May 1, 2023

I did some performance analysis against store reading, which basically amounts to reading txs. This was improved slightly by the new reader, but the bulk of work is in the input/witness/output and these three components are still using the std::iostream reader because (only) these using the chain object serialization. The tools showed that the bulk of time was spent in these specific areas. This is to be expected, as this is the bulk of the store data - but this should then results in comparable improvement in both download/store and read (for validation).