Vectorized Operations¶

Operations process vectors of values instead of single values.

Vectorization Benefits¶

• SIMD - Single instruction, multiple data
• Cache - Better cache utilization
• Overhead - Amortized interpretation overhead

Example: Filter Operation¶

// Scalar (row-at-a-time)
for row in rows:
    if row.age > 30:
        output.append(row)

// Vectorized
ages = column["age"]
mask = ages > 30  // SIMD comparison
result = rows[mask]

Data Chunk¶

Operations work on DataChunks:

struct DataChunk {
    columns: Vec<Vector>,
    selection: Option<SelectionVector>,
    count: usize,
}