halide 21.0.0

[BrewTestBot]

Created by brew bump

---

Created with brew bump-formula-pr.

Details

release notes

## Release highlights
We have deliberately skipped version 20.0.0 to align with the LLVM version we are now using. Note that LLVM 21.1.1 or higher is required as LLVM 21.1.0 has a major bug in the NVPTX backend.
Major changes

The rfactor scheduling directive was rewritten and enhanced. It is now compatible with autoschedulers.
The Mullapudi2016 autoscheduler now supports experimental GPU scheduling.
The Python bindings have been substantially improved, with many missing bindings filled in.
HL_DEBUG_CODEGEN gained a new filtering mode. Debug levels can now be set on a per-file/per-function basis.
Support was added for AMD Zen5 and the iOS Simulator.
The strict_float feature has been reimplemented and should be much more reliable.
Lots of bugfixes, performance improvements, and build system improvements. We spent a lot of time fixing issues with our testing infrastructure and are looking forward to implementing a more stable contribution experience going forward.

Deprecations

LLVM 19 and below are no longer supported, in keeping with our support policy.
Halide_BUNDLE_STATIC will be removed in the next release. If you are using it, please migrate to the shared library instead.
Support for Python 3.8 has been dropped.

Changelog
Scheduling

The rfactor scheduling directive was rewritten and enhanced.

Rewrite the rfactor scheduling directive by @alexreinking in Rewrite the rfactor scheduling directive halide/Halide#8490
Dequalify names when constructing RVars in rfactor by @alexreinking in Dequalify names when constructing RVars in rfactor halide/Halide#8560
Add promise_clamped in rfactor by @alexreinking in Add promise_clamped in rfactor halide/Halide#8608
Add rfactor patterns for NaN-propagating min/max by @alexreinking in Add rfactor patterns for NaN-propagating min/max halide/Halide#8587


The Mullapudi2016 autoscheduler now supports experimental GPU scheduling.

GPU autoscheduling with Mullapudi2016: the reference implementation by @antonysigma in GPU autoscheduling with Mullapudi2016: the reference implementation halide/Halide#7787
Mullapudi2016-GPU: Reorder to avoid for-loops to be sandwiched between gpu_blocks. by @antonysigma in Mullapudi2016-GPU: Reorder to avoid for-loops to be sandwiched between gpu_blocks. halide/Halide#8647
Enable experimental Mullapudi2016 GPU scheduler for test-bench by @antonysigma in Enable experimental Mullapudi2016 GPU scheduler for test-bench halide/Halide#8650
Highlight Metal GPU code in stmt_html by @antonysigma in Highlight Metal GPU code in stmt_html halide/Halide#8659
Always ensure gpu_threads count >= warp size of 32 by @antonysigma in Always ensure gpu_threads count >= warp size of 32 halide/Halide#8656


Fix incorrect natural vector size on Zen4 by @abadams in Fix incorrect natural vector size on Zen4 halide/Halide#8570
Make it an error to use a device extern stage without target support by @abadams in Make it an error to use a device extern stage without target support halide/Halide#8794
Add support for adding tuple outputs in the configure() method by @abadams in Add support for adding tuple outputs in the configure() method halide/Halide#8649

Python

Fix argument order in rpow by @alexreinking in Fix argument order in __rpow__ halide/Halide#8677
Drop support for Python 3.8 by @alexreinking in Drop support for Python 3.8 halide/Halide#8678
Fix segfault in RDom's operator<< by @alexreinking in Fix segfault in RDom's operator<< halide/Halide#8679
Use ruff to format and lint Python code by @alexreinking in Use ruff to format and lint Python code halide/Halide#8684
Get raw Runtime::Buffer from Buffer in Python rather than use PyBuffer by @alexreinking in Get raw Runtime::Buffer from Buffer in Python rather than use PyBuffer halide/Halide#8682
Bind in-place update operators (e.g. +=) in Python by @alexreinking in Bind in-place update operators (e.g. +=) in Python halide/Halide#8683
Clean up Python dependencies; document uv usage by @alexreinking in Clean up Python dependencies; document uv usage halide/Halide#8694
Fix several printing segfaults. by @alexreinking in Fix several printing segfaults. halide/Halide#8700
Add Python bindings for serialization by @alexreinking in Add Python bindings for serialization halide/Halide#8718
Add all remaining IROperator ops to Python bindings by @alexreinking in Add all remaining IROperator ops to Python bindings halide/Halide#8771
Fix up memoize; bind to Python by @alexreinking in Fix up memoize; bind to Python halide/Halide#8778
Fix invalid Python type annotation and return types (#8772) by @rtzam in Fix invalid Python type annotation and return types (#8772) halide/Halide#8773
Expose Runtime::Buffer::cropped to C++ and Python Buffer by @rtzam in Expose Runtime::Buffer::cropped to C++ and Python Buffer halide/Halide#8787

Debugging

New feature flag to allow for stack backtrace/unwind by @mcourteaux in New feature flag to allow for stack backtrace/unwind halide/Halide#8703
Add filtering capabilities to HL_DEBUG_CODEGEN by @alexreinking in Add filtering capabilities to HL_DEBUG_CODEGEN halide/Halide#8627
Adding worker_thread_idle() for more informative profiling by @slomp in Adding worker_thread_idle() for more informative profiling halide/Halide#8719
Color IR output in cout and cerr.  by @mcourteaux in Color IR output in cout and cerr.  halide/Halide#8635
Improve output format for lowering passes timing. by @mcourteaux in Improve output format for lowering passes timing. halide/Halide#8749
fix(stmt-html): Fix embedded Buffer processing performance issue. by @mcourteaux in fix(stmt-html): Fix embedded Buffer processing performance issue. halide/Halide#8748
Use AArch64 assembly syntax on macOS with LLVM<22 by @alexreinking in Use AArch64 assembly syntax on macOS with LLVM<22 halide/Halide#8710

CodeGen

Mark our PTX kernels as kernels, to stop them from being stripped by @abadams in Mark our PTX kernels as kernels, to stop them from being stripped halide/Halide#8571
Math functions renaming table for GPU backends to support vectorized evaluation of math functions. by @mcourteaux in Math functions renaming table for GPU backends to support vectorized evaluation of math functions. halide/Halide#8595
Apply version constraints to iOS objects by @alexreinking in Apply version constraints to iOS objects halide/Halide#8546
Redirect bitwise ops to logical ops in case the arguments are bool. by @mcourteaux in Redirect bitwise ops to logical ops in case the arguments are bool. halide/Halide#8597
scalarize select condition for LLVM where possible by @abadams in scalarize select condition for LLVM where possible halide/Halide#8575
Add missing addition simplifier rules by @abadams in Add missing addition simplifier rules halide/Halide#8630
Bounds and alignment analysis through bitwise ops by @abadams in Bounds and alignment analysis through bitwise ops halide/Halide#8574
Make the vld2 pattern more obviously profitable by @abadams in Make the vld2 pattern more obviously profitable halide/Halide#8765
Fix vector shuffle for Vulkan CodeGen by @derek-gerstmann in [vulkan] Fix vector shuffle for Vulkan CodeGen halide/Halide#8621
Suppress warning on Windows for duplicate constant symbols. by @mcourteaux in Suppress warning on Windows for duplicate constant symbols. halide/Halide#8555
Use lossless_cast for saturating casts from unsigned to signed on x86 by @abadams in Use lossless_cast for saturating casts from unsigned to signed on x86 halide/Halide#8527
AMD Zen5 support by @changhoon-sung in AMD Zen5 support halide/Halide#8612

Compiler

Rework strict_float to use individual op intrinsics instead by @abadams in Rework strict_float to use individual op intrinsics instead halide/Halide#8641
Don't cache mutations of Exprs that have only one reference to them by @abadams in Don't cache mutations of Exprs that have only one reference to them halide/Halide#8518
Only use the nodes-visited set for nodes with multiple refs by @abadams in Only use the nodes-visited set for nodes with multiple refs halide/Halide#8547
In graph_equal(), call the correct implementation for comparing equalities between statements and expressions by @BachiLi in In graph_equal(), call the correct implementation for comparing equalities between statements and expressions halide/Halide#8611

Runtime

Support copying the overlapping region from one buffer to another. by @mcourteaux in Support copying the overlapping region from one buffer to another. halide/Halide#8463
Add (iOS) simulator target feature. by @alexreinking in Add (iOS) simulator target feature. halide/Halide#8623
Opt out of JIT exceptions by @abadams in Opt out of JIT exceptions halide/Halide#8615
Experimental: support removing unused runtime functions via

HL_RUNTIME_DROP_FUNCS environment variable.

PoC feature: drop functions from the runtime by @mcourteaux in PoC feature: drop functions from the runtime halide/Halide#8653



Apps

The onnx app now builds with CMake:

Add CMake for onnx app by @vawale in Add CMake for onnx app halide/Halide#8707
Fix halide_as_onnx_backend_test by @alexreinking in Fix halide_as_onnx_backend_test halide/Halide#8784



Documentation

Add note about relative paths to readme by @abadams in Add note about relative paths to readme halide/Halide#8613

Bugfixes

Fix #8534 [Buffer serialization does not match deserialization] by @abadams in Fix #8534 halide/Halide#8535
Fix CUDA HTML code printing bug. by @mcourteaux in Fix CUDA HTML code printing bug. halide/Halide#8558
Fix halide_get_cpu_features() linkage to avoid name mangling issues by @derek-gerstmann in [bug] Fix halide_get_cpu_features() linkage to avoid name mangling issues halide/Halide#8573
Fix for #8578 by @mcourteaux in Fix for #8578 halide/Halide#8579
Fix shuffle bug in CodeGen C. by @mcourteaux in Fix shuffle bug in CodeGen C. halide/Halide#8567
Check if expression is defined before trying to compute its constant_integer_bounds by @vksnk in Check if expression is defined before trying to compute its constant_integer_bounds halide/Halide#8599
Drop invalid "in-bounds" GEP for constant offsets by @alexreinking in Drop invalid "in-bounds" GEP for constant offsets halide/Halide#8768
Record trace_loads directly on ImageParam. by @alexreinking in Record trace_loads directly on ImageParam. halide/Halide#8803
RewriteLoadsAs32Bit should use the mutated index by @rootjalex in [PTX] RewriteLoadsAs32Bit should use the mutated index halide/Halide#8581
Set any_strict_float for wrapper module if target has strict_flag feature by @vksnk in Set any_strict_float for wrapper module if target has strict_flag feature halide/Halide#8681
Fix wrong type of the bound by @vksnk in Fix wrong type of the bound halide/Halide#8781
Fix UB-introducing rewrite in FindIntrinsics by @abadams in Fix UB-introducing rewrite in FindIntrinsics halide/Halide#8539
Fix rewrite that doesn't preserve type by @abadams in Fix rewrite that doesn't preserve type halide/Halide#8674
Fix nested select handling in remove_undef by @abadams in Fix nested select handling in remove_undef halide/Halide#8669
Add an underlying type to the halide_buffer_flags to prevent UB in C++ by @mcourteaux in Add an underlying type to the halide_buffer_flags to prevent UB in C++ halide/Halide#8690

Testing / CI

Limit depth more strictly in CSE fuzz test by @abadams in Limit depth more strictly in CSE fuzz test halide/Halide#8512
Skip fast exp/log/pow/sin/cosine tests without sse 4.1 by @abadams in Skip fast exp/log/pow/sin/cosine tests without sse 4.1 halide/Halide#8541
Hopefully fix flaky mullapudi reorder test by @abadams in Hopefully fix flaky mullapudi reorder test halide/Halide#8542
Skip test when code could be using x87 by @abadams in Skip test when code could be using x87 halide/Halide#8537
Fix stale GPU lifetime management tests for Vulkan. by @derek-gerstmann in [vulkan] Fix stale GPU lifetime management tests for Vulkan. halide/Halide#8601
Upgrade runner for cmake_cmake_file_lists job by @alexreinking in Upgrade runner for cmake_cmake_file_lists job halide/Halide#8609
Buildbot fixes by @alexreinking in Buildbot fixes halide/Halide#8706
Fix the pip packaging workflow by @alexreinking in Fix the pip packaging workflow halide/Halide#8708
Fix complexity of bounds of nested pure intrinsics by @abadams in Fix complexity of bounds of nested pure intrinsics halide/Halide#8689
Skip two sub-tests on llvm 21.1 by @abadams in Skip two sub-tests on llvm 21.1 halide/Halide#8782
Speed up simd_op_check_wasm by @abadams in Speed up simd_op_check_wasm halide/Halide#8780
Reduce the beam size in the adams2019 apps test to avoid timeouts by @abadams in Reduce the beam size in the adams2019 apps test to avoid timeouts halide/Halide#8786
Workaround llvm slow compile time bug in Mullapudi overlap test by @abadams in Workaround llvm slow compile time bug in Mullapudi overlap test halide/Halide#8793
Restore concurrent behavior to gpu_allocation_cache test by @abadams in Restore concurrent behavior to gpu_allocation_cache test halide/Halide#8792
Revert "Skip two sub-tests on llvm 21.1" by @abadams in Revert "Skip two sub-tests on llvm 21.1" halide/Halide#8806
Fix WASM splat op check test. by @mcourteaux in Fix WASM splat op check test. halide/Halide#8705

Build

Fix workflow for next release by @alexreinking in Fix workflow for next release halide/Halide#8514
Fix Debian packaging by @alexreinking in Fix Debian packaging halide/Halide#8524
Remove llvm version check from Makefile by @abadams in Remove llvm version check from Makefile halide/Halide#8533
Drop deprecated / unsupported setups for Halide 20 by @alexreinking in Drop deprecated / unsupported setups for Halide 20 halide/Halide#8508
Fix check for Windows never having aligned_alloc available. by @mcourteaux in Fix check for Windows never having aligned_alloc available. halide/Halide#8551
Don't include CMAKE_INSTALL_PREFIX when LIBDIR is absolute by @alexreinking in Don't include CMAKE_INSTALL_PREFIX when LIBDIR is absolute halide/Halide#8552
Add target-nvptx to target-all in vcpkg.json by @alexreinking in Add target-nvptx to target-all in vcpkg.json halide/Halide#8562
Fix top of LLVM, and remove upper limit of LLVM version from CMakeLists. by @mcourteaux in Fix top of LLVM, and remove upper limit of LLVM version from CMakeLists. halide/Halide#8568
build_halide_h asserts that every header it slurps in is one of the args by @abadams in build_halide_h asserts that every header it slurps in is one of the args halide/Halide#8559
Upgrade pybind11 to 2.11.1 by @alexreinking in Upgrade pybind11 to 2.11.1 halide/Halide#8616
Drop check for LLVM_LIBCXX in FindHalide_LLVM.cmake by @alexreinking in Drop check for LLVM_LIBCXX in FindHalide_LLVM.cmake halide/Halide#8617
Fix finding LLD on Homebrew when multiple versions are installed. by @alexreinking in Fix finding LLD on Homebrew when multiple versions are installed. halide/Halide#8619
Fix build on GCC 15 (Comes with Fedora 42). by @mcourteaux in Fix build on GCC 15 (Comes with Fedora 42). halide/Halide#8626
Constrain Clang and LLD searches to LLVM version by @alexreinking in Constrain Clang and LLD searches to LLVM version halide/Halide#8634
Disallow empty CMAKE_BUILD_TYPE on single-config generators by @alexreinking in Disallow empty CMAKE_BUILD_TYPE on single-config generators halide/Halide#8651
Add missing outputs to add_halide_library; fix advice in Lesson 21. by @alexreinking in Add missing outputs to add_halide_library; fix advice in Lesson 21. halide/Halide#8660
Bump the LLVM version in the pip package to 20.1.8 by @alexreinking in Bump the LLVM version in the pip package to 20.1.8 halide/Halide#8698
Prefer to build against libjpeg-turbo and document this. by @alexreinking in Prefer to build against libjpeg-turbo and document this. halide/Halide#8775
Add C++17 requirement to RunGenMain CMake target by @alexreinking in Add C++17 requirement to RunGenMain CMake target halide/Halide#8795
Allow llvm-ar in BundleStatic.cmake by @alexreinking in Allow llvm-ar in BundleStatic.cmake halide/Halide#8799
Fix dubious find_package logic in test/generator by @alexreinking in Fix dubious find_package logic in test/generator halide/Halide#8804
Warning when extra-output is requested w/o filename by @FabianSchuetze in Warning when extra-output is requested w/o filename halide/Halide#8671
Makefile linker flag fixes and cleanups by @abadams in Makefile linker flag fixes and cleanups halide/Halide#8764

Ongoing maintenance

Fix clang-tidy-19 errors by @steven-johnson in Fix clang-tidy-19 errors halide/Halide#8509
Remove unused function in HexagonOptimize by @steven-johnson in Remove unused function in HexagonOptimize halide/Halide#8511
Fix two non-idiomatic uses of node_type by @abadams in Fix two non-idiomatic uses of node_type halide/Halide#8520
Handle some misc TODOs by @abadams in Handle some misc TODOs halide/Halide#8528
Use a consistent idiom for visit_let by @abadams in Use a consistent idiom for visit_let halide/Halide#8540
Upgrade to clang-format 19 by @alexreinking in Upgrade to clang-format 19 halide/Halide#8543
Suppress clang-tidy warning for make_with_shape_of() by @steven-johnson in Suppress clang-tidy warning for make_with_shape_of() halide/Halide#8545
Remove debugging print left in by @abadams in Remove debugging print left in halide/Halide#8572
Fixes for llvm trunk by @abadams in Fixes for llvm trunk halide/Halide#8590
Another fix for llvm trunk by @abadams in Another fix for llvm trunk halide/Halide#8591
Our internal error macros were redesigned:

Accurately annotate Error system with [[noreturn]] by @alexreinking in Accurately annotate Error system with [[noreturn]] halide/Halide#8564
Teach compilers that internal_error does not return. by @alexreinking in Teach compilers that internal_error does not return. halide/Halide#8807
Use a new macro trick to avoid throwing in destructors. by @alexreinking in Use a new macro trick to avoid throwing in destructors. halide/Halide#8774


Move to opaque llvm pointers by @abadams in Move to opaque llvm pointers halide/Halide#8614
Avoid throwing from a destructor in PartitionLoops.cpp by @alexreinking in Avoid throwing from a destructor in PartitionLoops.cpp halide/Halide#8767
Bump version to 21.0.0 by @alexreinking in Bump version to 21.0.0 halide/Halide#8810
remove superfluous overload that causes compile errors by @ongjunjie in remove superfluous overload that causes compile errors halide/Halide#8654
Remove obsolete WasmExecutor specific debug macro. by @zvookin in Remove obsolete WasmExecutor specific debug macro. halide/Halide#8670
Add missing header by @vksnk in Add missing header halide/Halide#8680
Attempted fix for LLVM change by @abadams in Attempted fix for LLVM change halide/Halide#8642
Fix top LLVM: renamed NVPTX barrier intrinsics. by @mcourteaux in Fix top LLVM: renamed NVPTX barrier intrinsics. halide/Halide#8631

New Contributors

@changhoon-sung made their first contribution in AMD Zen5 support halide/Halide#8612
@vawale made their first contribution in Add CMake for onnx app halide/Halide#8707
@rtzam made their first contribution in Fix invalid Python type annotation and return types (#8772) halide/Halide#8773

Full Changelog: halide/Halide@v19.0.0...v21.0.0

View the full release notes at https://github.com/halide/Halide/releases/tag/v21.0.0.

---

[chenrui333]

Unhandled exception: Error: For SVE/SVE2 support, target_vector_bits=<size> must be set in target.

[cho-m]

Added workaround for now. Should only impact arm64 linux code https://github.com/halide/Halide/blob/main/src/Target.cpp#L250-L267

i.e. guarded by both

#if defined(__arm__) || defined(__aarch64__) || defined(_M_ARM64) || defined(_M_ARM64EC)

#ifdef __linux__

[github-actions]

🤖 An automated task has requested bottles to be published to this PR.

Caution

Please do not push to this PR branch before the bottle commits have been pushed, as this results in a state that is difficult to recover from. If you need to resolve a merge conflict, please use a merge commit. Do not force-push to this PR branch.