crates/aptos-crypto/examples/README.md
Run from the crates/aptos-crypto directory via:
cargo run --release --example is_zkcrypto_constant_time
cargo run --release --example is_blstrs_constant_time
Quoting from the original README here: The program output will look like:
bench constant_time::blstrs_scalar_mul::run_bench seeded with 0xc41137dfdf1b2629
bench constant_time::blstrs_scalar_mul::run_bench ...
# of 1 bits in scalars for "left" class is in [1, 4)
# of 1 bits in scalars for "right" class is always 250
: n == +0.010M, max t = +1.86799, max tau = +0.01869, (5/tau)^2 = 71538
It is interpreted as follows. Firstly, note that the runtime distributions are cropped at different percentiles and about 100 t-tests are performed. Of these t-tests, the one that produces the largest absolute t-value is printed as max_t. The other values printed are:
n, indicating the number of samples used in computing this t-valuemax_tau, which is the t-value scaled for the samples size (formally, max_tau = max_t / sqrt(n))(5/tau)^2, which indicates the number of measurements that would be needed to distinguish the two distributions with t > 5t-values outside (-5, 5) (i.e., of absolute value > 5) are generally considered a good indication that the function is not constant time. t-values in (-5, 5) does not necessarily imply that the function is constant-time, since there may be other input distributions under which the function behaves significantly differently.