wide

chore: Release wide version 1.6.1

• Reverted a compile time optimization that triggered a compiler bug. See this issue for more information.
• Fixed code path errors that resulted when avx512f was enabled without avx512dq, which is a valid potential build. CI checks have been added for this combination to prevent recurrence.

chore: Release wide version 1.6.0

• Potential Change: i64xN functions to_bitmask, any, all and none no longer guarantee that the sign-bit is used to decide whether an element is considered "true" or "false". This is being counted as a "bug" in the docs for previously being too specific.
• Added cast functions cast_unsigned, cast_signed, to_bits, from_bits.
• Added u8x32/i8x32 swizzle and swizzle_relaxed: a full-width 32-entry byte table lookup (vpermb on AVX-512-VBMI, vqtbl2 on NEON, emulated on AVX2/SSSE3).
• Fixed i8x32/u8x32 swizzle_half on AVX2: out-of-range indices now correctly zero their output lane (previously leaked self[..][0]).
• Made to_array, as_array and as_mut_array available in const contexts.
• Added four unbounded-shift functions for integers.
• Renamed float function pow_{simd-type-name} to powf_simd and deprecated powf.
• Added conversions between wide types and native intrinsics SIMD types.
• Added reduce_mul for integers.
• Added integer functions reduce_mul and mul_keep_low_high.
• Added overflowing arithmetic for integers.
• Added float function round_ties_even.
• Fixed bugs in the fallback paths of any, all, none and fast_clamp.
• Added support for shifting SIMD vectors by values of isize and usize and SIMD vectors of "opposite-signedness".
• Deprecated blend and replaced it with select and bitselect.
• Fixed bug in powf.
• Fixed clamp and added guarantees for fast_clamp.
• Added small optimizations to saturating_add/sub.
• Fixed shift operators overflow behavior.
• Fixed round which previously behaved like round_ties_even.
• Fixed UpperExp formatting for floats.
• Renamed integer function mul_widen to widening_mul and added it for remaining types.
• Added integer function mul_keep_high for remaining types.
• Added missing #[must_use] annotations
• Updated documenattion.

chore: Release wide version 1.5.0

• Added several functions and trait implementations that previously were only implemented for some types inconsistently.
• Added inherent simd_* comparison functions and deprecated the Cmp* traits.
• Added integer Div and Rem implementations.
• Added integer functions clamp, saturating_mul and saturating_div.
• Added signed integer functions is_positive and signum.
• Added additional float constants.
• Fixed slight bug in f64xN functions round and round_int.
• Fixed simd_ne NaN behavior for floats.
• Fixed bug in u64x8::simd_lt.
• Fixed bug in i64x8 and u64x8 function to_bitmask
• Fixed bugs in u32x16 functions any, all and none.
• Optimized signed integer function is_negative.

chore: Release wide version 1.4.0

• Added more float functions.
• Implemented Rem for float types.
• Renamed float function sign_bit to is_sign_negative and added is_sign_positive.
• Corrected several bugs with non-finite inputs for various math functions.

chore: Release wide version 1.3.0

• Fixes the behaviour of f32x16 functions is_finite and round_int. They previously gave incorrect output in some cases.
• Added more f32x16 functionality.

chore: Release wide version 1.2.0

• added reduce operations and dot to i16x32

chore: Release wide version 1.1.1

• Further improvements to the Neg impls on non-x86 targets.

chore: Release wide version 1.1.0

• add i64x2::min, i64x2::max.
• add u64x2::min, u64x2::max.
• add u64x8::min.

chore: Release wide version 1.0.3

• fix floating point negation edge cases, which also allows optimizations in more cases.

chore: Release wide version 1.0.2

• fix edge case where wide was using the wrong avx512 sub-features to select when safe_arch functions could be called, causing build errors.

• initial stable version

• to_bitmask implemented for all unsigned int types.

• Fixed additional build errors on wasm targets.

• Fixed type errors in the fallback implementations of some methods that completely prevented compilation of the crate on some targets.

• Breaking:
  • move_mask returns unsigned values now.
  • move_mask and many other methods renamed to better align with the standard library portable simd types. See issue 209 for details.