🥮 mooncakes.io

    #Async stream benchmark

    Run from the repository root:

    moon run examples/async_bench --target native --release

    The benchmark measures the complete in-memory read, UTF-8 decode, JSON parse, and schema validation pipeline. It checks record counts, discards two warmups, and reports the median of seven runs. Payload construction is outside timing; reader creation and input encoding are included. No network or remote checks are involved. It is a manual benchmark, not a timing assertion in the test suite. Elapsed time uses the core monotonic clock and converts microseconds to milliseconds, so system clock adjustments cannot change the measured duration.

    One Windows run with MoonBit toolchain 0.1.20261012 compared the same workload before and after replacing per-byte accumulation with chunk views:

    WorkloadBeforeAfter
    JSONL, 20,000 short records81 ms57 ms
    SSE, 20,000 short frames66 ms49 ms
    JSONL, 500 records with 16 KiB strings556 ms154 ms

    These are local observations, not portable performance guarantees. Use the same workload, build mode, machine, and runtime when comparing changes.

    Source Files

    Powered by MoonBit

    Site sourceReport issueModulesSkillsCLIBuild queueStatistics

    © 2026 mooncakes.io