2.0.4 (2023-10-05)

Make

• Avoid uncommon Python version (c290513)

2.0.3 (2023-10-05)

Make

• Temporarily disable SVE checks (8ad641f)

2.0.2 (2023-10-05)

Make

• Use arch=armv8.2-a+dotprod flag (1f2dc65)

2.0.1 (2023-10-05)

Fix

• Revert to older neon_i8 implementation (3bb25d5)

2.0.0 (2023-10-05)

Add

• Parallel cdist API using OpenMP (153999a)

Break

• Remove pdist (ce04e5c)

1.5.2 (2023-10-05)

Fix

• Higher error tolerance (5c3b5b4)

1.5.1 (2023-10-05)

Fix

• fail-fast flags (5a87903)

Make

• Continue builds if one fails (195f79c)
• Prepare release (42a66af)

1.5.0 (2023-10-05)

Add

• cdist implementations with OpenMP (293504a)
• AVX-512 variants for i8 metrics (25412ee)
• Benchmark for SciPy (5c65142)
• Dynamic dispatch and hardware introspection (6864829)
• Hardware capabilities checks in Python (d1d51ab)
• New distances and project structure (9c833a5)
• Preceise calculation mode (b213ffb)
• Prototype for Intel AMX (c96f8f9)

Break

• New Python binding (b394995)
• Revert to header-only design (b0c1d32)

Docs

• Add bullet-points (5633785)
• Improve README (0f3df36)
• New README (a803ffd)

Fix

• AVX-512 masks undefined behaviour (857afba)
• Compilation (2d8422d)
• Correct and branchless masking in AVX-512 (0a32408)
• From SQRT macro to RSQRT (d6b6cee)
• From similarity to distance (a148d9e)
• Increase relative tolerance levels (5faee24)
• Low-precision accumulation in L2sq for i8 (b875c39)
• Remove false (5a66779)
• Skipped i8>i16>i32 upcasting step in AVX2 (2e43ab7)

Improve

• SIMSIMD_SQRT param to avoid LibC (196d679)
• Accumulating i8 vectors with integers (a7735aa)
• Add baseline comparison (53eecc3)
• Add GCC options directly to file (1296d78)
• Avoid _Pragma parameterized macros (e33e9f7)
• Benchmarking suite (0de1cb2)
• Equi-dimensional benchmarks (881091d)
• Expose pointers for USearch (d8acf83)
• Extend use of _mm_mask_rsqrt14_pd (61e401c)
• Jan Kadlec's magic numbers for rsqrt (ebbdb86)
• Multi-threaded benchmarks (6653c85)
• Normalize vectors before benchmarks (da3e8ed)
• Parameterizing benchmarks (fa220a7)
• Precompute baselines (bc48ff3)
• Repalce lib-wide compiler flags (53c15a4)
• Use rsqrtss in AVX2 metrics (47fac21)
• Use vrsqrte_f32 on Arm NEON (a5a868c)

Make

• Add avx512bw flag (bcba790)
• Add more Sapphire Rapids ISA subsets (77472b1)
• Compilation fixes (fbb3cca)
• Log environment (f641e08)
• Missing SciPy deps (f8ccb77)
• Switch to GCC 13 (2f24989)
• Switch to separate C files (2293ed3)

Refactor

• Reorganziation (6f78495)

Test

• Batch requests in Python (a594758)

1.4.0 (2023-09-05)

Add

• cos_i8x32_avx2 (e569561)
• dot_f16x4_neon (ff2d333)

Fix

• _mm_cvtph_ps to _mm256_cvtph_ps (5096928)

Improve

• #undef functionality overrides (c99747b)
• sqrt -> sqrtf (34d8885)

1.3.0 (2023-09-04)

Add

• f16 functions for AVX2 (88eb85c)

Fix

• Removing the square root (acf5a4d)

Make

• Build on macOS with GCC (07ca59a)
• Build python on MacOS with GCC (70a67ad)
• Disable python universal builds (636185b)
• Enable python MacOS build (e35fa8d)
• Install GCC 12 for MacOS (7dda672)
• Polish Release (4db6dae)

1.2.0 (2023-09-04)

Add

• Placeholders for string similarity (bd0d50c)

Fix

• avx512bw label for masked loads (ccce7c8)
• avx512f flag for __m512 registers (14ceef0)
• avx512fp16 compilation flag (f8640e0)
• fma flag for AVX2 cosine and dot-product (850a0e8)
• AVX-512 flag for _mm256_popcnt_epi64 (9055a02)
• C union instantiations (ebc4b12)
• Check only relevant symbols in tests (aa77e12)
• Compilation on non-SVE Arm machines (59cdc05)
• Disable AVX-512 for Windows and MacOS (352a5a1)
• Tanimoto out-of-range access (c21ecbc)

Improve

• Annotate GCC on how to compile code (4d4c267)

Make

• C compilation flags (027a0fd)
• Compile with gcc 12 (1deb593)
• Disable non-Linux packages (ae2fa18)
• Dont cancel jobs when one failed (7e79763)
• Install clang 16 (ad15f3b)
• Polish prerelease (81b7ab1)
• Relocate files (1e6c0a5)
• Set Flags env variables (84ceed6)

Refactor

• Dont set version to latest for MinGW (3d744ad)

1.1.2 (2023-08-22)

Make

• Disable versioning needs (ca9bb37)
• Restore temporary disabled jobs (953f7f3)
• Test build wheels without setup docker in Mac-Os (a952343)
• Update package classifiers (4d32c47)

1.1.1 (2023-08-22)

Make

• Change Docker setup action (4b7561b)

1.1.0 (2023-08-22)

Add

• l2sq_i8x16_neon (bb5f8ec)

Docs

• Update README.md (954b81b)

Fix

• simsimd_tanimoto_maccs_avx512 (f7ddbf5)

Improve

• Tanimoto coefficients with larger words (8ddedc6)

Make

• Prepare PyPi release (db19ab4)

1.0.3 (2023-07-14)

Fix

• Python deps & x86 benchmarks (6040a51)

1.0.2 (2023-07-13)

Make

• Disable Win32 and other deprecated builds (6fafb73)

1.0.1 (2023-07-13)

Make

• Fix parsing file on Windows (939ef60)

1.0.0 (2023-07-13)

Add

• cos_f16_sve common metric kind for Arm SVE (7b29466)
• simsimd_cos_i8x16neon (4dfe9d5)
• to_int method for capsules (c0ca719)
• Arm Neon f16 dot product, refactor visibility (23567db)
• Arm SVE acceleration for f16 dot-product, L2 (de34a04)
• AVX-512 for dot-product on Sapphire Rapids (401ed6b)
• Benchmarks for cosine similarity (e29c177)
• C99 version (06fb120)
• capsule for dot_f32sve (b5cb7b1)
• capsule for dot_f32sve (ff23d94)
• Capsules for Chemistry research (bf6ec11)
• configuration for test debug (8a82485)
• Graviton 3 results (8ea912a)
• Hamming distance with scalar, AVX-512, NEON (cbd2b85)
• Initial benchmark and AVX2, Neon, SVE (fb0aa11)
• Separate chemsitry-oriented measures (dc00183)
• SVE cosine distance of dynamic length vecs (3162f7d)
• SVE-accelerated Hamming distance (e407ab3)
• Tanimoto distances (168f026)
• test for module using (c5afa3a)

Build

• AVX-512 code for Hamming distance (d6a7820)
• cpython module and test (ac5985e)

Create

• CPython wrapper (54b5643)
• CPython wrapper (9612b04)

Docs

• Add functionality table (8b08703)
• Add functor annotation (a523a7b)
• Added AVX-FMA benchmarks on Threadripper (b683bf0)
• Indent (649e6a0)
• L2 annotation (2577dfc)
• Reordered (dde94a8)

Fix

• simsimd_cos_i8x16neon for Gravitons (ccad0b4)
• Arm Neon ordering FMA arguments (2ab1293)
• Benchmark compilation (9209bfb)
• Capsule destructors (12f756e)
• Casting to f16 (709e346)
• Casting to f16 (6245890)
• Compilation (905942a)
• Compilation error - unnamed argument (1311da4)
• Cosine similarity with Arm Neon (d2d8ad5)
• Duplicate symbol in cosine similarity (d791376)
• Hamming iteration step with SVE (e0194fa)
• IP error compensation with normalization (b577ad8)
• Missing includes for size_t (74760c7)
• Missing MSVC intrinsics (0fe4c1e)
• Module name (8a08231)
• Quantized i8 dot-product on AVX (725152d)
• Separate similarity and distance measures (3e922ce)

Improve

• Extended benchmarks to include chem.h (b222949)
• Using PyTest (8c29c45)

Make

• Downgrading to C++11, CMake imterface (add627d)
• Fix tests (a034a5c)
• Script to update version (9b12d67)
• Semantic Releases (208290e)
• Specifying the header sources (ce95f7c)
• Using different debuggers on MacOS (7390b7d)

Refactor

• From vmlaq_f32 to vfmaq_f32 (16f6f34)
• From similarities to distances (5b02e49)
• move to python folder (8ec854f)
• move to python folder (a507c56)
• Project name (8648ee0)
• Shorter function names (1d20d4d)
• Switching C++11 -> C99 (3bd535f)
• Type names (63a27ec)