Performance Concerns for Mersenne31Field::as_representative

[JulianGCalderon]

While adding QM31 support to starknet-io/types-rs, some concerns were raised on the performance impact of as_representative(). Original comment: starknet-io/types-rs#163 (comment).

The current implementation uses branching to check if the inner value is equal to MERSENNE_31_PRIME_FIELD_ORDER

lambdaworks/crates/math/src/field/fields/mersenne31/field.rs

Lines 30 to 35 in ad2a380

	fn as_representative(n: &u32) -> u32 {
	if *n == MERSENNE_31_PRIME_FIELD_ORDER {
	0
	} else {
	*n
	}

Would changing the implementation to be branchless improve performance? Using either:

• Modulo: n % MERSENNE_31_PRIME_FIELD_ORDER.
• Mask: n & (1u32 << 31) - 1.