A struct representing a target machine and os to generate code for. More...

#include <Target.h>

Public Types
enum	OS { OSUnknown = 0, Linux, Windows, OSX, Android, IOS, QuRT, NoOS, Fuchsia, WebAssemblyRuntime }
	The operating system used by the target. More...

enum	Arch { ArchUnknown = 0, X86, ARM, Hexagon, POWERPC, WebAssembly, RISCV }
	The architecture used by the target. More...

enum	Processor { ProcessorGeneric = 0, K8, K8_SSE3, AMDFam10, BtVer1, BdVer1, BdVer2, BdVer3, BdVer4, BtVer2, ZnVer1, ZnVer2, ZnVer3 }
	The specific processor to be targeted, tuned for. More...

enum	Feature { JIT = halide_target_feature_jit, Debug = halide_target_feature_debug, NoAsserts = halide_target_feature_no_asserts, NoBoundsQuery = halide_target_feature_no_bounds_query, SSE41 = halide_target_feature_sse41, AVX = halide_target_feature_avx, AVX2 = halide_target_feature_avx2, FMA = halide_target_feature_fma, FMA4 = halide_target_feature_fma4, F16C = halide_target_feature_f16c, ARMv7s = halide_target_feature_armv7s, NoNEON = halide_target_feature_no_neon, VSX = halide_target_feature_vsx, POWER_ARCH_2_07 = halide_target_feature_power_arch_2_07, CUDA = halide_target_feature_cuda, CUDACapability30 = halide_target_feature_cuda_capability30, CUDACapability32 = halide_target_feature_cuda_capability32, CUDACapability35 = halide_target_feature_cuda_capability35, CUDACapability50 = halide_target_feature_cuda_capability50, CUDACapability61 = halide_target_feature_cuda_capability61, CUDACapability70 = halide_target_feature_cuda_capability70, CUDACapability75 = halide_target_feature_cuda_capability75, CUDACapability80 = halide_target_feature_cuda_capability80, CUDACapability86 = halide_target_feature_cuda_capability86, OpenCL = halide_target_feature_opencl, CLDoubles = halide_target_feature_cl_doubles, CLHalf = halide_target_feature_cl_half, CLAtomics64 = halide_target_feature_cl_atomic64, OpenGLCompute = halide_target_feature_openglcompute, EGL = halide_target_feature_egl, UserContext = halide_target_feature_user_context, Profile = halide_target_feature_profile, NoRuntime = halide_target_feature_no_runtime, Metal = halide_target_feature_metal, CPlusPlusMangling = halide_target_feature_c_plus_plus_mangling, LargeBuffers = halide_target_feature_large_buffers, HexagonDma = halide_target_feature_hexagon_dma, HVX_128 = halide_target_feature_hvx_128, HVX = HVX_128, HVX_v62 = halide_target_feature_hvx_v62, HVX_v65 = halide_target_feature_hvx_v65, HVX_v66 = halide_target_feature_hvx_v66, FuzzFloatStores = halide_target_feature_fuzz_float_stores, SoftFloatABI = halide_target_feature_soft_float_abi, MSAN = halide_target_feature_msan, AVX512 = halide_target_feature_avx512, AVX512_KNL = halide_target_feature_avx512_knl, AVX512_Skylake = halide_target_feature_avx512_skylake, AVX512_Cannonlake = halide_target_feature_avx512_cannonlake, AVX512_SapphireRapids = halide_target_feature_avx512_sapphirerapids, TraceLoads = halide_target_feature_trace_loads, TraceStores = halide_target_feature_trace_stores, TraceRealizations = halide_target_feature_trace_realizations, TracePipeline = halide_target_feature_trace_pipeline, D3D12Compute = halide_target_feature_d3d12compute, StrictFloat = halide_target_feature_strict_float, TSAN = halide_target_feature_tsan, ASAN = halide_target_feature_asan, CheckUnsafePromises = halide_target_feature_check_unsafe_promises, EmbedBitcode = halide_target_feature_embed_bitcode, EnableLLVMLoopOpt = halide_target_feature_enable_llvm_loop_opt, WasmSimd128 = halide_target_feature_wasm_simd128, WasmSignExt = halide_target_feature_wasm_signext, WasmSatFloatToInt = halide_target_feature_wasm_sat_float_to_int, WasmThreads = halide_target_feature_wasm_threads, WasmBulkMemory = halide_target_feature_wasm_bulk_memory, WebGPU = halide_target_feature_webgpu, SVE = halide_target_feature_sve, SVE2 = halide_target_feature_sve2, ARMDotProd = halide_target_feature_arm_dot_prod, ARMFp16 = halide_target_feature_arm_fp16, LLVMLargeCodeModel = halide_llvm_large_code_model, RVV = halide_target_feature_rvv, ARMv81a = halide_target_feature_armv81a, SanitizerCoverage = halide_target_feature_sanitizer_coverage, ProfileByTimer = halide_target_feature_profile_by_timer, SPIRV = halide_target_feature_spirv, Vulkan = halide_target_feature_vulkan, VulkanInt8 = halide_target_feature_vulkan_int8, VulkanInt16 = halide_target_feature_vulkan_int16, VulkanInt64 = halide_target_feature_vulkan_int64, VulkanFloat16 = halide_target_feature_vulkan_float16, VulkanFloat64 = halide_target_feature_vulkan_float64, VulkanV10 = halide_target_feature_vulkan_version10, VulkanV12 = halide_target_feature_vulkan_version12, VulkanV13 = halide_target_feature_vulkan_version13, Semihosting = halide_target_feature_semihosting, FeatureEnd = halide_target_feature_end }
	Optional features a target can have. More...

Public Member Functions
	Target ()=default

	Target (OS o, Arch a, int b, Processor pt, const std::vector< Feature > &initial_features=std::vector< Feature >(), int vb=0)

	Target (OS o, Arch a, int b, const std::vector< Feature > &initial_features=std::vector< Feature >())

	Target (const std::string &s)
	Given a string of the form used in HL_TARGET (e.g. More...

	Target (const char *s)

bool	has_unknowns () const
	Return true if any of the arch/bits/os fields are "unknown"/0; return false otherwise. More...

void	set_feature (Feature f, bool value=true)

void	set_features (const std::vector< Feature > &features_to_set, bool value=true)

bool	has_feature (Feature f) const

bool	has_feature (halide_target_feature_t f) const

bool	features_any_of (const std::vector< Feature > &test_features) const

bool	features_all_of (const std::vector< Feature > &test_features) const

Target	with_feature (Feature f) const
	Return a copy of the target with the given feature set. More...

Target	without_feature (Feature f) const
	Return a copy of the target with the given feature cleared. More...

bool	has_gpu_feature () const
	Is a fully feature GPU compute runtime enabled? I.e. More...

bool	supports_type (const Type &t) const
	Does this target allow using a certain type. More...

bool	supports_type (const Type &t, DeviceAPI device) const
	Does this target allow using a certain type on a certain device. More...

bool	supports_device_api (DeviceAPI api) const
	Returns whether a particular device API can be used with this Target. More...

DeviceAPI	get_required_device_api () const
	If this Target (including all Features) requires a specific DeviceAPI, return it. More...

bool	operator== (const Target &other) const

bool	operator!= (const Target &other) const

bool	get_runtime_compatible_target (const Target &other, Target &result)
	Create a "greatest common denominator" runtime target that is compatible with both this target and `other`. More...

std::string	to_string () const
	Convert the Target into a string form that can be reconstituted by merge_string(), which will always be of the form. More...

int	natural_vector_size (const Halide::Type &t) const
	Given a data type, return an estimate of the "natural" vector size for that data type when compiling for this Target. More...

template<typename data_t >
int	natural_vector_size () const
	Given a data type, return an estimate of the "natural" vector size for that data type when compiling for this Target. More...

bool	has_large_buffers () const
	Return true iff 64 bits and has_feature(LargeBuffers). More...

int64_t	maximum_buffer_size () const
	Return the maximum buffer size in bytes supported on this Target. More...

int	get_cuda_capability_lower_bound () const
	Get the minimum cuda capability found as an integer. More...

int	get_vulkan_capability_lower_bound () const
	Get the minimum Vulkan capability found as an integer. More...

bool	supported () const
	Was libHalide compiled with support for this target? More...

const std::bitset< FeatureEnd > &	get_features_bitset () const
	Return a bitset of the Featuress set in this Target (set = 1). More...

Static Public Member Functions
static bool	validate_target_string (const std::string &s)
	Check if a target string is valid. More...

static std::string	feature_to_name (Target::Feature feature)
	Return the name corresponding to a given Feature, in the form used to construct Target strings (e.g., Feature::Debug is "debug" and not "Debug"). More...

static Target::Feature	feature_from_name (const std::string &name)
	Return the feature corresponding to a given name, in the form used to construct Target strings (e.g., Feature::Debug is "debug" and not "Debug"). More...

Public Attributes
enum Halide::Target::OS	os = OSUnknown

enum Halide::Target::Arch	arch = ArchUnknown

int	bits = 0
	The bit-width of the target machine. More...

int	vector_bits = 0
	The bit-width of a vector register for targets where this is configurable and targeting a fixed size is desired. More...

enum Halide::Target::Processor	processor_tune = ProcessorGeneric

Detailed Description

A struct representing a target machine and os to generate code for.

Examples: tutorial/lesson_11_cross_compilation.cpp, and tutorial/lesson_12_using_the_gpu.cpp.

Definition at line 19 of file Target.h.

Member Enumeration Documentation

◆ OS

enum Halide::Target::OS

The operating system used by the target.

Determines which system calls to generate. Corresponds to os_name_map in Target.cpp.

Enumerator
OSUnknown
Linux
Windows
OSX
Android
IOS
QuRT
NoOS
Fuchsia
WebAssemblyRuntime

Definition at line 23 of file Target.h.

◆ Arch

enum Halide::Target::Arch

The architecture used by the target.

Determines the instruction set to use. Corresponds to arch_name_map in Target.cpp.

Enumerator
ArchUnknown
X86
ARM
Hexagon
POWERPC
WebAssembly
RISCV

Definition at line 39 of file Target.h.

◆ Processor

enum Halide::Target::Processor

The specific processor to be targeted, tuned for.

Corresponds to processor_name_map in Target.cpp.

New entries should be added to the end.

Enumerator
ProcessorGeneric	Do not tune for any specific CPU. In practice, this means that halide will decide the tune CPU based on the enabled features.
K8
K8_SSE3	Tune for AMD K8 Hammer CPU (AMD Family 0Fh, launched 2003).
AMDFam10	Tune for later versions of AMD K8 CPU, with SSE3 support.
BtVer1	Tune for AMD K10 "Barcelona" CPU (AMD Family 10h, launched 2007).
BdVer1	Tune for AMD Bobcat CPU (AMD Family 14h, launched 2011).
BdVer2	Tune for AMD Bulldozer CPU (AMD Family 15h, launched 2011).
BdVer3	Tune for AMD Piledriver CPU (AMD Family 15h (2nd-gen), launched 2012).
BdVer4	Tune for AMD Steamroller CPU (AMD Family 15h (3nd-gen), launched 2014).
BtVer2	Tune for AMD Excavator CPU (AMD Family 15h (4th-gen), launched 2015).
ZnVer1	Tune for AMD Jaguar CPU (AMD Family 16h, launched 2013).
ZnVer2	Tune for AMD Zen CPU (AMD Family 17h, launched 2017).
ZnVer3	Tune for AMD Zen 2 CPU (AMD Family 17h, launched 2019).

Definition at line 61 of file Target.h.

◆ Feature

enum Halide::Target::Feature

Optional features a target can have.

Corresponds to feature_name_map in Target.cpp. See definitions in HalideRuntime.h for full information.

Enumerator
JIT
Debug
NoAsserts
NoBoundsQuery
SSE41
AVX
AVX2
FMA
FMA4
F16C
ARMv7s
NoNEON
VSX
POWER_ARCH_2_07
CUDA
CUDACapability30
CUDACapability32
CUDACapability35
CUDACapability50
CUDACapability61
CUDACapability70
CUDACapability75
CUDACapability80
CUDACapability86
OpenCL
CLDoubles
CLHalf
CLAtomics64
OpenGLCompute
EGL
UserContext
Profile
NoRuntime
Metal
CPlusPlusMangling
LargeBuffers
HexagonDma
HVX_128
HVX
HVX_v62
HVX_v65
HVX_v66
FuzzFloatStores
SoftFloatABI
MSAN
AVX512
AVX512_KNL
AVX512_Skylake
AVX512_Cannonlake
AVX512_SapphireRapids
TraceLoads
TraceStores
TraceRealizations
TracePipeline
D3D12Compute
StrictFloat
TSAN
ASAN
CheckUnsafePromises
EmbedBitcode
EnableLLVMLoopOpt
WasmSimd128
WasmSignExt
WasmSatFloatToInt
WasmThreads
WasmBulkMemory
WebGPU
SVE
SVE2
ARMDotProd
ARMFp16
LLVMLargeCodeModel
RVV
ARMv81a
SanitizerCoverage
ProfileByTimer
SPIRV
Vulkan
VulkanInt8
VulkanInt16
VulkanInt64
VulkanFloat16
VulkanFloat64
VulkanV10
VulkanV12
VulkanV13
Semihosting
FeatureEnd

Definition at line 82 of file Target.h.

Constructor & Destructor Documentation

◆ Target() [1/5]

Halide::Target::Target ( )

default

◆ Target() [2/5]

Halide::Target::Target	(	OS	o,
		Arch	a,
		int	b,
		Processor	pt,
		const std::vector< Feature > &	initial_features = `std::vector<Feature>()`,
		int	vb = `0`
	)

inline

Definition at line 173 of file Target.h.

References set_feature().

◆ Target() [3/5]

Halide::Target::Target	(	OS	o,
		Arch	a,
		int	b,
		const std::vector< Feature > &	initial_features = `std::vector<Feature>()`
	)

inline

Definition at line 181 of file Target.h.

◆ Target() [4/5]

Halide::Target::Target ( const std::string & s )

explicit

Given a string of the form used in HL_TARGET (e.g.

"x86-64-avx"), construct the Target it specifies. Note that this always starts with the result of get_host_target(), replacing only the parts found in the target string, so if you omit (say) an OS specification, the host OS will be used instead. An empty string is exactly equivalent to get_host_target().

Invalid target strings will fail with a user_error.

◆ Target() [5/5]

Halide::Target::Target ( const char * s )

explicit

Member Function Documentation

◆ validate_target_string()

static bool Halide::Target::validate_target_string ( const std::string & s )

static

Check if a target string is valid.

◆ has_unknowns()

bool Halide::Target::has_unknowns ( ) const

Return true if any of the arch/bits/os fields are "unknown"/0; return false otherwise.

◆ set_feature()

void Halide::Target::set_feature	(	Feature	f,
		bool	value = `true`
	)

Referenced by Target().

◆ set_features()

void Halide::Target::set_features	(	const std::vector< Feature > &	features_to_set,
		bool	value = `true`
	)

Examples: tutorial/lesson_11_cross_compilation.cpp.

◆ has_feature() [1/2]

bool Halide::Target::has_feature ( Feature f ) const

Referenced by Halide::SimdOpCheckTest::can_run_code(), has_feature(), has_large_buffers(), and Halide::Internal::schedule_scalar().

◆ has_feature() [2/2]

bool Halide::Target::has_feature ( halide_target_feature_t f ) const

inline

Definition at line 213 of file Target.h.

References has_feature().

◆ features_any_of()

bool Halide::Target::features_any_of ( const std::vector< Feature > & test_features ) const

◆ features_all_of()

bool Halide::Target::features_all_of ( const std::vector< Feature > & test_features ) const

◆ with_feature()

Target Halide::Target::with_feature ( Feature f ) const

Return a copy of the target with the given feature set.

This is convenient when enabling certain features (e.g. NoBoundsQuery) in an initialization list, where the target to be mutated may be a const reference.

Examples: tutorial/lesson_12_using_the_gpu.cpp.

Referenced by Halide::SimdOpCheckTest::SimdOpCheckTest().

◆ without_feature()

Target Halide::Target::without_feature ( Feature f ) const

Return a copy of the target with the given feature cleared.

This is convenient when disabling certain features (e.g. NoBoundsQuery) in an initialization list, where the target to be mutated may be a const reference.

Referenced by Halide::SimdOpCheckTest::get_run_target().

◆ has_gpu_feature()

bool Halide::Target::has_gpu_feature ( ) const

Is a fully feature GPU compute runtime enabled? I.e.

is Func::gpu_tile and similar going to work? Currently includes CUDA, OpenCL, Metal and D3D12Compute. We do not include OpenGL, because it is not capable of gpgpu, and is not scheduled via Func::gpu_tile. TODO: Should OpenGLCompute be included here?

Examples: tutorial/lesson_12_using_the_gpu.cpp.

Referenced by Halide::Internal::Autoscheduler::LoopNest::is_gpu_block(), Halide::Internal::Autoscheduler::LoopNest::is_gpu_serial(), Halide::Internal::Autoscheduler::LoopNest::is_gpu_thread(), and Halide::Internal::schedule_scalar().

◆ supports_type() [1/2]

bool Halide::Target::supports_type ( const Type & t ) const

Does this target allow using a certain type.

Generally all types except 64-bit float and int/uint should be supported by all backends.

It is likely better to call the version below which takes a DeviceAPI.

◆ supports_type() [2/2]

bool Halide::Target::supports_type	(	const Type &	t,
		DeviceAPI	device
	)		const

Does this target allow using a certain type on a certain device.

This is the prefered version of this routine.

◆ supports_device_api()

bool Halide::Target::supports_device_api ( DeviceAPI api ) const

Returns whether a particular device API can be used with this Target.

◆ get_required_device_api()

DeviceAPI Halide::Target::get_required_device_api ( ) const

If this Target (including all Features) requires a specific DeviceAPI, return it.

If it doesn't, return DeviceAPI::None. If the Target has features with multiple (different) DeviceAPI requirements, the result will be an arbitrary DeviceAPI.

◆ operator==()

bool Halide::Target::operator== ( const Target & other ) const

inline

Definition at line 264 of file Target.h.

References arch, bits, os, and processor_tune.

◆ operator!=()

bool Halide::Target::operator!= ( const Target & other ) const

inline

Definition at line 272 of file Target.h.

◆ get_runtime_compatible_target()

bool Halide::Target::get_runtime_compatible_target	(	const Target &	other,
		Target &	result
	)

Create a "greatest common denominator" runtime target that is compatible with both this target and other.

Used by generators to conveniently select a suitable runtime when linking together multiple functions.

Parameters

	other	The other target from which we compute the gcd target.
[out]	result	The gcd target if we return true, otherwise unmodified. Can be the same as *this.

Returns: Whether it was possible to find a compatible target (true) or not.

◆ to_string()

std::string Halide::Target::to_string ( ) const

Convert the Target into a string form that can be reconstituted by merge_string(), which will always be of the form.

arch-bits-os-processor-feature1-feature2...featureN.

Note that is guaranteed that Target(t1.to_string()) == t1, but not that Target(s).to_string() == s (since there can be multiple strings that parse to the same Target)... unless t1 contains 'unknown' fields (in which case you'll get a string that can't be parsed, which is intentional).

Examples: tutorial/lesson_12_using_the_gpu.cpp.

Referenced by Halide::SimdOpCheckTest::compile_and_check(), and Halide::SimdOpCheckTest::test_all().

◆ natural_vector_size() [1/2]

int Halide::Target::natural_vector_size ( const Halide::Type & t ) const

Given a data type, return an estimate of the "natural" vector size for that data type when compiling for this Target.

Referenced by Halide::Internal::GeneratorBase::natural_vector_size().

◆ natural_vector_size() [2/2]

template<typename data_t >

int Halide::Target::natural_vector_size ( ) const

inline

Given a data type, return an estimate of the "natural" vector size for that data type when compiling for this Target.

Definition at line 307 of file Target.h.

◆ has_large_buffers()

bool Halide::Target::has_large_buffers ( ) const

inline

Return true iff 64 bits and has_feature(LargeBuffers).

Definition at line 312 of file Target.h.

References bits, has_feature(), and LargeBuffers.

Referenced by maximum_buffer_size().

◆ maximum_buffer_size()

int64_t Halide::Target::maximum_buffer_size ( ) const

inline

Return the maximum buffer size in bytes supported on this Target.

This is 2^31 - 1 except on 64-bit targets when the LargeBuffers feature is enabled, which expands the maximum to 2^63 - 1.

Definition at line 319 of file Target.h.

References has_large_buffers().

◆ get_cuda_capability_lower_bound()

int Halide::Target::get_cuda_capability_lower_bound ( ) const

Get the minimum cuda capability found as an integer.

Returns 20 (our minimum supported cuda compute capability) if no cuda features are set.

◆ get_vulkan_capability_lower_bound()

int Halide::Target::get_vulkan_capability_lower_bound ( ) const

Get the minimum Vulkan capability found as an integer.

Returns 10 (our minimum supported Vulkan compute capability) if no Vulkan features are set.

◆ supported()

bool Halide::Target::supported ( ) const

Was libHalide compiled with support for this target?

◆ get_features_bitset()

const std::bitset<FeatureEnd>& Halide::Target::get_features_bitset ( ) const

inline

Return a bitset of the Featuress set in this Target (set = 1).

Note that while this happens to be the current internal representation, that might not always be the case.

Definition at line 343 of file Target.h.

◆ feature_to_name()

static std::string Halide::Target::feature_to_name ( Target::Feature feature )

static

Return the name corresponding to a given Feature, in the form used to construct Target strings (e.g., Feature::Debug is "debug" and not "Debug").

◆ feature_from_name()

static Target::Feature Halide::Target::feature_from_name ( const std::string & name )

static

Return the feature corresponding to a given name, in the form used to construct Target strings (e.g., Feature::Debug is "debug" and not "Debug").

If the string is not a known feature name, return FeatureEnd.

Member Data Documentation

◆ os

enum Halide::Target::OS Halide::Target::os = OSUnknown

Examples: tutorial/lesson_11_cross_compilation.cpp, and tutorial/lesson_12_using_the_gpu.cpp.

Referenced by Halide::SimdOpCheckTest::can_run_code(), and operator==().

◆ arch

enum Halide::Target::Arch Halide::Target::arch = ArchUnknown

Examples: tutorial/lesson_11_cross_compilation.cpp.

Referenced by Halide::SimdOpCheckTest::can_run_code(), and operator==().

◆ bits

int Halide::Target::bits = 0

The bit-width of the target machine.

Must be 0 for unknown, or 32 or 64.

Examples: tutorial/lesson_11_cross_compilation.cpp.

Definition at line 50 of file Target.h.

Referenced by Halide::SimdOpCheckTest::can_run_code(), has_large_buffers(), and operator==().

◆ vector_bits

int Halide::Target::vector_bits = 0

The bit-width of a vector register for targets where this is configurable and targeting a fixed size is desired.

The default of 0 indicates no assumption of fixed size is allowed.

Definition at line 55 of file Target.h.

◆ processor_tune

enum Halide::Target::Processor Halide::Target::processor_tune = ProcessorGeneric

Referenced by operator==().

The documentation for this struct was generated from the following file:

src/Target.h

Public Types

Public Member Functions

Static Public Member Functions

Public Attributes

Detailed Description

Member Enumeration Documentation

◆ OS

◆ Arch

◆ Processor

◆ Feature

Constructor & Destructor Documentation

◆ Target() [1/5]

◆ Target() [2/5]

◆ Target() [3/5]

◆ Target() [4/5]

◆ Target() [5/5]

Member Function Documentation

◆ validate_target_string()

◆ has_unknowns()

◆ set_feature()

◆ set_features()

◆ has_feature() [1/2]

◆ has_feature() [2/2]

◆ features_any_of()

◆ features_all_of()

◆ with_feature()

◆ without_feature()

◆ has_gpu_feature()

◆ supports_type() [1/2]

◆ supports_type() [2/2]

◆ supports_device_api()

◆ get_required_device_api()

◆ operator==()

◆ operator!=()

◆ get_runtime_compatible_target()

◆ to_string()

◆ natural_vector_size() [1/2]

◆ natural_vector_size() [2/2]

◆ has_large_buffers()

◆ maximum_buffer_size()

◆ get_cuda_capability_lower_bound()

◆ get_vulkan_capability_lower_bound()

◆ supported()

◆ get_features_bitset()

◆ feature_to_name()

◆ feature_from_name()

Member Data Documentation

◆ os

◆ arch

◆ bits

◆ vector_bits

◆ processor_tune