Skip to main content

Vectorized Operations

Vectorized operations process multiple field elements in parallel using SIMD instructions.

Type Definition​

typedef struct __vbn254fr {
    uint64_t __handle;
} vbn254fr_t[1];

Memory Management​

// Allocation and initialization
void vbn254fr_alloc(vbn254fr_t vfr, size_t len);
void vbn254fr_free(vbn254fr_t vfr);

Initialization​

// Set all elements to scalar value
void vbn254fr_set_ui_scalar(vbn254fr_t out, uint64_t x);

Arithmetic Operations​

Element-wise operations on vectors:

// Addition: out[i] = a[i] + b[i] mod p
void vbn254fr_addmod(vbn254fr_t out, const vbn254fr_t a, const vbn254fr_t b);

// Multiplication: out[i] = a[i] * b[i] mod p
void vbn254fr_mulmod(vbn254fr_t out, const vbn254fr_t a, const vbn254fr_t b);

Operations with Constants​

// Add constant to all elements: out[i] = a[i] + k mod p
void vbn254fr_addmod_constant(vbn254fr_t out, const vbn254fr_t a, 
                               const vbn254fr_constant* k);

// Multiply all elements by constant: out[i] = a[i] * k mod p
void vbn254fr_mulmod_constant(vbn254fr_t out, const vbn254fr_t a, 
                               const vbn254fr_constant* k);

// Montgomery multiplication with constant
void vbn254fr_mont_mul_constant(vbn254fr_t out, const vbn254fr_t a,
                                 const vbn254fr_constant* k);

Use Cases​

Vectorized operations are ideal for:

  • Batch Processing: Process multiple values simultaneously
  • Performance: Leverage SIMD instructions for speed
  • Parallel Computations: Element-wise operations on arrays

Example​

#include <ligetron/vbn254fr_class.h>

// Allocate vector of 1000 field elements
vbn254fr_t vec_a, vec_b, vec_result;
vbn254fr_alloc(vec_a, 1000);
vbn254fr_alloc(vec_b, 1000);
vbn254fr_alloc(vec_result, 1000);

// Initialize all elements
vbn254fr_set_ui_scalar(vec_a, 10);
vbn254fr_set_ui_scalar(vec_b, 20);

// Vectorized addition: result[i] = a[i] + b[i]
vbn254fr_addmod(vec_result, vec_a, vec_b);

// Clean up
vbn254fr_free(vec_a);
vbn254fr_free(vec_b);
vbn254fr_free(vec_result);