This adds bilrost, a little thing i've been working on.
The bilrost library supports encoding its messages both forward and backward. Backwards encoding is faster, but for obvious reasons doesn't support appending to existing data. both are included in benchmarks for comparison; hopefully this isn't too many annoying variations. It's possible we should remove at least some of these, like "populate-and-prepend," which aren't really informative on their own.
The encoding is a fork of protobuf, and the library is a fork of prost. Performance is largely similar but usually superior, in large part because i spent a couple hours profiling with rust_serialization_benchmark and restructuring & sprinkling #[inline] annotations where needed, and (in the "prepend" case) because i took the time to implement a novel encoding approach.
This adds bilrost, a little thing i've been working on.
The bilrost library supports encoding its messages both forward and backward. Backwards encoding is faster, but for obvious reasons doesn't support appending to existing data. both are included in benchmarks for comparison; hopefully this isn't too many annoying variations. It's possible we should remove at least some of these, like "populate-and-prepend," which aren't really informative on their own.
The encoding is a fork of protobuf, and the library is a fork of prost. Performance is largely similar but usually superior, in large part because i spent a couple hours profiling with rust_serialization_benchmark and restructuring & sprinkling
#[inline]
annotations where needed, and (in the "prepend" case) because i took the time to implement a novel encoding approach.