188 for (
const auto &f : initial_features) {
194 Target(
OS o,
Arch a,
int b,
const std::vector<Feature> &initial_features = std::vector<Feature>())
222 void set_features(
const std::vector<Feature> &features_to_set,
bool value =
true);
275 return os == other.
os &&
279 features == other.features;
283 return !(*
this == other);
316 template<
typename data_t>
372 std::bitset<FeatureEnd> features;
377 void validate_features()
const;
This file declares the routines used by Halide internally in its runtime.
halide_target_feature_t
Optional features a compilation Target can have.
@ halide_target_feature_large_buffers
Enable 64-bit buffer indexing to support buffers > 2GB. Ignored if bits != 64.
@ halide_target_feature_fma
Enable x86 FMA instruction.
@ halide_target_feature_wasm_bulk_memory
Enable +bulk-memory instructions for WebAssembly codegen.
@ halide_target_feature_tsan
Enable hooks for TSAN support.
@ halide_target_feature_msan
Enable hooks for MSAN support.
@ halide_target_feature_avx512_zen4
Enable the AVX512 features supported by Zen4 processors. This include all of the Cannonlake features,...
@ halide_target_feature_wasm_threads
Enable use of threads in WebAssembly codegen. Requires the use of a wasm runtime that provides pthrea...
@ halide_target_feature_trace_loads
Trace all loads done by the pipeline. Equivalent to calling Func::trace_loads on every non-inlined Fu...
@ halide_target_feature_enable_llvm_loop_opt
Enable loop vectorization + unrolling in LLVM. Overrides halide_target_feature_disable_llvm_loop_opt....
@ halide_target_feature_no_asserts
Disable all runtime checks, for slightly tighter code.
@ halide_target_feature_cl_doubles
Enable double support on OpenCL targets.
@ halide_target_feature_rvv
Enable RISCV "V" Vector Extension.
@ halide_target_feature_avx2
Use AVX 2 instructions. Only relevant on x86.
@ halide_target_feature_trace_realizations
Trace all realizations done by the pipeline. Equivalent to calling Func::trace_realizations on every ...
@ halide_target_feature_c_plus_plus_mangling
Generate C++ mangled names for result function, et al.
@ halide_target_feature_vulkan_float16
Enable Vulkan 16-bit float support.
@ halide_target_feature_no_runtime
Do not include a copy of the Halide runtime in any generated object file or assembly.
@ halide_target_feature_hvx_v65
Enable Hexagon v65 architecture.
@ halide_target_feature_debug
Turn on debug info and output for runtime code.
@ halide_target_feature_embed_bitcode
Emulate clang -fembed-bitcode flag.
@ halide_target_feature_armv86a
Enable ARMv8.6a instructions.
@ halide_target_feature_wasm_simd128
Enable +simd128 instructions for WebAssembly codegen.
@ halide_target_feature_vulkan
Enable Vulkan runtime support.
@ halide_target_feature_end
A sentinel. Every target is considered to have this feature, and setting this feature does nothing.
@ halide_llvm_large_code_model
Use the LLVM large code model to compile.
@ halide_target_feature_profile_by_timer
Alternative to halide_target_feature_profile using timer interrupt for systems without threads or app...
@ halide_target_feature_semihosting
Used together with Target::NoOS for the baremetal target built with semihosting library and run with ...
@ halide_target_feature_soft_float_abi
Enable soft float ABI. This only enables the soft float ABI calling convention, which does not necess...
@ halide_target_feature_sve2
Enable ARM Scalable Vector Extensions v2.
@ halide_target_feature_d3d12compute
Enable Direct3D 12 Compute runtime.
@ halide_target_feature_cuda_capability86
Enable CUDA compute capability 8.6 (Ampere)
@ halide_target_feature_armv89a
Enable ARMv8.9a instructions.
@ halide_target_feature_avx512_skylake
Enable the AVX512 features supported by Skylake Xeon server processors. This adds AVX512-VL,...
@ halide_target_feature_avx512_cannonlake
Enable the AVX512 features expected to be supported by future Cannonlake processors....
@ halide_target_feature_metal
Enable the (Apple) Metal runtime.
@ halide_target_feature_hvx_128
Enable HVX 128 byte mode.
@ halide_target_feature_cuda_capability70
Enable CUDA compute capability 7.0 (Volta)
@ halide_target_feature_fma4
Enable x86 (AMD) FMA4 instruction set.
@ halide_target_feature_cuda_capability30
Enable CUDA compute capability 3.0 (Kepler)
@ halide_target_feature_no_neon
Avoid using NEON instructions. Only relevant for 32-bit ARM.
@ halide_target_feature_cuda_capability61
Enable CUDA compute capability 6.1 (Pascal)
@ halide_target_feature_armv7s
Generate code for ARMv7s. Only relevant for 32-bit ARM.
@ halide_target_feature_spirv
Enable SPIR-V code generation support.
@ halide_target_feature_trace_pipeline
Trace the pipeline.
@ halide_target_feature_armv88a
Enable ARMv8.8a instructions.
@ halide_target_feature_cl_atomic64
Enable 64-bit atomics operations on OpenCL targets.
@ halide_target_feature_egl
Force use of EGL support.
@ halide_target_feature_hvx_v68
Enable Hexagon v68 architecture.
@ halide_target_feature_avx10_1
Intel AVX10 version 1 support. vector_bits is used to indicate width.
@ halide_target_feature_profile
Launch a sampling profiler alongside the Halide pipeline that monitors and reports the runtime used b...
@ halide_target_feature_strict_float
Turn off all non-IEEE floating-point optimization. Currently applies only to LLVM targets.
@ halide_target_feature_cuda_capability35
Enable CUDA compute capability 3.5 (Kepler)
@ halide_target_feature_armv8a
Enable ARMv8a instructions.
@ halide_target_feature_asan
Enable hooks for ASAN support.
@ halide_target_feature_armv87a
Enable ARMv8.7a instructions.
@ halide_target_feature_cl_half
Enable half support on OpenCL targets.
@ halide_target_feature_vulkan_float64
Enable Vulkan 64-bit float support.
@ halide_target_feature_arm_dot_prod
Enable ARMv8.2-a dotprod extension (i.e. udot and sdot instructions)
@ halide_target_feature_avx512_sapphirerapids
Enable the AVX512 features supported by Sapphire Rapids processors. This include all of the Zen4 feat...
@ halide_target_feature_vulkan_version13
Enable Vulkan v1.3 runtime target support.
@ halide_target_feature_vulkan_version12
Enable Vulkan v1.2 runtime target support.
@ halide_target_feature_sse41
Use SSE 4.1 and earlier instructions. Only relevant on x86.
@ halide_target_feature_power_arch_2_07
Use POWER ISA 2.07 new instructions. Only relevant on POWERPC.
@ halide_target_feature_opencl
Enable the OpenCL runtime.
@ halide_target_feature_trace_stores
Trace all stores done by the pipeline. Equivalent to calling Func::trace_stores on every non-inlined ...
@ halide_target_feature_hexagon_dma
Enable Hexagon DMA buffers.
@ halide_target_feature_avx512
Enable the base AVX512 subset supported by all AVX512 architectures. The specific feature sets are AV...
@ halide_target_feature_avx512_knl
Enable the AVX512 features supported by Knight's Landing chips, such as the Xeon Phi x200....
@ halide_target_feature_cuda_capability50
Enable CUDA compute capability 5.0 (Maxwell)
@ halide_target_feature_arm_fp16
Enable ARMv8.2-a half-precision floating point data processing.
@ halide_target_feature_armv82a
Enable ARMv8.2a instructions.
@ halide_target_feature_hvx_v62
Enable Hexagon v62 architecture.
@ halide_target_feature_armv84a
Enable ARMv8.4a instructions.
@ halide_target_feature_cuda
Enable the CUDA runtime. Defaults to compute capability 2.0 (Fermi)
@ halide_target_feature_armv81a
Enable ARMv8.1a instructions.
@ halide_target_feature_webgpu
Enable the WebGPU runtime.
@ halide_target_feature_sanitizer_coverage
Enable hooks for SanitizerCoverage support.
@ halide_target_feature_cuda_capability80
Enable CUDA compute capability 8.0 (Ampere)
@ halide_target_feature_wasm_mvponly
Disable all extensions to WebAssembly codegen (including +sign-ext and +nontrapping-fptoint,...
@ halide_target_feature_f16c
Enable x86 16-bit float support.
@ halide_target_feature_vulkan_int16
Enable Vulkan 16-bit integer support.
@ halide_target_feature_cuda_capability32
Enable CUDA compute capability 3.2 (Tegra K1)
@ halide_target_feature_armv85a
Enable ARMv8.5a instructions.
@ halide_target_feature_jit
Generate code that will run immediately inside the calling process.
@ halide_target_feature_avx
Use AVX 1 instructions. Only relevant on x86.
@ halide_target_feature_cuda_capability75
Enable CUDA compute capability 7.5 (Turing)
@ halide_target_feature_check_unsafe_promises
Insert assertions for promises.
@ halide_target_feature_vsx
Use VSX instructions. Only relevant on POWERPC.
@ halide_target_feature_vulkan_int8
Enable Vulkan 8-bit integer support.
@ halide_target_feature_armv83a
Enable ARMv8.3a instructions.
@ halide_target_feature_vulkan_int64
Enable Vulkan 64-bit integer support.
@ halide_target_feature_user_context
Generated code takes a user_context pointer as first argument.
@ halide_target_feature_no_bounds_query
Disable the bounds querying functionality.
@ halide_target_feature_vulkan_version10
Enable Vulkan v1.0 runtime target support.
@ halide_target_feature_fuzz_float_stores
On every floating point store, set the last bit of the mantissa to zero. Pipelines for which the outp...
@ halide_target_feature_sve
Enable ARM Scalable Vector Extensions.
@ halide_target_feature_x86_apx
Intel x86 APX support. Covers initial set of features released as APX: egpr,push2pop2,...
@ halide_target_feature_hvx_v66
Enable Hexagon v66 architecture.
This file defines the class FunctionDAG, which is our representation of a Halide pipeline,...
Target get_host_target()
Return the target corresponding to the host machine.
@ Internal
Not visible externally, similar to 'static' linkage in C.
Type type_of()
Construct the halide equivalent of a C type.
Target::Feature target_feature_for_device_api(DeviceAPI api)
Get the Target feature corresponding to a DeviceAPI.
Target get_jit_target_from_environment()
Return the target that Halide will use for jit-compilation.
DeviceAPI
An enum describing a type of device API.
Target get_target_from_environment()
Return the target that Halide will use.
unsigned __INT64_TYPE__ uint64_t
signed __INT64_TYPE__ int64_t
A struct representing a target machine and os to generate code for.
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 ot...
bool operator==(const Target &other) const
bool supports_type(const Type &t, DeviceAPI device) const
Does this target allow using a certain type on a certain device.
bool supports_device_api(DeviceAPI api) const
Returns whether a particular device API can be used with this Target.
int get_arm_v8_lower_bound() const
Get the minimum ARM v8.x capability found as an integer.
bool has_gpu_feature() const
Is a fully feature GPU compute runtime enabled? I.e.
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....
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....
Target(OS o, Arch a, int b, const std::vector< Feature > &initial_features=std::vector< Feature >())
void set_features(const std::vector< Feature > &features_to_set, bool value=true)
bool has_large_buffers() const
Return true iff 64 bits and has_feature(LargeBuffers).
enum Halide::Target::Arch arch
int64_t maximum_buffer_size() const
Return the maximum buffer size in bytes supported on this Target.
bool has_feature(Feature f) const
int natural_vector_size() const
Given a data type, return an estimate of the "natural" vector size for that data type when compiling ...
Target(const std::string &s)
Given a string of the form used in HL_TARGET (e.g.
enum Halide::Target::Processor processor_tune
int bits
The bit-width of the target machine.
Processor
The specific processor to be targeted, tuned for.
@ BdVer3
Tune for AMD Piledriver CPU (AMD Family 15h (2nd-gen), launched 2012).
@ ZnVer4
Tune for AMD Zen 3 CPU (AMD Family 19h, launched 2020).
@ AMDFam10
Tune for later versions of AMD K8 CPU, with SSE3 support.
@ ZnVer3
Tune for AMD Zen 2 CPU (AMD Family 17h, launched 2019).
@ K8_SSE3
Tune for AMD K8 Hammer CPU (AMD Family 0Fh, launched 2003).
@ ZnVer1
Tune for AMD Jaguar CPU (AMD Family 16h, launched 2013).
@ BdVer1
Tune for AMD Bobcat CPU (AMD Family 14h, launched 2011).
@ BtVer1
Tune for AMD K10 "Barcelona" CPU (AMD Family 10h, launched 2007).
@ ZnVer2
Tune for AMD Zen CPU (AMD Family 17h, launched 2017).
@ BdVer4
Tune for AMD Steamroller CPU (AMD Family 15h (3nd-gen), launched 2014).
@ BdVer2
Tune for AMD Bulldozer CPU (AMD Family 15h, launched 2011).
@ BtVer2
Tune for AMD Excavator CPU (AMD Family 15h (4th-gen), launched 2015).
@ ProcessorGeneric
Do not tune for any specific CPU. In practice, this means that halide will decide the tune CPU based ...
Target(OS o, Arch a, int b, Processor pt, const std::vector< Feature > &initial_features=std::vector< Feature >(), int vb=0)
enum Halide::Target::OS os
std::string to_string() const
Convert the Target into a string form that can be reconstituted by merge_string(),...
static bool validate_target_string(const std::string &s)
Check if a target string is valid.
Target without_feature(Feature f) const
Return a copy of the target with the given feature cleared.
const std::bitset< FeatureEnd > & get_features_bitset() const
Return a bitset of the Featuress set in this Target (set = 1).
Feature
Optional features a target can have.
bool operator!=(const Target &other) const
bool supported() const
Was libHalide compiled with support for this target?
DeviceAPI get_required_device_api() const
If this Target (including all Features) requires a specific DeviceAPI, return it.
bool has_feature(halide_target_feature_t f) const
bool has_unknowns() const
Return true if any of the arch/bits/os fields are "unknown"/0; return false otherwise.
bool features_any_of(const std::vector< Feature > &test_features) const
void set_feature(Feature f, bool value=true)
bool supports_type(const Type &t) const
Does this target allow using a certain type.
int get_vulkan_capability_lower_bound() const
Get the minimum Vulkan capability found as an integer.
int get_cuda_capability_lower_bound() const
Get the minimum cuda capability found as an integer.
Target with_feature(Feature f) const
Return a copy of the target with the given feature set.
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 ...
bool features_all_of(const std::vector< Feature > &test_features) const
Arch
The architecture used by the target.
OS
The operating system used by the target.
int vector_bits
The bit-width of a vector register for targets where this is configurable and targeting a fixed size ...
Types in the halide type system.