Expand description
SIMD-style i64 operations for integer-typed arrays.
Provides vectorized accumulation patterns for i64 data, avoiding the cost of i64->f64 conversion and preserving integer type fidelity. Uses chunks-of-4 manual unrolling for compiler auto-vectorization.
Operations that are inherently floating-point (mean, correlation, std dev, pct_change) remain f64-only.
Functionsยง
- cumsum_
i64 - Cumulative sum over i64 data. Uses wrapping arithmetic.
- diff_
i64 - First-difference over i64 data.
- rolling_
max_ i64 - Rolling maximum over i64 data using monotonic deque. O(n) amortized.
- rolling_
min_ i64 - Rolling minimum over i64 data using monotonic deque. O(n) amortized.
- rolling_
sum_ i64 - Rolling sum over i64 data with O(n) sliding window.
- simd_
max_ i64 - SIMD-style i64 maximum.
- simd_
min_ i64 - SIMD-style i64 minimum.
- simd_
sum_ i64 - SIMD-style i64 sum using 4-wide accumulator lanes.