Closed djkoloski closed 3 years ago
Fixed by 61cf6ecc713e87932dc2c4ee5872aa44b7b8b6ce. capnp
ended up not needing is serialize steps split because it actually doesn't support the same viral type pattern as prost
does. I thought it did because it had a similar builder type with similar functionality, but the builder type does serialization work behind the scenes as you set fields on it. This means that splitting the benchmarks wouldn't really separate the population and encoding steps.
Only relevant to
prost
andcapnp
, needs a README blurb about what the two numbers mean and how to determine which numbers apply to your case.