cmake_minimum_required(VERSION 3.20.0)

project( libclc VERSION 0.2.0 LANGUAGES CXX C)

set(CMAKE_CXX_STANDARD 17)

include( GNUInstallDirs )
set_property(DIRECTORY APPEND PROPERTY CMAKE_CONFIGURE_DEPENDS
  amdgcn-amdhsa/lib/SOURCES;
  amdgcn/lib/SOURCES;
  amdgcn-mesa3d/lib/SOURCES;
  amdgpu/lib/SOURCES;
  clspv/lib/SOURCES;
  clspv64/lib/SOURCES;
  generic/lib/SOURCES;
  ptx/lib/SOURCES;
  ptx-nvidiacl/lib/SOURCES;
  r600/lib/SOURCES;
  spirv/lib/SOURCES;
  spirv64/lib/SOURCES
)

# List of all targets
set( LIBCLC_TARGETS_ALL
  amdgcn--
  amdgcn--amdhsa
  clspv--
  clspv64--
  r600--
  nvptx--
  nvptx64--
  nvptx--nvidiacl
  nvptx64--nvidiacl
  spirv-mesa3d-
  spirv64-mesa3d-
)

set( LIBCLC_MIN_LLVM "3.9.0" )

set( LIBCLC_TARGETS_TO_BUILD "all"
    CACHE STRING "Semicolon-separated list of targets to build, or 'all'." )

option( ENABLE_RUNTIME_SUBNORMAL "Enable runtime linking of subnormal support."
OFF )

find_package(LLVM REQUIRED HINTS "${LLVM_CMAKE_DIR}")
include(AddLLVM)

message( "LLVM version: ${LLVM_PACKAGE_VERSION}" )

if( ${LLVM_PACKAGE_VERSION} VERSION_LESS ${LIBCLC_MIN_LLVM} )
	message( FATAL_ERROR "libclc needs at least LLVM ${LIBCLC_MIN_LLVM}" )
endif()

# mesa3d environment is only available since LLVM 4.0
if( ${LLVM_PACKAGE_VERSION} VERSION_GREATER "3.9.0" )
	set( LIBCLC_TARGETS_ALL ${LIBCLC_TARGETS_ALL} amdgcn-mesa-mesa3d )
endif()

if( LIBCLC_TARGETS_TO_BUILD STREQUAL "all" )
	set( LIBCLC_TARGETS_TO_BUILD ${LIBCLC_TARGETS_ALL} )
endif()

find_program( LLVM_CLANG clang PATHS ${LLVM_TOOLS_BINARY_DIR} NO_DEFAULT_PATH )
find_program( LLVM_AS llvm-as PATHS ${LLVM_TOOLS_BINARY_DIR} NO_DEFAULT_PATH )
find_program( LLVM_LINK llvm-link PATHS ${LLVM_TOOLS_BINARY_DIR} NO_DEFAULT_PATH )
find_program( LLVM_OPT opt PATHS ${LLVM_TOOLS_BINARY_DIR} NO_DEFAULT_PATH )
find_program( LLVM_SPIRV llvm-spirv PATHS ${LLVM_TOOLS_BINARY_DIR} NO_DEFAULT_PATH )

# Print toolchain
message( "clang: ${LLVM_CLANG}" )
message( "llvm-as: ${LLVM_AS}" )
message( "llvm-link: ${LLVM_LINK}" )
message( "opt: ${LLVM_OPT}" )
message( "llvm-spirv: ${LLVM_SPIRV}" )
message( "" )
if( NOT LLVM_CLANG OR NOT LLVM_OPT OR NOT LLVM_AS OR NOT LLVM_LINK )
	message( FATAL_ERROR "toolchain incomplete!" )
endif()

list( SORT LIBCLC_TARGETS_TO_BUILD )

if( "spirv-mesa3d-" IN_LIST LIBCLC_TARGETS_TO_BUILD OR "spirv64-mesa3d-" IN_LIST LIBCLC_TARGETS_TO_BUILD )
	if( NOT LLVM_SPIRV )
		message( FATAL_ERROR "SPIR-V targets requested, but spirv-tools is not installed" )
	endif()
endif()

set( CMAKE_MODULE_PATH ${CMAKE_SOURCE_DIR}/cmake )
set( CMAKE_CLC_COMPILER ${LLVM_CLANG} )
set( CMAKE_CLC_ARCHIVE ${LLVM_LINK} )
set( CMAKE_LLAsm_PREPROCESSOR ${LLVM_CLANG} )
set( CMAKE_LLAsm_COMPILER ${LLVM_AS} )
set( CMAKE_LLAsm_ARCHIVE ${LLVM_LINK} )

# Construct LLVM version define
set( LLVM_VERSION_DEFINE "-DHAVE_LLVM=0x${LLVM_VERSION_MAJOR}0${LLVM_VERSION_MINOR}" )


# LLVM 13 enables standard includes by default
if( ${LLVM_PACKAGE_VERSION} VERSION_GREATER "12.99.99" )
				set( CMAKE_LLAsm_FLAGS "${CMAKE_LLAsm_FLAGS} -cl-no-stdinc")
				set( CMAKE_CLC_FLAGS "${CMAKE_CLC_FLAGS} -cl-no-stdinc")
endif()

enable_language( CLC LLAsm )
# This needs to be set before any target that needs it
# We need to use LLVM_INCLUDE_DIRS here, because if we are linking to an
# llvm build directory, this includes $src/llvm/include which is where all the
# headers are not $build/include/ which is what LLVM_INCLUDE_DIR is set to.
include_directories( ${LLVM_INCLUDE_DIRS} )

# Setup prepare_builtins tools
set(LLVM_LINK_COMPONENTS
  BitReader
  BitWriter
  Core
  Support
)
add_llvm_executable( prepare_builtins utils/prepare-builtins.cpp )
target_compile_definitions( prepare_builtins PRIVATE ${LLVM_VERSION_DEFINE} )
# These were not properly reported in early LLVM and we don't need them
target_compile_options( prepare_builtins PRIVATE -fno-rtti -fno-exceptions )

# Setup arch devices
set( r600--_devices cedar cypress barts cayman )
set( amdgcn--_devices tahiti )
set( amdgcn-mesa-mesa3d_devices ${amdgcn--_devices} )
set( amdgcn--amdhsa_devices none )
set( clspv--_devices none )
set( clspv64--_devices none )
set( nvptx--_devices none )
set( nvptx64--_devices none )
set( nvptx--nvidiacl_devices none )
set( nvptx64--nvidiacl_devices none )
set( spirv-mesa3d-_devices none )
set( spirv64-mesa3d-_devices none )

# Setup aliases
set( cedar_aliases palm sumo sumo2 redwood juniper )
set( cypress_aliases hemlock )
set( barts_aliases turks caicos )
set( cayman_aliases aruba )
set( tahiti_aliases pitcairn verde oland hainan bonaire kabini kaveri hawaii
	mullins tonga tongapro iceland carrizo fiji stoney polaris10 polaris11
	gfx602 gfx705 gfx805
	gfx900 gfx902 gfx904 gfx906 gfx908 gfx909 gfx90a gfx90c gfx940 gfx941 gfx942
	gfx1010 gfx1011 gfx1012 gfx1013
	gfx1030 gfx1031 gfx1032 gfx1033 gfx1034 gfx1035 gfx1036
	gfx1100 gfx1101 gfx1102 gfx1103
	gfx1150 gfx1151
	gfx1200 gfx1201
)

# pkg-config file
configure_file( libclc.pc.in libclc.pc @ONLY )
install( FILES ${CMAKE_CURRENT_BINARY_DIR}/libclc.pc DESTINATION "${CMAKE_INSTALL_DATADIR}/pkgconfig" )
install( DIRECTORY generic/include/clc DESTINATION "${CMAKE_INSTALL_INCLUDEDIR}" )

if( ENABLE_RUNTIME_SUBNORMAL )
	add_library( subnormal_use_default STATIC
		generic/lib/subnormal_use_default.ll )
	add_library( subnormal_disable STATIC
		generic/lib/subnormal_disable.ll )
	install( TARGETS subnormal_use_default subnormal_disable ARCHIVE
		DESTINATION "${CMAKE_INSTALL_DATADIR}/clc" )
endif()

find_package( Python3 REQUIRED COMPONENTS Interpreter )
file( TO_CMAKE_PATH ${CMAKE_SOURCE_DIR}/generic/lib/gen_convert.py script_loc )
add_custom_command(
	OUTPUT convert.cl
	COMMAND ${Python3_EXECUTABLE} ${script_loc} > convert.cl
	DEPENDS ${script_loc} )
add_custom_target( "generate_convert.cl" DEPENDS convert.cl )

add_custom_command(
	OUTPUT clspv-convert.cl
	COMMAND ${Python3_EXECUTABLE} ${script_loc} --clspv > clspv-convert.cl
	DEPENDS ${script_loc} )
add_custom_target( "clspv-generate_convert.cl" DEPENDS clspv-convert.cl )

enable_testing()

foreach( t ${LIBCLC_TARGETS_TO_BUILD} )
	message( "BUILDING ${t}" )
	string( REPLACE "-" ";" TRIPLE  ${t} )
	list( GET TRIPLE 0 ARCH )
	list( GET TRIPLE 1 VENDOR )
	list( GET TRIPLE 2 OS )

	set( dirs )

	if ( NOT ${ARCH} STREQUAL spirv AND NOT ${ARCH} STREQUAL spirv64 AND
			 NOT ${ARCH} STREQUAL clspv AND NOT ${ARCH} STREQUAL clspv64)
		LIST( APPEND dirs generic )
	endif()

	if( ${ARCH} STREQUAL r600 OR ${ARCH} STREQUAL amdgcn )
		list( APPEND dirs amdgpu )
	endif()

	#nvptx is special
	if( ${ARCH} STREQUAL nvptx OR ${ARCH} STREQUAL nvptx64 )
		set( DARCH ptx )
	else()
		set( DARCH ${ARCH} )
	endif()

	# Enumerate SOURCES* files
	set( source_list )
	foreach( l ${dirs} ${DARCH} ${DARCH}-${OS} ${DARCH}-${VENDOR}-${OS} )
		foreach( s "SOURCES" "SOURCES_${LLVM_MAJOR}.${LLVM_MINOR}" )
			file( TO_CMAKE_PATH ${l}/lib/${s} file_loc )
			file( TO_CMAKE_PATH ${CMAKE_SOURCE_DIR}/${file_loc} loc )
			# Prepend the location to give higher priority to
			# specialized implementation
			if( EXISTS ${loc} )
				set( source_list ${file_loc} ${source_list} )
			endif()
		endforeach()
	endforeach()

	# Add the generated convert.cl here to prevent adding
	# the one listed in SOURCES
	if( NOT ${ARCH} STREQUAL "spirv" AND NOT ${ARCH} STREQUAL "spirv64" )
		if( NOT ENABLE_RUNTIME_SUBNORMAL AND NOT ${ARCH} STREQUAL "clspv" AND
		    NOT ${ARCH} STREQUAL "clspv64" )
			set( rel_files convert.cl )
			set( objects convert.cl )
			list( APPEND rel_files generic/lib/subnormal_use_default.ll )
		elseif(${ARCH} STREQUAL "clspv" OR ${ARCH} STREQUAL "clspv64")
			set( rel_files clspv-convert.cl )
			set( objects clspv-convert.cl )
		endif()
	else()
		set( rel_files )
		set( objects )
	endif()

	foreach( l ${source_list} )
		file( READ ${l} file_list )
		string( REPLACE "\n" ";" file_list ${file_list} )
		get_filename_component( dir ${l} DIRECTORY )
		foreach( f ${file_list} )
			list( FIND objects ${f} found )
			if( found EQUAL  -1 )
				list( APPEND objects ${f} )
				list( APPEND rel_files ${dir}/${f} )
				# FIXME: This should really go away
				file( TO_CMAKE_PATH ${CMAKE_SOURCE_DIR}/${dir}/${f} src_loc )
				get_filename_component( fdir ${src_loc} DIRECTORY )

				set_source_files_properties( ${dir}/${f}
					PROPERTIES COMPILE_FLAGS "-I ${fdir}" )
			endif()
		endforeach()
	endforeach()

	foreach( d ${${t}_devices} )
		# Some targets don't have a specific GPU to target
		if( ${d} STREQUAL "none" OR ${ARCH} STREQUAL "spirv" OR ${ARCH} STREQUAL "spirv64" )
			set( mcpu )
			set( arch_suffix "${t}" )
		else()
			set( mcpu "-mcpu=${d}" )
			set( arch_suffix "${d}-${t}" )
		endif()
		message( "	DEVICE: ${d} ( ${${d}_aliases} )" )

		if ( ${ARCH} STREQUAL "spirv" OR ${ARCH} STREQUAL "spirv64" )
			if( ${ARCH} STREQUAL "spirv" )
				set( t "spir--" )
			else()
				set( t "spir64--" )
			endif()
			set( build_flags -O0 -finline-hint-functions )
			set( opt_flags )
			set( spvflags --spirv-max-version=1.1 )
		elseif( ${ARCH} STREQUAL "clspv" )
			set( t "spir--" )
			set( build_flags "-Wno-unknown-assumption")
			set( opt_flags -O3 )
		elseif( ${ARCH} STREQUAL "clspv64" )
			set( t "spir64--" )
			set( build_flags "-Wno-unknown-assumption")
			set( opt_flags -O3 )
		else()
			set( build_flags )
			set( opt_flags -O3 )
		endif()

		add_library( builtins.link.${arch_suffix} STATIC ${rel_files} )
		# Make sure we depend on the pseudo target to prevent
		# multiple invocations
		add_dependencies( builtins.link.${arch_suffix}
			generate_convert.cl )
		add_dependencies( builtins.link.${arch_suffix}
			clspv-generate_convert.cl )
		# CMake will turn this include into absolute path
		target_include_directories( builtins.link.${arch_suffix} PRIVATE
			"generic/include" )
		target_compile_definitions( builtins.link.${arch_suffix} PRIVATE
			"__CLC_INTERNAL" )
		string( TOUPPER "-DCLC_${ARCH}" CLC_TARGET_DEFINE )
		target_compile_definitions( builtins.link.${arch_suffix} PRIVATE
			${CLC_TARGET_DEFINE} )
		target_compile_options( builtins.link.${arch_suffix} PRIVATE  -target
			${t} ${mcpu} -fno-builtin -nostdlib ${build_flags} )
		set_target_properties( builtins.link.${arch_suffix} PROPERTIES
			LINKER_LANGUAGE CLC )

		set( obj_suffix ${arch_suffix}.bc )

		# Add opt target
		add_custom_command( OUTPUT "builtins.opt.${obj_suffix}"
				    COMMAND ${LLVM_OPT} ${opt_flags} -o
				    "builtins.opt.${obj_suffix}"
				    "builtins.link.${obj_suffix}"
				    DEPENDS "builtins.link.${arch_suffix}" )
		add_custom_target( "opt.${obj_suffix}" ALL
		                   DEPENDS "builtins.opt.${obj_suffix}" )

		if( ${ARCH} STREQUAL "spirv" OR ${ARCH} STREQUAL "spirv64" )
			set( spv_suffix ${arch_suffix}.spv )
			add_custom_command( OUTPUT "${spv_suffix}"
					    COMMAND ${LLVM_SPIRV} ${spvflags}
					    -o "${spv_suffix}"
					    "builtins.link.${obj_suffix}"
					    DEPENDS "builtins.link.${arch_suffix}" )
			add_custom_target( "prepare-${spv_suffix}" ALL
			                   DEPENDS "${spv_suffix}" )
			install( FILES ${CMAKE_CURRENT_BINARY_DIR}/${spv_suffix}
				 DESTINATION "${CMAKE_INSTALL_DATADIR}/clc" )
		else()

			# Add prepare target
			add_custom_command( OUTPUT "${obj_suffix}"
				            COMMAND prepare_builtins -o
					    "${obj_suffix}"
					    "builtins.opt.${obj_suffix}"
					    DEPENDS "opt.${obj_suffix}"
					            "builtins.opt.${obj_suffix}"
					            prepare_builtins )
			add_custom_target( "prepare-${obj_suffix}" ALL
					   DEPENDS "${obj_suffix}" )

			# nvptx-- targets don't include workitem builtins
			if( NOT ${t} MATCHES ".*ptx.*--$" )
				add_test( NAME external-calls-${obj_suffix}
   					  COMMAND ./check_external_calls.sh ${CMAKE_CURRENT_BINARY_DIR}/${obj_suffix} ${LLVM_TOOLS_BINARY_DIR}
					  WORKING_DIRECTORY ${CMAKE_SOURCE_DIR} )
			endif()

			install( FILES ${CMAKE_CURRENT_BINARY_DIR}/${obj_suffix} DESTINATION "${CMAKE_INSTALL_DATADIR}/clc" )
			foreach( a ${${d}_aliases} )
				set( alias_suffix "${a}-${t}.bc" )
				add_custom_target( ${alias_suffix} ALL
						   COMMAND ${CMAKE_COMMAND} -E
						   create_symlink ${obj_suffix}
						   ${alias_suffix}
				                   DEPENDS "prepare-${obj_suffix}" )
				install( FILES ${CMAKE_CURRENT_BINARY_DIR}/${alias_suffix} DESTINATION "${CMAKE_INSTALL_DATADIR}/clc" )
			endforeach( a )
		endif()
	endforeach( d )
endforeach( t )