Fix: Jensen Shannon square roots (ashvardanian#233)

Infer boolean types in Python

Closes ashvardanian#202

Closes ashvardanian#184

Previous AVX-512 implementation of complex products
used an extra ZMM register for `swap_adjacent_vec`.
Moreover, they used the `vpshufb` instruction available
only with the Ice Lake capability and newer.

The replacement uses the `_mm512_permute_ps` and
its double-precision variant.

Relates to: albumentations-team/AlbumentationsX#118 (comment)

### Patch

- Improve: Round integer distances (c487b55)
- Fix: Absolute tolerance bound for integers (73a9ff7)
- Make: Skip flaky Arm failures (6be67bb)
- Fix: NEON guard for u8 dot dispatch (2c5876d)

### Patch

- Make: Same upload/download CI versions (ae9e567)

…n#296)

GCC and Clang do not recognize the  target attribute.
The correct flag for AVX-VNNI (AVX2-VNNI) instructions is .

This fix enables SIMSIMD_TARGET_SIERRA to compile successfully with
both GCC and Clang, allowing runtime dispatch to select optimal
SIERRA kernels on compatible CPUs.

Changes:
- include/simsimd/dot.h: avx2vnni -> avxvnni in pragma directives
- include/simsimd/spatial.h: avx2vnni -> avxvnni in pragma directives

Tested with:
- GCC 15.2.0
- Clang 20.1.8

Co-authored-by: Seungwon Yang <seungwon.yang@navercorp.com>

### Patch

- Fix: Replace `avx2vnni` with `avxvnni` for Sierra Forest (ashvardanian#296) (a8bb232)
- Make: Remove `NPM_TOKEN` for OIDC publishing (13cd5bc)
- Make: Sign rebase with GitHub Actions bot (e7b89b5)
- Fix: Revert to `atol=1` for test integer outputs vs SciPy (b75bdbd)

Co-authored-by: Markus Graf <24669860+markusalbertgraf@users.noreply.github.com>

### Patch

- Fix: Wrong predicate width in BF16 SVE L2 kernel (ashvardanian#301) (87ae846)
- Improve: FreeBSD comp-time target selection (ashvardanian#300) (cb11f8b)

…n#302)

`simsimd_capabilities` probes SIMD instructions with an uninitialized
`dummy_input` buffer and `n=0`. SVE implementations use `do { ... } while`
loops that always execute the body once. MemorySanitizer doesn't understand
SVE predicated loads and reports use-of-uninitialized-value.

Initialize the buffer to zero to silence the false positive.

Co-authored-by: Alexey Milovidov <18581488+alexey-milovidov@users.noreply.github.com>

### Patch

- Fix: Initialize `dummy_input` to fix MSan false positive (ashvardanian#302) (c2ad842)

MSan cannot track initialization through SIMD intrinsics (SVE, NEON,
SSE, AVX), causing false-positive "use-of-uninitialized-value" reports.
Add `__msan_unpoison` calls after every dispatch macro to mark results
as initialized.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

MemorySanitizer cannot track data flow through SIMD intrinsics
(SVE, NEON, SSE, AVX), causing false-positive "use-of-uninitialized-value"
reports. For example, SVE predicated loads only access memory for active
lanes, but MSan sees the full vector width as a memory read, flagging
tail elements as uninitialized.

Add `__attribute__((no_sanitize("memory")))` to `SIMSIMD_PUBLIC` and
`SIMSIMD_DYNAMIC` macros when MSan is detected via `__has_feature`.
This disables MSan instrumentation for all SimSIMD functions, which is
appropriate since they are entirely SIMD code that MSan cannot analyze.

The previous approach of unpoisoning results after dispatch (in lib.c)
was insufficient because MSan aborts inside the function body before
the dispatch wrapper can unpoison the output.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

`simsimd_capabilities` probes SIMD functions with n=0 to pre-initialize
dispatch function pointers. SVE implementations use `do { } while (i < n)`
loops that always execute the body once, even with n=0. MemorySanitizer
instruments SVE predicated loads (`svld1_f32` etc.) as full-width vector
reads regardless of the predicate mask, so it reports use-of-uninitialized
memory when the buffer is smaller than the SIMD register width.

Increase the dummy buffer from 8 bytes (`double[1]`) to 256 bytes
(`double[32]`) to cover the widest possible SVE vector (2048 bits).

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

Fix MSan false positive: enlarge dummy buffer for SVE predicated loads

Include `<unistd.h>` for ARM POSIX capability detection

LLVM's MSan does not instrument ARM SVE intrinsics — there is zero
handling of `aarch64_sve_*` in MemorySanitizer.cpp (as of LLVM 20),
only a file-level "FIXME: This sanitizer does not yet handle scalable
vectors". This causes false positives when `svaddv` produces scalar
results that MSan considers uninitialized.

Add `SIMSIMD_UNPOISON` calls after every `svaddv` reduction in SVE
function bodies across spatial.h, dot.h, binary.h, and sparse.h.
Define the macro in types.h with `__msan_unpoison` when MSan is active.

See also: llvm/llvm-project#165028