diff --git a/.github/workflows/cmake-tests.yml b/.github/workflows/cmake-tests.yml
new file mode 100644
index 000000000..b02385708
--- /dev/null
+++ b/.github/workflows/cmake-tests.yml
@@ -0,0 +1,153 @@
+name: cmake-tests
+# CMake-specific build and test workflows
+# This workflow validates zstd builds across different CMake configurations,
+# platforms, and edge cases to ensure broad compatibility.
+
+concurrency:
+  group: cmake-${{ github.ref }}
+  cancel-in-progress: true
+
+on:
+  pull_request:
+    branches: [ dev, release, actionsTest ]
+
+permissions: read-all
+
+env:
+  # Centralized test timeouts for consistency
+  QUICK_TEST_TIME: "30s"
+  STANDARD_TEST_TIME: "1mn"
+  # Common CMake flags
+  COMMON_CMAKE_FLAGS: "-DCMAKE_COMPILE_WARNING_AS_ERROR=ON -DZSTD_BUILD_TESTS=ON"
+
+jobs:
+  # Ubuntu-based cmake build using make wrapper
+  # This test uses the make-driven cmake build to ensure compatibility
+  # with the existing build system integration
+  cmake-ubuntu-basic:
+    name: "CMake Ubuntu Basic Build"
+    runs-on: ubuntu-latest
+    steps:
+    - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # tag=v4.2.2
+    - name: Install dependencies
+      run: |
+        sudo apt install liblzma-dev  # Required for compression algorithms
+    - name: CMake build and test via make
+      run: |
+        # Use make wrapper for cmake build with quick test timeouts
+        FUZZERTEST=-T${{ env.STANDARD_TEST_TIME }} ZSTREAM_TESTTIME=-T${{ env.STANDARD_TEST_TIME }} make cmakebuild V=1
+
+  # Cross-platform cmake build with edge case: source paths containing spaces
+  # This test ensures cmake handles filesystem paths with spaces correctly
+  # across different operating systems and build generators
+  cmake-cross-platform-spaces:
+    name: "CMake Cross-Platform (Spaces in Path)"
+    runs-on: ${{ matrix.os }}
+    strategy:
+      matrix:
+        include:
+          - os: ubuntu-latest
+            generator: "Unix Makefiles"
+            name: "Linux"
+          - os: windows-latest
+            generator: "NMake Makefiles"
+            name: "Windows NMake"
+          - os: macos-latest
+            generator: "Unix Makefiles"
+            name: "macOS"
+    env:
+      SRC_DIR: "source directory with spaces"
+    steps:
+    - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # tag=v4.2.2
+      with:
+        path: "${{ env.SRC_DIR }}"
+    - uses: ilammy/msvc-dev-cmd@0b201ec74fa43914dc39ae48a89fd1d8cb592756 # v1.13.0
+      if: ${{ matrix.generator == 'NMake Makefiles' }}
+    - name: "CMake build and install (${{ matrix.name }})"
+      run: |
+        # Test Release build with installation to verify packaging
+        cmake -S "${{ env.SRC_DIR }}/build/cmake" -B build -DBUILD_TESTING=ON -G "${{ matrix.generator }}" -DCMAKE_BUILD_TYPE=Release --install-prefix "${{ runner.temp }}/install"
+        cmake --build build --config Release
+        cmake --install build --config Release
+
+  # Windows-specific cmake testing with Visual Studio 2022
+  # Tests multiple generators and toolchains to ensure broad Windows compatibility
+  # including MSVC (x64, Win32, ARM64), MinGW, and Clang-CL with various architectures and optimizations
+  cmake-windows-comprehensive:
+    name: "CMake Windows VS2022 (${{ matrix.name }})"
+    runs-on: ${{ matrix.runner }}
+    strategy:
+      fail-fast: false
+      matrix:
+        include:
+          - generator: "Visual Studio 17 2022"
+            flags: "-A x64"
+            name: "MSVC x64"
+            runner: "windows-2022"
+            cmake_extra_flags: "-DCMAKE_COMPILE_WARNING_AS_ERROR=ON -DZSTD_BUILD_TESTS=ON"
+          - generator: "Visual Studio 17 2022"
+            flags: "-A Win32"
+            name: "MSVC Win32"
+            runner: "windows-2022"
+            cmake_extra_flags: "-DCMAKE_COMPILE_WARNING_AS_ERROR=ON -DZSTD_BUILD_TESTS=ON"
+          - generator: "Visual Studio 17 2022"
+            flags: "-A x64"
+            name: "MSVC x64 (No ZSTD_BUILD_TESTS)"
+            runner: "windows-2022"
+            # Intentionally omit ZSTD_BUILD_TESTS to reproduce the CXX language configuration bug
+            cmake_extra_flags: "-DCMAKE_COMPILE_WARNING_AS_ERROR=ON"
+          # - generator: "Visual Studio 17 2022"
+          #   flags: "-A ARM64"
+          #   name: "MSVC ARM64"
+          #   runner: "windows-2022-arm64"  # Disabled due to very long queue times
+          - generator: "MinGW Makefiles"
+            flags: ""
+            name: "MinGW"
+            runner: "windows-2022"
+            cmake_extra_flags: "-DCMAKE_COMPILE_WARNING_AS_ERROR=ON -DZSTD_BUILD_TESTS=ON"
+          - generator: "Visual Studio 17 2022"
+            flags: "-T ClangCL"
+            name: "Clang-CL"
+            runner: "windows-2022"
+            cmake_extra_flags: "-DCMAKE_COMPILE_WARNING_AS_ERROR=ON -DZSTD_BUILD_TESTS=ON"
+          - generator: "Visual Studio 17 2022"
+            flags: "-T ClangCL -A x64 -DCMAKE_C_FLAGS=/arch:AVX2"
+            name: "Clang-CL AVX2"
+            runner: "windows-2022"
+            cmake_extra_flags: "-DCMAKE_COMPILE_WARNING_AS_ERROR=ON -DZSTD_BUILD_TESTS=ON"
+    steps:
+    - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # tag=v4.2.2
+    - name: Add MSBuild to PATH
+      uses: microsoft/setup-msbuild@6fb02220983dee41ce7ae257b6f4d8f9bf5ed4ce # tag=v2.0.0
+    - name: "Configure CMake (${{ matrix.name }})"
+      run: |
+        cd build\cmake
+        mkdir build
+        cd build
+        cmake.exe -G "${{matrix.generator}}" ${{matrix.flags}} -DCMAKE_BUILD_TYPE=Debug ${{ matrix.cmake_extra_flags }} -DZSTD_ZSTREAM_FLAGS=-T${{ env.QUICK_TEST_TIME }} -DZSTD_FUZZER_FLAGS=-T${{ env.QUICK_TEST_TIME }} -DZSTD_FULLBENCH_FLAGS=-i0 ..
+    - name: "Build (${{ matrix.name }})"
+      run: |
+        cd build\cmake\build
+        cmake.exe --build .
+    - name: "Test (${{ matrix.name }})"
+      run: |
+        cd build\cmake\build
+        ctest.exe -V -C Debug
+
+  # macOS ARM64 (Apple Silicon) specific cmake testing
+  # Validates zstd builds and runs correctly on Apple Silicon architecture
+  # Uses native ARM64 hardware for optimal performance and compatibility testing
+  cmake-macos-arm64:
+    name: "CMake macOS ARM64 (Apple Silicon)"
+    runs-on: macos-14  # ARM64 runner
+    steps:
+    - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # tag=v4.2.2
+    - name: "CMake build and test (ARM64)"
+      run: |
+        # Configure and build with ARM64-specific optimizations
+        cd build/cmake
+        mkdir build
+        cd build
+        cmake -DCMAKE_BUILD_TYPE=Release ${{ env.COMMON_CMAKE_FLAGS }} -DZSTD_ZSTREAM_FLAGS=-T${{ env.QUICK_TEST_TIME }} -DZSTD_FUZZER_FLAGS=-T${{ env.QUICK_TEST_TIME }} -DZSTD_FULLBENCH_FLAGS=-i1 ..
+        make -j$(sysctl -n hw.ncpu)
+        ctest -V
diff --git a/.github/workflows/dev-short-tests.yml b/.github/workflows/dev-short-tests.yml
index 227b01a0c..31d44542f 100644
--- a/.github/workflows/dev-short-tests.yml
+++ b/.github/workflows/dev-short-tests.yml
@@ -72,40 +72,6 @@ jobs:
     # candidate test (for discussion) : underlink test
     # LDFLAGS=-Wl,--no-undefined : will make the linker fail if dll is underlinked
 
-  cmake-build-and-test-check:
-    runs-on: ubuntu-latest
-    steps:
-    - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # tag=v4.2.2
-    - name: cmake build and test
-      run: |
-        sudo apt install liblzma-dev
-        FUZZERTEST=-T1mn ZSTREAM_TESTTIME=-T1mn make cmakebuild V=1
-
-  cmake-source-directory-with-spaces:
-    runs-on: ${{ matrix.os }}
-    strategy:
-      matrix:
-        include:
-          - os: ubuntu-latest
-            generator: "Unix Makefiles"
-          - os: windows-latest
-            generator: "NMake Makefiles"
-          - os: macos-latest
-            generator: "Unix Makefiles"
-    env:
-      SRC_DIR: "source directory with spaces"
-    steps:
-    - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # tag=v4.2.2
-      with:
-        path: "${{ env.SRC_DIR }}"
-    - uses: ilammy/msvc-dev-cmd@0b201ec74fa43914dc39ae48a89fd1d8cb592756 # v1.13.0
-      if: ${{ matrix.generator == 'NMake Makefiles' }}
-    - name: cmake build on a source directory with spaces
-      run: |
-        cmake -S "${{ env.SRC_DIR }}/build/cmake" -B build -DBUILD_TESTING=ON -G "${{ matrix.generator }}" -DCMAKE_BUILD_TYPE=Release --install-prefix "${{ runner.temp }}/install"
-        cmake --build build --config Release
-        cmake --install build --config Release
-
   cpp-gnu90-c99-compatibility:
     runs-on: ubuntu-latest
     steps:
@@ -339,34 +305,6 @@ jobs:
         run: |
           meson install -C builddir --destdir staging/
 
-  cmake-visual-2022:
-    strategy:
-      matrix:
-        include:
-          - generator: "Visual Studio 17 2022"
-            flags: "-A x64"
-          - generator: "Visual Studio 17 2022"
-            flags: "-A Win32"
-          - generator: "MinGW Makefiles"
-          - generator: "Visual Studio 17 2022"
-            flags: "-T ClangCL"
-          - generator: "Visual Studio 17 2022"
-            flags: "-T ClangCL -A x64 -DCMAKE_C_FLAGS=/arch:AVX2"
-    runs-on: windows-2022
-    steps:
-    - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # tag=v4.2.2
-    - name: Add MSBuild to PATH
-      uses: microsoft/setup-msbuild@6fb02220983dee41ce7ae257b6f4d8f9bf5ed4ce # tag=v2.0.0
-    - name: Build & Test
-      working-directory: ${{env.GITHUB_WORKSPACE}}
-      run: |
-        cd build\cmake
-        mkdir build
-        cd build
-        cmake.exe -G "${{matrix.generator}}" ${{matrix.flags}} -DCMAKE_BUILD_TYPE=Debug -DCMAKE_COMPILE_WARNING_AS_ERROR=ON -DZSTD_BUILD_TESTS:BOOL=ON -DZSTD_ZSTREAM_FLAGS=-T30s -DZSTD_FUZZER_FLAGS=-T30s -DZSTD_FULLBENCH_FLAGS=-i0 ..
-        cmake.exe --build .
-        ctest.exe -V -C Debug
-
   msbuild-visual-studio:
     strategy:
       fail-fast: false  # 'false' means Don't stop matrix workflows even if some matrix failed.
@@ -497,6 +435,8 @@ jobs:
         make clean
         LDFLAGS="-static" CC=$XCC QEMU_SYS=$XEMU make -j check
         LDFLAGS="-static" CC=$XCC QEMU_SYS=$XEMU make -j -C tests test-cli-tests
+        CFLAGS="-march=armv8.2-a+sve2" LDFLAGS="-static" CC=$XCC QEMU_SYS=$XEMU make -j check
+        CFLAGS="-march=armv8.2-a+sve2" LDFLAGS="-static" CC=$XCC QEMU_SYS=$XEMU make -j -C tests test-cli-tests
 # This test is only compatible with standard libraries that support BTI (Branch Target Identification).
 # Unfortunately, the standard library provided on Ubuntu 24.04 does not have this feature enabled.
 #        make clean
diff --git a/build/cmake/CMakeLists.txt b/build/cmake/CMakeLists.txt
index f08905345..753cf4a9e 100644
--- a/build/cmake/CMakeLists.txt
+++ b/build/cmake/CMakeLists.txt
@@ -9,226 +9,73 @@
 
 cmake_minimum_required(VERSION 3.10 FATAL_ERROR)
 
-# As of 2018-12-26 ZSTD has been validated to build with cmake version 3.13.2 new policies.
-# Set and use the newest cmake policies that are validated to work
-set(ZSTD_MAX_VALIDATED_CMAKE_MAJOR_VERSION "3")
-set(ZSTD_MAX_VALIDATED_CMAKE_MINOR_VERSION "13") #Policies never changed at PATCH level
-if("${ZSTD_MAX_VALIDATED_CMAKE_MAJOR_VERSION}" EQUAL "${CMAKE_MAJOR_VERSION}" AND
-       "${ZSTD_MAX_VALIDATED_CMAKE_MINOR_VERSION}" GREATER "${CMAKE_MINOR_VERSION}")
-    set(ZSTD_CMAKE_POLICY_VERSION "${CMAKE_VERSION}")
-else()
-    set(ZSTD_CMAKE_POLICY_VERSION "${ZSTD_MAX_VALIDATED_CMAKE_MAJOR_VERSION}.${ZSTD_MAX_VALIDATED_CMAKE_MINOR_VERSION}.0")
-endif()
-cmake_policy(VERSION ${ZSTD_CMAKE_POLICY_VERSION})
-
-set(CMAKE_BUILD_WITH_INSTALL_RPATH on)
-
+#-----------------------------------------------------------------------------
+# Setup CMake environment
+#-----------------------------------------------------------------------------
+set(CMAKE_BUILD_WITH_INSTALL_RPATH ON)
 list(APPEND CMAKE_MODULE_PATH "${CMAKE_CURRENT_SOURCE_DIR}/CMakeModules")
+
+# Define project paths
 set(ZSTD_SOURCE_DIR "${CMAKE_CURRENT_SOURCE_DIR}/../..")
 set(LIBRARY_DIR ${ZSTD_SOURCE_DIR}/lib)
-# Parse version
-include(GetZstdLibraryVersion)
-GetZstdLibraryVersion(${LIBRARY_DIR}/zstd.h zstd_VERSION_MAJOR zstd_VERSION_MINOR zstd_VERSION_PATCH)
-
-set(ZSTD_SHORT_VERSION "${zstd_VERSION_MAJOR}.${zstd_VERSION_MINOR}")
-set(ZSTD_FULL_VERSION "${zstd_VERSION_MAJOR}.${zstd_VERSION_MINOR}.${zstd_VERSION_PATCH}")
-
-project(zstd
-  VERSION "${ZSTD_FULL_VERSION}"
-  LANGUAGES C   # Main library is in C
-            ASM # And ASM
-  )
-
-message(STATUS "ZSTD VERSION: ${zstd_VERSION}")
-set(zstd_HOMEPAGE_URL "https://facebook.github.io/zstd")
-set(zstd_DESCRIPTION  "Zstandard is a real-time compression algorithm, providing high compression ratios.")
-
-# Set a default build type if none was specified
-if(NOT CMAKE_BUILD_TYPE AND NOT CMAKE_CONFIGURATION_TYPES)
-  message(STATUS "Setting build type to 'Release' as none was specified.")
-  set(CMAKE_BUILD_TYPE Release CACHE STRING "Choose the type of build." FORCE)
-  # Set the possible values of build type for cmake-gui
-  set_property(CACHE CMAKE_BUILD_TYPE PROPERTY STRINGS "Debug" "Release" "MinSizeRel" "RelWithDebInfo")
-endif()
-
-include(GNUInstallDirs)
-
-# Always hide XXHash symbols
-add_definitions(-DXXH_NAMESPACE=ZSTD_)
 
 #-----------------------------------------------------------------------------
-# Installation variables
+# Configure CMake policies and version
+#-----------------------------------------------------------------------------
+include(ZstdVersion)
+
+#-----------------------------------------------------------------------------
+# Project declaration
+#-----------------------------------------------------------------------------
+project(zstd
+    VERSION "${ZSTD_FULL_VERSION}"
+    LANGUAGES C ASM  # Main library is in C and ASM
+    HOMEPAGE_URL "${zstd_HOMEPAGE_URL}"
+    DESCRIPTION "${zstd_DESCRIPTION}"
+)
+
+#-----------------------------------------------------------------------------
+# Build type configuration
+#-----------------------------------------------------------------------------
+if(NOT CMAKE_BUILD_TYPE AND NOT CMAKE_CONFIGURATION_TYPES)
+    message(STATUS "Setting build type to 'Release' as none was specified.")
+    set(CMAKE_BUILD_TYPE Release CACHE STRING "Choose the type of build." FORCE)
+    set_property(CACHE CMAKE_BUILD_TYPE PROPERTY STRINGS "Debug" "Release" "MinSizeRel" "RelWithDebInfo")
+endif()
+
+#-----------------------------------------------------------------------------
+# Include standard modules
+#-----------------------------------------------------------------------------
+include(GNUInstallDirs)
+
+#-----------------------------------------------------------------------------
+# Display installation information
 #-----------------------------------------------------------------------------
 message(STATUS "CMAKE_INSTALL_PREFIX: ${CMAKE_INSTALL_PREFIX}")
 message(STATUS "CMAKE_INSTALL_LIBDIR: ${CMAKE_INSTALL_LIBDIR}")
 
 #-----------------------------------------------------------------------------
-# Options
+# Configure build options
 #-----------------------------------------------------------------------------
-
-# Legacy support
-option(ZSTD_LEGACY_SUPPORT "LEGACY SUPPORT" ON)
-
-if (ZSTD_LEGACY_SUPPORT)
-    message(STATUS "ZSTD_LEGACY_SUPPORT defined!")
-    set(ZSTD_LEGACY_LEVEL 5 CACHE STRING "")
-    add_definitions(-DZSTD_LEGACY_SUPPORT=${ZSTD_LEGACY_LEVEL})
-else ()
-    message(STATUS "ZSTD_LEGACY_SUPPORT not defined!")
-    add_definitions(-DZSTD_LEGACY_SUPPORT=0)
-endif ()
-
-if (APPLE)
-    option(ZSTD_FRAMEWORK "Build as Apple Frameworks" OFF)
-endif ()
-
-if (ANDROID)
-    set(ZSTD_MULTITHREAD_SUPPORT_DEFAULT OFF)
-    # Old versions of bionic libc don't have fseeko/ftello
-    if ((NOT ${ANDROID_PLATFORM_LEVEL}) OR ${ANDROID_PLATFORM_LEVEL} VERSION_LESS 24)
-        message(STATUS "Setting compile definitions for old Android API")
-        add_compile_definitions(LIBC_NO_FSEEKO)
-    endif ()
-else()
-    set(ZSTD_MULTITHREAD_SUPPORT_DEFAULT ON)
-endif()
-
-# Multi-threading support
-option(ZSTD_MULTITHREAD_SUPPORT "MULTITHREADING SUPPORT" ${ZSTD_MULTITHREAD_SUPPORT_DEFAULT})
-
-if (ZSTD_MULTITHREAD_SUPPORT)
-    message(STATUS "ZSTD_MULTITHREAD_SUPPORT is enabled")
-else ()
-    message(STATUS "ZSTD_MULTITHREAD_SUPPORT is disabled")
-endif ()
-
-option(ZSTD_BUILD_PROGRAMS "BUILD PROGRAMS" ON)
-option(ZSTD_BUILD_CONTRIB "BUILD CONTRIB" OFF)
-
-# Respect the conventional CMake option for enabling tests if it was specified on the first configure
-if (BUILD_TESTING)
-    set(ZSTD_BUILD_TESTS_default ON)
-else()
-    set(ZSTD_BUILD_TESTS_default OFF)
-endif()
-option(ZSTD_BUILD_TESTS "BUILD TESTS" ${ZSTD_BUILD_TESTS_default})
-if (MSVC)
-    option(ZSTD_USE_STATIC_RUNTIME "LINK TO STATIC RUN-TIME LIBRARIES" OFF)
-endif ()
-
-# Enable C++ support for testing.
-set(ZSTD_ENABLE_CXX ${ZSTD_BUILD_TESTS})
-
-if(ZSTD_ENABLE_CXX)
-    enable_language(CXX)
-endif()
+include(ZstdOptions)
 
 #-----------------------------------------------------------------------------
-# Add extra compilation flags
+# Configure compilation flags
 #-----------------------------------------------------------------------------
 include(AddZstdCompilationFlags)
-ADD_ZSTD_COMPILATION_FLAGS(ON ZSTD_ENABLE_CXX ON) # C CXX LD
+ADD_ZSTD_COMPILATION_FLAGS(ON ZSTD_ENABLE_CXX ON)
 
 #-----------------------------------------------------------------------------
-# External dependencies
+# Configure dependencies
 #-----------------------------------------------------------------------------
-# Define a function to handle special thread settings for HP-UX
-# See https://github.com/facebook/zstd/pull/3862 for details.
-function(setup_hpux_threads)
-    find_package(Threads)
-    if (NOT Threads_FOUND)
-        set(CMAKE_USE_PTHREADS_INIT 1 PARENT_SCOPE)
-        set(CMAKE_THREAD_LIBS_INIT -lpthread PARENT_SCOPE)
-        set(CMAKE_HAVE_THREADS_LIBRARY 1 PARENT_SCOPE)
-        set(Threads_FOUND TRUE PARENT_SCOPE)
-    endif()
-endfunction()
-
-if (ZSTD_MULTITHREAD_SUPPORT AND UNIX)
-    if (CMAKE_SYSTEM_NAME MATCHES "HP-UX")
-        setup_hpux_threads()
-    else()
-        set(THREADS_PREFER_PTHREAD_FLAG ON)
-        find_package(Threads REQUIRED)
-    endif()
-    if (CMAKE_USE_PTHREADS_INIT)
-        set(THREADS_LIBS "${CMAKE_THREAD_LIBS_INIT}")
-    else()
-        message(SEND_ERROR "ZSTD currently does not support thread libraries other than pthreads")
-    endif()
-endif ()
+include(ZstdDependencies)
 
 #-----------------------------------------------------------------------------
-# Add source directories
+# Configure build targets
 #-----------------------------------------------------------------------------
-add_subdirectory(lib)
-
-option(ZSTD_PROGRAMS_LINK_SHARED "PROGRAMS LINK SHARED" OFF)
-
-if (ZSTD_BUILD_PROGRAMS)
-    if (NOT ZSTD_BUILD_STATIC AND NOT ZSTD_PROGRAMS_LINK_SHARED)
-        message(SEND_ERROR "You need to build static library to build zstd CLI")
-    elseif(NOT ZSTD_BUILD_SHARED AND ZSTD_PROGRAMS_LINK_SHARED)
-        message(SEND_ERROR "You need to build shared library to build zstd CLI")
-    endif ()
-
-    add_subdirectory(programs)
-endif ()
-
-if (ZSTD_BUILD_TESTS)
-    enable_testing()
-    if (NOT ZSTD_BUILD_STATIC)
-        message(SEND_ERROR "You need to build static library to build tests")
-    endif ()
-
-    add_subdirectory(tests)
-endif ()
-
-if (ZSTD_BUILD_CONTRIB)
-    add_subdirectory(contrib)
-endif ()
+include(ZstdBuild)
 
 #-----------------------------------------------------------------------------
-# Add clean-all target
+# Configure package generation
 #-----------------------------------------------------------------------------
-add_custom_target(clean-all
-   COMMAND ${CMAKE_BUILD_TOOL} clean
-   COMMAND rm -rf ${CMAKE_BINARY_DIR}/
-)
-
-#-----------------------------------------------------------------------------
-# Generate Package Config files
-#
-# This section is based on the boiler plate code from:
-# https://cmake.org/cmake/help/latest/manual/cmake-packages.7.html#creating-packages
-#-----------------------------------------------------------------------------
-include(CMakePackageConfigHelpers)
-write_basic_package_version_file(
-    "${CMAKE_CURRENT_BINARY_DIR}/zstdConfigVersion.cmake"
-    VERSION ${zstd_VERSION}
-    COMPATIBILITY SameMajorVersion
-    )
-
-# A Package Config file that works from the build directory
-export(EXPORT zstdExports
-    FILE "${CMAKE_CURRENT_BINARY_DIR}/zstdTargets.cmake"
-    NAMESPACE zstd::
-    )
-
-# A Package Config file that works from the installation directory
-set(ConfigPackageLocation ${CMAKE_INSTALL_LIBDIR}/cmake/zstd)
-install(EXPORT zstdExports
-    FILE zstdTargets.cmake
-    NAMESPACE zstd::
-    DESTINATION ${ConfigPackageLocation}
-    )
-configure_package_config_file(
-    zstdConfig.cmake.in
-    "${CMAKE_CURRENT_BINARY_DIR}/zstdConfig.cmake"
-    INSTALL_DESTINATION ${ConfigPackageLocation}
-)
-install(FILES
-    "${CMAKE_CURRENT_BINARY_DIR}/zstdConfig.cmake"
-    "${CMAKE_CURRENT_BINARY_DIR}/zstdConfigVersion.cmake"
-    DESTINATION ${ConfigPackageLocation}
-    )
+include(ZstdPackage)
diff --git a/build/cmake/CMakeModules/AddZstdCompilationFlags.cmake b/build/cmake/CMakeModules/AddZstdCompilationFlags.cmake
index 25231b5ec..3e7bcce59 100644
--- a/build/cmake/CMakeModules/AddZstdCompilationFlags.cmake
+++ b/build/cmake/CMakeModules/AddZstdCompilationFlags.cmake
@@ -1,5 +1,7 @@
-include(CheckCXXCompilerFlag)
 include(CheckCCompilerFlag)
+if(CMAKE_CXX_COMPILER)
+    include(CheckCXXCompilerFlag)
+endif()
 
 if (CMAKE_VERSION VERSION_GREATER_EQUAL 3.18)
     set(ZSTD_HAVE_CHECK_LINKER_FLAG true)
@@ -21,7 +23,7 @@ function(EnableCompilerFlag _flag _C _CXX _LD)
             set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} ${_flag}" PARENT_SCOPE)
         endif ()
     endif ()
-    if (_CXX)
+    if (_CXX AND CMAKE_CXX_COMPILER)
         CHECK_CXX_COMPILER_FLAG(${_flag} CXX_FLAG_${varname})
         if (CXX_FLAG_${varname})
             set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} ${_flag}" PARENT_SCOPE)
diff --git a/build/cmake/CMakeModules/ZstdBuild.cmake b/build/cmake/CMakeModules/ZstdBuild.cmake
new file mode 100644
index 000000000..ada44a966
--- /dev/null
+++ b/build/cmake/CMakeModules/ZstdBuild.cmake
@@ -0,0 +1,42 @@
+# ################################################################
+# ZSTD Build Targets Configuration
+# ################################################################
+
+# Always build the library first (this defines ZSTD_BUILD_STATIC/SHARED options)
+add_subdirectory(lib)
+
+# Validate build configuration after lib options are defined
+if(ZSTD_BUILD_PROGRAMS)
+    if(NOT ZSTD_BUILD_STATIC AND NOT ZSTD_PROGRAMS_LINK_SHARED)
+        message(SEND_ERROR "Static library required to build zstd CLI programs")
+    elseif(NOT ZSTD_BUILD_SHARED AND ZSTD_PROGRAMS_LINK_SHARED)
+        message(SEND_ERROR "Shared library required to build zstd CLI programs")
+    endif()
+endif()
+
+if(ZSTD_BUILD_TESTS AND NOT ZSTD_BUILD_STATIC)
+    message(SEND_ERROR "Static library required to build test suite")
+endif()
+
+# Add programs if requested
+if(ZSTD_BUILD_PROGRAMS)
+    add_subdirectory(programs)
+endif()
+
+# Add tests if requested
+if(ZSTD_BUILD_TESTS)
+    enable_testing()
+    add_subdirectory(tests)
+endif()
+
+# Add contrib utilities if requested
+if(ZSTD_BUILD_CONTRIB)
+    add_subdirectory(contrib)
+endif()
+
+# Clean-all target for thorough cleanup
+add_custom_target(clean-all
+    COMMAND ${CMAKE_BUILD_TOOL} clean
+    COMMAND ${CMAKE_COMMAND} -E remove_directory ${CMAKE_BINARY_DIR}/
+    COMMENT "Performing complete clean including build directory"
+)
diff --git a/build/cmake/CMakeModules/ZstdDependencies.cmake b/build/cmake/CMakeModules/ZstdDependencies.cmake
new file mode 100644
index 000000000..4e90c5f1b
--- /dev/null
+++ b/build/cmake/CMakeModules/ZstdDependencies.cmake
@@ -0,0 +1,30 @@
+# ################################################################
+# ZSTD Dependencies Configuration
+# ################################################################
+
+# Function to handle HP-UX thread configuration
+function(setup_hpux_threads)
+    find_package(Threads)
+    if(NOT Threads_FOUND)
+        set(CMAKE_USE_PTHREADS_INIT 1 PARENT_SCOPE)
+        set(CMAKE_THREAD_LIBS_INIT -lpthread PARENT_SCOPE)
+        set(CMAKE_HAVE_THREADS_LIBRARY 1 PARENT_SCOPE)
+        set(Threads_FOUND TRUE PARENT_SCOPE)
+    endif()
+endfunction()
+
+# Configure threading support
+if(ZSTD_MULTITHREAD_SUPPORT AND UNIX)
+    if(CMAKE_SYSTEM_NAME MATCHES "HP-UX")
+        setup_hpux_threads()
+    else()
+        set(THREADS_PREFER_PTHREAD_FLAG ON)
+        find_package(Threads REQUIRED)
+    endif()
+    
+    if(CMAKE_USE_PTHREADS_INIT)
+        set(THREADS_LIBS "${CMAKE_THREAD_LIBS_INIT}")
+    else()
+        message(SEND_ERROR "ZSTD currently does not support thread libraries other than pthreads")
+    endif()
+endif()
diff --git a/build/cmake/CMakeModules/ZstdOptions.cmake b/build/cmake/CMakeModules/ZstdOptions.cmake
new file mode 100644
index 000000000..3fca543a7
--- /dev/null
+++ b/build/cmake/CMakeModules/ZstdOptions.cmake
@@ -0,0 +1,68 @@
+# ################################################################
+# ZSTD Build Options Configuration
+# ################################################################
+
+# Legacy support configuration
+option(ZSTD_LEGACY_SUPPORT "Enable legacy format support" ON)
+
+if(ZSTD_LEGACY_SUPPORT)
+    message(STATUS "ZSTD_LEGACY_SUPPORT enabled")
+    set(ZSTD_LEGACY_LEVEL 5 CACHE STRING "Legacy support level")
+    add_definitions(-DZSTD_LEGACY_SUPPORT=${ZSTD_LEGACY_LEVEL})
+else()
+    message(STATUS "ZSTD_LEGACY_SUPPORT disabled")
+    add_definitions(-DZSTD_LEGACY_SUPPORT=0)
+endif()
+
+# Platform-specific options
+if(APPLE)
+    option(ZSTD_FRAMEWORK "Build as Apple Framework" OFF)
+endif()
+
+# Android-specific configuration
+if(ANDROID)
+    set(ZSTD_MULTITHREAD_SUPPORT_DEFAULT OFF)
+    # Handle old Android API levels
+    if((NOT ANDROID_PLATFORM_LEVEL) OR (ANDROID_PLATFORM_LEVEL VERSION_LESS 24))
+        message(STATUS "Configuring for old Android API - disabling fseeko/ftello")
+        add_compile_definitions(LIBC_NO_FSEEKO)
+    endif()
+else()
+    set(ZSTD_MULTITHREAD_SUPPORT_DEFAULT ON)
+endif()
+
+# Multi-threading support
+option(ZSTD_MULTITHREAD_SUPPORT "Enable multi-threading support" ${ZSTD_MULTITHREAD_SUPPORT_DEFAULT})
+
+if(ZSTD_MULTITHREAD_SUPPORT)
+    message(STATUS "Multi-threading support enabled")
+else()
+    message(STATUS "Multi-threading support disabled")
+endif()
+
+# Build component options
+option(ZSTD_BUILD_PROGRAMS "Build command-line programs" ON)
+option(ZSTD_BUILD_CONTRIB "Build contrib utilities" OFF)
+option(ZSTD_PROGRAMS_LINK_SHARED "Link programs against shared library" OFF)
+
+# Test configuration
+if(BUILD_TESTING)
+    set(ZSTD_BUILD_TESTS_default ON)
+else()
+    set(ZSTD_BUILD_TESTS_default OFF)
+endif()
+option(ZSTD_BUILD_TESTS "Build test suite" ${ZSTD_BUILD_TESTS_default})
+
+# MSVC-specific options
+if(MSVC)
+    option(ZSTD_USE_STATIC_RUNTIME "Link to static runtime libraries" OFF)
+endif()
+
+# C++ support (needed for tests)
+set(ZSTD_ENABLE_CXX ${ZSTD_BUILD_TESTS})
+if(ZSTD_ENABLE_CXX)
+    enable_language(CXX)
+endif()
+
+# Set global definitions
+add_definitions(-DXXH_NAMESPACE=ZSTD_)
diff --git a/build/cmake/CMakeModules/ZstdPackage.cmake b/build/cmake/CMakeModules/ZstdPackage.cmake
new file mode 100644
index 000000000..5e40dd2c1
--- /dev/null
+++ b/build/cmake/CMakeModules/ZstdPackage.cmake
@@ -0,0 +1,42 @@
+# ################################################################
+# ZSTD Package Configuration
+# ################################################################
+
+include(CMakePackageConfigHelpers)
+
+# Generate version file
+write_basic_package_version_file(
+    "${CMAKE_CURRENT_BINARY_DIR}/zstdConfigVersion.cmake"
+    VERSION ${zstd_VERSION}
+    COMPATIBILITY SameMajorVersion
+)
+
+# Export targets for build directory
+export(EXPORT zstdExports
+    FILE "${CMAKE_CURRENT_BINARY_DIR}/zstdTargets.cmake"
+    NAMESPACE zstd::
+)
+
+# Configure package for installation
+set(ConfigPackageLocation ${CMAKE_INSTALL_LIBDIR}/cmake/zstd)
+
+# Install exported targets
+install(EXPORT zstdExports
+    FILE zstdTargets.cmake
+    NAMESPACE zstd::
+    DESTINATION ${ConfigPackageLocation}
+)
+
+# Configure and install package config file
+configure_package_config_file(
+    zstdConfig.cmake.in
+    "${CMAKE_CURRENT_BINARY_DIR}/zstdConfig.cmake"
+    INSTALL_DESTINATION ${ConfigPackageLocation}
+)
+
+# Install config files
+install(FILES
+    "${CMAKE_CURRENT_BINARY_DIR}/zstdConfig.cmake"
+    "${CMAKE_CURRENT_BINARY_DIR}/zstdConfigVersion.cmake"
+    DESTINATION ${ConfigPackageLocation}
+)
diff --git a/build/cmake/CMakeModules/ZstdVersion.cmake b/build/cmake/CMakeModules/ZstdVersion.cmake
new file mode 100644
index 000000000..fceb0ec0c
--- /dev/null
+++ b/build/cmake/CMakeModules/ZstdVersion.cmake
@@ -0,0 +1,31 @@
+# ################################################################
+# ZSTD Version Configuration
+# ################################################################
+
+# Setup CMake policy version
+set(ZSTD_MAX_VALIDATED_CMAKE_MAJOR_VERSION "3")
+set(ZSTD_MAX_VALIDATED_CMAKE_MINOR_VERSION "13")
+
+# Determine appropriate policy version
+if("${ZSTD_MAX_VALIDATED_CMAKE_MAJOR_VERSION}" EQUAL "${CMAKE_MAJOR_VERSION}" AND
+   "${ZSTD_MAX_VALIDATED_CMAKE_MINOR_VERSION}" GREATER "${CMAKE_MINOR_VERSION}")
+    set(ZSTD_CMAKE_POLICY_VERSION "${CMAKE_VERSION}")
+else()
+    set(ZSTD_CMAKE_POLICY_VERSION "${ZSTD_MAX_VALIDATED_CMAKE_MAJOR_VERSION}.${ZSTD_MAX_VALIDATED_CMAKE_MINOR_VERSION}.0")
+endif()
+
+cmake_policy(VERSION ${ZSTD_CMAKE_POLICY_VERSION})
+
+# Parse version from header file
+include(GetZstdLibraryVersion)
+GetZstdLibraryVersion(${LIBRARY_DIR}/zstd.h zstd_VERSION_MAJOR zstd_VERSION_MINOR zstd_VERSION_PATCH)
+
+# Set version variables
+set(ZSTD_SHORT_VERSION "${zstd_VERSION_MAJOR}.${zstd_VERSION_MINOR}")
+set(ZSTD_FULL_VERSION "${zstd_VERSION_MAJOR}.${zstd_VERSION_MINOR}.${zstd_VERSION_PATCH}")
+
+# Project metadata
+set(zstd_HOMEPAGE_URL "https://facebook.github.io/zstd")
+set(zstd_DESCRIPTION "Zstandard is a real-time compression algorithm, providing high compression ratios.")
+
+message(STATUS "ZSTD VERSION: ${zstd_VERSION_MAJOR}.${zstd_VERSION_MINOR}.${zstd_VERSION_PATCH}")
diff --git a/build/cmake/lib/CMakeLists.txt b/build/cmake/lib/CMakeLists.txt
index 0a93e8d31..a6b894c4d 100644
--- a/build/cmake/lib/CMakeLists.txt
+++ b/build/cmake/lib/CMakeLists.txt
@@ -97,9 +97,11 @@ if (ZSTD_LEGACY_SUPPORT)
             ${LIBRARY_LEGACY_DIR}/zstd_v07.h)
 endif ()
 
-if (MSVC)
+if (MSVC AND NOT (CMAKE_CXX_COMPILER_ID STREQUAL "Clang"))
     set(MSVC_RESOURCE_DIR ${ZSTD_SOURCE_DIR}/build/VS2010/libzstd-dll)
     set(PlatformDependResources ${MSVC_RESOURCE_DIR}/libzstd-dll.rc)
+else()
+    set(PlatformDependResources)
 endif ()
 
 # Explicitly set the language to C for all files, including ASM files.
diff --git a/build/meson/meson.build b/build/meson/meson.build
index 5c35478d5..bae4d2501 100644
--- a/build/meson/meson.build
+++ b/build/meson/meson.build
@@ -10,7 +10,7 @@
 
 project('zstd',
   ['c', 'cpp'],
-  license: ['BSD', 'GPLv2'],
+  license: 'BSD-3-Clause OR GPL-2.0-only',
   default_options : [
     # There shouldn't be any need to force a C standard convention for zstd
     # but in case one would want that anyway, this can be done here.
diff --git a/contrib/largeNbDicts/largeNbDicts.c b/contrib/largeNbDicts/largeNbDicts.c
index 6502f1227..6b6cc34ca 100644
--- a/contrib/largeNbDicts/largeNbDicts.c
+++ b/contrib/largeNbDicts/largeNbDicts.c
@@ -739,6 +739,8 @@ static int benchMem(slice_collection_t dstBlocks, slice_collection_t srcBlocks,
     /* BMK_benchTimedFn may not run exactly nbRounds iterations */
     double speedAggregated =
         aggregateData(speedPerRound, roundNb + 1, metricAggregatePref);
+    free(speedPerRound);
+    
     if (metricAggregatePref == fastest)
       DISPLAY("Fastest Speed : %.1f MB/s \n", speedAggregated);
     else
diff --git a/contrib/seekable_format/tests/seekable_tests.c b/contrib/seekable_format/tests/seekable_tests.c
index f89bdc930..809ea6425 100644
--- a/contrib/seekable_format/tests/seekable_tests.c
+++ b/contrib/seekable_format/tests/seekable_tests.c
@@ -316,6 +316,10 @@ int main(int argc, const char** argv)
         for (pos = 0; pos < inSize; pos += 2) {
             size_t const decStatus = ZSTD_seekable_decompress(stream, outBuffer, 1, pos);
             if (decStatus != 1 || outBuffer[0] != inBuffer[pos]) {
+                free(seekBuffer);
+                free(outBuffer);
+                ZSTD_seekable_freeCStream(zscs);
+                ZSTD_seekable_free(stream);
                 goto _test_error;
             }
         }
@@ -323,6 +327,10 @@ int main(int argc, const char** argv)
             /* We read more than the compressed size, meaning there were some rereads.
                This is unneeded because we only seeked forward. */
             printf("Too much data read: %zu read, with compressed size %zu\n", buffWrapper.totalRead, seekSize);
+            free(seekBuffer);
+            free(outBuffer);
+            ZSTD_seekable_freeCStream(zscs);
+            ZSTD_seekable_free(stream);
             goto _test_error;
         }
 
@@ -342,6 +350,10 @@ int main(int argc, const char** argv)
         for (idx = 0; idx < sizeof(tests) / sizeof(tests[0]); idx++) {
             size_t const decStatus = ZSTD_seekable_decompress(stream, outBuffer, tests[idx].size, tests[idx].offset);
             if (decStatus != tests[idx].size || memcmp(outBuffer, inBuffer + tests[idx].offset, tests[idx].size) != 0) {
+                free(seekBuffer);
+                free(outBuffer);
+                ZSTD_seekable_freeCStream(zscs);
+                ZSTD_seekable_free(stream);
                 goto _test_error;
             }
         }
diff --git a/lib/common/compiler.h b/lib/common/compiler.h
index 944774a7a..6131ad068 100644
--- a/lib/common/compiler.h
+++ b/lib/common/compiler.h
@@ -218,6 +218,15 @@
 #  if defined(__ARM_NEON) || defined(_M_ARM64)
 #    define ZSTD_ARCH_ARM_NEON
 #  endif
+#  if defined(__ARM_FEATURE_SVE)
+#    define ZSTD_ARCH_ARM_SVE
+#  endif
+#  if defined(__ARM_FEATURE_SVE2)
+#    define ZSTD_ARCH_ARM_SVE2
+#  endif
+# if defined(__riscv) && defined(__riscv_vector)
+#   define ZSTD_ARCH_RISCV_RVV
+# endif
 #
 #  if defined(ZSTD_ARCH_X86_AVX2)
 #    include <immintrin.h>
@@ -227,6 +236,12 @@
 #  elif defined(ZSTD_ARCH_ARM_NEON)
 #    include <arm_neon.h>
 #  endif
+#  if defined(ZSTD_ARCH_ARM_SVE) || defined(ZSTD_ARCH_ARM_SVE2)
+#    include <arm_sve.h>
+#  endif
+#  if defined(ZSTD_ARCH_RISCV_RVV)
+#    include <riscv_vector.h>
+#  endif
 #endif
 
 /* C-language Attributes are added in C23. */
diff --git a/lib/compress/hist.c b/lib/compress/hist.c
index 4ccf9a90a..3692bc250 100644
--- a/lib/compress/hist.c
+++ b/lib/compress/hist.c
@@ -19,6 +19,12 @@
 #include "../common/error_private.h"   /* ERROR */
 #include "hist.h"
 
+#if defined(ZSTD_ARCH_ARM_SVE2)
+#define HIST_FAST_THRESHOLD 500
+#else
+#define HIST_FAST_THRESHOLD 1500
+#endif
+
 
 /* --- Error management --- */
 unsigned HIST_isError(size_t code) { return ERR_isError(code); }
@@ -65,6 +71,244 @@ unsigned HIST_count_simple(unsigned* count, unsigned* maxSymbolValuePtr,
 
 typedef enum { trustInput, checkMaxSymbolValue } HIST_checkInput_e;
 
+#if defined(ZSTD_ARCH_ARM_SVE2)
+FORCE_INLINE_TEMPLATE size_t min_size(size_t a, size_t b) { return a < b ? a : b; }
+
+static
+svuint16_t HIST_count_6_sve2(const BYTE* const src, size_t size, U32* const dst,
+                             const svuint8_t c0, const svuint8_t c1,
+                             const svuint8_t c2, const svuint8_t c3,
+                             const svuint8_t c4, const svuint8_t c5,
+                             const svuint16_t histmax, size_t maxCount)
+{
+    const svbool_t vl128 = svptrue_pat_b8(SV_VL16);
+    svuint16_t hh0 = svdup_n_u16(0);
+    svuint16_t hh1 = svdup_n_u16(0);
+    svuint16_t hh2 = svdup_n_u16(0);
+    svuint16_t hh3 = svdup_n_u16(0);
+    svuint16_t hh4 = svdup_n_u16(0);
+    svuint16_t hh5 = svdup_n_u16(0);
+    svuint16_t hh6 = svdup_n_u16(0);
+    svuint16_t hh7 = svdup_n_u16(0);
+    svuint16_t hh8 = svdup_n_u16(0);
+    svuint16_t hh9 = svdup_n_u16(0);
+    svuint16_t hha = svdup_n_u16(0);
+    svuint16_t hhb = svdup_n_u16(0);
+
+    size_t i = 0;
+    while (i < size) {
+        /* We can only accumulate 15 (15 * 16 <= 255) iterations of histogram
+         * in 8-bit accumulators! */
+        const size_t size240 = min_size(i + 240, size);
+
+        svbool_t pred = svwhilelt_b8_u64(i, size);
+        svuint8_t c = svld1rq_u8(pred, src + i);
+        svuint8_t h0 = svhistseg_u8(c0, c);
+        svuint8_t h1 = svhistseg_u8(c1, c);
+        svuint8_t h2 = svhistseg_u8(c2, c);
+        svuint8_t h3 = svhistseg_u8(c3, c);
+        svuint8_t h4 = svhistseg_u8(c4, c);
+        svuint8_t h5 = svhistseg_u8(c5, c);
+
+        for (i += 16; i < size240; i += 16) {
+            pred = svwhilelt_b8_u64(i, size);
+            c = svld1rq_u8(pred, src + i);
+            h0 = svadd_u8_x(vl128, h0, svhistseg_u8(c0, c));
+            h1 = svadd_u8_x(vl128, h1, svhistseg_u8(c1, c));
+            h2 = svadd_u8_x(vl128, h2, svhistseg_u8(c2, c));
+            h3 = svadd_u8_x(vl128, h3, svhistseg_u8(c3, c));
+            h4 = svadd_u8_x(vl128, h4, svhistseg_u8(c4, c));
+            h5 = svadd_u8_x(vl128, h5, svhistseg_u8(c5, c));
+        }
+
+        hh0 = svaddwb_u16(hh0, h0);
+        hh1 = svaddwt_u16(hh1, h0);
+        hh2 = svaddwb_u16(hh2, h1);
+        hh3 = svaddwt_u16(hh3, h1);
+        hh4 = svaddwb_u16(hh4, h2);
+        hh5 = svaddwt_u16(hh5, h2);
+        hh6 = svaddwb_u16(hh6, h3);
+        hh7 = svaddwt_u16(hh7, h3);
+        hh8 = svaddwb_u16(hh8, h4);
+        hh9 = svaddwt_u16(hh9, h4);
+        hha = svaddwb_u16(hha, h5);
+        hhb = svaddwt_u16(hhb, h5);
+    }
+
+    svst1_u32(svwhilelt_b32_u64( 0, maxCount), dst +  0, svshllb_n_u32(hh0, 0));
+    svst1_u32(svwhilelt_b32_u64( 4, maxCount), dst +  4, svshllt_n_u32(hh0, 0));
+    svst1_u32(svwhilelt_b32_u64( 8, maxCount), dst +  8, svshllb_n_u32(hh1, 0));
+    svst1_u32(svwhilelt_b32_u64(12, maxCount), dst + 12, svshllt_n_u32(hh1, 0));
+    svst1_u32(svwhilelt_b32_u64(16, maxCount), dst + 16, svshllb_n_u32(hh2, 0));
+    svst1_u32(svwhilelt_b32_u64(20, maxCount), dst + 20, svshllt_n_u32(hh2, 0));
+    svst1_u32(svwhilelt_b32_u64(24, maxCount), dst + 24, svshllb_n_u32(hh3, 0));
+    svst1_u32(svwhilelt_b32_u64(28, maxCount), dst + 28, svshllt_n_u32(hh3, 0));
+    svst1_u32(svwhilelt_b32_u64(32, maxCount), dst + 32, svshllb_n_u32(hh4, 0));
+    svst1_u32(svwhilelt_b32_u64(36, maxCount), dst + 36, svshllt_n_u32(hh4, 0));
+    svst1_u32(svwhilelt_b32_u64(40, maxCount), dst + 40, svshllb_n_u32(hh5, 0));
+    svst1_u32(svwhilelt_b32_u64(44, maxCount), dst + 44, svshllt_n_u32(hh5, 0));
+    svst1_u32(svwhilelt_b32_u64(48, maxCount), dst + 48, svshllb_n_u32(hh6, 0));
+    svst1_u32(svwhilelt_b32_u64(52, maxCount), dst + 52, svshllt_n_u32(hh6, 0));
+    svst1_u32(svwhilelt_b32_u64(56, maxCount), dst + 56, svshllb_n_u32(hh7, 0));
+    svst1_u32(svwhilelt_b32_u64(60, maxCount), dst + 60, svshllt_n_u32(hh7, 0));
+    svst1_u32(svwhilelt_b32_u64(64, maxCount), dst + 64, svshllb_n_u32(hh8, 0));
+    svst1_u32(svwhilelt_b32_u64(68, maxCount), dst + 68, svshllt_n_u32(hh8, 0));
+    svst1_u32(svwhilelt_b32_u64(72, maxCount), dst + 72, svshllb_n_u32(hh9, 0));
+    svst1_u32(svwhilelt_b32_u64(76, maxCount), dst + 76, svshllt_n_u32(hh9, 0));
+    svst1_u32(svwhilelt_b32_u64(80, maxCount), dst + 80, svshllb_n_u32(hha, 0));
+    svst1_u32(svwhilelt_b32_u64(84, maxCount), dst + 84, svshllt_n_u32(hha, 0));
+    svst1_u32(svwhilelt_b32_u64(88, maxCount), dst + 88, svshllb_n_u32(hhb, 0));
+    svst1_u32(svwhilelt_b32_u64(92, maxCount), dst + 92, svshllt_n_u32(hhb, 0));
+
+    hh0 = svmax_u16_x(vl128, hh0, hh1);
+    hh2 = svmax_u16_x(vl128, hh2, hh3);
+    hh4 = svmax_u16_x(vl128, hh4, hh5);
+    hh6 = svmax_u16_x(vl128, hh6, hh7);
+    hh8 = svmax_u16_x(vl128, hh8, hh9);
+    hha = svmax_u16_x(vl128, hha, hhb);
+    hh0 = svmax_u16_x(vl128, hh0, hh2);
+    hh4 = svmax_u16_x(vl128, hh4, hh6);
+    hh8 = svmax_u16_x(vl128, hh8, hha);
+    hh0 = svmax_u16_x(vl128, hh0, hh4);
+    hh8 = svmax_u16_x(vl128, hh8, histmax);
+    return svmax_u16_x(vl128, hh0, hh8);
+}
+
+static size_t HIST_count_sve2(unsigned* count, unsigned* maxSymbolValuePtr,
+                              const void* source, size_t sourceSize,
+                              HIST_checkInput_e check)
+{
+    const BYTE* ip = (const BYTE*)source;
+    const size_t maxCount = *maxSymbolValuePtr + 1;
+
+    assert(*maxSymbolValuePtr <= 255);
+    if (!sourceSize) {
+        ZSTD_memset(count, 0, maxCount * sizeof(*count));
+        *maxSymbolValuePtr = 0;
+        return 0;
+    }
+
+    {   const svbool_t vl128 = svptrue_pat_b8(SV_VL16);
+        const svuint8_t c0 = svreinterpret_u8(svindex_u32(0x0C040800, 0x01010101));
+        const svuint8_t c1 = svadd_n_u8_x(vl128, c0, 16);
+        const svuint8_t c2 = svadd_n_u8_x(vl128, c0, 32);
+        const svuint8_t c3 = svadd_n_u8_x(vl128, c1, 32);
+
+        svuint8_t symbolMax = svdup_n_u8(0);
+        svuint16_t hh0 = svdup_n_u16(0);
+        svuint16_t hh1 = svdup_n_u16(0);
+        svuint16_t hh2 = svdup_n_u16(0);
+        svuint16_t hh3 = svdup_n_u16(0);
+        svuint16_t hh4 = svdup_n_u16(0);
+        svuint16_t hh5 = svdup_n_u16(0);
+        svuint16_t hh6 = svdup_n_u16(0);
+        svuint16_t hh7 = svdup_n_u16(0);
+        svuint16_t max;
+        size_t maxSymbolValue;
+
+        size_t i = 0;
+        while (i < sourceSize) {
+            /* We can only accumulate 15 (15 * 16 <= 255) iterations of
+             * histogram in 8-bit accumulators! */
+            const size_t size240 = min_size(i + 240, sourceSize);
+
+            svbool_t pred = svwhilelt_b8_u64(i, sourceSize);
+            svuint8_t c = svld1rq_u8(pred, ip + i);
+            svuint8_t h0 = svhistseg_u8(c0, c);
+            svuint8_t h1 = svhistseg_u8(c1, c);
+            svuint8_t h2 = svhistseg_u8(c2, c);
+            svuint8_t h3 = svhistseg_u8(c3, c);
+            symbolMax = svmax_u8_x(vl128, symbolMax, c);
+
+            for (i += 16; i < size240; i += 16) {
+                pred = svwhilelt_b8_u64(i, sourceSize);
+                c = svld1rq_u8(pred, ip + i);
+                h0 = svadd_u8_x(vl128, h0, svhistseg_u8(c0, c));
+                h1 = svadd_u8_x(vl128, h1, svhistseg_u8(c1, c));
+                h2 = svadd_u8_x(vl128, h2, svhistseg_u8(c2, c));
+                h3 = svadd_u8_x(vl128, h3, svhistseg_u8(c3, c));
+                symbolMax = svmax_u8_x(vl128, symbolMax, c);
+            }
+
+            hh0 = svaddwb_u16(hh0, h0);
+            hh1 = svaddwt_u16(hh1, h0);
+            hh2 = svaddwb_u16(hh2, h1);
+            hh3 = svaddwt_u16(hh3, h1);
+            hh4 = svaddwb_u16(hh4, h2);
+            hh5 = svaddwt_u16(hh5, h2);
+            hh6 = svaddwb_u16(hh6, h3);
+            hh7 = svaddwt_u16(hh7, h3);
+        }
+        maxSymbolValue = svmaxv_u8(vl128, symbolMax);
+
+        if (check && maxSymbolValue > *maxSymbolValuePtr) return ERROR(maxSymbolValue_tooSmall);
+        *maxSymbolValuePtr = maxSymbolValue;
+
+        /* If the buffer size is not divisible by 16, the last elements of the final
+         * vector register read will be zeros, and these elements must be subtracted
+         * from the histogram.
+         */
+        hh0 = svsub_n_u16_m(svptrue_pat_b32(SV_VL1), hh0, -sourceSize & 15);
+
+        svst1_u32(svwhilelt_b32_u64( 0, maxCount), count +  0, svshllb_n_u32(hh0, 0));
+        svst1_u32(svwhilelt_b32_u64( 4, maxCount), count +  4, svshllt_n_u32(hh0, 0));
+        svst1_u32(svwhilelt_b32_u64( 8, maxCount), count +  8, svshllb_n_u32(hh1, 0));
+        svst1_u32(svwhilelt_b32_u64(12, maxCount), count + 12, svshllt_n_u32(hh1, 0));
+        svst1_u32(svwhilelt_b32_u64(16, maxCount), count + 16, svshllb_n_u32(hh2, 0));
+        svst1_u32(svwhilelt_b32_u64(20, maxCount), count + 20, svshllt_n_u32(hh2, 0));
+        svst1_u32(svwhilelt_b32_u64(24, maxCount), count + 24, svshllb_n_u32(hh3, 0));
+        svst1_u32(svwhilelt_b32_u64(28, maxCount), count + 28, svshllt_n_u32(hh3, 0));
+        svst1_u32(svwhilelt_b32_u64(32, maxCount), count + 32, svshllb_n_u32(hh4, 0));
+        svst1_u32(svwhilelt_b32_u64(36, maxCount), count + 36, svshllt_n_u32(hh4, 0));
+        svst1_u32(svwhilelt_b32_u64(40, maxCount), count + 40, svshllb_n_u32(hh5, 0));
+        svst1_u32(svwhilelt_b32_u64(44, maxCount), count + 44, svshllt_n_u32(hh5, 0));
+        svst1_u32(svwhilelt_b32_u64(48, maxCount), count + 48, svshllb_n_u32(hh6, 0));
+        svst1_u32(svwhilelt_b32_u64(52, maxCount), count + 52, svshllt_n_u32(hh6, 0));
+        svst1_u32(svwhilelt_b32_u64(56, maxCount), count + 56, svshllb_n_u32(hh7, 0));
+        svst1_u32(svwhilelt_b32_u64(60, maxCount), count + 60, svshllt_n_u32(hh7, 0));
+
+        hh0 = svmax_u16_x(vl128, hh0, hh1);
+        hh2 = svmax_u16_x(vl128, hh2, hh3);
+        hh4 = svmax_u16_x(vl128, hh4, hh5);
+        hh6 = svmax_u16_x(vl128, hh6, hh7);
+        hh0 = svmax_u16_x(vl128, hh0, hh2);
+        hh4 = svmax_u16_x(vl128, hh4, hh6);
+        max = svmax_u16_x(vl128, hh0, hh4);
+
+        maxSymbolValue = min_size(maxSymbolValue, maxCount);
+        if (maxSymbolValue >= 64) {
+            const svuint8_t c4 = svadd_n_u8_x(vl128, c0,  64);
+            const svuint8_t c5 = svadd_n_u8_x(vl128, c1,  64);
+            const svuint8_t c6 = svadd_n_u8_x(vl128, c2,  64);
+            const svuint8_t c7 = svadd_n_u8_x(vl128, c3,  64);
+            const svuint8_t c8 = svadd_n_u8_x(vl128, c0, 128);
+            const svuint8_t c9 = svadd_n_u8_x(vl128, c1, 128);
+
+            max = HIST_count_6_sve2(ip, sourceSize, count + 64, c4, c5, c6, c7,
+                                    c8, c9, max, maxCount - 64);
+
+            if (maxSymbolValue >= 160) {
+                const svuint8_t ca = svadd_n_u8_x(vl128, c2, 128);
+                const svuint8_t cb = svadd_n_u8_x(vl128, c3, 128);
+                const svuint8_t cc = svadd_n_u8_x(vl128, c4, 128);
+                const svuint8_t cd = svadd_n_u8_x(vl128, c5, 128);
+                const svuint8_t ce = svadd_n_u8_x(vl128, c6, 128);
+                const svuint8_t cf = svadd_n_u8_x(vl128, c7, 128);
+
+                max = HIST_count_6_sve2(ip, sourceSize, count + 160, ca, cb, cc,
+                                        cd, ce, cf, max, maxCount - 160);
+            } else if (maxCount > 160) {
+                ZSTD_memset(count + 160, 0, (maxCount - 160) * sizeof(*count));
+            }
+        } else if (maxCount > 64) {
+            ZSTD_memset(count + 64, 0, (maxCount - 64) * sizeof(*count));
+        }
+
+        return svmaxv_u16(vl128, max);
+    }
+}
+#endif
+
 /* HIST_count_parallel_wksp() :
  * store histogram into 4 intermediate tables, recombined at the end.
  * this design makes better use of OoO cpus,
@@ -73,8 +317,8 @@ typedef enum { trustInput, checkMaxSymbolValue } HIST_checkInput_e;
  * `workSpace` must be a U32 table of size >= HIST_WKSP_SIZE_U32.
  * @return : largest histogram frequency,
  *           or an error code (notably when histogram's alphabet is larger than *maxSymbolValuePtr) */
-static size_t HIST_count_parallel_wksp(
-                                unsigned* count, unsigned* maxSymbolValuePtr,
+static UNUSED_ATTR
+size_t HIST_count_parallel_wksp(unsigned* count, unsigned* maxSymbolValuePtr,
                                 const void* source, size_t sourceSize,
                                 HIST_checkInput_e check,
                                 U32* const workSpace)
@@ -151,11 +395,17 @@ size_t HIST_countFast_wksp(unsigned* count, unsigned* maxSymbolValuePtr,
                           const void* source, size_t sourceSize,
                           void* workSpace, size_t workSpaceSize)
 {
-    if (sourceSize < 1500) /* heuristic threshold */
+    if (sourceSize < HIST_FAST_THRESHOLD) /* heuristic threshold */
         return HIST_count_simple(count, maxSymbolValuePtr, source, sourceSize);
+#if defined(ZSTD_ARCH_ARM_SVE2)
+    (void)workSpace;
+    (void)workSpaceSize;
+    return HIST_count_sve2(count, maxSymbolValuePtr, source, sourceSize, trustInput);
+#else
     if ((size_t)workSpace & 3) return ERROR(GENERIC);  /* must be aligned on 4-bytes boundaries */
     if (workSpaceSize < HIST_WKSP_SIZE) return ERROR(workSpace_tooSmall);
     return HIST_count_parallel_wksp(count, maxSymbolValuePtr, source, sourceSize, trustInput, (U32*)workSpace);
+#endif
 }
 
 /* HIST_count_wksp() :
@@ -165,10 +415,15 @@ size_t HIST_count_wksp(unsigned* count, unsigned* maxSymbolValuePtr,
                        const void* source, size_t sourceSize,
                        void* workSpace, size_t workSpaceSize)
 {
+#if defined(ZSTD_ARCH_ARM_SVE2)
+    if (*maxSymbolValuePtr < 255)
+        return HIST_count_sve2(count, maxSymbolValuePtr, source, sourceSize, checkMaxSymbolValue);
+#else
     if ((size_t)workSpace & 3) return ERROR(GENERIC);  /* must be aligned on 4-bytes boundaries */
     if (workSpaceSize < HIST_WKSP_SIZE) return ERROR(workSpace_tooSmall);
     if (*maxSymbolValuePtr < 255)
         return HIST_count_parallel_wksp(count, maxSymbolValuePtr, source, sourceSize, checkMaxSymbolValue, (U32*)workSpace);
+#endif
     *maxSymbolValuePtr = 255;
     return HIST_countFast_wksp(count, maxSymbolValuePtr, source, sourceSize, workSpace, workSpaceSize);
 }
diff --git a/lib/compress/hist.h b/lib/compress/hist.h
index bea2a9ebf..e526e9532 100644
--- a/lib/compress/hist.h
+++ b/lib/compress/hist.h
@@ -35,7 +35,11 @@ unsigned HIST_isError(size_t code);  /**< tells if a return value is an error co
 
 /* --- advanced histogram functions --- */
 
+#if defined(__ARM_FEATURE_SVE2)
+#define HIST_WKSP_SIZE_U32 0
+#else
 #define HIST_WKSP_SIZE_U32 1024
+#endif
 #define HIST_WKSP_SIZE    (HIST_WKSP_SIZE_U32 * sizeof(unsigned))
 /** HIST_count_wksp() :
  *  Same as HIST_count(), but using an externally provided scratch buffer.
diff --git a/lib/compress/zstd_compress.c b/lib/compress/zstd_compress.c
index c8f6b2865..9b7aaf9f4 100644
--- a/lib/compress/zstd_compress.c
+++ b/lib/compress/zstd_compress.c
@@ -7284,6 +7284,93 @@ static size_t convertSequences_noRepcodes(
     return longLen;
 }
 
+#elif defined ZSTD_ARCH_RISCV_RVV
+#include <riscv_vector.h>
+/*
+ * Convert `vl` sequences per iteration, using AVX2 intrinsics:
+ *   - offset -> offBase = offset + 2
+ *   - litLength -> (U16) litLength
+ *   - matchLength -> (U16)(matchLength - 3)
+ *   - rep is ignored
+ * Store only 8 bytes per SeqDef (offBase[4], litLength[2], mlBase[2]).
+ *
+ * @returns 0 on succes, with no long length detected
+ * @returns > 0 if there is one long length (> 65535),
+ * indicating the position, and type.
+ */
+static size_t convertSequences_noRepcodes(SeqDef* dstSeqs, const ZSTD_Sequence* inSeqs, size_t nbSequences) {
+    size_t longLen = 0;
+
+    /* RVV depends on the specific definition of target structures */
+    ZSTD_STATIC_ASSERT(sizeof(ZSTD_Sequence) == 16);
+    ZSTD_STATIC_ASSERT(offsetof(ZSTD_Sequence, offset) == 0);
+    ZSTD_STATIC_ASSERT(offsetof(ZSTD_Sequence, litLength) == 4);
+    ZSTD_STATIC_ASSERT(offsetof(ZSTD_Sequence, matchLength) == 8);
+    ZSTD_STATIC_ASSERT(sizeof(SeqDef) == 8);
+    ZSTD_STATIC_ASSERT(offsetof(SeqDef, offBase) == 0);
+    ZSTD_STATIC_ASSERT(offsetof(SeqDef, litLength) == 4);
+    ZSTD_STATIC_ASSERT(offsetof(SeqDef, mlBase) == 6);
+    size_t vl = 0;
+    for (size_t i = 0; i < nbSequences; i += vl) {
+
+        vl = __riscv_vsetvl_e32m2(nbSequences-i);
+        // Loading structure member variables
+        vuint32m2x4_t v_tuple = __riscv_vlseg4e32_v_u32m2x4(
+            (const int32_t*)&inSeqs[i], 
+            vl
+        );
+        vuint32m2_t v_offset = __riscv_vget_v_u32m2x4_u32m2(v_tuple, 0);
+        vuint32m2_t v_lit = __riscv_vget_v_u32m2x4_u32m2(v_tuple, 1);
+        vuint32m2_t v_match = __riscv_vget_v_u32m2x4_u32m2(v_tuple, 2);
+        // offset + ZSTD_REP_NUM
+        vuint32m2_t v_offBase = __riscv_vadd_vx_u32m2(v_offset, ZSTD_REP_NUM, vl); 
+        // Check for integer overflow
+        // Cast to a 16-bit variable
+        vbool16_t lit_overflow = __riscv_vmsgtu_vx_u32m2_b16(v_lit, 65535, vl);
+        vuint16m1_t v_lit_clamped = __riscv_vncvt_x_x_w_u16m1(v_lit, vl);
+
+        vbool16_t ml_overflow = __riscv_vmsgtu_vx_u32m2_b16(v_match, 65535+MINMATCH, vl);
+        vuint16m1_t v_ml_clamped = __riscv_vncvt_x_x_w_u16m1(__riscv_vsub_vx_u32m2(v_match, MINMATCH, vl), vl);
+
+        // Pack two 16-bit fields into a 32-bit value (little-endian)
+        // The lower 16 bits contain litLength, and the upper 16 bits contain mlBase
+        vuint32m2_t v_lit_ml_combined = __riscv_vsll_vx_u32m2(
+            __riscv_vwcvtu_x_x_v_u32m2(v_ml_clamped, vl), // Convert matchLength to 32-bit
+            16, 
+            vl
+        );
+        v_lit_ml_combined = __riscv_vor_vv_u32m2(
+            v_lit_ml_combined,
+            __riscv_vwcvtu_x_x_v_u32m2(v_lit_clamped, vl),
+            vl
+        );
+        // Create a vector of SeqDef structures
+        // Store the offBase, litLength, and mlBase in a vector of SeqDef
+        vuint32m2x2_t store_data = __riscv_vcreate_v_u32m2x2(
+            v_offBase,          
+            v_lit_ml_combined   
+        );
+        __riscv_vsseg2e32_v_u32m2x2(
+            (uint32_t*)&dstSeqs[i], 
+            store_data,             
+            vl                      
+        );
+        // Find the first index where an overflow occurs
+        int first_ml = __riscv_vfirst_m_b16(ml_overflow, vl);
+        int first_lit = __riscv_vfirst_m_b16(lit_overflow, vl);
+
+        if (UNLIKELY(first_ml != -1)) {
+            assert(longLen == 0);
+            longLen = i + first_ml + 1;
+        }
+        if (UNLIKELY(first_lit != -1)) {
+            assert(longLen == 0);
+            longLen = i + first_lit + 1 + nbSequences;
+        }
+    }
+    return longLen;
+}
+
 /* the vector implementation could also be ported to SSSE3,
  * but since this implementation is targeting modern systems (>= Sapphire Rapid),
  * it's not useful to develop and maintain code for older pre-AVX2 platforms */
@@ -7451,6 +7538,70 @@ BlockSummary ZSTD_get1BlockSummary(const ZSTD_Sequence* seqs, size_t nbSeqs)
     }
 }
 
+#elif defined ZSTD_ARCH_RISCV_RVV
+
+BlockSummary ZSTD_get1BlockSummary(const ZSTD_Sequence* seqs, size_t nbSeqs)
+{
+    size_t totalMatchSize = 0;
+    size_t litSize = 0;
+    size_t i = 0;
+    int found_terminator = 0; 
+    size_t vl_max = __riscv_vsetvlmax_e32m1();
+    vuint32m1_t v_lit_sum = __riscv_vmv_v_x_u32m1(0, vl_max);
+    vuint32m1_t v_match_sum = __riscv_vmv_v_x_u32m1(0, vl_max);
+
+    for (; i  < nbSeqs; ) {
+        size_t vl = __riscv_vsetvl_e32m2(nbSeqs - i); 
+
+        ptrdiff_t stride = sizeof(ZSTD_Sequence); // 16
+        vuint32m2x4_t v_tuple = __riscv_vlseg4e32_v_u32m2x4(
+            (const int32_t*)&seqs[i], 
+            vl
+        );
+        vuint32m2_t v_offset = __riscv_vget_v_u32m2x4_u32m2(v_tuple, 0);
+        vuint32m2_t v_lit = __riscv_vget_v_u32m2x4_u32m2(v_tuple, 1);
+        vuint32m2_t v_match = __riscv_vget_v_u32m2x4_u32m2(v_tuple, 2);
+
+        // Check if any element has a matchLength of 0
+        vbool16_t mask = __riscv_vmseq_vx_u32m2_b16(v_match, 0, vl);
+        int first_zero = __riscv_vfirst_m_b16(mask, vl);
+
+        if (first_zero >= 0) {
+            // Find the first zero byte and set the effective length to that index + 1 to 
+            // recompute the cumulative vector length of literals and matches
+            vl = first_zero + 1;
+            
+            // recompute the cumulative vector length of literals and matches
+            v_lit_sum = __riscv_vredsum_vs_u32m2_u32m1(__riscv_vslidedown_vx_u32m2(v_lit, 0, vl), v_lit_sum, vl);
+            v_match_sum = __riscv_vredsum_vs_u32m2_u32m1(__riscv_vslidedown_vx_u32m2(v_match, 0, vl), v_match_sum, vl);
+
+            i += vl;
+            found_terminator = 1; 
+            assert(seqs[i - 1].offset == 0);
+            break;
+        } else {
+
+            v_lit_sum = __riscv_vredsum_vs_u32m2_u32m1(v_lit, v_lit_sum, vl);
+            v_match_sum = __riscv_vredsum_vs_u32m2_u32m1(v_match, v_match_sum, vl);
+            i += vl;
+        }
+    }
+    litSize = __riscv_vmv_x_s_u32m1_u32(v_lit_sum);
+    totalMatchSize = __riscv_vmv_x_s_u32m1_u32(v_match_sum);
+
+    if (!found_terminator && i==nbSeqs) {
+        BlockSummary bs;
+        bs.nbSequences = ERROR(externalSequences_invalid);
+        return bs;
+    }
+    {   BlockSummary bs;
+        bs.nbSequences = i;
+        bs.blockSize = litSize + totalMatchSize;
+        bs.litSize = litSize;
+        return bs;
+    }
+}
+
 #else
 
 BlockSummary ZSTD_get1BlockSummary(const ZSTD_Sequence* seqs, size_t nbSeqs)
diff --git a/lib/decompress/huf_decompress_amd64.S b/lib/decompress/huf_decompress_amd64.S
index 656aada95..dc1f3d921 100644
--- a/lib/decompress/huf_decompress_amd64.S
+++ b/lib/decompress/huf_decompress_amd64.S
@@ -38,6 +38,16 @@
 
 #endif
 
+// There appears to be an unreconcilable syntax difference between Linux and Darwin assemblers.
+// Name of a private label (i.e. not exported to symbol table) on Darwin has to start with "L",
+// on Linux has to start with ".". There's no way to have a name start with both "." and "L", so
+// we have to use a macro.
+#if defined(__APPLE__)
+#define LOCAL_LABEL(label) L_ ## label
+#else
+#define LOCAL_LABEL(label) .L_ ## label
+#endif
+
 #if ZSTD_ENABLE_ASM_X86_64_BMI2
 
 /* Calling convention:
@@ -117,22 +127,55 @@ ZSTD_HIDE_ASM_FUNCTION(_HUF_decompress4X1_usingDTable_internal_fast_asm_loop)
 _HUF_decompress4X1_usingDTable_internal_fast_asm_loop:
 HUF_decompress4X1_usingDTable_internal_fast_asm_loop:
     ZSTD_CET_ENDBRANCH
+    .cfi_startproc
+    .cfi_def_cfa_offset 8
+    .cfi_offset %rip, -8
     /* Save all registers - even if they are callee saved for simplicity. */
     push %rax
+    .cfi_def_cfa_offset 16
+    .cfi_offset rax, -16
     push %rbx
+    .cfi_def_cfa_offset 24
+    .cfi_offset rbx, -24
     push %rcx
+    .cfi_def_cfa_offset 32
+    .cfi_offset rcx, -32
     push %rdx
+    .cfi_def_cfa_offset 40
+    .cfi_offset rdx, -40
     push %rbp
+    .cfi_def_cfa_offset 48
+    .cfi_offset rbp, -48
     push %rsi
+    .cfi_def_cfa_offset 56
+    .cfi_offset rsi, -56
     push %rdi
+    .cfi_def_cfa_offset 64
+    .cfi_offset rdi, -64
     push %r8
+    .cfi_def_cfa_offset 72
+    .cfi_offset r8, -72
     push %r9
+    .cfi_def_cfa_offset 80
+    .cfi_offset r9, -80
     push %r10
+    .cfi_def_cfa_offset 88
+    .cfi_offset r10, -88
     push %r11
+    .cfi_def_cfa_offset 96
+    .cfi_offset r11, -96
     push %r12
+    .cfi_def_cfa_offset 104
+    .cfi_offset r12, -104
     push %r13
+    .cfi_def_cfa_offset 112
+    .cfi_offset r13, -112
     push %r14
+    .cfi_def_cfa_offset 120
+    .cfi_offset r14, -120
     push %r15
+    .cfi_def_cfa_offset 128
+    .cfi_offset r15, -128
 
     /* Read HUF_DecompressAsmArgs* args from %rax */
 #if defined(_WIN32)
@@ -154,13 +197,18 @@ HUF_decompress4X1_usingDTable_internal_fast_asm_loop:
     movq 88(%rax), %bits3
     movq 96(%rax), %dtable
     push %rax      /* argument */
+    .cfi_def_cfa_offset 136
     push 104(%rax) /* ilowest */
+    .cfi_def_cfa_offset 144
     push 112(%rax) /* oend */
+    .cfi_def_cfa_offset 152
     push %olimit   /* olimit space */
+    .cfi_def_cfa_offset 160
 
     subq $24, %rsp
+    .cfi_def_cfa_offset 184
 
-.L_4X1_compute_olimit:
+LOCAL_LABEL(4X1_compute_olimit):
     /* Computes how many iterations we can do safely
      * %r15, %rax may be clobbered
      * rbx, rdx must be saved
@@ -207,19 +255,19 @@ HUF_decompress4X1_usingDTable_internal_fast_asm_loop:
     /* If (op3 + 20 > olimit) */
     movq %op3, %rax    /* rax = op3 */
     cmpq %rax, %olimit /* op3 == olimit */
-    je .L_4X1_exit
+    je LOCAL_LABEL(4X1_exit)
 
     /* If (ip1 < ip0) go to exit */
     cmpq %ip0, %ip1
-    jb .L_4X1_exit
+    jb LOCAL_LABEL(4X1_exit)
 
     /* If (ip2 < ip1) go to exit */
     cmpq %ip1, %ip2
-    jb .L_4X1_exit
+    jb LOCAL_LABEL(4X1_exit)
 
     /* If (ip3 < ip2) go to exit */
     cmpq %ip2, %ip3
-    jb .L_4X1_exit
+    jb LOCAL_LABEL(4X1_exit)
 
 /* Reads top 11 bits from bits[n]
  * Loads dt[bits[n]] into var[n]
@@ -280,7 +328,7 @@ HUF_decompress4X1_usingDTable_internal_fast_asm_loop:
 
     .p2align 6
 
-.L_4X1_loop_body:
+LOCAL_LABEL(4X1_loop_body):
     /* Decode 5 symbols in each of the 4 streams (20 total)
      * Must have called GET_NEXT_DELT for each stream
      */
@@ -318,7 +366,7 @@ HUF_decompress4X1_usingDTable_internal_fast_asm_loop:
 
     /* If op3 < olimit: continue the loop */
     cmp %op3, 24(%rsp)
-    ja .L_4X1_loop_body
+    ja LOCAL_LABEL(4X1_loop_body)
 
     /* Reload ip[1,2,3] from stack */
     movq 0(%rsp), %ip1
@@ -326,20 +374,25 @@ HUF_decompress4X1_usingDTable_internal_fast_asm_loop:
     movq 16(%rsp), %ip3
 
     /* Re-compute olimit */
-    jmp .L_4X1_compute_olimit
+    jmp LOCAL_LABEL(4X1_compute_olimit)
 
 #undef GET_NEXT_DELT
 #undef DECODE_FROM_DELT
 #undef DECODE
 #undef RELOAD_BITS
-.L_4X1_exit:
+LOCAL_LABEL(4X1_exit):
     addq $24, %rsp
+    .cfi_def_cfa_offset 160
 
     /* Restore stack (oend & olimit) */
     pop %rax /* olimit */
+    .cfi_def_cfa_offset 152
     pop %rax /* oend */
+    .cfi_def_cfa_offset 144
     pop %rax /* ilowest */
+    .cfi_def_cfa_offset 136
     pop %rax /* arg */
+    .cfi_def_cfa_offset 128
 
     /* Save ip / op / bits */
     movq %ip0,  0(%rax)
@@ -357,41 +410,105 @@ HUF_decompress4X1_usingDTable_internal_fast_asm_loop:
 
     /* Restore registers */
     pop %r15
+    .cfi_restore r15
+    .cfi_def_cfa_offset 120
     pop %r14
+    .cfi_restore r14
+    .cfi_def_cfa_offset 112
     pop %r13
+    .cfi_restore r13
+    .cfi_def_cfa_offset 104
     pop %r12
+    .cfi_restore r12
+    .cfi_def_cfa_offset 96
     pop %r11
+    .cfi_restore r11
+    .cfi_def_cfa_offset 88
     pop %r10
+    .cfi_restore r10
+    .cfi_def_cfa_offset 80
     pop %r9
+    .cfi_restore r9
+    .cfi_def_cfa_offset 72
     pop %r8
+    .cfi_restore r8
+    .cfi_def_cfa_offset 64
     pop %rdi
+    .cfi_restore rdi
+    .cfi_def_cfa_offset 56
     pop %rsi
+    .cfi_restore rsi
+    .cfi_def_cfa_offset 48
     pop %rbp
+    .cfi_restore rbp
+    .cfi_def_cfa_offset 40
     pop %rdx
+    .cfi_restore rdx
+    .cfi_def_cfa_offset 32
     pop %rcx
+    .cfi_restore rcx
+    .cfi_def_cfa_offset 24
     pop %rbx
+    .cfi_restore rbx
+    .cfi_def_cfa_offset 16
     pop %rax
+    .cfi_restore rax
+    .cfi_def_cfa_offset 8
     ret
+    .cfi_endproc
 
 _HUF_decompress4X2_usingDTable_internal_fast_asm_loop:
 HUF_decompress4X2_usingDTable_internal_fast_asm_loop:
     ZSTD_CET_ENDBRANCH
+    .cfi_startproc
+    .cfi_def_cfa_offset 8
+    .cfi_offset %rip, -8
     /* Save all registers - even if they are callee saved for simplicity. */
     push %rax
+    .cfi_def_cfa_offset 16
+    .cfi_offset rax, -16
     push %rbx
+    .cfi_def_cfa_offset 24
+    .cfi_offset rbx, -24
     push %rcx
+    .cfi_def_cfa_offset 32
+    .cfi_offset rcx, -32
     push %rdx
+    .cfi_def_cfa_offset 40
+    .cfi_offset rdx, -40
     push %rbp
+    .cfi_def_cfa_offset 48
+    .cfi_offset rbp, -48
     push %rsi
+    .cfi_def_cfa_offset 56
+    .cfi_offset rsi, -56
     push %rdi
+    .cfi_def_cfa_offset 64
+    .cfi_offset rdi, -64
     push %r8
+    .cfi_def_cfa_offset 72
+    .cfi_offset r8, -72
     push %r9
+    .cfi_def_cfa_offset 80
+    .cfi_offset r9, -80
     push %r10
+    .cfi_def_cfa_offset 88
+    .cfi_offset r10, -88
     push %r11
+    .cfi_def_cfa_offset 96
+    .cfi_offset r11, -96
     push %r12
+    .cfi_def_cfa_offset 104
+    .cfi_offset r12, -104
     push %r13
+    .cfi_def_cfa_offset 112
+    .cfi_offset r13, -112
     push %r14
+    .cfi_def_cfa_offset 120
+    .cfi_offset r14, -120
     push %r15
+    .cfi_def_cfa_offset 128
+    .cfi_offset r15, -128
 
     /* Read HUF_DecompressAsmArgs* args from %rax */
 #if defined(_WIN32)
@@ -413,25 +530,33 @@ HUF_decompress4X2_usingDTable_internal_fast_asm_loop:
     movq 88(%rax), %bits3
     movq 96(%rax), %dtable
     push %rax      /* argument */
+    .cfi_def_cfa_offset 136
     push %rax      /* olimit */
+    .cfi_def_cfa_offset 144
     push 104(%rax) /* ilowest */
+    .cfi_def_cfa_offset 152
 
     movq 112(%rax), %rax
     push %rax /* oend3 */
+    .cfi_def_cfa_offset 160
 
     movq %op3, %rax
     push %rax /* oend2 */
+    .cfi_def_cfa_offset 168
 
     movq %op2, %rax
     push %rax /* oend1 */
+    .cfi_def_cfa_offset 176
 
     movq %op1, %rax
     push %rax /* oend0 */
+    .cfi_def_cfa_offset 184
 
     /* Scratch space */
     subq $8, %rsp
+    .cfi_def_cfa_offset 192
 
-.L_4X2_compute_olimit:
+LOCAL_LABEL(4X2_compute_olimit):
     /* Computes how many iterations we can do safely
      * %r15, %rax may be clobbered
      * rdx must be saved
@@ -495,19 +620,19 @@ HUF_decompress4X2_usingDTable_internal_fast_asm_loop:
     /* If (op3 + 10 > olimit) */
     movq %op3, %rax    /* rax = op3 */
     cmpq %rax, %olimit /* op3 == olimit */
-    je .L_4X2_exit
+    je LOCAL_LABEL(4X2_exit)
 
     /* If (ip1 < ip0) go to exit */
     cmpq %ip0, %ip1
-    jb .L_4X2_exit
+    jb LOCAL_LABEL(4X2_exit)
 
     /* If (ip2 < ip1) go to exit */
     cmpq %ip1, %ip2
-    jb .L_4X2_exit
+    jb LOCAL_LABEL(4X2_exit)
 
     /* If (ip3 < ip2) go to exit */
     cmpq %ip2, %ip3
-    jb .L_4X2_exit
+    jb LOCAL_LABEL(4X2_exit)
 
 #define DECODE(n, idx)              \
     movq %bits##n, %rax;            \
@@ -534,7 +659,7 @@ HUF_decompress4X2_usingDTable_internal_fast_asm_loop:
 
     .p2align 6
 
-.L_4X2_loop_body:
+LOCAL_LABEL(4X2_loop_body):
     /* We clobber r8, so store it on the stack */
     movq %r8, 0(%rsp)
 
@@ -551,21 +676,29 @@ HUF_decompress4X2_usingDTable_internal_fast_asm_loop:
     FOR_EACH_STREAM(RELOAD_BITS)
 
     cmp %op3, 48(%rsp)
-    ja .L_4X2_loop_body
-    jmp .L_4X2_compute_olimit
+    ja LOCAL_LABEL(4X2_loop_body)
+    jmp LOCAL_LABEL(4X2_compute_olimit)
 
 #undef DECODE
 #undef RELOAD_BITS
-.L_4X2_exit:
+LOCAL_LABEL(4X2_exit):
     addq $8, %rsp
+    .cfi_def_cfa_offset 184
     /* Restore stack (oend & olimit) */
     pop %rax /* oend0 */
+    .cfi_def_cfa_offset 176
     pop %rax /* oend1 */
+    .cfi_def_cfa_offset 168
     pop %rax /* oend2 */
+    .cfi_def_cfa_offset 160
     pop %rax /* oend3 */
+    .cfi_def_cfa_offset 152
     pop %rax /* ilowest */
+    .cfi_def_cfa_offset 144
     pop %rax /* olimit */
+    .cfi_def_cfa_offset 136
     pop %rax /* arg */
+    .cfi_def_cfa_offset 128
 
     /* Save ip / op / bits */
     movq %ip0,  0(%rax)
@@ -583,20 +716,51 @@ HUF_decompress4X2_usingDTable_internal_fast_asm_loop:
 
     /* Restore registers */
     pop %r15
+    .cfi_restore r15
+    .cfi_def_cfa_offset 120
     pop %r14
+    .cfi_restore r14
+    .cfi_def_cfa_offset 112
     pop %r13
+    .cfi_restore r13
+    .cfi_def_cfa_offset 104
     pop %r12
+    .cfi_restore r12
+    .cfi_def_cfa_offset 96
     pop %r11
+    .cfi_restore r11
+    .cfi_def_cfa_offset 88
     pop %r10
+    .cfi_restore r10
+    .cfi_def_cfa_offset 80
     pop %r9
+    .cfi_restore r9
+    .cfi_def_cfa_offset 72
     pop %r8
+    .cfi_restore r8
+    .cfi_def_cfa_offset 64
     pop %rdi
+    .cfi_restore rdi
+    .cfi_def_cfa_offset 56
     pop %rsi
+    .cfi_restore rsi
+    .cfi_def_cfa_offset 48
     pop %rbp
+    .cfi_restore rbp
+    .cfi_def_cfa_offset 40
     pop %rdx
+    .cfi_restore rdx
+    .cfi_def_cfa_offset 32
     pop %rcx
+    .cfi_restore rcx
+    .cfi_def_cfa_offset 24
     pop %rbx
+    .cfi_restore rbx
+    .cfi_def_cfa_offset 16
     pop %rax
+    .cfi_restore rax
+    .cfi_def_cfa_offset 8
     ret
+    .cfi_endproc
 
 #endif
diff --git a/programs/README.md b/programs/README.md
index 43ef07a45..2c2e94dc6 100644
--- a/programs/README.md
+++ b/programs/README.md
@@ -153,7 +153,7 @@ Usage: zstd [OPTIONS...] [INPUT... | -] [-o OUTPUT]
 Options:
   -o OUTPUT                     Write output to a single file, OUTPUT.
   -k, --keep                    Preserve INPUT file(s). [Default]
-  --rm                          Remove INPUT file(s) after successful (de)compression.
+  --rm                          Remove INPUT file(s) after successful (de)compression to file.
 
   -#                            Desired compression level, where `#` is a number between 1 and 19;
                                 lower numbers provide faster compression, higher numbers yield
diff --git a/programs/util.c b/programs/util.c
index d11f13cfd..652530b12 100644
--- a/programs/util.c
+++ b/programs/util.c
@@ -928,6 +928,7 @@ static int UTIL_prepareFileList(const char* dirName,
     hFile=FindFirstFileA(path, &cFile);
     if (hFile == INVALID_HANDLE_VALUE) {
         UTIL_DISPLAYLEVEL(1, "Cannot open directory '%s'\n", dirName);
+        free(path);
         return 0;
     }
     free(path);
diff --git a/programs/zstdcli.c b/programs/zstdcli.c
index fa7ea37b3..124ea47aa 100644
--- a/programs/zstdcli.c
+++ b/programs/zstdcli.c
@@ -147,7 +147,7 @@ static void usage(FILE* f, const char* programName)
     DISPLAY_F(f, "Options:\n");
     DISPLAY_F(f, "  -o OUTPUT                     Write output to a single file, OUTPUT.\n");
     DISPLAY_F(f, "  -k, --keep                    Preserve INPUT file(s). [Default] \n");
-    DISPLAY_F(f, "  --rm                          Remove INPUT file(s) after successful (de)compression.\n");
+    DISPLAY_F(f, "  --rm                          Remove INPUT file(s) after successful (de)compression to file.\n");
 #ifdef ZSTD_GZCOMPRESS
     if (exeNameMatch(programName, ZSTD_GZ)) {     /* behave like gzip */
         DISPLAY_F(f, "  -n, --no-name                 Do not store original filename when compressing.\n\n");
diff --git a/tests/cli-tests/basic/help.sh.stdout.glob b/tests/cli-tests/basic/help.sh.stdout.glob
index 21bc28c3f..66c6906cf 100644
--- a/tests/cli-tests/basic/help.sh.stdout.glob
+++ b/tests/cli-tests/basic/help.sh.stdout.glob
@@ -6,7 +6,7 @@ Usage: zstd *OPTIONS...* *INPUT... | -* *-o OUTPUT*
 Options:
   -o OUTPUT                     Write output to a single file, OUTPUT.
   -k, --keep                    Preserve INPUT file(s). *Default*
-  --rm                          Remove INPUT file(s) after successful (de)compression.
+  --rm                          Remove INPUT file(s) after successful (de)compression to file.
 
   -#                            Desired compression level, where `#` is a number between 1 and 19;
                                 lower numbers provide faster compression, higher numbers yield
diff --git a/tests/cli-tests/run.py b/tests/cli-tests/run.py
index 011c8478e..0addd6b1a 100755
--- a/tests/cli-tests/run.py
+++ b/tests/cli-tests/run.py
@@ -640,7 +640,7 @@ if __name__ == "__main__":
         help="Preserve the scratch directory TEST_DIR/scratch/ for debugging purposes."
     )
     parser.add_argument("--verbose", action="store_true", help="Verbose test output.")
-    parser.add_argument("--timeout", default=200, type=int, help="Test case timeout in seconds. Set to 0 to disable timeouts.")
+    parser.add_argument("--timeout", default=800, type=int, help="Test case timeout in seconds. Set to 0 to disable timeouts.")
     parser.add_argument(
         "--exec-prefix",
         default=None,
diff --git a/tests/fullbench.c b/tests/fullbench.c
index 12a27f48c..5841e5db5 100644
--- a/tests/fullbench.c
+++ b/tests/fullbench.c
@@ -890,6 +890,7 @@ static int benchMem(unsigned scenarioID,
             if (!BMK_isSuccessful_runOutcome(bOutcome)) {
                 DISPLAY("ERROR: Scenario %u: %s \n", scenarioID, ZSTD_getErrorName(BMK_extract_errorResult(bOutcome)));
                 errorcode = 1;
+                BMK_freeTimedFnState(tfs);
                 goto _cleanOut;
             }
 
diff --git a/tests/fuzz/fuzz_helpers.c b/tests/fuzz/fuzz_helpers.c
index f47ff2eb4..5c530f0e9 100644
--- a/tests/fuzz/fuzz_helpers.c
+++ b/tests/fuzz/fuzz_helpers.c
@@ -31,12 +31,11 @@ void* FUZZ_malloc_rand(size_t size, FUZZ_dataProducer_t *producer)
         return mem;
     } else {
         uintptr_t ptr = 0;
-        /* Add +- 1M 50% of the time */
+        /* Return junk pointer 50% of the time */
         if (FUZZ_dataProducer_uint32Range(producer, 0, 1))
-            FUZZ_dataProducer_int32Range(producer, -1000000, 1000000);
+            ptr += FUZZ_dataProducer_int32Range(producer, -1000000, 1000000);
         return (void*)ptr;
     }
-
 }
 
 int FUZZ_memcmp(void const* lhs, void const* rhs, size_t size)
diff --git a/tests/fuzz/fuzz_helpers.h b/tests/fuzz/fuzz_helpers.h
index f21ec4751..c5d75b402 100644
--- a/tests/fuzz/fuzz_helpers.h
+++ b/tests/fuzz/fuzz_helpers.h
@@ -66,6 +66,7 @@ void* FUZZ_malloc(size_t size);
 /**
  * malloc except returns random pointer for zero sized data and FUZZ_ASSERT
  * that malloc doesn't fail.
+ * WARNING: Only free the returned pointer if size > 0!
  */
 void* FUZZ_malloc_rand(size_t size,  FUZZ_dataProducer_t *producer);
 
diff --git a/tests/fuzzer.c b/tests/fuzzer.c
index b74460bb5..da380aced 100644
--- a/tests/fuzzer.c
+++ b/tests/fuzzer.c
@@ -44,6 +44,13 @@
 /* must be included after util.h, due to ERROR macro redefinition issue on Visual Studio */
 #include "zstd_internal.h" /* ZSTD_WORKSPACETOOLARGE_MAXDURATION, ZSTD_WORKSPACETOOLARGE_FACTOR, KB, MB */
 #include "threading.h"    /* ZSTD_pthread_create, ZSTD_pthread_join */
+#include "compress/hist.h" /* HIST_count_wksp */
+
+
+/*-************************************
+*  Macros
+**************************************/
+#define COUNTOF(array)    (sizeof(array) / sizeof(*(array)))
 
 
 /*-************************************
@@ -567,6 +574,123 @@ static void test_decompressBound(unsigned tnb)
     DISPLAYLEVEL(3, "OK \n");
 }
 
+static unsigned test_histCountWksp(unsigned seed, unsigned testNb)
+{
+    static const unsigned symLowLimits[] =  {   0,  27,   0,   0,  27,  42,   0,   0, 27, 42, 27, 42 };
+    static const unsigned symHighLimits[] = { 255, 255, 210, 110,  42,  42, 210, 110, 42, 42, 42, 42 };
+    static const unsigned symMaxLimits[] =  { 255, 255, 255, 255, 255, 255, 230, 130, 99, 99, 42, 42 };
+    static const size_t inputSizes[] = { 3367, 1761, 893, 117 };
+    unsigned workspace[HIST_WKSP_SIZE_U32];
+    size_t res, i, is, il;
+
+    DISPLAYLEVEL(3, "test%3u : HIST_count_wksp with empty source : ", testNb++);
+    {
+        /* With NULL source UBSan of older Clang could fail: applying zero offset to null pointer. */
+        static const unsigned char source[4] = { 0 };
+        unsigned count[1] = { 0 };
+        unsigned maxSym = 0;
+        res = HIST_count_wksp(count, &maxSym, source, 0, workspace, sizeof(workspace));
+        CHECK_EQ(res, 0);
+        CHECK_EQ(maxSym, 0);
+        CHECK_EQ(count[0], 0);
+    }
+    DISPLAYLEVEL(3, "OK \n");
+
+#if HIST_WKSP_SIZE_U32
+    DISPLAYLEVEL(3, "test%3u : HIST_count_wksp with small workspace : ", testNb++);
+    {
+        unsigned count[1] = { 0 };
+        unsigned maxSym = 0;
+        res = HIST_count_wksp(count, &maxSym, NULL, 0, workspace, sizeof(workspace) - 1);
+        CHECK_EQ(res, ERROR(workSpace_tooSmall));
+        CHECK_EQ(maxSym, 0);
+        CHECK_EQ(count[0], 0);
+    }
+    DISPLAYLEVEL(3, "OK \n");
+
+    DISPLAYLEVEL(3, "test%3u : HIST_count_wksp with wrong workspace alignment : ", testNb++);
+    {
+        unsigned count[1] = { 0 };
+        unsigned maxSym = 0;
+        res = HIST_count_wksp(count, &maxSym, NULL, 0, (unsigned*)(void*)((char*)workspace + 1), sizeof(workspace));
+        CHECK_EQ(res, ERROR(GENERIC));
+        CHECK_EQ(maxSym, 0);
+        CHECK_EQ(count[0], 0);
+    }
+    DISPLAYLEVEL(3, "OK \n");
+#endif
+
+    DISPLAYLEVEL(3, "test%3u : HIST_count_wksp with symbol out of range, small size : ", testNb++);
+    {
+        /* For less elements HIST_count_parallel_wksp would fail. */
+        static const unsigned char source[4] = { 1, 4, 0, 2 };
+        static const unsigned expected[6] = { 0 };
+        unsigned count[6] = { 0 };
+        unsigned maxSym = 2;
+        res = HIST_count_wksp(count, &maxSym, source, sizeof(source), workspace, sizeof(workspace));
+        CHECK_EQ(res, ERROR(maxSymbolValue_tooSmall));
+        CHECK_EQ(maxSym, 2);
+        for (i = 0; i < COUNTOF(expected); ++i) CHECK_EQ(count[i], expected[i]);
+    }
+    DISPLAYLEVEL(3, "OK \n");
+
+    DISPLAYLEVEL(3, "test%3u : HIST_count_wksp with symbol out of range, medium size : ", testNb++);
+    {
+        unsigned char source[3407];
+        unsigned count[6] = { 0 };
+        unsigned maxSym = 2;
+        for (i = 0; i < COUNTOF(source); ++i) {
+            source[i] = (48271 * (i + 1)) & 3;
+        }
+        res = HIST_count_wksp(count, &maxSym, source, sizeof(source), workspace, sizeof(workspace));
+        CHECK_EQ(res, ERROR(maxSymbolValue_tooSmall));
+        CHECK_EQ(maxSym, 2);
+        for (i = 0; i < COUNTOF(count); ++i) CHECK_EQ(count[i], 0);
+    }
+    DISPLAYLEVEL(3, "OK \n");
+
+    for (il = 0; il < COUNTOF(symMaxLimits); ++il) {
+        unsigned symMax = symMaxLimits[il];
+        unsigned symLow = symLowLimits[il];
+        unsigned symHigh = symHighLimits[il];
+        unsigned symRange = symHigh - symLow + 1;
+
+        for (is = 0; is < COUNTOF(inputSizes); ++is) {
+            unsigned char source[4000];
+            size_t inputSize = inputSizes[is];
+            assert(inputSize <= sizeof(source));
+            DISPLAYLEVEL(3, "test%3u : HIST_count_wksp test in [%u..%u], symMax: %u, inputSize: %u : ",
+                         testNb++, symLow, symHigh, symMax, (unsigned)inputSize);
+            {
+                unsigned count[260] = { 0 };
+                unsigned expected[COUNTOF(count)] = { 0 };
+                unsigned maxSym = symMax;
+                unsigned realMaxSym = symMax;
+                unsigned maxCount = 0;
+                for (i = 0; i < inputSize; ++i) {
+                    unsigned prng = (48271 * (i + seed)) % symRange + symLow;
+                    source[i] = (unsigned char)prng;
+                    ++expected[prng];
+                }
+                /* for basic buffer overwrite checks */
+                for (i = maxSym + 1; i < COUNTOF(count); ++i) expected[i] = count[i] = ~0u;
+                for (i = 0; i <= maxSym; ++i) maxCount = MAX(maxCount, expected[i]);
+                for (i = realMaxSym; i > 0; --i) {
+                    if (expected[i]) break;
+                    --realMaxSym;
+                }
+                res = HIST_count_wksp(count, &maxSym, source, inputSize, workspace, sizeof(workspace));
+                CHECK_EQ(res, maxCount);
+                CHECK_EQ(maxSym, realMaxSym);
+                for (i = 0; i < COUNTOF(expected); ++i) CHECK_EQ(count[i], expected[i]);
+            }
+            DISPLAYLEVEL(3, "OK \n");
+        }
+    }
+
+    return testNb;
+}
+
 static void test_setCParams(unsigned tnb)
 {
     ZSTD_CCtx* const cctx = ZSTD_createCCtx();
@@ -712,6 +836,8 @@ static int basicUnitTests(U32 const seed, double compressibility)
     }
     DISPLAYLEVEL(3, "OK \n");
 
+    testNb = test_histCountWksp(seed, testNb);
+
     DISPLAYLEVEL(3, "test%3u : compress %u bytes : ", testNb++, (unsigned)CNBuffSize);
     {   ZSTD_CCtx* const cctx = ZSTD_createCCtx();
         if (cctx==NULL) goto _output_error;
diff --git a/tests/largeDictionary.c b/tests/largeDictionary.c
index 998fd9fa8..30ff98d98 100644
--- a/tests/largeDictionary.c
+++ b/tests/largeDictionary.c
@@ -73,6 +73,7 @@ int main(int argc, const char** argv)
     int _exit_code = 1;
     (void)argc;
     (void)argv;
+    int _exit_code = 0;
 
     if (!buffer || !out || !roundtrip || !cctx || !dctx) {
         fprintf(stderr, "Allocation failure\n");
@@ -127,4 +128,5 @@ cleanup:
     ZSTD_freeDCtx(dctx);
     ZSTD_freeCCtx(cctx);
     return _exit_code;
+    return _exit_code;
 }
diff --git a/tests/legacy.c b/tests/legacy.c
index 3be386495..8fe291c5a 100644
--- a/tests/legacy.c
+++ b/tests/legacy.c
@@ -56,6 +56,7 @@ static int testSimpleAPI(void)
             } else {
                 DISPLAY("ERROR: %s\n", ZSTD_getErrorName(ret));
             }
+            free(output);
             return 1;
         }
         if (ret != size) {
@@ -64,6 +65,7 @@ static int testSimpleAPI(void)
     }
     if (memcmp(EXPECTED, output, size) != 0) {
         DISPLAY("ERROR: Wrong decoded output produced\n");
+        free(output);
         return 1;
     }
 
diff --git a/tests/longmatch.c b/tests/longmatch.c
index 547b26190..8e4e51dc1 100644
--- a/tests/longmatch.c
+++ b/tests/longmatch.c
@@ -39,24 +39,39 @@ int main(int argc, const char** argv)
   unsigned windowLog = 18;
   (void)argc;
   (void)argv;
+  int _exit_code = 0;
   /* Create stream */
   ctx = ZSTD_createCCtx();
   if (!ctx) { return 1; }
   /* Set parameters */
-  if (ZSTD_isError(ZSTD_CCtx_setParameter(ctx, ZSTD_c_windowLog, windowLog)))
-    return 2;
-  if (ZSTD_isError(ZSTD_CCtx_setParameter(ctx, ZSTD_c_chainLog, 13)))
-    return 2;
-  if (ZSTD_isError(ZSTD_CCtx_setParameter(ctx, ZSTD_c_hashLog, 14)))
-    return 2;
-  if (ZSTD_isError(ZSTD_CCtx_setParameter(ctx, ZSTD_c_searchLog, 1)))
-    return 2;
-  if (ZSTD_isError(ZSTD_CCtx_setParameter(ctx, ZSTD_c_minMatch, 7)))
-    return 2;
-  if (ZSTD_isError(ZSTD_CCtx_setParameter(ctx, ZSTD_c_targetLength, 16)))
-    return 2;
-  if (ZSTD_isError(ZSTD_CCtx_setParameter(ctx, ZSTD_c_strategy, ZSTD_fast)))
-    return 2;
+  if (ZSTD_isError(ZSTD_CCtx_setParameter(ctx, ZSTD_c_windowLog, windowLog))) {
+    _exit_code = 2;
+    goto _clean_ctx;
+  }
+  if (ZSTD_isError(ZSTD_CCtx_setParameter(ctx, ZSTD_c_chainLog, 13))) {
+    _exit_code = 2;
+    goto _clean_ctx;
+  }
+  if (ZSTD_isError(ZSTD_CCtx_setParameter(ctx, ZSTD_c_hashLog, 14))) {
+    _exit_code = 2;
+    goto _clean_ctx;
+  }
+  if (ZSTD_isError(ZSTD_CCtx_setParameter(ctx, ZSTD_c_searchLog, 1))) {
+    _exit_code = 2;
+    goto _clean_ctx;
+  }
+  if (ZSTD_isError(ZSTD_CCtx_setParameter(ctx, ZSTD_c_minMatch, 7))) {
+    _exit_code = 2;
+    goto _clean_ctx;
+  }
+  if (ZSTD_isError(ZSTD_CCtx_setParameter(ctx, ZSTD_c_targetLength, 16))) {
+    _exit_code = 2;
+    goto _clean_ctx;
+  }
+  if (ZSTD_isError(ZSTD_CCtx_setParameter(ctx, ZSTD_c_strategy, ZSTD_fast))) {
+    _exit_code = 2;
+    goto _clean_ctx;
+  }
   {
     U64 compressed = 0;
     const U64 toCompress = ((U64)1) << 33;
@@ -81,22 +96,28 @@ int main(int argc, const char** argv)
     }
     printf("Compressing, trying to generate a segfault \n");
     if (compress(ctx, out, srcBuffer, size)) {
-      return 1;
+       _exit_code = 1;
+       goto _clean_buffer;
     }
     compressed += size;
     while (compressed < toCompress) {
       const size_t block = rand() % (size - pos + 1);
       if (pos == size) { pos = 0; }
       if (compress(ctx, out, srcBuffer + pos, block)) {
-        return 1;
+        _exit_code = 1;
+        goto _clean_buffer;
       }
       pos += block;
       compressed += block;
     }
     printf("Compression completed successfully (no error triggered)\n");
+
+_clean_buffer:
     free(srcBuffer);
     free(dstBuffer);
   }
+
+_clean_ctx:
   ZSTD_freeCCtx(ctx);
-  return 0;
+  return _exit_code;
 }