halide/Halide v15.0.0

Halide v15.0.0

on GitHub

What's Changed

General Notes

• Support for RISC V Vector architectures.

• Python-related:

  • Halide builds for Python are now being built and provided to PyPI, so it is now possible to use the Halide Python bindings simply by pip install halide
  • Major improvements were made to the Python bindings, with many missing or incomplete sections of the API added or filled in.
  • We now support the use of Generators from Python (for both JIT and AOT usage).
  • The standard CMake rules now support generating a Python extension directly.
  • Support for Python was removed from Halide's Makefiles; you must use CMake to build the Python bindings

• Halide::Func now allows you to (optionally) constrain the type(s) of Exprs that the Func can contain, and/or the dimensionality of the Func.

• Added a new way to use the JIT (compile_to_callable) that allows calling a jitted function with the same syntax as for AOT-compiled functions, allowing more control over JIT lifespan, as well as thread-safe arguments without requiring ParamMap

• General improvements to SIMD codegen

• Several rarely-used parts of the C++ Generator API were deprecated, and the way that autoschedulers are specified for AOT compilation is now completely different (but better for future expandability).

• CMake builds now require >= v3.22

• WABT usage requires >= v1.0.30

• LLVM 12 is no longer supported

Notable bug fixes

• Make Halide::round behave as documented (#7012)
• Incorrect folding of saturating_sub (#6883)
• The check for race conditions didn't consider where clauses (#6808)
• Performance regression for x86 for certain LLVM versions (#6783)
• Fusing a specialization drops compute_withs from generated code (#6770)
• Incorrect output when realize condition depends on tuple call (#6915)
• Python extensions should default to throwing exceptions rather than calling abort() for errors (#6986)
• Python bindings didn't support bool buffers (#7006)
• Python bindings didn't support float16 buffers (#7060)
• Python extensions that executed on GPU didn't copy back to host properly (#6869)
• Fix bugs in div_round_to_zero and fast_integer_divide_round_to_zero (#7008)
• Bugs in add_requirement() (#7045)

Major changes

• Augment Halide::Func to allow for constraining Type and Dimensionality by @steven-johnson in #6734 and #6735
• Add Target support for architectures with implementation specific vector size. by @zvookin in #6786
• Add support for vscale vector code generation. by @zvookin in #6802
• Remove Python bindings from Makefiles by @alexreinking in #6821
• Add a new, alternate JIT-call convention by @steven-johnson in #6777
• Pip packaging by @alexreinking in #6886 and #6938
• Define a Generator framework in Python by @steven-johnson in #6764
• Make Halide::round behave as documented by @abadams in #7012

Minor changes

• -mtune=/-mcpu= support for x86 AMD CPU's by @LebedevRI in #6655
• Enable deprecations warnings by @steven-johnson in #6555
• Fix GPU depredication/scalarization by @shoaibkamil in #6669
• Allow PyPipeline and PyFunc to realize() scalar buffers by @steven-johnson in #6674
• Future-proof 'processortotune processor` by @LebedevRI in #6673
• Fix ctors for Realization by @steven-johnson in #6675
• -mtune=native CPU autodetection for AMD Zen 3 CPU by @LebedevRI in #6648
• Clean up Python extensions in python_bindings by @steven-johnson in #6670
• Halide::Tools::save_image() should accept buffers with const types by @steven-johnson in #6679
• Fix "set but not used" warnings/errors by @steven-johnson in #6683
• Drop support for LLVM12 by @steven-johnson in #6686
• Upgrade to clang-format 13 by @steven-johnson in #6689
• Always mark _ucon as 'unused' in Codegen_C by @steven-johnson in #6691
• Add break to avoid 'possible unintentional fallthru' warning by @steven-johnson in #6694
• Silence "unknown warning" in Clang 13 by @steven-johnson in #6693
• Fixes for top-of-tree LLVM by @steven-johnson in #6697
• Python: make Func implicitly convertible to Stage (#6702) by @steven-johnson in #6704
• llvm no longer wants a type suffix on vst intrinsics by @abadams in #6701
• Fix type-mangling for vst on arm32 for LLVM15 by @steven-johnson in #6705
• Remove the last remaining call to getPointerElementType() by @steven-johnson in #6715
• ARM vst mangling needs to be conditional on opaque ptrs by @steven-johnson in #6716
• Combine string constants in combine_strings() by @steven-johnson in #6717
• Update CodeGen_PTX_Dev to use new PassManager by @steven-johnson in #6718
• Closure functions for parallel tasks should be internal, not external by @steven-johnson in #6720
• Smarten type_of<> for fn ptrs; fix async_parallel for C backend by @steven-johnson in #6719
• Remove legacy::FunctionPassManager usage in Codegen_PTX_Dev by @steven-johnson in #6722
• get_amd_processor(): implement detection for the rest of supported AMD CPU's by @LebedevRI in #6711
• Add Func::output_type() method by @steven-johnson in #6724
• Grab-bag of minor Python fixes by @steven-johnson in #6725
• Remove rounding_halving_sub and non-existent arm rhsub instructions by @rootjalex in #6723
• Faster widening_mul(int16x, int16x) -> int32x for x86 (AVX2 and SSE2) by @rootjalex in #6677
• Add missing #include in ThreadPool.h by @steven-johnson in #6738
• Fix regression from #6734 by @steven-johnson in #6739
• Add forwarding for the recently-added Func::output_type() method by @steven-johnson in #6741
• Silence "unscheduled update stage" warnings in msan_generator.cpp by @steven-johnson in #6740
• Add pycache to toplevel .gitignore file by @steven-johnson in #6743
• Silence "may be used uninitialized" in Buffer::for_each_element() by @steven-johnson in #6747
• Update WABT to 1.0.29 by @steven-johnson in #6748
• Update hannk README link to hosted models page by @steven-johnson in #6749
• Add a HalideError base class to Python bindings by @steven-johnson in #6750
• Add GeneratorFactoryProvider to generate_filter_main() by @steven-johnson in #6755
• Minor metadata-related cleanups by @steven-johnson in #6759
• Expand the x86 SIMD variants tested in correctness_vector_reductions by @steven-johnson in #6762
• Fix Param::set_estimate for T=void by @steven-johnson in #6766
• add_python_aot_extension should use FUNCTION_NAME for the .so output … by @steven-johnson in #6767
• Fix fundamental confusion about target/tune CPU by @LebedevRI in #6765
• Fix annoying typo in Func.h by @steven-johnson in #6774
• Add execute_generator() API by @steven-johnson in #6771
• Allow overriding of Generator::init_from_context() for debug purposes by @steven-johnson in #6760
• Convert some assert-only usage of output_types() -> types() by @steven-johnson in #6779
• [miscompile] Don't de-negate and change direction of shifts-by-unsigned by @LebedevRI in #6782
• Move some options from execute_generator back to generate_filter_main by @steven-johnson in #6787
• LLVM codegen: register AA pipeline if LLVM is older than 14 by @LebedevRI in #6785
• Update the list of fused_pairs and run validate_fused_group for specalization definitions too by @vksnk in #6770
• halide_type_of<>() should always be constexpr by @steven-johnson in #6790
• Define an AbstractGenerator interface by @steven-johnson in #6637
• hexagon_scatter test should run only if target has HVX by @steven-johnson in #6793
• slow tests should support sharding by @steven-johnson in #6780
• Add missing include to test_sharding.h by @steven-johnson in #6795
• Pacify clang-tidy by @steven-johnson in #6796
• Silence a "possibly uninitialized" warning by @steven-johnson in #6797
• Make all tests default to -fvisibility=hidden by @steven-johnson in #6799
• Minor typedef cleanup by @steven-johnson in #6800
• Fix auto_schedule/machine_params parsing by @steven-johnson in #6804
• Rewrite strided loads of 4 in AlignLoads by @vksnk in #6806
• Fix two minor bugs triggered by an or reduction with early-out by @abadams in #6807
• [CMake] Mark multi-threaded tests as such by @LebedevRI in #6810
• Rework .gitignore by @alexreinking in #6822
• Update presets to format version 3 by @alexreinking in #6824
• Fix for top-of-tree LLVM by @steven-johnson in #6825
• Tweak python apps for better Blaze/Bazel compatibility by @steven-johnson in #6823
• Apply CMAKE_C_COMPILER_LAUNCHER to initmod clang calls by @alexreinking in #6831
• Scrub Python from Makefile after buildbot update by @alexreinking in #6833
• Remove unused function in callable_generator.cpp by @steven-johnson in #6834
• Disable testing for apps/linear_algebra on x86-32-linux/Make by @steven-johnson in #6836
• Rearrange subdirectories in python_bindings by @steven-johnson in #6835
• Better lowering of halving_sub and rounding_halving_add by @abadams in #6827
• Check RDom::where predicates for race conditions by @alexreinking in #6842
• Remove Generator::value_tracker and friends by @steven-johnson in #6845
• Add placeholder code for bfloat16 in Python (#6849) by @steven-johnson in #6850
• Fix the PLUGINS argument to properly join multiple arguments by @steven-johnson in #6851
• Add autoscheduling to the generator_aot_stubuser test by @steven-johnson in #6855
• Silence Adams2019 Autoscheduler by @steven-johnson in #6854
• [vulkan phase0] Add adts for containers and memory allocation to runtime by @derek-gerstmann in #6829
• Promote Reinterpret Intrinsic into an Reinterpret IR Node by @LebedevRI in #6853
• Python source reorg by @alexreinking in #6867
• Fix simd_op_check for top-of-tree LLVM by @steven-johnson in #6874
• Use pmaddubsw 8-bit horizontal widening adds (Fixes #6859) by @rootjalex in #6873
• [Codegen_LLVM] Radically simplify visit(const Reinterpret *op) by @LebedevRI in #6865
• [Codegen] Fail to codegen Call::undef, just like Call::signed_integer_overflow by @LebedevRI in #6871
• Fix error in Makefile for Adams2019 on OSX by @steven-johnson in #6877
• Refactor/cleanup in Autoscheduler code by @steven-johnson in #6858
• Ensure $CMAKE_{lang}_OUTPUT_EXTENSION is set before using it by @shoaibkamil in #6879
• #6863 - Fixes to make address sanitizer happy for internal runtime classes by @derek-gerstmann in #6880
• [Codegen_LLVM] Define all the things by @LebedevRI in #6866
• Add set-host-dirty/copy-to-host to PythonExtensionGen by @steven-johnson in #6869
• Rewrite PythonExtensionGen to be C++ based by @steven-johnson in #6888
• Fixes to allow compiling with LLVM16 by @steven-johnson in #6889
• Add support for generating x86 sum-of-absolute-difference reductions by @abadams in #6872
• Remove (most) of the env var usage from Adams2019 by @steven-johnson in #6861
• [vulkan phase1] Add SPIR-V IR by @derek-gerstmann in #6882
• Add auto_schedule label to Adams2019 and Li2018 tests in CMake by @steven-johnson in #6898
• [Simplify] Drop no-op single-input identity shuffles by @LebedevRI in #6901
• [Codegen_LLVM] Annotate LLVM IR functions with nounwind/mustprogress attributes by @LebedevRI in #6897
• Don't try to fold saturating_sub of VectorReduce by @rootjalex in #6896
• Upgrade clang-format and clang-tidy to v14 (v2) by @steven-johnson in #6902
• Allow AMX instructions with K dimension larger than 4 bytes by @frengels in #6582
• Fix autoscheduling trivial lut wrappers by @abadams in #6905
• Fix broken Makefile rules for autoschedulers on OSX by @steven-johnson in #6906
• LICENSE.txt: Include full text of Apache 2.0 license (not just the 'header' version) by @steven-johnson in #6912
• LICENSE.txt: add spirv license by @steven-johnson in #6913
• LICENSE.txt: add BLAS license. by @steven-johnson in #6914
• Upgrade CMake minimum version to 3.22 by @steven-johnson in #6916
• Remove unused GHA and packaging workflows. by @alexreinking in #6917
• Fix two warnings found with clang 16 by @steven-johnson in #6918
• Fix bug when realize condition depends on tuple call by @abadams in #6915
• Fix wrong install path for *.py files by @steven-johnson in #6921
• Make use of CMake 3.22 features by @alexreinking in #6919
• Make saturating_cast an intrinsic by @rootjalex in #6900
• Halide::Error should not extend std::runtime_error by @steven-johnson in #6927
• Rework internal PYTHONPATH maintenance by @steven-johnson in #6922
• Tutorial 10 needs to be skipped for Python when targeting Wasm (just as non-Python does) by @steven-johnson in #6932
• Add build & test presets for release and debug CMake builds by @steven-johnson in #6934
• Add ASAN support to CMake via toolchain file by @steven-johnson in #6920
• Fix badly-merged CMakePresets.json file by @steven-johnson in #6936
• Add minimal useful implementation of extracting and concatenating bits by @abadams in #6928
• Export HalidePythonExtensionHelpers.cmake for installs by @steven-johnson in #6941
• Add/update Python Readme by @steven-johnson in #6939
• Don't throw an exception from generate_filter_main by @steven-johnson in #6946
• Handle saturating_cast in compute_expr_cost() by @rootjalex in #6947
• Two quick build fixes by @alexreinking in #6950
• Remove add_python_aot_extension() rule in CMake by @steven-johnson in #6949
• Build fixes for manylinux2014 by @alexreinking in #6953
• Remove add_python_stub_extension(), adding the functionality to add_halide_generator() instead by @steven-johnson in #6952
• [HVX] Fix state_var issue by @rootjalex in #6894
• Fix RPATH for Python wheels on macOS by @alexreinking in #6958
• Python: don't crash for repr(Expr()) by @steven-johnson in #6962
• Some minor top-level CMakeLists.txt reorganization by @alexreinking in #6957
• CMake packaging fixes by @alexreinking in #6966
• Use CMake target to handle vendored SPIRV headers by @alexreinking in #6968
• Don't cache Halide_ASAN_ENABLED by @alexreinking in #6969
• Lower saturating_cast in bounds inference by @rootjalex in #6970
• Small refactor to remove confusion between CodeGen_LLVM and CodeGen_Internal. by @zvookin in #6973
• Fix XCode by wrapping weights in an OBJECT library by @alexreinking in #6977
• Add test for _Halide_target_export_single_symbol by @steven-johnson in #6983
• Fix markdown links by @alexreinking in #6988
• Improve error-handling in Python Extensions by @steven-johnson in #6986
• Refactor buffer-unpacking code in PythonExtensionGen by @steven-johnson in #6991
• Fixes for Xcode "new" build system. by @alexreinking in #6993
• Fix compiler warnings in Elf.cpp by @steven-johnson in #6992
• [Codegen] Adapt ModuleAddressSanitizerPass/ModuleSanitizerCoveragePass renaming by @MaskRay in #6996
• Apply _Halide_place_dll() to _Halide_gengen (#6999) by @steven-johnson in #7000
• Log target info in performance_fast_pow (#6997) by @steven-johnson in #6998
• Clean up Adams2019 CMake file by @steven-johnson in #7003
• Prohibit C99 VLA usage in runtime code by @steven-johnson in #7005
• Couple small fixes to update RISC V to current LLVM flags and enable vscale use. by @zvookin in #6995
• Fix Python handling of boolean buffers by @steven-johnson in #7006
• Fix some bugs in div_round_to_zero by @abadams in #7008
• [HVX] Simplify constant factor before distributing by @rootjalex in #7009
• Add one-sided widening intrinsics. by @rootjalex in #6967
• Rework Python Extension C++ code (again) by @steven-johnson in #7010
• Add minimum GitHub token permissions for workflow by @varunsh-coder in #7011
• Revert "[HVX] Simplify constant factor before distributing" by @steven-johnson in #7013
• Fix SpecificExpr canonicalization by @rootjalex in #7016
• Appease Python linter by @steven-johnson in #7022
• Don't use -g for EMCC by @steven-johnson in #7025
• Temporarily disable testing for apps/fft (#7033) by @steven-johnson in #7035
• Add reinterpret simplifications by @rootjalex in #7029
• Codegen_C for user_context by @steven-johnson in #7031
• Fix Wasm BulkMemory Codgen + Minor fixes to apps/HelloWasm by @steven-johnson in #7026
• Add stack-size-canary test to apps/fft's CMake file by @steven-johnson in #7034
• Handle widen_right_* intrinsics in bounds inference by @vksnk in #7039
• Revert "Temporarily disable testing for apps/fft (#7033)" by @steven-johnson in #7040
• Fix PyExt error handling by @steven-johnson in #7042
• add_requirement() maintenance by @steven-johnson in #7045
• Fix false positive use after free warning. by @zvookin in #7046
• Allow call_intrin to call an LLVM intrinsic with void return type. by @zvookin in #7048
• Allow CodeGen_LLVM::codegen_buffer_pointer to support vectors. by @zvookin in #7049
• Don't mutate GeneratorParams in PythonGenerators by @steven-johnson in #7052
• Allow redefinition of Generators when in interactive mode by @steven-johnson in #7053
• Upgrade wabt to 1.0.30 by @steven-johnson in #7058
• Add support for float16 buffer in python extension by @stevesuzuki-arm in #7060
• Add a terminate_handler to try to report unhandled exceptions by @steven-johnson in #7038
• Improve MSAN under JIT by @steven-johnson in #7059
• Autoscheduler test reorg, part 1 by @steven-johnson in #7064
• Autoscheduler test reorg, part 2 by @steven-johnson in #7065
• Autoscheduler test reorg, part 3 by @steven-johnson in #7067
• pacify clang-tidy by removing unused "using" by @steven-johnson in #7071
• Add pip packaging workflow to GHA by @alexreinking in #6938
• [HVX] Fix DistributeShiftsAsMuls by @rootjalex in #7083
• Support added for dot() instructions in Metal backend

Changes to public API since last release

• Add add_halide_python_extension_library() rule by @steven-johnson in #6979
• Add add_halide_runtime rule by @steven-johnson in #6985
• Remove deprecated Halide::Output type by @steven-johnson in #6685
• Remove deprecated build() support from Generators by @steven-johnson in #6684
• Remove deprecated versions of Func::prefetch() by @steven-johnson in #6698
• Remove deprecated JIT handler setters by @steven-johnson in #6699
• Drop support for Matlab extensions by @steven-johnson in #6696
• Revise PyStub calling convention for GeneratorParams by @steven-johnson in #6742
• Change stub module names in Python to be _pystub rather than _stub by @steven-johnson in #6830

New Deprecations (Upcoming API changes)

• Deprecate variadic-template version of Realization ctor by @steven-johnson in #6695
• Deprecate GeneratorContext getters with get_ prefix by @steven-johnson in #6753
• Deprecate disable_llvm_loop_opt (#4113) by @steven-johnson in #6754
• Add Func::type()/types(), deprecate Func::output_type()/output_types() by @steven-johnson in #6772
• Deprecate/remove Generator::get_externs_map() and friends by @steven-johnson in #6844
• Rework autoscheduler API (#6788) by @steven-johnson in #6838

Other Notes

• Although there are commits relating to a Vulkan backend, this release of Halide doesn't provide Vulkan support (it's still a work in progress)
• It's possible that the changes in #6754 can cause performance degradation (but usually only for poorly-schedule Halide code).

New Contributors

• @frengels made their first contribution in #6582
• @varunsh-coder made their first contribution in #7011