Halide 12.0.0

[alexreinking]

We are pleased to announce the release of Halide 12.0.0!

This is mostly a quality of life and bugfix release to set the stage for larger changes in Halide 13 (which will require C++17).

You can download one of our binary releases here, or check one of the following package repositories:

• Vcpkg: https://github.com/microsoft/vcpkg/tree/master/ports/halide
• Homebrew: https://formulae.brew.sh/formula/halide

Language and Compiler

• Added align_extent scheduling directive Add align_extent(), to align extent but not min #5829
• Added TailStrategy::Predicate as an alternative to TailStrategy::GuardWithIf to use predicated loops unconditionally Add TailStrategy::Predicate #5856
• Added scatter() and gather() expressions to support reading from and writing to multiple locations in update definitions Prototype of multiple scattering update definitions #5553
• Added internal memoization to Adams2019 autoscheduler (performance improvement) [adams2019] Add caching to autoscheduler #5697 [adams2019] Restructure autoscheduler + add timer #5654
• Removed old-style realize() methods which had been deprecated Deprecate old-style realize() methods #5676
• Removed deprecated scheduling directive overloads Remove all deprecated methods for Halide 12 #5656
• Many simplifier and bounds inference improvements and bugfixes is_positive_const and is_negative_const broken for (some) casts #5615 Add fixes to overflow analysis in bounds inference #5618 Add fuzzer to bounds_of_expr_in_scope + fix discovered overflow bugs #5895 Guard against overflow in constant folding for EQ rewrite rules (Fixes #5998) #6002

Backends

• Added support for AVX512 VNNI instructions Add support for AV512 VNNI instructions #5725 Add support for AVX-512 VNNI saturating dot products #5807
• Removed OpenGL/GLSL backend Remove OpenGL support (part 1) #5626
• Fixed various errors with large_buffers runtime: fix pointer calculations to avoid overflows. Fixes #5713. #5716 Fix flattening of ramps involving 64-bit mins #5940
• Improved support for sdot and udot instructions on ARM (where supported) Use dot products for factor 4 vector reduction sums on ARM #5954
• Improved support for WebAssembly SIMD ops, when compiling with LLVM 13 Add support for widening_mul in wasm #5849 Add support for pairwise_widening_add in wasm #5850 Add support for i16x8.q15mulr_sat_s in wasm #5853 Add support for wasm-simd saturating-narrow ops. #5854 Add support for wasm dot-product instruction #5861 Add wasm support for int32->f64 and f32->f64 simd ops #5863
• PyStub generators must now choose to use either only positional arguments or only keyword arguments. This is an ABI break Constrain calling convention for PyStub #5761

Build

• Added scripts to create Ubuntu packages Add Ubuntu package scripts #5754 CMake clean-up #5967
• Added experimental support for ClangCL on Windows Improve ClangCL support by disabling, fixing warnings #5876
• Added support and pre-built binaries for macOS ARM64
• Halide headers no longer inject stack space linker flags on Windows; now, the compiler runs on a fiber with enough stack space Use fibers to guarantee stack size on Windows #5873
• Halide shared library no longer exposes LLVM symbols on macOS and Linux. Help wanted for Windows! Use linker tools on OSX & Linux to limit exports (#4651) #5659

---

This discussion was created from the release Halide 12.0.0.