1.3.0

Add:

• AddLower, PairwiseAdd/Sub, MaskedAbsOr, BitsFromMask
• AVX10_2 and Loongson LASX/LSX targets
• AVX3_SPR F16, WASM_EMU256 F64 types
• CeilInt/FloorInt, DemoteToNearestInt and F16/F64 NearestInt
• Complex number operations, F16/BF16 assignment operators
• emulated bf16/f16 Load/StoreInterleaved
• hwy::Warn/HWY_WARN, use instead of fprintf
• HWY_UNREACHABLE, HWY_VISIT_TARGETS
• i16 Dot, AverageRound, RoundingShiftRight/RoundingShr
• InterleaveEvenBlocks/InterleaveOddBlocks, MinMagnitude/MaxMagnitude
• masked comparisons, promote, round, GetBiasedExponent
• MulByPow2/MulByFloorPow2, MulRound, MulLower/MulAddLower
• PositiveInfOrHighestValue/NegativeInfOrLowestValue
• RVV groundwork for runtime dispatch, enable tuples
• spin wait, NanoSleep, Counter2/4 barrier, Divisor64, perf_counters

Improvements:

• dpbf16 WidenMulPairwiseAdd Exp2, AVX10.2 float->int, AVX3 GetExponent
• header-only abort.h/cc, tests runnable with Bazel8
• HWY_BROKEN_*: allow individual override
• Lanes: 'optional constexpr', AllBits1
• MaskedEq/Ne, NEON SumOfMulQuadAccumulate, MaskedReduceMin/Max, MulEven
• Profiler: report concurrency stats, 1.36x less overhead
• RVV various ops via superoptimizer
• SetThreadName: support more systems
• SVE2 SatWidenMulPairwiseAccumulate, SSE2/SSSE3 U16 Min/Max
• TargetName: no longer returns unknown for other arch
• ThreadPool autotune, avoid WakeAll
• topology: add NUMA node, support Windows/Apple

Fixes:

• avoid wraparound for -ftrapv, topology for offline CPUs/RVV
• warnings from -Wmissing-declarations/prototypes
• AdvSIMD_HPFPCvt on OSX
• f32->bf16 rounding: avoid unspecified built-in cast
• MSAN, PPC InvariantTicksPerSecond on QEMU, HWY_RCAST_ALIGNED, IsNaN
• vqsort for ascending order, add 8-bit test

Thanks to all contributors, especially johnplatts and eustas!

1.2.0

• Add InterleaveEven/InterleaveOdd, BitShuffle, GatherIndexNOr

• Add IsNegative, IfNegativeThenElseZero, IfNegativeThenZeroElse

• Add NEON_BF16, HWY_VERSION_GE/LT, HWY_EXPORT_T/HWY_DYNAMIC_DISPATCH_T

• Add PromoteInRangeTo/ConvertInRangeTo/DemoteInRangeTo

• Add Rol/Ror, RotateLeft/RotateLeftSame/RotateRightSame

• Add SatWidenMulPairwiseAccumulate, SatWidenMulAccumFixedPoint

• Add stats.h, bit_set.h, IsEitherNaN

• Add UI8/UI32/UI64 MulHigh, I64 MulEven/MulOdd/Mul128

• Add WidenMulAccumulate, MulEvenAdd, MulOddAdd

• contrib/bit_pack: support 32/64-bit lanes

• contrib/math: Add Exp2, Hypot

• contrib/matvec: Add MatVecAdd

• contrib/sort: Add VQ/HeapSelect, partial sort

• contrib/topology: add affinity, detect topology/cache size/CPU name

• Enable runtime dispatch for NEON/RVV, bazel modules, abort handler

• Remove DASSERT for negative Gather indices

• Support opting out of GUnit dependency

• Use SPR/ZEN4 bf16 dot product

• Known GCC 13 RVV issue: parts of sort_test and bit_pack_test disabled

• Known Clang RVV/QEMU issue: incorrect rounding mode in upper/lower halves

1.1.0

• Add BitCastScalar, DispatchedTarget, Foreach
• Add Div/Mod and MaskedDiv/ModOr, SaturatedAbs, SaturatedNeg
• Add InterleaveWholeLower/Upper, Dup128VecFromValues
• Add IsInteger, IsIntegerLaneType, RemoveVolatile, RemoveCvRef
• Add MaskedAdd/Sub/Mul/Div/Gather/Min/Max/SatAdd/SatSubOr
• Add MaskFalse, IfNegativeThenNegOrUndefIfZero, PromoteEven/OddTo
• Add ReduceMin/Max, 8-bit reductions, f16 <-> f64 conversions
• Add Span, AlignedArray, matrix-vector mul
• Add SumsOf2/4, I8 SumsOf8, SumsOfAdjQuadAbsDiff, SumsOfShuffledQuadAbsDiff
• Add ThreadPool, hierarchical profiler
• Build: use bazel_platforms
• Enable clang16 Arm/PPC runtime dispatch, F16 for GCC AVX3_SPR
• Extend Dot to f32*bf16, FMA to integer
• Fix: RVV 8-bit overflow, UB in vqsort, big-endian bugs, PPC HTM
• Improved codegen in various ops, fp16/bf16 tests and conversions
• New targets: HWY_Z14, HWY_Z15
• Test: add foreign_arch builders, CodeQL

1.0.7

• Add LoadNOr, GatherIndexN, ScatterIndexN
• Add additional float<->int conversions
• Codegen improvements for 8-bit shift, PPC Compress/Expand
• Fixes for MSVC, PPC, RVV, WASM, GCC 13, GCC 8.2, i686, f16 type, QEMU 7.2
• Support CMake args in Debian packaging

1.0.6

• Add MaskedGatherIndex, MaskedScatterIndex, LoadN, StoreN
• Add SatWidenMulPairwiseAdd, SumOfMulQuadAccumulate, PromoteUpperLowerTo
• Add F64 for Wasm, F64 AbsDiff
• Add F16 support to AVX3_SPR, RVV tuple (both not yet enabled)
• Validate all D args in x86 function signatures
• License: now dual Apache2/BSD3
• Doc: new users, vcpkg install instructions, AVX10 plans
• Doc: advice on dynamic dispatch plus -march flags
• Build: avoid installing hwy_test if !HWY_ENABLE_TESTS
• Codegen: improved PPC9 Find*True, variable-length CopyBytes
• Fix: GCC 8.2, MSVC, ICC, PPC9, SVE, arm64 MSVC issues
• Fix: IfNegativeThenElse, MulFixedPoint15, Debian changelog format
• Tests: faster builds (split up), use release builds

1.0.5

• Add Insert/ExtractBlock, BroadcastBlock/Lane, NumBlocks
• Add integer Le/Ge and [Neg]MulAdd, extend DemoteTo/PromoteTo
• Add Leading/TrailingZeroCount, HighestSetBitIndex, ReverseBits
• Add MaskedLoadOr, tuple Get/Set/Create, ReduceSum, WidenMulPairwiseAdd
• Add [ZeroExtend]ResizeBitCast, BitwiseIfThenElse, Find[Known]LastTrue
• Add AESRoundInv, AESKeyGenAssist
• Add contrib/math Atan2/SinCos, contrib/unroller
• Add fp16/bf16 support (Armv8, SVE, RVV), HWY_DYNAMIC_POINTER
• Add OrderedTruncate2To, Per4LaneBlockShuffle, TwoTablesLookupLanes
• Add SlideUp/Down[Blocks/Lanes], Slide1Up/Down, ReverseLaneBytes
• Add SetBeforeFirst, SetAtOrBefore/AfterFirst, SetOnlyFirst
• Add 8-bit Reverse2/4/8, Shl/Shr, RotateRight, Reverse, Mul
• Add 8/16-bit DupEven/Odd, TableLookupLanes
• Add F64 ApproximateReciprocal[Sqrt], 32/64-bit SaturatedAdd/Sub
• Build: Support Bazel modules
• Codegen improvements
• Compiler: support Clang 15/16
• Doc: add Github pages, support policy, evaluation
• Doc: publish AVX-512 throttling/startup findings
• Release: add signing
• Test: add GCC to Github Actions
• VQSort: small N speedups: fix seeding, func ptr, 8-wide network.
• VQSort: add BenchAllColdSort, VQSortStatic
• VQSort: fix subnormal/inf/NaN, support fp16, fix KV types
• Workarounds: RVV VXRM, x87 excess precision, missing intrinsics

1.0.4

• Add PPC8..10, SSE2, AVX3_ZEN4, NEON_WITHOUT_AES targets
• Add Expand, LoadExpand, integer AbsDiff, SumsOf8AbsDiff
• Improved Half/Twice support, codegen for Shift*Same
• Support Wasm in Godbolt
• Faster KV128 sorting
• Fix armv7 build config, CMake config mode
• Update RVV intrinsics for 1.0-draft

1.0.3

• Add RearrangeToOddPlusEven, Xor3, 8-bit CompressStore, HWY_ASSUME
• Add contrib/bit_pack for 8/16-bit lanes
• Add WASM_EMU256 target
• Documentation improvements
• Allow opting out of C++ stdlib usage for Compiler Explorer
• Update for new RVV intrinsics; faster WASM min/max and extmul/q15mul
• Fix UB, GCC atomic

1.0.2

• Add ExclusiveNeither, FindKnownFirstTrue, Ne128
• Add 16-bit SumOfLanes/ReorderWidenMulAccumulate/ReorderDemote2To
• Faster sort for low-entropy input, improved pivot selection
• Add GN build system, Highway FAQ, k32v32 type to vqsort
• CMake: Support find_package(GTest), add rvv-inl.h, add HWY_ENABLE_TESTS
• Fix MIPS and C++20 build, Apple LLVM 10.3 detection, EMU128 AllTrue on RVV
• Fix missing exec_prefix, RVV build, warnings, libatomic linking
• Work around GCC 10.4 issue, disabled RDCYCLE, arm7 with vfpv3
• Documentation/example improvements
• Support static dispatch to SVE2_128 and SVE_256

1.0.1

• Add Eq128, i64 Mul, unsigned->float ConvertTo
• Faster sort for few unique keys, more robust pivot selection
• Fix: floating-point generator for sort tests, Min/MaxOfLanes for i16
• Fix: avoid always_inline in debug, link atomic
• GCC warnings: string.h, maybe-uninitialized, ignored-attributes
• GCC warnings: preprocessor int overflow, spurious use-after-free/overflow
• Doc: <=HWY_AVX3, Full32/64/128, how to use generic-inl