Decoder results
Published measurements compare mopenzl with the pinned OpenZL v0.3.0 C implementation on identical complete Silesia files. The matrix varies graphs and typed views at default level 6, format 27 and both checksums. Every inner frame was checked byte-identical before either implementation was timed.
AMD Ryzen 9 7950X (KVM, 2 vCPUs) · rustc 1.99.1 · C -O3 · 2026-11-10.
| Graph / input view | Files | Input bytes | Compressed bytes | Ratio | C MB/s | mopenzl MB/s | / C |
|---|---|---|---|---|---|---|---|
| zstd | 12 | 211,938,580 | 61,240,193 | 3.461 | 1,303.9 | 1,230.7 | 0.94× |
| lz4 | 12 | 211,938,580 | 78,385,876 | 2.704 | 2,551.8 | 2,393.6 | 0.94× |
| lz | 12 | 211,938,580 | 64,385,308 | 3.292 | 676.5 | 1,036.2 | 1.53× |
| huffman | 12 | 211,938,580 | 130,419,862 | 1.625 | 1,591.3 | 1,234.5 | 0.78× |
| fse | 12 | 211,938,580 | 130,057,228 | 1.630 | 1,089.7 | 955.0 | 0.88× |
| entropy | 12 | 211,938,580 | 130,379,952 | 1.626 | 1,507.1 | 1,197.8 | 0.79× |
| huffman_pivco | 12 | 211,938,580 | 130,590,079 | 1.623 | 2,325.3 | 2,422.0 | 1.04× |
| flatpack | 12 | 211,938,580 | 196,426,372 | 1.079 | 2,733.3 | 4,072.7 | 1.49× |
| generic | 12 | 211,938,580 | 61,240,193 | 3.461 | 1,541.5 | 1,486.7 | 0.96× |
| field_lz_struct4 | 3 | 50,891,072 | 10,944,133 | 4.650 | 1,946.2 | 1,398.1 | 0.72× |
| bitpack_num16 | 2 | 18,444,804 | 18,444,872 | 1.000 | 7,665.6 | 17,191.8 | 2.24× |
| numeric16 | 2 | 18,444,804 | 8,318,978 | 2.217 | 1,019.6 | 809.7 | 0.79× |
| numeric32 | 3 | 50,891,072 | 10,944,138 | 4.650 | 1,840.4 | 1,241.1 | 0.67× |
| generic_num16 | 2 | 18,444,804 | 8,318,978 | 2.217 | 963.9 | 799.4 | 0.83× |
| generic_num32 | 3 | 50,891,072 | 10,944,138 | 4.650 | 1,809.9 | 1,188.4 | 0.66× |
Throughput is uncompressed MB/s (10⁶ bytes/s), aggregated as total bytes over the sum of per-file median times. Compressed bytes are equal for both sides. The linked graph pages contain file-by-file numbers and charts. This dataset is a graph sweep, not a per-level benchmark; no unmeasured levels are implied.
Read the measurements
Use the full comparison for per-file charts and ratios, raw CSV for analysis, and commands to reproduce the API mode and corpus. Numeric/record views use only aligned compatible files, so compare each graph within its own corpus rather than ranking unrelated rows as one workload.
Other API modes
The parallel API records segment/task scheduling separately. The framed measurements record container overhead and independent-frame C timings. End-to-end results include different setup/allocation costs from reused per-graph benchmarks.