Triton is a virtual machine that comes with Algebraic Execution Tables (AET) and Arithmetic Intermediate Representations (AIR) for use in combination with a STARK proof system.
I wrapped the two calls tondarray's zeros in low_degree_extend_all_columns in the timing-profiler to measure how long they use, and found that for a FRI domain length of $2^{20}$ those functions take 5 % of total prover time to execute. This might be a good place to use unsafe code, or do something else. Because we're writing zeros to memory cells that will be overwritten again before being read.
I wrapped the two calls to
ndarray
'szeros
inlow_degree_extend_all_columns
in the timing-profiler to measure how long they use, and found that for a FRI domain length of $2^{20}$ those functions take 5 % of total prover time to execute. This might be a good place to useunsafe
code, or do something else. Because we're writing zeros to memory cells that will be overwritten again before being read.Cf.