https://github.com/halide/Halide
Tip revision: f9e4c7878385f43cf88cca23d5bd663233e9e7da authored by Steven Johnson on 27 April 2021, 19:14:54 UTC
Add support for dynamic tensors to hannk (#5942)
Add support for dynamic tensors to hannk (#5942)
Tip revision: f9e4c78
CMakeLists.txt
##
# Lists of source files. Keep ALL lists sorted in alphabetical order.
##
# The externally-visible header files that go into making Halide.h.
# Don't include anything here that includes llvm headers.
set(HEADER_FILES
AddAtomicMutex.h
AddImageChecks.h
AddParameterChecks.h
AlignLoads.h
AllocationBoundsInference.h
ApplySplit.h
Argument.h
AssociativeOpsTable.h
Associativity.h
AsyncProducers.h
AutoScheduleUtils.h
BoundaryConditions.h
Bounds.h
BoundsInference.h
BoundSmallAllocations.h
Buffer.h
CanonicalizeGPUVars.h
Closure.h
CodeGen_C.h
CodeGen_D3D12Compute_Dev.h
CodeGen_GPU_Dev.h
CodeGen_Internal.h
CodeGen_LLVM.h
CodeGen_Metal_Dev.h
CodeGen_OpenCL_Dev.h
CodeGen_OpenGLCompute_Dev.h
CodeGen_Posix.h
CodeGen_PTX_Dev.h
CodeGen_PyTorch.h
CodeGen_Targets.h
CompilerLogger.h
ConciseCasts.h
CPlusPlusMangle.h
CSE.h
Debug.h
DebugArguments.h
DebugToFile.h
Definition.h
Deinterleave.h
Derivative.h
DerivativeUtils.h
DeviceAPI.h
DeviceArgument.h
DeviceInterface.h
Dimension.h
EarlyFree.h
Elf.h
EliminateBoolVectors.h
EmulateFloat16Math.h
Error.h
Expr.h
ExprUsesVar.h
Extern.h
ExternFuncArgument.h
FastIntegerDivide.h
FindCalls.h
FindIntrinsics.h
FlattenNestedRamps.h
Float16.h
Func.h
Function.h
FunctionPtr.h
FuseGPUThreadLoops.h
FuzzFloatStores.h
Generator.h
HexagonOffload.h
HexagonOptimize.h
ImageParam.h
InferArguments.h
InjectHostDevBufferCopies.h
Inline.h
InlineReductions.h
IntegerDivisionTable.h
Interval.h
Introspection.h
IntrusivePtr.h
IR.h
IREquality.h
IRMatch.h
IRMutator.h
IROperator.h
IRPrinter.h
IRVisitor.h
JITModule.h
Lambda.h
Lerp.h
LICM.h
LLVM_Output.h
LLVM_Runtime_Linker.h
LoopCarry.h
Lower.h
LowerWarpShuffles.h
MainPage.h
MatlabWrapper.h
Memoization.h
Module.h
ModulusRemainder.h
Monotonic.h
ObjectInstanceRegistry.h
OffloadGPULoops.h
OutputImageParam.h
ParallelRVar.h
Param.h
Parameter.h
ParamMap.h
PartitionLoops.h
Pipeline.h
Prefetch.h
Profiling.h
PurifyIndexMath.h
PythonExtensionGen.h
Qualify.h
Random.h
RDom.h
Realization.h
RealizationOrder.h
RebaseLoopsToZero.h
Reduction.h
RegionCosts.h
RemoveDeadAllocations.h
RemoveExternLoops.h
RemoveUndef.h
runtime/HalideBuffer.h
runtime/HalideRuntime.h
Schedule.h
ScheduleFunctions.h
Scope.h
SelectGPUAPI.h
Simplify.h
SimplifyCorrelatedDifferences.h
SimplifySpecializations.h
SkipStages.h
SlidingWindow.h
Solve.h
SplitTuples.h
StmtToHtml.h
StorageFlattening.h
StorageFolding.h
StrictifyFloat.h
Substitute.h
Target.h
ThreadPool.h
Tracing.h
TrimNoOps.h
Tuple.h
Type.h
UnifyDuplicateLets.h
UniquifyVariableNames.h
UnpackBuffers.h
UnrollLoops.h
UnsafePromises.h
Util.h
Var.h
VectorizeLoops.h
WasmExecutor.h
WrapCalls.h
)
set(SOURCE_FILES
AddAtomicMutex.cpp
AddImageChecks.cpp
AddParameterChecks.cpp
AlignLoads.cpp
AllocationBoundsInference.cpp
ApplySplit.cpp
Argument.cpp
AssociativeOpsTable.cpp
Associativity.cpp
AsyncProducers.cpp
AutoScheduleUtils.cpp
BoundaryConditions.cpp
Bounds.cpp
BoundsInference.cpp
BoundSmallAllocations.cpp
Buffer.cpp
CanonicalizeGPUVars.cpp
Closure.cpp
CodeGen_ARM.cpp
CodeGen_C.cpp
CodeGen_D3D12Compute_Dev.cpp
CodeGen_GPU_Dev.cpp
CodeGen_Hexagon.cpp
CodeGen_Internal.cpp
CodeGen_LLVM.cpp
CodeGen_Metal_Dev.cpp
CodeGen_MIPS.cpp
CodeGen_OpenCL_Dev.cpp
CodeGen_OpenGLCompute_Dev.cpp
CodeGen_Posix.cpp
CodeGen_PowerPC.cpp
CodeGen_PTX_Dev.cpp
CodeGen_PyTorch.cpp
CodeGen_RISCV.cpp
CodeGen_WebAssembly.cpp
CodeGen_X86.cpp
CompilerLogger.cpp
CPlusPlusMangle.cpp
CSE.cpp
Debug.cpp
DebugArguments.cpp
DebugToFile.cpp
Definition.cpp
Deinterleave.cpp
Derivative.cpp
DerivativeUtils.cpp
DeviceArgument.cpp
DeviceInterface.cpp
Dimension.cpp
EarlyFree.cpp
Elf.cpp
EliminateBoolVectors.cpp
EmulateFloat16Math.cpp
Error.cpp
Expr.cpp
FastIntegerDivide.cpp
FindCalls.cpp
FindIntrinsics.cpp
FlattenNestedRamps.cpp
Float16.cpp
Func.cpp
Function.cpp
FuseGPUThreadLoops.cpp
FuzzFloatStores.cpp
Generator.cpp
HexagonOffload.cpp
HexagonOptimize.cpp
ImageParam.cpp
InferArguments.cpp
InjectHostDevBufferCopies.cpp
Inline.cpp
InlineReductions.cpp
IntegerDivisionTable.cpp
Interval.cpp
Introspection.cpp
IR.cpp
IREquality.cpp
IRMatch.cpp
IRMutator.cpp
IROperator.cpp
IRPrinter.cpp
IRVisitor.cpp
JITModule.cpp
Lambda.cpp
Lerp.cpp
LICM.cpp
LLVM_Output.cpp
LLVM_Runtime_Linker.cpp
LoopCarry.cpp
Lower.cpp
LowerWarpShuffles.cpp
MatlabWrapper.cpp
Memoization.cpp
Module.cpp
ModulusRemainder.cpp
Monotonic.cpp
ObjectInstanceRegistry.cpp
OffloadGPULoops.cpp
OutputImageParam.cpp
ParallelRVar.cpp
Parameter.cpp
ParamMap.cpp
PartitionLoops.cpp
Pipeline.cpp
Prefetch.cpp
PrintLoopNest.cpp
Profiling.cpp
PurifyIndexMath.cpp
PythonExtensionGen.cpp
Qualify.cpp
Random.cpp
RDom.cpp
Realization.cpp
RealizationOrder.cpp
RebaseLoopsToZero.cpp
Reduction.cpp
RegionCosts.cpp
RemoveDeadAllocations.cpp
RemoveExternLoops.cpp
RemoveUndef.cpp
Schedule.cpp
ScheduleFunctions.cpp
SelectGPUAPI.cpp
Simplify.cpp
Simplify_Add.cpp
Simplify_And.cpp
Simplify_Call.cpp
Simplify_Cast.cpp
Simplify_Div.cpp
Simplify_EQ.cpp
Simplify_Exprs.cpp
Simplify_Let.cpp
Simplify_LT.cpp
Simplify_Max.cpp
Simplify_Min.cpp
Simplify_Mod.cpp
Simplify_Mul.cpp
Simplify_Not.cpp
Simplify_Or.cpp
Simplify_Select.cpp
Simplify_Shuffle.cpp
Simplify_Stmts.cpp
Simplify_Sub.cpp
SimplifyCorrelatedDifferences.cpp
SimplifySpecializations.cpp
SkipStages.cpp
SlidingWindow.cpp
Solve.cpp
SplitTuples.cpp
StmtToHtml.cpp
StorageFlattening.cpp
StorageFolding.cpp
StrictifyFloat.cpp
Substitute.cpp
Target.cpp
Tracing.cpp
TrimNoOps.cpp
Tuple.cpp
Type.cpp
UnifyDuplicateLets.cpp
UniquifyVariableNames.cpp
UnpackBuffers.cpp
UnrollLoops.cpp
UnsafePromises.cpp
Util.cpp
Var.cpp
VectorizeLoops.cpp
WasmExecutor.cpp
WrapCalls.cpp
)
##
# Build and import the runtime.
##
add_subdirectory(runtime)
##
# Build the Halide mono-header.
##
set(HALIDE_H "${Halide_BINARY_DIR}/include/Halide.h")
set(LICENSE_PATH "${Halide_SOURCE_DIR}/LICENSE.txt")
add_custom_command(OUTPUT "${Halide_BINARY_DIR}/include/Halide.h"
COMMAND ${CMAKE_COMMAND} -E make_directory "$<SHELL_PATH:${Halide_BINARY_DIR}/include>"
COMMAND build_halide_h "$<SHELL_PATH:${LICENSE_PATH}>" ${HEADER_FILES} > "$<SHELL_PATH:${HALIDE_H}>"
DEPENDS build_halide_h "${LICENSE_PATH}" ${HEADER_FILES}
WORKING_DIRECTORY "${CMAKE_CURRENT_LIST_DIR}"
VERBATIM)
add_custom_target(HalideIncludes ALL DEPENDS "${Halide_BINARY_DIR}/include/Halide.h")
##
# Define the Halide library target.
##
add_library(Halide
${SOURCE_FILES}
${HEADER_FILES}
# Including these as sources works around the need to "install" Halide_initmod
$<TARGET_OBJECTS:Halide_initmod>)
add_library(Halide::Halide ALIAS Halide)
target_link_libraries(Halide PRIVATE Halide::LLVM)
target_link_libraries(Halide PUBLIC Halide::LanguageOptions)
target_compile_definitions(Halide PRIVATE $<$<STREQUAL:$<TARGET_PROPERTY:TYPE>,STATIC_LIBRARY>:Halide_STATIC_DEFINE>)
target_compile_features(Halide PUBLIC cxx_std_11)
include(TargetExportScript)
## TODO: implement something similar for Windows/link.exe
# https://github.com/halide/Halide/issues/4651
target_export_script(Halide
APPLE_LD "${CMAKE_CURRENT_LIST_DIR}/exported_symbols.osx"
GNU_LD "${CMAKE_CURRENT_LIST_DIR}/exported_symbols.ldscript")
set_target_properties(Halide PROPERTIES
POSITION_INDEPENDENT_CODE ON
VERSION ${Halide_VERSION}
SOVERSION ${Halide_VERSION_MAJOR})
target_include_directories(Halide INTERFACE "$<BUILD_INTERFACE:${Halide_BINARY_DIR}/include>")
add_dependencies(Halide HalideIncludes)
option(Halide_WITH_INTROSPECTION "Enable use of debugging symbols for default Func, Var, etc. names" ON)
if (Halide_WITH_INTROSPECTION)
target_compile_definitions(Halide PRIVATE WITH_INTROSPECTION)
endif ()
if (TARGET wabt-obj)
target_link_libraries(Halide PRIVATE wabt-obj)
target_compile_definitions(Halide PRIVATE WITH_WABT)
endif ()
##
# Set compiler options for libHalide
##
target_compile_options(
Halide
PRIVATE
$<$<CXX_COMPILER_ID:GNU,Clang,AppleClang>:-Wall>
$<$<CXX_COMPILER_ID:GNU,Clang,AppleClang>:-Wcast-qual>
$<$<CXX_COMPILER_ID:GNU,Clang,AppleClang>:-Wignored-qualifiers>
$<$<CXX_COMPILER_ID:GNU,Clang,AppleClang>:-Woverloaded-virtual>
$<$<CXX_COMPILER_ID: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>:-Wno-deprecated-declarations>
$<$<CXX_COMPILER_ID:GNU,Clang,AppleClang>:-Wno-double-promotion>
$<$<CXX_COMPILER_ID:GNU,Clang,AppleClang>:-Wno-float-conversion>
$<$<CXX_COMPILER_ID:GNU,Clang,AppleClang>:-Wno-float-equal>
$<$<CXX_COMPILER_ID:GNU,Clang,AppleClang>:-Wno-missing-field-initializers>
$<$<CXX_COMPILER_ID:GNU,Clang,AppleClang>:-Wno-old-style-cast>
$<$<CXX_COMPILER_ID:GNU,Clang,AppleClang>:-Wno-shadow>
$<$<CXX_COMPILER_ID:GNU,Clang,AppleClang>:-Wno-sign-conversion>
$<$<CXX_COMPILER_ID:GNU,Clang,AppleClang>:-Wno-switch-enum>
$<$<CXX_COMPILER_ID:GNU,Clang,AppleClang>:-Wno-undef>
$<$<CXX_COMPILER_ID:GNU,Clang,AppleClang>:-Wno-unused-function>
$<$<CXX_COMPILER_ID:GNU,Clang,AppleClang>:-Wno-unused-macros>
$<$<CXX_COMPILER_ID:GNU,Clang,AppleClang>:-Wno-unused-parameter>
$<$<CXX_COMPILER_ID:Clang,AppleClang>:-Wno-c++98-compat-pedantic>
$<$<CXX_COMPILER_ID:Clang,AppleClang>:-Wno-c++98-compat>
$<$<CXX_COMPILER_ID:Clang,AppleClang>:-Wno-cast-align>
$<$<CXX_COMPILER_ID:Clang,AppleClang>:-Wno-comma>
$<$<CXX_COMPILER_ID:Clang,AppleClang>:-Wno-covered-switch-default>
$<$<CXX_COMPILER_ID:Clang,AppleClang>:-Wno-documentation-unknown-command>
$<$<CXX_COMPILER_ID:Clang,AppleClang>:-Wno-documentation>
$<$<CXX_COMPILER_ID:Clang,AppleClang>:-Wno-exit-time-destructors>
$<$<CXX_COMPILER_ID:Clang,AppleClang>:-Wno-global-constructors>
$<$<CXX_COMPILER_ID:Clang,AppleClang>:-Wno-implicit-float-conversion>
$<$<CXX_COMPILER_ID:Clang,AppleClang>:-Wno-implicit-int-conversion>
$<$<CXX_COMPILER_ID:Clang,AppleClang>:-Wno-implicit-int-float-conversion>
$<$<CXX_COMPILER_ID:Clang,AppleClang>:-Wno-missing-prototypes>
$<$<CXX_COMPILER_ID:Clang,AppleClang>:-Wno-nonportable-system-include-path>
$<$<CXX_COMPILER_ID:Clang,AppleClang>:-Wno-reserved-id-macro>
$<$<CXX_COMPILER_ID:Clang,AppleClang>:-Wno-return-std-move-in-c++11>
$<$<CXX_COMPILER_ID:Clang,AppleClang>:-Wno-shadow-field-in-constructor>
$<$<CXX_COMPILER_ID:Clang,AppleClang>:-Wno-shadow-field>
$<$<CXX_COMPILER_ID:Clang,AppleClang>:-Wno-shorten-64-to-32>
$<$<CXX_COMPILER_ID:Clang,AppleClang>:-Wno-undefined-func-template>
$<$<CXX_COMPILER_ID:Clang,AppleClang>:-Wno-unused-member-function>
$<$<CXX_COMPILER_ID:Clang,AppleClang>:-Wno-unused-template>
$<$<CXX_COMPILER_ID:MSVC>:/W3>
$<$<CXX_COMPILER_ID:MSVC>:/wd4018> # 4018: disable "signed/unsigned mismatch"
$<$<CXX_COMPILER_ID:MSVC>:/wd4141> # 4141: 'inline' used more than once
$<$<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>:/wd4291> # 4291: No matching operator delete found
$<$<CXX_COMPILER_ID:MSVC>:/wd4503> # 4503: disable "decorated name length exceeded, name was truncated"
$<$<CXX_COMPILER_ID:MSVC>:/wd4800> # 4800: forcing value to bool 'true' or 'false' (performance warning)
$<$<CXX_COMPILER_ID:MSVC>:/wd4996> # 4996: compiler encountered deprecated declaration
)
if (CMAKE_GENERATOR MATCHES "Visual Studio")
# We could expose the /MP flag to all targets, but that might end up saturating the build
# since multiple MSBuild projects might get built in parallel, each of which compiling their
# source files in parallel; the Halide library itself is a "knot" point of the build graph,
# so compiling its files in parallel should not oversubscribe the system
target_compile_options(Halide PRIVATE $<$<CXX_COMPILER_ID:MSVC>:/MP>)
endif ()
target_compile_definitions(Halide
PRIVATE
# Disable warnings about standard C functions that have more secure replacements
# in the Windows API.
$<$<CXX_COMPILER_ID:MSVC>:_CRT_SECURE_NO_WARNINGS>
$<$<CXX_COMPILER_ID:MSVC>:_SCL_SECURE_NO_WARNINGS>
)
##
# Set up additional backend options for Halide
##
option(TARGET_OPENCL "Include OpenCL-C target" ON)
if (TARGET_OPENCL)
target_compile_definitions(Halide PRIVATE WITH_OPENCL)
endif ()
option(TARGET_METAL "Include Metal target" ON)
if (TARGET_METAL)
target_compile_definitions(Halide PRIVATE WITH_METAL)
endif ()
option(TARGET_D3D12COMPUTE "Include Direct3D 12 Compute target" ON)
if (TARGET_D3D12COMPUTE)
target_compile_definitions(Halide PRIVATE WITH_D3D12)
endif ()
option(TARGET_OPENGLCOMPUTE "Include OpenGLCompute target" ON)
if (TARGET_OPENGLCOMPUTE)
target_compile_definitions(Halide PRIVATE WITH_OPENGLCOMPUTE)
endif ()
##
# Add autoschedulers to the build.
##
if (BUILD_SHARED_LIBS)
message(STATUS "Building autoschedulers enabled")
add_subdirectory(autoschedulers)
else ()
message(STATUS "Building autoschedulers disabled (static Halide)")
endif ()