webp-container-spec.txt: replace &amp; with &
+db870881 Merge "webp-container-spec.txt: make reserved 0 values a MUST" into main
+01d7d378 webp-lossless-bitstream-spec: number all sections
+337cf69f webp-lossless-bitstream-spec: mv Nomenclature after Intro
+79be856e Merge changes I7111d1f7,I872cd62c into main
+5b87983a webp-container-spec.txt: make reserved 0 values a MUST
+bd939123 Merge changes I7a25b1a6,I51b2c2a0,I87d0cbcf,I6ec60af6,I0a3fe9dc into main
+04764b56 libwebp.pc: add libsharpyuv to requires
+7deee810 libsharpyuv: add pkg-config file
+1a64a7e6 webp-container-spec.txt: clarify some SHOULDs
+bec2c88a webp-container-spec.txt: move ChunkHeader to terminology
+c9359332 webp-container-spec.txt: clarify 'VP8 '/'XMP ' fourccs
+70fe3063 webp-container-spec.txt: rightsize table entries
+ddbf3f3f webp-container-spec.txt: update 'key words' text
+c151e95b utils.h,WEBP_ALIGN: make bitmask unsigned
+748e92bb add WebPInt32ToMem
+3fe15b67 Merge "Build libsharpyuv as a full installable library." into main
+4f402f34 add WebPMemToInt32
+a3b68c19 Build libsharpyuv as a full installable library.
+b4994eaa CMake: set rpath for shared objects
+94cd7117 Merge "CMake: fix dylib versioning" into main
+e91451b6 Fix the lossless specs a bit more.
+231bdfb7 CMake: fix dylib versioning
+bfad7ab5 CMakeLists.txt: correct libwebpmux name in WebPConfig.cmake
+c2e3fd30 Revert "cmake: fix webpmux lib name for cmake linking"
+7366f7f3 Merge "lossless: fix crunch mode w/WEBP_REDUCE_SIZE" into main
+84163d9d lossless: fix crunch mode w/WEBP_REDUCE_SIZE
+d01c1eb3 webp-lossless-bitstream-spec,cosmetics: normalize capitalization
+8813ca8e Merge tag 'v1.2.3'
+3c4a0fbf update ChangeLog (tag: v1.2.3)
+56a480e8 dsp/cpu.h: add missing extern "C"
+62b45bdd update ChangeLog (tag: v1.2.3-rc1)
+8764ec7a Merge changes Idb037953,Id582e395 into 1.2.3
+bcb872c3 vwebp: fix file name display in windows unicode build
+67c44ac5 webpmux: fix -frame option in windows unicode build
+8278825a makefile.unix: add sharpyuv objects to clean target
+14a49e01 update NEWS
+34b1dc33 bump version to 1.2.3
+0b397fda update AUTHORS
+c16488ac update .mailmap
+5a2d929c Merge "unicode.h: set console mode before using wprintf" into main
+169f867f unicode.h: set console mode before using wprintf
+a94b855c Merge "libsharpyuv: add version defines" into main
+f83bdb52 libsharpyuv: add version defines
+bef0d797 unicode_gif.h: fix -Wdeclaration-after-statement
+404c1622 Rename Huffman coding to prefix coding in the bitstream spec
+8895f8a3 Merge "run_static_analysis.sh: fix scan-build archive path" into main
+92a673d2 Merge "Add -fvisibility=hidden flag in CMakeLists." into main
+67c1d722 Merge "add WEBP_MSAN" into main
+1124ff66 Add -fvisibility=hidden flag in CMakeLists.
+e15b3560 add WEBP_MSAN
+ec9e782a sharpyuv: remove minimum image size from sharpyuv library
+7bd07f3b run_static_analysis.sh: fix scan-build archive path
+5ecee06f Merge "sharpyuv: increase precision of gamma<->linear conversion" into main
+f81dd7d6 Merge changes I3d17d529,I53026880,I1bd61639,I6bd4b25d,Icfec8fba into main
+2d607ee6 sharpyuv: increase precision of gamma<->linear conversion
+266cbbc5 sharpyuv: add 32bit version of SharpYuvFilterRow.
+9fc12274 CMake: add src to webpinfo includes
+7d18f40a CMake: add WEBP_BUILD_WEBPINFO to list of checks for exampleutil
+11309aa5 CMake: add WEBP_BUILD_WEBPMUX to list of checks for exampleutil
+4bc762f7 CMake: link imageioutil to exampleutil after defined
+0d1b9bc4 WEBP_DEP_LIBRARIES: use Threads::Threads
+20ef48f0 Merge "sharpyuv: add support for 10/12/16 bit rgb and 10/12 bit yuv." into main
+93c54371 sharpyuv: add support for 10/12/16 bit rgb and 10/12 bit yuv.
+53cf2b49 normalize WebPValidatePicture declaration w/definition
+d3006f4b sharpyuv: slightly improve precision
+ea967098 Merge changes Ia01bd397,Ibf3771af into main
+11bc8410 Merge changes I2d317c4b,I9e77f6db into main
+30453ea4 Add an internal WebPValidatePicture.
+6c43219a Some renamings for consistency.
+4f59fa73 update .mailmap
+e74f8a62 webp-lossless-bitstream-spec,cosmetics: normalize range syntax
+5a709ec0 webp-lossless-bitstream-spec,cosmetics: fix code typo
+a2093acc webp-lossless-bitstream-spec: add amendment note
+86c66930 webp-lossless-bitstream-spec: fix BNF
+232f22da webp-lossless-bitstream-spec: fix 'simple code' snippet
+44dd765d webp-lossless-bitstream-spec: fix ColorTransform impl
+7a7e33e9 webp-lossless-bitstream-spec: fix TR-pixel right border note
+86f94ee0 Update lossless spec with Huffman codes.
+a3927cc8 sharpyuv.c,cosmetics: fix indent
+6c45cef7 Make sure the stride has a minimum value in the importer.
+0c8b0e67 sharpyuv: cleanup/cosmetic changes
+dc3841e0 {histogram,predictor}_enc: quiet int -> float warnings
+a19a25bb Replace doubles by floats in lossless misc cost estimations.
+42888f6c Add an option to enable static builds.
+7efcf3cc Merge "Fix typo in color constants: Marix -> Matrix" into main
+8f4b5c62 Fix typo in color constants: Marix -> Matrix
+90084d84 Merge "demux,IsValidExtendedFormat: remove unused variable" into main
+ed643f61 Merge changes I452d2485,Ic6d75475 into main
+8fa053d1 Rename SharpYUV to SharpYuv for consistency.
+99a87562 SharpYuvComputeConversionMatrix: quiet int->float warnings
+deb426be Makefile.vc: add sharpyuv_csp.obj to SHARPYUV_OBJS
+779597d4 demux,IsValidExtendedFormat: remove unused variable
+40e8aa57 Merge "libsharpyuv: add colorspace utilities" into main
+01a05de1 libsharpyuv: add colorspace utilities
+2de4b05a Merge changes Id9890a60,I376d81e6,I1c958838 into main
+b8bca81f Merge "configure.ac: use LT_INIT if available" into main
+e8e77b9c Merge changes I479bc487,I39864691,I5d486c2c,I186d13be into main
+7e7d5d50 Merge ".gitignore: add Android Studio & VS code dirs" into main
+10c50848 normalize label indent
+89f774e6 mux{edit,internal}: fix leaks on error
+2d3293ad ExUtilInitCommandLineArguments: fix leak on error
+ec34fd70 anim_util: fix leaks on error
+e4717287 gif2webp: fix segfault on OOM
+e3cfafaf GetBackwardReferences: fail on alloc error
+a828a59b BackwardReferencesHashChainDistanceOnly: fix segfault on OOM
+fe153fae VP8LEncodeStream: fix segfault on OOM
+919acc0e .gitignore: add Android Studio & VS code dirs
+efa0731b configure.ac: use LT_INIT if available
+0957fd69 tiffdec: add grayscale support
+e685feef Merge "Make libsharpyuv self-contained by removing dependency on cpu.c" into main
+841960b6 Make libsharpyuv self-contained by removing dependency on cpu.c
+617cf036 image_dec: add WebPGetEnabledInputFileFormats()
+7a68afaa Let SharpArgbToYuv caller pass in an RGB>YUV conversion matrix.
+34bb332c man/cwebp.1: add note about crop/resize order
+f0e9351c webp-lossless-bitstream-spec,cosmetics: fix some typos
+5ccbd6ed vp8l_dec.c,cosmetics: fix a few typos
+c3d0c2d7 fix ios build scripts after sharpyuv dep added
+d0d2292e Merge "Make libwebp depend on libsharpyuv." into main
+03d12190 alpha_processing_neon.c: fix 0x01... typo
+d55d447c Make libwebp depend on libsharpyuv.
+e4cbcdd2 Fix lossless encoding for MIPS.
+924e7ca6 alpha_processing_neon.c: fix Dispatch/ExtractAlpha_NEON
+0fa0ea54 Makefile.vc: use /MANIFEST:EMBED
+29cc95ce Basic version of libsharpyuv in libwebp, in C.
+a30f2190 examples/webpmux.c: fix a couple of typos
+66b3ce23 Fix bad overflow check in ReadTIFF()
+54e61a38 Markdownify libwebp docs and reorganize them.
+b4533deb CMakeLists.txt,cosmetics: break long line
+b9d2f9cd quant_enc.c: use WEBP_RESTRICT qualifier
+ec178f2c Add progress hook granularity in lossless
+26139c73 Rename MAX_COST to MAX_BIT_COST in histogram_enc.c
+13b82816 cmake: fix webpmux lib name for cmake linking
+88b6a396 webp-container-spec.txt,cosmetics: normalize formatting
+6f496540 Merge tag 'v1.2.2'
+4074acf8 dsp.h: bump msvc arm64 version requirement to 16.6
+b0a86089 update ChangeLog (tag: v1.2.2)
+6db8248c libwebp: Fix VP8EncTokenLoop() progress
+827a307f BMP enc: fix the transparency case
+db25f1b4 libwebp: Fix VP8EncTokenLoop() progress
+286e7fce libwebp: do not destroy jpeg codec twice on error
+6e8a4126 libwebp: do not destroy jpeg codec twice on error
+faf21968 Merge "BMP enc: fix the transparency case" into main
+480cd51d BMP enc: fix the transparency case
+9195ea05 update ChangeLog (tag: v1.2.2-rc2)
+4acae017 update NEWS
+883f0633 man/img2webp.1: update date
+567e1f44 Reword img2webp synopsis command line
+1b0c15db man/img2webp.1: update date
+17bade38 Merge "Reword img2webp synopsis command line" into main
+a80954a1 Reword img2webp synopsis command line
+f084244d anim_decode: fix alpha blending with big-endian
+b217b4ff webpinfo: fix fourcc comparison w/big-endian
+ec497b75 Merge "anim_decode: fix alpha blending with big-endian" into main
+e4886716 anim_decode: fix alpha blending with big-endian
+e3cb052c webpinfo: fix fourcc comparison w/big-endian
+a510fedb patch-check: detect duplicated files
+f035d2e4 update ChangeLog (tag: v1.2.2-rc1)
 7031946a update NEWS
 973390b6 bump version to 1.2.2
 abd6664f update AUTHORS
@@ -35,7 +310,7 @@ e23cd548 dsp.h: enable NEON w/VS2019+ ARM64 targets
 42592af8 webp,cmake: Remove unnecessary include dirs
 e298e05f Add patch-check steps in PRESUBMIT.py
 29148919 Merge tag 'v1.2.1'
-9ce5843d update ChangeLog (tag: v1.2.1, origin/1.2.1)
+9ce5843d update ChangeLog (tag: v1.2.1)
 d9191588 fuzzer/*: normalize src/ includes
 c5bc3624 fuzzer/*: normalize src/ includes
 53b6f762 fix indent
@@ -214,7 +489,7 @@ a99078c1 remove call to MBAnalyzeBestIntra4Mode for method >= 5
 6a0ff358 Enc: add a qmin / qmax range for quality factor
 0fa56f30 Merge tag 'v1.1.0'
 6cf504d0 PNM decoding: handle max_value != 255
-d7844e97 update ChangeLog (tag: v1.1.0-rc2, tag: v1.1.0, origin/1.1.0)
+d7844e97 update ChangeLog (tag: v1.1.0-rc2, tag: v1.1.0)
 7f006436 Makefile.vc: fix webp_quality.exe link
 cf047e83 Makefile.vc: fix webp_quality.exe link
 c074c653 update NEWS

Makefile.am

@@ -1,5 +1,5 @@
 ACLOCAL_AMFLAGS = -I m4
-SUBDIRS = src imageio man
+SUBDIRS = sharpyuv src imageio man
 EXTRA_DIST = COPYING autogen.sh
 
 if BUILD_EXTRAS

Makefile.vc

@@ -5,6 +5,7 @@ LIBWEBPDECODER_BASENAME = libwebpdecoder
 LIBWEBP_BASENAME = libwebp
 LIBWEBPMUX_BASENAME = libwebpmux
 LIBWEBPDEMUX_BASENAME = libwebpdemux
+LIBSHARPYUV_BASENAME = libsharpyuv
 
 !IFNDEF ARCH
 !IF ! [ cl 2>&1 | find "x86" > NUL ]
@@ -31,12 +32,11 @@ CCNODBG    = cl.exe $(NOLOGO) /O2 /DNDEBUG
 CCDEBUG    = cl.exe $(NOLOGO) /Od /Zi /D_DEBUG /RTC1
 CFLAGS     = /I. /Isrc $(NOLOGO) /W3 /EHsc /c
 CFLAGS     = $(CFLAGS) /DWIN32 /D_CRT_SECURE_NO_WARNINGS /DWIN32_LEAN_AND_MEAN
-LDFLAGS    = /LARGEADDRESSAWARE /MANIFEST /NXCOMPAT /DYNAMICBASE
+LDFLAGS    = /LARGEADDRESSAWARE /MANIFEST:EMBED /NXCOMPAT /DYNAMICBASE
 LDFLAGS    = $(LDFLAGS) $(PLATFORM_LDFLAGS)
 LNKDLL     = link.exe /DLL $(NOLOGO)
 LNKEXE     = link.exe $(NOLOGO)
 LNKLIB     = lib.exe $(NOLOGO)
-MT         = mt.exe $(NOLOGO)
 RCNODBG    = rc.exe $(NOLOGO) /l"0x0409"  # 0x409 = U.S. English
 RCDEBUG    = $(RCNODBG) /D_DEBUG
 
@@ -82,6 +82,7 @@ OUTPUT_DIRS = $(DIRBIN) $(DIRINC) $(DIRLIB) \
               $(DIROBJ)\extras \
               $(DIROBJ)\imageio \
               $(DIROBJ)\mux \
+              $(DIROBJ)\sharpyuv \
               $(DIROBJ)\utils \
 
 # Target configuration
@@ -96,6 +97,7 @@ LIBWEBPDECODER_BASENAME = $(LIBWEBPDECODER_BASENAME)_debug
 LIBWEBP_BASENAME = $(LIBWEBP_BASENAME)_debug
 LIBWEBPMUX_BASENAME = $(LIBWEBPMUX_BASENAME)_debug
 LIBWEBPDEMUX_BASENAME = $(LIBWEBPDEMUX_BASENAME)_debug
+LIBSHARPYUV_BASENAME = $(LIBSHARPYUV_BASENAME)_debug
 !ELSE IF "$(CFG)" == "release-dynamic"
 CC        = $(CCNODBG)
 RC        = $(RCNODBG)
@@ -109,6 +111,7 @@ LIBWEBPDECODER_BASENAME = $(LIBWEBPDECODER_BASENAME)_debug
 LIBWEBP_BASENAME = $(LIBWEBP_BASENAME)_debug
 LIBWEBPMUX_BASENAME = $(LIBWEBPMUX_BASENAME)_debug
 LIBWEBPDEMUX_BASENAME = $(LIBWEBPDEMUX_BASENAME)_debug
+LIBSHARPYUV_BASENAME = $(LIBSHARPYUV_BASENAME)_debug
 !ENDIF
 
 !IF "$(STATICLIBBUILD)" == "TRUE"
@@ -118,13 +121,14 @@ LIBWEBPDECODER = $(DIRLIB)\$(LIBWEBPDECODER_BASENAME).lib
 LIBWEBP = $(DIRLIB)\$(LIBWEBP_BASENAME).lib
 LIBWEBPMUX = $(DIRLIB)\$(LIBWEBPMUX_BASENAME).lib
 LIBWEBPDEMUX = $(DIRLIB)\$(LIBWEBPDEMUX_BASENAME).lib
+LIBSHARPYUV = $(DIRLIB)\$(LIBSHARPYUV_BASENAME).lib
 !ELSE IF "$(DLLBUILD)" == "TRUE"
-DLLINC = webp_dll.h
+CC     = $(CC) /I$(DIROBJ) $(RTLIB) /DWEBP_DLL
-CC     = $(CC) /I$(DIROBJ) /FI$(DLLINC) $(RTLIB) /DWEBP_DLL
 LIBWEBPDECODER = $(DIRLIB)\$(LIBWEBPDECODER_BASENAME)_dll.lib
 LIBWEBP = $(DIRLIB)\$(LIBWEBP_BASENAME)_dll.lib
 LIBWEBPMUX = $(DIRLIB)\$(LIBWEBPMUX_BASENAME)_dll.lib
 LIBWEBPDEMUX = $(DIRLIB)\$(LIBWEBPDEMUX_BASENAME)_dll.lib
+LIBSHARPYUV = $(DIRLIB)\$(LIBSHARPYUV_BASENAME)_dll.lib
 LIBWEBP_PDBNAME = $(DIROBJ)\$(LIBWEBP_BASENAME)_dll.pdb
 CFGSET = TRUE
 !ENDIF
@@ -174,6 +178,15 @@ CFLAGS = $(CFLAGS) /D_UNICODE /DUNICODE
 # A config was provided, so the library can be built.
 #
 
+SHARPYUV_OBJS = \
+    $(DIROBJ)\sharpyuv\sharpyuv.obj \
+    $(DIROBJ)\sharpyuv\sharpyuv_cpu.obj \
+    $(DIROBJ)\sharpyuv\sharpyuv_csp.obj \
+    $(DIROBJ)\sharpyuv\sharpyuv_dsp.obj \
+    $(DIROBJ)\sharpyuv\sharpyuv_gamma.obj \
+    $(DIROBJ)\sharpyuv\sharpyuv_neon.obj \
+    $(DIROBJ)\sharpyuv\sharpyuv_sse2.obj \
+
 DEC_OBJS = \
     $(DIROBJ)\dec\alpha_dec.obj \
     $(DIROBJ)\dec\buffer_dec.obj \
@@ -335,12 +348,13 @@ UTILS_ENC_OBJS = \
     $(DIROBJ)\utils\quant_levels_utils.obj \
 
 LIBWEBPDECODER_OBJS = $(DEC_OBJS) $(DSP_DEC_OBJS) $(UTILS_DEC_OBJS)
-LIBWEBP_OBJS = $(LIBWEBPDECODER_OBJS) $(ENC_OBJS) $(DSP_ENC_OBJS) \
+LIBWEBP_OBJS = $(LIBWEBPDECODER_OBJS) $(ENC_OBJS) \
-               $(UTILS_ENC_OBJS) $(DLL_OBJS)
+               $(DSP_ENC_OBJS) $(UTILS_ENC_OBJS) $(DLL_OBJS)
 LIBWEBPMUX_OBJS = $(MUX_OBJS) $(LIBWEBPMUX_OBJS)
 LIBWEBPDEMUX_OBJS = $(DEMUX_OBJS) $(LIBWEBPDEMUX_OBJS)
+LIBSHARPYUV_OBJS = $(SHARPYUV_OBJS)
 
-OUT_LIBS = $(LIBWEBPDECODER) $(LIBWEBP)
+OUT_LIBS = $(LIBWEBPDECODER) $(LIBWEBP) $(LIBSHARPYUV)
 !IF "$(ARCH)" == "ARM"
 ex: $(OUT_LIBS)
 all: ex
@@ -368,7 +382,7 @@ $(DIRBIN)\anim_dump.exe: $(EX_GIF_DEC_OBJS) $(LIBWEBPDEMUX) $(LIBWEBP)
 $(DIRBIN)\anim_dump.exe: $(IMAGEIO_ENC_OBJS)
 $(DIRBIN)\cwebp.exe: $(DIROBJ)\examples\cwebp.obj $(IMAGEIO_DEC_OBJS)
 $(DIRBIN)\cwebp.exe: $(IMAGEIO_UTIL_OBJS)
-$(DIRBIN)\cwebp.exe: $(LIBWEBPDEMUX)
+$(DIRBIN)\cwebp.exe: $(LIBWEBPDEMUX) $(LIBSHARPYUV)
 $(DIRBIN)\dwebp.exe: $(DIROBJ)\examples\dwebp.obj $(IMAGEIO_DEC_OBJS)
 $(DIRBIN)\dwebp.exe: $(IMAGEIO_ENC_OBJS)
 $(DIRBIN)\dwebp.exe: $(IMAGEIO_UTIL_OBJS)
@@ -409,17 +423,16 @@ $(EX_UTIL_OBJS) $(IMAGEIO_UTIL_OBJS): $(OUTPUT_DIRS)
 $(IMAGEIO_DEC_OBJS) $(IMAGEIO_ENC_OBJS) $(EXTRAS_OBJS): $(OUTPUT_DIRS)
 !ENDIF  # ARCH == ARM
 
+$(LIBSHARPYUV): $(LIBSHARPYUV_OBJS)
 $(LIBWEBPDECODER): $(LIBWEBPDECODER_OBJS)
-$(LIBWEBP): $(LIBWEBP_OBJS)
+$(LIBWEBP): $(LIBWEBP_OBJS) $(LIBSHARPYUV)
 $(LIBWEBPMUX): $(LIBWEBPMUX_OBJS)
 $(LIBWEBPDEMUX): $(LIBWEBPDEMUX_OBJS)
 
-$(LIBWEBP_OBJS) $(LIBWEBPMUX_OBJS) $(LIBWEBPDEMUX_OBJS): $(OUTPUT_DIRS)
+$(LIBWEBP_OBJS) $(LIBWEBPMUX_OBJS) $(LIBWEBPDEMUX_OBJS) $(LIBSHARPYUV_OBJS): \
+    $(OUTPUT_DIRS)
 
 !IF "$(DLLBUILD)" == "TRUE"
-$(LIBWEBP_OBJS) $(LIBWEBPMUX_OBJS) $(LIBWEBPDEMUX_OBJS): \
-    $(DIROBJ)\$(DLLINC)
-
 {$(DIROBJ)}.c{$(DIROBJ)}.obj:
 	$(CC) $(CFLAGS) /Fd$(LIBWEBP_PDBNAME) /Fo$@  $<
 
@@ -429,20 +442,20 @@ $(LIBWEBP_OBJS) $(LIBWEBPMUX_OBJS) $(LIBWEBPDEMUX_OBJS): \
 	$(RC) /fo$@ $<
 {src\mux}.rc{$(DIROBJ)\mux}.res:
 	$(RC) /fo$@ $<
+{sharpyuv}.rc{$(DIROBJ)\sharpyuv}.res:
+	$(RC) /fo$@ $<
 
-$(LIBWEBP): $(DIROBJ)\$(LIBWEBP_BASENAME:_debug=).res
+$(LIBSHARPYUV): $(DIROBJ)\sharpyuv\$(LIBSHARPYUV_BASENAME:_debug=).res
+$(LIBWEBP): $(LIBSHARPYUV) $(DIROBJ)\$(LIBWEBP_BASENAME:_debug=).res
 $(LIBWEBPDECODER): $(DIROBJ)\$(LIBWEBPDECODER_BASENAME:_debug=).res
 $(LIBWEBPMUX): $(LIBWEBP) $(DIROBJ)\mux\$(LIBWEBPMUX_BASENAME:_debug=).res
 $(LIBWEBPDEMUX): $(LIBWEBP) $(DIROBJ)\demux\$(LIBWEBPDEMUX_BASENAME:_debug=).res
 
-$(LIBWEBPDECODER) $(LIBWEBP) $(LIBWEBPMUX) $(LIBWEBPDEMUX):
+$(LIBWEBPDECODER) $(LIBWEBP) $(LIBWEBPMUX) $(LIBWEBPDEMUX) $(LIBSHARPYUV):
 	$(LNKDLL) /out:$(DIRBIN)\$(@B:_dll=.dll) /implib:$@ $(LFLAGS) $**
 	-xcopy $(DIROBJ)\*.pdb $(DIRLIB) /y
-
-clean::
-	@-erase /s $(DIROBJ)\$(DLLINC) 2> NUL
 !ELSE
-$(LIBWEBPDECODER) $(LIBWEBP) $(LIBWEBPMUX) $(LIBWEBPDEMUX):
+$(LIBWEBPDECODER) $(LIBWEBP) $(LIBWEBPMUX) $(LIBWEBPDEMUX) $(LIBSHARPYUV):
 	$(LNKLIB) /out:$@ $**
 	-xcopy $(DIROBJ)\*.pdb $(DIRLIB) /y
 !ENDIF
@@ -450,13 +463,6 @@ $(LIBWEBPDECODER) $(LIBWEBP) $(LIBWEBPMUX) $(LIBWEBPDEMUX):
 $(OUTPUT_DIRS):
 	@if not exist "$(@)" mkdir "$(@)"
 
-# generate a helper include to define WEBP_EXTERN suitable for the DLL build
-$(DIROBJ)\$(DLLINC):
-	@echo #ifndef WEBP_DLL_H_ > $@
-	@echo #define WEBP_DLL_H_ >> $@
-	@echo #define WEBP_EXTERN __declspec(dllexport) >> $@
-	@echo #endif  /* WEBP_DLL_H_ */ >> $@
-
 .SUFFIXES: .c .obj .res .exe
 # File-specific flag builds. Note batch rules take precedence over wildcards,
 # so for now name each file individually.
@@ -482,6 +488,8 @@ $(DIROBJ)\examples\gifdec.obj: examples\gifdec.c
 	$(CC) $(CFLAGS) /Fd$(DIROBJ)\extras\ /Fo$(DIROBJ)\extras\ $<
 {imageio}.c{$(DIROBJ)\imageio}.obj::
 	$(CC) $(CFLAGS) /Fd$(DIROBJ)\imageio\ /Fo$(DIROBJ)\imageio\ $<
+{sharpyuv}.c{$(DIROBJ)\sharpyuv}.obj::
+	$(CC) $(CFLAGS) /Fd$(DIROBJ)\sharpyuv\ /Fo$(DIROBJ)\sharpyuv\ $<
 {src\dec}.c{$(DIROBJ)\dec}.obj::
 	$(CC) $(CFLAGS) /Fd$(LIBWEBP_PDBNAME) /Fo$(DIROBJ)\dec\ $<
 {src\demux}.c{$(DIROBJ)\demux}.obj::
@@ -502,13 +510,9 @@ LNKLIBS     = $(LNKLIBS) Shell32.lib
 
 {$(DIROBJ)\examples}.obj{$(DIRBIN)}.exe:
 	$(LNKEXE) $(LDFLAGS) /OUT:$@ $** $(LNKLIBS)
-	$(MT) -manifest $@.manifest -outputresource:$@;1
-	del $@.manifest
 
 {$(DIROBJ)\extras}.obj{$(DIRBIN)}.exe:
 	$(LNKEXE) $(LDFLAGS) /OUT:$@ $** $(LNKLIBS)
-	$(MT) -manifest $@.manifest -outputresource:$@;1
-	del $@.manifest
 
 clean::
 	@-erase /s $(DIROBJ)\*.dll 2> NUL

NEWS

@@ -1,10 +1,32 @@
-- 12/15/2021: version 1.2.2
+- 12/16/2022: version 1.3.0
+  This is a binary compatible release.
+  * add libsharpyuv, which exposes -sharp_yuv/config.use_sharp_yuv
+    functionality to other libraries; libwebp now depends on this library
+  * major updates to the container and lossless bitstream docs (#448, #546,
+    #551)
+  * miscellaneous warning, bug & build fixes (#576, #583, #584)
+
+- 8/4/2022: version 1.2.4
+  This is a binary compatible release.
+  * restore CMake libwebpmux target name for compatibility with 1.2.2 (#575)
+  * fix lossless crunch mode encoding with WEBP_REDUCE_SIZE
+    (chromium: #1345547, #1345595, #1345772, #1345804)
+
+- 6/30/2022: version 1.2.3
+  This is a binary compatible release.
+  * security fix for lossless encoder (#565, chromium:1313709)
+  * improved progress granularity in WebPReportProgress() when using lossless
+  * improved precision in Sharp YUV (-sharp_yuv) conversion
+  * many corrections to webp-lossless-bitstream-spec.txt (#551)
+  * crash/leak fixes on error/OOM and other bug fixes (#558, #563, #569, #573)
+
+- 1/11/2022: version 1.2.2
   This is a binary compatible release.
   * webpmux: add "-set bgcolor A,R,G,B"
   * add ARM64 NEON support for MSVC builds (#539)
   * fix duplicate include error in Xcode when using multiple XCFrameworks in a
     project (#542)
-  * doc updates and bug fixes (#538, #544)
+  * doc updates and bug fixes (#538, #544, #548, #550)
 
 - 7/20/2021: version 1.2.1
   This is a binary compatible release.

PRESUBMIT.py

@@ -41,6 +41,7 @@ _BASH_INDENTATION = "2"
 _GIT_COMMIT_SUBJECT_LENGTH = 65
 _INCLUDE_BASH_FILES_ONLY = [r".*\.sh$"]
 _INCLUDE_MAN_FILES_ONLY = [r"man/.+\.1$"]
+_INCLUDE_SOURCE_FILES_ONLY = [r".*\.[ch]$"]
 _LIBWEBP_MAX_LINE_LENGTH = 80
 
 
@@ -72,6 +73,35 @@ def _CheckCommitSubjectLength(input_api, output_api):
   return output_api.PresubmitResult("%s\n (%4.2fs) success" % (name, duration))
 
 
+def _CheckDuplicateFiles(input_api, output_api):
+  """Ensures there are not repeated filenames."""
+  all_files = []
+  for f in input_api.change.AllFiles():
+    for include_file in _INCLUDE_SOURCE_FILES_ONLY:
+      if re.match(include_file, f):
+        all_files.append(f)
+        break
+
+  basename_to_path = {}
+  for f in all_files:
+    basename_file = input_api.basename(f)
+    if basename_file in basename_to_path:
+      basename_to_path[basename_file].append(f)
+    else:
+      basename_to_path[basename_file] = [f]
+
+  dupes = []
+  for files in basename_to_path.values():
+    if len(files) > 1:
+      dupes.extend(files)
+
+  if dupes:
+    return output_api.PresubmitError(
+        "Duplicate source files, rebase or rename some to make them unique:\n%s"
+        % dupes)
+  return output_api.PresubmitResult("No duplicates, success\n")
+
+
 def _GetFilesToSkip(input_api):
   return list(input_api.DEFAULT_FILES_TO_SKIP) + [
       r"swig/.*\.py$",
@@ -154,6 +184,7 @@ def _CommonChecks(input_api, output_api):
       input_api.canned_checks.CheckChangeHasNoStrayWhitespace(
           input_api, output_api))
   results.append(_CheckCommitSubjectLength(input_api, output_api))
+  results.append(_CheckDuplicateFiles(input_api, output_api))
 
   source_file_filter = lambda x: input_api.FilterSourceFile(
       x, files_to_skip=_GetFilesToSkip(input_api))

README

@@ -1,795 +0,0 @@
-          __   __  ____  ____  ____
-         /  \\/  \/  _ \/  _ )/  _ \
-         \       /   __/  _  \   __/
-          \__\__/\____/\_____/__/ ____  ___
-                / _/ /    \    \ /  _ \/ _/
-               /  \_/   / /   \ \   __/  \__
-               \____/____/\_____/_____/____/v1.2.2
-
-Description:
-============
-
-WebP codec: library to encode and decode images in WebP format. This package
-contains the library that can be used in other programs to add WebP support,
-as well as the command line tools 'cwebp' and 'dwebp'.
-
-See https://developers.google.com/speed/webp
-
-The latest source tree is available at
-https://chromium.googlesource.com/webm/libwebp
-
-It is released under the same license as the WebM project.
-See https://www.webmproject.org/license/software/ or the
-"COPYING" file for details. An additional intellectual
-property rights grant can be found in the file PATENTS.
-
-Building:
-=========
-
-Windows build:
---------------
-
-By running:
-
-  nmake /f Makefile.vc CFG=release-static RTLIBCFG=static OBJDIR=output
-
-the directory output\release-static\(x64|x86)\bin will contain the tools
-cwebp.exe and dwebp.exe. The directory output\release-static\(x64|x86)\lib will
-contain the libwebp static library.
-The target architecture (x86/x64) is detected by Makefile.vc from the Visual
-Studio compiler (cl.exe) available in the system path.
-
-Unix build using makefile.unix:
--------------------------------
-
-On platforms with GNU tools installed (gcc and make), running
-
-  make -f makefile.unix
-
-will build the binaries examples/cwebp and examples/dwebp, along
-with the static library src/libwebp.a. No system-wide installation
-is supplied, as this is a simple alternative to the full installation
-system based on the autoconf tools (see below).
-Please refer to makefile.unix for additional details and customizations.
-
-Using autoconf tools:
----------------------
-Prerequisites:
-A compiler (e.g., gcc), make, autoconf, automake, libtool.
-On a Debian-like system the following should install everything you need for a
-minimal build:
-$ sudo apt-get install gcc make autoconf automake libtool
-
-When building from git sources, you will need to run autogen.sh to generate the
-configure script.
-
-./configure
-make
-make install
-
-should be all you need to have the following files
-
-/usr/local/include/webp/decode.h
-/usr/local/include/webp/encode.h
-/usr/local/include/webp/types.h
-/usr/local/lib/libwebp.*
-/usr/local/bin/cwebp
-/usr/local/bin/dwebp
-
-installed.
-
-Note: A decode-only library, libwebpdecoder, is available using the
-'--enable-libwebpdecoder' flag. The encode library is built separately and can
-be installed independently using a minor modification in the corresponding
-Makefile.am configure files (see comments there). See './configure --help' for
-more options.
-
-Building for MIPS Linux:
-------------------------
-MIPS Linux toolchain stable available releases can be found at:
-https://community.imgtec.com/developers/mips/tools/codescape-mips-sdk/available-releases/
-
-# Add toolchain to PATH
-export PATH=$PATH:/path/to/toolchain/bin
-
-# 32-bit build for mips32r5 (p5600)
-HOST=mips-mti-linux-gnu
-MIPS_CFLAGS="-O3 -mips32r5 -mabi=32 -mtune=p5600 -mmsa -mfp64 \
-  -msched-weight -mload-store-pairs -fPIE"
-MIPS_LDFLAGS="-mips32r5 -mabi=32 -mmsa -mfp64 -pie"
-
-# 64-bit build for mips64r6 (i6400)
-HOST=mips-img-linux-gnu
-MIPS_CFLAGS="-O3 -mips64r6 -mabi=64 -mtune=i6400 -mmsa -mfp64 \
-  -msched-weight -mload-store-pairs -fPIE"
-MIPS_LDFLAGS="-mips64r6 -mabi=64 -mmsa -mfp64 -pie"
-
-./configure --host=${HOST} --build=`config.guess` \
-  CC="${HOST}-gcc -EL" \
-  CFLAGS="$MIPS_CFLAGS" \
-  LDFLAGS="$MIPS_LDFLAGS"
-make
-make install
-
-CMake:
-------
-With CMake, you can compile libwebp, cwebp, dwebp, gif2webp, img2webp, webpinfo
-and the JS bindings.
-
-Prerequisites:
-A compiler (e.g., gcc with autotools) and CMake.
-On a Debian-like system the following should install everything you need for a
-minimal build:
-$ sudo apt-get install build-essential cmake
-
-When building from git sources, you will need to run cmake to generate the
-makefiles.
-
-mkdir build && cd build && cmake ../
-make
-make install
-
-If you also want any of the executables, you will need to enable them through
-CMake, e.g.:
-
-cmake -DWEBP_BUILD_CWEBP=ON -DWEBP_BUILD_DWEBP=ON ../
-
-or through your favorite interface (like ccmake or cmake-qt-gui).
-
-Use option -DWEBP_UNICODE=ON for Unicode support on Windows (with chcp 65001).
-
-Finally, once installed, you can also use WebP in your CMake project by doing:
-
-find_package(WebP)
-
-which will define the CMake variables WebP_INCLUDE_DIRS and WebP_LIBRARIES.
-
-Gradle:
--------
-The support for Gradle is minimal: it only helps you compile libwebp, cwebp and
-dwebp and webpmux_example.
-
-Prerequisites:
-A compiler (e.g., gcc with autotools) and gradle.
-On a Debian-like system the following should install everything you need for a
-minimal build:
-$ sudo apt-get install build-essential gradle
-
-When building from git sources, you will need to run the Gradle wrapper with the
-appropriate target, e.g. :
-
-./gradlew buildAllExecutables
-
-SWIG bindings:
---------------
-
-To generate language bindings from swig/libwebp.swig at least swig-1.3
-(http://www.swig.org) is required.
-
-Currently the following functions are mapped:
-Decode:
-  WebPGetDecoderVersion
-  WebPGetInfo
-  WebPDecodeRGBA
-  WebPDecodeARGB
-  WebPDecodeBGRA
-  WebPDecodeBGR
-  WebPDecodeRGB
-
-Encode:
-  WebPGetEncoderVersion
-  WebPEncodeRGBA
-  WebPEncodeBGRA
-  WebPEncodeRGB
-  WebPEncodeBGR
-  WebPEncodeLosslessRGBA
-  WebPEncodeLosslessBGRA
-  WebPEncodeLosslessRGB
-  WebPEncodeLosslessBGR
-
-See swig/README for more detailed build instructions.
-
-Java bindings:
-
-To build the swig-generated JNI wrapper code at least JDK-1.5 (or equivalent)
-is necessary for enum support. The output is intended to be a shared object /
-DLL that can be loaded via System.loadLibrary("webp_jni").
-
-Python bindings:
-
-To build the swig-generated Python extension code at least Python 2.6 is
-required. Python < 2.6 may build with some minor changes to libwebp.swig or the
-generated code, but is untested.
-
-Encoding tool:
-==============
-
-The examples/ directory contains tools for encoding (cwebp) and
-decoding (dwebp) images.
-
-The easiest use should look like:
-  cwebp input.png -q 80 -o output.webp
-which will convert the input file to a WebP file using a quality factor of 80
-on a 0->100 scale (0 being the lowest quality, 100 being the best. Default
-value is 75).
-You might want to try the -lossless flag too, which will compress the source
-(in RGBA format) without any loss. The -q quality parameter will in this case
-control the amount of processing time spent trying to make the output file as
-small as possible.
-
-A longer list of options is available using the -longhelp command line flag:
-
-> cwebp -longhelp
-Usage:
- cwebp [-preset <...>] [options] in_file [-o out_file]
-
-If input size (-s) for an image is not specified, it is
-assumed to be a PNG, JPEG, TIFF or WebP file.
-Note: Animated PNG and WebP files are not supported.
-
-Options:
-  -h / -help ............. short help
-  -H / -longhelp ......... long help
-  -q <float> ............. quality factor (0:small..100:big), default=75
-  -alpha_q <int> ......... transparency-compression quality (0..100),
-                           default=100
-  -preset <string> ....... preset setting, one of:
-                            default, photo, picture,
-                            drawing, icon, text
-     -preset must come first, as it overwrites other parameters
-  -z <int> ............... activates lossless preset with given
-                           level in [0:fast, ..., 9:slowest]
-
-  -m <int> ............... compression method (0=fast, 6=slowest), default=4
-  -segments <int> ........ number of segments to use (1..4), default=4
-  -size <int> ............ target size (in bytes)
-  -psnr <float> .......... target PSNR (in dB. typically: 42)
-
-  -s <int> <int> ......... input size (width x height) for YUV
-  -sns <int> ............. spatial noise shaping (0:off, 100:max), default=50
-  -f <int> ............... filter strength (0=off..100), default=60
-  -sharpness <int> ....... filter sharpness (0:most .. 7:least sharp), default=0
-  -strong ................ use strong filter instead of simple (default)
-  -nostrong .............. use simple filter instead of strong
-  -sharp_yuv ............. use sharper (and slower) RGB->YUV conversion
-  -partition_limit <int> . limit quality to fit the 512k limit on
-                           the first partition (0=no degradation ... 100=full)
-  -pass <int> ............ analysis pass number (1..10)
-  -qrange <min> <max> .... specifies the permissible quality range
-                           (default: 0 100)
-  -crop <x> <y> <w> <h> .. crop picture with the given rectangle
-  -resize <w> <h> ........ resize picture (after any cropping)
-  -mt .................... use multi-threading if available
-  -low_memory ............ reduce memory usage (slower encoding)
-  -map <int> ............. print map of extra info
-  -print_psnr ............ prints averaged PSNR distortion
-  -print_ssim ............ prints averaged SSIM distortion
-  -print_lsim ............ prints local-similarity distortion
-  -d <file.pgm> .......... dump the compressed output (PGM file)
-  -alpha_method <int> .... transparency-compression method (0..1), default=1
-  -alpha_filter <string> . predictive filtering for alpha plane,
-                           one of: none, fast (default) or best
-  -exact ................. preserve RGB values in transparent area, default=off
-  -blend_alpha <hex> ..... blend colors against background color
-                           expressed as RGB values written in
-                           hexadecimal, e.g. 0xc0e0d0 for red=0xc0
-                           green=0xe0 and blue=0xd0
-  -noalpha ............... discard any transparency information
-  -lossless .............. encode image losslessly, default=off
-  -near_lossless <int> ... use near-lossless image
-                           preprocessing (0..100=off), default=100
-  -hint <string> ......... specify image characteristics hint,
-                           one of: photo, picture or graph
-
-  -metadata <string> ..... comma separated list of metadata to
-                           copy from the input to the output if present.
-                           Valid values: all, none (default), exif, icc, xmp
-
-  -short ................. condense printed message
-  -quiet ................. don't print anything
-  -version ............... print version number and exit
-  -noasm ................. disable all assembly optimizations
-  -v ..................... verbose, e.g. print encoding/decoding times
-  -progress .............. report encoding progress
-
-Experimental Options:
-  -jpeg_like ............. roughly match expected JPEG size
-  -af .................... auto-adjust filter strength
-  -pre <int> ............. pre-processing filter
-
-
-The main options you might want to try in order to further tune the
-visual quality are:
- -preset
- -sns
- -f
- -m
-
-Namely:
-  * 'preset' will set up a default encoding configuration targeting a
-     particular type of input. It should appear first in the list of options,
-     so that subsequent options can take effect on top of this preset.
-     Default value is 'default'.
-  * 'sns' will progressively turn on (when going from 0 to 100) some additional
-     visual optimizations (like: segmentation map re-enforcement). This option
-     will balance the bit allocation differently. It tries to take bits from the
-     "easy" parts of the picture and use them in the "difficult" ones instead.
-     Usually, raising the sns value (at fixed -q value) leads to larger files,
-     but with better quality.
-     Typical value is around '75'.
-  * 'f' option directly links to the filtering strength used by the codec's
-     in-loop processing. The higher the value, the smoother the
-     highly-compressed area will look. This is particularly useful when aiming
-     at very small files. Typical values are around 20-30. Note that using the
-     option -strong/-nostrong will change the type of filtering. Use "-f 0" to
-     turn filtering off.
-  * 'm' controls the trade-off between encoding speed and quality. Default is 4.
-     You can try -m 5 or -m 6 to explore more (time-consuming) encoding
-     possibilities. A lower value will result in faster encoding at the expense
-     of quality.
-
-Decoding tool:
-==============
-
-There is a decoding sample in examples/dwebp.c which will take
-a .webp file and decode it to a PNG image file (amongst other formats).
-This is simply to demonstrate the use of the API. You can verify the
-file test.webp decodes to exactly the same as test_ref.ppm by using:
-
- cd examples
- ./dwebp test.webp -ppm -o test.ppm
- diff test.ppm test_ref.ppm
-
-The full list of options is available using -h:
-
-> dwebp -h
-Usage: dwebp in_file [options] [-o out_file]
-
-Decodes the WebP image file to PNG format [Default].
-Note: Animated WebP files are not supported.
-
-Use following options to convert into alternate image formats:
-  -pam ......... save the raw RGBA samples as a color PAM
-  -ppm ......... save the raw RGB samples as a color PPM
-  -bmp ......... save as uncompressed BMP format
-  -tiff ........ save as uncompressed TIFF format
-  -pgm ......... save the raw YUV samples as a grayscale PGM
-                 file with IMC4 layout
-  -yuv ......... save the raw YUV samples in flat layout
-
- Other options are:
-  -version ..... print version number and exit
-  -nofancy ..... don't use the fancy YUV420 upscaler
-  -nofilter .... disable in-loop filtering
-  -nodither .... disable dithering
-  -dither <d> .. dithering strength (in 0..100)
-  -alpha_dither  use alpha-plane dithering if needed
-  -mt .......... use multi-threading
-  -crop <x> <y> <w> <h> ... crop output with the given rectangle
-  -resize <w> <h> ......... scale the output (*after* any cropping)
-  -flip ........ flip the output vertically
-  -alpha ....... only save the alpha plane
-  -incremental . use incremental decoding (useful for tests)
-  -h ........... this help message
-  -v ........... verbose (e.g. print encoding/decoding times)
-  -quiet ....... quiet mode, don't print anything
-  -noasm ....... disable all assembly optimizations
-
-WebP file analysis tool:
-========================
-
-'webpinfo' can be used to print out the chunk level structure and bitstream
-header information of WebP files. It can also check if the files are of valid
-WebP format.
-
-Usage: webpinfo [options] in_files
-Note: there could be multiple input files;
-      options must come before input files.
-Options:
-  -version ........... Print version number and exit.
-  -quiet ............. Do not show chunk parsing information.
-  -diag .............. Show parsing error diagnosis.
-  -summary ........... Show chunk stats summary.
-  -bitstream_info .... Parse bitstream header.
-
-Visualization tool:
-===================
-
-There's a little self-serve visualization tool called 'vwebp' under the
-examples/ directory. It uses OpenGL to open a simple drawing window and show
-a decoded WebP file. It's not yet integrated in the automake build system, but
-you can try to manually compile it using the recommendations below.
-
-Usage: vwebp in_file [options]
-
-Decodes the WebP image file and visualize it using OpenGL
-Options are:
-  -version ..... print version number and exit
-  -noicc ....... don't use the icc profile if present
-  -nofancy ..... don't use the fancy YUV420 upscaler
-  -nofilter .... disable in-loop filtering
-  -dither <int>  dithering strength (0..100), default=50
-  -noalphadither disable alpha plane dithering
-  -usebgcolor .. display background color
-  -mt .......... use multi-threading
-  -info ........ print info
-  -h ........... this help message
-
-Keyboard shortcuts:
-  'c' ................ toggle use of color profile
-  'b' ................ toggle background color display
-  'i' ................ overlay file information
-  'd' ................ disable blending & disposal (debug)
-  'q' / 'Q' / ESC .... quit
-
-Building:
----------
-
-Prerequisites:
-1) OpenGL & OpenGL Utility Toolkit (GLUT)
-  Linux:
-    $ sudo apt-get install freeglut3-dev mesa-common-dev
-  Mac + Xcode:
-    - These libraries should be available in the OpenGL / GLUT frameworks.
-  Windows:
-    http://freeglut.sourceforge.net/index.php#download
-
-2) (Optional) qcms (Quick Color Management System)
-  i. Download qcms from Mozilla / Chromium:
-    https://hg.mozilla.org/mozilla-central/file/0e7639e3bdfb/gfx/qcms
-    https://source.chromium.org/chromium/chromium/src/+/main:third_party/qcms/;drc=d4a2f8e1ed461d8fc05ed88d1ae2dc94c9773825
-  ii. Build and archive the source files as libqcms.a / qcms.lib
-  iii. Update makefile.unix / Makefile.vc
-    a) Define WEBP_HAVE_QCMS
-    b) Update include / library paths to reference the qcms directory.
-
-Build using makefile.unix / Makefile.vc:
-$ make -f makefile.unix examples/vwebp
-> nmake /f Makefile.vc CFG=release-static \
-    ../obj/x64/release-static/bin/vwebp.exe
-
-Animation creation tool:
-========================
-The utility 'img2webp' can turn a sequence of input images (PNG, JPEG, ...)
-into an animated WebP file. It offers fine control over duration, encoding
-modes, etc.
-
-Usage:
-
-  img2webp [file-level options] [image files...] [per-frame options...]
-
-File-level options (only used at the start of compression):
- -min_size ............ minimize size
- -loop <int> .......... loop count (default: 0, = infinite loop)
- -kmax <int> .......... maximum number of frame between key-frames
-                        (0=only keyframes)
- -kmin <int> .......... minimum number of frame between key-frames
-                        (0=disable key-frames altogether)
- -mixed ............... use mixed lossy/lossless automatic mode
- -v ................... verbose mode
- -h ................... this help
- -version ............. print version number and exit
-
-Per-frame options (only used for subsequent images input):
- -d <int> ............. frame duration in ms (default: 100)
- -lossless  ........... use lossless mode (default)
- -lossy ... ........... use lossy mode
- -q <float> ........... quality
- -m <int> ............. method to use
-
-example: img2webp -loop 2 in0.png -lossy in1.jpg
-                  -d 80 in2.tiff -o out.webp
-
-Note: if a single file name is passed as the argument, the arguments will be
-tokenized from this file. The file name must not start with the character '-'.
-
-Animated GIF conversion:
-========================
-Animated GIF files can be converted to WebP files with animation using the
-gif2webp utility available under examples/. The files can then be viewed using
-vwebp.
-
-Usage:
- gif2webp [options] gif_file -o webp_file
-Options:
-  -h / -help ............. this help
-  -lossy ................. encode image using lossy compression
-  -mixed ................. for each frame in the image, pick lossy
-                           or lossless compression heuristically
-  -q <float> ............. quality factor (0:small..100:big)
-  -m <int> ............... compression method (0=fast, 6=slowest)
-  -min_size .............. minimize output size (default:off)
-                           lossless compression by default; can be
-                           combined with -q, -m, -lossy or -mixed
-                           options
-  -kmin <int> ............ min distance between key frames
-  -kmax <int> ............ max distance between key frames
-  -f <int> ............... filter strength (0=off..100)
-  -metadata <string> ..... comma separated list of metadata to
-                           copy from the input to the output if present
-                           Valid values: all, none, icc, xmp (default)
-  -loop_compatibility .... use compatibility mode for Chrome
-                           version prior to M62 (inclusive)
-  -mt .................... use multi-threading if available
-
-  -version ............... print version number and exit
-  -v ..................... verbose
-  -quiet ................. don't print anything
-
-Building:
----------
-With the libgif development files installed, gif2webp can be built using
-makefile.unix:
-$ make -f makefile.unix examples/gif2webp
-
-or using autoconf:
-$ ./configure --enable-everything
-$ make
-
-Comparison of animated images:
-==============================
-Test utility anim_diff under examples/ can be used to compare two animated
-images (each can be GIF or WebP).
-
-Usage: anim_diff <image1> <image2> [options]
-
-Options:
-  -dump_frames <folder> dump decoded frames in PAM format
-  -min_psnr <float> ... minimum per-frame PSNR
-  -raw_comparison ..... if this flag is not used, RGB is
-                        premultiplied before comparison
-  -max_diff <int> ..... maximum allowed difference per channel
-                        between corresponding pixels in subsequent
-                        frames
-  -h .................. this help
-  -version ............ print version number and exit
-
-Building:
----------
-With the libgif development files and a C++ compiler installed, anim_diff can
-be built using makefile.unix:
-$ make -f makefile.unix examples/anim_diff
-
-or using autoconf:
-$ ./configure --enable-everything
-$ make
-
-Encoding API:
-=============
-
-The main encoding functions are available in the header src/webp/encode.h
-The ready-to-use ones are:
-size_t WebPEncodeRGB(const uint8_t* rgb, int width, int height, int stride,
-                     float quality_factor, uint8_t** output);
-size_t WebPEncodeBGR(const uint8_t* bgr, int width, int height, int stride,
-                     float quality_factor, uint8_t** output);
-size_t WebPEncodeRGBA(const uint8_t* rgba, int width, int height, int stride,
-                      float quality_factor, uint8_t** output);
-size_t WebPEncodeBGRA(const uint8_t* bgra, int width, int height, int stride,
-                      float quality_factor, uint8_t** output);
-
-They will convert raw RGB samples to a WebP data. The only control supplied
-is the quality factor.
-
-There are some variants for using the lossless format:
-
-size_t WebPEncodeLosslessRGB(const uint8_t* rgb, int width, int height,
-                             int stride, uint8_t** output);
-size_t WebPEncodeLosslessBGR(const uint8_t* bgr, int width, int height,
-                             int stride, uint8_t** output);
-size_t WebPEncodeLosslessRGBA(const uint8_t* rgba, int width, int height,
-                              int stride, uint8_t** output);
-size_t WebPEncodeLosslessBGRA(const uint8_t* bgra, int width, int height,
-                              int stride, uint8_t** output);
-
-Of course in this case, no quality factor is needed since the compression
-occurs without loss of the input values, at the expense of larger output sizes.
-
-Advanced encoding API:
-----------------------
-
-A more advanced API is based on the WebPConfig and WebPPicture structures.
-
-WebPConfig contains the encoding settings and is not tied to a particular
-picture.
-WebPPicture contains input data, on which some WebPConfig will be used for
-compression.
-The encoding flow looks like:
-
--------------------------------------- BEGIN PSEUDO EXAMPLE
-
-#include <webp/encode.h>
-
-  // Setup a config, starting form a preset and tuning some additional
-  // parameters
-  WebPConfig config;
-  if (!WebPConfigPreset(&config, WEBP_PRESET_PHOTO, quality_factor)) {
-    return 0;   // version error
-  }
-  // ... additional tuning
-  config.sns_strength = 90;
-  config.filter_sharpness = 6;
-  config_error = WebPValidateConfig(&config);  // not mandatory, but useful
-
-  // Setup the input data
-  WebPPicture pic;
-  if (!WebPPictureInit(&pic)) {
-    return 0;  // version error
-  }
-  pic.width = width;
-  pic.height = height;
-  // allocated picture of dimension width x height
-  if (!WebPPictureAlloc(&pic)) {
-    return 0;   // memory error
-  }
-  // at this point, 'pic' has been initialized as a container,
-  // and can receive the Y/U/V samples.
-  // Alternatively, one could use ready-made import functions like
-  // WebPPictureImportRGB(), which will take care of memory allocation.
-  // In any case, past this point, one will have to call
-  // WebPPictureFree(&pic) to reclaim memory.
-
-  // Set up a byte-output write method. WebPMemoryWriter, for instance.
-  WebPMemoryWriter wrt;
-  WebPMemoryWriterInit(&wrt);     // initialize 'wrt'
-
-  pic.writer = MyFileWriter;
-  pic.custom_ptr = my_opaque_structure_to_make_MyFileWriter_work;
-
-  // Compress!
-  int ok = WebPEncode(&config, &pic);   // ok = 0 => error occurred!
-  WebPPictureFree(&pic);  // must be called independently of the 'ok' result.
-
-  // output data should have been handled by the writer at that point.
-  // -> compressed data is the memory buffer described by wrt.mem / wrt.size
-
-  // deallocate the memory used by compressed data
-  WebPMemoryWriterClear(&wrt);
-
--------------------------------------- END PSEUDO EXAMPLE
-
-Decoding API:
-=============
-
-This is mainly just one function to call:
-
-#include "webp/decode.h"
-uint8_t* WebPDecodeRGB(const uint8_t* data, size_t data_size,
-                       int* width, int* height);
-
-Please have a look at the file src/webp/decode.h for the details.
-There are variants for decoding in BGR/RGBA/ARGB/BGRA order, along with
-decoding to raw Y'CbCr samples. One can also decode the image directly into a
-pre-allocated buffer.
-
-To detect a WebP file and gather the picture's dimensions, the function:
-  int WebPGetInfo(const uint8_t* data, size_t data_size,
-                  int* width, int* height);
-is supplied. No decoding is involved when using it.
-
-Incremental decoding API:
-=========================
-
-In the case when data is being progressively transmitted, pictures can still
-be incrementally decoded using a slightly more complicated API. Decoder state
-is stored into an instance of the WebPIDecoder object. This object can be
-created with the purpose of decoding either RGB or Y'CbCr samples.
-For instance:
-
-  WebPDecBuffer buffer;
-  WebPInitDecBuffer(&buffer);
-  buffer.colorspace = MODE_BGR;
-  ...
-  WebPIDecoder* idec = WebPINewDecoder(&buffer);
-
-As data is made progressively available, this incremental-decoder object
-can be used to decode the picture further. There are two (mutually exclusive)
-ways to pass freshly arrived data:
-
-either by appending the fresh bytes:
-
-  WebPIAppend(idec, fresh_data, size_of_fresh_data);
-
-or by just mentioning the new size of the transmitted data:
-
-  WebPIUpdate(idec, buffer, size_of_transmitted_buffer);
-
-Note that 'buffer' can be modified between each call to WebPIUpdate, in
-particular when the buffer is resized to accommodate larger data.
-
-These functions will return the decoding status: either VP8_STATUS_SUSPENDED if
-decoding is not finished yet or VP8_STATUS_OK when decoding is done. Any other
-status is an error condition.
-
-The 'idec' object must always be released (even upon an error condition) by
-calling: WebPDelete(idec).
-
-To retrieve partially decoded picture samples, one must use the corresponding
-method: WebPIDecGetRGB or WebPIDecGetYUVA.
-It will return the last displayable pixel row.
-
-Lastly, note that decoding can also be performed into a pre-allocated pixel
-buffer. This buffer must be passed when creating a WebPIDecoder, calling
-WebPINewRGB() or WebPINewYUVA().
-
-Please have a look at the src/webp/decode.h header for further details.
-
-Advanced Decoding API:
-======================
-
-WebP decoding supports an advanced API which provides on-the-fly cropping and
-rescaling, something of great usefulness on memory-constrained environments like
-mobile phones. Basically, the memory usage will scale with the output's size,
-not the input's, when one only needs a quick preview or a zoomed in portion of
-an otherwise too-large picture. Some CPU can be saved too, incidentally.
-
--------------------------------------- BEGIN PSEUDO EXAMPLE
-     // A) Init a configuration object
-     WebPDecoderConfig config;
-     CHECK(WebPInitDecoderConfig(&config));
-
-     // B) optional: retrieve the bitstream's features.
-     CHECK(WebPGetFeatures(data, data_size, &config.input) == VP8_STATUS_OK);
-
-     // C) Adjust 'config' options, if needed
-     config.options.no_fancy_upsampling = 1;
-     config.options.use_scaling = 1;
-     config.options.scaled_width = scaledWidth();
-     config.options.scaled_height = scaledHeight();
-     // etc.
-
-     // D) Specify 'config' output options for specifying output colorspace.
-     // Optionally the external image decode buffer can also be specified.
-     config.output.colorspace = MODE_BGRA;
-     // Optionally, the config.output can be pointed to an external buffer as
-     // well for decoding the image. This externally supplied memory buffer
-     // should be big enough to store the decoded picture.
-     config.output.u.RGBA.rgba = (uint8_t*) memory_buffer;
-     config.output.u.RGBA.stride = scanline_stride;
-     config.output.u.RGBA.size = total_size_of_the_memory_buffer;
-     config.output.is_external_memory = 1;
-
-     // E) Decode the WebP image. There are two variants w.r.t decoding image.
-     // The first one (E.1) decodes the full image and the second one (E.2) is
-     // used to incrementally decode the image using small input buffers.
-     // Any one of these steps can be used to decode the WebP image.
-
-     // E.1) Decode full image.
-     CHECK(WebPDecode(data, data_size, &config) == VP8_STATUS_OK);
-
-     // E.2) Decode image incrementally.
-     WebPIDecoder* const idec = WebPIDecode(NULL, NULL, &config);
-     CHECK(idec != NULL);
-     while (bytes_remaining > 0) {
-       VP8StatusCode status = WebPIAppend(idec, input, bytes_read);
-       if (status == VP8_STATUS_OK || status == VP8_STATUS_SUSPENDED) {
-         bytes_remaining -= bytes_read;
-       } else {
-         break;
-       }
-     }
-     WebPIDelete(idec);
-
-     // F) Decoded image is now in config.output (and config.output.u.RGBA).
-     // It can be saved, displayed or otherwise processed.
-
-     // G) Reclaim memory allocated in config's object. It's safe to call
-     // this function even if the memory is external and wasn't allocated
-     // by WebPDecode().
-     WebPFreeDecBuffer(&config.output);
-
--------------------------------------- END PSEUDO EXAMPLE
-
-Bugs:
-=====
-
-Please report all bugs to the issue tracker:
-    https://bugs.chromium.org/p/webp
-Patches welcome! See this page to get started:
-    https://www.webmproject.org/code/contribute/submitting-patches/
-
-Discuss:
-========
-
-Email: webp-discuss@webmproject.org
-Web: https://groups.google.com/a/webmproject.org/group/webp-discuss

README.md

@@ -0,0 +1,53 @@
+# WebP Codec
+
+```
+      __   __  ____  ____  ____
+     /  \\/  \/  _ \/  _ )/  _ \
+     \       /   __/  _  \   __/
+      \__\__/\____/\_____/__/ ____  ___
+            / _/ /    \    \ /  _ \/ _/
+           /  \_/   / /   \ \   __/  \__
+           \____/____/\_____/_____/____/v1.3.0
+```
+
+WebP codec is a library to encode and decode images in WebP format. This package
+contains the library that can be used in other programs to add WebP support, as
+well as the command line tools 'cwebp' and 'dwebp' to compress and decompress
+images respectively.
+
+See https://developers.google.com/speed/webp for details on the image format.
+
+The latest source tree is available at
+https://chromium.googlesource.com/webm/libwebp
+
+It is released under the same license as the WebM project. See
+https://www.webmproject.org/license/software/ or the "COPYING" file for details.
+An additional intellectual property rights grant can be found in the file
+PATENTS.
+
+## Building
+
+See the [building documentation](doc/building.md).
+
+## Encoding and Decoding Tools
+
+The examples/ directory contains tools to encode and decode images and
+animations, view information about WebP images, and more. See the
+[tools documentation](doc/tools.md).
+
+## APIs
+
+See the [APIs documentation](doc/api.md), and API usage examples in the
+`examples/` directory.
+
+## Bugs
+
+Please report all bugs to the issue tracker: https://bugs.chromium.org/p/webp
+
+Patches welcome! See [how to contribute](CONTRIBUTING.md).
+
+## Discuss
+
+Email: webp-discuss@webmproject.org
+
+Web: https://groups.google.com/a/webmproject.org/group/webp-discuss

README.mux

@@ -1,258 +0,0 @@
-          __   __  ____  ____  ____  __ __  _     __ __
-         /  \\/  \/  _ \/  _ \/  _ \/  \  \/ \___/_ / _\
-         \       /   __/  _  \   __/      /  /  (_/  /__
-          \__\__/\_____/_____/__/  \__//_/\_____/__/___/v1.2.2
-
-
-Description:
-============
-
-WebPMux: set of two libraries 'Mux' and 'Demux' for creation, extraction and
-manipulation of an extended format WebP file, which can have features like
-color profile, metadata and animation. Reference command-line tools 'webpmux'
-and 'vwebp' as well as the WebP container specification
-'doc/webp-container-spec.txt' are also provided in this package.
-
-WebP Mux tool:
-==============
-
-The examples/ directory contains a tool (webpmux) for manipulating WebP
-files. The webpmux tool can be used to create an extended format WebP file and
-also to extract or strip relevant data from such a file.
-
-A list of options is available using the -help command line flag:
-
-> webpmux -help
-Usage: webpmux -get GET_OPTIONS INPUT -o OUTPUT
-       webpmux -set SET_OPTIONS INPUT -o OUTPUT
-       webpmux -duration DURATION_OPTIONS [-duration ...]
-               INPUT -o OUTPUT
-       webpmux -strip STRIP_OPTIONS INPUT -o OUTPUT
-       webpmux -frame FRAME_OPTIONS [-frame...] [-loop LOOP_COUNT]
-               [-bgcolor BACKGROUND_COLOR] -o OUTPUT
-       webpmux -info INPUT
-       webpmux [-h|-help]
-       webpmux -version
-       webpmux argument_file_name
-
-GET_OPTIONS:
- Extract relevant data:
-   icc       get ICC profile
-   exif      get EXIF metadata
-   xmp       get XMP metadata
-   frame n   get nth frame
-
-SET_OPTIONS:
- Set color profile/metadata/parameters:
-   loop LOOP_COUNT            set the loop count
-   bgcolor BACKGROUND_COLOR   set the animation background color
-   icc  file.icc              set ICC profile
-   exif file.exif             set EXIF metadata
-   xmp  file.xmp              set XMP metadata
-   where:    'file.icc' contains the ICC profile to be set,
-             'file.exif' contains the EXIF metadata to be set
-             'file.xmp' contains the XMP metadata to be set
-
-DURATION_OPTIONS:
- Set duration of selected frames:
-   duration            set duration for each frames
-   duration,frame      set duration of a particular frame
-   duration,start,end  set duration of frames in the
-                        interval [start,end])
-   where: 'duration' is the duration in milliseconds
-          'start' is the start frame index
-          'end' is the inclusive end frame index
-           The special 'end' value '0' means: last frame.
-
-STRIP_OPTIONS:
- Strip color profile/metadata:
-   icc       strip ICC profile
-   exif      strip EXIF metadata
-   xmp       strip XMP metadata
-
-FRAME_OPTIONS(i):
- Create animation:
-   file_i +di+[xi+yi[+mi[bi]]]
-   where:    'file_i' is the i'th animation frame (WebP format),
-             'di' is the pause duration before next frame,
-             'xi','yi' specify the image offset for this frame,
-             'mi' is the dispose method for this frame (0 or 1),
-             'bi' is the blending method for this frame (+b or -b)
-
-LOOP_COUNT:
- Number of times to repeat the animation.
- Valid range is 0 to 65535 [Default: 0 (infinite)].
-
-BACKGROUND_COLOR:
- Background color of the canvas.
-  A,R,G,B
-  where:    'A', 'R', 'G' and 'B' are integers in the range 0 to 255 specifying
-            the Alpha, Red, Green and Blue component values respectively
-            [Default: 255,255,255,255]
-
-INPUT & OUTPUT are in WebP format.
-
-Note: The nature of EXIF, XMP and ICC data is not checked and is assumed to be
-valid.
-
-Note: if a single file name is passed as the argument, the arguments will be
-tokenized from this file. The file name must not start with the character '-'.
-
-Visualization tool:
-===================
-
-The examples/ directory also contains a tool (vwebp) for viewing WebP files.
-It decodes the image and visualizes it using OpenGL. See the libwebp README
-for details on building and running this program.
-
-Mux API:
-========
-The Mux API contains methods for adding data to and reading data from WebP
-files. This API currently supports XMP/EXIF metadata, ICC profile and animation.
-Other features may be added in subsequent releases.
-
-Example#1 (pseudo code): Creating a WebPMux object with image data, color
-profile and XMP metadata.
-
-  int copy_data = 0;
-  WebPMux* mux = WebPMuxNew();
-  // ... (Prepare image data).
-  WebPMuxSetImage(mux, &image, copy_data);
-  // ... (Prepare ICC profile data).
-  WebPMuxSetChunk(mux, "ICCP", &icc_profile, copy_data);
-  // ... (Prepare XMP metadata).
-  WebPMuxSetChunk(mux, "XMP ", &xmp, copy_data);
-  // Get data from mux in WebP RIFF format.
-  WebPMuxAssemble(mux, &output_data);
-  WebPMuxDelete(mux);
-  // ... (Consume output_data; e.g. write output_data.bytes to file).
-  WebPDataClear(&output_data);
-
-
-Example#2 (pseudo code): Get image and color profile data from a WebP file.
-
-  int copy_data = 0;
-  // ... (Read data from file).
-  WebPMux* mux = WebPMuxCreate(&data, copy_data);
-  WebPMuxGetFrame(mux, 1, &image);
-  // ... (Consume image; e.g. call WebPDecode() to decode the data).
-  WebPMuxGetChunk(mux, "ICCP", &icc_profile);
-  // ... (Consume icc_profile).
-  WebPMuxDelete(mux);
-  free(data);
-
-
-For a detailed Mux API reference, please refer to the header file
-(src/webp/mux.h).
-
-Demux API:
-==========
-The Demux API enables extraction of images and extended format data from
-WebP files. This API currently supports reading of XMP/EXIF metadata, ICC
-profile and animated images. Other features may be added in subsequent
-releases.
-
-Code example: Demuxing WebP data to extract all the frames, ICC profile
-and EXIF/XMP metadata.
-
-  WebPDemuxer* demux = WebPDemux(&webp_data);
-  uint32_t width = WebPDemuxGetI(demux, WEBP_FF_CANVAS_WIDTH);
-  uint32_t height = WebPDemuxGetI(demux, WEBP_FF_CANVAS_HEIGHT);
-  // ... (Get information about the features present in the WebP file).
-  uint32_t flags = WebPDemuxGetI(demux, WEBP_FF_FORMAT_FLAGS);
-
-  // ... (Iterate over all frames).
-  WebPIterator iter;
-  if (WebPDemuxGetFrame(demux, 1, &iter)) {
-    do {
-      // ... (Consume 'iter'; e.g. Decode 'iter.fragment' with WebPDecode(),
-      // ... and get other frame properties like width, height, offsets etc.
-      // ... see 'struct WebPIterator' below for more info).
-    } while (WebPDemuxNextFrame(&iter));
-    WebPDemuxReleaseIterator(&iter);
-  }
-
-  // ... (Extract metadata).
-  WebPChunkIterator chunk_iter;
-  if (flags & ICCP_FLAG) WebPDemuxGetChunk(demux, "ICCP", 1, &chunk_iter);
-  // ... (Consume the ICC profile in 'chunk_iter.chunk').
-  WebPDemuxReleaseChunkIterator(&chunk_iter);
-  if (flags & EXIF_FLAG) WebPDemuxGetChunk(demux, "EXIF", 1, &chunk_iter);
-  // ... (Consume the EXIF metadata in 'chunk_iter.chunk').
-  WebPDemuxReleaseChunkIterator(&chunk_iter);
-  if (flags & XMP_FLAG) WebPDemuxGetChunk(demux, "XMP ", 1, &chunk_iter);
-  // ... (Consume the XMP metadata in 'chunk_iter.chunk').
-  WebPDemuxReleaseChunkIterator(&chunk_iter);
-  WebPDemuxDelete(demux);
-
-
-For a detailed Demux API reference, please refer to the header file
-(src/webp/demux.h).
-
-AnimEncoder API:
-================
-The AnimEncoder API can be used to create animated WebP images.
-
-Code example:
-
-  WebPAnimEncoderOptions enc_options;
-  WebPAnimEncoderOptionsInit(&enc_options);
-  // ... (Tune 'enc_options' as needed).
-  WebPAnimEncoder* enc = WebPAnimEncoderNew(width, height, &enc_options);
-  while(<there are more frames>) {
-    WebPConfig config;
-    WebPConfigInit(&config);
-    // ... (Tune 'config' as needed).
-    WebPAnimEncoderAdd(enc, frame, duration, &config);
-  }
-  WebPAnimEncoderAssemble(enc, webp_data);
-  WebPAnimEncoderDelete(enc);
-  // ... (Write the 'webp_data' to a file, or re-mux it further).
-
-
-For a detailed AnimEncoder API reference, please refer to the header file
-(src/webp/mux.h).
-
-AnimDecoder API:
-================
-This AnimDecoder API allows decoding (possibly) animated WebP images.
-
-Code Example:
-
-  WebPAnimDecoderOptions dec_options;
-  WebPAnimDecoderOptionsInit(&dec_options);
-  // Tune 'dec_options' as needed.
-  WebPAnimDecoder* dec = WebPAnimDecoderNew(webp_data, &dec_options);
-  WebPAnimInfo anim_info;
-  WebPAnimDecoderGetInfo(dec, &anim_info);
-  for (uint32_t i = 0; i < anim_info.loop_count; ++i) {
-    while (WebPAnimDecoderHasMoreFrames(dec)) {
-      uint8_t* buf;
-      int timestamp;
-      WebPAnimDecoderGetNext(dec, &buf, &timestamp);
-      // ... (Render 'buf' based on 'timestamp').
-      // ... (Do NOT free 'buf', as it is owned by 'dec').
-    }
-    WebPAnimDecoderReset(dec);
-  }
-  const WebPDemuxer* demuxer = WebPAnimDecoderGetDemuxer(dec);
-  // ... (Do something using 'demuxer'; e.g. get EXIF/XMP/ICC data).
-  WebPAnimDecoderDelete(dec);
-
-For a detailed AnimDecoder API reference, please refer to the header file
-(src/webp/demux.h).
-
-
-Bugs:
-=====
-
-Please report all bugs to the issue tracker:
-    https://bugs.chromium.org/p/webp
-Patches welcome! See this page to get started:
-    https://www.webmproject.org/code/contribute/submitting-patches/
-
-Discuss:
-========
-
-Email: webp-discuss@webmproject.org
-Web: https://groups.google.com/a/webmproject.org/group/webp-discuss

README.webp_js

@@ -1,75 +0,0 @@
-     __   __ ____ ____ ____     __  ____
-    /  \\/  \  _ \  _ \  _ \   (__)/  __\
-    \       /  __/ _  \  __/   _)  \_   \
-     \__\__/_____/____/_/     /____/____/
-
-Description:
-============
-
-This file describes the compilation of libwebp into a JavaScript decoder
-using Emscripten and CMake.
-
- - install the Emscripten SDK following the procedure described at:
-   https://emscripten.org/docs/getting_started/downloads.html#installation-instructions-using-the-emsdk-recommended
-   After installation, you should have some global variable positioned to the
-   location of the SDK. In particular, $EMSDK should point to the
-   top-level directory containing Emscripten tools.
-
- - configure the project 'WEBP_JS' with CMake using:
-
- cd webp_js && \
- emcmake cmake -DWEBP_BUILD_WEBP_JS=ON \
-       ../
-
- - compile webp.js using 'emmake make'.
-
- - that's it! Upon completion, you should have the webp.js and
-   webp.wasm files generated.
-
-The callable JavaScript function is WebPToSDL(), which decodes a raw WebP
-bitstream into a canvas. See webp_js/index.html for a simple usage sample
-(see below for instructions).
-
-Demo HTML page:
-===============
-
-   The HTML page webp_js/index.html requires an HTTP server to serve the WebP
-   image example. It's easy to just use Python for that.
-
-cd webp_js && python -m SimpleHTTPServer 8080
-
-and then navigate to http://localhost:8080 in your favorite browser.
-
-
-Web-Assembly (WASM) version:
-============================
-
-  CMakeLists.txt is configured to build the WASM version when using
-  the option WEBP_BUILD_WEBP_JS=ON. The compilation step will assemble
-  the files 'webp_wasm.js', 'webp_wasm.wasm' in the webp_js/ directory.
-  See webp_js/index_wasm.html for a simple demo page using the WASM version
-  of the library.
-
-  You will need a fairly recent version of Emscripten (at least 2.0.18,
-  latest-upstream is recommended) and of your WASM-enabled browser to run this
-  version.
-
-Caveat:
-=======
-
-  - First decoding using the library is usually slower, due to just-in-time
-    compilation.
-
-  - Some versions of llvm produce the following compile error when SSE2 is
-    enabled.
-
-"Unsupported:   %516 = bitcast <8 x i16> %481 to i128
- LLVM ERROR: BitCast Instruction not yet supported for integer types larger than 64 bits"
-
-    The corresponding Emscripten bug is at:
-    https://github.com/kripken/emscripten/issues/3788
-
-    Therefore, SSE2 optimization is currently disabled in CMakeLists.txt.
-
-  - If WEBP_ENABLE_SIMD is set to 1 the JavaScript version (webp.js) will be
-    disabled as wasm2js does not support SIMD.

build.gradle

@@ -105,6 +105,14 @@ model {
       sources {
         c {
           source {
+            srcDir "sharpyuv"
+            include "sharpyuv.c"
+            include "sharpyuv_cpu.c"
+            include "sharpyuv_csp.c"
+            include "sharpyuv_dsp.c"
+            include "sharpyuv_gamma.c"
+            include "sharpyuv_neon.c"
+            include "sharpyuv_sse2.c"
             srcDir "src/dec"
             include "alpha_dec.c"
             include "buffer_dec.c"

cmake/WebPConfig.cmake.in

@@ -3,6 +3,11 @@ set(WEBP_VERSION ${WebP_VERSION})
 
 @PACKAGE_INIT@
 
+if(@WEBP_USE_THREAD@)
+  include(CMakeFindDependencyMacro)
+  find_dependency(Threads REQUIRED)
+endif()
+
 include ("${CMAKE_CURRENT_LIST_DIR}/@PROJECT_NAME@Targets.cmake")
 
 set(WebP_INCLUDE_DIRS "@CMAKE_INSTALL_FULL_INCLUDEDIR@")

cmake/cpu.cmake

@@ -18,7 +18,8 @@ function(webp_check_compiler_flag WEBP_SIMD_FLAG ENABLE_SIMD)
   unset(WEBP_HAVE_FLAG_${WEBP_SIMD_FLAG} CACHE)
   cmake_push_check_state()
   set(CMAKE_REQUIRED_INCLUDES ${CMAKE_CURRENT_SOURCE_DIR})
-  check_c_source_compiles("
+  check_c_source_compiles(
+    "
       #include \"${CMAKE_CURRENT_LIST_DIR}/../src/dsp/dsp.h\"
       int main(void) {
         #if !defined(WEBP_USE_${WEBP_SIMD_FLAG})
@@ -26,7 +27,8 @@ function(webp_check_compiler_flag WEBP_SIMD_FLAG ENABLE_SIMD)
         #endif
         return 0;
       }
-    " WEBP_HAVE_FLAG_${WEBP_SIMD_FLAG})
+    "
+    WEBP_HAVE_FLAG_${WEBP_SIMD_FLAG})
   cmake_pop_check_state()
   if(WEBP_HAVE_FLAG_${WEBP_SIMD_FLAG})
     set(WEBP_HAVE_${WEBP_SIMD_FLAG} 1 PARENT_SCOPE)
@@ -52,13 +54,10 @@ if(MSVC AND CMAKE_C_COMPILER_ID STREQUAL "MSVC")
   endif()
   set(SIMD_DISABLE_FLAGS)
 else()
-  set(SIMD_ENABLE_FLAGS
+  set(SIMD_ENABLE_FLAGS "-msse4.1;-msse2;-mips32;-mdspr2;-mfpu=neon;-mmsa")
-      "-msse4.1;-msse2;-mips32;-mdspr2;-mfpu=neon;-mmsa")
+  set(SIMD_DISABLE_FLAGS "-mno-sse4.1;-mno-sse2;;-mno-dspr2;;-mno-msa")
-  set(SIMD_DISABLE_FLAGS
-      "-mno-sse4.1;-mno-sse2;;-mno-dspr2;;-mno-msa")
 endif()
 
-set(WEBP_SIMD_FILES_TO_NOT_INCLUDE)
 set(WEBP_SIMD_FILES_TO_INCLUDE)
 set(WEBP_SIMD_FLAGS_TO_INCLUDE)
 
@@ -75,8 +74,8 @@ list(LENGTH WEBP_SIMD_FLAGS WEBP_SIMD_FLAGS_LENGTH)
 math(EXPR WEBP_SIMD_FLAGS_RANGE "${WEBP_SIMD_FLAGS_LENGTH} - 1")
 
 foreach(I_SIMD RANGE ${WEBP_SIMD_FLAGS_RANGE})
-  # With Emscripten 2.0.9 -msimd128 -mfpu=neon will enable NEON, but the
+  # With Emscripten 2.0.9 -msimd128 -mfpu=neon will enable NEON, but the source
-  # source will fail to compile.
+  # will fail to compile.
   if(EMSCRIPTEN AND ${I_SIMD} GREATER_EQUAL 2)
     break()
   endif()
@@ -108,7 +107,7 @@ foreach(I_SIMD RANGE ${WEBP_SIMD_FLAGS_RANGE})
   # Check which files we should include or not.
   list(GET WEBP_SIMD_FILE_EXTENSIONS ${I_SIMD} WEBP_SIMD_FILE_EXTENSION)
   file(GLOB SIMD_FILES "${CMAKE_CURRENT_LIST_DIR}/../"
-            "src/dsp/*${WEBP_SIMD_FILE_EXTENSION}")
+       "src/dsp/*${WEBP_SIMD_FILE_EXTENSION}")
   if(WEBP_HAVE_${WEBP_SIMD_FLAG})
     # Memorize the file and flags.
     foreach(FILE ${SIMD_FILES})
@@ -142,11 +141,9 @@ foreach(I_SIMD RANGE ${WEBP_SIMD_FLAGS_RANGE})
             set(COMMON_PATTERNS)
           endif()
           set(CMAKE_REQUIRED_DEFINITIONS ${SIMD_COMPILE_FLAG})
-          check_c_source_compiles("int main(void) {return 0;}"
+          check_c_source_compiles(
-                                  FLAG_${SIMD_COMPILE_FLAG}
+            "int main(void) {return 0;}" FLAG_${SIMD_COMPILE_FLAG} FAIL_REGEX
-                                  FAIL_REGEX
+            "warning: argument unused during compilation:" ${COMMON_PATTERNS})
-                                  "warning: argument unused during compilation:"
-                                  ${COMMON_PATTERNS})
           if(NOT FLAG_${SIMD_COMPILE_FLAG})
             unset(HAS_COMPILE_FLAG)
             unset(HAS_COMPILE_FLAG CACHE)

cmake/deps.cmake

@@ -10,24 +10,30 @@
 
 # Check for compiler options.
 include(CheckCSourceCompiles)
-check_c_source_compiles("
+check_c_source_compiles(
+  "
     int main(void) {
       (void)__builtin_bswap16(0);
       return 0;
     }
-  " HAVE_BUILTIN_BSWAP16)
+  "
-check_c_source_compiles("
+  HAVE_BUILTIN_BSWAP16)
+check_c_source_compiles(
+  "
     int main(void) {
       (void)__builtin_bswap32(0);
       return 0;
     }
-  " HAVE_BUILTIN_BSWAP32)
+  "
-check_c_source_compiles("
+  HAVE_BUILTIN_BSWAP32)
+check_c_source_compiles(
+  "
     int main(void) {
       (void)__builtin_bswap64(0);
       return 0;
     }
-  " HAVE_BUILTIN_BSWAP64)
+  "
+  HAVE_BUILTIN_BSWAP64)
 
 # Check for libraries.
 if(WEBP_USE_THREAD)
@@ -37,15 +43,17 @@ if(WEBP_USE_THREAD)
     if(CMAKE_USE_PTHREADS_INIT AND NOT CMAKE_SYSTEM_NAME STREQUAL "QNX")
       set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -pthread")
     endif()
-    check_c_source_compiles("
+    check_c_source_compiles(
+      "
         #include <pthread.h>
         int main (void) {
           int attr = PTHREAD_PRIO_INHERIT;
           return attr;
         }
-      " FLAG_HAVE_PTHREAD_PRIO_INHERIT)
+      "
+      FLAG_HAVE_PTHREAD_PRIO_INHERIT)
     set(HAVE_PTHREAD_PRIO_INHERIT ${FLAG_HAVE_PTHREAD_PRIO_INHERIT})
-    list(APPEND WEBP_DEP_LIBRARIES ${CMAKE_THREAD_LIBS_INIT})
+    list(APPEND WEBP_DEP_LIBRARIES Threads::Threads)
   endif()
   set(WEBP_USE_THREAD ${Threads_FOUND})
 endif()
@@ -59,14 +67,17 @@ set(WEBP_HAVE_GL ${OPENGL_FOUND})
 
 # Check if we need to link to the C math library. We do not look for it as it is
 # not found when cross-compiling, while it is here.
-check_c_source_compiles("
+check_c_source_compiles(
+  "
     #include <math.h>
     int main(int argc, char** argv) {
       return (int)pow(argc, 2.5);
     }
-  " HAVE_MATH_LIBRARY)
+  "
+  HAVE_MATH_LIBRARY)
 if(NOT HAVE_MATH_LIBRARY)
   message(STATUS "Adding -lm flag.")
+  list(APPEND SHARPYUV_DEP_LIBRARIES m)
   list(APPEND WEBP_DEP_LIBRARIES m)
 endif()
 
@@ -74,12 +85,17 @@ endif()
 set(WEBP_DEP_IMG_LIBRARIES)
 set(WEBP_DEP_IMG_INCLUDE_DIRS)
 foreach(I_LIB PNG JPEG TIFF)
+  # Disable tiff when compiling in static mode as it is failing on Ubuntu.
+  if(WEBP_LINK_STATIC AND ${I_LIB} STREQUAL "TIFF")
+    message("TIFF is disabled when statically linking.")
+    continue()
+  endif()
   find_package(${I_LIB})
   set(WEBP_HAVE_${I_LIB} ${${I_LIB}_FOUND})
   if(${I_LIB}_FOUND)
     list(APPEND WEBP_DEP_IMG_LIBRARIES ${${I_LIB}_LIBRARIES})
     list(APPEND WEBP_DEP_IMG_INCLUDE_DIRS ${${I_LIB}_INCLUDE_DIR}
-                ${${I_LIB}_INCLUDE_DIRS})
+         ${${I_LIB}_INCLUDE_DIRS})
   endif()
 endforeach()
 if(WEBP_DEP_IMG_INCLUDE_DIRS)
@@ -100,13 +116,15 @@ if(GIF_FOUND)
   cmake_push_check_state()
   set(CMAKE_REQUIRED_LIBRARIES ${GIF_LIBRARIES})
   set(CMAKE_REQUIRED_INCLUDES ${GIF_INCLUDE_DIR})
-  check_c_source_compiles("
+  check_c_source_compiles(
+    "
       #include <gif_lib.h>
       int main(void) {
         (void)DGifOpenFileHandle;
         return 0;
       }
-      " GIF_COMPILES)
+      "
+    GIF_COMPILES)
   cmake_pop_check_state()
   if(GIF_COMPILES)
     list(APPEND WEBP_DEP_GIF_LIBRARIES ${GIF_LIBRARIES})
@@ -138,10 +156,7 @@ check_include_files(windows.h HAVE_WINDOWS_H)
 
 # Windows specifics
 if(HAVE_WINCODEC_H)
-  list(APPEND WEBP_DEP_LIBRARIES
+  list(APPEND WEBP_DEP_LIBRARIES shlwapi ole32 windowscodecs)
-              shlwapi
-              ole32
-              windowscodecs)
 endif()
 
 # Check for SIMD extensions.
@@ -153,17 +168,12 @@ set(PACKAGE_NAME ${PROJECT_NAME})
 
 # Read from configure.ac.
 file(READ ${CMAKE_CURRENT_SOURCE_DIR}/configure.ac CONFIGURE_AC)
-string(REGEX MATCHALL
+string(REGEX MATCHALL "\\[([0-9a-z\\.:/]*)\\]" CONFIGURE_AC_PACKAGE_INFO
-             "\\[([0-9a-z\\.:/]*)\\]"
-             CONFIGURE_AC_PACKAGE_INFO
              ${CONFIGURE_AC})
 function(strip_bracket VAR)
   string(LENGTH ${${VAR}} TMP_LEN)
   math(EXPR TMP_LEN ${TMP_LEN}-2)
-  string(SUBSTRING ${${VAR}}
+  string(SUBSTRING ${${VAR}} 1 ${TMP_LEN} TMP_SUB)
-                   1
-                   ${TMP_LEN}
-                   TMP_SUB)
   set(${VAR} ${TMP_SUB} PARENT_SCOPE)
 endfunction()
 

configure.ac

@@ -1,4 +1,4 @@
-AC_INIT([libwebp], [1.2.2],
+AC_INIT([libwebp], [1.3.0],
         [https://bugs.chromium.org/p/webp],,
         [https://developers.google.com/speed/webp])
 AC_CANONICAL_HOST
@@ -9,7 +9,8 @@ dnl === automake >= 1.12 requires this for 'unusual archivers' support.
 dnl === it must occur before LT_INIT (AC_PROG_LIBTOOL).
 m4_ifdef([AM_PROG_AR], [AM_PROG_AR])
 
-AC_PROG_LIBTOOL
+dnl === AC_PROG_LIBTOOL is deprecated.
+m4_ifdef([LT_INIT], [LT_INIT], [AC_PROG_LIBTOOL])
 AC_PROG_SED
 AM_PROG_CC_C_O
 
@@ -27,7 +28,8 @@ AC_ARG_ENABLE([everything],
               AS_HELP_STRING([--enable-everything],
                              [Enable all optional targets. These can still be
                               disabled with --disable-target]),
-              [SET_IF_UNSET([enable_libwebpdecoder], [$enableval])
+              [SET_IF_UNSET([enable_libsharpyuv], [$enableval])
+               SET_IF_UNSET([enable_libwebpdecoder], [$enableval])
                SET_IF_UNSET([enable_libwebpdemux], [$enableval])
                SET_IF_UNSET([enable_libwebpextras], [$enableval])
                SET_IF_UNSET([enable_libwebpmux], [$enableval])])
@@ -747,10 +749,13 @@ fi
 
 dnl =========================
 
+dnl Add an empty webp_libname_prefix variable for use in *.pc.in.
+AC_SUBST([webp_libname_prefix])
 AC_CONFIG_MACRO_DIR([m4])
 AC_CONFIG_HEADERS([src/webp/config.h])
 AC_CONFIG_FILES([Makefile src/Makefile man/Makefile \
                  examples/Makefile extras/Makefile imageio/Makefile \
+                 sharpyuv/Makefile sharpyuv/libsharpyuv.pc \
                  src/dec/Makefile src/enc/Makefile src/dsp/Makefile \
                  src/demux/Makefile src/mux/Makefile \
                  src/utils/Makefile \

doc/README

@@ -1,29 +0,0 @@
-
-Generate libwebp Container Spec Docs from Text Source
-=====================================================
-
-HTML generation requires kramdown [1], easily installed as a
-rubygem [2].  Rubygems installation should satisfy dependencies
-automatically.
-
-[1]: https://kramdown.gettalong.org/
-[2]: https://rubygems.org/
-
-HTML generation can then be done from the project root:
-
-$ kramdown doc/webp-container-spec.txt --template doc/template.html > \
-  doc/output/webp-container-spec.html
-
-kramdown can optionally syntax highlight code blocks, using CodeRay [3],
-a dependency of kramdown that rubygems will install automatically.  The
-following will apply inline CSS styling; an external stylesheet is not
-needed.
-
-$ kramdown doc/webp-lossless-bitstream-spec.txt --template \
-  doc/template.html --coderay-css style --coderay-line-numbers ' ' \
-  --coderay-default-lang c > \
-  doc/output/webp-lossless-bitstream-spec.html
-
-Optimally, use kramdown 0.13.7 or newer if syntax highlighting desired.
-
-[3]: https://github.com/rubychan/coderay

doc/api.md

@@ -0,0 +1,385 @@
+# WebP APIs
+
+## Encoding API
+
+The main encoding functions are available in the header src/webp/encode.h
+
+The ready-to-use ones are:
+
+```c
+size_t WebPEncodeRGB(const uint8_t* rgb, int width, int height, int stride,
+                     float quality_factor, uint8_t** output);
+size_t WebPEncodeBGR(const uint8_t* bgr, int width, int height, int stride,
+                     float quality_factor, uint8_t** output);
+size_t WebPEncodeRGBA(const uint8_t* rgba, int width, int height, int stride,
+                      float quality_factor, uint8_t** output);
+size_t WebPEncodeBGRA(const uint8_t* bgra, int width, int height, int stride,
+                      float quality_factor, uint8_t** output);
+```
+
+They will convert raw RGB samples to a WebP data. The only control supplied is
+the quality factor.
+
+There are some variants for using the lossless format:
+
+```c
+size_t WebPEncodeLosslessRGB(const uint8_t* rgb, int width, int height,
+                             int stride, uint8_t** output);
+size_t WebPEncodeLosslessBGR(const uint8_t* bgr, int width, int height,
+                             int stride, uint8_t** output);
+size_t WebPEncodeLosslessRGBA(const uint8_t* rgba, int width, int height,
+                              int stride, uint8_t** output);
+size_t WebPEncodeLosslessBGRA(const uint8_t* bgra, int width, int height,
+                              int stride, uint8_t** output);
+```
+
+Of course in this case, no quality factor is needed since the compression occurs
+without loss of the input values, at the expense of larger output sizes.
+
+### Advanced encoding API
+
+A more advanced API is based on the WebPConfig and WebPPicture structures.
+
+WebPConfig contains the encoding settings and is not tied to a particular
+picture. WebPPicture contains input data, on which some WebPConfig will be used
+for compression. The encoding flow looks like:
+
+```c
+#include <webp/encode.h>
+
+// Setup a config, starting form a preset and tuning some additional
+// parameters
+WebPConfig config;
+if (!WebPConfigPreset(&config, WEBP_PRESET_PHOTO, quality_factor)) {
+  return 0;   // version error
+}
+// ... additional tuning
+config.sns_strength = 90;
+config.filter_sharpness = 6;
+config_error = WebPValidateConfig(&config);  // not mandatory, but useful
+
+// Setup the input data
+WebPPicture pic;
+if (!WebPPictureInit(&pic)) {
+  return 0;  // version error
+}
+pic.width = width;
+pic.height = height;
+// allocated picture of dimension width x height
+if (!WebPPictureAlloc(&pic)) {
+  return 0;   // memory error
+}
+// at this point, 'pic' has been initialized as a container,
+// and can receive the Y/U/V samples.
+// Alternatively, one could use ready-made import functions like
+// WebPPictureImportRGB(), which will take care of memory allocation.
+// In any case, past this point, one will have to call
+// WebPPictureFree(&pic) to reclaim memory.
+
+// Set up a byte-output write method. WebPMemoryWriter, for instance.
+WebPMemoryWriter wrt;
+WebPMemoryWriterInit(&wrt);     // initialize 'wrt'
+
+pic.writer = MyFileWriter;
+pic.custom_ptr = my_opaque_structure_to_make_MyFileWriter_work;
+
+// Compress!
+int ok = WebPEncode(&config, &pic);   // ok = 0 => error occurred!
+WebPPictureFree(&pic);  // must be called independently of the 'ok' result.
+
+// output data should have been handled by the writer at that point.
+// -> compressed data is the memory buffer described by wrt.mem / wrt.size
+
+// deallocate the memory used by compressed data
+WebPMemoryWriterClear(&wrt);
+```
+
+## Decoding API
+
+This is mainly just one function to call:
+
+```c
+#include "webp/decode.h"
+uint8_t* WebPDecodeRGB(const uint8_t* data, size_t data_size,
+                       int* width, int* height);
+```
+
+Please have a look at the file src/webp/decode.h for the details. There are
+variants for decoding in BGR/RGBA/ARGB/BGRA order, along with decoding to raw
+Y'CbCr samples. One can also decode the image directly into a pre-allocated
+buffer.
+
+To detect a WebP file and gather the picture's dimensions, the function:
+
+```c
+int WebPGetInfo(const uint8_t* data, size_t data_size,
+                int* width, int* height);
+```
+
+is supplied. No decoding is involved when using it.
+
+### Incremental decoding API
+
+In the case when data is being progressively transmitted, pictures can still be
+incrementally decoded using a slightly more complicated API. Decoder state is
+stored into an instance of the WebPIDecoder object. This object can be created
+with the purpose of decoding either RGB or Y'CbCr samples. For instance:
+
+```c
+WebPDecBuffer buffer;
+WebPInitDecBuffer(&buffer);
+buffer.colorspace = MODE_BGR;
+...
+WebPIDecoder* idec = WebPINewDecoder(&buffer);
+```
+
+As data is made progressively available, this incremental-decoder object can be
+used to decode the picture further. There are two (mutually exclusive) ways to
+pass freshly arrived data:
+
+either by appending the fresh bytes:
+
+```c
+WebPIAppend(idec, fresh_data, size_of_fresh_data);
+```
+
+or by just mentioning the new size of the transmitted data:
+
+```c
+WebPIUpdate(idec, buffer, size_of_transmitted_buffer);
+```
+
+Note that 'buffer' can be modified between each call to WebPIUpdate, in
+particular when the buffer is resized to accommodate larger data.
+
+These functions will return the decoding status: either VP8_STATUS_SUSPENDED if
+decoding is not finished yet or VP8_STATUS_OK when decoding is done. Any other
+status is an error condition.
+
+The 'idec' object must always be released (even upon an error condition) by
+calling: WebPDelete(idec).
+
+To retrieve partially decoded picture samples, one must use the corresponding
+method: WebPIDecGetRGB or WebPIDecGetYUVA. It will return the last displayable
+pixel row.
+
+Lastly, note that decoding can also be performed into a pre-allocated pixel
+buffer. This buffer must be passed when creating a WebPIDecoder, calling
+WebPINewRGB() or WebPINewYUVA().
+
+Please have a look at the src/webp/decode.h header for further details.
+
+### Advanced Decoding API
+
+WebP decoding supports an advanced API which provides on-the-fly cropping and
+rescaling, something of great usefulness on memory-constrained environments like
+mobile phones. Basically, the memory usage will scale with the output's size,
+not the input's, when one only needs a quick preview or a zoomed in portion of
+an otherwise too-large picture. Some CPU can be saved too, incidentally.
+
+```c
+// A) Init a configuration object
+WebPDecoderConfig config;
+CHECK(WebPInitDecoderConfig(&config));
+
+// B) optional: retrieve the bitstream's features.
+CHECK(WebPGetFeatures(data, data_size, &config.input) == VP8_STATUS_OK);
+
+// C) Adjust 'config' options, if needed
+config.options.no_fancy_upsampling = 1;
+config.options.use_scaling = 1;
+config.options.scaled_width = scaledWidth();
+config.options.scaled_height = scaledHeight();
+// etc.
+
+// D) Specify 'config' output options for specifying output colorspace.
+// Optionally the external image decode buffer can also be specified.
+config.output.colorspace = MODE_BGRA;
+// Optionally, the config.output can be pointed to an external buffer as
+// well for decoding the image. This externally supplied memory buffer
+// should be big enough to store the decoded picture.
+config.output.u.RGBA.rgba = (uint8_t*) memory_buffer;
+config.output.u.RGBA.stride = scanline_stride;
+config.output.u.RGBA.size = total_size_of_the_memory_buffer;
+config.output.is_external_memory = 1;
+
+// E) Decode the WebP image. There are two variants w.r.t decoding image.
+// The first one (E.1) decodes the full image and the second one (E.2) is
+// used to incrementally decode the image using small input buffers.
+// Any one of these steps can be used to decode the WebP image.
+
+// E.1) Decode full image.
+CHECK(WebPDecode(data, data_size, &config) == VP8_STATUS_OK);
+
+// E.2) Decode image incrementally.
+WebPIDecoder* const idec = WebPIDecode(NULL, NULL, &config);
+CHECK(idec != NULL);
+while (bytes_remaining > 0) {
+  VP8StatusCode status = WebPIAppend(idec, input, bytes_read);
+  if (status == VP8_STATUS_OK || status == VP8_STATUS_SUSPENDED) {
+    bytes_remaining -= bytes_read;
+  } else {
+    break;
+  }
+}
+WebPIDelete(idec);
+
+// F) Decoded image is now in config.output (and config.output.u.RGBA).
+// It can be saved, displayed or otherwise processed.
+
+// G) Reclaim memory allocated in config's object. It's safe to call
+// this function even if the memory is external and wasn't allocated
+// by WebPDecode().
+WebPFreeDecBuffer(&config.output);
+```
+
+## WebP Mux
+
+WebPMux is a set of two libraries 'Mux' and 'Demux' for creation, extraction and
+manipulation of an extended format WebP file, which can have features like color
+profile, metadata and animation. Reference command-line tools `webpmux` and
+`vwebp` as well as the WebP container specification
+'doc/webp-container-spec.txt' are also provided in this package, see the
+[tools documentation](tools.md).
+
+### Mux API
+
+The Mux API contains methods for adding data to and reading data from WebP
+files. This API currently supports XMP/EXIF metadata, ICC profile and animation.
+Other features may be added in subsequent releases.
+
+Example#1 (pseudo code): Creating a WebPMux object with image data, color
+profile and XMP metadata.
+
+```c
+int copy_data = 0;
+WebPMux* mux = WebPMuxNew();
+// ... (Prepare image data).
+WebPMuxSetImage(mux, &image, copy_data);
+// ... (Prepare ICC profile data).
+WebPMuxSetChunk(mux, "ICCP", &icc_profile, copy_data);
+// ... (Prepare XMP metadata).
+WebPMuxSetChunk(mux, "XMP ", &xmp, copy_data);
+// Get data from mux in WebP RIFF format.
+WebPMuxAssemble(mux, &output_data);
+WebPMuxDelete(mux);
+// ... (Consume output_data; e.g. write output_data.bytes to file).
+WebPDataClear(&output_data);
+```
+
+Example#2 (pseudo code): Get image and color profile data from a WebP file.
+
+```c
+int copy_data = 0;
+// ... (Read data from file).
+WebPMux* mux = WebPMuxCreate(&data, copy_data);
+WebPMuxGetFrame(mux, 1, &image);
+// ... (Consume image; e.g. call WebPDecode() to decode the data).
+WebPMuxGetChunk(mux, "ICCP", &icc_profile);
+// ... (Consume icc_profile).
+WebPMuxDelete(mux);
+free(data);
+```
+
+For a detailed Mux API reference, please refer to the header file
+(src/webp/mux.h).
+
+### Demux API
+
+The Demux API enables extraction of images and extended format data from WebP
+files. This API currently supports reading of XMP/EXIF metadata, ICC profile and
+animated images. Other features may be added in subsequent releases.
+
+Code example: Demuxing WebP data to extract all the frames, ICC profile and
+EXIF/XMP metadata.
+
+```c
+WebPDemuxer* demux = WebPDemux(&webp_data);
+uint32_t width = WebPDemuxGetI(demux, WEBP_FF_CANVAS_WIDTH);
+uint32_t height = WebPDemuxGetI(demux, WEBP_FF_CANVAS_HEIGHT);
+// ... (Get information about the features present in the WebP file).
+uint32_t flags = WebPDemuxGetI(demux, WEBP_FF_FORMAT_FLAGS);
+
+// ... (Iterate over all frames).
+WebPIterator iter;
+if (WebPDemuxGetFrame(demux, 1, &iter)) {
+  do {
+    // ... (Consume 'iter'; e.g. Decode 'iter.fragment' with WebPDecode(),
+    // ... and get other frame properties like width, height, offsets etc.
+    // ... see 'struct WebPIterator' below for more info).
+  } while (WebPDemuxNextFrame(&iter));
+  WebPDemuxReleaseIterator(&iter);
+}
+
+// ... (Extract metadata).
+WebPChunkIterator chunk_iter;
+if (flags & ICCP_FLAG) WebPDemuxGetChunk(demux, "ICCP", 1, &chunk_iter);
+// ... (Consume the ICC profile in 'chunk_iter.chunk').
+WebPDemuxReleaseChunkIterator(&chunk_iter);
+if (flags & EXIF_FLAG) WebPDemuxGetChunk(demux, "EXIF", 1, &chunk_iter);
+// ... (Consume the EXIF metadata in 'chunk_iter.chunk').
+WebPDemuxReleaseChunkIterator(&chunk_iter);
+if (flags & XMP_FLAG) WebPDemuxGetChunk(demux, "XMP ", 1, &chunk_iter);
+// ... (Consume the XMP metadata in 'chunk_iter.chunk').
+WebPDemuxReleaseChunkIterator(&chunk_iter);
+WebPDemuxDelete(demux);
+```
+
+For a detailed Demux API reference, please refer to the header file
+(src/webp/demux.h).
+
+## AnimEncoder API
+
+The AnimEncoder API can be used to create animated WebP images.
+
+Code example:
+
+```c
+WebPAnimEncoderOptions enc_options;
+WebPAnimEncoderOptionsInit(&enc_options);
+// ... (Tune 'enc_options' as needed).
+WebPAnimEncoder* enc = WebPAnimEncoderNew(width, height, &enc_options);
+while(<there are more frames>) {
+  WebPConfig config;
+  WebPConfigInit(&config);
+  // ... (Tune 'config' as needed).
+  WebPAnimEncoderAdd(enc, frame, duration, &config);
+}
+WebPAnimEncoderAssemble(enc, webp_data);
+WebPAnimEncoderDelete(enc);
+// ... (Write the 'webp_data' to a file, or re-mux it further).
+```
+
+For a detailed AnimEncoder API reference, please refer to the header file
+(src/webp/mux.h).
+
+## AnimDecoder API
+
+This AnimDecoder API allows decoding (possibly) animated WebP images.
+
+Code Example:
+
+```c
+WebPAnimDecoderOptions dec_options;
+WebPAnimDecoderOptionsInit(&dec_options);
+// Tune 'dec_options' as needed.
+WebPAnimDecoder* dec = WebPAnimDecoderNew(webp_data, &dec_options);
+WebPAnimInfo anim_info;
+WebPAnimDecoderGetInfo(dec, &anim_info);
+for (uint32_t i = 0; i < anim_info.loop_count; ++i) {
+  while (WebPAnimDecoderHasMoreFrames(dec)) {
+    uint8_t* buf;
+    int timestamp;
+    WebPAnimDecoderGetNext(dec, &buf, &timestamp);
+    // ... (Render 'buf' based on 'timestamp').
+    // ... (Do NOT free 'buf', as it is owned by 'dec').
+  }
+  WebPAnimDecoderReset(dec);
+}
+const WebPDemuxer* demuxer = WebPAnimDecoderGetDemuxer(dec);
+// ... (Do something using 'demuxer'; e.g. get EXIF/XMP/ICC data).
+WebPAnimDecoderDelete(dec);
+```
+
+For a detailed AnimDecoder API reference, please refer to the header file
+(src/webp/demux.h).

doc/building.md

@@ -0,0 +1,213 @@
+# Building
+
+## Windows build
+
+By running:
+
+```batch
+nmake /f Makefile.vc CFG=release-static RTLIBCFG=static OBJDIR=output
+```
+
+the directory `output\release-static\(x64|x86)\bin` will contain the tools
+cwebp.exe and dwebp.exe. The directory `output\release-static\(x64|x86)\lib`
+will contain the libwebp static library. The target architecture (x86/x64) is
+detected by Makefile.vc from the Visual Studio compiler (cl.exe) available in
+the system path.
+
+## Unix build using makefile.unix
+
+On platforms with GNU tools installed (gcc and make), running
+
+```shell
+make -f makefile.unix
+```
+
+will build the binaries examples/cwebp and examples/dwebp, along with the static
+library src/libwebp.a. No system-wide installation is supplied, as this is a
+simple alternative to the full installation system based on the autoconf tools
+(see below). Please refer to makefile.unix for additional details and
+customizations.
+
+## Using autoconf tools
+
+Prerequisites: a compiler (e.g., gcc), make, autoconf, automake, libtool.
+
+On a Debian-like system the following should install everything you need for a
+minimal build:
+
+```shell
+$ sudo apt-get install gcc make autoconf automake libtool
+```
+
+When building from git sources, you will need to run autogen.sh to generate the
+configure script.
+
+```shell
+./configure
+make
+make install
+```
+
+should be all you need to have the following files
+
+```
+/usr/local/include/webp/decode.h
+/usr/local/include/webp/encode.h
+/usr/local/include/webp/types.h
+/usr/local/lib/libwebp.*
+/usr/local/bin/cwebp
+/usr/local/bin/dwebp
+```
+
+installed.
+
+Note: A decode-only library, libwebpdecoder, is available using the
+`--enable-libwebpdecoder` flag. The encode library is built separately and can
+be installed independently using a minor modification in the corresponding
+Makefile.am configure files (see comments there). See `./configure --help` for
+more options.
+
+## Building for MIPS Linux
+
+MIPS Linux toolchain stable available releases can be found at:
+https://community.imgtec.com/developers/mips/tools/codescape-mips-sdk/available-releases/
+
+```shell
+# Add toolchain to PATH
+export PATH=$PATH:/path/to/toolchain/bin
+
+# 32-bit build for mips32r5 (p5600)
+HOST=mips-mti-linux-gnu
+MIPS_CFLAGS="-O3 -mips32r5 -mabi=32 -mtune=p5600 -mmsa -mfp64 \
+  -msched-weight -mload-store-pairs -fPIE"
+MIPS_LDFLAGS="-mips32r5 -mabi=32 -mmsa -mfp64 -pie"
+
+# 64-bit build for mips64r6 (i6400)
+HOST=mips-img-linux-gnu
+MIPS_CFLAGS="-O3 -mips64r6 -mabi=64 -mtune=i6400 -mmsa -mfp64 \
+  -msched-weight -mload-store-pairs -fPIE"
+MIPS_LDFLAGS="-mips64r6 -mabi=64 -mmsa -mfp64 -pie"
+
+./configure --host=${HOST} --build=`config.guess` \
+  CC="${HOST}-gcc -EL" \
+  CFLAGS="$MIPS_CFLAGS" \
+  LDFLAGS="$MIPS_LDFLAGS"
+make
+make install
+```
+
+## CMake
+
+With CMake, you can compile libwebp, cwebp, dwebp, gif2webp, img2webp, webpinfo
+and the JS bindings.
+
+Prerequisites: a compiler (e.g., gcc with autotools) and CMake.
+
+On a Debian-like system the following should install everything you need for a
+minimal build:
+
+```shell
+$ sudo apt-get install build-essential cmake
+```
+
+When building from git sources, you will need to run cmake to generate the
+makefiles.
+
+```shell
+mkdir build && cd build && cmake ../
+make
+make install
+```
+
+If you also want any of the executables, you will need to enable them through
+CMake, e.g.:
+
+```shell
+cmake -DWEBP_BUILD_CWEBP=ON -DWEBP_BUILD_DWEBP=ON ../
+```
+
+or through your favorite interface (like ccmake or cmake-qt-gui).
+
+Use option `-DWEBP_UNICODE=ON` for Unicode support on Windows (with chcp 65001).
+
+Finally, once installed, you can also use WebP in your CMake project by doing:
+
+```cmake
+find_package(WebP)
+```
+
+which will define the CMake variables WebP_INCLUDE_DIRS and WebP_LIBRARIES.
+
+## Gradle
+
+The support for Gradle is minimal: it only helps you compile libwebp, cwebp and
+dwebp and webpmux_example.
+
+Prerequisites: a compiler (e.g., gcc with autotools) and gradle.
+
+On a Debian-like system the following should install everything you need for a
+minimal build:
+
+```shell
+$ sudo apt-get install build-essential gradle
+```
+
+When building from git sources, you will need to run the Gradle wrapper with the
+appropriate target, e.g. :
+
+```shell
+./gradlew buildAllExecutables
+```
+
+## SWIG bindings
+
+To generate language bindings from swig/libwebp.swig at least swig-1.3
+(http://www.swig.org) is required.
+
+Currently the following functions are mapped:
+
+Decode:
+
+```
+WebPGetDecoderVersion
+WebPGetInfo
+WebPDecodeRGBA
+WebPDecodeARGB
+WebPDecodeBGRA
+WebPDecodeBGR
+WebPDecodeRGB
+```
+
+Encode:
+
+```
+WebPGetEncoderVersion
+WebPEncodeRGBA
+WebPEncodeBGRA
+WebPEncodeRGB
+WebPEncodeBGR
+WebPEncodeLosslessRGBA
+WebPEncodeLosslessBGRA
+WebPEncodeLosslessRGB
+WebPEncodeLosslessBGR
+```
+
+See also the [swig documentation](../swig/README.md) for more detailed build
+instructions and usage examples.
+
+### Java bindings
+
+To build the swig-generated JNI wrapper code at least JDK-1.5 (or equivalent) is
+necessary for enum support. The output is intended to be a shared object / DLL
+that can be loaded via `System.loadLibrary("webp_jni")`.
+
+### Python bindings
+
+To build the swig-generated Python extension code at least Python 2.6 is
+required. Python < 2.6 may build with some minor changes to libwebp.swig or the
+generated code, but is untested.
+
+## Javascript decoder
+
+Libwebp can be compiled into a JavaScript decoder using Emscripten and CMake.
+See the [corresponding documentation](../README.md)

doc/specs_generation.md

@@ -0,0 +1,26 @@
+# Generate libwebp Container Spec Docs from Text Source
+
+HTML generation requires [kramdown](https://kramdown.gettalong.org/), easily
+installed as a [rubygem](https://rubygems.org/). Rubygems installation should
+satisfy dependencies automatically.
+
+HTML generation can then be done from the project root:
+
+```shell
+$ kramdown doc/webp-container-spec.txt --template doc/template.html > \
+  doc/output/webp-container-spec.html
+```
+
+kramdown can optionally syntax highlight code blocks, using
+[CodeRay](https://github.com/rubychan/coderay), a dependency of kramdown that
+rubygems will install automatically. The following will apply inline CSS
+styling; an external stylesheet is not needed.
+
+```shell
+$ kramdown doc/webp-lossless-bitstream-spec.txt --template \
+  doc/template.html --coderay-css style --coderay-line-numbers ' ' \
+  --coderay-default-lang c > \
+  doc/output/webp-lossless-bitstream-spec.html
+```
+
+Optimally, use kramdown 0.13.7 or newer if syntax highlighting desired.

doc/tools.md

@@ -0,0 +1,512 @@
+# WebP tools
+
+## Encoding tool
+
+The examples/ directory contains tools for encoding (cwebp) and decoding (dwebp)
+images.
+
+The easiest use should look like:
+
+```shell
+cwebp input.png -q 80 -o output.webp
+```
+
+which will convert the input file to a WebP file using a quality factor of 80 on
+a 0->100 scale (0 being the lowest quality, 100 being the best. Default value is
+75).
+
+You might want to try the `-lossless` flag too, which will compress the source
+(in RGBA format) without any loss. The `-q` quality parameter will in this case
+control the amount of processing time spent trying to make the output file as
+small as possible.
+
+A longer list of options is available using the `-longhelp` command line flag:
+
+```shell
+> cwebp -longhelp
+Usage:
+ cwebp [-preset <...>] [options] in_file [-o out_file]
+```
+
+If input size (-s) for an image is not specified, it is assumed to be a PNG,
+JPEG, TIFF or WebP file. Note: Animated PNG and WebP files are not supported.
+
+Options:
+
+```
+-h / -help ............. short help
+-H / -longhelp ......... long help
+-q <float> ............. quality factor (0:small..100:big), default=75
+-alpha_q <int> ......... transparency-compression quality (0..100),
+                         default=100
+-preset <string> ....... preset setting, one of:
+                          default, photo, picture,
+                          drawing, icon, text
+   -preset must come first, as it overwrites other parameters
+-z <int> ............... activates lossless preset with given
+                         level in [0:fast, ..., 9:slowest]
+
+-m <int> ............... compression method (0=fast, 6=slowest), default=4
+-segments <int> ........ number of segments to use (1..4), default=4
+-size <int> ............ target size (in bytes)
+-psnr <float> .......... target PSNR (in dB. typically: 42)
+
+-s <int> <int> ......... input size (width x height) for YUV
+-sns <int> ............. spatial noise shaping (0:off, 100:max), default=50
+-f <int> ............... filter strength (0=off..100), default=60
+-sharpness <int> ....... filter sharpness (0:most .. 7:least sharp), default=0
+-strong ................ use strong filter instead of simple (default)
+-nostrong .............. use simple filter instead of strong
+-sharp_yuv ............. use sharper (and slower) RGB->YUV conversion
+-partition_limit <int> . limit quality to fit the 512k limit on
+                         the first partition (0=no degradation ... 100=full)
+-pass <int> ............ analysis pass number (1..10)
+-qrange <min> <max> .... specifies the permissible quality range
+                         (default: 0 100)
+-crop <x> <y> <w> <h> .. crop picture with the given rectangle
+-resize <w> <h> ........ resize picture (*after* any cropping)
+-mt .................... use multi-threading if available
+-low_memory ............ reduce memory usage (slower encoding)
+-map <int> ............. print map of extra info
+-print_psnr ............ prints averaged PSNR distortion
+-print_ssim ............ prints averaged SSIM distortion
+-print_lsim ............ prints local-similarity distortion
+-d <file.pgm> .......... dump the compressed output (PGM file)
+-alpha_method <int> .... transparency-compression method (0..1), default=1
+-alpha_filter <string> . predictive filtering for alpha plane,
+                         one of: none, fast (default) or best
+-exact ................. preserve RGB values in transparent area, default=off
+-blend_alpha <hex> ..... blend colors against background color
+                         expressed as RGB values written in
+                         hexadecimal, e.g. 0xc0e0d0 for red=0xc0
+                         green=0xe0 and blue=0xd0
+-noalpha ............... discard any transparency information
+-lossless .............. encode image losslessly, default=off
+-near_lossless <int> ... use near-lossless image
+                         preprocessing (0..100=off), default=100
+-hint <string> ......... specify image characteristics hint,
+                         one of: photo, picture or graph
+
+-metadata <string> ..... comma separated list of metadata to
+                         copy from the input to the output if present.
+                         Valid values: all, none (default), exif, icc, xmp
+
+-short ................. condense printed message
+-quiet ................. don't print anything
+-version ............... print version number and exit
+-noasm ................. disable all assembly optimizations
+-v ..................... verbose, e.g. print encoding/decoding times
+-progress .............. report encoding progress
+```
+
+Experimental Options:
+
+```
+-jpeg_like ............. roughly match expected JPEG size
+-af .................... auto-adjust filter strength
+-pre <int> ............. pre-processing filter
+```
+
+The main options you might want to try in order to further tune the visual
+quality are:
+
+-preset -sns -f -m
+
+Namely:
+
+*   `preset` will set up a default encoding configuration targeting a particular
+    type of input. It should appear first in the list of options, so that
+    subsequent options can take effect on top of this preset. Default value is
+    'default'.
+*   `sns` will progressively turn on (when going from 0 to 100) some additional
+    visual optimizations (like: segmentation map re-enforcement). This option
+    will balance the bit allocation differently. It tries to take bits from the
+    "easy" parts of the picture and use them in the "difficult" ones instead.
+    Usually, raising the sns value (at fixed -q value) leads to larger files,
+    but with better quality. Typical value is around '75'.
+*   `f` option directly links to the filtering strength used by the codec's
+    in-loop processing. The higher the value, the smoother the highly-compressed
+    area will look. This is particularly useful when aiming at very small files.
+    Typical values are around 20-30. Note that using the option
+    -strong/-nostrong will change the type of filtering. Use "-f 0" to turn
+    filtering off.
+*   `m` controls the trade-off between encoding speed and quality. Default is 4.
+    You can try -m 5 or -m 6 to explore more (time-consuming) encoding
+    possibilities. A lower value will result in faster encoding at the expense
+    of quality.
+
+## Decoding tool
+
+There is a decoding sample in examples/dwebp.c which will take a .webp file and
+decode it to a PNG image file (amongst other formats). This is simply to
+demonstrate the use of the API. You can verify the file test.webp decodes to
+exactly the same as test_ref.ppm by using:
+
+```shell
+cd examples
+./dwebp test.webp -ppm -o test.ppm
+diff test.ppm test_ref.ppm
+```
+
+The full list of options is available using -h:
+
+```shell
+> dwebp -h
+Usage: dwebp in_file [options] [-o out_file]
+```
+
+Decodes the WebP image file to PNG format [Default]. Note: Animated WebP files
+are not supported.
+
+Use following options to convert into alternate image formats:
+
+```
+-pam ......... save the raw RGBA samples as a color PAM
+-ppm ......... save the raw RGB samples as a color PPM
+-bmp ......... save as uncompressed BMP format
+-tiff ........ save as uncompressed TIFF format
+-pgm ......... save the raw YUV samples as a grayscale PGM
+               file with IMC4 layout
+-yuv ......... save the raw YUV samples in flat layout
+```
+
+Other options are:
+
+```
+-version ..... print version number and exit
+-nofancy ..... don't use the fancy YUV420 upscaler
+-nofilter .... disable in-loop filtering
+-nodither .... disable dithering
+-dither <d> .. dithering strength (in 0..100)
+-alpha_dither  use alpha-plane dithering if needed
+-mt .......... use multi-threading
+-crop <x> <y> <w> <h> ... crop output with the given rectangle
+-resize <w> <h> ......... resize output (*after* any cropping)
+-flip ........ flip the output vertically
+-alpha ....... only save the alpha plane
+-incremental . use incremental decoding (useful for tests)
+-h ........... this help message
+-v ........... verbose (e.g. print encoding/decoding times)
+-quiet ....... quiet mode, don't print anything
+-noasm ....... disable all assembly optimizations
+```
+
+## WebP file analysis tool
+
+`webpinfo` can be used to print out the chunk level structure and bitstream
+header information of WebP files. It can also check if the files are of valid
+WebP format.
+
+Usage:
+
+```shell
+webpinfo [options] in_files
+```
+
+Note: there could be multiple input files; options must come before input files.
+
+Options:
+
+```
+-version ........... Print version number and exit.
+-quiet ............. Do not show chunk parsing information.
+-diag .............. Show parsing error diagnosis.
+-summary ........... Show chunk stats summary.
+-bitstream_info .... Parse bitstream header.
+```
+
+## Visualization tool
+
+There's a little self-serve visualization tool called 'vwebp' under the
+examples/ directory. It uses OpenGL to open a simple drawing window and show a
+decoded WebP file. It's not yet integrated in the automake build system, but you
+can try to manually compile it using the recommendations below.
+
+Usage:
+
+```shell
+vwebp in_file [options]
+```
+
+Decodes the WebP image file and visualize it using OpenGL
+
+Options are:
+
+```
+-version ..... print version number and exit
+-noicc ....... don't use the icc profile if present
+-nofancy ..... don't use the fancy YUV420 upscaler
+-nofilter .... disable in-loop filtering
+-dither <int>  dithering strength (0..100), default=50
+-noalphadither disable alpha plane dithering
+-usebgcolor .. display background color
+-mt .......... use multi-threading
+-info ........ print info
+-h ........... this help message
+```
+
+Keyboard shortcuts:
+
+```
+'c' ................ toggle use of color profile
+'b' ................ toggle background color display
+'i' ................ overlay file information
+'d' ................ disable blending & disposal (debug)
+'q' / 'Q' / ESC .... quit
+```
+
+### Building
+
+Prerequisites:
+
+1.  OpenGL & OpenGL Utility Toolkit (GLUT)
+
+    Linux: `sudo apt-get install freeglut3-dev mesa-common-dev`
+
+    Mac + Xcode: These libraries should be available in the OpenGL / GLUT
+    frameworks.
+
+    Windows: http://freeglut.sourceforge.net/index.php#download
+
+2.  (Optional) qcms (Quick Color Management System)
+
+    1.  Download qcms from Mozilla / Chromium:
+        https://hg.mozilla.org/mozilla-central/file/0e7639e3bdfb/gfx/qcms
+        https://source.chromium.org/chromium/chromium/src/+/main:third_party/qcms/;drc=d4a2f8e1ed461d8fc05ed88d1ae2dc94c9773825
+    2.  Build and archive the source files as libqcms.a / qcms.lib
+    3.  Update makefile.unix / Makefile.vc
+        1.  Define WEBP_HAVE_QCMS
+        2.  Update include / library paths to reference the qcms directory.
+
+Build using makefile.unix / Makefile.vc:
+
+```shell
+$ make -f makefile.unix examples/vwebp
+> nmake /f Makefile.vc CFG=release-static \
+    ../obj/x64/release-static/bin/vwebp.exe
+```
+
+## Animation creation tool
+
+The utility `img2webp` can turn a sequence of input images (PNG, JPEG, ...) into
+an animated WebP file. It offers fine control over duration, encoding modes,
+etc.
+
+Usage:
+
+```shell
+img2webp [file_options] [[frame_options] frame_file]...
+```
+
+File-level options (only used at the start of compression):
+
+```
+-min_size ............ minimize size
+-loop <int> .......... loop count (default: 0, = infinite loop)
+-kmax <int> .......... maximum number of frame between key-frames
+                        (0=only keyframes)
+-kmin <int> .......... minimum number of frame between key-frames
+                        (0=disable key-frames altogether)
+-mixed ............... use mixed lossy/lossless automatic mode
+-v ................... verbose mode
+-h ................... this help
+-version ............. print version number and exit
+```
+
+Per-frame options (only used for subsequent images input):
+
+```
+-d <int> ............. frame duration in ms (default: 100)
+-lossless  ........... use lossless mode (default)
+-lossy ... ........... use lossy mode
+-q <float> ........... quality
+-m <int> ............. method to use
+```
+
+example: `img2webp -loop 2 in0.png -lossy in1.jpg -d 80 in2.tiff -o out.webp`
+
+Note: if a single file name is passed as the argument, the arguments will be
+tokenized from this file. The file name must not start with the character '-'.
+
+## Animated GIF conversion
+
+Animated GIF files can be converted to WebP files with animation using the
+gif2webp utility available under examples/. The files can then be viewed using
+vwebp.
+
+Usage:
+
+```shell
+gif2webp [options] gif_file -o webp_file
+```
+
+Options:
+
+```
+-h / -help ............. this help
+-lossy ................. encode image using lossy compression
+-mixed ................. for each frame in the image, pick lossy
+                         or lossless compression heuristically
+-q <float> ............. quality factor (0:small..100:big)
+-m <int> ............... compression method (0=fast, 6=slowest)
+-min_size .............. minimize output size (default:off)
+                         lossless compression by default; can be
+                         combined with -q, -m, -lossy or -mixed
+                         options
+-kmin <int> ............ min distance between key frames
+-kmax <int> ............ max distance between key frames
+-f <int> ............... filter strength (0=off..100)
+-metadata <string> ..... comma separated list of metadata to
+                         copy from the input to the output if present
+                         Valid values: all, none, icc, xmp (default)
+-loop_compatibility .... use compatibility mode for Chrome
+                         version prior to M62 (inclusive)
+-mt .................... use multi-threading if available
+
+-version ............... print version number and exit
+-v ..................... verbose
+-quiet ................. don't print anything
+```
+
+### Building
+
+With the libgif development files installed, gif2webp can be built using
+makefile.unix:
+
+```shell
+$ make -f makefile.unix examples/gif2webp
+```
+
+or using autoconf:
+
+```shell
+$ ./configure --enable-everything
+$ make
+```
+
+## Comparison of animated images
+
+Test utility anim_diff under examples/ can be used to compare two animated
+images (each can be GIF or WebP).
+
+Usage:
+
+```shell
+anim_diff <image1> <image2> [options]
+```
+
+Options:
+
+```
+-dump_frames <folder> dump decoded frames in PAM format
+-min_psnr <float> ... minimum per-frame PSNR
+-raw_comparison ..... if this flag is not used, RGB is
+                      premultiplied before comparison
+-max_diff <int> ..... maximum allowed difference per channel
+                      between corresponding pixels in subsequent
+                      frames
+-h .................. this help
+-version ............ print version number and exit
+```
+
+### Building
+
+With the libgif development files installed, anim_diff can be built using
+makefile.unix:
+
+```shell
+$ make -f makefile.unix examples/anim_diff
+```
+
+or using autoconf:
+
+```shell
+$ ./configure --enable-everything
+$ make
+```
+
+## WebP Mux tool
+
+The examples/ directory contains a tool (webpmux) for manipulating WebP files.
+The webpmux tool can be used to create an extended format WebP file and also to
+extract or strip relevant data from such a file.
+
+A list of options is available using the -help command line flag:
+
+```shell
+> webpmux -help
+Usage: webpmux -get GET_OPTIONS INPUT -o OUTPUT
+       webpmux -set SET_OPTIONS INPUT -o OUTPUT
+       webpmux -duration DURATION_OPTIONS [-duration ...]
+               INPUT -o OUTPUT
+       webpmux -strip STRIP_OPTIONS INPUT -o OUTPUT
+       webpmux -frame FRAME_OPTIONS [-frame...] [-loop LOOP_COUNT]
+               [-bgcolor BACKGROUND_COLOR] -o OUTPUT
+       webpmux -info INPUT
+       webpmux [-h|-help]
+       webpmux -version
+       webpmux argument_file_name
+
+GET_OPTIONS:
+ Extract relevant data:
+   icc       get ICC profile
+   exif      get EXIF metadata
+   xmp       get XMP metadata
+   frame n   get nth frame
+
+SET_OPTIONS:
+ Set color profile/metadata/parameters:
+   loop LOOP_COUNT            set the loop count
+   bgcolor BACKGROUND_COLOR   set the animation background color
+   icc  file.icc              set ICC profile
+   exif file.exif             set EXIF metadata
+   xmp  file.xmp              set XMP metadata
+   where:    'file.icc' contains the ICC profile to be set,
+             'file.exif' contains the EXIF metadata to be set
+             'file.xmp' contains the XMP metadata to be set
+
+DURATION_OPTIONS:
+ Set duration of selected frames:
+   duration            set duration for all frames
+   duration,frame      set duration of a particular frame
+   duration,start,end  set duration of frames in the
+                        interval [start,end])
+   where: 'duration' is the duration in milliseconds
+          'start' is the start frame index
+          'end' is the inclusive end frame index
+           The special 'end' value '0' means: last frame.
+
+STRIP_OPTIONS:
+ Strip color profile/metadata:
+   icc       strip ICC profile
+   exif      strip EXIF metadata
+   xmp       strip XMP metadata
+
+FRAME_OPTIONS(i):
+ Create animation:
+   file_i +di[+xi+yi[+mi[bi]]]
+   where:    'file_i' is the i'th animation frame (WebP format),
+             'di' is the pause duration before next frame,
+             'xi','yi' specify the image offset for this frame,
+             'mi' is the dispose method for this frame (0 or 1),
+             'bi' is the blending method for this frame (+b or -b)
+
+LOOP_COUNT:
+ Number of times to repeat the animation.
+ Valid range is 0 to 65535 [Default: 0 (infinite)].
+
+BACKGROUND_COLOR:
+ Background color of the canvas.
+  A,R,G,B
+  where:    'A', 'R', 'G' and 'B' are integers in the range 0 to 255 specifying
+            the Alpha, Red, Green and Blue component values respectively
+            [Default: 255,255,255,255]
+
+INPUT & OUTPUT are in WebP format.
+
+Note: The nature of EXIF, XMP and ICC data is not checked and is assumed to be
+valid.
+
+Note: if a single file name is passed as the argument, the arguments will be
+tokenized from this file. The file name must not start with the character '-'.
+```

doc/webp-container-spec.txt

@@ -4,8 +4,8 @@ Although you may be viewing an alternate representation, this document
 is sourced in Markdown, a light-duty markup scheme, and is optimized for
 the [kramdown](https://kramdown.gettalong.org/) transformer.
 
-See the accompanying README. External link targets are referenced at the
+See the accompanying specs_generation.md. External link targets are referenced
-end of this file.
+at the end of this file.
 
 -->
 
@@ -25,7 +25,7 @@ to compress image data in a lossy way, or (ii) the WebP lossless encoding
 (and possibly other encodings in the future). These encoding schemes should
 make it more efficient than currently used formats. It is optimized for fast
 image transfer over the network (e.g., for websites). The WebP format has
-feature parity (color profile, metadata, animation etc) with other formats as
+feature parity (color profile, metadata, animation, etc.) with other formats as
 well. This document describes the structure of a WebP file.
 
 The WebP container (i.e., RIFF container for WebP) allows feature support over
@@ -36,7 +36,7 @@ for:
   * **Lossless compression.** An image can be losslessly compressed, using the
     WebP Lossless Format.
 
-  * **Metadata.** An image may have metadata stored in EXIF or XMP formats.
+  * **Metadata.** An image may have metadata stored in Exif or XMP formats.
 
   * **Transparency.** An image may have transparency, i.e., an alpha channel.
 
@@ -46,15 +46,16 @@ for:
   * **Animation.** An image may have multiple frames with pauses between them,
     making it an animation.
 
-The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT",
+The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD",
-"SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this
+"SHOULD NOT", "RECOMMENDED", "NOT RECOMMENDED", "MAY", and "OPTIONAL" in this
-document are to be interpreted as described in [RFC 2119][].
+document are to be interpreted as described in BCP 14 [RFC 2119][] [RFC 8174][]
+when, and only when, they appear in all capitals, as shown here.
 
 Bit numbering in chunk diagrams starts at `0` for the most significant bit
 ('MSB 0') as described in [RFC 1166][].
 
-Terminology & Basics
+Terminology & Basics
-------------------------
+--------------------
 
 A WebP file contains either a still image (i.e., an encoded matrix of pixels)
 or an [animation](#animation). Optionally, it can also contain transparency
@@ -83,18 +84,25 @@ _uint32_
 _FourCC_
 
 : A _FourCC_ (four-character code) is a _uint32_ created by concatenating four
-  ASCII characters in little-endian order.
+  ASCII characters in little-endian order. This means 'aaaa' (0x61616161) and
+ 'AAAA' (0x41414141) are treated as different _FourCCs_.
 
 _1-based_
 
 : An unsigned integer field storing values offset by `-1`. e.g., Such a field
   would store value _25_ as _24_.
 
+_ChunkHeader('ABCD')_
+
+: This is used to describe the _FourCC_ and _Chunk Size_ header of individual
+  chunks, where 'ABCD' is the FourCC for the chunk. This element's size is 8
+  bytes.
+
 
 RIFF File Format
 ----------------
 
-The WebP file format is based on the RIFF (resource interchange file format)
+The WebP file format is based on the RIFF (Resource Interchange File Format)
 document format.
 
 The basic element of a RIFF file is a _chunk_. It consists of:
@@ -106,7 +114,7 @@ The basic element of a RIFF file is a _chunk_. It consists of:
     +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
     |                          Chunk Size                           |
     +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
-    |                         Chunk Payload                         |
+    :                         Chunk Payload                         :
     +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
 
 Chunk FourCC: 32 bits
@@ -115,19 +123,13 @@ Chunk FourCC: 32 bits
 
 Chunk Size: 32 bits (_uint32_)
 
-: The size of the chunk not including this field, the chunk identifier or
+: The size of the chunk in bytes, not including this field, the chunk
-  padding.
+  identifier or padding.
 
 Chunk Payload: _Chunk Size_ bytes
 
-: The data payload. If _Chunk Size_ is odd, a single padding byte -- that
+: The data payload. If _Chunk Size_ is odd, a single padding byte -- that MUST
-  SHOULD be `0` -- is added.
+  be `0` to conform with RIFF -- is added.
-
-_ChunkHeader('ABCD')_
-
-: This is used to describe the _FourCC_ and _Chunk Size_ header of individual
-  chunks, where 'ABCD' is the FourCC for the chunk. This element's
-  size is 8 bytes.
 
 **Note:** RIFF has a convention that all-uppercase chunk FourCCs are standard
 chunks that apply to any RIFF file format, while FourCCs specific to a file
@@ -163,9 +165,11 @@ File Size: 32 bits (_uint32_)
 
 A WebP file MUST begin with a RIFF header with the FourCC 'WEBP'. The file size
 in the header is the total size of the chunks that follow plus `4` bytes for
-the 'WEBP' FourCC. The file SHOULD NOT contain anything after it. As the size
+the 'WEBP' FourCC. The file SHOULD NOT contain any data after the data
-of any chunk is even, the size given by the RIFF header is also even. The
+specified by _File Size_. Readers MAY parse such files, ignoring the trailing
-contents of individual chunks will be described in the following sections.
+data. As the size of any chunk is even, the size given by the RIFF header is
+also even. The contents of individual chunks will be described in the following
+sections.
 
 
 Simple File Format (Lossy)
@@ -180,9 +184,11 @@ Simple WebP (lossy) file format:
      0                   1                   2                   3
      0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
     +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
+    |                                                               |
     |                    WebP file header (12 bytes)                |
+    |                                                               |
     +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
-    |                          VP8 chunk                            |
+    :                          VP8 chunk                            :
     +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
 
 VP8 chunk:
@@ -191,20 +197,24 @@ VP8 chunk:
      0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
     +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
     |                      ChunkHeader('VP8 ')                      |
+    |                                                               |
     +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
-    |                           VP8 data                            |
+    :                           VP8 data                            :
     +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
 
 VP8 data: _Chunk Size_ bytes
 
 : VP8 bitstream data.
 
+Note the fourth character in the 'VP8 ' FourCC is an ASCII space (0x20).
+
 The VP8 bitstream format specification can be found at [VP8 Data Format and
 Decoding Guide][vp8spec]. Note that the VP8 frame header contains the VP8 frame
 width and height. That is assumed to be the width and height of the canvas.
 
-The VP8 specification describes how to decode the image into Y'CbCr
+The VP8 specification describes how to decode the image into Y'CbCr format. To
-format. To convert to RGB, Rec. 601 SHOULD be used.
+convert to RGB, Rec. 601 SHOULD be used. Applications MAY use another
+conversion method, but visual results may differ among decoders.
 
 
 Simple File Format (Lossless)
@@ -221,9 +231,11 @@ Simple WebP (lossless) file format:
      0                   1                   2                   3
      0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
     +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
+    |                                                               |
     |                    WebP file header (12 bytes)                |
+    |                                                               |
     +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
-    |                          VP8L chunk                           |
+    :                          VP8L chunk                           :
     +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
 
 VP8L chunk:
@@ -232,8 +244,9 @@ VP8L chunk:
      0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
     +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
     |                      ChunkHeader('VP8L')                      |
+    |                                                               |
     +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
-    |                           VP8L data                           |
+    :                           VP8L data                           :
     +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
 
 VP8L data: _Chunk Size_ bytes
@@ -261,7 +274,7 @@ An extended format file consists of:
 
   * Image data.
 
-  * An optional 'EXIF' chunk with EXIF metadata.
+  * An optional 'EXIF' chunk with Exif metadata.
 
   * An optional 'XMP ' chunk with XMP metadata.
 
@@ -279,7 +292,7 @@ details about frames can be found in the [Animation](#animation) section.
 
 All chunks SHOULD be placed in the same order as listed above. If a chunk
 appears in the wrong place, the file is invalid, but readers MAY parse the
-file, ignoring the chunks that come too late.
+file, ignoring the chunks that are out of order.
 
 **Rationale:** Setting the order of chunks should allow quicker file
 parsing. For example, if an 'ALPH' chunk does not appear in its required
@@ -293,9 +306,12 @@ Extended WebP file header:
      0                   1                   2                   3
      0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
     +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
+    |                                                               |
     |                   WebP file header (12 bytes)                 |
+    |                                                               |
     +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
     |                      ChunkHeader('VP8X')                      |
+    |                                                               |
     +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
     |Rsv|I|L|E|X|A|R|                   Reserved                    |
     +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
@@ -306,7 +322,7 @@ Extended WebP file header:
 
 Reserved (Rsv): 2 bits
 
-: SHOULD be `0`.
+: MUST be `0`. Readers MUST ignore this field.
 
 ICC profile (I): 1 bit
 
@@ -317,9 +333,9 @@ Alpha (L): 1 bit
 : Set if any of the frames of the image contain transparency information
   ("alpha").
 
-EXIF metadata (E): 1 bit
+Exif metadata (E): 1 bit
 
-: Set if the file contains EXIF metadata.
+: Set if the file contains Exif metadata.
 
 XMP metadata (X): 1 bit
 
@@ -332,25 +348,25 @@ Animation (A): 1 bit
 
 Reserved (R): 1 bit
 
-: SHOULD be `0`.
+: MUST be `0`. Readers MUST ignore this field.
 
 Reserved: 24 bits
 
-: SHOULD be `0`.
+: MUST be `0`. Readers MUST ignore this field.
 
 Canvas Width Minus One: 24 bits
 
 : _1-based_ width of the canvas in pixels.
-  The actual canvas width is '1 + Canvas Width Minus One'
+  The actual canvas width is `1 + Canvas Width Minus One`.
 
 Canvas Height Minus One: 24 bits
 
 : _1-based_ height of the canvas in pixels.
-  The actual canvas height is '1 + Canvas Height Minus One'
+  The actual canvas height is `1 + Canvas Height Minus One`.
 
 The product of _Canvas Width_ and _Canvas Height_ MUST be at most `2^32 - 1`.
 
-Future specifications MAY add more fields.
+Future specifications may add more fields. Unknown fields MUST be ignored.
 
 ### Chunks
 
@@ -368,6 +384,7 @@ animation.
      0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
     +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
     |                      ChunkHeader('ANIM')                      |
+    |                                                               |
     +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
     |                       Background Color                        |
     +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
@@ -383,8 +400,8 @@ Background Color: 32 bits (_uint32_)
 
 **Note**:
 
-  * Background color MAY contain a transparency value (alpha), even if the
+  * Background color MAY contain a non-opaque alpha value, even if the _Alpha_
-    _Alpha_ flag in [VP8X chunk](#extended_header) is unset.
+    flag in [VP8X chunk](#extended_header) is unset.
 
   * Viewer applications SHOULD treat the background color value as a hint, and
     are not required to use it.
@@ -397,8 +414,8 @@ Loop Count: 16 bits (_uint16_)
 : The number of times to loop the animation. `0` means infinitely.
 
 This chunk MUST appear if the _Animation_ flag in the VP8X chunk is set.
-If the _Animation_ flag is not set and this chunk is present, it
+If the _Animation_ flag is not set and this chunk is present, it MUST be
-SHOULD be ignored.
+ignored.
 
 ANMF chunk:
 
@@ -409,6 +426,7 @@ If the _Animation flag_ is not set, then this chunk SHOULD NOT be present.
      0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
     +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
     |                      ChunkHeader('ANMF')                      |
+    |                                                               |
     +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
     |                        Frame X                |             ...
     +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
@@ -418,26 +436,26 @@ If the _Animation flag_ is not set, then this chunk SHOULD NOT be present.
     +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
     |                 Frame Duration                |  Reserved |B|D|
     +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
-    |                         Frame Data                            |
+    :                         Frame Data                            :
     +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
 
 Frame X: 24 bits (_uint24_)
 
-: The X coordinate of the upper left corner of the frame is `Frame X * 2`
+: The X coordinate of the upper left corner of the frame is `Frame X * 2`.
 
 Frame Y: 24 bits (_uint24_)
 
-: The Y coordinate of the upper left corner of the frame is `Frame Y * 2`
+: The Y coordinate of the upper left corner of the frame is `Frame Y * 2`.
 
 Frame Width Minus One: 24 bits (_uint24_)
 
 : The _1-based_ width of the frame.
-  The frame width is `1 + Frame Width Minus One`
+  The frame width is `1 + Frame Width Minus One`.
 
 Frame Height Minus One: 24 bits (_uint24_)
 
 : The _1-based_ height of the frame.
-  The frame height is `1 + Frame Height Minus One`
+  The frame height is `1 + Frame Height Minus One`.
 
 Frame Duration: 24 bits (_uint24_)
 
@@ -448,7 +466,7 @@ Frame Duration: 24 bits (_uint24_)
 
 Reserved: 6 bits
 
-: SHOULD be 0.
+: MUST be `0`. Readers MUST ignore this field.
 
 Blending method (B): 1 bit
 
@@ -493,8 +511,9 @@ Disposal method (D): 1 bit
     if blend.A = 0 then
       blend.RGB = 0
     else
-      blend.RGB = (src.RGB * src.A +
+      blend.RGB =
-                   dst.RGB * dst.A * (1 - src.A / 255)) / blend.A
+          (src.RGB * src.A +
+           dst.RGB * dst.A * (1 - src.A / 255)) / blend.A
 ~~~~~
 
   * Alpha-blending SHOULD be done in linear color space, by taking into account
@@ -521,22 +540,23 @@ _padded_ chunks as described by the [RIFF file format](#riff-file-format).
      0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
     +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
     |                      ChunkHeader('ALPH')                      |
+    |                                                               |
     +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
     |Rsv| P | F | C |     Alpha Bitstream...                        |
     +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
 
 Reserved (Rsv): 2 bits
 
-: SHOULD be `0`.
+: MUST be `0`. Readers MUST ignore this field.
 
 Pre-processing (P): 2 bits
 
-: These INFORMATIVE bits are used to signal the pre-processing that has
+: These _informative_ bits are used to signal the pre-processing that has
   been performed during compression. The decoder can use this information to
   e.g. dither the values or smooth the gradients prior to display.
 
-    * `0`: no pre-processing
+    * `0`: No pre-processing.
-    * `1`: level reduction
+    * `1`: Level reduction.
 
 Filtering method (F): 2 bits
 
@@ -569,14 +589,14 @@ where `clip(v)` is equal to:
   * v    otherwise
 
 The final value is derived by adding the decompressed value `X` to the
-predictor and using modulo-256 arithmetic to wrap the \[256-511\] range
+predictor and using modulo-256 arithmetic to wrap the \[256..511\] range
-into the \[0-255\] one:
+into the \[0..255\] one:
 
 `alpha = (predictor + X) % 256`
 
-There are special cases for left-most and top-most pixel positions:
+There are special cases for the left-most and top-most pixel positions:
 
-  * Top-left value at location (0,0) uses 0 as predictor value. Otherwise,
+  * The top-left value at location (0, 0) uses 0 as predictor value. Otherwise,
   * For horizontal or gradient filtering methods, the left-most pixels at
     location (0, y) are predicted using the location (0, y-1) just above.
   * For vertical or gradient filtering methods, the top-most pixels at
@@ -638,14 +658,15 @@ The formats of VP8 and VP8L chunks are as described in sections
 [Simple File Format (Lossy)](#simple-file-format-lossy)
 and [Simple File Format (Lossless)](#simple-file-format-lossless) respectively.
 
-#### Color profile
+#### Color Profile
 
      0                   1                   2                   3
      0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
     +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
     |                      ChunkHeader('ICCP')                      |
+    |                                                               |
     +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
-    |                       Color Profile                           |
+    :                       Color Profile                           :
     +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
 
 Color Profile: _Chunk Size_ bytes
@@ -665,8 +686,7 @@ If this chunk is not present, sRGB SHOULD be assumed.
 Metadata can be stored in 'EXIF' or 'XMP ' chunks.
 
 There SHOULD be at most one chunk of each type ('EXIF' and 'XMP '). If there
-are more such chunks, readers MAY ignore all except the first one. Also, a file
+are more such chunks, readers MAY ignore all except the first one.
-may possibly contain both 'EXIF' and 'XMP ' chunks.
 
 The chunks are defined as follows:
 
@@ -676,13 +696,14 @@ EXIF chunk:
      0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
     +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
     |                      ChunkHeader('EXIF')                      |
+    |                                                               |
     +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
-    |                        EXIF Metadata                          |
+    :                        Exif Metadata                          :
     +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
 
-EXIF Metadata: _Chunk Size_ bytes
+Exif Metadata: _Chunk Size_ bytes
 
-: image metadata in EXIF format.
+: Image metadata in Exif format.
 
 XMP chunk:
 
@@ -690,13 +711,16 @@ XMP chunk:
      0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
     +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
     |                      ChunkHeader('XMP ')                      |
+    |                                                               |
     +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
-    |                        XMP Metadata                           |
+    :                        XMP Metadata                           :
     +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
 
 XMP Metadata: _Chunk Size_ bytes
 
-: image metadata in XMP format.
+: Image metadata in XMP format.
+
+Note the fourth character in the 'XMP ' FourCC is an ASCII space (0x20).
 
 Additional guidance about handling metadata can be found in the
 Metadata Working Group's [Guidelines for Handling Metadata][metadata].
@@ -720,49 +744,73 @@ A file MAY contain unknown chunks:
 Readers SHOULD ignore these chunks. Writers SHOULD preserve them in their
 original order (unless they specifically intend to modify these chunks).
 
-### Assembling the Canvas from frames
+### Assembling the Canvas From Frames
 
-Here we provide an overview of how a reader should assemble a canvas in the
+Here we provide an overview of how a reader MUST assemble a canvas in the case
-case of an animated image. The notation _VP8X.field_ means the field in the
+of an animated image.
-'VP8X' chunk with the same description.
 
-Displaying an _animated image_ canvas MUST be equivalent to the following
+The process begins with creating a canvas using the dimensions given in the
-pseudocode:
+'VP8X' chunk, `Canvas Width Minus One + 1` pixels wide by `Canvas Height Minus
+One + 1` pixels high. The `Loop Count` field from the 'ANIM' chunk controls how
+many times the animation process is repeated. This is `Loop Count - 1` for
+non-zero `Loop Count` values or infinitely if `Loop Count` is zero.
+
+At the beginning of each loop iteration the canvas is filled using the
+background color from the 'ANIM' chunk or an application defined color.
+
+'ANMF' chunks contain individual frames given in display order. Before rendering
+each frame, the previous frame's `Disposal method` is applied.
+
+The rendering of the decoded frame begins at the Cartesian coordinates (`2 *
+Frame X`, `2 * Frame Y`) using the top-left corner of the canvas as the origin.
+`Frame Width Minus One + 1` pixels wide by `Frame Height Minus One + 1` pixels
+high are rendered onto the canvas using the `Blending method`.
+
+The canvas is displayed for `Frame Duration` milliseconds. This continues until
+all frames given by 'ANMF' chunks have been displayed. A new loop iteration is
+then begun or the canvas is left in its final state if all iterations have been
+completed.
+
+The following pseudocode illustrates the rendering process. The notation
+_VP8X.field_ means the field in the 'VP8X' chunk with the same description.
 
 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 assert VP8X.flags.hasAnimation
 canvas ← new image of size VP8X.canvasWidth x VP8X.canvasHeight with
          background color ANIM.background_color.
 loop_count ← ANIM.loopCount
-dispose_method ← ANIM.disposeMethod
+dispose_method ← Dispose to background color
 if loop_count == 0:
-    loop_count = ∞
+  loop_count = ∞
 frame_params ← nil
 assert next chunk in image_data is ANMF
 for loop = 0..loop_count - 1
-    clear canvas to ANIM.background_color or application defined color
+  clear canvas to ANIM.background_color or application defined color
-    until eof or non-ANMF chunk
+  until eof or non-ANMF chunk
-        frame_params.frameX = Frame X
+    frame_params.frameX = Frame X
-        frame_params.frameY = Frame Y
+    frame_params.frameY = Frame Y
-        frame_params.frameWidth = Frame Width Minus One + 1
+    frame_params.frameWidth = Frame Width Minus One + 1
-        frame_params.frameHeight = Frame Height Minus One + 1
+    frame_params.frameHeight = Frame Height Minus One + 1
-        frame_params.frameDuration = Frame Duration
+    frame_params.frameDuration = Frame Duration
-        frame_right = frame_params.frameX + frame_params.frameWidth
+    frame_right = frame_params.frameX + frame_params.frameWidth
-        frame_bottom = frame_params.frameY + frame_params.frameHeight
+    frame_bottom = frame_params.frameY + frame_params.frameHeight
-        assert VP8X.canvasWidth >= frame_right
+    assert VP8X.canvasWidth >= frame_right
-        assert VP8X.canvasHeight >= frame_bottom
+    assert VP8X.canvasHeight >= frame_bottom
-        for subchunk in 'Frame Data':
+    for subchunk in 'Frame Data':
-            if subchunk.tag == "ALPH":
+      if subchunk.tag == "ALPH":
-                assert alpha subchunks not found in 'Frame Data' earlier
+        assert alpha subchunks not found in 'Frame Data' earlier
-                frame_params.alpha = alpha_data
+        frame_params.alpha = alpha_data
-            else if subchunk.tag == "VP8 " OR subchunk.tag == "VP8L":
+      else if subchunk.tag == "VP8 " OR subchunk.tag == "VP8L":
-                assert bitstream subchunks not found in 'Frame Data' earlier
+        assert bitstream subchunks not found in 'Frame Data' earlier
-                frame_params.bitstream = bitstream_data
+        frame_params.bitstream = bitstream_data
-        render frame with frame_params.alpha and frame_params.bitstream on
+    render frame with frame_params.alpha and frame_params.bitstream
-            canvas with top-left corner at (frame_params.frameX,
+      on canvas with top-left corner at (frame_params.frameX,
-            frame_params.frameY), using dispose method dispose_method.
+      frame_params.frameY), using blending method
-        canvas contains the decoded image.
+      frame_params.blendingMethod.
-        Show the contents of the canvas for frame_params.frameDuration * 1ms.
+    canvas contains the decoded image.
+    Show the contents of the canvas for
+    frame_params.frameDuration * 1ms.
+    dispose_method = frame_params.disposeMethod
 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 
 
@@ -798,7 +846,7 @@ RIFF/WEBP
 +- XMP  (metadata)
 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 
-An animated image with EXIF metadata may look as follows:
+An animated image with Exif metadata may look as follows:
 
 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 RIFF/WEBP
@@ -817,3 +865,4 @@ RIFF/WEBP
 [metadata]: https://web.archive.org/web/20180919181934/http://www.metadataworkinggroup.org/pdf/mwg_guidance.pdf
 [rfc 1166]: https://datatracker.ietf.org/doc/html/rfc1166
 [rfc 2119]: https://datatracker.ietf.org/doc/html/rfc2119
+[rfc 8174]: https://datatracker.ietf.org/doc/html/rfc8174

examples/Android.mk

@@ -1,3 +1,5 @@
+# Ignore this file during non-NDK builds.
+ifdef NDK_ROOT
 LOCAL_PATH := $(call my-dir)
 
 ################################################################################
@@ -91,3 +93,4 @@ LOCAL_STATIC_LIBRARIES := example_util imageio_util webp
 LOCAL_MODULE := webpinfo_example
 
 include $(BUILD_EXECUTABLE)
+endif  # NDK_ROOT

examples/Makefile.am

@@ -47,7 +47,7 @@ anim_dump_LDADD += ../imageio/libimageenc.la
 anim_dump_LDADD += $(PNG_LIBS) $(GIF_LIBS) $(TIFF_LIBS) -lm
 
 cwebp_SOURCES  = cwebp.c stopwatch.h
-cwebp_CPPFLAGS  = $(AM_CPPFLAGS)
+cwebp_CPPFLAGS  = $(AM_CPPFLAGS) -I$(top_srcdir)
 cwebp_LDADD  =
 cwebp_LDADD += libexample_util.la
 cwebp_LDADD += ../imageio/libimageio_util.la

examples/anim_util.c

@@ -241,7 +241,7 @@ static int ReadAnimatedWebP(const char filename[],
   image->bgcolor = anim_info.bgcolor;
 
   // Allocate frames.
-  if (!AllocateFrames(image, anim_info.frame_count)) return 0;
+  if (!AllocateFrames(image, anim_info.frame_count)) goto End;
 
   // Decode frames.
   while (WebPAnimDecoderHasMoreFrames(dec)) {
@@ -558,7 +558,10 @@ static int ReadAnimatedGIF(const char filename[], AnimatedImage* const image,
     }
   }
   // Allocate frames.
-  AllocateFrames(image, frame_count);
+  if (!AllocateFrames(image, frame_count)) {
+    DGifCloseFile(gif, NULL);
+    return 0;
+  }
 
   canvas_width = image->canvas_width;
   canvas_height = image->canvas_height;

examples/cwebp.c

@@ -27,6 +27,7 @@
 #include "../imageio/webpdec.h"
 #include "./stopwatch.h"
 #include "./unicode.h"
+#include "sharpyuv/sharpyuv.h"
 #include "webp/encode.h"
 
 #ifndef WEBP_DLL
@@ -571,7 +572,7 @@ static void HelpLong(void) {
   printf("  -qrange <min> <max> .... specifies the permissible quality range\n"
          "                           (default: 0 100)\n");
   printf("  -crop <x> <y> <w> <h> .. crop picture with the given rectangle\n");
-  printf("  -resize <w> <h> ........ resize picture (after any cropping)\n");
+  printf("  -resize <w> <h> ........ resize picture (*after* any cropping)\n");
   printf("  -mt .................... use multi-threading if available\n");
   printf("  -low_memory ............ reduce memory usage (slower encoding)\n");
   printf("  -map <int> ............. print map of extra info\n");
@@ -620,6 +621,7 @@ static void HelpLong(void) {
   printf("  -af .................... auto-adjust filter strength\n");
   printf("  -pre <int> ............. pre-processing filter\n");
   printf("\n");
+  printf("Supported input formats:\n  %s\n", WebPGetEnabledInputFileFormats());
 }
 
 //------------------------------------------------------------------------------
@@ -831,8 +833,12 @@ int main(int argc, const char* argv[]) {
 #endif
     } else if (!strcmp(argv[c], "-version")) {
       const int version = WebPGetEncoderVersion();
+      const int sharpyuv_version = SharpYuvGetVersion();
       printf("%d.%d.%d\n",
              (version >> 16) & 0xff, (version >> 8) & 0xff, version & 0xff);
+      printf("libsharpyuv: %d.%d.%d\n",
+             (sharpyuv_version >> 24) & 0xff, (sharpyuv_version >> 16) & 0xffff,
+             sharpyuv_version & 0xff);
       FREE_WARGV_AND_RETURN(0);
     } else if (!strcmp(argv[c], "-progress")) {
       show_progress = 1;
@@ -1139,9 +1145,10 @@ int main(int argc, const char* argv[]) {
       }
 
       picture.use_argb = 1;
-      if (!ReadWebP(memory_writer.mem, memory_writer.size, &picture,
+      if (!ReadWebP(
-                    /*keep_alpha=*/WebPPictureHasTransparency(&picture),
+              memory_writer.mem, memory_writer.size, &picture,
-                    /*metadata=*/NULL)) {
+              /*keep_alpha=*/WebPPictureHasTransparency(&original_picture),
+              /*metadata=*/NULL)) {
         fprintf(stderr, "Error! Cannot decode encoded WebP bitstream\n");
         fprintf(stderr, "Error code: %d (%s)\n", picture.error_code,
                 kErrorMessages[picture.error_code]);

examples/dwebp.c

@@ -96,7 +96,7 @@ static void Help(void) {
          "  -alpha_dither  use alpha-plane dithering if needed\n"
          "  -mt .......... use multi-threading\n"
          "  -crop <x> <y> <w> <h> ... crop output with the given rectangle\n"
-         "  -resize <w> <h> ......... scale the output (*after* any cropping)\n"
+         "  -resize <w> <h> ......... resize output (*after* any cropping)\n"
          "  -flip ........ flip the output vertically\n"
          "  -alpha ....... only save the alpha plane\n"
          "  -incremental . use incremental decoding (useful for tests)\n"

examples/example_util.c

@@ -103,7 +103,10 @@ int ExUtilInitCommandLineArguments(int argc, const char* argv[],
     }
     args->own_argv_ = 1;
     args->argv_ = (const char**)WebPMalloc(MAX_ARGC * sizeof(*args->argv_));
-    if (args->argv_ == NULL) return 0;
+    if (args->argv_ == NULL) {
+      ExUtilDeleteCommandLineArguments(args);
+      return 0;
+    }
 
     argc = 0;
     for (cur = strtok((char*)args->argv_data_.bytes, sep);
@@ -111,6 +114,7 @@ int ExUtilInitCommandLineArguments(int argc, const char* argv[],
          cur = strtok(NULL, sep)) {
       if (argc == MAX_ARGC) {
         fprintf(stderr, "ERROR: Arguments limit %d reached\n", MAX_ARGC);
+        ExUtilDeleteCommandLineArguments(args);
         return 0;
       }
       assert(strlen(cur) != 0);

examples/gif2webp.c

@@ -314,8 +314,11 @@ int main(int argc, const char* argv[]) {
           frame.use_argb = 1;
           if (!WebPPictureAlloc(&frame)) goto End;
           GIFClearPic(&frame, NULL);
-          WebPPictureCopy(&frame, &curr_canvas);
+          if (!(WebPPictureCopy(&frame, &curr_canvas) &&
-          WebPPictureCopy(&frame, &prev_canvas);
+                WebPPictureCopy(&frame, &prev_canvas))) {
+            fprintf(stderr, "Error allocating canvas.\n");
+            goto End;
+          }
 
           // Background color.
           GIFGetBackgroundColor(gif->SColorMap, gif->SBackGroundColor,

examples/img2webp.c

@@ -35,8 +35,7 @@
 
 static void Help(void) {
   printf("Usage:\n\n");
-  printf("  img2webp [file-level options] [image files...] "
+  printf("  img2webp [file_options] [[frame_options] frame_file]...\n");
-         "[per-frame options...]\n");
   printf("\n");
 
   printf("File-level options (only used at the start of compression):\n");
@@ -66,6 +65,8 @@ static void Help(void) {
          "arguments will be\n");
   printf("tokenized from this file. The file name must not start with "
          "the character '-'.\n");
+  printf("\nSupported input formats:\n  %s\n",
+         WebPGetEnabledInputFileFormats());
 }
 
 //------------------------------------------------------------------------------
@@ -187,7 +188,7 @@ int main(int argc, const char* argv[]) {
         verbose = 1;
       } else if (!strcmp(argv[c], "-h") || !strcmp(argv[c], "-help")) {
         Help();
-        goto End;
+        FREE_WARGV_AND_RETURN(0);
       } else if (!strcmp(argv[c], "-version")) {
         const int enc_version = WebPGetEncoderVersion();
         const int mux_version = WebPGetMuxVersion();

examples/unicode.h

@@ -16,11 +16,15 @@
 #ifndef WEBP_EXAMPLES_UNICODE_H_
 #define WEBP_EXAMPLES_UNICODE_H_
 
+#include <stdio.h>
+
 #if defined(_WIN32) && defined(_UNICODE)
 
 // wchar_t is used instead of TCHAR because we only perform additional work when
 // Unicode is enabled and because the output of CommandLineToArgvW() is wchar_t.
 
+#include <fcntl.h>
+#include <io.h>
 #include <wchar.h>
 #include <windows.h>
 #include <shellapi.h>
@@ -55,8 +59,16 @@
 
 #define WFOPEN(ARG, OPT) _wfopen((const W_CHAR*)ARG, TO_W_CHAR(OPT))
 
-#define WPRINTF(STR, ...) wprintf(TO_W_CHAR(STR), __VA_ARGS__)
+#define WFPRINTF(STREAM, STR, ...)                    \
-#define WFPRINTF(STDERR, STR, ...) fwprintf(STDERR, TO_W_CHAR(STR), __VA_ARGS__)
+  do {                                                \
+    int prev_mode;                                    \
+    fflush(STREAM);                                   \
+    prev_mode = _setmode(_fileno(STREAM), _O_U8TEXT); \
+    fwprintf(STREAM, TO_W_CHAR(STR), __VA_ARGS__);    \
+    fflush(STREAM);                                   \
+    (void)_setmode(_fileno(STREAM), prev_mode);       \
+  } while (0)
+#define WPRINTF(STR, ...) WFPRINTF(stdout, STR, __VA_ARGS__)
 
 #define WSTRLEN(FILENAME) wcslen((const W_CHAR*)FILENAME)
 #define WSTRCMP(FILENAME, STR) wcscmp((const W_CHAR*)FILENAME, TO_W_CHAR(STR))
@@ -65,6 +77,8 @@
 
 #else
 
+#include <string.h>
+
 // Unicode file paths work as is on Unix platforms, and no extra work is done on
 // Windows either if Unicode is disabled.
 
@@ -83,7 +97,7 @@
 #define WFOPEN(ARG, OPT) fopen(ARG, OPT)
 
 #define WPRINTF(STR, ...) printf(STR, __VA_ARGS__)
-#define WFPRINTF(STDERR, STR, ...) fprintf(STDERR, STR, __VA_ARGS__)
+#define WFPRINTF(STREAM, STR, ...) fprintf(STREAM, STR, __VA_ARGS__)
 
 #define WSTRLEN(FILENAME) strlen(FILENAME)
 #define WSTRCMP(FILENAME, STR) strcmp(FILENAME, STR)

examples/unicode_gif.h

@@ -45,18 +45,19 @@ static GifFileType* DGifOpenFileUnicode(const W_CHAR* file_name, int* error) {
   }
 
 #if defined(_WIN32) && defined(_UNICODE)
-
+  {
-  int file_handle = _wopen(file_name, _O_RDONLY | _O_BINARY);
+    int file_handle = _wopen(file_name, _O_RDONLY | _O_BINARY);
-  if (file_handle == -1) {
+    if (file_handle == -1) {
-    if (error != NULL) *error = D_GIF_ERR_OPEN_FAILED;
+      if (error != NULL) *error = D_GIF_ERR_OPEN_FAILED;
-    return NULL;
+      return NULL;
-  }
+    }
 
 #if LOCAL_GIF_PREREQ(5, 0)
-  return DGifOpenFileHandle(file_handle, error);
+    return DGifOpenFileHandle(file_handle, error);
 #else
-  return DGifOpenFileHandle(file_handle);
+    return DGifOpenFileHandle(file_handle);
 #endif
+  }
 
 #else
 

examples/vwebp.c

@@ -292,6 +292,19 @@ static void PrintString(const char* const text) {
   }
 }
 
+static void PrintStringW(const char* const text) {
+#if defined(_WIN32) && defined(_UNICODE)
+  void* const font = GLUT_BITMAP_9_BY_15;
+  const W_CHAR* const wtext = (const W_CHAR*)text;
+  int i;
+  for (i = 0; wtext[i]; ++i) {
+    glutBitmapCharacter(font, wtext[i]);
+  }
+#else
+  PrintString(text);
+#endif
+}
+
 static float GetColorf(uint32_t color, int shift) {
   return ((color >> shift) & 0xff) / 255.f;
 }
@@ -396,7 +409,7 @@ static void HandleDisplay(void) {
 
     glColor4f(0.90f, 0.0f, 0.90f, 1.0f);
     glRasterPos2f(-0.95f, 0.90f);
-    PrintString(kParams.file_name);
+    PrintStringW(kParams.file_name);
 
     snprintf(tmp, sizeof(tmp), "Dimension:%d x %d", pic->width, pic->height);
     glColor4f(0.90f, 0.0f, 0.90f, 1.0f);

examples/webpinfo.c

@@ -125,16 +125,16 @@ static void WebPInfoInit(WebPInfo* const webp_info) {
   memset(webp_info, 0, sizeof(*webp_info));
 }
 
-static const char kWebPChunkTags[CHUNK_TYPES][4] = {
+static const uint32_t kWebPChunkTags[CHUNK_TYPES] = {
-  { 'V', 'P', '8', ' ' },
+  MKFOURCC('V', 'P', '8', ' '),
-  { 'V', 'P', '8', 'L' },
+  MKFOURCC('V', 'P', '8', 'L'),
-  { 'V', 'P', '8', 'X' },
+  MKFOURCC('V', 'P', '8', 'X'),
-  { 'A', 'L', 'P', 'H' },
+  MKFOURCC('A', 'L', 'P', 'H'),
-  { 'A', 'N', 'I', 'M' },
+  MKFOURCC('A', 'N', 'I', 'M'),
-  { 'A', 'N', 'M', 'F' },
+  MKFOURCC('A', 'N', 'M', 'F'),
-  { 'I', 'C', 'C', 'P' },
+  MKFOURCC('I', 'C', 'C', 'P'),
-  { 'E', 'X', 'I', 'F' },
+  MKFOURCC('E', 'X', 'I', 'F'),
-  { 'X', 'M', 'P', ' ' },
+  MKFOURCC('X', 'M', 'P', ' '),
 };
 
 // -----------------------------------------------------------------------------
@@ -644,7 +644,7 @@ static WebPInfoStatus ParseChunk(const WebPInfo* const webp_info,
       return WEBP_INFO_TRUNCATED_DATA;
     }
     for (i = 0; i < CHUNK_TYPES; ++i) {
-      if (!memcmp(kWebPChunkTags[i], &fourcc, TAG_SIZE)) break;
+      if (kWebPChunkTags[i] == fourcc) break;
     }
     chunk_data->offset_ = chunk_start_offset;
     chunk_data->size_ = chunk_size;
@@ -939,7 +939,13 @@ static WebPInfoStatus ProcessChunk(const ChunkData* const chunk_data,
     LOG_WARN(error_message);
   } else {
     if (!webp_info->quiet_) {
-      const char* tag = kWebPChunkTags[chunk_data->id_];
+      char tag[4];
+      uint32_t fourcc = kWebPChunkTags[chunk_data->id_];
+#ifdef WORDS_BIGENDIAN
+      fourcc = (fourcc >> 24) | ((fourcc >> 8) & 0xff00) |
+               ((fourcc << 8) & 0xff0000) | (fourcc << 24);
+#endif
+      memcpy(tag, &fourcc, sizeof(tag));
       printf("Chunk %c%c%c%c at offset %6d, length %6d\n",
              tag[0], tag[1], tag[2], tag[3], (int)chunk_data->offset_,
              (int)chunk_data->size_);

examples/webpmux.c

@@ -329,7 +329,7 @@ static void PrintHelp(void) {
   printf("\n");
   printf("DURATION_OPTIONS:\n");
   printf(" Set duration of selected frames:\n");
-  printf("   duration            set duration for each frames\n");
+  printf("   duration            set duration for all frames\n");
   printf("   duration,frame      set duration of a particular frame\n");
   printf("   duration,start,end  set duration of frames in the\n");
   printf("                        interval [start,end])\n");
@@ -348,7 +348,7 @@ static void PrintHelp(void) {
   printf("\n");
   printf("FRAME_OPTIONS(i):\n");
   printf(" Create animation:\n");
-  printf("   file_i +di+[xi+yi[+mi[bi]]]\n");
+  printf("   file_i +di[+xi+yi[+mi[bi]]]\n");
   printf("   where:    'file_i' is the i'th animation frame (WebP format),\n");
   printf("             'di' is the pause duration before next frame,\n");
   printf("             'xi','yi' specify the image offset for this frame,\n");
@@ -460,7 +460,8 @@ static WebPMux* DuplicateMuxHeader(const WebPMux* const mux) {
     if (err == WEBP_MUX_OK && metadata.size > 0) {
       err = WebPMuxSetChunk(new_mux, kFourccList[i], &metadata, 1);
       if (err != WEBP_MUX_OK) {
-        ERROR_GOTO1("Error transferring metadata in DuplicateMux().", End);
+        ERROR_GOTO1("Error transferring metadata in DuplicateMuxHeader().",
+                    End);
       }
     }
   }
@@ -684,7 +685,7 @@ static int ParseCommandLine(Config* config, const W_CHAR** const unicode_argv) {
           ERROR_GOTO1("ERROR: Multiple features specified.\n", ErrParse);
         }
         arg->subtype_ = SUBTYPE_ANMF;
-        arg->filename_ = argv[i + 1];
+        arg->filename_ = wargv[i + 1];
         arg->params_ = argv[i + 2];
         ++feature_arg_index;
         i += 3;

extras/extras.c

@@ -18,8 +18,8 @@
 #include <string.h>
 
 #define XTRA_MAJ_VERSION 1
-#define XTRA_MIN_VERSION 2
+#define XTRA_MIN_VERSION 3
-#define XTRA_REV_VERSION 2
+#define XTRA_REV_VERSION 0
 
 //------------------------------------------------------------------------------
 

extras/get_disto.c

@@ -223,7 +223,8 @@ static void Help(void) {
           "  -o <file> . save the diff map as a WebP lossless file\n"
           "  -scale .... scale the difference map to fit [0..255] range\n"
           "  -gray ..... use grayscale for difference map (-scale)\n"
-          " Also handles PNG, JPG and TIFF files, in addition to WebP.\n");
+          "\nSupported input formats:\n  %s\n",
+          WebPGetEnabledInputFileFormats());
 }
 
 int main(int argc, const char* argv[]) {

extras/vwebp_sdl.c

@@ -75,7 +75,7 @@ int main(int argc, char* argv[]) {
       fprintf(stderr, "File too large.\n");
       goto Error;
     }
-    ok = WebpToSDL((const char*)webp, (int)webp_size);
+    ok = WebPToSDL((const char*)webp, (int)webp_size);
     free((void*)webp);
     if (!ok) {
       WFPRINTF(stderr, "Error decoding file %s\n", (const W_CHAR*)file);

extras/webp_to_sdl.c

@@ -29,7 +29,7 @@
 #endif
 
 static int init_ok = 0;
-int WebpToSDL(const char* data, unsigned int data_size) {
+int WebPToSDL(const char* data, unsigned int data_size) {
   int ok = 0;
   VP8StatusCode status;
   WebPDecoderConfig config;

extras/webp_to_sdl.h

@@ -14,9 +14,9 @@
 #ifndef WEBP_EXTRAS_WEBP_TO_SDL_H_
 #define WEBP_EXTRAS_WEBP_TO_SDL_H_
 
-// Exports the method WebpToSDL(const char* data, int data_size) which decodes
+// Exports the method WebPToSDL(const char* data, int data_size) which decodes
 // a WebP bitstream into an RGBA SDL surface.
 // Return false on failure.
-extern int WebpToSDL(const char* data, unsigned int data_size);
+extern int WebPToSDL(const char* data, unsigned int data_size);
 
 #endif  // WEBP_EXTRAS_WEBP_TO_SDL_H_

imageio/Android.mk

@@ -1,3 +1,5 @@
+# Ignore this file during non-NDK builds.
+ifdef NDK_ROOT
 LOCAL_PATH := $(call my-dir)
 
 ################################################################################
@@ -52,3 +54,4 @@ LOCAL_STATIC_LIBRARIES := imageio_util
 LOCAL_MODULE := imageenc
 
 include $(BUILD_STATIC_LIBRARY)
+endif  # NDK_ROOT

imageio/image_dec.c

@@ -11,6 +11,24 @@
 
 #include "./image_dec.h"
 
+const char* WebPGetEnabledInputFileFormats(void) {
+  return "WebP"
+#ifdef WEBP_HAVE_JPEG
+         ", JPEG"
+#endif
+#ifdef WEBP_HAVE_PNG
+         ", PNG"
+#endif
+         ", PNM (PGM, PPM, PAM)"
+#ifdef WEBP_HAVE_TIFF
+         ", TIFF"
+#endif
+#ifdef HAVE_WINCODEC_H
+         ", Windows Imaging Component (WIC)"
+#endif
+         "";
+}
+
 static WEBP_INLINE uint32_t GetBE32(const uint8_t buf[]) {
   return ((uint32_t)buf[0] << 24) | (buf[1] << 16) | (buf[2] << 8) | buf[3];
 }

imageio/image_dec.h

@@ -41,6 +41,9 @@ typedef enum {
   WEBP_UNSUPPORTED_FORMAT
 } WebPInputFileFormat;
 
+// Returns a comma separated list of enabled input formats.
+const char* WebPGetEnabledInputFileFormats(void);
+
 // Try to infer the image format. 'data_size' should be larger than 12.
 // Returns WEBP_UNSUPPORTED_FORMAT if format can't be guess safely.
 WebPInputFileFormat WebPGuessImageType(const uint8_t* const data,

imageio/image_enc.c

@@ -280,7 +280,7 @@ int WebPWrite16bAsPGM(FILE* fout, const WebPDecBuffer* const buffer) {
 }
 
 //------------------------------------------------------------------------------
-// BMP
+// BMP (see https://en.wikipedia.org/wiki/BMP_file_format#Pixel_storage)
 
 static void PutLE16(uint8_t* const dst, uint32_t value) {
   dst[0] = (value >> 0) & 0xff;
@@ -293,8 +293,11 @@ static void PutLE32(uint8_t* const dst, uint32_t value) {
 }
 
 #define BMP_HEADER_SIZE 54
+#define BMP_HEADER_ALPHA_EXTRA_SIZE 16  // for alpha info
 int WebPWriteBMP(FILE* fout, const WebPDecBuffer* const buffer) {
   const int has_alpha = WebPIsAlphaMode(buffer->colorspace);
+  const int header_size =
+      BMP_HEADER_SIZE + (has_alpha ? BMP_HEADER_ALPHA_EXTRA_SIZE : 0);
   const uint32_t width = buffer->width;
   const uint32_t height = buffer->height;
   const uint8_t* rgba = buffer->u.RGBA.rgba;
@@ -303,8 +306,9 @@ int WebPWriteBMP(FILE* fout, const WebPDecBuffer* const buffer) {
   uint32_t y;
   const uint32_t line_size = bytes_per_px * width;
   const uint32_t bmp_stride = (line_size + 3) & ~3;   // pad to 4
-  const uint32_t total_size = bmp_stride * height + BMP_HEADER_SIZE;
+  const uint32_t image_size = bmp_stride * height;
-  uint8_t bmp_header[BMP_HEADER_SIZE] = { 0 };
+  const uint32_t total_size =  image_size + header_size;
+  uint8_t bmp_header[BMP_HEADER_SIZE + BMP_HEADER_ALPHA_EXTRA_SIZE] = { 0 };
 
   if (fout == NULL || buffer == NULL || rgba == NULL) return 0;
 
@@ -312,30 +316,37 @@ int WebPWriteBMP(FILE* fout, const WebPDecBuffer* const buffer) {
   PutLE16(bmp_header + 0, 0x4d42);                // signature 'BM'
   PutLE32(bmp_header + 2, total_size);            // size including header
   PutLE32(bmp_header + 6, 0);                     // reserved
-  PutLE32(bmp_header + 10, BMP_HEADER_SIZE);      // offset to pixel array
+  PutLE32(bmp_header + 10, header_size);          // offset to pixel array
   // bitmap info header
-  PutLE32(bmp_header + 14, 40);                   // DIB header size
+  PutLE32(bmp_header + 14, header_size - 14);     // DIB header size
   PutLE32(bmp_header + 18, width);                // dimensions
-  PutLE32(bmp_header + 22, -(int)height);         // vertical flip!
+  PutLE32(bmp_header + 22, height);               // no vertical flip
   PutLE16(bmp_header + 26, 1);                    // number of planes
   PutLE16(bmp_header + 28, bytes_per_px * 8);     // bits per pixel
-  PutLE32(bmp_header + 30, 0);                    // no compression (BI_RGB)
+  PutLE32(bmp_header + 30, has_alpha ? 3 : 0);    // BI_BITFIELDS or BI_RGB
-  PutLE32(bmp_header + 34, 0);                    // image size (placeholder)
+  PutLE32(bmp_header + 34, image_size);
   PutLE32(bmp_header + 38, 2400);                 // x pixels/meter
   PutLE32(bmp_header + 42, 2400);                 // y pixels/meter
   PutLE32(bmp_header + 46, 0);                    // number of palette colors
   PutLE32(bmp_header + 50, 0);                    // important color count
+  if (has_alpha) {  // BITMAPV3INFOHEADER complement
+    PutLE32(bmp_header + 54, 0x00ff0000);         // red mask
+    PutLE32(bmp_header + 58, 0x0000ff00);         // green mask
+    PutLE32(bmp_header + 62, 0x000000ff);         // blue mask
+    PutLE32(bmp_header + 66, 0xff000000);         // alpha mask
+  }
 
   // TODO(skal): color profile
 
   // write header
-  if (fwrite(bmp_header, sizeof(bmp_header), 1, fout) != 1) {
+  if (fwrite(bmp_header, header_size, 1, fout) != 1) {
     return 0;
   }
 
-  // write pixel array
+  // write pixel array, bottom to top
   for (y = 0; y < height; ++y) {
-    if (fwrite(rgba, line_size, 1, fout) != 1) {
+    const uint8_t* const src = &rgba[(uint64_t)(height - 1 - y) * stride];
+    if (fwrite(src, line_size, 1, fout) != 1) {
       return 0;
     }
     // write padding zeroes
@@ -345,11 +356,11 @@ int WebPWriteBMP(FILE* fout, const WebPDecBuffer* const buffer) {
         return 0;
       }
     }
-    rgba += stride;
   }
   return 1;
 }
 #undef BMP_HEADER_SIZE
+#undef BMP_HEADER_ALPHA_EXTRA_SIZE
 
 //------------------------------------------------------------------------------
 // TIFF

imageio/jpegdec.c

@@ -336,7 +336,11 @@ int ReadJPEG(const uint8_t* const data, size_t data_size,
   pic->width = width;
   pic->height = height;
   ok = WebPPictureImportRGB(pic, rgb, (int)stride);
-  if (!ok) goto Error;
+  if (!ok) {
+    pic->width = 0;   // WebPPictureImportRGB() barely touches 'pic' on failure.
+    pic->height = 0;  // Just reset dimensions but keep any 'custom_ptr' etc.
+    MetadataFree(metadata);  // In case the caller forgets to free it on error.
+  }
 
  End:
   free(rgb);

imageio/pnmdec.c

@@ -20,6 +20,10 @@
 #include "webp/encode.h"
 #include "./imageio_util.h"
 
+#if defined(_MSC_VER) && _MSC_VER < 1900
+#define snprintf _snprintf
+#endif
+
 typedef enum {
   WIDTH_FLAG      = 1 << 0,
   HEIGHT_FLAG     = 1 << 1,
@@ -111,8 +115,9 @@ static size_t ReadPAMFields(PNMInfo* const info, size_t off) {
       break;
     } else {
       static const char kEllipsis[] = " ...";
+      const size_t kLen = strlen(kEllipsis) + 1;  // +1 = trailing \0
       int i;
-      if (out_size > 20) sprintf(out + 20 - strlen(kEllipsis), kEllipsis);
+      if (out_size > 20) snprintf(out + 20 - kLen, kLen, kEllipsis);
       for (i = 0; i < (int)strlen(out); ++i) {
         // isprint() might trigger a "char-subscripts" warning if given a char.
         if (!isprint((int)out[i])) out[i] = ' ';

imageio/tiffdec.c

@@ -188,7 +188,9 @@ int ReadTIFF(const uint8_t* const data, size_t data_size,
     fprintf(stderr, "Error! Cannot retrieve TIFF samples-per-pixel info.\n");
     goto End;
   }
-  if (samples_per_px < 3 || samples_per_px > 4) goto End;  // not supported
+  if (!(samples_per_px == 1 || samples_per_px == 3 || samples_per_px == 4)) {
+    goto End;  // not supported
+  }
 
   if (!(TIFFGetField(tif, TIFFTAG_IMAGEWIDTH, &image_width) &&
         TIFFGetField(tif, TIFFTAG_IMAGELENGTH, &image_height))) {
@@ -212,7 +214,7 @@ int ReadTIFF(const uint8_t* const data, size_t data_size,
       TIFFGetField(tif, TIFFTAG_TILELENGTH, &tile_height)) {
     if ((tile_width > 32 && tile_width / 2 > image_width) ||
         (tile_height > 32 && tile_height / 2 > image_height) ||
-        ImgIoUtilCheckSizeArgumentsOverflow(
+        !ImgIoUtilCheckSizeArgumentsOverflow(
             (uint64_t)tile_width * sizeof(*raster), tile_height)) {
       fprintf(stderr, "Error! TIFF tile dimension (%d x %d) is too large.\n",
               tile_width, tile_height);

infra/run_static_analysis.sh

@@ -85,10 +85,10 @@ scan_build make -j4
 
 index="$(find "${OUTPUT_DIR}" -name index.html)"
 if [[ -f "${index}" ]]; then
-  mv "$(dirname "${index}")/"* .
+  mv "$(dirname "${index}")/"* "${OUTPUT_DIR}"
 else
   # make a empty report to wipe out any old bug reports.
-  cat << EOT > index.html
+  cat << EOT > "${OUTPUT_DIR}/index.html"
 <html>
 <body>
 No bugs reported.

iosbuild.sh

@@ -133,10 +133,9 @@ for PLATFORM in ${PLATFORMS}; do
     CFLAGS="${CFLAGS}"
   set +x
 
-  # run make only in the src/ directory to create libwebp.a/libwebpdecoder.a
+  # Build only the libraries, skip the examples.
-  cd src/
+  make V=0 -C sharpyuv
-  make V=0
+  make V=0 -C src install
-  make install
 
   LIBLIST+=" ${ROOTDIR}/lib/libwebp.a"
   DECLIBLIST+=" ${ROOTDIR}/lib/libwebpdecoder.a"
@@ -144,7 +143,6 @@ for PLATFORM in ${PLATFORMS}; do
   DEMUXLIBLIST+=" ${ROOTDIR}/lib/libwebpdemux.a"
 
   make clean
-  cd ..
 
   export PATH=${OLDPATH}
 done

makefile.unix

@@ -3,7 +3,8 @@
 # It will not install the libraries system-wide, but just create the 'cwebp'
 # and 'dwebp' tools in the examples/ directory, along with the static
 # libraries 'src/libwebp.a', 'src/libwebpdecoder.a', 'src/mux/libwebpmux.a',
-# 'src/demux/libwebpdemux.a' and 'extras/libwebpextras.a'.
+# 'src/demux/libwebpdemux.a', 'extras/libwebpextras.a' and
+# 'sharpyuv/libsharpyuv.a'.
 #
 # To build the library and examples, use:
 #    make -f makefile.unix
@@ -125,6 +126,15 @@ endif
 ANIM_UTIL_OBJS = \
     examples/anim_util.o \
 
+SHARPYUV_OBJS = \
+    sharpyuv/sharpyuv.o \
+    sharpyuv/sharpyuv_cpu.o \
+    sharpyuv/sharpyuv_csp.o \
+    sharpyuv/sharpyuv_dsp.o \
+    sharpyuv/sharpyuv_gamma.o \
+    sharpyuv/sharpyuv_neon.o \
+    sharpyuv/sharpyuv_sse2.o \
+
 DEC_OBJS = \
     src/dec/alpha_dec.o \
     src/dec/buffer_dec.o \
@@ -282,11 +292,12 @@ EXTRA_OBJS = \
     extras/quality_estimate.o \
 
 LIBWEBPDECODER_OBJS = $(DEC_OBJS) $(DSP_DEC_OBJS) $(UTILS_DEC_OBJS)
-LIBWEBP_OBJS = $(LIBWEBPDECODER_OBJS) $(ENC_OBJS) $(DSP_ENC_OBJS) \
+LIBWEBP_OBJS = $(LIBWEBPDECODER_OBJS) $(ENC_OBJS) \
-               $(UTILS_ENC_OBJS)
+               $(DSP_ENC_OBJS) $(UTILS_ENC_OBJS)
 LIBWEBPMUX_OBJS = $(MUX_OBJS)
 LIBWEBPDEMUX_OBJS = $(DEMUX_OBJS)
 LIBWEBPEXTRA_OBJS = $(EXTRA_OBJS)
+LIBSHARPYUV_OBJS = $(SHARPYUV_OBJS)
 
 HDRS_INSTALLED = \
     src/webp/decode.h \
@@ -296,6 +307,11 @@ HDRS_INSTALLED = \
     src/webp/mux_types.h \
     src/webp/types.h \
 
+SHARPYUV_HDRS_INSTALLED = \
+    sharpyuv/sharpyuv.h \
+    sharpyuv/sharpyuv_cpu.h \
+    sharpyuv/sharpyuv_csp.h \
+
 HDRS = \
     src/dec/alphai_dec.h \
     src/dec/common_dec.h \
@@ -304,6 +320,7 @@ HDRS = \
     src/dec/vp8li_dec.h \
     src/dec/webpi_dec.h \
     src/dsp/common_sse2.h \
+    src/dsp/cpu.h \
     src/dsp/dsp.h \
     src/dsp/lossless.h \
     src/dsp/lossless_common.h \
@@ -334,6 +351,7 @@ HDRS = \
     src/utils/utils.h \
     src/webp/format_constants.h \
     $(HDRS_INSTALLED) \
+    $(SHARPYUV_HDRS_INSTALLED) \
 
 OUT_LIBS = examples/libexample_util.a
 OUT_LIBS += imageio/libimageio_util.a
@@ -341,6 +359,7 @@ OUT_LIBS += imageio/libimagedec.a
 OUT_LIBS += imageio/libimageenc.a
 OUT_LIBS += src/libwebpdecoder.a
 OUT_LIBS += src/libwebp.a
+OUT_LIBS += sharpyuv/libsharpyuv.a
 EXTRA_LIB = extras/libwebpextras.a
 OUT_EXAMPLES = examples/cwebp examples/dwebp
 EXTRA_EXAMPLES = examples/gif2webp examples/vwebp examples/webpmux \
@@ -386,6 +405,7 @@ src/libwebpdecoder.a: $(LIBWEBPDECODER_OBJS)
 src/libwebp.a: $(LIBWEBP_OBJS)
 src/mux/libwebpmux.a: $(LIBWEBPMUX_OBJS)
 src/demux/libwebpdemux.a: $(LIBWEBPDEMUX_OBJS)
+sharpyuv/libsharpyuv.a: $(LIBSHARPYUV_OBJS)
 
 %.a:
 	$(AR) $(ARFLAGS) $@ $^
@@ -403,6 +423,7 @@ examples/webpinfo: examples/webpinfo.o
 examples/anim_diff: examples/libanim_util.a examples/libgifdec.a
 examples/anim_diff: src/demux/libwebpdemux.a examples/libexample_util.a
 examples/anim_diff: imageio/libimageio_util.a src/libwebp.a
+examples/anim_diff: sharpyuv/libsharpyuv.a
 examples/anim_diff: override EXTRA_LIBS += $(GIF_LIBS)
 examples/anim_diff: EXTRA_FLAGS += -DWEBP_HAVE_GIF
 examples/anim_dump: examples/libanim_util.a examples/libgifdec.a
@@ -411,12 +432,14 @@ examples/anim_dump: examples/libexample_util.a
 examples/anim_dump: imageio/libimageio_util.a
 examples/anim_dump: imageio/libimageenc.a
 examples/anim_dump: src/libwebp.a
+examples/anim_dump: sharpyuv/libsharpyuv.a
 examples/anim_dump: override EXTRA_LIBS += $(GIF_LIBS) $(DWEBP_LIBS)
 examples/cwebp: examples/libexample_util.a
 examples/cwebp: imageio/libimagedec.a
 examples/cwebp: src/demux/libwebpdemux.a
 examples/cwebp: imageio/libimageio_util.a
 examples/cwebp: src/libwebp.a
+examples/cwebp: sharpyuv/libsharpyuv.a
 examples/cwebp: override EXTRA_LIBS += $(CWEBP_LIBS)
 examples/dwebp: examples/libexample_util.a
 examples/dwebp: imageio/libimagedec.a
@@ -424,13 +447,16 @@ examples/dwebp: src/demux/libwebpdemux.a
 examples/dwebp: imageio/libimageenc.a
 examples/dwebp: imageio/libimageio_util.a
 examples/dwebp: src/libwebp.a
+examples/dwebp: sharpyuv/libsharpyuv.a
 examples/dwebp: override EXTRA_LIBS += $(DWEBP_LIBS)
 examples/gif2webp: examples/libexample_util.a imageio/libimageio_util.a
 examples/gif2webp: examples/libgifdec.a src/mux/libwebpmux.a src/libwebp.a
+examples/gif2webp: sharpyuv/libsharpyuv.a
 examples/gif2webp: override EXTRA_LIBS += $(GIF_LIBS)
 examples/gif2webp: EXTRA_FLAGS += -DWEBP_HAVE_GIF
 examples/vwebp: examples/libexample_util.a src/demux/libwebpdemux.a
 examples/vwebp: imageio/libimageio_util.a src/libwebp.a
+examples/vwebp: sharpyuv/libsharpyuv.a
 examples/vwebp: override EXTRA_LIBS += $(GL_LIBS)
 examples/vwebp: EXTRA_FLAGS += -DWEBP_HAVE_GL
 examples/webpmux: examples/libexample_util.a imageio/libimageio_util.a
@@ -438,7 +464,9 @@ examples/webpmux: src/mux/libwebpmux.a src/libwebpdecoder.a
 examples/img2webp: examples/libexample_util.a imageio/libimageio_util.a
 examples/img2webp: imageio/libimagedec.a
 examples/img2webp: src/demux/libwebpdemux.a
-examples/img2webp: src/mux/libwebpmux.a src/libwebp.a
+examples/img2webp: src/mux/libwebpmux.a
+examples/img2webp: src/libwebp.a
+examples/img2webp: sharpyuv/libsharpyuv.a
 examples/img2webp: override EXTRA_LIBS += $(CWEBP_LIBS)
 examples/webpinfo: examples/libexample_util.a imageio/libimageio_util.a
 examples/webpinfo: src/libwebpdecoder.a
@@ -448,16 +476,19 @@ extras/get_disto: imageio/libimagedec.a
 extras/get_disto: src/demux/libwebpdemux.a
 extras/get_disto: imageio/libimageio_util.a
 extras/get_disto: src/libwebp.a
+extras/get_disto: sharpyuv/libsharpyuv.a
 extras/get_disto: override EXTRA_LIBS += $(CWEBP_LIBS)
 
 extras/webp_quality: extras/webp_quality.o
 extras/webp_quality: imageio/libimageio_util.a
 extras/webp_quality: $(EXTRA_LIB) src/libwebp.a
+extras/webp_quality: sharpyuv/libsharpyuv.a
 
 extras/vwebp_sdl: extras/vwebp_sdl.o
 extras/vwebp_sdl: extras/webp_to_sdl.o
 extras/vwebp_sdl: imageio/libimageio_util.a
 extras/vwebp_sdl: src/libwebp.a
+extras/vwebp_sdl: sharpyuv/libsharpyuv.a
 extras/vwebp_sdl: EXTRA_FLAGS += -DWEBP_HAVE_SDL $(SDL_FLAGS)
 extras/vwebp_sdl: override EXTRA_LIBS += $(SDL_LIBS)
 
@@ -468,12 +499,15 @@ dist: DESTDIR := dist
 dist: OUT_EXAMPLES += $(EXTRA_EXAMPLES)
 dist: all
 	$(INSTALL) -m755 -d $(DESTDIR)/include/webp \
-	           $(DESTDIR)/bin $(DESTDIR)/doc $(DESTDIR)/lib
+	            $(DESTDIR)/include/webp/sharpyuv \
+	            $(DESTDIR)/bin $(DESTDIR)/doc $(DESTDIR)/lib
 	$(INSTALL) -m755 -s $(OUT_EXAMPLES) $(DESTDIR)/bin
 	$(INSTALL) -m644 $(HDRS_INSTALLED) $(DESTDIR)/include/webp
+	$(INSTALL) -m644 $(SHARPYUV_HDRS_INSTALLED) $(DESTDIR)/include/webp/sharpyuv
 	$(INSTALL) -m644 src/libwebp.a $(DESTDIR)/lib
 	$(INSTALL) -m644 src/demux/libwebpdemux.a $(DESTDIR)/lib
 	$(INSTALL) -m644 src/mux/libwebpmux.a $(DESTDIR)/lib
+	$(INSTALL) -m644 sharpyuv/libsharpyuv.a $(DESTDIR)/lib
 	umask 022; \
 	for m in man/[cdv]webp.1 man/gif2webp.1 man/webpmux.1 \
                  man/img2webp.1 man/webpinfo.1; do \
@@ -489,6 +523,7 @@ clean:
               examples/*.o examples/*~ \
               extras/*.o extras/*~ \
               imageio/*.o imageio/*~ \
+              sharpyuv/*.o sharpyuv/*~ \
               src/dec/*.o src/dec/*~ \
               src/demux/*.o src/demux/*~ \
               src/dsp/*.o src/dsp/*~ \

man/cwebp.1

@@ -1,5 +1,5 @@
 .\"                                      Hey, EMACS: -*- nroff -*-
-.TH CWEBP 1 "November 17, 2021"
+.TH CWEBP 1 "March 17, 2022"
 .SH NAME
 cwebp \- compress an image file to a WebP file
 .SH SYNOPSIS
@@ -90,15 +90,17 @@ additional encoding possibilities and decide on the quality gain.
 Lower value can result in faster processing time at the expense of
 larger file size and lower compression quality.
 .TP
-.BI \-resize " width height
-Resize the source to a rectangle with size \fBwidth\fP x \fBheight\fP.
-If either (but not both) of the \fBwidth\fP or \fBheight\fP parameters is 0,
-the value will be calculated preserving the aspect\-ratio.
-.TP
 .BI \-crop " x_position y_position width height
 Crop the source to a rectangle with top\-left corner at coordinates
 (\fBx_position\fP, \fBy_position\fP) and size \fBwidth\fP x \fBheight\fP.
 This cropping area must be fully contained within the source rectangle.
+Note: the cropping is applied \fIbefore\fP any scaling.
+.TP
+.BI \-resize " width height
+Resize the source to a rectangle with size \fBwidth\fP x \fBheight\fP.
+If either (but not both) of the \fBwidth\fP or \fBheight\fP parameters is 0,
+the value will be calculated preserving the aspect\-ratio. Note: scaling
+is applied \fIafter\fP cropping.
 .TP
 .B \-mt
 Use multi\-threading for encoding, if possible.

man/img2webp.1

@@ -1,10 +1,10 @@
 .\"                                      Hey, EMACS: -*- nroff -*-
-.TH IMG2WEBP 1 "November 17, 2021"
+.TH IMG2WEBP 1 "January 5, 2022"
 .SH NAME
 img2webp \- create animated WebP file from a sequence of input images.
 .SH SYNOPSIS
 .B img2webp
-[file_level_options] [files] [per_frame_options...]
+[file_options] [[frame_options] frame_file]...
 .br
 .B img2webp argument_file_name
 .br

sharpyuv/Makefile.am

@@ -0,0 +1,41 @@
+AM_CPPFLAGS += -I$(top_builddir) -I$(top_srcdir)
+AM_CPPFLAGS += -I$(top_builddir)/src -I$(top_srcdir)/src
+
+lib_LTLIBRARIES = libsharpyuv.la
+
+noinst_LTLIBRARIES =
+noinst_LTLIBRARIES += libsharpyuv_sse2.la
+noinst_LTLIBRARIES += libsharpyuv_neon.la
+
+libsharpyuvinclude_HEADERS =
+libsharpyuvinclude_HEADERS += sharpyuv.h
+libsharpyuvinclude_HEADERS += sharpyuv_csp.h
+noinst_HEADERS =
+noinst_HEADERS += ../src/dsp/cpu.c
+noinst_HEADERS += ../src/dsp/cpu.h
+noinst_HEADERS += ../src/webp/types.h
+
+libsharpyuv_sse2_la_SOURCES =
+libsharpyuv_sse2_la_SOURCES += sharpyuv_sse2.c
+libsharpyuv_sse2_la_CPPFLAGS = $(libsharpyuv_la_CPPFLAGS)
+libsharpyuv_sse2_la_CFLAGS = $(AM_CFLAGS) $(SSE2_FLAGS)
+
+libsharpyuv_neon_la_SOURCES =
+libsharpyuv_neon_la_SOURCES += sharpyuv_neon.c
+libsharpyuv_neon_la_CPPFLAGS = $(libsharpyuv_la_CPPFLAGS)
+libsharpyuv_neon_la_CFLAGS = $(AM_CFLAGS) $(NEON_FLAGS)
+
+libsharpyuv_la_SOURCES =
+libsharpyuv_la_SOURCES += sharpyuv_cpu.c sharpyuv_cpu.h
+libsharpyuv_la_SOURCES += sharpyuv_csp.c sharpyuv_csp.h
+libsharpyuv_la_SOURCES += sharpyuv_dsp.c sharpyuv_dsp.h
+libsharpyuv_la_SOURCES += sharpyuv_gamma.c sharpyuv_gamma.h
+libsharpyuv_la_SOURCES += sharpyuv.c sharpyuv.h
+
+libsharpyuv_la_CPPFLAGS = $(AM_CPPFLAGS)
+libsharpyuv_la_LDFLAGS = -no-undefined -version-info 0:0:0 -lm
+libsharpyuv_la_LIBADD =
+libsharpyuv_la_LIBADD += libsharpyuv_sse2.la
+libsharpyuv_la_LIBADD += libsharpyuv_neon.la
+libsharpyuvincludedir = $(includedir)/webp/sharpyuv
+pkgconfig_DATA = libsharpyuv.pc

sharpyuv/libsharpyuv.pc.in

@@ -0,0 +1,11 @@
+prefix=@prefix@
+exec_prefix=@exec_prefix@
+libdir=@libdir@
+includedir=@includedir@/webp
+
+Name: libsharpyuv
+Description: Library for sharp RGB to YUV conversion
+Version: @PACKAGE_VERSION@
+Cflags: -I${includedir}
+Libs: -L${libdir} -l@webp_libname_prefix@sharpyuv
+Libs.private: -lm @PTHREAD_CFLAGS@ @PTHREAD_LIBS@

sharpyuv/libsharpyuv.rc

@@ -0,0 +1,41 @@
+#define APSTUDIO_READONLY_SYMBOLS
+#include "winres.h"
+#undef APSTUDIO_READONLY_SYMBOLS
+
+#if !defined(AFX_RESOURCE_DLL) || defined(AFX_TARG_ENU)
+LANGUAGE LANG_ENGLISH, SUBLANG_ENGLISH_US
+
+VS_VERSION_INFO VERSIONINFO
+ FILEVERSION 0,0,2,0
+ PRODUCTVERSION 0,0,2,0
+ FILEFLAGSMASK 0x3fL
+#ifdef _DEBUG
+ FILEFLAGS 0x1L
+#else
+ FILEFLAGS 0x0L
+#endif
+ FILEOS 0x40004L
+ FILETYPE 0x2L
+ FILESUBTYPE 0x0L
+BEGIN
+    BLOCK "StringFileInfo"
+    BEGIN
+        BLOCK "040904b0"
+        BEGIN
+            VALUE "CompanyName", "Google, Inc."
+            VALUE "FileDescription", "libsharpyuv DLL"
+            VALUE "FileVersion", "0.2.0"
+            VALUE "InternalName", "libsharpyuv.dll"
+            VALUE "LegalCopyright", "Copyright (C) 2022"
+            VALUE "OriginalFilename", "libsharpyuv.dll"
+            VALUE "ProductName", "SharpYuv Library"
+            VALUE "ProductVersion", "0.2.0"
+        END
+    END
+    BLOCK "VarFileInfo"
+    BEGIN
+        VALUE "Translation", 0x409, 1200
+    END
+END
+
+#endif    // English (United States) resources

sharpyuv/sharpyuv.c

@@ -0,0 +1,526 @@
+// Copyright 2022 Google Inc. All Rights Reserved.
+//
+// Use of this source code is governed by a BSD-style license
+// that can be found in the COPYING file in the root of the source
+// tree. An additional intellectual property rights grant can be found
+// in the file PATENTS. All contributing project authors may
+// be found in the AUTHORS file in the root of the source tree.
+// -----------------------------------------------------------------------------
+//
+// Sharp RGB to YUV conversion.
+//
+// Author: Skal (pascal.massimino@gmail.com)
+
+#include "sharpyuv/sharpyuv.h"
+
+#include <assert.h>
+#include <limits.h>
+#include <stddef.h>
+#include <stdlib.h>
+#include <string.h>
+
+#include "src/webp/types.h"
+#include "sharpyuv/sharpyuv_cpu.h"
+#include "sharpyuv/sharpyuv_dsp.h"
+#include "sharpyuv/sharpyuv_gamma.h"
+
+//------------------------------------------------------------------------------
+
+int SharpYuvGetVersion(void) {
+  return SHARPYUV_VERSION;
+}
+
+//------------------------------------------------------------------------------
+// Sharp RGB->YUV conversion
+
+static const int kNumIterations = 4;
+
+#define YUV_FIX 16  // fixed-point precision for RGB->YUV
+static const int kYuvHalf = 1 << (YUV_FIX - 1);
+
+// Max bit depth so that intermediate calculations fit in 16 bits.
+static const int kMaxBitDepth = 14;
+
+// Returns the precision shift to use based on the input rgb_bit_depth.
+static int GetPrecisionShift(int rgb_bit_depth) {
+  // Try to add 2 bits of precision if it fits in kMaxBitDepth. Otherwise remove
+  // bits if needed.
+  return ((rgb_bit_depth + 2) <= kMaxBitDepth) ? 2
+                                               : (kMaxBitDepth - rgb_bit_depth);
+}
+
+typedef int16_t fixed_t;      // signed type with extra precision for UV
+typedef uint16_t fixed_y_t;   // unsigned type with extra precision for W
+
+//------------------------------------------------------------------------------
+
+static uint8_t clip_8b(fixed_t v) {
+  return (!(v & ~0xff)) ? (uint8_t)v : (v < 0) ? 0u : 255u;
+}
+
+static uint16_t clip(fixed_t v, int max) {
+  return (v < 0) ? 0 : (v > max) ? max : (uint16_t)v;
+}
+
+static fixed_y_t clip_bit_depth(int y, int bit_depth) {
+  const int max = (1 << bit_depth) - 1;
+  return (!(y & ~max)) ? (fixed_y_t)y : (y < 0) ? 0 : max;
+}
+
+//------------------------------------------------------------------------------
+
+static int RGBToGray(int64_t r, int64_t g, int64_t b) {
+  const int64_t luma = 13933 * r + 46871 * g + 4732 * b + kYuvHalf;
+  return (int)(luma >> YUV_FIX);
+}
+
+static uint32_t ScaleDown(uint16_t a, uint16_t b, uint16_t c, uint16_t d,
+                          int rgb_bit_depth) {
+  const int bit_depth = rgb_bit_depth + GetPrecisionShift(rgb_bit_depth);
+  const uint32_t A = SharpYuvGammaToLinear(a, bit_depth);
+  const uint32_t B = SharpYuvGammaToLinear(b, bit_depth);
+  const uint32_t C = SharpYuvGammaToLinear(c, bit_depth);
+  const uint32_t D = SharpYuvGammaToLinear(d, bit_depth);
+  return SharpYuvLinearToGamma((A + B + C + D + 2) >> 2, bit_depth);
+}
+
+static WEBP_INLINE void UpdateW(const fixed_y_t* src, fixed_y_t* dst, int w,
+                                int rgb_bit_depth) {
+  const int bit_depth = rgb_bit_depth + GetPrecisionShift(rgb_bit_depth);
+  int i;
+  for (i = 0; i < w; ++i) {
+    const uint32_t R = SharpYuvGammaToLinear(src[0 * w + i], bit_depth);
+    const uint32_t G = SharpYuvGammaToLinear(src[1 * w + i], bit_depth);
+    const uint32_t B = SharpYuvGammaToLinear(src[2 * w + i], bit_depth);
+    const uint32_t Y = RGBToGray(R, G, B);
+    dst[i] = (fixed_y_t)SharpYuvLinearToGamma(Y, bit_depth);
+  }
+}
+
+static void UpdateChroma(const fixed_y_t* src1, const fixed_y_t* src2,
+                         fixed_t* dst, int uv_w, int rgb_bit_depth) {
+  int i;
+  for (i = 0; i < uv_w; ++i) {
+    const int r =
+        ScaleDown(src1[0 * uv_w + 0], src1[0 * uv_w + 1], src2[0 * uv_w + 0],
+                  src2[0 * uv_w + 1], rgb_bit_depth);
+    const int g =
+        ScaleDown(src1[2 * uv_w + 0], src1[2 * uv_w + 1], src2[2 * uv_w + 0],
+                  src2[2 * uv_w + 1], rgb_bit_depth);
+    const int b =
+        ScaleDown(src1[4 * uv_w + 0], src1[4 * uv_w + 1], src2[4 * uv_w + 0],
+                  src2[4 * uv_w + 1], rgb_bit_depth);
+    const int W = RGBToGray(r, g, b);
+    dst[0 * uv_w] = (fixed_t)(r - W);
+    dst[1 * uv_w] = (fixed_t)(g - W);
+    dst[2 * uv_w] = (fixed_t)(b - W);
+    dst  += 1;
+    src1 += 2;
+    src2 += 2;
+  }
+}
+
+static void StoreGray(const fixed_y_t* rgb, fixed_y_t* y, int w) {
+  int i;
+  assert(w > 0);
+  for (i = 0; i < w; ++i) {
+    y[i] = RGBToGray(rgb[0 * w + i], rgb[1 * w + i], rgb[2 * w + i]);
+  }
+}
+
+//------------------------------------------------------------------------------
+
+static WEBP_INLINE fixed_y_t Filter2(int A, int B, int W0, int bit_depth) {
+  const int v0 = (A * 3 + B + 2) >> 2;
+  return clip_bit_depth(v0 + W0, bit_depth);
+}
+
+//------------------------------------------------------------------------------
+
+static WEBP_INLINE int Shift(int v, int shift) {
+  return (shift >= 0) ? (v << shift) : (v >> -shift);
+}
+
+static void ImportOneRow(const uint8_t* const r_ptr,
+                         const uint8_t* const g_ptr,
+                         const uint8_t* const b_ptr,
+                         int rgb_step,
+                         int rgb_bit_depth,
+                         int pic_width,
+                         fixed_y_t* const dst) {
+  // Convert the rgb_step from a number of bytes to a number of uint8_t or
+  // uint16_t values depending the bit depth.
+  const int step = (rgb_bit_depth > 8) ? rgb_step / 2 : rgb_step;
+  int i;
+  const int w = (pic_width + 1) & ~1;
+  for (i = 0; i < pic_width; ++i) {
+    const int off = i * step;
+    const int shift = GetPrecisionShift(rgb_bit_depth);
+    if (rgb_bit_depth == 8) {
+      dst[i + 0 * w] = Shift(r_ptr[off], shift);
+      dst[i + 1 * w] = Shift(g_ptr[off], shift);
+      dst[i + 2 * w] = Shift(b_ptr[off], shift);
+    } else {
+      dst[i + 0 * w] = Shift(((uint16_t*)r_ptr)[off], shift);
+      dst[i + 1 * w] = Shift(((uint16_t*)g_ptr)[off], shift);
+      dst[i + 2 * w] = Shift(((uint16_t*)b_ptr)[off], shift);
+    }
+  }
+  if (pic_width & 1) {  // replicate rightmost pixel
+    dst[pic_width + 0 * w] = dst[pic_width + 0 * w - 1];
+    dst[pic_width + 1 * w] = dst[pic_width + 1 * w - 1];
+    dst[pic_width + 2 * w] = dst[pic_width + 2 * w - 1];
+  }
+}
+
+static void InterpolateTwoRows(const fixed_y_t* const best_y,
+                               const fixed_t* prev_uv,
+                               const fixed_t* cur_uv,
+                               const fixed_t* next_uv,
+                               int w,
+                               fixed_y_t* out1,
+                               fixed_y_t* out2,
+                               int rgb_bit_depth) {
+  const int uv_w = w >> 1;
+  const int len = (w - 1) >> 1;   // length to filter
+  int k = 3;
+  const int bit_depth = rgb_bit_depth + GetPrecisionShift(rgb_bit_depth);
+  while (k-- > 0) {   // process each R/G/B segments in turn
+    // special boundary case for i==0
+    out1[0] = Filter2(cur_uv[0], prev_uv[0], best_y[0], bit_depth);
+    out2[0] = Filter2(cur_uv[0], next_uv[0], best_y[w], bit_depth);
+
+    SharpYuvFilterRow(cur_uv, prev_uv, len, best_y + 0 + 1, out1 + 1,
+                      bit_depth);
+    SharpYuvFilterRow(cur_uv, next_uv, len, best_y + w + 1, out2 + 1,
+                      bit_depth);
+
+    // special boundary case for i == w - 1 when w is even
+    if (!(w & 1)) {
+      out1[w - 1] = Filter2(cur_uv[uv_w - 1], prev_uv[uv_w - 1],
+                            best_y[w - 1 + 0], bit_depth);
+      out2[w - 1] = Filter2(cur_uv[uv_w - 1], next_uv[uv_w - 1],
+                            best_y[w - 1 + w], bit_depth);
+    }
+    out1 += w;
+    out2 += w;
+    prev_uv += uv_w;
+    cur_uv  += uv_w;
+    next_uv += uv_w;
+  }
+}
+
+static WEBP_INLINE int RGBToYUVComponent(int r, int g, int b,
+                                         const int coeffs[4], int sfix) {
+  const int srounder = 1 << (YUV_FIX + sfix - 1);
+  const int luma = coeffs[0] * r + coeffs[1] * g + coeffs[2] * b +
+                   coeffs[3] + srounder;
+  return (luma >> (YUV_FIX + sfix));
+}
+
+static int ConvertWRGBToYUV(const fixed_y_t* best_y, const fixed_t* best_uv,
+                            uint8_t* y_ptr, int y_stride, uint8_t* u_ptr,
+                            int u_stride, uint8_t* v_ptr, int v_stride,
+                            int rgb_bit_depth,
+                            int yuv_bit_depth, int width, int height,
+                            const SharpYuvConversionMatrix* yuv_matrix) {
+  int i, j;
+  const fixed_t* const best_uv_base = best_uv;
+  const int w = (width + 1) & ~1;
+  const int h = (height + 1) & ~1;
+  const int uv_w = w >> 1;
+  const int uv_h = h >> 1;
+  const int sfix = GetPrecisionShift(rgb_bit_depth);
+  const int yuv_max = (1 << yuv_bit_depth) - 1;
+
+  for (best_uv = best_uv_base, j = 0; j < height; ++j) {
+    for (i = 0; i < width; ++i) {
+      const int off = (i >> 1);
+      const int W = best_y[i];
+      const int r = best_uv[off + 0 * uv_w] + W;
+      const int g = best_uv[off + 1 * uv_w] + W;
+      const int b = best_uv[off + 2 * uv_w] + W;
+      const int y = RGBToYUVComponent(r, g, b, yuv_matrix->rgb_to_y, sfix);
+      if (yuv_bit_depth <= 8) {
+        y_ptr[i] = clip_8b(y);
+      } else {
+        ((uint16_t*)y_ptr)[i] = clip(y, yuv_max);
+      }
+    }
+    best_y += w;
+    best_uv += (j & 1) * 3 * uv_w;
+    y_ptr += y_stride;
+  }
+  for (best_uv = best_uv_base, j = 0; j < uv_h; ++j) {
+    for (i = 0; i < uv_w; ++i) {
+      const int off = i;
+      // Note r, g and b values here are off by W, but a constant offset on all
+      // 3 components doesn't change the value of u and v with a YCbCr matrix.
+      const int r = best_uv[off + 0 * uv_w];
+      const int g = best_uv[off + 1 * uv_w];
+      const int b = best_uv[off + 2 * uv_w];
+      const int u = RGBToYUVComponent(r, g, b, yuv_matrix->rgb_to_u, sfix);
+      const int v = RGBToYUVComponent(r, g, b, yuv_matrix->rgb_to_v, sfix);
+      if (yuv_bit_depth <= 8) {
+        u_ptr[i] = clip_8b(u);
+        v_ptr[i] = clip_8b(v);
+      } else {
+        ((uint16_t*)u_ptr)[i] = clip(u, yuv_max);
+        ((uint16_t*)v_ptr)[i] = clip(v, yuv_max);
+      }
+    }
+    best_uv += 3 * uv_w;
+    u_ptr += u_stride;
+    v_ptr += v_stride;
+  }
+  return 1;
+}
+
+//------------------------------------------------------------------------------
+// Main function
+
+static void* SafeMalloc(uint64_t nmemb, size_t size) {
+  const uint64_t total_size = nmemb * (uint64_t)size;
+  if (total_size != (size_t)total_size) return NULL;
+  return malloc((size_t)total_size);
+}
+
+#define SAFE_ALLOC(W, H, T) ((T*)SafeMalloc((W) * (H), sizeof(T)))
+
+static int DoSharpArgbToYuv(const uint8_t* r_ptr, const uint8_t* g_ptr,
+                            const uint8_t* b_ptr, int rgb_step, int rgb_stride,
+                            int rgb_bit_depth, uint8_t* y_ptr, int y_stride,
+                            uint8_t* u_ptr, int u_stride, uint8_t* v_ptr,
+                            int v_stride, int yuv_bit_depth, int width,
+                            int height,
+                            const SharpYuvConversionMatrix* yuv_matrix) {
+  // we expand the right/bottom border if needed
+  const int w = (width + 1) & ~1;
+  const int h = (height + 1) & ~1;
+  const int uv_w = w >> 1;
+  const int uv_h = h >> 1;
+  uint64_t prev_diff_y_sum = ~0;
+  int j, iter;
+
+  // TODO(skal): allocate one big memory chunk. But for now, it's easier
+  // for valgrind debugging to have several chunks.
+  fixed_y_t* const tmp_buffer = SAFE_ALLOC(w * 3, 2, fixed_y_t);   // scratch
+  fixed_y_t* const best_y_base = SAFE_ALLOC(w, h, fixed_y_t);
+  fixed_y_t* const target_y_base = SAFE_ALLOC(w, h, fixed_y_t);
+  fixed_y_t* const best_rgb_y = SAFE_ALLOC(w, 2, fixed_y_t);
+  fixed_t* const best_uv_base = SAFE_ALLOC(uv_w * 3, uv_h, fixed_t);
+  fixed_t* const target_uv_base = SAFE_ALLOC(uv_w * 3, uv_h, fixed_t);
+  fixed_t* const best_rgb_uv = SAFE_ALLOC(uv_w * 3, 1, fixed_t);
+  fixed_y_t* best_y = best_y_base;
+  fixed_y_t* target_y = target_y_base;
+  fixed_t* best_uv = best_uv_base;
+  fixed_t* target_uv = target_uv_base;
+  const uint64_t diff_y_threshold = (uint64_t)(3.0 * w * h);
+  int ok;
+  assert(w > 0);
+  assert(h > 0);
+
+  if (best_y_base == NULL || best_uv_base == NULL ||
+      target_y_base == NULL || target_uv_base == NULL ||
+      best_rgb_y == NULL || best_rgb_uv == NULL ||
+      tmp_buffer == NULL) {
+    ok = 0;
+    goto End;
+  }
+
+  // Import RGB samples to W/RGB representation.
+  for (j = 0; j < height; j += 2) {
+    const int is_last_row = (j == height - 1);
+    fixed_y_t* const src1 = tmp_buffer + 0 * w;
+    fixed_y_t* const src2 = tmp_buffer + 3 * w;
+
+    // prepare two rows of input
+    ImportOneRow(r_ptr, g_ptr, b_ptr, rgb_step, rgb_bit_depth, width,
+                 src1);
+    if (!is_last_row) {
+      ImportOneRow(r_ptr + rgb_stride, g_ptr + rgb_stride, b_ptr + rgb_stride,
+                   rgb_step, rgb_bit_depth, width, src2);
+    } else {
+      memcpy(src2, src1, 3 * w * sizeof(*src2));
+    }
+    StoreGray(src1, best_y + 0, w);
+    StoreGray(src2, best_y + w, w);
+
+    UpdateW(src1, target_y, w, rgb_bit_depth);
+    UpdateW(src2, target_y + w, w, rgb_bit_depth);
+    UpdateChroma(src1, src2, target_uv, uv_w, rgb_bit_depth);
+    memcpy(best_uv, target_uv, 3 * uv_w * sizeof(*best_uv));
+    best_y += 2 * w;
+    best_uv += 3 * uv_w;
+    target_y += 2 * w;
+    target_uv += 3 * uv_w;
+    r_ptr += 2 * rgb_stride;
+    g_ptr += 2 * rgb_stride;
+    b_ptr += 2 * rgb_stride;
+  }
+
+  // Iterate and resolve clipping conflicts.
+  for (iter = 0; iter < kNumIterations; ++iter) {
+    const fixed_t* cur_uv = best_uv_base;
+    const fixed_t* prev_uv = best_uv_base;
+    uint64_t diff_y_sum = 0;
+
+    best_y = best_y_base;
+    best_uv = best_uv_base;
+    target_y = target_y_base;
+    target_uv = target_uv_base;
+    for (j = 0; j < h; j += 2) {
+      fixed_y_t* const src1 = tmp_buffer + 0 * w;
+      fixed_y_t* const src2 = tmp_buffer + 3 * w;
+      {
+        const fixed_t* const next_uv = cur_uv + ((j < h - 2) ? 3 * uv_w : 0);
+        InterpolateTwoRows(best_y, prev_uv, cur_uv, next_uv, w,
+                           src1, src2, rgb_bit_depth);
+        prev_uv = cur_uv;
+        cur_uv = next_uv;
+      }
+
+      UpdateW(src1, best_rgb_y + 0 * w, w, rgb_bit_depth);
+      UpdateW(src2, best_rgb_y + 1 * w, w, rgb_bit_depth);
+      UpdateChroma(src1, src2, best_rgb_uv, uv_w, rgb_bit_depth);
+
+      // update two rows of Y and one row of RGB
+      diff_y_sum +=
+          SharpYuvUpdateY(target_y, best_rgb_y, best_y, 2 * w,
+                          rgb_bit_depth + GetPrecisionShift(rgb_bit_depth));
+      SharpYuvUpdateRGB(target_uv, best_rgb_uv, best_uv, 3 * uv_w);
+
+      best_y += 2 * w;
+      best_uv += 3 * uv_w;
+      target_y += 2 * w;
+      target_uv += 3 * uv_w;
+    }
+    // test exit condition
+    if (iter > 0) {
+      if (diff_y_sum < diff_y_threshold) break;
+      if (diff_y_sum > prev_diff_y_sum) break;
+    }
+    prev_diff_y_sum = diff_y_sum;
+  }
+
+  // final reconstruction
+  ok = ConvertWRGBToYUV(best_y_base, best_uv_base, y_ptr, y_stride, u_ptr,
+                        u_stride, v_ptr, v_stride, rgb_bit_depth, yuv_bit_depth,
+                        width, height, yuv_matrix);
+
+ End:
+  free(best_y_base);
+  free(best_uv_base);
+  free(target_y_base);
+  free(target_uv_base);
+  free(best_rgb_y);
+  free(best_rgb_uv);
+  free(tmp_buffer);
+  return ok;
+}
+#undef SAFE_ALLOC
+
+#if defined(WEBP_USE_THREAD) && !defined(_WIN32)
+#include <pthread.h>  // NOLINT
+
+#define LOCK_ACCESS \
+    static pthread_mutex_t sharpyuv_lock = PTHREAD_MUTEX_INITIALIZER; \
+    if (pthread_mutex_lock(&sharpyuv_lock)) return
+#define UNLOCK_ACCESS_AND_RETURN                  \
+    do {                                          \
+      (void)pthread_mutex_unlock(&sharpyuv_lock); \
+      return;                                     \
+    } while (0)
+#else  // !(defined(WEBP_USE_THREAD) && !defined(_WIN32))
+#define LOCK_ACCESS do {} while (0)
+#define UNLOCK_ACCESS_AND_RETURN return
+#endif  // defined(WEBP_USE_THREAD) && !defined(_WIN32)
+
+// Hidden exported init function.
+// By default SharpYuvConvert calls it with SharpYuvGetCPUInfo. If needed,
+// users can declare it as extern and call it with an alternate VP8CPUInfo
+// function.
+SHARPYUV_EXTERN void SharpYuvInit(VP8CPUInfo cpu_info_func);
+void SharpYuvInit(VP8CPUInfo cpu_info_func) {
+  static volatile VP8CPUInfo sharpyuv_last_cpuinfo_used =
+      (VP8CPUInfo)&sharpyuv_last_cpuinfo_used;
+  LOCK_ACCESS;
+  // Only update SharpYuvGetCPUInfo when called from external code to avoid a
+  // race on reading the value in SharpYuvConvert().
+  if (cpu_info_func != (VP8CPUInfo)&SharpYuvGetCPUInfo) {
+    SharpYuvGetCPUInfo = cpu_info_func;
+  }
+  if (sharpyuv_last_cpuinfo_used == SharpYuvGetCPUInfo) {
+    UNLOCK_ACCESS_AND_RETURN;
+  }
+
+  SharpYuvInitDsp();
+  SharpYuvInitGammaTables();
+
+  sharpyuv_last_cpuinfo_used = SharpYuvGetCPUInfo;
+  UNLOCK_ACCESS_AND_RETURN;
+}
+
+int SharpYuvConvert(const void* r_ptr, const void* g_ptr,
+                    const void* b_ptr, int rgb_step, int rgb_stride,
+                    int rgb_bit_depth, void* y_ptr, int y_stride,
+                    void* u_ptr, int u_stride, void* v_ptr,
+                    int v_stride, int yuv_bit_depth, int width,
+                    int height, const SharpYuvConversionMatrix* yuv_matrix) {
+  SharpYuvConversionMatrix scaled_matrix;
+  const int rgb_max = (1 << rgb_bit_depth) - 1;
+  const int rgb_round = 1 << (rgb_bit_depth - 1);
+  const int yuv_max = (1 << yuv_bit_depth) - 1;
+  const int sfix = GetPrecisionShift(rgb_bit_depth);
+
+  if (width < 1 || height < 1 || width == INT_MAX || height == INT_MAX ||
+      r_ptr == NULL || g_ptr == NULL || b_ptr == NULL || y_ptr == NULL ||
+      u_ptr == NULL || v_ptr == NULL) {
+    return 0;
+  }
+  if (rgb_bit_depth != 8 && rgb_bit_depth != 10 && rgb_bit_depth != 12 &&
+      rgb_bit_depth != 16) {
+    return 0;
+  }
+  if (yuv_bit_depth != 8 && yuv_bit_depth != 10 && yuv_bit_depth != 12) {
+    return 0;
+  }
+  if (rgb_bit_depth > 8 && (rgb_step % 2 != 0 || rgb_stride %2 != 0)) {
+    // Step/stride should be even for uint16_t buffers.
+    return 0;
+  }
+  if (yuv_bit_depth > 8 &&
+      (y_stride % 2 != 0 || u_stride % 2 != 0 || v_stride % 2 != 0)) {
+    // Stride should be even for uint16_t buffers.
+    return 0;
+  }
+  // The address of the function pointer is used to avoid a read race.
+  SharpYuvInit((VP8CPUInfo)&SharpYuvGetCPUInfo);
+
+  // Add scaling factor to go from rgb_bit_depth to yuv_bit_depth, to the
+  // rgb->yuv conversion matrix.
+  if (rgb_bit_depth == yuv_bit_depth) {
+    memcpy(&scaled_matrix, yuv_matrix, sizeof(scaled_matrix));
+  } else {
+    int i;
+    for (i = 0; i < 3; ++i) {
+      scaled_matrix.rgb_to_y[i] =
+          (yuv_matrix->rgb_to_y[i] * yuv_max + rgb_round) / rgb_max;
+      scaled_matrix.rgb_to_u[i] =
+          (yuv_matrix->rgb_to_u[i] * yuv_max + rgb_round) / rgb_max;
+      scaled_matrix.rgb_to_v[i] =
+          (yuv_matrix->rgb_to_v[i] * yuv_max + rgb_round) / rgb_max;
+    }
+  }
+  // Also incorporate precision change scaling.
+  scaled_matrix.rgb_to_y[3] = Shift(yuv_matrix->rgb_to_y[3], sfix);
+  scaled_matrix.rgb_to_u[3] = Shift(yuv_matrix->rgb_to_u[3], sfix);
+  scaled_matrix.rgb_to_v[3] = Shift(yuv_matrix->rgb_to_v[3], sfix);
+
+  return DoSharpArgbToYuv(r_ptr, g_ptr, b_ptr, rgb_step, rgb_stride,
+                          rgb_bit_depth, y_ptr, y_stride, u_ptr, u_stride,
+                          v_ptr, v_stride, yuv_bit_depth, width, height,
+                          &scaled_matrix);
+}
+
+//------------------------------------------------------------------------------

sharpyuv/sharpyuv.h

@@ -0,0 +1,103 @@
+// Copyright 2022 Google Inc. All Rights Reserved.
+//
+// Use of this source code is governed by a BSD-style license
+// that can be found in the COPYING file in the root of the source
+// tree. An additional intellectual property rights grant can be found
+// in the file PATENTS. All contributing project authors may
+// be found in the AUTHORS file in the root of the source tree.
+// -----------------------------------------------------------------------------
+//
+// Sharp RGB to YUV conversion.
+
+#ifndef WEBP_SHARPYUV_SHARPYUV_H_
+#define WEBP_SHARPYUV_SHARPYUV_H_
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+#ifndef SHARPYUV_EXTERN
+#ifdef WEBP_EXTERN
+#define SHARPYUV_EXTERN WEBP_EXTERN
+#else
+// This explicitly marks library functions and allows for changing the
+// signature for e.g., Windows DLL builds.
+#if defined(__GNUC__) && __GNUC__ >= 4
+#define SHARPYUV_EXTERN extern __attribute__((visibility("default")))
+#else
+#if defined(_MSC_VER) && defined(WEBP_DLL)
+#define SHARPYUV_EXTERN __declspec(dllexport)
+#else
+#define SHARPYUV_EXTERN extern
+#endif /* _MSC_VER && WEBP_DLL */
+#endif /* __GNUC__ >= 4 */
+#endif /* WEBP_EXTERN */
+#endif /* SHARPYUV_EXTERN */
+
+// SharpYUV API version following the convention from semver.org
+#define SHARPYUV_VERSION_MAJOR 0
+#define SHARPYUV_VERSION_MINOR 2
+#define SHARPYUV_VERSION_PATCH 0
+// Version as a uint32_t. The major number is the high 8 bits.
+// The minor number is the middle 8 bits. The patch number is the low 16 bits.
+#define SHARPYUV_MAKE_VERSION(MAJOR, MINOR, PATCH) \
+  (((MAJOR) << 24) | ((MINOR) << 16) | (PATCH))
+#define SHARPYUV_VERSION                                                \
+  SHARPYUV_MAKE_VERSION(SHARPYUV_VERSION_MAJOR, SHARPYUV_VERSION_MINOR, \
+                        SHARPYUV_VERSION_PATCH)
+
+// Returns the library's version number, packed in hexadecimal. See
+// SHARPYUV_VERSION.
+SHARPYUV_EXTERN int SharpYuvGetVersion(void);
+
+// RGB to YUV conversion matrix, in 16 bit fixed point.
+// y = rgb_to_y[0] * r + rgb_to_y[1] * g + rgb_to_y[2] * b + rgb_to_y[3]
+// u = rgb_to_u[0] * r + rgb_to_u[1] * g + rgb_to_u[2] * b + rgb_to_u[3]
+// v = rgb_to_v[0] * r + rgb_to_v[1] * g + rgb_to_v[2] * b + rgb_to_v[3]
+// Then y, u and v values are divided by 1<<16 and rounded.
+typedef struct {
+  int rgb_to_y[4];
+  int rgb_to_u[4];
+  int rgb_to_v[4];
+} SharpYuvConversionMatrix;
+
+// Converts RGB to YUV420 using a downsampling algorithm that minimizes
+// artefacts caused by chroma subsampling.
+// This is slower than standard downsampling (averaging of 4 UV values).
+// Assumes that the image will be upsampled using a bilinear filter. If nearest
+// neighbor is used instead, the upsampled image might look worse than with
+// standard downsampling.
+// r_ptr, g_ptr, b_ptr: pointers to the source r, g and b channels. Should point
+//     to uint8_t buffers if rgb_bit_depth is 8, or uint16_t buffers otherwise.
+// rgb_step: distance in bytes between two horizontally adjacent pixels on the
+//     r, g and b channels. If rgb_bit_depth is > 8, it should be a
+//     multiple of 2.
+// rgb_stride: distance in bytes between two vertically adjacent pixels on the
+//     r, g, and b channels. If rgb_bit_depth is > 8, it should be a
+//     multiple of 2.
+// rgb_bit_depth: number of bits for each r/g/b value. One of: 8, 10, 12, 16.
+//     Note: 16 bit input is truncated to 14 bits before conversion to yuv.
+// yuv_bit_depth: number of bits for each y/u/v value. One of: 8, 10, 12.
+// y_ptr, u_ptr, v_ptr: pointers to the destination y, u and v channels.  Should
+//     point to uint8_t buffers if yuv_bit_depth is 8, or uint16_t buffers
+//     otherwise.
+// y_stride, u_stride, v_stride: distance in bytes between two vertically
+//     adjacent pixels on the y, u and v channels. If yuv_bit_depth > 8, they
+//     should be multiples of 2.
+// width, height: width and height of the image in pixels
+SHARPYUV_EXTERN int SharpYuvConvert(const void* r_ptr, const void* g_ptr,
+                                    const void* b_ptr, int rgb_step,
+                                    int rgb_stride, int rgb_bit_depth,
+                                    void* y_ptr, int y_stride, void* u_ptr,
+                                    int u_stride, void* v_ptr, int v_stride,
+                                    int yuv_bit_depth, int width, int height,
+                                    const SharpYuvConversionMatrix* yuv_matrix);
+
+// TODO(b/194336375): Add YUV444 to YUV420 conversion. Maybe also add 422
+// support (it's rarely used in practice, especially for images).
+
+#ifdef __cplusplus
+}  // extern "C"
+#endif
+
+#endif  // WEBP_SHARPYUV_SHARPYUV_H_

sharpyuv/sharpyuv_cpu.c

@@ -0,0 +1,14 @@
+// Copyright 2022 Google Inc. All Rights Reserved.
+//
+// Use of this source code is governed by a BSD-style license
+// that can be found in the COPYING file in the root of the source
+// tree. An additional intellectual property rights grant can be found
+// in the file PATENTS. All contributing project authors may
+// be found in the AUTHORS file in the root of the source tree.
+// -----------------------------------------------------------------------------
+//
+#include "sharpyuv/sharpyuv_cpu.h"
+
+// Include src/dsp/cpu.c to create SharpYuvGetCPUInfo from VP8GetCPUInfo. The
+// function pointer is renamed in sharpyuv_cpu.h.
+#include "src/dsp/cpu.c"

sharpyuv/sharpyuv_cpu.h

@@ -0,0 +1,22 @@
+// Copyright 2022 Google Inc. All Rights Reserved.
+//
+// Use of this source code is governed by a BSD-style license
+// that can be found in the COPYING file in the root of the source
+// tree. An additional intellectual property rights grant can be found
+// in the file PATENTS. All contributing project authors may
+// be found in the AUTHORS file in the root of the source tree.
+// -----------------------------------------------------------------------------
+//
+#ifndef WEBP_SHARPYUV_SHARPYUV_CPU_H_
+#define WEBP_SHARPYUV_SHARPYUV_CPU_H_
+
+#include "sharpyuv/sharpyuv.h"
+
+// Avoid exporting SharpYuvGetCPUInfo in shared object / DLL builds.
+// SharpYuvInit() replaces the use of the function pointer.
+#undef WEBP_EXTERN
+#define WEBP_EXTERN extern
+#define VP8GetCPUInfo SharpYuvGetCPUInfo
+#include "src/dsp/cpu.h"
+
+#endif  // WEBP_SHARPYUV_SHARPYUV_CPU_H_

sharpyuv/sharpyuv_csp.c

@@ -0,0 +1,110 @@
+// Copyright 2022 Google Inc. All Rights Reserved.
+//
+// Use of this source code is governed by a BSD-style license
+// that can be found in the COPYING file in the root of the source
+// tree. An additional intellectual property rights grant can be found
+// in the file PATENTS. All contributing project authors may
+// be found in the AUTHORS file in the root of the source tree.
+// -----------------------------------------------------------------------------
+//
+// Colorspace utilities.
+
+#include "sharpyuv/sharpyuv_csp.h"
+
+#include <assert.h>
+#include <math.h>
+#include <stddef.h>
+
+static int ToFixed16(float f) { return (int)floor(f * (1 << 16) + 0.5f); }
+
+void SharpYuvComputeConversionMatrix(const SharpYuvColorSpace* yuv_color_space,
+                                     SharpYuvConversionMatrix* matrix) {
+  const float kr = yuv_color_space->kr;
+  const float kb = yuv_color_space->kb;
+  const float kg = 1.0f - kr - kb;
+  const float cr = 0.5f / (1.0f - kb);
+  const float cb = 0.5f / (1.0f - kr);
+
+  const int shift = yuv_color_space->bit_depth - 8;
+
+  const float denom = (float)((1 << yuv_color_space->bit_depth) - 1);
+  float scale_y = 1.0f;
+  float add_y = 0.0f;
+  float scale_u = cr;
+  float scale_v = cb;
+  float add_uv = (float)(128 << shift);
+  assert(yuv_color_space->bit_depth >= 8);
+
+  if (yuv_color_space->range == kSharpYuvRangeLimited) {
+    scale_y *= (219 << shift) / denom;
+    scale_u *= (224 << shift) / denom;
+    scale_v *= (224 << shift) / denom;
+    add_y = (float)(16 << shift);
+  }
+
+  matrix->rgb_to_y[0] = ToFixed16(kr * scale_y);
+  matrix->rgb_to_y[1] = ToFixed16(kg * scale_y);
+  matrix->rgb_to_y[2] = ToFixed16(kb * scale_y);
+  matrix->rgb_to_y[3] = ToFixed16(add_y);
+
+  matrix->rgb_to_u[0] = ToFixed16(-kr * scale_u);
+  matrix->rgb_to_u[1] = ToFixed16(-kg * scale_u);
+  matrix->rgb_to_u[2] = ToFixed16((1 - kb) * scale_u);
+  matrix->rgb_to_u[3] = ToFixed16(add_uv);
+
+  matrix->rgb_to_v[0] = ToFixed16((1 - kr) * scale_v);
+  matrix->rgb_to_v[1] = ToFixed16(-kg * scale_v);
+  matrix->rgb_to_v[2] = ToFixed16(-kb * scale_v);
+  matrix->rgb_to_v[3] = ToFixed16(add_uv);
+}
+
+// Matrices are in YUV_FIX fixed point precision.
+// WebP's matrix, similar but not identical to kRec601LimitedMatrix.
+static const SharpYuvConversionMatrix kWebpMatrix = {
+  {16839, 33059, 6420, 16 << 16},
+  {-9719, -19081, 28800, 128 << 16},
+  {28800, -24116, -4684, 128 << 16},
+};
+// Kr=0.2990f Kb=0.1140f bits=8 range=kSharpYuvRangeLimited
+static const SharpYuvConversionMatrix kRec601LimitedMatrix = {
+  {16829, 33039, 6416, 16 << 16},
+  {-9714, -19071, 28784, 128 << 16},
+  {28784, -24103, -4681, 128 << 16},
+};
+// Kr=0.2990f Kb=0.1140f bits=8 range=kSharpYuvRangeFull
+static const SharpYuvConversionMatrix kRec601FullMatrix = {
+  {19595, 38470, 7471, 0},
+  {-11058, -21710, 32768, 128 << 16},
+  {32768, -27439, -5329, 128 << 16},
+};
+// Kr=0.2126f Kb=0.0722f bits=8 range=kSharpYuvRangeLimited
+static const SharpYuvConversionMatrix kRec709LimitedMatrix = {
+  {11966, 40254, 4064, 16 << 16},
+  {-6596, -22189, 28784, 128 << 16},
+  {28784, -26145, -2639, 128 << 16},
+};
+// Kr=0.2126f Kb=0.0722f bits=8 range=kSharpYuvRangeFull
+static const SharpYuvConversionMatrix kRec709FullMatrix = {
+  {13933, 46871, 4732, 0},
+  {-7509, -25259, 32768, 128 << 16},
+  {32768, -29763, -3005, 128 << 16},
+};
+
+const SharpYuvConversionMatrix* SharpYuvGetConversionMatrix(
+    SharpYuvMatrixType matrix_type) {
+  switch (matrix_type) {
+    case kSharpYuvMatrixWebp:
+      return &kWebpMatrix;
+    case kSharpYuvMatrixRec601Limited:
+      return &kRec601LimitedMatrix;
+    case kSharpYuvMatrixRec601Full:
+      return &kRec601FullMatrix;
+    case kSharpYuvMatrixRec709Limited:
+      return &kRec709LimitedMatrix;
+    case kSharpYuvMatrixRec709Full:
+      return &kRec709FullMatrix;
+    case kSharpYuvMatrixNum:
+      return NULL;
+  }
+  return NULL;
+}

sharpyuv/sharpyuv_csp.h

@@ -0,0 +1,60 @@
+// Copyright 2022 Google Inc. All Rights Reserved.
+//
+// Use of this source code is governed by a BSD-style license
+// that can be found in the COPYING file in the root of the source
+// tree. An additional intellectual property rights grant can be found
+// in the file PATENTS. All contributing project authors may
+// be found in the AUTHORS file in the root of the source tree.
+// -----------------------------------------------------------------------------
+//
+// Colorspace utilities.
+
+#ifndef WEBP_SHARPYUV_SHARPYUV_CSP_H_
+#define WEBP_SHARPYUV_SHARPYUV_CSP_H_
+
+#include "sharpyuv/sharpyuv.h"
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+// Range of YUV values.
+typedef enum {
+  kSharpYuvRangeFull,     // YUV values between [0;255] (for 8 bit)
+  kSharpYuvRangeLimited   // Y in [16;235], YUV in [16;240] (for 8 bit)
+} SharpYuvRange;
+
+// Constants that define a YUV color space.
+typedef struct {
+  // Kr and Kb are defined such that:
+  // Y = Kr * r + Kg * g + Kb * b where Kg = 1 - Kr - Kb.
+  float kr;
+  float kb;
+  int bit_depth;  // 8, 10 or 12
+  SharpYuvRange range;
+} SharpYuvColorSpace;
+
+// Fills in 'matrix' for the given YUVColorSpace.
+SHARPYUV_EXTERN void SharpYuvComputeConversionMatrix(
+    const SharpYuvColorSpace* yuv_color_space,
+    SharpYuvConversionMatrix* matrix);
+
+// Enums for precomputed conversion matrices.
+typedef enum {
+  kSharpYuvMatrixWebp = 0,
+  kSharpYuvMatrixRec601Limited,
+  kSharpYuvMatrixRec601Full,
+  kSharpYuvMatrixRec709Limited,
+  kSharpYuvMatrixRec709Full,
+  kSharpYuvMatrixNum
+} SharpYuvMatrixType;
+
+// Returns a pointer to a matrix for one of the predefined colorspaces.
+SHARPYUV_EXTERN const SharpYuvConversionMatrix* SharpYuvGetConversionMatrix(
+    SharpYuvMatrixType matrix_type);
+
+#ifdef __cplusplus
+}  // extern "C"
+#endif
+
+#endif  // WEBP_SHARPYUV_SHARPYUV_CSP_H_

sharpyuv/sharpyuv_dsp.c

@@ -0,0 +1,103 @@
+// Copyright 2022 Google Inc. All Rights Reserved.
+//
+// Use of this source code is governed by a BSD-style license
+// that can be found in the COPYING file in the root of the source
+// tree. An additional intellectual property rights grant can be found
+// in the file PATENTS. All contributing project authors may
+// be found in the AUTHORS file in the root of the source tree.
+// -----------------------------------------------------------------------------
+//
+// Speed-critical functions for Sharp YUV.
+//
+// Author: Skal (pascal.massimino@gmail.com)
+
+#include "sharpyuv/sharpyuv_dsp.h"
+
+#include <assert.h>
+#include <stdlib.h>
+
+#include "sharpyuv/sharpyuv_cpu.h"
+
+//-----------------------------------------------------------------------------
+
+#if !WEBP_NEON_OMIT_C_CODE
+static uint16_t clip(int v, int max) {
+  return (v < 0) ? 0 : (v > max) ? max : (uint16_t)v;
+}
+
+static uint64_t SharpYuvUpdateY_C(const uint16_t* ref, const uint16_t* src,
+                                  uint16_t* dst, int len, int bit_depth) {
+  uint64_t diff = 0;
+  int i;
+  const int max_y = (1 << bit_depth) - 1;
+  for (i = 0; i < len; ++i) {
+    const int diff_y = ref[i] - src[i];
+    const int new_y = (int)dst[i] + diff_y;
+    dst[i] = clip(new_y, max_y);
+    diff += (uint64_t)abs(diff_y);
+  }
+  return diff;
+}
+
+static void SharpYuvUpdateRGB_C(const int16_t* ref, const int16_t* src,
+                                int16_t* dst, int len) {
+  int i;
+  for (i = 0; i < len; ++i) {
+    const int diff_uv = ref[i] - src[i];
+    dst[i] += diff_uv;
+  }
+}
+
+static void SharpYuvFilterRow_C(const int16_t* A, const int16_t* B, int len,
+                                const uint16_t* best_y, uint16_t* out,
+                                int bit_depth) {
+  int i;
+  const int max_y = (1 << bit_depth) - 1;
+  for (i = 0; i < len; ++i, ++A, ++B) {
+    const int v0 = (A[0] * 9 + A[1] * 3 + B[0] * 3 + B[1] + 8) >> 4;
+    const int v1 = (A[1] * 9 + A[0] * 3 + B[1] * 3 + B[0] + 8) >> 4;
+    out[2 * i + 0] = clip(best_y[2 * i + 0] + v0, max_y);
+    out[2 * i + 1] = clip(best_y[2 * i + 1] + v1, max_y);
+  }
+}
+#endif  // !WEBP_NEON_OMIT_C_CODE
+
+//-----------------------------------------------------------------------------
+
+uint64_t (*SharpYuvUpdateY)(const uint16_t* src, const uint16_t* ref,
+                            uint16_t* dst, int len, int bit_depth);
+void (*SharpYuvUpdateRGB)(const int16_t* src, const int16_t* ref, int16_t* dst,
+                          int len);
+void (*SharpYuvFilterRow)(const int16_t* A, const int16_t* B, int len,
+                          const uint16_t* best_y, uint16_t* out,
+                          int bit_depth);
+
+extern void InitSharpYuvSSE2(void);
+extern void InitSharpYuvNEON(void);
+
+void SharpYuvInitDsp(void) {
+#if !WEBP_NEON_OMIT_C_CODE
+  SharpYuvUpdateY = SharpYuvUpdateY_C;
+  SharpYuvUpdateRGB = SharpYuvUpdateRGB_C;
+  SharpYuvFilterRow = SharpYuvFilterRow_C;
+#endif
+
+  if (SharpYuvGetCPUInfo != NULL) {
+#if defined(WEBP_HAVE_SSE2)
+    if (SharpYuvGetCPUInfo(kSSE2)) {
+      InitSharpYuvSSE2();
+    }
+#endif  // WEBP_HAVE_SSE2
+  }
+
+#if defined(WEBP_HAVE_NEON)
+  if (WEBP_NEON_OMIT_C_CODE ||
+      (SharpYuvGetCPUInfo != NULL && SharpYuvGetCPUInfo(kNEON))) {
+    InitSharpYuvNEON();
+  }
+#endif  // WEBP_HAVE_NEON
+
+  assert(SharpYuvUpdateY != NULL);
+  assert(SharpYuvUpdateRGB != NULL);
+  assert(SharpYuvFilterRow != NULL);
+}

sharpyuv/sharpyuv_dsp.h

@@ -0,0 +1,28 @@
+// Copyright 2022 Google Inc. All Rights Reserved.
+//
+// Use of this source code is governed by a BSD-style license
+// that can be found in the COPYING file in the root of the source
+// tree. An additional intellectual property rights grant can be found
+// in the file PATENTS. All contributing project authors may
+// be found in the AUTHORS file in the root of the source tree.
+// -----------------------------------------------------------------------------
+//
+// Speed-critical functions for Sharp YUV.
+
+#ifndef WEBP_SHARPYUV_SHARPYUV_DSP_H_
+#define WEBP_SHARPYUV_SHARPYUV_DSP_H_
+
+#include "sharpyuv/sharpyuv_cpu.h"
+#include "src/webp/types.h"
+
+extern uint64_t (*SharpYuvUpdateY)(const uint16_t* src, const uint16_t* ref,
+                                   uint16_t* dst, int len, int bit_depth);
+extern void (*SharpYuvUpdateRGB)(const int16_t* src, const int16_t* ref,
+                                 int16_t* dst, int len);
+extern void (*SharpYuvFilterRow)(const int16_t* A, const int16_t* B, int len,
+                                 const uint16_t* best_y, uint16_t* out,
+                                 int bit_depth);
+
+void SharpYuvInitDsp(void);
+
+#endif  // WEBP_SHARPYUV_SHARPYUV_DSP_H_

sharpyuv/sharpyuv_gamma.c

@@ -0,0 +1,113 @@
+// Copyright 2022 Google Inc. All Rights Reserved.
+//
+// Use of this source code is governed by a BSD-style license
+// that can be found in the COPYING file in the root of the source
+// tree. An additional intellectual property rights grant can be found
+// in the file PATENTS. All contributing project authors may
+// be found in the AUTHORS file in the root of the source tree.
+// -----------------------------------------------------------------------------
+//
+// Gamma correction utilities.
+
+#include "sharpyuv/sharpyuv_gamma.h"
+
+#include <assert.h>
+#include <math.h>
+
+#include "src/webp/types.h"
+
+// Gamma correction compensates loss of resolution during chroma subsampling.
+// Size of pre-computed table for converting from gamma to linear.
+#define GAMMA_TO_LINEAR_TAB_BITS 10
+#define GAMMA_TO_LINEAR_TAB_SIZE (1 << GAMMA_TO_LINEAR_TAB_BITS)
+static uint32_t kGammaToLinearTabS[GAMMA_TO_LINEAR_TAB_SIZE + 2];
+#define LINEAR_TO_GAMMA_TAB_BITS 9
+#define LINEAR_TO_GAMMA_TAB_SIZE (1 << LINEAR_TO_GAMMA_TAB_BITS)
+static uint32_t kLinearToGammaTabS[LINEAR_TO_GAMMA_TAB_SIZE + 2];
+
+static const double kGammaF = 1. / 0.45;
+#define GAMMA_TO_LINEAR_BITS 16
+
+static volatile int kGammaTablesSOk = 0;
+void SharpYuvInitGammaTables(void) {
+  assert(GAMMA_TO_LINEAR_BITS <= 16);
+  if (!kGammaTablesSOk) {
+    int v;
+    const double a = 0.09929682680944;
+    const double thresh = 0.018053968510807;
+    const double final_scale = 1 << GAMMA_TO_LINEAR_BITS;
+    // Precompute gamma to linear table.
+    {
+      const double norm = 1. / GAMMA_TO_LINEAR_TAB_SIZE;
+      const double a_rec = 1. / (1. + a);
+      for (v = 0; v <= GAMMA_TO_LINEAR_TAB_SIZE; ++v) {
+        const double g = norm * v;
+        double value;
+        if (g <= thresh * 4.5) {
+          value = g / 4.5;
+        } else {
+          value = pow(a_rec * (g + a), kGammaF);
+        }
+        kGammaToLinearTabS[v] = (uint32_t)(value * final_scale + .5);
+      }
+      // to prevent small rounding errors to cause read-overflow:
+      kGammaToLinearTabS[GAMMA_TO_LINEAR_TAB_SIZE + 1] =
+          kGammaToLinearTabS[GAMMA_TO_LINEAR_TAB_SIZE];
+    }
+    // Precompute linear to gamma table.
+    {
+      const double scale = 1. / LINEAR_TO_GAMMA_TAB_SIZE;
+      for (v = 0; v <= LINEAR_TO_GAMMA_TAB_SIZE; ++v) {
+        const double g = scale * v;
+        double value;
+        if (g <= thresh) {
+          value = 4.5 * g;
+        } else {
+          value = (1. + a) * pow(g, 1. / kGammaF) - a;
+        }
+        kLinearToGammaTabS[v] =
+            (uint32_t)(final_scale * value + 0.5);
+      }
+      // to prevent small rounding errors to cause read-overflow:
+      kLinearToGammaTabS[LINEAR_TO_GAMMA_TAB_SIZE + 1] =
+          kLinearToGammaTabS[LINEAR_TO_GAMMA_TAB_SIZE];
+    }
+    kGammaTablesSOk = 1;
+  }
+}
+
+static WEBP_INLINE int Shift(int v, int shift) {
+  return (shift >= 0) ? (v << shift) : (v >> -shift);
+}
+
+static WEBP_INLINE uint32_t FixedPointInterpolation(int v, uint32_t* tab,
+                                                    int tab_pos_shift_right,
+                                                    int tab_value_shift) {
+  const uint32_t tab_pos = Shift(v, -tab_pos_shift_right);
+  // fractional part, in 'tab_pos_shift' fixed-point precision
+  const uint32_t x = v - (tab_pos << tab_pos_shift_right);  // fractional part
+  // v0 / v1 are in kGammaToLinearBits fixed-point precision (range [0..1])
+  const uint32_t v0 = Shift(tab[tab_pos + 0], tab_value_shift);
+  const uint32_t v1 = Shift(tab[tab_pos + 1], tab_value_shift);
+  // Final interpolation.
+  const uint32_t v2 = (v1 - v0) * x;  // note: v1 >= v0.
+  const int half =
+      (tab_pos_shift_right > 0) ? 1 << (tab_pos_shift_right - 1) : 0;
+  const uint32_t result = v0 + ((v2 + half) >> tab_pos_shift_right);
+  return result;
+}
+
+uint32_t SharpYuvGammaToLinear(uint16_t v, int bit_depth) {
+  const int shift = GAMMA_TO_LINEAR_TAB_BITS - bit_depth;
+  if (shift > 0) {
+    return kGammaToLinearTabS[v << shift];
+  }
+  return FixedPointInterpolation(v, kGammaToLinearTabS, -shift, 0);
+}
+
+uint16_t SharpYuvLinearToGamma(uint32_t value, int bit_depth) {
+  return FixedPointInterpolation(
+      value, kLinearToGammaTabS,
+      (GAMMA_TO_LINEAR_BITS - LINEAR_TO_GAMMA_TAB_BITS),
+      bit_depth - GAMMA_TO_LINEAR_BITS);
+}

sharpyuv/sharpyuv_gamma.h

@@ -0,0 +1,35 @@
+// Copyright 2022 Google Inc. All Rights Reserved.
+//
+// Use of this source code is governed by a BSD-style license
+// that can be found in the COPYING file in the root of the source
+// tree. An additional intellectual property rights grant can be found
+// in the file PATENTS. All contributing project authors may
+// be found in the AUTHORS file in the root of the source tree.
+// -----------------------------------------------------------------------------
+//
+// Gamma correction utilities.
+
+#ifndef WEBP_SHARPYUV_SHARPYUV_GAMMA_H_
+#define WEBP_SHARPYUV_SHARPYUV_GAMMA_H_
+
+#include "src/webp/types.h"
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+// Initializes precomputed tables. Must be called once before calling
+// SharpYuvGammaToLinear or SharpYuvLinearToGamma.
+void SharpYuvInitGammaTables(void);
+
+// Converts a gamma color value on 'bit_depth' bits to a 16 bit linear value.
+uint32_t SharpYuvGammaToLinear(uint16_t v, int bit_depth);
+
+// Converts a 16 bit linear color value to a gamma value on 'bit_depth' bits.
+uint16_t SharpYuvLinearToGamma(uint32_t value, int bit_depth);
+
+#ifdef __cplusplus
+}  // extern "C"
+#endif
+
+#endif  // WEBP_SHARPYUV_SHARPYUV_GAMMA_H_

sharpyuv/sharpyuv_neon.c

@@ -0,0 +1,181 @@
+// Copyright 2022 Google Inc. All Rights Reserved.
+//
+// Use of this source code is governed by a BSD-style license
+// that can be found in the COPYING file in the root of the source
+// tree. An additional intellectual property rights grant can be found
+// in the file PATENTS. All contributing project authors may
+// be found in the AUTHORS file in the root of the source tree.
+// -----------------------------------------------------------------------------
+//
+// Speed-critical functions for Sharp YUV.
+//
+// Author: Skal (pascal.massimino@gmail.com)
+
+#include "sharpyuv/sharpyuv_dsp.h"
+
+#if defined(WEBP_USE_NEON)
+#include <assert.h>
+#include <stdlib.h>
+#include <arm_neon.h>
+
+static uint16_t clip_NEON(int v, int max) {
+  return (v < 0) ? 0 : (v > max) ? max : (uint16_t)v;
+}
+
+static uint64_t SharpYuvUpdateY_NEON(const uint16_t* ref, const uint16_t* src,
+                                     uint16_t* dst, int len, int bit_depth) {
+  const int max_y = (1 << bit_depth) - 1;
+  int i;
+  const int16x8_t zero = vdupq_n_s16(0);
+  const int16x8_t max = vdupq_n_s16(max_y);
+  uint64x2_t sum = vdupq_n_u64(0);
+  uint64_t diff;
+
+  for (i = 0; i + 8 <= len; i += 8) {
+    const int16x8_t A = vreinterpretq_s16_u16(vld1q_u16(ref + i));
+    const int16x8_t B = vreinterpretq_s16_u16(vld1q_u16(src + i));
+    const int16x8_t C = vreinterpretq_s16_u16(vld1q_u16(dst + i));
+    const int16x8_t D = vsubq_s16(A, B);       // diff_y
+    const int16x8_t F = vaddq_s16(C, D);       // new_y
+    const uint16x8_t H =
+        vreinterpretq_u16_s16(vmaxq_s16(vminq_s16(F, max), zero));
+    const int16x8_t I = vabsq_s16(D);          // abs(diff_y)
+    vst1q_u16(dst + i, H);
+    sum = vpadalq_u32(sum, vpaddlq_u16(vreinterpretq_u16_s16(I)));
+  }
+  diff = vgetq_lane_u64(sum, 0) + vgetq_lane_u64(sum, 1);
+  for (; i < len; ++i) {
+    const int diff_y = ref[i] - src[i];
+    const int new_y = (int)(dst[i]) + diff_y;
+    dst[i] = clip_NEON(new_y, max_y);
+    diff += (uint64_t)(abs(diff_y));
+  }
+  return diff;
+}
+
+static void SharpYuvUpdateRGB_NEON(const int16_t* ref, const int16_t* src,
+                                   int16_t* dst, int len) {
+  int i;
+  for (i = 0; i + 8 <= len; i += 8) {
+    const int16x8_t A = vld1q_s16(ref + i);
+    const int16x8_t B = vld1q_s16(src + i);
+    const int16x8_t C = vld1q_s16(dst + i);
+    const int16x8_t D = vsubq_s16(A, B);   // diff_uv
+    const int16x8_t E = vaddq_s16(C, D);   // new_uv
+    vst1q_s16(dst + i, E);
+  }
+  for (; i < len; ++i) {
+    const int diff_uv = ref[i] - src[i];
+    dst[i] += diff_uv;
+  }
+}
+
+static void SharpYuvFilterRow16_NEON(const int16_t* A, const int16_t* B,
+                                     int len, const uint16_t* best_y,
+                                     uint16_t* out, int bit_depth) {
+  const int max_y = (1 << bit_depth) - 1;
+  int i;
+  const int16x8_t max = vdupq_n_s16(max_y);
+  const int16x8_t zero = vdupq_n_s16(0);
+  for (i = 0; i + 8 <= len; i += 8) {
+    const int16x8_t a0 = vld1q_s16(A + i + 0);
+    const int16x8_t a1 = vld1q_s16(A + i + 1);
+    const int16x8_t b0 = vld1q_s16(B + i + 0);
+    const int16x8_t b1 = vld1q_s16(B + i + 1);
+    const int16x8_t a0b1 = vaddq_s16(a0, b1);
+    const int16x8_t a1b0 = vaddq_s16(a1, b0);
+    const int16x8_t a0a1b0b1 = vaddq_s16(a0b1, a1b0);  // A0+A1+B0+B1
+    const int16x8_t a0b1_2 = vaddq_s16(a0b1, a0b1);    // 2*(A0+B1)
+    const int16x8_t a1b0_2 = vaddq_s16(a1b0, a1b0);    // 2*(A1+B0)
+    const int16x8_t c0 = vshrq_n_s16(vaddq_s16(a0b1_2, a0a1b0b1), 3);
+    const int16x8_t c1 = vshrq_n_s16(vaddq_s16(a1b0_2, a0a1b0b1), 3);
+    const int16x8_t e0 = vrhaddq_s16(c1, a0);
+    const int16x8_t e1 = vrhaddq_s16(c0, a1);
+    const int16x8x2_t f = vzipq_s16(e0, e1);
+    const int16x8_t g0 = vreinterpretq_s16_u16(vld1q_u16(best_y + 2 * i + 0));
+    const int16x8_t g1 = vreinterpretq_s16_u16(vld1q_u16(best_y + 2 * i + 8));
+    const int16x8_t h0 = vaddq_s16(g0, f.val[0]);
+    const int16x8_t h1 = vaddq_s16(g1, f.val[1]);
+    const int16x8_t i0 = vmaxq_s16(vminq_s16(h0, max), zero);
+    const int16x8_t i1 = vmaxq_s16(vminq_s16(h1, max), zero);
+    vst1q_u16(out + 2 * i + 0, vreinterpretq_u16_s16(i0));
+    vst1q_u16(out + 2 * i + 8, vreinterpretq_u16_s16(i1));
+  }
+  for (; i < len; ++i) {
+    const int a0b1 = A[i + 0] + B[i + 1];
+    const int a1b0 = A[i + 1] + B[i + 0];
+    const int a0a1b0b1 = a0b1 + a1b0 + 8;
+    const int v0 = (8 * A[i + 0] + 2 * a1b0 + a0a1b0b1) >> 4;
+    const int v1 = (8 * A[i + 1] + 2 * a0b1 + a0a1b0b1) >> 4;
+    out[2 * i + 0] = clip_NEON(best_y[2 * i + 0] + v0, max_y);
+    out[2 * i + 1] = clip_NEON(best_y[2 * i + 1] + v1, max_y);
+  }
+}
+
+static void SharpYuvFilterRow32_NEON(const int16_t* A, const int16_t* B,
+                                     int len, const uint16_t* best_y,
+                                     uint16_t* out, int bit_depth) {
+  const int max_y = (1 << bit_depth) - 1;
+  int i;
+  const uint16x8_t max = vdupq_n_u16(max_y);
+  for (i = 0; i + 4 <= len; i += 4) {
+    const int16x4_t a0 = vld1_s16(A + i + 0);
+    const int16x4_t a1 = vld1_s16(A + i + 1);
+    const int16x4_t b0 = vld1_s16(B + i + 0);
+    const int16x4_t b1 = vld1_s16(B + i + 1);
+    const int32x4_t a0b1 = vaddl_s16(a0, b1);
+    const int32x4_t a1b0 = vaddl_s16(a1, b0);
+    const int32x4_t a0a1b0b1 = vaddq_s32(a0b1, a1b0);  // A0+A1+B0+B1
+    const int32x4_t a0b1_2 = vaddq_s32(a0b1, a0b1);    // 2*(A0+B1)
+    const int32x4_t a1b0_2 = vaddq_s32(a1b0, a1b0);    // 2*(A1+B0)
+    const int32x4_t c0 = vshrq_n_s32(vaddq_s32(a0b1_2, a0a1b0b1), 3);
+    const int32x4_t c1 = vshrq_n_s32(vaddq_s32(a1b0_2, a0a1b0b1), 3);
+    const int32x4_t e0 = vrhaddq_s32(c1, vmovl_s16(a0));
+    const int32x4_t e1 = vrhaddq_s32(c0, vmovl_s16(a1));
+    const int32x4x2_t f = vzipq_s32(e0, e1);
+
+    const int16x8_t g = vreinterpretq_s16_u16(vld1q_u16(best_y + 2 * i));
+    const int32x4_t h0 = vaddw_s16(f.val[0], vget_low_s16(g));
+    const int32x4_t h1 = vaddw_s16(f.val[1], vget_high_s16(g));
+    const uint16x8_t i_16 = vcombine_u16(vqmovun_s32(h0), vqmovun_s32(h1));
+    const uint16x8_t i_clamped = vminq_u16(i_16, max);
+    vst1q_u16(out + 2 * i + 0, i_clamped);
+  }
+  for (; i < len; ++i) {
+    const int a0b1 = A[i + 0] + B[i + 1];
+    const int a1b0 = A[i + 1] + B[i + 0];
+    const int a0a1b0b1 = a0b1 + a1b0 + 8;
+    const int v0 = (8 * A[i + 0] + 2 * a1b0 + a0a1b0b1) >> 4;
+    const int v1 = (8 * A[i + 1] + 2 * a0b1 + a0a1b0b1) >> 4;
+    out[2 * i + 0] = clip_NEON(best_y[2 * i + 0] + v0, max_y);
+    out[2 * i + 1] = clip_NEON(best_y[2 * i + 1] + v1, max_y);
+  }
+}
+
+static void SharpYuvFilterRow_NEON(const int16_t* A, const int16_t* B, int len,
+                                   const uint16_t* best_y, uint16_t* out,
+                                   int bit_depth) {
+  if (bit_depth <= 10) {
+    SharpYuvFilterRow16_NEON(A, B, len, best_y, out, bit_depth);
+  } else {
+    SharpYuvFilterRow32_NEON(A, B, len, best_y, out, bit_depth);
+  }
+}
+
+//------------------------------------------------------------------------------
+
+extern void InitSharpYuvNEON(void);
+
+WEBP_TSAN_IGNORE_FUNCTION void InitSharpYuvNEON(void) {
+  SharpYuvUpdateY = SharpYuvUpdateY_NEON;
+  SharpYuvUpdateRGB = SharpYuvUpdateRGB_NEON;
+  SharpYuvFilterRow = SharpYuvFilterRow_NEON;
+}
+
+#else  // !WEBP_USE_NEON
+
+extern void InitSharpYuvNEON(void);
+
+void InitSharpYuvNEON(void) {}
+
+#endif  // WEBP_USE_NEON

sharpyuv/sharpyuv_sse2.c

@@ -0,0 +1,201 @@
+// Copyright 2022 Google Inc. All Rights Reserved.
+//
+// Use of this source code is governed by a BSD-style license
+// that can be found in the COPYING file in the root of the source
+// tree. An additional intellectual property rights grant can be found
+// in the file PATENTS. All contributing project authors may
+// be found in the AUTHORS file in the root of the source tree.
+// -----------------------------------------------------------------------------
+//
+// Speed-critical functions for Sharp YUV.
+//
+// Author: Skal (pascal.massimino@gmail.com)
+
+#include "sharpyuv/sharpyuv_dsp.h"
+
+#if defined(WEBP_USE_SSE2)
+#include <stdlib.h>
+#include <emmintrin.h>
+
+static uint16_t clip_SSE2(int v, int max) {
+  return (v < 0) ? 0 : (v > max) ? max : (uint16_t)v;
+}
+
+static uint64_t SharpYuvUpdateY_SSE2(const uint16_t* ref, const uint16_t* src,
+                                     uint16_t* dst, int len, int bit_depth) {
+  const int max_y = (1 << bit_depth) - 1;
+  uint64_t diff = 0;
+  uint32_t tmp[4];
+  int i;
+  const __m128i zero = _mm_setzero_si128();
+  const __m128i max = _mm_set1_epi16(max_y);
+  const __m128i one = _mm_set1_epi16(1);
+  __m128i sum = zero;
+
+  for (i = 0; i + 8 <= len; i += 8) {
+    const __m128i A = _mm_loadu_si128((const __m128i*)(ref + i));
+    const __m128i B = _mm_loadu_si128((const __m128i*)(src + i));
+    const __m128i C = _mm_loadu_si128((const __m128i*)(dst + i));
+    const __m128i D = _mm_sub_epi16(A, B);       // diff_y
+    const __m128i E = _mm_cmpgt_epi16(zero, D);  // sign (-1 or 0)
+    const __m128i F = _mm_add_epi16(C, D);       // new_y
+    const __m128i G = _mm_or_si128(E, one);      // -1 or 1
+    const __m128i H = _mm_max_epi16(_mm_min_epi16(F, max), zero);
+    const __m128i I = _mm_madd_epi16(D, G);      // sum(abs(...))
+    _mm_storeu_si128((__m128i*)(dst + i), H);
+    sum = _mm_add_epi32(sum, I);
+  }
+  _mm_storeu_si128((__m128i*)tmp, sum);
+  diff = tmp[3] + tmp[2] + tmp[1] + tmp[0];
+  for (; i < len; ++i) {
+    const int diff_y = ref[i] - src[i];
+    const int new_y = (int)dst[i] + diff_y;
+    dst[i] = clip_SSE2(new_y, max_y);
+    diff += (uint64_t)abs(diff_y);
+  }
+  return diff;
+}
+
+static void SharpYuvUpdateRGB_SSE2(const int16_t* ref, const int16_t* src,
+                                   int16_t* dst, int len) {
+  int i = 0;
+  for (i = 0; i + 8 <= len; i += 8) {
+    const __m128i A = _mm_loadu_si128((const __m128i*)(ref + i));
+    const __m128i B = _mm_loadu_si128((const __m128i*)(src + i));
+    const __m128i C = _mm_loadu_si128((const __m128i*)(dst + i));
+    const __m128i D = _mm_sub_epi16(A, B);   // diff_uv
+    const __m128i E = _mm_add_epi16(C, D);   // new_uv
+    _mm_storeu_si128((__m128i*)(dst + i), E);
+  }
+  for (; i < len; ++i) {
+    const int diff_uv = ref[i] - src[i];
+    dst[i] += diff_uv;
+  }
+}
+
+static void SharpYuvFilterRow16_SSE2(const int16_t* A, const int16_t* B,
+                                     int len, const uint16_t* best_y,
+                                     uint16_t* out, int bit_depth) {
+  const int max_y = (1 << bit_depth) - 1;
+  int i;
+  const __m128i kCst8 = _mm_set1_epi16(8);
+  const __m128i max = _mm_set1_epi16(max_y);
+  const __m128i zero = _mm_setzero_si128();
+  for (i = 0; i + 8 <= len; i += 8) {
+    const __m128i a0 = _mm_loadu_si128((const __m128i*)(A + i + 0));
+    const __m128i a1 = _mm_loadu_si128((const __m128i*)(A + i + 1));
+    const __m128i b0 = _mm_loadu_si128((const __m128i*)(B + i + 0));
+    const __m128i b1 = _mm_loadu_si128((const __m128i*)(B + i + 1));
+    const __m128i a0b1 = _mm_add_epi16(a0, b1);
+    const __m128i a1b0 = _mm_add_epi16(a1, b0);
+    const __m128i a0a1b0b1 = _mm_add_epi16(a0b1, a1b0);  // A0+A1+B0+B1
+    const __m128i a0a1b0b1_8 = _mm_add_epi16(a0a1b0b1, kCst8);
+    const __m128i a0b1_2 = _mm_add_epi16(a0b1, a0b1);    // 2*(A0+B1)
+    const __m128i a1b0_2 = _mm_add_epi16(a1b0, a1b0);    // 2*(A1+B0)
+    const __m128i c0 = _mm_srai_epi16(_mm_add_epi16(a0b1_2, a0a1b0b1_8), 3);
+    const __m128i c1 = _mm_srai_epi16(_mm_add_epi16(a1b0_2, a0a1b0b1_8), 3);
+    const __m128i d0 = _mm_add_epi16(c1, a0);
+    const __m128i d1 = _mm_add_epi16(c0, a1);
+    const __m128i e0 = _mm_srai_epi16(d0, 1);
+    const __m128i e1 = _mm_srai_epi16(d1, 1);
+    const __m128i f0 = _mm_unpacklo_epi16(e0, e1);
+    const __m128i f1 = _mm_unpackhi_epi16(e0, e1);
+    const __m128i g0 = _mm_loadu_si128((const __m128i*)(best_y + 2 * i + 0));
+    const __m128i g1 = _mm_loadu_si128((const __m128i*)(best_y + 2 * i + 8));
+    const __m128i h0 = _mm_add_epi16(g0, f0);
+    const __m128i h1 = _mm_add_epi16(g1, f1);
+    const __m128i i0 = _mm_max_epi16(_mm_min_epi16(h0, max), zero);
+    const __m128i i1 = _mm_max_epi16(_mm_min_epi16(h1, max), zero);
+    _mm_storeu_si128((__m128i*)(out + 2 * i + 0), i0);
+    _mm_storeu_si128((__m128i*)(out + 2 * i + 8), i1);
+  }
+  for (; i < len; ++i) {
+    //   (9 * A0 + 3 * A1 + 3 * B0 + B1 + 8) >> 4 =
+    // = (8 * A0 + 2 * (A1 + B0) + (A0 + A1 + B0 + B1 + 8)) >> 4
+    // We reuse the common sub-expressions.
+    const int a0b1 = A[i + 0] + B[i + 1];
+    const int a1b0 = A[i + 1] + B[i + 0];
+    const int a0a1b0b1 = a0b1 + a1b0 + 8;
+    const int v0 = (8 * A[i + 0] + 2 * a1b0 + a0a1b0b1) >> 4;
+    const int v1 = (8 * A[i + 1] + 2 * a0b1 + a0a1b0b1) >> 4;
+    out[2 * i + 0] = clip_SSE2(best_y[2 * i + 0] + v0, max_y);
+    out[2 * i + 1] = clip_SSE2(best_y[2 * i + 1] + v1, max_y);
+  }
+}
+
+static WEBP_INLINE __m128i s16_to_s32(__m128i in) {
+  return _mm_srai_epi32(_mm_unpacklo_epi16(in, in), 16);
+}
+
+static void SharpYuvFilterRow32_SSE2(const int16_t* A, const int16_t* B,
+                                     int len, const uint16_t* best_y,
+                                     uint16_t* out, int bit_depth) {
+  const int max_y = (1 << bit_depth) - 1;
+  int i;
+  const __m128i kCst8 = _mm_set1_epi32(8);
+  const __m128i max = _mm_set1_epi16(max_y);
+  const __m128i zero = _mm_setzero_si128();
+  for (i = 0; i + 4 <= len; i += 4) {
+    const __m128i a0 = s16_to_s32(_mm_loadl_epi64((const __m128i*)(A + i + 0)));
+    const __m128i a1 = s16_to_s32(_mm_loadl_epi64((const __m128i*)(A + i + 1)));
+    const __m128i b0 = s16_to_s32(_mm_loadl_epi64((const __m128i*)(B + i + 0)));
+    const __m128i b1 = s16_to_s32(_mm_loadl_epi64((const __m128i*)(B + i + 1)));
+    const __m128i a0b1 = _mm_add_epi32(a0, b1);
+    const __m128i a1b0 = _mm_add_epi32(a1, b0);
+    const __m128i a0a1b0b1 = _mm_add_epi32(a0b1, a1b0);  // A0+A1+B0+B1
+    const __m128i a0a1b0b1_8 = _mm_add_epi32(a0a1b0b1, kCst8);
+    const __m128i a0b1_2 = _mm_add_epi32(a0b1, a0b1);  // 2*(A0+B1)
+    const __m128i a1b0_2 = _mm_add_epi32(a1b0, a1b0);  // 2*(A1+B0)
+    const __m128i c0 = _mm_srai_epi32(_mm_add_epi32(a0b1_2, a0a1b0b1_8), 3);
+    const __m128i c1 = _mm_srai_epi32(_mm_add_epi32(a1b0_2, a0a1b0b1_8), 3);
+    const __m128i d0 = _mm_add_epi32(c1, a0);
+    const __m128i d1 = _mm_add_epi32(c0, a1);
+    const __m128i e0 = _mm_srai_epi32(d0, 1);
+    const __m128i e1 = _mm_srai_epi32(d1, 1);
+    const __m128i f0 = _mm_unpacklo_epi32(e0, e1);
+    const __m128i f1 = _mm_unpackhi_epi32(e0, e1);
+    const __m128i g = _mm_loadu_si128((const __m128i*)(best_y + 2 * i + 0));
+    const __m128i h_16 = _mm_add_epi16(g, _mm_packs_epi32(f0, f1));
+    const __m128i final = _mm_max_epi16(_mm_min_epi16(h_16, max), zero);
+    _mm_storeu_si128((__m128i*)(out + 2 * i + 0), final);
+  }
+  for (; i < len; ++i) {
+    //   (9 * A0 + 3 * A1 + 3 * B0 + B1 + 8) >> 4 =
+    // = (8 * A0 + 2 * (A1 + B0) + (A0 + A1 + B0 + B1 + 8)) >> 4
+    // We reuse the common sub-expressions.
+    const int a0b1 = A[i + 0] + B[i + 1];
+    const int a1b0 = A[i + 1] + B[i + 0];
+    const int a0a1b0b1 = a0b1 + a1b0 + 8;
+    const int v0 = (8 * A[i + 0] + 2 * a1b0 + a0a1b0b1) >> 4;
+    const int v1 = (8 * A[i + 1] + 2 * a0b1 + a0a1b0b1) >> 4;
+    out[2 * i + 0] = clip_SSE2(best_y[2 * i + 0] + v0, max_y);
+    out[2 * i + 1] = clip_SSE2(best_y[2 * i + 1] + v1, max_y);
+  }
+}
+
+static void SharpYuvFilterRow_SSE2(const int16_t* A, const int16_t* B, int len,
+                                   const uint16_t* best_y, uint16_t* out,
+                                   int bit_depth) {
+  if (bit_depth <= 10) {
+    SharpYuvFilterRow16_SSE2(A, B, len, best_y, out, bit_depth);
+  } else {
+    SharpYuvFilterRow32_SSE2(A, B, len, best_y, out, bit_depth);
+  }
+}
+
+//------------------------------------------------------------------------------
+
+extern void InitSharpYuvSSE2(void);
+
+WEBP_TSAN_IGNORE_FUNCTION void InitSharpYuvSSE2(void) {
+  SharpYuvUpdateY = SharpYuvUpdateY_SSE2;
+  SharpYuvUpdateRGB = SharpYuvUpdateRGB_SSE2;
+  SharpYuvFilterRow = SharpYuvFilterRow_SSE2;
+}
+#else  // !WEBP_USE_SSE2
+
+extern void InitSharpYuvSSE2(void);
+
+void InitSharpYuvSSE2(void) {}
+
+#endif  // WEBP_USE_SSE2

src/Makefile.am

@@ -36,7 +36,7 @@ libwebp_la_LIBADD += utils/libwebputils.la
 # other than the ones listed on the command line, i.e., after linking, it will
 # not have unresolved symbols. Some platforms (Windows among them) require all
 # symbols in shared libraries to be resolved at library creation.
-libwebp_la_LDFLAGS = -no-undefined -version-info 8:3:1
+libwebp_la_LDFLAGS = -no-undefined -version-info 8:6:1
 libwebpincludedir = $(includedir)/webp
 pkgconfig_DATA = libwebp.pc
 
@@ -48,7 +48,7 @@ if BUILD_LIBWEBPDECODER
   libwebpdecoder_la_LIBADD += dsp/libwebpdspdecode.la
   libwebpdecoder_la_LIBADD += utils/libwebputilsdecode.la
 
-  libwebpdecoder_la_LDFLAGS = -no-undefined -version-info 4:3:1
+  libwebpdecoder_la_LDFLAGS = -no-undefined -version-info 4:6:1
   pkgconfig_DATA += libwebpdecoder.pc
 endif
 

src/dec/vp8i_dec.h

@@ -31,8 +31,8 @@ extern "C" {
 
 // version numbers
 #define DEC_MAJ_VERSION 1
-#define DEC_MIN_VERSION 2
+#define DEC_MIN_VERSION 3
-#define DEC_REV_VERSION 2
+#define DEC_REV_VERSION 0
 
 // YUV-cache parameters. Cache is 32-bytes wide (= one cacheline).
 // Constraints are: We need to store one 16x16 block of luma samples (y),

src/dec/vp8l_dec.c

@@ -178,7 +178,7 @@ static WEBP_INLINE int PlaneCodeToDistance(int xsize, int plane_code) {
 
 //------------------------------------------------------------------------------
 // Decodes the next Huffman code from bit-stream.
-// FillBitWindow(br) needs to be called at minimum every second call
+// VP8LFillBitWindow(br) needs to be called at minimum every second call
 // to ReadSymbol, in order to pre-fetch enough bits.
 static WEBP_INLINE int ReadSymbol(const HuffmanCode* table,
                                   VP8LBitReader* const br) {
@@ -253,11 +253,11 @@ static int ReadHuffmanCodeLengths(
   int symbol;
   int max_symbol;
   int prev_code_len = DEFAULT_CODE_LENGTH;
-  HuffmanCode table[1 << LENGTHS_TABLE_BITS];
+  HuffmanTables tables;
 
-  if (!VP8LBuildHuffmanTable(table, LENGTHS_TABLE_BITS,
+  if (!VP8LHuffmanTablesAllocate(1 << LENGTHS_TABLE_BITS, &tables) ||
-                             code_length_code_lengths,
+      !VP8LBuildHuffmanTable(&tables, LENGTHS_TABLE_BITS,
-                             NUM_CODE_LENGTH_CODES)) {
+                             code_length_code_lengths, NUM_CODE_LENGTH_CODES)) {
     goto End;
   }
 
@@ -277,7 +277,7 @@ static int ReadHuffmanCodeLengths(
     int code_len;
     if (max_symbol-- == 0) break;
     VP8LFillBitWindow(br);
-    p = &table[VP8LPrefetchBits(br) & LENGTHS_TABLE_MASK];
+    p = &tables.curr_segment->start[VP8LPrefetchBits(br) & LENGTHS_TABLE_MASK];
     VP8LSetBitPos(br, br->bit_pos_ + p->bits);
     code_len = p->value;
     if (code_len < kCodeLengthLiterals) {
@@ -300,6 +300,7 @@ static int ReadHuffmanCodeLengths(
   ok = 1;
 
  End:
+  VP8LHuffmanTablesDeallocate(&tables);
   if (!ok) dec->status_ = VP8_STATUS_BITSTREAM_ERROR;
   return ok;
 }
@@ -307,7 +308,8 @@ static int ReadHuffmanCodeLengths(
 // 'code_lengths' is pre-allocated temporary buffer, used for creating Huffman
 // tree.
 static int ReadHuffmanCode(int alphabet_size, VP8LDecoder* const dec,
-                           int* const code_lengths, HuffmanCode* const table) {
+                           int* const code_lengths,
+                           HuffmanTables* const table) {
   int ok = 0;
   int size = 0;
   VP8LBitReader* const br = &dec->br_;
@@ -321,7 +323,7 @@ static int ReadHuffmanCode(int alphabet_size, VP8LDecoder* const dec,
     // The first code is either 1 bit or 8 bit code.
     int symbol = VP8LReadBits(br, (first_symbol_len_code == 0) ? 1 : 8);
     code_lengths[symbol] = 1;
-    // The second code (if present), is always 8 bit long.
+    // The second code (if present), is always 8 bits long.
     if (num_symbols == 2) {
       symbol = VP8LReadBits(br, 8);
       code_lengths[symbol] = 1;
@@ -362,8 +364,7 @@ static int ReadHuffmanCodes(VP8LDecoder* const dec, int xsize, int ysize,
   VP8LMetadata* const hdr = &dec->hdr_;
   uint32_t* huffman_image = NULL;
   HTreeGroup* htree_groups = NULL;
-  HuffmanCode* huffman_tables = NULL;
+  HuffmanTables* huffman_tables = &hdr->huffman_tables_;
-  HuffmanCode* huffman_table = NULL;
   int num_htree_groups = 1;
   int num_htree_groups_max = 1;
   int max_alphabet_size = 0;
@@ -372,6 +373,10 @@ static int ReadHuffmanCodes(VP8LDecoder* const dec, int xsize, int ysize,
   int* mapping = NULL;
   int ok = 0;
 
+  // Check the table has been 0 initialized (through InitMetadata).
+  assert(huffman_tables->root.start == NULL);
+  assert(huffman_tables->curr_segment == NULL);
+
   if (allow_recursion && VP8LReadBits(br, 1)) {
     // use meta Huffman codes.
     const int huffman_precision = VP8LReadBits(br, 3) + 2;
@@ -434,16 +439,15 @@ static int ReadHuffmanCodes(VP8LDecoder* const dec, int xsize, int ysize,
 
   code_lengths = (int*)WebPSafeCalloc((uint64_t)max_alphabet_size,
                                       sizeof(*code_lengths));
-  huffman_tables = (HuffmanCode*)WebPSafeMalloc(num_htree_groups * table_size,
-                                                sizeof(*huffman_tables));
   htree_groups = VP8LHtreeGroupsNew(num_htree_groups);
 
-  if (htree_groups == NULL || code_lengths == NULL || huffman_tables == NULL) {
+  if (htree_groups == NULL || code_lengths == NULL ||
+      !VP8LHuffmanTablesAllocate(num_htree_groups * table_size,
+                                 huffman_tables)) {
     dec->status_ = VP8_STATUS_OUT_OF_MEMORY;
     goto Error;
   }
 
-  huffman_table = huffman_tables;
   for (i = 0; i < num_htree_groups_max; ++i) {
     // If the index "i" is unused in the Huffman image, just make sure the
     // coefficients are valid but do not store them.
@@ -468,19 +472,20 @@ static int ReadHuffmanCodes(VP8LDecoder* const dec, int xsize, int ysize,
       int max_bits = 0;
       for (j = 0; j < HUFFMAN_CODES_PER_META_CODE; ++j) {
         int alphabet_size = kAlphabetSize[j];
-        htrees[j] = huffman_table;
         if (j == 0 && color_cache_bits > 0) {
           alphabet_size += (1 << color_cache_bits);
         }
-        size = ReadHuffmanCode(alphabet_size, dec, code_lengths, huffman_table);
+        size =
+            ReadHuffmanCode(alphabet_size, dec, code_lengths, huffman_tables);
+        htrees[j] = huffman_tables->curr_segment->curr_table;
         if (size == 0) {
           goto Error;
         }
         if (is_trivial_literal && kLiteralMap[j] == 1) {
-          is_trivial_literal = (huffman_table->bits == 0);
+          is_trivial_literal = (htrees[j]->bits == 0);
         }
-        total_size += huffman_table->bits;
+        total_size += htrees[j]->bits;
-        huffman_table += size;
+        huffman_tables->curr_segment->curr_table += size;
         if (j <= ALPHA) {
           int local_max_bits = code_lengths[0];
           int k;
@@ -515,14 +520,13 @@ static int ReadHuffmanCodes(VP8LDecoder* const dec, int xsize, int ysize,
   hdr->huffman_image_ = huffman_image;
   hdr->num_htree_groups_ = num_htree_groups;
   hdr->htree_groups_ = htree_groups;
-  hdr->huffman_tables_ = huffman_tables;
 
  Error:
   WebPSafeFree(code_lengths);
   WebPSafeFree(mapping);
   if (!ok) {
     WebPSafeFree(huffman_image);
-    WebPSafeFree(huffman_tables);
+    VP8LHuffmanTablesDeallocate(huffman_tables);
     VP8LHtreeGroupsFree(htree_groups);
   }
   return ok;
@@ -1281,7 +1285,7 @@ static int ExpandColorMap(int num_colors, VP8LTransform* const transform) {
     uint8_t* const new_data = (uint8_t*)new_color_map;
     new_color_map[0] = transform->data_[0];
     for (i = 4; i < 4 * num_colors; ++i) {
-      // Equivalent to AddPixelEq(), on a byte-basis.
+      // Equivalent to VP8LAddPixels(), on a byte-basis.
       new_data[i] = (data[i] + new_data[i - 4]) & 0xff;
     }
     for (; i < 4 * final_num_colors; ++i) {
@@ -1336,7 +1340,7 @@ static int ReadTransform(int* const xsize, int const* ysize,
        ok = ok && ExpandColorMap(num_colors, transform);
       break;
     }
-    case SUBTRACT_GREEN:
+    case SUBTRACT_GREEN_TRANSFORM:
       break;
     default:
       assert(0);    // can't happen
@@ -1358,7 +1362,7 @@ static void ClearMetadata(VP8LMetadata* const hdr) {
   assert(hdr != NULL);
 
   WebPSafeFree(hdr->huffman_image_);
-  WebPSafeFree(hdr->huffman_tables_);
+  VP8LHuffmanTablesDeallocate(&hdr->huffman_tables_);
   VP8LHtreeGroupsFree(hdr->htree_groups_);
   VP8LColorCacheClear(&hdr->color_cache_);
   VP8LColorCacheClear(&hdr->saved_color_cache_);
@@ -1673,7 +1677,7 @@ int VP8LDecodeImage(VP8LDecoder* const dec) {
 
   if (dec == NULL) return 0;
 
-  assert(dec->hdr_.huffman_tables_ != NULL);
+  assert(dec->hdr_.huffman_tables_.root.start != NULL);
   assert(dec->hdr_.htree_groups_ != NULL);
   assert(dec->hdr_.num_htree_groups_ > 0);
 

src/dec/vp8li_dec.h

@@ -51,7 +51,7 @@ typedef struct {
   uint32_t*       huffman_image_;
   int             num_htree_groups_;
   HTreeGroup*     htree_groups_;
-  HuffmanCode*    huffman_tables_;
+  HuffmanTables   huffman_tables_;
 } VP8LMetadata;
 
 typedef struct VP8LDecoder VP8LDecoder;

src/dec/webp_dec.c

@@ -179,7 +179,7 @@ static VP8StatusCode ParseOptionalChunks(const uint8_t** const data,
       return VP8_STATUS_BITSTREAM_ERROR;          // Not a valid chunk size.
     }
     // For odd-sized chunk-payload, there's one byte padding at the end.
-    disk_chunk_size = (CHUNK_HEADER_SIZE + chunk_size + 1) & ~1;
+    disk_chunk_size = (CHUNK_HEADER_SIZE + chunk_size + 1) & ~1u;
     total_size += disk_chunk_size;
 
     // Check that total bytes skipped so far does not exceed riff_size.

src/demux/Makefile.am

@@ -13,6 +13,6 @@ noinst_HEADERS =
 noinst_HEADERS += ../webp/format_constants.h
 
 libwebpdemux_la_LIBADD = ../libwebp.la
-libwebpdemux_la_LDFLAGS = -no-undefined -version-info 2:9:0
+libwebpdemux_la_LDFLAGS = -no-undefined -version-info 2:12:0
 libwebpdemuxincludedir = $(includedir)/webp
 pkgconfig_DATA = libwebpdemux.pc

src/demux/anim_decode.c

@@ -23,6 +23,14 @@
 
 #define NUM_CHANNELS 4
 
+// Channel extraction from a uint32_t representation of a uint8_t RGBA/BGRA
+// buffer.
+#ifdef WORDS_BIGENDIAN
+#define CHANNEL_SHIFT(i) (24 - (i) * 8)
+#else
+#define CHANNEL_SHIFT(i) ((i) * 8)
+#endif
+
 typedef void (*BlendRowFunc)(uint32_t* const, const uint32_t* const, int);
 static void BlendPixelRowNonPremult(uint32_t* const src,
                                     const uint32_t* const dst, int num_pixels);
@@ -209,35 +217,35 @@ static uint8_t BlendChannelNonPremult(uint32_t src, uint8_t src_a,
   const uint8_t dst_channel = (dst >> shift) & 0xff;
   const uint32_t blend_unscaled = src_channel * src_a + dst_channel * dst_a;
   assert(blend_unscaled < (1ULL << 32) / scale);
-  return (blend_unscaled * scale) >> 24;
+  return (blend_unscaled * scale) >> CHANNEL_SHIFT(3);
 }
 
 // Blend 'src' over 'dst' assuming they are NOT pre-multiplied by alpha.
 static uint32_t BlendPixelNonPremult(uint32_t src, uint32_t dst) {
-  const uint8_t src_a = (src >> 24) & 0xff;
+  const uint8_t src_a = (src >> CHANNEL_SHIFT(3)) & 0xff;
 
   if (src_a == 0) {
     return dst;
   } else {
-    const uint8_t dst_a = (dst >> 24) & 0xff;
+    const uint8_t dst_a = (dst >> CHANNEL_SHIFT(3)) & 0xff;
     // This is the approximate integer arithmetic for the actual formula:
     // dst_factor_a = (dst_a * (255 - src_a)) / 255.
     const uint8_t dst_factor_a = (dst_a * (256 - src_a)) >> 8;
     const uint8_t blend_a = src_a + dst_factor_a;
     const uint32_t scale = (1UL << 24) / blend_a;
 
-    const uint8_t blend_r =
+    const uint8_t blend_r = BlendChannelNonPremult(
-        BlendChannelNonPremult(src, src_a, dst, dst_factor_a, scale, 0);
+        src, src_a, dst, dst_factor_a, scale, CHANNEL_SHIFT(0));
-    const uint8_t blend_g =
+    const uint8_t blend_g = BlendChannelNonPremult(
-        BlendChannelNonPremult(src, src_a, dst, dst_factor_a, scale, 8);
+        src, src_a, dst, dst_factor_a, scale, CHANNEL_SHIFT(1));
-    const uint8_t blend_b =
+    const uint8_t blend_b = BlendChannelNonPremult(
-        BlendChannelNonPremult(src, src_a, dst, dst_factor_a, scale, 16);
+        src, src_a, dst, dst_factor_a, scale, CHANNEL_SHIFT(2));
     assert(src_a + dst_factor_a < 256);
 
-    return (blend_r << 0) |
+    return ((uint32_t)blend_r << CHANNEL_SHIFT(0)) |
-           (blend_g << 8) |
+           ((uint32_t)blend_g << CHANNEL_SHIFT(1)) |
-           (blend_b << 16) |
+           ((uint32_t)blend_b << CHANNEL_SHIFT(2)) |
-           ((uint32_t)blend_a << 24);
+           ((uint32_t)blend_a << CHANNEL_SHIFT(3));
   }
 }
 
@@ -247,7 +255,7 @@ static void BlendPixelRowNonPremult(uint32_t* const src,
                                     const uint32_t* const dst, int num_pixels) {
   int i;
   for (i = 0; i < num_pixels; ++i) {
-    const uint8_t src_alpha = (src[i] >> 24) & 0xff;
+    const uint8_t src_alpha = (src[i] >> CHANNEL_SHIFT(3)) & 0xff;
     if (src_alpha != 0xff) {
       src[i] = BlendPixelNonPremult(src[i], dst[i]);
     }
@@ -264,7 +272,7 @@ static WEBP_INLINE uint32_t ChannelwiseMultiply(uint32_t pix, uint32_t scale) {
 
 // Blend 'src' over 'dst' assuming they are pre-multiplied by alpha.
 static uint32_t BlendPixelPremult(uint32_t src, uint32_t dst) {
-  const uint8_t src_a = (src >> 24) & 0xff;
+  const uint8_t src_a = (src >> CHANNEL_SHIFT(3)) & 0xff;
   return src + ChannelwiseMultiply(dst, 256 - src_a);
 }
 
@@ -274,7 +282,7 @@ static void BlendPixelRowPremult(uint32_t* const src, const uint32_t* const dst,
                                  int num_pixels) {
   int i;
   for (i = 0; i < num_pixels; ++i) {
-    const uint8_t src_alpha = (src[i] >> 24) & 0xff;
+    const uint8_t src_alpha = (src[i] >> CHANNEL_SHIFT(3)) & 0xff;
     if (src_alpha != 0xff) {
       src[i] = BlendPixelPremult(src[i], dst[i]);
     }

src/demux/demux.c

@@ -24,8 +24,8 @@
 #include "src/webp/format_constants.h"
 
 #define DMUX_MAJ_VERSION 1
-#define DMUX_MIN_VERSION 2
+#define DMUX_MIN_VERSION 3
-#define DMUX_REV_VERSION 2
+#define DMUX_REV_VERSION 0
 
 typedef struct {
   size_t start_;        // start location of the data
@@ -614,7 +614,6 @@ static int IsValidExtendedFormat(const WebPDemuxer* const dmux) {
 
   while (f != NULL) {
     const int cur_frame_set = f->frame_num_;
-    int frame_count = 0;
 
     // Check frame properties.
     for (; f != NULL && f->frame_num_ == cur_frame_set; f = f->next_) {
@@ -649,8 +648,6 @@ static int IsValidExtendedFormat(const WebPDemuxer* const dmux) {
                             dmux->canvas_width_, dmux->canvas_height_)) {
         return 0;
       }
-
-      ++frame_count;
     }
   }
   return 1;

src/demux/libwebpdemux.pc.in

@@ -8,4 +8,4 @@ Description: Library for parsing the WebP graphics format container
 Version: @PACKAGE_VERSION@
 Requires: libwebp >= 0.2.0
 Cflags: -I${includedir}
-Libs: -L${libdir} -lwebpdemux
+Libs: -L${libdir} -l@webp_libname_prefix@webpdemux

src/demux/libwebpdemux.rc

@@ -6,8 +6,8 @@
 LANGUAGE LANG_ENGLISH, SUBLANG_ENGLISH_US
 
 VS_VERSION_INFO VERSIONINFO
- FILEVERSION 1,0,2,2
+ FILEVERSION 1,0,3,0
- PRODUCTVERSION 1,0,2,2
+ PRODUCTVERSION 1,0,3,0
  FILEFLAGSMASK 0x3fL
 #ifdef _DEBUG
  FILEFLAGS 0x1L
@@ -24,12 +24,12 @@ BEGIN
         BEGIN
             VALUE "CompanyName", "Google, Inc."
             VALUE "FileDescription", "libwebpdemux DLL"
-            VALUE "FileVersion", "1.2.2"
+            VALUE "FileVersion", "1.3.0"
             VALUE "InternalName", "libwebpdemux.dll"
-            VALUE "LegalCopyright", "Copyright (C) 2021"
+            VALUE "LegalCopyright", "Copyright (C) 2022"
             VALUE "OriginalFilename", "libwebpdemux.dll"
             VALUE "ProductName", "WebP Image Demuxer"
-            VALUE "ProductVersion", "1.2.2"
+            VALUE "ProductVersion", "1.3.0"
         END
     END
     BLOCK "VarFileInfo"

src/dsp/Makefile.am

@@ -24,6 +24,7 @@ commondir = $(includedir)/webp
 COMMON_SOURCES =
 COMMON_SOURCES += alpha_processing.c
 COMMON_SOURCES += cpu.c
+COMMON_SOURCES += cpu.h
 COMMON_SOURCES += dec.c
 COMMON_SOURCES += dec_clip_tables.c
 COMMON_SOURCES += dsp.h

src/dsp/alpha_processing_neon.c

@@ -83,7 +83,7 @@ static void ApplyAlphaMultiply_NEON(uint8_t* rgba, int alpha_first,
 static int DispatchAlpha_NEON(const uint8_t* WEBP_RESTRICT alpha,
                               int alpha_stride, int width, int height,
                               uint8_t* WEBP_RESTRICT dst, int dst_stride) {
-  uint32_t alpha_mask = 0xffffffffu;
+  uint32_t alpha_mask = 0xffu;
   uint8x8_t mask8 = vdup_n_u8(0xff);
   uint32_t tmp[2];
   int i, j;
@@ -107,6 +107,7 @@ static int DispatchAlpha_NEON(const uint8_t* WEBP_RESTRICT alpha,
     dst += dst_stride;
   }
   vst1_u8((uint8_t*)tmp, mask8);
+  alpha_mask *= 0x01010101;
   alpha_mask &= tmp[0];
   alpha_mask &= tmp[1];
   return (alpha_mask != 0xffffffffu);
@@ -135,7 +136,7 @@ static void DispatchAlphaToGreen_NEON(const uint8_t* WEBP_RESTRICT alpha,
 static int ExtractAlpha_NEON(const uint8_t* WEBP_RESTRICT argb, int argb_stride,
                              int width, int height,
                              uint8_t* WEBP_RESTRICT alpha, int alpha_stride) {
-  uint32_t alpha_mask = 0xffffffffu;
+  uint32_t alpha_mask = 0xffu;
   uint8x8_t mask8 = vdup_n_u8(0xff);
   uint32_t tmp[2];
   int i, j;
@@ -157,6 +158,7 @@ static int ExtractAlpha_NEON(const uint8_t* WEBP_RESTRICT argb, int argb_stride,
     alpha += alpha_stride;
   }
   vst1_u8((uint8_t*)tmp, mask8);
+  alpha_mask *= 0x01010101;
   alpha_mask &= tmp[0];
   alpha_mask &= tmp[1];
   return (alpha_mask == 0xffffffffu);

src/dsp/alpha_processing_sse2.c

@@ -26,8 +26,8 @@ static int DispatchAlpha_SSE2(const uint8_t* WEBP_RESTRICT alpha,
   uint32_t alpha_and = 0xff;
   int i, j;
   const __m128i zero = _mm_setzero_si128();
-  const __m128i rgb_mask = _mm_set1_epi32(0xffffff00u);  // to preserve RGB
+  const __m128i rgb_mask = _mm_set1_epi32((int)0xffffff00);  // to preserve RGB
-  const __m128i all_0xff = _mm_set_epi32(0, 0, ~0u, ~0u);
+  const __m128i all_0xff = _mm_set_epi32(0, 0, ~0, ~0);
   __m128i all_alphas = all_0xff;
 
   // We must be able to access 3 extra bytes after the last written byte
@@ -106,8 +106,8 @@ static int ExtractAlpha_SSE2(const uint8_t* WEBP_RESTRICT argb, int argb_stride,
   // value is not 0xff if any of the alpha[] is not equal to 0xff.
   uint32_t alpha_and = 0xff;
   int i, j;
-  const __m128i a_mask = _mm_set1_epi32(0xffu);  // to preserve alpha
+  const __m128i a_mask = _mm_set1_epi32(0xff);  // to preserve alpha
-  const __m128i all_0xff = _mm_set_epi32(0, 0, ~0u, ~0u);
+  const __m128i all_0xff = _mm_set_epi32(0, 0, ~0, ~0);
   __m128i all_alphas = all_0xff;
 
   // We must be able to access 3 extra bytes after the last written byte
@@ -178,7 +178,7 @@ static int ExtractAlpha_SSE2(const uint8_t* WEBP_RESTRICT argb, int argb_stride,
 static void ApplyAlphaMultiply_SSE2(uint8_t* rgba, int alpha_first,
                                     int w, int h, int stride) {
   const __m128i zero = _mm_setzero_si128();
-  const __m128i kMult = _mm_set1_epi16(0x8081u);
+  const __m128i kMult = _mm_set1_epi16((short)0x8081);
   const __m128i kMask = _mm_set_epi16(0, 0xff, 0xff, 0, 0, 0xff, 0xff, 0);
   const int kSpan = 4;
   while (h-- > 0) {
@@ -267,7 +267,7 @@ static int HasAlpha32b_SSE2(const uint8_t* src, int length) {
 }
 
 static void AlphaReplace_SSE2(uint32_t* src, int length, uint32_t color) {
-  const __m128i m_color = _mm_set1_epi32(color);
+  const __m128i m_color = _mm_set1_epi32((int)color);
   const __m128i zero = _mm_setzero_si128();
   int i = 0;
   for (; i + 8 <= length; i += 8) {

src/dsp/alpha_processing_sse41.c

@@ -26,7 +26,7 @@ static int ExtractAlpha_SSE41(const uint8_t* WEBP_RESTRICT argb,
   // value is not 0xff if any of the alpha[] is not equal to 0xff.
   uint32_t alpha_and = 0xff;
   int i, j;
-  const __m128i all_0xff = _mm_set1_epi32(~0u);
+  const __m128i all_0xff = _mm_set1_epi32(~0);
   __m128i all_alphas = all_0xff;
 
   // We must be able to access 3 extra bytes after the last written byte

src/dsp/cpu.c

@@ -11,7 +11,7 @@
 //
 // Author: Christian Duvivier (cduvivier@google.com)
 
-#include "src/dsp/dsp.h"
+#include "src/dsp/cpu.h"
 
 #if defined(WEBP_HAVE_NEON_RTCD)
 #include <stdio.h>
@@ -212,7 +212,7 @@ VP8CPUInfo VP8GetCPUInfo = wasmCPUInfo;
 #elif defined(WEBP_HAVE_NEON)
 // In most cases this function doesn't check for NEON support (it's assumed by
 // the configuration), but enables turning off NEON at runtime, for testing
-// purposes, by setting VP8DecGetCPUInfo = NULL.
+// purposes, by setting VP8GetCPUInfo = NULL.
 static int armCPUInfo(CPUFeature feature) {
   if (feature != kNEON) return 0;
 #if defined(__linux__) && defined(WEBP_HAVE_NEON_RTCD)

src/dsp/cpu.h

@@ -0,0 +1,256 @@
+// Copyright 2022 Google Inc. All Rights Reserved.
+//
+// Use of this source code is governed by a BSD-style license
+// that can be found in the COPYING file in the root of the source
+// tree. An additional intellectual property rights grant can be found
+// in the file PATENTS. All contributing project authors may
+// be found in the AUTHORS file in the root of the source tree.
+// -----------------------------------------------------------------------------
+//
+//   CPU detection functions and macros.
+//
+// Author: Skal (pascal.massimino@gmail.com)
+
+#ifndef WEBP_DSP_CPU_H_
+#define WEBP_DSP_CPU_H_
+
+#include <stddef.h>
+
+#ifdef HAVE_CONFIG_H
+#include "src/webp/config.h"
+#endif
+
+#include "src/webp/types.h"
+
+#if defined(__GNUC__)
+#define LOCAL_GCC_VERSION ((__GNUC__ << 8) | __GNUC_MINOR__)
+#define LOCAL_GCC_PREREQ(maj, min) (LOCAL_GCC_VERSION >= (((maj) << 8) | (min)))
+#else
+#define LOCAL_GCC_VERSION 0
+#define LOCAL_GCC_PREREQ(maj, min) 0
+#endif
+
+#if defined(__clang__)
+#define LOCAL_CLANG_VERSION ((__clang_major__ << 8) | __clang_minor__)
+#define LOCAL_CLANG_PREREQ(maj, min) \
+  (LOCAL_CLANG_VERSION >= (((maj) << 8) | (min)))
+#else
+#define LOCAL_CLANG_VERSION 0
+#define LOCAL_CLANG_PREREQ(maj, min) 0
+#endif
+
+#ifndef __has_builtin
+#define __has_builtin(x) 0
+#endif
+
+#if !defined(HAVE_CONFIG_H)
+#if defined(_MSC_VER) && _MSC_VER > 1310 && \
+    (defined(_M_X64) || defined(_M_IX86))
+#define WEBP_MSC_SSE2  // Visual C++ SSE2 targets
+#endif
+
+#if defined(_MSC_VER) && _MSC_VER >= 1500 && \
+    (defined(_M_X64) || defined(_M_IX86))
+#define WEBP_MSC_SSE41  // Visual C++ SSE4.1 targets
+#endif
+#endif
+
+// WEBP_HAVE_* are used to indicate the presence of the instruction set in dsp
+// files without intrinsics, allowing the corresponding Init() to be called.
+// Files containing intrinsics will need to be built targeting the instruction
+// set so should succeed on one of the earlier tests.
+#if (defined(__SSE2__) || defined(WEBP_MSC_SSE2)) && \
+    (!defined(HAVE_CONFIG_H) || defined(WEBP_HAVE_SSE2))
+#define WEBP_USE_SSE2
+#endif
+
+#if defined(WEBP_USE_SSE2) && !defined(WEBP_HAVE_SSE2)
+#define WEBP_HAVE_SSE2
+#endif
+
+#if (defined(__SSE4_1__) || defined(WEBP_MSC_SSE41)) && \
+    (!defined(HAVE_CONFIG_H) || defined(WEBP_HAVE_SSE41))
+#define WEBP_USE_SSE41
+#endif
+
+#if defined(WEBP_USE_SSE41) && !defined(WEBP_HAVE_SSE41)
+#define WEBP_HAVE_SSE41
+#endif
+
+#undef WEBP_MSC_SSE41
+#undef WEBP_MSC_SSE2
+
+// The intrinsics currently cause compiler errors with arm-nacl-gcc and the
+// inline assembly would need to be modified for use with Native Client.
+#if ((defined(__ARM_NEON__) || defined(__aarch64__)) &&       \
+     (!defined(HAVE_CONFIG_H) || defined(WEBP_HAVE_NEON))) && \
+    !defined(__native_client__)
+#define WEBP_USE_NEON
+#endif
+
+#if !defined(WEBP_USE_NEON) && defined(__ANDROID__) && \
+    defined(__ARM_ARCH_7A__) && defined(HAVE_CPU_FEATURES_H)
+#define WEBP_ANDROID_NEON  // Android targets that may have NEON
+#define WEBP_USE_NEON
+#endif
+
+// Note: ARM64 is supported in Visual Studio 2017, but requires the direct
+// inclusion of arm64_neon.h; Visual Studio 2019 includes this file in
+// arm_neon.h. Compile errors were seen with Visual Studio 2019 16.4 with
+// vtbl4_u8(); a fix was made in 16.6.
+#if defined(_MSC_VER) && ((_MSC_VER >= 1700 && defined(_M_ARM)) || \
+                          (_MSC_VER >= 1926 && defined(_M_ARM64)))
+#define WEBP_USE_NEON
+#define WEBP_USE_INTRINSICS
+#endif
+
+#if defined(WEBP_USE_NEON) && !defined(WEBP_HAVE_NEON)
+#define WEBP_HAVE_NEON
+#endif
+
+#if defined(__mips__) && !defined(__mips64) && defined(__mips_isa_rev) && \
+    (__mips_isa_rev >= 1) && (__mips_isa_rev < 6)
+#define WEBP_USE_MIPS32
+#if (__mips_isa_rev >= 2)
+#define WEBP_USE_MIPS32_R2
+#if defined(__mips_dspr2) || (defined(__mips_dsp_rev) && __mips_dsp_rev >= 2)
+#define WEBP_USE_MIPS_DSP_R2
+#endif
+#endif
+#endif
+
+#if defined(__mips_msa) && defined(__mips_isa_rev) && (__mips_isa_rev >= 5)
+#define WEBP_USE_MSA
+#endif
+
+#ifndef WEBP_DSP_OMIT_C_CODE
+#define WEBP_DSP_OMIT_C_CODE 1
+#endif
+
+#if defined(WEBP_USE_NEON) && WEBP_DSP_OMIT_C_CODE
+#define WEBP_NEON_OMIT_C_CODE 1
+#else
+#define WEBP_NEON_OMIT_C_CODE 0
+#endif
+
+#if !(LOCAL_CLANG_PREREQ(3, 8) || LOCAL_GCC_PREREQ(4, 8) || \
+      defined(__aarch64__))
+#define WEBP_NEON_WORK_AROUND_GCC 1
+#else
+#define WEBP_NEON_WORK_AROUND_GCC 0
+#endif
+
+// This macro prevents thread_sanitizer from reporting known concurrent writes.
+#define WEBP_TSAN_IGNORE_FUNCTION
+#if defined(__has_feature)
+#if __has_feature(thread_sanitizer)
+#undef WEBP_TSAN_IGNORE_FUNCTION
+#define WEBP_TSAN_IGNORE_FUNCTION __attribute__((no_sanitize_thread))
+#endif
+#endif
+
+#if defined(__has_feature)
+#if __has_feature(memory_sanitizer)
+#define WEBP_MSAN
+#endif
+#endif
+
+#if defined(WEBP_USE_THREAD) && !defined(_WIN32)
+#include <pthread.h>  // NOLINT
+
+#define WEBP_DSP_INIT(func)                                         \
+  do {                                                              \
+    static volatile VP8CPUInfo func##_last_cpuinfo_used =           \
+        (VP8CPUInfo)&func##_last_cpuinfo_used;                      \
+    static pthread_mutex_t func##_lock = PTHREAD_MUTEX_INITIALIZER; \
+    if (pthread_mutex_lock(&func##_lock)) break;                    \
+    if (func##_last_cpuinfo_used != VP8GetCPUInfo) func();          \
+    func##_last_cpuinfo_used = VP8GetCPUInfo;                       \
+    (void)pthread_mutex_unlock(&func##_lock);                       \
+  } while (0)
+#else  // !(defined(WEBP_USE_THREAD) && !defined(_WIN32))
+#define WEBP_DSP_INIT(func)                               \
+  do {                                                    \
+    static volatile VP8CPUInfo func##_last_cpuinfo_used = \
+        (VP8CPUInfo)&func##_last_cpuinfo_used;            \
+    if (func##_last_cpuinfo_used == VP8GetCPUInfo) break; \
+    func();                                               \
+    func##_last_cpuinfo_used = VP8GetCPUInfo;             \
+  } while (0)
+#endif  // defined(WEBP_USE_THREAD) && !defined(_WIN32)
+
+// Defines an Init + helper function that control multiple initialization of
+// function pointers / tables.
+/* Usage:
+   WEBP_DSP_INIT_FUNC(InitFunc) {
+     ...function body
+   }
+*/
+#define WEBP_DSP_INIT_FUNC(name)                                            \
+  static WEBP_TSAN_IGNORE_FUNCTION void name##_body(void);                  \
+  WEBP_TSAN_IGNORE_FUNCTION void name(void) { WEBP_DSP_INIT(name##_body); } \
+  static WEBP_TSAN_IGNORE_FUNCTION void name##_body(void)
+
+#define WEBP_UBSAN_IGNORE_UNDEF
+#define WEBP_UBSAN_IGNORE_UNSIGNED_OVERFLOW
+#if defined(__clang__) && defined(__has_attribute)
+#if __has_attribute(no_sanitize)
+// This macro prevents the undefined behavior sanitizer from reporting
+// failures. This is only meant to silence unaligned loads on platforms that
+// are known to support them.
+#undef WEBP_UBSAN_IGNORE_UNDEF
+#define WEBP_UBSAN_IGNORE_UNDEF __attribute__((no_sanitize("undefined")))
+
+// This macro prevents the undefined behavior sanitizer from reporting
+// failures related to unsigned integer overflows. This is only meant to
+// silence cases where this well defined behavior is expected.
+#undef WEBP_UBSAN_IGNORE_UNSIGNED_OVERFLOW
+#define WEBP_UBSAN_IGNORE_UNSIGNED_OVERFLOW \
+  __attribute__((no_sanitize("unsigned-integer-overflow")))
+#endif
+#endif
+
+// If 'ptr' is NULL, returns NULL. Otherwise returns 'ptr + off'.
+// Prevents undefined behavior sanitizer nullptr-with-nonzero-offset warning.
+#if !defined(WEBP_OFFSET_PTR)
+#define WEBP_OFFSET_PTR(ptr, off) (((ptr) == NULL) ? NULL : ((ptr) + (off)))
+#endif
+
+// Regularize the definition of WEBP_SWAP_16BIT_CSP (backward compatibility)
+#if !defined(WEBP_SWAP_16BIT_CSP)
+#define WEBP_SWAP_16BIT_CSP 0
+#endif
+
+// some endian fix (e.g.: mips-gcc doesn't define __BIG_ENDIAN__)
+#if !defined(WORDS_BIGENDIAN) &&                   \
+    (defined(__BIG_ENDIAN__) || defined(_M_PPC) || \
+     (defined(__BYTE_ORDER__) && (__BYTE_ORDER__ == __ORDER_BIG_ENDIAN__)))
+#define WORDS_BIGENDIAN
+#endif
+
+typedef enum {
+  kSSE2,
+  kSSE3,
+  kSlowSSSE3,  // special feature for slow SSSE3 architectures
+  kSSE4_1,
+  kAVX,
+  kAVX2,
+  kNEON,
+  kMIPS32,
+  kMIPSdspR2,
+  kMSA
+} CPUFeature;
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+// returns true if the CPU supports the feature.
+typedef int (*VP8CPUInfo)(CPUFeature feature);
+WEBP_EXTERN VP8CPUInfo VP8GetCPUInfo;
+
+#ifdef __cplusplus
+}    // extern "C"
+#endif
+
+#endif  // WEBP_DSP_CPU_H_

src/dsp/dec_sse2.c

@@ -158,10 +158,10 @@ static void Transform_SSE2(const int16_t* in, uint8_t* dst, int do_two) {
       dst3 = _mm_loadl_epi64((__m128i*)(dst + 3 * BPS));
     } else {
       // Load four bytes/pixels per line.
-      dst0 = _mm_cvtsi32_si128(WebPMemToUint32(dst + 0 * BPS));
+      dst0 = _mm_cvtsi32_si128(WebPMemToInt32(dst + 0 * BPS));
-      dst1 = _mm_cvtsi32_si128(WebPMemToUint32(dst + 1 * BPS));
+      dst1 = _mm_cvtsi32_si128(WebPMemToInt32(dst + 1 * BPS));
-      dst2 = _mm_cvtsi32_si128(WebPMemToUint32(dst + 2 * BPS));
+      dst2 = _mm_cvtsi32_si128(WebPMemToInt32(dst + 2 * BPS));
-      dst3 = _mm_cvtsi32_si128(WebPMemToUint32(dst + 3 * BPS));
+      dst3 = _mm_cvtsi32_si128(WebPMemToInt32(dst + 3 * BPS));
     }
     // Convert to 16b.
     dst0 = _mm_unpacklo_epi8(dst0, zero);
@@ -187,10 +187,10 @@ static void Transform_SSE2(const int16_t* in, uint8_t* dst, int do_two) {
       _mm_storel_epi64((__m128i*)(dst + 3 * BPS), dst3);
     } else {
       // Store four bytes/pixels per line.
-      WebPUint32ToMem(dst + 0 * BPS, _mm_cvtsi128_si32(dst0));
+      WebPInt32ToMem(dst + 0 * BPS, _mm_cvtsi128_si32(dst0));
-      WebPUint32ToMem(dst + 1 * BPS, _mm_cvtsi128_si32(dst1));
+      WebPInt32ToMem(dst + 1 * BPS, _mm_cvtsi128_si32(dst1));
-      WebPUint32ToMem(dst + 2 * BPS, _mm_cvtsi128_si32(dst2));
+      WebPInt32ToMem(dst + 2 * BPS, _mm_cvtsi128_si32(dst2));
-      WebPUint32ToMem(dst + 3 * BPS, _mm_cvtsi128_si32(dst3));
+      WebPInt32ToMem(dst + 3 * BPS, _mm_cvtsi128_si32(dst3));
     }
   }
 }
@@ -213,10 +213,10 @@ static void TransformAC3(const int16_t* in, uint8_t* dst) {
   const __m128i m3 = _mm_subs_epi16(B, d4);
   const __m128i zero = _mm_setzero_si128();
   // Load the source pixels.
-  __m128i dst0 = _mm_cvtsi32_si128(WebPMemToUint32(dst + 0 * BPS));
+  __m128i dst0 = _mm_cvtsi32_si128(WebPMemToInt32(dst + 0 * BPS));
-  __m128i dst1 = _mm_cvtsi32_si128(WebPMemToUint32(dst + 1 * BPS));
+  __m128i dst1 = _mm_cvtsi32_si128(WebPMemToInt32(dst + 1 * BPS));
-  __m128i dst2 = _mm_cvtsi32_si128(WebPMemToUint32(dst + 2 * BPS));
+  __m128i dst2 = _mm_cvtsi32_si128(WebPMemToInt32(dst + 2 * BPS));
-  __m128i dst3 = _mm_cvtsi32_si128(WebPMemToUint32(dst + 3 * BPS));
+  __m128i dst3 = _mm_cvtsi32_si128(WebPMemToInt32(dst + 3 * BPS));
   // Convert to 16b.
   dst0 = _mm_unpacklo_epi8(dst0, zero);
   dst1 = _mm_unpacklo_epi8(dst1, zero);
@@ -233,10 +233,10 @@ static void TransformAC3(const int16_t* in, uint8_t* dst) {
   dst2 = _mm_packus_epi16(dst2, dst2);
   dst3 = _mm_packus_epi16(dst3, dst3);
   // Store the results.
-  WebPUint32ToMem(dst + 0 * BPS, _mm_cvtsi128_si32(dst0));
+  WebPInt32ToMem(dst + 0 * BPS, _mm_cvtsi128_si32(dst0));
-  WebPUint32ToMem(dst + 1 * BPS, _mm_cvtsi128_si32(dst1));
+  WebPInt32ToMem(dst + 1 * BPS, _mm_cvtsi128_si32(dst1));
-  WebPUint32ToMem(dst + 2 * BPS, _mm_cvtsi128_si32(dst2));
+  WebPInt32ToMem(dst + 2 * BPS, _mm_cvtsi128_si32(dst2));
-  WebPUint32ToMem(dst + 3 * BPS, _mm_cvtsi128_si32(dst3));
+  WebPInt32ToMem(dst + 3 * BPS, _mm_cvtsi128_si32(dst3));
 }
 #undef MUL
 #endif   // USE_TRANSFORM_AC3
@@ -477,11 +477,11 @@ static WEBP_INLINE void Load8x4_SSE2(const uint8_t* const b, int stride,
   // A0 = 63 62 61 60 23 22 21 20 43 42 41 40 03 02 01 00
   // A1 = 73 72 71 70 33 32 31 30 53 52 51 50 13 12 11 10
   const __m128i A0 = _mm_set_epi32(
-      WebPMemToUint32(&b[6 * stride]), WebPMemToUint32(&b[2 * stride]),
+      WebPMemToInt32(&b[6 * stride]), WebPMemToInt32(&b[2 * stride]),
-      WebPMemToUint32(&b[4 * stride]), WebPMemToUint32(&b[0 * stride]));
+      WebPMemToInt32(&b[4 * stride]), WebPMemToInt32(&b[0 * stride]));
   const __m128i A1 = _mm_set_epi32(
-      WebPMemToUint32(&b[7 * stride]), WebPMemToUint32(&b[3 * stride]),
+      WebPMemToInt32(&b[7 * stride]), WebPMemToInt32(&b[3 * stride]),
-      WebPMemToUint32(&b[5 * stride]), WebPMemToUint32(&b[1 * stride]));
+      WebPMemToInt32(&b[5 * stride]), WebPMemToInt32(&b[1 * stride]));
 
   // B0 = 53 43 52 42 51 41 50 40 13 03 12 02 11 01 10 00
   // B1 = 73 63 72 62 71 61 70 60 33 23 32 22 31 21 30 20
@@ -540,7 +540,7 @@ static WEBP_INLINE void Store4x4_SSE2(__m128i* const x,
                                       uint8_t* dst, int stride) {
   int i;
   for (i = 0; i < 4; ++i, dst += stride) {
-    WebPUint32ToMem(dst, _mm_cvtsi128_si32(*x));
+    WebPInt32ToMem(dst, _mm_cvtsi128_si32(*x));
     *x = _mm_srli_si128(*x, 4);
   }
 }
@@ -908,10 +908,10 @@ static void VE4_SSE2(uint8_t* dst) {    // vertical
   const __m128i lsb = _mm_and_si128(_mm_xor_si128(ABCDEFGH, CDEFGH00), one);
   const __m128i b = _mm_subs_epu8(a, lsb);
   const __m128i avg = _mm_avg_epu8(b, BCDEFGH0);
-  const uint32_t vals = _mm_cvtsi128_si32(avg);
+  const int vals = _mm_cvtsi128_si32(avg);
   int i;
   for (i = 0; i < 4; ++i) {
-    WebPUint32ToMem(dst + i * BPS, vals);
+    WebPInt32ToMem(dst + i * BPS, vals);
   }
 }
 
@@ -925,10 +925,10 @@ static void LD4_SSE2(uint8_t* dst) {   // Down-Left
   const __m128i lsb = _mm_and_si128(_mm_xor_si128(ABCDEFGH, CDEFGHH0), one);
   const __m128i avg2 = _mm_subs_epu8(avg1, lsb);
   const __m128i abcdefg = _mm_avg_epu8(avg2, BCDEFGH0);
-  WebPUint32ToMem(dst + 0 * BPS, _mm_cvtsi128_si32(               abcdefg    ));
+  WebPInt32ToMem(dst + 0 * BPS, _mm_cvtsi128_si32(               abcdefg    ));
-  WebPUint32ToMem(dst + 1 * BPS, _mm_cvtsi128_si32(_mm_srli_si128(abcdefg, 1)));
+  WebPInt32ToMem(dst + 1 * BPS, _mm_cvtsi128_si32(_mm_srli_si128(abcdefg, 1)));
-  WebPUint32ToMem(dst + 2 * BPS, _mm_cvtsi128_si32(_mm_srli_si128(abcdefg, 2)));
+  WebPInt32ToMem(dst + 2 * BPS, _mm_cvtsi128_si32(_mm_srli_si128(abcdefg, 2)));
-  WebPUint32ToMem(dst + 3 * BPS, _mm_cvtsi128_si32(_mm_srli_si128(abcdefg, 3)));
+  WebPInt32ToMem(dst + 3 * BPS, _mm_cvtsi128_si32(_mm_srli_si128(abcdefg, 3)));
 }
 
 static void VR4_SSE2(uint8_t* dst) {   // Vertical-Right
@@ -946,10 +946,10 @@ static void VR4_SSE2(uint8_t* dst) {   // Vertical-Right
   const __m128i lsb = _mm_and_si128(_mm_xor_si128(IXABCD, ABCD0), one);
   const __m128i avg2 = _mm_subs_epu8(avg1, lsb);
   const __m128i efgh = _mm_avg_epu8(avg2, XABCD);
-  WebPUint32ToMem(dst + 0 * BPS, _mm_cvtsi128_si32(               abcd    ));
+  WebPInt32ToMem(dst + 0 * BPS, _mm_cvtsi128_si32(               abcd    ));
-  WebPUint32ToMem(dst + 1 * BPS, _mm_cvtsi128_si32(               efgh    ));
+  WebPInt32ToMem(dst + 1 * BPS, _mm_cvtsi128_si32(               efgh    ));
-  WebPUint32ToMem(dst + 2 * BPS, _mm_cvtsi128_si32(_mm_slli_si128(abcd, 1)));
+  WebPInt32ToMem(dst + 2 * BPS, _mm_cvtsi128_si32(_mm_slli_si128(abcd, 1)));
-  WebPUint32ToMem(dst + 3 * BPS, _mm_cvtsi128_si32(_mm_slli_si128(efgh, 1)));
+  WebPInt32ToMem(dst + 3 * BPS, _mm_cvtsi128_si32(_mm_slli_si128(efgh, 1)));
 
   // these two are hard to implement in SSE2, so we keep the C-version:
   DST(0, 2) = AVG3(J, I, X);
@@ -970,11 +970,12 @@ static void VL4_SSE2(uint8_t* dst) {   // Vertical-Left
   const __m128i abbc = _mm_or_si128(ab, bc);
   const __m128i lsb2 = _mm_and_si128(abbc, lsb1);
   const __m128i avg4 = _mm_subs_epu8(avg3, lsb2);
-  const uint32_t extra_out = _mm_cvtsi128_si32(_mm_srli_si128(avg4, 4));
+  const uint32_t extra_out =
-  WebPUint32ToMem(dst + 0 * BPS, _mm_cvtsi128_si32(               avg1    ));
+      (uint32_t)_mm_cvtsi128_si32(_mm_srli_si128(avg4, 4));
-  WebPUint32ToMem(dst + 1 * BPS, _mm_cvtsi128_si32(               avg4    ));
+  WebPInt32ToMem(dst + 0 * BPS, _mm_cvtsi128_si32(               avg1    ));
-  WebPUint32ToMem(dst + 2 * BPS, _mm_cvtsi128_si32(_mm_srli_si128(avg1, 1)));
+  WebPInt32ToMem(dst + 1 * BPS, _mm_cvtsi128_si32(               avg4    ));
-  WebPUint32ToMem(dst + 3 * BPS, _mm_cvtsi128_si32(_mm_srli_si128(avg4, 1)));
+  WebPInt32ToMem(dst + 2 * BPS, _mm_cvtsi128_si32(_mm_srli_si128(avg1, 1)));
+  WebPInt32ToMem(dst + 3 * BPS, _mm_cvtsi128_si32(_mm_srli_si128(avg4, 1)));
 
   // these two are hard to get and irregular
   DST(3, 2) = (extra_out >> 0) & 0xff;
@@ -990,7 +991,7 @@ static void RD4_SSE2(uint8_t* dst) {   // Down-right
   const uint32_t K = dst[-1 + 2 * BPS];
   const uint32_t L = dst[-1 + 3 * BPS];
   const __m128i LKJI_____ =
-      _mm_cvtsi32_si128(L | (K << 8) | (J << 16) | (I << 24));
+      _mm_cvtsi32_si128((int)(L | (K << 8) | (J << 16) | (I << 24)));
   const __m128i LKJIXABCD = _mm_or_si128(LKJI_____, ____XABCD);
   const __m128i KJIXABCD_ = _mm_srli_si128(LKJIXABCD, 1);
   const __m128i JIXABCD__ = _mm_srli_si128(LKJIXABCD, 2);
@@ -998,10 +999,10 @@ static void RD4_SSE2(uint8_t* dst) {   // Down-right
   const __m128i lsb = _mm_and_si128(_mm_xor_si128(JIXABCD__, LKJIXABCD), one);
   const __m128i avg2 = _mm_subs_epu8(avg1, lsb);
   const __m128i abcdefg = _mm_avg_epu8(avg2, KJIXABCD_);
-  WebPUint32ToMem(dst + 3 * BPS, _mm_cvtsi128_si32(               abcdefg    ));
+  WebPInt32ToMem(dst + 3 * BPS, _mm_cvtsi128_si32(               abcdefg    ));
-  WebPUint32ToMem(dst + 2 * BPS, _mm_cvtsi128_si32(_mm_srli_si128(abcdefg, 1)));
+  WebPInt32ToMem(dst + 2 * BPS, _mm_cvtsi128_si32(_mm_srli_si128(abcdefg, 1)));
-  WebPUint32ToMem(dst + 1 * BPS, _mm_cvtsi128_si32(_mm_srli_si128(abcdefg, 2)));
+  WebPInt32ToMem(dst + 1 * BPS, _mm_cvtsi128_si32(_mm_srli_si128(abcdefg, 2)));
-  WebPUint32ToMem(dst + 0 * BPS, _mm_cvtsi128_si32(_mm_srli_si128(abcdefg, 3)));
+  WebPInt32ToMem(dst + 0 * BPS, _mm_cvtsi128_si32(_mm_srli_si128(abcdefg, 3)));
 }
 
 #undef DST
@@ -1015,13 +1016,13 @@ static WEBP_INLINE void TrueMotion_SSE2(uint8_t* dst, int size) {
   const __m128i zero = _mm_setzero_si128();
   int y;
   if (size == 4) {
-    const __m128i top_values = _mm_cvtsi32_si128(WebPMemToUint32(top));
+    const __m128i top_values = _mm_cvtsi32_si128(WebPMemToInt32(top));
     const __m128i top_base = _mm_unpacklo_epi8(top_values, zero);
     for (y = 0; y < 4; ++y, dst += BPS) {
       const int val = dst[-1] - top[-1];
       const __m128i base = _mm_set1_epi16(val);
       const __m128i out = _mm_packus_epi16(_mm_add_epi16(base, top_base), zero);
-      WebPUint32ToMem(dst, _mm_cvtsi128_si32(out));
+      WebPInt32ToMem(dst, _mm_cvtsi128_si32(out));
     }
   } else if (size == 8) {
     const __m128i top_values = _mm_loadl_epi64((const __m128i*)top);
@@ -1062,7 +1063,7 @@ static void VE16_SSE2(uint8_t* dst) {
 static void HE16_SSE2(uint8_t* dst) {     // horizontal
   int j;
   for (j = 16; j > 0; --j) {
-    const __m128i values = _mm_set1_epi8(dst[-1]);
+    const __m128i values = _mm_set1_epi8((char)dst[-1]);
     _mm_storeu_si128((__m128i*)dst, values);
     dst += BPS;
   }
@@ -1070,7 +1071,7 @@ static void HE16_SSE2(uint8_t* dst) {     // horizontal
 
 static WEBP_INLINE void Put16_SSE2(uint8_t v, uint8_t* dst) {
   int j;
-  const __m128i values = _mm_set1_epi8(v);
+  const __m128i values = _mm_set1_epi8((char)v);
   for (j = 0; j < 16; ++j) {
     _mm_storeu_si128((__m128i*)(dst + j * BPS), values);
   }
@@ -1130,7 +1131,7 @@ static void VE8uv_SSE2(uint8_t* dst) {    // vertical
 // helper for chroma-DC predictions
 static WEBP_INLINE void Put8x8uv_SSE2(uint8_t v, uint8_t* dst) {
   int j;
-  const __m128i values = _mm_set1_epi8(v);
+  const __m128i values = _mm_set1_epi8((char)v);
   for (j = 0; j < 8; ++j) {
     _mm_storel_epi64((__m128i*)(dst + j * BPS), values);
   }

src/dsp/dec_sse41.c

@@ -23,7 +23,7 @@ static void HE16_SSE41(uint8_t* dst) {     // horizontal
   int j;
   const __m128i kShuffle3 = _mm_set1_epi8(3);
   for (j = 16; j > 0; --j) {
-    const __m128i in = _mm_cvtsi32_si128(WebPMemToUint32(dst - 4));
+    const __m128i in = _mm_cvtsi32_si128(WebPMemToInt32(dst - 4));
     const __m128i values = _mm_shuffle_epi8(in, kShuffle3);
     _mm_storeu_si128((__m128i*)dst, values);
     dst += BPS;

src/dsp/dsp.h

@@ -18,6 +18,7 @@
 #include "src/webp/config.h"
 #endif
 
+#include "src/dsp/cpu.h"
 #include "src/webp/types.h"
 
 #ifdef __cplusplus
@@ -43,225 +44,6 @@ extern "C" {
 #define WEBP_RESTRICT
 #endif
 
-//------------------------------------------------------------------------------
-// CPU detection
-
-#if defined(__GNUC__)
-# define LOCAL_GCC_VERSION ((__GNUC__ << 8) | __GNUC_MINOR__)
-# define LOCAL_GCC_PREREQ(maj, min) \
-    (LOCAL_GCC_VERSION >= (((maj) << 8) | (min)))
-#else
-# define LOCAL_GCC_VERSION 0
-# define LOCAL_GCC_PREREQ(maj, min) 0
-#endif
-
-#if defined(__clang__)
-# define LOCAL_CLANG_VERSION ((__clang_major__ << 8) | __clang_minor__)
-# define LOCAL_CLANG_PREREQ(maj, min) \
-    (LOCAL_CLANG_VERSION >= (((maj) << 8) | (min)))
-#else
-# define LOCAL_CLANG_VERSION 0
-# define LOCAL_CLANG_PREREQ(maj, min) 0
-#endif
-
-#ifndef __has_builtin
-# define __has_builtin(x) 0
-#endif
-
-#if !defined(HAVE_CONFIG_H)
-#if defined(_MSC_VER) && _MSC_VER > 1310 && \
-    (defined(_M_X64) || defined(_M_IX86))
-#define WEBP_MSC_SSE2  // Visual C++ SSE2 targets
-#endif
-
-#if defined(_MSC_VER) && _MSC_VER >= 1500 && \
-    (defined(_M_X64) || defined(_M_IX86))
-#define WEBP_MSC_SSE41  // Visual C++ SSE4.1 targets
-#endif
-#endif
-
-// WEBP_HAVE_* are used to indicate the presence of the instruction set in dsp
-// files without intrinsics, allowing the corresponding Init() to be called.
-// Files containing intrinsics will need to be built targeting the instruction
-// set so should succeed on one of the earlier tests.
-#if (defined(__SSE2__) || defined(WEBP_MSC_SSE2)) && \
-    (!defined(HAVE_CONFIG_H) || defined(WEBP_HAVE_SSE2))
-#define WEBP_USE_SSE2
-#endif
-
-#if defined(WEBP_USE_SSE2) && !defined(WEBP_HAVE_SSE2)
-#define WEBP_HAVE_SSE2
-#endif
-
-#if (defined(__SSE4_1__) || defined(WEBP_MSC_SSE41)) && \
-    (!defined(HAVE_CONFIG_H) || defined(WEBP_HAVE_SSE41))
-#define WEBP_USE_SSE41
-#endif
-
-#if defined(WEBP_USE_SSE41) && !defined(WEBP_HAVE_SSE41)
-#define WEBP_HAVE_SSE41
-#endif
-
-#undef WEBP_MSC_SSE41
-#undef WEBP_MSC_SSE2
-
-// The intrinsics currently cause compiler errors with arm-nacl-gcc and the
-// inline assembly would need to be modified for use with Native Client.
-#if ((defined(__ARM_NEON__) || defined(__aarch64__)) && \
-     (!defined(HAVE_CONFIG_H) || defined(WEBP_HAVE_NEON))) && \
-    !defined(__native_client__)
-#define WEBP_USE_NEON
-#endif
-
-#if !defined(WEBP_USE_NEON) && defined(__ANDROID__) && \
-    defined(__ARM_ARCH_7A__) && defined(HAVE_CPU_FEATURES_H)
-#define WEBP_ANDROID_NEON  // Android targets that may have NEON
-#define WEBP_USE_NEON
-#endif
-
-// Note: ARM64 is supported in Visual Studio 2017, but requires the direct
-// inclusion of arm64_neon.h; Visual Studio 2019 includes this file in
-// arm_neon.h.
-#if defined(_MSC_VER) && \
-  ((_MSC_VER >= 1700 && defined(_M_ARM)) || \
-   (_MSC_VER >= 1920 && defined(_M_ARM64)))
-#define WEBP_USE_NEON
-#define WEBP_USE_INTRINSICS
-#endif
-
-#if defined(WEBP_USE_NEON) && !defined(WEBP_HAVE_NEON)
-#define WEBP_HAVE_NEON
-#endif
-
-#if defined(__mips__) && !defined(__mips64) && \
-    defined(__mips_isa_rev) && (__mips_isa_rev >= 1) && (__mips_isa_rev < 6)
-#define WEBP_USE_MIPS32
-#if (__mips_isa_rev >= 2)
-#define WEBP_USE_MIPS32_R2
-#if defined(__mips_dspr2) || (defined(__mips_dsp_rev) && __mips_dsp_rev >= 2)
-#define WEBP_USE_MIPS_DSP_R2
-#endif
-#endif
-#endif
-
-#if defined(__mips_msa) && defined(__mips_isa_rev) && (__mips_isa_rev >= 5)
-#define WEBP_USE_MSA
-#endif
-
-#ifndef WEBP_DSP_OMIT_C_CODE
-#define WEBP_DSP_OMIT_C_CODE 1
-#endif
-
-#if defined(WEBP_USE_NEON) && WEBP_DSP_OMIT_C_CODE
-#define WEBP_NEON_OMIT_C_CODE 1
-#else
-#define WEBP_NEON_OMIT_C_CODE 0
-#endif
-
-#if !(LOCAL_CLANG_PREREQ(3,8) || LOCAL_GCC_PREREQ(4,8) || defined(__aarch64__))
-#define WEBP_NEON_WORK_AROUND_GCC 1
-#else
-#define WEBP_NEON_WORK_AROUND_GCC 0
-#endif
-
-// This macro prevents thread_sanitizer from reporting known concurrent writes.
-#define WEBP_TSAN_IGNORE_FUNCTION
-#if defined(__has_feature)
-#if __has_feature(thread_sanitizer)
-#undef WEBP_TSAN_IGNORE_FUNCTION
-#define WEBP_TSAN_IGNORE_FUNCTION __attribute__((no_sanitize_thread))
-#endif
-#endif
-
-#if defined(WEBP_USE_THREAD) && !defined(_WIN32)
-#include <pthread.h>  // NOLINT
-
-#define WEBP_DSP_INIT(func) do {                                    \
-  static volatile VP8CPUInfo func ## _last_cpuinfo_used =           \
-      (VP8CPUInfo)&func ## _last_cpuinfo_used;                      \
-  static pthread_mutex_t func ## _lock = PTHREAD_MUTEX_INITIALIZER; \
-  if (pthread_mutex_lock(&func ## _lock)) break;                    \
-  if (func ## _last_cpuinfo_used != VP8GetCPUInfo) func();          \
-  func ## _last_cpuinfo_used = VP8GetCPUInfo;                       \
-  (void)pthread_mutex_unlock(&func ## _lock);                       \
-} while (0)
-#else  // !(defined(WEBP_USE_THREAD) && !defined(_WIN32))
-#define WEBP_DSP_INIT(func) do {                                    \
-  static volatile VP8CPUInfo func ## _last_cpuinfo_used =           \
-      (VP8CPUInfo)&func ## _last_cpuinfo_used;                      \
-  if (func ## _last_cpuinfo_used == VP8GetCPUInfo) break;           \
-  func();                                                           \
-  func ## _last_cpuinfo_used = VP8GetCPUInfo;                       \
-} while (0)
-#endif  // defined(WEBP_USE_THREAD) && !defined(_WIN32)
-
-// Defines an Init + helper function that control multiple initialization of
-// function pointers / tables.
-/* Usage:
-   WEBP_DSP_INIT_FUNC(InitFunc) {
-     ...function body
-   }
-*/
-#define WEBP_DSP_INIT_FUNC(name)                             \
-  static WEBP_TSAN_IGNORE_FUNCTION void name ## _body(void); \
-  WEBP_TSAN_IGNORE_FUNCTION void name(void) {                \
-    WEBP_DSP_INIT(name ## _body);                            \
-  }                                                          \
-  static WEBP_TSAN_IGNORE_FUNCTION void name ## _body(void)
-
-#define WEBP_UBSAN_IGNORE_UNDEF
-#define WEBP_UBSAN_IGNORE_UNSIGNED_OVERFLOW
-#if defined(__clang__) && defined(__has_attribute)
-#if __has_attribute(no_sanitize)
-// This macro prevents the undefined behavior sanitizer from reporting
-// failures. This is only meant to silence unaligned loads on platforms that
-// are known to support them.
-#undef WEBP_UBSAN_IGNORE_UNDEF
-#define WEBP_UBSAN_IGNORE_UNDEF \
-  __attribute__((no_sanitize("undefined")))
-
-// This macro prevents the undefined behavior sanitizer from reporting
-// failures related to unsigned integer overflows. This is only meant to
-// silence cases where this well defined behavior is expected.
-#undef WEBP_UBSAN_IGNORE_UNSIGNED_OVERFLOW
-#define WEBP_UBSAN_IGNORE_UNSIGNED_OVERFLOW \
-  __attribute__((no_sanitize("unsigned-integer-overflow")))
-#endif
-#endif
-
-// If 'ptr' is NULL, returns NULL. Otherwise returns 'ptr + off'.
-// Prevents undefined behavior sanitizer nullptr-with-nonzero-offset warning.
-#if !defined(WEBP_OFFSET_PTR)
-#define WEBP_OFFSET_PTR(ptr, off) (((ptr) == NULL) ? NULL : ((ptr) + (off)))
-#endif
-
-// Regularize the definition of WEBP_SWAP_16BIT_CSP (backward compatibility)
-#if !defined(WEBP_SWAP_16BIT_CSP)
-#define WEBP_SWAP_16BIT_CSP 0
-#endif
-
-// some endian fix (e.g.: mips-gcc doesn't define __BIG_ENDIAN__)
-#if !defined(WORDS_BIGENDIAN) && \
-    (defined(__BIG_ENDIAN__) || defined(_M_PPC) || \
-     (defined(__BYTE_ORDER__) && (__BYTE_ORDER__ == __ORDER_BIG_ENDIAN__)))
-#define WORDS_BIGENDIAN
-#endif
-
-typedef enum {
-  kSSE2,
-  kSSE3,
-  kSlowSSSE3,  // special feature for slow SSSE3 architectures
-  kSSE4_1,
-  kAVX,
-  kAVX2,
-  kNEON,
-  kMIPS32,
-  kMIPSdspR2,
-  kMSA
-} CPUFeature;
-// returns true if the CPU supports the feature.
-typedef int (*VP8CPUInfo)(CPUFeature feature);
-WEBP_EXTERN VP8CPUInfo VP8GetCPUInfo;
 
 //------------------------------------------------------------------------------
 // Init stub generator
@@ -550,15 +332,6 @@ extern void WebPConvertARGBToUV_C(const uint32_t* argb, uint8_t* u, uint8_t* v,
 extern void WebPConvertRGBA32ToUV_C(const uint16_t* rgb,
                                     uint8_t* u, uint8_t* v, int width);
 
-// utilities for accurate RGB->YUV conversion
-extern uint64_t (*WebPSharpYUVUpdateY)(const uint16_t* src, const uint16_t* ref,
-                                       uint16_t* dst, int len);
-extern void (*WebPSharpYUVUpdateRGB)(const int16_t* src, const int16_t* ref,
-                                     int16_t* dst, int len);
-extern void (*WebPSharpYUVFilterRow)(const int16_t* A, const int16_t* B,
-                                     int len,
-                                     const uint16_t* best_y, uint16_t* out);
-
 // Must be called before using the above.
 void WebPInitConvertARGBToYUV(void);
 

src/dsp/enc_neon.c

@@ -764,9 +764,14 @@ static WEBP_INLINE void AccumulateSSE16_NEON(const uint8_t* const a,
 
 // Horizontal sum of all four uint32_t values in 'sum'.
 static int SumToInt_NEON(uint32x4_t sum) {
+#if defined(__aarch64__)
+  return (int)vaddvq_u32(sum);
+#else
   const uint64x2_t sum2 = vpaddlq_u32(sum);
-  const uint64_t sum3 = vgetq_lane_u64(sum2, 0) + vgetq_lane_u64(sum2, 1);
+  const uint32x2_t sum3 = vadd_u32(vreinterpret_u32_u64(vget_low_u64(sum2)),
-  return (int)sum3;
+                                   vreinterpret_u32_u64(vget_high_u64(sum2)));
+  return (int)vget_lane_u32(sum3, 0);
+#endif
 }
 
 static int SSE16x16_NEON(const uint8_t* a, const uint8_t* b) {

src/dsp/enc_sse2.c

@@ -156,10 +156,10 @@ static void ITransform_SSE2(const uint8_t* ref, const int16_t* in, uint8_t* dst,
       ref3 = _mm_loadl_epi64((const __m128i*)&ref[3 * BPS]);
     } else {
       // Load four bytes/pixels per line.
-      ref0 = _mm_cvtsi32_si128(WebPMemToUint32(&ref[0 * BPS]));
+      ref0 = _mm_cvtsi32_si128(WebPMemToInt32(&ref[0 * BPS]));
-      ref1 = _mm_cvtsi32_si128(WebPMemToUint32(&ref[1 * BPS]));
+      ref1 = _mm_cvtsi32_si128(WebPMemToInt32(&ref[1 * BPS]));
-      ref2 = _mm_cvtsi32_si128(WebPMemToUint32(&ref[2 * BPS]));
+      ref2 = _mm_cvtsi32_si128(WebPMemToInt32(&ref[2 * BPS]));
-      ref3 = _mm_cvtsi32_si128(WebPMemToUint32(&ref[3 * BPS]));
+      ref3 = _mm_cvtsi32_si128(WebPMemToInt32(&ref[3 * BPS]));
     }
     // Convert to 16b.
     ref0 = _mm_unpacklo_epi8(ref0, zero);
@@ -185,10 +185,10 @@ static void ITransform_SSE2(const uint8_t* ref, const int16_t* in, uint8_t* dst,
       _mm_storel_epi64((__m128i*)&dst[3 * BPS], ref3);
     } else {
       // Store four bytes/pixels per line.
-      WebPUint32ToMem(&dst[0 * BPS], _mm_cvtsi128_si32(ref0));
+      WebPInt32ToMem(&dst[0 * BPS], _mm_cvtsi128_si32(ref0));
-      WebPUint32ToMem(&dst[1 * BPS], _mm_cvtsi128_si32(ref1));
+      WebPInt32ToMem(&dst[1 * BPS], _mm_cvtsi128_si32(ref1));
-      WebPUint32ToMem(&dst[2 * BPS], _mm_cvtsi128_si32(ref2));
+      WebPInt32ToMem(&dst[2 * BPS], _mm_cvtsi128_si32(ref2));
-      WebPUint32ToMem(&dst[3 * BPS], _mm_cvtsi128_si32(ref3));
+      WebPInt32ToMem(&dst[3 * BPS], _mm_cvtsi128_si32(ref3));
     }
   }
 }
@@ -481,7 +481,7 @@ static void CollectHistogram_SSE2(const uint8_t* ref, const uint8_t* pred,
 // helper for chroma-DC predictions
 static WEBP_INLINE void Put8x8uv_SSE2(uint8_t v, uint8_t* dst) {
   int j;
-  const __m128i values = _mm_set1_epi8(v);
+  const __m128i values = _mm_set1_epi8((char)v);
   for (j = 0; j < 8; ++j) {
     _mm_storel_epi64((__m128i*)(dst + j * BPS), values);
   }
@@ -489,7 +489,7 @@ static WEBP_INLINE void Put8x8uv_SSE2(uint8_t v, uint8_t* dst) {
 
 static WEBP_INLINE void Put16_SSE2(uint8_t v, uint8_t* dst) {
   int j;
-  const __m128i values = _mm_set1_epi8(v);
+  const __m128i values = _mm_set1_epi8((char)v);
   for (j = 0; j < 16; ++j) {
     _mm_store_si128((__m128i*)(dst + j * BPS), values);
   }
@@ -540,7 +540,7 @@ static WEBP_INLINE void VerticalPred_SSE2(uint8_t* dst,
 static WEBP_INLINE void HE8uv_SSE2(uint8_t* dst, const uint8_t* left) {
   int j;
   for (j = 0; j < 8; ++j) {
-    const __m128i values = _mm_set1_epi8(left[j]);
+    const __m128i values = _mm_set1_epi8((char)left[j]);
     _mm_storel_epi64((__m128i*)dst, values);
     dst += BPS;
   }
@@ -549,7 +549,7 @@ static WEBP_INLINE void HE8uv_SSE2(uint8_t* dst, const uint8_t* left) {
 static WEBP_INLINE void HE16_SSE2(uint8_t* dst, const uint8_t* left) {
   int j;
   for (j = 0; j < 16; ++j) {
-    const __m128i values = _mm_set1_epi8(left[j]);
+    const __m128i values = _mm_set1_epi8((char)left[j]);
     _mm_store_si128((__m128i*)dst, values);
     dst += BPS;
   }
@@ -722,10 +722,10 @@ static WEBP_INLINE void VE4_SSE2(uint8_t* dst,
   const __m128i lsb = _mm_and_si128(_mm_xor_si128(ABCDEFGH, CDEFGH00), one);
   const __m128i b = _mm_subs_epu8(a, lsb);
   const __m128i avg = _mm_avg_epu8(b, BCDEFGH0);
-  const uint32_t vals = _mm_cvtsi128_si32(avg);
+  const int vals = _mm_cvtsi128_si32(avg);
   int i;
   for (i = 0; i < 4; ++i) {
-    WebPUint32ToMem(dst + i * BPS, vals);
+    WebPInt32ToMem(dst + i * BPS, vals);
   }
 }
 
@@ -760,10 +760,10 @@ static WEBP_INLINE void LD4_SSE2(uint8_t* dst,
   const __m128i lsb = _mm_and_si128(_mm_xor_si128(ABCDEFGH, CDEFGHH0), one);
   const __m128i avg2 = _mm_subs_epu8(avg1, lsb);
   const __m128i abcdefg = _mm_avg_epu8(avg2, BCDEFGH0);
-  WebPUint32ToMem(dst + 0 * BPS, _mm_cvtsi128_si32(               abcdefg    ));
+  WebPInt32ToMem(dst + 0 * BPS, _mm_cvtsi128_si32(               abcdefg    ));
-  WebPUint32ToMem(dst + 1 * BPS, _mm_cvtsi128_si32(_mm_srli_si128(abcdefg, 1)));
+  WebPInt32ToMem(dst + 1 * BPS, _mm_cvtsi128_si32(_mm_srli_si128(abcdefg, 1)));
-  WebPUint32ToMem(dst + 2 * BPS, _mm_cvtsi128_si32(_mm_srli_si128(abcdefg, 2)));
+  WebPInt32ToMem(dst + 2 * BPS, _mm_cvtsi128_si32(_mm_srli_si128(abcdefg, 2)));
-  WebPUint32ToMem(dst + 3 * BPS, _mm_cvtsi128_si32(_mm_srli_si128(abcdefg, 3)));
+  WebPInt32ToMem(dst + 3 * BPS, _mm_cvtsi128_si32(_mm_srli_si128(abcdefg, 3)));
 }
 
 static WEBP_INLINE void VR4_SSE2(uint8_t* dst,
@@ -782,10 +782,10 @@ static WEBP_INLINE void VR4_SSE2(uint8_t* dst,
   const __m128i lsb = _mm_and_si128(_mm_xor_si128(IXABCD, ABCD0), one);
   const __m128i avg2 = _mm_subs_epu8(avg1, lsb);
   const __m128i efgh = _mm_avg_epu8(avg2, XABCD);
-  WebPUint32ToMem(dst + 0 * BPS, _mm_cvtsi128_si32(               abcd    ));
+  WebPInt32ToMem(dst + 0 * BPS, _mm_cvtsi128_si32(               abcd    ));
-  WebPUint32ToMem(dst + 1 * BPS, _mm_cvtsi128_si32(               efgh    ));
+  WebPInt32ToMem(dst + 1 * BPS, _mm_cvtsi128_si32(               efgh    ));
-  WebPUint32ToMem(dst + 2 * BPS, _mm_cvtsi128_si32(_mm_slli_si128(abcd, 1)));
+  WebPInt32ToMem(dst + 2 * BPS, _mm_cvtsi128_si32(_mm_slli_si128(abcd, 1)));
-  WebPUint32ToMem(dst + 3 * BPS, _mm_cvtsi128_si32(_mm_slli_si128(efgh, 1)));
+  WebPInt32ToMem(dst + 3 * BPS, _mm_cvtsi128_si32(_mm_slli_si128(efgh, 1)));
 
   // these two are hard to implement in SSE2, so we keep the C-version:
   DST(0, 2) = AVG3(J, I, X);
@@ -807,11 +807,12 @@ static WEBP_INLINE void VL4_SSE2(uint8_t* dst,
   const __m128i abbc = _mm_or_si128(ab, bc);
   const __m128i lsb2 = _mm_and_si128(abbc, lsb1);
   const __m128i avg4 = _mm_subs_epu8(avg3, lsb2);
-  const uint32_t extra_out = _mm_cvtsi128_si32(_mm_srli_si128(avg4, 4));
+  const uint32_t extra_out =
-  WebPUint32ToMem(dst + 0 * BPS, _mm_cvtsi128_si32(               avg1    ));
+      (uint32_t)_mm_cvtsi128_si32(_mm_srli_si128(avg4, 4));
-  WebPUint32ToMem(dst + 1 * BPS, _mm_cvtsi128_si32(               avg4    ));
+  WebPInt32ToMem(dst + 0 * BPS, _mm_cvtsi128_si32(               avg1    ));
-  WebPUint32ToMem(dst + 2 * BPS, _mm_cvtsi128_si32(_mm_srli_si128(avg1, 1)));
+  WebPInt32ToMem(dst + 1 * BPS, _mm_cvtsi128_si32(               avg4    ));
-  WebPUint32ToMem(dst + 3 * BPS, _mm_cvtsi128_si32(_mm_srli_si128(avg4, 1)));
+  WebPInt32ToMem(dst + 2 * BPS, _mm_cvtsi128_si32(_mm_srli_si128(avg1, 1)));
+  WebPInt32ToMem(dst + 3 * BPS, _mm_cvtsi128_si32(_mm_srli_si128(avg4, 1)));
 
   // these two are hard to get and irregular
   DST(3, 2) = (extra_out >> 0) & 0xff;
@@ -829,10 +830,10 @@ static WEBP_INLINE void RD4_SSE2(uint8_t* dst,
   const __m128i lsb = _mm_and_si128(_mm_xor_si128(JIXABCD__, LKJIXABCD), one);
   const __m128i avg2 = _mm_subs_epu8(avg1, lsb);
   const __m128i abcdefg = _mm_avg_epu8(avg2, KJIXABCD_);
-  WebPUint32ToMem(dst + 3 * BPS, _mm_cvtsi128_si32(               abcdefg    ));
+  WebPInt32ToMem(dst + 3 * BPS, _mm_cvtsi128_si32(               abcdefg    ));
-  WebPUint32ToMem(dst + 2 * BPS, _mm_cvtsi128_si32(_mm_srli_si128(abcdefg, 1)));
+  WebPInt32ToMem(dst + 2 * BPS, _mm_cvtsi128_si32(_mm_srli_si128(abcdefg, 1)));
-  WebPUint32ToMem(dst + 1 * BPS, _mm_cvtsi128_si32(_mm_srli_si128(abcdefg, 2)));
+  WebPInt32ToMem(dst + 1 * BPS, _mm_cvtsi128_si32(_mm_srli_si128(abcdefg, 2)));
-  WebPUint32ToMem(dst + 0 * BPS, _mm_cvtsi128_si32(_mm_srli_si128(abcdefg, 3)));
+  WebPInt32ToMem(dst + 0 * BPS, _mm_cvtsi128_si32(_mm_srli_si128(abcdefg, 3)));
 }
 
 static WEBP_INLINE void HU4_SSE2(uint8_t* dst, const uint8_t* top) {
@@ -875,14 +876,14 @@ static WEBP_INLINE void HD4_SSE2(uint8_t* dst, const uint8_t* top) {
 
 static WEBP_INLINE void TM4_SSE2(uint8_t* dst, const uint8_t* top) {
   const __m128i zero = _mm_setzero_si128();
-  const __m128i top_values = _mm_cvtsi32_si128(WebPMemToUint32(top));
+  const __m128i top_values = _mm_cvtsi32_si128(WebPMemToInt32(top));
   const __m128i top_base = _mm_unpacklo_epi8(top_values, zero);
   int y;
   for (y = 0; y < 4; ++y, dst += BPS) {
     const int val = top[-2 - y] - top[-1];
     const __m128i base = _mm_set1_epi16(val);
     const __m128i out = _mm_packus_epi16(_mm_add_epi16(base, top_base), zero);
-    WebPUint32ToMem(dst, _mm_cvtsi128_si32(out));
+    WebPInt32ToMem(dst, _mm_cvtsi128_si32(out));
   }
 }
 

src/dsp/lossless.c

@@ -49,7 +49,7 @@ static WEBP_INLINE uint32_t Clip255(uint32_t a) {
 }
 
 static WEBP_INLINE int AddSubtractComponentFull(int a, int b, int c) {
-  return Clip255(a + b - c);
+  return Clip255((uint32_t)(a + b - c));
 }
 
 static WEBP_INLINE uint32_t ClampedAddSubtractFull(uint32_t c0, uint32_t c1,
@@ -66,7 +66,7 @@ static WEBP_INLINE uint32_t ClampedAddSubtractFull(uint32_t c0, uint32_t c1,
 }
 
 static WEBP_INLINE int AddSubtractComponentHalf(int a, int b) {
-  return Clip255(a + (a - b) / 2);
+  return Clip255((uint32_t)(a + (a - b) / 2));
 }
 
 static WEBP_INLINE uint32_t ClampedAddSubtractHalf(uint32_t c0, uint32_t c1,
@@ -293,10 +293,10 @@ void VP8LTransformColorInverse_C(const VP8LMultipliers* const m,
     const uint32_t red = argb >> 16;
     int new_red = red & 0xff;
     int new_blue = argb & 0xff;
-    new_red += ColorTransformDelta(m->green_to_red_, green);
+    new_red += ColorTransformDelta((int8_t)m->green_to_red_, green);
     new_red &= 0xff;
-    new_blue += ColorTransformDelta(m->green_to_blue_, green);
+    new_blue += ColorTransformDelta((int8_t)m->green_to_blue_, green);
-    new_blue += ColorTransformDelta(m->red_to_blue_, (int8_t)new_red);
+    new_blue += ColorTransformDelta((int8_t)m->red_to_blue_, (int8_t)new_red);
     new_blue &= 0xff;
     dst[i] = (argb & 0xff00ff00u) | (new_red << 16) | (new_blue);
   }
@@ -395,7 +395,7 @@ void VP8LInverseTransform(const VP8LTransform* const transform,
   assert(row_start < row_end);
   assert(row_end <= transform->ysize_);
   switch (transform->type_) {
-    case SUBTRACT_GREEN:
+    case SUBTRACT_GREEN_TRANSFORM:
       VP8LAddGreenToBlueAndRed(in, (row_end - row_start) * width, out);
       break;
     case PREDICTOR_TRANSFORM:

src/dsp/lossless.h

@@ -182,9 +182,9 @@ extern VP8LPredictorAddSubFunc VP8LPredictorsSub_C[16];
 // -----------------------------------------------------------------------------
 // Huffman-cost related functions.
 
-typedef double (*VP8LCostFunc)(const uint32_t* population, int length);
+typedef float (*VP8LCostFunc)(const uint32_t* population, int length);
-typedef double (*VP8LCostCombinedFunc)(const uint32_t* X, const uint32_t* Y,
+typedef float (*VP8LCostCombinedFunc)(const uint32_t* X, const uint32_t* Y,
-                                       int length);
+                                      int length);
 typedef float (*VP8LCombinedShannonEntropyFunc)(const int X[256],
                                                 const int Y[256]);
 
@@ -198,7 +198,7 @@ typedef struct {        // small struct to hold counters
 } VP8LStreaks;
 
 typedef struct {            // small struct to hold bit entropy results
-  double entropy;           // entropy
+  float entropy;            // entropy
   uint32_t sum;             // sum of the population
   int nonzeros;             // number of non-zero elements in the population
   uint32_t max_val;         // maximum value in the population

src/dsp/lossless_enc.c

@@ -402,7 +402,7 @@ static float FastLog2Slow_C(uint32_t v) {
 // Compute the combined Shanon's entropy for distribution {X} and {X+Y}
 static float CombinedShannonEntropy_C(const int X[256], const int Y[256]) {
   int i;
-  double retval = 0.;
+  float retval = 0.f;
   int sumX = 0, sumXY = 0;
   for (i = 0; i < 256; ++i) {
     const int x = X[i];
@@ -418,7 +418,7 @@ static float CombinedShannonEntropy_C(const int X[256], const int Y[256]) {
     }
   }
   retval += VP8LFastSLog2(sumX) + VP8LFastSLog2(sumXY);
-  return (float)retval;
+  return retval;
 }
 
 void VP8LBitEntropyInit(VP8LBitEntropy* const entropy) {
@@ -522,11 +522,11 @@ static void GetCombinedEntropyUnrefined_C(const uint32_t X[],
 void VP8LSubtractGreenFromBlueAndRed_C(uint32_t* argb_data, int num_pixels) {
   int i;
   for (i = 0; i < num_pixels; ++i) {
-    const int argb = argb_data[i];
+    const int argb = (int)argb_data[i];
     const int green = (argb >> 8) & 0xff;
     const uint32_t new_r = (((argb >> 16) & 0xff) - green) & 0xff;
     const uint32_t new_b = (((argb >>  0) & 0xff) - green) & 0xff;
-    argb_data[i] = (argb & 0xff00ff00u) | (new_r << 16) | new_b;
+    argb_data[i] = ((uint32_t)argb & 0xff00ff00u) | (new_r << 16) | new_b;
   }
 }
 
@@ -547,10 +547,10 @@ void VP8LTransformColor_C(const VP8LMultipliers* const m, uint32_t* data,
     const int8_t red   = U32ToS8(argb >> 16);
     int new_red = red & 0xff;
     int new_blue = argb & 0xff;
-    new_red -= ColorTransformDelta(m->green_to_red_, green);
+    new_red -= ColorTransformDelta((int8_t)m->green_to_red_, green);
     new_red &= 0xff;
-    new_blue -= ColorTransformDelta(m->green_to_blue_, green);
+    new_blue -= ColorTransformDelta((int8_t)m->green_to_blue_, green);
-    new_blue -= ColorTransformDelta(m->red_to_blue_, red);
+    new_blue -= ColorTransformDelta((int8_t)m->red_to_blue_, red);
     new_blue &= 0xff;
     data[i] = (argb & 0xff00ff00u) | (new_red << 16) | (new_blue);
   }
@@ -560,7 +560,7 @@ static WEBP_INLINE uint8_t TransformColorRed(uint8_t green_to_red,
                                              uint32_t argb) {
   const int8_t green = U32ToS8(argb >> 8);
   int new_red = argb >> 16;
-  new_red -= ColorTransformDelta(green_to_red, green);
+  new_red -= ColorTransformDelta((int8_t)green_to_red, green);
   return (new_red & 0xff);
 }
 
@@ -569,9 +569,9 @@ static WEBP_INLINE uint8_t TransformColorBlue(uint8_t green_to_blue,
                                               uint32_t argb) {
   const int8_t green = U32ToS8(argb >>  8);
   const int8_t red   = U32ToS8(argb >> 16);
-  uint8_t new_blue = argb & 0xff;
+  int new_blue = argb & 0xff;
-  new_blue -= ColorTransformDelta(green_to_blue, green);
+  new_blue -= ColorTransformDelta((int8_t)green_to_blue, green);
-  new_blue -= ColorTransformDelta(red_to_blue, red);
+  new_blue -= ColorTransformDelta((int8_t)red_to_blue, red);
   return (new_blue & 0xff);
 }
 
@@ -636,17 +636,17 @@ void VP8LBundleColorMap_C(const uint8_t* const row, int width, int xbits,
 
 //------------------------------------------------------------------------------
 
-static double ExtraCost_C(const uint32_t* population, int length) {
+static float ExtraCost_C(const uint32_t* population, int length) {
   int i;
-  double cost = 0.;
+  float cost = 0.f;
   for (i = 2; i < length - 2; ++i) cost += (i >> 1) * population[i + 2];
   return cost;
 }
 
-static double ExtraCostCombined_C(const uint32_t* X, const uint32_t* Y,
+static float ExtraCostCombined_C(const uint32_t* X, const uint32_t* Y,
                                   int length) {
   int i;
-  double cost = 0.;
+  float cost = 0.f;
   for (i = 2; i < length - 2; ++i) {
     const int xy = X[i + 2] + Y[i + 2];
     cost += (i >> 1) * xy;

src/dsp/lossless_enc_mips32.c

@@ -103,8 +103,8 @@ static float FastLog2Slow_MIPS32(uint32_t v) {
 //     cost += i * *(pop + 1);
 //     pop += 2;
 //   }
-//   return (double)cost;
+//   return (float)cost;
-static double ExtraCost_MIPS32(const uint32_t* const population, int length) {
+static float ExtraCost_MIPS32(const uint32_t* const population, int length) {
   int i, temp0, temp1;
   const uint32_t* pop = &population[4];
   const uint32_t* const LoopEnd = &population[length];
@@ -130,7 +130,7 @@ static double ExtraCost_MIPS32(const uint32_t* const population, int length) {
     : "memory", "hi", "lo"
   );
 
-  return (double)((int64_t)temp0 << 32 | temp1);
+  return (float)((int64_t)temp0 << 32 | temp1);
 }
 
 // C version of this function:
@@ -148,9 +148,9 @@ static double ExtraCost_MIPS32(const uint32_t* const population, int length) {
 //     pX += 2;
 //     pY += 2;
 //   }
-//   return (double)cost;
+//   return (float)cost;
-static double ExtraCostCombined_MIPS32(const uint32_t* const X,
+static float ExtraCostCombined_MIPS32(const uint32_t* const X,
-                                       const uint32_t* const Y, int length) {
+                                      const uint32_t* const Y, int length) {
   int i, temp0, temp1, temp2, temp3;
   const uint32_t* pX = &X[4];
   const uint32_t* pY = &Y[4];
@@ -183,7 +183,7 @@ static double ExtraCostCombined_MIPS32(const uint32_t* const X,
     : "memory", "hi", "lo"
   );
 
-  return (double)((int64_t)temp0 << 32 | temp1);
+  return (float)((int64_t)temp0 << 32 | temp1);
 }
 
 #define HUFFMAN_COST_PASS                                 \
@@ -347,24 +347,24 @@ static void GetCombinedEntropyUnrefined_MIPS32(const uint32_t X[],
 static void AddVector_MIPS32(const uint32_t* pa, const uint32_t* pb,
                              uint32_t* pout, int size) {
   uint32_t temp0, temp1, temp2, temp3, temp4, temp5, temp6, temp7;
-  const uint32_t end = ((size) / 4) * 4;
+  const int end = ((size) / 4) * 4;
   const uint32_t* const LoopEnd = pa + end;
   int i;
   ASM_START
   ADD_TO_OUT(0, 4, 8, 12, 1, pa, pb, pout)
   ASM_END_0
-  for (i = end; i < size; ++i) pout[i] = pa[i] + pb[i];
+  for (i = 0; i < size - end; ++i) pout[i] = pa[i] + pb[i];
 }
 
 static void AddVectorEq_MIPS32(const uint32_t* pa, uint32_t* pout, int size) {
   uint32_t temp0, temp1, temp2, temp3, temp4, temp5, temp6, temp7;
-  const uint32_t end = ((size) / 4) * 4;
+  const int end = ((size) / 4) * 4;
   const uint32_t* const LoopEnd = pa + end;
   int i;
   ASM_START
   ADD_TO_OUT(0, 4, 8, 12, 0, pa, pout, pout)
   ASM_END_1
-  for (i = end; i < size; ++i) pout[i] += pa[i];
+  for (i = 0; i < size - end; ++i) pout[i] += pa[i];
 }
 
 #undef ASM_END_1

src/dsp/lossless_enc_sse2.c

@@ -54,8 +54,8 @@ static void TransformColor_SSE2(const VP8LMultipliers* const m,
   const __m128i mults_rb = MK_CST_16(CST_5b(m->green_to_red_),
                                      CST_5b(m->green_to_blue_));
   const __m128i mults_b2 = MK_CST_16(CST_5b(m->red_to_blue_), 0);
-  const __m128i mask_ag = _mm_set1_epi32(0xff00ff00);  // alpha-green masks
+  const __m128i mask_ag = _mm_set1_epi32((int)0xff00ff00);  // alpha-green masks
-  const __m128i mask_rb = _mm_set1_epi32(0x00ff00ff);  // red-blue masks
+  const __m128i mask_rb = _mm_set1_epi32(0x00ff00ff);       // red-blue masks
   int i;
   for (i = 0; i + 4 <= num_pixels; i += 4) {
     const __m128i in = _mm_loadu_si128((__m128i*)&argb_data[i]); // argb
@@ -239,7 +239,7 @@ static void AddVectorEq_SSE2(const uint32_t* a, uint32_t* out, int size) {
 
 static float CombinedShannonEntropy_SSE2(const int X[256], const int Y[256]) {
   int i;
-  double retval = 0.;
+  float retval = 0.f;
   int sumX = 0, sumXY = 0;
   const __m128i zero = _mm_setzero_si128();
 
@@ -273,7 +273,7 @@ static float CombinedShannonEntropy_SSE2(const int X[256], const int Y[256]) {
     }
   }
   retval += VP8LFastSLog2(sumX) + VP8LFastSLog2(sumXY);
-  return (float)retval;
+  return retval;
 }
 
 #else
@@ -376,7 +376,7 @@ static void BundleColorMap_SSE2(const uint8_t* const row, int width, int xbits,
       break;
     }
     case 2: {
-      const __m128i mask_or = _mm_set1_epi32(0xff000000);
+      const __m128i mask_or = _mm_set1_epi32((int)0xff000000);
       const __m128i mul_cst = _mm_set1_epi16(0x0104);
       const __m128i mask_mul = _mm_set1_epi16(0x0f00);
       for (x = 0; x + 16 <= width; x += 16, dst += 4) {
@@ -427,7 +427,7 @@ static WEBP_INLINE void Average2_m128i(const __m128i* const a0,
 static void PredictorSub0_SSE2(const uint32_t* in, const uint32_t* upper,
                                int num_pixels, uint32_t* out) {
   int i;
-  const __m128i black = _mm_set1_epi32(ARGB_BLACK);
+  const __m128i black = _mm_set1_epi32((int)ARGB_BLACK);
   for (i = 0; i + 4 <= num_pixels; i += 4) {
     const __m128i src = _mm_loadu_si128((const __m128i*)&in[i]);
     const __m128i res = _mm_sub_epi8(src, black);

src/dsp/lossless_sse2.c

@@ -27,23 +27,22 @@ static WEBP_INLINE uint32_t ClampedAddSubtractFull_SSE2(uint32_t c0,
                                                         uint32_t c1,
                                                         uint32_t c2) {
   const __m128i zero = _mm_setzero_si128();
-  const __m128i C0 = _mm_unpacklo_epi8(_mm_cvtsi32_si128(c0), zero);
+  const __m128i C0 = _mm_unpacklo_epi8(_mm_cvtsi32_si128((int)c0), zero);
-  const __m128i C1 = _mm_unpacklo_epi8(_mm_cvtsi32_si128(c1), zero);
+  const __m128i C1 = _mm_unpacklo_epi8(_mm_cvtsi32_si128((int)c1), zero);
-  const __m128i C2 = _mm_unpacklo_epi8(_mm_cvtsi32_si128(c2), zero);
+  const __m128i C2 = _mm_unpacklo_epi8(_mm_cvtsi32_si128((int)c2), zero);
   const __m128i V1 = _mm_add_epi16(C0, C1);
   const __m128i V2 = _mm_sub_epi16(V1, C2);
   const __m128i b = _mm_packus_epi16(V2, V2);
-  const uint32_t output = _mm_cvtsi128_si32(b);
+  return (uint32_t)_mm_cvtsi128_si32(b);
-  return output;
 }
 
 static WEBP_INLINE uint32_t ClampedAddSubtractHalf_SSE2(uint32_t c0,
                                                         uint32_t c1,
                                                         uint32_t c2) {
   const __m128i zero = _mm_setzero_si128();
-  const __m128i C0 = _mm_unpacklo_epi8(_mm_cvtsi32_si128(c0), zero);
+  const __m128i C0 = _mm_unpacklo_epi8(_mm_cvtsi32_si128((int)c0), zero);
-  const __m128i C1 = _mm_unpacklo_epi8(_mm_cvtsi32_si128(c1), zero);
+  const __m128i C1 = _mm_unpacklo_epi8(_mm_cvtsi32_si128((int)c1), zero);
-  const __m128i B0 = _mm_unpacklo_epi8(_mm_cvtsi32_si128(c2), zero);
+  const __m128i B0 = _mm_unpacklo_epi8(_mm_cvtsi32_si128((int)c2), zero);
   const __m128i avg = _mm_add_epi16(C1, C0);
   const __m128i A0 = _mm_srli_epi16(avg, 1);
   const __m128i A1 = _mm_sub_epi16(A0, B0);
@@ -52,16 +51,15 @@ static WEBP_INLINE uint32_t ClampedAddSubtractHalf_SSE2(uint32_t c0,
   const __m128i A3 = _mm_srai_epi16(A2, 1);
   const __m128i A4 = _mm_add_epi16(A0, A3);
   const __m128i A5 = _mm_packus_epi16(A4, A4);
-  const uint32_t output = _mm_cvtsi128_si32(A5);
+  return (uint32_t)_mm_cvtsi128_si32(A5);
-  return output;
 }
 
 static WEBP_INLINE uint32_t Select_SSE2(uint32_t a, uint32_t b, uint32_t c) {
   int pa_minus_pb;
   const __m128i zero = _mm_setzero_si128();
-  const __m128i A0 = _mm_cvtsi32_si128(a);
+  const __m128i A0 = _mm_cvtsi32_si128((int)a);
-  const __m128i B0 = _mm_cvtsi32_si128(b);
+  const __m128i B0 = _mm_cvtsi32_si128((int)b);
-  const __m128i C0 = _mm_cvtsi32_si128(c);
+  const __m128i C0 = _mm_cvtsi32_si128((int)c);
   const __m128i AC0 = _mm_subs_epu8(A0, C0);
   const __m128i CA0 = _mm_subs_epu8(C0, A0);
   const __m128i BC0 = _mm_subs_epu8(B0, C0);
@@ -94,8 +92,8 @@ static WEBP_INLINE void Average2_uint32_SSE2(const uint32_t a0,
                                              __m128i* const avg) {
   // (a + b) >> 1 = ((a + b + 1) >> 1) - ((a ^ b) & 1)
   const __m128i ones = _mm_set1_epi8(1);
-  const __m128i A0 = _mm_cvtsi32_si128(a0);
+  const __m128i A0 = _mm_cvtsi32_si128((int)a0);
-  const __m128i A1 = _mm_cvtsi32_si128(a1);
+  const __m128i A1 = _mm_cvtsi32_si128((int)a1);
   const __m128i avg1 = _mm_avg_epu8(A0, A1);
   const __m128i one = _mm_and_si128(_mm_xor_si128(A0, A1), ones);
   *avg = _mm_sub_epi8(avg1, one);
@@ -103,8 +101,8 @@ static WEBP_INLINE void Average2_uint32_SSE2(const uint32_t a0,
 
 static WEBP_INLINE __m128i Average2_uint32_16_SSE2(uint32_t a0, uint32_t a1) {
   const __m128i zero = _mm_setzero_si128();
-  const __m128i A0 = _mm_unpacklo_epi8(_mm_cvtsi32_si128(a0), zero);
+  const __m128i A0 = _mm_unpacklo_epi8(_mm_cvtsi32_si128((int)a0), zero);
-  const __m128i A1 = _mm_unpacklo_epi8(_mm_cvtsi32_si128(a1), zero);
+  const __m128i A1 = _mm_unpacklo_epi8(_mm_cvtsi32_si128((int)a1), zero);
   const __m128i sum = _mm_add_epi16(A1, A0);
   return _mm_srli_epi16(sum, 1);
 }
@@ -112,19 +110,18 @@ static WEBP_INLINE __m128i Average2_uint32_16_SSE2(uint32_t a0, uint32_t a1) {
 static WEBP_INLINE uint32_t Average2_SSE2(uint32_t a0, uint32_t a1) {
   __m128i output;
   Average2_uint32_SSE2(a0, a1, &output);
-  return _mm_cvtsi128_si32(output);
+  return (uint32_t)_mm_cvtsi128_si32(output);
 }
 
 static WEBP_INLINE uint32_t Average3_SSE2(uint32_t a0, uint32_t a1,
                                           uint32_t a2) {
   const __m128i zero = _mm_setzero_si128();
   const __m128i avg1 = Average2_uint32_16_SSE2(a0, a2);
-  const __m128i A1 = _mm_unpacklo_epi8(_mm_cvtsi32_si128(a1), zero);
+  const __m128i A1 = _mm_unpacklo_epi8(_mm_cvtsi32_si128((int)a1), zero);
   const __m128i sum = _mm_add_epi16(avg1, A1);
   const __m128i avg2 = _mm_srli_epi16(sum, 1);
   const __m128i A2 = _mm_packus_epi16(avg2, avg2);
-  const uint32_t output = _mm_cvtsi128_si32(A2);
+  return (uint32_t)_mm_cvtsi128_si32(A2);
-  return output;
 }
 
 static WEBP_INLINE uint32_t Average4_SSE2(uint32_t a0, uint32_t a1,
@@ -134,8 +131,7 @@ static WEBP_INLINE uint32_t Average4_SSE2(uint32_t a0, uint32_t a1,
   const __m128i sum = _mm_add_epi16(avg2, avg1);
   const __m128i avg3 = _mm_srli_epi16(sum, 1);
   const __m128i A0 = _mm_packus_epi16(avg3, avg3);
-  const uint32_t output = _mm_cvtsi128_si32(A0);
+  return (uint32_t)_mm_cvtsi128_si32(A0);
-  return output;
 }
 
 static uint32_t Predictor5_SSE2(const uint32_t* const left,
@@ -192,7 +188,7 @@ static uint32_t Predictor13_SSE2(const uint32_t* const left,
 static void PredictorAdd0_SSE2(const uint32_t* in, const uint32_t* upper,
                                int num_pixels, uint32_t* out) {
   int i;
-  const __m128i black = _mm_set1_epi32(ARGB_BLACK);
+  const __m128i black = _mm_set1_epi32((int)ARGB_BLACK);
   for (i = 0; i + 4 <= num_pixels; i += 4) {
     const __m128i src = _mm_loadu_si128((const __m128i*)&in[i]);
     const __m128i res = _mm_add_epi8(src, black);
@@ -208,7 +204,7 @@ static void PredictorAdd0_SSE2(const uint32_t* in, const uint32_t* upper,
 static void PredictorAdd1_SSE2(const uint32_t* in, const uint32_t* upper,
                                int num_pixels, uint32_t* out) {
   int i;
-  __m128i prev = _mm_set1_epi32(out[-1]);
+  __m128i prev = _mm_set1_epi32((int)out[-1]);
   for (i = 0; i + 4 <= num_pixels; i += 4) {
     // a | b | c | d
     const __m128i src = _mm_loadu_si128((const __m128i*)&in[i]);
@@ -285,12 +281,12 @@ GENERATE_PREDICTOR_2(9, upper[i + 1])
 #undef GENERATE_PREDICTOR_2
 
 // Predictor10: average of (average of (L,TL), average of (T, TR)).
-#define DO_PRED10(OUT) do {               \
+#define DO_PRED10(OUT) do {                         \
-  __m128i avgLTL, avg;                    \
+  __m128i avgLTL, avg;                              \
-  Average2_m128i(&L, &TL, &avgLTL);       \
+  Average2_m128i(&L, &TL, &avgLTL);                 \
-  Average2_m128i(&avgTTR, &avgLTL, &avg); \
+  Average2_m128i(&avgTTR, &avgLTL, &avg);           \
-  L = _mm_add_epi8(avg, src);             \
+  L = _mm_add_epi8(avg, src);                       \
-  out[i + (OUT)] = _mm_cvtsi128_si32(L);  \
+  out[i + (OUT)] = (uint32_t)_mm_cvtsi128_si32(L);  \
 } while (0)
 
 #define DO_PRED10_SHIFT do {                                  \
@@ -303,7 +299,7 @@ GENERATE_PREDICTOR_2(9, upper[i + 1])
 static void PredictorAdd10_SSE2(const uint32_t* in, const uint32_t* upper,
                                 int num_pixels, uint32_t* out) {
   int i;
-  __m128i L = _mm_cvtsi32_si128(out[-1]);
+  __m128i L = _mm_cvtsi32_si128((int)out[-1]);
   for (i = 0; i + 4 <= num_pixels; i += 4) {
     __m128i src = _mm_loadu_si128((const __m128i*)&in[i]);
     __m128i TL = _mm_loadu_si128((const __m128i*)&upper[i - 1]);
@@ -336,7 +332,7 @@ static void PredictorAdd10_SSE2(const uint32_t* in, const uint32_t* upper,
   const __m128i B = _mm_andnot_si128(mask, T);                         \
   const __m128i pred = _mm_or_si128(A, B); /* pred = (pa > b)? L : T*/ \
   L = _mm_add_epi8(src, pred);                                         \
-  out[i + (OUT)] = _mm_cvtsi128_si32(L);                               \
+  out[i + (OUT)] = (uint32_t)_mm_cvtsi128_si32(L);                     \
 } while (0)
 
 #define DO_PRED11_SHIFT do {                                \
@@ -351,7 +347,7 @@ static void PredictorAdd11_SSE2(const uint32_t* in, const uint32_t* upper,
                                 int num_pixels, uint32_t* out) {
   int i;
   __m128i pa;
-  __m128i L = _mm_cvtsi32_si128(out[-1]);
+  __m128i L = _mm_cvtsi32_si128((int)out[-1]);
   for (i = 0; i + 4 <= num_pixels; i += 4) {
     __m128i T = _mm_loadu_si128((const __m128i*)&upper[i]);
     __m128i TL = _mm_loadu_si128((const __m128i*)&upper[i - 1]);
@@ -384,12 +380,12 @@ static void PredictorAdd11_SSE2(const uint32_t* in, const uint32_t* upper,
 #undef DO_PRED11_SHIFT
 
 // Predictor12: ClampedAddSubtractFull.
-#define DO_PRED12(DIFF, LANE, OUT) do {            \
+#define DO_PRED12(DIFF, LANE, OUT) do {              \
-  const __m128i all = _mm_add_epi16(L, (DIFF));    \
+  const __m128i all = _mm_add_epi16(L, (DIFF));      \
-  const __m128i alls = _mm_packus_epi16(all, all); \
+  const __m128i alls = _mm_packus_epi16(all, all);   \
-  const __m128i res = _mm_add_epi8(src, alls);     \
+  const __m128i res = _mm_add_epi8(src, alls);       \
-  out[i + (OUT)] = _mm_cvtsi128_si32(res);         \
+  out[i + (OUT)] = (uint32_t)_mm_cvtsi128_si32(res); \
-  L = _mm_unpacklo_epi8(res, zero);                \
+  L = _mm_unpacklo_epi8(res, zero);                  \
 } while (0)
 
 #define DO_PRED12_SHIFT(DIFF, LANE) do {                    \
@@ -402,7 +398,7 @@ static void PredictorAdd12_SSE2(const uint32_t* in, const uint32_t* upper,
                                 int num_pixels, uint32_t* out) {
   int i;
   const __m128i zero = _mm_setzero_si128();
-  const __m128i L8 = _mm_cvtsi32_si128(out[-1]);
+  const __m128i L8 = _mm_cvtsi32_si128((int)out[-1]);
   __m128i L = _mm_unpacklo_epi8(L8, zero);
   for (i = 0; i + 4 <= num_pixels; i += 4) {
     // Load 4 pixels at a time.
@@ -468,7 +464,7 @@ static void TransformColorInverse_SSE2(const VP8LMultipliers* const m,
   const __m128i mults_b2 = MK_CST_16(CST(red_to_blue_), 0);
 #undef MK_CST_16
 #undef CST
-  const __m128i mask_ag = _mm_set1_epi32(0xff00ff00);  // alpha-green masks
+  const __m128i mask_ag = _mm_set1_epi32((int)0xff00ff00);  // alpha-green masks
   int i;
   for (i = 0; i + 4 <= num_pixels; i += 4) {
     const __m128i in = _mm_loadu_si128((const __m128i*)&src[i]); // argb
@@ -532,7 +528,7 @@ static void ConvertBGRAToRGB_SSE2(const uint32_t* src, int num_pixels,
 
 static void ConvertBGRAToRGBA_SSE2(const uint32_t* src,
                                    int num_pixels, uint8_t* dst) {
-  const __m128i red_blue_mask = _mm_set1_epi32(0x00ff00ffu);
+  const __m128i red_blue_mask = _mm_set1_epi32(0x00ff00ff);
   const __m128i* in = (const __m128i*)src;
   __m128i* out = (__m128i*)dst;
   while (num_pixels >= 8) {
@@ -561,7 +557,7 @@ static void ConvertBGRAToRGBA_SSE2(const uint32_t* src,
 static void ConvertBGRAToRGBA4444_SSE2(const uint32_t* src,
                                        int num_pixels, uint8_t* dst) {
   const __m128i mask_0x0f = _mm_set1_epi8(0x0f);
-  const __m128i mask_0xf0 = _mm_set1_epi8(0xf0);
+  const __m128i mask_0xf0 = _mm_set1_epi8((char)0xf0);
   const __m128i* in = (const __m128i*)src;
   __m128i* out = (__m128i*)dst;
   while (num_pixels >= 8) {
@@ -596,8 +592,8 @@ static void ConvertBGRAToRGBA4444_SSE2(const uint32_t* src,
 
 static void ConvertBGRAToRGB565_SSE2(const uint32_t* src,
                                      int num_pixels, uint8_t* dst) {
-  const __m128i mask_0xe0 = _mm_set1_epi8(0xe0);
+  const __m128i mask_0xe0 = _mm_set1_epi8((char)0xe0);
-  const __m128i mask_0xf8 = _mm_set1_epi8(0xf8);
+  const __m128i mask_0xf8 = _mm_set1_epi8((char)0xf8);
   const __m128i mask_0x07 = _mm_set1_epi8(0x07);
   const __m128i* in = (const __m128i*)src;
   __m128i* out = (__m128i*)dst;

src/dsp/lossless_sse41.c

@@ -25,11 +25,12 @@ static void TransformColorInverse_SSE41(const VP8LMultipliers* const m,
                                         int num_pixels, uint32_t* dst) {
 // sign-extended multiplying constants, pre-shifted by 5.
 #define CST(X)  (((int16_t)(m->X << 8)) >> 5)   // sign-extend
-  const __m128i mults_rb = _mm_set1_epi32((uint32_t)CST(green_to_red_) << 16 |
+  const __m128i mults_rb =
-                                          (CST(green_to_blue_) & 0xffff));
+      _mm_set1_epi32((int)((uint32_t)CST(green_to_red_) << 16 |
+                           (CST(green_to_blue_) & 0xffff)));
   const __m128i mults_b2 = _mm_set1_epi32(CST(red_to_blue_));
 #undef CST
-  const __m128i mask_ag = _mm_set1_epi32(0xff00ff00);
+  const __m128i mask_ag = _mm_set1_epi32((int)0xff00ff00);
   const __m128i perm1 = _mm_setr_epi8(-1, 1, -1, 1, -1, 5, -1, 5,
                                       -1, 9, -1, 9, -1, 13, -1, 13);
   const __m128i perm2 = _mm_setr_epi8(-1, 2, -1, -1, -1, 6, -1, -1,

src/dsp/quant.h

@@ -21,10 +21,15 @@
 
 #define IsFlat IsFlat_NEON
 
-static uint32x2_t horizontal_add_uint32x4(const uint32x4_t a) {
+static uint32_t horizontal_add_uint32x4(const uint32x4_t a) {
+#if defined(__aarch64__)
+  return vaddvq_u32(a);
+#else
   const uint64x2_t b = vpaddlq_u32(a);
-  return vadd_u32(vreinterpret_u32_u64(vget_low_u64(b)),
+  const uint32x2_t c = vadd_u32(vreinterpret_u32_u64(vget_low_u64(b)),
-                  vreinterpret_u32_u64(vget_high_u64(b)));
+                                vreinterpret_u32_u64(vget_high_u64(b)));
+  return vget_lane_u32(c, 0);
+#endif
 }
 
 static WEBP_INLINE int IsFlat(const int16_t* levels, int num_blocks,
@@ -45,7 +50,7 @@ static WEBP_INLINE int IsFlat(const int16_t* levels, int num_blocks,
 
     levels += 16;
   }
-  return thresh >= (int32_t)vget_lane_u32(horizontal_add_uint32x4(sum), 0);
+  return thresh >= (int)horizontal_add_uint32x4(sum);
 }
 
 #else