Merge remote-tracking branch 'origin/main' into abadams/faster_inlining

# Conflicts:
#	test/correctness/CMakeLists.txt

Author: abadams

.pre-commit-config.yaml

@@ -112,8 +112,14 @@ repos:
     hooks:
       - id: taplo-format
 
+  - repo: https://github.com/halide/cmtk-pre-commit
+    rev: v0.1.0
+    hooks:
+      - id: cmtk
+
   - repo: local
     hooks:
+
       - id: check-cmake-file-lists
         name: check CMake file lists
         entry: python3 tools/check_cmake_file_lists.py

CMakeLists.txt

@@ -1,21 +1,27 @@
 cmake_minimum_required(VERSION 3.28)
-project(Halide
-        VERSION 22.0.0
-        DESCRIPTION "Halide compiler and libraries"
-        HOMEPAGE_URL "https://halide-lang.org")
+project(
+    Halide
+    VERSION 22.0.0
+    DESCRIPTION "Halide compiler and libraries"
+    HOMEPAGE_URL "https://halide-lang.org"
+)
 
 enable_testing()
 
 ##
 # Disable find_package(Halide) inside the build
 ##
 
-file(CONFIGURE OUTPUT "${CMAKE_FIND_PACKAGE_REDIRECTS_DIR}/HalideConfig.cmake"
-     CONTENT [[set(Halide_FOUND 1)
-               set(Halide_VERSION @Halide_VERSION@)]])
+file(CONFIGURE
+    OUTPUT "${CMAKE_FIND_PACKAGE_REDIRECTS_DIR}/HalideConfig.cmake"
+    CONTENT [[set(Halide_FOUND 1)
+              set(Halide_VERSION @Halide_VERSION@)]]
+)
 
-file(CONFIGURE OUTPUT "${CMAKE_FIND_PACKAGE_REDIRECTS_DIR}/HalideHelpersConfig.cmake"
-     CONTENT "set(HalideHelpers_FOUND 1)\n")
+file(CONFIGURE
+    OUTPUT "${CMAKE_FIND_PACKAGE_REDIRECTS_DIR}/HalideHelpersConfig.cmake"
+    CONTENT "set(HalideHelpers_FOUND 1)\n"
+)
 
 ##
 # Set up project-wide properties
@@ -37,9 +43,10 @@ option(BUILD_SHARED_LIBS "Build shared libraries" ON)
 get_property(is_multi_config GLOBAL PROPERTY GENERATOR_IS_MULTI_CONFIG)
 if (NOT is_multi_config AND CMAKE_BUILD_TYPE STREQUAL "")
     string(JOIN "" error_message
-           "Halide cannot be built when CMAKE_BUILD_TYPE is empty. "
-           "Please set CMAKE_BUILD_TYPE to one of the standard types: "
-           "Debug, Release, RelWithDebInfo, or MinSizeRel.")
+        "Halide cannot be built when CMAKE_BUILD_TYPE is empty. "
+        "Please set CMAKE_BUILD_TYPE to one of the standard types: "
+        "Debug, Release, RelWithDebInfo, or MinSizeRel."
+    )
     message(FATAL_ERROR "${error_message}")
 endif ()
 
@@ -58,11 +65,17 @@ set(CMAKE_WINDOWS_EXPORT_ALL_SYMBOLS ON)
 
 # Require standard C++17
 set(CMAKE_CXX_STANDARD 17 CACHE STRING "The C++ standard to use. Halide requires 17 or higher.")
-option(CMAKE_CXX_STANDARD_REQUIRED "When enabled, the value of CMAKE_CXX_STANDARD is a requirement." ON)
-option(CMAKE_CXX_EXTENSIONS "When enabled, compiler-specific language extensions are enabled (e.g. -std=gnu++17)" OFF)
+option(CMAKE_CXX_STANDARD_REQUIRED
+    "When enabled, the value of CMAKE_CXX_STANDARD is a requirement." ON
+)
+option(CMAKE_CXX_EXTENSIONS
+    "When enabled, compiler-specific language extensions are enabled (e.g. -std=gnu++17)" OFF
+)
 
 if (CMAKE_CXX_STANDARD LESS 17)
-    message(FATAL_ERROR "Halide requires C++17 or newer but CMAKE_CXX_STANDARD=${CMAKE_CXX_STANDARD}")
+    message(
+        FATAL_ERROR "Halide requires C++17 or newer but CMAKE_CXX_STANDARD=${CMAKE_CXX_STANDARD}"
+    )
 endif ()
 
 # Build Halide with ccache if the package is present and the user requested it
@@ -74,11 +87,14 @@ if (Halide_CCACHE_BUILD)
         CCACHE_CPP2=yes
         CCACHE_HASHDIR=yes
         CCACHE_SLOPPINESS=pch_defines,time_macros,include_file_mtime,include_file_ctime
-        CACHE STRING "Parameters to pass through to ccache")
+        CACHE STRING "Parameters to pass through to ccache"
+    )
     mark_as_advanced(Halide_CCACHE_PARAMS)
 
     set(CMAKE_C_COMPILER_LAUNCHER ${CMAKE_COMMAND} -E env ${Halide_CCACHE_PARAMS} ${CCACHE_PROGRAM})
-    set(CMAKE_CXX_COMPILER_LAUNCHER ${CMAKE_COMMAND} -E env ${Halide_CCACHE_PARAMS} ${CCACHE_PROGRAM})
+    set(CMAKE_CXX_COMPILER_LAUNCHER
+        ${CMAKE_COMMAND} -E env ${Halide_CCACHE_PARAMS} ${CCACHE_PROGRAM}
+    )
 
     # Per https://ccache.dev/manual/latest.html#_precompiled_headers,
     # we must set -fno-pch-timestamp when using Clang + CCache + PCH
@@ -93,7 +109,9 @@ endif ()
 # Detect whether or not ASAN is enabled. Don't cache the result to ensure this
 # check happens every time we reconfigure.
 unset(Halide_ASAN_ENABLED CACHE)
-check_cxx_symbol_exists(HALIDE_INTERNAL_USING_ASAN "${Halide_SOURCE_DIR}/src/Util.h" Halide_ASAN_ENABLED)
+check_cxx_symbol_exists(
+    HALIDE_INTERNAL_USING_ASAN "${Halide_SOURCE_DIR}/src/Util.h" Halide_ASAN_ENABLED
+)
 
 if (Halide_ASAN_ENABLED)
     set(Halide_ANY_SANITIZERS_ENABLED 1)
@@ -107,27 +125,22 @@ function(set_halide_compiler_warnings NAME)
         ${NAME}
         PRIVATE
         $<$<CXX_COMPILER_ID:GNU,Clang,AppleClang>:-Wall>
-
         # variable length arrays in C++ are a Clang extension, we don't want to use them
         $<$<CXX_COMPILER_ID:Clang,AppleClang>:-Wvla-extension>
-
         $<$<CXX_COMPILER_ID:GNU,Clang,AppleClang>:-Wcast-qual>
         $<$<CXX_COMPILER_ID:GNU,Clang,AppleClang>:-Wignored-qualifiers>
         $<$<CXX_COMPILER_ID:GNU,Clang,AppleClang>:-Wimplicit-fallthrough>
-
         # GCC warns when these warnings are given to plain-C sources
         $<$<COMPILE_LANG_AND_ID:CXX,GNU,Clang,AppleClang>:-Woverloaded-virtual>
         $<$<COMPILE_LANG_AND_ID:CXX,GNU>:-Wsuggest-override>
-
         $<$<CXX_COMPILER_ID:Clang,AppleClang>:-Winconsistent-missing-destructor-override>
         $<$<CXX_COMPILER_ID:Clang,AppleClang>:-Winconsistent-missing-override>
         $<$<CXX_COMPILER_ID:GNU,Clang,AppleClang>:-Wdeprecated-declarations>
-
         $<$<CXX_COMPILER_ID:MSVC>:/W3>
         $<$<CXX_COMPILER_ID:MSVC>:/wd4018>  # 4018: disable "signed/unsigned mismatch"
         $<$<CXX_COMPILER_ID:MSVC>:/wd4146>  # 4146: unary minus applied to unsigned type
         $<$<CXX_COMPILER_ID:MSVC>:/wd4244>  # 4244: conversion, possible loss of data
-        $<$<CXX_COMPILER_ID:MSVC>:/wd4267>  # 4267: conversion from 'size_t' to 'int', possible loss of data
+        $<$<CXX_COMPILER_ID:MSVC>:/wd4267>  # 4267: size_t -> int narrowing
     )
 endfunction()
 
@@ -137,13 +150,17 @@ endfunction()
 ##
 
 ## Threads
-option(THREADS_PREFER_PTHREAD_FLAG "When enabled, prefer to use the -pthread flag to explicit linking" ON)
+option(THREADS_PREFER_PTHREAD_FLAG
+    "When enabled, prefer to use the -pthread flag to explicit linking" ON
+)
 find_package(Threads REQUIRED)
 
 ## LLVM
-find_package(Halide_LLVM 21...99 REQUIRED  # Use 99 to fake a minimum-only constraint
-             COMPONENTS WebAssembly X86
-             OPTIONAL_COMPONENTS AArch64 ARM Hexagon NVPTX PowerPC RISCV)
+find_package(
+    Halide_LLVM 21...99 REQUIRED  # Use 99 to fake a minimum-only constraint
+    COMPONENTS WebAssembly X86
+    OPTIONAL_COMPONENTS AArch64 ARM Hexagon NVPTX PowerPC RISCV
+)
 
 _Halide_pkgdep(Halide_LLVM PACKAGE_VARS Halide_LLVM_SHARED_LIBS)
 
@@ -167,23 +184,27 @@ find_package(PNG)
 Halide_feature(Halide_BUILDING_IN_CI "Enable if building in CI" OFF ADVANCED)
 
 Halide_feature(Halide_ENABLE_EXCEPTIONS "Enable exceptions in Halide" ON)
-Halide_feature(Halide_ENABLE_RTTI "Enable RTTI in Halide" ON
-               DEPENDS LLVM_ENABLE_RTTI)
+Halide_feature(Halide_ENABLE_RTTI "Enable RTTI in Halide" ON DEPENDS LLVM_ENABLE_RTTI)
 
-Halide_feature(WITH_AUTOSCHEDULERS "Build the Halide autoschedulers" ON
-               DEPENDS BUILD_SHARED_LIBS)
+Halide_feature(WITH_AUTOSCHEDULERS "Build the Halide autoschedulers" ON DEPENDS BUILD_SHARED_LIBS)
 Halide_feature(WITH_DOCS "Halide's Doxygen documentation" OFF)
 Halide_feature(WITH_PACKAGING "Halide's CMake package install rules" TOP_LEVEL)
-Halide_feature(WITH_PYTHON_BINDINGS "Halide's native Python module (not the whole pip package)" ON
-               DEPENDS Halide_ENABLE_EXCEPTIONS AND Halide_ENABLE_RTTI)
+Halide_feature(
+    WITH_PYTHON_BINDINGS "Halide's native Python module (not the whole pip package)" ON
+    DEPENDS Halide_ENABLE_EXCEPTIONS AND Halide_ENABLE_RTTI
+)
 Halide_feature(WITH_SERIALIZATION "Include experimental Serialization/Deserialization code" ON)
-Halide_feature(WITH_SERIALIZATION_JIT_ROUNDTRIP_TESTING
-               "Intercepting JIT compilation with a serialization roundtrip, for test only"
-               OFF ADVANCED
-               DEPENDS WITH_SERIALIZATION)
+Halide_feature(
+    WITH_SERIALIZATION_JIT_ROUNDTRIP_TESTING
+    "Intercepting JIT compilation with a serialization roundtrip, for test only"
+    OFF ADVANCED
+    DEPENDS WITH_SERIALIZATION
+)
 Halide_feature(WITH_TESTS "Halide's unit test suite" TOP_LEVEL)
 Halide_feature(WITH_TUTORIALS "Halide's tutorial code" TOP_LEVEL)
-Halide_feature(WITH_UTILS "Optional utility programs for Halide, including HalideTraceViz" TOP_LEVEL)
+Halide_feature(
+    WITH_UTILS "Optional utility programs for Halide, including HalideTraceViz" TOP_LEVEL
+)
 
 
 ##

apps/CMakeLists.txt

@@ -28,7 +28,7 @@ message(STATUS "Halide_TARGET: ${Halide_TARGET}")
 if (Halide_TARGET MATCHES "wasm")
     message(WARNING "Skipping apps when building under wasm")
     return()
-endif()
+endif ()
 
 # keep-sorted start case=no ignore_prefixes=#
 # add_app(auto_viz)  # TODO(#5374): missing CMake build

apps/HelloBaremetal/CMakeLists.txt

@@ -12,10 +12,12 @@ add_subdirectory(cmake-twice)
 set(IMAGE ${CMAKE_CURRENT_LIST_DIR}/../images/gray_small.pgm)
 if (EXISTS ${IMAGE})
     configure_file(${IMAGE} gray_small.pgm COPYONLY)
-    add_test(NAME add_filter
-             COMMAND add_filter gray_small.pgm 16 out.pgm)
-    set_tests_properties(add_filter PROPERTIES
-                         LABELS HelloBaremetal
-                         PASS_REGULAR_EXPRESSION "Success!"
-                         SKIP_REGULAR_EXPRESSION "\\[SKIP\\]")
+    add_test(NAME add_filter COMMAND add_filter gray_small.pgm 16 out.pgm)
+    set_tests_properties(
+        add_filter
+        PROPERTIES
+        LABELS HelloBaremetal
+        PASS_REGULAR_EXPRESSION "Success!"
+        SKIP_REGULAR_EXPRESSION "\\[SKIP\\]"
+    )
 endif ()

apps/HelloBaremetal/cmake-external_project/CMakeLists.txt

@@ -17,12 +17,12 @@ set(GEN_BINARY_DIR ${CMAKE_CURRENT_BINARY_DIR}/generator)
 # Build generator executable with host compiler in external child project
 # and export as a package for CMake find_package()
 include(ExternalProject)
-ExternalProject_Add(gen_project
+ExternalProject_Add(
+    gen_project
     SOURCE_DIR ${CMAKE_CURRENT_SOURCE_DIR}/generator
     BINARY_DIR ${GEN_BINARY_DIR}
     INSTALL_COMMAND ""
-    CMAKE_ARGS
-      -DGEN_EXE=${GEN_EXE}
+    CMAKE_ARGS -DGEN_EXE=${GEN_EXE}
 )
 
 
@@ -39,27 +39,25 @@ set(SRC_DIR ${CMAKE_CURRENT_SOURCE_DIR}/..)
 # Instead, declare the generator executable as imported, which is built in external project
 add_executable(${GEN_EXE} IMPORTED)
 # Specify the location of exe file to import
-set_property(TARGET ${GEN_EXE}
-             PROPERTY IMPORTED_LOCATION ${GEN_BINARY_DIR}/${GEN_EXE})
+set_property(TARGET ${GEN_EXE} PROPERTY IMPORTED_LOCATION ${GEN_BINARY_DIR}/${GEN_EXE})
 # Associate with external project so that it invokes build process
 add_dependencies(${GEN_EXE} gen_project)
 
 # Generate filter function
-add_halide_library(add
-                   FROM ${GEN_EXE}
-                   GENERATOR add
-                   # You may set Target feature in addition to what is set in the toolchain file
-                   FEATURES debug
-
-                   # In baremetal, set parallelism=1 because thread support is unavailable.
-                   AUTOSCHEDULER Halide::Mullapudi2016
-                   PARAMS autoscheduler.parallelism=1)
+add_halide_library(
+    add
+    FROM ${GEN_EXE}
+    GENERATOR add
+    # You may set Target feature in addition to what is set in the toolchain file
+    FEATURES debug
+    # In baremetal, set parallelism=1 because thread support is unavailable.
+    AUTOSCHEDULER Halide::Mullapudi2016
+    PARAMS autoscheduler.parallelism=1
+)
 
 # Main executable
 add_executable(add_filter ${SRC_DIR}/filter.cpp)
-target_link_libraries(add_filter PRIVATE
-                      Halide::ImageIO
-                      add)
+target_link_libraries(add_filter PRIVATE Halide::ImageIO add)
 
 # Depending on the target CPU, we need to enable Arm NEON manually
 # if the executable contains any SIMD instructions.

apps/HelloBaremetal/cmake-external_project/generator/CMakeLists.txt

@@ -4,7 +4,7 @@ project(HelloBaremetal-gen)
 
 if (NOT DEFINED GEN_EXE)
     message(FATAL_ERROR "CMakeVariable GEN_EXE is not set")
-endif()
+endif ()
 
 # Set up C++ language settings
 set(CMAKE_CXX_STANDARD 17)
@@ -16,5 +16,4 @@ set(SRC_DIR ${CMAKE_CURRENT_SOURCE_DIR}/../..)
 find_package(Halide REQUIRED)  # We need Halide::Generator
 
 # Build generator
-add_halide_generator(${GEN_EXE}
-                     SOURCES ${SRC_DIR}/add_generator.cpp)
+add_halide_generator(${GEN_EXE} SOURCES ${SRC_DIR}/add_generator.cpp)

apps/HelloBaremetal/cmake-super_build/CMakeLists.txt

@@ -14,27 +14,29 @@ set(GEN_BINARY_DIR ${CMAKE_CURRENT_BINARY_DIR}/generator)
 # Step 1
 # Build generator executable with host compiler in external child project
 # and export as a package for CMake find_package()
-ExternalProject_Add(gen_project
+ExternalProject_Add(
+    gen_project
     SOURCE_DIR ${CMAKE_CURRENT_SOURCE_DIR}/generator
     BINARY_DIR ${GEN_BINARY_DIR}
     INSTALL_COMMAND ""
     CMAKE_ARGS
-      -DGEN_PACKAGE=${GEN_PACKAGE}
-      -DGEN_NAMESPACE=${GEN_NAMESPACE}
-      -DGEN_EXE=${GEN_EXE}
+    -DGEN_PACKAGE=${GEN_PACKAGE}
+    -DGEN_NAMESPACE=${GEN_NAMESPACE}
+    -DGEN_EXE=${GEN_EXE}
 )
 
 # Step 2
 # Build application with cross compiler,
 # where the generator executable built in Step 1 is imported and called
 
-ExternalProject_Add(app_project
+ExternalProject_Add(
+    app_project
     SOURCE_DIR ${CMAKE_CURRENT_SOURCE_DIR}/app
     INSTALL_COMMAND ""
     DEPENDS gen_project  # To make sure generator is ready
     CMAKE_ARGS
-      -DCMAKE_TOOLCHAIN_FILE=${APP_TOOLCHAIN_FILE}
-      -DCMAKE_RUNTIME_OUTPUT_DIRECTORY=${CMAKE_CURRENT_BINARY_DIR}/bin
-      -D${GEN_PACKAGE}_ROOT:FILEPATH=${GEN_BINARY_DIR}
-      -DGEN_TARGET=${GEN_NAMESPACE}${GEN_EXE}
+    -DCMAKE_TOOLCHAIN_FILE=${APP_TOOLCHAIN_FILE}
+    -DCMAKE_RUNTIME_OUTPUT_DIRECTORY=${CMAKE_CURRENT_BINARY_DIR}/bin
+    -D${GEN_PACKAGE}_ROOT:FILEPATH=${GEN_BINARY_DIR}
+    -DGEN_TARGET=${GEN_NAMESPACE}${GEN_EXE}
 )

apps/HelloBaremetal/cmake-super_build/app/CMakeLists.txt

@@ -5,7 +5,7 @@ project(HelloBaremetal-app LANGUAGES C CXX ASM)
 
 if (NOT DEFINED GEN_TARGET)
     message(FATAL_ERROR "CMakeVariable GEN_TARGET is not set")
-endif()
+endif ()
 
 # Set up C++ language settings
 set(CMAKE_CXX_STANDARD 17)
@@ -21,21 +21,20 @@ set(SRC_DIR ${CMAKE_CURRENT_SOURCE_DIR}/../..)
 find_package(HelloBaremetal-add_generator REQUIRED)
 
 # Generate filter function
-add_halide_library(add
-                   FROM ${GEN_TARGET}
-                   GENERATOR add
-                   # You may set Target feature in addition to what is set in the toolchain file
-                   FEATURES debug
-
-                   # In baremetal, set parallelism=1 because thread support is unavailable.
-                   AUTOSCHEDULER Halide::Mullapudi2016
-                   PARAMS autoscheduler.parallelism=1)
+add_halide_library(
+    add
+    FROM ${GEN_TARGET}
+    GENERATOR add
+    # You may set Target feature in addition to what is set in the toolchain file
+    FEATURES debug
+    # In baremetal, set parallelism=1 because thread support is unavailable.
+    AUTOSCHEDULER Halide::Mullapudi2016
+    PARAMS autoscheduler.parallelism=1
+)
 
 # Main executable
 add_executable(add_filter ${SRC_DIR}/filter.cpp)
-target_link_libraries(add_filter PRIVATE
-                      Halide::ImageIO
-                      add)
+target_link_libraries(add_filter PRIVATE Halide::ImageIO add)
 
 # Depending on the target CPU, we need to enable Arm NEON manually
 # if the executable contains any SIMD instructions.