Conversation
Performance: add fat LTO + codegen-units=1 on release/bench profiles,
inline the tape accessors and deserialize_value, and dispatch on the raw
tape tag byte instead of an Option<TapeType> round-trip. Large-100KB
struct path ~24% faster, value path ~12% faster; value path now beats
serde_json::Value on large inputs.
Add opt-in `native` feature (off by default) that compiles the bundled
simdjson C++ with -march=native; ship .cargo/config.toml.example for the
matching Rust target-cpu=native, gitignoring the local copy so published
builds stay portable.
Breaking: replace the single tape::Error::TapeCorrupted { message } with
typed variants (OutOfBounds, UnknownTapeType, UnexpectedTapeType,
StringOutOfBounds, MissingValueWord, InvalidUtf8) carrying positions and
lengths. serde::Error::TapeCorrupted is removed; tape failures now
surface transparently as serde::Error::Tape { source: tape::Error }.
Use ensure! for same-type checks in tape.rs; .build().into() across the
serde boundary. Also fixes stale serde_support doctest paths.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Perf-profiled the Large struct deserialization path and attacked the two biggest user-space hotspots: - get_string called str::from_utf8 on every string node (~11% of CPU), re-validating UTF-8 that simdjson already checked during stage2. Replace with from_utf8_unchecked under a documented SAFETY invariant: a tape is only reachable after parser_parse returned success, which means the string_buf is valid UTF-8. - Vec fields deserialized with no size hint, so serde grew them element by element (~27% in malloc/realloc). Add TapeRef::scope_count reading the 24-bit element count simdjson packs into the opening bracket tape word, and surface it via TapeSeqAccess::size_hint so serde preallocates with Vec::with_capacity (capped by serde's cautious hint). Large-100KB struct: 183 us -> 137 us (-25%), now beating simd-json (176 us) and serde_json (187 us). Medium-2KB: 840 ns -> 638 ns. from_utf8 is gone from the profile; _int_malloc and grow_one dropped sharply. Also add simd-json as a dev-dependency bench competitor (struct path only, three sizes) and a throwaway examples/perf_sample.rs for perf-record sampling of the hot loop. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
No description provided.