- c3407cc Unpoison AMX tile store destinations under MSAN. by Krzysztof Rymski · 12 hours ago upstream/master
- c2cc22e Merge pull request #3487 from Wint3rNight:partialsort-k-zero by Copybara-Service · 13 hours ago
- 63826d8 Merge pull request #3497 from kennethassogba:fix/iguana-standalone-main by Copybara-Service · 13 hours ago
- 43ee508 Add native support for pointer and 32/64-bit enum keys in hwy btree since they have a direct bijective order preserving mapping. by Nikhil Dev Goyal · 15 hours ago
- 46f4a61 Add exhaustive leaf sweep tests, shard btree tests, and add presubmit check by Nikhil Dev Goyal · 15 hours ago
- 84351f1 add MaskedLoadExpandOr by Jan Wassenberg · 18 hours ago
- 100c82c Fix standalone iguana_test build by Kenneth Assogba · 33 hours ago
- 1693ca7 Add iterator walk benchmarks + expand kSweepSize to a log scale between 10K and 4M which is useful in plotting graphs by Nikhil Dev Goyal · 4 days ago
- e8f6dfa Reduce dispatch overhead: avoid function call, relaxed atomic. by Jan Wassenberg · 5 days ago
- af18aa9 Merge pull request #3457 from LXYan2333:refactor/x86-target-detection by Copybara-Service · 5 days ago
- fca7c5e Faster Arm lookups: masking, special-case for known vector sizes by Jan Wassenberg · 6 days ago
- 98f80f8 LeMac: make the key and state plain data, usable from non-SIMD code. by Jan Wassenberg · 6 days ago
- 3cb81dc add safe arithmetic/cast helpers, HWY_NODISCARD, IsEnum by Jan Wassenberg · 6 days ago
- a821fce Fix VQPartialSort aborting in debug builds when k is 0 by Wint3rNight · 6 days ago
- 4d677fb Merge pull request #3482 from advitrocks9:i8-sum-of-lanes by Copybara-Service · 6 days ago
- 87330ce Merge pull request #3465 from xia-chao:fix/emu128-div-by-zero by Copybara-Service · 6 days ago
- ed1ea1b Faster LeMac finalization for non-secret keys by Jan Wassenberg · 6 days ago
- a966054 Merge pull request #3468 from advitrocks9:fix-loggamma-range by Copybara-Service · 6 days ago
- 9f030d9 Avoid signed overflow in EMU128 ReduceSum by Advit Arora · 7 days ago
- 8d07f07 add HWY_DISABLE_AUTOVEC by Jan Wassenberg · 7 days ago
- 85cb4a5 Improve float32 SinCos and Tan accuracy via 4-part Cody-Waite range reduction. by Nikhil Dev Goyal · 7 days ago
- a51433f This CL introduces a tier 2.5 in insert() path before going to leaf splitting. by Nikhil Dev Goyal · 7 days ago
- d6a091e fix build for LSX/LASX LoadInterleaved3: support bf16 by Jan Wassenberg · 7 days ago
- 092c614 add macro protect for AT_HWCAP/AT_HWCAP2 by LXYan2333 · 7 days ago
- 6d99a21 make resolve function static by LXYan2333 · 7 days ago
- 7ee9529 Use SumsOf8 for x86 i8 lane sums by Advit Arora · 7 days ago
- ab51ce7 do not enable ifunc test by default by LXYan2333 · 7 days ago
- e0cf749 remove ZeroBytesInline by LXYan2333 · 7 days ago
- 8ed00de fix for newer clang's unused warning by LXYan2333 · 7 days ago
- 242f7c2 use shared hwy::CapBits by LXYan2333 · 7 days ago
- a4221ed fix for old clang-format by LXYan2333 · 7 days ago
- 2191a34 Merge pull request #3467 from advitrocks9:fix-nearestint-saturation by Copybara-Service · 7 days ago
- 2ed1121 emu128, scalar: return +/-Inf on division by zero by Xia Chao · 7 days ago
- 16361f5 add LeMac-x2, a fast AES-based keyed hash for tree hashing. by Jan Wassenberg · 8 days ago
- 24909a6 use up to 256-bit vectors in highwayhash-inl.h by Jan Wassenberg · 8 days ago
- 358bfea [HWY] Support lazily changing the wait mode. by Trevor Gale · 8 days ago
- c752a44 Define a non TAP exhaustive SweepAll test suite, which sweeps through all possible f32 values and 10^8 values for f64(since testing all values for f64 is computationally infeasiable). by Nikhil Dev Goyal · 8 days ago
- 1f57bc9 Add AlignedPtr, replacing std::unique_ptr by Jan Wassenberg · 8 days ago
- 9051ce7 disable AMX for clang < 16 due to intrinsics bug by Jan Wassenberg · 8 days ago
- 6f925d3 add HWY_PRESERVE_MOST and HWY_INLINE_ATTR by Jan Wassenberg · 8 days ago
- 0358602 Fix CombineSlideDownLanes build for SSE2 by Jan Wassenberg · 8 days ago
- dfbaeab Fix DemoteTo(f64 -> f32) finite overflow by Jan Wassenberg · 8 days ago
- 96eddc4 contrib/math: fold -0.5 into LogGamma constant, merge inf blends by Advit Arora · 10 days ago
- b93ad44 Merge pull request #3466 from Subhan1794:fix-range-coder-oob by Copybara-Service · 11 days ago
- 79d25d0 Add seperate memory benchmark for memory comparison after 100K random keys (between [0, 20*N])insertion one by one using insert() for both hwy:: and absl:: to ensure a fair apples to apples comparison, since the existing memory comparison compares memory usage after a bulk build which is more in favor of hwy since we have a bottom up bulk build greedy algorithm which they dont.. by Nikhil Dev Goyal · 11 days ago
- d2f5981 Avoid TableLookupLanes in AVX3 16-bit SlideUpLanesOr when the slide by Jan Wassenberg · 11 days ago
- 67e8cc8 the stddef header is used in aarch64 ifunc test by LXYan2333 · 11 days ago
- 3e7a91b remove unused include by LXYan2333 · 11 days ago
- 1598461 add aarch64 ifunc test by LXYan2333 · 11 days ago
- c5eb14b change word by LXYan2333 · 11 days ago
- 571f813 make sure new seperated target detect logic works in ifunc resolver during program init by LXYan2333 · 11 days ago
- 50233b5 add test to make sure DetectTargetsImpl accept rvalue by LXYan2333 · 11 days ago
- aa7658b mark sve detect function maybe unused by LXYan2333 · 11 days ago
- f06ee41 contrib/math: fix LogGamma for huge and subnormal inputs by Advit Arora · 11 days ago
- 6588b1a Fix NearestInt saturation at 2^31 and 2^63 on SCALAR and EMU128 by Advit Arora · 11 days ago
- 4ef52f4 Fix OOB reads in range coder on corrupt/truncated input by Subhan1794 · 11 days ago
- 1a2ca8c Support bulk build from unsorted + duplicate keys as well in Build(). by Nikhil Dev Goyal · 11 days ago
- f444393 emu128: fix division by zero by Xia Chao · 11 days ago
- 4f1a7cf change word by LXYan2333 · 12 days ago
- e56c7c4 port loongarch by LXYan2333 · 12 days ago
- e754529 port rvv by LXYan2333 · 12 days ago
- 460364c port s390x by LXYan2333 · 12 days ago
- 3d84c0a port ppc by LXYan2333 · 12 days ago
- b3c4a63 add arm's comment by LXYan2333 · 12 days ago
- 411313b Merge master and resolve target detection conflict by LXYan2333 · 12 days ago
- ee91931 port arm by LXYan2333 · 12 days ago
- 7ce03ef Merge pull request #3454 from jdymitarai:fix-vqsort-near-sorted-probe-3139 by Copybara-Service · 12 days ago
- 8f7e1aa add a missing macro comment by LXYan2333 · 12 days ago
- ee53fe2 Address target detection header review feedback by LXYan2333 · 12 days ago
- 183ccd5 clean headers by LXYan2333 · 12 days ago
- 88c8aa5 Avoid library calls when zeroing target detection buffers by LXYan2333 · 12 days ago
- 700c1bd Expose target detection implementation in a public header by LXYan2333 · 12 days ago
- b5cc09a Refactor x86 target detection to isolate platform calls by LXYan2333 · 12 days ago
- 949f3ca vqsort: address review comments on presorted probe and deduplicate scalar fallback by jdymitarai · 14 days ago
- f30d0d2 vqsort: use BaseCase in scalar fallback and restrict narrow vector probe to sizeof(T) >= 4 by jdymitarai · 14 days ago
- 199eaec vqsort: add presortedness probe and scalar fallback for narrow vectors (fixes #3139) by jdymitarai · 14 days ago
- 60583da Merge pull request #3442 from Arthur031221:fix-f16-promote-inf-nan by Copybara-Service · 2 weeks ago
- dcd348c Merge pull request #3404 from dcsid:codex/highway-f16-binary by Copybara-Service · 3 weeks ago
- 58165b1 Reduce float16 binary math test repetitions on slower builds by dcsid · 3 weeks ago
- e195862 Small API, robustness and documentation cleanups for hwy BTree by Nikhil Dev Goyal · 3 weeks ago
- b84214b Make BTree move operations transfer contents, matching copy and swap BTree can run in two modes: owning (default-constructed, holds its own nodes) or adapter (constructed from a BTreeState*, operating on state owned by someone else, e.g. in btree_set.h). by Nikhil Dev Goyal · 3 weeks ago
- 741eaa5 Cap descriptor width to 64 bytes in FindChild. by Nikhil Dev Goyal · 3 weeks ago
- 459db33 Merge pull request #3420 from Wint3rNight:ne-unordered-nan by Copybara-Service · 3 weeks ago
- 343398d Add float16 two-input math functions by dcsid · 3 weeks ago
- 1332fe8 Add erase(iterator) to the Highway B-tree and document iterator invalidation by Nikhil Dev Goyal · 3 weeks ago
- 2b87575 Fix emulated PromoteTo(f32 <- f16) for infinity and NaN by Arthur031221 · 3 weeks ago
- ab9f4c1 Merge pull request #3431 from Arthur031221:fix-bins-bint by Copybara-Service · 3 weeks ago
- d6a05fc rename detail::Func to detail::FuncImpl to avoid conflicts by Jan Wassenberg CA · 3 weeks ago
- c80017c Use Load instead of LoadU in remainder handling. by Nikhil Dev Goyal · 3 weeks ago
- 6fd5bfb Print Bins counts as double instead of %llu by Arthur031221 · 3 weeks ago
- ffc0ed9 warning fix: anon namespace and missing header by Jan Wassenberg · 3 weeks ago
- 62df766 Merge pull request #3403 from dcsid:codex/highway-f16-hyperbolic by Copybara-Service · 3 weeks ago
- 2c1df41 Merge pull request #3402 from dcsid:codex/highway-f16-inverse-trig by Copybara-Service · 3 weeks ago
- 824a49c Fix hwy::Bins for 64-bit bin counts by Arthur031221 · 3 weeks ago
- 4ac866f Add float16 hyperbolic math functions by dcsid · 3 weeks ago
- e4f85a4 Add float16 inverse trig, cube root, and error functions by dcsid · 3 weeks ago
- 32ae733 Add worst-case boundary tests for FastPow, FastTanh, and FastAtan, and update FastPow's error bound to 0.021%. by Nikhil Dev Goyal · 3 weeks ago
- c4604a8 Add deterministic worst-case regression tests for FastLog and FastExp by Nikhil Dev Goyal · 3 weeks ago
- ae7ac9d Expand FastTan float32 range to [-39000, +39000] and fix asymptote sign flips. by Nikhil Dev Goyal · 4 weeks ago
- 62249a5 add more testdata from snappy plus enwik7 by Jan Wassenberg · 4 weeks ago