Crate varpulis_simd 

SIMD-optimized operations for high-performance event processing

This crate provides vectorized implementations of common operations:

• Aggregations (sum, min, max, avg)
• Batch comparisons for filtering
• Field extraction to contiguous arrays
• Incremental aggregation accumulators

Structs

IncrementalMinMax

Incremental min/max tracker using BTreeMap for O(log n) operations

IncrementalSum

Incremental sum accumulator - O(1) updates instead of O(n) recomputation

Functions

compare_gt_f64

SIMD-optimized greater-than comparison Returns a bitmask where bit i is set if values[i] > threshold

compare_lt_f64

SIMD-optimized less-than comparison

extract_field_f64

Extract float field values from events into a contiguous array Returns None values as NaN for SIMD processing

extract_field_f64_filtered

Extract float field values from events, filtering out None values Returns (values, indices) where indices maps back to original positions

max_f64

SIMD-optimized max of f64 values

min_f64

SIMD-optimized min of f64 values

simd_avg

Compute avg of a field across events using SIMD

simd_max

Compute max of a field across events using SIMD

simd_min

Compute min of a field across events using SIMD

simd_sum

Compute sum of a field across events using SIMD

sum_f64

SIMD-optimized sum of f64 values Uses AVX2 (256-bit) when available, falls back to scalar