AztecProtocol/barretenberg

Special handling of first subrelation value in `process_subrelation_evaluations`

Opened this issue · 0 comments

We allocate FF linearly_independent_contribution{ 0 } (40ns?) only to immediately overwrite it with an element * challenges[0] where challenges[0] is 1. Looking at the numbers in basics.bench.cpp, this costs about 60ns per row during the combiner computation in x86. There are about 1.6M rows in the Full/6 bench, which I think is about 1/4 of a second in WASM. This is small but not nothing, and the change is very local .